Handbook of Accelerator Physics and Engineering (PDFDrive) PDF

Handbook of Accelerator Physics and Engineering Downloaded from www.worldscientific.
com
by 91.213.8.236 on 03/24/16. For personal use only.
and
Handbook of
Second Edition
Accelerator Physics
Engineering
Handbook of Accelerator Physics and Engineering Downloaded from www.worldscientific.com
This page intentionally left blank

Handbook of
Accelerator Physics
and
Engineering
Second Edition
edited by
Alexander Wu Chao
SLAC National Accelerator Laboratory, USA
Karl Hubert Mess

CERN, Switzerland
Maury Tigner
Cornell University, USA
Frank Zimmermann
CERN, Switzerland
World Scientific
NEW JERSEY • LONDON • SINGAPORE • BEIJING • SHANGHAI • HONG KONG • TA I P E I • CHENNAI
Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE
British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.
HANDBOOK OF ACCELERATOR PHYSICS AND ENGINEERING

Second Edition
Copyright © 2013 by World Scientific Publishing Co. Pte. Ltd.
All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the Publisher.
For photocopying of material in this volume, please pay a copying fee through the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to
photocopy is not required from the publisher.
ISBN 978-981-4415-84-2
ISBN 978-981-4417-17-4 (pbk)
Printed in Singapore by Mainland Press Pte Ltd.

Preface
This Handbook is a product of the world community of accelerator physicists and engineers. The first
edition was issued September, 1998. With the continued advancing of the accelerator field, a second
edition appears now in order.
This is not a textbook but rather a collection of information useful to professionals in research,
design, construction, and operation of accelerators. The Handbook has been prepared by more than 200
experienced experts from across the spectrum of accelerator related institutions and to them great thanks
are due.
In addition to content, a high priority has been given to portability of the book. This has led to a
sacrifice of some aesthetics in order to make the text as compact as possible. For that, our apologies go
to users and authors alike.
Singularly important are the references to be found at the end of each subsection. Here the user will
find locations of tutorial material as well as reliable detail for further reading. The references for the
most part are not intended to be exhaustive or to indicate priority of discovery or invention, but rather
to provide a reliable lead into the literature. In addition, a detailed index gives access to occurrences of
important subjects and concepts to be found herein.
The fees and royalties that would normally be paid to authors and editors are donated to provide
scholarships to the CERN Accelerator School and to the US Particle Accelerator School.
Authors and Editors have made great efforts to find and eliminate errors. Nevertheless we recognize
that there will be errors and have provided for errata to appear on a Handbook website:
http://www.worldscientific.com/worldscibooks/10.1142/8543
Please help in this community effort by sending suggestions for corrections by e-mail to the address
achao@slac.stanford.edu.
Editors
Alexander Wu Chao, Stanford, California

Maury Tigner, Ithaca, New York
Frank Zimmermann, Geneva, Switzerland
Karl-Hubert Meß, Sülfeld, Germany
July 2012
v
Acknowledgments to Second Edition
The editors hereby acknowledge with appreciation the support of the Managements of SLAC National
Accelerator Laboratory at Stanford University, the Cornell Laboratory for Accelerator-based Sciences
and Education, and the European Organization for Nuclear Research, CERN, without which this work
would have been impossible. We also wish to give our special thanks to Jeanne Butler and World
Scientific Publishing Company staff for countless essential technical help.
Acknowledgments to First Edition
The editors hereby acknowledge with deepest appreciation and thank the support of the Directors of
SLAC at Stanford University and of the Laboratory of Nuclear Studies at Cornell University without
which this work would have simply been impossible. We also wish to give special thanks to Scott
Berg, Indiana University, and Tom Hays, Cornell, for essential help with the technical complexities
of typesetting software and to Jim Wahl, SLAC and Don Miller, Cornell for their indispensable help
with the illustrations. For continual guidance and help with the intricacies of computer hardware and
software, great thanks are due to Tom Knight, SLAC.
vi
Table of Contents
Preface v
1 INTRODUCTION 1
1.1 HOW TO USE THIS BOOK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 NOMENCLATURE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 FUNDAMENTAL CONSTANTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 UNITS AND CONVERSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4.1 Units A.W. Chao . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4.2 Conversions M. Tigner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.5 FUNDAMENTAL FORMULAE A.W. Chao . . . . . . . . . . . . . . . . . . . . . . 5
1.5.1 Special Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5.2 Curvilinear Coordinate Systems . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5.3 Electromagnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5.4 Kinematical Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5.5 Vector Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5.6 Relativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.6 GLOSSARY OF ACCELERATOR TYPES . . . . . . . . . . . . . . . . . . . . . . . 8
1.6.1 Antiproton Sources K. Gollwitzer, J. Marriner . . . . . . . . . . . . . . . . . 8
1.6.2 Betatron M. Tigner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.6.3 Colliders D. Hartill . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.6.4 Cyclotron H. Blosser, M. Craddock . . . . . . . . . . . . . . . . . . . . . . . 14
1.6.5 Electrostatic Accelerator J. Ferry . . . . . . . . . . . . . . . . . . . . . . . . 17
1.6.6 FFAG Accelerator M.K. Craddock . . . . . . . . . . . . . . . . . . . . . . . 18
1.6.7 High Voltage Electrodynamic Accelerators M.R. Cleland . . . . . . . . . . . 22
1.6.8 Induction Linac R. Bangerter . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.6.9 Industrial Accelerators R. Hellborg . . . . . . . . . . . . . . . . . . . . . . . 27
1.6.10 Laser, Wakefield and Plasma Accelerators J. Rosenzweig . . . . . . . . . . . . 31
1.6.11 Linear Accelerators for Electrons G.A. Loew . . . . . . . . . . . . . . . . . . 34
1.6.12 Linear Accelerators for Protons S. Henderson, A. Aleksandrov . . . . . . . . . 36
1.6.13 Livingston Chart D. Hartill . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.6.14 Medical Applications of Accelerators J. Alonso . . . . . . . . . . . . . . . . . 40
1.6.14.1 Radiation therapy . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.6.14.2 Radioisotopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
1.6.15 μ+ μ− Collider R. Palmer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
1.6.15.1 Collider . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
1.6.15.2 Muon storage ring neutrino factories . . . . . . . . . . . . . . . . . 46
1.6.15.3 Technical challenges . . . . . . . . . . . . . . . . . . . . . . . . . . 46
1.6.16 Neutron Sources J. Wei, H. Lengeler . . . . . . . . . . . . . . . . . . . . . . 47
1.6.17 Pulsed High Voltage Devices J.A. Nation, D.A. Hammer . . . . . . . . . . . . 51
1.6.18 Radio Frequency Quadrupole J. Staples . . . . . . . . . . . . . . . . . . . . . 52
1.6.19 Rhodotron Y. Jongen, M. Abs . . . . . . . . . . . . . . . . . . . . . . . . . . 55
1.6.20 Storage Rings W. Fischer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
vii
1.6.21 Synchrotrons C. Zhang, S.X. Fang . . . . . . . . . . . . . . . . . . . . . . . . 58
1.6.22 Two-Beam Accelerator J.P. Delahaye . . . . . . . . . . . . . . . . . . . . . . 60
1.7 ACCELERATOR COMPUTER CODES R. Ryne . . . . . . . . . . . . . . . . . . . . 62
2 BEAM DYNAMICS 65
2.1 PHASE SPACE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.1.1 Linear Betatron Motion D.A. Edwards, M.J. Syphers . . . . . . . . . . . . . 65
2.1.2 Longitudinal Motion D.A. Edwards, M.J. Syphers . . . . . . . . . . . . . . . 66
2.1.3 Linear Coupled System D.A. Edwards, M.J. Syphers . . . . . . . . . . . . . . 67
2.1.4 Orbital Eigen-Analysis for Electron Storage Ring J.A. Ellison, H. Mais, G. Ripken 68
2.2 OPTICS AND LATTICES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.2.1 Single Element Optics K. Brown . . . . . . . . . . . . . . . . . . . . . . . . 71

2.2.2 3-D Multipole Expansion, Calculation of Transfer Maps from Field Data, Fringe
Fields M. Venturini, A. Dragt . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.2.3 Lattices for Collider Storage Rings E. Keil . . . . . . . . . . . . . . . . . . . 77
2.2.4 Lattices for Low-Emittance Light Sources C. Steier . . . . . . . . . . . . . . 81
2.2.4.1 Lattice choices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
2.2.4.2 Chromaticity correction and nonlinear optimization . . . . . . . . . 83
2.2.4.3 Systematic lattice optimization techniques . . . . . . . . . . . . . . 83

2.2.4.4 Evolution of light source lattices . . . . . . . . . . . . . . . . . . . 83
2.2.4.5 Ultimate storage rings . . . . . . . . . . . . . . . . . . . . . . . . . 84
2.2.5 Betatron Motion with Coupling of Two Degrees of Freedom V. Lebedev, A. Burov 85
2.3 NONLINEAR DYNAMICS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
2.3.1 Hamiltonian K. Symon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
2.3.1.1 General case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
2.3.1.2 Transverse motion . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
2.3.1.3 Longitudinal motion . . . . . . . . . . . . . . . . . . . . . . . . . . 91
2.3.1.4 Synchrobetatron coupling . . . . . . . . . . . . . . . . . . . . . . . 92
2.3.2 Tune Dependence on Momentum and Betatron Amplitudes
D.A. Edwards, M.J. Syphers . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
2.3.3 Nonlinear Resonances D.A. Edwards, M.J. Syphers . . . . . . . . . . . . . . . 94
2.3.4 Synchrobetatron Resonances A. Piwinski . . . . . . . . . . . . . . . . . . . . 94
2.3.5 Taylor Map, Hénon Map, Standard Map A. Dragt . . . . . . . . . . . . . . . 97
2.3.6 Lie Algebraic Methods A. Dragt . . . . . . . . . . . . . . . . . . . . . . . . 99
2.3.7 Differential Algebraic Techniques M. Berz . . . . . . . . . . . . . . . . . . . 105
2.3.8 Symplectic Integration Methods H.Yoshida . . . . . . . . . . . . . . . . . . . 109
2.3.8.1 Methods of realization . . . . . . . . . . . . . . . . . . . . . . . . . 109
2.3.8.2 Symplectic method vs. nonsymplectic method . . . . . . . . . . . . 111
2.3.9 Dynamic Aperture A. Wolski . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
2.3.10 Decoherence M.A. Furman . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
2.3.11 Momentum Compaction and Phase Slip Factor K.Y. Ng . . . . . . . . . . . . 116
2.3.12 Nonlinear Dynamics Experiments W. Fischer . . . . . . . . . . . . . . . . . . 118
2.3.13 Echo Effects G.V. Stupakov . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
2.4 COLLECTIVE EFFECTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
2.4.1 Collective Effects in High Energy Electron Linacs K. Kubo, K. Yokoya,
K. Thompson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
2.4.1.1 Single bunch longitudinal dynamics . . . . . . . . . . . . . . . . . . 123
2.4.1.2 Multibunch longitudinal dynamics . . . . . . . . . . . . . . . . . . 123
2.4.1.3 Single bunch transverse dynamics . . . . . . . . . . . . . . . . . . . 124
2.4.1.4 Multibunch transverse dynamics . . . . . . . . . . . . . . . . . . . 125
2.4.1.5 Effects of structure misalignment . . . . . . . . . . . . . . . . . . . 126
2.4.2 Collective Effects in Energy Recovery Linacs G. Hoffstaetter . . . . . . . . . 127
viii
2.4.3 Beam Loading D. Boussard . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
2.4.3.1 Single-bunch passage in a cavity . . . . . . . . . . . . . . . . . . . 129
2.4.3.2 Cavity equivalent circuit . . . . . . . . . . . . . . . . . . . . . . . . 129
2.4.3.3 Transmission of small modulations (AM and PM) through a cavity
with beam loading . . . . . . . . . . . . . . . . . . . . . . . . . . 130
2.4.3.4 Periodic beam loading at multiples of f0 . . . . . . . . . . . . . . . 131
2.4.3.5 Rf power needed for transient beam-loading correction . . . . . . . 132
2.4.3.6 Traveling-wave cavities . . . . . . . . . . . . . . . . . . . . . . . . 132
2.4.4 Space-Charge Dominated Beams in Guns and Transport Lines M. Ferrario . . 133
2.4.5 Space Charge Effects in Circular Accelerators B. Zotter . . . . . . . . . . . . 137
2.4.5.1 Direct space charge effects . . . . . . . . . . . . . . . . . . . . . . 137
2.4.5.2 Betatron frequency shifts . . . . . . . . . . . . . . . . . . . . . . . 138

2.4.6 Beam Dynamics in Proton Linacs S. Henderson, A. Aleksandrov . . . . . . . . 140
2.4.7 Vlasov and Fokker-Planck Equations B. Zotter . . . . . . . . . . . . . . . . . 144
2.4.8 Potential Well Effect B. Zotter . . . . . . . . . . . . . . . . . . . . . . . . . . 145
2.4.9 Single-Bunch Instabilities in Circular Accelerators B. Zotter . . . . . . . . . . 147
2.4.10 Sacherer Formulae B. Zotter . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
2.4.11 Landau Damping K.Y. Ng . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
2.4.12 Intrabeam Scattering and Touschek Effect V. Lebedev . . . . . . . . . . . . . 155

2.4.13 Ion Trapping, Beam-Ion Instabilities, and Dust F. Zimmermann . . . . . . . . 159
2.4.13.1 Ion trapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
2.4.13.2 Dust particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
2.4.13.3 Single-pass ion effects in storage rings and linacs . . . . . . . . . . 161
2.4.14 Electron-Cloud Effect M.A. Furman . . . . . . . . . . . . . . . . . . . . . . 163
2.4.15 Coherent Synchrotron Radiation Instability G. Stupakov . . . . . . . . . . . . 167
2.5 BEAM-BEAM EFFECTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
2.5.1 Beam-Beam Effects in Storage Ring Colliders K. Hirata . . . . . . . . . . . . 169
2.5.1.1 Infinitely short bunches . . . . . . . . . . . . . . . . . . . . . . . . 169
2.5.1.2 Long bunches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
2.5.1.3 Dispersion at IP, crossing angle . . . . . . . . . . . . . . . . . . . . 173
2.5.2 Crab Waist Collision Scheme M. Zobov . . . . . . . . . . . . . . . . . . . . . 174
2.5.3 Beam-Beam Effects in Linear Colliders P. Chen, D. Schulte . . . . . . . . . . 175
2.5.3.1 Disruption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
2.5.3.2 Beamstrahlung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
2.5.3.3 Background and spent beam . . . . . . . . . . . . . . . . . . . . . . 178
2.5.4 Parasitic Beam-Beam Effects and Separation Schemes J.M. Jowett . . . . . . 180
2.5.4.1 Separation schemes . . . . . . . . . . . . . . . . . . . . . . . . . . 180
2.5.4.2 Long-range beam-beam effects . . . . . . . . . . . . . . . . . . . . 181
2.6 POLARIZATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
2.6.1 Thomas-BMT Equation T. Roser . . . . . . . . . . . . . . . . . . . . . . . . 183
2.6.2 Spinor Algebra T. Roser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
2.6.3 Spin Rotators and Siberian Snakes T. Roser . . . . . . . . . . . . . . . . . . . 185
2.6.4 Depolarizing Resonances and Spin Flippers T. Roser . . . . . . . . . . . . . . 186
2.6.5 Polarized Hadron Beams and Siberian Snakes A.D. Krisch, M.A. Leonova . . . 187
2.6.6 Radiative Polarization in Electron Storage Rings D.P. Barber, G. Ripken . . . 191
2.6.7 Computing Algorithms for e− /e+ Polarization in Storage Rings D.P. Barber,
G. Ripken . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
2.6.8 Spin Matching in e− /e+ Rings D.P. Barber, G. Ripken . . . . . . . . . . . . . 195
2.6.9 Lie Algebra for Spin Motion K. Yokoya . . . . . . . . . . . . . . . . . . . . . 198
2.7 BEAM COOLING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
2.7.1 Stochastic Cooling M. Blaskiewicz . . . . . . . . . . . . . . . . . . . . . . . 199
2.7.1.1 Cooling rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
ix
2.7.1.2 Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
2.7.1.3 New techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
2.7.2 Electron Cooling S. Nagaitsev . . . . . . . . . . . . . . . . . . . . . . . . . . 202
2.7.3 Laser Cooling J. Hangst, A. Noda . . . . . . . . . . . . . . . . . . . . . . . . 205
2.7.4 Ionization Cooling D. Neuffer . . . . . . . . . . . . . . . . . . . . . . . . . . 209
2.7.5 Crystalline Beam J. Wei . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
3 ELECTROMAGNETIC AND NUCLEAR INTERACTIONS 215

3.1 SYNCHROTRON RADIATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
3.1.1 Radiation of a Point Charge H. Wiedemann . . . . . . . . . . . . . . . . . . . 215
3.1.2 Coherent Radiation H. Wiedemann . . . . . . . . . . . . . . . . . . . . . . . 216
3.1.3 Bending Magnet Radiation H. Wiedemann . . . . . . . . . . . . . . . . . . . 217

3.1.4 Synchrotron Radiation in Storage Rings H.Wiedemann . . . . . . . . . . . . . 220
3.1.4.1 Radiation integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
3.1.4.2 Radiation damping . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
3.1.4.3 Quantum excitation . . . . . . . . . . . . . . . . . . . . . . . . . . 221
3.1.4.4 Equilibrium beam emittances . . . . . . . . . . . . . . . . . . . . . 221
3.1.4.5 Damping wigglers . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
3.1.4.6 Quantum lifetimes . . . . . . . . . . . . . . . . . . . . . . . . . . . 222

3.1.5 Undulator and Wiggler Radiation H. Wiedemann . . . . . . . . . . . . . . . . 223
3.1.5.1 Polarization of synchrotron radiation H. Wiedemann . . . . . . . . . 225
3.1.6 Transition and Diffraction Radiation C. Thongbai . . . . . . . . . . . . . . . 226
3.1.7 Coherent Synchrotron Radiation S. Krinsky . . . . . . . . . . . . . . . . . . . 227
3.1.8 Free-Electron Lasers Z. Huang, P. Schmüser . . . . . . . . . . . . . . . . . . 229
3.1.9 Ultrashort X-ray Pulse Generation A. Zholents . . . . . . . . . . . . . . . . . 233
3.1.10 Compton/Thomson Sources G.A. Krafft . . . . . . . . . . . . . . . . . . . . . 235
3.1.10.1 Luminosity description . . . . . . . . . . . . . . . . . . . . . . . 236
3.1.10.2 Nonlinear scattering . . . . . . . . . . . . . . . . . . . . . . . . . 236
3.1.11 “Short” Magnet and Edge Radiation H. Wiedemann . . . . . . . . . . . . . . 237
3.1.12 Beam Solid-Target Photon Physics K. Ispiryan . . . . . . . . . . . . . . . . . 237
3.2 IMPEDANCES AND WAKE FUNCTIONS . . . . . . . . . . . . . . . . . . . . . . . 242
3.2.1 Definitions and Properties of Impedances and Wake Functions T. Suzuki . . . 242
3.2.2 Impedance Calculation, Frequency Domain R.L. Gluckstern, S.S. Kurennoy . . 243
3.2.3 Impedance Calculation. Time Domain E. Gjonaj, T. Weiland . . . . . . . . . . 248
3.2.4 Explicit Expressions of Impedances and Wake Functions K.Y. Ng, K. Bane . . 252
3.2.5 Effective Impedance T. Suzuki . . . . . . . . . . . . . . . . . . . . . . . . . 262
3.2.6 Parasitic Loss P. Wilson, B. Zotter, Y.-H. Chin . . . . . . . . . . . . . . . . . . 263
3.2.7 Trapped Modes S.S. Kurennoy . . . . . . . . . . . . . . . . . . . . . . . . . . 265
3.3 PARTICLE-MATTER INTERACTION . . . . . . . . . . . . . . . . . . . . . . . . . 266
3.3.1 Basic Formulae M. Tigner, A.W. Chao . . . . . . . . . . . . . . . . . . . . . 266
3.3.2 Beam and Luminosity Lifetime . . . . . . . . . . . . . . . . . . . . . . . . 269
3.3.2.1 Protons N.V. Mokhov, V.I. Balbekov . . . . . . . . . . . . . . . . . . 269
3.3.2.2 Electrons M.S. Zisman . . . . . . . . . . . . . . . . . . . . . . . . 271
3.3.3 Bhabha Scattering (e+e− → e+e− ) J.E. Spencer . . . . . . . . . . . . . . . . 273
3.3.4 Compton Scattering (e±γ → e±γ) J.E. Spencer . . . . . . . . . . . . . . . . . 275
3.3.5 Limit of Focusing of Electron Beam due to Synchrotron Radiation K. Oide . . 277
3.3.6 Thermal Outgassing and Beam Induced Desorption V. Baglin, E. Mahner,
O. Gröbner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
3.3.7 Photoemission and Secondary Emission V. Baglin . . . . . . . . . . . . . . . 283
3.3.7.1 Photoemission . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
3.3.7.2 Secondary emission . . . . . . . . . . . . . . . . . . . . . . . . . 284
3.3.8 Ionization Processes F. Zimmermann . . . . . . . . . . . . . . . . . . . . . . 286
x
3.3.9 Beam Induced Detector Backgrounds and Irradiation in e+ e− Colliders
S.D. Henderson, M. Sullivan . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
3.3.9.1 Sources of detector backgrounds . . . . . . . . . . . . . . . . . . . 287
3.3.9.2 Detector and IR radiation tolerance and budget . . . . . . . . . . . . 290
3.3.9.3 Detector background shielding . . . . . . . . . . . . . . . . . . . . 290
3.3.9.4 Detector background and radiation estimation . . . . . . . . . . . . 291
3.3.10 Particle Interactions and Beam-Induced Backgrounds and Radiation
N.V. Mokhov, S.I. Striganov . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
3.3.11 Beam Collimation R. Assmann . . . . . . . . . . . . . . . . . . . . . . . . . 300
3.3.12 Atomic and Nuclear Properties of Materials . . . . . . . . . . . . . . . . . . 308
4 OPERATIONAL CONSIDERATIONS 311

4.1 LUMINOSITY M.A. Furman, M.S. Zisman . . . . . . . . . . . . . . . . . . . . . . . 311
4.2 BRIGHTNESS P. Elleaume, K.-J. Kim . . . . . . . . . . . . . . . . . . . . . . . . . 318
4.2.1 Particle Beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 318
4.2.2 Radiation Beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
4.2.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
4.2.2.2 Bending magnet radiation . . . . . . . . . . . . . . . . . . . . . . 319
4.2.2.3 Wiggler radiation . . . . . . . . . . . . . . . . . . . . . . . . . . 320

4.2.2.4 Undulator radiation . . . . . . . . . . . . . . . . . . . . . . . . . 320
4.2.2.5 Brightness comparison . . . . . . . . . . . . . . . . . . . . . . . 320
4.3 LINAC OPERATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
4.3.1 Operation of High Energy Electron Linacs T.O. Raubenheimer . . . . . . . . 321
4.3.2 Operation of Superconducting Linacs S. Schreiber . . . . . . . . . . . . . . . 326
4.3.3 Halo in High Intensity Proton Linacs T.O. Raubenheimer . . . . . . . . . . . 329
4.4 RECIRCULATED ENERGY RECOVERY LINACS S.A. Bogacz, D.R. Douglas,
G.A. Krafft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
4.4.1 Recirculation and Energy Recovery . . . . . . . . . . . . . . . . . . . . . . . 329
4.4.2 System Architecture and Beam Dynamics Issues . . . . . . . . . . . . . . . . 331
4.5 BUNCH COMPRESSION P. Emma . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
4.6 OPERATION OF LINEAR-COLLIDER FINAL FOCUS SYSTEMS A.A. Seryi,
G.R. White . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
4.6.1 Chromaticity Compensation . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
4.7 OPERATION OF CIRCULAR ACCELERATORS . . . . . . . . . . . . . . . . . . . 346
4.7.1 Error Sources and Effects D. Rice . . . . . . . . . . . . . . . . . . . . . . . 346
4.7.2 Orbit and Lattice Function Measurements D. Rice . . . . . . . . . . . . . . . 347
4.7.3 Orbit Correction J. Wenninger . . . . . . . . . . . . . . . . . . . . . . . . . . 350
4.7.3.1 Global orbit correction . . . . . . . . . . . . . . . . . . . . . . . . 350
4.7.3.2 SVD algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
4.7.3.3 MICADO algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . 351
4.7.3.4 Local orbit bumps . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
4.7.4 Measurement and Diagnosis of Coupling and Solenoid Compensation D. Rubin 352
4.7.4.1 Sources of transverse coupling . . . . . . . . . . . . . . . . . . . . 352
4.7.4.2 Solenoids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
4.7.4.3 Coupling matrix analysis . . . . . . . . . . . . . . . . . . . . . . . 353
4.7.4.4 Measurement of coupling . . . . . . . . . . . . . . . . . . . . . . . 354
4.7.4.5 Solenoid compensation . . . . . . . . . . . . . . . . . . . . . . . . 355
4.7.5 Modeling and Control of Storage Rings Using Orbit Measurements J. Safranek 357
4.7.6 Emittance Dilution Effects M. Syphers . . . . . . . . . . . . . . . . . . . . . 360
4.7.6.1 Injection mismatch . . . . . . . . . . . . . . . . . . . . . . . . . . 360
4.7.6.2 Diffusion processes . . . . . . . . . . . . . . . . . . . . . . . . . . 361
xi
4.7.7 Real-Time Measurement and Control of Tune, Coupling and Chromaticity
R. Jones, R. Steinhagen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
4.7.7.1 Tune measurement . . . . . . . . . . . . . . . . . . . . . . . . . . 362
4.7.7.2 Chromaticity measurement . . . . . . . . . . . . . . . . . . . . . 363
4.7.7.3 Coupling measurement . . . . . . . . . . . . . . . . . . . . . . . 363
4.7.7.4 Real-time control of tune, coupling & chromaticity . . . . . . . . 364
4.7.8 Measurement of Dispersion by Resonant Excitation D.L.Rubin . . . . . . . . 365
4.7.8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
4.7.8.2 Formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
4.7.8.3 Measurement of the coupling matrix . . . . . . . . . . . . . . . . 366
4.8 TEMPORAL AND SPATIAL CORRELATIONS IN BPM MEASUREMENTS . . . 367
4.8.1 Non-Invasive Measurement for Linacs J. Irwin, Y. Yan . . . . . . . . . . . . . 367

4.8.2 Invasive Measurement for e+ , e− Circular Accelerators J. Irwin, Y. Yan . . . . 370
4.8.3 Virtual Models for Proton Circular Accelerators Y. Yan . . . . . . . . . . . . . 373
4.9 TRANSITION CROSSING J. Wei . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
4.10 RF GYMNASTICS IN A SYNCHROTRON R. Garoby . . . . . . . . . . . . . . . . 376
4.10.1 Adiabaticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
4.10.2 Single Bunch Manipulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
4.10.3 Multi-Bunch Manipulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 378

4.10.4 Debunched Beam Manipulation . . . . . . . . . . . . . . . . . . . . . . . . . 380
4.10.5 Beam Manipulations with Broad-Band RF Systems . . . . . . . . . . . . . . . 380
4.11 ENERGY MEASUREMENT WITH ELECTRON BEAMS J. Seeman . . . . . . . . . 381
4.12 RING INJECTION AND EXTRACTION G.H. Rees . . . . . . . . . . . . . . . . . . 382
4.12.1 Aspects of Slow Extraction P.J. Bryant . . . . . . . . . . . . . . . . . . . . . 385
4.13 COLLISION SCHEMES FOR RING COLLIDERS M. Zobov, F. Zimmermann . . . . 387
4.14 BEAM-BEAM COMPENSATION SCHEMES J.P. Koutchouk, V. Shiltsev . . . . . . . 391
4.14.1 Compensation of the Head-On Beam-Beam Effect . . . . . . . . . . . . . . 391
4.14.2 Compensation of the Long-Range Beam-Beam Effect . . . . . . . . . . . . . 392
4.15 SPACE CHARGE COMPENSATION (SCC) IN HADRON BEAMS V. Shiltsev . . . . 394
4.15.1 Longitudinal SCC: Inductive Inserts . . . . . . . . . . . . . . . . . . . . . . 394
4.15.2 Transverse SCC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
4.16 OPERATIONAL CONSIDERATIONS ON COOLING S. Nagaitsev, M. Blaskiewicz . 396
5 MECHANICAL CONSIDERATIONS 399

5.1 MECHANICAL AND THERMAL PROPERTIES OF STRUCTURAL MATERIALS
M. Kuchnir . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
5.2 MECHANICAL AND THERMAL PROPERTIES OF COMPOSITE
SUPERCONDUCTORS R.M. Scanlan . . . . . . . . . . . . . . . . . . . . . . . . . 404
5.3 THERMODYNAMIC & HYDRODYNAMIC PROPERTIES OF COOLANTS &
CRYOGENS M. McAshan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 406
5.4 CREEP AND STRESS RELAXATION IN ACCELERATOR COMPONENTS
F. Markley . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410
5.5 ELECTRIC AND MAGNETIC FORCES M. Tigner . . . . . . . . . . . . . . . . . . 411
5.6 DEFLECTIONS AND BUCKLING M. Tigner . . . . . . . . . . . . . . . . . . . . . 412
5.7 PRACTICAL HEAT TRANSFER AND FLUID FLOW M. McAshan, M. Tigner . . . 414
5.8 FABRICATION OF NIOBIUM RF STRUCTURES T. Hays, H. Padamsee, D. Proch . 420
5.9 REFRIGERATION SYSTEMS C. Rode, R. Ganni . . . . . . . . . . . . . . . . . . . 425
5.9.1 Refrigerators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
5.9.2 Storage and Utilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
5.9.3 Transfer Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
5.10 VACUUM SYSTEMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
5.10.1 Requirements for Vacuum Systems N.B. Mistry, Y. Li . . . . . . . . . . . . . . 431
xii
5.10.2 Units, Conversions and Some Useful Formulae N.B. Mistry, Y. Li . . . . . . . 432
5.10.3 Conductance and Pressure Profiles N.B. Mistry, Y. Li . . . . . . . . . . . . . . 433
5.10.4 Pumping Methods N.B. Mistry, Y. Li . . . . . . . . . . . . . . . . . . . . . . . 435
5.10.5 Instrumentation N.B. Mistry, Y. Li . . . . . . . . . . . . . . . . . . . . . . . . 439
5.10.6 Vacuum Chamber Design and Fabrication N.B. Mistry, Y. Li . . . . . . . . . . 440
5.10.7 Special Components in the Vacuum System N.B. Mistry, Y. Li . . . . . . . . . 445
5.10.8 Ceramic Vacuum Chamber Design H.L. Phillips . . . . . . . . . . . . . . . . 447
5.11 ALIGNMENT R. Ruland . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 448
5.12 MAGNET SUPPORTS AND ALIGNMENT G. Bowden . . . . . . . . . . . . . . . . 451
5.13 GROUND VIBRATION C. Montag, J. Rossbach . . . . . . . . . . . . . . . . . . . . 453
5.13.1 Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453
5.13.2 Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454

5.13.3 Instruments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454
5.13.4 Linacs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
5.13.5 Circular Accelerators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 456
5.13.6 Numerical Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458
5.14 VIBRATION CONTROL IN ACCELERATORS C. Montag . . . . . . . . . . . . . . 459
5.14.1 General Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459
5.14.2 Passive Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459

5.14.3 Active Stabilization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 460
6 ELECTRICAL CONSIDERATIONS 463

6.1 PROPERTIES OF DIELECTRICS M. Tigner . . . . . . . . . . . . . . . . . . . . . . 463
6.2 PROPERTIES OF CONDUCTORS, NORMAL AND SUPERCONDUCTING
R.M. Scanlan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 465
6.3 PROPERTIES OF FERROMAGNETIC MATERIALS M.Tigner . . . . . . . . . . . . 471
6.4 PERMANENT MAGNET MATERIALS R.D. Schlueter . . . . . . . . . . . . . . . . 471
6.5 PROPERTIES OF LOSSY MATERIALS E. Chojnacki . . . . . . . . . . . . . . . . . 473
6.6 COMMON TRANSMISSION LINES AND CAVITIES M. Tigner . . . . . . . . . . . 475
6.7 RF PULSE COMPRESSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 481
6.7.1 Passive Pulse Compression Z.D. Farkas . . . . . . . . . . . . . . . . . . . . . 481
6.7.2 Active Pulse Compression S. Tantawi . . . . . . . . . . . . . . . . . . . . . . 485
6.7.3 Ultra-High-Power Multimoded Rf Components S. Tantawi, C. Nantista . . . . 486
6.8 RF WINDOWS AND CAVITY COUPLING R.M. Sundelin, H.L. Phillips . . . . . . . 491
6.9 MULTIPACTING D. Proch, J. Sekutowicz . . . . . . . . . . . . . . . . . . . . . . . 494
6.10 POLYPHASE POWER CIRCUITS M. Tigner . . . . . . . . . . . . . . . . . . . . . . 496
6.11 HIGH PRECISION POWER CONVERTERS F. Bordry, J.P. Burne,
M. Cerqueira Bastos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 498
6.11.1 Introduction to Magnet Power Converters . . . . . . . . . . . . . . . . . . . . 498
6.11.2 Main Parameters of Magnet Power Converters . . . . . . . . . . . . . . . . . 498
6.11.3 Power Converter Topologies . . . . . . . . . . . . . . . . . . . . . . . . . . . 498
6.11.3.1 Thyristor controlled rectifier . . . . . . . . . . . . . . . . . . . . . 498
6.11.3.2 Switch-mode power converter . . . . . . . . . . . . . . . . . . . . 499
6.12 HIGH ACCURACY IN POWER CONVERTERS FOR PARTICLE ACCELERATORS
F. Bordry, J.P. Burnet, M. Cerqueira Bastos . . . . . . . . . . . . . . . . . . . . . . . 502
6.12.1 Power Converter Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 502
6.12.2 Current Measurement in Particle Accelerators . . . . . . . . . . . . . . . . . . 503
6.13 HIGH-GRADIENT LIMITATIONS IN ROOM TEMPERATURE RF LINACS
G.A. Loew, S. Tantawi, J.W. Wang . . . . . . . . . . . . . . . . . . . . . . . . . . . . 507
6.14 HIGH VOLTAGE TECHNIQUE B. Goddard . . . . . . . . . . . . . . . . . . . . . . 511
6.15 COATING RECIPES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515
6.15.1 Recipes for Coating Windows R.M. Sundelin, H.L. Phillips . . . . . . . . . . 515
xiii
6.15.2 Recipes for Coating Ceramic and Metal Vacuum Chambers S.D. Henderson . 516
6.16 CAVITY MEASUREMENTS R. Rimmer, M. Tigner . . . . . . . . . . . . . . . . . . 521
6.16.1 Field Maps by Perturbation Methods . . . . . . . . . . . . . . . . . . . . . . . 521
6.16.2 Q and β Determination from Input Coupler . . . . . . . . . . . . . . . . . . . 522
6.17 MAGNETIC MEASUREMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523
6.17.1 Accelerator Magnets A.K. Jain, P. Wanderer . . . . . . . . . . . . . . . . . . 523
6.17.2 Insertion Device Measurement S. Marks, R.D. Schlueter . . . . . . . . . . . . 529
6.18 HIGH POWER SWITCHES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 533
6.18.1 DC Switches J. M. Sanders, T. Tang, M. A. Gundersen, G. Roth . . . . . . . . 533
6.18.2 Ultra-High-Power Rf Switches S. Tantawi . . . . . . . . . . . . . . . . . . . 540
7 SUBSYSTEMS 543
7.1 PARTICLE SOURCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 543
7.1.1 Electron Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 543
7.1.1.1 Thermionic cathodes H.G. Kirk . . . . . . . . . . . . . . . . . . . 543
7.1.1.2 Photocathodes (unpolarized) D.H. Dowell . . . . . . . . . . . . . . 544
7.1.1.3 Photocathodes (polarized) T. Maruyama, M. Poelker . . . . . . . . 546
7.1.1.4 DC thermionic guns and preinjectors H.G. Kirk . . . . . . . . . . . 547
7.1.1.5 Normal conducting rf photo guns D.H. Dowell . . . . . . . . . . . 548

7.1.1.6 Superconducting rf photo guns J. Sekutowicz . . . . . . . . . . . . 550
7.1.1.7 DC photo guns B. Dunham . . . . . . . . . . . . . . . . . . . . . . 552
7.1.1.8 Continuous duty preinjectors I. Bazarov . . . . . . . . . . . . . . . 553
7.1.2 Positron Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555
7.1.2.1 Tungsten targets S. Ecklund . . . . . . . . . . . . . . . . . . . . . . 555
7.1.2.2 Conversion of undulator radiation A. Mikhailichenko . . . . . . . . 558
7.1.3 Polarized Protons and Heavy Ions T.B. Clegg, W. Haeberli . . . . . . . . . . . 561
7.1.4 H− Ion Sources K.N. Leung . . . . . . . . . . . . . . . . . . . . . . . . . . . 563
7.1.5 Antiproton Production G. Dugan . . . . . . . . . . . . . . . . . . . . . . . . 565
7.1.6 Multi-Charged Heavy Ion Sources J. Alessi, A. Pikin . . . . . . . . . . . . . . 568
7.1.7 Foil Strippers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 571
7.1.7.1 Charge state strippers M.A. McMahan . . . . . . . . . . . . . . . . 571
7.1.7.2 Stripper foils for H− beams M.A. Plum . . . . . . . . . . . . . . . . 574
7.1.8 Lorentz Stripping of H− Ions M.A. Furman . . . . . . . . . . . . . . . . . . . 576
7.1.9 Laser-Assisted H− Conversion to Protons V. Danilov . . . . . . . . . . . . . . 576
7.2 CONFINEMENT AND FOCUSING . . . . . . . . . . . . . . . . . . . . . . . . . . . 577
7.2.1 Resistive Magnets F.E. Mills, D.J. Harding . . . . . . . . . . . . . . . . . . . 577
7.2.2 Consequences of Saturation of High Permeability Material K. Halbach,
R. Schlueter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 584
7.2.3 Special Topics in Magnetics . . . . . . . . . . . . . . . . . . . . . . . . . . . 585
7.2.3.1 Properties of 3D vacuum fields integrating along a straight line
K. Halbach, R. Schlueter . . . . . . . . . . . . . . . . . . . . . . . 585
7.2.3.2 Pole width necessary to obtain desired field quality in a 2-D magnet
K. Halbach. R. Schlueter . . . . . . . . . . . . . . . . . . . . . . . 585
7.2.3.3 Eddy currents K. Halbach, R. Schlueter . . . . . . . . . . . . . . . 586
7.2.3.4 Power dissipation in the dipole coils of a storage ring with iron poles
K. Halbach, R. Schlueter . . . . . . . . . . . . . . . . . . . . . . . 588
7.2.3.5 Alpha magnet H. Wiedemann . . . . . . . . . . . . . . . . . . . . . 588
7.2.4 Cosθ Superconducting Magnets P. Schmüser . . . . . . . . . . . . . . . . . . 588
7.2.5 Superferric Magnets A. Zeller . . . . . . . . . . . . . . . . . . . . . . . . . . 595
7.2.6 High Field Accelerator Magnets G. Sabbi . . . . . . . . . . . . . . . . . . . . 599
7.2.7 Kickers C. Burkhart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 602
7.2.8 Permanent Magnet Elements K. Halbach, R. Schlueter . . . . . . . . . . . . . 607
xiv
7.2.9 Electrostatic Separators J.J. Welch . . . . . . . . . . . . . . . . . . . . . . . . 614
7.2.10 Deflection and Crab Cavities K. Akai, M. Tigner . . . . . . . . . . . . . . . . 617
7.2.10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 617
7.2.10.2 Multicell deflection cavities . . . . . . . . . . . . . . . . . . . . . . 617
7.2.10.3 Crab cavity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 617
7.2.11 Electrostatic Lenses A. Faltens . . . . . . . . . . . . . . . . . . . . . . . . . 620
7.2.12 Lithium Lens G. Dugan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
7.2.13 Orbit Feedback Control Y. Tian, L.H. Yu . . . . . . . . . . . . . . . . . . . . . 624
7.2.13.1 Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624
7.2.13.2 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 626
7.2.13.3 Local orbit feedback . . . . . . . . . . . . . . . . . . . . . . . . . . 628
7.2.14 Feedback to Control Coupled-Bunch instabilities J.D. Fox . . . . . . . . . . . 628

7.2.14.1 Beam diagnostics via feedback signals . . . . . . . . . . . . . . . . 634
7.2.15 Beam Deflection and Collimation with Aligned Crystals R.A. Carrigan, Jr. . . 636
7.2.16 Septum Devices R. Keizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 639
7.2.17 Electron Lenses V. Shiltsev . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
7.2.18 Spin Manipulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
7.2.18.1 Siberian snake V. Ptitsyn . . . . . . . . . . . . . . . . . . . . . . . 641
7.2.18.2 Partial snake H. Huang . . . . . . . . . . . . . . . . . . . . . . . . 642

7.2.18.3 Spin rotator V. Ptitsyn . . . . . . . . . . . . . . . . . . . . . . . . . 643
7.2.18.4 RF spin rotator M. Bai, T. Roser . . . . . . . . . . . . . . . . . . . 643
7.3 ACCELERATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
7.3.1 RF System Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
7.3.1.1 RF system design for stability D. Boussard . . . . . . . . . . . . . 645
7.3.1.2 Low level RF J.D. Fox . . . . . . . . . . . . . . . . . . . . . . . . 649
7.3.2 Klystron Amplifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 651
7.3.2.1 Klystrons D. Sprehn . . . . . . . . . . . . . . . . . . . . . . . . . 651
7.3.2.2 Klystron amplifier systems A. Gamp, M. Tigner . . . . . . . . . . . 653
7.3.3 Tetrode Amplifiers J.M. Brennan . . . . . . . . . . . . . . . . . . . . . . . . 655
7.3.4 Inductive Output Tube J.M. Brennan, S. Belomestnykh . . . . . . . . . . . . . 659
7.3.5 Drift Tube Linacs J.M. Potter . . . . . . . . . . . . . . . . . . . . . . . . . . 660
7.3.6 Normal Conducting v = c Linac Structures G.A. Loew . . . . . . . . . . . . . 664
7.3.7 Inductively Loaded Accelerating Cavities M. Yoshii . . . . . . . . . . . . . . 667
7.3.8 Fixed Frequency Cavities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 670
7.3.8.1 Multicell cavities W. Schnell . . . . . . . . . . . . . . . . . . . . . 670
7.3.8.2 Single cell cavities R. Rimmer . . . . . . . . . . . . . . . . . . . . 672
7.3.9 Superconducting Cavities for vp = c Linacs, Storage Rings, & Synchrotrons
D. Proch, H. Padamsee . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 674
7.3.10 Superconducting Cavities for vp < c Linacs M.P. Kelly . . . . . . . . . . . . . 681
7.3.11 Superconducting Single Cell Cavities S. Belomestnykh, J. Kirchgessner . . . . 683
7.3.12 Millimeter-Wave Linacs D. Whittum . . . . . . . . . . . . . . . . . . . . . . 687
7.3.13 Plasma Accelerators E. Esarey, C. Schroeder . . . . . . . . . . . . . . . . . . 690
7.4 BEAM INSTRUMENTATION AND DIAGNOSTICS . . . . . . . . . . . . . . . . . 697
7.4.1 Composition — Ion Beams P. Forck, P. Strehl . . . . . . . . . . . . . . . . . 697
7.4.2 Longitudinal Phase Space Measurement — Ion Beams P. Forck, P. Strehl . . . 699
7.4.3 Heavy Ion Linacs — Emittance Measurements P. Forck, P. Strehl . . . . . . . 702
7.4.4 Charge State — Ion Beams P. Forck, P. Strehl . . . . . . . . . . . . . . . . . . 704
7.4.5 Beam Current Measurement J. Hinkson . . . . . . . . . . . . . . . . . . . . . 706
7.4.6 Beam Position Monitors J.A. Hinkson, S. Smith, G. Decker . . . . . . . . . . 709
7.4.7 Longitudinal Distribution Function, Electrons B. Schmidt . . . . . . . . . . . 713
7.4.7.1 Longitudinal diagnostics with coherent radiation . . . . . . . . . . . 713
7.4.7.2 Electro-optical bunch length monitors . . . . . . . . . . . . . . . . 714
xv
7.4.8 Transverse and Longitudinal Emittance Measurements J.T. Seeman . . . . . . 716
7.4.9 Transverse Beam Profile Measurements with Transition and Diffraction
Radiation P. Evtushenko . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 719
7.4.10 X-Ray Beam Size Monitor J.P. Alexander, D.P. Peterson . . . . . . . . . . . . 721
7.4.10.1 X-ray source and optics . . . . . . . . . . . . . . . . . . . . . . . . 721
7.4.10.2 Detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 721
7.4.10.3 Beam size measurements . . . . . . . . . . . . . . . . . . . . . . . 721
7.4.10.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 723
7.4.11 Streak Cameras J.M. Byrd . . . . . . . . . . . . . . . . . . . . . . . . . . . . 723
7.4.11.1 Principle of operation . . . . . . . . . . . . . . . . . . . . . . . . . 723
7.4.11.2 Temporal resolution . . . . . . . . . . . . . . . . . . . . . . . . . . 724
7.4.11.3 Applications to measurement of beam dynamics . . . . . . . . . . . 725

7.4.12 Laser Wire M. Ross . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 726
7.4.13 Laser Interference Methods T. Shintake . . . . . . . . . . . . . . . . . . . . . 728
7.4.14 Beam Size via Synchrotron Radiation T. Shintake . . . . . . . . . . . . . . . 730
7.4.15 Electron Cloud Measurements F. Caspers, F. Zimmermann . . . . . . . . . . 732
7.4.16 Beam Loss Monitors R. Jones, K. Wittenburg . . . . . . . . . . . . . . . . . . 736
7.4.16.1 Beam loss monitoring using ionisation detection . . . . . . . . . . . 737
7.4.16.2 Beam loss monitoring using light detection . . . . . . . . . . . . . . 739

7.4.17 Schottky Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 739
7.4.17.1 Transverse Schottky spectra and beam transfer functions
O. Boine-Frankenheim, V. Kornilov, S. Paret . . . . . . . . . . . . . 739
7.4.17.2 Longitudinal Schottky spectra and beam transfer function
E. Shaposhnikova . . . . . . . . . . . . . . . . . . . . . . . . . . . 742
7.5 IMPEDANCE DETERMINATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . 745
7.5.1 Bench Measurements F. Caspers . . . . . . . . . . . . . . . . . . . . . . . . 745
7.5.2 Beam-Based Characterization of Coupled Bunch Instabilities J.M. Byrd . . . . 750
7.5.2.1 Passive techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . 750
7.5.2.2 Active techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . 751
7.5.3 Other Beam Based Methods to Measure Impedances J. Gareyte, J. S. Berg . . 752
7.5.4 Direct Wakefield Measurement W. Gai, C. Jing . . . . . . . . . . . . . . . . . 754
7.6 POLARIMETERS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 756
7.6.1 Lepton Polarimeters E. Chudakov, D. Gaskell, J. Grames, M. Woods . . . . . 756
7.6.1.1 Mott polarimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 756
7.6.1.2 Møller polarimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . 756
7.6.1.3 Compton polarimetry . . . . . . . . . . . . . . . . . . . . . . . . . 757
7.6.2 Proton Polarimeters Y. Makdisi . . . . . . . . . . . . . . . . . . . . . . . . . 758
7.7 CONTROLS AND TIMING K. Rehlich . . . . . . . . . . . . . . . . . . . . . . . . . 760
7.8 FEMTOSECOND PRECISION OPTICAL SYNCHRONIZATION F. Löhl . . . . . . 763
8 RADIATION EFFECTS AND PROTECTION 767

8.1 RADIATION PROTECTION PRINCIPLES S. Roesler, M. Silari . . . . . . . . . . . . 767
8.1.1 Quantities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 767
8.1.1.1 Physical quantities . . . . . . . . . . . . . . . . . . . . . . . . . . . 767
8.1.1.2 Protection quantities . . . . . . . . . . . . . . . . . . . . . . . . . . 767
8.1.1.3 Operational quantities . . . . . . . . . . . . . . . . . . . . . . . . . 768
8.1.1.4 Dose conversion coefficients . . . . . . . . . . . . . . . . . . . . . 768
8.1.2 Health Effects of Ionizing Radiation . . . . . . . . . . . . . . . . . . . . . . . 768
8.1.3 Radiation Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 769
8.2 SOURCES OF RADIATION R.H. Thomas,S. Roesler, M. Silari . . . . . . . . . . . . 769
8.2.1 Electron Beams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 770
8.2.1.1 Bremsstrahlung . . . . . . . . . . . . . . . . . . . . . . . . . . . . 770
xvi
8.2.1.2 Electromagnetic cascade . . . . . . . . . . . . . . . . . . . . . . . 771
8.2.2 Proton Beams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 773
8.2.3 Ion Beams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 774
8.3 MONTE CARLO CODES S. Roesler, M. Silari . . . . . . . . . . . . . . . . . . . . . 776
8.3.1 FLUKA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 776
8.3.2 GEANT4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 776
8.3.3 MARS15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 777
8.3.4 MCNPX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 778
8.3.5 PHITS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 778
8.4 RADIATION TRANSPORT AND SHIELDING S. Roesler, M. Silari . . . . . . . . . 779
8.4.1 General Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 779
8.4.2 Semi-Empirical Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 779

8.4.3 The Monte Carlo “hybrid” Method . . . . . . . . . . . . . . . . . . . . . . . . 780
8.4.4 Photon Transmission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 781
8.4.5 Ducts and Labyrinths for Neutron Attenuation . . . . . . . . . . . . . . . . . 782
8.4.6 Ducts and Labyrinths for Photon Attenuation . . . . . . . . . . . . . . . . . . 783
8.4.7 Shielding Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 784
8.4.8 Muons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 785
8.4.9 Skyshine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 785

8.5 RADIATION MEASUREMENTS S. Roesler, M. Silari . . . . . . . . . . . . . . . . . 786
8.5.1 Neutron Dosimetry and Spectrometry . . . . . . . . . . . . . . . . . . . . . . 786
8.5.1.1 Rem counters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 786
8.5.1.2 Bonner sphere spectrometers . . . . . . . . . . . . . . . . . . . . . 787
8.5.1.3 Bubble detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 787
8.5.1.4 Track etched detectors . . . . . . . . . . . . . . . . . . . . . . . . . 788
8.5.2 Photon Dosimetry and Spectrometry . . . . . . . . . . . . . . . . . . . . . . . 788
8.5.2.1 BC501A liquid scintillator . . . . . . . . . . . . . . . . . . . . . . 789
8.5.2.2 BGO photon spectrometer . . . . . . . . . . . . . . . . . . . . . . . 789
8.5.2.3 Geiger Müller counter . . . . . . . . . . . . . . . . . . . . . . . . . 789
8.5.3 Detectors Based on Micro-Dosimetric Principles . . . . . . . . . . . . . . . . 789
8.6 INDUCED RADIOACTIVITY S. Roesler, M. Silari . . . . . . . . . . . . . . . . . . 790
8.6.1 Fundamental Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 790
8.6.2 Solid Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 791
8.6.3 Liquids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 792
8.6.4 Air . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 793
8.7 RADIATION DAMAGE THRESHOLDS H. Schönbacher . . . . . . . . . . . . . . . 793
8.7.1 Organic Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 793
8.7.2 Semiconductors and Electronic Devices . . . . . . . . . . . . . . . . . . . . . 797
8.7.3 Ceramics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 799
8.7.4 Vitreous Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 799
8.7.5 Metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 801
8.7.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 801
Author index I-1
Subject Index I-5
xvii
Chapter 1. INTRODUCTION
by UNIVERSITY OF BIRMINGHAM LIBRARY - INFORMATION SERVICES on 03/21/15. For personal use only.
1.1 HOW TO USE THIS BOOK NC normal conducting

RFQ rf quadrupole
This is an accelerator designer’s and operator’s rms root mean square
handbook of formulae, tables, figures and refer- RRR residual resistivity ratio
SC superconducting
ences. It is meant to be a concise working tool. SR synchrotron radiation
An effort has been made to provide an index SRF superconducting rf
which is as complete as possible. SW, TW standing wave, traveling wave
Each subsection (e.g. 2.3.4) is treated as a UHV ultrahigh vacuum
unit which is more or less self-contained. Num- Abbreviations of references
bering of all figures and tables are reset at each APAC Asian PAC
subsection, and references are found following APL Applied Physics Letters
each subsection. References are not meant to be EPAC European PAC
exhaustive but represent the experts’ recommen- IJMP International Journal of Modern Physics
dation about a reliable place to begin. IPAC International PAC
While the linear and circular accelerators for JAP Journal of Applied Physics
JETP Journal of Experimental and Theoretical Physics
high energy physics and synchrotron radiation ap- JINST Journal of Instrumentation
plications are our primary concern, we have tried JVST Journal of Vacuum Science and Technology
to provide connections to other types of accelera- NIM Nuclear Instruments & Methods
tors in the glossary section 1.6. NJP New Journal of Physics
PA Particle Accelerators
PAC Particle Accelerator Conference
1.2 NOMENCLATURE PL Physics Letters
PR Physical Review
Boldface symbol means matrix quantity. At is PRL Physical Review Letters
the transpose of A. A tilde, Ã, means Fourier PRST-AB Physical Review Special Topics –
is a vector. â is a unit vector.
transform. A Accelerators & Beams
√
For complex numbers, we use i = −1, RAST Reviews of Accelerator Science & Technology
and a sinusoidal time dependence is described by RMP Reviews of Modern Physics
e−iωt unless otherwise noted. RSI Review of Scientific Instruments
Abbreviations Tabulated below are symbols adopted through-

ac, dc alternating current, direct current out this Handbook unless otherwise noted. More
BBU beam break-up (instability) symbols are defined in the table of fundamental
BPM beam position monitor
CM center of mass
constants in Sec.1.3. and locally in the text.
c.w. continuous wave Symbol Quantity
EM electromagnetic αp momentum compaction factor
FEL free electron laser αT thermal expansion coefficient
FWHM full-width at half-maximum αx,y horizontal (x), vertical (y) Courant-
HOM higher order mode Snyder alpha functions
IR, UV infrared, ultraviolet B brightness
IP, IR interaction point, region
B magnetic induction
LHe liquid helium (Bρ) = P/e, rigidity of a particle of charge e
l.h.s., r.h.s. left-hand side, right-hand side (Bρ) [T-m] = 3.335641 P0 [GeV/c]
n-D n-dimension(al)
β, β = (v, v )/c, (speed, velocity) relative
rf radio frequency to light
1
Sec.1.2: NOMENCLATURE
βx,y horizontal (x), vertical (y) Ω solid angle
Courant-Snyder beta functions P,P ,p,
p particle momentum
C circumference P0 , p0 design particle momentum
c v , cp specific heat P power

D displacement current Q quality factor of oscillator
δ = ΔP/P0 , relative momentum error q charge on a particle
Dx,y horizontal (x), vertical (y) R = C/(2π), average radius
dispersion functions ρ bending radius
Dx,y beam-beam disruption parameters ρ volume density

detA determinant of matrix A ρr resistivity
E Young’s modulus s longitudinal coordinate
E particle total energy along an accelerator
E0 design particle energy σ interaction cross-section

E0 = mc2 rest mass energy σc = 1/ρr , conductivity

E electric field σx,y,z horiz. (x), vert. (y),
x,y,L horizontal (x), vertical (y), long. (z) rms beam size
longitudinal (L) emittances σx ,y horiz. (x ), vert. (y )
(unnormalized) rms angular spread
N x,N y = βγx,y , normalized emittances σE,δ energy (E), relative
η = αp − (1/γ 2 ), phase slip factor momentum (δ) rms spread
f0 revolution sgn(x) sign function,
frequency
γ = 1/ 1 − β 2 , relativistic factor = 1 if x > 0, = −1 if x < 0
γx,y horizontal (x), vertical (y) t time
Courant-Snyder gamma functions T temperature
γt transition gamma T kinetic energy
H Hamiltonian trA trace of matrix A
H synchrotron integral H-function T0 revolution period
h rf harmonic number U0 synchrotron radiation loss

H magnetic field per revolution
H(x) step function, Υ beamstrahlung parameter
= 1 if x > 0, = 0 if x < 0 v, v speed, velocity
I0 = NB ef0 , average bunch current vg group velocity
Ib peak bunch current vp phase velocity
Jx,y,s horizontal (x), vertical (y), Vrf rf voltage
synchrotron (s) partition numbers Wm , W⊥m longitudinal, transverse
⊥
J current density (or Wm ,Wm ) wake function of mode m
k = 2π/λ, wave number (W if m = 0, W⊥ if m = 1)
κ thermal conductivity x horiz. displacement
L luminosity x = dx/ds, horiz. angular
λ wavelength deviation of a particle
μ Poisson’s ratio X0 radiation length
μ betatron phase advance per ξx,y = (dνx,y /dδ), horizontal (x),
turn or period vertical (y) chromaticity
N total no. of particles in beam y vert. displacement
NB number of particles per bunch y = dy/ds, vert. angular
in a bunched beam deviation of a particle
nB number of bunches in beam ψ distribution density in phase
νx,y,s horizontal betatron (x), space, normalized to unity
vertical betatron (y), ψx,y horiz. (x), vert. (y)
synchrotron (s) tune betatron phase
νsp spin tune z long. displacement of a particle
ω = 2πf , angular frequency relative to synchronous particle
ω0 = 2πf0 , angular revolution freq. (z > 0 ahead, z < 0 behind)
ωx,y,s = νx,y,s ω0 , betatron (x, y), Zm , Z⊥m longitudinal, transverse
⊥
synchrotron (s) angular freq. (or Zm ,Zm ) impedance of mode m
(Z if m = 0, Z⊥ if m = 1)
2
Ch.1: INTRODUCTION
1.3 FUNDAMENTAL CONSTANTS [1]
Quantity Symbol Value

pi π 3.141592653589793238
exponential constant e 2.718281828459045235
Euler’s constant γ 0.5772156649
speed of light c 2.99792458 E8 m s−1 (exact)
4π E-7 Henry m−1 (exact)
permeability of vacuum μ0
permittivity of vacuum 0 = 1/(μ0 c2 ) 8.854187817 E-12 Farad m−1
electronic charge e 1.6021765 E-19 C = 4.8032043 E-10 esu
Planck constant h 6.626069 E-34 J s
reduced Planck constant = h/(2π) 1.054572 E-34 J s = 6.582119 E-16 eV s

Boltzmann constant kB 1.380650 E-23 J K−1
Avogadro number NA 6.022142 E23 mole−1
gravitational constant G 6.67428 E-11 Newton m2 kg−2
std. grav. accel. g 9.80665 m s−2
electron mass me 9.1093822 E-31 kg
proton mass mp 1.6726216 E-27 kg
rest mass energy of
electron me c2 0.51099891 MeV
proton mp c2 938.27201 MeV
neutron mn c2 939.5653 MeV
deuteron md c2 1875.6128 MeV
muon mμ c2 105.65837 MeV
Z-particle mZ c2 91.188 GeV
W -particle mW c2 80.399 GeV
anomalous gyromagnetic ratio G = (g − 2)/2
electron 0.00115965219
muon 0.001165923
proton 1.79284739
deuteron -0.1429878
fine structure constant αF = e2 /(4π0 c) 1/137.0359997

impedance of free space Z0 = μ0 /0 = μ0 c 376.7303 Ω
classical radius of
electron re = e2 /(4π0 me c2 ) 2.8179403 E-15 m
proton rp = e2 /(4π0 mp c2 ) 1.534698 E-18 m
electron Compton wavelength λe = h/(me c) 2.4263106 E-12 m
λ̄e = λe /(2π) 0.3861593 E-12 m
Alfvén current IA = ec/re 17.045093 kA
Bohr radius a∞ = 4π0 2 /(me e2 ) 5.29177209 E-11 m
Thomson cross section σT = (8π/3)re2 6.65246 E-29 m2
Bohr magneton μB = e/(2me c) 5.7883818 E-5 eV/Tesla
nuclear magneton μN = e/(2mp c) 3.1524512 E-8 eV/Tesla
Stefan-Boltzmann constant σSB = (π 2 /60)kB
4 /(3 c2 ) 5.67040 E-8 W m−2 K−4
gas constant R = NA kB 8.3145 J K−1 mole−1
References
[1] K. Nakamura et al. (Particle Data Group), J. Phys. G37, 075021 (2010)
3
Sec.1.4: UNITS AND CONVERSIONS
1.4 UNITS AND CONVERSIONS 1.4.2 Conversions

M. Tigner, Cornell U.
1.4.1 Units
A.W. Chao, SLAC key:
Quantity, symbol, name in SI, Abbr., [dim.]
We use the SI (Système International, MKSA)
Quantity to be converted, conversion factor
units throughout this Handbook unless otherwise
noted. Table below gives the conversion of vari-
Quantity in SI = Quantity to be converted ×
ous physical quantities from Gaussian to SI unit

systems [1]. conversion factor, E1 = 10, E2 = 100, etc.
Quantity Gaussian SI
Length, , meter, m, [L]
speed of light c √1 inch, in. 2.54E-2

μ0 0
charge q √q foot, ft. 0.3048
4π0
charge density ρ √ρ angström, Å 1.0E-10
4π0
√I
fermi, fm 1.0E-15
current I 4π0 light year, ly 0.946E16

current density J √J Area, A, sq. meter, m2 , [L2 ]
√4π0
Φ sq. inch, in2 6.4516E-4
scalar potential 4π0 Φ
vector potential
A 4π
A sq. foot, ft2 9.2903E-2
√ μ0 acre 4.0468E3
voltage V 4π0 V
√
hectare, ha 1E4
electric field E 4π0 E sq. mile, mi2 2.5888E6
displacement
D 4π
D barn, bn 1E-28
0 Volume, V , cu. meter, m3 , [L3 ]

B 4π
magnetic induction μ B cu. inch, in3 1.639E-5
√ 0
magnetic field H 4πμ0 H gallon, gal. (liquid) 3.785E-3
conductivity σc σc cu. ft., ft3 2.832E-2
4π0
dielectric constant /0 Mass, M , kilogram, kg, [M]
permeability μ μ slug 14.59
μ0
resistance R 4π0 R Density, ρ, kg/m3 , [M/L3 ]
inductance L 4π0 L slug/cu ft 515.4
C pound/cu in, lb/in3 2.768E4
capacitance C 4π0 Time, t, second, s, [T]
Table below gives some numerical conver- year, yr. 3.156E7
sions between Gaussian and SI units. Speed, v, meter/sec, m/s, [L/T]
Quantity Gaussian SI
foot/sec, ft/s 0.3048
mile/hour, mi/hr 0.4470
Conductivity 8.9876E9 s−1 =1 mho/m Force, F , Newton, N, [ML/T2 ]
Resistance 1.1127E-12 s/cm =1 ohm dyne 1.0E-5
Capacitance 8.9876E11 cm =1 farad pound, lb 4.448
Inductance 1.1127E-12 =1 henry Pressure, P , Pascal, Pa, N/m2 , [M/T2 L]
If P is power and V is voltage, then atmosphere, atm 1.013E5
dB = 10 log10 (P1 /P2 ), or 20 log10 (V1 /V2 ) bar 1.0E5
Other notations: dBm (reference point 1 milli- dyne/cm2 0.1
watt); dBc (reference to the carrier power); dBa is pound/in2 , psi 6.895E3
used in acoustics in reference to a standard sound in. H2 O @ 4◦ C 2.491E2
pressure of 2E-4 microbar. Torr, mmHg @ 0◦ C 1.3332E2
Energy, W , Joule, J, [ML2 /T2 ]
References BTU 1055
erg 1.0E-7
[1] J.D. Jackson, Classical Electrodynamics, 3rd ed., foot-pound, ft lb 1.356
Wiley (1999) horsepower hour 2.685E6
4
Ch.1: INTRODUCTION
calorie, cal 4.186 Capacitance, C, farad, f, [T2 /L]
kilowatt hour, kWhr 3.6E6 abfarad 1.0E9
electron volt, eV 1.602E-19 statfarad 1.112E-11
liter atmosphere 101.31 Inductance, L, henry, hy, [L]
Power, P , watt, W, [ML2 /T3 ] abhenry 1.0E-9
erg/s 1.0E-5 stathenry 8.9874E11
BTU/hr 0.2930 Activity, bequerel, Bq, [1/T]
foot pound/sec, ft lb/s 1.356 Curie, Ci 3.7E10
horsepower 745.7 Dose, gray, Gy, [L2 /T2 ]

calorie/sec, cal/s 4.186 rad 1.0E-2
Thermal conductivity, κ, W/m K,[ML/T3 ] Exposure, see Ch.8
Watt/cm K 1.0E2 Temperature, T , Kelvin, K

BTU/hr/ft2 /◦ F/ft 1.73 T[◦ C]=T[K] - 273.16
cal/cm·s 418.6 T[◦ F]= 9/5(T[◦ C]) + 32
Specific heat, cv,p , J/kg K, [L2 /T2 ]
cal/gm ◦ C 4.186E3 1.5 FUNDAMENTAL FORMULAE
BTU/lb ◦ F 4.186E3 A.W. Chao, SLAC
Viscosity, h, kg/m·s, [M/LT]
poises, g/cm·s 0.1 1.5.1 Special Functions
slug/ft·s 4.79E1 Error function erf(x), erfc(x) = 1− erf(x):
x
Charge, q, Coulomb, C, [M1/2 L1/2 ] 2 2
abcoulomb 10 erf(x) = √ dt e−t , erf(∞) = 1

π 0
statcoulomb 3.336E-10
Bessel functions Jν (x), Iν (x), Nν (x), Kν (x):
Current, I, Ampere, A, [M1/2 L1/2 /T]
d 2 Jν ν2
dx2 + x dx + (1 − x2 )Jν = 0
1 dJν
abampere 10
∞
statampere 3.336E-10 e−ix cos θ = n=−∞ i−n Jn (x)einθ
Potential, V , volt, V, [M1/2 L3/2 /T2 ] 2π
Jn (x) = 2π 1
dθ e−inθ+ix sin θ
abvolt 1.0E-8
0
in 2π
statvolt 2.9979E2 = 2π 0 dθ einθ−ix cos θ
Elec. Field, E, volt/m, V/m, [M1/2 L1/2 /T2 ] In (x) = i−n Jn (ix)
abvolt/cm 1.0E-6 Jn (−x) = J−n (x) = (−1)n Jn (x)
statvolt/cm 2.9979E4
In (−x) = (−1)n I−n (x) = (−1)n In (x)
Mag. Field, H, A-turn/m, [M1/2 /L1/2 T]
Jn (x) ≈ In (x) ≈ n! 1
(x/2)n for |x| 1
Oersted 79.58
Jν (x) cos νπ−J−ν (x)
Displacement, D, Coul/m2 , [M1/2 /L3/2 ] Nν (x) = sin νπ
abcoulomb/cm2 7.958E3 Kν (x) = 2 sinπ νπ [I−ν (x) − Iν (x)]
statcoulomb/cm2 2.654E-7 J0 = −J1 , N0 = −N1
Mag. flux, Φ, Weber, volt-s, [M1/2 L3/2 /T] I0 = I1 , K0 = −K1
Tesla m2 1 For x 1,
maxwell, abvolt-s 1.0E-8
statweber, statvolt-s 2.9979 Jν (x) ≈ 2
πx cos(x − π2 ν − π4 )

Flux density, B, Tesla,T, [M1/2 /L1/2 T]
Nν (x) ≈ 2
πx sin(x − π2 ν − π4 )
Gauss 1.0E-4
esu 2.9979E6 Iν (x) ≈ √1
2πx
ex , Kν (x) ≈ √1
2πx
e−x
Conductivity, σc , siemens/m, S/m [T/L2 ] For roots of Jn and Jn , see Sec.6.6.
mho/cm 1.0E2 Gamma function Γ(x):
Resistivity, ρr , ohm meter, Ω-m, [L2 /T] ∞
microhm cm 1.0E-8 Γ(x > 0) = 0 dt e−t tx−1
Resistance, R, ohm, Ω, [L/T] Γ(n + 1) = n!, Γ(x + 1) = xΓ(x)
abohm 1.0E-9 Γ(x)Γ(1 − x) = sinππx
statohm 8.987E11 √
Γ(1/2) = π, Γ(1/4) ≈ 3.626
5
Sec.1.5: FUNDAMENTAL FORMULAE
1.5.2 Curvilinear Coordinate Systems Spherical:
General Orthogonal System (u1 , u2 , u3 ) [1]: ds2 = dr 2 + r 2 dθ 2 + r 2 sin2 θdφ2

System u1 u2 u3 h1 h2 h3 dV = r 2 sin θdrdθdφ
Cartesian x y z 1 1 1 ∇ψ = ∂ψ 1 ∂ψ 1 ∂ψ
∂r r̂ + r ∂θ θ̂ + r sin θ ∂φ φ̂
Cylindrical r θ z 1 r 1
∇·A = 12 ∂ (r 2 Ar ) + 1 ∂ (sin θAθ )
r ∂r r sin θ ∂θ
Spherical r θ φ 1 r r sin θ
1 φ∂A
x + r sin
Frenet-Serret x y s 1 1 1+ ρ θ ∂φ

∇×A = 1 r̂ ∂ (sin θAφ ) − ∂Aθ
r sin θ ∂θ ∂φ
ds ≡ h1 du1 û1 + h2 du2 û2 + h3 du3 û3
θ ∂φ − r ∂r (rAφ )
1 ∂Ar 1 ∂
ds2 = h21 du21 + h22 du22 + h23 du23 +θ̂ r sin
dV = h1 h2 h3 du1 du2 du3 ∂

+ 1r φ̂ ∂r (rAθ ) − ∂A r
∇ψ = h11 ∂u ∂ψ
û1 + h12 ∂u∂ψ ∂ψ
û2 + h13 ∂u û3 ∂θ
1 2 3
∇2 ψ = r12 ∂r
∂
r 2 ∂ψ 1 ∂
∂r + r 2 sin θ ∂θ sin θ ∂θ
∂ψ
∇·A
= 1 ∂ ∂
h1 h2 h3 ∂u1 (h2 h3 A1 ) + ∂u2 (h3 h1 A2 ) 2
1
+ r2 sin ∂ ψ
2 θ ∂φ2
+ ∂u∂ 3 (h1 h2 A3 )
∇×A = 1 × References
h1 h2 h3
h1 û1 ∂u∂ 2 (h3 A3 ) − ∂u∂ 3 (h2 A2 ) [1] J. Murphy, Synchrotron Light Source Data Book,
BNL 42333 (version 3.0) (1993)
+h2 û2 ∂u∂ 3 (h1 A1 ) − ∂u∂ 1 (h3 A3 )

1.5.3 Electromagnetism
+h3 û3 ∂u∂ 1 (h2 A2 ) − ∂u∂ 2 (h1 A1 )
∇·B =0
∂ψ
∇2 ψ = h1 h12 h3 ∂u∂ 1 hh2 h1 3 ∂u

1
∇·D =ρ

∂ψ h1 h2 ∂ψ
+ ∂u∂ 2 hh3 h2 1 ∂u 2
+ ∂
∂u3 h3 ∂u3 ∇×H − ∂D = J ∂t

∇×E + ∂B = 0
∂t
Cartesian:
D = E,
B = μH

ds2 = dx2 + dy 2 + dz 2
dV = dxdydz = ∇ × A,
B E = −∇Φ − ∂ A
∂t
∇ψ = ∂ψ ∂ψ ∂ψ
∂x x̂ + ∂y ŷ + ∂z ẑ
Continuity ∂ρ + ∇ · J = 0
∂t
q1 q2
∇·A = ∂Ax + ∂Ay + ∂Az static Coulomb force F = 4π 0r
2 r̂
∂x
∂y ∂z + v × B)
= x̂ ∂Az − ∂Ay + ŷ ∂Ax −
∇×A ∂Az Lorentz force F = q(E
∂y ∂z ∂z ∂x
energy density u = 1 (D ·E
+H
2
· B)
∂Ay
+ẑ ∂x − ∂y ∂Ax
×H
momentum density g = c12 E
∂2ψ ∂2ψ ∂2ψ
∇2 ψ = ∂x2
+ ∂y 2
+ ∂z 2
=E
Poynting vector S ×H
Cylindrical: Steady-state boundary conditions between two

media:
=ds2 + dr 2 r 2 dθ 2
+ dz 2 2 − D
(D 1 ) · n̂ = ρsurface
dV = rdrdθdz
2 − B
(B 1 ) · n̂ = 0
∇ψ = ∂ψ 1 ∂ψ ∂ψ
∂r r̂ + r ∂θ θ̂ + ∂z ẑ
(E2 − E 1 ) × n̂ = 0
∇·A = 1 ∂ (rAr ) + 1 ∂Aθ + ∂Az
r ∂r r ∂θ ∂z
(H2 − H1 ) × n̂ = −Jsurface

∇×A
= r̂ 1 ∂Az − ∂Aθ + θ̂ ∂Ar − ∂Az
r ∂θ ∂z ∂z ∂r
∂ where ρsurface is the surface charge density;
+ 1r ẑ ∂r (rA ) − ∂Ar Jsurface is the surface current on the boundary;
θ 2 ∂θ 2
n̂ is the unit vector normal to the boundary and
∇2 ψ = 1r ∂r
∂
r ∂ψ 1 ∂ ψ ∂ ψ
∂r + r 2 ∂θ 2 + ∂z 2 points into medium 2.
6
Ch.1: INTRODUCTION
1.5.4 Kinematical Relations

Relations between β, cp, rest mass energy E0 , kinetic energy T , E, and γ [1]:
β cp T E γ

β= β ep
E =
√ cp/E0
1 − (1 + E0 )
T −2 1−(E ) = E
E 0 2 cp
1 − γ −2
(cp/E0 )2 +1

cp = E0 / β −2 − 1 cp [T (2E0 + T )]1/2 E 2 − E02 E0 γ 2 − 1
1/2
= Eβ = T γ+1 γ−1 = Eβ

E0 = cp/βγ cp(γ − 1)
2 −1/2 T /(γ − 1) E 2 − c2 p2 E/γ
= E(1 − β 2 )1/2

T = [√ 1
− 1]E0 E02 + c2 p2 − E0 T E − E0 E0 (γ − 1)
1−β 2
1/2
= cp γ−1
γ+1
γ= (1 − β 2 )−1/2 cp
E0 β = [1 − ( Ecp0 )2 ]1/2 1 + T /E0 E/E0 γ
First derivatives:
dβ d(cp) dγ = dE/E0 = dT /E0
dβ = dβ [1+(cp/E0 )2 ]−3/2 d(cp)/E0 γ −2 (γ 2 − 1)−1/2 dγ
= γ −3 d(cp)/E0 = β −1 γ −3 dγ
d(cp) = E0 (1−β 2 )−3/2 dβ = E0 γ 3 dβ d(cp) √
E0 γdγ
= E0 dγ
β
2 γ −1
dγ = dE/E0 β(1 − β 2 )−3/2 dβ [1+(E0 /cp)2 ]−1/2 d(cp)/E0 dγ
= dT /E0 = = βγ 3 dβ = βd(cp)/E0
Logarithmic first derivatives:
dβ/β dp/p dT /T dE/E = dγ/γ
dβ/β = dβ/β γ −2 dp/p [γ(γ + 1)]−1 dT /T (γ 2 − 1)−1 dγ/γ
= dp/p − dγ/γ = (βγ)−2 dγ/γ
dp/p = γ 2 dβ/β dp/p [γ/(γ + 1)]dT /T β −2 dγ/γ
dT /T = γ(γ + 1)dβ/β (1 + γ −1 )dp/p dT /T γ(γ − 1)−1 dγ/γ
dE/E = (βγ)2 dβ/β β 2 dp/p (1 − γ −1 )dT /T dγ/γ
dγ/γ = = (γ 2 − 1)dβ/β = dp/p − dβ/β
Relations between bending magnetic field B, Variables differential relation

dp 2 dR dB
momentum p, mean orbit radius R, and rev- (B, p, R) p = γt R + B
olution frequency f : Of the four quantities (f, p, R) dp 2 df 2 dR
p =γ f +γ R
(B, p, R, f ), only two can be independently γ −γt2 dp
2
(B, f, p) dB 2 df
chosen, yielding the table on the right (γt is B = γt f + γ 2 p
2 df
B = γ f + (γ − γt ) R
the transition gamma). dB 2 2 dR
(B, f, R)
References
[1] C. Bovet et al., CERN/MPS-SI/Int. DL/70/4 (1970)
7
Sec.1.6: GLOSSARY OF ACCELERATOR TYPES
1.5.5 Vector Analysis Energy-momentum :

a · (b × c) = b · (c × a) = c · (a × b) Px = γ(Px − βc E), E = γ(E − cβPx )

Py = Py , Pz = Pz
a × (b × c) = (a · c)b − (a · b)c
(a × b) · (c × d)
= (a · c)(b · d)
− (a · d)( b · c) EM fields :
∇ × ∇ψ = 0 Ex = Ex , Bx = Bx
∇ · (∇ × a) = 0 Ey = γ(Ey − cβBz ), By = γ(By + βc Ez )

∇(φψ) = φ∇ψ + ψ∇φ Ez = γ(Ez + cβBy ), Bz = γ(Bz − βc Ey )

∇ × (∇ × a) = ∇(∇ · a) − ∇2a
∇ · (ψa) = a · ∇ψ + ψ∇ · a
∇ × (ψa) = ∇ψ × a + ψ∇ × a
1.6 GLOSSARY OF ACCELERATOR
∇(a · b) = (a · ∇)b + (b · ∇)a TYPES
+a × (∇ × b) + b × (∇ × a)
∇ · (a × b) = b · (∇ × a) − a · (∇ × b) 1.6.1 Antiproton Sources
K. Gollwitzer, J. Marriner, FNAL
∇ × (a × b) = a(∇ · b) − b(∇ · a)
Antiproton (p̄) sources are complete accelerator
+(b · ∇)a − (a · ∇)b
complexes utilizing many accelerator technolo-
a × (b × c) + b × (c × a) + c × (a × b) = 0

gies [1]–[4]. A primary proton beam is used to
produce p̄’s on a target. The production process
1.5.6 Relativity is inefficient, and the secondary p̄ beam is several
Let F be the stationary laboratory frame with orders of magnitude less dense than the primary
space time coordinates (x, t). Let F with (x , t ) proton beam. The design strategy consists of
= cβ with maximizing the p̄ phase space density by (a) ap-
be a frame moving with velocity V
propriate preparation of the proton beam and (b)
respect to F . Lorentz transformations:
beam cooling using some combination of stochas-
Coordinates : tic cooling (Sec.2.7.1) and electron cooling
(Sec.2.7.2).
x = x + γ β γ
γ+1 β · x − ct It is advantageous for the proton beam to have
t = γ(t − 1c β · x) a short time spread and a small spot size at the p̄

target since these properties are transferred to the
γ v ·V
v +γ V 2 −1
Velocity : v = γ+1 c secondary p̄ beam. The proton beam can obtain
γ 1− V 2·v a short bunch length by a bunch rotation process
c
Energy-momentum : (Sec.4.10). The rf voltage is lowered so that the

beam fills the bucket and then raised to its maxi-
P = P + γ β γ+1
γ
β · P − 1c E mum value. The bunch rotates 90◦ in phase space
E = γ(E − cβ · P ) and achieves a momentary short length. The pro-
ton beam is strongly focused transversely to a
EM fields :
minimum spot size at the target; see Fig.1. The
= γ(E
E − γ 2 (β · E)
+ cβ × B) β resulting beam focal length is comparable to typ-
γ+1
= γ(B
− 1 β × E)
− γ2 β ical nuclear interaction lengths, favoring the use
B c γ+1 (β · B) of high-Z targets, but excessive energy deposition
= V x̂, the above becomes from pair production limits their usefulness. Tab.1
When V
shows typical proton beam parameters.
Coordinates : The production cross-section has been param-
x = γ(x − V t), t = γ(t − Vx
c2
) eterized as [5]

y = y, z = z E d3 σ
vx −V = [0.065(1 − xr )8 exp(−3p2t )]
Velocity : vx = βvx σabs dp3
1−
× 1 + 24s−2 exp (8xr )
c
vy vz =
vy /γ vz /γ
=
1− βvcx
,
1− βvcx

× a exp bp2t exp (−cxr )
8
Ch.1: INTRODUCTION
Table 2: Beam parameters of the collected p̄ beam.
6π is for 95% beam, un-normalized. Δp/p is the full
momentum spread of the beam.
Parameter FNAL CERN
(Debunch.) (AD)
Momentum (GeV/c) 8.9 3.57
Figure 1: Antiproton Targeting. The proton beam is Δp/p (%)
given a short time spread via bunch rotation and fo- Before rotation 4.5 6
cused to a small spot size on the production target. The

After rotation 0.2 1.5
secondary beam is focused by a lithium lens or horn.
After cooling 0.1 0.18
6π (mm-mrad)
Table 1: Proton beam parameters for the FNAL Main Before cooling 35π 200π
Injector and CERN Proton Synchrotron. After cooling 3π 5π
Parameter FNAL CERN Stochastic cooling
(MI) (PS) bandwidth (GHz) 4-8 0.9-1.6
Momentum (GeV/c) 120 26
Protons/pulse (1012 ) 8 15
Cycle time (s) 2.2 100
# Bunches 81 5
RF bucket length (ns) 19 105
RF bunch length (ns) 1.5 5
Beam radius σ
at target (mm) 0.16 1.5
with σabs the absorption cross-section, pt the

transverse momentum [GeV/c], xr the radial scal-
ing variable (= E/Emax , p̄ energy in CM di- Figure 2: Antiproton Stacking. A small momentum
vided by its kinematically allowed value), s the spread is recovered from the secondary beam via a
bunch rotation and the beam is injected into an ac-
square of the CM energy [GeV2 ]. The parame-
cumulation ring where the longitudinal density is in-
terization consists of three multiplicative factors:
creased by stochastic stacking. Individual particles
(i) a cross section for a hydrogen target at infi-
are stochastically decelerated towards the stack core,
nite energy, (ii) deviations from scaling (including
where high density beams can be formed and ex-
s-dependence), and (iii) the nuclear dependence. tracted.
Values for a, b, c for various nuclei are given in
[5]. For copper, a = 1.50, b = 1.43, c = 1.56. where W = fmax −fmin is the system bandwidth,
The initial large angular divergence of the p̄ β is the relativistic velocity factor, E is the beam
beam is largely eliminated by a collection ele- energy, η = slip factor, Λ = ln (fmax /fmin ) ,
ment: a high-gradient lithium lens (Sec.7.2.12) or and E1d = V1 dV dE with V (E) the voltage gain per
a pulsed horn. The large momentum spread is return. In [6], V is exponentially decreasing with
duced as the beam is debunched through a sec- E while the particle density increases as a func-
ond bunch rotation. The beam is then pre-cooled tion of energy Ψ(E) ∝ eE/Ed . For the FNAL
with stochastic cooling. The parameters of the Accumulator, η = 0.02, Ed ∼ 10 MeV, E = 8.9
collected beam are given in Tab.2. GeV. Some parameters of the stacking process are
A difficult step is to accumulate p̄’s from 104 given in Tab.3.
or more pulses. The stacking process is accom- At FNAL, the p̄ beam is transferred from the
plished with stochastic cooling in an accumulator Accumulator to the Recycler, a large storage ring
ring; see Fig.2. The flux of p̄’s that can be accu- that stacks the beam azimuthally using barrier
mulated is [6] bucket rf manipulations (Sec.4.10). Both stochas-
tic and electron cooling are used to increase the p̄
Ed T0 W 2 |η| beam density. The electron cooling system is no-
Φ0 = (1)
β 2 EΛ table for its high energy electron beam (4.9 MeV).
Final p̄ beam parameters are in Tab.4.
9
Table 3: Beam properties in the accumulation ring. with the FNAL Accumulator. FAIR will also pro-
Parameter FNAL vide low energy p̄ beams (down to 30 MeV) for
(Accum.) experiments with low energy or stopped p̄’s.
Stack Rate (1010 hr−1 ) 28 Antiproton beams have been used to produce
p̄/pulse (107 ) 20 relativistic antihydrogen atoms at CERN [8] and
Yield (p̄/proton) (10−6 ) 24 Fermilab [9]. The relativistic antiatoms lasted a
Final Δp/p (%) 0.2 few tens of ns. Experiments at the AD have used
trapped p̄’s to produce antihydrogen [10, 11];
Final emit. 6π

(mm-mrad) 1.5π with the antihydrogen existing long enough for
Bandwidths (GHz) physics measurements [12]. p̄’s will be used at
Stacktail 2-4 the AD and FAIR in the future to study antihydro-
Core cooling 2-4,4-8 gen.

Total particles (1012 )
typical, max 0.25, 3 References
[1] M.D. Church, J. Marriner, Ann. Rev. Nucl. Part.
Table 4: Final antiproton beam properties.
Sci. 43 (1993) 253
Parameter FNAL CERN [2] B. Autin et al, CERN/PS/AA 78-3 (1978); E.J.
(Recyc.) (AD) Wilson et al, CERN 83-10 (1983)
Cycle time 20 hr 100 s [3] Fermilab Design Report, Tevatron I Project
Total particles 5.5x1012 1x107 (1984)
Final Δp/p (%) 0.04 0.01 [4] S. Maury, ed., CERN/PS 96-43 (1996)
Final emit. 6π [5] C. Hojvat, A. van Ginneken, NIM 206 (1983) 67
(mm-mrad) 0.2π 1π [6] S. van der Meer, CERN/PS/AA 78-22 (1978)
[7] H.H. Gutbrod, ed., FAIR Baseline Technical Re-
port (2006), http://www.gsi.de/fair/
At CERN the p̄ source consisted originally
reports/btr.html
of an antiproton accumulator (AA) to which was [8] G. Bauer et al, Phys. Lett. B 368 (1996) 251
added the antiproton collector (AC) to increase [9] G. Blanford et al, PRL 80 (1998) 3037
production by increasing the p̄ beam acceptance [10] M. Amoretti et al, Nature 419 (2002) 456
and by using the bunch rotation technique pio- [11] G. Gabrielese et al, PRL 89 (2002) 233401
neered at FNAL. The low energy antiproton ring [12] G.B. Andresen et al, Nature Physics 7 (2011)
(LEAR) was the first accelerator dedicated to par-
ticle experiments with p̄’s. LEAR has been de-
commissioned, but low energy p̄’s continue to be 1.6.2 Betatron
produced by the CERN Antiproton Decelerator M. Tigner, Cornell U.
(AD), a reconfiguration of the original CERN p̄ The Betatron [1] is a cyclic electron accelerator
source that emphasizes economical production of with a circular orbit of approximately constant ra-
low energy p̄’s [4]. Antiprotons are not accu- dius which provides acceleration through Mag-
mulated but each pulse of p̄’s is decelerated and netic Induction. Iron core machines of energies
cooled several times before extracting the entire up to 300 MeV have been constructed. The guide
beam for low energy/stopping beam experiments. field in the classical Betatron is weak focusing.
Stochastic cooling is used initially while electron A section perpendicular to the equilibrium orbit
cooling is used at the lower energies. Tab.4 gives plane is shown in Fig.1 [2].
typical beam parameters for the current CERN The beam travels in a doughnut shaped, evac-
and FNAL sources, which have now significantly uated, dielectric (e.g. glass) chamber with a thin
diverged in their primary uses. conducting film on the inside to prevent charging.
The FAIR project at GSI [7] will include The accelerating electric field is produced by the
a p̄ source and storage rings to support exper- changing magnetic flux within the equilibrium or-
iments with p̄ beams. FAIR will provide high bit. By combining Newton’s law and Faraday’s
energy p̄ beams (up to 14 GeV) interacting with law of induction together with the Lorentz force
an internal gas jet target at high luminosity (2 × law we can write
1032 cm−2 s−1 ), extending the technique that was dp e ∂Φ
first used briefly in the CERN ISR and then later = , p = Beρ (1)
dt 2πρ ∂t
10
Ch.1: INTRODUCTION
yyyyyyyy
Expansion
Side
Guide Package
Field Electron
Central Flux Orbit
yyyyyyyy
yyyyy
;;;;;
Bo Flux Core
Bc Vacuum
Chamber
Flux
Laminated Pole Piece

8–98
Magnet
8355A64
;yyy
;;;
yyy ;
y
yyyyyyyy
6 Field Magnets
Figure 1: Betatron schematic.
;;;
yyy
;; ;;
yy
and find that
yyyyy
;;;;;
2
;;
yy
;;;yy
;;
ΔΦ = 2πρ B (2)
Flux Core
the famous 2 to 1 condition that the flux change
yyy
within the orbit must be twice that which one
r0 = 10 in.
would obtain if the field were uniform throughout
the region inside the orbit and equal to the field at
the orbit. 12–97
8355A63
Since the field in an iron core magnet is lim-
ited by saturation, increase in maximum energy Figure 2: Separated function betatron.
means increase in radius of the core and thus the
volume of the iron. The iron of the 300 MeV Illi- a few microseconds at injection time. Fig.3 [2]
nois Betatron weighed > 300 tons. shows some typical power cycles of various beta-
There is another constraint on the maximum trons. Injection is complicated and involves space
energy which can be obtained using the betatron charge effects (Sec.2.4.5.1) in a central way [3].
principle. As the beam energy rises synchrotron Injection efficiency can be greatly enhanced by
radiation loss rises and competes with the energy adding pulsed coils to draw the equilibrium orbit
gain due to magnetic induction. This effect spoils temporarily away from the gun just after injection
the linearity of Eq.(1) and requires special means and slowly restoring it as the initial betatron os-
for adding the extra energy needed. In practice, cillations damp. Ejection or targeting is done in
the synchrotron radiation begins to become im- the same way by pulsewise distortion of the orbit
portant at ∼100 MeV and limits beam energies enough to drive the beam into a weak field region
that can be obtained with iron and copper magnet and out of the doughnut through a thin window or
technology to ∼300 MeV. into a tungsten bremsstrahlung target within the
While betatrons using the combined func- doughnut with subsequent extraction of the x-rays
tion, one magnetic circuit, design of Fig.1 have through a thin window.
been successfully operated at low energies, con- While some betatrons, particularly the 300
siderable efficiency in size and operation can be MeV machine were used for nuclear physics re-
achieved by separating the functions as shown in search, the majority were used for medical ther-
Fig.2 [2]. The scale refers to an 80 MeV machine. apy or diagnostic x-raying of industrial equip-
The magnetic elements are punched 0.355 mm ment. The 300 MeV machine was capable of
laminations of silicon transformer steel, wound producing 14,600 R/min. at 1 m in Pb while
with litz wire to minimize eddy current. Even so the machines used for therapy produce typically
the highest energy betatrons achieve only about about 100 R/min. at 1 m. These machines
5% duty factor due to the large hysteresis and have been largely supplanted by linacs which are
eddy current losses coupled with the difficulty of more powerful, more flexible, lighter and more
cooling the massive iron cores. Injection is ac- easily controlled. In an attempt to overcome
complished by insertion of an electron gun into the space charge engendered limitation to beam
the sealed, doughnut vacuum chamber, just out- current in the classical betatron, the concept of
side or just inside the equilibrium orbit, and puls- the “Modified Betatron” was developed. In the
ing it negatively at tens to more than 100 kV for Modified Betatron, a torroidal field and strong
11
Flux higher available energies. The center of mass en-
B (a)
Field ergy Ecm for the collision of two particles of mass
Field m1 and m2 with energies E1 and E2 with a cross-
O Bias t
ing angle θ is

–B
Inject Eject
Ecm = 2E1 E2 + (m21 + m22 )c4 (1)
1 1
1
2
+ 2(E12 − m21 c4 ) 2 (E22 − m22 c4 ) 2 cos θ
(b) Flux
B
Field m2, E2
Inject
O t
Flux m1, E1
–B Eject θ
Bias
Saturation 12–97
8355A18
(c) Pulse Interval
B
Field Typically m1 = m2 and E1 = E2 (= E) and
Field θ = 0 with Ecm = 2 E and the CM is station-
Bias
O t ary in the laboratory. With the advent of B fac-
–B
Inject Eject
Flux Flux tories, beams of e− ’s and e+ ’s are collided with
Bias
different energies so that the CM is moving with
12–97
8355A65
high velocity. Short lived particles produced in
the collisions then move a measureable distance
Figure 3: Typical power cycles. before decay enabling important experiments in
CP violation. In this case, Ecm ≈ 2 (E1 E2 )1/2
focusing both are added to the normal, weak fo- for θ = 0.
cusing, vertical (mirror) field of the classical be- Colliding beam storage rings were first de-
tatron. A beam of ∼1 kA has been accelerated scribed in 1956 [1]. Princeton-Stanford 500 MeV
this way to 20 MeV [4]. Other new ideas have two ring e− e− , the two ring 140 MeV e− e− ring
come forward recently, inspired by new materials VEP-1 at Novosibirsk, and the 250 MeV e+ e−
and new needs. One approach proposes very high AdA at Frascati were the first operating e± stor-
frequency operation to achieve high beam power age rings. The ISR at CERN was the first hadron
[5]. Another approach uses the FFAG principle collider that collided protons using two intersect-
(Sec.1.6.6) in a betatron configuration [6]. ing storage rings [2]. The first linear collider was
the SLAC Linear Collider (SLC) and was based
References on an idea that was first proposed in 1965 [3]. The
[1] M.S. Livingston, J.P. Blewett, Particle Accelera- SLC used a single linac with two beamline arcs to
tors, McGraw-Hill (1962) collide the particles.
[2] By permission, McGraw-Hill Book Co. Most colliders are based on storage rings
[3] L. Gonella, Supplement to Nuovo Cimento 3 where the beams are allowed to interact at one
(1966) 303 or more collision points and can be either dou-
[4] C.A. Kapetanakos et al, Phys. Fluids B5 (1993) ble rings or a single ring employing electrostatic
2295 separation to keep the oppositely charged beams
[5] G.V. Dolbilov et al, APAC 2007 (2007) 628 apart except at the collision points. The beam par-
[6] S. Boucher et al, EPAC08 (2008) 1860 ticles can interact many times in this configura-
tion. Another configuration is the use of one or
two linacs with the beam particles colliding only
1.6.3 Colliders once. The proposed ILC [4] and CLIC [5] col-
D. Hartill, Cornell U. liders use two linacs each aimed at the collision
Particle beam colliders were developed to in- point.
crease the center of mass energy available for new The interaction rate of a collider is measured
particle production and interaction. For a station- by its luminosity (Sec.4.1) with typical units of
ary target this energy scales only as the square 1032 cm−2 s−1 . Various instabilities can limit the
root of the beam energy while with a collider the performance of a collider but a variety of feed-
collision energy scales linearly allowing for much back systems and careful selection of operating
12
Ch.1: INTRODUCTION
Table 1: Colliding Beam Machines. [a] DR: Double storage ring. SR: Single storage ring. LC: Linear collider.
[b] Princeton-Stanford Colliding Beam Experiment. [c] 200 GeV per charge unit. [d] Planned.
Location Name (type[a] ) Beam Energies E (GeV) Start
Stanford/SLAC, USA CBX[b] (e− e− DR) .5 + .5 1963
Spear (e+ e− SR) 2.5 + 2.5 1972
PEP (e+ e− SR) 15 + 15 1980
SLC (e+ e− LC) 50 + 50 1989
PEP-II (e+ e− DR)
3.1 + 9.0 1999

Frascati, Italy AdA (e+ e− SR) 0.25 + 0.25 1962
Adone (e+ e− SR) 1.5 + 1.5 1969
DAΦNE (e+ e− SR) 0.5 + 0.5 1997
Super B (e+ e− DR) 3.1 + 9.0 2015[d]

Novosibirsk, Siberia VEP-1 (e− e− DR) 0.13 + 0.13 1963
VEPP-2/2M (e+ e− SR) 0.7 + 0.7 1974
VEPP-4 (e+ e− SR) 7+7 1979
Cambridge, USA CEA Bypass (e+ e− SR) 3.0 + 3.0 1971
Orsay, France ACO (e+ e− SR) 0.5 + 0.5 1966
DCI (e+ e− DR) 1.8 + 1.8 1976
DESY, Germany Doris (e+ e− SR) 3+3 1974
Petra (e+ e− SR) 19 + 19 1978
HERA (e± p DR) 30 (e) + 820 (p) 1992
CERN, Europe ISR (pp DR) 31.5 + 31.5 1971
Spp̄S (pp̄ SR) 315 + 315 1981
LEP (e+ e− SR) 104.5 + 104.5 1989
LHC (pp, PbPb DR) 7,000 + 7,000 2009
Brookhaven, USA RHIC (heavy ion, pp DR) 100/u[c] + 100/u[c] 1999
Cornell, USA CESR (e+ e− SR) 6+6 1979
KEK, Japan Tristan (e+ e− SR) 30 + 30 1986
KEK B (e+ e− DR) 3.5 + 8 1999
Super KEK B (e+ e− DR) 3.5 + 8 2015[d]
Beijing, China BEPC (e+ e− SR) 1.55 + 1.55 1989
BEPC II (e+ e− DR) 2.1 + 2.1 2008
Fermilab, USA Tevatron (pp-bar SR) 980 + 980 1987
points have enabled performance levels in excess Damping Final Final Damping
Ring Linac Focus Focus Linac Ring
of 200 times this unit at the asymmetric B factory x
at KEK with similar performance at the asymmet- Collision
e+ point e–
ric B factory at SLAC. An upgraded Super KEK 7–95 Source Source 7993A6
B [6] at KEK and a new Super B collider [7] to

be located near Frascati using parts from the de- Figure 1: Components of a Linear Collider.
commissioned SLAC B-factory both project per-
formance levels at 8,000 to 10,000 times this unit. and projects an integrated luminosity of more than
Both projects have been approved by their fund- 6 fb−1 over the next two years. The Tevatron col-
ing agencies and will take advantage of very low lider with a beam energy of 1 TeV has accumu-
emittance beams made possible by the damping lated over 8 fb−1 per interaction region by the end
ring technology developed for the SLC. The ex- of its operation in FY2011. The Tevatron collider
pected start of commissioning is in the 2014-2016 has been in operation over 23 years and its cur-
time frame. rent peak luminosity of 4 × 1032 cm−2 s−1 is 400
The LHC has begun operation with an initial times its design. The LHC collides protons on
beam energy of 3.5 TeV and has already achieved protons using two separate rings and the Tevatron
a peak luminosity in excess of 2 × 1032 cm−2 s−1 collides antiprotons on protons in a single ring
13
pp and ep Colliders Lepton Colliders
105 103 Cyclotrons are often referred to by the diam-
eter of the magnet pole (e.g. 27 inch, 184 , 2 me-
LEP
LHC
ter, U-400). More recently “K”, (≈ proton kinetic
104 102 SLC
energy in MeV) has become a designation (e.g.
Petra Tristan
PEP K500, K1200) particularly for multi-particle cy-
Ecm (GeV)
103
Tevatron
10 1 SPEAR
VEPP–4
CESR clotrons where the energy for an ion of charge Qe
PEPII
SppS CEA
Doris
DCI
KEKB and mass Am0 (where m0 is 1/12 of mass of 12 C,
RHIC(pp)
Adone BEPC m0 c2 = 931.48 MeV) is given nonrelativistically
Hera CBX VEPP–2M

102
ISR
10 0
ACO VEPP–2 Daφne by E = KQ2 /A. The maximum bending power
AdA
(Bρ) is related to K by K = (eBρ)2 /(2m0 ).
VEP–1
101
Many cyclotrons are referred to by a local name
10–1
1–2006
'70 '80 '90 2000 '60 '70 '80 '90 2000 or acronym (e.g. ORIC, AGOR). Typical beam
8355A20 Year Year
parameters achieved by cyclotrons: normalized
Figure 2: Colliders over the years. emittances N r ≈ 2 to 0.2 mm-mrad, N z ≈ 5 to
1 mm-mrad, and energy spread ΔE/E typically
using electrostatic separators to provide separa- 10−3 with best value 2 × 10−4 .
tion of the beams except at the two IPs. Cyclotrons have evolved in many, sometimes
Future colliders are likely to be either lin- overlapping, subclassifications. Their character-
ear colliders using e− and e+ to avoid high istics are best documented in the proceedings of
synchrotron radiation losses or storage rings us- a series of triennial conferences [2], and are sum-
ing muons to get to the highest collision ener- marized in [3].
gies. Electrons and muons are constituent parti- Classical cyclotron (now rare) Fig.1 shows the
cles while only the quarks in hadrons are the con- original cyclotron concept [1]. A static and uni-
stituent particles with only a fraction of the total form magnetic field B is applied perpendicular
hadron energy. It is the collision energy of the to D-shaped hollow electrodes (“dees”). The
constituent particles that can access new particles. dees are driven by an rf voltage whose frequency
matches the constant cyclotron frequency
References qB
f= (1)
[1] D. Kerst et al, PR 102 (1956) 590; G.K. O’Neill, 2πm
PR 102 (1956) 1418 of nonrelativistic ions. Ions from a central ion
[2] K. Johnsen, Proc. 8th Int. Conf. on High Energy source are repetitively accelerated in and out of
Acc. (1971) p.79; CERN AR/Int. SG/64-9 Design the dees on a spiral path to maximum energy. In
Study of ISR (1964) practice, the magnetic field must decrease with r
[3] M. Tigner, Nuovo Cimento 37 (1965) 1228 to assure stability in the axial direction, and the
[4] ILC Reference Design Report: http://media. accelerating voltage must be high for ions to reach
linearcollider.org/rdr draft v1.pdf (2007) the design energy before they get out of the ac-
[5] CLIC Design Report: CERN-OPEN-2008-021 celerating phase of the rf cycle (due to relativis-
(CLIC-Note-764) (2008) tic mass increase and to magnetic field decreasing
[6] T. Kageyama, Super KEKB, PANIC 05 Proceed- with r). These effects limit highest ion velocity.
ings, p.589 (2005)
The record for classical cyclotrons is β = 0.22
[7] Super B Conceptual Design Report (2007):
(Oak Ridge 86 , Stockholm 225-cm).
http://www.pi.infn.it/SuperB/CDR
Isochronous cyclotron Thomas [4] pointed out
1.6.4 Cyclotron that magnets with alternate strong and weak az-
H. Blosser, Michigan St. U. imuthal regions (“sectors” or “hills and valleys”)
M. Craddock, U. British Columbia & provided an additional axial focusing which could
TRIUMF offset the defocusing from a radially increasing
magnetic field. The average magnetic field can
Earliest [1] and most numerous of circular accel- therefore match the mass increase of the accel-
erators, cyclotrons are characterized by magnetic erated particle with positive axial focusing pro-
field and accelerating rf frequency which are con- vided by the azimuthal variation. Two electron
stant in time (c.w.). (Synchrocyclotrons are ex- cyclotrons with sinusoidal azimuthal variation of
ceptions. See later.) the magnetic field were built (Berkeley, 1950,
14
Ch.1: INTRODUCTION
over a broad range, and difficult extraction com-

ponents (electrostatic deflectors, magnetic chan-
nels) are avoided. Multiple beams can be ex-
tracted simultaneously at different energies us-
ing suitably shaped foils. The H− technique is
used in most radionuclide-production cyclotrons,
where high beam currents make extraction es-
pecially difficult. To avoid electric dissociation
(Lorentz stripping, Sec.7.1.8) of H− at high ve-

locities, the 520-MeV meson-factory cyclotron at
Figure 1: A classical cyclotron. If magnetic field TRIUMF requires weak magnetic fields so that
points into page, spiral line is trajectory of positive ion. this cyclotron is the world’s largest (17 m dia.
poles).
β = 0.5). In the 1950s, pole tips with a con-
stant gap in the hill region and a larger con- Superconducting compact cyclotron This
stant gap in the valleys came into use, pro- class includes cyclotrons with superconducting
viding a more trapezoidal azimuthal field vari- main magnet coils or rf systems (e.g. TRITRON
ation. Also, at that time, “strong-focusing” at Munich). These cyclotrons can be up to ×10
by spiralling the hills was introduced and des- lighter than room-temperature cyclotrons due
ignations “sector-focused cyclotron”, “spiral- to the unusual cyclotron scaling law that total
ridged cyclotron”, “azimuthally-varying-field cy- flux ∝ 1/B. Weight and size reduction leads to
clotron” are now used largely inter-changeably ∼×0.5 construction cost and ∼×0.9 operating
with isochronous cyclotron. (Thomas cyclotron costs. Examples: (1) 22-ton cancer therapy
is normally reserved for azimuthally sinusoidal cyclotron in Detroit produces 50 MeV deuterons,
fields.) More than 800 isochronous cyclotrons (2) 240-ton K1200 nuclear physics cyclotron
have been built (50-590 MeV for nuclear and par- at East Lansing produces beams >10 GeV for
ticle physics, ∼230 MeV for proton therapy, and heavier ions, (3) 90-ton Varian/Accel cancer
10-100 MeV for commercial production of ra- therapy cyclotrons at PSI and Munich produce
dionuclides). Over a dozen 210-ton IBA 230- 250-MeV protons.
MeV cyclotrons have been installed in hospitals Synchrocyclotron This largely outmoded form
worldwide in recent years (Sec.1.6.14). of the classical cyclotron uses an rf frequency
Separated sector cyclotron This is a subclass which varies with time to track the orbital fre-
of isochronous cyclotrons in which the valley re- quency. The energy limit of the classical cy-
gions are iron free. The concept was proposed clotron is relieved, but the c.w. characteristic
in the late 1950s by Oak Ridge for a 900 MeV is lost and beam current typically decreases by
isochronous cyclotron, and first used at PSI in ×1000. Auxiliary slow extraction systems were
the 590-MeV meson factory (operation 1974). A often used to stretch the duty cycle. Synchro-
simpler, “radial-sector” (i.e. non-spiral) formula- cyclotrons provided the energy-frontier beams
tion of this concept was adopted for the Indiana of the 1950s but have now been superseded
U. 200 MeV proton cyclotron (operation 1975). by synchrotrons for high-energy physics and by
The radial, separated sector design is also used isochronous cyclotrons for nuclear physics. The
in large heavy-ion cyclotrons (GANIL, RIKEN, highest beam velocity achieved by a cyclotron is
Lanzhou). The RIKEN K2600 SRC, using super- in the 1 GeV synchrocyclotron at Gatchina. A
conducting magnet coils, delivers 80-GeV heavy very compact synchrocyclotron for cancer ther-
ions, making it the highest energy c.w. accelerator apy with 250-MeV protons has been built by Still
of any type. River Systems. The 9-T superconducting magnet
H− cyclotron Cyclotrons to accelerate H− ions is light enough to be mounted on a rotating gantry.
were introduced in 1962 at U. Colorado [5]. At Orbit characteristics Magnetic Field
maximum energy a thin foil strips the electrons
B(r, θ) = B0 (r) + Bi (r) cos[iθ + ζi (r)] (2)
from the H− ion, reversing their bending radius
i
and thus sending the beam quickly out of the cy-
clotron. Moving the foil to a different radius eas- Near r = 0, Bi (r) ≈ (r/G)i where G is magnet
ily changes the output energy of such a cyclotron gap. B0 = B0 (r = 0) = 2πf0 m0 /q [Eq.(1)].
15
Choice of sector number N = 3, 4, 6, 8, ... Due to large energy gain per turn and rela-
depends on maximum E/A (energy/nucleon). tively rapid change of ν values with energy, cy-
N = 2 is radially unstable. Odd numbers other clotrons frequently pass through essential beta-
than 3 are too complicated. N = 3 is the most tron resonances at ν = N/3, N/4, etc. and im-
frequent choice below 200 MeV/A because of the perfection resonances at 1/1, 2/2. Imperfection
fast rise of B3 near r = 0. resonances are often intentionally used to steer the
Flutter is defined as beam in the central region and near extraction.
B 2 − B2 1 Space Charge Limits
F (r) = = fi (r)2 (3) Axial limit Assume fully overlapping turns

B 2 2
i (“current sheet” approximation),
where fi (r) = Bi (r)/B0 (r). Sometimes f 2
Ilim = 0 (2πf0 )Aνz,min Df V (7)
is called flutter, F (sometimes written F 2 ) called

mean square flutter. With flutter, orbit scallops, where 0 is the permittivity of free space, A is
⎡ ⎤ the full beam height, νz,min is the axial tune at
the radius of weakest focusing, Df is the frac-
r(θ) = r ⎣1 + gi cos(iθ + ζi )⎦ (4) tion of the machine circumference occupied by
i=1 the beam, and V the peak accelerating voltage per
where gi ≈ fi /(i2 − 1). turn. Ilim ≈ 10 mA is typical.
Field spiral ζN (r) is the angular location of Longitudinal limit In an isochronous device
maximum value of the main flutter component. the orbit period is independent of the energy so
Spiral angle α is the angle between radius vector the longitudinal length of bunch is unchanged but
and tangent line to spiral curve, tan α = rdζ/dr. space charge force increases the energy spread of
The acceleration system consists of dees (2 particles in a given turn and is the principal limit
gap drift tube) or cavities (1 gap drift tube), now for single-turn-extraction (see below). The linear
usually located in valleys of magnet (in all val- component of space charge force can be compen-
leys or, if N is even, in alternate valleys). Dees sated by moving beam slightly to side of voltage
usually operate at 50-200 kV (peak voltage dee to wave. Numerical calculations are required to es-
ground). Cavities operate up to 850 kV per gap. timate residual nonlinear component. A scaling
The acceleration system often runs on harmonic law shows that turn separation ∝ V 3 , so V is crit-
of beam orbital frequency frf = hf0 . Multiple ical. “Flat-topping” voltage by adding higher har-
dees are sometimes phased by selecting a natu- monic gaps is sometimes used.
ral mode of the resonator structure (0-mode or π- Beam extraction H− cyclotrons use stripping
mode) and sometimes by servoed phase shifters foils. Positive ion cyclotrons use an electrostatic
(as in 3 dee h = 1 systems at MSU, Texas A&M, deflector with thin (∼0.1-0.3 mm) septum fol-
Catania, & AGOR). lowed by magnetic channel(s).
Without flutter, Precessional extraction introduces coherent
m0 radial oscillation (amplitude a) prior to extraction
Bisoc = 2πf0 γ(r) to add precessional component (2πa|νr − 1|) to
q
−1/2 the acceleration radius-gain-per-turn [dr/dn =
m 2πf0 r 2 r(γ/(1 + γ))(qV /E)(νr )−2 where qV is the
= 2πf0 1− (5) energy-gain-per-turn]. Precession usually induces
q c
a νr = 1 transition at the edge region using a field
r dB bump (i.e. magnetic azimuthal first harmonic) of
kisoc = = γ 2 − 1 ⇒ νr = γ (6)
B dr ∼ 10−5 relative to main field. A radius gain of
As scalloping increases, the radial tune νr rises 3-5 mm is typical (limited by axial instability at
faster than γ, hitting resonances. The N/2 stop- the nearby νr = 2νz coupling resonance).
band limits maximum γ (e.g., γmax ≈ 1.2 ⇒ Single-turn extraction uses high accelerating
200 MeV/A for N = 3). voltage V and narrow rf phase interval at voltage
Axial focusing tune νz ≈ [−k + F (1 + peak to give turn spacing larger than turn width
2 tan2 α)]1/2 . Normally one picks α(r) to in- so that the deflector septum can be placed at a low
crease with r to give νz ≈ 0.2 to avoid reso- density point between last two turns (extraction
nances. If sectors are radial (α = 0), νz decreases efficiency of 99.97% at 590 MeV at PSI). This re-
with r and resonances limit usable energy band. quires highly stabilized magnet and rf (amplitude
16
Ch.1: INTRODUCTION
1 3 4 5 6 7 8 9 10 11 12
and frequency). Flat-topping the rf waveform by
adding a third harmonic is used at PSI to broaden
the usable phase interval. Extraction current is
limited by longitudinal space charge spreading of
the turn structure. Very high V is required at high
16
currents (∼3.4 MV/turn at PSI for a 2.2-mA beam 2 1 13 11 14
7
15
- until recently at 1.3 MW currently the highest
power beam from any accelerator, cw or pulsed).
Multi-turn extraction is the default situation

1. Pelletron Tank 9. Chain Idler Wheel
when single-turn requirements are not met. Par- 2. Beam Input Drift Tube 10. Pellet Charging Chain
ticles on either side of the minimum-turn-number 3. Insulating Plastic Support 11. Charging Inductor
4. Accelerating Tube-Cutaway view 12. Chain Drive Wheel

phase make an additional turn and so on to the 5. Potential Distribution Ring 13. Gas Stripper Assembly
edge of occupied phase interval. Particles at turn 6. Generating Voltmeter 14. High Voltage Terminal
number steps line up on the septum and a frac- 7. Discharging Inductor 15. Dessicant Tray
8. Capacitor Pickup 16. Beam Output Drift Tube
tion of (septum thickness)/(radius-gain-per-turn) 11–97 8355A46
is lost. Overall multi-turn extraction efficiency
ranges from 50% (typical) to 93% (PSI) to 97% Figure 1: 1 MV Tandem Pelletron accelerator.
(numerical optimization).
Table 1: Tandem PelletronR
basic specifications. V is
References
terminal voltage, D and L are pressure tank diameter
[1] Lawrence, Edlefsen, Science LXXII (1930) 376 and length, Iˆ is maximum charging current. (*Folded
[2] Proc. Int. Conf. on Cyclotrons and Their Aplica- tandem with 180◦ magnetic deflection in terminal.)
tions, http://jacow.org/
V (MV) Orient. D(m) L(m) ˆ
I(μA)
[3] M.K. Craddock, K. Symon, RAST 1, 65 (2008)
[4] L.H. Thomas, PR 54 (1938) 580 1 Horiz. .61 2.18 50
[5] M.E. Rickey, W.R. Smythe, NIM 18-19, 66 (1962) 3 Horiz. 1.22 5.11 300
5 Horiz. 2.13 8.43 1000
10 Vert. 4.27 18.3 300
1.6.5 Electrostatic Accelerator 15 Vert. 5.49 26.3 300
J. Ferry, NEC 20 Vert.* 8.23 23.77 900
An electrostatic accelerator is a single potential 25 Vert.* 10.06 29.87 900
drop system in which the high voltage potential
is generated by the mechanical transfer of charge from the high voltage terminal to ground gaining
from ground to the high voltage terminal. Ex- additional energy. Tab.1 shows tank size versus
amples: fabric belt charged Van de Graaff ac- terminal voltage for tandem pelletrons.
celerators built by High Voltage Engineering Eu-
The FWHM of the terminal voltage variation
ropa B.V., the Netherlands, by Nissin-High Volt-
of a Pelletron is ∼500 V without special ripple
age Co., Ltd., Japan, and by Vivirad High Voltage, reduction circuitry. Thus, for a proton beam the
USA, and pellet chain charged Pelletron R built by
FWHM energy spread is 500 eV. For a C+4 beam
National Electrostatics Corp (Fig.1).
it is 2000 eV.
Most electrostatic accelerators built today are
Final emittance depends on ion mass because
dual acceleration, tandem types. Negative ions are
of multiple scattering in the terminal stripper. For
injected and accelerated by a positive dc voltage
light ions (p’s and α’s), final emittance is ∼ 3π
on the high voltage terminal. At MeV energies,
mm-mrad (MeV)1/2 . For heavy ions like gold, the
the ions are directed through a turbo-molecular
final emittance will be limited by the acceptance
pumped [1], i.e. compressed, tube of argon or
of the high energy acceleration tube to about 20 π
nitrogen gas (microns Hg), which strips two or
mm-mrad (MeV)1/2 .
more electrons away changing them to positive
ions. Carbon foils of a few μgm/cm2 can also Acceleration tube All electrostatic accelerators
be used, however important heavy ions such as must have a highly evacuated tube for the ions to
C+3 at 2.6 MeV cause the foils to rapidly thicken traverse during acceleration. This tube is made up
and break. The positive ions are accelerated away of dense, low outgassing, insulating rings sealed
17
to metal electrodes. The voltage between the elec- at a wheel. The inductor electrodes are biased
trodes is established by resistors or corona points up to −60 kV for chain runs to and +60 kV for
draining current from the high voltage terminal. chain runs from a positive polarity high voltage
Glass insulating rings sealed with polyvinylac- terminal. As the grounded chain, pressing firmly
etate, together with aluminum, titanium or stain- against a wheel, moves into the inductor elec-
less steel electrodes are successfully used. NEC trode charge flows smoothly onto it. Then while
tube uses high density alumina ceramic insulat- the chain is still within the inducing field, contact
ing rings sealed with aluminum metal to titanium with ground is broken as the wheel turns. The in-
metal electrodes. For most acceleration tubes the duced charge is trapped on the chain and carried to
electric fields along the tubes are inclined or spi- the terminal or ground where the chain enters the
raled so that low energy secondary ions and elec- discharging inductor electrode. Contact with the
trons produced on the electrodes are deflected into wheel occurs as the charge on the pellet is bound
nearby electrodes and cannot gain high energies by the inductor to prevent sparks or corona dis-
[2, 3]. Some tubes also use low magnet fields to charges from the chain.
sweep low energy secondaries and electrons out
of the beam path [4]. To remove ions and elec- References
trons produced on the electrodes, one can also [1] F. Terrasi et al, NIM B259 (2007) 15
vary the electric field along the beam path to pro- [2] W.D. Allen, Nat. Inst. Res. Nucl. Sci. NIRL/R/21
duce cylindrically symmetric focusing fields that (1962)
deflect low energy particles into adjacent elec- [3] R.J. Van de Graaff, P.H. Rose, A.B. Wittkower,
trodes [5, 6]. It is generally found that a com- Nature 195 (1962) 1292
fortable reliable gradient of about 1.6 MV/m can [4] F.A. Howe, IEEE Trans. Nucl. Sci. N.S.-14, No.3
be achieved with a modern acceleration tube. See (1967) 122
also Secs.6.14, 7.2.9. [5] M.L. Sundquist, R.D. Rathmell, J.E. Raatz, NIM
A287 (1990) 87
Support column Below 6 MV terminal volt-
[6] J.E. Raatz et al, NIM A244 (1986) 104
age, one can use acrylic plastic plates as the in- [7] J.A. Ferry, NIM A328 (1993) 28
sulating mechanical support column for compo- [8] G.A. Norton et al, NIM B37/38 (1989) 403
nents inside the pressure tank. Aluminum hoops
are positioned along the insulating plates and are
connected to the acceleration tube resistor divider 1.6.6 FFAG Accelerator
[7]. This construction is suitable for columns up M.K. Craddock, U. British Columbia &
to about 29 ft. length. Above 6 MV, columns TRIUMF
with alumina ceramic insulators can be used. In Following the discovery of alternating gradient
this case ceramic cylinders are bonded with alu- (AG) focusing in 1952, Fixed-Field Alternating-
minum to titanium electrodes to form posts about Gradient accelerators (FFAGs) were proposed in-
18 in. long. The posts can support 1 MV when dependently by Ohkawa in Japan, Kolomensky in
immersed in pressurized SF6 gas. the USSR and Symon and Kerst in the US [1].
Charging system Many fabric belts are still in With fixed magnetic fields, modulated rf, and
use, although no new machines do so. Since the pulsed beams, FFAGs operate just like synchro-
fabric belt is an insulator, charge transfers to and cyclotrons, but are capable of reaching tens of
from the belt must be accomplished via corona GeV. The innovations were to break the magnet
discharge or physical rubbing. Such methods, al- into radial or spiral sectors to provide strong fo-
though effective, are difficult to control precisely, cusing, and (usually) to remove the central region
leading to large terminal voltage ripple and can — the same steps that convert a classic Lawrence
limit useful belt life. cyclotron into a separated-sector ring cyclotron.
Another method uses a chain made up from The FFAG is the most general type of fixed-field
steel cylindrical pellets linked by stainless steel accelerator (i.e. cyclotron [2], Sec.1.6.4).
pins to nylon insulators. A chain carries typically Fixed magnetic fields lead to spiral orbits, so
> 150 μA of current [8]. The chains are charged an FFAG’s vacuum chamber, magnets and rf cav-
and discharged by an induction scheme. Each ities tend to be larger and more costly than a syn-
chain is surrounded by a long, closely spaced in- chrotron’s. On the other hand, its beam intensity
ductor electrode where the chain leaves or arrives can be much higher, as the radial and momentum
18
Ch.1: INTRODUCTION
of F (1 + 2 tan2 α). MURA’s recipe was to keep

the flutter F (r) = constant by using constant pro-
file B(θ)/B̄, and, depending on the sector type:
- if spiral: α constant, sector axis r = r0 eθ cot α
- if radial: boost F by specifying BD = −BF .
Of course, reverse fields raise the average ra-
acceptances are larger, and the repetition rate, set dius. The “circumference factor” R/ρ ≥ 4.5 if
there are no straights [1], but smaller with them
purely by rf considerations, can be several kHz.

The most intensive studies were carried out (1.8 for the KEK 150 MeV ring).
by Symon, Kerst and others at MURA (the Scaling FFAGs operating or under construc-
Mid-western Universities Research Association) tion Recent years have seen the construction
in Wisconsin in the 1950s and 60s, and culmi- and successful operation of the first-ever FFAGs
nated in the construction and successful testing of for protons by Mori’s group at KEK [7, 8] and
electron models of radial-sector and spiral-sector several more following scaling principles (Tab.2).
designs [3]. But proposals for proton FFAGs Except for ERIT and NHV (FDF), the radial-
were not funded at that time, nor were those for sector designs employ DFD triplet magnets.
1.5 GeV spallation neutron sources in the 1980s. Mori introduced important innovations in
Recently, with improvements in magnet and both magnet and rf design. The DFD triplets are
rf technology, FFAGs have become the focus of built and powered as single units, without a steel
renewed attention. Ten machines have been built return yoke, forcing the return flux through the
and a muon cooling ring is under construction. In D and automatically providing reverse field. The
addition, ∼15 designs are under study for the ac- open structure also facilitates injection and extrac-
celeration of protons, heavy ions, electrons and tion. The rf innovation (avoiding the cumbersome
muons, with applications as diverse as treating rotary capacitors on synchrocyclotrons) is the use
cancer, irradiating materials, driving subcritical of FINEMET metallic alloy tuners, which offer
reactors, boosting high-energy proton intensity, (a) rf modulation (with a 1.5–4.6 MHz sweep) at
and producing neutrinos. Moreover, it has be- 250 Hz or more, and so high beam-pulse rep rates;
come apparent that FFAG designs need not be (b) high permeability, and so short cavities with
restricted to the “scaling” approach explored in high effective fields; (c) low Q (≈1), allowing
the 1950s. Dropping this restriction has revealed broadband operation.
a range of interesting new design possibilities,
which have been explored in a series of FFAG
Workshops [4]. For recent reviews see [5, 6].
Scaling FFAGs Resonance crossing was a big
worry in the early days of AG focusing, because
of the low energy-gain/turn. The scaling principle
was therefore adopted, whereby the orbit shape,
optics and tunes are kept the same at all energies.
To first order the tunes are given by
νr2 ≈ 1 + k (1)
νz2 ≈ −k + F (1 + 2 tan2 α) (2)
where
- average field index k(r) ≡ r(dB̄/dr)/B̄
- average field at radius r is B̄ ≡ B(θ)
- magnetic flutter F ≡ (B(θ)/B̄ − 1)2
- sector spiral angle α ≡ r(dθ/dr). A similar 150-MeV FFAG has been built
Clearly, constant νr requires k = constant, im- at the Kyoto University Research Reactor Insti-
plying a magnetic field profile B̄ = B0 (r/r0 )k and tute (KURRI), together with injector and booster
a momentum profile p = p0 (r/r0 )k+1 . As a large FFAGs. The world’s first tests of accelerator-
k is usually chosen to minimize the radial aper- driven sub-critical reactor (ADSR) operation were
ture, constant νz requires a constant, high, value carried out there in 2009. An 11-MeV 70-mA
19
FFAG proton storage ring has also been built at A first 12-triplet ring (radius 6.25 m) would pro-
KURRI for boron neutron capture therapy using vide 250-MeV protons, and a similar second ring
an internal Be target, the high current being main- (radius 9.3 m) 400-MeV/u C6+ ions.
tained by ionization cooling (Sec.2.7.4). Vertical scaling FFAGs A more radical ap-
FFAGs are also of interest for muons. PRISM proach has been taken by Brooks [4, 11] who pro-
(Phase-Rotated Intense Slow Muon source), a poses a “vertical FFAG” where the beam follows
68 MeV/c DFD storage ring, is under construc- a helical path at fixed radius in a ring of super-
tion at RCNP Osaka for eventual installation at
conducting magnets, and shows that a scaling law

J-PARC (Sato et al. [9]). The muon bunches it Bz ∝ ekz produces skew-quadrupole focusing
collects will be rotated in phase space, reducing and constant tunes.
their momentum spread from ±30% to ±3% for
Linear nonscaling (LNS) FFAGs In a study
ultra-sensitive studies of rare muon decays. Op-

of FFAG arcs for recirculating linacs in 1997,
tics measurements have been carried out with α-
Mills and Johnstone pointed out that the rapid
particles.
acceleration (<20 turns) essential for muons al-
Scaling FFAG studies In addition, a number of lows betatron resonances no time to damage beam
different scaling FFAG designs are being stud- quality. Scaling can therefore be abandoned, the
ied [6], some of spiral- and some of radial-sector tunes allowed to vary, and a wider variety of lat-
design. These range from a fist-sized prototype tices explored. Moreover, using constant-gradient
of a 10-MeV machine for electron irradiation, to “linear” magnets greatly increases dynamic aper-
medium-sized sources for 230-MeV proton and ture and simplifies construction, while employ-
400 MeV/u ion therapy (for which the high pulse ing the strongest possible gradients minimizes the
repetition rates are clinically advantageous), to physical aperture. Johnstone et al. [12] applied
a chain of four FFAGs accelerating muons from this nonscaling approach to a complete FFAG
0.3 to 20 GeV for a neutrino factory (see also ring, showing that it would be advantageous to
Sec.1.6.15) at J-PARC. use superconducting magnets in which the field
For the 3.6-12.6 GeV stage of the reference strengths decrease outwards - i.e. positively bend-
scheme of the International Design Study of a ing Ds and reverse-bending Fs (Fig.1). The radial
Neutrino Factory (IDS-NF) Mori’s group pro- orbit spread would be reduced (allowing the use
poses a 161-m radius ring with 225 FDF cells; of smaller vacuum chambers and magnets), and
with 1.8 GV per turn at a fixed frequency of the orbit length C(p) shortened and made to pass
200 MHz, acceleration would occur over 6 turns through a minimum instead of rising monotoni-
within a stationary bucket (T. Planche et al. [10] cally as p1/(k+1) (Fig.2 (left)). The variation in
p.4506). A major advantage of FFAGs over linacs orbit period is thereby reduced, allowing the use
– either single or recirculating – is that their large of high-Q fixed-frequency rf.
acceptances in r and p reduce the need for muon C(p)’s parabolic variation and its parametric
cooling or phase rotation. Moreover the accelera- dependence can be derived by treating the F and D
tors themselves are significantly less costly.
Fixed-frequency acceleration between sta-
tionary buckets and by harmonic number jump-
ing (HNJ) have also been considered. The for-
mer “serpentine” method (see below) is possible
in scaling
√ FFAGs if the orbital period minimum
at γ = k + 1 lies within the acceleration range,
and has been successfully demonstrated by
Yamakawa et al. [4] on the NHV 10-MeV elec-
tron FFAG.
Quasi-scaling FFAGs A novel technique for
creating a scaling field with superconducting
magnets has been proposed by Machida ([10]
p.112) for the PAMELA cancer therapy FFAG.
This uses nested 2-, 4-, 6- and 8-pole solenoids
to approximate the r k law over the beam aperture. Figure 1: Scaling and nonscaling FFAG orbits.
20
Ch.1: INTRODUCTION
Circumference Variation (cm)
40
Scaling E a LNS-FFAG complex of three concentric 48-
Non-scaling cell rings to produce 250-MeV protons and 400-
30
MeV/u C6+ for cancer therapy, and show that
dangerous resonances can be avoided. They also
20
propose ([10] p.124) a lightweight LNS-FFAG
gantry, capable of accepting the whole extracted
10
momentum range at fixed field. This would be
composed of either superconducting or permanent
0
magnet triplets, and weigh only ∼1.5 tons.
10 15 20 -π/2 0 +π/2
Ruggiero ([5] p.84) proposed a chain of 3
Energy (GeV) Phase
LNS-FFAGs in the AGS tunnel as a 19-MW, 12-
Figure 2: Linear nonscaling FFAGs: (left) circumfer- GeV proton driver. Operation could be either
ence variation with energy; (right) acceleration path pulsed (100 Hz) or c.w. (requiring harmonic num-
(yellow) in longitudinal phase space. ber jumping).
Nonlinear nonscaling (NLNS) FFAGs For
magnets as thin lenses of strength qi (gradient × cancer therapy, the high rf voltage needed to cross
length) [13]. For symmetric F0D0 or triplet cells, resonances quickly would be too expensive. John-
and assuming qF = qD = q, stone’s “tune-stabilized” design [18] keeps the be-
tatron tunes constant by using NLNS magnets
12π 2 where the pole gap varies with radius and the
C(p) = C(pm ) + (p −pm )2 (3)
e2 q 2 N LF D sector edges are straight, but not all radial, pro-
where N is the number of cells, and LF D is the viding extra edge focusing. There are separate
(shorter) F–D spacing. The minimum is at pm = 250-MeV proton and 400-MeV/u C6+ rings, each
(4pc +eqLF D )/6 where the pc closed orbit is such with 4 triplet cells. A similar 1000-MeV proton
that BF = 0. The orbit radii r(p) show similar ring could be added to serve as an ADSR driver.
dependence, with distinct pmin . Rees ([5] p.74) also uses nonlinear field pro-
Lattices along these lines have been devel- files but a more complicated dFDFd cell structure,
oped. The IDS-NF has adopted a LNS-FFAG for to design a muon ring that is isochronous from 8
the final 12.6-25 GeV stage of muon acceleration to 20 GeV – a muon cyclotron. The extra magnets
(67 FDF cells, radius 111 m, decay loss 7.0% over provide an additional degree of freedom, render-
11.6 turns) [14]. ing Eq.(2), derived for two-component cells, no
With the orbit length varying by only 0.02%, longer applicable, so that spiral edges are not re-
first falling and then rising, Berg and Kosciel- quired. Méot et al. ([5] p.44) have carried out
niak [15] have shown that by exceeding a crit- tracking studies without any attempt at fine tuning
ical rf voltage an acceleration path can be cre- and find only small losses at a few resonances.
ated (Fig.2 (right)) that stays close to the voltage Using a similar lattice Rees has also designed
peak (crossing it three times), snaking between a non-isochronous 4-MW 10-GeV proton driver
neighbouring buckets (rather than circulating in- (C = 624 m) ([5] p.102). This would operate at
side them) just as in an imperfectly isochronous 50 Hz, fed by a 3-GeV RCS.
cyclotron. In order to demonstrate the novel fea- Insertions Beam injection and extraction is of-
tures of such a design – particularly the serpentine ten a design challenge for FFAGs, because of
acceleration outside buckets, and the crossing of the regular cell structure. Insertions providing
many integer and half-integer resonances – a 10- longer drift spaces can however be designed, as
20 MeV electron model (EMMA) has been built first shown for scaling FFAGs by Meads [19].
[16] and successfully commissioned at Daresbury. Machida ([10] p.558) has recently presented a
Initial experiments on these two features have design with 4 straights that are well-matched to
shown that the beam behaves as expected [17], the arcs over the full 30-250 MeV energy range.
confirming the viability of the LNS approach. Lagrange and Mori ([10] p.4503) have proposed
LNS-FFAGs have also been considered for racetrack versions of the PRISM and ERIT stor-
lower-energy applications with slower accelera- age rings, and have built a single straight scaling
tion, where resonance crossing is of more con- FFAG triplet for beam tests [4]. For NS-FFAGs
cern. Trbojevic et al. ([5] p.115) have proposed Rees ([5] p.74) showed that four 9-cell straights
21
could be inserted in his isochronous muon ring, power from the electric service to multiple recti-
matched acceptably to the main arcs at all mo- fier stages through an array of high voltage capac-
menta. Concentration of the rf allowed the cir- itors, while others use an assembly of high voltage
cumference to be reduced from 1255 m to 903 m. transformers [1, 2, 3]. Although there are several
different designs of the high voltage power sup-
References plies, the particle sources and acceleration meth-
ods are usually similar.
[1] K.R. Symon et al, PR 103 (1956) 1837
[2] E.M. McMillan, in “Experimental Nuclear Particle acceleration Charged particles are ac-
Physics”, v.3, 639-786 (Wiley, 1959) celerated in highly evacuated tubes to minimize
[3] K.R. Symon, PAC’03, 452 (2003); F.T. Cole, collision with residual gases. Particles are ex-
Proc. Cyc’01, Suppl. (2001) tracted from their sources (ions from plasmas,
[4] FFAG’11 Workshop; electrons from hot cathodes), focused, and accel-
http://www.cockcroft.ac.uk/events/ffag11 erated by strong electric fields created by the high
[5] C.H Prior (ed.) ICFA Beam Dynamics Newslet- voltage potentials. The contours and strengths of
ter 43, 19-133 (2007); the electric fields are determined by the shapes
www-bd.fnal.gov/icfabd/Newsletter43.pdf and spacings of the accelerating electrodes with
[6] M.K. Craddock, K.R. Symon, RAST, 1, 65 intermediate potentials. The use of multiple over-
(2008). lapping electrodes prevents spark discharges in
[7] M. Aiba et al, EPAC’00, 299 the acceleration tube, protects the insulating rings
[8] S. Machida et al, PAC’03, 3452 between the electrodes from scattered particles
[9] A. Sato et al, EPAC’06, 2508; EPAC’08, 3389 and permits the use of high potentials and strong
[10] Proc. IPAC’10, http://www.jacow.org electric fields.
[11] S.J. Brooks, HB2010, 96,
http://hb2010.web.psi.ch/ High voltage generation The many applica-
[12] C. Johnstone, W. Wan, A. Garren, PAC’99, 3068 tions can require potentials from as low as 70 kV
[13] M.K. Craddock, FFAG2003; S. Koscielniak, to as high as 5 MV. For voltage ratings below
M.K. Craddock, EPAC’04, 1138 300 kV, conventional single-stage transformer-
[14] J.S. Berg et al, PAC’12, 241 rectifier systems can be used, but for higher volt-
[15] J.S. Berg, Proc. Snowmass 2001, T503 (2001); ages, multiple-stage cascaded rectifier systems
S. Koscielniak, C. Johnstone, ibid., T508 (2001) are needed to insulate the secondary dc circuits
[16] R. Barlow et al, NIM A624, 1 (2010) from the primary ac power source. A variety of
[17] S. Machida et al, Nature Physics 8, 243 (2012) | such multiple-stage systems have been developed.
doi:10.1038/nphys2179 The different methods for coupling ac power to all
[18] C. Johnstone et al, PAC’11, 2116 (2011) of the rectifier stages are illustrated in Fig.1 [2].
[19] P.F. Meads Jr., PAC’83, 244 In Fig.1a, ac power is inductively coupled in se-
ries from each of the transformer-rectifier stages
to the next stage. The dc outputs of each stage
1.6.7 High Voltage Electrodynamic
are connected in series to produce high voltage
Accelerators
dc power. In Fig.1b, ac power is inductively cou-
M.R. Cleland, IBA Industrial, Inc.
pled in parallel from a common primary winding
High voltage electrodynamic accelerators are also to all of the secondary windings. In Fig.1c, ac
called potential-drop or direct-action systems. power is capacitively coupled in series from each
They increase the kinetic energies of ions and rectifier stage to the next. Transformers are not
electrons by connecting particle sources to high needed in the rectifier stages. In Fig.1d, ac power
voltage generators and accelerating the particles is capacitively coupled in parallel to the rectifier
to a variety of targets at ground potential. In con- stages.
trast to electrostatic accelerators (Sec.1.6.5) that In series-coupled systems, the lower stages
transport charges mechanically from ground to in the cascaded rectifier circuit must transmit ac
the particle source, the high voltage power sup- power to the upper stages, so the dc voltage gen-
plies in electrodynamic accelerators convert low erated under load is reduced in the upper stages.
voltage alternating current (ac) to high voltage In parallel-coupled systems, all stages receive
direct current (dc) by means of cascaded recti- ac power directly from the primary source and
fier circuits. Some of these systems transfer ac the voltage generated under load is the same in
22
Ch.1: INTRODUCTION
(a) (b) (c) (d) Parallel-coupled systems The parallel-
coupled cascade circuit proposed by Schenkel
[11] predated the Greinacher circuit, but it was
3 2 2 not used for very high voltage generators because
3 2 3 2
3 of the difficulty of making capacitors that could
withstand the dc output voltage. During the 1950s
and 1960s, Radiation Dynamics, USA, developed
the parallel-coupled cascade circuit used in
1 1 1 1 Dynamitron accelerators [12]-[15]. High voltage,

1–98 high frequency ac power at about 100 kHz is
8355A169
generated by a triode-driven resonant system
Figure 1: Cascade generators. 1—Power supply; 2— consisting of an iron-free transformer and a pair
stage; 3—voltage rectifying and multiplying circuits. of semicylindrical electrodes which surround the
rectifier assembly. Power is capacitively coupled
from these electrodes to semicircular corona rings
all stages. Parallel coupling allows the use of connected to the rectifier junctions. The rectifiers
many more rectifier stages, reduces the internal are connected in series to produce high voltage
impedance of the high voltage generator and in- dc power. Insulation between components is
creases the amount of current and power that can provided with compressed SF6 gas. Dynamitrons
be provided for particle acceleration. with voltages up to 5.0 MV have been produced
with electron beam power ratings up to 300 kW.
Capacitive Cascade generators Two-stage tandem ion accelerators using this type
Series-coupled systems Cockcroft and Walton [4] of high voltage generator have been produced
used a high voltage generator with multiple rec- by Radiation Dynamics, Inc. [16, 17]. General
tifier stages capacitively coupled to a source of Ionex Corporation also produced parallel-coupled
ac power. This type of system is commonly tandem accelerators called Tandetrons for high
called a Cockcroft-Walton accelerator. Their energy ion implantation in silicon wafers [18].
series-coupled voltage multiplying rectifier circuit Similar systems are now made by High Voltage
was proposed earlier by Greinacher [5]. During Engineering Europa in the Netherlands.
the 1930s and 1940s, the firm N.V. Philips, the
Netherlands, produced many air-insulated high Inductive Cascade generators
voltage generators and accelerators based on this Series-coupled systems In the 1950s and 1960s,
concept with voltages up to 3.0 MV [6]. High Voltage Engineering Corporation developed
During the 1950s, Emile Haefely & Co the Insulating Core Transformer (ICT). This is a
Ltd, Switzerland, developed a symmetrical series- three-phase, multi-stage rectifier cascade circuit
coupled cascade circuit with three columns of causing magnetic coupling to transfer low voltage,
pacitors. The middle column reduced the ac ripple low frequency ac power from the primary wind-
voltage at the high voltage terminal. They pro- ings at the low voltage end of the transformer to
duced many accelerators using this type of gen- an array of high voltage secondary windings. The
erator for a variety of research applications such magnetic cores of the secondary windings are sep-
as electron microscopy, ion injection into higher arated by thin sheets of solid insulating material.
energy rf accelerators, and separation of high en- Rectifiers and filter capacitors convert ac power
ergy particle beams. Potentials up to 4.0 MV to dc at each stage. All of the dc circuits are con-
have been obtained with compressed gas insula- nected in series to produce the high voltage at the
tion [7, 8]. During the 1980s, the firm Nissin High output end of the assembly [19]-[21]. The largest
Voltage Co Ltd, in Japan, simplified the symmet- ICT accelerators produce voltages up to 3.0 MV
rical cascade circuit by omitting the middle col- with electron beam power ratings up to 100 kW.
umn of capacitors for industrial applications that ICTs are now made by Vivirad High Voltage in
do not need low ac ripple voltage. Voltages up to France and Wasik Associates in the USA. Sealed
5.0 MV have been obtained with compressed gas Air Corp. makes 0.5 MV ICTs for irradiating their
insulation. Their largest accelerator of this type heat-shrinkable food packaging films.
is rated for 30 mA of dc electron beam current or Parallel-coupled systems During the 1970s,
150 kW of beam power [9, 10]. the Budker Institute of Nuclear Physics developed
23
several types of single-phase, multi-stage trans- [8] G. Reinhold, K. Truempy, J. Bill, IEEE Trans.
former systems. All of the secondary windings Nucl. Sci., NS-12, No.3 (1965) 288
are coupled to an external coaxial primary wind- [9] K. Mizusawa et al, Radiation Phys. & Chem.,
ing which extends the full length of the high volt- Vol.31, Nos.1-3 (1988) 267
age assembly. The Elita is a resonant pulse trans- [10] S. Uehara et al, Radiation Phys. & Chem.,
former with a solenoidal high voltage secondary Vol.42, Nos.1-3 (1993) 515
winding. The ELT is a low frequency system [11] M. Schenkel, Elektrotechnische Zeitschrift,
for generating high voltage dc power. It has in- Vol.40 (1919) 333
sulated magnetic cores inside the multiple sec- [12] M.R. Cleland, Voltage Multiplication Apparatus,
ondary windings. The ELV is a similar multi- U.S. Patent No. 2,875,394 (1959)
[13] M.R. Cleland, P. Farrell, IEEE Trans. Nucl. Sci.,
stage, transformer-rectifier system without mag-
NS-12, No.3 (1965) 227
netic cores. ELV accelerators produce electron

[14] M.R. Cleland, C.C. Thompson, H.F. Malone, Ra-
beam power ratings of 50 kW with dc potentials
diation Phys. & Chem., Vol.9, Nos.4-6 (1977)
up to 2.0 MV and 400 kW of beam power at 547
0.4 MV [2, 22, 23]. A 2.5 MV dc system has been [15] C.C. Thompson, M.R. Cleland, NIM B40 (1989)
made by LBNL [24]. 1137
During the 1970s and 1980s, the Budker In- [16] P.R. Hanley et al, IEEE Trans. Nucl. Sci., NS-16
stitute, the Efremov Institute for Electrophysical (1969) 90
Apparatus in St. Petersburg, and the Institute for [17] S.A. Cox, P.R. Hanley, IEEE Trans. Nucl. Sci.,
High Temperatures in Moscow developed three- NS-18 (1971) 108
phase, transformer-rectifier high voltage systems. [18] K.H. Purser et al, SPIE Vol.530, Soc, Photo-
These use continuous long iron cores at ground Optical Instr. Eng. (1985) 14
potential with primary windings on each core. [19] R.M. Emanuelson, Insulating Core Transform-
The high voltage secondary windings are insu- ers, U.S. Patent No. 3,274,526 (1966)
lated from the cores and primary windings. Recti- [20] R.J. Van de Graaff, High Voltage Electromag-
fiers connected in three-phase bridge circuits con- netic Apparatus Having an Insulating Magnetic
vert ac to dc power. The Teus system uses a sin- Core, U.S. Patent No. 3,289,066 (1966)
gle bridge rectifier circuit with three high voltage [21] R. Emanuelson, R. Fernald, C. Schmidt, Radia-
windings and can produce 0.3 MV dc and 75 kW tion Phys. & Chem., Vol.14 (1979) 343
of electron beam power. The Aurora system uses [22] G.I. Budker et al, Charged Particle Accelera-
multiple bridge rectifier circuits with lower volt- tor, SU Patent No. 589698 (1973), U.S. Patent
age per stage and can produce 0.75 MV dc and No. 4,016,499, GB Patent No. 1454485
[23] R.A. Salimov et al, Radiation Phys. & Chem.,
100 kW of beam power [2, 25, 26, 27]. More pow-
Vol.57, Nos.3-6 (2000) 661
erful systems have been developed by the Efre-
[24] L.L. Reginato et al, AIP Conf. Proc. 392 (1997)
mov Institute for applications that do not need
1305
voltages higher than 1.0 MV. [25] B.I. Albertinsky, M.P. Svin’in, Cascade Genera-
tors, Atomizdat (1980)
References [26] M.P. Svinin, Radiation Phys. & Chem., Vol.18,
Nos.5-6 (1981) 1353
[1] W. Scharf, Particle Accelerators and Their Uses, [27] V.N. Lisin et al, High Voltage Transformer-
Part 1, Accelerator Design, Part 2, Applications Rectifier Device, U.S. Patent No. 4,338,657
of Accelerators, Harwood Academic (1986) (1982)
[2] E.A. Abramyan, Industrial Electron Accelerators
and Applications, Hemisphere (1988)
[3] M.R. Cleland, Chapter in Radiation Processing 1.6.8 Induction Linac
of Polymers, Hanser (1992) R. Bangerter, LBNL
[4] J.D. Cockcroft, E.T.S. Walton, Proc. Royal Soc.
London, Series A136 (1932) p.619 Induction linacs (IL) are employed in applica-
[5] H. Greinacher, Zeit. fur Physik, Vol.4 (1921) tions that require combinations of beam current,
195 beam energy, and pulse lengths that are not eas-
[6] A. Bouwers, A. Kuntke, Zeit. fur Technische ily achieved using rf accelerators. The first large
Physik, Vol.18 (1937) 209 IL was the Astron Injector at LLNL [1]. Al-
[7] W. Heilpern, Helvetica Phys. Acta, Vol.28 (1955) though induction acceleration had been used for
24
Ch.1: INTRODUCTION
Core Cavity
some time (e.g. the betatron, Sec.1.6.2), Ref.[1]
is usually credited for the invention of IL. Since Acceleration
then more than 40 ILs have been built. For a Gap
comprehensive review with listings of For a com- Beam
prehensive review with listings of major ILs and
their characteristics, see [2].
Conventional linacs produce long trains of
beam pulses at radio frequencies. The beam cur-
rent is usually < 1 A. In contrast, ILs produce Pulser Pulser

beam pulses at much lower repetition rates, but
much higher currents (e.g. >10 kA). The As- Acceleration Short
Gap Circuit
tron Injector originally produced 350 A of elec-
trons (3.7 MeV, pulse length 300 ns, burst rep- Beam
etition rate >1 kHz) to create a magnetic field
in the Astron magnetic fusion device. Subse-
quently ILs to produce beams for electron ring ac-
celerators (experimental collective accelerators)
were built at Berkeley and Dubna [3, 4]. Flash
radiography is another important application of Figure 1: upper: An IL with two induction modules
ILs, e.g. FXR and DARHT in the US, AIRIX (stages). The beam is the secondary of a series of
in France, and DRAGON-I in China [2]. The transformers. lower: Alternatively, one can think of
ATA, a 50-MeV IL, originally built for defense an induction module as a shorted transmission line.
studies, currently holds the record for the high-
est voltage (kinetic energy). Other applications with a changing magnetic field by V = −dΦ/dt,
that have been suggested or implemented in-
where Φ = B · dA is the magnetic flux in
clude FELs, microwave power generation, elec-
the core. The integral is over the cross-sectional
tron beam welding, food irradiation, pulsed neu- is uni-
tron sources, treatment of materials, treatment of area of the core. If one assumes that B
chemical and nuclear waste, and tunneling in rock form
over the area, then V = −AdB/dt, or
[2]. These applications use electron or proton V dt = A ΔB. This quantity is often referred to
beams. Heavy ion inertial confinement fusion simply as the volt-seconds of the core, while ΔB
and high energy density physics are applications is referred to as the flux swing.
that require beams of heavier ions (Li to U). The Another way to achieve induction accelera-
old hardware from the ATA has been modified to tion is by changing the area occupied by the mag-
build NDCX-II, a lithium accelerator designed for netic field rather than the field itself (line-type in-
basic studies of hot dense matter [5]. duction accelerator, in contrast to the core-type
[3]). In the line-type machines, an induction cav-
The IL concept Fig.1 (upper) illustrates the ba- ity is basically a radial or axial transmission line.
sic concept of induction acceleration. A pulser (or A ferromagnetic core is not used. The line-type
modulator) provides the power to energize an in- cavities usually have very low impedance. In-
duction module. The electric acceleration field is deed, a principal function of the core in a core-
confined primarily to the axis of the accelerator type module is to provide a high impedance in
by the conducting walls – the induction cavity – parallel with the load (beam). In both induction
surrounding the induction core. The core is usu- accelerators and rf accelerators, the accelerating
ally made of ferromagnetic (or ferrimagnetic) ma- electric field is associated with a changing mag-
terial although it could be simply air or vacuum. netic field. However, rf cavities are often driven
One may view the induction module as an electri- at resonance and induction cavities are not. Also,
cal transformer. The beam is the secondary of this induction cores are often large (∼1m dia.). The
transformer. An IL consists of one or more (often large size, together with the use of ferromagnetic
many) induction modules placed in series. One material, allows the induction accelerators to be
can also think of an induction cavity as a shorted used at longer pulse lengths than rf accelerators.
transmission line as shown in Fig.1 (lower). In For ILs, the pulse length lies in the range from
any case, the accelerating voltage V is associated tens of ns to tens of μs.
25
ILs are related to other devices, e.g., the in- losses. Lamination is usually achieved by wind-
ductive voltage adder, where a solid conductor re- ing the cores from thin tape (thickness ≤ 50 μm).
places the beam. For example, typical losses in iron-based metal-
Induction technology The main components of lic glass are ≈ 800 J/m3 at a pulse length of
an IL are the pulsers, the beam transport system, Δt ≈ 1 μs and a flux swing of 2.5 T. The losses
and, in the case of a core-type machine, the induc- scale approx. as ΔB 2 /Δt. (Because of hystere-
tion cores. sis, the losses at 2.5 T do not drop appreciably
below 100 J/m3 even at very long pulse lengths.)
Depending on the voltage, pulse length, and

other characteristics, the pulsers may be based A newer class of nanocrystalline materials such
on simple switched capacitors, pulse forming net- as Hitachi’s Finemet and Vacuumschmelze’s Vit-
works, or pulse forming lines. Switches for the roperm has losses that are typically several times
pulsers include thyratrons, spark gaps, magnetic lower than those of the older metallic glasses [7].
switches, and solid-state devices. To maximize Research topics Many ILs have now been built,
ΔB, a reset pulser normally magnetizes the core but IL technology is still not as highly devel-
in one direction before it is pulsed in the opposite oped as rf accelerator technology. For example,
direction by the main pulser. although DARHT has now achieved its design
The beam transport system of an IL must be goals, during construction there were some unan-
capable of carrying high current. One can esti- ticipated problems involving the behavior of fer-
mate the maximum current that a transport chan- romagnetic materials and the distribution of volt-
nel can carry by setting the beam self force equal ages within the induction cavities. This exam-
to the applied focusing force of the lattice. The ple points to the importance of continuing devel-
actual current will be less than this maximum be- opment of low-loss, low-cost ferromagnetic ma-
cause of emittance. A detailed calculation is re- terials having predictable, consistent properties.
quired in each individual case. For continuous In this regard, the nanocrystalline materials men-
solenoidal focusing the maximum current is ap- tioned above appear to be promising. Also, since
proximately ILs often have relatively large apertures to carry
high beam current, the development of large, in-
IS = 4 × 105 (Z/A)βγ(Ba)2 Amp. (1) expensive, highly reliable, high-gradient insula-
where Z and A are the charge and atomic mass tors is also an important research topic. More-
numbers, B is in Tesla, and a is beam size in m. over, some IL applications such as high energy
For a magnetic quadrupole transport system the density physics and inertial fusion will likely re-
maximum beam current is approximately quire the development of novel beam transport
systems such as compact, multi-beam arrays of
IQ = 8 × 105 (βγ)2 (ηBa) Amp. (2) superconducting quadrupoles. Research on this
topic is in its infancy [8]. Finally, the development
where η is the effective occupancy factor (the
of advanced pulsers, e.g., those using solid state
fraction of the lattice occupied by quadrupoles).
switches can be expected to have a profound in-
For electrostatic quadrupoles B is replaced by
fluence on IL design and applications. For exam-
E/(βc) where E is the electrical field in V/m.
ple, in one experiment [9], an uncooled induction
Solenoids are usually preferred for light particles
core was run for a few weeks at a repetition rate
(e− , p), but quadrupoles appear preferable for
of approximately 100 Hz. The pulser used capaci-
heavy ions, particularly at the high velocity end
tive storage and an array of field effect transistors.
of the machine. There has been research on neu-
The measured efficiency into a resistive load ap-
tralized and collective focusing systems [6].
proached 50%. With continued development, ILs
Magnetic materials used for induction cores may become competitive with rf accelerators for
include steel tape, nickel-iron tape, ferrite, and high average power applications.
a class of amorphous metallic glasses such as
Metglas R , originally produced by Allied Chem-
References
ical, but now produced by Metglas R Inc., a sub-
sidiary of Hitachi Metals. Core losses are an im- [1] N.C. Christofilos et al, RSI 35 (1964) 886
portant consideration, particularly for lower beam [2] Induction Accelerators, K. Takayama and R.J.
currents. For core materials other than ferrite, the Briggs (Eds.), Springer: NY, (2010)
core must be laminated to minimize eddy-current [3] J.E. Leiss, PAC 79, p.3870
26
Ch.1: INTRODUCTION
[4] A. Faltens, D. Keefe, Proc. Linac Conf., LANL 2. In the electrostatic accelerator a charging
Report, LA-9234-C (1981) p.205 system conveys the charge from ground to
[5] A. Friedman et al, Physics of Plasmas 17, 056704 the insulated high voltage terminal (origi-
(2010) nally by an insulating belt, today mostly by a
[6] S. Humphries, Jr. et al, PAC 81, p.3410 chain of metal cylinders connected by insu-
[7] A.W. Molvik et al, LINAC98, p.320 lating links), see Sec.1.6.5.
[8] G. Sabbi et al, NIM A 544 (2005) 285
3. In a linear accelerator (or linac) the beam
[9] W. Barletta et al, LBL-35960 (1994)
travels through a series of hollow tubular

electrodes. Along the Widerøe principle the
1.6.9 Industrial Accelerators electrodes are connected alternatively to op-
R. Hellborg, Lund U. posite poles of an rf-supply. Along the Al-
varez principle the structure consists of a set

The years around 1930 can be taken as the start- of resonator tubes having an rf-voltage of the
ing point of the accelerator era when people con- same phase applied to them. Inside each res-
ducted development work along different princi- onator tube, a potential distribution exists,
ples. In 1937 the first accelerator for applied use see Secs.1.6.11, 1.6.12.
was constructed. This accelerator could, in over 4. The RFQ accelerator has a symmetry
the energy range 0.5 – 1.2 MeV and electron beam corresponding to that of an electrostatic
up to 3 mA, produce x-rays for cancer treatment quadrupole. It combines the action of
up to 0.01 C/kgair per mA beam current. Today focusing, bunching and acceleration, see
accelerators are applied in very diverse fields. Of Sec.1.6.18.
the more than 20 000 accelerators in operation 5. In the cyclotron the beam is bent into a circu-
around the world only a few hundred are used in lar path by a magnetic field and the particles
(applied) physics research. One third is used in orbit inside two semicircular metal cham-
medical applications, two-thirds are used for in-
bers. These so called “dees” are connected
dustrial applications.
to a source of alternating voltage. In the gap
Electron beam processing is utilized by many between the dees, the particles feel an accel-
major industries, including the plastics, automo- erating voltage and gain energy twice during
tive, rubber goods, petrochemical, wire and ca- each circle, see Sec.1.6.4.
ble, electrical-insulation, textile, semiconductor,
packaging and pollution control industry. In food 6. In the synchrocyclotron the frequency of
industry electron beams are used for sterilization. the applied rf-field between the two dees is
Ion beam accelerators are used for a broad variety slowly decreased as the particle energy in-
of industrial applications like micro-machining creases to compensate for the relativistic ion
and for national security applications, which in- mass. Only one bunch at a time can be sent
clude x-ray inspection of cargo containers and through this accelerator. (An alternative to a
stewardship of civil and military nuclear materi- synchrocyclotron is to modify the cyclotron
als. and divide the magnetic field into sectors of
Tables 1, 2, 3, and 4 present some examples alternating high and low fields. This is called
of industrial and medical applications of acceler- a sector focusing or azimuthally varying field
ators. The applications are given in column 1. In cyclotron.) See Sec.1.6.4.
column 2 a suggested technique/method is given. 7. In the synchrotron the massive magnet (in
In column 3 the necessary accelerator parameters cyclotrons and synchrocyclotrons) is re-
are to be found. In column 4 suggestions for more placed by a ring of bending magnets. The
readings are given. rf-field and the magnetic field strength are
The different types of accelerators used in in- varied to keep the orbital radius of the beam
dustrial applications are: constant. Only one bunch at a time is sent
through the accelerator, see Sec.1.6.21.
1. The cascade accelerator (also called high
voltage accelerator) in which the high- Industrial and medical applications of accel-
voltage unit consists of a multiplying erators will increase in the future and new appli-
rectifier-condenser system (first used by cations can be foreseen [1]. New isotopes will
Cockcroft and Walton), see Sec.1.6.7. be developed for the PET-technique. The isotope
27
Table 1: Material processing by a high energy Ion Beam.
Industrial/medical applica- Suggested technique/ Accelerator parameters: To learn

tions method to be used typea /particle /current more, see
/energy e.g.
Fabrication of semiconductor Ion implantation of Casc, el stat, linac (rare cases ch.1 in [3]
devices and materials dopants and other ions RFQ) /several different /a
few μA to 100 mA /hundreds
of keV (or even less) to MeV
Filter and permeable mem- Irradiation of thin Casc, el stat /heavy ions /tens ch.17 in [2]
brane production films with heavy ions, of μA /10–100 MeV

followed by chemical
etching of the films
Material (nanoscale) engi- Irradiation by MeV Casc, el stat /p, He, light and ch.25 in
neering in fields like biomed- ions heavy ions up to Bi /1 fA – [4], several
ical tissue engineering, quan- 100 mA /0.1 keV – 20 MeV chapters in
tum devices, optical and [2]
magnetic information storage
technology etc.
Treatment of cancer tumors Irradiation with high Cycl, synchrotr /p, 12 C / e.g. [5]
by radiation doses of ions 12 C:140–400 MeV u−1
Mutogensis of seeds for plant Irradiation by keV – Casc, el stat /heavy ions /e.g. [6]
breeding MeV ions N2+ :1018 ions cm−2 /e.g.
N2+ :30 keV
Clinical use of radio nuclides Radio nuclides pro- Mainly cycl but even linacs ch.4 in [3],
for therapy duced by nuclear are used/ mostly light ions ch.19 in
reactions and placed like p, d, 3 He, 4 He /0.1 – 10 [4], [7]
close to the tumor, mA /mainly 10 – 20 MeV
either inside the body
“branchytherapy” or
close to the body
“teletherapy”
a The abbreviations used for different types of accelerators are: casc = cascade accelerator; el stat =
electrostatic accelerator; linac = linear accelerator; cycl = cyclotron (and synchro-cyclotron); synchrotr
= synchrotron; RFQ = Radio Frequent Quadrupole accelerator. For detailed descriptions of different
types of accelerators see sections in this handbook and [4].
99m Tc will be produced by accelerators. The need with energy- and intensity-modulated beams us-
for 99 Mo from nuclear reactors will therefore be ing 3-D scanning, Sec.1.6.14. To get quantitative
reduced. A number of hadron (p and 12 C) therapy information about sample composition and sur-
facilities are under construction. Hadron irradi- face structure of nanometer technology products,
ation offers better dose distribution than conven- a demand for MEIS (Medium Energy Ion Scat-
tional photon and electron beams do and for rela- tering) facilities can be foreseen. MEIS is per-
tively radio-resistant tumors, treatment with heav- formed at projectile energies of 100–300 keV. It
ier hadrons (12 C and 20 Ne) offers great potential involves very sophisticated/complicated detection
benefit. An excellent depth dose can be reached equipment, see Ch.7 in [2].
28
Ch.1: INTRODUCTION
Table 2: Material characterization by a high energy Ion Beam.

/energy e.g.
Dating extremely small sam- Accelerator Mass tandem casc, el stat/10 Be, [8]
Spectrometry (AMS) 14 C, 26 Al, 36 S, 41 Ca
ples (down to 1 μg or even
less) for geological, cosmo- etc./extremely low/tens
logical and archeological pur- of MeV or higher

poses
Quantifying extremely low Accelerator Mass tandem casc, el stat /10 Be, [8]
concentrations of traces in Spectrometry (AMS) 14 C, 26 Al, 36 S, 41 Ca etc. /ex-
small samples (down to 1 tremely low /tens of MeV or

μg or even less) for e.g. higher
biomedical and environmental
purposes
Industrial use of radio nu- Radio nuclide pro- Mainly cycl but even linacs ch.4 in [3],
clides, mainly for imaging and duction by suitable are used /mostly light ions [7]
for tracking physical or bio- nuclear reactions like p, d, 3 He, 4 He /0.1 – 10
logical processes in for exam- mA /mainly 10 – 20 MeV
ple plants or animals
Material analysis with bulk or Ion Beam Analysis single or tandem casc, el stat ch.5 in [3]
depth sensitive nuclear meth- (IBA) including /light ions /μA /a few MeV
ods methods like RBS,
PIXE, PIGE, ERDA,
CPA etc.
Pollution control (e.g. welding Analyze with the casc, el stat /protons /1 – 10 [9]
dust, combustion emissions, PIXE-technique μA /2 – 3 MeV
transport)
Geophysical exploration in the Detection of gamma- Casc or sealed tube generator ch.6 in [3]
petroleum industry, moisture rays emitted from a /neutrons produced in a target
content measurements in tim- neutron initiated nu- after acceleration of p or d
ber and construction industry clear reaction
Safeguard inspections for ex- Detection of neutrons Casc /neutrons produced in a ch.22 in
plosives, chemical weapons, or gamma-rays emit- target after acceleration of p [4], ch.6 in
fissionable material etc. ted from a neutron or d to a few hundred keV [3]
initiated nuclear re-
action
Biomedical use of radio nu- Radio-nuclide pro- Mainly cycl but even linacs ch.19 in
clides, mainly for imaging and duction by suitable are used /mostly light ions [4], [7]
radiotracers to track physical nuclear reactions like p, d, 3 He, 4 He /0.1 – 10
or biological processes mA /mainly 10 – 20 MeV
a For the abbreviations used for different types of accelerators, see footnote in Tab.1.
29
Table 3: Material processing by a high energy Electron Beam.

/energy e.g.
Sterilization, polymerization, Irradiation by an elec- casc, el stat /electrons /10 μA ch.28 in [4]
disinfection, vulcanization tron beam with doses – 100 mA or more /up to a
and several other similar up to 100 kGy few MeV
industrial processes
Material processing like Electron beam irradia- casc, el stat /electrons /up to ch.2 in [3],
welding, cutting, drilling, tion 1 A /up to 200 keV (in rare [10]
heat treatment and melting cases MeV)
Treatment of cancer tumors Irradiation with high linac /electrons /doses up to [11]
by an external beam doses of electrons or 2 Gy per treatment /10 – 20
by x-rays obtained af- MeV
ter the electron beam
has collided with a tar-
get
Table 4: Material characterization by a high energy Electron Beam.

/energy e.g.
Production, quality control Spectroscopic and Synchrotron producing x- ch.8 in [3],
and control of regulatory re- imaging technique by rays of 1 eV – 100 keV (far [12]
quirements x-ray irradiation infrared to hard x-ray)
Computed tomography (CT) x-ray or neutron irradi- linac /electrons /absorbed ch.7 in [3],
or 3-D scanning of industrial ation dose up to tens of Gray/up [13]
products. Metrology of in- to 10 MeV (quite low x-
ternal structures of complex ray energy for micro and
parts or assemblies. Non de- nanotomography)
structive testing and inspec-
tion of e.g. air cargos, haz-
ardous waste etc.
References [4] R. Hellborg (ed.), Electrostatic Accelerators –

Fundamentals and Applications, Springer, 2005
[1] R.W. Hamm, M.E. Hamm, Physics Today, June [5] IAEA-TECHDOC-1560, Intern. Atomic Energy
2011, p.46 Agency, Vienna (2007); K. Noda, CERN Courier
[2] R. Hellborg, H.J. Whitlow, Y. Zhang (eds.), 50:5, 22-24 (2010)
Ion Beams in Nanoscience and Technology, [6] Introduction to ion beam biotechnology, Trans.
Springer, 2009 L. Yu, T. Vilaithong, I. Brown, Springer, 2006
[3] R.W. Hamm, M.E. Hamm (eds.), Industrial Ac- [7] IAEA Technical report Series 468 (2009)
celerators and their Applications, World Scien- [8] R. Hellborg, G. Skog, Mass Spec Rev. 27:5, 398
tific (2012) (2008)
30
Ch.1: INTRODUCTION
[9] S.A.E. Johansson, J.L. Campbell, K.G.

Malmqvist, Particle-Induced X-ray Emission
Spectrometry (PIXE), Wiley, 1995
[10] D.E. Power et al, Welding Journal, 79-2, 35
(2000)
[11] B.R.R. Persson, Radiation Therapy, in Radia-
tion – at Home, Outdoors and in the Workplace
(eds. D. Brune, R. Hellborg, B.R.R. Persson and
R. Pääkkönen), Scandinavian Sci. Pub., Oslo

2001
[12] K. Nasta, C.-C. Kao, Synchrotron Radiation Figure 1: Schematic of laser-structure accelerator,
News, 20-4, 7 (2007)
showing laser injection and asymmetric (flat) acceler-

[13] R.H. Bossi et al (eds.), Non-destructive Testing ating beam, from [2].
Handbook, Vol.4, American Soc. for Non-destr.
Testing, Columbus, OH (2000)
when P is a limiting consideration, can be re-
laxed in favor of structures that closely resem-
1.6.10 Laser, Wakefield and Plasma ble laser resonators [2], as in Fig.1. These struc-
Accelerators tures, which rotate the laser’s transverse electric
J. Rosenzweig, UCLA field to give longitudinal acceleration, are fabri-
The rf linac has evolved continuously from its cated of low loss dielectric that gives breakdown
birth after the World War II, an innovation made limits in the GV/m range. To exploit higher fields,
possible by the development of high power mi- the particle trajectory may also be bent to have a
crowave sources for radar. The success of these periodic transverse component, as in the inverse
devices is undeniable; however, with the insis- free-electron laser (IFEL).
tent demand for increasingly high energy, both the Dielectric laser accelerator (DLA) structures
size and cost of these “big science” machines has have been developed in a wide variety of de-
reached the billion-dollar level. The problem of signs, many inspired by optical sciences, e.g. pho-
miniaturizing the linac has been recognized and tonic band-gap structures. Such devices have un-
has stimulated research aimed at reinventing the dergone testing at SLAC, with DLA-based ac-
accelerator scheme based on new physical princi- celeration demonstrated [3]. These studies have
ples that use lasers, plasmas and wakefields (ad- recently concentrated on pre-bunched beam in-
vanced acceleration techniques) [1]. jection, with the optical-scale bunching provided
Present rf linacs are limited in acceleration by an upstream IFEL. The IFEL interaction for
gradient to <∼100 MV/m by two effects: first, the micro-bunching is now a common tool for injec-
limit on maximum power and stored energy avail- tion into optical accelerators. IFEL staging was
able and, second, by structure breakdown. To shown at BNL [4]; two stages of IFEL were syn-
reach present limits, power and field energy con- chronized, the first giving micro-bunching and
siderations have yielded an optimization at shorter the next accelerating the beam with small energy
rf wavelength λEM , in X-band. Advanced accel- spread. High energy gain was shown later at
erators aim to enhance field gradients to >GV/m UCLA [5], with 14.5 MeV injected beam acceler-
(allowing a TeV machine in a km), arguing for diated to ∼35 MeV peak in 25 cm. This experiment
minishing λEM by at least an order of magnitude. also displayed the higher harmonic IFEL interac-
In this frequency regime, metals are too lossy to tion, which has extended the utility of IFEL for
be of use; the present approach to accelerator de- micro-bunching [6].
sign must be abandoned in favor of new, more While the IFEL has shown more experimen-
general “structures”. Further, to obtain high pow- tal progress than the DLA, its application to high
ers at short λEM new sources must be considered. energy physics accelerators is limited by syn-
The most obvious coherent EM power source chrotron radiation. It is now under study, how-
available is the laser, which easily produces co- ever, for sub-GeV applications, such as inverse-
pious infrared-to-optical power P , to >TW. With Compton scattering (ICS) production of MeV
such high P , the paradigm for accelerating struc- photons for nuclear materials identification. On
tures may be changed. Axisymmetry, essential the other hand, DLAs promise more control and
31
flexibility than IFELs, and scale to high-energy The first DWA experiments in the high fre-
application well. Indeed, with a structure, an op- quency rf regime (∼30 GHz) at ANL [8] achieved
tical undulator [7] may be built that permits an field gradients up to 100 MV/m. The field scaling
x-ray FEL using < 500 MeV beams. These may of the coherent Cerenkov radiation (CCR) emit-
be most attractively obtained from a laser structed in a DWA is Nb ω 2 , where Nb is the number of
ture or IFEL accelerator; this combination is now beam charges and ω is the EM frequency. To ex-
under study to give compact, all optical, coher- cite a high frequency mode, the drive beam pulse
ent x-ray source, i.e. the GALAXIE project, a (or train) must have significant Fourier compo-
UCLA-SLAC-BNL-Penn State consortium. nents at this frequency. With sub-100 fs beams

DLA research presents many challenges. As available in recent years, GV/m DWA fields be-
many schemes utilize geometries that are sub- came accessible. First measurements at SLAC
λEM in one transverse direction, the normalized FFTB have shown that > 5 GV/m fields can
emittance in this direction must be very small, in be excited in a SiO2 DWA before breakdown [9].
the N ∼10−9 m-rad range. Further, the longitudi- Subsequent experiments at UCLA have demon-
nal dynamics is very challenging, as relevant time strated that the CCR exiting a DWA tube is quite
scales are 4 orders of magnitude smaller than in narrow band [10], yielding a unique high power
current devices. Thus, while much shorter λEM THz source for frequency-specific applications.
is still demanded, options midway from rf to op- DWA research is continuing in many lab-
tical, in the mm-THz regime, are attractive. Here, oratories: the ANL AWA, BNL ATF, and the
the dynamics requirements are relaxed, and a key nascent SLAC FACET facility. These programs
breakdown mechanism in dielectrics, avalanche will explore breakdown mechanisms, resonant
ionization, is mitigated. However, there is no multi-pulse excitation, acceleration in multi-GV
readily available EM power source that can gen- wakes, enhanced transformer ratios, and are now
erate Ez > GV/m in a THz-scale structure. This developing asymmetric (slab) structures [11]. The
need has stimulated the invention of wakefield ac- use of slab structure with asymmetric beams per-
celeration, in which the EM fields of the acceler- mits acceleration of higher charge trailing beams,
ating wave are created inside of the structure it- and also strongly suppresses transverse wake-
self by an intense, relativistic particle beam. This fields [12]; these advantages are shared with many
drive beam may be of lower quality and energy optical accelerator structures.
than a trailing, accelerating beam. Further, the Ultimately, both the DLA and DWA are lim-
drive beam may be specially shaped (in, e.g. a ris- ited by breakdown. For Ez > 5 GV/m, one must
ing triangular current profile) to give much larger consider the accelerating “structure” to be broken
acceleration in the trailing beam than deceleration down, that is to be plasma. The scenario in which
in the driver; this relationship is termed the trans- a drive beam excites plasma waves that support
former ratio. When one uses a dielectric structure large Ez is termed the plasma wakefield acceler-
in such a scheme (Fig.2), it is called a dielectric ator (PWFA) [13]. In the linear regime, where
wakefield accelerator (DWA). the plasma density n0 greatly exceeds the beam
density nb , the wake wave is an electrostatic os-
cillation with plasma frequency ωp = kp c =

4πe2 n0 /me . As long as the wake is efficiently
excited (kp σz < 1, kp σr < 1), the excited field
Ez ∼= Nb e2 kp2 /2∼(nb /n0 )EW B , where the wave-

breaking field EW B ≈ 0.96 n0 (cm−3 ) V/cm.
When nb ≈ n0 Ez approaches EW B ; for a modest
plasma density of 1018 cm−3 , this is ∼100 GV/m.
The first PWFA was performed at ANL [14].
Simultaneously, it was proposed to use plasma
wakefields for creating lenses (Sec.7.3.13) with
extreme focusing strength for a linear collider fi-
Figure 2: Schematic of dielectric wakefield accelerator nal focus. The case with kp σz > 1, kp σr < 1 and
showing drive beam hollow dielectric tube generating nb > n0 (underdense plasma lens) is most attrac-
Cerenkov wakes. tive, producing nearly aberration-free focusing in
32
Ch.1: INTRODUCTION
witness beams. With even shorter beams avail-

able at the LCLS (σz ≈ 0.6 μm, Q = 20 pC) a
blowout regime PWFA with Ez > 1 TV/m may
be created (Fig.3). This option is under study
[21].
While the PWFA has yet to produce a low
energy spread and low N beam, the laser-driven
version of the scheme (laser wakefield accelera-
tor, LWFA) has done so, trapping and accelerat-

ing plasma electrons to yield N ∼mm-mrad, >kA
beams up to ∼1 GeV [22]. With some improve-
ments in beam quality, these beams may create

a “table-top” FEL [23]. Further information on
LWFA and other laser-driven plasma accelerators
is found in Sec.7.3.13 and [24].
References
Figure 3: PWFA field Ez excited by 2 fs, σx = 77 nm,
20 pC beam, with n0 ≈ 8 × 1019 cm−3 . [1] See, e.g. the Advanced Accelerator Concepts se-
ries proceedings, the last being 14th Advanced
Accelerator Concepts Workshop, Eds. S.H. Gold,
an electron-evacuated ion column, with fields lin- G.S. Nusinovich, AIP Proc. 1299 (2010)
ear in r and independent of ζ = z − vb t. For [2] T. Plettner et al, PRST-AB 9, 111301 (2006); J.B.
the example of a n0 = 1018 cm−3 plasma the Rosenzweig, A. Murokh, C. Pellegrini, PRL 74,
equivalent quad strength in the lens is 2πen0 = 2467 (1995)
30 MT/m, > 5 orders of magnitude stronger [3] T. Plettner et al, PRL 95, 134801 (2005)
than a conventional magnet. Overdense plasma [4] W.D. Kimura et al, PRL 92, 054801 (2004)
(nb < n0 ) lensing was demonstrated at UCLA [5] P. Musumeci et al, PRL 94, 154801 (2005)
[15]; underdense plasma lensing has been ob- [6] C.M.S. Sears et al, PRL 95, 194801 (2005)
served at FNAL [16]. [7] T. Pletter, R. Byer, PRST-AB 11, 030704 (2008)
[8] W. Gai et al, PRL 61, 2756 (1988)
Indeed, PWFA operation with nb n0 (and
[9] M.C. Thompson, PRL 100, 21 (2008)
kp σz < 1, kp σr < 1) where electrons are expelled [10] A.M. Cook et al, PRL 103, 095003 (2009)
from the beam channel has considerable advan- [11] G. Andonian et al, PRL 108, 244801 (2012)
tages [17]. In addition to the linear ion-column [12] A. Tremaine, J. Rosenzweig, P. Schoessow, PR
focusing inside the blowout “bubble” region, as E56, 7204 (1997)
the currents supporting the (pure EM) wave lay [13] P. Chen, PA 20, 171 (1985)
outside of the bubble, Ez (see Fig.3) is indepen- [14] J.B. Rosenzweig et al, PRL 61, 98 (1988)
dent of r just as in a standard rf linac. Ion column [15] G. Hairapetian et al, PRL 72, 1244 (1994)
focusing in the blowout regime was first shown at [16] M.C. Thompson et al, Phys. Plasmas 17, 073105
ANL [18], and acceleration at > 100 MeV/m ob- (2010)
served at FNAL. With the advent of short (σz ≈ [17] J.B. Rosenzweig et al, PRA – Rapid Comm. 44,
20 μm), high charge (3 nC) beams at the SLAC R6189 (1991)
FFTB, peak Ez in the blowout regime at 10- [18] N. Barov et al, PRL 80, 81 (1998)
100 GV/m could be reached [19]. These studies [19] M.J. Hogan et al, PRL 95, 054802 (2005)
demonstrated a wide variety of effects, including [20] I. Blumenfeld et al, Nature 741, 445 (2007)
betatron-oscillation x-ray production, and a dou- [21] J.B. Rosenzweig et al, NIM A,
bling of the beam energy at the drive beam tail doi:10.1016/j.nima.2011.01.073
[20]. [22] W.P. Leemans et al, Nature Physics 2, 696 (2006)
New experimental work in PWFA at FACET [23] M. Fuchs et al, Nature Physics 5, 826 (2009)
[24] E. Esarey, C.B. Schroeder, W.P. Leemans, Rev.
has begun, along with a DWA program. It will
Mod. Phys. 81, 1229 (2009)
explore wakefield acceleration with positrons, as
well the creation of ramped beams and separate
33
1.6.11 Linear Accelerators for Electrons High Voltage

Modulator High Power
G.A. Loew, SLAC Klystron
Electron (positron) linear accelerators (linacs) Peak Power P0
are axially rectilinear structures which capture a ~c/2 3c/4 c Load
beam from an electron injector (positron source),
and accelerate it to a desired energy [1, 2, 3]. Triode Buncher Accelerator
These linacs rely on rf (typically 0.5-30 GHz) Gun
Velocity
energy to produce the accelerating electric field. Modulating
In Out
Above a few MeV, dc machines no longer work ~30° ~5°
Prebuncher
because cumulative high voltages are difficult to Δφ 70
= ~ ° 2–98
8334A2
obtain.
The rf energy is used to launch a traveling Figure 2: Schematic of a single-section, single-power-

wave (TW) or a standing wave (SW) in an array of source linac, with typical injector. Δφ is bunch length
microwave cavities or cells (Fig.1). (Secs.1.6.12, in units of rf phase. The gridded triode gun is some-
2.4.6, 7.3.6). In both cases, the structure is de- times replaced by a laser-driven photocathode or an rf
signed so that the phase of the wave is synchro- gun where the cathode is embedded inside an rf cavity.
nized with the beam. Because of the small rest Progressive bunching is graphically illustrated. Multi-
mass of e± , above a few MeV, the needed rf phase section linacs are simply constructed by adding sec-
velocity vp ≈ c. Generally, when entering the tions linearly beyond the first one.
linac, particles are already considerably bunched,
and what little compression remains to be done exhibits a second upper-branch (there are an in-
occurs during the first few MeV of acceleration finity of such higher-order modes, HOM) which
(Fig.2). The bunches can then ride at a constant rf intercepts the vp = c line at a point with nega-
phase and corresponding acceleration field. tive slope (backward-wave HEM11 mode). These
The structures shown in Fig.1 are periodic (or modes are commonly called wakefields when they
quasi-periodic, see below) in that their cell ge- are excited by a beam bunch. The wakefields then
ometry is invariant with respect to longitudinal cause particles later in this bunch or subsequent
translation. The periodic loading is necessary be- bunches to be deflected transversely, producing
cause a smooth waveguide would have vp > c. emittance growth and eventually beam break-up
The irises or other periodic obstacles create an and particle loss. This problem can be controlled
infinite family of space-harmonic modes (Fig.3). by appropriate solenoidal focusing (at low en-
The fundamental mode (n = 0) generally has ergy) and quadrupole focusing (at higher energy),
the largest amplitude and is used for acceleration. or by cell detuning (see below).
The Brillouin diagram [4], Fig.3, corresponds to Most linac structures are made out of high
a TW structure in which this fundamental mode quality copper which, however, has a finite loss
has vp = c at a phase shift of 2π/3 per cell. Fig.3 which causes attenuation. For a given amount
of pulsed rf peak power P0 injected into an ac-
p celerator section, it is not economical to extend
(a) Load
its length beyond the point where ∼70% of the
power has been dissipated. In a TW structure, the
remaining power is dumped into a load. The in-
put is matched so that there is no reflection at the
P source. In a SW structure, there are also losses
(b)
and while some power is reflected from the end,
it is possible to match the input so that there is no
power reflection in steady state. Because a stand-
3–98
8334A1
ing wave can be decomposed into two oppositely
moving traveling waves, the only efficient way of
Figure 1: (a) Traveling-wave structure with matching operating a SW structure is in a mode with a π-
input iris and matched load at output. (b) Standing- phase shift per cell on axis, which means that both
wave structure with total reflection at output end and traveling waves are confluent and produce accel-
matching input iris (steady state). eration (in reality, for mode stability reasons, the
34
Ch.1: INTRODUCTION
11–97
8355A44
Figure 3: Typical Brillouin diagram for a disk-loaded waveguide, showing fundamental accelerating mode oper-
ating at 2π/3 phase shift per cell, and one branch of a higher-order HEM11 deflecting mode.
actual phase-shift per cell in e.g. the SW side- scheme is to equip every cell with four symmet-
coupled structure invented at Los Alamos [5] is rical side-openings which enable the HOMs with
π/2). Advantages and disadvantages of TW vs. both horizontal and vertical polarizations to leak
SW structures are discussed in [5]. out into manifolds while leaving the fundamental
All periodic linac structures are characterized accelerating mode undisturbed [6].
by a figure of merit called the shunt impedance While the accelerator structure is the heart of
per unit length, r. The steady-state no-load en- the linac, there are other essential components and
ergy acquired by a particle riding on top of the subsystems. Fig.2 shows a high power rf source
synchronous fundamental wave is which in most machines is a klystron (magnetrons
are used in single-section low energy machines)
V = K P0 r (1) with its associated high voltage modulator. Be-
where the proportionality constant K < 1 for a cause of the high peak power required (typically
TW structure depends on the attenuation of the 4-80 MW), these klystrons have low duty cycles
section, and K = 1 for the perfectly matched SW (∼100 Hz) and pulse lengths ∼ a few μs. To in-
structure, assuming no appreciable field reduction crease the peak power, modern linacs sometimes
due to beam loading. If the linac consists of N use rf pulse compression systems (Sec.6.7), which
identical sections, then the total energy gain is temporarily accumulate the rf energy in a storage
N V . With a peak beam current I, the steady-state device, thereby enhancing the peak power emit-
energy (i.e., after the appropriate filling times) is ted in a pulse at the expense of its width. Fig.2
reduced by a subtractive term ∝ rI. also shows an injector with its triode gun, pre-
TW structures may be designed to be quasi- buncher and buncher. Other important subsys-
periodic. Two examples stand out. The first is the tems include rf drive and phasing, focusing, vac-
constant-gradient structure in which dimensions uum, alignment, water cooling, ac power, instru-
are tapered so as to decrease the group velocity mentation and control [3].
linearly with length. The fields are thereby caused For a given total energy, the length of a linac
to be constant as opposed to exponentially decay- can be reduced by increasing its accelerating gra-
ing as a function of length, as is the case in reg- dient. Typical machines use gradients from a few
ular constant-impedance structures. The second MV/m up to proposed ∼100 MV/m. At low gra-
one is a variation of the former where the cavity dients, the structure can sometimes suffer from
dimensions are varied so as to keep the accelerat- a parasitic resonant phenomenon (multipactor) in
ing fields approximately constant while giving the which a surface-emitted electron gets accelerated,
HOM frequencies a Gaussian distribution which hits the surface elsewhere (or returns to the same
causes them to decohere rapidly in time after their point), and then ejects new secondary electrons
excitation by the beam. A further variation of this which produce an avalanche condition (Sec.6.9).
35
Table 1: Linac parameters for SLC (achieved) [7], an rf frequency of 1.3 GHz with superconducting
NLC (design) [8], and LCLS (design) [9]. technology, and the other (CLIC) using 12 GHz
SLC NLC LCLS with room temperature technology, are still being
Energy (GeV) 50 2×250 15 considered, among the many options proposed in
Gradient(MeV/m) 17 57 17 2003 [12].
frf (GHz) 2.856 11.424 2.856
klystron References
#/linac 224 3270 70 [1] Linear Accelerators, ed., A. Septier, P.M. Lapos-
peak power(MW) 65 75 65 tolle, North-Holland (1970)
pulse length(μs) 3.5 1.5 3.5 [2] G.A. Loew, R. Talman, AIP Conf. Proc. 105
e− /bunch (1010 ) 4 1.0 0.63 (1983) p.1
N x (mm-mrad) 45 4 1.0 [3] R.B. Neal et al, The Stanford Two-Mile Linear
N y (mm-mrad) 5 0.04 1.0 Accelerator, Benjamin (1968)
σz (mm) 1.0 0.15 0.03[a] [4] L. Brillouin, Wave Propagation in Periodic
σδ (10−3 ) 1.5 3.0 < 1[a] Structures, Dover (1953)
[5] R.H. Miller, Proc. Linear Acc. Conf., SLAC-303
[a] after full compression. (1986) p.200
[6] R.H. Miller et al, Proc. Linear Acc. Conf. (1996)
[7] P. Emma, PAC 95, p.606
At high fields, electron field emission takes place, [8] Zeroth Order Design for Next Linear Collider,
which can have a variety of deleterious effects: SLAC-474 (1996)
parasitic absorption of energy, wakefields, dark- [9] V.K. Bharadwaj, PAC 97
current producing spurious radiation and back- [10] G.A. Loew, Proc. Proton Linear Acc. Conf.
grounds, and eventually rf breakdown which can (1976) p.217
make the linac inoperative. Field emission is [11] J. Clendenin et al, Compendium of Scientific
caused by a variety of surface irregularities, im- Linacs (1996)
purities, dust, and contamination, and can be con- [12] G.A. Loew et al, Int. Linear Collider Tech. Rev.
trolled, up to a point, by proper surface fabrica- Comm. Second Report, SLAC-R-606 (2003)
tion, treatment and cleaning (Sec.6.13). Another
type of structure damage more recently discov-
1.6.12 Linear Accelerators for Protons
ered at high magnetic fields is excessive rf pulse
S. Henderson, FNAL
heating resulting in copper melting.
A. Aleksandrov, ORNL
Tab.1 gives the parameters achievable by the
full SLAC e− linac, as well as originally proposed First demonstrated in 1947 by Alvarez and co-
design parameters for the NLC and the LCLS. workers at Berkeley [1], proton linear accelera-
Electron and positron linacs are used for tors [2, 3] find use as dedicated linacs for nuclear
many purposes [10]. The largest number of elec- physics, injectors for high-energy synchrotrons
tron linacs (many thousands) are radiotherapy ma- for nuclear and particle physics, and drivers for
chines which can be found in hospitals worldwide spallation neutron production. Often, H− ions
(energy ∼4-30 MeV, x-rays produced by electrons (Sec.7.1.4) are accelerated instead of protons to
impinging on targets, or direct electron radiation, make use of charge-exchange injection in a down-
Sec.1.6.14). Low energy linacs are also used stream synchrotron or accumulator ring. Tab.1
in industry for sterilization of various materials summarizes parameters of existing proton/H− lin-
and products, x-ray radiography, etc. (Sec.1.6.9). ear accelerators. Beam energies extend to 1 GeV
Most of these linacs consist of single sections and and duty factors to several percent. Peak beam
power sources. Linacs of higher energy (50 MeV- currents reach ∼200 mA for protons and ∼50 mA
50 GeV) are used in laboratories for nuclear and for H− beams, with pulse lengths ranging from
particle physics and as injectors into e± storage ∼10 μs to 1 ms.
rings of various types [11]. As this article is being Proton linacs generally utilize RFQs
written, the international particle physics com- (Sec.1.6.18) or Cockroft-Waltons (Sec.1.6.7) as
munity is still assessing the possibility of build- injectors, and therefore provide acceleration for
ing an e+ e− linear collider in the TeV CM en- particle velocities of β 0.03 to β 0.9. Accel-
ergy range. Two main designs, one (ILC) using eration is provided by TM010 -like fields (Sec.6.6)
36
Ch.1: INTRODUCTION
−
Table 1: Parameters of operating proton/H linacs (taken from [4] except where noted).
Facility Species Design Structure RF Peak Beam Rep

Energy (energy Frequency Beam Pulse Rate
[MeV] [MeV]) [MHz] Current Length [Hz]
[mA] [μs]
IPNS/ANL [5] H− 50 DTL 200.06 12 90 30
FNAL [6] H− 400 DTL (116) 201.25 50 60 15
CCL 805
Indiana U. H− 7 DTL 425 1 400 5
LANSCE/LANL p/H− 800 DTL (100) 201.25 17(p) 825 120
CCL 805.0 11(H− )
BNL [7] H− 200 DTL 201.25 37 530 7.5

IHEP/Beijing p 35 DTL 201.25 40 150 12.5
KEK H− 40 DTL 201.07 18 80 20
Kyoto U. p 7 DTL 433 0.6 50 180
DESY H− 50 DTL 202.56 20 30 1
MMF/INR p/H− 600 DTL (100) 198.2 50 80 100
DAW 991.0
I-2/ITEP p 25 DTL 297 200 20 2
ISTRA/ITEP p 36 DTL 297 150 10 1
I-100/Protvino p 103 DTL 148.5 100 100 1
URAL-30/Protvino [8] p 30 RFQ-DTL 148.5 100 10 16.6
Linac 2/CERN p 50 DTL 202.56 180 120 2
ISIS/RAL H− 70 DTL 202.5 25 500 50
J-PARC [9] H− 400 DTL (50) 324 50 500 50
SDTL (191) 324
ACS 972
SNS/ORNL[10] H− 1000 DTL (87) 402.5 38 1000 60
CCL (186) 805.0
SCL 805.0
established in accelerating gaps (or cells) which Rf structures

are arranged in standing-wave multicell cavities Drift tube linac The Drift Tube Linac (DTL)
of two types. In the first, a long structure with (Sec.7.3.5), is the most common structure for ac-
tens or even hundreds of cells is excited by a celeration of proton/H− beams in the velocity
single rf power source. The physical spacing range 0.03 < β < 0.4 (∼0.5 MeV to ∼100
between cells varies along the length of the MeV) and so constitutes the low-energy portion
structure in accordance with the design particle of nearly all proton linacs. DTLs generally oper-
velocity profile. In the second, the distance ate in the frequency range ∼200-400 MHz. A full
between cavities is fixed, and relative beam/rf DTL system consists of one or more cylindrical
phase is adjusted by varying the rf generator pillbox resonant tanks, each excited in a TM010 -
phase. like standing wave mode. Drift tubes, placed with
The principal challenge in modern high- a nominal spacing of βλ, provide a field-free drift
intensity proton linac design and operation is the region during the decelerating portion of the rf cy-
minimization of beam loss due to halo growth cle. The accelerating field in each gap oscillates
[11, 12, 13]. An often stated loss criteria [14, 15] at the same phase and frequency (the zero-mode).
demands particle loss rates less than 1 watt of i is
The length of cell
beam power per meter in order to limit residual φs,i+1 − φs,i
li = + 1 βi λ (1)
activation dose rates to less than ∼100 mrem/hr 2π
at 30 cm after 4 hr cooldown. See Sec.2.4.6 for a where φs,i is the synchronous phase for cell i and
description of the beam dynamics in proton linacs. βi is the velocity at the exit of the cell.
37
The accelerating gradient is usually constant is in use at the Moscow Meson Factory [19]. An
from cell to cell, although a field ramp is some- Annular Coupled Structure (ACS) is being built
times included to maintain smooth longitudinal for J-PARC [20].
focusing. The synchronous phase at each cell Elliptical superconducting cavities The
(the “phase law”) is selected to provide ade- SNS utilizes elliptical SC multicell cavi-
quate longitudinal focusing (typically between ties (Sec.7.3.10) with geometric-beta (the
−30◦ and −20◦ ) and may also be ramped to ad- synchronous particle velocity) βg < 1, for
just phase acceptance, or for longitudinal match- acceleration from 186 to 1000 MeV [21, 22].
ing. Either permanent magnet or electromagnetic Two structures with βg = 0.61 and βg = 0.81
quadrupoles are contained within drift tubes to provide acceleration in the velocity range
provide transverse alternating gradient focusing. 0.55 < β < 0.90. The cavity is excited in
At higher energy, the effective shunt the π mode with cell spacing equal to βg λ/2.
impedance begins to decrease, so that the DTL The benefits of SC cavities realized in other
structure begins to become less efficient than applications, namely, high accelerating gradients,
other structures above ∼100-200 MeV. compact designs, large aperture, lower rf power
Coupled-cavity structures The various “high- requirements, etc., are gained in application to
energy”’ rf structures exploit the remarkable proton/H− linacs as well.
properties of the π/2 mode of excitation in a Other structures A large number of other
biperiodic array of coupled resonant cavities [16]. structure types have been studied and several have
In the π/2 mode, every other cavity is excited, been built. These include the CCDTL structure
while adjacent cavities are unexcited, and there- [23], and the crossed-bar H-cavities (CH-cavity)
fore do not contribute to beam acceleration. The under study for the FAIR project [24]. A seper-
π/2 mode has the following properties: i) the field ated drift tube linac structure (SDTL) consisting
amplitude in excited cells is independent of cell of short five-cell tanks with quadrupoles between
frequency errors in first order, therefore making tanks, is used at the J-PARC facility for accel-
the structure less sensitive to construction toler- eration from 50-190 MeV [25]. A DTL with
ances and tuning errors; ii) rf losses do not pro- space-periodic RFQ focusing (a RFQ-DTL) oper-
duce a phase shift in excited cavities; and iii) rf ates at IHEP Protvino [8]. Additionally, SC spoke
losses produce a cavity amplitude decrease only resonators are under consideration for the low-
in second-order. energy portion of future high-power proton linacs
[26].
Efficient acceleration is achieved by placing
the nominally unexcited cells off the beam axis, so Rf power considerations For a shunt
that the beam traverses every other cell in the long impedance defined by Rs = V02 /P , the shunt
array of coupled cavities. The relative rf phase impedance per unit length is Z = E02 /(P/L) and
shift between accelerating cells is therefore π, and the effective shunt impedance per unit length is
the (excited) cell spacing is nominally βλ/2. A ZT 2 = (E0 T )2 /(P/L), where E0 is the peak
single klystron powers a long coupled cavity linac axial electric field, T is the transit-time factor
(CCL) module, which consists of many multi- and P is the rf power. For example, SNS DTL
cell tanks or segments (each of which may have tank 6 (length 6.34 m), has ZT 2 = 39 MΩ/m,
∼10 accelerating cells) forming a long chain of E0 T = 2.8 MV/m, and so requires an rf structure
as many as 200 coupled oscillators. Segments are power of 1.25 MW to establish the field.
coupled one to the next by off-axis “bridge cou- Design codes The most commonly used de-
plers” that span the intercavity drift spaces where sign codes for rf cavities and proton linac
quadrupoles and diagnostic devices are located. structures are the POISSON/SUPERFISH codes
A high shunt impedance at higher velocity [27] for electromagnetic field calculation, and
(β > 0.4) makes these structures useful for ac- PARMILA [28] for generating the linac layout
celeration of a beam delivered by a DTL. The and performing beam dynamics computations.
most common coupled cavity structure is the side- Rf setpoint determination To ensure good
coupled structure used at the LANSCE linac [16], quality of the accelerated beam, the phase and
the FNAL linac [17] and the SNS linac [18], all amplitude of the rf cavities must be set and
of which operate at 805 MHz, and which follow a maintained very close to their design values
∼100 MeV DTL. The Disk and Washer structure (Sec.2.4.6). Typical requirements are less than
38
Ch.1: INTRODUCTION
1
0.95
0.9
RFQ transmission [rel. u.]
0.85
Unom = .172
I = 22.7mA
0.8
0.75
0.7
0.65
0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15
RFQ field [rel. u.]
Figure 2: Measured beam current vs SNS DTL rf

Figure 1: Measured transmission vs rf field amplitude
phase for several different DTL rf amplitudes. Curves
for the SNS RFQ compared to a model prediction.
with larger phase width correspond to higher rf field
amplitude.
1◦ phase error, and less than 1% amplitude error.
Beam-based tuning techniques that make use of
model-based methods are best suited.
RFQ tuning RFQ tuning requires setting the
rf field amplitude only. A fit of the measured
beam transmission vs rf power to a curve calcu-
lated from the model allows the correct set point
to be determined. An example is shown in Fig.1
[29].
DTL tuning The rf phase and amplitude must
be set for each DTL tank. One method, the ac-
ceptance scan, determines set points by compar-
ing the measured beam transmission vs rf phase to
the phase width predicted by a model (Sec.2.4.6).
Low-energy particles are absorbed in an energy
degrader, so that only properly accelerated beam
Figure 3: Measured difference in beam phase recorded
current is measured with a faraday cup. An exam-
by two beam phase monitors vs SNS DTL rf phase
ple SNS DTL acceptance scan is shown in Fig.2.
for two different DTL rf amplitudes. The solid curves
A second method, phase scan signature matching
show measured data and the points show results of the
[30], makes use of the measured beam phase at model-based fit.
one or two locations downstream of the tank vs
the tank rf phase and amplitude. The rf set points
and energy of the incoming beam are obtained by SCL tuning The rf phase and amplitude have
comparing the measured curve to the model. An to be set for each cavity in the SCL. A phase
example is shown in Fig.3 [31]. scan analogous to the one described for the DTL
is used to determine the input beam energy and
CCL tuning The rf phase and amplitude have
phase, and the cavity field amplitude [34].
to be set for each CCL module. In the Delta-
T scan method, the beam phase at two locations
downstream of the module is measured vs the References
module rf phase with the rf on and off. The [1] L.W. Alvarez et al, RSI 26 (1955) 111
set points are calculated using the measured data [2] T. Wangler, Principles of RF Linear Accelera-
and coefficients derived from the model [32, 33]. tors, Wiley (1998)
Phase scan signature matching analogous to that [3] P. Lapostolle, A. Septier, eds., Linear Accelera-
described for the DTL can also be used. tors, North-Holland (1970)
39
[4] J. Clendenin et al, CERN/PS 96-32(DI);

http://linac96.web.cern.ch/Linac96/
Compendium/COMPENDI.PDF
[5] J.C. Dooling et al, PAC 01, p.1653
[6] L. Allen et al, LINAC 96, p.329
[7] J. Alessi et al, LINAC 96, p.773
[8] Yu. A. Budanov et al, LINAC 04, p.285
[9] Acc. Technical Design Report for J-PARC,
JAERI-Tech 2003-44
[10] N. Holtkamp, LINAC 04, p.837
[11] Proc. 29th ICFA Adv. Beam Dyn. Wkshp,
HALO 2003, AIP Conf Proc. 693
[12] S. Nath et al, PAC 03, 1515

[13] F. Gerigk et al, LINAC 02, p.569
[14] 7th ICFA Mini-Workshop on High Intensity,
High-Brightness Beams (1999)
[15] N. Catalan-Lasheras et al, EPAC 02, p.1013
[16] E.A. Knapp, in [3], p.601
[17] C.W. Schmidt, PAC 93, p.1655
[18] J.H. Billen, PAC 01, p.1104
[19] V.G. Andreev et al, LINAC 1972, LA-5115,
p.114
Figure 1: Livingston Chart.
[20] V.V. Paromonov, KEK Report 2001-14 (2001)
[21] C. Rode, PAC 01, p.619
[22] G. Ciovati et al, PAC 01, p.484 where the center of mass energy is twice the beam
[23] J.H. Billen et al, LINAC 94, p.341 energy for equal energy beams (Sec.1.6.3)
[24] L. Groening et al, LINAC 04, p.42
[25] Y. Yamazaki, LINAC 04, p.554 References
[26] S.H. Kim, LINAC 04, p.549
[27] POISSON/SUPERFISH, LANL LA-UR-96- [1] M.S. Livingston, J.P. Blewett, Particle Accelera-
1834 tors, McGraw-Hill (1962)
[28] H. Takeda, LANL LA-UR-98-4478
[29] A. Aleksandrov, PAC 05, p.97
[30] T.L. Owens et al, PA 48 (1994) 169 1.6.14 Medical Applications of Accelerators
[31] J. Galambos et al, PAC 05, p.1491; J. Galambos, J. Alonso, LBNL
private communication 1.6.14.1 Radiation therapy
[32] K. Crandall, LANL Report LA-6374-MS (1976)
[33] A. Feschenko et al, PAC 05, p.3064 Ionizing radiation in sufficient doses kills cells.
[34] S. Henderson et al, PAC 05, p.3423 As a result of much biological and clinical re-
search, radiation is now highly effective for can-
cer therapy [1]-[4]. Precise tumor definition with
1.6.13 Livingston Chart
CT and MRI imaging is crucial, as is dose-
D. Hartill, Cornell U.
control to the percent level. Fig.1 shows energy-
The historical rise of the beam energy of accel- deposition characteristics of radiation modalities
erators when new technologies were developed used, beam-delivery systems must work within
to accelerate particle beams can be graphically these constraints to achieve 3-D conformation of
illustrated by the Livingston Chart. The chart dose to the tumor and avoid unacceptable damage
is named after Stanley Livingston who first used to nearby healthy tissue.
this method of showing this history. (Fig.1) [1]. X-ray therapy [5, 6] is the most widely used
The important parameter for new particle produc- radiation treatment. The x-ray depth-dose rela-
tion and interactions is the center of mass energy tion is basically exponential. Treating a tumor
which depends on the fixed target particle mass in deep inside a patient involves significant doses
addition to the beam energy. The chart in Fig.1 as- up- and down-stream of the treatment field. These
signs an equivalent beam energy for the colliders doses can be mitigated by multi-port treatments in
40
Ch.1: INTRODUCTION
scattering and range-straggling of protons, the tail

of the carbon curve comes from nuclear breakup
of the projectile into lighter (longer-ranged) frag-
ments. Ionization density for charged particles
varies as Z2 , so heavier ions in addition to having
sharper stopping points are more lethal to malig-
nant cells. Initial ion treatments used neon, but to-
day carbon is the ion of choice for sparing of nor-
mal tissue on the entry path and good cell killing

in the stopping region.
The well-defined stopping point also makes
it easier to conform the radiation dose to an

irregularly-shaped tumor by independently vary-
ing the position and energy of stopping parti-
Figure 1: Energy deposition vs depth for various radi- cles. Pencil-beam scanning systems have been
ation modalities. designed for this purpose, and are now in clinical
use at several facilities.
which beams are brought in from several angles Even with less sophisticated delivery sys-
and overlap at the tumor. Computer-controlled tems, dose distributions of protons and ion beams
IMRT (Intensity-Modulated Radiation Therapy) are considerably better than even the best x-ray
adjusts the entry angle, dose per field and field systems, and successful clinical programs with
shape (with multileaf collimators) to achieve ex- these beams have been ongoing for more than 50
cellent 3-D conformation for fields as large as years. Up until recently, the primary hurdle to
30 cm. widespread application has been limited availabil-
ity due to its high cost.
Accelerators for x-ray therapy 5-30 MeV S-
band electron linacs are the mainstay of radi- Accelerators for proton therapy Proton en-
ation therapy today (∼5000 worldwide, princi- ergy of 250 MeV allows penetration to 30 cm in
pal manufacturers: Varian, Elekta, GE, Mit- tissue. Average beam current of a few nA yields
subishi, Toshiba). X-rays are produced by elec- dose rates adequate for treatment times of ∼1-2
trons striking a heavy-metal target. The very min for all but the largest therapy fields. Early
broad bremsstrahlung spectrum is “hardened” by proton therapy was performed with fixed-energy
using absorbers to filter out contributions from synchrocyclotrons (e.g. Harvard, Uppsala); the
lower energies. S-band linacs are a highly suc- large therapy fields (≈ 20 cm dia.) at the required
cessful spin-off from high-energy and nuclear uniformity (≈ 5%) were produced using sophis-
physics programs. Their compactness, efficiency ticated scattering foil techniques; the range of the
and reliability have been key to their acceptance beam was varied by energy degraders in front of
for clinical applications. With overall lengths of the patient. These “passive” delivery systems are
1-2 m, and low rigidity of the electron beams, decoupled from the accelerator, requiring only a
very compact accelerator and beam transport sys- fast and reliable beam cutoff system. In these
tems are possible. The development of the isocen- first facilities, treatment port orientations were ad-
tric gantry (patient lies stationary while the x- justed by moving the patient.
ray beam is rotated around the tumor) facili- The more advanced delivery systems now be-
tates IMRT. Powered by either a magnetron or a ing developed require energy variability, precise
klystron, these accelerators operate at repetition beam-intensity control and above all high sta-
rates up to 1 kHz. bility and rapid response. The control systems
Proton and ion-beam therapy [7]-[13] Fig.1 for beam-delivery and accelerator must be tightly
reveals the advantage of using heavy-charged par- coupled. A continuous beam is preferred, with a
ticles (instead of e− ) for therapy. First recog- duty factor > 25%. Interfacing an advanced de-
nized by Bob Wilson in 1946, the Bragg Peak at livery system with a short-pulsed beam presents
the end of the particle’s range can provide a sig- difficult problems.
nificant concentration of dose into a tumor. The The first hospital-based proton therapy ac-
wider proton curve arises from the higher multiple celerator was the synchrotron at Loma Linda,
41
California, built by Fermilab. It has a weak- stallation of units started in 2012. The CPAC-
focusing lattice, injected by a 2 MeV RFQ with LLNL high-gradient DWA (dielectric wall accel-
a single-turn kicker. Operating on a 2-s cycle, the erator) induction linac aims for 100 MeV/m accel-
half-integer resonant extraction provides reason- erating gradient, compact enough to also fit in the
ably flat spills with a 25% duty factor at any de- treatment room. Proton beams generated by high-
sired energy up to 250 MeV. Beam is transported power lasers are being developed, but are a long
to two fixed-beam rooms and three gantry rooms. way from meeting clinical-beam specifications.
The field is advancing rapidly. For “conven- The “star dose” boost from capture of stopped
tional” layouts (one accelerator, and beam-lines π − and p̄ excites interest in the physics commu-
serving several independent treatment rooms de- nity. Several thousand patients were treated with
livered as a “turnkey” system) IBA (Belgium) π − in the 70’s/80’s at LAMPF, TRIUMF, PSI but
leads the pack with ten or more installations of disappointing clinical results halted trials. p̄ is out
their 235-MeV cyclotron-based system. Hitachi of economic reach for today’s technology.
and Mitsubishi have between them eight slow- FFAG’s (Sec.1.6.6) could provide energy
cycling synchrotron-based systems in operation. variability with cyclotron-like beam quality, and
Accel Gmbh developed a compact superconduct- rapid-cycling compact synchrotrons could pro-
ing cyclotron (based on a Henry Blosser design) vide cost savings. Linac energy-boosters are be-
and has two operating facilities now. Accel’s cy- ing developed in Italy.
clotron rights have been purchased by Varian that
Accelerators for ion-beam therapy An energy
is now marketing this technology.
of 400 MeV/u is needed for a carbon beam to pen-
Energy changes in today’s facilities are etrate 30 cm tissue, corresponding to a magnetic
mostly done outside the treatment room to avoid rigidity of 6 T-m. Isocentric delivery presents
quality degradation and neutrons. Cyclotron- a formidable challenge. Treatments with helium
based systems use a degrader and collimated beams began in the mid 1950’s at Berkeley’s
energy-selection channel in the transport line, ad- 184 Synchrocyclotron, while trials with neon
justments take a few seconds. Ion source current (and other ions) at the Bevalac took place be-
in the cyclotron must be increased (by up to 103 ) tween 1978 and 1993. All existing facilities
to maintain brightness (dose rate). Synchrotron- are synchrotron-based (slow-cycling). Most now
based systems are capable of pulse-to-pulse en- strive to deliver both proton and carbon beams
ergy variation. from the same accelerator, for clinical intercom-
Isocentric delivery is now standard. Because parisons.
of the high rigidity of the proton beam (up to The HIMAC facility in Chiba, Japan com-
2.5 T-m) gantries are very large (10-13 m dia). As pleted in 1994 has two 16 T-m synchrotrons
much as 3 m between the last bending magnet and (over/under) capable of 30 cm range with silicon
the patient (isocenter) is needed for treating large beams and has treated 7000 patients with carbon.
fields, and locating dosimetry and field definition Hyogo, completed in 2001 has treated 4000 pa-
equipment. Gantry diameter can be reduced by tients with carbon and protons. A third Japanese
starting the spreading process before the last mag- carbon facility at Gunma, is now running, and
net, but this increases the size and weight of this to more are under construction. GSI, starting in
magnet. 1997 treated 440 patients with carbon, employ-
While passive (scattering-foil spreading) sys- ing a sophisticated scanning system with control
tems still account for the majority of installed of all accelerator parameters for each pulse, from
delivery systems, active scanning is rapidly ad- ion source out. The Italian CNAO facility in Pavia
vancing. Next major technological hurdle is pa- came online in 2011. All these offer fixed field
tient/organ motion during treatment, requiring on- orientation: either horizontal, vertical, or oblique
line imaging instrumentation and live tracking by for their carbon beams, but some do have gantries
beam. for proton delivery.
Developing compact accelerators for reduc- HIT in Heidelberg treated its first patient in
ing size and cost of proton delivery is a hot topic 2010. This is a synchrotron-based facility with a
today. Noteworthy is the Mevion 9T supercon- 7 MeV/u RFQ/IH linac injection chain and two
ducting synchrocyclotron, mounted directly on a ECR-based ion-source front ends. It has two
gantry in the treatment room. Delivery and in- fixed-beam rooms, and one gantry room capable
42
Ch.1: INTRODUCTION
of the full rigidity beam from the accelerator. The specificity of the pharmaceutical. Though clini-
gantry is 13 m diameter, 25 m long, weights over cal success has been sparse, research continues.
600 tons and has its scanning magnets upstream Reactor neutrons have been principally used in
of the last 90◦ magnet. these studies. Initiatives now look to high-current
Two new European projects: MedAUSTRON (10-100 mA) low-energy (2.5 MeV) proton beams
in Austria and ETOILE in France are under con- producing neutrons via the {p-Li} reaction. Also
struction. At this time there are no active projects studied is driving subcritical reactor assemblies
in the US. with high-current proton beams from FFAG struc-
Mitsubishi built the Gunma facility, and is ac- tures.

tively pursuing further projects around the world.
IBA has designed a superconducting cy- 1.6.14.2 Radioisotopes
clotron capable of 400 MeV/u Q/A = 1/2 beams

(C6+ , H+2 ), offering significant cost reduction and Radioactive isotopes are widely used in both di-
reduced facility size for ion beam therapy. Pro- agnostic and therapeutic applications [14]. Trac-
tons are extracted by stripping foil dissociation of ers, isotopes either alone or attached to physio-
the hydrogen molecule. logically relevant molecules, are used for func-
Neutron therapy Fast neutrons (14-70 MeV) tional imaging, detecting activity or hyperactiv-
have been used for therapy for over 50 years. ity concentrations in tissues. Imaging is possible
With depth dose similar to lower-energy x-rays, using positron emitters (PET) such as 11 C, 18 F,
localization of dose into a well-defined volume or single photon emitters and suitably collimated
is difficult. However, the very high ionization gamma-ray detectors (SPECT). PET isotopes are
density (owing to low-energy proton knockons short-lived, and are produced with small single- or
from nuclear scattering) produces favorable re- dual-particle (H or H/D) cyclotrons (≤ 18 MeV)
sults in radioresistant tumors, located at shallow close to the end-use clinic.
depths, particularly salivary-gland tumors. Inter- While diagnostic isotopes emit hard gammas
est in this modality is not widespread in the med- to minimize absorption in the body, alpha and
ical community. D-T generators (14 MeV neu- beta emitters find therapeutic applications, max-
trons) were widely used for many years. Few imizing dose to the volume close to the iso-
remain today because of maintenance issues and topic application, delivered either by uptake of
recognized need for higher-energy neutrons. Cy- injected/ingested material (e.g. iodine in thyroid
clotrons producing neutrons via either {p-Be} or treatments) or by surgical implantation of radioac-
{d-Be} reactions in the 60-70 MeV range were tive seeds (brachytherapy).
employed in the 1970-80. Of these, only the U. Accelerator-produced isotopes are made with
Washington machine remains. Also still oper- proton (or alpha) beams of 35 MeV or lower
ating today are the neutron therapy programs at from cyclotrons (e.g. 201 Tl, 123 I, 67 Ga, 111 In).
Fermilab (60-MeV protons deflected onto a neu- Commercial production is concentrated in a few
tron target partway down the injector linac), and centers with elaborate distribution networks to
iThemba LABS, South Africa, using a 66 MeV provide rapid delivery of short-lived isotopes.
extraction port from their 200 MeV separated sec- Research isotopes are also produced at higher-
tor cyclotron. A 60 MeV SC deuteron cyclotron energy accelerators [e.g. 100-800 MeV protons
mounted on a ring in the treatment room provides from TRIUMF, BLIP (AGS-BNL), LANSCE
a compact source of neutrons from dissociation (LANL) and iThemba LABS (South Africa)].
of the deuterons striking an internal target. This 99 Mo/99m Tc is the workhorse of nuclear
machine at Harper-Grace Hospital in Detroit is medicine today, over 50,000 procedures are per-
currently off line for economic reasons, but may formed daily in the US. It is a prominent fission
restart in the future. fragment from HEU (highly-enriched uranium)
Slow neutron “capture therapy” (BNCT) has reactor cores. The supply of this isotope is not
a small though faithful following. Boron-loaded, secure today owing to the age and reliability of
tumor-seeking pharmaceuticals are administered the few production reactors, none of which are
to the patient, the tumor area is flooded with ep- located on US soil. Interesting options exist for
ithermal neutrons which are selectively absorbed accelerator-production of this isotope using high-
by the boron causing more radiation damage to current SC linacs producing milliampere beams
the tumor tissue. Critical to success is the tumor- of > 200 MeV protons.
43
Accelerators for isotope production PET iso- [6] F.M. Khan, The Physics of Radiation Therapy,
topes are most commonly produced by (p,n) re- 2nd Ed., Williams & Wilkins (1994)
actions with low-energy (11-15 MeV) cyclotron [7] Particle Therapy Cooperative Group (PTCOG)
beams. The commercially-available cyclotrons (http://ptcog.web.psi.ch/)
are compact, self-shielded, highly reliable, and [8] Ion Beams in Tumor Therapy, U. Linz,
totally automated. Targetry and autochemistry ed, Chapman & Hall (1995); http://juwel.fz-
units are usually included, providing complete juelich.de:8080/dspace/handle/2128/659
hands-off preparation of isotopes in a form ready [9] Ion Beam Therapy, U. Linz, ed, Springer, Bio-
for administration. Manufacturers include CTI logical and Medical Physics series (2012)
(Knoxville, TN), IBA (Belgium) and Ebco (Van- [10] Hadrontherapy in Oncology, Proc. 1st Int. Symp.
on Hadrontherapy (1993), Elsevier Excerpta
couver).
Medica, Int. Congress Series 1077 (1994)
Small cyclotron technology has been revolu-

[11] W.T. Chu, B.A. Ludewigt, T.R. Renner, RSI 64
tionized by the development of high-quality H−
(1992) 2055
ion sources, which solved the thermal, mechani- [12] W.T. Chu et al, LBL-33749 (1993);
cal and activation problems associated with beam http://www.osti.gov/bridge/product.biblio.
extraction. For energies up to 30 MeV, magnetic jsp?query id=0&page=0&osti id=10163935
fields in the cyclotron can be high, leading to very [13] A.M. Sessler, convenor, Workshop on Hadron
compact structures, but for higher energies the Beam Therapy of Cancer, Erice, 2009;
magnetic field must be reduced to avoid Lorentz http://erice2009.na.infn.it/programme.htm
stripping (Sec.7.1.8) of the H− ions. TRIUMF [14] NSAC Isotopes Subcommittee; http://www.
(500 MeV) has a maximum field of 5 kG. phy.anl.gov/mep/NSACI/. Relevant reports:
Beam currents for the PET isotope systems http://www.sc.doe.gov/np/nsac.html, “Com-
are modest (e.g. 50 μA). The higher-energy cy- pelling Research Opportunities using Isotopes”
clotrons used for production of longer-lived iso- (Apr 2009); and “Isotopes for the Nation’s
topes push the limits of current (up to 1 mA). Future” (Nov 2009)
Advanced radiography [15] The advent of [15] W. Thomlinson, NIM A319 (1992) 295
high fluxes of high quality x-rays from syn-
chrotron radiation sources has opened up opportu- 1.6.15 μ+ μ− Collider
nities for diagnostics with monochromatic x-rays. R. Palmer, BNL
A notable example is the coronary angiography
The idea of muon colliders was introduced by
program started at SSRL and continued at NSLS
Skrinsky et al [1]. Neutrino Factories [2] use
(BNL) and HASYLAB (DESY), in which an ex-
many of the same components.
posure taken with x-rays just above and just be-
low the K edge are subtracted, producing a de-
tailed image of coronary arteries with little con- 1.6.15.1 Collider
trast agent. Relevant x-ray energies are in the 10’s The main advantages of muons for a collider, as
of keV range. Storage rings e− or e+ beam ener- opposed to electrons, are:
gies >∼ 2.5 GeV are adequate for this purpose. For
lower energy rings, high magnetic field dipoles • Synchrotron radiation, ∝ E 4 /m4 , requiring
and wigglers are required. high energy electron colliders to be linear
and long, is suppressed, so that a muon col-
lider can be circular and smaller.
References
• Because circular, the muon bunches collide
[1] R.R. Wilson, Radiology 47 (1946) 487
many times, allowing larger emittances for
[2] W. Scharf, Biomedical Particle Accelerators,
given luminosity. The number of such col-
AIP Press (1993)
lisions is limited by the muon lifetime to
[3] W.H. Scharf, O.A. Chomicki, Physica Medica
XII (1996) 199
nturns ≈ 150 Bave ; where Bave is the aver-
[4] Reviews of Accelerator Science and Technology, age ring bending field in T.
A. Chao, W. Chou, ed, Vol.2, World Scientific • Synchrotron radiation, as bunches cross
(2009) (beamstrahlung, Sec.2.5.3), is suppressed,
[5] C.J. Karzmark, C.S. Numan, E. Tanabe, Medical resulting in much smaller collision energy
Electron Accelerators, McGraw-Hill (1993) spreads.
44
Ch.1: INTRODUCTION
Table 1: Parameters of Collider Rings.
C of m Energy TeV 1.5 3

Luminosity 1034 cm2 sec−1 1.1 4
Beam-beam Tune Shift .087 .087
Muons/bunch 1012 2 2
Ring <bending field> T 5.8 8.4
β ∗ at IP = σz mm 10 5
rms momentum spread % 0.1 0.1

Repetition Rate Hz 15 12
Proton Driver power MW ≈4 ≈4
⊥ Trans Emittance μm 25 25
Long Emittance mm 72 72
• S-channel Higgs production is enhanced by

a factor of (mμ /me )2 ≈ 40000.
But there are challenges: selecting polarized
muons is very inefficient; ring magnets and detec-
tors must be shielded from decay electrons; ac-
celeration and cooling must be fast to avoid decay
losses; and, at higher energies, neutrino radiation
is a significant hazard. Figure 1: schematic of Muon Collider.
Representative parameters of colliders at 1.5
and 3 TeV CM energies are given in Tab.1. The
initial muon 6-D emittance (2⊥ ) is ∼ 4 × 106
times the specified final value.
Synchrotron, electron, and stochastic cooling
are too slow, but ionization cooling (Sec.2.7.4)
appears feasible. Muons passing through an ab-
sorber, lose momentum in 3 directions, while only
the longitudinal momentum is restored by rf. In
the linear case, the equilibrium emittance in this
process is

β⊥ 1 14.1 2 1
o ≈
βv 2 mμ LR dγ/ds
where β⊥ is the beam Courant-Snyder parameter, Figure 2: Longitudinal vs. transverse emittances in
βv = v/c, mμ is the muon mass in MeV, LR is beam manipulations prior to acceleration.
the material radiation length, mμ dγ/ds is the en-
ergy loss per unit length. Hydrogen is the best ab- acceleration. All components of this scheme have
sorber material. Efficient cooling requires β⊥ to been simulated at some level. The main compo-
be tapered to lower values as the emittance falls. nents, numbered as in Figs.1 & 2, are
For 6-D cooling, the lattice must have dis- 1. A high intensity 8 GeV proton source, and
persion and an absorber geometry that yields a buncher that forms intense short (∼2 ns)
greater energy loss at higher momenta than lower. bursts.
This reduces the longitudinal emittance, but in- 2. A liquid metal target, able to withstand the
creases the transverse, giving ‘emittance ex- 4 MW beam, in a 20 T hybrid solenoid (wa-
change’. Combined with transverse cooling this ter cooled copper coils inside superconduct-
gives 6-D cooling. ing) to capture the pions. A tapered solenoid
The schematic of the most complete concep- transports them to a lower field decay region.
tual design [3] is shown in Fig.1. Fig.2 gives the
3. A system of rf cavities bunch the muons and
longitudinal vs. the transverse emittances in the
phase rotate [4] them into a train of both
sequence of beam manipulations prior to muon
muon signs.
45
4. Charge separation in a bent solenoid.

5. Separate 6-D cooling of each sign in peri-
odic lattices following slow (Guggenheim)
helices [7] (Fig.3b). Energy loss is in wedge
shaped liquid hysdogen absorbers, focus-
ing and bending by tilted solenoids, and re-
acceleration in vacuum rf cavities. The Heli-
cal Cooling Channel (HCC) [8] (Fig.3c) is

an alternative 6-D cooler. It is filled with
high pressure hydrogen gas that acts as the
absorber and suppresses rf breakdown.
Figure 3: 6-D cooling lattices: a) HFOFO Snake; b)

6. The multiple bunches are merged into single Guggenheim; c) HCC.
bunches, one of each sign. Merging in lon-
gitudinal phase space is done by phase ro- 1.6.15.2 Muon storage ring neutrino
tations using rf and drifts. Merging in the factories
transverse dimensions is done by kicking dif-
The main components of a neutrino factory [6] are
ferent bunches into transports of differing
the same as for a collider, but the requirements are
lengths (a trombone[9]) to bring them at the
less severe: more smaller muon bunches; far less
same time to differing transverse positions.
cooling (a factor ∼ 10 vs. ∼ 106 ); no emittance
7. 6-D re-cooling, of the now larger combined exchange, and acceleration to lower energies (4-
bunches, are continued to the lowest tech- 40 GeV vs. 0.75-1.5 TeV). Instead of a collider
nically feasible transverse emittance. The ring, the beam is injected into one or two long race
transverse emittances at this point are still track shaped storage ring, whose orientations and
∼10 times worse than those required, but the tilts are such that one of the long straight sections
longitudinal emittances are ∼100 times bet- points to distant neutrino detectors.
ter than needed.
8. Final 4D transverse cooling to the required 1.6.15.3 Technical challenges
transverse emittance is achieved with liquid A liquid mercury target has been tested [12] in
hydrogen in high field (30-40 T) solenoids, a proton beam at CERN and has demonstrated
and low energies, where the longitudinal multi-megawatt capability.
emittance rises rapidly from the adverse de- An experiment to demonstrate ionization
pendence of energy loss on energy. But this cooling [13] is under construction at Rutherford
is acceptable. Appleton Lab (RAL) in the UK.
9. The muons of the two signs can now be re- Experiments with rf in magnetic fields, at
combined, with suitable spacing. 805 [14] and 201 [15] MHz, have shown damage
10. Initial acceleration is in linacs, with frequen- and/or limited acceleration gradients. It is pro-
cies first low, but rising as the bunches be- posed [18] that field emitted electrons are acceler-
come shorter. By 200 MHz, the cavities ated by the rf and focused by the magnetic fields,
should be superconducting. The linacs are causing damage by cyclic heating. Several possi-
followed by Recirculating Linear Acceler- ble solutions are under study:
ators, and then one or more pulsed syn- 1. rf in high pressure hydrogen gas does not
chrotrons [10]. [16] show the problem, and the effects of a
11. The collider ring [11] must be isochronous relatively low intensity proton beam did not
to allow very short (1 cm) bunches. Chro- cause breakdown [17]. With a small admix-
maticity from the IP is locally corrected to ture of an electro-negative gas, the loss of rf
allow low β ∗ insertions. Tungsten shield- power appeared acceptable [17].
ing is needed to keep the decay electrons 2. Tests, in fields [14], showing severe damage
from heating and quenching the ring’s SC of copper surfaces, have shown little or no
magnets. The detector also requires special damage on beryllium surfaces. A cavity with
shielding from these electrons. all beryllium end walls is under construction
to test if this is a solution.
46
Ch.1: INTRODUCTION
3. Improvements in surface treatments, includ- 1.6.16 Neutron Sources

ing Atomic Layer Deposition [19] should J. Wei, Michigan State U.
suppress the initial electron production, and H. Lengeler, CERN
there are plans to test this.
Another challenge is to build the small bore Neutron production methods include fission
30-40 T ‘all superconducting’ solenoids for the (reactor-based), isotope and excited-state decay,
final cooling. Lower fields give somewhat lower fusion (accelerated deuteron beam bombarding
performance. HTS materials have sufficient cur- deuteron or tritium targets), e− -bremsstrahlung-
rent density for even higher fields, but it has not induced photon production and photonuclear re-
been demonstrated. A test that could reach 40 T action, charged-particle reaction, and spallation
is under construction [20]. [1]. The latter three accelerator-driven sources
rely on charged particle beams bombarding metal
References targets to produce n beams. Targets are sur-

rounded by a moderator-reflector layout matched
[1] V.V. Parkhomchuk, A.N. Skrinsky, Proc. 12th to produce neutrons in the desired energy range
Int. Conf. on High Energy Acc. (1983) 485; A.N. (cold, thermal, epithermal). Beam channels bring
Skrinsky, V.V. Parkhomchuk, Sov. J. of Nucl. moderated neutrons to the experimental stations
Phys. 12 (1981) 3 surrounding the shielded target station.
[2] N. Holtkamp, D. Finley. eds., Fermilab-Pub- Because of the weak interaction of neutrons
00/108E (2000) with matter and the engineering heat transfer
[3] R.B. Palmer, Muon Collider, Proc. AHIPA09
limit on n production methods, n-scattering is
[4] D. Neuffer, MUC-NOTE-269 (2003)
an intensity limited field. In contrast to reactor
[5] Y. Alexahin, http://www.cap.bnl.gov/mumu
sources, which are essentially c.w., accelerator-
/conf/collider-091201/talks/YAlexahin-1-
091201.ppt
driven sources have the advantage that they can
[6] J.S. Berg et al, PRST-AB 9 (2006) 011001 be pulsed. High peak flux is of interest for most
[7] R.B. Palmer et al, PRST-AB 8 (2005) 061003; n-scattering types of measurements. Accordingly,
Pavel Snopok, there are efforts in:
http://www.cap.bnl.gov/mumu/conf/ MUTAC- • Increasing the power of the incident beam,
090406/talks/PSnopok1-090406.pdf
realized by increasing the beam energy,
[8] Y. Derbenev, R. P. Johnson, PRST-AB 8,
beam intensity, and using a pulsed time
041002(2005)
structure.
[9] C. Ankenbrandt,
http://indico.fnal.gov/getFile.py/access? • Increasing the n target yield. Neutron yield
contribId=65 &ses- increases significantly with the incident-
sionId=28&resId=0&materialId=slides particle energy, and is a strong function of
&confId=2854 the target material and incident beam type.
[10] D.J. Summers et al, PAC 07, THPMS082
[11] Y. Alexahin, E. Gianfelice-Wendt, • Increasing the n transmission efficiency.
http://www.cap.bnl.gov/mumu/conf/collider The target-moderator-reflector design is op-
-091201/talks/YAlexahin-2-091201.ppt timized; guides with high-reflectivity, some-
[12] http://proj-hiptarget.web.cern.ch/proj-hiptarget/ times with focusing properties, are used.
[13] MICE,http://mice.iit.edu/
[14] A. Moretti et al, PRST-AB 8, 072001 (2005) According to the incident beam’s time struc-
[15] A. Bross, http://www.fnal.gov/projects/muon ture, there are three types of accelerator-driven
collider sources (Tab.2)[6, 7, 8]:
/FridayMeetings/ (2009)
[16] P. Hanlet et al, EPAC 2006, TUPCH147 • C.w. sources driven by high-intensity elec-
[17] K. Yonehara, et al, MOPPC036, Proc. IPAC12, trostatic accelerators or cyclotrons. Exam-
New Orleans,LA May 21-25, 2012 ple: the isochronous-cyclotron of PSI, which
[18] R.B. Palmer et al, PRST-AB 12, 031002 (2009) delivers a c.w. proton beam of 1.3 MW at
[19] S.M. George, A.W. Ott, J.W. Klaus, J. Phys. 590 MeV [9].
Chem. 100, 13121 (1996)
[20] R. Weggel et al, NFMCC-DOC-553 • Sources with long (ms) pulses driven by
high-intensity linacs. Examples: the 800
47
–
MeV LAMPF p-linac at Los Alamos [5], the H
13 MeV LENS p-linac at Indiana University, (a)

6 9
1 2 4 5
and the proposed 2.5 GeV ESS p-linac in 3
7
Lund.
p
i(t) i(t) 8
• Sources with short (μs) pulses driven by a
combination of high-intensity linacs and (b)
rings. Example: the J-PARC 3 GeV pro- t t
1 ms 1 μs
ton rapid-cycling-synchrotron (RCS) sup- 10–97
plied by the 181 MeV H− linac and the SNS

20 ms 20 ms 8355A36
1 GeV proton accumulator ring (AR) sup- Figure 1: (a) Schematic of a short-pulsed neutron
plied by the 1 GeV H− linac.
source. 1) H− source, 2) RFQ, 3) chopper, 4) DTL or

CCDTL, 5) nc or sc high energy linac, 6) AR or RCS
A dominant design criterion is ultra low beam with H− -H+ charge exchange injection, 7) beam trans-
losses for avoiding component, air, and tunnel ac- port to target, 8) target with moderators and reflectors,
tivation and for allowing maintenance and repair 9) neutron channels. (b) Time structure of beams at
on short notice. A typical guideline is to limit linac and at ring.
the average uncontrolled beam loss to be below
1 W/m. Linac Linacs for long-pulse sources accelerate
Compact neutron sources In fusion-based (D- high-intensity protons; up to 200 mA peak current
D, D-T, T-T) n generators, beams of currents up is at the limit of present-day technology. Linacs
to 350 mA are accelerated with a voltage typ- for short-pulse sources accelerate high-intensity,
ically from 60 to 120 kV. The n yield covers low-emittance H− for low-loss charge-exchange
a wide range from 106 to 1011 n/s. Compact injection into the ring; up to 100 mA peak cur-
sources based on electron and proton linacs pro- rent and down to ∼0.1 π mm-mrad normalized
duce yields up to 1014 n/s (Tab.1). rms emittance are at the limit of present-day tech-
Spallation neutron sources In spallation nology. Duty cycles of up to 100% has been
sources [5], neutrons are produced by the interac- demonstrated for RFQ and superconducting rf
tion of high energy (GeV) protons with a target. linac (SCL).
The number of n produced is proportional to the A linac starts with a proton or H− source
proton energy in range 0.2-10 GeV; in a typical (Sec.7.1.4) followed by one or more sections of
lead target a 1 GeV proton produces 20 neutrons. SC and/or non-SC accelerating or transport struc-
Pulsed sources allow time-of-flight correla- tures. The 1-GeV SNS linac consists of H−
tion of the incident neutron energy. This avoids source, LEBT, RFQ (Sec.1.6.18), MEBT, DTL
the monochromatization needed in c.w. sources (Sec.7.3.5), and SCLs of two different β’s. The
which greatly reduces the useful n-flux. Time 400-MeV J-PARC linac consists of H− source,
of flight measurements require small repetition LEBT, RFQ, MEBT, DTL, SDTL, and ACS.
rate (≤ 60 Hz), in order to avoid the overlap of SCL is preferred to non-SC options espe-
slow n from one pulse with fast n from the next. cially for duty cycles higher than 5%, and for re-
Long-pulse (ms) facilities have the advantage of tuning and operational robustness. For a pulsed
a simpler driver layout (i.e., a pulsed linac with p beam with its velocity changing during accel-
high proton intensity), coupled target-moderator eration, effects of Lorentz detuning, microphon-
configuration, and potentially higher integrated ics, beam transients and injecting-energy offsets
n-flux. Short-pulse (μs) facilities have the ad- require careful rf amplitude and phase control [2].
vantage of much higher peak n-flux. However, The SNS’s SCL (6% duty) starts at 186 MeV with
they require an H− linac combined with an AR each cavity driven by its own klystron and rf con-
or RCS filled by multiturn injection and emptied trol.
by fast one-turn extraction (Fig.1). Existing and To keep average linac losses low (∼1 W/m)
planned spallation sources use a proton energy emittance growth and halo generation have to be
range 0.5-5 GeV. Different options for linac and carefully controlled at all stages of the linac, in
ring energies have been used or proposed (Tab.2) particular at structural and rf frequency transi-
[5, 6, 8], depending on the goal beam power. tions. Low loss ring injection requires new ways
48
Ch.1: INTRODUCTION
Table 1: A few existing and planned compact sources.
Name Status Accelerator Target Ave. Rep. Pulse Ave.
type & type beam rate length at n-yield
energy power (Hz) target (1013 )
(kW) (ms) (n/s)
Hokkaido operat. 45MeVe-linac W+Pb 1 50−100 10−8 −3×10−3 0.16
LENS,Indiana operat. 13MeVp-linac Be 13 20 2 4
CPHS,Tsinghua constr. 13MeVp-linac Be 16 50 0.5 5
Table 2: A few existing and planned spallation sources.

Name Status Accelerator Target Average Rep. Protons Pulse

type & type beam rate per length
energy power (Hz) pulse at tar-
(MW) (1013 ) get(μs)
IPNS,ANL [5] 1981− 50 MeV linac U 0.0075 30 0.3 0.1
2008 500 MeV RCS
ISIS,RAL [10] operat. 70 MeV linac Ta, W, 0.16 50 2.5 0.45
1985− 800 MeV RCS U
SINQ,PSI [9] operat. 590MeVcyclotron Zr, 1 cw - -
1996− Pb
LANSCE 1985−PSR 800 MeV linac W 0.08 20 3 0.27
LANL [5] 1975−linac AR
SNS, ORNL operat. 1GeVlinac(nc/sc) Hg 1 60 10 0.7
[11, 12] AR
J-PARC operat. 181(400)MeVlinac Hg 0.3 (1) 25 8.3 1
[13] Tokai 3 GeV RCS
CSNS constr. 80 MeV linac W 0.1 25 1.6 0.8
[16] Dongguan 1.6 GeV RCS
ESS planned 2.5 GeV linac 5 20 62 2000
Lund
of halo containment, diagnostics, collimation and At ultra-high energies (e.g. 8-GeV linac proposed
beam cleaning. Extensive space-charge analy- by Fermilab), black-body stripping further limits
sis of resonance and equipartition conditions and the maximum field strength of the magnets [3].
comprehensive Monte-Carlo simulations (includ- Rings [19] The SNS AR of 248-m circumfer-
ing nonlinear space charge forces) have to be ap- ence accumulates 1 MW p-beam (2.9×10−6 duty,
plied with up to 107 macro-particles for a realistic 35 A peak current). The J-PARC RCS presently
layout between H− -source and ring injection [17]. accelerates 300 kW p-beam from 181 MeV to
Intrabeam stripping (Sec.2.4.12) is identified to be 3 GeV. The average uncontrolled beam loss is
a loss mechanism for H− beams and studied at the ∼1 W/m limiting the performance. At very high
SNS linac [18]. beam power (e.g. >5 MW) the use of multiple
The ultra low injection losses also require rings may become mandatory.
the transport (HEBT) between linac and rings to Upon charge-exchange injection using a strip-
have a precise control of energy (energy-deviation ping foil, elaborate painting with correlated 4-D
correction by a rotator cavity and energy-width or 6-D phase space fills the ring acceptance as uni-
spreading by ramping or a spreader cavity) and formly as possible. Large ring acceptance (typi-
a removal of halo particles by betatron and mo- cally >400 π mm-mrad) and good magnetic field
mentum scraping. At high injection energies, low quality (∼10−4 deviation level) are needed for
magnetic fields have to be used to avoid magnetic keeping effects of space charge and magnetic res-
stripping (Lorentz stripping, Sec.7.1.8) of H− s. onances at a tolerable level. Stripping foils have
49
an efficiency of about 98-99% so that a high in- supply neutrons to a large number of experiments
tensity of partially stripped H0 particles has to be (∼1000-2000/yr). High availability, high reliabil-
handled in the injection region. This low emit- ity and short down-times are essential.
tance beam presents an interesting source of high Neutron targets [5] For spallation sources,
energy protons which can be used e.g. for the pro- water-cooled solid (Ta, W, Pb, depleted U) and
duction of radioactive beams or muon beams. liquid (Hg, Pb-Bi) heavy metal targets with hor-
The ring lattice is generally based on a high izontal or vertical injection are in use up to MW
periodicity and transition energy (Sec.4.9 should range. Rotating solid targets are studied for multi-
not be passed. Long straight sections are used for MW class sources.
rf, injection, extraction and scraping systems. The Targets have to contain the nuclear cascade
efficiency of rf trapping can be increased by using produced by protons and withstand high radiation
a low frequency, dual-harmonic rf-system and by damage. For MW targets this can become com-
chopping the injected H− -beam (at the linac front parable to the range expected in Tokamak fusion
end) at the ring revolution frequency. reactors (> 100 d.p.a. – displacements per atom –
Collective effects and instabilities are most produced by p and n knock-on, and 10,000 appm
relevant; low impedance beam lines are essen- – atomic parts per million – of He produced by
tial. For RCSs with ceramic vacuum chambers transmutation).
either internal rf-cages or external metal stripes For short pulses an additional load stems from
are used [10, 13]. The coupling impedance of ex- shock/stress waves produced by the high energy
traction kickers and the pulse-forming network is content (up to 100 kJ) of proton pulses. This com-
minimized [21, 8]. Effects of any beam-induced bined load and the absence of corrosion and tri-
electron cloud are mitigated by vacuum-chamber tium production linked to cooling water circuits
surface coating, solenoidal magnetic fields, and are the main reasons for developing liquid tar-
clearing electrodes to reduce the secondary emis- gets. Liquid targets in pulsed operations suffer
sion yield [22, 23, 8] (Sec.2.4.14). At ISIS, the cavitation-induced pitting damage to the surfaces
tunes are adjusted to (i) compensate for the nat- of target vessels. Stainless steel specially treated
ural chromaticity and the varying magnet field at and hardened by the kolsterizing process is in use.
injection, (ii) ramp up tunes to minimize effects For compact sources driven by lower energy
of space-charge depressions during beam capture, proton beams, water-cooled Li and Be targets
(iii) reduce tunes during the time from 2 to 4 ms are in use, and Ga-liquid-metal cooling has been
after injection to avoid transverse resistive wall in- proven out [26]. The short stopping range of low-
stability, and (iv) lower tunes to avoid coupling energy p beams causes complications.
resonances at extraction. Other applications The accelerator and target
Fast kicker magnets (< 200ns) (Sec.7.2.7) are technology necessary for spallation sources in the
needed to avoid beam losses at extraction. Par- MW-range has requirements which are in many
ticles are brought to the target station with large respects similar to the ones needed for future high
acceptance transfer lines which may contain mul- power proton accelerators envisaged for neutrino
tipole elements to flatten the beam density profile factories, muon storage rings (Sec.1.6.15), trans-
at the target. mutation of nuclear waste, accelerator production
Momentum scraping is essential for RCSs to of tritium, accelerator driven subcritical nuclear
control the beam loss during rf trapping and ramp- power generation, proton radiography, white neu-
ing. Betatron scraping efficiency is improved by tron source applications, and high intensity ra-
a two-stage system with acceptances significantly dioactive beams. Its use for irradiation facilities
smaller than the rest of the ring, but larger than the is also being considered.
beam core [24]. Beam-in-gap cleaning reduces
the beam loss at extraction for ARs [25, 24]. References
Adequate shielding, beam dumps and fast [1] J.M. Carpenter, Y.B. Yelon, Methods of Exper-
beam loss monitoring acting bunch-to-bunch are imental Physics, Ch.2, Vol.23, p.99 (Academic
essential. Radiation resistant materials are used in Press)
areas of high radio-activation. Remote handling [2] J. Wei et al, PAC 2001, 319
devices are used in areas demanding frequent [3] H.C. Bryant, G.H. Herling, J. Mod. Optics 53
maintenance. Large spallation sources have to (2006) 45
50
Ch.1: INTRODUCTION
[4] T.A. Gabriel, J.R. Haines, T.J. McManamy, output voltage, sub-1 Ω impedances, and power
J. Nucl. Materials 318 (2003) 1 levels of tens of TW. Applications of this technol-
[5] Proc. Int. Collaboration on Advances Spallation- ogy include x-ray generation and inertial confine-
Sources (ICANS); ICANS XII, Abingdon, UK ment fusion.
(1993), RAL Proc.94-025, ICANS XIII (1995)
PSI Proc. 95-02
Marx generators/Pulse lines The typical Marx
[6] G.H. Rees, PAC 93, p.731 Generator uses plus/minus charged columns of
[7] P. Bryant, PAC 95, p.322 capacitors that are charged in parallel and dis-
[8] J. Wei, RMP 75 (2003) 1383 charged in series, through triggered spark gaps,
[9] G.S. Bauer et al, PAC 97 into a transmission line. The charging and dis-
[10] D.J. Adams et al, EPAC 00, 975 charging of the transmission line occurs in ∼1 μs
[11] J. Wei et al, PRST-AB, 3 (2000) 080101 and ≤ 0.1 μs respectively. The transmission lines
[12] N. Holtkamp, Linac 04, p.837 use deionized water (for low impedance) or trans-
[13] JAERI/KEK Report JAERI-Tech 2003-044, former oil (for intermediate-to-high impedance)
KEK Report 2002-13 (2003) as the insulating material. Breakdown strengths
[14] ESS Study Final Report, Vol.III, ESS-96-53M of the dielectrics are ∼100 and 300 kV/cm re-
(1996) spectively for pulse durations ∼1 μs and increase
[15] H. Lengeler, NIM B139 (1998) 82 slowly (∝ t−1/3 for sub-μs pulses) with decreas-
[16] IHEP Report IHEP-CSNS-Report/2004-01E ing pulse durations. The transmission line is usu-
(2004) ally connected to the load by an overvolted gas
[17] K. Bongerdt, M. Pabstand, A. Letchford, NIM or water spark gap. Modest changes in the pulse
A451/1 (2000) 287 duration and the generator output impedance are
[18] V. Lebedev et al, Linac 2010, p.929 achieved through the use of tapered transmission
[19] G.H. Rees, EPAC 94, p.241 lines connecting the pulse line to the load. In
[20] R. Macek et al, PAC 01, p.688 sub-1 Ω impedance generators the basic Marx
[21] D. Davino et al, EPAC 02, p.1467 generator-pulse line configuration may be re-
[22] P. He et al, EPAC 04, p.1804
peated many times (e.g., the Z-machine at Sandia
[23] L. Wang et al, PRE 036501 (2004)
uses 36 modules in parallel). Very short duration
[24] N. Catalan-Lasheras et al, PRST-AB, 4, 010101
high power pulses require low inductance power
(2001)
[25] R.L. Witkover et al, PAC 99, p.2250
feeds, which is accomplished by the the use of
[26] B.W. Blackburn, MIT Thesis (2002) water-dielectric transfer capacitors as an interme-
diate low inductance circuit element between the
Marx generator and the transmission line. The lat-
1.6.17 Pulsed High Voltage Devices ter can be charged with a faster rise time from
J.A. Nation, D.A. Hammer, Cornell U. the water capacitor than directly from the Marx
The engineering science of Pulsed High Voltage generator, leading to a more compact system as
devices [1, 2] was first developed by J.C. Martin well as to the formation of multiple channel dis-
of AWRE (Atomic Weapons Research Establish- charges, and hence low series impedance, con-
ment, Aldermaston, UK). He used a Marx gener- necting the transmission line to the load. In ul-
ator to impulse charge a solid dielectric, or oil- tra high current devices the load is connected
filled, or water dielectric transmission line as a to the transmission line by a Magnetically In-
lumped parameter capacitor. The discharge of the sulated Transmission Line (MITL), an evacuated
transmission line, now as a distributed line, into transmission line in which the wave electric field
a vacuum diode reduced the pulse duration, com- causes electron emission from the negative line
pared to the charging time of ∼1 μs, by an or- conductor. However, the current in the MITL pro-
der of magnitude and hence increased the avail- duces a sufficiently large transverse magnetic field
able power correspondingly. Electron beams were between the line conductors to cause the emitted
produced by field emission cathodes. Simple sys- electrons to flow parallel to the electrodes instead
tems of this type produce output pulses of 1-10 of across the electrode gap.
MV with pulse durations of order 20-100 ns, at Transmission line loads The load depends on
impedances of a few to 50 Ω, depending on the the application. Three commonly used loads are:
dielectric used in the pulse line. More recent (i) Vacuum diodes with field emission cathodes
extensions of the technology produced ∼30 MV that produce electron beams; (ii) Diodes with
51
plasma electrodes that are used to produce elec- high power, as has been done in the design for the
tron or ion beams; and (iii) Z pinches. next larger Z-pinch driver by Stygar et al. [4].
The above arrangements are commonly used Voltage and current fluctuations For many
for hard x-ray production, high power microwave pulsed power applications, square voltage and
generation, inertial confinement fusion research current waveforms are of secondary importance
and for the generation of intense soft x-ray pulses. compared to the peak power output. This is es-
Recent experiments at Sandia have reported the pecially true in multi-TW devices where the low
production of 2 MJ x-ray pulses with peak pow-
impedance of the generator leads to a relatively

ers of up to 280 TW. slow increase in the load current. For example,
Diodes naturally generate electron beam cur- the Z-machine delivers a ∼5 MV, 26 MA, 100 ns
rents when a suitable polarity high voltage pulse current rise-time pulse to a Z-pinch. The voltage
is applied across the diode. Ion beams can also be fluctuations can be ∼ 50% and the current rise
generated if there is a suitable ion source, such time is comparable to the pulse half width. The
as a plasma, on the anode, but the ion genera- device is well matched, however, to the dynam-
tion efficiency is low unless the electron current ics of the imploding Z-pinch. The fluctuations in
is suppressed. This is accomplished by applying the output of higher impedance devices are, how-
a transverse magnetic field in the diode such that ever, much lower and a degree of tuning is pos-
the electron excursion from the cathode is less sible, especially in the beam production mode of
than the anode cathode gap spacing. Proton and operation.
other low atomic number ion beams have been
Pulse transformers Many of the above devices
produced this way.
are not well suited to high repetition rate op-
Inductive addition Induction accelerators are eration, except in burst modes. For applica-
discussed in Sec.1.6.7. High voltage pulses for tions requiring more modest beam currents, (e.g.
radiography or for use in electron beam injectors klystrons), pulse transformers offer the preferred
are frequently produced by the use of a single cen- modulator configuration. Thyratron switching
tral cathode conductor as the secondary of several permits high repetition rate use more readily than
induction modules. The voltage of the cathode that achievable with pressurized gas switches.
is then equal to the sum of the secondary volt- Pulse durations are typically ∼1 μs and output
age outputs of each of the modules; i.e. the con- voltages in the range 300-500 kV. The achievable
ductor adds the voltages from the modules. For rise time is strongly affected by the transformer
hard x-ray production, which scales as I·V 2.8 for step-up ratio and the core material selection. Rise
a given high-Z target material, the high voltage times of 100-200 ns are achievable with step-up
output is essential for efficient use of the acceler- ratio’s of <
∼ 8.
ator power. The Hermes III accelerator at Sandia,
e.g., uses 20 1-MV induction modules to produce References
a 20 MeV electron beam in a single diode. For
electron beam production the increased injector [1] J.C. Martin, Pulsed Power, Advances in Pulse
energy allows larger space charge limited beam Power Vol.3, Plenum Press (1996)
currents, and reduced beam divergence. [2] J.A. Nation, PA 10 (1970) 1
[3] A.A. Kim et al, PRST-AB 12, 050402 (2009)
Linear Transformer Driver (LTD) The fun-
[4] William Stygar et al, PRST-AB 10, 030401 (2007)
damental unit of LTD technology is a capacitor
switched into a low inductance circuit with soft
iron core isolation. In order to achieve a very 1.6.18 Radio Frequency Quadrupole
high current in a low inductance load, such as a J. Staples, LBNL
Z-pinch, many such modules are switched in par-
allel in a circular array using a large, circular iron A Radio Frequency Quadrupole (RFQ) is a com-
core to isolate all of them, with the load in the pact and versatile accelerator operating over a
center. Several 1 MA pulsers with rise times of mass range of protons to low charge state heavy
150-300 ns have been built using this approach ions, accelerating from a few keV/n for heavy
[3]. Such circular arrays can then be stacked in ions with a total voltage integral of up to several
series in order to drive a higher impedance load, MV. The operating frequency ranges from 6 MHz
or in series-parallel arrangements to reach ultra- (for Bi+2 ) to over 400 MHz (for protons). Duty
52
Ch.1: INTRODUCTION
factors range from 0.1% to 100%, and transmit- cupying more than half of the physical length of
ted current to greater than 200 mA of protons. the RFQ, followed by an accelerating section.
Beam dynamics An RFQ comprises an The d.c. input beam is matched at the RFQ en-
alternating-gradient time-varying electric trance to the time-varying transverse beam pro-
quadrupole strong focusing transport chan- file by the radial matcher, typically 4-8 cells, fol-
nel with an accelerating field Ez (z) added as a lowed by a short shaper section that initiates the
perturbation by modulating the vanetip profile. formation of the bunch. The helper code RFQUIK
assembles a cell table for use in PARMTEQM. To
The Ez (z) profile may be chosen arbitrarily,

giving wide design freedom of the capture reduce the truncation error of the two-term field
and acceleration sections. The alternating- expansion, PARI adds higher-order terms and ad-
gradient electric-quadrupole field provides justs the cell parameters to maintain the accelera-
velocity-independent focusing and will transport tion of the reference particle. PARI uses a look-up
unaccelerated or partially accelerated beam to the table for a limited number of vanetip profiles.
exit. The TOUTATIS code, requiring signifi-
As a Sloan-Lawrence accelerator, the shunt cantly more computational resources, solves the
impedance drops off as β −2 in the accelerating boundary-value problem for the actual vanetip ge-
section, restricting output beam energies to less ometry in each cell, allowing arbitrary cell shapes
than 2-3 MeV for protons for efficient designs. and gaps in the vane, and a PIC formulation of
Ez (z) is controlled by the vane modulation the space charge forces. TOUTATIS is based on
parameter m ≥ 1, where the displacement of the PARMTEQM, but removes the paraxial approx-
vanetip from the axis varies from a to ma along imation, adds field maps, and treats transverse
the length of the cell. Ez (z) scales approximately beam loss on the physical location of the vane
as (m−1)2/3 . The upper limit of m is constrained boundary. The LIDOS and RFQTRAK codes also
by the minimum longitudinal radius which sets include field maps for a more accurate evaluation
the size of the tool used to cut the vane profile, of the beam characteristics. Jameson [5] presents
as well as the magnitude of the multipole com- a lengthy summary comparing the above codes,
ponents of the fields. As m and thereby Ez in- used during the design of the IFMIF c.w. RFQ.
crease, the transverse focusing strength and the The KT formalism is suitable for the design
clear beam aperture decrease, limiting geometric of high-current RFQs, but greater bunching effi-
acceptance. ciency, lower emittance growth and shorter struc-
The transverse phase space acceptance of an tures may be generated using other design ap-
RFQ increases rapidly with the design field gra- proaches. None of these approaches is straight-
dient. The peak surface field on the vanetip is forward, but rather ad-hoc, without a specific de-
Es = κV /r0 , where κ is the field enhancement sign recipe. The Hofmann diagram [6], which il-
factor. V is the peak rf voltage between vanetips, lustrates bunch resonances as a function of beta-
and r0 is the average vanetip displacement from tron phase advance and tune depression, is of use
the beam axis. Depending on the detailed vanetip in determining an optimum parameter space for
geometry, κ ≈ 1.25 − 1.55. The sustainable field high-current, low emittance-growth designs.
is expressed in units of kilpatrick, found by solv- Space charge The KT approach used in
ing the implicit equation f = 1.643Es2 e−(8.5/Es ) PARMTEQM uses the helper code CURLI to op-
for Es , the surface field in MV/m, f is the fre- timize the beam dynamics design at the end of the
quency in MHz. One kilpatrick at 200 MHz is gentle buncher section where the bunch is formed,
14.7 MV/m, and 19.4 MV/m at 400 MHz. Short- the charge density is high and the energy is still
pulse RFQs may be safely pushed to greater than low, so the tune depression is at a maximum.
2.0 kilpatrick, with c.w. RFQs held to less than The transverse and longitudinal current limits It
1.8 kilpatrick. and I are defined as the current that depresses
Beam dynamics codes RFQ design and simula- the tune by typically 60%. The transverse limit
tion codes include PARMTEQM [1], TOUTATIS It scales as βλ2 V 2 φx q/(a2 A) and the longitudi-
[2], LIDOS [3] and RFQtrak [4]. PARMTEQM nal I as V φ2s a/λ, with a the minimum vane tip
uses the Kapchinskii-Teplyakov (KT) design pro- radius, V the vane-to-vane peak voltage, φs the
cedure where the beam is bunched adiabatically stable phase, and q/A the charge-to-mass of the
with a long gentle buncher section, typically oc- ion, indicating that a low-frequency machine is
53
preferred for a large transverse current limit, but a dipole modes higher in frequency and away from
high-frequency machine would have a higher lon- the quadrupole mode.
gitudinal current limit. Maintaining an adequate Rf structure codes The 4-vane RFQ operates at
focusing phase advance for mass greater than pro- the waveguide cutoff frequency, which may be es-
ton requires a longer operating wavelength λ, as timated with 2-D codes such as SUPERFISH [1]
the vane voltage V and aperture a are already at a or URMEL. However, the presence of mode sta-
practical limit. The operating frequency of heavy- bilizers and vane end cutbacks require a full 3-D
ion RFQs may be in the 50-200 MHz range, or calculation for accurate estimate of the resonant
even lower. frequency, field distribution and wall power den-

Rf structures All RFQs have in common a sity.
time-varying electric quadrupole focusing field on Advances in computational capabilities have
axis. The choice of rf structure will depend on the resulted in the development of powerful finite-
operating frequency, length, and power efficiency element and finite-difference codes that permit the
required. characteristics of complex rf structures to be de-
The 4-vane structure excites a waveguide in termined. RFQs, particularly ones that employ
the TE210 mode, producing a quadrupole E-field, various types of stabilizers and tuners, require a
adding vanes to concentrate the electric field near large range of mesh density to include small de-
the axis. A typical frequency range is 100 to 425 tails over a large cavity volume.
MHz. A major drawback of the 4-vane structure Example codes include CST-Microwave Stu-
is the presence of almost degenerate dipole TE110 dio [9], HFSS [10], and Omega3P [11], which
modes which may mix with the TE210 quadrupole have been used to model RFQ mode structure and
mode. Several methods have been used to in- surface power density. CAD modeling codes such
crease the dipole-quadrupole frequency separa- as ANSYS [12], which also include 3-D elec-
tion. Opposite vanes may be strapped together tromagnetic solvers as an additional module are
with vane coupling rings (VCRs) in low duty- used to model the time-dependent thermal and
factor structures. or π-mode stabilizers for high thermally-induced stresses in the structure.
duty-factor RFQs, used at JPARC and SNS. The Tuning The 4-vane RFQ is a standing-wave
LEDA RFQ uses resonant coupling between lon- structure where the deviation of local field
gitudinal RFQ regions to effectively shorten each dE0 (z)/E0 as a function of the local detuning
section, reducing the field perturbations due to lo- δf0 (z)/faverage is given by the solution of
cal frequency error.
∂ 2 δE0 (z) 8π 2 δf0 (z)
The 4-rod structure avoids the TE110 mix- = (1)
∂z 2 E0 λ2 faverage
ing problem by supporting opposing vane pairs
on quarter-wave stubs, shortened due to the addi- where δf0 (z) is the localfrequency variation due
tional capacitive loading of the vanes themselves. to mechanical errors and (δf0 ) = 0. Field errors
The electric field energy is almost entirely within scale as the square of the length of the RFQ. This
the vane region, and the magnetic field energy restricts the practical length of an RFQ to less than
mainly around the stubs. The field profile may about 5 free-space wavelengths above which me-
be modulated by the periodic structure of the sup- chanical tolerances become severe.
port stubs, and there may be a non-zero poten- The field distribution is usually determined
tial between the ends of the vanes and the end- by pulling a metallic or dielectric bead through
walls, affecting bunching. Both of these potential the RFQ and noting the change in resonant fre-
problems are manageable. This structure is more quency. The metallic bead removes both E- and
compact than the 4-vane structure and has been H-field, the dielectric bead only E-field stored en-
frequently applied to lower-frequency heavy-ion ergy. Sensing loops may be placed along the RFQ
accelerators. and calibrated by the bead pull to measure the
The split-coaxial structure has been promoted fields during operation.
for even lower frequency accelerators for low- Tuners are introduced along the outer walls
charge-state heavy ions and for superconducting of 4-vane RFQs usually in the form of pistons
RFQ structures. Here, the 4-vane structure is which when moved inward, increase the local res-
modified with large cut-outs in the vane base that onant frequency by removing H-field energy. A
modify the mode structure, moving the TE110 program that specifies tuner settings by solving
54
Ch.1: INTRODUCTION
Eq.(1), based on bead pull measurements, eases [8] H. Vernon. Smith et al, PAC 2001, p.3297
tuner adjustment. [9] www.cst.de
The field distribution in 4-rod RFQs concen- [10] www.ansoft.com
trates the magnetic field energy in the volume [11] K. Ko, Proc. ACES 2002 Conf, 2002
around the support studs. Tuners may be in the [12] www.ansys.com
form of capacitive plates near the rods, lowering [13] A. Pisent et al, Linac 2004, p.69
the local frequency, or in metallic blocks in the [14] A. Pisent et al, EPAC 2008, p.3542
vicinity of the support studs, raising the local fre- [15] A. Pisent, Linac 2010, p.372
quency. [16] S. Virostek et al, Linac 2000, p.908

Rf power couplers The rf may be introduced
through multiple or single loop or iris couplers,
1.6.19 Rhodotron
which couple with the rf H-field, maximum at the

Y. Jongen, M. Abs, IBA
wall of 4-vane RFQs. Loop couplers are easily
adjustable, but are limited in the peak rf power The Rhodotron, invented by Jacques Pottier,
that can be introduced in each coupler to the few CEA, 1987, is a novel accelerator principle suit-
hundred kW range. Multiple couplers driven from able for electron acceleration up to 10-12 MeV.
the same rf source should be isolated from each This new accelerator principle was the answer to
other to ease balancing the power flow through the needs of high power electron beams of more
each. than 2-3 MeV for industrial applications or for in-
Higher average power may require the use of tense x-ray production. Energies below 3 MeV
iris couplers. The high power density at the edge are still well served by electrostatic accelerators
of the iris will require special attention. (Sec.1.6.5). At more than 3-5 MeV industrial
The measured Q-value of typical 4-vane linac’s have been used so far but with a relatively
structures runs from 50 to 80% of the theoretical poor electrical efficiency and limited beam con-
value calculated by the electromagnetics codes. trol and intensity.
High average power RFQs RFQs designed for Ion Beam Applications, S.A. (Belgium) de-
c.w. operation include LEDA [8], TRASCO [13], veloped a range of 5 to 10 MeV Rhodotrons
IFMIF [14], Project-X(PXIE) [7] and many oth- with average beam power ratings from 25 kW to
ers. Pisent [15] describes several specific de- 700 kW at working frequencies of 107 MHz and
signs. Designs tend to lower frequency to reduce 215 MHz. Thirty machines have been sold so far.
the wall power density to keep it below the 15 Beam acceleration The Rhodotron is a recircu-
W/cm2 range with larger aperture to reduce beam lating rf accelerator based on a unique beam pat-
loss in the structure. High duty-factor RFQs with tern that looks like a flower. See Figs.1 and 2.
high average wall power densities (in excess of The beam crosses a number of times at different
1 W/cm2 ) may change shape due to thermal ex- azimuths a half wave coaxial cavity in its median
pansion, resulting in a redistribution of the field plane. Machines performing up to 12 crossings
profile. At 400 MHz, an isothermal copper cavity have been manufactured. This seems to be a prac-
shifts 7 kHz per degree centigrade. Active tuners tical limit. Each time the beam crosses the cavity
may be required, or in the case of LEDA and it gains up to 1 MeV. The beam is then reinjected
the SNS RFQs, the vanes and the RFQ body are by external DC magnets towards the center of the
supplied with separate coolant temperatures [16] cavity. The magnets’ position and field are chosen
to differentially control the vane length from the in such that the beam is re-injected with the right
body dimensions to hold the frequency constant. phase with respect to the rf field. The first magnet
is usually of a different size as the beam is not yet
References relativistic.
[1] http://laacg1.lanl.gov The big accelerating cavity has a fairly high
[2] R. Duperrier et al, XX Int’l Linac Conf, p.839 Q factor (50, 000 @ 107.5 MHz) and so has a
[3] B. Bondarev et al, PAC2001, p.2947 relatively low power consumption thus enabling
[4] L. Young, PAC 1997, p.2752 c.w. operation, even if pulse operation is possi-
[5] R.A. Jameson, ORNL/TM-2007/001 ble. The central conductor of the cavity has two
[6] I. Hofmann, PAC 1981, p.2399 cones at the extremities thereby improving the Q
[7] D. Li et al, IPAC 2012 and avoiding HOMs to be excited by the beam
55
fact that the magnets lengthen the beam trajecto-

ries and shift the phase when the energy increases.
The electromagnetic focusing forces due to
the rf field have limited effects except on the first
crossing. The vertical and horizontal focusing is
ensured mainly by the magnets. No other ele-
ments are needed to control the beam size, making
the machine simple and robust.
The extraction of the beam is obvious and al-

lows placing different exits at different energies
by selecting to switch off certain magnets.
Figure 1: Horizontal section of the cavity showing the
beam path and magnet positions. Machine properties summary

• Energy range: up to 10-12 MeV per accel-
erator (more cavities can be put in series to
increase energy)
• Energy spread: < 300 keV at 10 MeV
• Beam current control < 0.1% precision in
less than 500 μs
• Electrical efficiency: up to 55% at full beam
power
Figure 2: Vertical section showing the tapered central 1.6.20 Storage Rings
conductor. W. Fischer, BNL
Storage rings are circular machines that store par-
harmonics. The cavity that is excited by an exter- ticle beams at a constant energy. Beams are stored
nal power amplifier chain based on tetrodes does in rings without acceleration for a number of rea-
not require a tuning element. The small reso- sons (Tab.1). Storage rings are used in high-
nance frequency drifts due to thermal changes in energy, nuclear, atomic, and molecular physics, as
the cavity dimensions are followed by the rf gen- well as for experiments in chemistry, material and
erator based on phase measurements on the final life sciences. Parameters for storage rings such
amplifier. An amplitude regulation is required to as particle species, energy, beam intensity, beam
maintain the accelerating field stable at better than size, and store time vary widely depending on the
1% as the beam power is changed. application. The beam must be injected into a
storage ring but may not be extracted (Fig.1). Ac-
Beam injection The injection of the beam into celerator rings such as synchrotrons (Sec.1.6.21)
the cavity is done with an external electron gun. are used as storage rings before and after acceler-
The injection energy is 30-60 keV. The electron ation.
train must be pulsed at working frequency to al-
Particles stored in rings include electrons and
low the beam transmission ∼100%. The typ-
positrons; muons; protons and antiprotons; neu-
ical phase acceptance is large and around 60◦ .
trons; light and heavy, positive and negative
This means that the injected peak current is 8
atomic ions of various charge states; molecular
to 10 times the average current. Peak current
and cluster ions [1], and neutral polar molecules.
up to 1 A, with average currents up to 100 mA,
Spin polarized beams of electrons, positrons, and
has been successfully injected in the most pow-
protons were stored. The kinetic energy of the
erful model. The IBA e-guns are based on com-
stored particles ranges from 10−6 eV [2] to 4.0 ×
mercial cathode-grid assemblies that allow good
1012 eV (LHC, 7 × 1012 eV planned), the num-
beam control in time and amplitude. A beam cur-
ber of stored particles from one (ESR [4]) to 1015
rent control accuracy of 0.1% is easily achieved.
(ISR [5]).
Beam focalization Unlike linacs, the phase sta- To store beam in rings requires bending
bility is of synchrotron type. This is due to the (dipoles) and transverse focusing (quadrupoles).
56
Ch.1: INTRODUCTION
Table 1: Storage ring applications with examples of The time dependent beam intensity I(t) can
past, existing, and planned machines. often be approximated by an exponential function
Beam accumulation: AA, AR, CR, EPA, I(t) = I(0) exp(−t/τ )
MIMAS, PAR, PIA, PSR, RESR, RR, SNS where the decay time τ and, correspondingly,
Beam quality improvement: LEIR, ILC DR the store time ranges from a few turns to 13
Stretcher, slow extraction: ELSA, KSR days (ISR [6]). τ can be dominated by a va-
Synchrotron light source: ALS, APS, AS, riety of effects including lattice nonlinearities
(Sec.2.3), beam-beam (Sec.2.5), space charge
BESSY, CLS, DIAMOND, ESRF, INDUS,

NSLS, PLS, SOLEIL, SLS, Spring-8 (Sec.2.4.5.1), intrabeam and Touschek scattering
Collision with internal target: COSY, HERAe, (Sec.2.4.12), interaction with the residual gas or
HESR, IUCF, MIT-Bates, Nuclotron target (Sec.3.3.2.1), or the lifetime of the stored
Collider: AdA, BEPC, HERA, ISR, KEKB, particle. In this case, the beam lifetime measure-
RHIC, LHC, Tevatron, VEPP-2000 ment itself can be the purpose of a storage ring
Stored beam experiments: ASTRID, experiment [1].
ESR, g − 2, pEDM, TARN, TSR, UMER The main consideration in the design of a
storage ring is the preservation of the beam qual-
ity over the store length. The beam size and mo-
mentum spread can be reduced through cooling
(Sec.2.7), often leading to an increase in the store
time. For long store times vacuum considera-
tions are important since the interaction rate of the
stored particles with the residual gas molecules
is proportional to the pressure, and an ultra-high
vacuum system may be needed (Sec.5.10). Dis-
tributed pumping with warm activated NEG sur-
faces or cold surfaces in machines with supercon-
ducting magnets are ways to provide large pump-
Figure 1: Small storage ring (CRYRING at the Manne
Siegbahn Laboratory) with main components labeled.
ing speeds and achieve low pressures even under
conditions with dynamic gas loads.
The largest application of storage rings to-
Higher order multipoles are used to correct chro-
day are synchrotron light sources (Sec.2.2.4), of
matic aberrations, to suppress instabilities, and to
which about 50 exist world wide. Storage ring
compensate for nonlinear field errors of dipoles
light sources are continuously improved and will
and quadrupoles. Magnetic multipole functions
remain the dominant form for the foreseeable fu-
can be combined in magnets. Beams are stored
ture [7].
bunched with radio frequency (rf) systems, and
In experiments where the beam collides with
unbunched. The magnetic lattice and rf sys-
an internal target or another beam (Sec.1.6.3), a
tem are designed to ensure the stability of trans-
storage ring allows to reuse the accelerated beam
verse and longitudinal motion (Sec.1.6.21), re-
many times if the interaction with the target is suf-
spectively. New technologies allow for better
ficiently small. In hadron colliders and ion storage
storage rings. With strong focusing the beam
rings store times of many hours or even days are
pipe dimensions became much smaller than previ-
realized, corresponding to up to 1011 turns and
ously possible. For a given circumference super-
thereby target passages. Ref.[3] is the first pro-
conducting magnets (Sec.7.2.4) make higher en-
posal for a collider storage ring.
ergies possible, and superconducting rf systems
A number of storage rings exist where the
(Sec.7.3.9) allow for efficient replenishment of
beam itself or its decay products are the object of
synchrotron radiation losses of large current elec-
study.
tron or positron beams. Storage rings have instru-
mentation to monitor the electrical and mechan-
References
ical systems, and the beam quality. Computers
are used to control the operation (Sec.7.7). Large [1] M. Larsson, Rep. Prog. Phys. 58, p.1267 (1995)
storage rings have millions of control points from [2] K.-J. Kügler, W. Paul, U. Trinks, Phys. Lett. B,
all systems. Vol.72, Issue 3, p.422 (1978)
57
[3] G. O’Neill, PR 102, 1418 (1956) the magnetic field ramps along with the beam mo-
[4] Yu.A. Litvinov et al, Nucl. Phys. A 756 3 (2005) mentum increase; the ejection elements extract
[5] K. Johnsen, CERN 84-13 (1984) the beam when it reaches the desired energy.
[6] W. Scharff, Particle Accelerators and their uses’,
Harwood (1986)
[7] M. Bei et al, NIM A 622, 518 (2010)
1.6.21 Synchrotrons
C. Zhang, S.X. Fang, IHEP
Synchrotrons are characterized by the magnetic
field synchronizing with momentum of particle

beams and the electric field synchronizing with
their circulating frequency, so that beams may Figure 1: Schematic of a synchrotron.
circulate along stationary orbits and be continu-
ously accelerated. Particle beams can be acceler- As the energies and the arrival times to the rf
ated to higher momentum in synchrotrons, where cavity of the particles in a beam are slightly differ-
the phase slipping is avoided for synchronization ent, there are energy and phase deviations in ref-
and magnets get less massive as in cyclotrons erence to the idealized synchronous particle. The
(Sec.1.6.4). In a synchrotron, the guide field B nonsynchronous particles will be lost in the ac-
varies with time as celerator if there is no longitudinal focusing. In
p(t)c 1944, Veksler [2] in the USSR and McMillan [3]
B(t) = (1) in the US independently discovered the principle
Zeρ
of phase stability in synchrotrons. Figure 2 illus-
where p is particle momentum, c is speed of light, trates the phase oscillation, or synchrotron oscil-
ρ is orbit radius, and Ze denotes particle charge. lation, below and above the transition energy.
As seen in Eq.(1), ρ stays constant if variation of
B matches the increase in p.
Another requirement in synchrotrons is for
the frequency of rf field frf to synchronize with
circulating frequency f0 of the particle beam,
frf = h · f0 (2)
Here h is harmonic number (integer). There may
have up to h synchronizing particle bunches in a
synchrotron.
Synchrotrons posed technical challenges to
magnets, microwave as well as their control and
these became possible after World War II. In Figure 2: Phase oscillation in synchrotrons, stable
1946, F. Goward and D. Barnes at the Telecom- phase is in the range (0, π/2) below the transition en-
munications Research Laboratory of UK modified ergy and (π/2, π) above the transition energy.
its small betatron to operate as a synchrotron [1].
In the following year, a team at General Electric There are two factors affecting the revolution
Co. constructed a dedicated 70 MeV electron syn- frequency of particles in a synchrotron, i.e. speed
chrotron. In 1952, BNL finished a proton syn- and orbit length: the higher the energy of a par-
chrotron of cosmic ray energy range, the 3 GeV ticle, the faster it moves and the longer is its or-
Cosmotron. bit. The speed is dominating factor at low energy,
Figure 1 shows the schematic of a syn- while orbit length dominates when approaching
chrotron. As shown, particle beam is injected into the speed of light. This shows that there is a tran-
the vacuum pipe through a deflector; bending and sition energy γt : higher energy particles circulate
focusing magnets confine the particles to move faster when γ < γt , and slower when γ > γt .
along and around the central orbit; an rf cavity in- Below transition, the particle “b” circulates with
stalled in a straight section accelerates the beam; higher frequency for Eb > Ea and gains more
58
Ch.1: INTRODUCTION
energy for φb > φa than the synchronized par- are combined in the same magnets, while in the
ticle “a”, so that it oscillates around the particle latter case dipoles and quadrupoles play roles of
“a” anticlockwise in the φ-ΔE space when γ< γt . bending and focusing separately, so that it is more
Similarly, particle “d” oscillates around the par- flexible providing even stronger focusing.
ticle “c” clockwise when γ> γt . In case when The alternating gradient focusing quickly su-
Δφ = φ − φs is small enough, the “restore force”, perseded constant focusing in synchrotron design.
ΔVrf = Vrf (φ) − Vrf (φs ) ≈ 2πfrf V cos φs · Δφ, CERN immediately abandoned its 10 GeV weak
is nearly constant and the synchrotron oscillation focusing plan and constructed a 25 GeV strong
behaves as a simple pendulum, focusing proton synchrotron. In 1959 and 1960,

d2 Δφ the CERN Proton Synchrotron (CPS), and the
+ ωs2 Δφ = 0 (3) BNL’s Alternating Gradient Synchrotron (AGS)
dt2
with oscillation frequency of 30 GeV completed successively. Their mag-

netic gaps are only 70-80 mm.
ωs f0 hη cos φs eV
fs = = − (4) For the electric and magnetic fields to be
2π βs 2π Es modulated in synchronism with beam momen-
where η = 1/γt2 − 1/γ 2 , βs and Es are relative tum, synchrotrons are operated in pulsed mode
velocity and energy of the synchronous particle with typical average beam current of the or-
respectively. der of μA’s. The maximum beam current in
The transverse focusing was explored in cy- synchrotrons is limited by space charge effects
clotrons (Sec.1.6.4). Particles get focusing in both (Sec.2.4.5) and a variety of other instabilities,
horizontal and vertical planes if the guide field in- and also by beam-beam interactions in colliders
dex n satisfies 0 < n < 1 with (Sec.2.5).
ρ ∂By Synchrotrons are applied for nuclear and par-
n=− ( )r=ρ (5)
B0 ∂r ticle physics experiments of fixed targets, and
However, this constant-gradient focusing is rather serve as booster injectors for higher energy ac-
weak which makes the beam’s cross section large celerators. Synchrotrons are used to make colli-
and leads to bulky vacuum chambers and massive sions between oppositely directed beams, known
magnets. The typical size of vacuum chamber is as colliders, which in turn can provide high ef-
800-mm wide by 200-mm height, while the total fective interaction energy (Sec.1.6.3). The high
weight of magnates gets to ∼10000 tons in weak energy frontier accelerators constructed in re-
focusing synchrotrons of GeV scale such as the cent years are all collider type. The light emit-
Cosmotron. ted from electron synchrotrons was harmful in
In the same year when Cosmotron was com- the history of e+ -e− colliders, it has now be-
pleted, E. Courant, M. Livingston and H. Snyder come an important tool for scientific research (see
proposed the concept of strong focusing (or alter- also Sec.1.6.20). Dozens of synchrotron radia-
nating gradient focusing) [4]. The idea had been tion sources have been constructed in the world.
visualized by Chritofilos in an unpublished paper Proton synchrotrons are used as spallation neu-
[5]. The strong focusing opened a new era of syn- tron sources (Sec.1.6.16). Proton and heavy ion
chrotrons, which allowed using smaller magnets synchrotrons are also widely applied for medical
and reaching higher beam energies. In a strong treatment (Sec.1.6.14).
focusing synchrotron, horizontal focusing and de-
focusing (vertical focusing and defocusing) mag- References
nets are alternatingly arranged to make a global
strong focusing like in an optical system, [1] F.K. Goward, D.E. Barnes, Natrure, 158 (1946),
413
1 1 1 d
= + − (6) [2] V.I. Veksler, Dokl. Akad. Nauk SSSR 43, 346 and
F f1 f2 f1 f2 44, 393 (1944)
If the lenses have equal and opposite focal [3] E.M. McMillan, PR 68 (1945) 143
lengths, i.e. f1 = −f2 , the overall focal length [4] E.D. Courant, M.S. Livingston, H.S. Snyder, PR
F = d/f 2 > 0, with d distance between two 88 (1952)
lenses. There are two catalogues of strong focus- [5] N.C. Chrostofilos, U.S. Patent no. 2.736,799
ing, i.e. combined function and separated func- (1956)
tion. In the former case, the bending and focusing
59
1.6.22 Two-Beam Accelerator Up to 90% of the beam energy is transformed in rf

J.P. Delahaye, CERN power after which the remaining beam is dumped.
Acceleration of the main beam is pursued in a fol-
Novel schemes of Two-Beam Acceleration (TBA) lowing sector using a fresh drive beam.
have been proposed [1, 2] as power source for
high-energy facilities and especially for e± linear Drive beam generation The drive beam gen-
colliders. In the TBAs, a high-current, low-energy eration complex is shown in Fig.2 [7]. It produces
drive beam is used to generate rf power that is one after the other all required drive beams for
applied to a high-gradient acceleration structure, each linac. The drive beams are generated as one
where a low-current beam is accelerated to high long train with a bunch spacing of 60 cm. A fully-
energy (Fig.1). Many variations of the TBA con- loaded normal-conducting linac operating at a low
cept have been investigated, with early work [1] frequency (1 GHz) is used to accelerate the drive
centered on using FELs to extract rf power from beams to 2.4 GeV. The bunch spacing is then re-
the drive beam. However, most rf extraction con- duced to 2.5 cm in three successive stages in a de-
cepts (FELs, klystrons, gyrotrons, wake-field) can lay loop and two combiner rings using funneling
be configured into a TBA. There are also several techniques to repetitively interleave 240 ns-long
choice options for the drive beam source, and the slices of the trains. As a result, the bunch repe-
method of drive-beam acceleration. The TBA has tition frequency and the beam intensity are mul-
the great advantage of high efficiency for power tiplied by a factor of 24. Operating the linac in
conversion from the drive beam to rf power. In ad- the fully-loaded mode enables the beam to be ac-
dition, TBAs scale [3] favorably to high frequen- celerated with an rf-power-to-beam efficiency of
cies (≥ 11.4 GHz) and high accelerating gradi- ≈ 96%. The rf power for each drive-beam accel-
ents (≥ 100 MV/m). erator is supplied by efficient 15 MW multi-beam
klystrons with long rf pulse at low frequency. A
particularly attractive and cost effective feature of
the CLIC scheme is that energy upgrading of the
collider only requires a pulse lengthening of the
modulators which drive the klystrons and not an
increase in the number of klystrons.
Rf power production By initially sending the
drive beam trains in the opposite direction to
the main beam, different time pulses in the train
Figure 1: Conceptual layout of Two-Beam Accelerator are used to power different sections of the main
(TBA). linac. The structures to extract the power from
the high-intensity drive beam are referred to as
Two main approaches of TBA research have PETS (Power Extraction and Transfer Structures)
been pursued so far: [8]. Each 0.5-m structure extracts a rf power of
The relativistic klystron approach (RK-TBA) 130 MW from the 100 A drive beam. The struc-
[4] developed by LBNL/LLNL uses induction ac- tures are passive microwave devices in which the
celeration of the drive beam and to maintain the drive beam bunches interact with a large aperture
energy at 10 MeV throughout most of the device. (25 mm diameter) structure with a shallow peri-
A klystron-like output structure extracts power at odically corrugated inner surface to preferentially
11.4 GHz from the drive beam. The primary tech- excite the synchronous TM01 mode at 12 GHz.
nical challenge of the RK-TBA lies in propagat- In the process, the beam kinetic energy is con-
ing the intense drive beam (hundreds of amperes) verted into electromagnetic energy at the mode
at low energy (10 MeV) over long distances. frequency. This energy travels along the structure
The Compact Linear Collider (CLIC) scheme with the mode group velocity and the rf power
[5, 6] is based on a relativistic drive beam which produced is collected at the downstream end by
is not reaccelerated to avoid active elements in the a power extractor and conveyed to the main linac
main tunnel. The drive beam is characterized by structure by rectangular waveguides. For the
a 12 GHz bunch structure, low energy (2.4 GeV) sake of beam stability, these structures have to be
and high current (100 A). Successive drive beam damped to reduce long-range transverse wakefield
trains supply power to a linac sector 876 m long. effects. PETS are made of eight octants separated
60
Ch.1: INTRODUCTION
Figure 2: Layout of CLIC RF power generation scheme.
by HOM damping slots connected to broad-band bunched beam transport through two reaccelera-
SiC rf damping loads. In case of problems, the tion induction cells and three traveling-wave ex-
power generated by these structures is turned off traction cavities. These experiments resulted in
by a remotely adjustable external reflector. a total rf output of > 200 MW, with phase and
Main beam acceleration The main linac is amplitude stable over a significant portion of the
made of a succession of two beam modules beam pulse. At LBNL a RK-TBA version [12]
integrating all necessary components including was designed as a power source for a linear col-
rf structures, beam instrumentation, vacuum, lider with 1.5-TeV CM collision energy.
quadrupoles, alignment & stabilization (Fig.3). Two generations of CLIC Test Facilities
(CTF1 1990-1995 and CTF2 1996-2002) have
demonstrated the technical feasibility of the CLIC
scheme. In CTF1, a peak power of 76 MW
was extracted from the drive beam by a high-
impedance 30-GHz traveling-wave section, and
used to reaccelerate the same beam. In CTF2,
a string of four power-extracting structures driv-
ing five accelerating structures increased the en-
ergy of a single electron bunch of the probe
beam by 55 MeV [13]. Both the high intensity
drive beam and the probe beam were generated
by laser-illuminated photo-cathodes in rf guns.
A new CLIC Test Facility CTF3 [14] has been
Figure 3: Two Beam Acceleration module. built at CERN to address the major key CLIC-
technology-related feasibility issues. In particular
Experiments The earliest TBA experiments CTF3 has demonstrated the generation of a 130-
[9] were performed on ETA-I addressing issues of ns 150-MeV 28-A drive beam with 2 cm bunch
power extraction. Work on the RK version started spacing using a fully-loaded linac and two stages
shortly after, using the ARC facility. These exper- of bunch interleaving resulting in an intensity and
iments [10] used a 1-MeV 1-kA 70-ns induction frequency multiplication by a factor 8 [15]. This
accelerator beam to produce 300 MW of rf power beam will be used to produce up to 1.5 GW of
level at 11.4 GHz. Reacceleration experiments 12 GHz power in a 20-m Test Beam Line deceler-
[11] were performed on ATA that demonstrated ator in order to carry out beam stability studies
61
Sec.1.7: ACCELERATOR COMPUTER CODES
and benchmark simulation codes. It will also Community standards

be used to power accelerating structures at their • Accelerator Lattice Descriptions: Complex
nominal gradient of 100 MV/m in prototype Two accelerator lattices may contain many thou-
Beam modules thus addressing the feasibility of sands of elements, so community standards
Two Beam Acceleration [16]. for lattice descriptions are important to
facilitate sharing of lattice information
References among codes. An overview of accelera-
[1] D.B. Hopkins, A.M. Sessler, J.S. Wurtele, NIM tor description formats can be found at
Phys. Res. 228 (1984) 15 http://cern.ch/AccelConf/ICAP06/PAPERS/
[2] A. Sessler et al, PRL 58, 2439 (1987); NIM A THM2IS01.PDF. The first, and most widely
306,592 (1991) used, standard for the specification of lat-
[3] S.M. Lidia et al, Snowmass Workshop (1996) tices is the Standard Input Format (SIF); this
[4] T.L. Houck, G.A. Westenskow, IEEE Trans. On is described in the documentation for the
Plasma Sci. 24 (1996) 938 MAD-X code, http://mad.web.cern.ch/mad.
[5] The CLIC Conceptual Design Report, The SIF definition of beamline elements
http://clic-study.org/accelerator/CLIC- has been very widely adopted. For real-
ConceptDesignRep.php istic analysis (as contrasted with idealized
[6] R. Tomas et al, PRST-AB 13,014801 (2010) design) two extensions of SIF have been
[7] The CLIC RF Power Source, CLIC note 364 found to be important: full instantiation
[8] I. Syratchev, 7th Workshop on High Energy (where every element has its own pa-
Density and High Power RF, Kalamata, Greece rameters, errors, etc.) and compatibility
(2005) with modern software tools. The latter
[9] D.B. Hopkins et al, SPIE High Intensity Laser reduces mainly to standard eXtensible
Processes 664 (1986) 73 Markup Language (XML), preferably
[10] M.A. Allen et al, PRL 63 (1989) 2472 disciplined by an XML Schema; see, e.g.,
[11] G.A. Westenskow, T.L. Houck, Proc.10th Int. http://code.google.com/p/ual/source/browse/
Conf. High Energy Part. Beams (1994)
trunk/doc/adxf. A Universal Accel-
[12] Zeroth-order Design Report for Next Linear Col-
erator Parser has also been devel-
lider, Appendix A, SLAC-474 (1996) p.925
oped that parses several formats; see
[13] H.H. Braun, CERN-PS-2001-008
[14] G. Geschonke et al, CTF3 design report,
http://www.lns.cornell.edu/∼dcs/aml/
CERN/PS 2002-008 • Self-Describing Data Sets (SDDS): SDDS
[15] P. Skowronski et al, IPAC10 is a self-describing file protocol that
[16] J.P. Delahaye et al, Mod. Phys. Lett. A, Vol.26, has been widely adopted in the light
No.40, 2997 (2011) source modeling community; see www.
aps.anl.gov/Accelerator Systems Division
/Operations Analysis/SDDSInfo.shtml
1.7 ACCELERATOR COMPUTER
CODES • Experimental Physics and Industrial Con-
R. Ryne, LBNL trol System (EPICS): EPICS is widely used
in accelerator control systems. Given
Simulation is essential to accelerator design, the ubiquity of EPICS, it is important
commissioning, operation, and upgrades. At to accelerator modelers who integrate on-
the present time several web sites maintain line accelerator codes into control sys-
lists of codes used by various groups, including: tems. Information about EPICS is lo-
http://oraweb.cern.ch/pls/hhh/code website.disp allcat/, cated at http://www.aps.anl.gov/epics/ and
http://www.slac.stanford.edu/accel/ilc/codes/, http://www-csr.bessy.de/epics/
http://pbpl.physics.ucla.edu/Computing/Code Overview/,
• Radiation safety modeling and standards:
http://www.cap.bnl.gov/mumu/
See http://rsicc.ornl.gov/index.html and
In the future it is expected that a list of ac-
http://irs.inms.nrc.ca/software/egsnrc/
celerator codes will be maintained at the website
for Physical Review Special Topics Accelerators The following tables list some widely used (non-
& Beams, http://prst-ab.aps.org/ commercial) codes and contact information.
62
Ch.1: INTRODUCTION
Beam Dynamics Codes:

(Below, PIC refers to codes with particle-in-cell space-charge capability.)
Ȁ
ǤǤȀ̱ ͵ǡǤ
ǤȀȀȀ
ǤǤ ǣǡǡ
ǡ ǤǤǤȀ̱ȀȀ
Ϊ
ǤǤ Ǧ
ǤǤȀǦȀ ͵ǢǢǦ
Ȁ ǤǤȀǤ ǢǡǢǢǢ

ǦǤǤȀ
Ǥ ̷ Ǥ Ǥ ǡ
ǤǤ ͵Ǧ

ǤǤ ǡǡǡʹȀ͵
ǤǤǤȀ̱Ȁ ǢǢ
ǦǢ
̷ǤǤ Ǣ
ǤǤǤȀȀȀ ǦǦǦ
ǤǤǤȀ ǡǡǡǡ
Ȁ ǤǤ ͵ǢΪ
Ǧ ǤǤǤȀ

ǤǤȀ̱ ΪΪ
ǤǤǤ ͵Ǣǡ
ǡǢǦǤ
̷Ǥ
Ǥ̷Ǥ ͵ǢǢ
ǦǤǤȀȀǤ ǡǡƬ
ǤǤȀȀ ͵Ǣǡ
ǤǤǤȀȀ Ǣ
ǦǤǤȀȀ Ȁ
ǣȀȀǤǤȀ ͵ǣǡ
̷Ǥ ǢǢǢ
ǤǤǤȀȀ ͵ǡ
libtracy.sourceforge.net/
ǤǤǤ ͵ǢǦǦǦ
ǤǤȀȀȀ

̷Ǥ ͵

ǤȀȀȀ ǢǢǢǦ
Beam Dynamics w/ emphasis on specific phenomena (beam-beam, e- -cloud, spin):

ǦǤǤȀ̱ȀȀǤ ǡ
ǡ ̷ǤǤ ǦǦǦ
ǦǦǤǤǤȀǦǦȀ ǦǦ
͵ ǤǤ ǢǦǢǦǢǦǢǦ
ǡ ǦǦǤǤǤȀǦǦ ǢǦǢ Ǣ͸Ǧ
Ǧ ̷ǤǤ ǦǢǦ
ǦǦǤǤǤȀǦǦǦ ǦǦǦǢǦǡǢǦ
ǦǦǤǤǤȀǦǦǦ ǤǢǦǡǦǦǢ
ǡ ̷ǤǤ ǦǡǢǦ
̷Ǥ ʹǦǢ
ǤǤȀȀ Φ

Ǥ̷Ǥ ǡǤǡǡ
Ȁ ̷Ǥ ͵ǦǦǢ
Electromagnetics:
ǤǤ ʹǤͷ
͵ ǤǤǤȀȀ̴ȀȀ ͵ Ǣ͵Ȁ͵ǡ͵ǡ͵ǡ͵ǡ͵
ǤǤǤ ͵
ǤǤǦǤȀǤǫαʹ͵ͷ Ǧ
ǡ ǤǤǤȀǤ ǦǦǦ
Ȁ ǤǤ ʹ
ǤǤȀȀȀ ͵Ǣ
ǤǤȀȀǤ ͵Ǣ
63
Sec.1.7: ACCELERATOR COMPUTER CODES
Free Electron Laser Codes:

Ǥ̷Ǥ ͵Ǧ

Ǥ̷Ǥ ͵Ǧ ǡ

̷Ǥ ǦǦǢ͵ǢǤƬǤ
ǤǤ ̷Ǥ ͵Ǧ ƬǢǦ
ǤǤǤ ǦǢͳ
ǤǤǤǤȀȀ ͵Ǧ ǡ
Ǧ
Synchrotron Radiation Modeling Codes:

ʹ ǤǤȀȀǤ

ǤǤǤǤȀ
Ǥ̷Ǥ ǦǡǡǤ
ǤǤȀȀȀ Ǥ
Ȁ͵ Ǥ̷Ǥ

ǤǤȀȀȀȀȀʹǤ͵ Ǧ
Beam/Material Interactions:
ǤǤ

Ͷ ǤͶǤȀͶȀ

Ͷ ͶǤǤ
Ͷ
ǤǤȀ̱ȀȀ
ǦǤǤȀȀ ͵
ǦǤǤȀǤ

Pre-, Post-Processors, Parsers, Auxiliary Codes:

Ȁ ǤǤǤȀ̱ȀȀ ǦǤǡǤ
ͷ ͷǤǤȀǡǤǤȀ ͷǡȀ
ǤǤǤȀ
Other Web Resources Since 2006, the proceedings of the International Computational Accelerator
Physics (ICAP) conference series have been administered by the Joint Accelerator Conferences Web-
site (JACoW); they are accessible and searchable at http://www.jacow.org. As of this writing, Physical
Review Special Topics - Accelerators and Beams is planning to be a resource for information about ac-
celerator codes; see http://prst-ab.aps.org/. Another searchable resource is the CERN Document Server,
http://cdsweb.cern.ch/.
64
Errata and Additions to Handbook, 2nd Edition
January 18, 2016
p.170, left column, Dynamic emittance

Original text
New text A more complete expression for the dynamic emittance was computed in:
• A.V.Otboyev, E.A.Perevedentsev, PRST-AB 2, 104401 (1999)
p.261, second row from the top, longitudinal impedance and wake of a BPM consisting of two strip lines
2 2
φ0
Original text Z0 = 2Zc 2π 2 sin2 kL−i sin2kL , W0 = 2Zc c 2π
φ0
δ(z) − δ(z +2L)
2 2

New text Z0 = Zc 2π φ0
2 sin2 kL−i sin2kL , W0 = Zc c 2πφ0
δ(z) − δ(z +2L)
p.572, right column, Lifetimes of carbon foils

Original text kfoil = 0.018 for foils produced by vapor deposition
New text kfoil = 0.0018 for foils produced by vapor deposition
p.576, right column, References

Original text
New text Remove Ref.[1]. Add two additional references:
• L.R. Scherk, Can. J. Phys. 57, 558 (1979)

• P.B. Keating et al., Phys. Rev. A 52, 4547 (1995)
1
Chapter 2. BEAM DYNAMICS
2.1 PHASE SPACE α, β, γ may be selected to match the x x distri-

bution of the input beam. In a recursive system,

2.1.1 Linear Betatron Motion [1, 2]
by CHINESE UNIVERSITY OF HONG KONG on 02/10/15. For personal use only.
the parameters are usually defined by the structure

D.A. Edwards, DESY rather than by the beam.
M.J. Syphers, Michigan State U. Matrix parameterization

Equations of motion x2 x
= M(s1 , s2 ) 1 (5)
e ∂B 1 x2 x1
x + Kx x = 0, Kx ≡ + 2 (1)
p ∂x ρ β2
e ∂B m11 = (cos Δψ + α1 sin Δψ)
y + Ky y = 0, Ky ≡ − β1
p ∂x √
m12 = β1 β2 sin Δψ
z = −x/ρ
1 + α1 α2 α1 −α2
m21 = − √ sin Δψ + √ cos Δψ
where x, y are the transverse displacements from β1 β2 β1 β2
the reference particle of another particle of equal
momentum p, and z the distance by which it leads β1
m22 = (cos Δψ − α2 sin Δψ)
the reference particle, which is itself a distance β2
s along the reference trajectory from the origin.
When s1 and s2 are separated by the repeti-
The reference trajectory is flat in the x, s plane,
tion length of the system, e.g. circumference C in
with its bending radius of curvature ρ due only to
a synchrotron, M reduces to
a static magnetic field in the y direction.
Solution in phase-amplitude form Below α β
MC = I cos ψC + J sin ψC , J =
when results are expressed for x, x , typically −γ −α
they apply for y, y as well. (6)
√ Note that J2 = −I and so MnC = I cos nψC +
x = A β cos (ψ + δ) (2) J sin nψC . The transformation of J from
A one location to another is given by J2 =
x = − √ [α cos (ψ + δ) + sin (ψ + δ)]
β M(s1 , s2 )J1 M(s1 s2 )−1 , the associated phase
advance is
where A and δ are constants of integration, α ≡
− 12 dβ(s)/ds. −1 m12
Δψ = tan (7)
The amplitude function, or β-function, satis- β1 m11 − α1 m12
fies where the mij are elements of M. The condition
2ββ − β 2 + 4β 2 K = 4 (3) for stable motion is | cos μ = 12 trMC | ≤ 1.
and the phase advances according to dψ/dx = Emittance The characterization of the area oc-
1/β. Expressing A in terms of x, x yields cupied by the beam in x, x space or y, y space
is termed the emittance in that degree-of-freedom.
A2 = γx2 + 2αxx + βx2 (4) Various definitions are in use. For a Gaussian
1 2
beam with standard deviation σ in x and βx +αx,
= x + (αx + βx )2
β the area containing a fraction F of the beam is
with γ ≡ (1 + α2 )/β. In a single pass system 2πσ 2
such as a linac, the Courant-Snyder parameters ε=− ln(1 − F ) (8)
β
65
Sec.2.1: PHASE SPACE
A more general definition which reduces to the Tune and chromaticity The number of oscilla-
15% case of Eq.(8) is tions per turn of circumference C in a cyclic ac-
√ celerator is the tune,
ε = x2 x2 − xx 2 (9) "
ΔψC 1 ds
and the beam “sigma” matrix is defined by ν≡ = (16)
2 2π 2π β
x xx β −α The momentum dependence of restoring
Σ= =ε (10)
xx x2 −α γ force leads to a tune variation characterized by the
The normalized emittance, εN ≡ γ(v/c)ε, chromaticity, ξ,
where here γ is the Lorentz factor, is proportional Δp
Δν = ξ (17)
to the area in x, px phase space and so is ex- ps
pected to exhibit adiabatic invariance under en- The natural chromaticity associated with the lin-
ergy change. ear optics is "

Momentum dispersion Trajectories of parti- 1
ξ=−
Kβds (18)
cles having momenta different from that of the 4π
reference trajectory by Δp contain the additional A single sextupole magnet of length and B ≡
term D(s)Δp/ps , where D is the dispersion func- ∂ 2 By /∂x2 will contribute
tion. 1 B
D satisfies an inhomogeneous Hill’s equation Δξ = ± Dβ (19)
4π (Bρ)
ps 1 Δp 1 ps where the + sign is associated with the bend
D + Kx − 2 D= (11) plane; this is the basis for chromaticity adjustment
p ρ p ρ p
where ρ is the radius of curvature for ps . using sextupoles.
The matrix M (s1 , s2 ) for the bend plane
may be enlarged to include propagation of off- References
momentum rays, [1] E.D. Courant, H.S. Snyder, Ann. Phys. 1 (1958) 1
⎡ ⎤ ⎡ ⎤ ⎡x ⎤
x2 m11 m12 0 m14 1 [2] D.A. Edwards, M.J. Syphers, An Introduction to
⎢ x2 ⎥ ⎢m21 m22 0 m24⎥ ⎢ x1 ⎥
the Physics of High Energy Accelerators, Wiley
⎢ ⎥=⎣ ⎢ ⎥
m31 m32 1 m34⎦ ⎣ z1 ⎦
⎣ z2 ⎦ (12) (1993)
Δp Δp
ps 0 0 0 1 ps
where the elements of Eq.(5) are unchanged and 2.1.2 Longitudinal Motion [1, 2]
D.A. Edwards, DESY
m14 = D2 − m11 D1 − m12 D1 (13)
M.J. Syphers, Michigan State U.
m24 = D2 − m21 D1 − m22 D1
m31 = D1 + m21 D2 − m11 D2 Equations of motion The only role of the trans-
verse degrees-of-freedom taken into account in
m32 = −D1 + m22 D2 − m12 D2 this section is the bend-plane momentum disper-
m34 = −D2 (m21 D1 + m22 D1 ) sion function D defined in Sec.2.1.1, which leads
s2 − s1 to changes in z. Momentum change is due to an
+ D2 (m11 D1 + m12 D1 ) +
γ2 RF electric field E in the longitudinal direction
Particles differing in momenta will follow acting on charges of magnitude e,
dz D(s)
paths of differing lengths. For a sufficiently long z ≡ =− Δp (1)
path S, e. g. the circumference of a large syn- ds ρ(s)ps
chrotron, with the use of Eqs.(12-13) the relative dΔp eE(s + z) − eE(s)
path length difference ΔS becomes Δp ≡ = (2)
! ds vs
ΔS z2 −z1 D 1 Δp Bunch compression The principle of a bunch
=− = − 2 compressor may be illustrated by a magnetic chi-
S s2 −s1 ρ γ ps
(14) cane of four magnets with bends θ, −θ, −θ, θ
The compaction factor, αp , and the transition separated by distances L, S, L. The result is
gamma, γt , are defined by Δz = −2LθΔp/ps in the limit of magnets of
! zero length. If the particle momentum increases
1 D
αp = 2 ≡ (15) from bunch front to rear, the bunch becomes
γt ρ shorter in passing through this system.
66
Ch.2: BEAM DYNAMICS
Synchrotron oscillations Here it is assumed Bucket area and longitudinal emittance The
that the rate of energy change is sufficiently low separatrix passes through the unstable fixed points
so that the electric field may be represented by a located at z1 and the turning point z2 found from
traveling wave traveling with the reference parti- cos k(s + z2 ) + kz2 sin ks = cos k(s + z1 ) +
cle. The difference equation approach suitable to kz1 sin ks. The area bounded by the separatrix
abrupt energy change is discussed in [2]. In the may be calculated by numerical integration. In the
slow case, Eqs.(1) and (2) become particular case of sin ks = 0, a stationary bucket,
z2 = −z1 = −λ/2, and the bucket area is
z = −
η 1
Δp, η ≡ 2 − 2
1
(3)
ps γt γ 8λ Es eV0
A0 = (10)
eE0 πc 2πhη
Δp = [sin(ks + kz)−sin(ks)] (4)
vs The stable range of z is that for which Δp/ps does

where k ≡ 2π/λ; for a cyclic accelerator the not exceed the bucket height

wavelength λ is the circumference divided by an Δp 2eV0
integer h. Transition gamma, γt , is defined in = (11)
Sec.2.1.1. Combining Eqs.(3) and (4) gives ps max πhηps vs
A longitudinal emittance may be defined us-
ps ηeE0
z + z + [sin(ks + kz)−sin(ks)] = 0 ing Eq.(8) in analogy with its transverse counter-
ps ps vs part, and in z, Δp coordinates is expected to ex-
(5) hibit adiabatic invariance with energy.
For sufficiently small z and ps = 0, The intrinsic nonlinearity of Eq.(2) can be
2πηeV0 ameliorated through the use of harmonic RF
z + cos(ks)z = 0 (6) structures. An example may be found in the in-
ps vs hλ2
stallation at DESY of a superconducting 3.9 GHz
where V0 = E0 hλ. The number of oscillations in structure in the 1.3 GHz-based FLASH free elec-
a distance hλ – the synchrotron oscillation tune – tron laser [3].
is
1/2
hη References
νs = eV0 cos ks (7)
2πβ 2 Es [1] E.D. Courant, H.S. Snyder, Ann. Phys. 1 (1958) 1
where Es is the energy of the reference parti- [2] D.A. Edwards, M.J. Syphers, An Introduction to
cle and here β ≡ vs /c. So, for stable oscilla- the Physics of High Energy Accelerators, Wiley
(1993)
tions above transition energy it is necessary that
[3] E.R. Harms et al, Proc. SRF 2009, http:// accel-
cos ks > 0. This condition differs in sign from
conf.web.cern.ch/AccelConf/srf2009 /index.htm
that in [1, 2] due to the use here of z rather than
the phase at rf station passage. A first integral of
Eq.(5) for ps = 0 is 2.1.3 Linear Coupled System
D.A. Edwards, DESY
1 2 1 M.J. Syphers, Michigan State U.
Δp + U = C (8)
2 2
Linear coupling between two of the three degrees-
where C is a constant, and of-freedom may for the purposes of this section be
−1 eV0 ps characterized by a 6 × 6 matrix of the form
U= [cos k(s + z)+kz sin ks] (9) ⎡ ⎤
πhη vs ADF
The “potential energy” U has zero slope at z = M = ⎣E B 0 ⎦ (1)
z1 ≡ (λ/2 − 2smodλ ); this is the unstable fixed G 0 C
point in the neighborhood of which the differen- where A–G are 2 × 2 matrices. M operates
tial equation approach becomes invalid. The sta- on a column vector (x, x , y, y , z, Δp/p) or
ble region in longitudinal phase space is termed a (x, px , y, py , z, Δp). With either choice M is
bucket; note that U has the form of a descending symplectic (Sec.2.3.6); the second, related to the
series of buckets above transition and an ascend- canonical formalism, may be preferable if conser-
ing series below. vation of phase space volume is of interest.
67
Transverse coupling The equations of motion: where the fields are those of the vector potential
1 As = −(E /ω) sin(ωt) · x. The reference particle
x + Kx (s)x = S(s)y + R(s)y + R(s) y is at the midpoint of each cell at the zero-crossing
2
of the field, and E is the amplitude of ∂E/∂x at
1
y + Ky (s)y = S(s)x − R(s)x − R(s) x x = 0. For an n-cell structure the matrix is
2 ⎡ ⎤

S ≡ Bskew /(Bρ), R ≡ Bs /(Bρ) 0 0 − α1 − 4D
nλ
− nαλ
4
⎢ 0 0 −D1
−α ⎥
∂B ∂Bx ∂R ⎢ ⎥ (4)
B ≡
, R ≡ ⎣ −α − nαλ 0 ⎦
y
, Bskew ≡ (2) 4 0
∂x ∂x ∂s
−D1
− α1 − 4D
nλ
0 0
The S and R terms arise from skew quadrupole
and solenoidal fields. Skew quadrupole fields are where α is the magnitude of the bend angle of
common in synchrotrons due to magnet fabrica- each of the four magnets that create the disper-
tion or alignment errors. sion D after the first two, and each cell of the pill-
A perturbative treatment of the transverse box cavity has length λ/2. The matrix used for
oscillations x = a(βx /β0 )1/2 cos ψx , y = the cavity is limited to lowest order in λ/(2πnD),
b(βy /β0 )1/2 cos ψy shows that at the sum reso- and the exchange is ideal as in [5]. Experimental
nance νx + νy = integer, a2 − b2 = constant, results cited in [6] were made with a 5-cell struc-
so the sum resonance leads to instability [1]. In ture of TESLA style. This process is also finding
contrast, for the difference resonance a2 + b2 = use in beam diagnostics [7]. See also Sec.2.2.5.
constant and the motion though coupled is stable.
A potential benefit of complete coupling for References
an electron synchrotron is discussed in Sec.2.2.5. [1] E.D. Courant, H.S. Snyder, Ann. Phys. 1 (1958) 1
A positive use of solenoidal coupling for phase [2] R.A. Brinkmann, Ya. Derbenev, K. Flötttmann,
space manipulation is found in the round-to-flat PRST-AB 4, 053501 (2001)
beam transformation [2, 3, 4]. [3] Yin-e Sun, PhD thesis, U. Chicago (2005),
Transverse-longitudinal coupling In a syn- Fermilab-thesis-2005-17,
chrotron, the frequency modulation of the beta- http://lss.fnal.gov/archive/thesis/fermilab-thesis-
tron oscillations arising from the natural chro- 2005-17.shtml
maticity, ξ, leads to sidebands νx ± nνz , an in- [4] A.W. Chao, Proc. 11th Chinese Accel. Phys.
stance of synchrobetatron resonances (Sec.2.3.4). symp., Fuzhou (2011)
The natural chromaticity is usually compensated [5] K.J. Kim, A. Sessler, AIP Conf. Proc. Vol.821,
by the introduction of sextupole magnets; this p.115 (2005)
step in itself implies the irreversible introduction [6] T.W. Koeth, PhD thesis, Rutgers U. (2009),
of nonlinearities into the dynamics. Also dis- http://hdl.rutgers.edu/1782.2/
rucore10001600001.ETD.000051364
cussed in Sec.2.3.4 are the sidebands due to pas-
[7] Y-E. Sun et al, http://arxiv.org/abs/1003.3126
sage through accelerating structures in which the
dispersion D = 0, as well as the sidebands asso-
ciated with the transverse fields, Ex and By . 2.1.4 Orbital Eigen-Analysis for Electron
That these same fields can be used to advan- Storage Ring
tage is illustrated by the transverse-longitudinal J.A. Ellison, U. New Mexico
emittance interchange process, in which a deflect- H. Mais, DESY
ing mode rf structure is placed at the midpoint of G. Ripken, Deceased 2004
a 4 bend double “dogleg” arrangement [5]. For a
TM110 π-mode pillbox structure the equations of a. Basic problem A general 6-D formalism
motion for this process are is presented for the calculation of the bunch pa-
rameters (e.g. 6-D stationary beam-envelope ma-
px eE
x = , px = cos(ks) · z trix) for electron storage rings including radiation
ps c damping and quantum excitation. The problem
eE is formulated in terms of a stochastic differential
z = sin(ks) · x
ps ck equation (SDE) and basic to our approach is the
eE eE orbital eigen-analysis first introduced in [1]. The
pz = cos(ks) · x + sin(ks) · x (3) latter gives a more general framework than that of
c kc
68
Ch.2: BEAM DYNAMICS
Courant-Synder. An SDE approach was first in- grained sense. This article presents results of an
troduced in [2] and developed further in [3]. At analysis of this single particle probability density.
a later stage work [4] was incorporated and the More precisely, the (periodic) 6-D closed orbit
starting point is the SDE in Frenet-Serret coordi- xco , satisfying (1) with b = 0 will be defined by
nates with respect to a design orbit as described in an integral equation, (1) will be linearized around
[5]. The 6D-SDE is xco and the linearized equation analyzed.
1 b. The equation x = A(s)x and its Eigen-FSM
x = A(s)x+c(s)+g (x, s)+ 2 b(s)ξ(s)e6 (1)
The solutions of the linear periodic Hamiltonian
where x = (x, px , y, py , z, pz )t . We have ex- system
panded up to second order in the dynamical vari- x = A(s)x, At J+JA = 0, A(s+C) = A(s) (3)
ables and retained only the leading nonlineari-
are central to our analysis. Here J =
ties due to sextupoles and due to radiation effects

diag(J2 , J2 , J 2 ) is the unit symplectic matrix
in quadrupoles. All functions except ξ are C-
0 1
periodic in s where C is the ring circumference. where J2 = .

−1 0
The matrix A(s) is Hamiltonian and gives rise
to the linear symplectic synchrobetatron oscilla- The transfer map, M(s, s0 ), is a funda-
tions. The x-independent term c describes closed mental solution matrix (FSM) in s which sat-
orbit distortions induced by dipole field errors and isfies M(s0 , s0 ) = I (often called the princi-
by the fact that the energy losses in the bending pal solution matrix). The basic properties are
magnets and quadrupoles are not replaced at the (P1) M(s2 , s1 )M(s1 , s0 ) = M(s2 , s0 ) (transport
location in the ring where they occur. Our anal- property), (P2) Mt JM = J (symplecticity) and
ysis is perturbative. The parameter is inserted (P3) M(s + C, s0 + C) = M(s, s0 ) (periodicity).
to indicate the perturbation size and to discuss the We assume all solutions of (3) are bounded.
perturbation procedure and the nature of the error This is the case if and only if the monodromy ma-
estimates. In the end, can be taken to be one in trix, M(C, 0), has six linearly independent eigen-
applications of the formulas. The first perturba- vectors, w k , and its eigenvalues (characteristic
multipliers), ρk = exp(ı2πνk ), have modulus one
tion term is g (x, s) = δA(s)x + f(x, s). Here
where the νk are tunes (characteristic exponents),
δA models both the energy losses from radiation
see [7]. To avoid resonance, we further assume
and the energy gain in the rf cavities, and f takes that the ρk are distinct. Because M is real we can
into account the nonlinear terms due to sextupoles choose the νk such that
and due to radiation effects in quadrupoles. The
0 < ν1 < ν3 < ν5 < 1/2, ν2l = −ν2l−1 , (4)
last term in (1) simulates the stochastic excitation ∗
of the particle motion due to the quantum nature and the w k such that w 2l = w 2l−1 . Thus (1 −
∗
ρj ρk )w H
j Jw k = 0, so that w jH Jw
k = 0 for
of the radiation. Here ξ is Gaussian white noise, b
is an amplitude function proportional to and ek j = k. Here H denotes Hermitian conjugate. Let
is the unit vector with 1 in the k-component, thus w k = ak + ibk , ak and bk real then w kH Jw
k = iγk
the stochastic excitation only affects the pz com-
where γk = 2ak Jbk and γ2l = −γ2l−1 . More
t
ponent directly. The explicit form of these quanti- compactly
ties can be found in [5] and details of our analysis WH JW = ıĨ, where (5)
below will be given in [6].
The main quantity of interest is the N particle W = [w 1, · · · , w
6 ], Ĩ = diag(γ1 , · · · , γ6 ).
random bunch density Finally, the γk are nonzero (since W is nonsingu-
lar) and we normalize the w 2l−1 so that γ2l−1 =
1
N
ρN (x, s) := δ(x − xn (s)) (2) ±1 (it appears the sign is not known a priori).
N The phase space density of a bunch can be ef-
n=1
ficiently approximated in terms of the Eigen-FSM
where the xn (s) are independent and identically
for (3),
distributed random variables determined by (1).
Let p be the single particle probability density de- Ψ(s) := M(s, 0)W (6)
fined by (1), then ρN = p. Here, and in the fol- which was first introduced in [1] and generalizes
lowing, angular brackets will denote the expected the Courant-Synder formalism. It follows from
value of stochastic quantities. We will assume (5) and (P2) that
for large N that ρN (x, s) ≈ p(x, s), in a coarse ΨH (s)JΨ(s) = ıĨ (7)
69
Define Ψ̂ by Because of the non-resonant tune condition

(4), the averages D̄ and Ē are diagonal with
Ψ(s) =: Ψ̂(s) exp(ıωsN), ω := 2π/C (8)

where N = diag(ν1 , · · · , ν6 ), then Ψ̂ is C- D̄jj = λj := −ıγj ψ̂jH (s)JB(s)ψ̂j (s) (14)
periodic and (8) is a Floquet representation of the Ējj = b(s)|Ψ̂(s)5j |2 (15)
Eigen-FSM. Furthermore Ψ̂ satisfies (7). The k-
where the overbar denotes the s average of the un-
th column of Ψ̂ will be denoted by ψ̂k .
derlying periodic function.
c. Closed orbit and associated linearized SDE The solution of the IVP for (13) is
Since no characteristic multiplier, ρk , of (3) is
one, (1), with b = 0, has a unique C-periodic so- Va (s)jk = exp[(λj + λ∗k )s](V0 − Vae )jk
lution for sufficiently small. This closed orbit +Vae jk

solution is defined by the integral equation Vae jk = −δjk Ējj /2λj (16)
xco (s, ) = xco (s, 0) Here V0 is the initial condition for (12) and Vae
C
+ 0 G(s, t)g (xco (s + t, ), s + t)dt (9) is the unique equilibrium solution of (13).
In summary, the stochastic process x, de-
Here G(s, t) = (M(s, s + C) − I)−1 M(s, s + t) fined by (1), is given in a linear approximation
C
and xco (s, 0) = 0 G(s, t)c(s + t)dt. See Thm. by xco + yL and the moment matrix of yL is
2.1, p.154 of [8]. The closed orbit can be deter- U = ΨVΨH , where the averaging approxima-
mined approximately by iterating (9). tion to the moment matrix is
Let y := x − xco then for y small, y ≈ yL U(s) ≈ Ua (s) := Ψ(s)Va (s)ΨH (s) (17)
where yL satisfies the linear SDE
1 e. Dissipative case For λj < 0, the dis-
yL = (A(s)+B(s))yL + 2 b(s)ξ(s)e6 (10)
sipative case, Va (s) → Vae as s → ∞ and
Here B(s) := D1g (xco (s), s) is the Jacobian the approximate moment matrix in (17) becomes
matrix of g (·, s). The most important informa- Ua (s) = Ψ(s)Vae ΨH (s) = Ψ̂(s)Vae Ψ̂H (s).
tion about the bunch is contained in the moment Using (14), (15) and (16), this stationary, C-
(beam-envelope) matrix U(s) = yL (s)yLt (s). periodic Ua can be written
We now determine an approximation to U. ∗
Ua (s)jk = Gl [(ψ̂2l−1 (s))j (ψ̂2l−1 (s))k ]
d. Equation for moment matrix and averaging l=1,2,3
approximation The transformation yL → z via C
yL =: Ψ(s)z gives 2 1
Gl = − b(s)|ψ̂2l−1 (s)5 )|2 ds (18)
αl C 0
z = D(s)z + ξ(s)d(s)
1
2
−1
The quantities αl := 2λ2l−1 are called the

d(s) := b(s)Ψ (s)e6 damping constants and are given by
D(s) := Ψ−1 (s)B(s)Ψ(s) (11)
λ2l−1 = λ2l
Now U = ΨVΨH where V = z zH and the
γ2l−1 C H
differential equation for V is = [ψ̂j (s)JB(s)ψ̂j (s)]ds < 0
C 0
V = [D(s)V + VDH (s) + E(s)] (12) f. Robinson sum rule Since D and B are
where E(s) = d H (s).
d(s) related by the similarity transformation in (11),
TrD(s) =TrB(s) and thus the C-periodicity of
Applying averaging methodology [9, 6] to
B gives
(12) we obtain V(s) ≈ Va (s) where

1 C 2U0
Va = [D̄Va + Va D̄H + Ē] (13) TrD̄ = TrD1 g(xco (s), s)ds = (19)
C 0 E0
Here D̄ and Ē denote the average of the quasiperi-
where U0 and E0 are the energy gain in the cav-
odic functions D(s) and E(s), e.g.,
ity and beam energy respectively
(See [5]). From

1 L (14), λ∗2l = λ2l−1 so that 31 αl = TrD̄ and thus
D̄ = lim D(s)ds (19) is the Robinson sum rule [10].
L→∞ L 0
70
Ch.2: BEAM DYNAMICS
g. Remarks References
[1] A.W. Chao, J. Appl. Phys. 50, 595 (1979) and
1. The averages of E and D were computed EPAC 2008
under the nonresonance condition of (4). [2] A.A. Kolomensky, A.N. Lebedev, Theory of
However the standard averaging error bound cyclic accelerators, North-Holland, 1966
|V(s) − Va (s)| < O() for 0 ≤ s < O(1/) [3] H. Mais, G. Ripken, DESY 83-062 (1983)
requires a sufficient, O(1), separation be- [4] J. Jowett, SLAC-Pub-4033 (1986)
tween 0, ν1 , ν3 , ν5 , and 1/2. In the dissipa- [5] D.P. Barber et al, DESY M-94-09 (1994)
tive case, the s-interval of validity of the av- [6] J.A. Ellison, H. Mais, K. Heinemann, Details
eraging approximation can be extended to all of Orbital Eigen-analysis for Electron Storage
s ≥ 0. Details are given in [6]. The resonant Rings, in preparation (to be posted on arXiv).
case is considered in [11]. [7] K.R. Meyer et al, Introduction to Hamiltonian

Dynamical Systems and the N-Body Problem,
2. The mean of yL is easily handled and in
Springer, NY, 2009

the averaging approximation yL (s) ≈ [8] J.K. Hale, Ordinary Differential Equations,
Ψ(s) exp(D̄s)Ψ−1 (0)yL (0) and the co- Krieger, Florida, 1980
variance matrix is easily computed. Of [9] J.A. Sanders et al, Averaging Methods in Nonlin-
course, in the dissipative case the mean ap- ear Dynamical Systems, 2nd Ed., Springer, 2007;
proaches zero for large s. If yL (0) is a Gaus- J.A. Ellison, H.-J. Shih, AIP Conf. Proc. 326,
sian random vector, then yL is a Gaussian 1995
process. [10] K.W. Robinson, PR 111, 373 (1958)
[11] J. Wu et al, PAC 2005; B. Nash et al, PRST-AB
3. Spin-orbit motion in electron storage rings
9, 032801 (2006)
and especially spin diffusion due to quan-
[12] F. Ruggiero et al, Ann. Phys., 197, 396 (1990)
tum fluctuations can be treated approxi-
[13] K. Ohmi et al, PRE, 49, 751 (1994)
mately by introducing an 8-D matrix formal- [14] M. Sands, SLAC-121 (1970)
ism (Secs.2.6.7, 2.6.8). The codes SLIM [15] A. Wolski, PRST-AB 9, 024001 (2006)
and SLICK discussed in Sec.2.6.7 are based
on the orbital eigen-analysis of this section
and can be used to calculate beam polariza-
tion, as well as the orbital dynamics of this 2.2 OPTICS AND LATTICES
section.
2.2.1 Single Element Optics
4. A general formalism for treating the linear
K. Brown, Deceased 2002
electron beam dynamics with radiation ef-
fects taken into account is also presented Charged particle optics The following of a
in [12] and [13]. Whereas [12] starts from charged particle through a system of magnetic
a kinetic description (Fokker-Planck equa- lenses may be represented by matrix multiplica-
tion), [13] uses (as we do) the SDE for tion. At any position s measured along a refer-
the particle motion. The second order mo- ence trajectory, a charged particle is represented
ments (beam envelopes) are calculated di- by a vector (single column matrix) X(s) with
rectly (rather than using the orbital-FSM) Xt (s) = (x(s), x (s), y(s), y (s), l(s), ΔP/P0 ).
and these results are used in the computer where x = the horizontal displacement of a trajec-
code SAD. Furthermore generalized radia- tory (with respect to the central trajectory). The
tion integrals are derived which in the lim- x plane is typically defined as the midplane of
iting case of a completely uncoupled ma- a bending magnet; x = dx/ds is the angle this
chine reduce to the well known results of trajectory makes in the horizontal plane; y = the
Sands, [14] (Sec.3.1.4.1). This is also true vertical displacement; y = dy/ds is the vertical
in our case if we separate the six dimen- angle; l = the path length difference between the
sional dynamics into fast betatron and slow trajectory and the central trajectory as measured
synchrotron components via the dispersion from the beginning of the system, s = 0; ΔP/P0
(Sec.2.6.8). See [15] for a discussion of the is the momentum deviation of the trajectory from
beam envelope without radiation effects. that of the central trajectory. The central orbit is
defined as that trajectory whose initial conditions
71
Sec.2.2: OPTICS AND LATTICES
N S N
S Note that some of the R matrix elements have a
y
N N specific physical meaning,
y
y a a R21 = −1/Fx , R43 = −1/Fy (4)
g 0 x 0 x 0 x
w S S
where Fx,y are the focal lengths of the system in
S N S
the x and y planes; and R16 and R26 are the mo-
N mentum dispersion and the angular dispersion in
Dipole Quadrupole Sextupole
7–98
the x plane, respectively.
8355A256
Ray tracing using the R matrix The R matrix
Figure 1: Magnet profiles with midplane symmetry. allows one to trace individual trajectories (rays)
through any systems of magnets. To track a
are X(0) = 0. The momentum of the central trabeam of particles (assume Gaussian distribution)
jectory is P0 . Fig.1 shows a few magnet cross- through the system, the beam may be character-
ized by a “beam ellipse”. The equation of an
sections with midplane symmetry.

The vector X(0) at s = 0 is transformed to ellipse in n-D may be formulated by the matrix
another vector X(1) at position 1 by equation
X(1) = RX(0) (1) Xt Σ−1 X = 1 (5)
where R is a 6×6 matrix characterizing each where Σ is a positive definite n × n moment ma-
trix. For example, in 2-D,
magnet or drift distance.
All physics is contained in R representing x σ11 σ21
X= , Σ=
each type of magnet and R is derived from the x σ21 σ22
basic equations of motion of a particle passing =⇒ σ22 x2 − 2σ21 xx + σ11 x = detΣ (6)
2
through a static magnetic field. The traversing of
See Fig.2.
several magnets and interspersing drift spaces is
The area of the ellipse is
described by the same matrix equation (1) with a
total R given by Area = π(detΣ)1/2
R(s) = R(n) · · · R(i) · · · R(2)R(1) = πxmax xint = πxint xmax = π (7)
√ √
=⇒X(s) = R(s)X(0) (2) where σ11 = xmax , σ22 = xmax , √ = the
This formalism may be extended to second- and emittance of the beam, and r21 = σ21 / σ11 σ22
higher-order terms [1, 2]. is the correlation between x and x .
Eq.(1) can be expanded as X(s) = RX(0) = The transformation of the beam ellipse from
⎡ ⎤⎡ ⎤ position s = 1 to a position s = 2 in a beam line
R11 R12 R13 R14 0 R16 x(0)
is given by
⎢ R21 R22 R23 R24 0 R26 ⎥⎢ x (0) ⎥
⎢ ⎥⎢ ⎥ Σ2 = RΣ1 Rt (8)
⎢ R31 R32 R33 R34 0 R36 ⎥⎢ y(0) ⎥
⎢ ⎥⎢ ⎥
⎢ R41 R42 R43 R44 0 R46 ⎥⎢ y (0) ⎥
⎣ R R R R 1 R ⎦⎣ l(0) ⎦ The same equation applies for the y,√y plane
51 52 53 54 56
and also for the 6-D ellipsoid where σ11 =
0 0 0 0 0 1 ΔP/P0 √ √ √
xmax , √ σ22 = xmax√ , σ33 = ymax , σ44 =
For static magnetic fields, P0 and ΔP/P0 are con-
ymax , σ55 = lmax , σ66 = (dP/P0 )max , and
stants of the motion. The determinant det R = 1. √
rij = σij / σii σjj is the correlation between the
The 0 elements in row six occur because ΔP/P0
i and j variables in the 6-D phase space. As a
is a constant of the motion in static magnetic result of the fact that the Σ matrix is positive def-
fields. The 0 elements in column five are because
inite, the rij fall in the range −1 to +1.
x, x , y, y and ΔP/P0 do not depend upon l.
Courant-Snyder notation (See also Sec.2.1.1)
If all magnets in the system have midplane
The Σ matrix notation is rewritten
as
symmetry about y = 0, then in first order optics
σ11 σ21 β −α
the x- and y-motions are decoupled, and Σ= = T, T ≡ (9)
⎡ ⎤ σ21 σ22 −α γ
R11 R12 0 0 0 R16
⎢ R21 R22 0 0 0 R26 ⎥ with det T = 1. The equation of the ellipse
⎢ ⎥ (Courant-Snyder invariant) is then generated by
⎢ 0 0 R33 R34 0 0 ⎥
R=⎢ ⎥ (3) the matrix equation
⎢ 0 0 R43 R44 0 0 ⎥
⎣R R 0 0 1R ⎦ Xt T−1 X = , or γx2 + 2αxx + βx =
2
51 52 56
0 0 0 0 0 1 (10)
72
Ch.2: BEAM DYNAMICS
r21 xmax
X'
1 σ22 γ
slope = r σ11 = – α
21
x' max = σ 22 = γε σ
slope = r21 σ22 = – α
11 β
2 ε
x'int = σ22 (1–r21) = r21 x'max
β
x
σ11 = xmax = βε
Centroid
) = x int = γε
2
8–98 σ11 (1–r21
8355A255
Figure 2: A two-dimensional beam phase ellipse.

The area of the ellipse is A = π and the trans- the thick quad. Then, with L = thick quad length,
formation of the ellipse parameters from position Rthin quad, pp =
1 to position 2 is ⎡ ⎤
1 0 0 000
T2 = RT1 Rt (11) ⎢ −k sin kL 1 0 0 0 0⎥
⎢ ⎥
⎢ 0 0 1 0 0 0⎥
which gives the transformation of (β, α, γ). ⎢ ⎥ (14)
⎢ 0 0 k sinh kL 1 0 0 ⎥
⎣ 0 0 0 010 ⎦
R matrices for important optical elements
Drift space L = The length of the drift space. 0 0 0 001
⎡ ⎤ Note that for both the thick lens and thin lens,
1L0 0 00 R21 = −1/Fx and R43 = −1/Fy .
⎢0 1 0 0 0 0⎥
⎢ ⎥
⎢0 0 1 L 0 0⎥ Wedge bending magnet Rwedge bend =
Rdrift = ⎢ ⎥ (12) ⎡
⎢ 0 0 0 1 0 0 ⎥ 1
⎣0 0 0 0 1 0⎦ cos kx L kx sin kx L 0
⎢ −kx sin kx L cos kx L 0
0 0 0 0 01 ⎢
⎢ 0 0 cos ky L
⎢
⎢ −ky sin ky L
Thick lens quadrupole Rthick quad = ⎢ h 0 0
⎡ ⎤ ⎣ k sin kx L kh2 (1−cos kx L) 0
coskL sinkL
x x
k 0 0 00 0 0 0
⎢ −ksinkL coskL 0 0 0 0⎥
⎢ ⎥
⎢ sinhkL
0⎥ ⎤
⎢ 0 0 cosh kL k 0 ⎥ (13) 0 0 h
(1−cos kx L)
⎢ 0 ksinhkL coshkL 0 0 ⎥ kx2
⎢ 0 ⎥ h ⎥
⎣ 0 0 0 0 1 0⎦ 0 0 kx sin kx L ⎥
1 ⎥
0 0 0 0 01 ky sin ky L 0 0 ⎥
⎥ (15)
cos ky L 0 0 ⎥
This represents a quadrupole which is focusing ⎥
0 1 h2
3 (kx L − sin kx L) ⎦
in the x plane and defocusing in the y plane. kx
This R matrix is measured from the input face 0 0 1
of the quadrupole to the exit face. L = the ef- where ρ0 = the bending radius of the central tra-
fective magnetic length of the quadrupole; a = jectory; h = 1/ρ0 , kx2 = (1 − n)h2 , ky2 = nh2 ,
the radius of the aperture of the quadrupole; α = hL = the angle of bend of the central
B0 = the magnetic field strength at the radius a; orbit, L = the path length of the central orbit,
k 2 = (B0 /a)(1/Bρ), where (Bρ) is the magnetic n = −(dB/dx)(ρ0 /B0 ) is the normalized field
rigidity (momentum) of the central reference tra- gradient of the bending magnet measured on the
jectory, (Bρ) = 33.3564 P0 kG-m when P0 is in optical axis (x = 0, y = 0), R21 = −1/Fx ,
GeV/c. R43 = −1/Fy .
Thin lens quadrupole The two principal planes for Most of the time uniform field (n = 0,
a thin lens quad are both located at the center of kx = h = 1/ρ0 , ky = 0) magnets are used
73
x1 x2 Principal Plane
β1 β2 of Wedge Bending Magnet
R2 pp
R1
1
ρ0 = Central
h
Trajectory
4–98
7–98 8355A272
8355A257
Figure 4: Uniform wedge bend whose two principal

Figure 3: A bending magnet. The sign conventions
plans meet at the center of magnet.

for x, β, R and h are all positive as shown in the fig-
ure. Positive y is out of the paper. Positive β’s imply

2
transverse focusing. Positive R’s (convex curvatures) ψ = k(hg) 1+sin β
cos β , k is typically of the order 1/2
represent negative sextupole components of strength to 1, so if hg is less than 1%, this correction term
S = (−h/2R) sec3 β [1]. A wedge bending magnet can usually be ignored.
corresponds to β1 = β2 = 0.
Note that a pole-face rotation on a bending
magnet is equivalent to a thin quad located at the
for optical systems. Then for wedge magnets, input or exit face of the bending magnet having a
Runiform wedge bend =
⎡ ⎤ focal length 1/Fx = −h tan β in the x plane and
C S/h 0 0 0 (1 − C)/h 1/Fy = +h tan β in the y plane.
⎢ −hS C 000 S ⎥ For a bending magnet having rotated input
⎢ ⎥
⎢ 0 0 1L0 0 ⎥ and output pole faces, the total R Matrix for the
⎢ ⎥ (16)
⎢ 0 0 010 0 ⎥ magnet is
⎣ S (1 − C)/h 0 0 1 (α − S)/h ⎦
0 0 000 1 R = R(β2 )Rwedge bend R(β1 ) (19)
where C = cos α, S = sin α. If the R matrix is Solenoid (see also Sec.4.7.4) Rsol =
measured at the center of the wedge magnet, i.e. ⎡ 2 ⎤
SC
between the principal planes of the wedge magnet C2 K SC SK 0 0
⎢ −KSC C 2 −KS 2 SC 0 0 ⎥
(Fig.4), then it becomes simple: Here we separate ⎢ ⎥
⎢ −SC − S 2 C 2 SC ⎥
the Rx and Ry matrices and leave out the path ⎢ K 0 0⎥ (20)
⎢ KS 2 −SC −KSC C 2 0 0 ⎥
K
length terms, ⎢ ⎥
⎡ ⎤⎡ ⎤ ⎣ 0 0 0 0 1 0⎦
1 00 x
0 0 0 0 01
Rx,pp = ⎣ −hS 1 S ⎦ ⎣ x ⎦
0 01 ΔP/P0 where L = effective length of solenoid, K =
B0 /(2Bρ), B0 is the field inside the solenoid,
1L y
Ry,pp = (17) (Bρ) is the momentum of the central trajectory,
01 y
C = cos KL, S = sin KL.
Pole-face Rotation Matrix The first-order R ma- Rotating the transverse coordinates x and y
trix for a pole-face rotation as shown in Fig.3 is about the optical axis at the exit of the solenoid
R(β)⎡= ⎤ by an angle (−KL) decouples the x and y first-
1 0 0 000 order terms, i.e. Rrot (−KL)Rsol =
⎢ h tan β 1 0 0 0 0⎥ ⎡ ⎤
⎢ ⎥ C S/K 0 0 00
⎢ 0 0 1 0 0 0⎥
⎢ 0 0 −h tan(β −ψ) 1 0 0 ⎥
(18) ⎢ −KS C 0 0 0 0⎥
⎢ ⎥ ⎢ ⎥
⎣ 0 0 0 010 ⎦ ⎢ 0 0 C S/K 0 0 ⎥
⎢ 0 0 −KS C 0 0 ⎥
(21)
0 0 0 001 ⎢ ⎥
⎣ 0 0 0 0 1 0⎦
where β = angle of rotation of pole, 1/h = ρ0 = 0 0 0 0 11
bending radius of central orbit, g = total gap of
the dipole magnet, ψ = a correction term result- The focal length of a solenoid is 1/Fx =
ing from the finite extent of the fringing fields, 1/Fy = −R21 = −R43 = KS = K sin KL.
74
Ch.2: BEAM DYNAMICS
Beam rotation matrix Cylindrical multipole analysis In a current-

⎡ ⎤ free region the magnetic field B = ∇ψ can be
C 0 S 0 00
⎢ 0 described in terms of a magnetic scalar potential
C 0 S 0 0⎥
⎢ ⎥ ψ obeying ∇2 ψ = 0. For straight beam-line
⎢ −S 0 C 0 0 0⎥
Rrot = ⎢
0 0⎥
(22) elements (solenoids, quadrupoles, higher-order
⎢ 0 −S 0 C ⎥
⎣ 0 0 0 0 1 0⎦
multipole magnets, wigglers, and dipoles with
small sagitta) a cylindrical system with the z-axis
0 0 0 0 01
parallel to the magnet axis represents a natural
where α = the angle of rotation about the z axis coordinate choice yielding for ψ the repre-
∞
(the direction the beam is going), C = cos α, sentation ψ =
m=0 ψm,c (ρ, z) cos mφ +
S = sin α. For example a quadrupole rotated ∞
m=1ikz ψm,s (ρ, z) sin mφ, with ψm,s =
clockwise by an angle α about the z axis is dke bm (k)Im (kρ) and ψm,c being the
Rrot (−α)Rquad Rrot (+α) = Rrot. quad (23) same as ψm,s but with am (k) replacing bm (k).
The quantities Im are the modified Bessel

A skew quad is generated by setting α = ±45◦ . functions. The arbitrary functions bm and am
If the convention is that Rbend represents a measure the amplitude of the ‘normal’ and ‘skew’
magnet bending to the right, then a bend magnet field components. Purely solenoidal, dipole,
bending the beam up is given by the coordinate quadrupole, sextupole, · · · field components
rotation correspond to m = 0, 1, 2, 3, · · · , respect-
BEND UP = R(+α)Rbend R(−α) (24) ively. Equivalently we can write ψm,α (ρ, z) =
∞ [2] 2+m , where d
BEND DOWN = R(−α)Rbend R(+α) (25) =0 d,m Gm,α (z)ρ ,m = (−1) /
[22 !( + m)!], and α = s, c [1, 2, 3]. The
where α = 90◦ . The case α = 180◦ represents a [n]
magnet bending to the left in either of the above functions Gm,α (z), called the on-axis generalized
[n+1] d [n]
equations. gradients, obey Gm,α (z) = dz Gm,α (z), and
Accelerator matrix The transport matrix for an ac- uniquely determine ψ. The resulting magnetic
celerator section is listed in [3]. field for a normal pure multipole of order m ≥ 1
reads
∞
References Bρ = sin mφ (2 + m)d,m G[2] 2+m−1
m (z)ρ
[1] K. Brown et al, SLAC-75-rev-4 (1982); SLAC-91- =0
rev-2 (1977) ∞
[2] K. Brown, R. Servranckx, SLAC-PUB-3381 Bφ = cos mφ md,m G[2]
m (z)ρ
2+m−1
(1984) =0
[3] D.C. Carey, K.L. Brown, F. Rothacker, SLAC- ∞

R/95/462 (1995) p.179-182 Bz = sin mφ d,m G[2+1]
m (z)ρ2+m
=0
The expressions for the skew multipoles
2.2.2 3-D Multipole Expansion, Calculation follow from the replacements sin →cos and
of Transfer Maps from Field Data, cos → − sin. For pure solenoids Bz = ∞ =0 d,0
Fringe Fields [2+1] ∞ [2]
M. Venturini, LBNL G0 (z)ρ2 , Bρ = =1 2d,0 G0 (z)ρ2−1 ,
A. Dragt, U. Maryland Bφ = 0.
In a few cases, certain iron-free or rare-earth-
The behavior of orbits in charged-particle beam cobalt (REC) magnets [3, 4, 5], it is possible to
transport and accelerator systems can depend sen- determine the generalized gradients directly in an-
sitively on nonlinear fringe-field and high-order- alytic form or in terms of integrals over current or
multipole effects in the various beam-line ele- REC material distributions. When iron is present,
ments. Calculation of transfer maps that cap- the problem is more difficult. Then usually the
ture these effects requires a detailed and realistic available information is in the form of field data
model of the interior and fringe fields, including at various discrete points obtained either by mea-
their high spatial derivatives. We present methods surement or the use of 3-D finite element codes.
to determine field representations from available In the following we discuss how to determine the
field data that are suitable for this purpose. generalized gradients from field data.
75
(x)| ∼

Fitting methods A common past approach has |1/Im 2π|x|e−|x| for |x| 1. This
been to use on-axis or mid-plane 3-D field data rapid decay provides a powerful intrinsic smooth-
to fit some kind of profile function specified in ing effect that attenuates any uncorrelated er-
analytic form with various free parameters. Ex- rors/noise present in the field data [1, 2, 7]. The
amples of fitting profiles are the Enge functions smoothing increases with R, hence motivating the
[6], of the form 1/(1 + eSk (z) ), with Sk (z) be- use of field data on the largest radius cylinder that
ing a polynomial of order k in z. Other ex- can be accommodated within the aperture of the
amples include the following model functions device. Expressions similar to (1) can be worked
(corresponding to the exact fields generated by out for Bz or Bφ and may be more useful depend-
certain cylindrical sheet currents [3, 4]). For ing on circumstances. Variants employ sets of
solenoids G0 (z) = g[F0 (z + ) − F0 (z − )], dipoles
[1] orthonormal basis functions solving the Laplace
equation when cylinders with elliptic or rectangu-

G1 (z) = g[F1 (z + ) − F1 (z − )], and quadrupoles
[0]
lar cross sections are used. Their use improves the
G2 (z) = g[F2 (z + ) − F2 (z − )]; where F0 (t) =
[0]
smoothing property of the method for modelling

f1 (t), F1 (t) = 2f1 (t) − f3 (t), F2 (t) = 9f1 (t) − magnets (e.g. wigglers) with an aperture that is
8f3 (t) + 3f5 (t), with fn (t) = tn /(a2 + t2 )n/2 significantly larger in one direction [2, 7]. Analo-
and z ± = z ± L/2. The fitting parameters are gous methods are applicable to rf cavities [8].
the field/gradient strength g and the radius a and
length L of the cylindrical sheet current. Once More general surfaces In the case of curved
the parameters have been determined the profile elements with large sagitta, it is necessary to
functions are repeatedly differentiated as needed use more general surfaces (e.g. a bent box with
for map computation. However, differentiation straight ends). For geometries in which the
amplifies the unavoidable errors present either be- Laplace equation is not separable, it is possible
cause of imperfect fitting or noise in the field data. to obtain explicit expressions for interior fields
Use of these methods is therefore not expected to in terms of surface fields providing that all three
be reliable for calculations beyond first or perhaps components of the surface field are employed (vs.
second order. one field component required for geometries in
which the Laplace equation is separable, as in
Surface methods (circular cylinder) The
the previous discussion). One method is based
problem of noise and its amplification by dif-
on the Helmholtz decomposition theorem for a
ferentiation can be overcome with the use of
source-free vector field in a volume V bounded
surface methods. Suppose that the normal field
by a surface S [9, 10]. Let n̂(r ) be the out-
component Bρ (R, φ, z) is known on the surface
ward unit normal to S at the point r ∈ S. In
ρ = R of a virtual circular cylinder (in practice,
the values of the field on the surface are obtained
this decomposition B(r) = ∇χ + ∇ × A with
χ(r) = (4π)−1 S dS [n̂(r )·B(r )]/|r −r | and
by interpolation of the field data on nearby grid
points). The cylinder, long enough to include A(r) = −(4π)−1 S dS [n̂(r )×B(r )]/|r−r |.
the fringe regions, is to be contained within A second method, which employs only a vector
the aperture of the magnetic device and aligned potential so that B(r) = ∇×A, utilizes the Dirac
with the z-axis. This surface information is monopole vector potential and is suitable for a
sufficient to uniquely determine the full magnetic Hamiltonian formulation [2, 7, 11]. It is based on
field within the surface. From Bρ (R, φ, z) the
decomposition A = An +At where An (r) =

S dS [n̂(r ) · tB(r )]K (r, r ) and A (r) =
n t
we compute the quantities B̂ρ,m,s (k) =
∞
−ikz 2π dφ sin(mφ)B (R, φ, z)/(2π)2 S dS ψ(r )K (r, r ). Here ψ(r ) is the scalar
dze
−∞ 0 ρ
potential on S, and the kernels K n and K t are
with a similar expression for B̂ρ,m,c (k) with sin given by the relations K n (r, r ) = {n̂(r )× (r −
[n]
replaced by cos. Then the Gm,α (z) are given in r )}/{4π|r − r |[|r − r | − n̂(r ) · (r − r )]} and
terms of surface data by the relations K t (r, r ) = [n̂(r )×(r−r )]/[4π|r−r |3 ]. Both
∞ methods are intrinsically smoothing and yield an-
[n] in kn+m−1
Gm,α (z) = m dk B̂ρ,m,α (k)eikz alytic interior fields. The second yields interior
2 −∞ Im (kR) fields that satisfy the Maxwell equations exactly
(1) regardless of field data quality and integration ac-
The integrand in (1) decays rapidly at large curacy.
|kR| because of the Bessel function property
76
Ch.2: BEAM DYNAMICS
Hard-edge fringe-field maps The accurate the design-orbit rigidity, and ± refers to entrance
treatment of high-order fringe-field effects is and exit. Numerical evaluation of the map for a
complicated. They can be computed reliably if in- REC quadruple in the limit of zero inner radius
terior fields are known in analytic form, as is the verifies that the exact f4 does indeed approach
case for iron-free or REC magnets, or if surface the hard edge limit, and it does so fairly rapidly.
methods are used. In some few cases, hard-edge This example shows that, for quadrupoles, there
models provide a useful first approximation. Sup- are cases where the hard-edge approximation can
pose the transfer map for a magnet is written in be usefully employed even when the aperture is
the factorized Lie form M = R exp(: f3 :) exp(: quite large.
f4 ) · · · (Sec.2.3.6) and that some model is pro-
posed for the fringe-field transition from no field References
outside to some s-independent field inside. With

such a model one may study the factors of M in [1] M. Venturini, A. Dragt, NIM-A 427, 387 (1999)
the limit in which the transition region shrinks to [2] A. Dragt, Lie Methods for Nonlinear Dynamics
zero. Do the various factors then approach def- with Applications to Accelerator Physics (2010);
inite finite limits, and are the limits model inde- www.physics.umd.edu/dsat/
pendent? It can be shown that, for solenoids, R [3] M. Bassetti, C. Biscari, PA 52, 221 (1996)
and f3 do take on limiting values with these de- [4] P. Walstrom, NIM-A 519, 216 (2004)
sirable properties. However, the f4 and higher- [5] A. Dragt et al, MaryLie Manual (2003);
order polynomials become infinite. In the case of www.physics.umd.edu/dsat/
[6] H.A. Enge, Rev. Sci. Instrum. 35, 278 (1964)
dipoles, R and f3 also take on finite and model-
[7] C. Mitchell, A. Dragt, PRST-AB 13, 064001
independent limiting values. However, some of
(2010)
the f4 take on finite model independent values,
[8] D. Abell, PRST-AB 9, 052001 (2006)
some take on finite but model dependent values, [9] R. Plonsey, R.E. Collin, Principles and Appli-
and some become infinite. Thus, in the case cations of Electromagnetic Fields, McGraw-Hill
of solenoids and dipoles, there is no meaningful (1961)
hard-edge fringe-field approximation beyond sec- [10] S. Manikonda, M. Berz, NIM-A 558, 175 (2006)
ond order; detailed knowledge of the field pro- [11] A. Dragt et al, arXiv:1012.1647 (2000)
file is essential to go beyond second order. For [12] E. Forest, Beam Dynamics, Harwood (1998)
the case of quadrupoles it can be shown that R, [13] G. Lee-Whiting, NIM 83, 232 (1972)
f3 , and f4 take on finite and model-independent
limiting values. Thus, for quadrupoles, there is
a meaningful hard-edge approximation through 2.2.3 Lattices for Collider Storage Rings
third order. The limiting behavior of still higher- E. Keil, CERN
order multipoles is also known [12]. However,
use of such information would seem warranted Symbols α, β, γ Courant-Snyder parameters,
only if all other relevant beam-line elements are K = (1/Bρ)(dBy /dx), f focal length of
treated accurately, including their fringe-field ef- quadrupoles, Lp Length of whole cell, μ Phase
fects, through the same order. advance per cell, ϕ Bending angle of whole cell,
In the dipole case the limiting form of R is z Transverse coordinate x or y.
built into most codes and is referred to as edge fo- Lattice matching A lattice typically is com-
cusing [12]. The hard-edge form of f3 for dipoles posed of unit cells and insertions. The points
is available as an option in MaryLie [5]. No de- where cells and insertions meet require lattice
tailed studies have been carried out to see how matching.
well the hard-edge approximation holds in the Insertions transport the beam from a point
case of realistic dipoles, but such studies should with one set of α, β, D, D to another point with a
now be possible with the use of surface methods. different set of these parameters. Phase advances
The hard-edge form of f4 for quadrupoles through the insertion may also be constrained.
is given by f4± = ±(k/12)(y 3 py − x3 px + Splitting an insertion into modules helps, with
3x2 ypy − 3y 2 xpx ), and its use is equivalent to each module achieving part of the matching.
Lee-Whiting’s earlier result [12, 13]. Here k = Matching is done by matching programs
Q/Bρ where Q is the quadrupole strength, Bρ is (Sec.1.7), which adjust quadrupole and dipole
77
strengths, drift space lengths. The number of vari- 5

ables ≥ that of conditions. 4
(a)
Separated-function FODO cell Two quads,

3
β /Lp
with equal and opposite f , assumed to have van-
±
ishing length, separated by length Lp /2. Spaces 2
between quadrupoles are filled by homogeneous-
field dipoles. Lp applies to both focusing and 1
bending. Assume that edge focusing in dipoles 0
is negligible (when ρ Lp ) and that ϕ 1. In
terms of μ (assumed the same in x- and y-planes) 4 (b)
and Lp [1],
D /L p ϕ
3
Lp
±
f =± (1) 2
4 sin μ2
Lp (1 ± sin μ2 ) 1
β± = (2)
sin μ 0
0 0.1 0.2 0.3 0.4 0.5
± ∓1 − sin μ2 10–97
8355A37 μ /2π
α = (3)
cos μ2
Figure 1: (a) β ± /Lp vs μ/2π, (b) D± /Lp ϕ vs μ/2π.
±Lp ϕ[1 ± 12 sin μ2 ]
D = (4)
4 sin2 μ2 1.0
2.0
The sign of α applies to the quadrupole entrances.
1.6 0.8
At the exits, α has the opposite sign. The up- βy
per (lower) signs apply to the focusing (defocus-
βx,y L p
D L pϕ
1.2 0.6
ing) quadrupole in the plane under consideration.
0.8 D 0.4
From the focusing quadrupole (s = 0) to the de-
focusing quadrupole (s = Lp /2), 0.4 βx 0.2
μ
2s(1+sin ) 4s2 μ
β(s) = β+ − cos μ
2
+ Lp tan 2 (5) 0 0
0 0.2 0.4 0.6 0.8 1.0
2
μ 10–97 s/Lp
2s sin s2 ϕ
D(s) = D + 1 − Lp
2
+ 2Lp (6) 8355A39
Figure 2: βx /Lp , βy /Lp , and D/Lp ϕ vs s/Lp for the

See Figs.1 and √2. β + /Lp has a minmum near FODO case μ/2π = 0.25.
μ/2π = sin−1 ( 5−12 )/π ≈ 0.21. The dispersion
averaged over the FODO cell is
# $ per cell are

Lp ϕ 1 1 ϕ5 1− 34 sin2 μ2 + 60
1
sin4 μ2
D = − (7) I5,FODO = (10)
4 sin2 μ2 12 4Lp sin2 μ2 sin μ
# $
The momentum compaction factor is (see also ϕ2 4
Sec.2.3.11) I6,FODO = 1+ (11)
2Q sin2 μ2
# $
ϕD ϕ 2 1 1 where Q = quad length.
αc = = − (8)
Lp 2 sin2 μ2 12 In a flat ring, the dispersion
√ function often
scales approximately with β,
The natural chromaticity per FODO cell is
αc R
1 μ D(s) ≈ βx (s) (12)
ξFODO = − tan (9) νx
π 2 where 2πR is the ring circumference and νx is the
If a storage ring consists of n FODO cells, it has a horizontal tune of the ring.
total natural chromaticity ξ = nξFODO . The syn- The average β- and dispersion functions are
chrotron radiation integrals I5 and I6 (Sec.3.1.4) given approximately by, using Eqs.(2) and (4) for
78
Ch.2: BEAM DYNAMICS
relatively small μ, Dispersion suppressor insertion These inser-

tions match D without affecting α, β. The idea
βx,y ≈ R/νx,y , D ≈ R/νx2 (13)
is to bring D from the value in the FODO cells
where 2πR = nLp , and νx,y = nμx,y /2π. It to zero by launching a forced oscillation of D
follows from Eq.(13) that around a value ∼ 12 of that in the FODO cells. (i)
If π/μ = integer, D is suppressed by π/μ FODO
αc ≈ 1/νx2 , γt ≈ νx (14) cells with bending angle ϕ/2, using Eq.(4). (ii) It
(For nonlinear momentum compaction factor for is possible to suppress D within two FODO cells,
a FODO lattice, see Sec.2.3.11.) if their bending angles are chosen as [3]
# $
Combined-function FD cell Two combined- 1 ϕ
function dipoles of length Lp /2 each, with equal ϕ1 = ϕ 1 − , ϕ2 = (19)
and opposite strengths K, and no drift spaces. 4 sin2 μ2 4 sin2 μ2

√ √
Lp K Lp K where ϕ1 and ϕ2 are bending angles of the FODO

cos μ = cos cosh cell on the side where D = 0, and D = 0, respec-
2 2
tively. The betatron oscillations are unchanged
L4p K 2 L8p K 4
=1− + + · · · (15) in both styles of dispersion suppressor. When
96 645120 μ = π/2, both styles coincide. For μ ≤ π/3,
Equating the leading two terms in Eq.(15) to the there are no reversed dipoles in the dispersion
leading two terms of cos μ yields (overestimating suppressor. Both styles can start at horizontally
K by <10% for phase advances μ/2π ≤ 1/4) [2] focusing or defocusing quadrupoles [4]. How-
√ ever, they both work for just one phase advance.
4μ 3 (iii) It is possible to suppress D by just modifying
K≈ (16)
L2p the focal length of two quadrupoles. In this case,
α, β and μ are also modified. (iv) In hybrid dis-
β ± at the centers of the F and D quadrupoles are
persion suppressors, some bending magnets are
(< 10% error for μ/2π ≤ 1/4)
# $ left out, and D is halved by halving the product
√
Lp π 3 μ Lp ϕ, using Eq.(4). By giving them enough free
±
β ≈ 1± (17) parameters, they can be matched to FODO cells
sin μ 2 2π
over a range of phase advances [5].
while (underestimating D ± by <13% for μ/2π ≤ Low-beta insertion These insertions reduce the
1/8) # β-functions to small values, assuming D = 0 al-
√ $
Lp ϕ μ 3 ready. They are used in colliders to achieve small
±
D ≈ 2 1± (18) beam sizes at the IPs. Lattice matching requires
μ 8
four or six free parameters, typically quadrupole
See Fig.3. gradients, depending on whether only two α’s and
two β’s, or whether also two μ’s are matched.
Usually αx = αy = D = D = 0 at the IP.
2.0 2.0
Most critical are the quadrupole doublet or triplet
βy near IP.
1.6 1.6
The β-function near a waist is given by
D / L pϕ
D
β x,y Lp
1.2 1.2
β(s) = β0 + (s − s0 )2 /β0 (20)
0.8 βx 0.8 where β0 is the β-function at waist position s0 .
0.4 0.4
Doublets are often used in e+ e− colliders
with flat beams, the first quadrupole from the IP
0 0 focuses vertically, the second horizontally. In thin
0 0.2 0.4 0.6 0.8 1.0 lens approximation, their focal lengths d1 and d2
10–97 s/Lp
8355A38
are found by imposing point-to-parallel focusing
Figure 3: Computed (no Taylor series approximation) in both planes,
√
βx /Lp , βy /Lp and D/Lp ϕ vs s/Lp for the combined s1 s2 √ √
FD case μ/2π = 0.125. d1 = ± √ , d2 = ∓ s2 s1 +s2 (21)
s1 +s2
79
where s1,2 are the drift spaces from the IP Φ > 1. Synchrotron radiation from magnets near
to the first quadrupole, and between the two IP should not get into detector in two-ring collid-
quadrupoles, respectively. The two solutions cor- ers with at least one e± beam. Cf. [8, 9] for re-
respond to interchanging x and y. In LEP, the views of high-luminosity e+ e− colliders.
point-to-parallel approximation overestimates the Separation schemes (Sec.2.5.4.1) depend on
focal length by < 8%, and the thin-lens approxi- whether the beams have equal or opposite charge
mation by another 12%. and equal or different momenta. In a pp collider
Triplets are often used in pp and pp̄ collid- like LHC [10], beams are in a common vacuum
ers with round beams. Assuming a symmetrical chamber, and focused by common quadrupoles
triplet with the first and third quadrupole having between separating dipoles on either side of IP.
the same strength and equi-distant from the cen- In e+ e− B factories like KEK-B and PEP-II,
tral quadrupole, and in thin lens approximation, and ep colliders like HERA [11], different mo-
the focal lengths t1 and t2 are found by imposing menta of beams are used to achieve separation.
point-to-parallel focusing in both planes, In e+ e− colliders like BEPC-II [12] and DAΦNE
with beams of opposite charge and equal mo-
−r + s2 (2 s1 2 + 3 s1 s2 + 2 s2 2 ) mentum, the beams cross horizontally at an an-
t1 = ± gle (Sec.2.5.2). The crossing angle between the
2 (s1 + s2 )
beams is enhanced by the common first horizon-
r − s2 2 s1 + 5 s1 s2 + 2 s2 2
2
tally defocusing quadrupole, such that the sec-
t2 = (22)
4 (s1 + s2 ) t1 ond horizontally focusing quadrupoles can be in-
1 stalled side-by-side.
2
r = s22 (2s21 +3s1 s2 +2s22 )2 −8s21 s32 (s1 +s2 ) Crossing angles θ are used in almost all two-
ring colliders to avoid head-on beam-beam col-
where s1,2 are the drift spaces from the IP to the lisions (Sec.2.5.4.1) other than those at the IP.
first quadrupole, and between the quadrupoles, re- Long-range beam-beam collisions still occur in
spectively. In LHC (Version 4.2) [6], the point- the common vacuum chamber at every half bunch
to-parallel approximation overestimates ti by < spacing near the IP. The larger θ, the smaller the
8%, and the thin-lens approximation by another long-range kicks and tune shifts. In LHC [10]
< 36%. crossings are vertical and horizontal in diametri-
cally opposite IPs. Separation at long-range colli-
Two-ring Colliders Two rings are needed in sions is about 7σ⊥ . See also Sec.2.5.4.
colliders of particles with the same sign of charge, Bunch overlap at Φ > 0, improved by crab
e.g. LHC and RHIC, and in colliders of particles crossings and crab waists, increases L. In a crab
with opposite charge, but different momentum. crossing [13], used in KEK-B [14], a transversely
Since luminosity (Sec.2.5) L is proportional to deflecting RF cavity in each beam causes an or-
beam current I, two rings are also needed in “fac- bit distortion that varies along the bunch. Hence,
tories” of particles with opposite charge and high the bunches have a correlation at the IP between
L, when I bunch current, limited by collec- offset in the crossing plane and longitudinal po-
tive effects (Sec.2.4). The rings must have equal sition along the bunch such that θ is cancelled.
circumferences [7]. Arcs often consist of FODO The crab waist [15] in DAΦNE (Sec.2.5.2) uses
cells and dispersion suppressors, described above, the fact that a test particle with an offset in the
and sometimes of separated function lattices with plane of crossing encounters the centre of the op-
bending period length a multiple of focusing pe- posite bunch at a distance from the IP along the
riod length. Two-ring interaction regions consists beam that is proportional to the offset. Sextupoles
of low-β insertions, described above, for each satisfying conditions on phases and strengths are
beam, possibly with common elements, and sep- used to optimize the geometric overlap between
aration schemes that funnel the two beams from the two colliding bunches, and more importantly,
separate arcs through the interaction point IP with to mitigate synchrobetatron resonances driven by
low β-values and back into separate arcs. Beams beam-beam collisions with Φ = 0 (cf. Sec.2.3.4).
often cross at an angle 2θ, often measured in units
of Piwinski parameter Φ = θσs /σ⊥ (Sec.2.3.4)
References
with rms bunchlength σs and rms radius σ⊥
in crossing plane. Large crossing angle means [1] E. Keil, CERN 77-13 (1977) 11
80
Ch.2: BEAM DYNAMICS
[2] H. Bruck, Accélérateurs circulaires de particules, tron coupling, avoids beam size increase due to
Presses universitaires de France (1966) energy spread and results in the largest possible
[3] E. Keil, CERN 77-13 (1977) 22 reduction of the equilibrium emittance due to ra-
[4] R.H. Helm, SLAC-PUB-3278 (1984) diation emitted in insertion devices. All early
[5] LEP Design Report, Vol.II, CERN-LEP/84-01 3rd generation light sources employed achromat
(1984) lattices. Over time, however, nearly all of the
[6] The Large Hadron Collider: Conceptual Design, early 3rd generation rings moved away from the
CERN/AC/95-05(LHC) (1995) achromatic condition resulting in lower equilib-
[7] K. Hirata, E. Keil, Phys. Lett. B232 (1989) 413 rium emittances [3]. In most cases the dispersion
[8] Y. Funakoshi (ed.), ICFA Beam Dynamics in the straights is still small enough that insertion
Newsletter No.31 (2003) devices continue to reduce the overall equilibrium
[9] M.E. Biagini (ed.), ICFA Beam Dynamics
emittance and the effective emittance is smaller

Newsletter No.48 (2009)
than in the achromatic case.
[10] LHC Design Report, Vol.1, CERN-2004-
003 (2004) http://lhc.web.cern.ch/lhc/LHC-

The wavelength λ of radiation emitted by an
DesignReport.html undulator is given by

[11] M. Seidel, F. Willeke, EPAC00 (2000) 379 λu K2
[12] J.Q. Wang et al, PAC07 (2007) 53 λ= 1 + (1)
2nγ 2 2
[13] R.B. Palmer, Snowmass Conf. High Energy
Physics in the 1990s, 1988, Ed. S. Jensen, World where λu is the undulator period, γ is the rel-
Scientific (1989) 642 ativistic Lorentz factor, K ≈ λU Be/(2πm0 c)
[14] Y. Funakoshi et al, EPAC08 (2008) 1893 (the “undulator parameter”), B is the magnetic
[15] http://arxiv.org/ftp/physics/papers/0702/0702033. field, and n = 1, 3, 5, ... denotes the harmonic.
pdf The term λu /(2γ 2 ) shows that one can arrive at
the same wavelength using higher beam energies
2.2.4 Lattices for Low-Emittance Light and longer periods, or lower beam energies and
Sources shorter periods. Many parameters enter in the
C. Steier, LBNL evaluation of the optimum beam energy, includ-
Third generation light sources have been ex- ing cost, natural emittance, intrabeam scattering,
tremely popular since the first ones went into op- beam instabilities, heat load on optics, magnetic
eration around 1992. They are based on lepton material properties, and the desired photon wave-
storage rings, in most cases using electron beams, length range. In general, even with advanced un-
in very few cases positron beams. The fundamen- dulator technology, low energy rings (∼2 GeV)
tal properties of these facilities are high beam cur- are unchallenged at low photon energies (below
rents (≥ 100 mA), small emittances (≤ 10 nm 20 eV) and provide excellent performance up to
horiz.), moderate beam energy spreads (≤ 10−3 ) 4 keV, intermediate energy rings (∼3 GeV) from
and the use of insertion devices (mostly undula- a few 100 eV to above 10 keV, and, if higher pho-
tors) in long (∼5 m) straight sections to produce ton energies are needed, higher electron energies
high brightness (Sec.4.2) photon beams. Origi- are necessary (∼4.5 - 7 GeV).
nally there were two main lattice types that were Brightness and coherence The spectral photon
optimized to meet these requirements: the double brightness, B(λ) (Sec.3.1.5), of light emitted at
bend achromat (DBA [1]), first used in the NSLS wavelength λ from an insertion device in a syn-
at BNL, and the triple bend achromat (TBA [2]), chrotron light source is
first developed for the ALS at LBNL. Later on, F (λ)
facilities evolved further by detuning the achro- B(λ) = (2)
mat lattices to allow dispersion leakage into the (2π)2 σTx σTy σTx σTy
straight sections. Newer light sources also use where F (λ) is the photon flux and σTx , σTy , σTx ,
multi-bend achromats, as well as damping wig- and σTy are the wavelength dependent convolu-
glers to reduce the emittance further. tions of the respective electron and photon beam
sizes and divergences. For a given flux, maximum
2.2.4.1 Lattice choices
brightness is achieved when the electron beam
Traditionally the straights were designed with emittance is reduced toward and beyond the in-
zero dispersion, which minimizes synchrobeta- trinsic diffraction-limited emittance λ/(4π) of the
81
photons and when the electron beam beta func-

tion is close to the equivalent beta function of the
diffraction ellipse.
Coherence is a measure of the degree to
which the radiation can exhibit interference pat-
terns. The fraction fcoh of photon flux at wave-
length λ that is transversely coherent is related to
the ratio of the intrinsic photon emittance to the
total emittance of the photon beam,
Fcoh,T (λ) λ/(4π) λ/(4π)
fcoh = = (3) Figure 2: Example of a TBA lattice with distributed
F (λ) σTx σTx σTy σTy

dispersion: ALS at LBNL.
Flux, brightness, coherent flux, and coherent
fraction are not the only figures of merit for a syn- The minimum emittance of a DBA lattice is
chrotron light source, but they are always impor- given by
tant considerations. Which of the four quantities 1 Cq γ 2 θb3
is most important depends upon each particular DBA,min = √ (6)
4 15 Jx
experiment. where θb is the bending angle per magnet. Lat-
Equilibrium emittance The horizontal emit- tices of facilities in operation are detuned sig-
tance x results from the combined effect of nificantly from these minimum values, to make
an excitation of horizontal betatron oscillations optimization of nonlinear effects easier (see be-
(Sec.3.1.4), Sx , and its damping, τx , low). Instead of striving for minimum emittance
in an achromatic condition, further optimization
1 Jx Cγ E 3 1 Jx Cγ E 3
x = Sx τx , = ds = I2 , (4) was achieved by allowing dispersion to leak into
τx 4πT0 ρ2 4πT0
the straight sections, thereby lowering the disper-
γx ηx2 +2αx ηx ηx +βx ηx
2
γ 2 I5
x = Cq , I5 = ds (5) sion in the arcs. One example of a distributed dis-
Jx I2 |ρ3 |
persion lattice is shown in Fig.2. The ALS lattice
with ρ the bending radius, Jx the horizontal is a triple bend achromat which is operated with
damping partition number. The integrals along s ηx = 6 cm in the straights. The use of horizon-
are nonzero only in bending magnets and inser- tally defocusing gradients in the dipole magnets
tion devices, while they are zero in drift spaces allows for more compact lattices and also shifts
and negligible in quadrupoles and sextupoles. In the damping partition numbers, further reducing
most machines the contributions of the insertion the horizontal emittance at the expense of the lon-
devices to Sx and τx are negligible and the es- gitudinal damping.
sential contribution comes from the bending mag- How low one can get the natural emittance is
nets. A small emittance requires the use of a lat- related to the bending angle of individual mag-
tice with small ηx and βx in the bending magnets. nets. Therefore newer rings with lower emittances
The first lattices to achieve this and to provide generally have larger circumferences for a given
the space for insertion devices were the double beam energy, resulting in more unit cells. Alter-
and triple bend achromats [1, 2]. The principle natively, one can segment the bending into more
is shown in Fig.1. magnets per unit cell and refocus both βx and
η between each pair of successive bending mag-
Achromat Symmetry Point Achromat Symmetry Point
nets. The resulting lattice is called a multiple bent
Dispersion Function
Dispersion Function achromat (MBA). For MBAs, the arcs usually re-
semble the so called theoretical minimum emit-
tance structure [7]. The emittance from such a
1/2 Insertion Straight 1/2 Insertion Straight 1/2 Insertion Straight 1/2 Insertion Straight lattice scales asymptotically as
E2
x ∝ 3 (7)
Achromat
Insertion Symmetry Point
Achromat
Insertion Symmetry Point N
where N is the number of bending magnets. Fig.3
Figure 1: Principle of double and triple bend achromat shows the lattice structure for a seven bend achro-
lattices. mat lattice (Max-IV).
82
Ch.2: BEAM DYNAMICS
Achromat Symmetry Point
Beyond correction of the chromaticity, while
maintaining maximum dynamic and momentum
aperture, lattice designers must also verify: (i)
sensitivity of the dynamic (momentum) aperture
to magnet imperfections or other lattice symme-
Dispersion Function try distortions; (ii) sensitivity to insertion de-
1/2 Insertion Straight 1/2 Insertion Straight vices as well as optimum compensation schemes;
(iii) beam lifetime, in particular the influence
of Touschek scattering; (iv) intrabeam scattering
(Sec.2.4.12), which will be a very substantial ef-
Achromat
Insertion Symmetry Point fect in the latest light sources.
Figure 3: Example of a multi-bend achromat lattice, in 2.2.4.3 Systematic lattice optimization

this case a seven bend achromat. techniques

Historically, lattice design depended strongly on
2.2.4.2 Chromaticity correction and the experience of the designer often choosing
nonlinear optimization from a menu of known lattice choices. Many
The low emittance design of light sources in- choices in this approach are subjective and usu-
herently relies on strongly focusing quadrupoles. ally the optimization of the nonlinear dynamics
These quadrupoles generate large chromatic aber- properties is carried out separately from the linear
rations that need to be corrected with sextupoles. lattice design.
The sextupoles in turn generate geometrical and One newer technique to find the globally opti-
nonlinear chromatic aberrations, exciting nonlin- mal lattices in terms of emittance or brightness is
ear resonances that can make the motion of the GLASS [4]. It uses a global grid scan of the few
electrons unstable. This inherent feature has been parameters of a simplified standard cell. It then
the major design challenge from the beginning analyzes the properties of all lattices that were sta-
and continues to be a dominant topic for lattice ble. This allows to search for potential lattices
upgrades as well as new rings. Many techniques with certain properties. Another technique in-
have been developed to accurately predict the dy- volves multi-objective, genetic algorithms. Those
namic aperture (Sec.2.3.9) as well as the dynamic algorithms are fairly old, but applications to ac-
momentum aperture and to allow their optimiza- celerators are new. The first major one involved
tion, thereby maximizing injection efficiency and the optimization of a high brightness DC photoin-
Touschek lifetime (Sec.2.4.12). The techniques jector. More recently they have been used for de-
include design optimization tools, as well as mea- tailed linear lattice design and for multi-parameter
surement methods useful to optimize existing ac- optimization of the nonlinear dynamics of com-
celerators based on beam based measurements. plex or low-periodicity lattices. They also allow
In terms of lattice designs, the trend has been for the simultaneous optimization of linear and
to include more quadrupole families (the newest nonlinear lattice properties [5].
lattices include 6-10 quadrupole families) to al- These new techniques are now used for many
low full control of the linear lattice, and more sex- purposes, including the optimization of special in-
tupole families (state-of-the-art is 5-10 families) sertions (e.g. low beta straights, convergent beta
to allow control of detuning with amplitude terms straights, fs-slicing facilities, crab cavity inser-
as well as resonance driving terms, or a compre- tions), development of lower emittance upgrades
hensive optimization of the dynamic (momentum) for existing facilities, optimization of linear and
aperture. Many rings have each magnet powered nonlinear lattices for new facilities, and the eval-
by an individual power supply to allow flexibil- uation of candidate designs for ultimate storage
ity and improved correction of insertion device rings.
effects. The newest trends are the addition of
2.2.4.4 Evolution of light source lattices
weak octupole magnets to manipulate detuning
with amplitude terms directly as well as the use 3rd generation light sources have seen a steady
of more relaxed lattices combined with damping evolution in the >15 years since their initial op-
wigglers. eration. This does not just include new and more
83
Table 1: Advanced storage ring facilities that are in early operation, under construction, or under study [8].
Project Energy Circ. x Curr. Lattice Design Status
[GeV] [km] [nm] [A]
PETRA-III 6 2.3 1 0.1 7/8 FODO + 1/8 DBA + DW In operation
NSLS-II 3 0.792 0.6 0.5 30 x DBA + DW Under const.
MAX IV 3 0.528 0.24 0.5 20 x 7BA + DW Under const.
USRLS 7 2 0.3 0.5 50 x 4BA Design study
XPS7 7 2.2 0.08 1.0 80 x 6BA Design study
PEP-X 4.5 2.2 0.1 1.5 DBA + TME + DW Design study
Tsumaki 6 1.44 0.07 0.1 20x10BA,4x5BA,4xLSS+DW Design study
USR7 7 3.1 0.015 0.2 40 x 10BA Design study
advanced facilities, but also continuous improve- in low energy rings. Other examples are the fs-
ments of existing ones. Lattice developments slicing facilities, that use horizontal or vertical
have evolved into two main directions. One is to dispersion manipulation to provide spatial sepa-
allow for reduction of the natural emittance. The ration of energy modulated bunch slices to gen-
other is to incorporate complex local lattice inser- erate fs duration x-ray pulses. Intentional sym-
tions or breaks of the global lattice symmetry, to metry breaks were also introduced to allow for
enable radiation sources with improved character- individual straight sections to be lengthened, en-
istics, while minimizing the negative effects on abling longer undulators, as well as for straight
the global nonlinear beam dynamics. Some ex- sections with smaller (or convergent) beta func-
amples of those developments are listed below. tions, to optimize the photon beam properties for
Advances to enable lower equilibrium emit- specific beamlines. The newest proposal is a crab
tances started with the move to distributed dis- cavity insertion, which allows for two sets of crab
persion lattices. Later on, the addition of more cavities with precise control of phase advance and
sextupole families and most recently the inclu- local coupling between them. Combined with
sion of weak octupoles help to control the non- proper beamline optics, this will allow to generate
linear dynamics and enabled lattices closer to the high brightness, high repetition rate x-ray pulses
minimum emittance. Multi-bend achromats al- with ps duration.
low for smaller emittances for a given number of
unit cells at the expense of a smaller fraction of 2.2.4.5 Ultimate storage rings
available straight section space. More recently, While storage rings are a “mature” technology,
damping wigglers have seen a resurgence for light they nevertheless still have the potential for sig-
sources. They allow to use a more relaxed lat- nificantly further enhanced performance. One can
tice in the arc cells, resulting in larger momentum imagine an “ultimate” storage ring [9, 10] that
compaction factors and less lattice related nonlin- produces high-brightness, transversely coherent
ear dynamics challenges while still achieving ex- x-rays. For such a source to maximize transverse
tremely small emittances. However, they use up photon coherence, the beam emittance must be
space, require more expensive rf and cooling sys- extremely small in both transverse planes, around
tems, increase the equilibrium energy spread and the wavelength-dependent diffraction limit. Stor-
use up straight section space. Furthermore, non- age ring sources have achieved diffraction lim-
linear dynamics challenges due to wigglers can be ited emittances for hard x-rays in the vertical
significant. So to determine the optimum use of plane by minimizing beam coupling, but the hor-
damping wigglers requires a careful design trade- izontal emittance must be reduced by a factor of
off study. Tab.1 lists lattice parameters of the 100 or more from the lowest values achieved to-
newest light sources in early operation or under day to reach that limit. Ultimate rings would
construction, and of design studies for future fa- necessarily have large circumferences (compare
cilities. Tab.1) to reach low emittance values. Some of the
Examples of modifications to provide special main ideas are the use of more bending magnets,
photon sources include Superbends [6], a cost ef- smaller physical apertures to allow for higher gra-
fective way to provide hard x-rays with moder- dients of magnets, operation with large (or full)
ate brightness using low beta function locations coupling to alleviate intrabeam scattering effects
84
Ch.2: BEAM DYNAMICS
and possible on-axis injection with beam replace- the Hamiltonian, Hik = ∂ 2 H/∂xi ∂xk . In the
ment to mitigate small dynamic apertures. Ulti- absence of longitudinal magnetic field, θx = x
mate rings would have brightness and coherent and θy = y are the horizontal and vertical angles;
flux one or two orders of magnitude higher than otherwise θx = x − Ry/2, θy = y + Rx/2,
the highest performance ring-based light sources where R = eBs /P c, Bs is the longitudinal mag-
in operation or under construction. netic field and P is the total momentum. For a flat
horizontal orbit,
References ⎡ 2 2 ⎤
K +k+ R4 0 N − R2
⎢ 0 1 R ⎥
[1] M. Sommer, LAL/RT/83-15 (1983) H=⎢ 2 2 0 ⎥ (2)
⎣ 0 ⎦
2 −k+ 4
[2] A. Jackson, PA 22 (1987) 111 R R
N
[3] L. Farvacque et al, EPAC 1994 −R/2 0 0 1
[4] D. Robin et al, PRST-AB 024002 (2008) where K = eBy /P c, k = eG/P c, N =

[5] L. Yang et al, NIM A 609 (2009) 50; M. Borland,
eGs /P c, By is the vertical component of mag-

ICAP 2009
netic field; G and Gs are the normal and skew
[6] D. Robin et al, NIM A 538, 1-3 (2005) 65
(tilted by +45◦ around the s axis) components of
[7] S.Y. Lee, L. Teng, PAC 1991, p.2679
the magnetic field gradient. Note that a quad tilted
[8] K. Balewski et al, PETRA III TDR, DESY 2004-
035; J. Ablett et al, NSLS-II CDR, BNL 2006;
by an arbitrary angle can be represented by a lin-
M. Eriksson et al, EPAC 2008; A. Ropert et ear combination of normal and skew quads. For
al, EPAC 2000; M. Borland, NIM A 557, 230 any two solutions of Eq.(1), xt1 Ux2 = const.
(2005); R. Hettel et al, PAC 2009; K. Tsumaki, These constants are called the Lagrange invari-
N. Kumagai, NIM A 565, 394 (2006); M. Bor- ants.
land, Proc. SRI 2009 Let M(0, s) be a transfer matrix: x(s) =
[9] M. Bei et al, NIM A, 622(3) (2010) 518 M(0, s)x(0). Conservation of the Lagrange in-
[10] Y. Cai, NIM A645,168 (2011); Y. Cai et al, variant yields that M is symplectic:
PRST-AB 14, 054002 (2012) Mt UM = U ⇔ MUMt = U (3)
This leaves only 10 independent parameters for 16
2.2.5 Betatron Motion with Coupling of Two matrix elements. In a ring, its revolution matrix
Degrees of Freedom M has four eigenvalues λi and eigenvectors vi ,
V. Lebedev, A. Burov, FNAL Mvi = λi vi , i = 1, . . . , 4 (4)
A stable betatron motion requires all eigenvalues
Betatron coupling was initially considered as an
to be confined to a unit circle, none of them equal
undesired effect, and efforts were made to sup-
to ±1; thus, the eigenvalues are split into two
press it. However, over recent three decades the
complex conjugated pairs: λ3 = λ∗1 , λ4 = λ∗2 , and
betatron coupling has become an intrinsic part
similarly the eigenvectors. For any two normal-
of several proposed and built accelerators [1]-[4].
ized eigenvectors, condition (3) yields their sym-
(See also Sec.2.1.3.) Presently, two different ba-
plectic orthonormality,
sic representations are mostly used. The first was
proposed by Edwards and Teng [5, 6] and the sec- vl† Uvl = −2i , l = 1, 2
(5)
ond by Mais and Ripken [7, 8]. The Mais-Ripken vk† Uvm = 0 , k = m = 1, . . . , 4
representation was later extended by Lebedev and
Bogacz [9], where the equivalence of the two ba- where † stands for the Hermit conjugation. The
sic representations was established. choice of Eq.(5) normalization constants makes
Motion equations and symplecticity condition the matrix
For a linear Hamiltonian system with two degrees V = [Rev1 , −Imv1 , Rev2 , −Imv2 ] (6)
of freedom, equations of motion can be written in symplectic and selects one (prime) eigen-vector
the following matrix form: from each pair.
Eigen-vectors and ellipsoid in 4-D phase space
dx J 0 0 1
= UHx , U = , J= (1) Turn-by-turn particle positions and angles can be
ds 0J −1 0
describedas
Here s = βct is the conventional distance used
xn = Re A1 e−i(ψ1+nμ1 ) v1 +A2 e−i(ψ2 +nμ2 ) v2
as time variable, x = (x, θx , y, θy )t is a vector
of canonical variables, H is a Hessian matrix of = VSn ξA (7)
85
where four real parameters, A1 , A2 , ψ1 and ψ2 Eq.(14), the emittances and the eigenvectors can
represent the betatron amplitudes and phases, and be found from the matrix of moments Σ,
⎡ ⎤
cos μ1 sin μ1 0 0 det (ΣU+iεl I) = 0 ; (ΣU+iεl I) v̂l = 0 (17)
⎢− sin μ1 cos μ1 0 0 ⎥
S=⎣
0 0 cos μ2 sin μ2 ⎦ The mode emittances ε1 and ε2 are the motion in-
0 0 − sin μ2 cos μ2 variants, i.e. they cannot be changed in the course
⎡ ⎤ of linear Hamiltonian motion. In particular, if ini-
A1 cos ψ1 tial and final states are uncoupled, the only possi-
⎢−A1 sin ψ1 ⎥
ξA = ⎣
A2 cos ψ2 ⎦
(8) ble emittance exchange is an emittance permuta-
tion. This statement is correct for three degrees of
−A2 sin ψ2 freedom as well.
The 3-D surface inscribing an ensemble of par- For circular machines, the eigen-vectors vi
ticles, whose turn-by-turn coordinates are con- are determined by the revolution matrix, Eq.(4).
tained in a 4-D ellipsoid, can be determined by For transfer lines, they are determined by the
the bilinear form phase space ellipsoid of the incoming beam, as
given by Eqs.(14) or (17). The downstream eigen-
xt Ξx = 1 (9)
vectors and phase advances can be found with
It is related to the mode emittances ε1 and ε2 by Eq.(7).
[9] Mais-Ripken parameterization Symplecticity
t
Ξ = UVΞ̂V Ut (10) leaves only 8 independent real parameters for the
where eigenvectors of stable betatron motion. An exten-
sion [9] of Mais-Ripken (MR) parameterization
Ξ̂ = diag (1/ε1 , 1/ε1 , 1/ε2 , 1/ε2 ) (11)
[7] suggests the form
and the function diag(. . .) forms a diagonal 4 × 4 t
√
matrix. The inversion of Eq.(10) is v1 = β1x ,− i(1−u)+α
√ 1x
, β1y e √ 1y eiν1
iν1 ,− iu+α
β1x β1y
Ξ̂ = Vt ΞV t
(12) √
v2 = β2x eiν2 ,− iu+α√ 2x eiν2 , β2y ,− √ 2y
i(1−u)+α
i.e. a symplectic transform V reduces matrix Ξ to β2x β2y
its diagonal form. The beam emittance is a prod-
Here βix,y , αix,y are generalized Courant-Snyder
uct of the semi-axes of the ellipsoid described by
functions; three other real functions, u and ν1,2 ,
Ξ̂ so that
−1/2 −1/2 can be expressed in terms of the Courant-Snyder
ε4D = Ξ̂11 Ξ̂22 Ξ̂33 Ξ̂44 = det Ξ̂ functions. The transfer matrix M(s1 , s2 ) between
two points can be expressed through the eigenvec-
= (det Ξ)−1/2 = ε1 ε2 (13) tors of these points, V(s1 ) and V(s2 ):
where we omitted the factor π 2 /2
correcting for M(s1 , s2 ) = −V(s2 )SUV(s1 )t U
the actual 4-D volume of the ellipsoid. The = V(s2 )SV(s1 )−1 (18)
quadratic form Ξ determines the emittances and
eigenvectors, where the matrix S is given by Eq.(8). Since all
matrices in Eq.(18) are symplectic, this equation
Ξ − iε−1
l U vl = 0 ; det Ξ − iε−1
l U =0 yields a symplectic transfer matrix for any incom-
(14) ing and outgoing eigenvectors, Consequently, the
For a Gaussian distribution transformation M(s1 , s2 ) can always be imple-
−1
f (x) = 4π 2 ε1 ε2 exp −xt Ξx/2 (15) mented with conventional linear optics elements.
and the second-order moments are Edwards-Teng parameterization Edwards-
Teng (ET) parameterization [6] is based on
Σij ≡ xi xj = xi xj f (x)dx4 a canonical transform which reduces a 4 × 4
−1
revolution matrix M to its block-diagonal normal
= VΞ̂ Vt = Ξ−1 ij (16) modes form,
ij
Consequently, a symplectic transform VU re- −1 A0
M = RM̂R , M̂ = (19)
duces matrix Σ to its diagonal form. Similarly to 0B
86
Ch.2: BEAM DYNAMICS
Matrices A and B are parameterized similar to of the particle canonical angular momentum with-
the uncoupled motion. Thus, matrix M̂, describ- out breaking the rotational symmetry of the en-
ing particle motion in new coordinates, is param- tire ensemble [11]. It was found by Derbenev [1]
eterized by six real parameters. A symplectic ma- that a conventional planar beam state, with β2x =
trix R is parameterized as β1y = α2x = α1y = u = 0, can be transformed
into a round beam and back by proper optical
I cos ϕ D−1 sin ϕ schemes, called him as beam adapters. (See also
R= (20)
−D sin ϕ I cos ϕ Sec.2.1.3) To illustrate the main idea, let eigen-
states of the incoming beam be represented by two
where I is the unit 2 × 2 matrix, and D is a 2 × 2
circular modes described above with αlx,y = 0,
symplectic matrix. Consequently, ten parameters
and βlx,y = β. Let this beam undergo an un-
fully describe the transfer matrix M. They are
coupled transformation T, which horizontal part

the six parameters of the decoupled motion, β1,2 ,
is the identity transformation I, and vertical part
α1,2 , μ1,2 , Teng’s phase φ, and three parameters
gives 90◦ phase advance,

of matrix D. Their relation to the MR parameters
√
can be found in [9], in particular sin ϕ = ± u,
I 0 0 2β
β1,2 = β1x,2y / (1 − u), α1,2 = α1x,2y / (1 − u). T= ; N=
−(2β) −1 0
0 N
For ET, the phase advances of the betatron motion
are defined as It is straightforward to check that this transforma-
v̂i (s)e−iμi (s) = M̂(0, s)v̂i (0) (21) tion converts two circular modes into two planar
modes tilted by 45◦ . Since the initial beam state
where v̂i (s) are the decoupled eigenvectors in the is axially symmetric, the transfer line tilt by 45◦
normal mode coordinates. Comparison of Eq.(19) rotates the modes making them conventional pla-
with Eq.(18) shows that ET and MR present the nar modes. The simplest adapter implementation
same eigen-modes with different sets of parame- requires a system of three quadrupoles [10].
ters. Therefore their phase advances are identical.
Usually optical codes support an analysis of Perturbation theory Let the revolution matrix
coupled optics using either MR or ET formalism: M be perturbed: M = (I + P)M0 , where the
MAD8 [13] supports MR and ET; MAD-X [14] perturbation P is small, but not necessarily sym-
supports ET, and OptiM [15] supports MR. plectic. In the first order of perturbation theory,
Round beams and beam adapters Some the complex tune shifts are [16]
applications require round (axially-symmetric) Δμl /(2π) = −(4π)−1 vl† U P vl , l = 1, 2
beams. These beams are conveniently described (23)
in a basis of circular modes, which are given by In particular, this formula allows calculation of
Eq.(18) with βlx = βly = β, αlx = αly = α, the incoherent beam-beam and space charge tune
u = 1/2, and ν1,2 = ±π/2. In a matched shifts for arbitrary-coupled optics [17]. It also
solenoid β = 1/R and one of modes is a Lar- yields that a sum of the two growth rates is in-
mor motion with center at the solenoid axis, and dependent on the eigenvectors,
another one is a pure offset, x, y = constant. A
product of mode emittances is equal to the total 4- Im(Δμ1 + Δμ2 ) = Tr(P)/2 (24)
D emittance (see Eq.(13)), and their difference is
This statement is sometimes called as a rate-sum
the beam angular momentum xθy −yθx . If a lin-
theorem.
ear transformation M preserves axial symmetry, it
also preserves an absolute value of the canonical Coherent motion There is a general rule re-
angular momentum of every particle. The general ducing a problem of coherent motion for arbi-
form of this transformation is trary coupled optics to uncoupled one. If the
growth rates are smaller than the tune difference
A 0
M = R(θ) (22) the two problems are analytically identical for any
0 ±A beam distribution function, rf configuration and
where R(θ) describes a rotation with an arbitrary impedances Zx,y , with the following rule of cor-
angle in both x, y and θx , θy subspaces, and respondence [18]:
A is an arbitrary 2 × 2 symplectic matrix. A βx,y Zx,y → βlx Zx + βly Zy ; l = 1, 2 (25)
transformation cannot change an absolute value
87
Coupling adjustments also affect the chromatici- proposed in [22] reduces the tune chromaticity in
ties of betatron modes and their space charge tune vicinity of integer resonance resulting in an in-
shifts; all these factors make beam stability non- crease of the cooling channel momentum accep-
trivially dependent on coupling [19]. tance.
Acceleration Hereinabove, no acceleration is
Round beams for circular colliders were pro-
assumed; however, acceleration is not actually an
posed [23, 24] and successfully implemented
obstacle to the described formalism. Indeed, let
[25, 26] to mitigate the beam-beam effects. If the
P0 and P be the initial and final momenta. Then
revolution map in the interaction point (IP) is ro-
the transfer matrices are symplectic in the vector
tationally invariant, than the eigen-vectors are ro-
space x̃ = P/P0 x, making the above formal- tationally invariant too (β1x = β2x = β1y = β2y ,
ism applicable to the case of beam acceleration. α1x = α2x = α1y = α2y , u = 1/2); and the
Applications Applications of coupled optics betatron tunes are equal to μ1,2 = μ ± θ, where
formalism and the planar-circular adapters in-
μ is the betatron phase advance for matrix A in

clude relativistic electron cooling, round beams Eq.(22). In this case, the angular momentum
for circular colliders, flat electron beams for linear is preserved in the course of beam transport.
colliders, low-energy hadron storage rings, and For θ = 0 the fractional parts of both tunes
ionization cooling of muons [11, 12]. ν1,2 = μ1,2 /(2π) are equal, allowing to avoid
dangerous beam-beam resonances. Note that for
FNAL Electron cooling Because of very high
electron beam energy, the beam transport in Fer- θ = 0 the map rotational invariance does not
milab electron cooler uses conventional optics automatically results in the beam being round.
The degeneracy of eigenmodes yields that they
elements instead of normally used longitudinal
can be chosen circular or planar with arbitrary
magnetic field [2]. To keep the electron beam
rotation angle. In the first case the beam is round
focused in the cooling section and minimize its
effective temperature, both the cathode and the for any ratio of the mode emittances, while in
the second case only for equal mode emittances.
cooling section are immersed in the longitudinal
If the beam is not round, the beam-beam col-
magnetic field, so that the magnetic fluxes through
lisions do not preserve rotational symmetry,
the beam at the cathode and cooler are identi-
cal. It requires the beam transport from cathode to and the axially-symmetric optics does not bring
solenoid to be rotationally invariant, Eq.(22). The significant benefits for beam-beam effects sup-
pression. In reality, the block-diagonal x and y
cooling section is 20 m long, and the total length
components A in the map (22) cannot be exactly
of beam transport (from cathode to collector) is
identical; differences in the corresponding Twiss
∼100 m. In addition to solenoids, the beam trans-
port includes 10 dipoles, which do not provide parameters βx,y and phase advances μx,y remove
degeneracy and make modes planar for θ = 0. To
an axially-symmetric transformation. However,
provide mode circularity, the solenoidal coupling
with appropriate strengths and locations of focus-
angle θ has to exceed uncoupled mismatch,
ing solenoids, the transport map from the cathode
to the cooling section and back to the collector is |θ| |μx − μy | , |1 − βx /βy |
rotationally invariant. Matching electron and an- To get round beams in IP of an e+ e− collider,
tiproton beam radii in the cooler imposes an addi- R. Talman suggested the Möbius revolution map
tional requirement to the cathode-cooler electron [27]
supply line [11]. 0 −Y
M= (26)
Neutrino factories and muon colliders require 3- X 0
D ionization cooling of muons. It cannot be where X and Y are 2 × 2 matrices with arbitrary
achieved without coupling between degrees of phase advances μx and μy . Experimental investi-
freedom. To maximize the cooling range and ef- gation of that lattice at CESR showed that it is not
ficiency, most schemes use solenoidal focusing as successful as the axially-symmetric optics [25].
which introduces x-y coupling at the fundamen- Indeed, for μx = μy the map (26) does not belong
tal level. To couple the transverse and longitudi- to the rotation-invariant group (22) and does not
nal degrees of freedom, one needs to excite the preserve the angular momentum. Although the
dispersion, i.e. to make a curvilinear beam or- two emittances are equal for that lattice, advan-
bit [20, 21]. The resonance dispersion excitation tages of the rotation-invariant optics do not apply.
88
Ch.2: BEAM DYNAMICS
Instead, the axially-symmetric Möbius map, μx = 2.3 NONLINEAR DYNAMICS

μy , βx = βy , αx = αy , preserves angular mo-
mentum, so from that point of view it should be 2.3.1 Hamiltonian
as good as uncoupled axially-symmetric optics. K. Symon, U. Wisconsin
The axially-symmetrical Möbius corresponds to
θ = π/2; thus, the tunes are different by 12 .
2.3.1.1 General case
Although the beam-beam resonance suppression
may be as good as for θ = 0, it still requires
an experimental verification. The Möbius op- The Hamiltonian for a particle of charge e, mass
tics also requires a redesign of a chromaticity m, moving in an electromagnetic field described
correction. em , φ is [1, 2, 3, 4,
by electromagnetic potentials A
6, 7]
References
Ht (x, px , y, py , s, ps ; t) = eφ + T (1)
[1] Ya. Derbenev, U. Michigan, UM HE 93-20
(1993); UM HE 98-04 (1998)
where
[2] S. Nagaitsev et al, PRL 96, 044801 (2006)
[3] R. Talman, PRL 74, 1590 (1995)
[4] R.C. Fernow et al, Proc. AIP 372, 146 (1995) T = c m2 c2 + (px − eAx )2 + (py − eAy )2
[5] L.C. Teng, Fermilab, FN-229 0100 (1971)
[6] D.A. Edwards, L.C. Teng, IEEE Trans. Nucl. Sci. 2 1/2
ps − eAs
20, 3, p.885 (1973) + = γmc2 (2)
[7] I. Borchardt et al, DESY 87-161 1 + ρ−1 x
[8] F. Willeke, G. Ripken, AIP Conf. Proc. 184
(1989) is the kinetic energy. The coordinates are in Fig.1.
[9] V. Lebedev, S. Bogacz, e-print JLAB-ACC-99- The canonical vector potentials are
19 (1999), JINST, v.5 (2010)
[10] A. Burov, V. Danilov, Fermilab-TM-2043 (1998)
Ax = x̂ · A
em
[11] A. Burov et al, PRST-AB 3, 094002 (2000)
[12] A. Burov, S. Nagaitsev, Ya. Derbenev, PRE 66, Ay = ŷ · A
em

016503 (2002) x
[13] http://mad.web.cern.ch/mad/mad8web/mad8.html As = 1 + ŝ · A
em (3)
ρ
[14] http://mad.web.cern.ch/mad/madx manual.pdf
[15] http://www-bdnew.fnal.gov/pbar/
organizationalchart/lebedev/OptiM/optim.htm For a circular accelerator, all quantities are peri-
[16] A. Burov, PRST-AB 9, 120101 (2006) odic functions of s.
[17] A. Burov, Ya. Derbenev, Fermilab-PUB-09-392- With s as the independent variable the Hamil-
AD (2010) tonian is
[18] A. Burov, V. Lebedev, PRST-AB 10, 044402
(2007)
x
[19] R. Cappi, E. Métral, D. Möhl, CERN/PS 2001- Hs (x, px , y, py , t, −E; s) = −eAs − 1 + Ps
010 (AE) (2001) ρ
[20] Ya. Derbenev, R. Johnson, PRST-AB 8, 041002
(4)
(2005) where −E is the momentum conjugate to the co-
[21] R.C. Fernow, R.B. Palmer, PRST-AB 10, 064001 ordinate t and
(2007) 2
[22] Y. Alexahin, AIP Conf. Proc. 1222, 313 (2010)
Ps = E−eφ
c − (px − eAx )2
[23] S. Krishnagopal, R. Siemann, PAC’89, p.836
[24] V. Danilov et al, EPAC’96, p.1149 1/2
[25] S. Henderson et al, PAC’99, p.410 − (py − eAy )2 − m2 c2 = βs γmc (5)
[26] D. Berkaev et al, ICFA Beam Dyn. Newslett. 48,
235 (2009)
[27] R. Talman, PRL 24, 1590 (1995) is the kinetic momentum in the s direction.
89
Sec.2.3: NONLINEAR DYNAMICS
y
The canonical potentials for magnetic lattice x
elements, expanded to 4th order in x, y, are:
y
=0, s
straight section: A (6)
x
s
= ŝ − Bx + B 1 sin α B
bend: A − x 2
ρ
2 2ρ

B1 sin α 2 B1 sin α B2 sin2 α
− y + − x3
2 3ρ 6 Figure 1: Coordinates Relative to Reference Orbit: x

B2 sin2 α B1 sin α is distance measured horizontally outward, y measured
+ − xy 2
2 2ρ vertically, s measured along the reference orbit from

B3 sin3 α B2 sin2 α some reference point, and ρ is the local radius of curva-
+ − x4 ture of the reference orbit, which lies in the x-s plane.
24 8ρ

B2 sin2 α B3 sin3 α 2 2 B3 sin α 4 defined near the reference orbit by:
+ − x y + y
2ρ 4 24
= ŷB(s)
bend: B (13)
B1 cos α 2 B2 sin α cos α 2
+x̂ − y + xy = ŷ[B(s) + G(s)x]
grad. bend: B (14)
2 2
= ŷQ(s)x + x̂Q(s)y
B3 sin2 α cos α 2 2 B3 cos α 4 quad: B (15)
− x y + y (7)
4 24 skew quad: B = x̂Qs (s)x − ŷQs (s)y (16)

gradient bend: A = ŝ − Bx − B + G x2 = ŷ S(s) (x2 − y 2 )
sextupole: B
2ρ 2 2
+x̂S(s)xy (17)
G 2 G G
+ y − x3 + xy 2 = x̂O(s)(y 3 − 3x2 y)
octupole: B
2 3ρ 2ρ
# $
−ŷO(s)(x3 − 3y 2 x) (18)
B B G
2
G
+ + − + y4
24Bρ 12ρ 24 24ρ2 In Eq.(7), α(s) is the angle between the local tan-
gent to the reference orbit and the perpendicular
B 2 B G
+x̂ − y + − xy 2 to the magnet edge. The parameters in Eq.(7) are
2 2ρ 2
given by the field and its gradients on the refer-
B G B G
− 2
− x2 y 2 + + y4 (8) ence orbit,
2ρ 2ρ 24 24ρ B
B1 = (19)
cos α
quad: A = ŝ Q (y 2 − x2 ) B B sin α
2 B2 = − (20)
cos α ρ cos3 α
2
Q 4 Q
+ (x − 6x2 y 2 − y 4 ) − x̂ xy 2 (9) B 3B sin α
48 2 B3 = 3
− (21)
cos α ρ cos4 α

B (1 + 2 sin2 α) B 2 sin α
skew quad: A = ŝ Qs xy − Qs x3 y + 2 5
−
6 ρ cos α Bρ cos4 α
3 2
Eq.(8) is for a bend with a gradient. Primes denote
y x y
+x̂Qs − (10) derivatives with respect to s and appear in edge
6 2
corrections.
sextupole: A = ŝS 1 xy 2 − 1 x3 Edge corrections depend on the edge design.
2 6 Eq.(7) assumes a straight edge (no field parallel

1 2 2 1 4 to the edge and no dependence of the fields on
+x̂S − x y + y (11)
4 24 the coordinate parallel to the edge). It is accu-

= ŝO 1 x4 − 3 x2 y 2 + 1 y 4 (12) rate if edge fields do not extend past the magnet
octupole: A
4 2 4 center. For a thin edge perpendicular to the refer-
ence orbit, α = 0. Eq.(8) assumes the magnet
where B(s), G(S), Q(s), S(s), O(s) are the ver- contours run perpendicular to the reference or-
tical field and its gradients on the reference orbit, bit at each point, so that in the median plane the
90
Ch.2: BEAM DYNAMICS
field gradient depends only on s. The formulae and we have rescaled px , py , HsT to be 1/p (not
for multipoles assume the edge fields preserve the 1/p0 ) times their values in Eq.(4). The electric
multipole symmetry about the reference orbit. For potential φ has been dropped. Space charge can
other edge designs, these formulae should give be included by adding
approximate values of the coefficients, particu-
x 1 λg(x, y)
larly if the edges are thin, but they do not include As = − 1+ −f (28)
any terms which violate the assumed symmetries. ρ γ2 4π0 cβ
The potential for an accelerating gap may be which is not the canonical potential, but gives the
written, to second order, space charge field including the φ term under the
t % conditions of Eqs.(23)-(25).
=− Δ 2
A V (t)dt ŝ Δ(s) − (x + y 2 ) The expansion of Eq.(26) to fourth order in
4
x, y, px , py is

& −2
Δ Δc V̇ 2
− (x̂x + ŷy) + ŝ (x + y 2 ) (22) 1 2 1 2 x p0

2 4 HsT = p + p − 1−
2 x 2 y ρ p
where V (t) is gap voltage; Δ(s) describes the
4
field shape with gap dsΔ(s) = 1. Usually, one eAsk k eB 2
− x y + y px
p 2p
approximates Δ(s) = δ(s − sgap ). The terms k+=2
in x, y are edge field corrections appropriate to a x(p2x + p2y ) e2 B 2 4
circular gap perpendicular to the reference orbit; + + y
2ρ 8p2
they are usually neglected.
eB eAx12 eAx21 2
The potentials for space charge fields are + − xy 2 px − x ypx
2pρ p p
(1 − f )λ eAx03 3 1
φ= g(x, y) (23) − y px + (p2x + p2y )2
4π0 p 8
= ŝ βλ (1 + x )g(x, y)
A (24) 2
+Csc (x + y ) +2 Csc 3
(x + xy 2 ) (29)
4π0 c ρ ρ
where λ is charge line density, f is the neutraliza- where p0 = eBρ is the momentum of the ref-
tion fraction. Time rates of change are assumed erence particle, Ask and Axk are the coeffi-
slow relative to light transit time across the vac- cients multiplying xk y in the appropriate terms
uum chamber. For a uniform cylindrical charge of of Eqs.(7)-(12), and
radius a, centered in a circular perfect conductor
of radius b, the function g(x, y) is (r 2 = x2 + y 2 ) 1 eλ
Csc = − f (30)
' 2
γ2 4π0 a2 βp
1+2 ln ab − ar 2 if r ≤ a
g(x, y) = b2 (25) Only linear space charge terms have been in-
ln r2 if a ≤ r ≤ b cluded.
a and b may be functions of s and t. Some other
cases are found in [2, 3, 5]. 2.3.1.3 Longitudinal motion
The Hamiltonian, to second order in w = E − Es
2.3.1.2 Transverse motion [2] is
From Eqs.(4)-(12), we obtain the Hamiltonian for K 2
transverse motion: HsL (ϕ, w; s) = w +eV̂ (cos ϕ+ϕ sin ϕs )Δ(s)
2
eAs (31)
HsT (x, px , y, py ; s) = − Subscript s refers to the synchronous energy (e.g.
p
1 Es , ps , φs , γs ), K = cβωrf3 Eηs , V̂ is gap voltage am-
x eAx 2 2
plitude, and
−(1 + ) 1 − px − − py
2
(26)
ρ p
Δ(s) = δ(s − sj ) (32)
where the kinetic momentum is j
2 1
E − m2 c4 2 For a linac, η = −γ −2 .
For a circular accelerator,
p= (27) η = γt−2 − γ −2 , sj = sgap + 2πRj.
c2
91
The Hamiltonian (31) implies the difference ps is treated as a constant and k(s) = ρ12 − pQ0 .
equations We assume gradients Q(s), like bend fields B(s),
ϕj+1 = ϕj + (sj+1 − sj )Kj wj (33) change slowly with time in proportion to p0 (t).
The Hamiltonian for synchrobetatron cou-
wj+1 = wj + eV̂j+1 (sin ϕj+1 − sin ϕs,j+1) pling is

For a linac, if V̂ and the distance Δs between h2 1 D
gaps is a smooth function of s, we may replace HSBC (φ, W, x, px ; s) = − W2
2R2 ps γs2 ρ
Δ(s) by its average value,

1 eV̂ eV̂s hD hD
HsL (ϕ, w; s) = Kw2 + (cos ϕ + ϕ sin ϕs ) − Δ(s) cos φ − x+ px
2 Δs ωrfs R Rps

hD
1 eωrf g0 eV̂s hD
− −f λ(ϕ) (34) − φ− x+ px sin φs
γ2 4π0 βc 2πRωrfs R Rps
1 2 kps 2
λ(ϕ) is line charge density at the point in the beam + p + x
2ps x 2
with phase ϕ, and g0 = g(0, 0) = 1 + 2 ln ab .
For a circular accelerator, with time as the in- +HNLS + HNLB + HNLC + HSVC (41)
dependent variable, the smoothed Hamiltonian is
where h = harmonic number, φs is defined by
1 eV̂ ω0
HtL (ϕ, w; t) = βcKw2+ (cos ϕ+ϕ sin ϕs ) dEs eV̂s
2
2π =− sin φs (42)
1 eωrf g0 ds 2πR
− −f λ(ϕ) (35) and the voltage across the accelerating gap is
γ2 4π0
t
where V̂ is the maximum voltage gain per turn. V (t) = V̂ (t) sin ωrf (t)dt (43)
0
2.3.1.4 Synchrobetatron coupling
We have (i) assumed Eq.(32), (ii) assumed the
The complete Hamiltonian for synchrotron and bend edges ⊥ the reference orbit (α = 0), (iii)
horizontal betatron motion is [8, 1] included quadrupoles, sextupoles, and straights,
Hs (x, px , t, −E; s) (iv) kept nonlinear terms up to third degree in
W, x, px , φ − φs , and the time derivatives of the
1/2
x px 2 slowly varying quantities ps , V̂ . The nonlinear
= −eAs − p 1 + 1− 2 (36) synchrotron term is
ρ p # $
where p is given by Eq.(27), x is the horizontal h3 D 2 hm2 ωrfs
2
D
HNLS = + − 1
distance measured from the reference orbit, and 2R3 p2s 2Rp4s ρ
px is the corresponding momentum. 3 3

Make a canonical transformation to syn- eSh D
+ W3 (44)
chrotron, betatron variables: 6R3 p3s
t
h(s − sgap ) The cubic terms in the betatron variables are
φ=− ωrf (t)dt +
0 R eS 3 1
HNLB = x + xp2 (45)
hD hD 6 2ps ρ x
+ x− px (37)
R Rps The nonlinear coupling terms are
E − Es
W = (38) eShD 2
ωrf HNLC = x W
2Rp
hD # s $
x=x− W (39) eSh2 D2 2
ωrfs m2 h2 D 2
Rps + + + xW 2
hD 2R2 p2s 2p3s ρ 2R2 ps ρ
px = px − W (40)
R hD h D
+ xpx W + − 1 p2x W
Subscript s denotes synchronous particle. From Rps ρ 2Rp2s ρ

here on ps = p0 (t = ts ). Dispersion func- h2 D D
tion D(s) satisfies D + kps D = 1/ρ, where + 2 2 − 1 px W 2 (46)
R ps ρ
92
Ch.2: BEAM DYNAMICS
The coupling terms linear in the time derivatives [4] E.D. Courant, R.D. Ruth, W.T. Weng, AIP Proc.
of slowly varying quantities are 127 (1985) p.294
[5] L.J. Laslett, BNL-7534 (1963) p.324
˙ [6] A.J. Dragt, AIP Proc. 87 (1982) p.147
eV̂s
HSVC = 2 Δ(s){− sin φ + (φ − φs ) cos φ} [7] K. Symon, AIP Proc. 249 (1992) p.277
ωrfs
% [8] T. Suzuki, KEK Report 96-10 (1996)
ṗs ps h2 D
− x(φ − φs ) − 2 px W
ps ωrfs ρ R ps

h D 1 h3 DD 2 2.3.2 Tune Dependence on Momentum and
+ − 2 W (φ − φs ) − W Betatron Amplitudes
R ρ γs 2R3 ps
D.A. Edwards, DESY
eV̂s m2 c4
+ − Δ(s) sin φ M.J. Syphers, Michigan State U.
ωrfs Es2
& The multipole expansion used here is
sin φs
+Δ(s)(φ − φs ) cos φ + (φ − φs )2
∞
4πR
(47)
ΔBy + iΔBx = B0 (bn + ian )(x + iy)n (1)
n=1
where where B0 is the field on the reference trajectory,
hD hD and bn (an ) are the normal (skew) multipole co-
φ=φ− x+ px (48)
R Rps efficients (see also Sec.6.17.1). This form asso-
Some handy formulae: ciates n with a 2n + 2 pole magnet.
If a field gradient ΔB ≡ ∂By /∂x is in-
dts Es 1 troduced at a location s over a distance ds in a
= = (49)
ds ps c2 βs c closed trajectory of a synchrotron, the bend-plane
βs c hps c2 betatron oscillation tune will be perturbed by an
ωrfs = h = (50) amount
R REs "
1 ΔB (s)
ωrfs m2 c4 p˙s dν = β(s) ds (2)
ω̇rfs = (51) 4π (Bρ)
ps Es2
with an effect of the same magnitude but opposite
h ω 2 m2 in sign in the other transverse degree-of-freedom.
p = ps + W − rfs 3 W 2 (52)
R 2ps Then using Eq.(1) the tune change due to a mul-
hm2 ωrfs
2
hω̇rfs tipole of order n for a closed path differing from
+ W3 − 2 (φ − φs )W the reference trajectory by Δx(s) will in lowest
2Rp4s Rωrfs
order be [1]
"
1 ∞ n B0 (s)
1 (s − sgap ) Δν = β(s) bn (s)Δx(s)n−1 ds (3)
Δ(s) = + cos (53) 4π (Bρ)
2πR πR R
=1 In the case of a displaced equilibrium orbit due
! to a momentum offset, Δx(s) in Eq.(3) becomes
D 1 D(s)Δp/ps where D is the dispersion function
= (54)
ρ av γt2 (Sec.2.1.1). Though a betatron oscillation is not

h2 1 1 a closed path, Eq.(3) may be used for an estimate
−Kωrfs = 2 − (55) with Δx = a (β(s)/β0 ) cos ψ(s).
R ps γs2 γt2
Applying Eq.(2), the tune shift due to an oc-
topole distribution is
References
3 a2
Δν = β 2 b3 (4)
[1] K.R. Symon, ANL/APS/TB-28 (1997) (derivation 8 β0
of formulae in this section)
[2] D.A. Edwards, M.J. Syphers, An Introduction to
the Physics of High Energy Accelerators, Wiley References
(1993) p.129 [1] D.A. Edwards, M.J. Syphers, An Introduction to
[3] H. Bruck, Accélérateurs Circulaires de Particules, the Physics of High Energy Accelerators, Wiley
Presses Universitaires de France (1966); English (1993)
translation: LANL Report LA-TR-72-10 (1972)
93
2.3.3 Nonlinear Resonances References

D.A. Edwards, DESY
[1] E.D. Courant, H.S. Snyder, Ann. Phys. 1 (1958) 1
M.J. Syphers, Michigan State U.
[2] Vol.1, Sec.4.7, LHC Design Report,
http://cdsweb.cern.ch/record/782076
Recursive passage through field imperfections in
the magnets of a cyclic accelerator can lead to un-
wanted resonant behavior resulting in beam in-
(1993)
stabilities and loss [1]. For example, in low-
est nonvanishing order of a perturbative treat-
ment, introduction of a sextupole field can pro- 2.3.4 Synchrobetatron Resonances
duce instability at betatron oscillation tunes if A. Piwinski, DESY
3νx , 2νx + νy , νx + 2νy or 3νy are integers. The

ubiquity in tune-space of potential resonances for Synchrobetatron resonances (or satellite reso-
realistic magnet designs leads to reliance on com- nances) are excited when
puter simulation for dynamic aperture prediction kνx + νy + mνs = n (1)
(Sec.2.3.9 and the relevant material in [2]).
with k, , m = 0, and n integers. These res-
A useful aspect of nonlinear resonances is
onances reduce the lifetime of the beams dras-
found in the slow extraction process (Sec.4.12.1).
tically and limit the performance of circular ac-
As an illustration, the equations for oscillation de-
celerators. Sometimes they are single particle ef-
velopment with turn number n in the neighbor-
fects, but more often they depend strongly on the
hood of a half-integer resonance as given in [3]
bunch current.
are
The main mechanisms for the excitation of
dx 1 3 synchrobetatron resonances are dispersions in
= px − Q + 2πδ + D(x2 + p2x ) (1) sextupoles, dispersions in rf cavities, transverse
dn 2 8
fields in rf cavities, and the beam-beam interac-
dpx 1 3
= x − Q − 2πδ − D(x2 + p2x ) (2) tion (Sec.2.5.1) at a crossing angle. The excita-
dn 2 8 tion by the natural chromaticity [1, 2], where all
with px ≡ βx + αx, Q is the quadrupole half- sidebands of the integer resonances are excited, is
integer resonance driving term, δ = |Q/(2π)| weaker and compensated by sextupoles.
the stop-band width resulting from Q, and D is Dispersion in sextupoles The changes of x and
a zero-harmonic B (octupole) term providing a y in a sextupole depend on ks [(xβ + Dx δ)2 −
tune dependence with amplitude. For B > 0 (yβ + Dy δ)2 ] and on ks (xβ + Dx δ)(yβ + Dy δ)
and stable motion of small amplitude oscillations, where δ = ΔE/E, ks = integrated sextupole
δ < 0 and |Q| < 4πδ so for extraction the strength, and Dx,y = horizontal and vertical dis-
phase stable region will shrink as |Q| increases. persion at the sextupoles. These kicks excite the
These equations have a first integral (equivalent third order betatron resonances 3νx = n and
to a Hamiltonian), νx ± 2νy = n, and the following satellite reso-
nances [3]:
H = (px −u)2 +x2 −r 2 (px +u)2 +x2 −r 2 Horizontal satellites
2νx ± νs = n (excited by Dx )
with r 2 = u2 + v 2 where u and v are the fixed
points found by setting dx/dn = 0, dpx /dn = 0 νx ± 2νs = n (excited by Dx2 − Dy2 )
in Eqs.(1,2). Thus the separatrix is formed by two Vertical satellites
circles the centers of which are the stable fixed
points at ±u. The intersections are at the un- 2νy ± νs = n (excited by Dx )
stable fixed points ±v, from which particles are νy ± 2νs = n (excited by Dx Dy )
“squeezed” out as extraction proceeds. Satellites of the betatron coupling resonances
Though developed to provide external beams
νx ± νy ± νs = n (excited by Dy )
for HEP experiments, resonant extraction has
application in medical accelerator facilities It is assumed that the synchrotron oscillation
(Sec.4.12.1), where the third-integer sextupole- energy is much larger than the betatron oscillation
driven process is outlined. energy, so that the synchrotron coordinates are not
94
Ch.2: BEAM DYNAMICS
changed by the betatron coordinates. For a more with

precise calculation one has to take into account 1 A2 βx
the path lengthening due to the betatron oscilla- A1 = − Dx sin μx − D̃x (1−cos μx )+
βx 2
tion [4, 5].
The rise times for the satellites of the betatron A2 = −Dx (1 − cos μx ) − D̃x sin μx
coupling resonances are [3] 1
D̃x = Dx βx − Dx βx
1 y,x 2
= f0 δ̂ (2)
τ1x,y x,y μx = betatron phase advance, αp = momentum
( ( compaction factor, C = circumference. xβ , xβ ,
( (
× (( ksn Dyn βxn βyn ei(φxn ±φyn ) (( βx , βx , Dx , and D̃x refer to the beginning of the
n revolution.
where δ̂ is the energy oscillation amplitude, f0 is For linear betatron and synchrotron oscilla-
the revolution frequency. For the satellites with tions an exact solution for the coupled motion
a distance of νs /2 from integer and half integer can be found. The phases μ1,2 of the eigenval-
resonances, ues of the revolution matrix, on the resonance
( (
1 ( ( νx ± νs = n, are [4]
= f0 δ̂ (( ksn Dxn βx,yn e2iφx,yn (( (3)
τ2x,y n −(Dx2 + D̃x2 )
and for satellites with a distance of 2νs from inte- μ1,2 = μx,s ± (1−cos μx ) (8)
αp Cβx sin μx
ger resonances,
( ( Eq.(8) shows that an instability occurs, i.e. the
1 f0 δ̂2 ( (
= √ (( ksn (Dxn2
− Dyn
2
) βxn eiφxn (( amplitudes grow exponentially, when sin μx >
τ3x 2 x n 0. This is a difference resonance, since νs =
( ( (4) μx /2π − n is always smaller than 0.5.
1 f0 δ̂2 ( ( Nonlinear satellite resonances with |m| > 1
= √ (( ksn Dxn Dyn iφyn (
βyn e ( (5)
τ3y y [Eq.(1)] can be investigated approximately. An
n
invariant of motion is [4]
Note that only the rise times on 2νx,y ∓ νs = n
give an exponential increase. The other rise times αp Cβx 2
x̂2β + δ̂ = const. (9)
depend on the amplitudes and change with in- 2πmνs
creasing amplitude. All these satellites can be
compensated by varying the sextupole strength so where x̂β and δ̂ are the oscillation amplitudes. For
that the sums in Eqs.(2) to (5) vanish. most cases the second term in Eq.(9) dominates,
Dispersion in cavities In a cavity the energy of and the change of the betatron amplitude is larger
a particle is changed and its closed orbit is shifted than the change of the synchrotron amplitude.
by Dx δE/E. This changes the betatron coordi- Above transition, for satellite resonances
nate by Eq.(1), an instability occurs for a difference res-
onance. This is opposite to the case of x-y cou-
δE
δxβ = −Dx (6) pling resonances, for which difference resonances
E are stable and sum resonances are unstable. Be-
eV 2πz low transition (αp γ 2 < 1), one has to replace αp
= −Dx sin φs + − sin φs
E λ by (αp γ 2 − 1)/(γ 2 − 1); then satellite resonances
with V = cavity voltage, φs = synchronous phase, behave similarly to the x-y coupling resonances.
λ = rf wavelength, z = longitudinal position. A Synchrobetatron resonances due to a disper-
similar equation is obtained for δxβ . Only that sion, as satellites of integer resonances, were first
part of the energy change which varies with the observed in NINA [6] and later in SPEAR [7], PE-
synchrotron frequency plays a role for a reso- TRA [4], DORIS III [8], and LEP [9]. They can
nance, and not the constant radiation losses. be compensated with dispersion bumps.
The change of z per revolution is [4, 5] Transverse fields with longitudinal variation
"
1 ΔE Transverse fields which vary in longitudinal direc-
δz = xβ + Dx d
ρ E tion inside the bunch can be produced by the ac-
ΔE celerating voltage [10] or by the bunch itself [11].
= A1 xβ + A2 xβ − αp C (7) In both cases they are caused by asymmetries of
E
95
the cavities or by an off-centered passage of the with f the function describing the beam-beam
bunch. force and 2φ the total crossing angle. The syn-
The excitation can be calculated exactly for chrotron oscillation is also influenced by the beta-
resonances νx,y ± νs = n. For horizontal excita- tron oscillation, δpz = φ δpx , and
tion the kick is δE δp δpz δpx
≈ ≈ ≈φ = φδxβ = φf (xβ + φz)
e E p p p
δx = (Ex + vBy ) dt (10) (16)
p
The complete coupling is described by
The integral is taken along the path of the particle Eqs.(15) and (16). For small oscillation ampli-
between two limits where the fields vanish. The tudes one gets a linear coupling
kick can then be linearized to yield
4πξx
f (xβ + φz) ≈ − (xβ + φz) (17)

e ∂Ez βx∗
δx = G z where G = dt (11)
p ∂x
with
The same fields also perturb synchrotron mo- r0 NB βx∗
ξx =
tion. The corresponding energy change which 2πγσx,eff (σx,eff + σy )

varies linearly with the transverse position of a σx,eff = σx2 + φ2 σz2 , NB = number of parti-
particle in the bunch is cles per bunch, βx∗ = β-function at the IP, σz,x,y
= rms longitudinal, horizontal and vertical bunch
δE e ∂Es v2
=x v dt = 2 Gx ≈ Gx (12) sizes. With linear coupling, the phases μ1,2 of the
E E ∂x c
eigenvalues are [4]
Eqs.(11) and (12) describe a linear coupling. On
the linear resonance νx ± νs = n the phases of the −αp C
μ1 = μx + 2πξ ± 2πξφ (18)
eigenvalues of the revolution matrix are βx∗ sin μx

G αp Cβxc 2 αp C −αp C
μ1,2 = μx,s ± (13) μ2 = μs −2πξφ ±2πξφ (19)
2 − sin μx μs βx∗ βx∗ sin μx
where βxc is the β-function at the cavity. The os- See also Sec.2.5.1.3.
cillation amplitudes increase exponentially on a Computer simulations [12] can be used to de-
difference resonance. Assuming a constant syn- termine the strength of higher order satellites. In
chrotron amplitude, the growth rate for the beta- case of a horizontal crossing all satellites can be
tron amplitude x̂β is [4] excited for which k + m is even (for antisymmet-
ric space charge force and beam centers colliding
dx̂β 1 head-on). Many resonances are suppressed if the
= Gβx f0 ẑ (14)
dt 2 IPs are equidistant. If the phase advances between
where ẑ is the synchrotron amplitude. the IPs are different, e.g. due to machine errors,
more resonances can be excited.
These satellites can be compensated with help
Experiments with DORIS I [12] as well as the
of orbit bumps in the cavity sections. Since the
simulations show that the synchro-betatron reso-
excitation by a dispersion and the excitation by
nances excited by the beam-beam interaction with
transverse fields are additive, both mechanisms
a crossing angle reduce mainly the lifetime and
can amplify or compensate each other depend-
do not affect seriously the beam size or the lumi-
ing on the phases of the dispersion and the orbit.
nosity. Four parameters determine the reduction
Since most of the satellites are current dependent,
of the lifetime: (i) the normalized crossing angle
the compensation is complete only for a specific
φσz /σx , (ii) the beam-beam parameter ξ, (iii) the
bunch current.
synchrotron tune νs , and (iv) the physical or dy-
Beam-beam interaction with a crossing angle namical aperture.
A particle with longitudinal distance z from the
center of a bunch and a betatron displacement xβ References
receives a beam-beam transverse kick (Dx = 0)
[1] Y. Orlov, Sov. Phys. JETP, V5 (1957) 45
δxβ = f (xβ + φz) (15) [2] K.W. Robinson, CEA-54 (1958)
96
Ch.2: BEAM DYNAMICS
[3] A. Piwinski, DESY 93-187 (1993) Given initial conditions z i at reference position
[4] A. Piwinski, 11th Int. Conf. on High-Energy si , (3) can be integrated to yield unique final con-
Acc. (1980) p.638 ditions zf at position sf . We write this relation
[5] A. Piwinski, A. Wrulich, DESY 76/07 (1976) between z i and zf as (1). Integrating backward
[6] M.C. Crowley-Milling, I.I. Rabinowitz, PAC 77, in time, the map M arising from (3) is always in-
p.1052 vertible, z i = M−1 z f . Theorem of Poincaré: If
[7] SPEAR Group, PAC 75, p.1366 the fj (z ; λ; s) are analytic (have convergent Tay-
[8] H. Nesemann, 15th Int. Conf. on High Energy lor series) in s and z, then z f is analytic in z i . For
Acc. (1992) p.549
charged particle motion, the fj involve electro-
[9] J.M. Jowett et al, EPAC 90, p.1612
magnetic fields obtained from the Maxwell equa-
[10] N.A. Vinokurov et al, 10th Int. Conf. on High
tions and the conditions for Poincaré’s theorem
Energy Acc. (1977) p.272
[11] R.M. Sundelin, PAC 79, p.3604

are met. Therefore the expansion (2) always exists
[12] A. Piwinski, PAC 77, p.1408 and converges for sufficiently small ζ i . Analytic
dependence on parameters λ, e.g. field strength,

can be treated similarly.
2.3.5 Taylor Map, Hénon Map, Standard
Map Calculation of Taylor maps In the approxi-
A. Dragt, U. Maryland mation that extended fringe-field effects are ne-
glected, there are s-independent fj for each ide-
Definition A transfer map M is a relation that alized beamline element. The design orbit for
sends initial conditions z i to final conditions z f , each such element is then known, and the Taylor
which we write symbolically in the form map about the design orbit can be found by let-
ting the associated Lie transformation (Sec.2.3.7)
z f = Mz i (1) If s-dependence is taken into account, as
act on ζ.
A Taylor map is a representation of M that is necessary for the accurate treatment of fringe-
presents final conditions as a Taylor series in the field effects and wigglers/undulators, one can in-
initial conditions. Let ζ refer to deviation coordi- tegrate (3) to find a design orbit and integrate
nates about some reference orbit. Then M has a the associated complete variational equations to
Taylor representation [1] find the Taylor expansion (2) of M about this
design orbit. There is a machine-implementable
ζjf = Rjk ζki + Tjk ζki ζi algorithm that will generate the complete varia-
k

k tional equations to any desired order. If there is
+ Ujkm ζki ζi ζm
i
+ ··· (2) a Hamiltonian formulation for the equations of
km motion, one may also use a symplectic integrator
(Sec.2.3.8) to find the Taylor coefficients or may
The number of monomials of degree 0 through
directly integrate equations for the Lie generators
D in N variables is (D + N )!/(D!N !). There-
of M. The equations of motion for the Lie gener-
fore, if (2) is truncated beyond terms of degree
ators are known through 8th order. This approach,
D, the number of coefficients required to describe
which exploits the symplectic condition, involves
a truncated Taylor map is N (D + N )!/(D!N !).
integrating roughly 3-times fewer equations than
In the Hamiltonian case the coefficients are in-
those required to find the Taylor coefficients [2].
terconnected by nonlinear relations arising from
the symplectic condition, and are thus not inde- Manipulation of Taylor maps Multiplication
pendent. Extensive tools have been developed for (concatenation): If maps for individual elements
manipulating truncated power series (Sec.2.3.7). are known in Taylor form, these series may be
Existence of Taylor maps The equations of substituted into each other to find a net Taylor
motion for a charged particle in an EM field, in- map. Factorization: Maps in Taylor form may
cluding the average (but not the stochastic) ef- be factored into the product of nonsymplectic and
fect of synchrotron radiation, can be written in the symplectic parts, each in Lie form and unique. In-
form version: Taylor maps may be inverted by series re-
dzj version implemented by inverting R followed by
= fj (z; s) = fj (z1 , · · · zN ; s), j = 1, N recursive back substitution. Normal forms: Let
ds
(3) M be a given map. One seeks normalizing maps
97
A such that the map N = AMA−1 has a partic- symplectic condition can be enforced by use of
ularly simple form called a normal form. mixed-variable generating functions or Cremona
maps. Both can be computed from the truncated
Applications of Taylor maps Maps can be used Taylor map. Among generating functions, that
to look at aberration (nonlinear) content, to mon- of Poincaré seems optimal. In summary, the use
itor a design process, and to replace and speed of Taylor maps, expressed in different coordinate
up element-by-element tracking. To look at aber- systems, and their transformation to Lie and nor-
ration content one converts the Taylor map to mal forms, have become standard tools of con-
a Lie map, in which each Lie coefficient corre- temporary optical system design. These maps
sponds to a unique and readily identifiable aber- provide insight into the presence, nature, origin,
ration (Sec.2.3.6). For storage rings, these aberra- and effect of aberrations/nonlinearities, and ways
tions can also be specified in an action-angle reso-
to compensate/exploit them [2, 3].

nance basis. To monitor a design process one ex-
Hénon map and standard map [4] Requir-
amines the aberration coefficients at each step to

ing that the general quadratic Taylor map in two
ascertain hoped-for improvements. Designs and
variables be symplectic yields the much studied
design codes can be compared and cross-checked
Hénon map. It can be written as
by comparing aberration strengths. In a beamline,
one or more segments may be of special interest. xn+1 = xn cos μ + (pn + Sx2n ) sin μ
For example, in a final focus system, sf is cho- pn+1 = −pn sin μ + (pn + Sx2n ) cos μ (4)
sen to be the IP. To avoid having the map coeffi-
In Lie form (Sec.2.3.6), it is given by M =
cients depend on the choice of si , one can factor
R(μ) exp(:Sx3 /3:). This map corresponds to a
the map as a linear map from si to the IP, followed
clockwise phase-space rotation by angle μ (phase
by a nonlinear map from the IP to itself. For a
advance), followed by a thin sextupole kick of
circular machine the one-turn map can be found
strength S. Although simple in appearance, when
by concatenation and then brought to normal form
viewed as a one-turn map it is complicated un-
N . N contains all information about tunes, phase
der iteration and has a tune-dependent dynamic
slip, chromaticities, and anharmonicities. A con-
aperture (Sec.2.3.9). Thus, it provides a simple
tains all information about lattice (betatron) func-
example that can be studied to gain insight into
tions. All couplings are automatically included.
the general question of dynamic apertures. It has
The δ-dependent chromatic terms, which are iso-
been generalized to higher dimensions [5].
lated by this procedure, often need special atten-
The simplest synchrotron motion map, asso-
tion because the longitudinal emittance is typi-
ciated with a thin rf cavity kick, can be written as
cally orders of magnitude larger than the trans-
(Sec.2.1.2)
verse. It is important to understand how these
δ-dependent terms arise/cancel in beam-line sec- eV
δn+1 = δn + 2 (sin φn − sin φs )
tions and to devise schemes that minimize this de- β E
pendence. Together A and the amplitude depen- φn+1 = φn + 2πhη(δn+1 )δn+1 (5)
dent terms (anharmonicities) and chromaticities
If one defines y ≡ h|η(δ)|δ, x ≡ φ/2π, νs = the
determine the tune footprint for any given beam
synchrotron tune, then (5) becomes for stationary-
distribution. The tune footprint can also be mini-
bucket synchrotron motion the much studied stan-
mized.
dard map:
When due attention is paid to the symplec-
tic condition, experience shows that tracking with yn+1 = yn − 2πνs2 sin 2πxn
a high-order one-turn map, or a few high-order xn+1 = mod(xn + yn+1 , 1) (6)
lumped maps, gives the same results (comparable The Lie form for the standard map is Ms =
dynamic aperture) as element-by-element track- exp(:νs2 cos 2πx:) exp(: − y 2 /2:). This map ex-
ing. Use of the one-turn map or a few lumped hibits chaotic behavior that becomes large-scale
maps can be at least an order of magnitude faster, rather quickly as νs is varied.
and realistic maps based on 3-D field data and in-
cluding all fringe-field and high-order multipole
References
effects, rather than the approximate thin-element
kick maps employed in element-by-element track- [1] K. Brown, SLAC-75-REV-4 (1982) and SLAC-R-
ing, can be used without sacrificing speed. The 95-462 (1995)
98
Ch.2: BEAM DYNAMICS
[2] A. Dragt, Lie Methods for Nonlinear Dynamics Hamiltonian Lie operators are when N is even
with Applications to Accelerator Physics (2010); and gb (z) = a (∂f /∂za )Jab . Non-Hamiltonian
www.physics.umd.edu/dsat/ vector fields are used for dissipative effects (e.g.
[3] M. Berz, Advances in Imaging and Electron synchrotron radiation).
Physics 108, P. Hawkes, Edit., Academic (1999) Lie transformations: Associated with :f : or f
[4] Google “Henon Map” and “Standard Map” is the Lie transformation denoted by exp(:f :) and
[5] J. Moser, Mathematische Zeitschrift 216, 417 defined by
(1994)
∞

exp(:f :) = e:f : = :f :n /n! (7)
2.3.6 Lie Algebraic Methods n=0
A. Dragt, U. Maryland
More precisely, exp(:f :) is a Hamiltonian Lie
Map and Lie methods provide a framework for transformation. Operators exp(Lg ) are also called
abstract concepts and related computational al- Lie transformations. The quantities f , or :f :, or
gorithms. They provide a powerful extension of Lg are called Lie generators.
linear concepts and algorithms into the nonlinear General facts/theorems Symplectic matrices
regime [1, 2, 3, 4, 5]. form a Lie group, called Sp(2n, R), of dimension
Definitions Let f and g be functions of the n(2n+1) where 2n is the phase-space dimension.
phase-space variables z = (q1 , p1 , q2 , p2 , · · · ). Symplectic maps form an infinite dimensional Lie
Although results are general, we will assume that group. If M arises from integrating Hamiltonian
phase space is 6-D. differential equations, it will be symplectic, and
Poisson bracket: vice versa. For any g(z, t), Lie operators have the
3 properties
∂f ∂g ∂f ∂g
[f, g] = − (1)
∂qj ∂pj ∂pj ∂qj :f :0 g = g, :f :g = [f, g], :f :2 g = [f, [f, g]], · · ·
j=1
Fundamental Poisson brackets: exp(:f :)g = g+[f, g]+[f, [f, g]]/2!+· · · (8)
[za , zb ] = Jab , a, b = 1, 6 where (2) The sum on the right side of (8) is called a Lie
⎡ ⎤ series.
J2 0 0
0 1 Suppose f (z i , t) is any function of z i and
J = ⎣ 0 J2 0 ⎦ , J2 = (3)
−1 0 possibly the time t. Define
0 0 J2
Symplectic matrices: Matrices N that satisfy z f = exp(:f :)z i (9)
N t JN = J are called symplectic. Then z f and z i are related by a symplectic map
Symplectic maps: Given a transfer map M, M,
form the Jacobi matrix M (z i ),
z f = Mz i with M = exp(:f :) (10)
Mab (z i ) = ∂zaf /∂zbi (4)
M is called symplectic if M (z ) is symplectic for
i That is, Lie transformations provide an endless
all z i . Truncating a Taylor map generally violates supply of symplectic maps. Conversely, suppose
the symplectic condition. M is an analytic symplectic map having a Taylor
Lie operators: Associated with any function expansion of z f in terms of z i . Then M has the
f (z, t) of phase-space variables z and time t is a Dragt-Finn factored product representation
differential Lie operator denoted by :f :, M = exp(:f1 :) exp(:f2c :) exp(:f2a :)
3
exp(:f3 :) exp(:f4 :) · · · (11)
∂f ∂ ∂f ∂
:f : = − (5)
∂qj ∂pj ∂pj ∂qj Here the fm are homogeneous polynomials of de-
j=1
gree m. The quadratic polynomials f2c,a are of the
More precisely, :f : is a Hamiltonian Lie operator.
form
Let x = (x1 · · · xN ) and let g1 (x) · · · gN (x) be 1 c,a
a collection of N functions of x. A general Lie f2c,a = − Sde zd ze (12)
2
operator (or vector field) Lg is defined by de
where S c and S a are any symmetric matrices that

N
Lg = gb (x)(∂/∂xb ) (6) commute or anticommute with J, respectively.
b=1
We remark that any Taylor map can be uniquely
99
factored into a symplectic (Hamiltonian) and non- tors, for which we use the notation {A, B}, etc.,
symplectic (non-Hamiltonian) part [4]. The sym- C(s, t) = sA + tB + (1/2){A, B}
plectic part then has the factorization (11) and the 1 1
nonsymplectic part has an analogous Lie factor- + s2 t{A, {A, B}} + st2 {B, {B, A}}
12 12
ization in terms of non-Hamiltonian vector fields.
Finally, unlike the Taylor representation, the prod- −(1/24)s2 t2 {A, {B, {A, B}}} + · · · (14)
uct (11) can be truncated at any stage and the re- That C depends only on commutators is the
sult will still be a symplectic map. Baker-Campbell-Hausdorff (BCH) theorem. The
Introduce the shorthand notation series (14) is called the BCH series. In the set-
ting of Lie transformations, exp(:f :) exp(:g:) =
Rf = exp(:f2c :) exp(:f2a :) (13) exp(:h:) with
h = f + g + (1/2)[f, g] + (1/12)[f, [f, g]]

Each factor in (11) plays a definite role. The fac- 1 1
+ [g, [g, f ]] − [f, [g, [f, g]]] + · · · (15)
tor exp(:f1 :) produces phase-space translations

and Rf produces the linear part of the Taylor map 12 24
Inversion formulas: It follows from the BCH
described by the matrix R. See Sec.2.2.1. The
formula that the inverses of (10),(11) are
factor exp(:f3 :) produces the quadratic nonlinear
part described by T and also contributes to still M−1 = exp(−:f :)
M−1 = · · · e−:f4 : e−:f3 : e−:f2 : e−:f2 : e−:f1 : (16)
a c
higher-order parts described by U , · · · ; the factor
exp(:f4 :) produces the cubic nonlinear part de- Map multiplication: Define the matrix Rf as-
scribed by U and also contributes to still higher- sociated with Rf by the rule
order parts, etc. Indeed, there is a well-defined f
procedure for passing between Taylor series and Rf za = Rab zb (17)
Lie product representations of symplectic maps. b
To reproduce the Taylor coefficients for a sym- Two symplectic maps Mf =
plectic map M through terms of degree D, it Rf exp(:f3 :) exp(:f4 :) · · · and Mg =
is necessary to specify the generators f1 , f2 · · · Rg exp(:g3 :) exp(:g4 :) · · · can be multiplied
fD+1 . In the case of N variables this requires (concatenated) to find Mh = Mf Mg with
[−1 + (D + 1 + N )!/(D + 1)!N !] coefficients, Mh = Rh exp(:h3 :) exp(:h4 :) · · · . Use of the
which is considerably fewer that the number of BCH series gives the results
Taylor coefficients. In summary, exp(:f1 :) and Rh = Rg Rf
Rf describe translations and linear transforma- h3 = f3tr + g3
tions, and the remaining factors describe nonlin- h4 = f4tr + g4 + (1/2)[f3tr , g3 ]
ear effects. And this is done in an optimal way
since any set of fm produces a unique symplectic h5 = f5tr + g5 − [g3 , f4tr ] − (1/6):f3tr :2 g3
map, and any analytic symplectic map can be ex- +(1/3):g3 :2 f3tr , etc. (18)
pressed as (11) with the fm uniquely determined. tr
Here fm denotes the transformed polynomial
Later, after introducing suitable coordinates spe- fmtr (z) = f [(Rg )−1 z]. Results are known
m
cific to accelerator physics applications, we will through h8 and when Mf and Mg also contain
summarize some of the simpler Lie maps that f1 and g1 factors. Map multiplication in Lie form
are applicable to various idealized beam-line el- is much faster than in Taylor form. The price to be
ements. paid for this speed is that formulas of the kind just
quoted have to be derived and coded in advance
Lie algebraic facts There are deep properties of for any specified order whereas substitution of
the exponential function which facilitate the ma- Taylor series into Taylor series can be done using
nipulation of maps in Lie form. a machine-implementable algorithm that can be
Combining exponents: Let A and B be any initiated to any specified order at run time. Con-
two square matrices of the same dimension. Form catenation formulas are used to find the net map
matrices exp(sA) and exp(tB) where s and t are for successive beam-line elements and for gen-
parameters. For s and t sufficiently small, a ma- eral map manipulation including the conversion
trix C that satisfies exp(sA) exp(tB) = exp(C) of M−1 as given by (16) into standard factored
exists, is unique, and is a sum of elements formed product form. In a circular machine, they are also
only from A and B and their multiple commuta- used to find the one-turn (or multiple-turn) map.
100
Ch.2: BEAM DYNAMICS
Canonical coordinates Here we introduce straight beamline element (length L) is indepen-

canonical coordinates that are useful for accelera- dent of s, the map for the body of the element is
tor physics applications. We will work with a full given in single exponent form
6-D phase space. M = exp(−L:H:) (23)
Unscaled coordinates: z = (x, px , y, py , t̃, p̃t )
(There is an analogous result for curved ele-
where p̃t = pt −p0t with pt = −(m2 c4 +p2 c2 )1/2 = ments.) Here H is assumed to be expanded in
−γmc2 and p0t = −[m2 c4 + (p0 c)2 ]1/2 and p0 homogeneous polynomials,
the design momentum. The coordinate t̃ is time
of flight relative to the synchronous particle. H = H2 + H3 + H4 + · · · (24)
Dimensionless scaled coordinates: Here is Table 1 gives a few examples of single-exponent
a convenient scale length. maps in lowest order for some simple beam-line
elements/transformations.
X = x/ , Px = px /p0
Y = y/ , Py = py /p0 Table 1: Example Lie maps for some ideal beam-line
τ = ct̃/ , Pτ = p̃t /(p0 c) (19) elements.
Relation to noncanonical coordinates: With Beamline element Lie Map

β0 = v 0 /c = −p0 c/p0t , Pτ and the momentum Drift length L L
exp[− 2 :Px2
deviation δ = (p − p0 )/p0 are related by the for- Pτ2
(lowest order) + Py2 + (β0 γ0 )2
:]
mulae
Pτ = −(1/β0 ){[1 + (2δ + δ2 )β02 ]1/2 − 1} Thin quad. Q exp(− Q
2 :X − Y :)
2 2
=−β0 δ + (δ2 /2)(β03 − β0 ) + · · · (20) Thin sext. S exp(− S3 :X 3 − 3XY 2 :)

δ = [1 − 2Pτ /β0 + Pτ2 ]1/2 − 1 X
Kick ΔPx = f (X) exp[: 0 f (x)dx:]
=−Pτ /β0 + (Pτ2 /2)(1 − β0−2 ) + · · · (21)
x-deflection d exp(d:X:)
With s = z as the independent variable and
(X, Px , Y, Py , τ, Pτ ) as dependent variables, the Thick quad. k, L exp(−L:H:) where H =
2
Hamiltonian for motion in a drift space is 1
( 2 )[(Px2 +Py2 + βP2 τγ 2 )
0 0
1 2Pτ 1 Pτ 1 +k2 (X 2 − Y 2 )]
H=− (1− +Pτ2 −Px2 −Py2 ) 2 + − 2
β0 β0 β0 x-phase advance θ exp(− θ2 :Px2 + X 2 :)
(22)
Using H, we find that x = dx/dz = X = x-phase advance exp(− θ2 :βPx2
(∂H/∂Px ) and hence with Courant- +2αXPx + γX 2 :)
− 1 Snyder parameters
2Pτ 2
x = Px 1 − + Pτ2 − Px2 − Py2
β0 A single exponent Lie map can be converted
to Taylor form by letting it act on z. In general
Px Pτ Px 3
= Px + + Pτ2 ( 2 −1)+Px2 +Py2 +· · · many terms need to be retained in the resulting
β0 2 β0
Lie series and the coefficients 1/n! must be relied
The TRANSPORT variable x equals Px in lowest on to yield convergence. To convert a single expo-
order, but has second-order chromatic and higher- nent form to a factored product form, one way is
order geometric and chromatic differences. The to convert the Taylor series just described to fac-
variables X, x do not form a canonical pair, tored product Lie form. Other ways, faster, and
[X, x ] = 1. Consequently, the use of TRANS- better convergent for large LH, involve scaling,
PORT variables makes it difficult to verify or ex- splitting, and squaring methods.
ploit the symplectic condition. Nonautonomous case: When the Hamilto-
Computation of maps For each beamline ele- nian is not constant (as occurs in reality due to
ment, M obeys the equation of motion dM/ds = fringe fields and field nonuniformity including
M: − H: with the initial condition M(0) = I. wigglers), the transfer map is first written in the
Autonomous case: In the approxima- reverse factorized form
tion/idealization that the Hamiltonian H of a M = · · · exp(:f4 :) exp(:f3 :)R. (25)
101
let S be the symmetric matrix with H2 =

Next, of numerical noise, the process of continuing in-
(1/2) ab Sab za zb . Then the matrix R that de- ward using the Maxwell equations is smoothing.
scribes R obeys This smoothing is related to the fact that harmonic
dR/ds = JSR, initial cond. R(0) = I (26) functions take their extrema on boundaries. When
using surface methods, all fits are made on such
Similarly, the fm obey boundaries. Therefore, if these fits are accurate,
df3 /ds = −H3int interior data based on these fits will be even more
accurate. Thus, with the use of surface meth-
df4 /ds = −H4int + (1/2):f3 :(−H3int ) ods, it is possible to compute for the first time
df5 /ds = −H5int + :f3 :(−H4int ) accurate transfer maps for realistic beam-line ele-
+(1/3):f3 :2 (−H3int ), etc. (27) ments. [4, 6, 7].
Applications of maps There are four well-

with the initial conditions fm (0) = 0. Results are developed uses for maps: analysis, optimiza-
known through f8 . Here the interaction picture tion/fitting, tracking, and the treatment of mo-
quantities Hm int are the polynomials H int (z) =
m ments including emittance questions.
Hm (Rz). Eqs.(26) and (27) are to be integrated Analysis (single-pass systems): Write the to-
numerically. The work involved in this Lie ap- tal transfer map M in the form M = LT where
proach is considerably less than that involved in L is a map (often linear) describing the desired
integrating the complete variational equations for behavior for the system and T describes aberra-
the corresponding Taylor map. Again, the price tions. Table 2 gives examples of aberration maps
to pay for this speed advantage is that formulae of a spot forming or final focus system. Suppose
of this kind have to be derived and coded in ad- the system is made of two sections M1,2 and a
vance for any specified order whereas there is a corrector C is placed between them. Then T is re-
machine-implementable algorithm that can be ini- placed by T with T = T M−1 2 CM2 . Analogous
tiated at run time to generate the complete vari- formulae apply for the case of several correctors
ational equations to any specified order. (Both at several locations. The correctors and their lo-
in this case and for concatenation formulas it is cations are selected so that T is free of offensive
pleasant that there are two complementary ap- generators.
proaches for checking each other.) Once R and
the fm are determined, (25) can be converted to Table 2: Example aberration maps.
standard factored product form by concatenation.
Realistic maps In the discussion above it is as- Aberration Aberration map T
sumed that the quantities Hm in (24) are known. Chrom. horiz. steering exp(e:P P :)
Since the Hamiltonian for charged-particle mo- x τ
(dispersion), strength e
tion in an EM field involves the vector poten-
tial A, this assumption is equivalent to assum- Chrom. horiz. focusing exp(e/2:P 2 P :)
x τ
ing high-order Taylor expansions of A are known. strength e
Equivalently, high-order derivatives of the field exp(:aPx4
are assumed known. Realistic field data can be Third-order spherical
provided on a grid with the aid of various 3- + bPx2 Py2 + cPy4 :)
D finite element codes, sometimes spot checked
against measured data. But the computation of Analysis (circulating systems): Here, normal
high-order transfer maps based on this data ap- form analysis is useful [2, 3, 5]. Let M be the
pears to pose an insurmountable problem: the di- one-turn map. Its normal form is N = AMA−1
rect calculation of high derivatives based only on where A is a symplectic map selected to make N
grid data is intolerably sensitive to noise (due to as simple as possible. In case of a static map (no
truncation or round-off) in the grid data. This rf), A can be selected so that N = exp(−:h:)
problem can be solved by the use of surface meth- with
ods (Sec.2.2.2). The effect of numerical noise can h = (wx + wx Pτ + wx Pτ2 )hx
be overcome by fitting field data on a bounding + (wy + wy Pτ + wy Pτ2 )hy
surface far from the beam axis and continuing in-
ward using the Maxwell equations. While the pro- + ah2x + bhx hy + ch2y + dPτ2
cess of differentiation serves to amplify the effect + ePτ3 + f Pτ4 + · · · (28)
102
Ch.2: BEAM DYNAMICS
Here hx = (X 2 + Px2 )/2, hy = (Y 2 + Py2 )/2. In Optimization/Fitting: Procedures can be ex-

principle (28) can be extended to arbitrary order ecuted in which various beam-line parameters
(∼ 12th order in practice) providing the tunes are or other parameters are varied to drive selected
not resonant to that order. Coefficients in (28): transfer map parameters, or quantities computed
• wx , wy give the “tunes” (= w/2π) of the on- from them, to specified values. This includes fit-
energy closed orbit, ting tunes, anharmonicities, chromaticities, dis-
• wx , wy are first-order chromaticities, persions, and other lattice functions. It also in-
cludes focusing and aberration correction and the
• wx , wy are second-order chromaticities, construction of high-order achromats. Because
• a, b, c are “anharmonicities” i.e. how tunes Lie generators are independent, while Taylor co-
depend on betatron amplitudes. efficients are not, there is less likelihood of de-
• d, e, f are linear and higher order “phase- manding the impossible [3].
slip” terms, Tracking: To compute single-particle trans-
N sends circles in the X, Px and Y , Py planes port, it is necessary to evaluate Mz i with M

into themselves, i.e. it describes idealized beta- given by (11). The effect of exp(:f1 :) is to pro-
tron motion. A gives complete information about duce a phase-space translation, and the exp(:f2 :)
all linear and nonlinear Courant-Snyder parame- factors produce the linear transformation (13).
ters, lattice functions (including nonlinear distor- The effect of the remaining product of factors
tion functions) and their chromatic behavior, the N = exp(: f3 :) exp(: f4 :) · · · exp(: fD+1 :),
closed orbit, and coupling effects. Thus, A and the nonlinear part of M, is more difficult to eval-
A−1 describe the transformation between the ac- uate. Several approaches are currently available.
tual and the idealized betatron motion. First, a Taylor series map T of the form (2)
The quantities I x,y = A−1 hx,y are invariant in Sec.2.3.5, truncated beyond terms of degree D,
under the action of M, i.e. MI x,y = I x,y , and may be used. Although truncation generally vio-
generalize Courant-Snyder invariants to the non- lates the symplectic condition, such an approach
linear case. is usually adequate for single-pass systems such
Applying map A−1 to turn-by-turn tracking as linacs, beam lines, telescopes, and final focus
data, the transformed data lie on circles in the X, systems. Symplecticity can be improved by re-
Px and Y , Py planes to the extent that the nor- taining more terms in the expansion (7). This can
mal form process succeeds. This presentation of be done for each of the factors in the product N .
tracking data provides a test for the existence of Alternatively, we can combine the generators :f3 :,
KAM tori. :f4 :, · · · into a single exponent using the BCH se-
Let A1 and A2 be the factors in A that remove ries, and then use a truncated version of (7) to
closed-orbit offsets and bring the linear part of evaluate the exponential of this single exponent.
the remaining map to normal form, respectively. This method often proves satisfactory for a mod-
Write the partially normalized map as est number of turns in a circulating system.
For long-term tracking it is necessary to
M = (A2 A1 )M(A2 A1 )−1 maintain the symplectic condition exactly. For
= R2 exp(:g3 :) exp(:g4 :) · · · element-by-element tracking, which is slow, this
may be done by using symplectic integrators
R2 = exp(wx :hx : + wy :hy : + d:Pτ2 :) [4, 5] (Sec.2.3.8). When a one-turn or few-lump
The generators g3 , g4 , · · · can be expanded in map is employed, which is much faster if feasi-
eigenvectors of the Lie operators :hx : and :hy :. ble, mixed-variable generating function or Cre-
These eigenvectors form the static resonance ba- mona map approximations to these maps may be
sis. When such an expansion is made, the coef- used.
ficients of the various resonance basis vectors are For 2n-dimensional phase space there is a
the resonance driving terms in the map M. 2n(4n + 1) parameter family of generating func-
Methods of this section can be generalized to tions [4, 8]. Among them the best for tracking
time dependent dynamic maps (powered rf cavi- purposes appears to be that of Poincaré. Form the
ties). All synchrotron oscillation effects are then quantities Σ = (z f + z i )/2 and Δ = z f − z i .
taken into account including synchrobetatron cou- In the Poincaré generating function approach the
pling and resonance effects. action of N is represented by the implicit (and
103
guaranteed symplectic) map polynomials in z through terms of some fixed de-

z f = z i + J∂Σ F (29) gree. Then one can define initial moments miα by
where F (Σ, Δ) is a polynomial that can be calcu-
the rule
lated by formula from the polynomials f3 , f4 · · · . miα = d6 zh(z)Pα (z) (31)
Alternatively, J∂Σ F can be computed algorithmi-
Next suppose the particle distribution is trans-
cally from the associated Taylor map T . The rela-
ported through some system described by a map
tions (29) can be solved numerically using New-
M. Then the final distribution at the end of the
ton’s method. There are three advantages to the
system is h(M−1 z). Correspondingly, the final
use of the Poincaré generating function: First,
moments are given by
suppose T is initially inverted and then Poincaré
symplectified. The result is the same as that found mfα = d6 zh(M−1 z)Pα (z)
when T is initially symplectified and the result

then inverted. Second, suppose the map RT R−1
is formed, where R is any linear symplectic map, = d6 zh(z)Pα (Mz) (32)

and then Poincaré symplectified. The result is
Since the Pα are complete, there is an expansion
the same as symplectifying T and then sandwich-
ing the result between R and R−1 . We may say of the form

that Poincaré symplectification is covariant under Pα (Mz) = Dαβ (M)Pβ (z) (33)
inversion and linear symplectic transformations. β
Finally, it can be shown that the Jacobian of T ,
where the Dαβ (M) are coefficients that can be
which is the identity matrix I for z i = 0, can
calculated for any transfer map M. It follows that
take on values all the way out to −I before the
moments transform linearly according to the rule
Poincaré symplectification process fails.
A Cremona map C is a map that is both mfα = Dαβ (M)miβ (34)
polynomial and exactly symplectic [9]. A Cre- β
mona map approximation consists of a series of
Note that by this method one can find the evolu-
Cremona maps C1 , C2 · · · Cm such that N
tion of moments without tracking particle distri-
C1 C2 · · · Cm . The construction of satisfactory Cre-
butions. We also remark that with the aid of A
mona approximations requires a knowledge of
one can manufacture matched beam distributions,
group theory and cubature formulae on the mani-
even in the nonlinear case, without tracking.
folds S 2 and SU (3)/SO(3). In general the num-
Let m be a vector with components mα , and
ber of maps m depends on the degree D. For
let D(M) be a matrix with entries Dαβ (M).
4 and 5 dimensional phase space one has m =
12, 24, 30, 32, 50, and 72 for D = 4, 6, 7, 8, 10, Write (34) in the more compact form mf =
and 13, respectively. Because Cremona maps are D(M)mi . A function of moments I[m] is said
polynomial and exactly symplectic, they can be to be a kinematic invariant if it obeys the rela-
evaluated rapidly on a computer to give results tion I[D(M)m] = I[m] for all symplectic maps
that are symplectic to machine precision. It has M. Rather little is known about the existence and
been found for the LHC that tracking using a Cre- properties of kinematic invariants for the set of all
mona approximation to the one-turn map satisfac- symplectic maps. However, all kinematic invari-
torily reproduces its dynamic aperture, and is at ants have been found when the symplectic maps
least an order of magnitude faster than element- M are restricted to be linear [10]. Of particular
by-element tracking. interest are kinematic invariants that can be con-
[2]
Moments and emittance analysis: Suppose structed from the second-order moments Zab . For
h(z) is the initial distribution function describing a given particle distribution, Let Z be the 6 × 6
[2]
[2]
a collection of particles in phase space [3, 10]. matrix with entries Zab . For a 3-degree of free-
Define initial moments Zabc··· i by the rule dom system, there are 3 such functionally inde-
pendent invariants,
i
Zabc··· = za zb zc · · · = d6 zh(z)za zb zc · · · (n)
I2 [Z (2) ] = tr[(Z (2) J)n ], n = 2, 4, 6 (35)
(30)
Alternatively let Pα (z), where α is some running Even more can be said. Z [2]
can be shown
index, denote a complete set of homogeneous to be positive definite, from which it follows
104
Ch.2: BEAM DYNAMICS
that there is a symplectic matrix A such that [5] E. Forest, Beam Dynamics: A New Attitude and
AZ [2] At = diag(λ1 , λ1 , λ2 , λ2 , λ3 , λ3 ) with all Framework, Harwood Academic (1998)
λj > 0. The quantities λ2j are mean-square eigen [6] C. Mitchell, U. Maryland PhD Thesis (2007);
emittances that generalize the 1-D mean-square www.physics.umd.edu/dsat/
emittances 2i = qi2 p2i − qi pi 2 to the fully [7] C. Mitchell, A. Dragt, PRST-AB 13, 064001
coupled case. In terms of the λj the kinematic (2010)
invariants have the values [8] B. Erdélyi, Michigan State U. PhD Thesis (2001)
(n) [9] D. Abell, U. Maryland PhD Thesis (1995);
I2 = 2(−1)n/2 [λn1 + λn2 + λn3 ], n even (36) www.physics.umd.edu/dsat/; D. Abell, E. McIn-
There are symplectic matrix routines for finding tosch, F. Schmidt, PRST-AB 6, 064001 (2003)
A and the λj . If only the λj are required, they [10] A. Dragt et al, PRA 45, 2572 (1992); N. Yampol-
sky et al, arXiv:1010.1558 (2010)
can be found from the eigenvalues of JZ [2] . Fi-

nally, while the entries in Z [2] evolve as a particle
distribution propagates through a beam line, the 2.3.7 Differential Algebraic Techniques
eigen emittances remain unchanged (in the lin- M. Berz, Michigan State U.
ear approximation). Thus, given an initial particle
distribution, one can compute the initial second The study of the behavior of motion in the vicinity
moments za zb i , and from them the eigen emi- of a chosen reference solution is a central prob-
tances. Let λ2min be the minimum of the quanti- lem arising in many subfields of dynamical sys-
ties λ2j . Also, let za zb f be the set of quadratic tems, including beam dynamics. The Taylor ex-
moments of the beam after it has passed through pansions of these solutions can be obtained by
some beam processing system. Then there is solving the so-called variational equations, which
a classical uncertainty principle and a minimum in beam physics has been carried out to orders
emittance theorem which state that, no matter two and three in the code Transport[1], to orders
what is done to the beam (ignoring nonlinear and three for example in the codes TRIO [2], GIOS
nonsymplectic effects), there are the restrictions [3] and MaryLie [4], and to order five in COSY
(for i = 1, 2, 3.) 5.0 [5]. This approach is laborious in practice,
and the development of the DA techniques has
qi2 f p2i f ≥ λ2min and (2i )f ≥ λ2min (37) greatly simplified this endeavor in beam physics
This information is useful when designing and other fields. In their latest versions [6, 7, 8, 9],
sources (e.g. magnetic field immersed and/or oth- the unprecedented accuracy these methods afford
erwise tailored electron guns) and when designing for the solution of differential equations has been
a beam line to perform emittance manipulations awarded the Moore prize for rigorous computing.
on a beam. It should also be useful in analyzing A review of DA as used in our field can be
the results of beam cooling experiments. In this found in [10]. The DA techniques allow the con-
case one can measure all quadratic moments be- venient computation of high-order Taylor expan-
fore and after a cooling channel. Next compute sions of the transfer map M which relates final
the eigen emittances of Z [2] before and Z [2] after. particle coordinates zf to initial coordinates zi
Ideally, one would like to find that all the λ2j have and parameters δ,
decreased, or at least the minimum of the λ2j has zf = M(zi , δ) (1)
decreased.
relative to one (in most beamlines, microscopes,
or synchrotrons) or several (in FFAGs, cyclotrons
References and some spectrographs) reference orbits. The
[1] A. Dragt, AIP Conf. Proc. 87, R. Carrigan et al map M is the flow (solution depending on ini-
ed. (1982) tial conditions) of ODEs
[2] A. Dragt et al, Ann. Rev. Nucl. Part. Sci. 38, 455 z = f(z, s), independent variable s (2)
(1988)
[3] A. Dragt et al, MaryLie 3.0 Users’ Manual Since their introduction [11, 12], DA techniques
(2003); www.physics.umd.edu/dsat/ have been utilized in most newly developed codes
[4] A. Dragt, Lie Methods for Nonlinear Dynamics [13, 14, 15, 16, 17, 18] and the DA engines form-
with Applications to Accelerator Physics (2010); ing the core of COSY INFINITY [13] also form
www.physics.umd.edu/dsat/ the backbone of various other codes. DA methods
105
have their origin in the algebraic study of prob- ation” [24], although in that field derivatives are
lems involving differentiation and integration for typically only obtained to first or second order and
the purpose of solving complicated integrals, the challenge lies in the efficient handling of large
ODEs, and PDEs. Pioneered by Liouville [19] numbers of independent variables. An impor-
in connection with the search of elementary in- tant practical problem of this map integration ap-
tegrals for elementary functions, it was put on proach is to obtain higher order terms accurately,
a solid foundation and significantly enhanced by which can be done by reducing the step size. This
Ritt [20] and Kolchin [21]. Now the methods have approach has been used to retrofit several existing
gained prominence in the field of formula manip- tracking codes, including recently MAD-X for the
ulation, where they provide the backbone of the extraction of high-order DA maps.
theory of analytic quadrature and integration of However, using DA techniques it is possible
ODEs [22]. to entirely avoid the approximate time stepping of

For a given function f of v variables, we form conventional integrators, but rather develop new
a vector that contains all Taylor expansion coeffi- integrators of arbitrary order in time. Strikingly,
cients at x = 0 up to a certain order n. Know- they require only a single evaluation of the r.h.s
ing this vector for two functions f and g allows per step, resulting in greatly increased efficiency
to compute the respective vector for f + g and and robustness [10, 6, 7]. One way is based on
f · g, since the derivatives of the sum and product the common rewriting of the ODE as a fixed point
function is uniquely defined from those of f and problem
sf
g. The resulting operations of addition and multi-
zf = zi + f(z, s̃)ds̃ (3)
plication lead to an algebra, the Truncated Power si
Series Algebra (TPSA) [12, 23]. One can also in-
Utilizing the operation ∂ −1 for the integral, the
troduce elementary functions like exp, sin etc. on
problem can be iterated in DA with s̃ as an ad-
TPSA. TPSA allows the convenient computation
ditional variable. It can be shown that iteration
of derivatives of any functional dependency on a
converges to the exact result in n steps, where n
computer [24].
is the order of the DA operations; moreover, this
For the solution of ODEs and PDEs it is nec- is not affected if in the ith iteration step the over-
essary to introduce another operation. For any all order is reduced to i. The result is an nth order
fixed function g with g(0) = 0, it is possible to integrator; for a given accuracy demand, the inte-
determine the Taylor coefficients of g · ∂f /∂xi = grator typically results in a speed-up of about an
∂g,i f from those of f . Including the operation order of magnitude. Moreover, the integrator also
∂g,i and its inverse with addition and multiplica- affords a rigorous and sharp estimate of all inte-
tion leads to a differential algebra (DA). Details gration errors [6, 7, 8, 9].
on this particular DA can be found in [10, 24]. In Similarly, it is also possible to solve PDEs in
passing it is worth mentioning that the DA struc- finitely many steps. For this purpose, one elimi-
tures also provide a novel way to a theory of dif- nates differentiation with respect to one variable
ferentials as infinitely small numbers [10]. For by integration. For example, the PDE
practical work with DA, care has to be taken to
∂ ∂V ∂ ∂V ∂ ∂V
provide elementary operations +, ·, ∂g,i that work a1 (a2 )+b1 (b2 )+c1 (c2 )=0
to any order and any number of variables. Since ∂x ∂x ∂y ∂y ∂z ∂z
(4)
usually many derivatives vanish due to symmetry,
which describes the scalar potential in particle op-
the active support of sparsity is essential [23].
tical relative coordinates, is rewritten as
( %
Solutions of ODEs and PDEs To determine the ( 1 ∂V ((
map Eq.(1), it is necessary to solve the ODEs V =V( + ( (5)
y=0 b2 y ∂y y=0
Eq.(2). Since usually fields are known only in &
a1 ∂ ∂V c1 ∂ ∂V
the midplane, it is often also necessary to solve − (a2 )+ (c2 )
their PDEs in the process. The crudest approach y b1 ∂x ∂x b1 ∂z ∂z
to solve Eq.(2) is to replace all operations in a and again, iteration converges to the exact result in
tracking code by the corresponding ones in DA finitely many steps [10]. In addition to producing
[12, 11]. This replacement is similar in flavor general PDE solvers, this approach allows the ex-
to the modification of existing code through ecution of the commonly performed out-of-plane
pre-processors performing “automatic differenti- expansion to arbitrary order [25].
106
Ch.2: BEAM DYNAMICS
It is important to observe that the accuracy of Mn . Then we have

the out-of-plane expansion rests on the quality of
In = (M1 + M∗n ) ◦ Nn
field derivatives in the midplane. If the field arises
from models, DA methods can directly be used to = M1 ◦ Nn + M∗n ◦ Nn ⇒
obtain these derivatives accurately to any order of Nn = M−1 ∗
1 (I − Mn ◦ Nn ) (7)
interest. if on the other hand, the field is based on
a fixed point problem for Nn . Beginning iteration
measurements, it is highly advantageous to repre-
with Nn = In yields convergence to the exact re-
sent the field in space through integrals over sur-
sult in n steps because M∗n is purely nonlinear.
face fields based on the Helmholtz theorem which
The map of the reversion [27, 10] of a system, i.e.
resembles the Cauchy formula in complex anal-
the system traversed backwards, is related to the
ysis. DA methods can be used elegantly to ex-
inverse of the map; but since angles change di-
pand fields simultaneously in the coordinates in

rections when going backwards, the reversed map
the midplane, as well in the surface coordinates,
Mr is Mr = R ◦ M−1 ◦ R, where R is the linear
and the integration is carried out directly utilizing

map that changes the signs of angles.
the DA operation ∂ −1 .
For ODEs that are time independent and for Symplectic tracking with maps Inverse maps
which z = 0 is a solution, which is the case in par- can also be used to establish mixed-variable gen-
ticle optical relative coordinates, another method erating functions that represent the map, which
can be applied. For a given function on phase are useful for symplectic tracking. Beginning
space g (z, s) , it is possible to obtain its deriva- from the n-th order map Mn , we create Nn,1 ,
tive along the true solution via consisting of the position part of Mn , and
an identity for the momenta; thus ( qf , pi ) =
d + ∂ g = L g Nn,1( qi , pi ). Similarly we construct Nn,2 via
g (z, s) = f · ∇g (6)
f −1
ds ∂t (qi , pf ) = Nn,2( qi , pi ). Inversion of Nn,1 (if pos-
Apparently we also have dn /dsn g = Ln g. If −1
f sible) yields ( qi , pi ) = Nn,1 ( qf , pi ), and compo-
g is not explicitly time dependent, the operator sition with Nn,2 from the right yields
Ln can be evaluated directly within DA for any −1
f qi , pf ) = Nn,2 ◦ Nn,1
( (
qf , pi ) (8)
n; using it for g = zν , the components of the
vector z, we obtain an integrator of adjustable or- From this mixed-variable representation, the un-
der. This method is utilized for the main fields derlying generating function can be obtained by a
of elements in COSY; an element is typically tra- mere integration along a suitable path, again us-
versed in one step, and orders of 25-30 are usu- ing the DA operation ∂ −1 ; in a similar way as the
ally chosen to obtain integration to nearly ma- case shown here, also all the other three common
chine precision. Note that for certain particle op- generating functions can be determined.
tical systems, including the notoriously improp- However, in practice it is frequently observed
erly treated fringe fields, there are other efficient that different symplectification schemes result in
and fast perturbative methods to obtain approxi- different tracking results; so it is highly desirable
mate solutions of high accuracy, like the method to achieve symplectification with minimal modi-
of symplectic scaling [26]. fication of the prediction of Taylor transfer map.
Advanced DA operations and manipulation of Utilizing Hofer’s metric on spaces of Hamiltoni-
maps Given the nth order representations of ans, this can actually be achieved in a unified way
two maps Mn and Nn , it is possible to determine using DA techniques, resulting in the so-called
the nth order representation of their composition EXPO symplectification scheme [28].
Mn ◦ Nn as long as Nn has no constant parts. To Normal forms Another important manipulation
this end, one simply inserts the components of Nn of maps is the transformation to normal form
into the polynomial defined by the coefficients of [10]. The first step is the transformation to the
Mn . This allows to determine the map of a com- parameter-dependent fixed point z(δ) which sat-
bined system from the maps of the subsystems. isfies (z(δ), δ) = Mn (z (δ), δ). Subtracting the
It is also possible to obtain the nth order rep- non-parameter identity map Inz on both sides we
resentation Nn of the inverse of Mn as long as have (0, δ) = (Mn − Inz )(z (δ), δ), and thus
its linear part is invertible. To this end, one writes
Mn = M1 + M∗n , where M1 is the linear part of (z (δ), δ) = (Mn − Inz )−1 (0, δ) (9)
107
from which we read off z(δ) in the non-parameter Some other DA-based methods not discussed
lines. above shall be mentioned briefly. For the analy-
The linear part of the normal form algorithm sis of spin dynamics, there are methods that allow
itself is based on a study of the eigenvalue spec- the computation of the spin map, spin tracking, as
trum of the map; if they are distinct and complex, well as invariant spin axis n̂ There are also exten-
then it amounts to a diagonalization to the form sions of the DA methods to allow for a mathemat-
R, where the phases of the diagonal terms cor- ically rigorous treatment of the remainder terms
respond to the tunes. The nonlinear part of the of Taylor’s formula [6], which allow for the de-
normal form algorithm consists of a sequence of velopment of rigorous error bounds for integra-
transformations Sn that to order n have the form tors [7, 8] and also a rigorous bounding of stabil-
Sn = I + Sn . The orders higher than n in Sn can ity times of nonlinear motion [29, 6] when com-
in principle be picked freely, and frequently they bined with methods of verified global optimiza-
are chosen so that all transformations Sn belong tion [6]. There are methods for the solution of al-
to the same symmetry group as the original map. gebraically constrained differential equations, so-
Up to order n, we have Sn−1 =n I − Sn ; let- called differential algebraic equations [30], and
ting Nn = R + On denote the map that has been methods to develop high order versions of the Fast
normalized to order n, we have Multipole Method for 3-D space charge computa-
Nn = Sn ◦Nn−1 ◦Sn =n (I+Sn )◦Nn−1 ◦(I−Sn ) tion [31].
=n R + On + {Sn ◦ R − R ◦ Sn } (10) Applications The differential algebraic meth-
Apparently, the Sn in the commutator {Sn ◦ R − ods form the core tools for the computation of
R ◦ Sn } can now be chosen to remove nth or- maps and aberrations for a large number of codes.
der terms in On . In the symplectic case, it turns First developed and used in the code COSY IN-
out that the remaining terms are just those that FINITY [13] which currently has about 1,500 reg-
describe motion on circles, with a frequency that istered users, the methods also form the core en-
depends on the radius and possible parameters δ, gines in many other codes [14, 15, 16, 17, 18].
allowing direct computation of amplitude depen- The methods have been used for design, op-
dent tune shifts. timization and analysis of numerous beam opti-
The radius vectors in each phase space pair cal devices, of which we only showcase a limited
represent invariants, the quality of which is di- number for reasons of space. Maps are being used
rectly determined by the magnitude of the coef- for the efficient symplectic tracking of dynamics
ficients of Sn , the so-called resonance strengths. in synchrotrons resulting in speed increases for
Trying to minimize their size is an efficient way the LHC [32] and the Tevatron; at the latter, they
to enhance the overall nonlinear behavior of the have also been used for purposes of linear decou-
system. pling [33] and a rigorous estimate of long-term
Other DA-based tools It is also possible to ob- stability [31]. Another important application is
tain Lie factorizations, the tracking and correction based on more sophis-
ticated models for fringe fields or other nonlin-
Mn =n {M1 exp(: f3 :) exp(: f4 :)...}I earities in large acceptance rings [34], symplec-
or Mn =n {... exp(: f4 :) exp(: f3 :)M1 }I tic tracking in light sources under consideration
(11) of wigglers and undulators, all the way to vari-
of symplectic maps Mn [10], where =n means ous applications for novel accelerator types like
two functions are equal up to order n; in fact, the FFAG.
the DA approach currently represents the only The methods have also been used for the
method to obtain them to arbitrary order. The pro- design and correction of fragment separators
cedure is order-by-order; in the ith step, all orders and particle spectrographs, including the MSU
less then i have already been taken care of, and FRIB, S800 and A1900, those at TJNAF, Cata-
the problem is reduced to finding fi+1 such that nia, RIKEN and OSAKA. In addition to com-
exp(:fi+1 :)I = I + ∇f i+1 · Jˆ agrees with the putation and hardware correction, a particularly
given map to order i. This is an integration prob- important application is the reconstructive cor-
lem similar to the case of the generating functions rection of aberrations in high resolution spectro-
above, which is resolved once more with the DA graphs [35]. This technique uses combinations of
operator ∂ −1 . the map and its inverse to eliminate aberrations
108
Ch.2: BEAM DYNAMICS
by measurement data in two planes, i.e. knowl- [25] S. Manikonda, PhD thesis, Michigan State U.,
edge of (xf , yf , af , bf ). Since this is uniquely de- 2006
termined as a function of (yi , δi , ai , bi , ), utilizing [26] G. Hoffstaetter, M. Berz, PR E54, 4
map inversion techniques discussed above yields [27] W. Wan, M. Berz, PR E54, 3, 2870 (1996)
reaction energy and angles to precisions that are [28] B. Erdélyi, M. Berz, PRL 87, 11, 114302 (2001)
sometimes two orders of magnitude higher than [29] K. Makino et al, ECMI Newsletter 39
without correction. [30] J. Hoefkens, PhD thesis, Michigan State U.
Finally, over the last years, use is also be- (2001)
ing made of the methods in electron microscopy, [31] H. Zhang, M. Berz, NIM submitted
where they afford the ability to compute and cor- [32] M.L. Shashikant, M. Berz, B. Erdélyi, IOP CP
175, 299 (2002)
rect high-order aberrations when the fields of the
[33] P.V. Snopok et al, NIM 558, 142 (2005)
devices are known.

[34] M. Berz, B. Erdélyi, K. Makino, PRST-AB 3,
124001 (2000)
References
[35] M. Berz et al, PR C 47, 2, 537 (1993)
[1] K.L. Brown, SLAC Tech. Rep. 91 (1979)
[2] T. Matsuo et al, Mass Spectrometry 24
2.3.8 Symplectic Integration Methods
[3] H. Wollnik et al, AIP CP 177 (1988) 74
[4] A.J. Dragt et al, IEEE Tran. Nuc. Sci. NS-3, 5,
H.Yoshida, Nat. Astron. Obs., Japan
2311 (1985)
The exact time evolution (solution) of a Hamilto-
[5] M. Berz et al, NIM A258 (1987) 402
nian system
[6] K. Makino, PhD thesis, Michigan State U.,
MSUCL-1093 (1998) dq ∂H dp ∂H
[7] M. Berz, K. Makino, Reliable Computing 4 (4) = , =− (1)
dt ∂p dt ∂q
(1998) 361
[8] K. Makino, M. Berz, Int. J. Diff. Eq. & Appl. 10, is symplectic [1, 2, 3]. This means that the map-
4 (2005) 353 ping from (q(0), p(0)) to (q(t), p(t)) along the
[9] K. Jackson, Proc. 5-th Int. Workshop on Taylor solutionconserves the symplectic 2-form dp ∧
Models, in preparation dq ≡ i dpi ∧ dqi exactly. This symplectic 2-
[10] M. Berz, Modern Map Methods in Particle Beam form represents the sum of oriented areas of phase
Physics, Academic Press, San Diego, 1999; also space element projected onto the plane of (pi , qi ),
available at http://bt.pa.msu.edu/pub, together and for 1-D systems, “symplectic” is identical to
with more recent publications. “area-preserving” property. Traditional integra-
[11] M. Berz, PA 24, 109 (1989) tion methods, such as the Euler method and the
[12] M. Berz, NIM A258, 431 (1987) classical Runge-Kutta method, do not respect this
[13] M. Berz et al, http://cosyinfinity.org fact and after a long term integration, fictitious
[14] Y. Yan, AIP Conf. 297 (1993) p.279 damping or excitation occurs, making the result
[15] J. van Zeijts, AIP Conf. 297 (1993) p.285 of integration unreliable. One needs integration
[16] L. Michelotti, Fermilab Tech. rep. (1990)
methods which are designed to keep the sym-
[17] W. Davis et al, Proc. Workshop on Nonlinear Ef-
plectic property of the original Hamiltonian flow.
fects in Accelerators, 1993
These integration methods are in general called
[18] F.C. Iselin, AIP Conf. 297 (1993) p.325
[19] J.F. Ritt, Integration in Finite Terms - Liouville’s
the symplectic integrators (symplectic integration
Theory of Elementary Methods, Columbia U. methods). With use of symplectic integrators the
Press, 1948 error of energy does not grow monotonically. This
[20] J.F. Ritt, Differential Algebra, APS, Washington, comes from the existence of a conserved quantity
DC, 1950 which is close to the original Hamiltonian [3, 4].
[21] E.R. Kolchin, Differential Algebraic Groups,
Academic, NY, 1985 2.3.8.1 Methods of realization
[22] R.H. Risch, APS Bulletin 76, 605 (1970)
[23] M. Berz, Forward algorithms for high orders and There are explicit symplectic schemes and im-
many variables, Automatic Differentiation of Al- plicit schemes for general Hamiltonian systems.
gorithms: Theory, Implementation and Applica-
tion SIAM Implicit schemes for general Hamiltonian
[24] M. Berz et al, SIAM 1996 systems The classical 4th order Runge-Kutta
109
method is generalized as follows. For a given sys- there exist explicit symplectic algorithms. The
tem of ordinary differential equations, idea to construct explicit schemes is simply to
dz approximate the original Hamiltonian flow by a
= f (z) (2) composition of trivial symplectic mappings,
dt
∂T
coefficients ki are determined by the simultaneous q = q + ci τ , p = p (9)
algebraic equations ∂p
⎛ ⎞
s and
∂V
ki = f ⎝z + τ aij kj ⎠ (3) q = q, p = p − di τ (10)
j=1
∂q
repeatedly. Here constants (ci , di ), (i = 1, ...k)
(i = 1, ..., s). Then the one-step mapping z → z
are determined so that the composed mapping

is defined by (τ = Δt) (q, p) → (q , p ) coincides with the Taylor series

s expansion of the solution up to the desired order.
z = z + τ bj kj (4) The simplest, 1st order scheme (q, p) →
j=1 (q , p ) is given by

Here, aij and bj are scalar constants which char- ∂T ∂V
acterize the scheme. This scheme is called the Im- q = q + τ , p = p − τ
∂p p=p ∂q q=q
plicit Runge-Kutta (IRK) method in general. (11)
If the given system is a Hamiltonian system The 2nd order scheme (q, p) → (q , p ) is attained
(1), and if the constants aij and bj satisfy the con- by
ditions
τ ∂T ∂V
Mij ≡ bi aij + bj aji − bi bj = 0, (1 ≤ i, j ≤ s) q∗ = q + , p = p − τ ,
2 ∂p p=p ∂q q=q∗
(5)
identically, then the mapping (4) is known to be τ ∂T
symplectic [5, 2]. An implicit symplectic integra- q = q∗ + (12)
2 ∂p p=p
tion method is thus obtained.
The simplest case (s = 1) which satisfies the and this scheme is known as the leap-frog method.
conditions (5) is given by a11 = 1/2, b1 = 1, This problem to derive explicit symplectic in-
and we have the scheme, tegrators is well reformulated in terms of Lie al-
τ
gebraic language (Sec.2.3.6). First rewrite the
k1 = f z + k1 , z = z + τ k1 Hamilton equation (1) in the form
2
dz
or, more concisely, = [z, H(z)] (13)
dt
z + z
z = z + τ f (6) where brackets stand for the Poisson bracket.
2 If we introduce a differential operator :G: by
which is known as the implicit midpoint rule. This :G:F ≡ [G, F ], then (13) is written as ż =
scheme has 2nd order accuracy. For s = 2, an −:H:z, so the formal exact solution of z(t) from
example is given by t = 0 to t = τ is given by
√
z(τ ) = [exp(−τ :H:)]z(0) (14)
1 1
− 3
(aij ) = 1 4 √3 4 1 6 , (bj ) = 12 12 For a Hamiltonian of the form (8), we have
4 + 6 4
(7) z(τ ) = exp[τ (A + B)]z(0) (15)
and this scheme has order 4 accuracy.
For general Hamiltonian systems, it is also where A ≡ −:T : and B ≡ −:V :. Operators A
possible to design symplectic schemes using the and B do not commute in general.
“generating function” [6]. Suppose (ci , di ) is a set of constants which
satisfies exp[τ (A + B)] =
Explicit schemes for special Hamiltonian sys-
tems For Hamiltonians of the form -
k
exp(ci τ A) exp(di τ B) + O(τ n+1 ) (16)
H = T (p) + V (q) (8) i=1
110
Ch.2: BEAM DYNAMICS
for a given integer n, which corresponds to the 2.3.8.2 Symplectic method vs. nonsymplectic
order of the integrator. Then consider a mapping method
from z = z(0) to z = z(τ ), given by
k For the harmonic oscillator, H = 12 (p2 + q 2 ), the
-
Euler method (nonsymplectic) gives the mapping
z = exp(ci τ A) exp(di τ B) z (17)
i=1 q 1 τ q
This mapping is symplectic because it is a product = (25)
p −τ 1 p
of elementary symplectic mappings, and approx-
imates the exact solution (15) up to the order of and one finds easily that at each step, the value of
τ n . Furthermore (17) is explicitly computable. In energy is multiplied by (1+τ 2 ), i.e., (p 2 +q 2 ) =
fact (17) gives the succession (1+τ 2 )(p2 +q 2 ). On the other hand, the 1st order
ofthe mappings
symplectic scheme (11) gives the mapping
∂T
qi = qi−1 + τ ci
∂p p=pi−1 q 1 τ q
p =
−τ 1 − τ 2 p
(26)
∂V
pi = pi−1 − τ di (18)
∂q q=qi and the energy does not grow monotonically. This
for i = 1 to i = k, with (q0 , p0 ) = z and is because (26) has the exact conserved quantity
(qk , pk ) = z . An n-th order symplectic integra- 1 τ
tor is thus realized. For example, when n = 1, a H̃ = H + τ H1 = (p2 + q 2 ) + pq (27)
2 2
trivial solution is c1 = d1 = 1, (k = 1), which
and the numerical solution cannot deviate from
corresponds to the identity
the true solution curve. For general Hamiltonians,
exp[τ (A + B)] = exp(τ A) exp(τ B) + O(τ 2 ) there exists always a conserved quantity
(19)
and gives the 1st order integrator (11). When n = H̃ = H + τ H1 + τ 2 H2 + τ 3 H3 + · · · (28)
2, we find a solution c1 = c2 = 1/2, d1 = 1, d2 = which is given as a formal power series of τ , as far
0, (k = τ2),i.e. exp[τ (A + B)] = as symplectic algorithm is employed. This pre-
τ
exp A exp(τ B) exp A + O(τ 3 ) (20) vents the occurrence of fictitious damping or ex-
2 2 citation, which is common when a nonsymplectic
and yields the 2nd order integrator (12).
Composition method to obtain higher order method is used.
schemes Once a lower order scheme is given, a
higher order scheme is obtained by a composition References
of lower order ones [7, 3]. [1] E. Hairer, S.P. Nørsett, G. Wanner, Solving
For example, let S2 (τ ) be a 2nd order sym- Ordinary Differential Equations (2nd rev. ed.),
plectic scheme, (12) or (6). Then a 4th order in- Springer (1993)
tegrator S4 (τ ) is obtained by a symmetrical com- [2] J.M. Sanz-Serna, M.P. Calvo, Numerical Hamilto-
position of 2nd order ones as nian Problems, Chapman & Hall (1994)
S4 (τ ) = S2 (x1 τ ) ◦ S2 (x0 τ ) ◦ S2 (x1 τ ) (21) [3] H. Yoshida, Celest. Mech. 56 (1993) 27
with [4] G. Benettin, A. Giorgilli, J. Statistical Phys. 74
−21/3 1 (1994) 1117
x0 = , x1 = (22)
2 − 21/3 2 − 21/3 [5] J.M. Sanz-Serna, BIT 28 (1988) 877
In order to have a 6th order integrator, S6 (τ ), [6] P.J. Channell, J.C. Scovel, Nonlinearity 3 (1990)
composition of 7 2nd order one is necessary: 231
S6 (τ ) = S2 (y3 τ ) ◦ S2 (y2 τ ) ◦ S2 (y1 τ ) ◦ S2 (y0 τ ) [7] H. Yoshida, PL A150 (1990) 262
◦S2 (y1 τ ) ◦ S2 (y2 τ ) ◦ S2 (y3 τ ) (23)
with numerical constants 2.3.9 Dynamic Aperture
y1 = −1.1776 79984 17887 A. Wolski, U. Liverpool
y2 = 0.23557 32133 59357
As a particle travels around a storage ring, the am-
y3 = 0.78451 36104 77560 plitude of its motion with respect to the closed or-
y0 = 1 − 2(y1 + y2 + y3 ) (24) bit may increase due to certain dynamical effects.
Much higher order integrator can be obtained by a The dynamic aperture may be defined in terms of
composition of lower order ones in a similar way. the variation in amplitude over time: a particle is
111
said to be outside the dynamic aperture of the stor- effects: first, the elements will cause oscillations
age ring if its amplitude exceeds some large value about the closed orbit to grow in amplitude for
after a sufficient number of turns. particular values of the tunes (resonances); sec-
Although it usually refers to an idealized or ond, the tunes will depend on the amplitudes of
simplified system, dynamic aperture is a useful the oscillations. The result will be a limited dy-
concept because it characterizes dynamical ef- namic aperture. In addition, ripple in magnet
fects that may limit the acceptance of a storage strengths due to power supply defects will affect
ring. The acceptance refers to the range of trajec- the dynamic aperture through the existing nonlin-
tories of all particles that can remain indefinitely earities.
in a storage ring, taking into account physical lim- Nonlinearities can come from a variety of
itations (e.g. vacuum chamber aperture) as well as sources. First, the equations of motion for par-
dynamical effects, and is important for injection ticles in “linear” elements such as dipoles and
efficiency and beam lifetime. quadrupoles are only linear under certain approx-
Although analytic formulae have been devel- imations. Second, nonlinear elements such as
oped for some cases [1, 2], dynamic aperture is sextupoles may be deliberately introduced to cor-
generally computed using long-term (many hun- rect such effects as chromaticity (chromatic sex-
dreds, or thousands of turns) tracking. Usually, tupoles are often the dominant limitation on dy-
only single-particle dynamical effects are con- namic aperture). Third, all magnets in practice
sidered, but collective effects (e.g. beam-beam have fields with higher-order components. These
forces, Sec.2.5.1) may be included. Longitudinal components may be intrinsic in the design (sys-
dynamics may be included in different ways: they tematic errors), or the result of variations in the
may be ignored altogether; it may be assumed construction (random errors).
that the energy deviation is constant; or, an accu- The behaviour of nonlinear systems can be
rate model including synchrotron motion may be very complicated, but some understanding in the
used. For lepton storage rings, synchrotron radia- case that the nonlinearity is a perturbation ap-
tion (Sec.3.1) can have a significant impact on the plied to an integrable system is provided by the
dynamic aperture, because of damping and quan- Kolmogorov-Arnol’d-Moser (KAM) theorem [3,
tum excitation. However, including radiation ef- 4]. The KAM theorem states that for a weak
fects in the model can significantly slow the track- perturbation, periodic trajectories sufficiently far
ing, so such effects are often neglected or checked from resonant frequencies will remain stable (al-
only once at the end. beit with modified frequencies) while other tra-
The main issues associated with dynamic jectories close to resonance become unstable
aperture are as follows. First, the computa- (Arnol’d diffusion [4]). The Chirikov criterion [5]
tional results can have a strong dependence on states that where resonances overlap, particle mo-
the model (which dynamical effects are included, tion becomes chaotic.
what boundaries are used to determine unstable
Computing dynamic aperture In principle,
trajectories, number of turns tracked, etc.) Sec-
any tracking code can be used to compute the dy-
ond, long-term tracking can be computationally
namic aperture of a storage ring. Tracking codes
expensive. Third, it is difficult to represent the dy-
in common use include: MAD [6], DIMAD [7],
namic aperture (which, in its most general sense,
BMAD [8], SIXTRACK [9], LEGO [10], Merlin
refers to a set of points in 6-D phase space) in a
[11], and AT [12]. The first four of these pro-
complete and meaningful way. Finally, although
grams follow instructions prepared in a command
some techniques have been developed for iden-
file. LEGO and Merlin are C++ class libraries. AT
tifying and minimizing limitations on dynamic
is a set of tracking and beamline analysis routines
aperture, the problem of optimizing the dynamic
working within Matlab.
aperture subject to other design constraints has
A basic estimate of dynamic aperture can be
not been completely solved.
obtained in two steps. First, the lattice is defined
Factors affecting dynamic aperture A storage with the required configuration (magnet strengths,
ring lattice in which the equations of motion are rf voltage, magnetic field and alignment errors
perfectly linear and lead to stable oscillations etc.) Second, particles are tracked over a range of
has infinite dynamic aperture. The introduction initial conditions. The initial conditions of those
of nonlinear elements will generally have two particles whose trajectories stay within specified
112
Ch.2: BEAM DYNAMICS
hundreds of thousands of turns. It is also possi-

ble to estimate the dynamic aperture from a sur-
vival plot, showing the number of turns a parti-
cle survives, as a function of the initial (coordi-
nate) amplitude: the data points can often be fit-
ted by a line asymptotic to some amplitude, below
which particles remain in the storage ring indefi-
nitely [14]. For long-term tracking, accuracy is
important, since small errors can become signifi-
cant over many turns.
Since computation of the dynamic aperture
usually includes only single-particle effects, par-

allelization can reduce the computation time in
Figure 1: Dynamic aperture computation for SPEAR3 proportion to the number of processors available.
[13], for six seeds of machine errors. Solid lines: zero Another way to speed the computation is to rep-
energy deviation. Dashed line: 3% energy deviation. resent groups of elements (or even the entire lat-
tice) by a single dynamical map [15]. However,
computation of the map can be time-consuming,
limits (usually taken at the limit of validity for the
since it is often necessary to include terms to very
maps used for individual components) for a spec-
high order to achieve sufficient accuracy. Unless
ified number of turns are deemed to lie within the
the maps are symplectic (i.e. the symplectic er-
dynamic aperture.
ror is small, compared to physical nonsymplec-
Performing the tracking to carry out a thor- tic effects such as radiation) then artificial growth
ough exploration of phase space can be very time or damping of oscillation amplitudes can lead to
consuming, and can be a significant issue if many a poor prediction of the dynamic aperture. To
lattice configurations (and sets of random errors) enforce symplecticity, it is possible to use sym-
need to be studied. The computation time can be plectic representations of dynamical maps, such
reduced either by reducing the size of the prob- as Lie transforms, or mixed-variable generating
lem, or by using techniques to speed the tracking. functions (Secs.2.3.6, 2.3.8); however, this adds
The most common way to reduce the size of complexity, and such maps can be difficult or ex-
the problem is to limit the volume, or the dimen- pensive to compute (and evaluate) to sufficiently
sionality (or both) of phase space used for initial high order.
conditions. For example, a search can be made for
the largest initial amplitudes in transverse coordi- Measuring dynamic aperture An experimen-
nate space, for which the subsequent trajectories tal estimate of the dynamic aperture in an oper-
are stable (Fig.1). However, there is a risk that ating storage ring may be obtained by observing
the search procedure may miss points within the the particle loss that results from kicking a bunch
final boundary that lead to unstable trajectories: to excite coherent oscillations at different ampli-
note that the dynamic aperture is not necessarily a tudes. It can be difficult to distinguish dynamical
connected region in phase space. effects from the effects of physical apertures in
In the case that a storage ring lattice includes such experiments, although use of adjustable col-
only normal (i.e. not skew) multipole magnets, limators can give information. Studies (e.g. at
then the equations of motion are symmetric un- ESRF [16]) have shown that there can be good
der the transformation y → −y (though not un- agreement between the measured and computed
der x → −x), where y is the vertical coordinate. dynamic apertures, if a sufficiently complete and
In that case, tracking is necessary to cover only detailed model is used for the computation. See
y > 0. also Sec.2.3.12.
If damping processes are present (e.g. syn- Characterizing the underlying dynamics
chrotron radiation, or adiabatic damping during During the design process, it is often impor-
acceleration), then there is little value in tracking tant to understand the reasons for the dynamic
for more than one or two damping times. How- aperture limits. This involves characterizing the
ever, in hadron storage rings operating at fixed underlying dynamics at a more detailed level,
energy, it may be necessary to track for tens or in particular to understand the tune shifts with
113
values of the coordinates, and then determining

the frequency of oscillation around the closed or-
bit with high precision. Strictly speaking, the
tunes are only defined for regular orbits, lying on
an invariant torus (an N -dimensional closed sur-
face in 2N -dimensional phase space). However,
numerical techniques can be used to determine os-
cillation frequencies even for quasi-periodic mo-
tion. For example, using numerical analysis of
the fundamental frequencies (NAFF [21]), one
searches for the quantities ν and φ0 that maxi-

mize the sum N n

n=1 xn cos(2πν N + φ0 ), where
xn is the coordinate on turn n. Applying NAFF
to a range of initial coordinates, and plotting the

horizontal and vertical frequencies on a tune grid
produces a frequency map, Fig.2. Tune shifts with
amplitude are visible from the area of tune space
covered. Resonance strengths can be seen from
the clustering of points around resonance lines.
The “diffusion rate” in tune space (the rate of
change of the tunes over the period of the track-
ing) can be indicated on a color scale. Regions
Figure 2: Frequency Map Analysis for the Advanced of poor stability (where small errors are likely to
Light Source [22]. lead to trajectories becoming unstable) are char-
acterized by high diffusion rates, and are often as-
sociated with the intersection of resonance lines
amplitude, and the strengths of resonances. A in tune space. One of the strengths of FMA lies
phase space portrait (showing the coordinates and in the indications it gives for ways to improve the
momenta of a set of particles on successive turns) dynamic aperture; for example, it may be possible
can give a simple indication of these features. to avoid a strong resonance simply by making a
Nonlinear effects appear as distortions of the set change in the working point in tune space (though
of ellipses that would be produced in a linear lat- such a change may also cause changes in the tune
tice; resonances may appear as “islands”, though shifts with amplitude that lead to particles at large
this depends on the details of the dynamical map. amplitudes crossing other resonances).
Phase space portraits produce mostly qualitative Dynamical stability can also be characterized
information, and they may be difficult to interpret by calculating the Lyapunov exponent, which de-
in the presence of coupling between different scribes the rate at which two initially close trajec-
degrees of freedom. tories diverge from each other [14].
The limits on the dynamic aperture may also Optimizing dynamic aperture A design goal
be investigated by inspecting the strengths of var- for any storage ring is to achieve sufficient accep-
ious nonlinear terms in the dynamical map for a tance to meet injection efficiency and lifetime cri-
complete turn. The map may be computed, e.g., teria. The dynamic aperture must be large enough
using a differential algebra code (e.g. COSY [17], to meet the required acceptance, with some mar-
Sec.2.3.7) or a Lie algebra code (e.g. Zlib [18], gin provided e.g. in case magnet errors are larger
or MARYLIE [19], Sec.2.3.6). The strengths of than expected. Despite various attempts to de-
tune shifts and resonance driving terms may be velop systematic procedures for improving the
deduced from different terms in the map. dynamic aperture subject to various other design
One of the most powerful and widely-used constraints, a definitive technique has not yet been
techniques for characterizing the dynamics under- produced. It is usually necessary to approach
lying the dynamic aperture is frequency map anal- each lattice on an individual basis, although there
ysis (FMA [20]). FMA involves tracking a single are some general rules and principles that can be
particle over many turns, starting with specified applied.
114
Ch.2: BEAM DYNAMICS
One key aspect is the choice of working point: [2] J. Gao, NIM A451 (2000) 545
this can be selected, for example, based on the [3] V.I. Arnol’d, Mathematical Methods of Classical
results of FMA. Alternatively, a tune scan can be Mechanics, Springer (1978)
performed, with the size of the dynamic aperture [4] M. Tabor, Chaos and Integrability in Nonlinear
at different working points represented by points Dynamics: An Introduction, Wiley (1989)
of different sizes or colors on a tune diagram (a [5] B.V. Chirikov, J. Nucl. Energy, Part C Plasma
swamp plot). Phys. 1, 253 (1960)
If the tune shifts with amplitude are large and [6] http://mad.web.cern.ch/mad/
there are strong resonances, it may not be possi- [7] http://www.slac.stanford.edu/accel/ilc/codes/dimad/
ble to achieve sufficient dynamic aperture at any [8] http://www.lepp.cornell.edu/ dcs/bmad/
[9] F. Schmidt, CERN SL/94-56 (AP), 1994
working point. Then, it is necessary to reduce
[10] Y. Cai et al, PAC 1997, 2583
the combined strengths of the nonlinear compo-

[11] http://www.desy.de/ merlin/
nents in the lattice. The need for strong chromatic
[12] http://www-ssrl.slac.stanford.edu/at/
sextupoles can be avoided by designing the linear [13] J. Corbett et al, PAC 1999, 2364
lattice for low chromaticity (small beta functions [14] M. Giovannozzi, W. Scandale, E. Todesco, AIP
and quadrupole strengths), and by locating the Proc. 395 (1997) 243
sextupoles at appropriate points (high dispersion, [15] A.W. Chao et al, PAC 1991, 336
good separation between the beta functions) to [16] A. Ropert, PAC 1998, 921
maximize their chromatic effects. In some cases, [17] http://bt.pa.msu.edu/index cosy.htm
it may be possible to design the chromatic cor- [18] Y.T. Yan, PAC 1991, 333
rection so that sextupoles occur in pairs with π [19] http://www.physics.umd.edu/dsat/dsatmarylie.html
phase advance (in x and y) between them (a −I [20] D. Robin et al, PRL 85 (2000) 558
transformer) [23]; in that case, for particles with [21] J. Laskar, Icarus 88 (1990) 266
zero energy deviation, the geometric aberrations [22] C. Steier et al, EPAC 2000, 1077
(tune shifts with amplitude, resonance terms) [23] Y. Cai et al, PAC 1995, 576
cancel out.
Chromaticity can make it particularly diffi-
cult to achieve good dynamic aperture in low- 2.3.10 Decoherence
emittance storage rings (e.g. for third-generation M.A. Furman, LBNL
synchrotron light sources, Sec.2.2.4) or colliders.
Such machines generally require low dispersion If a stored beam is kicked transversely by angle
and low beta functions in the dipoles; if the lattice Δx (or is injected offset), its centroid betatron
also needs to be compact, then it can be difficult signal decoheres due to betatron tune spread. We
to provide good locations for sextupoles for chro- define q ≡ x/σx and p ≡ (αx x + βx x )/σx
matic correction. Colliders generally need very where αx , βx and σx are the lattice functions
low beta functions at the interaction points, and and rms beamsize at the observation point (=kick-
the beta functions can reach very large values in point). We assume: (1) the beam is Gaussian in
the final quadrupoles, resulting in high chromatic- (x, x ) and in the relative momentum deviation
ity. δ ≡ ΔP/P0 ; (2) there is no x-y coupling and no
It may be beneficial to include nonlinear com- synchro-betatron coupling; (3) damping, quantum
ponents in the lattice (e.g. sextupoles at lo- excitation and the mutual interactions of the parti-
cations with low or zero dispersion; octupoles cles can be ignored; and (4) the tune dependence
etc.) specifically to provide control over the on amplitude and energy offset is
nonlinear dynamics. However, reducing certain ν = ν0 − μ(q 2 + p2 ) + ξδ (1)
nonlinear terms in the map often comes at the
cost of introducing additional higher-order terms, where ξ = chromaticity. Then the time evolution
that may lead to an overall reduction in dynamic of the beam centroid is [1, 2]
aperture. iZF (n)
q + i p =
(1 − iθ)2
References
Z 2 iθ
× exp −2πinν0 + (2)
[1] G. Guignard, J. Hagel, Lecture Notes in Physics 2 1 − iθ
247, 367, Springer (1986)
115
where n = turn number, θ = 4πμn, Z = of head-tail damping at the TRISTAN MR is an-

βx Δx /σx and the chromatic form factor alyzed in [8]. A full 3-D analysis is provided in
[9], and was used as a tool to measure the emit-
ξσδ 2 2 tance in the TRISTAN ring. An application to the
F (n) = exp −2 sin (πnνs ) (3)
νs Photon Factory is presented in [10], and to a ver-
sion of the muon collider in [11]. A more general
where νs = synchrotron tune. The second mo- discussion is provided in [12].
ments after the kick are
⎛. 2 /⎞ ⎛⎞
q 2
1 References
⎝qp⎠ = 1 + Z ⎝0⎠
. 2/ 2 [1] R.E. Meller et al, SSC-N-360, 1987
p 1
⎛ ⎞ [2] M.G. Minty et al, PAC95, p.3037

2Z 2 θ 2 ⎝− cos ψ⎠
[3] I.C. Hsu, PA 34, 43 (1990)
Z 2 F 2 (n)
+ exp − sin ψ (4) [4] S.Y. Lee, Proc. Int. Workshop on Nonlinear
2(1+4θ 2 )3/2 1+4θ 2 cos ψ Problems in Acc. Phys., Berlin, 1992, Inst. Phys.
Conf. Series, p.249
Z 2θ
ψ ≡ 4πnν0 − − 3 tan−1 2θ (5) [5] H. Moshammer, PR E48(3), 2140 (1993)
1 + 4θ 2 [6] J. Shi, S. Ohnuma, PAC 93, p.3603
[7] G.V. Stupakov, A.W. Chao, PAC 95, p.3288
The
. /normalized rms size is . σ/x (n)/σ
. /x (0) = [8] N. Akasaka, S. Kamada, EPAC 96, p.1141
( q 2 − q2 )1/2 . Note that q 2 + p2 = 2 + [9] S. Kamada, N. Akasaka, K. Ohmi, PAC 97,
Z 2 = constant. The amplitude A ≡ |q + i p| p.1831
of the beam centroid is [10] Y. Kobayashi, K. Ohmi, EPAC 98
[11] E.-S. Kim, M. Yoon, Jpn. J. Appl. Phys. 40, 4237
ZF (n) Z 2θ2
A(n) = exp − (6) (2001)
1 + θ2 2(1 + θ 2 ) [12] K.-Y. Ng, Physics of Intensity-Dependent Beam
Long after the kick, θ 1, the cen- Instabilities, World Scientific, 2006, Ch.13
troid amplitude decoheres as A ∼ θ −2 , while
the rms beam size approaches an equilibrium
2.3.11 Momentum Compaction and Phase
σx (∞)/σx (0) = (1 + Z 2 /2)1/2 .
Slip Factor
As time elapses, F (n) periodically comes K.Y. Ng, FNAL
back to its peak value of unity. Therefore, if
μ = 0 (i.e. θ = 0), the beam centroid “recoheres” The phase slip factor η is the relative slip in rev-
with the synchrotron period. This effect provides olution period T for a particle with fractional
a way [3] to measure the product ξσδ (assuming off-momentum δ = Δp/p0 , i.e. ΔT /T0 = ηδ,
νs 1). If μ = 0, the recoherence is still par- where the subscript zero stands for on-momentum
tially present. (Sec.2.1.2). The various orders of momentum-
The formulae above apply to 1-D. Extension compaction factor αi give the relative increase
to 2-D, including x-y coupling in the tune de- in closed-orbit length Cfor an off-momentum
pendence with amplitudes, is addressed in [4]. particle, or ΔC/C0 = ∞
i=0 αi δ
i+1 . With η =
∞ i
Ref.[5] treats the decoherence phenomenon in- i=0 ηi δ , we have [1]
cluding synchrobetatron coupling, damping and
ηi−1 3β02 ηi−2 (1−5β02 )β02 ηi−3
quantum excitation. Ref.[6] applies the canon- ηi = αi − + +
ical Hamiltonian perturbation formalism to 2-D γ02 2γ02 2γ02
decoherence in the presence of an arbitrary non- 5(3−7β02 )β04 ηi−4
− +· · ·, ηi = 1, i = −1 (1)
linear tune dependence on amplitudes; this for- 8γ02 0, i < −1
malism allows computing the decoherence rate
of a beam trapped in a resonant island. Ref.[2] where β0 and γ0 are the on-momentum Lorentz
presents data on the dependence of decoherence factors. The transition gamma is defined as γt =

rate on beam intensity in the SLC (in particular, 1/α0 . To lowest order, all off-momentum
the dependence on the sign of ξ through head-tail particles have the same transition gamma when
damping), while Ref.[7] analyzes this effect using α1 /α0 ≈ − 12 , and cross transition at the same time
a two-particle model. Experimental observation when α1 /α0 ≈ − 32 .
116
Ch.2: BEAM DYNAMICS
For a FODO lattice with thin quadruples of

strength B /(Bρ) = ±S/L, where L is the
half cell length with dipole bending angle θ, we
have [2, 3] (see also Sec.2.2.3)
S(D̂0 − Ď0 ) S(D̂1 − Ď1 )

α0 ≈ 1 − , α1 ≈ − ,
Lθ Lθ
S(D̂2 − Ď2 ) S (D̂0 − Ď0 )
3 3 3
α2 ≈ − − (2)
Lθ 6L3 θ
where the dispersions at the F- and D-quadrupoles

have been power expanded, respectively, as D̂ =
∞ ∞
i=0 Ďi δ . When S 12,

i i
i=0 D̂i δ and Ď =
which is usually true because S = 2 sin μ2 and μ With further reduction of |η0 /η1 |, the buck-
is the phase advance per cell, α1 /α0 → + 32 and ets become α-like (lower figure), which shrink to
zero when |η0 /η1 | = 0. The total bucket height
reduces to + 12 after chromaticities are corrected
|3η0 /(2η1 )| is small. It is asymmetric in momen-
by sextupoles.
tum spread and the beam is susceptible to lon-
For an isochronous or quasi-isochronous gitudinal head-tail instability. If the η1 term is
ring, we must require the spread in η for off- eliminated, the Hamiltonian will be dominated by
momentum particles to be small also. Therefore, η0 and η2 and the bucket becomes pendulum-like
α1 and α2 need to be controlled in addition to again [3]. If the Hamiltonian is dominated by
α0 . In fact, first-order effect of sextupoles alters the η2 term alone, the kinetic term is similar to
α1 , that of octupoles alters α2 , etc. For exam- a quartic potential providing maximal amount of
ple, let Sn = B (n) /(Bρ) be the strength of a thin synchrotron-frequency spread and therefore Lan-
quadrupole (n = 1), thin sextupole (n = 2), or thin dau damping.
octupole (n = 3) of length at a location where The first three orders of the slip factor can be
the horizontal and vertical dispersions are, respec- extracted by measuring the synchrotron tune νs
tively Dx and Dy . Their first-order effects are [4] while changing the rf frequency frf [2]:
Δα0 = −S1 (Dx2 − Dy2 )/C0 , Δα1 = −S2 (Dx3 −
heVrf |η0 cos φs | s1 Δfrf s2 Δfrf 2
3Dx Dy2 )/C0 , Δα2 = −S3 (Dx4−6Dx2 Dy2+Dy4 )/C0 . νs2 ≈ 1+ +
2πβ02 E0 η0 frf η02 frf
The Hamiltonian describing the longitudinal
rf phase difference Δφrf is (Sec.2.3.1) [5] 2η1 −η02 1
s1 = − + 2
η0 γ0
H = h 12 η0 δ2 + 31 η1 δ3 + 14 η2 δ4 +· · · + 3η2 η0 −2η12 η1 3γ02 β02 +2
eVrf s2 = − + (5)
+ [cos(φs +Δφrf )+Δφrf sin φs ] (3) η02 η0 γ02 2γ04
2πβ02 E0
Notice that Δfrf /frf is typically O(η0 ).
For the application of THz near-field imag-
where Vrf is the rf voltage with synchronous phase ing, THz spectroscopy, and others, bunch length
φs and harmonic h, while E0 is the on-momentum compressed to the order of στ ∼ 1 ps is desired.
energy. If only the η0 and η1 terms are considered, An obvious advantage is to store the bunch in the
the two series of distorted pendulum-like buckets α-like buckets, where the bucket half width
in the top figure (Δφrf vs δ with φs = 0 or π, see
[2] for nonzero or non-π φs ) begin to merge to the Δφrf ≈ |η0 /η1 | 2πβ02 E0 h|η0 |/(3eVrf | cos φs |)
middle figure when |η0 /η1 | is lowered to
is intrinsically narrow. Low-α operation modes
( ( (
have been implemented in many light sources. Es-
( η0 ( ( π (( sentially, α0 is reduced by making the dispersion
( (= ( 6eVrf −φ sin φ − cos φ (
( η1 ( ( πβ 2 hη E 2
s s s ( outside the achromats of the Chasman-Green lat-
0 0 0
tice negative by scaling the quadrupole strengths.
(4)
At BESSY II, a reduction from α0 = 7.3 × 10−4
117
100-fold or even more is possible [6]. At SPEAR and h(Ix , Iy ) is the resonance driving term
III, α0 has been reduced from 1.18 × 10−3 240- (RDT). If the motion is governed by multiple res-
fold. However, smaller α0 implies shorter bucket onances, h(Ix , Iy ) has to be replaced by a series
height and therefore shorter beam lifetime [7]. of terms. The particle motion is completely deter-
To increase bucket height, sextupoles are used mined by g and h, which can be calculated from
to minimize |α1 |. For a more reliable operation higher order multipoles (Sec.2.3.3), or obtained
of the machine, the low-α mode of BESSY II is from simulations. Deviations from pure Hamil-
compromised to α0 = 3.5×10−5 with zero-current tonian motion occur due to synchrotron radiation
rms bunch length reduced 5-fold to στ = 3.5 ps. damping (Sec.3.1) in lepton or very high energy
For such an operation, no injection tuning of the hadron rings, parameter variations, and diffusion
optics is required and beam accumulates at a good processes such as residual gas and intrabeam scat-
rate up to a 200-bunch current of 5 mA with a 40-h tering (Sec.2.4.12). The time scale of the non-
lifetime. At SPEAR III, the α/21-operation mode Hamiltonian process determines the applicability
incorporates a 21-fold α0 -reduction at 100 mA of the Hamiltonian analysis.

in 280 bunches with a 30-h lifetime, and a mea- Transverse nonlinearities are introduced
sured bunch length στ = 6.9 ps. The shortest through sextupoles or higher order multi-
bunch length achieved has been 2.5 ps at the sin- poles and magnetic field errors in dipoles and
gle bunch current 3.5 μA, when α0 is reduced quadrupoles. Sextupoles can already drive all
240-fold. The beam lifetime is mostly limited resonances. The beam-beam interaction and
by Touschek effect because of the short bucket space charge also introduce nonlinear fields.
height. When the bunch length is narrow enough, Intentionally introduced nonlinearities are
beam instability often occurs due to coherent syn- used to extract beam on a resonance or through
chrotron radiation. capture in stable islands [1] (Sec.4.12.1). Lo-
calization and minimization of nonlinearities in
References a ring is a general strategy to decrease emittance
[1] J. Wei, PhD thesis, Stony Brook (1990, rev. 1994) growth and increase the beam lifetime. The min-
[2] K.Y. Ng, Fermilab Reports FN-0578, FN-0852 imization of nonlinear effects can be done locally
[3] K.Y. Ng, NIM A404 (1998) 199 or globally. Except for resonant extraction, ampli-
[4] D. Robin et al, PR E48, 2149 (1993) tude increase and particle loss is often the result of
[5] K.R. Symon, A.M. Sessler, Proc. Int. Conf. High chaotic particle motion. Large chaotic regions al-
Energy Acc. (1956) p.44 low particles to increase their amplitudes, leading
[6] J. Feikes et al, EPAC 04, p.2291 to their ultimate loss. However, chaotic particles
[7] X. Huang et al, PAC 07, p.1308 can, on average, still survive the time period of
interest, i.e. the storage time.
2.3.12 Nonlinear Dynamics Experiments Nonlinear dynamics experiments aim to de-
W. Fischer, BNL termine either the detuning and driving terms g
and h directly, or their effect on other quanti-
The goal of nonlinear dynamics experiments is ties. Nonlinear phenomena observed in experi-
to improve the understanding of single particle ments include phase space deformations and res-
effects that increase the particle amplitude and onant islands in Poincaré surfaces of section, non-
lead to loss. Particle motion in storage rings linear phase advances, amplitude detuning g, de-
is nearly conservative and for transverse dynam- coherence (Sec.2.3.10), resonance driving terms
ics the Hamiltonian in action angle variables h, smear, halo formation, echoes (Sec.2.3.13),
(Ix , Iy , φx , φy ) near an isolated resonance kνx + the tune response matrix [2], dynamic aperture
lνy ≈ p is (Sec.2.3.9), emittance growth, and particle loss.
H = Ix νx0 + Iy νy0 + g(Ix , Iy ) + Nonlinear experiments can also be done in the
+ h(Ix , Iy ) cos(kφx + lφy − pθ) (1) longitudinal plane [3].
where k, l, p are integers, θ = 2πs/L is the az- Surface of section The properties of a non-
imuth, and s and L are the path length and cir- linear Hamiltonian system can be visualized by
cumference respectively. The amplitude depen- a Poincaré surface of section, where the phase
dent tunes are given by space variables of the particle trajectory are plot-
νx,y (Ix , Iy ) = νx0,y0 + ∂g(Ix , Iy )/∂Ix,y (2) ted turn-by-turn (TBT). This is the experimental
118
Ch.2: BEAM DYNAMICS
Figure 2: Typical amplitude dependence of betatron

tune measured in VEPP-4M (courtesy V. Sajaev).
Figure 1: Experimental Poincaré surface of section

near a third integer resonance obtained from BPM data
after kicking the beam to different betatron amplitudes
in the IUCF (reprinted with permission from [5], copy-
right by APS).
determination of the 1-turn map. It reveals dis-

tortions of trajectories and resonant islands. Pho-
tographs of synchrotron light from as early as
1968 show beam trapped in transverse resonance
islands [4]. The surface of section can be re-
constructed from two TBT readings of a pair of
BPMs, where the ideal phase advance between the
2 BPMs is an odd multiple of π/2. For Fig.1 a
45 MeV proton beam was cooled in the IUCF and Figure 3: Measurement of sextupolar local term
kicked to different horizontal amplitudes [5]. |χ3000 | in RHIC with an ac dipole. This term is propor-
Detuning and RDTs A Hamiltonian system is tional to a driving term h for the resonance 3νx = p.
completely characterized by the amplitude depen- The bottom plot shows the sextupolar components in
dent tune shift, and the resonance driving terms. the ring (reprinted with permission from [7], copyright
To measure the amplitude detuning, the beam is by APS).
kicked to different amplitudes and the tune ob-
tained from a spectral analysis of the TBT data [5, With the reliable measurement of s-
6, 7, 8]. The tune error Δν in a Fast Fourier Trans- dependent changes in resonance driving terms,
formation (FFT) is proportional to 1/N , where N multipole fields can be inferred and the correc-
is the number of turns used, and can be improved tion of single or multiple resonances becomes
to be proportional to 1/N 4 [9, 10]. Figure 2 shows possible [12].
a detuning measurement in VEPP-4M, where a Tune and amplitude diffusion Mapping the
1.8 GeV electron beam was kicked and observed frequencies (tunes) as a function of initial con-
in BPMs for up to 4096 turns [6]. ditions (Ix , Iy ) is often referred to as Frequency
Resonance driving terms can also be derived Map Analysis (FMA) [9]. Precise tune measure-
from TBT BPM data (Fig.3). With an ac dipole a ments [9, 10] allow the experimental determi-
coherent dipole motion of indefinite length can be nation of frequency maps that reveal potentially
induced [11] and thereby the signal-to-noise ra- harmful resonances (Fig.4). Also accessible is
tio increased compared to a kick-based measure- the tune change over time Δν = ν(T1 ) − ν(T2 ),
ment [7]. The length of the TBT measurement is where T1 and T2 are consecutive intervals, which
then limited by the BPM system. is a measure of the tune diffusion.
119
Figure 6: Measured and simulated DA in SPS with 8

strong sextupoles and tune modulation with parame-

ters q = 1.87 × 10−3 and νM = 2.1 × 10−4 (reprinted
Figure 4: Experimental frequency map for the ALS. with permission from [8], copyright APS).
At a fixed set tune the beam is kicked to different
amplitudes and the beam oscillation frequency is ob- is displayed as survival plots (Sec.2.3.9). In an ex-
tained from turn-by-turn BPM data. The dotted lines periment a single large kick places a large number
are resonances of order ≤ 5 (reprinted with permission of particles across the DA. The DA is then deter-
from [14], copyright by APS). mined as the maximum amplitude where particles
can be observed with a transverse profile monitor,
e.g. with a wire scanner (Fig.6). Increasing the
emittance with many small kicks is also possible
but requires a higher sensitivity in the transverse
profile measurement. In lepton machines, where
the survival times of particles only need to be of
the same order as the synchrotron radiation damp-
ing time, the beam is usually kicked until beam
loss is observed.
Tune modulation is caused by the synchrotron
motion and non-zero chromaticity as well as
power supply ripples in the quadrupoles. Tune
Figure 5: Particle loss rate at a HERAp collimator after
retraction by 100 μm, fitted time-dependent loss curve
modulation affects the long-term stability of par-
and diffusion constant R (courtesy M. Seidel [17]). ticle motion. In the presences of an isolated reso-
nance with island tune νI , the modulated tune
The time evolution of a particle distribution ν(N ) = ν0 + q sin(2πνM N ), (4)
f (I, t) with amplitude diffusion is given by [13] where N is the turn number, and q and νM are the
∂ ∂ ∂ modulation depth and tune, leads to four different
f (I, t) = D(I) f (I, t) (3)
∂t ∂I ∂I phases in the (νM /νI , q/νI ) diagram (Fig.7).
where D(I) = ΔI /(2Δt) is the amplitude de-
2
100.00
pendent diffusion coefficient. Over small ampli-
Tune Modulation Depth q / νI
Strong
tude ranges, such as those created when a scraper 10.00 Chaos
Sidebands
is moved from position Ic , D(I) can be assumed

constant and the change in the loss rate at the 1.00
scraper can be fitted to obtain R = D(Ic )/Ic2 Amplitude Phase
(Fig.5). Larger regions of I are sampled when 0.10 Modulation Modulation
the time evolution of the transverse profiles are

recorded. To access large amplitudes the beam is 0.01
0.01 0.10 1.00 10.00 100.00
kicked, often creating a hollow beam [8, 15]. Dif- Tune Modulation Frequency νM / νI
fusion rates, caused by a number of mechanisms,
can span many orders of magnitudes [16]. Figure 7: Approximate boundaries between dynamical
Dynamic aperture and tune modulation The phases in tune modulation space (reprinted with per-
dynamic aperture (DA) determined in simulations mission from [19], copyright by APS).
120
Ch.2: BEAM DYNAMICS
Massive chaos occurs when sidebands created by

the modulation overlap (Chirikov criterion [18]).
Tune modulation effects have been studied exten-
sively [8, 19, 20].
References
[1] R. Cappi, M. Giovannozzi, PRST-AB 7, 024001
(2004); M. Giovannozzi et al, PRST-AB 12,
024003 (2009)
[2] G. Franchetti, A. Parfenova, I. Hofmann, PRST-
AB 11, 094001 (2008)
[3] M. Ellison et al, PRL 70, 591 (1993)

[4] G.N. Kulipanov et al, Novosibirsk Preprint INP
68-251 (1968) Figure 1: Transverse echo phase space (q, p) of the

[5] D.D. Caussyn et al, PRA 46, 7942 (1992) beam. Note the intricate folding pattern which devel-
[6] V. Kiselev et al, PA 57, 65 (1997) ops due to dependence of the rotational speed on the
[7] R. Tomás et al, PRST-AB 8, 024001 (2005) betatron amplitude. At a critical moment the structures
[8] W. Fischer, M. Giovannozzi, F. Schmidt, PRE in the phase line up to give a significant echo signal.
55, 3507 (1997)
[9] H.S. Dumas, J. Laskar, PRL 70, 2975 (1993) Assume that at t = 0 the beam is deflected√in
[10] R. Bartolini et al, PA 52, 147 (1996) the transverse direction by√ a. Define q ≡ x/ β
[11] M. Bai et al, PR E56, 6002 (1997) and p ≡ (αx x + βx x )/ β. Assume no x-y
[12] R. Bartolini et al, PRST-AB 11, 104002 (2008) coupling and neglect synchrotron damping and
[13] K.-H. Mess, M. Seidel, NIM A 351, 279 (1994) quantum excitation. The initial distribution func-
[14] D. Robin et al, PRL 85, 558 (2000) tion of the beam
is assumed Gaussian,
ψ(q, p) =
[15] T. Chen et al, PRL 68, 33 (1992) (2π)−1 exp −(q 2 + p2 )/2 , with the beam
[16] F. Zimmermann, PA 49, 67 (1995) emittance. At time t = τ a quadrupole kick
[17] M. Seidel, PhD thesis, Hamburg U., DESY 94- of strength Q = (β-function at the quad)/(focal
103 (1994) length of the quad) is applied to the beam. We
[18] B.V. Chirikov, Physics Report 52, 263 (1979) assume Q 1.
[19] T. Satogata et al, PRL 68, 1838 (1992) Let the betatron tune be
[20] O.S. Brüning, F. Willeke, PA 55, 237 (1996) q 2 + p2
ν = ν0 − μ (1)

2.3.13 Echo Effects where μ is the dimensionless detuning parameter.
G.V. Stupakov, SLAC The second order perturbationtheory [1, 2] pre-
dicts the echo amplitude η = q 2 + p2 ,
Transverse echo An echo effect in accelerators
τ0 t − 2τ
was first proposed in [1, 2] for the betatron mo- η = aF , (2)
tion of a beam. It could be observed in a situation τd τd
where the beam in the ring is deflected off-orbit where τ0 = Qτ , τd = T0 /4πμ is of the order of
at time t = 0, causing its centroid to undergo be- the decoherence time with T0 the revolution pe-
tatron oscillations. After these oscillations have riod, and
x
completely damped out due to beam decoher- F (x, y) = (3)
ence (Sec.2.3.10), the beam is excited again by ((1 + x2 − y 2 )2 + 4y 2 )3/4
a quadrupole kick at t = τ . This kick does As follows from Eqs.(2) and (3), the echo occurs
not produce any visible displacement of the beam at about t ≈ 2τ , and the duration of the echo pulse
at that time, but it turns out that close to time is of the order of the decoherence time τd . It de-
t = τecho = 2τ the beam centroid undergoes tran- cays as ∝ |t − 2τ |−3/2 for large |t − 2τ |. A char-
sient betatron oscillations with an amplitude that acteristic phase space of the beam at the moment
is a fraction of the initial beam offset. The echo of the maximal echo signal is shown in Fig.1.
arises as a result of nonlinear beam dynamics in In addition to the echo effect at t = 2τ higher
the phase space due to the tune dependence on the order echos (with smaller amplitudes) can be ob-
amplitude of the betatron oscillations. served at even multiples of τ [2].
121
0.6 0.025
0.5
0.015 t t peak
0.4
η max (a.u.)
Amplitude [volts]
0.3
0.005
0.2
0.1 -0.005
0
0 0.5 1 1.5 2 2.5 3 -0.015
τ(8D0ω 2revμ2 /3ε)1/3
t echo
-0.025
Figure 2: Echo signal (in arbitrary units) as a function
0.0 0.2 0.4 0.6 0.8 1.0

Time [sec]
of delay τ .
Figure 3: Beam response to impulse at n = 9, fol-

Collective effects can affect the decoherence lowed by m = 10.
mechanism, and therefore the echo signal, signif-
icantly through an impedance in the accelerator
developed in [6, 7, 8]. First, a short-duration rf
[3].
pulse is applied to the beam at a revolution har-
Effect of diffusion The echo effect is sensitive
monic nω0 . After a delay τ , a second pulse is
to diffusion processes in the phase space. This
applied at frequency mω0 . The lowest order echo
is because in echo generation the phase space
can be observed at the frequency (n − m)ω0 at a
of the beam develops structures with characteris-
time τecho = mτ /|n − m|. The longitudinal echo
tic size inversely proportional to the delay time
can also be viewed as a nonlinear mixing of two
τecho , see Fig.1. These fine structures can be eas-
waves propagating in the beam.
ily smeared out even by a weak diffusion, whose
strength is amplified with diminishing dimensions Echo experiments First echo experiments have
of the structures. Hence, the echo can be a useful been carried out in the Fermilab Antiproton Accu-
diagnostic tool for measuring week diffusion pro- mulator [6] and at CERN SPS [9]. The echo sig-
cesses in the phase space of the beam. nal measured in the Fermilab experiment is shown
The effect of diffusion on echo for a bunched in Fig.3. A theoretical fit to the data gave an ef-
beam was studied in [4] when τ0 τd . In this fective diffusion rate in the beam corresponding
limit the maximum value η max is achieved at t = to the diffusion time of 300 s. An extremely small
τecho , diffusion rate corresponding to the doubling of the
aQ τ energy spread after 107 s has been measured by
η max = (4) means of echo in the CERN experiments [10].
τd 1 + 8D0 μ2 ω02 τ 3 /3
Transverse echo was measured in RHIC by
where ω0 = 2π/T0 and D0 is the diffusion co-
applying a dipole kick, followed by a one-turn
efficient. See Fig.2. For small τ , the maximum
quadrupole kick [11].
echo signal linearly increases with τ . For large τ ,
the deformation of the distribution function due to Using echo for seeding FEL The echo scheme
the diffusion becomes strong enough to compete can also be used for generation of high harmonics
with the formation of the echo, and causes η max in seeding of free electron lasers [12]. This echo-
to decay ∝ τ −2 in the limit τ → ∞. enabled harmonic generation (EEHG) mechanism
Longitudinal echo The echo effect can also be offers a remarkable up-frequency conversion effi-
observed in the longitudinal direction for both ciency, with the amplitude of the generated har-
bunched and coasting beams. The case of a monics only slowly decreasing with the harmonic
bunched beam is analogous to the transverse echo number [13].
[5], with an analog of the transverse dipole kick Spin echo in accelerators An echo occurs in
being an instantaneous change in the rf phase, and the spin polarization of a beam with finite energy
an analog of the quadrupole kick realized as a spread among its particles when two spin depo-
sharp increase in the rf amplitude. larization resonances are crossed (or a single res-
Most of the echo studies in accelerators were onance crossed twice). If the two crossings are
actually carried out with coasting beams. The the- spaced by a time interval τ , the spin echo occurs
ory of longitudinal echo for such beams has been at a time t = 2τ [14].
122
Ch.2: BEAM DYNAMICS
References which is proportional to the bunch charge and in-

verse of the bunch length. Here, φ0 > 0 means
[1] G. Stupakov, SSCL Report 579 (1992)
ahead of rf crest and
[2] G. Stupakov, S. Kauffmann, SSCL Report 587 ∞
d
(1992) V (z) ≡ ρ(z )W (z − z)dz (2)
[3] G.V. Stupakov, A.W. Chao, PAC 1995 dz z
[4] G. Stupakov, A.W. Chao, PAC 1997
[5] N. Mahale et al, SSCL-N-817, 1993
[6] L.K. Spenzouris, J.-F. Ostigy, P.L. Colestock,
2.4.1.2 Multibunch longitudinal dynamics
PRL 76, 620 (1996) For multi-bunch longitudinal dynamics, only fun-
[7] O. Brüning, CERN SL/95-83 (AP), 1995 damental accelerating mode is important. Its ef-
[8] E. Shaposhnikova, CERN SL/Note 95-125 (RF), fect is called beam loading.
1995 Symbols: q = bunch charge, ωrf = rf fre-

[9] O. Brüning et al, CERN SL-MD Note 217, 1996 quency, ω = resonance frequency of structure,
[10] O. Brüning et al, EPAC 96, 1996

κ1 = fundamental mode loss factor per length,
[11] W. Fischer, T. Satogata, R. Tomas, PAC 2005 v
Tf = structure filling time, fv ≡ c−vg g where vg
[12] G. Stupakov, PRL 102 (2009) 074801
[13] D. Xiang, G. Stupakov, PRST-AB 12 (2009) is group velocity in the structure, Ls = structure
030702 length, tb = bunch spacing, nB = total number
[14] A.W. Chao, E. Courant, PRST-AB 10 (2007) of bunches in the bunch train, Ib = q/tb = aver-
014001 age beam current in the bunch train. We assume
tb ωrf /2π = integer and |ωrf − ω| ω.
Standing wave structures In a standing wave
2.4 COLLECTIVE EFFECTS structure, beam induced voltage by a bunch train
at time t after the first bunch enters the structure
2.4.1 Collective Effects in High Energy is [3]

Electron Linacs 2κ1 Tf Ib Ls −(1−i tan ψ) Tt
Vb (t) = 1−e f (3)
K. Kubo, K. Yokoya, KEK 1 − i tan ψ
K. Thompson, Archimedes Inc.
where tan ψ = (ω − ωrf )Tf (ψ is the detuning
Symbols: Average longitudinal and transverse angle).
wake functions per unit length along the beam Traveling wave structures In a constant gradi-
line are W (z) [unit V/C/m] and W⊥ (z) [unit ent structure, induced gradient at time t after the
V/C/m2 ] (Sec.3.2.1); z is the longitudinal sepa- first bunch enters the structure is [3]

ration between the exciting and the witness point rs Ib τ Ls 1 − e−2τ x −2τ
charges. We only consider the monopole mode Vb (t) = − xe (4)
1 − e−2τ 2τ
wake for longitudinal motion and the dipole mode
wake for transverse motion. We ignore depen- where rs is shunt impedance per unit length
dence of these functions on position s in the beam (Ω/m), τ the attenuation parameter, I = eN/tb
line. the average current, x = min(t/Tf , 1). Vb is con-
stant for t > Tf .
2.4.1.1 Single bunch longitudinal dynamics Multi-bunch energy compensation methods
If the bunch train length <∼ Tf , transient beam
Energy variation and minimization z-
loading must be compensated to obtain uniform
dependence of particle energy can be controlled
acceleration.
by choosing the rf phase at bunch center, φrf ,
For standing wave structures, energy gain can
along the linac. Energy spread can be also con-
be made uniform by adjusting time of rf pulse,
trolled by shaping the longitudinal distribution
starting rf before bunch train enters the structure
ρ(z) before injection [1].
by time trf ,
For a short bunch, assuming sinusoidal rf
voltage, off-crest phase, φ0 , for minimum energy Vc
trf = Tf ln 1 + (5)
dependence is approximately [2] 2κ1 Ls Tf Ib
∞
c where Vc is the uniform accelerating voltage for
sin φ0 = − ρ(z)V (z)dz (1) all bunches.
ωrf V̂ −∞
123
Sec.2.4: COLLECTIVE EFFECTS
For SC cavities (Q0 QL ), almost all rf 2.4.1.3 Single bunch transverse dynamics
power can be given to beam (no reflection power)
keeping the uniform acceleration condition, if The equation of motion is [6]

trf = Tf ln 2. In such case, assuming ψ = 0, d d
E(s) x(z, s) + E(s)k2 (s)x(z, s)
1 ds ds
Vc = −Vb (t → ∞) = Vg (t → ∞) (6) ∞
2 =e 2
W⊥ (z − z)x(z , s)dz (8)
z
where Vb (t → ∞) is voltage induced by continu-
ous beam with the same current, and Vg (t → ∞) The energy spread and discreteness of focusing
is voltage generated by continuous rf with the elements are omitted.
same power. For uniform ρ = N/l (l is length of bunch),
For traveling wave structures, several meth- W⊥ (z) = W⊥ z, uniform acceleration E(s) =
ods are possible. E0 + Gs, k(s) = k0 , initial condition x(z, 0) =
x0 , k0 s 1,
1. Matched filling: By adjusting timing and
bunch spacing of train, one may cancel most ∞
E0 ik0 s 1 η n
bunch-to-bunch energy variation in a short train x(z, s) ≈ Re x0 e
E( s) n!(2n)! 2i
[10]. For constant impedance structure, assuming n=0
nB tb Tf , the first and last bunch have equal (9)
energy gains if where

eN Wperp z2s E(s)
tb 2κ1 q/(1 + fv ) η= ln (10)
=√ (7) k0 (E(s) − E0 )l E0
Tf 2rs Prf lA e−(1+fv )s0 /lA + nB κ1 q
For large η 1, the sum has a asymptotic form,
Here s0 is the distance rf pulse has propagated in
the structure when the first bunch is injected, and E0 η −1/6 3√3 η1/3
lA = L/τ is the attenuation length of rf. This x(z, s) ≈ x0 e 4 (11)
E(s) 6π
method works well only for short bunch train.
2. Pre-filling (Sometimes called ΔT This beam breakup instability is strong for
method.): Supplying shaped rf power before linacs with large single-bunch charge and small
beam injection can be done in such a way that the aperture accelerating structures, such as normal
energy gain of each bunch during the transient pe- conducting linear colliders. This problem is usu-
riod approximates the energy gain in steady state ally cured by BNS damping.
[11]. In the simplest form, the amplitude of the Two-macroparticle model With two
input rf is linearly ramped during the first filling macroparticles (spacing l, charge N e/2) in-
time, then bunch train is injected during second jected with initial displacement x̂, solutions
filling time with constant rf amplitude. Shaping are given in Tab.1 [3, 4]. For the case with
of rf wave form can be done, e.g., by controlling no acceleration (E = constant) and constant
phases at which two or more klystron outputs are focusing (k1 , k2 = constants), if k1 = k2 ,
combined.
2 −x1
xn.a. n.a. N e2 W⊥ (l)
3. Staggering timing: Timing of rf pulses into = s sin ks (12)
a certain number of accelerating sections are dis- x̂ 4kE
tributed over filling time (or bunch train length, If k1 = k and k2 = k + Δk with Δk k,
whichever shorter) [12, 13, 14].
2 −x1
xn.a. n.a. N e2 W⊥ (l)
4. Δf method: This is achieved by mixing ≈− 1−
x̂ 4kEΔk
structures of slightly different frequencies. Re-
quired total amplitude of such compensation rf ×2 sin(Δks/2) sin[(k + Δk/2)s] (13)
field is proportional to inverse of frequency dif- BNS damping and autophasing [7] In two-
ference, while the frequency difference, Δω < ∼ macroparticle model, to make head and tail
1/nB tb . macroparticles follow the same orbit, Eq.(13) re-
The last two methods have problem that com- quires
pensation is nonlocal and energy spread may be N e2 W⊥ (l)
too large in the low energy part of linac. Δk = (14)
4kE
124
Ch.2: BEAM DYNAMICS
Table 1: Two-macroparticle model solutions.
xn.a.
1 = x̂ cos ks
No acceleration
N e2 W⊥ ()
xn.a.
2 = x̂ cos ks + s sin ks
4kE
Case p = 0:
1/2
xp=0
1 (s) = x̂ E0
E(s) cos k0 s
Uniform acceleration 1/2
and smooth focusing: N e2 W⊥ ()
xp=0 (s) = x̂ E0
cos k s+ ln(1 + αs) sin k s
2 E(s) 0 0
4k0 E0 α
0 (1 +
E(s) = E αs) Case p = 1/2:
p
E0 p=1/2 E0 1/4 E0 1/4
k(s) = k0 x1 (s) = xn.a.
E(s) E 1 (seff ) = E x̂ cos k0 seff
p=1/2 E0 1/4 n.a.
(adiabatic, i.e. α k0 ) x2 (s) = E x2 (seff )

E0 1/4 N e2 W⊥ ()
= E x̂ cos k0 seff + seff sin k0 seff
4k0 E

where effective length: seff ≡ k10 0 k(s )ds
s
or, with E1 = E, E2 = E + ΔE, ΔE E, 2.4.1.4 Multibunch transverse dynamics

Δk ΔE
k = ξ E (ξ is related to chromaticity and usu- The cumulative BBU, first studied at SLAC [15,
ally ∼ −1),
16, 17, 18], has been treated for various cir-
N e2 W⊥ (l) cumstances; long/short beam pulse, strong/weak
ΔE = (15)
4ξk2 focusing system, rf/induction/superconducting
linacs, etc. [19, 20] See Sec.4.4 for the case of
These are autophasing conditions; also sometimes
energy recovery linacs (ERLs).
called BNS damping conditions, but BNS damp-
Equation of motion One can treat each bunch
ing may also refer to cases where betatron oscil-
as a point charge (since electron bunches are
lation is actually damped in parts of the bunch.
short). (Finite length bunch is discussed in
To achieve autophasing for an arbitrary bunch
[21].) For nB equally-charged (bunch charge
distribution ρ(z) [8] (no acceleration),
N e), equally-spaced (spacing l) bunches, each
∞
Δk(z) e2 bunch represented
as a single
macroparticle,
= 2 ρ(z )W⊥ (z − z )dz (16) d d
k 2k E z E(s) xn (s) + E(s)kn2 (s)xn (s)
ds ds
These autophasing conditions also apply to uni-
formly accelerated case with p = 1/2.
n−1
Choice of rf phase profile involves consider- = N e2 W⊥ ((n − j)l)xj (s) (17)

ation of minimizing energy spread, optimization j=1
of BNS damping, and amount of sacrificed accel- A difference from single-bunch BBU is that W⊥
eration by putting bunches off rf crests. One so- is dominated by one or a few resonator modes
having large shunt impedance,
R
lution, as considered in normal conducting linear ω z
colliders, is to run behind rf crests early in linac, − k ωk z
W⊥ (z) = ωk e 2cQk sin (18)
introducing BNS energy spread, then run ahead of Q ⊥,k c
k
crests later in linac so as to reduce energy spread In case of a single isolated resonator with in-
at the end. finite Q, the amplitude blowup factor of the last
Quadrupole beam break-up can occur if beam bunch takes a form similar √ to Eq.(10) with domi-
is sufficiently intense and transverse beam size is nant exponential factor e η (η 1),
L
comparable to beam pipe radius [9]. In contrast to ds
dipole BBU, this can occur even if beam is well η = e2 nB N W⊥0 (19)
0 E(s)k(s)
steered to axis of accelerator.
125
where W⊥0 is the coefficient of sine function for 2. Effects to the next bunch are small if Qs of
the resonator mode in Eq.(18). The condition η < ∼ dominant deflecting modes are small enough
1 crudely determines the required W⊥0 . (Called “damped structure”). When wake
For constant acceleration and focusing that can reach only the next bunch (daisy chain
scales as E −p , solutions in WKB approximation, model), the condition that the blowup be
for an arbitrary wake, are small is
% s
N e2 N e2 W⊥ (l) L ds
xn (s) = xn (0) + ds <1 (24)
2iE0 kn (0) 0 2 0 E(s)k(s)
1−p
E0 2 One of the most strongly damped structures
×
exp[−iψn (s , 0)] is choke mode structure [25], in which Q can
E(s )
⎫
be as small as <
∼ 10 for all deflecting modes.

n−1 ⎬
× W⊥ ((n − j)l)xj (s )
3. If deflecting modes have a frequency spread,

⎭
j=1 the sum of the wake along a long struc-
1−p ture (or more than one structures) can be re-
E0 2
× exp[iψn (s, 0)] (20) duced [26, 27]. Spread can come from ran-
E(s) dom construction errors [28]. Or, structure
where s
can be designed so that cell dimensions dif-
ψn (s, s ) ≡ kn (s )ds (21) fer slightly from cell to cell so as to give
s spread of mode frequency, causing deco-
These equations for xn (s) may readily be solved herence of wake function (“detuned struc-
numerically. ture”). For sufficient reduction in one bunch
Daisy chain model If the wake field is negligi- spacing (tb ), required frequency spread is
ble beyond more than one bunch spacing, σf > ∼ 1/(2tb ). For avoiding re-coherence
x1 + k2 x1 = 0 at the end part of bunch train, typical dif-
ference of two frequencies next each other
N e2 W⊥ (l)
xn + k2 xn = xn−1 (n > 1) (22) is required as δf < ∼ 1/(nB tb ), which re-
E quires many cells with different frequencies
ignoring acceleration and variation of focusing for long bunch train. Combining method
along the linac [23]. For initial conditions 2 with this method (damped-detuned struc-
xn (0) = 1, xn (0) = 0, ture), this re-coherence effects can be miti-
(−iσs)j
n−1 gated [29].
xn (s) ≈ eiks (23)
j! 2.4.1.5 Effects of structure misalignment
j=0
N e2 W⊥ (l)
where σ ≡ 2kE For sufficiently large |σs|,
. So far we discussed BBU induced by wakefield
the amplitude of oscillation of bunch n grows as perfectly aligned wakefield sources (accelerating
sn−1 . The criteria for little or no blow-up in linac structures), where injection error is the seed of
is |σL| < 1 where L is length of linac. the oscillation. Misalignment of structures can
Analytic daisy chain results given above still be treated as additional effects, assuming typical
hold for case of constant acceleration and a focus- misalignment is much larger than amplitude of
ing characterized by p = 1/2, provided that s is beam oscillation. Orbit offset of particle of lon-
interpreted as the effective length (See Tab.1) and gitudinal position s at the end of linac (length L)
xn (s) is interpreted as bunch offset with adiabatic is
L
damping (E0 /E)1/4 factored out. R12
Control of multibunch BBU For large scale x(z, L) = −eW⊥,sum (z) dsa(s) (25)
0 E(s)
linacs, several methods (and their combinations)
where a(s) is offset misalignment of wake source
have been proposed.
at s, R12 the transfer matrix from s to the end of
1. Effects of wake are minimized if bunches are linac, E(s) the beam energy at s, and we define
near zero crossing of wake, i.e. dominant de- “sum wake” as
z
flecting mode frequency and bunch spacing
satisfy sin ωk tb ≈ 0. W⊥,sum (z) ≡ ρ(z )W⊥ (z − z )dz (26)
0
126
Ch.2: BEAM DYNAMICS
For multi-bunch dynamics, it may be [3] P. Wilson, AIP Proc. 87(1981)

[4] A. Chao, Physics of Collective Instabilities in
W⊥,sum (n) = eN W⊥ ((n − j)ctb ) (27) High Energy Accelerators, Wiley (1993)
j<n [5] A. Piwinski, DESY report HERA 92-11(1992)
Beam size increase is expressed as [6] A.W. Chao, B. Richter, C.Y. Yao, NIM 178
∞ (1980)
1 [7] V. Balakin, S. Novokhatsky, V. Smirnov, Proc.
Δσ 2 (L) ≈ ρ(z)x2 (z, L)dz
Ne 0 12th Int. Conf. on High Energy Acc. (1983)
L 2 [8] V. Balakin, Proc. Workshop on Linear Colliders
R12 (s)
= e2 W⊥,div dsa(s) (28) (1988) p.55
0 E(s) [9] A. Chao, R.K. Cooper, PA 13(1083)1
[10] R.D. Ruth, Proc. ICFA/INFN Workshop on
R12 (s) is 1-2 element of transfer matrix from s

to the end, E(s) beam energy at s and we define Physics of Linear Colliders (1988)
[11] K.A. Thompson, R.D. Ruth, PAC93
“divergence of sum wake” as

[12] R.B. Neal, ed., The Stanford Two-Mile Acceler-
∞ 2
W⊥,sum (z) ator, Benjamin (1968) p.84
W⊥,div ≡ dzρ(z) 2 [13] R.H. Helm et al, IEEE Trans. Nucl. Sci. NS-16,
0 (N e)
∞ No.3(1969)311
W⊥,sum (z) 2 [14] B. Aune, J. Leroy, A. Mosnier, PAC83, p.2995
− dzρ(z) (29)
0 Ne [15] R.B. Neal, W.H. Panofsky, Science (1966)1353
[16] R.H. Helm, G.A. Loew, Linear Accelerators,
For random and independent misalignment of
North-Holland (1970)
each structure of length Ls (this may be replaced [17] W.K.H. Panofsky, M. Bander, RSI 39 (1968)206
by length of independently aligned unit), [18] V.K. Neil, L.S. Hall, R.K. Cooper, PA 9 (1979)
3 2 4
L
R12 R12,i 2 213
dsa(s) = L2s a2i [19] Y.Y. Lau, PRL 63 (1989) 1141
0 E(s) Ei [20] G.H. Hoffstätter, I.V. Bazarov, PRST-AB 7
i
(30) (2004) 054401
where i is index of structure, and s dependence [21] C.L. Bohn, J.R. Delayen, PR A45 (1992) 5964
is replaced by i dependence for every term. In- [22] R.L. Gluckstern, R.K. Cooper, P.J. Channell, PA
crease of projected, normalized emittance is eval- 16 (1985) 125
uated from Δσ 2 (L) dividing by beta-function and [23] K.A. Thompson, R.D. Ruth, PR D41 (1989) 964
multiplying energy factor at the end, Δγ = [24] G.F. Caparaso, Linac86, 1986, p.17; G.F. Ca-
γ(L)Δσ 2 (L)/β(L). Expressions above are suit- paraso et al, PRL 57 (1986) 1591
able for numerical calculation. [25] T. Shintake, Jap. J. Appl. Phys. 31, L1567 (1992)
For a long linac, one may transform a2i → [26] K. Yokoya, DESY 86-084 (2986)
[27] R.L. Gluckstern, F. Neri, R.K. Cooper,
i → Ls , and
2 L
a (a is rms misalignment),
2 LINAC86, p.540
R12,i
Ei → β(L)/2(β/E)E(L) (¯ denotes aver- [28] R.L. Gluckstern, F. Neri, R.K. Cooper, PA 23
age along linac), then (1988) 37
[29] K.A. Thompson et al, PA 47 (1994) 65

e2 2 β
Δγ = a L s LW ⊥,div (31)
2mc2 E
2.4.2 Collective Effects in Energy Recovery
For multibunch beam of very long pulse, Linacs
since the wake function eventually dies away as G. Hoffstaetter, Cornell U.
z → ∞ due to finite Q, W⊥,sum (n) ≈ constant
for large n, bunches in tail part of the beam have Energy Recovery Linacs (ERLs) accelerate in a
the same orbit [22]. recirculating linac with Np passes to high energy
where the beam is used for experiments. Sub-
sequently the spent beam is decelerated through
References
Np passes to low energy where it is dumped.
[1] F.J. Decker, R. Holzapple, T. Raubenheimer, ERLs are therefore nonperiodic systems and all
LINAC94 (1994) p.47 collective effects of one-pass systems (Sec.2.4.1)
[2] K. Bane, AIP Proc. 153(1987) can be relevant, e.g. single-bunch beam breakup,
127
accumulative beam breakup, space charge forces, first turn, and 2 indicates the return pass, with the
coherent synchrotron radiation, and ion trapping. energies Ei , the transverse beta functions βi , and
While linac beam currents are usually lim- the betatron phase advance Ψ between 1 and 2.
ited by the available power, this limitation is lifted If there is only one HOM and one recircula-
when the energy of the spent beam is recov- tion turn, the threshold current can be determined
ered, and much larger currents become available. by the following three cases:
Some collective effects common in ring accelera- 2E1 1 1
tors therefore can become relevant, e.g. Touschek Imin = (2)
ec τλ ( R )[2n] |D sin Ψ12 |
scattering, intrabeam scattering, fast ion instabil- Q λ
ity and electron cloud. Imin
tr τλ , D sin(ωλ tr ) < 0 : Ith =
However, there is a class of instabilities that | sin(ωλ tr )|
is pertinent to ERLs and specific to the nature tr τλ , D sin(ωλ tr ) > 0 :

of multi-turn, nonperiodic systems. This is the
τλ
transverse and longitudinal recirculative beam- Ith = Imin 1 + [ mod(ωλ tr , π)]2

breakup instability and the recirculative orbit distr
tortion. tr τ λ :

For currents higher than a threshold Ith , τλ π
the power in HOMs can resonantly increase. Ith = Imin 1 + [ mod(ωλ tr ± , 2π)]2
tr 2
The HOM indexed with subscript λ has angu- Occasionally, additional factors are found in the
lar frequency ωλ , quality factor Qλ , impedance −
tb
[2n]
(R/Q)λ Qλ (in circuit definition, 12 of linac def- literature, e.g. [7], notably e τλ . In [1] it has
been shown that such additional factors are not
inition, and units of mΩ2n ) for a HOM of multi- required.
pole order n, polarization angle θλ and polariza- General transverse recirculative BBU A
tion vector eλ = cos θλex + sin θλey . The damp- general theory for Np passes and N HOMs can
ing time is τλ = 2 Q λ
ωλ . The bunch repetition rate be formulated in the form
is tb . The wake potential (Sec.3.2.1) is λk = kth eigenvalue of [W(ω)U] (3)
R [2n] − t and the threshold current is the smallest real 1/λk
W (t) = c( )λ e τλ sin ωλ t (1)
Q for ω ∈ [0, tπb ]. The matrix W(ω) has Np2 N 2 co-
A general formulation of recirculative BBU efficients WikLI that describes how the ith HOM in
for dipole modes can be found in [1] for uncou- th
the L turn accumulates fields from orbit oscilla-
pled, and in [2] for coupled optics. A general tions excited by the kth HOM in the I th turn, and
theory for quadrupole and 2n-pole modes is out- UkjIJ describes how transverse momenta changed
lined in [3]. For 1 degree of freedom, [4] has a at the HOM for j and J produce orbit offsets at the
theory of BBU instability in recirculating linacs, k th HOM in turn I. U therefore depends on cou-
i.e. energy is added in each pass, and where ev- pled motion and HOM polarization angles. The
ery bucket is filled; [5] adds subharmonic bunch- form of W(ω) and U can be found in [1].
ing. Tracking simulations [6] compared well with The growth rate α can be obtained by
this theory. Coupling and polarized modes were
1 dλk −1
considered for one HOM and one turn in first or- α = − 2 Im{( ) } (4)
der theory [8]. Comparisons between experiments Ith dω
and theory have been successful, including polar- evaluated for the k and the ω in Eq.(3) that lead to
ized modes and coupled optics, e.g. in [9]. Lon- the threshold current.
gitudinal recirculative BBU is analyzed in [10]. Transverse higher order recirculative BBU
One dipole HOM and one recirculation A multipole of order n produces the transverse
The return time is tr and the linear beam optics in kick fx + ify = n(x − iy)m einθλ with m = n − 1
the transverse plain is described by Δx = MΔx , and the HOM excitation [3]

so that the excitation of HOM μ by a kick in HOM fx
ν is characterized by Dμν = eTμ Meν and for one D = (fx , fy )(x2 )M (x1 )1 (5)
fy
mode D = Dλλ. For horizontal HOMs this re-
E1 √
where x2 is expressed in terms of x1 , and the aver-
duces to D = E2 β1 β2 sin Ψ, where the in- age is performed over the phase space at location
dex 1 refers to the particle after the HOM in the 1. For dipole HOMs this amounts to D in Eq.(2).
128
Ch.2: BEAM DYNAMICS
For round beams with decoupled and rota- 2.4.3 Beam Loading
tionally symmetric transport, D. Boussard, CERN
D = M12 n2 Re{[(x2 −iy2 )(x1 +iy1 )]m }1 2.4.3.1 Single-bunch passage in a cavity
E1
D = 4 β1 β2 x sin(2Ψ) for quad modes, The amplitude of beam-induced voltage is
E2
3 q R
= ωc q
Vb0 = (1)
E1 2
C Q
D = 32 8 β1 β2 2x sin Ψ cos2 Ψ
E2
where q = single bunch charge (assumed short
for sext modes, bunch with σz λrf ), R/Q = cavity geometric
n
2 E1 2 parameter, C = (ωc R/Q)−1 = equivalent shunt
D(nΨ) = n m! β1 β2 mx sin(nΨ) (6)
E2 capacitance of cavity, fc = ωc /2π = cavity reso-

nant frequency.
The last line for a general 2n-pole shows only the Energy deposited by the bunch in the cavity is
term with leading betatron frequency nΨ as an or-
der of magnitude estimate. These D are inserted 1 q2 1R 1
Wb = = ωc q 2 = qVb0 (2)
into Eq.(2) to determine the threshold current in 2C 2Q 2
the applicable regimes. The net energy received by the bunch is
Longitudinal recirculative BBU In the sim-
1
plest model, monopole modes drive energy oscil- W = qVg sin φb − Wb = q Vg sin φb − Vb0
lations above the threshold current 2
(3)
2E1 where Vg = generator voltage, and φb = beam
Ith = R
(7)
ηtr ωλ ( Q )λ Qλ phase angle (counted from zero crossing of rf
wave). The net voltage seen by the bunch is
with the slip factor η describing the relative time
delay per relative energy difference for one pass. Vb0
V = Vg sin φb − (4)
Recirculative orbit distortion When the ith 2
HOM is offset by x0i and a current I0 is acceler- Observations: (i) The bunch sees only half
ated, the orbit distortions xi − x0i in equilibrium of its own induced voltage (“Fundamental The-
are given by orem of Beam Loading” [1]), and (ii) Loss factor
11 1R
x − x0 = [I0 W(0)U − 1]−1 x0 (8) of cavity on a particular mode is = ωc .
2C 2Q
As shown in [1], the matrix is invertible for any A phasor diagram at the rf frequency is shown
I0 smaller than the dipole BBU threshold current below with V − and V + = rf voltages before and
Ith . after the bunch passage, V = net voltage seen by

the bunch, and ib = rf component of bunch cur-
References rent. The vectors Vb0 and ib are in phase.
ib
[1] G.H. Hoffstaetter, I. V. Bazarov, PRST-AB 7,
φb +
054401 (2004) V
[2] G.H. Hoffstaetter, I.V. Bazarov, C. Song, PRST- V bo
AB 10, 044401 (2007)
[3] C. Song, G.H. Hoffstaetter, Report Cornell-ERL- V
07-10 (2007)
[4] J.J. Bisognano, R.L. Gluckstern, PAC87, p.1078 1–98
8355A135
–
V
[5] B.C. Yunn, PAC91, p.1785 1787 (1991)
[6] G.A. Krafft, J.J. Bisognano, PAC87, p.1356
[7] B.C. Yunn, PRST-AB 8, 104401 (2005) 2.4.3.2 Cavity equivalent circuit [2]
[8] E. Pozdeyev, PRST-AB 8, 054401 (2005)
In the figure below, V = cavity voltage, Rg =
[9] D.R. Douglas et al, PRST-AB 9, 064403 (2006)
[10] J.J. Bisognano, M.L. Fripp, CEBAF-PR-89-018 shunt resistance of tetrode seen from cavity
(1989) (transformed to the gap), ig = pure current source
(transformed to the gap), and ib = rf component
129
of beam current. Note that maximum power and (transformed to the gap). The cavity coupling co-
efficiency of the tetrode does not correspond to efficient βc is given by βc = R /Rg .
Rg = R . The forward power Pf in the line (i.e. power
flowing from klystron + circulator towards cavity)
it
is given by
Ig Rg V R' L C ib 1 1
Rg i21 = Rg i2g
Pf = (8)
2 8
1–98
Tetrode Cavity Beam At optimum coupling, βc = βopt , defined as
R R ib sin φb
8355A31
The phasor diagram (steady state, above tran- βopt = =1+ (9)
Rg V
sition case) is shown below.

and optimum detuning Δω = ωDO , the combina-
ib tion (cavity + beam) is matched to the generator

ϕb and the required rf power is minimum,
it
1 V2 1V2 1
Pf = Pf min = = + V ib sin φb
ϕz 2 Rg 2 R 2
V (10)
ϕL io where the first term on the r.h.s is cavity loss,
1–98
ig
while the second term is the power transferred to
8355A136
the beam. There is no reflected power dissipated
Notation: i0 = V /R = generator current needed in the circulator load (see figure below).
to produce V when the cavity is at resonance and Equivalent coupler power Peq is the traveling-
there is no beam current, R = R Rg /(R + Rg ), wave power in the line which produces the same
Y = ib /i0 = Vb /V = beam loading parameter, peak fields. It is an important parameter for the
Vb = beam-induced voltage at resonance, φb = design of rf couplers, usually tested in traveling
stable phase angle (short bunches) measured from wave conditions. For a (cavity + beam) undercou-
zero crossing of rf wave, φz = cavity impedance pled i.e. βc < βopt , Peq is independent of ib ,
phase angle (tan φz = Δω/σ, Δω = ωc − ωrf , 1 V2
R Peq = (11)
and σ = ωc /2QL ), QL = R/( Q ) = loaded qual-
2 Rg
ity factor of the cavity, and φL = phase angle be-
tween generator current (or grid voltage) and V. The following figure shows the rf powers vs
beam current for a cavity driven by a matched
Steady state equations From the phasor dia-
generator and optimum detuning Δω = ωD0 or
gram,
φL = 0:
tan φz − Y cos φb
tan φL = (5) RF
Equivalent
1 + Y sin φb Power
Coupler Forward
i0 (1 + Y sin φb ) Power Power
ig = (6)
cos φL 1 V
2 Beam
2 Rg Power
Optimum detuning (real load) is obtained when Cavity
Losses Reflected
φL = 0, or Power
tan φz = Y cos φb , or Δω = ωDO = Y σ cos φb i b sin ϕb

(7) –V(1/Rg +1/R') –V/R' 0 V(1/Rg –1/R') 1–98
8355A29
This is the usual operating condition of a cavity
which is automatically obtained by its servo tuner.
Case of a matched generator (e.g. klystron 2.4.3.3 Transmission of small modulations
and circulator) Use the same equivalent cir- (AM and PM) through a cavity with
cuit, where Rg is the transmission line impedance beam loading [2]
(transformed to the gap) and replace ig by 2i1 To evaluate the dynamic behaviour of an rf sys-
where i1 = forward traveling wave current in line tem including feedback loops (amplitude, phase,
130
Ch.2: BEAM DYNAMICS
tuning, etc.), it is useful to know the transmission 2.4.3.4 Periodic beam loading at multiples of
of small modulations around the static operating f0 [3, 4]
point. The inputs are modulations of ig (generator
current) in amplitude (AM) and phase (PM), mod- Modulated beam current (e.g. unequal filling of
ulations of ib (RF component of beam current), bunches) at multiples of the revolution frequency
and tune modulations. The outputs are the ampli- f0 results in AM and PM of V . Self-consistent
tude and phase modulations of the cavity voltage solution, valid for only one cavity in the ring or
V. identical, equal-voltage cavities, is given by
AM
Generator current ig to cavity voltage V av
= −ωDO × (22)
PM to PM ab
1 2
ωDO (1 + tan2 φb ) − Δω + (s + σ) tan φb

GG
pp = σ (1 + tan2 φz ) + σ(1 + Y sin φb )s (s + σ)2 − Δω[ωDO (1 + tan2 φb ) − Δω)]
D
PM
+σ 2 Y (sin φb − tan φz cos φb ) (12)
Pv
= −ωDO × (23)
PM to AM ab
1 [ωDO (1 + tan2 φb ) − Δω] tan φb − (s + σ)
GG
pa = − σ 2 Y (cos φb + tan φz sin φb ) (s + σ)2 − Δω[ωDO (1 + tan2 φb ) − Δω]
D

+σ(tan φz − Y cos φb )s (13) Optimal detuning For optimal detuning with
Δω = ωDO = σY cos φb , which corresponds to
AM to AM a real load to the generator, one has
GG G
aa = Gpp (14) AM
AM to PM av ωDO (s + σ + ωDO tan φb ) tan φb
=−
ap = −Gpa
GG G
(15) ab (s + σ)2 − ωDO 2 tan2 φ
b
(24)
where D = s2 + 2σs + σ 2 (1 + tan2 φz ) with s
PM
the Laplace variable.
Pv ωDO (s + σ − ωDO tan3 φb )
Beam current ib to cavity voltage V =− (25)
ab (s + σ)2 − ωDO 2 tan2 φ
b
PM to PM
Magic detuning For “magic” detuning with
Y 2
Δω = ωDO (1+tan2 φb ), one obtains simple, first
GB
pp = σ (tan φz cos φb −sin φb )−σ sin φb s
D order responses
(16)
AM
PM to AM av ωDO tan φb
=−
Y 2 (26)
ab s+σ
GB
pa = σ (tan φz sin φb +cos φb s)+σ cos φb s
D PM
(17) Pv ωDO
AM to AM =− (27)
ab s+σ
GBaa = Gpp
B
(18) With magic detuning, perfect compensation of
AM to PM phase modulations due to beam gaps in collider
ap = −Gpa
GB B rings is possible [3].
(19)
Uniform bunch train For a uniform bunch
Tune modulations from x = Δω/σ to cavity train of length t0 followed by a gap and Δω =
voltage V ωDO , φb ≈ 0, there is no amplitude modulation,
PM and the phase modulation corresponds to a single
σ 2 + σs first-order response with maximum phase excur-
Gxp = (20) sion (for σ 1/T0 )
D
AM 1 R ωc
Δφmax = īb (T0 −t0 ) = ωDO (T0 −t0 )
−σ 2 tan2 φz 2Q V
Gxa = (21)
D (28)
131
where i¯b = average rf component of beam cur- Case b

rent, T0 = revolution period. Periodic modulation 1 R ib
of beam phase (w.r.t. an unmodulated RF) follows Δω = ωc
4Q V
exactly the phase modulation of V , as the energy 2
gain per turn must be identical for all bunches. 1 ib V ib
Pi = V i20 + ≈ (30)
This is shown below for several values of σT0 . 2 2 4
i b (t) for ib i0
ib Installed peak power for a klystron and circu-

lator (φb = 0)
t 2
t0 T0 Rg 1 1 2 V
ϕ( t )
Pi = V 2
+ + − ib (t)
8 Rg R x
Δϕmax
(31)
where x = 12 QR ωc
Δω = cavity reactance.
With cavity on tune for zero beam current
σT0 = 50
(Δω = 0) and optimum matching, one has for
σT0 = 0 case (a),
2
t 1 ib 1
1–98 t0 T0 = + 2 (32)
8355A30
Rg2 V R
2.4.3.5 Rf power needed for transient V2 V2 V ib
Pi =
+ ≈ for R ib V (33)
beam-loading correction [4] 4R 4Rg 4
The objective is to keep V constant independent Pi fluctuates between V ib /8 (when ib = 0) and
V ib /4 (when ib = ib max ).
of ib . The solution is to provide via the rf genera-
With cavity half detuned, Δω =
tor an additional current Δig = −ib to cancel out (1/4)(R/Q)(ωc ib /V ), and optimum match-
ib variations which cannot be corrected by cavity ing, one has for case (b),
detuning (usually slow).
For a fully modulated beam (e.g. injection 1 ib 2 1
2
= + 2 (34)
transient, beam gap), the required power is also Rg 2V R
generally modulated (ig = ig1 or ig2 ), e.g. if the V2 V2 V ib
cavity is tuned for no beam [Δω = 0, φz = 0, Pi = constant =
+ ≈ (35)
4R 4Rg 8
case (a) of figure below]. However, for a half-
for R ib V
detuned cavity [case (b)], the power is unmodu-
lated (the required peak power is in this case min- 2.4.3.6 Traveling-wave cavities [5, 6]
imum).
ib φz = 0
In smooth approximation the transient response of
ib
a matched traveling-wave cavity to a point charge
t
i o = i g1 having synchronous velocity is an rf burst starting
Pi at time z/vp and of duration z/vg (vp , vg = phase
(a)
–ib
t i g2 and group velocities in the structure, z = longitu-
ib
Pi
i g1
dinal coordinate).
t (b)
φz | i g1|= | i g2| In frequency domain,
io φz = 0
1–98
8355A32 R2 −iωz/vp 1 − e−iτ z/L
Eb (z, ω) = ib e (36)
i g2 2 iτ /L
Installed peak power for a tetrode (φb = 0) where Eb = beam-induced rf field, R2 = effective
Consider two cases: (a) cavity on tune for zero shunt impedance of the structure (R2 = E 2 /2P ),
current, and (b) cavity half detuned. L = total length of structure, β = relativistic factor,
and τ = transit time factor,
Case a

1 V ib Δω vg
Pi = V i20 + i2b ≈ for ib i0 (29) τ =L 1− (37)
2 2 vg βc
132
Ch.2: BEAM DYNAMICS
The beam-induced voltage is [4] D. Boussard, CERN SL/91-16 (RFS)

[5] G. Dôme, Proton Linac Conf. (1976) p.138
R2 L2 sin(τ /2) 2 τ − sin τ
Vb = ib − 2i [6] D. Boussard, CERN Acc. School (1996)
4 τ /2 τ2
(38) 2.4.4 Space-Charge Dominated Beams in
For traveling-wave cavities the transfer Guns and Transport Lines
impedances from generator and beam are differ- M. Ferrario, INFN-LNF
ent. This is in contrast with standing wave cav-
ities, which for a given mode exhibit the same Laminarity parameter The dynamics of high
transfer function (circle in figure below) from peak current low emittance beams is strongly
generator or from beam to cavity voltage. Perfect dominated by space charge effects [1]. Elec-
correction of beam loading is not possible with tron emission and transport is affected by self-
traveling-wave structures. This is illustrated by fields produced by the electron bunch itself and
the transient response of a traveling-wave cavity by the image charge induced on metallic bound-
to a step in generator power (linear rise of volt- aries, which oppose the applied fields.
A beam
age) and to a step in beam current (parabolic rise characterised by an rms envelope σ = x2 and
of voltage). transverse normalised rms emittance
Figure below shows the impedance seen by
the beam. Solid curve is for traveling-wave struc- εn = γεg = γ x2 x2 − xx 2 (1)
ture. Dotted curve is for standing-wave structure.
(εg = rms geometrical emittance) is space charge
Im Z τ=–π dominated as long as the space charge collective
forces are largely dominant over the emittance
pressure, or the laminarity parameter
τ = – 2π
Iˆ σ 2
τ=0
ρ= (2)
2IA γ ε2n
Re Z
greatly exceeds unity, where IA = 17 kA is the
τ = 2π
Alfvén current. Under this condition the beam be-
(b) haves like a laminar flow (all beam particles move
on trajectories that do not cross) and transport and
(a)
acceleration require a careful tuning of focusing
τ=π
7–98
8355A27
and accelerating elements in order to keep lami-
Figure below shows the transient response of narity. Correlated emittance growth is typical in
a traveling-wave structure. this regime which can be conveniently made re-
versible [2] if proper beam matching conditions
If, Ig
are fulfilled [3]. When ρ < 1 the beam is emit-
t
tance dominated (thermal regime) and the space
charge effects can be neglected. The transition to
Vg
thermal regime occurs when ρ ≈ 1 corresponding
Linear to the transition energy
t
L/vg
Vb Iˆ σ 2
γtr = (3)
Parabolic 2IA ε2n
t
10–97
L/vg
For example a beam with Iˆ = 100 A, εn =
8355A28
1 μm and σ = 300 μm has a transition energy
of 131 MeV. This energy limit defines also the
References physical extension of the injection system. Space
[1] P. Wilson, IXth Int. Conf. High Energy Acc. charge effects may recur above transition if bunch
(1974) p.57 compressors are active at higher energies and a
[2] F. Pedersen, PAC 75, p.1906 new energy threshold with higer Iˆ has to be con-
[3] F. Pedersen, CERN/PS 92-59 (RF) sidered.
133
DC gun In a DC gun electrons are emitted by The emitted electrons are rapidly accelerated
the cathode and are accelerated in the static elec- to relativistic energies thus partially mitigating the
tric field until they reach an anode with opening to emittance growth due to space charge force ef-
allow beam passage. The limiting current density fects.
is given by the modified Child’s law [4] At the gun exit an approximate expression for
the space charge induced emittance for a Gaussian
4 2e V 3/2 bunch of longitudinal and transverse dimensions
J = ε0 F (4)
9 m d2 σz and σx is [6]
where the enhancement factor F ≥ 1 accounts c Iˆ σz
for the finite cathode size. In a high voltage DC εsc = (8)
8ανrf IA (3σx + 5σz )
gun (V = 500 kV) over a gap of d = 5 cm, the
eEo
field is about 10 MV/m, giving a limiting current where α = 2mc 2 k , Eo being the peak rf field, k =
density of 0.33 A/mm2 with F = 1. The thermal 2πνrf
with ν the rf frequency. Notice that α ≥ 1
c rf
emittance for a thermionic cathode operating at a is the typical operational condition of an rf gun.
temperature T , is ] An important contribution to the beam emit-
tance comes also from time-dependent rf fields
Rc kB T
εth = (5) which can be evaluated as
2 mc2
k3 α 2 2
which gives, per unit cathode size Rc , εth = x = √ σx σz
εrf (9)
0.3 μm/mm at T = 2500 K. 2
In the Pierce gun geometry the cathode has Particles are emitted from the cathode with a
a nonplanar design to provide initial focusing for spread in phase Δϕo that at the gun exit results
the beam. DC injectors are characterized by low to be
cos ϕo
beam energy extraction and long drift regions Δϕe = Δϕo − Δϕo (10)
used for ballistic bunch compression. For high- α sin2 ϕo
current sources there can be significant emittance showing that for ϕo < π/2 (off crest) rf bunch
growth and beam expansion downstream the gun compression occurs.
unless care is taken to confine and control the The induced energy spread σγ = Δγ 2 is
beam with applied magnetic or electrostatic fields. σγ = αkσz (11)
Rf gun In an rf gun electrons are emitted by a
photocathode, located inside an rf cavity, typically with σz = k Δϕe . The final longitudinal
1 2
emittance is
a one and a half cells standing wave structure, il- √
luminated by a laser pulse so that the bunch length εrf
z = 3 (γ − 1) k2 σz3 (12)
and shape can be controlled on a sub-ps time scale An additional contribution to the transverse emit-
via the laser pulse. The limiting current density is tance comes from the chromatic effects in a
given by (4) or in the short pulse limit at the cath- solenoid of effective length lsol and field strength
ode [15] by 2
2 = eBsol
ksol , usually placed near the gun exit,
eε0 E02 2γmc
J= (6) that in the thin lens approximation scales as
me c
2
In an S-band rf gun the peak field E0 is around εch = σγ ksol lsol σx2 (13)
100 MV/m, giving an estimated maximum current Neglecting correlations among all the contribu-
density in the range of 50 A/mm2 . The thermal tions the total normalized emittance at the gun exit
emittance for a metallic photocathode is [5] is

ω − φef f εn = ε2th + ε2sc + ε2rf + ε2ch (14)
εth = σx (7)
3mc2 Envelope equation Under the paraxial ray ap-
where φef f = φw − φSchottky , φw being the ma- proximation px pz , the transverse beam dy-
terial work function and φSchottky the Schottky namics can be conveniently described by the rms
work function. With the typical parameters of a envelope equation that for an axisymmetric beam
Copper photocathode illuminated by a UV laser is [1]
the expected thermal emittance per unit spot size γ Ksc ε2g
is around 0.5 μm/mm. σ + σ + kext
2
σ= + 3 (15)
γ σ σ
134
Ch.2: BEAM DYNAMICS
The first term is the change in the envelope slope, are the two limiting solutions for the space charge
the second term drives the adiabatic damping due and thermal regime respectively. These solutions
to acceleration, the third term accounts for lin- represent the matching conditions for which the
ear external focusing forces, the fourth represents external focusing exactly balances the internal
the defocusing space charge effects and the fifth forces.
the internal pressure due to the emittance. In the Correlated emittance When transverse-
space charge dominated regime, where ρ > 1, the longitudinal correlations within the bunch are
emittance term can be neglected. In the transi- important, as the one induced by the space charge
tion regime, ρ ≈ 1, both terms should be con- and rf fields, a simple model can be considered
sidered. When ρ < 1 the thermal regime oc- by slicing the bunch in N longitudinal slices
curs and the space charge term can be neglected. of envelope σs (z, ζ) and divergence σs (z, ζ)
ˆ
Ksc = 2IAI γ 3 is the beam perveance, Iˆ is the peak (ζ = z−zLtail is the normalized longitudinal
coordinate along a bunch of length L) whose
current and γ = eE acc
mc2 , Eacc being the accelerat-
evolution can be computed from a slice envelope

ing field. The rms effective geometrical emittance
equation equivalent to (15) provided that all
ε2g = ε2th + ε2B includes the contribution of cath-
bunch parameters refer to each individual slice
ode thermal emittance εth2 and
an equivalent emit- (γs , γs , Ksc,s ). Correlations within the bunch
eBσ
tance term: εB = 8mc , derived from Busch may cause (correlated) emittance oscillations that
theorem, that accounts for the centrifugal poten- can be evaluated once an analytical or numerical
tial when the canonical momentum pϑ , induced [8] solution of the slice envelope equation is
by a residual magnetic field on the cathode is dif- known by using the definition
ferent from zero.
2 = k 2 + k 2 , in a
The external focusing kext sol rf
εg,cor = σs2 σ 2s − σs σs 2 (19)
typical photoinjector design, is due to the solenoid where the average is performed over the entire
field and to the ponderomotive rf focusing effects bunch slices ensemble. In the simplest case of a 2
described by [7] slices model the previous definition reduces to
( (
2 1 eEacc 2 1 γ 2 εg,cor = (σ1 σ2 − σ2 σ1 ( (20)
krf = η = η
8 γmc2 8 γ that represents a very useful formula for an esti-
where on-crest acceleration only has been consid- mation of the emittance scaling.
ered. The quantity η is a measure of the higher Emittance oscillation An interesting example
spatial harmonic amplitudes of the rf wave and is the beam dynamics in the space charge dom-
is generally quite close to unity in standing wave inated regime, nearly matched to an external fo-
structures and close to 0 in traveling wave struc- cusing channel, without acceleration (kext =
tures. ksol ) [3], described by the slice envelope equation
Brillouin flow An important case is an intense Ksc,s
uniform-density beam matched to an external fo- σs + ksol
2
σs = (21)
σs
cusing channel without acceleration (kext = ksol ,
ˆ
γ = 0, εg = εth ) described by the envelope equa- A stationary solution is σs,B = k21 Ig(ζ)
2γ 3 IA
sol
tion
Ksc ε2th where the local dependence of the current Iˆs =
σ + kext
2
σ= + 3 (16) ˆ (ζ) within the bunch has been explicitly indi-
Ig
σ σ
cated (g (ζ) is the space charge field form fac-
The Brillouin flow [1] is the equilibrium solution
tor). Since ksol has a slice independent constant
(σ = 0) of (16) given by
value, the Brillouin matching condition cannot
be achieved at the same time for all the bunch
1 2 4 4
σ= σsc + σsc + 4σth (17) slices. Assuming there is a reference slice per-
2 fectly matched with an envelope σr,B , the match-
where ing condition for theother slices can be written as
√ σsB = σrB + σrB δIs
with respect to the ref-
Ksc εth 2 Iˆ
σsc = , σth = (18) erence slice. Considering a small perturbation δs
kext kext
from the equilibrium in the form σs = σs,B + δs
135
and substituting in Eq.(21) we obtain an equation which is formally identical to Eq.(15) with the
2
offset: δs +2k
for the slice √ 2
ext δs = 0. The solution scaled parameters k̂ext2 2
= kext + 14 γγ =
δs = δo cos 2kext z , where δo = σso − σsB is 2
2 + γ 1 η
the amplitude of the initial mismatch that we as- ksol γ 4 + 8 , εn = γεg and K̂sc =
sume for simplicity the same for all slices, leads γKsc . Equilibrium solutions in the reduced vari-
to the expression ables (“invariant envelope” in the literature [3])
√ √
σs = σs,B + δo cos 2kext z (22) are σ̂sc = K̂sc and σ̂th = εn
in the space
k̂ext k̂ext
Equation(22) shows that slice envelopes oscillate charge and thermal regime respectively, corre-
around the equilibrium solution with the same sponding to the matching conditions for the beam
envelope
(plasma) wavelength for all slices which depends

only on the external focusing forces. Using the ˆ A
2I/γI
σsc = 2 for ρ > 1 (25)
two slices model and substituting solution (22),

with s = 1, 2, in Eq.(20) the emittance evolution Θ +γ 2 η2 +1
results, eB
( √ ((
where Θ = mc and
1 ( ΔI
εg,cor = ksol σrB (( δo sin 2kext z (( (23) εn
4 Iˆ σth = 1/2 for ρ < 1 (26)
2
Θ + γ η2 +1
2
where ΔI = Iˆ1 − Iˆ2 . Note that in this simple case
envelope oscillations of the mismatched slices in- Note that condition (25) scales like γ −1/2 while
duce undamped correlated emittance oscillations (26) is independent of γ. For the transition regime
which periodically goes back to zero. Emittance (ρ ≈ 1) the matching condition can be derived
oscillations are the basic features of the emittance from an equivalent expression of (17) with re-
compensation technique. duced variables.
Emittance compensation Space charge in- An approximate expression for emittance os-
duced emittance growth is partially correlated and cillations in the space charge dominated regime,
can be reduced in a drift downstream the elec- i.e. when γ < γtr , can be derived in a similar
tron source with a solenoid located at the exit of way as (23) taking into account acceleration [3],
the gun [2]. In order to prevent additional space giving
charge emittance growth in the subsequent accel-
erating sections, emittance oscillations have to be δ0 I/IA (( √ (
(
damped and the final emittance minimum has to Δεn (z) ∝ (cos ψ − 2 sin ψ ( (27)
γ 2γ
be reached at the transition energy so that space √
charge forces become ineffective. To this end the where ψ ≡ ln(γ/γ0 )/ 2. Before the transi-
beam has to be properly matched to the accelerat- tion energy is achieved the emittance performs
ing sections (Brillouin like flow) in order to keep damped oscillations with wavelength depending
under control emittance oscillations via the pon- on the external fields and with amplitude depend-
deromotive rf focusing force [3]. In many cases ing on the current profile. A careful tuning of the
rf focusing is too weak to provide sufficient beam external fields and bunch charge profile can mini-
containment. A long solenoid around the accel- mize the value of the emittance at the injector ex-
erating structure is a convenient replacement to traction. A successful application of the emittance
provide the necessary focusing. This possibility compensation technique can be seen in [10, 11].
is included in the following discussion. The previous description of emittance com-
The matching conditions for the beam subject pensation for space charge effects applies equally
to acceleration (γ = γo + γ z) can be derived fol- well to both rf and DC guns.
lowing the previous example. With the substitu-
√ Velocity bunching (VB) Emittance compensa-
tion of the reduced variable σ̂ = σ γ [9], the tion is a key feature of the Velocity Bunching
envelope equation (15) becomes technique [12, 13], a method to compress the
K̂sc ε2 bunch using rectilinear trajectories in an rf struc-
σ̂ + k̂ext
2
σ̂ = + n3 (24) ture by means of the longitudinal rf focusing ef-
σ̂ σ̂
fect. Under the assumption that the beam current
136
Ch.2: BEAM DYNAMICS
Table 1: Rf and wavelength scaling laws. 2.4.5 Space Charge Effects in Circular
Charge Wavelength Accelerators
Q ∝λ B. Zotter, CERN
σx , σy , σz ∝ Q1/3 ∝λ 2.4.5.1 Direct space charge effects
Eo ∝ λ−1
B ∝ λ−1 Impedances The electric and magnetic forces
εsc ∝ Q2/3 ∝λ (FE and FM = −β 2 FE ) due to a straight beam
of charged particles nearly cancel when their ve-
εth ∝ Q1/3 ∝λ
locity v = βc is ultrarelativistic (β ≈ 1), both in
εrf ∝ Q4/3 ∝λ free space and in a perfectly conducting, concen-
tric smooth pipe. The factor 1 − β 2 = 1/γ 2 1
then reduces the direct space-charge force and

grows as a linear function of the energy like Iˆ = hence the longitudinal space-charge impedance
γ ˆ
γo Io the matching condition in the space charge (See also Sec.3.2.4)
dominated regime is given by
Z iZ0
= g (1)
2Iô n 2βγ 2
σsc = (28)
γo IA Θ + γ 2 η2 + 1
2 where n = ω/ω0 . The g-factor is g = 1 +
2 ln(b/a) for a circular cylindrical beam of radius
where γo and Io are the values for the current a in a concentric pipe of radius b > a.
and the energy respectively at injection into the Space-charge force is quite weak in pro-
VB structure. Also this solution is independent ton machines with GeV energies, and com-
of γ. pletely negligible in high-energy electron ma-
Charge and wavelength scaling laws Once an chines. However, in proton or ion machines in
optimized rf injector design has been found, a the MeV range, space charge is often the domi-
simple procedure allows to keep the optimized nant impedance. It corresponds to a negative in-
performances with a different bunch charge Q ductance, sometimes called capacitive although it
and/or rf wavelength λ, by simply scaling the is ∝ ω like an inductance. In machines operating
relevant parameters as shown in Tab.1 [14]. The above transition this may lead to negative mass
resulting emittance scaling is also reported. instability (Sec.2.4.9).
The transverse space-charge impedance is
References
RZ0 1 1
Z⊥ = i 2 2 − (2)
[1] M. Reiser, Theory and Design of Charged Parti- β γ a2 b2
cle Beams, Wiley, New York, 1994 where R = average machine radius. For a uni-
[2] B.E. Carlsten, NIM A285, 313 (1989) form beam with elliptic cross section [1], half
[3] L. Serafini, J.B. Rosenzweig, PR E55, 7565 axes a1 and a2 , the first term in brackets should be
(1997) replaced by 2/a1,2 (a1 + a2 ), and by 1/σx,y (σx +
[4] K. Togawa et al, PRST-AB 10, 020703 (2007) σy ) for a Gaussian distribution.
[5] D.H. Dowell et al, PRST-AB 12, 074201 (2009)
Betatron tune shifts are proportional to Z⊥ Ib
[6] K.J. Kim, NIM A275, 201 (1989)
and can bring tunes onto resonances leading to
[7] J.B. Rosenzweig, L. Serafini, PR E49, 1599
particle losses. In high-current machines, the tune
(1994)
[8] M. Ferrario et al, Int. Journal of Modern Physics
variation can be reduced by surrounding the beam
A, Vol. 22, No. 3 (2007) with an rf shield following the beam profile such
[9] C. Wang, PRE E4, 046502 (2006) that the ratio b/a remains constant. In bunched
[10] R. Akre et al, PRST-AB 11, 030703 (2008) beams, tunes are different in the center and at the
[11] M. Ferrario et al, PRL 99, 234801 (2007) edge of a bunch, and the whole tune region should
[12] L. Serafini, M. Ferrario, AIP Conf. Proc. 581, 87 avoid low order resonances.
(2001) The direct space charge force is centered on
[13] M. Ferrario et al, PRL 104, 054801 (2010) the beam and not the chamber axis. Hence it is
[14] J.B. Rosenzweig, E. Colby, report TESLA-95-04 independent of any transverse displacement of the
[15] J.B. Rosenzweig et al, NIM A57, 87 (2006) beam and does not affect rigid dipole oscillations.
However, it changes the external focusing forces
137
which leads to the beam envelope equation for the while the additional transverse forces are
beam half-width ax (see also Sec.2.4.4), 2
Nb E0 r0 − s 3x F1 (s/σs )
Fx (s) = − √ 2e 2σs2 +
d2 ax 2 4λr0 2πRσs (9Rσs2 )1/3
+Kx (s)ax − x3 − 2 3 = 0 (3)
ds 2 ax β γ (ax + ay ) Nb E0 r0 y s
Fy (s) = − √ F1 (7)
The equation for the beam half-height ay is ob- 2πRσs (9Rσs2 )1/3 σs
tained by interchanging all subscripts x and y. where
∞
Here Ki is the external focusing strength, includ- du
(v − u)e−(v − u)
2
ing possible gradient errors, εx,y the emittances F0 (v) = 1/3
u
in the x or y plane, λ the (constant) line density, 0 ∞
du −(v − u)2
r0 the classical particle radius, and γ the energy F1 (v) = e (8)
u2/3
factor. The space charge force leads to an en- 0
velope modulation which reduces the tune shift These transverse forces as well as the longi-
(0)
to 34 Δνsc for the antisymmetric (“quadrupole”) tudinal energy gradient in dispersive regions con-
(0) tribute to emittance growth.
mode, and to 12 Δνsc for the symmetric (“breath-
(0)
ing”) mode, where Δνsc is the space-charge tune 2.4.5.2 Betatron frequency shifts
shift when the modulation is neglected.
In addition to direct space-charge fields, induced
Curvature effects The compensation of elec- fields due to charges and currents in the surround-
tric and magnetic forces is perturbed when the ing vacuum chamber walls (“image fields”) shift
beam trajectory and the surrounding vacuum the betatron frequencies, which is often called
chamber are curved. The impedance of a perfectly Laslett tune shifts [8]. Below we present the verti-
conducting chamber of rectangular cross section cal tune shifts; the horizontal ones can be obtained
(half height h, half width w) bent in a circle of by exchanging y ↔ x and height ↔ width.
radius R is purely reactive [2], Incoherent tune shifts refer to changes of be-
tatron frequencies of single particles in a beam.
Z 2h 2 2hω 2
Im ≈ Z0 A − 3B (4) Let ωy = νy ω0 be the unperturbed betatron fre-
n πR πc
quency, Fy the vertical forces due to induced and
where the constants A and B are close to unity space-charge fields, I0 the beam current, and y0
when w ≥ 2h, and about 0.7 when the offset of beam centroid. To first order,
w = h.
Eq.(4) holds for k = ω/βc < (π/h) R/w, cor- ∂Fy (( ∂Fy ((
Fy = y ( + y0 ( (9)
responding typically to a frequency of 100 GHz. ∂y y=y0 =0 ∂y0 y=y0
At higher frequencies, there are resonances with The first term gives rise to a change of the inco-
R/Q values given in terms of Bessel function ze- herent betatron tune νinc with
1 ∂Fy ((
ros of large order [3]. Based on a simpler model
of a circular beam between two infinite plates 2
νinc = νy2 − ( (10)
separated by 2g, the real part of the radiation m0 γω02 ∂y y=y0 =0
impedance can be approximated by [4] When the shift Δνinc νy is small, Δνinc ≈
Z g (νy2 − νinc
2 )/2ν . Incoherent tunes of different
inc
Re = 300 [Ω] (5) particles in a beam usually vary over a finite range
n R called tune spread.
The effect of curvature on transverse beam mo- Coherent tune shifts change the frequency of
tion due to the centrifugal space charge force was the beam centroid. Measurements of tunes, e.g.
first considered important [5], but it is cancelled with pick-up electrodes sensitive only to average
almost completely by that of the transverse elec- motion, show the coherent tune which satisfies

∂Fy (( ∂Fy ((
tric field on particles with different energies [6].
1
Nevertheless, curved trajectories can lead to emit- 2
νcoh = νy2 − ( + (
tance growth due to additional forces [7]. The en- m0 γω02 ∂y y=y0 =0 ∂y0 y=y0
ergy gradient along a Gaussian bunch becomes (11)
For small shifts, Δνcoh ≈ (νy2 − νcoh
2 )/2ν
coh .
dE 2Nb E0 r0 s The difference between ωinc and ωcoh is pro-
=√ F0 (6)
cdt 2π(3R2 σs4 )1/3 σs portional to ∂Fy /∂y and ∂Fy /∂y0 and equals
138
Ch.2: BEAM DYNAMICS
Table 1: Laslett coefficients for simple geometries.
Coefficient circular elliptic cross section par.plate comments
εsc,V 1/2 b/(a + b) - direct space charge

εsc,H 1/2 b2 /a(a
+ b) -

h2 2 2K 2
ε1V /H 0 ± (1 + k ) − 2 ±π 2 /48 incoherent electric
12d2 π
ε2V /H - - ±π 2 /24 incoherent magnetic
h2 2
2K
ξ1V 1/2 2
−1 π 2 /16 coherent electric
4d π

h 2 2Kk 2
ξ1H 1/2 1 − 0
4d2 π
ξ2V - - π 2 /16 coherent magnetic
ξ2H - - 0
U + (1 + i)V , where U and V are the transverse and w are √ the chamber half height and width,
dispersion relation coefficients, proportional to k = 1 − k2 is the complementary modu-
the transverse impedance Z⊥ , lus, and K = K(k ). [This relation is much
Z⊥ r0 I0 simpler when expressed with the nome, defined
U + (1 + i)V = i (12) as q = exp(−πK /K), which is a common
Z0 γνy e
parameter in most computer codes for elliptic
Unbunched beams The vertical tune shifts for integrals.]
unbunched beams in a perfectly conducting vac- When a beam is partially neutralized, and η
uum chamber of half-height h, between perfect expresses the ratio of densities of particles of op-
magnetic pole pieces at a distance ±g from the posite charge to particles in the beam, the elec-
axis, can be written tric fields due to direct space-charge and due to

2r0 RI0 β ε1 2 ε2 εsc induced charges are reduced, while the magnetic
Δνinc=− +β + (13) field remains unchanged. This can strongly in-
ec β 3 γ h2 g2 γ 2 b2
⎧ crease the direct space-charge field, which is no
⎪
⎪ ξ1 β ξ2
2
longer multiplied by 1 − β 2 = 1/γ 2 but by
2r0 RI0 β ⎨h2 + g2
Δνcoh=− 1 − β 2 − η. To avoid excessive tune shifts when
ec β 3 γ ⎪ ⎪ β 2 ε1 ξ1 β 2 ε2 η 1/γ 2 one needs to reduce neutralization, e.g.
⎩ 2 + 2 2+ 2
h γ h g with clearing electrodes or lower rest gas pres-
where β is the average β-function (= R/ν in sure. Tune shifts of unbunched beams with neu-
smooth approximation), r0 the classical radius tralization are given in Eq.(14) with B = 1. Im-
of the particle. The first expression for Δνcoh , age coefficients for off-axis beams can be found
called integer formula, is valid for very thin walls in [10, 11].
through which ac magnetic fields can penetrate, The image coefficients ε2 and ξ2 for closed
while the second one, called half-integer formula, magnetic boundaries (e.g. circular or elliptic) can-
is for the usual case of thicker walls where these not be calculated for μ → ∞ since the induced
fields do not penetrate [9]. The transition is dis- magnetic field would not permit a charged beam
cussed in [9]. to pass through. They have therefore been calcu-
The Laslett coefficients for beams in the cen- lated for perfect magnetic yokes with gaps [12].
ter of some common geometries are shown in Closed magnetic yokes are used e.g. in super-
Tab.1, where K(k) is the complete elliptic inte- conducting magnets, but there the coefficients
gral of the first kind with modulus k, given by ε2 = ξ2 → 0 since the magnetic material is driven
exp(−πK /K) = (w − h)/(w + h), where h completely into saturation (μ → 1).
139
Bunched beams For bunched beams, several influence the beam dynamics, iii) longitudinal
terms in the expression for the tune shifts are pro- phase-stability is maintained by accelerating off-
portional to the peak rather than the average cur- crest with negative synchronous phases, and iv)
rent. This can be taken into account by division radial defocusing in accelerating gaps must be
with the bunching factor B < 1, defined as the taken into account [1, 2, 3, 4].
ratio of average to peak current. This yields Energy gain and transit time factor For ex-
citation of an accelerating gap (considered a
2r0 RI0 β 1 − β 2 − η ε1
Δνinc = − β 2
+ cylindrical pillbox cavity) in a TM010 stand-
ec β 3 γ B h2
ing wave mode, the non-zero fields are written
ε2 1 − β − η εsc
2
Ez (r, z) cos(ωt + φ), Er (r, z) cos(ωt + φ) and
+ β2 2 + (14)
g B b2 Bθ (r, z) sin(ωt + φ). Assume an accelerating gap
2r0 RI0 β of length L with the origin z = 0 at the electrical

Δνcoh = − center of the cell. Consider a particle traversing
ec β 3 γ
⎧ the accelerating gap such that at t = 0 it is lo-

⎪ 2 1 − β 2 − η ξ1
⎪ ξ2 cated at z = 0. The phase of the rf field is φ. The
⎨ β + 2
+ β2 2
× B h g energy gain in the gap is
⎪ β 2 ε1 + 1 − β − η ξ1 + β 2 ε2
⎪ 2
⎩ 2
ΔW = qV0 T (r, k) cos φ
h B h2 g2
where k = ω/cβ = 2π/βλ,
The neutralization in bunched beams is usually L/2
weaker than in unbunched ones, and can be fur- V0 = E0 L = Ez (0, z)dz
ther reduced by leaving a gap in the filling pattern. −L/2
For machines with vacuum chambers con- E0 is the peak axial field, and T (r, k) is the
sisting of sections with unequal dimensions, one transit-time factor,
needs to calculate the tune shifts for each cham- L/2
1
ber type separately, and add the contributions in T (r, k) = Ez (r, z) cos kz dz
proportion to their lengths. V0 −L/2
With cylindrical symmetry,
References T (r, k) = T (0, k)I0 (Kr) = T (k)I0 (Kr)
[1] L. Teng, Report ANLAD-59 (1963) where T (k) is the axial transit-time factor, and

[2] Y.K. Ng, R. Warnock. PRD 40 (1989) 231 2π 2
[3] R. Warnock. P. Morton, SLAC PUB 4562 (1988) K 2 = k2 − k02 =
γβλ
[4] A. Faltens, L.J. Laslett, PA 4 (1973) 152
[5] R. Talman, PRL 56 (1986) 1429 where k0 = ω/c.
[6] E. Lee, PA 25 (1990) 241 For Ez constant within the accelerating gap
[7] V. Derbenev, Y. Shiltsev, SLAC-PUB at the drift tube bore radius r = a [Ez (a, |z| <
7181(1996); P. Emma, R. Brinkman, PAC L/2) = const] but zero outside the gap,

97; R. Li, C.L. Bohn, J.I. Bisognano, PAC 97 2πa I0 (Kr) sin(πL/βλ)
[8] L.J. Laslett, BNL Report 7534 (1963) p.324 T (r, k) = J0 (1)
λ I0 (Ka) πL/βλ
[9] B. Zotter, CERN/ISR-TH/72-8 (1972)
[10] B. Zotter, CERN/ISR-TH/74-11 and 38 (1972) which gives for the axial transit-time factor, in the
[11] G. Guignard, CERN Report 77-10 (1977) limit a/λ → 0, the familiar result
[12] L.J. Laslett, Report UCRL 18892 (1969) sin(πL/βλ)
T (k) =
πL/βλ
2.4.6 Beam Dynamics in Proton Linacs When considering multiple accelerating gaps,
S. Henderson, FNAL it is useful to define the synchronous particle as
A. Aleksandrov, ORNL that ideal particle which arrives at each gap with
the design beam/rf relative phase (known as the
Proton linacs differ from electron linacs in sev- synchronous phase) prescribed in the “phase law,”
eral ways, all of which are due to the fact that and which therefore acquires the design energy
the proton beam is not ultrarelativistic: i) the gain in each gap. The synchronous particle has
spacing between accelerating gaps increases with phase and energy coordinates φs and Ws and ve-
particle velocity, ii) space-charge effects strongly locity cβs .
140
Ch.2: BEAM DYNAMICS
0.8
The transit time factor for a multicell struc-
ture is given by Eq.(1) with K replaced by Ks = 0.6
2π/γs βs λ.
0.4
The transit time factor for a multicell π-mode
structure (e.g. elliptical multi-cell cavity) with 0.2
W − Ws (MeV)
field Ez (z) = E0 cos ks z, where ks = π/L, is 0

π sin((βs /β − 1)π/2) −0.2
T (r, k) = I0 (Ks r)
4 (βs /β − 1)π/2

−0.4
sin((βs /β + 1)π/2)
+ (2) −0.6
(βs /β + 1)π/2
−0.8
−30 −20 −10 0 10 20 30 40 50
φ − φs (degrees)
which gives, for β = βs , T (0, ks ) = π/4.
Longitudinal dynamics (See also Sec.2.1.2) Figure 1: Longitudinal phase space trajectories and
Consider the motion of a particle with phase separatrix (red) with φs = −20◦ .
and energy coordinates φ(s) and W (s). Define
conjugate particle coordinates relative to those of The half-height of the separatrix at φs , the en-
the synchronous particle: ϕ(s) = φ(s) − φs (s) ergy acceptance, is
and w(s) = W (s) − Ws (s). In smooth approxi-
mation with small acceleration rate, and assuming 2qE0 T mc2 βs3 γs3 λ
2
wmax = (φs cos φs − sin φs )
that E0 T, φs , and βs are constant, the equations of π
motion become For small amplitude oscillation, the equation
2π ∂H of motion for the phase becomes ϕ + kz0 2 ϕ = 0
ϕ (s) = − w= with the longitudinal zero-current phase advance
mc2 βs3 γs3 λ ∂w
per meter
2πqE0 T sin(−φs )
w (s) = −qE0 T [cos φs (1−cos ϕ)+sin φs sin ϕ] 2
kz0 =
mc2 βs3 γs3 λ
∂H
=− The phase advance per focusing period of length l
∂ϕ
is νz0 = kz0 l. The phase-space trajectory is an el-
which derive from a Hamiltonian lipse of the form w2 /ŵ2 + ϕ2 /ϕ̂2 = 1 with ŵ2 =
qE0 T mc2 βs3 γs3 λ sin(−φs )ϕ̂2 /2π. The area of
2π w2
H=− −qE0 T [sin φs (cos ϕ−1) the phase space ellipse is an adiabatic invariant,
mc2 βs3 γs3 λ
2 so that after acceleration, an ellipse with initial
+ cos φs (sin ϕ − ϕ)] (3) phase and energy widths ϕ̂i , ŵi is transformed to
an ellipse with final widths given by ϕ̂f /ϕ̂i =
Fig.1 shows the separatrix and resulting phase
(βi γi /βf γf )3/4 and ŵf /ŵi = (βf γf /βi γi )3/4 .
space trajectories, which display stable motion
The phase motion is damped at the expense of an
about the synchronous phase φs . The phase ac-
increase in energy spread.
ceptance of the separatrix extends from ϕmax =
When acceleration is included, that is, when
φmax − φs = −2φs to ϕmin = φmin − φs , which
βγ =const in Eq.(3), the shape of the input accep-
is given by solution of the equation
tance resembles a golf-club, as shown in Fig.2.
sin φmin − φmin cos φs + sin φs − φs cos φs = 0 Also shown in Fig.2 are two trajectories which
demonstrate the phase damping.
The total phase acceptance, Ψ = ϕmax − ϕmin , is Transverse beam dynamics Transverse defo-
related to the design synchronous phase by cusing terms arise from the Er and Bθ rf field
components in the accelerating gaps. The radial
sin Ψ − Ψ
tan φs = defocusing is given by
1 − cos Ψ
qE0 T (k)LI1 (Kr) sin φ
which for Ψ 1 is tan φs ≈ −Ψ/3. Note that Δ(γβr ) = −
mc2 βs γs
the phase acceptance shrinks to zero as φs → 0.
141
1.5
and
qV0
1
Wc = Wi + [T (rc , kc ) cos φc
2
+S(rc , kc ) sin φc ]
L 2π qV0 kc
W − W (MeV)
0.5
φc = φi + + T (rc , kc ) sin φc
2 βi λ 2mc2 βc2 γc3
s

0
− S (rc , kc ) cos φc
L qV0
rc = ri + ri − T (rc , kc ) cos φc
−0.5
2 2mc2 βc2 γc3 r

−Sr (rc , kc ) sin φc
−1
−40 −20 0 20 40 60 80 qV0
φ − φs (degrees) rc = ri − [Tr (rc , kc ) sin φc
2mc2 βc2 γc3

Figure 2: Longitudinal phase space trajectories and +Sr (rc , kc ) cos φc ] (5)
input acceptance (red) for adiabatic acceleration with
where
φs = −20◦. L/2
2
S(r, k) = dzEz (r, z) sin kz
V0 0
which for Kr 1 gives L/2
2
Tr (r, k) = dzEr (r, z) sin kz
πqE0 T (k)L sin φ V0 0
Δ(γβr ) = − r
L/2
mc2 βs2 γs2 λ 2
Sr (r, k) = dzEr (r, z) cos kz (6)
The rf defocusing is more severe at low energies, V0 0
and must be taken into account in the linac de- and a prime denotes taking derivative with re-
sign. The combined focusing from quadrupoles spect to k. For cylindrical symmetry, Tr (r, k) =
or solenoids and rf gaps is included in determin- T (k)kI1 (Kr)/K. The full and mid-gap transfor-
ing transverse stability of the motion via stability mation equations are coupled and may be solved
diagrams [5, 6]. by iteration. In the above transformations, (small)
Gap transformation equations Proton dynam- gap correction terms proportional to r have been
ics in an accelerating gap may be calculated using neglected. See [7, 3] for the full treatment.
a thin-lens method in which the beam is assumed Space charge dynamics The effect of space
to drift at constant velocity to the accelerating gap charge forces on beam dynamics decreases pro-
center, where the energy, phase and radial coordi- portional to 1/β 2 γ 3 and is therefore important for
nates are advanced, followed by drift at the new nonrelativistic beams in high-intensity accelera-
velocity to the end of the cell [7, 3, 8]. The ini- tors.
tial coordinates at the entrance to the accelerating The Kapchinskiy-Vladimirskiy (K-V) enve-
cell are Wi , φi , ri , ri , the coordinates at the exit lope equations [9] describe horizontal, X, and
of the cell are We , φe , re , re , and those at the gap vertical, Y , beamsizes (ellipse semi-axes) of a
center are Wc , φc , rc , rc . The gap transformations continuous uniform density beam in a periodic
are quadrupole-focusing channel,
d2 X 2 2K
We = Wi + qV0 T (rc , kc ) cos φc + kx (s)X − x3 − =0
ds 2 X X +Y
L 2π 1 1
φe = φi + + d2 Y 2y 2K
2 λ βi βe + ky (s)Y − − =0 (7)
ds2 Y3 X +Y
qV0
+ 2 2 3 kc T (rc , kc ) sin φc where kx,y (s) represents the external focusing
mc βc γc
force, x,y is the total emittance, and
L qV0
re = ri + (ri +re )− 2 2 3 Tr (rc , kc ) cos φc qI 1
2 mc βc γc K=
qV0 2π0 mc3 β 3 γ 3
re = ri − Tr (rc , kc ) sin φc (4) is the generalized perveance.
mc2 βc2 γc3
142
Ch.2: BEAM DYNAMICS
The K-V envelope equations can be gener- f is the geometric form-factor which is a func-
alized to an arbitrary charge density distribution tion of the ellipsoid aspect-ratio parameter p =
√
with elliptical symmetry in x-y space. Second γσz / σx σy . For 0.8 < p < 5, a useful approxi-
moments of the distribution, or rms sizes σx,y sat- mation is f = 1/3p [12]. The Trace-3D code cal-
isfy the rms envelope equations [10, 11] culates beam envelopes for a bunched beam with
linear space charge force [13].
d2 σx 2x,rms K/2
2
+ kx (s)σx − − =0 (8) The periodic solution of Eqs.(11) provides en-
ds σx3 σx +σy velopes of a beam matched to the focusing chan-
d2 σy 2y,rms K/2 nel. Structure resonances can occur when oscilla-
2
+ ky (s)σ y − 3
− =0 (9) tion of the beam size deviation from the matched
ds σy σx +σy
values is in resonance with the periodic focus-

ing force for some beam envelope modes. To

where σx = x2 , σy = y 2 are the beam
avoid the strongest resonance, the transverse zero-
2 current phase advance per period should be kept

sizes, and x,rms = x2 x2 − xx , y,rms =
below 90◦ [14].
2
y 2 y 2 − yy are the rms emittances. The collective field of the beam exhibits a de-
The fact that Eqs.(7,7) and (8,9) have the focusing force on individual particles within the
same form justifies the concept of equivalent uni- beam thus reducing the focusing action of ex-
form beams: rms envelopes of elliptically sym- ternal fields. The tune depression η quantifies
metric distributions are identical, independent of the strength of the space-charge defocusing force.
the details of the distributions. Therefore, a uni- The longitudinal and transverse tune depressions
form distribution can be used to describe real are
beams. This concept does not hold exactly in the 2
case of bunched beams, although it was shown νz f K3 L
ηz = = 1−
that the dependence on the details of the distri- νz0 σx σy σz νz0
bution is weak [11]. The electric field of a 3-D el-
lipsoidal bunch with semi-axes rx , ry , rz and uni- νx,y (1 − f )K3 L 2
ηx,y = = 1−
form charge distribution, given by νx0,y0 (σx +σy )σx,y σz νx0,y0
3Iλ(1 − f ) x where L is the length of the focusing period,
Ex = νz , νx,y are the phase advances per focusing pe-
4π0 c(rx + ry )rz rx
riod with space-charge, and νz0 , νx0,y0 are the
3Iλ(1 − f ) y zero-current phase advances.
Ey =
4π0 c(rx + ry )rz ry Halo and emittance growth The nonlinear
3Iλf z time-dependent space-charge forces in a beam
Ez = (10)
4π0 crx ry rz with non-stationary distribution can cause the rms
emittance to increase. The free-energy concept
can be used to obtain the “equivalent” rms en-
states that a stationary beam is in its lowest en-
velope√equations for a bunched beam [1] (where
ergy state and any deviation from equilibrium, or
rx = 5σx , etc.),
mismatch, provides a source of free energy which
d2 σx 2x,rms K3 (1 − f ) can transfer to an emittance increase. This con-
+ kx (s)σ x − − =0 cept provides an estimate of the upper limit for
ds2 σx3 (σx +σy )σz
the emittance growth in a continuous beam [14],
d2 σy 2y,rms K3 (1 − f )
+ ky (s)σy − − =0
ds 2 σy3 (σx +σy )σz f 2h
= 1+ 2
i η
d2 σz 2z,rms K3 f
+ kz (s)σ z − − =0 (11) where i , f are the initial and final emittances
ds2 σz3 σx σy
(average for x and y for non-round beams), η
where is the tune depression, and h is a dimensionless
3qIλ 1
K3 = √ free-energy parameter which can be calculated for
20 5π0 mc3 β 2 γ 3 each particular type of mismatch [14].
is the 3-D space charge parameter, I = qN c/λ In the case of strong anisotropy between
is the beam current averaged over the rf period, longitudinal and transverse planes in a bunched
143
beam, the degree of freedom with higher tem- TraceWin [19], and Track [20] are examples of
perature can be a source of free energy for the more popular codes [21].
emittance growth in the other plane. To prevent The principal challenge in modern high cur-
emittance redistribution the condition of equipar- rent proton linac design is the minimization of
tition, nx kx = ny ky = nz kz , has to be met beam loss due to halo growth [22, 23, 24, 25].
[14], where n is the normalized emittance, and
k is the phase advance per unit length including References
space charge.
Another approach is to slow down the redis- [1] T. Wangler, RF Linear Accelerators, Wiley
tribution process by avoiding space charge cou- (2008)
pling resonances. The so-called ’Hofmann dia- [2] P. Lapostolle, A. Septier, eds., Linear Accelera-
gram’ is a useful tool for choosing an optimal tors, North-Holland (1970)

[3] B. Schnizer, PA 2 (1971) 141
working point trajectory when designing a high
[4] N. Pichoff, Proc. CERN Acc. School (2000)

intensity linac [15].
[5] L.W. Smith, R.L. Gluckstern, RSI 26 (1955) 220
Space-charge-induced emittance growth is
[6] M. Prome, in [2]
often associated with the formation of low density [7] A. Carne et al, in [2]
tails, or halo, which can cause particle loss and [8] P. Lapostolle, Los Alamos LA-11601-MS
is therefore especially undesirable for high inten- [9] I.M. Kapchinsky, V.V Vladimirsky, Proc. Int.
sity linacs. The principal halo-formation mecha- Conf. on High Energy Acc. (1959) p.274
nism is a resonant interaction between the parti- [10] P. Lapostolle, IEEE Trans. Nucl. Sci. NS-18
cles in the beam and a beam core that is oscillat- (1971) 1101
ing because of mismatch. The particle-core model [11] F.J. Sacherer, IEEE Trans. Nucl. Sci. NS-18
uses the envelope equation for calculating space- (1971) 1105
charge fields in a mismatched core and single par- [12] P.M. Lapostolle, CERN report AR/Int. SG/65-15
ticle equations for calculating individual particle (1965)
motion in that field. In the case of a round contin- [13] K.R. Crandall, D.P. Rusthoi, Los Alamos LA-
uous beam [16], UR-97-886
[14] M. Reiser, Theory and Design of Charged Parti-
d2 r η2 1 − η2
2
+r− 3 − =0 (12) cle Beams, Wiley (1994)
ds r r [15] I. Hofmann, PRST-AB 6 (2003) 024202
% & % [16] T.P. Wangler et al, PRST-AB 1 (1998) 084201
d2 x 2 x/r 2 |x| < r
+x−(1−η )× = 0 (13) [17] H. Takeda, LANL LA-UR-98-4478
ds2 1/x |x| ≥ r
[18] J. Qiang et al, J. Comp. Phys. 163 (2000) 434
where r is the ratio of the core radius to the [19] D. Uriot et al, PAC 03, p. 3491
matched core radius, x is the ratio of the parti- [20] V. Aseev et al, PAC 05, p. 2053
cle displacement to the matched core radius, and [21] R. Ryne, ICFA HB 08
s = k0 z is the phase advance. [22] Proc. 29th ICFA Adv. Beam Dyn. Wkshp,
The maximum resonant particle amplitudes HALO03, AIP Conf Proc. 693
can be found by solving Eqs.(12, 13) numerically [23] S. Nath et al, PAC 03, p.1515
as a function of mismatch parameter μ, defined as [24] F. Gerigk et al, LINAC 02, p.569
the ratio of the initial beam radius to the radius [25] R. Duperrier, IPAC 10, p. 1246
of the matched beam. In the range of tune de-
pression 0.5 < η < 0.9 and mismatch parameter 2.4.7 Vlasov and Fokker-Planck Equations
0.5 < μ < 0.95 and 1.05 < μ < 2 the maximum B. Zotter, CERN
amplitude is given by an approximated formula:
xmax 4(1 + | ln μ|) [16]. Liouville theorem The Vlasov Equation de-
There is no complete theory to describe dy- scribes the evolution of a continuous distribution
namics of non-stationary beam distributions an- function ψ in phase space. It is essentially a col-
alytically and numerical simulations should be lisionless Boltzmann equation, or an expression
used for detailed study. Several codes have of the Liouville theorem, which states that, in
been developed to calculate linac beam dynam- the absence of collisions and dissipation, the lo-
ics including space-charge forces using particle- cal density in phase space must remain constant
in-cell methods. PARMILA [17], IMPACT [18], dψ/dt = 0. Expressed in conjugate variables p
144
Ch.2: BEAM DYNAMICS
and
q [1], In case damping and noise come from syn-
chrotron radiation in electron storage rings, we
∂ψ ∂ψ ∂ψ have [3, 4]
q˙ ·
+ + p˙ · =0 (1)
∂t ∂q ∂
p Horiz. Vert. Long.
c c ηc
Written in spatial and velocity coordinates x, v , (q, p) (− x , x) (y, y ) (z, − δ)
ωx ωy ωs
∂ψ ∂ψ 1 ∂ψ U0 U0 U0
+ v · + F (x, t) · =0 (2) α
∂t ∂x m ∂v 2ETrev 2ETrev ETrev
with d
p/dt replaced by the force on a particle of ρ2 Du c2 Du η 2 c2 Du
D
charge e and mass m. The Lorentz force F = 2
E νx 4 2
2ωy E γ 2 2 ωs2 E 2
+ v × B)
e(E should be obtained from the self- cσx cσy ω s σz
consistent EM fields. rms σx = σy = σδ =
ωx ωy ηc

Linearized Vlasov equation Consider 2-D Cq γ 2 ρ Cq ρ Cq γ 2
phase space (p, q), and write ψ as the sum = 2
= √ =
νx 2νy 2ρ
of a time independent (stationary) part ψ0 (p, q) 2 γ 4 /ρ, D
and a small time dependent (perturbation) part where U0 = (4π/3)r √ m
e e c u =
ψ1 (p, q, t). For a Hamiltonian system, q̇ = Ṅ u2 /2 = (55/48 3)r √ e m e c4 γ 7 /ρ3 (u photon
∂H/∂p and ṗ = −∂H/∂q, ψ0 satisfies energy), Cq = (55/32 3)/me c, E = me c2 γ,

η slip factor, ρ bending radius. The rms size is
∂ψ0 ∂H ∂ψ0 ∂H given by σq2 = σp2 = D/2α. In the x-motion, x
− = [ψ0 , H] = 0 (3)
∂q ∂p ∂p ∂q is identified with p rather than q because quantum
excitation occurs in x rather than x .
where the square brackets are Poisson brackets. It
follows that ψ0 = ψ0 (H), and does not depend References
explicitly on p or q.
The linearization consists in neglecting terms [1] A. Hofmann, CERN 95-06, Vol.I (1995) p.259
of second order in small quantities, [2] J. Jowett, AIP Proc. 153 (1985) p.864
[3] M. Sands, SLAC-121 (1970)
∂ψ1 ∂ψ1 1 ∂ψ1 1 ∂ψ0 [4] A.W. Chao, Lecture Notes in Phys. 296, Springer-
+ q̇ + F0 = − F1 (4) Verlag (1986) p.51
∂t ∂q m ∂p m ∂p
where F0 is the stationary part of the force (due
to the applied external EM fields or force gen- 2.4.8 Potential Well Effect
erated by ψ0 , e.g. through wake fields); F1 is B. Zotter, CERN
the small, time-dependent part of the force gen- Symbols Ib beam average current, nB number of
erated by ψ1 through wake fields. If the pertur- bunches in beam, Vrf rf voltage, νs0 = ωs0 /ω0 ,
bation is oscillatory ∼ exp(−iωt), one can re- νs = ωs /ω0 zero-current and perturbed syn-
place (∂/∂t) → −iω. The equation permits chrotron tunes, b vacuum chamber pipe radius, τ
the determination of ω (complex). The distribu- time displacement of particle from bunch center,
tion (and thus the beam) becomes unstable if ω h harmonic number, η slip factor, ϕs synchronous
has a positive imaginary part exceeding radiation phase, 0 zero-current bunch length, and ΔE
damping. bunch length and energy spread, ξ = Ib /V̇rf =
Fokker-Planck equation [2] In the presence of αp eIb /νs2 E = 2πIb /hVrf cos ϕs scaling parame-
dissipative force or random noise, the phase space ter (dimension Ω−1 ). We have
density is no longer conserved. This leads to the
hηVrf cos ϕs |η|c ΔE
FP equation νs0 = , = (1)
2πβ 2 E/e βωs E
∂ψ ∂ψ ∂ψ ∂2ψ
q˙ ·
+ + p˙ · = 2αψ + D 2 (5) Equation of motion [1]
∂t ∂
q ∂
p ∂
p
d2 τ 2
2 ξ
ωs0 Zhom
α, D = the damping and diffusion coefficients. + ω s0 τ = + τ Zpot (2)
dt2 2π ω0
145
where the higher-order-mode impedance Zhom This leads to a cubic equation for x = σ/σ0 ,
and the potential-well impedance Zpot are √
ξ 2π
∞ x3 − x + Zc 3 3 = 0 (9)
Zhom = ReZ(pω0 )J0 (pω0 τ̂ )λ̃(pω0 ) 2π ω0 σ0
p=−∞ Note: (i) Eq.(9) does not apply if b where
∞ resonant impedances dominate. (ii) Coherent syn-
≈ ReZ(pω0 )λ̃(pω0 ) chrotron radiation can also produce bunch length-
p=−∞ ening [3], although usually weaker than effect of
∞ the impedance.
J1 (pω0 τ̂ )
Zpot = ImZ(pω0 ) λ̃(pω0 ) Turbulent threshold The energy spread of an
ω0 τ̂ /2
p=−∞ electron bunch remains unchanged with increas-
∞ ing current up to the turbulent threshold, where

≈ ImZ(pω0 )pλ̃(pω0 ) (3) also the slope of bunch length versus currents in-
p=−∞ creases. Dynamics in this regime is not com-

where τ̂ is amplitude of synchrotron oscillations, pletely understood. The bunch length above
and λ̃ is Fourier transform of bunch line density. threshold can be expressed in terms of the scal-
The second expressions are valid for small ampli- ing parameter ξ. For an impedance described by a
tude synchrotron oscillations. The perturbed syn- power law Z (ω) = (Z )0 (ω/ω0 )a , one gets [4]
chrotron frequency is ( ( 1/(2+a)
σz ( Z (
ξ ∝ (( (( ξ (10)
νs2 = νs02
1− Zpot (4) R n 0
2π
while the center of the oscillation is shifted by For an inductive wall a = 1, and σ ∼ ξ 1/3 .
Δϕ = hξZhom /(2π). However, turbulence is mainly due to the high fre-
Proton bunch lengthening Without syn- quency part of the impedance evaluated at ωcrit =
chrotron radiation, the longitudinal emittance c/σz . Then [5],
(∝ ΔE) is invariant. It follows from Eq.(1) that ( (
σ 3 e αp I0 (( Z (( Z
z
2 ωs = 20 ωs0 (5) =√ 2 E/e ( n ( +
R 2πnB νs0 crit n pot
For > b, the impedance is mainly inductive with (11)
Z /n = iω0 L = constant (n = ω/ω0 ), and for The turbulent threshold current is
bunch with parabolic line density, [2] √
2πkb νs0 2 E/e (Z/n)pot
3 Ib Zc Ith = 1+
νs2 = νs0
2
1+ 2 3 (6) e αp |Z /n|crit |Z/n|crit
π B hVrf cos ϕs
( (
( Z ( gZ0 (12)
Zc = (( (( − The bunch length at threshold σth /σ0 = 1 +
n 2βγ 2
1/2
where B = full /2πR the bunching factor. The (Z/n)pot /|Z/n|crit . Above threshold,
second term in Zc is from direct space charge, g =

1 + 2 ln(b/a) (Sec.2.4.5.1). (Z/n)pot −1/2
Eliminating the synchrotron frequency yields νs ∝ 1 +
|Z/n|crit
a quartic equation for the bunch-lengthening fac- 1
tor x = /0 , σE (Z/n)pot − 2 σ
= 1+ (13)
ξ 3 σE0 |Z/n|crit σ0
x4 + Zc x − 1 = 0 (7)
2π π 2 B03 Haissinski equation For given wake function,
where B0 = 0 /(2πR). the equilibrium line density satisfies [6]
τ
Electron bunch lengthening For electrons, the
energy spread is constant for low currents. The λ(τ ) = K exp −U0 (τ ) − ξ dtS(τ + t)λ(t)
−∞
energy distribution is Gaussian. For linear mo- (14)
tions the line density is also Gaussian. Eq.(1) where U0 (τ ) = τ 2 /2σ02 for a linear rf voltage
yields τ
(parabolic potential), S(τ ) = 0 W (t)dt is the
σωs = σ0 ωs0 (σ = σz /c) (8) step-function wake potential with W the wake
146
Ch.2: BEAM DYNAMICS
[10] K. Bane, SLAC-PUB-5517 (1990)

[11] K. Oide, KEK Report 90-21 (1990) p.64
bunch
current:
high high
2.4.9 Single-Bunch Instabilities in Circular
low
Accelerators
low
B. Zotter, CERN
These instabilities can occur either in the longitu-
dinal or in one of the transverse directions (usu-
Figure 1: Electron bunch in resonator impedance for ally more limiting vertically).
various currents. Robinson instability The detuning of a cav-
ity by the induced current of a bunched beam

constant K is deter-
function (Sec.3.2.1), The leads to stability limits for the tuning angle ψ =
mined by normalization λ(t)dt = 1. tan−1 [2Qr (ωr − ωrf )/ωr ] [1, 2],
For a purely resistive impedance Rs , the Vrf cos ϕs
Haissinski equation can be solved [7], 0 < sin(2ψ) < (1)
Rs I0
2 (2/π) exp τ 2 /2σ02 for machines operating below transition. The first
λ(τ ) = √ (15)
ξσ0 Coth(ξ) − erf(τ /σ0 2) condition requires tuning of the fundamental res-
where ξ ∝ R onance ωr below the rf frequency ωrf . The sec-
s Ib ; it can be obtained from the nor- ond criterion limits the tuning angle as function
malization dτ λ(τ ) = 1. The solution corre-
sponds to an asymmetric bell shape. For an in- of the rf voltage Vrf , the synchrotron phase angle
ductive impedance S(τ ) = Lδ(τ ), hence [8] ϕs , shunt impedance Rs and average current I0 .
Up to now, the second limit has not been clearly
τ2 ξ observed. Above transition, the opposite inequal-
λ(τ ) = K exp − 2 + λ(τ ) (16)
2σ0 2 ities hold. See also Sec.7.3.1.1.
Eq.(16) can be solved numerically and corre- Higher-order cavity modes may occur on ei-
sponds to bunch lengthening above transition. For ther side of a harmonic p of the revolution fre-
a capacitive impedance, bunch shortens above quency, contribute to damping, or drive instabili-
transition. ties with growth rates
For more realistic resonator impedances, so- 1 2ωs I0
= (2)
lutions can be found by a perturbation algorithm τR hVrf cos ϕs
[9]. See Fig.1. For other impedances, the Haissin- ×Re {Z [(p + ν)ω0 ] − Z [(p − ν)ω0 ]}
ski equation is integrated numerically. A diver-
gence may occur for higher currents which has The fundamental resonance is usually strongest,
no physical significance [10]. The bunch defor- its correct tuning yields Robinson damping. The
mation is important for determining instability sum over p and all impedances determines beam
thresholds [11]. stability.
Negative mass instability [3] Particles with
References v ≈ c cannot travel much faster by acceleration
but will increase their momentum and thus move
[1] B. Zotter, CERN Report SPS/81-14 (1981) with a larger radius of curvature. Above transi-
[2] S. Hansen et al, PAC 75, p.1381 tion energy, particles which normally repel each
[3] C. Pellegrini, A. Sessler, Nuovo Cimento 53 B other, seem to experience an attractive force. This
(1968) 198 is known as negative mass effect for its similar-
[4] A. Chao, J. Gareyte, PA 25 (1990) 229 ity with the formation of planetary rings with an
[5] A. Hofmann, J. Maidment, LEP-Note 168 (1979) otherwise attractive gravitational force outside the
[6] J. Haissinski, Nuovo Cimento 18 B (1973) 72
Roche limit. In uniform beams this may lead to
[7] A. Ruggiero, PAC 77, p.1205
self-bunching and particle loss when the current
[8] G. Besnier, B. Zotter, KEK Report 90-21 (1990)
exceeds a threshold
p.52
[9] J. Hagel, B. Zotter, KEK Report 90-21 (1991) E0 |η| ΔpFWHM 2
p.59 Ithresh = F (3)
e γZ/n m0 c
147
where ΔpFWHM is the momentum spread (full Table 1: Form factors F for various distributions.
width at half maximum). The form factor F , of F arg Z = tan−1 (Im Z /Re Z)
order unity, depends on the particle distribution
(see Tab.1). Distrib. −π/2 0 π/2 Remarks
Resistive wall instability Currents are induced sharp-
Parabolic 1.047 0.798 0
in metallic vacuum chamber walls by the beam. edge
Due to the finite resistivity of the walls, they 3/2
extend behind the position of ultrarelativistic 1.061 1.061 1.061 Keil-
power Schnell
charges. Their EM fields act on charged parti-
cles arriving later and may increase their oscilla- Quartic 1.073 1.194 2.203 rounded
tion amplitudes, and cause instabilities in bunched Gaussian 0.942 1.359 ∞ rounded
or unbunched beams, in the longitudinal [4] or a

Triangular sharp-
transverse direction [5]. 0 2.0 0
edge
The skin depth in a metalof resistivity ρc
and permeability μ is δs = 2ρc /(μω). For Rounded with 1%
0.555 1.98 1.265
metal walls thicker than δs , the resistive wall triangle rounding
impedances of a circular pipe of radius b are
Z μZ0 μRZ0 For long bunches, e.g. of proton beams, the
(ω) = δs , Z⊥ (ω) = δs (4) impedance is mainly inductive and the threshold
n 2μ0 b μ0 b3
current is given by the Boussard criterion [8],
where 2πR is accelerator circumference. Since
Z /n ∝ Z⊥ ∝ δs ∝ ω −1/2 , the largest impedance Z E0 |η|γ σz ΔpFWHM 2
≤ F (6)
occurs at the lowest frequency. For transverse in- n e Ib cT0 p
stabilities, this lowest frequency occurs at where E0 = mc2 , Ib = NB e/T0 . This agrees
ωmin = ω0 min{ν − [ν], [ν + 1] − ν} (5) with the (simplified) unbunched-beam or Keil-
Schnell criterion [9],
where [ν] = integer part of the betatron tune ν.
Z E0 |η| ΔpFWHM 2
See Eq.(11), Sec.2.4.10 for growth rate. This sin- ≤ F (7)
gle spectral line has narrow band width, and thus n e γI0 m0 c
oscillates for a long time and may cause coupled applied to the local values of current and energy
bunch instabilities. Impedances at all other beta- spread in a bunch. The form factor F , shown in
tron frequencies may be grouped into a wide fre- Tab.1, depends on particle distribution and ratio of
quency band. These fields decay rapidly, hence real to imaginary part of the impedance [10, 11].
drive only single bunch instabilities such as head- For short bunches, the impedance seen by the
tail modes discussed below. beam is dominated by resonances at higher fre-
Longitudinal coupled-bunch instabilities can quencies, and may lead to turbulent bunch length-
also be caused by a resistive wall, but are usu- ening. A tentative explanation for this is longitu-
ally weaker. In general, impedances due to all el- dinal mode coupling [12] (Sec.2.4.10).
ements should be added to determine beam stabil- Operating with η < 0 can reduce
ity, as can be done with computer programs [6, 7]. bunch lengthening since the usually inductive
Microwave instability In addition to potential impedance then shortens it [13]. However, theo-
well bunch lengthening (Sec.2.4.8), an increase retical and experimental investigations show that
of bunch current Ib can lead to longitudinal in- the turbulent threshold is lower with the shorter
stabilities of a single bunch, independent of the bunch, giving a larger energy spread.
presence of other bunches in the machine. [See Head-tail instability Short-range transverse
also Eqs.(12-15), Sec.2.4.10] Because it is often wake fields, excited by particles at the head of
accompanied by high-frequency signals, it is also a bunch, may excite oscillations at its tail. Syn-
called microwave instability. Oscillation frequen- chrotron motion brings these particles again to the
cies change with amplitude due to non-linearities, head and they continue to excite particles behind.
hence this instability is usually self-limiting and These oscillations will grow (head-tail instability)
only rarely leads to particle loss, but may reduce if they add in phase due to a finite chromaticity
luminosity in colliders since bunches do not keep ξ = Δν/δ [14, 15] and if the growth rate exceeds
an optimum distribution. radiation and Landau damping.
148
Ch.2: BEAM DYNAMICS
In the lowest mode, m = 0, all particles in a where σ = ω ẑ/(2c), ẑ is maximum z-amplitude

bunch oscillate in phase, which corresponds to a of synchrotron oscillation of all particles (wa-
rigid dipole oscillation at the betatron frequency. terbag model), η = η0 +η1 δ+· · · is the phase slip
The m = 0 mode is unstable for ξ < 0. Since the factor (Sec.2.3.11). For the case m = 1 (dipole
natural chromaticity of circular machines is neg- mode), Eq.(11) can be rewritten as
ative, one must correct it with sextupoles. Bunch 1 cη1 ẑ dΔE
shape modes with m ≥ 1 oscillate at synchrotron =− (12)
τ1 3η0 CE dẑ
side-band frequencies ωβ ± mωs . In the m = 1
mode, particles at head and tail have opposite where ΔE is the energy loss of the beam bunch to
phases (when ξ = 0), the bunch “toggles” about the impedance. Its dependence on bunch length ẑ
its middle, while higher modes (m ≥ 2) oscil- characterizes the growth rate of the longitudinal
late with m nodes over the bunch length. Modes head-tail instability.
with m ≥ 1 may become unstable for ξ > 0, Transverse mode coupling instability (TMCI)
but their growth rates are usually small and easily TMCI, also called fast head-tail instability or
stabilized by damping except for large machines strong head-tail instability, occurs when the fre-
with very strong transverse wakes. quencies of two neighboring head-tail modes ap-
For short bunches, the growth rate of the m-th proach each other due to detuning with increas-
mode is (ignoring radial modes) [16] ing current during accumulation. The original
1 NB r0 c ξωβ ẑ name transverse turbulence [19] is more appro-
= priate for beams being injected into machine well
τm 2π 2 γ cηνβ2 above threshold, when a large number of modes
∞
ω ẑ ω ẑ may become coupled simultaneously.
× dωReZ⊥ (ω)Jm ( )Jm ( ) (8)
0 c c Its threshold current is lowest in the y-plane
with NB number of particles per bunch, ẑ the since many elements are less high than wide,
z-amplitude of synchrotron oscillation (airbag 2πνs E/e
model), r0 classical particle radius, ωβ the beta- Ithresh = F (σz ) (13)
βi Z⊥ i
tron frequency. i
Example For resistive wall [15], The form factor F ≈ 1 for short bunches.
√ √ For longer bunches, it increases proportionally
1 2Jm r0 c NB ξ ẑ
= − 5/2 √ (9) to σz , and is essentially the ratio of the ma-
τm π μσc ηγb3 νβ chine impedance to the effective impedance.
with μ permeability, σc conductivity, b For Gaussian bunches and broad-band resonator
chamber half-height. The factor Jm =
√ impedances, Ithresh can be expressed with the
π/2
dx sin x
π ψ transverse loss or kick factor κ (σz ) which elim-
0 dψ cos mψ sin 2 for ⊥
0 inates the need for a bunch length correction fac-
m = 0,1,2 is approximately 2.9, −0.57, −0.21. tor,
For negative chromaticity ξ < 0, the m = 0 mode C1 fs E/e
Ithresh = (14)
is strongly unstable, while the higher modes are βi κ i (σz )
⊥
weakly unstable for ξ > 0. i
Example For the case with constant wake- The constant C1 ≈ 8, but is often replaced by
function W1 (z) = −W0 , 2π [11] as in Eq.(13). For more exact calcula-
1 8 NB r0 W0 ξ ẑ tions of the threshold one should use computer
= (10) programs solving for the coherent modes [20], in-
τm π(4m2 − 1) cηγT0 Z0 νβ
cluding bunch lengthening [21] and potential-well
Longitudinal head-tail instability This insta- deformations with current.
bility appears when the momentum compaction For the case of constant wake with W1 (z) =
has an appreciable nonlinear dependence on en- −W0 and a two-particle model beam bunch, the
ergy. It has been seen at the CERN-SPS [17]. The TMCI threshold occurs at
growth rate of the instability can be written [18] 8 ωβ ωs (E/e)

1 16m2 η1 NB r0 ∞ ReZ (ω) Ithresh =
π cW0
(15)
= dω (11)
τm 3η0 γCZ0 −∞ σ2 where ωβ,s are the betatron and synchrotron fre-
× [σJm (σ)Jm+1 (σ) + (1 − m)Jm
2
(σ)] quencies.
149
TMCI is a severe limitation of single bunch total beam current Itot = nB IB , the coherent
current for large high-energy electron storage frequency shift of for m ≥ 1 is [2]
rings which have a high transverse impedance (m)
m ωs Itot Zeff
Z⊥ ≈ (2R/b2 )Z/n because of their large radius Δωm = −i (1)
R and the large number of rf cavities. m + 1 3B 3 hVT cos ϕs
The incoherent synchrotron frequency ωs = νs ω0
References is a function of beam current, ωs0 is its value for
vanishing current, VT = Vrf ωs2 /ωs0 2 = the total
[1] K. Robinson, CEA Report CEAL-1010 (1964) voltage seen by the beam, Vrf = applied rf voltage
[2] Yu. F. Orlov, E.K. Tarasov, JEPT 34(7), No.3, with harmonic number h = ωrf /ω0 , and ϕs =
339 (1958) stable phase angle. Above transition, 90◦ < ϕs <
180◦ is needed for stability, hence cos ϕs < 0. An
[3] X. Nielsen, A. Sessler, Proc. Int. Conf. High En-

ergy Acc. (1959) p.239 inductance, for which Zeff = −iω0 L, then gives
[4] K. Neil, A. Sessler, RSI 36 (1965) 429 a positive coherent frequency shift, compensating
[5] J. Laslett, K. Neil, A. Sessler, RSI 36 (1965) 436 almost exactly its negative incoherent frequency
[6] A. Hofmann et al, PAC 79, p.1514 shift. The bunching factor B = nB f0 τL is the
[7] M. Zisman et al, LBL-21270 (1985) ratio of bunch length L = βcτL to bunch spacing
[8] D. Boussard, CERN/PS-BI (1972)
2πR/nB .
[9] E. Keil, W. Schnell, CERN/ISR-TH/69-48
Assuming the modes do not depend on
(1969)
impedance or beam current, the effective
[10] A. Ruggiero, V. Vaccaro, CERN/ISR-RF/68-33
(1968)
impedance (Sec.3.2.5) can be expressed
∞
[11] B. Zotter, CERN Report ISR-GS/76-11 (1976) (m) Z (ωmp )
[12] F. Sacherer, PAC 77, p.1393 Zeff = ω0 Hm (ωmp ) (2)
p=−∞
ωmp
[13] S.X. Fang et al, KEK 94-190 (1995)
[14] C. Pellegrini, Il Nuovo Cimento 44 (1969) p.447 where the spectral power density Hm (ω) ∝

[15] M. Sands, SLAC Report TN-69-8 and 10 (1969) |λ̃m (ω)|2 is normalized by p Hm (ωmp ) = 1.
[16] A.W. Chao, Physics of Collective Beam Instabil- The oscillation spectrum is ωmp = ω0 (nB p + n +
ities in High Energy Accelerators, Wiley (1993)
mνs ). The summation over all frequencies can be
[17] D. Boussard, T. Linnecar, EPAC 90, p.1560
replaced by one over only positive harmonics with
[18] B. Chen, Thesis, U. Texas at Austin (1995)
upper and lower sidebands, as illustrated in Fig.1
[19] R. Kohaupt, Int. Conf. Part. Acc. (1980) p.562
[20] Y. Chin, CERN/SPS 84/2 (1984)
for the case of 4 Gaussian bunches in a resonator
[21] B. Zotter, CERN/LEP/AP 89-82 (1989) impedance.
2.4.10 Sacherer Formulae

B. Zotter, CERN 0 1 2
3 3
2
In longitudinal phase space x = r cos θ, ẋ = 1
ωs0 r sin θ, a bunch shape mode ∼ exp(−imθ) 0
is specified by the azimuthal mode number m 3 2 1 0 3 2 1 0
0
(m = 1 dipole, m = 2 quadrupole etc.), when 1–98
8355A52 ωs ωo
only the lowest radial mode is considered. Its pro-
jection onto real space λm (x) is called line den- Figure 1: Effective impedance as summation over
sity mode. The coherent frequency ωm can be positive harmonics with upper and lower sidebands,
obtained from the (complex) coherent frequency weighted by spectral power density of a Gaussian.
shift Δω = ωcoh − ωinc , if the incoherent oscil- Indices 0,1,2,3 are multibunch mode number for
lation frequency ωinc is known [1]. For nB > 1 nB = 4.
bunches, the coupled bunch mode number 0 ≤
n < nB gives the phase-shift 2πn/nB between Proton bunches have mostly a parabolic distri-
oscillations of adjacent bunches. bution, and Legendre modes λm ∼ Pm (τ ) are
Longitudinal bunch oscillations For nB eigenmodes in an inductive impedance. How-
equally spaced and populated bunches, with ever, sinusoidal modes λ(τ ) ∼ sin[πmω0 τ ] or
150
Ch.2: BEAM DYNAMICS
cos[π(m + 1)ω0 τ /2] (for m odd or even) give Transverse bunch oscillations The transverse
better agreement with observation. Their spectral mode frequencies are ωmp = ω0 (nB p + n −
power density is mt νβ + mνs ). The transverse mode number mt
1 − (−)m+1 cos ωτL is usually taken to be unity (dipole oscillations),
Hm (ω) = Cm (m+1)2 2 (3) while the longitudinal one (m ≥ 0) should in-
( ωτπL )2 −(m+1)2 clude the rigid bunch oscillation m = 0, which
usually has the largest frequency shift and is most
with Cm = 2nB ω0 τL /π 3 for normalization. Pro- likely to become unstable [6, 7].
ton bunches are usually long compared to the vac- The coherent frequency shift of a bunch with
uum chamber diameter; the impedance is mainly total current Itot = nB IB and energy E is [3, 4]
inductive, Z (ω)/n = −iω0 L is constant and is
(m)
equal to Zeff , hence the frequency shift is mostly
iec Itot Z⊥ eff
real. When the bunches are short compared to the Δωm = − (8)
m + 1 4πνβ EB
rf wavelength, the coherent shift of the dominant

dipole mode (m = 1) is equal and opposite to the with the effective transverse impedance
incoherent one, and thus ω1,coh remains constant ∞

(m)
with current. Z⊥ eff = Z⊥ (ωmp )Hm (ωmp − ωξ ) (9)
Electron bunches usually have a Gaussian distri- p=−∞
bution, and oscillate in Hermitian modes, Hm is evaluated at the mode frequency ωmp
Hm (ωmp ) = Dm (ωmp σ)2m exp(−ωmp 2
σ 2 ) (4) shifted by the chromatic frequency ωξ = ξω0 /ηνβ
Dm = Γ(m + 1/2)/nB ω0 for normalization. where ξ = dνβ /δ.
Neglecting potential well deformation and (i) For a constant impedance Z⊥ eff ≈
turbulence, the frequency spread due to a sinu- Z⊥ Hm (ωξ ), one obtains the unbunched beam re-
soidal rf voltage is sult divided by the bunching factor, and by (m+1)
since higher modes are more difficult to drive.
π 2 1 + 23 sin2 ϕs 2 (ii) For a narrow-band impedance, when only
S = ωs B (5)
16 cos2 ϕs a single resonance at frequency ωR contributes to
The effective spread for a parabolic bunch is ap- the sum for the effective impedance, the coherent
√
proximately mS/2(m+1), while it is mS/4 for frequency shift becomes
Gaussian bunches. iec Itot Z⊥ (ωR )
When modes with different mode numbers do Δωm = − F (χ − ωR τL )
m + 1 4πνβ E0
not couple, an instability may occur if the total
(10)
frequency shift |Δω| is larger than the effective
where χ = ωξ τL is the chromatic phase-shift
frequency spread; Landau damping is lost, and if
across the bunch, and F (ωτL ) = Hm (ω)/B.
the growth rate 1/τ = −ImΔω exceeds radiation
(iii) The resistive wall impedance has a sin-
damping.
gle large peak at ωmin [Eq.(5), Sec.2.4.9]. The
Resonator impedance With resonant fre- narrow-band (or long range) contribution is ap-
quency ωR , shunt impedance Rs and quality proximately
factor Q, the coherent frequency shift is
Δωs 0.159 Rs I0 iec Z⊥ (ω0 )Itot π
= Fm (ωR τL )D(αd ) (6) Δωm = − (11)
ω0 BhV cos ϕs m + 1 4πνβ E0 τL nB B
The form factor Fm (x) is proportional to xm for The short range contribution can be approximated
short bunches, while it is 64m/πx2 for long ones. by replacing the infinite sum for the effective
The factor D depends on the attenuation αd = impedance by an integral over all frequencies, and
πωR is usually negligible.
nB Qω0 between adjacent bunches,
Longitudinal mode coupling Different oscil-
1 1
D(x) = −iαd − (7) lation modes can strongly couple when detun-
1 − ex+ 1 − ex− ing makes two mode frequencies approach each
where x± = −2πi(n ± ωR /ω0 )/nB − αd . The other. A single bunch will become unstable for
maximum value of |D| is about unity for nar- currents above a threshold given essentially by the
row band, and approaches zero for wide band res- unbunched beam expression applied to the peak
onators. current. For longitudinal oscillations, turbulent
151
bunch lengthening will occur until the bunch is 2.4.11 Landau Damping
stable again. This is also called microwave insta- K.Y. Ng, Fermilab
bility (Sec.2.4.9). Its threshold can be expressed
in terms of matrix elements which are functions Wake force excites a number of collective waves
of impedance and beam current, in a beam and displaces it from its equilibrium po-
sition. These waves of the beam center exchange
(m,k) energy among themselves, resulting in growth in
m ωs Itot Zeff
Mmk = −i (12) amplitude for some and damping for some. The
m + 1 3B 3 hVT cos ϕs
spread in oscillation frequency accelerates the
with the generalized effective impedance damping and decelerates the growth. This process
∞ is often called Landau damping [1, 2, 3, 4].
(m,k) Z(ωmp )
Zeff = Hmk (ωmp ) (13) Transverse oscillation of unbunched beam

ω
p=−∞ mp
/ω0 Consider a coasting beam of energy E0 and beta-
tron frequency ωβ under the influence of a trans-

where Hm,k (ω) = λ̃m (ω)λ̃∗k (ω) is the cross verse wake W1 (Sec.3.2.1). The transverse dis-
power density. The coherent frequency ω is de- placement y(θ, t) of a particle at azimuthal angle
termined by the condition θ around the accelerator ring is
( (
( ω − mωs − Mm,m −Mm,m+1 ( ∂ ∂ 2
( ( + ω0 + ωβ2 y(θ, t)
( (
( ( = 0 (14) ∂t ∂θ
( −Mm,m+1 ω − (m + 1)ωs ( 0
( −M ( eI0
m+1,m+1 =− dt W1 (t − t)y(θ, t ) (1)
γmβcT0 −∞
The threshold current is determined by
where T0 = 2π/ω0 is the revolution period, βc the
|Mm,m+1 | = 12 |Mm+1,m+1 − Mm,m − ωs | (15) particle velocity, I0 is the average beam current,
and · · · implies averaging over all particles ac-
To couple different azimuthal modes requires cording to the distribution ρ(ωβ ). Solution can be
strong impedances at frequencies above the beam- obtained in the harmonic-frequency
pipe cutoff, which occur rarely in practice. How- (n-Ω) space
via the transformation (1/4π 2 ) dθdte−inθ+iΩt .
ever, impedances at lower frequencies can cou- In the upper part of the Ω-plane (Im Ω > ωg , with
ple radial modes with the same azimuthal mode ωg being the fastest growth rate of y(θ, t)),
number m, when the bunch shape is asymmet-
ric, e.g. due to potential well deformation in re- 2iω̄β κỹn (Ω) i(Ω−2nω0 )yn0
ỹn (Ω) = − (2)
sistive impedances [8, 9]. Contrary to its trans- ωβ2 − ω̂ 2 2π(ωβ2 − ω̂ 2 )
verse counterpart (TMCI, Sec.2.4.9), coupling of
radial modes occurs at lower beam currents than where
the initial beam displacement is y(θ, 0) =
inθ , ẏ(θ, 0) = 0, ω̂ = Ω−nω , ω̄ = ω ,
coupling of different azimuthal modes. n yn0 e 0 β β
κ = ecI0 Z1⊥ (Ω)/(2ω̄β E0 T0 ), and Z1⊥ the trans-
verse impedance (Sec.3.2.1). Physically, with
References
κ = κR + iκI , κI is the betatron frequency shift
[1] F. Sacherer, CERN/SI-BR/72-5 (1972) due to impedance and κR is the growth rate. Ana-
[2] F. Sacherer, CERN/MPS-BR/73-3 (1973); PAC lytic continuation into the whole Ω-plane gives
73, p.825
ρ(ωβ )
[3] F. Sacherer, Proc. 9-th Int. Conf. on Part. Acc. iyn0 (ω̂−nω0 ) dωβ
C β − ω̂
ω 2 2
(1974) p.347
ỹn (Ω) = −
[4] F. Sacherer, Erice School (1976), CERN 77-13 H(Ω)
(1977) p.198; CERN/PS-BR/76-21 (1976)
ρ(ωβ )
[5] F. Sacherer, CERN/PS-BR/77-5 and 6 (1977); H(Ω) = 1 − i2ω̄β κ dωβ 2 (3)
PAC 77, p.1393 C ωβ − ω̂ 2
[6] A. Hofmann, K. Hübner, B. Zotter, PAC 79, with integration path C going below the pole at ω̂
p.1514
and above the pole at −ω̂. Back to the θ-t space,

[7] B. Zotter, CERN/LEP/89-74 (1989)
[8] K. Oide, K. Yokoya, KEK-Preprint 10 (1990)
y(θ, t) = dΩ ei(nθ−Ωt) ỹn (Ω) (4)
[9] M. D’yachkov, R. Baartman, EPAC 94, p.1075 W
n
152
Ch.2: BEAM DYNAMICS
with path W above all poles. Thus the center of

1.0 (1)Lorentzian (4)Parabolic
the beam consists of many harmonic waves at fre-
(2)Rectangular (5)Cosine−sq.
quencies determined by the zeroes of H(Ω). (3)Elliptical (6)Gaussian
As an example, consider the Lorentzian dis- 0.5
tribution ρ(ωβ ) = (S 12 /π)/[(ωβ − ω̄β )2 + S 21 ],
−Im(Δω)0/S1/2
2
where S 12 is the HWHM. Since S 21 /ω̄β 1 and
|κ|/ω̄β 1, keeping the lowest order, the solution 0.0
(3) (4) (5)
simplifies to (2)
∞
ω̄β −nω0 −0.5
y(θ, t) = yn0 (6)
n=−∞
2ω̄β
−(S 1 + κ )t
(1)
× einθ−i(ω̄β +nω0 +κI )t e 2 R + c.c. (5) −1.0
which are betatron waves corresponding to beta- −1.0 −0.5 0.0 0.5 1.0
−Re(Δω)0/S1/2
tron sidebands of the revolution harmonics. Here
n > 0 corresponds to fast waves, which are stable Figure 1: Plot of stability contours for various fre-
because κR > 0. For the slow waves with n < 0, quency distributions in the −(Δω)0 /S 12 -plane.
κR < 0 and there is stability only when S 12 > |κR |.
We see that the growth initiated by Z1⊥ is counter- where ℘ denotes principal value, ρ̂(v) = S 12 ρ(ωβ ),
acted by the spread in betatron frequency. v = (ωβ − ω̄β )/S 12 so that v = 1 is at the HWHM.
Note that even when y(θ, t) is damped to
The stability contour is the locus of
zero, the displacements of individual particles are
−(Δω)0 /S 12 as u is varied from −∞ to ∞.
not. In practice, any small initial displacement of
the beam center will be damped immediately if The beam is stable for points lying on the side of
S 12 > |κR |, ensuring that Z1⊥ will stop driving the the locus containing the origin, or Im Δω0 < S 21
individual displacements. In other words, Landau for the Lorentzian distribution. Let us con-
damping nips any instability growth in the bud. sider next the generalized elliptical distribution
H(Ω) = 0 is called the dispersion relation be- ρ̂(v) = (An /an )(1−v 2 /a2n )n when |v|√ ≤ an and
cause it gives frequency Ω as a function of har- zero otherwise, where An=Γ n+32 /[ πΓ(n+1)]
monic number n. Since ±ω̂ are far apart, the de- and the parameter a2n = 1/(1−2−1/n ) is so chosen
nominator can be linearized to give that v = 1 is the HWHM. Note that n needs not
be an integer. For example, n = 0, 12 , 1, 32 , ∞
ρ(ωβ )dωβ
H(Ω) = 1 − iκ =0 (6) correspond, respectively, to the rectangular,
C ωβ − ω̂ elliptical, parabolic, tri-elliptical, and Gaussian
To obtain the stability contour, ω̂ is considered to distributions. In addition, n = 2.36 reproduces
be real, but with a positive infinitesimal imaginary closely the cosine-square distribution. Their
part added. With u = (ω̂ − ω̄β )/S 12 , the dispersion stability contours are plotted in Fig.1.
relation is normalized to the HWHM spread, Notice that all the contours, except the
(Δω)0 Lorentzian, intersect the − Im(Δω)0 /S 12 -axis at
√
1+ f (u) + ig(u) = 0 (7) roughly the same point −1/ 3. A simplified sta-
S 12
bility criterion (Keil-Schnell criterion) can there-
where (Δω)0 = ω − nω0 − ω̄β = −iκ. In fact, fore be represented by a circle centered at the ori-
ieβI0 gin with radius equal to the intercept [6, 7],
(Δω)0 = − β⊥ Z1⊥ (ω̄β +nω0 ) (8)
2E0 T0 1
|(Δω)0 | < √ S 12 F⊥ (10)
is the Z1⊥ -driven complex frequency shift in the √
3
absence of Landau damping (subscript 0) and β⊥ The form factor is F⊥ = 3 for the Lorentzian
is the betatron function at the impedance. f (u) + distribution. For√ the generalized elliptical distri-
ig(u) is called the transfer function and can be bution, F⊥ = 3an /(πAn ), or 1.103, 1, 1.040,
measured (Sec.7.4.17): 1.068, 1.097, and 1.174, respectively, for the rect-

ρ̂(v)dv angular, elliptical, parabolic, tri-elliptical, cosine
f (u) = ℘ , g(u) = π ρ̂(u) (9) square, and Gaussian distributions.
v−u
153
If the frequency spread S 12 is due to a momen- and therefore its revolution frequency. The col-
tum spread Δδ 12 , the simplified stability criterion lective frequency shift, (Δω)0 , is similar to a syn-
becomes chrotron oscillation and is given, in the absence of
4πβE0 ξeff Δδ 12 Landau damping, by [5]
|Z⊥ (nω0 + ω̄β )| < √ F⊥ (11) ienηI0 ω02
3eI0 β⊥ (Δω)20 = (Ω−nω0 )2 = Z (nω0 ) (14)
2πβ 2 E0 0
where ξeff = ξ−η(n+νβ ) with ξ the chromaticity, Landau damping arises from a spread in the rev-
νβ the betatron tune, and η the slip factor. For a olution frequency. For a distribution ρ(ω0 ) with
broad-band impedance rolling off at frequency ωc , mean ω̄0 , the dispersion
relation is [6]
the substitution n = ωc /ω0 can be made. ρ(ω0 )dω0
1 − (Δω)0 2
=0 (15)
Transverse oscillation of a single bunch When (Ω − nω0 )2
C
the bunch is very much shorter than the wave- where C goes below the pole at Ω/n. The above,
length of the perturbing Z1⊥ , the bunch can be including the correct path of integration C can
viewed as a single macro-particle [3], oscillating be derived by solving the Vlasov equation as an
transversely with frequency Ω. The dispersion re- initial-value problem. At the onset of instabil-
lation is the same as Eq.(6), but with ω̂ = Ω. The ity, integration by parts and normalization to the
simplified stability criterion is still Eq.(10). The HWHM revolution frequency spread S 12 result in

complex mode frequency shift due to wake effect (Δω)2 ρ̂ (v)dv
in the absence of Landau damping is now 1 − 2 20 =0 (16)
n S1 v − u − i
e2 NB β⊥ W
2
(Δω)0 = (Ω− ω̄β )0 = (12) where → 0+ , v = (ω0 − ω̄0 )/S 12 , nu = (Ω −

2β 2 E0 T0 nω̄0 )/S 12 , and ρ̂(v) = S 12 ρ(ω0 ) so that v = 1 is
∞
the HWHM. If the spread in ω0 comes from the
W= W1 (−kC)eiω̄β kT0 spread ΔE, then S 12 = (|η|ω̄0 /β 2 )(ΔE/E0 ) 12 .
k=1 The dispersion relation becomes, in the U -V no-
∞
iβ ⊥ tation,
=− Z (pω0 + ω̄β ) (13) 2i ρ̂ (v)dv
T0 p=−∞ 1 1 − sgn(η)(U + iV ) = 0 (17)
π v − u − i
with NB the number of particles in the bunch.
eI0 β 2 Z0 (nω̄0 )/n
When the length of the bunch is taken into U + iV =
4|η|E0 (ΔE/E0 )21
account, there will be many more modes of trans- 2
verse oscillation along the bunch. Head-tail insta- The stability contour can be traced by varying u
bilities can happen (Sec.2.4.9) and can be Landau from −∞ to ∞, and is depicted in Fig.2 for vari-
damped when the growth rate < ous distributions when η < 0.
∼ S 12 .
Stronger impedance may cause two modes Except for the Lorentzian distribution, all
to merge, resulting in transverse mode-coupling contours intersect the V -axis at roughly Vin = −1.
instability (Sec.2.4.9). The growth is fast once A simplified stability criterion will therefore be
above threshold and Landau damping usually the approximation of the contours as circles pass-
does not help. ing through( Vin(, or [6]
( Z ( 4|η|E ΔE 2
When the bunch is very much longer than the ( 0( 0
( (< F (18)
wavelength of Z1⊥ and the growth rate is much ( n ( eβ 2 I0 E0 1
2
faster than the synchrotron frequency, locally the For the Lorentzian distribution, the form factor is
bunch can be viewed as unbunched. Therefore F = π/2. For the generalized elliptical distribu-
the stability criterion of Eq.(10) applies [7]. How-
tion F = πa2n /(4n+2), which amounts to 1.047,
ever, we must replace I0 by the local peak current
1.047, 1.061, 1.080, 1.097, and 1.133, respec-
and Δδ 12 by the local momentum spread (Bous-
tively, for the elliptical, parabolic, tri-elliptical,
sard criterion). cosine square, (1 − v 2 /a24 )4 , and Gaussian dis-
Longitudinal oscillation of unbunched beam tributions. The simplified stability criterion for
This case is unique in that there is no external the Gaussian distribution can also be written in

focusing frequency. However, the longitudinal the close form as |Z0 /n| < 2π|η|E0 σE2 /(eβ 2 I0 ),

impedance Z0 does alter the the particle’s energy where σE is the fractional rms energy spread.
154
Ch.2: BEAM DYNAMICS
10.0
When the bunch is much longer than the
From inside out (4)Cosine−sq.
(1)Elliptical
2 2 4
(5)(1−v /a4 )
wavelength of Z0 and the synchrotron oscillation
7.5 (2)Parabolic (6)Gaussian period is much longer than the instability growth
(3)Tri−elliptical (7)Lorentzian time, the beam can be viewed as unbunched lo-
2
V=Re(π/2)[(Δω)0/nS1/2]
cally. Thus dispersion relation (17) and stability

5.0 condition (18) apply [7]. However, the average
current I0 must be replaced by the local peak cur-
rent and the energy spread ΔE by the local energy
2.5
spread (Boussard criterion).
References
0.0
[1] L.D. Landau, J. Phys. USSR 10 (1946) 25

−2.5 [2] A. Hofmann, CERN 89-01 (1989)

−6 −4 −2 0 2 2
4 6
U=−Im(π/2)[(Δω)0/nS1/2] [3] A.W. Chao, Physics of Collective Beam Instabili-
ties in High Energy Accelerators, Wiley (1993)
Figure 2: Stability contours for longitudinal un- [4] K.Y. Ng, Physics of Intensity Dependent Beam In-
bunched beam in the π(Δω)20 /(2n2 S 21 )-plane. stabilities, World Scientific (2006)
2
[5] K.Y. Ng, AIP Conf. Proc. 184, Vol.2, p.1969; Fer-
milab Report FN-0438 (1986)
Longitudinal oscillation of a single bunch [6] E. Keil, W. Schnell, CERN/TH-RF/69-48 (1969);
When the bunch is very much shorter than the V.K. Neil, A.M. Sessler, RSI 36 (1965) 429

wavelength of Z0 , it can be approximated by a [7] D. Boussard, CERN Lab II/RF/Int 75-2 (1975)
macro-particle [3]. In the presence of the syn-
chrotron frequency ωs , the problem is similar to
that of the transverse. As a result, the dispersion 2.4.12 Intrabeam Scattering and Touschek
relation is still Eq.(6) but with ω̂ = Ω. (Δω)0 is Effect
the shift from the mean synchrotron frequency ω̄s V. Lebedev, Fermilab
in the absence of Landau damping (potential-well
distortion included, Sec.2.4.8), Coulomb scattering of charged particles in a beam
e2 NB ηcW results in an exchange of energy between different
(Δω)0 = (Ω − ω̄s )0 = (19) degrees of freedom. The total cross section of the
2βE0 T0 ω̄s two-particle scattering in vacuum diverges; how-
∞
ever it has a finite value for collisions in plasma
W= eiω̄s kT0 W0 (−kC) (or beam) due to field screening by other parti-
k=1 cles [1]. Usually two scattering regimes are con-
∞
i
sidered: the single scattering when a rare single
= (pω0 + ω̄s )Z0 (pω0 + ω̄s ) (20) collision makes large change of particle momen-
C p=−∞
tum (Touschek effect), and the multiple scattering
W0 being the longitudinal wake function and W0 when multiple frequent collisions cause a diffu-
its derivative. Stability contours is given by Fig.1, sion. The former is usually responsible for cre-
and Eq.(10) remains a simplified stability crite- ation of distribution tails and beam loss in elec-
rion with S 21 being the HWHM spread in ωs . tron machines, while the latter for changes in the
For finite bunch length, there will be many distribution core. Although such approach is use-
longitudinal modes of oscillation. Longitudinal ful in many applications there are cases when it
head-tail instability can occur (Sec.2.4.9) but can fails to deliver an accurate result.
be Landau damped by the spread in ωs . Multiple scattering in single component

When Z0 is strong enough, two modes of plasma The temperature exchange in plasma
oscillation can merge and longitudinal mode- is driven by interplay of friction and diffusion
coupling instability occurs (Sec.2.4.9). Once [2]. In uniform nonrelativistic one-component
above threshold, the growth is fast and Landau plasma with Gaussian distribution for each
damping usually does not help. degree of freedom but different rms velocities
155

σvk ≡ vk , k = (x, y, z) it results in [3],
2 longitudinal temperature is much smaller than the
transverse one and its growth
rate in the course of
⎛ 2 ⎞
σ beam transport is [6] (σvz v 2 )
d ⎝ vx2 ⎠ (2π)3/2 nr02 c4 Lc
σvy = dTz dσ 2 (2π)3/2 nmr02 c4 Lc
dt 2
σvz 2 + σ2 + σ2
σvx vy vz
≡ m vz = (5)
dt dt v2
⎛ ⎞
Ψ (σvx , σvy , σvz ) The temperature relaxation is suppressed for a
×⎝Ψ (σvy , σvz , σvx )⎠ (1) beam moving along strong solenoidal magnetic
Ψ (σvz , σvx , σvy ) field [7].
where r0 is the particle classical radius, n is Note that Eq.(1) is not self-consistent; i.e. it
the plasma density, Lc = ln(ρmax /ρmin ) is the implies that the initial distribution remains Gaus-
Coulomb logarithm, sian; but an interplay of the diffusion and friction

due to multiple scattering does not allow for the
ρmin = r0 c2 /v 2 , ρmax = v 2 /4πnr0 c2 (2) beam distribution to remain Gaussian during the
process. Additionally the single scattering creates
where v 2 = σvx2 + σ 2 + σ 2 . Function Ψ (x, y, z) non-Gaussian tails. However deviations from the
vy vz
can be expressed through the symmetric elliptic Gaussian distribution are comparatively small and
integral of the second kind, RD (x, y, z), Eq.(1) represents a good approximation in most
√ practical cases.
2r 2
Ψ(x, y, z) = y RD (z 2 , x2 , y 2 ) (3) Multiple scattering in a storage ring Al-
3π though the beam focusing and finite dispersion
in circular accelerators significantly complicate
+ z 2 RD (x2 , y 2 , z 2 ) − 2x2 RD (y 2 , z 2 , x2 )
equations they do not change the essence of the
∞ process. However the time dependent nature of
3 dt
RD (x, y, z) = transverse focusing results in an absence of ther-
2
0 (t+x) (t+y) (t+z)3 mal equilibrium and a continuous growth of 3-D
emittance. In the absence of x-y coupling for a
where r = x2 + y 2 + z 2 ; x, y, z ≥ 0. An algo- bunched beam nonrelativistic in the beam frame
rithm for fast numerical evaluation of RD (x, y, z) (BF) the emittance growth rates can be expressed
is discussed in [4]. Ψ (x, y, z) is chosen so in a compact matrix form
that it depends on the ratios of its variables
3
but not on r. It is symmetric relative to the √ 3 Lc Bkij Rij 4
2 2
variables y and z, and is normalized so that dεk N r0 c / εx εy i,j=1
Ψ (0, 1, 1) = 1. The energy conservation yields = √ (6)
dt 4 2σz β 2 γ 4 βx βy FD tr(Σ) s
Ψ(1, 0, 1) = Ψ(1, 1, 0) = −1/2 and Ψ(x, y, z) +
Ψ(y, z, x) + Ψ(z, x, y) = 0. In thermal equilib- coinciding with the results of [4, 8, 9] for the
rium Ψ (1, 1, 1) = 0. If the first parameter is zero case of zero vertical dispersion. Note that the
it can be approximated to ∼0.5% accuracy by derivatives of Courant-Snyder parameters omitted
√ 2 2 2 22 in [10] can significantly change results if optics
2 x +y x −y has large β-function variations. Here . . .s de-
Ψ(0, x, y) 1+ ln −0.055 2 2
π 2xy x +y notes averaging over the machine circumference,
β and γ are the relativistic factors; N is the num-
The asymptotics are
⎧ ber of particles per bunch; Σ ≡ [vi vj ] is the ma-
⎨√1 ln 28r2 2 − 3z , x, y z trix of the second moments of local velocity dis-
Ψ(x, y, z) = √2 y2 +z 2
2π x +y r
tribution in the BF,
⎩
π ln 8r2 + 3x r , y, z x Σ = (γβc)2 Gt Ξ−1 G (7)
(4)
Ξ is the bilinear form of particle angles
Eq.(1) is obtained in the logarithmic approx-
(θx , θy , Δp/p),
imation and is justified if ln(ρmax /ρmin ) 1 ⎛ ⎞
where the plasma perturbation theory can be used βx /εx 0 −βx Φx /εx
[5]. It describes the temperature relaxation in a Ξ=⎝ 0 βy /εy −βy Φy /εy ⎠
wide class of electron and ion beams. Particularly, −βx Φx /εx −βy Φy /εy Ξ33
in the case of electrostatic acceleration the beam (8)
156
Ch.2: BEAM DYNAMICS
Ξ33 = 1/σp2 + Ax /εx + Ay /εy dσp2 /dt. Note that the factor of 2 in Bz reflects
the absence of the synchrotron motion, taken into
αx Dx αy Dy
Φx = Dx + , Φy = Dy + account in Eq.(6).
βx βy In many practical applications the longitudi-
nal temperature in the BF is much smaller than
Dx2 + (βx Φx )2 Dy2 + (βy Φy )2 the transverse one (θ /γ θ⊥ ) and the verti-
Ax = , Ay =
βx βy cal dispersion can be neglected. Eq.(6) (bunched
beam) then can be reduced to
FD = 1 + Dx2 σp2 / (εx βx ) + Dy2 σp2 / (εy βy ) ⎛ ⎞
ε
σz is the rms bunch length; βx , βy , αx , and αy d ⎝ x⎠ N r2c
εy = √ 0 (10)
are Courant-Snyder parameters; Dx , Dy , Dx and dt σ 2 4 2β 3 γ 3 σz
Dy are dispersions and their derivatives; εx , εy p

⎛ ⎞4
3
and εz = σz σp are the unnormalized transverse
Lc ⎝ Ψx (θx , θy ) ⎠
and longitudinal rms beam emittances; σp is the × Ψy (θx , θy )

σx σy θ⊥ Ψ (0, θ , θ ) s
relative rms momentum spread; matrices Bk and x y
G are
⎛ ⎞ where
βx 0 −Φx βx # $
σp2 (Φx βx )2
Bx = ⎝ 0 0 0 ⎠ εx
θ⊥ = θx2 + θy2 , θx2 = 1+
−Φx βx 0 Ax βx σx2
⎛ ⎞
0 0 0 εy βy
B = ⎝0 βy −Φy βy ⎠
y θy2 =
βy
, Ψy (θy , θx ) = Ψ (θy , θx , 0) 2
γ
0 −Φy βy Ay
βx
Bz = diag (0, 0, βz ) , G = diag (1, 1, 1/γ) Ψx (θx , θy ) = Ψ (0, θx , θy ) Ax + Ψ (θx , θy , 0)
γ2
the function diag(..) makes a diagonal matrix
For an ultrarelativistic machine, γ Qx , the
from a vector; βz = σz /σp is the longitudinal β-
terms with βx,y /γ 2 are small and can be ne-
function;
t glected.
R = G−1 TΨIBS Tt ΣT Tt G−1 (9) Similar to the suppression of IBS in an elec-
a rotational matrix T reduces Σ to its diago- tron beam by magnetic field the multiple IBS is
nal form σ = Tt ΣT; and the matrix function suppressed if σp becomes so small that the col-
ΨIBS (σ) with its argument of a diagonal matrix lision time becomes comparable to the betatron
σ is frequency [11].
√ √ √ Touschek effect Single scatterings become im-
ΨIBS (σ) = diag (Ψ( σ11 , σ22 , σ33 ),
√ √ √ √ √ √ portant when there is a large difference between
Ψ( σ22 , σ33 , σ11 ), Ψ( σ33 , σ11 , σ22 )) rms velocities of different degrees of freedom in
the BF. In this case a single scattering can result
The Coulomb logarithm is computed similarly to
in a momentum transfer significantly exceeding
the plasma case with the following correction af-
the rms value of the coldest degree of freedom
fecting the value of maximum impact parameter
thus creating non-Gaussian tails and particle loss.
of Eq.(2),
For the case when the longitudinal temperature is
ρmax = min σmin , γσz , tr (Σ) /4πnr0 c2 much smaller than the transverse ones it is called
the Touschek effect. It was first investigated in
where electron storage rings for a flat beam and nonrel-
2 ativistic energies in the BF [12, 13]. The calcu-
2
2σmin = σx2 + σy2 − σx2 −σy2 + 4Dx2 Dy2 σp4 lations were extended to arbitrary energies [14],
round beams [15] and the dispersion was included
σx2 = εx βx +Dx2 σp2 , σy2 = εy βy + Dy2 σp2
[16]. A general case for an arbitrary ratio of trans-
The above equations can be used for a√coasting verse emittances and arbitrary energies was con-
beam with substitutions σz → L/ (2 π) and sidered in [17]. The lifetime T of the particles
Bz = diag (0, 0, 2) implying that dε3 /dt → having Gaussian distributions for all degrees of
157
freedom is [17] assumed in the Touschek lifetime, as well as in

1 N r2c the IBS growth rates of Eq.(6) implying a linear
= √ 2 40 × rf waveform, i.e. a short bunch. However it is
T 8 πβ γ εx εy σz σp usually not the case in hadron colliders where the
3 ∞ bunch fills most of the longitudinal acceptance.
1 1 2 u/um Initially the beam is well inside of the rf bucket
√ 2+ −1
Ξ33 u 1+u and the beam loss is dominated by single scatter-
um
ings. Later, the multiple scatterings result in the
beam distribution reaching the rf bucket bound-
1+u 1 1 u/um
+1 − − 4+ ln ary leading to domination of multiple scattering
u/um 2u u 1+u in the particle loss. For the pancake velocity dis-
√ !
tribution in the BF σp /γ θ⊥ (which is always

udu
e−B1 u I0 (B2 u) √ (11) the case for colliders) the evolution of longitudi-
1+u s nal distribution and the particle loss from the rf

where um = β 2 (Δpm /p)2 , Δpm is the maxi- bucket due to multiple scattering can be described
mum stable momentum deviation and we assume by a diffusion equation

it does not depend on a betatron motion ampli- ∂f 1 ∂ D(I) ∂f
tude, = I (13)
∂t 2 ∂I ω(I) ∂I
# $
1 βx βx Φ2x βy βy Φ2y where I and ω(I) are the action and frequency of
B1 = 2 2 1− + 1− synchrotron motion. For a single harmonic rf the
2β γ εx Ξ33 εx εy Ξ33 εy
dimensionless Hamiltonian and the diffusion are

Dy2
p̂2 ϕ 2
βx βy 1 Dx2 H= + 2 sin (14)
B2 = B12 − 4 4 + + 2 2
β γ εx εy Ξ33 σp2 εx βx εy βy "
ÃLc
and I0 (x) is the modified Bessel function. For D(I) = p̂2 n(ϕ)dψ
Dy = 0 (planar orbit) B1,2 simplify to πIω(I)
( ( 3 4
( ( √ π 3 N r02 βz2 fRF 3
Ψ(0, θx , θy )
1 (( βx βy (( Ã = 2 2π
B1,2 = 2 2 ( ± ( β 6 γ 3 c2
2β γ ( ε 1 + (σ β Φ /σ )2 εy ( σx σy θx2 +θy2 s
x p x x x

where n (ϕ) = f (I(ϕ, p̂))dp̂ is the linear den-
For flat beams εx /βx εy /βy and non-
sity normalized so π that at the process beginning
relativistic energies in the BF (γ 2 εx /βx 1) the
2π f (I)dI = −π n(ϕ)dϕ = 1, ψ is the mo-
Bessel function can be expanded for large argu-
ments and one obtains tion phase (dψ = ω(I)dt), and fRF is the rf fre-
3 3
4 quency. An integration of Eq.(13) for the linear rf
1 N r02 c (βx /εx ) 2 C (εm ) and Gaussian distribution yields the momentum
=
T 8πβ 3 γ 5 σz 2 32 growth rate of Eq.(10). To take into account the
σ β Φ
1+ p σxx x σx σy εm s single scattering one needs to replace Eq.(13) by
integro-differential equation [18]
(12)
∞
with ∂f (I, t)
∞ = Ã W (I, I ) f (I , t) − f (I, t) dI
3 3x x u du ∂t
C (x) = − e−x + 1+ + ln e−u 0
2 2 2 x u (15)
ε The kernel is a symmetric function, W (I, I ) =
Δp2m βx W (I , I), which for I ≥ I can be expressed as
εm =
p2 γ 2 ε 1 + (σ β Φ /σ )2 ωω
x p x x x W (I, I ) =
π (H(I) − H(I ))3
IBS in hadron colliders The radiation damping ⎡ ⎤

a(I)
a(I)
in electron synchrotrons typically results in the ⎢ dϕ ⎥
rms momentum spread being much smaller than ×⎣ H(I)−H(I ) n(ϕ) + 2 n(ϕ)p̂dϕ⎦
p̂
the rf bucket size, σp Δpm /p. It has been 0 0
158
Ch.2: BEAM DYNAMICS
where a(I) is the motion amplitude. This equa- [19] M. Chanel et al, PL B192, n.3,4, p.475 (1987)
tion has to be solved numerically together with the [20] V. Lebedev et al, Linac 2010; A. Shishlo et al,
equations describing the evolution of transverse PRL (2012)
emittances.
Intrabeam stripping There is another mecha- 2.4.13 Ion Trapping, Beam-Ion Instabilities,
nism of particle loss related to the intrabeam col- and Dust
lisions of H− ions. It results in stripping one of F. Zimmermann, CERN
two H− ions in a binary collision. It was first ob-
served in the CERN antiproton ring LEAR oper- Positively charged ions trapped in the beam po-
ating with H− ions during its commissioning [19]. tential affect performance of electron and antipro-
It was recently found that the intrabeam stripping ton storage rings. Size of the trapped ions ranges
is a major mechanism of particle loss in the su- from single atoms, small molecules, to micron-
perconducting part of the SNS linac making al- size dust particles. Effects of trapped ions on
most uniform relative loss of ∼10−6 m−1 [20]. the beam include: increased emittance, betatron
The particle loss cross section has a flat top of tune shifts and broadening, collective instabilities,
∼4 × 10−15 cm2 in the relative velocity range of reduced beam lifetime. In future storage rings
β ⊂ [2× 10−4 , 5× 10−3 ]. It can be approximated single-pass effects are important, i.e. the ions do
by a fitting formula uniting analytic, numeric and not have to be “trapped” to be harmful.
experimental data [20], Symbols: C storage ring circumference, f0
2 revolution frequency, Lsep bunch spacing (in me-
240a2∞ (β −βm )6 αF β +αF ter), beam distribution assumed bi-Gaussian with
σ≈ ln 1.8
(β −βm )6+βm 6 β +αF αF rms σx,y , x,y unnormalized emittances, ltrain =
nB Lsep bunch-train length, b beam-pipe radius, I
where a∞ is the Bohr radius, αF is the fine struc- average beam current, βx,y average β-functions,
ture constant, and β ≥ βm ≈ 7.5 × 10−5 . λion ion line density, σion ionization cross-section,
Σpipe vacuum pipe cross-sectional area, Nion to-
References tal number of trapped ions (Ne total number of
trapped electrons), N = NB nB total number of
[1] L.P. Pitaevskii, E.M. Lifshitz, Physical Kinetics,
v.10, Pergamon (1981)
particles in beam, Q ion charge in units of elec-
[2] L.D. Landau, JETP, 7, 203 (1937); Phys. Zs. tron charge e (in general Q = 1 for regular ions,
Sowjet, 10, 154 (1936) Q 1 for dust particles), A ion mass in units
[3] S. Ichimaru, M.N. Rosenbluth, Phys. Fluids, 13, of proton mass mp , Aatom and Zatom the atomic
2778 (1970) mass and atomic number of the dust particle con-
[4] S. Nagaitsev, PRST-AB 8, 064403 (2005) stituents, fi = ωi /2π ion oscillation frequency
[5] S. Ichimaru, Basic principles of plasma physics in the beam potential, τth thermal lifetime of dust
(1973) particle, Tmp melting point, Teq equilibrium tem-
[6] H. Boersch, Z. Phys. 139, 115 (1954) perature of trapped dust particle, r0 classical ra-
[7] V. Kudelainen et al, JETP, 56, n.6, 1191 (1982) dius of the beam particles (in most cases r0 = re
[8] J.D. Bjorken, S.K. Mtingwa, PA 13, p.115 (1983) or rp ), dgas = pgas /(kB T ) the residual gas den-
[9] M. Conte, M. Martini, PA 17, 1 (1985) sity, pgas the gas pressure. More symbols are de-
[10] A. Piwinski, Proc. 9th Int. Conf. on High Energy fined in Tabs.1 and 2.
Accelerators, Stanford (1974) p.405
[11] V. Parkhomchuk, Proc. Workshop on Electron 2.4.13.1 Ion trapping
Cooling and Related Applications (1984)
[12] C. Bernardini et al, PRL 10, 407 (1963) Average beam potential for bi-Gaussian distribu-
[13] J. Haissinski, LAL Orsay report 41-63 (1963) tion (round pipe, elliptical beam) is [1]
[14] U. Voelkel, DESY 67/5 (1965) ∞
eNB
[15] Y. Miyahara, Jap. Journal of Appl. Phys., 24, n.9, U (x, y) ≈ − dq
p.1742 (1985) 4π0 Lsep 0
[16] J. Le Duff, CERN Acc. School (1987) x2 y2 b2
− 2 +q − 2σ 2 +q −
[17] A. Piwinski, DESY-98-179 (1998) e 2σx y −e 2 +q
2σx
[18] V. Levedev, HB-2004 (AIP Conf. Proc., v.773, ×
2005) p.440 2σx2 + q 2σy2 + q
159
Table 1: Selected properties and effects of trapped ions.
property/effect formula
QNB rp Lsep
trapping condition (no gap) [4] A ≥ Acrit ≡ 2σy (σx +σy ) or πLsep fi ≤ c
trapping with gap Tg [1] | cos(ωi ltrain /c) − ωi Tg sin(ωi ltrain /c)/2| < 1
1/2
c 2QNB rp
single-ion oscillation frequency [4] fi = 2π ALsep σy (σx +σy )
equil. ion distr. (no gap) [5] ρy = √1 K (ξ)e−ξ

π 2πσy 0
with ξ ≡ [y/(2σy )]2
coherent tune shift Δνx,y;coh ≈ βx,y r0 Qλion C/[4πγσx,y (σx + σy )]

incoherent tune shift Δνx,y;inc ≈ 2Δνx,y;coh
inc. phase shift along train (1 pass) Δψx,y (s) ≈ βx,y rp Qλion s/[γσx,y (σx + σy )]
Table 2: Instability growth

(multi-turn) ion-related instabilities; n closest integer, τion =
rates for classical
2λ r R2 r βy λion R
1/(σion cdgas ), νp ≡ γp σy e(σpx +σy ) , νe ≡ πσyN(σrxe+σ
R
y)
, Δνy ≡ γσby (σx +σy )
, η desorption yield, and sp linear
pumping speed (in m2 s−1 ).
instability growth rate 1/τ
proton-ion oscillations [9] (n − νx,y )A/(8νx,y γτion )

νe
proton-electron (or antiproton-ion) inst. w/o freq. spread [11] πf0 νp n−ν e
electron-ion & proton-electron inst. w freq. spread [10] π 2 f0 Δνy /(Δfi /fi )FWHM
pressure bump [14, 15] (ηIσion /e − sp )/Σpipe
⎧
⎪
⎪
2 2
constant + σx (σxx +σy ) + σy (σyx +σy ) Techniques for ion removal include (i) clear-
⎪
⎪
⎪
⎪ ing gap; (ii) beam shaking (rf knock-out) [6, 7, 8];
⎪
⎪ σx σy
− (σx +σ x2 y 2 x4 2σx
⎪
⎪ 2 2σx2 σy2 + 12σx4 1 + σy (iii) clearing electrodes.
⎪
⎪ y)

⎪
⎪ y4 σ
⎪
⎨ + 12σ 4 1 + 2 σyx
IZ0 y 2.4.13.2 Dust particles
≈ for x,y ≤ σ x , σy
4π ⎪
⎪ √
⎪
⎪ constant + x2
+ 2 σ|y|y Capture of positively charged dust particles has
⎪
⎪ σ (σ +σ )
⎪
⎪
x x y been observed at many electron and antiproton
⎪
⎪ for σy y σx , x σx storage rings [16, 17, 18, 19, 20]. To remove or
⎪
⎪
⎪
⎪ study dust, specific devices have been built, which
⎪ ln x2 +y2
⎩ for x, y σ , σ
b2 x y shake the vacuum chamber [21] or create an elec-
(1) tric discharge [22]. Dust trapping in an electron
where Z0 = 377 Ω. ring has been directly recorded by video camera,
Refs.[1, 2, 3, 4] give overview and discussion showing a luminous body resembling a shooting
of ion stability, accumulation, removal and effects star [22]. Falling or negatively charged dust par-
on beam; Ref.[1] discusses longitudinal ion mo- ticles can also lead to temporary high local losses
tion; Refs.[6, 9, 11, 12, 13] present theories and from proton beams [23].
observations of conventional (i.e. multi-turn) co- The estimated equilibrium temperature Teq of
herent beam-ion/electron instabilities. Tabs.1 and a trapped dust particle allows predicting whether
2 give a summary of these effects. it explodes after being trapped by the beam, or,
160
Ch.2: BEAM DYNAMICS
in case it survives (when Teq < Tmp ), what its beam-lifetime reduction that such a particle may
thermal lifetime will be. Teq is determined by a induce, are determined by its equilibrium charge.
balance of ionization energy deposition and heat The equilibrium charge is calculated by equating
radiation, i.e. the ionization rate
( (
dE (( ! dE (( 4π Zatom re R4
= ⇒ Teq (2) Q̇ioniz ≈ f0 N NA re (6)
dt ( dt (
ion rad 3 Aatom σx σy Q
Very roughly, for a spherical particle of radius R and the field evaporation of ions [25]
(R is not the ring radius here!),
( ( Aatom mp 8π 2 kB2 T2
ΔE (( 2R3 NB nB f0 ρ dE (( Q̇ev ≈ − 3
≈ (3) h √
Δt (ion 15 σx σy d(ρx) (min U + V − Φ− e2 Q
× exp − +
where ρ = mass density, dE/d(ρx)|min ≈ 1.5 kB T 4π0 RkB T

MeV cm2 g−1 , and T T

1 dT
( ∞ − dT C (T ) (7)
ΔE ((
p
R2 ω 3 /(πc2 ) kB 0 T 2 0
( ≈ − dω ω Q̄(ω, R)
Δt rad 0 e kB T − 1 where Cp is the heat capacity at constant pres-
(4) sure, NA is Avogadro’s number, U and V ion-
Absorption coefficient Q̄(ω, R) for small parti- ization and vaporization energy, respectively, and
cles can be found according to Mie theory [19, Φ− workfunction. The trapping of photoelectrons
24]. Q̄ depends on the material and also on shape is another discharging mechanism.
of the dust particle. It is calculated from the com- Finally, there are two effects by which a
plex dielectric function, e.g., using program BH- trapped dust particle can reduce the beam lifetime:
MIE [24]. If Teq > Tmp , particle explodes. If (i) Bremsstrahlung on nuclei field [26]
Teq < Tmp , the thermal lifetime due to evapora- # $
tion is 1 2
cAZatom 8re2 α ΔE 183
≈ − ln ln
1 Ȧ 4πAatom τbn Aatom σx σy C 3π E 1/3
Zatom
≡ ≈ −
τth A 2πmp Aatom kB
where ΔE/E is energy acceptance;
3mp 2/3 p(Teq (A)) [Pa] (ii) Bremsstrahlung in collective field of dust
× (5)
4πρ Teq (A) A1/3 particle [19]
For needle-shaped particle of transverse ra-
where p(T ) is the vapor pressure as a function of dius R and length h, aligned in beam direction,
temperature. Eqs.(2-5) are solved numerically. Il- rmax
lustrative example shown in Fig.1 [19]. 1 c
≈ π Nγ (r)r dr (8)
The dynamic stability of a trapped dust par- τbcoll 0 2πσx σy C
ticle, as well as the betatron tune shift and the 5 cαΥ(r) max(r, h)
Nγ (r) ≈ √
2 3 λe γ [1 + Υ(r)2/3 ]1/2
1020
1016
2γQre min(r/h, 1)
SiO2 Needle Υ(r) ≈
1012
c max(R, r)2
τth (s)
10 8
SiO2 Sphere where rmax is defined by Υ(rmax ) = 0.02 [27].
10 4
In most cases, this second effect is not important.
Ti Needle
10 0 2.4.13.3 Single-pass ion effects in storage
10–4 rings and linacs
10–8
108 1010 1012 1014 1016
Nonlinear coupling effect due to ions can cause
3–95 A 7900A1
vertical emittance growth. It is contained if beta-
tron phases have |μx −μy |/|μx | > 5% [28]. How-
Figure 1: Thermal lifetime τth of trapped dust particles ever, resonant single-bunch tail excitation can still
in the HERA electron ring as a function of mass. occur (Fig.2). An initial perturbation y0 (z, s) =
161
Table 3: Rise-time estimates for fast beam-ion instability and equilibrium with feedback and noise; ω̃i ≡
1/2
4N r
c 3ALsep σyB(σpx +σy ) the coherent angular ion frequency (Q = 1 assumed), (Δω̃i )rms its rms variation around

the ring circumference, and η̄(z) ≡ z τd /τc /(nB Lsep ), where z denotes the longitudinal position along the
bunch train (z = 0 at the start of the train), τd is the (feedback) damping time, and y 2 (0) the rms oscillation
amplitude of the first bunch in the train.
approximation asymptotic behavior characteristic rate 1/τ

1
3/2 1/2 1/2
4dgas σion βy NB n2B re rp Lsep c
linear treatment (y σy ) [30] y ∼ exp( t/τc ) ≡ √ 3/2
τc 3 3 γσy (σx + σy )3/2 A1/2
1 1 c
with decoherence & spread Δfi [32] y ∼ exp(t/τe ) ≈ √
τe τc 2 2 ltrain (Δω̃i )rms
1 1 c
nonlinear regime (y σy ) [32] y ∼ t/τH ≈
τH τc ω̃i Lsep n3/2
B

equilibrium state with y 2 (z)
≈ τd /τc y 2 (0)η̄ 3 /(24 nB ) (for η̄ 1)
2 /2
feedback and noise [33] y (z) ≈ τd /τc y (0)e
2 2 η̄ /(4 η̄nB ) (for η̄ 1)
Y Tail Head Y
100
ions 80
60
90
Z Z 40
10–2
Jy (εy)
ions
Head Tail
6-94 7709A1
10–4 1
20
Figure 2: Schematic picture of resonant tail excitation
for two different betatron phases; beam moves to the 10–6
;;;;;;;;;
left. 0 1000 2000 3000 4000
3–95 s (meter) 7906A1
;;;;;;;;;
zo Figure 4: Simulated growth of the action of the ver-
+ + +
+
+ + e– + + + + + + + tical centroid for every twentieth bunch in the NLC
;;;;;;;;;
+ + +
e–
e–
+
+ + +
e–
+ ++ + + + Damping Ring for a vacuum of 10−8 Torr of CO gas;
e– +
+
e–
+ growth saturates at ∼ σy due to nonlinearity of beam-
;;;;;;
+
z ion force.
s
;;;;;;
Electron Bunch train Ions ŷz/σz cos(s/βy + φ0 ), e.g., due to wakefields,
zo
leads to emittance growth after distance L [29],
– –
e+
–
–
– –
– – – e+ 2
e+ e+ 1 ŷLβy λion re
z Δy (L) ≈ (9)
2πβy 2γσy (σx + σy )
Atomic
4-95
7947A1 Positron Bunch Electrons
where λion is the ion density at end of bunch.
Figures 3 and 4 illustrate the fast beam-ion
Figure 3: Schematic of fast beam-ion instability which instability mechanism. The theory and simula-
can arise due to ion trapping in electron bunch train tions, are discussed in [30]-[35] and summarized
or due to trapping of free electrons in single positron in Tab.3. (Possible) experimental evidence is dis-
bunch. cussed in [36]-[38].
162
Ch.2: BEAM DYNAMICS
References [37] J. Huang et al, PRL 81 (1998) 4388

[38] Y. Ohnishi et al, EPAC 00, p.1167
[1] D. Villevald, S. Heifets, PEP-II AP-18-93 (1993)
[2] A. Poncet, Lecture Notes in Phys. 425, Springer
Verlag (1994) p.202 2.4.14 Electron-Cloud Effect
[3] Y. Baconnier, A. Poncet, P. Tavares, CERN/PS M.A. Furman, LBNL & Cornell U.
94-40 (1994)
[4] Y. Baconnier, G. Brianti, CERN/SPS/80-2 The electron-cloud effect (ECE) was first iden-
(1980) tified at the Photon Factory (PF) at KEK [1] as
[5] T.O. Raubenheimer, F.-J. Decker, J.T. Seeman, a fast transverse coupled-bunch instability that
PAC 95, p.3291 arose only when PF was operated with a positron
[6] R. Alves Pires et al, PAC 89, p.800 e+ beam. Unlike the ion-induced instability
[7] J.C. Lee et al, PAC 95, p.2968 (Sec.2.4.13) observed when operated with an
[8] S.J. Werkema, PAC 95, p.3397 electron beam, the e+ beam instability persisted
[9] H.G. Hereward, CERN MPS/Int. DL 64-8 (1964) even with a substantial gap in the bunch train. The
[10] H.G. Hereward, CERN 71-15 (1971) coupled-bunch mode spectrum for the e+ beam
[11] E. Keil, B. Zotter, CERN-ISR-TH/71-58 (1971)
was qualitatively different from that for an e−
[12] G. Koshkarev, P. Zenkevich, PA 3 (1972) 1
beam under otherwise similar conditions. The
[13] L.J. Laslett, A.M. Sessler, D. Möhl, NIM A121
phenomenon disappeared when the bunch spac-
(1974) 517
[14] E. Fischer, K. Zankel, CERN/ISR-VA/73-
ing was sufficiently large, and it could not be at-
52(1973) tributed to known machine impedances. The am-
[15] O. Gröbner, R.S. Calder, PAC 73, p.760 plitude of the unstable motion reached saturation
[16] H. Saeki et al, RSI 62, No.4, 874 and No.11 and was accompanied by the excitation of vertical
(1991) 2558 coupled-bunch oscillations, and possibly of verti-
[17] D. Sagan. NIM A330 (1993) 371 cal emittance growth.
[18] E. Jones et al, PAC 85, p.2218 Experimental analysis [1], simulations [2]
[19] F. Zimmermann et al, PAC 95, p.517 and analytical work [3] showed that the cause of
[20] D.R.C. Kelly et al, PAC 95, p.2017 the instability is an electron cloud (EC) that de-
[21] H. Frischholz, T. Wikberg, CM-P00072614, veloped inside the vacuum chamber generated by
CERN/ISR-VA/TW/ck (1971) photoelectron emission by synchrotron radiation
[22] Y. Tanimoto et al, PRST-AB 12, 110702 (2009) from the beam striking the walls of the cham-
[23] T. Baer et al; F. Zimmermann et al, IPAC’11 ber. This photoelectron instability (PEI) [2] is
[24] C.F. Bohren, D.R. Huffman, Absorption and one of many ECEs investigated in positron stor-
Scattering of Light by Small Particles, Wiley age rings since then. The phenomenon was later
(1983) studied in dedicated experiments at BEPC and
[25] E.U. Condon, H. Odishaw, Handbook of Physics, APS [4, 5]. The ECE led to serious performance
2nd ed., McGraw-Hill (1967) limitations at PEP-II and KEKB [6]. A closely
[26] A. Piwinski, CERN 85-19 (1985) related coupled-bunch instability was previously
[27] K. Yokoya, P. Chen, Lect. Notes in Phys. 400, observed at CESR, although in this case the pho-
Springer Verlag (1990) p.415
toelectrons were trapped and localized in a section
[28] T. Raubenheimer, P. Chen, LINAC 92
of the ring rather than spread out over most of the
[29] P. Emma, T. Raubenheimer, F. Zimmermann,
circumference. A comprehensive program dedi-
EPAC 94, p.1162
[30] T.O. Raubenheimer, F. Zimmermann, PR E52,
cated to measurements and analysis of ECE’s for
No.5 (1995) 5487 e+ e− storage rings is now ongoing at CESR [7].
[31] G.V. Stupakov et al, PR E52, 5 (1995) 5499 The above-mentioned ECE’s are related to
[32] G.V. Stupakov, Proc. Int. Workshop on Collec- previously observed electron-proton dynamical
tive Effects and Impedance for B factories, KEK effects such as beam-induced multipacting (BIM),
Proc. 96-6 (1996) p.243 first observed at ISR [8] when operated with
[33] A.W. Chao, G.V. Stupakov, Proc. MBI97 work- bunched beams. Closely related to BIM is
shop (1997) trailing-edge multipacting observed at the LANL
[34] G.V. Stupakov, PRST-AB 3 (2000) 019401 spallation neutron source PSR [9], where electron
[35] R. A. Bosch, PRST-AB 3 (2000) 034402 detectors register a large signal during the passage
[36] J. Byrd et al, PRL, Vol.79, No.1 (1997) 79 of the tail of the bunch even for stable beams.
163
All ECEs in e+ e− as well as in hadron storage when the net number of electrons generated by
rings have precursors in the e-p instabilities for primary mechanisms balances the net number of
bunched and unbunched beams first seen at BINP electrons absorbed by the walls. If δeff > 1, the
in the mid-60s [10]. EC grows exponentially. The exponential growth
slows down as the space-charge fields from the
Phenomenology In e+ or e− rings, the EC typ- EC effectively neutralize the beam field, reduc-
ically starts when the synchrotron radiation (SR) ing the electron acceleration. Ultimately, the pro-
emitted by the beam yields photoelectrons upon cess stops when the EC space-charge fields are
striking the vacuum chamber. In proton rings, the strong enough to repel the electrons back to the
EC is typically initiated by ionization of residual walls of the chamber upon being born, at which
gas, or from electron generation when stray beam point δeff becomes = 1. At this point, the EC dis-
particles strike the chamber. The LHC is the first
tribution reaches a dynamical equilibrium charac-

proton storage ring in which the beam emits copi- terized by rapid temporal and spatial fluctuations,
ous SR (critical energy ∼44 eV), leading to sub-
determined by the bunch size and other variables.

stantial photoemission, hence the mechanism for For typical present-day storage rings, whether us-
the formation of the EC is analogous to presenting e+ or p beams, the spatio-temporal average ne
day positron rings [11]. reaches a level ∼1010−12 m−3 . The energy spec-
The above-mentioned primary mechanisms trum of the electrons typically peaks at an energy
are typically insufficient to lead to a significant below ∼100 eV, and has a high-energy tail reach-
EC density ne . However, these primary elec- ing out to keV’s. If there is a gap in the beam, or if
trons are kicked by successive bunches, strik- the beam is extracted, the cloud dissipates with a
ing the chamber walls with a broad energy spec- falltime that is controlled by the low-energy value
trum and leading to secondary electron emission of δ(E) [16]. In general, there is no simple, direct
(SEE). This can lead to a large amplification (∼ a correlation between the risetime and the falltime.
few orders of magnitude) of the primary electron Electron generation, trapping and ejection from
density and to strong temporal and spatial fluctu- quadrupole magnets at the PSR is significant, and
ations in the electron distribution [12]. This com- these electrons act as seeds for the EC buildup in
pounding effect of SEE is usually the main deter- nearby drift regions [17].
minant of the strength of the ECEs, and is particu-
In regions of the storage ring with an external
larly strong in positively-charged bunched beams. magnetic field, such as dipoles, quadrupoles, etc.,
The ECE combines many parameters of a the EC distribution develops characteristic geo-
storage ring such as bunch current, bunch spacing, metrical patterns. For typical B-fields in the range
beam energy [13], vacuum chamber geometry, 0.01–5 T and typical EC energies < 100 eV, the
vacuum pressure, and electronic properties of the electrons move in tightly-wound spiral trajecto-
chamber surface material such as photon reflectiv- ries about the field lines. In practice, in a bending
ity Rγ , effective photoelectric yield (or quantum dipole, the electrons are free to move in the verti-
efficiency) Yeff , secondary electron yield (SEY), cal (y) direction, but are essentially frozen in the
characterized by the function δ(E) (E =electron- horizontal (x). As a result, the y-kick imparted
wall impact energy), secondary emission spec- by the beam on a given electron has an x depen-
trum [14, 15], etc. The function δ(E) has a dence that is “remembered” by the electron for
peak δmax typically ranging in 1 − 4 at an en- many bunch passages. It often happens that the
ergy E = Emax typically ranging in 200 − 400 electron-wall impact energy equals Emax at an x-
eV. A convenient phenomenological parameter is location less than the chamber radius. At this lo-
the effective SEY, δeff , defined to be the aver- cation δ(E) = δmax , hence ne is maximum, lead-
age of δ(E) over all electron-wall collisions during to characteristic high-density vertical stripes
ing a relevant time window. Unfortunately, there symmetrically located about x = 0 [18]. For
is no simple a-priori way to determine δeff , be- quadrupole magnets, the EC distribution develops
cause it depends in a complicated way on a com- a characteristic four-fold pattern, with character-
bination of many of the above-mentioned beam istic four-fold stripes [19].
and chamber parameters. If δeff < 1, the cham-
ber walls act as net absorbers of electrons and Diagnostics Dedicated electron detectors have
ne grows linearly in time following beam injec- been designed and used to study the intensity
tion into an empty chamber. The growth saturates and energy spectrum of the electrons striking the
164
Ch.2: BEAM DYNAMICS
vacuum chamber. The EC average density can A high-current instability that has been ob-
be inferred via simulations, in some cases, from served for many years at the PSR is also an ECE
the measured coherent tune shift of successive [9]. The phenomenon has been studied in intense,
bunches along a train. Other techniques, notably long-pulse, heavy-ion fusion drivers [32], at the
the use of microwaves injected into the vacuum J-PARC proton rings [33], and at the FNAL Main
chamber, have been employed. See Sec.7.4.15. Injector [34].
If a train of bunches, separated by a time in-
Consequences One consequence of the EC is a
terval τ , traverses a region with a dipole magnetic
strong increase in the vacuum pressure as a func-
field B, beam-EC resonances leading to an en-
tion of bunch current. The pressure rise exhibits
hancement of ne occur when τ is an integer mul-
a threshold behavior, and is sensitive to the bunch
tiple of the electron cyclotron period, i.e. when
fill pattern at fixed total current, as observed in
ωτ /2π = integer, where ω = eB/m, m being

the B factories and the SPS [20]. When RHIC
the electron mass [35]. The strength of these res-
was operated with ion beams, the pressure rise
onances is characterized by the phase-averaging

was rather dramatic at transition energy due to the
factor e−ω σ /2 , where σ = rms bunch length (in
2 2
short bunch length, often triggering a beam abort
by the machine protection system [21] (the prob- units of time). Consequently, they are well de-
lem has been controlled via low-SEY coatings— fined only for short bunches, i.e., for ωσ <∼ 1.
see below). Mitigating mechanisms A low value of Yeff
In 1996-97, calculations showed that the LHC is favorable, as ne is a monotonically increasing
will be subject to an ECE [11, 22, 23]. The main function of Yeff . At the LHC small transverse
concern here is the power deposited by the elec- grooves were engraved on the vacuum chamber,
trons rattling around the chamber, which must be effectively resulting in normal photon incidence
dissipated by the cryogenic system. Since the thus leading to a reduction of Yeff by a factor of
cryogenic system was designed before the discov- 2–4, and Rγ by a factor ∼ 10 [36].
ery of the ECE, substantial effort has been de- Yeff can also be reduced by an antechamber
voted since 1997 to better estimate the power de- on the outboard side of the vacuum chamber. At
position, to identify the conditions under which PEP-II [37], the antechamber allows for ∼99% of
the cooling capacity may be exceeded, and to de- the photons to escape out of the vacuum chamber.
vise mitigation mechanisms if necessary. As part The remaining 1% of the photons, however, are
of this effort, the ECE has been studied at SPS and radiated at wide angle and low energy, generat-
PS at the high beam intensities required for nom- ing photoelectrons more efficiently [38] than the
inal LHC operation. A comprehensive reference high-energy photons that escape.
database can be found in [24]. In practice, lowering Yeff is effective only
EC-induced single-bunch head-tail instabili- when SEE is not dominant. If SEE off the cham-
ties [25, 26, 27] have been predicted and observed ber surface is significant, achieving a low value of
at several storage rings. Before EC mitigation δeff is essential. Lowering Yeff via antechambers
mechanisms were implemented at KEKB, such is a practical option only when implementable, as
instabilities led to an effective beam blowup, lim- is often the case in Al chambers made by extru-
iting luminosity performance. Bunch-to-bunch sion. Although pure Al has a low peak value of δ,
tune shift that grows towards the tail of the bunch δmax <∼ 1, its surface is normally covered with a
train, sometimes leading to coupled-bunch insta- layer of Al2 O3 with δmax ∼ 2.5 − 3, among the
bilities, were observed [1, 28, 29]. highest of all metals used in practice to build vac-
Electrons from the cloud getting “sucked uum chambers. Copper and stainless steel have
into” the body of a passing bunch typically lead δmax ∼ 1.3 − 1.5 when adequately conditioned,
to a rather significant head-tail tune spread. When but this value may not be low enough in many
this tune spread is combined with synchrotron cases. Beryllium, used in rf cavity windows and
oscillation and/or space-charge forces, incoher- in the vicinity of IP of a collider, has similar fea-
ent effects result, such as slow emittance growth tures to Al2 O3 : its oxide has δmax ∼ 4 − 5, but its
and/or particle losses. The bunch current thresh- elemental form has δmax ∼ 0.6 − 0.9.
old for such effects can be lower than for the The ECE is a self-conditioning effect in the
above-mentioned single-bunch instabilities [30, sense that δmax gradually decreases owing to
31]. the electron-wall bombardment during normal
165
machine operation [9]. The important practical are being studied to control the single-bunch EC-
question is how long it takes for δmax to fall below induced instability observed at the SPS [54].
a value where the EC is no longer an operational Besides the usual accelerator conference pro-
limitation. ceedings and internet websites of the ECLOUD
Low-emission coatings are usually employed and HB workshop series, other useful websites
to reduce SEE. For example, the Al chambers in can be found at:
the arcs of the PEP-II e+ ring were coated with –Pressure Rise 2003 (BNL):
TiN [39]. This coating, once properly condi- http://www.c-ad.bnl.gov/icfa/
tioned, has δmax ∼ 1.1. TiN coatings are used in –Stopping & range of electrons in matter:
the SNS [40] and are recommended for the damp- http://www.srim.org/SREM.htm
ing rings of a future linear collider [41]. Other –UT-Knoxville metrology & lithography:
coatings, such as the NEG substance TiZrV, have http://web.utk.edu/∼srcutk/

been shown to reduce SEE and have been, or will –EC in the LHC:
be, used in the warm regions of RHIC and the http://ab-abp-rlc.web.cern.ch/ab-abp-rlc-ecloud/

LHC [42, 43], and possibly in the ILC damping –Comparison of EC Simulations:
rings [44]. Amorphous (glassy) carbon is being http://wwwslap.cern.ch/collective/ecloud02/ ec-
actively investigated as an effective coating for the sim/ index.html
LHC [45, 46]. –CERN website on SEE:
Longitudinal grooves (parallel to the beam di- http://at-div-vac.web.cern.ch/at-div-vac/
rection) of pitch and depth on the mm scale etched VACPAGES/ps/Phys&tech/Phys/Ecloud/
on the vacuum chamber walls have shown to sup- SEY/SEY paper.html
press SEY [47]. –NIST database on electron impact cross-
sections:
Reducing δmax , however, is sometimes not
http://physics.nist.gov/PhysRefData/Ionization/
enough: the low-energy (E < ∼ 10 eV) value of –LBNL X-ray interactions with matter:
δ(E), and certain details of the emission energy
http://henke.lbl.gov/optical constants/
spectrum, primarily related to the intensity of the
rediffused component [15, 48, 14], can have a sig-
nificant effect on the survival of the EC during a References
beam gap and hence on the average ne . [1] M. Izawa et al, PRL 74, 5044 (1995)
Active mechanisms have been used to control [2] K. Ohmi, PRL 75, 1526 (1995)
EC-induced instabilities, such as raising the verti- [3] S. Heifets, Proc. CEIBA95, p.295
cal chromaticity above its nominal value, or using [4] Z.Y. Guo et al, PAC97, 3B012
octupoles [49, 50]. Solenoidal windings wrapped [5] K.C. Harkay, PAC99, TUAL4
around most of the circumference of the e+ rings [6] H. Fukuma, ICFA BDNL 48, p.112 (Apr. 2009)
of PEP-II and KEKB, proved essential for these [7] M.A. Palmer et al, PAC09, FR1RAI02
machines to reach, and exceed, their specified per- [8] O. Gröbner, 10th Intl. Conf. on Part. Accel, 1977,
formance [6, 51]. Solenoids have also been used p.277
in parts of the circumference of RHIC [21]. A rel- [9] R.J. Macek et al, ECLOUD04, p.93
[10] V. Dudnikov, PAC2001, TPPH094
atively low solenoidal field (B ∼ 20 − 30 G) is
[11] F. Zimmermann, SLAC-PUB-7425, LHC Project
sufficient to trap the electrons near the walls of the
Report 95 (1997)
vacuum chamber, thereby minimizing their effect
[12] G. Rumolo et al, PRST-AB 4, 012801 (2001);
on the beam. Elaborate bunch fill patterns were Erratum: 4, 029901 (2001)
used at PEP-II [52] and RHIC [21]. The many [13] G. Rumolo et al, PRL 100, 144801 (2008)
gaps in these patterns have the effect of promoting [14] M.A. Furman, LBNL-60512/CBP Note-736
the dissipation of the EC. At PSR it was found that (2006); Proc. HB2006, http://hb2006.kek.jp/
increasing the momentum spread of the beam in- [15] R. Cimino et al, PRL 93, 014801 (2004)
creases the current instability threshold [9]. Clear- [16] M.A. Furman, PAC03, TOPC001
ing electrodes have been designed and tested at [17] R. Macek et al, ECLOUD10
KEKB [53]. To the extent that the EC-induced [18] F. Zimmermann, ECLOUD02, p.47
instability has some degree of coherence, it is in [19] G. Arduini et al, EPAC04, WEPLT044
principle possible to combat it via a feedback sys- [20] G. Arduini et al, ECLOUD04, p.31; J.M.
tem. High-bandwidth (up to ∼1 GHz) systems Jiménez, p.49
166
Ch.2: BEAM DYNAMICS
[21] W. Fischer et al, PRST-AB 11, 041002 (2008)

[22] O. Gröbner, PAC97, p.3589
[23] M.A. Furman, LBNL-41482/CBP Note
247/CERN LHC Report No. 180 (1998)
[24] EC in the LHC, see URL list above.
[25] K. Ohmi, F. Zimmermann, PRL 85(18), 3821
(2000)
[26] K. Ohmi et al, PRE 65, 016502 (2002)
[27] F. Zimmermann, PRST-AB 7, 124801 (2004)
[28] J.A. Crittenden et al, IPAC10, TUPD024
[29] K. Ohmi et al, ECLOUD04, p.313
[30] G. Franchetti et al, PRST-AB 12, 124401 (2009)
[31] J.-L. Vay et al, IPAC10, WEOBRA02 Figure 1: CSR wake W as a function of distance z.
[32] A.W. Molvik et al, POP 14, 056701 (2007) The wake is normalized by Z0 cγ 4 /3πρ2 and the dis-
[33] T. Toyama et al, ECLOUD04, p.77 tance is measured in units 2ρ/3γ 3. The wake vanishes
[34] K.G. Sonnad et al, PAC07, FRPMS028 for z < 0.
[35] C.M. Celata et al, PRST-AB 11, 091002 (2008)
[36] V. Baglin et al, EPAC 98, TUP18H
[37] M.A. Furman et al, MBI97, p.170
[38] O. Gröbner et al, J. Vac. Sci. Technol. A7, 223
(1989)
[39] K. Kennedy et al, PAC97, 8C009
[40] P. He et al, PAC01, p.2159
[41] M. Pivi et al, IPAC10, WEPE097
[42] LHC Design Report, CERN-2004-003, 1, p.346
[43] Y. Suetsugu et al, PAC07, FRPMN042
[44] M. Pivi et al, PAC05, ROPB001
[45] E.N. Shaposhnikova et al, PAC09, MO6RFP008
[46] C. Yin Vallgren et al, IPAC11, TUPS027 and
TUPS 028
Figure 2: CSR wake (per unit length, solid curve) for
[47] J.S.T. Ng et al, ICFA BDNL 48 (2009) p.102; Y.
a Gaussian bunch (dashed curve) with rms length σz .
Suetsugu et al, p.118 4/3
[48] M.A. Furman, V.H. Chaplin, PRST-AB 9, The wake is normalized by Z0 c/4πσz ρ2/3 and the
034403 (2006) distance is measured in units σz . The head of the
[49] Z.Y. Guo et al, MBI97, p.150 bunch corresponds to z > 0.
[50] H. Fukuma, Two-Stream2000
[51] J.T. Seeman, EPAC08, TUXG01 coordinate z is measured in the direction of mo-
[52] F.-J. Decker et al, PAC01 TPPH126 tion, and the wake is localized in front of the par-
[53] Y. Suetsugu et al, IPAC10, WEOAMH01 ticle in contrast to “traditional” wakes which trail
[54] J.D. Fox et al, IPAC10, WEPEB052 the source charge. This is because the charge fol-
lows a circular orbit and the radiation propagates
along chords getting ahead
∞ of the source. The
2.4.15 Coherent Synchrotron Radiation
wake has the property 0 W (z)dz = 0. For dis-
Instability
G. Stupakov, SLAC tances such that ρ z ρ/γ 3 (corresponding
to abscissa values >∼5 in Fig.1) the wake (per unit
Radiation reaction force—CSR wake field length of path) is
The collective force acting on a beam due to its
coherent synchrotron radiation (CSR) is described Z0 c 1
W (z) = − (1)
in terms of the so called CSR longitudinal wake 2π 34/3 ρ2/3 z 4/3
[1, 2, 3]. For an ultrarelativistic particle and a
long magnet, the plot of the wake (Sec.3.2.1) in The CSR wake convoluted with a Gaussian bunch
the vicinity of a point charge as a function of the is shown in Fig.2. The negative value of this wake
distance z along the particle’s orbit is shown in in the head of the bunch corresponds to energy
Fig.1, where ρ denotes the bending radius. The gain. From Eq.(1) one finds the CSR longitudinal
167
impedance where I is the peak beam current, IA = 17.5

kA, σδ is the rms relative energy spread, η is the
1 ∞
Z (ω) = dzW (z)e−iωz/c momentum compaction factor (assumed positive),
c 0 and ρ = C/2π with C the ring circumference.

Z0 eiπ/6 2 ω 1/3 Wall shielding suppresses the instability at long
= 1/3
Γ (2) wavelengths and limits the region of instability to
2π 3 3 c ρ2/3
1/3
k> 1/2 3/2
∼ ρ /h . Further theoretical studies of the
where Γ is the gamma function. The real part CSR instability included detailed computer simu-
of this impedance is related to the spectrum of lations of the instability in a nonlinear “bursting”
the energy loss of a charge due to the radiation: phase [8].
dP /dω = (e2 /4π 2 0 )ReZ , where dP /dω is the A comparison of the theory with observations
power of synchrotron radiation per unit length in
has been carried out in the experiment at the ALS

unit spectral interval. Eq.(2) is valid for c/ρ [9]. The observed thresholds for the CSR bursts
ω cγ 3 /ρ.
agreed well with the instability thresholds from

Wall shielding and formation length The the coasting beam model.
wake Fig.2 is valid for long magnets, when the CSR instability in bunch compressors Mi-
magnet length lm is much larger than the so called crobunching due to the CSR induced instabil-
formation length of the radiation (24σz ρ2 )1/3 for ity has been identified as a potential danger in
a given bunch length. In the opposite limit the bunch compressors. Experimental observations
wake is dominated by transient effects at the en- [10] showed an increased energy spread and hor-
trance and the exit from the magnet [4]. izontal emittance growth of the beam caused by
Close proximity of metallic walls to the beam the coherent radiation. The effect has been stud-
orbit suppresses the CSR wake. This wall shield- ied numerically [11], and several theoretical mod-
ing effect becomes important when the distance els of the instability in bunch compressors were
from the beam orbit to the walls h is such that developed [14, 12, 13].
h< 2 1/3
∼ (σz ρ) .
Curvature induced transverse electromagnetic References
forces In addition to the longitudinal force there [1] L.V. Iogansen, M.S. Rabinovich, Sov. Phys.
is also a transverse electromagnetic force F⊥ JETP, 37 (1960) 83
caused by the curvature of the beam trajectory. [2] Y.S. Derbenev et al, DESY FEL Report TESLA-
This force varies in radial direction within the FEL 95-05 (1995)
bunch however the gradient ∂F⊥ /∂r does not [3] J.B. Murphy, S. Krinsky, R.L. Gluckstern, PAC
cause additional beam focusing to the lowest or- 95
der in curvature [5]. This effect is referred to as [4] E.L. Saldin, E.A. Schneidmiller, M.V. Yurkov,
the cancellation effect, and is due to the compen- NIM A398 (1997) 373
sation of the gradient of F⊥ by the particle’s en- [5] E.P. Lee, PA 25 (1990) 241
ergy variation in the potential field of the beam [6] Y.S. Derbenev, V.D. Shiltsev, SLAC-PUB-7181
space charge [6]. (1996)
[7] G. Stupakov, S. Heifets PRST-AB 5 (2002)
Instability The CSR wakefield can cause a mi- 054402
crowave instability of a beam in a ring. Theo- [8] M. Venturini, R. Warnock, PRL 89 (2002)
retical analysis of the instability based on the as- 224802
sumption of short wavelength of the perturbation, [9] J. Byrd et al, PRL 89 (2002) 224801
kσz 1, and the free space CSR wakefield (1) [10] H. Braun et al, PRST-AB 3 (2000) 124402
leads to the following conclusions [7]. The beam [11] M. Borland, PRST-AB 4 (2001) 070701
is unstable for wavenumbers k < 2.0Λ3/2 /ρ with [12] E.L. Saldin, E.A. Schneidmiller, M.V. Yurkov,
the maximum growth rate at k = 0.68Λ3/2 /ρ NIM A490 (2002) 1
equal to (Im ω)max = 0.43Λ3/2 cηδ0 /ρ. The di- [13] Z. Huang, K.-J. Kim, PRST-AB 5 (2002) 074401
mensionless parameter Λ is defined by [14] S. Heifets, G.V. Stupakov, S. Krinsky, PRST-AB
5 (2002) 064401
1 I ρ
Λ= (3)
ηγσδ2 IA ρ
168
Ch.2: BEAM DYNAMICS
2.5 BEAM-BEAM EFFECTS Here w is the complex error function (Sec.1.5.1)

w(z) = e−z {1 − erf(−iz)}
2
2.5.1 Beam-Beam Effects in Storage Ring
Colliders See also Sec.2.5.4.
K. Hirata, Graduate U., Sokendai Define the (incoherent) beam-beam parame-
We assume a two-ring e+ e− collider with one inter ξ ± as
teraction point (IP). Modifications for other types N ∓ re βq0±
of colliders are straightforward. The beam-beam ξq± = (6)
γ± 2πσq (σx∓ + σy∓ )
∓
effects consist of (i) the beam-beam interaction at
the IP (s = 0) and (ii) the transformation from IP where βq0 is the nominal (= without beam-beam
to IP. The two are independent of each other dy- effect) β-function at the IP.
namically but all beam-beam phenomena are their When both beams are Gaussian,
combined effects [1, 2, 3, 4, 5].
+ −x̄− )2 (ȳ + −ȳ − )2

N + N − f0 − (x̄ −
L0 = LG
0 = e 2Σ2
x 2Σ2y
2.5.1.1 Infinitely short bunches 2πΣx Σy

(7)
One can ignore the bunch length σz when σz with the effective beam size
βq at the IP (q stands for x or y). Let ρ± (x, y)
be the transverse distribution of e± beams. The Σq = (σq+ )2 + (σq− )2
luminosity (Sec.4.1) for very short bunches, L0 , The coherent beam-beam parameter is
is
N∓ re βq0±
L0 = N + N − f 0 dxdy ρ+ (x, y)ρ− (x, y) (1) Ξ±
q = (8)
γ± 2πΣq (Σx + Σy )
At the IP, a particle is kicked by the beam- The luminosity, Eq.(7), can be expressed in terms
beam force of Ξ as
p± ± ±
x,y → px,y − fx,y (x, y) −x̄ ) + − 2 + −ȳ − )2
f0 N+ N− γ+ γ− − (x̄ 2Σ − (ȳ
LG
2 2Σ2
± ± ∓
fx,y (x, y)= −N∓ ∂x,y φ (x, y; ρ ) (2) 0 =
r N γ + N− γ−
e x y
where φ is the “potential” which depends on the ρ #e + + $

Ξ+ Ξ− Ξ+ Ξ−
of the other beam, φ± (x, y; ρ) = × x
+ x
+
y
+
y
(9)
βx0+ βx0− βy0+ βy0−
re
dx∗ dy ∗ G(x−x∗ , y−y ∗ )ρ(x∗ , y ∗ ) (3)
γ± which shows that the β 0 ’s should be small to ob-
where G(x, y) = − ln(x2 + y 2 ). tain large luminosity, because there exists an em-
Gaussian beams If ρ± is Gaussian, pirical upper bound for Ξ’s.
(x−x̄± )2 (y−ȳ ± )2 Linear dynamical effects In the linear approx-
1 − ± )2 − 2(σ ± )2
ρ± (x, y) = e 2(σx y
imation, the dynamics can be treated as 1-D sys-
2πσx± σy±
tem. We denote (q, pq ) simply as (q, p) and omit
where x̄± = x± and σx± = (x − x̄± )2 ± , ± designations. Here we use the weak-strong pic-
etc, the force felt by a particle can be put in the ture, where one (weak) beam consists of a single
Bassetti-Erskine formula [6], particle so that it does not affect the other (strong)
± N∓ re beam.
fx,y (x, y) = Fx,y (x − x̄∓ , y − ȳ ∓ ; σx∓ , σy∓ ) tune shift When x̄+ = x̄− and ȳ + = ȳ − at the IP,
γ±
(4) the head-on collision reads

Fy (x, y, σx , σy ) + iFx (x, y, σx , σy ) q q 1 0
⎡ ⎛ ⎞ →K ,K = 4πξ0 (10)
p p − β0 1
2π ⎣ ⎝ x + iy ⎠
= w (5) Here ξ0 and β0 are the nominal values. It is
σx2 − σy2 2(σx2 − σy2 ) combined with the one-turn matrix without beam-
⎛ ⎞⎤
σy σx beam effects,
x2 y2 x + i y
− exp − 2 − 2 w⎝ x ⎠⎦
σ σy
q q cos μ0 β0 sin μ0
2σx 2σy 2(σx2 − σy2 ) →M ,M=
p p − sinβ0μ0 cos μ0
169
Sec.2.5: BEAM-BEAM EFFECTS
From the eigenvalues of MK, we get

cos μ = cos μ0 − 2πξ0 sin μ0 (11)
For small ξ0 , we have ν = μ/2π = ν0 + ξ0 . Thus
ξ0 is also called the beam-beam tune shift param-
eter. Stability requires
1
ξ0 ≤ cot πν0 (0 ≤ ν0 < 1/2) (12)
2π

which is periodic in ν0 with period 1/2.
Figure 1: Dynamic beta, emittance and beam size ef-
Dynamic beta [1] The new β-function, β, at the
fects as functions of ν0 for ξ0 = 0.05. Solid curve is

IP can be found from
σ/σ0 ; dashed curve is /0 ; dotted curve is β/β0 .
cos μ β sin μ
K1/2 MK1/2 =
− sin μ/β cos μ Offset collision When there are offsets (x̄+ −
as x̄− = dx and ȳ + − ȳ − = dy ), the beam-beam
kick, Eq.(10), is modified as [8]
sin μ0 β0 ⎡ ±⎤ ⎡ ⎤⎡ ±⎤
β = β0 = x 1 0 0 0 x
sin μ 1 + 4πξ0 cot μ0 − 4π 2 ξ02 ±
⎢ px ⎥ ⎢ −kxx 1 −kxy 0 ⎥ ⎢ p± ⎥
(13) ⎣ y ± ⎦ → ⎣ 0 0 1 0 ⎦ ⎣ yx± ⎦ (18)
Dynamic emittance [7] Synchrotron radiation ef-
p±y −kxy 0 −kyy 1 p±
y
fects can be included by putting
where
q q r̂ ±
ξ0y 4πσy
→ ΛMK + 1 (14) kyy + ikxy = 1 − e−A σσxy
p p r̂2 β ± σx −σy
0y
√
where Λ is 2 × 2 matrix representing the average − √ 2π 2 (dy − idx ) w(B) − e−A w(C)
2(σx −σy )
effect of radiation on the one-turn matrix and the
ξ ± 4πσx −A σy
r̂’s represent its stochastic effect. Define the en- −ikxy + kxx = β0x
± σx −σy 1 − e σx
0x
velope matrix √
π
2 + √ 2 2 (−dy + idx ) w(B) − e−A w(C)
2(σx −σy )
q qp
R= 2 (15) d2x d2y d +id
qp p with A = 2σx2
+ 2σy2
,B = √ x 2 y 2 and
2(σx −σy )
σy σ
The equilibrium value of R is the solution of d +i σy dy
σx x
C= √ x
.
2(σx −σy2 )
2
t
R = (ΛMK)R(ΛMK) + D, D = r̂i r̂j Nonlinear 1-D models
√ (16) Round beam For σx = σy (round beam) and x̄ =
The equilibrium emittance is = det R. ȳ = 0, and if we observe a particle always with
Λ and D are determined by the lattice in the y = 0, we get
arc and do not depend on the beam-beam interac-
tion. If p± ± ± ∓
x → px − fr (x; σx ) (19)
2

10 β 0 2N∓ re 1 x
Λ = e−δ , D = 0 1−λ2 0 −1 fr± (x; σx ) = exp − 2 − 1
01 0 β0 γ± x 2σx
Flat beam For σy σx and y σx , and we
then observe a particle always with x = 0, then [6]
(1 + 2πξ0 cot μ0 )0 p± ± ± ∓
= + O(δ) (17) y → py − ff (y; σy )
1 + 4πξ0 cot μ0 − 4π 2 ξ02 √
± N∓ re 2π y
The dynamic beta, ff (y; σy ) = erf √ (20)
√ dynamic emittance and γ± σx∓ 2σy
dynamic beam size ( β +O(δ)) effects are illus-
trated in Fig.1. By Eqs.(12) and (13), ν0 slightly Both models show flip-flop hysteresis and cusp
above a half-integer is preferred. catastrophe behaviors [6].
170
Ch.2: BEAM DYNAMICS
Nonlinear Perturbative Effects Under the as- The beam-beam kick matrix R is
sumption of linear betatron oscillation and the
I− A(Ξ + ) A( Ξ+ Ξ− )
beam-beam kick at the IP, the first order perturba- + −
R(Ξz , Ξz )= z z z
− −
tion shows the amplitude dependence of the tune A( Ξ+ z Ξz ) I − A(Ξz )

shifts [1] 0 0
A(Ξ) = (25)
0 0 1 + 1/a 4πΞ 0
νx = νx + ξ x (21)
2 The perturbed tunes are given by

∞ Z1 1+uαx αy
Z2 1+u/a 1
2 −
× du cos μ = (cos μ+ 0 + cos μ0 ) (26)
(1 + u) (1 + u/a )1/2
3/2 2 2
0
− − 1 √
0 sin μ0 − πΞ0 sin μ0 ±
− πΞ+ +
D
1+a 2
νy = νy0 + ξy0 (22)
2 where
D = 16π 2 Ξ− + + −
∞ Z1 1+u
αy
Z2 1+aαx 0 Ξ0 sin μ0 sin μ0
2u
× − − − 2
0 −cos μ0 −2πΞ0 sin μ0 +2πΞ0 sin μ0 ]
+[cos μ+
du + +
(1 + u) 3/2 (1 + a2 u)1/2
0
Here The mapping is unstable if (i) ν0+ or ν0−
<
∼+half-integers (half integer resonances); or (ii)
Z1 (x) = e−x [I0 (x) − I1 (x)] ν0 + ν0− <
∼ integer (sum resonances) [6].
Z2 (x) = e−x I0 (x) When ν0+ = ν0− ≡ ν0 , M is simply
√
where αq is the amplitude normalized by 2σq , U − UA UA
a = σy /σx is the aspect ratio, and I0 and I1 are M=
UA U − UA
modified Bessel functions.
Coherent beam-beam motion Assume the which can be diagonalized as
Rigid Gaussian Model (RGM) in which the beams
U O
are Gaussian with the nominal beam sizes and LML−1 =
O U − 2UA
only their barycenter can change. The beam-beam
kick for the barycenters are [6] with
1 I I
N ∓ re L= √
p̄± ±
q → p̄q − Fq (x̄± − x̄∓ , ȳ ± − ȳ ∓ ; Σx , Σy ) 2 −I I
γ±
(23) Thus the dynamics is split into two modes:
where F is in Eq.(5). A consequence is that the (i) σ-mode: the mode is
linear approximation of the kick for p̄q is 1
Zσ = √ (Z+ + Z− )
± ± ±
δp̄± ∓
q = −4πΞ0 /β0 (q̄ − q̄ )
2
instead of Eq.(10). whose eigentune is unaffected (νσ = ν0 ).
The one-turn map is (ii) π-mode: the mode is
1
M = Û(ν0+ , ν0− )R(Ξ+ −
0 , Ξ0 ) (24) Zπ = − √ (Z+ − Z− )
2
Û is the block-diagonal 4 × 4 matrix through the whose eigentune is gives by
arc,
cos μπ = cos μ0 − 2π(Ξ+ −
U(ν0+ ) 0 0 + Ξ0 ) sin μ0
+ −
Û(ν0 , ν0 ) =
0 U(ν0− ) When two beams are identical, Ξ = ξ/2 so that
νπ ν0 + ξ 0 .
cos μ sin μ
U(ν) = Choice of working point In more general cases
− sin μ cos μ
where there are N e+ bunches and N e− bunches,
M operates on the dynamical variables (Z+ , Z− )t periodically placed and colliding at NIP periodi-
where cally placed collision points (denoted as N ⊕N =
⎡ ⎤
q̄ ± / β0± NIP ), we know that [10]: (i) All the cases can
Z± = N± γ± ⎣ ⎦ be reduced to one of the two irreducible cases,
β0± p̄±
q NIP = 2N and NIP = N with N odd; (ii)
171
In irreducible cases, with perfect symmetry be- where

tween bunches, beams, IPs, and arcs between IPs, Λ(r) = 1.33 − 0.37 r + 0.279 r 2 (29)
the instability occurs for ν0 <∼ NIP /(2N ) with For a very flat beam, r = 0 and G is 1.33 (hori-
= 1, 2, ..., 2N (This pattern repeats for ν0 at ev- zontal) and 1.24 (vertical).
ery NIP ); (iii) With broken symmetries between For higher order coherent motions, refer to
bunches and IPs, the instability pattern for ν0 is [2, 6] and references quoted therein. When the
similar to that of the 1 ⊕ 1 = 1 case. Examples deflection is large, one can observe a hysteresis in
for 3⊕3=3 and 3⊕3=6 cases are shown in Fig.2. the frequency response of the beam [6].
2.5.1.2 Long bunches

When σz is comparable to βq , we should consider

the bunch length effect (hour-glass effect) [15].
The interaction now depends also on the longitu-

dinal position z. The collision between particles
with z + and z − takes place at
s = S(z + , z − ) = (z + − z − )/2 (30)
where the beam sizes are different from those at
s = 0 (the IP).
Luminosity The geometric reduction of luminos-
ity due to the hour glass effect is discussed in
Sec.4.1.
Dynamics The only known symplectic beam-
beam map usable for a long bunch is the synchro-
Figure 2: The threshold values of Ξ as functions of ν0 beam mapping [16]. The map for the collision of
(the “one turn” tune). Upper: 3⊕3=3 case. Lower: an e+ having z + and a 3-D Gaussian e− bunch is
3⊕3=6 case. as follows: The strong e− beam is cut into lon-
Spontaneous beam separation, etc. The RGM as- gitudinal slices: each slice is represented by z − ,
sumes each beam behaves as a single particle. its longitudinal coordinate at s = 0. Let n− i be
Other features related to RGM are the number of electrons in the i-th slice. Then the
(i) an explanation why the four-beam compensa- beam-beam kick is represented by the Lie opera-
tion scheme of DCI experiment failed [11]. tor -
(ii) showing serious instability if the two collider exp n− − −
i :φ X, Y, σx (S), σy (S) : (31)
rings have different circumferences [6]. i
(iii) beyond the threshold of the half integer which acts on the 6 phase space coordinates of the
resonances, two beams tend to separate from e+ at s = 0. Here, S is defined by (30) and
each other at the IP (Spontaneous Separation of X = x + px S(z + , z − ), Y = y + py S(z + , z − )
Beams) [6]. The σ’s should be evaluated at s = S so that, e.g.,

(iv) analysis of parasitic collisions [12, 13]. σx− (S) = x2 0 + 2xpx 0 S + p2x 0 S 2
(v) extension of Eq.(23) for the case where the Here 0 is the average at s = 0.
transverse axes of both beams are tilted [14]. The map Eq.(31) is the successive application
Visible frequencies When we deflect beams with of the following map:
a single frequency, we can observe peaks in its
xnew = x + S(z, z∗ )fX
response corresponding to the above two modes
[6]. When Ξ± pnew = px − f X
0 1 the peaks sit on ν0 and ν0 +
x
− y new = y + S(z, z∗ )fY
Ξ+0 + Ξ 0 ν 0 + ξ0 . This is a result of the RGM.
A more accurate estimate of the visible tune shift pnew
y = py − f Y
can be summarized by [6] z new = z
νπ = ν0 + G(r)ξ0 (27) 1 1
where r = σy /(σ pnew = pz − fX [px − fX ]
% x + σy ) and z
2 2
Λ(r), horizontal 1 1
G= (28) − fY [py − fY ] − g (32)
Λ(1 − r), vertical 2 2
172
Ch.2: BEAM DYNAMICS
where
n−
i re
fX = fx (X, Y ; σx (S), σy (S))
γ0
−
n re
fY = i fy (X, Y ; σx (S), σy (S))
γ0
n− re
g= i R22 (0; z ∗ )gx (X, Y, σx (S), σy (S))
γ0

+ R44 (0; z ∗ )gy (X, Y, σx (S), σy (S)) S
1
gx (x, y, σx , σy ) = −
2(σx2 − σy2 )
' ;
x2 y2
σy − 2σ 2 + 2σ 2
× xfx + yfy + 2 e x y −1
σx
1 Figure 3: The geometrical luminosity reduction factor
gy (x, y, σx , σy ) = (33) L/L0 as a function of Φ ≡ φσz /σx and βy /σz .
2(σx2 − σy2 )
' ;
x2 y2
σx − 2σ 2 + 2σ 2 Here l stands for u, v, w and γl = (1 + α2l )/βl .
× xfx + yfy + 2 e x y −1
σy R2 is a 2×2 horizontal-vertical coupling matrix
defined and calculated from the physical one-turn
2.5.1.3 Dispersion at IP, crossing angle matrix [17]. This factorization can express any
symplectic matrix. The η’s (generalized energy
These two are known to introduce synchrobeta-
dispersion) reduce to the usual dispersion when
tron resonances (Sec.2.3.4) .
there is no dispersion in cavities; otherwise, the
Dispersion Here, we review the properties of
ζ’s (generalized time dispersion) may appear [18].
the arc, which are represented by the one-turn The simplest case where there is a vertical
map around the IP (s = 0). The revolution matrix
dispersion at the IP and nothing else, i.e., where
M(0) can be obtained by any 3-D tracking code
ηy = ηx = ηx = 0, α’s vanish, ζ’s vanish, the
but usually it is useful to represent it in terms of
beam-beam force is linear as K, Eq.(10), and in
several parameters (3, 10, 21 for 1-D, 2-D, 3-D).
the weak-strong situation, then the synchrotron
One way to do it is as follows [17]:
⎡ ⎤ tune is modified as [19]
mu 0 0 (ηy0 )2 σpz
M(0) = H R −1 −1 ⎣
0 mv 0 ⎦ RH νz0 −→ νz0 + ξy0 χ, χ = (35)
σz βy0
0 0 mw
⎡
⎤ as well as the usual transverse tune shifts.
1 − detH x
I 1+a 1
Hx H̃y −Hx Crossing angle The crossing angle can be
⎢ 1+a
⎥ treated by using the Lorentz boost for both beams.
H=⎢ 1
⎣ 1+a Hy H̃x
detH
1 − 1+ay I −Hy ⎦
⎥
At the IP (s = 0), we perform
−H̃x −H̃y aI D BB D −1 A
(34) x(0)−→x∗ (0∗ )−→x∗ (0∗ ) −→ x (0)−→x(0) · · ·
Here Here “BB” is the (symplectic) beam-beam kick
for the head-on collision. In the ultrarelativistic
ζ η −ηq ηq
Hq = q q , H̃q = limit, D can be written as [20]
ζq ηq ζq −ζq
x∗ = tan φz + (1 + h∗x sin φ)x
a = 1 − detHx − detHy
⎡ ⎤ y ∗ = y + h∗y x sin φ
bI −R̃2 0 z ∗ = z/cos φ + h∗z x sin φ
R = ⎣ R2 bI 0 ⎦ , b = 1 − detR2
0 0 I p∗x = (px − h tan φ)/cos φ
p∗y = py / cos φ
cos μl + αl sin μl βl sin μl
ml = p∗z = pz − px tan φ + h tan2 φ
−γl sin μl cos μl − αl sin μl
173
Here φ is the half crossing angle and the horizon- bunch length (see Sec.4.13 for different collisions
tal crossing was assumed, schemes).

The CW scheme can substantially increase
h(px , py , pz ) = pz+1− (pz +1)2 −p2x −p2y (36)
collider luminosity since it combines several
and h∗q = ∂h∗ /∂p∗q . See also Eqs.(15-19), potentially advantageous ideas. Consider two
Sec.2.3.4. bunches colliding under a horizontal crossing an-
The geometrical luminosity reduction factor gle θ (Fig.1). The CW principle can be explained
can be obtained from Eq.(6), Sec.4.1. This factor in three steps. The first step is large Piwinski
is shown in Fig.3 for the case with two identical angle Φ = (σz /σx ) tan(θ/2) 1. In the CW
beams. scheme, Φ is increased by decreasing σx and in-
creasing θ. In this way we can gain in lumi-
References nosity and the horizontal tune shift decreases (as

discussed in Sec.4.13); parasitic collisions (PC)
[1] A.W. Chao, AIP Proc.127 (1983) p.201

(Sec.2.5.4) become negligible since with larger
[2] A.W. Chao, SSCL-346 (1991)
θ and smaller σx the beam separation at the PC
[3] AIP Proc. 214 (1990)
[4] Proc. 3rd ICFA Beam Dynamics Workshop on
is larger in terms of σx . But the most impor-
Beam-Beam Effects in Circular Colliders (1989) tant effect is that the overlap area of the collid-
[5] Proc. 7th ICFA Beam Dynamics Workshop on ing bunches is reduced, since it is proportional to
Beam-Beam Issues for Multibunch, High Lumi- σx /θ (see Fig.1).
nosity Circular Colliders, JINR (1996) As the second step, the vertical β-function βy
[6] K. Hirata, [3] p.175 can be made comparable to the overlap area size
[7] K. Hirata, F. Ruggiero, PA 28 (1990) 137 (i.e. σz ),
[8] B.W. Montague, CERN/ISR-GS/75-36 (1975) 2σx ∼ σz
[9] A. Chao, AIP Proc. 57 (1979) p.42 βy∗ ≈ = σz
θ Φ
[10] K. Hirata, E.Keil, PA 56 (1996) 13
[11] E. Keil, Proc. 3rd Advanced ICFA Beam Dynam- So, reducing βy∗ at the IP gives us several advan-
ics Workshop (1989) p.85 tages:
[12] K. Hirata, AIP Proc. 214 (1990) p.441 • Luminosity increase with the same bunch
[13] M. Furmann et al, LBL-32435, ESG-193, DAP- current.
NIA/SPP 92-03, ABC-49 (1992)
[14] V. Ziemann, SLAC-PUB-5479 (1991)
[15] S. Krishnagopal, R. Siemann, PR D41 (1990)
x EY
2312 e+ e-
[16] K. Hirata, H. Moshammer, F. Ruggiero, PA 40 VxT
(1993) 205
T
[17] K. Ohmi, K. Hirata, K. Oide, PR E49 (1994) 751 Vz*T z
[18] S. Petracca, K. Hirata, PAC 97
2Vz
[19] S. Petracca, K. Hirata, AIP Proc. 395 (1997) 2Vx
p.369
[20] K. Hirata, PRL 74 (1995) 2228
a) Crab sextupoles OFF
2.5.2 Crab Waist Collision Scheme

x EY
M. Zobov, INFN LNF e+ e-
VxT
The crab waist (CW) collision scheme was pro-
posed [1, 2] and successfully tested at the Φ- T
Vz*T z
factory DAΦNE [3]. At present this scheme is
considered to be most attractive for the next gen- 2Vz
2Vx
eration lepton factories since it holds the promise
of increasing the luminosity of the storage-ring
colliders by 1-2 orders of magnitude beyond the b) Crab sextupoles ON
current state-of-art, without any significant in-
Figure 1: Crab Waist collision scheme.
crease in beam current and without reducing the
174
Ch.2: BEAM DYNAMICS
• Possibility of the bunch current increase (if it Δνy

OFF
Δνy
ON
is limited by ξy ), thus farther increasing the
luminosity.
• Suppressing vertical synchrobetatron reso-
nances [4].
Furthermore, there is no need in decreas-
Δνx Δνx
ing the bunch length to increase the luminos-
ity as required in standard collision schemes
(Sec.4.13). This will help solve the problems of Figure 3: DAΦNE beam-beam footprint with crab sex-
HOM heating, coherent synchrotron radiation of tupoles off (left) and on (right) obtained by FMA [6].
short bunches and excessive power consumption.
However, implementation of these two steps DAΦNE has provided a factor of 3 luminosity in-
introduces new beam-beam resonances which crease [3] in good agreement with numerical sim-
may strongly limit the maximum achievable tune ulations [7].

shifts. At this point the crab waist transformation All the projects of the next generation lep-
[1, 2] enters the game. This is the third step. As ton factories are based on the crab waist scheme.
seen in Fig.1b, the β-function waist of one beam Their designs rely on collisions of flat beams with
is oriented along the central trajectory of the other very small emittances [8] comparable to those ob-
beam. In practice the CW vertical β-function ro- tained in modern synchrotron light sources. So is-
tation is provided by sextupoles placed on both sues of achieving and preserving small emittances
sides of the IP in phase with the IP horizontally at high beam current become of primary impor-
and at π/2 phase vertically (Fig.2). tance. This requires a careful study of emittance
The integrated strength K of these sextupoles growth due to wake fields, e-clouds, ion-related
should satisfy a condition depending on θ and the phenomena, intrabeam scattering etc.
β-functions at the IP and the sextupole locations,
References
1 1 βx∗
K= ∗ [1] P. Raimondi, 2nd SuperB Workshop, 2006
θ βy βy βx
[2] P. Raimondi et al, e-Print:physics/0702033 (2007)
The crab waist transformation gives a small [3] M. Zobov et al, PRL104, 174801 (2010)
geometric luminosity gain due to the vertical β- [4] D.V. Pestrikov, NIM A336, 427 (1993)
function redistribution along the overlap area. It [5] P. Raimondi, D. Shatilov, M. Zobov, EPAC08,
is estimated to be of the order of several percent. p.2620
However, the dominating effect comes from the [6] D. Shatilov et al, PRST-AB 14, 014001 (2011)
suppression of betatron and synchrobetatron res- [7] M. Zobov et al, ICFA Beam Dyn. Newslett. 48,
onances (Sec.2.3.4) arising (in collisions without p.34
CW) due to the vertical motion modulation by the [8] ICFA Beam Dynamics Letter 48 (2009)
horizontal betatron oscillations [5]. The collision
of flat beams with σy σx is the essential con-
2.5.3 Beam-Beam Effects in Linear Colliders
dition for the resonance suppression with the crab
P. Chen, Taiwan U. & SLAC
sextupoles [6].
D. Schulte, CERN
Fig.3 demonstrates the resonance suppression
applying the frequency map analysis (FMA) for In linear colliders, the colliding beams must have
the beam-beam interaction in DAΦNE in CW small transverse dimensions to reach high lumi-
collisions [6]. The crab waist collision test at nosity. This leads to a strong electromagnetic
force exerted by each beam on the particles of the
Sextupole IP (Anti)sextupole other. This gives rise to strong disruption effects
Ex,E y E *x , E *y Ex,E y
and leads to the emission of beamstrahlung. A full
S S
analytic treatment of the beam-beam interaction
'P y 'P y is usually not possible, so it is most often stud-
2 2
'P x S 'P x S ied using simulation codes, e.g. CAIN [1] and
GUINEA-PIG [2]; for the latter also a C++ ver-
Figure 2: Crab sextupole locations. sion exists [3].
175
The beams in linear colliders are normally 1

analyt.
Dy=2.5
flat, σx σy . We imply this in the following; Dy=5
the reason for this choice is beamstrahlung. 0.8 Dy=10
Dy=20
Dy=40
0.6
L/L0
2.5.3.1 Disruption
0.4
Crossing angle Often a crossing angle θc is
0.2
foreseen in the horizontal plane between the two
colliding beam lines. The resulting luminosity 0
loss is avoided by using a crab crossing scheme 0 1 2 3 4 5 6
(Sec.4.1). In this scheme an x -z correlation is Δy/σy
introduced in the beam before the collision, such

Figure 2: Luminosity as a function of beam-beam
that it is aligned with the longitudinal axis of the
offset for different disruption parameters. The sim-
laboratory system at the collision, see Fig.1. For

ple analytic expression is compared to simulation with
this case the beam-beam effects can be treated as
GUINEA-PIG.
if no crossing angle were present.
Disruption parameter We first consider the Kink instability The dependence of the lumi-
motion of a single particle moving parallel to the nosity on the beam-beam offset is strongly af-
beam axis near x = y = 0 in the field of the on- fected by the beam-beam interaction. For rigid
coming bunch assumed to have a Gaussian profile bunches and assuming βx,y σz , the luminosity
with rms sizes σx,y,z . as a function of beam-beam jitter Δx,y has
# $
Important are the disruption parameters Δ2x Δ2y
L = L0 exp − 2 − 2 (2)
2N re σz 4σx 4σy
Dx,y = (1)
γσx,y (σx + σy ) For limited disruption parameters, the luminosity
where N is the bunch particle number, re the clas- is less sensitive to offsets since the beams attract
sical electron radius, γ the Lorentz energy factor. each other. But for larger vertical disruption pa-
For Dx,y 1 particles close to the beam axis will rameters the collisions become more complex, see
be deflected toward the axis Fig.2. This is due to the development of a two-
x y stream instability during the collision. Such an
x = Dx , y = Dy instability can even occur if the mean offset of the
σz σz
beam particles is zero but if the different longi-
If D 1 the particle will move transversely dur- tudinal slices have offsets. In this case the pro-
ing the passage through the oncoming bunch and jected beam sizes are not a good measure for the
the description of the beam-beam interaction be- achieved luminosity. By adjusting the mean off-
comes complex. set and angle of the incoming beam slightly the
luminosity can however be recovered [4].
Luminosity enhancement Since the particles
of each beam are focused by the other beam, Multi-bunch kink instability Parasitic cross-
the beam sizes decrease during collision. This ing of the outgoing and incoming bunches can
increases luminosity by the luminosity enhance- lead to a multi-bunch kink instability. If bunches
ment factor HD , which is typically in the range collide with a small vertical offset, they receive a
1–2. While formulae exist that approximate HD strong vertical deflection. After the collision these
as a function of beam parameters, it is generally bunches can have parasitic crossings with the next
calculated using one of the simulation codes. ones of the incoming beams and because of their
now large vertical offset can kick them vertically.
In the likely case that all bunches in the trains have
the same offset Δy0 this effect will increase the
offset at the collision point along the train. In case
of an infinitely long train and small offset this will
lead to a maximum offset Δy,
Figure 1: Conceptual sketch of the crab crossing
Δy0
Δy = re ∂y
(3)
scheme. 1 − nc 4N
γθ 2 ∂Δy
c
176
Ch.2: BEAM DYNAMICS
Here, nc is the number of parasitic crossings be- and the mean field strength E + B, Υ can be
fore the beams are separated, θc is the full cross- expressed as
ing angle and ∂y /∂Δy is the deflection at the IP. E + B 5 re2 γN
To limit the offset amplification to a factor two Υ=γ ≈ (9)
Bc 6 ασz (σx + σy )
one requires Υ is a measure of the field strength in the elec-
4N re ∂y 1 tron’s rest frame, i.e. γE + B, in units of Bc .
nc < (4) As Bc signals the onset of nonlinear QED effects,
γθc2 ∂Δy 2
Υ 1 corresponds to the classical limit, while
The slope ∂y /∂Δy of beam-beam deflection as Υ 1 the deep quantum regime, of the beam-
a function of offset can be obtained from simula- strahlung effects.
tion. In order to obtain a first order of magnitude The fact that the beam field is not uniform
expression, one can simplify a formula in [19], across the beam indicates that the mean field
which is valid for rigid beams (i.e. small disrup- strength introduced above may not be sufficient
tion parameters), small vertical and no horizontal in describing certain beamstrahlung phenomena,
offset: and a local definition is necessary. The maximum
∂y 1 Dy beam field corresponds to a maximum Υmax ,
≈ (5)
∂Δy 2 σz which is related to the mean value by
12
This leads to the requirement: Υmax ≈ Υ (10)
5
σx2 1 4N 2 re2 1 1 In principle, all beamstrahlung effects should be
nc Dx Dy ≈ nc 2 < (6)
2 2
σz θ c γ σx σy θc2 2 calculated locally, and integrated over the beam
cross section. However, the global beamstrahlung
In case of high disruption the vertical angles tend
parameter, Eq.(7), often suffices, at least to under-
to be smaller than indicated, so the effect will be
stand the fundamental scaling of effects.
smaller. Note that significant production of co-
Number of photons and average energy loss
herent pairs can lead to a stronger parasitic de-
The photon emission is described by the Sokolov-
flection of the beams, since the coherent pairs are
Ternov spectrum ⎡∞ ⎤
deflected more than the beam particles due to their

low energy. ∂ ẇ α ⎣ ω ω
=√ K 5 (x )dx + K 2 (x)⎦
∂ω 3πγ 2 3 E E −ω 3
2.5.3.2 Beamstrahlung x
The bending of the trajectories during beam-beam Here, x = ωωc E−ω

E
and K5/3 and K2/3 are modi-
interaction leads to production of beamstrahlung, fied Bessel functions.
which is photon emission similar to synchrotron The most important physical parameter in
radiation. In typical linear collider cases the num- beamstrahlung is the average number of photons
ber of photons emitted per beam particle is O(1), emitted per electron during the collision,
α2 σ
so one has to consider the quantum nature of the z
nγ ≈ 2.54 Υ U0 (Υ) (11)
process. The emitting particle will collide fur- re γ
ther, contributing to the luminosity at a lower than 1
U0 (Υ) ≈
nominal center-of-mass energy. [1 + Υ2/3 ]1/2
Beamstrahlung parameter All beamstrahlung The average fractional energy loss due to
effects can be described by the Lorentz-invariant, beamstrahlung is
dimensionless beamstrahlung parameter Υ, de- α2 σ
z
fined as δB ≈ 1.24 Υ ΥU1 (Υ) (12)
re γ
e 1
Υ ≡ 3 4 (pμ F μλ pν Fλν )1/2 (7) U1 (Υ) ≈
me c [1 + (1.5Υ)2/3 ]2
where pμ is the four-momentum of the particle The average
√ energy of all % the emitted photons is
and Fμν is the mean energy-momentum stress ten- 4 3 U1 (Υ) 0.462Υ (Υ → 0)
δγ = Υ ≈
sor of the beam field. In terms of the Schwinger 15 U0 (Υ) 0.254 (Υ → ∞)
critical field, (13)
m2 c3 Note that δγ is finite (about 1/4 of the initial elec-
Bc ≡ ≈ 4.4 × 1013 Gauss (8) tron energy) when Υ → ∞.
e
177
Luminosity vs beamstrahlung Beamstrahlung Vertical beam size and luminosity Limitations

is one factor that limits luminosity in linear col- for σy can be of different nature. One limitation
liders. The luminosity spectrum affects the qual- is the hour glass effect. If βy < σz , the beam
ity of the physics event analysis in the detector, is focused to a small size but only over a small
hence one limits nγ . The luminosity spectrum is length, which reduces luminosity. Hence one re-
also degraded by initial state radiation, which is quires βy ≈ σz . Applying this limit the luminos-
the emission of photons as a part of the physics ity can be rewritten for the classical case as
process. One usually tries to limit the spectrum
nγ nγ Eγ
degradation due to beamstrahlung to the level of L∝ √ ηPwall ∝ √ ηPwall
degradation due to initial state radiation. This typ- y σz y
ically leads to a required nγ = O(1). In the case of short bunches βy can be limited
Flat beams allow to maximize the luminosity, by the design of the final focus system. Or the
which is ∝ 1/(σx σy ) and at the same time limit beam-beam jitter resulting from dynamic effects
the nγ , which ∝ 1/(σx + σy ) in the classical case. may pose a lower limit on beam size.
In the following, we consider the classical case Depolarization There are two major mecha-
U0 ≈ U1 ≈ 1 and flat beams σx σy . The total nisms that cause the spin depolarization of the
luminosity is e− and e+ beams during beam-beam interac-
N2 tion. One is the classical precession under the
L = HD nb fr beam field described by the Thomas-BMT equa-
4πσx σy tion (Sec.2.6.1). The other is the spin-flip, or the
with nb the number of bunches per train and fr Sokolov-Ternov, effect (Sec.2.6.6) upon the emis-
the repetition rate for the trains. Neglecting HD sion of beamstrahlung photons.
and using σx + σy ≈ σx , we find The average final depolarization due to the
nγ Thomas-BMT effect is [17]
L∝
σy
ηPwall 3 nγ 2
ΔP BM T = (14)
50π 2 U0 (Υ)
where Pwall is the total power consumption of the
collider and η the efficiency of transferring this while that due to Sokolov-Ternov effect is [17]
power into beam power. The average energy of 7
ΔP ST ≈ Υ2 nγ (1 < ∼Υ < ∼ 100) (15)
the photons is given by 12
1 The net average depolarization is then the sum
Eγ ∝ nγ of the two contributions. It is also important to
σz
estimate the luminosity weighted depolarization,
The most relevant property of the luminosity [ΔP ], which is related to ΔP by a simple nu-
spectrum is usually the fraction of luminosity in merical factor,
the peak close to the nominal center-of-mass en-
ergy. This peak depends on nγ and to some extent [ΔP ] = 0.273ΔP (16)
on Eγ . Usually the relevant luminosity spectrum
is derived by simulation. 2.5.3.3 Background and spent beam
Luminosity vs. beamstrahlung in the non- Beam-beam effects are an important source for
classical case In this case the radiation is sup- background in the detectors.
pressed to some extent since the critical energy is Disruption angles The beam-beam disruption
above the beam energy. For Υ 1, one can ap- angles of the particle are important to define the
proximate U0 (Υ) with Υ−1/3 . This leads to the apertures of the post collision line. The angle is
expression characterized by the nominal deflecting angle
√
0.024 3/2 γ 1 2re N Dx σx Dy σy
L≈ n √ N nb fr θ0 ≡ = = (17)
α2 γ σz re σy γ(σx + σy ) σz σz
In this case a shorter bunch suppresses the beam- For a flat beam one can estimate the maximum
strahlung, which allows to reduce σx to have the deflection in the horizontal plane. In the limit of
same number of beamstrahlung photons. Con- large β-functions and small disruption parameter
sequently, the luminosity can be increased for one finds
shorter bunches. θx,max ≈ 0.77 θ0
178
Ch.2: BEAM DYNAMICS
which is reached by particles at x ≈ 1.3 σx . In the it is focused in the same way by the positrons. Or
vertical direction the maximum field is reached at it can fly in the direction of the positrons, in which
y ≈ 0.0065 σx with case it will be deflected to larger angles by the
electron beam. The detector has to be protected
θy,max ≈ 1.25 θ0
from these pairs by carefully choosing the aper-
For head-on collision the number of particles that tures for the spent beam, which can impact the
experience the full deflection in the vertical plane choice of crossing angle between the two beam
can be negligible. But collisions with a beamlines, and by introducing masking systems. Both
beam offset can lead to a number of particles beam-beam codes allow the simulation of coher-
having large angles. Further note that particles ent pairs from beamstrahlung, GUINEA-PIG also
may be deflected to larger angles if they radi- includes the trident cascade process.
ate beamstrahlung and that the initial beam diver- Incoherent pairs Two colliding photons can
gence needs to also be taken into account. In ad- produce an e+ e− pair. These incoherent pairs
dition secondary particles, in particular coherent have an important impact on the detector design.
and incoherent pairs, discussed below, can be im- Most of the pair particles are produced with low
portant for the post collision line aperture require- energies and at small angles, but some can have
ments. larger angles at production. The particles are de-
Coherent Pairs A photon propagating through flected by the beam fields in the same fashion as
a transverse EM field has a probability of turn- the coherent pairs and can hit the vertex detec-
ing into e+ e− pairs. This process has been called tor and produce background that prevents proper
coherent pair creation [20]. Such a process can event reconstruction. This is avoided by the use
be contributed by either the real, beamstrahlung, of a high solenoid field in the detector which con-
photons, or the virtual photons comoving with the fines the particles and by use of a large enough
high energy particle. For beamstrahlung coherent radius of the innermost layer of the vertex de-
pair creation, the average number of e+ e− pairs tector, which results in a limited vertex resolu-
created per primary electron is tion. Beam-beam codes allow to study the gen-
ασ 2 eration of incoherent pairs and their deflection by
z
nbe+ e− = Υ Ξ(Υ) the beams.
λ̄c γ
% Hadronic events The collision of two photons
(7/128) exp(−16/3Υ), (Υ < ∼ 1)
Ξ(Υ) = can also produce hadrons. The cross section for
0.295Υ−2/3 (ln Υ−2.488)(Υ 1)
this is significantly lower than for incoherent pair
The energy spectrum of the coherent pairs can be production, but the produced particles can im-
approximated by (when Υ < ∼ 10) pact the reconstruction of the physics events. Dif-
dnbe+ e− ασ 2 1 − x ferent parametrizations of the cross sections for
z hadronic events exist. For given beam parame-
= 0.2 Υ
dx λ̄c γ Υx ters, the rates can be calculated with GUINEA-

2 1−x 4 PIG, which also produces the information neces-
× exp − + (18)
3Υ x 1−x sary to generate them for further analysis.
When Υ 1, the contribution from virtual pho-
tons becomes significant. The virtual process is References
sometimes called trident cascade [20], [1] P. Chen et al, NIM A 355, 107 (1995)
ασ
nve+ e− =
z
Υ Ω(Υ) [2] D. Schulte, DESY-TESLA-97-08 (1996)
λ̄c γ [3] D. Schulte et al, PAC07-THPMN010
Ω(Υ) ≈ 0.23 α ln Υ (Υ 1) [4] R. Brinkmann, O. Napoly, D. Schulte, DESY-
TESLA-2001-16, CLIC-Note-505
In the multi-TeV regime, coherent pair production [5] D. Schulte, CLIC-Note-560
can lead to significant pair production. The num- [6] R. Hollebeek, NIM 184 (1981) 333
ber of pairs can be a significant fraction of the [7] P. Chen, K. Yokoya, PRD 38 (1988) 987
number of beam particles. Particles from coherent [8] R.J. Noble, NIM A256 (1987) 427
pairs have significantly lower energies than the [9] M. Bell, J.S. Bell, PA 24 (1988) 1; R. Blanken-
beam particles. The electron of a pair can fly in becler, S.D. Drell, PRL 61 (1988) 2324; P. Chen,
the direction of the electron beam, in which case K. Yokoya, PRL 61 (1988) 1101; M. Jacob, T.T.
179
Wu, Nucl. Phys. B303 (1988) 389; V.N. Baier, orbits within a single chamber, over some part of
V.M. Katkov, V.M. Strakhovenko, Nucl. Phys. the circumference. (Two rings with different cir-
B328 (1989) 387 cumferences and a common section are also pos-
[10] P. Chen, Lecture Notes in Phys. 296, Springer- sible.) Separation at the IP may be required, e.g.,
Verlag (1988) during injection at lower energy.
[11] K. Yokoya, P. Chen, Lecture Notes in Phys. 400, Typical separation schemes If the beam
Springer-Verlag (1992) momentum-to-charge ratios are equal and oppo-
[12] P. Chen, Nucl. Phys. B 51A (1996) 179 site at the IP, p+ /q + = p− /q − (e.g., equal-
[13] K. Yokoya, KEK Report 85-9 (1985) energy e+ e− -collider), orbit separation requires
[14] P. Chen, PAC 93, p.617 transverse electrostatic or time-dependent mag-
[15] K. Yokoya, P. Chen, PAC 89, p.1438
netic fields. Otherwise, static magnetic fields can
[16] P. Chen, PRD 46 (1992) 1186
also separate (e.g., unequal-energy e+ e− -collider,

[17] K. Yokoya, P. Chen, AIP Proc. 187 (1988) p.938
pp-collider). If there are no common sections of
[18] I.F. Ginzburg et al, NIM 205 (1983) 47
[19] O. Napoly, B. Zotter. CLIC-Note 289 (1995) the orbits (e.g., two-ring collider with crossing an-
[20] P. Chen, in Research Directions of the Decade, gle) a separating field may not be necessary.
Snowmass, World Scientific (1990); P. Chen, V. Near an IP, design of a separation scheme is
Telnov, PRL 63 (1990) 1796 intimately connected with interaction region de-
sign [1, 2, 3, 4, 5, 6]; see Figs.1-2 for some typical
schemes. Schemes with crossing angles can al-
2.5.4 Parasitic Beam-Beam Effects and low smaller bunch spacings than those with head-
Separation Schemes on collisions and are an essential feature of most
J.M. Jowett, CERN modern high-luminosity colliders.
2.5.4.1 Separation schemes
IP 5
Encounter points Counter-rotating (+,−)
bunches of equal speed, starting at azimuths
s+ , s− , at a given initial time, in a collider of IP 4 IP 6
circumference C, encounter each other (and may
collide) at two azimuths separated by C/2,

+ − C+s+ +s−
(ŝ, s̃) = s +s
2 , 2
(mod C) (1)
All encounters between beams of ar- IP 3 IP 7
bitrary bunch structure (filling scheme)
− −
(s+ +
1 , ...sM + ), (s1 , ...sM − ) are given by a
matrix of such pairs
⎡ ⎤
(ŝ, s̃)11 · · · (ŝ, s̃)1j · · · (ŝ, s̃)1M −
IP 2 IP 8
⎢ .
.. .. .
.. .
.. ⎥
⎢ . ⎥
⎢ ⎥
⎢ (ŝ, s̃)i1 · · · (ŝ, s̃)ij · · · (ŝ, s̃)iM − ⎥ (2)
⎢ .. .. .. ⎥ e+ IP 1 e-
⎣ .
..
. . . ⎦
(ŝ, s̃)M + 1 · · ·(ŝ, s̃)M + j · · ·(ŝ, s̃)M + M − Figure 1: Horizontal orbit of one beam in a “pret-
zel” scheme in equal-energy e+ e− -collider LEP with
The 2M − encounters experienced by bunch i of many potential locations for parasitic encounters in the
the “+” beam are listed in row i of the matrix. The arcs where the beams are separated in a single cham-
s±i differ amongst themselves by multiples of the
ber. Head-on collisions occur at the even-numbered
rf wavelength λrf and the ŝij , s̃ij by λrf /2. Un- IPs. Anti-symmetry about the IPs helps to equalize
equal speeds or circumference can lead to moving the global optical parameters between the two beams.
encounter points. (Courtesy J.-P. Deluen)
Unwanted collisions can be suppressed en-
tirely, by separating the beams into two separate Side-effects Separation schemes which involve
vacuum chambers at some distance from the inter- displacements of the ideal orbit from a nomi-
action point (IP), and/or partly, by separating the nal orbit (typically passing through the centers of
180
Ch.2: BEAM DYNAMICS
2.5.4.2 Long-range beam-beam effects
Beam-beam kick Consider the beam-beam

force on test particle (or ion) (mass m, charge
Ze) at position x = (x, y, z) due to the charge
distribution, ρ(x), of the opposing beam. (The
coherent behavior of beams must be obtained by
summing these forces over the beam and is not
amenable to straightforward analytical treatment.)
The potential [7] is

−1 ρ(x )
Φ(x) = dx
4π0 |x − x |
Figure 2: Differential bending separation by mag- ∞

net (B1), horizontal plane of unequal-energy e+ e− - −1 dt
−|
x− x |2
= √ ρ(
x )e t dx
collider PEP-II. (Courtesy M. Sullivan) 4π0 π 0 t3/2
∞
−N e
= √ dt (4)
4π0 π 0
2 2 2

exp − 2σx2 +t − 2σy2 +t − 2σz2 +t
× x y z
(2σx2 + t)(2σy2 + t)(2σz2 + t)
where the last form holds when ρ(x) is a 3-D

gaussian with rms σx,y,z and normalized to N e.
In the short bunch limit, the components of
the beam-beam kick
∞
Ze
Δpx,y = − ∂x,y Φ(x)dz (5)
p0 c −∞
are conveniently computed in complex form [8,

Figure 3: 7σ beam envelopes in 2D separation around
9] iΔpx + Δpy = − ZNγ r0 Π(x, y, σx , σy ), where
ALICE IP in equal-energy pp-collider LHC at 450
GeV; yellow planes are beam-beam encounters; hori- r0 = e2 /(4π0 mc2 ), Π(x, y, a, b) =
zontal separation at IP for injection, vertical separation '
⎧
at other encounters from crossing angle. ⎪
⎪
⎪ 2π √ x+iy
⎪
⎪ a2 −b2
w
⎪
⎪
2(a2 −b2 )
⎪
⎪
⎪ ;
magnets) require detailed evaluation of changes ⎪
⎪ x2 y2 xb
+i ya
⎨ − e− 2a2 − 2b2 w √ a 2 b 2 , (a > b, y ≥ 0)
to the optical functions, tunes, etc. Moreover 2(a −b )
these may differ between beams. E.g., in a pret- ⎪
⎪ −Π∗ (x, −y,a, b),
⎪
⎪ (a > b, y < 0)
zel scheme with horizontal separation xp (s) in a ⎪
⎪ ix+y
⎪
2 2
⎪ − x +y
FODO lattice the perturbation of the dispersion ⎪
⎪ 2 1 − e a 2 , (a = b)
function Dx0 → Dx = Dx0 + Dx1 gives rise to a ⎪
⎪
x2 +y 2
⎩ ∗
number of effects including a change in the longi- iΠ (y, x, b, a), (a < b)
tudinal damping partition number (Sec.3.1.4) for (6)
e+ e− beams, [See also Sec.2.5.1.] Additional terms must be
added [10] to make the full 6-D map symplectic.
For non-Gaussian charge distributions see [11].
dU (δ) 2
Js ≈ = Js0 + K 2 xp Dx1 ds (3) See [12] and Sec.2.5.1.2 for cases where the beam
dδ I2
is thick longitudinally. Figs.4-5 show the behav-
where K is the quad gradient. ior of the beam-beam kick.
181
0.01 0.01
2 y / mm y 0 y 0.1 mm
0.008 0.008
0.006 0.006
[y
0.004 U 0.004
U
[y
0.002 0.002
x / mm x / mm
1
0 [x 0
[x
-4 -2 0 2 4 -4 -2 0 2 4
0.01 0.01
[y x 0 x 1 mm
0.008 0.008
0 0.006 0.006 [y
0.004 U 0.004
U
0.002
[x y / mm 0.002
[x y / mm
0 0
-4 -2 0 2 4 -4 -2 0 2 4
-1
Figure 6: Parasitic beam-beam tune-shifts for a parti-

cle on the closed orbit, calculated perturbatively, for
-2 x / mm the same conditions as Fig.5 with βx = βy = 1 m.

-2 -1 0 1 2 The charge density is also shown.
Figure 4: Kick vector field where the length (and

Beam-beam tune-shift The beam-beam inter-
color) of the arrows indicates the kick strength. Gray
shading indicates the charge density. Same parameters
action produces tune-shifts
βx,y
as Fig.5. ξx,y = − ∂x,y Δpx,y (8)
4π
where (x, y) is the position of the closed orbit
20
p x y0 20
p x y 0.1 mm (center) of the weak beam relative to the opposing
10 7 10 7 beam. In the case of a Gaussian beam, the change
0
p y
x / mm 0
x / mm of variables t = u/(1 − u) transforms Eq.(4)
-10 -10 p y into a form convenient for numerical evaluation
-20 -20 (Gaussian case can be evaluated analytically from
-4 -2 0 2 4 -4 -2 0 2 4
Eq.(6)). Usually it is best to integrate separately
x0 x 1 mm
over the neighborhoods of u = 0 and u = 1. Ex-
20 20
p y p y
10 10
7 y / mm 7 y / mm amples are shown in Fig.6. For large separations,
0 0
p x Eqs.(7) and (8) give
-10 -10
p x ZN r0 (βx , −βy )(x2 − y 2 )
(ξx , ξy ) =
-20 -20
(9)
-4 -2 0 2 4 -4 -2 0 2 4
2πγ (x2 + y 2 )2
Figure 5: Beam-beam kicks in μrad felt by a 1 GeV, The distribution of tune-shifts with betatron
Z = −1, particle (e.g., electron) as a function of hor- amplitudes inside a bunch can be derived analyti-
izontal (left) and vertical (right) separations around a cally, extending Eqs.(21-22), Sec.2.5.1 in the case
bunch with N = 1010 , σx = 1 mm, σy = 0.2 mm of a single encounter [14, 15]. Beyond this, sim-
(e.g., positron bunch). In the plots on the left, the other ulation techniques [13, 16, 5] are generally more
separation is zero, while on the right a non-zero sepa- practical.
ration is included. Each plot has the same scales and Systems of many bunches Since different
the (scaled) Gaussian charge density is also shown. bunches of each beam may see different se-
quences of encounters Eq.(2), there may be differ-
ent orbits, optical functions, tunes and chromatic-
At large separations, min(x, y) ities for each bunch. Bunch pairs of bunches may
max(σx , σy ), the beam looks like a point see different separations, sizes and optical func-
charge, the kick is purely radial and Eq.(6) can be tions at the same encounter point. In practice, a
approximated as self-consistent solution for the closed orbit and
2ZN r0 (x, y) optics of a two-beam many-bunch system has to
(Δpx , Δpy ) = (7) be found [19, 2] by simulation.
γ x2 + y 2 A weak-strong [17, 18, 2, 13] simulation
This formula is useful in preliminary design of a treats a single test particle in the fields of an op-
separation scheme. posing beam, regarded as a fixed charge distri-
bution. Beam-beam encounters are simulated as
182
Ch.2: BEAM DYNAMICS
special optical elements, using the kicks (6) with [10] K. Hirata, H. Moshammer, F. Ruggiero, PA 40
fixed ρ(x) with maps to propagate the orbit and (1993) 205
optics between beam-beam encounters. The op- [11] V. Ziemann, SLAC-PUB-5582 (1991)
tics and tune of the test particle can be found [12] J. Irwin et al, PAC 95, p.2904
as functions of its amplitude with respect to the [13] D. Shatilov, PA 52 (1996) 65
closed orbit, giving tune footprints for the weak [14] T. Sen et al, PRST-AB 7 (2004) 041001
beam. A strong-strong simulation [5] computes [15] M. Meddahi, R. Schmidt, CERN SL/90-15
mutual effects of the two beams. Some bunches (1990)
in the system may be unstable and be lost, chang- [16] W. Herr, PA 50 (1995) 69
ing the stability of others which may subsequently [17] Proc. 7th ICFA Beam Dynamics Workshop
(1995), JINR Report 1996
be lost (e.g. the “PACMAN” effect) [20].
[18] A. Temnykh, D. Sagan, PAC 97, p.1768
Simple Criteria for adequate separation [19] E. Keil, CERN SL/95-75 (1995)
Present understanding provides no universal [20] D. Neuffer, S. Peggs, SSC-63 (1986)
criteria for the adequacy (beam lifetime, tolerable [21] D. Siergiej et al, PRE 55 (1997)
effects on beam sizes, etc.) of a separation [22] Y. Papaphilippou et al, PRST-AB 5 (2002)
scheme. Simple criteria [1] can be used in a 074001
design phase, pending thorough simulation, [23] E.G. Stern et al, PSRT-AB 13 (2010) 024401
analysis or experiment. If Xi is the separation [24] M.A. Furman et al PAC 85, p.2297
at encounter i (head-on excluded), σi a relevant
beam size (e.g. max(σx , σy )), then some of the
simplest criteria are 2.6 POLARIZATION
• Xi > nσ σi , where nσ ≈ 5.5–10
2.6.1 Thomas-BMT Equation
• ξxi , ξyi < ξmax , where ξmax ≈ 10−4 –10−3 T. Roser, BNL

• ξxi , ξyi < ξmax , where ξmax ≈ 10−3
i i Precession of polarization vector P of a particle
Applicability depends on the nature of the beam with mass m and charge Ze is given by [1]-[4]
(generally stricter for protons than electrons) and dP 0 × P
=Ω (1)
other conditions. Experiments have shown that dt
none of these criteria is strictly necessary or suffi- 0 = − Ze (1 + Gγ)B
Ω ⊥ + (1 + G)B
cient [3]. In hadron colliders, a strategy of mini- mγ
mizing the tune-spread in all the bunches is often
γ E × v
followed [16, 3], possibly involving the compen- + Gγ +
sation of tune-shifts between different parasitic γ+1 c2
encounters. Parasitic encounters may also induce P is defined in the particle rest frame, E and B
amplitude-dependent chromaticity [14], drive ad- in the laboratory frame. B = B ⊥ + B , B
=
ditional resonances and diffusion, coherent ef- (v · B)
v /v 2 .
fects, affect dynamic aperture, lifetime, combine It is instructive to compare with the Lorentz
with impedance effects, etc [21, 22, 24, 23]. force equation (E = 0)
dv c × v
References =Ω
dt # $
2 × v
[1] A.B. Temnykh et al, PAC 93, p.2007 c = − Ze B
Ω ⊥ + γ E
(2)
[2] LHC99, CERN-SL-99-039 AP (1999) mγ γ 2 − 1 c2
[3] V. Shiltsev et al, PRST-AB 8 (2005) 101001
In a frame rotating with v :
[4] M. Sullivan et al, EPAC 04 MOPLT141 (2004)
for B̃⊥ = 0 :
[5] M. Tawada et al, PAC 05 TPPP006 (2005)
[6] R. Bailey et al, PAC 93, p.2013 dP Ze
[7] K. Takayama, Lett. Al Nuovo Cimento 34 (1982)
=− Gγ B ⊥ × P (3)
dt mγ
190
[8] M. Bassetti, G.A. Erskine, CERN-ISR-TH/80-06
for Ẽ⊥ = 0 :

(1980) dP Ze γ E × v
=− Gγ − 2 ×P

[9] R. Talman, AIP Proc. 153 (1987) p.827 dt mγ γ −1 c2
183
Sec.2.6: POLARIZATION
G = g−22 is the anomalous magnetic moment; Thomas-BMT equation can be written in the
μ spinor or unitary representation,
g = 2m
Ze J is the gyromagnetic ratio.
( 2( ( ZeG ( rad
( mc ( ( ( dP
G ( G ( [GeV] mc Tm =Ω × P
dt
i
e 0.00115965 0.440649 0.680342
dψ
μ 0.00116592 90.6220 0.00330816 → = − σ · Ω ψ (3)
p 1.79285 0.523341 0.572843 dt 2
d σ · P i
d -0.142987 13.1173 0.0228547
3
He -4.18396 0.671228 0.893266 or =− σ · Ω , σ · P
3
H 7.91819 0.354744 0.845096 dt 2
Solution for constant Ω (Axis n̂, |Ω|
= ω) is
Note:
(i) In the rest frame, ψ (t) = M (n̂, ωt) ψ (0) , or (4)

0 = − Zeg B = − Ze (1 + G) B.
Ω †
σ · P (t) = M (n̂, ωt) σ · P (0) M (n̂, ωt)

2m m
(ii) For B = 0 and E⊥ = 0 or γ 1, one has

(Ω0 − Ωc ) /Ωc = Gγ. This is the spin tune in this where the rotation operation M (angle ϕ = ωt
around axis n̂) is
special case. ϕ
(iii) For the negative hydrogen ion the magnetic M (n̂, ϕ) = exp −i (σ · n̂) (5)
moment is given by the proton and the charge by ϕ 2 ϕ
the extra electron. For B = 0 and E ⊥ = 0, one = cos − i (σ · n̂) sin
2 2
therefore has (Ω0 − Ωc ) /Ωc = − (2 + Gγ).
Conversely, given the rotation operator M , the ro-
Ω0 − Ωc is in-
(iv) The “magic” energy for which
tation angle and axis can be found by the inverse

dependent of E is given by γ = 1 + 1 . operation,
G
ϕ 1
References cos = tr (M ) (6)
2 2
[1] B.W. Montague, Phys. Rep. 113 (1984) 1 i
n̂ = tr (σ M ) (7)
[2] L.H. Thomas, Phil. Mag. 3 (1927) 1 2 sin ϕ2
[3] V. Bargmann, L. Michel, V.L. Telegdi, PRL 2
The net spin rotation for one turn is given by
(1959) 435
[4] S.Y. Lee, Spin Dynamics and Snakes in Syn-
the one-turn matrix
chrotrons, World Scientific (1997) M0 (θ) = Mn · · · M2 M1 (8)
where θ is the starting (and ending) azimuth. The
2.6.2 Spinor Algebra spin tune νsp and the spin closed orbit n̂0 (also
T. Roser, BNL called n̂0 axis or stable spin direction) are
1
The coordinate frame is specified by the indices: cos (πνsp ) = tr (M0 (θ)) (9)
(1,2,3) = (radial outward, longitudinal forward, 2
vertical up) = (x̂, ŝ, ŷ). The Pauli matrices independent of θ
σ = (σ1 , σ2 , σ3 )
i/2
n̂0 (θ) = tr (σ M0 (θ)) (10)
01 0 −i 1 0 sin (πνsp )
= , ,
10 i 0 0 −1 With a preferred axis of symmetry, usually
satisfy the following properties: vertical,

σ1 σ1 = σ2 σ2 = σ3 σ3 = I 1 01
σ+ = (σ1 + iσ2 ) =
σ3 σ1 = −σ1 σ3 = iσ2 (cyclic perm.) 2 00

tr(σi ) = 0, det(σ i ) =−1 1
σ− = (σ1 − iσ2 ) =
00
(11)
(σ · a) σ · b = a · b + iσ · a × b
(1) 2 10
The normalized vector P has spinor representa- are a more useful representation. These matrices
tion obey the following relations:
P = ψ †σ ψ σ+ σ+ = σ− σ− = 0

1 1 + P3 σ± σ3 = ∓σ±
↔ψ= (2)
2 (P3 + 1) P1 + iP2 σ3 σ± = ±σ± (12)
184
Ch.2: BEAM DYNAMICS
and the following transformations: Full Siberian snake: rotates P by 180◦ (ϕ = π)

ϕ ϕ around an axis in the horizontal plane with angle
exp i σ3 σ± = exp ±i σ± (13) α from x̂ (Snake axis angle) [2]:
2 2
ϕ ϕ
σ± exp i σ3 = exp ∓i σ± MSnake = −i (σ1 cos α + σ2 sin α) (5)
ϕ 2 ϕ 2
exp i σ3 σ± exp −i σ3 = e±iϕ σ± Type 1 snake: snake axis is long. (α = 90◦ );
2 2 Type 2 snake: snake axis is radial (α = 0◦ ) Note:
MDipole MSnake MDipole = MSnake .
2.6.3 Spin Rotators and Siberian Snakes Examples of one-turn matrices for rings:
T. Roser, BNL
“Spin Rotators” are devices that rotate P , prefer- 1. Ideal ring without spin rotators or snakes:
ably without changing v .
M0 (θ) = exp (−iσ3 πGγ) , νsp = Gγ (6)
Examples:
Wien Filter: Transverse Ex and By with condition
v 2 2. Figure-8 ring without spin rotators or snakes:
E×
c2
= 1−γ
γ2
⊥ , the spin rotation is described by
B

Ze (1 + G) M0 (θ) = 1, νsp = 0
ϕ= B⊥ ds (7)
mcβγ 2
ϕ ϕ
MWien = cos − iσ3 sin (1) 3. Ring with solenoid (partial type 1 Siberian
2 2 snake) at θ0 = 0 and spin rotation ϕ [4]:
Solenoid: ϕ
M0 (θ) = cos cos (πGγ)
Ze (1 + G) 2
ϕ= B ds ϕ
mcβγ − iσ1 sin sin ((π − θ) Gγ)
ϕ ϕ 2
MSolenoid = cos − iσ2 sin (2) ϕ
2 2 − iσ2 sin cos ((π − θ) Gγ)
2
Example:
ϕ = 90◦ and p = 1 GeV/c requires ϕ
B ds = 1.88 T-m for protons and B ds = − iσ3 cos sin (πGγ) (8)
2
5.23 T-m for electrons. 1 ϕ
νsp = cos−1 cos cos (πGγ)

Dipole: In lab frame: π 2

Ze (1 + Gγ) 4. Ring with full Siberian snake with axis angle
ϕ= By ds
mcβγ α at θ0 = 0:
In frame rotating with v : M0 (θ) = −i[σ1 cos(α − (π − θ)Gγ)

ZeG + σ2 sin(α − (π − θ)Gγ)] (9)
ϕ= By ds
mcβ 1
ϕ ϕ νsp =
MDipole = cos − iσ3 sin (3) 2
2 2
Note: with a type 1 snake (α = 90◦ ) at θ =
Example: ϕ = 90◦ and β ≈ 1 requires By ds = 0, n̂0 (θ = π) = ŝ, i.e. the beam polarization
2.74 T-m for protons and By ds = 2.31 T-m for is longitudinal.
electrons.
Full twist helical dipole: [1] 5. Ring with two full Siberian snakes with axis

(s) = B0 sin 2πs , 0, cos 2πs , λ > s > 0:
B angles αa and αb at θa and θb :
λ λ

M0 (θ) = − exp(−iσ3 χ) (10)
ϕ = 2π 1 + χ2 − 1
χ ≡ αb − αa + (π − θb + θa )Gγ
ϕ ϕ
M = cos − i(σ2 + χσ3 ) sin (4)
Note: M0 is energy independent for θb −
2 2
1 ZeG B0 λ θa = π; then νsp = π1 (αb − αa ). For
χ≡ 1+ (αb − αa ) = π2 then νsp = 12 .
Gγ mc 2π
185
6. Ring with N pairs of full Siberian snakes For an isolated resonance, ξ =

with axis angles αia and αib at θai and θbi : K exp (−iKθ). In a frame rotating around
ŷ with tune K,
M0 (θ) = (−)N exp(−iσ3 χ) (11)
N i i
ψK = exp Kθσ3 ψ
where χ = i=1 αb − αa +
i
2
N
π − i=1 θb − θa Gγ.
i i
dψK i Gγ − K −K
=− ∗ ψ (3)
Note: M is energy independent dθ 2 −K K − Gγ K
N 0i
for θ − θai = π; then with solution (for small betatron amplitudes, n̂0
N i
i=1 b
νsp = π i=1 αb − αia .
1
There are and νsp are evaluated along the closed-orbit),
many parameter sets that result in νsp = 12 .
i
See [5, 6] for parameter sets that minimize ψK (θ) = exp − νsp θ (n̂0 · σ ) ψK (0)
2
the amplitude dependence of the spin tune.
νsp = (Gγ − K)2 + |K |2 (4)

References [(Gγ − K) σ3 − K σ+ − ∗K σ− ]
(n̂0 · σ ) =
[1] V. Ptitsin, Y. Shatunov, NIM A398 (1997) 126 νsp
[2] Ya.S. Derbenev, A.M. Kondratenko, PA 8 (1978) Solutions for a single isolated resonance in a ring
115 with snakes also exist, but more complicated [3].
[3] Operating example at HERA, D.P. Barber et al,
Passage through an isolated resonance with
Phys. Lett. 343B (1995) 436
a linear crossing pattern is described by the
[4] T. Roser, Proc. Workshop on Siberian Snakes and
Froissart-Stora Equation [4],
Depolarizing Techniques (1989) p.1442 # $
[5] G.H. Hoffstätter, High-Energy Polarized Proton Pfinal π |K |2
Beams, Springer (2006) = 2 exp − −1 (5)
Pinitial 2α
[6] S.R. Mane, NIM A587 (2008) 188
with α = d(Gγ)dθ (crossing speed). Fast passage
→ Pfinal ≈ Pinitial . Slow passage → Pfinal ≈
2.6.4 Depolarizing Resonances and Spin
−Pinitial → spin flip. Passage with a piecewise
Flippers
linear zigzag pattern is described in [5].
T. Roser, BNL
Artificial resonance from local oscillating
Thomas-BMT equation with azimuthal coordi- longitudinal field [6, 7] (ω = applied frequency,
nate θ as independent variable and the fields ex- ω0 = revolution frequency):
pressed in terms of the particle coordinates: B = B̂ cos (ωt)

dψ i Gγ −ξ
=− ∗ ψ (1) (1+G) B̂ ds ω
dθ 2 −ξ −Gγ →K = ; K = n+ (6)
4π Bρ ω0
ξ = −ρy (1 + Gγ)
Spin flip by ramping artificial resonance through
y resonance condition with speed α:
−i (1 + Gγ) y − ρ (1 + G)
ρ Kend − Kstart
α= (7)
where ρ is the bending radius. Resonance strength 2πN
is [1] " where N is number of turns during ramp. For
1 more than 99% spin flip,
K = ξ exp (iKθ) dθ (2)
2π 1
Kend − Kstart ≥ 14K ; K ≥ (8)
where the loop integral keeps only the e−iKθ fre- 140N
quency component in ξ. The case K = kP ± νy For local radial oscillating field (AC dipole) reso-
gives intrinsic resonance driven by vertical beta- nance strength is sum of direct contribution
tron motion; P is super periodicity. The case
K = k gives imperfection resonance driven by B⊥ = B̂⊥ cos (ωt)

vertical closed orbit distortions. One way to re- (1+Gγ) B̂⊥ ds ω
duce resonance strength is in [2]. →K = ; K = n+ (9)
4π Bρ ω0
186
Ch.2: BEAM DYNAMICS
plus contribution from oscillating vertical orbit

motion in the lattice quadrupoles. The latter con-
tribution is large near a strong intrinsic resonance
[8] or in high energy rings with many quadrupoles
[9].
In a ring with pairs of Siberian snakes addi-
tional higher order “snake” resonances [10] occur
when the fractional vertical betatron tune
νsp − k 2k − 1 1
Δνy = ± = for νsp =
(2l − 1) 2(2l − 1) 2
(10)
With a single snake or vertical closed orbit distor-

tions snake resonances also occur when
νsp − k 2k − 1 1
Δνy = ± = for νsp = (11)
(2l) 2(2l) 2
Snake resonances are particularly strong at ener-
gies near strong intrinsic resonances of the ring
without snakes.
References Figure 1: The individual depolarizing resonance cor-

rection hardware installed in AGS [5].
[1] E.D. Courant, R. Ruth, BNL-51270 (1980)
[2] A.W. Chao, Ya.S. Derbenev, PA 36 (1991) 25
60
[3] S.R. Mane, NIM A498 (2003) 1
[4] M. Froissart, R. Stora, NIM 7 (1960) 297 18.5 GeV/c Gγ = 9
sin 9
[5] A.W. Chao, PRST-AB 8, 104001 (2005) 40
[6] D.D. Caussyn et al, PRL 73 (1994) 2857
[7] M. Bai et al, PRST-AB 8 (2005) 99001
[8] M. Bai et al, PRL 80 (1998) 4673 20 FWHM = 10
[9] V. Ptitsyn et al, NIM A608 (2009) 225 Set
P (%) –4
[10] S.Y. Lee, S. Tepikian, PRL 56 (1986) 1635
0
2.6.5 Polarized Hadron Beams and Siberian

Snakes –20
–10 0 10
A.D. Krisch, U. Michigan
αk
M.A. Leonova, FNAL
Figure 2: Individually correcting the Gγ = 9 imper-
In 1973, the first high energy polarized proton
fection resonance in AGS [5].
beam was successfully accelerated in the Argonne
ZGS. The depolarizing resonances in the 12 GeV
weak-focusing ZGS were fairly weak; thus, main- AGS has strong depolarizing resonances. As
taining polarization only needed careful orbit con- shown in Fig.1, AGS needed many complex hard-
trol to overcome the imperfection resonances and ware modifications. To maintain the polarization
moderately fast betatron tune jumps to overcome up to 22 GeV, 39 imperfection resonances were
the intrinsic resonances [1]. Polarized protons corrected individually by using 96 small dipoles
were then successfully accelerated to several GeV to form the correct harmonic wave of horizon-
at Saturne [2] and KEK [3]. The plan for higher tal magnetic field as each resonance was passed;
energy polarized protons started at a 1977 work- moreover, 5 intrinsic resonances were jumped
shop [4] two years prior to the 1979 ZGS shut- (only partly successfully) using 10 formidable
down. 22 MW 1.6 μs pulsed quadrupoles. Fig.2 shows
Polarized protons were first accelerated at a typical AGS imperfection resonance correction
AGS in 1984. Maintaining the polarization was curve [5]. Each polarized beam tune-up required
more difficult, since the 30 GeV strong-focusing 3-7 weeks of dedicated AGS operation.
187
This individual resonance correction tech-

0.6
nique was impractical at higher energy, since the
number of imperfection resonances to be crossed 160 MeV
is E/(0.52 GeV). Thus, it became clear that the 0.4 20 % snake
untested Siberian snakes [6] (Secs.2.6.3, 2.6.4)
P
were needed [7] to accelerate polarized protons νx
0.2
3.640
above 30 GeV. 3.700
The first Siberian snake tests were performed 0
at IUCF’s 500-MeV Cooler Ring. The snake was 4.782 4.841
a 2 T-m superconducting solenoid installed in a 6- 4.76 4.8 4.84 4.88 4.92

m straight section. Fig.3 shows data when cross- νy
ing the Gγ = 2 imperfection resonance, with

and without the snake. Without the snake, full Figure 5: Observation of 7 − νy and 1 + νy − νx spin
polarization was achieved only when the imper- resonances at IUCF [11].
fection magnetic fields were exactly corrected;
any small imperfection field destroyed the po-
larization. With the snake, full polarization was 2.1 GeV/c protons
0.8
maintained over the entire range of imperfection
|P|
fields [8]. Fig.4 shows the snake also overcoming
an intrinsic resonance [9]. 0.7
0.6
0 10 20 30 40 50 60
n
Figure 6: Multiple spin flipping at COSY [17, 14].
During 10 adiabatic ramps of a partial

Siberian snake [10], there was no polariza-
tion loss when Gγ was a half-integer (2 12 ), as
suggested by Courant. Fig.5 shows the first
firm evidence for a 2nd -order snake resonance
(Gγ = 1 + νy − νx ) [11]. Experiments were also
performed at IUCF involving: synchrotron side-
band resonances [9]; partial Siberian snakes [12];
Figure 3: Siberian snake overcoming the Gγ = 2 im- snake resonances [13]; and rf-induced spin res-
perfection resonance at IUCF [8]. onances, which can calibrate the beam energy
or flip the spin direction with over 99.9% effi-
ciency [14]. In 2002, the Cooler Ring’s final year,
177 MeV striking and unexpected behavior was found [15]
80 in the first spin manipulation of simultaneously
vector- and tensor-polarized deuterons, which are
spin-1 bosons.
40 Pradial After Cooler Ring’s shutdown, its spin pro-
P (%) Pvert. Snake ON gram was transferred to the COSY Cooler Syn-
Snake
OFF chrotron in Jülich. COSY had earlier accelerated
0
polarized protons to 3 GeV using moderately fast
betatron tune jumps [16].
5.10 5.14 5.18
The new SPIN@COSY team then demon-
νy strated that one could also have proton spin-
flip efficiencies as high as 99.92 ± 0.04% [17],
Figure 4: Siberian snake overcoming the strong intrin- as shown in Fig.6, at energies 10 times that at
sic resonance Gγ = 3 + νy at IUCF [9]. IUCF [14]. They also confirmed [18], with better
188
Ch.2: BEAM DYNAMICS
1 The Chao matrix (CM) formalism [19] al-

(PV, PT) lows analytic calculations of a beam’s polariza-
0.5 (-2/3, 0)
tion behavior inside a spin resonance, where the
PV / PVi
(-1/3, -1)
0 (-1, +1) Froissart-Stora formula [20] is not applicable.
(+1, +1)
-0.5
The formalism’s prediction of polarization oscil-
lations when crossing an isolated spin resonance
-1 were tested at COSY [21]. The good fits to the
1 experiment’s precise data (Fig.8) demonstrate the
formalism’s ability to obtain experimental values
PT / PTi
0.5 for a spin resonance’s position, its strength, and its

0 frequency spread due to beam momentum spread.
Kondratenko et al [22] proposed a new tech-

-0.5
nique for overcoming medium-strength depolar-
0 10 20 30 40 50
Δt (s) izing resonances when Siberian snakes are not
practical and Fast Crossing (FC) is too weak.
Figure 7: Spin-manipulating deuterons’ vector and Each resonance is crossed using the Kondratenko
tensor polarizations at COSY [18, 15]. Crossing (KC) pattern (Fig.9). KC was tested at
COSY with deuterons using an rf-induced spin
resonance [23]. KC reduced the depolarization
1.85 GeV/c 25 s e-cooling far better than FC at the same fast crossing rate,
0.6 Δt = 100 ms
Δf = 400 Hz
by factors of 4.7 ± 0.3 and 19+12
−5 for unbunched
0.4 ε = 1.06 x 10 -5 and bunched beams, respectively.
(+1, +1) χ2/(N-2) = 3.5
0.2 (-1/3, -1) χ2/(N-2) = 1.1
The SPIN Collaboration also produced de-
(-2/3, 0) χ2/(N-2) = 1.1 tailed plans for accelerating polarized protons
PV
0
(-1, +1) χ2/(N-2) = 4.4 at Fermilab (6 snakes in Tevatron and 2 in the
-0.2
120-GeV Main Injector) [24-a], in the 920-GeV
-0.4
fr = 916 985.3 +
_ 0.5 Hz
HERA proton ring (4 snakes) [24-b] and in the 20-
-0.6 δ fΔ p = 23 +_ 1 Hz TeV SSC (26 snakes in each 20-TeV ring) [24-c].
These projects were not implemented.
916.9 917.0 917.1 Brookhaven’s RHIC-SPIN Collaboration in-
fend (kHz) stalled 2 Siberian snakes in each RHIC ring,
with 25 GeV/c polarized protons injected from
Figure 8: Measured deuteron vector polarizations vs. AGS [25, 26] (Fig.10). With help from RIKEN
rf-solenoid end frequency, with CM fits [21]. in Japan, this 100 GeV/c on 100 GeV/c polarized
p-p collider started operating in 2000.
KC FC f In the mid-1990s, polarized beam accelera-
fast slope link slope
tion again started at AGS to prepare for inject-
ing polarized protons into RHIC. However, the
Δffast fKC
Δfslow AGS problems discussed earlier have remained;
slow slope t
Δfgap
its 2-25 GeV/c range is too low for full Siberian
Δt fast snakes to be practical and too high for the reso-
nance jumping techniques to work efficiently. To
Δt slow
help solve this problem another workshop was
Figure 9: Kondratenko Crossing (KC) {solid line} and held in 2002 [27]. As suggested by Roser [28],
Fast Crossing (FC) {dashed} patterns [23]. a cold partial (15%) snake, was installed in AGS.
The result of using two partial helical dipole
snakes [29] to overcome about 40 imperfection
precision, the striking behavior [15] of deuterons resonances in AGS is shown in Fig.11 [26]. The
when their spins are manipulated; their tensor po- final AGS polarization during the 2006 run was
larization’s quantum mechanical behavior (Fig.7) about 65% [26, 29].
may have some significance outside of accelerator More importantly, RHIC has demonstrated
physics. that the Siberian snake technique worked well
189
Figure 10: RHIC accelerator complex in 2008 with polarized proton elements highlighted [25-31].
18 60
16
50
14
Luminosity [ × 10 30 cm – 2 s – 1 ]
12 40
Beam Polarization [ % ]
10
30
8
6 20
4
10
2
0 0
–3 2 7 12 17 22
Elapsed time from beginning of ramp [ × 10 3 s]
Figure 11: The measured asymmetry in AGS with two
partial helical dipole snakes [26]. Figure 12: A typical polarized
√ proton store in both
RHIC rings during 2005 at s = 200 GeV [30].
at high energy; when properly tuned, its beams
were accelerated from 25 GeV/c at injection [7] A.D. Krisch et al, AIP Conf. Proc. 145 (1986)
to 100 GeV/c and then stored for many hours [8] A.D. Krisch et al, PRL 63 (1989) 1137
with essentially no polarization loss, as shown [9] J.E. Goodwin et al, PRL 64 (1990) 2779
in Fig.12 [30]. This allowed measurements of [10] R.A. Phelps et al, PRL 72 (1994) 1479
spin-spin effects in 100 GeV/c p-p scattering [11] C. Ohmori et al, PRL 75 (1995) 1931
√
( s = 200 GeV). Moreover, in recent runs, a [12] B.B. Blinov et al, PRL 73 (1994) 1621
45% polarization was reached in brief efforts to [13] R.A. Phelps et al, PRL 78 (1997) 2772
accelerate and store 250 GeV/c protons [26, 31]. [14] B.B. Blinov et al, PRL 88 (2002) 014801
[15] V.S. Morozov et al, PRL 91 (2003) 214801
References [16] A. Lehrach et al, AIP Proc. 675 (2003) 153
[17] M.A. Leonova et al, PRL 93 (2004) 224801
[1] T.L. Khoe et al, PA 6 (1975) 213 [18] V.S. Morozov et al, PRST-AB 8 (2005) 061001
[2] J.L. Laclare et al, JP 46 (1985) C2-499 [19] A.W. Chao, PRST-AB 8 (2005) 104001
[3] H. Sato et al, NIM A 272 (1988) 617 [20] M. Froissart, R. Stora, NIM 7 (1960) 297
[4] A.D. Krisch et al, AIP Conf. Proc. 42 (1978) [21] V.S. Morozov et al, PRL 100 (2008) 054801
[5] F.Z. Khiari et al, PRD 39 (1989) 45 [22] A.M. Kondratenko et al, Phys. of Particles and
[6] Ya.S. Derbenev, A.M. Kondratenko, PA 8 (1978) Nuclei Lett. 1 (2004) 266
115 [23] V.S. Morozov et al, PRL 102 (2009) 244801
190
Ch.2: BEAM DYNAMICS
[24] -a SPIN Colaboration, U. Michigan Report UM Baier-Katkov flip rate For electron spins ini-
HE 95-09 (1995); -b ibid. UM HE 96-20 (1996); ˆ
tially aligned along an arbitrary unit vector ξ,
-c ibid. UM HE SSC-LOI (1990) Eq.(1) generalizes to [2]
[25] T. Roser, PAC95, 3154
1 2 8
[26] T. Roser, AIP Proc. 1149 (2009) 180; H. Huang W = 1 − (ξ̂ · ŝ)2 + √ ξ̂ · b̂ (6)
et al, ibid. 767 2τ0 9 5 3
[27] A.D. Krisch et al, AIP Conf. Proc. 667 (2003) where ŝ = direction of motion and b̂ = (ŝ ×
[28] T. Roser, ibid. 130 ˙ ŝ|.
˙ b̂ is the magnetic field direction if the elec-
[29] H. Huang et al, PRL 99 (2007) 154801
ŝ)/|
[30] T. Roser, PAC05 358; H. Huang et al, ibid. 1404 tric field vanishes and the motion is perpendicular
[31] M. Bai et al, PRL 96 (2006) 174801 to the magnetic field. The corresponding instan-
taneous rate of polarization along ξ̂ is

−1 2
2.6.6 Radiative Polarization in Electron τbk = τ0−1 1 − (ξ̂ · ŝ)2 (7)
Storage Rings 9
D.P. Barber, DESY Thomas-BMT equation Neglecting radiative
G. Ripken, Deceased 2004 spin flip, the motion of the CM spin expecta-
tion value ξ of a charged particle traveling in EM
Sokolov-Ternov (ST) effect [1] Relativistic fields is governed by the Thomas-BMT equation
electrons in a storage ring emit synchrotron radia- × ξ (Sec.2.6.1).
dξ/dt =Ω
tion. A very small fraction of the radiated photons
We write
cause spin flip. For electron spins aligned along
a uniform magnetic field, the ↑↓ and ↓↑ flip rates =Ω
Ω co +
ω sb (8)
differ and then spin polarization builds up antipar- where Ω co is due to the fields on the closed orbit,
allel to the field. Positrons become polarized par- whence Ω co (s+C) = Ω co (s). Ω
co = Ω
ref +
ω imp ,
allel to the field. The transition rates for electrons ref
where Ω contains the design fields and ω imp
are
√ represents the effects of magnet misalignments,
5 3 re γ 5 8 correction fields etc. ω sb is due to 6-D synchro-
W↑↓ = 1 + √
16 me |ρ|3 5 3 betatron motion with respect to the closed orbit.
√ On the closed orbit the T-BMT equation
5 3 re γ 5 8
W↓↑ = 1 − √ (1) d co
16 me |ρ|3 5 3 ξ = Ω ×ξ (9)
For positrons, interchange plus and minus signs dt
here and elsewhere. can be solved in the form
The equilibrium polarization in a uniform = Rco (s, s0 )ξ(s
ξ(s) 0) (10)
3×3
magnetic field is independent of γ, where Rco 3×3 is a rotation matrix. The real unit
W↑↓ − W↓↑ 8 eigenvector (rot. axis) for the 1-turn matrix
Pst = = √ = 0.9238 (2)
W↑↓ + W↓↑ 5 3 Rco3×3 (s + C, s), denoted by n̂0 (s), is the periodic
spin solution on the closed orbit. For a perfectly
For a beam initially unpolarized, the time depen-
aligned flat ring with no solenoids, n̂0 (s) = ±ŷ.
dence for build-up to equilibrium is
The 1-turn matrix has a complex conjugate pair
P (t) = Pst [1 − exp (−t/τ0 )] (3) of eigenvalues e±i2πνsp . Given n̂0 , we define the
with build-up rate unit vectors (m̂0 , ˆl0 ) such that m̂0 = ˆ
l0 × n̂0 and
√ ˆl0 = n̂0 × m̂0 fulfill Eq.(9), and such that
5 3 re γ 5
τ0−1 = (4) m̂0 (s0 + C) + iˆl0 (s0 + C)
8 me |ρ|3
The time τ0 depends strongly on γ and ρ but is = ei2πνsp m̂0 (s0 ) + iˆl0 (s0 ) (11)
typically minutes or hours. In a flat ring in which
all bending magnets have the same ρ just average The (m̂0 , ˆl0 ) are usually not periodic in s. But
(4) over the circumference C, with a further rotation by an angle ψsp (s) around
2π E[GeV]5 n̂0 we can construct the vectors (m̂, ˆl),
τ0−1 [s−1 ] ≈ (5)
99 C[m]ρ[m]2 m̂(s) + i ˆl(s) = e−iψsp (s) m̂0 (s) + iˆl0 (s) (12)
191
By choosing ψsp (s + C) − ψsp (s) = 2πνsp , the azimuth and the polarization is aligned along the
set (n̂0 , m̂, ˆl) is then periodic in s with period C. Derbenev-Kondratenko vector n̂ [7, 8].
The vectors (m̂, ˆl) are needed in Sec.2.6.7. The unit vector field n̂ (the “invariant
The closed orbit spin tune νsp is the number spin field”) [9, 10] depends on s and u ≡
of spin precessions per turn around n̂0 . For a per- (x, px , y, py , z, δ). n̂(u; s) satisfies the T-BMT
fectly aligned flat ring without solenoids νsp = equation at (u; s) and is periodic: n̂(u; s) =
aγ0 , where a = (g − 2)/2 (Sec.2.6.1) and γ0 is n̂(u; s + C). On the closed orbit n̂(u; s) reduces
the Lorentz factor for the beam energy.1 Only the to n̂0 (s).
fractional part of νsp can be extracted from the Derbenev–Kondratenko–Mane formula Tak-
numerical values of the eigenvalues e±i2πνsp . ing into account radiative depolarization due to
Baier-Katkov-Strakhovenko (BKS) equation photon-induced longitudinal recoils, the equilib-
Neglecting the effect of stochastic (synchrotron rium electron polarization along the n̂ field is
radiation) photon emission on the trajectory and [7, 11, 8]
= >
imagining that all particles remain on the closed <

− 5√ 8
3
× ds 1
|ρ(s)| 3 b̂ · (n̂ − ∂δ )
∂ n̂
orbit, the equation of motion for electron polar- = s>
2 11 ∂ n̂ 2
Pdk = <
ization is [3] ds |ρ(s)| 1
(1− 2
(n̂ · ŝ) + )
3 9 18 ∂δ
dP co 1 2 8 s
= Ω ×P − P − ŝ(P ·ŝ)+ √ b̂(s) (15)
dt τ0 (s) 9 5 3
In the case of horizontal motion in a vertical mag- where s denotes an average over phase space at
= (aγc/ρ)ŷ, and b̂(s) = ŷ. azimuth s. This formula differs from Eq.(13) by
netic field, we have Ω the inclusion of the terms with ∂∂δn̂ and use of n̂
From the BKS equation, the generalized instead of n̂0 . The ensemble average of the polar-
Sokolov-Ternov formula for the asymptotic elec- ization is
tron polarization in arbitrary magnetic fields
along the closed orbit is Pens,dk (s) = Pdk n̂s (16)
< n̂0 (s)·b̂(s)
ds |ρ(s)|3 and n̂s is very nearly aligned along n̂0 (s) (see
8
Pbks = − √ n̂0 < (13) the angle estimate below). The value of the en-
5 3 [1− 29 (n̂0 (s)·ŝ)2 ] semble average, Pens,dk (s), is essentially inde-
ds |ρ(s)| 3
See [4] for time scales. Usually, in rings contain- pendent of s.

ing dipole spin rotators (Sec.2.6.3) the polariza- The effect of transverse recoil can also be in-
cluded but contributes derivative terms analogous
tion |Pbks | cannot reach 0.9238 [5].
to ∂∂δn̂ which are typically a factor γ smaller than
The BKS polarization build-up rate is
√ " ∂ n̂ ∂ n̂
5 3 re γ 5 1 1 − 29 (n̂0 · ŝ)2 ∂δ and can be neglected unless ∂δ is very small
−1 ∂ n̂
τbks = ds [12, 13]. If ∂δ were to vanish, a Pdk of 99.2 %
8 me C |ρ(s)|3 could be reached [12, 13, 8].
(14)
In the presence of radiative depolarization
This is in accord with Eq.(7) by replacing ξ̂ → n̂0
Eq.(14) becomes
and averaging. √
Radiative depolarization The stochastic ele- −1 5 3 re γ 5 1
τdk = (17)
ment of photon emission together with damping 8 me C
determines the equilibrium phase-space density " 3 4
∂ n̂ 2
1 − 29 (n̂ · ŝ)2 + 11
distribution. The same photon emission also im- × ds 18 ∂δ
parts a stochastic element to ω sb and then, via |ρ(s)| 3
s
the T-BMT equation, spin diffusion (and thus de- This can be written in the form
polarization) can occur [6]. The attainable po-
1 1 1
larization is the result of a balance between the = + (18)
Sokolov-Ternov effect and this radiative depolar- τdk τst τdep
ization. In the approximation that the orbital mo- where τst−1 can be (very well) approximated by
tion is linear, the value of the polarization is es- −1
τbks in (14) and
sentially the same at each point in phase space and √ " 3 4
11 ∂ n̂ 2
1
In this and the next sections, we use the symbol “a” in- −1 5 3 re γ 5 1 18 ∂δ
τdep = ds (19)
stead of “G” used in the rest of the Handbook. 8 me C |ρ(s)|3
s
192
Ch.2: BEAM DYNAMICS
The time dependence for build-up from an initial masses of the Z and of vector mesons such as the
polarization P0 to equilibrium is Υ’s can be obtained [20, 21, 22, 23, 24]. High
polarizations are not needed for these measure-
P (t) = Pens,dk 1 − e−t/τdk + P0 e−t/τdk (20) ments.
This formula can be used to calibrate polarime- Notes For a treatment of phase space and po-
ters (Sec.7.6) [14]. However, the calibration will larization evolution equations using an equation
be imprecise if ∂∂δn̂ in the numerator of Eq.(15) is of the Fokker-Planck type (Sec.2.4.7), see [16,
not well enough known. For examples of build-up 17, 18] For an overview of measurements see
curves see [5]. [25, 5, 26]. For more information and an overview
Resonances Away from the spin–orbit reso- of the theoretical background see [8, 18].
nances2 (see also Sec.2.6.7)
References
νsp = k + kx νx + ky νy + kz νz (21)
[1] A.A. Sokolov, I.M. Ternov, Sov. Phys. Dokl. 8
n̂(u; s) ≈ n̂0 (s). But near resonances n̂(u; s) No.12 (1964) 1203
deviates from n̂0 (s) by typically tens of milli- [2] V.N. Baier, V.M. Katkov, Sov. Phys. JETP. 25
radians at a few tens of GeV and the deviation (1967) 944
increases with distance in phase space from the [3] V.N. Baier et al, Sov. Phys. JETP. 31 (1970) 908
closed orbit. The “spin orbit coupling function” [4] B. Montague, Physics Reports 113 (1984)
∂ n̂ [5] D.P. Barber et al, Phys. Lett. 343B (1995) 436
∂δ , which quantifies the depolarization, can then
be large and the equilibrium polarization can be [6] V.N. Baier, Yu. Orlov, Sov. Phys. Dokl. 10 (1966)
low. Note that even very close to resonances, 1145
|n̂s | ≈ 1: Pens,dk (Eq.(16)) is mainly influenced [7] Ya.S. Derbenev, A.M. Kondratenko, Sov. Phys.
JETP. 37 (1973) 968
by the value of Pdk in Eq.(15).
[8] D.P. Barber et al, ICFA workshop on Quan-
To get high polarization, one must have
tum Aspects of Beam Physics, World Scientific
(∂ n̂/∂δ)2 1 in dipole magnets. The machine
(1998)
optimization required to make ∂∂δn̂ small is called [9] D.P. Barber, J.A. Ellison, K. Heinemann, PRST-
“spin matching” (Sec.2.6.8). AB 7 (12), 124002 (2004)
Kinetic polarization The (numerator) term lin- [10] G.H. Hoffstaetter, M. Vogt, D.P. Barber, PRST-
ear in ∂∂δn̂ in Eq.(15) is due to a correlation be- AB 2 (11) 114001 (1999)
tween the spin orientation and the radiation power [11] S.R. Mane, PR A36 (1987) 105
[4]. In rings where n̂0 is horizontal due, say, to the [12] D.P. Barber, S.R. Mane, PR A37 (1988) 456
presence of a solenoid Siberian Snake (Sec.2.6.3) [13] L.N. Hand, A. Skuja, PRL 59 (1987) 1910
[15], ∂∂δn̂ has a vertical component in the dipole [14] D.P. Barber et al, NIM A338 (1994) 166
fields. This can lead to a build-up of polariza- [15] C.W. de Jager, V. Ptitsin, Yu.M. Shatunov, Proc.
tion (“kinetic polarization”) even though the pure 12th Int. Symp. High Energy Spin Physics,
−1 World Scientific (1997)
Sokolov-Ternov effect vanishes. The rate is τdk .
Beam energy calibration A polarized electron [16] K. Heinemann, DESY Report 97-166 (1997) and
beam can be depolarized by applying a weak os- arXiv: physics/9709025
cillating magnetic field perpendicular to n̂0 with a [17] K. Heinemann, D.P. Barber, NIM A463 (2001)
62 and A469 (2001) 294
frequency frf related to the fractional part of the
[18] D.P. Barber, G. Ripken, arXiv:
spin tune ν̃sp by [19]
physics/9907034v2
frf = f0 ν̃sp or frf = f0 (1 − ν̃sp ) (22) [19] Kh.A. Simonian, Yu.F. Orlov, Sov. Phys. JETP.
18 (1964) 123
Thus the required frf gives an accurate measure-
[20] Yu.M. Shatunov, A.N. Skrinsky, Particle World 1
ment of ν̃sp and this gives high relative precision
(1989) 35
on νsp . By relating νsp to the average energy of
[21] R. Assmann et al, Z. f. Physik C66 (1995) 567
each beam, high precision measurements of the [22] M. Placidi et al, EPAC98
CM energy of colliding e+ e− beams and of the [23] D.P. Barber et al, PL B135 (1984) 498
2
The resonance condition should be more precisely ex- [24] W.W. MacKay et al, PR D29 (1984) 2483
pressed in terms of the “amplitude dependent spin tune” [25] D.P. Barber, Proc. 12th Int. Symp. High Energy
[8, 9]. But for typical e− /e+ rings the amplitude dependent Spin Physics, loc.cit.
spin tune differs only insignificantly from νsp . [26] Yu.M. Shatunov, PA 32 (1990) 139
193
2.6.7 Computing Algorithms for e− /e+ The combined orbit and spin motion is then
Polarization in Storage Rings described by 8 × 8 transport matrices of the form
D.P. Barber, DESY
M6×6 06×2
G. Ripken, Deceased 2004 M̂ = (4)
G2×6 D2×2
There are two classes of computer algorithm for acting on the vector (u, α, β), where M6×6 is a
estimating the equilibrium e− /e+ polarization in symplectic matrix describing orbital motion and
real rings (for some notations see Sec.2.6.6): G2×6 describes the coupling of the spin variables
(i) Methods based on evaluating ∂∂δn̂ in (α, β) to the trajectory and depends on m̂(s) and
the Derbenev-Kondratenko-Mane (DKM) for- ˆl(s) (Eq.(3), Sec.2.6.8). D2×2 is a rotation matrix
mula (Sec.2.6.6) given the ring layout and magnet associated with the spin basis rotation of Eq.(11)
strengths; and in Sec.2.6.6 [8, 9].

(ii) algorithms such as SLICKTRACK [1] es- The eigenvectors for one turn defined by
timate τdep using Monte-Carlo tracking. M̂(s0 + C, s0 ) · qμ = λ̂μ · qμ are written in the
The class (i) algorithms are further divided form
according to the degree of linearization of the spin
v (s )
and orbital motion: qk (s0 ) = k 0 ,
qk (s0 )]∗
q−k (s0 ) = [
w k (s0 )
(ia) The SLIM family (SLIM [2, 3] - thin
for k = I, II, III,
lenses, SLICK [4] - thick lenses), and ASPIRRIN
06 (s0 )
[5]: linearized orbital and spin motion; qk (s0 ) =
, qk (s0 )]∗
q−k (s0 ) = [
(ib) SMILE [6] and SODOM [7]: Linearized w k (s0 )

orbital motion but |ξ|-preserving spin motion; for k = IV (5)
(ic) SpinLie: Nonlinear orbital motion and for arbitrary s0 . The vk are the eigenvectors for
orbital motion with eigenvalues λk = e−i2πνk

|ξ|-preserving spin motion (Sec.2.6.9).
The linear approximation – SLIM The orbital and with ν−k = −νk (k = I, II, III). These
and spin motion are linearized and the compo- eigenvectors obey the orthogonality and normal-
nents ωzsb , ωxsb and ωysb of ω sb are linearized in ization relations [2, 8]. The corresponding eigen-
the form [8, 9] values of M̂(s0 + C, s0 ) are λ̂k = λk (k =
⎡ ⎤ I, II, III) and λ̂IV = e−i2πνIV with νIV =
x
⎡ sb ⎤ ⎢ px ⎥ νsp and with ν−IV = −νIV .
ωz ⎢ ⎥ The spin parts of the eigenvectors w k (s0 )
⎣ ωxsb ⎦ = F3×6 ⎢ ⎢ ⎥
y ⎥
(1)
⎢ py ⎥ (k = I, II, III) and w IV (s0 ) can be written
ωysb
⎣z ⎦ −1
δ w k (s0 ) = − D(s0 + C, s0 ) − λ̂k
× G(s0 + C, s0 )vk (s0 )
where u ≡ (x, px , y, py , z, δ) describes motion
with respect to the closed orbit. In particular for k = I, II, III (6)

px = x and py = y (except in solenoids). 1 1
IV (s0 ) = √
w e−i ψsp (s0 )
The detailed forms of the matrix F3×6 for 2 −i
the various machine elements can be found in [9]. for k = IV
The orbital motion in sextupoles is linearized. For
example for a quadrupole, defining g̃ = −(1 + and w −k (s0 ) = [w k (s0 )]∗ , (k = I, II, III, IV ).
∂B In this linear approximation n̂(u; s) can be
aγ0 ) g where g = pe0 ∂xy obtained via [10, 11, 9]
⎡ ⎤
000000 α(u; s)
n̂(u; s) − n̂0 (s) ≡
F(s) = ⎣ 0 0 g̃ 0 0 0 ⎦ (2) β(u; s)
g̃ 0 0 0 0 0
= {Ak w k (s) + A−k w −k (s)} (7)
In linear approximation we write k=I,II,III
for amplitudes Ak determined by the orbit via
n̂(u; s) = n̂0 (s)+α(u; s)m̂(s)+β(u; s)ˆl(s) (3)
u(s) = {Akvk (s)+A−k v−k (s)} (8)
valid for α2 + β 2 1. k=I,II,III
194
Ch.2: BEAM DYNAMICS
Then with respect to the (n̂0 , m̂, ˆl) frame, damping, and delivers estimates of τdep . The po-
∂ n̂ larization is then obtained from the approximation
∗
≡i {vk5 w k − vk5 w k∗ } (Sec.2.6.6)
∂δ
k=I,II,III τtot 1 1 1
Peq = Pbks with = + (12)
∗
= −2 Im vk5 wk (9) τbks τtot τbks τdep
k=I,II,III
This ignores the (normally small) term with ∂∂δn̂
The depolarization time is then in the numerator of the DKM formula. SLICK-
√
TRACK calculates with full |ξ|-preserving spin
−1 55 3 re γ05 s0 +C 1
τdep,lin ≈ ds̃ motion and, in contrast to the analytical algo-
36 me C s0 |ρ(s̃)|3 rithms, it has the advantage that it can handle
# $2
2 the strongly nonlinear orbital motion caused by

∗
× Im [vk5 (s̃)wkμ (s̃)] (10) beam-beam forces.
μ=1 k=I−III
References
In this linearized theory the vectors n̂(u; s)
and ∂∂δn̂ display only the first order resonances [1] D.P. Barber, Proc. 16th Int. Spin Physics Symp.,
World Scientific (2005) 738
νsp = k + kI νI + kII νII + kIII νIII (11) [2] A.W. Chao, NIM 180 (1981) 29(∗)
with |kI | + |kII | + |kIII | = 1. They arise from [3] A.W. Chao, AIP Proc. 87 (1981) 395(∗)
the denominator matrix in Eq.(6). The formalism [4] SLICK is a thick lens version of SLIM by D.P.

is only valid for α2 + β 2 1. In a fully un- Barber (1982)
coupled optic I → x, II → y and III → z. [5] C.W. de Jager, V. Ptitsin, Yu.M. Shatunov, Proc.
This formalism forms the natural language for 12th Int. Symp. High Energy Spin Physics,
World Scientific (1997)
maximizing the polarization by “spin matching”
[6] S.R. Mane, PRA 36 (1987) 105
(Sec.2.6.8).
[7] K. Yokoya, KEK Report 92-6 (1992), DESY Re-
Other computer codes [12] The SMILE algo- port 99-006 (1999)
rithm works in the thin lens approximation and [8] H. Mais, G. Ripken, DESY Report 83-62
calculates ∂∂δn̂ by extending the first-order pertur- (1983)(∗)
bation theory of SLIM to high-order for the spin [9] D.P. Barber, K. Heinemann, G. Ripken, Z. f.
motion. High-order resonance effects are man- Physik C64 (1994) 117
ifested by high-order resonance denominators. [10] S.R. Mane, DESY Report 85-125 (1985); Cor-
SODOM represents n̂ in terms of spinors and en- nell Report CLNS-86/750
codes the condition n̂(u; s) = n̂(u; s + C) and [11] D.P. Barber, K. Heinemann, G. Ripken, DESY
the similar periodicity of the 1-turn 2×2 spinor Report M-92-04 (1992), revision 1999
transfer matrix, in terms of 2π-periodicity in the [12] G.H. Hoffstaetter, M. Vogt, D.P. Barber, PRST-
three orbital phases. n̂(u; s) is then constructed AB 2 (11) 114001 (1999)
(∗)
by solving for the Fourier coefficients and ∂∂δn̂ is Modern notation: replace n̂ by n̂0
obtained by numerical differentiation. The algo-
rithm SpinLie utilizes Lie-algebraic methods to 2.6.8 Spin Matching in e− /e+ Rings
provide a perturbation expansion for n̂ and can D.P. Barber, DESY
handle moderately nonlinear orbit motion. The G. Ripken, Deceased 2004
vector n̂(u; s) can also be obtained by “strobo-
scopic averaging” using the code SPRINT [12]. Spin matching in the SLIM formalism Syn-
This algorithm automatically includes all orders chrotron radiation causes depolarization and to
−1
of resonance. maximize the polarization we must minimize τdep
The above analytical algorithms all exploit (See Sec.2.6.6 for some notations). In lowest ap-
the DKM formula (Sec.2.6.6) but SLICKTRACK proximation this is quantified by the SLIM for-
(an extended version of SLICK) uses the transfer malism (Sec.2.6.7). The manipulations are called
matrices of SLICK to simulate the depolarization spin matching and in real rings this takes place in
process directly using Monte-Carlo tracking sim- stages as follows.
ulations of the effects on the trajectories, and then Stage 1: Strong synchrobeta spin matching of the
on the spins, of stochastic photon emission and perfectly aligned ring
195
−1 ∗
To minimize τdep we must minimize vk5 (k = This is strong synchrobeta spin matching. A sec-
I, II, III) or the components of w k at az- tion of the ring satisfying a condition in Eq.(1) is
imuths where 1/|ρ(s)|3 is large. The vk5 ∗ deter- spin transparent for mode k. The interpretation is
mine the orbit excitation due to synchrotron ra- immediate: the overall spin-orbit coupling for the
diation [1]. In particular, for rings without x-y section vanishes for mode k. Clearly, the exact
coupling, vII5 ∗ usually vanishes in the arcs since spin matching conditions depend on the layout of
the vertical dispersion ηy vanishes. However, vII5 ∗ the ring and each case must be handled individu-
does not vanish inside spin rotators (Sec.2.6.3) ally. In thin lens approximation the G matrix for
containing vertical bends. On the other hand vI5 ∗ a quadrupole of length lq is

tends not to vanish in the arcs since the hori- −q̃ly 0 −q̃lx 0 0 0
zontal dispersion ηx = 0. Finally, vIII5 ∗ essen- G= (3)
+q̃my 0 +q̃mx 0 0 0
tially never vanishes. Each case must be eval- where q̃ = (1 + aγ) g lq . The thin- and thick-lens
uated individually but the minimal recipe is to forms of G for other magnet types are given in
try to minimize w k for (k = I, II, III) only [3, 4, 5].

at azimuths where |vk5 (s)|2 /|ρ(s)|3 is sufficiently If the Gij (sμl+1 , sμl ) cannot be brought to
large. This in turn requires (Eq.(6), Sec.2.6.7) zero while maintaining an acceptable optic, then
that G(s + C, s) · vk (s) be minimized for k = the G(sμl + C, sμl )vk (sμl ) themselves should be
I, II, III. The ring layout must be designed minimized. This essentially means that the ef-
with this in mind and sufficient independently fects of elements of the G matrices of sections of
powered quadrupoles be provided to ensure flexi- the ring are made to partially cancel one another.
bility in the optics. Subsequent calculations with Spin matching of a ring with a solenoid Siberian
SLIM will indicate whether the match criteria Snake (Sec.2.6.3) is an example of this [6]. By
for the adopted design suffice. Consider, for ex- Eq.(6), Sec.2.6.7, reduction of G(s + C, s)vk (s)
ample, a specific mode k. Label those bending for (k = I, II, III) also reduces the angle be-
magnets at which |vk5 (s)|2 /|ρ(s)|3 is large by tween n̂ and n̂0 at azimuth s.
μ1 (k) , μ2 (k) , ..., μnk (k) . Then the suppression of
depolarization associated with the kth mode re- Alternative Stage 1: Harmonic synchrobeta spin
quires that w k (sμi ) = 0 for all (i = 1 to nk ). In matching of the perfectly aligned ring
general this in turn requires [2] If the strong spin matching methods are imprac-
tical for some reason, another approach aimed
G(sμ2 , sμ1 ) vk (sμ1 ) = 0 at minimizing the strengths of depolarizing res-
G(sμ3 , sμ2 ) vk (sμ2 ) = 0 onances can be adopted [7, 8].
.. Reformulation in terms of beta functions and dis-
. persion [9]
G(sμ1 + C, sμn ) vk (sμn ) = 0 (1) We can reformulate Stage 1 by transforming the
where we suppressed the superscript label “k”. To coordinates from u ≡ (x, px , y, py , z, δ) to ũ ≡
fulfill Eq.(1) we then require the Gij (sμl+1 , sμl ) (x̃, p̃x , ỹ, p̃y , z̃, δ) defined by
to vanish when the jth component of vk does not ũ = Ku
vanish. The matrix G can be written in the form with the symplectic matrix
s2 ⎡ ⎤
1 0 0 0 0 −η1
G(s2 , s1 ) = ds̃ D(s2 , s̃) G0 (s̃) M(s̃, s1 )
⎢ 0 1 0 0 0 −η2 ⎥
s1 ⎢ ⎥
⎢ 0 0 1 0 0 −η3 ⎥
where K(s) = ⎢ ⎥
⎢ 0 0 0 1 0 −η4 ⎥
ls lx ly ⎣ η −η η −η 1 0 ⎦
G0 = ·F (2) 2 1 4 3
−ms −mx −my 0 0 0 0 0 1
Thus Gij (sμl+1 , sμl ) depends on the orientation where the η ≡ (η1 , η2 , η3 , η4 ) = (ηx , ηx , ηy , ηy )
are components of the dispersion vector. Then
of (m̂, ˆ
l) so that in some cases some elements
of Gij (sμl+1 , sμl ) vanish automatically. But in x̃ = x − δηx , ỹ = y − δηy
general these conditions can only be fulfilled by The formalism remains canonical and in particu-
adjusting quadrupole strengths – while maintain- lar, there are new eigenvectors
ing other necessary features of the orbital optics. =⇒ ṽ μ (s) = K(s) vμ (s)
196
Ch.2: BEAM DYNAMICS
for which the eigenvalues and orthogonality con- 0 in analogy with Eq.(1). Then in a fully un-
ditions are unchanged. Furthermore G̃(s2 , s1 ) = coupled optic and by taking into account only
G(s2 , s1 )K−1 (s1 ). the quadrupoles this is equivalent to requiring
The depolarization rate then takes the form [11, 12]:
√ s0 +C For horizontal motion:
−1 55 3 re γ 5 1 1
τdep,lin= ds̃ (1 + aγ) 1 sμl+1
36 me C s0 |ρ(s̃)|3 − √ ds̃ βx (s̃)g(s̃)e−i ψx (s̃)
# $2 2 C sμl
2
× Im [ fk (s̃)w̃kμ (s̃) ] × [ ly (s̃) ± i my (s̃) ] e∓iψsp (s̃) = 0 (4)
μ=1 k=I−III
For vertical motion:
with fk = 6n=1 [K −1 ]5n ṽkn ∗ = v ∗ and w̃ k =
k5 (1 + aγ) 1 sμl+1
w
k . In the special case, or the approximation, of − √ ds̃ βy (s̃)g(s̃)e−i ψy (s̃)
no orbital coupling, the 6 × 6 orbit matrices just 2 C sμ
consist of three 2 × 2 matrices on the diagonal. × [ lx (s̃) ± i mx (s̃) ] e∓iψsp (s̃) = 0 (5)
The eigenvectors ṽk (s) of the revolution matrix
are then written in the form For longitudinal motion:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ sμ
tx 02 02 (1 + aγ) 1 l+1 [αz (s̃) + i]
− √ ds̃
ṽI = ⎣ 02 ⎦ , ṽII = ⎣ ty ⎦ , ṽIII = ⎣ 02 ⎦ 2 C sμl βz (s̃)
02 02 tz
×g(s̃)e−iψz (s̃) {ηy [lx (s̃) ± imx (s̃)]

tr = 1 βr (s) +ηx [ly (s̃) ± imy (s̃)]}e∓iψsp (s̃) = 0 (6)
e−iψr (s)
2βr (s) −[α r (s) + i]
Harmonic synchrobeta spin matching in
(r ≡ x, y, z) and the fk are given by fI ≡ fx = terms of beta functions and dispersion follows the
−(ṽI1 η2 − ṽI2 η1 ) ; fII ≡ f y = −(ṽII3 η4 − path detailed earlier under “Alternative Stage 1”
βz −iψz (s) but with the eigenvectors ṽk and the matrices F̃.
ṽII4 η3 ) and fIII (s) ≡ fz = 2 e . The
Typical expressions can be found in [12, 13, 7].
|fI | and |fII | are just the factors
2 2
Commentary
ηr 2 + (αr ηr + βr ηr )2 A detailed commentary on the practical use of this
(r = x, y)
βr formalism, with examples, can be found in [7, 8].
in [10]. fIII is in practice almost independent of Stage 2: Harmonic closed orbit spin matching
s. Note that α’s and β’s here are Courant-Snyder Once the perfectly aligned ring has been spin
parameters, not the spin components. With these matched, the effects of misalignment must be ad-
coordinates we can write dressed. In a perfectly aligned flat ring with no
−1 −1
solenoids, τdep,lin is automatically zero [7, 8]. In
k (s0 ) = − D(s0 + C, s0 ) − λ̂k
w̃ real misaligned rings there is a vertical closed or-
× G̃(s0 + C, s0 )ṽk (s0 ) bit distortion and n̂0 is tilted from the vertical in
the arcs (see below). Then parts of G and G̃
for (k = I, II, III). for the arc quadrupoles do not vanish so that nor-
−1
In thin lens approximation the G̃ matrix for a mally τdep,lin = 0. In practice the tilts can be
quadrupole is tens of milliradians but even these small angles

−q̃ly 0 −q̃lx 0 0 κ1 can lead to strong depolarization so that it is es-
G̃ = sential that the ring be very well aligned from the
+q̃my 0 +q̃mx 0 0 κ2
beginning. Note that vertical closed orbit distor-
where κ1 = −q̃ly η1 − q̃lx η3 and κ2 = +q̃my η1 + tion leads primarily to depolarization due to hor-
q̃mx η3 . By separating the transverse coordinates izontal synchrobetatron motion in the arcs. Note
into betatron and dispersion contributions, col- also that tilts of tens of milliradians cause a neg-
umn six of G̃ contains terms depending on dis- ligible decrease of the underlying ST polarization
persions. (Eq.(13), Sec.2.6.6).
The strong spin matching condition w̃ k = 0 The key to minimizing depolarization due to

now means setting the G̃(sμl+1 , sμl ) ṽk (sμl ) = distortions is reduction of the tilt of n̂0 in the arcs.
197
This is conveniently done using SLIM-like con- [4] A.W. Chao, AIP Proc. 87 (1981) 395(∗)
cepts [14]. Viewed from the (n̂0 , m̂, ˆl) frame cal- [5] D.P. Barber et al, PA 17 (1985) 243(∗)
culated for the design orbit, the first order devia- [6] C.W. de Jager, V. Ptitsin, Yu.M. Shatunov, Proc.
tion of n̂0 from the design orientation can be writ- 12th Int. Symp. High Energy Spin Physics,
ten as World Scientific (1997)
[7] D.P. Barber, G. Ripken, arXiv: physics/
C ei2πks/C 9907034v2
[δn01 (s) − iδn02 (s)] = −i hk
2π k − νsp [8] Handbook of Accelerator Physics and Engineer-
k
ing, 1st ed., World Scientific, 2006
where the hk are Fourier coefficients given by [9] H. Mais, G. Ripken, DESY Report 86-29
(1986)(∗)
1 s0 +C
hk = ds̃ [d1 (s̃) − id2 (s̃)] e−ik2πs̃/C [10] M. Sands, SLAC 121 (1970)
C s0 [11] A.W. Chao, K. Yokoya, KEK Report 81-7 (1981)

[12] J. Buon, K. Steffen, NIM A245 (1986) 248
Here
[13] K. Yokoya, KEK Report 81-19 (1982)
d1 ls lx ly [14] D.P. Barber et al, DESY Report 85-44 (1985)(∗)
=
d2 −ms −mx −my [15] D.P. Barber et al, NIM A338 (1994) 166
⎧ ⎡ ⎤⎫
⎨ ΔBs 1+aγ
1+γ0
0
⎬ [16] R. Assmann et al, Proc. 11th Int. Symp. High En-
· F · uco − e ⎣ ΔBx (1 + aγ0 ) ⎦ ergy Spin Physics, AIP Proc. 343 (1995)
⎩ p0 ⎭
ΔBy (1 + aγ0 ) (∗) Modern notation: replace n̂ by n̂0 .
where the ΔBx,y,s are field errors and uco is the
deviation of the 6-D closed orbit from the design 2.6.9 Lie Algebra for Spin Motion
orbit. δn̂0 can be minimized by using correction K. Yokoya, KEK
coils to adjust the closed orbit (e.g. by generat- The Lie algebra method (Sec.2.3.6) can include
ing closed bumps so that the luminosity is not af- the classical spin motion [1]. The Hamiltonian is
fected) in such a way that the real and imaginary written in general as
parts of hk , with k near νsp , are small. This is har-
monic closed orbit spin matching. Since the ring H = Horb (z, θ) + s · W(z, θ)
misalignments are usually not well known and where z represents the six orbital variables, θ is
the closed orbit cannot usually be measured well the machine azimuth, Horb the orbit Hamiltonian,
enough, the closed orbit correction is carried out s the spin vector, and W is written in terms of
empirically by observing the polarization [15]. If the electromagnetic field along the orbit. The
the misalignments and the orbit are well enough Thomas-BMT equation (Sec.2.6.1) is then
known the correction coil strength can be calcu- ds
lated ab initio (deterministic harmonic closed or- = [s, H] = W × s
bit spin matching) [16]. The correction scheme dθ
should be chosen so that it achieves the maximum where [ ] is the Poisson bracket. Note [si , sj ] =

effect on δn̂0 with the smallest possible additional k ijk sk . The Lie map of an element of length
orbit distortion. Δθ is e−:HΔθ: (valid if H is independent of θ
Higher order resonances and Beam-beam spin or if Δθ is infinitesimally small). If one ignores
matching the Stern-Gerlach force (the effect of spin on the
By calculating beyond the linearizations of the orbit motion), the map can be decomposed as
SLIM formalism one finds higher order spin-orbit e−s·W e−:H: where H and W are polynomials of
resonances The beam-beam interaction is equiva- z. One can concatenate the Lie maps to obtain
lent to a nonlinear lens and can spoil a spin match. the map of a beamline. In contrast to the orbit
For comments and more details see [7, 8]. case the concatenation of the spin map does not
require the Baker-Campbell-Hausdorf formula; it
References can be written explicitly as
[1] A.W. Chao, J. Appl. Phys. 50 (1979) 595

e−s·W1 e−s·W2 = e−s·W
[2] H. Mais, G. Ripken, DESY Report M-84-04 cos φ = cos φ1 cos φ2 − a1 · a
(1984)(∗) W
[3] A.W. Chao, NIM 180 (1981) 29(∗) sin φ = a2 cos φ1 +a1 cos φ2 +a1 × a2
|W |
198
Ch.2: BEAM DYNAMICS
where φ1 = |W1 |/2, a1 = (W1 /|W1 |) sin φ1 , where is the beam emittance, N is the number
etc. When the orbit one-turn map can be trans- of particles in the ring, W = fmax − fmin with
formed to the normal form e−:V : e−:H: e:V : where fmax,min the frequency limits of the bandwidth
2 +
H is a function of the orbit action variables (z2i W . The fastest cooling is obtained at the opti-
2
z2i−1 )/2 (i = 1, 2, 3) and V = V (z, θ), the total mum system gain, i.e., g = 1/(M + U ). The
map can also be written in normal form with the mixing factor M may be defined to be the ratio
same H and V , of the peak Schottky power density to the aver-
e−:V : e−:s·U : e−:H+s·W :e:s·U : e:V : age Schottky power density (averaged over all the
Schottky bands in the cooling system bandwidth).
where W is a function of orbit action and is iden- U is the ratio of electronic noise power to the av-
tified as the spin tune (times 2π). The Derbenev- erage Schottky power density.
Kondratenko vector n(z, θ) is given by s · n =
Transverse cooling rate Applying kinetic the-

e−:V : e−:s·U : s3 . A computer code to evaluate the ory and feedback theory, the betatron emittance
equilibrium polarization in electron storage rings cooling rate for particles with revolution fre-
using Lie algebra is available [2]. quency f = ω/2π is [5]
'
References G[(m ± Q)ω] eim(θp −θk )±iψβ
Sω =
[1] K. Yokoya, NIM A258(1987)149 T [(m ± Q)ω] ±i
m,±Q
[2] Yu. Eidelman, V. Yakimemko, PA 45(1994)17; ( (2 ;
50(1995)261 πN f (ω) (( G[(m ± Q)ω] ((
+ (2)
|m ± Q| ( T [(m ± Q)ω] (
with Q the betatron tune, f (ω) the normalized
2.7 BEAM COOLING frequency distribution, G the electronic gain, and
ψβ the betatron phase between pickup and kicker.
2.7.1 Stochastic Cooling
The dielectric response is
M. Blaskiewicz, BNL
T [(m ± Q)ω] = 1 + N G((m ± Q)ω)
Stochastic Cooling was invented by Simon van f (ω1 ) exp(i[θp − θk ])
der Meer [1] and was demonstrated at the CERN × dω1 + (3)
ISR and ICE (Initial Cooling Experiment) [2]. ,±ν
i((m±Q)ω)−i(±ν)ω 1 +0
Operational systems were developed at Fermilab
[3] and CERN [4]. A complete theory of cooling where 0+ means take the limit as this term goes
of unbunched beams was developed [4, 5, 6], and to zero though positive numbers. When pickup
was applied at CERN and Fermilab. Several new noise is included the equation for the transverse
and existing rings employ coasting beam cooling emittance of particles with revolution frequency
[7]. ω is
( (
Bunched beam cooling was demonstrated in dω πN ( G[(m±Q)ω] (2
=Sω ω+ ( ( noise,m±Q
ICE and has been observed in several rings de- dt ω0 ( [(m±Q)ω] (
m,±Q
signed for coasting beam cooling. High energy (4)
bunched beams have proven more difficult. Sig-
For low gain and resolution bandwidth larger than
nal suppression was achieved in the Tevatron [8],
ω0 (m±
ω0 the signal to noise ratio near frequency
though operational cooling was not pursued at
Q) is rms /noise,m±Q with rms = f (ω)ω dω
Fermilab. Longitudinal cooling was achieved in
the rms emittance. Changing the gain downstream
the RHIC collider [9]. More recently a vertical
of the noise source does not change noise,m±Q .
cooling system in RHIC cooled both transverse
As cooling proceeds noise,m±Q remains fixed and
dimensions via betatron coupling [10].
the signal to noise ratio drops.
Longitudinal cooling Taking x = E − E0 as
2.7.1.1 Cooling rates energy variable with no measurement noise the
Unbunched beam A simple, but useful, esti- equation of motion for particle k is [5]
mate of stochastic cooling rates is
N ∞

1 d W ẋk = G(nω(xj ), xj )ein(θk −θj ) (5)
= − [2g − g2 (M + U )] (1) j=1 n=−∞
dt N
199
Sec.2.7: BEAM COOLING
where N is the number of particles in the ring, full rf buckets, M is the ratio of peak Schottky
G(Ω, x) = GF (Ω) + xGP (Ω) with GF (Ω) the power to average Schottky power, as in coasting
filter cooling gain and GP (Ω) the electrical part beams [11]. Mixing from intra-beam scattering
of the Palmer cooling gain. A damped diffusion (IBS, Sec.2.4.12) is important in the FNAL recy-
equation can be obtained, cler [12]. For higher accuracy the damped diffu-
sion equations for bunched beams are much more
∂f (x, t) ∂ ∂f
=− F (x, t)f − D(x, t) (6) complicated than those for coasting beams [13]
∂t ∂x ∂x
though calculations without synchrotron sideband
where overlap [13], or neglecting signal shielding [11],
G(mω(x), x) have been done. For application to RHIC it was
F (x, t) = eim(θp −θk ) found that multiparticle simulations proved both
L (mω(x))
m
fast and reliable [9]. To simulate Nr real particles
In the notation of [5], section 4, Gm (x) = using Ns simulation particles one simply multi-
G(mω(x), x)eim(θp −θk ) . When summed over all plies the real gain by Nr /Ns and tracks for Ns /Nr
revolution lines the dielectric response is purely a fewer turns [14]. The relevant algorithms are
function of frequency, identical to those used for the simulation of co-
exp[im(θp − θk )] herent instabilities. As an example consider lon-
L (Ω) = 1 + dxN gitudinal cooling. The first update will take place
m iΩ − imω(x) + 0+
at the rf cavity and, since synchrotron tunes are
×G(Ω, x) ∂ (ω(x)f (x, t)) (7) small compared to one, we may place the stochas-
∂x
The diffusion coefficient is tic cooling kicker at the same spot,
( (
N π (( dx (( (( G(mω(x), x) ((2
x̄n = xn + qV (τn )
D(x, t) = ( (( f (x,t) τ̄n = τn + κ1 x̄n
m |m| dω L (mω(x)) (
( (2 where xn = E − E0 for particle n, τn is the ar-
( G(mω(x), x) (
+ N π (( ( U (8) rival time of particle n with respect to the syn-
m L (mω(x)) ( m chronous particle, and the bars denote updated
where Um is the ratio of noise to signal power variables. The parameter κ1 = f ηTrev /β 2 E0 ac-
measured with a large resolution bandwidth. counts for particle slip when traversing a fraction
Longitudinal stacking Solutions of Eq.(6) with of the ring f between kicker and pickup. At the
∂f /∂t = 0 approximate the central part of the pickup one accumulates the line density array for
solution during the accumulation of antiprotons. k = 1, 2, . . . , Ngrid ,
For no measurment noise and L = 1 one has Ns
( ( λ(tk ) = δ̂(tk − τ̄n )

∂f N π (( dx ((
−f (x) + Gm (x)f |Gm (x)|2 n=1
m
∂x m
|m| ( dω ( where δ̂(t) is the triangle function for linear in-
= Ψ0 (9) terpolation on the grid tk = kΔt. Next one
transports from the pickup to the kicker via τ̄¯n =
where Ψ0 is the flux of cooling particles. Set-
τ̄n + (1 − f )κ1 x̄n . The kick for simple filter cool-
ting Gm (x) = α/f (x) for m− < |m| ≤ m+
ing starts with Δλ(tk ) = λ(tk ) − λold (tk ) where
and maximizing the cooling with respect to α one
the array λold was accumulated on the previous
finds
⎧ ⎫ turn. Then Δλ(t) is Fourier transformed, mul-
⎨(m −m )2 dω/dx ⎬ tiplied by the frequency dependent gain, and in-
f (x) = f (x1 ) exp
+
− (x−x1 )
⎩ 2πN ln m+ Ψ ⎭ verse transformed. The voltage obtained is added
m− 0 to the rf voltage and the update sequence is re-
(10) peated.
where x1 is an energy in the constant flux range. Additional effects such as betatron coupling
Bunched beam cooling For rough estimates and IBS are easily included, though one must be
one can use Eq.(1) with N corresponding to the careful not to make Ns too small and introduce
number of particles in the ring that would create unphysical mixing due to the increased IBS rates.
a current equal to the peak beam current. The The results with several competing factors operat-
major unknown is the mixing factor, M . With ing can be quite different from the estimates of in-
large synchrotron sideband overlap, as occurs in dividual effects. For instance, longitudinal action
200
Ch.2: BEAM DYNAMICS
diffusion from IBS turns out to be very important Longitudinal filter cooling usually employs
for transverse cooling in RHIC. This is because one turn delays between pickup and kicker. The
particles with small synchrotron amplitude have simplest and most common type is of the form
poor cooling [13, 14] but longitudinal diffusion S1 (t) = S0 (t) − S0 (t − Trev ) where S0 (t) is
causes all particles to have signifcant synchroton the pickup signal, Trev is a highly accurate one
amplitudes at least some of the time, leading to turn delay and S1 (t) is the output signal. Sin-
even cooling. gle mode optical fibers allow for multi-GHz band-
widths and are inexpensive. Careful temperature
2.7.1.2 Hardware control is required to keep length variations well
under 1/4 wavelength and various sorts of feed-
The cooling system is a wide-band feedback loop
back on the delay are standard.
with bandwidths of a few hundred MHz to a few
GHz. Pickups are generally wide-band devices Kickers for coasting beams look much like
and employ multiple slots to extract sufficient pickups, as expected from the Lorentz reciprocity
beam power. The individual slot signals can be theorem. One simply reverses the beam direction
combined using a meander line [7]. Alternately, and uses the same solution to Eq.(11) [16]. For
the coupling slots can modify the phase veloc- bunched gold beams in RHIC, longitudinal cool-
ity of the waveguide mode leading to resonant ing requires an rms voltage of 3 kV. Fortunately,
buildup over a wide frequency range [15]. Both the bunch spacing of 107 ns is much larger than
of these technologies require careful design but the 5 ns bunch length. This allowed for a Fourier
yield wide band high signal to noise devices. The series based approach using resonant cavities to
very strong Schottky signals of the gold ions in generate the kick [9, 14].
the RHIC collider allowed the longitudinal pickup
to be very simple. A pair of commercial waveg- 2.7.1.3 New techniques
uide launchers on either side of the beam pipe are Microwave stochastic cooling systems are limited
combined in sum mode. A ceramic window keeps to bandwidths of a few GHz. While this is ad-
all the electronic components outside the vacuum, equate for condensing antiprotons and other rare
resulting in an inexpensive, robust design. particles or cooling the gold beams in RHIC it is
During design both the Panofsky-Wenzel the- not useful for cooling high density proton beams
orem and the potential theorem [16] significantly in colliders. For such beams bandwidths of sev-
augment the usefulness of electromagnetic simu- eral hundred GHz are required. Two technologies
lation codes. Let s = z + vt denote the longi- have been suggested for cooling such beams.
tudinal position of the kicked particle as a func- Optical stochastic cooling employs a wiggler
tion of time. For an isolated kicker the Panofsky- pickup, optical amplifier, and wiggler kicker to
Wenzel theorem (Sec.3.2.1) states there exists a close the cooling feedback loop [18, 19]. Op-
function Φ(x, y, z) such that the momentum kick erations at micron wavelengths with 10% band-
is Δ(px , py , pz ) = ∂Φ/∂r where r = (x, y, z). width are envisioned. The bandwidth is of order
The potential theorem states that Φ obeys 1013 Hz but there are significant challenges. The
∂2Φ ∂2Φ 1 ∂2Φ short wavelengths require linear but nearly achro-
2 + 2 + 2 =0 (11) matic optics and the laser amplifiers push the state
∂x ∂y γ ∂z 2
of the art. For a laser power limited system the en-
where γ is the Lorentz factor. For coasting beam ergy cooling rate is [19]
cooling the front end noise in the pickup is a se- 2 2
rious concern. Another difficulty is maintaining dσE σE 1 P̄ cZ0 K 2
=− with 2 ≈
the appropriate gain and phase functions over very dt nE Trev nE ¯
I/qλσ 2 2
E /q
wide bandwidth. Various sorts of equalizers are (12)
often needed and feedback on delays is common- where σE is the rms energy spread, q is the charge
place. For bunched beams there is the additional of the ion, P̄ is the average laser power, I¯ is the
complication of beam driven coherent lines [17]. average beam current, K = qcBλu /(2πM c2 ) is
In RHIC, the signal power from coherent lines is the undulator parameter, Z0 = 377 Ω, and λ is
often 100 times larger than the Schottky power. A the central waveleth for the laser. The maximum
large dynamic range is needed to keep intermod- laser power is Pmax = P̄ Imax /I¯ where Imax is
ulation products under control. the peak beam current.
201
Coherent electron cooling [20] involves an [14] M. Blaskiewicz, J.M. Brennan, COOL07
electron bunch comoving with the hadron bunch. [15] D. McGinnis, PAC99, p.1713
In the modulator the hadrons induce a density [16] G. Lambertson, AIP Proc. 153 (1987) p.1414
modulation on the electrons. The modulation is [17] D. Boussard, Lecture Notes on Phys. 296,
then amplified in a high gain free electron laser. p.289
In the kicker the electron bunch is again merged [18] A.A. Mikhailichenko, M.S. Zolotorev, PRL 71,
with the hadron bunch and the electrons kick the p.4146 (1993).
hadrons. Appropriate optics and timing result in [19] M.S. Zolotorev, A.A. Zholents, PR E50, p.3087
cooling the hadrons. Central wavelengths of order (1994)
10 microns are envisioned and appropriate optics [20] V.N. Litvinenko, Y.S. Derbenev PRL 102,
114801 (2009)
are being designed. High energy, high current en-
ergy recovery linacs appear capable of supplying

the electron beams to cool proton colliders. The
energy cooling rate is [20] 2.7.2 Electron Cooling

S. Nagaitsev, FNAL
1 4πGL rc Lk D sin ψk
≈ 2 (1 − cos ψp ) (13)
nE λγσx,e ψk Electron cooling, which has become a conven-
tional tool in accelerator technology, makes it pos-
where GL is the gain of the free electron laser, sible to increase or to control the phase space den-
rc = q 2 /4π0 mc2 is the classical particle radius, sity of heavy charged particle (protons, ions, an-
Lk is the length of the kicker, D = d/d(Δp/p) is tiprotons) beams in a storage ring. About ten such
the path length variation with momentum for the storage rings, referred to as coolers, are presently
ions, γ is the Lorentz factor, σx,e is the rms radius operating in Europe, US, China and Japan. These
of the electron beam, ψp ≈ π/2 is the plasma have greatly expanded the experimental capabili-
phase advance in the pickup and ψk ≈ π/2 is ties in the study of elementary particles, nuclear
the plasma phase advance in the kicker. It has and atomic physics.
been assumed that the light wavelength, λ is long
The initial idea belongs to Budker [1] and is
enough, λ/2π ≥ DΔp/p, so that all ions are
based on the model of two-component electron-
cooled.
ion plasma. A monochromatic electron beam is
merged with a straight section of the storage ring.
References The average longitudinal velocities of the ion and
[1] S. van der Meer, CERN/ISR-PO/72-31 (1972); electron beams are made equal and the two beams
S. van der Meer, RMP 57 (1985) 689 are made to overlap transversely, thus creating
[2] G. Carron et al, PAC 79, p.3456; G. Carron et al, electron-ion plasma co-propagating for a period
PL B77 (1978) 353 of time until the electron beam is separated from
[3] G. Lambertson et al, Proc. Int. Conf. on High En- the ion beam and discarded. The ions, while trav-
ergy Acc. (1980) p.794; R.L. Hogrefe et al, PAC eling with electrons, undergo Coulomb scattering
81, p.2455 and, under certain conditions, transfer some of
[4] D. Mohl et al, Phys. Rep. 58 (1980) 73 their thermal energy to electrons. Ultimately, in
[5] J. Bisognano, C. Leemann, AIP Proc. 87, p.583 the absence of other heating mechanisms, the ion
(1981) temperature in the beam frame (BF) of reference
[6] John Marriner, David McGinnis, AIP conf. 249 becomes equal to the effective electron tempera-
p.693 (1992) ture. As a result, the ion beam angular spread,
[7] See e.g. S. Nagaitsev, R.J. Pasquinelli eds., AIP θi , becomes substantially smaller than the elec-
conf. 821 (2005); COOL07 tron beam angular spread, θe ,
[8] R. Pasquinelli, PAC95, p.2379
[9] M. Blaskiewicz, J.M. Brennan, F. Severino, PRL p⊥ m
100 174802 (2008) θi = ≈ θe (1)
p0 M
[10] M. Blaskiewicz et al, IPAC11, p.913 (2011)
[11] J. Wei, CERN 94-03 (1994) where m and M are the electron and ion masses
[12] D. Broemmelsiek et al, PRST-AB 8, 124404 respectively. Early theoretical investigations of
(2005) the electron cooling process [2] demonstrated the
[13] S. Chattopadhyay, LBL-14826 (PhD thesis) importance of achieving low electron beam tem-
(1982); also several papers in PAC83 perature (angular spread).
202
Ch.2: BEAM DYNAMICS
Friction force In a simplified electron-ion

plasma model, an ion traveling through the elec-
tron “gas” experiences a friction force

+∞
2 2 ve −v 3
F(v) = 4πne m(Zre c ) Λ f (ve ) d ve
|ve −v|3
−∞
Figure 1: Schematic diagram of the NAP-M electron
(2)
cooling system: 1 – electron gun, 2 – solenoids, 3 –
with ne the BF electron density, Λ is the Coulomb
anodes of the electron gun, 4 – toroids, 5 – cooling
log (≈ 5-10) and f (ve ) is the electron velocity
section, 6 – vacuum chamber, 7 – electron beam col-
distribution function in the BF. For example, an lector, 8 – vacuum pumps, and 9 – dipole correctors.
anisotropic Gaussian distribution function has
2

1 2
ve⊥ ve
f (ve ) = exp − − (3) To express these values through LF quantities one
3
2σ⊥2 2σ2
2σ
(2π) 2 σ⊥
has to use proper Lorentz transformations and to
recall that the cooling section occupies only a
with σ⊥ and σ the BF transverse and longitudi-
fraction, η, of the ring circumference. Thus we
nal rms velocity spreads, which can be expressed
obtain the following cooling rates:
through laboratory-frame (LF) quantities as
δE 2 Ie Zre 2 m mc2
σ⊥ = βγθe c , σ ≈ (4) λ = 4 ηΛ (9)
βγmc π e a M β γ 3 θe2 δE
2
2
where θe is the LF rms electron beam angular √ Ie Zre m 1
spread and δE is the LF rms energy spread of λ⊥ = 2πηΛ (10)
e a M β 4 γ 5 θe3
the electron beam. Typically σ⊥ σ lead-
ing to what is called a “flattened” electron dis- where Ie and a are the electron beam current
tribution [3]. The electron beam angular spread and radius. A uniform electron density distribu-
is mainly determined by (a) the cathode temper- tion is assumed. A numerical example: consider
ature, (b) the electron beam space charge and (c) 100 MeV protons being cooled by a 54.5-keV, 1-
the transport line imperfections. The electron en- A electron beam with a 1 cm radius and assume
ergy spread has three major contributions: (a) the that the electron beam rms energy spread is 10 eV
energy ripple, (b) the multiple-Coulomb scatter- and its rms angular spread is 2 mrad. Suppose
ing, and (c) the electron beam density fluctuations that η = 0.02 and Λ = 10. The cooling times
(aka the longitudinal-longitudinal relaxation). are 0.13 and 4.5 s correspondingly. For large ion
Approximate expressions for the cooling beam angular spreads, λ ∝ (θi )−3 .
force (2) can be found in [3, 4]. Consider two To obtain a more complete picture of the cool-
simplified cases when an ion has only one coming process one needs to average Eq.(2) over the
ponent of its velocity vector either longitudinal or ion beam distribution, taking into account both the
transverse: v σ and v⊥ σ⊥ . We have betatron and synchrotron motion.
(σ⊥ σ ) Magnetic field Electron cooling was first real-
ized in 1974 in experiments on cooling 65-MeV
√ v c2
F ≈ −4 2πne m (Zre c)2 Λ 2 (5) protons at the NAP-M storage ring at BINP [5].
σ⊥ σ Fig.1 shows the schematic of the system.
An electron gun was placed into a solenoid
√ v⊥ c2
F⊥ ≈ −π 2πne m (Zre c)2 Λ 3 (6) producing the longitudinal guiding magnetic
σ⊥ field, which accompanies the beam until it reaches
Cooling rates In these cases the friction forces the collector. The longitudinal magnetic field en-
(5) and (6) are proportional to the ion velocity and abled passing an electron beam over the required,
thus the particle loses its thermal energy in the BF rather long, (a few meters) distance without dilu-
exponentially with a time constant tion. Moreover, it allowed to form an intense elec-
√ m c3 tron beam with a small angular spread. The lon-
τ−1 = 4 2πne (Zre )2 cΛ 2σ (7) gitudinal magnetic field transforms the electron
M σ⊥ motion, perpendicular to it, into gyrations about
√ m c3 the magnetic field lines. Cooling by “magnetized”
τ⊥−1 = π 2πne (Zre )2 cΛ (8)
M σ⊥3 electrons differs substantially from the simplified
203
Figure 2: Schematic layout of the Recycler electron cooling system and accelerator cross-section (insert).
description of Eq.(2): it makes electron cooling exceeding in energy of “low-energy” coolers by

faster (at small relative velocities) and the equi- more than an order of magnitude. Fig.2 shows the
librium beam temperatures lower. The reason is schematic layout of the Fermilab system. The dc
as follows. The Larmor radius of the electron gy- electron beam is generated by a thermionic gun,
ration can be smaller than the impact parameter of located in the high-voltage terminal of the electro-
the electron – ion collision. This makes the elec- static (Van-de-Graaff type) accelerator. This ac-
tron interact with an ion longer and thus increases celerator is incapable of sustaining dc beam cur-
the friction force substantially. The effect was first rents to ground in excess of about 100 μA. To at-
discovered experimentally at NAP-M [6] and then tain the electron dc current of 500 mA, a recircu-
studied theoretically [7, 8, 9]. A dedicated ex- lation scheme is employed. A typical inefficiency
perimental study of this effect was performed at of such a process is 20 ppm.
BINP with a single-pass setup MOSOL [10]. A The Fermilab system employs a unique beam
factor of > 100 enhancement in the cooling force transport scheme [15]. The electron gun is im-
at small relative velocities was observed. These mersed in a solenoidal magnetic field, which cre-
studies lead to the realization that quality (field ates a beam with large angular momentum. Af-
line straightness) of the solenoidal field is critical ter the beam is extracted from the magnetic field
to fast magnetized cooling. A typical field uni- and accelerated to 4.3 MeV, it is transported to
formity achieved in state-of-art electron coolers is the 20-m long cooling section solenoid using con-
ΔB⊥ /B < 10−4 at a field level of B ∼1 kG. ventional focusing elements. The cooling section
Practical realizations Following the NAP-M solenoid removes this angular momentum and the
success, a number of coolers were built in 1970- beam is made round and parallel such that the
90 [3, 11, 12] with key features similar to that beam radius, a, produces the same magnetic flux,
of the first cooler Fig.1. The required electron Ba2 , as at the cathode. The magnetic field in
beam currents of ∼1 A and beam energies of the cooling section is quite weak (100 G) and
up to 300 keV demanded a beam recirculation the kinetics of the electron-antiproton scattering
scheme in which the electron beam is decelerated is weakly affected by the magnetic field.
in the collector, giving most (> 99%) of its en- Applications In elementary particle physics,
ergy back to the high-voltage power supply. In the most important application of electron (as well
practice, this allowed achieving very low energy as stochastic) cooling is in accumulation and stor-
ripple (< 100 ppm) and acceptable vacuum gas age of antiprotons destined for the colliding beam
loads in the cooling section. A description of the experiments. In nuclear physics the appearance
recently built LEIR electron cooler can be found of coolers has revealed qualitatively new possibil-
in [13]. ities: precision experiments with extremely small
Relativistic electron cooling As of now, Fer- (down to 10−6 ) relative energy spreads, thin inter-
milab’s Electron Cooler [14] remains unique in nal targets, mass spectrometry of super-high reso-
its ability to cool relativistic (8 GeV) antiprotons, lution and the separation and storage of long lived
204
Ch.2: BEAM DYNAMICS
isotopes and isomers [16]. Highlights include

using electron cooling to prepare dense bunches v0 v 0 +v r
hν
of lead ions for ion-ion collisions at the LHC (
!
) (
#
#
$
* + ,
[13] and very demanding cooling schemes for the

FAIR project at GSI [17]. In atomic physics the
electron beam can be used as the “target”, allow-

ing for a precision study of ion recombination. Fi-

nally, electron cooling has led to the development

of ordered (crystalline) ion beams [18, 19].

Isotropic
References <v>=v 0 +vr
[1] G.I. Budker, At. Energy 22, 346 (1967) [Sov.

Atomic Energy 22, 438 (1967)]
[2] Ya.S. Derbenev, A.N. Skrinsky, PA 8, 1 (1977)

[3] I.N. Meshkov, Phys. Part. Nucl. 25(6), 631 Figure 1: Mechanism of absorption and spontaneous
(1994) emission of photons in a laser cooling process.
[4] H. Poth, Phys. Reports 196 (1990) 135
[5] G.I. Budker et al, PA 7, 197 (1976) by irradiation of the co-propagating or counter-
[6] G.I. Budker et al, USSR High Energy Ac- propagating laser with the Doppler shifted fre-
cel. Conf., 1976; translation CERN/PS/DL/Note quency, ν, satisfying such a condition as hν co-
76-25 incides with the band gap between upper state, B
[7] Ya.S. Derbenev, A.N. Skrinsky, PA 8, 235 (1977)
and lower state, A. The ion obtains the momen-
[8] M. Bell, PA 10, 101 (1980)
tum of hν/c in the direction of laser irradiation by
[9] V.V. Parkhomchuk, NIM A 441, 9 (2000)
absorption of the photon. Thus the ion is slightly
[10] N.S. Dikansky et al, Preprint INP 88-61, 1988
[11] M. Bell et al, NIM 190, 237 (1981)
accelerated (co-propagating case) or decelerated
[12] T. Ellison et al, IEEE Trans. Nucl. Sci. NS-30, (counter-propagating case). If the lifetime of the
2370 (1983) excited state is short enough, the ion immediately
[13] G. Tranquille et al, COOL07 returns back to the ground state by emitting a pho-
[14] S. Nagaitsev et al, PRL 96, 044801 (2006) ton. Such a spontaneous emission process occurs
[15] A. Burov et al, PRST-AB, 094002 (2000) isotropically and recoil momentum given to the
[16] H.-J. Kluge et al, NIM A 532, 48 (2004) ion is statistically zero. So after the excitation and
[17] http://www.gsi.de/fair/ de-excitation processes Fig.1, the ion on the aver-
[18] M. Steck et al, PRL 77, 3803 (1996) age only receives a small velocity gain in the laser
[19] H. Danared et al, PRL 88, 174801 (2002) direction keeping other conditions unchanged. In
such a case as a co-propagating laser is irradiated
and the laser frequency is at first detuned to red
2.7.3 Laser Cooling (longer wavelength) side and then swept toward
J. Hangst, CERN the resonant frequency, the velocity distribution
A. Noda, Kyoto U. of the ion can be modified as shown in Fig.2 cre-
Laser cooling [1], where an ion is excited by ating a sharp peak in the velocity distribution.
absorption of a photon and returns back to the The laser cooling force of a single laser in its
ground state by spontaneous emission of a pho- propagation direction is [3]
ton, has been successfully applied for trapping k ( Γ2 )2
ions and cooling them to a temperature below F1 (v , k) = SΓ (1)
2 (Δ−v · k)2 +( Γ )2 (1+S) 2
1 mK [2]. A moving beam in a storage ring is
typically hot and its constituent particles have a where , Γ = 1/τ , Δ = ω − ω0 , v , k and S are
large fluctuation in kinetic energies in the CM sys- reduced Planck constant, the natural linewidth of
tem. This fluctuation can be reduced by absorp- the cooling transition, the angular frequency de-
tion and spontaneous emission of the photons. As tuning of the laser relative to the resonance, ion
shown in Fig.1, an ion in a lower energy state, velocity, the wave vector of the laser and the sat-
A, moving with a velocity v0 in a certain direc- uration parameter, S = 2( μE 2
Γ ) (E: laser electric
tion, will get excited to an upper energy state, B, field, μ: dipole moment for the transition).
205
Figure 2: (a) Irradiation of co-propagating laser, re-

sults in acceleration of the ions in the velocity range.
(b) The ion is given the energy as large as hν from the
laser and corresponding velocity change of the ion oc- Figure 4: Total cooling force given by a combined sys-
curs. (c) After sweeping the laser frequency through tem of two lasers co- and counter-propagating with the
the whole velocity range of the ion beam, ions are col- ion beam.
lected in a sharp velocity range.

Table 1: Main parameters of laser cooling of moving
ions.
and the corresponding damping time is ∼10−6 s

Figure 3: Total cooling force superposing the laser
for such ion beams as 7 Li+ , 9 Be+ and 24 Mg+ ,
cooling force and the induction voltage by INDAC.
ever laser cooled. The equilibrium temperature
Stable point and capture region are indicated.
of laser cooling is determined by the balance of
the cooling force and other heating mechanisms.
For laser cooling with a single laser, a coun- Without external heating, equilibrium is estab-
teracting force to the above laser cooling force is lished between the cooling force and the spon-
needed in order to keep the ion beam at a cer- taneous emission, and is known as the Doppler
tain stable orbit. Such a force is provided with limit,
an induction accelerator or rf acceleration voltage Γ
for the cases of coasting or bunched beams, re- TD = (4)
2kB
spectively. In Fig.3, total cooling force created by
where kB is the Boltzmann constant (8.62 ×
combination of a single laser and a counteracting
10−5 eV/K).
deceleration voltage applied through an Induction
The Doppler limit for laser cooled ions in
Accelerator (INDAC) is given together with an in-
a storage ring is listed in Tab.1 together with
dication of stable point.
other main parameters. They are in the range of
For two counter propagating lasers with the
10−4 ∼10−3 K.
same intensity and detuning, the total cooling
force is given by superposition (Fig.4), Experiments Laser cooling has been first ap-
plied to a moving beam at the TSR, MPI-K,
2 (v , k) = F1 (v , k) + F1 (v , −k)
F (2) Heidelberg for coasting 7 Li+ ion beam [4]. Its
In the small velocity region, the force can be ap- metastable 2s 3 S1 (F=5/2) state is excited to the
proximated by a linear function of the relative ve- upper level 2p 3 P2 (F=7/2) with a mean radia-
locity, tive lifetime of 43 ns, which has a branching ra-
dv dF tio less than 10−5 to the ground state 1 S0 . The
F =m ≈ −βv (β = |v=0 ) (3)
dt dv lower metastable 3 S1 state has a radiative life-
From Eq.(3), the velocity v is known to be rep- time of 50 s, long enough compared to ion stor-
resented as an exponential function of −(β/m)t age life time ∼2s. Thus, the 3 S1 (F=5/2) → 3 P2
206
Ch.2: BEAM DYNAMICS
IBS and cooling efficiency is not so good to ap-

proach a cold beam. Another scheme to utilize
longitudinal-horizontal coupling through ring dis-
persion (transverse laser cooling through disper-
sive coupling) has also been demonstrated at TSR.
This acts as a single-particle effect independent of
the ion density in contrast to IBS cooling. The
transverse cooling rate realized, however, is ∼1
s−1 , still far below typical longitudinal laser cool-
ing rates [9].
As an efficient way of horizontal and longi-
tudinal coupling, a scheme to utilize an rf cav-

ity located at a dispersive position in the ring
Figure 5: Scheme of bunched beam laser cooling.

has been proposed. The synchrotron motion and
Laser frequency detuning is swept from the bottom of
horizontal betatron oscillation are coupled by a
the separatrix to the center of resonance condition.
Synchro-betatron resonance [10]. In order to cou-
ple the horizontal and vertical directions, the usual
(F=7/2) transition approximately plays the role horizontal-vertical coupling near a difference res-
of two level system. Ring dye lasers co- and onance is applied using a solenoid or a skew
counter-propagating with the ion beam with fre- quadrupole magnet. The synchro-betatron cou-
quencies which coincide with the level difference pling scheme is now under experimental research
of 548.5 nm after Doppler shift, are utilized for at S-LSR, ICR, Kyoto Univ. and experimental
excitation of the ion, resulting in the longitudinal indication of such a resonant coupling between
temperature of 3 K. Laser cooling is applied also two degrees of freedom, longitudinal an horizon-
to 9 Be+ and 24 Mg+ ions at TSR and ASTRID, tal directions, has been obtained [11]. In this ex-
IPA in Aarhus for coasting and bunched beams, periment, the optical measurement system of the
respectively [5, 6]. Fig.5 illustrates the scheme transverse beam profile by detection of the laser-
of bunched beam laser cooling [6]. The laser fre- induced fluorescence (LIF) light with a CCD cam-
quency is red shifted to the lowest energy level era basically developed at ASTRID [12] played
covered by the rf separatrix and then the laser fre- an essential role. As shown in Fig.6, S-LSR is
quency is shifted to the central value of the sepa- designed to have a higher symmetry of 6 in order
ratrix according to the cooling requirement. The to satisfy the maintenance condition to keep beam
equilibrium temperature of the cooled beam is de- crystal structure stable:
termined by the balance between the laser cooling N
force and the heating process due to intra-beam νh,v ≤ √
scattering, which limits the attained beam temper- 2 2
ature around 1K for intense beam. where νh,v and N are betatron tunes in horizon-
tal and vertical directions and number of super-
3-D cooling Concerning longitudinal cooling, a periodicity of the ring [13]. As illustrated in
beam temperature less than 1 mK can be attained Fig.7, S-LSR also provides a capability to keep
for a small fraction (∼10−4 ) of 7 Li+ beam, which the crystalline beam structure to be stable sup-
is in a metastable 3 S1 state [7] and beam tempera- pressing shear heating by superposing an electro-
ture lower than 30 mK is attained for 9 Be+ beam static field, E, with the bending magnetic field, B,
which is pre-cooled by electron cooling [5]. At- satisfying the relation
tainable equilibrium temperature depends on the
1
ion beam density. (1 + )E = −v0 × B

In order to extend the laser cooling force to γ2
3-D, various approaches have been applied. At where γ and v0 are Lorentz factor and velocity of
TSR, transverse laser cooling with the use of the beam, respectively [14, 15].
intrabeam scattering has been demonstrated for Circulating crystalline ion beam (Sec.2.7.5)
9 Be+ ion beam [8]. The transverse cooling force, has been realized with a circular RFQ storage
however, depends on the passive energy transfer ring, PALLAS at LMU, Munich, with application
between transverse and longitudinal directions by of laser cooling to 24 Mg+ ion with the energy of
207
Figure 8: (a) Illustration of ion trapping with the rf

electric fields at a linear ion trap, and (b) realized
multi-layer ion crystal in a linear ion trap [21].
Figure 6: Layout of S-LSR. Six-fold symmetry is
adopted for its lattice to satisfy the so-called mainte-

nance condition.
Particle Trapping Laser cooling has already
been successfully applied to neutral atoms and
charged ions trapped inside a trap, which has been

utilized for spectroscopy, fundamental quantum
physics, quantum computation etc. Neutral atoms
are captured and trapped after laser cooling, in
many cases, with the aid of magnetic field confin-
ing them by magnetic moment [19]. An ion trap
is an apparatus for confining charged particles uti-
lizing electromagnetic forces, e.g. Penning trap,
Figure 7: Illustration of shear heating and the deflec- utilizing static electric and magnetic fields, and rf
tion element combining the electrostatic and magnetic trap, utilizing rf electric fields. Rf ion traps have
fields to suppress the shear heating adopted at S-LSR. configuration of 3-D quadrupoles (Paul Trap [20])
and a linear trap utilizing four rods with applica-
1 eV [16]. The crystal structure is found to be tion of rf electric fields to confine charged ions in
destroyed in case the ion energy is increased by radial direction as shown in Fig.8(a), while con-
acceleration to higher than 1 eV with a drift tube finement in the axial direction is attained by static
set inside the PALLAS due to shear heating. potentials applied to end plates. With a linear ion
Laser cooling of relativistic ions has also been trap, a crystalline beam of 40 Ca+ has already been
performed at ESR in GSI, Darmstadt applying the realized as shown in Fig.8(b) [21].
Ar+ ion laser with the wave length of 257.34 nm
to the highly Doppler shifted C3+ ions with the References
energy of 1.47 GeV (122 MeV/u) to excite the
transition from 2S1/2 to 2P1/2 (transition wave [1] T. Hänsch, A. Schawlow, Opt. Commun. 13, 68
length: 155.07 nm) and from 2S1/2 to 2P3/2 (tran- (1975)
sition wave length: 155.81 nm). Such a result of [2] S. Chu et al, PRL 55, 48 (1985)
laser cooling is to be applied at the FAIR project [3] J. Gordon, A. Ashkin, PRA A21, 1606 (1980)
[17]. [4] S. Schröder et al, PRL 64, 2901 (1990)
[5] W. Petrich et al, PRA A48, 2127 (1993)
Perspective Although the possibility to couple
[6] J.S. Hangst et al, PRL 74, 4432 (1995)
the transverse and longitudinal motions dynami-
[7] J.S. Hangst et al, PRL 67, 1238 (1991)
cally in order to “share” the laser cooling force [8] H.J. Meisner et al, PRL 77, 623 (1996)
among three dimensions [10] has been shown ex- [9] I. Lauer et al., PRL 81, 2052 (1998)
perimentally, its efficiency is not so high as be [10] H. Okamoto, A. Sessler, D. Möhl, PRL 72, 3977
able to realize crystalline beam for the rather in- (1994)
tense beam (∼107 per ring) [11]. Experimental [11] M. Nakao et al, to be published
approach to increase the transverse cooling ef- [12] N. Madsen et al, PRL 83 (1999) 4301
ficiency by reducing the ion beam intensity to [13] J. Wei, X.-P. Li, A M. Sessler, PRL 73 (1994),
the level as MD simulation predict realization of 3089.; X.-P. Li et al, PRST-AB 9 (2006),
string beam (∼104 per ring) [18] is under way 034201
which requires a more sensitive transverse profile [14] W. Henneberg, Annalen der Physik (Leipzig) 19
detection system. (1934) 335; W.E. Millett PR 74 (1948) 1058
208
Ch.2: BEAM DYNAMICS
[15] A. Noda, M. Ikegami, T. Shirai, New Journal of

Physics 8, 288 (2006)
[16] T. Schätz, U. Schramm, D. Habs, Nature 412,
717 (2001)
[17] U. Schramm, D. Habs, Progress in Particle and
Nuclear Physics 53, 583 (2004).
[18] Y. Yuri, H. Okamoto, PRST-AB 8 (2005) 114201
[19] W.D. Phillips, RMP 70 (1998) 721 Small emittance
Large emittance
[20] V.W. Paul, H. Steinwedel, Zeitschrift für Natur-
forschung A8, 448 (1953) Absorber Accelerator
[21] K. Ito et al, IPAC’10, p.3622 Momentum loss is Momentum gain
opposite to motion, is purely longitudinal
p, p x, p y, ΔE decrease
2.7.4 Ionization Cooling

Figure 1: Concept of ionization cooling.

D. Neuffer, FNAL
In ionization cooling (μ-cooling), particles pass The equation for longitudinal cooling with
through a material medium and lose energy (mo- energy loss is
mentum) through ionization interactions, and this . /
is followed by beam reacceleration in rf cavities dσE2 ∂ dE d ΔErms2
= −2 ds 2
σ + (2)
(Fig.1). The losses are parallel to the particle mo- ds ∂E E dt
tion, and therefore include transverse and longi- The first term is the cooling term; the second is
tudinal momentum losses; reacceleration restores the heating term caused by random fluctuations in
only longitudinal momentum, and the transverse the particle energy loss. Beam cooling can occur
cooling results from that loss of transverse mo- if the derivative ∂(dE/ds)/∂E > 0. This energy
mentum. The cooling effect is opposed by the loss can be estimated by the Bethe-Bloch equation
random process of multiple scattering in the mate- (Sec.3.3.1),
rial medium, which increases the rms beam diver-
dE Z
gence and thereof re the emittance. This cooling = 4πNA ρ re2 me c2
method is not generally useful for protons, which ds A
would have frequent nuclear interactions, or elec- 1 2me c2 γ 2 β 2 δ
× ln − 1 − (3)
trons, which would have bremstrahlung, but is β2 I(Z) 2β 2
practical for muons, and cooling rates compatible where NA is Avogadro’s number, ρ, A and Z
with muon lifetime are possible. are the density, atomic weight and number of the
The differential equation for rms transverse absorbing material, me and re are the mass and
cooling is [1, 2, 3] classical radius of the electron, (4πNA re2 me c2 =
. 2 /
dεN 1 dE βγβ⊥ d θrms 0.3071 MeV cm2 /gm). The ionization constant
=− 2 εN + I(Z) is approximately 16Z 0.9 eV, and δ is the
ds β E ds 2 ds
density effect factor which is small for low-energy
g⊥ dE β⊥ Es2
=− 2 εN + 3 (1) muons. The derivative is negative (or naturally
β E ds 2β mμ c2 LR E heating) for Eμ < ∼ 0.3 GeV, and is only slightly
where the first term is the energy-loss cooling ef- positive (cooling) for higher energies.
fect and the second is the multiple-scattering heat- In the long-pathlength Gaussian-distribution
ing term, εN is the normalized emittance, E is limit, the second term in Eq.(2) is approximately
the beam energy, β and γ are the Lorentz fac- . /
d ΔErms 2 2 β2
tors, dE/ds is the energy loss rate, θrms is the = 4π re me c2 ne γ 2 1− (4)
rms multiple scattering angle, LR is the material ds 2
radiation length, β⊥ is the betatron function at where ne is the electron density in the material.
the absorber, and Es is the characteristic scatter- This expression increases rapidly with higher en-
ing energy (∼13.6 MeV). (see also Sec.3.3.1) The ergy (larger γ), opposing the cooling process. Af-
symbol g⊥ is introduced to include coupling with ter adding this energy straggling, ionization cool-
longitudinal cooling (see Eq.(7)); for uncoupled ing does not naturally provide adequate longitudi-
cooling g⊥ = 1. nal cooling.
209
A B
Σg
Figure 2: Overview of emittance exchange. A dis-

persion is introduced into the beam transport. A:
Wedge technique: A “wedge” absorber is inserted that Momentum Pμ (MeV/c)
presents more path length, and therefore more energy

loss, for higher-energy particles is. B: Homogeneous Figure 3: The sum of the cooling partition numbers
Absorber: The dispersion makes total path length en- Σg = (gx + gy + gL ) as a function of momentum Pμ
ergy dependent; an absorber that fills the transport then (0-500 MeV/c). gx and gy are naturally 1 while gL
obtains energy-dependent energy loss. becomes strongly negative for Pμ < 200 MeV/c. Σg
remains greater than 0, which means that ionization
loss remains intrinsically cooling at low momenta.
However, the cooling term can be enhanced
by placing the absorbers where transverse posi-
tion depends upon energy (nonzero dispersion) The longitudinal partition number is
dεL /ds @ ∂(dp/dt)
and where the absorber density or thickness also ∂( dE dp
εL ds ) ds ∂p
depends upon energy, such as in a wedge ab- gL = dp/ds = = (dp/dt) (8)
sorber. This makes the beam particle path length ∂E p
p p
through the absorber material energy dependent and is a function of muon energy. With δ = 0 in
(Fig.2). In that case the cooling derivative can be the energy loss formula we find
rewritten as
( 2(1 − βγ 2 )
2
∂ dE ∂ dE ( dE ηρ 2
ds ( gL = − + (9)
ds
⇒ ( + (5) γ2 c2 β 2 γ 2
∂E ∂E ( ds βcpρ0 ln 2meI(Z) − β2
0
where ρ"/ρ0 is the change in density with respect The sum of partition numbers Σg = gx +gy +gL is
to transverse position, ρ0 is the reference density 2
associated with dE/ds, and η is the dispersion. (1 − βγ 2 )

Σg = 2β 2 + 2 (10)
Increasing the longitudinal cooling rate in this c2 β 2 γ 2
ln 2meI(Z) − β2
manner decreases the transverse cooling by the
same amount. The transverse cooling term is This sum is a function of muon momentum, and
changed to is displayed in Fig.3 (evaluated at H2 absorber
parameters). Σg is approximately 2 for pμ >
dεN 1 dE ηρ
=− 2 1− εN (6) 0.3 GeV/c, but is smaller for lower energies.
ds β E ds ρ0 However, Σg does remain positive for all ener-
Note that the coupled transverse cooling (and gies, which indicates that cooling remains possi-
heating) changes occur in the same direction ble even at low muon energies.
(i.e. horizontal or vertical) as the dispersion and With (horizontal) wedge enhancement of lon-
wedge. However the sum of the cooling rates gitudinal cooling, gL increases by ηρ"/ρ0 , and gx
(over x, y, and z) remains constant. This sum can becomes gx = 1 − ηρ /ρ0 , leaving the sum of
be represented, as with radiation damping, as a the partition numbers Σg constant. (This coupling
sum of cooling partition numbers, defined as the also mixes the heating terms; in initial approxima-
ratios of the cooling rate to the fractional momen- tions we neglect this complication.)
tum loss rate. For x and y emittance cooling the The longitudinal cooling Eq.(2) only tracks
partition numbers are both naturally 1, energy spread. It can be transformed into a longi-
? tudinal emittance cooling equation by adding lon-
gy = gx = dεxεx/ds dp/ds
p = 1 (7) gitudinal rf focusing that places the beam within
210
Ch.2: BEAM DYNAMICS
a bunch, coupling of x, y, and z motion with integrated ab-

. 2
/ sorbers enables efficient 3-D cooling [6]. Parti-
dεL gL dE βz d ΔErms
=− 2 εL + (11) cles follow helical orbits at a radius a and period
ds β E ds 2 ds
λ, where a depends on the particle momentum p,
Here βz is a focusing function, defined by following√
. 2/
δz 1 λ0 αp 1 + k2 a2 eBh (1 + k2 a2 )
βz2 = = 3 (12) p(a) = eBs −
ΔE 2 β γeV sin φs 2π mc2 k ka
where λ0 is the focusing rf wavelength, eV sin φs where k = 2π/λ is the wave number of the helical
is the mean focusing rf gradient, and αp is the mo- period, and Bs and Bh are the solenoidal and heli-
mentum compaction. (αp = 1/γ 2 − 1/γt2 in a cal magnetic fields on the reference orbit, respec-
synchrotron and αp = 1/γ 2 in a linac.) The lon- tively. The dispersion η = p da/dp can be calcu-
gitudinal bunch emittance εL is in δE-δcτ units. lated from the equation, including the divergence
of Bs and Bh with respect to radius. By embed-
Guidelines for optimum cooling can be ob-

tained from Eqs.(1)-(4) and the partition func- ding a homogeneous dense hydrogen gas and rf
tions. It is desirable to obtain small β⊥ (from cavities into such a continuous dispersion mag-
strong focusing) in the absorbers. Strong focusing net, longitudinal phase space cooling takes place
systems with high-field solenoids or quads or cur- without a wedge absorber as shown in Fig.2, since
rent carrying lenses (Li or Be lens [4], Sec.7.2.12, the path-length is momentum dependent.
where the focusing element is also the energy Simulation codes that track the full dynam-
loss medium) are needed. To minimize multi- ics of ionization cooling have been developed
ple scattering emittance dilution, materials with (ICOOL [7] and G4BeamLine [8]). An inter-
large values of the product LR dE/ds are pre- national collaboration (MICE) is constructing a
ferred, and this is maximal for light element ab- cooling section with absorbers, rf and solenoids
sorbers. Eq.(4) indicates that energy straggling at RAL and will test it in a muon beam to deter-
increases greatly with high energies, while the mine the practicality of ionization cooling [9].
partition function becomes small for low energies Scenarios that can obtain ionization cooling
(Fig.2). An optimum for cooling with minimal of muons by the large factors needed for μ+ − μ−
heating occurs at intermediate values; i.e., γ ≈ 3 colliders (Sec.1.6.15) are under investigation [10].
(pμ ≈ 300 MeV/c or Bρ ≈ 1 T-m), where the In those studies transverse cooling from εN ≈
partition number sum is ∼2. 0.015 to 0.000025 m-rad, and longitudinal cool-
The above formalism assumes uncoupled x ing by an order of magnitude are required. Sce-
and y motion, and ignores potentially compli- narios for muon storage ring based neutrino fac-
cating higher-order effects. Solenoid mixes x tories use cooling by much smaller factors (i.e.,
and y and complicates the motion with angular transverse cooling by a factor of ∼10) [11].
momentum effects. Wang and Kim have devel- It has also been noted that ionization cool-
oped rms equations for ionization cooling with ing beam dynamics can be useful in low-energy
quad and solenoid focusing as well as disper- ion and proton storage rings, increasing the life-
sion with wedge focusing, and these include cou- time of beams that are stored for multi-turn pas-
pling terms in heating and cooling, and angular sages through thin targets, and thus enhancing the
momentum damping [5]. The additional terms nuclear production of useful low-energy secon-
are small in “well-designed” cooling transports daries, such as neutrons for medical treatment and
(with periodical solenoid field flips and optimized unstable nuclei for beta-beams [12].
wedge/dispersion insertions), and the 1-D cool-
ing Eqs.(1) and (11) remain good first approxima- References
tions. In solenoid focusing systems with disper- [1] A.N. Skrinsky, V.V. Parkhomchuk, Sov. J. Nucl.
sion/wedge longitudinal cooling, the x and y mo- Physics 12, 3 (1981)
tions can be tightly coupled with approximately [2] D. Neuffer, PA 14, 75 (1983)
equal focusing and cooling; the transverse anti- [3] D. Neuffer, NIM A 532, 26 (2004)
damping due to the dispersion/wedge is equally [4] B.F. Bayanov et al, NIM 190, 9 (1981)
split between x and y. [5] C.X. Wang, K.J. Kim, NIM A 532, 260 (2004)
Derbenev and Johnson have designed a heli- [6] Y. Derbenev, R.P. Johnson, PRST-AB 8, e041002
cal solenoid transport with rf cavities, where tight (2005)
211
[7] R. Fernow, PAC 1999, p.3020, Measure dimensions in units of the characteristic
http://pubweb.bnl.gov/people/fernow/ distance ξ with ξ 3 = r0 ρ2 /β 2 γ 2 , time in units of
[8] T. Roberts et al, (2009), ρ/βγc, and energy in units of β 2 γ 2 Z02 e2 /4π0 ξ,
http:// g4beamline.muonsinc.com where r0 = Z02 e2 /4π0 M0 c2 is the classical ra-
[9] R. Sandstrom, MICE Collab., AIP Conf. Proc. dius, βc and γM0 c2 are the velocity and energy of
981:107 (2008) the reference particle, and ρ is the radius of cur-
[10] R. Palmer et al, PAC 2009 vature in bending regions of magnetic field B0 . In
[11] J.S. Berg et al, PRST-AB 9, 011001 (2006) a bending region with pure dipole magnetic field,
[12] Y. Mori, NIM A562, 591 (2006); C. Rubbia, et the Hamiltonian for particles of the ith species is
al, NIM A568, 475 (2006); D. Neuffer, NIM
A585, 109 (2008) 1 2
Hi = P + P 2 + Pz2 (2)
2 x y
1 2 Z̄i
2.7.5 Crystalline Beam + x + − 1 γ 2 z 2 + VCi
J. Wei, Michigan State U. 2 mi

Z̄i Z̄i ρ
Since mid 1980’s, there has been experimental − 2− γxPz + −1 x
[1, 2] and theoretical [3] efforts to achieve crys- mi mi ξ
talline beams. When the beams in storage rings where the Coulomb potential is
are sufficiently cold in the beam rest frame, the Z̄i Z̄j /mi
ions may “lock into” a position where the re- VCi= (3)
pelling Coulomb force on the average balances j
(x j −x)2 +(y −y)2 +(z −z)2
j j
the external focusing force. The interest, besides
The summation, j, is over all the other particles.
intrinsically on this new state of matter, is primar-
In a non-bending region with longitudinal electric
ily on studying the physics of completely space-
field and non-dipole magnetic fields,
charge dominated beams, and the possibility of
obtaining high luminosity in colliders [4]. 1 2 n1 Z̄i 2 2
Hi = P +Py2 +Pz2 − (x −y ) (4)
Conditions of crystallization There are two 2 x 2 mi
necessary conditions [5] to form and maintain a Z̄i n2 ξ Z̄i 3
multidimensional crystalline beam: (i) The ring is − n1s xy− (x −3xy 2 )+VCi +Us
mi 6 mi
alternating-gradient (AG) focusing operating be-
low transition, where the quadrupole, skew quadrupole, and
√ and (ii) The ring lattice periodicity sextupole strengths are represented by n1 =
is at least 2 2 as high as the maximum betatron
tune. Condition (i) arises from the criterion of sta- −(ρ/B0 )(∂By /∂x), n1s = −(ρ/B0 )(∂By /∂y),
ble kinematic motion under Coulomb interaction n2 = −(ρ/B0 )(∂ 2 By /∂x2 ), respectively, the
when particles are subject to bending in a storage electrical force Fs is expressed in terms of elec-
ring. Condition (ii) arises from the criterion that tric field Es in the laboratory frame,

there is no linear resonance between the phonon ∂Us Z̄i Z0 eEs ξ ρ 2
modes of the crystalline structure and the machine Fs ≡ − = (5)
∂z mi M0 c2 ξβγ
lattice periodicity. A ring lattice periodicity as
high as 4 times the maximum betatron tune is pre- In the usual case of a single species of ion, Z̄i =
ferred so that linear resonance stopbands are not mi = 1.
crossed during the cooling process as the beam Numerical methods Numerical study of the
density is increased [6]. Condition (ii) is relaxed crystalline state has been performed with the
for the ordering of 1-D states in the azimuthal di- molecular dynamics (MD) method [3, 5, 9] it-
rection [7, 8]. erating the equations of motion derived from
Beam rest-frame Hamiltonian [5] Consider a the Hamiltonian. Ewald-type summation is per-
system of multi-species of ions using the rest formed in the azimuthal direction to evaluate the
frame (x, y, z, t) of a circulating reference par- long-ranged Coulomb forces among particles and
ticle. Define the reference particle with electric their image charges modelled in periodic “super-
charge Z0 e and atomic mass M0 , and define for cells” for computing efficiency.
the ith species of ions with charge Zi e and mass Ground-state In a crystalline ground state, the
Mi , motion of the circulating particles is periodic in
Z̄i ≡ Zi /Z0 , and mi ≡ Mi /M0 (1) time [5] with the period of the machine lattice.
212
Ch.2: BEAM DYNAMICS
B D 20
F
8
6 10
x, y, z (ξ)
4 x
y (ξ)
y
0
2 z
0
-10
-2
0.0 0.2 0.4 0.6 0.8 1.0
Time (FODO periods) -20
RF -20 -10 0 10 20
4 x (ξ)
2 Figure 2: A multi-shell structure with particle posi-

tions projected into the x-y plane (λ = 25γξ −1 ).
Px, y, z
0
Px
-2 Py
0
Pz 10
Heating Rate, ΔT/T per lattice period

-1
-4 10
0.0 0.2 0.4 0.6 0.8 1.0
Time (FODO periods) 10
-2
-3
10
Figure 1: Particle trajectory of a bunched crystalline
-4
beam. The machine consists of 10 FODO cells with 10 IBS scaling
-1
νx = 2.8, νy = 2.1, and γ = 1.4. Lattice compo- 10
-5
λ=25 γξ
-1
λ=1 γξ
nents in each cell are displayed on the figure: B is a 10
-6 -1
λ=1 γξ (refined time step)
-1
λ=0.25γξ
bending section, F and D are focusing and de-focusing 10
-7
-4 -3 -2 -1 0 1 2 3
10 10 10 10 10 10 10 10
quadrupoles, and RF is the bunching rf cavity. Temperature, T
Figure 3: Typical heating rates as functions of temper-

As shown in Fig.1, particle trajectory in the trans-
ature obtained by MD simulation at various line densi-
verse direction conforms to AG focusing (breath- ties λ.
ing), and in the longitudinal direction conforms
to the change in bending radius (shear). In the
presence of a longitudinal electric field, momen- nonzero temperature the crystalline beam absorbs
tum Pz also varies periodically conforming to the energy and heats up under time-dependent exter-
energy gain at the cavity. nal forces caused by variations in lattice focus-
The ground state structure is a 1-D chain ing and bending. In the high temperature limit,
when the beam line density is low [5, 9]. The this intrabeam scattering (Sec.2.4.12) results in a
structure becomes 2-D lying in the plane of growth rate ∝ λT −5/2 for coasting beams and
weaker transverse focusing if the line density λ ∝ λT −3 for bunched beams, where the normal-
in the machine is ized temperature T (with its components defined
1/3 as the deviation of Px , Py and Pz from their
λ > 0.62γξ −1 min(νy2 , νx2 − γ 2 ) (6) ground-state values, squared and averaged over
where νx,y are the transverse tunes. For even particles) is related to the conventional beam tem-
higher density, the particles arrange themselves perature TB at high temperature by
into 3-D crystals, becoming helices and then he- 2kB ρ2
T ≈ 2 2 TB (7)
lices within helices. Fig.2 shows such a multi- β γ M0 c2 ξ 2
shell structure at a defocusing location of the with kB the Boltzmann constant. The peak heat-
lattice. The average ratio between the hor- ing rate occurs at the temperature of about T ≈ 1
izontal and vertical particle spacing is about when the ordering starts to occur, as shown in
[νy2 /(νx2 − γ 2 )]1/3 . The maximum spatial den- Fig.3.
sity in the laboratory frame is approximately Typically, strong spacial correlation appears
γνy νx2 − γ 2 /(2ξ 3 ). If a sinusoidal electric field in all directions when the temperature T < ∼ 0.05.
is present, the crystalline structure can be bunched Lattice shear and AG focusing have similar effects
azimuthally. on beam heating. Heating behavior is similar for
Lattice heating A crystalline beam in its both bunched and coasting beams. Effects of ma-
ground state, despite breathing and shear motion, chine lattice imperfection, ion neutralization, and
remains in the zero-temperature state [5]. At any envelope instability have been studied.
213
Cooling methods To attain a crystalline state, ized in storage rings of moderate circumference.
the beam must be effectively cooled in 3-D with High-density 1-D strings were proposed for high-
a sufficient speed to overcome the heating. Both luminosity ion-electron collisions with rare ra-
electron (Sec.2.7.2) and laser (Sec.2.7.3) cooling dioactive ions [8, 22].
provide high cooling efficiency in the longitudinal
direction, reaching a beam temperature of less References
than 1 K, but not in the transverse directions
(around 100 K) [2]. “Sympathetic cooling” due [1] E.E. Dement’ev et al, Zh. Tekh. Fiz. 50 (1980)
1717; N.S. Dikanskiĭ, D.V. Pestrikov, Proc.
to intrabeam scattering does produce transverse
Workshop on Electron Cooling and Related Ap-
cooling [10], but the heat exchange becomes in-
plications, KfK 3846 (1984); V.V. Parkhomchuk,
effective as the beam approaches an ordered state.
A.H. Skrinsky, Reports on Progress in Physics,
Coupling cavities operating on a synchrobetatron 54 (1991) 919

resonance or regular rf cavities in a dispersive re- [2] S. Schroder et al, PRL 64 (1990) 2901; J.S.
gion can provide effective 3-D cooling [11], but Hangst et al, PRL 67 (1991) 1238
the coupling mechanism ceases to work before an [3] J.P. Schiffer, P. Kienle, Z. Phys. A 321 (1985)
ordering can be reached. Realization of multidi- 181; A. Rahman, J.P. Schiffer, PRL 57 (1986)
mensional crystalline beams requires cooling that 1133
provides the ions with constant angular velocity, [4] J. Wei, A.M. Sessler, EPAC 98, p.862; J. Wei et
rather than constant linear velocity (so-called ta- al, COOL 07, p.91
pered cooling) [12]. [5] J. Wei, X-P. Li, A.M. Sessler, PRL 73 (1994)
Experiments Experimentally, 3-D ion crystal- 3089; BNL-52381 (1993)
lization has been observed in rf traps [13], static [6] I. Hofmann, J. Struckmeier, Proc. Workshop on
[14] traps, and rf quadrupole rings [15] using laser Crystalline Ion Beams (1988) p.140; B. Yang
et al, Phys. Plasmas, 3 (1996) 688; Y. Yuri, H.
cooling (Sec.2.7.3). Ion traps were used to ex-
Okamoto, PRST-AB, 8, 114201 (2005)
perimentally simulate features of an AG-focusing
[7] R. Hasse, PRL 83 (1999) 3430
storage ring [16]. “Moving” crystals were gener-
[8] I. Meshkov et al, RIKEN Report: RIKEN- AF-
ated in a ring-shaped Paul trap system [17]. AC-34 (2002)
In storage rings of sizable circumference with [9] R.W. Hasse, J.P. Schiffer, Ann. Phys. 203 (1990)
beams of significantly high energy, 3-D crystal- 419
lization has not been successful due to the lack of [10] H.-J. Miesner et al, PRL 77 (1996) 623
effective transverse cooling, and the lack of ring [11] H. Okamoto, A.M. Sessler, D. Möhl, PRL 72
lattice superperiodicity. Evidence of anomaly in (1994) 3977
the Schottky signal was observed on an electron- [12] J. Wei, H. Okamoto, A.M. Sessler, PRL 80
cooled proton beam at NAPM [1]. 1-D ordering (1998) 2606; H. Okamoto, J. Wei, PR E58, 3817
were reported using electron cooling with both (1998)
proton and heavier ions at storage rings ESR, [13] F. Diedrich et al, PRL 59 (1987) 2931; D.J.
[18] SIS, CRYRING, and S-LSR. Shear-free ring Wineland et al, PRL 59 (1987) 2935
lattices consisting of both magnets and electrodes [14] S.L. Gilbert et al, PRL 60 (1988) 2022
were proposed [19] and constructed [20]. Fur- [15] H. Walther, Proc. Workshop on Light Induced
ther efforts include developing storage rings of Kinetic Effects on Atoms, Ions and Molecules
high lattice super-periodicity [21], attempting (1991) p.261
ultra-low temperature (or emittance) three- [16] H. Okamoto et al, Proc. 10th Inter. Comp. Accel.
dimensionally using laser cooling and transverse- Conf. (2009, to be published); S. Ohtsubo et al,
longitudinal resonance coupling, and exploring PRST-AB 13, 044201 (2010)
tapered cooling. [17] T. Schätz, U. Schramm, D. Habs, Natur (London)
412, 717 (2001); U. Schramm, M. Bussmann, D.
Applications The crystalline states correspond Habs, NIM A 532, 348 (2004)
to ultralow emittance states of charged-particle [18] M. Steck et al, PRL 77, 3803 (1996)
beams. Ordered multidimensional beams were [19] R. E. Pollock, Z. Phys. A 341, 95 (1991)
proposed for ion-ion colliders for increased lu- [20] M. Ikegami et al, PRST-AB 7, 120101 (2004)
minosity [4]. Machine lattices of high or imag- [21] A. Nada, NIM A532, 150 (2004)
inary transition energy were proposed so that [22] T. Katayama, D. Möhl, RIKEN Report RIKEN-
high-energy or colliding crystals may be real- AF-AC-39 (2002)
214
Chapter 3. ELECTROMAGNETIC AND
NUCLEAR INTERACTIONS
3.1 SYNCHROTRON RADIATION Specific radiation characteristics are determined

by the particular temporal variation of β and n̂
3.1.1 Radiation of a Point Charge along the particle path.
H. Wiedemann, Stanford U./SSRL Radiation power The total radiation power per
particle in its own system is
Radiation fields Electromagnetic radiation
emitted from relativistic charged particles, es- 2re me c2 ∗2
pecially electrons, as they are accelerated is P∗ = β̇ (3)
3c
called synchrotron radiation. Radiation fields, at and in the laboratory system
= −∂A
location P , E ∂t
− ∇φ and B = ∇×A 2
with retarded Lienard-Wiechert potentials 2re me c2 2 d 2 dγ
( P = γ γβ − (4)
( 3c dt dt
1 e β (
A(P, t) = ( Defining acceleration vectors parallel and or-
4π0 c R 1 + n̂ · β (
(r ˙ ˙ ˙
( thogonal to the particle motion β = β + β⊥ splits
1 e 1 (
φ(P, t) = (1) the total radiation power into parts due to lon-
4π0 R 1 + n̂ · β (r
gitudinal and transverse acceleration respectively.
˙
r ) is the distance vector from observer
where R(t With β = c3 dγ for longitudinal acceleration,
γ ds

to source, n̂ = R/R. All quantities on the r.h.s. 2
2 dγ
are taken at the retarded time ctr = ct − R(tr ). P|| = cre me c2 (5)
= (E × n̂)r the Poynting vector be- 3 ds
With cB
comes Sr (t) = −0 c2 E 2 (1 + β · n̂)n̂|r and the
In case of transverse deflection by magnetic field
˙
instantaneous differential radiation power is and bending radius ρ we have β⊥ = β 2 c/ρ and
dP (t) 2 (1 + β · n̂)R2 |r 2 β4γ4
= −n̂ · SR
2 |r = 0 c2 E P⊥m = cre me c2 2 (6)
dΩ 3 ρ
re me c2 R5 (( × β]
(2
˙ (( (2) and for transverse deflection by electric field E
= 5 (n̂ × [(n̂ + β) ˙
4πc r r
the acceleration is β⊥ = γc meE 2 and
ec

where r = R(1 + n̂ · β) = R(1 − β cos θ)
2 eE 2
between β and −n̂. Replacing
and θ the angle P⊥e = cre me c2 γ 2
3 me c2
(7)
dW/dΩ = (dP/dΩ)dt and taking the Fourier
transform of the electric field results in the spatial The classical result is slightly modified by a
and spectral distribution of radiation energy emit- quantum mechanical correction due to the recoil
ted by one electron during a single pass [1, 2], of the photon on the particle [3],

d2 W re me c2 2 55 εc
= ω Pqm = Pclassical 1 − √ (8)
dω dΩ 4π 2 c 16 3 E
( ∞ (2
( ( where εc is the critical photon energy [Eq.(6),
× (( e−iω(tr + Rc ) dtr (
n̂ × (n̂ × β) (
−∞
Sec.3.1.3].
r
215
Sec.3.1: SYNCHROTRON RADIATION
Radiation power depends on the mass of the

radiating particle like 1/m4 . For protons and elec- P(sinζ,ψ)
trons of the same total energy,

Pp me 4 ρ ψ
= = 8.80 × 10−14 Θ
ζ
Pe mp
Relativistic transformations Transformation
diffraction
of frequency ω and unit vectors n̂ from particle pattern

frame (∗ ) (particle moving in +z direction) to source area
laboratory frame produces the relativistic Doppler
effect
Figure 1: Optical diffraction.

ω = ω ∗ γ(1 + βn∗z ) (9)
and efficient collimation into the forward direc- The intensity distribution in ζ is still Gaussian and
tion, since for small angles ζ ≈ σr we find the stan-
n∗x,y β + n∗z dard width of the diffraction limited radiation dis-
nx,y = , nz = (10) tribution to be σr = 1/(kσr ). The actual ge-
γ(1 + βn∗z ) 1 + βn∗z
ometric intensity distribution in an image plane
Angles with respect to z-axis transform with nz = depends on the distance and particular focusing
cos Θ and n2x + n2y = sin2 Θ etc. like arrangement.
The diffraction limited photon emittance in
sin Θ∗
sin Θ ≈ (11) the horizontal and vertical plane is wavelength de-
γ(1 + β cos Θ∗ ) pendent and given by (see also Sec.4.2.2)
Here Θ is the emission angle independent of ob- λ
server. Radiation emitted into all forward space ph,u = 12 σr σr = . (2)
4π
(Θ∗ ≤ π/2) in particle system is collimated to Radiation from a linear source of length L (e.g.
within an angle Θ = ±1/γ in laboratory system. undulator) appears to emerge from a disk of diam-
eter D = Lσr in the middle of the source. From
References diffraction theory it follows for a round aperture
[1] J.S.Schwinger, PR 75 (1949) 1912 that D ≈ λ/σr and therefore with eq.(2)
[2] H. Wiedemann, Particle Accelerator Physics, 3rd √
λ λL
Edition, Springer, Berlin Heidelberg 2007 σr = , and σr = . (3)
[3] J.S. Schwinger, Proc. Nat. Acad. of Sci. USA 40 L 2π
(1954) 132 Photon distribution in 6-D phase space is
called the photon beam brightness, defined for
Gaussian distributions by (Sec.4.2)
3.1.2 Coherent Radiation Φph
H. Wiedemann, Stanford U./SSRL B = , (4)
4π 2 σT x σT x σT y σT y (dω/ω)
Diffraction limit and brightness Consider a where Φph is the photon flux (1/s) and σT u ’s are
round Gaussian source (σr ) and with angular the effective photon source sizes. The√diffraction
emission distribution (σr ). Further, we define a limited brightness is, with σx,y = σr / 2, etc.,
diffraction pattern where a point P is defined by 4Φph
the coordinates sin ζ, ψ, where ζ is the emission Bmax = . (5)
angle with respect to the optical axis (Fig.1). The λ2 dω
ω
intensity distribution in this pattern is then given For a real particle beam in an undulator, the
by Fraunhofer’s diffraction integral [1] diffraction limited photon emittance is diluted
∞ 2 by finite beam size and divergence. Further-
− ρ
Ig (ζ, ψ) ∝ e 2σr2 J0 (kρη) ρdρ more, geometrical considerations increase appar-
0
ent source size in the middle of the undulator, par-
∝ exp − 12 (kσr ζ)2 (1) ticularly due to (i) path oscillation amplitude a =
λp K/(2πγ) in the deflecting plane, where λp and
216
Ch.3: ELECTROMAGNETIC AND NUCLEAR INTERACTIONS
+0
K are the undulator period length and strength 1E
(Eq.(6), Sec.3.1.5), respectively; (ii) variation of Rectangular
distribution
beam size along the undulator (the 1/12...-terms); 1E
–1
and (iii) oblique observation at a finite horizontal Gaussian
g2 ( )
distribution
ψ and vertical θ angle with respect to the optical
–2
axis (the 1/36...-terms): 1E
2 2
2 σr2 2
σb0,x L ψ 2 L2
+ a2 +
σT,x = + σb0,x + 1E
–3
2 12 36
2 1 2 2
σT,x = σr + σb0,x
1E
–4
2 0.1 1
1 1 2 1 2 2 /λ 7–98
σT,y = σr2 + σb0,y

2 2 2 8355A134
+ σb0,y L + θ L
2 12 36
1 Figure 2: Form factors for Gaussian and uniform par-
2
σT,y = σ 2 + σb0,y
2
(6) ticle distribution.
2 r
where σb0,u and σb0,u are the betatron beam size
and divergence at the undulator entrance. To transverse distribution Ψ(r) = ψ(z) and ei-
maximize the photon beam brightness the focus- ther a Gaussian
2 particle distribution ψ(z) =
ing of the electron beam must be adjusted such
√ 1
2πσ
exp −z /(2σ 2 ) with standard width σ or
z
z
that 6-D phase space is minimized. For this case a uniform distribution ψ(z) = 1/ for − 12 ≤ z ≤
σT,u depends only on σr and σb0,u etc. The op- 1
2 the form factors become (see Fig.2).
timum matching condition occurs with σb,u 2 =
2 σ2
u βu and σb0,u = u /βu when σT,u σT,u = g(λ, σ) = exp −2π 2 2 (10)
λ
u βx + 12 σr2 u /βu + 12 σr2 becomes a min- sin(π/λ)
imum for g(λ, ) = (11)
π/λ
σr L
βu = = , where u = x, y (7) The duration δt and spectral width δν of a
σr 2π
wavepacket are connected by δt δν ≥ 1/(4π),
Spatial coherence The ratio where δt is called the coherence time and the co-
Φcoh λ2 herence length is derived from this for the general
= 2
(8) case and undulator radiation, respectively
Φph 16π σT,x σT,x σT,y σT,y
defines the fraction of spatially coherent photon λ2
coh = and coh,und = Np λ (12)
flux. Significant spatial coherent radiation is emit- Δλ
ted into the forward direction if the beam emit- The coherence length can be increased by reduc-
tance is reduced to u < λ/(4π). ing the spectral width Δλ in the monochromator.
Temporal coherence Particles confined into
<
very small volumes of dimensions ∼ λ can emit References
spatial and temporal coherent radiation [2]. The
[1] M. Born, E. Wolf, Principles of Optics, Pergamon
electrical field from electron m at frequency ω =
(1975)
ck has the form Em ∝ ei(ωt+krm n) where n
[2] H. Wiedemann, Particle Accelerator Physics, 3rd
the unit vector in the direction from observer to
Edition, Springer, Berlin, Heidelberg 2007
source center and rm is the vector from bunch
center to the electron m . The radiation power from
e ∗
all Ne electrons is with P (ω) ∝ N m,n=1 Em En , 3.1.3 Bending Magnet Radiation
H. Wiedemann, Stanford U./SSRL
P (ω) = p(ω)Ne 1 + (Ne − 1) g2 (λ, r) (9)
where p(ω) is the radiation power from a sin- Bending radius of charged particle path in homo-
( (2
gle electron, g(λ, r) = ( eikrn Ψ(r) dr ( the geneous field B at energy E is
form factor and Ψ(r) is the three-dimensional, 1 eB B[T]
normalized particle distribution. For a uniform [m] = = 0.2998 (1)
ρ βE βE[GeV]
217
1000
Instantaneous radiation power becomes, with
β̇⊥ = β 2 c/ρ, 100 K2/3(ξ)
10
c Cγ E 4
Pγ = , with (2) 1 K1/3(ξ)
2π ρ2
0.1
4π re m
= 8.8460 × 10−5
0.01
Cγ =
3 (me c2 )3 GeV3
0.001
10–4 10–3 10–2 10–1 100 101
for electrons. The total energy loss to radiation 6–98
8355A143 ξ
per 360◦ deflection is
" Figure 1: Functions K1/3 (ξ) and K2/3 (ξ).
Cγ 4
U0 = Pγ dt = E I2 (3)
2π
where I2 is defined in Sec.3.1.4. For an isomag-
netic ring U0 = Cγ E 4 /ρ. For a circulating beam
current I the total radiation power is
E 4 [GeV]
Ptot [W] = 109 Cγ I[A] (4) –3
eρ
The spatial radiation power distribution is (ψ –2
= deflection angle, θ = observation angle with re-

spect to the plane of deflection)
c)
–1
/ω
(ω
3
d2 Pγ 21 Pγ 1
10
= γ
g
Lo
1
dψ dθ 32 2π (1 + γ 2 θ 2 )5/2 0
γθ
5 γ 2 θ2 –1
× 1+ (5)
7 1 + γ 2 θ2 –3 1 12–97
8355A69
The first term represents the σ-mode with the

electrical field orthogonal to the deflecting field.
The second term is the π-mode for which the
electrical field is in the plane of deflecting field –3
and line of observation. After integration, Pσ =
–2
(7/8)Pγ and Pπ = (1/8)Pγ . 3
/ω
c)
2
The critical photon energy is –1
(ω
1
10
0
g
0
2 cγ /ρ ≡ ωc
3 3 γθ
Lo
εc = (6) –1
–2
–3
1 12–97
or (for electrons) 8355A70
E[GeV]3 Figure 2: Spatial and spectral distribution of radiation.

εc [keV] = 2.218 Upper: σ-mode. Lower: π-mode.
ρ[m]
2
= 0.665 E [GeV] B[T] (7)
Ki (ξ) is modified Bessel function (Fig.1), ξ =
Spatial spectral photon flux distribution is [1] 1 ω 2 2 3/2 , and
2 ωc (1 + γ θ )
d2 Φph Δω ω 2 2 2

= CΩ E 2 I K (ξ)F (ξ, θ) (8) γ 2 θ 2 K1/3 (ξ)
dθdψ ω ωc2 2/3 2 2 2
F (ξ, θ) = (1 + γ θ ) 1 + 2 (ξ)
1 + γ 2 θ 2 K2/3
where for electrons,
(10)
3α Again, the two terms in the square bracket
CΩ =
4π 2 e(me c2 )2 are for the σ- and π- mode polarization respec-
photons tively. The angular and spectral distributions of
= 1.3255 × 1022 (9) both modes are shown in Fig.2, relative intensity
s rad2 GeV2 A
218
100 10
σ–mode 1/3
1.333 ξ
80 1
S (ω/ωc)
60
(percent)
ω/ωc=0.001 0.1
1/2 ξ
0.777 ξ /e
40 0.01
0.1
0.01
π–mode
20 1
0.001
0 0.0001 0.001 0.01 0.1 1 10
0 2 4 6 8 10
7–98
8355A149 γθ
6–98
8455A144 ξ = ω /ωc
Figure 3: Relative magnitude of σ- and π-mode radia- Figure 4: Universal Function S(ω/ωc ).
tion.
f(ξ) = σθ (mrad) E (GeV)

10
distributions for different photon energies in
Fig.3. 0.5463/ξ
1/3
Integrating over the nondeflecting plane gives 1

the total photon flux per unit deflection angle,
0.1
dΦph Δω ω
= Cψ EI S (11)
dψ ω ωc – – – –
10 4 10 3 10 2 10 1 100 101
7–98
8355A145 ξ = ω/ωc
with
4α Figure 5: Function f (x) = σθ E.
Cψ =
9 eme c2
photons
= 3.9614 × 1019 (12) Harmonic representation Synchrotron radia-
s rad A GeV tion is emitted at frequencies which are νth har-
monics of the revolution frequency up to very
The universal function (Fig.4) is
high orders determined by the critical photon fre-
√ ∞ quency at harmonic number νc = ωc /ωL =
9 3 ω (3/2)γ 3 where ωL = c/ρ is the Larmor fre-
S(ω/ωc ) = K5/3 (x) dx (13) quency. For νc 1 the modified Bessel func-
8π ωc
ω/ωc tions K1/3 and K2/3 in the general expression (8)
' can be approximated by ordinary Bessel functions
1.3333 (ω/ω√c )1/3 for ω ωc
≈ of high order ν [3, 4] and the photon flux from a
0.77736 exx for ω ωc circulating beam current I becomes
where x = ω/ωc . dṄph I 2 Δν

= αν 2 (J ν + θ 2 Jν2 ) (15)
The rms spatial distribution σθ close to dΩ e ν
the forward direction (θ ≈ 0) is, with where the arguments of the Bessel functions are
dṄph d2 Ṅph √ νβ cos θ.
dψ ≈ dθdψ 2πσθ ,
Shielding Radiation formulas generally hold
Cψ S(x) 1 f (x) only in free space. When radiation is emitted
σθ = √ 2 (x/2) E = E (14)
2πCΩ x2 K2/3 within metallic pipes, long wavelength radiation
>
with λ ∼ pipe dimensions is cut off. For the case
f (x) is shown in Fig.5. of a beam between two infinite metallic plates at a
219
distance d, this shielding effect gives a differential also the numbering scheme in [4]):
photon flux in the forward direction (θ = 0) [2], "
I1 [m] = (κx Dx + κy Dy ) dz (1)
dṄph 2πρ Δν I "
= αν 2 (16)
dΩ dν ν e I2 [m−1 ] = κ2 dz (2)
j≤νdβ/πρ "
jπρ Jν2 (x)
J ν (x) +
2
× I3 [m−2 ] = |κ|3 dz (3)
d x2
j=1,3,5... "
2
I4x [m−1 ] = κ κx Dx
where x = (νβ)2 + (jπρ/d)2 and ν 1. For
d → ∞, it becomes Eq.(15). +2κx (kDx + kDy )] dz (4)
"
Shielding is particularly significant for short −1
2
bunches at frequencies ω ωc . For a uniform I4y [m ]= κ κy Dy
particle distribution of length , the total coherent
+2κy (kDx − kDy )] dz (5)
radiation power for circulating beam current I in "
absence of shielding is −1
I5u [m ] = |κ| 3 Hu dz (6)
#√ $4/3 "
o
Pcoh =
π 3ρ
I2 (17) I6x [m−1 ] = (kDx + kDy )2 dz (7)
0 c "
I6y [m−1 ] = (kDx − kDy )2 dz (8)
and with shielding by two infinite plates at dis-
tance d, where k, k are the strengths for normal and skew
√
∞ π 3ρd 2 quadrupoles, respectively, u = x or y, Du is the
Pcoh = I (18) dispersion, I4u is for a sector magnet, Hu (z) =
0 c 22
βu Du 2 + 2αu Du Du + γu Du2 , κx ≡ 1/ρx denotes
∞ /P o
The shielding efficiency is Pcoh
√ coh = the horizontal curvature, κy ≡ 1/ρy the vertical
2/3 2/3
3d/(23 ρ ). 1/3 one, and κ2 = κ2x + κ2y . The above integral ex-
pressions include vertical bending, but are miss-
ing terms related to coupled normal mode disper-
References sions (Sec.4.7.4.3) [5].
For a wedge magnet the integrals I4u must be
[1] H. Wiedemann, Particle Accelerator Physics II,
Springer, Berlin Heidelberg 1995.
modified to
[2] J.S. Nodvick, D.S. Saxon, PR 96 (1954) 180 I4u |wedge = κ2u Du0 tan θ0 + I4u + κ2u Due tan θe
[3] A.A. Sokolov, I.M. Ternov, Synchrotron Radia- (9)
tion, Pergamon, Oxford 1968 where Du0 , Due the dispersion functions at the
[4] G.A. Schott, Phil. Mag. 13 (1907) 194 entrance and exit, respectively and θ0 , θe are the
entrance and exit angles as measured from a
pole face normal to the beam orbit and the signs
3.1.4 Synchrotron Radiation in Storage
are chosen such that for a symmetric rectangular
Rings
magnet θ0 and θe are negative.
H.Wiedemann, Stanford U./SSRL
3.1.4.2 Radiation damping
3.1.4.1 Radiation integrals
Radiation damping occurs due to average energy
Beam parameters in a storage ring are modified loss into synchrotron radiation. Damping for en-
by the emission process of synchrotron radiation ergy deviation and bunch length derives from en-
[1, 2]. See also Sec.2.1.4. Far from linear cou- ergy dependence of radiation. In the transverse
pling resonances, they can be expressed approx- planes damping comes from the loss of transverse
imately through the following radiation integrals momentum with the emission of photons while rf-
[3], where we use the coordinate system and sign system replenishes only the longitudinal momen-
convention of Sec.2.1.1 and with u = x or y (note tum.
220
Damping decrements Betatron and syn- 3.1.4.3 Quantum excitation

chrotron oscillation amplitudes are damped like Emission of photons into synchrotron radiation
Ai = Ai,0 e−αi t , where i = x, y, z. With the occurs statistically leading to quantum fluctua-
circumference C "and tion of beam parameters [6]. The effect of quan-
1 cCγ 4
Pγ = Pγ dz = E I2 (10) tized emission of photons depends on the follow-
C 2πC ing quantities:
where Cγ is defined in eq.(3.1.3) (2) and for a flat
ring of circumference C the damping decrements mean photon energy ε = 158√3 cr
. 2 / 11 2
[6] are rms photon energy ε = 27

√ cr
Cα 3 I4x P
total photon flux Ṅph = 158 3 crγ
αx = E I2 1 −
C I2 . 2/
Ṅph ε = 2455 √ cr Pγ
Cα 3 I4y product 3
αy = E I2 (1 − ) √ c2 re me c2 |κ|3 γ 7
= 2455
C I2 3

Cα 3 I4x + I4y Here cr is the critical photon energy [Eq.(6),
αz = E I2 2 + (11) Sec.3.1.3]. Quantum excitation of beam energy
C I2
c rc 2 3 spread along path of length L is with the fine
where Cα = 3(mc2 )3 = 2113.1 m /GeV /sec.
structure constant α
If vertical bending magnets are present then the
2(
( 55α(c)2 7 L 3
vertical damping decrement must be modified ac- ΔσE q
= √ γ |κ| dz . (17)
cordingly similar to the horizontal decrement. 48 3 0
Damping partition numbers Ji are defined, Statistical emission of photons causes sudden
I4x I4y change in particle orbit at locations where disper-
Jx = 1 − , Jy = 1 − sion is finite. This effect leads to statistical in-
I2 I2
I4x + I4y crease of betatron oscillations and emittance. In-
Jz = 2 + , (12) crease of beam emittance u (u = x, y) due quan-
I2 tum excitation along distance L is
with Robinson’s sum rule [7, 8]
55 re c 5 L (( 3 ((
Ji = 4 (13) Δu = √ γ κu Hu ds . (18)
i 24 3mc2 0
The damping time τi is the reciprocal of the damp-
ing decrement αi . 3.1.4.4 Equilibrium beam emittances
Dependence on rf frequency The electron
beam energy in a storage ring can be changed Energy spread and bunch length Equilib-
rium energy spread is reached when the av-
by varying the rf-frequency Δδ ≡ Δp =
p0 erage z quantum
( excitation rate around the
− η1c Δf 2 /dt(
frf , where ηc = αc − γ 2 is the slip-
1
ring dσE
rf
z is equal to damping rate
page factor. When displacing the particle orbit ( q
2 /dt( = −2α σ 2 ,
dσE d z z E
transversely, quadrupoles also act like bending
magnets with κx = (kDx + kDy ) δ and κy =
2
σE I3
= Cq γ 2 (19)
(−kDx + kDy ) δ. The change in I4u is [9]with E2 2I2 + I4x + I4y
u = x, y with
ΔI4u = 2I6u δ (14) 55 c
Dependence on synchrotron oscillation Oscil- Cq = √ = 3.8319 × 10−13 m (20)
32 3 mc2
lating particle energy δ = δmax sin ωs t due to syn-
for electrons. The bunch length is
chrotron oscillations causes a periodic variation of √
damping decrements by [9] c|ηc | σE 2π c αc E0 σE
Δαz Δαu I6u σz = = ,
= − = δmax sin ωs t (15) ωs E0 ω0 heVrf cos φs E0
αz αu I2 (21)
where u = x, y. These oscillations result in where φs is the synchronous phase determined
a “breathing” variation of betatron (and syn- by Vrf sin φs = U0 with the sign convention
chrotron oscillation) amplitudes au by η cos φs < 0 and U0 given by Sec.3.1.3, ωs is
Δau I6u δmax
= 4 (16) the synchrotron oscillation frequency and the slip-
a0,u τu ω s I 2 page factor ηc = αc − 1/γ 2 .
221
Transverse beam emittance Equilibrium where the indices 0 and w indicate that the inte-
transverse emittance is reached when gral be taken over regular ring bending magnets
quantum excitation is equal to damping and wiggler magnets separately. A horizontally
du /dt|d z = −2αu u : deflecting wiggler magnet with Np periods and si-
γ 2 I5u nusoidal field variation along the electron path in
u = Cq (22) a dispersion-free section of a circular accelerator
Ju I2 generates a new equilibrium emittance
where u = x or y. In an ideal, flat accelerator
Dy = Dy = 0 and therefore y becomes zero. In 1+

8Cq βx 2 ρ0 3
xw 30πJx Np x0 ρw γ ρw Θw
this case, a transverse recoil due to photon emis- = (28)
sion into an angle 1/γ cannot be ignored anymore x0 1+ 12 Np ρρw0 Θw
and defines a lower fundamental limit on beam

emittance, where Θw = λp /(2πρw ), ρ0 > 0, the bending
radius ρw > 0 is determined by the wiggler peak
Cq βy z I3
y = (23) field. The resulting beam emittance is reduced if
2Jy I2 8 Cq βx
15π Jx 0 ρw γ Θw ≤ 1. For very long wiggler
2 2
Due to construction and alignment imperfections magnets (Np → ∞) the emittance approaches
both horizontal and vertical betatron oscillations
can become coupled. The degree of emittance xw 8 Cq βx 2 2
coupling depends on the tunes and the cou- −→ γ Θw (29)
x0 15π Jx 0 ρw
pling coefficient which is for linear coupling by
quadrupoles The change of equilibrium energy spread due to
( s+L (
( 1 ( damping wigglers is
κ− = ( ( k βx βy exp (iΨ) dz (( (24)
2π s 2
σw 1 + I3w /I30
where Ψ = ψx + ψy − (νx + νy + q) 2πz/L with = (30)
2
σ0 1 + I2w /I20
ψu (z) the betatron phases, νu the tunes, m, q inte-
gers and k the strengths of the skew quadrupoles. which approaches (for Np → ∞)
In case of a linear sum resonance, both emittances
grow indefinitely if |νx + νy − p| ≤ κ with p be- 2
σw ρ0 Bw
ing an integer. In case of a difference resonance, 2 −→ ρ
σ0
=
B0
(31)
w
νx − νy is close to an integer and we observe a
constant exchange of horizontal and vertical
√ beam
emittances at a frequency Ω = κ2 + Δ2 /2 3.1.4.6 Quantum lifetimes
where Δ = νx −νy −p. The maximum beam emit-
tances obey in accordance with Liouville’s theo- Particle losses occur for Gaussian particle distri-
rem the sum rule butions in 6-D phase space due to finite transverse
apertures or energy acceptance. Lifetimes due to
x + y = x0 (25)
these effects are [10, 11]
where x0 is the uncoupled horizontal emittance.
The ratio of the beam emittances is 1 eξ A2u
τq = τu , with ξ= (32)
y κ2 /2 2 ξ 2σu2
= 2 − (26)
x κ− /2 + Δ2
where u = x, y or E and Au the limiting half
apertures. For the synchrotron dimension, As
3.1.4.5 Damping wigglers might be determined by the rf-bucket height (see
Wiggler magnets cause additional damping and also Sec.2.1.2),
quantum excitation modifying the equilibrium
beam emittance to [9] 2 max 2 U0
ξ = max , = F (q)
γ 2 I50 + I5w
2
2σE E0 π |η| hE0
w = Cq , or
Ju I20 + I2w with F (q) = 2 q 2 − 1 − cos−1 (1/q) (33)
w 1 + I5w /I50
= (27)
0 1 + I2w /I20 and q = eVrf /U0 is the overvoltage factor.
222
References A periodic magnetic field device is called an

<
[1] A. Chao, JAP 50 (1979) 595 undulator if K ∼ 1 and a wiggler magnet if
[2] K. Hirata, K. Ohmi, K. Oide, PR E49 (1994) 751 K 1. Radiation is emitted in a line spectrum at
[3] R.H. Helm et al, PAC 73, San Francisco, 1973 a fundamental wavelength λ1 and its harmonics i.
[4] J. Jowett, AIP Proc. 153 (1985) p.934 The wavelength [3] is the result of Lorentz con-
[5] D. Sagan, BMAD manual vers.16.11, Sec.10.15, traction and Doppler shift of the magnetic period
https://www.lepp.cornell.edu/∼dcs λp ,
/bmad λp
[6] M. Sands, in: Physics with Intersecting Stor- λi = 1 + 12 K 2 + γ 2 (θ 2 + ψ 2 ) (7)
age Rings, ed. B.Touschek, Academic Press, NY, 2 γ2 i
1971 where θ, ψ are the horizontal and vertical obser-
[7] K.W. Robinson, PR 111 (1958) 373 vation angles with respect to the axis. The wave-
[8] Iu.F. Orlov, E.K. Tarasov, JETP 34 (1958) 651 lengths for θ = ψ = 0 are in practical units for
[9] H. Wiedemann, Particle Accelerator Physics, electrons
3rd. Edition, Springer, Berlin, Heidelberg 2007 λp
[10] A.W. Chao, Lecture Notes Phys., V.296, λi [Å] = 1305.6 2 1 + 12 K 2 (8)
Springer, Berlin, 1988, p.51 iE
[11] Y.H. Chin, DESY 87-062 (1987) and the corresponding photon energies
i E2
i [eV] = 9.4963 (9)
3.1.5 Undulator and Wiggler Radiation λp 1 + 12 K 2
The spectral bandwidth of undulator radiation is
Field expansion for sinusoidal field variation Δλ 1
By = B0 cos(kp z) along axis [1]: = (10)
λ i Np
Bx = 0
and the polar opening angle
By = B0 cosh kp y cos kp z (1)
Bz = −B0 sinh kp y sin kp z 1 1 + 12 K 2
σu ≈ , (11)
where λp = 2π/kp is the period length. Undula- γ 2iNp
tor field strength for hybrid magnet design [2] is,
where u = x or y. The total energy loss of an
for small gap apertures g < λp ,
electron passing through an Np -period wiggler or
g g undulator magnet of length Lu = Np λp is
By ≈ 3.33 exp − 5.47 − 1.8 (2)
λp λp 1 2 2 2 2
Erad = 3 re me c γ K kp Lu . (12)
Equation of motion n̂ = ρ
e
v×B
2
me γv0
has solu-
E[GeV ]2 K 2
tions Erad [eV] = 0.07257 Lu [m]
λp [m]2
K
x(t) = cos kp β̄ct A circulating beam current I generates a total
γkp
photon beam power of
K2
z(t) = β̄ct + sin(2kp β̄ct) (3) 4π 2 re E 2 K 2 Np I
8kp γ 2 P = , (13)
3 eme c2 λp
with average drift velocity
E[GeV]2 K 2 Np I[A]
K2 P [W] = 0.07257 .
β̄ = β 1 − 2 , (4) λp [m]
4γ
maximum oscillation amplitude a and deflection Spectral and spatial photon flux from beam cur-
angle ϑ with respect to axis rent I into harmonic i and σ- and π-mode polar-
ization (Fig.1)
K B0 λp
a= and ϑ = (5) dΦph (ω) Δω I
γkp Bρ 2π = αγ 2 Np2
and strength parameter dΩ ω e
∞
eB0 × i2 Sinc(Fσ2 + Fπ2 ) (14)
K= = γϑ = 93.4 Bλp (6)
βme ckp i=1
223
with In forward direction (θ = 0) and at harmonic

frequencies (Δωi = 0) the spatial photon flux dis-
sin πNp Δωi /ω1 2
Sinc = , tribution for the i-th harmonic is
πNp Δωi /ω1 (
2γθΣ1 cos ϕ − KΣ2 dΦph (ω) (( Δω I K 2 i2 [JJ]
Fσ = , ( = αγ 2 Np2
1 + 12 K 2 + γ 2 θ 2 dΩ i ω e (1 + 12 K 2 )2
2γθΣ1 sin ϕ Δω
Fπ = , = 1.7466 × 1023 E[GeV]2 I[A]Np2 fi (K)
ω
1 + 12 K 2 + γ 2 θ 2
with
where Δωi = ω − ωi , and
2
∞
[JJ] = J i−1 (x) − J i+1 (x) (15)
Σ1 (i) = J−m (u)Ji−2m (v), 2 2
m=−∞ iK 2
∞
and x = 4+2K 2 . The functions fi (K) are shown
in Fig.2.
Σ2 (i) = J−m (u)[Ji−2m−1 (v)
m=−∞
+Ji−2m+1 (v)], Table 1: Tabulation of functions fi (K).
ω β̄ K 2
u= 1 , K f1 f3 f5 f7 f9 f11
ω1 4(1 + 2 K 2 + γ 2 θ 2 )
0.1 0.010 0.000 0.000 0.000 0.000 0.000
ω 2β̄ K 2 γθ cos ϕ 0.2 0.038 0.000 0.000 0.000 0.000 0.000
v= , 0.4 0.132 0.004 0.000 0.000 0.000 0.000
ω1 1 + 12 K 2 + γ 2 θ 2
0.6 0.238 0.027 0.002 0.000 0.000 0.000
and the angle θ is taken with respect to the beam 0.8 0.322 0.087 0.015 0.002 0.000 0.000
axis and ϕ is the azimuthal angle counted counter- 1.0 0.368 0.179 0.055 0.015 0.004 0.001
clockwise while viewing the source against beam 1.2 0.381 0.276 0.128 0.051 0.019 0.007
direction with ϕ = 0 in the deflection plane. 1.4 0.371 0.354 0.219 0.118 0.059 0.028
1.8 0.320 0.423 0.371 0.286 0.206 0.142
σ–mode π–mode 2.0 0.290 0.423 0.413 0.354 0.285 0.220
5.0 0.071 0.139 0.188 0.228 0.261 0.290
10.0 0.019 0.037 0.051 0.068 0.075 0.085
i=1 20.0 0.005 0.010 0.013 0.016 0.019 0.022
x y
i =1
i=2 3
0.1 5
7
f i (K)
9
11
0.01
i=3
0.001
0.1 1 10 100
1–98
8355A146 K
4–98
8377A148
i2 K 2 [JJ]
Figure 2: Functions fi (K) = 2 .
Figure 1: Spatial undulator radiation distribution for (1+ 12 K 2 )
σ- and π- mode radiation in lowest order harmonics.
224
3.1.5.1 Polarization of synchrotron radiation

Polarization of synchrotron radiation occurs in the
direction of acceleration and radiation from bend-
K=0.1 K=1.0 ing magnet deflecting the beam in the horizontal
plane is polarized to a high degree in the hor-
izontal plane. This polarization is also known
as the σ-polarization (orthogonal or in German

”senkrecht” to magnetic field) in contrast to π-
mode polarization which is parallel to the mag-
netic field. For intensities of both components see
K=0.5 K=2.0 Sec.3.1.3.

Due to variations of source characteristics and
–2 0 2 4 –2 0 2 4
1–98
In ωph In ωph observation, a variety of polarization conditions
8355A147 can be generated. A radiation field at the ob-
server E (t) = (Ex (t) , Ey (t) , 0) has a gen-
Figure 3: Line spectra of wiggler/undulator radiation erally complex spectral amplitude of A (ω) =

for increasing strength parameter K. E (t)eiωt dt = (Ax (ω) , Ay (ω) , 0) and an in-
tensity I (ω) = |A (ω)|2 . With these definitions,
Total spectral photon flux of the harmonic i the Stokes parameters [1] are defined by
into forward cone of angle ±σu is S0 = Ix + Iy = Ax A∗x + Ay A∗y
Δω I iK 2 [JJ] S1 = Ix − Iy = Ax A∗x − Ay A∗y
Φph (ω)|i = παNp (16) S2 = I45 − I−45 = Ax A∗y + Ay A∗x
ω e 1 + 12 K 2
S3 = IR − IL = i Ay A∗x − Ax A∗y
Δω
= 1.431 × 1017 INp gi (K) where I45 , I−45 are the intensities with polariza-
ω
tion of ±45◦ ; IR , IL are the intensities of right
where
√ polarization; and A45 , A−45√=
and left handed
iK 2 [JJ] (Ax ± Ay ) / 2; AR , AL = (Ax ± iAy ) / 2,
gi (K) = (17)
1 + 12 K 2 respectively. In ordinary synchrotron radiation
Ay = 0 and S0 = S1 = Ix while S2 = S3 = 0.
Fig.3 shows the line spectrum as a function of K.
The electric field of synchrotron radi-
For K < 1 only few harmonics are obvious while
ation from
a positron
is proportional
to
for K 1 the line spectrum approaches the con-
tinuous synchrotron radiation spectrum of a wig- E ∝ n × n × β̇ = n nβ̇ − β̇ =

gler or bending magnet. nx nβ̇ − β̇x , ny nβ̇ − β̇y , nz nβ̇ −β̇z ,
The critical photon energy of wiggler radia-
tion depends on the observation angle ψ in the de- or with nβ̇ = nx β̇x + ny β̇y + nz β̇z
flecting plane because of a varying magnetic field, and nz
≈ 1 we get (E x , Ey , 0) ≈
−β̇x , −β̇y + ny nx β̇x + ny β̇z , 0 . Here, n
2
γψ is the unit vector from observer to the source.
εc (ψ) = εc (0) 1− . (18) On-axis, Ey = 0 and we get only linearly
K
polarized radiation. Off-axis (ny= 0) a vertical
field component Ey ≈ ny β̇z appears for
References nx = 0. Looking from above the midplane
[1] H. Wiedemann, Particle Accelerator Physics, 3rd on an arc section of the particle path, we ob-
Edition, Springer, Berlin Heidelberg 2007 serve on the far away part of the arc a positive
[2] K. Halbach, J. de Physique, C1, suppl.no 2, Tome β̇z > 0 and at the nearby part of the arc a
44 (1983) negative acceleration β̇z < 0 with an associated
[3] H. Motz, JAP, vol. 22, no. 5 (1951) 527 antisymmetric vertical electric field while the
horizontal electric field is symmetric along this
225
arc section. The amplitude for ny > 0 is then 3.1.6 Transition and Diffraction Radiation
A (ω) = (Ex cos ωt + iEy sin ωt)dt or Ax is C. Thongbai, Chiang Mai U./ThEP
real and Ay is imaginary. For bending magnet
radiation, S2 = 0 and S3 = 0 and depending Transition radiation (TR), first predicted by
on the sign of S3 , we have right (S3 > 0) or left Ginzberg and Frank [1], is emitted when a
handed (S3 < 0) circularly polarized radiation. charged particle passes through an interface be-
In wiggler and undulator magnets, we com- tween two media with different dielectric con-
bine the radiation from a pair of adjacent poles stants. The angular spectral energy density for
and notice that the radiation is equal from both backward transition radiation emitted by an elec-
poles except for the sign of the acceleration. tron travelling from a medium with dielectric con-
Therefore, the property of elliptically polarized stant ε1 to ε2 [2]
1/2
re me c2 β 2 ε1 sin2 θ1 cos2 θ1
radiation is cancelled. Ordinary flat wigglers and dW

= ×
undulators exhibit only linear polarization in the dΩdω π 2 c(1 − β 2 ε1 cos2 θ1 )
horizontal plane. |F1 F2 |2 . (1)
Insertion devices for elliptically polarized radi- (ε − ε2 )
F1 = 1 .
ation: (1 + β ε2 − ε1 sin2 θ1 )
a) In an asymmetric wiggler field one can
disturb the above mentioned cancellation by large (1 − β 2 ε1 + β ε2 − ε1 sin2 θ1 )
and short positive fields followed by lower and F2 = .
longer negative fields. Such a magnet produces (ε2 cos θ1 + ε1 ε2 − ε21 sin2 θ1 )
elliptically polarized radiation off-axis like that For forward radiation, exchange subscript 1 and 2
from bending magnets. and replace β by −β. Applied to the transition
b) A helical undulator includes the fields from vacuum (ε1 = 1) to metal (ε2 → ∞), the
B 0x sin
kp z, B 0y cos kp z causing accelerations radiation energy becomes
β̇x , β̇y and the fields on-axis are (Ex , Ey , 0) ≈ dW re me c2 β 2 sin2 θ1
= 2 . (2)
dΩdω π c(1 − β 2 cos2 θ1 )
−β̇x , −β̇y , 0 generating circularly polarized The radiation intensity increases from zero at
radiation. Here S2 = 0 and S3 = 0, indicat- θ1 = 0 to a broad peak at an angle sin θ1 =
ing circularly or elliptically polarized radiation at 1/(βγ).
λ
a fundamental wavelength of λ = 2γp2 1 + K 2 For a ψ-incidence on a thin metal foil placed
λ on the xy-plane, the forward radiation is emitted
for B0x = B0y and λ = 2γp2 1 + 12 Kx2 + 12 Ky2 from the metal-to-vacuum interface with the radi-
for B0x = B0y . ation cone in the direction of the electron (at an
c) Cross field undulators: Specific insertion angle ψ to the z axis). The backward radiation is
devices of a great variety have been designed to emitted from vacuum-to-metal interface with the
produce a variation of polarization characteristics. radiation cone in the direction of an angle π − ψ
similar to the helical device. with respect to the z axis. Its spectral-angular dis-
d) Variable polarization undulator: Here tribution is given by the contribution of parallel
the fields can be shifted
in phase against
and perpendicular polarization radiation with
each other B0x sin kp z , B0y cos kp z + Δ dW
re me c2 β 2 cos2 ψ
or Ex ∝ β̇x ∝ B0y cos kp z + Δ and = (3)
dΩdω π2 c
Ey ∝ β̇y ∝ B0x sin kp z . From that Ax ∝ [sin θ − β cos ϕ sin ψ]2
.
B0y (cos Δ − i sin Δ) and Ay ∝ B0x . With this [(1 − β sin θ cos ϕ sin ψ)2 − β 2 cos2 θ cos2 ψ]2
and for low values of K the Stokes parameters are
for the fundamental wavelength S1 ∝ B0y 2 − B2 ,
0x dW ⊥ re me c2 β 2 cos2 ψ
S2 ∝ 2B0x B0y cos Δ, and S3 ∝ 2B0x B0y sin Δ. = (4)
dΩdω π2 c
Variation of Δ by mechanical or electrical means [β cos θ sin ϕ sin ψ]2
.
allows the free choice of polarization. [(1 − β sin θ cos ϕ sin ψ)2 − β 2 cos2 θ cos2 ψ]2
The radiation angular distribution for oblique-
References incidence is azimuthally asymmetric. The asym-
[1] G.G. Stokes, Trans.Cambridge Phil.Soc. 9 (1852) metry, however, vanishes for highly relativistic
399
226
electrons as the distribution becomes closer to that angle between the radiation direction and −v . For
for normal incidence. Further details and discus- a single electron, α = 1 and the radiated intensity
sions of other interface conditions can be found in approaches that of transition radiation when the
[2, 3]. aperture radius becomes very small (a → 0).
Although the radiation has its maximum in- For a uniform beam of radius b, α =
tensity around an angle 1/γ, most of the radiation (2cβγ/ωb)I1 (ωb/cβγ).
is emitted at larger angles (> 96% for γ > 10) For an electron passing the aperture at a
and collection of the radiation within an exper- distance r from the center [1], add the term
imental acceptance angle is therefore important. dWT R /dωdΩ [(r/a) J1 (ωa sin θ/c)]2 to Eq.(6).
Collectable radiation energy within an acceptance For a discussion of diffraction radiation from
angle θa is a slit, see [2, 5] and references therein.

dW re me c2 2(1 − β 2 ) cos θa
(θa ) = × References
dω 2πc (1 − β 2 cos2 θa )
[1] V. Ginzberg, I. Frank, J.Phys. 9 (1945) 353
(1 + β 2 ) (1 − β cos θa )(1 + β)
+ ln −2 . [2] M.L. Ter-Mikaelian, High-energy Electromag-
β (1 + β cos θa )(1 − β) netic Process in Condensed Media, Wiley-
While the total radiation emitted into half space is Interscience, New York, 1972
[3] V.L. Ginzburg, V.N. Tsytovich, Transition Radia-
dW re me c2 (1 + β 2 ) (1 + β) tion and Transition Scattering, Adam Hilger, Bris-
= ln −2 ,
dω 2πc β (1 − β) tol, 1990
(5) [4] Yu.N. Dnestrovskii, D.P. Kostomorov, Sov. Phys.
which for relativistic electrons is proportional to Dokl. 4 (1959) 132; Sov. Phys. Dokl. 4 (1959) 158
ln γ. [5] M. Castellano, NIM A394 (1977) 275
As for the spectral radiation distribution, the
transition radiation spectrum from a single elec-
tron is uniform
√ up to a very high frequency of 3.1.7 Coherent Synchrotron Radiation
around γc 4πZN re , where N is the atomic den- S. Krinsky, BNL
sity and Z is the atomic number. For coherent
(See also Secs.2.4.15, 3.1.12) Consider a bunch
transition radiation, this uniform spectrum folds
containing N highly relativistic electrons passing
with the form factor of the electron distribution in
through a magnet with bending radius ρ. The
a bunch.
power radiated per unit wavelength P (λ) is [1]
Diffraction radiation (DR) is emitted when a
charged particle travels in the neighborhood of P (λ) = p (λ) [N + N (N − 1) g (λ)] , (1)
some inhomogeneity. Specifically, the radiation where p (λ) is the power radiated per unit wave-
can be generated when an electron travels past a length for a single electron and g (λ) is the abso-
metallic structure such as a transition in beam line lute square of the Fourier transform of the bunch
cross sections. The theory of DR was developed distribution. For a Gaussian bunch of rms length
in the late 1950’s, examining charged particles
σ, g (λ) = exp −4π 2 σ 2 /λ2 . The first term in
passing through circular apertures [4]. The radia- Eq.(1) with linear dependence on N is the inco-
tion emitted from an electron moving through the herent power and the second term with quadratic
middle of a circular aperture of radius a in an ideal dependence is the coherent power. Coherent syn-
conducting screen can be expressed as chrotron radiation (CSR) has been observed using
dW dWT R 2 linacs [2] and storage rings [3, 4]. For a Gaussian
= α
dωdΩ dωdΩ bunch in free space, the total coherent power ra-
ωa
2 diated is
ωa ωa
J0 sin θ K1 . (6) N2
c cβγ cβγ FS
Pcoh [W] = 2.42 × 10−20 2/3 . (2)
Here, dWT R /dωdΩ is the transition radiation de- ρ [m] σ 4/3 [m]
fined in Eq.(2), θ is the observation angle with re- In the case of electrons moving in the midplane
spect to the beam axis, J0 is the Bessel function of between parallel conducting plates with full sep-
order zero and K1 , I1 are modified Bessel func- aration h, the single electron power spectrum is
tions of the first order. For backward DR, θ is the cut off for wavelength greater than the cutoff
227
p
Figure 2: Equilibrium beam distribution for CSR wake

and positive momentum compaction.
Figure 1: For a Gaussian bunch, the ratio of the co-
herent power radiated between parallel plates to that in
At low current in a storage ring, the bunch
free space is plotted versus the ratio of the rms bunch
distribution is Gaussian. At higher currents, the
length to the cutoff wavelength.
CSR impedance can result in potential well dis-
tortion, which modifies the bunch distribution.
1/2
wavelength λco = 2h (h/ρ) [5] according to Below the microwave threshold, this is a repro-
p (λ) ∝ exp −λ2 /2λ2co . The total radiated ducible steady-state phenomenon that for positive
power is reduced relative to that radiated in free momentum compaction sharpens the fall of the
space by the factor Pcoh /Pcoh FS = F (σ/λ ) [6]
co bunch density at the front of the bunch and ex-
shown in Fig.1. We see that if σ/λco < 0.1, then tends the CSR power spectrum to higher frequen-
the total coherent power is very closely equal to cies. The modification of the bunch distribution
that in free space, but for larger bunch length, the is described by the Haissinski equation. We de-
total coherent power drops rapidly. fine the dimensionless
variables, x = s/σ0 and
Coherent emission results from the individ- y (x) = Z0 /σ0 Vrf (ρ/3σ0 )1/3 I (s), where σ0
ual electrons moving in the electric fields radi-
ated by other electrons. In accelerator physics is the rms length of the low-current Gaussian
bunch, Vrf = dV /ds is the spatial derivative of
this type of phenomena is generally treated us- rf
ing the formalism of wakefields or impedance. the rf voltage, and I (s) is the current distribution.
In addition to enhancing the radiated power, the The step-function CSR wake potential is given by
CSR impedance can result in an alteration of the S (s) = −Z0 (ρ/3s)1/3 . The Haissinski equation
electron bunch distribution. In a storage ring, the is [8, 9]
CSR impedance can give rise to the microwave in- 2
x
stability (Secs.2.4.9, 2.4.15). This instability can y (x) = K exp − (4)
2
result in the generation of a high-frequency den- ⎤
sity modulation of the bunch and in an increase ∞
of the CSR power at the modulation frequencies. +sgn(αp ) dzy (x − z) z −1/3 ⎦ ,
One problem in utilizing this mechanism for a 0
radiation source is that it is not a reproducible
where the constant K is determined by the nor-
steady state phenomenon. The stability condition
malization condition ds I (s) = N ec and hence
for the microwave instability driven by the CSR
depends on the number of electrons in the bunch.
impedance is [7]
For positive momentum compaction, the distor-
Ip γαp δ02 C tion of the bunch distribution is illustrated in
< , (3) Fig.2.
IA 2 (πρλ2co )1/3
where γ is the relativistic factor, αp is the mo-
References
mentum compaction, δ0 is the low current energy
spread, C is the storage ring circumference, Ip is [1] J.S. Nodvick, D.S. Saxon, PR 96 (1954) 180
the peak current of the bunch and IA = ec/re = [2] T. Nakazato et al, PRL 63 (1989) 1245
17, 000 A. [3] G.L. Carr et al, NIM A463 (2001) 387
228
[4] M. Abo-Bakr et al, PRL 88, 254801 (2002); 90, vx electron trajectory
094801 (2003)
[5] R.L. Warnock, SLAC-PUB-5375 (1990) vx z
[6] J.B. Murphy et al, PA 57 (1997) 9 Ex
[7] G. Stupakov, S. Heifets, PRST-AB 5, 054402
(2002)
[8] K. Bane et al, AIP Conf.Proc. 367 (1995) p.191 light wave
Ex
[9] F. Sannibale et al, PRL 93, 094801 (2004)
Figure 1: Condition for sustained energy transfer from

electron to light wave.
3.1.8 Free-Electron Lasers
Z. Huang, SLAC
P. Schmüser, U. Hamburg The fundamental wavelength of undulator radia-

tion in forward directionis
Overview The free-electron laser (FEL) is an λu K2
accelerator-based coherent light source of ex- λ = 2 1 + . (4)
2γ 2
treme brilliance. In contrast to conventional Low-Gain FEL Theory A low-gain FEL con-
lasers, the electrons in an FEL are not bound to sists of a storage ring or linac, a short undulator,
matter, but travel as free particles
at relativistic and an optical cavity. Upon each passage through
speed (β = v/c ≈ 1, γ = 1/ 1 − β 2 1) the undulator the light intensity grows by only a
through an undulator magnet. The light wave- few per cent, but after very many round trips a
length is inversely proportional to the square of large average FEL beam power can be achieved,
the electron energy and can be tuned at will. FELs e.g. more than 10 kW in the infrared FEL at JLAB
may operate as oscillators, starting from spon- [5].
taneous radiation, or as amplifiers, boosting the The light wave co-propagating with the elec-
light from a seed laser. FELs in the THz, infrared tron beam is taken as a plane wave Ex (z, t) =
and visible regime are equipped with an optical E0 cos(k z − ω t + ψ0 ) with wavelength λ and
cavity and work in the low-gain regime. At ul- wave number k = 2π/λ = ω /c. The trans-
traviolet and x-ray wavelengths, optical cavities verse electron velocity vx and the field Ex must
are presently not available and a large FEL gain point in the same direction to enable energy trans-
must be achieved in a single passage through a fer. Sustained energy transfer from electron to
very long undulator magnet. Of particular im- light wave requires that the light wave slips for-
portance is the Self-Amplified Spontaneous Emis- ward with respect to the electron by λ /2 per
sion (SASE) process. SASE FELs are capable of half period of the electron trajectory, see Fig.1.
producing coherent ultraviolet and x-ray pulses This is fulfilled if λ and γ are related by Eq.(4).
of femtosecond duration with a peak brilliance Slippages by 3λ /2, 5λ /2 . . . are also permitted,
106 − 1010 times higher than in third-generation leading to odd higher harmonics (λ /3 , λ /5 . . .)
synchrotron light sources. For a systematic pre- of the FEL radiation. However slippages of
sentation of FEL theory see Refs.[1, 2, 3, 4] and 2λ /2, 4λ /2 . . . yield zero net energy transfer,
the literature quoted therein. hence even harmonics are absent.
Undulator Radiation The undulator axis is The time derivative of the electron energy is
chosen along the beam direction z, the magnetic me c2 dγ/dt = −evx Ex
field in y direction (vertical). The magnet period
= −ecKE0 /(2γ) [cos ψ + cos χ]
λu is a few cm. The on-axis field in a planar un-
dulator is nearly harmonic, By = −B0 sin(ku z), with ψ = (k + ku )z − ω t + ψ0 . (5)
with ku = 2π/λu . The electrons travel on a sine- ψ is called ponderomotive phase, and χ =
like trajectory x(z) = K/(βγku ) sin(ku z) with ψ − 2ku z. If Eq.(4) holds, cos ψ remains con-
the undulator parameter stant along the undulator, while cos χ is rapidly
oscillating and cancels out. Choosing ψ =
K = eB0 λu /(2πme c) , K ≈ 1 . (1)
0, ±2π, ±4π.. yields optimum energy transfer
The transverse and longitudinal velocities are along the entire electron trajectory.
vx (z) = K c/γ cos(ku z) (2) FEL pendulum equations. Let the lasing pro-

vz ≈ c 1 − (1 + K 2 /2)/(2γ 2 ) (3) cess be started by light of wavelength λ . The
229
resonance electron energy γr me c2 and the rela-

tive energy deviation η are defined by

λu (1 + K 2 /2) γ − γr
γr = , η= (6)
2λ γr
The FEL pendulum equations read
ψ = dψ/dz = 2ku η , (7)

η = dη/dz = −eE0 K/(2m A 2 2
e c γ r ) cos ψ . (8)
In a planar undulator, vz contains an oscillatory
term which leads to odd higher harmonics and in-
fluences the energy transfer. This is accounted for
by modifying the undulator parameter in Eq.(8)

KA = K [J0 (ξ) − J1 (ξ)] , ξ = K 2 /(4 + 2K 2 ) .
Figure 2: Exponential growth and saturation of the
(9)
FEL power in LCLS at λ = 1.5 Å as a function
The FEL gain function, defined as the relative
of active undulator length [7]. The progressing mi-
increase in light intensity during one passage of
crobunching is indicated schematically.
the undulator, G = ΔI/I, is proportional to the
negative derivative of the line-shape curve of un-
dulator radiation (Madey theorem) [6]. Electrons with a slowly growing complex amplitude j̃1 (z).
with η > 0 enhance the light intensity, those with Combining Eqs.(10), (11) and (12) yields
η < 0 reduce it, see Fig.4 below. A
Ẽx (z) = dẼx /dz = −μ0 cK/(4γ r ) j̃1 (z) .
1D High-Gain FEL Theory High-gain FEL ra-
(13)
diation has a much higher intensity than undulator
Coupled first-order equations. To study the
radiation because many electrons radiate coher-
phase space motion of many electrons, the bunch
ently. Microbunching is essential: electrons los-
is subdivided into slices. In the slice 0 ≤ ψ < 2π
ing energy to the light wave travel on a sinusoidal
we choose N electrons which phases ψn . In the
trajectory of larger amplitude than electrons gain-
periodic model the distribution is continued in the
ing energy. The resulting modulation of vz leads
other slices. Then from Fourier series expansion
to the formation of microbunches. The particles in
a microbunch radiate like a single particle of high
N
charge. The growing radiation field enhances the j̃1 = 2j0 exp(−iψn )/N . (14)
microbunching further and leads to an exponen- n=1
tial growth and eventual saturation of the radia-
exp(−iψn )/N is called bunching factor.
tion power. Results from LCLS (Linac Coherent Combining Eqs.(7), (8) and (13) one obtains a set
Light Source) are shown in Fig.2 [7]. of coupled first-order equations
Wave equation. In the one-dimensional (1D)
FEL theory the dependencies of bunch current ψn = 2ku ηn , n = 1...N (15)

and electromagnetic fields on the transverse co- ηn A

= −eK/(2me c γr ) Re Ẽx exp(iψn )
2 2
ordinates x, y are neglected. The radiation field
obeys the 1D inhomogeneous wave equation A
Ẽx = −μ0 cK/(4γr ) · j̃1 ,
2
∂ 1 ∂2 ∂ j̃x which, together with Eq.(14), describe the evolu-
− Ẽx (z, t) = μ0 , (10)
∂z 2 c2 ∂t2 ∂t tion of the phases ψn and energy deviations ηn
where j̃x is transverse current density resulting of the N electrons, as well as the growth of Ẽx (z)
from the sinusoidal motion. We make the ansatz and j̃1 (z). Longitudinal space charge forces are of
Ẽx (z, t) = Ẽx (z) exp[i(k z − ω t)] (11) minor importance in short-wavelength FELs and
with a complex amplitude function Ẽx (z). The are neglected here and in Eq.(16) below.
microbunching effect is anticipated by assuming Third-order equation. The main physics of
a small periodic modulation of the current density the high-gain FEL is contained in the first-order
K Eqs.(15) but these can only be solved numerically.
j̃x (ψ, z) ≈ cos(ku z) j0 + j̃1 (z) exp(iψ)
γ If the modulation current j̃1 remains small a lin-
(12) ear third-order differential equation for the elec-
230
P [W]
tric field can be derived (see e.g. [4]):
i 9
Ẽx + 4iku η Ẽx − 4ku2 η 2 Ẽx − √
10
Ẽx = 0
( 3Lg0 )3 seed
(16) 6
10
with the 1D power gain length SASE
1/3
1 4γr3 me
Lg0 = √ . (17) 10
3
A 2 e2 ku ne
3 μ0 K
Here ne is the electron density.
1
Exponential gain and saturation. The solu-
0 5 10 15 20 25 30
tion of Eq.(16) is of the form Ẽx (z) = z/L
3 g0
j=1 Aj exp(αj z). For the special case η = 0

Figure 3: FEL power as a function of z/Lg0 in a
(electrons√ are on resonance) one √ finds α1,2 = seeded FEL (blue) and a SASE FEL (red). Solid
(±1+i/ 3)/(2Lg0 ), α3 = −i/( 3Lg0 ). In case
curves: numerical integration of coupled first-order
of laser seeding with an initial field E0 , all ampli-
Eqs.(15). Dots: analytic solution of third-order
tudes are equal, Aj = E0 /3. The light power Eq.(16).
stays almost constant in the “lethargy regime”,
<
0 ≤ z ∼ 2Lg0 , but then it grows exponentially 0.06 400
G(K) K
(see Fig.3) 0.04 300
P (z) ∝ exp(2Re[α1 ]z) ≡ exp(z/Lg0 ) . (18) 0.02

200
0
The Eqs.(15) yield the same result in the lethargy 100
0.02
and exponential gain regimes but describe FEL 0
0.04
saturation in addition. The saturation power is
0.06 100
Psat ≈ ρ Pb ,
10 0 10 10 0 10
(19) KU KU
where Pb the electron beam power, and ρ is the Figure 4: FEL gain function G(η) plotted vs. η/ρ at
dimensionless FEL (Pierce) parameter [9] two positions in a long undulator: z = 1Lg0 , 8Lg0 .
1/3 Red curves: high-gain theory. Blue dots/curve: low-
λu π I0 K A2 gain theory (Madey theorem).
ρ= √ = . (20)
4π 3Lg0 8 IA γr3 Ab ku2
(I0 peak current, IA ≈ 17 kA Alfvén current, Ab which has a spectral component within the FEL
beam cross section). For short-wavelength FELs, bandwidth. The effective shot-noise power and
ρ ≈ 10−3 . modulated current density are [10, 4]
FEL gain-function and bandwidth. For a
short undulator (length≤ Lg0 ), the high-gain FEL Pn = ργme c2 σω /(2π) , j̃1 (0) ≈ e I0 σω /Ab
theory agrees perfectly with the low-gain theory, (22)
but in long undulators strong differences are seen: The computed power rise for typical parameters
the gain is much larger and the gain-function of the soft x-ray FEL FLASH (see e.g. [4]) is
approaches a Gaussian (Fig.4). The high-gain shown in Fig.3. Saturation is achieved at an undu-
FEL acts as a narrow-band amplifier with an rms lator length Lu ≈ 20Lg0 . The SASE bandwidth
bandwidth at saturation is σωs /ω ≈ ρ.
SASE radiation exhibits shot-to-shot fluctua-
σω /ω = 0.83ρλu /z . (21) tions in wavelength. The coherence length at sat-
uration is
SASE [8, 9] permits the startup of lasing √ √
Lcoh ≈ πc/σωs = λ /(2 πρ) . (23)
without seed radiation. Intuitively, spontaneous
undulator radiation produced in the first section For a bunch length Lb > Lcoh , the average num-
of a long undulator serves as seed radiation in the ber of spikes in the wavelength spectra is M =
remaining section. Because of the random elec- Lb /Lcoh . In the exponential gain regime the nor-
tron distribution, the current contains a noise term malized radiation pulse energy u = Urad / Urad
231
0.01
(fifth) harmonic can reach 1% (0.1%) of the fun-
relative energy deviation η
z = 12 Lg0 z = 14 Lg0
damental power.
0
3D Effects The realistic description of high-
gain FELs has to be based on a three-dimensional
(3D) theory, taking into account electron beam
-0.01
2 2
emittance and energy spread, and optical diffrac-
norm. charge density
tion. For idealized cases, e.g. round beam

with uniform longitudinal charge density, an FEL
1 1
eigenmode equation including all these effects

0 0 can be developed. Realistic cases require sophisti-
0.01 z = 16 Lg0 z = 18 Lg0 cated simulation codes such as FAST [12], GEN-
relative energy deviation η
ESIS [13] or GINGER [14]. These are indispens-

0
able for the design of short-wavelength FELs.
3D gain length. The 3D gain length Lg is typ-
ically 30 − 50% longer than the 1D gain length
-0.01
10 2 Lg0 . In [15] Lg is expressed in terms of three di-
norm. charge density
mensionless parameters: Xγ = Lg0 4πση /λu (en-

5 1 ergy spread par.), Xd = Lg0 λ /(4πσr2 ) (diffrac-
tion par., σr rms beam radius) and Xε =
0 0
-2π 0
ponderomotive phase ψ
2π -2π 0
ponderomotive phase ψ
2π Lg0 4πε/(βav λ ) (angular spread par., ε emit-
tance, βav average beta function).
Figure 5: Evolution of the microbunch structure at Lg = Lg0 (1 + Λ) (25)
z = 12Lg0 , 14Lg0 , 16Lg0 , 18Lg0 . Upper subplots: a2 a4 a6 a8 a9
Λ = a1 Xd + a3 Xε + a5 Xγ + a7 Xε Xγ
distribution of particles in (ψ, η) phase space. Three
FEL buckets are indicated by dashed curves. Lower + a10 Xda11 Xγa12 + a13 Xda14 Xεa15
subplots: normalized charge density as function of ψ. + a16 Xda17 Xεa18 Xγa19
a1 = 0.45, a2 = 0.57, a3 = 0.55, a4 = 1.6,
fluctuates according to the gamma distribution a5 = 3.0, a6 = 2.0, a7 = 0.35, a8 = 2.9,
a9 = 2.4, a10 = 51, a11 = 0.95, a12 = 3.0,
M M uM −1 −M u
e , σu2 = 1/M .
pM (u) = a13 = 5.4, a14 = 0.7, a15 = 1.9, a16 = 1140,
Γ(M )
(24) a17 = 2.2, a18 = 2.9, a19 = 3.2 .
Phase space and simulation of microbun- Gain guiding counteracts the diffractive
ching. The FEL dynamics resembles the widening of the FEL beam since most of the light
synchrotron oscillations of a stored proton is generated in the central core of the electron
beam. In the (ψ, η) phase space the particles beam [16]. Gain guiding permits the FEL beam to
rotate clockwise, hence particles in the right follow slow, “adiabatic” motions of the electron
half of an FEL bucket transfer energy to the beam and is thus crucial for the long undulator in
light wave, while those in the left half withdraw an x-ray FEL.
energy. Eqs.(15) are well suited for modelling Transverse coherence. The fundamental
the microbunching, see Fig.5. For z ≥ 12Lg0 Gaussian mode TEM00 has its highest intensity
pronounced microbunches evolve in the right on the beam axis while higher modes extend to
halves of the FEL buckets and increase the light larger radial distances. The TEM00 mode grows
intensity, while beyond 18Lg0 they move into fastest along the undulator, owing to its superior
the left halves and reduce it. The FEL power overlap with the electron beam. Near saturation
oscillations in Fig.3 are caused by this rotation in it dominates and the FEL radiation possesses a
phase space. high degree of transverse coherence, as verified
Higher harmonics. Close to saturation, the by double-slit diffraction experiments [11].
periodic sequence of narrow microbunches (see Seeding Various seeding methods have been
Fig.5) corresponds to a modulation current with proposed to improve the coherence length of
rich harmonics contents. In a planar undulator, SASE radiation. Direct seeding requires a co-
odd higher harmonics will be amplified. The third herent signal well above the shot-noise level. In
232
the VUV such signals may be obtained by high References

harmonic generation (HHG) in a gas. At shorter
[1] J.B. Murphy, C. Pellegrini, Introduction to the
wavelengths, self-seeding [17] may be applied: a
physics of free electron lasers, Laser Handbook
SASE signal, produced in a short undulator, is
Vol. 6 (1990) 115
passed through a monochromator and serves as [2] E.L. Saldin, E.A. Schneidmiller, M.V. Yurkov,
seed radiation in the main undulator. In a high- The Physics of Free Electron Lasers, Springer
gain harmonic generation (HGHG) FEL [18], the 2000
electron beam is energy-modulated in an undu-
[3] Z. Huang, K.-J. Kim, PRST-AB 30, 034801

lator by interaction with a powerful laser. A (2007)
magnetic chicane converts the energy modula- [4] P. Schmüser, M. Dohlus, J. Rossbach, Ultraviolet
tion to a density modulation. A second undulator and Soft X-Ray Free-Electron Lasers, Springer
causes the density-modulated beam to emit co- 2008

herent radiation at a higher harmonic frequency. [5] FEL at JLAB: see www.jlab.org/FEL
In an echo-enabled harmonic generation (EEHG) [6] J.M.J. Madey, Nuovo Cim. 50B (1979) 64
FEL [19], a second modulator followed by a [7] P. Emma et al, PAC 09
second chicane are inserted before the radiator. [8] A.M. Kondratenko, E.L. Saldin, PA 19 (1980)
The electron beam interacts twice with two laser 207
pulses in the two modulators. The longitudinal [9] R. Bonifacio, C. Pellegrini, M. Narducci, Opt.
phase space becomes highly nonlinear, leading to Comm. 50 (1984) 373
density modulations at a very high harmonic num- [10] K.-J. Kim, PRL 57 (1986) 1871
ber for a modest energy modulation. [11] R. Ischebeck et al, NIM A507 (2000) 175
Beam requirements Very bright electron [12] E.L. Saldin et al, NIM A429 (1999) 233
beams are required to drive ultraviolet and x-ray [13] S. Reiche, NIM A429 (1999) 243
FELs. Higher peak current and smaller cross [14] W. Farlay, Report LBNL-49625, LBL (2002)
[15] M. Xie, NIM A445 (2000) 59
section reduce the gain length (see Eq.(20)).
[16] G. Moore, NIM A239 (1985) 19
High peak currents require longitudinal bunch
[17] J. Feldhaus et al, Opt. Com. 140 (1997) 341
compression (see Sec.4.5), but the energy spread
[18] L.H. Yu, PR A44 (1991) 5178
is increased by this process (which affects [19] G. Stupakov, PRL 102, 074801 (2009)
the gain length through Xγ in Eq.(25)). Very
low-emittance beams can be generated with
photocathode rf guns or specially designed 3.1.9 Ultrashort X-ray Pulse Generation
thermionic guns. The beam cross section in the A. Zholents, ANL
undulator can be reduced by stronger focusing Apart from squeezing the electron bunch [1, 2],
(i.e., smaller βav ), but the increased angular the other way to obtain ultrashort x-ray pulses is
spread will eventually degrade the FEL gain to select the radiation produced by a small sec-
(through Xε in Eq.(25)). The FEL design tion of the electron bunch. In the case of spon-
optimization is therefore multi-dimensional taneous emission, this is accomplished by physi-
and beyond our scope here, but the typical cally pointing this radiation into a different direc-
requirements on electron beams are tion than the radiation of other electrons. In the
I0 ≥ 1 kA , ση ≤ ρ/2 , ε ∼ λ /(4π) . (26) case of coherent emission, this is accomplished
These requirements apply to the “slice” beam by providing a condition when this radiation dom-
qualities defined on the scale of the coherence inates the radiation of all other electrons by a large
length (see Eq.(23)). For harmonic generation factor.
FELs, the slice energy spread should be much Spontaneous emission “Slicing” of the elec-
smaller than the ρ−parameter of the final ampli- tron bunches has been implemented in storage
fier because the additional energy modulation im- rings [3] where electron beam interaction with
posed on the beam becomes the effective energy a laser in a wiggler magnet has been used to
spread there. Beam current, slice emittance and produce energy modulation of electrons with
energy spread should be “flat” along the bunch modulation amplitude ΔE several times larger
in order not to increase the final radiation band- than the electron beamenergy spread σ:
width. Such high-quality beams can be produced ΔE(q, ν, σ̂τ ) = 2 AL αωL ξ (J0 (ξ/2)
with linacs but not with storage rings.
−J1 (ξ/2)) f (q, ν, σ̂τ ) (1)
233
where f (q, ν, σ̂τ ) =

0.5
cos(2πν ẑ−tan−1 (qẑ)) −ẑ 2 /4σ̂2
2q √ e τ dẑ,
(2π)1/2 σ̂τ 1+(qẑ)2
−0.5
and α is the fine structure constant; AL the
laser pulse energy; ωL = 2πc/λL with λL the
laser wavelength; q = Lw /zR with Lw the
wiggler length and zR the Rayleigh length; στ
the rms width of the laser intensity distribution;

ẑ = z/Lw with z the longitudinal coordinate
centered at the wiggler center; ξ = K 2 /(2 + K 2 ),
where K = eB0 λW /(2πme c) with B0 the
wiggler peak magnetic field and λW the wiggler Figure 1: Density plot showing a fragment of the lon-
period; ν = N 2(δγ/γr ) is the detuning from gitudinal phase space of the electron bunch after inter-
the resonance energy
with
relativistic factor action with a few-cycle laser pulse in one period wig-
γr2 = λW /2λL 1 + K 2 /2 ; σ̂τ = στ c/N λL ; gler magnet. The width of the curve is defined by the
and N is the number of wiggler periods. Then electron energy spread. Two regions of interest (see
lattice dispersion is used to separate the emission text) are highlighted by red color and labelled (1) and
from electrons with large ΔE. The actual source (2).
of the synchrotron radiation (SR) can be bending
magnet, wiggler, or undulator. The duration of can prepare the electron bunch entering the FEL
the x-ray pulse is defined by the laser pulse and by undulator in such way that an ultrashort slice of
stretching of the electron slice accumulated on a the electron bunch will lase, and all the rest of the
path from the wiggler to the radiation source. The electrons will not lase or lase only with a much
timing for the x-ray pulse is defined by the arrival slower pace, thus producing a much weaker sig-
of the laser pulse and, thus, is synchronized with nal. One possibility is to use a slotted foil and
the laser pulse. spoil the emittance of the entire electron bunch
Another technique uses a deflecting cavity in except for a small fraction that goes through a slot
a storage ring to produce a time-dependent kick in the foil [5]. Another possibility is to use a few-
to the electron bunch that changes the sign in the cycle laser pulse with a carrier-envelope phase
middle of the electron bunch and reaches a large stabilization [6] and a wiggler with one or two pe-
magnitude at the bunch edges which exceeds by riods upstream of the FEL undulator, and thus ob-
many times the angular divergence of the x-ray tain energy modulation of electrons as shown in
emission in the undulator σr and the rms beam Fig.1. Two special regions are seen there: one
angular spread σy [4]. This kick is eventually with the largest energy offset and one with the
canceled by a second deflecting cavity, but the x- largest energy chirp. The uniqueness of each re-
ray beamline located between these two cavities gion can be explored in the FEL in a number of
can use the time-dependent variation imprinted on ways [7, 8, 9, 10, 11, 12, 13, 14], each favoring
the x-ray pulse from the SR source either by se- lasing either by electron group (1) or group (2),
lecting a short slice of the radiation with a slit or and damping lasing of the remaining electrons.
by compressing the entire radiation pulse using an For example, a tapered undulator producing z de-
asymmetrically cut crystal: pendence of the undulator parameter K defined in
B # $
C [10] as
Eb 1 C 2
D εy 1 + σr . (2)
στ −xray ≈ d ln K λx 1 + K 2 /2 d ln γ
eU 2πfrf βrf σy2 =− , (3)
dz λu K 2 /2 cdt
Here U is the “deflecting voltage”, frf is the de- with λx the x-ray wave length and λu the undu-
flecting cavity frequency, Eb is the electron beam lator period, will only support lasing of electrons
energy, εy is the vertical electron beam emittance, with the energy chirp dγ/dt from region 2. De-
and βrf is the vertical beta function at the deflect- pending on the actual technique, FEL type, and
ing cavity. x-ray photon frequency, the minimum duration of
Coherent emission In the case of coherent x-ray pulses can vary from a few femtoseconds
emission from free-electron lasers (FELs), one to a few tens of attoseconds. However, the time
234
bandwidth product of these pulses is close to the

Fourier transform limit in most cases.
The most attractive feature of all techniques
involving the laser is synchronization of the ul-
trashort x-ray pulse to a modulating laser pulse.
Moreover, two or more successive laser pulses
can produce two or more electron beam modu-
lations in different sections of the electron bunch
and, therefore, create two or more ultrashort x-ray Figure 1: Scattering geometry and angle definitions.
pulses with a precise time delay between them. It
is even possible to have two-color ultrashort x-ray
Table 1: Energy Reach of Scenarios.
pulses [14].
Scenario Maximum Energy
References Backscattering 4γ 2 Elaser
[1] J. Feikes et al, EPAC 04, 1954 Sidescattering 2γ 2 Elaser
[2] S. Reiche et al, NIM A593 (2008) 39 Small Angle Scattering (1 + γ 2 Φ2 )Elaser
[3] R.W. Schoenlein et al, Science 287 (2000) 2237
[4] A.A. Zholents et al, NIM A425 (1999) 385
[5] P. Emma et al, PRL 92, 74801 (2004) incident laser beam is in the x-z plane at angle Φ,
[6] F. Krausz, M. Ivanov, RMP 81 (2009) 163 and the laser polarized linearly in y-direction. θ
[7] A.A. Zholents, W.M. Fawley, PRL 92, 224801 and φ are spherical polar angles for the scattered
(2004) radiation. Definitions: Backscattering Φ = π,
[8] E.L. Saldin, E.A. Schneidmiller, M.V. Yurkov, Sidescattering Φ = π/2, Small Angle Scattering
Opt. Commun. 239 (2004) 161 Φ << 1.
[9] A.A. Zholents, G. Penn, PRST-AB 8, 050704 The incident photon energy in the beam frame
(2005) is Elaser γ(1−β cos Φ) where Elaser is the incident
[10] E.L. Saldin, E.A. Schneidmiller, M.V. Yurkov, photon energy in the lab. In terms of the angles in
PRST-AB 9, 050702 (2006) Fig.1, the scattered energy Eγ is
[11] A.A. Zholents, M.S. Zolotorev, NJP 10, 025005
(2008) Elaser (1 − β cos Φ)
Eγ (θ, φ) =
[12] Y. Ding et al. PRST-AB 12 (2009) 060703 1 − β cos θ + Elaser (1 − cos ΔΘ)/Ee
[13] D. Xiang, Z. Huang, G. Stupakov, PRST-AB 12,
060701 (2009)
where ΔΘ is the angle between the incident
[14] A.A. Zholents, G. Penn, NIM A612 (2010) 254 and scattered photons satisfying cos ΔΘ =
cos(Φ) cos(θ)−sin Φ sin θ cos φ, and Ee = γmc2
is the total energy of the electrons. The Thomson
3.1.10 Compton/Thomson Sources limit, implying the neglect of electron recoil, is
G.A. Krafft, TJNAF [2]
When a relativistic electron beam passes through Elaser (1 − β cos Φ)
Eγ (θ, φ) = .
a high-field laser beam, the electrons radiate in- 1 − β cos θ
tense and highly collimated electromagnetic ra- There are non-linear corrections at high laser in-
diation by the Compton/Thomson scattering pro- tensities discussed separately.
cess. Due to relativistic upshifting of the pho- A plane wave expansion for the incident pho-
tons and the relativistic Doppler effect, highly en- ton beam in terms of normalized vector potential
ergetic polarized photons are radiated along the (a = eA/mc) is
electron beam motion. By this means effective
a(x, t) = {a(ωt − k · x)ei(ωt−k·x) }ŷ.
x-ray and gamma ray sources can be built [1].
High energy Compton scattering is discussed in The peak value of the normalized vector poten-
Sec.3.3.4. The physics and quantitative discus- tial is generally denoted a0 . This field strength
sion of the scattered radiation is largely the same parameter is analogous to K for undulators. A
as in undulators (Sec.3.1.5). In Fig.1, the z-axis plane wave formulation implies the neglect of
is aligned with the relativistic electron beam (rel- hour-glass effects in the description of the collid-
ativistic factors β and γ), the wave vector of the ing photons.
235
The spectral energy density of the output where the total Thomson scattering cross section
pulse per electron is [3] is σT = 8πre2 /3. For overlapped Gaussians of
r 2 0 (( ω(1 − β cos θ) ((2
rms size σe and σp ,
d2 Uγ
= e (Ẽ ( Np Ne
dωdΩ 2πc c(1 − β cos Φ) Nγ = σT
2π 2 + σ2
σex 2 2
cos2 φ(1 − β cos θ)2 + sin2 φ(cos θ − β)2 px σey + σpy
× ,
γ 2 (1 − β cos θ)4 where Np is the number of incident photons and
Ne is the number of electron scatterers. The lumi-

where Ẽ is the spatial Fourier transform of the
nosity is
transverse electric field as a function of the lon-
gitudinal variable of the incident laser evaluated Np Ne
L=f ,
in the lab frame (E = −∂A/∂t), and the nota- 2π 2 + σ2

σex 2 2
px σey + σpy
tion means evaluate the transform at the Doppler
shifted wave number ω(1 − β cos θ)/c(1 − β where f is the collision repetition rate. A related
cos Φ). formula is
The number of photons scattered per electron Ne σT
in a flat incident pulse is (a0 << 1) Uγ = γ 2 (1+β) Ulaser
2π σex 2 + σ2 2 2
px σey + σpy
2παNλ a20
Nper e− = ,
3 where Uγ is the total scattered energy per pulse
where α is the fine-structure constant and Nλ and Ulaser is the total energy in the incident pulse.
is the number of wavelengths in the incident
pulse. 3.1.10.2 Nonlinear scattering
As a0 approaches one, frequency red-shifting and

3.1.10.1 Luminosity description harmonic generation become prominent [5]. For
a longitudinally flat illuminating laser, computa-
Compton sources are an example of a photon- tions and results are very similar to those in stan-
lepton collider. A description in terms of “lumi- dard undulator theory.
nosity” is appropriate. Results in this section are E.g., a modified resonance condition
specific to the Thomson Backscatter case.
In a single collision, the total number of scat- 4γ 2 Elaser
Eγ ≈
tered photons is Lorentz invariant, as is σ, the to- 1 + a20 /2
tal scattering cross section in the rest frame of the
applies for linear polarization. As in undulators,
electrons [3, 4]. The lepton current 4-vector is
a20 /2 → a20 in this formula for a circularly polar-
Φμe = J μ /e. The incident laser photons are de-
ized incident laser.
scribed by the 4-vector Φμp = (np , np k̂p ). The For linear polarization, odd harmonics are
number of scattered photons Nγ is proportional generated. The strengths of the harmonic lines
to both np and ne , and d4 x is Lorentz invariant. are quantified by “Bessel function factors”. These
Consequently, Nγ must be factors have the same functional form as in un-

dulators, but the relativistic invariants in the argu-
Nγ = Φpμ Φμe σd4 x ments of the functions are modified. For example,
the flux factor into the nth harmonic (n odd) is [6]
%
= (1 − βp · βe )np (x, y, z, t)ne (x, y, z, t)σd4 x. n2 a20 na20
Fn (a) = J (n−1)/2
(1 + a20 /2)2 4(1 + a20 /2)
When the hourglass effect is negligible &
the distributions are separable, np (x, y, z, t) = na20
−J(n+1)/2 ,
np (x, y)Ip (z + ct), ne (x, y, z, t) = ne (x, y) 4(1 + a20 /2)
Ie (z − βct), where Ip , Ie are unit normalized, and where J denotes the Bessel function of integer or-

der. At low a0 , the strength of the higher harmon-
Nγ = np (x, y)ne (x, y)σT dxdy, ics grows with much higher powers of the field
strength than the fundamental. As a0 increases
236
beyond 1, most of the emission is radiated in the Radiation with fundamentally different character-
harmonics. istics is being emitted by relativistic electrons at
For non-flat illumination pulses the situation the entrance and exit edge of otherwise uniform
is more complicated. Because of electron slow- bending magnets. If ϑ and ψ are the horizon-
down and red-shifting, the spectrum emitted be- tal and vertical observation angles with respect
comes broad, exhibits interference structures, and to the beam axis and horizontal being the deflect-
is generally maximum at the maximum red shift ing plane, we get for the spatial distribution of the
within the pulse [7]. photon flux [3]:
dṄph (ω) α I dω
References = 2 γ2 (Fσ êσ + Fπ êπ )2
dΩ π e ω
[1] P. Sprangle et al, JAP 72 (1992) 5032
[2] F. V. Hartemann, et al, PRST-AB 8, 100702 (2005)

where

[3] G.A. Krafft, PAC 97, p.739 iπ ∞
[4] W.J. Brown, F.V. Hartemann, PRST-AB 7, 060703
Fσ = (θ − βx ) exp(iΦ) dz
λγ −∞
(2004) ∞
iπψ
[5] E. Esarey, S.K. Ride, P. Sprangle, PR E48 (1993) Fπ = − exp(iΦ) dz
3003 λγ −∞
[6] K.-J. Kim, AIP Conf. Proc. 184 (1989) 565 π 2
Φ= (θ + ψ 2 + 1/γ 2 )z
[7] G.A. Krafft, PRL 92, 204802 (2004) λ
z z
−2θ βx dz + βx2 dz
0 0
3.1.11 “Short” Magnet and Edge Radiation
H. Wiedemann, Stanford U./SSRL βx = vx /c the transverse particle velocity,

1 s
The source length of synchrotron radiation from βx (z) = f (z )dz + βx (0) (2)
long uniform magnets is ∼ ±ρ/γ setting an up- ρ 0
per limit for the photon energy spectrum ωc ≈
with ρ = −me c2 γ/(eB0 ) and f (z) = B(z)/B0 ,
4πcγ 3 /ρ. If a magnet is shorter than this source
assuming that the particles oscillate in the deflect-
length (L < ρ/γ), the radiation pulse is also
ing/horizontal plane only.
shorter and the spectrum extends to higher fre-
quencies of about ωsh ≈ 4πcγ 2 /L. The spatial There is no radiation in the forward direc-
distribution of the photon flux into a band width tion. The intensity of the σ-mode increases with
dω/ω is [1, 2] θ reaching a maximum at θ = 1/γ. The π-mode
radiation peaks at ψ = 1/γ. For very long wave-
dṄph (ω) 80 c3 re2 I dω 6 2 lengths the photon flux levels off at
= γ f |B̃(ω)|2 (1)
dΩ π e ω (λ→∞
with dṄph (ω) (( α γ2 I
( → (3)
(1 + γ 2 θ 2 )2 sin2 ϕ + (1 − γ 2 θ 2 )2 cos2 ϕ dΩ ( 2 2 4π 2 e
f2 = θ +ψ =1/γ 2
(1 + γ 2 θ 2 )6
where the two terms in the denominator indicate References
the σ- and π- mode polarization respectively, θ the
angle between observation (n) angle with respect [1] R. Coisson, PR A20 (1979) 524
to the beam axis (v) and ϕ the angle between [2] R. Bossart et al, NIM 164 (1979) 375
plane (n, v) and plane including acceleration and [3] O.V. Chubar, N.V. Smolyakov, PAC 93, 1626
beam (a, v) , B̃(ω) is the Fourier transform of the
magnetic field B(t). In case of a Lorentzian field
distribution B(z) = B0 /(1 + 4z 2 /L2 ), Eq.(1) 3.1.12 Beam Solid-Target Photon Physics
becomes K. Ispiryan, Yerevan Physics Institute
dṄph (ω) π0 cre2 I dω 2 2 2 2 This section reviews Cherenkov radiation,
= γ B0 L f
dΩ 2 e ω bremsstrahlung, coherent bremsstrahlung, chan-

L(1 + γ 2 θ 2 ) neling radiation, parametric x-ray radiation, and
×(1 + γ θ ) exp −
2 2 2
ω Smith-Purcell radiation [1, 2, 3, 4, 5, 6].
2γ 2 c
237
Cherenkov radiation (CR): Electromagnetic As it is seen the dipole like ∼ cos2 θ angular distri-
radiation [1] produced by particles with charge z bution of B valid for E mc2 becomes sharply
passing through a medium with constant velocity forward directed with maximum at θ ∼1/γ, while
v exceeding the phase velocity v/n of light in the the spectral distribution is ∼1/ω.
medium with refractive index n(ω), i.e. CR has a At higher energies the spectral distribution of
threshold βthr = vthr /c = 1/n. With the help of B is given by the following QED formula (also
Weizsäcker-Williams method of pseudophotons relevant for other types of radiation)
1
accompanying the particles one can interpret the
dσB
production of CR photons as W-W photons sepa- = 4αZ 2 r02 1 + (1 − x)2 ψ1 (δ)
rated from the charge due to their slowness. CR dω ω
has been discovered in 1934 [7], while the QED 2
− (1 − x) ψ2 (δ) , (4)
theory has been developed in [8]. From simple 3

Huygens principles, the CR photons are emitted
under angle where x = ω/E, r0 = e2 /mc2 theclassical elec-
cos θ = 1/βn (1) tron radius,
with spectral distribution of number of photons ωmc2 2π 1 2π 1 x

δ= = (5)
per unit length equal to 2E(E − ω) λC 2 λC γ 1 − x
d2 NCR z2α the minimal value of the longitudinal component
= sin2 θ , (2) q|| of the momentum transfer q, λC = (h/mc) the
dld(ω) c
electron Compton wavelength,
where α = e2 /hc = 1/137, hc = 1.9733 ×
10−5 eVcm. CR spectrum is extended in certain
mc
ψ1 (δ) = 1 + fscr (q)(q − δ)2 qdq , (6)
wide or narrow spectral regions from radio up to δ
X-ray frequencies where n(ω) > 1. CR has found

mc
wide application in particle detection and identifi- ψ2 (δ) = 5
6 + (q 3 − 6δ2 q ln qδ
cation as well as in some advanced particle accel- δ
eration methods. +3δ2 q − 4δ3 )fscr (q)dq , (7)
Bremsstrahlung (B) Electromagnetic radiation
[2, 3] produced when the velocity of charged par- and fscr (q) is a factor taking into account the
ticles (mainly of light electrons) is changed due screening by the atomic electrons when the
to scattering or deceleration in the field of nuclei screening parameter ζ = 100/[Z 1/3 γ(1/x − 1)]
and electrons. Quantum electrodynamical (QED) is not much greater than unity. In Moliere approx-
theory of B has been developed in [9]. One can imation fscr (q) is given by
interpret and even obtain some results consid- 3
αi
ering the Compton scattering of pseudophotons fscr (q) = , (8)
on electrons and nuclei. When the electron en- 1
βi2 + q 2
ergy is higher than the so called critical energy,
Ecr ≈ 800 MeV/(Z + 1), radiation energy losses where α1 = 0.1; α2 = 0.55; α3 = 0.35; βi =
due to B are higher than ionization energy losses (Z 1/3 bi /121); b1 = 6; b2 = 1.2; b3 = 0.3. In
and increase proportionally to exp(x/X0 ) with x the case of complete screening ψ1 = ln 183Z −1/3
the target thickness and X0 the so called radiation and ψ2 = ln 183Z −1/3 − 1/6. The angular distri-
length, which at γ = E/mc2 1 is equal to bution of B is narrow, θ ∼1/γ.
In certain cases it is necessary to take into
A716.4
X0 = √ , (3) account 1) The medium polarization or Ter-
Z(Z + 1) ln(287/ Z) Mikaelian effect resulting in suppression of the
where A and Z are the mass and atomic num- spectral distribution at low photon energies ω <
ber of the radiator nuclei. At electron energies hωp γ where ωp is the plasma frequency and 2)
mc2 E ≤ 50 MeV one can estimate the The Landau-Pomeranchuk-Migdal (LPM) effect
angular-spectral distribution classically: at very high electron energies resulting in suppres-
sion of the spectra at low photon energies ω <
d2 σ 8Z 2 r02 γ 2 233 1 + γ 4 θ4 E 2 /ELP M , where ELP M = αm2 X0 /8π c =
= ln .
dωdΩ 137πω Z 1/3
(1 + γ 2 θ 2 )4 3.8 × 1012 eV X0 [cm].
238
Coherent Bremsstrahlung (CB): CB [2, 4] is

produced when charged particles pass through
crystal under small angle (but larger than the
channeling critical angles, see below) to the crys-
tallographic axes and/or planes in crystals and
when in some frequency regions the intensity of B
is enhanced due to interference of coherent pho-
tons. For history of initial CB theory and ex-
periments see [2, 4]. The production of CB can

be considered as follows [2, 4]: CB photons are
formed over a longitudinal distance called forma-
tion zone,
1 1−x
Zcoh = = 2λC γ . (9)
δ x
For values of γ sufficiently high, Zcoh becomes
larger than the periodic distances, d, between the
atoms in the axes or between the planes. There-
fore, the production of B photons in certain en-
ergy regions on ∼Ncoh ∼Zcoh /d nuclei takes place
coherently, and one can expect enhancement ap-
proximately ∼Ncoh 2 times.
Now the radiation cross section is the sum of

coherent and incoherent parts. As the calculations Figure 1: The spectral distribution of the intensity of
show [2, 4] both parts are given by expressions of CB measured at YerPhI using 4.5 GeV electron and
type (4) with 100 μm thick diamond (points with errors) and calcu-
2
lated (solid curve, top figure) and the calculated po-
ψ1coh (δ) = (2π)
N a3
|S(g)|2 larization dependence on photon energy (solid curve,
g bottom figure) (by courtesy of R.O. Avakian).
δg⊥2
e−2W f scr (g) g 2 , (10)
||
2 (1 − e−2W ) which decreases the incoherent cross

ψ2coh (δ) = 6 (2π)
N a3
|S(g)|2
g section by ∼10% compared with (4).
2 (g −δ)
δ 2 g⊥
e−2W f scr (g) 4
⊥
g||
. (11) CB photons are polarized, and degree of lin-
ear polarization is equal to
In (10) and (11) N is the number of atoms per
unit cell, a the crystal constant, g|| and g⊥ are the dσ|| − dσ⊥ 2uv 2
P = = ,
longitudinal and transverse components of the re- dσ|| + dσ⊥ 1 + u − 4uv(1 − v)
2
ciprocal lattice vector g with g = 2π/dhkl and (13)
dhkl the interplanar spacing defined by the Miller with u = 1 − x, v = δ/q|| , q|| the longitudinal
indices h, k, l, and the summation is carried out component of the transfer momentum. From (13)
over all g with g|| > δ. the polarization is maximal at v = 1. Usually the
( (2 polarization is calculated from the measured spec-
( (
( N
( tral distribution because the direct measurement
|S(g)|2 = (( eigrj (( , (12) of polarization is much more difficult. Figure 1
( j=1 (
shows the dependence of intensity and polariza-
is the structure factor and the factor e−2W the tion on the energy of the CB photons produced
Debye-Waller factor, where W = 0.5 < u21 > g2 , by 4.5 GeV electrons ipassing through a 100 μm
with < u21 > the mean square of 1D thermal thick diamond crystal oriented for maximal polar-
amplitude of the crystal atoms. For the incoher- ization.
ent part of the cross section the expressions under CB photons have found wide application in
the integrals (6) and (7) are multiplied by a factor photoproduction experiments.
239
Channeling Radiation (ChR) Planar and ax-

ial ChR [4, 11] is produced by channeled par-
ticles entering into crystal under angles smaller
than the so called critical or Lindhard angles,
θcr = (2U/E)1/2 , with respect to crystallo-
graphic planes and axes with averaged potential
depth U, respectively. When the electron energy
is higher than a few hundred MeV, and the num-
ber of transverse energy levels of the particles in

the potential well is much larger than one, one can
consider ChR classically as radiation of a mov-
ing dipole. At lower energies quantum approach

is necessary, and ChR is produced as a result of
transitions between a finite number of transversal
energy levels of electrons in crystal. Apparently
ChR has been observed for the first time in the
works [10], while the corresponding theory has Figure 2: The spectral distribution of ChR of positrons
been developed by many authors gradually up to in 100 μm thick diamond. The circles are the results
its transparent interpretation (see [11]). of measurements at SLAC, the curves are the results
In dipole approximation, applicable roughly of calculation for the first 3 harmonics (dashed curves)
when θcr γ < 1, the spectral distribution of ChR and for the sum (solid curve) (by courtesy of R.O.
like the one for undulator radiation consists of Avakian).
many harmonics with numbers n = 1, 2, 3 . . ..
and for a parabolic potential U (x) = U0 x2 is
given by [11] of ChR are close to those of transition radia-
( (2 tion for electron energies of a few tens of MeV.
∞ ((β̇ (
( Like the transition radiation ChR has found wide
dIChR e2 ⊥n ω ω2
= ω 1 − 2 + 2 application.
dω c Ω2n ωn ωn2
n=1
(14) Parametric X-ray Radiation (PXR): PXR [4,
where ωn = 2Ωn γ 2 , Ωn = n(2U0 /m)1/2 the 6] is produced when charged particles pass the
crystallographic planes. It can be considered as
oscillation frequency, β̇⊥n the transverse Fourier
Bragg diffraction of pseudophotons. Differently
component of the acceleration (for examples see
from most of the other types of radiation, PXR
[11]). In the “dipole” case the first harmonic is
is monochromatic and emitted under large angles
dominant. For undulator radiation the particle
θ 1/γ, close to Bragg angles. It was predicted
oscillation period is given by the undulator pe-
as a special case of resonance transition radiation
riod, and ωmax ∼γ 2 , while for ChR the particle
in 1969 in the Russian publication of the mono-
oscillation period depends on particle energy and
graph [2] using the perturbation theory of the clas-
ωmax ∼γ 2/3 . As in the case of undulator and many
sical electrodynamics and was first observed in
other types of radiation ChR photons are emitted
[12]. Like the theory of Bragg diffraction the PXR
under angles ∼1/γ, and there is a relation between
theory was developed in kinematical (correct for
the frequency and the emission angle.
thin crystals when multiple reflections from the
Figure 2 shows the spectral distribution of
crystallographic planes can be neglected) and dy-
the number of ChR photons produced by 6 GeV
namical approaches [13].
positrons channeled in a 100 μm thick diamond.
According to the kinematical approximation
At electron energies lower than a few hun-
of PXR (the more general dynamical and quan-
dreds MeV ChR spectral-angular distribution
tum mechanical theories of PXR give no essential
is calculated quantum mechanically by solving
corrections) the number of photons NPXR pro-
the Schrödinger equations to find the energy
duced by an electron in a crystal with thickness L
levels, the population and transition probabili-
and absorption length Labs , per unit angles θx and
ties over all the crystal thickness. Also multiple
θy — the parallel and perpendicular components
scattering, dechanneling, polarization of ChR etc.
must be taken into account. The characteristics of the PXR emission angle θ = θx2 + θy2 with
240
respect to the Bragg reflection direction — is

d2 NPXR α
= ωB Fgeom (σ̂, v̂, k̂) (15)
dθx dθy 4π
|χg (ωB )|2 θx2 cos2 (2θB ) + θy2
2 ,
sin2 θB 2
θx2 + θy2 + θph
where ωB and θB are the Bragg frequency and
angle for the given crystal orientation,

( (
( (
(σ̂ k̂(
Fgeom (ĝ, v̂, k̂) = Labs (16)
|σ̂v̂|
⎡ ⎛ ⎞⎤ Figure 3: Relative intensity of SPR vs θ for various
⎣1 − exp ⎝− L values of h/d indicated next to the curves.
( ( ⎠⎦
( (
Labs (σ̂ k̂ (
Many models for analytical and numerical calcu-
is a geometrical factor; σ̂, v̂ and k̂ are unit vec- lations for the angular distribution of the intensity
tors in directions perpendicular to the surface, of d2 I/dzdΩ of SPR per unit length of the grating
the particle particle v, and of the emitted photon, have been developed [6]. At relativistic energies
respectively; [15]

ω 2 F (g) d2 ISP 4e2 4πh/d
Xg (ω) = S(g) exp(−W ) P 2 (17) = sin θ exp −
2
ω Z dzdΩ πd2 γ (1 − β cos θ)
is the coefficient of the Fourier expansion of the 1
dielectric susceptibility; F (g) the atomic scatter- (1 − β cos θ)3
ing factor; and
Figure 3 shows SPR angular distribution.
2 1 ωp2 At present SPR has been experimentally stud-
θph = 2+ 2. (18)
γ ω ied in the region from millimeter up to optical
In Eq.(15) the frequency is related to the PXR wavelengths using electrons with E = (0.02 −
angle through 855) MeV. SPR can be used for constructing tun-
c |v̂g | able compact sources (also SPR FELs) and for di-
ω= √ , (19) agnostics of particle beams.
1/β − εv̂ k̂
which is close to the Bragg formula for X-ray References
diffraction. In the last expression ε = 1−(ωp /ω)2
is the dielectric constant. [1] J.D. Jackson, Classical Electrodynamics, 3rd
For various crystals and geometries the total edition, J. Wiley&Sons, New York, 1999
number NPXR of emitted PXR photons per elec- [2] M.L. Ter-Mikaelian, The Influence of the
tron increases with γ up to γcr ∼ωB /ωp ∼200, sat- Medium on High Energy Processes at High En-
urating at NPXR ∼10−5 − 10−7 . Beginning from ergies, Academy of Science of Armenia, Yerevan
1969; High Energy Electromagnetic Processes
γcr the contribution of the diffracted transition ra-
in Condensed Media, Wiley Interscience, New
diation (DTR) photons with almost the same fre-
York, 1972
quency and emission angle slightly exceeds the
[3] V.B. Berestetski, E.M. Lifshitz, L.P. Pitaevski,
pure PXR yield. Kvantovaya Electrodynamika, Moscow, Nauka,
Smith-Purcell Radiation (SPR) SPR [6] is 1989 (in Russian)
the coherent diffraction radiation produced by [4] P. Rullhusen, X. Artru, P. Dhez, Novel Radiation
charged particles flying over a grating with pe- Sources Using Relativistic Electrons, World Sci-
riod d at distance h of the order γβλSP /(4π). It entific, Singapore, 1998
was discovered and studied in [14] in which using [5] H. Wiedemann, Synchrotron Radiation,
Huygens principles it has been shown that for the Springer, Berlin, 2003
n-th harmonic [6] A.P. Potylitsin, Electromagnetic Radiation of
d 1 Electrons in Periodic Structures, Springer,
λSP = − cos θ Berlin-Heidelberg, 2011
n β
241
Sec.3.2: IMPEDANCES AND WAKE FUNCTIONS
[7] P.A. Cherenkov, Doklady Akad.Nauk SSSR, v.2 is a function of only z = s − vt, and satisfies the
(1934) 451 Panofsky-Wenzel theorem [3],
[8] I.E. Tamm, I.M. Frank, Doklady Akad.Nauk ∂
SSSR, v.14 (1937) 107 ∇⊥ F = F⊥ (5)
[9] H. Bethe, W. Heitler, Proc.Roy.Soc. 146 (1934) ∂z
83 Here denotes longitudinal and ⊥ denotes trans-
[10] R.L. Walker et al, PRL 25 (1970) 5; PR A11 verse.
Definition of wake functions Wm , Wm ,W ,W
(1975) 736 ⊥
[11] M.A. Kumakhov, Izluchenie Kanalirovannikh

Chastits v Kristallakh, Energoizdat, Moscow, F⊥ = −eIm Wm (z)mr m−1
1986 (In Russian) ×(r̂ cos mθ − θ̂ sin mθ)
[12] S.A. Vorobev et al, Zh.Eksp.Teor.Fiz. 41 (1985)

F = −eIm Wm (z)r m cos mθ (6)
3
[13] V.G. Barishevski, I.D. Feranchuk, A.P. Ulyanov, Here a prime denotes d/dz. Wm is sometimes
called transverse wake function and Wm longi-
Parametric X-Ray Radiation in Crystals: Theory,
Experiments and Application, Springer Tracts in tudinal wake function. Similarly, for a general
Modern Physics, Heidelberg, 2005 cross-section, W⊥ and W are defined as
[14] S.J. Smith, E.M. Purcell, PR 92 (1953) 1069
[15] J.E. Walsh, K. Woods, S. Yeager, NIM A341 ⊥ = −eq 2x1 W⊥ (z)x̂ (transverse dipole)
F
(1994) 277 F = −eq W (z) (longitudinal monopole)
(7)
3.2 IMPEDANCES AND WAKE Properties of wake functions [1]

• Wm (z) = 0, Wm (z) = 0 for z > 0 (causal-
FUNCTIONS
ity).
3.2.1 Definitions and Properties of (z) ≥ 0 for z → 0− .
• Wm (z) ≤ 0, Wm
Impedances and Wake Functions • Wm (0) = 0 (in most cases, except space
T. Suzuki, KEK
charge).
Driving current • Wm (0) = 1 W (0− ) (fundamental theorem
2 m
Axially symmetric case [1] of beam loading [4]).
[coordinates (r, θ, s)] Charge q, offset r = a, θ = • Wm (0− ) ≥ |W (z)| for all z.
m
0, velocity v in s-direction. 0 (z)dz ≥ 0.
q • −∞ Wm
ρ = δ(r − a)δ(θ)δ(s − vt) (1) Dimensionalities are
a
∞
Im cos mθ [Wm ] = [Ωs−1 m−2m+1 ] = [VC−1 m−2m+1 ]
= δ(s − vt)δ(r − a)
m=0
πa m+1 (1 + δ
m0 ) [Wm ] = [Ωs−1 m−2m ] = [VC−1 m−2m ]
j = vρŝ, Im = qam (m−th moment) [W⊥ ] = [Ωs−1 m−1 ] = [VC−1 m−1 ]
[W ] = [Ωs−1 ] = [VC−1 ]
General cross-section [2]
Definition of impedances
[coordinates (x, y, s)] ∞
dz −iωz/v
js = qvδ(x)δ(y)δ(s − vt) (2) Zm (ω) = e Wm (z)
−∞ v
(longitudinal monopole)
i dz −iωz/v
js = qvδ(y)[δ(x − x1 ) − δ(x + x1 )]δ(s − vt) Z⊥m (ω) =
v/c v
e Wm (z) (8)
(x1 → 0) (transverse dipole) (3)
Time dependence of e−iωt is assumed. Some au-
Wake potential The wake force F = e(E + thors use ejωt (j = −i). Z ⊥ (ω) and Z (ω) are
v × B)
is integrated over a period L (holding z = defined in the same way as the Fourier transforms
s − vt fixed). The wake potential of W⊥ (z) and W (z) for a general cross-section.
L/2 Dimensionalities are
=
F dsF (4) [Zm ] = [Ωm−2m ], [Z⊥m ] = [Ωm−2m+1 ]
−L/2
242
Properties of impedance [1] [2] R.L. Gluckstern, J. van Zeijts, B. Zotter, PR E47
• Zm (ω) = ωc Z⊥m (ω) (1993) 656
(Panofsky-Wenzel theorem [3]). [3] W.K.H. Panofsky, W.A. Wenzel, RSI 27 (1956)
% ∗ 967
Zm (ω) = Zm (−ω)
• ∗ (ω) = −Z [4] P.B. Wilson, AIP Proc. 87 (1981) 450
Z⊥m ⊥m (−ω)
[5] R. Gluckstern, PR D39 (1989) 2780
(reality of wake functions).

⎧ ∞ [6] A. Novokhatski, A. Mosnier, PAC 97 (1997)
⎨ 0 dωImZ⊥m (ω) = 0 p.1661
∞ ImZm (ω)
• dω =0 [7] K. Bane, A. Novokhatski, SLAC-AP 117 (1999)
⎩ 0 ω
[8] K.-Y. Ng, PR D42 (1990) 1819
ReZm (0) = 0
(Wm (0) = 0, in most cases). [9] H. Henke, O. Napoli, EPAC 90 (1990) p.1046
' ∞
ImZ (ω )
ReZm (ω) = π1 P.V. −∞ dω ωm −ω
• ∞ ReZ (ω )
ImZm (ω) = − π1 P.V. −∞ dω ωm −ω 3.2.2 Impedance Calculation, Frequency
(causality, Hilbert transform) Domain
The same expressions apply to Z⊥m . R.L. Gluckstern, Deceased 2008
% S.S. Kurennoy, LANL
ReZm (ω) ≥ 0 for all ω
•
ReZ⊥m (ω) > 0 if ω > 0, < 0 if ω < 0 The impedance calculation is reduced to the solu-
• Z⊥1 ≈ 2c
b2 ω Z0 , Z⊥m ≈ 2c
b2m ω Z0 , tion of an electrodynamics problem of finding the
fields (voltages, forces) produced in the vacuum
Zm ≈ b2m
2
Z0
chamber by a given beam current. Depending
These are approximate expressions relating
upon the way chosen to formulate and solve the
Z⊥ and Z . They are exact for resistive
problem one may classify the calculation meth-
round pipe, b = vacuum chamber radius (in
ods (i) by the initial formulation of the problem
appropriate freq. range, see Sec.3.2.4).
— in frequency (excitation by a current harmonic
Limiting value of wake For periodic, cylin- with a given frequency) or in time domain (first
drically symmetric structures whose closest ap- calculating wakes, and then finding impedances
proach to the axis is b, the steady state wakes have via Fourier transform); (ii) by the method of solu-
the properties tion — analytic, semianalytic, or numerical; (iii)
Z0 c 2Z0 c by the approach used — electrotechnical (a cham-
W0 (0− ) = , W1 (0− ) = (9) ber element is replaced by an equivalent circuit) or
πb2 πb4
electrodynamic (solving the Maxwell equations).
where W0 , W1 , are respectively the longitudinal,
The convenience of a method depends on the
transverse (dipole) wakes per unit length and 0−
geometry of a chamber element and on the fre-
signifies a test particle position right behind the
quency range of interest. Below we briefly list
exciting charge; with W1 (0− ) = 0 and Z0 =
the main calculation methods in the frequency do-
377 Ω. Eq.(9) is true for a resistive pipe [1], a
main mentioning some of their applications and
disk-loaded accelerator structure [5], a pipe with
limitations. More detailed analyses and refer-
small periodic corrugations [6, 7], and a dielec-
ences can be found in specialized reviews and
tric tube within a metallic pipe [8]. For a non-
books [1, 2, 3, 4, 5, 6, 7, 8]. Explicit expres-
round periodic structure the result will be a dif-
sions for the impedances of some discontinuities
ferent constant [e.g. for 2 parallel metallic plates
are given in Sec.3.2.4.
separated by 2b, W|| (0− ) = πZ0 c/(16b2 ) [9]],
but again a constant dependent on transverse di- Electrotechnical approach In this approach an
mensions only and independent of material prop- element of the vacuum chamber, typically having
erties. We see that for short bunches the longitu- a complicated geometry, is replaced by an equiv-
dinal wake approaches a fixed maximum, and the alent circuit, such as a transmission line or RLC-
transverse wake approaches zero. circuit, with circuit parameters (e.g., its resonance
frequency, Q-value, or line length) determined
from heuristic considerations. The approach typi-
References
cally allows one to obtain results in a simple form,
[1] A.W. Chao, Physics of Collective Beam Instabili- but is limited to the frequency range (usually low
ties in High Energy Accelerators, Wiley (1993) frequencies) in which the replacement is valid.
243
The broad-band resonator model for the The matching technique was used to calculate
impedance of a vacuum chamber can be regarded the impedance of a cylindrical chamber with an
as an example of this approach [9]. Another ex- axisymmetric pill-box cavity with fields matched
ample is the calculation of the impedance of strip- on the pipe radius: for periodic structures [18],
line BPMs [10, 11]. or, with added radiation conditions at infinity, for
a single cavity [19]. The matrix equation usually
Electrodynamic methods listed below solve
allows an analytic solution in the low-frequency
the Maxwell equations with a given current to find
limit and gives approximate expressions for res-
the fields excited in the chamber. While doing so,

onances; in a general case, the truncated sys-
proper boundary conditions (BCs) should be im-
tem of equations is solved numerically. Two
posed on the fields. In addition to the metal BC
codes for periodic pill-box structures are well-
on the vacuum chamber walls, other conditions
known: KN7C [20] for axisymmetric modes, and

may be necessary to “close” the problem: (i) pe-
TRANSVRS [21] for azimuthal ones.
riodicity conditions; (ii) “closure” conditions —
The field matching in transverse planes was
BCs on imaginary surfaces closing the region of
used to calculate the impedance of an arbitrary ax-
interest such as imaginary metal planes in beam
isymmetric structure with a discontinuity approx-
pipes far from the cavity to study its impedance at
imated by a sequence of stepwise transitions [22].
frequencies below the pipe cut-off; (iii) radiation
The technique is applied naturally to collimators
conditions for open nonperiodic structures, such
[23] and irises [24]. Besides, the approach can
as a cavity coupled to a beam pipe.
be used for an abrupt step between waveguides of
Eigenfunction Method Since a solution of the different sizes [25].
inhomogeneous Maxwell equation can be repre- Perturbation methods For small variations of
sented by a series in eigenfunctions (EFs), the the transverse dimensions of the chamber pertur-
impedance can be expressed as a formal se- bation methods are applicable. If the radius of
ries. The method allows one to evaluate ei- an axisymmetric chamber is r = b(z) = b(1 +
ther resonance impedance values for the lowest εs(z)), where ε = h/2b 1 is a small parame-
modes with frequencies below the chamber cut- ter, the BC on the perfectly-conducting wall
off whose EFs can be calculated (usually numer-
ically, in the “closed cavity” approximation, us- Ez + b (z)Er r=b(z) = 0, (1)
ing standard codes, e.g., in [12], or analytically
as well as the fields, can be expanded in powers
[13]), or the low-frequency impedance in the sim-
of ε. Here h is the maximal depth of variations,
plest models where all EFs can be found analyti-
and b is the mean chamber radius. It leads to a
cally [14]. For a numerical solution one can apply
recurrent sequence of equations with BC at r = b,
any frequency-domain eigenvalue solver like Su-
and to the impedance in the form of a series Z =
perfish [15] or CST [16].
Z (0) + ε2 Z (2) + ε3 Z (3) + . . ., where Z (0) is the
A modification of the EF-method allows one
impedance of the smooth chamber and Z (1) = 0.
to calculate resonances also at frequencies above
The technique is known in mathematics as “BC
the cutoff, but only for periodic structures [17].
transfer to the regular boundary”. It was applied
However, it requires computing numerically the
in the lowest order to calculate the low-frequency
structure dispersion characteristics which is time-
impedance and resonances [26]; the higher orders
consuming, especially for long periods.
in ε were considered in [27].
Field matching method In structures which The ε-expansion works when b(z) is a single-
can be subdivided into a few simple subdomains valued function of z and its derivative | b (z) | is
such that the wave equations in each of them al- limited. An additional restriction applies at low
low the separation of variables, the field matching frequencies: D h, where D is the structure
method, also called the partial domain method, is period [28]; the method fails for bellows cor-
applied. Solutions in each subregion are expanded rugations, in which the period D is comparable
into a complete set of EFs with unknown coeffi- with h.
cients. Field matching on the subregion interfaces The impedance of a corrugated vacuum
and BCs lead to an infinite system of linear equa- chamber can be calculated at low frequencies [28]
tions for these coefficients, which usually can be and in the resonance region [29] by the matrix
truncated to a finite size. method. It is applicable to periodic structures
244
with an axial or flat symmetry, when the boundary D obstacles protruding into the chamber like a
b(z) is a single-valued function, | b (z) | is limited, post or a mask [38]. Another example of using
but the boundary variation h = max(b(z) − b) impedance BCs for calculating the impedances of
is not necessarily small. Substituting the solution perforated beam pipes is [39].
— a series in space harmonics with unknown co- Integral-equation methods The impedance
efficients — into the BC, and expanding it into a calculation for a discontinuity in an infinite pipe
complete set gives an (infinite) matrix equation. can be reduced to the solution of an integral equa-
For h b it reduces to the ε-expansion. The ma-
tion. For an axisymmetric cavity an integral equa-

trix equation is solved numerically after the trun- tion was derived [40, 41], in which the unknown
cation. The method is convenient for short-period function is the field at the cavity-pipe boundary,
structures like bellows because truncated matrices and the kernel is a series over the cavity EFs.
as small as 5 × 5 are sufficient: the truncation size An integral over its solution gives the impedance.
required is proportional to the period length [29]. When the discontinuity is small, the impedance is
Method of effective boundary conditions The expressed in an analytical form [42]. The same is
idea is to impose some effective BCs, usually on possible for some cavities in the high-frequency
a smooth surface, which simplify the solution for limit. The results can also be generalized to sev-
the fields. One example is “impedance BCs” in eral discontinuities and, in the limit, to periodic
a cylindrical waveguide of radius b with the rect- structures [41].
angular corrugation having pitch D and depth h The impedance can frequently be calculated
(h b) as a variational expression in the unknown field

g ωh in the integral equation. Using a truncated series
Ez = iZ0 1 − tan Hϕ ; Eϕ = 0 ; with arbitrary coefficients for the unknown field
D c
(2) leads to rapidly convergent numerical results for
at r = b, where g/D is the filling coefficient [30]. the impedance, e.g., [24, 36].
The other effective BCs, obtained by conformal Another integral equation was obtained for
mapping of the corrugation cell into a rectangle axisymmetric discontinuities of a rather general
[31], replace the corrugated surface by an equiv- shape, with the only limitation that b(z) is a
alent smooth one at r = bef f having a “dielec- single-valued function [43]. The impedance is ex-
tric coating” with permittivity depending on z. pressed as its solution at a particular point. For
The high-frequency impedance resonance due to simple boundaries the kernel can be evaluated an-
a synchronous wave slowed down by the corru- alytically, but in the general case it is a definite in-
gations was analytically calculated [31] with this tegral. A more refined development [44] leads to
approach. an integro-algebraic equation, which works better
A version of the effective BCs approach was above the pipe cut-off.
applied for calculating the coupling impedances The impedance calculation for a cylindrical
of small pumping holes in the chamber walls semi-infinite waveguide with perfectly conduct-
[32, 33] for wavelengths larger than a typical hole ing walls was reduced to a solution of a set of two
size. The method, based on the Bethe theory integral equations [45, 46], which can be solved
of diffraction by small holes [34], replaces the by the factorization technique.
hole by effective “magnetic” currents to satisfy Diffraction methods Methods of the diffrac-
BCs. For small holes, the fields radiated by these tion theory are used to calculate impedances at
currents are approximated by those due to elec- high frequencies, ωb/c 1, where b is a typi-
tric and magnetic dipoles, which allows an an- cal transverse size of the chamber, following the
alytical evaluation of the impedances. The ap- pioneer work [47]. A contradiction between two
proach works for all small discontinuities, and the results on the high-frequency impedance behavior
impedance and parameters of the trapped modes — either Re Z(ω) ∼ ω −1/2 [47, 14] or ∼ ω −3/2
(Sec.3.2.7) are expressed in terms of the electric [48] — existed for a long time. It has been re-
and magnetic polarizabilities of the discontinuity solved first by the matching technique [49]: for a
[35]. The polarizabilities have been calculated for single pill-box cavity the high-frequency behav-
holes in a wall of a finite thickness by the varia- ior Re Z(ω) ∼ ω −1/2 holds, whereas for an in-
tional method [36], for axisymmetric cavities and finite periodic structure Re Z(ω) ∼ ω −3/2 . For
irises using conformal mapping [37], and for 3- a finite number M of cavities there is a transi-
245
tion from ω −3/2 -behavior (which takes place in anomalous skin effect [71].
√ A synchronous mode
the frequency range L/b ωb/c M 2/3 L/b) [68, 69] with fres ∝ 1/ h, where h is the corru-
to ω −1/2 when ωb/c M L/b, where L is the gation (roughness) depth, is similar to one found
distance between elements. The result was con- in a dielectric-layer model [72]; it can be related
firmed by the diffraction model in [50], as well as to an even earlier result [31] obtained using the
in [41], where the asymptotic behavior of Im Z effective BCs. The high-frequency impedances of
was also studied, see [1]. a finite-length resistive insert into a smooth pipe
were calculated in [73, 74]; various frequency
Other methods A few methods which solve the

Maxwell equations in the frequency domain di- regimes are studied in [75]. The impedances of
rectly do not fit the above classification. They tapered transitions were originally derived for an
include analytical calculations of the impedance axisymmetric case in the low-frequency approxi-
of a smooth chamber at finite beam energy (the mation [76]. More recently, the perturbation cor-
space-charge impedance) [51], the resistive-wall rections were calculated [77], and the derivations
impedance for an axisymmetric chamber [52] or were extended to the rectangular [78] and ellip-
the chamber with an arbitrary cross section [53], tical [79] geometry. Many recent results are re-
and the impedance due to the curvature of a viewed in [8]. Related to the impedance, the
smooth chamber [54]. One should mention here problem of beam shielding has been addressed —
two proofs that the impedances of discontinuities shielding by an array of concentric wires [80] and
asymmetrical in the longitudinal direction are in- by thin conductors [81].
dependent of the beam direction [40, 55]. Codes Some codes used for impedance calcula-
The impedance of an arbitrarily-shaped cav- tions are listed in Tab.1; for an extensive list of
ity with semi-infinite pipes can be calculated us- older and specialized codes see [82]. The table
ing known frequency-domain codes for field com- gives the vacuum chamber layout, where the code
putations [56], with the radiation BCs simulated is applicable, and the code method in frequency
by introducing a medium with a small imaginary (f -) or in time (t-) domain; W → Z means find-
permittivity, which fills the smooth pipe. Compu- ing impedances Z by Fourier transforming com-
tations, however, are to be performed separately puted wakes W . Eigensolvers find frequencies
for each frequency. and fields of resonant modes, from which Rsh /Q
Recent developments The developments have and loss factors can be computed, e.g. [61], usu-
been mostly driven by requirements of projects ally in post processing. The ‘Output’ column
under construction and in design stage. For low- shows the code output: the longitudinal (L) or
energy high-intensity accelerators, it is important transverse (T ) impedance Z or loss factor k; R/Q
to know the impedances of non-ultrarelativistic refers to the impedance of resonant modes and
(fixed β = v/c < 1) beams. The β-dependence Z(f ) means the impedance in some frequency
was studied analytically for the impedances of range. The code references include their web ad-
small holes [57, 58] and small discontinuities of dresses; codes available free are marked by *.
arbitrary shape [59], for the resistive-wall wakes The symbol ‘p’ shows the codes with parallel ver-
and impedances [60], and for loss factors [61]. A sions implemented, FIT stands for Finite Integra-
detailed knowledge of the impedances at very low tion Technique (Sec.3.2.3), FD for Finite Differ-
frequencies is important for the beam stability in ence, and FE for Finite Element methods. Time-
large hadron colliders. The calculations of the domain codes are generally more universal and
low-frequency impedances in multi-layer cham- applicable for various discontinuities. They com-
bers [62, 63, 64, 65], based on the field match- pute wake potentials created by a bunch of a fixed
ing and including skin effects, improve upon the shape. However, one can encounter difficulties in
earlier results, see in [7]. High-energy electron calculating high-frequency impedances due to a
linacs and colliders, as well as x-ray FELs, require limited bunch spectrum: wakes of short bunches
very short bunches, which stimulated renewed are needed to accurately calculate impedances at
interest in short-range wakes and correspond- high frequencies, while computing such wakes re-
ing very-high-frequency impedances. The high- quires very small mesh steps. Code NOVO [83]
frequency impedances have been calculated for calculates the wakes of very short bunches more
the wall roughness [66, 67], small wall corruga- efficiently due to its implicit scheme for solving
tions [68, 69], and resistive wall [70] including the field equations. On the other hand, to resolve the
246
Table 1: Some codes for impedance computations (see text for symbol definitions).
Layout Code Method Output Reference

periodic axisymm. KN7C f , matching fres , kL [20]*
rectang. corrugations TRANSVRS f , matching fres , kT [21]*
axisymmetric Superfish f , EF fres , R/Q [15]*
discontinuities ABCI t, W → Z ZL,T (f ), k(f ) [86]p*
NOVO t, W → Z WL,T (s) [83]*
CST MWS(E) f , EF, FIT fres , R/Q [16]p

general CST PS t, FIT, W → Z ZL,T (f ) [16]p
discontinuities GdfidL t, FD, W → Z ZL,T (f ) [87]p
HFSS f , EF, FE fres , R/Q [88]p
impedances at very low frequencies, one has to [20] E. Keil, NIM 100 (1972) 419;
compute very long wakes, e.g. [84], which can keil.web.cern.ch/keil/Fortran/
cause the computations to become numerically [21] K. Bane, B. Zotter, Proc. 11th
unstable. A new computational scheme [85] that HEACC (1980) 581 www.slac.stan-
eliminates numerical dispersion in the beam di- ford.edu/grp/ara/groups/BIG/codes.shtml
rection and therefore is more stable for calculating [22] L. Vos, CERN SPS/86-21(MS) (1986)
long wakes was implemented in the code ECHO. [23] S. Kheifets, K. Bane, H. Bizek, SLAC-PUB-
4097 (1987)
[24] H. Okamoto, S. Jiang, R.L. Gluckstern, PR E50
References (1994) 1501; PA 52 (1996) 179
[25] S. Heifets, S. Kheifets, SLAC-PUB-3965 (1986)
[1] S. Heifets, S. Kheifets, RMP 63 (1991) 631 [26] S. Krinsky, Proc. 11th HEACC (1980) 576
[2] S.S. Kurennoy, Phys. Part. Nucl. 24, (1993) 380; [27] S. Krinsky, R. Gluckstern, PAC 81, 2621
CERN SL/91-31(AP)(1991) [28] S. Kheifets, B. Zotter, NIM A243 (1986) 13
[3] R.L. Gluckstern, PAC 93, 3219; CERN-2000- [29] S.S. Kurennoy, S.V. Purtov, PA 34 (1990) 155
011 (2000) [30] B.Z. Katzenelenbaum, High-Frequency Electro-
[4] A.W. Chao, Physics of Collective Beam Instabil- dynamics, Nauka (1966)
ities in High Energy Accelerators, Wiley (1993) [31] V.I. Balbekov, IHEP 85-128 (1985)
[5] K.-Y. Ng, AIP Proc. 184 (1989) 472; P. Wilson, [32] S.S. Kurennoy, PA 39 (1992) 1
ibid 525 [33] R.L. Gluckstern, PR A46 (1992) 1106; 1110
[6] L. Palumbo, V.G. Vaccaro, M. Zobov, Proc. CAS [34] H.A. Bethe, PR 66 (1944) 163
(1993), CERN 95-06, 331 [35] S.S. Kurennoy, R.L. Gluckstern, G.V. Stupakov,
[7] B. Zotter, S. Kheifets, Impedances and Wakes in PR E52 (1995) 4354
High-Energy Particle Accelerators, World Scien- [36] R.L. Gluckstern J.A. Diamond, IEEE Trans.
tific (1998) MTT 39 (1991) 274
[8] G.V. Stupakov, PAC 09, 4270 [37] S.S. Kurennoy, G.V. Stupakov, PA 45 (1994) 95
[9] A. Hofmann, Proc. 11th HEACC (1980) 540 [38] S.S. Kurennoy, PR E55 (1997) 3529
[10] R.E. Shafer, UM-HE 84-1 (1984) 155 [39] S. Petracca, PR E60 (1999) 6030
[11] K.-Y. Ng, PA 24 (1988) 93 [40] R.L. Gluckstern, B. Zotter, CERN LEP-Note 613
[12] K.-Y. Ng, Physics of the Superconducting Super- (1988)
collider, Snowmass (1986) 592 [41] R.L. Gluckstern, PR D39 (1989) 2773; 2780
[13] G.V. Stupakov, S.S. Kurennoy, PR E49 (1994) [42] R.L. Gluckstern, F. Neri, PAC 89, 1271
794 [43] S.S. Kurennoy, S.V. Purtov, PA 36 (1992) 223
[14] G. Dome, CERN SPS/85-27 (ARF) (1985) [44] R.L. Warnock, SLAC-PUB-6038 (1993)
[15] K. Halbach, R.F. Holsinger, PA 7 (1976) 213; [45] S. Heifets, L. Palumbo, V.G. Vaccaro, CERN-
laacg.lanl.gov LEP-TH/85-23 (1985)
[16] T. Weiland et al, SLAC-303 (1986) 282; see [46] S. Heifets, L. Palumbo, CERN LEP Note 580
Sec.3.2.3; CST Studio Suite, www.cst.com (1987)
[17] S.S. Kurennoy, S.V. Purtov, IHEP 88-11 (1988) [47] J.D. Lawson, Report RHEL/M144 (1968); PA 25
[18] E. Keil, B. Zotter, PA 3 (1972) 11 (1990) 107
[19] H. Henke, PA 25 (1990) 183
247
[48] K. Bane, P. Wilson, Proc. 11th HEACC (1980) [86] Y.H. Chin, PAC 93, 3414;
592 http://abci.kek.jp/abci.htm
[49] S. Heifets, S. Kheifets, PR D39 (1989) 960 [87] W. Bruns, PAC 97, 2651; www.gdfidl.de
[50] S.A. Heifets, PR D40 (1989) 3097 [88] High Frequency Structure Simulator,
[51] C.E. Nielsen, A.M. Sessler, K.R. Symon, Proc. www.ansys.com
Int. Conf. High Energy Acc. (1959) 239
[52] V.K. Neil, A.M. Sessler, RSI 36 (1965) 429
[53] R.L. Gluckstern, J.B.J. van Zeijts, B. Zotter, PR 3.2.3 Impedance Calculation. Time Domain
E47 (1993) 656 E. Gjonaj, T. Weiland, TU Darmstadt

[54] R.L. Warnock, P. Morton, PA 25 (1990) 113; K.-
Y. Ng, ibid 153
The primary quantity of interest in time domain
[55] S.A. Heifets, SLAC/AP-79 (1990)
calculations is the longitudinal wake potential,

[56] R.L. Gluckstern, F. Neri, Proc. HEACC (1987)
v.2, 170
1 ∞ s+z
[57] S. De Santis et al, PR E54 (1996) 800 W|| (x, y, s) = − dzEz (x, y, z, ),
[58] A.M. Al-Khateeb et al, J. Phys. G 27 (2001) 2471 q −∞ c
[59] S.S. Kurennoy, PRST-AB 9 (2006) 054201 (1)
[60] F. Zimmermann, K. Oide, PRST-AB 7, 044201 where q is the charge of a wakefield inducing
(2004) bunch and s is the relative position of a wit-
[61] S.S. Kurennoy, PRST-AB 2,014201 (1999) ness particle with respect to the bunch. Coupling
[62] A. Burov, V. Lebedev, EPAC 2002, 1452 impedances and other related parameters, such
[63] E. Métral, B. Zotter, B. Salvant, PAC 07, 4216 as transverse kick factors are determined from
[64] M. Ivanyan et al, PRST-AB 11, 084001 (2008) W|| (e.g., using the Panofsky-Wenzel theorem
[65] H. Hahn, PRST-AB 13, 012002 (2010) [1]).
[66] G.V. Stupakov, PRST-AB 1, 064401 (1998) The calculation of (1) is based on the tran-
[67] G.V. Stupakov et al, PRST-AB 2, 060701 (1999) sient solution of the full set of Maxwell’s equa-
[68] A. Mostacci et al, PRST-AB 5, 044401 (2002) tions. Usually, no simplifying assumptions are in-
[69] K. Bane, G.V. Stupakov, PRST-AB 6, 024401 volved which would limit the applicability of the
(2003) method. Accelerator structures of arbitrary 3D-
[70] K. Bane, G.V. Stupakov, PAC 05, 3390
geometry with resistive and rough surface walls
[71] B. Podobedov, PRST-AB 12, 044401 (2009)
as well as bunches of arbitrary distribution can
[72] A.V. Novokhatsky, A. Mosnier, PAC 97 (1997)
be handled. The drawback resulting from such
1661
[73] S. Krinsky, B. Podobedov, R.L. Gluckstern,
a general approach is the large numerical effort
PRST-AB 7, 114401 (2004) which is often needed in terms of simulation time
[74] G.V. Stupakov, PRST-AB 8, 044401 (2005) and memory needs. In the important case of the
[75] Y. Shobuda, Y.H. Chin, K. Takata, PRST-AB 12, short-range coupling of ultra-relativistic bunches,
094401 (2009) however, specialized numerical algorithms exist
[76] K. Yokoya, CERN SL/90-88(AP) (1990) which make time domain calculations feasible for
[77] B. Podobedov, S. Krinsky, PRST-AB 9, 054401 most practical purposes. In the following, a brief
(2006) description of these algorithms is given.
[78] G.V. Stupakov, PRST-AB 10, 094401 (2007)
Space and time discretizations The starting
[79] B. Podobedov, S. Krinsky, PRST-AB 10, 074402
point is a set of semidiscrete (time-continuous-
(2007)
space-discrete) matrix equations of the form,
[80] T.-S.F. Wang, S.S. Kurennoy, R.L. Gluckstern,
PRST-AB 4, 104201 (2001)
d M e 0 CT e j
[81] R.L. Gluckstern, B. Zotter, PRST-AB 4, 024402 = · − , (2)
(2001) dt M μ h −C 0 h 0
[82] Los Alamos Accelerator Code Group, LA-UR-
90-1766 (1990) where e, h denote discrete electromagnetic de-
[83] A.V. Novokhatski, SLAC-PUB-11556 (2005); grees of freedom and j is the discrete representa-
www.slac.stanford.edu/ñovo/ tion of the bunch current on a computational grid.
[84] S.S. Kurennoy, EPAC 2002, 1526 For the sake of simplicity, conducting currents in-
[85] I. Zagorodnov, T. Weiland, PRST-AB 8, 042001 duced in accelerator structures of finite electrical
(2005) conductivity are neglected.
248
The matrices M , M μ and C in (2) repre- Dispersion free schemes The purpose of these
sent the discrete mass- and curl-operators, respec- schemes is to suppress numerical dispersion for
tively, as obtained by the application of an ap- plane wave solutions propagating in the longitu-
propriate discretization method. The traditional dinal direction. Two classes of dispersion free
method used for this purpose is the Finite Integra- schemes have been reported. The first includes
tion Technique (FIT) on Cartesian grid [2]. The methods which modify the time update (3), while
method provides a second order accurate and nu- leaving the discrete operators in (2) unchanged. In
merically highly efficient framework for the solu- [9], this is realized by introducing an approximate
tion of Maxwell’s equations in the time domain. splitting of the update operator G into longitu-
A number of major 3D-wakefield codes, includ- dinal and transversal updates. The Longitudinal-
ing MAFIA [3], CST PARTICLE STUDIO# R Transversal (LT) splitting reads

(CST PS) [3], GdfidL [4], ECHO [5] and PBCI Δt Δt

[6] are based on this formulation. Furthermore, G(Δt) ≈ GT · GL (Δt) · GT ,
wakefield calculations using Finite Element-type 2 2
(4)
Methods (FEM) of higher order accuracy on un-
where GT and GL contain only spatial deriva-
structured grids have been reported. This dis-
tives in the transversal and in the longitudinal di-
cretization approach is implemented in the code
rections, respectively. The LT-scheme is shown to
Tau3P [7] developed at SLAC.
be stable and longitudinally dispersion free at the
Equations (2) are further discretized in time “magic time step”, Δt = Δz/c. This is demon-
using an explicit (or implicit) time update scheme. strated in Fig.1, where numerical phase velocities
The explicit update can be generally written as in the zx-plane for different propagation angles
n+1 n n are shown. The LT-scheme is implemented in the
e e ΔtM −1 j 3D-wakefield codes PBCI and GdfidL.
= G(Δt) · − ,
h h 0 Another approach based on operator splitting
(3) is the TE/TM scheme [10, 11]. The explicit for-
where n and n + 1 are two subsequent time lev- mulation introduced in [11] uses the splitting,
els in the computation and G is a scheme-specific
G(Δt) ≈ GT E (Δt) · GT M (Δt), (5)
time update operator in (tn , tn+1 ] depending on
the discrete operators M , M μ and C. A numer- where GT E and GT M are (partial) time update
ically efficient and, thus, commonly used update operators affecting only TE- and TM-wave solu-
scheme is the leap-frog integrator [8]. This ap- tions, respectively. Also this scheme is stable and
proach is used in the wakefield codes MAFIA and longitudinally dispersion free at the “magic time
CST PS. step”. When a uniform Cartesian grid is used, the
It has been early recognized that, when the method is dispersion free along the two transver-
leap-frog integrator is applied, the maximum nu- sal directions as well. The TE/TM scheme is
merical dispersion error occurs for plane wave so- implemented for 2D- and 3D-wakefield calcula-
lutions propagating parallel to the grid axes (cf. tions in the code ECHO. In Fig.2, longitudinal
[9]). This property represents a drawback of the wake potentials in a TESLA cavity computed with
scheme, since the dominant wakefield propaga-
tion direction coincides with the main grid axis, (a) 90 (b) 90
120 60 120 60
z, which is commonly aligned with the bunch di-
rection of motion. The resulting numerical error is 150 30 150 30
comparatively large, especially, in calculations in-

180 0 180 0
volving short bunches. Furthermore, the scheme
imposes a stability condition which restricts the
210 330 210 330
time step below the longitudinal periodicity im-
plied by the grid, Δt < Δtz = Δz/c. A moving 240 300 240 300
270 270
window implementation (see below) based on a
leap-frog update is, thus, impracticable. The sit- Figure 1: Normalized phase velocities in the zx-plane
uation improves when so called dispersion free at the stability limit for the (a) leap-frog scheme and
schemes are applied. These are described in the (b) LT-splitting. Solid and dashed lines represent dif-
following. ferent grid resolutions.
249
W|| [V/pC] E/M-2.5D This behavior is observed even when approximate

0
( z = σ / 5) absorbing boundary conditions are applied. Dis-
( z = σ /10) persion free schemes at the “magic time step” are
-10
able to cope with the problem in a straightfor-
ward manner. No special boundary conditions
-20
TE/TM-3D are required for terminating the window. Grid
z=
x
=
y σ
=
and field solution can be simply shifted by Δz in
-30 POT-2.5D
3 3 2.5 every time step without loss of accuracy. Mov-
ing window implementations based on dispersion

-40 20 cells free schemes are realized in the 3D-codes ECHO,
-0.5 0 0.5 1 1.5 PBCI and GdfidL.
s [cm]
Boundary conformal methods Impedance cal-
Figure 2: Longitudinal wake potentials presented in culations in the time domain are mostly based on
[10] for an electron bunch with σz = 1 mm in a 20-cell a Cartesian grid discretization. This approach is
TESLA structure. The solutions obtained for different numerically efficient and it enables the applica-
grid resolutions using the leap-frog scheme (dashed tion of dispersion free schemes. However, the
lines) and the TE/TM splitting (solid lines), respec- numerical error resulting from the crude approx-
tively, are shown. imation of curved cavity boundaries (staircasing
error) may become large. The idea of bound-
ary conformal methods is to improve calculation
ECHO are shown. The immense advantage of the
accuracy by including additional information on
scheme with respect to numerical accuracy com-
the boundary geometry in the spatial discretiza-
pared to the classical leap-frog approach is clearly
tion step.
observed.
In a second class of dispersion free methods A well established boundary conformal
the spatial discretization represented by M , M μ method in the general context of electromag-
and C is constructed such that the numerical dis- netic field computation is the Partially Filled Cell
(PFC) approximation by Thoma-Weiland [14].
persion is minimum in the longitudinal direction,
The method is implemented for impedance calcu-
even when a leap-frog update is applied. A Finite
Volumes (FV) based discretization method using lations in the fixed window code CST PS, and in
this idea is reported in [12]. The method is stable the moving window code ECHO. The high accu-
racy of the method compared to the standard stair-
at the “magic time step” as well as dispersion free
case approximation is illustrated in Fig.3, which
along the three grid axes. The non-split, disper-
sion free FV-method is implemented in the code shows wake potentials for a rotationally symmet-
PBCI. ric collimator structure computed with ECHO-
2D.
Moving window When interested in the head-
tail interaction of ultra-relativistic bunches,
causality implies that the integration range in (1)
can be restricted within a small comoving frame
including the particle bunch. This idea was first
expressed in [13] and it has since then been used
in time domain calculations to reduce computa-
tional effort. The actual computational domain
consists of a small window which is shifted at the
speed of light in vacuum to account for the bunch
motion.
In order for this approach to work, how-
ever, the longitudinal numerical phase velocity
must exactly match the speed of light in vacuum. Figure 3: Longitudinal wake potentials for a collima-
Otherwise, numerical reflections at the window tor structure calculated with ECHO-2D. The numerical
boundaries occur which sooner or later pollute the convergence with respect to grid resolution in staircase
wakefield solution in the vicinity of the bunch. and conformal approximation, respectively, is shown.
250
While the application of a boundary confor- y'

outgoing pipe
mal method clearly improves accuracy, it also
increases the computational load. This is par-
ticularly the case for long accelerator structures, indirect path
where the operation count needed for extract- y R
ing detailed geometrical information may become direct path
z
prohibitively large. Another drawback of the Zp zof
method is the time step reduction resulting from
stability constraints. This makes the application Figure 4: Schematic of the indirect integration path.
of dispersion free schemes in this combination
difficult. A solution to this problem was found
in [15]. The Uniformly Stable Conformal (USC) range (−∞, Zp ], containing the accelerator struc-
method implemented in ECHO allows for a con- ture a (direct) numerical calculation is used. In
formal discretization without time step reduction. the outgoing pipe, for (Zp , ∞), a semianalytical
Resistive wall wakefields The numerical tech- procedure (indirect integration) is applied.
niques mentioned above refer primarily to the A well known indirect integration approach
wakefields due to accelerator chamber geome- consists in deforming the integration path such
try. In certain situations, however, the wake- that it may be continued to infinity along the con-
fields induced by the finite conductivity of the ducting wall of the pipe. The situation is illus-
chamber walls become important. The effect of trated in Fig.4. Using the notations therein, the
these so called resistive wakefields tends to be following relation holds in the general 3D-case:
dominant, in particular, in the “warm” accelera- ∞
tor sections of coherent light sources such as un- s+z
dzEz (x, y, z, )= (6)
dulators (cf. [16]). In order to handle resistive Zp c
wakefields numerically, a broadband frequency R
s + Zp
dependent impedance model for the electrical dy (Ey − cBx )T M x, y , Zp , .
y c
conductivity of lossy accelerator walls is neces-
sary, which makes the numerical implementation Thus, the improper wake integral can be trans-
of dispersion free schemes, moving window and formed into a proper one along a short path on
boundary conformal approaches extremely cum- the cross-section of the pipe. The relation (6) is
bersome. The calculation of resistive wakefields due to Henke and Bruns [18]. It represents the
is still possible in the fixed-window codes such generalization of the indirect integration method
as MAFIA and CST PS, although the applica- for rotationally symmetric cavity-like structures
tion of these tools is limited to comparatively long which was earlier introduced by Weiland [19] and
bunches and short accelerator structures. Moving then further developed by Napoly, Chin and Zot-
window and dispersive free implementations for ter in [20]. The indirect integration by path de-
resistive wakefields remain a challenge for future formation is implemented in the 2D-codes BCI,
developments. A recent work presenting such an ABCI and TBCI. The 3D-variant given by (6)
implementation for rotationally symmetric prob- is implemented in the codes GdfidL and ECHO.
lems can be found in [17]. Note the subscript T M appearing in (6) which in-
Indirect integration refers to the procedure dicates that, in the 3D-case, the TM-wave com-
used for calculating the wakefield contribution in ponent needs to be extracted from the full field
the potential (1) after the bunch has passed the solution in the beam pipe before the indirect inte-
accelerator structure. The distance needed for the gration is performed.
wakefields to “catch-up” with the bunch in an out- A second approach is based on the decom-
going beam pipe behind the structure scales as position of outgoing wakefields into waveguide
a2 /σ, where a is the pipe radius and σ the bunch modes. Given such a decomposition, the wake in-
length. This distance is, usually, several meters tegral in (Zp , ∞) can be performed analytically.
long which makes the computation of (1) based on While this approach does not represent an indi-
one of the previously described numerical meth- rect integration in the literal sense, it provides a
ods impossible. In order to deal with this problem, detailed information on the outgoing wakefields
the wake integral (1) is split into two parts. In the which can be used for purposes other than indirect
251
30 Direct potential (z = 150mm) [18] H. Henke, W. Bruns, EPAC 06, 2170

Transition potential
Steady state potential [19] T Weiland, NIM 216 (1983) 31
10 [20] O. Napoly et al, NIM A334 (1993) 255
[21] N. Watson et al, EPAC 06, 697
-10
Wz / [V / pC]
-30
-50 3.2.4 Explicit Expressions of Impedances and

-70
Wake Functions
K.Y. Ng, FNAL, K. Bane, SLAC
-90
-4 -3 -2 -1 0 1 2 3 4
See tables in the next pages.
s/V
Figure 5: Direct vs. indirect integration for a 0.3 mm-

bunch in the ILC collimator #8 [21]. “Transition po- References
tential” refers to the wakefield integral contribution in
the outgoing pipe. 150 waveguide modes were used in [1] A.W. Chao, Physics of Collective Beam Instabil-
the calculation. ities in High Energy Accelerators, Wiley (1993)
ch.2
[2] K.Y. Ng, Fermilab-FN-0756, 2004
integration (cf. [6]). The indirect integration by [3] K. Bane, M. Sands, Proc. Micro Bunches Work-
shop 1995, 131
modal decomposition is implemented in the code
[4] S. Krinsky et al, PRST-AB 7, 114401 (2004);
PBCI. Figure 5 illustrates the application of the
Proc. Micro Bunches Workshop 1995, 131
method in the calculation of the wake potential
[5] A. Piwiniski, DESY 94-068 (1994); DESY 84-
for one of the rectangular collimators used in the 097 (1984)
ILC-ESA test beam program [21]. [6] A. Piwiniski, PAC 77, 1364; X.E. Lin, SLAC-
PUB-7924 (1998)
References [7] L. Palumbo and V.G. Vaccaro, Nuovo Cimento A
89 (1985) 3
[1] W.K.H. Panofsky, W.A. Wenzel, RSI 27 (1956) [8] R. Gluckstern et al, PR E47 (1993) 656
967 [9] K. Yokoya, PA 41 (1993) 221
[2] T. Weiland, AEÜ 31 (1977) 116 [10] A. Lutman et al, PRST-AB 11, 074401 (2008)
[3] CST AG, http://www.cst.com. [11] K.Y. Ng, PA 16 (1984) 63
[4] W. Bruns, http://www.gdfidl.de [12] N. Mounet, E. Métral, CERN-BE-2009-039
[5] I. Zagorodnov et al, J.Comp.Phys. 191 (2003) (2009) For a Mathematica code to compute
525 multi-layer impedances, visit https://
[6] E. Gjonaj et al, ICAP 06 (2006) 29 impedance.web.cern.ch/impedance/Codes
[7] M. Wolf et al, ACES 02 (2002) /ReWall/Rewall to date.zip and https://
[8] K.S. Yee, IEEE Trans. Antennas and Propagation impedance.web.cern.ch/impedance
AP-14 (1966) 302 Note that we separate impedances into the usual
[9] T. Lau et al., FREQUENZ 59 (2005) 210 space-charge (self-field plus wall image, both
[10] I. Zagorodnov, T. Weiland, J.Comp.Phys. 207 → 0 as γ → ∞) and wall impedances, whereas
(2005) 69 Mounet separates them instead into self-direct
[11] M. Dohlus, I. Zagorodnov, J.Comp.Phys. 228 (due to beam only but excluding wall image) and
(2009) 2822 wall (our usual wall impedance plus wall image).
[12] E. Gjonaj et al., PAC 09 (2009) FR5PFP091 [13] B. Zotter, CERN-AB-2005-043 (2005)
[13] K. Bane, T. Weiland, HEACC 83 (1983) 314 [14] A. Burov, V. Lebedev, EPAC 02 (2002) 1402
[14] P. Thoma, Ph.D. dissertation D17 (1997) TH [15] G. Stupakov et al, PRST-AB 10, 054401 (2007);
Darmstadt K. Bane et al, PRST-AB 10, 074401 (2007)
[15] I. Zagorodnov et al, Int.J.Num.Mod. 16 (2003) [16] S.A. Heifets, PR D40 (1989) 3097; S.A. Heifets,
127 S.A. Kheifets, Rev. Mod. Phys. 63 (1990) 631
[16] K.L.F. Bane, G. Stupakov, PAC 05, 3390 [17] E. Gianfelice, L. Palumbo, IEEE Tr. NS 37, 2
[17] A. Tsakanian, M. Dohlus, I. Zagorodnov, PAC (1990) 1081
09, 3342 [18] I. Zagorodnov, K. Bane, EPAC 06, 2859
252
General Remarks and Notation:

In cylindrically symmetric structures Wm (z) and W (z) denote, respectively, m-th azimuthal
m
multipole longitudinal and transverse wake functions, generated by point charge Q, at distance −z >
0 behind. Wm (z) ≡ dW (z)/dz. W (z) = 0 and W (z) = 0 when z > 0 when particle travels at the
m m m
speed of light. Wm (0) = 1 lim W (z). Longitudinal and transverse momentum kicks on test charge
2z→0− m
q near pipe axis: Δp (z) = −qQW0 (z)/c, Δp⊥ (z) = −qQr̃⊥ W1 (z)/c, where r̃⊥ is (small) offset of
the source or exciting charge.


The m-th multipole longitudinal impedance Zm (k)= e−ikz/β Wm (z) dz/(βc) is related to

⊥ (k)= i e−ikz/β W ⊥ (z) dz/(β 2 c), by Z = kZ ⊥
the m-th multipole transverse impedance, Zm m m m
∗ ⊥ (−k) = −Z ⊥ ∗ (k).
(m = 0), where k = ω/c. Note that Zm (−k) = Zm (k), Zm m
For periodic or translationally invariant structures: steady-state results are given per length L.
Unless otherwise stated, structures are cylindrically symmetric with perfectly conducting metallic
walls, and with beam pipes of radius b. In many cases, β = v/c has been set to 1. Z0 = μ0 /0 ≈
377 Ω is impedance, 0 electric permittivity, and μ0 magnetic permeability of free space. ‘Pill-box
cavity’ signifies a pill-box with beam pipes. Here [α±i|β|]1/n (with α, β real, n = 2 or 3) is in the
1st/4th quadrant. H(x) = 0, 1 for x < > 0.
For 3D structures with mirror symmetry in x and y, near axis momentum kick in y, Δpy =
−qQ(ỹWdy + yWqy ), with ỹ (y) offset of exciting (test) charge, and Wdy (Wqy ) dipole (quad) wake
terms. Total y wake Wy = Wdy + Wqy ; total y impedance Zy = Zdy + Zqy .
Description Impedances wake
Space-charge: [1]
Z0 Z0 kg0 b W0 Z0 c b
beam radius a in a =i g0 = 1 + 2 ln = 1 + 2 ln δ (z)
perfectly conducting L 4πβ 2 γ 2 a L 4πγ 2 a

beam pipe of radius b, Z ⊥ Z0 1 1 Wm=0 Z0 c 1 1
transverse distribution
m=0
= i − = − δ(z)
L 2πβ 2 γ 2 m a2m b2m L 2πγ 2 m a2m b2m
uniform.

Nonuniform distri-

Distribution λ(r) g0 (m = 0) aeff (m = 0) a⊥ eff (m = 1)
butions: [2] aeff is 1/2
3 r 2 8 b 2
equivalent-uniform- 1− 2 H(r̂−r) + 2 ln 0.8692r̂ r̂
beam radius, g0 = 2πr̂ r̂ 3 2r̂ 3

1 + 2 ln(b/aeff ), while 1 r2 3 b 1
1− 2 H(r̂−r) + 2 ln 0.7788r̂ √ r̂
⊥
aeff = [πλ(0)] −1/2 is 2πr̂ 2 r̂ 2 r̂ 2
√
the same when self- 2π πr b π 2 −4
force part written as cos2 H(r̂−r) 1.921+2 ln 0.6309r̂ √ r̂
⊥ π 2 −4 2r̂ r̂ 2π
1/aeff , γe ≈ 0.57721
2
1 −r2 /(2σr2 ) b √
is Euler’s constant. e γe +2 ln √ 1.7647σr 2σr
2πσr2 2σr
Image part of Z1⊥ can be written in terms of Laslett’s electric image
coefficients as 1/b2 → 2(ξ1x,y − 1x,y )/h2 with h denoting half height of
vacuum chamber. See Sec.2.4.5.
Resistive wall: [1, 3] Zm

Z0 /(πb2m+1 )
relative magnetic
wall thickness t, dc = , μr =
L permeability
and ac conductivities iZ σ̃
0 c ibk
(1+δm0 ) − σ̃c = σc /(1−ikcτ )
σc , σ̃c , relaxation time kμr m+1
τ ; assume |k|b
Typically, τ = 27/40/8 fs for Cu/Ag/Al. Valid for cτ /s0 1, characteristic
(s0 /b)3 , thick walls:
distance s0 = [2b2 μr /(Z 1/3
0 σc )] , α# = [(m+1)(1+δ
$ √ m0 )/2]2/3 .
t δc = √ ∞
Wm 4Z0 c(m+1) e αz/s 0 3αz 2
2
x2 eαzx /s0
2/(|k|Z0 μr σc ), the = cos − dx
skin depth. L πb 2m+2 3 s0 π 0 x6 + 8
253
Description Impedances Wakes

Low frequency: [1] Zm 1 − sgn(k)i Wm −c Z0 μr 1
k 1/s0 , long range = =
L (1+ δ0m )πσc δc b2m+1 L 2πb2m+1 (1+ δm0 ) πσc (−z)3/2
|z| s0 .
2 2
Note : Z1⊥ = 2 Z0 , W1 = 2 W0 dz
b k b
⊥
Low frequency, thin Z0
=−
Z0 kt
,
Z1 Z0 t
= −i 3 W0 Z0 tc W1 Z0 ct
wall: [1]√t δc and L 2πb L πb =− δ (z), =− 3 δ(z)
L 2πb L πb
|k| 1/ bt.
Wm Z0 c(m + 1) z/4cτ
= e
High frequency: [3] Zm 4Z0 cτ (m + 1)
= L πb2m+2
k> πb2m+1
∼ 1/s0 , short range L 2kp
|z| < 0 , with cτ ∼ s0 .
∼ s > 1 − 4ikcτ × cos αz , for α see above
× b
kp = Z0 σc /cτ is b(1 − 4ikcτ )2 + 32kp (αcτ )2
plasma frequency/c.
These formulae depend only on the plasma frequency of the metal. Ef-
fects of relative magnetic permeability have
not been considered.
Finite length, lossy 1/2 Z
2 0 iL
insert: [4] of length L, If b2 /s30 k2 L kb2 , Z0 = 3/2 , else Z0 as given above
in lossless pipe π b k

Displaced beam: [5] Wall impedances in last section multiplied by fz for Z0 and fx,y for Z1x,y
at a = (ax , ay ), rms b2 +a2 b(b2 −a2 +4a2x ) b(b2 −a2 +4a2y )
bunch length σ , aver- with fz = 2 2 , fx = , fy = Power
age current Ib , and b −a (b −a )
2 2 3 (b2 −a2 )3
3 2
(b/k 2 , b, b − a) δc P Γ( 4 )Ib
loss per length traversed is = fz
and γ 1. L 4π 2 bσ
3/2
2μr σc /Z0

Displaced beam 1−sgn(ω)i |ω|μr Z0 π(sgn(ω)1−i)

between two infi- Z0 = fz , Z1⊥ = f⊥
nite plates: [5] at πh 2cσc 2|ω|σc /(cμr Z0 )
y = ±h/2. γ 1, πy0 πy0 fz
fz = 1+ tan , f⊥ = 3 , beam at y = y0
[h/k2 , h−2y0 ] δc . h h h cos2 (πy0 /h)
Thin dielectric coating iωZ0 (r μr −1)Δh iπZ0 (r μr −1)Δh
Z0 = − fz , Z1⊥ = − f⊥
of thickness Δh. πcr h r

Metallic coating A+tanh(νtm ) 1 1−sgn(ω)i
on ceramic pipe: [6] Z0 = Z0 (met) , A = 1− νtc , ν =
1+A tanh(νtm ) r δc
compared with all P Z0 Ib2 tc (r −1) √ 2 V2 r σ
= √ V − πV e erfc(V ) , V =
metal pipe Z0 (met). L 4 πbσ2 r (r −1)Z0 σc tm tc
tm,c = metal/ceramic Ez,out 1
thickness b. γ 1, Field penetration through pipe, = ,
Ez,in 1+4(1−1/r )tm tc /δc2
[(r−1)t2c , (1−−1
r )btc ] <
σ2 . Loss P/L is becomes significant when tm ∼ tcrit = δc2 /tc . P/L is at maximum at tcrit .
max. at V = 0.82.
Elliptical beam pipe: Low frequency, see [7, 8, 5], high frequency, see [9, 10].
Rectangular beam Low frequency, see [8], high frequency, see [9, 11].
pipe:
254
Multi-layer pipe wall impedances: [12, 13] Cylindrical beam pipe with N layers, pth layer between
b(p−1) <r < b(p) and b(N ) → ∞. Layer 1 is vacuum, a < r < b(1) , with particle beam of charge Q at
r = a and θ = 0. r < a is called the 0-th layer. Each layer has its own wavenumber ν = k 1−β 2 ε1 μ1 ,
k = ω/v and own properties = 0 ε1 = 0 r (1 + i tan ϑE ) − iω(1−iωτ σdc
) , μ = μ0 μ1 = μ0 μr (1 +
i tan ϑM ); ϑE , ϑM are loss angles, σdc dc conductivity, and τ relaxation time. Actually any frequency
dependent , μ, and conductivity can be assumed. Inside vacuum, ν = k/γ; inside conducting metal
of skin depth δc , ν ≈ (1−i)/δc . A user-friendly Mathematica code for computation is available [12].
The derivation is outlined briefly below. In terms of Bessel and Kelvin functions, mth multipole
longitudinal fields inside p-th layer:
(p) (p)
Es(p) = cos mθ eiks CIe Im ν (p) r +CKe Km ν (p) r , E is electric field

(p) (p) = Z0 H, H is magnetic field

Gs(p) = sin mθ eiks CIg Im ν (p) r +CKg Km ν (p) r , G
(p)
Matching
⎡ ⎤Es , Eθ , G
⎡s , and⎤Gθ at boundary
⎡ ⎤r = b ⎡between ⎤ p-th and (p+1)-th layers gives
C
(p+1)
C
(p)
C
(N )
C
(1) ⎧
⎢ Ie N −1
(p+1) ⎥ ⎢ Ie (p) ⎥ ⎢ Ie(N ) ⎥ ⎢ Ie(1) ⎥ ⎪
⎨M ≡ MN −1 MN −2 · · · M1
N 2
⎢CKe ⎥ ⎢ C ⎥ ⎢ C ⎥ ⎢ C ⎥
⎢ (p+1) ⎥ = Mp ⎢ (p) ⎥ iteratively ⎢ (N ) ⎥ = M⎢ (1) ⎥ where See [12] for explicit
p+1 Ke Ke Ke
⎢C ⎥ ⎢C ⎥ =⇒ ⎢C ⎥ ⎢C ⎥ ⎪
⎣ Ig ⎦ ⎣ Ig ⎦ ⎣ Ig ⎦ ⎣ Ig ⎦ ⎩expression of M p+1
(p+1) (p) (N ) (1) p
CKg CKg CKg CKg
(N ) (N ) (1)
Since the last layer goes to infinity, CIe = CIg = 0. From the beam region, CKg = 0 and
(1)
CKe = −ikQZ0 Im (ka/γ)/[πβγ 2 (1+δm0 )], one can easily solve for
(1)
(1)
Km (1) (1) M12 M33 −M32 M13 (1) (1)
CIe ≡−α1 (1) CKe = −CKe , withIm = Im ν (1) b(1) , Im = Im ν (1) b(1) .
Im M11 M 33 −M 13 M 31
at r = a1 , θ = 0, reduced forces on a unit test charge at r = a2 > a1 and θ = θ2 are
With beam
Z = − ds Es (a2 , θ2 , s; ω)e−iks , Zx = −i ds [Ex (a2 , θ2 , s; ω)−βGy (a2 , θ2 , s; ω)] e−iks .
Space-charge contributions for all multiples (α1 = 1 or perfectly
conducting at r =b(1) ):
∞ (1)
ikZ0 L cos mθ2 Km kai
ZSC = 2
Im(x2 )Km (x1 ), Km (xi ) = Km (xi ) − (1) Im (xi ) , xi =
πβγ (1+δm0 ) I m
γ
m=0
∞
ikZ0 L mγ
ZxSC = Im (x1 ) cos θ2 cos mθ2 Km (x2 )+ sin θ2 sin mθ2 Km (x2 )
πβγ 3 (1+δm0 ) a2
m=0
The rest are from wall impedances. To any order an1 1 an2 2 , they are

iLμ0 ω ka1 n1 ka2 n2
(1) (1)
W,n1 ,n2 cos mθ2 ᾱ1 (m)Km /Im
Z =− 2 2
πβ γ 2γ 2γ (1+ δm0 )( 1 2 )!( 1 2 )!( n2 −m
n −m n +m n2 +m
2 )!( 2 )!
n1 n2 (1) (1)
iZ0 L ka1 ka2 (n2 cos θ2 cos mθ2 +m sin θ2 sin mθ2 )ᾱ1 (m)Km /Im
ZxW,n1,n2=− 2 n −m n +m n −m n +m
πβγ a2 2γ 2γ 1 1
(1+ δm0 ) ( 2 )!( 2 )!( 2 )!( 2 )! 2 2

where ᾱ1 (m) ≡ 1−α1 (m), implies from m = 0 to min(n1 , n2 ) with n1 −n2 and n1 −m even.
The usual monopole and dipole pipe-wall impedances are
(1) (1)
ikZ0 L ᾱ1 K0 ⊥ ZxW,1,1 iLZ0 k2 ᾱ1 K1
Z0 = ZW,0,0 = and Z 1 = =
2πβγ 2 I (1) a1 4πβγ 4 I (1)
0 1
Multi-layer special cases: [13] Pipe wall: b(1) < r < b(2) = b(1) +t.
Thin wall: Good for low frequencies. t → 0 and Es does not change across wall. At r = b(3) , Case
PC: perfectly conducting, Case PM: perfectly magnetic, and Case INF: b(2) → ∞.
# $2
γ 2 β 2 (1−α2 )+2ixγβ/mζ Km (y)Im (x) b(1)
ᾱ1 = − , α2 = −α2 =
PC PM
≈ , αINF
2 =0
2
1+ x 2 − ixγβ 2
+ ζ(1−α2 ) Km (x)Im (y) m= 0 b(3)
m m ζ(1−α2 ) 2
x = kb(1) /γ, y = kb(3) /γ, ζ = Z0 σc t, and m = 0.
255
Thick wall: Good for high frequencies. At r = b(2) , Case PC: perfectly conducting, Case PM:
INF: b → ∞. For m ≥ 1,
perfectly magnetic, and Case (2)
(1+i)ΔQη
−2β γ 1− 2mβγ 2
2 2
k2 δ2
ᾱ1 = , p= , Δ = μ1 βkδc ,
(1−i)Qα (1+i)ΔQη Qα Qη −m2 p2 2
1−2ip−β mΔ − 2m + m2 γ 2
Q2 −α2 P2 Q2 −η2 P2 K (2) I (2)

Qα = kb(1) , Qη = kb(1) , Q2 = m(2) , P2 = m(2)
1−α2 1−η2 Km Im
(2,3) (2)
Km Im Km (2,3) I (2)
m
Boundary conditions require αPC PM

2 = η2 = (2,3) (2)
, η2
PC
= αPM
2 = ,
(2,3) (2)
Im Km Im Km
(p+1,p) (p) ,
αINF INF
2 = η2 = 0, with Im = Im ν (p+1) b(p) , Im = Im ν (p) b(p) and similar definitions for Im
Km , and Km .
Electric- and magnetic-dipole approximation: ᾱ1 can also be derived [14] by approximating beam
dipole motion as a superposition of oscillating electric and magnetic dipoles.
High frequency optical model: [15] High frequency k 1/h, short-range −z h, transition
length L kh2 , h is minimum aperture. For tapered transition of angle θ, need k 1/hθ.
Transitions, shallow Z and kZ ⊥ are both constants W = −Z cδ(z)
cavities, collimators, similar for kZd⊥ , kZq⊥ , Wd⊥ , Wq⊥
irises: W⊥ = −kZ ⊥ cH(−z)

(a) Axially symmet- Z0 d Z0 1 1
(i) Z0 = Z1⊥ = 0, (ii) Z0 = ln , kZ1⊥ = −
ric examples: [16]- π b π b2 d2
[18] (i) step-in transi-
Z0 d Z0 1 b2
tion (from d to b), (ii) (iii) Z0 = ln , kZ1⊥ = −
step-out (from b to d), π b 2π b2 d4
long collimator, shal- where b is small iris or pipe radius, d is large pipe radius. Note: for shallow
low cavity with gap g, cavity, waves reflect from outer wall ⇒ g > ∼ k(d − b) ; for collimator,
2
(iii) thin iris bottom length kb2
(b) 3D, mirror sym-
metric in x, y: [15] (i) π 1 1 1 1
(i) kZy = Z0 2 − 2 , Zqy = Zdy = Zy
flat step-out transition, 8 b d 2 3
aperture 2b to 2d, (ii) Z0 1
any step-in transition; (ii) Z = Z⊥ = 0, (iii) kZy = , Zqy = Zdy = Zy
2πb2 2
iris with small (iii) flat
(height 2b), (iv) ellipti- Z0 y Z0 b2 y Z0 b2
(iv) kZy = , kZ d = 1+ , kZ q = 1−
cal (axes w by b), aper- 2πb2 4πb2 w2 4πb2 w2
ture
High frequency √ √
diffraction formu- 2Z0 ig 2Z0 c g
Zm = Wm =
lae: k 1/b (1+δm0 )π 3/2 b2m+1 k (1+δm0 )π 2 b2m+1 −z
(a) Deep cavity (Fres- 2
Z1⊥ = 2 Z0 2
nel diffraction)
b k W1 = 2 W0 dz
[19, 1], cavity radius d b
Note: no reflections from outer
and gap g.
wall ⇒ g <∼ k(d − b) .
2
256
(b) Periodic array of −1

deep cavities (model W0 Z0 c 2
Z0 iZ0 α(g/L)L 2πi = 2 eη(z) erfc η(z)
for linear accelerator = 1+ L πb
L πkb2 b kg
structures): [20]-[24]
αL −2πz
period L, gap g, outer η(z) =
α(ζ) ≈ 1 − 0.465 ζ − 0.070ζ b g

cavity radius d, with

2
g<∼ k(d − b) . W1 = 2 W0 dz
2
2
Z1⊥ = Z b
b2 k 0
Numerical fit:[24, 25]

valid over larger z Z0 c z 4Z0 cz1 z z
W0 = 2 exp − , W1 = 1 − 1+ exp −
range: −z/L ≤ 0.15, πb z0 πb4 z1 z1
0.34 ≤ b/L ≤ 0.69,
0.54 ≤ g/L ≤ 0.89. b1.8 g1.6 b1.79 g0.38
z0 = −0.41 2.4 , z1 = −0.17
L L1.17
Bethe’s dipole Electric and magnetic dipole 20 3 4 3
d= − a E, m =− a B
moments of a hole of moments when wavelength a: 3 3μ0
radius a on beam pipe and B
E are electric and magnetic flux density at hole when hole is
wall [26]. absent. This is a diffraction solution for a thin-wall pipe.
Small 3D obstacle on

Z0 = −ikcL, W0 = −c2 L δ (z)

beam pipe: [27, 28] 4 4
size b, low freq. Z1⊥ = 2 Z0 cos φ W1 = 2 W0 cos φ dz
b k b
k 1/(size); φ a-
Z0 (αe + αm )
zimuthal angular posi- Inductance L =
tion of object. 4π 2 b2 c
αe is electric polarizability, αm magnetic susceptibility
⎧ 3 2 ⎧ 4
Elliptical hole: major ⎪ πa m1 [K(m)−E(m))] ⎪ πd [ln(4a/d)−1] beam
⎪
⎪ ⎪
⎪
and minor radii are a ⎨3E(m)[E(m)−m1 K(m)]m→1 ⎨ 3a db
and d. K(m) and αe + αm = =⇒
⎪
⎪ 3 long ⎪ 3
E(m) are complete ⎪πa [E(m)−m1 K(m)] ellipse⎪
⎩ ⎪
⎩
πa ⊥ beam
elliptical functions of 3[K(m)−E(m)] 3 [ln(4a/d)−1] a b
the first and second 3
kind, with m = 1−m1 αe + αm −→ 2a
circular
circular hole a = d b
and m→0 3
m1 = (d/a)2 . For Above are for t a. When t ≥ a, ×0.56 when hole is circular and
long ellipse perpen- ×0.59 when hole is long-elliptic.
dicular to beam, ma- For higher frequency correction, add to αe + αm the extra term,
⎧
jor axis a b, beam ⎪
⎪ πad2 k2 a2 beam
pipe radius, because ⎪−
⎨ 3
2πa3 11k2 a2 5 long ellipse
the curvature of the + circular,
⎪
3 30 ⎪ 2πa3 2k2 a2
beam pipe has been ⎪
⎩+ ⊥ beam
neglected here [29]. 3 5[ln(4a/d) − 1] long ellipse
Rectangular slot: αe + αm = w3 (0.1814 − 0.0344w/L) t a, ×0.59 when t ≥ a
length L, width w.
Rounded-end slot:
length L, width w. αe + αm = w3 (0.1334 − 0.0500w/L) t a, ×0.59 when t ≥ a
257
Description Impedances wake

Annular-ring-shaped
cut: inner and outer π 2 d2 a π 2 w2 (a + d)
αe + αm = − td
radii a and d = a + w 2 ln(32d/w) − 4 16
with w d. αe + αm = πd2 w − 12 w2 (a + d) t ≥ d

Half ellipsoidal pro- 1 1
αe + αm = 2πahd +
Ic − 3
trusion with semi I

b
axes h radially, a h2 2
Ib =2F1 1, 1; 2 ; 1− a2 , Ic =2F1 1, 2 ; 2 ; 1− ha2 , if a = d
5 1 5
longitudinally, and d
azimuthally. 2 F1 is 2πh3
αe + αm = πa3 if a = d h
if a = d = h ,
the hypergeometric 3[ln(2h/a) − 1]
function.
8h3 4 π a
αe + αm = 1+ − if a h = d
3 π 4 h

8πh4 2a
αe + αm = ln −1 if a h = d
3a h

2 2
Small inductive Z0 = −ikcL, Z1⊥ = Z0 W0 = −c2 Lδ (z), W1 = 2 W0 dz
objects-2D: [27, 30] b2 k b
small cavities, shal- 2
Z0 g
low irises, and transi- Pill box g <∼ h b : L = 2πcb gh − 2π
tions at low freq.
(h b, k 1/h); Z0 h2
Shallow iris g < ∼ h b : L = 4cb
h is height of object,
g is gap of cavity Z0 h2 2πb 1
Transition pair g h, h b : L = 2 ln +
or length of iris; L
2π cb

h 2

is inductance. For Z0 h2 bθ 3 θ π π
tapered transition Tapered :L = 2 ln −2θ cot θ + −γe −ψ − cot θ−
π cb h 2 π 2 2θ
pair: θ is taper angle.
γe ≈ 0.57721 is Euler’s constant, ψ(x) is psi function.

2ikZ0 ∞
Wall roughness 1-D: Z0 = − κ|h̃(κ)|2 dκ
b 0
inductive model: [35] ∞
1
1-D axisymmetric with spectrum h̃(k) = h(z)e−ikz dz
2π −∞
bump on beam pipe, ∞ ∞
h(z) or 2-D bump 4ikZ0 κ2
2-D: Z0 = − |h̃m (κ)|2 dκ
h(z, θ). Valid for low b m=−∞ −∞ κ + m2 /b2 2
frequency k (bump 2π ∞
length or width)−1 , 1
with spectrum h̃m (k) = dθ dz h(z, θ)e−ikz−imθ
h b, and |∇h| 1. (2π)2 0 −∞
See also [36] Note: small periodic corrugations model is also used for wall roughness
impedance estimation.
Small periodic

corrugations: (a) [31, Z0 Z0 k W0 Z0 c
= 2 πkr δ(k2 − kr2 ) + i ℘ 2 , = 2 cos kr z
32] L < ∼ h b, L πb k − kr2 L πb
k 1/h; L period, h
2 2L 4hg 2
depth, g gap, ℘ prin- Z1⊥ = Z , kr = ; (1 − βg ) = , W1 = 2 W0 dz
cipal value; βg c group b2 k 0 bgh bL b
velocity.
(b) [33] L h, L b,
Z0 Z0 h2 kL
3/2 W0 Z0 ch2 kL3 1
= (−ik)1/2 =−
k 1/h; kL = 2π/L. L 8πb L 16π 3/2 b (−kL z)3/2
258

Thin dielectric or fer- 1/2
2r
rite layer on pipe: Like small periodic corrugations (a), but kr = , with
(r μr − 1)bh
[34] thickness h b. relative dielectric constant r and magnetic permeability μr .
Coherent syn-
1/3 W0 Z0 c 1
chrotron radiation Z0 Z0 2 ik =−
(CSR): [37, 38, 39] = Γ L 2 · 34/3 πR2/3 z 4/3
2 · 31/3 π R2
L 3
Bunch moves in free
space on a circle of Γ(2/3) ≈ 1.3541. Note: non-zero wake for test particle ahead of driv-
radius R; k γ 3 /R. ing particle. W0 (0+ )/L ≈ 0.1Z0 cγ 4 /R2 . This is also used to approx-
See Sec.2.4.15. imate effect at high k for beam in beam pipe; shielded (suppressed) for
k< 1/2 −3/2 .
∼R b
Round collimator:
Z0 Z0 c
(a) [40] low frequen- Z1⊥ = −0.3i W1 = −0.3 δ(z)
cy k 1/d. d d
collimator radius d b. collimator radius d b.
(b) High frequency
k 1/d; if tapered, See optical model formulae (a) above
angle θ 1/(kd).
(c) [41] For any fre-
−iZ0 k Z0 c
quency, small angle, Z0 = ds (d )2 W0 = ds (d )2 δ (z)
d (s) 1, kdd 4π
2 4π
−iZ0 d 2
1, with d(s) pipe pro- ⊥
Z1 = ds Z0 c d
file versus longitudinal 2π d W1 = − ds δ(z)
2π d
position s, and d is ⇒ symm. tapers of angle θ 1 :
derivative of d with re-
−iZ0 1 1 Z0 c 1 1
spect to s. Z1⊥ = θ − W1 = − θ − δ(z)
π d b π d b
Flat collimator: [42]
low frequency, small
angle, h (s) 1,
−iZ0 w (h )2
h w , with h(s) Zy = ds Z0 cw (h )2
4 h3 Wy = − ds δ(z)
vertical profile, w 4 h3
width, length
Z0 g Z0 cg
Pill-box cavity —low Z0 = −ik ln S W0 = − ln S δ (z)
frequency: [43] cavity 2π 2π
Z0 g S − 1
2
Z0 cg S 2 − 1
radius d, gap g; S = Z1⊥ = −i 2 2 W1 = − 2 2 δ(z)
d/b. When g 2(d − πb S + 1 πb S + 1
b), replace g by d − b.
Valid for k 1/d. Effect will be one half for a step in the beam pipe from radius b to radius
d, or vice versa, when g 2(d − b).
Resonator model: (m) (m)
[1] for m-th azimuthal Rs Rs ckr αz
Zm = Wm = e sin k̄r z
mode, with shunt im- 1 + iQ (kr /k − k/kr ) Qk̄r
(m) (m)
pedance Rs , ⊥ Rs /k where α = kr /(2Q)
quality factor Q, and Zm =
1 + iQ (kr /k − k/kr ) k̄r = |kr2 − α2 |
resonant frequency
kr .
Valid only close to kr . As k → ∞, Z0 → k−1/2 for non-periodic
cavities and → k −3/2 for an infinite array of cavities. [16, 46]
259

Closed pill-box 2 x2mn p2 π 2
cavity: [44] res- kmnp
+ 2 =
g d2
onant frequencies ⎧
⎪
⎪ sin2 gk0np × 1
8 ⎨
kmnp and “circuit” p even
(Rs /Q)mnp [45], Rs Z0 2β 1 + δ0p
= 2 2
where m, n, p, are Q 0np x0n J0 (x0n ) πgk0np ⎪
⎪ cos2 gk0np
⎩ p odd
azimuthal, radial, 2β
longitudinal mode ⎧
⎪ gk1np
numbers. Cavity
Rs Z0 2 ⎨ sin2 p = 0 and even
2β
radius d and length
= 2
g; xmn is nth zero of
2
Q 1np J1 (x1n ) πgd2 k1np ⎪
⎩ cos2 gk1np p odd
Bessel function Jm . 2β
Curvature impe- Valid from zero

frequency up to just below synchronous resonant modes,
dance: [47] Smooth i.e., 0 < ν < R/h with ν = kh,
toroidal chamber '
of rectangular cross ikZ0 h2 ν 2
−2π(b−R)/h −2π(R−a)/h
section, width b − a, Z0 = 1− e −e 1−3
π2R π
height h, inner radius ;
a, outer radius b, ν 2
and R = 12 (a + b). +0.05179 − 0.01355 + ρkR
π
As Lorentz factor ν 2
γ → ∞, a contribu- ikZ0 h2
≈ A − 3B .
tion remains. π2R π
where ρ is quadratic in ν. As (b−a)/h increases, ρ vanishes exponen-

tially and A ≈ B ≈ 1. In√general, A/B ≈ 1 implying ImZ0 changes
sign (a node) near ν = π/ 3.
Kicker with window- k2 c2 μ20 L2 x20 c3 μ20 L2 x20

Z0 = W0 = − δ (z)
frame magnet: [49] 4a2 Zk 4a2 Zk 0
width a, height b, kc2 μ20 L2 c3 μ2 L2
length L, beam offset Z1⊥ = W1 = − 20 δ (z)
4a2 Zk 4a Zk
x0 horizontally, and
all image current Zk = −ikcL + Zg with L ≈ μ0 bL/a the inductance of the windings
carried by conducting and Zg the impedance of the generator and the cable. If the kicker is of

current plates. C-type magnet, x0 in Z0 should be replaced by (x0 + b).
Traveling-wave
kicker [49] with char- Zc θ Zc c L
Z0 = 2 sin2 −i sin θ W0 = δ(z)−δ z+
acteristic impedance 4 2 4 βph
Zc for the cable, and

a window magnet of Zc L 1−cos θ sin θ Zc βc L
width a, height b, Z1⊥ = −i W1 = H(z) −H z+
4ab θ θ 4ab βph
and length L. Valid
for frequency below
cutoff. θ = kL/βph denotes the electrical length of the kicker windings and
βph c = Zc ac/(Z0 b) is the matched transmission-line phase velocity of
the capacitance-loaded windings. Here, βph β → 1, the beam velocity.
260

2 2
Strip-line BPMs φ0 φ0
Z0 = 2Zc 2 sin2 kL−i sin2kL W0 = 2Zc c δ(z) − δ(z+2L)
(pair): [48] length L, 2π 2π
angle each subtending

⊥ Z0 1 4 2 2 φ0 8Zc c φ0
to pipe axis φ0 , form- Z1 = 2 φ
sin W1= 2 2 sin2 H(z)−H(z+2L)
ing transmission lines k b 0 2 π b 2
pair
of characteristic impe-
dance Zc with pipe. The strip-lines are assumed to terminate with impedance Zc at the up-
stream end.
Wakes for a Gaussian Bunch:
The bunch wakes of a bunch with longitudinal charge distribution λz , are given by Wm (z) =
0
(x)λ (z−x) dx, W (z) = 0 W (x)λ (z−x) dx. In the following we give bunch wakes
−∞ W m z m −∞√ m z
of a Gaussian bunch [λz = e−(z/σz ) /2 /( 2πσz ), with σz the rms bunch
2
∞ length] for wakefield forms
found in the tables above, and also give their first moments W = −∞ W(z)λz (z)dz and the rms

Wrms = W 2 − W2 . Here the z dependence alone is considered and the wake coefficient is
scaled out; for a specific problem, the appropriate coefficients, found in the tables above, need to be
included at the end.
Note: for power law wakes with −2 < α < −1, W is obtained using ∞ integration by parts [38]. It is
assumed that in the range |z| σz the wake form changes so that −∞ W (z) dz = 0. Consequently,
W can be obtained without knowing the details of W at very short range.
Wake form, W Bunch wake, W W Wrms
1 1 0.111
e−(z/σz ) /2
2
Circuit Models: √ √
2πσz 2 πσz σz
Resistive: δ(z)
z 1
e−(z/σz ) /2
2
√ 0 √
Inductive: δ (−z) 2πσ 3 6π 31/4 σz2
z
1 −z 1 1
Capacitive: H(−z) 1 + erf √ √
2 2σz 2 2 3
Power Law: (−z)α f (−z/σz )σzα , with f (x) given by
(upper/lower sign for x > < 0):
Low freq. resistive wall
(Wm ) and Fresnel dif- π|x| −x2/4 (( 0.723 0.292
): α = − 1 e I−1/4 ±I1/4 ( 2 √ √
fraction (Wm 2 8 x /4 σz σz
x
Fresnel diffraction (Wm ): π (( √ √
|y|1/2 e−y /4 I−1/4 ± I1/4 ( 2 dy
2
α = 12 0.489 σz 0.374 σz
32 −∞ y /4
Low freq. resistive wall
(Wm ) and small periodic
π|x|3 −x2/4 (( −0.489 0.516

corrugations (W0 ): [50] e I 1/4 −I −3/4 ±I −1/4 ∓I 3/4 ( 2 3/2 3/2
8 x /4 σz σz
α = − 32
∞
(x + y)e−(x+y) /2
2
CSR (W0 ): z α with 3 −0.758 0.532
−√ dy
α = − 3 (note: z > 0)
4
2π 0 y 1/3 4/3
σz
4/3
σz
1
W = f (−z/σz ) , with f (x) = e−(kr −αr )σz /2−αr σz x
2 2 2
Resonator
% Model:
& % & % 2 &
sin(−kr z) αr z Im ikr σz (x−αr σz ) (ikr − αr )σz + x
cos(kr z)
e , with × e 1 + erf √
Re 2
% &
kr , αr , real 1 −(kr2 −α2r )σz2 Im −i2kr αr σz2
W = e e 1 + erf (ikr −αr )σz

2 Re
261
[19] J. Lawson, Rutherford Report RHEL/M 144 3.2.5 Effective Impedance [1, 2, 3]
(1968); K. Bane, M. Sands, SLAC-PUB-4441 T. Suzuki, KEK
(1987)
[20] R. Gluckstern, PR D39 (1989) 2773, 2780 The effective impedance is used to estimate com-
[21] G. Stupakov, PAC 95, 3303 plex coherent tune shifts approximately. Analytic
[22] K. Yokoya, K. Bane, PAC 99, 1725 sums are found for a sinusoidal mode [4] and ap-
[23] A. Fedetov, R. Gluckstern, M. Venturini, PRST- proximately for a Gaussian bunch [5, 6] in the
AB 2 (1999) 064401 case of a resonator impedance.
[24] K. Bane et al, ICAP 98, 137 Definition longitudinal

[25] K. Bane, SLAC-PUB (2003) 9663

∞ Z0 (ωp )
[26] M. Sands, SLAC note PEP-253 (1977); [m] ωp hm (ωp )
H.A. Bethe, PR 66 (1944) 163 Z0
p=−∞
= (1)
[27] S.S. Kurennoy, PR E55 (1997) 3529; S. Kuren- ω eff

∞
noy, R. Gluckstern, ibid (1997) 3533 hm (ωp )
p=−∞
[28] S.S. Kurennoy, PA 39 (1992) 1; PA 50
(1995) 167; R.L. Gluckstern, PR A46, 1106, ωp = ω0 (pnB + μ + mνs )
1110 (1992); S.S. Kurennoy, R.L. Gluckstern, μ = coupled bunch mode number
G.V. Stupakov, PR E52 (1995) 4354 m = azimuthal mode number
[29] A. Fedotov, PhD Thesis, U. Maryland (1997)
[30] S.S. Kurennoy, G.V. Stupakov, PA 45 (1994) transverse
95
∞
[31] A. Novokhatski, A. Mosnier, PAC 97, 1661 Z⊥1 (ωp )hm (ωp − ωξ )
[32] K. Bane, A. Novokhatski, SLAC-AP-117 (1999) [m] p=−∞
(Z⊥1 )eff = (2)
[33] G. Stupakov, Proc. 19th Advanced ICFA Beam
∞
hm (ωp − ωξ )
Dynamics Workshop (Arcidosso, 2000), 141 p=−∞
[34] K.Y. Ng, PR D42 (1990) 1819; A. Burov, A. No-
ωp = ω0 (pnB + μ + νβ + mνs )
vokhatski, HEACC 92, 537
1
[35] G.V. Stupakov, PRST-AB 1, 064401 (1998) η = α − 2 , νβ betatron tune
[36] K. Bane et al, PAC 97, 1738 γ
[37] J.B. Murphy et al, PAC 95, 2980; PA 57 (1997) 9 ξ Δνβ Δp
[38] Y. Derbenev et al, DESY-TESLA-FEL 95-05 ω ξ = νβ ω 0 , ξ = /
η νβ p
(1995)
Tab.1 gives the spectral density hm (ω) ∝
[39] R. Warnock, P. Morton, SLAC-PUB-4562 ( (2
(1988); R. Warnock, Proc. 4th Advanced ICFA ( (
(λ̃m (ω)( , where λ̃m (ω) is the Fourier transform
Beam Dynamics Workshop (1990) 151
[40] K. Bane, P. Morton, LINAC 86, 490 of the line density, τ is the full bunch length in
[41] K. Yokoya, CERN-SL-90-88-AP (1988) time, στ is the rms bunch length in time. hm is
[42] G.V. Stupakov, SLAC-PUB-7167 (1996) chosen to be dimensionless. See also Tabs.2 & 3.
[43] E. Keil, B. Zotter, PA 3 (1972) 11; K.Y. Ng, Fer-
milab Report FN-389 (1981) Table 1: Spectral densities hm (ω) for various modes.
[44] T. Weiland and B. Zotter, PA 11 (1981) 143
[45] G. Dome, PAC 85, 2531 mode hm (ω)
[46] S.A. Heifets, S.A. Kheifets, PA 25 (1990) 61;
A. Hofmann, T. Risselada, B. Zotter, Proc. 4th 1 + (−1)m cos ωτ
sinusoidal (m + 1)2
π ) − (m + 1) ]
[( ωτ
ICFA Beam Dynamics workshop (1990) 138 2 2 2
[47] K.Y. Ng, R. Warnock, PAC 89, 798; PR D40 2
[Jm+1/2 (ωτ /2)]
(1989) 231 Legendre
[48] K.Y. Ng, PA 23 (1988) 93 |ωτ /2|
[49] T. Toyama, et al, HB 2006, 140; G. Nassib-
(ωστ )2m e−ω
2 σ2
Hermite τ
ian, CERN/PS 84-25 (BR) (1984); CERN 85-68
(BR) (1986)
[50] A. Piwinski, DESY Report 72/72 (1972)
262
Table 2: Complex longitudinal frequency shifts. ωs = 3.2.6 Parasitic Loss
νs ω0 , β = v/c. P. Wilson, SLAC, B. Zotter, CERN
Y.-H. Chin, KEK
mode Ωm − mωs
Total loss factor The energy change (parasitic
im
sinusoidal × loss) of a bunch of charge q and normalized line
m+1 density λ(t), traversing a structure with longitudi-

12NB e2 η Z0 [m] nal impedance Z (Sec.3.2) is
(parabolic bunch)
T0 Eβ 2 ωs τ 3 ω eff ΔE = −κ q 2 (1)
12i Γ(m + 12 ) Exact expressions for κ exist for special cases

Legendre √ × [1, 2]. Otherwise, one calculates numerically, e.g.

π (m − 1)!
by integrating the longitudinal electric field in a

NB e2 η Z0 [m] frame moving with the bunch (Secs.3.2.2, 3.2.3).
(parabolic bunch) If the bunch moves with a constant speed v < c,
T0 Eβ 2 ωs τ 3 ω eff
as e.g. in a proton linac, κ depends on v.

i Γ(m + 12 ) The loss factor κ , usually stated in units
Hermite ×
2π 2m (m − 1)! V/pC = 1012 V/C, is

1 ∞
NB e2 η Z0 [m] κ (σ) = dωZ (ω)h(ω, σ)
(Gaussian bunch)
2π −∞
T0 Eβ 2 ωs στ3 ω eff ∞
1
= dωReZ (ω)h(ω, σ) (2)
π 0
Table 3: Complex transverse frequency shifts for vari- where h(ω, σ) = λ̃(ω)λ̃∗ (ω) is the spectral power
ous beam models. ωβ = νβ ω0 . density of the bunch of rms length σ.
For a line spectrum, the integral becomes an
model Ωm − ωβ − mωs infinite sum. For a single bunch in a circular ac-
celerator,
i ∞
sinusoidal − × ω0
1+m κ (σ) =
Z (pω0 )h(pω0 , σ) (3)
2π p=−∞
NB e2 c [m]
(square bunch) (Z⊥1 )eff For short bunches in large machines (ω0 1/σ),
2ωβ T0 Eτ
the sum can be replaced by an integral.
i Γ(m + 12 ) In terms of wake function W (t) (Sec.3.2),
Legendre − √ ×
2 π m! the loss factor
∞
is ∞
NB e2 c [m] κ =
dτ Wλ (τ )λ(τ ) = dtW (t)S(t)
(square bunch) (Z⊥1 )eff −∞ 0
ωβ T0 Eτ (4)
∞
i Γ(m + 12 ) where Wλ (τ ) = 0 dtW (t)λ(τ − t) is the
Hermite − × ∞
4π 2m m! wake potential, S(t) = −∞ dτ λ(τ )λ(τ − t) is
the auto-correlation function of the line density.
NB e2 c [m]
(Gaussian bunch)
ωβ T0 Eστ
(Z⊥1 )eff For a Gaussian√ bunch with rms 2σ 2 = σz /βc,
λ = e−t /2σ /( 2πσ), h(ω) = e−ω σ ,
2 2
1
e−τ /4σ
2 2
S(τ ) = √ (5)
2 πσ
References Examples of auto-correlation functions for other
bunch distributions can be found in [3].
[1] A.W. Chao, Physics of Collective Beam Instabili-
In a resonator with shunt resistance Rs (≡
ties in High Energy Accelerators, Wiley (1993) 2 /(2P
[2] F.J. Sacherer, PAC 77, 1393
Vcav wall )), quality factor Qr , and resonant
frequency ωr ,
Rs
[3] F.J. Sacherer, 9th Int.Conf.High Energy Acc.
(1974) 347 κ = ω 1 w(ω 1 σ) − ω 2 w(ω 2 σ)
4Qr

[4] B. Zotter, CERN/ISR-TH/78-16 (1978)

[5] B. Zotter, CERN/ISR-TH/80-03 (1980) Rs
= Re ω 1 w(ω 1 σ) for Qr ≥ 21 (6)
[6] K. Balewski, R.D. Kohaupt, DESY 90-152 (1990) 2Qr
263
where ω1,2 = (ωr /Qr )[−i/2 ± Qr ], Qr 2 = Longitudinal Loss Factor V/pC
Q2r − 1/4, and w(z) is the complex error function
[4]. The loss factor, normalized by ωr Rs /2Qr , 10
is shown in Fig.1a. Approximate expressions for (a) Qr = 0.50
κ are found below, while the relation between κ

Qr = 1.00
8
and Z /n is discussed in [1]. Qr = 2.00
Qr = 5.00
Qr = ∞
κ (σ) ≈
Cases
6
high-Q cavity mode ωr Rs −ωr2 σ2

e
(Q > 10 or so) 2Qr 4

short bunches ωr Rs 2 ωr σ
1−
ωr σ 1 2Qr π Qr 2
low-Q cavity and Rs
√
long bunches 4 πQ2 ωr2 σ 3 0
1 2 3
ωrσ
The kick factor The transverse loss factor for Transverse Loss Factor
the dipole (m = 1) mode, also called the kick
factor, with dimensions V/(pC-m), is defined by 0.3 (b) fr = 10.00 GHz
∞
1 Qr = 0.50
κ (σ) =
⊥ dωZ⊥ (ω)h(ω, σ) Qr = 1.00
2πi −∞
∞ Qr = 2.00
1 Qr = 10.00
= dωImZ⊥ (ω)h(ω, σ) 0.2
π 0
∞
= dtW⊥ (t)S(t)
0 ∞
0.1
= dτ W⊥λ (τ )λ(τ ) (7)
−∞
in terms of the transverse impedance Z⊥ (ω), wake

potential W⊥λ (τ ), and wake function W⊥ (τ )
0
(Sec.3.2). For a Gaussian bunch traversing a res- 0 2 4 6 8 10
onator the kick factor, normalized by ωr R⊥ /Qr , ωrσ 11–97
8355A45
is shown in Fig.1b (for Qr ≥ 1/2),

ωr R⊥ Figure 1: Normalized (a) longitudinal and (b) trans-
κ (σ) = w(ω1 σ) − w(ω2 σ) verse loss factors of Gaussian bunch in resonator im-
⊥
4Qr
pedances with various quality factors Qr versus ωr σ.
ωr R⊥
=
Im w(ω1 σ) (8)
2Qr The modal loss factors For the n-th longitudi-
nal mode, the modal loss factor is defined kn =
ωr R⊥ ωr σ
For short bunches κ (σ) ≈
⊥ √ , while for Vn Vn∗ /(4Un ), where Vn is the induced voltage
Qr π across the impedance and Un the stored energy.
R⊥ 1
long ones κ (σ) ≈
⊥ √ . For a resonator impedance Z ,
Qr 2 πσ
The parameters of an equivalent transverse 1 ∞ ωn Rn
kn = dωReZ (ωn ) = (9)
broad-band resonator can be estimated from the π 0 2 Qn
dependence of the kick√factor on bunch length. In
Behind a point-bunch (τ > 0), the wake function
particular, R⊥ /Qr = 2 π limσ→∞ (ωr σ) κ , and
⊥
is
ωr = αc/σ̂, where σ̂ is the location of the maxi- ∞

mum κ (σ), and α is a correction factor of order
⊥
W (τ ) = 2 kn cos (ωn τ ) (10)
unity [5]. n=1
264
while it vanishes in front of it (τ < 0). For a where b is the vacuum chamber radius, g is the
Gaussian bunch, the total loss factor is axial length of the cavity.
∞
∞ For perfectly conducting walls, Re Z is zero
kn e−ωn σ (11)
2 2
κ (σ) =
kn h(ωn , σ) = below the first cavity resonance at ω01 /c = j01 /d,
n=1 n=1 and κ decreases exponentially with σ, rather than

For the n-th dipole mode, the modal kick fac- σ −3 in the resonator model.
∗
V1n V1n
tor is k1n = where V1n is the voltage at For a cavity-like structure with a small open-
4U1n ing gap (smaller than the chamber radius) and for
the iris or beam tube radius. The transverse wake a short bunch (shorter than the chamber radius),
function is the loss factor can be estimated in a good accuracy
∞
2c k1n only from the gap size, the chamber radius and the
W⊥ (τ ) = 2 sin (ω1n τ ) (12)
b ω1n bunch length, regardless of the exact shape and

m=1
size of the structure. This is because the bunch
For a Gaussian bunch, the total kick factor is
does not have enough time to see the details of
2c k1n ∞ the structure in one passage, and the interaction
κ (σ) = 2
⊥ dt S(t) sin ω1n t
b ω1n 0 takes place almost solely between the gap and the
2c k1n bunch. The analytical expression of the loss fac-
=√ 2 D(ω1n σ) (13)
πb ω1n tor in this case is available in a form suitable for
z numerical evaluations [9].
where D(z) = exp (−z 2 ) 0 dt exp (t2 ) is the
Dawson integral [4]. Structures without cylindrical symmetry For
vacuum chambers not cylindrically symmetric
Examples other than resonator impedance
but with mirror symmetry about two orthogonal
For further examples (other vacuum chamber dis-
planes, intersecting at an axis, the loss factors de-
continuities, such as tapers and steps) [2, 6].
pend on beam displacement. The kick factors of
Resistive wall For a cylindrical resistive wall
off-axis beams have finite cross terms and can be-
with conductivity σc and pipe radius b, the loss
come negative (focusing).
factor per unit length for a Gaussian bunch is
For structures with less or no symmetry, no
κ (σ)

Γ( 34 )c Z0 1/2 axis is defined, and the loss factors must be com-
= 3/2
(14)
L 4π 2 bσz 2σc puted along straight trajectories, e.g. parallel to
Γ(3/4) ≈ 1.225. Eq.(14) holds also for a beam the vacuum chamber wall.
moving (a) centered between two parallel plates
References
spaced by 2b, and (b) a distance b from a single
resistive plate. For an elliptical pipe, κ is min-
[1] A.W. Chao, Physics of Collective Beam Instabili-
imum when (major axis)/(minor axis)≈1.4, and ties in High Energy Accelerators, Wiley (1993)
the minumum value is 0.87 of Eq.(14) [7]. [2] B. Zotter, S. Kheifets, Impedances and Wakes in
Collimator For a collimator of aperture a High Energy Particle Accelerators, World Scien-
in a pipe of radius b, and cσ a, κ ≈
tific (1997)
Z0 ln(b/a)/(2π 3/2 σ). [3] R. Bracewell, The Fourier Transform and Its Ap-
Very short bunches For very short bunches in plications, McGraw-Hill (1965) 152
a linac structure, κ is not determined by the col- [4] M. Abramowitz, I. Stegun, Handbook of Mathe-

lective interaction of the beam with the periodic matical Functions (1965)
cavity array, but by diffractive losses in the indi- [5] B. Zotter, CERN Report LEP-TH/87-34 (1987)
[6] S.A. Heifets, S.A. Kheifets, PR D39 (1989) 960
vidual cells [6]. In this regime (σ < a2 /(N p),
[7] A. Piwinski, DESY 94-068 (1994)
with a iris aperture radius, p periodic length, N
[8] K. Bane, M. Sands, SLAC-PUB-4441 (1987)
number of cells in the structure), κ diverges as

[9] Y. Shobuda, Y.H. Chin, K. Takata, PRST-AB 10,
σ −1/2 (see Secs.3.2.2, 3.2.4). 044403 (2007)
For a resonant cavity in a storage ring, κ is
given by a sum over modes up to the cut-off fre-

quency, plus a high frequency diffraction contri- 3.2.7 Trapped Modes
bution (diffraction model [8], also Sec.3.2.4) S.S. Kurennoy, LANL

Γ( 14 )Z0 cg The term “trapped mode” refers to a mode which
κ ≈
(15) cannot propagate in the beam pipe. Often this
4π 5/2 b σ
265
Sec.3.3: PARTICLE-MATTER INTERACTION
term is reserved only for modes whose frequency measurements of the trapped modes in perforated
is above the pipe cutoff frequency [1], since pipes [6]. The trapped modes produced by small
modes below the pipe cutoff cannot propagate in discontinuities in vacuum chambers with arbitrary
the pipe anyway. An example of a trapped mode homogeneous cross sections have been investi-
is a mode in a long cavity, e.g., multi-cell cavity or gated in [7].
IR, with fields concentrated near the cavity center
and vanishing toward its ends. Such a mode does References
not propagate into the pipe due to its very weak
coupling to the pipe modes. The trapped modes [1] S. Heifets, S. Kheifets, RMP 63 (1991) 631
usually produce narrow resonance peaks of the [2] T. Weiland et al, PRST-AB 2, 042001 (1999)
coupling impedance. In normal-conducting cavi- [3] R. Schuhmann, T. Weiland, PRST-AB 3, 122002
(2000)
ties such resonances can be damped (coupled out),

[4] G.V. Stupakov, S.S. Kurennoy, PR E49 (1994) 794
but supressing them in superconducting cavities is
[5] S.S. Kurennoy, PR E51 (1995) 2498
complicated. A detailed analysis of the trapped
[6] F. Caspers, T. Scholz, PA 51 (1995) 251
modes in multi-cell cavities, supported both by
[7] S.S. Kurennoy, R.L. Gluckstern, G.V. Stupakov,
numerical and experimental studies, can be found PR E52 (1995) 4354
in [2, 3].
In other cases the term “trapped mode” is gen-
eralized by including the modes with frequencies
slightly below the cutoff [4]. A small discontinu- 3.3 PARTICLE-MATTER
ity (e.g. a pillbox or a hole, but not an iris or a INTERACTION
post) on a smooth pipe can produce such a mode
localized near the discontinuity [4]. For a small In this section, effects caused by beam-beam col-
axisymmetric bulge with the area A in its lon- lisions and by beam interactions with residual
gitudinal cross section, the frequency shift Δω01 gas and accelerator components are considered.
down from the cutoff frequency ω01 = μ01 c/b of These effects lead to decrease of beam intensity,
the TM01 mode in the cylindrical pipe of radius luminosity deterioration in colliders, and forma-
b is tion of beam halo. Loss of beam particles fur-
2
μ2 A ther leads to (i) beam induced gas desorption; (ii)
Δω01 = ω01 01 (1) irradiation of conventional and superconducting
2 b2
components with a possible quench; (iii) increase
where μ01 is the first root of the Bessel function
of background rates in the detectors; (iv) possi-
J0 (x). The damping rate γ01 for the case of a ble radiation damage, overheating or destruction
finite, though large, conductivity σ of the walls, of components; (v) increased prompt and residual
is γ01 = ω01 δs /(2b), where δs is the skin depth. radiation fields in the machine and environment.
The trapped mode disappears when γ01 becomes
larger than Δω01 , but in the limit of perfectly con- 3.3.1 Basic Formulae [1]
ducting walls it exists even for a very small per- M. Tigner, Cornell U.
turbation. This trapped mode leads to a resonance A.W. Chao, SLAC
of the longitudinal impedance with the shunt im-
pedance Symbols M incident particle mass, E = γM c2
Z0 μ301 A3 incident particle energy, T kinetic energy, ze
R01 = (2) charge of incident particle, Z, A atomic num-
πδs b5
The results for an enlargement remain valid for ber and atomic mass of medium, I mean exci-
a hole if one replaces A → αθ /(4πb) in Eqs.(1) tation energy, δ density effect
correction to ion-
and (2), where αθ is the magnetic susceptibility energy loss, ωp = 4πNe re3 me c2 /α =
ization
of the hole (Sec.3.2.4). The resonances due to 28.816 ρZ/A [eV] plasma energy (ρ in g/cm3 ),
the trapped modes can be dangerous in large su- Ne electron volume density (m−3 ), K/A =
perconducting colliders like the LHC, where the 4πNA re2 me c2 /A, X0 radiation length in g/cm2 ,
vacuum chamber incorporates the inner thermal Ec critical energy in MeV.
screen (liner) with many pumping holes. The Single Coulomb scattering of spin- 12 paticles
trapped-mode effect for many holes has been Neglecting atomic electrons and nuclear form fac-
studied analytically [5], and was confirmed by tor effects and assuming scattering nucleus at rest
266
in lab [2], 10
2 8
dσ Z2 2 me c 1 θ
= r 1 − β sin
2 2 6
4 e
H2 Liquid
- dE/dx (MeV g-1cm2)

θ
dΩ βp sin4 2
2 5
(1) 4
When screening is included, and for small angles, He Gas
3
Eq.(1) reads Al C

Fe
dσ me c 2 1 2
≈ 4Z 2 re2 2 2 )2 (2) Pb
dΩ βp (θ + θmin Sn
where θmin ≈ αZ 1/3 (me c/p).

1
0.1 1.0 10 100 1000 10 000
βγ = p/Mc
Thomson cross section Unpolarized EM radia- 12–97
8355A48
tion on free electron or proton [3],
Figure 1: Energy loss rate in liquid hydrogen, gaseous
2
re,p
dσ helium, carbon, aluminum, tin, and lead.
= (1 + cos2 θ) (3)
dΩ 2
8π 2
σT = r (4) 50000
3 e,p
20000 C
Klein-Nishina cross section For high energies 10000 Fe
ωx−ray me c2 [3], 5000 Pb
2000
R/M (g cm-2 GeV-1)
H2 liquid
σKN 3 mc2 1 1 2ω 1000
≈ + ln (5) He gas
σT 4 ω 4 2 me c2 500
200
See also Sec.7.6.1. 100
50
Ionization energy loss Bethe-Bloch equation 20
for mean rate of energy loss (stopping power) 10
for moderately relativistic charged particles other 5
than electrons, 2
1
dE Z 1 0.1 2 5 1 2 5 10 2 5 100
− = Kz 2 (6) 12–97
βγ = p/Mc
dx A β2 8355A61

1 2me c2 β 2 γ 2 Tmax δ Figure 2: Range of heavy charged particles in liquid
× ln − β2 −
2 I2 2 (bubble chamber) hydrogen, helium gas, carbon, iron,
and lead.
Here
K/A = 4πNA re2 me c2 /A Multiple Coulomb scattering through small
angles If we define θ0 = θ rms √1 θ rms ,
= 0.307 MeV cm2 g−1 for A = 1 g/mol plane = 2 space
then it is often sufficient to use a Gaussian ap-
2me c2 β 2 γ 2
Tmax = proximation for the central 98% of the projected
1 + 2γme /M + (me /M )2 angular distribution, with a width

is the maximum kinetic energy which can be im- 13.6 MeV x x
parted to a free electron in a single collision. The θ0 = z 1+0.038 ln (7)
βcp X0 X0
units are so that dx is measured in mass per unit
area (g cm−2 ). The mean excitation energy I is Here p, βc, and z refer to the incident particle; x
(10 ± 1 eV) × Z for elements heavier than oxy- is the thickness of the scattering medium. This
gen. At very high energies, the density effect cor- value of θ0 is from a fit to Molière distribution for
rection δ/2 → ln(ωp /I) + ln βγ − 1/2. For singly charged particles with β = 1 for all Z, and
pions on various material, see Fig.1. Eq.(6) may is accurate to 11% or better for 10−3 < (x/X0 ) <
be integrated to find the total range R. See Fig.2. 100.
267
200 1000
Copper
X0 = 12.86 g cm−2 l
ta
Ec = 19.63 MeV
l
to
ta
dE/dx (MeV g −1 cm 2 )
100 100
Fe
to
dE/dx × X0 (MeV)
U
ng
l
70 ta
lu
To
ah
Rossi:
tr
50
Ionization per X0 10
ss
tb E
= electron energy H total
m
40
s≈
re
em
30 Fe ion
Br
ac
Ex
Ionization
s
1
em
ir
cl
20
pa
br
nu
Fe
Fe
Fe
Brems = ionization
0.1
10
2 5 10 20 50 100 200 1 10 102 103 104 105
Muon energy (GeV)
Electron energy (MeV)
Figure 3: Two definitions of the critical energy Ec . Figure 4: Average energy loss of a muon in hydrogen,
iron, and uranium as a function of muon energy.
Radiation length This is the mean distance

over which a high-energy electron loses all but 1/e Bremsstrahlung on gas nuclei [2]
of its energy by bremsstrahlung, and is the scale dσ
length for high-energy electromagnetic cascades ≈ 4αre2 Z(Z + 1)
du
(other than hydrogen), 4 184.15
× (1 − u + .75u2 ) ln
1 NA 2
3u Z 1/3
= 4αre2 Z Lrad −f (Z) + ZLrad
X0 A
(8) with u = k/E the fractional energy loss by radia-
−1
tion, where k is the energy of the photon radiated
f (Z) ≈ a (1 + a ) + 0.20206
2 2
in the bremsstrahlung event. Integral forms are
found in Secs.3.1.12, 3.3.2 and 4.6.
−0.0369 a2 + 0.0083 a4 − 0.002 a6
(a = αZ) (9) Beam-beam bremsstrahlung loss cross section
Lrad = ln 184.15 , L = ln 1194
(10) [2] (See also Sec.2.5.3.2.)
Z 1/3 rad Z 3/2
E
For A = 1 g/mol, 4αre2 NA /A = (716.4 g/cm2 )−1 . σbbb loss =
dσ
dk
An electron loses energy by bremsstrahlung kmin dk
at a rate nearly proportional to its energy, while
the ionization loss rate varies only logarithmically 16αre2 E2 1 Ebeam 5
with the electron energy. The critical energy Ec ≈ ln 2cm4 − ln −
3 me c 2 kmin 8
is sometimes defined as the energy at which the
two loss rates are equal, Ec = (800 MeV)/(Z + 2
1 Ebeam 3 π2
1.2). Among alternate definitions of Ec is that of + ln − − (11)
Ref.[4], which defines Ec as the energy at which 2 kmin 8 6
the ionization loss per radiation length is equal where kmin is the largest energy change that can
to the electron energy. Equivalently, it is the be tolerated by the acceptance of the accelerator.
same as the first definition with the approxima-
tion |dE/dx|brems ≈ E/X0 . These definitions
are illustrated in the case of copper in Fig.3. References
Muon energy loss at high energy At suffi- [1] K. Nakamura et al (Particle Data Group), J. Phys.
ciently high energies, radiative processes become G37, 075021 (2010)
more important than ionization for all charged [2] F.C. Porter, NIM A302 (1991) 209
particles. For muons and pions in materials such [3] J.D. Jackson, Classical Electrodynamics, 3rd ed.,
as iron, this “critical energy” occurs at several Wiley (1999)
hundred GeV. See Fig.4. [4] B. Rossi, High Energy Particles, Prentice-Hall
(1952)
268
0
3.3.2 Beam and Luminosity Lifetime 10
P (torr)
3.3.2.1 Protons −1
4.5E−7
1.0E−6
10
N.V. Mokhov, V.I. Balbekov, FNAL 3.0E−6
2.0E−5
Particle density
One needs to distinguish: Beam lifetime, Emit-
tance lifetime, and (in colliders) Luminosity life- 10
−2
time. Involving phenomenon is beam halo forma-

tion that depends on partial effects i. The total

lifetime τtotal is calculated from 10
−3
1 1
= (1)
τtotal τi −4
i 10
0 10 20 30 40
Partial effects exclusively due to beam interac- x (mm)
tions with residual gas and particle collisions

at the interaction point (IP) are considered in Figure 1: Transverse beam profile in 70 GeV proton
synchrotron for several values of residual gas pressure.
this section. Other related effects are consid-
ered elsewhere: Touschek effect (Sec.2.4.12)
and intra-beam scattering (Sec.2.4.12); Beam usually a mixture of H, C, N and O atoms. At high
cooling (Secs.2.7,4.16); Beam-beam scattering energies, the differential cross-section of proton-
in colliders (Sec.4.7.6.2); and other beam-beam hydrogen (antiproton-hydrogen) elastic scattering
effects (Sec.2.5.1,4.13,4.14). is well described by a simple exponent
dσ/dt ∼ e−b|t| (3)
Beam lifetime is determined by single-acting
processes: beam-gas interactions and (in col- where t ≈ −p2 θ 2
is the 4-momentum transfer
liders) by beam-beam collisions at the IP, squared, p is the proton momentum, in units of
1/τbeam = 1/τgas + 1/τIP . As a result of beam- (GeV/c)2 . At 0.001 < |t| < 0.8 (GeV/c)2 ,√b in-
creases with energy from 12 (GeV/c) −2 at s=
gas single electromagnetic and nuclear elastic √
scatterings with angles exceeding the angular 10 GeV up to 16.4 (GeV/c)−2 at s = 1.8 TeV,
beam size, and nuclear inelastic interactions, where s is the CM collision energy squared [3].
particles leave the core creating beam halo and The proton-nucleus elastic differential cross-
ultimately are lost on the limiting apertures of the section on light and medium nuclei (9 < A < 62)
machine. Diffusion caused by multiple Coulomb at p > 5 GeV/c is described as [4]
scattering (MCS) on electrons and nuclei of dσ
[mb/str] = 38.0 p2 A1.63 e−14.5A |t|
0.66
residual gas atoms is negligible at E > a few dΩ

GeV [1] (Sec.4.7.6.2). The resulting “beam pro- + 53.2 p2 A0.33 e−10|t| (4)
file” can be drastically different from Gaussian.
Fig.1 shows a transverse profile measured at the Corresponding rms angle θel = C/p, where

70 GeV proton synchrotron at IHEP (Protvino) 0.0595A−0.02 + 0.175
and fitted for different residual gas pressures P C=
0.862A0.64 + 1.75
by [2]
For nitrogen C = 0.19 GeV/c. Using Eq.(4), one
P finds that, e.g. at Tevatron (p = 980 GeV/c), most
N (x) = e−x
2 /2σ 2
+ C0 ln(P/P0 ) (2)
x2 of elastic (and certainly all inelastic) scatterings
where x is a distance from the beam axis in mm, force particles to leave the core creating halo and
(3 < x < 40 mm), σ = 2.2 mm, C0 = beam loss, with
−4.8 × 104 mm2 /Torr, P < 10−3 Torr, and θel ∼ 0.2 mrad θbeam = 0.011 mrad
P0 = 10−3 Torr.
where θbeam is rms beam divergence.
At high energies elastic and diffractive scat-
Summing over species of circumferentially
terings contribute to the beam halo and emittance
averaged gas densities ni , one gets
growth, while inelastic nuclear interactions result
1 1 dN
in multi-particle production and shower develop- =− =v σi n i (5)
ment along one or two magnets. Residual gas is τgas N dt
i
269
pp
W
Cu
3
Fe
Total cross section (mbarn)

2 10
10 Al
Cross section (mb)
N
C
total Be
2
10
p
1 elastic
10
1
−1 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
p (GeV/c) Laboratory beam momentum (GeV/c)
Figure 2: Total and elastic pp cross-sections [5]. Figure 4: Total proton-nucleus cross-sections [6].
pipe regions of “constant” pressure and tempera-

tures, Eq.(5) for molecular nitrogen becomes
pbarp
1 wk Pk
≈ 2.4 × 105 β (7)
10
2 τgas [s] Tk
k
Cross section (mb)
total where Pk (Torr) is the gas pressure in the pipe of

the kth region of the fraction wk of the total cir-
cumference at the temperature Tk (K), relative ve-
locity β ≈ 1. For a gas mixture one should use the
1 elastic
equivalent nitrogen pressure
10
2/3
Nucl
Peq = 0.0861 Pi kij Aij (8)
−1 0 1 2 3 4 5 6 7 8 9
i j
10 10 10 10 10 10 10 10 10 10 10
p (GeV/c)
where kij is the number of j atoms in i molecule
Figure 3: Total and elastic pp̄ cross-sections [5].
and Aij is atomic mass of j atom in i molecule.
In colliders, all inelastic beam-beam interac-
tions remove particles from each beam. Only a
fraction of elastic events at IP has the same effect.
where v = βc is the proton velocity, and σi is the
For example, using Eq.(3) for the IP of Tevatron,
total proton or antiproton cross-section on species
i. Experimental data on the total and elastic cross- θel ∼ 0.17 mrad, θbeam ∼ 0.09 mrad (9)
sections for pp or pp̄ collisions are presented in
Figs.2-3 along with the Regge description [5] and about 40% of elastically scattered protons
(anti-protons) remain in the 3σ core after a bunch-
σtot [mb] = 22 s0.079 + Y s−0.46 (6) bunch collision at the IP. So, the beam lifetime
where Ypp = 56.1, Ypp̄ = 98.2. Fig.4 shows due to collisions is given by a sum over interac-
total proton-nucleus cross-section calculated in tion rates at each of the n IP in the collider,
[6] within a multichannel generalization of the 1 1
Glauber model using σpp parameterizations of ex- = σpp Ln (10)
τIP nB NB n
perimental data and the Woods-Saxon distribu-
tion of nucleons in nuclei. The total p-A cross- where nB is the number of bunches, NB is the
section remains approximately constant at proton number of particles per bunch, σpp is the inelas-
momenta 1 GeV/c < p <50 TeV/c. tic plus a fraction of elastic (60% in the Tevatron
For nitrogen σ ≈ 410 mb at proton momenta case, 9 mb) pp or pp̄ cross-section, and Ln is the
3 GeV/c < p < 5 TeV/c. Summing over the beam average luminosity at the nth IP.
270
Emittance lifetime in colliders, assuming where β ∗ and L are the β-function and luminos-
identical beams, is defined as 1/τ = (1/)d/dt, is the fraction of the elastic scat-
ity at the IP, σel
with being the rms normalized beam emittance. tering cross-section resulting in emittance growth,
Under normal conditions the growth of is the and θel is the rms projected elastic scattering an-
dominant cause of luminosity deterioration [7]. gle, obtained from Eq.(3). This gives (see also
In an “ideal” machine, the growth of emittance is Eq.(14), Sec.4.7.6)
due to beam-gas MCS, beam-beam elastic scatter- 1 dN 1 f0 NB γ 2
ing at the IPs, and the intrabeam scattering. The = σel θel (16)
N dt 8 2N
beam-gas emittance growth over a given time pe-
riod is proportional to the residual gas pressure where f0 is the revolution frequency.
and the average β-function [8]. Summing again Luminosity lifetime in colliders is a sum of
over the beam tube regions of “constant” pressure (7), (10), (11) and (16),
and temperatures, the emittance growth rate due
1 1 1 1
to MCS, e.g. in x-plane, is (see also Sec.4.7.6) = + + (17)
1 dx π τL τbeam1 τbeam2 τ
2
= βγ wk βxk θtk (11)
x dt 2x As stated at the beginning, the role of partial con-
k
tributions is different depending on the machine.
where x is the normalized rms beam emittance
For example, for the Tevatron collider, beam-gas
(phase space area with a factor of π included).
2 is the rms angle of MCS per time unit interactions and pp̄ collisions at the IPs are the
Here θtk
dominant sources in this equation for the pro-
in the pipe of the kth region, and βxk is average
cesses considered in this section. In reality, pro-
β-function in x-plane in the kth region. Summing
cesses listed at the beginning as well as the abort
over species of average gas densities ni in each
kicker system performance, the low-β quadrupole
region, one gets
squeeze process, Main Ring Lambertson power
4πrp2 c −1/3 supply noise, the main dipole/quadrupole power
θt2 = 3 2 ni Zi (Zi + 1) ln(183Zi )
β γ supply ripple and the ambient ground motion [9]
i
(12) all could result in reduction of the luminosity life-
where Zi is atomic number of species, and rp = time. Fortunately, many of these sources can be
1.54 × 10−16 cm. For molecular nitrogen at pres- (and in the Tevatron have been) mitigated through
sure P (Torr) and temperature T (K), studies and improvements.
1 dx 2.2 × 107 wk βxk Pk
≈ (13) References
x dt x β 2 γ Tk
k
where time is in seconds, βxk in meters and x in [1] A.N. Kalinovskii, N.V. Mokhov, Yu.P. Nikitin,
π mm-mrad. Passage of High-Energy Particles through Matter,
For arbitrary mixture of i gases with par- AIP (1989)
tial pressures Pi , one uses the nitrogen-equivalent [2] A.A. Asseev et al, IHEP 79-91, Protvino (1979)
pressure referred to temperature of 273 K in each [3] R. Rubinstein, Fermilab-Conf-90/160-E (1990)
[4] J. Ranft, PA 3 (1972) 129
machine region
[5] Review of Particle Properties, PR D54 (1996)
PeqCoul
≈ 2 × 10−3 Pi (14) [6] S.I. Striganov, IHEP 94-14, Protvino (1994)
i [7] G. Jackson et al, Fermilab-Conf-88/80 (1988)
−1/3
× kij Zij (Zij + 1) ln(183Zij ) [8] C. Bovet et al, CERN/MPS-SI/Int. DL/70/4
(1970)
j
[9] G. Jackson, AIP Proc. 255 (1992) 105
In a collider, elastic scattering off the oncom-
ing beam particles at the IP also causes emittance
growth. (Intra-beam scattering is discussed in 3.3.2.2 Electrons
Sec.2.4.12.) Eq.(11) can be applied replacing β- M.S. Zisman, LBNL
functions averaged over machine regions with β ∗
at each IP and using the following expression for Symmetric collider If the beam sizes remain
the rms scattering angle in one of the beams constant as the luminosity decreases (not true
2 L when beam-beam limit is reached), the time de-
θt2 = σel θel (15) pendence of L(t) is contained entirely in the time
nB NB
271
dependence of N (t). (See also Sec.4.1.) There where

are then two distinct processes for luminosity
σ+ σ− N0+ σ−
loss: (i) those due to collisions between the two G = L0 − , r= (23)
N0+ N0− N0− σ+
beams, and (ii) single-beam loss processes. The
processes in (i) relevant to a high-energy e+ e− The time for the luminosity to decay to a fraction
collider are Bhabha scattering (BS) and radiative f of its initial value is
%
Bhabha scattering (RBS) (Sec.3.3.3). In BS pro- 1 1
cess, a beam particle is lost if its angular deflec- tf = ln (1 − r)2 + 2f r
G 2f
tion is beyond the ring’s transverse acceptance; in
RBS process, loss occurs if the beam particle’s + (1 − r) (1 − r)2 + 4f r (24)

momentum change is outside the longitudinal ac-
ceptance of the ring. Other loss processes A full treatment of

For a symmetric collider (two beams of iden- luminosity lifetime must also take into ac-
tical sizes and intensity) with a loss cross section count the single-beam losses that can inde-
σ (Sec.3.3.1), the particle loss rate is pendently decrease N+ and N− . These in-
clude gas scattering (both elastic and inelas-
dN
(t) = −σL(t) = −σkN 2 (t) (18) tic i.e. gas bremsstrahlung), Touschek scattering
dt (Sec.2.4.12), and, at high energy, Compton scat-
where k = L0 /N02 = fc /(4πσx∗ σy∗ ) with fc the tering off thermal photons (Sec.3.3.4) and beam-
collision frequency, σx,y ∗ the beam sizes at the IP,
strahlung (Sec.2.5.3). In a simple model [2], the
and subscript 0 means initial values at t = 0. gas scattering rate is independent of beam inten-
Solving Eq.(18) yields [1] sity, leading to exponential decay,

1 1 1 dN
N (t) = N0
1 + (σL0 /N0 )t =− = ng βcσi (25)
τi N dt
2
1 where ng = 9.656 × 1024 nZ P [Torr]/T [K] is
L(t) = L0 (19)
1 + (σL0 /N0 )t the number density of residual gas atoms, βc is
The time for the luminosity to decay to a frac- the beam velocity, nZ is the number of atoms
tion f of its initial value is per gas molecule, and σi is the cross section,
where i denotes either the elastic scattering (el)
N0 1
tf = √ −1 (20) or bremsstrahlung (Br) process. The elastic cross
σL0 f section is [3]
Asymmetric collider Generalizing to a two-
2πre2 Z 2 β⊥
ring asymmetric collider (still assuming constant σel = (26)
beam sizes), the particle loss is described by [1] γ 2 A⊥
dN+ where A⊥ and β⊥ are the ring limiting accep-
(t) = −σ+ kN+ N− tance (beam chamber radius squared divided by
dt
dN− β⊥ ) and corresponding transverse β-function, re-
(t) = −σ− kN+ N− (21) spectively. The bremsstrahlung cross section in
dt terms of the radiation length X0 (Sec.3.3.1) is [1]
L0 fc /2π
k= + − = 4 A 1 1 5
N0 N0 (σ + σ ∗2 )(σ ∗2 + σ ∗2 )
x+
∗2
x− y+ y−
σBr = ln − (27)
3 NA X0 (Δp/p)lim 8
where we no longer assume that the loss cross sec- where A is the atomic weight of the scattering nu-
tions for the two beams are identical. The solution cleus, NA is Avogadro’s number, and (Δp/p)lim
to Eq.(21) is [1] is the limiting momentum acceptance of the ring

1−r (either a longitudinal limit from the rf bucket
N+ (t) = N0+
eGt − r height, or a transverse limit from scattering in a
dispersive section of the ring). Note that, in a typ-
1−r
N− (t) = N0− ical ring the residual gas pressure, and thus the de-
1 − re−Gt cay rate, includes a term proportional to the beam

1−r 2 intensity, leading to non-exponential decay. The
L(t) = L0 eGt
(22)
eGt − r decay rate for Touschek scattering also depends
272
on the beam intensity [2]. For typical collider Differential cross section with one photon ex-
parameters, beam-gas bremsstrahlung is the most change for unpolarized beams in CM [1]
important of the single-beam loss mechanisms. QED
dσBorn
At highest energy circular e+ e− colliders [4] the (e+e− → e+e− ) (1)
beam lifetime can be limited by beamstrahlung dΩ
[5]. α2h̄2 c2 1+cos4 2θ 2 cos4 2θ 1
= − + (1+cos θ) 2
2s sin4 2θ sin2 2θ 2
Lifetime mitigation For synchrotron light √
when s ≡ (p1 + p2 )c me c (typically s =
2
sources, it is desirable to keep a nearly constant
current circulating in the ring to minimize cyclic 4Eb2 ).
thermal changes in precision optical components. The large forward cross section is useful for
For this reason, most modern light sources have monitoring e± luminosity while the larger angle
adopted the technique of top-up injection [6], scattering is used for electroweak measurements
whereby small amounts of beam are injected on e.g. using the Z annihilation diagram. Based on
a quasi-continuous basis to keep the circulating lepton universality, the e± annihilation diagram
current essentially constant. Especially for corresponds to e+e− → μ+μ− and more generally,
asymmetric colliders, an identical approach, for elementary fermions f of charge Qf e,
which mitigates the decrease in luminosity due dσ + − α2h̄2 c2
to current decay in one or both rings has been (l l → f +f − ) = (1 + cos2 θ)Q2f
dΩ 4s
implemented [7, 8] (also see Sec.3.3.9.1). By (2)
making the average and peak beam currents more Integrating gives the (partial) total cross section
nearly the same, the gain in integrated luminosity of the expected luminosity or background rates in
can be substantial, up to 30–50%, and this has annular detectors. For example, for the u-quark
contributed to the world-record luminosity of with Qf = 2/3 and muons,
2.1 × 1034 cm−2 s−1 at KEKB [9]. 4πα2h̄2 c2 θ 1
σ(e+e− → u+u− ) = cos θ+ 31 cos3 θ
18s θ2
References 9
σ(e e → μ μ ) = σ(e e → u u )
+ − + − + − + −
(3)
4
[1] F.C. Porter, NIM A302 (1991) 209 2 2 2 2
where 4πα h̄ c = 261 GeV nb.
[2] M.S. Zisman, J. Bisognano, S. Chattopadhyay, PA
23 (1988) 289; M.S. Zisman, J. Bisognano, S. Møller scattering The corresponding cross
Chattopadhyay, ZAP User’s Manual, LBL-21270, section for e±e± → e±e± is [2]
QED
UC-28 (1986) dσBorn α2h̄2 c2
[3] J. Le Duff, NIM A239 (1985) 83 (e±e± → e±e± ) = (4)
# dΩ 2s $
[4] A. Blondel, F. Zimmermann, arXiv: 1112.2518
1 + cos4 2θ 1 + sin4 2θ 2
[hep-ex] × 4 θ
+ + 2θ
[5] V. Telnov, arXiv: 1203.6563 [hep-ex] sin 2 cos4 2θ sin 2 cos2 2θ
[6] H. Ohkuma, EPAC 08, 36, and references therein. Luminosity monitoring with Bhabha (Møller)
[7] U. Wienands, PAC 05, 149 scattering BS is used to monitor or calibrate
[8] N. Iida et al, PAC 09, 2769 luminosity by selecting events with two oppo-
[9] Y. Funakoshi et al, IPAC 10, 2372 sitely charged tracks with high momentum (small
pt elastic) and small acollinearity angle that orig-
inate at the IP. Various contaminations such as
3.3.3 Bhabha Scattering (e+e− → e+e− ) muon pairs are statistically subtracted.
J.E. Spencer, SLAC A quite good approximation for the singles
rate (or pair coincidence rate) for an annular de-
The elastic scattering of electrons and positrons tector subtending an angular range δθ (θ1 to θ2 )
proceeds via one boson exchange and annihilation near the IP – ignoring field effects from the detec-
diagrams (the u and t channels for e−e− → e−e− ): tor or beam optics is
θ1
4πα2h̄2 c2 1 θ
γ,Z,... σ(δθ) ≈ + 4 ln(sin )
γ,Z,... s sin2 ( θ2 ) 2
θ2
(5)
273
For small angles, A detector with area Lx Ly might cover the region
from x 1 = X/R12 to x 2 = (X + Lx )/R12 and
250 GeV 2 θ22 − θ12
σ(δθ) ≈ 4.2 pb (6) from y1 = −Ly /(2R34 ) to y2 = Ly /(2R34 ). The
Eb θ12 θ22
(θ values of X (and/or Y ) and Lx,y are chosen to
The expected Bhabha rate is given by Lσ(δθ)(θ21 . avoid the incoming and outgoing disrupted beams
If the predicted rate is reasonable, we can then and backgrounds.
consider the simulated backgrounds to estimate Radiative Bhabha scattering (e+e− → e+e− γ)
the signal-to-noise ratio. Increasing θ2 is usually
not effective. k
It is fair to ignore the solenoid field if there pe pe’
is no crossing angle between the colliding beams, q
nor strong disruption from beam-beam effects. To

account for beam-beam effects, solenoid fields,
finite beam size and divergence effects there are
Monte Carlo codes (e.g. ABEL, BHLUMI and A free, charged particle radiates when scattered
others) that can be used to simulate the beams and or accelerated: “it never happens that a scatter-
their interactions at the IP for use with tracking ing event is unaccompanied by the emission of
codes such as MAD or TRANSPORT for optics quanta” [3]. Pure Bhabha scattering occurs at ex-
design and electromagnetic shower codes such as tremely small and presumably unresolvable an-
GEANT or EGS4 for detector design and back- gles for leptons. For the RBS photons, θk ∼
ground simulation. (See also Sec.3.3.9.4.) O(1/γ). For the radiating lepton,
A semi-analytical way to use such codes is 11−w 1 Ek
to transport the generated events (for the assumed θl ∼ ∼ (9)
γ w γ Eb − Ek
conditions and perturbed by the beam-beam inter-
action, Sec.2.5) outside the primary detector. The where w = Pe /Pe ≈ Ee /Eb + O(1/γ 2 ). The
resulting rate of scattered pair events and singles RBS differential cross section in lowest order in
is then a function of the effective, disrupted lu- the ultrarelativistic limit for the outgoing lepton is
minosity, the acceptance of the detector and the [4]
intervening beam optics. A standard approxima- dσe− ± ∓
tion, for the rate in a rectangular detector outside (e e → e±e∓ γ) (10)
dw
a solenoid, is
4αre2
QED 16α2h̄2 c2 dΩ s
1 + w2 − 23 w ln 2
w
−
1
RBS =L (7) =
1−w me 1 − w 2
s θ4

16α2h̄2 c2 dx dy and for the photons [4], with y = 1−w = Ek /Eb ,
=L 2 2 2
s x y (x + y ) dσγ ± ∓
(e e → e±e∓ γ) (11)
where θ = x 2 + y 2 , tanφ = y /x , and x , y dy
are the angles in cartesian coordinates.
4αre2 2 4 s 1−y 1
For a transverse point source and dispersion = y + 3 (1 − y) ln 2 −
free transport from the IP to the detector, the po- y me y 2
sition of a particle at the detector is xd = R12 x where 4αre2 = 2.32 mb. Using a finite bunch size
and yd = R34 y where R12 and R34 can be calcu- correction [5], a good approximation expected for
lated with an optics code, (Rij = first order trans- the radiated photon power for some bandwidth δy
port matrix elements, Sec.2.2.1). Substituting the into a forward detector is
BS detector’s coordinates in the optical reference
4 2 σx σy
system gives the corresponding integration limits Pγ ≈ LEb αr 2 + ln
3 e λc (σx + σy )
to compute the expected rates,
3 ymax
4α2h̄2 c2 1 × y − 2y + 4y y
2
(12)
QED
=L −
min
RBS (8)
s xy where λc = re /α and σ is the rms beam size.
(
y2 (x2 The value ymin →0 is never required in the
1 1 −1 y (
− + tan ( above expressions because of cuts or detector res-
x 2 y 2 x y ( olution.
1 x1
274
Some conclusions are:

• Elastic BS decreases inversely with s =
4Eb2 .
• RBS effects typically increase logarithmi-
cally with s/m2 . The relativistic cross section for free electrons
• Photons are emitted within an angle θγ = was given by Klein and Nishina [2]. Including
m/Eb with the particles. helicities, σe,e = ±1/2, λγ,γ = ±1, in the CM
√
• Their spectrum will run up to ∼Eb and vary for s ≡ (p1 + p2 )cme c2 gives
QED
as 1/Eb . dσBorn
(e±γ → e± γ ) (1)
• Hard scattering corresponds to angles out- dΩ
%
2
α h̄ c2 2 −2 θ 1
cos 2 σe,e = ± 2 , λγ,γ = ±1
side the m/Eb cone. =
• Typical beam divergence angles and lumi- s cos2 2θ σe,e = ± 21 , λγ,γ = ∓1
nosity monitors are θγ . CS of polarized laser photons on electron beams is
used to measure e-beam polarization and provide
• Virtually all scattered particles that are de- higher energy polarized photons (Sec.7.6.1).
tected are hard. Classical Compton scattering (e± +ω→e± +γ)
• Initial state radiation cannot be explicitly dis- CS as a frequency upshifting technique occurs in
tinguished. This is true for the final state many ways, e.g. in wigglers, undulators and FELs
when the radiative event is untagged. This where the initial photons are the low frequency,
implies the outgoing scattered particles are virtual components of static fields. Single-photon
not generally collinear. It also implies that CS (order re2 ) in the laboratory is
the cm and lab systems are no longer the
dσ 2πre2 1 4y
same. = (1 − y) + −
dy x (1 − y) x(1 − y)
• Beam-beam disruption will broaden the
4y 2
acollinearity distribution. + 2 + 2σ λ
e,e γ,γ FP (2)
x (1 − y 2 )
• Most experiments based on BS need to in- with FP in terms of relativistic
clude radiative effects. invariants xand y
y 2y
FP = (2 − y) 1 −
Electrodynamics suggests other radiative effects (1 − y) x(1 − y)
that need to be considered for BS in an external and x and y in terms of 4-vectors p and k
field or non-free environment. A more complete s 2p · k Eb ω
discussion is available [6] including comparisons x= −1= = 2 2 (1 − β cos θ)
p·p p·p m
to data, ranges of applicability of expressions and
≤ 0.0153 Eb [GeV] ω[eV] (3)
the importance of such radiative effects.
θ = 0 defines head-on scattering in the lab,
u 2p · k
References y =1− = (4)
p·p p·p
[1] H.J. Bhabha, Proc. Roy. Soc. 154 (1935) 195 2Eb ω ω
(1 − β cos θ ) = + O(θ )
2
[2] C. Møller, Ann. d. Physik 14 (1932) 568 = 2
m Eb
[3] J. Schwinger, PR 75 (1949) 898
[4] G. Altarelli, F. Bucella, Nuovo Cim. 34 (1964) 4πre2 ≈ 1 barn. Integration gives the total cross
section σC = σNP +2σe,e λγ,γ σP where
1337; G. Altarelli, B. Stella, Lett. Nuovo Cim. 9
(1974) 416 2πre2 4 8 1 8
σNP = 1− − 2 ln(1 + x) + +
[5] A.E. Blinov et al, PL B113 (1982) 423 x x x 2 x
[6] J.E. Spencer, SLAC-ARDB Rept. (1997)
1 2πre2 1
− → ln x + (5)
2(1 + x)2 x 2

3.3.4 Compton Scattering (e±γ → e±γ) 2πre2 2 5 1
J.E. Spencer, SLAC σP = 1+ ln(1 + x) − +
x x 2 1+x

Compton scattering (CS) [1] is a two-body, elastic 1 2πre2 5
process proceeding, in lowest order, via − → (ln x − ) (6)
2(1 + x)2 x 2
275
The arrows give the ultrarelativistic case x 1.

For an electron at rest in the lab (x = 2ω/m), σNP
and σP both fall rapidly compared to the Thomson
2
limit for x > 1. For x 1, σNP = 8πr 3 (1 − x),
e
the Thomson result. The total cross section typ-

ically depends less on the polarization than does
the outgoing photon energy distribution. σP = 0
for ln x ≈ 2.5, but for y = 0.7 near ymax =
x/(1 + x), the spectrum can differ by a factor of

6. The relation between incoming and outgoing
photons is:
ω2 1 − β cos θ1
= (7)
ω1 1 − β cos θ2 + ω11 (1 − cos φ)
where θ1 , θ2 and φ are the angles between the vec-
p1 ,k1 ), (
tors ( p1 ,k2 ), and (k1 ,k2 ) and 1 =Eb . Fig.1
shows some characteristic distributions.
Multiphoton Compton scattering (e±+nω →
e± +γ) Intense laser fields modify the expressions
above. Multiphoton effects can appear in many
ways e.g. with a laser having nγ λ3 >1 where
nγ = I/(ch̄ω) is the number density of nearly
monochromatic photons. Using I = E 2 /377Ω
for 1 eV photons shows that present terawatt
tabletop lasers greatly exceed this. One effect is to
increase the outgoing photon energy. This regime
can be characterized by two dimensionless, classi-
cal and quantal strong-field invariants: Υ (Beam-
strahlung parameter, Sec.2.5.3.2) involving a par-
ticle’s Compton wavelength and η involving the
photon’s wavelength,
|(F pν )2 | 1 E∗
Figure 1: Unpolarized Compton distributions for the
μν 1 2
Υ = eλ̄C ≡ rms outgoing photons and electrons in the lab.
(p1 · p1 )2 Ec

e |Aμ Aμ | eErms m2e c4
η= = λ̄ = Υ (8) An electron in the wave field behaves as if it
me c2 me c2 2h̄ω1 1
has an effective mass
Aμ is the 4-vector potential of the incident field.
∗
Erms is the rms electric field in the electron rest m̄2e c4 = 2 − (p c)2 = m2e c4 (1 + η 2 ) (10)
frame. η =1 (Υ=1) corresponds to an energy gain
of one electron mass over one photon (Compton) This effect is identifiable by a shift in the kine-
reduced wavelength λ̄ (λ̄C ). Ec is called the “crit- matic edge for CS and the term η 2 me c2 /2 can be
ical field” of QED, identified as the “ponderomotive potential” in the
m2 c3 equivalent Hamiltonian for the problem. The ef-
Ec = = 1.3 × 1016 V/cm (9) fective mass of the electron in the strong field is
eh̄
taken into account by replacing pμ by the “quasi-
While it is not possible to create static, critical momentum” q μ (and pμ → q μ ),
fields in the lab frame, highly relativistic elec-
trons channeling through a crystal or electrons
η 2 m2 μ
and positrons at the IP of the next-generation q μ = pμ + k (11)
linear colliders may experience such fields from 2(k · p)
their opposing beams (Sec.2.5.3.2).
276
The kinematics of the scattering process are given This can also be integrated over y ≤ymax =x/(1 +
by q μ + nkμ = q μ + kμ . Here n is the number x). Using ω̄ in the expression for x and averag-
of absorbed photons and ing over incoming angles gives (see also Eq.(15),
1 Sec.4.6)
edge (n, η) = (12) ΔN 1
1 + 2n(k · q)/m̄2 ≈ nγ LσC (x̄) (15)
N 2
The fact that the kinematic edge decreases with
increasing n is essential in distinguishing elec-
References
trons scattered via n > 1 processes from ordinary
n = 1 CS. Nonlinear CS has been observed [3]. [1] A.H. Compton, PR 21 (1923) 715
The transition probability for nonlinear [2] O. Klein, Y. Nishina, Z. Physik 52 (1929) 853
Compton scattering is [4] [3] C. Bula, et al, PRL 76 (1996) 3116

∞ [4] N.B. Narozhnyi, A.I. Nikishov, V.I. Ritus, JETP
dW dWn 20 (1965) 622
=
du du [5] V.L. Telnov, NIM A260 (1987) 304; B. Dehning
n=1
∞ % et al, PL B249 (1990) 145
2πr02 4 u2
= − 2 Jn2 (z) + 2 +
n=1
u1 η 1+u
2 : 3.3.5 Limit of Focusing of Electron Beam due
× Jn−1 (z) + Jn+1 (z) − 2Jn (z)
2 2
to Synchrotron Radiation
where K. Oide, KEK

k · k 2k · q 2η u(un − u) Symbols Assume focusing in one plane (y).
u= ; u1 = ; z = Incoming rms invariant emittance γεy , focus-
k · q m2 u1 1 + η 2 ing strength and length of the final lens K =
and un = nu1 . An expansion, for small values of B /(Bρ), L. Courant-Snyder parameters at the IP
η, shows that the contributions from the nth order βy∗ , α∗ = 0. Distance between the FP and the fi-
multiphoton process scale as η 2n . nal lens ∗ βy∗ . Unperturbed orbit at the IP
Compton scattering on background radiation (y0∗ , y ∗0 ). Deviation of the orbit at the IP due to
Feenberg and Primakoff (1948) first considered radiation Y . Photon energy emitted in the lens,
the scattering of fast electrons on various sources u(s). Expected number of photons per unit length
of cosmic radiation. With the discovery of 2.7 K of the lens N (s). Fourier transform of photon
thermal background radiation, others considered spectrum S(P ).
how this restricted the range and energy of dif- Minimum rms spot size The rms spot size at
ferent particles that are observable in the vicin- the FP is [1]
ity of the earth. Later, Telnov and Dehning et 110
al. [5] considered how scattering from thermal σy∗2 = βy∗ εy + √ re λ̄e (1)
background photons can limit storage ring life- 3 6π
2 5/2
times and increase detector backgrounds. This is √ √ γ εy
× F ( KL, K∗ )
a straightforward application of above formulae βy∗
and the Planck distribution for spectral density of
thermal photons, The second term represents the blow-up due to en-
ergy fluctuation caused by synchrotron
√ radiation
√
dnγ ω2 1 in the final lens. Dimensionless F ( KL, K∗ )
= ω (13)
dω π 2 (h̄c)3 e( kT ) − 1 depends on the configuration of the focusing
assumed to be isotropic in space. Number density (Fig.1),
√ √
nγ = 20.2 T 3 cm−3 with average photon energy F ( KL, K∗ ) (2)
ω̄ = 2.7 kB T . The electron energy loss distribu- √KL ( √ (3
( (
tion for a beam of N particles over a distance L ≡ dφ (sin φ + K∗ cos φ(
is 0
φ 2 2
dN dσ dΩ √
= NL (1 + cos θ1 )dnγ (ω, T ) × sin φ + K∗ cos φ dφ
dy dy 4π 0
(14)
277
5αγ ∗
N (s) = √ |Kg(s)y 0 |
2 3
√
3( 1 + P 2 + P )5/3
S(P ) = √
10P 1 + P 2
√
3( 1 + P 2 − P )5/3
− √
10P 1 + P 2
3 ∗2
P = λ̄γ 2 ky 0 K 2 G(s)|g(s)|
2 √
sin Ks √
g(s) = √ + ∗ cos Ks
K
s
√ √
Figure 1: Function F ( KL, K∗ ). G(s) = g(s)2 ds
0
s
∗ u(s)
Y ≈ Ky0 G(s) ds
The minimum rms spot size 0 E

7
σy∗ min = (γεy )5/7 (3) References
5
1/7 [1] K. Oide, PRL 61 (1988) 1713
275 √ √
× √ re λ̄e F ( KL, K∗ ) [2] J. Irwin, Eq.(11.111), SLAC-474 (1996)
3 6π [3] K. Hirata, B. Zotter, K. Oide, PL B224 (1989) 437
is achieved when
2
275 √ √ 7 3 3.3.6 Thermal Outgassing and Beam
βy∗ = √ re λ̄e F ( KL, K∗ ) γ(γεy ) 7 Induced Desorption
3 6π
(4) V. Baglin, E. Mahner, CERN
In [1], the horizontal beam size σx was ig- O. Gröbner, ret. CERN
nored. A modified expression including a correc- Thermal outgassing rates [1, 2]
tion can be found in [2]. The additional contri- Q (Torr l s−1 cm−2 )
bution from σx can be made small by decreasing
the strength of the horizontally focusing magnet Q = C exp(−Eb /kB T )
in the final focus lens.
depends strongly on surface coverage C, tem-
Reduction of luminosity Since the resulting perature T and molecular binding energy Eb [3].
distribution of the beam differs from Gaussian Molecules may exist with different binding states
(unperturbed core, plus long tails), the luminos- between 0.8 and 2 eV [4]. A low outgassing rate
ity is better than that calculated from Eq.(3). The requires a chemical cleaning procedure adapted
minimum effective beam size determined by lu- to the material [5]. Coverage is also determined
minosity is typically 12 σy∗ min . by diffusion from the bulk, a particularly strong
Luminosity L is reduced from the value L0 effect for hydrogen in most metals [6]. For
without radiation in the lens by the factor [3] clean, unbaked Al, Cu and stainless steel (SS) sur-
faces after 100 h pumping at room temperature,

βy∗ εy ∞ the main gas species desorbed are fairly similar,
D=2 dk exp(−βy∗ εy k2 ) Tab.1.
π 0
The quoted rates may decrease by several or-
× |ρ̃Y (k)|2y∗ ,y ∗0 (5) ders of magnitude with pumping time as lower
0
binding energy states empty. In unbaked systems
where |ρ̃Y (k)|y∗ ,y ∗0 is the Fourier transform of
0 water dominates and Q remains sensitive to tem-
the distribution of Y , averaged over (y0∗ , y ∗0 ), and perature [7]. H2 O is found to decreases approxi-
L mately inversely with time
|ρ̃Y (k)| = exp N (s)(S(P ) − 1)ds QH2 O ∼Qo t−1
0
278
Table 1: Unbaked outgassing rates Q. 10-2
Desorption yield (molecules/photon)

150ºC Baked OFHC Copper
Gas Q (Torr l s−1 cm−2 ) -3
3.35 keV Critical Energy
10
H2 7 10−12
CH4 5 10−13
10-4
H2 O 3 10−10
CO 5 10−12 H2
10-5
CO2 5 10−13 CO
CO2
-6
10
Table 2: Outgassing rates (Torr l s−1 cm−2 ) for Cu, H2O
Al, and SS after a bakeout followed by 50 hours of 10-7

O2
pumping. CH4
10-8
Gas Cu Al and SS Ar
H2 1 10−12 5 10−13 10-9 19 20 21 22 23
CH4 5 10−15 5 10−15 10 10 10 10 10 1024
CO 1 10−14 1 10−14 Dose (photons/m)
CO2 5 10−15 1 10−14
Figure 1: Desorption yield for baked OFHC Cu at
3.35 keV critical energy versus dose.
For chemically cleaned and baked surfaces
(150◦ C for Al and Cu, 300◦ C for SS), and fol- step process, whereby first photoelectrons are pro-
lowing 50 h pumping, outgassing rates are dom- duced, which excite molecules bound in near sur-
inated by H2 , Tab.2. Within practical limits, it is face layers. Subsequently, excited more weakly
preferable to increase temperature instead of ex- bound molecules desorb thermally. The thermal
tending baking time. H2 outgassing of SS below desorption may persist long after the direct photon
10−13 Torr l s−1 cm−2 can be obtained by firing irradiation has stopped. Conversely, it takes time
the steel at 950◦ C in a vacuum furnace. to re-establish a quasi-steady rate after a stop.
At cryogenic temperatures below 20 K, ther- As a function of dose D (photons/m)
mal outgassing vanishes for all species apart from
H2 and He. η = ηo D −α where α∼0.6 to 1 .
Photon induced gas desorption at room tem- As a function of Q (Torr l/m)
perature The desorption yield η = desorbed η = η0 exp(−Q/Q0 ) .
molecules/incident particle. The yield for photons
varies with gas species, photon energy, photon an- Values for baked OFHC Cu are listed in Tab.3.
gle of incidence, surface cleanliness and tempera- In a circular machine photons originate pre-
ture. With increasing photon dose, D, the surface dominantly in bending magnets. As function of
becomes cleaner and the initial ηo decreases at a critical energy, Ec , η scales roughly like η ∼Ecα
rate which may be different for different species, with α∼0.7 to 1.2 below 1 keV. Fig.2 has been
see Fig.1 for baked Cu [8]. D is expressed in compiled from different measurements for Cu and
photons/m of irradiated vacuum chamber since in Al over a wide range of Ec . At large D (>
an accelerator the photons impinge on part of the 1023 photons/m), the surface becomes very clean
vacuum chamber wall and the remaining surface and a pumping effect of ∼100 l/s/m for CO, O2
is irradiated by scattered and reflected photons. and CO2 has been measured while the surface re-
The main species desorbed are H2 , CH4 , CO and saturates with gas. Cu desorption yields of less
CO2 . Even with baked chambers, H2 O and O2 than 10−6 molecules/photon have been measured
are released as the surface carbon concentration at 11 keV critical energy with an accumulated
is reduced. For baked SS ηo is similar to Cu; for dose of 2 × 1025 photons/m [9].
Al it is about a factor 10 higher. For activated, Photon induced gas desorption at low temper-
TiZrV coated chambers ηo is 2 orders of mag- ature At 77 K and 4.2 K the same species H2 ,
nitude less compared to Cu. Molecular desorp- CH4 , CO and CO2 are desorbed as at RT. Figure 3
tion by photons is generally attributed to a two- shows η of Cu at 77 K, 10 mrad grazing incidence
279
Table 3: Values of η0 and Q0 for baked OFHC Cu.
H2 CH4 CO CO2
ηo (molecules/photon) 9.2 10−4 2.3 10−4 3.7 10−4 5.5 10−4
Qo (Torr l/m) 3.0 10−2 4.5 10−4 8.4 10−3 1.1 10−2
10-1
Desorption yield (molecules/photon)
LEP
DCI
EPA
10-2
10-3
INP
EPA (Cu)
10-4 INP (Cu)
DCI (Al)
LEP (Al)
10-5
101 102 103 104 105 106
Critical Energy (eV)
Figure 2: Desorption yield versus Ec compiled from

different experiments.
10-2 Figure 4: Recycling of weakly bound, physisorbed

molecules versus coverage.
H2
Yield (molecules/photon)
10-3
the photon induced recycling, κ, of physisorbed
molecules becomes important, particularly for
10-4 CO H2 , see Figure 4 [13]. Condensed molecules such
CO2
as CH4 and CO2 may also be cracked into H2 ,
CH4
CO and O2 by photons [14].
10-5 In a cold vacuum system external pumps are
inefficient since longitudinal gas flow is strongly
10-6
suppressed. In the LHC, molecules are effec-
tively pumped on the 1.9 K magnet cold bore.
1018 1019 1020 1021 1022 Molecules can reach the cold bore through per-
Dose (photons/m)
forations in the beam screen, i.e., a fraction, f , of
Figure 3: Desorption yield of Cu at 77 K, 10 mrad the specific surface area, F . The volume density
grazing incidence and Ec = 50 eV. n (molecules/cm3 ) and surface coverage Θ on the
screen held at 5 to 20 K interact through mutual
desorption and re-adsorption [15]
and Ec = 50 eV [10]. At 4.2 K η0 is reduced by
dn dΘ
about one order of magnitude compared to room V = q − an + bΘ and F = cn − bΘ .
temperature and α∼1/3 [11]. dt dt
At low temperature the surface is Here q = ηΓ is the desorption rate of strongly
‘sticky’ since the mean sojourn time bound molecules by the photon flux, Γ.
τ (s) ≈ 10−13 exp(Eb /kb T ) becomes very b = F/τ + κΓ describes physisorbed
long [12]. Thus, under photon irradiation, molecules, which desorb spontaneously after a so-
molecules desorb initially from a strongly journ time τ , as well as by photons.
bound state and accumulate in a weakly bound c = vm sF/4 gives the wall pumping with the
physi-sorbed state (Eb ∼20 to 100 meV). As the sticking probability s, and the mean velocity for
surface coverage, Θ (molecules/cm2 ) increases, molecules with mass m, vm = 8kb T /πm.
280
Figure 6: Ion induced desorption yield for unbaked SS

as a function of ion impact energy.
The yields for baked SS are about 5 times

lower. Since ions are produced from the resid-
ual gas, η is a net yield, as ions have a probability
Figure 5: Pressure response to synchrotron radiation
to be implanted in the surface. In a clean vacuum
in a perforated Cu-coated SS liner at 10-20 K inside a
chamber, the overall effect may become negative
cold bore at ∼3 K.
resulting in ‘beam-pumping’.
a = c + νm F f /4 represents the surface To improve surface cleanliness many cleaning
pumping and the pumping by the perforation methods (bakeout, Ar-O2 glow discharge, surface
in the beam screen. The equilibrium gas den- coating) can be used. S is ultimately limited by
4ηΓ the specific molecular conductance cm (m4 s−1 )
sity nequ = vm F f depends only on the desorp- of the vacuum pipe. In a room-temperature sys-
tion yield, which decreases gradually with pho- tem with regularly spaced, ‘lumped’ pumps (of
ton dose, and on the pumping through the beam speed 2S) at distance 2L the limit is
screen perforations. Θ and κ determine the time
response to reach the steady state. The pressure (ηI)crit < π 2 /4(e/σi )(cm /L2 ) .
response to synchrotron radiation in a test system
for the LHC is shown in Figure 5. Temporary For a uniform circular pipe cm = (2π/3)vm r 3 .
warming to 52 K can be used to remove the ph- The conductance limitation can be overcome
ysisorbed H2 layer. with linear, distributed pumping. With the large
Desorption by residual gas ions With a posi- molecular mass, CO becomes the most critical
tively charged beam of current I, residual gas ions gas. Since the residual gas is a mixture of H2 ,
are accelerated to the wall and cause ion induced CH4 , CO and CO2 the single gas stability limit
desorption [16]. The flux of desorbed molecules can be significantly reduced by the desorption of
one species by another. Thus an H2 dominated
Q(I) = ησi P I/e + Q0 .
gas composition will change to CO and the simple
For relativistic particles the ionisation cross- expression may overestimate the stability limit.
section σi ∼10−18 cm2 . With the thermal out- In a cold-bore vacuum system the stability
gassing Q0 and the linear pumping speed S limit is given by the wall-pumping (ηI)crit =
(l s−1 m−1 ) the dynamic pressure is (νm /4)(e/σi )sF .
P (I) = Q(I)/S = Q0 /(S − ησi I/e) . For the LHC and a molecular sticking proba-
The vacuum stability limit is given by (ηI)crit = bility s∼1, (ηI)crit ∼5 103 A. This effect is partly
Se/σi . During operation, when the product ηI offset by η which increases approx. proportionally
approaches the stability limit, the local pressure to the surface coverage; η > 103 molecules/ion at
increases drastically and ultimately leads to beam a monolayer of condensed H2 [18]. To guaran-
loss [17]. tee stability, a minimum pumping speed is pro-
η is a function of gas species, ion energy and vided by the perforations in the beam screen,
surface cleanliness. Measured yields for unbaked which ensures a coverage below a monolayer and
SS are shown in Fig.6. (ηI)crit ∼102 A.
281
Table 4: η for differently prepared targets bombarded with 1.5 109 Pb53+ ions at 14 mrad grazing incidence.
Material Vac. firing Treatment Coating Bakeout η [mol/ion]

SS (316LN) 950◦ C, 2h - - 300◦C, 24h 2.0 104
SS (316LN) 950◦ C, 2h Ar-O2 glow disch. - 350◦C, 24h 6.2 104
SS (316LN) 950◦ C, 2h - TiZrV (1.5 μm sput.) 200◦C, 24h 2.7 102
SS (316LN) 950◦ C, 2h - TiZrV (1.5 μm sput.) 300◦C, 24h 59
950◦ C, 2h 350◦C, 24h 2.9 104
SS (316LN) Ar-O2 glow disch. -

SS (316LN) 950◦ C, 2h 50 μm electropol. - 300◦C, 24h 1.0 104
SS (316LN) 950◦ C, 2h 150 μm electropol. - 400◦C, 24h 9.3 103
SS (316LN) 950◦ C, 2h - Au (30 μm galvanic) 300◦C, 24h 1.4 103

SS (316LN) 950◦ C, 2h 50 μm chem. pol. - 300◦C, 24h 6.5 103
SS (316LN) 950◦ C, 2h Liner: Si coated SS Si (0.4 μm evap.) 300◦C, 24h 3.5 104
SS (316LN) 950◦ C, 2h He-O2 glow disch. - 350◦C, 24h 1.7 104
SS (316LN) 950◦ C, 2h Liner: 127 μm Mo - 300◦C, 24h 2.4 104
foil
SS (316LN) 1050◦C, 2h 50 μm chem. pol. - 300◦C, 24h 6.2 103
SS (316LN) 1050◦C, 2h Vented after scrubb. - 300◦C, 24h 5.3 103
SS (316LN) 1050◦C, 2h - Ag (2 μm galvanic) 300◦C, 24h 1.2 103
SS (316LN) 1050◦C, 2h 50 μm chem. pol., - 300◦C, 24h 6.9 103
getter purified
SS (316LN) 950◦ C, 2h Liner: St707 - 350◦C, 24h 2.5 102
getter strips +400◦C, 3h
SS (316LN) 950◦ C, 2h 150 μm electropol. Pd (0.6 μm sput.) 300◦C, 24h 2.6 102
SS (304L) - - - 300◦C, 24h 3.2 103
Cu (OFE) - - - 150◦C, 24h 2.2 104
Al (6082) - - - 150◦C, 24h 2.1 104
Heavy-ion induced gas desorption Large pres- the in situ bakeout temperature (300◦ C or 400◦ C)
sure rises by several orders of magnitude can have no significant influence on η. Nitrogen vent-
be caused by lost beam ions that impact under ing of a heavy-ion scrubbed SS surface to atmo-
grazing angle onto vacuum chamber walls. The sphere results in the complete loss of the cleaning
heavy-ion induced gas desorption can seriously effect. Perpendicular lead ion impact results in
limit the ion intensity, luminosity and beam life- the lowest η, which increases for grazing angles in
time, as observed at CERN, GSI, and BNL [19] the mrad range. The dominant gases desorbed by
Measurements of η (molecules/ion) using lead ions are CO, CO2 , and H2 . A clear correla-
4.2 MeV/u lead ions from CERN LINAC3, bom- tion exists between the surface oxygen and carbon
barding differently prepared accelerator-type vac- content and the measured η.
uum chambers under 14 mrad grazing incidence Heavy ion-induced η, measured as a function
angle, are summarized in Tab.4 [20, 21]. The of the ion energy and comprising different types
desorption yield depends critically on the surface of ions, charge states, target types, materials, and
properties of the baked chambers. For bare SS η impact angles are summarized in Fig.7, where
varies between 3000 and 2 × 104 molecules/ion. molecular desorption yields derived from dedi-
Noble metal or getter coatings (Au, Ag, Pd, cated beam line experiments (HCX: K+ , LINAC
TiZrV) reduce η by a factor of up to 100. Glow 3: Pb53+ , HLI: Pb27+ , Zn10+ , Xe18+...21+ , TSL:
discharged (Ar-O2 , He-O2 ) SS, Al, Cu, and Mo Ar8+...12+ , HHT: Ar10+ , U73+ , SPS: In49+ )
have η ≥ 104 molecules/ion. The ex situ vacuum are compared with results obtained from ma-
firing temperature (950◦ C or 1050◦ C) of a SS sub- chine experiments (AGS: Au31+ , SIS 18: U28+ ,
strate, the thickness of a noble metal coating, and RHIC: U73+ , Cu29+ ). The yield scales with the
282
8
10
Grazing impact angle Perpend. impact angle Au
79+ [19] E. Mahner, PRST-AB 11, 104801 (2008)
BNL (AGS) BNL (RHIC)
10
7
BNL (RHIC) BNL (RHIC) [20] E. Mahner et al, PRST-AB 6, 013201 (2003)
BNL (RHIC) GSI (HLI)
10
6 CERN (LINAC 3) GSI (HLI) [21] E. Mahner et al, PRST-AB 8, 053201 (2005)
CERN (SPS) GSI (HLI)
GSI (SIS 18) GSI (HHT)
In
49+ [22] A. Molvik et al, PRL 98, 064801 (2007)
ηeff [molecules/ion]
LBNL (HCX) 53+

10
5 Pb GSI (HHT)
Au
79+
K
+ Au
31+ U
28+
Uppsala (TSL) [23] H. Kollmus et al, JVST A27 (2009) 245
10
4 [24] K. Wien, Radiation Effects and Defects in Solids
3
Pb
27+ U
73+ Au
79+
109 (1989) 137
10
18+...21+
Xe
29+
10
2
10+
Cu
Zn
8+...12+
Ar
10
1
Ar
10+
3.3.7 Photoemission and Secondary Emission
0
V. Baglin, CERN
10
0.01 0.1 1 10 100 1000 10000 100000

Ion energy [MeV/u]
3.3.7.1 Photoemission
Photoemission in a circular machine is mainly due
Figure 7: Heavy-ion induced desorption yields classi- to bending magnet radiation. Photoemission is
fied by experiments with either grazing or perpendicu- a 3-step phenomenon [1]. First, a photon is ab-
lar ion-impact angles [19].
sorbed inside the material over a few optical at-
tenuation lengths (∼5 nm for metalic surfaces at
electronic energy loss of the projectile [22] 10 mrad incidence angle). Second, an electron is
excited and diffuses through the solid while los-
dE n
η=k . ing energy and creating secondary electrons by
dxel inelastic collisions during the process. Third, the
For Ar and U ions (5–100 MeV/u), impacting electrons escape from the solid across the po-
on SS targets, one finds kAr = 1.7, kU = 0.9, tential barrier. The escape depth of the excited
nAr ≈ 2.1, and nU ≈ 2.9 [23]; n > 2 indicates a electrons is roughly independent of the materi-
thermally moderated desorption process [24]. als. Electrons with kinetic energies up to 1 keV
Mitigation techniques to overcome high η have escape depth of the order of 3-5 nm. Pho-
values were first applied in LEIR at CERN, which toemission is therefore surface sensitive. When
is equipped with NEG-coated SS vacuum cham- a photon irradiates a surface, a photoelectron can
bers and gold-coated absorbers to catch charge- be emitted if the photon energy, hν, is larger than
exchanged lead ions under perpendicular impact. the work function, WF , of the material. The ki-
netic energy of the emitted electrons varies from
References 0 to (hν − WF ). Most of the photoelectrons are
secondary electrons with low energies and only
[1] CAS, CERN 99-05 19 (1999) 0.1-1% have high energies. Fig.1 shows the en-
[2] CAS, CERN-2007-003 (2007) ergy distribution curves (EDC) of evaporated gold
[3] A.G. Mathewson et al, 7th Int. Vacuum Congress when irradiated by photon energies of 35 and
(1977) 1027 100 eV. Electrons originating from the valence
[4] M. Moraw, Vacuum 36, 7-9 (1986) 523 band and core levels can be identified from the
[5] CAS, CERN 94-01 v 2 (1994) 535
EDCs. X-ray photoelectron spectroscopy (XPS)
[6] M.D. Malev, Vacuum 23, 2 (1973) 43
is widely used in surface science to identify the
[7] A. Berman, Vacuum 47, 4 (1996) 327
components on the material’s surface.
[8] O. Gröbner et al, JVST A 12(3) (1994) 846
[9] K. Kanazawa et al, NIM A499 (2003) 66
When irradiated by synchrotron radiation, the
[10] R. Calder et al, JVST A 14(4) (1996) 2618 vacuum chamber is subjected to irradiation rang-
[11] V. Baglin et al, Vacuum 67 (2002) 421 ing from a few tenth of eV to a few keV. There-
[12] CAS, CERN-2004-008 (2004) 396 fore, the produced photoelectron spectrum is the
[13] V.V. Anashin et al, Vacuum 53, 1-2 (1999) 269 sum of the monochromatic EDCs in the range
[14] V.V. Anashin et al, Vacuum 60, 1-2 (2001) 15 of interest. This spectrum is dominated by sec-
[15] O. Gröbner, Vacuum 60, 1-2 (2001) 25 ondary electrons whose energy is below 10 eV.
[16] I.R. Collins et al, e+ e− Factories’99, KEK Pro- The photoelectron yield, PY∗ , which is the
ceedings 99-24 (2000) 88 ratio of photoelectrons per absorbed photon, is a
[17] O. Gröbner, R. Calder. PAC 73, (1973) 760 function of the material, its surface cleanliness, its
[18] J.C. Barnard et al, Vacuum 47, 4 (1996) 347 surface treatment, and of the incidence angle.
283
Table 1: Forward scattering photon reflection and photoelectron yield for materials subjected to synchrotron
radiation with 45 and 194 eV [3].
45 eV 194 eV
Material Status R (%) PY∗ (e/ph) R (%) PY∗ (e/ph)
Al unbaked - 0.11 - 0.32
Cu-smooth unbaked 81 0.11 77 0.32
Cu-electrodeposited unbaked 5 0.08 7 0.08
Cu-sawtooth unbaked 8 0.03 7 0.04

TiZr unbaked 20 0.06 17 0.08
TiZr activated at 350◦ C 20 0.02 17 0.03
Figure 2: EDCs of Cu at 10 K when bombarded by

electrons with energies of 11 and 312 eV [6].
Figure 1: Photoelectron spectra of gold when irradi-
ated by photon energies of 35 and 100 eV [2].
3.3.7.2 Secondary emission
Primary electrons impinging on a surface cause

Since the photoelectron yield measurements
the emission of secondary electrons. Secondary
of interest are performed at grazing angle on sam-
emission is a 3 steps process. First, a primary
ples of a few meters in order to mimic the machine
electron of energy EP penetrates into the solid
operation condition, the forward reflectivity of the
and generates secondary electrons. Second and
photons, R, must be taken into account to derive
third, secondary electrons diffuse trough the solid
PY∗ . Tab.x1 shows typical results measured in
and are evacuated into the vacuum [5]. The pene-
a photon beam line when a vacuum chamber is
tration depth of the primary electrons is described
subjected to synchrotron radiation at grazing an-
by the stopping power law. It increases with the
gle (11 mrad). For smooth Cu, at 4 keV critical
electron kinetic energy and is in the range of 1-
energy, R and PY∗ equals 33% and 0.43 e/ph.
5 nm for 200-500 eV electrons. The emitted elec-
When a “sawtooth structure” is impressed on trons are divided into 3 populations. The true
the horizontally outward side of the beam pipe secondary electrons have a kinetic energy of less
[3], the photoelectron yield and the forward re- than 50 eV. The elastically reflected primary elec-
flectivity are reduced to a few %. Photon condi- trons have a kinetic energy EP . The inelasti-
tioning also decreases PY∗ [3]. cally backscattered electrons have kinetic ener-
Since the photoelectrons spectrum is domi- gies from 50 eV to EP . Auger electrons with de-
nated by low energy electrons, a magnetic field fined peak energy can also be visible in the EDC.
of about 50 G parallel to the emitting surface can Fig.2 shows an example of two EDCs of Cu at
also be used to significantly reduce the photoelec- 10 K when bombarded by electrons with primary
tron emission [4]. energies of 11 and 312 eV. The ratio of reflected
In a cryogenic environment, the photoelec- electrons to the true secondary electrons decrease
tron yield is slightly reduced. The effect of con- with the increase of the kinetic energy of the pri-
densed gas is negligible up to 10 monolayers. mary electrons. Above 50 eV, this ratio is less
284
One mitigation of the electron cloud in accel-

erator is obtained by thin films coatings. These
coatings are applied onto the vacuum chamber
wall to reduce δ. TiN, activated TiZrV and un-
baked amorphous carbon coatings have respec-
tively a δmax of 1.5, 1.1 and 1.0. Beside a sur-
face treatment which reduces the oxide layer, the
surface morphology also plays an important role
and allows further reduction of the SEY below

unity. Production of macroscopic grooves is an-
other mean to reduce the SEY [7, 10].
In accelerator machines, external means to

Figure 3: SEY of some technical surfaces [7]. reduce and control the SEY such as magnetic
fields, clearing electrodes and beam conditioning
than 10% whereas the contribution of the reflected are also used. A 50 G magnetic field parallel to
electrons is dominant at very low energy. the electron emitting surface reduces the emission
The secondary electron yield (SEY), δ, is de- of the true secondary electrons, allowing the mit-
fined as the ratio of the number of produced elec- igation of the electron cloud by solenoid magnets
trons to the number of incident electrons. Fig.3 in field free regions. In superconducting cavity
shows the SEY curves for different technical sur- rf couplers, a bias voltage of 2.5 kV is often ap-
faces measured at perpendicular incidence. The plied on clearing electrodes to avoid multipact-
maximum of the SEY, (δmax ), is assumed at a cer- ing. Beam conditioning with electrons is used
tain value of the primary incident electron energy, to reduce the SEY of technical surfaces such as
called Emax , which typically is in the range 200- copper and stainless steel. During this process, a
300 eV. The SEY of technical surfaces can differ graphitization of the oxide carbon is observed and
greatly from an atomically clean surface due to δmax is typically reduced to 1.1-1.3 after a dose
the presence of a surface layer of oxide and con- of 10 mC/mm2 [11]. The conditioning efficiency
taminants: sputter cleaned Cu has a δmax of 1.3 of low energy electrons with less than 50 eV is
whereas as received Cu has > 2. The SEY of weaker than for higher energy electrons. Some
Cu is the same at room temperature and at cryo- beam conditioning is also observed under photon
genic temperature. However, the condensation of irradiation.
gas affects the SEY: 200 monolayers of water on
sputter cleaned Cu increase δmax to 2.3. References
The SEY curve can be described by the sum
[1] C.N. Berglund et al, PR 136, A1030 (1964)
of true secondary electrons and reflected electrons
[2] R. Cimino et al, PRST-AB 2, 063201 (1999)
[6, 8]:
[3] I.R. Collins et al, EPAC 98; Y. Suetsugu et al,
δ (EP ) = δtrue (EP ) + Rδelastic (EP ) JVST A 21 (1) (2003) 186
where [4] V.V. Anashin et al, Vacuum 60 (2001) 225
s × (EP /Emax ) [5] H. Bruining. Physics and applications of sec-
δtrue (EP ) = δmax ondary electron emission, Pergamon Press
s − 1 + (EP /Emax )s
(1954); O. Hachenberg et al, Adv. Electron Phys.
with s∼1.4. When decreasing the incidence an-
11 (1959) 413-499
gle, the SEY increases. Introducing the angle θ [6] R. Cimino et al, App.Surf.Sci. 235 (2004) 231
with respect to the surface normal, the parameters [7] N. Hilleret et al, EPAC 00, 217
δmax , Emax and s may vary like [8, 9]: [8] M.A. Furman. CERN LHC Project Report 180
δmax (θ) = δmax e0.4(1−cos θ) (1998) or LBNL-41482/CPB note 247; G. Ru-
molo et al, PRST-AB 4, 012801 (2001)
Emax (θ) = Emax × (1 + 0.7 (1 − cos θ)) [9] R.E. Kirby, F.K. King, NIM A469 (2001) 1
[10] I. Montero et al, App.Surf.Sci. 253 (2007) 7627;
s (θ) = s × (1 − 0.18 (1 − cos θ))
M. Pivi et al, NIM A551 (2005) 187
The azimuthal distribution of the true sec- [11] C. Scheuerlein et al, App.Surf.Sci. 202 (2002)
ondary electrons follows the Beer-Lambert law. 57; M. Nishiwaki et al, ECLOUD 07
The reflection of the elastic electrons is specular.
285
3.3.8 Ionization Processes where C is the √ ring circumference or beam-line

F. Zimmermann, CERN length, li ≈ 2bρ is the average path of photons
until they hit the chamber (ρ is the bending radius,
Cross section for collisional ionization is [1]
2 b the horizontal chamber half size), and n()d is
λ̄e 1 the number of quanta emitted per unit time per
σcoll ≈ 4π C1 +2C2 ln βγ −
β 2 electron with energies between and + d, eval-
(1) uated at location s. If only ions inside the rms
where C1 and C2 depend on gas properties, and beam size are of interest, replace b by σx in the
v = βc is the velocity of the ionizing charged argument of li .

particle. Assuming the latter is a singly charged For the lifetime of ion beams, ion-atom im-
e− or proton, for carbon monoxide C1 = 35 and pact ionization cross sections [9] are important.
C2 = 3.7, and σcoll ≈ 2 Mbarn, at highly rela-
tivistic energies (γ = 103 –106 ). The ion density Limits on ion accumulation [10, 11] Accumu-
per meter λion after the passage of N beam parti- lation of ions is limited by: (i) residual-gas den-
cles is (at 300◦ K) sity dgas and multiple ionization (dion ≤ dgas );
(ii) neutralization [dion ≤ nB NB /(2πσx σy C)];
λion [m−1 ] = σion ρm N ≈ 6 N Pgas [Torr] (2)
(iii) heating by beam-gas collisions; (iv) ion drifts
where Pgas gas pressure, and ρm molecular den-
or recombination with photoelectrons. Accumu-
sity. In future accelerators, tunneling ionization
lation limit in undulators is not clear [6, 11].
may be significant; the corresponding ionization
rate is approximately [2, 3]
References
α2 c Eion 4 α Eion
W ≈8 2 exp − (3)
λ̄e eE 3 λ̄e eE [1] F.F. Rieke, W. Prepejchal, PR A6 (1972) 1507
where E ≈ eλbeam /[2π0 (σx + σy )] the collec- [2] T.O. Raubenheimer, PAC 95, 2752
tive bunch field and Eion the ionization energy. [3] L.D. Landau, E.M. Lifshitz, Quantum Mechan-
Tunneling ionization thresholds measured in ex- ics: Non-Relativistic Theory, Pergamon (1981)
periments on beam-driven plasma wake-field ac- [4] C.L. O’Connell et al, PRST-AB 9, 101301
celeration [4] are consistent with first-order calcu- (2006)
lations of the “ADK” approximation for the field- [5] M.V. Ammosov et al, JETP 64 (1986) 1191
ionization rate [5]. [6] Y. Miyahara, Jpn. J. Appl. Phys. 26, no.9 (1987)
In e± storage rings, many ions are produced 1544
via photoionization [6]. Photoionization cross [7] E.L. Kasarev, E.R. Dodolyak, Opt. Spectrosk. 56
section for hydrogen is [7, 6] (1984) 643
[8] D.E. Cullen et al, UCRL–50400-Vol.6-Rev.4-
100 3.228
σpi ≈ 5.35 × 10−20 (cm2 ) Pt.A, DE90 010470 (1989)
[9] I.D. Kaganovich et al, NJP 8 (2006) 278
for > th ≈ 18 eV (4) [10] A. Poncet, Lecture Notes in Phys. 400, Springer
and for carbon monoxide Verlag (1994)
σpi≈9.12 × 10−14 −2.48 − 4.80 × 10−12 −4.05 [11] Y. Baconnier, A. Poncet, P. Tavares, CERN/PS
94-40 (1994)
(cm2 ) for > th ≈ 14 eV (5)
where is the photon energy and th the threshold
energy, both in eV. Photoionization cross sections
3.3.9 Beam Induced Detector Backgrounds
for many elements are tabulated in [8]. Inside
and Irradiation in e+ e− Colliders
the beam, collisional ionization produces about 3
S.D. Henderson, FNAL
times more ions than photoionization. Most pho-
M. Sullivan, SLAC
toionization processes occur between beam ra-
dius and vacuum chamber. The total number of Consideration of detector backgrounds and irra-
ions generated by photoionization is typically 2-4 diation is important in modern e+ e− colliders
times that produced by collisional ionization. Av- which require unprecedented colliding beam cur-
erage ion production rate (ions produced per sec- rents and precision charged particle tracking de-
ond and per electron) due to photoionization is tectors near the IP [1, 2, 3, 4, 5, 6, 7, 8]. High
C ∞
1 current e− p colliders also fall into this category.
= li ρm ds σi ()n(, s) d (6)
τpi 0 th
A masking system is incorporated to shield the
286
detector from beam-generated background. Col- where N is the gas density and σT = σTCoul +
limators are required upstream of the IP to re- σTBrem is the total cross-section folded with the
move errant beam particles or to shield the detec- transport efficiency from source to target. If the
tor from synchrotron radiation produced by parti- transport efficiency from source to target is 100%
cles at high transverse amplitudes. The degree of [compare Eq.(6), Sec.3.3.3],
background suppression is set by detector track-
0.2606 Z 2 1 1 1
ing and trigger requirements, and by component σTCoul [barns] = 2 −
lifetime. Interaction region shielding designs and E [GeV2 ] θ22 θ12 [mrad2 ]
operational performance for the B-factories are (2)

Brem 183
discussed in [9] for PEP-II, [10] for KEK-B and σT [barns] = 0.00309 Z(Z + 1) ln 1/3
[11] for CESR. The BEPCII e+ e− collider is de- Z
δ1 3 2
scribed in [12], and the Super B-factory designs in × ln + (δ1 − δ2 ) + (δ1 − δ2 ) 2

(3)
[13] for the Italian SuperB and in [14] for Super- δ2 8
KEKB. Finally,
3.3.9.1 Sources of detector backgrounds dN [particle/s]
= 202 σT [barn] I[A] P [nTorr]
ds[cm]
Detector backgrounds arise from the following (4)
processes: (i) the interaction of beam particles The power striking the target region is
with residual gas (beam-gas backgrounds), (ii)
dP [GeV/s] = 202 I[A] P [nTorr] ds[cm]E0 [GeV]
synchrotron radiation (SR) generated in nearby δ2
magnets, (iii) Touschek scattering, (iv) beam- dσ Brem
× σTCoul [barn] + (δ + 1) (δ) dδ
beam interactions, (v) interaction with thermal δ1 dδ
photons, and (vi) operational particle losses due to (5)
injection, tuning, beam loss, etc. In general, (i)- where (δ) is the transport efficiency. It should be
(v) may be estimated through simulation, while emphasized that the residual gas pressure has both
(vi) is difficult to estimate. a static and current dependent component.
Beam-gas interactions The most important Coulomb scattering (Sec.3.3.1) Consider an
beam-gas interactions are Coulomb scattering and elliptical vacuum chamber having walls at ±H(s)
bremsstrahlung. These interactions produce beam in x and ±V (s) in y. A Coulomb scattering at
particles with large transverse oscillation ampli- s0 with scattering angle θ and azimuthal angle φ
tudes that may follow trajectories which strike the gives a trajectory x(s) = xc (s) + Sx θ cos φ and
vacuum chamber near the IR, generating shower y(s) = yc (s)+Sy θ sin φ where xc (s), yc (s) is the
debris which reaches the detector. closed orbit, and Sx = βx (s)βx (s0 ) sin ψx (s −
Lost particle and energy fluxes We estimate s0 ). The scattering angle required to strike the
the particle flux in a “target” region (the inter- vacuum chamber at s (where a collimator or mask
action region mask) extending from s1 to s2 , is located) is given by solution of
due to beam-gas interactions at s0 (s0 < s1 <
s2 ). Let the Coulomb scattering angle required V 2 (s)[xc (s) + Sx θ cos φ]2 (6)
to just strike the aperture at s1 be θ1 and that re- +H 2 (s)[yc (s) + Sy θ sin φ]2 = H 2 (s)V 2 (s)
quired to just strike at s2 be θ2 (θ1 > θ2 ). For (The trajectory may strike the vacuum chamber
bremsstrahlung, the electron energy loss, δ = upstream of s.) Scatterers π/2, 3π/2, · · · in beta-
(E − E0 )/E0 , required to strike the aperture at tron phase from the IP may be regions with large
s1 is δ1 , and that required to strike at s2 is δ2 , Coulomb contribution.
where −1 < δ1 < δ2 < 0. These limits may Bremsstrahlung (Sec.3.3.1) results in an off-
be easily determined with a particle tracking code energy particle (with energy error δ) and a photon
that includes full off-momentum beam transport (with energy |δ|E). The scattering angle of order
capability. Typical limits for the energy loss range 1/γ may generally be neglected.
from many times the beam energy spread to sev- Scatters located near π, 2π, · · · in horizontal
eral percent. The flux of particles striking the “tar- betatron phase from the IP may be regions with
get” region from a source region between s0 and large bremsstrahlung contribution. Additionally,
s0 + ds is large amplitudes result when the source or target
dN (s0 ) = Ibeam σT (s0 ) N (s0 ) ds (1) are in a region of high dispersion.
287
A bremsstrahlung interaction at s0 gives cross sections must be used. The Rayleigh scat-

x(s) = xc (s) + δ[D(s) − Cx D(s0 ) − Sx D (s0 )], tering cross-section is given by
where Cx = βx (s)/βx (s0 )[cos ψx (s − s0 ) + dσR /dΩ = (dσT /dΩ)[F (x, Z)]2
αx (s0 ) sin ψx (s − s0 )]. The vertical amplitude
where F (x, Z) is the atomic form-factor [15] as
remains unchanged (assuming no y dispersion).
a function of momentum transfer x, and atomic
The particle strikes the chamber wall at s when
number Z, and σT is the Thomson cross-section.
x2 (s) = H 2 (s). The energy loss required to
The Compton scattering cross-section is similarly
strike the chamber at s is
modified,
±H(s) − xc (s) dσC /dΩ = (dσKN /dΩ)[S(x, Z)]
δ± (s) = (7)
D(s) − Cx D(s0 ) − Sx D (s0 )
where S(x, Z) is the incoherent scattering func-
where + is for an “outside” hit and − for an “in- tion [16], and σKN is the Klein-Nishina cross-
side.” (There is only a single physical solution section (Sec.3.3.1). Total cross-sections are tab-
since −1 < δ < 0.) ulated in [17]. The photoabsorption cross-section
Eq.(7) is valid in linear approximation only. goes as Z 5 /E 7/2 , forward Rayleigh as Z 2 , and
For δ <∼ −0.01, one has to include higher order the Compton as Z. Therefore, the ratio of total
chromatic effects. Such a treatment is easily per- absorbed to scattered radiation goes as ∼ Z 3 .
formed with first order matrix multiplication, tak- The critical energy kc of the photon energy
ing into account the energy loss by proper adjust- spectrum of a bending magnet is
ment of magnet strengths. kc [keV] = 0.666E 2 [GeV]B[T]
Synchrotron radiation backgrounds SR Although half of the total energy of the photon
(Sec.3.1) generated in bends or quads (due to spectrum is above this number only 8.7% of the
finite beam size or displaced trajectories) near total photons are above this energy where the pho-
the IP strikes the vacuum chamber and detector ton flux is falling exponentially. The photon spec-
shielding and may be scattered into the detector trum from focusing elements is more complicated
beam pipe by any of the usual x-ray scattering but the spectrum generally falls exponentially due
mechanisms (Rayleigh scattering, Compton to most of the critical energy values being small
scattering, or x-ray fluorescence; Secs.3.3.1, and because the beam particle density is falling
3.3.4). In addition, x-rays generated from beam off rapidly as the transvese position of the beam
particles at large amplitudes may directly strike particle increases.
the detector beam pipe. Some of the SR backgrounds arise from scat-
Fig.1 shows the various transport mechanisms tering mechanisms in the extreme forward (θ < ∼
that give rise to SR backgrounds and dose. The 5◦ ), or backward (θ ∼ > 175◦ ) directions. F (x, Z)
mechanisms are grouped into three categories: (i) has a maximum at x = 0 (forward scattering),
direct SR, (ii) backscattering either from the near- where F (0, Z) = Z, and decreases to zero for
est masking surfaces or from downstream vacuum large x (backscattering of high-energy x-rays).
chamber surfaces, and (iii) forward scattering ei- S(0, Z) = 0, and increases to S Z for
ther from mask tips (tip scattering) or from the large x. As a result, forward scattering is dom-
upstream vacuum chamber surfaces. inated by Rayleigh scattering. Backscattering of
To calculate SR fluxes, we need to know: (i) low-energy x-rays is dominated by Rayleigh scat-
the number of photons emitted in path length ds, tering, and backscattering of high-energy x-rays
is dominated by Compton scattering. The two
dn photons 5 αγ E[GeV]
= √ = 1.29×1020 backscattering cross-sections are approximately
ds m−A−s 2 3 ρe ρ[m] equal for Eγ 20 keV in Cu, and Eγ 40 keV
(8) in Ta. The function S is always less than Z, so
and (ii) the normalized x-ray spectrum that the real bound electron Compton scattering is
(Sec.3.1.3), always less than the KN formula. In the backward
direction, S = 0.9Z at 30 keV in Cu, and 60 keV
1 dn 8 1 E
= √ S (9) in Ta.
n dE 15 3 E Ec
Photoabsorption followed by x-ray fluores-
For estimates of SR scattering in the range of cence contributes to both forward and back scat-
interest here, Eγ <∼ 100 keV, realistic scattering tering, since photoemission is isotropic. K-shell
288
3411297-006
Quadrupole
Detector
Dipole
Masks
Beam
Beam
Vacuum
Chamber Profile
Wall
Tip Scattering
Back Scattering
Direct SR
Upstream Bounce - in
Figure 1: Synchrotron radiation scattering processes.
Table 1: K-shell photoelectric data. see this background due to the outgoing beams
Element K edge μphoto /ρ ωK EK having a first bend in an offset trajectory in the
keV cm2 /g keV shared final focus magnet that is nearly 2 m from
Al 1.56 4890 0.039 1.49 the IP. A comparison of B-factory IR designs
Cu 8.98 300 0.440 8.14 can be found in [21]. Both very high luminosity
Ag 25.51 56.4 0.831 22.59 Super-B factory designs must keep beam bending
W 69.53 10.8 0.958 60.94 in the IR to a minimum in order to minimize this
Au 80.73 8.46 0.964 70.69 background source. Another QED process that
becomes important at high luminosity is e+ e−
pair production. These low energy particles are
edge energies, total photoelectric mass attenua- generally wrapped up by the detector magnetic
tion coefficients, fluorescence yields ωK [18] and field, but the process can set a lower limit on the
weighted K x-ray energies are given for a few el- beam pipe radius at the collision point.
ements in Tab.1 [17]. The L-shell emission must Beamstrahlung At proposed highest-energy
be taken into account for high-Z materials. The circular e+ e− colliders (e.g. LEP3, TLEP, Su-
L-shell fluorescence yields are much smaller; for perTRISTAN) beamstrahlung (Sec.2.5.3), gener-
Ta they are ∼0.25. Absorption in the L edge can ating off-energy particles, may become an impor-
be an effective technique for reducing characteris- tant background source [22].
tic x-ray emission. If the K-shell energy is above Touschek scattering Touschek scattering is a
the critical energy of the photon spectrum the ma- source of off-energy beam particles arising from
terial is considered a good absorber of the incident the elastic scattering of particles within a bunch
spectrum. (Sec.2.4.12) Scattering results in two particles
For tip scattering, only interactions within a (with energy errors +δ and −δ) which follow
few λmfp of the tip contribute to the scattering; cosine-like betatron trajectories about the off-
tip scattering rates are dominated by x-rays which energy closed orbit. General techniques for
strike within ∼10 μm of the mask tip. shielding off-momentum particles generated by
Beam-beam interactions Detector “back- bremsstrahlung are applicable to the Touschek
grounds” may arise during e+ e− collisions from background. The Touschek lifetime scales as γ 2 .
ordinary QED processes. These scale with lumi- So it is a concern in low-energy colliders such as
nosity rather than beam current. For example, DAΦNE [23] and CESR-C [24], where a set of
radiative Bhabha scattering (Sec.3.3.3) was an movable collimators is (was) installed well away
important process at PEP-II due to the presence from the detector. New collider designs and new
of the dipole located 20 cm from the IP [19, 20]. light source rings use very low emittance beams
BELLE, the KEKB detector, essentially did not to achieve desired performance, so that Touschek
289
lifetimes and backgrounds have become impor- The following are radiation dose limits for com-
tant design criteria. mon detector/IR components:
Thermal photon compton scattering An in- Crystal Scintillators CsI, CsI(Tl): 1-100
elastic scattering process resulting in off-energy krad, depending on dopant, impurities, manufac-
beam particles (Sec.3.3.4), this is important at turer [32, 33, 34]. For other crystals, see [35].
very high energy colliders (LEP, LEP2) [25, 26]. Drift Chamber: 0.5 C/cm accumulated on
sense wire [36, 37, 38]
Injection losses Radiation during injection can
Silicon Detectors: ∼1 Mrad (e− , γ) [39, 40]
arise from either the injected charge or from the

stored beam. The injected charge may gener- Readout Electronics: See Sec.8.7; CAMEX
ate radiation that could be more effectively trans- chip (CMOS): ∼25 krad powered (∼100 krad
ported to the detector by virtue of the large oscil- unpowered) [41]; JAMEX chip (JFET): ∼170
lation amplitude. The manipulation of the stored krad powered (∼200 krad unpowered) [42]; SVX
beam may generate SR or lost particle back- chip: ∼20 krad [43]; Rad hard processes: see
grounds. Experience at CESR with two different [44, 45, 46].
IR configurations has shown that radiation during Permanent Magnet Material: SmCo, ∼104
injection accounted for ∼50% of the total IR radi- Mrad. NdFeB, ∼50 Mrad (e− , γ), 5 × 1014
ation dose for CESR Phase I operation and ∼20% n/cm2 [47, 48, 49]. See also Sec.6.4.
for Phase II [5, 7]. Radiation budget Whereas the detector occu-
Continous injection Both B-factories devel- pancy rates are relevant during normal high-
oped methods of continously injecting charge into energy physics (HEP) data-taking, the detector ra-
the storage rings while the detector is opera- diation dose that determines the lifetime of com-
tional [27, 28]. The injection rates were usually ponents includes all sources. The total accumu-
3-10 Hz depending on stored beam lifetime. In- lated dose is the sum of HEP, injection, machine
jecting while the detector was taking data was studies, accelerator startup (initial vacuum system
achieved by careful study of the detector back- processing), beam-loss events, etc. CESR Phase
grounds on a pulse by pulse basis looking for II measured the following: HEP (63%), injection
corelations between large background pulses and (17%), machine studies (12%), other (8%) [5, 7].
errant injection bunches. Light source facilities
now have adopted continuous injection (or top- 3.3.9.3 Detector background shielding
up) as a requisite of the accelerator design. A Synchrotron radiation shielding [20, 50, 51]
nice summary of light source top-up was given by The detector is shielded from SR by the placement
Ohkuma [29] (also see Sec.3.3.2.2). of a mask tip near the detector beam pipe. This
Ion trapping Electron beam destabilization due mask tip is placed to intercept the SR fan from the
to trapped ions (Sec.2.4.13) leads to poor stored nearby dipole which would ordinarily sweep di-
e− lifetimes and may produce sudden large de- rectly across the detector beam pipe, as well as the
tector background rates, depending on the source radiation generated in IR quads. In high current
location. B-factory designs incorporate a clear- machines, a thin metallic coating (usually high-Z)
ing gap in the bunch train to reduce this ef- is applied to the beam pipe for further reduction of
fect [30]. Electron-cloud effects (Sec.2.4.14) on the SR flux entering the detector.
the positron beam usually affect the luminosity of The placement and size of the innermost
the collider before the detector sees an increase mask tip (or tips for asymmetric machines) de-
in background levels. However, in the p-p col- termines the effectiveness of SR shielding, and
lider LHC, electron-cloud build up leads to higher the ultimate SR background rates that will be
detector background through increased vacuum achieved, once the accelerator lattice is fixed. In
pressure [31]. addition to respecting the beam stay clear criteria,
the following are general guidelines for choosing
3.3.9.2 Detector and IR radiation tolerance the mask tip radius and position considering SR
and budget shielding (a similar analysis is required for beam-
gas):
A detector background budget is based on three 1. The mask tip should be placed so that no di-
factors: (i) detector component radiation sensitiv- rect radiation within ±10σx , ±10σy may strike
ity, (ii) detector occupancy, and (iii) trigger rates. the beam pipe.
290
2. Placement of the mask tip in s is a tradeoff 2. Collimators may be placed upstream of the
between two factors: i) moving the tip closer IP to remove scattered beam particles. In cases
to the IP decreases the portion of upstream vac- where lost particle trajectories pass outside of
uum chamber visible by the beam pipe, thereby the beam stay clear envelope it may be possible
protecting the detector beam pipe from scat- to significantly reduce the flux by placement
tered radiation upstream, and ii) tip scattering of a collimator at points of high D, β. This
and backscattering rates from the mask increase technique is not possible for trajectories which
rapidly as the mask tip is moved closer to the remain within the beam-stay clear envelope,
IP. These two concerns must be weighed, and however. The PEP-II design calls for graded
a general guideline is to place the tip as close apertures, becoming progressively larger to-
to the IP as acceptable background levels will ward the IP [20]. Likewise, the DAΦNE de-
allow. sign [23] calls for large tungsten beam scrap-

3. The mask tip radial position should be made ers upstream of each IR to be inserted to 10σH
small enough to shield the beam pipe from di- reducing the flux of off-momentum particles
rect SR, but not so small that it intercepts SR striking the KLOE IR by a factor of ∼200. LEP
flux that would otherwise pass through the IR. used a system of distributed adjustable tungsten
collimators to remove off-energy beam parti-
4. No surface within ∼3 m of the IP (for a B-
cles in the arcs, and for detector shielding in
factory) on which SR lands should be visible by
the IR straights [8].
the detector beam pipe (except the one irredu-
cable backscattering surface). A consequence 3. The detector beam pipe and vertex detector
of this rule is that mask surfaces must either be are shielded with nearby high-Z masking ma-
sufficiently sloped so as to remain invisible to terial. The inner profile is designed by consid-
the central beam pipe, or smooth tapers must be ering SR shielding requirements.
stepped.
3.3.9.4 Detector background and radiation
A thin metallic coating is applied to the inner estimation
surface of the beam pipe [52] (Sec.6.15.2) reduc-
ing the x-ray flux entering the detector by 2-3 or- Detector background estimation is a three step
ders of magnitude (for a 0.3% X0 thick coating process: (i) production of background generating
of Au). The combination of masking material and particles, (ii) propagation of these particles to the
beam pipe coating must be well-matched to the detector region, and (iii) interaction and shower
incident SR spectrum, and the mask characteristic development in the detector and surrounding ma-
x-ray energies. Material selection has been stud- terial. Generally, only the first step may be at-
ied in detail for CESR Phase II [53, 7], showing tacked analytically; the remainder of the problem
that the specific choice of materials can produce is simulated with Monte-Carlo methods.
variations of 2 orders of magnitude in background Lost particle simulations The general program
rates. proceeds as follows. Coulomb, bremsstrahlung,
A moveable radiation shield was incorporated Touschek or Thermal Compton scattering events
at CESR to shield the detector from SR during are generated along the beam orbit with a uni-
non-HEP operations [54]. form scattering angle or energy loss probabil-
Sets of collimators were used at LEP to assure ity. Scattered particle trajectories are tracked to
that no “single-bounce” sources of SR from the the IP, and then trajectories striking within a few
arcs may reach the detector [8]. meters of the IP are collected and used as in-
put to a full detector Monte-Carlo code includ-
Lost particle shielding The detector is shielded ing the solenoidal detector field and IR mag-
from lost particles by: netic elements. The events are weighted by
1. Maintaining low pressure to reduce beam- their production probabilities and the gas pres-
gas scattering at its source. The source effec- sure at the point of scattering. The program
tiveness for beam gas scattering extends from DECAY TURTLE [55] can be used for generat-
the IP to 30-60 m upstream (in B-factories and ing and tracking of beam-gas scattered particles.
in super B designs). Special care is taken in this Detector specific Monte-Carlo codes (generally
region to reduce the pressure; Secs.3.3.6,5.10. GEANT-based [56]) are used for the charged par-
ticle/photon interactions and detector response.
291
SR simulations The general program proceeds Haba, e+ e− Factories 1999, KEK Proc. 99-24,
as follows. The beam is tracked through nearby p.237
magnetic elements and SR photons are generated [11] S. Henderson, D. Cinabro, in [10]
tangent to the trajectory, with a spectrum deter- [12] C. Zhang, Proc. of the 40th ICFA ABDW 2008;
mined by the local radius of curvature. Beamsize BEPCII group, BEPCII Design Report, IHEP
and x, x correlation are taken into account [50]. Proc. 2001
SR photon trajectories are tracked to vacuum [13] M. Biagini et al, arXiv:1009.6178v1 (2010)
chamber surfaces and collected. These are then [14] S. Hashimoto et al, KEK-REPORT-2004-4
used as input to Monte-Carlo scattering codes for [15] J.H. Hubbell, I. Øverbo, J.Phys.Chem.Ref.Data
propagation of the x-rays from the vacuum cham- 8 (1979) 69
ber surfaces to the detector beam pipe and into [16] J.H. Hubbell et al, J.Phys.Chem.Ref.Data 4
(1975) 471
the active volume of the detector. There is little

[17] E. Storm, H.I Israel, LA-3753 (1967)
uniformity in SR background calculation codes.
[18] M.O. Krause, J.Phys.Chem.Ref.Data 8 (1979)
A package of codes for generating (SRGEN) and 307
scattering (SRSIM) are in use at CESR [50]. The [19] D. Kirkby, in Ref.[6]
scattering code is a single-scattering Monte-Carlo [20] Asymmetric B-factory Conceptual Design Re-
optimized for SR background calculations, and port, SLAC-372
includes all K and L shell fluorescence physics. [21] M. Sullivan, PAC97 (1997) 296
At PEP-II, QSRAD was used for photon gen- [22] V. Telnov, arXiv: 1203.6563 [hep-ex]
eration, and the interaction was simulated with [23] S. Guiducci, in Ref.[6]
EGS [57]. Likewise, “home-grown” SR simula- [24] R.A. Briere et al, Cornell CLNS 01/1742
tion codes were in use at LEP [8]. [25] B. Dehning et al, PL B2349 (1990) 145
Comparison with measurements Since the [26] V.I. Telnov, NIM A260 (1987) 304
background calculations are quite complex, com- [27] J. Turner et al, EPAC 04 Lucerne
[28] Y. Funakoshi et al, EPAC 04
parison with measurements of detector back-
[29] H. Ohkuma, EPAC 08; J. Corbett et al, IPAC 10;
grounds are extremely valuable. Iterations of cal-
C. Cristou et al, IPAC 10
culation and measurement are necessary. Detailed
[30] U. Wienands et al, EPAC 08
comparisons of detector background simulation [31] G. Bregliozzi, Report at LHC Background Study
predictions with measurements have been under- Group, 18 Oct. 2010; cern.ch/lbs
taken at several labs. Generally good agreement [32] C.L. Woody et al, IEEE TNS (1992) 524
was obtained at CESR/CLEO [4, 5, 7, 11], PEP- [33] Z. Wei, R. Zhu, NIM A326 (1993) 508
II/BaBar [9], KEK-B/Belle [10], and LEP [8]. [34] G. Bratta, H. Newman, R.Y. Zhu,
Ann.Rev.Nucl.Part.Sci. 1994, 14:453
References [35] S. Majewski, C. Zorn, Instrumentation in High
Energy Physics, World Scientific (1992) p.157
[1] H. DeStaebler, AIP Proc. 214 (1990) 59 [36] J. Kadyk, NIM A300 (1991) 436
[2] N. Toge, Proc. Int. Workshop on B-Factories: [37] J. Kadyk et al, IEEE TNS 37 (1999) 478
Accelerators and Experiments (1992) 44 [38] J. Va’vra, SLAC-PUB-5207
[3] S. Khan, B Factories, SLAC-400 (1992) 290 [39] E. Fretwurst et al, NIM A288 (1990) 1-12
[4] D. Cinabro, S. Henderson, H. Yamamoto, Int. [40] G. Hall, NIM A368 (1995) 199
Conf. High Energy Phys. (1994); D. Cinabro, 8th [41] J. Alexander et al, NIM A337 (1993) 171
Mtg. of DPF (1994), World Scientific, p.1474 [42] C. Ward, Cornell CBX-96-71
[5] S. Henderson, PAC 97 [43] N. Bacchetta et al, NIM A324 (1993) 284
[6] Proc. 2nd Workshop on Backgrounds at the [44] L. Blanquart et al, IEEE TNS 41 (1994)
Machine-Detector Interface, World Sci. (1997) [45] M. Raymond et al, NIM A351 (1994) 449
[7] S. Henderson, in Ref.[6] [46] M. Millmore et al, 2nd Workshop on Electronics
[8] G. von Holtey et al, CERN-SL/97-40 (EA) for LHC Experiments (1996) p.23
[9] B. Aubert et al, NIM A479 (2002) 1; T. Mattison [47] H.B. Luna et al, NIM A285 (1989) 349
et al, PAC 99; W. Kozanecki et al, NIM A446 [48] W.V. Hassenzahl et al, NIM A291 (1990) 378
(2000) 59; T. Fieguth et al, PAC 05; W.S. Lock- [49] J. Pfluger, G. Heintze, I. Vasserman, RSI 66
man et al, PAC 05 (1995) 1946
[10] K. Akai et al, NIM A499 (2003) 191; T. Abe, [50] CESR-B Conceptual Design for a B-factory
H. Yamamoto, PRST-AB 7, 072802 (2004); J. Based on CESR (1993)
292
[51] KEK-B Conceptual Design Report where Z and A are the absorber atomic and
[52] S. Henderson, 8th Mtg. of DPF (1994) World mass numbers, βc is the particle velocity, s the
Scientific, p.1480 path length in g/cm2 and εmax the maximum en-
[53] S. Henderson, XXVI ICHEP (1992) 2022 ergy transferred in a single collision. The log-
[54] D.J. Dumas et al, NIM A404 (1998) 17 normal distribution fits well the Vavilov function
[55] D. Carey, K.L. Brown, Ch. Iselin, SLAC- for κn > 0.3 [9], and its use simplifies simulation
246/UC-28/Fermilab PM-31 (1982) drastically.
[56] R. Brun et al, CERN DD/EE/84-1 (1987) To simulate δ-electron production at any step,
[57] W.R. Nelson, H. Hirayama, D.W.O. Rogers, one calculates εG = ξ/0.3 and εδ = min(εG , εc ).
SLAC-265 (1985) Then, the restricted energy loss with E < εδ
is sampled from log-normal distribution, and the
3.3.10 Particle Interactions and number of δ-electrons with E > εδ is simulated

using a Poisson distribution. The coordinates of
Beam-Induced Backgrounds
δ-electrons are calculated recursively. The elec-
and Radiation
N.V. Mokhov, S.I. Striganov, FNAL tron energies are sampled from the Bhabha for-
mulae. Total energy loss of a particle is the sum
Hadronic and Electromagnetic Showers of the δ-electron energies and of the restricted en-
Electromagnetic (EM) interactions, decays of ergy loss. Correlations of ionization energy loss
unstable particles and strong inelastic and elastic and angular deflection due to Coulomb scattering
nuclear interactions all affect the passage of high- need to be taken into account in some applica-
energy particles through matter [1, 2, 3, 4]. At tions. The algorithm is described in [9].
E > a few GeV the main feature of interactions Radiative energy loss of heavy particles For
is multi-particle production. Energetic particles muons (pions) with E > a few hundred GeV
are concentrated around the primary beam axis and protons with E > a few tens TeV, the ra-
forming the shower core. Neutral particles diative mechanisms (bremsstrahlung and direct
(mainly neutrons) and photons dominate with a e+ e− pair production) dominate over the ioniza-
cascade development when E < 100 MeV. The tion losses, especially in heavy materials. The ap-
shower dimensions grow with primary energy. proximation [6] to the exact Kel’ner formulae can
The nuclear interaction rate is determined by be used for the pair production cross-section,
hadron-nucleus cross-sections (Fig.2, Sec.3.3.2).
dσp a(1 + a) 1 dE
The EM showers induced mainly by π 0 -decays = bp , bp = (2)
have different scale, but several common features dν ν(ν + a)2 E dx p
with hadron cascades as listed in Tab.1. where (dE/dx)p is the mean energy loss per unit
Ionization energy loss Conventional ap- length, ν = (pair energy)/(particle energy). With
proach to EM processes is based on a continuous a few percent accuracy,
energy loss approximation using the mean 2
stopping power for charged particles in material cm me Z(Z + 1)
bp ≈ 1.689 × 10−5
[1, 2, 3, 5]. A detailed Monte-Carlo approach g M A
to EM interactions of heavy particles is de- −1/3

b3 Z
scribed in [6]. It is used in modern Monte-Carlo × b1 ln − b2
codes [7], where collisions of charged particles 1 + 4b3 Z −1/3 M/E
and atoms with energy transfer ε greater than Here M is the incident particle mass, respectively;
a cutoff εc are considered as discrete events a is determined from the second moment of the
involving production of δ-electrons, e+ e− -pair, cross-section,
bremsstrahlung [6]. The secondaries with E > εc ' −3 mμ
are followed explicitly. Energy losses with ε < εc 1 ·10 M ,
a E ≤ a3
a= −3 mμ (3)
(so-called restricted losses) are considered as a1 + a2 ln( a3 ) ·10 M , E > a3
E
continuous. The redefined Vavilov parameters [8]

for restricted loss distribution are ai , bi for some charged particles are given in
Tab.2. See also Sec.3.3.12.
Z
ξ = Bs, B = 0.1536 2 There are a number of different approaches
Aβ for the calculation of the muon bremsstrahlung
κn = ξ/εc , βn2 = β 2 εc /εmax (1) cross-section. These methods differ mainly in
293
Table 1: Comparison of the properties of electromagnetic and hadron cascades.
[∗]
See Sec.3.3.12 for X0 (radiation length), and λI (interaction length) for various substances. See Sec.3.3.1 and
[1] for Ec (critical energy). Emin is minimal energy of detected particles.
Characteristic Electromagnetic Hadron cascade

cascade
Multiplication Production of e− e+ pairs Inelastic hadron-
processes and electron bremsstrahlung nucleus collisions
Mean free path (9/7)X0 for photons,[∗] Mean free path of

before the X0 / ln(Ee /Eγ ) a hadron before the
interaction for electrons inelastic intteractions:
λI ≈ A/NA σin [∗]

Mean inelasticity K ∼ 1 for γ, K ∼ 0.5
for secondaries K ∼ 0.07 for e−
Secondary particles Electrons and photons Principally π mesons
and nucleons
Longitudinal development Depends weakly on matter Depends weakly on matter if
of the cascade if the depth t is expressed the depth t is expressed in
in units of radiation units of the mean free path
length, x = t/X0 . of a hadron before inelastic
interaction, x = t/λI .
Energy dependence of cascade ∼ ln(E0 /Emin ) ∼ ln(E0 /Emin )
maximum position
Characteristic length Λ = (3 − 4)X0 , does not Λ = (1 − 2)λI , increases
of the cascade decay depend on E0 with increasing E0
Length of the effective teff (10 − 30)X0 teff (5 − 10)λI
region of development ·ln(E0 /Emin ) · ln(E0 /Emin )
of the cascade
Transverse development Depends weakly on matter Depends weakly on matter
of the cascade if the distance from the if the distance from the
shower axis is expressed cascade axis is expressed
in Moliere units, rM in g/cm2
= Es X0 /Ec with Es
≡ me c2 4π/α ≈ 21.1 MeV
Transverse dimension reff 2rM reff ∼ λI
of the effective
region of development
of the cascade
the treatment of screening corrections. A gen- can be approximated by

eral expression [6] allows arbitrary nuclear and dγ s 1
atomic form-factors to be applied. For a small fcγ (Δc , E, s) d s 1−d s
(5)
εc γ Δc γ
energy transfer ε < εγ = 10−3 E at a muon en-
The continuous energy loss Δc at a step s is sam-
ergy E > 10 GeV, the bremsstrahlung differential
pled from here providing Δc < εc < εγ . The
cross-section reaches the complete screening limit
% 4b
production of bremsstrahlung photons with ener-
dγ dγ = 3γ gies > εc is considered as a discrete process.
Σγ (E, ε) = , where (4)
ε bγ = ( E1 dE
dx )γ Deep inelastic interactions Existing models
With this, the restricted energy loss distribution of deep inelastic muon-nucleus scattering are
294
Table 2: Parameters of the approximation for pair production cross section.
Muon Pion Kaon Proton

E > 20 GeV E > 20 GeV E > 50 GeV E > 90 GeV
a1 5.2 5.5 5.45 5.4
a2 .13 .18 .35 .43
a3 (GeV) 200 200 100 100
b1 .787 .791 .819 .833
b2 1.1 1.09 1.14 1.17

b3 2986 3017 2773 2532
consistent at the 30% accuracy level (see, e.g., independent of the pion charge. The follow-
[10]). However, the relative mean energy loss for ing parameterization was proposed [13] for p0 ≥
this process is ≤ 10% of the total, even at very 70 GeV/c and p⊥ ≤ 1 GeV/c:
high energies. The corresponding mean free path αg = 0.8−0.75 xF +0.45 x3F /|xF |+0.1 p2⊥ (7)
exceeds ∼ 100 m of iron for muons in the TeV
energy region. It turns out that Eq.(7) describes data well at
Hadron-nucleus interactions Elastic hA inter- p0 ≥ 24 GeV/c and can be successfully used at
actions are described in Sec.3.3.2.1. Inelastic hA lower momenta (5 ≤ p0 ≤ 24 GeV/c) if it is re-
interactions are described by either Monte Carlo placed with
event generators or phenomenological models [2]. α = αg − 0.0087 (24 − p0 ) (8)
Theoretical calculations based on the intranuclear ±
cascade model are reliable at proton momenta The RpA→π X ∼ Aα
form does not extrapo-
p0 < 5 GeV/c. Microscopic models, such as DP - late well to A = 1 because of the difference in the
MJET [11] (based on the dual topological unita- π-yield in pp and pn collisions. This difference
rization approach) do a decent job at higher ener- can be taken into account if one uses the follow-
±
gies > 10 GeV/c. The entire kinematical region ing form for RpA→π X [13]:
for the inclusive spectra can be reliably described α
± A
with phenomenological models such as [12]. RpA→π X = f (p0 , Y ) (9)
2
Many reliable data and parameterizations ex-
± dσ ±
ist on pion yield in pp-collisions. One can com- dp (pd → π )/ dp (pp → π ).
where f (p0 , Y ) = dσ
pensate for the lack of data for pA reactions by It turns out that pion yields in pd and pp collisions
using the following form for the double differen- are not very different, i.e. f (p0 , Y ) ≈ 1. One
tial cross section of the pA → π ± X reaction: finds f (p0 , Y )π− = 1 + 0.225/Nπ− − aπ− Ycms ,
±
d2 σ pA→π X where Nπ− is mean π − multiplicity in pp colli-
sions and Ycms is pion rapidity in the CM. Data
dpdΩ
show linear dependence of Nπ− on free energy
d2 σ pp→π X
±
√ √ 0.25 √
= RpA→π
±X
(A, p0 , p, p⊥ ) (6) W = ( s − 2 · mp )0.75 / s , where s is
dpdΩ the CM collision energy. Our fit to the data gives
where p and p⊥ are the total and transverse mo- Nπ− = 0.81(W − 0.6). The other parameter
menta of π ± , and A is an atomic mass of the target aπ− = 0.16 for p0 ≤ 20 GeV/c, and depends on
±
nucleus. The function RpA→π X , measured with energy for higher momenta as aπ− = −0.055 +
much higher precision than the absolute yields, is 0.747/ ln(s). f (p0 , Y )π− is forced to be 1 if it be-
almost independent of p⊥ and its dependence on comes less than 1. For π + production the approx-
p0 and p is much weaker than for the differen- imation is much simpler f (p0 , Y )π+ = 0.85 +
tial cross-section itself. Because of rather differ- 0.005 p0 for p0 ≤ 30 GeV/c and f (p0 , Y )π+ = 1
ent properties of pion production on nuclei in the for higher momenta.
forward (xF >0) and backward (xF <0) hemi- R(xF < 0.05) In this region, due to the lack
spheres, where xF is the Feynman’s longitudinal of experimental data on α, one uses the following
variable, one treats these two regions differently. expression for the function R in Eq.(6):
R(xF > 0.05) In this region one assumes ±X dN/dY (pA)
RpA→π
±X
∼ Aα . The power α is almost RpA→π = (10)
dN/dY (pp)
295
The following scaling law was proposed Table 3: Parameters in Eq.(14).

in [14] for charged shower particle (β > 0.7) pro-
A B C D E F
duction in pA collisions at 20–400 GeV/c:
π+ 60.1 1.9 0.18 0.3 12 2.7
Y0 dN
= f A,
η
(11) π− 51.2 2.6 0.17 0.3 12 2.7
Ns dη Y0
where Ns is a mean multiplicity of shower
particles, Y0 is rapidity of primary proton and collisions one can use the formula

±
η = − ln(tan(θ/2)) is pseudorapidity of a sec- d3 σ pp→π X p∗ B
ondary particle. One finds that this approxima- E = A 1 −
dp3 p∗max
tion is in a reasonable agreement with data at ∗
p
× exp(− √ )V1 (p⊥ )V2 (p⊥ )
p0 > 7.5 GeV/c. Unfortunately, η is not a conve- (14)

nient variable to describe forward pion production C s
(θ ≈0). Analysis [12] of the pA → π − X data at where p∗ and p∗max are pion momentum and max-
10 < p0 < 100 GeV/c shows that replacing η in imum momentum transfer in CM and parameters
Eq.(11) with rapidity Y are given in Tab.3. The best description of the p⊥

dN Nπ Y dependence is obtained with:
= F A, (12)
V1 (p⊥ ) = (1 − D)e−Ep⊥ + De−F p⊥ ,
2 2
dY Y0 Y0
provides better description of the pion yield in the p⊥ ≤ 0.933 GeV/c
entire kinematic range. Here Nπ is mean pion
multiplicity, Y0 = ln( E0m+pp
0
) is rapidity of inci- V1 (p⊥ ) = 0.2625/(p2⊥ + 0.87)4 ,
dent proton and Y = ln( Eπm+p z
) is π rapidity, p⊥ > 0.933 GeV/c
⊥
2 2
m⊥ = p⊥ + mπ . One chooses the Gaussian (15)
form for the scaling function V2 (p⊥ ) = 0.7363e0.875p⊥ ,
2
Y Y p⊥ ≤ 0.35 GeV/c
F A, = c1 exp − − c2 /c3
Y0 Y0 (16)
(13)
V2 (p⊥ ) = 1, p⊥ > 0.35 GeV/c
where for π − : c1 = 1.149 A0.0479 , c2 =
0.492 A−0.0565 , and c3 = 0.214 A−0.121 . Reliable γA→hX reactions Hadroproduction in
rapidity distributions for π + at xF <0 are mea- photon-nucleus interactions at Eγ ≥ 0.14 GeV
sured only for p0 ≥ 100 GeV/c. Assuming that can be simulated approximately by replacing the
the scaling (15)-(16) is valid for π + also, one finds photon with a real pion of random charge with
the following parameters from data: c1 = 1.6, the same kinetic energy. The total cross-section
c2 = 0.4833 − 0.0006A, A < 108 and c2 = is calculated as σγA = RA [Zσγp + (A − Z)σγn ].
0.418, A > 108, c3 = 0.1873 − 0.0027A, A < For the total γp cross-section experimental
24.3 and c3 = 0.122, A > 24.3. The data on data are used [6]. The A-dependence of the
dN/dY for pp → π ± X reaction is well de- cross-section RA is extracted from experimental
scribed by: dN/dY = Cpp exp(−Ycm 2 /2σ 2 + data, with RA ≈ 1.047A−0.085 at high energies.
√
d Ycm ),√where σπ+ = −2.725 + 2.065 ln(√ s) +
4 A comparison of this description with data is
√ s and σπ− = −0.732 + 1.1 ln( s) +
4.205/ shown in Fig.1. Photoneutron production in the
1.3/ s. The parameter d depends on energy as giant resonant energy region 6 ≤ Eγ ≤ 60 MeV
d = w/sγ − 1/R, where wπ+ = 6.68, γπ+ = 2.1, is described according to the algorithm [15, 16]
wπ− = 496 and γπ− = 4. Coefficient R reads extended for light nuclei 4 ≤ A ≤ 56 on the
as Rπ+ = 41.5 + 0.151(ln(s))3.29 and Rπ− = basis of the latest data. An interpolation is used
63.2 + 0.052(ln(s))3.77 . The normalization pa- between 60 and 140 MeV.
rameter in (13), combined of Nπ , Cpp etc, is Space-time source term A detailed compari-
chosen to match the functions (12) and (13) at son of detector backgrounds for hadron, electron
xF = 0.05. and muon colliders is given in [17]. LHC, NLC
pp → π ± X To describe the invariant cross (or ILC nowadays) and 2 × 2 TeV μ+ μ− collid-
section of charged pion production in pp ers are considered representatives of their classes.
296
of detector components, and the radiation envi-

ronment in the experimental hall. High instan-
taneous particle fluxes complicate track recon-
struction, cause increased trigger rates, and affect
detector occupancy. Therefore, the beam’s time
structure must be taken into account in the com-
parison of expected background environments at
different machines.
Instantaneous and integrated rates Parame-

ters of three types of future colliders are presented
in Tab.4. The integrated luminosities are obtained
from the design peak values by multiplying by

107 s. The instantaneous or effective luminosity,
which determines the detector performance, is
defined for the amount of radiation in the detector
active element over the drifting time Δtd or the
bunch train length, whichever is smaller. Collider
detector elements most susceptible to occupancy
problems have the drifting/integration time in
Figure 1: Calculated photon-nucleus cross section in the 40 to 300 ns range. Taking conservatively
comparison with experimental data vs photon momen- Δtd = 300 ns, one gets Leff = L × Δtd for LHC
tum for lead, copper, carbon, beryllium and hydrogen. and Leff = L/f for the other colliders, where f is
the repetition rate (collision rate for μ+ μ− ). The
NLC (ILC) γγ mode is 10 to 20 times worse than
Despite different colliding particle types and ma- the e+ e− one because of the large backscattered
chine parameters, there are common considera- laser cross-section [18]. Multiplying the above
tions for these accelerators. luminosities by the appropriate hadron production
Collision remnants from the IP are often the cross-sections, one can compare pp, e+ e− and
major source of background and radiation levels μ+ μ− collisions as a source of background. The
in hadron collider detectors, in experimental halls last two lines of Tab.4 show that LHC produces
and in final focus quads. Small aperture collima- at least 106 times more background hadrons
tors on either side of the IP in front of the first from the IP annually than the lepton machines.
final focus quads protect the accelerator compo- At the same time the instantaneous background
nents from IP radiation. Shielding around colli- productions are not so drastically different.
mators reduces radiation levels in detectors and in The situation is very different with the accel-
collision halls. erator related backgrounds. This component is
Beam loss in the IP vicinity is the second due to elastic and inelastic beam-gas interactions
background source. Without protection, the num- in the collider lattice, and due to quasi-local beam
ber of hits from halo particles in the detector can halo loss in the IR components, mainly diffrac-
be greater than that from the IP and their prod- tive protons from another IP, and halo tails from
ucts. Reduction of beam loss in the IR improves the beam cleaning system. At the high-energy
this situation. Multi-stage high-performance col- lepton colliders the backgrounds generated in the
limation system (Sec.3.3.11) is the major way to machine are a major concern. Synchrotron radi-
do that. Tertiary collimators at the non-IP sides ation and muons produced in beam halo interac-
of the IR with additional collimators as close as a tions along the lattice create serious backgrounds
few centimeters from the IP (at lepton colliders) in the detectors at linear e+ e− colliders. These
are mandatory nowadays. Plugging the accelera- can be reduced with an appropriate final focus de-
tor tunnel at the entrance to the experimental hall sign and a set of collimators (Secs.3.3.9,3.3.11).
further helps reduce this component. At high energy muon colliders the situation is
Temporal considerations in the background much worse. Unavoidable μ→eν ν̃ decays oc-
analysis are important. Integrated levels deter- curring in the beam-pipe are the major source
mine radiation damage, aging and radioactivation [6, 17, 19, 20, 21, 22].
297
Table 4: Collider parameters and calculated integrated and effective luminosities.
Parameters LHC NLC-500 NLC-1000 μ+ μ−

Ecm (TeV) 14 0.5 1 4
L (1034 cm−2 s−1 ) 1 0.71 1.45 4.55
Rep. (collision) rate f (Hz) - 180 120 4.04×104
Particles/bunch (1011 ) 1 0.07 0.11 20
Bunch/pulse - 90 75 1
Bunch separation (ns) 25 1.4 1.4 18.6×103

Yearly Ly (fb−1 ) 100 71 145 455
σh (μb) 80×103 0.045 0.034 0.054
Δtd or bunch train length (ns) 300 126 105 -
Leff (cm−2 ) 3.00×1027 3.94×1031 1.21×1032 1.13×1030

(σh × Ly )/(σh × Ly )LHC 1 4.00×10−7 6.16×10−7 3.07×10−7
(σh × Leff )/(σh × Leff )LHC 1 7.39×10−3 1.71×10−2 2.54×10−4
Hadron colliders The mean energies of particles coming to the

Interaction point The charged particle fluxes detector from the LHC tunnel are [23] 6.6 GeV
in the detector tracking cavity, determined by the (μ), 8.1 GeV (h± ), 310 MeV (n), 150 MeV (e± ),
primary events, decrease as 1/r 2 , whereas neu- and 30 MeV (γ). The mean distance from
tron fluxes are more uniform in the cavity, de- the beam axis is 1.5 m, the mean angles are
pending strongly on the calorimeter material. The ∼ 600 mrad for neutrons and ∼ 130 mrad for all
damage induced in semiconductor components other particles. Most of these particles can be
is linearly dependent on the non-ionizing energy intercepted with a concrete plug at the tunnel-
loss (NIEL ∼100 MeV mb), determined by the in- hall interface, except for muons, which penetrate
tegrated flux of > 100 keV neutrons and charged through shielding, accelerator and detector, creat-
hadrons. The hit rates are proportional to the ing a rate of a few Hz/cm2 in the detector.
charged particle flux (primary and that created e+ e− linear colliders
by neutrals) in the sensitive volume and are re- Interaction point Tab.5 shows that the average
lated to the effective luminosity. The rates in the integrated hadronic fluxes produced at the IP at a
endcap calorimeters are much higher, especially linear collider (ILC project) are about 106 lower
at small radii. Neutron fluxes in the electromag- compared to LHC. However, the difference in in-
netic calorimeter can reach high values severely stantaneous rates is less drastic.
restricting the lifetime of silicon detectors and Beam halo Synchrotron radiation, beam-gas
readout electronics. In the forward muon system, and beam halo interactions with the components
the signal is composed of charged particles, pho- of the final focus and adjacent sections of the lin-
tons with ∼1% efficiency and neutrons (∼0.3% ear colliders create fluxes of muons and other sec-
efficiency). For the CMS and ATLAS detectors, ondaries which can exceed the tolerable levels at
the signal rate ranges from 600 Hz/cm2 at r = 1 m a detector by a few orders of magnitude. A multi-
to a few Hz/cm2 at r > 5 m. stage collimation set and a system of magnetized
Beam halo The crucial issue here is the beam iron spoilers which fill the tunnel can meet the
loss distribution in the IR [23, 24]. The first com- ILC design goal of allowing a continuous 0.1%
ponent is calculated via modeling beam interac- beam loss, resulting in a few muons at the detec-
tions with residual gas in the beam pipe using tor [26, 27].
a detailed gas pressure map for the entire ma- μ+ μ− colliders
chine [25]. For the LHC, the nuclear inelastic Interaction point A muon collider is the
interaction rates are approximately 500 m−1 s−1 “cleanest” machine with respect to both integrated
in warm and cold straights and 2 ×104 m−1 s−1 and instantaneous particle background from the
in cold arcs. The second component is formed IP (see Tab.5).
by interactions with the IR limiting apertures of a Beam decays One of the most challenging
tertiary halo of protons escaping the betatron and problems in the design of a muon collider arises
momentum cleaning collimation systems. from muon decay [6, 17, 20, 22]. With 2 × 1012
298
−2
Table 5: Background fluxes (cm ) from the IP ac- (μ). There is a rather uniform distribution of neu-
cumulated over 1 year (1) and effective (2) in central trals in the cavity with charged fluxes being sub-
tracker, endcap calorimeter and forward muon spec- stantially lower. The maximum hit rate density
trometer at different radii. in a vertex detector of a 1.5-TeV muon collider is
Detector r (cm) LHC ILC μ+ μ− calculated to be about 1 MHz/mm2 , same as in the
(1) CMS detector at the LHc at the same luminosity.
At the same time, total the peak hit rates per bunch
Tracker 30 2×1013 107 6×106 crossing at the muon collider are higher than at the
ECAL 50 1014 108 108 LHC. It is assumed that a reliable beam cleaning
Forward 100 1011 5×103 8×103 system is in the lattice far upstream from the IP.
(2) Studies show that the loss of even a small fraction
Tracker 30 0.6 0.01 2×10−4 of the beam within a few hundred meters of the IP
ECAL 50 0.9 0.8 2×10−2 results in backgrounds in a detector comparable
to those from μ→eν ν̃ decays.
Background particle spectra and space distri-
Table 6: Accumulated over 1 year and effective accel- butions are not very different in similar detector
erator related fluxes (cm−2 ) in detector components at configurations at hadron, e+ e− and μ+ μ− col-
r = 50 cm, with all the protective measures on. liders. Expected background levels are summa-
rized in Tab.5 for IP and in Tab.6 for accelera-
LHC ILC-1000 μ+ μ− tor backgrounds. The integrated fluxes are con-
14 verted to the NIEL or “equivalent 1 MeV neutron”
Integrated 108 1.6×106 10 values, and instantaneous (effective) fluxes (sig-
Effective 3×10−6 10−3 5 nals) are defined as the charged particle flux plus
0.003× (Fn +Fγ ).
muons in a 2-TeV bunch, there are 2 × 105 References

μ→eν ν̃ decays per meter in a single pass through [1] Review of Particle Physics, PLB 592 (2004)
an IR, or 6 × 109 decays/m/s. Both the decay [2] A.N. Kalinovskii, N.V. Mokhov, Yu.P. Nikitin,
electrons (E ≈ 700 GeV) and the synchrotron Passage of High-Energy Particles through Mat-
photons emitted by these electrons in a strong ter, AIP (1989)
magnetic field induce EM showers in the col- [3] J.D. Jackson, Classical Electrodynamics, 3rd ed.,
lider and detector components. Almost 15 MW of Wiley (1999)
power is deposited in the storage ring, or about [4] A. Ferrari, P. Sala, The Physics of High Energy
2 kW/m. The resulting heat load to the cryo- Reactions, Int. Centre for Theoretical Physics
genic systems and the background levels in the (1996)
collider detectors are serious issues. The intense, [5] D.E. Groom, N.V. Mokhov, S.I. Striganov,
directed neutrino beam generated is perhaps an Atomic Data and Nuclear Data Tables, 78 (2001)
issue more serious than the high radiation level 183
from EM showers. Neutrino interactions in the [6] N.V. Mokhov, S.I. Striganov, AIP Proc. 372
soil intended to shield the collider may produce (1995) 234
unacceptable radiation levels at large distances. [7] N.V. Mokhov, FNAL-FN-628 (1995);
Fermilab-Conf-04/053 (2004); http://www-
By carefully designing the final focus system ap.fnal.gov/MARS/
with special spoiling dipoles, by embedding so- [8] P.V. Vavilov, JETP 5 (1957) 749
phisticated collimators in the immediate IP vicin- [9] S.I. Striganov, Rad. Prot. Dosimetry 116 (2005)
ity and shielding and other protective measures 293-296
the heat load and backgrounds can in principle be [10] L.B. Bezrukov, E.V. Bugaev, Sov. J. Nucl. Phys
mitigated by several orders of magnitude [17, 20]. 33 (1981) 635
All particles over a wide energy range contribute [11] S. Roesler, R. Engel, J. Ranft, ICRC-2001,
to the background levels. Mean momenta of par- Copernicus Gesellschaft (2001) 439
ticles entering a 1.5-TeV muon collider detector [12] N.V. Mokhov, S.I. Striganov, AIP Proc. 435
are[22]: 0.9 MeV/c (γ), 6 MeV/c (e± ), 45 MeV/c (1997) 453
(n) 0.5 GeV/c (charged hadrons) and 23 GeV/c [13] W.M. Geist, Nucl. Phys. A525 (1991) 149
299
1000
[14] E. Stenlund, I. Otterlund, CERN-EP/82-42 LHC
(1982) 100
Stored Energy [MJ]

[15] I.S. Baishev, I.A. Kurochkin, N.V. Mokhov, LHC
(2010)
IHEP-91-118, Protvino (1991) 10 ISR
SPS HERA-p
[16] B.L. Berman, S.C. Fultz, RMP 47 (1975) 713 ILC
1 Tevatron
[17] N.V. Mokhov, Nucl. Phys. B (Proc. Suppl.) 51A RHIC
PEP-II
(1996) 210 0.1
SNS LEP-2 CLIC
[18] R. Engel et al, Proc. 2nd Workshop on KEKB
Simulating Accelerator Radiation Environments

10000
(SARE2) (1995)
Energy Density [MJ/mm2]

LHC
[19] μ+ μ− Collider Feasibility Study, BNL-52503;
ILC CLIC
1000
LHC
Fermilab-Conf-96/092; LBNL-38946 (1996)
100 (2010)
[20] C. Johnstone, N. Mokhov, Fermilab-Conf- 10
96/366 (1996) SPS
HERA-p
1 KEKB
[21] N. Mokhov et al, PAC 11 (2011); also RHIC Tevatron
arXiv:1202.3979 0.1 PEP-II

LEP-2
SNS
[22] N. Mokhov S. Striganov, Physics Procedia 0.01 ISR
(2011), arXiv:1204.6721 1 10 100 1000 10000

[23] A. Drozhdin, M. Huhtinen, N. Mokhov, NIM Beam Momentum [GeV/c]
A381 (1996) 531
[24] N. Mokhov, T. Weiler, CERN-2009-003 (2009) Figure 1: Stored energy (top) and stored energy den-
37 sity (bottom) for different electron (open symbol) and
[25] A. Drozhdin, N. Mokhov, S. Striganov, PAC 09, proton (filled symbol) accelerators. ILC and CLIC are
Fermilab-Conf-09/172-APC design studies for e+ e− colliders. Stored energy den-
[26] A. Drozhdin et al, Fermilab-TM-2200 (2003) sity is calculated at locations of collimators or other
[27] N.V. Mokhov, A.I Drozhdin, M.A. Kostin, PAC targets.
05
Estored /(2πσx σy ) is increased. Energy densities
for various accelerators are shown in Fig.1 [1].
3.3.11 Beam Collimation Modern accelerators operate with beam densities
R. Assmann, CERN in the range of 10 kJ/mm2 to 4 GJ/mm2 . This is
Stored energy and density of beams The beam often far above the typical damage limit of metals
current ib is maximized by increasing the repeti- (50 kJ/mm2 ).
tion frequency (or revolution frequency) fr of the Collimator concepts Collimators place special
accelerator and by increasing the number of parti- blocks of materials (“jaws”) at strategic locations
cles Np = ib /(fr q), with q being the charge of a such that beam losses are safely absorbed, pro-
beam particle. Considering single charge particles tecting against damage and loss-induced pertur-
(q = e) with relativistic momentum p (γ 1), bations of the accelerator operation. The colli-
a beam with Np particles carries a stored energy mator jaws are the materials which are closest to
Estored = p c e Np . Here, c is the light veloc- the beam and usually must be especially robust.
ity. Fig.1 compares the stored energy for various Many different concepts exist. A few different
electron and proton accelerators [1]. Modern ac- collimator schemes are illustrated in Fig.2. Col-
celerators operate with beams between 10 kJ and limators can be as simple as a fixed vacuum pipe
500 MJ. Most of the listed accelerators are storage with thick walls, sometimes movable as a whole.
rings where the beam is exploited for many hours. At the other extreme they can be advanced high
However, linac-based facilities (SNS, ILC, CLIC) tech devices which consist of a vacuum tank, two
use the produced beam only once and regenerate fully movable jaws, cooling, stepping motors and
it at 5–100 Hz. Loss-induced heating can be in redundant position diagnostics. This design type
the range from 1 kW to 500 kW if 0.1% of beam (as developed for the LHC) is illustrated in Fig.3.
is lost over 1 s. Collimator jaws are either slightly curved
Most accelerators require a high brightness, along the beam direction (to reduce sensitivity to
achieved by small transverse beam sizes σx and beam-jaw angular alignment) or have a fully flat
σy . As a consequence the energy density ρe = top. Magnetized jaws have been considered and
300
vacuum tank (water cooled)
inside gap
e holderr
der
holder
ed holde
Beam
Beam
water-cooled
cooled
water-cooled
e cooled
jaw
jjaw jaw
er coo
with
it with
temp.
temp
beam temp.
water-
wa er-
prob
o e
probe and probe
wat
wat
halo
support with water & cable feed-through

flexible
(a) Fixed elliptical mask (b) Fixed rectangular mask bellows
reference plane
outside gap
reference plane
Vacuum tank pos sensor pos sensor
gap sensor
motor motor
+ resolver + resolver
roller cages
Beam Beam roller cages
mechanical table (fixed on vacuum tank)
vacuum tank (water cooled)

jaw
with flange
temp.
probe
jaw assembly
with 2 temperature probes per jaw
(c) One-sided, L-shaped (d) Two parallel movable with water cooling

beam
direction
movable jaw jaws
flexible
bellows
Figure 2: Illustration of possible collimator concepts.
All designs can be installed in different azimuthal
mechanical table
angles to provide collimation in horizontal, vertical mechanical table
(downstream) (upstream)
and/or skew planes.

Figure 3: Sketch of an advanced collimator design
with two movable jaws inside a vacuum tank (top:view
might have some potential for very high beam en- along beam direction; bottom: side view). The jaws
ergies. So far, however, magnetized jaws have not are supported by two mechanical tables, one at each
been used. end of the tank. Reference planes allow precise cal-
System design goals Several strategically ibration during production and outside monitoring of
placed collimators form a collimation system that absolute jaw positions and gaps.
has to fulfill various critical tasks [1, 2, 3]:
(i) Background control: Beam halo, beam- Heating from synchrotron-radiation is intercepted
gas scattering and emitted synchrotron photons with local collimators.
can induce spurious signals in physics detectors. (iv) Ultra-high vacuum: Beam loss induced
Materials with good absorption quality are placed desorption can reduce beam lifetimes. This is
around the experiment and shield the sensitive de- particularly important for partially ionized beams:
tectors. ions lose an electron by beam gas scattering, af-
(ii) Hands-on maintenance: Collimators in- fected ions get lost in a high dispersion point, des-
tercept beam losses at locations optimized for ra- orbed molecules degrade the vacuum, more ions
diation impact. A typical design goal is to keep experience beam-gas events, . . . and a vacuum in-
beam loss outside of collimators below the 1 stability develops. Collimators with low desorp-
W/m level allowing hands-on maintenance (resid- tion materials are designed to intercept losses in-
ual dose rate < 1 mSv/h at 30 cm). Collimators stead.
must have good absorption quality and are often (v) Passive protection: High power beams can
heavily shielded. potentially destroy parts of the accelerator. Colli-
(iii) Protection against heating: Direct beam mators are placed at strategic locations to inter-
loss or synchrotron radiation losses can lead to cept mis-directed beam, for example from kicker
local heating. Typical quench limits for super- errors (timing, amplitude) and trips of accelera-
conducting magnets range from 1 mW/cm3 to 100 tor equipment (dipoles, quadrupoles, correctors,
mW/cm3 , to be compared with possible power rf). Exposed materials must be robust. Additional
loss in the kW to MW regime. Beam loss related collimators are often placed behind the exposed
heating is minimized by highly efficient multi- collimators for absorption of diluted beam parti-
stage collimation sections (leakage 0.1%). cles and showers.
301
0.01
Measured Beam Loss [Gy/s]
Impact on 1m carbon
collimator block
0.001
Impact on 1m tungsten
collimator block
0.0001
1e-05
1e-06
-50 0 50 100 150 200 250 300
Distance from first impact [m]
Figure 4: Measured beam loss profile (BLM) down-

stream of a single collimator for two different materi- Figure 5: Photograph of a collimator gap defined by
als when hit by 450 GeV protons. two parallel, flat jaws. The image currents of the beam
are guided by rf fingers.
Effect of a single collimator A collimator jaw

surface before the whole jaw length is traversed.
intercepts stray particles, but can also give rise to
Collimation is an edge process and becomes a
a wakefield (or impedance).
multi-turn process in storage rings, with more
When a particle hits a collimator jaw then the than one impact on the primary collimator.
particle experiences a number of elastic and in- Collimator jaws usually are the closest ma-
elastic processes (Secs.3.3.1,3.3.10). An inelas- terials to the beam in order to fulfill their func-
tic interaction initiates a downstream shower and tion. Being so close, the electro-magnetic fields
the initial particle is usually lost locally (often of the beam interact with the surface of the colli-
called “absorbed”). The probability for inelastic mator jaw. The jaws must be tapered and rf con-
interaction and the amount of energy absorbed in tacts must be properly included into the collima-
the collimator block depend on the type of beam, tor design to guide the image currents. The mag-
its energy, the jaw material used, the jaw length nitude of the wakefields depend on the resistiv-
and transverse dimensions [4]. They are calcu- ity of the jaw material and the distance from the
lated with dedicated specialist tools like FLUKA, beam. For short bunches, dipolar deflections can
MARS and GEANT. Measurements are shown in be induced and were indeed observed and charac-
Fig.4 for the LHC beam, comparing measured terized for the SLAC linac [6]. For long bunches
beam loss for 450 GeV protons impacting on a (like the LHC) the resistive wall impedance Z⊥
1-m long carbon block or on a 1m long tungsten can be dominated by collimator-induced impe-
collimator with jaws fully closed. An attenuation dance (Z⊥ ∝ 1/gap3 for a two-sided collima-
by a factor 100 is reached within 15 m behind a tor) [7, 8]. For low frequencies the classical thick
tungsten block, while the same takes 60 m for a wall theory is modified, so-called “inductive by-
carbon block. High Z materials are therefore used pass” (Secs.3.2.2,3.2.4).
for absorption while much more robust, low Z
The collimator volumes can in bad circum-
materials are used for diluting any impacting
stances lead to trapped modes and collimator
beam.
heating. Modern 3D simulations are applied to
Single collimators cannot achieve very high calculate any possible trapped modes and to ad-
efficiencies. The limited absorption power of dress them by either modifying the collimator de-
a single collimator is further reduced by out- sign or by placing ferrites for absorption. So-
scattering from the jaw surface. Particles diffuse called rf fingers are required to guide the image
slowly to large amplitudes in storage rings. They currents of the beam and to prevent discontinu-
hit the collimator jaw as limiting aperture with ities. An example of a collimator gap and rf fin-
small transverse offsets (impact parameter), in the gers is shown in Fig.5.
oder of a few 100 nm to a few μm [5]. Particles
experience multiple Coulomb scattering in the Collimator-induced phase space cuts Beam
jaw and many can escape through the collimator collimation is defined with normalized
302
coordinates zn , zn (z = x or y):

zn = z/ βz z (1) Dx
Dx [m]
αz · z + βz · z
βx [m]
zn = √ (2)
βz z
βx
Ideally a particle describes a circle
with con-
stant normalized amplitude az = zn2 + zn 2 .
However, in practice az is stochastically increased

by diffusion processes: space charge, scatter- jaw at +n1
ing (IBS, beam-gas, beam-beam, . . . ), rf noise,
in σx(δ=0)
Dx = 0
ground motion and other effects. A beam halo

(beam tails) forms around the beam core. A two-
sided collimator shall be placed at a normalized Dx = 0
ax
jaw at -n1
D (δ)
distance ±n1 fromthe beam center. Then its jaws =D
x
x,0 βx (δ)
=β
are set at ±n1 · βz,0 z where βz,0 is the on-
x,0
momentum beta function (δ = 0). The follow-

δ
ing combinations of betatron amplitudes (az ) and
momentum offsets (δ) have particles impacting on Figure 6: Energy dependence of beta and dispersion
the collimator edge [9]: functions at a collimator location in the LHC (top) and
its effect on the normalized betatron cut, Eq.(3) (bot-
βz (δ) Dz (δ)
±n1 = az + δ (3) tom) [9].
βz,0 βz,0 z
A pure betatron cut az = n1 is achieved when large dispersion tilts the betatron cut versus en-
the collimator is placed at a location with zero ergy offset δ, while the chromatic change of the
dispersion and chromatically well-corrected op- Twiss functions introduces a curvature in the be-
tical functions (no dependence on δ). Typically tatron cut. Energy offsets can be large in the
one uses 4σz < n1 < Az with Az being the halo and chromatic effects can become limiting
normalized machine aperture that shall be pro- factors.
tected against primary beam loss. Locations with
high beta functions are advantageous for reduc- System design guidelines (i) The number of re-
ing wakefield and impedance effects, relaxing me- quired primary collimators per plane depends on
chanical tolerances, increasing impact parameters the details of the beam and the accelerator. A
and lowering beam density in case of erroneous stored beam in a circular accelerator features a
hits (better robustness against error cases). slow diffusion of primary beam halo that appears
Maximizing the dispersion makes the colli- locally at the limiting aperture bottleneck. A sin-
mator sensitive to off-momentum losses. n1 can gle primary collimator (in the extreme case with
be increased and the collimator intercepts pre- one jaw) is sufficient to establish primary colli-
dominantly off-momentum particles, especially mation of beam halo in a given plane at n1 . Dis-
when another collimator at zero dispersion cuts tributed losses can also arise from synchrotron
the betatron halo. An optimum is achieved with radiation, beam gas induced ionization of par-
a maximum normalized dispersion at the primary tially ionized beams, machine errors and other
momentum collimator [11]: sources. A large number of distributed collima-
tors or masks is required to intercept these losses.
1 dD α
=− (4) Finally, single pass accelerators and transfer lines
D ds β require several primary collimators distributed in
The chromatic correction of the optics is essen- phase advance for constraining the maximum ex-
tial in order to guarantee phase space cuts at the tent of beam halo [10]. The minimal system
foreseen depths. This is illustrated in Fig.6. The requires two two-sided primary collimators per
energy dependence of the dispersion and the beta plane, separated by 90◦ in phase advance. √ This
function is shown for the location of an LHC mo- would constrain the extent of beam halo to 2 n1 .
mentum collimator. The resulting change of the More primary collimators separated by less phase
normalized betatron cut ax is shown as well. The advance can reduce this maximum extent further.
303
Primary C Secondary C Tertiary W SC triplet
(ii) Gaps for a given n1 shall be maximized. collimator collimator collimator
This requires locations with a high beta func-
tion for betatron collimators and a high disper-
sion for momentum collimators. This guideline n3
reduces wakefield and impedance effects, relaxes n1 n2
Primary beam & halo
mechanical tolerances for collimators, increases
impact parameters and lowers beam density in Secondary halo
Tertiary halo
case of erroneous hits (better robustness against
Quartiary halo
error cases).
(iii) A betatron collimation section shall be Cleaning insertion Experimental insertion
followed by a momentum collimation section,
if high efficiency is important. The interac- Figure 7: Illustration of multi-stage cleaning as used
tions in betatron collimators produce primary off- for the LHC. The primary collimators intercept the pri-
momentum particles. mary beam halo and leak a secondary halo. The sec-
(iv) Collimators can become highly radioac- ondary collimators intercept the secondary halo and
tive. Large transverse space should be foreseen leak a tertiary halo that is intercepted by tertiary col-
to provide possibilities for shielding and/or opti- limators. A fourth stage also exists, but is not shown
mized handling. here.
Multi-stage collimation systems Highly effi-
cient collimation systems rely on several stages. Primary Secondary collimators Primary Secondary collimators
collimator collimator
The classical two-stage collimation system [11, Retraction
Δx
Retraction
Δx
12, 13] relies on a primary collimator for scat-
tering and secondary collimators for absorption.
Orbit offsets
The recent design at the LHC [14] has extended Beta beating
the two stages into four stages. Stages are de-

fined by different normalized settings to the beam.
For example, the primary collimators are clos- Figure 8: Illustration of machine errors (left: orbit er-
est to the beam and have a distance of n1 σ from ror; right: beta beating) that can affect the collimation
the beam center. Defining the normalized setting hierarchy in normalized phase space and reduce clean-
of primary, secondary and tertiary collimators as ing efficiency.
n1 , n2 , n3 we have the following condition:
n1 < n2 < n3 < Az (5) analytically [11, 12, 13]. An optimal efficiency
where Az is the normalized machine aperture as can be achieved by placing two secondary col-
introduced above. The system designed for the limators at unique phase advance locations be-
LHC at 7 TeV is illustrated in Fig.7. hind the primary collimator. This is illustrated in
The values for n1 , n2 , n3 must be chosen such Fig.9. In normalized phase space a particle de-
that the critical aperture Az is effectively shad- scribes a circle of amplitude n1 . At max zn it is
owed. n1 must be large enough not to reduce intercepted by a primary collimator which is set
the beam lifetime or the transmission, normally to n1 (position 1 in Fig.9). The particle receives
n1 > 4σ. The difference Δx1 = n2 − n1 is called a kick that increases its normalized amplitude to
secondary retraction, Δx2 = n3 − n2 is the ter- n2 . At the exit of the primary collimator it has
tiary retraction. It is important that the multi-stage position 2 in Fig.9. The particle must then be
hierarchy is fulfilled at all times. The retraction transported through some phase advance before
values must therefore be large enough to allow it again reaches max zn and can be intercepted
for drifts of the beam center and changes in beam by a secondary collimator set to n2 (position 3 in
size due to variations in the betatron function [1]. Fig.9). The phase advance depends on the sign of
The impact of orbit and betatron errors are illus- the kick received in the primary collimator.
trated in Fig.8. The collimation system must be The optimal phase advance μopt from the
designed in coherence with a machine tolerance exit of the primary collimator to the first primary
budget for proper functioning. collimator is calculated to be
The one-dimensional betatron collimation n1
system with two-stages has been fully described cos μopt = (6)
n2
304
]Q¶ ]Q¶ Table 1: The extension Amax2 of secondary beam halo
is listed for three primary collimators and for different
Q
Q numbers Nsec of secondary collimators per primary.
Collimators with two flat, parallel jaws are assumed.
The last column shows a realistic example. [13]
]Q Q ]Q
Nsec Amax
2 Amax
2
Q (n1 = 6, n2 = 7)
3 4n22 − 3n21 9.4

4 2n2 − n 2
√1
7.9
Figure 9: Illustration of collimator induced particle
4n22 −2n21 + 2n21
kicks and optimal phase advance to a secondary col- 8 √ 7.2
2+ 2
limator. See discussion in text. ∞ n2 7.0
The two secondary collimators should then be

placed at phase advances μ1 and μ2 downstream 1
of the primary collimator: 0.1

Secondary halo
Normalized population
μ1 = μopt and μ2 = π − μopt (7) 0.01
For a realistic system design, however, it is not 0.001
sufficient to consider a 1D collimation system as 0.0001 Tertiary halo

sketched above [15]. A particle impacting on
1e-005
a horizontal collimator will be scattered both in
horizontal and vertical directions. It is therefore 1e-006
required to design a more sophisticated system. 1e-007
The maximum extent Amax 1 of the primary halo is 6 8 10 12 14 16 18 20
Amplitude [σr]
constrained by the primary collimators. For two
orthogonal primary collimators at n1 (usually in
Figure 10: Extension of secondary and tertiary beam
horizontal and vertical orientation) [13]: halo at a zero dispersion location, as simulated for
√
Amax
1 = 2 n1 ≈ 1.41 n1 (8) the LHC multi-stage collimation system at 7 TeV with
n1 = 6 and n2 = 7.
For three primary collimators with 45◦ coverage
(horizontal, vertical, skew) Amax
1 becomes:
30 jaws). The system is only efficient if the sec-
Amax
1 = n1 / cos(π/8) ≈ 1.08 n1 (9)
ondary collimators are placed at the correct phase
It is therefore advantageous to design a system advances from the primary collimator. The best
with three primary collimators. The need for the solution requires a modulated optics with unequal
skew collimator must be assessed with a study on horizontal and vertical phase advance. Detailed
the expected beam halo densities: skew collima- solutions and tables for optimal theoretical phase
tors are effective in cases where there are many advance locations are listed in [13]. The simu-
halo particles with simultaneously large horizon- lated shapes of secondary and tertiary halos are
tal and vertical amplitudes. shown in Fig.10 for 7 TeV LHC collimation [14],
The optimal number, orientation and phase here at a location of zero dispersion.
advance locations of secondary collimators can be It is noted that the quoted extension of the
calculated with full collimation theory. The exten- secondary beam halo is correct for particles that
sion of secondary beam halo is quoted in Tab.1 as are scattered in a jaw, but do not lose signifi-
a function of the number Nsec of secondary colli- cant amounts of energy. This is not fully realis-
mators per primary collimator. tic. Even at multi-TeV beam energies some par-
The secondary beam halo can be efficiently ticles will lose a significant energy in the primary
constrained by placing four secondary collimators jaw without being stopped. These particles will
per primary collimator. The betatron collimation escape as off-momentum particles. At locations
system then consists of three primary and 12 sec- of high dispersion the particles’ amplitude will
ondary collimators (in total 15 collimators with extend beyond Amax 2 . This is shown in Fig.11
305
1
0.1
Secondary halo n1
Normalized population
n1
0.01
0.001
0.0001 Tertiary halo
1e-005
1e-006 n1
n1
1e-007
6 8 10 12 14 16 18 20
Amplitude [σr]
Figure 11: The same simulation as shown in Fig.10,

but now at a location of maximum dispersion in the Figure 12: Simulation of multi-turn cleaning in the
LHC. It is seen that the secondary halo contains off- LHC storage ring. The vertical phase space is shown at
energy protons. These originate mostly from single- the start, after 1, 5 and 50 turns. Primary collimation at
diffractive scattering in the primary collimator jaw. n1 is indicated. The collimator jaw surfaces show up
as straight lines which define the scattering locations
(see 1 turn).
for the same case as in Fig.10. It is therefore im-
portant to append betatron collimators to an off-
momentum collimation section. Performance The inefficiency ηc of a collima-
tion system is defined as the ratio between the
Simulation of beam halo and collimation number Nleak of particles that leak out with az >
Modern collimation systems are designed based acut
z and the number Nimpact of impacting parti-
on simulations of optics, phase space coverage, cles:
impedance, beam halo and cleaning efficiency: Nleak (az > acut
z )
(i) Accelerator optics programs are used to ηc = (10)
Nimpact
design the collimation insertion and its optics,
minimizing the extent of downstream halo ampli- with acut
z > n1 . The value of acut z is given by
tudes [16]. the available machine aperture and is often around
(ii) Special routines are used to constrain im- 10 sigma. Modern collimation systems can reach
pedance from collimators at the same time, trying quite low inefficiencies with ηc in the range of
to place collimators at the locations of larger beta 10−2 (1%) to 10−4 (0.01%). Efficiency is de-
values. fined as η = 1 − ηc and is in the range of 99%
(iii) Tracking programs are used to track up to 99.99%.
to 20 million halo particles for several 100 turns, Inefficiency is not sufficient to characterize
while simulating the particle interaction in col- the performance of collimation. Large amplitude
limator materials [17, 18]. The calculated effi- particles are not lost at one location but are spread
ciency is passed back into (i) for a next design over some dilution length Ldil . A local cleaning
iteration. Simulations with reduced numbers of inefficiency η̃c [1] is defined as
collimators identify the most and least efficient ηc
collimators, allowing further optimization. η̃c = (11)
Ldil
(iv) Special aperture analysis programs pass
with units of 1/m. Dilution is not uniform and
the halo found in (iii) through a model of the ma-
simulations are used to predict the local cleaning
chine hardware to predict the loss patterns. Partic-
inefficiency η̃c along the accelerator. Local clean-
ular loss peaks in sensitive equipment are identi-
fied and local collimators can be placed for further ing inefficiency is minimized by reducing global
enhancing collimation efficiency. inefficiency (overall leakage) and maximizing di-
lution. It is crucial to work on both aspects for
An example of multi-turn tracking for verti-
achieving best performance.
cal halo in the LHC is shown in Fig.12. These
A maximum local cleaning inefficiency
simulation data yielded the results of Figs.10
max[η̃c ] is defined over all critical locations.
and 11.
For example, in a super-conducting storage ring
306

max[η̃c ] describes the peak loss per m in super- VXSHUFRQGXFWLQJUHJLRQV
EHDP EHWDWURQFOHDQLQJ
FROOLPDWRUV
conducting magnets. Alternatively, in a linac
5HODWLYH/RVV0HDVXUHPHQW
URRPWHPSHUDWXUHUHJLRQV

max[η̃c ] may describe the peak loss per m in the GXPSSURWHFWLRQ
,5
regions that must be protected for hands-on main- PRPHQWXPFOHDQLQJ

tenance. ,5 ,5
,5
The maximum allowed loss rate Rloss at the H ,5
primary collimators then depends on the allow- H
able maximum beam loss rate Rlim for critical lo- H
cations and the maximum local inefficiency of the H

system [1]: /RQJLWXGLQDO5LQJ3RVLWLRQ>P@
Rlim Figure 13: Example for a measurement of collimation

Rloss = (12) performance in the LHC at 450 GeV. The data shows
max[η̃c ]
peak integrated losses over 1.3 s. A beam loss is pro-
Considering a stored beam and assuming that voked for beam 1 in the horizontal plane (emittance
particles are lost at collimators we can relate blowup). Losses are normalized to the peak loss in
Rloss = ΔN/ΔT to the number of particles the ring. The peak loss appears as expected at the be-
Nmax and beam lifetime τmin : tatron collimators and falls off exponentially over the
betatron cleaning insertion. Leakage around the ring
ΔT Nmax is measured. The measurement resolution is limited
τmin = − ≈ (13) by noise in the beam loss monitors (6 orders of magni-
Rloss ·ΔT Rloss
ln 1 − Nmax tudes below the peak loss).
The maximum achievable beam intensity can
be expressed as a function of the maximum local
cleaning inefficiency, the minimum beam lifetime cut is present all around the ring after phase space
that must be sustained and the limit of beam loss mixing (particles oscillating around the closed or-
in critical regions [1]: bit, sweeping around the whole allowed phase
space volume). The second jaw of the reference
τmin · Rlim collimator can then be moved to the same cut. A
Nmax = (14) sudden spike in beam loss measured downstream
max[η̃c ]
of the collimator is used to detect the halo edge.
This equation is used during the design phase
Successively all collimators around the ring are
to determine the required collimation perfor-
set up to the same cut in normalized phase space.
mance once beam intensity, minimum beam life-
In the end all jaws are centered on the beam and
time and loss limits have been fixed. Similar
any beta variations have been calibrated.
equations can be introduced for single-pass accel-
erators. Measurement of collimation performance
Beam-based setup of collimation Collimators Collimation performance can be measured if
must be centered around the beam with an accu- a distributed beam loss measurement (BLM)
racy that is a fraction of the collimator retraction. system has been installed around the ring. Ideally
Tolerances for collimator settings can then be in beam loss is measured at all collimators, all
the range of a few 10’s of μm. However, the exact quadrupoles (here the beta functions are max-
beam position and size are not known a priori with imal) and other critical locations. An example
this accuracy. Collimators are therefore set up in a measurement of collimation performance in
beam-based process. This beam-based procedure the LHC at 450 GeV is shown in Fig.13. The
differs for one-pass or stored beams. procedure for such a measurement is described:
In a single pass accelerator or in transfer lines (i) The measurement shall not disturb the or-
a collimator jaw is moved through the beam while bit nor the beta functions, as this would decrease
observing the transmission of beam and down- the cleaning efficiency. Therefore one induces a
stream showers due to beam impact on the colli- strong diffusion process that rapidly increases the
mator material. In a storage rings a reference col- beam emittance. In a storage ring one can move
limator (most often a primary collimator) is used the beam onto a resonance, for example the 1/3
to define a betatron cut in normalized phase space. resonance, or one can blow up individual bunches
Assuming zero dispersion, the same phase space by gated noise excitation.
307
(ii) The integrated beam losses are moni- 3.3.12 Atomic and Nuclear Properties of
tored around the ring as the beam emittance is Materials [1]
blown up.
(iii) The loss data is normalized to the high- The table on the next page is abridged from
pdg.lbl.gov/AtomicNuclearProperties
est loss all around the ring, which by construction
should occur at a collimator. by D.E. Groom (2007). See web pages for
Such a measurement is shown in Fig.13 [19] more detail about entries in this table including
and is closely related to the local cleaning in- chemical formulae, and for several hundred other
efficiency defined above. Differences arise due entries. Quantities in parentheses are for NTP
to BLM response characteristics. The measured (20◦ C and 1 atm), and square brackets indicate
maximum “local cleaning inefficiency” in a crit- quantities evaluated at STP. Boiling points are at
1 atm. Refractive indices n are evaluated at the
ical region (super-conducting magnets) is about

2 × 10−5 in Fig.13. It is seen that leakage is very sodium D line blend (589.2 nm); values 1 in
small. brackets are for (n − 1) × 106 (gases).
References References
[1] K. Nakamura et al, (Particle Data Group), J. Phys.
[1] R. Assmann, CERN-AB-2003-008 ADM (2003)
G37, 075021 (2010), and 2011 partial update for
[2] N. Mokhov et al, FERMILAB-Conf-03/220
the 2012 edition
[3] M. Seidel, DESY 94-103 (1994)
[4] P.J. Bryant et al, CERN SL/93-15(AP) (1993)
[5] R. Assmann et al, CERN LHC PR 592 (2002)
[6] D. Onoprienko et al, SLAC-PUB-10192 (2002)
[7] E. Metral et al, CERN LHC PR 1015 (2007)
[8] F. Zimmermann, CERN-AB-Note-2006-007
[9] C. Bracco, R. Assmann, CERN-ATS-2009-033
[10] S. Di Mitri, PRST-AB 13, 052801 (2010)
[11] P.J. Bryant, E. Klein, CERN SL/92-40(AP)
(1992)
[12] T. Trenkler, J.B. Jeanneret, PA 50 (1995) 287
[13] J.B. Jeanneret, PRST-AB 1, 081001 (1998)
[14] R. Assmann et al, CERN LHC PR 919 (2006)
[15] P.J. Bryant, CERN SL/92-24(AP) (1992)
[16] D. Kaltchev et al, CERN LHC PR 134 (1997)
[17] G. Robert-Demolaize et al, CERN-AB-2005-033
(2005)
[18] S. Redaelli et al, CERN LHC PR 938 (2006)
[19] D. Wollmann et al, IPAC 10
308
Material Z A dZ/A Nucl.coll. Nucl.inter. Rad.len. dE/dx|min Density Melting Boiling Refract.
length λT length λI X0 MeV {g cm−3 } point point index
{g cm−2 } {g cm−2 } {g cm−2 } {g−1 cm2 } ({g−1 }) (K) (K) (@ Na D)
H2 1 1.00794(7) 0.99212 42.8 52.0 63.04 (4.103) 0.071(0.084) 13.81 20.28 1.11[132.]
D2 1 2.01410177803(8) 0.49650 51.3 71.8 125.97 (2.053) 0.169(0.168) 18.7 23.65 1.11[138.]
He 2 4.002602(2) 0.49967 51.8 71.0 94.32 (1.937) 0.125(0.166) 4.220 1.02[35.0]
Li 3 6.941(2) 0.43221 52.2 71.3 82.78 1.639 0.534 453.6 1615.
Be 4 9.012182(3) 0.44384 55.3 77.8 65.19 1.595 1.848 1560. 2744.
C diamond 6 12.0107(8) 0.49955 59.2 85.8 42.70 1.725 3.520 2.42
C graphite 6 12.0107(8) 0.49955 59.2 85.8 42.70 1.742 2.210

N2 7 14.0067(2) 0.49976 61.1 89.7 37.99 (1.825) 0.807(1.165) 63.15 77.29 1.20[298.]
O2 8 15.9994(3) 0.50002 61.3 90.2 34.24 (1.801) 1.141(1.332) 54.36 90.20 1.22[271.]
F2 9 18.9984032(5) 0.47372 65.0 97.4 32.93 (1.676) 1.507(1.580) 53.53 85.03 [195.]
Ne 10 20.1797(6) 0.49555 65.7 99.0 28.93 (1.724) 1.204(0.839) 24.56 27.07 1.09[67.1]
Al 13 26.9815386(8) 0.48181 69.7 107.2 24.01 1.615 2.699 933.5 2792.

Si 14 28.0855(3) 0.49848 70.2 108.4 21.82 1.664 2.329 1687. 3538. 3.95
Cl2 17 35.453(2) 0.47951 73.8 115.7 19.28 (1.630) 1.574(2.980) 171.6 239.1 [773.]
Ar 18 39.948(1) 0.45059 75.7 119.7 19.55 (1.519) 1.396(1.662) 83.81 87.26 1.23[281.]
Ti 22 47.867(1) 0.45961 78.8 126.2 16.16 1.477 4.540 1941. 3560.
Fe 26 55.845(2) 0.46557 81.7 132.1 13.84 1.451 7.874 1811. 3134.
Cu 29 63.546(3) 0.45636 84.2 137.3 12.86 1.403 8.960 1358. 2835.
Ge 32 72.64(1) 0.44053 86.9 143.0 12.25 1.370 5.323 1211. 3106.
Sn 50 118.710(7) 0.42119 98.2 166.7 8.82 1.263 7.310 505.1 2875.
Xe 54 131.293(6) 0.41129 100.8 172.1 8.48 (1.255) 2.953(5.483) 161.4 165.1 1.39[701.]
W 74 183.84(1) 0.40252 110.4 191.9 6.76 1.145 19.300 3695. 5828.
Pt 78 195.084(9) 0.39983 112.2 195.7 6.54 1.128 21.450 2042. 4098.
Au 79 196.966569(4) 0.40108 112.5 196.3 6.46 1.134 19.320 1337. 3129.
Pb 82 207.2(1) 0.39575 114.1 199.6 6.37 1.122 11.350 600.6 2022.
U 92 [238.02891(3)] 0.38651 118.6 209.0 6.00 1.081 18.950 1408. 4404.
Air (dry, 1 atm) 0.49919 61.3 90.1 36.62 (1.815) (1.205) 78.80
Shielding concrete 0.50274 65.1 97.5 26.57 1.711 2.300
Borosilicate glass (Pyrex) 0.49707 64.6 96.5 28.17 1.696 2.230
Lead glass 0.42101 95.9 158.0 7.87 1.255 6.220
Standard rock 0.50000 66.8 101.3 26.54 1.688 2.650
Methane (CH4 ) 0.62334 54.0 73.8 46.47 (2.417) (0.667) 90.68 111.7 [444.]
Ethane (C2 H6 ) 0.59861 55.0 75.9 45.66 (2.304) (1.263) 90.36 184.5
Propane (C3 H8 ) 0.58962 55.3 76.7 45.37 (2.262) 0.493(1.868) 85.52 231.0
Butane (C4 H10 ) 0.59497 55.5 77.1 45.23 (2.278) (2.489) 134.9 272.6
Octane (C8 H18 ) 0.57778 55.8 77.8 45.00 2.123 0.703 214.4 398.8
Paraffin (CH3 (CH2 )n≈23 CH3 ) 0.57275 56.0 78.3 44.85 2.088 0.930
Nylon (type 6, 6/6) 0.54790 57.5 81.6 41.92 1.973 1.18
Polycarbonate (Lexan) 0.52697 58.3 83.6 41.50 1.886 1.20
Polyethylene ([CH2 CH2 ]n ) 0.57034 56.1 78.5 44.77 2.079 0.89
Polyethylene terephthalate (Mylar) 0.52037 58.9 84.9 39.95 1.848 1.40
Polyimide film (Kapton) 0.51264 59.2 85.5 40.58 1.820 1.42
Polymethylmethacrylate (acrylic) 0.53937 58.1 82.8 40.55 1.929 1.19 1.49
Polypropylene 0.55998 56.1 78.5 44.77 2.041 0.90
Polystyrene ([C6 H5 CHCH2 ]n ) 0.53768 57.5 81.7 43.79 1.936 1.06 1.59
Polytetrafluoroethylene (Teflon) 0.47992 63.5 94.4 34.84 1.671 2.20
Polyvinyltoluene 0.54141 57.3 81.3 43.90 1.956 1.03 1.58
Aluminum oxide (sapphire) 0.49038 65.5 98.4 27.94 1.647 3.970 2327. 3273. 1.77
Barium flouride (BaF2 ) 0.42207 90.8 149.0 9.91 1.303 4.893 1641. 2533. 1.47
Bismuth germanate (BGO) 0.42065 96.2 159.1 7.97 1.251 7.130 1317. 2.15
Carbon dioxide gas (CO2 ) 0.49989 60.7 88.9 36.20 1.819 (1.842) [449.]
Solid carbon dioxide (dry ice) 0.49989 60.7 88.9 36.20 1.787 1.563 Sublimes at 194.7 K
Cesium iodide (CsI) 0.41569 100.6 171.5 8.39 1.243 4.510 894.2 1553. 1.79
Lithium fluoride (LiF) 0.46262 61.0 88.7 39.26 1.614 2.635 1121. 1946. 1.39
Lithium hydride (LiH) 0.50321 50.8 68.1 79.62 1.897 0.820 965.
Lead tungstate (PbWO4 ) 0.41315 100.6 168.3 7.39 1.229 8.300 1403. 2.20
Silicon dioxide (SiO2 , fused quartz) 0.49930 65.2 97.8 27.05 1.699 2.200 1986. 3223. 1.46
Sodium chloride (NaCl) 0.55509 71.2 110.1 21.91 1.847 2.170 1075. 1738. 1.54
Sodium iodide (NaI) 0.42697 93.1 154.6 9.49 1.305 3.667 933.2 1577. 1.77
Water (H2 O) 0.55509 58.5 83.3 36.08 1.992 1.000(0.756) 273.1 373.1 1.33
Silica aerogel 0.50093 65.0 97.3 27.25 1.740 0.200 (0.03 H2 O, 0.97 SiO2 )
309
Material Dielectric Young’s Coeff. of Specific Electrical Thermal

constant (κ = /0 ) modulus thermal heat resistivity conductivity
() is (κ–1)×106 [106 psi] expansion [cal/g-◦ C] [μΩcm(@◦ C)] [cal/cm-◦ C-sec]
for gas [10−6 cm/cm-◦ C]
H2 (253.9) — — — — —
He (64) — — — — —
Li — — 56 0.86 8.55(0◦) 0.17
Be — 37 12.4 0.436 5.885(0◦) 0.38
C — 0.7 0.6–4.3 0.165 1375(0◦) 0.057

N2 (548.5) — — — — —
O2 (495) — — — — —
Ne (127) — — — — —
Al — 10 23.9 0.215 2.65(20◦) 0.53
Si 11.9 16 2.8–7.3 0.162 — 0.20
Ar (517) — — — — —
Ti — 16.8 8.5 0.126 50(0◦ ) —
Fe — 28.5 11.7 0.11 9.71(20◦) 0.18
Cu — 16 16.5 0.092 1.67(20◦) 0.94
Ge 16.0 — 5.75 0.073 — 0.14
Sn — 6 20 0.052 11.5(20◦) 0.16
Xe — — — — — —
W — 50 4.4 0.032 5.5(20◦) 0.48
Pt — 21 8.9 0.032 9.83(0◦) 0.17
Pb — 2.6 29.3 0.038 20.65(20◦) 0.083
U — — 36.1 0.028 29(20◦) 0.064
310
Chapter 4. OPERATIONAL CONSIDERATIONS
4.1 LUMINOSITY In the y plane, the beam-beam tune shift pa-

M.A. Furman, M.S. Zisman, LBNL[1] rameter of an on-axis particle in the positron beam
due to its interaction with the opposing electron
Definition When two bunches (+ and −) hav- beam is
by UNIVERSITY OF BIRMINGHAM on 08/30/15. For personal use only.
ing N± particles and distributions ρ± (x, t) (nor- ∗

re N− βy,+
malized by d3 xρ± (x, t) = N± ) collide, the ξy,+ = ∗ ∗ ∗ ) (2)
2πγ+ σy,− (σx,− + σy,−
single-collision luminosity Lsc for a particular re-
action is defined as the number of events produced Expressions for the remaining 3 tune shift param-
per unit cross section σ for this reaction, and is eters are obtained by x ↔ y and/or + ↔ −.
given by the overlap integral [2, 3] Transparency symmetry In a two-ring e+ e−
collider, beam parameters need not be identical
1
Lsc = d3 xdt ρ+ (x, t)ρ− (x, t) in both rings. In the initial design of the two
c
asymmetric B factories PEP-II [6] and KEKB
× c2 (v+ − v− )2 − (v+ × v− )2 (1) [7], it was suggested to restrict the available
where v+ (v− ) is the common velocity of all the parameter space [8, 9, 10] by choosing values
particles in bunch +(−). Eq.(1) is a relativistic to mimic the situation in a symmetric (single-
invariant, has dimensions of 1/area, and is valid ring) collider. The “transparency” conditions thus
for arbitrary velocities v± (see [4] for generaliza- reached for two-ring colliders include: (i) pair-
tion to the case when velocity distributions are not wise equality of beam-beam tune shift param-
homogeneous). eters (ξx,+ = ξx,− ; ξy,+ = ξy,− ); (ii) pair-
wise equality of beam sizes (σx,+ ∗ ∗ ;
= σx,−
For a storage ring collider in which the ∗ ∗ ); (iii) equality of tune modulation
bunches are spaced by a distance sB , bunches σy,+ = σy,−
collide periodically with frequency fc = βc/sB amplitudes associated with synchrotron oscilla-
∗ ) = (σ ν /β ∗ ) ; and some-
tions ((σz νs /βx,y
yielding Ṅ reaction events per unit time. For a + z s x,y −
linear collider, fc = (repetition rate)×(number of times (iv) equality of radiation damping decre-
bunches per bunch train). The peak luminosity is ments for the two rings. In particular, choosing
given by L = Ṅ /σ = fc Lsc [5]. It is traditionally parameters such that all four beam-beam param-
expressed in cgs units, cm−2 s−1 . eters are equal is called “optimal coupling.” This
Tab.1 gives expressions for L in various situ- case requires [6, 8]
∗ ∗ ∗ ∗
ations for head-on e+ e− collisions1 and σz small σy σy βy βy
compared to βx∗ , βy∗ . These expressions are valid σx∗ +
=
σx∗ −
=
βx∗ +
=
βx∗ −
even with nonzero dispersion at the IP, unless oth-
y y
erwise noted. For initial estimates using Tab.1, we = = ≡r (3)
assume the nominal emittances and beam sizes, x + x −
but these values generally evolve in time owing to Operational experience at PEP-II and KEKB
the beam-beam force, and the luminosity should [11], as well as at the Tevatron and HERA,
be modified accordingly [6]. has shown that transparency symmetry is not an
1
For particle species other than e+ e− these expressions essential requirement for stable operation; the
remain valid if one replaces the classical electron radius observed beam-beam dynamics generally varies
re by the appropriate value, provided the colliding particle smoothly when conditions depart from this sym-
species have equal mass. metry.
311
Sec.4.1: LUMINOSITY
Table 1: Head-on luminosity expressions for short e+ e− upright Gaussian bunches.
Expression for L Conditions for validity

N+ N− fc
general
2π ∗2 + σ ∗2 )(σ ∗2 + σ ∗2 )
(σx,+ x,− y,+ y,−
N+ N− fc ∗ ∗
σx,+ = σx,− ≡ σx∗ , σy,+
∗ ∗
= σy,− ≡ σy∗
4πσx∗ σy∗
∗ ∗
EI σx,+ = σx,− ≡ σx∗ , σy,+
∗ ∗
= σy,− ≡ σy∗ ,
K(1 + r)ξy
βy∗ +,− ξx,+ = ξx,− ≡ ξx , ξy,+ = ξy,− ≡ ξy
∗ ∗ ≡ σx∗ , σy,+
∗ ∗ ≡ σy∗ ,
EI σx,+ = σx,− = σy,−

K(1 + r) ξ ∗
βy +,− ξx,+ = ξy,+ ≡ ξ+ , ξx,− = ξy,− ≡ ξ−
∗
σx,+ ∗
= σx,− ∗
= σy,+ ∗ ,
= σy,−
N fc γξ
∗
βx,+ ∗
= βx,− ∗
= βy,+ ∗
= βy,− ≡ β∗,
re β ∗ N+ = N− ≡ N, E+ = E− ≡ E
x,+ = x,− = y,+ = y,− ≡ ,
N 2 fc ∗ ∗ ∗ ∗
βx,+ = βx,− = βy,+ = βy,− ≡ β ∗ , N+ = N− ≡ N,
4πβ ∗ E+ = E− ≡ E, Dx,± ∗ ∗
= Dy,± =0
∗
x,+ = x,− ≡ x , σy,+ ∗ ,
= σy,−
πfc γ 2 x ξx ξy (1 + r)2 ∗ ∗
βx,+ = βx,− ≡ βx∗ , βy,+
∗ ∗
= βy,− ≡ βy∗ ,
re2 βy∗ N+ = N− ≡ N, E+ = E− ≡ E, Dx,± ∗ =0
Alternative expressions Because the luminos- factor for Gaussian beams is [13]
ity in a circular collider is limited by the value of L
ξ, it is useful to write L explicitly in terms of ξ R(tx , ty ) ≡
L0
as seen in the third row of Tab.1. Here E and I ∞
are the beam energy and total beam current in one dt exp(−t2 )
= √ (5)
ring and K = 1/(2e3 ) = 1/(2ere me c2 ). With π (1 + t2 /t2 )(1 + t2 /t2 )
E in GeV, I in A, βy∗ in cm, and L in cm−2 s−1 , −∞ x y
we have K = 2.17 × 1034 . The symbol ( )+,− with

means that the enclosed parameters may be taken ∗2 + σ ∗2 )
2(σx,+ x,−
from either beam, on account of the transparency t2x =
2 + σ 2 ) σ ∗2 /β ∗2 + σ ∗2 /β ∗2

(σz,+ z,− x,+ x,+ x,− x,−
conditions.
For a linear collider, and correspondingly for ty . The nominal lumi-
nosity, L0 , is that represented by Tab.1, which
H N P
L= (4) corresponds to the limit σz → 0. See Fig.1 [13].
4πE σx∗ σy∗ Another reduction factor comes from a non-
where N = N+ = N− , P is the aver- zero crossing angle, typically in the horizontal
age beam power, and H is the pinch enhance- plane. For the symmetric-collider case with σy∗
ment factor [12]. The factor N/σx∗ determines σx∗ we obtain [14]
the number of beamsstrahlung photons emit-
L 2 b
ted (constrained by background considerations). RL ≡ = ae K0 (b), (6a)
L0 π
The factor P/σy∗ represents the major technical ∗
challenge—providing high beam power and small βy
a= √ , (6b)
bunch size. 2σz
2
Reductions and enhancements of the luminos- σz
ity When σz > ∗
∼ β , the luminosity decreases rel- b = a2 1 + tan φ (6c)
σx∗
ative to the case when σz β ∗ due to the re-
duced overlap of the two bunches during the col- where K0 is a Bessel function and φ is half the
lision. This geometrical (“hourglass”) reduction crossing angle. For the symmetric-collider case
312
Ch.4: OPERATIONAL CONSIDERATIONS
energy and vice versa. Thus the energy spread

of the c.m. of any given colliding e+ e− pair is
effectively reduced relative to the standard (zero
dispersion) case, hence the name “monochroma-
tization” [18].
Crab crossing scheme The simplest design for
the interaction region (IR) of all particle colliders
implies that the beams cross at an angle, hence the
luminosity is subject to the reduction factor RL ,
Eq.(8), relative to the idealized case φ = 0. The
PEP-II design [6] was perhaps the only exception,
achieving φ = 0 by means of a rather unusual

IR geometry that involved additional vertical and
Figure 1: Hourglass reduction factors, Eq.(5). horizontal bends in the positron beam. In the
more generic case, however, reducing φ implies

with round beams σy∗ = σx∗ and βx∗ = βy∗ ≡ β ∗ long IRs, which may only be practical for very
(typical for hadron colliders) one has [15] large machines such as hadron colliders. Even
then, if the beam is made up of many closely-
cos φ ∞ e−As
2
L spaced bunches, as is the case in the design of
RL ≡ =√ ds ,
L0 πσz −∞ 1 + (s/β ∗ )2 all present and future colliders, a small value of
(7a) φ has the drawback of giving rise to a number
sin2 φ cos2 φ of undesirable “parasitic,” or “long-range,” colli-
A≡ + . (7b) sions between the bunches in the opposing beams
σx∗ 2 (1 + (s/β ∗ )2 ) σz2
as they travel towards the IP in a common vac-
When σz βy∗ , Eqs.(6) and (7) reduce to [4] uum pipe. A way to compensate for φ is to tilt
−1/2 the bunches by an angle φ such that they will col-
2
σz lide with zero effective crossing angle. Such tilt-
RL = 1 + tan φ (8)
σx∗ ing can be achieved by applying a time-dependent
transverse dipolar kick to the bunches as they ap-
If the beams are, in addition, offset transversely proach the IP [19, 20, 21]. See Sec.4.13.
by δx and δy, Eq.(8) acquires the extra multiplica-
tive factor Crab waist scheme An alternative to the crab-
' ; crossing scheme has been developed and success-
(δx/2)2 δy 2 fully tested at the DAΦNE e+ e− collider. In this
exp − ∗2 −
σx cos2 φ + σz2 sin2 φ 2σy∗ scheme the bunches collide in the x − z plane
(x=horizontal dimension, z=longitudinal). The
Monochromatization If the purpose of the col- bunches are tilted so that φ is relatively large, but
lider is the study of a very narrow resonance, i.e. the collision is arranged such that the longitudinal
narrow compared to the beam energy spread, such location of the vertical waist is a function of x.
as a τ -charm factory, it is possible to enhance Here all components of the bunch collide at the
the luminosity by means of a monochromatiza- minimum of βy , hence the hourglass reduction is
tion scheme. For example, the J/ψ resonance has lessened, and so are detrimental beam-beam dy-
a width-to-mass ratio Γ/mc2 = 2.8× 10−5 , small namical effects [22, 23, 24]. See Sec.4.13.
compared to the energy spread of the beam, typ-
ically σδ < −3
∼ 10 . The improved energy resolu- Peak and average luminosity The history of
tion also allows detailed measurement of thresh- peak luminosity achieved or designed for many
olds and branching rates in the e+ e− c.m. en- colliders is illustrated in Figs.2 and 3. Roughly
ergy range 3 − 5 GeV [16, 17]. The scheme calls speaking, the typical peak luminosity for e+ e−
for large y dispersion such that Dy,+∗ = −Dy,−∗ , colliders is 1–2 orders of magnitude larger than
which induces a correlation between the y posi- for hadron colliders. This difference can be at-
tion of the particles and their energy in such a way tributed to several factors: leptons, being point-
that e+ ’s with higher-than-average energy prefer- like objects, tend to yield cleaner reactions upon
entially collide with e− ’s with lower-than-average collision than hadrons, which are composite.
313
Sec.4.1: LUMINOSITY
length of time during which the beams are in colli-

sion). Following injection, the luminosity decays
in time due to particle losses from various sources.
If it takes a time tf to refill the machine and bring
the beams back into collision, during which time
the beams are not colliding, one wants to deter-
mine the value of tc that leads to the largest av-
erage luminosity. If we make the approximation
L(t) = L0 exp(−t/τ ) where τ is the characteris-
tic lifetime, then the average luminosity is given
by2
tc
1 1 − e−x
L = dtL(t) = L0 (9)
tc + tf x+a
Figure 2: Peak luminosity of e+ e− colliders. 0
where x = tc /τ and a = tf /τ . If tf is indepen-

dent of the number of particles left in the machine
Therefore, leptons can be collided at a lower en- at the end of the luminosity run, the equation for tc
ergy than hadrons to study specific physical pro- (or, equivalently, x) that maximizes L that fol-
cesses. As a result, lepton colliders tend to be lows from (9) is [25]
smaller, hence less costly, than hadron collid- ex = 1 + x + a (10)
ers for a given physics-equivalent beam energy. Given a, Eq.(10) can be readily solved numeri-
This advantage of lepton colliders diminishes as cally by iterating the map xn+1 = ln(1 + xn + a)
the beam energy increases owing to the much starting from any x0 > −a. An approximate so-
stronger emission of synchrotron radiation than lution is given by√the first iteration, x1 , assuming
hadrons, related to the relative lightness of the lep- the choice x0 = 2a,
tons. Thus, the LHC has a much wider physics √
x ln 1 + 2a + a
reach than LEP, which was of the same size as the
LHC, owing to the overwhelming effect from syn- This expression becomes exact when a → 0. For
chrotron radiation at LEP. As a positive side effect larger a, its relative error is at most ∼ 7%, and
of the higher synchrotron radiation, lepton beams this worst case occurs for a 1.1. Thus, if the
are intrinsically more stable than hadron beams, condition (10) is satisfied, the maximum average
hence lepton colliders can typically accommodate luminosity is
higher beam intensities than their hadron counter- L
Lmax = L0 e−x √0 (11)
parts. In addition, electron beams are typically 1 + 2a + a
less expensive to produce than hadron beams. If tf does depend on the number of particles
While peak luminosity is typically a basic left in the machine at the end of the prior luminos-
performance parameter specified a priori in the ity run, the optimal condition is, of course, more
collider design, it is the average luminosity, or its complicated [25], although a similar analysis is
integral over a period of time, that ultimately de- applicable.
fines the success of the detailed physics obtained Figs.4, 5, 6 and 7 show the luminosity his-
from the collider. To account for down time, tory of CESR, PEP-II, KEKB and DAΦNE, re-
injection, beam lifetimes, etc., one experimental spectively; Figs.8, 9 and 10 give the luminosity
“year” is taken by convention to be 107 s. Then, history of the Tevatron and RHIC; and Figs.11, 12
the expected integrated luminosity for a collider and 13 the luminosity history of the LHC during
−2
luminosity of 1 × 10 cm
delivering a peak 33 2011.
s−1 would be Ldt = 1 × 1040 cm−2 = 10 The time evolution of the luminosity perfor-
fb−1 =10,000 pb−1 (1 b= 10−24 cm2 ). In practice, mance of many past and present colliders ex-
however, the long-term integrated luminosity de- hibits systematic features that can be interpreted
pends on many more variables than peak luminos- in terms of the complexity of the machine [27].
ity, and many of them are unrelated to beam dy-
namics. Perhaps the simplest variable that can be 2
The exponential decay is a convenient approximation;
controlled is the duration of a “store” tc (i.e., the for a more detailed analysis see M.S. Zisman, Sec.3.3.2.2.
314
Figure 3: Peak luminosity of hadron and e-p colliders. Arrows indicate polarized beams.
104 CESR/CLEO Luminosity History 105

Nov. 1979 - Mar. 2008
Last updated:
4/8/2008
10:04
PEP-II Monthly Integrated Luminosity
Peak (left, 10**30 cm**-2 s**-1) 21
Monthly integral (left, pb**-1) 4 20
Total integrated (right, pb**-1) 10 19
103 18
17
16
15
103 14
13
102
12
fb−1
11
10
102 9
8
7
6
101 5
101
4
3
2
1
0
Jun-99
Jun-00
Jun-01
Jun-02
Jun-03
Jun-04
Jun-05
Jun-06
Jun-07
Mar-99
Sep-99
Dec-99
Mar-00
Sep-00
Dec-00
Mar-01
Sep-01
Dec-01
Mar-02
Sep-02
Dec-02
Mar-03
Sep-03
Dec-03
Mar-04
Sep-04
Dec-04
Mar-05
Sep-05
Dec-05
Mar-06
Sep-06
Dec-06
Mar-07
Sep-07
Dec-07
Mar-08
100 100
1/1/86 1/1/96 1/1/06 Last Updated:
Total PEP-II Delivered Luminosity
Date 4/8/2008 10:04
550
Figure 4: Luminosity history of CESR. The lower lu- 500

557.4
450
minosity after mid-2003 was due to a reduction in 400
beam energy from 5.3 to 1.9 GeV, which was under- 350
taken to run CESR in charm-factory mode [26]. 300

-1
fb
250
200
References 150
100
50
[1] We are grateful to H. Burkhardt, W. Fischer, O. 0
Napoly, F. Porter, A. Zholents and F. Zimmer- Jan-99 Jan-00 Jan-01 Jan-02 Jan-03 Jan-04 Jan-05 Jan-06 Jan-07 Jan-08
mann for valuable comments. We are indebted to

D. Rice, N. Phinney, J. Marriner, V. Shiltsev, R. Figure 5: History of PEP-II luminosity. Top: monthly
Moore, Y. Funakoshi and S. Guiducci for provid- integral; bottom: total integrated.
ing the luminosity data.
[2] C. Møller, K.Danske Vidensk.Selsk.Mat.-
Fys.Medd. 23 (1945) 1 [6] “PEP-II: An Asymmetric B Factory – Concep-
[3] M. Furman, LBNL-53553 (2003) tual Design Report,” LBL-PUB-5379, SLAC-
[4] O. Napoly, PA 40 (1993) 180 418, (1993) p.184
[5] D. Edwards, M. Syphers, An Introduction to the [7] “KEKB B-Factory Design Report,” KEK Report
Physics of High-Energy Accelerators, John Wi- 95-7
ley & Sons, NY, 1993, p.9 [8] A.A. Garren et al, PAC 89, 1847
315
Sec.4.1: LUMINOSITY
103
Tevatron Luminosity History 104
Oct. 1992 Ð Sep. 2011
Peak (left, 10**30 cm**-2 s**-1)
Monthly integral (left, pb**-1)
Total integrated (right, pb**-1)
102
103
1
10
Run II
102
0
10
Run I
10-1 101
1/1/94 1/1/99 1/1/04 1/1/09
date
Figure 8: Luminosity history of the Tevatron. The 4-

year shutdown between Runs I and II was undertaken
to carry out improvements that included: (a) construc-
tion of a new synchrotron (the Main Injector) to in-
crease the proton bunch intensity in the Tevatron, and
the proton flux on the p̄ target; (b) construction of a
new p̄ storage ring (the Recycler), originally intended
to store leftover p̄’s at the end of collider stores; (c)
a variety of improvements to the existing Antiproton
Source, namely doubling the bandwidth of the stochas-
tic cooling systems; changing the transition γ in the
Antiproton Acumulator; and moving the injection lo-
Figure 6: KEKB luminosity history. Top: peak lumi- cation in the Tevatron in order to accept beams from
nosity and other quantities, as indicated. Bottom: total the Main Injector [28].
integrated.
Figure 7: DAΦNE luminosity history. Dots indicate

peak luminosity at the various detectors employed
(left-hand scale). The solid line is the total integrated
luminosity (right-hand scale). The increase in luminos-
ity delivered to SIDDHARTA was achieved with the
crab-waist scheme (see text). Figure 9: RHIC luminosity for polarized proton
beams.
316
Figure 10: RHIC luminosity for ion beams. The “ion-

pair luminosity” LN1 N2 is defined to be N1 N2 L,
where Ni (i = 1, 2) is the number of nucleons per
Figure 12: LHC integrated luminosity delivered to the
ion in beam i, and L is the conventional luminosity for
four detectors during 2011, for p–p collisions.
ion-pair collisions.
Figure 13: LHC integrated luminosity delivered to

Figure 11: LHC peak luminosity for p–p collisions as three of the four detectors during 2011, for Pb82+ –
observed during 2011 at the four detectors. Pb82+ collisions.
317
Sec.4.2: BRIGHTNESS
[9] Y.H. Chin, AIP Conf. Proc. 214 (1990) p.424 When γ 1, the phase space coordinates
[10] S. Krishnagopal, R. Siemann, PR D41 (1990) ≈ (x, γx , y, γy , z, Δγ), where z is the particle
1741 position relative to beam center, and Δγ = γ − γ0
[11] Y. Funakoshi, ICFA Beam Dynamics Newsletter with γ0 energy of the reference particle. Without
No.31 (2003) p.78 acceleration, another convenient set of the phase
[12] K. Yokoya, P. Chen, US-CERN School on Parti- space coordinates is (x, x , y, y , z, Δγ). B is de-
cle Accelerators, Frontiers of Particle Beams: In- fined as the density in the appropriate phase space.
tensity Limitations (Hilton Head Island, S. Car- Assume γ 1 with no acceleration. As-
olina, USA, Nov. 714, 1990) Springer Verlag sume B factorizes in the three dimensions, and
LNP 400, p.415 consider x-dimension. (Extension to general case
[13] M.A. Furman, PAC 91, 422 is straightforward.)
[14] K. Hirata, PRL 74 (1995) 2228
d2 F
[15] B. Muratori, CERN AB-Note-2003-026 (ABP)
B(x, x ; s) = (1)
[16] J. Jowett, Springer-Verlag Lecture Notes in Phys. dxdx
425 (1992), p.79 where F may be considered as the flux or longitu-
[17] J. Kirkby, AIP Conf. Proc. 349, p.11 dinal particle density. The B distributions at two
[18] A. Renieri, Frascati preprint INF-75/6(R), 1975 different s are related by the coordinate transfor-
[19] R. Palmer, SLAC-PUB-4707 (1988), unpub- mation between them,
lished.
[20] K. Oide, K. Yokoya, PR A40 (1989) 315 B x2 , x2 ; s2 = B x1 , x1 ; s1 (2)

[21] K. Ohmi, ICFA Beam Dynamics Newsletter 52 x1 x2
= M−1
(2010) p.33 x1 x2
[22] P. Raimondi, 2nd SuperB Workshop, Frascati,
The spatial and the angular densities of the
Mar. 2006,
flux are
http://www.lnf.infn.it/conference
/superb06/talks/raimondi1.ppt dF
S(x; s) = = B(x, x ; s)dx (3)
[23] P. Raimondi et al, LNF-07/003; arXiv: dx

physics/0702033 dF
A(x ; s) = = B(x, x ; s)dx (4)
[24] M. Zobov et al, PRL 104, 174801 (2010) dx
[25] F.C. Porter, NIM A302 (1991) 209
[26] D. Rice, D. Rubin, ICFA Beam Dynamics F = S (x; s)dx = A (x ; s)dx = B (x, x )dxdx
Newsletter No. 48 (2009) p.152
[27] V. Shiltsev, Mod. Phys. Lett. A 26, 11 (2011) 761 S and A are not invariant along the particle trajec-
[28] S. Holmes et al, JINST 6, T08001 (2011) tory. In the absence of aperture, F is conserved
and is an invariant characterization of the global
4.2 BRIGHTNESS strength of the beam.
P. Elleaume, Deceased 2011 Brightness For a well-designed beam, B is a
smooth function peaked at the phase space ori-
K.-J. Kim, ANL gin. Thus B(origin) is often referred to as the
brightness. A related quantity is the emittance
Particle density in phase space is generally re-
ferred to as the brightness distribution. The (phase space area). The brightness is flux di-
brightness distribution plays an important role in vided by emittance. There are different defini-
tions of the emittance, and hence of the bright-
beam transport calculation and an invariant char-
ness. One definition is the rms emittance [2]
acterization of the source strength. The concept
applies to both particle and photon beams. x = x2 x 2 − xx 2 where means av-
eraging with B as the weight function.
4.2.1 Particle Beam Near the beam waist at s = 0, one may have
Brightness distribution The brightness distri- a Gaussian brightness distribution
# $
bution B is the density in 6-D phase space [1] F 1 x2 x 2

(x, px , y, py , t, E) where t is the arrival time, E is B(x, x ; 0) = exp − + (5)
2πx 2 σx2 σx2
the kinetic energy canonical conjugate to t. When
the system has a Hamiltonian, B is invariant along with brightness B0 = F/(2πx ). See also
each particle trajectory in an accelerator. Sec.3.1.2.
318
Emittance The rms emittance is invariant for polarization state can be defined by introducing
relativistic beam with no acceleration. With ac- four brightness functions in analogy with the four
celeration, the normalized emittance N x = γβx Stokes parameters (Sec.3.1.5.1) defining the po-
is invariant (Sec.2.2.1). larization of incoherent photon beams [5]. In most
For applications involving bunch compres- cases of interest, the spectral brightness is maxi-
sion, consider longitudinal brightness distribution mum on axis of the electron beam (x, x , y, y ) =
[3] (0, 0, 0, 0). In the following we shall restrict the
d2 N discussion to on-axis brightness which only de-
BL (z, Δγ; s) = (6)
dz dΔγ pends on ε and in most cases corresponds to linear
which in the relativistic limit is invariant under polarization with horizontal electric field.
beam compression and acceleration. For a Gaus-
sian distribution, the peak longitudinal brightness 4.2.2.2 Bending magnet radiation
is BL = N/(2πσz σΔγ ) and z = σz σΔγ the nor-
malized longitudinal emittance. We consider an electron beam circulating along
an horizontal planar circular trajectory in a bend-
ing magnet with vertical magnetic field. The on

References
axis brightness BBM of the synchrotron radiation
[1] E.D. Courant, H.S. Snyder, Ann. Phys. 3 (1958) 1 produced in such a bending magnet is expressed
[2] P.M. Lapostolle, CERN-ISR-DI/71-6 (1971) as :
[3] C. Pellegrini, NIM A272 (1988) 364 dΦ 1
BBM = (1)
dΩdε/ε 2π σ 2 + σ 2 σ 2 + σ 2
x R y R
4.2.2 Radiation Beam
Where σx (σy ) is the rms horizontal (vertical)
4.2.2.1 Introduction dΦ
electron beam size at the source point and dΩdε/ε
The concepts in Sec.4.2.1 also apply for the rais the angular spectral flux generated per unit solid
diation beams in geometric optics. In wave op- angle dΩ in the plane of the orbit per unit relative
dΦ
tics, the concept of phase space and brightness do energy bandwidth dε/ε of the photons. dΩdε/ε is
not have direct physical meaning. Nevertheless, it expressed as a function of the electron energy E,
is possible to apply these concepts by identifying the electron current I, the bending magnet field B
the Wigner Distribution of the electric fields as the and photon energy ε according to [6]:
brightness distribution [4]. For a Gaussian mode
dΦ
(TEM00) propagating in free space, Eq.(5) holds. [photons/s/mr 2 /0.1%] = (2)
The rms radiation size and angular divergence are dΩdε/ε
≡ σ = σ (cylindrical 2
σR ≡ σx = σy and σR x y ε 2
K2/3 ε
symmetry). For a coherent Gaussian mode , the εc εc
σ is equal to λ/4π.
1.327 × 1013 E2 [GeV] I[A] 2
“radiation rms emittance” σR R σ
The quantity corresponding to the beta function 1 + σy
R
in radiation optics is called the Rayleigh length
zR = σR /σR . See also Sec.3.1.2. In general, the where K2/3 (y) is modified Bessel function of
brightness distribution of synchrotron radiation order 2/3, εc the critical energy of the bending
is the convolution of single electron brightness magnet radiation, εc [keV]=0.665E2 [GeV]B[T],
(also called diffractive brightness) with the elec- σy the vertical rms divergence of the electron
beam and σR (σ ) the natural rms divergence (rms
tron brightness distribution. The radiation beam R
brightness for synchrotron radiation is usually ex- size) produced by a single electron (diffractive
pressed in units of photons/sec/mm2 /mr2 /0.1% contributions), which depends on photon energy
spectral bandwidth. In the literature, there exist ε = hc/λ, and is approximated as [7]
several denominations such as brightness, spec- −x
320 ε
tral brightness, brilliance or spectral brilliance σR [μrad] = (3)
E[GeV] εc
which all refer to the same quantity. Radiation
brightness is in general a function of the trans- and σR σ ∼ λ/(4π). with x = 0.354 (0.549) if
R =
verse phase space (x, x , y, y ) and photon en- ε/εc < 1(> 1), h the Planck constant and c the
ergy ε. The brightness of radiation with specific speed of light.
319
Sec.4.2: BRIGHTNESS
4.2.2.3 Wiggler radiation the most intense odd harmonics of the spectrum,
it is approximated by [8]:
In a planar sinusoidal vertical field undulator or
wiggler, the electron beam performs an oscillating Φn
BU = (8)
motion characterized by a deflection parameter K (2π)2 Σx Σy Σx Σy
which is related to the spatial period λ0 of the
magnetic field and the peak magnetic field B̂ ac- Where Φn is the total spectral flux generated on
cording to K = 0.0934λ0 [mm]B̂[T]. One differ- the nth harmonic which depends on the electron
entiates two limiting cases namely the undulator current I, number of period N of the magnetic
case (K < 2.5) in which the radiation from each field and deflection parameter K:
period interferes producing an harmonic spectrum Φn [Photons/s/0.1%] (9)
and the wiggler case (K > 3) for which the radia-

nK 2
tion originating from each period has a nearly ran- = 1.431 × 1014 N I[A] K2
dom phase justifying the derivation of brightness 1+ 2
as the simple sum of brightness from each source 2
nK 2 nK 2
point. J n−1 − J n+1
The on axis brightness BW of the wiggler ra-

2 4 + 2K 2 2 4 + 2K 2
diation is [6]
and Σx = σx2 + σR2 and Σ = 2 ;
σx2 + σR
x
dΦ 2N similarly for the y direction. The size σR and di-
BW = (4)
dΩdε/ε 2πΣx Σy vergence σR correspond to single electron emis-

where dΦ
is the angular spectral flux gener- sion (diffractive contribution) with σR = ∼ λ
dΩdε/ε 2L
ated on axis by a bending magnet of field B̂ given and σR ∼ λL
= 8π 2.
by Eq.(2), and N the number of periods of the
As can be seen from Eq.(8), the undulator
magnetic field. Σx and Σy are the effective hori-
brightness increases as the beam size and diver-
zontal and vertical rms source sizes averaged over
gence of the electron beam shrinks. Since the
the length L of the wiggler
brightness is an important figure of merit, syn-
L2 2 chrotron light sources are optimized for a small
Σ2x = σx2 + a2 + 2
σ + σR and
12 x emittance of the electron beam. A facility in-
L2 corporating many undulators in a storage ring
Σ2y = σy2 + σ 2 + σR
2
(5) designed for high radiation brightness featuring
12 y
small emittance, high current, and long straight
with σR defined by Eq.(3) and a the amplitude of sections is referred to as “3rd generation” facility.
the oscillation of the horizontal trajectory inside
the wiggler, which is given by :
4.2.2.5 Brightness comparison
λ0 [mm]K
a[μm] = . (6) Fig.1 compares on axis brightness from a bend-
2π 1.957 E[GeV] ing magnet, a wiggler and an undulator installed
in the ESRF storage ring. Note that the brightness
4.2.2.4 Undulator radiation
from the wiggler is computed for the maximum
The spectrum of the radiation produced by an un- field of 1.5 T. On the other hand the brightness
dulator is made of an harmonic series of peaks. curves for the undulator require to adjust the field
For the most commonly used planar field un- to the targeted photon energy defined in Eq.(7).
dulator, the energy εn of these peaks observed on Each harmonic covers a different photon energy
axis of the trajectory, is given by: range. Note that there exists a lack of overlap
of the spectral range covered by harmonics 1 and
nE 2 [GeV]
εn [keV] = 9.50 (7) 3 which is linked to the low value (K = 1.8)
2
λ0 [mm] 1 + K2 of the U23 undulator. High K undulators pro-
vide increased tunability at the cost of larger in-
Where n is an integer corresponding to the har- tegrated angle, lower energy of the fundamental
monic number. The Brightness BU of undulator peak and higher harmonic content. On the other
radiation is maximum on axis of the trajectory for hand low K undulators provide limited tunability
320
22
10
Undulator U23 4.3 LINAC OPERATION
21 Harm. # 5
10
4.3.1 Operation of High Energy Electron

[Phot/s/.1%/mr /mm ]
2
20
10
Linacs
2
10
19 Undulator U23 Undulator U23
Harm. # 1 Harm. # 3 T.O. Raubenheimer, SLAC
18
10
17
(See Refs.[1, 2, 3].)
10
Brightness
16
Wiggler W150 Linac model The longitudinal equation of mo-
10
tion in a long linac is
10
15 Bending Magnet 0.85T
d e
10
14 γ(s, z) = Erf (s) cos(krf z + φrf )
ds mc2
3 4 5 6 7 8 9 2 3 4 5 6 7 8 9
10keV 100keV
Photon Energy ∞
+4π0 NB re W (s, z − z)λ(z )dz
Figure 1: Comparison of brightness from several z
sources of radiation installed at the ESRF. The elec-
where s, z = longitudinal position in the acceler-

tron beam energy is 6 GeV, the ring current 200 mA,
the horizontal and vertical emittances are 4 nm and
ator and in the bunch, Erf = acceleration gradi-
10 pm. The U23 undulator has 174 periods of 23 mm ent, and krf = rf wavenumber, φrf = rf phase at
and a peak field of 0.83 T. The wiggler W150 has a the bunch center, W = longitudinal wakefunc-
150 mm period and a length of 1.5 m. tion (V/C/m), λ = longitudinal beam distribution.
The transverse equation of motion is
and are useful only on the lowest harmonic num- 1 d d
γ y(s, z) − (1 − δ)K(y − yq )
bers. Note that some brightness can also be ob- γ ds ds
tained on the even harmonics of the spectrum but
their value is usually one or two orders of magni- 1 1−δ
tude lower and it is difficult to derive an analytical = (1 − δ) − 4π0 NB re (1)
ρ γ0
estimate. ∞
The brightness from other third generation is × W⊥ (s, z − z)(y(z) − ya )λ(z )dz
similar but in general shifted to lower photon en- z
ergies for lower electron energy facilities (ALS, where δ(s, z) = (γ − γ0 )/γ, γ0 mc2 = nominal
SLS,..) and to higher photon energies for higher beam energy, K = normalized quadrupole field,
electron energy facilities (APS, SPring-8). The ρ = bending radius, yq , ya = quadrupole and
emerging “4th” generation facilities employing accelerator structure misalignments, and W⊥ =
free electron lasers have much higher peak bright- transverse wakefunction (V/C/m2 ).
ness. Detailed information on the source charac- For emittance dilution, we calculate Δ(γ).
teristics of these facilities can be found in [9]. The projected rms normalized emittance is then

References γ = γ 0 1 + 2 Δ(γ)
γ0 before filamentation
γ0 + Δ(γ) after filamentation
[1] E.D. Courant, H.S. Snyder, Ann.Phys. 3 (1958) 1 (2)
[2] P.M. Lapostolle, CERN-ISR-DI/71-6 (1971) where γ0 is the initial emittance.
[3] C. Pellegrini, NIM A272 (1988) 364 Assume the linac consists of thin-lens FODO
[4] K.-J. Kim, NIM A246 (1986) 71
cells with cell length Lcell (s) ∝ γ(s)α and phase
[5] M. Born, E. Wolf, Principles of Optics, Pergamon
advance per cell φc kept constant. The average
Press (1986)
β-function over the cell is
[6] Undulators, Wigglers and their Applications,
Edited by H. Onuki & P. Elleaume, Taylor and Lcell φc 2 φc
Francis (2003) β= cot + tan (3)
2 2 3 2
[7] K.-J. Kim, X-ray Data Booklet, Center for X-ray
Optics and Advanced Light Source, PUB-490 rev The wakefunctions of a periodic accelerator
2 (http://xdb.lbl.gov/) structure have been parameterized by a number
[8] K.-J. Kim, AIP Proc. 184 (1989) 565 of authors [4, 5, 6]. Here, we will present sim-
[9] http://www.lightsources.org ple expressions that are based on the theoretical
321
Sec.4.3: LINAC OPERATION
short-range dependence plus an empirical term to Injection Errors A beam injected with a trajectory
describe the behavior at longer distances: error (y0 , y0 ) performs a betatron oscillation along
Z c the linac. If the autophasing condition is met, the
W (s) ≈ 0 (4)
πa a2 + 8.6sλrf effect of the transverse wakefields will be mini-
mal. But, if the injected beam has an uncorre-
2Z cs
W⊥ (s) ≈ 0 (5) lated energy spread, it will filament because of the
πa2 a2 + sλrf chromatic dependence of the phase advance. The
where a = iris radius of the structure, λrf = rf emittance dilution due to this error is
wavelength. This model holds over the range s <
∼ γ (1 + α2y ) 2
a/4 and 0.1 < ∼ a/λrf <∼ 0.2. Δ(γ) ≈
y0 + 2αy y0 y0 + βy y02
φrf is chosen to minimize the energy spread 2 βy
(see also Sec.2.4.1). The residual energy spread (8)

after compensation is found from the convolution where βy and αy are evaluated at the injec-
of the bunch with the longitudinal wakefield, and tion point. The case without autophasing energy
the acceleration rf, spread is discussed in Sec.2.4.1.

ΔE 1.25 A similar effect arises if the beam is injected
≈ 2π0 NB re W (1.5σz ) with a β-function mismatch [14],
E G cos φrf

3 Δ(γ) ≈ γ(Bmag − 1) (9)
− σz κrf G sin φrf (FWHM) (6)
2 with
where G = eErf /(me c2 ).
1 β β β β 2
There are other techniques of reducing Bmag ≡ + + α −α
ΔE/E. One example is to shape the longitudinal 2 β β β β
beam distribution [9]; this technique was used in (10)
the SLC linac to reduce ΔE/E by a factor of ∼2- Subscripts denote mismatched lattice parame-
3 [10]. Another approach is to rotate the bunch ters describing the beam. Parameters without
in longitudinal space by > 90◦ part way through denote the natural lattice functions of the linac.
linac. This flips the sign of the energy correla- Accelerator Misalignments Emittance dilution
tion and then the longitudinal wakefield can re- due to random accelerator structure misalign-
move the energy spread that it had previously in- ments is
troduced [11]; this technique allows control of Δ(γ) ≈ ya2 [π0 NB re W⊥ (2σz )]2
ΔE/E of very short bunches. α
Autophasing [7, 8] can be used to control the Lacc β i γf
× −1 (11)
single-bunch beam break-up effect (Sec.2.4.1). In 2αG γi
the case of a FODO lattice, this requires where Lacc is the length of the accelerator struc-
βLcell tures (assumed constant along the linac), β i is the
σδ,auto ≈ π0 NB re W⊥ (2σz ) (7) initial average β-function.
4γ tan(φc /2)
σδ,auto scales as γ 2α−1 and is constant along the Equation (11) assumes that the alignment of
linac for α = 1/2. the individual structures is random. In case all the
Eq.(7) is derived for a rigid offset of the structures between a pair of quadrupoles have a
bunch. A similar condition can be found for a systematic misalignment (e.g. when quadrupoles
bunch offset generated by a misaligned acceler- are used for alignment reference),
ator section. Here, the offset of the bunch de- Δ(γ) ≈ ya2 [π0 NB re W⊥ (2σz )]2
pends on z; the autophasing condition (condition 2α
Lcell i β i γf
to preserve x-z correlation as the bunch propa- × −1 (12)
gates down the linac) is ≈ 1/6 of Eq.(7) [12]. 4αG γi
Emittance dilution effects, single bunch [13] where Lcell i is the initial cell length. In gen-
The primary sources of dilution are transverse eral, Eq.(12) contributes more than Eq.(11) be-
wakefields, dispersive errors, and betatron cou- cause more structures are contributing.
pling. Additional contributions include rf de- BPM Misalignments Consider a 1-to-1 trajectory
flections, beam-gas scattering, quadrupole wake- correction where one minimizes the readings on
fields, and synchrotron radiation. BPMs located at all of the quadrupoles. If these
322
BPM’s are misaligned, the trajectory will be off- i.e., the same roll or same pole error, the dilution
set in the accelerator structures and quadrupoles. will be small because the errors cancel.
Assuming that the autophasing condition is met, RF Deflections Rf deflections occur if the time
one has varying acceleration field is not oriented in the di-
Δ(γ) ≈ yBPM
2
[π0 NB re W⊥ (2σz )]2 rection of beam propagation. The misalignment
2α can arise from a misaligned accelerator structure,
cos ψ2c L2cell i γf an angular trajectory through a structure, or asym-
× − 1 (13)
sin3 ψc 16αG
2
γi metries in a structure such as tilted irises or the
couplers. In addition to deflecting the beam, the rf
If the BPM readings are only located at the deflections will increase the projected emittance
focusing quads, and the defocusing quads are well by causing a deflection that is a function of z,
aligned,
φc β Lacc G γf α
Δ(γ) ≈ yBPM
2
[π0 NB re W⊥ (2σz )]2 tan3 Δ(γ) ≈ g 2 (σz krf )2 i −1
2 α γi
2α (17)
φc L2cell i γf
× 1−sin −1 (14) where g is the longitudinal-transverse coupling of
2 36αG γi the deflection,
Eq.(14) contributes significantly less than Eq.(13) ⎧
⎨ Θacc /2 structure misalignment
because the trajectory offset is demagnified in the g = Θtra /2 trajectory misalignment (18)
defocusing quads, leading to smaller offsets on ⎩ < Θ tilted irises
average. ∼ iris
Effect of misalignments of the defocusing Beam-Gas Scattering Beam-gas scattering is
quads can be included by multiplying Eq.(14) by rarely an issue for the emittance in a linac,
α
yQD 2 φc 2 φc 160πZ(Z + 1)re2 β i γf
1+4 1 − sin tan2 Δ(γ)y ≈ ngas −1
yBPM 2 2 αG γi
where yQD is the misalignment of the defocusing (19)
quads. where Z is the atomic number of the gas and
Betatron Coupling In future linear colliders, typ- ngas is the density of the residual gas. At 20◦ C,
ically y x , and there are tight tolerances on ngas ≈ 3.22 × 1022 nmol P [m−3 ]. Here, nmol is
the allowable skew quad fields. Assuming purely the number of atoms per molecule of gas and P is
random errors [3], the vacuum pressure in torr.
E 2 The primary effect is to contribute to a halo.
K The number of particles scattered out to an ampli-
Δ(γ)y ≈ γx 4Ncell (15)
K tude greater than k times the rms beam size can
E is the skew quad gradient normalized by be estimated,
where K α
the magnetic rigidity Bρ and K is the normalized ΔN ngas β 4πZ 2 re2 γf
E = 2Θq K. Ncell ≈ 2 i −1 (20)
quad gradient. For roll errors, K N k γ αG γi
is the total number of FODO cells.
Similarly, assuming systematic errors, Emittance dilutions effects, multi-bunch
E Beam Break-up Multi-bunch BBU [15, 16, 17]
KQF + K E QD 2 sin2 Ncell Δφc leads to an amplification of the incoming tra-
Δ(γ)y ≈ γx 2
|K1 | sin2 Δφ c jectory jitter. One solution is reduction of the
2
(16) long-range transverse wakefield in the struc-
where subscripts QD and QF denote the skew ture design. Assuming the daisy chain model
components of the focusing and defocusing (Sec.2.4.1), the criterion for little or no blowup
quads, Δφc = φxc − φyc is the difference be- for the case α = 1/2 is [18]
( (
tween the x- and y-phase advances per cell. Note ( 2π0 NB re W⊥ (SB )β i ( γf α
that Δ(γ)y ∝ Ncell 2 when Δφ < 2/N
c ∼ cell . It is
( ( <1 (21)
( αG ( γi
thus useful to split φxc and φyc by a few degrees.
Also note that the dilution depends on the sum where W⊥ (SB ) is the wakefield at the following
KE QF + K E QD . If all elements have the same error, bunch.
323
Assuming the linac is operated with mini- equal to [20]

mized multi-bunch BBU and close to the au- Δy 2 = A × T × L
tophasing condition, one can estimate the emit-
tance dilution due to misaligned accelerator struc- The high-frequency jitter can be used to estimate
tures and a corrected trajectory similarly to the the motion of the beam centroid while the ATL-
single bunch case. One can use the single drifts primarily result in an increase in the beam
bunch emittance dilution formulas, simply replac- emittance, since the actual motion of the centroid
ing W⊥ (2σz ) with 4Wsum , where Wsum 2 ≡ can be compensated with beam-based feedback
2 systems, and thereby specify the time interval be-
nB −1 nB −1
1 1 tween component re-alignment.
W⊥ (iSB ) −
2
W⊥ (iSB ) Uncorrelated vibration of the linac quads yq
nB nB
i=0 i=0 will deflect the trajectory, causing a trajectory mo-

(22) tion, at the end of the linac, of
φc
nB is the number of bunches.
Δy 2 2 4γi Ncell tan 2 γf 1−α
Component jitter and stability There are three ≈ y + 1
q
σy2 γy Lcell i γi
issues: (i) jitter where the beam trajectory
(24)
changes from pulse-to-pulse, (ii) emittance degra-
Similarly, if uncorrected, the ATL motion will
dation due to shifts in the position of the accelera-
cause a trajectory displacement of
tor components, and (iii) the degradation of the di-
agnostic resolution that arises from the beam cen- Δy 2 Ncell γf sin2 φ4c 1 − α
troid jitter which can lead to poor convergence of ≈ 8AT (25)
σy2 γy sin φc 2 − α
tuning or correction procedures. We consider is-
sues (i) and (ii). Issue (iii) depends upon the diag- where A depends only the specifics of the site
nostics and the operational procedures. but typically A ranges from 0.1 to 100 nm2 /m/s.
Sources of vibration include natural seismic Of course, this trajectory drift could be corrected
motion and man-made cultural noise (Sec.5.13). with beam-based feedback systems at the end of
The ground motion and vibration can be divided the linac (or portion of the linac with multiple
into three regimes: high frequency where there is feedback systems) but it will still result in emit-
little or no spatial correlation of the vibration, the tance degradation. The emittance increase due to
intermediate regime where the lower frequency the trajectory oscillation is
φc
ground motion tends to be relatively well corre- sin2 1−α
lated, and slow drifts where the motion is uncor- Δ(γ) ≈ 8AT Ncell
3
γi δe2 4
cos2 φ2csin φc 2 − α
related.
(26)
In general, the effect of the ground vibration where δe is the rms deviation from the autophas-
is described with a 2-D power spectrum P (ω, k) ing energy spread. Typically, δe ∼ 20 ∼ 30% of
[19]. The motion of the beam at a point can be δauto since the energy variation along the bunch,
described by induced by the rf and longitudinal wakefield, usu-

1 ally does not exactly match the desired form un-
Δy(t)2 = P (ω, k)G(k)F (ω)dωdk less the bunch charge distribution is specially
(2π)2
(23) shaped.
where G(k) is the spatial response function for Distributions The preceding estimates have
the focusing lattice, and F (ω) is the temporal re- been for the expected emittance dilution due to a
sponse of the feedback systems. The growth of random distribution of errors. For a specific set of
the beam spot size can be evaluated in a similar errors, the dilution can deviate from the expected
manner with a different function G. value significantly.
Evaluating Eq.(23) requires detailed com- Assuming a Gaussian distribution of errors
puter calculation, but one can estimate the two which cause a single dominant cause of dilution,
limiting extremes: uncorrelated high-frequency the dilution will have the distribution [3],
jitter and slow uncorrelated drifts of components − Δ(γ)
which can be described with the “ATL” relation, f (Δ(γ)) = (1/μ)e μ (27)
which states that after a time T , the relative mo- Here, μ is the expected dilution. Strictly, Eq.(27)
tion of two points, separated by a distance L, is applies only when the distribution of errors is
324
Gaussian, but in a long linac with many errors, is of two, or more, trajectories while some param-
applies approximately regardless of the detailed eter is varied. For example, to align the quads
distribution for the errors. and minimize the residual dispersion, one can
Distribution (27) has a long tail. There is a 5% measure the trajectory while changing the quad
probability that any specific case will exceed 3μ. power supplies or the beam energy. A few of
These large dilutions arise when the errors add these steering techniques were tested at the SLC
in phase at the betatron frequency. Fortunately, [22, 23], including the “Dispersion-Free” [24] and
this situation is easy to measure and correct by the “Wake-Free” [25] algorithms. A version of
advanced correction techniques (see later). the Dispersion-Free steering technique was im-
If there are N forms of emittance dilution that plemented as part of the standard trajectory cor-
are all contributing to the total dilution with equal rection on the SLC. Similarily, to measure the
magnitude, the distribution function for the emit- short-range transverse wakefield due to acceler-
tance is described with a χ-squared distribution ator structure misalignments, one can vary the
with 2N degrees-of-freedom. For example, a to- bunch intensity or length [12, 26], and, to measure
the long-range transverse wakefield due to accel-
tal emittance dilution of 50% would have a χ-

squared distribution with 4-d.o.f. if it consisted erator structure misalignments, one can vary the
of 25% dilution due to transverse wakefields and bunch charge or the bunch spacing [27].
25% dilution due to betatron coupling. As N in- In all cases, the best alignment that can be at-
creases, the probability of having an emittance tained using local techniques is determined by the
much larger than the expected value decreases. It BPM or diagnostic precision and is roughly inde-
is difficult to calculate the distribution function in pendent of the magnitude of the misalignments;
the general case, but typically one obtains a close this arises because the techniques measure the dif-
approximation with N = 1 or 2. ferences between trajectories. To estimate the ef-
fectiveness of these techniques, there are analytic
Advanced trajectory and emittance correction approaches [28, 29], or one can perform com-
techniques A number of more advanced beam- puter simulations which include many of the envi-
based techniques have been developed or pro- sioned errors. This is important because the tech-
posed to reduce emittance dilution. The most niques frequently rely on measuring small differ-
straightforward is to use “-bumps”. Here, the ences between the beam trajectories and can be
measured beam emittance is minimized by adjust- sensitive to beam jitter or power supply fluctua-
ing some dilution source upstream of the mea- tions. Finally, another difficulty with these ap-
surement point. For example, one can reduce the proaches is that the beam emittance is not actu-
dilution due to the transverse wakefields by ap- ally measured and, instead, the dilution sources
propriately misaligning a few accelerator struc- are inferred from other measurements—thus, it is
tures. Such techniques are “non-local” because desirable to include emittance diagnostic stations
the sources of dilution are not reduced; instead, and use non-local correction to remove any resid-
additional dilutions are added to cancel the effect ual dilutions.
of the others.
The -bump technique is routinely used at the References
SLC to reduce Δ(γ) by ∼ an order of magnitude
[1] R.D. Ruth, Proc. 1986 CERN/US Part. Acc.
[21]. Some of the difficulties with the non-local
School (1986)
correction techniques are: (i) the beam emittance
[2] H. Henke, LINAC 88
and tails need to be measured accurately, (ii) mul-
[3] T.O. Raubenheimer, PhD Thesis, Stanford U.
tiple measurement stations may be needed to pre- (1991)
vent the dilutions from filamenting, and (iii) being [4] P. Wilson, AIP Proc. 87 (1982) p.450
nonlocal, it is sensitive to small changes in beta- [5] R.B. Palmer, Ann. Rev. Nucl. Part. Sci. 40 (1990)
tron phase advance that can arise from changes in 529
the quadrupole strengths or energy profile; this re- [6] K. Yokoya, private communication (1995)
sults in large fluctuations in the beam emittance at [7] V. Balakin, SLAC-355 (1989) p.55
the SLC over a time scale of hours. [8] V. Balakin et al, HEACC 83, 119
“Local” correction techniques measure and [9] G.A. Loew, J.M. Wang, PAC 85, 3228
correct the emittance dilution sources locally. [10] F.J. Decker et al, LINAC 94, 47
Typically, this is done by measuring the difference [11] K.L. Bane et al, PAC 93, 3933
325
[12] T.O. Raubenheimer, K. Kubo, NIM A370 (1996) cycle:

303 f0 W G
Q0 = 2π = , (1)
[13] T.O. Raubenheimer, PAC 93, 11 Pdiss Rsurf
[14] F.J. Decker et al, PAC 91, 6 f0 is the resonance frequency, Pdiss the dissipated
[15] R. Helm, G.A. Loew, Linear Accelerators, N. rf power. Q0 can also be expressed as the ratio
Holland (1970) of a geometrical factor G (with a dimension of
[16] R.L. Gluckstern, R.K. Cooper, P.J. Channell, PA resistance) to the surface resistance Rsurf .
16 (1985) 125
The TESLA type 9-cell cavities used at
[17] K. Yokoya, DESY 86-084 (1986)
FLASH are built with solid niobium and have
[18] K. Thompson, R. Ruth, PR D41 (1990) 964
a length of 1.038 m. They are operated at
[19] For example see: A. Sery, O. Napoly, PR E53
(1996) 5323
1.3 GHz in the fundamental mode TEM010 (π-
[20] V. Shiltsev, 4th Int. Workshop on Acc. Align. mode). Higher harmonics are sufficiently damped
(1995) using higher-order mode couplers. With a ge-
[21] J.T. Seeman, F.J. Decker, I. Hsu, Proc. 15th Int. ometrical factor of G = 270 Ω and a residual
Conf. High Energy Acc. (1992) 879 surface resistance of less than 10 nΩ, the qual-
[22] C. Adolphsen et al, PAC 89, 977 ity factor is well above 2 × 1010 . The design
[23] R. Assmann et al, KEK Proc. 95-12 (1995) R/Q is 518 Ω (using the definition of dissipated
power Pdiss = Vcav 2 /R with cavity voltage V
[24] T.O. Raubenheimer, R.D. Ruth, NIM A302 sh cav
(1991) 191 and shunt impedance Rsh ). For a comprehensive
[25] T.O. Raubenheimer, NIM A306 (1991) 61 overview of TESLA type superconducting cavi-
[26] C. Fischer, PAC 95, 731 ties see [3].
[27] R. Assmann et al, PAC 97, 503 The loaded quality factor QL is given by
[28] C. Adolphsen, T.O. Raubenheimer, PAC 93, 417 f0
[29] A. Sery, A. Mosnier, DAPNIA/SEA-96-06 QL = (2)
ΔfL
(1996)
with the (loaded) bandwidth ΔfL . The bandwidth
is defined as the full width at half height of the res-
4.3.2 Operation of Superconducting Linacs onance curve. At FLASH, the coupling of the rf
S. Schreiber, DESY to the cavity is adjusted such that QL = 3 × 106
with a loaded bandwidth of ΔfL = 430 Hz. For
The operation of Superconducting (SC) linear ac- superconducting cavities, the loaded QL is much
celerators (linacs) is illustrated by the example of smaller than the unloaded Q0 , so that the coupling
FLASH. β, defined as β = Q0 /QL − 1, is with ≈ 3000
Superconducting accelerator FLASH much larger than 1 – in contrast to normal con-
FLASH is the free-electron laser at DESY ducting cavities. Therefore, the properties of su-
in Germany in user operation since 2005. It perconducting cavities are mainly determined by
produces laser-like radiation in the VUV and the coupling or the loaded QL , while the cryo-
soft x-ray wavelength range with unprecedented genic load is determined by the unloaded Q0 .
brilliance.[1, 2] The radiation pulses in the Accelerating module Eight cavities are assem-
wavelength range of 4.1 to 47 nm are extremely bled in a 12.2 m long cryostat module cooled with
short, from 10 to 100 fs, with a peak power superfluid helium. Operation at 2 K is assured by
exceeding 1 GW. The FEL is driven by a high stabilization of the helium pressure at 31 mbar to
quality electron beam with an energy of up to the 3×10−4 level (rms). The cryostat design takes
1.25 GeV. FLASH uses TESLA superconducting care that the static heat load, important for effi-
acceleration technology. cient operation, is less than 1 W/m.
State-of-the-art superconducting cavities Superconducting technology allows acceler-
have accelerating gradients well above 25 MV/m. ation of many bunches grouped in bunch trains
At the same time, the (unloaded) quality factor with a large number of bunches per second.
Q0 is very high exceeding 1010 . A high quality FLASH is designed to accelerate a beam current
factor allows efficient operation with high duty of up to 9 mA – within a bunch train of 800 μs
cycle. in length, at 10 Hz repetition rate. For a single
The quality factor is proportional to the ratio bunch charge of 1 nC, this leads to 72000 acceler-
of stored energy W to the energy lost in one rf ated bunches per second.
326
After the upgrade in 2009/10, FLASH oper- Many noise signals distorting the amplitude
ates seven modules. Two modules with together and phase flatness are repetitive from rf pulse to
16 cavities are powered by one klystron. A 5 MW rf pulse. For instance drifts along the pulse due
klystron is sufficient to provide a drive rf power to thermal loads, beam loading effects, Lorentz-
of 250 kW per cavity required for 25 MV/m op- Force detuning, and others. A learning feed-
erational gradient. The first module has its own forward algorithm recognizes the repetitive dis-
5 MW klystron, the last two modules exceed- tortions and compensates them applying appro-
ing 25 MV/m are powered with a 10 MW multi- priate feedforward signals. The energy stability
beam klystron. The energy reach of FLASH is achieved at FLASH during operation with long
1.25 GeV. The electron source is a photoinjec- bunch trains is better than 10−4 (rms) from train
tor with a warm L-band copper rf gun together to train and along the bunch train. The phase sta-
bility is better than 0.02 ◦ .
with a suitable drive laser providing low emit-

tance beams. Stability in arrival time is important for free-
Important for successful acceleration of electron lasers, since pump and probe experi-
bunch trains with hundreds of bunches is the sta- ments rely on a measurable delay between the
bility and flatness of the accelerating gradient and pump and probe beam to the femtosecond level.
phase along the rf pulse, together with a reliable The goal is to achieve an arrival time stability
exception handling and machine protection sys- shorter than the radiation pulse length or, in other
tem to prevent catastrophic beam losses. words, at the 10 fs level.
Stability of acceleration There are several An already excellent energy stability of 10−4
sources of errors which affect the amplitude and of the beam entering the first bunch compressor at
phase of the accelerating field – such as errors in FLASH with an R56 of 0.18 m induces an arrival
cavity tuning, Lorentz force detuning, microphon- time jitter of 60 fs. The arrival time change δt of a
ics, errors and fluctuations of the drive amplitude bunch with a momentum deviation of δp from the
and phase, errors of relative phase of beam and nominal momentum p is given by δt = R56 /c ·
cavity rf for individual cavities, errors in injection δp/p. Other sources like the drive laser and rf gun
phase. Also the stability of the total beam charge add to the arrival time jitter. The measured arrival
affects the stability of the beam energy, phase, and time jitter at FLASH is usually 80 fs (rms) after
arrival time. the first bunch compressor.
Noise from the environment induces vibra- A reduction of the arrival time jitter to the
tions of the cavities (microphonics). The fre- 10 fs level is difficult to achieve with the low-level
quencies of microphonics are typically below rf feedback alone, and a beam based feedback is
1 kHz, which does not much affect the accelera- required. The beam based feedback system at
tion within a bunch train, but may lead to fluctua- FLASH measures the arrival time of the bunches
tions of the beam energy and phase from train to by comparing the signals obtained by a pick-
train. up antenna with an ultra-stable fiber laser. The
A feedback system stabilizing amplitude and amplitude of acceleration upstream of the bunch
phase within a pulse train and from train to train compressor is corrected for subsequent bunches
in the train. The resolution achieved is 6 fs [5].
is mandatory. The so called low-level rf sys-
So far, promising results have been achieved, and
tem analyses the signals from pick-ups installed
in each cavity to measure amplitude and phase of the arrival time has been stabilized to within 20–
the accelerating field. Since one klystron drives 40 fs.
many cavities (16 at FLASH), the system calcu- Beam loading The forward rf power Pfwd re-
lates the vector sum of amplitude and phase, and quired to accelerate a beam with current Ib at
provides feedforward and feedback for stabiliza- phase φ and detuning Δf is [4]
tion. The goal is to achieve stable acceleration of
2
all bunches along the pulse train and from train to V2 2RL Ib
train at the 10−4 level. Low latency and fast real Pfwd = cav 1+ cos φ
8RL Vcav
time computation abilities with high performance
ADC’s are required. FLASH uses modern FPGA 2
Δf 2RL Ib
based controllers with 14 bit ADC’s and DAC’s + 2QL + sin φ (3)
f0 Vcav
with 81 MHz clock rate.
327
with the loaded resistance RL = (R/Q)QL . For Pfwd = 233 kW, QL = 2.8 × 106 and get a gen-
the case of zero beam current, an accelerating gra- erator current of Ig = 18 mA, a filling time of
dient of 25 MV/m (Vcav = 26 MV) requires a for- τ = 681 μs, and an optimal time to inject the
ward power of 54 kW per cavity, a current of 9 mA beam of 472 μs after start of rf power. The single
requires 288 kW for QL = 3 × 106 , on-crest ac- bunch transient, the voltage drop for a bunch with
celeration and no detuning. a charge of 1 nC is 1.6 × 10−4 .
The coupling of the rf to the cavities is ad- In practice, FLASH is operated with many
justed to optimize the drive power for a given different bunch charges and bunch train currents,
beam current. The couplers at FLASH are usually ranging from 100 pC to 2 nC, and currents from
optimized for 9 mA operation. The optimum cou- 10 μA to 4.5 mA. The rf power needs to be ad-
pling is such that the sum of beam induced voltage justed according to the actual current. This is
and cavity voltage is matched to the generator for done by measuring the bunch train current with
minimum rf power and zero reflected power (see toroids (resolution 3 pC) and applying an appro-
Sec.2.4.3.2). A derivation of the optimum cou- priate feedforward rf power (beam loading com-
pensation).
pling is given in [4] and can be expressed by

Lorentz force detuning The Lorentz force be-
Vcav
Qopt
L = R
. (4) tween the rf magnetic field and the induced cur-
2Q Ib cos(φ) rents in a thin surface layer of the cavity walls
leads to a mechanical deformation. This induces
For a cavity voltage of Vcav = a detuning proportional to the square of the cavity
26 MV (25 MV/ m) and on-crest accelera- field Eacc
tion (φ = 0) the optimum coupling for a beam Δf = −KL Eacc 2
(9)
current of 9 mA is Qopt
L = 2.8 × 10 . Optimizing
6
the loaded quality factor reduces the forward with the detuning constant KL =
power needs from 288 kW to 233 kW. 1 Hz/( MV/ m)2 (for the mechanically stiffened
TESLA cavities).
Filling a cavity on resonance with constant
In pulsed operation, the detuning changes
forward power yields in an increase in cavity volt-
dynamically along the rf pulse. Measure-
age Vcav by [4]
ments give a linear detuning (rf flat-top) of

0.5 Hz/( MV/ m)2 / ms [6]. As a result, in
Vcav (t) = 2RL Ig 1 − e−t/τ (5)
pulsed operation with a field of 25 MV/m, the cav-
ity is detuned by 260 Hz within the pulse train du-
with a filling time of τ = 2QL /ω0 . The generator
ration of 800 μs. Since this is in the order of the
current Ig is given by
cavity bandwidth, the effect has to be accounted
for. Using Eq.(3) with Δf = 130 Hz, this would
Pg
Ig = 2 R
. (6) lead to an energy reduction along a 9 mA bunch
2Q QL train of 6.8%. To keep the energy stable, an ad-
ditional rf power of 7.7 % would be required to
The beam with a current Ib induces a voltage in compensate the detuning.
the cavity of For low gradients, pre-detuning together with
the low-level rf feedback is sufficient to regulate
Vb (t) = −2RL Ib 1 − e−(t−tinj )/τ (7) against Lorentz force detuning. However, for op-
eration with higher gradients above 25 MV/m, the
The optimum injection time to achieve immediate required additional power would be too high; a
steady state operation is tinj = τ ln 2, then when fast tuning system is required [7]. At FLASH,
the cavity voltage reaches half of its maximum. Piezo-stacks are mounted in parallel to the mo-
The cavity field at the optimum injection time is torized tuning system of the cavities. They allow
R a fast tuning within the rf pulse and compensate
Vcav (t = tinj ) = 2 Q L Ib . (8) efficiently the Lorentz force effect.
Q
Beam loss detection At FLASH the total power
To summarize, for optimized operation with of the design beam is 72 kW, the beam for ILC
a beam current of 9 mA and an accelerating gra- will exceed 10 MW. A permanent loss of beam
dient of 25 MV/m, we need a forward power of of the order of 0.1% into the cryostat would lead
328
to an unacceptable heat load. For many beam- 4.3.3 Halo in High Intensity Proton Linacs
line components like the SASE undulators, al- T.O. Raubenheimer, SLAC
lowed losses are one or two orders of magnitude
smaller. Only a few Gy per day are tolerated for In high intensity proton linacs, beam loss must
the FLASH undulators translating in a tolerable be carefully controlled to avoid activation of the
loss of approximately 5 × 10−5 (nominal beam, accelerator components. This requires control-
10 years undulator lifetime). ling the beam halos. Halos are most commonly
Beam losses are detected with two systems generated during bunching when rapidly chang-
[8]. Toroids along the machine measure the ing space charge forces can generate transverse
charge of the electron bunches with a single bunch tails and incomplete bunching can generate lon-
resolution of 3 pC. An alarm is triggered, when- gitudinal tails. In addition, halos can be popu-
lated when the space charge force is modulated
ever a single bunch loss, or a loss of the inte-

grated charge along a bunch train occurs. A sec- due to phase space mismatches which cause beam
ond system is based on beam loss detection using density oscillations [1, 2]; this modulation of the
scintillators placed at critical positions along the space charge force can resonantly drive particles
accelerator. Special care has been taken to pro- out to large transverse amplitudes. Relatively
tect the undulator system from beam losses larger simple models [3, 4, 5, 6] predict the qualitative
than 10−5 . Sudden losses trigger an alarm block- behavior and show good agreement with detailed
ing the drive laser of the electron source within simulation models.
3 μs. Small permanent losses lead to warning and
inhibit beam for a certain period. References
In case of sudden beam loss, the beam load- [1] M. Reiser, PAC 91, 2497
ing compensation of the low-level rf system is [2] A. Cucchetti et al, PAC 91, 251
triggered to turn off immediately (exception han- [3] J. O’Connell, PAC 93, 3657
dling). Otherwise the overhead on rf power would [4] J. Lagniel, NIM A345 (1994) 46
eventually quench the superconducting cavities. [5] R. Gluckstern, PRL 73 (1994) 1247
A proper exception handling limiting the for- [6] T. Wangler et al, LINAC 96, 372
ward power for individual cavities allows opera-
tion close to the quench limit. Since 16 cavities
are driven by one klystron and only the vector sum
of all cavities is regulated, some cavities may have 4.4 RECIRCULATED ENERGY
increasing, others, for compensation, decreasing RECOVERY LINACS
amplitudes along the rf pulse. Proper tuning to S.A. Bogacz, D.R. Douglas, G.A. Krafft,
resonance and adjustment of the loaded quality Jefferson Laboratory
factor is required to keep all amplitudes flat. At
FLASH, an operable gradient of about 5 to 10 % 4.4.1 Recirculation and Energy Recovery
below the actual quench limit has been achieved.
Linacs that are recirculated have several advan-
tages in achieving electron beam parameters out-
References side of the scope of the traditional ring acceler-
[1] FLASH web site http://flash.desy.de/ ators or linacs [1]. Beam quality emerging from
[2] W. Ackermann et al., Nature Photonics 1 (2007) a recirculated linac is not limited by synchrotron
336 radiation effects as in electron storage rings; with
[3] B. Aune et al, PRST-AB 3, 092001 (2000) the advent of energy recovery, currents can be
[4] T. Schilcher, PhD Thesis, U. Hamburg, TESLA- raised to within an order of magnitude of the best
1998-20 (1998) lepton storage rings in existence now.
[5] F. Löhl et al, PRL 104, 144801 (2010) The original impetus for recirculated super-
[6] M. Liepe, PhD thesis, U. Hamburg, DESY- conducting linac development came from the
THESIS-2001-045 (2001) electron scattering community of nuclear physi-
[7] M. Liepe, W.D. Moeller, S.N. Simrock, PAC 01, cists [2]. Coincidence measurements involving
1074 multiple-particle final states are most advanta-
[8] L. Froehlich, PhD thesis, U. Hamburg, DESY- geously run at limited peak currents to reduce
THESIS-2009-012 (2009) effects of accidental coincidences. To maintain
329
Sec.4.4: RECIRCULATED ENERGY RECOVERY LINACS
high event rates, integrated at low peak current, the beam load from the decelerating beam passes,
high average current is desired, implying c.w. op- there is no limit to the average current that may
eration. Early c.w. superconducting linacs were be accelerated due to rf source capacity. An al-
built to perform electron scattering experiments ternate method to provide energy recovery is to
at University of Illinois and Stanford University. recirculate the beam back through the accelerat-
These devices were upgraded in energy by the ing cavities in the opposite direction [12]. Be-
simple expedient of multipass beam recirculation cause electrons are relativistic and travel at the
where beam emerging from the end of the linac is velocity of light largely independently of beam
taken back to the beginning of the linac and run energy, coordinating beam acceleration and en-
at the same accelerating phase as the first pass [3, ergy recovery over a linac consisting of multiple
4]. More recently, a recirculated superconducting accelerating cavities is straightforward for elec-
linac was built at Darmstadt, initially for electron trons, but difficult for non-relativistic particles.
scattering experiments [5]. Multipass beam re- Because the beam transit time through the recir-
circulation has also been achieved at normal con- culated linac is much smaller than the radiation-
ducting linacs at Massachusetts Institute of Tech- induced emittance growth times in the bending
nology’s Bates Laboratory [6], the United States arcs, the beam longitudinal and transverse emit-
National Bureau of Standards [7], the Mainz cas- tances can be much smaller in energy recovered
caded racetrack microtrons [8] and at the Budker linacs than in storage ring accelerators that oper-
Institute [9]. Similarly, economic considerations ate at the same energy. It should also be noted that
drove the design of the 6 GeV CEBAF c.w. elec- energy recovery is an important element in the de-
tron accelerator at Jefferson Lab [10] to multipass sign of high average current electrostatic acceler-
recirculation, where a relatively expensive super- ators.
conducting linac could be replaced by relatively Beam energy recovery was first proposed as
inexpensive recirculation beam lines to reduce the a way to construct high luminosity colliders for
overall project cost. high energy physics [13]. Although never realized
Aside from economic considerations, appli- in this application, energy recovered accelerators
cations requiring the following elements should have been built as electron cooling drivers and
generally be well suited to deploying a recircu- high power free electron laser drivers [14, 15, 16].
lated and/or energy recovered linac: c.w. or other Many proposed applications utilize the advan-
high duty factor operation, high beam average tages of energy recovered linacs. For example,
current, low delivered beam energy spread, and Cornell University is investigating the energy re-
low delivered beam emittance. C.w. beam ac- covered linac as an undulator driver yielding su-
celeration with high accelerating gradient (5-10 perior, high average brilliance x-ray sources as an
MV/m) generally requires deploying a (recircu- upgrade to their conventional synchrotron light fa-
lated) linac consisting of superconducting accel- cility [17]. Similar programs exist at Argonne
erator structures. GeV-scale recirculated linacs at and Daresbury Laboratories [18], and in Japan
100 mA average current would ordinarily require [19, 20]. Brookhaven National Laboratory and
at least 100 MW of installed rf power merely to CERN are investigating the use of high average
accelerate the beam load. Beam energy recovery current energy recovery linacs as electron sources
allows the rf beam loading of the cavities to be for high-luminosity electron ion colliders [21].
substantially lowered. In applying this idea with
a back to front beam recirculation, the beam re- Applications of ERLs Applications of ERL in-
circulation path length is chosen to be a half inte- clude high average power FELs (Fig.1), syn-
gral number of rf wavelengths long. Because the chrotron light sources, Terahertz and Compton
beam sees an accelerating phase on the lower ac- sources as well as electron cooling devices and
celerating beam passes through the linac, after a electron-ion colliders for nuclear and high en-
phase shift of 180 degrees energy it is delivered ergy physics. Past and present ERLs are listed
back to the (SC) rf cavity on the higher deceler- in Tab.1. FEL ERLs promise high overall system
ating beam passes, energy is extracted and trans- efficiency, unprecedented average laser power
ferred directly to the accelerating beams without (∼100 kW) a nd reduced beam dump activa-
the need for power to be provided by the rf sources tion. Typical beam parameters for high power
[11]. To the extent that the average beam load FEL ERLs are bunch charge ∼0.1-1 nC, average
from the accelerating passes completely cancels current ∼100 mA, and energy ∼100-200 MeV.
330
Figure 1: JLab IR-UV Demo - Example of ERL-based FEL.
Table 1: Parameters of operating ERLs.

E [MeV] Iav [mA] Qb [pC] N [μm] Rep [MHz] Duty [%]

Chalk River [40] 25 30 10 50 3000 0.1
Bates [6] 400 10 3.5 10 2856 1
HEPL [41] 48 0.6 50 10 11.8 pulsed
CEBAF-FET [42] 45 0.3 0.2 5 1497 100

JLab IR Demo [43] 20-50 5 60 10 75 100
CEBAF-ER [29] 1050 0.09 0.2 1 500 100
JAERI [44] 17 40 500 30 80 pulsed
BINP [45] 22 30 2000 30 22.5 100
JLab IR Upgrade [39] 165 9 135 10 75 100
ALICE [46] 27.5 8.125 100 1.2 81.25 0.1
JLab UV Demo [47] 135 2.5 60 5 37.5 100
Synchrotron light ERLs promise to produce ra- coupling of beam to the environment via wakes
diation with much enhanced average brightness and impedances), and synchrotron radiation emit-
and peak flux compared to third generation SR ted during beam bending. ERLs are transport
sources, very short pulses (1 ps to 100 fs), and lines, not closed systems. The beam does not
high coherence. ERLs also are proposed as part reach equilibrium; beam quality is thus source-
of future electron-ion colliders for nuclear and/or limited and dominated by excitation processes
particle physics research [26]. The linac-ring op- (e.g. quantum excitation, wakes, noise etc.). Ra-
tion of Electron-Ion Colliders with target lumi- diation damping has limited palliative effect; ac-
nosity around 1033 cm−2 s−1 , such as LHeC [36], tive stabilization (feedback) requires very fast di-
would solely rely on the ERL concept to limit agnostics and controls. Beam quality and stability
the total wall-plug power to 100 MW, exploiting are therefore of paramount concern.
the unique features of ERLs. An ERL accelerator ERL dynamics are fully 6-dimensional and
combines characteristics of both storage rings and benefit from use of emittance exchanges amongst
linacs, and is potentially capable of accelerating projected subspaces [31]. Though the beam may
tens of milliamperes of average current to several reside in common regions of configuration space
tens of GeV. As a result of energy recovery, the on multiple passes, the energy differs from pass to
rf power required for acceleration becomes nearly pass so that (in contrast to storage rings) there is
independent of the beam current. no ”closed orbit”.
4.4.2 System Architecture and Beam As the system is not cyclic and is of fi-
Dynamics Issues nite length, betatron stability is not inherently
required; Twiss parameterizations of the beam
Energy recovery is used to mitigate cost in rf- (beam betatron functions) cannot, consequently,
and SRF-driven linacs; economics and perfor- be uniquely identified with Twiss parameteriza-
mance are dominated by fundamental physical tions of the lattice transfer map: the beam and
constraints and by management of power flow the lattice are different. There can in fact be
through the rf drive system (transient manage- benefits — for the purposes of aberration control
ment), the beam itself (losses from halo and — from deliberate mismatch of the beam to the
331
Sec.4.4: RECIRCULATED ENERGY RECOVERY LINACS
lattice [32]. The particular scheme used to man- High power, high brightness SRF systems
age the longitudinal phase space through beam ac- The use of SRF cavities in high-brightness sys-
celeration, delivery, use, and recovery cycle — or tems engenders numerous issues associated with
“longitudinal match” — is a key feature of any beam stability and beam quality preservation. The
ERL design, inasmuch as it dictates much of the most evident, historically, is the beam break-up
susceptibility of the system to collective effects (BBU) instability (Sec. 2.4.2) [25]; recent devel-
and defines nearly all features of the rf drive [33] opments in SRF cavity and accelerator transport
and other aspects of power flow through the sys- system design have rendered this effect largely
tem. In this regard, ERLs are essentially equiva- manageable [35] though means to do so (e.g. H/V
lent to time-of-flight spectrometers. phase space exchanges [30]) remain key design
The absence of equilibrium, closed orbit, and features.
uniquely defined/equivalent beam/lattice func- Efforts to preserve beam quality during ac-
tions elevates halo to the role of a key constraint celeration, delivery, and recovery must be ob-
on ERL performance. High beam power in such servant of numerous effects driven by interac-
tion of the beam with itself and its environ-
systems demands that halo-related losses be lim-

ited to a few parts per million of the beam cur- ment, some common to storage rings, others more
rent. However, the nonequilibrium beam behavior closely associated with linacs and transport sys-
renders conventional characterizations of beam tems. Space charge (both transverse (SC) and lon-
and aperture in terms of Gaussian distributions gitudinal (LSC) (Sec.2.4.4) has significant impact
and rms spot sizes (“σ”) are generally inappli- on beam quality early in the acceleration cycle
cable [34]. Instead, halo sources must be indi- and strongly influences the longitudinal match,
vidually catalogued, their evolution through the limiting the minimum tolerable injection energy
system evaluated, and mitigation methods (like and forcing injection of a long bunch with pro-
collimation, Sec.3.3.11) implemented if needed. cessing to shorter lengths at high energy. Longitu-
Numerous phenomena are of concern in the de- dinal matching is also strongly influenced by co-
sign and operation of ERLs. These may be cat- herent synchrotron radiation (CSR) (Sec.2.4.15),
egorized as 1) source performance, 2) issues re- which degrades beam quality during bunch length
lated to beam-quality preservation in high power, compression (Sec.4.5) and increases achievable
high brightness SRF systems, 3) effects in large- emittance. As ERLs often compress beams to
acceptance transport systems, and 4) problems re- high peak currents (e.g. for FEL driver opera-
lated to power-flow management. tion), wakefield (Sec.2.4.1) and resistive wall ef-
fects (Sec.2.4.9) are evident and of concern both
Sources Performance of ERLs is source lim- for their impact on beam quality and due to local-
ited. An essential element of an ERL is a high per- ized heating of beam line hardware. Beam and en-
formance source capable of production of highest vironmental impedances must be carefully char-
quality beam and then appropriate beam transport acterized and managed.
to assure emittance preservation in the low en- The combination of LSC, wakes, and
ergy regime, where the space charge forces have CSR can result in microbunching instabilities
paramount effect on emittance growth. There are (e.g. Secs.2.4.15,3.1.7,3.1.8). High energy ERLs
presently four major source technologies: SRF, with small geometric emittance must (as in high-
NC rf, dc and thermionic. The use of high ac- brightness electron strorage rings) contend quan-
celerating gradients is required to rapidly accel- tum excitation (incoherent synchrotron radia-
erate the electrons from the cathode to alleviate tion) (Secs.3.1,3.1.4.1), which can degrade the
beam quality degradation. Operating at high rep- beam quality. Preservation of small emittance
etition rate gives rise to significant thermal ef- also forces consideration of intrabeam scatter-
fects. Optimization study done at Cornell for a ing (Sec.2.4.12), Touschek effect (Sec.2.4.12),
DC gun where space-charge induced emittance and beam-gas scattering (Sec.4.3.1) as sources of
growth is most severe has demonstrated emittance emittance degradation and halo formation; the lat-
compensation for a wide range of bunch charges ter can potentially lead to intolerable levels of
(0.1 mm-mrad emittance for 0.1 nC charge, 0.7 beam scraping. Given the possible use of high
mm-mrad for 1 nC) [27]. Nuclear and/or particle- repetition rate with modest or low bunch charges,
physics applications may require a polarized elec- ion accumulation and trapping (Sec.2.4.13.1) are
tron source. of concern.
332
Accelerator transport An ERL transport sys- radiative processes, and interaction with targets,
tem must meet a number of implicit requirements: is a critical aspect of the overall system design
it must deliver a properly configured phase space and operation so as to avoid beam instability and
to the user (transverse and longitudinal matching); beam/power-loss induced damage to hardware.
it must support the decelerating energy-recovery Numerous power-management issues must
beam; these requirements must be met in a man- be addressed. High gradient SRF cavities un-
ner assuring beam stability and limiting beam loss dergo ponderomotive (or Lorentz force) detun-
to tolerable levels. This generally will require ing, which changes under varying beam condi-
manipulation of phase space to avoid beam loss tions - such as during beam off/on transients or
through the use of appropriate transverse and lon- when a user process (such as an FEL) extracts en-
gitudinal matching. ergy from the beam. The recovery phase - cou-
The use of the beam results in extraction of pled to the energy through the longitudinal match
energy (e.g. via an FEL interaction or emission of - then shifts, altering the degree of beam loading
synchrotron radiation) with attendant degradation and shifting cavity resonances. The rf drive must
of beam quality. As a result, ERLs often require therefore be properly configured to assure stabil-
compression of energy spread during energy re- ity [33]. The available beam power creates poten-
covery to limit losses. Additionally, conservation tial for the emission of high power electromag-
of energy demands that either the rf drive sys- netic radiation through the coupling of the beam
tem compensate for the energy draw by user and electromagnetic field to the accelerator environ-
collective processes, or that the beam be recov- ment. This of course can be intentional (e.g. ERL-
ered to a final energy below that at which it was driven FELs), but can lead to unanticipated power
injected, with outcome determined by choice of loss/deposition through a number of mechanisms.
longitudinal matching scheme. In the event of a Propagating HOM fields from SRF cavities can
severely user-degraded phase space, energy com- convert and transmit substantial power [25], as
pression requirements may demand that acceler- can the interaction of the beam with any envi-
ated and recovered beams be less than 180 de- ronmental impedance (Secs.3.2.1,2.4.1) – just as
grees out of phase with one another, leading to in storage rings. Resistive wall effects can, with
“incomplete energy recovery” [38], with the re- short bunches and high peak currents, cause sig-
covered final energy higher than that at injection. nificant heating [39].
The details of the longitudinal matching process Coherent synchrotron radiation (CSR)
define rf power requirements [33]. (Secs.2.4.15,3.1.2,3.1.7), can extract extremely
Given the potential for large amplitude beam large amounts of power from the beam and may
components (halo) and degradation of beam qual- result in significant heating of various system
ity during transport and use, ERL transport sys- components with deleterious effect. This has
tems must typically be of large acceptance and limited performance in existing oscillator-based
incorporate means for the management of aber- ERL-driven FELs [39]. In very high current
rations. This is consistent with the view of an ERLs operated at high energy, substantial power
ERL as a time-of-flight spectrometer system, and will be emitted as incoherent synchrotron radi-
emphasizes that magnetic field quality imposes ation, at power levels potentially sufficient to
significant constraints on ERL performance. In warrant engineered controls to avoid heating and
particular, field imperfections can lead to energy- hardware damage; the associated degradation
spread-inducing phase errors, which will aggra- of beam quality and loss of energy must be ac-
vate losses during beam recovery; tolerance to counted for in the longitudinal matching scheme
these errors scales inversely with energy, so that and design of the recovery process. Use of
the absolute field quality of an ERL must improve incomplete energy recovery [38] may be required
as the full energy increases [37]. to adequately manage the longitudinal phase
Power-flow management ERLs are intended space during deceleration; rf power transients
for use as wall-plug-efficient sources of extremely will be at issue as described above.
high power beams (“linac quality beam at storage As beam powers are very high, beam loss
ring electrical efficiency” [1]); as a consequence must be limited to a small fraction of total beam
the management of power flow via rf drive in- current (typically a few parts per million) so as
puts, beam-environment interactions, beam loss, to avoid radiation, activation of and/or damage
333
Sec.4.5: BUNCH COMPRESSION
to beamline components. Beam halo is therefore [31] M. Rihaoui et al, PAC 09 (2009)
of paramount concern and processes leading to [32] B. Blind, A. Jason, PAC 05 (2005) 225
the evolution of halo (field emission in high volt- [33] T. Powers, C. Tennant, ERL 07, 75
age/rf structures, incomplete extinction ratio in [34] S. Bernal et al, PRST-AB 14, 104202 (2011)
gated photocathode drive lasers, beam dynamical [35] C. Tennant et al, PRST-AB 8, 074403 (2005)
effects during beam formation and handling, in- [36] S.A. Bogacz et al, IPAC 11 (2011)
trabeam scattering (Sec.2.4.12), Touschek effect [37] D. Douglas, BIW 10, 506
(Sec.2.4.12), ion accumulation (Sec.2.4.13.1)) [38] S.V. Benson, D. Douglas, U.S. Patent Number
must be evaluated to analyze beam loss patterns, 7,166,973 (2007)
generate schemes for halo tuning, and/or and pro- [39] S.V. Benson et al, PAC 07, 79
[40] S.O Schriber et al, PAC 77, 1061
vide collimation systems (Sec.3.3.11) for con-
[41] T.I. Smith, NIM A259 (1987) 1
trolled management of beam spill.

[42] N. Sereno et al, PAC 93, 3246
[43] G. R. Neil et al, PRL 84 (2000) 662
References
[44] R. Hajima et al, FEL 06, 312
[1] L. Merminga, D.R. Douglas, G.A. Krafft, Annu. [45] O.A. Shevchenko, ERL 11
Rev.Nucl.Part.Sci. 53 (2003) 387 [46] Y. Saveliev, ERL 11
[2] R.E. Rand, Recirculating Electron Accelerators, [47] R. Legg et al, PAC 11 (2011)
Accelerators and Storage Rings, Vol. 3, Harwood
Academic, New York (1984)
[3] P. Axel et al, IEEE Tr.NS-24 (1977) 1133 4.5 BUNCH COMPRESSION
[4] C.M. Lyneis et al, IEEE Tr.NS-28 (1981) 3445 P. Emma, LBNL
[5] H.-D. Graef, A. Richter, LINAC 88, 231
[6] J. Flanz, C. Sargent, IEEE Trans NS-32 (1985) Linear compression Future 4th-generation
3213 light sources and linear colliders require very
[7] S. Penner et al, IEEE Trans.NS-32 (1985) 2669 short bunches of high-brightness electron beams.
[8] H. Herminghaus et al, IEEE Trans.NS-30 (1983) Linac-based free-electron lasers create a strong
3274 motivation to compress below the 100-fsec level,
[9] E.A. Antokhin et al, NIM A528 (2004) 15 which is 2-3 orders of magnitude shorter than in
[10] C.W. Leemann, D.R. Douglas, G.A. Krafft,
present storage rings. The physics and technical
Annu. Rev.Nucl.Part.Sci. 51 2001) 413
challenges of electron bunch compression are
[11] T.I. Smith et al, NIM A259 (1987) 1
described briefly here.
[12] J. Sekutowicz et al, PRST-AB 8, 010701 (2005)
[13] M. Tigner, Nuovo Cim. 37 (1965) 1228
RF Acceleration and Energy Chirp An elec-
[14] G.R. Neil et al, PRL 84 (2000) 662 tron bunch is compressed in length by first accel-
[15] R. Hajima et al, NIM A507 (2003) 115 erating in a linac at an rf phase, φrf , which is not
[16] E.J. Minehara, NIM A557 (2006) 16 at crest (i.e., φrf = 0), producing a nearly lin-
[17] S.M. Gruner et al, RSI 73 (2002) 1402 ear energy ‘chirp’ (energy correlation along the
bunch), h, defined as
[18] Y. Saveliev et al, IPAC 10 (2010) 2350
[19] K. Umemori et al, SRF-2009 (2009) 896 2πeV0 2π Ei
[20] I. Sakanaka et al, IPAC 10 (2010) 2338
h=− sin φrf = − 1− tan φrf .
λEf λ Ef
[21] V.N. Litvinenko, IPAC 10 (2010) 2364 (1)
[22] J. Flanz, S. Kowalski, C. Sargent, IEEE Here λ is the rf wavelength and V0 is the on-crest
Trans.NS-28 (1981) 2847 voltage gain of the linear accelerator, which, at
[23] J. Flanz, PAC 89 (1989) 1349 an off-crest phase of φrf , accelerates the electron
[24] V.I. Veksler, USSR Acad. Sci. 43 (1944) 346; J. charge, e, from an energy of Ei to Ef . Signs are
Phys. USSR 9 (1945) 153 chosen such that an rf phase of 0 < φrf < π sets
[25] G.A. Krafft, J.J. Bisognano, PAC 87, 1356 the head of the bunch (z > 0) at a lower energy
[26] I. Ben-Zvi et al, ERL 2005 Workshop
than the tail (h < 0). Longitudinal wakefields
[27] I.V. Bazarov, C.K. Sinclair, PRST-AB 8, 034202
of the rf structures typically ‘fight’ this rf-induced
(2005)
chirp, adding an (estimated) opposing chirp of
[28] P. Piot, D.R. Douglas, G.A. Krafft, PRST-AB 6, √
030702 (2003) 2NB e2 Z0 cs0 L − lb /s0
hw ≈ [1−(1+ lb /s 0 )e ],
[29] S.A. Bogacz et al, PAC 03 (2003) 195 πa2 Δz 2 Ef
[30] R. Rand, T. Smith, PA 11 (1980) 1 (2)
334
where lb is the FWHM bunch length (unifiorm effect on R56 ). The sign convention here has the
distribution), NB is the bunch population, Z0 is bunch head at z > 0, which means a high energy
the free-space impedance, c is the speed of light, particle passing through a chicane (R56 > 0) is
L is the rf structure length, a is its mean iris ra- advanced, arriving early (Δz > 0).
dius, and s0 is the characteristic wakefield length. For |θ| 1, the R56 approximates to
Magnetic Compression With an initial rms
2
bunch length of σzi , and an initial uncorrelated R56 ≈ 2θ ΔL + LB .
2
(7)
3
rms relative energy spread of σδi at energy Ei , the
final rms bunch length after a magnetic bend sys- This can be substituted into R56 of Eq.(3) to cal-
tem with path length dependence on particle en- culate the final bunch length after the chicane.
ergy, to first order, is Bunch compressors also produce non-linear
terms, such as: Δz = R56 δ + T566 δ2 + U5666 δ3 +
σzf = (1 + hR56 )2 σz2i + (Ei /Ef )2 R56 2 σ2 , . . .. In a chicane, their ratios to the linear term are
δi
(3) 3 4
T566 ≈ − R56 , U5666 ≈ + R56 , etc. (8)
where R56 is the transfer matrix element which 2 2
maps relative energy (momentum) deviations, A Simple Dog-Leg Bend A ‘dog-leg’ bend
δ ≡ Δp/p, into longitudinal position deviations system is composed of two dipole magnets of
within the bunch (Δz = R56 δ). Full compression opposite (or same) sign with quadrupole mag-
is achieved at R56 = −1/h, ‘under-compression’ nets between and the linear dispersion fully sup-
at 1 + hR56 > 0, and ‘over-compression’ at pressed after the 2nd bend. The total R56 over the
1 + hR56 < 0. A compression factor can be de- system for |θ| 1 is
fined as C0 ≡ σzi /σzf ≈ |1 + hR56 |−1 , with typ- R56 ≈ −θ 2 LB /3, (9)
ically C0 > 1. The final rms energy spread is also
where the sign has flipped with respect to a chi-
necessarily increased by the linac energy chirp to
cane. Unlike the chicane, the higher-order terms
σδf = h2 σz2i + (Ei /Ef )2 σδ2i . (4) are not so simply related, and the R56 is typically
quite small compared to a chicane.
Velocity Compression For a bunch which is A FODO Arc An arc composed of FODO
not ultra-relativistic, the value of R56 over a drift cells is a common compressor with compression
of length L at a constant energy γ = E/mc2 is matrix element
1 θT2 LT
R56 = L 2 . (5) R56 ≈ − , (10)
γ 4Nc2 sin2 (μ/2)
This can be substituted into R56 of Eq.(3) to cal- where θT is the net bend angle of the arc, LT is the
culate the compression over a drift at a constant arc length, Nc is the number of FODO cells, and μ
low energy. It is typically ignorable in high en- is the bend-plane betatron phase advance per cell.
ergy linacs, but can be used in the injector [1]. The sign is opposite to a chicane and T566 /R56 is
Types of magnetic compressors The compres- typically > 1.
sion description above assumes a particular R56 Nonlinear compression limitations The ef-
value generated by a bending system. fects of an inevitable non-linear energy chirp due
A Magnetic Chicane The simplest, and there- to the sinusoidal rf are examined here, along with
fore most common bunch compressor is the chi- the second-order compression term, T566 .
cane, which consists of just 4-dipole magnets and Second-Order Limits The second-order chirp,
is transversely achromatic to all orders (rectangu- h2 , due to sinusoidal rf, is defined as
lar magnets), generating an R56 value of 2π 2 eV0
h2 ≡ − 2 cos φrf . (11)
R56 = 4LB [sin(θ)/θ − cos(θ)] + 2ΔLθ sin(θ), λ Ef
(6) With the further definition r ≡ T566 /R56 , the fi-
where ±θ is the bend angle of each dipole mag- nal rms bunch length (squared), to second order,
net, ΔL is the drift distance between the 1st and is
2nd, as well as the 3rd and 4th magnets, and LB E2 2 2
is the magnetic length of each dipole, each mea- σz2f = i2 R56 σδi + (12)
Ef
sured along the beam path with E → ∞. (The
distance between the 2nd and 3rd magnets has no (1 + hR56 )2 σz2i + R56
2
(h2 + rh2 )2 ζ ,
335
Sec.4.5: BUNCH COMPRESSION
where ζ ≡ zi4 − σz4i , the 4th moment of the dis- several sextupoles and care in balancing the geo-
tribution. A Gaussian (uniform) distribution pro- metric and chromatic aberrations. It is not easily
duces ζ = 2σz4i (ζ = 4σz4i /5). The term involv- achieved in a chicane.
ing h2 and r introduces a 2nd-order compression Synchrotron radiation effects An electron
limit, as can wakefields (not included here). If beam passing through a bend magnet generates
h2 = r = 0, then Eq.(12) reduces to Eq.(3). synchrotron radiation, which may degrade the
Second-Order Correction by RF Phasing The beam brightness.
second-order aberration will vanish if h2 + rh2 = Incoherent Synchrotron Radiation (ISR) An
0. Equation (11) shows that h2 < 0 for an ac- electron beam of energy E passing through a bend
celerating rf phase (i.e., −π/2 < φrf < π/2 and magnet of length LB , bend angle θ, and bend ra-
eV0 > 0). Since h2 > 0, h2 + rh2 cannot be dius ρ = LB / sin θ, generates synchrotron radia-
zero for an accelerating system if r < 0. Equa- tion, which adds rms relative energy spread to the
tion (8) shows that a chicane, as well as any other beam of [4]
compressor with only dipole magnets, produces 55 re c E 5 LB
σδISR = √ (15)
r < 0. Therefore, the second-order compression

24 3 (mc2 )6 |ρ3 |
aberration cannot be canceled with simple rf phas-
1
ing for an accelerating system ending in a chicane. ≈ (4.13 × 10−11 m2 · GeV−5 )E 5 |θ 3 | .
A decelerating phase can be used, but the energy LB
extracted may make this an unattractive solution. This energy spread is stochastic in character and
There are, however, other compressors, such as a adds in quadrature to the initial energy spread. A
simple FODO-cell arc which have r > 0, where chicane with four dipole magnets, each of length
it is possible to remove the second-order aberra- LB , will add twice this level, 2σδISR . The en-
tion simply by choosing the right rf accelerating ergy spread is generated within the chicane and
phase, given by induces an additive bend-plane emittance growth
of [5]
Ei2 + 8e2 V02 r(1 + 2r) − Ei β̂ + β̌
cos φrf = > 0. Δγ ≈ (8×10−8 ·GeV−6 )·E 6 (ΔL+LB + ),
2eV0 (1 + 2r) 3
(16)
(13)
where β̂ and β̌ are the maximum and minimum of
Second-Order Correction by Harmonic RF
the bend-plane beta function within the chicane.
Harmonic rf can also be used to compensate 2nd-
Coherent Synchrotron Radiation (CSR) As
order limits [2, 3]. The higher frequency section
the electron bunch is compressed, it may begin
is typically set at the decelerating ‘crest’ phase
to radiate coherently as well, generating further
(φn = π). The harmonic rf voltage needed for
bend-plane emittance growth [6]. A very rough
2nd-order compensation of bunch compression is
estimate for a Gaussian bunch, including only
2
Ef 1 − 12 r π2λR2 (1 − σzf /σzi )2 − E0 steady-state longitudinal forces over just the 4th
eVn ≈ 56
, bend of a chicane (where the bunch is shortest in
(λ/λn )2 − 1 ‘under-compression’), is
B
(14)
C # $2/3
where E0 is the initial energy before rf accelera- C (0.22)2 re2 NB2 |θ 5 |LB
D
≈ 1+ ψ
tion at wavelength λ (e.g., E0 ≈ 0), and the har- 0 16 γN β σz4f
monic voltage, Vn > 0, is decelerating for r < 0
and is reduced by the square of the harmonic num- ψ ≡ L2B (1 + α2 ) + 4β 2 + 4αβLB , (17)
ber (n ≡ λ/λn ). It is therefore an advantage where N is the initial normalized bend-plane
to use the highest frequency harmonic available emittance, NB is the bunch population, and β, α
(provided that 2πσzi /λ 1). It is also possible are the nominal bend-plane Twiss parameters at
to use a harmonic rf phase which is not at the de- the end of the 4th bend. The effect can be mini-
celerating crest in order to add some linear chirp mized by keeping α ≈ 0 and using weak bends,
or even correct third-order correlations. This pos- but this formula is too rough to be used to opti-
sibility is not addressed here. mize α and β. Equation (17) is typically an under-
Second-Order Correction by Sextupoles Sex- estimate, since it ignores radiation from the other
tupole magnets can also be used to compen- chicane bends. More information is available in
sate 2nd-order limits, but this typically requires Secs.2.4.15,3.1.7.
336
Compressor tolerances Bunch compression [4] M. Sands, SLAC-121 (1970)

systems may also introduce many tight tolerances, [5] T.O. Raubenheimer, P. Emma, S. Kheifets, PAC
with two such issues highlighted here. 93, 635
Compressor Stability A small rf phase error [6] E.L. Saldin, E. A. Schneidmiller, M.V. Yurkov,
can produce a large compression error, demand- NIM A483 (2002) 516
ing tight rf phase stability tolerances. The relative [7] K.L. Bane et al, PAC 93, 596
compression jitter is related to the rf phase jitter [8] M. Dohlus, T. Limberg, FEL 05, Stanford
[7]
Δσzf ΔC 4.6 OPERATION OF
=− = (C0 ∓ 1)Δφrf cot φrf
σz f C0 LINEAR-COLLIDER
FINAL FOCUS SYSTEMS
Δφrf
≈ C0 , (18) A.A. Seryi, JAI, G.R. White, SLAC
φrf
with under-compression expressed by the minus The Final Focus System (FFS) is part of the Beam
sign and over-compression by the plus sign. The Delivery System (BDS) which is responsible for
approximation at right is for a high compression transporting the e+ /e− beams from the exit of
factor (C0 1) and the rf phase not too far off the high energy linacs, focusing them to the sizes
crest (|φrf | 1) [8]. required to meet the luminosity goals, bringing
Compressor Dipole Field Quality An energy them into collision, and transporting the spent
chirped beam can become very large transversely beams to the main beam dumps. In addition, the
at the center of a chicane (or similar compressor) BDS must measure the linac beam and match it
where the bend-plane dispersion is largest, requir- into the FFS; protect the beamline and detector
ing demanding dipole field quality tolerances. against mis-steered beams from the main linacs;
The tolerance for quadrupole field compo- remove any large amplitude particles (beam-halo)
nent, b1 , and sextupole field component, b2 , (both from the linac to minimize background in the de-
measured at radius r0 ) in the 2nd and 3rd chicane tectors; measure and monitor the key physics pa-
dipole magnets, each with bend angle θ, disper- rameters such as energy and polarization before
sion D, bend-plane beta function β, and initial and after the collisions. The BDS must provide
geometric bend-plane emittance 0 , expressed as sufficient instrumentation, diagnostics and feed-
a fraction of the dipole field, b0 , is back systems to achieve these goals.
( ( The main subsystems of the BDS needed to
( b1 ( 1 r 2Δ/0
( (< 0
, (19) fulfill the listed goals are the diagnostics region,
( b0 ( |θ| β 1 + ξ2 fast extraction and tuneup beamline, betatron and
( ( energy collimation, FFS proper, interaction region
( b2 ( 2 Δ/0
( ( < 1 r0 , (20) (IR) and extraction line. The main task of a FF is
( b0 ( |θ|
β 3 0 1 + ξ 2 to focus the beams to the small sizes required at
where ξ 2 ≡ D 2 σδ2 /(β0 ) and Δ/0 is the tolera- the interaction point (IP). To achieve this, the FFS
ble emittance growth (e.g., 2%). forms a large and almost parallel beam at the en-
These field errors generate 1st and 2nd or- trance to the Final Doublet (FD), which contains
der dispersion, which can be corrected with small two or more strong quadrupole lenses. The high
quadrupole and/or sextupole magnets added to the energy linear colliders require σx σy at the IP
compressor chicane, if the tight field quality tol- in order to maximize the luminosity while han-
erances cannot be met. dling the beam-beam interaction. Typical designs
have a demagnification of a few hundred in y with
an order-of-magnitude less in x. The y-focusing
References is usually set close to the limit from the depth of
[1] L. Serafini, M. Ferrario, AIP CP 581 (2001) 87 focus, i.e. βy∗ ∼ σz , which can be extended by a
[2] P. Emma, SLAC-TN-05-004, factor of several by travelling focusing, but other
http://www.slac.stanford.edu/pubs/slactns/slac-tn- constraints, e.g. chromaticity or tolerances may
05-004.html limit the achievable depth of focus.
[3] F. Zimmermann, T.O. Raubenheimer, AIP Most next generation linear colliders have
Conf.Proc.367 (1996) 84 beam collisions with a crossing angle Θc to avoid
337
Sec.4.6: OPERATION OF LINEAR-COLLIDER FINAL FOCUS SYSTEMS
parasitic bunch collisions and to provide a method

of directing the spent beams to dumps. The
crossing-angle will reduce the luminosity unless
the beams are “crabbed”, i.e. tilted in x − z plane
so that they pass through each other head-on.
Significant advances in the optical design and
chromatic correction of the linear-collider FFS
have been made since described in earlier editions
of this handbook [1].
Figure 2: Optics of ILC beam delivery system [7].
4.6.1 Chromaticity Compensation
compensation of chromaticity is achieved by in-

Even for a beam with a minor energy spread of
terleaving a pair of sextupole magnets with the
a fraction of a percent, the focused beam size
quadrupoles of the final doublet. The dispersion
will be diluted by the chromaticity of these strong
throughout the FD is created by upstream bends,

lenses. The design of a FF is therefore driven pri-
and is designed to be zero at the IP. Geomet-
marily by the necessity of compensating the chro-
ric aberrations, generated by FD sextupoles are
maticity of the FD.
cancelled by two or more sextupoles located up-
There are two primary approaches for chro-
stream. Sextupoles placed in FD generate second
maticity compensation — the non-local scheme,
order dispersion, which, however,can be compen-
implemented particularly at FFTB [2] and B fac-
sated simultaneously with x and y chromaticity
tories [3, 4] and the local compensation scheme
provided that half of the total horizontal chro-
[5] presently being implemented at ATF2 [6],
maticity of the whole final focus is generated up-
which is used as a basis for the FFS design of pro-
stream. The optics of the ILC Beam Delivery Sys-
posed linear colliders.
tem [7] was designed according to this described
In the non-local FFS the chromaticity is com- approach aiming for up to 1 TeV CM. This is
pensated in dedicated chromatic correction sec- shown in Fig.2.
tions by sextupole magnets placed at maxima
Luminosity The luminosity is
of dispersion and beta-functions. The geometric
aberrations generated by the sextupoles are can- L = L0 HD (1)
celed by using them in pairs with a minus identity
where HD is the luminosity enhancement
transformation between them. The non-local FFS
(Sec.2.5.3), L0 is the geometric luminosity which
is built from separated optics blocks with strictly
depends on the beam properties as well as the
defined functions, and is relatively simple for de-
FFS properties such as the depth of focus and the
sign and analysis. The major challenge of the
bunch crossing angle θc (θc = 0 with crab cross-
non-local FFS is in its applicability to high en-
ing; θc = Θc without crab crossing),
ergy colliders — the FFS bend magnets have to
be sufficiently long and weak to minimize the ad- frep nb Nb2
ditional energy spread generated, lengthening the L0 = η(Cθ , Ay ) (2)
4πσx σy
system to the range of several kilometers which
makes it impractical to use it for TeV scale e+ /e− where frep , nb , and Nb are the repetition rate,
colliders. the number of bunches per pulse, and the num-
are the rms beam
ber of particles per bunch, σx,y
Fig.1 shows conceptual optics layout of FFS
with local chromaticity compensation. Local sizes at the IP and function η describes the ef-
fects due to depth of focus and the crossing an-
gle. Assuming flat Gaussian beams (σx σy ),
η(Cθ , Ay ) is given by the quantity RL of Eq.(6),
Sec.4.1. The geometric luminosity loss is 14%
when Ay = σz /βy∗ = 1 and another 10% when
Cθ = θc σz /σx = 1. Typically, FFSs are designed
with Ay < ∼ 1 and Cθ < ∼ 1. Crab crossing and
Figure 1: Schematic FFS with local chromatic correc- “travelling focus” [8] have been proposed to re-
tion. duce the sensitivity to the crossing-angle and to
338
the depth of focus, respectively, and are being im- The tolerance on the bunch timing is much looser
plemented in the FFS design. and just shifts the longitudinal IP position with a
Crossing-angle and crab-crossing We address luminosity loss that depends on the depth of fo-
four issues with respect to a crossing angle: cus.
(i) Geometry If Θc is large, it may be possi- (iv) Solenoidal field effects With a crossing
ble to have separate quads for the incoming and angle, the beams do not travel parallel to the
outgoing beams while for small Θc an exit port solenoidal field. This results in y deflections of
is usually needed in the FD quads. In all cases, the beams and vertical dispersion at the IP. As-
Θc needs to be sufficiently large to separate the suming the FD quads are outside of the solenoidal
beams and to allow them to escape the IR. field, the offset is
(ii) Multibunch crossing instability [9] An in- Bs L2 Θc
Δy = (7)
stability can arise from the parasitic bunch in- Bρ 4

teractions. Assuming flat beams, the criterion where Bs is the solenoidal field. This deflection
to avoid significant vertical displacement of the can be corrected by steering but it is more difficult
bunches due to this instability is to correct the associated dispersion. There are ad-
ditional effects arising in the practical case when
CΘ2
1 Dy
(mB − 1) + (3) solenoidal field is partly overlapping with the FD,
Dx Dy 2 3 which is discussed below.
where Dx,y are the beam-beam disruption pa- Aberrations In the FFS with local compensa-
rameters (Sec.2.5.3), CΘ = Θc σz /σx , mB is tion of chromaticity, the second order aberrations
the number of bunches that interact parasitically. are canceled when the x and y pairs of sextupoles
Assuming no additional masking, mB = 1 + are separated by transfer matrices MF,D which
2L /SB with SB the bunch spacing and L the have block-diagonal structure
⎛ ⎞
free distance to the IP. This instability decreases F 0 0
rapidly with larger Θc . ⎜ F 1/F 0 0 ⎟
M = ⎝ 12
0 0 F 0 ⎠
, (8)
(iii) Crab crossing When CΘ > ∼ 1, the beam
can be crabbed [10]. Here, the bunches are given 0 0 F34 1/F
a z-dependent offset so that they collide head-on: where all non-zero elements are arbitrary. In order
dx /dz = Θc /2. There are two approaches: ei- to cancel the second order chromatic aberrations
ther use an rf deflecting cavity or use the corre- the sextupole integrated strength has to satisfy [5]:
lated energy spread along the bunch and a small KSF 2 = −F 3 KSF 1 ; KSD2 = −D 3 KSD1
amount of residual dispersion Dx at the IP [11]. ξx1 + ξx2 ξy
The required Dx is KSF 1 = 3 ; KSD1 = 3 ,
RF 12 D RD34 DIP
σδ
Dx c = Θc /2 (4) where ξx1 is the horizontal chromaticity of the
σz system upstream of the bend, ξx2 is the chromatic-
where σδc is the correlated component of the rms ity downstream, ξy is the vertical chromaticity,
relative energy spread, while the required rf volt- DIP is the angular dispersion at the IP and the
age is transfer matrices are as shown in the figure above.
Θc Eλrf The final-focus chromaticity ξ is defined as inte-
eVrf = (5)
4πR12 gral over s of the product of quadrupole strength
where E is the beam energy and R12 is the trans- K1 and beta function, without any factor of 4π.
port matrix element (Sec.2.2.1) from the crab cav- The condition ξx1 = ξx2 means that half of the
ity to the IP. horizontal chromaticity is generated upstream, al-
Tolerances on the alignment and stability of lowing simultaneous compensation of the x and y
the crab cavity tend to be loose except for the rel- chromaticity together with the second order dis-
ative phase tolerance between the crab cavities on persion. The flexibility to modify the coefficients
either side of the IP. This phase difference Δφrf in Eq.(9) via a proper choice of optics allows
can cause an x offset between the beams. To limit compensation of third and fourth order aberra-
the luminosity dilution to 2%, the bunch separations. Further improvements can be achieved with
tion must be < 0.3σx rms, which in turn requires an additional bend upstream of the second sex-
0.6 σx tupole pair, a further sextupole and an additional
Δφrf < 2π [radians] (6) decapole.
Θc λrf
339
Synchrotron radiation The FFS with local zero at the IP. The beam will reach the IP with
compensation of chromaticity is more amenable the vertical angle δy ∼ B0 L∗ θc /(Bρ)/8, anti-
for its scaling into the multi TeV region of ener- symmetrically with the opposite beam (for e+/e-
gies. For a wide range of parameters, the IP beam case) resulting in zero overall crossing angle in
size dilution for this FFS is dominated by the en- vertical plane.
ergy loss due to synchrotron radiation in the bends The more realistic case when the detector
(this additional rms energy spread can be obtained solenoid field overlaps with the FD means that the
using Eq.(17), Sec.3.1.4), which scales as [5]: symmetry is broken, resulting in anomalous cou-
pling and dispersion created at the IP, as well as
Δσy2 γ5 3
∝ 2 D B ∝ (9) non-zero vertical offset of the beam at the IP. The
σy L effects can be calculated exactly [13], and can be
# $3/2
2 estimated as
3/2 ∗3 DIP x 3/2 γ 7/2
(γy ) L . Bl θc ∗ σxp0 BL ∗
x y L5 yIP ≈ L ; < yx >≈ L ;
Bρ 2 σy0 Bρ
where DB is the angular dispersion produced

σE Bl ∗
by the bend magnets which scales as DB ∼ < yE >≈ D L . (10)
L∗ /L and the bend length is assumed to be
σy0 Bρ x0
DIP
proportional to the total length of the system L. These effects need to be compensated and this
The last two terms in parentheses are constant if can be done most efficiently with use of weak
DIP is proportional to the horizontal beam diver- antisolenoids. Such compensation removes the
gence and if the ratio between horizontal and ver- anomalous effects in the entire energy range of
tical emittance is constant with energy. We also the collider.
assume constant IP β-functions and x and y chro- The detector solenoid and resulting vertical
maticity, which is equivalent to holding L∗ and curved orbit produce synchrotron radiation gener-
the length of the final quadrupoles fixed. This ated energy spread and corresponding growth of
implies that the scaling is valid only in the range the beam size at IP

where the gradients required can be provided by dz
increased pole-tip field or decreased aperture. If (δσySR )2 = CE γ 5 R362
(z) , (11)
|ρ(s)|3
the normalized emittance stays constant, which is
a worst case assumption, then the length of the where CE = 55re λe /(31/2 24). The beam size
FFS system scales with energy as L ∼ γ 7/10 . growth is proportional to (B0 Lθc )5/2 and does not
If however the achievable normalized emittance depend on the beam energy. However, the nomi-
decreases inversely with energy as is assumed in nal beam size typically decreases with energy, and
[12], then the length of the FFS scales as L γ 2/5 . therefore this SR beam size growth must be kept
In this case, it is possible to design a FFS for a 3 small enough not to limit performance at higher
TeV center of mass energy collider in a length of energies.
about 600 m. The Detector Integrated Dipole (DID) is a coil
Another effect due to synchrotron radiation integrated into the detector solenoid, producing
(Oide effect) is discussed in Sec.3.3.5. sine-like transverse field through the detector, al-
Effects of the detector solenoid The field of lowing compensation of the angle at the IP and
the detector solenoid produces beam coupling minimization of the vertical orbit variation (thus
and, when including a crossing angle, it also gen- also the reduction of the synchrotron radiation ef-
erates additional dispersion and vertical trajectory fects) [14]. A negative polarity of DID is also
through the IP region. The latter will also result in considered in order to reduce the effect of back-
additional IP beam size growth from synchrotron ground due to beam-beam pairs — in this case
radiation. the anti-DID field is optimized in such a way that
In the ideal case when the solenoid field does most of low energy pairs are guided into the exit
not overlap with the FD quadrupoles, the symme- aperture of the extraction line [15].
try is preserved, and correspondingly various cou- Wakefields Wakefields are dangerous in the
pling terms are exactly zeroed. The vertical or- FFS because the β-functions are large in places
bit deviation reaches δy B0 L∗2 θc /(Bρ)/(32) at and due to the necessity to install collimators and
the maximum deviation, however will be exactly protection masks close to the beam.
340
Wakefields from the transverse geometric and Thermal photons The inverse-Compton scat-
resistive wall can be important through two ef- tering with thermal photons is discussed in
fects: (a) increase of the beam emittance, and (b) Sec.3.3.4. To calculate the number of large am-
amplification of the centroid jitter. To fully in- plitude scatterings, the differential cross-section
vestigate the impact of component and geometry needs to be integrated numerically. However, an
choices in the FFS, calculated wakes are typically estimate can be obtained over a distance L:
inserted into beam tracking codes. [16], [17] and ΔN
[18] describe analyses investigating: geometric = 0.5 nLσc (15)
N
and resistive wall wakefields of the tapered and where ΔE/E = xave /(1 + xave ) is the en-
flat parts of collimators; resistive wall wakes of ergy deviation of the particle due to scattering,
the beam pipe, which are especially important in and σc is given by substituting x by xave =
the regions of the final quadrupoles (where the β 10.8 EkB T /(m2e c4 ) in Eqs.(5-6), Sec.3.3.4. For
functions are very large); trapped modes in crab typical parameters, Eq.(15) tends to overestimate
cavities (which can be used to tilt the bunches at the number of large amplitude scatterings by a
the IP to compensate for a crossing angle).
factor of 2-5.
Effect (b) often imposes the more restrictive Elastic Coulomb scattering off nucleus The
constraint. The centroid jitter amplification due energy change of the incident particle is rela-
to the resistive wall wakefield is tively small and the primary effect is an angular

Δy Nb re LβF D δs Δy deflection that may cause the particle to exceed
= 1.6 (12)
σy γσz b3 σy F D the beam-pipe aperture. The differential cross-
section for Coulomb scattering on atomic nuclei
with L the length of the aperture, b the beam- is found in Eq.(2), Sec.3.3.1. The neglect of the
pipe radius, βF D the β-function in the FD, δs =
nuclear form factor and the recoil of nucleus
σz /Z0 σc skin depth. causes a slight overestimate of the particle loss.
The effect due to the discontinuity of a gentle Assuming the aperture is limited at a single
taper is [19] location by bap , such as the final doublet, the num-
ber of particles scattered to an amplitude greater
Δy Nb re LΘ2 βF D Δy
= 0.56 (13) than the aperture over a distance L is
σy γσz b1 b2 σy F D
ΔN 2πZ 2 r 2 2
where L and Θ are the length and angle of the = ngas L 2 2 e ds R12 + R34 2
(16)
taper, b1 and b2 are the initial and final beam-pipe N γ bap
radii. Both these tolerances can be minimized by where R12 , R34 are from the scattering position
enlarging b. to the aperture, and we assume b2ap (R12 2 +
Scattering 2 2
R34 )θmin .
Bremsstrahlung The cross-section for Elastic Coulomb scattering off atomic e− [21]
bremsstrahlung at high energies (assuming The angular deflection can be accounted for by
complete screening) is [20] replacing Z 2 with Z(Z + 1) in Eq.(16). However,
the recoil of the electron results in a significant
16 183
σBr ≈ αZ(Z + 1.35)re2 ln energy change to the incident particle. The
3 Z 1/3
differential cross-section for a relative energy
δmax change of δ is
× ln + δmin − δmax (14)
δmin dσee 2πZre2 1
= (17)
where δmin and δmax are the minimum and max- dδ γ δ2
imum photon energy in units of the beam en- and the number of particles scattered beyond a
ergy. The factor Z(Z + 1.35) accounts for the limiting energy aperture δmin is
nuclear charge and approximates the atomic elec-
trons. Typical cross-section for CO with large ΔN 2πZre2 1
= ngas L (18)
energy losses ∼ a few barns. The number of N γ δmin
scattered beam particles is ΔN/N = ngas LσBr , Tolerances The “bare” tolerances one can de-
with L the distance the beam travels, ngas [m−3 ]= fine for elements of the FFS are extremely restric-
3.2 × 1022 P [Torr]Natom at 300◦ K, Natom = the tive, there is no expectation that real-world align-
number of atoms per gas molecule. ment and mechanical fabrication processes can be
341
controlled well enough to suffice. One designs where Θb is the bending angle of the bend, βy is
tuning knobs to specifically remove the expected evaluated at the magnet, and k1 is the integrated
aberrations produced. For all classes of elements quadrupole strength K1 . For most magnets in the
in the FFS, we can assign two different types of FFS, | sin(ψy − ψy )| ≈ 1.
tolerance. The tightest sensitivity is usually that on the
Capture tolerances define the tolerances to final doublet. When it moves as a unit, the trajec-
which the elements must be tuned before global tory motion at the IP is roughly equal to the mo-
tuning knobs are applied, in order to ensure that tion of the magnets. The sensitivity is even tighter
the knobs will reduce the IP spot size / obtain when the F and D magnets move in opposite di-
required luminosity within a finite number of it- rections. Fortunately, ground motion tends to be
erations. These tolerances strongly depend upon correlated over long distances, and the sensitivi-
the specifics of the hardware and algorithmic im- ties are greatly reduced [22]. Also, for a linear
plementation of the global tuning knobs, and are collider operating in multi-bunch mode, there will
thus hard to specify in a general sense. For exam- always be the requirement to have a fast feedback
ple, tuning knobs for ATF2 require coordinated
system operating at the IP which also greatly re-

movements of the FFS sextupole magnets to in- laxes this sensitivity [23].
troduce known aberrations at the IP. The range of Dispersion Dispersion generated by deflec-
motion of the magnets is limited by the physical tions (in a FFS with locally corrected chromatic-
range of motion of the cam-based magnet mover ity) is mostly due to the chromatic dependence of
system. This provides a capture range for spe- the deflection rather than any chromatic depen-
cific IP aberrations that the mover-based tuning dence of the displaced downstream trajectory,
system can deal with, which in turn provides cap- % &
ture tolerances for FFS elements which produce yq k1
Δσy = σδ |R34 | (20)
these aberrations. Also, imperfections in the mo- θb Θb
tions of the magnets, field errors in the magnets
and non-linearities of the knobs themselves mean Vertical dispersion can also be generated by rolls
that the knobs are not perfectly orthogonal over of the quads θq or displacements of the sextupoles
their entire range. This provides for a soft capture ys ,
% &
tolerance that is hard to uniquely define. 2θq Dx k1
Stability tolerances specify how far a given Δσy = σδ |R34 | (21)
ys Dx k2
element may drift from its post-tuneup state be-
fore enlarging the IP beam size by a given amount. where Dx is the horizontal dispersion evaluated at
These are easier to define in generic terms. the magnet.
We consider here effects that contribute to the Skew coupling Skew coupling is generated by
vertical IP spot size. direct skew fields or by trajectory offsets in sex-
Incoming beam If the geometric aberrations tupoles. The direct skew fields can arise from rolls
in the FFS are well corrected, the nonlinear aber- of the quads or displacements of the sextupoles,
rations will not be very sensitive to incoming beta- and from any detector solenoid.
tron oscillations or changes in the incoming beam % &
emittance. Provided the design optical transfor- 2θq k1
Δσy = σx |R34 | (22)
mations between the FFS sextupoles exist, these ys k2
will cause most aberrations to cancel. However,
the Oide effect can impose constraints on the in- where σx is the x beam size at the magnet. The
coming trajectory jitter. tightest sensitivities are usually on the FD mag-
Trajectory errors The dominant source of y nets. The FFS must be designed with tuning ele-
trajectory errors are movements of the quads yq ments to correct for the skew coupling as part of
and rolls of the bends θb . The position at the IP is the design optics including the detector solenoid.
% & With flat beams, the skew correction only needs to
Δy yq K1 be applied to the two phases that enlarge the ver-
= dsR (19)
σy
34
θb /ρ tical beam size at the IP, i.e. the x -y and x -y
% & terms. As most of the FFS elements are at a phase
=
yq k1 βy
| sin(ψy − ψy )| x 90◦ from the IP, the dominant correlation term to
θb Θb y be corrected is usually the < x y > term.
342
Magnet field quality Unwanted higher-order procedures, but these should be considered guides
multipoles in the magnets of the FFS can be a more than a set of detailed steps, and require a
problem due to the large beta functions in parts of certain amount of intuition in their application.
the system. This was found to be especially true at For example, the recipe for optimizing the ILC
ATF2 for example, where the relatively low beam FFS design [26] uses a semi-automatic design tool
energy (1.3 GeV) means a much larger beam size based on Matlab together with MAD [27] for ba-
to magnet aperture diameter ratio compared to the sic matching and a combination of TRANSPORT
high-energy LC designs that it is testing. Normal and tracking codes such as TURTLE [28] to de-
and skew multipole tolerances were set based on termine third and fourth order matrices and calcu-
IP beam size growth estimates from particle track- late beam spot sizes at the IP. As one tries to push
ing codes such as described in the section on op- the demagnification of the system to its ultimate,
timization and tuning [24]. Skew multipole sensi- and aim for very small spot sizes ≤ 1 nm, for ex-
tivities are the most restrictive for vertical beam ample as required by the CLIC [29] design, then
size at the IP, especially skew sextupole. It is one needs to take into account even higher order
possible to ameliorate the tightness of this sen- aberrations. The recipe in [30] deals with aber-
sitivity given precise measurements of the mul- rations up to ninth-order. Here a code based on
tipole strengths and angles. This is done through PYTHON, MAPCLASS [31] uses MADX [32]
non-linear retuning techniques using the FFS sex- together with the polymorphic tracking code PTC
tupole strengths and matching quads and/or in- [33] to compute transfer maps up to arbitrary or-
creasing βx∗ . The latter reduces the horizontal ders across the FFS beamline. Tracking codes
beam size in the magnet apertures, which causes such as TURTLE, PLACET [34], DIMAD [35],
less coupling into the vertical plane at the IP. Lucretia [36] etc are required to determine the
Magnet field strength Initial errors in the field beamsize, GUINEA-PIG [37] is used to deter-
strength of the FFS magnets, drifts of field mine the luminosity for the non-Gaussian beam
strength over time due to temperature variation profiles at the IP. The tracking code is also essen-
or power supply setting errors. These cause di- tial for the higher demagnification systems to in-
rect growth of the vertical beam size at the IP clude the effects of synchrotron radiation. These
due to introduced mismatches from the design recipes require much back-and-forth between the
optics, especially in the sextupoles which per- various codes to determine the aberrations, beam
form fine cancellations of chromatic, geometric sizes and to create and use complex knobs to per-
and chromo-geometric terms. They also cause er- form the optimizations. This is why to help au-
rors in the tuning knobs that rely on the manip- tomate the process an intermediary scripting lan-
ulation of those magnets to generate orthogonal guage such as PYTHON or Matlab are used, both
sets of aberrations at the IP. Typically, the mag- to handle the optimization process and to transfer
net fields need to be accurate at the 1e-4 to 1e- the lattice and changes thereof between the vari-
3 level (dB/B). Some magnets are more sensitive ous codes.
than others; the strong FD magnets, magnets at Online tuning of the FFS [38] As discussed in
locations with high beta functions and the sex- the tolerance section, the tolerances of the FFS are
tupoles are especially sensitive. usually sufficiently tight that there is no hope that
Optimization and tuning The design of the magnet manufacturing and alignment procedures
compact FFS optics, and any optimization of that can be good enough by themselves. It is essential
design, for example to change the focal length, to design online tuning procedures that use live,
or demagnification of the system, or energy band- beam-based information to perform the final tun-
width etc, requires many free parameters to be ing of the system to achieve the design beam pa-
tweaked in order to control the cancellation of rameters at the IP.
various important aberrations. Specific auto- We assume here an incoming matched, dis-
matic design and analysis tools have been avail- persion and coupling corrected beam using the
able in the past to handle generic FFS’s, such upstream BDS optics, such that the remaining
as FFADA [25]. The optimization of the com- aberrations are within the capture tolerances of
pact FFS however is much more tricky and only the FFS tuning knobs. The tuning of the FFS
semi-automated tools exist for this task. Pub- starts with the traditional tasks of beam-based
lished “recipes” exist which describe the design alignment (BBA), aligning the electrical readouts
343
of the BPMs with the magnet field centers of realistic errors becomes more difficult the more
the magnets they are attached to [39], and using aggressively one tries to focus the beam at the IP;
global steering algorithms [40] to put the beam ie the difficulty scales with the chromaticity of the
in a straight trajectory through the system. Us- system. For systems with very small IP spot sizes
ing beam tracking code, we model the FFS op- such as required by CLIC, one needs a more non-
tics, including expected error sources and magni- linear approach to tuning making additional use
tudes and the expected performance of the BBA of octupole and decapole magnets to correct for
process. This modeling is done in a Monte Carlo the more dominant higher-order aberrations dur-
fashion, each seed of the Monte Carlo simulation ing the latter stages of tuning.
contains a different spread of possible errors using
a distribution based upon expectations from the Operational aspects In order to operationally
test the design theories behind the compact FFS
various hardware systems. Where possible, mea-

surements and direct experience from beam oper- optics, the test accelerator ATF2 [41] was con-
ation are included. The outcome of this modeling structed. ATF2 has a scaled version of the ILC
is to derive a list of the most likely aberrations at FFS optics designed to accept an extracted 1.3
the IP. We then design a set of tuning knobs to GeV electron beam from the ATF damping ring
remove these. The knobs are designed to each or- with 12 pm vertical emittance and focus it down to
thogonally target one of the expected aberrations. 35 nm vertically at the IP. Configured in this way,
To test how robust these knobs are likely to be the ATF2 FFS optics has a level of chromatic-
in the face of a beamline with realistic errors, the ity equal to that of the ILC design and provides
tuning procedure is performed in simulation using a proof of concept test of its operation. There
the Monte Carlo model developed for this. The is also the possibility to further squeeze the beta
basis for the tuning knobs is the deliberate offset functions at the IP to produce a vertical beam size
of the chromaticity correction sextupole magnets. down to about 22 nm that would test our abil-
Horizontal and vertical translations, together with ity to tune a more highly chromatic system more
roll about the beam axis DOF of these sextupoles relevant to a machine such as CLIC. At the time
generates different sets of first and second-order of writing, the accelerator is still being commis-
aberrations at the IP which are calculated to form sioned; a summary of progress so far is available
a response matrix. This is then inverted to provide [42].
a set of approximately orthonormal knobs to cor- In addition to testing the optics itself, ATF2
rect the desired list of aberrations. The simulation provides a test bed for the various high-precision
is then used to estimate a sequence of knobs to use diagnostics that are required for a future Lin-
to remove all important IP aberrations. The most ear Collider to operate. These are such devices
dominating terms are usually the vertical disper- as high resolution cavity BPMs (<100 nm res-
sion, waist shift and < x y > coupling terms. olution), laserwire systems for non-destructive
Horizontal and vertical motion of the sextupoles beam size measurements, fast feedback systems
generates these terms as follows (considering the for multi-bunch orbit control. It is not enough
more sensitive vertical dimension here). just to deliver a small beam to the IP to achieve
Horizontal offset of a sextupole produces a high luminosity, the beam at the IP must remain in
feed-down normal quad field, which can be used collision with its colliding partner beam at a level
to generate a waist shift at the IP by, small compared to the beam size itself. This typ-
Δα∗y ≈ −K2 LΔxβx∗ βy∗ cos 2Δμsy (23) ically means a tolerance of the IP beam position
where K2 is the sextupole strength, s denotes the at the sub-nm level. Such stability is only possi-
location of the sextupole and ∗ the IP. ble through a set of low and high bandwidth orbit
feedback systems. A secondary aim of the ATF2
Vertical motion of the sextupoles generates
a feed-down skew-quad field component that is project is to demonstrate the required stability re-
used to couple the x-y motion of the beam par- quired for the maintainable, high luminosity oper-
ticles, and in areas of design horizontal dispersion ation of the FFS by demonstrating both a contin-
additionally produce vertical uous small beam spot size at the IP and holding it
dispersion. stable at the few-nm level.
ΔDy∗ = −K2 LΔyDx2 βxs βys sin Δμsy (24) Simulations of the ATF2 system with ex-
It is noted from simulation studies that the tun- pected levels of mechanical vibration and ground
ing of the FFS beamline in the presence of motion show an expected growth rate of the beam
344
size at the IP of about 0.5nm per hour. In addition beam disruption. The last three items usually re-
to holding the orbit steady with beam feedbacks, quire separating the particle beam from the pho-
this means we also have the requirement to pe- ton beam.
riodically re-tune the beam using our developed The dump must be able to withstand the aver-
tuning knobs. It is envisioned that we will apply age beam power as well as instantaneous energy
these using a knob dither-feedback system. deposition from a single bunch or bunch train.
Machine Detector Interface The MDI design The heating due to the very high energy density in
aimed to provide the necessary luminosity while a single pulse generates a shock wave that can ex-
also providing satisfactory background environ- ceed the material stress limit and thus determines
ment, and beam instrumentation at the IP includ- the minimum beam size at the dump. The main
ing the fast luminosity monitor, as well as beam approach considered is to use a liquid such as wa-
position monitors required for the intra-train feed- ter for the primary absorbing material similar to
back. The ILC luminosity can be measured with a the beam dump developed for the high current 25
precision 0.1% or better by measuring the Bhabha GeV beam at SLAC [43, 44].
rate in the polar-angle region from 30-90 mrad,

and at smaller polar angles of 5-30 mrad the rate References
or energy deposition of beamstrahlung e+e- pairs
can be measured for a fast luminosity diagnostic. [1] T.O. Raubenheimer, F. Zimmermann, Operation
The spatial distributions of pairs in this region can of Final-Focus Systems in Linear Colliders, in
be used to determine beam collision parameters Handbook of Accelerator Physics and Engineer-
such as transverse sizes and bunch lengths. The ing, 1st edition (1999); RMP 72 (2000) 95
MDI design must be compatible with stringent re- [2] V. Balakin et al, PRL 74 (1995) 2479
quirements on stability of the elements, in particu- [3] M. S. Zisman, LBL-34556 (1993)
[4] E. Kikutani et al, KEK Preprint 2001-157
lar the final doublets. For the particular case when
[5] P. Raimondi, A. Seryi, PRL 86 (2001) 3779
the linear collider will have a single interaction re-
[6] P. Bambade et al (ATF Collaboration), PRST-AB
gion but will be equipped with two experimental
13, 042801 (2010)
detectors in the so called push-pull configuration, [7] ILC Reference Design Report, v.3, SLAC-R-857
the MDI region needs to be designed to be com- (2007)
patible with rapid disconnection and exchange of [8] V.E. Balakin, 77th ICFA Workshop on Beam Dy-
the detectors. namics, May 1991, Los Angeles
Extraction and beam dump The last element [9] P. Chen, K. Yokoya, SLAC-PUB-4653 (1988)
of the FFS is the primary beam dump. In the [10] R. Palmer, Snowmass DPF Summer Study 1988
next generation of linear colliders [7, 12], typi- (1988) p.613
cal average beam powers are many MW. After the [11] R. Brinkmann, 5th Int. Work. Next Generation
beam-beam collision, a significant fraction of this Linear Colliders, SLAC-436 (1993)
power is in the form of photons. It is very im- [12] “A 3 TeV e+e- Linear Collider Based on CLIC
portant that the dump line transmits most parti- Technology,” CERN 2000-008
cles to the dump. This also reduces background [13] Y. Nosochkov, A. Seryi, PRST-AB 8, 021001
(2005)
sources in the detector. Complete transmission
[14] B. Parker, A. Seryi, PRST-AB 8, 041001 (2005)
is difficult to attain because of the enlarged beam
[15] A. Seryi, T. Maruyama, B. Parker, Nanobeam
emittance and energy tail due to beam-beam dis-
2005, SLAC-PUB-11662 (2006)
ruption. Thus one requirement of the extraction [16] G. Rumolo, A. Latina, D. Schulte, EPAC 06
line is to have large apertures and a large energy [17] K.L.F. Bane, A. Seryi, PAC 07
bandwidth while capturing the disrupted beam. [18] G. Burt et al, EPAC 06
Another consideration is that the dump line [19] K. Yokoya, CERN SL/90-88 (1988)
must include the diagnostics necessary to moni- [20] H. Bethe, J. Ashkin, Experimental Nuclear
tor and tune the beam collisions. The primary di- Physics, Vol. 1, Wiley (1953)
agnostics needed are: (i) the beam position and [21] I. Reichel et al, ICAP 98, eConf C980914 (1998)
angle after IP which are needed to interpret the 222
beam-beam deflection scans to determine the IP [22] F. Poirier et al, EPAC’08; W. Bialowons et al,
beam sizes, (ii) the beam energy and energy spec- EPAC06; A. Seryi, L. Hendrickson, G. White,
trum, (iii) the beam polarization, and (iv) the Nanobeam 2005, SLAC-PUB-11661
345
Sec.4.7: OPERATION OF CIRCULAR ACCELERATORS
[23] J. Resta-López et al, JINST 5 P09007 (2010) deflection source is

[24] B.I. Grishanov et al, KEK-REPORT-2005-9 θi
(2005) Δx(si ) = β(si ) cot(πν) (3)
2
[25] B. Dunham, O. Napoly, DAPNIA/SEA 94-06
+ θi
[26] A. Seryi, P. Raimondi, SLAC-PUB-9895, (2003) Δx (si ) = [1 − α(si ) cot(πν)] (4)
[27] H. Grote, F. C. Iselin, CERN SL-90-13-AP- 2
REV.2 (1991) where Δx (s+ i ) is the slope of the orbit immedi-
[28] D.C. Carey, FERMILAB-NAL-064 (1971) ately after the kick. The maximum orbit excur-
[29] R. Tomas, PRST-AB 13, 014801 (2010), and ref- sions
erences therein β(si )β(s)
[30] R. Tomas, PRST-AB 9, 081001 (2006) Δx̂(s) = θi (5)
2 sin(πν)
[31] R. Tomas, CERN-AB-Note-2006-017 (2006)
[32] H. Grote, F. Schmidt, CERN-AB-2003-024 ocur at Δφ = ±π(ν −m), m = 1, 2, 3, ... < ν.

[33] E. Forest, F. Schmidt, E. McIntosh, KEK Report The probable displacement of the closed or-
2002-3 bit at location s from a random distribution of N
[34] D. Schulte, CERN PS-2000-028-AE (2000), deflections is

https://savannah.cern.ch/projects/placet/ N β(si )
[35] R. Servranckx et al, SLAC-R-285 (1990) Δx(s)rms = β(s) √ θi rms (6)
[36] P. Tenenbaum, PAC 05, 4197 2 2 sin(πν)
[37] D. Schulte, TESLA-97-08 (1996) If the deflections are caused by displaced
[38] Y. Nosochkov et al, SLAC-PUB-9255 (2005); G. quadrupoles,
White et al, EPAC 08
[39] J. Smith et al, PAC 05 θi = (Kl)i δxi (7)
[40] P. Tenenbaum et al, PAC 01
where 1/(Kl)i is the focal length [K =
[41] B.I. Grishanov et al, SLAC-R-771 (2005)
(c/E0 )(∂By /∂x)] and δxi is the quad displace-
[42] P. Bambade et al., PRST-AB 13, 04280 (2010)
ment. K > 0 for a defocusing quad.
[43] D.R. Walz, IEEE Tr.Nucl.Sci. 3 NS-14 (1967)
923 Quadrupole errors The perturbation to β-
[44] Zeroth-Order-Design Report for the Next Linear function from N quadrupole errors at positions si
Collider, LBNL-PUB-5424, SLAC-474 (1996) (i = 1, ...N ) is
Δβ(s) 1 N
= (ΔKl)i β(si )
β(s) 2 sin(2πν)
4.7 OPERATION OF CIRCULAR i=1
ACCELERATORS × cos(2|φ(s) − φ(si )| − 2πν) (8)

The betatron tune change Δν = ν − ν0 is
4.7.1 Error Sources and Effects [1, 2, 3]
D. Rice, Cornell U. cos(2πν) − cos(2πν0 ) (9)
s+C
1
Dipole errors Transverse orbit displacement at = sin(2πν0 ) β(s )ΔK(s )ds
2 s
position s due to deflections θi is
For a single gradient error and small Δν,
β(s)
N
Δx(s) = θi β(si ) β(si )
2 sin(πν) Δν = − (ΔKl)i (10)
i=1 4π
× cos(|φ(s) − φ(si )| − πν) (1) Vertical dispersion errors [4]
s+C
Resulting change in orbit circumference is βy (s)
Dy (s) = F (s ) βy (s )
2 sin(πνy ) s

N
ΔL = θi D(si ) (2) × cos[φ(s ) − φ(s) − πν]ds (11)
i=1 F (s) = (K + SDx )yc − Ksq Dx − Gy
with D(s) the dispersion function. where S, Ksq , Gy are sextupole (m−2 ), skew quad
The displacement and slope of the closed or- (m−1 ) and dipole (m−1 ) strengths, and yc is the
bit due to a single deflection, at the location of the closed orbit displacement.
346
The vertical emittance from vertical disper- References

sion is
2J Dy2 2 [1] E.D. Courant, H.S. Snyder, Ann. Phys. 3 (1958) 1
y (disp) = σ (12) [2] C. Bovet et al, CERN/MPS-SI/Int. DL/70/4
Jy βy δ (1970)
with J,y the damping partition numbers, σδ the [3] H. Wiedemann, Particle Accelerator Physics,
rms beam relative energy spread. Vol.1, Springer-Verlag (1993)
There are five terms in the expression for [4] T.O. Raubenheimer, R.D. Ruth, PAC 89, 1435
Dy2 /βy [4]. The first three of which arise from [5] R.E. Shafer, Fermilab report TM-991 (1980)
(1) quad rotations θ, (2) sextupole misalignments [6] D.A. Edwards, M.J. Syphers, An Introduction to
ym , and (3) vertical dipole kicks, respectively: the Physics of High Energy Accelerators, Wiley
1 (1993) p.113
(Kl)2 θ 2 βy Dx2
(1) 2 [7] A. Chao, SSC-N-434 (1987)

2 sin πνy quads
1
(2) 2 (Sl)2 ym
2
βy Dx2 4.7.2 Orbit and Lattice Function
8 sin πνy sexts

Measurements
yc2 D. Rice, Cornell U.
(3) with corrected closed orbit
βy Beam position measurement Beam position is
The remaining terms include the closed orbit cor- usually measured by coupling to the EM field of
relation function [4] and decrease with the num- the beam. Insulated plates or buttons, striplines,
ber of correctors. and rf cavities are the most commonly used cou-
Eddy current multipoles Eddy currents in- plers (Sec.7.4.6). The difference between two but-
duced in a changing magnetic field generate mul- ton pickups is normalized to the sum as [1]:
tipoles which sometimes need to be considered.
For example, in a rippling dipole field B(t) =
B sin ωt, a round beam pipe shields the field and
the rippling field seen in the beam pipe by the
beam becomes [5]
B
B(t) = √ sin[ωt − tan−1 (ωτ )] (13)
1 + ω2τ 2
with τ [s] = μ0 σc bt/2 = 2π × 10−13
b[mm]t[mm]/ρr [Ωm], where b and t are the ra- R−L
x
b 4 sin φ2
dius and thickness of the beam pipe, and σc its = (1)
R+L φ
conductivity. Similarly, in a rippling quad K(t) = with variables defined in the figure.
K sin ωt, the quad field seen by the beam is
K ωτ
Alternatively, the phase and amplitude of a
K(t) = sin ωt − tan−1 transverse mode in an rf cavity are processed to
2 2
1 + ω τ /4 2 yield a position. A comparison of advantages
(14) and resolutions among these couplers is given
In case the beam pipe is not round, a chang- in Tab.1. The resolution quoted is without av-
ing multipole can also generate higher multipole eraging or numeric processing and is primarily
fields. The sextupole field generated in a changing determined by intrinsic bandwidth and coupling
dipole, e.g., is an operational consideration during strength. Electronics noise figure, processing al-
ramping (particularly in fast cycling machines). If gorithms, and mechanical stability of the pickup
the beam pipe is a wide chamber with gap height will ultimately determine usable resolution.
g and thickness t, the sextupole field is [6] Processing electronics vary according to
∂ 2 By t needs (bunch-by-bunch, multi-turn vs. high-
= 2μ0 σc Ḃ (15) resolution averaging) and pickup type. Direct
∂x2 g
If the beam pipe is slightly elliptical with = sampling (better time resolution) and rf down
[(width)-(height)/(height)] (|| 1), then [7] conversion (better position resolution) methods
∂ 2 By t are shown in Fig.1.
= μ0 σc Ḃ (16) Cavity BPM electronics are conceptually sim-
∂x2 4b
ple amplifier/down conversion circuits. However,
347
Table 1: Beam Position Monitor pickups compared (electrode spacing is for reference cited).
Pickup Type Resolution (μm) Electrode Spacing Time Response (ns) Cost / monitor
Button [2] 10 50 mm ≈1 modest
Stripline [3] .15 35 mm ≈ 2L/c medium
RF Cavity [4, 5] 0.5 – ≈ 50 higher
must be exercised when Qβ −Qs separation is not

large [9].)
Beam centering in quadrupoles Modulating a
quadrupole by a small amount ΔK when the

beam is displaced Δx from the magnetic center
ĂͿ
will cause a deflection ΔθQ
ΔθQ = −ΔxΔKl
where l is the magnet length. Δx may then be cal-

culated from the measured change in the closed
orbit. In practice, because of the change in optics,
closed orbit bumps are successively applied un-
til the orbit no longer changes as the quadrupole
current is changed. This method is difficult to
implement when the BPM pickups are located
ďͿ any significant distance from the quadrupoles.
The change in beam closed-orbit coordinates at
Figure 1: a) Direct sampling method [2], b) Frequency
a quadrupole (thin lens approximation) resulting
down conversion [6].
from a change in the quadrupole gradient is [11]:
βΔKcot(πυ)
phase and amplitude stability throughout is criti- Δxco|exit = −xbq
2 + βΔKcot(πυ)
cal to realize full resolution.
Appropriate signal processing methods such ΔK [1 − αcot(πυ)]
as averaging or Model Independent Analysis Δpco|exit = −xbq
2 + βΔKcot(πυ)
(MIA) using SVD (Sec.4.8) can give one or two where xbq is the initial displacement of the closed
orders of magnitude improvement in usable beam orbit from the quad center, β, α are the usual
position measurement resolution as well as iden- twiss parameters at the quadrupole exit, ΔK is the
tify faulty detectors [6, 7]. Resolutions below 30 change in quadrupole strength, and ν is the beta-
nm have been achieved using cavity BPM pickups tron tune. A transfer map may be constructed re-
and MIA/SVD. lating measured changes in closed orbit to the (un-
Dispersion Measurement The dispersion func- known) xbq ’s and (known) ΔK’s, from which the
tion D(s) can be measured at each BPM by sub- xbq ’s may be found. Adding betatron phase ad-
tracting 2 orbits taken at different rf frequencies. vance measurements to the orbit difference mea-
Knowing the phase slip factor η, surements improves knowledge of the optics beta
ΔE functions, enhancing the resolution of this method
Δx (s) = D (s) (2)
E0 [12].
D(s) ΔC D(s) Δfrf Beam centering in sextupoles A sextupole of
= =− . strength S is modulated by ΔS. The changes in
η C η frf
betatron tunes are proportional to the horizontal
A variation of this method is to drive the beam displacement of the beam,
longitudinally at its synchrotron frequency and to βx,y (s)
process as in a coupling measurement (Sec.4.7.8), Δνx,y = ∓ ΔxΔSl
using the horizontal signal in a dispersive region 4π
and with vertical displacement the minimum split
as a reference [8, 9, 10]. In addition to being between the transverse normal modes is given by
faster, this method is useful when aperture limi- 1
tations restrict the swing in rf frequency. (Caution δνmin = Δy (ΔSl) βx βy
2π
348
Several accelerators have skew-quad windings on where θ is an overall arbitrary constant, p(i) is an
sextupoles, permitting the magnetic center of the integer chosen so that 2πp(i) − θrev (i) gives the
skew quad (presumed close to the sextupole cen- correct phase shift, adjusting for time of flight and
ter) to be determined by techniques similar to cable delays within a factor of π, and m(i) is an
those for quadrupoles [13] with the appropriate integer chosen so that φ (measured) − φ (theory)
cross terms: < π. ωsb is the measured betatron frequency in
the relevant plane. If ωsb < 0 (lower sideband)
θx,y = −KS Δ (y, x)
is measured, the sign of θs (i) must be changed.
where θx,y is the kick in x or y plane from a dis- With digital sampling of the BPM signal the time
placement in the y or x plane respectively and KS of flight correction is intrinsic in the sampling pro-
the sextupole strength, to be varied. 80 μm typical cess.
error is estimated in reference [13]. All buttons should be measured and averaged
Beta function measurement The β function with proper sign since local coupling will affect
may be directly measured at each quadrupole by the betatron phase measurement at a single but-
ton. Typical achievable resolution is 1◦ for a sin-
modulating the current in that quadrupole and

measuring the change in betatron tune using gle measurement and ∼ 0.2◦ with averaging [14].
Tune measurement The transverse and longi-
β(s)
Δν = − ΔK(s)l tudinal motions of the beam in a periodic accel-
4π erator are sampled by a monitor at a single az-
The accuracy of this measurement is degraded by imuth. For a bunched beam this sampling pro-
hysteresis in the magnet field and potentially by duces a series of spectral lines at frequencies
coupling resonances. It is particularly unsatisfac- fi = |if0 + fβ,s| (i=integer; the negative frequen-
tory at locations where β(s) is small. cies become the lower sidebands). The useful
The beam may be “kicked” and its position frequency spectrum is limited by the finite beam
measured on several successive turns, in which bunch length or pickup electrode characteristics.
case the envelope of the free betatron oscillation A simple but highly sensitive detection
follows βx,y . Alternatively measurements of scheme is shown in Fig.2 [15]. The fast signals
several non-degenerate closed orbit errors may be from a pair of pickups are peak detected then dif-
fitted with β and the corresponding phase ϕx,y (s) ferentially amplified. The time constant Rf Cf
as free parameters (Sec.4.7.5). is matched to the required bandwidth. Measure-
A direct measurement of betatron phase ad- ment resolution can approach 10−5 in tune frac-
vance may be accomplished by shaking the beam tion (sideband frequency) with this method.
at its betatron frequency and measuring the signal Hadron storage rings require special consider-
phase θS at each BPM [12]. The betatron phase ation [16] to minimize emittance growth. Proton
advance between a point si and a reference point and ion storage rings may use Schottky spectra
s0 is: for both bunched and unbunched beam diagnos-
si tics. (Sec.7.4.17)
1
ϕ (si ) − ϕ (s0 ) = ds Once a high resolution tune spectrum is in
s0 β(s)
hand, a wide variety of measurements follow, for
This method is particularly useful where β is example: chromaticity (dν/(dp/p0 )) (Sec.2.1.1);
small or where beam aperture is limited, and can beta functions (see above); optics nonlinearity
be carried out with minimal perturbation to ma- (Sec.2.3.2); impedances (Secs.7.5.2 and 7.5.3).
chine operation.
In analogue systems the measured phase
θS (i) contains contributions from the beam time
of flight, as well as cable delays. Here one must
make a baseline measurement, θrev (i), at the rev-
olution frequency (without excitation). The beta-
tron phase is then:
ϕ (i) = θs (i) + 2πm (i) +
ωs b Figure 2: Direct Diode Detection for tune measure-
[θrev (i) − 2πp (i)] +θ ment [13].
ω0
349
References 4.7.3.2 SVD algorithm

[1] R. Shafer, AIP Proc. 249 (1992) 601
When M ≥ N , the Singular Value Decomposi-
[2] M. Palmer et al, IPAC 10, 1191
tion (SVD) [1] of matrix A has the form A =
[3] X.J. Wang, I Ben-Zvi, Z. Segalov, EPAC 96, t
1576
UWV ⎡ , ⎤
(1) (2) (N )
u1 u1 · · · u1 ⎡ ⎤
[4] R. Lill et al, PAC 07, 4366 w1
⎢ (1) (2) (N ) ⎥
[5] S.T. Boogert et al, IPAC 10, 1140 ⎢ u2 u2 · · · u2 ⎥ ⎢ w2 ⎥
[6] M. Wendt et al, DIPAC 09 A=⎢ ⎢ .. .. .. ⎥⎥⎣
⎢ .. ⎥ ⎦
[7] J. Irwin et al, PRL 82 (1999) 8 ⎣ . . . ⎦ .
(1) (2)
u u ··· u
(N ) wN
[8] J. Borer et al, EPAC 92, 1082 M M M
[9] F. Ruggiero, CERN SL-91-38-AP ⎡ ⎤
(1) (1) (1)

[10] J. Shanks et al, PAC 09, WE6PFP104 v1 v2 · · · vN
[11] A. Wolski, F. Zimmermann, CERN-AB-2004- ⎢ ⎥
⎢ (2)
v1
(2)
v2 ⎥ (2)
· · · vN
010, LBNL Report - 54360, ATF-03-08 (2004) ×⎢
⎢ .. ..
⎥
⎥ .. (4)
⎣ . ⎦
[12] J. Shanks et al., IPAC 10, 4620 . .

[13] N. Terunuma et al, ICFA Beam Dynamics Panel (N ) (N ) (N )
v1 v2 · · · vN
Newsletter 50 (2009) 36
[14] D. Sagan et al, EPAC 96, 1335 U is the M × N matrix whose column vectors
[15] M. Gasior et al, BIW 10, TUCNB01 u(α) , (α = 1, ..., N ) form an orthonormal set,
[16] J. Klute et al, DIPAC 97 Frascati, 230
Ut U = I. W is N × N diagonal matrix with
non-negative elements. Vt is the transpose of
4.7.3 Orbit Correction the N × N matrix V, whose column vectors
J. Wenninger, CERN v (α) , (α = 1, ..., N ) form an orthonormal set,
Vt V = VVt = I.
4.7.3.1 Global orbit correction
From Eq.(4), it follows that (α = 1, ..., N ),
Consider a storage ring with M beam position
monitors (BPM) and N correctors. Orbit dis- Av (α) = wαu(α) , Atu(α) = wαv (α) (5)
placements d (M -component vector) arising from
corrector kick angles θ (N -component vector) are and
determined by the M × N linear response matrix
AAtu(α) = wα2 u(α) , At Av (α) = wα2 v (α) (6)
A,
A θ = d (1)
√ When none of the diagonal elements wα van-
βm βn θ = −VW−1 Ut d.
ish, the solution of Eq.(2) is
Amn = cos (|φm − φn | − πν)
2 sin πν
d may be expanded in terms of eigenvectors
The elements of A may be obtained from the ma-
u(α) [2],
chine model or be determined experimentally by
measuring deviation at each BPM resulting from
N
exciting each corrector individually. d = Cα u(α) + d0 (7)
The task of the orbit correction is to find a α=1
set of corrector kicks θ that satisfy the following
relation, where Cα = d · u(α) , while d0 corresponds to
d + Aθ = 0 . (2) the uncorrectable part of the orbit. The corrector
In general the number of BPMs (M ) and the num- strength required for correction is
ber of correctors (N ) are not identical and Eq.(2)
is either over- (M > N ) or under-constrained N
Cα (α)
(M < N ). In the former and most frequent θ = − v (8)
α=1
wα
case, Eq.(2) can not be solved exactly. Instead,
an approximate solution must be found, and com-
If a given wα = 0 indicating that the matrix is sin-
monly used least square algorithms minimize the
gular, one discards the corresponding term from
quadratic residual
Eq.(8). An example for an eigenvalue spectrum is
S = d + Aθ2 . (3) given in Fig.1 for LEP.
350
0
on few eigenvectors with the largest eigenvalues
log10(wn/w1)
Vertical Plane
-1 are similar to corrections of the main harmonics.
Such a scheme spreads out the correction of a few
-2 kicks over the whole machine which can be an as-
set when the strength of correctors is limited. To
-3
compensate an isolated kick locally, a large num-
-4 ber of eigenvectors must be included in the cor-
rection such that the linear combination forming
-5
0 100 200 300
θc converges to a single nonzero corrector.
Eigenvalue Number Singularities of the response matrix, associ-
ated to very small eigenvalues, are handled more
Figure 1: Vertical orbit eigenvalue spectrum for LEP. easily with SVD, since it is sufficient to avoid us-
The last 4 eigenvalues correspond to singular solu-
ing the corresponding eigenvectors in the correc-
tions in the low-beta sections around the interaction
tions procedure. For the MICADO algorithm, it
points.
is necessary to regularize matrix A by removing

redundant correctors.
In practice one may want to limit the num-
ber of eigenvalues used for the correction to con- 4.7.3.4 Local orbit bumps
trol the r.m.s. strength of the orbit correctors or to
avoid small eigenvalues that are very sensitive to A local bump may be built from three correctors
the accuracy of the model. with deflections θ1,2,3 at locations 1, 2, 3. The de-
flections may be expressed in terms of the lattice
The SVD algorithm is ideally suited for feed-
parameters,
back application since the correction can be cast
in the simple form of a matrix multiplication once
the SVD decomposition has been performed. This θ2 β1 sin(φ3 − φ1 )
= −
provides a fast and reliable correction procedure θ1 β2 sin(φ3 − φ2 )
for a realtime feedback.
θ3 β1 sin(φ2 − φ1 )
= − (9)
4.7.3.3 MICADO algorithm θ1 β3 sin(φ2 − φ3 )
MICADO [3] is a least square correction algo- At a target point t between 1 and 2, the position
rithm based on Householder transformations. MI- and angular displacements are
CADO performs an iterative search for the most
effective corrector and is, together with SVD, one dt = θ1 β1 βt sin(φt − φ1 )

of the most common orbit correction algorithms. β1

For a non-singular matrix, a MICADO correction dt = θ1 cos (φt − φ1 )
with all N correctors and an SVD correction with βt

all N eigenvectors yield identical solutions. For −αt sin (φt − φ1 ) (10)
corrections with a limited number of correctors or
eigenvectors, and for singular matrices, the two At a point t between 2 and 3,
algorithms converge differently.
A major difference between SVD and MI- dt = θ3 β3 βt sin (φ3 − φt )

CADO is the corrector strength distribution, MI-
β3
CADO using fewer but also much stronger kicks. dt = −θ3 cos (φ3 − φt )
βt
The corrector strength r.m.s. can be easily con-
trolled with SVD over the number of eigenvalues +αt sin (φ3 − φt ) (11)
that are included in a correction. A correction
of a small number of localized kicks is very ef- To control both position dt and angle dt at
fectively handled by MICADO, particularly when the source point, a four-magnet local bump is re-
the response matrix is accurate, in which case quired. The four-magnet local bump with correc-
MICADO can be used to identify the sources of tor locations 1, 2, 3, 4, where the source point t
the kicks. On the other hand, corrections based is located between correctors 2 and 3 is given in
351
terms of optics functions by: Skew quadrupoles A skew quad is a quad ro-
dt (cos(φt − φ2 ) − αt sin(φt − φ2 )) tated by 45◦ ,
θ1 = √
βt β1 sin(φ2 − φ1 ) 1 Kf + Kd −Kf + Kd
Qskew = (4)

d βt /β1 sin(φt − φ2 ) 2 −Kf + Kd Kf + Kd
− t √ √
sin(φ2 − φ1 ) For a thin skew quad, l → 0 and k sin( kl) → f1 ,
−dt (cos(φt − φ1 ) − αt sin(φt − φ1 ))
θ2 = √ 00
βt β2 sin(φ2 − φ1 ) Qthin =
I Kt
, Kt = 1 (5)

d βt /β2 sin(φt − φ1 )
Kt I f 0
+ t
sin(φ2 − φ1 )
−dt (cos(φ4 − φt ) − αt sin(φt − φ4 )) 4.7.4.2 Solenoids

θ3 = √
βt β3 sin(φ4 − φ3 ) Longitudinal fields The transfer matrix for the

dt βt /β3 sin(φt − φ4 ) motion through a longitudinal field is
+
sin(φ4 − φ3 ) Ms1 Ms2
Mlong = (6)
dt (cos(φ3 − φt ) − αt sin(φt − φ3 )) −Ms2 Ms1
θ4 = √
βt β4 sin(φ4 − φ3 ) where ks = e
2pc Bz
and
dt βt /β4 sin(φt − φ2 ) 1
− (12) 1 ks sin ks z
sin(φ4 − φ3 ) Ms1 = (7)
0 cos ks z
1
References Ms2 =
0 ks (cos ks z − 1) (8)
0 − sin ks z
[1] W.H. Press et al, Numerical Recipes, Cambridge
U. Press (1988) p.52 Radial fringe The fringe field of a solenoid is
[2] A. Friedman, E. Bozoki, NIM A344 (1994) 269 radial, of equal magnitude and opposite direction
[3] B. Autin,Y. Marti, CERN report ISR MA/73-17 at each end. The elements of the transfer matrix
(1973) for the radial fringe are
(x|x0 ) = cos χ cosh χ
4.7.4 Measurement and Diagnosis of (x|x0 ) = √ 1 (sin χ cosh χ
2kr
+ sinh χ cos χ)
Coupling and Solenoid Compensation
D. Rubin, Cornell U. (x|y0 ) = sin χ sinh χ
(x|y0 ) = √ 1 (sin χ cosh χ − sinh χ cos χ)
In e+ e− storage ring colliders, luminosity is in- 2kr
versely proportional to the vertical dimension of (y|x0 ) = −(x|y0 )

the ribbonlike beams. The most important source (y|x0 ) = −(x|y0 )
of vertical beam size is from its coupling to hori- (y|y0 ) = cos χ cosh χ
zontal and longitudinal motion.
(y|y0 ) = √2k
1
(sin χ cosh χ + cos χ sinh χ) (9)
r
4.7.4.1 Sources of transverse coupling
kr 1 e
where χ = 2 z, kr = 2a pc Bz , and a is the
Tilted quadrupoles The 4 × 4 transfer matrix
for a horizontally focusing quad with strength k length (along z) of the pole tips. [(x |x0 ), (x |x0 ),
and length l can be written as (Sec.2.2.1) etc. are obtained by differentiating (x|x0 ), (x|x0 ),
etc. with respect to z.] In the limit of a thin radial
Kf 0
Mquad = (1) fringe a → z → 0 and kr z = k2a sz
→ k2s , the
0 Kd
transfer matrix becomes
where Kf,d are appropriate 2 × 2 matrices. The
matrix for a quad rotated about an angle θ is I Ks
Mfringe = , (10)
−Ks I
Qrot = R−1 (θ)Mquad R(θ) (2)
where where
I cos θ I sin θ 0 0
R(θ) = (3) Ks = (11)
−I sin θ I cos θ ks
2 0
352
Symplecticity of solenoid maps Eqs.(6), (9) If kr2 > kq2 , then

are not symplectic, but the solenoid matrix
Msol = Mfringe Mlong M−1
fringe is symplectic. See (x|x0 ) = 2 2
h (φ cos ζ cosh ζ + kq sin ζ sinh ζ)
Eq.(21), Sec.2.2.1. (x|x0 ) = 2α − +
hφ2 (−g sin ζ cosh ζ + g sinh ζ cos ζ)
Solenoid lens Msol can also be written as a (x|y0 ) = 2kr
h (sin ζ sinh ζ)
combination of rotations of an angle θ = ks l/4
(x|y0 ) = −2kr α
2 (cos ζ sinh ζ − sin ζ cosh ζ)
and a thick lens that focuses in both planes with hφ
e
focusing strength k = (ks /2)2 , where ks = pc Bz (y|x0 ) = −(x|y0 )
and l is the solenoid length, (y|x0 ) = −(x|y0 )
2 (y|y0 ) = h (φ cos ζ cosh ζ − kq sin ζ sinh ζ)
2 2
Msol = R k4s l F k4s , l R k4s l (12)
(y|y0 ) = −
hφ2 (g cosh ζ sin ζ − g sinh ζ cos ζ)
2α +

Kf (k, l) 0
F= 1
0 Kf (k, l) where φ = |kq2 − kr2 | 4 , h = 2φ2 , g± =
√
Kf is as in Eq.(1). ±φ2 − kq , α = φ/ 2 and ζ = αz. Again

(x |x0 ), (x |x0 ), etc. are obtained by differentiat-
Superimposed solenoid and quadrupole fields ing (x |x0 ), (x |x0 ), etc. with respect to z. Matri-
The matrices MQL and MQR are given for super-
ces MQL and MQR are not symplectic, but the
imposed quadrupole and longitudinal fields and thin −1 is symplec-
for superimposed quadrupole and radial fields. combination Mthin QR MQL (MQR )
The elements of MQL are tic. MthinQR is M QR in the limit of zero length
fringe. See Eq.(11).
(x|x0 ) = − f 1kq (g + θ−
2
cos θ+ z − g − θ+
2
cosh θ− z)
Sextupoles When there is a vertical closed orbit
(x|x0 ) = 1 +
f kq (−g θ− sin θ+ z + g − θ+ sinh θ− z)
y0 at a sextupole, the sextupole behaves as a skew
(x|y0 ) =
kq ks
f |kq | (−θ− sin θ+ z + θ+ sinh θ− z) quad with strength k = 2eB0 /(pcr02 ), where B0
and r0 are the field and the radius at the pole tip.
(x|y0 ) = ks
f (cos θ+ z − cosh θ− z)
The sensitivity of the luminosity in e+ e− colliders
(y|x0 ) = (x|y0 ) to vertical orbit details is related to this property.
(y|x0 ) = −(x|y0 )
−
f (g cos θ+ z − g cosh θ− z)
1 +
(y|y0 ) = 4.7.4.3 Coupling matrix analysis
(y|y0 ) = 1 − +
f kq (g θ+ sin θ+ z + g θ− sinh θ− z) (13)
Normal modes The one-turn transfer matrix T
is decomposed into normal modes as [1, 2]
The focusing strength of the quadrupole is kq ,
e
ks4 + 4kq2 , θ± = M n
ks = pc Bz , f = T= = VUV−1 (15)
mN
| 12 (ks2 ± f )|, and g± = kq − 12 (ks2 ± f ). El-

ements (x |x0 ), (x |x0 ), etc. are obtained by dif- A0 γI C
where U = , V= (16)
ferentiating (x|x0 ), (x|x0 ), etc. with respect to z. 0 B −C† γI
If kq2 > kr2 (where kr = ks /(2a)), elements
of MQR are and γ 2 + detC = 1. A and B are the one-
turn matrices for each of the two normal modes.
(x|x0 ) = 1 + −
h (g cos φz − g cosh φz)
Symplectic conjugate C† = −SCt S (see also
(x|x0 ) − Sec.2.1.3). The laboratory phase space coordi-
h2 (g sin φz − g sinh φz)
2φ +
=
nates x = (x, x , y, y ) are related to the nor-
h (cosh φz − cos φz)
kr
(x|y0 ) = mal mode coordinates w = (w, w , v, v ) by
(x|y0 ) = h2 (sinh φz − sin φz)
2kr φ
x = Vw. The same relation holds for energy
(y|x0 ) = −(x|y0 ) displacements so the normal mode dispersions
may be calculated from (Du , Du , Dv , Dv )t =
(y|x0 ) = −(x|y0 )
−
V−1 (Dx , Dx , Dy , Dy )t . Given the normal mode
h (g cosh φz − g cos φz)
1 +
(y|y0 ) = Courant-Snyder parameters and dispersions the
(y|y0 ) = 2φ + −
h2 (g sinh φz − g sin φz) (14) normal mode emittances can be calculated.
353
Phase space normalization of the normal Coupling resonances Suppose that a coupling
mode vectors yields error δa, δb is introduced at s1 . Then propagate a
GA 0 and b around the ring and back to s1 according to
w̄ = Gw = w (17)
⎡
0 GB
⎤ e−iΣ a = a + δa, e−iΔ b = b + δb (26)
√1 0 Σ = 2π(νA + νB ) and Δ = 2π(νA − νB ). The
GA,B = ⎣ αA,B ⎦
βA,B
√ βA,B solution to Eq.(26) for a and b at s1 is
βA,B
ie−i 2 ie−i 2
Σ Δ
We define
γI C̄ a= δa, b = δb (27)
V̄ ≡ GVG−1 = (18) 2 sin Σ2 2 sin Δ
2
−C̄† γI
Note that near the sum resonance the fast wave (a)
Propagating the coupling matrix Elements of

the coupling matrix can be propagated through is magnified and near the difference resonance the
a lattice given the intervening transfer matrices. slow wave (b) tends to dominate.
Consider the one-turn 4 × 4 matrix at s1 with
T1 = V1 U1 V1−1 . At another point s2 , we have

4.7.4.4 Measurement of coupling
−1 M12 m12 Tune split If the full turn betatron phase ad-
T2 = V2 U2 V2 . Let T12 = be
n12 N12 vance for both horizontal and vertical motion can
the matrix that propagates from s1 to s2 , then be tuned to the coupling resonance, the splitting
U2 = WU1 W−1 with W = V2−1 T12 V1(19) of the normal mode tunes can be a useful measure
Since U1,2 are block of machine coupling. Eqs.(15, 16) give
W is also block
−1diagonal,
E 0
diagonal, W =
0 F−1
. (For very strong tr(A − B) = tr(M − N)2 + 4 det(m + n† )
coupling, W may be off block diagonal [3].) The
coupling matrix at s2 can be written in terms of m + n† = −tr(A − B)γC (28)
the elements of the coupling matrix at s1 by At the coupling resonance, where νh = νv ,
C2 = (M12 C1 + γ1 m12 )F−1 12 (20) tr(M − N) = 0, and Eq.(28) becomes
where
cos 2πνA − cos 2πνB = det(m + n† ) (29)
γ22 = det(n12 C1 + γ1 N12 )
E12 = (γ1 M12 − m12 C†1 )/γ2 Thus the splitting of the normal mode tunes is
proportional to det(C). If the full turn matrix
F12 = (n12 C1 + γ1 N12 )/γ2 (21)
is block diagonal, then m = n = 0, and the
When there are no couplers between s1 and splitting of normal mode tunes can be reduced to
s2 , n and m are zero and zero. By adjusting coupling correctors, such as
C2 = M12 C1 F−1 12 = M12 C1 N12
−1
(22) skew quads, to minimize the normal mode tune
2
Define M̄12 = GA M12 (GA ) 1 −1 and N̄12 = split near the sum resonance we are reducing a
G2B N12 (G1B )−1 where G1A is the normalization and minimizing the tune split at the difference res-
matrix for horizontal motion at s1 , G2A is the nor- onance reduces b.
malization matrix at s2 , etc. Then Eq.(22) gives Tune split due to a thin skew quad or solenoid
G2A C2 (G2B )−1 = C̄2 = M̄C̄1 N̄−1 (23) Consider a machine tuned to the difference reso-
The normalized transfer matrices M̄ and N̄ are nance that is described by the full turn block di-

simple rotations in the normalized phase space M 0
and C̄2 = R(φA )C̄1 R(φB )−1 , and R is a rota- agonal matrix at some point s by F = .
0 N

tion matrix. Define complex coupling coefficients cos μh βh sin μh
a = c̄11 − c̄22 + i(c̄12 + c̄21 ) We can write M = , and
−γ sin μh cos μh
b = c̄11 + c̄22 + i(c̄21 − c̄12 ) (24) similarly for N. Here μh = 2πνh . At the dif-
Then ference resonance, tr(M − N) = 0. Then intro-
a2 = e−i(φA +φB ) a1 , b2 = e−i(φA −φB ) b1 (25) duce a thin skew quad Qthin as in Eq.(5). The
where φA + φB is the sum of the normal mode perturbed full turn matrix is

phase advances between 1 and 2, and φA − φB is M MKt
the difference. P = FQ = (30)
NKt N
354
From Eq.(29) we have =⇒ √

⎡ ⎤ ⎡ ⎤
x γ A βA cos φ
cos μA −cos μB = det[MKt + (NKt )† ] (31)
⎢x ⎥ ⎢ √ x ⎥
⎣ y ⎦ = ⎣− β (c̄ cos φ + c̄ sin φ)⎦
A A 22 12
Since μh ∼ μv , y y
√ =⇒
1 βh βv
Δν = νA − νB ∼ (32) x = γ A βA cos φ
2π f
To compute the tune split due to a thin solenoid, y = − A βB (c̄22 cos φ + c̄12 sin φ)

we substitute = − A βA c̄222 + c̄212 cos(φ + δφ) (34)

I I sin θ
Mthin sol = where δφ = tan−1 c̄c̄12 . The ratio of the horizontal
−I sin θ I 22
and vertical amplitudes of the normal mode mo-
for Q in Eq.(30) and then for the thin solenoid tion is

1 βB
1 (βh + βv ) (y/x)Amp = c̄212 + c̄222

Δν = θ √ γ βA
(35)
2π βh βv
The ratios of the component of the vertical motion
where θ = ks l, and ks defined above.
that is in phase and out of phase with the horizon-
Coupling wave due to thin skew quads and tal motion are
solenoids Eqs.(27, 28) and thin coupler matri-
(y/x)in phase = (1/γ) βB /βA c̄22 (36)
ces are combined to give a first order expression
for the complex coupling parameters a and b. The (y/x)out of phase = (1/γ) βB /βA c̄12 (37)
coupling at j due to skew quads and/or solenoids The in-phase coupled motion corresponds to a tilt
at k is of the real space beam ellipse and the out of phase
+ − motion to an increase in the height of the ellipse.
ρk e−i(ξjk + 2 ), bj = χk e−i(ξjk + 2 )
Σ Δ
aj =
Excitation of the orthogonal mode (B = 0)
k k
yields
where
(x/y)in phase = (1/γ) βA /βB c̄11
# $
βhk βvk 2 sin Δ 2i cos Δ (x/y)out of phase = (1/γ) βA /βB c̄12
2 2
ρk = +
tr(A − B)γ fk gk Measurement The elements of the coupling
# $
matrix can be measured at BPM’s. A mag-
βhk βvk 2 sin Σ2 2i cos Σ2 netic shaker excites one or the other of the
χk = + normal modes. The transfer function between
tr(A − B)γ fk gk shaker drive and horizontal and vertical motion
at the BPM is measured. The relative phase
±
ξjk = φjh − φkh ± (φjv − φkv ) of the vertical and horizontal motion δφ =
Σ = μh + μv , Δ = μh − μv c̄12 /c̄22
and the relative amplitude is (y/x)Amp =
(1/γ) βB /βA c̄212 + c̄222 . See also Sec.4.7.5.
In the limit of weak coupling tr(A − B) = The measurement of c̄12 in CESR is repro-
tr(M − N) = 2(cos μh − cos μv ), and γ ∼ 1. ducible at the 0.5% level. Coupling errors are typ-
Relative amplitude and phase When a normal ically reduced to < 1%. c̄22 is somewhat more
mode is excited, the relative amplitude and phase difficult to measure since the vertical amplitude is
of horizontal and vertical motion can be measured typically the horizontal amplitude.
at a BPM [7]. Consider the motion in the lab-
oratory coordinate system as a consequence of 4.7.4.5 Solenoid compensation
excitation of the A mode. If the initial normal-
Large detector solenoids are strong coupling ele-
ized, normal mode vector is Gw0 = w̄0 =
√ ments. The compensation consists of coupling el-
( A , 0, 0, 0), after some number of turns when
ements deployed to globally decouple the motion
the normal mode A has propagated through some
outside the compensation region and to preclude
phase advance φ, we have
feedthrough of horizontal motion to beam height
x = Vwn = VUn w0 = G−1 V̄Ūn w̄0 (33) at the IP.
355
Compensation with anti-solenoids The cou- Compensation with three pairs of rotated
pling of the solenoid can be compensated by quadrupoles But it is not necessary that
anti-solenoids of equal but opposite integrated Tright→IP be block diagonal. Suppose that the
strength. If a pair of half strength anti-solenoids focusing quadrupoles are deployed symmetrically
are placed symmetrically about the main solenoid, about the IP and the compensating rotated quads
then there is vanishing coupling at the mid-point antisymmetrically, (equal but opposite rotation
of the main solenoid. The anti-solenoids may be angles). (Note that the radial fringe at the left end
displaced from the ends of the main solenoid by a of the solenoid is of opposite sign to the fringe at
field free drift. However, the solenoid matrix does the east end.) It is shown [5] that if
not commute with that of a quadrupole. And anti- 1. T31 = 0
solenoids are ineffective if there are intervening
2. T32 = 0
quadrupole focusing elements.

3. −T41 T12 + T42 T11 = 0
Compensation with skew quads Alternatively,
where T ≡ Tright→IP , then Tthrough is block di-
skew or rotated quadrupoles can be used to effect
agonal. The three constraints imposed on the ma-

compensation. In general, four pairs of such cou-
trix can be realized with three independent pairs
pling elements are required. In a symmetric IR
of rotated quadrupoles. Independence of course
(where the final focus elements are mirror sym-
implies nondegenerate betatron phase advance be-
metric about the IP), three pairs of coupling ele-
tween the rotated quads.
ments are sufficient.
In CESR (5 GeV/beam), the solenoid that
Transfer matrix Define the compensation re- is centered at the interaction point has a field
gion of the lattice to include the detector solenoid of 1.5 T and length of 3.5 m. The final fo-
and all of the compensation elements. cus quadrupoles are rotated antisymmetrically
to compensate the coupling of the solenoid.
1. The coupling is compensated if the 4 × 4
The three pairs of quadrupoles are rotated by
matrix (Tthrough ) through the compensation re-
±4.5◦ , ±6.8◦ and ±14.0◦ .
gion is block diagonal.
Coupling matrix The compensation conditions
2. Vertical displacement at the IP is indepen- can be cast in terms of the C-matrix elements.
dent of the horizontal phase space outside the According to Eqs.(36, 37), the blowup of the
compensation region. beam cross section is proportional to c̄12 and the
The second criterion imposes constraints on tilt of the beam is proportional to c̄22 . The require-
the matrix (Toutside→IP ) that propagates trajecto- ment of no vertical enlargement or twist at the IP
ries from outside the compensation region to the implies that
IP, Toutside→IP = Toutside→IP = 0.
31 32 c̄ 0
Let Tright→IP and Tleft→IP be the matrices C̄IP = 11 (38)
c̄21 0
that transport trajectories through the right and
left halves of the compensation region to the IP. And the machine is globally decoupled if at any
Then Tthrough = Tright→IP JT−1 left→IP J, where
and all points outside the insertion

⎡ ⎤ C̄ =
00
(39)
1 0 0 0 00
⎢ 0 −1 0 0 ⎥
J=⎣
0 0 1 0 ⎦ Eqs.(38, 39) are true if and only if the compensa-
0 0 0 −1 tion criteria described above are satisfied.
The three pair compensation scheme can also
Compensation with four pairs of rotated be described in terms of the C-matrix elements.
quadrupoles It is clear that if Tright→IP and If the compensating elements are placed antisym-
Tleft→IP are block diagonal then Tthrough is metrically about the IP, then by symmetry, the tilt
block diagonal and the compensation criteria are of the beam c̄22 at the IP is zero. Similarly, by
satisfied. Symplectic matrices have the property symmetry, c̄22 = c̄11 at a symmetry point outside
that if one off diagonal 2 × 2 block is all zero, the compensation region, (such as the point dia-
then so is the other. Therefore four independent metrically opposite the IP). So, if antisymmetric
coupling elements (skew or rotated quadrupoles) pairs of rotated or skew quads are used to compen-
are sufficient to satisfy the compensation criteria. sate for the solenoid, then three constraints must
356
be satisfied to ensure that there is no coupling of a BPM is given by Eq.(1), Sec.4.7.1. Averaging
horizontal motion into the vertical plane at the IP, the data for many turns, these closed orbit shifts
and that the full turn matrix evaluated everywhere can often be measured with accuracies of 1 μm or
outside the compensation region is block diago- better, so they provide a wealth of precise optics
nal. They are information.
1. c̄∗12 = 0 Beam-based alignment In this application,
θ is generated by changing the strength of a
2. c̄outside
12 =0 quadrupole rather than a steering magnet. If the
3. c̄outside
21 =0 initial closed orbit of the beam relative to the
magnetic center of the quad is x, then changing
where c̄outside
refers to coupling parameters at the
the quad strength by ΔKL gives θ = xΔKL.
symmetry point outside of the compensation re-
Beam-based alignment simply entails steering the

gion. Then Eqs.(38, 39) are satisfied.
beam toward the quad center by minimizing the
Dispersion Vertical dispersion at the IP in- measured orbit shift associated with changing the
creases the vertical beam size and is a source
quad strength [1, 2].

via the beam-beam interaction of synchrobeta-
Beam-based alignment is useful because
tron coupling. As long as there are no bending
BPMs are more accurate for measuring shifts in
magnets and zero horizontal dispersion inside of
the closed orbit, but quads are more accurate for
the compensation region, then the compensation
measuring the absolute orbit. The typical align-
schemes described above will generate no verti-
ment error of a quad magnetic center is 200 μm,
cal dispersion at the IP. But if there is horizontal
while the typical alignment error of a BPM elec-
dispersion inside of the compensation region then
trical center is 500 μm. Furthermore, centering
additional constraints will be required.
the beam through in the quads avoids orbital os-
cillation due to power supply ripple in the quad.
References Orbit response matrix analysis In this appli-
[1] D. Edwards, L. Teng, PAC 73, 885 cation, the closed orbit shifts are considered a re-
[2] M. Billing, Cornell Report CBN 85-2 (1985) sponse matrix A [Eq.(1) of Sec.4.7.3]. Given the
[3] D. Sagan, D. Rubin, Cornell Report CBN 96-20 design gradients of all quads, one can calculate
(1996) the expected response matrix.
[4] S. Peggs, PA 12 (1982) 219 Conversely, given the measured response ma-
[5] D. Rubin, Cornell CON 84-4 (1984) trix from the real accelerator, one can calculate
[6] D. Rubin, Cornell CON 84-5 (1984) the the real gradients in each of the quads [3, 4].
[7] P. Bagley, D. Rubin, PAC 87, 1301 This is done by finding the computer model of
the storage ring that best reproduces the mea-
sured response matrix. The model quad gradients
4.7.5 Modeling and Control of Storage Rings
are varied to minimize the χ2 difference between
Using Orbit Measurements
the model and measured response matrices. The
J. Safranek, SLAC
steering magnet calibrations (the θ’s) and BPM
Due to magnet strength and misalignment errors, gains are also varied in the χ2 fit.
a real storage ring differs from its design. To The accuracy with which the fit parameters
adjust the real ring to achieve the design perfor- can be determined depends on the particulars of
mance, one can apply methods which are based on a storage ring, particularly on the ratio of BPM
measured orbits. Descriptions here are for storage number to quad number and on the BPM noise.
rings, but many of the techniques are also useful The example for the NSLS storage ring is shown
for linacs. We first describe methods using closed in Tab.1.
orbits. Then methods using turn-by-turn measure- In some cases the orbit response matrix
ments of betatron oscillations are described. Apli- data is insufficient to accurately constrain the
cation of a Model Independent Analysis to storage fit quadrupole gradients, and the χ2 fit either
rings will be discussed in Sec.4.8.2. does not converge or results in unrealistic fit
Closed orbit analysis When the strength of a quadrupole strengths. In such cases it has been
steering magnet is changed to deflect the beam shown that modified χ2 fit, which includes con-
by an angle θ, the resulting closed orbit shift at straints on the quadrupole gradient variation,
357
Table 1: The rms uncertainty in the fit parameters due
to noise in the orbit response matrix measurement.
Also included is the resulting uncertainty in the β-
functions around the ring.
parameter rms uncertainty

quadrupole gradients .04%
BPM gains .5%
steering-magnet .5%
calibrations
β-functions .08%
Figure 2: The gradient variations in ALS predicted by

fitting the response matrix.
Figure 1: The measured horizontal dispersion in the

NSLS X-Ray Ring compared to the design model and
to the model fit to the response matrix.
yields a fit optics that reliably reproduces the ac-

tual accelerator optics [4]. Figure 3: The VUV Ring optics before correction.
Often there are significant discrepancies be-
tween the measured and design optics before ap-
plying response-matrix analysis. The model fit focusing in the insertion devices. With this under-
to the response matrix greatly improves predic- standing, changes in quadrupole gradients could
tion of the measured dispersion. See e.g. Fig.1, be derived to best restore the design optics. Fig.4
where the response matrix analysis revealed that shows the optics fit to the response matrix after
the biggest source of optics distortion was hori- these changes were implemented.
zontal orbit offsets in sextupoles. The orbit response matrix method can be ex-
Fig.2 shows the result of the response ma- tended to include the coupling terms such as the
trix analysis for ALS [5]. The 24 QD mag- orbit shift in y when steering in x. These coupling
nets have the same gradients by design, but the terms determine the roll misalignments of the
fit model indicated large variations, which were quads, steering magnets and BPMs (Sec.4.7.2).
subsequently confirmed by power supply current Coupling correction (Sec.4.7.4) As an alter-
measurements. native to using the coupling terms in the orbit re-
Once the source of optics distortions are un- sponse matrix for finding the rolls of the quads,
derstood, the gradient errors can be eliminated or one can find instead the strengths of skew quad
compensated to restore the design optics. Fig.3 correction magnets that best cancel the coupling
shows the NSLS VUV Ring optics fit to the orbit terms. The coupling of closed orbit shifts from
response matrix that had been in operation. The the x- to y-plane is not quite the same as the cou-
β-function distortions were associated with the pling of betatron oscillations; however, it is shown
358
Figure 5: The measured vertical beam size in the

Figure 4: The VUV Ring optics with the design peri-
NSLS X-Ray Ring before and after coupling correc-
odicity restored.
tion.
experimentally [6] that reducing the former helps

phase at the BPMs [7]. An example of such mea-
reduce the later.
surements at LEP is shown in Sec.7.5.3, and led
Minimizing the vertical emittance requires
to a measure of the reactive transverse impedance.
minimizing both coupling and vertical dispersion.
The problem can be reduced to linear algebra [6], The harmonic analysis alone does not provide
an accurate measurement of βb because the am-
MKs = −V (1) plitude information is corrupted by BPM gain er-
where the vector V contains all the measured ror. The β-function, however, is the inverse of the
shifts in vertical orbit with horizontal steering derivative of the betatron phase (Sec.2.1.1), so the
magnets (total number of elements equal to the measured betatron phases at consecutive BPMs
number of horizontal steering magnets times the can be used to determine the β-functions [9].
number of vertical BPMs) in addition to the mea- Coupling correction (Sec.4.7.4) Turn-by-turn
sured vertical dispersion (an additional number of measurements are useful in characterizing the
elements equal to the number of vertical BPMs). coupling in two ways: by exciting the beam at its
The matrix M is the measured change in V with two normal mode frequencies, or by giving an im-
each of the skew quads. Eq.(1) is solved for the pulse kick in x and monitoring the coupling into
desired strengths of the skew quads, Ks . The rela- y.
tive weighting of coupling and vertical dispersion With the normal mode technique the beam is
correction can be adjusted empirically to give the excited resonantly at either one of the two trans-
best reduction of vertical emittance. verse modes [10]. One mode is primarily hori-
Fig.5 shows the measured vertical beam size zontal and one primarily vertical, but both will
before and after implementing this coupling cor- show some excitation in each plane due to cou-
rection algorithm in the NSLS X-Ray Ring. The pling. Turn-by-turn digitization of the data pro-
measured ratio of vertical to horizontal emittance vides information that can be used to set the skew
was reduced by a factor of ∼18 to <0.1%. quads to best minimize the coupling.
Turn-by-turn orbit analysis Phase and With the impulse method the beam tunes are
β-function measurements The betatron oscilla- adjusted to the coupling resonance, and the beam
tions of the beam about the closed orbit measured is kicked horizontally [11]. Due to coupling, the
turn by turn at a BPM are oscillations shift back and forth between x and y.
Maximizing the period of this beating minimizes
xb,n = a βb cos (2πnν + φb ) (2) the coupling.
where n is the turn number and a is the oscillation
amplitude. A harmonic analysis of many consec- References
utive turns of digitized betatron oscillation data
can yield an accurate measure of φb , the betatron [1] P. Röjsel, NIM A343 (1994) 364
359
[2] A. Wolski, F. Zimmermann, CERN-AB-2004- We assume there is no beam damping system

010, LBNL Report-54360, ATF-03-08 (2004) present. Often times, feedback systems are em-
[3] J. Safranek, NIM A388 (1997) 27 ployed to damp injection oscillations before deco-
[4] ICFA Beam Dynamics Newletter, No. 44 (2007) herence occurs. It is also assumed that the time-
p. 43, http://icfa-usa.jlab.org/archive/newsletter/ stationary beam distribution has cylindrical sym-
icfa bd nl 44.pdf metry in the normalized phase space (x, αx+βx )
[5] D. Robin et al, EPAC 96, 971 – i.e. no nearby resonances other than an am-
[6] J. Safranek, S. Krinsky, AIP Proc. 315 (1994) plitude dependent tune. For longitudinal phase
163 space, this assumption implies (beam emittance)
[7] J. Borer et al, EPAC 92, 1082 (bucket area).
[8] D. Brandt et al, PAC 95, 570
[9] P. Castro et al, PAC 93, 2103 Steering errors Let σ0 be the incoming rms
[10] P. Bagley, D. Rubin, PAC 89, 874 beam size at a location in the accelerator where
[11] P.J. Bryant et al, PAC 77, 1440 the Courant-Snyder parameters are β0 , α0 . If the
beam enters with an initial offset (Δx, Δx ), the
rms beam size after decoherence is x2 1/2 with

4.7.6 Emittance Dilution Effects [1] 1 2
x2 = σ02 + Δx +(β0 Δx + α0 Δx)2 (1)
M. Syphers, FNAL 2
Important sources of emittance dilution in cir- Optical errors Focusing strength errors (“β-
cular accelerators can be grouped into two cate- mismatch”) Suppose a beamline delivers β and
gories. The first type, mismatch errors, occurs α to a point in an accelerator where its periodic
during the transfer of beam from one accelera- parameters are β0 and α0 . The relative amplitude
tor to another. Impulses, such as from kicker function mismatch, Δβ/β0 , propagates about the
magnets, can also generate dilution in the same accelerator (in the absence of gradient errors)
manner. The second type is from scattering and according to
diffusion processes, such as intrabeam scatter- d2 Δβ Δβ
ing (Sec.2.4.12), beam-gas scattering (Sec.3.3.2), + (2ν0 )2 = −2ν02 det ΔJ (2)
dφ2 β0 β0
power supply noise, etc. Due to synchrotron radi-
ation damping (Sec.3.1.4.2), the beam emittance where ν0 is the unperturbed tune, φ = ψ0 /ν0
in electron synchrotrons damps to an equilibrium is the unperturbed
reduced phase advance, J =
value. Proton synchrotrons are less forgiving, α β
, where γ = (1 + α2 )/β. The elements
however, and the preservation of small emittance −γ −α
is often a serious issue. See Sec.4.3 for emittance of ΔJ are Δβ = β − β0 , etc. It can be shown that
dilution effects in linacs. det ΔJ is an invariant,
2 2
Δβ
4.7.6.1 Injection mismatch β0 + Δα − α0 Δβ β0
det ΔJ = − <0
1 + Δβ/β0
If a beam is injected into a circular accelera- (3)
tor away from the closed orbit, or if the in- After decoherence the rms transverse beam size is
coming beam ellipse is not matched to the peri-
1
odic Courant-Snyder parameters of the accelera- x2 = σ02 1 + | det ΔJ| (4)
tor, the beam would coherently oscillate and tum- 2
ble about the closed orbit, thus occupying more Dispersion function errors If a beamline de-
phase space area than desired. Because real accel- livers a dispersion function D, D to a point in an
erator components have some degree of nonlinear accelerator where its periodic dispersion function
fields present, the motion of the particles will de- is D0 , D0 , then off-momentum particles will see
cohere (Sec.2.3.10), generating an emittance in- injection steering errors which after decoherence
crease. Figs.1 and 2 illustrate the effects of steer- will result in a new transverse beam size
ing and optical mismatch errors. The phase space
evolution is depicted, illustrating the effects of 1 σp 2
x2 = σ02 + ΔD 2 + (β0 ΔD + α0 ΔD)2
decoherence until an equilibrium distribution is 2 p
reached, and an increase in the emittance. (5)
360
Figure 1: Steering error: Beam enters accelerator off-centered, and oscillates about the closed orbit.
Figure 2: Amplitude function error: The same distribution as above is injected centered, but the incoming beam
ellipse is mismatched with the periodic β-function of the accelerator (circular in this illustration).
Longitudinal errors For bunch areas rf function in terms of the individual particle emit-
bucket area, similar expressions as above ap- tance, W ≡ [x2 + (βx + αx)2 ]/β, ψ satisfies the
ply for longitudinal emittance growth due to er- diffusion equation (see also the related Fokker-
rors in synchronized beam transfers between syn- Planck equation in Sec.2.4.7)
chrotrons.
∂ψ ∂ ∂ψ
Energy mismatch An energy error δE at in- =R W (9)
jection leads to a final energy spread (originally ∂t ∂W ∂W
σE ), where R = dW /dt, the average being taken
1 over the particle distribution. The rate of change
ΔE 2 = σE2
+ δE 2 (6)
2 of the transverse (normalized, rms) emittance
Phase mismatch Let φs be the synchronous would be dN /dt = 12 (γv/c)R.
phase, Vrf the rf voltage per turn, η the slip factor, Dipole noise For a single dipole steering error
h the harmonic number, and E0 the synchronous randomly fluctuating each revolution of the accel-
energy. A phase error at injection of δφ leads to a erator with rms value θrms , the emittance growth
final energy spread of rate is
dN 1 2
δφ2 E0 (v/c)eVrf cos φs = f0 (γv/c)β0 θrms (10)
ΔE 2 = σE
2
+ − (7) dt 2
2 2πhη where β0 is the β-function at the location of the
error, and f0 is the revolution frequency.
Bucket mismatch To conserve longitudinal
emittance, the buckets of the two synchrotrons Rf noise If vrms is the rms voltage noise
should be matched. This requires K = K0 , (summed over all the cavities), uncorrelated from
where K = (V | cos φs |)/(h|η|). K corresponds turn to turn, the transverse emittance growth rate
to the incoming beam and K0 to the receiving syn- is
dN 1 e2 vrms
2
chrotron. If the two synchrotrons are mismatched = f0 γH (11)
then the final energy spread is dt 2 (v/c)3 E02
# $ where E0 is the synchronous energy, H = [D 2 +
1 + K/K0 (βD + αD)2 ]/β at the cavities.
ΔE =
2
2
σE (8)
2 K/K0 Coulomb scattering If the scattering is due to
small angle Coulomb interactions between the
4.7.6.2 Diffusion processes beam particles and other material in the beam
chamber, then
Let ψ(x, x , t) be the particle transverse phase
space density function at time t. Suppose some dN 1 13.6 MeV 2 z
= f0 β
process is randomly altering the particles’ phase dt 2 mc2 γ(v/c)3 X0
space amplitude over time. Writing the density (12)
361
Table 1: Relative and absolute increase in (normalized, rms) emittance due to injection errors. Here, N =
(σ02 /β0 )(γv/c), s = σE σφ /ωrf .
Error Source Relative Increase, Δ/ rms Emittance Increase, Δ

Transverse, N
1 Δx2 +(β0 Δx +α0 Δx)2 1 Δx2 +(β0 Δx +α0 Δx)2
Steering Error 2 σ02 2 β0 (γv/c)
2 2
1 ΔD 2 +(β0 ΔD +α0 ΔD)2 σp 1 ΔD 2 +(β0 ΔD +α0 ΔD)2 σp
Disp. Mismatch 2 σ02 p 2 β0 (γv/c) p
β - Mismatch 1
|
2 det ΔJ| 2 | det ΔJ|
N
Longitudinal, s
2 2
2πhηE0
Energy Mismatch 1 δE 1 E0
− (v/c)eV δE
2 σE 2 ωrf rf cos φs E0
2
Phase Mismatch 1
2
δφ
σφ
1 E0
2 ωrf − (v/c)eV rf cos φs
2πhηE0 (δφ)2
√ 2 √ 2
1− 1+ΔK/K0 1− 1+ΔK/K0
Bucket Mismatch 1
2
√ s
2
√
1+ΔK/K0 1+ΔK/K0
where mc2 is the rest energy of a beam particle, References

z its charge, and X0 is the radiation length of the
material which is distributed over a total distance
about the circumference. For example, a high
(1993)
energy proton beam scattering with the residual
gas (assumed to be air) gives
dN P [μTorr] 4.7.7 Real-Time Measurement and Control
= β (1.6 × 10−7 /s) (13) of Tune, Coupling and Chromaticity
dt γ
R. Jones, R. Steinhagen, CERN
Beam-beam elastic scattering In a collider,
elastic scattering of the on-coming beam particles Accurate control of tune, coupling and chromatic-
causes emittance growth. The emittance growth ity is an integral part of modern accelerator oper-
rate due to scattering at a single IP is ation, reducing beam losses associated with insta-
bilities and preserving small beam sizes. The reli-
dN 1 1
= γβ ∗ θrms
2
Lσel able implementation of such a system must over-
dt 2 nB NB come the problems related to the interdependency
1 f0 NB of these measurements. As examples; poor con-
= γ σel θrms
2
(14)
2 4N trol of coupling will lead to tune changes which
where β ∗ is the β-function at the IP, σel is the elas- cannot be compensated by quadrupole magnets
tic scattering cross section, θrms is the rms pro- alone, while changes in chromaticity can lead to
jected elastic scattering angle. Expressions for σel modifications in the observed tune spectra which
and θrms can be found in Secs.3.3.1,3.3.2. can complicate the accurate measurement of the
Longitudinal effects Voltage noise in the rf tune. As with most feedback systems the per-
system, with rms value vrms , causes a growth of formance is ultimately determined by the reliabil-
the longitudinal emittance at the rate ity of the initial measurements. Experience has
shown that in order to provide such measurements
ds 1 E0 2πhηE0 evrms 2 all three of these properties - tune, coupling and
= f0 −
dt 2 ωrf (v/c)eVrf cos φs E0 chromaticity - must be treated simultaneously so
(15) as to provide a robust real-time control for one or
Likewise, rf phase noise causes an emittance all [1, 2].
growth rate of
4.7.7.1 Tune measurement
ds 1 E0 (v/c)eVrf cos φs All betatron tune measurements are based on
= f0 − (δφrms )2
dt 2 ωrf 2πhηE0 measuring the characteristic frequency of the
(16) transverse motion of the beam. In the simplest
362
case the beam is given a single kick using a pow- change in betatron tune as a function of the beam
erful stripline or magnetic kicker, and allowed to energy. This is usually done by varying the rf
oscillate freely. A position pick-up is used to frequency, keeping the magnetic field static. The
measure the resulting beam motion. For stable equations are
beam conditions, the betatron tune is then usu- Δp Δp ΔR
ally the frequency which has the highest ampli- ΔQ = (ξQ) = Q = Q γt2
p p R
tude response in the power density spectrum ob- 2 γ 2 Δf
tained from this data using Fast Fourier Transform −γ
= Q t
,
(FFT) techniques. Where powerful kickers are γ 2 − γt2 f
unavailable or when it is important to minimise
where ΔQ is the change in tune, Δp/p the rel-
the frequency range over which power is put into
ative change in momentum, ΔR/R the relative
the beam, swept frequency “chirp excitation” or

change in radius, Δf /f the relative change in rf
band limited white noise can be used to replace
frequency and Q (ξ) the (normalised) chromatic-
the single kick. In all these cases the measure-
ity. This method lends itself well to continuous
ment rate is determined by the re-triggering ca-
chromaticity measurement, simply requiring the

pability of the exciter and the damping time of the
addition of a small rf modulation while tracking
beam, typically in the Hz range and achieving res-
the tune. To avoid longitudinal emittance blow-up
olutions of 10−3 . . . 10−4 in tune.
the modulation is typically applied either well be-
Recent advances in detector technology and
low or well above the synchrotron tune. Fast mod-
acquisition systems [3] have allowed unprece-
ulation is less frequently used due to uncertainties
dented sensitivity to be obtained with detection
on the bunch rigidity and its high power require-
of oscillations possible at the nanometre level.
ments especially for superconducting high-Q cav-
Such systems typically do not require any ad-
ities.
ditional external excitation sources, providing
Other methods for measuring chromaticity in-
excellent, continuous tune measurements from
clude its determination from the width of the tune
residual beam motion. This is of great advan-
peak, the height of the synchrotron sidebands
tage in hadron accelerators, where any applied ex-
around the main tune peak, and the phase of the
citation leads to irreversible transverse emittance
head-tail motion within a bunch. Although each
blow-up.
has been used to measure and correct for chro-
Where the residual beam motion is not suf- maticity in various machines, none has proven re-
ficient to provide clean continuous tune spectra liable enough for inclusion in an automatic feed-
a Phase Locked Loop (PLL) system can be im- back system.
plemented. In this case the exciter is fed with a
sine wave from a reference oscillator. The max- 4.7.7.3 Coupling measurement
imum amplitude response from the beam, at the
tune frequency, is obtained when the phase differ- The amount of coupling in a circular machine
ence between the observed beam motion and the is described by the complex coupling coefficient
sinusoidally driven exciter is 90◦ . By continually C − = |C − | eiχ (Sec.3.1.4.4), where |C − | is the
adjusting the reference oscillator to meet this con- coupling magnitude and χ is the coupling phase.
dition, the system automatically tracks any tune It can be shown [4] that
changes, resulting in a continuous tune measure- ⎧
⎨ Q = Q − 1 Δ + 1 Δ2 + |C − |2
ment. Tracking precisions down to 10−6 can be 1 x,0 2 2
(1)
achieved. ⎩ Q = Q + 1 Δ − 1 Δ2 + |C − |2
2 y,0 2 2
For particle colliders, where beam-beam
%
forces due to collisions produce complex coherent Δφ1 = φ1,y − φ1,x = χ
(2)
tune spectra, Schottky monitors measuring the in- Δφ2 = φ2,x − φ2,y = ±π − χ
coherent tune have been successfully used to pro-
vide reliable measurements. where Q1 and Q2 are the measured horizontal and
vertical tunes (eigenmodes), while Qx,0 and Qy,0
are the so called “unperturbed tunes”, i.e. the
4.7.7.2 Chromaticity measurement
tunes that would be measured in the absence of
The most common method of measuring the chro- coupling. Δφ1 is the phase of the vertical motion
maticity of a circular machine is to measure the with respect to the horizontal motion and Δφ2
363
that of horizontal with respect to the vertical mo- This individual parameter control complex-
tion. Δ is the difference between the fractional ity is dwarfed by the challenge of operating par-
part of the unperturbed tunes (Δ = Qx,0 − Qy,0 − allel feedback loops on tune, chromaticity, cou-
p, with p an integer). This can be re-written as pling, orbit, radial position and transverse bunch-
( − ( 2√r1 r2 |Q1 − Q2 | by-bunch motion. Even in a fully optimised
(C ( = , scheme, some cross-talk is inevitable: the mo-
(1 + r1 r2 ) mentum modulation required to measure chro-
|Q1 − Q2 | (1 − r1 r2 ) maticity induces tune and radial offsets that are
Δ=
(1 + r1 r2 ) seen by the tune, orbit and radial position feed-
where backs; transverse feedback, by design, minimises
⎧ the very same beam oscillations required to mea-
⎨ r1 = |C − |
A1,y βy
= · sure the tunes. If not addressed at an early design
A1,x
βx 2ν+Δ
,
⎩r = A2,x βx |C − | stage, a naı̈ve one-by-one implementation of these
2 A2,y = βy · 2ν+Δ feedback loops can lead to serious interferences,
coupling and instabilities.
1
ν= Δ2 + |C − |2
2 There are various classic de-coupling strate-
with A1,x and A1,y the amplitudes of eigenmode gies such as: diagonalisation, e.g decoupling
Q1 in the horizontal and vertical plane respec- of horizontal and vertical planes; suppression
tively; A2,x and A2,y representing the amplitudes of known cross-terms, i.e. allowing certain vari-
of eigenmode Q2 in the horizontal and vertical ations which are required for measurements;
plane respectively; βx and βy the optical beta- dead-bands to limit the operational ranges of one
functions at the point of observation. feedback in favour of another; time-scheduling
All of the coherent tune measurement tech- between feedback actions, such as alternating
niques outlined above are detecting the eigentune measurements with transverse feedback op-
modes and their amplitudes, from which the un- eration; choosing different bandwidths for each
perturbed tunes and coupling magnitude can be loop.
inferred. Using PLL or chirp excitation can, in An improved de-coupling strategy is to de-
addition, give access to the coupling phase. rive the dependent variable from the compensated
feedback actuator control signal. In this case the
4.7.7.4 Real-time control of tune, coupling & tune feedback is operated at the maximum desired
chromaticity bandwidth, fully compensating radial modulation
In a first-order control approach the individual pa- induced tune changes. Chromaticity is in turn de-
rameter deviations measured by one of the above rived and corrected from the amplitude of the ac-
techniques are converted into corrections that, tuator signal required to stabilise the tunes. Due
when applied to the magnet circuits, minimise the to the finite bandwidth and gain of the feedback,
observed errors. For most accelerators the rela- the actuator signal does not typically contain the
tionship between magnetic field, required circuit full modulation. An accurate chromaticity esti-
current and induced tune, chromaticity and cou- mate needs to account for this and should be com-
pling changes is sufficiently linear to be cast into plemented by the demodulation of the residual
matrices. The correction to be applied is then tune frequency oscillation remaining on the beam.
typically calculated using the pseudo-inverse of The required dispersion orbit variation and cor-
these matrices (e.g. using singular value decom- responding momentum mismatch need to be ad-
position). Iterating continuously compensates for dressed differently. This is done by subtracting
dynamic parameter deviations and residual errors them dynamically from the orbit and radial-loop
that may arise due to transfer function uncertain- feedback reference targets. In machines running
ties. However, due to intrinsic limitations such as with transverse feedback systems, the tune can
the bandwidth of the magnetic circuits and non- similarly be derived from its actuator signal while
linear effects such as delays and rate-limits, the keeping beam oscillations and potential instabili-
stabilisation is typically not instantaneous. To ties under control.
cope with these effects requires a more complex Because of the various inter-loop depen-
form of controller to optimise temporal conver- dencies, it is beneficial to implement the tune,
gence (see e.g. [5, 6]). chromaticity, coupling, orbit and radial-loop
364
feedbacks in one global controller to minimise 4X4 full turn map T . We suppose that there is a
data exchange and synchronisation requirements. single rf cavity with matrix

I 0 1 0
Crf = , where A =
References 0A − ωc VE 1
[1] M. Minty et al, PAC 11, 1394 Then

[2] R. Steinhagen, CAS on Beam Diagnostics, Dour- X Xz
dan, France (2008) 317 T = RCrf where R = (1)
Zx Z
[3] M. Gasior, R. Jones, DIPAC 2005, 212
[4] R. Jones, P. Cameron, Y. Luo, BNL-C-A/AP/204 Write

[5] D. C. Youla et al, IEEE Trans. on Automatic Con- cos θx − α sin θx βx sin θx
X=
trol, Vol.21-1 (1976) 3-13 & 319-338 −γx sin θx cos θa x + α sin θx
[6] G. Goodwin, S. Graebe, M. Salgado,Control Sys- (2)
tem Design, Prentice Hall (2000) and
1 Lαp
Z=
0 1
4.7.8 Measurement of Dispersion by Since δ(η ) is the closed orbit for energy offset δ
Resonant Excitation we have that
D.L.Rubin, Cornell U. ⎛ ⎞ ⎛ ⎞
η η

4.7.8.1 Introduction ⎜η ⎟ ⎜ η ⎟
R⎝ ⎠ = ⎝ ⎠
l l
Dispersion is the dependence of the closed or- 1 1
bit on the beam energy. In a storage ring with
horizontal but no vertical bends, the horizontal η l η
→X + X =
dispersion characterizes the optics. Absent mis- η z
1 η
alignments and transverse coupling the first order We know that the transverse position is indepen-
vertical dispersion is zero. Residual vertical dis- dent of l as long as there is no RF. Therefore we
persion results from vertical kicks due to offset can write that
quadrupoles and tilted dipoles, and from coupling
0 η
of horizontal dispersion via tilted quadrupoles and Xz = (I − X)
0η
offset sextupoles. Vertical dispersion is a princi-
ple source of vertical emittance. Measurement of 0 η(1 − cos θx + α sin θx ) − η βx sin θx
=
vertical dispersion is essential to identifying and 0 γη sin θx + η (1 − cos θx − α sin θx )
correcting its sources. (3)
The traditional technique for determining the and using the symplecticity of R we find that
dispersion is to measure the difference in closed
orbits of beams with different energies. The en- η (1 − cos θx + α sin θx ) − ηγ sin θx
Zx =
ergy is changed by adjusting the frequency of 0
⎞
the rf cavities. Alternatively, we recognize that η(cos θx + α sin θx − 1) − η βx sin θx
dispersion represents the coupling of longitudi- ⎠
nal and transverse motion. This allows us to 0
exploit the techniques developed for measuring
horizontal-vertical coupling, in particular reso- Then if
Mm
nant excitation of the normal mode frequencies T =
n N
and then measurement of the relative phase and
amplitude of the vertical and horizontal response we find that

at each beam position monitor [1]. −m12 ωV
cE 2η(1 − cos θx )
H = m + n† =
cE 2η (1 − cos θx )
−m22 ωV
4.7.8.2 Formalism (4)
The coupling matrix defined as
We begin with an illustration of horizontal dis-
persion as coupling of longitudinal and horizontal −Hsgn(tr[M − N ])
C= (5)
motion. The linear motion is characterized by a γ (tr[M − N ])2 + 4|H|
365
where γ 2 = 1 − |C|, so that where

−1 Ga 0
T = V UV (6) U = G−1 Ū G, G = (9)
0 Gb
with
and
A 0 γI C R(θa ) 0
U= and V = Ū = , (10)
0 B −C † γI 0 R(θb )

(A and B are 2x2 matrices. I is the 2x2 identity). cos θ sin θ
R(θ) = , (11)
The determinant of H as given in Eq.(4) is − sin θ cos θ
2ωV and # $
|H| = − (1 − cos θx )(m12 η − m22 η) √1
0
cE Gi = βi √ (12)
(2πQz )2 − √αβi βi
= −2 (1 − cos θx )(m12 η − m22 η) . i
Lαp We imagine that the a-mode is horizon-

tal(vertical) motion, and the b-mode is syn-

L is the circumference, and αp the momentum
chrotron motion. The beam is resonantly excited
compaction. In the weak coupling limit where
at the synchrotron tune Qs . Then
4|H| (tr[M − N ])2 , that is far from the cou-
⎛ ⎞ ⎛ ⎞
pling resonance, γ ∼ 1 and wa 0

⎜ a⎟
w ⎜0⎟
H ⎝ w ⎠ = w0 ⎝ 1 ⎠ (13)
C∼ b
tr[M − N ] wb 0
H
∼ where we have chosen without loss of generality
2(cos θx − cos θz )
to specify w at a time when wb = 0. Then at any
1 later time and at location i,
∼
2(cos θx − cos θz ) ⎛ ⎞ ⎛ ⎞
wa 0

⎜ wa ⎟ ⎜ 0 ⎟
⎝ w ⎠ = w0 ⎝ cos(θ − φ ) ⎠ (14)
cE 2η(1 − cos θx )
−m12 ωV b iz
sin(θ − φiz )
−m22 cE 2η (1 − cos θx )
ωV wb

0 η where θ = ωs t, and ωs is the synchrotron tune.
∼ (7) From a measurement of the time dependence of
0η
the position signal at the ith beam position moni-
where we have assumed that cos θz ∼ 1, that tor we extract the transverse amplitude and phase
is synchrotron tune near zero. We see that where
the dispersion can be identified with elements
of the coupling matrix. A measurement of the xi = Aix cos(θ − φix ) (15)
longitudinal-horizontal and longitudinal-vertical (or x → y). We can similarly write the longi-
coupling yields the dispersion. tudinal displacement in terms of the longitudinal
amplitude and phase,
4.7.8.3 Measurement of the coupling matrix
zi = Aiz cos(θ − φiz ) (16)
We need to relate the coupling matrix elements where φi(x/y) is the horizontal (vertical) normal
to the quantity that we can measure, namely the mode betatron phase advance and φiz is the lon-
vertical and horizontal amplitude and phase of the gitudinal phase advance at θ = 2πnQs . From
signal modulated at the synchrotron tune, at each Equation 8 we get that
bpm. It is convenient to use normalized phase
space coordinates. We remember that the phase Gx = GV G−1 w
⎛ ⎞
space 4-vector x is related to the normalized, nor- 0
mal mode representation w according to γ C̄ ⎜ 0 ⎟
Gx = ⎝ cos(θ − φ ) ⎠
−C̄ † γ iz
x = V G−1 Ū w
(8) sin(θ − φiz )
366
where C̄ = Ga CG−1
b and then 4. Use the measured horizontal and computed
⎛ 1 ⎞ longitudinal phase and amplitude to deter-
√ Aix cos(θ + φix )
βx mine a quantity proportional to the horizon-
⎜ ∼ ⎟
⎜ 1 ⎟ tal dispersion and fit to the model dispersion
⎝ √ Aiz cos(θ + φiz ) ⎠ to determine the amplitude of energy oscilla-
βz
∼ tions
⎛ ⎞ 5. From the measured vertical phase and ampli-
C̄11 cos(θ − φiz ) + C̄12 sin(θ − φiz )
⎜ ∼ ⎟ tude, the computed longitudinal phase, and
=⎝ ⎠ the fitted amplitude of energy oscillations,
γ cos(θ − φiz )
γ sin(θ − φiz ) determine the vertical dispersion.
With some rearrangement we find that In the CesrTA at the Cornell Electron Storage
Ring, the technique yields a measurement of the
βiz Aix vertical dispersion with few millimeter resolution
sin(φix − φiz )
C̄12 = in a few seconds.

βix Aiz

βiz Aix References
C̄11 = cos(φix − φiz )
βix Aiz
[1] D. Sagan et al, PRST-AB 3, 092801 (2000)
√
Finally, C12 = βa βb C̄12 and then according to
Equation 7
4.8 TEMPORAL AND SPATIAL
Ai(x/y) CORRELATIONS IN BPM
ηx/y = C12 = βiz sin(φi(x/y) − φiz )
Aiz MEASUREMENTS
Ai(x/y)
= βiz sin(φi(x/y) − φiz ) One may easily and rapidly take buffered BPM
az
√ data pulse by pulse in linear or turn by turn in cir-
where Aiz = az βi z. Since there is no good cular accelerators. These BPM readings will be
way to measure the longitudinal phase and am- correlated both temporally and spatially. Among
plitude of the beam at each of the BPMs, we de- many advantages, the derived information follows
pend on the model calculation. Since the longi- a statistical square root N rule in enhancing the
tudinal focusing is typically very weak, the lon- measurement resolution, where N can be the num-
gitudinal phase advances very slowly and any er- ber of pulses or turns and/or the number of BPMs.
rors that might arise due to discrepancy between Depending on the accelerator and the physical
model and measurement tend to be very small. quantities to be measured, taking these BPM read-
The longitudinal amplitude is not measured. ings can often be non-invasive.
In practice we determine the longitudinal ampli-
tude (the amplitude of the energy oscillation) by 4.8.1 Non-Invasive Measurement for Linacs
fitting the measured C12 (horizontal dispersion J. Irwin, Y. Yan, SLAC
data) to the model dispersion. The fitted ampli-
tude is used to determine vertical dispersion. The pulse-by-pulse incoming jitter can offer ex-
In summary, to measure the dispersion by res- tractable correlated signals through use of a
onant excitation singular-value decomposition (SVD). Using M
BPMs together has the immediate advantage
1. Drive the beam at the synchrotron tune of improving the resolution on the motion by
√
M [1]. The number of independent eigenvec-
2. Measure the amplitude and phase of the hori- tors with eigenvalues above the noise floor de-
zontal and vertical components of the motion termines the number of physical variables that
at each beam position monitor. are changing and affecting beam centroid motion.
The linear space of the eigenvectors provides an
3. Compute the longitudinal beta and phase at orthogonal coordinate system on the linear space
each BPM from the machine model. spanned by the underlying changes, and the time
367
Sec.4.8: TEMPORAL AND SPATIAL CORRELATIONS IN BPM MEASUREMENTS
patterns of the occurrence of these eigenvectors large, the order of 104 , if the control system is ca-
provide an orthogonal coordinate system on the pable of buffering pulse-by-pulse BPM data.
linear space of the time patterns of the physical If a set of time patterns are known, and it is
changes. See Sec.4.8.2 for application to storage known or suspected that this set is not correlated
rings. to variables outside the set, then the shape patterns
Let bmp represent the reading of the mth BPM corresponding to all these variables can be found
(m = 1 to M ) on the pth pulse (p = 1 to P ). The from
vector b(tp ) = (b1p , b2p , ..., bM
p ) ≡ bp will represent At σ
FS = C−1S QS B + O( √ )
the complete set of readings on the pth pulse, and P
B = [bm p ] will be the P × M matrix of BPM
where CS = Q At Q
A S is the correlation matrix
readings with bp as the pth row. The expectation S
is that of the variables in S . If there was an unknown

variable correlated to one or more of the time pat-

S
S
terns in S , the shape pattern corresponding to
bp = b0 + Δqps ∂sb + Δqpr Δqps ∂r ∂sb
that variable will be mixed into the above calcula-

s=1 r,s=1
tion such that each deduced fs may contain a bit
+ · · · + np of that pattern. Even in this case it is usually a
good strategy to use the known time patterns in
Here b0 is a constant arising from centroid or this way and later attempt to identify and remove
BPM offsets, Δqps are the values of the sth vari- unknown patterns. Additional time patterns can
able on the pth pulse, ∂sb are the first order deriva- be put into the Q A matrix without negative conse-
tives of the beam centroid due to the change in quence. If the patterns are not present, the corre-
the sth variable, and ∂r ∂sb are the second order sponding shapes √ will be zero, modulo a vector of
changes. np is random noise in each BPM on the magnitude σ/ P . Errors which arise if there are
pth pulse. Anything that is not random will oc- random errors in the time patterns are second or-
cur in the sum of variables. It is useful to divide der in the magnitude of the error, i.e. 10% random
each variable change Δqps by its rms value to get error in the time pattern gives a 1% mixing of the
“normalized” change patterns Δq Hs , and multiply space patterns.
p

the derivatives ∂s b by the same rms change to get Of course some variables may be totally un-
rms shape functions, fs ≡ Δqrms s ∂ known, and some variables may be known as im-
s b. The same is
portant, but not measured. The motions from
done for the product variables. The fs all have the
these variables may be found by analyzing the B
dimension of length and the normalized change
matrix directly. What is known should first be re-
patterns are dimensionless. The normalized time moved, then one performs an SVD of the remain-
patterns can be assembled
as a [P × S(S + 1)/2] der. The SVD yields a product decomposition
A H s
variable matrix Q = Δqp , ..., Δq r s
B = UΛVt . Here the matrix U will be a P × M
p Δqp , ... . And
the space pattern functions fs can be put into an matrix of normalized (and orthogonal) time pat-
S(S + 1)/2 × M matrix F. With this notation we terns, V will be an M × M matrix of normal-
A ized orthogonal (eigenvector) space patterns, and
may write our expectation for B as B = QF+N,
Λ are eigenvalues of the space patterns. Multiply-
where N is the collection of noise vectors. If
ing the normalized space patterns by the eigenval-
any variable shapes are known to be of negligi-
A and ues (which are the rms amplitudes of these mo-
ble magnitude, they can be dropped from Q tions) the product ΛVt can be interpreted as the
F. rms space patterns. So we end up with a decom-
If one of the time patterns Δq Hps is known (a
position B = U[ΛVt ] resembling the expecta-
suspected 60 Hz signal, or an important beam A + N. The linear space of F vectors
tion B = QF √
parameter has been measured) and furthermore
with rms amplitude greater than σ/ M will be
known or suspected to be uncorrelated to the other
spanned by the V vectors and the linear space of
time patterns, the corresponding shape pattern can A vectors will be spanned by the
be found by multiplying the measurements matrix corresponding Q
t √ U vectors. A variety of techniques must now be
B by Q Is t = Δq Hps . The error is only σ/ P
employed to identify the vector direction of the
where σ is the rms BPM noise. P can be very physical motions.
368
Horizontal Eigenvalue
Degree of freedoms with increasing number of BPMs
0.015 200
(a)
180
0.010
160
Singular Values in um (unnormalized)

0.005
140
0 120
0.02 100
Vertical Eigenvalue
(b) 80
0.01 60
40
20
0
2–98
0 50 100 150
0
8355A209 Eigen Mode Index 0 20 40 60 80 100 120
BPM #
Figure 1: Typical graph of eigenvalues for several di-

Figure 3: Degree of freedom measured at the SLAC
mensions of motion plus BPM noise.
linac (with a corrector dithered around BPM 60).
Horizontal Eigenvector Vertical Eigenvector

0.5 0.5 In this way we have also identified the number
0 0 of degrees of freedom, i.e. the number of underly-
–0.5 –0.5
0.5 0.5 ing physical variables that are changing and giv-
0 0 ing rise to changes in the beam √ centroid greater
–0.5 –0.5
1 0.5
than or the order of about σ/ M . In this first
0 0 step one should glance at the eigenvectors of the
–1 –0.5 retained motion. See Fig.2. If there is a spike in
1 1
0 0
these vectors, it usually indicates a bad BPM at
–1 –1 this location. In such a case, the data from the bad
0.5 0.5 BPM should be dropped from the analysis, and
0 0
– 0.5 –0.5
this step repeated.
0.5 1 A second step often consists of identifying
0 0
– 0.5 –1
initial betatron amplitudes. This can be done by
5–98
0 50 100 150 0 50 100 150 using the noise-reduced data set and restricting
BPM Index BPM Index
8355A207
attention to the first few BPMs. Computing an
Figure 2: The leading 6 eigenvectors measured at the SVD of this restricted data set will typically reveal
SLAC linac. only two major eigenvectors. They can be identi-
fied as the betatron vectors and the corresponding
time patterns will be the sought betatron ampli-
The first step consists of identifying the noise, tude time patterns.
and subtracting out the noise floor. A typical The third step consists of forming an SVD for
eigenvalue curve is shown in Fig.1. The noise the data restricted to the first m BPMs for ev-
will be flat and for the most part, linear. As P ery m between 1 and M . A plot of detectable
increases the slope of the noise floor gets smaller eigenvalues as a function of m , called a “de-
reaching an asymptote that corresponds to the dis- gree of freedom plot” and shown in Fig.3, re-
tribution of the resolutions of the BPMs. In this veals where in the beamline each new degree of
way one has a direct measure of the BPM per- freedom arises. By studying the SVD factoriza-
formance. The average value of the eigenvalues tion in the region where new degrees of freedom
of the noise√floor is the average BPM resolution arise, one can identify time patterns associated
divided by M . By setting the eigenvalues of with this motion. There are certain limitations
what has been identified as the
noise floor to zero, that arise here. The time pattern will necessarily
and re-multiplying U ΛVt one obtains a data be perpendicular to previously determined time
set with noise removed, getting a step closer to patterns, and can differ from a physical time pat-
A
QF. tern in that regard, and the space pattern will be
369
/home/yan/mia/ler/2004/jan13/
unique at its origin in the beamline, but may be- 2 1
come mixed downstream with motion originating
1
Orbit X
Orbit Y
0 0
downstream.
−2 −1
Ideally one would make a concerted effort to 2
0 100 200 300 400
0.5
0 100 200 300 400
measure all physical variables that could influence
2
0
Orbit X
Orbit Y
0
input conditions and have diagnostics to measure −0.5
relevant beam parameters on each pulse. Then, −2

0 100 200 300 400
−1
0 100 200 300 400
2 5
together with the betatron amplitudes, they would
3
Orbit X
Orbit Y
form a complete set of input conditions. In addi- 0 0
tion, variables can be dithered as the data is taken, −2

0 100 200 300 400
−5
0 100 200 300 400
and the corresponding vectors easily isolated and 5 5
4
identified. Combining these techniques, one can
Orbit X
Orbit Y
0 0
hope to limit the unknown shapes to those arising −5 −5

0 100 200 300 400 0 100 200 300 400
from variables in the beamline itself. By their na- BPM sequence number BPM sequence number
ture these are often well localized and identifiable

by the methods of the preceding paragraph. Figure 1: Four independent orbits extracted from PEP-
II LER BPM buffer data. The first two orbits (x1 , y1 )
The betatron vector, which may be deter-
and (x2 , y2 ) are extracted from beam orbit excitation at
mined very accurately from these procedures, can
the horizontal tune while the other two orbits (x3 , y3 )
be used to study the betatron emittance pattern in and (x4 , y4 ) are from excitation at the vertical tune.
the linac [which is intimately related to wake ef-
fects, BNS profiles, etc. (Sec.4.3)]. The betatron
vector also can be used to differentiate shape pat- location a to location b. A complete set of data
terns to find “kick” representations of these vec- will allow for the extraction of the 4 independent
tors. The kick patterns often facilitate identifica- orbits.
tion and localization of sources. Unlike linacs where there is often enough in-
coming jitter in the beam to measure and identify
References betatron modes, in the e± rings, to offset radiation
damping, the most economic process for data ac-
[1] J. Irwin, C.X. Wang, Y.T. Yan et al, PRL 82 (1999) quisition is through two orthogonal resonance ex-
1684 citations, one at the horizontal betatron tune and
the other at the vertical tune, each time buffering
BPM data. Since a betatron motion has two de-
4.8.2 Invasive Measurement for e+ , e− grees of freedom, each excitation would generate
Circular Accelerators a pair of conjugate (cosine- and sine-like) beta-
J. Irwin, Y. Yan, SLAC tron motion orbits from a tune-matched FFT after
An accelerator is built based on an ideal lattice removing non-physical BPM data [3]. Therefore,
design. However, its optics will not be exactly a complete set of 4 independent linear orbits in X
the same as the ideal lattice. On the other hand, and Y can be extracted from the two eigen-mode
one may be able to construct a virtual model with excitations. A typical set of such orbits for PEP-
very much the same linear optics as the accelera- II LER is shown in Fig.1. Note that if the stor-
tor, particularly an e+ or e− storage ring, through age ring is not strongly coupled, taking the sum
betatron-motion orbit measurement [1] and SVD- of the squares of the cosine- and sine-like orbits
enhanced Least-square fitting [2]. from the two excitations approximately gives the
horizontal and vertical betatron amplitudes in the
A complete set of data acquisition for geomet- BPM measurement space. Note that one can also
ric optics measurement The linear geometric perform SVD of the buffered BPM data. How-
optics can be well determined if one gets 4 in- ever, the tune-matched FFT yields more accurate
dependent linear orbits and knows the acceler- extraction of the 4 independent linear orbits in the
ator structure. These can be obtainable from case of tune-resonance excitation.
the linear mapping, Z b = Rab Z a , written as
Rab = Z b Z a−1 , where the 4 × 4 matrix, Z a = Fitting constraints Since the storage ring is
[z1a , z2a , z3a , z4a ], represents 4 independent linear based on an ideal lattice, that ideal lattice can
orbits at location a, and Rab is the linear map from serve as the initial guess for setting its linear
370
magnetic components. While it is clear that all Coupling ellipses For each double-view BPM,
quadrupole strengths and sextupole feed-downs as one can trace the extracted high-resolution real-
well as the BPM linear gains and cross-couplings space orbits to obtain a coupling ellipse in
should be used as a complete set of variables, one real space for each resonance (eigen) excitation.
can be flexible in choosing fitting constraints. The Therefore, one can calculate coupling-ellipse tilt-
key to successful fitting is dealing with degenera- angles and axis-ratios for all double-view BPMs
cies which are well handled by an SVD-enhanced in each of the two eigen planes [9]. For a very
Least-square fitting [2]. small βy location such as the interaction point
Convenient constraints, derivable from the 4 (IP) in a collider, the tilt angle of the coupling el-
independent orbits, include phase advances [7], lipse for the horizontal eigen plane is very close
Green’s functions between any two BPMs [1] and to the real tilt angle of the beam. One can also
for the same BPMs between any two turns [8], and calculate the corresponding coupling parameters
eigen-plane coupling ellipse tilt-angles and axis- from the linear map of a lattice model. Therefore,
ratios [9]. These constraints are not necessarily these quantities can also be used as fitting parame-
independent, but for better convergence, one may ters although they can be replaced with additional
include many such constraints. On the other hand, Green functions.

one may omit the eigen-plane coupling-ellipse Dispersion measurement and fitting The lin-
tilt-angle and axis-ratio constraints in order to ear magnet component fitting should be solely de-
check that the fitted result is self-consistent [8]. termined by geometric optics. That is, the disper-
Phase advances One can derive the orbit beta- sion is not part of the completeness of the con-
tron phase at each BPM location by simply tak- straints that determine the quads, the skews, and
ing the arctangent of the ratio of the imaginary the sextupole feed-downs. However, if the accel-
part to the real part of the resonance excitation erator orbit is well confined, the geometric-optics-
FFT mode [7]. Phase advances between adjacent fitted model dispersion should be close to the
BPMs can then be calculated by subtraction. Note measurement which can be obtained from turn-
that the ratio of the imaginary part to the real part by-turn BPM buffer data by exciting synchrotron
of the FFT will cancel in the linear BPM gains oscillations. If one wishes, the same variables
but not in the BPM cross-couplings. Therefore can be further fitted to include the dispersion con-
the phase advances among BPMs are repeatedly straints (with a moderate weighting) without hurt-
calculated during the Least Square fitting process. ing the geometric-optics fitting.
Linear Green’s functions The linear Green’s
function are simply the R12 ab , Rab , Rab , Rab of the Virtual Accelerator for Real Accelerator optics
34 14 32
linear transfer matrix between any two BPMs la- improvement The well-fit lattice model can be
belled as a and b. They are given in the data mea- viewed as a virtual accelerator matching the op-
surement space as [1] tics of the real accelerator [8]. As an example,
(xa1 xb2 −xa2 xb1 )/Q12 +(xa3 xb4 −xa4 xb3 )/Q34 = Rab shown in Fig.2 are optical characteristics plots for
12
a well-fit PEP-II LER. Since one has the virtual
(xa1 y2b −xa2 y1b )/Q12 +(xa3 y4b −xa4 y3b )/Q34 = Rab
32 accelerator in a computer, it is convenient to test
(y1a xb2 −y2a xb1 )/Q12 +(y3a xb4 −y4a xb3 )/Q34 = Rab
14 for solutions that involve changing a limited num-
(y1a y2b −y2a y1b )/Q12 +(y3a y4b −y4a y3b )/Q34 = Rab ber of key normal and skew quadrupoles, and/or
34
where Q12 and Q34 are the two invariants relat- orbit correctors (for producing sextupole bumps)
ing to the two resonance excitation amplitude; to improve the virtual accelerator linear optics.
(x1 , y1 ), (x2 , y2 ), (x3 , y3 ), (x4 , y4 ) are the 4 in- One could then apply the changes to the real ac-
dependent linear orbits; and R12 , R32 , R14 , R34 celerator. These processes have been successfully
are given as applied to PEP-II LER and HER to reduce cou-
pling effects, beta beating, and bring the working
Rab b ab a b ab a b ab a b ab a
12 = gx R12 gx +gx R14 θxy +θxy R32 gx +θxy R34 θxy
tune to near half integers for luminosity enhance-
32 = gy R32 gx +gy R34 θxy +θyx R12 gx +θyx R14 θxy ment [10].
14 = gx R14 gy +gx R12 θyx +θxy R34 gy +θxy R32 θyx
34 = gy R34 gy +gy R32 θyx +θyx R14 gy +θyx R12 θyx
References
where gx ’s, gy ’s are the BPM gains, and θxy ’s and [1] J. Irwin, Y.T. Yan, EPAC 00, 151
θyx ’s are the BPM cross-coupling multipliers [1]. [2] Y.T. Yan, Y. Cai, NIM A558 (2006) 336
371
red: virtual, blue: design, green: measured (38.5511, 36.5887) /home/yan/mia/2005/nov22/ler

max(β ) = 150.4264 meter max(β ) = 251.4695 meter
β (m)
β (m)
1 2
200
100 max(βx) = 130.9422 meter max(βy) = 141.3741 meter
1
2
0 0
−50 0 50 −50 0 50
150
β (m)
100 *
β1 = 42.1553 cm
1
50
0
−1000 −500 0 500 1000
150
β (m)
100 *
β2 = 11.0665 mm
2
50
0
−1000 −500 0 500 1000

200 θ* = −0.28126 degree 50 θ* = −77.2096 degree
β (mm)
β (cm)
1 * 2
(b/a)1 = 0.034207 *
(b/a)2 = 0.73452
100 *
βx = 41.5908 cm β* = 11.0036 mm
x
y
* y *
Smin = −1.4429 cm, αx = −0.034694 Smin = 0.12038 mm, αy = 0.010942
0 (cm)
0
−50 0 50 −20 −10 0 10 20 (mm)
0.4
1
sin(Φ)
sin(Φ)
sin(φ*) = 0.22412
0.5 0.2
0 0
−1000 −500 0 500 1000 −1 −0.5 0 0.5 1
1
dispx(m)
0
−1
−2
−1000 −500 0 500 1000
0.1
tilt angle 1 (deg) dispy(m)
0
−0.1
tilt angle 2 (deg)
−1000 −500 0 500 1000

100 100
avg(Δ θ1)=0.63484 avg(Δ θ )=0.54261
2
0 0
−100 −100
−1000 −500 0 500 1000 −1000 −500 0 500 1000
1 1
avg(Δ (b/a)1)=0.0087507 avg(Δ (b/a)2)=0.010531
(b/a)1
2
(b/a)
0.5 0.5
0 0
−1000 −500 0 500 1000 −1000 −500 0 500 1000
S (meter) S (meter)
xrayEllipseTiltAngle = 9.316 synchroEllipseTiltAngle = 4.9096
Figure 2: Plots to show a virtual accelerator linear optics characteristics (red) compared with those of the designed
lattice (blue) and/or measurement (green). The top two plots show the two eigen beta functions in the vicinity of
IP followed by two plots that show the β-functions around the ring and then the β-function plots at IP, which show
the β ∗ ’s, the waists, the coupling ellipse tilt angles (θ∗ ’s), and axis ratios ((b/a)∗ ’s). The next two plots show the
phase-space coupling angles in the generating matrix for decoupling normalization, followed by horizontal and
vertical dispersion plots. The bottom 4 plots show the coupling eigen-plane ellipse tilt angles and axis ratios at all
double-view BPMs.
372
[3] tune-matched FFT means varying the number of of the transfer matrices which do not cross any
turns included in the FFT so that the principle of the oscillating ac dipoles. One may use two
FFT mode is sharp. Alternatively, one may use pairs of ac dipoles located as far as possible from
a time-correlation process to identify the peak each other. Turn on the first pair of ac dipoles.
modes [4]. One may also perform an SVD to Let one dipole of this pair oscillate at near the
extract dominant modes for Lattice parameter horizontal tune while the other at near the verti-
mesurements [5] [6]. cal tune. Then take turn by turn beam orbit data
[4] X. Huang et al, PRST-AB 8 (2005) 064001 with BPMs. Performing two focused FFTs [3]
[5] C.X. Wang, V. Sajaev, C-Y. Yao PRST-AB 6, of the orbit data one for the mode near the hor-
104001 (2003) izontal tune and the other for the mode near the
[6] R. Calaga et al, EPAC 04, 1541 vertical tune, one would obtain two pairs of con-
[7] Y.T. Yan et al, SLAC-PUB-9368 (2002)
jugate (cosine-like and sine-like) orbits for cal-

[8] Y.T. Yan et al, PAC 05, 3426
culating local linear Green’s functions between
[9] Y.T. Yan, Y. Cai, PAC 05, 3459
[10] Y.T. Yan et al, SLAC-PUB-10369 (2003)
BPMs (Sec.4.8.2). Discarding any Green’s func-
tion that crosses either one ac dipoles or both of
them, one would get the first set of local Green’s

4.8.3 Virtual Models for Proton Circular functions, none of which are modified by the in-
Accelerators serted ac dipoles. However, this reduced set of
Y. Yan, SLAC Green’s functions is not complete because there
is no Green’s function in the vicinity of the two
As described in Section 4.8.2 for invasive mea- active ac dipoles. To make up for the discarded
surement of an electron circular accelerator, one local Green’s functions in the area near the first
can construct a virtual model with very much the pair of ac dipoles, one turns off the first pair and
same linear optics as the real machine. Such a turns on the second pair of ac dipoles. In the same
virtual model may be obtained for a proton circu- manner one would be able to get another reduced
lar accelerator, too. However, it requires a more set of local Green’s functions, none of which are
elaborate procedure. modified by the inserted ac dipoles. Combining
Unlike an electron circular accelerator where the two reduced sets of local Green’s functions,
one can resonantly excite clean betatron motion one has a complete set of Green’s functions that
because of synchrotron radiation damping, in a covers the entire circular accelerator in its original
proton circular accelerator such as the LHC, one configuration without active ac dipole. This com-
may only excite modified betatron motion by in- bined set of Green’s functions provides a com-
serting an ac dipole [1] oscillating at a frequency plete set of fitting constraints for deriving a vir-
close but not exactly equal to the horizontal or the tual model by following the fitting procedures of
vertical betatron tune. The modification is due to Section 4.8.2
the ac dipole itself which modifies the circular ac-
celerator as if there is an insertion of a symplectic
linear map or simply a quad in one dimensional References
case [2]. While measuring the modified betatron [1] M. Bai et al, PR E56 (1997) 6002; S. Peggs,
motion is straightforward as one can follow a pro- C. Tang, BNL Report No. RHIC/AP/159 (1998)
cedure similar to the one described in Sec.4.8.2, [2] R. Miyamoto et al, PRST-AB 11, 084002 (2008);
accurately subtracting the inserted linear map due For nonlinear effects, one may refer to R. Tomas,
to the ac dipole for establishing a virtual model PRST-AB 5, 054001 (2002)
can be tedious. However, taking advantage of the [3] Reference [3] in Sec.4.8.2
Green’s functions from Sec. 4.8.2, one may con-
sider using ac dipoles for beam orbit excitation
but avoiding direct measurement of the modified
optics quantities 4.9 TRANSITION CROSSING
To avoid measuring optics quantities modi- J. Wei, Michigan State U.
fied by the ac dipole insertion, one may not di-
rectly measure the globally affected quantities During acceleration in a synchrotron, the longi-
such as beta functions, phase advances but the lo- tudinal particle motion is non-adiabatic within a
cal linear Green’s functions, R12, R14, R32, R34 characteristic time ±Tc near transition energy γT
373
Sec.4.9: TRANSITION CROSSING
[1, 2], where the total nonlinear time ±Tnl is given [5, 9,
13 10] by
πEs βs2 γT3 ( (√
Tc = (1) ( 3βs2 (( 6σ̂δ γT
qeV | cos φs |γ̇hωs2 (
Tnl = ((α1 + )
2 ( γ̇
where the subscript s denotes the synchronous
value. Upon crossing, the synchronous phase (φs ) This effect was experimentally observed, [9, 11,
typically needs to be shifted by π − 2φs in a 12] and α1 was obtained by measuring the syn-
time much shorter than Tc . Define the momen- chrotron frequency or minimum-loss timing as a
tum compaction factors αi and the phase slippage function of the beam radial position. Reducing
factors ηi , i = 0, 1, 2, · · · , according to Sec.2.3.11 the chromatic nonlinearity using sextupole fami-
[3, 4, 5]. lies was proposed [7] and demonstrated. [9, 12]
Linear equations of motion With a normalized Umstätter effects Transverse space-charge

time dτ = kdt, k = qeV | cos φs |/2πh, the lon- force changes the tune of each individual particle,
gitudinal motion is described by a Hamiltonian making γT dependent on the azimuthal beam
[5] H(ϕ, J; τ ) = ±J/βL . The action-angle vari- density. The amount of subsequent mismatch
ables (ϕ, J) are related to the rf phase φ and is inversely proportional to βs γs2 , and usually
W ≡ −ΔE/hωs by negligible if γT is much higher than the injection
energy.
Δφ = ∓ 2J/βL (sin ϕ + αL cos ϕ) Multi-particle mismatch Emittance growth

W = − 2JβL cos ϕ due to bunch mismatch [13] under a reactive im-
pedance Z at the bunch frequency is proportional
where the upper (or lower) sign is for below (or to the ratio of the beam-induced force to the accel-
above) γT , αL = −βL /2, and denotes the erating force,
derivative with respect to τ . The amplitude func- ( (
tion βL is given by ΔS hIˆ (Z /n(
≈ (2)
1 1 S 3V | cos φs | σ̂φ2
βL βL − βL2 + KβL2 = 1
2 4
where Iˆ is the peak current at γT . Eq. 2 is valid
−2πh3 ωs2 η0 exactly for a parabolic distribution [14] under the
K=
qeV cos φs Es βs2 space charge force.
For a constant γ̇ near transition, A longitudinal resistive impedance R at
π bunch frequency causes energy dissipation,
βL
= x J2− 1 (y) + N2− 1 (y) ≈ 1.58 − 1.15x shifting the synchronous phase by Δφs ≈
kTc 3 3 3 ˆ
IR/V | cos φs | while producing a growth
where y = 2x3/2 /3, x = |Δt|/Tc , and Δt is the ˆ
ΔS IR
time delay [5]- [7] from γT . The synchrotron fre- ≈√
quency is Ωs = kβL−1 . The maximum excursions S 6V | cos φs | σ̂φ
√ √
in φ and W are φ̂ = 2γL J and Ŵ = 2βL J, The change in φs at transition can cause se-
2
where 1 + αL = βL γL . For a bunch of rms bunch vere beam loading stress [15, 16] while the rf cav-
area S = 2πJ, the rms phase and momentum ity tuning system changes the sign of the reactive
deviations at γT are σ̂φ = 0.52 (S/kTc )1/2 and beam loading compensation.
σ̂δ = 0.71hωs (kTc S)1/2 /Es βs2 . Instabilities A capacitive (or inductive) longi-
Single-particle effects Single-particle effects tudinal coupling impedance Z at a broad-band
include mismatching to the accelerating rf bucket, frequency will cause a microwave instability [14,
coupling to transverse motion, [8] and various 17, 18] during a time Tmw ≈ 1.37 (D − 1) Tc
kinds of mis-timing in a time comparable to Tc . after (or before) transition if
( (
Emittance growth due to chromatic nonlinear- 4hIˆ (Z /n(
ities (Johnsen effect [3]) is [5] D ≈ ≥1 (3)
⎧ 9V | cos φs | σ̂φ2
⎪ Tnl
ΔS ⎨ 0.76 T , for Tnl Tc ; Eq. 3 is valid exactly for a parabolic distribution
≈ c3/2 under negative-mass instability above γT . A resis-
S ⎪
⎩ 43 TTnlc
e − 1, for Tnl ≥ Tc , tive longitudinal impedance may cause instability
374
both below and above γT . Microwave instabil- Other compensation methods Other methods
ity near γT has been experimentally observed [19] attempted or proposed include (a) minimizing the
and simulated. impedance at γT by adding reactive loading [30]
The transverse microwave instability [20] (also see Sec.4.15.1), (b) rf system feedback, (c)
threshold at γT is avoiding phase jump [7] by continuously vary-
1.5σ̂δ βZ ing V and φs , (d) rf manipulation [7] to eliminate
D⊥ ≈ D ≥ 1
b bunch-length oscillation, (e) artificial blow-up [7]
where βZ is the average β function at the impe- of the longitudinal emittance, (f) reducing rf volt-
dance location, and b is the beam pipe radius. age [5] to alleviate chromatic effects, (g) tem-
When the beam stays near γT for a relatively porarily changing the orbit circumference using
long time, longitudinal head-tail [21] and other programmed V and φs [5], (h) using a flattened rf
slow-growing instabilities may also occur. wave [31] to reduce σ̂δ and Iˆ and to provide equal
Simulations Macro-particle method [14, 17, acceleration for all the particles near γT .
22] has been used to construct beam-induced Methods to avoid transition include (a) rais-
forces in both the space and frequency domain.
ing injection energy (b) reducing γT along with

For a given numerical accuracy, the number of transverse tunes (c) creating a large or imaginary
macro particles needed to simulate a reactive (or γT by either using negative bends [32] or by in-
resistive) coupling is proportional to the cubic (or troducing negative dispersion in bending regions
linear) power of the highest frequency considered using either the harmonic approaches (e.g. en-
[23]. hancing higher-order Fourier component of the
Transition jump A γT -jump [13, 24] has been momentum compaction choosing the horizontal
demonstrated on many machines to improve tune just below the number of lattice superperiod
crossing efficiency by effectively increasing the [33]) or the modular approaches (e.g. using a π-
crossing rate. Without varying the tunes, [25] straight section to link two FODO cells containing
a sudden change of γT is achieved by pulsing dipoles [34]). Issues include possible poor pack-
quadrupoles, often grouped in π-doublets, at lo- ing factor, tunability, and dynamic aperture, large
cations of high dispersion. In order to min- dispersion and β-function oscillations, and exces-
imize optical distortion and chromatic nonlin- sive number of magnet families [35]. The oper-
earity enhancement, [12] “matched, first-order” ating J-PARC Main Ring’s imaginary-γT lattice
schemes [26] have been adopted for recently pro- uses supercell modules containing three FODO-
posed accelerators incorporating two families of cells of 90◦ phase advance, omitting dipoles in the
quadrupoles at regions of different dispersion. For middle FODO-cell at the high positive-dispersion
a maximum allowable fractional growth of bunch location [36].
area GS ≡ ΔS/S, the minimum size ΔγT and Applications Operating storage rings under a
speed |γ˙T | of γT jump is [5, 28]
quasi-isochronous condition (very small α0 ) has
|γ̇ − γ̇T | 0.76 Tnl 6/5 been proposed to achieve very short bunches for
ΔγT > 2γ̇Tnl , >
γ̇ GS Tc free electron drivers, synchrotron light sources,
to compensate for chromatic nonlinear effect, next generation e+ e− colliders, and muon collid-
# ( ( $4/3
31Es βs2 qeV γT3 I¯ (Z /n( ers. These designs require both an accurate con-
ΔγT > 1/3 trol of α1 to provide the necessary momentum
h | cos φs |1/3 ωs2 S 2 GS V
acceptance (∼ α−1
# ( ( $2 1 ) [37], and effective ways to
|γ̇ − γ̇T | 2hGS Iˆ (Z /n( damp instabilities. Obtaining short bunches by
> extracting near γT has also been proposed for a
γ̇ V | cos φs | φ̂2 proton driver of muon collider and for proton-
for self-field mismatch, [27, 28] and [28, 29]
# ( ( $4/3 driven plasma wakefield acceleration.
46Es βs2 qeV γT3 I¯ (Z /n(
ΔγT > 1/3
h | cos φs |1/3 ωs2 S 2 V References
# ( ( $2
|γ̇ − γ̇T | 8hIˆ (Z /n( [1] N.M. Blachman, E.D. Courant, RSI 20 (1949)
> 596
γ̇ 3V | cos φs | φ̂2 [2] E.D. Courant, H.S. Snyder, Ann.Phys. 3 (1958) 1
for microwave instability, where Iˆ and φ̂ are val- [3] K. Johnsen, CERN Symposium on High-Energy
ues in the absence of jump, I¯= N0 qeωs /2π. Acc., 1 (1956) 106
375
Sec.4.10: RF GYMNASTICS IN A SYNCHROTRON
[4] K. Takayama, PA 14 (1984) 201 [30] R.J. Briggs, V.K. Neil, J. Nuclear Energy, Part C,
[5] J. Wei, Ph.D. dissertation, Stony Brook (1990); 8 (1966) 255; The CERN Study Group on New
rev. 1994 Accelerators, AR/Int. SG/64-15 vol.1 (1964) 62;
[6] J.C. Herrera, PA 3 1972) 49 A.M. Sessler, V.G. Vaccaro, CERN 68-1 (1968)
[7] A. Sørenssen, PA 6 (1975) 141 [31] J.E. Griffin, Fermilab TM-1734 (1991)
[8] H.G. Hereward, A. Sørenssen, MPS/Int. DL 68-3 [32] V.V. Vladimirski, E.K. Tarasov, Theoretical
(CERN, 1968) Problems of the Ring Accelerator, USSR
[9] P. Faugeras et al, IEEE Trans.Nucl.Sci.NS-26 Academy of Science (1955)
(1979) 3577 [33] R. Gupta et al, IEEE Trans. Nucl. Sci. 32:2308
[10] S.Y. Lee, J. Wei, EPAC 88, 764 (1985); E.D. Courant et al, PAC 91, 2829
[11] J.E. Griffin et al, IEEE Trans.Nucl.Sci.NS-30 [34] L.C. Teng, PA 4 (1972) 81; R.V. Servranckx et
(1983) 2630; I. Kourbanis et al, PAC 91, 111; al, PAC89, 1355; D. Trbojevic, EPAC 90, 1536
E. Ciapala et al, IEEE Trans.Nucl.Sci.NS-26 [35] G. Guignard, PAC 89, 915; S.Y. Lee et al, PR
(1979) 3571 E48 (1993) 3040
[12] J. Wei et al, EPAC 94, 976; J. Wei et al, PAC 95, [36] Y. Ishi et al, APAC 98, 5D002
3334 [37] C. Pellegrini, D. Robin, NIM A301 (1991) 27;

[13] A. Sørenssen, 6th Int.Conf.High-Energy Acc. H. Bruck, IEEE Trans.Nucl.Sci. 20 (1973) 822;
(1967) 474 D. Robin et al, Micro Bunches Workshop, AIP
[14] J. Wei, S.Y. Lee, PA 28 (1990) 77 Proc. 376 (1995) 181
[15] D. Boussard, IEEE Trans.Nucl.Sci.NS-32, 1852
(1985)
[16] S.Y. Lee, J. Wei, PAC 89, 1169
4.10 RF GYMNASTICS IN A
[17] W.W. Lee, L.C. Teng, 8th Int.Conf.High-Energy SYNCHROTRON
Acc. (1971) 327 R. Garoby, CERN
[18] S.Y. Lee, J.M. Wang, IEEE Trans.Nucl.Sci.NS-
52 (1985) 2323; J. Wei, S.Y. Lee, Workshop on 4.10.1 Adiabaticity [1, 2]
RHIC Performance, BNL-41604 (1988) 339
[19] K. Takayama et al, PRL 78 (1997) 871
An rf gymnastic is called “adiabatic” when the pa-
[20] A.W. Chao, Physics of Collective Beam Instabil- rameters of the synchrotron motion are changed
ities in High Energy Accelerators, Wiley (1993) slowly enough for the distribution of particles to
p.263 be at equilibrium all along the process. The lon-
[21] H. G. Hereward, RL-74-062, EPIC/MC/48, gitudinal emittance is preserved and the process
Rutherford Lab (1974); D. Boussard, T. Lin- is reversible. “Adiabaticity” is quantified by the
necar, EPAC 90, 1560 dimensionless adiabaticity parameter:
( (
[22] J.A. MacLachlan, FN-446 (FermiLab, 1987) 1 ( dωS ((
ε = 2 (( (1)
[23] J. Wei, S.Y. Lee, A.G. Ruggiero, PA 24 (1989) ω S dt (
211
[24] W. Hardt and D. Möhl, CERN ISR-300/GS/69- Adiabatic means: ε 1 (typically ε < 0.1)
16 (1969); D. Möhl, CERN-ISR/300/GS/69-
62 (1969); W. Hardt et al, 8th Int. Conf. 4.10.2 Single Bunch Manipulations
High Energy Acc. (1971) 329; W.W. Lee, These manipulations involve changing bunch
L.C. Teng, IEEE Trans.Nucl.Sci.NS-18 (1971) length and energy spread.
1057; W. Hardt, 9th Int. Conf. High Energy Acc.
Controlled longitudinal blow-up [3]-[7] This
(1974) 434; W. Merz et al, PAC87, 1343; W.K.
non-adiabatic technique is used to stabilize the
van Asselt et al, PAC95, 3022
beam by decreasing the longitudinal density of
[25] L.C. Teng, FN-207/400 (1970)
[26] T. Risselada, CERN PS/90-15 (1990); A. Bogacz
particles while providing a well-defined (typically
et al, Fermilab III Instabilities Workshop (1990) parabolic) distribution with sharp edges.
177; S. Peggs, S. Tepikian, D. Trbojevic, PAC The basic ingredients of this rf gymnastics
93, 168 are:
[27] D. Möhl, PEP Note-41 (1973) (i) Main rf system (V0 at h0 ) keeping the
[28] J. Wei, S. Peggs, EPAC 94, 973 beam bunched; and (ii) Modulated rf at high fre-
[29] S.Y. Lee, K.Y. Ng, Fermi Lab III Instabilities quency (V1 at h1 h0 )
Workshop (1990) 170 The principle of this operation is to excite lon-
gitudinal resonances inside the bunch by phase
376
Table 1: Choice of parameters in a controlled longitudinal blow-up.
Parameter V̂1 /V̂0 h1 /h0 α (rad) ωm /ωs Duration
• h1 integer • high values tend

to create tails
∼ propor-
tional to • ∼ even integer blow up
Comment • bunch length increases
blow-up larger than 1 for uniform effect
speed on all bunches with time
period at h1 ,
lb > h2π (independence of
1 ω0 θ)
Typical 10 to 60 for max- 0.8π

0.1 to 0.3 imum blow-up to 3 to 12 ≥ 20 · 2π
ωs
range
speed 1.2π
1 principle is to stretch the bunch along the phase

(4) Compression efficiency (/10)
(time) axis, and let it rotate in the center of a high
(2) Optimum elongation (bunch and strongly mismatched bucket.
length/bucket length) before rotation
Voltage reduction
or
0.1
(3) Compressed length (bunch length/
Jump on unstable phase
bucket length) after rotation (x10)
Initial Bunch Stretched Bunch Bunch Rotation
Elongation techniques include: adiabatic or

(1) Equilibrium
(bunch length/bucket length) non-adiabatic voltage reduction, or stretch along
0.01
separatrix around the unstable phase. During
0.0001 0.001 0.01 0.1 bunch rotation, the bunch ends rotate slower than
Normalized Bunch Emittance 12–97
the centre, because of the nonlinearity of the rf
8355A101
waveform, giving an “S-shape” to the bunch in the

Figure 1: Equilibrium bunch length and bunch rotation phase plane and tails in the time-domain. For a
parameters vs normalized bunch emittance. given normalised emittance, the shortest bunch is
approximately obtained when the bunch extremes
modulation of V1 (h1 ), filamentation being accel- along the time axis and the extreme along the
erated by the nonlinearities due to V1 (h1 ). Phase energy axis before rotation end at the same po-
modulation of V1 : sition in time after ∼ 1/4 of a synchrotron pe-
V1 = V̂1 sin(h1 ω0 t + α sin ωm t + θ) (1) riod. The optimum elongation (curve {2}) and
with α is the peak phase modulation, ωm the the corresponding minimum bunch length (curve
modulation frequency and θ a constant phase off- {3}) are given in Fig.1, where compression effi-
set. The filamentation resonances correspond to ciency (curve {4}) is defined as (adiabatic bunch
ωm = kωS (k integer). The effects and the typi- length/bunch length) after optimum bunch rota-
cal range of parameters are given in Tab.1. tion.
Application of Fig.1 to a bunch filling 1%
Bunch compression [8, 9] (i) Adiabatic vari-
of the bucket leads to (i) equilibrium (adiabatic)
ation of parameters (energy, voltage, etc . . .) can
bunch length ∼ 0.073 of bucket length; (ii) opti-
reduce bunch length and increase energy spread.
mum elongation before rotation ∼ 0.27 of bucket
Curve {1} in Fig.1 gives (bunch length/bucket
length; (iii) minimum length after rotation ∼ 0.02
length) of a bunch versus its normalised emittance
bunch emittance . of bucket length; and (iv) compression efficiency
defined as: bucket acceptance ∼ 3.7.
(ii) Non-adiabatic methods of “Bunch rota- Further improvements can be obtained by re-
tion” are faster and attain smaller length. The ducing nonlinearities using multiple rf harmonics.
377
Table 1: Choice of parameters in a debunching or rebunching operation.
Parameter Voltage reduction func- Duration Final voltage Vfinal

tion Vred (t) tred
Comment Vinitial function of • determines emittance blow-

ε (adiab. up due to debunching
param.) and
ωs1 (synch. • when the bunch fills the
Vfinal freq. at bucket,
Vfinal ) debunched beam emittance =
π
bunched beam emittance
t= tred tred+tdeb 2
• the next figure shows that

ratio versus normalised
emittance in the bucket at

Vfinal
V (ε < 0.1)
Typical initial 2 10
range 1− 1−
Vinitial
V t
t tred ≈ ωs1
1
ε
final red
1
0.01 0.1 1 10
Normalised bunch emittance
4.10.3 Multi-Bunch Manipulations In practice, density modulation is still too

large after tdeb (e.g. tdeb = 0 when the bunch
These manipulations involve changing harmonic fills the bucket) and more time is necessary for
number and/or number of bunches. a good quality debunching with a small residual
Debunching-(Re)bunching [9, 10] modulation.
(i) Debunching transforms a bunched beam into (ii) (Re)bunching transforms a continuous
a continuous beam without azimuthal modulation beam into a bunched beam. Iso-adiabatic
of density. Iso-adiabatic debunching does it with (re)bunching is the time reversed equivalent of
a “slow” and constant adiabaticity (Eq.(1)) reduc- iso-adiabatic debunching. The same rules and for-
tion of the voltage from Vinitial to Vfinal , followed mulae apply.
by a step change from Vfinal to 0. The effects of
parameters are summarized in Tab.1. Merging and splitting [11],[12] Merging is
“Classical” debunching time tdeb is defined as used to divide the number of bunches. It can
the drift time required for the beginning of overlap be quasi-adiabatic and preserve the type of distri-
in azimuth of particles from adjacent rectangular bution if parameters are changed slowly enough.
bunches. Splitting is the reverse process.
π − Δϕ Basic ingredients for merging pairs of
tdeb = ,
hω|η|Δp/p bunches: two rf systems, the first one at the initial
where 2Δφ and 2Δp are the full spreads in phase frequency (V0 at h0 ) and the second one typically
and momentum. at half that frequency (V1 at h1 = h0 /2).
378
Table 2: Choice of parameters in a merging or splitting operation.
Parameter V0 (h0 ) at the beginning of merging Voltage functions Duration

linear time varia- function of adi-
Comment • determines the nor- tions can be used, abaticity and
malised bunch emittance: provided the nor- tolerable blow up
single bunch emittance
bucket acceptance withV0 (h0 ) malised bunch
emittance at merg-
• the next figure shows the ratio ing is large enough
V1 /V0 at which bunches begin (∼ 1/3)
to merge as a function of nor-
malised emittance
Typical ≥ 5 · ω2π where

s0
range ωs0 is the syn-
chrotron frequency
in the bucket due to
V0 (h0 ) at the begin-
ning of merging
Normalised bunch emittance
Principle: Time variations of V0 and V1 as il- (ii) step increase of V1 (h1 ) and cancellation of
lustrated in the following figure translate into such V0 (h0 ), to trigger a collective rotation of the set
an evolution in the phase plane that particles from of bunches around its centre during 1/4 of a
both bunches merge into a single one. Tab.2 gives turn (similar to bunch compression); (iii) suppres-
the effects of parameters in this operation. sion of V1 (h1 ) and step increase to a large volt-
age V0 (h0 ) for capturing all particles in a single
V1=0 V0(h0) V0= 0
bucket. See next figure.
V1(h0/ 2)
Time
V1/ V 0 = 0.5 V1/V0 =1.2
V0 = V0 initial V0= 0 V0>n2 V0 initial
V1= 0 V1>(h0/h1)V0 initial V1= 0
With a similar technique using three simulta- n initial
Rotation of Capture in 1
neous rf harmonics, quasi-adiabiatic splitting of bunches on h0
bunches on bucket on h0
bunches in three can be obtained without measur- h1 (< h 0 /2n)
able emittance blow-up [12].

Coalescing [10, 13] Coalescing is a non- Slip stacking [14, 15] Slip stacking is a non-
adiabatic process (large increase of the total lon- abiabatic scheme, whereby two sets of bunches
gitudinal emittance), by which n bunches are con- are finally concentrated into a single set.
centrated into a single one. Rf frequency and har- Basic ingredients: two rf systems operating
monic number are the same in the initial and final with slightly different frequencies.
states. Principle: (i) initial state: two sets of n
Basic ingredients: two rf systems, the first bunches separated in azimuth and energy, held by
one at the initial frequency (V0 at h0 ) and the two rf systems (peak voltage on each system =
second one at a much smaller frequency (V1 at Vdrift ) at h0 · f0 ± Δf ; (ii) drift (“slip”) of the 2
h1 < h0 /2n). beams with respect to each other until azimuthal
Principle: (i) initial state: n almost filled superposition is obtained; (iii) suppression of both
buckets with V0 initial (h0 ) and V1 (h1 ) = 0; rf’s and step increase to a large voltage V0 (h0 ) for
379
Table 3: Choice of parameters in a slip stacking operation.
Parameter Vdrift 2πΔf V0

• must be large to minimize
acceptance of drift- bunch distortion and final
ing bucket (Adrift ) large enough for
Comment emittance the final bucket to
bunch emittance
hold both bunches
(εl ) • limited by energy accep-
tance
> 2ωs drift with ωs drift /2π the > 4Vdrift for
Typical range Adrift ≥ 2εl synchrotron frequency in an un- 2πΔf = 2ωs drift
perturbed drifting bucket
capturing two bunches in each bucket. See figure 4.10.4 Debunched Beam Manipulation
below.
Phase displacement acceleration [17, 18] Us-
+ Δf ing “Phase displacement acceleration” the energy
h0f Rev
– Δf of a debunched beam can be changed at a slow
2 n bunches separated Azimuthal Capture on h0 of
rate with a limited rf voltage. It can be quasi-
in energy and azimuth slip 2 bunches/bucket adiabatic.
Basic ingredient: a single low voltage rf sys-
The effects of parameters on the slip stack- tem with a small frequency swing (a few %).
ing operation are shown in Tab.3. The emittance Principle: sweeping empty rf buckets (V0
blow-up can be reduced with a progressive reduc- at h0 ) through a debunched beam changes the
tion of Δf during the final approach. mean beam energy by: ΔE = Abucket frf where
Batch compression [16] Batch compression is Abucket is the acceptance of the sweeping bucket
applied to squeeze a set of n bunches into a in eVs . Effects of parameters are summarized in
smaller fraction of the circumference. The num- Tab.1.
ber of bunches is unchanged. It can be quasi- Limitations: small rate of acceleration and
adiabatic, preserving shape and emittance of progressive degradation of the stack edges with
bunches. the repeated passages of the buckets.
Basic ingredients: >2 rf systems with a large
enough tuning range.
4.10.5 Beam Manipulations with
Principle: progressively change the harmonic
Broad-Band RF Systems
number of the rf experienced by the beam by
“slowly” decreasing the voltage on the first har- Broad-band rf systems are able to generate volt-
monic h0 after having raised the voltage on h1 > age waveforms with a large spectral content. Be-
h0 . See figure below. yond the techniques described in the previous sec-
tions which are all applicable, they make feasible
2π / (h 0 ω 0 )
a large variety of specific manipulations.
Voltages
Barrier / Isolated bucket [10, 19, 20] The ba-
V0(h0)
sic building block is the beam barrier resulting
from a unipolar voltage pulse.
Principle: two voltage pulses of opposite po-
V1(h1) larity with a periodicity of one turn create a sin-
Time gle bucket. Depending on the polarity of the field
2π / (h 1 ω 0 ) and the sign of η, it is an “isolated” or a “bar-
rier” bucket (see figure below). In the case of
Limitations: reduced acceptance of the buck- a single sine-wave (two contiguous half-sines),
ets displacing the edge bunches. A fraction of the bucket height is given by the conventional for-
circumference (∼ 1/5) must not contain particles. mula (Sec.2.1.2).
380
Table 1: Choice of parameters in phase displacement acceleration.
Parameter V0 (h0 ) Γ = sin ϕs Duration

as large as possible as small as possible
Comment to minimize blow to minimize blow proportional to: (Abucket V0 Γ)−1
up up
Typical range Γ ∼ 0.1 a few hundred sweeps
Voltage Revolution Period

Waveform
[16] R. Garoby, IEEE NS-32, No. 5 (1985) 2332
T
η<0
Accel.
η>0
[17] K.N. Henrichsen, M.J. de Jonge, CERN-ISR-
Decel. Barrier RF-MA/74-21

Isolated
Bucket Bucket
Δp/p [18] E.W. Messerschmid, CERN/ISR-TH/73-31
[19] J.E. Griffin et al, IEEE NS-30, No. 4 (1983) 3502
[20] M. Blaskiewicz, J.M. Brennan, EPAC 96, 2373
Applications of barrier buckets [21] Combin- [21] C.M. Bhat, RPIA 2006, FERMILAB-CONF-06-
ing multiple pairs of voltage pulses and varying 102-AD
their amplitudes and phases as a function of time,
bunches can be quasi-adiabatically accumulated
or extracted from a well-defined part of a high in- 4.11 ENERGY MEASUREMENT
tensity stack. Such techniques have been exten- WITH ELECTRON BEAMS
sively used in the FNAL Recycler Ring [21]. J. Seeman, SLAC
Similarly, bunches can be merged or split,
and their length and energy spread can be quasi- Single pass energy measurement Single pass
adiabatically modified [21]. energy measurements are typically made by mea-
Limitations: need for specific rf systems suring the bend angle of a beam through a dipole
which cannot be used for fast acceleration be- magnet. However, in general, the beam is made to
cause of their limited peak voltage resulting from pass through a transport line to define a “straight”
the intrinsically low impedance of broad-band trajectory and then is made to pass into a trans-
resonators. port section with dispersion D from several dipole
magnets. If transverse beam positions relative to
References the nominal trajectory before and after the dipole
magnet are zero, the nominal energy E0 is defined
[1] W.T. Weng, AIP Conf.Proc. 184, Summer by the geometry. For a drift-dipole magnet(s)-
Schools 1987 & 1988, p.242
drift geometry,
[2] M.H. Blewett (editor), CERN 77-13, 63-81
[3] V.V. Balandin, M.B. Dyachkov, E.N. Chapoc- 0.29979 Tm
E0 [GeV] = Bdl
hnikova, PA 35 (1991) 1 θ rad
[4] R. Cappi, R. Garoby, E. Chapochnikova, where θ is the measured geometrically surveyed
CERN/PS 92-40 (RF) bend angle. A 1 m long dipole with a 1 T field
[5] R.K. Reece et al, PAC 89, 1934 bends a 10 GeV beam by 30 mrad. Integrated
[6] R. Cappi et al, PAC 93, 3570
dipole fields can be measured in the range of 0.1
[7] J.M. Brennan et al, EPAC 94, 1897
to 0.5% and angles can be measured to about
[8] J. Griffin et al, IEEE NS-30, No. 4 (1983) 2630
0.01%. The magnet power supplies must be reg-
[9] R. Garoby, CERN PS/RF/Note 93-17
[10] J.E. Griffin, J.A. MacLachlan, Z.B. Qian, IEEE
ulated and calibrated to provide the required res-
NS-30, No. 4 (1983) 2627 olution. If the position readings before the dipole
[11] R. Garoby, S. Hancock, EPAC94, 282 magnet(s) or in the dispersive region are nonzero
[12] R. Garoby, CERN/PS 98-048(RF); HEACC98 arising from either betatron oscillations or from
Dubna, 1998 the beam being off the standard energy, the en-
[13] D. Wildman et al, PAC 87, 1028 ergy must be corrected using these offsets [1, 2].
[14] F.E. Mills, BNL Int.Rep. AADD 176 (1971) The correction ΔE/E0 to E0 above is
[15] D. Boussard, Y. Mizumachi, IEEE NS-26, No. 3 ΔE x3 − m31 x1 − m32 x2
(1979) 3623 =
E0 D3
381
Sec.4.12: RING INJECTION AND EXTRACTION
where x1 and x2 are transverse position measure- lutely. See Sec.2.6.1 for the connection between
ments in the plane of the bend in the nondispersive the beam energy and the spin tune. See Sec.7.6
transport lines (D1 = D2 = 0) defining the initial for polarimeters .
trajectory and x3 a position measurement in the In very large storage rings the energy can vary
dispersive region with D3 nonzero. mij represent due to these causes and others such as earth tides.
the R11 transport matrix elements (Sec.2.2.1) be- Detailed reports concerning the energy calibration
tween positions i and j. It is assumed here that of the LEP storage ring, taking these effects into
the positions 1 and 2 are 90 degrees apart in be- account, can be found in [5, 6].
tatron phase, else the proper phase must be used. Another method useful for an e+ e− collider
For example, with mij = 1, D3 = 1 m, and x is to measure the ring energies when the machine
measurements on the order of 1 mm, energy cor- is producing particles on a narrow excitation en-
rections are typically of order 0.1%. ergy. For example while operating on the Y(2S)
Storage ring energy measurement A stored resonance, the beam energies are known to be at
beam passes through a series of bends and quads 10.02326 ± 0.00031 or one part in about 30,000
[7, 8].
which create a closed orbit, approximately circu-

lar. Usually, the beam is made to pass through the
centers of the quadrupoles verified by beam posi- References
tion measurements. The beam energy is
[1] J. Seeman, AIP Proc. 249 (1992) 49
0.29979
E0 [GeV] = n Bdl [Tm] [2] J. Frisch et al, BIW 08 (2008) 17
2π
[3] M. Sullivan et al, PAC 01, 3570
where Bdl is the integrated dipole field of one [4] R. Assmann et al, Z. für Physik C, Springer-
of the n dipole magnets around the ring. For ex- Verlag, C66 (1995) p.567
ample, a ring with 100 dipoles of length 1 m and [5] LEP Energy Working Group, CERN-EP/98-040,
field 0.628 T produces a beam energy of 3 GeV. CERN-SL/98-012 (1998)
Again, integrated dipole fields can be measured in [6] http://www.cern.ch/LEPCAL/reports/reports.html
the lab from 0.1 to 0.5% and magnet power sup- [7] Y. Cai et al, PRST-AB 12, 061002 (2009)
plies must match the required precision. When [8] Particle Data Group, Particle Physics Booklet,
other dipole fields are present, corrections to the July 2010, 125
energy calculation must be made [3, 4]. Several
common sources for energy corrections including
typical relative contributions are: (i) Dipole cor- 4.12 RING INJECTION AND
rection magnets in the dispersive arcs which are EXTRACTION
used for orbit correction that have non-zero av- G.H. Rees, Rutherford Appleton Lab.
erage strength (0.02%). (ii) A beam that passes
off axis in the quadrupoles in the arcs, perhaps Single turn injection with a septum and kicker
due to a changed rf frequency, from misalign- [1] Input line, lattice functions are set to match
ments, or from tidal forces (0.05%). (iii) Tem- those for the ring at the septum output (αx = αs ,
perature changes on the lengths and the pole gaps βx = βs ). Kicker unit is downstream at (αx =
of the dipole magnets (0.02%). (iv) Extraneous αk , βx = βk ), and deflection θ brings the beam on
currents which flow in the vacuum chambers or axis, after betatron phase shift, μk . Input beam at
magnets ground paths (0.05%). (v) Energy losses output of septum, at (x, x , y = y = 0), requires:
from around the ring due to synchrotron radiation
(0.05%). (vi) Hysteresis effects from standardiza- x = −(αs + cot μs )x/βs (1)

tion errors (0.05%). θ = x/(sin μk (βs βk )) (2)
Precise NMR measurements (0.001%) of the
main dipole fields and rotating coil measurements For kickers and septa see Secs.7.2.7 and 7.2.16,
of the relative integrated dipole fields (<0.01%) respectively.
allow estimation of time varying effects on the Conventional, one plane multi-turn injection
beam energy. with septum and orbit bump [1] (Fig.1) As-
Spin tune depolarization measurements using sume un-normalized, beam emittances and sep-
backscattered laser photons can be used to de- tum output, lattice parameters of (εi , αi , βi ) for
termine the beam energy to about 1 MeV abso- input beam and (ε, α, β) for final ring beam. The
382

X' = ( x + x')/ size amplitudes by the end. For the latter, beam
motion is large in both planes at the start and is
septum or foil small by the end of injection, or vice-versa. The
orbit painting bumps have to be programmed so
that the beam traversals of the foil are minimized.
X = x/ Collapsing bumps offer advantages.
One scheme uses a fixed orbit bump, in a low
<<
i dispersion region; it may have a corner foil and
extra x, y orbit bumps for the painting (SNS) or a
Figure 1: Optimized, 1-plane multi-turn injection. side foil with input steering for y and an x-orbit
√ bump (J-PARC). Another scheme (ESS, KAON)
Mismatched input ellipse touches ε circle once.
has the injection foil at the waist points of a low

field, ring dipole, and uses a corner foil, a y-orbit
number of injected turns, n, is maximized by ad- bump and a p-momentum ramp for the y- and p-,
justing input mismatch and collapse rate of in- x-painting. The dispersion at the foil, for a ramp
jection orbit bump. An electrostatic septum and of Δp/p up to Δ, requires a Dx value:

a non-dispersive, injection straight are beneficial. √ √
Input beam at (x, x , y = y = 0), relative to the Dx ≤ ( (εx βx ) − 2 (εix βix ))/Δ (5)
orbit bump, at the septum output, requires: Charge exchange injection for FFAG rings
α/β = αi /βi = −x /x ; (3) Most FFAGs have finite dispersion straights, too
short for an injection orbit bump. An exception is
β/βi = α/αi = (ε/εi )1/3 ; a non-linear, scaling FFAG with high cell tunes,
qx ≈ 0.78, qy ≈ 036, which allow low dispersion
n (at low space charge) ≈ 0.5 − 0.7(ε/εi ) straights, of a length up to ≈ 8 m. The non-linear,
Two-plane, multi-turn injection with a tilted x − y ring coupling influences the distributions
septum and orbit bumps [2] The number of painted by the orbit bumps.
injected turns, n, is maximized by adjusting the Resonant injection scheme (ZGS) [4] In-
input mismatching, the tilt of the septum, and the volved are a fast kicker and a pulsed, bump unit
collapse rates of the orbit bumps. The require- with dipole, quadrupole and octupole fields. Two
ments are (using earlier symbols, but with y = 0, stable regions are formed beside an unstable fixed
y = 0 and adding subscripts x, y): point. After beam is injected into the outer one,
fields are set to merge the regions. Injection is
βx /βix = αx /αix = (εx /εix )1/3 ; repeated by re-setting conditions for the outer re-
βy /βiy = αy /αiy = (εy /εiy )1/3 ; gion, leaving injected beam in the inner region.
αx /βx = αix /βix = −x /x ; Electron and Positron (e± ) Injection Fast,
single-turn injection for low, electron or positron
αy /βy = αiy /βiy = −y /y
energies is similar to that for most ions. At high
εx εy energies, the radiation damping plays a significant
n ≈ (0.1 − 0.125) (4) and useful injection role, as indicated in Fig.2 (X
εix εiy
and X are as for the Figure 1). Here, I is the
Multi-turn, charge exchange injection [1, 3] mismatched injected beam, Xs the inner edge of
Key items are foil, or laser or gas jet strippers, the injection septum, Xco the collapsing horizon-
foil scattering, foil temperatures, injection mis- tal orbit bump, and S shows the stored beam af-
match and painting, and the removal of unwanted ter damping. If Xσi is the rms beam radius after
beam from the ring. After the stripper, input and injection, and Xσ is the rms beam radius at maxi-
circulating beams merge, and stripped electrons mum energy, then:
pass to a nearby collector. High intensity needs
Xs > Xco + 6Xσi and Xs ≥ 10Xσ (6)
εix εx . For H− beams, fields at the foil are set
to optimize the lifetimes of excited H ◦ states. Pulsed sextupole magnet for e± injection [5]
An anti-correlated, or correlated, x − y beam A conventional, pulsed local orbit bump, with
distribution may be painted. Beam motion in the four kickers sited along one long straight section,
former is large in the x- and small in the y-plane may be replaced by the action of a single, pulsed
at the start of injection, transposing to opposite sextupole magnet. The parabolic waveform of the
383

X' = ( x + x')/ extraction efficiencies found for MTE at the PS
ring are in the range from 97 to 98%. An order of
Xco
magnitude lower loss would have to be obtained
Xs if the scheme were to be used for a MW proton
driver.

X = x/ Multi-pass halo extraction by bent crystals [7]
S Multi-turn halo extraction may be obtained for a
high energy ring beam via a gently bent crystal at
I the beam edge. Halo moves out to the crystal by
beam-gas scattering, beam-beam collisions or by
Figure 2: Multi-turn, electron-positron injection. noise. Use of a kicker may speed up the move.
Positively charged, beam particles may pass into

sextupole field allows injected beam to receive a a crystal several times before channeling. Crystal
required input deflection, while the effects on a dislocations and defects, and multiple scattering
stored beam are small due to the field reduction to in a channel influence total extraction-bending ef-
zero on the sextupole axis. Key parameters are (x, ficiencies, Eo , given by:
x ) values at the output of an injection septum, lat- Eo = Ee (1 − RT /Rm ) exp(−s/λdb ) (7)
tice functions at the septum and downstream sex-
tupole and the horizontal betatron phase shift be- where Ee is the extraction efficiency, RT =
tween the septum and the sextupole. Practical is- pβc/Ec is the Tsyganov radius, Ec is the inter-
sues involve the reduction of sextupole, eddy cur- atomic field where particle trajectories become
rents, and the protection from the high power syn- unstable, Rm is the minimum bending radius of
chrotron radiation. curvature, s is the crystal length and λdb is the
de-channeling length. Silicon crystals are used as
Single turn ejection with kicker and septum
they are free of large dislocations. A typical effi-
This is the reverse of the single turn injection pro-
ciency of 30% is found at Fermilab when using a
cess, described earlier. A phase shift from the
40 mm length of crystal for a 1 TeV beam.
kicker output to septum input of π/2 is optimum.
Kickers may be electrostatic, or have a delay line Slow, third order resonant extraction [8]
or transmission line form (terminated or shorted) Slow extraction on the third order resonance,
or be of a lumped, inductive, pure magnetic type. 3Qx = n, is summarized here. Details are
given in the next section. For the S(s) =
Fast, multi-turn extraction (MTE) [6] Beam
∂ 2 By (s)/∂x2 /(2Bρ), sextupole excitation, the
to be extracted is divided into beamlets, one on
equation for x is:
the axis and four in stable outer horizontal phase
space islands. The scheme is the reverse of the d2 x/ds2 + K(s)x = −S(s)(x2 − y 2 ) (8)
ZGS, resonant injection noted earlier, but the use √
A normalization to u = x/ βx and ϕQx =
of sextupole and octupole fields creates four outer
∫ ds/βx , with r 2 = ε = u2 + p2 and pQx =
islands, not one. The five islands separate after
du/dϕ, leads to
sweeping through the resonance involved, which
is at 4Qh = 25 for the PS ring at CERN. The d2 u/dϕ2 + Q2x u = An Qx u2 cos(nϕ + ϕo ) + . . .
required separation is controlled by the Qh value (9)
at the end of the resonance crossing and by the where An = nth harmonic component of
strength of the non-linear magnets. SQx (βx )5/2 .
The separated, circulating beamlets are Adjusting Qx and S(s) transforms circular
moved turn by turn, beyond the septum of the ex- orbits in (u, ϕ) space into triangular separatrices,
traction magnet by the use of slow and fast closed along which there is an exponential increase of
orbit bumps. In comparison with CT, the con- x to the extraction septa, sited in dispersion free
tinuous transfer of beam, previously used at the regions. Rf knock-out increases the emittance,
PS ring, the MTE scheme has a beam loss at the ε, and is often used. Tune adjusts may be di-
septum reduced by a factor of two. The unifor- rect, or by rf steering (chromatic ejection). Other
mity of the extracted beam is also improved, as schemes use stochastic extraction or a high-β in-
the trapped beamlets may have similar intensities, sertion. For all momenta, separatrices need to be
emittances and focusing parameters. The typical, superimposed [8]. The unstable, fixed points are
384
at (r, γ) from the origin, given by r = 8((n/3) − A X'

Qx )/An , and by γ = nϕ/3 + tan−1 (p/u) =
(0, −2π/3. − 4π/3) − ϕo /3.
P3
References
P1
[1] G.H. Rees, CERN 94-01 (1994) 731 h
[2] C.R. Prior, G.H. Rees, NIM A415, No. 1, 2 (1998)
357 X
[3] G.H. Rees, RAST, Vol. 1 (2008) 45 Stable
[4] T.K. Khoe, R.J. Lari, VIIIth Int.Conf.High Energy region
C
Acc., CERN (1971) 98
[5] H. Takaki et al, PAC 07, 230 P2 Unstable region

[6] A. Franchi, S. Gilardoni, M. Giovannozzi, surounding stable
http://ab-project-mte.web.cern.ch & CERN Triangle
Courier, 2009 B
[7] V.M. Biryukov, V.I. Kotov, Yu.A. Chesnokov, 12–97
8355A114
Crystal Chanelling and its Applications at High
Energy Accelerators, Springer, 1997 Figure 1: Phase-space geometry at the sextupole.
[8] W. Hardt, CERN PS/DL/LEAR, Note 81-6 h = 2ε/(3S) =√4πδν/S. Separatrix: A: X =
−2ε/(3S). B: − 3 X /2 + X/2 = 2ε/(3S). C:
√
3 X /2 + X/2 = 2ε/(3S). Fixed√points: P1 :
4.12.1 Aspects of Slow Extraction
(4ε/(3S), 0). P√2 : (−2ε/(3S), −2ε/( 3 S)). P3 :
P.J. Bryant, CERN (−2ε/(3S), 2ε/( 3 S)).
(See also Sec.4.12.) For irradiation of tumors
in medical synchrotrons (Sec.1.6.14), the survival n ± 1/3. Fig.1 summarises the phase-space ge-
of the tissue surrounding the tumour depends not ometry derived from Eq.(1) at the sextupole for
only on the spatial precision of the beam delivery, mono-energetic particles whose equilibrium orbit
but also on the ability of the machine to deliver an is centred on the origin.
exact radiation dose. The use of slow resonant ex- The sense of rotation in the stable region is
traction from a synchrotron [1] extends the beam clockwise when the particles are above resonance
spill time sufficiently to perform on-line dosime- and the whole diagram will rotate clockwise if
try at the patient and to switch the beam on and the observer moves downstream. The resonance
off according to the dose required. Either the half- sextupole is best put in a dispersion-free region.
integer or third-integer resonance can be used, but Elsewhere in the machine the dispersion will shift
the current trend is towards using the slower and the equilibrium orbit and the stable triangle. The
more controllable spills from the third-integer res- equation of a separatrix anywhere in the machine,
onance. Higher order resonances are not used be- including the effect of dispersion, is
cause the angular separation of the separatrices (X − Dn δ) cos α + (X − Dn δ) sin α = h (2)
becomes too small.
where (Dn , Dn ) is the normalised dispersion
Basic theory A possible technique for design-
function and α is the angle measured from the
ing an extraction system is to apply the simpli-
X-axis in the counter-clockwise direction to the
fied Kobayashi Hamiltonian, H, for the motion in
perpendicular, h, to the extraction separatrix.The
the x plane [2, 3] and then to refine the design
spiral step, ΔXES , at the electrostatic septum is
with tracking programs that include higher-order 2
effects, 4ε 3S 2 2 π
ΔXES = − (XES +XES ) | cos(α− )|
ε S 3S 4 2
H = (X 2 +X ) + (3XX −X 3 ) (1)
2 2
(3)
2 4
Hardt condition When the beam enters the res-
where (X, X ) are the normalised coordinates of onance, there is a range of momentum and emit-
a particle, the normalised sextupole strength is tance values that defines a boundary between the
3/2
S = (1/2)βx,sext (d2 By /dx2 )sext /(Bρ), ε = stable and unstable regions (Fig.2). The large am-
6πΔν, δν is the distance from the resonance ν = plitude/emittance particles will enter resonance
385
Betatron Uncorrected Hardt condition gives single X'
amplitude separatrices extraction separatrix
Half of the Machine Lattice Resonance
Momentum spread X' X' sextupole
of extracted beam X
D = D '= 0
Resonance dividing
stable and unstable
regions
Beam moving into X X Δμ = 225°

resonance Electrostatic
septum (ES) X'
12–97 D > 0, D '< 0 Kick
8355A115 Qresonance Radial position/tune/momentum
X
Figure 2: Hardt condition seen at electrostatic septum.
Δμ < 90°
Magnetic septum '
X
first, while still some distance from the reso- (MS)
nant tune, and there is a continuous range of D = D '=0
X
stable-triangle sizes down to zero for the zero-
amplitude/emittance particles that must be exactly
Kick becomes
on resonance to be unstable. In general, the dis- a gap for (MS)
3–98
8355A116
persion function will spread out this range of

phase-space triangles and separatrices so the par- Figure 3: Possible layout of the extraction elements.
ticles of different momenta will reach the electro-
static septum with different angles. The condition 105◦ before the following separatrix hits the sep-
to set the optics such that all the extraction sepa- tum, but in practice an angle ∼ 60◦ is sufficient
ratrices are aligned to give minimum losses on the and more convenient.
electrostatic septum is known as the Hardt condi-
Sensitivity to ripple [6] Irregularities of spill
tion [4, 5],
often quantified by duty factor F ,
Dn cos α + Dn sin α = −4πξ/S (4)
F = φ2 /φ2 over time Tspill (5)
where ξ = dν/(dp/p) is the chromaticity.
Effectively the Hardt condition moves the sta- where φ = dN/dt, the particle flux in the spill.
ble triangles along the direction of the dispersion The name duty factor is an unfortunate accident
vector by varying ξ. Fig.2 describes a general sce- of history and should not be confused with duty
nario, but the Hardt condition is only effective if cycle.
the resonance is stationary and the beam is moved Typically, the transit time in the resonance
e.g. by a betatron core or stochastic rf noise. varies from 50 to 4000 turns and most particles
Medical machines work below transition and are extracted within 2000. For a revolution time
ideally ξ < 0 to ensure the transverse stability of 0.5 μs, this represents a delay of 1 ms. Thus,
(Sec.2.4.9) of the “waiting” beam. Extraction is for common power supply, ripple frequencies be-
best made to the outside of the ring and injection low 1 kHz, it is reasonable to assume that the par-
to the inside. These choices limit the possibili- ticle flux entering the resonance appears instan-
ties for applying the Hardt condition to positions taneously in the spill. The particle flux in the
where the electrostatic septum has Dn > 0, Dn < spill is then directly proportional to the relative
0, e.g. as in the second half of a dispersion bump. motion, ν̇, between the resonance and the wait-
Configuration of the resonance Fig.3 shows a ing beam, i.e. φ = Ṅ = λ(ν̇0 + ν̇ripple ) =
possible layout for exciting the νx = 1.666 reso- λ(ν̇0 + ωδν cos ωt). Substitution in Eq.(5) gives
nance. The lattice comprises two arcs with closed duty factor for ripple below 1 kHz,
dispersion bumps linked by two dispersion-free −1
1 (ωδν)2
insertions. The resonance sextupole is placed in F = 1+ (6)
2 ν̇02
one dispersion-free region. The electrostatic sep-
tum is placed in the second half of the arc where The above assumes that ν̇0 > ωδν. The contrary
D > 0 and D < 0 for the Hardt condition. The is likely to be true and, once ν̇0 ≤ ωδν, the res-
phase advance to the electrostatic septum must onance plunges periodically into the stack, chop-
not be less than 210◦ or the next separatrix will ping the spill into a series of spikes.
be aperture limited on the septum. The phase ad- Above 1 kHz, the storage of particles in the
vance to the magnetic septum can be as high as resonance applies some smoothing and modifies
386
Modulated by
low frequency
φ0 ripple Poisson
spill
φav
φav/2
Frequency
Half of the spill Over-modulated by
forms a uniform medium frequency High frequency ripple
base line ripple into series of smoothed by over–
12–97 individual spikes lapping first plateau,

8355A117
shoulders & then peaks
Figure 4: Typical characteristics of ripple at different frequencies.

Eq.(6) into (7). This occurs because the spikes ratio. However, in measurements that are inte-
have a characteristic shape with a narrow leading grated over the spill, the true emittance may be
peak, a small shoulder and then a long plateau and masked by movements of the separatrix, changes
the plateaux overlap [7]. Above 1 kHz, it is very in machine parameters, or dispersion effects. One
likely that ν̇0 > ωδν and the spill becomes a se- effect that is not included in Eq.(9) but does give
ries of spikes. Duty Factor for ripple 1-100 kHz: a true emittance in the extracted beam is the cou-
−1 pling from the vertical betatron motion in the ex-
1 (ωδν)2
F = 1+ (7) citation sextupole [9]. This effect can be esti-
4 ν̇02 mated using
√
Above 100 kHz, the leading spikes overlap x 3π 3 βy 2 2 2
and the spill is smoothed until the frequency is ≈ S (XSE + XSE ) sin 2πνy(10)
y 4 βx
high enough to reveal the Poisson statistics in the where x,y are the geometric extracted emittances.
beam density,
F = N̄ /(N̄ + 1) (8) References
where N̄ is the expected number of particles in [1] H.G. Hereward, Proc. Vth Int. Conf. on High En-
a single measurement bin. Fig.4 summarises the ergy Acc. (1963)
characteristics of spills affected by low, medium, [2] Y. Kobayashi, H. Takahashi, Proc. VIth Int. Conf.
high and very high frequency ripple. High Energy Acc. (1967) 347
Eq.(6) shows that it is possible to reduce the [3] M.Q. Barton, Proc. VIIIth Int. Conf. High Energy
influence of ripple by increasing the speed, ν̇0 , at Acc. (1971) 85
which the particles cross the boundary between [4] W. Hardt, CERN/PS/DL/LEAR Note 81-6
the stable and unstable regions. This can be done [5] C. Steinbach, EPAC 92, 857
by adding a band of stochastic noise to accelerate [6] R. Cappi, C. Steinbach, PAC 81
the beam [8], or by channelling the particles past [7] M. Pullia, CERN/PS 97-50 (DI)
an empty rf bucket [6]. [8] S. van der Meer, CERN/PS/AA 78-6 (1978)
[9] Private communication M. Pullia, TERA Founda-
Emittance of the extracted beam In a per-
tion and CERN
fectly adjusted extraction scheme, the emittance
x of the extracted beam will be extremely small.
This can be seen from the Liouville theorem and
the conservation of the combined longitudinal and
4.13 COLLISION SCHEMES FOR
transverse phase spaces, RING COLLIDERS
M. Zobov, INFN Frascati
x,stack δstack T0 = x,spillδspill Tspill (9)
F. Zimmermann, CERN
The large ratio between the revolution time and
the spill time (∼ 106 ) is compensated by an Present generation lepton factories have been very
equally large reduction in the transverse emittance successful [1]. Both B-factories, KEKB in Japan
387
Sec.4.13: COLLISION SCHEMES FOR RING COLLIDERS
and PEP-II in USA, largely exceeded their design angle φ = σz /σx∗ tan(θ/2) avoided excessive ge-
goals. The Italian Φ-factory DAΦNE exceeded its ometric luminosity reduction and diminished the
phase-I design luminosity and further increased strength of synchrobetatron resonances excited by
it by a factor 3 after implementing a novel crab beam-beam interaction with a crossing angle.
waist collision scheme. The recently commis- However, further substantial luminosity in-
sioned Tau-Charm factory in Beijing is well ad- crease based on the standard collision scheme
vanced towards its design luminosity. Similar is hardly possible due to several limitations im-
progress is seen for hadron colliders, where, with posed by beam dynamics requirements: A) In or-
much ingenuity, Tevatron and RHIC pushed their der to minimize the luminosity reduction due to
luminosities to new limits, and, only in its sec- the hour-glass effect (the variation of the vertical
ond year of operation (2011), the Large Hadron beam size within the longitudinal position along
Collider (LHC) at CERN exceeded its design lu- the collision region) the vertical beta function at
minosity if scaled to the actual beam energy of 3.5 the IP cannot be much smaller than the bunch
TeV [2] (Sec.4.1). length. B) A drastic bunch length reduction is
All the present generation ring colliders re-
impossible without incurring single bunch insta-

lied, at least in the beginning of their operation, on bilities: bunch lengthening and microwave insta-
the standard strategy choice for high-luminosity bilities due to the beam interaction with the sur-
parameters, which can be understood from ex- rounding vacuum chamber. Besides, too short
pressions for luminosity L (Sec.4.1) and beam- bunches tend to produce coherent synchrotron ra-
beam parameter ξ x,y (Sec.2.5.1). In the case of diation (CSR) affecting beam quality and leading
head-on collisions of identical short bunches they to a dramatic increase of the power losses. C) A
are multibunch current increase would result in dif-
Nb2 ferent kinds of (coupled-bunch) beam instabili-
L = nb f0 (1)
4πσx∗ σy∗ ties, in excessive power loss due to interactions
with parasitic higher order modes (HOM), resis-
πγ 2 ξx ξy x σy∗ 2 tive wall or due to electron cloud (especially for
= nb f0 1 +
re2 βy∗ σx∗ SC hadron rings), in an increase of the required
with wall plug power, or in unrealistic rf system pa-
∗
βx,y
Nb re rameters; D) Higher emittances conflict with stay-
ξx,y = (2)
2πγ σx,y (σx∗ + σy∗ )
∗ clear and dynamic aperture limitations, and re-
quire again higher currents to exploit the emit-
where the asterisk indicates quantities at the inter-
tance increase for the luminosity enhancement. F)
action point (IP). This shows that the luminosity
Tune shifts saturate and beam lifetime drops due
can be increased by:
to a strong nonlinear beam-beam interaction.
1. raising number of particles per bunch, Nb ; 2.
more colliding bunches, nb ; 3. increase of the In order to overcome these (and other) limi-
emittance, x and y (at constant ξx,y ); 4. smaller tations several novel collisions concepts and new
∗ ; 5. colliding beams
beta functions at the IP, βx,y collision schemes were proposed. The most pop-
with equal rms sizes at the IP, σy∗ = σx∗ ; and 6. ular approaches are the following:
pushing up the beam-beam parameters, ξx,y . 1. round beam collision preserving additional
The present (and past) colliders obtained their integral of motion (e.g. [3] ); 2. crab crossing
good luminosity performances by trying to ful- [4, 5, 6]; 3. collision with large Piwinski angle
fill the above conditions as much as possible ex- [7] (including “superbunch” schemes in hadron
cept that: A) for lepton factories transversely flat colliders [8, 9]); 4. collisions with alternating
bunches are collided, with σy∗ σx∗ , since the planes of crossing in different IPs [10]; 5. (other)
transverse emittances are naturally unequal and it beam-beam compensation schemes (Sec.4.14);
is rather difficult to provide a good dynamic aper- 6. colliding hadron bunches with longitudinally
ture in case the IP beta functions are very low in flat profile [9, 11, 12]; 7. crab waist collision
both transverse planes; for hadron colliders the [13, 14]; 8. collision with traveling waist [15];
beams are indeed round; B) in order to eliminate 9. longitudinal strong rf focusing [16]; 10. colli-
parasitic collisions in multibunch operation a non- sions with negative momentum compaction fac-
zero horizontal crossing angle θ has been neces- tor [17], e.g. for shortening the bunches and
sary. In the factories a relatively small Piwinski mitigating longitudinal beam-beam effects; and
388
11. collisions at much higher beam energy with KEK B-factory [24], where a world record lumi-
extremely large crossing angle [18, 19], e.g. for nosity of 2.1x1034 cm−2 s−1 was obtained in this
reducing the beam-beam tune shift, suppressing configuration. However, the achieved luminosity
instabilities, and (for leptons) enhancing radiation remained lower than predicted by numerical sim-
damping. ulation.
The idea of round-beam collision was pro- Colliding with a large Piwinski angle had
posed more than 20 years ago for the Novosi- been discouraged after the experience of harmful
birsk Φ-factory design. It requires equal emit- synchrobetatron resonances with 24-mrad cross-
tances, equal and small fractional tunes, equal ing angle at the double storage ring DORIS-I [23],
beta functions at the IP, no betatron coupling in until, in 1995, Hirata suggested that a large angle
the arcs. A 90◦ rotation at each passage of the might have several merits for future lepton fac-
transverse oscillation plane by means of solenoids tories [7]. Also for hadron colliders increases in
in the interaction regions (IR) provides conserva- the bunch length or the crossing angle have been
tion of the longitudinal component of the angular proposed for luminosity optimization at the beam-
moment Mz = ypx − xpy . Thus the transverse beam limit [8, 9]. The advantages of a large Φ can
motion becomes one-dimensional. In addition to be understood by writing down the formulae for
the obvious advantages from Eq.(1), the round the luminosity and tune shifts with a horizontal
beam scheme helps to eliminate all betatron cou- crossing angle. Neglecting the hour-glass effect,
pling resonances that are of crucial importance these expressions can be obtained from Eq.(1)
for tune shift saturation and lifetime degradation. simply by substituting the horizontal beam size
The synchrobetatron resonances are also weak- σx∗ by the effective transverse size σx∗ (1 + Φ2 )1/2 .
ened since the transverse tune shift is almost in- Then, for large Piwinski angle, Φ 1, the lumi-
dependent of the particle’s longitudinal position nosity and the tune shifts scale as [25].
along the bunch. The round beam concept was
successfully tested at the electron-positron col- Nb ξ y Nb βy∗ /y Nb
L∝ ; ξ y ∝ ; ξx ∝
lider VEPP2000 in 2007 at 510 MeV beam en- βy∗ σz θ (σz θ)2
ergy [20]. Despite the low energy a high single (3)
bunch luminosity of 1031 cm−2 s−1 was achieved If in such a case Nb is increased proportionally
together with a maximum tune shift as high as 0.1. to σz θ, the vertical tune shift ξy remains constant,
Another round beam collisions scheme, “Möbius and the horizontal tune shift drops like 1/(σz θ),
accelerator” [21] was tested at CESR. It provided while the luminosity grows as σz θ. In hadron col-
a tune shift of 0.09 in agreement with simulations liders with round beams and alternating crossing
[22]. at two IPs the scaling of both ξx and ξy is identical
In 1988 R. Palmer proposed the crab-crossing to the one for ξy in the above formulae [8, 9].
collision scheme for linear colliders [4]. Soon In the case of hadron bunches colliding at two
it was adopted and further developed for stor- IPs with alternating crossing at a large Piwinski
age ring colliders [5]. This idea makes it pos- angle, for the same bunch charge and the same
sible to collide bunches at a large crossing an- beam-beam tune shift the luminosity of a uni-
gle without luminosity loss and without excita- form
√ (or flat) longitudinal distribution is exactly
tion of synchrobetatron resonances. In the crab 2 times higher than for a Gausian bunch pro-
crossing scheme both bunches are tilted prior to file [11]. Flat bunches could be generated by us-
the collision, by half the crossing angle θ/2, ef- ing higher harmonic rf systems [12], by accelerat-
fectively providing a head-on collision at the IP. ing and/or generating one or several long super-
The tilt is created by a transverse rf deflector bunches [8, 9] using a barrier bucket rf system
(crab cavity) imparting opposite transverse kicks (Sec.4.10.5), by introducing empty phase space
to the bunch head and tail. The rf deflector is in the centre of a coasting beam before bunch-
placed at a point where the betatron phase in the ing [26], by recombining with an empty rf bucket
crossing plane is −π/2 from the IP. In the clas- [27], or by redistributing surfaces in phase space
sic crab crossing scheme another rf deflector af- [27].
ter the collision point is used to restore the tilt. The idea of using a “travelling” waist (focus)
Crab crossing collision, with a single crab cav- to compensate the luminosity reduction due to the
ity per ring, was successfully implemented at the hour-glass effect in circular colliders came from
389
Sec.4.13: COLLISION SCHEMES FOR RING COLLIDERS
linear colliders [28]. In the travelling waist col- CRAB OFF CRAB ON
lision scheme, the optical focal point depends on
the longitudinal position of a particle within the
bunch. In other words, particles with different Vy = 398 Pm Vy = 143 Pm
longitudinal coordinates along the bunch in col-

lision “see” the same and minimal beta functions
when they encounter the center of the opposing
beam. In circular colliders the travelling waist can
be realized by a combination of accelerator com- Figure 1: Transverse beam profiles in DAΦNE with
ponents that provides a transformation described crab-waist on and off.
by a Hamiltonian H = H0 − (zp2y )/2 relating the
longitudinal position z and the vertical momen-

are discussing Super B-factories (SuperB in Italy
tum py . For example, the travelling waist with the
[32] and SuperKEKB in Japan [33]) and a Su-
crab crossing can be obtained by combining crab
per tau-charm factory (Novosibirsk, Russia [34])
cavities and sextupole magnets [15].
with luminosities about two orders of magnitude

The longitudinal strong rf focusing is an al- beyond those achieved at present or past B and
ternative way to obtain short bunches at the IP tau-charm factories, respectively, an e+ e− Higgs
[16]. It consists in realizing a large momentum factory in the LHC tunnel [35], a 500-GeV e+ e−
compaction of the lattice together with a strong collider in a 233-km tunnel [36], and an upgrade
rf gradient. In this regime the bunch length is no of the LHC, called HL-LHC, with a “virtual”
longer constant, but it is modulated along the ring peak luminosity of 2 × 1035 cm−2 s−1 [37], all
[29] and can be minimized at the IP. In turn, if the of which might make use of the crab-waist colli-
main impedance generating elements of the ring sion scheme, or, at least, incorporate some of its
are located where the bunch remains long, it is components.
possible to minimize the strength of wake fields.
This helps avoiding microwave instabilities and
excessive bunch lengthening due to the potential References
well distortion. This concept was proposed as one
[1] ICFA BDN 48, April 2009
possible option for the DAΦNE upgrade [30].
[2] https://cern.ch/lpc
Contrary to the conventional strategy, the crab [3] V.V. Danilov et al, EPAC 96, 1149
waist collision scheme requires small emittance, [4] R.B. Palmer, SLAC-PUB-4707, 1988
large Piwinski angle and larger crossing angle. [5] K. Oide, K. Yokoya, PR A40 (1989) 315
There is no need to decrease the bunch length [6] R. Calaga, R. Tomas, F. Zimmermann, CERN-
and to push beam currents beyond the values 2007-002, 77
already achieved in the present factories. This [7] K. Hirata, PRL 74 (1995) 2228
scheme can greatly enhance the luminosity of a [8] K. Takayama et al, PRL 88, 144801 (2002)
collider since it combines several potentially ad- [9] F. Ruggiero, F. Zimmermann, PRST-AB 5,
vantageous ideas: collisions with a large Piwinski 061001 (2002)
angle, micro-beta insertions and suppression of [10] D. Neuffer, S. Peggs, SSC-063 (1986)
beam-beam resonances using the dedicated “crab [11] F. Ruggiero et al, KEK Proc. 2002-30, p. 131
waist” sextupoles. For details of beam dynamics [12] C. Bhat, F. Zimmermann, IPAC 11, 1879
in crab-waist collision see Sec.2.5.2. [13] P. Raimondi, in 2nd SuperB Workshop (2006)
[14] P. Raimondi et al, physics/0702033 (2007)
The crab-waist collision scheme has been
[15] Y. Ohnishi,

Handbook of Accelerator Physics and Engineering (PDFDrive) PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Handbook of Accelerator Physics and Engineering (PDFDrive) PDF

Uploaded by

Copyright:

Available Formats

Handbook of Accelerator Physics and Engineering Downloaded from www.worldscientific.

This page intentionally left blank

Karl Hubert Mess

British Library Cataloguing-in-Publication Data

HANDBOOK OF ACCELERATOR PHYSICS AND ENGINEERING

Printed in Singapore by Mainland Press Pte Ltd.

Alexander Wu Chao, Stanford, California

Acknowledgments to First Edition

1.4.2 Conversions M. Tigner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.2.1 Single Element Optics K. Brown . . . . . . . . . . . . . . . . . . . . . . . . 71

2.2.4.3 Systematic lattice optimization techniques . . . . . . . . . . . . . . 83

2.4.5.2 Betatron frequency shifts . . . . . . . . . . . . . . . . . . . . . . . 138

2.4.12 Intrabeam Scattering and Touschek Effect V. Lebedev . . . . . . . . . . . . . 155

3 ELECTROMAGNETIC AND NUCLEAR INTERACTIONS 215

3.1.3 Bending Magnet Radiation H. Wiedemann . . . . . . . . . . . . . . . . . . . 217

3.1.4.6 Quantum lifetimes . . . . . . . . . . . . . . . . . . . . . . . . . . . 222

4 OPERATIONAL CONSIDERATIONS 311

4.2.2.3 Wiggler radiation . . . . . . . . . . . . . . . . . . . . . . . . . . 320

4.8.1 Non-Invasive Measurement for Linacs J. Irwin, Y. Yan . . . . . . . . . . . . . 367

4.10.3 Multi-Bunch Manipulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 378

5 MECHANICAL CONSIDERATIONS 399

5.13.2 Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454

5.14.2 Passive Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459

6 ELECTRICAL CONSIDERATIONS 463

7.1.1.5 Normal conducting rf photo guns D.H. Dowell . . . . . . . . . . . 548

7.2.14 Feedback to Control Coupled-Bunch instabilities J.D. Fox . . . . . . . . . . . 628

7.2.18.2 Partial snake H. Huang . . . . . . . . . . . . . . . . . . . . . . . . 642

7.4.11.3 Applications to measurement of beam dynamics . . . . . . . . . . . 725

7.4.16.2 Beam loss monitoring using light detection . . . . . . . . . . . . . . 739

8 RADIATION EFFECTS AND PROTECTION 767

8.4.2 Semi-Empirical Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 779

8.4.9 Skyshine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 785

Author index I-1

Subject Index I-5

1.1 HOW TO USE THIS BOOK NC normal conducting

Abbreviations Tabulated below are symbols adopted through-

Dx,y beam-beam disruption parameters ρ volume density

E0 design particle energy σ interaction cross-section

1.3 FUNDAMENTAL CONSTANTS [1]

Quantity Symbol Value

reduced Planck constant = h/(2π) 1.054572 E-34 J s = 6.582119 E-16 eV s

1.4 UNITS AND CONVERSIONS 1.4.2 Conversions

ous physical quantities from Gaussian to SI unit

speed of light c √1 inch, in. 2.54E-2

horsepower 745.7 Dose, gray, Gy, [L2 /T2 ]

Watt/cm K 1.0E2 Temperature, T , Kelvin, K

abcoulomb 10 erf(x) = √ dt e−t , erf(∞) = 1

1.5.2 Curvilinear Coordinate Systems Spherical:

General Orthogonal System (u1 , u2 , u3 ) [1]: ds2 = dr 2 + r 2 dθ 2 + r 2 sin2 θdφ2

dV = h1 h2 h3 du1 du2 du3 ∂ 

Cylindrical: Steady-state boundary conditions between two

1.5.4 Kinematical Relations

Relations between bending magnetic field B, Variables differential relation

1.5.5 Vector Analysis Energy-momentum :

∇(φψ) = φ∇ψ + ψ∇φ Ez = γ(Ez + cβBy ), Bz = γ(Bz − βc Ey )

Energy-momentum : (Sec.4.10). The rf voltage is lowered so that the

cused to a small spot size on the production target. The

with σabs the absorption cross-section, pt the

Final emit. 6π

Core cooling 2-4,4-8 gen.

Laminated Pole Piece

and find that

(c) Pulse Interval

dV = h1 h2 h3 du1 du2 du3 ∂

∇(φψ) = φ∇ψ + ψ∇φ Ez = γ(Ez + cβBy ), Bz = γ(Bz − βc Ey )

Final emit. 6π