Day 1

Theory and Practice of
Free-Electron Lasers
Particle Accelerator School

Day 1
Dinh Nguyen, Steven Russell

& Nathan Moody
Los Alamos National Laboratory
US Particle Accelerator School 2009

LA-UR 09-01205 University of New Mexico - Albuquerque NM 1-1
Course Content
Chapter 1. Introduction to Free-Electron Lasers
Chapter 2. Basics of Relativistic Dynamics
Chapter 3. One-dimensional Theory of FEL
Chapter 4. Optical Architectures
Chapter 5. Wigglers
Chapter 6. RF Linear Accelerators
Chapter 7. Electron Injectors

Course Schedule
Monday Tuesday Wednesday Thursday Friday
9:00
Intro. to FEL
Optical Wigglers RF Linac Final Exam
Architectures
Relativistic
10:30 Dynamics
10:45
Final Exam
Optical
1-D FEL Theory Wigglers RF Linac
Architectures
Lab Report Due
12:15
1:15
Simulation Simulation Simulation Simulation

Lab Lab Lab Lab
3:15
3:30
Electron
Optical
1-D FEL Theory RF Linac Injectors
Architectures
5:30

Chapter 1
Introduction to Free-Electron Lasers

Introduction to Free-Electron Lasers
• The nature of light
• Gaussian beam
• Laser beam emittance
• Longitudinal coherence
• How a quantum laser works
• How an FEL works
• Basic features of FEL
• RF-linac FEL
• Fourth-generation Light Sources
• Applications of FEL

Light can be described as both
particles (photons) and waves
• Light consists of photons each having energy E  hv where h = Planck’s
constant (h = 6.626 x 10-34 J-s) and  = frequency of the light; v  c
Photon energy can be calculated from wavelength as follows
1.24eV 12.4keV
E E
 ( )  (Å)
• Light can also be described as a travelling electromagnetic (EM) wave.
c B m
c  2.9979 108
E s
We can treat the EM wave as a sinusoidal plane wave. In our convention,
the electric field is in the x direction and magnetic field in the y direction. For
a wave travelling in the positive z direction, the fields are given below
E( z, t )  xE0 cos(kz  t   )
ˆ where k = wavenumber in m -1
= angular frequency in s-1

B( z, t )  yˆ B0 cos(kz  t   ) = phase in radians

Gaussian Laser Beam
I0
rms radius in x

 x2 


I ( x, y ) x 2 dxdy x
 Intensity


I ( x, y )dxdy FWHM
w .135 I0
1/e2 radius -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0
    x    
w  2 x 2 2
x  y
2P 2  2 
I ( x, y )  e  w
e  w
Full width at half max (FWHM) w 2
FWHM  2 2 ln 2 x
FWHM  2.355 x

Gaussian Beam Propagation
Parabolic expansion of 1/e2 radius with z Rayleigh length

 z2   w02
w  w 1  2 
2 2
0 zR 
 zR  
At large z the divergence angle scales with /w0 

 w0
The product of the waist radius and converging angle of Diffraction limit
a diffraction limited beam is the wavelength divided by . 
Focusing the beam to small spots requires large angles. w  
0

Laser Beam Emittance
( x, x) ( x, x)
dx z
photons
x 
dz
x x x
converging waist diverging
x x x
Light phase space area =  times x

(rms radius) times x’ (rms angle) Photon beam emittance
 w0  
Arms   x x '    x x ' 
4 4 4

Longitudinal Coherence
9
210 1
0.8
9
1.510
intensity (W/cm2)
Fourier transform 0.6

f( )
9
g( t ) 110
 0.4

8
510 0.2
0
0
0 4 2 0 2 4
4 2 0 2 4
time (ps) 5  (THz)
frequency 5
t
If  and  are the full-width at half max (FWHM), the transform limit becomes
  0.44
Gaussian pulse
2
2
 t   t 
2
 
 t 
4ln 2 
  
Coherence length Lc 
I  I 0e  I 0e 
An optical pulse with length  is fully coherent if its coherence length ≥ 2 c 

How a quantum laser works
An external source of energy excites electrons from the ground state to an
excited state
Electrons from excited state decay to a metastable energy level with long
lifetime (transition from this level to the ground state is quantum
mechanically forbidden) → population inversion
A co-propagating light beam stimulates emission of radiation → amplification
of co-propagating light beam (Light Amplification by Stimulated Emission of
Radiation)
e
m
g
Absorption Population Inversion Stimulated Emission

How an FEL works
light (electromagnetic wave)
wiggler
magnets
bunched electron beam pulse
Bw v║
electron trajectory
y z
x
v┴
unbunched electron beam pulse
Electrons in an FEL are not bound to atoms or molecules. The “free” electrons
traverse a series of alternating magnets, called a “wiggler,” and radiate light at
wavelengths depending on electrons’ energy, wiggler period and magnetic field.

How an FEL works (cont’d)
The wiggler induces transverse sinusoidal velocity in electron beam
Energy exchange occurs between the transverse electron current and

transverse electric field of a co-propagating light beam
W  e  Es
Depending on the phase of the light beam with the electrons’ wiggling motion,
some electrons gain energy while others lose energy → energy modulation →
bunching of electrons along the axial direction into microbunches with period
equal to an optical wavelength
Microbunched electron beams radiate coherently at higher power →

amplification of the co-propagating light beam.
Note: The subscript ┴ denotes transverse and s stands for signal.

Basic features of FEL
• Wavelength tunable
• Diffraction limited optical beam
• Longitudinally and transversely coherent
• High power (GW peak, 100kW to MW average)
• Efficient (with energy recovery)
Gamma X-rays VUV Visible IR THz mm-wave wave
 0.1nm 1nm 10nm 100nm 1 10 100 1mm 10mm 100mm

Eb 10GeV 1GeV 100MeV 10MeV 1MeV 100keV

Wavelength Tunability
Select coarse wavelength by choosing the electron beam energy,
wiggler period and wiggler magnetic field. Fine-tune wavelength
by adjusting electron beam energy or wiggler magnetic field.
w wiggler period
 resonant wavelength
w
  2 1  aw2 
2
aw (also Krms) rms wiggler parameter
 relativistic factor eB0
aw   0.66 B0 T  w  cm 
For electrons (m0c2 = 0.511 MeV) 2k w m0 c
T
 2
 1  2T  MeV  Another convention uses peak parameter K
m0c eB0 w  K 2 
K  2 aw   2 1  
kw m0 c 2  2 

Radio-frequency Linac FEL
Wiggler
Outcoupler Electron High Reflector

Injector Beam Dump
Booster
FEL
Oscillator
Electron Injector
Energy Recovery Linac
RF Linac
Bunch Compressor
Wiggler
Single-pass Amplifier FEL Beam
Self-Amplified Spontaneous Emission (SASE)
Beam Dump
RF-Linac FEL Pulse Structure
FEL macropulse
Tmacro
FEL micropulses
n tRF
RF wave train
1
t RF 
f RF

Temporal & Spectral Structures
SASE FEL have spiky temporal and spectral features.
SASE
SASE spectral
time domain domain

 

Unsaturated oscillator/amplifier FEL have smooth temporal and spectral profiles.

Oscillator/
Oscillator/
Amplifier
Amplifier
time domain
spectral
domain

FEL optical beam properties
2 N p hv
• Intensity I W/cm2
 w t
2
0
N p hv
• Brightness B W/m2
  x  y t
2
• Spectral bandwidth  1

 Nw
• Brilliance Np
B photons/(m2 s 0.1% BW)
  x  y t 
2

4th Generation Light Source (4GLS)
Injector
at 2-km point
1 km S-band linac
e Transport
Undulator
Experiment Hall
Linac Coherent Light Source (LCLS)
Peak brilliance of linac-based 4th generation light sources (XFEL) is 8-10

orders of magnitude higher than that of 3rd generation light sources and
>20 orders of magnitude above Bremsstrahlung sources.

Some examples of 4GLS
LCLS European XFEL SCSS
Institution SLAC DESY Spring-8
Location Palo Alto, CA Hamburg Hyogo
Country USA Germany Japan
Wavelength 0.15 nm 0.1 nm 0.1 nm
X-ray energy 8 keV 12.4 keV 12.4 keV
Beam energy 14.3 GeV 20 GeV 8 GeV
Linac type, frequency NCRF, 2.856 GHz SRF, 1.3 GHz NCRF, 5.712 GHz
Length 1 km 3.4 km 0.75 km
Gun type, frequency NCRF, 2.856 GHz L-band RF gun Pulsed DC gun
Cathode Cu photocathode Cs2Te photocathode CeB6 thermionic
Bunch charge 0.25 nC 1 nC 1 nC
Bunch length 75 fs 80 fs 250 fs
rms emittance 0.4 m 1.4 m 2 m
Wiggler period 3 cm 3.56 cm 1.5 cm
aw (K) 2.62 (3.7) 2.33 (3.3) 1.3 (1.838)
Length 55 m 200 m 50 m
Peak brilliance of 4GLS
Pulse energy ~ 1 mJ
Photon energy ~ 1 keV
# of photons ~ 1013
rms emittance ~ 10-4 m
rms bunch length ~ 10-13 s
Energy spread ~ 0.01% BW
Brilliance ~ 1033 (s m2 0.1% BW)-1

High-average-power FEL
• Ground-based FEL Program (Boeing/LANL, LLNL/TRW)
• Energy-recovery FEL (e.g. Jefferson Lab FEL)
Jefferson Lab FEL holds the world record in cw average power (14 kW).

Applications of FEL and 4GLS
FEL Features Wavelengths Examples of applications
• Ultrashort tunable pulses

– Medicine 1-6 m Laser surgery
– Physics XUV Ultrafast spectroscopy
– Chemistry XUV, UV Chemical dynamics
– Biology X-rays Protein structures
• High peak power

– High-density physics X-rays Warm dense matter
– Materials sciences near-IR Laser machining
• High average power

– Directed energy IR Defense
– Space near-IR Power beaming
– Material processing UV Lithography

Chapter 2
Basics of Relativistic Dynamics

Basics of Relativistic Dynamics
• Special relativity
• Lorentz transformation
• Relativistic Doppler shifts
• Wavelength dependence on angle
• Relativistic velocity, momentum & energy
• Lorentz force law
• Curvilinear coordinate system
• Linear beam dynamics
• Emittance
• Emittance & energy spread requirements

Special Relativity
1. All inertial frames are completely equivalent with
regard to physical phenomena
2. The speed of light in vacuum is the same for all

observers in inertial frames of reference.
e- beam v
y’ y
Beam Frame Lab Frame
z’ z
x’ x

Lorentz Transformation
y’ Beam coordinates
y
c
x'= x Lab
coordinates
z z’
y'= y
x x’
z ' = g ( z - b ct ) Lorentz factor g=
1
1- b 2
ct ' = g (ct - b z ) Velocity
u
relative to c b=
Transverse dimensions are unchanged. c
Lengths of moving objects along direction of motion appear to be
contracted in the Lab frame by a factor  (Lorentz-FitzGerald contraction)
Clocks in the moving objects run slower by  as observed in the Lab
frame (time dilation).

Wiggler period contracts in beam frame
w
Lab frame
y
x w
Beam frame Wiggler period in
beam frame
y’
' lw
z’ l =
g
w
x’

Lorentz Transformation of Fields
Electric field Magnetic field
Ex' = g ( Ex + u z By ) æ u ö÷
B = g çç Bx - 2 E y ÷÷
' z
x çè c ø
E y' = g ( E y - u z Bx )
æ u ö÷
B = g çç By + 2 Ex ÷÷
' z
'
E = Ez
z
y çè c ø
Bz' = Bz
Transverse electric and magnetic fields are different in the beam frame.
Pure electric (and magnetic) fields in the Lab frame transform into mixed
electric and magnetic fields in the beam frame. Longitudinal (along the
direction of motion) electric and magnetic fields remain the same.

Electromagnetic Field Transformation
Lab frame Wiggler magnetic field deflects electrons in x direction
y F = -e (v x B)
v
B
z
Force is due to magnetic field in Lab frame
x
Beam frame Electromagnetic field deflects electrons in x’ direction
y’ F’ = -e (E’ + v’ x B)
B’
z’
E’
v’ ~ 0 in beam frame
x’ Force is almost entirely due to electric field

Radiation in Beam Frame
Beam frame Wiggler electromagnetic wave

behaves like virtual photons
y’ impinging on the electrons
B’
z’
E’
x’
View from the top Real photons are scattered off the electrons. They
can also be seen in the beam frame as circular
waves radiated from the electrons at frequency ’
c gc
n'» =
lw' lw
Lorentz contraction causes ’ to be increased
by a factor of compared to Lab frame

Useful Relativistic Relations
Exact relations Approximations for  ~ 1
1 1
2
g = b » 1- 2
1- b 2 2g
1 1
2
b = 1- 2 1- b » 2
g 2g
2 2 2 1 1
b g = g -1 » 1+ 2
b 2g
1 1
= 2 -1 1 1
2 2
b g b -1 » 2
b 2g
Lorentz Transformation of
Frequency and Angle
Relativistic Doppler shift depends on Lab frame observation angle
n'
n ' = g (1- b cos q ) n n=
g (1- b cos q )
Backward ( = ) Forward ()
1
Use approximation 1- b »
n' 2
2g
n=
2g Relativistic Doppler shift
in the forward direction n = 2gn '
For  >>1 Lorentz transformation yields 1/ emission angle
1- cos q = g 2 (1- b cos q ) (1- cos q ' )

2
1/
For small angles 1- cos q '

q=
g

Longitudinal Doppler Shift (Forward)
Beam frame Doppler effect causes 
y’ up-shift in frequency and
narrowing of emission angle
n = 2gn '
z’
Consider radiation emitted in the forward

x’ æ g c ö÷
direction (same direction as electrons) n = 2gn ' = 2g ççç ÷÷
Lab frame è l ÷ø
w
y 2g 2 c
n=
lw
z
Combined effect of Lorentz contraction and Doppler
lw
l= 2
x shift gives a factor of 2 increase in frequency 2g

Longitudinal Doppler Shift (Backward)
Beam frame
Doppler effect causes 
y’ down-shift in frequency
n'
z’ n=
2g
x’ Consider radiation emitted in the backward
direction (opposite to beam direction)
n' 1 æç g c ö÷ c
Lab frame n= = çç ÷ = ÷
y 2g 2g è lw ÷ø 2lw
z Lorentz contraction is negated by Doppler

shift. Frequency is reduced by a factor of 2.
x l = 2lw

Wavelength Dependence on Angle
The wavelength of wiggler (undulator) radiation depends on emission angle.
Shortest wavelengths are radiated in the forward direction ( = 0). Radiation
at larger angles have longer wavelengths. The opening half angle of wiggler
radiation,  is given by
2 aw
q=
w g
lw
l = lw (1 + b ) l » 2 (1 + g 2q 2 )
2g
Backward wave Forward wave

Relativistic Energy & Momentum
2 2 2
Total energy E = T + m0 c = mc = g m0 c
Kinetic energy
T = (m - m0 ) c 2 = m0 c 2 (g -1)
Momentum p = mu = bg m0 c Energy is in unit of MeV or GeV.
Momentum is in unit of MeV/c or GeV/c
(cp ) = b g (m0c )
2 2
= (g -1)(m0 c )
2 2
Multiply by c and square 2 2 2 2
E = (cp ) + (m0 c )
2 2
Energy right triangle 2 2
moc2 E
cp

Parameter Variation Table
d dp d
 p 
d 1 1
 1
  2
 
2 2
dp 1
  2
1
p 2
d
  2 2 2 1


Relative velocity differences
become smaller at high energy
1
0.8
db 1 dg
0.6
= 2 2

b b g g
0.4
0.2
0
1 2 3 4 5 6 7 8 9 10
Most electron accelerators are speed-of-light (=1) machines

At large , it becomes very hard to perform ballistic bunch compression
because all electrons travel nearly at the speed of light.

Relative momentum change is
same as energy spread at high energy
12
10
1 dg
p (moc)
dp
6
= 2
4 p b g
2
0
1 2 3 4 5 6 7 8 9 10
Bunch compression via momentum spread can be done at any energy

Given sufficient energy spread and dispersive elements such as magnetic
chicanes, electron bunches can be compressed to ultrashort pulses.

Lorentz Force Law
F  e E   v  B  
In MKS units, e = 1.6 x 10-19 coulomb, electric field is in

volts/m and magnetic field is in tesla.
Electric force acts on electrons along their direction of motion
and thus changes the electrons’ kinetic energy.
T   F  ds    eE  ds
Magnetic force is perpendicular to direction of motion and
does not change the electrons’ kinetic energy. Magnetic field
can be used to change momentum, i.e. bend electron beams.
 
p   Fdt  e    B  dt

Bending Relativistic Beams
dipole magnet
Bend angle and radius incident beam
s
p^ 
tan q = p║
p bent beam
p^ = F Dt = -eu BDt = -eBDs  p┴
bend radius
Eb
p = bg m0 c =
c
Ds -ecBDs
tan q = = 1 B(T )
r Eb

 m   299.8
1
Eb  MeV 
1 ecB
= Magnetic rigidity
r Eb
1
B  T  m   Eb  MeV 
299.8

Curvilinear Coordinate
y y s
Trajectory of
reference particle
x
x
Electrons travel in the s direction. Use (x, y, s) coordinate system to follow
the reference electron, an ideal particle at the beam center with a curvilinear
trajectory. The reference particle trajectory takes into account only pure
dipole fields along the beam line. The x and y of the reference trajectory are
thus affected only by the placement and strength of the dipole magnets.
For other electrons, define x’ and y’ as the slopes of x and y with respect to s
dx dy
x  y 
ds ds

Paraxial Rays & Trace Space
Paraxial ray approximation deals with non-crossing trajectories near the axis.
( x, x)
z
x’ x’ x’
x x x
In a drift space, converging beams come to a waist and then diverge

Lorentz Forces
Lorentz force in x Lorentz force in y
d 2x d2y
 m0 2  e By  ecBy  m0 2  e Bx  ecBx
dt dt
Slope of x with respect to s
dx 1 dx
x  
ds c dt
Curvature of x with respect to s
d 2x 1 d 2x
x  2  2 2
ds c dt
ecBy eBy eBx eB
x   y   x
 m0 c 2
p  m0 c p

Quadrupole Lens
A quadrupole is a focusing element in one plane (e.g., x) and defocusing in
the other plane (e.g., y). Its magnetic field, and thus the focusing force,
increases linearly with distance from the center. .
y
Quadrupole Quadrupole field
Before quadrupole Quadrupole focusing After drift

x’ x’ x’
x x x

Linear Beam Dynamics
Linear beam dynamics is valid if the restoring forces in x and y are linear.
Quadrupoles are linear focusing (and defocusing) elements since the
restoring forces are linear with distance from the center.
Mathieu-Hill Equations
x  K x x  0 y  K y y  0
A system of alternating focusing and defocusing quadrupoles separated
by drift space (abbreviated FODO) is used to transport electron beams.
Rx
Ry

Phase space concept
x x x
x x x
converging waist diverging
Beams are treated as a statistical distribution of particles in x’-x (also in y’-y

and -ct) phase space (trace space, to be exact). We can draw an ellipse
around the particles such that 50% of the particles are found within the ellipse.
The area of this ellipse is a measure of rms spread of electron distribution in
phase space. The rms emittance is area of the ellipse divided by . Emittance
has dimension of length (e.g. microns) since x’ is dimensionless. Traditionally,
emittance has unit of mm-mrad.
A
 rms 


Beam Emittance
Root-mean-square x emittance (for y emittance, replace x with y)
2
 rms , x  x 2
x2  xx
Emittance is defined using ensemble averages, denoted by < >, of x2 and x’ 2

and x’-x correlation. The correlation vanishes at the waist (upright ellipse) and
rms beam emittance becomes xx’ where  x  x 2 is the rms radius in x
and  x  x2 is the rms spread in x’.
Ensemble average of x2 Ensemble average of x’2 x’-x Correlation

1 1 1
x  x0    x   x  x0  xj
N N N
2 2
x 2
 j x 2
 j 
xx  j
N j 1 N j 1 N j 1

Liouville’s Theorem
Liouville’s theorem : In the absence of non-linear forces or acceleration, the
beam ellipse area in x-px phase space is conserved. If the forces acting on the
beam are linear, its emittance is also conserved.
x  px  const.
If the beam is accelerated, emittance (defined by x and x’) is not a conserved
quantity because x’ decreases as the axial momentum increases by .
px x’ = px/pz px x’ = px/pz
accelerated
pz pz
By accelerating the beam (increasing pz), we reduce the “un-normalized”
emittance (also known as Lab frame emittance). The conserved quantity is the
normalized emittance, un-normalized emittance multiplied by . Normalized
emittance is used to specify the quality of electron beams regardless of energy.
 n   u

Electron Beam Emittance Requirement
Electrons’ phase-space area must be less than photons’ phase space
area for efficient energy exchange between electrons and photons
photons
x x x
electrons
x x x

Ae  u 
4
Accelerating the electron beam reduces its
un-normalized emittance (adiabatic damping). n
Beams with large (bad) normalized emittance u 
need to be accelerated to high energy.

At a fixed wavelength and beam energy, the 
required normalized rms emittance for FEL is n 
4

Energy Spread Requirement
Electron beam’s energy spread must be smaller than the electrons’
velocity spread over the interaction length.
 1

For oscillator FEL, interaction length ~ wiggler length  2Nw

For SASE and amplifier FEL, interaction length ~ gain length 

Uncompressed electron beams have small energy spread and low peak
current. Compressed beams have high current and large energy spread.
 
 ct ct 

Chapter 3
1-D Theory of FEL

One-dimensional Theory of FEL
• Transverse motion in a wiggler
• Figure 8 motion and harmonics
• Pendulum equation
• FEL bunching
• Bunched beam radiation
• Spontaneous emission spectrum
• Madey’s theorem
• Low-gain FEL
• Synchrotron oscillation
• Saturation
• Extraction efficiency
• High-gain FEL
• Self-consistent FEL equations

Equations of Motion

ˆ 0 cos  k w z 
By  yB
y
By
z
x
vz 2
kw 
w
Lorentz force laws
 
x  e  zˆ z  yˆ  Bo cos  k w z 
Fx   m0 
 
z  e  xˆ x  yˆ  Bo cos  k w z 
Fz   m0 
For most FEL, vx is much smaller than vz . We can ignore the second
force equation and consider only motion in x (the wiggle plane).

Equations of Motion (cont’d)
Small-angle approximation: transverse motion is small; axial velocity is almost c
x  c Transverse velocity Transverse acceleration

z   c  x   z x d
x   z x    z2 x

dt
Consider only on-axis magnetic field
Second derivative of x with respect to z
B( z )  Bo cos  k w z 
d 2 z 
x 
x
x  2  2  2
Lorentz force equation dz z c
d 2x e Rewrite Lorentz force equation in term of
x  2   z Bo cos  k w z 
 second derivative with respect to z
dt  m0
eBo
x   cos  kw z 
 m0 c

Solution to Transverse EOM
Integrate Lorentz force equation once to obtain deflection angle
eBo x’0 = initial deflection angle

x    cos  kw z  dz x0 = initial position
 m0c
Integrate again to obtain position
eBo  2aw 
x   sin  kw z   x0 x     sin  kw z   x0 dz
 kw m0 c   
2 aw 2aw
x   sin  kw z   x0 x cos  kw z   x0 z  x0
  kw
Transverse motion is periodic with wiggler wavenumber kw. Wiggler
magnetic force is harmonic oscillator’s restoring force. Transverse motion
in the absence of field errors is given by
Wiggler wavenumber
x  k w2 x  0 2
kw 
w
B field, deflection and position
Wiggler magnetic field Bo cos  kw z 
+ error - error
2 aw
First integral of field (deflection)  Bo cos  kw z  dz  
sin  k w z 
 2 aw
Second integral of field (position)
 
 Bo cos  kw z  dz dz  kw
cos  kw z 

Transverse and longitudinal velocities
Transverse velocity is oscillatory with period equal to the wiggler period
2caw vx v = c
x   sin  k w z 
 vz
Longitudinal velocity Find the square root and use small x
 z2   2 c 2   x2 approximation (1 + x)½ ≈ 1 + ½ x
é 1 ù
 2a 2
 u z = c ê1- 2 (1 + 2aw sin (kw z ))ú
2 2
 z2  c 2   2  w
sin 2  kw z   êë 2g úû
  2

Use sine squared identity
2 1 2aw2 
 z  c 1  2  2 sin  kw z  
2 2
2sin 2 (k w z ) = 1- cos (2k w z )
   
Axial velocity oscillates with a period equal to one-half the wiggler period
 1  aw2  a 2 
 z  c 1   2 cos  2kw z   caw2
 z   z  2 cos  2kw z 
w
 2 2
2  2
 

Figure 8 Motion
In the reference frame that travels at the electrons’ average axial velocity,
vz as given by
z
 
1  aw2 
 c 1 

Motion in reference electron’s rest frame
 2 2 
 
Electrons’ transverse and axial motions are
coupled. At zero crossing, transverse speed
is at a maximum and axial speed a minimum.
At the edges, transverse speed is zero and
axial speed is at a maximum. Electrons’
motion on the x-z plane follows the figure 8.
2caw
'  sin  kw' z ' 
x

caw2
 z  ' 2 cos  2kw' z ' 
'
2kw Figure 8 motion gives rise to harmonics

in spontaneous (incoherent) radiation
Energy exchange between
electrons and FEL beam
dW
  j  Es
dt
Transverse electron current Plane-wave transverse electric field
j  ecx  
ecaw
sin  kw z  Es ( z, t )  Es ,0 cos(kz  t )

d   m0 c 2  ecaw E0
 sin  kw z  cos(kz  t )
dt 
Rate of energy exchange depends on the phase of the “ponderomotive wave”
d   m0 c 2 
sin   kw  k  z  t 
ecaw E0

dt 2

Resonance Condition
Question: How can an optical wave traveling at the speed of light interact with
slower electrons in a fast wave device (e.g., FEL)?
Answer: If the optical wave slips ahead of the electrons exactly one wavelength
every wiggler period, the sum of wiggler phase and optical phase is constant,
and energy exchange can occur.
   kw  k  z  t  const. k  1  aw2 
kw  k   k k 2 
d   1  aw 2
 2  
 kw  k   0 1 
z 2 
dz  2  
Resonance wavelength satisfies this condition

w    w c
z
 1  aw2   1  aw2 
kw  k  2 
  w  2 
 2    2  
w
Ponderomotive phase = -/2
kwz = 0 j
kz - t = -/2 dW
0
(kw + k)z – t = -/2 dt Es
Electrons gain energy
kwz =  Es
kz - t = -3/2
(kw + k)z – t = -/2 j
Electrons gain energy (light is absorbed) j
Optical wave slips ahead by  every w

Es

Ponderomotive phase = 0
kwz = 0 j
kz - t = 0 dW
0
(kw + k)z – t = 0 dt Es
No energy gain or loss

kwz =  Es
kz - t = -
(kw + k)z – t = 0 j
No energy gain or loss

j
Optical wave slips ahead one 
Es

Ponderomotive phase = /2
kwz = 0
j
kz - t = /2 dW
0
(kw + k)z – t = /2 dt Es
Electrons lose energy
kwz = 
Es
kz - t = -/2
(kw + k)z – t = /2 j
Electrons lose energy (FEL gains energy)

j
Optical wave slips ahead by  every w
Es
Ponderomotive Wave
The electrons interact with the so-called ponderomotive wave with
frequency  and wavenumber kw + k. The ponderomotive wave is
synchronous with the resonant electrons, i.e. those at the zero phase of
the ponderomotive wave. The ponderomotive phase velocity,  divided
by kw + k, is slightly less than the speed of light. The phase of the
ponderomotive wave is defined by average arrival time of the electrons
   kw  k  z   t where k
2

Taking derivative with respect to z
d 
  kw  k  
dz z
Average electron axial velocity
 1  aw2 
 z  c 1  2 
 2 

Phase Equation
Evolution of phase along the wiggler
d  
  kw  k   1 
1  a 2
w  
  kw  k
1  a w
2
dz c  2 2
 2 2
 
Using the definition for resonance condition in k space kw 
k
2 R2
1  a 2
w 
d  R2
 kw  kw 2
dz 
Define an energy difference relative to the resonant energy R
  R R 
2
  
 1    1  2  
R R 
  
 R 
The phase of individual electrons evolves
d   
along the wiggler according to their energy  2k w  
difference relative to the resonance energy dz 
 R 

Energy Exchange Equation
eEs ,0
Define a dimensionless signal field parameter, as as 
km0 c 2
Energy exchange rate depends on the phase of electrons in the
ponderomotive potential. Electrons with phase between –and 0 gain
energy. Electrons with phase between 0 and  lose energy.
d ckas aw
 sin 
dt 
Rewrite the above equation in terms of derivative with respect to z of the
energy difference relative to the resonant energy, R
The energy of an electron relative
to the resonance energy evolves d    kas aw
according to the sine of its phase in     2 sin 
the ponderomotive wave
dz   R  R

Coupled First-Order
Differential Equations
Evolution of relative energy difference and phase along the wiggler
d    kas aw
Rate of energy gain/loss along z     2 sin 
dz   R
 R
Rate of phase change along z d   
 2k w  
dz 
 R 
Define new variables,  and a
  2kw  angular phase
Pendulum equations
    = angular velocity
v
 R 
 v   a sin 
|a| = height of potential well
  v
kas aw
a    2
2
R  = oscillation frequency

Hamiltonian System
Hamiltonian mechanics is useful in representing beam physics because it
relies on something being conservative. In the case of a pendulum, the
conserved quantity is the total energy of the system of two canonical
conjugate variables , the angular momentum, and , the angular phase.
Hamiltonian = Total energy
2
H  a cos 
2
Kinetic energy Potential energy
Hamiltonian equations
H  H
     a sin    
 

Pendulum Equations
v   a sin 
  

  

 v   a sin 
Coupled non-linear 1st order differential equations
  v
Particles rotate clockwise in  phase space as the rate of change of  is
proportional to -sin  and the rate of change of  is . Particles follow
elliptical trajectories each of which corresponds to a constant energy.
Higher energies occupy larger ellipses up to phase angle of ± .

Small-angle Solutions

  

 Second-order differential equation

Small-angle approximation, i.e.
sin ~  leads to harmonic v    2    2  0
solutions with oscillation
frequency , square root of |a|   v and its solution
  sin  0 
The small-angle oscillation frequency is known as
the synchrotron frequency 0. The synchrotron 1
frequency is proportional to the square root of W0 = kas aw
dimensionless optical field (fourth root of intensity). gR

Large-angle Close-orbit Solutions
 
 
 

Solutions corresponding to large-angle oscillations can be solved numerically.
The large-angle oscillation frequency is lower than the small-angle synchrotron
frequency and approaches zero at  = ± Oscillation frequency is given by
p
W 2
=
where K : elliptic function. W0 æ æ z öö
K ççsin 2 çç 0 ÷÷÷÷÷÷
çè èç 2 øø
W z 02
Oscillation frequency for initial angle  up to  » 1-
W0 16

Separatrix

  

Motion at the two nodes, , vanishes. These are unstable equilibrium
points, corresponding to the pendulum at the top. The separatrix is the
boundary separating trapped and un-trapped trajectories. The region inside
the separatrix is called the “bucket.” The bucket height is proportional to the
square root of the optical field (fourth root of optical intensity).
Separatrix for a uniform wiggler Bucket half-height
v  0 2  cos   1 as aw
vmax 
1  aw2
Laser Field and Bucket Height
eEs ,0
Dimensionless optical (signal) field parameter, as as 
km0 c 2
The electric field of the FEL beam depends on
the optical intensity and free space impedance Es ,0  2 Z 0 I L
Z 0  377
Laser intensity depends on power and mode radius 2 PL
IL 
 w02
X-ray FEL at 1.5 Å
Peak power (W) 1.5 x 1010
Intensity (W/cm2) 5 x 1014 Bucket half-height
Electric field (V/m) 6 x 1010 as aw
as 3 x 10-6  max 
1  aw2
max 1 x 10-3

Open Orbits

 
 

Motion has large angular velocity. The pendulum rolls over the top and
librates about the pivot point. The corresponding phase space trajectories
are not elliptical. These represent un-trapped electrons outside the
“bucket.” The un-trapped electrons also provide FEL gain. The electrons
at small phases near the top of the “bucket” flow down into the “troughs”
and lose energy to the optical field. As the optical field grows, the bucket
also grows in height and eventually capture these electrons.

Synchrotron Oscillation Animation
Phase Space
20
20
10
 PHSP 
2 0 
Energy
 w.1  i
0
 10
 20
 20
2 2 2 2 2 2 4 4 4 4 4 40 6 6 6 6 6 62 8 8

.51  
.75
.6
.422
.365
.3  PHSP
  
.1 
1 
i
2.6 
2.422
2.365
2.75
2.5
2.3
Theta

Synchrotron Oscillation Animation
Phase Space
60
50
 PHSP 0 
2
Energy
Change scale  w.1  i

0
 50
 60
0 00 2 2 2 22 4 4 4 44 6 6 6

.3 .2
..15   
.15  PHSP
  
.1 
1 
i
2. 
2.15
1.85
2.3
1.8
Theta

Microbunching
The FEL interaction causes the electrons to gain or lose energy, depending
on their ponderomotive phase. Electrons with positive ponderomotive
phase lose energy and migrate to the bottom of the bucket. Electrons with
negative ponderomotive phase gain energy and move to the top of the
bucket. The resulting energy modulation causes the electrons to develop
density modulation with period of the radiation wavelength. The bunched
electrons radiate higher power, i.e. it amplifies the electromagnetic wave.
As the electric field of the electromagnetic wave increases, the height of the
bucket also increases. When the electrons are completely bunched, FEL
power is saturated. Microbunching is responsible for harmonic generation
(the Fourier transform of short bunches has high frequency components).
Courtesy of S. Reiche

Radiation from bunched beam
w  E  Ne
(a)
I incoherent  N e
Nw 
E  Ne
(b)
I coherent  N e2
lb lb lb
Nw 
z
Electrons at the wiggler entrance are randomly distributed (a). Randomly

distributed electrons radiate incoherently, i.e. the electric fields of Ne randomly
distributed wave trains with Nw (Nw is the number of wiggler periods and  is
the wavelength) add incoherently. The total electric field is proportional with
square root of Ne. The spontaneous radiation intensity scales with Ne.
Near saturation, the electrons are bunched into microbunches with bunch
length z less than radiation wavelength (b). The electric fields of Ne wave
trains scales with Ne, and the coherent radiation intensity scales with Ne2.

Spontaneous Emission
Spectral and angular energy fluence of spontaneous emission radiation from a
planar wiggler as a function of frequency detuning from resonance condition
2
2 æ sin D ö÷
dW e 2 g 2 N w2 N e 2æç aw ö÷ çç 2 ÷÷
= [ JJ (aw )] çç 2÷
÷ çç ÷
d wd W 2pe0 c è1 + aw ø÷ çè D ÷ø÷
Difference between J0 and J1 Bessel functions
JJ (x ) = J 0 (x ) - J1 (x ) aw2 0.25
x=
Approximation for small  2 (1 + aw2 )
2
2  sin  
é JJ (x )ù » 1- x - x 2f ( ) 0.125 Dw 1
 =
ë û Frequency detuning   
 
w Nw
2 4

  2 N w
 0
 20  15  10 5 0 5 10 15 20


Spontaneous emission is peaked at zero detuning (resonant wavelength)

Spontaneous Emission (cont’d)
Consider only photons within coherent spectral bandwidth and solid angle

Coherent spectral bandwidth Coherent angle Solid angle

Dw 1 l pl
= q= pq 2 =
w Nw Lw N wlw
Number of coherent spontaneous photons per electron does not depend on Nw
2
N photon 2æç aw ö÷
= pa [ JJ (aw ) ] çç 2÷
÷
Ne è1 + a ø÷ w
where  = fine structure constant

e2 1 For typical values of aw, on average we need
a= » 200 electrons to generate 1 spontaneous
c 4pe0 137 photon within coherent angle and bandwidth

Madey’s Theorem
Madey’s Theorem: The small-signal gain spectrum (gain versus energy detuning)
for a low-gain FEL is the derivative of the spontaneous emission spectrum. The
small-signal gain is positive (amplification) at positive detuning, zero on resonance
and negative (absorption) at negative detuning.
3
4 (4pr N w ) æ D ö÷
g ss (D) = ç
ççè1- cos D- sin Dø÷÷
3
D 2
0.5 Maximum gain is at = 2.6

 E  DE 2.6 1
  4 N w   = »
 E  E 4p N w 5 N w
gss()
g(  ) 0
Maximum gain occurs at positive energy

detuning (higher energy) than resonance, or
at a fixed energy, longer wavelength.
 0.5
 10  7.5 5  2.5 0 2.5 5 7.5 10
3
 g max » 2 (2pr N w )

ss

Visualization of Madey’s Theorem
On resonance R No gain or loss
1
R 2Nw
Positive detuning R Amplification

1
5Nw
Negative detuning R Absorption
1
5Nw

Small-Signal Gain
The small-signal gain for a planar wiggler at the peak of the gain curve,
assuming the electron beam radius  is smaller than the optical beam, is
2
æ 2p N w ö÷
3
æ[ JJ ] aw ö÷ æI ö÷
g ss = çç ÷÷ ççç ÷÷ çç
çè I
÷÷
èç g ø çè s k ÷ø w Aø
÷
2p
where k w =
lw Fundamental
Power versus z inpower
a low-gain FEL
Pout
and IA (Alfven current) = 17 kA
gss scales with z3
gssPin
Peak Power (W)
(W)
Small-signal gain in a low-gain
FEL is proportional to Nw3 Peak Power
Pin
Pout = (1 + g ss ) Pin
0 0.06 0.12 0.18 0.24 0.3 0.36 0.42 0.48 0.54 0.6
Wiggler zlength
(m)
(m)
1h

Large-Signal Gain
FEL gain is reduced when optical intensity approaches the saturation intensity,
2
1 mc æç ö÷ 4
1 3 æ g ö÷
IS = ç
ç
÷÷ ççç ÷÷
8p s çè[ JJ ] awlw ÷ø èNwø
÷
Large-signal gain
g ss Small-signal gain
g (I ) =
æI ö÷
(W) 2)
1 + çç
(W/cm
÷÷
Large-signal gain
çè I ÷
S ø
Peak Power
Saturation Intensity
Peak Intensity
At high intensity, more electrons Peak Intensity
reside at the bottom of the bucket
and FEL gain decreases. Saturation
intensity is the intensity at which FEL z (m)
Wiggler Length (m)
gain reduces to one-half of gss.

Synchrotron Oscillation
Energy and phase equations Synchrotron period
d    kas aw
    2 sin  w 1  aw2
dz   R R
 S 
d 2
d   
 2k w 
2 as aw

dz 2
dz   R 
Plot of power vs z showing synchrotron oscillations
2nd-order differential equation
of phase evolution with z S
d 2
Peak Power (W)
 S sin   0
2
2
K
dz
Synchrotron oscillation wavenumber
2kw kas aw as aw
KS   2k w
 R2 1  aw2  0 0.4 0.8 1.2 1.6 2
z (m)
2.4 2.8 3.2 3.6 4

Extraction Efficiency
Wiggler length ~ synchrotron period
w 1  aw2
Lw  S  1
2 as aw  2Nw
 
w
Lw 
2 max

w
 max  At saturation, the wiggler length is about the same
2 Lw as a synchrotron oscillation period. The electrons
rotate to the bottom of the “bucket.” The bucket
half-height is inversely proportional to 2Nw.
1
 max 
2Nw
High-Gain FEL
Dimensionless Pierce parameter as a function of kw (left) or w (right)
2 1 2 1
1 æç[ JJ ] aw ö÷ 3 æI
çç
ö÷ 3 1 çæ[ JJ ] awlw ö÷ 3 æ ö
çç I ÷÷
3
r= çç ÷÷ ÷÷ r = çç ÷÷
2g çè s kw ÷ø çè I ÷ g çè 4 2ps ÷ø çè I ÷÷ø
Aø A
Recall JJ is the difference between J0 and J1 Bessel functions of argument 

[ JJ ] = J 0 (x ) - J1 (x )
x x2 where
[ JJ ] » 1- -
x2 x 2 4 aw2
J 0 (x ) » 1 - J1 (x ) » x=
4 2 2 (1 + aw2 )
High gain FEL is applicable in a long wiggler driven by a high-brightness
electron beam (one with high peak current and small emittance). The wiggler
length must be significantly longer than the power gain length, given by
Power gain length w
LG 
4 3
Power Growth in High-Gain FEL
Power vs distance Natural log of FEL power vs
Fundamental z (wiggler length)
power
1  z 
P( z )  P0 exp  Lsat
Psat
9  LG 
Synchrotron
Log Power
Saturation length Exponential growth
Oscillation
 9 Psat 
Lsat  LG ln  
 0 
P
P0
Saturation power
0 0.4 0.7 1.1 1.4 1.8 2.2 2.5 2.9 3.2 3.6
IEb
Psat   z (m)
e 1h
Power grows exponentially with distance by one e-folding (2.7) every power
gain length. Starting from noise, the FEL saturates in 20 power gain lengths.
FEL saturation power, Psat, is approximately  times the electron beam power.

Slowly Varying Envelope Approximation
So far, we’ve considered only the electron phase-space motion. To be complete,
we must write self-consistent FEL equations for N electrons and the optical field.
We’ll treat the optical field as a slowly varying phasor (ignoring the optical
frequency oscillation). The phasor’s amplitude is the usual dimensionless optical
field as. This is known as the Slowly Varying Envelope Approximation (SVEA).
Optical electric field with fast oscillations SVEA phasor

E (t )  E0 exp i (kz  t   ) a = as e-ifs
Wave equation without the fast time scale terms (e.g. 2nd order derivatives)
(4p ) [ JJ ] æç I ö÷ éê e-iq 1 ùú
3
é¶ 1 ¶ù
ê + úa= çç ÷÷ ê aw - ia
êë ¶z c ¶t úû 2kSb è I A ÷ø ë g g úû
The electron bunch is assumed to be many wavelengths long, so the beam
current density is assumed to be independent of z over many wavelengths.

Self-Consistent FEL Equations
Evolution of the jth electron’s phase and energy
d j k
 k w  2 1  aw2  2aw as  JJ  cos  j 
dz 2 j
d j kaw as
  JJ  sin  j
dz j
Evolution of optical phasor’s phase and amplitude
d fs (4p ) æç I ö÷ éê aw [ JJ ] cos q 1 ùú
3
= çç ÷÷ ê -
dz ÷
2 k Sb è I A ø ë a s g g úû
3
das ( ) aw [ JJ ] æç I ö÷ sin q
4 p
= çç ÷÷
dz 2 k Sb è I A ÷ø g
The <cos> term corresponds to the real part of the e-beam’s susceptibility
(refractive index) and <sin> term corresponds to the imaginary part (gain).

Scaled Variables
• Scaled axial position z

t=
Lw
• Dimensionless current density
3
j = 2 ( 4p N w r )
• Scaled phasor equation
da
= - j e-iq
dt
t
da j
= ò t a (t - t ) e
' ' -in 0 t '
dt'
dt 2 0

Cubic Equation
Take the derivative of the last equation successively
d 3 a (t ) ja (t )
3
=-
dt 2
Assuming solutions are of the form ei and at resonance condition, we obtain
the characteristic cubic dispersion relation
j
l3 + =0
2
Note:  are roots of the cubic equation, not wavelength
Solutions of the cubic equation are of the form
a (t ) = a0 eilt

Solutions to Cubic Equation
Three roots of the cubic equation 1
Im  j 3 1 i 3
Complex root 1      
3 2  2 2 
2
1
-1 Re Complex root  j 3 1 i 3
½ 2      
2  2 2 
1
Real root
 j 3
3    
2
Solutions in electric field
1
i 
j 3
  j  3   3   j  3   3   j 13  i 3  
1 1

22
E ( ) 
E0 e  
 e 2   2   e 2   2   e 2   2  
3  
 growing decaying oscillatory
 mode mode mode 

Exponential Growth
In the limit of large z, only the growing mode needs to be considered.
The optical field vs scaled length is given by
1 1
i  j   3  j 
3 3
2 2  2  2 
E ( )  13 E0 e  
e  
Multiplying the electric field by its complex conjugate yields the FEL
intensity versus the scaled length 
2 1

E0   3
exp  3   
2 j
E ( ) 
9  2 
 
Plug in the expressions for  and j, we arrive
at the expression for intensity vs. distance in
the wiggler. This equation gives the I0  4 3 z 
I ( z )  exp  
 w 
exponential growth with wiggler length and 9
the initial 1/9 reduction in signal intensity.

References
Books and Articles
Free-Electron Lasers C.A. Brau
Particle Accelerator Physics I & II H. Wiedmann
Physics of Free Electron Lasers E.L. Saldin, E.A. Schneidmiller
and M.V. Yurkov
“Free Electron Lasers” S. Khan (2008) J. of Modern Optics, 55:21,3469 – 3512
“Development of X-ray Free-Electron Lasers” C. Pellegrini and S. Reiche (2004)
J. Quantum Electronics, 10(6) 1393-1404
URL
UC Santa Barbara WWW FEL http://sbfel3.ucsb.edu/www/vl_fel.html
Linac Coherent Light Source http://www-ssrl.slac.stanford.edu/lcls/
European XFEL http://xfel.desy.de/


Day 1

Uploaded by

Document Information

Original Description:

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

Day 1

Uploaded by

Copyright:

Available Formats

Theory and Practice of

Particle Accelerator School

Dinh Nguyen, Steven Russell

US Particle Accelerator School 2009

US Particle Accelerator School 2009

Simulation Simulation Simulation Simulation

US Particle Accelerator School 2009

US Particle Accelerator School 2009

US Particle Accelerator School 2009

= angular frequency in s-1

US Particle Accelerator School 2009

Full width at half max (FWHM) w 2

US Particle Accelerator School 2009

Parabolic expansion of 1/e2 radius with z Rayleigh length

Light phase space area =  times x

US Particle Accelerator School 2009

Fourier transform 0.6

US Particle Accelerator School 2009

US Particle Accelerator School 2009

US Particle Accelerator School 2009

Energy exchange occurs between the transverse electron current and

Microbunched electron beams radiate coherently at higher power →

Note: The subscript ┴ denotes transverse and s stands for signal.

US Particle Accelerator School 2009

Gamma X-rays VUV Visible IR THz mm-wave wave

 0.1nm 1nm 10nm 100nm 1 10 100 1mm 10mm 100mm

US Particle Accelerator School 2009

US Particle Accelerator School 2009

Outcoupler Electron High Reflector

Single-pass Amplifier FEL Beam

Self-Amplified Spontaneous Emission (SASE)

US Particle Accelerator School 2009

Unsaturated oscillator/amplifier FEL have smooth temporal and spectral profiles.

US Particle Accelerator School 2009

US Particle Accelerator School 2009

Linac Coherent Light Source (LCLS)

Peak brilliance of linac-based 4th generation light sources (XFEL) is 8-10

US Particle Accelerator School 2009

Photon energy ~ 1 keV

rms emittance ~ 10-4 m

rms bunch length ~ 10-13 s

Energy spread ~ 0.01% BW

Brilliance ~ 1033 (s m2 0.1% BW)-1

US Particle Accelerator School 2009

• Ground-based FEL Program (Boeing/LANL, LLNL/TRW)

• Energy-recovery FEL (e.g. Jefferson Lab FEL)

US Particle Accelerator School 2009

• Ultrashort tunable pulses

• High peak power

• High average power

US Particle Accelerator School 2009

US Particle Accelerator School 2009

US Particle Accelerator School 2009

2. The speed of light in vacuum is the same for all

US Particle Accelerator School 2009

US Particle Accelerator School 2009

US Particle Accelerator School 2009

US Particle Accelerator School 2009

Lab frame Wiggler magnetic field deflects electrons in x direction

US Particle Accelerator School 2009

Beam frame Wiggler electromagnetic wave

US Particle Accelerator School 2009

1- cos q = g 2 (1- b cos q ) (1- cos q ' )