i

Name Symbol Name Symbol Name Symbol

Actinium A, Hafnium Hr Praseodymium P,

Aluminum
Ame ri cium
Antimon"
Argon
Arsenic
-
AI
Am
Sb
M
A.
Helium
Holmium
Hydrogen
inclllJln
Jodine
H.
Ho
Ii
I"
I
Promethium
Protactinium
Radium
Radon
Rhen ium
..
Pm
Po

R"
R,
Astatme At Iridium h Rhodium Rh
, Banum B. Iron F. Rubidium Rb
Berkelium Bk Kryp ton K, Ruthenium R..
Beryllium Be Lanthanum Lo Samarium Sm
BI L,
Bismuth
....00 B
B,
Lav.'Tcncimn
Lo.J Pb
Scandium
Selenium
Silicon
"S.SI
~
Bromine Lithium
Cd Silver
Cadmium
Calcium C.
Lutetium
:-'lagnesiurn \lg Soc!ium "S,N.
Californium Cf \Ianganese \ 111 Strontium
Carbon C \ Ienaele\ jum \Id Sulfur S
Cerium C. \1en:urV Hg Tantalum T,
Cesium C. \loIybdenum \10 Techne ti um T,
Chlorine CI Keodymium Nd Tellurium T.
Chromium C, Kron !'\e Tl:rbium Tb
Cobalt Co "'''ptuniurn :\p Thallium 11
Copper C.. Klcke\ No TII(lrium Th
Curium Coo r\iobiurn Nb TIlLl];lJIn Tm
Dysprosium Dy "'I trogen N Tin Sn
Einstein ium E. ;':obelium :\0 Titanium Tt
Erbium E, Osmium 0. Tungsten W
Oxygen Uranium U
EurOIJium
Fermium
F1uorillC
Ell
Fm
F
PalladIUm
Phosphorus
°PPd Vanadium
Xenon
V
X,
Francium F, Platinum PI Ytterbium Vb
Gadolinium Cd Plutonimn p" Yttrium V
Gallium C, Polonium Po Zinc Z"
C. Zirconium Z,
Germanium
Cold .,"
Potassium
"

---=- . . . . - ~-- .. .
.,...~-
... -_. ..
 H' HT
Periodic Table, with the Outer Electron Configurations of Neutral
I., Atoms in Their Ground Siaies h'
U' Be' The nolation u\cd to describe the elc(:lwnic configuration of' ~lt()m' B' C· N' 0" F' Ne'o
,md ioo, i' di~cu'~cd in all textbook., of introductory alomH; phy_ie"
The letter~ $, fl. d, .. .\ ignify cledron~ having orhital angular
0, ",' 2.\.2211 2.~~2pl 2s72p~ 2s'2p' 2.1:'2,,5 2.,'2p l
momentum D, I, 2, .. . ill unih -t\; the numher to the lei! of the
, Na il Mg " letter denute, the pTlncipaJ quantum numher ot Olle urbit, and the AP] SI" P" 5" CJ'1 At "
,uper~cript to the right dcnole~ the nl1mllcr IIi' electroll~ in the orhit

3, 3I.·'3p 3s'3p' ~"3p' :1.~'31'· 3.,'3p' 3.I:'3p'

K" 'CaN
" Se" Tj irl V" Ct" Mn" Fe"" Co" Nili' Cu" Z,' Ga" Ge" As]1 Se H Br·l~ K,'
3d 3d' :JrJl 3(/5 3d' 3d" 3d' 3d" 3d'~ 3el l "
4,\" 4.s! 4,\,! 4.," 4,~l 4.' 4." 4.' 4,' 4.~2 4, 4.' 4s 2 4p 4s'4 p 2 4s14p'l 4.r4,,· 4s 24p' 4s'4p'
Rb 17 5,- Y" Z,· Nb" Mon Tc" Au" Rh'l Pd'· Ag<l Cd" In" 5," Sb" Te u I" Xe o•

,d 4cP 4d' 4d' 4d' 4d' 4d" 4d " 4d'" 4d"

s.' 5., 5, 5, 5, - 5., 5.~2 58 25p 58 251" 5,~25p , 5.15,)' 58 25p' 5.~25p l
5," s.'
5,
CS~I 8a l - La l l HIll '"
Ta 7'] W" Re fS OS'I Ir71 Pt" Au" Hg" n"' Pb"' Bi") Po" At'S An'"
4f"
5d 5d' S,P 5d' 5d' 5d' Sri" Sd' Sel'· 5d'~
fu
Fr"'
""Ra" fu'
Ac"
fu'
I'
"'. .,' 6,'
"" - fu 6, 6.' ""6p 00 261'2 6s'6p·1 6s 26p' OO 26p· 6s 26p'

Ce- Pro' Nd" Pm" Sm u E," Gd Tb" Oy' Ha"l E..• Tm" Yb7~ LU ll
6<1 4f' 4f' 4f' 41' 41' 4/, 41' 4f' 4fl~ 4fll 4fIJ 4/" 4fH 4/"
7, 7.\.'1 7~ 5d 5d 5d
6.,.2 6.~2 fu' 6.' ",' .,'
""Cm" fu'
6.\,1

Th" Pa'l U" Np·

""p,' "" Am'! Sk'] CI"
""Es" Fm'"
"" ""
Md 'o, NolO! Lr '"
- Sf' S!", 5f' Sf' Sf' Sf'
6(P 6d 6d 6d
7._2
- ---..:.:-.
7 .•2 7~·1 7 .•'
--- 7,·2 7.v2 .
L:C.. __
7,.2
-
)
 )

CHARLES KITTEL

Introduction
to
Solid State
Physics \

SEVENTH EDITION

John Wiley & Sons, Inc., New York, Chichester,

Brisbane, Toronto, Singapore

7
 J 5.1.1
- \Li-r
/ '>410 o[)::
C c7
About the Author
Charles Kittel taught solid state physics at Berkeley from 1951 to 1978;
earlie r he was a member of the solid state group at the Bell Laboratories. His
undergraduate work was at M.tT. and at Cambridge University, followed by
graduate work at the University ofWiS<..'Onsin. He is a member or tile National
Academy of Sdem;e and of the American Academy of Arts and Sdern.:es.
His research in solids began with studies of ferromagnetic, antiferromag-
netic, and paramagnetic resonance, along with work on magnetic domllins, spin
waves, and domain boundaries in ferromagnets and ferroelectrics. His work on
the single domain strudure of flne particles has Imd broad llpplil~ltion in mag-
ne tic re«>rding, geomagnetism, and biomagnetism. Along with collaborators at
Berkeley he did the fi rst work onl.'yciotron resonance in semi<.'onductors, which
led to the understanding of the band structure of silicon, germanium, and in-
dium antimonide, together with the theory of their impurity states. He also
worked on the interpretation of magnetoplasma resonance in scmieonductors
and of A]fven resonance in eleetron-hole drops in germanium.
The frrst edition of [SSP integrated the elementary aspeds of solid state
physics for study by seniors and beginning graduate shldcnts. Now in its sev-
enth edition, ISSP plays the same part for the current gencmtion of students.

Copyright C 1953, 1956, 1966. 1971, 1976, 1966, 1996 by John Wiley &; Sons, Inc.
All righu resen·ed. Published simultaneously in Canada.
Reproduction or t ranslation of any part of this work heyond that pcnniUed hy Section. 107 or lOS
of the 1976 United States Copyright Act without the permission of the copyright owner is
unla"fu l. Requests for permis~ion or furth"r information shouk! be addressed to the Permissions
Deparbnent, John Wiley &; s.:.os. Inc.

U brary of Congress Cataloging in PubilcD/lon Data:

Charles Kittel.
Introduction to solid state physics I Charles Kittel. - 7th ed.
p. cm.
Includes index.
ISBN 0-411_1118 1-3 (cloth: alk. paper)
I Solid state (ilysicS. L Title.
QC I76.KS 1996
530.4' I-dc20 05-18445
e lP
Printed irl the United States of AfIlt!rica
15 14 13 12 11
 Preface

This hook is the seventh edition of an elementary text on solid state physics
fOf senior and beginning graduate studen ts of physical science and engineering.
The book is an update of the sixth edition of 1986 and includes additions,
improvements, and (:orrections made in that edition in 13 successive printings-
which it was time to pull together-and a num ber ofnew topics besides. Signn-
i<.:ant ad\'am:.~ in the field have been added or discussed more fully: thus 11igh
temperature superrondu<.iors are treated, and results of scanning tunneling
rniQ"OS(.'OPY are displayed; the treatme nt of fiber optics is expanded. nlere are
di5(.'lssions. among other topics, of nanostrudllres, supcrlattices. Bloch!
Wannier levels, Zener hmneling, light-emitting diodes. and new magnetic
m.1teria1s. The additions have been made within a boundary (.'()nditi(m intended
to keep the text within one volume and at a reasonable price.
The theoretical level of the text itself has not been changed. There is more
diS<.1lssion of useful materials. TIle treatment of elastic constants and elastic
waves which was dropped after the fourth edition has now been returned be-
cause, as many Ilave pointed out, the matter is IIseful and not easily accessible
elsewhere. TIle treatment of superconductors is much more extensive tllan is
IIsual in a text at this level: either )-'OU do it or you don'!.
Solid state physics is concerned witll the properties, often astonishing and
often of great utility, that result from the distribution of eledrons in metals,
semiconductors. and insulators. TIle book also tells how the excitations and
imperfections of real solids can be understood with simple models whose power
and scope are now firmly established. The subjed matter supports a profltable
interplay of experiment, application. and theory. The book, in English and in
many translations, has helped give several generations of students a pil.1ure of
tile process. Students a1so flnd the field attractive because of the frequent possi-
bility of working in small groups.
lnstn](;tors will lise the book as the foundation of a course in their own
way. yet there are m 'O general patten"]s to the introduction. scledion and order
of the basic material. If students have a signiAcant preparation in elementary
quantum meek. D. they will like to begin willI the quantum theory of clec-

ii,
 )
trOlls in one-dimensional solids, starting with the frcc electron gas in Olaptcr 6
and energy bands in Chapter 7. One will need to treat the rcciprocallattice in
three dimensiOns (Chapter 2) before plunging into semiconductors (Chapter 8)
and Fermi smfaces (Chapter 9). Crystal stOlchITCS, crystal binding, and pho-
nons ("ould be considered as recreational reading. In a more gradual approach.
the first eight chapters through the physics of semi<:onductors are read consel.'U-
lively as a One-semester introdudion to the field.
What abollt the necessary statistical mechanics? A vague discomfort at the
thought Qf the chemical potential is still charaetcristic of a physics education.
TIlis intelledual gap is due to the obs<"urity of the writings of 1- Willard Gibbs.
who discovered and understood the matter 100 years ago. Herbert Kroemer
and I have darif'ied the physics of the chemical potential in the early chapters of
our book on thermal physics.
Review series give excdlent extended treatments of all the subjects
treated in this book and many more besides; thus with good conscience I give
few references to original papers. In these omissions no lack of honor is in-
tended to those who first set sail on thesc seas.
Thc crystallographic notation conforms with (.·urrent usage in physics.
Important equations are repeated in 51 and CCS-Caussian units, where these
differ. Exceptions arc figurc captions, chapter summaries, some problems, and
any long sedion of text where a single indicated substitution will translate from
CCS to 51. Chapter COntcnts pages diS<.'USs (.'Olwentions adopted to make paral-
lei usage Simple. The dna! usagc in this book has been found useful and accept-
able.
Tables are in conventional units. The symbol e denotes the charge on the
proton and is positive. The notation (18) refers to Equation (18) of thc (.'Urrent
chapter, but (3.18) refers to Equation 18 of Chapter 3. A caret - over a vector
refers to a unit vedor. Few of thc problems are exactly easy; most were dc\<ised
to carry forward the subjed of the chapter. With a fcw exceptions, the prob-
lems are those of the orib'inai sixth edition.
This edition owes much to the advice of Professor Steven C. Louie. For
collected corrections, data, and illustrations I am grateful to P. Allcn, M. Beas-
ley, D. Chemla, T.-C. Chiang, M. L. Cohen. M. C. Craford, A. E. Cur£on,
D. Eiglcr, L. M. Falimv, R. B. Frankel , 1- Friedel, T. H. Ceballe, D. M.
Cinsberg, C. Herring, H . F. Hess, N. HolonY.lk, Jr., M. j1lmb, J. Mamin,
P. McEuen, J. C . Mullen, J. C. Phillips, D. E. Prober, Marta Puebla, D. S.
Rokhsar, L. Takacs, Tingye Li, M. A. Van Hove, E. R. Wcber. R. M. Wllite,
J. P. Wolfc, and A. Zeul. Of the Wiky staff I have particularly great debts to
Clifford Mills for publication supervisiOn, to Cathy Donovan for her ingenuity
in processing the additions between the thirteen successive printings, and to
Suzanne Ingrao of Ingrao Associatcs for hcr skill and understanding during the
editoriul prO<.'Css.
 )
Corrections and suggestions will be gratefully received and may be ad-
dressed to the autilOr at the Department of Physics, University of California,
Berkeley, CA 94720-7300; by email to kiue l@uclillk4.l3erkelcy.cdu; and by fax
to (510) 643-9473.

C. KiHel

An Instructor's Manual is available for this revision; several problems have

been added (to Chapter 3 and Chapter 6); one dropped (from Chapter 4), and
several corrections made. Instructors who have adopted the lext for classroom
use should direct a request on departmental letterhead to Jolm Wiley & Sons,
Inc., 605 Third Avenue, New York, NY 10158-0012. Limited requests for- per-
mission to copy figures or other material should be addressed to the Permis-
sions Editor at tllis address.
 Contents

Cuide to Tables i.

Ceneral References .i

1 CRYSTAL STRUCrURE 1

2 REC1PROCAL LArnCE 27

3 CRYSTAL BiNDING AND EUsnC CONSTANTS 53

4
5
6
PHONONS I. CRYSTAL VIBRATIONS

PHONONS U. THERMAL PROPER'flES

FREE ELECfRON FERMI CAS

...
91

141

7 ENERGY BANDS 113

8 SEMICONDUCfOR CRYSTALS 191

9 FERMI SURFACES AND METAL't 2"

10 PLASMONS, POURITONS. AND POlARONS 269

11
12
OPTICAL PROCESSES AND EXOTONS

SUPERCONDUCnvITV
"'"
333

13
14
15
DlELECfRICS AND FERROELECTRICS
DlAMAGNETISM AND PARAMAGNE'fl5M ...
310

.s....
FERROMACNE'fl5M AND ANTIFERROMACNE'fiSM 441

16 MAGNETIC RESONANCE

17 NONCRYSTALLINE SOLIDS

18 POINT DEFFrvrS .39

"",
viii

(
19 SURFACE AND INTERFACE PHYSICS
20 DISLOCATIONS
""
58'
21 ALWYS 609

Appendix

A TEMPERATURE DEPENDENCE OF TilE nEFLECIlON UNES 631

B EWAlD CALCULATION OF LATIlCI': SUMS 634

C QUANTIZATION OF ELASTIC WAVES, PHONONS 638

D FERMI-DIRAC DlSCR1BUTlON FUNCTION 643

E DERIVATION OF dkJdt EQUATION 646

F BOLTZMANN HlANSPORT EQUATION 647

G VECrOR POTENTIAL, FIELD MOMENTUM, AND GAUGE

TRANSFORMATIONS 651

H COOPER PAIRS
I GINSBURG-LANDAU EQUATION ...
656

J ELECTRON-PHONON COLUSIONS 002

Subject Index 667

Table of SI Prefixes 673
 (

Guide to Tables

1.I Fourteen lattice types in three dimensions 10

I.'
1.3
Characteristics of cubic lattices
Crystal Struchll"C$ of the c1emCflts
12
23
1.4 Density and atomic cooccntration of the clements 24

3.1 Cohcsi\'c energies of the elements 57

3.2 Mdting temperatures of the elements 58
3.3 Hulk modulii and oompressibilities of the clements 59
3.4 Properties of inert gas crystals 60
3.5 Ionization energies 61
3.6 Electron affinities of ncgati,'e k.tns 68
3.7 Properties of alkali halide crystals 73
3.8 Fractional ionic character of bonds 76
3.9 Atomic and ionic radii 78
3. 10 Standard radii of ions 79
3. 11 t:lastic stiffness rorutants of cuhic crystals at 0 K and 300 K 91
3. 12 t:lasHc sti ffness constants of cubic crystals at 300 K 92

5. 1 Dchyc temperatuTe and thermal conductivity 126

5.' Phonon mean free paths 133

6. 1 Free electron Fermi ~-urface parnmctcn for metals 1S<l

6.' Electron heat capaci ty of metals 157
6.3 Electrical oondocthi ly and J"e$istivity of metaL,; 160
6.4 !lall coefficients 167
6.5 Lorenz numbers 168

8. 1 Energy gaps in semiconductors 201

8.2 Elfccth'C masses of holes and e lectrons 21'
8.3 Carrier mobi liti~ at room temj)er.l.ture 221
8.4 Stalk dielectric constants of semkonductOrlii 223
8.5 Donor ionization energies 224
8.6 Acceptor ioni7.ation c~es 225
8.7 Electron and llOlc concentrations in scmimctals 229

;,
•
10. 1 Ultraviolet transmi~sion limib of alkali metals 275
10,2 Volume plasmon energies 278
10,3 Lattice frequencies 292
lOA I'olaron ma:sscs and coupling COI1~lants 298

11.1 Exciton (,lulling energies 314

11.2 Electron-hole liquid pm'1lmctcJ'li 322
11.3 Acronyms of current expelimcntal methods for Land structure studies 328

12. 1 Superconductivity parameters of the clements 336

1~2 Superconductivity of selected l'Ompounrls 336
12,3 Energy gaps in superconductors 34.
12.4 Isotope effect in supcroonducton 347
12.5 Coherence length and penctratjon depth 353

13.1 Ekctronic polari:vlbilitics of ions 391

13.2 Ferroelectric crystals 396
13.3 Antiferroclcctric crystals 406

14. 1 Magneton numbers of L"1nlhanide group kllls 425

14.2 Magneton numbers of iron group ions 426

15. 1 eritia! point c.>;poncnts of fcrromagncts 445

15.2 Ferromagnetic: crystals 4.9
15.3 Antifcrromab'llctk crystals 465

Hi. I Nuclear magnetic resonance data 489

16. 2 Knight shifts in NMR 5()2

18. 1 Diffusion cons tants and <lCI; ivation cnergies 545

18. 2 Activation energy for a prn;ith'c ion vac:.\ncy 547
18.3 Bpcrimental F center absorption energies 549

19.1 'lop layer atomic TClaxation 558

19.2 Eiedn.m work functions 56.

20.1 Elastic lim it and shear modulus 588

20.2 Dislocation dcmities 598

21.1 E lectron/atom ratios of electron compounds 616

(
 Selected General References

S'aljstical ph!llii~ background

C. Kittel aud H . Kroemer, Thermal plJys1cs, 2nd 00., freeman, 1980. Has II [u11,
d ear discussion of tile chemical potential and of semiconductor statistics; cited as
TP.
Intermediate len
J. M. Ziman, Principle. of the I/,oory of soIitU, Cam bridge, 1972.
Adoonud tertI
C. Kittel, Qmmtum Iheory of 101ids, 2nd rc\iscd printing. Wiley, 1987, with solu·
tions appendix by C. Y. Fong; ciled as QTS .
j. Callaway, Quantum Iheory of the solid stale. 2nd ed., Acadcmk. 1991.

Applied solid Ii'ale

It Dalvcn, Introdlwlioll to applied liOlld state ph~ia. 2nd cd., Plenum, 1990. A
readahle Introduction to representative areas in a vast field.

Review senn
F. Seitz and others, ~l/d state ph~ks, adconce, il1 reselJrch olld appllcatfOrl$,
Vols. 1-(48), plus supplements. This valuable rontinu!ng series is often cataloged as
II serial. as if it were II journal. and is cited here as Solid state physics.

Literature guidtl
There are many good databases, libraJ)' and organizational, this is the way to go for
monograph and Journal searches.

"
(
 1
Crystal Structure

PERIODIC ARRAYS OF ATOMS 3

Lattice translation l't.'C:tors 4
Basis and the crystal structure 5
Primitj\'c lattice cell 6

FUNDAMENTAL TYPES OF LATTICES 8

Two-dimt.>JlsionaJ lattice types 8
'nm..'C-dimClls;onallallice types 10
INDEX SYSTEM FOR C RYSTAl. PLANES lZ
SIMPLE CHYSfAL STRUCrURES 15
SOOium chloride structure 15
Cesium chloride structure 17
He~ooal close-packed ~truclurc 17
Dian)()nd structure 19
Cubic zinc 5\LUidc structure 20
DIRECT IMAGING OF ATOMIC STRUCrURE 20
NONlDEAL CRYSTAL STRVe-fURES 21
Random stacking and poIytypism 22

CRYSTAL STRUcrURE DATA 22

SUMMARY 25

PROBLEMS 25
I. Tetram-dral 811g1C$ 25
2. Indices of planes 25
3. Hcp structure 25

REFERENCES 25

UNITS, I A= 1 angstrom - 10- 8 em = 0.1 nm - 10- 10 m.

 (.) <h)

<,)
Jo'igure I Relation of the ~d.m...J kino of Ct)'Stals to the bm ofl~ ~leme"laty building bIods_
TIle buildinli: blocks arc identical In (a) . ok! (b), but d;ff"re"nt <T>... ~I &.:cs are Je..~Iopcd. (e) CIea,"·
Inp; a "'j..... 1 cf ~ I.

,
 CHArTER 1: CRYSTAL STRUCTURE

Solid state physics is largely concemed with crystals and electrons ill crys-
tals. TIle study of solid state physics began in the early years of this century
following the discO\'ery of x-ray diffraction by crystals and the publication of a
series of simple ca1culations rola successful predictions of tile properties of
crystals.
When a crystal !:,trows in a constant environment, the fonn develops as if
identical building blocks were added continuOlisly (-Fig. 1). The bUilding blocks
are atoms or groups of atoms, so that a crystal is a three-dimensional periodic
array of atoms.
111is was known in the 18th century when mineralogists discovered that
the index numbers of the directions of all faces of a crystal are exact integers.
O nly the arrangement of identical particles in a periodic array can account for
the law of in tegral indices, I as discussed below.
In 1912 a paper entitled "Inte rference effects with Rontgen rays" was pre-
sented to the Bavarian Academy of Sciences in Munich. In the first part of the
paper, Laue developed an elementary theory of the diffraction of x-rays by a
periodic array. In tile second part, Friedrich and Knipping reported the first
experimental observations of x-ray diffraction by crysta1s.2
The work proved decisively that crystals are composed of a per iodic array
of atoms. Willi an established atomic model of a crystal, physicists now could
til ink much further. The studies have been extended to include amorpholls or
noncrystalline solids, glasses, and liquids. The wider field is known as con-
densed matter physics, and it is now the largest and probably the most vigorous
area of physics.

PERlODIC ARRAYS OF ATOMS

An ideal crystal is constnlcted by the infinite repetition of identical stmc-

tural units in space. In the simplest crystals the stnLctural un it is a single atom,
as in copper, silver, gold, iron, aluminum, and the alkali metals. But the small-
est stmctural unit may comprise many atoms or molecules.
The structure of aU crystals can be i:l.escribed in lenns of a lattice, with a
group of atoms attaclLed to every lattice point. The group of atoms is called the I.
basis; when repeated in spaCe it forms tile crystal structure.

'R. I. ~l a"'Y, £.sai .fUM thtone sur Ia $/n,cture des cristoux, Paris. 1784; TroiM de cri$lal-
iographie. Parl~. 1501.
"For penonal ac-roU'l" of the early years ofx-ray diffraction stud;es of cr),stals. see P. P. Ewald.
ed .• Fifty yean of x_m , [raction. A. Oosthoek's Uitge,"e rsmij .• UIrC<.11t. 1962.

3
•
Lattice Trllfl.!llation Vectors
The lattice is defined by three fundamental translation vectors 81, 82. ~
such that the atomic arrangement looks the same in every respect when viewed
from the point r as when viewed from the point
(1)

where lit. tl2. 113 are arbitrary integers. The set of points r' defined by (1) for all
tll> 1/2. u3 defines a lattice.
A lattice is a regular periodic array of points in space. (The analog in two
dimensions is called a net, as in Chapter 18.) A lattice is a mathematical abstrac-
tion; the crystal structure is formed when a basis of atoms is attached identically
to every lattice point. The logical relation is
lattice '+ basis"'" crystal structure (2)
The lattice and the translation vectors 8j, 82, ~ are said to be primitive if
any two points r, r' from which the atomic arrangement looks the same always
satisfy (1) with a suitable choice ofthe integers Ill> 1i2, 1/3' With this definition of
the primitive translation vectors, there is no cell of smaller volume that can
serve as a building block for the crystal structure.
We often use primitive translation vectors to define the crystal axes. How-
ever, nonprimitive crystal axes are often used when they have a Simpler rela-
tion to the symmetry of the structure. ·The crystal axes a), a2, a3 fonn three
adjacent edges of a parallelepiped. If there are lattice points only at the corners,
then it is a primitive parallelepiped.
A lattice translation operation is defined as the displacement of a crystal by
a crystal translation vector
(3)

Any two lattice points are connected by a vector of this form.

1b describe a crystal structure, there are three important questions to
answer: What is the lattice? What choice of aj, 82. 83 do we wish to make? What
is the basis?
More than one lattice is always possible for a given structure, and more
than one set ofaxes is always possible for a given lattice. The basis is identified
once these choices have been made. Everything (including the x-ray diffraction
pattern) works out correctly in the end provided that (3) has been satisfied.
The symmetry operations of a crystal carry the crystal structure into itself.
These include the lattice translation operations. Further, there are rotation and
reflection operations, cruled point operations. About lattice points or certain
special points within an elementary parallelpiped it may be possible to apply
rotatiOns <11"1 reflections that carry the crystal into itself.
Finally, there may exist compound operations made up of combined trans-
lation and point operations. Textbooks on crystallography are largely devoted to
 I Cryol<l/ St~!ldure 5

~"
. ~., ~

Figu.e 2 1'<Jrlion of a c~'$tal of an imaginary prote in loo\eC\Ile, in a tW<HlifllC'n~ional "orld. (We

picked a prote in molecu~ because it is not likely to have a special symmriryol" its own.)The atomic
ammgemt'nt in the t.T}'ilal lookll Cl<8ctly tk SlImt' 10 an OOlerver at r' as 10 an OOM'ner at r ,
pro,~ded that th c ,·cdorT whid, t.'OOn«.·b r ' and r may be c~pressed as an integral multiple of the
\ooon a , and ••. In tloi. iI1""tral ion. T ... - a, + la,. ·floe ,"«tors., and a, are pl·imiti'"(" trwosla-
lion vectors of the tu"O-dimt'nsiOOlli iatti("f:.

0 •
0
• • • "

C •
0
0 0
Figure 3 Similar to Fig. 2, but \Ioith protein moleculeli owodatcd in Pllin. 1lw: crystal ,.a" dation
a,. A rotatlo" of 1T radi."s about an)· point marked )( ,,·ilI carry the cl).",1 into
n :don a ..... , and
i'....lf. "mis QCC:tln
also for equi\'alent points in ot her celis. but we have marked the points x onl)"
"ithin one cell.

the description of symmetry operations. 11lC crystal structure of Fig. 2 is drawn

to have only translational symmetry operations. The crystal structure of Fig. 3
allows both translational and point symmctry operations.

Bflsi:s and the Crystal Stntcillre

A basis of atoms is attached to every lattice point, with (..'Vcry basis identical
in composition, arrangement, and orientation . Figure 4 shows how a crystal
structure is form ed by adding a hasis to every lattice point. The lattice is indi-
cated b y dots ill Figs. 2 and 3, but in Fig. 4c the dots arc omitted.
6

Figure 4 ',1'e cry~ tal structure is formed

by the addition of the bru;is (b) 10 .. very
latti~'" point ofth .. latlice (a). By JooI<ingat
(e), you can feC<Jgni7.e the bru;is and then
you can absl~1 the space lattice. II docs
nol matler ",here the basis is put in 1"c~1-
lion 10 a lanke point.

The number of atoms in the basis may be one, or it may be more than one.
The po:;ition of the center of an atom j of the basis relative to the associated
lattice point is

(4)

We may arrange the origin, which we have called the associated lattice point. so
that 0 :S Xj' I)j' ZJ :S L

l'rimitice utttice Cell

The parallelepiped defined by primitive axes Rh a2. 33 is called a primitive
cell (Fig. 5b). A primitive cell is a type ofrell 0T unit cell. (The adjective unit is
supcrlluou5 and not needed.) A cell will fill all space by the repetition of suita-
ble crystal translation operations. A primitive cell is a minimum-volume cell.
There are many ways of choosing the primitive a.xes and primi.t ive cell ror a
given lattice. The lIumber of atoms in a primitive cell or primitive basis is
always the same ror a given crystal structure.
\
 J Crylt(ll StrlOCturfI 7

• • • • • • • • • •
•
[] . "~. ~ .

• ., / 7r~--:7 "0-
~, 'J
<;
, ; ; ;r,
; ,
,
"

• • • • • ~ w
"
• •

o
-
•• •• ••
0,
~,
•• •• ••
"
.e:::: -D - --"
•• •• •• ,

OJ "~I

Figure Sa Lattice pointli of Q space latt ice in two dimensions. All pain of "ecton: a " a, a ....
translation VCl'lCln of the lalli~. But a,''' , ~'" are not primitivc tran)\atton ~fttorJ becau~ we
cannot rofln the lattice ttanslation T from integral combinatio", oh, ,OJ and ..' ''. All othe.- pain
siw;MTI of I, and a, fM)' be takea IllI the primit;'.., tn".. \ation ' "!!don of the lattice. nlC parallelo-
grams I, 2. 3 are equal in area and any of tile," wuid be taken lUi the primitive «>11. The parallelo-
gram 4 ha$ mice the area of a primitivc cell.

rogure sb Primili\'c cd! of a space lattice in three dirnenliions.

Fig....., 50: Suppose tt-..ose points are Mkntical atOlllS . • kctch in On thefigu~ ald oflallice points. a
cho;ce of primitivc a:ces, a primiti~'e .,..,11, and the basis of atoms associated with a lattice point.

There is always one lattice point per prim itive cell. If the primitive cell is
a parallelepiped with lattice points at each of the eight comers, eaeh lattice
point is shared among eight cells, so that the total number oflaltice points in
the cell is one: 8 x • "" 1.
The volume of a par-.tllelc pipccl with a.xes ai, a2. a3 is
(5)
by elementary vector analysis. The basis associated with a primitive ceil is
called a primitive basis. No basis contains fewer atoms than a primitive basis
contains.
8

Figure6 A primil;,·c cell may also be chosen follO\<-·

i"~ this procedure. (\ ) draw line! 10 amnc;-ct II givt"n
lattice point 10 all nearby lattice points. (2) at the
midpoint and normal to these lint'll. draw new lines
Or pla nes. The smallest volum .. end()l;ed in this way
is the \\'igne r-Seil>: primiU\-e <:0"11 . All spaot" may be
filled br t!>eSt" ~'t'IIs . just as by the cells of I:";g. 5.

Another way of choosing a primitive cell is shown in Fig. 6. This is known

to physicists as a Wigner-Scitz cell .

•' UNDAMENTAL TI'PES OF LATIlCES

Crystal lattices can be carried or mapped into themselves by tIle lattice
trdnslations T and by various other symmetry opcmtions. A typical symmetry
operation is that of rotation about all axis that passes through a lattice point.
Lattices can be found such that one-, two-, three-. four-, and sixfold rotation
axes carry the lattice illlo itself, corrcspollding to rOtations by 21T, 2m2 , 2m3,
21T/4, and 2m6 radians and by integral multiples of these rotations. The rotation
axes arc denott..a by the symhols 1, 2, 3, 4, and 6.
We cannot find a lattice that goes into itself under other rotations, such as
by 21Tn radians Or 2nl5 'ddians. A single molecule properl)' dt.'Signcd can have
any degree of rotational symmetry, but an infinite periodic lattice cannot. We
can make a crystal from molecules that individually have a fivefold rotation axis,
but ,",,'C should not expect the lattice to have a fivefold rotation axis. In Fig. 7 we
show what happens if we try to construct a periodic lattice having fivefold
symmetry: the pentagons do not fit together to fill all space, showing that
we cannot combine fivefold point symmetry with the required translational
periodicity.
By lattice point group we mean the colk>etion of symmetry operations
which, applk><l about a lattice point, C"drry the lattice into itself. The possible
rotations have been listoo. \Ve can have mirror rellections m about a plane
through a lattke point. The inversion operation is composed of a rotation of 1T
followed by reflection in a plane normal to the rotation axis; the total effect is to
replace r by - r. 'l1)e symmetry axes and symmetry planes of a cube arc shown
in Fig. 8.

Two-DimensiolUd Lattice lypes

llICre is an unlimitoo number of possible lattices because there is no natu-
ral restriction on the lengths of the lattice translation vectOrs or on the angle If>
hetween them . The lattice in Fig. 5a was drawn for arbitrary 8 1 and az. A
general lattice such as this is known as an oblique lattice and is invariant only
under rotation of 1T and 21T abou t any lattice point.
 1 Cryotal Struclure

Figu.e 1 It f1w:fold "-~i$ of ~)'I)lmctl')' can-

not exilil in It periodic lattice ~ It is
not IlOi!Iibl.. to fill t~ artII of a p~ ,,11h •
connected array of pentagons. We urn, how-
~-er. IlII all 100 area or a plane " ith just t-o.'O
distioct deliignl of ~til..,~ ...... elementar}'
pol)·go,,$. It quasicrystal is a qUll5iperiodic
nonrandom a..semblyof",,"() t)"pcI offlgur~
Quasier}'Sla1s are diKuual at the end of
Chapter 2.

(-(

figure 8 (II) A plane of s~mmetry panlld to the ~ of. cu~. (b) It ~1 plane of $ymmetry
in aeub!-. (c)Thc thro:e tetrad IUeli ofa t'l.l~. (d ) ihe four triad axes of a cubo,o. (e) Th .. l ix diad lL~es
of a t'l.lbc.
 But speciaJ lattices of the oblique type can be inva riant unde r rotation of
21T13. 27T/4. or 21T16. or under m irror renection. 'Ve must impose restrictive
conditio ns on a1 and a2 if we want to construct a lattice that will be invariant
under one or more of these new operations. 1llere are four distinct types of
rest riction. and each leads to what we maycalJ a special lattice type. Tllus tllere
are five distinct lattice types in two d imensions. the oblique 1,IUicc and the four
special lattices shown in Fig. 9. Bravais lattice is the common phrase for a
distinc t lattice type; we say that the re arc five Bravais lattices or lIets in two
dimensions.

Tllrce-Dimellsional Lattice Types

The point symmetry groups in thrce dimensions requirc the 14 d ifferent
lallice types listed in Tablc l. The general latticc is frielinic, and there are 13
special lattices. These are grouped for con\-cnicllce into systems classified ac-
cording to sevCII types of cells, wh ich are triclinic. monoclinic, orthorhombic,
tetragonal. cubic. trigonal, and hexagonal. The division into systems is ex-
pressed in the table in terms of the axial relations tha t desc ribe the cells.
The cells in F ig. 10 are conventionaJ cells; of these onl y tIle SC is a primi-
tive cell. Often a non primitive ccll has a more obvious relation with the point
symmetry operations than has a primitive cell.

Table 1 nlC 14 laUice l)'pes in three t1imensions

Number Restrictions On
of <:on\'cntional (-ell
System lattices axes and ang ......

T,iclinic [ al#a2#a3
a""{3 ¢ ¥
]'.Ionociinic 2 a, #al¢a3
a = y=fU -.F{3
OrtllorllO Lnbic 4 a,;o! a 2;o! a3
a={3=y = OO"
Tctr.lgonal 2 a . = al""a.l
a = {3 = y = 9O"
Cubic 3 a L - a ~= a3

a={3:y"9O"
Trigonal al = a2 = a3
Q = f3 = ¥ < 120". >F 00"
a, = aZ""a3
a = {3 = 9C1'
y = 1 20~
 "

• •

•
!2 • •

• • •
• • •
II) S(luar" LaUke
11,1'" I..t., .. 90"

r~EJ •
• ~
"

,
• • •
• •
"1 ~1aItn (rl) c..nlHed red.",... lar bitloe;
'~Lt .. JootI; .-110" UPS are ........" lor hath tIM!
prim;"';'" ",,11 -.I for lhe
r<:<.~anguW llbi! celL for
~h [a,I" laJ; rp- 9(1.

Figure 9

.'
,

Figu,'c 10 TIle cubic .paC!! lattit~. The noll. ~howll are the C'Otwent ionai cells.
"
TaMe 2 Olaractenslics of cubic 1311io::c~

Simp": 11o<k-cenlered Fatt-(.'Cn tcred

VoluHle , oolwcntiouru cell

LaUK."C potllts per cdl
n'
", "4
In' In'
Volume, primit ive cell
Lattice points per unit vlllume "Un' V,, 4/«'
NumLcr of nearest neighbors"" 12
NcnrC$ t-ncighbor d istance ,6 8
3 111012 = 0,86& (l12 i12 - 0_7ff7o
NIHllLcr of sl'COncl neighbors
SC(."(l11d neighbor di~lunce
I'
21fl!O
6
, 6
,
J>ad in~ fnl<.1iOn b I. bYJ hrV2
-= 0.524 - 0.680 =0.740

"Tables of nUlllbcTli ofncighbors and distances in sc, Ix:c, fe<:, hcp . and dlQl1lond s(rul1ul"elI arc
given ~ pp. ]007- 1009 of}. IIinchfelder, C. F. Curtii' and It 8 . Bird, MoW;u /ar ,1ooory of FIlM.
Gnd liqI,id!, Wiley, 1964.
"The pac1.ing fraction i. the ma~i mu 'n propl)l1 ion of the a"ailable wllumc Ihat can be filled
with hard iphelft.

'!11Crc arc thr(..'C lattices in the cubic system: the simple cubic (sc) laltk-c,
til(' body-ccntcrcd cubic (Ix:c) lattice, and the fa(;'C-('Cntc red cubic (fcc) lattice.
The characteristics of the three cubic lattices are 5ummari7.ed in Tablc 2.
A primitive cell of the bee lattice is shown in Fig. ll , and the primitive
trm)Slalion vectors ure shown in Fig. 12. 'nlC primitive translation vectors of the
fcc IdUice are shown in Fig. 13. Primitive cells by defin ition contain only one
lattice point , but the conventional bee cell contains two lattice points, and the
!toe cell <:ontains four lattice points.
The position of a point in a cell is spt."Cific<.l by (4) in terms of the atomic
coordinates x. y, z. lIere each <:oordinatc is a fraction of the axiallcngth a .. 1l2.
III in the direction of the coordinate axis. willi thc origin I<lken at one corner of
the cell . Thus the l-oordinates of the body center ofa cell are iii. and the face
<'Cnters include BO, OH; IOi.
III the hexagonal system the primitive cell is a right prism based on a
rhombus with an included angle of 12(t. Figure 14 shows the relationship of the
rhombic cell to a hexagonal prism.

I NDI~X SYSTEM .."'On CnYSTAL l'lANES

The orie ntation or

a crystal plane is determined by three points in the
plalle, provided they arc not collinear. If each point lay 011 a different crystal
a\is. the plane could be specified by giving the coordinates of the points in
terms of the laUke constants (I], U 2. IT:! .
 CrydtU Slruclu ..

F"ogure It Body-cenlered cubic Iallice, .bowing 11 Figure 12 Primiti~'e tranWation vectors or the body-o::n-
primitive l-cll. n.e
erimith'e a.'Ulihown is 11 rhomoo- tered cubic lattice; these "cctOrii connect the latlice point
or
be<lron edge i Va
0, and the IUlglc between adja- al th e origin 10 latlice poinU; al the body cen teno The
emt edga is 109"28'. primiU"e cdl is obtained on completing the rhombohe-
d ron. In lemu clthe cube edge" the prin';tin" tran.Jalion
vectors arc
., - IOI(i+y- f.) ; aa - Io(- i+y+i} ;
aJ - lo (i - ~+i).

Figure 13 TI,,:, rhombohedral primilr.-e cell of lhe facc-cen - Figure 14 J\t>lalion clille primit ive eel
Icreel (:ublc cr)'l;lal. TIM' primitive Irnnslalion ve~1ors a" aa. a 3 in the be~agonol $y$lem (lleavy lines) 1<
con .....-d the Ialtia.' point at the OI"igin "~th IaUice poinU; al the a pri.ttl or hexagonal .ymrJlCtry. lien
face l't'llten. As drlM-Tl. the prim(t;\'e "eclOn a~: OIl - 0 , " ' 03-

·r - Ia{i+t) ; aa - Io (t+t) ; .~ - I ,,(i+i)

n.e angles ber....'I'Cfl the a.,,~ure 60". lIe-re 1, t, t. are the Carte-
sian unit ' ·e<."tOIS.
..
•

"
" "
•
,
t'ig.. ~c 15 This plane intercepts Ih" aj, a~, "3 "",cs al 311" 21111> 211 3, TIle
Il.OC:proaI11 of tJl<.>JiC nllmben are l , .. f. "nlC smal lC!>t Ihr~ inlefo(en ha" ing
the u.mc r;l.lio are 2. 3. 3, lind Ih.~ the indices of the plane are (233).

Howe \ cr, it turns out to be more useful for slTucture analysis to specify the
orie ntation of a plane by the indices determined by the following rules
(Fig. 15).

• Find the intercepts on the axes in tcnm of the lattice constants Gh~, G3 . The
axes may be those of a pri mitive or nonprimiti\'c cell.
• TOlkc th e reciprocals of these numbers and thell reduce to three integers
having the same ratio, usually the smallest thrcc integers. 11\C result , e n-
closed in pare ntheses (hkl), is called the index of the plane.

For the plane whose intercepts arc 4, I, 2, the reciprocals arc ". I, and I;
the sma1!cst three integers having the same rdl io arc (142). For an intercept at
infinity. the corresponding index is ;.:ero. 11le indices orsoOle important planes
in a cubie crystal are iIJustratecl by Fig. 16.
TIle indices (1Ik1) may denote a single planc or a set of parallel planes. If a
plane cuts an axis on thc ncgativc side ofthc origin. thc corrcsponding index is
ncgative. indicated by placi ng a minus sign abovc the index: (hkl). '!11C cube
faces of a cubic crystal arc (100). (010). (00 1). (foo). (010). and (001). Plancs
equivalent by symmetry may be de noted by curly brackets (braces) around
indices; the sct of cube faces is {IOO}. Whcn we speak of the (200) plane we
mean 3 planc paraJ\cI to (t OO) but cutting the 31 axis at la.
The indices (mac] of li direction in a crystal are the set of the smallest
integers that have the ratio of the components of a vector in the desired d irec-
tion , referred to the axes. The al axis is the (loo] direction; the -a2 axis is the
\
 C'1l'lal Struct"re 15

f
---- i
(100) (J 10) (III)

''''''
Figure 16 Indices of nnportaut plaue$;n a cubic CT} ~tal. 1hc plane (200)" parallcl to (100) and to
(TOO).

LOIO} direction. In cubic crystals the di rection LhklJ is perpendicular to a plane

(hkl) having the same indices. but this is not generally true in other crystal
systems.

SIMPLE CRYSTAL STRUcrURES

We discuss simplc crystal structures of gener.d interest: the sodium chlo-

ride, cesium chloride. hcxagonal close-packed , diamond, and cubic zinc sulfide
structures.

SOllimn Chloride Structure

°l11e sodium chloride. NaCI. structure is shown in Figs. 17 and 18. 'l11e
lattice is face-centcred cubic; the basis consists of one Na atom and one C l atom
separ.l.tcd by onc.ha.1fthe body diagonal of a unit cube. 11lCre are four units of
Nael in each unit cube. with atoms in the positions

000 1l0; 101 ; 01/

III 001; 010 ; 100 i
16
,
,~--~--
~,,~..- ~ .. . !:~': . ...'..:-.-
j .
'. - -.
• t;"-\ ,..... .- . •.;-.~ .. '. .....
j.
, •
. . ..~ ...
-l
'
-
; '. . ,---. .. ~
· · · · 0 · 1'0 · ;·······
.::." '. ' .. .-:'.' . . ..
. . ;'.. '

,-.-.:
. . ".'. 4!t
,.
, •. ' . '. ' f!'.~
,.-.. ....,. ... ~ ..~ .
(: tit \..' (" ,. .". ' .' .
.

. ..':" ,. , .

l~tf'·4; .. :,'. '~.:.:

t'igurc 11 We ",a~ ~'()nstr"t1 the sod ium chloride crys"

tallilnoct"n: b)· alT'anl(ing N,,·
and CI ions alternately
"tEl" 6;' . ~r.
.'i(,f: l'.:;... ..,',v~ ' ~ '
,: .: .
..
III tho: lattice pointb' of 11 simpk:: cubic bltice. In the ~"li­
tal cach ion iii su rrlllmded by . ix nearest neighbors oft!..,
V.~.(l, · C .'
opposite d w-gc. "nit: Iipll('C lAttice is f('C, and the basis
hasune CI ion al 000 and one /'\0 ' ion at Hi_ Thcllgurc
,heM'S (N1.C (.()twen' ;onal cuok cell. TIle Mmic: diamet"rs .";gure 18 Mood of K01ium chloride. TIoe sodIum ;0<11
htt", nrc n...!,,,,,d on rd~tioll 10 th e (.-ell In o, dcr to clnrify MrC smaUcr tlnn the cblorine ions. (Courtesy or A. N.
tht: sp"liai armnt;(.·"'''nl. 1101&..... and P_ Singer.)

"' igun. 20 7hc ccsium d llorirJc CJ}"S1a1 struc-

tu ... 1lx- 5pat'"e lattke i~ ~inlple rubic. and the
J<'igurc 19 Natur.l.1 " .... '>tab oflead ,,,mile, l'bS, whk-h has Ihe ln~is I,.." onl· Cs ' lOll at 000 and un" CI ion at
Nad CJ"ystal , Iructure. (PhoIogrnph b)· 8 . UurlCR>n.) III.
 1 CrtPfol S,ru<:,..... 17
,
Each atom has as nearest neighbors si~ atoms of the opposite kind. Reprcsenta-
tin: crystals having the NaG1 ammgement include those in the folloY/ing table.
1be cube edge a is given in angstroms; 1 }.. _ lO- 1! cm - 10- 10 m _ .Q.l nlll.

Figure 19 is a ph otograph of crystals of lead sulfide (PbS) from Joplin,

Missouri. 1be Joplin specimens form in beautiful cubes.

Cesium Chloride Structure

TIle ce~iulll chloride structure is shown in Fig. 20. There is one molecule
per plimilive cell, with atoms at the comers 000 and body-ccntered positions
iU of the simple cubic space lattice. Each atom may be viewed as at the center
ofaeube of atoms of the opposite kind, so that the number of nearest neighbors
or coordination number is e ight.

BeCu
C",,\4\ •
2.70 A
Cl)"tal
UHg
•
3.29 A
AlNI 2.88 NH...CI 3.87
CUZO (Il-L..u) 2.94 nBr 3.97
CuPd
AgMg
2.99
3.28
esCI
nl
4"
4.20

IInaganal Close-packed Structure (hcp)

Thcre are an infinite number of ways of arranging identical spheres in a
regular array that maxitnizes the packing fraction (Fig. 21). One is the face-
centered cubic stnlcture; another is the hexagonal close-packed structure
(Fig. 22). The fraction of the total volume occupied by the spheres is 0.74 for
both structures. No str ucture. rCJ!:ular or not. has denser packing.
Spheres arc arranged in a single closest-packed layer A by placing cacll
sphere in contact with six others. This layer may serve as either the basal plane
ol"an hcp structure or the (111) plane of the fcc structure. A second similar layer
B may be added by placing each spllcre of B in contact with three spheres of the
bottom layer. as in Fig. 21. A third layer C may he added in two ways. We
obtain the fcc structure if the spheres ofthe third layer are added over the holes
in the fi rst layer that are not occupied by B. We obtain the hcp structure when
the spheres in the third layer arc placed directly ove r the centers of the spheres
in the first layer.
 IS

"
fo'igure 21 A close-P"K'kcd la)'erol" ~phcres is sh",vn , with (;enters at points marh..! A. A scrood and
identical layer or spheres em be p'''''-1:d on top of thi~, above and paranel 10 the plane of the
dmwing. with centers o,'cr the points marh..! B. TIler" are tw" choices lOr a third tarcr. It can go
in over A o r over C. If il goes in over A the SCQUCOl''e;S A8A8AB. . and the Slndure is hexagonal
closc-p,wkell. If the third 1.1)'<!r goes in ""CT C the sequence i~ ABCABCABC . . . and the SIru<.'lurc
i~ r~nlcred C1,Lic.

, a 1\
I'
, ,
, ,,
"
:;, ,
.! ' 'If l
: ,
-<>: L,L: i ' 'KJI'-
a "-... IY--,"'-.,"V I','

t' igurc 22
" ---
The IICX3gonal clO!>C-l'acL...d slnlL'"
"

t-'ig,,~c 23 The p~imili"" L..,II has u, ~ "2.

lure. Th" atom po$ition~ in Ih~ ~lru<.111rc do wilh an incl"ded angle of 120". The c axiS (or
not C(ln~titllie a sp:wc lalli,..,. l'11e space lattice R,) i~ oo~mal 10 the plane of 8, and R•. TIlt'

I is simple hc~agona' wilh a basis of two idcnli.

,'Ill atoms associated with ea(h Iauice point.
11.., [alike parameters a and c arc indi.:alcd,
II whcre a is in the basa.l'lanc and c is the mag-
ideal hcp ~1r"" I "rc has c ~ 1.633 u. The two
atoms of one basis arc shOWll as !iOIid drclcs.
On., atom of the basis is at tI,e origin; the
other alom is at lB. whi...!, mcans at the po:;i-
lion ~ '" la, + Ia. + III;).

II
nit "de of lhe axis 83 of Fig. 14.

'i11e hep structure has the primitive cell of the bexagonallattice. but with a
basis of two atoms (Fig. 23). The fcc primitive cell has a basis of one atom
(Fig. 13).
The ratio da (or a3"al) for hexagonal closest-packing of spheres has the
value (0" 2 = 1.633, as in Problem 3. It is usual to refer to crystals as hcp even if
the actual da ratio departs somewhat from this theoretical vaiue.
'i11e number of nearest-neighbor atoms is 12 for both hcp and fcc st~uc­
lures. If the binding energy (or f~ee energy) de per '~d only on the number of
\.
II
 e:::-::------ --- ------a
,":,
,
:
------ /
~/
:
:,
,,I ,,
,, 1,
,, ,
:

"i~'re 24 Atomic po6ltioni in the cubic reO ci the diamond

"ructure projed ..-d on /I cube m ; f'racI:iOIls Ilcnote height
i,

Hgu",
____

~
---------
Cry~ lal
. ,,
I

','

structure cl diamond,
soo..ing the: tctraht:dra\ bond arranllC-
.l.Jo,.e the base in unit, of IIl1,bc t:~lge. "The poinu Rl 0 and I m'.'Tlt .
are 00 the fcc Lotl;':e; those at f and f are on II similar IlIttk'C
di~placffi along the body diagonal by one-lOurll, of it51t:ngth.
With ~ It'C space lattice, the IJUSis CQOlsi~ts r;l tv.'o idt!ntical
moms at 000; HI.

nearest-neighbor honds per atom, there \I,Quld be no difference in energy be-

tween the fcc and hcp structures.

Cl)'slal da C!')"l1ll1l da Crystal da

lie 1.633 Zn 1.861 Z, 1.594
Be 1.581 Cd I S86 Gd 1.592
'Ig 1.623 Co I.fi22 Lu 1,586
T; 1.586 l' t.57IJ

Diamond Structure
The space lattice of diamond is fcc. -111c primitive basis has two identical
atoms at 000; tU associated \\1Ih each point of the fcc lattice. as in Fig. 24. 11ms
the conventional unit cube contains eight atoms. There is no •....ay to choose the
primitive cell such that the basis of diamond contains only one atom.
TIIC tctrahedral bonding ch:uactcristic of the diamond stPlcturC is shown
in Fig. 25. Each atom has 4 nearest ncighbors and 12 next nearest ncighbors.
The diamond structure is rdatively cmpty: thc maximum proportion of the
availablc volumc which may be filled by hard spllcrcs is only 0.3-1. which is 46
pcrccnt ofthc filling'" 'tor for a closest-packed structure such as fce or hcp. TIle
I
20

(
diamond structure is an cKamplc of the directional covalen t bonding found in
column IV of the periodic table of clements.
Carhon, silicon, gcnnanium, and tin can crystallize in the diamond struc-
ture, ",;Ih lattice constants a"" 3.56,5.43, 5.65. and 6.46 A. respeclively. Here
(/ is the edge of the conventional cubic cell.

Cllbic Zi'lC Slilfide Stn/clure

1bc diam()JJd structure may be viC\\w as twu fcc structures displaced from
each other by one-(lUUTler of a bod)' diagonal. The cubic zinc sulfide (zinc
blcnde) strudure results when Zn atoms arc placed on one fcc lattice and S
atoms on the other fcc lattice. as in Fig. 26. '111c conventional cell is a cube. The
coordinates of the Zn atoms arc 000; OU ; ~Ol; ao; the coordinates of the S
aloms arc iH ; Hi; iU ; JU. '111c lattice is fC<'. '111crc arc four molecules o f ZnS
per conventional ccJ1. About each atom there are four equally distant atoms of
the opposite kind arranged at the wmcrs uf a regular tetrahedron.
'I1lC di.1mond structure al lo,",,'S a center-of-inversion symmetry operation at
the midpoint uf evel)' line between nearest-neighbor atums. The inversion
uperation C'oInics all atom at r inlu an atum al - r . The cubic ZnS structu re d oes
not ha\'c inversion ~ymmc t ry . Examples oflhe cubic zinc sulfide structure arc

Crystnl Crystal a
C uF
"
4.26 ,\ ZnSe S.65 A
SiC 4.35 C"'" 5.65
CllCI 5.41 AlA, 5.66
ZnS 5.4 1 CdS 5.82
AlP 5.45 InSb 6.46
Cal) 5.45 Agl 6.47

'(be close ccJuaiity of sevcral pairs, notably (AI,CalP and (A I.CalAs, makes pos-
sible thc construction uf semiconductor hcterojunctions (Chapler 19).

DIRECT IMACINC OF ATOMIC STRUCTURE

Direct images of crystal structure have been produced by transmission

electron microscopy. Perhaps the most beautiful images arc produced by scan-
ning tunneling microscop); in STM (Chapter 19) one cxploits the large varia-
tiuns in {juantum tunneling as a function of the height uf a £Inc me tal tip above
the surface uf a cl)'Stal. The image of Figure 27 wa,<; produced in this way; see
alsO' Figu rcs 12.19 and 19.21. An ST~I melhod has been de\·eloped Ihal will
assemble singlc atO'ms intO' an urganized layer nanometer structure un a crystal
substrate: sec the electron corroll in Figure 19.21.
 \

l'igure 26 Cr)'stal structure of cubic %inc sulfKlc.

Figure 27 A SUlnninJ,: tunneling micr~ ;rna).,,,, of atoms 0fI a (III ) lurk'C of platinum at 4 K.
TIle nlllral neighbor spacing is 2.78 A. (PI,oIo coo rtesy ofl) , M. ElgllT. IB M RCK'IU't:h Division .)

NONIDEAL CRYSTAL STRUcrURES

The ideal crystal of classical crystaUographcrs is funned by the pcri(xlic

repetition of identical units in space. But no general proof has been given that
the ideal crystal is the stale of minimum energy of idcntiml atoms at absolute
zero. At finit e temperatures this is not likely to be tme-sec the discussion of
lattice defects in Chapter 18. Further, it is nol alwar.; possible for a structure to
attain the equilihrium state in a reasonable .time- see the discussion of glasses
in Chapter 17. Many structures Ihat occur in nature arc not entirely pcri<xlic;
sec the quasicrystals treated at the end of Chapter 2. We give some examples
here that supplement those in the chapters just cited.
"
Rarulom Stacking anti Polyt!lpisfU
TIle fcc and bep structures arc made up of close-packed planes of atoms.
The structures differ in the stacking sequence of the planes. fcc having the
sequence ABCABe .. . and hcp having the sequence ABABAB . . Struc-
tures are known in which the stacking sequence of close-packed plane5 is ran-
dom. This is known as random stacking and may be thought of as crystalline in
two dimensions and noncrysta1linc or glasslikc in the third.
Polytypism is characterized by a stacking sequence with a long repeat unit
along the stacking axis. The best k'TlOWll el(ampie is zinc sulfide, ZnS, in which
more than 150 polytypes have be en identified, with the longest periodicity
being 360 layers. Another example is silicon carbide, SiC, which occurs with
more than 45 stacking sequences of the close-packed layers. The polytype of
SiC known as 393R has a primitive cell with a =: 3.079 A and c =: 989.6 A.
The longest primitive cell observed for SiC has a repeat distance of 594 layers.
A given sequence is repeated many times within a single crystal. The mecha-
nism that induces such long-range crystallographic order is not a long-range
force as such, but is associated with the presence of spiral steps due to
dislocations in the growth nucleus (Chapter 20).

CRYSTAL STRUcruRE DATA

In Table 3 we list tbe more common crystal structures and lattice struc-
tUres of the clements. Values of the atomic concentration and the density are
given in Table 4.
Many clements occur in several crystaJ structures and transform from one
to the other as tbe tempcratureor pressure is varied. Sometimes two structures
coexist at the same temperature and pressure, although one may be slightly
more stable .
TIle reader who wishes to look up the crystal structure of a substance may
consult the excellent compilation by Wyckoff listed in the references at the end
of the chapter. Stnlctllre Reports and the journals Acta Crystallographica and
Zcitschrift wr Kristallographie are valuable aids.

\
 T.ble 3 Crystal structures oCthe elements
H '~ H e' 2K
ho, The data given are at room temperature Cor the most common ronn, or I t hop .
the stated temperature in deg K. F or further descriptions or the clements
3.75 3.57
see Wyckoff, Vol. 1, Chap. 2. Structures labeled complez are described
6. 12 S.83

-..
there.
u'" B.
boo
3.491
hop
2.27
....
B
...... ...
C
3.567
N "" 0
S." <0,)
F N. ~
foe
4.46
3.59 (N:)

~a !oI(

'"
4.225
M,
hop
3.21
5.21,
,
,
,
. _. - ..:l '"""Crystal structure.
a lattice parameter, in A
c lattice parameter, in A,
,
,
,
AI
f"
4.05
......
SI
5.430
-,. -. -..
P S CI
(CI1 )
Ar 'K
f"
5.31

C. So Tf C, M, F. Co NI C. G. G. A. S. B,
...•• , -,,,
K" V l' Kr 4K

'"
5.225
f"
S." ""
3.31 ""
2.95 '" "',
3.03
2.88
tIIb~
WIIplu '" ""
2.87 2.51
foe
3.52
f"
3.61
ho,
2.66
ClGI1l~lu dllmond
5.658
'hoInb.
(Brl )
foe
S.64

. ..."", -...
5.27 4.68 4.07 4SS

Rb " S, Y l, Nb Mo T, R. Rh Pd A, Cd I, Sn (a) Sb T. I
-.. ....
Xe 'K
foe ho, ho, foe foe ho, telt.
'"
5.585 608
hop
3.65
5.73
3.23
boo
3.30 '"
3.15 2.74
hop
2.71
4.28
f"
3.80 3.89 4.09 2.98 3.25
4.95
6.49 (11)
f"
6.13
5.15 4.0 562
...
C. "
boo
,04S
B.
'"
5.02
h".
3.77
ABAC
3.19
S.OS
Hf
hop
T.
boo
3.30
W
boo
3.16
R•
hq>
2.76
4.46
O.
ho,
2.74
4.32
I,
foe
3.84
PI
foe
3.92
A.
fe<
4.08
H,
...... TI
ho,
3.46
5.52
Pb
foe
4.95 -
BI Po
"
3.34
At

-
R,

-
-
F,

-
R.

-
'"foe
5.31
I"
C.
'" Nd Pm Sm
.,.. E. Gd Tb Oy
hop
Ho E< Tm
hop
Yb l. I
f
f" h". h". boo ho, hop hop ho, foe hop
5. 16 3.67
ABAC
3.66 - 4." 3.63
5.78
3.60
5.70
3.59
5.65
3."
S.62
3.56
S.59
3.54
S.56
S." 3.SO
5.55

Th
foe
Po
tetr. .... -.... -""
U N, P. Am
"',.
Cm Bk Cf E. Fm

-
Md No L,

-
- .- .'.- -
5.08 3.92
3.24
3.64
ABAC
-

---.......--
- - - - -
eJ
, Table 4 Density and atomic concentration He 2~
H" ~
0.088 The data are given at atmospheric pressure and room temperature, or at the 0.205
stated temperature in dcg K. (Crystal modiflcatlons as for Table 3.) Itllltm)

li 78J( B. B C N ~ 0 F Ne 41(
0.542 1.82 2.47 3.516 1.03 l.SI
4.700 12.1 13 .0 17.6 4.36
3.023 2.22 1.54 1.44 3.16

Na 51( M, AI 51 P 5 CI 93K Ar 4K
1.013 1.74 , Density in g tm -) (lO'kg m-1 ) • 2.70 2.33 2.03 1.77
2.652 4.30 , Concentration in l Ol' cm -1 (10" m - 1) 6.02 5.00 2.66

."
3.659 3.20 • Nearest-neighbor distance, in A (lO-lom) 2." 2.35 2.02 3.76

0.910
1.402
C,
1.53
2.30
So
2.99
4.27
TI
4.51
5.66
V
6.09
7 .22
C,
7.19
8.33
M,
7.47
8.18
F.
7,87
8.50
..,
Co
8.97
NI
8.91
9.14
Co
8.93
8.45
Z,
7.13
6.55
G,
5.91
5.10
G.
5.32
4.42
As
5.77
4.65
S.
4.81
3.67
Br l23K
4.05
2.36
Kr
3.09
2.17
41(

4.525 3.95 3.25 2.89 2.62 2.50 2 .24 2.48 2.50 2.49 2.56 2.66 2.44 2.45 3.16 2.32 4.00

Rb 51( 5, Y Z, Nb Mo To Ro Rh Pd Ag Cd I, 5, Sb T. I Xc 41(

1.629 2.58 4.48 6.51 8.58 10.22 11.50 12.36 12.42 12.00 10.50 8.65 7.29 5.76 6.69 6 .25 4.95 3 .78
1.148 1.78 3.02 4.29 5.56 6.42 7.04 7.36 7.26 6.80 5.85 4.64 3.83 2.91 3.31 2.94 2.36 1.64
4.837 4.30 3.55 3.1 7 2." 2.72 2.71 2.65 2.69 2.75 2.89 2.98 3.25 2.81 2.91 2." 3.54 4.34

Cs 51<. B, l, Hf T, W R. Os I, pt Ao Hg227 TI Pb BI Po At R,
1.997 3.59 6.17 13.20 16.66 19.25 21.03 22.58 22.55 21.47 19.28 14.26 11.87 11.34 9.80 9.31
0.905 1.60 2.70 4.52 5.55 6.30 6.80 7.14 7.06 6.62 5.90 4.26 3.50 3.30 2.82 2.67 - -
5.235 4.35 3.73 3.1 3 2." 2.74 2.74 2.68 2.71 2 .77 2.88 3.01 3A6 3.50 3 .07 334
F, R, Ao I"
10.07 C. P, Nd Pm Sm Eo Gd Tb 0, Ho E, Tm Yb lo
- - 2.66 6. 77 6.78 7.00 7.54 5 .25 7.89 8.27 8.53 8.80 '.04 9.32 6.97 '.84
3.76 2.91
3.65
2.92
3 .63
2.93
3.66
- 3.03
3.59
2.04
3.96
3.02
3.58
3.22
3 .52
3.17
3.51
3.22
3.49
3.26
3.47
3.32
3.54
3.02
3.88
3.39
3.43

Th P, U N. Po Am Cm Bk Cf Es Fm Md No l'
11 .72 15.37 19.05 20.45 19.81 11.87
3.04 4.01 4.80 5.20 4.26 2.96 - - - - - - - -
3.60 3.21 2.75 2.62 3.1 3.61
- -
"
 SUMMARY

• A lattice is an army of points related by the lattice translation operator T =

Ulal + U2a2 + u3a3' where UI, U2. U3 arc integers and a" a2, 33 arc the crystal
axcs.
• To form a crystal \.\'e attach to every lattice point an identical basis composed
of s atoms at the positions rJ = :rJal + YP2 + Z.rt3, wHhj ... 1, 2, .. .• s. Here
x, y. z may be sclected to have vaJues between 0 and 1.
• The axes al. a2, a3are primitive for the minimum cell volume 18J • 32 )C 831 for
which the crystaJ can be constructed from a lattice translation operator T and
a basis at every lattice point.

Problems
I. Tetralledrol anglu. The angles between the tetrahedral bonds of diamond are the
same as the angles betwcen the body diagonals of a cube, as in Fig. 12. Use elemcn-
tary vedor anal)'$is to fiw\ the value of the angle.

2. Indica o{,Jlnna. Consider thtl plantls with inilires (100) aoo (001); thtllatticc is k<:,
aoo thtl iooioos refer to the convtlntiooal cuhic rell. What are IllC indices of thestl
plantls when referred to the primitive lUes of Fig. 13?

3. lIep . ' ruclure. Show that the da ratio for an kJeal hel(~al d ose-packed stnll1ure
is (1)lrt "" 1.633. If do. is slgllilleantiy larger than this value. the crystal structure may
be thought of as composcil of planes of cll-ly packed atoms. the planes being klosely
stacked.

References
ELEMENTARY
W. 8 . Pear$<,)n. CryslDI chnnisl'1l oOld physics of rmtDb Dnd Dlloys. Wiley, 1912-
II. I) . MegII..... Crys/lll structures: D workillg (/JIf'f"06Cn. Saundcn. 1913.

CRYSTALLOCI\A.PIIY
M. J. Buerger. Introduaion 10 Cf'JIStlU g<!Omet'1l, McGraw· mll , 197!.
G Bums allil A. M. Claser. Spoce groups for «>lid IIDle physldJts. Acatlcmic. 1918.
F. C . Phill.t», An introauclion 10 crysIDIlogl"lll"'fJ. 4th .:t.I .• Wiley. 1911. A W,."j place 10 begin.
II J JlJret~kt.'. C~IDI phrJSia: mDCfOSCOplc phyria of Dni.wlrDlnc w/ids. Bcujamin, 1914
B K. Vainshtein. Modern cn;ltllllogrnphy. Sp""ger. 1961.
J. F. ,,)..., Physiall proprl"fon of cn;stDls, Oxbd. 19M.
CRY~TAL GROWTl I
W G. f'&nn . Zone melting. Krwgn. 1978. I96&:.
A. w. Verc. C~ItI/ grou.II,. princirla Dnd pr0gn5l. F1enum. 1987.
J. C . Brice. Cl"fJstDI IlrDlj:lh procme•• IlaMed. 1986.
S. II Lill . ·· Fractals anil the" applieahon ill condt:nso<i1 matter PO)·s......- SoI,i1 ~lalc physics 39, '1tr1
(1986).
 .
Series, Jourrn'l\ ufCrystai Growth, includes procet."<.Iillgsofthc International Conference, OIl Cry~I9.1
Croo.-1h.
Springer Seri.,. , Crystals- CrO'.''th, P'roprrtie,. and App\l(:aIiollll.

CLASSICAl. TABLES AND I lANDBOOKS

InterTllltklno / tablu jor "'-ray crJI.It(lUogrophV. K)~,ocIt Press, 4 volumes, Birmin!..... m, 1952- 19'74.
J. F. Nye; Phll$iaJl prc,~rtiu of c~tol" their rq>~se"lal/{)n b" len.ron (HId lI\6/rica, O"lOnl,
1984,
P. Villar. am.! L. D. Calvert. Peors()II" IlIIndbook of crystallographic dalo for Inlennetallic ""O$>U,
Amer. Soc. Met als, 3 Yols., 1985.
A. F. Wells, Structurol il'\Org<'nk dremtst'll. 5th ed., 0d0n1 Uni''''mty Preu, 1000, 19S4c:.
W. C . 'V)"Ckolf, Crystal "ructura. 2nd ed., K......ger, IQliI.

(
 2
Rec;iprocal Lattice

DlFFRAcnON OF WAVES BY CRYSTALS 29

Bragg Jaw 29

SCAlTERED WAVE AMPUTUDE 30

Fourier analY5is 30
Reciprocal lattice vectors 33
Diffraction conditions 34
Laue etjuations 36
BRILLOUIN ZONES 31
Reciprocallallice to §c laUice 4()
Reciprocal lattice to bee lattice
Recillrocallaltice to fcc lattice

FOURIER ANALYSIS OF THE BASIS

"'"42
Structure factor of tile bee lattice 4.
Structure factor of the fcc lattice 45
Atomic fonn factor .5
QUASICRYSTALS ...
SUMMARY 49

PlmBLEMS 51
1. lnterplanar separation 51
2. Hexagonal Spite<! bttice 51
3. Volume of Brillouin ;tone 51
•• Width of difTUClion maximum 51
5. Structure factor oC diamond
6. Form factor of atomic hydrogen .,.,
51

1. Diatomic line
REFERENCES 5.

\
 0

5
'·n
..........

.0
I 11'
.5
[ IedIl)l'lS

i'--
,

o•••
5 "
Photon energy. keY
." Figure I \\'~ ..d eogth ~'ersU$ parli.
Neutron _II)', 0.01 eV de eIlCf'gy. lOr pho lom, lI.,uITOtl S,
Electron energy. 100 eV and dectroflS.

Figure 2 Dt:rivation of the Bragg C<lualiOll 2d jim 9 - nA . here d is the .pacing ri parallel atomic
n,.,.
planes am.l21Tn is the lMfen:llw in .,I".se \.>et".et.'n reflect ions from suca::lisivc plauM. refledmg
pl<llleS Ioa,·c nothing to do wi tll the surliore planes boomdflll: the pa.liNbr specimen .

28
 'JAPTER 2: RECIPROCAL LATTICE

DIFFRACTIOi'lO OF WAVt::S BY C RYSTALS

Bragg Law
We study crystal stmeture through the diffraction of photons, neut rons,
and cJectrons (Fig. 1). The diffraction depends on the crystal structure and on
the wavelength. At optical wavelengths such a.~ sooo A the superposition of the
v..aves scattered elastically by the individ ual atoms of a crystal results in ordi-
nary optical refraction . When the v..avclen~,'th of the radiation is comparable
with or smalle r thaq the lattice constant, we may find diffracted beams in direc-
tions quite different from the incident direction.
W. L. Bragg presented a simple explanation of the diffracted beams from a
crystal. The Bragg deri vation is simple but is conv incing only because it repro-
duces the corn'Ct .. result. Suppose that the incident waves are reflected
spccularly from parallel planes of atoms in the crystal, with eaeh plane refll.'Ct-
ing only a very small fraction of tile radiation, like a lightly silvered mirror. In
specular (mirrorlike) reflection the angle of incidence is equal to the angle of
reflection. The diffracted beams arc found when the refle<:tions from parallel
planes of atoms interfcre constructivcl)', as in Fig. 2. \Ve treat clastic SC""aUer-
ing. in wllich the e nergy of the x-ray is !lot changed on refle<:tion. Inelastic
scatteri ng, with excitation of elastic waves, is discussed in Appendix A.
Qmsider para1Ic1lattice planes spaced d apart. The radiation is incident in
the plane of the paper. The path difference for rays reflected from adjacent
planes is 2d sin e. where e is measu red from the plane. Constructive interfer-
ence of the radiation fro m successive planes otX"Urs when the path diflcrence is
an integral number 11 of wavelengths A. SO that

I';d sin e = flA . (I)

11lis is the Bragg law. Bragg refl ection can occur only for .....avelength A :s 2tI.
This is why we cannot usc visible light.
Although the reflection from each plane is specular, for only certain values
e
of \\ill the reflections from all parallel planl.'!i add up in phase to give a strong
reflected beam. If cach plane were pcrfl.'Ctl y rcfkcting. only the first plane of a
parallel set \\Uuld see the radiation, and any wavelength would be refleckd.
But each plane rcfll.'Cts 10- 3 to 10- 5 ofthe incident radiation, so that 1()3 to l OS
planes may contribute to the formation of tile Bragg-reflected beam in a perfect
Cry5tal. Reflection by a single plane of atoms is treak...:! in Chapter 19 on surface
physics.
111e Bragg law is a consequence of tile periodicity of the lattice. Notice that
the law docs not refer to the composition of the basis of atoms associatl.-d wilh
h-ery lattice point. We shall see, however, that the composition of the basis

"
 30

determines the relative intensity of the various orders of diffraction (dcnott.-d by

II above) from a given sct of paraJlcI planes. Experimental results for Bragg

rcncction from single crystals aTe shown in Figs. 3 and 4, for rotation about a
fixed axis.

SCATTERED WAVE AMPLITUDE

"11lC Bragg derivation of the diffraction condition (I) gives a neat statement
of the condition for the constructive inlc.fcrcnce of waves scattered from the
lattice points. We need a deeper analysis 10 determine the scattering intensity
from the basis of atoms, whidl means from the spatial distribution of eit.'Ctrons
within each ceIL
From (1.3), a crystal is invariant under any translation of the form T =
UIII, + 112112 + U3US. where u" 11 2. 113 arc integers and 8], 32. a3 arc the crysta1
axes. Any local physical property of tile crystal is invariant unde r T. such as the
charge concentration. electron number density, or magnetic moment density.

Fourier Analysis
What is most important to us here is that the electron number density ,,(r)
is a periodic function of r, with periods a •• a2, a3 in the directions of the thrcc
crystal axes. TI1US
lI(r + T) = nCr) . (2)
Such periodicity creates an ideal situation for Fourier analysis. The most inter-
esting properties of crystals arc directly related to the Fourier components of
the electron dcnsity.
We consider first a function fI(X) with period 0 ill the direction x, in onc
dimension. We expand 11(.1') in a Fourier series of sines and cosines:

n(x) = '10 + L. [C p cos(2'7Tpda) + Sp sin(211pXia))

,> , (3)

II where the p's are positive intcgers and Cpo S" arc real constants, called the
Fourier coefficients of the expansion. The factor 2nlo in the argumellts ensures
that n(x) has the period tJ:
I 11(.1' + a) = "0 + :rIC,. cos(211pXia + 2'7Tp) + S" sin(2'7Tpda + 2'7Tp»)
(4)
= no + :rIC" eos(2'7Tpda) + 5" sin(2npxJa») = fI(X) .

II We say that 2'7Tpla is a point ill the rcciprocallaUice or Fourier space of tile
crystal. In one dimension these points lie on a line. The reciprocallatticc points
tell us the allowed terms in the Fourier series (4) or (5). A term is allowed if it
is consistcnt with the periodicity of tllc crystal, as in Fig. 5; othe r points in the
reciprocal space are not allo .....ed in the Fourier ex~nsion of a periodic function.
\
 , lluiprocal Lall ice 31

-- - --- I'--~
....
~ - I .. A

- 'lMA
I
I
l
u.o.ooo~ .. ..
....
. · I 16.l

~ ' 00
1IlOi •. -

• .. ~
Br-uOOfOlt'
j
'" -----..

- U...J.....iatcd
rompoocnt. of
mal~ """In
figu" J Sketch of a nl<XlOChromator which by Bragg rc flt'ct loll .ei•.'d S a narrow sp<.'ctnlnl orx-ray
<X' nelliron " ....,,,lengths from a broAtl sp«lru m ludden! beam. TIle up.,.,' part oflhe figure shows
the analysis (obtained by rc:flt'ctiotl from a S«Ond crystal) dlhe puril )'o£. 1. 16 J.. beamof nCulrOr1$
frorro a C'alcium fluoride Clj'flal monod,romator. The main beam i. Ihal nol reflccted from the
~olld ~-~al. (After C . ~,.)

Figure of X-ray oJiffllldomcter rewnling cl pOW<\ered sili(ml, dMJWing a ~unter R'COroing oflhe
diffracted beams.. (CclUrte$Y of w. Parrish.)

Figure:;: A periodk: function n(l) of o

period II, Qnd the tenns 2fr{J1a that
may appear in the Fourier tram/orm
n~) .. I nl' eo:p(I2"F"a). The mag-
nitude. or the iooivio.!uaI tern,. n"are
not plotted.

\
"
It is a greal convenie nce 10 write the series (4) in the l'Orn p nc t form

n(x) = L,. 11" cxp(i2'1TJlxlal • (5 )

w here the sum is over all inte~rs 1): positive , nCWltiv(;'. and ze ro. Thl' cocm-
cit' llls II" now arc complex numbers. Tn e nsure th"t nCr ) i5 a real fundion. we
require
(6)

for tllen the sum of the terms in r and - p is Teal. The asterisk 011 fl!" d enotes
the complex conjugate of 11_,..
With ql "" 2'1T'IlXftl , the sum of the terms in lJ and - p in (5) can he shuwl1 to
he real if (6) is satis{lal . 11le sum is
n,,(ros It' + i sin !p) + " - IACOS IP - I sin \pI
= (II,. + IJ _1,)cos It' + j(rI,. - 1/ _,,)sin IP
which in turn is equal to the real function
2Rc{n,,} cos IP - 2Im{n,.} sin \fl. (8)

if (6) is satisfied . Here nc{!!,,1and Im{",,} denote the real and imaginary parts of
II". Thus the number density II (X) is a real function, as desired.
TIle extension o r tile Fourier analysis to periodic [unctions fI (r) in tJJn,.,<,
dimension s is strai ghtrol'"\\~lrd. We mus t find a sd or vectors C such that

n(r) ::: Lc lIe; exp(iC · r ) (9)

is invarian t under all crystal translations T that le.wc the crystal invariant. It
\vill he shown below that the set of Fourier coefficie nts lie; determines the x-ray
scatteri ng a mplitude.
Inversion of Fourier Series. We now show that thc Fourier cocffident ""
in the series (5) is given by

n,,::: (I - I r dx n(x) exp(- i2'IT1,x/a) (10)

Substitute (5) in (10) to obtain

Ill' :: a- I L ",,"f" dx exp{i2'1T(J)'- JI ).rJa] (II)

I'· ~)
I r J1' ¢ p the value or the integral is

•
--;--;-,;--,- ,(~"'P"- p)
·i2'1T(p' 1)
- I) ;; 0
because ')' - " is an integer and exp{i2n(integer)) "" 1. For the term,,, :: J, the
illh,-grand is exp(iO) = I , and the vaJue of the integral is a, so that II" =
a- I"IP = "", which is an identity, so that (10) is an identity.
Similarly, the inversion of (9) givcs

11(; :: V ; I 1
cdl
dv nCr) cxp(- iC, r) (12)

Here Vc is the volume of a cell of the Cl)'stal.

Reci1)1"(Xn/ Lattice Vectors

To proceed further with the Fourier analysis ofthe electron concentration
we must find the vectors C of tile FouriL"f" sum:£11(; exp(iC' r)as in (9). There is
a powerful. somewhat abstract p rocedure for doing this. The procedure form s
the theoretical basis for mueh of solid state physics, where Fourier analysis is
the order of the day.
We construct the mtis vectors bl> bi!, h:J of the reci procal lattice:
,
. (13)

The factors 211'" are not used by crystallographers but arc convenient in sol id
state pllysics.
If a .. ~, 13 are primitive vectors of the crystal lattice, then hI> ~, b:, arc
primitive vectors of the reciprocal lattice. Each vector defined by (13) is orthog-
onal to two axis VL'CtOrs of the crystal lattice. Thus ht • b~. b:, have the property
bi ' I} = 2rrliu ' (14)
where liu = 1 iff = j and liu = 0 ifi ¢ j.
Points in the reciprocal lattice arc mapped by the set of vectors
C = v1b. + ~bl + V:\b:, • (15)
where VI. '1, tl3 are integers. A vector C of this form is a reciprocal lattice
vector.
Every crystal structure has two lattices as.wciated \\;th it, the crystal lattice
and the reciprocal lattice. A diffraction pattern of a crystal is, as we shall show, a
map of the rcciprocallattice of the crystal . A microscope image. if it could be
resolved on a finc enough scale, is a map of the crystal structure in real space.
The two lattices are related by the definitions (13). 'I1lUS when we rotate a
crystal in a holder, we rotate both the direct lattice and the rcciprocallatticc .
Vcetors in the direct lattice have the dimensions of[ length]; vectors in the
redprocallattice have the dimensions ofli/length]. The rcciprocallatticc is 'I.
lattice in the Fourier space associated with the crystal. 11le term is motivated
below. \VavcVt.'Cl:ors arc always drawn in FOUl;cr space, so that cvcry position
in Fourier space may have a meaning as a description of a wave, but thcre is a
special significance to the points defhu..>d by the set of C's associated with a
crystal structure.
TIlt.' vt..octors C in the Fourier scrit..'s (9) are just the n .ociprocal lattice vocton
(15), for then the Fourier series represcntation of the dt..dron dcnsity has the
desirt...o invariance under any crystal translation T = u .a. + 112a2 + 11383 as de-
fint..>d by (1.3) . From (9),

nCr + T) ::: LC nc exp(iC' r) exp(iC" T) (J6)

But exp(iC' T) ::: 1, because

exp(iC . T) ::: cxp[i(v.b, + V2i>2 + V:3b:,) . (u,a, + 1I~2 + 113a3»
(17)
= c,p{i2-n{v,lI, + V2112 + tlJlI.:J]
The argument of the exponential has the fonn 2m timt..'S an integer, because
V,II , + V:2U2 + lIJU3 is all integer. being the sum of products of integers. Thus by
(9) we have the desired invariance. nCr + T) ::: nCr).
This result PTO\'es that the Fourie r representation of a function periodic in
the crystal lattice can contain components nc cxp(iC' r) only at the wciprocal
lattice vectors C as defined by (15).

Dif/raction Conditions
T/leorem. The set of reciprocal lattice vt..'Ctors G detelmincs the possible
,..·ray reflections.
We St..'C in Fig. 6 that the difference in phase factors is exp[i(k - k'), r1
bctwt..'Cn Ix..'ams scattered from volume clcmcnts r apart. The wavevectors of
thc incoming and outgoing beams are k and k' o 111C amplitude of the wave
scaUert.>d fmm a volume element is proportional to the local electron concentra-
tion n(r). The total amplitude of the scatlert..>d wavc in the direction of k' is
proportional to the integral over the crystal of n(r) dV limes the phase factor
exp[j(k - It) . r J.
In other \"'OM, the amplitude of the eit..'Ctric or magnetic field vectors in
the scatlen.>d ck-'Ctmmagnetic wave is proportional to the foUowing int egral
which defiIlt..'S the quantity ,," that we call the scattering am plitude:
F ::: f dv nCr) exp[i(k - k') , r) ::: f dV nCr) exp(- iAk' r) • (18)
where
k+Ak ::: k' . (J9)

Here Ak measures the change ill w~l\Ie\'t..'<"tor and is called the scatte ring vector
(Fig. 7). We add Ak to k to obtain k', the wavevector of the scattered bt,.'llI1l.
~
 2 Rmprocol Lallice 33

Figure 6 'Ibc dilferen,'e in path len~th of lhe inei<lcnt w:',-e k at the points 0 , r i~,. sin <p, and the
.:Iifference in "hase angle is (2 .... sin <pYA, which is e(Joal to k . r. Fer the uilfraclt...l ,,",we the
.:Iiffere'K'e in phase an1'(le is - k' . r. The tolal differen<.'e in phase angle is (k - k') . r . an.:lthe wa"e
scattereU from dV at r has the phase foc-tor exp[i (k - k') ' rl relati,-e to the "m'e s<:alte, ~..I fTDltl a
volume elemen t at the origi n 0_

Figure 7 Definition oflhe s,... tt ering vector tr.k such that

k + tr.k '"' k' . In elastic S<-1Itt ering the m."IgIliludes satisfy
k' - k. Further, in Bragg SUlltering from a pcriotlic latli<:e
anyaUowro tr.k must equal some redpro<:al lattice ve<.ior C.

We introduce into (18) the Fourier components (9) of n(r) to obtain for the
scattering amplitude

F "" L: f dV nc; exp[i(C - 6k)' rl (20)

C

When the scattering vector 6k is equal to a particular reciprocal lattice vector,

(21)

the argument of the exponential vanishes and f' = Vnc;. It is a simple exercise
(Problem 4) to show that " is negligibly small when .6.k differs sib'llmcantly from
any rt;.'Ciprocal laUice vector.
In elastic scaUeling of a photon its ene rgy flw is con5el"voo. so that the
frequency w' = ck' of the emergent beam is equal to the freque ncy of the
incident beam. Thus the magnitudt;.-s k and k' are equal, and /.?- = k'2. a result
that holds also for ek-'Ctron and neutron beams. From (21) we found 6k = C or
\
36

k + G :::: k' , 50 thai the diffraction condition is written as (k + C)2 = Jc2, or

(22)

This is the ccntral result of the theory of e lastic scattering of wavt.'S in a

periodic lattice. If G is a reciprocal lattice vector, so is -G. and with this
substitution we can write (22) as
r - - -.....
!!IL'G - G2 (23)
11115 particular expression is oftell used as the conditioll for diffraction .
Equation (23) is another statement of the Bragg condition (I). The result of
Problem 1 is that the spacing d(Md) between parallel lattice planes that are
nonnallo the direction G :: Ilb l + k~ + lb 3 is d(/lkl) = 21f~GI. Thus the result
2k • G = C 2 may be wrille n as
2(21T1A) sin 0 = 2md(~kl) ,
or 2d(hkl) sin fJ::: A. Here 8 is the angle betwt.'C1l the incident bt.>am and the
crystal plane.
The integers 'Ikl that dcfUle G are not necessarily identical with the indices
of an actual crystal plane. because the hkl may contain a common facto r fl,
whereas in the dcfinition of the indices in Chapter J the common factor has
bccn eliminatal. We thus obtain the Bragg result:
2d sin (J "" n'\ , (24)
where d is the spacing between adjacent parallel planes with indices lun, kin,
Vn.

lAue EqlmtiOWl
The ol"iginal result (21) of diffraction theory, namely that Ak = G, may be
expressed in another way to give what are called the Laue equations. These are
, 'aluable IX'C3use of their gL'Omctrical rtlpl'esentation (s(''C Chapter 19).
Take the scalar product of both Ak and G successively with 3. , 32, 33 .
From (14) and (15) we get
a,; · Ak = 2~ ; 3:J. Ak = 211"V3 • (25)
These C(luations have a simple gt:ometrica1 interpretation. The first C(luation
3, • Ak = 21TV1 tells us that 6k lies on a certain cone about the direction of a,.
I hc second equation tells us that Ak lies on a cone about 3z as well, and the
third equation requires that Ak lies on a cone about 3:).
lhus, at a reflection !lk must satisfy all three equations; it must lie at the
rommonline of int ers~'Ction of three concs, which is a severe condition that can
be satisfk>d only by systematic SW~'epiIlg or searching in wa,·elength or crystal
orientation--or e lse by sh(;.'Cr accident .
 !J Reciprocal lAlli« :r.

Figure S 11)(' IJOin ts OIl t}w, right.h;.nd ~idc we rcdpnx;>llaUice IlOin ts of the crystal. 111e ~'CC'or k
is dmwn in the direction orlhe intident K·my hCllm , and the origin is chost:'11 SUL-h that k terminates
II.t any rL"<:il'l"u."al lall ke IlOint. We dmw • sphere of radius k '"' 2",111 about the origin of k. A
dilfr:>dcd beam will be formed Ir lhis .rh're intersect! any other poinl in the n'CillrOl"&l !alike. The
' IJhcreas drawn in tCt"Ct'PIS a ll'ointl"Onneclcd with the M"od ofk by a reciJlfUC'llllallice veclor C . 111e
d,If"",lcd ~·ray br.-am i. in II..., dirCdi<Jn It '"' k + C . The angl., (J is lhe nragg Illgle of Fig. 2. Th is
romtnx.1ion is due to P. P. t-;wII.ld .

A redutiful construction. the Ewald construction, is exhibited in Fig. 8.

This helps us visualize the nature of the accidellt that must occur in order to
Sll.tisfy the diffraction "cOndition in three dimensions. The condition in Iwo di-
mensions (diffraction from a surf:lCC layer) is tn:atl.'d in Chapter 19.
Reflt.'ction fmm a single plane of atoms takes p lace in the directions of the
lines of inte rsection of two cones, for example the cones definl.-<:i by the first two
of the Laue c<)uations (25). Now two cones will in general intercept each othe r
prcwidl.,,(\ the waVevt.'Ctor of the particles in the incident beam exceeds some
threshold value determined by the first two Laue equations. No accidental
coincidence is requin..><l. unlike the problem of diffraction in 3~. 'Illis matter is
of prime imp0l1ance in the diffraction of low energy e lectrons from the surface
of a crystal.

BRILLOUIl" ZOl"ES

Brillouin gave the stntement of the diffmction condition that is most Widely
used in solid state phy5ics, which mt.'8ns in the description of c1ectl"On energy
band theory and of the e lementary excitations of other kinds.
 38

\
" 11(;, --c.:

Figure 9. Recipo'OClll "'lIice ,JOints near ,h., poin t 0 ml Figure 9h Square ~iprocal wtK'e "il" recip1lX:'a!
the origin of the recil'rocallaUiCC'. The rt'<.iproc."3! lat - lalliec \'CCIors ~hown as fine black lines. -Ole Jines
Ii<..'e ,.,.,IOf C c conn<.'C.1S pointl OC, lind C" con n('(.is shown in white are perpendicular biSl,.'(io/"$ of the reo
00. Two plant'l I and 2 ~ dra"'" ",Ilk'" .are the p" .... l'iprorol loltl i<:e H.'CIOJS. TIle centrol square is the small-
pendicular biSC'<'tors of ~ and Cn, rt')Jlectf\"(~I )'_ An y ~ volume about the orqon which is bounded en lin~ly
,-edor from the origin to the plane I , suchlU k" wiH by ,.mile li l\Cli. The square i$ the W;gner-SE'ilz p rimi-
satisfy the difTra'1ioo «",cl it ion Ii,' (ICd - (lGrjl. li'e rell of Ihe rec;l'ro<.-~1 IaltlCe. It is "'llL1ed the IIrst
Any ndor- from the urigin 10 the plane 2, ~Udl as ~ •• Boliouin 7.0111'.
will sal i~fy the dilTl"lIClion condi lioo lei ' (lel)) -
(lC"t.
A Brillouin zolle is defined as a Wigner-Seitz plimiti\'e cell in the recipro-
cal lattice. (TIle construction in the ruH..-'Ct lattice was shown in Fig. 1. 6.) The
value of the Brillouin zone is that it gives a \'h id geometrical interpretation of
the d iffraction condition 2k· G = C'l of Eq. (23). We divide both sides by 4 to
obtain
(26)
We work in reciprocal space, the space of the k's and C's. Sclt:ct a , 'ector G
from the origin to a r<.'CiprocallatHce point. Construct a plane normal to this
vector G at its midpoint . 'Ibis plane forms a part ofthe zone boundary (Fig. 9a),
An x-ray beam in the crystal will be diffracted if its \Yavel'ooor k has the magni-
tude and din_'CIion requir<.>d by (2:6). The diffracted bt:am will then he in the
I dirt.'ction k - G, as we see from (19) with Ak = - G. l1lUs the Brillouin con-
struction exhibits all the wav<.,,·ectors k which <.'an be Bragg-rc(1ccted by the
crystal.
The set of p lanes that arc the pcrpcll(Jicular bist.'Ctors of the reciprocal
lattice vectors is of general importance in the theory of ,va,·e propagation in
crystals. A wave whose wavevcctor drawn from the origin tennillah.'S on an)' of
these plant.'S will satisfy the condition ror diffrnction .
These planes divide the Fourier space of the crylital into fragments. as
shown in Fig. gb for a square lattice. The central square is a primitive cell of the
n:ciprocal lattice. It is a Wigner-Seitz cell of the ( .)rocal lattice.
 ---~"

•
F"'re 10 Consl~ion of the Onl Bn110uin
zone 10.- an oblique bUice in ....vu dimcnsiom. We •
first draw a numllt!. of \-eciors rnll n 0 10 nearby
points in the rctiprucal bttice. Ned We oonstnl<.i •
linn llerpendicul:t.r tolhese \'t"I.1on at their mid·
point•. The .maIla! enclosed area is the 6rst
Brillouin zone, •

r·=·~·
j
'.
"-:/t,- ,
.-~. .
.... -'""T,"'~
.....-,'"' ,
.....
Figure II Cl)1lal and l«'iproc:'ill lattices In one di,neflsion . The basis vcdQr in the rec.';prucal
lallice is b, oflcngth equal to 2",'<1. The dlOrtest re<.iprocallaliloe vecto"" from the origin are band
-b. The perpendicular bisectoc"' of these "edoni form the oound;uie$ of the tint Brillouin woe.
11w:: boundaries are a' '" - :1::""11.

The central cell in the reciprocal lattice is of special importancc in the

theory of solids, and we call it the first Brillouin zone. The first Brillouin zone is
the smallest volume entirely enclos(..(1 by planes that are the perpendicular
biSL'Ctors of the reciprocallaltice vectors drawn (!'Om the origin .
The first Brillouin zone of an oblique lattice in two di mensioIlS is COIl-
sh'Uctcd in Fig. 10 and of a linear lattice in one dimension in Fig. 11 . 'Ibe zone
boundaries of the linear lattice are at k ". ±7rla, where a is the primitive axis of
the crystal lattice. (
"
Hislorlca1ly, Brillouin zo nes a re 1101 part of the lall~'\lage of x-ray diffraction
analysis of crystal structures, but the zones are all eSSential part of the analysis
orlhe electronic energy-band structure of crystals. The spt.>Cia1 utility of the first
Bl"illouin zone is dcvc!opt..(I in Chapter 9.

Reciprocal Lattice to 3C Lattice

nlC primitive translation \'(..'Ctors of a simple cuhic lattice may be taken as
the sel
(27a)
Here X, y, 1: arc orthogonal vectors of unit length . The volume of the ceU is
a,' 32 X a3 = (13. The plimitive translation vectors orlhe rt:ciprocallattice ano!
found from the standard prescription (13):
(21b)
l'Je re the reciprocallattire is ilse lf a simple cubic lattice, noworlattire collstant
2'TTla.
The boundaries orlhe first BriUouin ...ones are the planes lIonnal to the six
rt.-'Ciprocallaltice vectors ± blo :t b 2 • ±b:! at their midpoints:
±{b. = ±(lI'la)x ;
111e six plancs bound a cube of led ge 211'10 and of volume (27Tla'f; this cube is the
first B..iIIouin zone of the 5C crystal lattice.

Reciprocal Lattice to bee Lattice

The primitive t!"anslation vectors of the Lee lattice (Fig. 12) arc
a.""ia(-x+Y+i); fl2 = la(x-y+i); 1"13=ia(x+y - i). (28)
where a is the sidc of the cOllvcntional cube and x, y, i are orthogonal unit
\·edors pal"allcl 10 the cube edges. The volume of the primitive cell is
V= la •• fI2: )( aJl = 1a3 . (29)
111t: primitive translaltollS oftne reciprocallaUice are defint.'d by (13). We
have, using (28),
h. = (211'Ia)(5' + i ) ; ~ = (2-m'a)(i + i) ; b, ~ (2"'a)(' + Y) • (30)
NOle by comparison with Fig. 14 (p. 42) that these are just the primitive vectors
of an fcc lattice, so that an fcc lattice is the reciprocallaltice of the bee lattice.
111e general n.'Ciprocal lattice vector is, for integra] VJ, vt, 0:1,
C "" v.b l + V:zbl! + VJb3 = (2m'a)[(~ + ~ + (VI + Oa)y + (VI + ~J (31)
Ihe shortest C's arc the following 12 vectors, where all choices of sign are
independent :
(2r.la)(±y ± 2) ; (2m'a)(±x ± 2) ; (2ma)(±x ± y) . (32)
 Figure 13 Fir.;1 Brillouin zone of the body..rentered
Figurt: 12 I'rimitive buil \'e<;IQU Qf III<' body-«'nlcred (....,bic buice. The figure Is I. regular rhombic
(.l1bic lallke. dod~cdron .

The primitive cell of the reciprocal lattice is the parrulelepipL'<1 dcscribL'tl

by the b l> b,;, b3 d efim.>d by (30). The volume of this cell in rL'Ciprocru space is
hi' b 2 ~ b:J "" 2(2'11'10)3. The cell contains one rt.'Ciproca1 lattice point, bt.'Ca.use
each of the eight comer points is shart.-d among eight parallelepi(l<.-ds. Each
parallelepipt.-d contains one-eighth of each of eight corner points.
In solid state physics we take the centr.d (WignL"T-Scit"..:) cell ofthe recipro-
cal lattice as the first Brillouin zone. Each such cell contains one lattice point at
the central point of the cell . This zone (for the bee lattice) is bounded by the
planes nonna! to the 12 vectors of Eq. (32) at their m idpoints. The zone is a
reb'lliar 12-fac<.>d solid , a rhombic dodt.>caht.>dro n, as shown in Fig. 13. 111C
vectors from the origin to the center of each face are
('lTlaX :!: »:!: i) ; (33)
All choices of sign are independent, giving 12 vt.'Cton.

Reciprocal Lattice to f cc Lattice

The plimitive translation vectors of the fcc lattice of Fig. 14 are
al :c: ~a(y + i); ~ =! a(i + 2); ~ "" ia{i +y) (34)
The volume of the primith-e cell is
V = la, . a
2 )( aJ "" in' (35)
nle primitive tra nslation vectors of the lattice recip rocal to the fcc laUice

hi = (2'ITla)(-i + y + i); ~ = (21rlaXR - y + i)

(36)
b:J "" (21TIa)(i + y - i) .
.,

r
"

1 --
Figure 14 Primitive basis ,"ectors of the fa.:e -
.,entered cubic I~ttice.

These are p rimitive translation vectors of a bee lattice, so thai the bee lattice is
reciprocal to the fcc lattice. -111e volume of the primitive cdl of the rccipl'OC"<l1
lattice is 4(21Tla)3_
The shortest C's are the eight vectors:
(21Tlll)(±X ± 5' ± i) (37)

111e boundalif;.'S of the ccnlral cell in the reciproca1latticc are detcnnined for
the most prut by the eight pla nes normal to these v(.'cto rs at their midpoints.
But the corners oCthe octahedron thus funned are cut by the planes that are the
perpendicular bist;.'Ctors of six other R'Ciprocal lattice vectors;
(2maX±2i) ; (2wla)(:t 2y) ; (21Tla){±22) . (38)
Note thai (21Tla)(2i) is a rt;.'Ciprocallattice v(.'ctor 1)(;.'Cause it is equal to ~ + b:J.
The first Blillouill zone is the smallest bounded volume about the origin, the
truncated octahOOl"On shown in Fig. 15. The six planes bound a cube of edge
47T/a and (before truncation) of volume (47T/a)3 .

•' OURIER ANALYSIS 0.' THE BASIS

When the diffraction condition .6."

= C of Eq. (21) is satisfied, the scatter-
ing amplitude is determined by (18), which for a crystal of N cells may be
written as

Fe = Nf dV n (r) exp( - iC·r) = NSe (:39)

~"
l11C quantity Se is called the structure factor and is defined as an integral over
a Single cell , with r = 0 at one cOllle r.
(
 2 Rtciprocd Lufiu 43

Figure IS BrillOllin zones of the r:lCe-C't'nterro

cubic I:l.ttice, 'OJe cells are in reciprocal sllace.
and the rcciprornl lan ice is bod)' eentered,

Often it is useful to write the electron concentration n(r) as the superposi.

tion of electron concentration functions nJ associalt..-d with each atom i of the
cell. IffJ is the vector to the centcrofatomi, then the function nJ(r - rJ) defines
the contribution of that atom to the electron concent ration at r. TIle total elec-
tron concentration at f due to all atoms in the cell is the sum
•
nCr) = 2: nAr - rJ) (40)
!-'

O\'cr the s atoms of the basis. The dt.'<'Omposition of n(r) is not unique. for we
cannot always say how mllch charge is associatt.'<i with each atom, '111is is not an
important difficulty.
The structure factor defined by (39) may now be written as integrals over
the s atoms of a cell:

(41)
L exp(-iG' rj} f dv nj.p) exp(-iG' p)
}

wherep - r - fj.
\Ve 1I0W define the atomic form factor as

h • f dY "}(p) .xp(-IG' p), (42)

integratcdo,'er all space. If IlJ(P) is an atomic pmpcrty,h is an atomic property.

(
 We combine (4 1) and (42) 10 obtain the structure factor of the basis in tile
form

5c = L h exp(-/C' fJ) . (43)

J

The usual form of this res ult follows on writing for atom j:
fJ = XJlIl + !/j82 + Z;,83 • (44)

as in (104 ). Then, for the reflection labelled by VI> '-1!. 1':J we have
G' fJ = (vlb l + '-1!b2 + ti;)b:J· (::I)al + YJai + z.,a:J
(45)
= 21J{Vllj + '-1!YJ + V:JZj)
so that (43) becomes

5c(VIV2tJ:J) = L h e;\:p(- f2r.{V IXj + '-1!YJ + ~)J (46)

J
The stmcture factor 5 nt:ed not be real because the scattered intensity will
involve 5*5, where 5* is the complex conjugate of 5 so that 5*5 is real.
At a :rero of Sc the scattered intensity will be :rero, even though C is a
perfectly good reciprocal lattice vector. What happens if we choose the cell in
another ".'ay, as a conventional cell instead of a primitive cell, for example? The
basis is changed, but in such a ",oay that tile physiml scattering is unchanged..
Thus for two choices, 1 and 2, it is not hard to satisfy yourself from (39) that

Structure FactCTl' of tile bee LAttice

The bec basis referred to tile cubic cell has identical atoms at x, = !Il ==
%1 = 0 and at X2 = !Ii "" ~ == i. Thus (46) becomes
(47)
where f is the form factor of an atom . 'fhe value of 5 is zero whe never tile
exponential has the value - I, which is whenever the argument is -i1T X (odd
integer). Thus we have
S "" 0 when VI + tljj + V;J "" odd integer ;
5 :: 2[ when VI + Vi + V:l == even integer
1I.-Ietallic sodium has a bee structure. 1be diffraction pattern does Ilot con-
tain lines such as (100), (300), (til), or (221), hut lines such as (200), (110), and
(222) will be present; here the indices (VIV2v:J are refelTed. to a cubic cell. What
is the physical interpretation of the result that the (100) reflection vanishes?
The (00) reflecti on normally occurs when reflections from the planes that
bound tile cubic cell differ in phase b)' 211. In the bee lattice there is an inter·
vening plane (Fig. 16) of atoms. labeled the second plane in the figure, which is
 ! RocipF"fX'tll LIIttice 45
(

Figure 16 F..xlllanation of the ~ of a (100) reflection rrom . bodf-oentcred culJic. bttic<!. 1"he
lilase difference between suca:ssi"e "lanes is .., ~ that the reflected amplitude from two .djacent
plan~ ii I +.,- ''' . ) - 1 - 0.

equal in scattering }lOV>'er to the other planes. Situated midway between them.
it gives a reflection retarded in phase by 11 with respect to the first plane.
thereby canceling the contribution from that plane. The Cl.mccl1a.tion of the
(100) reflection OCCurs in the bec lattice because the planes are identical in
composition. A similar cancellation can easily be found in the hcp structure.

Structure Factor of tlte fcc Lattice

The basis of the k .c structure referred to the cubic cell has identic-.ll atoms
at 000; Oil; iOi; UO. 'Inus (46) becomes
S(t1tt12V~ "" f{l + expl - i1r(v2 + t1;))] + expl - i11(OI + vill
(48)
+ exp[ -i1r(vl + ~]}
If all indices are even integers. S "" 4/; similarly if all indices are odd integers.
But if only one of the integers is even, two of the exponents will be odd multi-
ples of -i11 and S will vanish. If only one of the integers is odd, the same
argument applies and S will also vanish.
Thus in the fcc lattice no reflections can occ::ur for whidl the indice.~ are
prutly even and partly odd. The point is beautifully illustrated by Fig. 17: both
KCI and KilT have an fcc lattice. but KC\ simulates an sc lattice because the K+
alld CI- ions have equoll numbers of e lectrons.

Atomic Form Factor

In the expreSSion (46) for the structure factor. there occ::urs the quantity jj,
which is a measure of the scattering power oftbejth atom in the unit cel\. The
value of/involves the number and distribution of atomic electrons, and the
wavelength and angle of scattering of the radiation. We now give a classical
calculation of the scattering factor .
 ~--

.u

...
l I

~~j
Figure 17 Comparison or.-my reflecti",,! from KCl
",. and Kfl. powder.;. In KC I the " un ,ben of c lL'CIrons
"f K' "ud (.1 iones "re Il<jual, The scattering ampli-
t"dc~ ! (" oJ "nd I (CI - ) arc almost If'xaclly e<lutd. 00
tbat the CI)"W looks to .-raY' as if it were a mona-
tomic' ~imple cubic bllke 0( lattice constant 012.
Only C"cn integers O(:CU' in the renmiol)lI ;"dk-es
when these ""' ba5ed on . cubic lattice on att ire coo-
stanl D. In KB. the form factor of Br- is (Juite differ-
ellt than that of K'. and all ren ..",!ions of thll fcc lal-

-" ti<.-e are present. (COIlI10:5)' of R. "a" No rdstraml. )

-f11e scattered radiation from a single atom takes account of inte rfere nce
effects within the atom. We defined the rorm factor ill (42);
h """ f dV ¥li T) exp( - iG' r) , (49)

with the integral e"tended over the electron concentration associated wjlh a
single atom. Let r make an angle a with C ; then G· r "" C,. cos Q. lfthe d ec·
tron distribution is spheriC3lly symmetric about the origin, then
h '" 211" f d,.,.2 d(cos a) ,~,.) exp(- iC,. cos a)
ec r _ c- 1C ,.
"" 211" f d,. ,.211,;( ,.) • ~---;.;;-'---
,C,
after integration over d (cos a) betwee n - 1 and L TIlUS the form factor is
given by
si.'l' C,"
Jj "" 411 f (1,. 11j(r)r2 , (50)
C,
If the same total election density were concentrated at r "" 0, only Gr "" 0
would contrihute to the integrand. In this limit (sin Cr)/C,. "" I , and
h :<% 411" f d,. nJ..r),-2 = Z ! (51)
 2 Rcciproool umia 47

" A • 0: 109 A

-Ilartrco;- t'oo:\: <;>\.; ..... _

"
, (111)0
0

""OJ
f
, on 0

"'" 1m('
{3JI}O 0
(~20) (42'~ 0
(.5 11 )

.-
0
,
0.' 0,
" .j""filA
0.5
" ' .7 0.6

figure 18 AbliOlute ~'~lletill1t,~llal atomic scattering &etors (Ot metallic alum inum, after Bal.
terman. Chillman, and IJeM a~. Each observed rt'fk,clion il; labeled. No reflL'Ctions occu r ~r
indk-el p;ortly even and partly odd. as predicted b an fcc 0)'5ta1.

the number of atomic electrons. Therefore I is the ratio of the rawation ampli-
tude scattered by the actual electron distribution in an atom to that scattered by
one electron localized at a point.
In the forward direction G ::= 0 , and f reduces again to the value Z. Values
of the atomic fonn factor I for atoms may be found in the l llternational lables
lor x-ray crystallography. Vol. 3.
The overall electron distribution in a solid as seen in x-ray d iffraction is
fairly close to that of the appropriate free atoms. This statement does not mean
that the outermost or valence electrons are not redistributed in forming the
solid; it means only that the K-ray reflection intensities are represented well by
the free atom values of the fonn factors and are not very sensitive to small
redistributions of the electrons.
As an example, Batterman and CQ-workers find agreement within 1 per-
cent in a comparison of the x-ray intensities of Bragg reflections of metallic iron,
COpper. and al uminum with the theoretical free atom values from wavefunction
calculations. 1be results for aluminum arc s.hown itl Fig. 18.
TIlere have been many attempts to obtain direct x-ray evidence about the
actual e lectron distribution in a covalent chemic..! bond, particularly in crystals
•
haVing the diamond structure. The question now lies within the limits of what
can be explored by ~ .ly diffraction methods. In siliCOn at a point midway
between two nearest-neighbor atoms, there is an appreciable increase in elec-
tron concentration over what is e"pected from the overlap of the electron densi-
ties calculated for two free atoms.
Scattering from crystal surlaces is treated in Chapter 19. It is shown in
Appendix A that thermal motion does not broaden a diffraction line, but only
reduces the intensity. Th e lost intensity reappears as long, low wings about the
position of the diffraction line.

QUASICRYSTALS

In 1984 quasicrystals were firs t observed;' these are structures which can-
not be indexed to any Bravais lattice and "which have sym metries intermediate
between a crystal and a liquid." They were first observed in grains of size 2 ILm
in an alloy of AI with 14 at pet Mn. The smaller Mn atoms are each surrounded
by 12 AI atoms arranged at the corneTS of an icosahedron. The structure is made
up of parallel icosahedra attached at their edges. Crystals cannot exhibit the
fivefold symmetry of an icosahedron, but a crystal can be constructed by nucle-
ation at a center cell, followed by outward growth from there. All of the space of
a nodule alnnot be filled by repeating the basic unit (see Figures 19 and 1. 7 fo r
the picture in two dimenSions), although the "parallel" part of the specification
does give a long-range orientational order to the structure. It is perhaps sur-
prising that the ,,-ray diffraction pattem of such a structure can have fivefold
symmetry; that is how they we re first observed.
The known quasicrystals are intermetallic alloys and are very poor electri-
cal COllductors; they are ncarly insulators with a somewhat well-defined band
gap (Chapter 7) at the Fermi level. They are of great interest inte llectually in
expanding the definition of crystal lattice.
A distinctly different crystal difl'raction pattern results fTOm an almost peri-
odic structure, one that is neither rigorously periodic nor simply amorphous (as
for a glass, Chapter 17). An almost periodic structure can he e"pressed in one
dimension if we are given the electron charge density Wdve:
(i,X) "" L[ e" cos[2rrn( 1 + T)xlu1 , (52)
where T is an irrational fraction . TIle terms in 2rrnla by themselves give the
usual lattice with translational periodicity a. When the terms in T are added,
the charge density is almost periodic; that is, the period (I + T)a is not an
integral multiple of the period D, because T is irrational. The period gives a
long-range nonrandom order to the structure. and the long-range order gives a
diffraction pattem, which appears split off from the pattern defined by the
short-range order. 1bis is dominated by the reciprocal lattice points in nJ. but

'D. 1.("1 ltIe and I' J. Steinhardt. 1'I1)'s. Re\ . l.cll 53, 24TI(I984), lllY" Rev. B.1~ , 596(1986),
D. S. Schcchtman and olh"o, 11l)'s. RL'V . l.ett 53. 1951 (1984).
Figurt! 19 A quaskr)'stal tiling in two <Iimcmions. after the "llI"l< or Penrosc. 1llc Iong-r.mge
OrJCntationai order 8.l\d the kmg-l1ongc nonpo..Tiodic order arc shown.

will appear to be clustered and spread Ollt (broadened). l1le diffraction pattern
of a three·dimensiollal quasicrystal is quite different. however; the pattern is
well defined and tan have the fivefold symmetry by which quasicrystals were
first discovered. A computer-generated diffraction pattern with fivefold sym-
metry is sh~wn in F igure 20.

SUMMARY

• Various statements of the Bragg condition:

2d sin 0 = nA .6.1.:"" G.;
• Laue conditions:
~· .6.1.: =21Tt1 ;
• The prim iti\"e translation vectors of the reciprocal lattice are

lIere a .. A:z, 83 are the primitive translation vectors of the crystal lattice.
• A reciprocal lattice vector has the limn
G = vlb j + 112b2 + U;)b3
where v,. VJ!. li3 are integers or zero.
• 'Ole scattered amplitude in the directK>1l 1.:' "" L: + .6.1.: == k +G is propor-
tional to the geometrical structure factor:
 Sc - k i.J e!1p(-irj ' G) "" l:i.J cxpi - i211{xj v l + Yj~ + ljVJ)] ,
whe re j Tuns over the s atoms of the basis, and i.J is the atomic form factOT (49)
of the jth atom of the basis, The expression on the right.hand side is written
for a reflecti on (VIV:2V:Ll, for which C "" vlb l + V2~ + 1J:lba,
• Any function invlUiant undeT a lattice translation T may he expanded in a
Fourie r series of the form

n(r) "" L 'Ie exp(iC ' r)

c
• n le flrst Brillouin zone is the Wigner-Seilz primitive cell of the reciprocal
lattice, Only waves whose wavevector k drBy.'1\ from the origin terminates on
a surface of the Brillouin zone can be diffracted by the crystal.
• Crystal lattice I'irst B"illouill zone
Simple cubic Cube
Body-rentered cubic Hhombic dodecahed.-on (Fig, 13)
Face-centered cubic Truncated octahedron (Fig, 15)

•
• • • •
• •• • •• •• •
• • •
•
• • • • •• •• • •• •
• •
• •• •
• • •• • • • •• • •
• • • •
• •• •• • •
••
• • • •• •• •• ••
• • •• • •
• •• •• ••
• • •• • • • • •• •
• •
• • • • •• •• •
• • •• • •• •
• •• • • • •• •• • • •
•
• • • • •• • • •• •
• • •• •
• • • • • • •• •
• • ••
• • • • •
• • •• • •
• •• • •

Figure 20 I'holograph of the calco,takod FOll.;"r transform (diffl'llclion pattern) of an irosahedrol

','I8,\icr}'Stal along one of the fivefold axes, itlusl ratiro'!: tI,e Il>'-fold .yotm.:lry. TI,e t ransform Is
caku l~tL'(1 from a theoret ical compllle .... gener:oted model, b) ( Noel JaL'<Ib.
 Problems
I. lnterplanllr uparntion. Consider a plane Itkl in a crystal lattice. (a) Prove that the
reciprocallauice vector C ... hb, + kb,.; + lb;, is perpendicular to this plane. (b) Prove
that the distance between two adjacent parn.l1e1 planes of the lattice is d(hkJ )-
2n11Gl . (e) Show for a simple cubic lattice that d 2 = crl(hl + k2 + f) .

2. Hf!%agonal space tallice. The primiti ... e translation vcctors of the hexagonal space
lattice may he taken as
I1t
al - -(3 al2)i + (af2)Y ;
(a) Show that the volume of the primitive cell is (3 11tI2p1c.
(1)) Show that L11e primitive translatioo! of the redprocal lattice are

b:i - (21'rlc)i •
so that the lattice [5 its own recillrocal, but with a rotation of axes.
(e) Describe and sketch the first Brillouin zone of the hexagonal space lattice.

3. Volume of Brillouin zone. Sho\v that the \'Oiume of the first BrillOlJin zone is
(2.,;fNt:. where Vt: is the volume of a crystal primitive cell. Hint: The "t:>!.ume of a
Brillouin zone is equal to the vOluine a
lhe primitive parallelepiped in Fourier
space. Recall the vector identity (c " a) " (a " b) = (c' a )( b)a .

4. Width of diffraction marimum. We suppose that in a linear crystal there are identical
point scattering centers at every lattice point p,., = ma, where m is an integer. 8y
analogy with (20) the total scattered radiation amplitude will he proportional to F =-
I expl - ima ' 6kJ. The sum over At lattice points is

F = J - CIIJ?!-IM(a . 6It))
J t::qJ[ - I(a . 6.k)]
by the ure of the series
"-1
'" ,..----
/>l
~ 1-:1:
.. --0 1-:1:
.

(a) The scattered intensity is proportional to IFf Show that

t
IFlt • PF ... sin iM(a • 6k)
sintl{a . Ilk)

(b) We know that a difTrnction maximum appears when a • Ilk - 21Th, where h is an
integer. We change Ak slightly and define ~ in a' 6k = 21Th + ~ such that IE gives the
po5ition of the first zero in sin IM(a' l!.k). Show that IE = 2rr1M. 50 that the width cI
the diffractiun maximum is proportional to liM and can he extremely narrow lOr
macroscopic values ofU. The same resu lt holds true for a three-dimensional crystal.

5. StruC/UTe factor of diamond. The crystal structure of diamond is described in Cilap-

ter I . The basi3( sists of eight atoms if the cell is taken as the conventional cube.
"
•~, '"
!.
i ..,
(I ' I )
.."
!
~ .., - PH)
.~

f
z
. - - o.4S'
0.65" - I-
-, \-0.70"
....=i\::
.,.
Counter .,.,.11000 2f

fo"jgure 2 1 Neutron diITI"..ction pallen. for powtlered diamond. (After G . &con.)

(a) Find the struchll'c factor S rL this basis. (b) Find the zeros of S and show that the
allowed reflections of the diamond structure satisfy VI + vt + V3 '" 4n, whtlre all in-
d ices are even Rnd ,. is any integer, or else all indices arc odd (Fig. 21). (Notice that
h, k, I may be writte n for VI , Vi. V:J and this is oRen done.)

6. Form fadQr of atomic hvdrogen. For the hydrogen alom in it5 ground state, the
numbe r de nsity is n(r) = ( ~ - I exp(- 2rlUf). where flO is the Bohr rad ius. Sh(1l.ll that
the form f:K"tor is Ie = 161(4 + C2~'l.

7. Diatomic line. Cln~ 'd er a linc of atoms ARAB. . AD, with an A- 8 bond lenb<th of
ia. The form fadou are flo. ffj for atoms A , B, re spectively. The incident beam or
x-rays is pcrpendiC\Jlar to the li ne oiatoms. (a) Show that the interference condition Is
1'1" "" a cos e,
where 6 i5 the angle between the dilTr.lCI:ed beam and the line of
atoms. (b) Show that tile intensity of the d iffracted beam is proportional to !fA - 11112
for 1'1 odd, and to IJA + fll12 for 11 even . (c) Explain what happells if IA - lB.

References
X-HAY DIFFRACTION
C. S. Barrell and T. B. Massal~ki , Strwturf' of metau: CTyStaUograpl'lc mffllOdJ, prin.cfplu, datil .
3 rd. eel. rev., Oxfonl 1960. EXCt:lIenl guide 10 tbe pract ical solution of relatively ~implll struc-
tures .
M. J. fluergcr. Cont£mpOrcry crydci/ogrul'h!l , McCr.l.w-lIi1l , 1970. A Hne introd"ction.
B. E. \\'''lTtln , X-rll!l diffrlldlct1 . Addi.oo-\\'Cli]cy. 1900.
C . JallOt. Qt<mN-ryslIILr: II primer. Oxfurd , 1993.
D. I'. DiVincen7:0 and I'. J. Steinhard t , ..ods., Quasicryslll/s, World ScientiFIc, 1991.
F\. O'01It and T. lansen , ~ E><titation! in inoomll1en:1urnle CT)·~lai I,hases, Solid slate ph ysics 41,
R

201 ( 1988).

NEUTIION DIFFRACTION
G . E . Baron. Neulrem diffraction. 3rd <!d., Od"ord , 1975.
S. w . !..oo.'csey, Tlu:ory of Ill;Ulron SCClflerizlt!,from eontkmed maltcr, 2 vols., Oxford , 1965.
W. Marshall and S . w . Lovescy, TI,/IOTy of llumMI neutron xutlerl7lg, O!tford, 1971.
 (

3
Crystal Binding and Elastic Constants

CRYSTALS 01<' IJIOERT CASES

IO""IC CRYS'rALS 00

OOVALE"~r CRYSTALS 72

ME'fALS

HYDROGEN BONDS 7.
ATOMIC RADII 77

ANALYSIS OF ELASTIC STRAINS 80

ELASTIC COMPLlANCE AND S'fIH'NESS CONSTANTS 83

ELASTIC WAVES IN CUBIC CRYSTALS 87

SUP.tMARY 00

PROBLEMS .,
1. Quantum solid U2
2.
3.
4.
Cohesil'c energy of bee and fcc neon
Solid molecular hydrogen
Possibility of ionic crystals R+R- ,
.,"'
!l3
5. Linea r ionic crystal " 93
6. Cubic znS structure 93
7. Divalent ionic cl"}'$tals !l3
8. Youngs modulus and Poisson's ratio .3
9. Longitudinal wave velocity .3
10. Transverse W3\'C yciocity
••••
.
11. Effective shear constant
12. Determinantal approach 94
13. Cenernl propagation diredioll
14. Stability criteria
REFERENCES
""
 CJ)-.blb" .. a.g<II1
(..... de.. Waal.)
.j

Fiplrfll n.e priflci~ typr$ of t'1')'1ital1ifl<l bifldi'!&- 10 (al neulrallltOfflS with ck>sal ..k-ctroo shells
are bound togeIllflf" ...."",.kl), h). the 'an der Waals fon.?s aDOdated wil h nuctuatioos m the charge
di~t rib"tioo •. In (b) electrons are tnnsfnn:d from the alkali MOO ns 10 the h.logen atoms, and the
re.ulting io<lf ""' heM tflg.,ther by IIttr.>di,"C electrostatic fi:wt.es beI"''fIffIII~ posilh-e and ncgat;'-e
ions. In (e) the , ..knee .,Iect.."" 1I~ t'&.,n IIway from each .. U...I; 1110111 to form II comm"....1electron
sea in whicb the po.,,'i,·., ions are di>pcrsed. In (d) the neutral ~t", 'Ilfl"ar to be bound together
by th., ""CI"lapping pru-ts of lh';-;r electron t1i~lnbut;onli . ~
 CHAPTER 3: CRYSTAL BINDING

In this chapte r we are concerned with thejJuestion: What holds a crystal

together? 11le attractive electrostatic intemctipn between the negative charges
of the electrons and the positive charges of the nuclei is entirely responsible for
the cohesion of solids. Magnetic forces have only a weak effect 011 cohesion. and
gravitational forces are negligible. Specialized terms categorize distinctive situ-
ations: exchange energy, vall der Waals forces. and covalent bonds. TIle ob-
served difiere nces between the forms of condensed matter are ca\ISed in the
final analysis by differences in the distribution of the outennost electrons and
the ion cores (Fig. 1).
l11e cohesive energy of a crystal is defined as the energy that must be
added to the crystal to separate its components into neutral free atoms at rest,
at infinite separation, with the same electronic configuration. The term lattice
energy is used in the discussion of ionic crystals and is defined as the energy
that must he added to the crystal to separate its component ions into free ions at
rest at infinite separation.
Values of the cohesive energy of the crystalline elements are given ill
Table 1. Notice the wide variation in cohesive energy hem-een different col-
umns of the periodic table. The inert gas crystals are weakly bound. with cohe-
sive energies less than a few percent of the cohesive energies of the elements in
the C, Si. Ge ... column. The alkali metal crystals have intermediate values of
the cohesive energy. The transition element metals (in the middle columns) are
quite strongly bound. 11le melting temperatures (Table 2) and hulk modulii
(fable 3) vary rough ly as the cohesive energies.

CRYSTALS OF INERT CASES

l1le inert gases form the simplest crystals. 111e electron distribution is very
close to that of the free atoms. Their properties at absolute zero are summa-
rh.ed in Table 4. The crystals are transparent insulators, weakly bound, with
low melting temperatures. The atoms have very high ionization energies (see
Table 5). 111e outermost electron shells of the atoms are completely filled. and
the dislribut'ion of electrol"l charge in the free atom is spherically symmetric. In
the crystal the inert gas atoms pack together as closely as possible l : the crystal
structures (fig. 2) are all cubic close-packed (fcc), except He3 and He"'.
'Zero-poinl ,nol;on athe atolTls <Idnrlic ,m",rgy ~I absolule 7.ero) is a (luan lulTI dfed that Ill.~,
• tb.-ninant role in lie.) and He~. They tJo not 5()lidif~ at ~ro pre» ure e,'",n at "">wlule ~.em. The
a"..rage f1l1ctu~lion al 0 K a a He atom from iU ~lIilibriUln poSition Is a the onler of 30 10 40
percent of th", neare!;t-Ildghbor dista.nce. The heR\"ier the Rlom_ the less import&llt ~re the :tero-
s
(>Oint effects. lf we olTlit :wro-poinl motion, we calculllte a ,nola. "olome 01"9 ern mol- ' ror solid
helillm, as compared with theobser"\'ed ,,,]u~of'l:1. 5 &lid 36.8 con.) ,nor' fw liquid Ile~ and liqtUd
~, respecti'"el~. In Ill<" '""OUnd ~ta.te ofhcliu.n ""e must ta1.e ..:wunt of the ?rI"O-point lllOlion of
tho: atoms. ~
Fisure 2 Cubic doJ.e.packed (kc) crysud structure of (he inert gl\M'S Ne , Ar, Kr, and Xe. l he
lattice pll...."neters of the cubic cells arc 4.46, 5.31 , 5.64, and 6.13 A, respedi"ely, ,,( 4 K.

What holds an inert gas crystal together? nle electron distribution in the
crystal cannot btl significantly distorted from the electron distribution around
tlle free atoms, because the cohesive energy of an atom ill the crystal is only 1
percent or less of the iOlli7.ation energy of an atomic electron. 11ws not much
enerb'Y is available to distort the free atom charge distributions. Part of this
distortion gives the van der \Vaals interaction.
Vall de.- Waals-London "Iteraction
Consider tv.'O identical inert gas atoms at a separation R large in compari-
SOil with the radii of thtl atoms. \Vhat inttlractiOlls exist between the rn.'O neutral
atoms? If the charge distributions on the atoms were rigid, the interaction
between atoms would btl zero, bt:Causc the electrostatic potential of a spherical
distribution of electronic charge is canceled outside a neutral atom by the elec-
trostatic potential of the charge Oil the nucle us. Then the inert ga.~ atoms could
show no cohesion and could not condense. Dut the atoms induce dipoltl mo-
ments in each other, and the induced moments cause an attractive interaction
betv.'Cen the atoms.
 Table 1 Cohe~i "e e nergie~

Energy required to form separated neutral atoms In their ground electronic B e N 0 F

Ne
LI Be
state from the solid at 0 K at I atm . The data were supplied by Prof. Leo 561 711 474 251 81 0
1 92
158. 320.
1.63 3.32 Brewer in units kcal per mole, revised to May 4, Hm, after LBL Report ~~1 1:3t.: 4.92'; -2:W o.s;c
0:026
37 .7
N,
107,
1.113
2567
76 .5

M.
145.
1.51
34,7
3720 Rev.

.~,.,---
kJ/mol
= =';'=====O-e:VlatOm
kcallmol
-134 - 170. -113.4

~
AI SI
.~ 1~i3~ 4.Sr
P
327 . 446. 331.

78.1 106,7 79.16

60.03- 19 .37 0.46

S
275,

65,75 32,2
cr
135 .
J43." ·Z~ }~ ·ra~ §oaol
I"
7.74

185

K Ca S, T; V C, M, Fe Co NI C, Z, G, G, A, S, B, K,
901 178 . 376 468 . 512 395. 282. 413, 424 :....- 428 . 336, . .E0__ _ 2Z.L. 372 285 ,3 237 118. 11 .2
OJi3~ 1.84..... 3.9~ 4.85 sr3f.... 4 . 1 1f~ -~.92 4.28-' 4.39 4.44 .... -3.49- 1,35 2.81 3.8~ 235s 2.4F T22 O)ftli
21.54 425 899 11 1,8 122.4 945 6 7.4 987 1013 102 .4 80.4 31 .04 648 88.8 68.2 56.7 28 .18 2.68
Rb S, V Z, Nb Mo T, R, Rh Pd AS Cd I, S, Sb T, I X,
822 166 422 603. 730 . 658 661 650. 554. 376 284. "2 303 265. 211
243. 107. 15.9
0.852 1.72 4.37 6.25 7.57 6.82'" 6.85_ o.74~ 5,75_ 3.89 '2.95"" JJ13 232- ·3.1. r 27~_ 2:"1 9- rfr" 1).'1"6"'
1964 39 .7 100 ,8 1442 174.5 157.2 158. 155.4 1325 89 .8 -58- 0 26.73 58.1 724 63.4 50.34 ~2562 3.80

--
C, Ba HI T, W R, 0, I, PI A, HS TI Pb BI Po AI R,
776 183, "
431. 621. 782. 859. 775. 788 670. 564 . 368. 65. 182 . 196. 210. 144.
1 . SO~
195
0.804 1.90 4.4.1 644 8l0..... 8.90_ ~93 8.17 6.9~__ ~.84 . 3.81- Q:ij7 ..188~ J.Q:f' 2·tB 0..202
1854 437 103 1 148.4 186.9 205.2 185.2 188.4 160. 1 134 7 8796 15.5 43.4 46.78 SO.2 345 466
F, Ra Ao r"- "
-- 160
1.66
410
4.25
Ice
417
Pr
357 .
Nd
328 .
Pm Sm
206
E,
179
Gd
400
Tb 0,
391. 294. 302
Ho
317 . 233
E,
154. 428
Tm Vb
" i
4.32a. ~.70 _3A.Q.... ~...... 1.86>0, ':"4J tlC A.O ~ 3&4: .:U~ ~29;, :-2,42._ 1.130 4~3-""

I
382 98 ... 14
99 .7 85.3 785 493 428 95.5 93.4 70 .2 72 .3 75.8 55.8 37 .1 1022
Th p, Np p, Am Cm CI E, Fm Md No
U 8k
"
598
o.20""f
1429
- 536.
5:$.
128,
456 347 . 264 .
;(7j .,.J,60_ -2..13
109. 83 63 °
385
3:gg
92 ,1
- ~':!~":":: ~ ~

~
 ~

Table 2 Melting points, in K.

Ct B. B C N F N.
°
453.7 1562

N, M9
- (After R. H. Lamoreaux, LBL Report 4995)
- - 2365

AI 51
63.15 54.36 53.48 24.56

P 5 CI ..
371.0 922 933.5 1687 w 317 388.4 172.2 83.81
, 863

K C, So TI V C, M, F. Co Nt C, Z, G, G. A, 5. B, K,
336.3 1113 1814 1940 2202 2133 1520 1811 mo 1728 1358 692.7 302.9 1211 1OS9 494 265.9 115.8

Rb S, V Z, Nb Mo To R, Rh Pd A9 Cd I, S, Sb T. I Xo
312.6 1042 1801 2128 2750 2895 2477 2527 2236 1827 1235 594.3 429.8 505.1 903.9 722.7 386.7 161.4

C, B, C, HI T, W R. 0, I, PI A, H9 TI Pb BI Po AI R,
301 .6 1002 1194 2504 3293 3695 3459 33<>6 2720 2045 1338 234.3 577 600.7 544.6 527

~
F, R, Ao I"
973 1324 C. p, Nd Pm 5m E, Gd Tb oy Ho E, Tm Vb C,
1072 1205 1290 1340 1091 1587 1632 1684 1745 1797 1820 1098 1938

Th p, U Np p, Am Cm Bk ' CI E, Fm Md No lw
2031 1848 1406 910 913 1449 1613 1562
 T.able 3 h:othc nnal bulk mod uli! and oompreuibilitlc5 at room
H,. te mperatu re Held.
0.002
r~1
Mer K. Gschnl....dner, Jr., Solid state phYSics 16, 275-426 (1964); several
data are frnm F. Birch, in Handbook of phyrical coNlantl, Geological Soci-
to 00
miLl
ety of America Memoir 97, 107-173 (1966). Origin:u refe rences should be
consulted when values are nceded for research purposes. Values in paren- B C [41 N 1.1 0 I ,~ i Ne If)
U IBe theses are estimates. Letters in parentheses refer to the crystal form. Let· 1.78 4.43 0.012
;=U~ -k~t~ &9:: =--I~I 0,010
~1-+---1
leTS in brackeu refer to the temperature:
raj - 77 K; [bl - 273 K: leI - 1 K: (d ) = 4 K; leI = 81 K.
~~ -o:-~.
fc;=I;':
Na Mg
Bulk modulus in units 10lJ dynlc~ or 10" Nlm>
AI lSi p ,w IS "' CI At 1,1

I~.t~' Compressibility in units 1O" f cm'/dyn or 10' " m'1N

Q.72,2 Q.,9.?8 O}04. 0.178,
1.385 l ..ol.2 3.29 ~ 5.62- ..... 0.013
1 7.a.:~

0.032
" ~0.152 0.435
'l"-S -6:58
Sc Ti
"..,... IV
.2.",... .51 ==
"IC' 1
M

-" '''',. -n-526 -n.,-

1.051 1.619 1.901 0.596 1.683 1.914 1.86
~ -~......
C, IZ" G" " ~G' lAS
IF. ICo IN'n.m 0."
.6;-5~ 'O~S22
...~~ .~ ~--
1.37
-"-- "-0- -:
0.598 0.569 0.772 0.394 0.091
' 1~6Z: l"'~
~
,:>o~ '2~5'-
~~
ISe I"d"'"
~ ~ --~---
.
0.Ql8
56- -

"b
~ I' ID:no._INb= I~ Ik
0.031 0.116 _D.l6§.. _D.Bl! 1-702 2.725, (2.~7) 3: 208 2.70~.
32~ J!.li,;... 3-1L ~ ]©~ 1).= O~_.~
I~ I~ I~ 1~8_08 1- I'"
}.007 0.467
Cd 0.411
,1.]) _
S' "" Sb
O.2~~ :.='
2-1!J.l
ASSJ 0".9. U-:C :2;R.- ..Q.9tll .2..!ilA ~;35
,T' ,
. -..
Xe
1"'-'--1
I"' I'" ~ Ih IW ~ ~ '" ,~ ,~ Hg'O' ITi
0.020 0.103 0.243 1.09 2.00 3.232 3.72 (4.18) 3.55 2.783 1.732 a.3S!. 0.359 0.430 D.lI? (0.26)
Cs
'5'1,
Po
~w.. 3.l?. O.~?; ,O.M[! .QJ~ .Pi-209' (O~ ·0.281 Q.,352 O.S7Z, I.~§.Q h79... kll... 3,.17,., (3~L
- :;
At IPb I"'
-=--:
I""
"
Fr

~
I Ra

Q..
l Ac
,t2.020)1CQYl2!\O.J~
l4l.':.
I'
Ce (,J
0.239 '" INd Ipm Ism ;:IE' Gd
0.306 0.327 (0.35) 0.294 0.147 0.3B1. 0Tb I lit
Ho I E, Tm
.3_99 0 .384 0.397 _0.411 0.392 _0.133
Vb I I O.~ll
Lu
i
<l.l1lZ ~ 2Q<;;: .12M! ,p O -§.80_ .~zm; .. 2~S4~ .2~,.. 2 52.? .1.52..: .2.§.3 _
L

J
INP ~ ' lAm I_I~_I_ -I ~I
"k Cm Cf Es Fm Md No IJ
Th IP' IU
D.54! (O}.., 0.987 -,0 .•8) 0.54
Il K:n;m: 1.6'r; tr.st: ·~A.-..~ -I ~ -
_
--
- I'~I -
~t---I-
1-1 ~
.
1--- ·1-,· -I
--- ----1-
Figure 3 Coordinates of tbe twn oscillators.

As a model we consider two identical linear harmonic oscillators I and 2

separated by R. Each oscillator bears charges ±e with separations x, and X2, as
in Fig. 3. The particles oscillate along the x axis. Let p , and P2 d enote the
momenta. The force constant is C. TIlen the hamiltonian of the un perturbed
system is

1 2 .2 1 2 .2
"0"" - PI
2m
+ iC1J + -P2
2m
+ i Cri . (1)

Each uncoupled oscillator is assumed to have the frequency ~ of the strongest

optical absorptioll line of the 3tom. l1lUs C = ~.
Let 1ft be the coulomb interaction energy of tlle two oscillators. The g€Qm~
elry is shown in the fi gure. TIle internuclear coordinate is R. TIlen

(CCS) (2)

in the approximation Ix.l. IX21 .c:s;: R we expand (2) to obtain in lowest order:

" 2e2XIX2
I il"! - R3 (3)

Table 4 Properlin or inert gas crystals

(ExtrapolalL'tI loOK and zero pressure)

PllramelcB in
looi-
Lennard-Jones
7.<ltJOIl

.....,
ExperiHlent'll potential.
Nearest- potential
collt'sive Eo,. 10
neighbor "'r~
dista~. Melting RtOHl. <. u.
in A 1e)lrool eViatolll poont. K .v in 10-1& e rg in A
II, (liquid at zero prCSSUf(') 24.58 14 2.56
N,
., 3. 13
3 .76
1.88
7.74
0.02
0.0fJ0
24
84
21. 56
15.76
50
167
2.74
3.40
K. 4.01 11. 2 0. 116 117 14.00 225 3.65
X. 4.35 16.0 0. 17 161 12. 13 320 3.98
 Table 5 Joni7,ation energies
H
13.595
The total enerlO' required to remove the first two electrons is the sum or the ~
24.58
flrst and seco nd ionizatiOn potentials. Source: National Bureau of Standarus

. -
7898
CilUllar 467.

LI
• B C N 0 F N.
5.39 9.32 8.30- .1T.Z€" 14.54 T3.61 17A2 21.56'
81.01 27.53 33.45 35.64 44.14 48.76 52.40 62.63

N,
S.14
52.43
M,
7.64
22.67
, - . Energy 'to remove one electron, In eV ...-
Energy to remove two electrons. in eV
- -- ,
AI SI P S CI
'S.9!C If.lS.... f.O-:SS'· Jlr3r J3.~.1- 15.i"tf
24.80 24.49 30.20 340 36.S1 43.38
.
K
36.15
C, S.
4.34- ·6.[f- 6:56
17.98 19.45
n
6.sT
20.46
V
6.74
21.39
M.
C,
6.7G:'
M,
7.'43":
23.25 23.07
F. Co NI
7:-90.." 7.~ l.W 7!fT"" 9.39 ....
Co
24.08 24,91 25.78 27.93 27.35 26.51
Z, G,
JOO:
G. A,
Las.;..,. 9.§C
23,SI 30.0
-.-
S.
9.75.
31.2
B,
~ 14.00
33.4
K,

38.56

Rb S, Y Z, Nb Mo T. Ro Rh Pd Ag Cd I, S, Sb Te I X.
4.1S 5,69 6.5 .,... 6.95 6.77 7.1S 7.28 7.36 7.46 _ 8.33 7.57 8.99]" -5.78 7. ~ 864 9.6T"" iM5 12)3
31.7 16,72 IS.9 20.98 21.22 23.25 22.54 24.12 25.53 27.75 29.05 25.S9 24.64 21.97 25.1 27.6 29.54 ".3

C, B, HI T, R. 0, I, Ao Hg
- ",
-
La W PI TI Ph BI Po At
3.S9 .. 5.21 5.61 7. :: 7.88 7.98 .: 727 8.7 ...... 9.- : : 8.% 9.22 '" IO.43' 6:11 nr T29 8.43 : 10.74_
29.0 15.2 1 17.04 22. 24.1 25.7 24.5 26. 27.52 29.7 29,IS 26.53 2244 23.97
"
~
F, R, A.
"' p, 1
-- -- --
C. Nd Pm Sm Eo Gd Tb Dy Ho E, Tm Yb Lo
5.28
15.42
6.9 "!.
19,0 6.9f'" 5.76 ~ - 5~ 5.67..... 6.t6 6.74 . ~ 6.2--=- S.O=.
1-
Th p, U
4.
Np Po Am Cm Bk

-
CI
1=
E,
-- --Fm -Md --No -
----
L,

- "
"
The total hamiltonian with the approximate form (3) for:H1 can be c.Iiago!lal~
ized by the normal mode transformatioll

(4)

or, 0 11 solving for XI and X2 ,

(5)

TIle subscripts s and a denote symmetric and anlisymmetric mooes of motion .

F\lrthe r, we have the momenta P., V.. associated with the Iv.o modes;

(6)

TIle total hamiltonian 110 +!Jl, after the transformations (5) and (6) is

"= [ ~nP:+i(c - ~)x!]+[~p:+~(c+~)x:] . (1)

TIle two frequencies of the coupled oscillators are found by inspt.'Ct:ioll of (7) to
be

with Ct\l given by (Clm)"'l. In (8) -we have expandc.>d the square root.
The zero point energy of the system is ili(tu, + w,,); because of the interac-
tion the sum is lowered from the uncoupled value 2 . iliWo by

• "
'( 2<1')' :
t:..U = M(!J.w + 6w ) = -/i,.,- . - - -
-u 8 cn3
-A
- .
nO (9)

TIlis attractive interaction varies as the m inus sixth power of the separation of
the tv.·o oscillators.
111is is called the van der waals interaction, kTl()Wll also as the London
interaction or the induced dipole-dipole interaction. It is the principal attrac-
tive interaction in crystals of inert ~es and also in crystals ot many organic
molecules. The interaction is a quantum effect. in the sense lhat 6.U _ 0 as
Ii - O. Thus the 7.ero point energy of the system is lowered' by the dipole-
dipole coupling of Eq. (3). 11le van der Waals interaction does not depend for
its existe nce on any overlap of the change densities of the t\'.-o atoms.
An approximate value of A for identical atoms is given by Ii~. where
IiWo is the energy of the strongest optical absorption line and a is the electronic
polarizability, Chapter 13.
(
 3 Cty"'" Binding 63

• •
~~
• •
~
Figure 4 Eiloctm' IIc dwrg.: distributions
overlap as 1110111$ approadl. The solio d,...
des dt:oote the IIlIc:I'~i .

Repulsive Interaction
As the h\1) atoms are brought together their charge distributions gradually
overlap (Fig. 4), thereby changing the electrostatic encrgy of the system. At
sufficiently close separations the overlap energy is repulsive, in large part be·
cause of the Pnuli exclusion principle. TIle e lementary statement of the princi.
pie is that two e lectrons cannot have all their quantum numbers equal. When
the charge distributions of two atoms overlap, there is a tendency for electrons
from atom B to occupy in part states of atom A already occupied by electrons of
atom A, and vice versa.
The Pauli principle prevents mu ltiple occupancy, and electron distribu·
lions of atoms with closed shells can overlap only if accompanied by the partial
promotion of electrons to unoccupied high energy states of the atoms. Thus the
electron overlap increases the total energy of the system and gives a rep\llsive
contribution to the interaction. An extreme example in which the overlap is
complete is shown in Fig. 5.
\Ve make no attempt here to evaluate the repulsive interaction 2 from first
principles. Experimental data on the inert gases can be fitted well by an empiri·
cal repulsive potential of the fonn wnl2, where 11 is a positive constant, when
used together with a long·range attractive potential of the form of (9). 111e
COnstants A and Jj are empirical parameters determined from independent
measurements made in the gas phase; the data used include the vinal coeffi·
dents and the viscosity. It is uSIJ.'l.1 to write the total potential energy of h\1)
atoms at separation 11 as

U(R) ~ 4'[(~)" - (~n ' (10)

where E and (T are the new parameters, with 4Ed' - A and 4Eif2 - n. 111e
potential (lO) is known as the Lelll13nJ.jones poteQtial, Fig. 6. 'The rorce be-
tween the h\1) atoms is given by -dU/dR. Values of E and ugiven in Table 4 (:an
~ ",'erlap energy naturaU)' depends 00 the radial dhtribulion of charge about."...h _ .
The mathemaHcai calculation is always complicateO a'Cf'l ir the ch;lrge distribution is kflOw...
 ·
O 0 O II
hi
II

"I
II ~
.,
Total """I«,WI
"'\"'8Y·- ~.311 ~ \'

.' igure 5 n.eeffect of F\iuli principle Oil the rqrobive energy: i ll an extreme eUm l)Ie, two
h)·drogcn atoms ~ ptJllhed together u"lil the protons arc almost ill o;»r,tad . ' nc energy of the
electron syste m alone can be taken from obser\'atious on atomic He, ..filch has two electrons. In
(a) the e lectrons nne anlipara11el sp ills and the Pauli principle ha.s no effect: IJIIl electron s are
bound by -78,98 e \'. In (b) the spills are parallel: the Pauli principle I"orces the promotwn of an
ckctroo from a Is t oo-bilal of H 10 11 2.r t orl;. biI. of He. The eioctrons now are bound by
- 59.38 e\', k.u than (a) by 19.00 e\'. 11lis is the amount by which the ~li lH'inciple hal increased
lhe repulsion. We ha,'c om illecl thc repuJsi\"l~ coulomb e nergy of the two proton s. ....h ich is thc
same In l>oth (Il) and (b).

,
,
•
,
U(El)/w
,
,
, \
, ., " " " L , ,,
FIgUre 6 . 'onn orlhe Lennard-Jone! pote ntial (l O} .. hkh describes the I"ten etio" or lWl,l lncrt gas
atom s. '[be ml,,;mum occurs at RIfT "" 2''''.
1. 12. Notico: how sl CCp the curve i'I" iruide the mini-
mum, and how flat it iii outside the m llumum , The .'IIl ue of U at the m ini mum hi - ~ ; and U - 0 at
R - u.
 3 Cryslal Binding 65

be obtained from gas-phase data, so that calculations on properties of the solid

do not involve disposable pardmeters.
O ther empirical forms for the repulsive interdction are widely used. in
particular the exponential fonn ). exp(- R1p), where p is a mea~ure of the r.mge
of the interaction. This is generally a~ ea~y to handle analytically as the inverse
power law form.

EquiUbrium Lattice Constants

If we neglect the kinetic energy of the inert gas atoms, the cohesive energy
of an inert gas crystal \s given by summing the Lennard-Jones potential (10)
over all pairs of atoms in the crystal. If there arc N atoms in the crystal, the total
potential energy is

U~ ~ IN(4')[L' (..5:-.)" - L' (..5:-.)'] , (ll)

J p!iR J PuR
where P!lR is the distance between reference atom i and any other atom j,
expressed in terms of the nearest neighbor distance R. '111e factor i occurs with
the N to compensate for counting twiee each pair of atoms.
The summations in (11) have been evaluated, and for the fcc structu re

~' -12 '<" - , = 14.45392 .

.c... P" = 12. 13188 ; £... P!I (12)
J J

There arc 12 nearest-neighbor sites in the fcc stnlcture; we sec that the series
are mpidly converging and have values not far from 12. TIle Ilearest neighbors
contribute most of the interaction energy of inert ga~ crystals. TIlC correspond-
ing sums for the hcp structure are 12.13229 and 14.45489.
If we take V",. in (ll) as the total energy of the crystal, the e<Juilibrium
value Ro is given by requiring that Vtot be a minimum with respect to variations
in the Ilearest neighbor distance n:
dUo. [ if'
~ = 0 = - 2NE (12)(12.13) RI3 - (6XI4.45) R7
,j'] • (13)

whence
f4ja = 1.09 • (14)
the same for all elements with an fcc structure . The observed values of f4ju,
using the independently determined values of a given in Table 4, are:
Nc Ac Xc
1.14 I.ll 1.09
lbe agreement with (14) is remarkable. The slight departure of IVa for the
lighter atoms from the universal value 1.09 predicted for inert gases can be
"
explained by 7.cro-point quantum effects. From measurements on the gas phase
we have predicted the lattice constant of the crystal.

Col,esive Energy
The cohesive energy of inert gas crystals at absolute zero and at 7.ero pres-
SUfe is obtained by substituting (12) and (14) in (11):

Uro.(R) = "'V'[(1213)(~)" - (1445)(;),] . (15)

and, at R = flo.
(16)
the same for all inert gases. This is tbe calculated cohesive cnerb'Y when the
atoms arc at re.~t. Quantum-mechanical corrections act to reduce the binding
by 28, 10,6, and 4 percent of Eq. (I6) for Nc. AT, Kr, and Xc, respectively.
"Ole heavier the alom, the s maller the quantum correction. \Ve can under-
stand the origin of the Ijuantum correction by consider-llion of a simple model
in which an atom is confined by fixed boundaries. If the particle has the quan-
tum wavelength A, where A is dt;!terminoo by the boundaries, then thc particlc
has kinctic energy p'l/2M:= <,"A)'l/2Af ..vith the de Broglie relation p:= hlA for
the connection between the momentum and the wavelength of a particle. On
tltis model the quafltIl1ll1.cro-point (:orrectioll to the energy is inversely propor-
tional to the mass. Thc final calculated cohcsivc energics agrcc with the cxperi-
mcntal values of Tablc 4 within 7 to 1 percent.
One consequence of the quantum kinetic energy is that a cl)'Stal of the
isotope Ne 2ll is observed to have a largcr lattice constant than a crystal of Nc22 .
°nIe highcr quantum kinetic e nergy of thc lighte r isotope expands the lattice,
because the kinetic energy is reduced by expansion. The observed lattice oon-
stants (extnlpolated to absolute 1.cro from 2.5 K) are Ne !1V, 4.4644 A; Ne2\!,
4.4559 A.
IONIC C1\YSTALS

Jonie crystals are made up of positive and negative iOlls. The ionic bond
results from the electrostatic interaction of oppositely charged ions. Two com-
mon crystal structures found for ionic crystals, the sodium chloride and the
ccsium chloride structures, were shown in Chapter L
nle electronic configurations of all ions of a simple ionic crystal correspond
to closed electronic shells, as in thc incrt gas atoms. III lithium fluoridc thc
configuration of the ncutral atoms are, according to the periodic table in the
front endpapers of thi..~ book. Li: Ih, F: 1;2.s'2p iS. T1\e Singly chargcd ions
have the configurations Li+: 1;. F-: I I2s~p6. as for helium and neon, respc<....
tively. fnert gas atoms have closed shells. and the charge distributions are
spherically symmetric. We expect that the charge di! uHons on each ion in
\
 3 Cry!lal Dinding 61

)-"igu,.., 1 electron density distribution in the base

plane of NaCl, ..ner x-ny studies by C. Sehoknecht.
'I1le numbe.. on the contou .. gin' the relat;"e elec-
tron COIlCCflt ..... tion.

c.
+ 5. 14 cV

looi ... tiot,

cn<:rtO'
-a c.,
+ GJ

Figure 8 The energy per molecule unil of a crystal

GJ +
c..
- c..
+ 3.61 co\'

'''"'"'
affinity
of wdium chloride is (7.9-5. 1 +3.6) " 6.4 eV
\o-.>\."I" than the energy of sepanted neutral atoms.
'I1le lattice energy with respect to separaled ions is
7.9 eV per molecule unit. All \-alue~ on 11,e rtgure are
experimental. Values of 11,., ionization energy are
given ill Table 5, and \-alues of the electro" affinity
are given ill Table 6.
G) +
c.. c.
- l,<)· .. al
+ 79,·\
Co,hn;
rnt.,.~\

an ionic CryStal will have approximately spherical symmetry, with some distor-
tion near the region of contact with neighboring atoms. l1lis picture is con-
finned by x-ray studies of electron distributions (Fig. 7).
A qUick estimate suggcsts that wc are not misguided in looking to electro-
static intcractions for a large part of the binding energy of an ionic crystal. The
distance between a positive ion and the nearest negative ion in crystalline
sodium chloride is 2.81 X 10- 8 cm, and the attmctive coulomb part of the p0-
tential enel"b'Y of the h'lO ions by themselves is 5.1 eV. This value mdY be
compared (Fig. 8) with the experimental value of7.9 eV per molecular UJlit for
the lattice eJlergy of crystalline NaCI with reSpect to separated Na+ and CI-
ions. \Ve now ca' 'ate the energy more dosely.
\
G8
(
Table 6 Electron affinities of negative iOlls
The e lectron affinity is posi tive for a stable negative ion. Source: H . Hotop and \ V. C.
Lineberger, J. Phys. Chern. Rd. Data 4, 539 (1975).

Atom Electron affinity en""rgy '«>m Electron affinity energy

H 0.7542 eV s; L3geV
U 0.62 P 0.74
C 1.27 S 2.08
0 1.46 CI 3.61
F 3.40 a, 3.36
N. 0.55 I 3.06
AI 0.46 K 0.50

Eleclrostalic or Madelung EnerWJ

The !ong-rdnge intcrJction beN'cen ions with charge -z.q is the electro-
static interaction ±t/lr, attmctivc between iOlls of opposite charge and repul-
sive bct\veen ions ohhe same charge. TIle ions arrange themselves in whatever
cryst.d structure gives the strongest attract!v,. interaction eompalible with the
repulsive inlera(.1ion at short distmu:es hetween ion cores. TIle repulsive inter-
actions I>ctween ions with inert ga.~ configurations arc similar to those hetv.·cen
inert gas atoms. nle van der Waals part of the attractive interaction in ionic
crystals makes a rciath'cly small contribution to the cohesive ellerh'Y in ionic
crystals. of the order of 1 or 2 percent . The main contribution to the IJinding
energy of ionic crystals is electrostatic and is called the Madelung energy.
If VI} is the intera<..1ion energy betwcen ions i and j. we define a sum V,
which includes all interactions ill\'olving the Klfl i :

U.=L'UI} , (17)
J

where the summation includes all ions except} "" i. \\'e suppose that VI) ma}, be
written as the slim of a central field repulsive potential of the form A exp(-rip),
where A and p are empirical parameter.~, and a coulomb potential -z.rllr. nlUs

(CCS) (18)

where the + sign is {'<lken for tllO like charges and the - sign for unlike dl3rges.
In 51 units the coulomb illteroctioll is ±r/"I47r€oT; ue write this section in CC5
IInits in which the coulomb intt!raction is -z.q 2 /r.
 3 C'1l"al 8indir>« 69

nle repulsive lenn describes the fact that each ion resists overlap with thc
electron distributions of neighboring ions. We trcat thc strength A and range p
as (.'onst-tnts to bt: dc tennined from observed values of the lattice constant and
compressibility; we have used the exponential form of tbe empirical repulsive
potential rather than the R- 12 form used for the inert gases. The change is
made because it may give a better representation of the repulsive inICr'<l<.1ion.
For the ions, we do not have gas-phase data available to permit the indepen-
dent determination of A and p. We note that p is a measure of the range of the
repulsive interaction: when,. == p, the repulsive interoc1ion is reduced to e - lor
the value at r = O.
In the NaCl strudurc the vallie of U, does not depend on whether the
refere nce ion i is a positive or a negative ion. l1lC sum in (17) can be arranged to
(.'Onverge rapidly, so that its ,'alue 'will not depend on the site of lhe re ference
ion in the crystal, as long as it is not ncar the surface. We neglect Sll rface e ffects
and write the total lattice energy U..,. of a crystal composed of N molecules or
2N ions as UI ... ... NU,. He re N, rather than 2/1.1, occurs because we mils! coun t
each pair" of interactions only once or each bond only once. n le total lattice
energy is defined as the energy required to separate the crystal into indiVidual
iOnS at an infinite distance apart.
It i~ collvenient aglin to introduce quantities VII slleh that r.., - vvR, where
R i... the nearest-neighbor separation in the crystal. Ir we include the repulsive
interaction onl)' among nearest neighbors, we have

(ces) A ex",-RJp) - -'f (nearest neighbors) (19)

R

(otherwise).

(ces) Ulo t=' " (

""U = N zAe- wp - -R
-
'
al) (20)

where z is the number of nearest neighbors of any ion and

""
a - '" -
(~)
. Madclung constant (21)
J PIj

The sum should include the nearest-Ilcighbor contribution, which is just z. TI10
(±) sign is discllssed jllst before (25). The value of the Madelung constant is of
cenh"",1 importaf)(;c in the theory of an ionic crystal. Methods for it.. calculation
are discus.~ed below.
70

At the equilibrium sepamtion dU,ot1dR = 0, so that

dUj N:.A Nul
(CCS) N- - - -cxp( - Rlp) +-- = 0 (22)
dR P Ri
0'

(CCS) RJ cxp(-RJp):::: paq2tzA. . (23)

n.is dete rmines the equilibrium separation no

if the parameters p. A of the
repulsive interaction arc known. For 51, replace q'fl. by lI47r~.
TIle lotallatticc energy of tile crystal 0(21\' ions at their C(l"ilibrium sepa-
ration Ro may be written, using (20) and (23), as

(CCS) U
.01
= - Nacl
flo
(1 _...e...)
~)
(24)

'l11e term - Nar/IRo is the ~·ladelungcllcrgy . We shall find that p is ofthe order
of O. IRo, so that the repulsive interaction has a vcry short range.

EGoluotioJl of ti,e Madelung COfI.!lanl

'l1lt~ first calculation of the coulomb energy constant a was made by
Madclung. A powerful gencm! method for iaUic:c slim calculations W 'd S devel-
oped by E\vald and is developed in Appendix B. Computers arc 110W used for
the culclllations.
111e definition of the Madclllng constant a is, by (21).

a = L:' (±) .
j P,
For (20) to give a stable crystal it is neccs...ary that a be positive. If \\IC take the
reference ion as a negative charge the pIllS sign will appl)' to posith·c ions and
the minus sign to ncgative ions.
An cquivalcnt definition is

~= L:' (±) (25)

R J rJ

whcre TJ is the distance of thc Jlh ion from thc reference ion and R is thc
nearest-neighbor distance. It must be emphasized that thc valllC given ror a
will dcpend on whethcr it is defined in tcrms of the ncarest-neighbor distance
R or in tenus of the latticc parameter a or in terms of somc other rclevant
length.
As an example, we (.'o mputc the ~Iadelung constant for the infinite line of
ions of alternating sign in F ig. 9. Pick a negative iOIl as rcfcrence iOll, and let R
l
 ..
"
,>

Figure 9 Line of ;'.lnS of alternating signs, with dlstau(.'e R between ion5.

denote the distance Letwccn adjacent ions. TIlen

: = 2[~ - 2~ + ~ - 4~ + ... J•
0'

a=2[1 - .!..+.!.
2 3
- .!..+
4
... J. •

the factor 2 occurs because there aTC two ions, ont: to the right and one to the
left, at e<Jual distances rJ. We slim this series by the expansion

..c- i1 1'4
In(1 +x)=x--+ - - - + ...
2 3 4

Thus the Madelung constant for the one-dimensional chain is a = 2 In 2-

In three dimensions the series presents greater difficulty. It is not possible
to write down the successive tcnns by a casual inspcdion . Morc important, the
series will not oolwerge unless the successive terms ill the series arc arranged
so that the contributions from the poJ.itivc and nCJ;llive terms Iw.arly <.'3Il('"e1.
Typical values of the Madelung constant are listoo below, ooscd on unit
charges and referred to the nearest-neig1lllOr llistance:

Structure ({
Sodium ~hloride. NaCI 1.747565
Ce1illm cMorlde. {:SCI 1.162675
\,,;Zi;:;'n~<~b;;l;;
oo;;;,d;;c,",.;,";;;I,lc
:-.;;..;;7.n
;;S; :...__ 1. 638 l

l1)c Madclung and repuL~ive contributions to the binding of a KCI cr-ystal

are shown in Fig. 10. Properties of alkali halide crystals having the sodium
chloride structure are given in Table 7. 11te cak:ulated values of the lattice
energy are in exceedingly good agreement with the observed values.

l
7t

"
"
" l\t·pul<i.,., rn~<gY
(24" 10') c.p(- RiO.3O).,V

,•
•
, ,
>
•

j -, , • , 6 RinIO"<'In

-.
--,.
-"
-"
-"
Figure 10 E"c~ pcr molecule ~ KCI crystal, showing Madelung lind repu lsh .. COlll ribuUOIIS.

COVALENT CIO'STALS

TIle covaltmt bond is the classical electron pair or homopolar bond of

chemistry, particularl)' of organic chemistry. It is a strong bond: the bond Oc-
tween two carbon atoms in diamond with respect to separtlled neutral atoms is
comparaLle \\ith the bond strength in io nic crystals.
TIle covalent bond has strong directional properties (Fig. 11). Thus car-
bon. silicon. and germanium have the diamond structure, with atoms joined to
four nearest neighbors at tetrahedral angles, enm tho ugh this arrangement
glYCS a low filling ofspa<.'t!. 0.34 orlhe available space, compared with 0.74 fora
close-packed structure. The tetrahedral bond allows on1), four nearest neigh-
bors, whereas a close-packed structure has J 2. We should not overemphasize
the sim ilarity of the oonding of carbon and silicon. Carbon gives biology, but
silicon gives geology and stlmioonductor technology.
The co\'alent oond is IIsually formtld from two electrons, one from each
atom purtidpating in the bond. The electrons fanning the bond tend to be
partl y localized in the region between the two atoms joined by the bond. TIlC
spins of the two electrons in the bond arc antiparaJlel.
 Table 7 Properties of alkali halide crystnls with the NaCI structure
All values (except those in brackcts) at room temperature and atmospheric pressure, \\ith no correction for changes in and U from no
absolute zero. Values in brackets at absolute zero temperature and 1.ero pressure, from private communication by L. Brewer.

Nearest- Repuhive Repuhive

Bulk modulus B, Lattice energy compared
neighbor energy range !(I free Ions, in kealJmol
separation in 10\1 dyn/cm 2 parameter parameter
Ro. in A or 1010 Nlm2 zA, in 1O -~ erg p, in A Experimental Calculated

UF 2.014 6.71 0.296 0.291 242.3[246.8J 242.2

LiCI 2.570 2.98 0.490 0.330 198.9[201 .8J 192.9
LiSr 2.75 1 2.38 0.591 0.340 189.8 181.0
Lil 3.000 (1.71 ) 0.599 0.366 177.7 166.1
NaF 2.317 4.65 0.641 0.290 214.4[217.9) 215.2
NaCI 2.820 2.40 1.05 0.321 182.6[185.3) 178.6
NaBr 2.989 1.99 1.33 0.328 173.6[174.3) 169.2
Nal 3.237 1.51 1.58 0.345 163.2[162.3] 156.6
KF 2.674 3.05 1.31 0.298 189.8[194.5] 189.1
KCI 3. 147 1.74 2.05 0.326 165.8{I69.5) 161.6
KB, 3.298 1.48 2.30 0.336 158.5[159.3) 154.5
KI 3."" LI7 2.85 0.348 149.9[151.1] 144.5
RbF 2.815 2.62 1.78 0.301 181.4 180.4
RbCI 3.291 1.56 3.19 0.323 159.3 155.4
RbBr 3.445 1.30 3.ro 0.336 152.6 148.3 "
RbI 3.671 1.06 3.99 0.348 144.9 139.6
i
Data from various tables by M. P. Tosl, Solid nate physil;s 16. 1 (1964).
~
Ii

~
Figure II Cakulatro ,...Jcnceelectron roncent.-ation in gl. rrnanium . 1l>c numbcn Ul thc CQn tOUI'll
give the electron concentnttion per primitive cell, with b.tr ,-alene<" cle(.trons per atom (..;ght
e lectrons pcr primit ive ceO). l\'ote the high concentnltion midway along the c...--ce
bond, as we
expect lOr oow1ent bonding. (Afie.' J. R. Clte hkowsl<), and M. L. Cohen.)

The binding of molecular hydrogcn is a simple examplc of a covalent rond.

The strongcst binding (Fig. 12) occurs when .tlle spins of the two electrons are
antiparallcl. The binding depends on the relative spin orientation not because
there arc strong magnetiC dipolc forces betwccn thc spins, but because the
Pauli principle modifies the distribution of charge according to the spin oricnta~
tion. lbis spin-dcpendent coulomb encrgy is called the exchange interaction.
The Pauli principlc givcs a strong repulsivc intcraction between atoms
with Hned shclls. If the shells are not Illled, electron overlap can be accommo-
dated without excitation of electrons to high encq,'Y states and the bond will be
shortcr. Compare the bond length (2 A) of CI2 with thc interatomic distance
(3.76 J..) of Ar in solid Ar; also compare the cohesivc cnergies givcn in Tablc L
The diffcrencc between CI2 and Ar2 is that thc CI atom has five electrons in thc
3p shell and the Ar atom has six, filling thc shcll, so that the repulsivc interac-
tion is strongcr in Ar than in Cl.
The clemcnts C, Si, and Ce lack four elcctrons with respect to Hlled shells,
and thus these elements (for example) can have an attractivc intef'dction associ~
ated with charge overlap. 111C electron configuration of carbon is 1~2p2. To
form a tctrahedral systcm of covalent bonds the carbon atom must first be
promoted to the electronic configuration ls"2.t2p3. l "his promotion from the
ground state requircs 4 cV. an amount more than regained when the bollds are
formed.
There is a colltinuous rdnge of crystals betwecn the ionic and the covalcnt
limits. It is OftCIl important to estimate the extcnt a given bond is ionic or
covalent. A semiempirical theory of the fractional ionic or covalent charactcr of
\.
 3 Crystal Binmng 7:;

'-'r -\IT-,-i-- -;-:::r:-:---r -;----' :-;;;;;;;;;;;,

:t
o
g
O
.2i----\-\-----+...:".t: -
-~

Figure 12 Energy of mok cular hydrogL'Tl (Hoz) rdern,d to >eparat....! neutral atoms. A negative
energy COJTcSponds to binding. The .... u·ve N refers to a dassical OIkulation ....ith free atom dunge
densities; A is the result for pBI-allci electron spins, takin g the Pauli e~d"sion principle into ac-
count, and S (the stable stat,,) lOr antiparallcl spins. 11,edL'Tlsity of dun-ge is represt.'TltOO by contour
lines fur the stat"" A and S.

a bond in a dielectric crystal has been developed with considemble success by

J. C. Phillips, Table 8, p. 76.

METALS

Metals are characteri7.cd by high electrical conductivity, and a large num-

ber of electrons in a metal arc frce to move about, usually one or two per atom.
The electrons available to move about are calk--d conduction electrons. The
valence electrons of the atom become the conduction c!ectrons of the metal.
In some metals the interaction of the ion cores with the conduction elec-
trons always makes a large contribution to the binding enerb'Y. but the chamc-
tcristic feature of metallic binding is the lo\\'Cring of the enelb'Y of the valence
electrons in the metal as compared with the frcc atom. This lowering is exhib-
ited by several simnle models treated in Chapters 7 and 9.
(
76

Table 8 Fractional Ionic character or bonds in binary crystals

Fndioroal F_
"",..J ionic character "",..J mic cJlItraclcr

S; 0.00
SiC 0.18 C"",, 0.31
Co 0.00 CaSh 0.26
ZoO 0.62 Agel 0.86
ZnS 0.62 AgO, 0.85
ZnSe 0.63 AsJ 0.77
ZnTe 0.61
"sO 0.84
edO 0.7Il MgS 0.79
CdS 0.69 MgSe 0.79
CdSc 0.70
CdTc 0.67 !..iF 0.92
NaCi 0.94
JoP 0.42 RbF 0."
JOM 0.36
InSh 0.32

After J. C. Phill ip$, Vonds and lHmds in femiamduc1f1t"I. Academic Pre.s, 1973, Chap. 2.

The binding energy of an alkaJi metal crystal is considerably less than that
of an alkali halide crystal: the bond fOlTllcd by a conduction electron is not vcry
strong. nlC interatomic distances arc relatively large in the alkali metals be-
cause the kinetic cncrb'Y of the conduction electrons is lower at large inter-
atomic distances. This leads to weak binding. MctaL~ tend to crystalli7.c in rela-
tivc!y close packed structures: hcp, fcc, bee, and some other closely related
structures, and not in loosely-packed structures such as diamond.
In the transition metals there is additional binding from inner electron
shells. Tronsition metals and the metals immediately following them in the
periodic table have large d-clectron shells and are characterized by high bind-
ing energy.

HYDROCEN BONDS

Because neutral hydrogen has only one electron, it should form a covalent
bond with onl}' one other atom. It is known, however, that undcr certain condi-
tions an atom ofhydrogcn is attracted by mtller strong forces to hvo atoms, thus
forming a hydrogen bond between them, with a bond CJlcrgy of the order of
0.1 cV. It is belicved that the hydrogen bond is largely ionic in chamctcr , being
fonned ani)' between the most electronegative atoms, particularly F, 0, and N.
fig,," 13 l1lc hydrogen difluoride ion HY;
it 5tDbiUzOO by a hydrogLTl bond. The sketch
is of an extreme model or the bond, extccme
in the sense that the proton is 5hown bare of
electrons.

In the extreme ionic fonn of the hydrogen bond, the hydrogen atom loses its
electron to another atom in the molecule; the bare proton fonns the h}'drogen
bond. The atoms adjacent to the protoo arc so close that more than two orthem
would get in each other's wa)'. thus the hydrogen bond connects only two atoms
(Fig. 13).
The hydrogen bond is an important part of the interaction between H 20
molecules and is responsible together with the electrostatic attraction of the
electric dipole moments for the striking ph)'sical properties of ,vuter and ice. It
is important in certain ferroelectric crystals (Chapter 13).

ATmnc RADll

Distances between atoms in crystals can be measured very accurately by

x-ray diffraction, often to 1 part in l<f. Can we say that the observed d istance
betwcen atoms may be assigned partly to atom A and partly to atom B? Can a
definite meaning be assigned to the radius of an atom or an ion, irrespective of
the nature and composition of the crystal?
Strictly, the answer is no. The charge distribution around an atom is not
limited by a rigid spherical boundary. Nonetheless, the conccpt of an atomic
r.w.ius is fruitful in predicting interatomic spacing. The existence and probable
lattice constants of phases that have not yet been synthesized can be predicted
from the additive properties of the atomic radii. Further, the electronic conHg-
uration of the constituent atoms often can be inferred by comparison of mea-
sured and predictccl values of tlle lattice constants.
To make predictions of lattice constants it is convcnient to assign (Table 9)
sets of sclf-consistcllt radii to various types of bonds: one set for ionic crystals
with thc constituent iOns 6-coordi nated in inert gas closed-shell configurations,
ano ther sct for the ions in tetruhedrally-<.'OOrdinated structures, and anot her set
for 12-coordinatt.-d metals (close-packed).
The prcdictt.>d sc1f-ronSistent mdii of the C'.ltion Na + and the anion F- as
given in Table 9 would lead to 0.97 A. + 1.36 A. = 2.33 A for the intemtomic
scpamtion in the crystal NaF, as comparcd with the observed 2.32 A. T1lis
agreement is much better than if we assume atomic (neutral) configumtions for
 Table 9 Atomic and ionic radii
H
Values approximate only. Uni ts arc 1 ).= 10- 10 m. For original r{!fcrences
H. ~
2.08
see W. B. Pearson, Crystal chemistry Gnd physics of metals and alloys,
Wiley, 1972.

B. N.
Li
0.68 0.35
1.06
0.23
B
0.88
ICO} 5 IN
1.71
,0.77 0.70
1°1.40
0.66
IF
1.36
0.64
1.58

1.56 1.13 0.98 0.92

N'
0.97 1M,
0.65
lAO
_Stan dard radii for ions in inert gas (fil led shell) conflguration _ _
!,;;;;,""" )ladii of atOIEs wQ!W in tfj:np.,oorai co.Y1\cnt bond~ -
AI
0.50
1.26
5' IP
10.41
1.17
2.12
1.10
5
1.84
1.04
I CI
1.81
0.99
I"
1,88

1.91 1.60 - Radii of ions in 12-coordinate<l metals _ 1.43 1.32

Ti v C, M, F. Co Ni C, G, G. A, 5. B, K,
K
1.33
Ca
0.99
Sc
0.81 0.68
- "
0. 74 0.62 0.53 2.22 1,98 1.95 2.00

- ~1.~5.. 1~i181~1~6~l l~~~1 f"25 1. 35 1.31 1.26 1.22 1.18 1.14 loll,.,.
2.38

Rb
1.48
1.98

5,
1.13
1.&4
y
0.93
1.46

Zr
0.80

1.60
Nb
0.67

t4:/
I Mo

i40 1.36
T, R, Rh
1.25

Pd
1.28

A,
1.26
1.52
1.39

Cd
0.97
lAS
1. 41

"0.81 0.71
J,44.. 1.40
1.37
5,
1.39
50
2.45
1.36
T.
2.21 2.16
1.34.- 1.28
, X.
2,17

2.55 2.15 1.80 1.34 1.35 1.38 1. 45 1.57 1.66 1.55 1.59
C, B, T, w R. 0, I Pt A, H, TI I At R,
" Hf Ir Pb Bi Po
1.67
~

2.73
-
1.35

2.24
1.15

1.88
1-
1.58 1.47 1- 1- 1-
.J..~~ 1.38 1.35 1.36.. 1~
1.37

1.44
1.10
1.48
1.57
0.95

1.72
0.84

1.75 1.70
1- 1 -
1.76
I R,
A,
1.1 1 "I p, 5m 0, Ho
1.75 I 1.37 I Ce
1.01
Nd Pm
" Gd Tb
" Tm Vb
"
.1...
71
1 82=---1
'. 1. 83' -I -1-
1.82~ ~i . 8 1 ( SO
2.04:· J - I~
I.sol· l 1 -:-80 - - 1.78 i.77- 1.77·17:"76 1.75
.J.~~' -'
1 7410

p, Am em I Bk Fm Md I No
IP'
_0~90 IU
Np Cf
" "
::"O;~
Th

1.80 1.63
0.83,
1.56 1.56 ""'"1-"1 -' I ~
\.64 '1.81 ·· - -
h - --
I - 'I ~

J
 -,
'-

Table 10 Use of the standard radii of ions gi\'en in Table 9

The interiOnic distance D is represented by DN - Rc + RA + tJ. N, for ionic crystals,
whE:re N is the coordination number of the cation (positive ion), Rc and RA are thE:
standard radii of tile catiOO and anion, and tJ. N is a correction for coordination number.
Room temperature. (Afier Zachariascn.)

N o..,(A) N o.,,{A) .v 0.".(A)

I -0.50 5 -0.05 9 +0. 11

2 -0.31 6 0 10 +0. 14
3 -0.19 7 +0.04 II +0.17
• - O. ll 8 +0.08 12 +0.19

Na and F. for this woultl lcad to 2.58 A for the interatomic separation in the
crystal. The latter value is i(n.n. distance in metallic Na + interdtomic distance
in gaseous F J.
11lC interatomic distance betwcen C atoms in diamond is 1.54 A; one-half
of this is 0.77 A. In silicon, which IlllS the same crystal structure, one-half the
interatomic distance is 1.17 A. In SiC each atom is surrounded by four atoms of
the opposite kind. If we add the C and Si radii just given, we predict 1.94 Afor
the length of the C-Si bond, in fair agreement with the 1.89 ).. obsclVed for the
bond length. l1tis is the kind of agrt.'Cment (a few percent) that we shall filld in
using tables of atomic rddii. 3

Ionic Cry:Jtal Rodii

In Table 9 we include a set of ionic crystal radii in inert gas configurations.
The ionic radii are to be used in conjunction with Table 10. Let us consider
SaTiO" in Fig. 13.10, with a lattice constant of 4.004 ).. at room temperature.
Each Sa+ + ion has 12 nearest 0 - - iOnS, so that the coordination number is 12
and the correction 6. 12 of Table 10 applies. IT we suppose that the structu re is
dctemlincd by the Ba-O contacts, we have D 12 "", 1.35 + 1.40 + 0.19 =
2.94 A, or a = 4.16 A; if the Ti-O contact dctermincs the structure, we have
D6 z: 0.68 + 1.40 = 2.08 or a "" 4.16 A. The actual lattice COnstant is some-
what smaller than the estimates and may perhaps suggest that the bonding is
not purely ionic, but is partly covalent.

'Fvr referena:s on atomic and ion" F.IIlJi , $Ce Pear$On, Table 9, p. 76. 1.. Pauling, T~ nature
oJlhf cMmiCll1 bond, 3rd ed., Cotnell, 1960; J. C. Slater. J. Chern. Pl-Lys. 41. 3199 (1964); B. J.
Austin and V. Heine, J. C1",m. Ph)'1i. 4S, 928 (1966); R C . PIInoru and v . F. Weisslop(, Zeits. f.
Ph),>,,,, 202, 492 (1967), S. GeDer. Z. Kriswlognphie 125, I (l967); fl. D. Shannon. Acta Cryot.
Ali. 151 (1976); and, foro.-idcs and llU<riles, R D. Shannon anu C. T. P~"'tt. Acta Cryol_ 825.
925(1969). An early siudy if by W. L. B'"AAo -The arr.mgemcnl ofo!<lIns in crystals,· Ph,IO$. Mag.
(S) 40, 269 (1920). I
"
ANALYSIS OF ELASTIC STRAINS

We collsider the c lastic properties of a crystal viewed as a homogencous

continuous medium rather than as 11 periodic array of atoms. The continuum
approximation is usually vaJid for CISlitiC waves of wavelengths A longer than
(

I
10- 6 em , which means for frequencies bel(w,. 1011 or 10 12 Hz. Some of the
material below looks complicated because of the unavoidable multiplicity of
subscripts 011 the s)'mbol~. The basic physical ideas arc simple: we usc Hooke's
law and Newton's second law. Hooke's law states that in an clastic solid the
strain is directly proportional to the stress. The law applics to small strains only.
\Vc say that we arc ill the nonlinenr region when the strains arc so large that
Hooke's law is no longer satisllcd.
"Vc Specify the strain in terms of the components ell" tw' e:u. 0:11/' ejP-' e=
which arc defined below. We treat infinitesimal strains only. We shall not dis-
tinguish in our notation between isothermal (constant temperature) and adia-
batic (constant entropy) deformations. The small diiTerences between the iso- .'
thermal and adiabatic clastic constants are not often of importance at room
temperature and below.
We imagine that three orthogonal vectors X. Y. 2 of unit length are emb<..>d-
ded securely in lhe ullStraincd solid, as shown ill Fig. 14. After a small unifonn
dcfonnation of the solid has taken place the axes arc distorted in orientatiOll and
in length. In a unifonn defonnation each primitive cell of the crystal h de-
fonned ill the same way. 11lC new axes ,,'. y'. z' ma), be writteJl ill terms of the
old axes:
x' -= (l + E;.-...)X + ~-'1Iy + ~~~2 ;
y' = ~If.'x + (1 + ~w)Y + £vz2 ; (26)
z' = V + £=!I)' + (1 + Euri. ,
The coefficients £afJ define the deformation; the)' arc dimensionless and have
values ~ I if the strain is small. The original axes were of unit lenS:,rlh. but the
new axes will not neCt.'Ssarily be of unit length. For example.
x"x' = 1 +2£..... + E!.,+ ~.. + E!..
whence %' iii: I + £= + .... The fractional changes of length of the x. y, and i
a"{es are En. EW' ~~~, rcspt.'Ctively. to the first order.
What is the effect of the deformation (26) Oil an atom originally at r ,., xX +
yy + xi? The origin is taken at some other atom. If the deformation is unifonn.
then after deformation the point will be at the position 4 r' = XlI' + yy' + "0.'.
11m displacement n of the deformation is defined by
R a r' - r =x{x' - x)+ y(y' - y)+z(z' -2). (27)

'This is <Jb.,iolls[y COrf'C.'C"t tr"""e choolic tlIC.i Il.~is suel, tllM r - ri. thcn r' - xx' by dc..../lnition
of.'.
 , "

~-- "
, ;

(~
"
(b)

Figure 14 Coorditlate a<e5 for tire description 0( tile state of strain; the
ortf.:,gonal unit lUeS in the unslrained state (I.) an: t1erormed in tI,."
stra1ned state (b).

or, from (26),

R(r) • (u.u + yEll" + u ....,)x + (XE..y + !lEw + U .q/)Y
+ (U.n + !IE..,. + udi (28)
'n lis rna)' be written ill a more general form by introducing II, v, w such that
the displacement is givCll by

I R{r) "" u( r}i + v(r}5' + w(r}Z . I (29)

Ir the deformation is nonuniform we must relate u, v, w to the local stmins. We

take the origin of r close to the region of interest; the n comparison of (28) and
(29) gives, by Taylor series expansion of R using R(O) 0, 0::::

OU
x'" a x - · ctc. (30)
~ ilx'

It is usual to work with coelHcicnts eop rather than Eol1' " 'e define the
strain components e...... ew> e= by the relations

Ov Ow
e - e = - ez.o . -u
.. =az
- (31)
IN lit' iJy

using (30), TIle other strain components e-'1/' ev<, e..., arc defined in terms of the
changes in angle between the axes: using (26) we rna}' define
.
" OIlOV
e"Y Iii X'Y ;:;:" '" +E,.=-+ -
u.-" oyox
, •
,
.. ~
. y·z a1 E___
-~
+€1/& - -clv
i1z
+ilw
ay- (32)

, . all ale
e=
.. z·x n =-+-
. a t:u+E
azdx

\Vc may replace the ~ signs b)' = sigllS if we neglect terms of order e 2, The six
dimensionless coefficients e..,J.=efJo} completely define the strain.

Dila/ioll
The fractiona l increase of volume associated with a deformation is called
the dilation. The dilation is negative for hydrostatic pressure. The unit cube of
edges i, Y. 1 has a volume after dcfonnation of
V' = x' .)" I< Z' • (33)
by virtue of a well-known result for the volume of a pwaJlelcpipcd havingcdgcs
x', y'. z' . From (26) we have
1+ IE.,.

""y'I<Z'= a l+eu+ew+eu' (34)

Products of two stmin components have been Ilt.'glcdcd. The dilation l> is thell
given by

(35)

St ress Comporlent8
111e force acting on a unit area in the solid is defined as the stress. There
arc nine stress components: Xr • X~. X&. Yro Y~. Y2 • z...
t;.. Z&. The capitallct-
ter indicates the direction of the force. and the subscript indicates the nonnal to
the plane to which the force is applied. In Fig. 15 the stress component Xr
represents a force applied in thc " direction to a unit area of a planc whose
nonnallics in thc " direction; thc stress component X~ represents a force ap-
plied in thc x direction to a unit area of a planc whose nonnallies in thc y
direction. Thc number of indcpendcnt strcss rompooents is reduced from nine
to six by applying to an cJemcnt,lry cube (as in Fig. 16) the condition that the
angular acceleration vanish;5 and h~ncc that the total tonluc must be 7..ero. It

s-n"s Out.'$ nol mean we umnot 1........1 .,ro/.lc",.!n ... h,ch Ihere Is ,gular a<.'Ct'1er:alion; it jusl
mt.'1lnS th.tl we ron use Ih" st.ltic SIIUM ..", to de"ne the cbstoc con
"
 3 ElaotM: C_ot.."u

,,
,, I'
'- ~
"~ ,
"
, ' ,"
,
, I
I
I
"
I ~~ "I
--r
x " .,',, ---
::J;::"«::7'"- ,
-- figu~ 15 Stress comporlC.'nt X. IS a bu: II{lplieC I"

11--
.,
'-,'
'. X,
,, ,,
, ,
" '"
", the 11 direction 10 a unit area of a rlane ",llOSe nnrm ..1
liell in the:l din."Clion; X~ is a{lf>hed in the:t din." ,c>n
to I unit area of a plane .... hose 1I0rmal lk'S ill II,,) "

" di ...."CIion.

,
x,--

"----+1T'- <
1
Y,
Figure 16 o.,.monstration that for a body in static a Juilib-
rium Y . .. X... The sum 0( the fo~. in the :t direction is
7-l'ro. The slim 01' the 1'on'Cll in the", direction is a1iO uro.
1111tl total hee "lmishes. The total touIIIC aboot the origin is
a1iO zero if Y, - X~.

rollo\\"s that
y~ = ZII ; z.,=X~ ; X!I= ¥". (36)

l1lC six independent stress components may be taken as X". YII' z." ¥~, Z~, XI/""
Stress componenl~ have the dimensions or rorce per unit area or energy
per unit volume. The strain components are ratios lengths and are dimen- or
sionless.

ELASTIC COM PLIANCE AND STIFFNESS CONSTANTS

Hooke's law states that ror sufficiently small deronnations the strnin is
directly proportionaJ to the stress, so that the str.lin components are linear
rUJlctions or the st.. ~~s t.:o mponenls:
, ,
.. )
e~ = SllX~ + 5 12Y, + Sl~a + SI4Y~ + SI:;Z~ + Sl.y:-,
eW = SZIX. + S22 Y.. + S~z + SUY" + S:z..v.... + Szt;X, ;
eu = S31 X• + S3ZY 1/ + S33 Z• + 534Y. + SasZ..- + S~ .. ;
(37)
ell" = S41X. + 542ft' + 5..:)2. + S.S. + S4~ + S4~1I ;
eu = SSlX. + 552Y.. + S~z + S.5<\Y. + Ss.-;Z. + SsoXll ;
eq/ = 56 .X. + S62YII + 5m2. + 5{>jY. + SosZ. + SooX,I
I
x" = Cllen + C1Zew + C1#u + C l4e ll1 + else.... + CU;e,,'I ;
y 1/ = Clne"" + C:ue w + Cd~ + C",ell" + C~ + ewe... ;
z. = e3l en: + C3~w + C3#u + C:J,4eF + C~u: + C~''1 ; (38)
y. = Cue,." + Cuevv + C43e,.. + C44ep + C~u; + C4#.t}I ;
Zr = CSI£u. + CSz€w + C~ + C,54e p + C:sseu. + C~ ... ;
XII = C61 e"" + C!>'#w + Croe~ + C64eV<' + C6.¢u + C6f!!."JI .
The quantities Sil. S 12 • • • arc called clastic compliance constants or clas-
tic constants; the quantities Gil. C I2 • . . arc called the elastic stiffness con-
stants or moduli of elasticity. 1bc 5's have the dimensions of (arcaVlforccl or
(volumc1tlcncrgy]. The C's have the dimensions oflforcc]l[arca} Or [cncrgy¥
[volume] .

EMstie E,lIJrgy Density

The 36 constants in (37) or in (38) may be reduced in number by several
considerations. The clastic energy density U is a quadratic function of the
strains, in the approximation of Hooke's law (recall the expression for the cn-
CJ"b'Y of a stretched spring). Thus we may write

I • •
U ="2 L L CA,.eAe,.
A- I ,._ 1
, (39)

v.ilere the indices 1 through 6 arc defined as:

6 - xy. (40)
The C's arc related to the C's of (38), as we sec in (42) below.
The stress components arc found from the deri\"ath'e of U with respect to
the associated strain component. This result foll(ly,."s from the definition of p0-
tential energy. Consider the strcss X. applied to one face of a unit cube, the
opposite face being held at rest:
 3 """'ie eo...,,,,,,.
)
Note that only the combination 1(C"13 + CrsJ enters the stress-strain relations.
It follows that the elastic stiffness constants are symmetrical:
C"13 :: i{C"1l + Cp..) :: CjJl> . (42)
Thus the thirty-six clastic stiffness constants are reduced to tv.'enty-one,

Elastic Stiffness Constanl$ of Cubic Crystals

The number of independent clastic stiffness constants is reduced further if
the crystal possesses symmetry clements, We now show that in cubic crystals
there are only three independent stiffness constants.
We assert that the elastic energy density of a cubic crystal is
U = iC,,{~ + e!v + e~ + iC.u{c.... + ~ + ~v)
+ C 12{e.".e= + e=ej(X + e"..ew ) ' (43)
and that no other quadratic terms occur; that is
(en-e.... + ...) (44)
do not occur.
The minimum symmetry requirement for a cubic structure is the existence
offour three-fold rotation a... es. The axes arc in the (Ul] and equivalent direc-
tions (Fig. 17). The clrcct of a rotation of 211"13 about these four a... es is to
interchange the x, y. z a...es according to the schemes
-x_z_-y_-x;
- x-y-z--x, (45)
according to the axis chosen. Under the first of these schemes, for example,
e;.+e:v+t{,-e!v+e~+~" ,

,,3
i ,
,,
'b

,
,,,
,

Figure 17 Rotation by 2m'3 "bo'" the axis

mark.,.! 3 changes r ..... y. y ...... ~ and z_r.
86

and similarly for the other terms in parentheses in (43). Thus (·13) is invariant
under the operations considercO. But each of the tcnns exhibited in (44) is odd
in one or more indices. A rotation in the sct (45) can be found which will change
the sign of the tCnTI, because e~JI = -eo:(-..). for example. Thus the lenTIli' (44)
arc not invariant under the required operations.
It remains to verify that the numerical factors in (43) arc (. orrect. By (41)
aU/iJen = Xx = e ll en + C I2(e JlJl + e~J . (46)
The appearance ofC1 1e,;x agrees \~;th (38). On further comparison, we sec that
(47)

Further, from (43),

aUlae~1I ,.. XII = C44e~1I ; (48)
on comparisOn ",;th (38) we have

C 61 :;; C62 = C63 = C64 = eM "" 0 (49)

Thus from (43) we find that the array of values of the clastic stiffness con-
stants is reduced for a cubic crystal to the matrix

X. C" C" C" 0 0 0

Y, C" C" C" 0 0 0
Z. C" C" C" 0 0 0
(SO)
Y. 0 0 0 C.. 0 0
Z. 0 0 0 0 C.. 0
X, 0 0 0 0 0 C..

For cubic crystals the stiffness and compliance constants are related by
C 44 = 11544 ; CII - C I2 = (5 11 - v-
51 1
;

CII + 2C I2 = (5 11 + 25,J- 1 (51)

These relations follow on cva1uating the iuverse matru to (SO).

Bulk Modulus and Compressibility

Consider the unifonn dilation en .... ew:O: eLl: = ia. For this defonnation
the energy density (43) of a cubie crystal is

U = i(C
We may define the bulL: modulus B by the relation
I + 2C I"~ . (52)

U ~ IB8' • (53)
 3 £In.ti.: Co... t4 ..,. 81

Figure 18 Cu"," of volume lu, 4y fJ.:t acted

on by a stress - X.(%) O il the face at %, and
Volu me /J.x t:.y /J.=.
X.{% + fJ.x) ill X.(%) + i1X. Ax on the parallel

face at % + fJ.%. "

TIle IIet force u
(i1~~ fJ.x) fJ.y fJ..:. Other forres in the J di~
lion arise from the ..... rlation across the clIl.>e<f
the stresscs X~ and X.. which lire not show".
'''e net % compoII""t of the forre on the
cub" is

"-. _ ax. =i1X +--

(~+ i1X.) 4x ...
• yflz..
i1x i1y i1:t
' 11e force equals the mll$S of the cube tin'e<
the rompollcnt of the accekrntion in the x di-
rection. ' 11e mass is pfJ.x l!.y t1zand the act",l-
eration is ;;ufilt~.

which is C<luivaient to the definition -V dpldV. For a cubic crystal

B = l(C l l + QCd . (54)
The compressibility K is defined as K '" lIB. Values of Band K arc given in
Table 3_

ELASTIC WAVES IN CUBIC CRYSTAlS

By (.'Onsidering as in Figs. 18 and 19 the forces acting on an clement of

volume in the crystal we obtain the C<luation of motion in the x direction
ilu ax~ aXil aXiL
p-,- ~--+ +- - (55)
at ax ay az
here p is the density and u is the displacement in the x direction. Thcre are
similar equations for the y and z directions. From (38) and (SO) it follows that for
a cubic crystal
2
p-aatu2 -c
-
aen c (ae""
--+
11 ax 12
ae=) + c (ae + -
~+ --
ax ax «ay
ae=)
az- '.
i !!
(56)

here the x. y. z directions are parallel to the cube edges. Using the definitions
(31) and (32) of the strain com[xments we have
2 2
a2 u iJ
2
u (ihl iJ'lu) ( (l v aw )
p at 2 :: Gil a:c + C 44 a!l + ae + (C 12 + C 4 J ax ay + ax az
(57a)
where 11, v. w arc ~ J components of the displacement R as defined by (29).
"

Figu", 19 lhpri"WiA and B are stretched equally. the block bel"-'Ccn them expcri-
ern.'eS no net force. ' 1,is iUu51raies the fact that" unil"onn stress X. in" solid d"",s ,lOt
gj,'e II. net force on" \'o\ume dement. If the spring a' B ;s stretched more than the
$pring lit A, the block between the", will be accelerated by the force X.(B) - X.(A).

nlC corresponding equations of motion for iPvlar and ;Pwlaf arc found
directly from (57a) by symmetry:
a2v
p iJf = Gil
a2 v
ar! + CH
(ilv 3 v)
2

ar + aZZ + (C 12 + GH )
(axiJ iJy + iJya w)
2
u
2

ilz ;
(57b)
2 2 2 2
aw
p ---:2 = C II
iJ w
----r- (iJ W
(l2W) ( au av )
+ C~4 ~_.2 + ~ + (C 12 + G«) - - + - -
2

or iJz OA iJy ax oZ ay iJz

(57<)
We now look for simple special solutions of these equations.

Waws in the (IOO] Direction

One solution of (57a) is given by a longitudinal wave
u= Uo cxp [i(Kx - wi)] , (58)
where u is the x component of the particle displacement. Both the v."a\'Cvcctor
and the particle motion arc along the x eube edge. Here K = 2m}" is the
wavcvector and w = 27Tl/ is the angular frequency. If we substitute (58) into
(57a) we find

w2 p = cllJ(l ; (59)
thus the velocity wlK of a longitudinal wave in the {tOO] direction is
v~ = II}" = cdK = (C lI /p)lf2 . (60)
Consider a transverse or shear wave with the wavevector along the x cube
edge and with the partide displacement v in the y direction:
v = flo exp [i(Kx - Wf)] . (61)
On substitution in (57b) this gives the dispersion relation
w2p = cHJ(l j (62)
thus the velocity fdK of a transverse wavc in thc [tOO] direction is
v~ = (C";p)112 . (63)
The identical velocity is obtained if the particle displacement is in the z direc-
 3 £leal.., Cono'an.U 69

(
tion. l1IUs for K palll!lcl to (100] the two independent shear wavcs have equal
velocities. This is not tnle for K in a general dircction in the crystal.

Wavu in tile [1101 Directioo

There is a special interest in \\'Uves that propagate in a face dia1,'O naJ direc-
tion of a cubic crystal. because the three clastic constants ean be found simply
from the three propagation velocities in this direction .
Consider a shear wave that propagatcs in the xy plane with particle dis-
placement w in the z; direction
W "<' tva exp 1f(K.x + K,;J - wt)] • (64)
whence (32c) gives
(65)
independent of propagation direction in the plane.
Consider other \\'BVCS that propagate in the xy plane with particle motion
in the xy plane: let
" "" 110 cxp [i(K.,.r + ~y - wt)] v = Va exp [i(K.x + ~y - wt)]
(66)
From (57a) and (57b),
2
w pu = (CIIK! + C.,.,K!)u + (C 12 + CoH)K"K,p ;
(67)
t.Jlpv = (CIIK! + coHK!)v + (C 12 + C••)K,,~II .
111is pair of equations has a partiCtilarly simple solution for a wave in the lItO]
direction, for which K" = Kv = KlVz. The condition for a solution is that the
dctenninant of thc coeffieicnts of II and v in (67) should equaJ zero:
-w2p + !(C II + c.J~
l Me 12 + C..Jk.02
(66)

'nIis equation has the roots

w2 p = i(C II + C l2 + 2C4.)~ ; (69)
The first root describes a longitudinal \\'Uve; the second root describes a
shear \v.wc. How do we dctennine the direction of particlc displacemCllt? The
first root when substituted into the upper equation of (67) gives
!(C II + C I2 + 2C44)~U "" I(C II + C44)k~u + !{C I 2 + CH)~V , (70)
whence the displacement components satisfy u = v. Thus the particle displace-
ment is along (ItO] and parallcl to the K vcctor {Fig. 20). The second root of(44)
when substituted into the upper equation of (67) gives
!{C u - CI~K'2t1 .,. 4(C Il + C 44 )K'2U + i{C 12 + C.JK!v, (71)

...
 I IL
K

/L
K
lr- T

\Va'''' in [11 1].Iirectioll

/ \Va'''''n ( 1001 dir«1ion W,,'" in [IIOJ .lircdion
L : Cll L : ~(CII + e12' 2C u ) I . : A(GI] • OC n + 4Cf,I)

T : C.. " 1' C44 r . J <CU - Clil. tC..,}

T2 : i(Cu-CIll

Figure 20 Efi'ecthoe elastic .:o" .tanls b Ihe th ree modes of ela$l~ waves in the principal
propag.atiof. diredion" in rub~ crystals. , ,,., two transvene modes are degenerate ror prop-
ag;ooon inlhe [1(0) and 1111 1directions.

whence " = - v. 1110 particle displacement i~ along (1101and perpendicular to

the K \"cctor.
Selected values of the adiabatic clastk stiffness constants of cubic crystals
at low temperatures and at room temperature arc given in Table 11. Notice the
general tendency for the clastic constants to decrease as the temperature is
increased. Further values at room temperature alon e Brc given in Table 12.
There arc three nonnal modes of wave motion in a crystal for a given
magnitude and diro::tion of the v.'Bvc"cctor K. In general, the polarizations
(directions of particle displacement) of thcse modcs are not exactly parallel or
perpendicular to K. In the special propagation directions 11001. [Ill}, and [110}
of a cubic cry5tal two of the three modes for a given K are such that the particle
motion is exactly tronsverse to K and in the third mode the motion is exactly
longitudinal (parallel to K). lbe ana1ysis is much sim pler in these special direc-
tions than in general directions.

SUMMARY

• Crystals of inert gas atoms arc bound by the van der Waals interaction (in-
duced dipole-dipole interaction). and this varies with distance as ll RI!.
• lbe repulsive interaction between atoms arises generally from the electro-
static repulsion of overlapping charge distributions and the Pauli principle.
which compels o\·erlapping electrons of l>arallel spin to enter orbitals of
higher energy.
• IOllie crystals arc bound by the electrostatic att~ ~ of charged ion5 of
 3 E1d.1it: Coo.",..", .
Table II Adiahatic: elw; lic ~Ii(fncn wnstants of cubic crystals
at low temperature "lid at room temperature
'The "alue. give n al 0 K Wt.TC oblaim:d by CJllnpoialion of meaSllfcments carried
out down 10 4 K. The table was com pilt.>d with the assistance of ProfclSOl' Charles S.
Smith.

Stif!iwu cons~b,
In lO, a dynclcm l (10" N/m 2)
C I),st;ol e" Coo C" Temperatun', K Ot.'rl$ity, gfem'"

W 5,326 ..,,, 1.631 0 19,317

5.233 2.015 1.607 JOO
Th ~663 1.582 0.874 0 16.696
2.609 1.574 0.818 300
ell 1. 762 1. 249 0.818 0 9.018
1. 684 1.2 14 0.754 300
Ag 1. 3 15 0.973 0.511 0 10.635
1.240 0.937 0.46 1 300
A" 2.016 1.697 0."'" 0 19.488
1.923 l. 631 0.420 300
AI 1.143 0.619 0.316 0 2.733
1.068 0.607 0.282 300
K 0.04 16 0.0341 0.02116 4
0.0370 0.03 14 0.0188 295
Ph 0.555 0.454 0, 194 0 11.599
0.495 0.423 0. 149 300
Ni 2.6 12 1.500 1.311 0 8.968
2.500 1.500 1. 235 300
Pd 2.34 1 1.761 0.712 0 12. 132
2.271 1.761 0.7 17 300

Opposite sign. The electrostatic energy of a structure of 2N ions of charge

±q is
q2 (±}q2
(ces) u~ -N.- = -N2.--
R fiJ •

where a is the Madclung constant and R is the distance between nearest

neighbors.
• Metals are bound by the reduction in the kinetic energy of the valence elec-
trons in the metal as compared with the free atom.
\
" (
Table 12 Adiabatic elastic 5tiiTneSl oonstanb
of several cubic crystals al room temperatun:
or 300 K.

Stiffness UlIl$tllll15,
in lO" dyne/ems or 10" N/ms
C" C" c.
Diamond 10.76 1.25 5.76
N. 0.073 0.062 0.042
U 0. 135 O.1l4 0 .....
Ce 1.285 0.463 0.680
S; 1.66 0.639 0.796
CoSb 0.885 0."" 0.433
JnSb 0.672 0.367 0.302
"gO 2.86 0.87 1.48
NaCI 0.487 0.124 0. 126

• A covalent bond is characterized by the overlap of charge distributions of

anti parallel e lectron spin. The Pauli contribution to the repulsion is reduced
for antiparallcl spins, and this makes possihle a greater degree of overlap. The
overlapping electrons bind their associated ion cores by electrostatic attrac-
tion.

Problems
I. Quantum .did. In a quantum Did the dominant repulsive energy 15 the zcro-point
energy of th e atoms. Consider a crude onc-dimL>Jl~ional model of crystalline He"
with each He atom oonflned to a Iinc segment of lengt h L. In the ground "Iatc the
wave function within each segment is taken as a h,llf ....·a"elcngth d a free partkk.
Find the zero-point kinetic ellergy per particle.

2. Cohesive energy of be<: ondfa: neon. Using the LI..'rInard-JQlle.s potential, calculate
the ratio d the cohesive energies d noon in the bee and fcc druelurCii (Am•. 0.958).
The lattice .urns for the bee ~tructun:s are:

L' ,,; 12= 9.11418 ;

J
3. Solid molecular hydrogen. For lis one finds from measurements on ti le ga:s tllat the
Lennard-lone.s parameten are t' - 50 X lO- 1f erg ami u - 2.96 A. Find the com.--
siVe energy in kJ per mole of H!; do the calculation for All fcc:: structure. Treat each
H2 mok'ellie as a sphere. The obserlll..>d value of the collcsi"e energy is 0 .751 kJ/
mol, much leu than we calclliatl..-d, so thllt (Juautum corn..'(:tions must be V(:I')' im·
portant .
 93

4. pUll/ibilitli of ionic crysWIf R+R -. Imaginc a crystal that cxploits for binding thc
coulombattractioll ofthc posi th'C and m..'g3livc ions orthe 5amc atom or mol(.'(!u le It
This is beli(.",,,,(} to OCCur with C(.>rtain organic mok'(!uleli, but it is not found ....i le n H
is a ~ingle atom. Usc the data in Tables 5 and 6 to ('valuate the stability of SUdl a
form uC Na in the NaCI structure relative to normal metallic toclium. Eval uate tlle
CnCI1o'Yat the obscT\lCd interatomic distance in melallic sodium, and use 0.78 eVas
the electron affin ity of Nil.

S. linear ionic CTJIIItal. Consider a line of 2N ions of alternating charge ±q with a

repulsive potential energy AlR n betwt.'Cn nearesl ncighboTli. (a) Show thai at the
t.'<lumbrium separation

(ces) U(~= 2N~ln2(1_~)

(b) Lt.'I the O)"5lal be compressed so that ~ _ Ro(l - 6). Show that the work done
in compressing a unit lellb<th of the crystal has the leading 1t.'TJ1l .. cd. wht.'TC

(ces) c ~ "(""--- .
--';1"'""-,1,,"-,,,
.

To obtain the ~IIItS i.n 51, replace q2 by q2/4Trto. Note: We should nol expt.'CI 10
oblai n thi.s result from the expression for U(Ro), but we must !lie the complete
expression for U(R).

6. Cubic znS structure. Using), and p from Table 7 and the ~Iadeillng COIlstants gi~'t!1l
in the text, calculate the cohesive energy of KCI in the cubic ZnS structure de-
scribed ill Chapter I. Compare with the value C"'uculat(.'(} fo r KCI in tile NaCI stnlc-
ture.

7. Divdenl ionic crystals. Barium oxide has the NaCI stru ctu re. Estimnte th e cohe-
~i ve energi(.'S per molecule of the hypothetical cryltals Bu+O- llnd 8a.+ +0-- re-
ferred 10 separaled neulral aloms. '111e observed neaTest-neighbor intem uclear dis-
tance i~ Hu - 2.76 A; the firsl and st.'"COnd ionization pote nliat~ of Ba arc 5. 19 and
9.96 eV; and Ihe eit.'(!lron affinities of the first and s(.'OOnd cl(.'(!trons added to the
neu tral Oxygt.' l atom arc 1.5 and - 9.0 eV. 'Joe first ek'Ctron affi nilY d" the n eu tral
oxygen atom is the energy released in the reaction 0 + e _ 0 -. '111e it.'(.'()nd elec-
tron affinity i.s the (''T1t.'"rgy rcl eased in the reaction 0 - + e -+ 0--. Which vale nce
state do you pn..-dict will occur? Assume ~ b the same for b«h forms, and n(.'gk'CI
the repu l!;, '!! t.'T1t."rgY.

8. Young', modu/u. and Poiuon', ratio. A cubic crystal is subject to 1t.,lsion in the
[100] di"-dion. Find expressions in t(.'Tt1lS of the clastic sliffncsSt.'S for ¥O\lng'l mod-
ulus and PoiSiOfl'S ralio as defin(.-d in Fig. 21.

9. /.(mgitudinal tOOve vdocity. Show that the velocity of a longiludillal wave in the
1111 1 direction of a cubic crystal is givcn by v. - (j(C 11 + 2C I2 + 4C••)fpJIII. l linl:
For such a wave 1I :.. V = w. Lctll = Ud111((>+~+ .o,V\l.ic- "", and usc Eq. (57a).
 ~+'''-1, \)ndetornwd body 7
,-- ---- -..l. -- - 1
,,
-- I
I I
I
=i
r- I

L.
I I I
TensIOn I I w_6w w I

- I
I
I
I
I
I
=--ll L _ _ ________ J
I

L
---- ~

Fogure 21 Young's moJ"lus is defined as streuf I'.gure 22 ·n.is

deformatioll is compouOOcd
straill ror a t<:. \li de .Iress adillg in one directi<m , wi lh from the two shean "'.. - - II"..,.
Ihe sides a spccim""l k.-ft rroc:, Pois..o,,'s mlio is do.'-
Ilnc.:d Illi (liuiw)/(fi/l) Sor this situation.

10. Tran.weru .votJe ve/Q(:ity, Show that the ,,'elocit y of transverse wa\·eS in the [1111
direction of a cuhic crystal is given by v. - II(C LI - CIS + C. 4)/pJUI. Hirlt: See
Prohlem 9.

II. E,ffedioe shear constml l. Show that thc shcar COIlstant t(C 11 - C IIll in a culJic crys-
tal is (.lcllncd by setting e." ... - e w = k and all other stmins equal to 7.ero, as in Fig.
22. Hint; Consider tile energy density (43); look for a C' sudl that U - IC'e",

1.2. Delerminantal apprf1fUh. II is known that an R·dimensional square matrix wi th all

e1emellts equal te unity has roots Rand 0, with the R occurring 0JlC(] and tile zero
occurring R - I limes. If all elements ha\"C the vaIue p, then the roots are Rp and o.
(a) Show that if the d~nlll elements are q and all other element!'> are 1', then there
is one root equ~ to (R - I)" + If and R - I roolS e<IlIal to (I - p. (b) Show from the
elast ic equation (57) for a wave in the [III J direction of a cubic crystal tim! the
de terminantal equation which gi\'cs Wi as a fUllction of K is

P p
'1 _ wtp P -0 ,
I' p 'I - ".}P
whe re q . jKZ(c n + 2C-w) lind I' . jKZ(c lt + C~J. 'I1l i~ exp~es the condition
that t hree linear hom(lgeneous algebraic equations for the three dispL'\CCmelit com-
ponents u, v, W !lave a solution. Use the result of part (a) to find the three roots of
".}; dleck with the results given for Problems 9 and 10.

13. ~ propagatian direction. (a) By sul~titution iu (57) find tile dcterminanta!

equation wh idl expresses the condition that the displacement

R(r) "' I~ + Our + %il exp II(K· .· - ...t)J

be a :'Iolution oflhe clastic wa\'e equations in a cubic crystal. (b) TIle sum oftlw roots
of a dcterminan tal equation is equal to the sum of tile dlap" ... aI clements a~ . Show
 3 Ela.tic Com,,,,,,.

from part (a) tlmt the slim ofthe squares oflile three elastic wa\-e velocities in any
dirl.'Ction in a cubic crystal is equal to (C II + 2C 44 )1P. Recall that v~.., .,}1K.'l.

14. Stability criteria. The criterion that a cuhic crystal with olle atom in the primith-e
l.ocll be stable against small homogeneous deformations is that the energy denSity
(43) be positiVe for all combinations of strolin components. \Vlmt restrictions are
tllereby imposed on tile e lastic stiffness constants? ([n matllematical language the
problem is to find tile conditions that a real symmetric qua..iratic form should be
posili ~-e definite. The solution is gi~-ell in books on algebra; see also Kom and Kom,
Mathematicnl Handbook, McCraw-Hili, 1961, Sec. 13.5-6.)Ans. C~4 > 0, C II > 0,
C~, - ~t > O. and C + 2C , t > 0. For an example or the instahility whicll results
when Cll "'" C,t see L. " R. Testardi ct al., l'hys. Re~- . Letters 15, 250 (1965).

References
M. 80m alld K. Iluang, Dynamical thoofY of r:rysltJl ftJlti('£! , o..ford, 1954. The das;ic work on
ionic crystals.
G. Leibfried, in Enc!}Clo. afphysics 7/ 1,1955. Superb rl.'Vil."W oflaUk:e properties.
M. P. Tre;i, "Cohesion of iunic solids in Ihe &1m model , ~ Solid stale phy5ics 16, 1 (1964).
L Pau li ng, NtJlurll oft/", du:mir.al bond, 3rt! ....J., CorneD, 1960.
J. llirllChfcldcr. C. F. Curti•• and R. B. BinI, Molecular lheory of gtlSI':$ and liquids, Wi"'')'. 1964.
G. K. Ilorton, Mld"'3.1 rare gas crystals." Amer. J. Phy. 36, 93 (1968). A good nwit.'W of the lattice
propt.'TIics of crystals of Ihe inl.'TI gases.
T. K. CaylonI. ..T.....sllr dl.-.cription of physical propertil.'S of crystal5." Arner. J. Phy$. 43, 861
(1975).
M. D. 'oe;len 'lIId L . Schaad. llydrogtn bonding, Dekker, 1974.
W. B. Pearson , Crystal dtemi.\try and phy81c8 of metals and alloys, Wiky- Inter.;cience, 1972.
F. A. Cotton and C. Wilkinson. A,k;anced Inorganic chemi.\try, 41h ....J .. Wiley, 1980.
G. Maitl:md et aI. , Intennolecular forc'.s: tlodr origin and determination. OKford, 1981.
II. , . ,u",t>dJkc. Crys/al physic,: mm;roscolnc l'Ioysic8 of anisolropic solids, Benjamin. 1974.
R A. Cuyer. "Physic~ of quantum crystals," Solid 5talc physics 23. 413, (1969).
J. F. Nyc, "IoyslClJI pro,w.rlies of crystals: fhdr reprnenta/km bylO!l'l3OTS and malrices, o..fonI,
1957.
II. B. Hunlington, ~Elaslic constants of cryslals,~ Solid stale phy~ics 7. 213 (1958).
A. E. II . Love, A trootise on Ihe mathematical/lowry of ellJ$ticity . Dover P"Pl.'J"oo"k, 1944.
w. P. Mason, "hys/ad llcoustic$ ami tlo" Pf'01",rlies ofsdlds, Van N05tranil, 19;;6.
W. P. Mason, ..-d .• "hysical tJCQlIsllcr, in s•.'v eral volu mes, Academic Press, Vol I, 1964.
C. z.." ICI", ";la$tidly and anellJ$ticily of fIUi/tJls. UniVl.',."ity of Chicago PrU's, 19-18.
T. A. Bak, ....J., l'honom an,1 phonon in/erne/ions, Benjamin, 1964.

\
(

•
 (

4
Phonons 1. Crystal Vibrations

VIBRATIONS OF CRYSTALS WITH MONATOMJC BASIS 99

First Brillouin wne 100
Group velocity 102
Long wa,,-elengtll limit 103
Derivation of force constanL~ from e1<Jlflriment 103

TWO ATOMS PER PRIMITIVE BASIS 104

QUANTIZATION OF ELASTIC WAVES 107

PHONON l\(QMENTUM 108

INELASTIC SCAITERING BY PHONONS 110

SUMMARY 112
PROBLEMS 112

1. Monatomic linear lattice 112

2. Continuulll wa,,-e equation 112
3. Ba.~is of two unlike atoms 112
4. Kohn anomaly 113
5. Diatomic chain 113
6. Atomic vibrations in a metal 113
7. Soft phonon modes 113
REFERENCES 114

Chapter 5 treats the thermal pmperties of phonons. Chapler to treats the interaction of
phullons and photons, and tim interaction of plJOnons and electrons.
 -
_~l- 11~oo
- I',""1roo - -------- ~'idtl

-./II!I"--- 11,,,,,,,,,
--ilt-- ....moo
, ___~.~':!:n"li""ion "a,,,
""Iamn Ek-<.-tro:n + d."t>c defor"," ion
.!
I
f'ib'Urc I Import:ml elementary excitations in solids. The origin" of the
CQI"""pts ancl the names of the excitations art': di$CU$.w hy C. T. Walker
and C. A. Slade. Am. J. Phys. 38, 1380 (1970).

,, ,
i
'f,, y,
,,
'i, 'j'
:
, ,,,
.. ~,
i
?'''.-. Y
~,
,
" ~,
,, " II•• ,
,, U•• t
,
, I
? .,,
,-
.;;, ~.
,, ,I
,i , :,, , Fig"re 2; (Dasheu lines) Planes of atoms
when in •.'q uilibrium . (Solid lincs) Planes

... ..
J,

.., ...l',
y, of atoms whf,.'" di$plaeed a.s for a IOllgit udi ·

.'-f,
(
, - .~ ,, , "al Wa'"c. The c.)(>r<linatc II measures the
• di$placcmcn l of the planes .

II. ,'It ..

"'igu..., 3 planes of atoms ll$ Jispl~ .lUTing

passage of a tronsv\ wave .

"
 CHAPTER 4: PHONONS t. CRYSTAL VIBRATIONS

VIBRATIONS OF CRYSTALS WITH MONATOMIC BASIS

Consider the elastic vibrations of a crystal with one atom in thc primitive
cell. We want to find the frC<Iuency of an clastic wave in tenns of the wavevcc-
tor that describes the wavc and in tcrms of the clastic constants.
111e mathematical solution is simplest in the [100], [110], and [I ll] propa-
gation directions in cubic crystals. 111cse arc the directions of thc cube edge,
face diagonal, and body diagonal. Wbcn a wave propagates along one of these
din.'(:tions, entire planes of atoms move in phase with displaet:mcnts either
parallel or perpendicular to the direction of the wavevcctor. Wc can describe
with a single coordinatc u~ the displacement of thc plane s from its equilibrium
position. The problem is then one dimensional. For each wavcn'dor there arc
three modes, one of longitudinal polarization (Fig. 2) and two of transverse
polari"l.ation (Fig. 3).
'''e asSllme that the clastic response of the crystal is a linear function of the
force5. 111at is equivalent to the assumption that the clast ic encrgy i5 a tluad-
ratic function of thc relative displacement of any two poin ts in the crystal.
Tenns in the energy that arc linear will vanish in etIuilibriIJIll- scc the mini-
mum in Fig. 3.6. Cubie and higher order terms may he negl('cted for suffi-
ciently small clastic deformations, but play some role al high temperatures, as
we see in Chapter 5.
We a.~sulil e accordingly that thc lorcc on the plane $ caused by the dis-
placement of the plane $ + P is proportional to the difference u.+" - u.oftheir
displacements. For brevity we considcr only nearest-neighbor interactions, so
that p '" ± I. The total force on $ comes from planes $ ± I:
(I)
This expression is linear in the displacements and is of the form of Hookc's law.
Tbe constant C is the force constant bern'ccn nearest-neighbor planes and
will diffcr for longitudinal and transverse wavcs. It is convenient hcrcafter to
regard C as defined for one atom of the plane. so that F. is thc force on one atom
in the plane s.
The etjuation or motion of the plane s is

(2)

where !II is the ml of an atom. We look for solutions with all displacements

.
100

having the ti me dependence cxp(- iUlt). Then cl2u/clfl = -urll., and (2) be-
rome<

-MiIfu, "" qUa+l + U._I - 211,) . (3)

This is a difference equation in the displacements II and has traveling wave
solutions of the form :
fl,% I = 1I exp(isKll) exp( ± iKa) • (4)
where tI is the spacing between planes and K is the wavevc«or. The value to
usc for a will depend on the direction of K.
With (4), we have from (3):
-w'Mu exp(i&Ka)= Cu{cxP(i(s+ 1)Ka]+ cxp[i(.t- l )Ka] - 2 exp(isKa)} . (5)
We cancel u exp(isKa) from both sides. to leave
w'lM -.:: -C{cXP(iKa) + cxp(- iKa) - 2J (6)
With the identity 2 cos Ka = cxp(iKa) + cxp{- iKa), we have the dispersion
relation connecting wand K
Ill- = (2CIM )(1 - cos Ka) . (7)
The boundary of the first Brillouin 7.onc lies at K = ± TrIa. We shO\y from
(7) that the slope of w versus K is zero at the zone boundary:

(rorlrlK=(2CalM) sin Ka :::: 0 (8)

at K = ± TrIa, fur here sin Ka =sin (± 11") = O. The special significance of
phonon wavcvectors that lie on the zone boundary is developed in (12) below.
By a trigonometric identity (T) may be written as
c.J '" (4CIM) sin2 iKa ; w "" (4CIM)I~sin iKaI (9)

A plot of w versus K is givcn in Fig. 4.

FiNJt Brillouin Zone

What range of K is physically SignifICant for elastic waves? Only those in
the first Brillouin zone. From (4) the ratio of the displacemcnts of two succes-
sive planes is given by

"...
- - • u expli(s+ l )KaJ • "",,UG»
J'
. (10)
u~ 1/ exp(isKa)

The range - Tr to +11" for the phase Ka covers all independent values of the
exponential.
There is absolutely no point in saying that two adjacent atoms are Ollt of
phase by more than 1T: a rclative phase of 1.211 is physically iden tical w ith a
,-
'"
(
u

' 0

0
••
(4Cf Af)1Ii
0.'

••
"
0
-. o
__ '__,1
I1-- - - - Fint U.illouin UJnc >oj

F"tgure 4 Plot of", versus K. n.e ..eglon or Ii: -C 1/11 or A .. II wrTe'!iponds to the <;OIllin-
uurn approximation; IlI.Te '" i5 directly proporti.)lIal to K.

f1gure 5 TIle ....'avc represcnlt.'t.! hy tIle iOl id curve COIlve>'s n(\ .nfonnalion not given hy IheJashed
curve. On ly wavl"l.:ngth. 1000ger Ib..11 2d ~ oec:ued to represent the motion .

relative pha.~e of - O.8n, and a relative phase of 4.2nis identical with O.21T_ We
need both positive and negative values of K because waves can propagate to the
right or to the left.
The range of independent values of K is 5pccilled by

- 'IT < Ka .:s;; Tr, 0' "•

- - < K $;-. "
•
This range is the first Brillouin 7.onc of the linear lattice, as defined in Chap-
ter 2. l1le extreme values arc K.,..,. = ± mll.
There is a real difference here from an clastic continuum: in the continuU lIl
limit a -+ 0 and K",u -+ ± 0:.. Vallies of K outside of Ihe first Brillouin zone
(Fig. 5) merely rep roduce lattice motions described by values with in the limits
± TrIa .
We rna)' treat a value of K outside these limits by subtracting the integral
multiple of2Trla that will give a wavevcctor inside these limits. Suppose K lies
102

outside the first zone, but a related wavcvooor K' defined by K' - K - 2mt1a
lies within the first zone, where II is an integer. nCIl the displacement ratio
(IO) becomes
11. .... /11. = exp(iKa) . exp(i21T11) exp[i(Ka - 21m)] _ cxp(iK'n). (11)
becallse cxp(i21Tn) = 1. 11ms the d isplacement can always be described by a
wavcvcctor within the first 'l.onc. We note that 2TTFlla is a reciprocal lattice
vector because 2mtJ is II reciprocal lattice vector. Thlls by subtraction of an
appropriate reciprocal latlk-e vector from K, we always obtain an ClJllivalent
wavcvcctor in the first zonc.
AI the boundaries K~ = :!: rna of the Brillouin zone the sohltion u. =
II cxp(isKa) docs not represent a traveling \\ll\VC, but 3 standing \\lave. At the
zone boundaries .~K". ...a = :!:S1T, wilcnl'C
II. = U cXp( ::!:: is?T) "" II ( - l}, (12)
This is a standing wave; alternate aloms oscillate in opposite phases, because
u. = ± 1 according to whether.t is an even or an odd integer. The wave moves
neither to tJle right nor to the left
This situation is equivalent to Bragg rcllcction of x·rays: when the Bragg
condition is satisfied a traveling wave cannot propagate in a lattice, but through
suocessive reflections back and forth, a standing wave is set up.
The critical value Kma:c = ±ma found here satisfies the Bragg condition
2t1 sin 6 = IIA: we have (J = iTT, d = n, K = 2Tr1A, II = 1. so that A = 2a. With
x-rays it is possible to have" C()ual to other integers besides unity because the
amplitude of the electromagnetic wave has a meaning in the space betwccn
atoms, but the displacement amplitude of an elastic wave mually has a meaning
only at the atoms themselves.

Group Velocity
The transmission velocity of a wave packet is the group velocity, given as
vII: = dcJdK ,
0'
(13)
the gradient of the frequency with respect to K. This is the velocity of energy
propagation in the medium.
With the particular dispersion relation (9), the group velocity (Fig. 6) is
vj!: = (Ca 2/M)l ll cos iKa . (14)

This is zero at the edge of the zone where K "" TrIo. Ilere the wave is a standing
wave, as in (12), and we expect zero net transmission velocity for a standing
wave.

• -
 .( pJocnon. I. C."./Dl VibratioN 103

' · "r~o:::c-----'-------'

(C"'f"O'~

J
"
Figure 6 Croup vel<Jrit)' ". VenU5 K. for modd d
Fig. 4. At the zone boundary the group >e Wry 15
tt."ro.
.,
,,~------------.}-------------'!
,

Long Wave-lengt11 Umit

When Ka <t I we expand cos IVI iii I - HKaf, so that the dispersion rela-
tion (7) becomes
(\5)
The result that the frC<luency is directly proportional to the wavcvcctor in the
long wavelength limit is equivalent to the statement that the velocity of sound
is independent offrctluenc)' in this limit. Thus v = wlK, exactly as in thecont in -
Ilum theory of el a.~tic waves- in the continuum limit a "" 0 and thus Ku = O.

Derivation of Force COllstarlts from ErlJer1merlt

In metals thc effective forces may be of quite long range, carried from ion
to ion through the conduction electro,! sea (Chapter 10). Interactions have been
found between planes of atoms scparoted by as many as 20 planes. We can
make a statement about the ronge of the lorccs from the obseTVcd dispersion
relation for Q). 11\e gcneralization of the dispersion relation (7) to V nearest
planes is easily found to be
w2 = (21M) L C,.(l - cos pKa) . (100)
",
We solve fo r the interplanar force constants C p by ffili itiplying both sides
by cos rKil, where,. is an integer, and integroting over the ronge of indepen-
dent values of K:

M
f"'·
- wid
dK Q)~ cos ,.Ku = 2 L
p >O
Cp f"·
- wi"
dK (l - cos pKa) cos rKu

= - 27rC,1o . (\6b)
The integral vanishes except lor p ;; r. Thus
MaJm. rlK w~ cos 7JKa
C,• = - -2 (17)
7T _ ."/,,

gives the fOrCe COl nt at rangc po, for a structure with a monatomic ba5is.
,0<

TI'VO ATOMS PER PRIMITIVE BASIS

The phonon dispersion relation shows new features in crystals with Iwo or
more atom~ per primiti\'c basis. Consider. for example, the NaG) or diamond
structures, with two atoms in the primiti\'c cell . For each poJariz.'ttion mode in
a given propagation direction the dispersion relation w versus K develops two
branches. known as the acoustical and optical branches. We have longitudinal
LA and transverse acoustical TA modes, and longitudinal LO alld transverse
optical TO modes, as ill Fig. 7.
if there arc p atoms in the primitive cell, there arc 3" branches to the
dispersion relation: 3 acoustical branches and 3p - 3 optical branches. Thus
germanium (Fig. Sa) and KBr (Fig. Bb). each with two atoms in a primitive cell,
have six branches: one LA, one La, two TA. and two TO.
The numerology of the branches follows from the number of degrees of
freedom of the atoms. With p atoms in the primitive cell and N primitive cells,
there are pN atoms. Each atom has three degrees of frct."<.!om, one for each of
the %, y, z directions, making a total of3pN degrees offrecclom for the crystal.
The number of allowcO K values in a single branch is just N for one Brillouin
mne. I Thus the LA and the two TA branches have a total of3N modes, thereby
accounting for 3N of the total degrees of freedom . The remaining (3p - 3)N
degrees of freedom are accommodated by the optical branches.
We consider acubic crystal where atoms of mass All lie on one set of planes
and atoms of mass M2 lie on planes interleaved between those of the first sel
(Fig. 9). It is not essential that the masses be different, but either the force
constants or the masses wiJl be different if the two atoms of the basis arc in
nonequivalent sites. Let a denote the repeat distanee of the lattice in the direc-
tion nonnal to the lauke planes considered. \\le treat waves that propagate in a
symmetry direction for which a single plane contains only a single type of ion;
such directions are (ill] in the Nael structure and llOO) in the CsCI structure.
We write the equations of motion under the a~sumption that each plane
interacts only wilh its nearest-neighbor planes and that the force constants are
identical between all pairs of nearest-neighbor planes. We refer to Fig. 9 to
obtain
dZu
M. d/= qv. + V. _ I - 211.) ;
(18)
d'v.
Mz df- = qu.... I + II. - 21).)

'We sho;nu in Chaplc:r:5 by applM.-atioo oi periodIC' boundary c:ond,hons to lhe: ITlo!lda oi lh",
crystal oi \olulTle V that lhere . one K value in the volume (2 ,,)'1V In Foul'ief space'. n.e volume of
a Brillouin zone is (2rrj'IV,. ",he..., V~ is the volume oi . ~rys;tdl llrim'tM: cell. nUl5 ,he number oi
aU(M\ed J( values in a BrillOUin zone is VlV~, ,,'Iuch IS just N, lhe number of primUlYe cells in the
crystal.
 - - - :(2CIM.)1I2
i
j (2CIMI)'"

Figure 7 O"tk:il and aOOOllitical branch« of the

dilJ~ion rewt..., fOl" • d iatomic linear l~ltk-c,
-.....
ph01l011 br.nMl
,,
,,,
,
showmg the hmiting frequen..;el at K " 0 and ~----------~---<
- ~

K - K.- - ../1f. 1he lattice comta"t il a. •

"',--------------,
, , • •• •'"

• 6
:" . • TO
•

"J,, ,
f•
'A
• •
•
• 1
• •
, J •
• • • TA
•

,. ,.,
" o.~
1>.'''-.. In \1 111 d"cdion

Figure Sa Phonon dlipt:TSion rcbtionlln the 11 11) direc-
lion ill gl'rm.,ullm at 80 K 11>e two TA phol1oo bnmchcs
are OOrW'nt.ll at the roue bound.uy posi tion . "-- -
(2..lo)(IUl· The LO aoo TO branmes t"Oilicide at K - O.
this IlIso ,. II t"OIISCqutnCe of U,e crystal symmetry of~.
1n.e resul ts '\l're obtamed with neutron inelastic" K-al te r-
in~ by G. "ih~ and G. Nciin
Figure 8b Dispersion ClUVt'S In Ihe \ 1111
di rt"Ction in KBr al 90 K. alIer A. D. B.
Woods, B S Broc..... OOUK. R A. eo..,'''."y,
aud W. Co<-hran 11>e extr1lllolJlion to
K .. 0 of lhe TO , LO branches lIlT (oalk-il
w" w,.; Iht·5t'l.fC diSl""ussed m Chapter 10.

We look for a solution in the form of a tra\'cling wan', now with different
amlllillldes II , t; on alternate plalles:
UA = 1/ exp(isKa) exp{-iwt) ; v. "" v exp(isKu) E';\.p{-iwt) (19)
Recall the definition of u in Fig. 9 as the distance bct\\'eE'n nearest identical
planes, not nearest-neighhor planes.

-
u.
0,_, .... ,
- - ".
-
Il...

- K

•
M, M.
• • •
L- .- J
.' igure 9 A d iatomic CI'}'lil m\ Sl ru<.'turc with mll'.cs lit" M~ l'On"cctcn by rOR'f: eOIll;talll C between
~t1jao..'en l plane s. The d;sl'\ac-o.menu of atoms M , are denoted by 11._,. It,." . .,
Il.. alld ofaloms
M~ by 1>._ , . v .. v•• ,. The repeat dislanl~ is .. in the direction of the w,wcv<:Ctor K. The: atom$ are
shown in their u"d ... plact.-d positions .

.On substitution of (19) in (18) we have

- w2 M,u = ev{ l + cxP( - iKo») - 2Cu
(20)
2
- w M2v = Cu(clql(iKo) + I) - 2Cu .
The homogeneous linear equations have a solution only if the detenninant of
the coefficients of the unknowns u, v vanishes:
2C - M,ur -C( I + c><p{-;Ka)J
~O (21)
-C(I + c><P{U(o)j 2C - M'JlW2
0'

\\le can solve this equation exactly for w2 , hut it is simpler to exam ine the
limiting cases Kn <4 1 and Ka = ± 1T at the zone boundary. For small Ka we
have cos Ka 3! 1 - j K2a 2 + . .. , and the two rools arc

(optical hrarlCh) (23)

(acoustical branch) . (24)

Thc cxtcnt ofthc first Brillouin zonc is - mo:S K :S mo, whcrc a is thc repeat
distance of the lattice. At K.nu = ± mCI the roots are
(25)
The dependence of w on K is shown in fig. 7 for M. > M 2 .
The particle displacements in the transverse acoustical (TA) and transverse
optical (fO) branches arc shown in Fig. 10. For the optical branch at K = a wc
find. on substitution of (23) in (20),
~= _ M2 (26)
V M.
 .. Phono... I. Cry.tcl Vi/m!tioru 1" 7

-
figure 10 TransvcnlC optical and tran~ .
....,ne k()ustical "''aVes in a di~tomic linear
lattice, ill",tratoo by the particle displace-
ments fw the hYO modes at the same
"'lI"dength.

111e atoms vibrate against each other, but their ccnter of mass is flXcd. If tbe
two atoms C.1rry opposite charges. as in Fig. 10, \\IC may excite a motion of this
type with tbe electric field of a light wave, so tbat tbe branch is called tbe
optical hranch. AI a general K the ratio ulv will he complex, as follows from
either of the equations (20). Another solution for the amillitude ratio at small K
is u = v, obtained as the K = 0 limit of (24). TIle atoms (and their center of
mass) move together, as in long \vavelength acoustical vibrations, whencc the
term acoustical branch.
Wavelike solutions do not exist for certain frequencies, here between
(2CIM 1)"'2 and {2CIM.JII'l. This is a characteristic feature of ela.~lic w<'!ves in poly-
atomic lattices. There is a frequency gap at the boundary K,,,,,. = ± ma of the
first Brillouin zone. If we look for solutions in the gap with w real, then the
wavevcctor K will he complex, so that the wave is damped in Sp..1CC.

QUANTIZATION OF ELASTIC WAVES

The energy of a lattice vibration is quantized. The qU.1lltUIll of energy is

called a phonon in analOb'Y with the photon of the electromagnetic wave. Elastic
waves in crystals are made up of phonons. Thermal vihrations in crystals arc
thermally excited phonons, like the thermally c\cited photons of black-body
electromagnetic radiation in a cavity.
Ine energy of an clastic mode of angular frequency w is
to = (n + j )Aw (27)
when thc modc is cxcitcd to quantum number n; that is, when thc modc is
occupied by n phonons. The term jAw is the zero point energy of tllc mooe.
H OCC\Irs for both phonons and photons as a consequcnce of their equivalcnce
to a quantum harmonic oscillator of frequency w, for which the encrgy cigcn-
valll(.'S are also (n + ! )Aw. The quantum theory of phonons is dcvelopcd in
Appendix C.
108
(
\ \'e can easily quantize the mean square phonon amplitude. Consider the
standing wave mode of amplitude
lJ = Uo cos Kx cos wt .
Here u is the displacement of a volume clement from its equilibrium position at
x in the crystal.
The energy in the mode. as in any harmonic oscillator. is halfkincUc en-
ergy and half potential energy. when avcr.lgcd over time. The kinetic energy
density is lP{iJu/ill:r. where p is the mass density. In a crystal orvolume V, the
volume integral of the kinetic energy is tPVcJuX sin 2 wI. The time average
kinetic energy is
tpVtJlug = HfI + i)hw (28)

and the square of the amplitude is

ug = 4(11 + l)/llpVw . (29)
This relates the displacement in a given mode to the phonon occupnncy n of the
mode.
\\'hat is the sign of w?Thc equations of motion such as (2) arc equations for
w2 , and if this is positive then w can have either sign, + or - , But the energy of
a phonon must be positive, so it is conventional and suitable to view w as
positive . (For Circularly polarized waves both signs are often used, to distin-
guish one sense of rotation from the other.) Uthe crystal structure i.s unstable,
or becomes unstable through an unusual temperature dependencc of the force
constants (Chapter 13), then w'l will be negative and w will be imaginar),.
A mode with w imaginary will be unstable, at least ifthc real part of w is
ncgative. The crystal will transform spontaneously to a more stablc structure.
An optiC'.!! mode with w closc to zero is called a soft mode, and thcsc arc oftcn
involvcd in phase transitions, as in ferroe lectric crystals.

PHONON MOMENTUM

A phonon of wavevCdor K will interact with particles such as photons,

neutrons, and electrons as if it had a momcntum 1zK. However, a phonon docs
not carry physical momeotum.
The reason that phonons on a lattice do nol carry momentum is that a
phonon coordinate (except for K = 0) involves relative coordinates ofthc atoms.
Thus in an H2 moleculc the internuclear vibrational coordinate T( - T2 is a
relative coordinate and docs not carry linear momcntum; the centcr of mass
coordinate H T( + ril corresponds to the unifonn mode K = 0 and can C'dlT}'
linear momentum .
- " I'hononll I . Cryolal VibratiOn.l 109

Thc physical momentum of a crystal is

p = Med/dl) L II.. (30)
When the cryslal earri(.'S a phonon K,
p = M(dllldt} L cxp{is Ka) = IlI(duldt}[1 - cxp(iNKa)!1 (1 - cxp{iKa)! (31)
•
where S Tuns over the N atoms. We have used the series
N- '
._
2:0,. = (I - '"Y(I - ,) . (32)

[n the next chapter we enumerate the discrete vaJues of K compatihle with

the. boundary conditions, to find that K = :!: 2-rrrlNo.. where,. is an integer.
Thus cxp(iNKa) = exp(±i21Tr) = I , and from (3 1) it follows that the crystal
momentum is zero;

p = M(duldt) L cxp(isKa) = 0 (33)

•
The only exception to (33) is the uniform mode K = 0, for which all II. equal II,
so that p = NM(dllfdt). This mode represents a uniform translation of the crys-
tal as a whole. and such a translation docs carry momentum.
All the same, for !nost practical purposes a phonon acts as ifits momentum
were 1iK. sometimes called the crystal momentum. In crystals there exist
wavevcctor selC<"tioll rules for allowed transitions between quantum states. Wc
saw that the clastic scattering of an x-ray photon by a crystal is governed by the
wavcveclOr selection rule
k' = k+G, (34)
wherc G is a vector in the rcciprocallatticc; k is the wavcvector ofthc inciuent
photon, anu k' is the wavevector of the scattcrcu photon. In the reflection
process the crystal as a whole will recoil with momcntum - hG, but this uni-
form moUe momentum is rarely collsiuered explicitly.
Equation (34) is an example of the nllc that the total wavevcclor of inter-
acting wavcs is conserved in a periodic lattice, with the possible adclition of a
reciprocal latticc veclor C. The true momentum of the whole system always is
rigorously conserved.
lflhe sC:lttering of the photon is inelastic, with the creation of a phonon of
wavevectOr K, then the wavevector selection rule hecomes
k'+K = k +C . (35)
If a phonon K is absorbed in the process, we have instead the relation
k'= k+K + C . (36)
Relations (35) and (36) are the natural extensions of (34).
."
We exhibit the mathematics involved in the wavevcclor selection rule.
SUPlXlSC two phonon! Kio K2 interact through cubic terms in the elastic energy
to create a third phonon ~. The probahility of the collision will involve the
product of the three phonon wave amplitudes, summed over all lattice sites:
(phonon KI in)(phonon K2 in}(phollon l\:) out} 0:

" "
This sum in the limit of a large number of lattice sites approaches zero unless
~ "" KI + K2 or ~ "" K J + K2 + C. Ifcithcr of these conditions is satisfied , of
which the first is merely a special case of the scrond, the sum is equal to the
number of lattice sites N. A similar sum was considered in Problem 2.4.

INELASTIC SCA'ITERINC BY PIiONONS

Phonon dispersion relations <t.{K) arc most often determined by the inelas-
tic scattering ofncutrons with the emission or absorption of a phonon. Further,
the angu lar width of the scattered neutron beam gives information bearing on
the lifetime of phonons.
A neutron sees the crystal lattice chiefly by interaction with the nudei of
the atoms. The kinematics of the scattering of a neutron beam by a crystal
lattice are described by the general wavevcctor selection rule:
k +G = k'±K, (37)
and by the TCcjuirement ofconscrvalion of enerb'Y. Here K is the waveycctor of
the phonon created (+) or absorbed (-) in the process, and G is any reciprocal
lattice vector. For a phonon we choose G such that K lies in the first Brillouin
zone.
The kinetic energy of the incident Ileutron is p2/2M ... where M" is the mass
of the neutron. The momentum p is given by ~k, where k is the wavcvcctOl' of
the neutron. Thus f/ 2k'J.12M" is the kinetic energy of the incident neutron. Ifk'
is the waveyector of the scattered neutron, the energy of the scattered neutron
is (,2k'2/2M ". The statement of conservation of energy is
~2k2 ~2k'2
--~--±('W (38)
2M.. 2M"
where (,w is the energy of the phonon created (+) OT absorbed (- ) in the
process.
To determine the dispersion relation using (37) and (38) it is necessary in
the experiment to find the energy gain or loss of the ~rattercd neutrons as a
function of the scattering direction k - k'. Results for { manilllll and KBr arc
 4 Phor.Of" I . CrydaJ VibrntiOfLl 111

.
•
U'"' •e·' "
• IIIOj
••
• II LlI
••
.., .
", ,
' 0
.• • ,0
,
•
0
00
-

•
•

,
0

,"
• Longatooi""j
• 0
, 0 0

oT,... ",.:"" , 0

"" .00 " 110 0 III

\\\\,':"'<.10< in "ni" 2n/Q
Figure II n.e dis)X'rsion curves of sodium for-phonons pro.,agating in rl,c (OO I J. [IIOJ. and (I J IJ
dio«iioni at 90 I( "Ii detcrmined by inelastic scattering of neutrons , by Wood" Brod.:oousc.
March and 1J,o..('JS.

Figurc 12 A lriple a,is neutron spectromete r al Brookhaven. (CnurlC1'y of B. H. Crier.)

given in Fig. 8; results for sodium are given in Fig. 11. A spectrometer used for
phonon studies is shown in Fig. 12.
Recently the concept of " minor symmetry" has heen introduccd into the
study of alkali-halide ion dynamics. 2 The idea is to conside r the crystal that
would be fonned jfthe signs of the ions in A+C- were reversed. Now A- B+
docs not exist, hut a nearest mass pair that docs exist will have. it turns out
experimentally. a phonon dispersion relation remarkably similar to that of
A ~ B-. Thus KF is the approximate mirror image of NaCI. for K+ is isoclcc-
Ironic with CI- and F- is isoclectronie with Na+.
2L. L. Fokly and P "'egall. Phys. Rev. HiS , 1200 (1982); L. L. Foldy and T. A. Witten, Solid
Stale Commun. 37. 7{ lS I).
 SUMMARY

• TILe quantum unit of a crystal vibration is a phonon. [fthe angular frequency

is w, the cncrgy of thc phonon is liw.
When a phonon of wavcvcctor K is created by the inelastic Sc.1ttering of a
photon or neutron from ....-avevcctor k to k'. the wave\'ector selection rule that
govcrns the process is
k=k'+K+C
where C is a reciprocal lattice vector.
• All clastic waves can be described by wavcvcctors that lie within thc first
Brillouin wnc in reciprocal SP.1CC.
• Ifthcre arc p atoms in thc primitivc cell, thc phonon dispersion relation will
have 3 amustka.1 phonon hrallchcs and 3p - 3 optical phonon branches.

Problems
1. Monatomic litlftlf lnttice. Consider a longitudinal wave
u, ~ u oos(wt - sKil)

,,"IIich propagates in a monatomic linear lattice of a!Om~ of mass M . spacing ll. and
lleare~t- neigM)Or illteraction C.
(a) Show d lat the tornl energy of Ule wave is

E- ,At L (du/d{f + jC L (u, - u.~ lf

wlle re s nlOs u""e r all ato ms.

(l» fiy slibstitution of II, in this e_pressioo, show Ulal the time·ave~e total cnergy
per atom is

where in the last step we have used the disper.;ion relation (9) for this problem.

2. C(mtinuurn wave equation. Show that for long wavelengths the e<luation of motion
(2) reduces to die continuum elastic wave C<llIation

",,,here v is the velocity of sound.

3. lIasis o!tlVolmlikeatom". For the problem treated by (18) to (26), fin{l the arnpli tll{lc
ratios I J V for dlC two bnmclle5 at K" ... "" Tria. show that at this value of K d Ie two
lattices act as if deroopled: oue lattice remains at rest while the other lattice moves.
 4 Phonon. l. C'lI",al Vibmtion.l 113

4. KoIm anomaly. \ Ve suppose d,al the interplanat fOrce constant C" between planes $
Qnll .f + Jl is of the fOrm

C , - A~inP'rkoo .
where A ami k.J are constants and p runs over all integers. Such a form is expect(.'(l in
metals. Usc this ant! E(l. (1&) to £lnll an c~prcssion for cJ and also for aw2/ilK. Prove
that aJ'iJK is infinite when K ., 4. Thus a plot of '" vcrsus K or of w versus K Illu a
vertil'ai tangent at "=t: there is a kink at ko in the phonon diSpersiorl relation w(K).

5. Diatomic choin. Consider the nonnal modes of a linear chain in ",ilich the force
constants between nearest-neighbor atoms are altemately C and IOC. Let the masses
be e<lual, 811tl let the nearest-neighbor separatkln he al2. Fin(l w(K) at K - 0 ami
K "" ma. Sketch in the disperston relation by ere. Tllis probJem simulates a crystal of·
diatomic molecules such as 11 2 .

6. Alomic vibrati<m8 ill a melal. Com iller point ions of mass M and charge e immerst'(l
in a uniform sea of conduction ek~trons, The ions Rre imaginal to be in shlble
C<luilibrilLm when at regular lattice points. rr one ion is llisplaced a smallliistan(:e r
from its e<luilibrium position, tile restoring force is largely due to the electric cllarge
within the spl'cre of radius r centered at the e<llliliblium position. Take the number
density of ions (or of oond uction electrons) as 314'11R', ",hid! defines R. (a) Show that
the fre<luency of a single ion .set into oscillation is w - (e t /,ol,fR')lI'2. (L) Estimate the
val ue of this frequeocy for sodium, roughly. (c) From (a), (b), and some oommOfi
sense, estimat e the orde r of magnitude of the velocity of sound in the metal.

·1. Soft p 11emon modea. Consider a line of ions of C<lual n"laSS but alternating in cilarge,
with c,, :;; e(- I)" as the charge on the rth ion. The interatomic potential is tlw su m of
two ('Ontriblltions: ( I) a short-range interaction of force constant C ,A - ")' tl,at acts
belwetm nearest neighbors only, anll (2) a OOlilomb interaction between all ions.
(a) Show that the contribution of the coulomb internctiOfi to the atomic force con-
slants is C,.c - 2(- lf rllrlct, where a is the C<lllilibrium nearest-neighbor d istance,
(h) From (IGa) show that the Jispenion relation may be written as

w'1~ - sin' iKa + 0' L (- I )" (I -

,-, cos pKa)p- 3 ,

where ~ - 4yM anll 0' - Cl-yol, (c) Show that J is negative (unstable mode) at the
zone boumlary Ka ,. '11 if 0' > 0.475 or 4n«3). where' is a Riemann ;r.eta funcliOfi.
Show further tI ,at the speet.! of soullll at small IW is imaginary if 0' > (2 In 2,- ' _
0.721. Thlls w~ goes to zero anll the lattice is IInnable for some value of Ko in the
interval (0. '11) if 0.475 < 0' < 0.721. Notice that ti le pllOnon spectnlm is not tI ,at of a
diatomic lattice hecame the interaction of any 100 with its neighbors is ti,e same as
that of any other ioo .

*This problem is rat her dilfoc·u lt.

--
114

Refertmces
II . llilz and W. Kn..u, Phc"'nn disJlf!nion rdation! in i"$llla/I)r" Sprin~r. 1979. An atlas.
C. A. Ilorton and A. A. Mal"'<ludin, als., Dllnam.iooi,lro/lertlt.J til w/i!h, North-Holland. A contin-
"jug n'ries 1974- {l900).
W. P. ~Iason. ~"'., Ph ysic4l /lCOII&/iGf, Acatlemic Press. A continuing 5eriel. ,'Ok 1- (20).
W. Cochran, Dynamic:! of lIta"', In r.rys/ou, Crane, nuSllak, N. Y., 1973. A good It,);! .
J. A. Ik~ ••[and, Phlj'ic$ vf plot»wns, Wi le)', 1973.
II . Botl~r, Prindpie, of the theory of lattice d!lnOlllicr, Weillheim: Ph)"ik Verlag. 1983.
P. lkuL~h . P/woo",. throry lind C¥/lelinuonI3, Springer, 1967

\
 5
Phonons II. Thermal Properties

PIIQI'\ON HEAT CAPACIIT 117

Planck distribution 117
1'\0.,11111 mode ellumeration 118
Density of stales in one dimension 119
Density of slates in three dimensions 120
Ocbye model for density of slates 122
Ocbre T3 law 124
Einstein model of the dcnsily of stales 124
Ceneral result (01" D(w) 127

ANHARMONIC CRYSTAL INTERACTIONS 129

Thermal expansion 130

THERMAL CONDUCflVITY 131

Thermal resistivity of phonon gas 133
Umklapp processes 135
ImllCrfcdions 131

PROBLEMS 139

I. Singuhllity in density of slales 139

2. RnlS thermal dilation of crystal cell 139
3. Zero point lattice dislJlaccmcnt and strain 139
4. Heat c(tIJacity of laycr lattice 139
5. Grunc;scn constant 139

nEFERENCES 140

(
 ,

• ,
< .> ,/
, 1//
,,/'
V
, / '/
/ V
, ,//'/
/ /
, //, , , ,
Figure I Plot of Plwd tli$t ribulion function. AI h igh temperatures the oc:cupancy of a .tate is
a pproximately Jifa'ar in IhtJ Icmper.. lt>re. TIle f"OOion ( .. ) + t which is not plott~..I> appl'OlllChe' the
dashed lioc as asymptote at high lemperntures. n.., uadled lint: i. lhe dassic8llrmit.

U6
 CHAPTEf - . PHONONS U . THERMAL PROPERTIES

We discuss the heat capacity of a phonon gas and then the e ffects of an-
harmonic laltke interactions on the phonons and on the crystal. The rmal prop-
e rties of metals are treated in Chapter 6, superconductors in C hapter 12, mag-
netic mate rials in ChapteTS 14 and 15, and noncrystalline solids in C hapter 17.

PHONON HEAT CAPACITY

By heat capacity we shall usually mean the heat capacity at constant vol-
ume, which is more fundamental than the heat capacity at constant pressure,
which is what the experiments iletermine. I TIle heat capacity at constant \'01-
ume is defined as C\, - (aUlaT}", where U is the energy and T the te mperature.
The contribution of the phonons to the heat capacity of a crystal is called
the lattice heat capacity and is denoted by C .....
The total energy of the phonons at a temperature -r{- kBT) in a crystal may
be written as the sum of the energies o\'Cr all phonon modes. here indexed by
the wa\'e\'cctor K and polarization index p:

(I)

where (11K,,) is the thermal equilibrium occupancy of phonons of wavevcctor K

and polari7.3Uon p . TIle form of (11K.,) is given by the Planck distribution func-
tion:

(0) (2)
exp(licd"T) 1

where the (... ) denotes the average in thennal eqUilibrium . A graph of (n) is
given in Fig. I .

PlatlCk Di.s,ri/ro'ion
Consider a sct of identical hannonic oscillators in thermal equilibri um.
TIle ratio of tile number of oscillators in their (11 + l)th quan tum state of excita-
tion to the number in the nth ljuantum state is
(3)

I A thelltllld)"namic rdation gi,'Cll C. - C,," 9cr81'T, ,,·here 0' il the te mpc:r~l lIn: rodJ'rienl
of lillt.'ar expano;ion, V I.... ,oIume. arld 8 the bulk modulllS. The fractional dilTen:nce betwt:t."l l C..
and C v iJ UJuaily small and ol'lt."I1 may be ru:glt:cted . As T .... 0 ,....... set." that C ..... C v , pro",dul 0'
and 1J IR." comlant.
liS

by usc orthe Bolt..:mann factor. Thus the fraction of the totaJ number of oscilla-
tors in the nth quantum state is
N" cxp(-nliwlT)
-.-- ~ (4)
L
~mO
N, L
~_ o
e<p(-,n",h)

\Ve sec thai the average excitation quantum number of an osl'illator is

L
s cxp(-s/iwl"
(n) ~ -','---- (5)
L, cxp( - sliwh)
The summations in (5) are

1
L"
•
~--
1- x
, (6)

wit h % = cxp{-liwl'T). Thus we may rewrite (5) as the Planck distrihution:

, 1
(n) ~ - - ~ -=::;;;:';:;-::-;- (7)
I - x cxP(licd7) 1

Normal Mode Enumeration

The ene rgy of a collection of oscillators of frequencies wK.p in the rmal
equ ilibrium is found from (1) and (2);

It is usually convenient to replace the summation over K by an integral. Sup-

pose that the crystal has Dp{w)dw modes of a given polari ...alion p in the fre-
quency range w to w + dw. Then the energy is

'" Jdw Dp{w) cxp(liwl.ry

U = -;- n", 1 (9)

The lattice heat capacity is found by differentiation with respect to tempera-

ture. Let x = liwh = IiwlkBT: then iJUliJT gives

Cia, = kB L.
'" Jdw D,,(w) ( Xi exp x
1)2 (10)
" exp x
The central problem is to find D(w), the number of modes pe r unit fre-
quency range. This function is called the density of ( 'Ies or, more often.
 S p"",""", II. TMrrnd f'ropmia I.J

F~-J . .. 0
. . . :]-: . r-·1-- 1-t", , ~
I 2. . .. 10
Figure 2 Elasl i(" line of N + I aloms, "ith N - 10. for llOU I.,Jary colkbt ioru; Ihal Ihe eutl atoms
s'" 0 and, ... 10 are fhed. TI)C particle .mpID"l'"llICnts in the normal rnodd for eilhe r 10ngit..,jinaJ
or h"llns...:ne displaceme nlli are of the form. II, .. sin sKiJ. TIlis form is aulom~l il."ll!]Y 7.ero at the
atom 11.1 the end s ... O. and we ch()O.';C K to make the displacement l<..""ro at Ihe ...nd s ... 10.

, • ,. 'W,0.
100 100

FIg"re 3 n.e boundary coudition sin sKiJ ... 0 for, ... 10 can be ~Iisn..'tl by dlOOl;ing K '"
""'10<1. 21rllO<l • .•. • 9wI10<l. where lOu is the length L of the lille. The prellellt Ilgurc i. in K
space. The dot. art: not atoms but are the al lowed ,.,.Iues of K. or
the N + I particles 011 the lint:.
()Illy N - I an: alJo..·ed to lTIO\·e. and their mosl general motion can be cxpreu ed In terms of the
N - 1 allo-o.·w , ..lut'S of K. ·o,is quanli:tal.ion of K has notlling to do wllh quant um mechanics hut
follows dauically from the boundary conditions lhal the eno.I atoms be nxal

density of states. The best practical way to obtain the densily of stales is to
measure the dispersion relation w versus K in sc1octoo crystal d irections by
inelastic neutron scattering and then to make a theoretical analytic fit to gi\'e
the d ispersion relation in a general d irection. from which D(w) may be calcu·
!ated.
Derl.$ily of Slalca in O ne Dimensi01I
Consider the boundary value problem for vibrations of a olle·dimensional
lille (Fig. 2) of le ngth L carrying N + 1 particles at scpa ration a. We suppose
that the particles s = 0 and s = N at the ends of tile line :"Irc held fixed. Each
!lonnal vibrational mode of polarization p has the fOrm of a stand ing wave,
where II~ is the (lisplacement of the particle s:
u~ = u(O) exp(-fwK."t) sin sKa (II )
where wJ<.1l is related to K by the app ropriate dispersion relat ion.
As in Fig. 3, the wave\'ector K is restricted by the fixed-end boundary
oonditions to the "alues
~ 2~ 3w (N - l)w
K:::- L •... , L (12)
L ' L
1be solution for K = 'TriL has
u. CIC sin ($1M1L) (13)
and vanishes for s = 0 and $ = N as required.
l1le solution fo r K = N7T/L = wIll = K".,.,. has u. CIC sin $11"; this permits no
motion of any atom, because sin 511" ,·anishes at each atom. TII US there are
N - I allowed inrl ....»endent values of K in (12). ntis Ilumber is equal to the
l
.
,
number of particles allowed to m()\'c. Each allowed value of K is associated with
a standing wave. For the one-dimensional line there is one mode for each
intervai llK "" m L. 50 that the number of modes per unit range of K is U'lf for
K:$ ,"In , and 0 for K > 7T/a .
The re are three polarizations p fcr each value of K; in one dimension two of
these arc transverse and Olle longitudinal. In three dimensions the polariza-
tions arc this simple only for wavt.,\'eclors in certain special crystal directions.
Another device for enumerating modes is orten used that is C(lually valid.
\Ve consider the medium as unbounded, bul require thai the solutions be
periodico\'cr a large distance L, so that u(sa) = u(sa + L). TIle method of peri-
odic boundary conditions (Figs. 4 and 5) docs not change the physics of the
problem in all y essen tial respect for a large system. In the running wave solu-
tion u. = u(O) exp[i(sKa - wxt)] the allowed values of K are
N~
K = O, .. . , L
(14)

TIlis method of enume ration gives the same number of modes (one per
mobile atom) as given by (12), but we have now both plus and m inus values d"
K, with the intcrval llK = 2'1t1L bctween suC('Cssivc valucs of K. For periodic
boundary conditions the number of modes per unit range of K is U21r for
-1rla :s K ::s TTla, and 0 otherwise. The situation in a two--dimensional lattice is
portrayed in Fig. 6.
·We need to know D(w), the number of modes per unit frequency range.
The number of modes D(w) d w in dw at w is given in one dimension by
LdK L dw
D(w)dw=--dw =- · - - . (15)
TT dw 1r cbldK
We can obtai n the group velocity dwldK from the dispersion relation w versus
K. There is a singularity in D (w) whcnever the dispersion relation w(K) is
horizontal; that is, whenever the group velocity is zero.

Density af Stl/fCS in Three Dimensions

We apply periodic boundary conditions O"er N 3 primitive cells within a
cube of side L, so that K is determined by the condition
exp[i(Krc + K~y + KLZ)) • exp{i{K.{x + L) + ~(I) + L) + K:.(z + L))} (16)
whencc
2~ 4~ N~
K, ,~ , K:. = 0; +- .
- L • +- - L • (1 7)
L
TherefOre there is one allowed value of K per ,·olume (2TT1L'j1 in K space, or

(2TTL)' = 8r.
V (18)
 S f'hlllU,)M II. Thenncll'roptrfin 121

.-,
FIgure 4 CuISi<kr N partie....s WflSlraineo;! lo .Iide on acircular ring. 11-.e
ptrIidesc:an otci llale ifconntded by el:a:llicsprings. In a normal mode lhe
displace.....,.,1 u, of alom , ....·mbe of I~ ro.m sin .KG 01" 00II IlKG; Ihne are
independent n't:>cIu. B)' the geometrical periodicil)' oflhe ring the boo.>d·
aN wndition is Ihat U,,·•• - u. fOl" all s, so 110.'11 NKG must be an inle,gral
.-. .
.-
m~lliple of 2... . 'or N - 8 Ihe allowed ir>dept.'1l<Jent .-alll'" of K ..r.: 0,
271"&, 411"1'&, 6 ..1&, and 8 ..1&. TIoe value K - 0 is m..aninglt'SS for the
sine i(mn, because ,i n tOe> ,. O. The value 871"/& has I meaning onl)' for
the eosine rorm, box,.IIse sin (..e"<lIBa) - sin nr'" O. noe three other val _
.-,
ues of K are 1I1I0-...."i! fo r bolh the sine And rosine model, giving a total of
eighl alkowed rnQ\Ie, ro r the eigJ,t particles. n,,,s Ihtl perlndic boundary
condition k:ads to one allowed mo<k p•.:1· partid", eQCtI)' lIS fur thtl fixed-
(.TK1 IJoundary <."On<lition of Fig. 3. Ir we had taken tloe IllO<Ie:. in the com-
plcK klnn cxp(IsKG), the periodic boundary eoodi tiof. would lead 10 the
dght modes \\i!h K - 0, ~2mNa, ~4 ..INa, ~fmfNa, and 8mNa, as in
Eq. (14).

•
N.
-1:
.
•
••
••,• ;..
••
, ..
• •
O'
••
W•
• ,
•
N.

,- •• •

Figure S AlIowt-.l ,·alut's of "a"e"eclor K ror periooic ooundary conditions applk-.l to a lin(... r
\:Iui<;t: of periodiclt)' N - 8 atoms on a line of length l... 'rl!e K - 0 solution is the unlfurm mO<le,
The spccW poiuts ~ NmL represent onl), a singk solution bec:atlliC exp(imj is identical to
exp( - Im); I\u,s there art' eighl allo .....ed mocks, "il\, u.s~m(.... ts of the s th atom Proportional to
I, exp(.:I:1.....4). e.p( ~ lmI2j, ~p(~i3mi4), up(lmj.

•
•
• • • • • • • •
• • • ~
•
• • • •
• • • • • • • •
• • • • • • • •
• • • •
•
• • • • •
• • • • • • • •
• • • • • • • •
• • • • • • •
• • • • • • •

Flgu.e 6 Alla..-ed "",lues in FOUrier space of the phonon wa"e\"OOor K for a square lattice oflattice
mnstant G, ....,th periodic boundary wnditious appli...! ()\'t::r a square of side L - 104. ,,"" unilOrm
mode is marl."t.od ""til a cross. nlC~ is one allm>~ ,mlle of K per IlI"l:3 (2mIOa)l- (2mL"f, so that
"'i !hin the Qrc:ie of ar~a .. Ks the ~moolhed number of allcJ.,.-ed points is ","'U2'ftf.
'22
allowed values of K per unit volume of K space, for each polarization and fOT
each branch. 'Inc volume of the specimen is V = L3.
TIle tolal number of modes with wavcvcctor less than K is found from (18)
to be (U27T)3 times the volume of a sphere of radius K. Thus
(19)
for each poiari7.3tion type. The density of states for each polarization is
D{w) = dNltlw ::::I (VK2/2-zr)(dKJdw) (20)

Del}!JC Model for Density of States

In the Ocbye approximation the velocity of sound is taken as constant for
each polari:.r.alion type, as it would be for a classical elas tic continuum. 'Inc
dispersion relation is written as
w = oK . (21)
wit h 0 the constant velocity of sound.
The density of states (20) becomes
D(w) = Vw'l/2.,rv 3 (22)
If there are N primili\'c cells in the specimen, the total number of acoustic
phonon modes is N. A cutoS' frequency Wv is determined by (19) as
Wb = 6-rr'lrlNlV (23)
To this frequency thcre corresponds a cutoff wavevector in K space:
Kv = wdv :::: (67fNIV)1I3 . (24)
On the Dcbye modcl wc do not allow modes of w3\'c\'eclor largcr than KI) . The
num ber of modes with K :s; Kv exhausts the number of dcgrees offrecdom of a
monatomic lattice.
The thennal encrgy (9) is given by

u:::: f dw D{wXn(w)liw - r'" dw (2";:3) (ij:~w_ 1)

for each polari7.3tion type. For brevity we assumc that the phonon vclocity is
independcnt of the polarization, so that we multiply by the factor 3 to obtain

(26)

whcre x • "wi., >II " wlkBT and

Xo ~ "wu1kBT iiiili 81T . (27)
This defincs tile Dcbye temperature 0 in terms of WI) dcfined by (23). We
m ay express 0 as
o ~ ~. (&n'N)'~ (28)
kll V • (
 5 Phonom II. Thernwll'ropert;"1 123

/
V

, /
, II
rogure 1 lIeat capa<.~ty C,. of a SQlid , according to I
!he Deb)"e approximation. The \"ertical scale is in J
mol- I K- I . The horizon tal sc:Ile is the temperature
roormaliZl.od w the Debye tem(X'rature e. The m·
')
,
I

gioll of the T a la..· is below 0, I e. TIle asymptotic , " ,. ., 0.8 1,0 'A
, ..1..., at high val"es of TIe is 24.943 J mol- I deg- I. .,- " "
, C,·.m.'1llium l
,
•
,
, / /
t/
/
/ --- SilK.>on
I
I
!
// l
.·iguu 8 lIeat capacit)" of s~icon anti genna·
ilium. Note the m'<"l"ea5e at low temperat",..,..
Tocolwert a "at..., in caVmol· K to J/nlQl·K. m"l·
tiply by 4. 186.
:.1/
, '00
I
I

so thai the total phonon energy is

U = 9Nk T(~)' LXV dx ~ (29)

B 8 0 e" - I
",..here N is tbe number of atoms in the specimen and Xv = BIT.
The heat capacity is found most easily by differentiating tbe middle expres-
siOn of (26) with respect to temperature. Then

C
v
"=
3Vh 2

27ilti'kBT2
I"b dw (e"w"'....
0
e"wlT
T 1'1
= 9NkB _ (T)' LX" dx
0 0
X4
(e" -
e"
1)2
(30)
The Debye heal capacity is plotted in Fig. 7. At T J> 8 tbe heal capacity ap-
proaches the classical \laJue of 3Nk B • Measured values for silicon and germa-
nium are plotted i ig. 8.
..,
Debye T3 Law
At very low temperatures we may approximate (29) by letting the upper
limit go to infinity. We have

L" <Ix - ''L""

o
-=
1
IT -
d:u L exp(- n) "" 6 L
3 1 =~
" -.
." 1
' •
I S 15
(3 1)

where the sum over 6- 4 is found in standard tables. 1'nu5 U >;II 3-rr"NksT4l5rfJ for
T <t 9, and

Cv £!! 12;4 Nks( ~r a 234 Nk8( ~r (32)

which is the Debye r 3 approximation. Experimental res ults for argon are plot-
ted in Fig. 9.
At sufficiently low temperatures the r 3 approxi mation is qUite good; that
Is, when only long wavelength acoustic modes arc thermally excited. These are
just the modes that may be treated as an elastic continuum with macroS<.'Opic
elastic constants. The energy of the short wavelength modes (for which this
approximation fails) is too high for them to be populated significantly at low
temperatures.
We unde rstand the r 3 result by a simple argument (Fig. 10), Only those
lattice modes having flw < k8T will be excited to any appreciable extent at a
low te mpemture T. The excitation of these modes will be approximately classi-
cal, each with an ene rgy close to k8T, according to Fig. 1.
Of the allowed volume ill K space, the fraction occupied by the excited
modes is of the order of (wtlWv>3 or (K~KD)3, where KT is a "thermal" wavevec-
tor defined such that fl vKT = k8 Tand Kv is the Debye cutoff wavevector. Thus
the fraction occupied is (TIO'jl of the total volume in K space. There are of the
order of 3N(TIO'jl excited mocles. each having energy kHT. The energy is
- 3Nk BT(TIO'jl, and the heat capacity is - l 2Nk B{TJO'jI.
For actual crystals the temperatures at which the T3 approximation holds
are Cluite low. It may be neocssary to be below T = 0150 to get reasonably pure
T3 behavior.
Selected values of 0 are given in Table 1. Note, for example, in the alkali
metals that the heavier atoms have the lowest (/s, because the vclocityof sound
decreases as the density increases.

Einllldn Model of the Density of States

Conside r N OScillators of the same freque ncy Wo and in one dimension. The
Einste in density of states is D(w) = N6(w - ~. where the delta function is
ocntered at Wo. 111C thermal energy of the system is
Nhw
(33)
 S Piooroom II. Thn-rnat Pro)H'rlia 125

17.78 /
1
..
;

, 13.33
/
•
~
•."
/
,i• V
" .... /
rI"

o 1/ ...
o ,." ..,
1". k/-itl'
Figure 9 Low t~"mpenturc heat <"ap"",t)" of solid argon . plotted ag..il15 t :r'. In this tcmper.oturc
region thc cxpcrhnental ra;u1ts arc in excel1...nt agreemen l "ilh lhe l>cbye T~bw "illl 8 - 92..0 K.
({""..ourlclY of L Finegokl and N. E . Phil1i&K.)

Figure 10 To obtain a qualil3ti'·c explanation cL the Deb)oe T 3 law. woe sUPIlOtc: that all phonon
mocks of """"1l'"ccIor Ics!. lhan Kr ha,·e theclassiatlthcn..al energy k"T and thai mock'S bel"'·.,.,n Kr
and the: Dcb)·., ""tofT K" are n(l( exc:itcd al all. Ofthc: 3N possiblc modes. tho: rract;on ocitcd Is
or
(K,.J",,), - (Tt"', ~ this is th., "";0 ofthc .. oIume Ihr InllCT spherc 10 thc outcr s(lncre.
The: energy is U " k~T' 3N{TfB),. and the: heat capacity b; Cv " aU/aT" J2Nkrl,Tf8'j'.
 Table 1 Debye temperature and thermal conductivity" ~
e.
- - e
-
U C N 0 F N.
344' 1<40 2230 ~
75
0.85 2.00 0.27 1.29
N. Mg AI $1 P $ CI A,

158
1,41
400
~
-1.56 ~ - ..":r.ow- t~P§.tur~~tof(j, in Kelvin "
Thermal conductivity -at 300 K. in W cm"' KO,
~- 428
2.37
- 645'
1.48
~ .... - 92

K C. $0 TI V C, M, F. Co NI Co Z, G, G. A, $. e, K,

91 230-"- 360-:" 420 - 3llO 630- -416'" 470'.... 445 450- 343 - 327 320 o .
374 282 90 72
1.02 0.16 0.22 0.31 0.94 0,08 0,80 1.00 0.91 4.01 1.16 0,41 0.60 0.50 0 02
Rb $, Y Z, Nb Mo To Ro Rh Pd Ag Cd I, $, • $b T. I X.
56 ;: T4:. <l!O , "291' 2'15 _ ]50 I
~
"606" 480 27. 225 209- lOS 200 211 153
-,
•• 64
0, 58 0.17 0.23 0.54 1.38 0,51 1.17 1.50 0.72 4.29 0.97 0.82 0.67 0,24 0.02
C, e, c. p Hf T, W R. 0, I, PI Ao Hg TI Pb el Po At R,
-~'- 110 ,.., -14;: l,l;; ~~ ,[00 ]~O
0.36 0,14 0.23 0.58 1.74 0.48
500
0,88
420
1.47
240-
0.72
165
3.17
71) .!8. ~ 105
0,46 0.35
11'
0.08
~
-
F, R, Ao .....
p,

-
C. Nd Pm $m Eo Gd Tb 0, Ho E, Tm Yb lo

- -
0.11
Th
0.12
p, U
-
0.16
N, Po
0.13
Am
200

Cm
0.11 0.11
ek
210
0.11
Cf
0.16
E,
~

0,14
Fm
- ~

Md
0.17
120
0.35
No
210
0.1 6
L,

i~
0.54 - _20'1'0-
0.28 0.06
..-
0.07
- -- --
-- ~ ~
~
~ '-.-..... -- .--. -
• ~t o~t of the 6 val ues were ~upp!icd by N, Pearlman. references arc !o:ive n th e A I,P, Hnm//XlOk . 3ro N1. the tb c rlll..J l'O"J lldi\ ,ty V,,!t' l" <l.n: Ir"Il' II. W
Powell a,,<l Y. S. ToulouK'J.n, ~clence 181, 999 (1973),
 6

,
/ -
• -- --'-
,
, ;)
,
, :;
./
,o /
0.1 0.2 0.3 o.~ O~ 06 0.1 0.8 09 10
.',
Figure J I Comparuon of experimental value' or the helll CIl.pacity of diamon d v.,th ""'ues calcu-
bled on th e earlle.t quantum (Einstein) model, ".;ng the ~ristic I•..'mpcrature 8" .. lib/A:...
1320 K. To convert to Jlmol-cleg. multipJy by 4. 186.

with w now written in pl::tee of ~ for convenience.

111C heat C'.lpacity of the oscillators is

C v ""
au) v
(aT = Nkll -:;
(r.w)' (Jt"""e''"'o 1)2 (34)

as plotted in Fig. 11 . This expresses the Einstein (1907) result for the contribu-
tion of N identical oscillators to the heat capacity of a solid.
In three dimensions N is replaced by 3N. The high temperature limit orc v
becomes 3Nk B , which is known as the Dulong and Petit vaJue.
At low temper.ltures (34) decreases as exp(-li«h), whereas the experi.
mental form of the phonon contribution is knov.'1l to be rl as accounted for by
the Debye model treated above. The Einstein mooel is often used to approxi-
mate the optical phonon part of the p honon spectrum.

General Remit for D(w)

We want to find a general expression for D{w), the number of states per
unit frequency range, given the phonon d ispersion relation lrJ/:K). TIle number
of allowed values of K for which the phonon frequency is between wand w +
c:k.J is

D(w) dw ~ (2L)'f d 'K. (35)

." >heLl

where the integral is extended over the volume of the shell in K space bounded
by the two surfaces on which the phonon frequency is constant, one surface on
which the frequency is wand the other on which the frequency is w + dw.
The real problem is to evaluate the volume of this shell. We let dS ... denote
an element of area (Fig. l2) on the surface in K space of the selected constant
'"

Figure 12 Element of area dS .. on • ~Iallt

frequency $"rface In K space. The \'11lume be--
tween two ,"rraces of constant ffeo:luenc)' at QI
and", + dDt Is equal to f dS .. tl""lV,,~.

fre<luency w. The element of volume between the constant frequency surfaces

wand w + dw is 3 right cylinder of base dS... and altitude dK~. so that

I..." d 3K = JdS...dKJ.. • (36)

Here dKJ.. is the perpendiCUl.1T distance (Fig. 13) between the surface w con-
stant and the surfuce w + dw constant. The value of dK.,t will vary from one
point to another on the surface,
The gradient of w, which is V"w, is also nonnal to the surface w constant,
and the quantity

IV"wI dKJ. = dw •

is the difference in frequ ency between the two surfaces connected by dKJ..
Thus the element of the volume is
dw dw
dS... dK J.. ::: dS... IV ..w! = dS... vol:

where VIC "" IV"wl is the magnitude orlhe group velocity of a phonon. For (35)
we have

D(w) dw ~ (~)'
211'
f dS. dw .
v.
We divide both sides by dw and write V = L3 for the volume of the crystal: the
result for the density of st:'ltcs is

D(w) ~--
v
(21T)l
f dS.
- -
VII
(37)
 5 phQnCfU ll. Thn-""" I'roperlin 129

Figure 13 The Iluant;ty dK J is the per]lendirular disilltice be-

1......'Cn Iwo (:0115l:01nl frCllu<o:ncy nufaces in K space, une at fre-
quency '" and the other al freo:luency '" + d ... ,

D{w D{w)

(.) (h)
rLgure 14 Denlily of stales II~ II functiw. of fi-eo:tucn(:)' tOr (al the ~'e wlid and (b) .n actual
~'S11i1 strudure. 'The 5PL'Clrum lOr IOc<:r)'SlIIl starts as w' for small "'. bul discontinuities d.,veioplII
singular poinb.

The integral is taken over the area of the surface w constant, in K space. The
result refers to a single branch of the di~-persion relation. We can use this result
also in electron band theory.
111ere is II special interest in the contribution to D(w) from points at which
the group velocity is zero. Such critica1 points produce singularities (known as
Van Hove singularities) in the distribution function (Fig. 14).
\
ANHARMONIC CRYSTAL 1l'\7ERAcrIONS

The theory of lattice vibrations discussed thus far has been limited in the
potential energy to terms Quadratic in the interatomic displacements. This is
the harmonic them)'; among its consequcJ)(:es are:
• Two IaUice waves do not interact; a single Wolve does not decay or change
fonn with time.
.. There is no thermal expansion .
.. Adiabatic and isothennal elastic constants arc e<lual .
• The elastic constants are indl.--pendenl of pressure and temperature.
• The heat capacity becomes constant at high temperatures T> 6.
I,.

In real crystals nOne of these oonsc{lucnccs is satisfied accurately. lbe devia-

tions may be attributed to the neglect of anharmonic (higher than quadratic)
terms in the interatomic displacements. We discuss some of the simpler aspects
of anharmonic effects.
Beautiful demOnstrations of anharmonic effccts are the experiments on the
interaction of two phonons to produce a third phonon at a frequency «l;; =
WI + Wz. Shiren described an experiment in which a beam of longitudinal
pbonolls of frequency 9.20 Gill. interacts in all MgO crystal with a parallel
beam of Jongitudinal phonons a19. 18 C ll z. '111e interaction oCthe two beams
produced a third beam of longitudinal phonons at 9.20 + 9. 18 = 18.38 CHz.
Three-phonon processes are caused by third-order terms in the lattice
po~ntial energy. A typical term might be Ul ::; Aencw€=. wherc the e's are
stram components and A is a constant. The A's have the same dimensions as
elastic stiffness COnstants but may havc values perhaps an order of magnitude
larger. The physics of the phonon interaction call be stated simply; the pres-
ence of one phonon ClUses a periodic clastic strai n which (through the an-
harmonic interaction) modulates ill space and timc the clastic constant of the
crystal. A second phonon perceives the modulation of the clastic constant and
thereupon is scattered to produce a third phonon, just as from a moving three-
dimensional grating.

Thermal E:tpansion
We may understand thermal expansion by considering for a classical oscil-
lator the effect of anharmonic terms in the potential energy On the mean separn-
lion of a p.1ir of atoms at a temperature T. \\'e take the potential energy of the
atoms at a displacement x from their equilibrium separ"dtiOn at absolute zero as

U(x) = 0:
2
- gx 3 - Ix· . (38)
with c. g. and I all pOSitive. lne term in x3 represents tbe asymmetry of the
mutual repulsion of the atoms aud the term in %4 represents tbe softening of tbe
vibration at large amplitudes. 111e minimum at x = 0 is not an absolute mini-
mum, but for small oscillations the fonn is an ade<lU3te representation of an
interatomic potential.
We calculate the average displacement by using the Boltzmann distribu-
tion function. which weights the possible values of % according to their thermo-
d),namic probability;

r- dx:r CKP{ - PU(x)]

.:.-:;;-;.-----
(,) ~

L: dx CKp( - PU(:r))
\
 , Phonom fl. The,mall'roperl;u
'"
'"
Triple pom! 1 1.62

~
5.42 ./ '"' -,
",,, .I
./ ~
'0< - "
-~
j / 1.70
~
a
"" // '"
/'
,-"
Figu", IS Lattice oo,..la,,1 of ~olid argon as a fUllc_
tion of temperature.
" .,
T~mper.'lure. K
"'

with f3 '" lIk nT. For displacements suclJ that the anharmonic terms in the en-
ergyare small in comparison with k'JT. we may expand the integrands as
f dxx exp(-{3U) a J dx [exp(-f3cx2)J(x + f3gx4 + f3f0~ :::: (3~14)(glc=-,p~312 ;
f d, e,p(-pU) ~ f d, e"P(-Jlcr~ ~ (-m(Jc),n (39)
whence the thermal expansion is

(40)

in the classical region. Note that in (39) we have left cx 2 in the exponential. hut
we have expanded exp(I¥.1X3 + pfr4) E!! 1 + /3gxl + 13ft" + .. ",
Measurements of the lattice constant of solid argon are shown in Fig. 15.
The slope of the curve is proportional to the thermal expansion coefficient. The
expansion coefficient vanishes a5 T - 0, as we expect from Problem 5. In low-
est order the thermal expansion does not involve the symmetric term fx4 in
U(x), but only the anlisymmetric term gx3,

THERI\JAL CONDVCfIVllY

The thermal conductivity coefficient K of a solid is defined with respect to

the steady-state flow of heat down a long rod with a tempel".lture gradient
,",<lx,
dT
iu=-K -d, (41)

where iu is the nux of thermal energy, or the energy transmitted across unit
area per unit time ,
132
(

This form implies that the process of lhennal energy transfer is a random
process. The energy does not simply enter one cnd of the specimen and pro-
ceed directly in a straight path to the other end, but diffuses through the speci-
men, suffering frequent collisions. If the energy were propagated directly
through the specimen without deflection, then the expression for the thermal
flux would not depend on the temperature gradient, but only on the difference
in temperature llT between the ends orthe specimen, regardless of the length
of the specimen. The random nature df the conductivity process brings the
temperature gradient and. as we shall sec,..a mean free path into the expression
for the thermal flux .
From the kinetic theory of gases we find below in a certain approximation
the following expression for the thennal conductivity:
(42)
where C is the heat capacity per unit volume, v is the ave rage particle velocity,
e
and is the mean free path of a particle between collisions. This result was
applied fmt by Debye to describe thermal conductivity in dielectric solids,
with C as the heat capacity of the phollons, v the phonon velocity, and the e
phonon mean free path. Several representative values of the mean free [>.1th are
given in Table 2.
We give the elementary kinetic theory which leads to (42). The flux of
particles in the :t direction is i!ll(lv~), where n is the concentr.ttion of molecules;
in eqUilibrium there is a flux of equal magnitude in the opposite direction. The
(... ) denote average value.
If c is the heat capacity of a particle, then in moving from a region at lo-
cal temperature T + t:J.T to a region at local temperature T a particle will
give up energy c t:J.T. Now t:J.T betv.'Cen the ends of a free [>.1th of the particle is
given by

where T is the average time between collisions.

The net flux of energy (from both senses of the particle flux) is therefore
dT <If
Ju = - n(v~ J; = -i'I(v2) CT J; (43)

If, as for phonons. v is constant, v..'C may write (43) as

<If
iu = -icve dx ' (44)

with t III! VT and C _ nco Thus K = !Cvt.

 S l'''onono II. Thumol l'~rtia 133

Tablc 2 Phonon mClln rree paths

[Calculated from (44), taking u = 5 x l ~ cmlsecas a representative sound ,·clocity. The
t"s obtained in tllis wlIy rdcr to umklapp pl""OCCsses. J

....·K~.·;·,·\Y
..om
---~·d·"'
·-··~....t·.·,·,~
A
- ,.
CC~"'
;.;:;I.-..,I!.._.T~.•·~
.
C_....;C
;.';;'~J~cm- adeg- I
Quartz" 0
- 190
2.00
0.55
0 . 13
0.50 '"
540
Nael o 1.88 0.07 23

- '00
- 190 1.00 0. 27

'Parallel to optic axis.

Therm al Resistivity of Phonon Cas

The phonon mean free path t is determincd principally by two processes.

gl.'Omctrical scattering and scattering by other phonons. If the forces between
atoms were purely harmonic. there would be no mechanism for collisions be-
twcen differcn t phonons, and the mean free path would be limited solely by
collisions of a phonon with the crystal boundary, and by lattice imperfections.
l nere are situations where these cffects arc dominant.
With anharmonic lattice intcractions, therc is a coupling between different
phonoru which limits the value of the mean free path. 'Ine exact states of the
anharmonic system arc no longer like pure phonons.
TIle theo l)' of the effect of anharmonic coupling on thermal reSistivity
predicts that t is proportional to liT at high temperatures. 2 in agreement with
many experiments. \ Ve can understand this dcpendence in tcrms of the num-
berof phonons with which a given phonon can interolct: at high temperature the
total number of excited phonon..s is proportional to T. The collision frequency of
a given phonon should be proportional to the number of phonons with which it
can collide, whence t cc liT.
To define a thermal conductivity the re must exist mechanisms in the crys-
tal whereby the distribution of phonons may be brought locally into thermal
equilibrium . Without such mechanisms we may not speak of the phonons at
one end of the crystal as be ing ill thermal equilibrium at a te mperolture T2 and
those at the other end in equilibrium at TI .
lt is not sufficient to have only a way or limiting the mean free path, but
there must also be a Woly of establishing a local thermal eqUilibrium distribution
of phonoru. Phonon collisions with a stat ic imperfection or a crystal boundary

' Sec J. M. Zim... u, Electrons lind pllOlKm8 , OdiHlI. 1960, I\. 6errnllfl. ~ He&l conductivi ty of
llOu·m"tal1ic C.T)'stals, ~ Conlemp. Ph)")' 14. 101 (1973).
"..

,..
..
•

Figure 1611 Flow <X gas molerules in II state of drifHng equIlibrIum down 1\ long open lube with
frictionless walls. Elasticco1iision proccnes a mong the gas molecules do not change the rTlQrncntum
or e nergy nUl( of the gas because in ~h rollision tl,e velocity dlhe center of mass ofthc colliding
particles and their "neq:y remain unchanged . Thull energy is tr.ulsported from len
to right wllhoul
being drivCfl by II temperature gradient. n~rdore the thnmal Tesistivity is zero and t1~ thcnnal
conduct;,·;!), is infinite.

rogun 16b The usual definition of thermal rondUCI;.ity lO .gas rcfen to 1\ situ.lllioo """'"" 00 ma$lI
now is permitted. Hen: the tube is closed at both (:neis, prc:vcniinc the ~ or enl rallC't: of
molccu\c1;. With" temperature gradient the collidin$( pain with abO\-e-avcrngc ttfller of masIi
"docitin willtcnd to be dirccted to the right, those w,th bdow-""cragc vcloctties "iHlend to be
directed to the left. A ~Iight ('Ort(.'Cntration gradient. high on the right, will be set "I) to enable tbe
net man InmlpOrt to be ~ro "hile allowing I net energy tnlnSport from the hot to the coLd end.

will not by themselvcs cstablish thermal equilibrium, bC(.'iluse such (.'Ollisions

do 110t change the encrgy or individual phonons: thc frequency W2 oCthe scat-
teroo phonon is equal to the rrcqucm:y WI of the incident phonon .
It is rather remarkable also that a three-phonon collision process
(45)
will not establish equilibrium. but for a subtle reason: the total momentum of
the phonon gas is not changed by such a collision. An equilibrium distribution
of phonons at a temperature T can move down the crystal with a drift velocity
whieh is not disturbed by three-phonon collisions of the form (45). For sueh
(."Ollisions the phonon momentum I
(46)

is conserved, because on collision the change in J is ~ - K:z - KJ = O. Here

nl( is the number of phonons having W3vcvcctor K.
 S f'hctwn. II. TMorwl f'rv,krliu U:

-
"-

Figure 1& In a crystal we mByammge to CK'''C pholJOfl5 chiclly at one end, as by ;Ruminating the
left rod with a lamp. From that Cf.o the.e \"n be. net Ou. or phonon. tuward the rigt,t coo .,rtlle
crystal , If on ly N processes (K, + K. '" K,,) occu r, the phonon fluJC Is unchanged in JJl()mentum 00
eollision and some phonon fl."" w,n pen;S! down the length nfthc crystal. On arrival c:f phonons ..
the . ight end we can arrange In principle to con\crt mOlt of their energy to ,.,.(I;II\;on, thereby
creating a sink ror the phoOOl\l. Just as in (a) the then nal reo:istivity is 7.cro.

Figure 16d In U proc...,sscs there is a large- net ehllnge in phnnon JJl()lIlCntum in caeh rollision
cvalt. An initial net phon<lfl flux wtll npidly d~y as we move 10 the right . The ends may act as
KllIrttS and sinh. NL1 energy t. lU1Spor! uoder • temperature graciicnt occurs as In (b).

For a distribution with J >* 0, colli sions such as (45) arc incap.."lble of estab-
lishing complete thenna! equilibrium because they leave J nnchangoo. If we
start a distribution of hot phonons down a rod with J ~ 0, the distribution will
propagate down the rod with J unchanged. Thcrcrorc there is no thermal resis-
tance. The problem as illustrated in Fig. 16 is like that of thc (:ollisions bchvccn
moleculcs of a gas in a straight tube with frictionless walls.

UrnkIapp Prousses
'rnc important three·phonon processes that causc thermal resistivity are
not of the fonn K t + ~ = K;, in which K is <:OllSetvcU, but are of the form
(47)
where G is a rcciprocal lattice vcctor (Fig. 11). These prOC"CS5eS, discovered by
Peierls, are called umklapp processes. We recall that G may occur in all mo-
mentum conscrvation laws in crystals. In all allOY.·cd proccsscs of the form of
(46) and (47), encrgy is consctvcd.
We have secn cxamplcs of wave intcraction processes in crystals for which
the total W3\·C\·CCt ...... ehange need not be zero, but may be a reciprocal lattice
l
'36

.,
F1gurt: 17 (al Normal 1\, + K., - ~ and (hl umkJapp 11:, + 1\2 - ~ + G phonon coIli~ion pro-
tnc
(It'SK'S In a two-dimensional MjuaJl:: Ialtice. The scll....... in e-.dl togure K'pO'senls Jim Brillouin
zone in the phonon K space. th is woe COI1tairu; all the possible independent ""lues ofthc ph(ll"lOll
"'a\lC''edor. \'ccl()l"!l K with arrowheads at the ceI1tcr 0( the r.onc represent phonons absorbed in
the C'Olli~ pI'Ol'n:5, th05C with ~ _way rrom the center oflhc zone reprncfll phonons
emlllcd in the collision. We sco:' in (b) that in the umkJ.w pn::oct5Ilhe direction of the r-componenl
ofll,e phooon flux ha. bren revc:-rscd. The rociprocallatlice \'t'C.tor C B shown is ofJeflgth 2m'o ,
whereo Is the lattice constant ofthccrystallauice. and is p.nDclto the K,axis. For all processes, N
or U, energy must be ro...en"Cd, so thai w, + "" - W;,

vector. Such processes are always possible in periodic lattices. The argument is
particularly strong for phonons: the only meaningful phonon K's lie in the first
Brillouin zone, so that ally longer K produced in a rollision must be brought
back into the first zone by addition of a C . A collision of two phonons both with
a negative value of K" <:alL by an umklapp process (C " 0) create a phonon with
positive K, . Umklapp processes arc also called U prtX'Csscs.
Collisions in which C = 0 arc called normal pnx.'esses or N processes. At
high temperJ.tures T > fJ all phOllOll modes arc excited because kilT> liw,n;a.

,
A substantial proportion of all phonon (;ollisions will then be U processes, with
the attendant high momentum erumge in the collision. In this regime we can
estimate the thermal resisti\·ity without p..1rticular distinction between Nand U
pn)(.'csses; by the earlier argumcnt ahout nonlinear effects we cxpect to find a
lattice thermal resistivity IX T at high temperatures.
The energy of phonons KJ, K2 suitable lOr umklapp to occur is of the order
of ikBfJ, because each of the phOlLOOS 1 mLd 2 must have wavcvcuors of the
order of iG io order fOf" the collision (47) to be possible. If both phollons ha\·e
low K, and therefOre low energy, there is 00 \\.'ay to gel from their collision a
phonon of wavevcetor outside the first wILe. 'me um)(tapp process must con-
serve energy, just as for the normal process. At low temperatures the number
of suitable phonons of the high energy ik/!6 rCfJuired may be expected to vary
roughly as exp(-612T). according to the Boltzmann rador. 111c exponential
fonn is in good agreement with experiment. IlL summary, the phOllon mean
 :; Phot,o", II. TMnnai Proptrlin 131

i 100
." 3 50
,

,
Fig\IrE 18 n.crmal ronductlvity of a highly
purir.cd crystal of SQd; um fluoriodc , after
II. E. Jackson. C. T. Walkc.-, and T. F. , " " Xl
McNdly.

free path which enters (42) is the mean free path for umklapp c.'O lIisiolls belwccn
phooolls and not for all collisions behveen phonons.

Impe''/ecfio,1S
Ceometrical effccts may also be important in limiting the mean free path.
We must consider scattering by crystal boundaries, the distributioll of isotopiC
masses in natural chemical clements, chemical impurities, lattice imperfec-
tions, and amorphous structures.
When at low tem peratures the mean free path f becomes comparable with
the width of the test specimen, the value of f is limited by the width, and the
thennal conuuctivity bcc.'Omes a function of the dimensions of the specimen.
This effect was discovered by de Haas and Bierma5z. The abrupt decrease in
thermal conductivity of pure crystals at low temperatures is caused by the size
cITed .
At low temperatures the umklapp pTOCCSS hecomes ineffec.'tive in limiting
the thcrmal conductivity, and the size effect becomes dominant, as shown in
Fig. 18. One would expect then that the phonon mean frcc path would be
constant and of the order of the diameter D of the specime n. 50 that
K -CvD. (48)
The only tempcraturc-dependent term on the right is C, the heat capac.ity,
which varics as r 3 at low temperatures. We expect the thennal c.'Onductivity to
vary as r 3 at low temperatures. The size effect enters whenevcr the phollon
mean free path becomes comparable with the diameter of thc specimen.

7
 l
136

00

~
f--
K - O.06T' / ~ ~n nd\t<..J c..-
7

'"
/J 1"-\
10 If.
, r/ Normal c;., ""'" j\
, 1/
I
I'\;
1
if ~
rJ I'"
,
o,
1
, , In 20
r,,·,nlK'r.l!lI ,e, K
.
50 100 200 500

figure 19 l""lope dk,<,1 o n thermal conduct ion in gcnmu,ium, amounting 10 a [aclor of three al
the rond ucli\'ily maximum. The enriched specimen is OOIlCR'Cnt Gc" , ""turd' germanium is 20
perrenl {;e7ll, 'l:1 percent Cc 70 , 8 percen t Gen. ~ percenl Gc", ,md 8 pe rcenl Gc7tl • Below 5 K
the cndchcd specimen has K - 0.060 Tl, which agR'CS well willl Casimir's th eory for thermal
r"s ~tliJ1"'" """oed by boundary scattering. (After T. H. Cerolle and C . W lIull.)

Dielectric crystals may have thennal (.'Onductivitics as high as metals. Syn-

thetic sapphire (AlzO;J has one of the highest vaJues of the conductivity; nearly
200 W em- I K- 1 at 30 K. The maximum of the thermal conductivity in sap-
phire is greater than the maximum of 100 W em- I K- 1 in copper. Metallic
gallium, howcvcr, has a conductivity of 845 \V cm - l K- l at 1.8 K. Thc elcc-
tronic (.t mtl'ibution to thc thcrmal (.'Onductivity of mctals is treatcd in
Chapter 6.
In an otherwisc pcrfect crystal, the distribution of isotopcs of thc chcmical
clcmcnts orten prmidcs an imf.K>rtant mechanism for phonon scattcring. The
random d istribution of isotopic mass d isturbs thc periodicity of thc dcnsity as
seen by an clastic wave. In some substanccs scattcring of p honons by isotopes is
comparable in importruK'C to scattering by other phonons. Rcsults for gcnna-
nium are shown in Fig. 19. Enhan(.'Cd thennal (.'Onductivity has been obscrved
also in isotopically purc silia)fl and diamond; the lattcr has dcvice importance
as a heat sink for laser SOUR'CS.

(
 S Phonrml II. Thrn",J PfVlWrtW. 139

Problems
I . Sillg"larily ill density of dold. (a) From tI,C dispcrsion relation deriv(.-u in Chapter
4 for a monatomic linear lattice ofN atoms with nearest ncighboc interactions, show
that the density of nlOdes is

D(w) - -
.v . 2 . .2\lfl
"1f (wt~ W /

wl,ere w", is the ma;o: imum frequency. (b) Suppose that an optical phonon hrancll has
the kn-nl w(K) - ~ - AK~ , ncar K - 0 in thrcc dimcns ions. Show tl I.1t 0 (,,",) =
(U2"1ft'(2-mA:WX~ - ,,",)'12 for ""' < ~ and D(,,",) "" 0 for w > ~. Ile rc the density of
modes is di.scontinuous.

2. Hrn.f thermal dilation 0/ cnJ$tol cell. (a) Es ti m.ate for 300 K tllC root mean square
thcrmal dilation .6.VI V·far a primitive cell of $Odium. Take the bulk modulus as 7 x
10 '0 erg an - 3. Note that the Ocbre temperature 158 K Is less than 300 K. so that the
thermal energy is of the vnl.CI" of kilT. (b) Use this rCiul1 to estimate the root mean
square thcrnlal fluctuation c.a1a of the lattice parameter .

3. 7Aro poin t lattice (li311i.acem,," t and . trai". (a) In the Oebye approximation, show
that the mean square displacement of an atom at ah salute wro is (R~ =
3Ii~-n'pv3, where I) is the velocity of sound. Start from the result (4.29) summcd

O\'e r the independent la tt ice modes: (n~ - (tll2pV )}:w - I. We have included a factor
of i to go from mcan squllre am plit udc to mean S(juare (lisplllCt'men t . (h) Show that
};w- I amI (R~ diverge for a one-dimcnsionallattiee, bu t tl,at the mean square strain
is fl nite. Consider «(aruaxf) = IIKtllS as the mean square 5train, and show that it is
L-qual to IIWbU4A1No 3 for a line of N atoms each of mass AI, counting longitudinal
modes only, The divCI"gencc of nt is not Significant for any ph)1'icaJ measuremen t.

4. Ileat copacjly o/layer lattice. (a) Conside r a d ielectric crystal made UI) of layers of
atoms, with rigid coupli ng bctv."CCn laycrs so that the motion of the atoms is re-
stricted to the plane of the Ia)'er. Show that tile phonon hmt capacity in the Oc hre
approximation in the low temperature limit is proportional to T i , (b) Suppose ill-
~tcad. :u in many laye r structu res, that adjacen t layers are very weakly bound to each
0111(.'1", What form would )"011 expect Ihe pllOllon heat capacity to approach at ex-
trL'rncly low temperatures?

-5. Gnmdsen constan'. (a) ShowtlUlt the free cnergyof a phonon modeclfrequcncy w
is ksT In 12 sinh (1Iw12kIlT )j. It is nl'C'Cssary to retai n the zero-point energy illw to
obtain this result. (b) If C. is the fractional volume c1Jange, then ti le fi-ce ene rgy of the
crystal may be written as
'"
where B is the hulk modulu.. Assume II,at tI,e volume dependence of W)( is liwlw =
- yd, where y is known as the Griinciscn constant. If'Y is taken as independent orthe
mode K, SIIOW that F is a min imum with respect to l!. wi,en BIJ.. ""
y}:i1iw (.'Otl! (hwl2k lJ T). and show that th is may be written in terms of ti le thermal
energy dens ity as IJ. = yU(1')JB. (e) Show that on the Debyc model y ""
-a In fJliJ In V. Note: Many approximations arc involved in this theory: the result (a)
is val id only if w is independent of tempcrature; ymay Ix: qUite different lOr different
modCli.

References
R A. Cowley, Anharmonic cry.tals,·· Repl•. Prog. Phys. 31 , pt . I, 123- 166 (1968).
M

C. Leibrri<.:d and W. Ludwig. 'lhcor~ of anharmonic effects in crystals, Solid slate phys'"" 12,
M

276-+14 (1961).
T. ruste, ed., Anharmonic lafllce8, RrtldL>ral lraMilfom and me/tlng. NoordholJ, 1964.
R. S. Krishnan, Thermal expcmsion of crysIol$, Plenum, 1980.
A. D. Broce and H. A. CowlC)', Sfrucfurll/ pilau frlln.t/fioIlS. Taylor and Francis, 1981.
M . Toda, Theory of nonlinear lGlticell, Springer , 1981.

THERMAL CONDUCflVITY
J. E. Parrott and A. D. StnclcC$. TI,ernw/ cvndllcfh,-;ty of ~olids. Academic" Pre.s, 1975.
P. C. K!cmellll, '"Thcrm."ll conducti'i ly and lattice vibration mode.:' Solid slate physks 7,1 -98
( 1958); we also E"cvdo. of ,Jiysit;8 14, 198 (1956).
C. Y. Ho, R W. Pov."ell and P. E. Liley, Thermal co"dllctivitv oflhe eumlfm l~; A CO/llprehensioo
review. J. of J'h.~'$. and a,,:m. Rd. Data, \ '01. 3, SUJlPlemcnt I.
H. r. Tre, cd., Thermal comluctivily, Acadcmk Pre". 1969.
j. M . Ziman , Electrons and phoOIOnB. Oxford, 1960, Cllllptc r 8.
H. Sennan, Thenual CO"dllction In solids, Oxford , U176.
C. M. Bhandari and D . M. Howe, ThenlWIl oo"J"clion if, $emlco,ul"cl~, W iley, 1958.
 6
Free Electron Fermi Gas

ENERGY LEVELS IN ONE DIMENSION 144

EFFECT OF TEMPERATURE ON TIlE FERMI·DI.RAC
DISTRIBUTION 146

FREE ELEcrRON GAS IN THREE DIMENSIONS 146

HEAT CAPACITY OF TIm ELECTRON CAS 151

Experimental heat capacity of metals 155
Heavy fermions 156
ELECfRICAL CONDUCllVrry AND OHM'S LAW 1S6
Experimelltal electriClll resistivity of metals 159
Umklapp scattering 162

MOTION IN MAGNETIC FIELDS 163

Hall effect 164
TIIERMAL CONDUcnVlIT OF METAI. S 166
Ratio of thermal to eled:rical conductivity 166
NANosrnucruREs 168
PROBLEMS IG9
L. Khlelic energy of e ltttron gas 169
2. Pressure and bulk modull15 of an electron gas 169
3. Chemical potential in two dimensions 169
4. Fermi gases in aslroph}"sics 169
5. Liquid lIe' 170
6. Frequency dependence of the clectrieal conductivity 170
7. Dynamic magnetoconductivity t('llsor for free electrons 170
8. Cohesh·e energy of (ree eleclron Fermi gas 170
9. Static magnelooondudivity tensor 171
10. Maximum surface rfiistance 111
11. SmaIl metal spheres 171
12:. Density of states-nanometric wire 172
13. QuantiUltion of ronductance 172
REFERENCES 172
 1' ;&11 .11' I Schematic model of a crystaJ. of sodium metal. 1lIc atomic cores are Na· ioll!; they are
Immersed in a rea of c:onductlon clcd:rons. TIle (~o",:ludion electrons nrC deri\'ro from the 3s
,'tdcnce " le<..1rons of the free al oms. Tnc alomic t'Or<;$ contain 10 clcdmm In the c,,"llgurolion
Il"'2.r"2pft. In an aJ ~"ali metal the atomic U>fCli na:upy a rdalj\'dy small plitt (- 15 IX'rccnt) of the
total \-olumc " rthe cr}'SbI., but in a ooblc metal (Cu, AI:. Au) the ~omic rotC>; arC relat ively \uscI'
and rna)' be in contact w ilh each oth«. The <:ammon cry.tal structu,( ' n)(Mn Icmpcr.Jfllre is b&:
{or the alkali metals and fcc {or the noble mcblb.

''' ---
 CHAPTER 6: FREE ELECTRON FERMI CAS

In a theory which Iws gicen n .. ,lt.l like tl,ese, there

mlt3f certainlll be a great deal cf trull,.
H. A. Lorentz

We can understand many physical properties of metals, all(lnot only of the

simple metros, in terms of the free electron model. Acoordillg to this model, the
valence electrons of the constituent atoms become conduction electrons and
move about freely through the m lume of the metal.
Even in metals for which the free electron model .,\lorks best, tile charge
distribution of the (.'o ll(ludion electrons relleds the strong electrostatic potell-
tial of the ion cores. The utility of the free electron model is greatest ror proper-
ties that depend essentially on the kinetic propcrtk'S of the conduction elC(.,-
trollS. The interaction of the conductiolL electrons with the ions of the lattice is
treated in Chapter 7.
The simplest metals are the alkali metals-lithium, sodium, pot ssium,
cesium, and rubidium. In a free atom of sodium the \·alell(.'C ciectroll is in a as
state; in the metal this electron becomes a conduction electron. We speak d the
as conduction band.
A monovalent crystal wh ich contains N atoms will have N (.'onduetion elec-
trons and N positive ion cores. n le Na+ ion core COIltains to electrons that
occupy the 1$, 21, and 2p shells of the frec ion , with a spatial distribution that is
essentially the same whell in the metal as in the frec ion.
The ion cores fill only about 15 percent of the volume of a sodium crystal,
as in Fig. 1. The radius of the free Na+ ion is 0.98 A, whereas one-half of the
ncarest-neighbor distan(.'C of the metal is 1.83 A.
The hlterpretation of metallic properties ill terms of the motion of frcc
electrons was developed long before the invention of quantum mechanics. TIlC
classical theory had se\'eral conspicuous su(.'(.'Csses. notably the deri\'ation of the
fo nn of Ohm's law and the relation between the electrical and thennal conduc-
thity. 'Ine classical theory fails to explain the heat capacity and the magnetic
susceptibility of the condudion electrons. (These arc not fuilures of the frcc
electron model, but fai lures of the Maxwell distribution function.)
There is a further difficulty. From many types of cxperiments it is clear
that a conduction electron in a metal can mo\'e freely in a straight path over
many atomie distances, ulldeflccted by collisions with other conduction elec-
trons or by collisions with the atom cores. In a vcry pure specimen at low
temperatures the mean frec path may be as long as lOS inte ratom ic spaeings
(more than 1 em).
\Vhy is condensed matter so trallsparent to conduction electrons? 111e
answer to the q. ion contains two parts: (a) A conduction electron is not
...
deflected by ion COTes arranged on a lJeriodic lattice because matter waves
prop.1gah: fr'-'ely in a periodic structure. {b} A conduction e lectron is scattered
only infrequently by other conduction electrons. This property is a conse-
quence ofthe Pauli exclusion principle. Bya free electron Fermi gas, we mean
a gas of free electrons subject to the Pauli principle.

ENERGY LEVELS IN 01'\E OU.IENSION

Consider a free electron gas in one dimension , taking account of quantum

theory and of the Pauli principle. An electron of mass m is confined 10 a length
L by infinite b.'lrriers (Fig. 2). TIle wavefunction ",,,(x) of the electron is a solu-
tion of the SchrOdinger equation 1£1/1 = ttl/!; witll the neglect ofpotentiaJ energy
we have 1£ = ,,212m, where p is the momentum. In quantum theory 1) may be
re p."csented by - ill (lIdx, so that

(I )

where E" is the energy of the electron in the orbital .

\Ve use the term orbital to denote a solution of the wave equation for a
system of only one electron. Tbe term allows us to distinguish between an exact
quantum state of the wave equation of a system of N electrons and an approxi-
mate quantum state wbich we construct b)' assigning the N electrons to N
different orbitals, where each orbital is a solution of a wave equation for one
electron. lhe orbital model is t!xact onl), if the re are no interactions between
electrons.
TIle boundar)' conditions are 1/1 .. (0) = 0; I/I,,(L) = 0, as imposed b)' the infi-
nite potential energy barriers. They are satisfied if the wavefunction is sinelike
with an integral number" of half-wa\'elengths between 0 and L:

"'~ = A sin(~7T.r) ; inA" = L , (2)

"
where A is a constant. We see that (2) is a solution of (I), becnuse

d~.
- (nr.)
- -_ A - - "" (nw)
- - r .
2
d 1/1" = _ A(nw)2 sin (~l")
cb: L L ' dr L L
whence the e nergy E" is given by

E .. = !!:....(
2m
nw)'
L
(3)

We want to accommodate N electrons on the line. According to the Pauli

exclusion principle no two electrons can ha\'e all their quantum numbers idcn-
 ~ - -E",,<IY ......,Is
--W"~""'"
~ 2; ....bth.., tede
~ A- 3" " _
Figuri:! Fint three cn<-Tgy !c...,.. and wa~c­
;J:i ,~- -- -=-/--
!unctions of I fm.- e lectron of mass ... confmed
10 • line of length L. n.c f'flcrgy k ..ds Ire
labeled aceord;ng to the 'l"lUltum number n
....hieh g;"cs the number of half.wa,..,j"ngtlu
in U", "'.....,function. lhc "'I.'denglhs arc
indialtrd on tM.- .....-cI'undions. n.e energy
~
"
. - - ---;;;> ' "
~,
i
On of the le,u of'i\lanlllm number n i$ <-'qu91
to (h"I2mXnI2L)".
°o~--------------~,

tical. That is, each orbital can be occupied by tit most one electron. nlis applies
to e1t:ctrons in atoms, molecules, 0 .. solids.
In a linear solid the quantum numbers of a conduction electron orbital tire
nand m., where n is any posilive integer and the magnetic quantum number
m. ::::; ::!:t, according to spin orientation. A pair of orbitals labeled by the quan-
tum number n can accommodate two e lectrons, one wi th spin up and one with
spin down.
If there are sil( electrons, then in the ground state of Ihe system the filled
orbitals are those given in the table;

Electron Electron
, m, oceupancy , m, occlIp:ln<-')'

I 3 I
I j 3 j I
2 I 4 I 0
2 j 4 j 0

More than one orbital may have the S.1l11e energy. The number of orhitals with
the same energy is called the degeneracy.
Let "F denote the topmost filled energy level, where we start filling the
levels from the bottom (n ::::; 1) and continue filling higher levels with electrons
until all N electrons are accommodated. It is convenient to suppose that N is an
e,'cn number. The condition 2nF = N detennines "F. the value of 11 for the
uppennost filled level.
The Fermi energy EF is defined as the energy of the topmost filled level in
the ground slate of the N electron system. J3y (3) with II ::::; IlF we have in one
dimension;

(4)
'"
EI'FECf OF TEMI'ERATURE ON TIfE FERMJ-DInAC DISTRIBUTION

The ground state is the state of the N elcctron syskm at absolute zero.
What hilPPCIlS .IS the temperature is increased? This is a standard problem in
elementary statistical mechanics. and the solution is given by the Fermi-Dirnc
distribution function (Appendix D <lnd TP, Clwpter 7).
The kinetic energy of the electron g,IS increases as the temperature is
increased: some energy levels are occupied which were vacant at absolute zero,
and some levels are vacant which were occupied at absolute 7£ro (Fig. 3). 111c
Fenni-Dirac distribu tion gives the prob..'Ihility that an orbital at energy £ will
be ClCCupied in an ide,11 dcctroll g.'IS in thermal equilibrium:

(5)

TIll! quantity IJ. is a function of the temperature; Ji is to be chosen for the

particular problem ill such a way that the total number of particles in the
system colnes out correctly- that is, equal to N. At absolute I'.ero p. = fEb be-
cause in the limit T-O the fUnction feE) changes discontinuously from the
value I (filled) to the w lue 0 (empty) at fE "" fEF "" Ii. At (IU tcm,wrotllresf(E) is
eqllal to i w/lell E"" Ii. for then the denominator of (5) has the value 2.
nle quantity Ii is the chemical potential (TP, Chapter 5), and we see that
at .lbM>iute zero the chemical potential is equal to the Fermi energy. defined as
the energy of the topmo~t filled ol'hihu at absolute zero.
nle high ellergy tail ohhe distribution is that P.·u1 for which e - Ii J>- kilT;
here the exponential term is dominnnt in the denominator of (5), so thalf(fE) ==
cxpl&-t - e)/kIlT]. This limi t is called the Boltl-mann or Maxwell distribution.

FREE ELECTRON CAS IN THHEE DIMENSIONS

nle free-particle SchrOdinger equation in three dimensions is

h' ( if' if' if' )

- 2m ax2 + ay2 + ail- ""-(r) "" ek ch(r-) . (6)

If the electrons are confined to a cuhe of edge L. the wavefunction is the

standing wave
(7)

where " •• flU> II : are positive integers. 111e origin is at one corner of the cube.
B is convenient til introduce wavefunctions that s'-l.tisfy periodic boundary
conditions, as we did for phonons in Chapter 5. We n( require Ihe wa\'cftmc-
 6 Fr«: EI«/ran "·rrm; en, ,.,.
,

r----:..~ .. ,
,
yo: /0' /(
,., i'.

,, ~4.' 1'-- JQ<

" "'\'\
, ~ t--..
'I-'['~<
,, --- ~ t--
,,
':f ,~/( f'.-
"""
1'--
, , , , - , ,
Figure 3 Fcrmi- Dirx llist n bulioll r"....-tion .t \llrious tcmpo.·raturfl , ror T, .~.II(" - 50,000 K.
111C results apilly to a J.!." illtllre<: dimensions. '111e total number of p'rtic\~') is constant. inllcpm-
dent of ten'llCr..tun: . TIle c}x, miall poIeuHal .. t each tempcntur" may be read oITthc. J:nil'h as the
t.-ncrgy at ",hic:h / = 0.5

tions to be periodic in x. y, z; with perioo L. 11lUS

l/A.,x + L. I). z) = ¢(x, !J, z) • (8)
and similarly for the y and Z ('()Ordinates. Wavefunctions s.llisfying the free-
particle Schrodinger equation and the periodicity condition are of the fonn of a
traveling plane wave:

!J!l,(r) = exp (ik . r) , (9)

provided that the components of d le wavevector k satisfy

2~
k~ = 0 +--
- L (10)

and simila.-ly fo r kl/ and k•.

Any component of k is of the form 2nmL. where n is a positive or negative
Integer. 111e components of k are the quantum numbers of the problem. along
with the quantum number m~ for the spin direction. We confirm that these
values of k, satisfy (8). for
exp{ik,,(x + L)j = exp[i2111'T{x + L)/LJ
...,. CXP(i2n71xIL) exP(iZII1T) = exp(iZlI"m1L) = c\ll(ik.,x) (I I)
'48
I
On subslituting (9) in (6) we have the energy Ek of the orbital with wave-
vector k:

(12)

11le magnitude of the wavevector is related to the wavelength" by k = 2m). .

The linear momentum p may be represented in quantum mechanics by
the operator p "" -i/tV. whence for the orbital (9)
p!/ll.(r) = - ihV~r) = likth(r) • (13)

so that the plane wave tJ!I. is an eigenfunction ofthe linear momen tum with the
eigenvalue Ill,. The particle velocity in the orbital k is given by v := IMm.
In the gl'Ound state of a system of N free electrons the occupied orbitals
may be represented as points inside a sphere in k space. nIt! eot!rgy at the
surface of the sphere is the Femli energy; the wavcvectors at the Fermi surface
have a magnitude kF such that (Fig. 4):
I,'
",..= - - /&. . (14)
2m
From (10) we see that there is one alJov.w wavevector-that is, one dis-
tinct triplet of quantum numbers kr' k~, k~-for the volume element (2rnL';' of
k space. Thus in the sphere of volume 41Tk]:I3 the total number of orbitals is
41Tk'}J3 V
2· (2"'Lf : 3,,' kj. : N (15)

where the factor 2 on the left comes from the two allowed vruucs of III., the spin
quantum number, for eaeh allowed value of k. Then

k" = (3:N ffJ (16)

which depends only on the particle concentration.

Using (14), E =~ (3~N)"'. (17)

" 2m V

TIlis relates thc Fermi energy to the e1cctron concentration NIV. Thc electron
velocity ~F at the Fermi surface is

(18)

Calculated values of k,.., VF. and E,.·arc given in TaLlc 1 for selected metals; also
given are values of the quantity T F which is defined as Eplk s . (The quantity TF
has nothing to do with the temperature of the ek'Ctron gas!)
Flgure.. In th e ground state of a system of N fn'C clec-
trtNl$ the O<'C\lpied orbitals of the system flU a sphere of
,adi'l$ Ie,., when: 1",. .. ~lIl.iI2m is the energy of an ek-ctwll
hiving I w..."'".,ctor Ie"

V(o '

~,

\\
FigureS Density of single-partide stalt.'S as I fundion
of encrgy, fur I rn:~· electron gas in three dimensions.
The dlUh~od Clll'Ve n:prescnt~ the density fj~ .nD(t) of
1lI1t'" orbitAb at a fl "ite temperature. but such that kilT \
\
Is sm,.]1 in l'Dmparison with to' The .h9lkod area repR~
scnb the IlIled orbit als at absolute 7£ro. The a"c.-rage
\ ,
(",)C,'g)' is increa5Cd when the temperature is illa..-a..ctl L-----__...L~'_'''"''~ ___ _
from 0 to T, fur dectrons are thenna]ly excited from
region 1 to rl"gion 2. "
E"", ~, ._

We now find an el{pression for the number of orbitals per unit energy
mnge, D(E), c.111ed the density of states. I We use (17) for the total number of
orbitals of energy S E:
_ V (2m<)'"
N- -- (19)
2
3.r 1i.
so that the density of states (Fig. 5) is

D(E) dNdE = -2.r

I!! - -
V
- . (2m)"
--
1,2
. EII2 (20)

'Strictly, Djf) i. th e dellsity of one-partie1c stat("s, or dell~ity of orbital'!. Often OIIe 5<.15 the
\'OluHle V " I ill ("xprcs.siolls for D(~).
 ~

Table 1 Calculated free electron Fenni surface parameters for metals at room temperature
(Except for Na. K. Rb. Cs at 5 K and Li at 78 K)
~

Fermi
Electron Radius" Fennl ~'ennl Fenni temperature
concentration. parame ter wa\'evec!or . velocity, energy, T,. (,1k".
in em - 3 In In em , - I
";deney Met;d '. em- I In eV in dcg K
~

1 Li 4.70 X lcP 3.25 1.11xldl 1.29 X loB 4.72 5.48 x 10 4

Na 2.65 3.93 0.92 1.01 3.23 3.75
K 1.'" 4.86 0.75 0.86 2. 12 2.46
Rh US 5.20 0.70 0.81 1.85 2.15
C, 0.91 5.63 0.64 0.75 1.58 1.63
Cu 8.45 2.67 1.36 1.57 7.00 8. 12
Ag 5.85 3.02 1.20 1.39 5.48 6.36
Au 5.00 3.01 1.20 1.39 S.SI 6.39
2 B, 24.2 1.88 1. 93 2.23 14. 14 16.41
Mg 8.60 2.85 1.37 1. 58 7. 13 8. 27
C, 4.60 3.27 1.11 1.28 4.68 5.43
S, 3.56 3.56 1.02 U8 3.95 4.58
B, 3.20 3.69 0.98 1.13 3.65 4.24
~
Z. 13.10 2.3 1 1.51 1.82 9.39 10.90
Cd 9.28 2.59 1.'" 1. 62 7.46 8.66
3 Al 18.06 2.07 1.75 2.02 11.63 13.49
G, lS.3O 2.19 1.65 1.91 10.35 12.01
In 11 .49 2.41 1.50 1.74 8.60 9.98
4 Ph 13.20 2.30 1.57 1.82 9.37 10.87
Sn{w) 14.48 '.23 1.62 1.88 10.03 11.64

-
'The dlmendon\eu r..J.lu. parameter iJ deRned ... ~, - rd~H' when: OH Is the Rrst Bohr ndlu. and~" II the ndlu. or a sphere that oontalns one deet""n.
 6 Frte .Ekctroo F""", CIlI 151

l1lis result may I~ obtained and expressed more simply by writing (19) as
3 dN 3 dE
ILl N::. "'2 In E + const;mt N="2'--;-
whence
dN 3N
D(E) a - = - . (21)
dE 2E
Within a factor of the order of unity. the number of orbitals pel' unit energy
range at the Fenni energy is the total number of conduction electrons divided
by the Fenni energy. just as we would expect.

H EAT CAPAClTI' 01" THE ELECTRON GAS

TIle question that caused the greatest difficulty in the early development
or the electron theory of metals concerns the heat cap.'lcity of the conduction
electrons. Classical statistical mechanics predicts that a rree particle should
have a bcat capacity of jkll • where ka is the BoltJ:mann constant. Ir N atoms
each give one valence electron to the electron gas. and the e lectrons are rreely
mobile, then the electronic contribution to the heat capacity should be jNka •
just as for the atoms of a monatomic gas. But the observed e lectron ic contribu-
tion at room temperature is usually less than 0.01 or this value.
111is discrepancy distracted the early v.urkers. such as Lorentz: how can
the electrons p..1rticip..1te in electrical conduction processes as ir they were mo-
bile. while not contributing to the heat capacity? The question was answered
only upon the discovcryorthe Pauli exclusion principle and the Fermi distribu-
tion runction . Fenni found the l.'Orrect equation. and he wrote, "One ret.'Og-
nizes that the specific heat vanishes at absolute zero and that at low tempera-
tures it is proportional to the absolute temperature,"
\Vhen we beat the specimen from absolute zero not every e lecllun gains
an energy - kaT as expected clasSically. but only those electrons in orbitals
within an energy range kaT or the Fenni level are excited thennally; these
electrons gain an energy which is itselr or the order of kilT, as in Fig. 5. This
gives an immediate qualitative solution to the problem or the heat capacity or
the conduction electron gas. Ir N is tile total number or electrons. only a fraction
of the order or TIT I' can be excited thennally at temperature T. because only
these lie within an energy range of the order of kilT of the top or the energy
distribution.
Each or these NTITF electrons has a thennal energy or the order of kaT.
n.e total electronic thennal kinetic energy U is of the order or
(22)
'" (
The ek"Ctronic heat cap.1cit)' is given by
Cd = aUlaT - NkJ.TITF) (23)

and is diredly proportional to T. in agreement with the experimental results

discussed in the following section. At room temperature Cd is smaller than the
classical value fNkR by a fUctor ofthe order orO.Ol or less, for T F - 5 X 104 K.
We now derive a quantitative expression for the dectronic heat capacity
valid at low temperatures k8T ~ Ef ·• TIle increase AU "" U{r) - V(O) in the
total energy (Fig. 5) of a system of N electrons when heated from 0 to T is

liV "" L'" elE ED(E) feE) - f' dE ED(€) . (24)

Here feE) is the Fermi-Dirac function and D(E) is the number of orb{tals pel"
unit energy range. We multiply the identity I

N=
L" i"
elE D(E) fiE) = () tiE D(E) (25)

by E" to obtain

(26)

We USe (26) to rewrite (24) tiS

l::.U = J'" dE(E - E,,)f(E)D(E} + L

e
,. clE(EF - E)[1 - !(E)]D(E) (27)
"
TIle first intt:gral on the righi-hand sidt: of(27) gi~'es the t:nergy needed to take
electrons from tE~· to the orbitals of energy tE > tEp, and the second intt:gral gives
the energy needed to bring the electrons to tE~· fl'Om orbitals below tE~·. Both
contributions to the energy are positive.
111t: product f(tE)D(tE}tl€ in tht: Ilrst integral is the number of electrons
elevated to orbitals in the energy range dtE at an ent:rgy tE. The factor (1 - fetE)]
in the second integral is the probability that an electron has been removed from
an orbital tEo The function l1U is plotted in Fig. 6. In Fig. 3 we ploUed the
Felmi-])irac distribution function versus tE fOr six values of the temperature.
The electron concentration of the Fermi gas ....-as taken such that E~'/klJ =
50,000 K, characteristic of the conduction electrons in a metal.
TIle heat capacity of the electron gas is found on differentiating l1U with
respect to T. The only temperature-dependent term in (27) is feE), whence we
can group terms to obtain

C~l :: (11U = L m

dt{E - Ep) d D(E) .r (28)

,T 0 ( tr
AI the temperatures of interest in metals TIEl' < 0.01, and we see from
Fig. 3 that (E - tE~.) dPdThas large positive peaks at energies near E~" It is a good
 6 F,.,., Eleclron Fermi Cw .50

L ,
/
is ' ,
/
•
v
Figu~e 6 Tempe",!urc dependence of the
energy of a noninlcracting fermion gas in
three dimensions_ The energy is plotted in
,, ./'
/
normali~.cd form as !J.UINEF. where N is the
-'lUmber of c1.,ctroll'i. The temperature is
,., 0.4 0.6
" ..,
plot l<.-d as k~TIE,-.

,
r IWgion of d.'g<-""'r"'~ quant um go>

I~ :.
' I ' ~.
,
-- r---.. Jogiofi of <-ta..L.. 1[01> _

. ~
\ !.'' ~
,I' '.• ~
'·V ·
,,I" · , ,
kMT -
Figure 7 .'lot of the chcmitil poknlial Ii ,'crsll< te mperature Ir."T for a gas of nonintcracting
fermion'S in three dimensions . For ronvcnic ncc in " lulling, the units of Ii and k"T arc (). 763E~"

approximation to evaluate the density of states D(E) at €~- and take it outside of
the integral:

(29)

Examination of the graphs in :Figs. 7 and 8 of Ihe variation of p, with T

suggests that when k8T <C fEr we ignore the tempem.ture dependence of the
chemical potential J1. in the Fermi-Dirac distribution function and replace Jl. by

..
..
,
I ",

"

.~,'----------c.l,c----------,",~
,
- -
Figure 8 \'arla.lion with Icmpel'llurc of the chemical p:>l cnlial /L. for free elcct ron Fcnni g;ues;n
ooc .nd three dimensions. In common mcl.al~ ",.~ - 0.01 . 1 room Ic mperalure , so that /L is closely
equal to ~F. These cun-es were cakubtcd from oeries "~pAnSions of the integral for the number of
particles in the 1)"Slem.

the constant £1'_ We have then. with T B k8T,

(30)

We set
(3 1)
and it follows from (29) and (30) that

en: kiT D (€r) f"

-~,h
til: r
e'
(~+ 1)' (32)

We may safely replace the lower limit by -00 because the factor fI? in the
integnilld is already negligible at x =. - E~T if we arc concemed with low tem-
pemtures such that EFh - 100 or more. 'nlC integnlfZ beromes

f"
_" dxi' (e'+ I)'
e'
-3 '
"..
(33)

tn.e integral is not elementAry. but may be eI.'alualtd from the m(lre &mili:1r result

L- '"-'
-1-~
r+ 12a t •

on d,fferentiation of both sides wilh respect 10 the pammeter /I

\
 .'r---------------------------------,

-.. .-..--- .
err - 2.ll'I + 2..57 P ~ _ _. - -
~\lm

-.~
_______. 4....~
.-.-.-.
·'~,--------------'.~I--------------".~,--------------'~,'3

Figure 9 F..xperimental heal capacity values for potassium, plotted as err versus 'f'l. (After W. H.
Lien and N. E. Phillips.)

whence the heat capacity of an electron gas is

Cd = !-rfD(£p)kiT . (34)
From (21) we have
D(£F) = 3NfMF = 3NI2kBT F (35)
fur a free electron gas with k8TF 51 "p. Thus (34) becomes
C.., = i.,(lNksTITF • (36)
Recall that although Tp is called the Fermi temperature, it is not an actual
temperature, hut only a convenient reference notation.

Experimental Heat Capacity of Metab

At temperatures much below both the Ocbye temperature and the Fermi
temperature, the heat capacity of metals may be written as the sum of electron
and phonon contributions: C = yT + AT3 , where 'fand A are constants charac-
teristic of the material. The electronic term is linear in T and is dominant at
sufficiently low temperatures. It is convenient to exhibit the experimental val-
ues of C as a plot of CIT versus T2;
CIT=y+AT 2 , (37)
fOr then the points should lie on a straight line with slope A and inteccept y.
Such a plot for potassium is shown in Fig. 9. Observed valuc.s of y, called the
Sommerfeld parameter, are given in Table 2.
The observed values of the coefficient yare of the expected magnitude,
but often do not agree very closely with the value calculated for free electrons
of mass m by use of (34). It is common practice to express the ratio of the
observed to the free electron values of the electronic heat capacity as a ..atio or
'" (
a thermal effecti"e mass m,1> to the electron mass JIl, where m,1> is defined by
the relation
y(obscrvcd)
(38)
")'(free)
This fOTm arises in a natur'dl way because ~I' is im'crsely proportional to the
mass of the electron, whence;, ex m. Values of the mHo are given i,\ Tablc 2.
The departure from unity iuvolves three scparalt: effects;

• The interaction of the conduction electrons with the periodic potential oflhe
rigid crystal lattice. The effective mass of an electron in this potential is called
the band effective mass and is treated later.
• The inlcmction of the conduction e lectrons with phonons. An electron tends
to polarize or distort the Janice in its neighborhood. so that the moving e1~
tron tries to dnlg nearby ions along, thereby increasing the effective mass of
the electron.
• The inler-dction of the conduction electrons with themselves. A moving elec-
hun causes an inertia] reaction in the surrounding electron gas, thereby in-
creasing the effective mass of the electron.

Heavy Jiermiom. Several metallic compounds have been discovered that

have enormous values, two or three orders of magnitude higher thlm usuul, of
the electrollic heat capacity. The heavy fcnnion compounds include UDe I 3.
CcAI:J, and CeCu2Si2' 11 has been suggested that f electrons in these com-
pounds may have inertial masses as high as 1000 m, because of the weak o\'er-
lap of wavefunctions off electrons on neighboring ions (see Chapter 9, "tight
binding'/. References are given by Z, Fisk. J. L. Smith, and H. R. Ott. Physics
Today, 38, S-2O Uanuary, 1985). The heavy fermion compounds form a class of
superconductors k"11o ...m as "exotic superconductors."

ELECTRICAL CONDUCTIVIlY AND OHM'S LAW

TIle momentum ofa free e lectron is related to the wavevcctor by "I\' = hk.
In an e lectric field E and magnetic field D the force F on an electron of charge
-e is - e[E + (llc)\' x DI. so that Newton's seoond law of motion becomes

(CCS) F:: m~
dl
= h~ :: -e(E + ~\, D) .
dl c
X (39)

In the absence of collisions the Fermi sphere (Fig. lO) in k space is displaced at
a ullifol"ln ratc by a constant applied electric field . We integrate with D = 0 to
obtain
k(l) - k(O) "" - eEIII! . (40)
r

Table 2 E:lperimenlal and free electron values of electronic: heat

LI B. capacity constant 'Y of metals B C N
163
0.749
2.18
0.17
O.SO<>
0.34
(From compil a t ion~
thermll] efTcdive m(lSs
kindly fll rni'hcd by N. Philli!,S <lnd N. Pearlma n. The
j, Jdlned by Eq. (38).
- _.......~
~
~

- - I.'" - -
N. M, AI SI P
1.38 13
1.094 "O':§92
1.26 1.3
Observed'Y in mJ mol-' K- '.
Ca~tited rree~ecrian: i~':m:rmOP"k-'-
m"lm - (observed y)/(Iree electror'l yl.
- . ~
1.35
(f91~ -
~-

C. So TI C, F. Co NI C, Z, G. G. A,

- -- --.- --
K V Mn(y)
2.08
1.668
1,25
2.'
flIT
I'
-
10.7
..-.:."
3.35
~-
..
9.26 1.40 9.20 4.98
...... ~.,.
4.73 7.02
-.......- 0.695 064 0.596
O:S~5' 1i:7S! "IT2S
1.38 0.85 0.58
~- -- 0.1 9

Rb S, Y Z, Nb Mo To R, Rh Pd A, Cd' I, Sn (>ol Sb

- - -- . - -
-
-
2.4 1 3.6 10.2 2SO 7,79 20 3.3 4, 9.42 0.646 0.688 1.69 1.78 0.11
,~,...
-'~.64j_ _Q:.~
~
1.9 11, 1.799. :--.. = ~ ~ I~33 1.4.\..Q.,
1.26 20 1.00 0,73 1.37 1.26

C, B. L. HI T. W R. 0, I, pt A, Hg(a) TI Pb BI •
-- -
5.' .....,., - -- - i
3.20 2.7 10. 2.16 1.3 2.3 2.4 3.1 6.8 0 .729 1.79 1.47 2.98 0.008
ici38~ -1"§Jt ~ ~ ."'"T'>

-
"Ilf-""" -,
6:642""- o~§~ }.29' 'f$69~ -":""""1"
"

f
\.43 1.4 1.1 4 1.88 1.14 1.97

~
f

, ~
'"

......
.. . .. . .
+---+-----+
.. .. .. .. .. .. •. +----t-+-- -+ •.

'"
Figure 10 (al The Fenni sphere t'ndmes the occu pied dcctnm mbitals in Ie space in the ground
~tate of the electron gotS. The ne t momentum is zero. because for C\'CO)' orbital k there is ill!
occu pied orbi tal at - k. (b) Under the influence of a constant ro~ F acting for a time intcn-alt
eH:ry orbital has its k "CdOT ir>erc:w:d by Ilk = FilA. This is equivalent to 11 di~pla~mcnl of the
whole Fenlli sphere by lik. TI,e total moment um is Nh6k. if there arc N clcclrons present. The
application of the force in~'rcascs the e nergy of the ~ystcm by N(/"SkI"IZm_

If the field is applied at time t = 0 to an electron gas that fills the Fermi
sphere centered al the origin ort. space, then at a later time t the sphere....,11 be
displaced to a new center at
oa.
= - eEtlh . (41)
Notice that the Fermi sphere is displaced as a whole.
Because or collisions of electrons with impurities, lattice imperfections.
and phonons. the displaced sphere may be maintained in a steady state in an
electric field. If collision time is T, the displacement of the Fermi sphere in the
stead y state is given by (4 1) with t = T. The incremental velocity is v ::::
-eE-rim. Ifin a constant electric field E there are n electrons of charge q = - e
per unit volume. the electric current density is
(42)
This is Ohm's law.
The electrical conductivity (T is defined by j = aE, so that

(43)

The electrical resistivity p is defined as the reciprocal of the conductivity,

so that
p = mlncT . (44)
 6 Free J-;fednm "er",' Cu. 159

Values of the electrical conductivity and resistivity of the elements are given in
Table 3. It is useful to remember that q in Gaussian units has the dimensions of
frequency.
It is easy to understand the result (43) for the conductivity. We expect the
charge transported to be proportional to the charge density ne; the factor elm
enters bec-duse the acceleration in a given electric field is proportional to e and
inversely proportional to the mass m. The time 7 describes the free time during
which the field acts on the carrier. Closely the same result for the electrical
conductivity is ohtained for a classical (Maxwellian) gas of electrons, as realized
at low carrier concentrdtion in many semiconductor pnlblems. The mathen lut-
ics of this similarity is developed in the section on tr.msport theory in TP,
Chapter 14.
It is possible to obtain crystals of copper so pure that their conductivity at
liquid helium temperatures (4 K) is nearly 10"5 times that at room temperature;
e
for these conditions 7"'" 2 X 10- 9 s at 4 K. The mean free path of a condUG
tion electron is defined as
(45)
where VF is the velocity at the Fenni smface, because all collisions involve only
electrons near the Fermi surface. From Table 1 we have Vt· == L57 x
1«' cm S- I for CUi thus the mean free path is
l(4K)=O.3cm; l(300 K) = 3 x 10- 6 cm .
Mean free paths as long as 10 cm have been observed in velY pure metals in the
liquid helium temperature range. For electron-electron collisions. see Eq.
(10.63).

E:qJerimental Electricm Resistivity of Metals

The electrical resistivity of most metals is dominated at room temperature
(300 K) by collisions of the conduction electrons with lattice phonons and at
liquid helium temperature (4 K) by collisions with impurity atoms and mechan-
ical imperfections in the lattice (Fig. 11). The rdtes of these collisions are often
independent to a good approxiulation, SO that if the electric field were switched
off the momentum distribution .....ould relax back to its ground state with the net
relaxation time
I 1 1
-:-+-, (46)
7 'TL 7,

where TL and 7, are the collision times for scattering by phonons and by imper-
fections. respectively.
The net resistivity is given by
P=PL+fJI (47)
 §
Table :l Electrical conductivity and rcsi~ tivity of metals at 295 K
LI B. B C N 0 F N.

- - -- --..
(Resistivity value~ as given by C. T. Meaden, Electrical resistance of

1.0"-
9.32
3158
3.25 -- -
metals, Plenum, 1965; res idual resistivities have been subtracted. )

- -- ~
-- ~

--
N, Mg AI SI P S CI

2. 11.
4.75
2,3r
430
Cc!ndtICt&i!Y I n+units-of l O~ (o1iii£!!!)1:-:-''''
Resistivit y in uM s 01 10-' ohm·em.
" Ii 3.I~:'
2.74
-- ::::T - ...- b
~

K C, S< TI V C, M, F. Co NI C, 2' G, G. A, S. B, K,
y",
7.19

Rb
n .IL 021 _023 Too
36

S,
46.8

Y
43.1

2,
19.9

Nb
~
12.9

Mo
§.rtJ2 T:~ T~
139.

To
9.8

R,
5.8

Rh
C43:: ~~ TW'
7.0
Pd
1.70

Ag
5.92

Cd I,
- - - -- ---
0.6t::: -~--
14.85

Sn (wJ
"' ~ -

Sb T. I x.
080
12 ,5
0:;;["
21.5
O-:-lL
58.5
0.24
42.4
6-6~~ T~89
14.5 5.3
':"'-0.7
-14.
f.3£
7.4
2·0lC
4B
1l~ ~~J:: T,j'g";
10.5 1.61 727
m~
~
8.75
- - -
~~ S'-:-Z'Ci
11.0 41.3
~ ~

C, B, L, Hf T, W R. 0, I, PI A, Hg 'IQ. TI Pb BI Po AI R,
O- SO
200
0.26-
39
013
79.
0.33
30 6
O~7r:;: -1.89
13.1 5.3
O.Se
!S6
l.l{)
91
1.%
5.1
o.96
~
10.4
4.-5~
2.20
bE
95.9
0.61-..
16.4
-o:~
21.0
o.OSEL
116.
O.n::
46 - --
F, R, Ao I'
C. p, Nd Pm Sm Gd Tb Ho Tm Yb L,

6:12~ >6.(5" 0.1'-

~
0.10
"
O:n- Omp 0090 o:rr- on-
" Oy

0.12... "O":lr" 1)'~ 'If[9''"

81 67. 59. 99. 89. 134. Ill . 90.0 77.7 81 62. 26.4 53.

-- --
Th p, Np p, Am Cm Bk Cf Fm Md No

-"
U L'
O.~
-- 6]~ 0.085 0.070.
-- -- -- --"
- - -- - ~
~

-
15.2 25.7 11 8. 143.
 6 Fru ,,; '«fron Fen..i C..... 161

Figuno II Eic'dricai re.isti,·ity in most metal. ariJe$ from coll isions of ele<.1rons with ir...,gularitics
in the b tt ice, as in (a) by phonon, and in (b) by bnpuritles and \x-ant latt ice sites.

• "

•-ogurc 12 It. .. istance of jlI)t.:usium below

20 K, as measured on two spec:imens by
I). M.cDonaid anc:l K. Me ndelssohn Thedif·
rerellt interrepts at 0 K arc attributed to dif-
ICrent roncentratioos d impwiti<... and statioc
imperl"<.'Ctions in the two specimens.

where PI... is the resistivity caused by the thennal phonons, and p, is the resistiv-
ity C'.lused by scattering of the eloctro n wa,'es by static defects that disturb the
periodicity of the lattice. Often PI... is independent of the number of defects
when their concentration is small, and often p, is independent of temper.lhlre.
This empirical observation expresses Matthiessen's 11..I1e, which is CQn,·cnic nt
in analyzing experimcntal data (Fig. J2).
The residual resistivity, p,(O), is the extrdpolated resistivity at 0 K because
P,. v.lnishes as T - O. The lattice resistivity, pdT) = p - p,{O), is the samc for
differcnt spccimens of a mctal, enm though pl..O) may itself wry widely. file
'"
rC5is th·ity ratiO of a specime n is usually defined as the ratio of its resistivity a t
room tempe rature to its residual resistivity. It is a convenient approximate
indicator of sample purity: for many materials an impurity in solid solution
creates a residual resistivity of about 1 p.ohm-cm (1 x 10- 6 ohm-em) per
atomic percent of impurity. A copper specimen with a resistivity ratio of 1000
will have a residual resisti\·ity of L 7 x 10- 3 .uohm-cm, corresponding to an
impurity concentration of about 20 ppm. In exceptionally pure specimens the
resistivity rJtio may be as high as lOS, whereas in some alloys (e.g., manganin) it
is as low as 1.1.
The temperature-dependent part of the electrical resistivity is propor-
tional to the rate at which an electron collides with thermal phonolls and ther-
mal electrons (Chapter 10). The coUision rate with phonons is proportional to
the concentration of thermal phonons. One simple limit is at temperdtures over
the Debye temperdture (J; here the phonon concentration is proportional to the
te mperdture T , so that pIX T for T> (). A sketch of the theory is given in
Appendix J.

Umklnpp Scattering
Umklapp scatte ring of electrons by phonons (Chapte r 5) accounts for most
of the electriCdl resistivity of me tals at low te mperatures. These are e1ectron-
phonon scattering processes in which a reciprocal lattice \'ector G is im'Olvoo,
so that electron momentum change in the process may be much larger than in
a normal electron-phonon sCdttering process at low te mperatures, (In an
umklapp process the wavevector of one particle may be "flipped o\'er. ")
Consider a section perpend icular to [100] through two adjacent Brillouin
m nes in bee potassium, with the equiva.lent Fermi spheres inscribed within
each (Fig. 13). The lower half of the figure shows the normal e1ectron-phonon
collision k' == k + q, while the upper ha.lf shows a possible scattering process
k' = k + q + G involving the same phonon and terminating outside the first
Brillouin 7.one, at the point A. This point is exactly equivalent to the point A'
inside the original zone, where AA' is a reciprocal lattice vector C.
This scattering is an umklapp process, in analogy to phonons. Such colli-
sions are strong sCdtterers because the scattering angle can be close to 7T, and a
single collision can practically restore the e1ectron to its ground orbital.
When the Fermi surface does not intersect the zone boundary, there is
some minimum phonon \I/a\'evector qo for umklapp scattering. At low enough
temperatures the numbe r of phonons available for umklapp scattering falls as
exp( ~ OdT), where (}v is a chamcteristic temperature calculable from the geom-
etry of the Fermi surface inside the Brillouin zone. For a sphericaJ Fermi
surface with one electron orbital per atom inside the bee Brillouin mne, one
can show oy geome try that qo = 0.267 k t ·.
 6 Frl'e l;;ledron Fern» Cal 163

.'igure 13 Two Fermi ~phcrc. in ailJ:'>ct'nt WIles, a

constroction to show the role of phonon umklapp
processes in e1ec:lricai resi.livity.

The experimental data for potassium have the expected exponential fonn
with 8u = 23 K compared with tbe Debye 8 "" 91 K. At the very lowest tem-
perdtures (below about 2 K in potassium) the number of umklapp processes is
negligible and the lattice resistivity is then caused only by small angle SC'dtter-
ing, which is the normal scattering.
Bloch obtained an analytic result for the nonnal scattering, with Pt, « T 5 /(f'
al vcr)' low temperatures. This is a classic limiting result. These nonnal pro-
loeSseS contribute to the resistivity in all metals, but they have not yet been
clearly isolated for any me tal bee-duse of the large competing effects of imper-
fection scattering, electron-electron S<;dttering, and umklapp scattering.

MOTlON IN MAGNETIC FIELDS

By the argument of (39) and (41) we are led to the equation of motion for
the displacement .skof a Fermi sphere of particles acted on by a force F and by
fril'Iion as represented by collisions:

'(~+~)&=F
dt T
. (48)

The free particle acceleration term is (lidJdt) & and the ·effect of collisions (the
friction) is represented by ".skIT, where T is the collision time.
Consider now the motion oCthe system in a unifonn magnetic field B. The
Lcrrentz force on an electron is

(CGS) F=-e(E +~VXB) (49)

(SI) F= -«E + .;. 8)

If mv "" h8k then the equation of motion is

(CGS) m(1 +~)v =-e(E+ :VXB) (SO)

'64

An important situation is the following; lei a slatic magnetic field B lie

along the % 8.1[is. Then the equations of motion arc

(ees) m(~t + ~)v% = -e(Ex + : vv)

m(.!!.+1.)v
dt T
= -e(E - ~v.)
1/ 1/ C
(51)

rl I) .
m ( dt + -;- v,. -eE,. . =

The results in SI arc obtained by replacing c hy 1.

In the steady state in a static electric field the time derivatives arc zero, so
that the drifi velocity is
n.
Vz = - - E• • (52)
m
where Wc . eBlmc is the cyclotron frequency, as discussed in Chapter 8 for
cyclotron resonance in semiconductors.

llall Effect
The Hall field is the electric field developed across two faces of a conduc-
tor, in the direction j )( n, wben a current j flows across a magnetic field B.
Consider a rod-shaped specimen in a longitudinal electric field E. anda truns-
verse magnetic field, as in Fig. 14. If current cannot nowou! of the rod in the y
direction we must have ~ = O. From (52) this is possible only if there is a
transverse electric field
(ee5) E\I - w'tE
en,
=
c" = - -
me- E~ (53)

(51)

The fJuantity dei'i.,oo by

(54)

is called the Hall coefficient. To evaluate it on our simple model we use

j~ = rW'TE.Jm and obtain
(ee5) RII = _ ce,B"'cE~'J,;":::'c'_ = __1_ (55)
ne'l-rE.Blm ne<
•
(51) nJl = -neI-
 (

I r-ri' I lL:.
-e ~J'
-...
perpo!ndiwlac
tot ub;
I
~. ~J'
drifI: , ...kri:y
ill>! SUrting liP.

-~
. :.

, ,
+++ + ++

Figure 14 '111C ~Ialldard geometry for the n ail effect: a rod-shaped Ipccimcn of n!dangulal' CI"QIO'-
scdion i. placro ill a magnetic field B•• ll$ in (a). An clo.'dric field E. applied across t he end
elect rodes aiOu:lCS ..., electric current d ensity J. to flow down the rod. The drift ,""Iocit y of the
negali,-cly-charged elect...,,,. immed iately aft<:r the electric !leld i>; applied as !.L.own in (h). The
deflection ill the Ii direction ii I;lIUSCd by the magllclic field. E~mlli accumulate OIl o nC,' bee of
the rod and. posit ive iol1 ('Keen i~ cstablid>ed Oil the opposite ~ u nlil , as in (el, the Irru15\-CI"SC
elect r iC fiekl (ll all Ileld) just cancel. the Loren tz ror~ dne to the magnetic fleld.

This is negative for free electrons. for e is positive by definition .

The loy,>er the carrier concentration, the greater the magnitude of the Hall
coefficient. Measuring H, / is an important way or measuring the carrier concen-
tration.
The symbol RII denotes the Hall coefficient (54). but it is sometimes used
\\i th a d ifferent meaning. that of Hall rcsi5tance in two-dimensional problems.
When we treat such problems in Chapter 19. we shall instead let
PII := BRit ;. E,jlx (55a)
denote the Hall resistance, where j~ is the surface current density.
TIle simple result (55) follows from the assumption that all rclaxation times
are equal, independent of the velocity of the electron. A numerical factor of
order unity entcrs ifthc rclaxation time is a function ofthc velocity. The expres-
sion becomes somewhat morc complicated if both electrons and holes contrib-
ute to the conductjvity. 111e theory of the Hall effect again becomes simple in
high magnetic fi elds such that WeT" 1, where We is the cyclotron freq ue ncy and
T the relaxation time. (See Q1"S. pp. 24 1- 244.)

..
".
In Table 4 observed valnes ohhe Hall coefficient arc compared with values
calculated from the carrier concentration. nlC most accurate measurements are
made by the method of helicon resonance which is treated as a problem in
Chapter 10. In the table "cony." stands for ·'conventional."
'Inc accurate values for sodium and pota~sium arc in excellent agreement
with values calculated for onc conduction electron per atom, using (55). Notice,
however. the experimental values for the trivalent elements aluminum and
indium: these agrre \\ith values calculated for onc positive charge carrier per
atom and thus disagree in magnitude and sign with values calculated for the
expected three negative charge carriers.
TIle problem of an apparent positive sign for the charge l'3rricrs arises also
for Be and As, as seen in the table. The anomaly of tile sign was explained by
I'eierls (1928). 11lC motion of carriers of apparenl positive sign, which lIeisen-
berg later calletl "holes," cannot be cxplainoo by a free electron gas, hut finds a
IMlural explanation ill terms of tile (;nel"gy bllld theory developed in Chapters
7- 9. Hand theory also accounts for the occurrellce of \'ery large values of the
Hall cocfHcient, as for As. Sb, ami Hi.

THERMAL CONDUCflVITY OF METALS

In Chapter 5 \"C found an expression K "" led for tile thermal conductiv-
ity of particles of velocity v, heat capacity C per unit volume. and mean free
path t. 'nle thermal conductivity of a Fermi gas follows from (36) for the heat
capacity. and with t:F = im4:

(56)

Here l = VF"T; the electron concentration is n, and "T is the collision time.
Do the electrons or the phonons carry the greater part of the heat current
in a metal? In pure metals the electronic contribution is dominant at all temper-
atures. In impure metals or in disordered alloys. the electron mean free path is
reduced by collision ,\oith impurities. and the phonon contribution may be
comparable with the electronic contribution.

Rotio of Thermol to Elect,.ical Cooductivity

The Wiedemann-Franz law states that for metals at not too low tempera-
tures the ratio of the thermal conductivity to the electrical conductivity is di-
rectly proportional to the temperature, with the value ofthcoonstant ofpl"opor-
tionality independent of the particular metal. This result wa~ important in the

(
 6 Frre t:krtron Fu-mi C... 16

Table 4, Comparison of observed Hall ~mcients with free eled-ron theory

{The experimental v"lues of RII as obtained by OOIlventional methods are Summarized
trom data at room temperature presented in the Landolt-Bomstein tables. The values
obtained by the helicon wa ... e metll(l(l at 4 K are by J. M. Goodman. The values of the
carrier concentration n are from Table 1.4 except for Na. K, AI, In, where Goodman'li
values arc used. To l.'OIlvert the lI1IIuc of IIH in CCS units to the value in voIt·cmlamp-
gauss, multiply by 9 X 1011; to con\'('rt fill in CCS to m'/cou:omb, multiply by 9 X
JOI3. J

uperirr.."'la\ A~umt:d C..lt:ulaloo

Metal Method
".,
in 10- 1-1 CCS ,olits
carrie"
per ~ t om
- J/n«.
in 10- '" CCS units

U con .... - J.89 1 e lectron - 1.48

N. helicon - 2.619 1 elt!(;trOIl -2.603
conv. - 2.3
K helicon - 4.~ 1 electl'Oll - 4.9«
conv. - 4.7
Rh conv. - 5.6 1 electron - 6.04
Cu conv. - 0.6 1 e lectron - 0.82
Ag ~" , - 1.0 I electron - 1.19
Au ,",W, - O.S 1 electron - US
Be conv. +2.7
Mg ~" . - 0.92
Al helicon +1.136 1 holc +1.135
10 helicon +1.774 1 holc +1 .780
COny. +50.
'"Sh ~". - 22.
m ~" . - 6000.
J..

Table 5 Experimental LoI'Cm: numbers

L X 10" WlIu-ohmldci" L )( If/' Wlltt-ohmltlcg"-

<1'C JOO'C <1'C JOO'C
Metal

Ag 2.3 1 2.37
"""
pb 2.41 2.56
A, 2.35 2.40 PI 2.5 1 2.60
Cd 2.42 2.43 S" 2.52 2.49
C, 2.23 2.33 W 3.04 3.20
Mo 2.61 2.79 Z" 2.31 2.33

history oC the theory ofmclals, for it supported the picture of an electron gas., It
can be explained by using (43) for CT and (56) for K
!i. = 'filkjTn'I-I3m "" ~ (kB)'r . (fiT)
u neTim 3 e
The Lorenz number L is defined as
L ~ KloT • (58)
and according to (51) should have the value

L= ~ (~r = 2.72 X 10- 13 (crglesu - dcg)2

0:2.45 X 10- 8 watt-<lhm/dci . \ (59)
11\15 remarkable result involves neither tl nor m. It docs not involve 'T if the
relaxation times arc identical for c lcctricv.l and thermal processes. Experimen-
tal values of L at (f'C and at lOO"C as given in Table 5 arc in good agreement
with (59), AI low temperatures (T ~ U) the Lorenz number tends to dccrea~e;
sec the book by Ziman.

NANOSTRVcrVRES

111c tcrm nanostructure dcnotcs a condensed matter structurc having a mini-

mum dimension approximately betwecn 1 nm (to 11.) and 10 nm (100 A). Thesc
structurcs may be finc particlcs, fine wircs. or thin film s. Finc partieles t ypi-
cally contain between 10 and 1.000 atoms. Scmiconductor technology (see sec-
tion "supcrlattices» in Chapter 8) has made it possible to fabricate small pools of
electrons called in various ways: single-electron transistors. quantum dots, arti-
ficial atoms, Coulomb islands, or quantum corrals (Chapter 19). '111e unusual
physical properties of nanostructures compared with bulk solids arc attributed
to several factors involving properties treated above and in latcr chapters:
 6 Frre £Itdron Fermi Gu 169

o Thc ratio of the number of atoms on the surface to the number of atoms in the
interior may be of the order of unity.
• l1le ratio of surface energy to total energy may be of the order of unity.
o The conduction or valenec electrons are con6ned to a small length or volume,
so that the quantum ....avelcngth of the lowest electronic state is constricted
and consequently the minimum ~avelength is shorter than in the bulk solid.
o A wavelength or boundary condition shift will change the optical absorption
spectrum (Chapter II ).
o Assemblies of nanoclusters of metals may have great hardness and yield
strength. because it is difficult to create and to move dislocations (ChapteT20)
in spatially confined regions.
• Magnetic monolayers, as of ahernating films of fenomagnetic iron and of
pllramagnetic chromium, may hav~ the magnetization (Chapter 15) of the
iron films coupled by tunneling of the magnetization through the chromium
oonicrs.

Problems
I . Killetic energy of electron ga&. Show that the kinetic energy of a thn._Iimcnsional
gas of N free electrons at 0 K is

(60)
2. Presfl'l"e and bulk moelllllls of on ~lectron gas. (a) Derive a re lation connecting tile
pressure and volume of an electron gas at 0 K. Hint: Use the result of Problem t mId
the relation bctv.-een f f and electron concentration. TIle tesult may be written as
p - HUM. (b) SilO\\' thai tIle bulk modulus B = - V(ilplilV) of an eit.'CIron gas at
o K is B = 5pI3 - lOU0f9V. (e) Estimate for potassium, u~ing Table I , the value of
the electron gas contribution to B.

3. Chemical potential ill two dffllen.sKms-. Show that the dlemical potential of a Fenni
gas in two dimensiOns is given by:

(61)

fot n ek,<:trons per unit area. Note: TIle density of orbitals of a free elcctron gas in
t .....o dimensiOns is Independent of l'llergy: D(E) = mJ.rrli 2 , per unit area ofspt.'dmen.

4. "~nni gases in rutropllYrics. (a) Ch'en /1/0 = 2 x l~ g for the mass of the Sun,
estimate the number of electrons in the Sun. In a white dwarf ~tar this number of
electrons may be IOllizerl and con tained in a spJlere of r:lfiius 2 X 109 em: finrl the
Fermi energy of the electrons in electron volts. (b) TILe energy of an electron in the
re lativistic limit ~ ~ mCl is reiate<l to the wave vector as f '" pc ... ftloc. Show that
the Fermi energy in this limit is fp '" Itc{NIV)'f3 , roughly. (e) If the above number
of electrons were contained within a pulsar of raclius 10 km, show that the Fermi
energy would be - to' eV. TIlis value e~ plains wIlY pulsars arc believal to he com-
170

posed largely of neutrons ruther dlun of protons and clt-d rOrls. for the energy re-
lease in tile reaction n _ p + e - is only. 0.8 x uf cV. which is oot large enough to
enable many electrons 10 fOrm a t'crmi sea. TIle nc utrOn decay proceeds only until
th e electron concentration builds up enough locreate a Fenn; level 0[0.8 x Iff eV,
at which point tile neutron, I)ruton, and electron conrentrations are in etJuillbrium .

5. Liquitl Jle', The atom lie' has spin i and is a fermiml. The density of liquid lie' is
0 .081 gem - 3 ncar aLsolu!e zero. Calculate the fermi energy Ef and the Fermi
tem)>emture TF'

6. f'requcnClJ dependence of tile ekctricGl COf1lluctit,ity. U.se the equation m(Jv/tlt +

vlT) - - eE for the elcctron drift velocity u to show that the conductivity at fre-
(luclley w is

1+
u(w) ... 0(0) ( 1 +
;=) .
(wi! (62)

where 0(0) - """'111 .

·7. Dynomic l1'IOg1Ietoconductivit y teR$Or for fru ekdrom. A metal with a OOllCl'lllrn-
tion n of free electrons of charge - fl is in 8 static mab'lletic field 82. The electric
current density in the ly plane is rclatet"1 to the electric field by

Assume that the freque ncy w ~ "'c and w ~ liT. where w" - e8/me and T is the
collision time. (a) Solve the drift velocity Ctjuation (51) to flrtd tile components of the
magnetooonduclivity tcnsor:

Un = U~V - i w!J4'iTW ;

whcre w: -41T11C/m. (b) Notc from a Maxwclll'ljUati01l tilat tile d ielectric function
tensor of the med ium is relall'll to the conductivity tensor as E - 1 + i(4mw)u .
Consider an elcctromllgnet ie w:we ....;th waveVl'CI:Or k - ki. Show thllt tile disper-
sion relaUon for th is wave in the medium is
(63)

At a givcn frc<juency there are two modes of propagation with different wavellcctors
and d ifferent velocit ies. The two modes .correspond 10 Circu larly polari:.r,erl waVl'S.
Because a linearly polari:I;Cd wave can be decomposCtI into two Circularly polarized
wa\'es. it follows that the plane of polari:.r.a.lion of II linearly polarized wave will be
rotated by the magnetic Reltl.

-8. Cohesive el"leTgy offree eke' ron Fer-nU g~. We dcfme tJle dhnensionil'S.S length r.
as rJan. w/lere ro is the radius of a sp/lere that contai ns one electron, Mel an is the
Bollr mdiU5 Iii/em. (a) Show that the average kinetic energy per cll'Ctron in a free
ele<.1: ron Fermi gas at 0 K is 2.211r!, where the encrgy is expressed In ryelbcrgs.

(
 6 l ' rH Ekctron Fnml Gel 111

..... ith I R)' - fne4/2Jil. (b) Show that the l""OUlomb ene rgy ofa point positive charge e
interacting wit h the uniform electron distributitm of one electron in the volume of
radius ro is -lIf/2rQ, or - 31r. in rydbergs. (c) Show that tile coulomb self-energy of
the eleclron e]islrihution in the ~phere is 3Cf5ro , or 615r, in 1)1Iberp. (el) The sum
of (b) aod (c) gives - l .80/r. for the tolal coulomb energy per electron. Show that the
l.'<juilibrium ... alue of r. is 2.45. Will sud, a metal be stable wit h rcspcd to separalce]
II atoms?

9. Stalic: rnalP'etoconducd vity lenlor. For tI,e drill velocit)' theory of (51), show that
the static current densit)' can be written in matri~ form lIS

1 +L.)(~:) (&I)

In the high magnetic field limit of w.:"r~ I, sho ..... that

Up = nedB .., -0'., . (65)
In this limit un '" 0, to order lIw~7. The eluant it)' Uv- is called the HaJl condudi,·ity.

10. Mllrimum ,ur/a« re.rislance. Consider a scluarc sheet of siele L, tllicL:nl.'SS d, and
elC(:trical resistivity p. The resistancc measured bct.....t:en opposite edges of tile
shet.>t is culled the surfucc resistanre: R..,. = pI) LtI "" pld, wliich is independent of
the area Lt 0( tile sheet. (R"I is called the re~istance per square and is expre.sscd in
ohms per sCluare, becausc pld has the dimensions of ohms.) If we eX)ress P by (44),
then 'lOCI '"" mJndC"f". Suppose now that the minimum ... al .. e of tile collision ti me is
dete rmined by scaUering from the surfaces of the sheet, so that.,. .... dJvp, wllere vr
is the t~erml velOCity. TIms the maximu m surface resl~t ivlty is ROIl '" mvrlrul'ltf.
Show for a mona tomic metal sheet one atom in thickness Illat noq ... tile! ... 4. 1 kn,
where I kfl is Id' ohms.

'II, Small m£tal. spheres. Consider free electrons in a spherical square well potential of
radius a, with lin infinitel y high boundary. (a) SllOw that the w:we function of an
orbital of angular momentum l ancl projeclion m has tile form

(66)
where the r.ul.ial wa\'C function hIlS the form

and Y is a sl>llCtical harmonic. Here I is a Bessel function ofilldf.integn.l order and

satisfies the boundary condition l tHn(ka) ,.. O. The roots give the energy eigenval-
ues ~ of the levels above the bottom of the well, where ~ - /i'lJel12m . (b) Show that
the order of the levels above the ground orbital is
11, II', Id, 2s, If, 21', Ig, 2d, I h, 3.f, 2f,
where B, 1', d, f g, II denote (= 0, 1, 2. 3, 4, 5.
(
172

12. Denlity of lIates-nlJnometric wire. (a) Consider II nanometric wire in the form of
a rectangular parallelepiped, with two sides L. .. Lv " I nm and the 10.lg axis L . -
I em. TIle single particle eigcnstatCli of tIle system may be written as

t/I - sin (n,."1TXILJ sin (n~TrYll..,) exp[i2nN.tl

The energy of the eigenstate is, with n - II~, '\.,

E(n, N) - t (n) = (2ntJNf/2Jn "" AN' - 'hmui ,

where v is the electron velocity alollg th e .tonxis. Here A - (2ntitl2m and N-
V((E eAJ. Then /jE "" 2AN6N, show that tllcdensity ofstates D"at Rxed n. witll
account of the tv."() spin Orientation! and thc tv."() ± values of N, is

(b) Sum over tile values of n for \\flich E z: € .. to show tlmt

D(E) = ID ..(E ) .. ~~7rll){ IIt:{E)J6(E - t.J •
where Sex) is the Heaviside unit function, "Zero for :t < 0 and unity for :t > O.

13. Quonti.::at ion of conductance. TIle curre"t in tIle nallometric wire ofProblcm 12 is
1 - (N+ - N _)ev", where N .. - N _ .. D (E ,,)eV, where V is the bias voltage. Show
that the current may be written as

where n_ is the nu mber of occupied states n" nw' hl't1('C tlle quantil.l.'d conduc-
tance is (~/7rll~ n<>=

References

D. N. l .:tngcnberg. "Resource letter OEPM-I on the onlinary elcclronic propertlCl< ofmclals,"

Amoc... J. Ph)'I. 36. TI7 {I968}. An cu'CJlcnt carl)' bibliogral,!')' on tn.n~port dfCdl , anomalous
~J.:in cited, Al-.bel-Kaner cyciotnlfl rcso... nce. ma.gnetoplasmll w,wes, liize elfcW, conduction
eicc.1ron spin rcsonar..:e , opli<:al sp:.'Clra Hlld pl'otocm ission, (Iua"tum oscillationl, rnlIgllctic
breaJ.:dowl1, ultrasoo>c effects, and the Kohn cfIK>t.
C. T. McatIen, fkclrictJ rntsta,," of metal", l'Ienu",. New Yorl<, 1965.
C. M. l1 un!. Thfl Hall effect In me lah and allQy$, 11cllllm, 1972.
C . I. CI,ien alld C. R WCstgatE', cds., Hall effecl and iU apP/i<:tllilmJ , l'lcnum, 1980.
C. M. Hun!. Ehctrom III fntfllU, Wiley. 1975.
S. Kagoshima. ef 01., OI\e-dimen.rlonaloonduClcrr, Springer, 19611.
J. Ziman, f/eclN"m . and pllUl"Mlrnl, Oxfonl , 1900.
 (

7
Energy Bands

NEARLY FREE ELECfRON MODEL 116

Origin Qf the energy gap 177
Magnitude of the energy gap 179
BLOCH FUNCTIONS 179
KRQNIC-PENNEY MODEL 180
WAVE EQUATION OF ELECTRON IN A PERIODIC POTEJ'liTlAL 183
Restatement of the Bloch theorem 185
Crystal momentum of an eit:dron 186
Solution of tile central eqUAtion 186
Kronig-Penney model in re<:iprocal space 187
Empty lattice approx imatioll 188
Appmxi rnale soIutkm near a zone boundary 189
NUMBER OF ORBITALS IN A BAND 111£
Metals lind insulators 193
SUMMARY 195

PROBLEMS 195

1. Square lattice, rree ek-ctron encl'8ics 195

2. Free electron e ne rgies in reduced wile 195
3, Kronig-Penney model 195
4. Potential energy in th e diamond structure 195
5. Complex wuveveclors in tile energy gap 196
6. Square lallke 196
REFERENCES 196

--
 D D D D
E.,,·rgy

Cd B QO D U
...
D- D 0
-- G D
"
.
Stmlrond...._
I ....bta Semiro,,""UQr

~,,~
"""
I Schematic .,Ie<'tron O<Ulpaocy of allo ....ed e.,."rgy bands fOr an Insulator. mt.'lal, scmi-
oncl:!l, and semiconductor. n,,! vertical e>.tent of the bm..os Indicates the allOWt.-d energy regions;
the shaded a.C'.tS indic-.. te thc regions filled with electrons. In a IoCmimcial (such acs bl.'imuth) one
ba"d is almost 1111•.'<1 and another band is nearl)· "mply at abwlute zero. but II pure _nkondudor
(such as silicon) bt,comes an il\$ulator lit absolute zero. 1'h., left orthe 1\<0} scmio:oudliClon ~hown;s
al a finite lempcrnlu..e, "illl Clllrien "x";ll'il thermally. 11,e oIhc:r w - ~ductor is electr(llJ.ddi-
dent hemu.., of impurities. (

'"
 CHAPTER 7: ENERGY BANDS
•
Wile" I , 'ar'etl to 'Milk about it, I fdt tllll t the mai"
problem lro, to ~pl"jll how th e electroll~ could 'nook
by all tire iol~ in a melal . ... By ' traigh t f'Otlrier
IlfWl'Ysis I focmtl to my delight that the 10000 dijJerelf
from tilt: ,'/"'1£ lcave of free electroru only by a p£ri-
odic mOlllllatiOlI.
F. Bloch

Thc rree clectron model or m<.>tals givcs us good insight into thc heat
capacity, thcnnal conductivity, electrical condllcth-ity, magnctic susceptibility,
and electrodynamics or metals. But the model rails to help us with other large
questions; the distinction between mcta1s, semimetals, semiconductors, and
insulators; the occurrence or positive values or the Ilall coeffici<''11t; the relation
of conduction electrons in the mctal to the \'alence electrons or rree atoms; and
many transport properties, particularly magnctotransport. We need a less naIve
theory. and rortunatcly it tUniS out that almost any simple attcmpt to improve
upon the rree electron modcl is enonnously profitable.
The differcncc hctween a good conductor and a good insulator is striking.
The electrical resistivity or a {'ure me tal may be as low as 10- 10 ohm-cm at a
te mperature or I K. apart rrom thc possibility or supe«.'Onductivity. The rcsis-
tivity or a good insulator may be as high as 1<f2 ohm-cm. This range or l(il2 may
he the widest or any commOn physical property or solids.
Evcry solid contains electrons. 111e important question ror electrical con-
ductivity is how the electrons rcspond to an applied electric field. We shall see
that electrons in crystills arc arranged in ene~ hands (Fig. l) separated by
regions in energy ror whieh no wavelike electron orbita1s exist. Stich rorbidden
regions are called e nergy gaps or band gaps, and result rrom the interaction or
the conduction electron waves with the ion corcs or thc crysta1.
The crystal behavcs as an insulator ir the allowed energy bands arc either
filled or empty, ror then no electrons can mo\'C in an electric field. lnc crystal
behavcs as a metal ir one or more bands arc partly filled . say between 10 and 00
perce-nt filled. The crystal is a semiconductor or a semimetal ir oneor two bands
are slightly £illcd or slightly cmpty.
To understand the difference between insu lators and conductors, we must
extend the free electron model to take acrount or the periodic lattice of thc
solid. The possibility or a band gap is thc most important new property tlmt
emergcs.
We simI! encounter other quitc remarkable properties or electrons in crys-
tals. For example, they respond to applied electric or magnetic fields as ir the
electrons were endowed with an eITective mass m*, which may be larger or
smaller than th... '-rcc electron mass, or may even be negatil·e. Electrons in
crystals rcspon. ,) applied fields as ir endowcd with ncgath'c or positive
17'
charges, -e or +e, and herein lies the explanation or the negative and positive
values of the Hall coefficient.

l\'t~ARLY .' REE t~ LEcrnON MODEL

On the frcc electron mooel the a110\..w energy values arc di~tributcd es-
sentially continuously from zero to infinity. We saw in Chapter 6 that

E~ = -
I,'
-(1.-; + ~ + k!) • (/)
2m
where. for periodic boundary conditions over a cube of side L,
2~
+ - ~ -
L ' .. .. (2)
- L '

The frcc electron wavcfunctiom arc of the form

!JtI..(r) = cxp(ik . r) ; (3)

they represent running waves and carry momentum p = Me

'nlc band structure of a crystal can orten be explained by the nearly frcc
ck-'ctron model for which the band clcctrom arc treated a.~ pertu rbed only
weakly by the periodic potential of the ion corcs. This model answers almost all
the qualitative questions aboll l the behavior of electrons in melals.
We know tllat Bragg rcnectioll is a characteristic feature of wave propaga-
tion in crystals . Bragg reflection of electron waves in crystals is the cause of
energy gaps. (At Bragg reflection wavelike solutions of the Schrodinger equa-
tion do not exist, as ill Fig. 2.) These energy gaps arc of decisive Significance in
determining whcther a solid is an insulator or a conductor.
We explain physically the origin of energy gaps in the simple problcm of a
linear solid of lattice o::ollstant o. Thc low cnergy portions of tll C band structure
arc shown qualitatively in Fig. 2, in (a) for cntirely free electrons and in (b) for
clectrons that arc llcarly free, but with an cnergy gap at k "" ± -rrIo . The Bragg
condition (I.: + G )2 = J,1l for diffraction of a wavc of wavevector I.: becomes in onc
dimcnsion
k ::: ± iC= ± wTTla, (4)
where C = 2'IT1l10 is a rcciprocal lattice ....ector and n is an integer. The 6rst
reflections and thc first energy gap occur at k = ± 1T/a. TIlC region in I.: space
betwcen - 1Tla and -rrIa is the first Brillouin zOlle of this lattice. Othcr encrgy
gaps occur for other v<llllcs of the integer n.
The wavcfunction5 at k = ± ""a
arc not the traveling waves exp(i'7T:tla) or
cxp{ - i1TXla) offrce electrons. At thesc special "alues of k the wavcfullctiolls arc
 7 E_rgV Hcndr 177

-+-----''''"'''--!c.
-. - ,
figure 2 (:i) Plot <£ energy ~ ,~rsus wa\l:\<t.'c tor k ror ar,ee elt.·t:tron. (b) riot of.,.,.,rgy \'Cr5\15
"'lI\~.'t:ctor IN WI electron in a monatomic linear lallict <£ bottk~ constant a. nle L'Ilergy gap f:~
sl>O\<'" is a~ated with the first U....ogg rcfu.:ction at k ... ::t: 'frla: oth ..'!" gaps are fuu"d at !: nma, for
int<.>gra' .aIucs of n.

made up of equal parts of "'"lives traveling to the right and to the left. When the
Bragg reflection CQndihon k = ± '1fla is satisfied by the wa..'cvcctor, a wave
traveling to the right is Bragg-reflected to tra,,'el to the left , and vk'C vcrsa.
Each subscquent Bragg reflection will revcrse the direction of travel of the
wavc. A wavc tllat travels neither to the right nor to the len is a standing wave:
it doesn't go anywhere.
The time-independent state is represented by standing wavcs. We can
fOTffi two different standing waves from thc tv..o travcling """lives cxp(± im:Ja),
namely
.p(+) = exp(i1TxJa) + exp(-imla) ::: 2 cos (m:lo.) ;
(5)
1/1(-) = cxp(i'lTXia) - cxp( - i7Tlia)::: 2i sin (m:l a) .
The standing waves arc labelctl (+) or (-) according to whether or not they
change sigl1 whcn -x is suhstituted for x. Both stand ing waves arc composed of
equal parts of right- aud left-dircctcd tra\'eling waves.

Origin 0/ rhe Energy Cap

The two standing wavc.s .p(+) and .p(- ) pilc up electrons at dilTerent re-
gions, and therefore the two wavcs have different "alucs of the p.ltcntial
cncrgy. This is tile origin of the energy gap. Thc prohability density p of a
particle is YI'"!JI ::: 1¥42. For a pure traveling wave cxP(ikx), we have p :::
exp( - ik%) cxP(ikx) = I. so that the charge density is constant. The charge dcn~
sity is 110t constant for linear comhinations of plane waves. Consider the stal1d~
ing wave .x. +) in (5); for this we have
p(+) = 1.x.+~2 <X cos2 n:tfa .
This function piles up electrons (negative charge) on the positive ions centered
at x ::: 0, a, 2a, . .. in Fig. 3. where thc potential cnergy is lowest.
178

Figure 3a picttlrcs the variation of the electrostatic potculial energy of a

conduction elcctron in the field of the positive iOIl cores. The ion cores bear a
net positive charge lxx:ause the atoms arc ionized in the metal, with the va-
lence electrons taken off to ronn the conduction band. The potential energy of
an electron in the field of a positive ion is neg-dtive, so that the force between
them is atlractive.
For the other standing wave ¢(-) the probability density is
p(-) = 11,11(_)1 2 0:: sin2 7rr/(J •

which conccntratc5 electrons away from the ion corcs. In Fig. 3b we show the
electron <.'Onccntration for the standing waves tJi(+), ¢( - ), and for a traveling
wave.
\Vhen we calculate the average or expectation values of the potential en-
ergy over these thrcc charge distributions, we find that the potentia] energy of

.J

Figure 3 (a) Variallon cl potential eTM!I1IY of a ronduction ek'Ctron in the rlC:ld nfthe ion cores of
'"
" IinL"aI" JattiC!:. (b) Distribution of probabilit y density p in the lallice lOr I+<-~· slnll miD;
14(+ ~1« cos' mia, and for a tra"eling w,we. TI,e wavt:f"""tion 4(+ ) piles up ek'Ctronic dill rile On
thecon."& of the positive ions, tht.'feby lowering the potential energy in cornp:u-ison ,dth the ""crage
pote ntial t."Ocrgy ......"O by II lra,eling wa,-e . n,e ~f"nc:tion 11(-) plks up charge In lhe ~gion
bt1wetn the joos. then.>by raising the potential energy in «:woparison with that seen by a lra"ding
Wave. n,ls figure u the h-y to understanding the angin of the COl gap.
 7 F.nergy Haruh 17!

Ii. +) is lower than that of the traveling wave, whereas the potential energy of
p{- ) is higher than the traveling wave, We have an energy gap of width E/: if~he
energies of{i.- ) andli.+) differ by E/:. Just below the energy g;tp at points A in
Fig. 2 the wavefunction is Vi{ +), and just above the gap at points B the wave~
function is ~ -).

Magnitllcle of tile Energy C6p

The wavefunctions at the Brillouin wne houndary k :: TrIa arc
Y2 cos 'm:la and Vz sill wa, normalized over IInit length of line. We write
the potential energy of an ek'Clron in the crystal at point x as
U(x) :: U cos 2wa .
TIle first-order energy difference between the two standing wave states is

£, ~ o
fdx U(x)ll"'+i'-I"'-i']

f
= 2 dx U co.s(2wa) (cos 2 wa - sin2"ID:la) :: U
(6)

• We sec that the gap is equal to the Fourier component of the crystal potential.
BLOCH FUNCnONS

F. Bloch pro\'cd the important theorem that the solutions of the Schro-
clinger Ct"\uation for a pcr;ooic potential mllst be of a special fonn;

I!her) = IIk{l") exp(ik . r) , I (7)

where "k(r) has the period of the crystal lattice \\;th 11 ..(r) = IId r + T). The
result (T) expresses the Bloch theorem:

The eigenfunctions of the wave equation for a periodic potential arc

the product of a plane wave exp(i1.: . r) times a function I/...{r) with the
periodicity of thc crystal lattice.

A one-electron wavefunclion of the form (7) is called a Bloch function and

can be decomposed into a sum of tra\'eling wavC5, as we sec later. Bloch func-
tions can hc assembled into localized wave packets to represent electrons that
propagate freel y through the potential field of the ion cores.
We give nQw a restricted proof of the Bloch thoorem , valid when I}I" is
nondcgcnerate. l1mt is, when there is no other wavefunetion with thc same
encrgy and wa,· "":ctor a'i I}I". TIlc general case will be treated later. We con-
\
,.,
sider N identical lattice pOints on a ring ofl cngth No. 11lC potential energy is
periodic in a, ,\;th U(x) :: U(x + $6). whCTe s is an integer.
We arc gU ided by the s)'InlTIctryofthc ring to look for solutions of the wave
C<}uation such that
1/1(% + oJ =0 CI/J(x) • (8)
where C is a constant. Then. on going on<.'C around the ring.
tJA"x + Nu) "" VJ(x) = eN .p(r) ,
because ¢(x) mllst be single·va1l1oo. It follows that C is one of the N roots of
unity, or
c= cxp(i2m1N) ; $=0, 1, 2, .. . ,N- I . (9)
We sec that
.p(x) :::: u,"x) cxri,f2-rrsxINa) (10)
satisfies (8), provided that ",,(x) has the periodicity a, so that II,,(X) :::: U,"X + oj.
With k"" 27tsINa, we have the Bloch result (7), For another de rivation, sec
(29) .

KRONlC·PENNEY MODEL

A periodic potential for which the wave equation can be solved in terms of
elementary functions is the square-weI! army of Fig. 4. ' J11C wave equation is
h2 cf",
- 2m Jr + V(x)1/I ::: E1/I (II)

where Vex) is the potential energy and E is the energy eigelwalue.

In the region 0 < x < a in which V::: 0, the eigenfunction is a linear
comhination,

or pfanc waves tmvcling to the right and to the lert, with encrgy
E ::: h 2K2/'bn . (13)
In the region - b < x < 0 within the barrier the solu tion is or the form
1/1 ::: CeQ~' + Dc- Q• , (14)
with
(l5)
 181

. ,_ _ u.

tip .... 4; Sq~"''ell periodic pollTLtiaI as in-

Irod llC«! 1», Krooig and PennL1'. - !«+b) - b 0 a at b

We want the complete solution to have thc Bloch fonn (7). Thlls the solll·
tion in the region a < % < II + b mllst he related to the solution (14) in the
region -b < x < 0 by the Bloch theorem:
~a <r<a + b) "" ~- b < x< O) ell«· ... ") • (16)
which serves to define the wa\'cvcctor k used as an index to label the solution.
l1lC constants A, B. C, D are chosen so that I/J and cl.pldx arc continuous at
x = 0 and x "" o. These arc the usual quantum mechanical boundary conditions
in problems that involve square pote ntial wells. At r '"' 0,
A+B=C+D; (17)

'K(A - B) = Q(C -D) . (18)

At x =: (I, with the usc of(16) for I/J{tJ) under the barrie r in tenns of 44,- b ).
A~t<. + Be - lid = (Ce- Qb + IJ4P"'J e"t" +"l ; (19)
iK(A~ - Be- I".,) ,£ Q(Ce- Qb - DcQh) tllt{tl+b) (20)
The fouf equations (17) to (20) have a solution only if the dctcn n inant of
the coeffICients of A, B, C, D vanishes, or if
«Q 2 _ K2p'2QKJ sinh Qb sin Ka + cosh Qb cos Ka "" cos k(a + b) . (21a)
It is rather tedious to ohtain this equation.
The result is simplified if we represent the potential by thc periodic delta
function obtained when we pass to the limit b := 0 and Uo = QQ in such a way
that Q2baJ2 = P, a finite quantity. In this limit Q J> K and Qb <!!I: L Then (21a)
reduces to 1

(PIKo)sin Ka + cos Ka = cos ka . (2 1b)

The r.mgcs of K for which this equation has solutions arc plou ed in Fig. 5,
for the case P "" :mI2. Thc rorrcsponding values of the energy are plotted in
Fig. 6. Note thc energy ~ps at the zone boundaries. 'DIC wa\'c",cctor k of the
Bloch function is the important inde,.;, not the K in (12), which is related to the
energy by (13). A treatment of this problem in wa\,C\·cctor space is given later
in this chapter.

b
' 82

Figure 5 Plot of the function (PI KD) sin KQ + COS 1Vl, lOr P - 3'1rl2. The " llowed ,'alues of the
energy e are give" by those 1'1lngl"J <£ Ka - (2mElft~'lIa Ibr w!lich the function lies between :!: 1.
For other nlues of the energy there are n.o 1....,·c1ing ".,.,-e 00" Bkdl-Ji).;e solutions 10 the wa,~
equatioo. so thai lOrb;dclcn gaps in the enc.-rgy spedrum an: 5:>r11'll..i

.,

~
:>•
i§,
.,
"
10 I
,,
•, ./
,,
~
/
, --- "-' ..'" '" ,-
Figure 6 Plot of .,nerg)' \'$. w;t'·I,... lUmbe.- lOr the Kronig.I'cnnL")' polential, with p _ 3'1T12. Notice
the energy gaps at kit - 'IT, 271', 3'IT .
 1 £nerp &n<h 183

WAVE EQUATION OF ELECTRON IN A PERIODIC POTENTlA.L

We considered in Fig. 3 the approximate fonn we expect for the solution of

the SchrOdinger C<luation ifthewa\levector is at a zone boundary, as at k = ma.
We treat in detail the wave equation fOT a general potential, at general values of
k. Let U(x) denote the potential energy ofall electron in a linear lattice oflattice
constant a. We know that the potential energy is invariant under a crystal
lattice translation: Vex) = U(x + a). A function invariant undcr a crystallaUicc
translation may be cxp.1.ndcd as a Fourier .series in the rcciprocallattice \I(.'ctors
G. We write the Fourier series for the potential energy as

U(x) = L Vc e'Cr . (22)

c
The \lalues of the coefficients Va for actual crystal potentials tend to decrease
rapidly with increasing magnitude of G. For a bare coulomb potential Vc de-
crca.scs as lIG:it.
We w.tnt the potcntial encrgy V(x) to be a real function:

V(x) = L Vc(e IC %+e- IC ') =2 L Vc cos Gx . (23)

c>o c >o
For convenience we have assumed that the crystal is symmetric about x "'" 0
and that Vo = o.
The wave equation of an electron in the crystal is:H1/! = EI/!, where:H is the
hamiltonian and Ii is the encrgy eigenvalue. 11le solutions I/! arc called eigen-
functions or orbitals or Bloch functions. Explicitly, the wave equation is

( _ 1_ 1)2 + V(x») I/f{x) = (_ I_ p2 +L Vc e/C%) I/I{x) = EI/I{X) • (24)

2m 2m c

Equation (24) is written in the one-electron approximation in which the orbital

~) describes the Illotion of one electron in the potential of the ion cores and in
the average potentiaJ of the other conduction electrons.
The wavefunction I/J{x) may be expressed as a Fourier series summed over
all valucs of the wavevector pennittcd by the houndary conditions, so that

~; L, C(k) ..... (25)

where k is real. (We could equally well write the index k as a subscript on C, as
in C".)
The set of values of k has the form 27mfL. because these values satisfy
periodic boundary conditions over length L. Here n is any integer. positive or
ncge.tive. We 110 (-. assume, nor is it generally true, that ~x) itself is periodic
.64

in the fundamcntallaUire translation 6. The translational properties of I/I(x) are

detennlned by the Bloch theorem (7).
Not an wavcvectors of the set 27ml L enter the Fourier expansion of any
one Bloch function. If one particular wavevector Ie is contained in a tP. then all
other wavevectors in the Fourier expansion of this'" will have the form k + G,
where G is any redproca1 lattice vector. We prove this result in (29) below.
We can label a waverunction '" that contains a component Ie as .pI< or,
equally well, as I/Ik+C. because if Ie enters the Fourier expansion then k + C
may enter. 'nlC wavcvectors k + G running over G arc a restricted subset of
the set 21TnlL, as shown in Fig. 7.
We shall usually choose as a label for the Bloch function that Ie which lies
within the fin.t Brillouin zone. \Vhen other conventions arc used. v,'t: shal1 say
so. This situation differs from the phonon problem, where there are no compo-
nents of the ion motion outside the first zone. The elcctron problem is like the
x-ray dilfraction problem because the electromagnetic field exists everywhere
within the crystal and not only at the ions.
To solve the wave equation, substitute (25) in (24) to obtain a set of lin ear
algebraic equations for the Fourier coefficients. TIle kinetic energy term Is

1
_p2.p(r):::: -
1 (
- ifi -
d )2I/I(r):::: -fi-
2 d-.p = -l:
2 h'1.
Jc2C(k) elL: ;
2m 2m dr 2mtlr 2m"
and the potential energy term is

(LC Ue ~G') </(x) = LC Lk UG~e'C(k) ~"

-nle wave equation is obtained as the sum:

L ~k'C(k) "'" + L L U"C(k) ",He" = <L C(k) "'" . (26)

" 2m C " "

Each Fourier component must have the same coefficient on both sides of the
equation. Thus

I (Ak - E)C(k) + t U(;C(k - G ) = II -I (27)

with the notation

(z.'l)

Equation (27) is a useful form of the wave equation in a periodic lattice,

although unfamiliar because a set of algebraic equations has taken the place of
 7 .E_rgy &nd.r IllS

.~ . 1.:.+ t-
o

.f· ······r·t..... . . ··t·

- 30 - 20 - 10
Tn
-......
0
t·· ....... f·•••·. . ··f·. ··. •. ··f···
10 20 30
k, in un;ts 2". L

Figu .... 7 The lower (lolnts rcprc...,nt \al",-,s of the "3\·cn."c tor k - 2"nlL 3Ucn.....t1 by the pcriodic
bouudary ,~onditjon on tbe "",\'elimction OH'T It nUl( of circumre!"Cnce L CQmpos.,d or20 I,rimitj,·c
«,Us. 11.c ,.]J,cn."C<.! \alut'! continue to ~"'. The uPl'er points rel'r~n t Ihe flnt rew"'a\'e"ecton
wh;ch mit)' n tter into the .-oo.,ric,r cxp"...1on of • w",-dunction \ltV), slarling rrom •
particular
"lI,-.,.«:tor k - k., ~ - 8(2.../L). The shorte' rcdj>rocal bttice ,-cdar ~ 2fT/a ~ 2012""')'

the usual differential equation. nle set appears unpleasant and fonnidable be-
cause there are, in principle, an infinite numberofC{k - G) to be determined,
In practice a small number will often suffice, perhaps h>.-'O or four . It takes some
experience to appreciate the practical advantages of the algebraic approach.

Rntatement of the Bloch Theorem

Once we determine the C's from (27), the wavefunction (25) is given as

~,(x) ~ L C(k - G) <",,-0). (29)

G

which rna)' be rearranged as

with the definition

flk{X) - L C(k - G) e - /C~ ,

G

Uecause fI~X) is a Fourier series over the reciprocal lattice vectors, it L!,
invariant under a crystal lattice translation T, so that u,Jr) "" + n. We 11*
verify this directly by evaluating IIA(r + T);
IIdx +n ... l: C(k - G) e - iG (.<+ T ) "" e- IGT(l: C{k - G) e- iC..-J "" e- I('T Ilk{X) .

Because exp(- iCT) :: I by (2. 17), it follows that u,rlr + T) "" lIi{X), thereby es-
tablishing the periodicity of Uk. lllis is an alternate and el[act proof of the Bloch
theorem and is va1id even when the iJ!" are degenerate.
'86

Crystal Momentum of an Electron

What is the significance of the wavevector k used to label the Bloch func-
tion? It bas several properties;

• Under a crystal lattice translation which carries r to r +T we have

(30)
because U"T + 1') = Ilk(r). Thus exp{ik . 1') is the phase factor l by which a
Bloch function is multiplied when we make a crystal lattice translation T.

• If the lattice potential vanishes, the centra1 equation (27) reduces to

(ft." - t)C(k) = 0, so that all C(k - G) are zero except C(k), and thus tl..{r)
is constant. We have ~(r) = eft<.r. just as for a free electron. (This assumes
we have had the foresight to pick the "right" k as the label. For tnany
purposes other choices of k, differing by a reciprocal lattice vector, will be
more convenient.)

• The quantity k enters in the conservation laws that govern collision processes
in crystals. (The conservation Jaws are really selection rules for transitions.)
Thus lik is called the crystal momentum of an electron. If an electron k
absorbs in a collision a phonon of wavevector q, the selection rule is k + q =
k' + G. In this process the electron is scattered from a state k to a state k'.
with G a reciprocal lattice vector. Any arbitrariness in labeling the Bloch
functions can be absorbed in the G without changing the physics of the
process.

Solution of the Central Equation

Equation (27) may be called the central equation:

(A, - .)C(k) +L UGC(k - G) ~ 0 (31)

c
represents a set of simultaneous linear equations that connect the coefficients
C(k - C) for aU reciprocal lattice vectors C . It is a set because there are as
many equations as there are coefficients C. -nlese equations are consistent if
the determinant of the coefficients vanishes.
Let us write out the equations for an explicit problem. We let g denote the
shortest C. We suppose that the potential energy U(x) contains only a single

'We may also say that exp(,"" . T) i~ the eigeowalue of the crystaltramlalion opernlion Y. and ""'-
is Ihe tigeo-.vector. Thai is. Yth(x) = th(:r + T) - EOII'(ik' T)oh(.t). so Ihat" i'i uuilabJe label fo~ the
eigenvalue. He~e we h'we used the Bloch theo ..,m.
Fouricr component U, = U_,. denotcd by U. Then a block orthc dctcrminant
of the cocff'JCients is given by:
A" ... 2,( - E U 0 0 0
U A,,_,; - E U 0 0
0 U A" - E U 0 (32)
0 0 U Ak +1! - E U
0 0 0 U A..... 2g- E

To sec this, write oot five successive cquations of the set (31). The determinant
in principle is infinite in extent, but it will oftcn be sofficient to set equal to zero
the portion we have shown.
At a given k, each root E or f::" lies on a different e nergy band, c.o.:cept in case
of coincidence. 111C solution ofthc detennlnant (32) gives a set of energy eigen-
...aloes E"", where n is an index for ordering the energies and k is the w:avcvector
that labels C".
Most often k will be taken in the first zone, to reduce possiblc confusion in
thc labeling. If we chose a k different from the originaJ by some reciprocal
lattice vector, we y"ould have obtained the same set of equations in a different
order- but having the same energy spectrum.

Kronig-Pe,mey Model in Reciprocal Space'

As an example of the use of the central equation (31) for a problem that is
e~ly solvable, we use the Kronig-Penney modcl of a periodic dclta-function
potential:

U(x) = 2 L
G>O
Uc cos ex = AaL 8(x - sa) , (33)

whcre A is a constant and a the lattice spacing. Thc sum is over aJl integers s
between 0 and 1/(1. The boundary conditions are periodic over a ring of unit
length, which means over lIa atoms. 1110S the Fourier coefficients of thc poten-
tial arc

Uc = LI dx U(x) cos ex = A(I~ LI dx 8(x - sa) cos ex

(34)

"" AaL cos Gsa"" A

2-J"his tn... tment "-all suggested by Surjit Singh, Am. J. pil)'I . ~I, 179 (1983).
188

We write the central equation with k as the Bloch Index. lllUs (31)
becomes
(Ak - E)C(k) + iL C(k - 21m/a) :: 0 • (35)
"
where Ak • fl 2k'lf2m and the sum is over all integers n. We want to solve (35)
for E(k).
We define
I(k) - L
C(k - 2mJ.) , (36)
"
so that (35) becomes
C(k) = _ (2mAlh~I(k) (37)
k'1- - (2f)lE/fr'lJ
Because the sum (36) is over all coefficients C, \o\-"e have. for any n,
I(k) - I(k - 2mJ.) . (38)

-nils relation leis us write

C(k - 2mJ.) ~ -(2mAlh')I(k)[(k - 2~o1.f - (2m"h~I-' (39)
We sum both sides over all n to obtain, using (36) and cancellingf(k) from both
sides.
(h'/2mA) - - L [(k - 2mJ.)' - (2m"h~r ' . (40)
"
'OIC sum can be carried out with the help of the standard relation

ctnx""L 1 . (4 1)
.. 017"+%

After trigonometric manipulations in which \\'e use relations for the difference
of two cotangents and the product of two sines, the sum in (40) becomes
if sin Ko.
(42)
4Ka{cos ka - cos Ka)
where \.\<'e write K2 >= 'bnE/h 2 as in (13),
TIle final result (or (40) is
(mAa 2 /2A2)(Ka) - 1 sin Ka + cos Ka = cos ka • (43)

which agrees with the Kronig-Penney result (21b) with P written for mAa2/2fl2.

Empty Lattice ApPrO%imation

ActuaJ band structures are usually exhibited as p lots of energy versus
wave vector in the first Brillouin zone. When \>.avevector!; happen to be given
outside the first zone, they are carried back into the first zone by subtracting a
 7 E"crgylland$ 169

suitable rcciprocallattice vector. SlIch a translation can always be found . 'Ille

operation is helpful in visualization and economical of graph paper.
When band e nergies are approximated fairly v.ell by free electron energies
lk :: h~k2l2m, It is advisable to start a calculation by carrying the free electron
energies back into the first wne. l 11c procedure is simple enough once you get
the hang of it. \ Ve look for a G such that a k' in the first wne satisfies
k'+G = k .
where k is unrestricted and is the true free eledron wavevector in the empty
lattice. (Once the plane wave is modulated by the lattice, there is no single
"true" wavevector (iw the state l/I.)
If we drop the prime on k' as unnecessary baggage. the free electron
energy can always be written as
l(k~,kl/ .kz) = (h 2 I2m)(k + G)2
= (h2I2m) [(k... + G...)2 + (kl/ + GIl + (Ie" + Gz)2)
with k in the first zone and G aHov.ed to ron over the appropriate reciprocal
lattice points.
We consider as an example the !OY.·-Iying free electron hands of a simple
cubic lattice. Suppose we want to exhibit the energy as a function of k in the
(looJ direction . For convenience, choose units such that ft212f1l == I. We show
several low-lying bands in this empty lattice approximation with their energies
l(OOO) at k = 0 and l(k...oo) along the k... axis in the first zone:

... ~ Cal21T ( (000) t (k,OO)

I
2,3
000
100,100
0
(2,",0)1
(27Tla)lI
"1..;+
•
(k... :t 2mof
(2mo)lI
4,5,6,7 O1o,oio,OO I ,OOI
8,9, 10, 11 11 0,101, 110,101 2I..27T10f (k... + 2maf + (27TJO)'
12, 13, 14, 15 JIO,Io l ,TIo,101 2I..21Tlo)1 (k. - 2ma)1 + (27T1a'f
16,17, 18, 19 OIl,OII,OIi,OTI 2(21'Tlo)t I..~ + 2(21'T/o)t

1l1ese free elcctron bands are plotted in Fig. 8. It is a good exercise to plot the
same bands for k parallel to the [II J) direction of wavevcctor space.

Approximate Solution Near a Zone BOIItlMry

\ Ve suppose Ihat the Fourier components Uc of the potential energy are
small in comparison with the kinetic energy of a free electron at the zone
boundary, We first consider a wavevector exactly at the zone boundary at ie,
that is, at wIn . Here ,
(k - cl' = (IC - C)' = (IC)' ,
100

Jo'jgure 8 Low-Iying free electron energy I~ a

L __ --=:::,_-;,-_""'---__-,!
• 0
• i._
the emjlty Ie laUice. "5 translOrmed to the fln! Bril-
louin ZOne ~nd plotted n. {k.OO). The rree electron
energy is "'(Ie + C),f2m, where the C'I are gi"eJl in
the second ro/umn of the table. The bold curves are
in the fin;t Brillouin zone, wilh -frla:iS Ie. :s mao
.. Energy bands drllwn In this way llTe uK! to be in the
"i" reducec:l :rone $Cherne.

so that at the zone boundary the kinetic energy of the two component waves
k = ::t Ie arc equal.
If C(jG) is an important coefftcient in the orbital (29) at the zone bound~
ary. then C(- iG) is also an Important coefficient. This result also follows from
the discusSion of(5). We retain only those equations in the central equation that
contain both coefficients C(IC) and G(-iC), and neglect all other coefficients.
One equation of (31) becomes, with k = IG and ,\ - li~tG)2/2m,
(> - .)C(IG) + uC(- IG) = 0 (44)
Another equation of (31) becomes
(> - .)C(-IG) + Uc(IG) = 0 (45)
These two equations have nontrivial solutions for the two coefficients if the
ener'b'Y E satisfies

U
A- E
I =0 • (46)
 7 Energy Bond. 191

whence,
h'
E = A ± U = _(!G)2 ± U . (47)
2m
The energy has two roots, one lower than the free electron kinetic energy by U.
and one higher by U. Thus the potential energy 2U cos Gx has created an
energy gap 2U at the zone boundary.
The ratio of the C's may be found from cither (44) or (45):
-"C~(-",I'O'Gcc
)~ _E_-_, ~ + 1
(48)
C(IG) U -,
where the last step uses (47). Thus the Fourier expansion of I./{x) at the zone
boundary has the two solutions
I{I{x) = exp(iGx/2) ± exp( - iG112) .
These orbitals are identical to (5).
One solution gives the wavcfunction at the bottom of the energy gap; the
other gives the wave function at the top of the gap. Which solution has the lower
energy depends on the sign of U.
We now solve for orhitals with wavevector k near the zone boundary iG.
We use the same two-component approximation, now with a wa,·efunction of
the form
I./{x) = C(k) e lh + C(k - G) ei(k - G)~ . (49)
As directed by the central equation (31), we solve the pair of equations
('. - E)C(k) + UC(k - G) ~ 0 ;
(Ak_C - E)C(k - G) + UC(k) = 0
with Ak defined as h 2J(l12m. These equations have a solution if the energy E
satisfies
U
Ak _ C -
I
E
- 0

whence ~ - E(Ak- G + A.v + Ak_CAk - U 2 = o.

The energy has two roots:
(SO)
and each root describes an energy band, plotted in Fig. 9. It is convenient to
expand the energy in terms of a quantity j( (the mark over the j( is called a
tilde), which measures the difference k !!!!!! k - iC in wavevector between k
and the zone boundary:
EI;:=(h2/2m)(lG 2 + j(2'j ± [4A(h 2k 2/2m) + lPl'f2
""' (h2/2m)(lG 2 + k 2) ± U[l + 2(AllP)(1i2j(2/2ml] (51)
in the region 1i2Cltf2m <ll luI. Here A = (1i2/2m)(iCt as before.
'"
,
•

~,
".-
(o'ig\lre 9 SoIul>onS ofr-;o) In the
"

pcriod~ woe sdK'me, In the repoll onr a boundary of the flllit

Brillouin zone. The \I"il~ are ,,,ell that U " - 0.45; C .. 2, and"'Im ..
J. n.e free electron curve Is
drawn for comparison. n.e i!f1erxY gap allhe zolle boundary '" O.90. The ... alue of U has deliberat ely
ken wOu:n large fot this illustrat ion, too large for the two-Ierm a l'l',.oo;j"",,;o1'l to be IK'CUnlle.

Writing the two zone boundary roots of (47) as E{:t), we may write (51) as

(52)

-nlCSC are the roots for the energy when the y,oavevcctor i5 very close to the
zone boundary at iG.
Note the quadratic dependence ofthe energy on the wavevedor i<.. For U
negative, the solution E(-) corresponds to the upper of the t ....u bands, and E(+)
to the lower of the t .....,o bands. 111C tv.'O C's are plotted in Fig. 10.

NUMBER QIo' ORBITALS IN A BAND

Consider a linear crystal constructed of an even number N of primitive

cells of lattice constant a. In order to count states we apply periodic boundary
conditions to the wavefunctions over the length of the crystal. The allowed
values of the e lectron wavevector k in the first Brillouin zone are given by (2);
2w Nw
~-. (53)
L ' L
We cut the series off at N1TIL = TrIa, for this is the zone boundary. Th e point
-N1TIL = -ma is not to be counted as an independent point because it is
 ,~--------t
..'"
Fin'

-----

, ,.,
A_ Ie

+ C(k - C) e~rti(k - C»:I "" ooc"lated

I'1g11re 10 Ratio of the roelr.cft,Tlts in~) - C(k) np(ikt)
near the boundary of the first Brillouin ZOne. OnecompOOlent dominates "" ...-e 1110\." away from the
boundary.

connected by a rcciprocallattice vector with TrIo . l11e tota1 numberofpoints is

exactly N, the number of primitive cells.
Each primitive cell contributes ell:actly one independent value of k to each
energy band . TIlis result carries over into three dimensions. \Vuh account
taken of the two independent orientations of the electron spin, there 81-e 2N
independent orbiwls in each energy band. If there is a single atom of valence
one in each primitive cell. the band can be half filled with electrons. If each
atom contributes two valence electrons to the band. the band can be exactly
filled. If there are two atoms of valence one in each primith-e cell. the '>and can
also be ell:actly filled .
Metal. and ,,,,ulatoTi
If the valence electrons exactly fi1l one or more bands, leaving others
empty, the crystal will be an insulator. An ell:temal electric field will not cause
current flow in an insulator. (We suppose that the electric field is not strong
enough to disrupt the electronic structure.) Provided that a filled band is sepa-
rated by an energy gap from the next higher band, there is no continuous way
to change the total momentum of the electrons if every accessible state is filled.
Nothing changes when the field is applied. 111is is quite unlike the situation for
free electrons for which k increases uniformly in a field (Chapter 6).
'94

o o
••
H
1- -,; 1- -
(0) ,b, '0'
Figure II Oo::\'pied stales and band slruclun,s giving (II) An lruuw.t..... , (bJ. mdal or Asemimctal
bt.><:ause ofband !I\IeI"Iap. and (el a m('tal bccaUIC of clt<ctron l'OOCenlllltion. III (b) tht! overlap need
not occur along the ... me direction. in the BrillDuin ZO""'. If the O\'crlap is small, with relatively few
slates Invol.-ed, we .peak or a semimc.1al.

A crystal can be an insulator only if the number of valence electrons in a

primitive cell of the crystal is an even integer. (An exception must be made for
elcctrons in tightly bound inner shells which cannot be treated by band thc-
ory.) If a crysta1 has an even number of valence electrons per primitive cell, it
is necessary to consider whether or not the bands overlap in energy. If the
bands overlap in energy, then instead of one filled band giving an insulator. we
can have two partly 6l1c(rbands giving a metal (Fig. 11).
The alkali metals and the noble metals have one valencc c1cctnm per
primitive cell. so that the)' have to be metals . The alkaline earth metals have
two valence electrons per primitive cell; the)' could be insulators, but the bands
overlap in energy to give metals, but not very good metals. Diamond, silicon,
and gennanium each have two atoms of valence four, so that there arc e ight
valence electrons per primitive cell ; the bands do not overlap, and the pure
crystals are insulators at absolute zero.
 7 Energy Band. 195

SVMMARY

• ne solutions of the wave equation in a periooic lattice arc of the Bloch form
~(r) = eilc . r u~(r), where uJr) is invariant under a crystal lattice translation.
• nere are regions of energy for which no Bloch function solutions of the wave
equation exist. These energies form forbidden regions in which the wave--
functions are damped in space and the values of the k's arc complex, as
pictured in Fig. 12. 111e existence offorbidden regions of energy is prerequi-
site to the existence of insulators.
• Energy bands may often be approximated by one or two plane waves: for
example, t/tAAx) 5': C(k)fth + C(k - C)el{k - G )x ncar the 7.Qne boundary at lG.
• The number of orbitals in a band is 2N. where N is the number of primitive
cells in the specimen.

Problems
I. SqtUJre lnttice. free electron energiu. (a) Show lOr a simple square lattice (two di-
a
mensions) that the kinetic energy a free electron at a (.'()rner of the (irst zone is
higher than that of an electron at midpoint of a side face of the wne by a factor of2.
(b) what is the (.'Orresponding fuetor for a simple cubic lattice (three dimensions)?
(c) What bearing might the result of(b) ha\'e on the oonducti\'ity of di\'a1ent metals?

2. Fru electron energies in reduced wne. Consider the free electron energy bnnds of
an fcc crystallatlice in the approximation of an empty lattice, but in the reduced zone
scheme in whidl allie's are transformed to lie in tile (irst Brillouin zone. Plot roughly
in the {Ill] direction the energies of all bands up to six times the lowest band energy
at the zone boundary at Ie = (2m'a)(I.U). Let thiS be the unit of energy. TIlls prob-
lem shows why balld edges need not necessarily be at the zone center. SC\·cml of the
degeneracies (band crossings) will be remo\·ed when account is taken of the crystal
potential.

3. Kronig.Penney model. (a) For the delta·function potential and with P ~ I, find at
k .. 0 the energy of the Iov,'cst energy band. (b) I<'or the same problem find the band
gap at k "" wla.

4. Potenlial energy in the dianwnd structure. (a) show that for the diamond structure
the I<'ourier oomponent U<; of the cry:o;tal potential seen by all electron is equal to zero
for G = 2A, where A is a basis ,·ector in the reciprocal lattice referred. to the com-en-
tional cubic cell. (b) Show that in the usual fh~t-on:ler approximation to the solutions
of the W""d,·e equation in a periodic lattice the energy gap \'alli~hes at tile zone bound-
ary plane normal to the end of the \'cdor A.
'" (

0.27

~• -... ,
. 0."

T ----.
" ,
.,i, O~ E",,'lY gap
,,
O~ 1 0 0."
-- /
0.01
ItMil""'>' part or I<IC
."0." ." 0."
Rc.l pan or l/G
0.51
'"
rlgUn: u.
In the energy gap tht.n a1st solutions ofttw. wave eqll8lion roc- romple!< values of the
wa\'e\~or. At the boundary of tho. fil"$t ZOIlt' the real pari of the WI~tttor Is iG. 1be imaginary
part of I: in the gap is plotted in the: apprmimation oflwo plan" """"., lOr U - 0.01 ""C~I2m. III
an infin ite unbounded <.T)'HaJ the wa'~r musl "" retJ, or ebe the amplitude win increase
wlthoullimit. But on • wr£.,e or at aju1)dion then can nist 5OIuliooli with complex \\IltYe\-ect<ll".

·S. Complex ICtIVI!VI!ctor. in 'he enugy gap. Find an exp.-ession (or the imaginary part of
the wave\'edor in the energy gap at the boundary of the first I3rillouln rone, in the
approximation that led to EQ. (46), Gil'e the result for the Im(k) at the center of the
energy gap. lhe result for small Im{k) is
(h~l2rn)!Im(k)J2 _ 2rnU 2/f1 2C t .

The fonn as plotted in Fig. 12 is of importance In the theory of Zener tunneling (rom
one oond to another in the presence of a strong electric neld.

6. Sqll6re lemice. Consider a sfluare lattice in tw() dimensIons with the crystal potential

U(,;.,,) - - 4U cos(21t:tla) cos(2wyla) .

Apply the centraJ equation to find approximately the energy gap al Ihc comer point
(wla. wIll) ofthe Brillouin zone. It will suffICe 10 soh'e a 2 x 2 determinantal equation.

References
J. M. lim.n, Principlt$ of the theorv of w1Wh, C.mbndge, 2nd ed., J97'2.
N. F. Moll and H. Jontc'S, TMo'll of the p~ie$ of tlltuW and alloy" Oxf~, J9J6. (Do..-er
pIIplTback reprint)
(Further rd'erencn 011 band theory an- gh"efl 8' the end of Chapter 9.)
 8
Semiconductor Crystals

BAND GAP 199

EQUATIO,,·S OF MOTION 203
Physk:a1 derivation of lil; ==- F 20:;
Holes 206
Effective nlass 209
Physical intCl"pretalion of the effective man 210
Effective masses in semiconductors 212
Silicon and germanium 214

INTRINSIC CARRIER CONCENTRATION 216

Intrinsic mobility 220

IMPURITY CONDUCTIVITY 221

Dunur states 222
Acceptor states 224
1llermal ionhation of domini and accepton 226
THERMOELECfflIC EFFECTS 221
SEMIMETALS 228

SUPERLATIlCES 228
SUMMARY 231

PROBLEMS 231
I. Impurity orbits 231
2. lonb:alilm (lr donotS 231
3. Hall effect with tv."\) carner types 231
4. Cyclotron resonance 232
S. Magneloresislanoe with two canier types 232
REFEREf\;CES 232

NOT£., 1be discussion of carrier orbits in applied fields is continued in Chapter 9.

Amorphou~ semiconductors are treated in Chapler 17. Junctions and barrierS are
treated in Chapter 19.

b
 Sb

.,
<

i
c------""--------,L---------------------' I~ ' ~
I
•
~
-HI'· d

_ 10"

-JO"

eo, I""",J

--~------------------------------ ,.,
Fixure I Carrier COf"ICftltnti.omlOr metals. semimetals. IlJ1d semiconductors. 11le W!mic(lllductOl'"
1'lII~ may be .,..tended upward by fncrca$ing the Impurity C'OIlCIffil
""III. and the range <:an be
extended downward to merge eventual ly with the Ins"L1ta" ~....
 CHAPTER 8: SEMICONDUCTOR CRYSTALS

Carrier concentrations representative of metals. semimetals. and semicon-

ductors arc shown in Fig. 1. Semiconductors are generally classified by their
electrical resistivity at room temperature. with values in the range of 10- 2 to
l(f ohm-cm. and strongly dependent on temperature. At absolute zero a pure.
perfect crystal of most semiconductors will he an insulator. if we arbitrarily
define an insulator as having a resistivity above 10'· ohm-em.
Devices b.ued on semiconductors include transistors. switches. diodes.
photovoltaic cells. detectors. and thennistors. These may be used as single
circuit clements or as components of integrated circuits. We discuss in this
chapter the central phYSical features of the classical semiconductor crystals.
particularly silicon. gcnnanium. and gallium arsenide.
Some useful nomenclature: the semiconductor compounds of chemical for-
mula AB. where A is a trivalent clement and B is a pentavalent clement. arc
called HI-V (three-five) compounds. Examples arc indium antimonide and gal-
lium arsenide. Where A is divalent and B is hexavalent. the compound is callt. d
a II-VI compound; examples are zinc sulflde and cadmium sulfide. Silicon and
germanium are sometimes called diamond-type semiconductors. because they
have the crystal structure of diamond. Diamond itself is more an insulator
rather than a semiconductor. Silicon carbide SiC is a IV-IV compound.
A highly purified semiconductor exhibits intrinsic conductivity. as distin-
guished from the impurity conductivity ofless pure specimcns. In the intrinsic
temperature range the c1ectrical properties of a semiconductor arc not essen-
tially modified by impurities in the crystal. An electronic band scheme leading
to intrinsic conductivity is indicated in Fig. 2. '[be conduction hand is vacant at
absolute zero and is separated by an energy gap Eg from the filled valence band.
The hand gap is the difference in energy between the lowest point of the
conduction band and the highest point of the valence band. The lowest point"in
the conduction band is ca1led the conduction band edge; the highest point in
the valence band is called the valence band edge.
As the temperature is increased. electrons are thennally excited from the
valence band to the conduction band (Fig. 3). Both thc electrons in the conduc-
tion band and the vacant orbitals or holes left behind in the valence band
contlibute to the electrical condudivity.

HAND GAP

The intrinsic conductivity and intrinsic carrier concentrations are largely

controlled by E/koT. the ratio of the band gap to the temperature. When this
ratio is large, the concentration of intrinsic carriers will be low and the conduc-
tivity will be low. Band gaps of representative semiconductors arc given in
Table L The be~ alues of the band gap arc obtained by optical absorption.

19-..
..
Figure 2 Band scheme for intrinsic oondudlvity in 11 $cmic:on.luctor. At 0 K the conduct";I)' is
zen) bemuse all stales in th., "alence baud are 6111.,d and all states in 1m, couduction band are
..aCILnt. As In., tempe .... lure is illl're'.ued. dectrolls are tho:mlal ly eM:itl-'<.i from th.: \1I\(,n« bond to
th., «>nd"dion band, where th,,), become mobl~ .

.
" "

V V
,
" 1/
V
'/ V , /
V •
"
, .,
"
V !! V
10 "
V
,. / J" I '"
-
HI'
V..
'""" ~ '" '" "" '" '" "" liS
.'" 4"- -,

(.j ",j
Figu .... 3 Intrinsic electron Wt>eeulration as • function 01' temp.' .... '..." fur (011) germanium and
(h) siliron. Under inlriru;jc conditions the holerotrtnlralion ISlOquallo lilt' eI..aronroncentralion.
The In triruic con«>n trahoo at a gh-en temperature Is higher In Ce than in SI bea.use the energy gap
is IW"1"\)Wt'r in Ce (0.66 eV) than in Si (1 . 11 e\'). (Afle.- W. C Dunlap.)
 (
• Srmkond..dor Cf")P'aU
""
Table J Energy gap between the valence and COIlduction bands

- (i ,. indi~ect gap; d .. direct !,?!p)

EC",N E.. .. v

-c.",",

DialOond
Si
Ge
c.p OK

5.4
1.17
0.744
300 K

I.ll
0.66
Crystal

HgT"
PbS
PbSe
G.p

d
d
OK

- 0.30
0.286
0.165
300K

0.34-0.37
0.'"
aSn d 0.00 0.00 PbTe 0.190 0.29
InSb d 0.23 0.17 COS d 2.582 2.42-
''''', d 0.43 0.36 CdSe d 1.840 1.74
I.P
G,p ,
d 1.42
2.32
1.27
2.25
COTe
Z.O
d 1.607
3.436
I ."
3.2
C..A, d 1.52 1.43 ZnS 3.91 3 .•
GaSh d 0.81 0.68 SnTe d 0.3 0.18
AISb 1.65 1.• Aget 3.2
SiQl1ex) 3.0 Agl 2.'
T. d 0.33 CuiO ,/ 2. 172-
'7..nsb 0.56 0.56 TiO~ 3.00

"lIg'Te is • temilllt:tal; the banos o~'t'r\ap_

or
TllC threshold continuous optical absorption at frequency wI( determines
the band gap EI( = flw,.. in Figs. 13 and Sa. In the direct absorption process a
11hol011 is absorbt. .d by the crystal with the (,'r cntion of an electron and a hole.
In the indired absorption pl'OCeS5 in figs. 41, and 5b the miuilnum energy
golp of the band structure involves electrons and holes separated by a substan-
tial wavevcctor kc. Here a dire<:t plloton transition at the energy of the mini-
mum gap cannot satisfy the requirement of conservatioll of wavevcctor, be-
cause photon wavcvcctors are negligible at the energy range of interest. But if
a phonon of wavcvcctor K and frequency 0 is created in the process, then we
can have
\.:(photon) = k.. + K e O; flw= E,,+ flO.
as required by the conservation laws. lne phonon energy fin will generally be
muell less than E ll: a phonon even of high wavevector is an easily accessible
source of crystal momentum because the phonon energies are characteristkally
small (- 0.01 to 0.03 eV) in comparison with the energy gap. Hthe temperature
is high enough that the necessary phonon is already thermally excited in the
crystal. it is possible also to have a photon absorption process in which thc
phonon is absorbed.
"2
Cln!>TAI . wrfll J)JRt:cT e ....' ow!>...... wrm t:.:onu:cr GAP

A1-""'VIion AIl.." pt;""

0"....1 of lnoJl,c:d
photon ! . . ...iI;""
.'"

' •.
11~~.,,, "K"' I:¥ ft .. _
':" t- An t::....
I'hoIuol '""-"1!Y ~ .. __

Figure " (}pI:kal ahso'l,tion in Po"'" insulators HI absolute zero. III (a) tN;, thrc)hold ddnmil1C$lhe
c llcrgy gap 11$ E. - '11"0_ In (Ill the optic:1IJ absorption Is wc..kr nt:llr the threshold : N
flw - £~ + fiO R photon is obliOrhed with the creation of Ihn-e. parlic\es: II frt.~ electron, a free
hole , Bnd a phonon of energy fin. III (h) the energy £._ marks the thredXlId for the creaUon oia
free electron and II &ee hole, wilh no phonOOI lo\'ol....,d. Such II transition Is CllI led "ertlcal; it is
similar to Ihe direct transition in (a). 1'he!,;e plots do not ~how absorptioo. linef thaI somelimu are
sl:t:n tying ju51 to the tov,. encrgy stoo of lht' th.t."hold. Such lint'S are dut' to thecreatioll of a bound
electron·hole pair, called au eliCUOII.

., ,,,
Fi8t1ff S In <al the lat.......... pOint of the c:.vnduclion band O<'C'Urs althe Same ''alut' of" '" the highe$!
point of the ''alene.. band. A direct optical!r.ulIition isdmwn vertically with no lignificant cha¥ J
Ie. bt."ClIuse the nbrorbcd photon has a very ,ma ll ....'a'"t:\·ector. n.e threshold r"-'fIuency "'. fOr
absorpt ion by tho;, direct 1r'3l11ition determinel the energy gap E.. - "1»•.
TIM: Indirect t...ulIition in
(b) ""'01",,1 both a photon and a phonon because the band ed~ of the eonduction and ,.'k-nce
bands are widdy scpouated in Ie spKC. 1'he thr."hokl en....gyfor the indirect process in (b) Is grnter
than the Irue band gap. n.e absorption thremold lOr the indired: tran.ilion between lhe bond
edges is lit ~"" .. E. + fin. wht'l'e n i5 thf! frequency of lin emiUI!<! vltonon of w.\vevt:dor
K a - k".. At higher tem~ .... tures phonons art, already present; If. phonon Is absorbt.d along with a
pho(on, t},., threahold CDeTb'Y 1$ "I.> ..
E. - "n.
/1.'01.: n.e rlgUre..haws only the threshold tnlnJi-
110m . T""",itlonl occur generally between almost all points of the two bands lOr .... hictl the
wn.,........tors and t'nerg)' ClIn be COIISCO·ed.
 8 S"micond"cfor Crlld"!' 203

'~'r--------------------------,

,,
, ,~

l!
i
l<
"p I,-

"I,-

,.,
" ~
"
... ineV_

rLgU~e 6 Optical absorption in pure indium antimonide. InSb. The transition is direct lK...:auSt'
or the Brillouin mne , \( - O. Notice the
hoth conduction and valenc.o. barn:! cdgtc'$ are at the cente~
sharp th~cshold . (Afle~ C. W. Cobeli and H. Y. Fan.)

ne band gap may also be deduced from the temperature dependence of

the conductivity or of the carrier concentration in the intrinsic range. The
canier concentration is obtained from measurements of the Hall voltage (Chap-
ter 6). sometimes supplemented by conductivity measurements. Optical meas-
urements determine whether the gap is direct or indirect. ne band edges in
Ce and in Si are connected by indirect transitions; the band edges in InSb arc
connected by a direct transition (Fig. 6). The gap in aSn is direct and is exactly
ZCro; HgTe and HgSe are semimctals and have negative gaps- the bands
overlap.

EQUATIONS OF MOTION

We derive the equation of motion of an electron in an energy band. We

look at the motion of a wave packet in an applied electric field. Suppose that the
wave packet is mar' up of wavefunctions near a particular wavevcctor k. ne
 (
group velocity is vI! '= dwldk. The frequency associated wh .. a v.'8vcfunction of
energy E is w = Eill, and so
0' (I)
TIle effects o(tile crystal on the e1cctron motion arc contained in the dispersion
relation E(k).
The work & done on the elcctron by the elcctric flCld E in the interval
fit. is
& "" -eEv, 8t . (2)
We observe that
& "" (dtldk)& "" flO, Bk (3)

using (1). On comparing (2) with (3) we have

Ek c -(,Elh", • (4)
whence IldkJdt "" - eE, the same relation as for free electrons.
We may write (4) in tcnns of the external force F as

(5)

TIlis is an important relation: in a crystal hdk/dt is aJval to the external force on

the electron. In free space d(mv)ldt is CClvaJ to the force. We have not over-
thrown Newton's scconcl law of motion: the electron in the crystal is subject to
forces from the crystal laltke as well as from extemaJ sources.
The force tcnn ill (5) also includes the Lorentz force on an electron in a
magnetic field , under ordinary conditions where the magnetic field is not so
strong that it breaks down the band structure. 11lUS the equation of motion of
an electron of group velocity v in a constant magnetic field B is

,II< e , dI<
(ees) h- ::o:--vx B (SI) A - == -ev x B (6)
dt , dt

wbere the right-hand side is the Lorentz rorcc 011 the electron. Witb the group
velocity 'IV '" grddkE, the rate of change of the wavevcctor is

(ees) (51) (7)

where now both sides of the equation refer to the coordinates in k space.
We see rrom the vector cross-product in (7) that in a magnetic field an
electron moves in k space in a direction nonna! to the direction or the gradient
of the e nergy f , so that the e lectron moves on a surface or conslant energy.
 or
The value of the 1 Jection ko Ie on B is constant during the motion . The
motion in k space is on a plane normal to the direction of B, and the orhlt is
or
defined by the intersection of this plane with a surface constant energy.

fhyrical Derivation of ~k = F
We consider the Bloch eigenfunction 4/Jo.. belonging to the energy eigen.
vdlue t:k and wavevector k:

'" = Lc C(I< + G) o>p[;(1< + G) • d (8)

The expectation value of the momentum of an electron in the state k is

Pd = (I<[-;AV[k) = Lc A(I< + G~C(I< + G~' = h(l< + Lc G[C(I< + G)I~

(9)

using L[C(k + G)[2 == 1.

or
We examine the transfer momentum between the electron and the lat-
tice when the state Ie of the electron is changed to k + t.k by the application of
an external force. We imagine an insulating crystal electrostatically neutral
except for a single electron in the state k of an otherwise empty band.
We suppose that a weak external force is applied for a time intervdl such
that the total impulse given to the entire crystal system is J = IF dl. If the
conduction electron were free (m* e m), the total momentum imparted to the
or
crystal system by the impulse ","'Quid appear in the change momentum of the
conduction electron:
J= ap.". = ap"1 = ~t.k . (10)
The neutral crystal suffers no net interaction with the electric Held, either
directly or indirectly through the free electron.
If the conduction electron interacts with the periodic potentiallJf the crys-
tal lattice, we must have
J= aptat 11:1 aPlat + ape! (11)
From the result (9) for Pel we have

"Pd = Mk +L AG[(V,IC(k + G)I~ . "k] (12)

c
The change aPIa! in the lattice momentum resulting from the change or
or
state the electron may be derived by an elementary physical consideration.
An electron reflected by the lattice tran5fe~ momentum to the lattice. If an
incident electron with plane wave component of momentum ~k is reflected

L
 with momentum Ii(" + G), thc lattice acquircs the momentum -IiG, as re-
quired by momentul11 conservation. The momentum transfer to the lattice
when thc state th goes over to th+d. is

6",.. ~ -,2: CI(V,IC(I< + C)I' . 6k) (13)

c
as the portion
V,]C(I< + C)I'· 6k (14)
of each individual component of the initial state is reflected during the state
change die.
"The total momentum change is therefore
(15)
exactly as for frcc electrons, E(I. (IO). Thus from the definition of J. we have
IidkJdt = F • (16)
derived in (5) by a different method. A rigorous derivation of(l6) by an en tirely
differcnt method is given In Appendix E.

HoieB
111C properties of vacant orbitals in an othcl'Vlisc filled band arc important
in semiconductor physics and in solid state electronics. Vacant orbitals in a
band are commonly called holes. A hole acts in applied electric and magnetic
fields as if it has a positive charge +e. 11lc reason is given in five steps in the
boxes that follow.

1. ",. = -k.,. . (17)

The total wavcvector of the electrons in a filled band is zero: Ik = O. This
result follows from the geometricaJ symmetry of the Brillouin zone: every
fundamentaJ lattice type has symmetry under the inversion opera tion
r _ - r about any lattice point; it follows that the Brillouin zone of the
lattice also has inversion symmetry. If the band is filled all pairs of orbitals
" and -k are filled. and the total wavC\lcctor is zero.
If an electron is missing from an orbital of wavcvcctor k,., the total
wavC'Vcctor of the system is - k.,. and is attributcd to the hole. This result
is surpl;sing: theelcctron is missing from k., and the position of the hole is
usuaJly indicated graphically as situated at k,., as in fig. 7. But the true
waycveclor Ie,. of the hole is -Ie.. which is the wavcvcctor of the point G
if the hole is at £ . Tlle wavcyector - It., e nters into selection rules for
photon absorption .

. .
 •

,-
,
, c

Figure 7 Absorption c:( 1\ photon of e"ergy f,(i) and neWigible wavevl'<:tor takes an electron from E
in the filled ~'1lIence '-lid 10 Q In the wndUdion hand Irk. was lhe Wl\\'C"ector mlhe electron lit
E. it Il«OffiCS the wan..,,'t'dor oflhc electron at Q. 'The total WlI\'cvedor of the .'llience band aftcr
the alw:wption is -k", and Ihis i. the Wln'l~'-ector we must ascril>e to the hole if we describe the
\U..-.ce hand a.'I occupied hy one hole. 'rhus k. - - k.; the wlw",..,.,torofthe hole is the """'"' as the
wa\l~-ectoror lhc electron which remains a' C. For the enlire s>'Slem the tola1 WlI\'evector-afier the
absorption of the photon is k. + Ie., - 0, so thai Ihe lotal wavcv<."Ctor is UI>changed by the a1JSOrp-
lion d the photon WId d ie creation of II free " Iectron am.! free hole.

The hole is an alternate description or a band with one missing dec-

tron, and we either say that the hole has w8vevector - .... or that the band
witb one missing electron has total wavcvcctor - k.,.
2. (18)
Let the zero or energy of the valence band be at the top of the band. The
lower in the band the missing electron lies. the higher the energy of the
system. nle energy of the hole is opposite in sign to the energy of the
miSSing electron, becausc it takes more work to remove an electron from
a low orbital than from a higb orbital. Thus if the band is symmetric, I
E.,(1:,,) = EeC- lc,,) = -E/.(-1:,,) = -E/,(kh). We construct in fig. 8 a band
scheme to represent the properties of a hole. This 'bole band is a helpful
representation bceause it appears right side up.

LBands are always symmetric under the in\l:nion 1c .... - I.: if the SpilKlrbit interadion Is ne-
glected. Even .... ith spin.()rbit interaction. bands arc a1Wl1)'S symmelr\c if the ~T)'stal sirudure
petrn illthe inv(.-nion operation . Withollt a c.mt.!rofs)'Imne try. bll t wilh spln-orbit interaction. the
bi.nds are symmetric if we comp.are sIIbban.b for which the spin direction i. reverseoJ:
f(l. f) - ..(- 10::. l)t " e QTS. Chapter 9.
 (

HgllTe 8 The upper ho.lf of the figure sl)Qw1 the hole baod thaI I;mu l. res the dynamics of. hole,
constructed by In.-enion r.:ilhe n.~nce t..nd in the origin. The ..... e ..ector a"d fllefgy clthe hole
are equal, but Of.'PO"ile in sign. to the WIovt:\'«'!Ot" and energy of the empty electron orbital In the
•..k'f1tt oolld. W... (10 nooI ••how thcdi'l'osUion of the ck"l1run removed Crolll the .'alen«' h.'l\d at 11,.

3. (19)
The velocity of the hole is equal to the velocity of the missing e lcctnm.
from Fig. 8 we sec thai V'i./,(k,,) = Vlfc(k,), so that v,,(k,,) ::::: v"(k,,).

•• (2()
We show below that the effective mass is inversely proportional 10 the
curvature cPdclJil. and for the hole band this has tht: opposite sign to thaI
for an e lcctron in the valence band. Ncar the lop of the valence band m" is
negative, so that m..,
is positive.
dk, I
5. IiTt ::: e(E + -;vJ. ,,8) (21)

This comes from the equation of motion

(CCS) It dk,. "" -...IE

(It "\
+ ~v
c"
II B) (22)

that applies to the missing eloctron when we substitute - ).;,. for k~ and V"
for v~. The equation of motioll for a hole is that of a partidc of positive
charge l!. The positive charge is consistent with the e lectric current car-
ried by the valence band of Fig. 9: the corrent is carried by the unpaired
electron in the orbital C:
j = (-e)v(C) = (- e)l - v(E)] = ~(E) • (2.3)
which is just the curI"Cnt of a positive charge moving with the velocity
ascribed to the missing elcctron at E. The currcnt is shown in Fig. 10.
 (
•
-c.
---:?.,{>'..,,--- '. ---..4!.';--- •.

,• J
,

,.) 0' .)

a,
Fiprc 9 (a) ....1 I .. 0 aU slales are fiUed noepl F at Ihe lop d the bAAd, the "eloeity 1>, Is :r.ero
F because dtldl:, - O. (hI An electric field E, is applied in the ~:r. dirt-dion. The ron:e on the
electrons Is in the - It, direction and all electrons make transitions together in the - k•.t\iredipn,
ffiOYing the hole '0 the stale E. (el After a further ,n'ervallhe electrons mO\'c tanher along in k
spaee lind the hole ;s now at D.

Figure 10 Molioo of e"-'droos In the conduclioll oond and holes

in the valence I.,>d in the ek<:tric foeld E. n.c hQlc and electron
drift "clodlles ~ in opposite direction., h"t their tledric cur-
renb arc ill the ",me dirL'Chon, th.: dircdlOll of the ck-ctric fleW .

Effective MaSl
When we look at the cncrgy-wavcvcctor relation € "" {fl2I2m)1..!J. for free
electrons, we sec that the coefficient of P determines the curvature of E versus
k. Tunlcd about. we can say that Um, the reciprocul mass, determines the
curvature. For electrons in a band there can be regions or unusually high curva-
ture ncar the band gap at the zone boundary, as we sec rrom thesolulions orthe
wave Ctluation ncar the wile boundary. lfthc energy gap is small in comparison
with the rree electron energy A at the boundary, the curvature is enhanccd by
the ractor AlE" and the reciprocal mass is enhanced by the same raclor.
In semiconductors the band width, which is like the rrec electron energy,
is ortJle order oH.o eV, while the band gap is orthe order orO.2 to 2 eV. Thus
the reciprocal mass is enhanced by a fuetor 10 to 100, and the effective mass is
reduced to 0. 1-0.01 of the rrec electron mass. These values apply ncar the
band gap; as we go away rrom the golP the curvatures are likely to approach
those or rree electrons.
'"
To summarize the solutions of Chapter 7 for U positive, an electron ncar
the lower edge of the second band has an energy that may be writtel) as
m)m = 1I{(2,vU) - IJ . (24)
H ere K is the wavcvector measured &om the zone boundary, and m" denotes
the effective mass of the electron ncar the edge of the second band. An electron
ncar the top of the first band has the CI1Crgy
m,/m = l1{(WU) + IJ . (25)
The curvature and hence the mass will be negative ncar the top of the first
band, but we have introduced a minus sign il110 (25) in order that the symbol m i.
for the hole mass will have a positive value-see (20) above.
The crystal docs not weigh any less if the effective mass of a carrier is less
than the free electron mass, nor is Newton's sec()nd law violated for the crystal
taken as a whole. ions p lus carriers. The important point is that an electron in
a periodic potential is accelerated relative to the lattice in an applied electric or
magnetic field as if the mass of the electron were equal to an effective mass
which we now define.
We diffe rentiate the result (I) for the group velocity to obtain

.!!5i = II_ '~=II_ ,( cf2f: dk) (26)

dt dk dt dP d~

We know from (5) that dkJdt = Flh, whence

dVg = (_'_ d") . 112 dv

0' F = d 2ddP (27)
dt 1i2 dP F, dill

If we identify 1i2/(d2ddP) as a mass, then (27) assumes the fonn of Newton's

second law. Wc define the e ffective mass m* by

(28)

It is easy to gcnera1ize this to take account of an anisotropic energy surface,

as for electrons in SI or Ce. We introduce the components of the reciprocal
effective mass tensor

dv. ~ (_
') F (29)
dt m·~~'"

where 1'. "arc Cartesian coordinates.

PlIysicallnterpretation of the Effective Masa

How can an electron of mass m when put into a crystal respond to applied
fields as if the mass were m*? It is helpful to thin l • the process of Eragg
reOcction of electron waves in a lattice. Consider the weak interaction approxi-
 r COO
I
I,
,,
I
,,
I

v.,
'"
F"tgllre 11 Explanation fA nt.-gal;n efICd;~'e mau<.'S ...hichOOOJr .......... Brillouin zone bounJary. In
(a) the energy of the elfttron beam Indd"_'ll on • thin U'}"ItaI is !lightly too low 10 satisry I~
rondilion for Bragg rene...'tion and the IJnm Is tnunmilled through !he CI)1taL The appUc.lion all
muo.lI ~"Oltagr across the grid UtAy, as in (b). Cl\U$e the Bragg condition to Ix: .... tid'.ed, and the
dectron \.learn will then he rellected rrom the IIppropriate ~I of crystal planes.

mation treated in Chapter 7. Ncar the bottom of the lower band the orbital is
represented quite adC(luately by a plane wave exp(ikx) with momentum f1k; the
wave component cxpli(k - G)x} with momentum h(k - G) Is small and in-
creases only slowly as k is increased, and in this region m· ... 1n. An increase in
the reflected component cxp[i(k - C)XI as k is increased represC'n ts momentum
transfer to the electron from the lattice.
Ncar the boundary the rcflcctcd component is quite large; al the boundary
it becomes equal in amplitude to the forward component. at which point the
eigenfunctions are standing waves, rather than running waves. Here the mo-
mentum component /i(-IG) cancels the momentum component /i(iG).
It is not surprising to find negative values for m* just below a zone bound-
ary. A Single electron in an energy band may have positive ur negative e ffective
mass; the states of positive effective mass occur ncar the bottom of a band
because positive effective mass means that the band has upward ctlrvature
(cPddk? is positive). States of'negative effective mass occur ncar the top of the
band. A negative effective mass means that on going from state k to state
k + llk, the momentum transfer to the lattice from the electron is larger than
the momentum transfer from the applied force to the electron. Although k is
increased by llk by the applied electric field. the approach tu Bragg reflectiun
can give an overall decrease in the forward momentum of the electron; when
this happens the cffective mass is negative (Fig. 11).
As we proceed in the second band away from the boundary, the amplitude
of cxp[;(k - G)r} decreases mpidly and m* assumes a small positive value.
Here the incrcase in electron velOCity resulting from a given external impulse is
larger than that which a frcc electron would experience. The lattice makes up
the difference t' ,ugh the reduced recoil it experiences whell the amplitude of
exp{i(k - G)r} ~, .liminished.
 Ifthc c nc rgy in a band dcpends only slightly on k, thcn the effective mass
will be very largc. That is, m*'m J> 1 whcn d'lEJdf<2 is very small. Thc tight-
binding approximation d iscusscd in Chaptcr 9 gives quick insight into the for-
mation of narrow bands. If thc wavcfunctions centered on neighboring atoms
overlap very little, thcn the overlap integral will be small; the width ofthc b and
narrow; and the effective mass large.
The overlap or wavefunctions cente red on neighboring atoms is small for
thc inner or (.'Ore electrons. The 41 electrons of the rare earth metaJs, for exam-
ple, overlap very littlc. TIlC overlap integral dctcrmincs thc rate of quantum
tunneling of an electron from one ion to another. \Vhen thc cffective mass is
large, the electron tunnels slowly from onc ion to an adjacent ion in thc lattice.

Effective Mmses in Semiconductors

In many semiconductors it has been possiblc to.dctc rmine by cyclotron
resonance the fonn of the e ne rgy surfaces of the conduction and vaJencc bands
near the band edges. The de termination of the energy surface is equivalent to a
de te rmination of the effectivc mass tcnsor (29). Cyclotron resonance in a scmi-
conductor is carried out with CCntimctcr wavc or millimctcr wave radiation at
low carricr conccntration.
The current carriers are accelcrated in helical orbits about the axis of a
static magnc tic ficld . The angular rotation frequ cncy we is

eB
(eeSI we :::-- <Ilil (301
m*c

whcrc m* is the appropriate cyclotron effective mass. Resonant abwrption of

cnergy from an rf elcctric field perpendicular to thc static magnetic field
(Fig. 12) occurs when thc rf frCtJucn{.1' is equal to the cyclotron frequ cncy.
Holes and electrons rotate in oppositc senses in a magnctic field .
,"Ve consider thc experimc nt for m*'m = O. L At Ie = 24 CHz, or we =
1.5 X 1011 S- I, wc have B ""' 860 C at resonance. The line width is determined
by the collision relaxation time 'T, and to obtain a distinctivc rcsonancc it is
necessary that We'T 2: L The mean free path must be long enough to penni! the
average carricr to gct onc radian around a circle bctvicen collisions. Thc re-
quiremcnts arc met with thc use of highcr frcqucncy radiation and higher
magnctic fields, with high purity crystals in liquid helium.
In direct-gap semiconductors with band edgcs at thc centcr of thc Bril-
louin ZOne, the bands havc thc structure shown in Fig. 13. The conduction
band edge is spherical with the effective mass Tn,,:
(311
referred to the valcnce band edge. The valence bands are characteristically
 8 S~ironduetor Coyal<lb 213

Figor" 12 Arnmgemen l of fields in • (.,....-,Iot ron r..sonallCe "XIl".iment ill il semicondlX1or. 1"b..
seme of the dtclJ1ation
i5 oppo$ite f.... electrons ..,d holes.

liJ:I,t hoi...

figu." 13 Simplilk-d V;"'W of the band ..-dg~ struclu." of a di .....,t gap semk-ooductor.

7
...
Table 2 Effective masses of dectrons and holes in !lired gap semioonduclOJrs

Electron Heavy hole Ught hole Split-off hole Spin -orbil

C~tal mJm m."lm mu,lm '"..wIm 4 , eV

InSh 0.015 0.39 0.021 (0. 11) 0.82

I,,,, 0.0"2"i 0.41 0.025 0.08 0.43
fli P 0.073 0.4 (0.078) (0. 15) 0.11
CaSh 0.047 0.3 0.06 (0. 14) 0.80
e,"" 0.066 0.5 0.082 0.17 0.34
CUiO 0.99 0.58 0.69 0.13

threefold ncar the edge, with the heavy hole Ir1r and light hole III bands degCll'
crate at the center, and a band soh split oIT by the spin-orbit splitting A:
~(M) 5!! - 1i'lk'lJ2mhh ; E~(lh) ill - 1i21?/2m'h
(32)
~(soll) .5! - IJ. - 1i,2k'l/2m.",

Values of the mass parameters arc given in Table 2. 111c fOl'105 (32) arc only
approximate, because even close to k ;;; 0 the heavy and light hole bands are
not spherical-sec the discussion below [or Cc and Si.
The perturbation theory of band edges (Problem 9.8) suggests that the
elcctron effective mass should be proportional to the band gap, approximately,
for a direct gap crystal. We use Tables I and 2 to find the fairly constant values
m/{mEJ = 0.063, 0.060, and 0.051 in {eV)- 1 for the series InSb. InAs, and
InP, in agreement with this suggestion.

Silicon and Cermanium

TIle conduction and valence bands of germanium are shown in Fig. 14,
based on a combination of theoretical and experimental results. TIle valence
band edge in both crystals is at k = 0 and is derived from V3!2 and Pm states of
the free atoms, as is clear from the tight-binding approximation (Chapter 9) to
the wavcfunctions.
The P3I2. level is fourfold degene rate as in the atom; the four states corre-
spond to m, vaJues ± t and ± i. The PII2 level is doubly degenerate, with
111, = ± i . 11Je P3I2 states are higher in energy than the Pm states; the e nergy
difference !J. is a measure of the spin-oruit interaction.
TIle vaJence band edges are not simple. lloles ncar the band edge arc
c1laracterh:ed by two effective masses, light and heavy. These arise from the
two bands fonned from the PM le\·e! of tlle atom. 1111~re is also a band fonned
from the PII2 level, split off from the P312 level by the s()in-orbit interaction. The
energy surfaces are not spheric-d.!. but warped (QTS , p 27 1);
(33)
 .;r-------,,-----,------~--------_,

~""IH "'.V
, 5.4 e\l
~--~
".
F.', - O.9<:V

3.8 c

-,
•~ -3

-,
-.
-,
-,
-,
-w •
-u

- "L,>.C,,~,,--------------_,JL,_-----------------------cr.c-l
k_ !'G!.(.!..!...!.)
2 2 2
,, -
0
[ 1111 ______~I----_ /1m)
. .. ~11l0)
G

F"",,~ 14 Calculated band structure of I!)enmlll ium, after C. 1'. Fong. The g..·'H.......J lealures are in
~ ~rlTltlll .... im
experiment. The IOOr ~'a!cnce bands are shown In gruy. The fine structureof'
II. vak~ band ed;e is ClU$OO by spin-(lrbil splitting. The energy gaP;$ indirect; the condUC1:ion
band ~ is al the pomt (2,.{oXIII). 'lheconstant crOl:rgy surfaces arou nd this puint are cllipooidal.
,os
(
The choice of sign di~tinguishcs tlle t\.\'O masses. The split-off band ha~ £{k) =
- /1 + AJ....2. The cxpcliments give, in units ,,212m,
5i : A::: - 4.29 ; 181- 0.68 ; ICI- 4.87 ; A = 0.044 eV
Ce: A = -13.38 ; 181 - 8.48 ; Ie! - 13.15 ; A = 0.29 eV
Roughly. the light and heavy holes in gcnnanium have ma~scs 0.043 m and
0.34 m; in silicon 0.16 m ami 0.52 m; in diamond 0.7 m and 2.12 m.
The conduction band edges in Ge are at the equivalent points L ~f the
Brillouin wne, Fig. 1Sa. Each Land edge has a spheroidal energy surface ori-
ented along a (ill) crystal axis, with a longitudinal mass m, "" 1. 59 m and a
o
transverse ma.'\s mt ::: 0.082 m. For a static magnetic fielcl at an angle with tlle
longitudinal axis of a spheroid, the effective cyclotron mass 'rlt: is
I co.Jle sin 2 0
-- --+-- (34)
~ ~ mIm,

Rcsull~ for Gc are shown in Fig. 16.

In silicon the conduction band edges arc spheroid" oriented along the
equivalent (100) directions in the Brillouin wne, with ma~s parametcrs ml =
0.92 m and m, = 0.19 m, as in Fig. 17a. The band edges lie along the lines
labeled 6. in the wne of Fig. 15a, a little way in from the boundary points X.
In CaA.~ we have A = -6.98, B = - 4.5, lei = 6 .2, 6. = 0.341 eV. The
band structuI'C is shown in Fig. 17b.

INTRINSIC CARRIER CONCEI\'TRAnON

We want the concentration of intrinsic cruTicrs in tcrms of the band gap.

We do tJle calculation for simple parabolic band edges . We first calculate in
terms of the chemical potential Jl the number of electrons excited to the con-
duction band at tcmperature 1'. In semiconductor physics Jl is called thc FeJ'lni
level. At the temperatures of interest ,ve may suppose fOr the conduction band
of a semiconductor that E - Jl » kilT, and the Fermi-Dirac distribution func-
tion reduces 10
(35)
TIl is is the probability that a conduction electron orbital is occupied, in an
approximation valid when f~ <C I.
111c cnergy of an electron in the conduction band is
(36)
where E.. is tJle energy at tile conduction b.1nd edge. as in Fig. 18. Here m" is
the effective ma~s of an electron. Thus from (6.20) the density of states at E is

(37)
 8 Se""«mduclar Cry.tal. 217
(

OJ
Figure 15 Standard labels of lhc symmetry points and .... <:$ofthc Brillouin WIll'!i of lhc iCcand bee
lattices. TIle wile cenh,.." are r . In (a) the boundary point at (2.,./a)( JOO) is X; the boundarypoinl at
(2.,.fll}(l l i) is L; the line!J. runS belwCt.'1l rand X . In (b) the COITCSp:mding srmbo1s are H . P and fl .

0."

'" IT
03 ,

'"
i , 0.2
II
~ 0.20
.~ ~ 1/
~• 0
" V II
" I~ t.:
,.'"
Figure 16 Eff..'Ctiv" <)dotron mass of dectrons in gt'1"ma.
niurn at " K ror magnetic field diTl.'CbOlls in II. 110 plane.
,..
Then: an:: lOur independ<.'1lt mass sph<.'ll)id. ill GO', olle along § I~
c....n III a,i~. but vi<,...wed in the 110 plane tWQ spheroids '=
- -
alW3)'S appear "'(Iu;valent. (A ller Oresselha"., Kip, and
KilteL )
" -"=
- 10 0 10 20 30
Angle in degree> In
4(}
"'W "',.
lIO pia •.., ffUm 10011
00
a>;l,'
,00
'"
•
'. ,
,
.>," ,
~

"• -,
'. -,
-,
-, ,
,. r
Ib,
Figure 17a Con~tanl L.,>erg»' ellipsoods b el~ figure 17b Band struCh'l"e ofC.lAs, aftcr S. C . Louie.
In)IlS in silicon , d",wn for m/ III, _ 5.

l 11c concent ration of electrons in Ihe conduction band is

(38)
f OO (€ _ Ec)112 cxp( - ElknT}dE
f;,

which integrates to give

(39)

The problem is solved for n when Ii is known. It Is useful to calcu late the
equilibrium concentration of holes p. 111c dishibution function flo for holes is
related to tlle e lectron distribution Cunction f~ by f" = 1 - f~. because a hole is
the absence of an electron. We have
I
flo = 1 - -e-"-;,-;-(.-,,.cY~k'--.'::T;-)-:-+--:-1 - -e-",,--;I-;-(,,- _.c.C,
Yk'.,"
-- T;-)-:-+-:-
1
(40)
a CXP{(E - ,u)lknT] ,
provided (J.t - E) ,. knT.
If the holes near the top of the valence band Ix:have as particles with
effective mass mh, the dcmity of hole states is given by

I
DIr(E) = 2"r
(2m"),,
7 (E" - £)112 , (4 1)
 ,-
''"r:;:r I
,.....~~""-'"

= __"", 1
(- )

Figure 18 Energy i>ClIlc lOr statistical cakula-

1;0.'$. The Fern,i distTibul ..... t\Joction is .hown
(IIl l he !>lime seal." for • Icmpt'r .ltu.-c k"T <c E~
1nc ."enni level jJ. I. tal:cn 10 lic ",,,II ""Illin thto
band gap. IliII b- lUI illlrin.ic semieondudOO" If
f - ,. . tI,l.'TI/ - I.

\\tlcre E" is the energy at the valence band edge. Proceeding as in (38) we
obtain
(42)

for the conccntration p of holes in the vaJencc band.

We multiply together the expressions fol' nand p to obtain the equilibrium
relation, with the cnCfb'Y gap £8 = Ec - Eo.

(43)

This useful result docs not involve the Fcnni level fJ.. It is an expression of the
law of mass action.
We have nowhere assumed in the derivation that the matcliaJ is intrin~ic:
the result holds in the presence of impurities as well. TIle only a55umption
made is that the distance of the Fcnni level from the edge of Loth bands is large
in comparison with kBT. At 300 K the value of np is 2.10 x 10 19 em-II, 2.89 x
1()26 em-II, and 6.55 x 1012 em-II, for the actual band structures ofSi, Ce. and
CaAs, respectively.
A simple kinetic argument shows wilY tile product fip is constant at a given
temperature. Suppose that the C(}uilibrium population of e lectrons and holes is
maintained by black-body photon radiation at temperature T. The photons gen-
crate e lectron·hol.... pairs at a rate A(D . while 8(J') np is the !'ate of the recombi·
nation reaction e +" "* photon. Then
d,Jdt ;;0 A(T) - B('l')np ". dpldt (44)
In equilibrium dnJ(it 0; dpldt = 0, whence lip = A(1·)B(1").
;;0

Because the product of tile electron and hole concentrations is a constant

independent of impurity concentration at a given temperature, the introcluc-
Bon of a small proportion of a suitable impurity to increase ", say, must de-
crease p. 111is r~-ult is important in practice-\vc can reduce the total canier
concentration 11 + P in an impure crystal. sometimes enormously, by the con-
trollcel introduction of suitable impmitics. Sudl a I~uction is called compen-
sation.
In an intrinsic semiconductor the number of electrons is equal to the num-
ber ofholcs, because tile thcrmal excitation of an e lectron leaves behind a hole
in the valence band. l1lUs from (43) we have, letting the subscript i denote
intlinsic and Eg = E" - E",

k,T ) '"
"I ." PI = 2 ( 2n1i2 y,
(m ..m,.) exp( - E/2k B'J') (45)

The intrinsic carrier dcpends exponentially on E,/2k8T, where Eg is the

encrgy gap. We set (39) C{Juai to (42) to obtain, for the Fermi level a... measured
fium the top of the valence band,
exp(2jUkIJT) = (mhlm~r exp{E,/ksT); (46)
po. = lEI: + ik8T In (mJm.J (47)

If Jlih "" m .., then po. = lEI! and thc Fenn! level is in the middle of the fOrbidden
gap. A thorough treatment of the statistical physics of sem iconducton is givcn
in TP, Chapter 13.

l"tnllSic Mobility
The mobility is the magnitude of tile drift velocity per unit electric field :
(48)
The mobility is defined to be positive fOr both electrons and holes, a1t110Ugh
tlleir drift velocities are opposite. By writing Jl.. or po./o for the e lectron or holc
mobility we can avoid any confusion between po. a.. the chemical potential and a.~
the mobility.
The electrical conductivity is the sum of the c1ectmn and hole contribu-
tions: (T = (nepo... + pep..,J , (49)
where nand p are the concenh·ations of e lectrons and holes. In Chapter 6 the
drift velocity of a charge q was found to be v "" qTEJm. whence
(SO)
 , ~,,w,o.,d""lOr Cry.fab "I
(
Table 3 Carrier mobilities at room temperature, in cm1fV-s

Cryslal Electrons Holes Crystal Electrons Hol.,s

• 1800 1200
Diamond eaAs 8000 300
Si 1350 48" GaSb 5000 1000
Ce 3600 IBOO PhS 550 600
lusb BOO 450 PbSc I"'" 930
I, A> 30000 450 PbTc 2500 1000
Inl' 4500 100 Agel 50
AlAs 280 KBr (HJO K) 100
AISb 900 400 SiC 100 10-20
•
The mobilities depend on temperature as a modest power law. The tempera-
ture dependence of the conductivity in the intJinsic region will be dominated
by the exponential dependence exp(-EJ2kB T) of the canier concentration,
Ell. (45).
Table 3 gives experimental values of the mobility at room temperature.
111c mobility in Sl unit~ is expressed in rn 2 N_s and is 10- 4 of the mobility in
practical units. For most substances the values quoted are limited by the scat-
tering of carriers by thermal phonons. nlc '!Ole mobilities typically arc smaller
than the electron mobilities oc"Cause of the occurrence of band dcgenerdC}' at
the vruence band edge at the zone center, thereby making possible interband
scattering pnx.'e sses that reduce the mobility conside rably.
In some crystals, particularly in ionic cr},strus, the holes are essentiruly
immobile and get about only by tllennruly-acti vated hopping from ion to ion.
111e principal cause of this "self-tl'llpping" is the lattice distortion a~sociated
with the Jahn-Tellel' elTect of degenerate states (Chapter 14). The necessary
orbita1 degenel'llcy is much more frequent fol' llOles than fol' e lectrons.
There is a tendency fol' crystals with small energy gaps at direct band
edges to 1mve high values of the electron mobility. By (9.41) smrul gaps lead to
light effective ma~ses, which by (50) favor high nlObilities. 11le highest mobility
observed in a semiconductor is 5 X lOS cm 21V-s in PbTe at 4 K, where tIle gap
is 0. 19 eV.

IMPURITY COl"iDUCTJVITY

Certain impurilies and imperfections drastically affect the electrical prop-

erties of a semiconductor . The addition of boron to silicon in the proportion of
I boron atom to 1<r> silicon atoms increases the conductivity of pure silicon by a
factor of 103 at room temperatu re. In a compound semiconductor a stoichiomet-
ric deficiency of one constituent will act as an impurity; such semiconductors
'"
arc known as deficit semiconductors. TIle deliberate addition of impurities to a
scmironductor is called doping.
We consider the effect ofimpUlitics in silicon and germanium. 111cse ele-
ments crystallize in the diamond structuJ'C. Each atom rOmlS four covalent
bonds, onc with each of its nearest neighbors, con"csponding to the chemical
valence four. Iran impurity atom ofvalencc fivc, such as phosphorus. arsenic,
or antimony. is substituted in the Jaltice in place ora normal atom, there will be
one valence electron from the impurity atom left over after the four covalent
bonds arc established with the nearest neighbors, Ul3t is, after tlle impurity
atom has been accommodated in the structure with as little disturbance as
possible.
Demar Stiltes. The structure in Fig. 19 has a positive charge on the impu-
rity atom (which has lost one electron). Lattice constant studies have verified
that the pentavalent impurities enter the lattice by substitution for noonal
atoms, and not in interstitial positions. Impurity atoms that can give up an
electron are called donors. The crystal as a whole remains neutral because the
electron remains in the crystal
The electron moves in the coulomb potential elf" of the impurity ion,
where f in a covalent crystal is the static dielectric constant or the medium. The
factor Uf takes account of the reduction in the coulomb force between charges
caused by the electronic polarization of the medium. This treatment is valid for
orbits large in comparison with the distance between atoms, and for slow mo-
tions of the electron such that the orbital frequency i... low in comparison with
the frequency Wg correspond ing to the energy Wlp. These conditions are satis-
fied quite well in Ge and Si by the donor electron of P, As, 01" 5b.
We estimate the ionization energy of the donor impUlily. l1le Bohr theory
of the hydrogen atom may be modified to take Into aerount the dielectric con-
stant of the medium and the effective mass of an elcctron in the pcriodiCJX)ten-
tial of the crystal. 'Ole ionization energy of atomic hydrogen is _e4mJ21i2 in
CG5 and -e4 Jn/2(47Tfoft)2 in 51.
In the semiconductor we replace r by r/f. and m by the cfiCctive mass file
to obtain

(5 1) FJ~.,' . ' (51)

~4'7Tft,;)r.r-

as the donor ionization energy of the semiconductor.

111e Bohr radius of the ground state of hydrogen is IIi/mil in CG5 or
47TEflt2/mr in 51 Thus the Bohr radius of the donor is

(CGS) ad = di
m~
2
= (O.53f)
m/m
A (SO (52)

. ,
Figure 19 Charga auociated "ith an arsenic impuril y alom in . ilicon. Arsenic hll$ five valence
electrons. hut silic:on hll$ only four valence ",I«irons. Thus lOu r electrons 011 the arst.'tIic form
telr2hedrall'OV1llcll t bonds ..milar 10 ~ilicon , and the fifth electron is ,1lI,tibble for conduction. n.e
Dl"knic atom is caJ.k>d a dono.- because "h..,., ionired it donates an elcrtron to the coodUCIioll band.

The application to germanium and silicon is complicated by the anisotropic

dfective mass of the conduction electrons. But the dielectric constant has the
more important effect Qn the donor energy because it enters as the square,
whereas the effecti ve ma~s enters only as the first power.
To obtain a general impression of the impurity levels we usc m~ - 0.1 m
for electrons in gennanium and m. "., 0.2 m in silicon. TIle static dielectric
constant is given in Table 4. The ionization energy of the free hydrogen atom is
13.6 eV. For germanium the donor ionization energy Ed on our mode l is
5 meV. reduced with respect to hydrogen by the factor mJm€l "" 4 x 10- 4 •
TIle colTesponding result for silicon is 20 meV. Calculations using the correct
anisotropic mass tensor predict 9.05 mcV for gCl1113nlum and 29.8 meV for
silicon. Observed values of donor ionization e nergies in Si and Gc arc givcn in
Table 5. Rec".I11 that I meV iii! 10- 3 eV. In eaAs donors have Ed - 6 meV.
l 1le radius of the first Bollf orbit is increa~cd by ~mJIlIc ovcr the value

Table" StatU:: relative dieleCIric COn!ltant or semiconductors

Crystal • CrySlll1 •
Diamond
51
•••
11 .7
c..sb
GaA,
15.69
13. 13
Cc
InSb
15.8
17.88
.'"
AISb
10. 1
10,3
I"", 14.55 SiC 10.2
I,P 12.37 7.1

-
CUtO
Table 5 Donor ionizatm energies Ed of pcnta"alent impurities in germanium and
silicon, in mcV

p Sb
s; 45. 39.
Co 12.0 9.6

0.53 A for the free hydrogen atom. The rorresponding radius is

(l60XO.53) "'" 80 Ain germanium and (60XO.53) "'" 30 Ain siliron. 'IneS€! are
large radii. so that donor orbits overlap at relatively low donor concentrations,
compared to the number of host atoms. With appreciable overlap. an "impurity
band" is fonned from the donor states: see the discussion of the metal-insulator
transition in Chapter 10.
'Ine semiconductor can conduct in the impurity band by electrons hopping
from donor to donor. The p rocess of impurity band ronduction sets in at lower
donor concentration levels if there are also some acceptor atoms present, so
that some of the donors are always ionized. It is easier for a donor electron to
hop to an ionized (unoccupied) donor than to an occupied donor atom, so t hat
two electrons wiU not have to occupy the same site during charge transport.
Acceptor States. A hole may be bound to a trivalent impurity in germa-
nium or silicon (Fig. 20). just as an electron is bound to a pentavalent impurity.
Trivalent impurities such a.~ B. AI , Ga, and In are caJled acceptors because they
accept electrons from the valence band in order to complete the covalent bonds
nith neighbor atoms, leaving holes in the band.
When an acceptor is ioni7.ed a hole is f~, which requires an input of
energy. On the usual energy band diagram, an electron rises wh en it gains
energy, wflereas a hole sinks in gaining energy.
Experimental ionization e nergies of acceptors in germanium amI silicon
are given in Table 6. The Bohr model applies Qualitatively for holes jU6t as for
electrons, but the degeneracy at the top of the valence band complkates the
effective mass problem.
The tables show' that donor and acreptor ioni7.ation energies in 5i are com-
parable with kBT at room temperature (26 meV), so that the thermal ioniz."1tion
of donors and acceptors is important in the e lectrical conductivit}' of silicon at
room temperature. If donor atoms are present in considerably greater numbers
than acceptors, the thennal ioni7.ation of donors will release electrons into the
conduction band. The conductivity of the specimen Ihen will be controlled by
electrons (ne~}1tive charges), and the material is said to be n type.
If acceptors are dominant, holes will be released into the valence band and
the conductivity will be controlled by holes (positive charges): the materia] is 1}
 8 Snnicond.. cto~ Cr!flllaU

,
Table 6 Acceptor ionization cncrgiC$ E. of trivalent impul'ities in gcnnanium and
silicon, in mcV

• AI G•
"
Si 45. 57. 65. 157.
Go lOA 10.2 10.8 11.2

Att"4'pIor b...."d .......L

- --- ---- E..
0 , j

Figure 20 Boron h as only three ~ el<'dTOn5; It can C'Omplete III. tetl"'lhed ral bonds only by
taking an electron from a Si-S i bond. leaving behind a hole in tile ~ilicon ,·alene<: band. Thr; positi,·e
hoi", is then available for C'Onduction. 1br; boron .. tom is called an RC«plor because whm Ioni:ted it
acrept~ an dect,on from the vale...,., band. At 0 K the hole is bound: remember t.... t holes f\oo;l.

type. The sign of the Hall voltage is a rough test for n or 1J type. Another handy
laboratory test is the sign of the thermoelectric potential, discussed below.
'The numbers of holes and electrons are equal in the intrinsic regime. 'The
intrinsic electron concentration n. at 300 K is 1.7 X 1013 cm - 3 in germanium
and 4.6 x 1<f cm- 3 in silicon. The electrical resistivity of intrinsic material is
43 ohm-em for germanium and 2..6 x J(fI ohm-cm for silicon.
Cermanium has 4.42 x IlP atoms per cm3. The purification of Ce has
been carried furth er than any other element. 2 'The concentration of the com-
mon electrically active impurities- the shallow donor and acceptor impurities-
has been reduced below 1 impurity atom in 1011 Ce atoms (Fig. 21). For exam-
ple, the concentration of Pin Ce can be reduced below 4 x 1010 cm- 3. The
experimental sensitivity for detection of these impurities is 10' cm - 3 by the
method ofpholothcrmal ionization spectroscopy.3 There are impurities (1-1, 0,

"E . E. Haller, W. L Hansc:n, and F. S. Gou lding, " l'bysic:s 0 ultl"'l'pun: gcrmanium,- Ad, .
Phys 30. 93-135 ( 19EH).
"s M. Kogan and T M. uf~hits. Phys. Statu, Solidi (.) 39. 11 (1977).
".
10"

Figu", iI Tefrlpern.lure dependence ufille r~", c-amer ronc-en l rahon in uil.-:optlre Ce, .ner R. N.
HaiL Th .. net ronct.'nlralion of electrically Idjye impUrities is 2 x )0'" cm- 3 , as ddermi rd by
Hall rodTocicnt me<l5Ureme nt 5. the ... pidomet oCinlrin*ucitation;5 evident allow \'ahlCS .£Iff.
The c:anicr conccn lnttion. ;~ closely corulanl b1:t~n 20 " and 200 K.

Si, C) whose concentrations in Ce cannot usually be reduced below 10 12 _

10 14 cm- 3 • but these do not affect electrical measurements and therefore may
be hanl to detect.

TI, ermlll Ionization of D(mors and Acceptors

The calculation of the equilibrium concentration of conduction elect rons
from ionized donors is identical with the standard calculation ill statistic.'11 me-
chanics orthe thermal ionization of hydrogen atoms ("fr, p. 369). Ifthere are no
acceptors present, the result in the low temperature limit ksT <il: Ed is

In at (noN.,)1I2 exp(-Ej2k BT) • I (53)

with no - 2i,m~k8TI27T1i2'f'2; here Nd is the concentration of donors. To obtain

(53) we apply the laws of ch emical equilibria to the concentration ra tio (e) [N,rV
{Nd], and then set [N,t] = {eJ = n. Identical results hold for acceptors, under
the assumption of no donor atoms.
If the donor and acceptor concentrations are comparable, affairs are com-
plicated and the equations are solved by numerical methods. However, th e law
of mass action (43) requires the np product to be constant at a given tempera-
ture. An excess of donors will increase the electron concentration and d ecrease
the hole concentration; the sum n + 1) will increase. 11le conductivity will in-
crease as n + p if the mobilities are equa), as in Fir "2.
 8 S"micontiuclor Crystal. 227

,
10" 'i

1
10" ~
I
•

Figure 22 Elect rical <;onduct i,·it y and hole concent ration p calculated as a funcl ion of eleclron
concentraliun n for a serniconduclor al a lernpernlu.c such Ih.ol lip '" uf" cm - e. The <:OI"Iducli'·ily
l~s)'.-n"'lClrical aboul n = 1010 cm - 3. fOT n > 10'" the specimen is n type ; fUl" n < lOw, il is P type .
We rowe laken II-. ,.. I'A, fur the mobi litie<.

THERMOELECTRIC EFFECTS

Consider a semk'Onductor maintained at a constant temperature while an

electric Geld drives through it an electric current density jq. If the current is
carried only by electrons, the charge flux is
(54)
where fJ.e is the electron mobility. The average e nergy transported by an elec-
tron is, referred to the Fenni level p..,
(Ec - p.) + ik8T ,
where Ec is the energy at the conduction band edge. We refer the e nergy to the
Fermi level because diffe re nt conductors in contact have the same fe rllli level.
The energy flux that accollipanies the charge flux is
jv = n(Ec - p. + ik8 TX-ILe)E (55)
The Peltier coefficient n is defined by the relation ju :: njq; it is the
energy carried per unit charge. For electrons,
n.. = - (Ec - p.. + ikBT)/e (56)
'"
and is negative because the energy nUl( is opposite to the charge 1111)(. For holes
(57)
where E~ is the energy at the valence band edge. Thus
Ill> = (}.t - Eo> + ikB'l)'e (58)
and is positive. Equations (56) and (58) are the result or our simple drift velocity
theory; a treatment by the BoIt7.Jnann transport equation gives minor nume ri·
cal differences. 4
The absolute thermoelectric power Q is defined from the open circuit
. electric field created by a temperature gradient:
E~Qll"'dT . (59)
The Peltier <:oefficient n is related to the thermoelectric power Q by
l1 =QT. (60)
This is the famous Kelvin relation of irreversible thermodynamics.S A measure-
ment of the sign of the voltage across a semiconductor specimen, one cnd of
which is heated. is a rough and ready wa)' to tell if the specimen is n type o r p
type (Fig. 23).

SE"'UMETALS

In semimetals the conduction band edge is very slightly lower in energy

than the valence band edge. A small overlap in energy of the conduction and
valence bands leads to small concentmtion of holes in the valence band and of
electrons in the conduction band O'able 7). Three of the sem imetais, arsenic,
antimony, and bismuth, are in group V of the periodic table.
'Their atoms associate in pairs in the crystal lattice, with two ions and ten
valence electrons per primitive cell. 'The even number of valence electrons
would allow these elements to be insulators , Like sem iconductors, the semi·
metals may be doped with suitable impurities to vary the relative numbel'S of
holes and electrons. TIleir concentrations may also be varied with pressure, for
the band edge overlap varies \\lith pressure,

SVPERI.ATIICES

Consider a multilayer crystal of alternating thin layers of different compo-

sitions. Coherent layers on a nanometer thickness scale may be deposited by

OR, A. Smith, Stmiconducton, 211d ed.; Cambr idge, 1978; .... Frltzscbl", Solid State Coro ln ull .
9, 1813 (197l). A simple discussioo or Boltzmann t ransport theory is gh'ell in Appr.ndi~ F.
511 . B. Callen , Thermc<lgnamla, 211d ed.; Wile>" 1965.
 (
,,
,.• .d "'\
,,
, "."
I ..
, ""']" \
(\n~)
iJ"
-Q
•
-Q , ~ .J.
-, • ... ""
•
.., ,., .. ....
T.... "pe .... u...,. in K ""
Figure 23 P~hi~r oocfficien t of n and p silicoo .5 II rUllc l ion of temperature. Above 600 K the
...,ecimcns ad as intrinsic S<':.nironducturs_ n.., ru .....·o:s.re c~k"lal~od amI the points are el<lIcrimcn.
tal. (After T. H. Geballe and C. W. Hun.)

Tflblc 7 Electron and hole concentrations in semimctals

Arsenic
Antimony
Bismuth
-
..... in cm -~

(2.12 ± 0.01) X J(jIO

(5.54 ± 0.(6) X 10 111
2.88 X 1017
(2. 12 ± 0.01) X JcPO
(5.49 ± a.m) X JOIIl
3.00 X 1017
Graphite 2.72 X lO'S 2.04 X 10 18

molecular-beam epitax)' or metai-organic vapor deposition, thus building up a

superperiodic structure on a large scale. Systems of alternate layers of eaAs
and CaAlAs have been studied 10 50 periods or more, with lattice spacing A of
perhaps 5 om (SO A). A superperiodic crystal potential arises from the SUPeT-
periodic stmcture and acts on the conduction electrons and holes to create new
(small) Brillouin 7.ones and mini energy bands superposed on the band struc-
lures of the constituent layers. Here we treat the Illotion of an electron in a
superlattke in an applied electric field ; magnetic fields are treated in
Chapter 19.

Blodl Oscillator
Consider a collisionlcss electron in a periodic lattice in one dimension.
with motion normal to the planes of the superlattice. The equation of motion in
a constant electric field pamllel to k is lidkidt =0 -eE or, for motion across a
Brillouin mne with reciprocal lattice vector G :: 2mA, we have hG = h2mA :=
eET. The Bloch rrequency of the motion is w8 ;; 27rIT = eEAlh. The electron
accelerates from k = 0 toward the mne boundary; when it reaches k = mA it
reappears (as by an Umklapp process) at the wne boundary at the identical
point -mA, using the argumeut of Chapter 2.
We l.'Onsider the motion in a model system in real space. We suppose that
the electron lies in a simple energy band of width £0:
£ "" £0(1 - cos kA) (6/)
The velocity in k-space (momentum space) is
v = h - Idslfk = (A&J/i) sin kA (62)
and the position of the electron in real space, with the initial cond ition z = 0 at
t := 0, is given by

z :=: fv dt = fdk v(kXdtldk ) = (A&JIi ) Jllk (- liIeE) sin kA

= (- £c/eEXl.'Os leA - 1) = (- £JeEXcos(-eE.-\ IIIi) - 1) . (63)
This conflrllls that the oscillation frequency in real space is w8 = eEAlIi. The
motion in the periodic lattice is quite d iffe rent from the motion in free space,
for which the acceleration is constant.

Wunnier lAdder
The electrostatic energy of the system is increased by eEA if the initial
position of the electron is displaced by one lattice constant from z = 0 to z = A.
For a displacement of the initial position by n lattice constants, the e nergy
eigenvalues are
£n :: neEA + £0(1 - cos kA) , (64)
This set of solutions defines the Wannier Ladder of energy levels, with adja-
cent level separation tu: :=: eEA = IiW8' The exact solution of the wave equation
gives a continuum of states, with density peaked sharply at the e igenvalues
(64). TIle ladders are obsel'Ved in optical experiments. The Bloch frequency has
not been observed directly (as of mid-l993).
Estimates, Let E = 3 X 104 Vern - I, £0 '" 0. 1 eV, and A = 10- 6 cm; then
WIl = 4.6 X 1013 S- I and the localization length is Ec/eE = 3 nm. For narrow
bands in strong fields it may be possible to localize the electron waverunction
dominantly on a single atom; changing the index n in En will change the localiza-
tion site of the electron.

Zener Tunneling
Thlls far we have considered the effect of the electrostatic potential - eEz
(or - eEnA) on one euergy band; the potential tilts" whole band. Higher
\
bands will also be tilted similarly, creatin~ the possibility of crossing between
ladder levels of different bands. The interaction he tween different band levels
at the same energy opens the possibility for an declron in one band at n to cross
to another band at n'. This field-induced interband tunneling h an example of
Zener breakdown, met most often at a single Junction as in the Zener diode.

SUMMARY

• 111e motion of il w:we packet C'entcK.'<.! lit wavcvector k h de.'lCribet! Ly F =

fldkJdt, where F is the applied force . The motion in real space is obtained
from the group velocity ve = f1 - ' \\t{k).
• The smaller the energy gaP. the smaller is Im·1near the gap.
• A crystal with one hole has one empty electron state in an otherwise filled
band. "Inc properties of the hole are those of the N - 1 electrons:
(a) If the electron is missing from the state of wa.. evector k.", then the
wavevector of the hole is k/, "'" - k." .
(b) The rate of change of Iv. in an applied fi eld requires the assign ment of a
positi\'e charge to the hole: ell = e = - ee'
(c) Ifv. is the velocity an electron would have in the state ke. then the veloc-
ity 10 be a.~ribed to the hole of wave\'eclor k/, = - k." is v/, = ve .
(d) The energy of the hole referred to :r.ero for a filled band is positive and is
£/,(k/,) :. -£(k.,) .
(e) The effective mass of a hole is opposite to the effective mass of an electron
at the same point on the energy band: mil = -me'

Problems
I. Impurit!l orbitl. Indium antimonidc has £. = 0.2.3 e V; dielectric constant t .. 18;
electron effective mass rn. - 0.01 5 m. Calculate (a) thc donor ionization energy;
(b) the raWus of the ground state orbit. (e) At what minimum donor coocentratKIIl will
appreciablc ""crlap effects be twCC' '1 the orbits of adjacent impurity atoms occur?This
m'crlap tends to produce an impurity band- a band of energy lel'cls which permit
conduethity prcsunuilily by a hopping medlo'\nism in whM:h electrons mm'c from ()fie
impu rity site to a neighboring ionized impurity site.

2. loniultion of donon. In a particular semicond uctor there are IOU donorsIcm 3 wit h
an ioniution en,"org)' £<1 of I meV and an effectivc mass 0.01 m. (a) Es!in,.tc the
Concent ration of conduction electrons al 4 K. (b) What is the value of the Hall coeffi-
cient? Assume no IICCei1tor atoms are p resent and that E. it> k/lT.

3. lIoU effect with liDO corrier t ype1f. Assuming concentrations n.p; klaxntKm times '7'••
1),; and masses 'Itc. lIiI.. show that the Hall coefficient in the drift velocity appro)(ima-
tion is
'"
1
(CCS) Ii ,f " " - -
"
where IJ = P-rip." u the mobilit)' ratio. In the dcrh-ation neglect terms of order 8 2 , In
51 we drop the c. lIinl : In the presence or II longitudinal electric field, Snd the
transverse dearie field s~h that the transverse current vlUl ishcs. The algebra ma)'
seem tedious, but the result is \\'Ol1h the trouble. Use (6.64), hut for two carrier
types; neglcd (wet,! ill comparison \vilh w,!r.

4. Cyclotron nsonanu for a ,plUlroidal energy 'Ur!fIU. Consider the energy surface

E(k) "" + ~ + k!) .

h2( ~2m, 21'111
where m,ls the transverse mass parameter and 1tl/ is the longitudinal mass parame ter.
A surface OIl which e(\c) is constant will be a spheroid. Usc the equat ion of Il\otion (6),
wi th v "" Ii - IV~", 10 show that We '"" eBl(m(m,)lI2c when the static magnetic field B lies
in the 1",1/ plane. This result agrees wilh (3-1) whe n e ..
'1112. The resu lt is in CCS; to
obtain 5 1, omit the c.

5. Magnetoresutal'lCl! u:id t IWf) ctl,-rie,. ' YPe1l. Problem 6.9 shows that in the driA veloc-
ity appro"imation the motion of charge carriers in electric and magnetic Gelds .lOCI
oot Icad to trallS\'crsc magnetoresistancc. The rC$ult is different with two car rier
types. Conside r a conductor with a concentration I I of clectrons of effective m.- 'llll5."
and relaxation time T~; an d a conccntration p ofholes of effective mass TIl" and relaxa-
tion time 1), . Treat the limit of \'ery strong magnetic fields, "leT"
1. (a) Show in t\ili;
lim it Ihat ()'~. - (n - p)edB . (b) Show that Ihe Hall field is gi\'en by. with Q - «0,,'1,

E~ --(n-p)(;, +;J-'E"
which vanishes if I I ... p. (c) Show that the effective conductivity in the :I: din.'CI.ion is

I f n .. " , ()' 0: 8 - 2 • If n ,-6 ". ()' saturates in strong fields; that is, it approachcs a limit
ind ependent of B 35 B_ 00.

Referen ces
K $eegr.r, Sem/cclllductor pilr;sia: lin ,,,troaUCliOlI. 4th cd. . Spril)b'er, 1969.
R. Dalvcn. Introduction 10 IIpplied wlid Itale phr;' /u. 2nd ed . . Plenum, 1990.
R. A. Smith, Slm/c.mducton. 2nd ed .. Cambridge, 1978.
S. M. Su, Semic(l"ductor devicU . P/'lISiCl . lind technologr;. Wi ley, 1965.
M. L. COOcn and J. R. Chr.li ko....sl;y, E1edronic $/1WI<'Te ond opt/ctJl prope:r1u" of _>condl'('-
.ors. 211.1 ed., Springer, 1989.
C. W. B<:enakker.nd H . vlUl Hou len, "Qual1tu,n tl1lfUCpUrl in %e.niconductor narlOSlrucItJrP.$,N
Solid slale physics 44 , I (1991).
C. Ba>lard d ai. , "Electronic 5tale~ In sem icond udor hetetooitructurel ," saltd ,tate physics <t4, 230
(1991).
 9
Fermi Surfaces and Metals

R(.-ducOO zone 5Chcmc 2.3:!i

Periodic zone scheme 231
CONS'"RUCTIO N OF FERMI SURFACES 238
Nearly free electrons 240
ELECfRON ORBITS, HOLE ORBITS. AND OPEN ORBITS 242-
CALCULATION OF ENERCY BANDS 244
Tight binding method for energy bands 245
Wigncr-Scitz mel hod 248
Cohcsi...~ ener gy 249
J\cudopotcntial metho.ts 252

EXPEIIIMENTAL METl-IOIJS IN FERMI SU RFAC E STUDIES

Quantization of orb i t.~ in II magnetic field
De Haas-\'aI! Alphen cffect
...
Z55

'"
btrcmal orbits
Fermi surface of copper
Example: Fermi sur ratt of gold
Magnetic br eakdown
...
200

,'",
SUMMARY
PROBLEMS
...
'66

...
I. Urillouin zoncs of rectangub.r lattice
2. Brillouin 7.llOC, rcdangular lattice .66
3. lIeugonal dose-packed structure

".....
4. Brillouin zones of two-dime nsional divalent m etal '66
5. Open orbits
6.
7.
8.
9.
Cohesive energy for II square wd l pote ntial
I)c HIUlS-van Alphen period of potassium
Band edge structure 00 k . p perturbation theor),
Wflnnicr runctions
.,
.66

26'
10.
II.

REFEREKCES
Open orbits flnd magncto .. csistancc
Landau levels ......
26'

---
 A1ummum

Figure t F",", dectroo Fermi surfaces for fi.x "",Iali .... ilh o~ (Cu) a/ld Ihree (Ai) valence ekdrw s
pcr primitive «II. 7hc Fermi surface lioo..,., for coppeo-lw been deformed from a ~phere 10 argro:e
wilh the exper imen lal results. TIll'! serond zone of aluminum is nc... rly h.lf·filled with .,kctrons.
(A. R. MlIICkinlosh.)
 CUAPTER 9: FERMI SURFACES ANI) METALS

Few ,>ropIe lcoulel define a ,,,etal a, Ma ,oIUllCjlh a

Fermi Iturface." 1'hi" may Iln;.erlhetelill be Ille m~1
mtllllj"gful lIefiJli/ivJI of a me/al olle C(l1I si~e loday; il
represell" a profoulI(i (lClt'(lllce ill liIe uullenta.Klillg
of ICilY melals /HiIa!:e IiII ,lley do. 1'Ile OO'lCejl / of Ihe
Fermi :mrface, fI& (le.;eloped by quun'"f11 physic"
1Jroddes a preciu errJlanalwm of Ih e maln p/lysicai
propel'Ijes of melals.
A. R. Mackintosh

1110 Fenni surface is the surface of constant ellerb'Y €t' ill k sp....t'e. TIle
Fermi surr., ce sepnrates the unfilled orbitals from the filled orbitals, at absolute
zero. TIle electrical properties of the metal are determined by the shape of the
Fermi surf.'l.Ce, becausc the current is due to changes in the occupancy of states
near thc Fermi surface.
11)c shape may be very intricate ill a reduced zone scheme and yet h:lVe a
simple interpretation when reconstructed to lie near the surface of a sphere.
We exllibit in Fig. 1 the free electron Fermi surfat'est'()nstructed ror two metals
that have the race-centered cubic crystal structure: coppcr, with one valence
electron, MId aluminum, with three. The rree e lectron Fenni surfaces were
developed rrom sphe res of radius kt' detennined by the valence electron con-
centration. How do we constnlct these sllrL'l.Ces rrom a sphere? The t"Onstruc-
tiolls require the reduced and pt:riodic zone schemes.

Reducetl Zmle Scheme

It is always possible to select the wa"evector index k or any Bloch runction
to lie within the first Brillouin zone. 'nle procedure is known as mapping the
balld in the reduced zone scheme.
If we encounter a Bloch function written as tf.\.(r) = e1k··'I/k·(r), with k'
outside the first zone, as in Fig. 2. we may always find a suitable reciprocal
lattice "ector G ~uch that k = 10;' + G lies within the first Brillouin zone. Then
I/Ii,.(r) = e'k"'udr) = ti....(e- 'c··u".(r »
(I)
iil e""'uk(r) ... I/Ii,(r) ,
where "k(r) .. e- IG·'uk.(r). Both e- .c·· :md u".(r) are periodic in the crystalla.t-
tice, so u,,(r) is also, whence "'J..~r) is or the Bloch rorm .
Even with rl'ee electrons it is useful to work in the reduced 7..one M:heme,
as in Fig. 3. Any 'rgy €k' for k' outside the fil"51 'Wile is equal to an €" in the
 (

Figure 2 First Bril100in zolle d a square lattice or side d . 'l1le ""lI\", ..n . e(or 10:' can be cwried into the
flrst wn.,IJy formillg k' + C . The W3'"C\'<.'Ctor III a poin' II on the xonc boundary il carried by C to
the point ... ' on th e opposi te houn(\;ory orihc same u)Jlc. shan we count both Ir. and A ' as 1~lng In
the fin' wne? ~me they call be co"noetoo by . rociprocalla.uice \"t.'dor, we count tloem as tme
identical point in the z:one.

"

,,
,,
,,
,,
, ,
\",
,.

_~>------C~--ok--<~------~
~

Lk-_____ ~~ I\ri~~'" - - -....i

FiJl;lIrl' 3 Enf'rg~'·wa'·c'·f!'d.or re lation fl - fi 1k"/2m £OT f ....'C electrons as drawn in the ••:<lured:w~
M:heme. n'i~ construction ofIen gh"es a lIiOC£ul idea d lhe o,",,:raI] appearance oflhe band Sll'1.l('lure
ol a Cl')slal. 111r: hrardl AC ifdisplattd b)' -2'1110 !;i\"Cl thc usualrr~ ck,rtroo run'c £".. '""-ptin: k,
as IU~f'Sled b,' the da.iht.-d C\In'l'. "The branch A 'C if dlspla<'eCl b,' 2""0 ghu the moW <'I,""'' '
for
po~i'i'c I. . II cry~lal potential U{r) will inlrodu<'o: band P Ili at the oog('s ortne xone (as at II a nd II')
and at the center orill., mne (as at C). The point C ""hen ,'le"'I.'<I in the .,xtend..'<1 zone sd,.,rne bib
at the ..dge. ofll",""'-"OfKI wne. Th., ..,...,nll widtll and gross fuaturcs ofuM.' band ~tructurc areof"ten
ind ...... ted ()I"Ope.ly~· such rroc electron ban<h in til., " ,d uc.-cd ZOOe scl!eme.
 \
fir.;1 zone, where k = k' + G. lllUS we need solve for the energy only in the
flrsl Brillouin zone, for ench band. An e nergy band is a single branch of the €"
versus k sun.'l.Cc.
III the reduced zone scheme we should not be surprised to find different
energies at the same value orthe waveveclor. Each diffe rent energy character-
izes a different band.
Two wnvcfunctions at the same k but of diHerent energies will be indepen-
dent of each other: the wavefunClions will be m ade up of different combinations
or the pL-mt: wave components exp[i(k + C). r] in the expansion of(7.29). Be-
cause the values of the <:oeffidenls C(k + G) will diller for the different bands
we should add a symbol, So'l)' n. to the C's to serve as a b.'lnd index: C ..(k + G).
nus the Bloch function for a state of't\'avcvt!(.1or k in the band II mil be written
~

"'.... :: exp(ik . r)u" ...(r) = Lc ClICk + G) cxp{i{k + G) . r]

Periodic Zone ScI.eme

We ron repeat a given Brillouin wne periodically through nil of wnvevcc-
tor sp.'l.ce. To repeat a zone , we tmuslate the zone by a reciprocal btti<:e vector.
If we can transL'l.te a band from other zones iuto the first zone, we can translate
a b.'Uld in the first zone into every other zone. In this scheme the energy «~ of
a band is n periodic function in the reciprocal lattice;
(2)
Here «he is unde rstood to refer to the same energy band as «I<.
The result of this construction is known as the periodic zone scheme. 11le
periodic property of the energy also can be seen Cc'lSily from the central equa-
tion (7.27).
Consider for example nn energy band of a simple cubic lattice as calcubted
in the tight-binding approximation in (13) below:
Ek = - a - 2y «.'Os kxU + cos ktf! + cos k~a) , (3)
""ilere a and y are constants. A reciprocal bttice vector of the sc lattice is
G ::t (2r.1a~; if we add this vector to k the only ch..'l.Oge in (3) is
(''OS k.a -+ cos (kr + 2r.1n)n = cos (k.a + 270) ,
but this is identically equal to cos k.a. 11le e nergy is unchanged when the
W3vevector is increased by a reciprocal lattice vector, so th;:.t the energy is a
pel'jodic fun ction of the w3vevector.
'"

Figu~ 4 Three energy band s of a linear lattice plotted in the extended (Brillou in), reduet:d, and
periodic 7,001" Khemes.

ll1fce diffe rent wne sche mes arc useful (Fig. 4):

• l1\C extended zone scheme in wh ich different b<'lnds are drawn in different
zones in wavcvcctor space.
• TIle reduced zone scheme ill wllich all bands are drawn in the first Brillouin
zone.
• 11lC periodic zone scheme in which every band is drawn in e very zone.

CONSTRUcnON Qi' FER~ 1l SURFACES

We conside r in Fig. 5 the analysi.. for a square laUke. TIle equation of lhe
zone boundaries is 2k . G + G 2 = 0 and is satisfied ifk terminates 011 the plan e
normal 10 G at the midpoint ore. The I1rst Brillouin Zf'-'" of the square laUice is
\
 9 "c.n"i Surfa,~, and lIfel<lh

I~
0 0 0 0 0 0 o

c. 0 0
, 2 ', :}
0

G,
• 0_ 0 o
0
" ,,,
0 0 0 0 0 0 0 o

0 0 0 0 0 0 0 0 0 o

.., (b)

Figure 5 (a) Consln lCtlon in Ie sp;!.C'e .,r Ille IlrS! thrtl" Dnl10uln :ameli nf a SQuare bUice. 111e tI1l"<.'C
shortcst £onn5 of the r..'ciprocallallice \"I~lon an: indlcuted III C" c., and G 3 . The lines dm"n arc
the pcrpcndi<:u lar blliecton of these G·s. (b) On c'mslr,,'etin~"l1 lines equi,..Icnt by symmetry to
the three lin.,lln (a) "-': obtain Ih., n."j1;ions in Ie 'p""" which funn the linllhree Brillouin "On.." . ·.11e
n.. nben d.,note the:t.Olle to which the n.-gions belong; the numben he...., are oroen. d acmrding to
the length.,r the , .......tor G Ill\vh-ed in the COIlllrl.ll1ion of the ou , ..... boundary of ,he region.

the area enclosed by the perpendicular bisectors ofC t and orthe three recipro-
cal lattice vectors equivalent by symmetry to C. in Fig. Sa. TIlese four recipro-
cal lattk'e vectors are :!: (2r.1a)k x and :!: (2r.1a)k ll •
111C second zone is constructed from Gs: alld the three vectors equ ivalent
to it by s)'mmetry. and similarly for the third zone. 11le pieces of the second
and third zones are drnwn in Fig. sb.
10 determine the boundaries of some zones we have to consider sets of
severnl Iloncqui\'alent reciprocal hiltiCC vectors. Thus the boundaries of sec-
tion 3., of the third zone are fonned from the perpendicular bisectors of three
C·s. namely (2r.1aik,; (4r.1al~; and (2r.1a)(k" + kv).
nle free electron Fermi surface for an arbitrary electron concentr.ttion is
shown in Fig. 6. It is inconvenient to have sections of the Fermi S\lrface that
belong to the same zonc appear detached from one another. The det<IChment
can be repaired by a trnnsformatioll to the reduced ZOlle scheme.
We take the trinnglc labeled 2" and move it by a reciprocallatlice vector
C :: -(2'IT/(llk" such that the triangle renppears in the area of the first Brillouin
zone (Fig. 7). Other reciprocallaltice vectors will shifl the triangles 21>. 2(". 2,/10
other parts of the first zone. completing the mapping of Ihe second zone il110
the reduced zone scheme. The parts of the Fermi SIIrt.'lCe falling in tl\{' second
ZOne are now co,""ected, as shown in Fig. 8.
 o•

I>l ....... f ..d ..,..., 3nI ZONe

Figu," 1 M3pJ'ing oft"'" fif$I , K<'Qnd. and th ird Bril10uin wn<'S in Ihe red"...,d zooc.. scheme. n.c
St.'d.iollS vi the .crond 1.Ono..' in f~ 6 arc put 1000000r ;1110 a 'Quare II\' h~nslalion th~ an
appropriate "'..... p.ocaI lattice ,'"",tot. A different G il; ncc.-dcd for <-~ pice<: of a zone.

,,\ 1/ r--..\
• •
'/1 \"
h i _'" 20MI ..,... "'"It
3.d
Figu • ., 8 'I'he fTt'" ck'dron Fe.,ni sunaco: of Fig. 6. as viewed in the ..oouc.,d zone scheme. The
Mwodcd "...:.-as represen t OC<'tlpied e lectron Iitalt.'S. Paruof thc ,,' coni ... r£:...'t! fall in the second, third ,
and fourth ZOOt'S 'Tho: 1Ou.lh ZOllO' ~ n()I shcw...,. 'n..: flTJ.I :woe is entirely IX'CUpied.

A third 7-one is a.~sembled into a square in F ig. 8, but the pnrls of the
Fermi sun.'K'e still appear disconnected. When we look at it ill the periodic
zolle scheme (Fig. 9). the Fermi surfat.'e fonns a lattice of rosettes.

Nellrly Free Eleclrons

Jlow do we go from Fermi surf.'lccs for free electrons to Fermi surfaces for
nc.lrly free e lect rons? \\le Colli make approximate constructions freehand by the
USI;: of rour f.lcts;
 ,
{

, ,

--ttl--~---~--
,, ON
,, ~
,,
I

--~---~---~--
,, ,,

"Itt;ure 9 TIle Fcnni ,,,,face in Ihe third 7.QnC as

d....wn in the periodic ,.one ~emc . 'Illc rlgUrc wu
constructed by repealing the th ird V>r\C of Fig. 8.
--tE---t1f---~--
, ,, , ,,

G
""'-
F"IgU", 10 Qualitali,'c impreSSion of the clT.."Ct of a weal: po..~iodk ('ryJlaI pol<.'flliai on the J-"crml
surb:e of Fig. 8. AI one POUlt on each Fermi surfao...e we h:wtl ~1' the ....'CtOJ gnd"t. In the
K'l:Ond:wnc the energy ina(.'3S<."III()\.>wO the inh..... ior of the figure. and in the third wne the e nerg)'
iOCTClllics tow. rd the exterior. The sllad..-d n:gJOOS are I1!Icd ....iul "k,.-t rons and an: lower in energy
than the unsll:ldcd regions. We _hall ~ that a . "cnnj 5Ur£acc like that o;f thc third lOne is clcctroo-
I,ke, whereas one like that of the !i<.'<."OfId 'lOne Is holcl;).,,,. It ~ said that cklctronS sink and holel
n...
• 11m interaction of the electron with the periodic potenlial or the crystal
causes energy g.'lpS at the zOlle boundarit..'S.
• Almost alv.ays the Fermi surface will intersect zone bound1ries perpendicu-
larly (see below).
• TIle crystal potential will round out sharp corners in the Fermi sun.'lct.'S.
• The lolal volume enclosed by Ihe Fe rmi surf.'ICe depends only on the electron
concentration and is independent or the details of tbe bttice interaction.

We cannot make quantit.'lth'e statements wilhout calculation, but qualitatively

we expect the Fermi sllIfuces ill the s('(.'Ol1d and third 7.ones or Fig. 8 to be
ch..'lngcd as shown in Fig. 10 .

..
'"
Freehand impressions of the Fermi surfaces derived from frcc t:leclron
surfaces arc useful. Fermi surfaces for free electrons are L'Onstrucled by a pro-
cedure credited to Harrison, Fig. II. TIle reciprocal lattice points are deter-
mined, and a free-electron sphere of radius appropriate to the electron concen-
tration is drown around each point. Any point in k sp.'K.'e that lies within at least
olle sphere corresponds to an occupied state in the first zone. Points within at
lenst two spheres correspond to occupied states in the sc<:ond zone, find simi-
larly for points in tlu"ee or more spheres.
We saki earlier that the nlk.:tli metals are the simplest metals, with weak
interactions between the conduction electrons and the lattice. Because the
alkalis have only one valence electron per atom, the first Brillouin zone bound-
aries are distant from the approximatciy spherical Fermi surface that flUs one-
llalr or the volume orthe zone. It is known by c..llctllation and experiment that
the Fermi surfa<.'e ofNa is closely sphe rical, and that ror Cs tile Fermi surface is
derormed by perhaps 10 percent rrom a sphere.
The divalent metals Be and Mg also have weak lattice interactions ruld
nearly spherical Fermi surf.'lCes. But because they h..wc two valence electrons
each, the Fenni surface encloses twice thc volume in k sprlce as ror the n1kalis.
11mt is, the volume enclosed by thc Fenni SUifacc is exactly equal to that of a
zone, but because the surf.'lcc is spherkal it extends out or the flrst ZOne ruld
into the set:ond zone.

ELECTRON ORBITS, HOLE ORBITS, AND OPEN ORBITS

We saw in Eq. (8.7) th.'lt electrons in a ~1atic magnetic fleld move on a

curve of constant ene'l,,), on a plane normal to B. An electron Of] the Fermi
surf.'lCC will move in a curvc 0 11 the Fermi surf.'l<''e. because this is a slll·ra<.'e of
constant energy. Three t)'l>es or orbits in a magnetic fleld are shown in Fig. 12.
TIle closed orbits or (a) and (1)) arc traversed in opposite senses. Because
particles of opposite chargc circulate in a magnetic fleld in opposite senses, we
s.1.y that onc orbit is electron like and the other orbit is holelike. Electrons in
holelike orbits move in a magnetic fleld as ir cndow(.>d with a positive charge.
TIl is is consistent "'ith the treatment or holes if] Chapter 8.
III (c) the orbit is 1I0t closed: the particle on reaching the zone boundary at
A is instantly umkbppecl back to B, where B is eqU ivalent to B' becnuse they
are connected by a reciprocal latlice vector. Such an orbit is called an Opell
orbit. Open orbits h.we an important effect on the magneto resistance.
Vacant orbitals near the top or an otherwise fliled band give rise to holelike
orbits, as in Figs. 13 and 14. A view or a possible encl'b'Y surt.'lCe in three
dimensions is given in Fig. 15.
O .. bits that enclose filled states are electron orbits. Orbits that enclose
empty states are hole orbits. Orbits that move rrom ZOo' to zone without clos·
ing are open orbits. \
 9 I-"e.",. Surfttcn .. 1Id Mdilh 2ta

Figure 11 Harrison <'''OlUlruction of frt'e

cl«1ron F<''f1fli surf:o.ccs on the sc."C'Ond,
third, a.nd IOurth ~.one, 1M • iill"....e lallk.~ .
n,., t-'crmi surfiK.e enclose, the "nlire Il .... t
zone, whid! Iherefore is Illled wilh eke-
trollS.

1101.: o.biI

'"
"-
B,," l

(. J (' J
t-'igurc 12 Molion in a magnetic Ileld oftlw " ,,\·cn.'Ctor of an electron on the Fenfl; surface: , in (a)
and (b) fo,- "',,roni $urra"..,llopoIogka!Iy cquh-alcnl to tllO$e of F ig. )0. In (a) the wa •..,,·<.'(."tor mo'cs
around the octi t in ac1ocl<w1le dUection ; in (b) th.cwa'·c' -,,'dorfI\O'\·cs around lheorbi t in a counter·
clockwise dirt"Clion. The dinxtion in (1)) II what ".., expert for II rrec elc."<."tron of <;h~ - e : the
smaller k ,.,.Iues ha.-c the lower energy, 50 that Ihe filled eiL-dron ~t~ tcs lie inside the Fermi
s" r&ce. We ....11 the orbi t in (b) electrunlilc. 1bc $Cnse of lhe motion in II magnetic Ileld is op~ile
in (a) to that In (bl, SO that we rder to the orbit in Cal as hoIclike. A holc 1IlO' ..,S as >I pnrtide of
pn<ih\"e charga: e. In (e) for a n.'C!angular nme ".., show Ihe motion on an open orbit In the: periodic
zone KilClTle. This is topologic.JJy inlenm.-diale betv.-cen a hole orbn and an elcctron orlHt.

F"'.gure 13 (a) Vacant Sillies at thl, <."(lrnCrs of an

al'llOIit·fillc.-d band, drawn In the reduced zone
...-hemc. (b) In the periodic zone Kht.'me the
," .
'

'-arious l)IIrh of Ihe Fermi ~"rfal"C are con·

n<.'Ctcd Ead! circle ronnl a holelike orbit. The
diFl"c.TCnt circles arc entirely t~lui\... IL"rlt to t-acb
othc.-, and ti!edens,ty of ~tatC$ is that of a ~nglc
arclc. (ll.., orb;b ",-oed not be true cirdcs: b-
the lattk.: "lO....n ,t is o nly required that the
orbits havc blrlOld 5)'mmctry )

OJ
....

Figu~ 14 Vacant stale1i ne:lr the lop of an almost filled balld in II two-
dimensional C1}'$tal. This flgure is cquivaleut to Fig. l2a.

'.
Ftgure 15 Olllstant energy surface in the Brillouin :KIne <i a simpl" cubic lattice, for the a!;.umcd
energy batKl E. " - 0 - 2}'(cos k", + COS ".,0 + COS "',.6). (II) Coustan' cnc'l,'Y 5"rflO(.'(' t .. - a . The
fllk-d \Vlume COIltains One ekct,on pcr primitive <.'en.(b) TIIC same .uri'ac<: exhibil(."Il in the peri-
od;c ,.:111<: schenw:. 111e rormocti\'ityofthc orbits is dearly showlI. Can you Jind electron, hole , and
open orbits for motion in II magnetic fleld Bi? (A. ScmmcrflJ1d and II. A. B<:the.)

CALCULATION Qt' ENERGY BANDS

Few masters of energy band calculation Icanled their methods entirei),

from books. Band calculation is a craft learned by cxperien(.-e, often developed
in groups, and needing access to computers. Wigner and Seitz, who perfOl'med
the first serious band C'cllculatiOlls in 1933, refer to afternoons spent on the
manual desk <'''alt~lators of those days. using one afternoon for a trial wavcfunc-
tion. Modern computers have." eased the pain . However. the fonnulation o f the
problcm re<luircs grcat care. and the <.'()mputer programs are not trivial.
Here we limit ourselves to three methods useful to beginners: the light-
billding method. useful for illterpoiation; the Wigner-Seitz method. useful for
the Visualization and understanding of the alkali metals; and the pseudopoten-
tia! method. utilizing the general theory of Chapter 7. which sho\'''s the simplic-
ity of many problems. Reviews of these and other mcthod~ are cited at the end
of this chapter.
 (

• • • •

,,'
tlgure 16 (a) Schc"....tic dn,y,it'8 '" "",'cf""ctioos or electronli on two h),}rogen atoms at large
KfJllralioo . (b) Croomd slate y,;a\-fl"unction at dosc~ liepantioo. Ie) Ettitcd Slate "'-a,·Junction.

Tight Bincling Method for Energy &mm

lei os start with neutral separated atoms and watch the changes in the
atomic energy levels as the charge distributions of adjacent atoms overlap when
the atoms arc brought together to form a (:rystal. Consider two hydrogen
atoms, each with an ciel'troo in the Is ground state. The wavcfunctions 1/111. 1/18
00 the separated atoms arc shown in Fig. 100.
As the atoms arc brought together, their wavcfunctions overlap. We con·
sider thc 1,"'0 combinations 1/1" ± !/lB' Each combination shares an electron with
the tv.-o protons, bUI an electron in the slale iJ!A + I/IIJ will have a somewhat
lower energy than in the state I/!A - !/In.
In 1/1... + 1/16 thc electron spends part of the ti me in the region midway
between the two protons, and in this region it is in the attractive potential or
roth protons at once, thereby in<:rcasing the binding energy. In ~A - ~B the
probability densit)' vanishes midway between the nuclei; an extra binding docs
not appear.
As tw() atoms arc brought together, two separated energy levels are
rormed ror each level of the isolated atom. For N atoms, N orbitals arc formed
ror each orbital of the isolated atom Wig. 17).
As free atoms arc brought together, the coulomb interaction between the
atom corcs and the electron splits the energy levels, spreading them into
hands. Each state of given quantum number of the free atom is spread in the
l'rystal into a band of energies. 111e width of the band is proportional to the
strength of the overlap interaction betv."CCn neighboring atoms.
There will be bal'lds ronned from 17, d• ... states (I "" 1.2. ..) ohhe free
atOms. States degenerate in the free atom will form different bands. Each will
not have the same energy as any other band over an)' substantia] range of the
wavevC(.'I.or. Bands may l'Oindde in energy at l'Crtain values ofk in the Brillouin
7.one.
'"
"...
0.'

l' - 0.2

I - 1.0
•
~ -"
~ -2.6

-3.4

- 4.2
, , I , ,

0 2 3 4 S
h"o:areot ~ d~. in &hr radii

Figure 11 The Is band or ~ ring ofm hydrogen atoms; the olle-t'ledron energy olk... laled in the
lighl-binding appro>: ;malion ,,~th the nea",st-ncighhor overlall intcgral or Eq. (9).

TIlC approximation that starts out from the wavcfunctions of the free at oms
is known as the tight binding approximation or the LeAD (linear combination
of atomic orbitals) approximation. The approximation is quite good for the inner
electrons of atoms, but it is not oftcn a good description of the l:onduction
electrons themselves. It is used to dCK'fibc approximately the d banos of the
Irans:tion metals and thc valence bands of diamondlikc and inert gas crystals.
Suppose thai the ground state of an elcctron moving in the potential U(r ) of
an isolated atom is 'J'{r), an s state. TIIC treatmen t of hands arising rrom degen-
erate (p,d, . . .)atomie levels is morecomplkal!.:d. Irthe influence or one atom
on anothe r is small, we obtain an approximate wavcfunction for one electron in
the whole crystal by taking

"M ~ L
J
C""" - 'J) (4)

where the sum is over aJi lattice points. We assume theprimitivc basis contains
one atom. This function is of the Bloch form (7.7) if ely::::: N- U2 FJ, which
gives. for a crystal of N aloms,

IJroJr) ::::: N - lf2 L exp(ilc . fJxP{r - rj) (5)

J

We prove (5) is of the Bloch form . Consider a translation T connecting two

lattice points:
 9 Ferm' S"r[acu and Metah 241

I/Jk(r + T) "" N- I12 L exp{ik . r)/pcr + T - rJ)

J

== exp(ik' T) N - I12
L exp{ik' (rJ - T)]I,O{r - (rJ - T») (6)
J

= exp(ik . T)I/Jk(r)

exactly the Bloch condition.

We find the first-order energy by (.-akulating the diagonal matrix clements
of the hamiltonian of the crystaJ:

(killik) ~ N - ' L L c'Pl.k . ('J - "")] ( ...IHI.)

J '"

where 1,0,,, s:; /pcr - r"J. Writing Pm = r,n - rJ'

(kIHlk) ~ L c,p( - 'k' P,J

'"
f dV .'(, - p",)HoI.') (8)

We now neglect aJl integrals in (8) except those on tllC same atom and
those between nearest neighbors (."()nnccted by p. We write
f dv 1,O*(r)lfq;{r) = -0: ; f dV 1,O*(r - p)Hlf{r) = - "")' ; (9)
and we have the first-(lrder energy, provided (klk) = 1:

(kllflk) = -0 -)' L exp(- ik· P"J = E. (10)

'"
The dependence of the overlap energy ), on the interatomic separation p
can be evaluated explicitly for two hydrogen atoms in Is states. In rydberg
energy units, Ry = me4 121r.2, we have
(ll)
where fl() = fr.2/mr. The overlap energy decreases exponentiaJly with the sepa-
ration.
For a simple cubic strudure the nearest-neighbor atoms are at
Pm = (±a,O,O); (0, ±a,O); (0,0, ±a) , (12)
SO that (lU) ba"Omes
E. = -0 - 2)'(l'OS k...n + cos kyO + cos k~a) (13)
Thus the energies are confined to a band of width 12"")'. The weaker the overlap,
the narrower is the energy band . A constant energy surfuce is shown in Fig. 15.
For ka 4: 1, f'k:' -0: - 6)' + )'k?-a2 . The effective mass is m* = fr.2/2)'a 2. \\'hen
...
the o\lerlap integral 'Y is small, the band is nalTOw and the effective mass is
high.
\Ve l.'o llsidered onc orbital of each free atom and ohtained one band £.,._The
numbcr of orbitals in the band that corresponds to a nondegencrate atomic
level is 2N, for N atoms. We see this directly: values of k within the first
Brillouin zone define independent w3vcfunctiolls. TI\C simple cubic zone has
-TrIa < kA < 'TrIa, etc. The m ne volume is 8'fillu 3, 111C number of orbitals
(counting both spin orientations) per unit volume ork space is V14,,3, so that the
n umber of orbitals is 2V1c? Here V is the volume orthe crystal, and lIal is the
number of atoms per unit volume. Thus there aTC 2N orbita1s.
For the bee structure with eight nearest neighbors.
EI< = -Q - 8')' cos ikrU cos jkv" cos tk=ou (14)

For the fcc structure with 12 nearest neighbors.

£1< = -0 - 4)'(cos ik~ cos ik~ + ('OS ik~ CQs ik..a +
cos ik~a ('Os lk~) (15)
A CQnstanl energy 5urfa('e is shoown in Fig. 18.

Wigtler-Seitz Method
Wigner and Seitz shoowed that at least foor the a1kaH metals there is nO'
iOC'Qnsistency between the electrO'n waverunctiO'ns O'rrroo atO'ms and the nearly
rroo ek'CtrO'n model or the band structure O'r a crystal. Over most oor a hand the
energy may depend O'n the wavevectO'r nearly as rO'r a rrec electron. HO'wever,
the Dloch waverunctioll, unlike a plane wave, will pile up charge O'n the positive
ion CQres as in the atO'mic waverunctiO'n.
A Bloch rUllction satisfles the wave equatiO'n

(16)

With p • -ill grad, we have

p ej""~II.,(r) = IIke· h ll.,(r) + en.·~pu,,(r)
"'·ru.,(r) + c'''-r{2hk· p)u"Jrl + e*Tp2u".{r)
p2 ei !<-rfl,,(r) "" (flk)2 e
thus (16) may be written as an equation fO'r II,,:

(2~ (p + IiIcf + VCr») u.,(r) "" EI.II,,(r) (17)

At k = 0 we have 1/10 "" uo(r), where uo(r) has the periodidty or the lattice. sees
the iO'n ('Orcs, and ncar them will look like the wavcrunction O'r tilt:: rrce atO'm.
It is much easie r to' find a solulion at k = 0 than roor a general k, I)(.:(:ause at
k "" 0 a nO'ndegellerate solution will have the full symmetry or the crystal. We
,.' jguR' 18 "mll~tant etle~ surt..ce of lin kc
Cf'Y't.&i 't roc:t""~. in the nc.... est·ndghbor t ~t­
hi"dinJl: 3I'Proxi"",t;OIl . The SII rf.K.c .ho.... n hall
t - - 0 + 21li.

atn then use t'OCr) to construct the apprOJdmate solution

V'k "" exp{ik - r)uo(r) . (18)
This is of the Bloch fonn, but 110 is not an exact solution of (17): it is a solution
onl}' if we drop the term ill k . p . Often this term is treated a~ a perturbation. as
in Problem 8. The k . p perturbatio n theory developed there is especiall}' uSC-
ful in finding the effective mass m· at a b.md edge.
Ihx.'ausc it takes account of the ion l..'Ore potential the function (18) is a
much oo"cr approximation than a plane Wdve to the 1..'OlTecl wavcfunction. The
enefl.,'Y of the approximate solution depends on k as (flk'fl2m, exactly as for the
pl.lIle wave, even though the modulation representc<l b y ttoCr) may be very
strong. Bcf..-ause Ill) is a solution of

( _ I_ p2 + VCr») 1I(~r) = EoIIp(r) , (19)

2m
the function (18) has the energy expectation value Eo + (h2 A:2l2m). The function
uoCr) often will give us a good picture of the charge distribution within a cell.
Wigner and Seitz developed a simple and fairly accurate method of mll-u-
lating " o(r). Figure 19 shows the Wigner-Se itz wavefunction for k "" 0 in the 3s
conduction band of metallic sod iulll . The function is practimlly 1..'Qllstant over
0.9 of the atomic volume. To the extent that tllc solutions for higher k may be
approximated by cxp(ik. . r)uo(r), the wavefunctions in the conduction band will
be similar to plane waves over most of the atomic volume. but increase mark-
edly and oscill,Ite within the ion core.
Cohuiue Energy. The stability of the simple metals with respect to free
atoms is caused hy the lowe ring of the ene rgy of the Bloch orbital with k "" 0 in
the crystal compared 10 the ground \·alence orbital of the free atom. The effect
is illustrated in Fig. 19 for sodium and in Fig. 20 for a linear peri<ldic potential
•

0
v- --. '.-'. ---.-
'. '"
", ..,k - O
•
~I

,.,.. alorn

- """""'"
kla!. ~ oJ Brillouio

v
I , , , -
o
r(8<ior u~u.)
•
Figure 19 Rad ial w"\'cfUnctions lin- the 31 orb ital offrce sodium atom and "r
& CQII.Iudion band
in sodium metal. The ",avefllnctiol15, which arc 1101 normalized here, are lOund by illl<:grJting the
S<;hrlldinger ~UQlioll lOr lin el...;lron in t he poten tial well of an Na' iOIl core. For the rrccalom the
"'''''cfundiDlI is integr.... ctl subject. 10 the usual Sdu6:lingcr boundary condition .x.r) _ 0 as r_ "';
the energy eigeo\i31ue is - 5. 15 eV. The wa""fu nction for ..-avevector k ... 0 in the melal is su bject
10 the Wigner_Seilz boundary <.'Qllditioo thai do/lfdr " 0 when r is mid".,." between nrighboring
atoms; the Clingy ofthis orbilal is -8. 2 eV, considl'r'A>I), lower than lOr the free atom. n.eorhitals
al the "«Inc booodar)' ........ nut mJed io $Odium; their enagy is -+2.7 eV. (ARn E. \Vlgr>el" and
F. SdtL.)

",," ,
.,•
~
-0.45

- 0.6
~
1
>
•~ , ,
b/d--

Figu re 20 Cround orbital (k .. 0) en ergy rOo" :.on electron in a periodic ,;qUlI", well potential or
depth !UJ - 'lIl1/mat . The energy Is lowe red as the ",-ells come doser togt:lher. Ikre a is hek'
comtan t and b is varied. Large bla cocretpond s to sel>al1ltcd atoms. (Courtesy ore. Y. Fong. )
 - ,'; IS,-"
,-
Cro..nd f "dt,·
CoI""i,'~ ~""'IY

-J---_____' .3 •.' .",===- A,~rage energy

FIgure 2 1 Coh('Si~ energy uf sodiu m 01<:1,,1is the

{lifl'erenw l>el~n th e Hvcrage energy of an dcc-
1«111 in the mela! (- 6.3 tV} and lhe Jo(rOund slale
energy (- 5.15 cV)ufthe valence 3" ek'Ctron in the
free: atom, relerr'l!d to Ilfl Na · ion .,ILI$ ireI' electron
- 8.2 )-:-::-:-:-:--
..r. infinite It'(l'Iralion. ~ _ 0 .....1'

of attral'tivc square wells. The ground orbital clIl!rgy is much lower at the actual
spacing in the metal than for isolated atoms.
A decrease in ground orbital energy will increase the binding. The de-
crease in ground orbital energy is a consCf.luence oflhc dmngc in the boundary
condition on the wavefunction: The Schrooinger boundary condition for the
frcc alom is .p(r) -). 0 as r __ CD, In the crystal the k :::: 0 wavcfunction uo{r) has
the symmetry of the lattice and is symmetric about r = O. To have this. the
normal derivat ive of I/J must vanish across every plane midway between adja-
cent atoms.
III a spherical approximation to the shape of the smallest Wigner-Seitz cell
we usc the Wigner-Seitz boundary condition
(20)
where '-0 is the radius of a sphere equal in volume to a primitive (.'ell of the
lattice. In sodium .,-o "" 3.95 Bohr units, or 2.08 A; the half-<listance to a near-
est neighbor is 1.86 A. The spherical approximation is not bad for ft."C and ix"C
slnK1ures. The bou ndary (.'ondition allows the ground orbital wavefu nl.1ion to
have much less curvature than the free atom boundary condition. Much less
curvature means much less kinetic energy.
In sodium the other ruled orbitals in the conduction band can be repre-
sented in a rough approximation by wavcfuncfions of the form (18). with
h'l<'
EIc=Eo+-- .
2m
TIle Fermi energy is 3. 1 e\', from Table 6.1. The average kinetic energy per
electron is 0.6 of the Fermi energy. or 1.9 e\'. Because Eo -8.2 eV at k :::. 0, =
the average electron energy is (E0 = - 8.2 + 1.9 = -6.3 e\', compared with
-5.15 eV for the valence electron of the free atom, Fig. 21.
We therefore estimate that sodium metal is stable by about 1.1 eV with
respect to the frc- "tom. This result agrees well with the experimental value
'"
J.13 eV. We have neglected sevcral corrections whose overall effect in sod ium
is small.

PSeOOopotential Mdhoo3
Conduction electron wavcfunctions arc usually smoothly varying in the
region between the iOIl cores, but have a complicated nodal structure in the
region of the cores. This behavior is illustrated by tIle ground orbital of sodium,
Fig. 19. It is helpful to view the nodes in thcoonduction electron wavefunction
in the core region as created by the re(juircmcnt that the function be orthogo-
nal to the wa\'cfunctions of the core electrons. This alll'Omes out of the Schro-
clinger equation, but we can see tha i we nc(.'d the flexibility ortwo nodes in the
as OOnclul'tion orbital of Na in order to be orthogollal both to the 1.1' corc orbitaJ
with no nodes and the 2s (.'ore orhitaJ with one node.
Outside the core the potential energy that acts on thc conduction electron
is relative ly weak the potential energy is only the <''QuJomb potential of the
Singly-charged positive ion cores and is red uced markedly by the e lectrostatic
screcning of the other (.'Onduction e1cctfO!lS, Chapter 10. In this outer region
the conduction e lectron wavefunctiolls are as smoothly varying as plane waves.
If the conduction orbitals in this outcr region are approximately plane
waves, the energy must depend on the wavcvcctor approlCimatcly as if" =
IJ'lk 212m as for free ciectron!L Dut hO\\I do we treat the conduction orbitaJs in the
<''Q re region where the orbitals arc not Rt all like plane waves?
What goes on in the <''Q re is largcl)' irrelevant to the dependence of ~Oll k.
RamI that we can calculate the enerJ.,'Y b), applying the hamiltonian operator to
an orhital at Rlly point in space. Applied in the outer region, this operation will
give an energy nearly equal to the free electron energy.
This argument leads naturally to the idea that we might replace the adual
potential energy (and filled shells) in the core region by an effective potential
energy' that gives the same wavefunctions outside the core as are give n by the
actual ion cores. It is startling to find that the effed.ive potentiaJ or pseudolJO-
tential that satisfies this req ui rement i~ nearly 7.cro. This conclusion about
pseudopolenhals is supported by a large amount of e mpirical experience as well
as by theoreti<.'al arguments. 111e result is referred to as the cancellation theo-
rem.

IJ. C. Phillip> and I. Kleinman , Phy). Rt,v. 116, 287 (19.59), E. Antonci!.:, J. Phys. Olem.
Sohds 10, 314 (1959). 11K: gene",1 theuyol pseodopotenlialHsdiscunai by B. J. AU5h n. V. Heine.
lOud L. J. Sham. ph" , Rev. 127. 276 (1962), see also \ 'ol 24 d'Solid , rolf! phI/des . TIle utilityofthe
emilly l'(lre model has been known ror many years, it goes l:.acl.: to £. Fcrrni, r..'uovo Cimenta 2. 157
(1934). H . Hellmanu, Ada J'hysiochlmd URSS I , 913 (1935). and II lIellffilllln r.and \\I Kas ·
J.atol~. in, J Chern. f'h)'S. 4, 324 (1936), " 'M wrole "S~ the field «the Oon determined in this
way runs a rather flat ('(lUrie, it is , uffid ent in the first apprw:imation to sd the Vlllence electron in
the lanke equal to a plane wa\,t',·'
 9 Fermi S"'r!acu and lIfdah
(
The pseudopotential for a problem is not ullique nor exact, bllt it may be
very good. On the Empty Core t-,·Iodel (EeM) we can even take the unscreened
pseudopotential to be ?-cro illside some radius R..:
forr<R.,;
(21)
forr>~ .

This potential should now be scrccned as described in Chapter 10. Each (.um-
ponent U(K) of VCr) is to be divided by the dielectric constant E{K) of the
electron gas. If, just as an example. we use the Thomas-Fermi dielet.1ric func-
tion (10.33), we obtain the screened pseudopotentiaJ plotted in F ig. 22a.
The p seudopotential as drnwn is much weaker than the true potential, but
the pseudopotential was adj us ted so that the wavefunctioll in the outer region is
nearly identical to that for the true potential. In the language of S(.-altering
theof),. we adjust the phase shifts of the pseudopotential to match those of the
true potential.
CaJculation of the band strudure depends only on the Fourier (.'Omponents
of the pseudo[XItential at the Tet.iprocal lattice vectors. Usually only a few val-
ues of the coefficients U(C) are needed to get a good balld structure: sec the
U(e) in Fig. 22b. TIlese cocHkients are sometimes calculated from model p0-
tentials, and sometimes they are obtained from Ats of te ntative band structures
to the results of optkal measurements. Cood values of V(O) ('"'all be estimated
from Arst principles; it is shown in (10.43) lhat for a screencU Coulomb poten-
tia] U(O) = -§E,.-.
In the remarkably successful Empirk-al Pscudopotential Method (EPM)
the band structure is calculated using a few cocfficiellts U(C) deduced from
theoretkal fits to measurements of the optical reflectance and absorption of
crystals, as discussed in Chapter n. Tables of vallles of V(C) are given in the
reviC\-Y by M. L. Cohen and v. Heine.
Charge density maps can be plotted from the wavcfunctions gcnernted by
the EPM-see Fig. 3. 11. TIle rc.~u hs are in cx(.'Clien t agreement with x-ray
diffraction determinations; such maps give an understanding of the bonding
alld have great predictive value for proposed new structures and (.'Ompoullds.
TIle EPM values of the (.'OCfficients U(e) often arc additive in the contribu-
tions of the several types of ions that are present. TIllis it may be possible to
(.'Onstruct the U(C) for entirely new strudures, starting from results on known
structures. Further, the pressure dependence of a band structure may be de-
termined when it is possible to estimate from the form of the U(r) curve the
dependence of U(C) on small variations of C.
It is often possible to calculate band structures. (.uhesive energy, lattice
constants. and bulk moduli from first principles. In such ab initio pseudopoten-
tial calculations the basic inputs arc the crystal !itru(.1ure type and tIle atomic
number, along with well-tested thcoretk-al approximations to exdlange energy
 r. in unit. of Bohr ,,01;;_

O·4-------+-----1f---'j--------~3-------__i.

P'I,-"Idopol<'fl.lla l ]

o ----- ------------- •: ti ro\.

y¢~._..I.,wt.'-"\t;il
· •• , . ..-n_-;
- .-
P"'.:,;,.;.;_.... -
.,0 ..............
...\:~..;;'".,
..........
..-_--

of
~, -, .. .
..
•
~
S
i Ionic potentia!
··
··
·:··
Figure 22.a Pseudopotential fur melallic sod ium, based 00 the empty core model and screened by
the Thomas·Fermi dielectric fnnction. The calculations Were made for an emply COre rod iu. R.. -
1.66ao. where 00 is the Bohr rodin • • and fur a >CJeening p"nunetcr 1:.00 "" 0.79. The dashed cu ...·e
.hows the a~.umed uoscreened potential, as from (21). The dotted curvt! is the actual potential of
the ion core. other value. of VCr) are -50.4, - B .6. and -4.6, for r " 0. 15, 0.4, and 0.7, reSIlf!C-
lively. Thus the actnal potcotial of lhe ion (chosen to fit the energy [e,'cls of the free atom) is "Cl)'
much larger than the pseudopotential, O,'el 200 timC$ larger at r - 0. 15.

c,
O~------~c~.--------~c~.~7'~-:;::O,.----

Figure 22b A typical rttipl"OClll spac<!'

pseudopotential. ValuC$ of U{I!) ror
w1,veveclors equal to the reciprocal
lattice vectors, G, are ind icated by the
dol$ . FOr "ery ••nall k Iht pnt totial
appr01lChe. (-213) limes t he Fermi
eoergy, which is the ' crecoed·ion limit
fur metals. This limit is derived io
Chapter ' " (Arter M. 1- whe,..,.)
(
terms, This is not the ~me as calculating from atomic number alone. but it is
the most reasonable basis for a first -principles calculation. The results of M. T.
Yin and M. L. Cohen. Phys, Rev. B 26. 5668 (1982). are compared with experi-
ment in the table that follows.

Lattice Cohe!iive Bulk

constant energy modulus
(AI (.VI (MbBr)

Silicon
Calculated 5.45 4.84 0.98
Experimental 5.43 4.63 0.99
Cermanium
CaIClliatt:d 5.66 4.26 0.73
E~perimclltal 5.65 3.85 0.77
Diamond
Calc... lah.-d 3.60 8.10 4.33
E~perimental 3.57 7.35 4.43

EXPERIMENTAL Ml!.IHODS IN F'ERMI SURFACE STUDIES

Powerful experimental methods have been developed for the detennina-

tion of Fermi surf.1ceS. "Mle methods include magnetoresistallce. anomalous
skin effect. cyclotron resonance. magneto-acoustic geometric effects, the Shub-
nikow-de Hans effect. and the de Haas-van Alphen effect. Further information
on the momentum distribution is given by positron annihilation, Compton scat-
tering. and the Kohn effect.
We propose to study one method rather thoroughly. All the methods are
useful, but all need detailed theoretical analysis and are not for beginners. We
seled the de Haas-van Alphen effect because it exhibits very well the charac-
teristic periodicity in lIB of the properties of a metal in a uniform magnetic
field.

QuantiwtKnI of Orbits in a Magnetic Field

The momentum p of a particle in a magnetic field is the sum (Appendix G)
of two parts, the kinetic momentum Plein = mv = file and the potential momen-
tum or field momentum Pfkkl = qAlc. where q is the charge. The vector poten-
tial is related to the magnetic field by 8 "" curl A. The totaimomentum is
(CGS) P "" Pki" + PIM<l :z: ~k + qAlc (22)
t
In Sl the factor r - is omitted.
 Following the sem icla~sicaI approach of Onsager ane. .fshitz. we a~sume
that the orbits in a magnetic field are quantized by the Bohr-Sommerfeld
relation
f P • dr "" ('1 + 'Y)211'f1 , (23)
when n is an integer and y is a phase correction that for free electrons has the
value i. Then

fp 'dr = fflk'dr + :f A ' dr (24)

The equation of motion of a particle of charge q in a magnetic Aeld is

dk q dr
h - - - - .8 (250)
dt c dt
We integrate with respect to time to give

Ilk .., !L r x B
c
apart from an additive coost3l1t wh;ch does not cootribute to the final result.
Thus one of the path integrals in (24) is

f"k'dr : : frx B·dr = - : B ·f r xdr = - ; ~ (25b)

where ¢I is the magnetic flux contained within the orbit in real space. We have
used the geometrical result that
f r I( dr = 2 x (area enclosed by the orbit) .
The other path integral in (24) is

-qf A ' dr =-qj curIA · du ""-qj B · du== -q¢l ("-5c)

c c c c
by the Stokes theorem; here du is the area e leme nt in real space. The momen-
tum path integra! is the sum of (25b) and (25c):

fp·dr =- :¢I=(n+ Y)2Trh. (26)

It follows that the orbit of an electron is quantized in such a way that the
flu x through it is
¢I" = (n + y)(2Trhcle) . (27)
The flux unit 211'f1de.:: 4.14 X 10- gauss 1 or T cm 2 m2 .
In the de Haas-van Alphen effect discussed below we need the area of the
orhit in W3vevcctor space. We obtained in (27) tile fl ux through the orbit in real
space. By (25a) we know that a line element llr in the plane nOl'mal to B is
related to 6k by 6 ("deB) 6k, so that the area S" in k space is related to the
area A" of the orbit in r space by
A" = (fldeB'fS" . (28)
It follows that

4>" == (-I>c)' 1
, -SOl
2"", ,
B = (n + 1')--
, (29)

from (27), whence the area of an orbit in k space will satisfy

2..
S,,=(n+1')~B . (30)

In Fermi surface experimenl~ we may be interested in the increment 6B

for which two successive orbits, n and n + 1, have the same area in k space on
the Fenni surface. TJ:Ie areas are equal when

S(B:+I - ;)== 2fl7 ' (31)

from (30). We have the important result that equal increments of liB reproduce
similar orbits-this periodicity in liB is a striking feature of the magneto-
oscillatory effects in metals at low teml~ratu res: resistivity, susceptibility,
heal capoo.ty.
The population of orbits on or near the Fermi surface oscillates as B is
varied, causing a wKle variety of effects. From the period of the oscillation we
ra'(lnstruct the Fermi surface. The result (30) is indel~ndeJlt of the gauge of
the vector potential used in the expression (22) for momentum~ that is, p is not
gauge invariant, but SOl is. Gauge invarian(.'C i~ discussed further in Chapter 12
and in Appendix G.

De llaas-t:4n Alphen Effect

The de I Ius-van Alphen effect is the oscillation of the magnetic moment of
a metal as a function of the static magnetic field intensity. The effect can be
observed in pure specimens at low temperatures in strong magnetic fields: we
do not want the quantiz.;ltiOfl of the electron orbits to be blurred by collisions,
a,xl we do not want the population oscillations to be averaged out by themlal
population of adjacent orbits.
The analysis of the dHvA effect is given for absolute zero in Fig. 23. The
electron ~pin is neglected. The treatment is given for a two-dimensional (20)
system; in 30 we need only multiply the 20 wavefunction by plane wave fac-
tors exp(ik~), where the magnetic field is para.llel to the :t axis. The area of an
orbit in k., kv space is quantized a~ in (30). The area between successive or-
bits is
(32)
 ·., " " • B- O
"

,,j ~I
'" "
Figure 23 E>;plination of the de 'I:us-'~ Alphen elfeet ... " rree electron W'S in two d;me~
'<I

in Il lTIIIW'eticfldd. The IlIled orbif;lb oCtile Fermi sea in the abscnccoh magnetic rw::1d are shaded
in p and J. n.e ene'lt)' levels in a IllIIgnelic roeklllll.' ' )lOWn In 11, c. and e. In b the field has" value
8, such that the total energy of the electrons is the _ &I In the al;.;ence or" magnetic fock!: II'"
man)' electrons have their energy nised as 1.....ere<! h)' the orbital quantiz:ltion in the magn..tic field
8,. When"," increase the field to 8~ the total elcctTon el>c'1(Y is ~~, because the upJlel'Tl'\m1
electrons ha,-c their I'llerg)' niseo:!. In e f(,.. lick! 8, the energy u; again equal to that ~r the field
8 " O. 1be total energy is a minimum at points such as 8 .. 8 3 • B,. ... , and " ma.:.imum near
points such as 8" B~, .

The area in k space occupied by a single orbital is (27T/L)2, neglecting spin,

for a square specimen of side L. Using (32) we flnd that the number of free
e lectron orbitals that coalesce in a single magnetic level is
D = (27Tf!Bllic)(U21T)'l = pB , (33)
2
where p = eL /27Thc, as in Fig. 24. Such a magnetic level is called a Landau
level.
lne dependence of the Fermi level on B is dramatic. For a system of N
electrons at absolute zero the Landau le\'els are e ntirely fllled up to a magnetic
quantum number we identify by s, where s is a positive integer. Orbitals at the
next higher level s + 1 will be partly filled to the extent needed to accommo-
date the electrons. The Fermi level w:illlie in the Landau level s + 1 if there
are e lectrons in Ihis level; as the magnetic field is increased the electrons move
to lower levels. When s + 1 is vacated. the Fermi I~'el moves down abruptly to
the next lower level s.
The electron transfer to io"..er Landau levels can occur because their de-
generacy D increases as B is increased. as shown in Fig. 25. As B is increa~ed
there occur values of l:I at which the quantulll number or the uppemlost filled
\
J,'igu re 24 (a) Allowed electron orbitals in two dimensions in absence ora magnetic lidd. (h) In a
magnetic field the points whicb repre.wnttne orbitals orrrce elcrtrons may be v>e~ as redrlcted
to ~ircles in the rormer k,k" plane. The SU<:"CeS$;"'e cirdcs COrffspond to StiO.'Ceui.. e \";I.lue~ of th.,
Quantum number n in the energy {n - Ill:...,.. The _ between sua:essi,.., circks II
(ceS)
n.e angular positiln of the points has flOlignif.call~e. "The numberrI o.bitab 00 acircle b ('()Clltant
and is t.'Quai to the area between Sll<:USSJ.-e circles times the nwnber of orbitab per unit area in (a).
or (2 .....BI fu:)(U2.1I"' '" L'"Bf}.Trfrc. neglecting electron .pin.

I t

.-
CO
%r-----1------J,------t,-----"t---
1000B-
0"
I'1gure 2.5 (a) "The hea\')' line gi"es the number or particles in Ic,"Cl~ which arc completelyoccupicd
in a magnetic field B, for a two-dimCO$ional I)'ltem with N .. 50 and p - 0.50. The . haded ;orca
gi"es tt... number of particles in Ic-.·ds partially occupied. The vaiue of. dcr"ll)tes t he qUllnt,,'n
number of the highel;t Ic-.'d which is C'(lmpietely filled. Thus at B - 40 we have . .. 2: the le""I,
n _ I and n ... 2 are filled and there are 10 pllrtlcle. in th., level n .. 3 AI B .. SO the \c-.'d n - 315
t'mpty. (b) The periodicity in liB i. e'\idenl when the lame point5 are plotted ag.;oinst l IB.

level decreases abruptly by unity. At the critical magnetic fields labeled B. no

level is partly occupied at absolute zero, so that
spB~=N . (34)
The number of filled levels limes the degeneracy at B. Illust equal the number
of electrons N.
 To show the periodicity of the energy as B is varied. we use the result that
the energy of the Landau leve l of magnetic quantum number n is E.. =
(n - j;)~ we, where We = eBlm*c is the cyclotron frequency. The result for E"
follows from the analogy between the cyclotron resonance orbits and the simple
harmonic oscillator, but now we have found it convenient to start counting at
n = 1 instead of at tI == O.
The total energy of the electrons in l evel.~ that are fully occupied is
•
L Dhw.{n -I): I D"w~
.-. , (35)

where D is the number of electrons in each leve l. The toral e nergy of the
electrons in the partly occupied leve l s + 1 is

where sV is the number of electrons in the lower fill ed levels. The total energy
of the N e lectrons is the sum of (35) and (36). as in Fig. 26.
The magnetic moment J.t of a system at absolu te zero is given by Jl ==
-aU/aB o-m e moment he re is an o!iCillatory fUnction of liB , Fig.~27. This oscil-
latory magnetic moment of the Fe rmi gas at low temperatures is the de Haas-
van Alphen effect. From (31) we see that the oscillations occur at equal intervals
of 11B such that

6(~)
B
: 2""
~ cS '
(37)

where S is the extremal area (see be low) of the Fermi sUlface normdl to the
direction of B. From measure me nts of 6 (JlB), we deduce the corresponding
extremal areas S; thereby much can be inferred about the shape and size of the
Fermi surface.
Extremal Orbits. One point in the interpretation of the dHvA effect is
subtle. For a Fermi surface of general shape the sections at differe nt value5 of
ko will have differe nt periods. The re5ponse will be the sum of contri butions
from al1 sections or all orbits. Bllt th e dominnnt response of rile system comes
f rom orbits whose periods n re statiorlnry with respect to smnll chnoges in kB.
Such orbits are called extrem al orbits. Thus in Fig. 28 the section AA' domi-
nates the obse rved cyclotron period.
The argument can be put in mathematical form, bu l we do not give the
proof here (QTS. p. 223; Ziman, p. 322). Essentially it is a question of phase
cancellation ; the contributions of differen t Ilonextremal orbits cancel, but near
the extrema the phase varies on ly slowly and there is a net signal from these
orbits. Sharp resonances are obtained even from complicated Femi surfaces
because the experiment selects the extremal orbits.
 9 " r r ml Surfoa. lInd Mdllb 261

ToI.I """rIP'

" , , • ,
10018-
Figure 26 -11m upper curve i.s the tolal electronic energ.' "enu~ l IB. lOc osciIlutionsln the ene'ltY
U ,nay be detected by nlCI1$urement of ille magnetic moment, ~i\'cn by - aUlaB. 11K: thamal and
tr.lnSilOrt properties of Ihe metal also OKillate as 51lCCcnh'c orb ital icvcb cut through the Fermi
b 'd whe n the field .5 h"-"'reased. The shaded region in the figure giv~ II..., contribution to the
1:'WIgY from levels that an only partly filled. The Ila~ters lOr the rogun: are the same as lOr
Fig. 25, and we h ave taken the un iU of 8 5u<:h that B .. "'"'€'

,i
•

Figu", 27 At absolute :tet'o the magnetic moment is g,,,'ll by - aUfitB. 11.c energy plotted in
F,g. 26 leads to the mllgllCtic moment shown here, an OS<'illatury fum;llon of liB In impure
Sp«~meru the 05CiI1allol>. aro smudged out in part bcca"u~ the energy Ie"eli are no longer shllrply
dcfinccl-
 Figure 28 The orbits in the 5edion AA' are extremal
orbits: the cyclotroo period is roughly c:-onstant O"er a
rf'Qonable $CCtion d the Feroni surface. Other sec' iolls
~uch Wi 88' have orbits that '111")' in period along the
section.

Fermi Surface of Copper. The Fermi surface of copper is distinclly non-

spherical: eighl necks make contact with the hexagonal faces of the firlit Dril-
louin zone of the fcc lattice. The e lectron concentration in a monovalent metal
with an fL'C structure is n = 41t?: the re are four e lectrons in acube of volume cr.
The radius of a free electron Fermi 5phere is
kF = {3w2n)' 13 = {J2w2/cfJ)tl:l iii (4.9fYa) , (38)
and the diameter is 9.8010.
The shortest distance across Ihe Brillouin zone (the distance between hex-
agonal faces) is (21Tla)(3)",z = 1O.881n, somewhat larger than the diameter of the
free electron sphere. Th e sphere does not touch the zone boundary, but we
know that the presence of a zone boundary tends to lower the band e nergy ' lear
the boundary. Thus it is plausible that the Fermi surface should neck out to
meet the closest (hexagonal) faces of the zone (Figs. 18 and 29).
The square faL'CS of the zone are more distant, with sep.1ration 12.57/0, and
the Fermi surface does not neck ou t to meet these faces.

EXAMPLE: Fermi Surface of Cold. In gold for quite a wide range offield dirt:diollS
Shocnbcrg flnds the magnetic moment has a period of 2 x 10- 8 gauss- I. This period
corresponds to an edremaI orbit cl area
1
S 21relflc 9.55 x 10 8 16 -2
• eo: .o.(IIB) iii 2 x 10- 8 ill 4. x 10 em

From Table 6. L. wc have k" - 1.2 x Hi' em- I lOr a fn:."e electron Fermi sphere for gold,
or a ll extremal area of 4 .5 x 10 16 em- t , in general agreement with the experimental
value. The actual periods re{>or((.-<1 by Shocnbcrg arc 2.05 x 10- 8 gauss- I for tilC orbit
Bil l of F ig. 2B anl! 1.95 x 10- 0 gauss- I lOr BlOo • In the Illil dil"(.'(:lioll in Au 11 large
pcriolI of 6 x 10- 8 gauss- I is abo founll ; the corresponlling orbital area is 1.6 x
1015 em- s . This is the "nc<:k~ orbit N. Another extremal orbit, the "dog's bone,~ is
shown in Fig. 30; its area in Au is about 0.4 of tile bell y area. Ellpc,;mcntal resul ts are
shown in Fig. 31. To do the example in 51, dlop c from the relation for S and usc 8S the
period 2 x 10- 5 tesla- I •

"' ..
 or
t'igure 29 Fennl surface copper, after Pipila rd. The
Brillouin :woe or the It:<: structure is the truf1Cll.ted oct,..
hror<.>n derh'Cd In Chapter 2. The Fenni surface makes
«Intact with the boundary at the center of the hexagonal
faces of thc zone , in the [ III } directioos in k space. Two
MbeUy" elltrcmal orbits are shown, denoted by B; the
eXh'emal " nedt orbi t is denoted by N.
•
. 'jgu re 30 Dog'~ bone orbit of an electron on the
Fermi surface of Wf'Pl'r or gold in a mllgJlctic
field . This orbit is c:Ia~ified as hoklike bee:tu$e
the energy increases toward the interior of the
orbit.

45.0kG 45.5 kG 46.0kG

Figure 31 De Haas·van Alphen effect 1o gold ""th B I [1I0}. The oscillation is from the dog's
bon., orbit of Fig. 30. The sibroal is rel:lIed to the second derivative of the magnetic moment with
~pect 10 rlCld. The results were obtaifled by I. field modulation technique in I. high,oolT><Jgene:ity
IUpereonducting wlenoid .t abou' 1.2 K. (Cou rtesy of I. M. Templeton.)
2M

Figure 32 Multipl)·-<.'OIlneded hole iiurface of

magnesium in hand~ I Bnd 2, aooJrding to
I... ~f FalicoV. (Dmwing by ~farta Pucbla.)

--- ~
-
I ( ) WI) (I ( ~
-
Figu re 33 Breakoown "rband ~Irudure by. strong n\3gJ1dic field. Brillouin zone boundaries IIIl:
the lig),t line, . The frcc ck.'d ron orbit ' (al in II strong field cllange ~~n nLdi\'ity in II weak field (b) to
bo:<:ome open orbits in the forst band and electron orbits in the lC-....:oOO '-00. Both bands an:
nlaI'Pt.-d together.

The free ek>ctron Fe rmi sph ere of aluminum fills the fl rs t 7.onc entirely
and ha!'i a large overlap into the second and third ZOIlCS, Fig. L The third zolle
Fenni sllIface is quite complicated. even though it is just made up of certain
pk-u'S of the surface of the free electron sphere. Ihc ff(..'e electron model also
gives smml sxx:kcl<; ofholc!> in the fourth zone, htlt when the lattice potenlial is
taken into act'Ou nt these e mpty out, the electrons heing added to the thi rd
wile. The general features of the pn.>dictt.>d Fermi surface of aluminum are
quite well verified b)' experiment. Figure 32 shows part of the frt.'C ek'Ctron
Fermi surface of magnesium.
Mllgnetic Brellkdown. Electrons in suflicientl)' high magnetic fields will
move in free parlicle orbits, the circular cyclotron orbits of Fig. 33a. J lere the
magnetic forces arc dominant, and the lattice potential is a slight perturbation.
In this limit the classification of the orbitals into bands may have little impor-
tan<.'C. However, we know that at low magnetic fields the motion is d~cribed
 9 , ........; SurjoQ!' and Mdoh 265

by (8.7) with the band structure £,. that obtains in the absem..'C of a magnetic
field .
The eventual breakdown of this description as the magnetic field is in-
creased is called magnetic breakdown .2 1'he passage to strong magnetic fields
rna)' drdStkally change tbe connectivity of the orbils, as in the figure . !he onset
of magnetic breakdown will be revealed by physical propcrtiC5 such as mag-
netorcsistance that depend sensitively on the connectivity.
The condition for magnetic breakdown is that "w~£,.· > E:.
approximately.
Here t:,.. is the free electron Fenni energy and Eg is the energy gap. !his
condition is much milder, espt!ciall)' in metals with small gaps, than the naive
condition that the magnetic splitting IIw~ exceed the Wip.
Small gflps may be found in hcp metals where the gap across the hexagonal
face of the 7..one would be zero except for a small splitting introduced by the
spin-orhit interadion. In Mg the splitting is ofthe order of 10- 3 eV; for this gap
and £,.. - 10 eV the breakdown {.'ondition is "w~ > 10- :1' eV, 01' 8 > tOOO G.

SUMMARY

• A Fermi surface is the surface in k space of constant energy equal to t:,... The
Fermi surface separatL'S filled states from empty states at absolute zero. The
form of the Fenni surface is usually exhibited ~t in the reduced zone
scheme, but the connectivity of the surfaces is clearest in the periodic zone
scheme.
• An energy band is a Single branch of the q, venus k surfat-e.
• 11le cohesion of simple metals is accounted for by the lowering of energy of
the k = 0 conduction band orbital when Ule boundary conditions on the
wavefunction arc changed from Schrlklingel' to Wigner-Seit£-.
• The periodicity in the de Ha:ls-van Alphen effect mcasurL"S the extremal
crO:'is-section area S in k spat.'C of tJle Fermi surfat'C, the cross section being
taken perpendicular to 8;

,,(-'-) ~ 2w, .
B lIeS

I R W. Slark and '- M. Falicov, Hp.,bgnc.1.ic Breakdown In Metals. Hin low tm.pn-oIu~ phys-
Ics, Vol. V, ,,"orth Holland, 1967. pp. 23&-256.
'"
Problems
I. Brillouin WI'IU of rrdllngulor lattice. Make a plot of Ihe fin! two Brillouin zones of
a IJrim itivc J'(.'(.1angular twu-dhncnsiooal luttice with ruces G, b - 36.

2. Brillouin ume, r«tangular lattice. A two-dimensional metal has one atom of \111.-
Icney onc in a simple rt!(;tangular prilllith,.;: cell a - 2 A. 1, - 4 A. (0) Ornw the first
Brillouin zone. Give its dimensions, in em - I , (1)) Calculate tIle nldius of the free
clcd:ron Feml! sphere, in em - I, (c) Omw this sillierI.' to SC'.uc on a drawing of the
first Brillouin 7Al11C. Make another sketch to show the first few periotls of the free
crt.:clron band in tl}c periodic zone scheme. fur both thc first and sl'COnd ene rgy
bands. Assume there is a small cnCIb'Y gap al the zone boundary.

3. Ile:tagonal clore-packed structure. C,.onsidcr the fin! Brillouin 7.onC of a cr}'lital with
a silnplc hexagonal lattice in thrL'C dimensions wilh IlIltice constant.<; a lind c. Let G c
denote the shorlcst r(:cipr'OClll latticc ,'(.'dor parallel to thc c axis of the crystal
laltice. (a) Show that for a hCJlab'Otlal cloS{'-packt.'(1 crystal structure thc Fourier
l'Omponent U(G c) of the l'J'}'slal potcntial U(r) is 7.Cro. (b) Is U(2G c) also zcro? (c)\Vhy
ls it possible in prirw.:iplc to obtain an insulator made up of div.J.!cnt atoms at the
bUicc points of a simple hcxilgonal 1a.Uk:c? (<I) Why is it not possiblc to obtain an
insul.ltor made up of monovalcnt atoms in a hexagolml-dost.... packl'tl structure?

4. Brillouin :one, of two-Jimenrional Jit."(llerlt metoi. A two-dimensional metal in the

form of a square lattil"e has two conduction electrons per atom. In tllC almost frt.'C
electron approximation, sketch carefull }' the electron and hole energy surfaces. For
the: eit.'ctron$ dwosc II zone sc:he:mc such that the Fermi surface is shown as ciQllt.'(\.

5. Open orbit,. An open orbit in a monovalent tctragon.·.1 metal oonnl'(.1s opposite

ract:s of the boundary of a Brillouin '1.one. TIle fllCCs are sqlarlltcd by C = 2 x
10" em- I. A magnetlt- field B ... l(il gallSlj := 1O- t tesla is normal to the plane of the
opcn orhit. (a) What is thc order of magnitude of tile peri()(l of the Illotion in k
SpIICC? Take 0 - 10" em/sec. (h) Describe in real space the motion of an electron on
this orbit in the pre5Cnce r:l tile magnetic field .

6. Collesi.:e energy for 41 tquare weU potential. (a) Find an expression for the binding
l'Ilergy r:l an eit.-ctron in one dimcnsio:l in a single square wt:1I of depth Ua and widtll
!I. (This i~ the standanl first problem in elementary quantum mechan ics.) Assume
that the solution is symmetric about the midpoi nt of the wd l. (b) Find a num(!rjc-.ll
n.-sult for the bilxling energy in tenns of Uo lOr the spt!cial U\SC lu~ ... 2Nl/ma l and
l'(lmparc with the appropriate limit of Fig. 20. In this limit of wWcly !K:pnratt.'(1 wells
the hand wid th goes to zero, so thc energy for k ,.. 0 is the samc as the e nergy fOl"
any other k in the l{)W(!St energy band. Other bands arc forml-d from the excited
statl'S of the well, in this limit.

7. Dc lIaa.-LYln Alphen period of potauiurn. (a) Cak.'u late the period 6 (IIB) expcclcJ
fOr pot~ium on thc free dcc;tron modd. (h) What is the area in real space d the
extremal orbit. ror 8 - 10 kG - 1 T? TIle same period apillics to osc:iI]ations in the
c1t.'CIriaJ resistivity. known as the Shubnikow-de Haas ('IT--et.
(
 9 "ami Sur/flcn flfld Mefllu 267

.8. Bond edge ,tructure on k· P peTturootion theory. Consider a nonde generatc or.
bital V1"}" at k - 0 ir the band n of a c\lbic ~tal. Usc sccond-onk-r perturbation
thL'Ory tu Gnd the result

(39)

where the Slim b OvtJ r all other orhiW.ls V1Jl. at k - O. The effi."CtivtJ mass at this
point is

..!:!...:o 1 + .!.. L' l(nOh>l.iO>I~ . (40)

m· 111 J f"(O) fj.O)
The mass at tll<: l'Onduction l;ond L"tige in a narrow gap semiconductor is oftell
duminatL-cl by the effi.'Ct r:l tll<: vak'llCC band L"tige, whence

(4 1)

where the 5l1m is over the vak'llCC hands; E~ is the energy gap. For gi"e n mollrix
clements, small galls lead to small ma!;SCs.

9. Wormier functions. The Wannier fundions of a band are definl-d in terms of the
Bloch functions of the same band by

w(r - rJ = N - 1I2 L, exp(- ik· r.) V1I.(r) (<2)

where r" is a lattice point. (a) Prove that Wannier functions abou t different lattice
points n,m arc orthogonal

" rm. (43)

This orthogonality llroperty makL'S the functiuns often nf grL'IIte r usc than atomic
orbitals ccnterL-d on differcnt lattice sites, because tile laller are not generally or-
thogorwl. (b) The Wannier functions are peaked aroun(lthe lattice sites. Show tha t
for 41. - N- 1!2 rfh udX) the Wannier function is

kJr N atoms on a line of lattice constant o.

10. Opm orbit$ and magnetoresistance. We considered the transvCrse magnetoresis·

tanL'e of(rcc c1L'Ctrons in Problem 6.9 and of electrons and holes in Prohle m 8 . .5. In
sOllie crystals the magru.:toresistance saturates eXl.'t:llt in sllL'Cial crystal orientations.
An open orbit carries current only ill a single din..ction in the plane normal to the
magnetic field : such C8lTiers arc not defiL'L,tc(1 by the Geld. In the arrangement of
Fig. 6.13, kt the opcn orhi.ts be parallel tu k. : in real space these orbits carry
us
(
current parallel 10 the !lll.~b. Lei un = gUo he the conductivity of the ope n orbits;
this <k.fincs the constant ,o TIle magnctocondultivlty tcoSOr in the h igh field limit
WeT» I is

willI Q _ WeT. (a) Show that the lIall ficld is 1:::~ .. -1:::.IIQ. (b) Show tllal the eff(.,<:-
tive resistivity in the % dirx.'CIion is p" «(jlIU(XS/f + I), so thaI the K'Si~ivity docs
not saturate, but increast..'S as 8~.

-II. Landau levels. The vt.'d or potential ora uniform magnetic field Ri is A " - Byi in
the Landau gauge. The hamiltonian of a free cit..'Ctron without spin is

H ::. -(Ii~!2JII}(ot/a" + ~/a-?; + (1121»)[ -ilia/ax - c!lBlc]' .

We will look fOr an cigcnfum:tioo of the wave equation H.p .. t'.p in IIle form

1/1 - MY) c"p[i{k.r + kaz))

(a) SIIOW thaI M!I) salisfit,.'S th e l'qualion

(,.if'lm)tPJld,r + h' - (li i I;!/2m) - lm~ 1I - !lJllx" 0

wllcrc ~ - eBlme ilI1cl Yo" dlkJeR (b) Show thai this is the wave (.'<luation of a
hanllQllic oscillator with frt:qucnl,), We. where

References
T. E. .'abeT. Introduction 10 tile llleOI"JI of f«"l,,/d mettlb, Aoldcmic Press. 1974.
W. A. Ibrri""", Electronic Itnu:'",. t1nd the propcrtiu of .wlldl, . ·n:eman. 1980.
, . M. Zhnlln, Principles of tIre fllWrIJ of .wlids. 2nd cd .• Cambridge, 1972.
V. 1_ Mnruu.i. J. F. Jan"". ~nd A. R. Williams. Ctlk"lale,1 eieclMl1e properties of nu!tah, Per-
Ilan"lnn, 1978. Compilation.
A. B. Pillpa,-d , Mogndoraisw1ta ,,, rnetnh, Cambridge, JIl89.
W. A. de Hccr, W. D. Knigh t, M. Y. Chou, and M. L.. Col.t:.... ~EJcctronic shell structure and
metal clust.:ts, ~ Solid Illltc ph),sics, 40, 93 (1987).
FF.RMl SURFAL"'E$
A. P. CrackncU lind K. C . Wong. Fermi '''ifac£. OxfOrd , 1973.
L. M. FiliiCU\'. " ~'enni surface studies:' Eleclroru In cl"JI&lnlllne wlids, IAEA. Vlcrl11ll. 1973.
I. M. Lifu'H~.• M. Va. A:dx:I. and M. I. Kaganuv, Electronic theory of Imtoll;, Consultunl.$ Bureau,
1973.
A. B. P;pp.onI, Vvrwmicl of conducllon eledrom. Gordon lind Bread>. 1965.
1'. B. ViuclK"r and 1- M. Fal icov, - Fen"i surface properties of m....tals. ~ Fb)·s. Stal us Solid i B 54, 9
(1972). 100.... to Ferm i IUri"atts IUld '-nd structures.
M. Springford, cd., Electronl Ol I~ FennilUrfoc:r, Cambridge. 1960.
D. Shoo:nbcrg. Magnetic O$d/ll1lionr in I/IcUlb. ea."bridgc. 198-1.
C. R. St~"Wart, "' kavy fermion s)·stcm •• ~ Revi. Mod. I"')'s. 56, 755 (I984).
A C . Hewson, Kondo problem 10 I~JV!I f enuloWl. Cambridge, 1993.
l'SEll DOPOTEf'iTIALS
Solid ~tlltc phpics 24 (I97Oj«lf\lains three articles oro Im:udopolcnt ial~ by V. Heine. M. l... CohCll,
and O . Wea;..,.
 10
Plasmons, Polaritons, and Polarons

DIELECfRIC FUNCTION OF THE ELEcrRON GAS 271

Plasma optics 212
Dispersion relation for electromagnetic W',1ve5 273
Transverse optical modes in a plasma 214
Tran.~parcncr of alkali metals in the ultraviolet 214
Longitudinal plasma oscillations 216
PLASMONS 218
ELECntoSTATlC SCREEI\'lNG 280
Screened coulomb potential 282
Pseudopotential compor'K:nl U(O) 283
Molt meml-insulator trnnsition 284
Screening and phonons in metals 286

POLARITONS 287
LST relation 291

ELECll\ON·ELECfRON INTERAGI10N 294

Fermi ii(luid 204

ELECTRON·PHONON 1"'TERAGnON: POLARONS 207

PEIERi.S INSTABILITY OF UNEAH METAl.S 300

SUMMAHV 301

PR08LEMS 302

I. Surface plasmons 302

2. Interface plasmons 302
3. Alfvtn W.lVC5 302
,. Helicon waves 303
5. ptasmQn mode of a sphere 303
6. Magneloplasma frt.''Qucocy
7. Photon bra nd, at low Wi\vt.'vt.'(.1or
,.,
303

•• Plasma frequl.'1lC)' and ekaTkal conductivity 303

". Hulk modulu.s o f the Fenni gas 303
10. Hesponsc of e\tttron gas
II. Cap plasmons ano the van oer Waals interact ion ,..
""
REFERENCES ~

NOTE, nlC h:~t and problems or this chaptL'r a»umc fM:ility in the use or ek'(.i:Tomag-
ndic thL'Ory at the k'Vclaa good senior COUr.iC.
 ...
,, ~--~~----~----~----~--
0,5 1.5
," "
•
I ".

-, I
j
,I
,I
I
_. L _ __
\
Figure L Dielectric function c(...} of a free cl.,ctron gas \-.:rs ... frequency In units of 1M pluma
frequency "'... E lectromagr>ctic WilYC$ Iln>pagale without damplngr" ..... hen .. is ~itive.nd real.
t: k·(:tromagnct." wavc) arc.: 1()ldlly rdl~'CIed fro,,, the Illediu m "It is ncg;.tive .

. ___________________
_
c
n
~
- CHAPTER 10: PLASMONS, POLARITONS, AND POLARQNS

DIELECTIUC FUNCTION OF TilE ELECTnON GAS

The dielectric flmction ~w,K) of the electron gas, with its strong dcpcnd-
en<.'C on frequ ency and W3vcvector, has significant consequences for the physi-
cal properties of solids. In one limit, E(w,O) describes the collc(.1ivc excitations
of the Fenni sea-the volume and suruu..-e plasmons. In another limit, E(O,K)
dcscribes the ek"Ctrostatic screening of the electron-electron. electron-lattice,
and c!cdron-impurity interactions in crystals.
We use the dick"Ctric fUllclion of an ionic crystal to derive the polariton
spectrum. Later we discuss the properties of polarons. But nrst we arc COIl-
ccm(.>ti with tile electron gas in metals.
Dejinition1l of the Dielectric Function. The d ick>ctric constant f' of electro-
statics is defined in lenns of the electric field E and the polari7.ation P, the
dipole momcnt density;

(CCS) D = E + 47TP = EE (51) 0 :::: E'oE + I) IC EEoE (l )

Thus defined. E is also known as the relativc permittivity.

The introduction of the dispial.'Cment D is motivated bl' the usefulness of
this vector related to the c"temal appli(!d charge density PU.I in the same way as
E is related to the total charge dcnsity P = Pexl + Ph..!. where Plnd is the charge
dt.'11sity induced in the system by Pcxl.
Thus the divergcnce relation of thc electric field is

(CGS)
div 0 = div EE = 47TP"11 div D = cliv EEOE =. Pco. l (2)

div E "" 4np = 4niAxI + Piool rov E =plEo = (P"11 + Pm.lYEo (3)

Paris oflhis chapter will be written in CCS; toohlain results in SI, write I/Eo for
'w.
."
(

\\le shall nl.'(:d relations bctw(."Cn th e Fourier colnponcnl5 of 0 , E, p , and the

clt.drostatic potential 1(1. For hn:vit)' we do out exhibit here the freque ncy dependence.
Define E(K) such that

("')
tm.'fl (3) ]x.-comes
<.Iiv E .. tliv I E(K) eiK... - 411" I p(K) !!,,'r • (31))

and (2) 1)(.'C;Qrncs

div 0,. di ... I E(K)E(K) e'J;,. - 417 I Pnt(l..l~" . (&)

Each of the .:qllations must be satisfit.'tI term by h.:rrn ; we divide one by the other to
obtain

~K) "" A-,,(K) .., I _ Ar,.J(K) (3d)

p(K) p().1 .
The clt:ctrostatic potcntirulpr,' dt:.fint:d by - Vtp..., .. D satisfies the Puiswn <-'qlla-
tion V2 /f'.... "" - 41Tp.,,, ; and the ck'Ctrostatic potcntialljl dc6nt.-d by - Vip "" E satisfies
Va"., = - 4r.p. The Fourier compont.'nts U the potentials must therefore satisfy

(3<)

by (3d), We use this relation in the treatment of the $Ul..-en<•.J coulomb potential.

plaslTUl O"tiC8
The long wavelength dielectric response f(W,O) or IE("') of an electron gas is
obtained from the equation of motion of a free electron in an electric fie ld:
d'r
fII dt 2 = -eE . (4)

If x and ,.~ have the time dependencc e- I "". then

-w2mx = - eE ; x = eElm(,r . (5)
The dipole moment of one ciochon is - ex "" -tilElmtJ1-, and the polaril"..ation,
defined as the dipole moment per unit volume, is

P =- 'lex = - --E
mw2
," (6)

where '1 is the electron conccntration.

 JO Pia.!mOl'LS, Polunlom, and PoWr<)'" 273
(
The dielectric function at frcqucm.y w is

D(w) P(w)
(CGS) f (w) - - - - l
E(w)
+ 4'7T- - '
E(w) ,

, (w) = D(G) pew)

(51) := 1 + - - -.
lEoE(w) ·oE(w)
The dielectric function of the free electron gas foll ows from (6) and (7):

47Tflt?'
(CGS) <{w) = 1 - "..! iSQ «w) = I (8)

The plasma fn."Qucncy wI' is defined by the relation

~...,.--..,
(ees) w! - 41Tnt!lm ; (SI) ~ _ n~t/~1I (9)

A plasma is a Inedium with equal concentration of positive and negative

d 'llll"ges, of which at Ica.~1 onc charge type is mobile. In a solid the negative
dmrgcs of the conduction cla1.rons arc balanced by an equal com:cnlrollion of
positive charge of the ion cores. We w r ile the dielectric function (8) as

I«W)= I - ~ , I (lO)

plotted in Fig. 1.
If the positive ion core background ha.~ a dick"Clric COllslnnt labcl(,.>(l f(OO)
t:ssl..... lially constant up to frequencies well a\xwe wI" then (8) becomes
E(w) == ,,(00) - 41Tnt!lmWl = £(oo){ l - wVw2] (ll)
wh{.'re wI' is defined as
(12)
Nolice that Ii "'" 0 3t w = wp-

Dispersion Relation for Electromagnetec WaveB

In a nonmagnelic isotropic medium tbe electromagnetic wuve equation is

(13)

We look fur il solution with E <X cxp(- iwt) exp(iK ' r) and D = E(w, K)£ ; the n
we have the dispersion relation for electromagnetic waves:

(CGS) o(w,K)w' = C'K' ; (14)

'"
This relation tells us a great deal. Consider
• If real and> O. For w real. K is real and a transverse electromagnetic wave
propagates with the phase velocity dE-lit,
• If real and < O. For w real, K is imaginary and the wave is damped with a
characteristic length l~KI .
• E complex. For w real, K is complex and the waves are ~ampcd in space.
• E = eo, This means the system has a flnit e resp:mse in the absence of an
applied force; thus the poles of ~w.K) deflne the frequencies of the free
oscillations of the medium.
• If = O. We shall see thai longitudinally polarized waves are possible only at
the u:ros of E.

Tnll1..!verse Optiazl Mrxk. in a PlunuJ

The dispersion relation (14) becomes, with (11) Cor E{w),
(CCS) (15)
For w < wp we have K2 < 0, so that K is imaginary. TOe solutions of the wave
equation are of the form exp( -IKIx) in the frequency region 0 < w :S w".
Waves incident on the medium in this frequency region do not propagate, but
will be totally reflected.
An electron gas is transparent when w > wp • for here the dielectric f\lIlC-
tion is positive real. The dispersion relation in this region may be written as
(CCS) (16)
this describes transverse electromagnetic waves in a plasma (Fig. 2).
Values of the plasma frequency wp and of the free space wavelength
Ap • 2m/wp for electron concentrations of interest are given below. A wave
will propagate if its free space wavelength is less than Ap; otherwise the wave is
reflected.
n, electrolls/cm3 10" 1018 1014 1010
W • S-1 5.7 x lOiS 5.7 X LO I3 5.7 X 1011 5.7xlcf
p
Ap. cm 3.3 X 10- 05 3.3 X 10- 3 0.33 33

Transparency of Alkali Metal. in 1M Ultraviolet. From the preceding dis-

cussion of the dielectric function we conclude that simple metaJs should reflect
light in the visible region and be transparent to ultraviolet light. 1be effect ",'llS
discovered by wood and explained by Zener. A comparison of calculated and
observed cutoff wavelengths is given in Table L The reflection of light from a
metaJ is entirely similar to the reflection of radio waves from the ionosphere. for
the free electrons in the ionosphere make the dielectric constant negative at
low frequencies. Experimental results for InSb wit1\ =4 X 10 18 cm- 3 are
shown in Fig. 3, where the plasma frequency is near _.119 cV.
 10 Pkmnon.., Polari,on.., oM Pola roN 2.5

•
",
Figure 2 DislJOenion relation for trans" cnecJectronw.gnet ic wan's in _,)lasma. The group velocit)'
". .. d""dK is the slope of the d.spersion cun,<:. Although t he dit:kctlk function is lJt,twttn n:ro
and une, th e group velocit)' is len than the \lCIQelt)' of light in vacuu ul .

F""UI"e 3 Rt..fk-ctance ofiodium antimonide with

n - 4 )( 10" ~ _3. (A rter J. N. Hodgson.)

u N. K Rb c•
•
Ap , calculated. mass m 1550 '090 287. 3""
Ap , observed 15S0 2100 31SO 3400
276

Longitudinal Plo8ma OsclllatiOM

The zeros of the dielectric fundion determine the frequencies of the longi-
tudinal modes of oscillation. That is, the condition
E(W,] = 0 (17)
determines the longitudinal frequency WI. ncar K = O.
By the geometry of a longitudinal polarization wave there is a depolari7.a-
lion ficld E - 41TP, discussed below. Thus 0 - E + 41TP = 0 fot'" a longitudi_
CI

nal wave in a plasma or more generally in a clysta). In SI units, 0 == EoE +

P=O.
For an electron gas, at the 7.cro of the dielectric function

E(WJ = 1 - wi/wt = 0 , (18)

whence W,. = Wp. Thus there is a free longitudinal oscillation mode (Fig. 4) of an
electron gas at the plasma frequency described by (15) as the low-frequency
cutoff of transverse electromagnetic wavcs.
A longitudinal plasma oscillation with K = 0 is shown in Fig. 5 asa uniform
displacemcnt of an electron gas in a thin metallic slab. The electron gas is
movcd as a whole with respect 10 the posilive ion background. The displace-
ment II of the ele<.1ron gas creates an electric fie ld E = 41T1lC1i thai acts /.IS a
restoring force on the gas.
111e equation of motion of a unit volume of the electron gas of concentra-
tion II is

(eGS) (19)

0'

(CCS)
d'u
-dt'- + W•211=O W .~
41Tncl)
(- --
m
'" . (20)

This is the equation of motion of a simplc harmonic oscillator of frequency w,..

the plasma fl·equenc),. -nlC expression for w" is identical y,ith (9), which arose in
a diffe rcnt connection. In SI, the IDsploc'Cmcnt II creatcs the electric fie ld E =
mJII/Eo, whence w,, = (nr/Eom)ll2.
A plasma oscillation of small wavcvectot.has approximately the frequency
wjl' The wavevector dependence of the dispersion relation for longitudinal
oscillations in a Fermi gas is given by

(21)

where VF is the electron velocity at the Fermi energy.

 lO PLa,mom, PolaritON, and PolaroN 277

"b _

•
Figu~" A plasma oscillation. The arrows Ind)rnt~ the direction of displllC(,mcnt oft~ electrons.

,
'. : t· ' .. ,.:,
...
-t-
. .................
........
.. .. ..... "" ' .....

.. ,
..
)

.. ... •
• • .. -o. :

Sur!at.. '-!wlll' d.....,II)'

., - - .....

., .. +"...,
I,'igure S In (3) is shown a thin slab or film of a m~t .. l. A cross Kdion III shovo... in (bl, "ith the
pa<ith'c Ion cores indicated b)' + signs:and the eledron sea indimtc,d by the gray background_ The
dab is ck:drically neutral. In (el the negative charge has been displaced ul"''IIrd uniformly by a
s"WI dis13nce U , shown exagg~rated in the figu re. M in (d), this diSl'larem<''1lt establishes a surface
charge d~nslty - f l f t j on the upper sur&ct: of the slab and +nl"U 00 tm:.. 1_"<.... sur1'x'C.". "I~re II Is
the electron COflC'mtr.ltion. An e1ednc rlCld f-: .. 4~ i1 productd inside the dab This rleld trod s
10 ~tOI"l." the e kClron sea to its equilibrium position (h). I n SI \lnits, f-: - roeu/€".

...
27'

PLASMONS

A plasma oscillation in a metal is a collective longitudinal excitation of the

conduction electron gas. A plasmon is a quantum of a pldSma oscillation; we
may excite a plasmon by passing an e1cctron through a thin metallic film (Figli.
Ii and 1) or by renccting an electron or a photon from a film . The charge of the
electron couples with the electrostatic field nuctuations of the plasma oscilla-
tions. The reflected or transmitted electron will show an energy loss equal to
integral multiples of the plasmon energy.
Experimental excitation spectra for AI and Mg are shown in Fig. 8. A
comparison of observed and calculated values of plasmon energies is given in
Table 2; further data are given in the reviews by Raether and by Daniels. Recall
that wp as deflned by (12) includes the ion core effects by lise of £(co).
It is equally possible to excite collective plasma oscillations in dielectric
mms; results for several dielectrk-s are included. The calculated plasma ener-
gies of Si, Ce, and luSb are based on four valenl..'C electrons per atom. In a
dielectric the plasma oscillation is physically tbe same as in a metal: the entire
valence electron sea oscillates back and forth with respect to the iOIl (.'Orcs.

Table 2 Volume plasmon energies, in eV

Calculated
Material Ohscl"ved
'". "w,.
MetCl/s
U 7.12 8.02 7.96
N. 5.71 5.95 5.58
K 3.72 4.29 3.86
r.lg 10.6 10.9
AI 15.3 15.8
Dielectrics
Si 16.4- 16.9 16.0
G. 16.0-16.4 16.0
InSb 12.0-13.0 12.0

(
 -- .'"""

!M::>tlt'l'<'<l ded",,,

figure 6 Creation of a plasmon in II metal film by iudas!ic srnl-

It,'fing of an ck"drun . The incioc'fll electron lyp;.,.."]I)' has an CIl-
<'lJ;Y I 10 10 keY, th., plasmon cncr-gy may tJe of the order of
10 eV. An evt,nt is al-o shown in which tW1) plasmon. arc creat(,d.

l\fianling
OM

l'igu .... 7 A spectrometer wit h cl.,ctrostatic analyzer

lOr the study of plasmon c)<Citation by clectron~.
(Alter J. Daniels et a1.)

" "
~
, ", j
",
-~ , .~ ,
.1 j"
~ •
, • ,•
'I>
'" .. .,
E""'troo
50
"" '"
(''''''gy lou (eV)
,
" '" W
" '"
[lrc(ron .. -.:rgy bs. leV)
.,
'"
'0 (b)

Figure 8 Energy 10.... sp<:dra IOrde<:lrons ..<:fleet ..,.! from mm~ of (a) aluminum and (b) magnesium,
h- prnnary ck>dron (....crgics of2020 eV. The 12105s JX,aksnbscrved in Al arc made up of combina-
lions of 10.3 aoo 15.3 c V \oS§(." $, where the 10.3 cV ]<>liS is due 10 surfaec plasmon. and the 15.3 cV
loss is due to ,-ohm,., plasmon •. The len lou peaks olJSCn'(,d in Mg arc made up of combinations of
7. 1 cV surface plasmon. and 10.6 cV volume plasmoU$. Sumcc plasmon. arC the . ubj<.>ct of Prob·
"'m I. (Mtcr C . ] . pov.·c!1 and ] . B. Swan.)

L l
280
I
ELEcrROSTAUC SCREENINC

The electric field of a positive charge embedded in an electron gas falls off
with increasing r taster than lIr, becausc the electron gas tends to gather
arollnd and thus to screen the positive charge. 111e static screening can be
described by the waveveclor dependence of the static dielectric function
E(O,K). We consider the response of the electrons to an applied external electro-
static field. We start with a uniform gas of electrons of charge concentration
-noe superimposed on a background of positive charge of concentration nac.
Let the positive charge background be deformed mechanica1ly to produce a
sinusoidal variation of positive charge density in the x direction:
(22)
The term p",,(K) sin Kx gives rise to an electrostatic field that we call the exter-
nal field applied to the electron gas.
The electrostatic potential (f of a charge distribution is found from the
Poisson equation Vlrp == - 41TP, by (3) with E == -Vrp. For the positive charge
we have

P = p"",(K) sin Kx (23)

The Poisson equation gives the relation

K 2rp.,.,(K) = 41T~,(K) (24)
The electron gas will be deformed by the combined influences of the elec-
trostatic potential rp",,(K) of the positive charge distrihution and of the as yet
unknown induced electrostatic potential rp,nd(K) sin Kx of the deformation of
the electron gas itself. The electron charge density is
(25)
where Pind(K) is the amplitude of the charge density variation induced in the
electron gas. We want to find p,,\d(K) in terms of p".,(K).
The amplitude of the total electrostatic potential (j'{K) = rpu,(K) + rp'o<,(K)
of the positive and negative charge distributions is related to the total charge
density variation P(K) = p",,(K) + And(1<) by the Poisson equation. Then, as in
Eq. (24),
K'<P(K) = 4np(K) . (26)
To go further we need another equation that relates the electron con(.'Cn-
tration to the electrostatic potential. We develop this connection in what is
called the Thomas-Fermi approximation. The approximation consists in assum-
ing that a local internal chemical potential can be deflned as a function of the
electron con(.'Cntration at that point. Now the total chemical potential of the
electron gas is constant in equilibrium, independent of position. In a region
where there is no electrostatic contribution to the chemic-.t1 potential we have
 10 1'Ittsmon.., l'olDritooll. and Polo""", 281

FiSU",9 In thermal and diffusive cqui),briurn th" ch~mical potential Is constant , to maiotain it
COO'ISIant we incn.-ase the .,Iectron CO<IC'l'I'Itntion in regions of If.'Il'CC ",hC.'TC the potential (:nelllY is
low. and we d ..'Cf'eaSe th., concentrat ion where th~ pokn tial is hish·

IJ,<=
r.'
4- = -(3.,rno'F (27)
2m
at absolute zero, according to (6. 17). In a region where the electrostatic poten-
tial is Ip{x), the total chemical potential (Fig. 9) is constant and equal to
2
Iii 2 213 1i 2 213
I-' = ~,.{x) - e<p{x) a -[31f n{x}] - e!f(x) a - [31f nol (28)
2m 2m
ere E.I-{x) is the local value of the Fermi energy.
\.\o....

The expression (28) is valid for static electrostatic potentials that vary
slowly compared with the wavelength of an electron at the Fermi level; specifi-
cally. the approximation is Cf ~ k,.·. By a Taylor series expansion of ~F ' Eq. (28)
may be written as
dE,..
- l - (n(x) - no] a e'P\'x) ("')
'no
From (27) we have d~Fldno = 2l':F13no. whence
3 elPCx)
n(x) - no iii! -110-- . (30)
2 ~F

111e left-ll3nd side is the induced part of tbe electron concentration; thus
the Fourier components of this equation are
(3 1)
By (26) this becomes
(32)

By (3d) we have

~O K) = 1 - p"w(K) = 1 + ICIK 2 (33)

~, " K) •
he re, after some rearmngement,
k: _ 6'T171oC2fEF = 4(311fjl/"J 1I~13/00 c 47TCD(E.F) (34)
where ao Is the Bohr radius and D(EF) is the density of stales ror a free electron
gas. The approximation (33) ror t:(O, K) is called the Thomas-Fermi dielectric
funct ion, and Uk. is the Thomas·Fermi screening length, as in (40) below. For
copper with no = 8.5 x 1022 cm- 3 , the screening length is 0.55 A.
We have derived 1"',0 limiting expressions for the dielectric function of an
electron gas:
,1 ,
« O,K) = 1+ _ . «W,O) = 1 - ~2. ' (35)
K' ' W

We notice that f(O,K) as K _ 0 does not approach the same limit as t:(w,O) as
w _ O. This means that great care must be taken with the dielectric function
near the origin of the w-K plane. The full theory ror the general function E(w,K)
is due to Lindhard. I
Screened Coulomb Potential. We consider a point charge q placed in a sea
of conduction electrons. The Poisson equation for the unscreened coulomb
potential is
V'llpo "" - 411f/l3{r) , (36)
and we know that IPu = (fIr. lei us write
\PQ(r) "" (211)- 3 f dK ipo(K) exp(iK· r) (37)

We use in (36) the Fourier representation of Ihe delta functi on:

I3{r) = (271",-3 f dK exp(iK' r) • (38)
whence J<lipo(K) :: 411q .
By (30),

ruK)fq{K) = «K) ,
where fP(K) is the lotal or screened potential. We use E(K) in the Thomas-Fermi
fonn (33) to find
Owq
(39)
K2 + k!
The screened coulomb potential is the transfonn of rp(K):

4wq
rp(r) "" - (\3
211/
1-
0
dK
21TK2
2.
K + "-
.11
J'_) d(oos 8) exp(iKr cos 8)

=-
2q I'"' dK K&in Kr 2. = -
q
exp( - k~ r) (40)
11r 0 J(2 + k. ,.

I" IJ)Od discussion of t he undhard dielectric runction is given by J. Zimao. Prindpla ofrhe
Ilw:ory 0/ aclids, 2nd 00_, ~mbrldge, 1912. Chaplt.T 5. ".., algeLnJc steps in Ihe cwJuation rJ
Ziman'$ equation (5- 16) are given in dewl by C. KiUd, Solid slate phv<'''S l2, I (1968), Sedioo 6-
,f
 ,-
,
"

.,,r-"
,

i - 02

Fi!,'tire lOa Comparison of screened and un$<.TCt,"fI(,..J eoubmb potentials of a Sialic unit positive
charge. The screening length Uk, is set (."<jual to unity. TIle static sen..," ....! inl<.'t"aClion is takn in
too lhomas. Fcrmi a[lpro~imalion, which holds for low wa"c\/(,'Coors q <t!: k..; more romplete calcu-
lationS with all wavf,.vectors indud.:d (,,,hihit spatial oscilblions, called Frit.... d oscillations, in 2Jr. .. r
and are plotted in QTS, p. 114.

as in Fig. lOa. The screening parameter k~ is defined by (34). The exponential

factor reduces the range of the coulomb potential. nlC bare potential l/Ir is
obtained on letting the charge concentration flo _ 0, for then k. _ O. In the
vacuum limit 1f(K) "" 4wqlK?-.
Qne application of the screened interaction is to the resistivity of certain
alloys. 'niC atoms of the series Cu, Zn, Ca, Ce, As have valences 1, 2, 3, 4, 5.
An atom of Zn, Ca, Ceo or As added substitutionally to metallic eu has an
excess charge, referred to Cu, of I, 2, 3, or 4 if all the valence electrons join the
conduction band of the host metal. The foreign atom scatters the conduction
electrons, with an interaction given hy the screened coulomb potential. This
scattering contributes to the residual electrical resistivity. and calculations by
Mott of the resistivity increase are in fair agreement with experiment.
Pseudopotential Component U(O). In the legend to Fig. 9.22h we stated a
result that is important in pseudopotential theory: "For very small k the poten-
tial approaches - j times the Fermi energy. ,. The result, which is known as the
S(.'J'Cened ion limit for metals, can be derived from Eq. (39). When converted to
the potential energy of an electron of charge e in a metal of valency z with '10
iOns per unit volume, the potential energy component at k = 0 becomes
V(O) = -ezn!#'CO) = -4'1TZ¥lk~ (41)
The result (34) for k; in this situation reads
k; =- 6'1TZnrrIEF (42)
whence
V(O) = - JEF . (43)
 (

MoU Metal-ln8llwtor Transition

A crystal composed of one hydrogen atom per primitive cell should always
be a metal, according to the independent-electron model, because there will
always be a half-filled energy band within which charge transport can take
place. A crystal with one hydrogen molecule per primitiw cell is a different
matter, because the two electrons can fill a band. Under extreme high pres-
sure, as in the planet Jupiter. it is possible thai hydrogen occurs in a metallic
fonn.
But let us imagine a lattice of hydrogen atoms at absolute 7£ro: will this be
a metal or an insulator? The answer depends on tlle lattice constant, with small
values of 0 giving a metal and large "a!.ues gi\'ing an insulator. Mott" made an
earl y estimate of the critical value G" of the lattice constant that separates the
metallic state from the insulating state: lie "" 4.5ao, where Go "" h 2/rn,r is the
radius of the fi rst Bohr orbit of a hydrogen atom.
On one approach to the problem. we start in the metallic state where a
conduction e lectron sees a screened coulomb interaction from each proton:
U(r) "" -(rlr) exp( - k.r) , (44)
where k~ "" 3.939nJ13lao.
as in (34). where Ito is the electron concentration. At
high concentrations k. is large and the potential has no bound state. so thut we
must have a metal.
The potent ial is known 3 to have a bound state wben k, i~ smaller tllan
1.19lao. With a bound state possible the electrons may condense about the
protons to form an insulator. The inequality may be written in tenns of no as
(45)
3
With no"" 1/0 for a simple cuhic lattice. we may have an insulator wh en
oc> 2.7&0, which is not far from the Motl result 4.50 0 found in a dilferent
way.
'J"e lenn metal-insulator transition has come 10 denote situations where
the electrical conductivity of a material changes from metal to insulator as a
function of some external parameter, which may he composition. pressure.
strain, or magnetic field. The metallic pha.~e may usually he pictured in tenus
of an independent-electron model; the insulator phase may suggest important
electron-electron interactions. Sites randomly oc;c upk'1l introduce new and in-
teresting aspects to the problem. aspects that lie within percolation theory. The
percolation tranSition is beyond the scope of our book.

'for an t'XCCtk"TIt !'C"jew d the field of metal·insulator transition s. see N. F. MOlt. Pro<: It
Soc. London A 382. I (1982).
~F. J. Rogers, II. C. Cral"osk, Jr., and D. J. Harwood . Ph)'I. Rev. AI, ISn...(l970).
 10 Plaomono. Po!Dritons. orwf PoI4rOfU 285

w',-----,---,--_,, __---,

10'

IO·· o'-----+---'"----;----~
2 468

Figure lob Semilog plot of ob$er\'ed ·zero IL'Tllperalurtt conduct;";I), U{0) ' ·HSU S donoe <."QIlcen-
1.... llon n for phosphorous donon in $ilil."Ol\. (Aftee T. F. Rosenbaum , t:I al.)

When a semiconductor is doped with increasing concentrations of donor

(or acceptor) atoms, a transition will occur to a conducting metallic phase. Ex-
perimental results for P atoms in silicon are shown in Figure lOb. Here the
insulator-metal transition takes pial.'e when the concentration is so high that the
ground state W8vefunctions of electrons on neighboring impurity atoms overlap
Significantly.
The observed value of the critical concentration in the Si: P alloy system is
"c Ie 3.74 X 10 18 em- 3 , as in the Agure. If we take 32 X 10- 8 cm as the radius
oftlle ground state or a donor in Si in the spherical approximation, then by the
l>.'lott criterion ac ::0 1.44 X 10- 6 em. The P atoms are believed to occupy laUice
sites at random, but if instead their lattice were simple cubic, the critical Mott
concentration would be
nr = lid: = 0.33 x 10 18 em- 3 • (46)
appreciably less than the ohserved \'alue. It is usual in the semiconductor liter-
...
ature to refer to a hea\'ily-doped semiconductor in the metallic range as a
degenerate semiconductor.

Screening and Phonorns in Meta13

An interesting application of our h'lo limiting [onTIS of the dielectric func-
tion is to longitudinal acoustic phonons in metals. For longitudinal modes the
total dielectric function, ions plus electrons, must be zero, by (11). Provided
the sound velocity is lesS than the Fermi velocity of the electrons. we may use
for the eledrons the Thomas-Fermi dielectric function
(47)
Provided also that the ions are weU-spaced and mo\'e independently, we may
use for them the plasmon E{w,O) limit with the approximate mass M.
The total dielectric function, lattice plus electrons. but without the ek-c.
Ironic polarizability of the iOIl cores, is
47171r I.';
.(wK) = i - - - + - (48)
• M~ Kt.

At low K and w we neglect the term l. At a zero of t:(w,K) we have, with

t:F - lnlV¥,

0'
w = vK; v == (m/3M)112 VF • (50)
This describes long wavelength acoustic phonons.
In the alkali metals the result is in quite good agreement with the observed
longitudinal wave velocity. For potassium we calculate v = l.a x l~ em S- I;
the observed longitudinal sound , 'elocity at 4 K in the [100] diredion is 2.2 x
lOS cm 5- 1.
There is another 7.ero of t:(w,K) for positive iOn!; imbedded in an eJectron
sea. For high frequencies we use the dielectric contribution -u/;)t.JJ of the
electron gas:

l'::(w,O) = 1 (51)

and this function has a 7.ero when

I I I
-=-+- (52)
p. AI m
This is the e lectron plasma frequency, but with the reduce<! mass correction fOr
the motion of the positi\'e ions.
 10 Plasma .... PoIDritOnl. and PoIDr",.. 281

U~r-,,-r---------------------,
'/ , '"
,,
0.050 A"'w

,:
, """"'"
.""' A;;;-I--------~-~-~-~-=-:-=-;-·--,,~o..-'-'-..--. .-~-~-"1
>

,; 0.044
, , 0

·U.., ,,
...'
0'

...
U"'"
1:,
~:

,
,,,
01 O.!l 0.3 0.4 0.5 0.6 0.1 0.11

Figure 11 A l»ot of the observed energies .nd ",..,..,,-ectors of the polariton • .:md oftlu: 1.0
pI-oonons ;n CaP. 1lIc tJ,e()t't:tical dispersion ("\)0'0:$ lite shown by tho: solid Hnes. The: di.persion
CUO"l!S for the uncollpJ..-d Ilhonom; and photons aTe dlOYo"TI by the short, ..whed lines. (Arter C. II.
fh,nry and J. J. lIopfldd.)

POLARITONS

Longitudinal optical phonons and transverse optical phonons were dis·

cussed in Chapter 4, but we deferred treatment of the interaction of transverse
optica1 phonons with transverse electromagnetic waves. At resonance the
phonon-photon coupling entirely changes the character of the propagll.lion, and
a forbidden band is established for reasons that have nothing to do with the
periodicity of the lattice.
By resonance we mean a condition in which the frequencies and W3VeVI!C-
tors of both waves are approximately equal. Tile region of the crossovcr of the
two dashed curves in Fig. 11 is the resonance region; the two dashed curves are
the dispersion relations for photons and transverse optical phonons in the ab-
sence of any coupling between them. In reality, however, there al\\.'Sys is cou-
pling implicit in Maxwell's equations and expressed by the dielectric function .
TIle quantum of the coupled phonon-photon transverse wave Ocld is called a
polariton.
In this section we see how the coupling is responsible for the dispersion
relations shown as solid curves in the figure. All takes place at very low values
of the wa\'Cvcctor in l.:omparison \\;th a 7.one boundary, because at croSsover
w(photon) = ck(pboton) = w{phonon) ""' 1013 S-I; thus k .. 300 em - I.
An early warning: although the symhol WI- will necessarily arise in the
theory, the effects do not concern longitudinal optical phonons. Longitudinal
phonons do not cou"le to transverse photons in the bulk of a crystal.
(
...
TIle coupling of the electric field E of the photon with the dielectric polar-
ization P of the TO phonon is described by the e lectromagnetic W3\"e equation:

(CCS) (53)

At low wavevectors the TO phonon frequency wr is independent of K. TIle

polaril'.ation is proportional to the displacement of the poSitive ions relative to
the negath·t: ions, so that the equation of motion of the polarization is like that
of an osciUator and rna)' be written a.~. with P = Nqu,

(54)
where there aTC N ion pairs of effective charge q and reduced mass M, per unit
volume. For simplicity we neglect the electronic contribution to the polariza-
tion .
The equations (53) and (54) have a solution when

(55)

This gives the polariton dispersion relation. similar to that plotted in Figs. 11
and 12. At K = 0 there are tv.·o roots, W = 0 for the photon and
w2 "" w¥- + 4nNq2/M (56)

for the polariton. Here wr is the TO phonon frequency in the absence of cou-
pling with photons.
The dielectric function obtained from (54) is:
4'tTNq'lIM
e(w) = I + 41TPIE - 1 +W·r
. • (51)
w-

If there is bn opticaJ electronic contribution to the polarization from the ion

cores, this should be included. In the frequency range from zero up through
the infrared, we write

.(w) c 0(00) (58)

in accord with the definition of e(oo) as the optical dielectric constant, obtaint.>d
as the square of the optica1 refractive index.
We set w = 0 to obtain the static dielectric function :
(59)
which is combined with (58) to obtain E(W) in terms of acwssible parameters;
 (
,

, ,-
- - "'''' K
- - - - Mognllud~ vf K
" 'M" K is pure

...,. - )IoCIO" rad . - 1

..~ . II<IO"nd . - 1

--- --- --- .., -~

,K

.,
;
---- --- ----
.. 't-------~/''---_::::=::::::::::::-=-==-=-==-=-==-~~~-~~~,'~~;-=-~-

o 1.0
"
Figure 12 Coupk-d Il'IOdca of photons and l •• "m:rse optp] phonons in an Ioniccryslal. The flne
horizontal line represents oscilJ.lors offrequ"n~ in the abKroce of eollilling to tho: cl«1l"01T1l1&-
netic field , and the fine line labeled .., - cKlVt (GO) corTespond" 10 elc:d.ra.n-.gnetic _Vel in tho:
cryftal, but u"coupled 10 the lallie" oscil1alOU "'7. TIlt: ]m.v>, lines arc Ih" diipilnion .dations in
th" presence of coupling bt:t~"n Ih" IaUice oscillators aod the c!l-ctromagllcUc wa,'e. On" & ..'Cl of
Ih" COUI,]ing is 10 creale the frequency gap bet"",,, "'t. and "'1'": within this glIp the Wlt\<e>'fX'Ior is
pure imaginary of mlgIlitude g1~~ D)' the brok.eo!me in lhe figun:. In Ihe gap the",.,.., I Uenuates
as eJp(-IK\.t). and ".., see from lhe plot that ~ allmnalien is much stl'Orlgel" near ..... than ne.."'r...
The d"tr.II:te' ort},., bmncIoes wrie$ with K. there il;. region of mixl':d "le('lric-mechanicalaspeds
Ilear the nomhUlI crosson', •• Note , fonally, II Is Intuitively OO"ious that th., group ve locity oI' light in
th" medIum ill always <c, becans., the slope i)<.I~K for the i>Ctuai di$penion Il:laliorJ$ (heavy linc:s) is
C\,.,rywhc:re less than the slope c lOr the uncoupled photon In free space.

0'

E{w) =
w.}E(O) - WllE(O:O)
2 2 ::: E{o:o)
(wI.. - (2)
.2 lI' (60)
wr-W wr- w
TIle 7.£ro of E{w) defines the frequency Wt., as the pole of E(W) defineS Wor. The
zero gives
(61)
Waves do not propagate in the frequency region for which E(w) is negative,
between its pole at w .., Worand its zero at w ,., wt., as in Fig, 13. For negative E,
waves do not propagate heause then K is imaginary for real w, and
290

:J
2 _ _. (ot)

1 ,
•
-,

FiAure 13. Plot of E(...) from (60) fot" E('"'J .. 2 and f (O) "" 3. The dielectric C<.HIS!anl II n"p;"'"
betWftfJ Col - "'Tand "'r.- (3/'l)III"'T. that 1$, between the pole(inSnily) of E(w) and the zero of E(oo).
I l>Ciden' electromagnetic WII\ "et oHrequencies """ < OIl < "'r. will nol p~te in the medium, but
" "ill be n:(]e«ed at the bou ndary.

, ,
,'",
..
,•
,
•
0
,
- I
•,
- I0
-., .. ~.~.W - •••. ... '~~
,~

..
t-""!"""")" in lit

" Igure 13b I);ekclric function (rea1 IN-rt) ofSrF. me~uud O'o"er a wide frequency 1"lIlge. exhibit-
ing the dOXTea5e of the ionic polanu.billl), at high rreq...:nc~ {A. 11011 lIippd.}

(
 - E
.- ,
L,
f- ............

<I .-,
L.
- E
,t
...............
-,
... + ...... +

'I
ro phc...., WI""""""
F~'re ,. Relali,"" dh:placements rithe peniti"" . nd nt:gati\'e ions at one ilUtant ri hm., fora w:we
in an optical moue tnn:lingaJong the z axis. 1ne pll..",s of nodes (zero displacement) are .hown. for
long ,,",..eiength Ilhonons the Ol)(lalp1ant:S aresepanleci I,y many plancs Ualom,. In Ihe tranwcne
OI)tiaolphonon mode the l);article tlisplact:IT"'"t is perpl'ndicular to the ""'·e\Il...::Ior K, th .. macro-
KOiJic ekctrie field il) an intlnite medium will He on ly in Ihe :!: :c dirt.-....110n for the ll1ol><.Ie ~hown, and
by the s}'mmeby of the problem iJEjiJ:c - O. It rollowli Ihat div E - 0 ror a TO phomm. II) the
Iongitlldma] "ptio:.tl pllOnon mode the IlBrticlc di'IlI~lnel)ts and he"ce the diekctricpoiarizaliol) P
~ J*1OILe I !Q the _·C\·ector. The macroscopic rle(tnc Grid E satisfies D - E + • .,p .. 0 in CCS
or EuE'" P - 0 il) SI; by symm<:try E and P lin' panIlc\to the z axil. and iJEJ az 1"1 0. Thus
tli,' E 1"1 0 lOr an 1.0 phooon, and ~ ...) d,v E is !ero ody if ~ ...) - O.

exp(iKx)_ exp(-IKlx), damped in space. 'nle 7.ero of E(W}, by our earlier argu-
ment, is the LD fre<luency at low K. Fig. 14. Just as with the plasma frequency
wp , the frequency WL has two meanings, one as the LO fre(IUency at low K and
the other as the upper cutoff frequency of the forbidden band for propagation of
an electromagnetic wave. The value of WI. is identical at both frequencies.

LST Relation
We write (61) as

(62)

where 1:(0) is the static dielectric constant and E("") is the high-frequency limit of
the dielectric function, defined to include the core electron contribution. TIlis
result is the Lyddane-Sachs-Teller relation. The derivation assumed a cubic
crystal with two atoms per primitive cell .4 For soft modes with Wor- 0 we see
that E(O) _ DO, a characteristic of ferroel cctricity.
Undamped electromagnetic waves with frequencies within the gap cannot
propagate in a thick crystal. The reflectivity of a crystal surface is expected to be
high in this frequ ency region. as in Fig. 15.
'For the extenslol)to<:<>mpiex structures, see A. S. Barker, Jr., " Inrrared dielectric beha"iurof
ferroek:d.ric crystal~:' il) Ptrroeieclricl'!I, E. F. Weller. ed . . Elsevier, 1967. Barker gi"es an in te r-
attng deri"alion or the LST relalion b)' a causali ty argument.
... (
For films of thickness Jess than a wavelength the situation is changed.
Because for frequencies in the gap the wave attenuates as exp(- IKIx). it is
possible for the radiation to he transmitt(..c through a film for the small values of
jKI near WI.-. but for the large values oflKI near wr the wave will be reflected. By
refleaioll at nonnormw incidence the frequency WI,. of longitudinal opt imJ
phonons can he observed. as in Fig. 16.
Experimental values of E(O), E(O:», and Wr are given in Table 3, with va\-
ues of w,. calculated using the LST relation. Eq. (62). We compare values of

Table 3 Lattit'e frequencies, chiefly al 300 K

"'Ok Opti(:ai
•
die1ed:rlc dielectric

- -
""r'. in 10 13 5- ' (I),.• in 10'3 , - I
""',- constan t
CO Il ~lanl

LiH
LiF
«01

12.9
8.9
" -I
3.6
1.9
II.-
experimental

5.'
LST relation

21.
12.
Liel 12.0 2.7 3.6 7.5
UBr 13.2 3.' 3.0 6.1
Nat- 5.1 1.7 4.5 7.6
NaCI 5.9 2.25 3.1 5.0
NaBr 6A 2.6 2.5 3.9
KF 5.5 1.5 3.6 6. 1
KCI 4.65 2.1 27 4.0
Kl 5.1 27 1.9 2.6
nbF 6.5 1.9 2.9 SA
Itbl 5.5 2.6 IA ),9
C,CI 7.2 2.6 ),9 3. 1
Cd 5.65 3.0 1.2 1.6
11C! 31.9 5. 1 1.2 3.0
11B, 29.8 SA 0.81 1.9
AgCl 12.3 4.0 ),9 3A
AgBr 13.1 4.6 1.5 2.5
MgO 9.8 2.95 7.5 14.
C,P 10.7 8.5 6.9 7.6
CaA, 12.9 10.9 5. 1 5.5
C.Sb 16. 1 14.4 4.3 4.6
InP 12.4 9.6 5.7 6.5
InA$ 14.9 12.3 4.1 4.5
inSb 17.7 15.6 3.5 3.7
SiC 9.6 6.7 14.9 17.{1
C 5.5 5.5 25.1 25. 1
Si 11.7 1l .7 9.9 99
ee 15.8 15.8 5.7 5.7

I)ata largdy due 10 E. Bunleill.

 10 l'I.umon.. 1'0000riloolI, and f'oIllrotll 293

(
'OO,-------,---~='_~--r_--,_--.__,__.
,.- -=:~
I,," "
.. r!
I ,i
I /
--lOOt.:

---"" ,
------ 200 t.:

f !
I I
I ,J
,I
I:
!
.
. ..
Phd"" " ..,.:length in 10-' ...,
.. '00

Fi8u re 15 Rdl"ctance of aCTyStal ofNaCl at Jeveral ternller:illllres. \o't:nu~ wavelength. n.e oomi-
nal 'lIll1el of ...... and...,.. at room temperJlure c:orn:s~ 10 wan,lengths ci 38 and 61 )( IO-~ an.

.. ..
respecth·o:Iy. (After 10.. Mibllishi el aI.)
\\"'-': I\u",l ..... in cu,

,.,'000
i"I': r
. ..I,.,. .•. ~ ,
I
I
I
• 4-
I
I
,
" P 11U!arir.ot...,
I • • ,lOJbri>OlIia,

.6
I ,
I I
I I
I
. I

I
I Tr..l'I'~

,•• r=·
I."ngil.,.]ina!
,,I
V opt;...!
, 12..5
,hmoo
~. 6
I
.,
'" "
Fizure 16 R"flectarK'e ,·.,nus wa".,kngth cli UF film bacl<ed by sih-er, for radiotlon i",tid.:nt nl"lll"
30". The longitudinal optical phonon abJorl» strongly the radi~t ion polariZl.,J (P) in th" ilia""
normal to th" film , but absorbs hard I)' at all th" miatlon polari7.ed (I) IlIUlIliel to the film. (After
D. W. B"rr.,man.)
 c
,,.
w,}wr obtained by inelastic neutro n scatteri ng with experim e ntal values of
(t:(O)/t:{oo)l'f2 obtained by d ielectric measurements;
NaI KU, eaAs
w,}w.( 1. 44 ± 0.05 1. 39 ± 0.02 1.07 ± 0.02
[E(0)/£(00)JII2 1.45 ± 0.03 1.38 ± 0.03 1. 08
The agreemen t with the LST relation is excellent.

ELECTRON-ELECTRON INTt;RACTlON

Fermi Liquid. Because of the inte raction of the conduction e lectrons with
each other through the ir electrostatic inte raction, the e1cctrons suffer colli-
sions. Further, a moving electron causes an ine rtial reaction in the surrou nding
electron gas, thereby increasing the effcctive mass of the electron. The effects
of electron-electron interactions are usually described within the framework of
the Landau theory of a Ferm i liquid.!> The ohject of the theory is to give a
unified account oCthe effect ofintcrnctions. A Fermi gas is a system of no ninter-
acting fe nnions; the same system with interactions is a Fermi liq uid.
lAndau's theory gives a good account of the low-lying single particle exci-
tations of the system of interacting electrons. These Single particle excitations
are called quasiparticles; they have a one-to-one correspondellce with the siu-
g1e particle excitations of the free electron gas, A quasiparticle may be t hought
of as a Single particle accompanied by a distortion cloud in the electron gas.
One effect of the coulomb interactions between electrons is to change the effec-
tive mass of the electron; in the alkali metals the increase is roughly of the onler
of 25 percent.
Electron-Electron CoUision!. It is an astonishing property of meta ls that
cond uction e lectrons, although crowded togethe r only 2 A apart, trave l long
d istanecs octween collisions with each othe r. 111e mean free paths for e1cctron-
electron collisiollS are longer than 104 A at room temperature and longer than
1OcmatlK.
Two fuctors are responsible for these long mean free paths, without which
the free electron model of metals would have little value. 'The most powerful
factor is the exclusion princi ple (Fig. 17), and the second factor is the scr eening
of the coulomb interaction between two electrons.
We show how the exclusion principle reduces the collision freq uency of an
electron that has a low excitation e ne rgy E I outside a filled Fermi sphe re
(F ig. 18). We esti mate the effect of the exclusion principle on the two-booy
collision 1 + 2 _ 3 + 4 he tween an electron in the excited orbital 1 and an
~ L Landau, '-I'he<lry of a Ferm i liqu id," Sovict Ph)'llics I ETP 3, 920 (1957); L. Landau,
"Os!;illalioos in a Ferm i liquid," S<;.viel Physic1 I ETP 5. IO (1957); see also D. Pines and
P. Nmiercs, Theory of quontum liquids, Benjamin. 1966. Vol. I.
 •

!
figure 17 A collision between t...u electrons of
t
wavcvcctor- k, and k,. After the collision the
!Wlides ha~"1l Wll.".:'Vttlor 1<, 1lIId k.,. 'The Pauli
eoo:clu.ioo principle .11",,'5 collisions only 10 final
.t~es k... Ie. which weI"() vacan t hcrure the ('<lli,-
~ .....

., • ,
-t~~~+-~~-+-, .

•• .,
(0) ~)

•,
Spll,-,.. of
final >tolla

'0'
figure 18 In {Ill lhe dechu'" in initial orbitals I and 2 collide. If the orhitals 3 and" lIJ"C inihally
\'8Cant. the electrons I and 2 ~ occupy orbitals 31Uld 4 &lin the collision. Energy and rnOml'nlum
Iln:! C'UnSf:t".-OO. In (b) the da1R)1l1 in inilialotbilab I and 2 have no ...acanllinai urbitals a\-aib.bIc
thai . llow energy to he (."OfOSeI'Ved in lhe collision. Orbitals such as 3 and 4 would COIl~'C energy
and momen tum, bullhcy are iLl reatly lilled wilh other electrons. In (e) we havedenotoo with x the
~~Ior of the ~nt~ gf mau I and 2. All polin of Ofbitals 3 and 4 wnoe ....'c momen tum and
energy If lhcy lie at tJPllOSile c'l(b <:l a diaoneler of the small sphere. The small spllel"<: was d"'''l1
from the «n ler of rrIIW to pa.s through I and 2. [lui not all pairs of poin!s 3, 4 ar~ allowed by the
l'.1Clu.1un principle, lOr both 3, 4 must lie oubide the Fenni 5phere. the fno::tion alkM-.:d iii -(l ,lt£~ .

\
electron in the filled orbitaJ 2 in the Fenni sea. It is convenient to refer all
energies 10 the Fermi level ~ taken as the zero or energy; thus EI will be
positive and E2 will be negative. Because of the exclusion principle the orbitals
3 and 4 of the electrons after collision must lie oulside the Fenni sphere, all
orbitals within the sphere being already occupied; thus both energics E3, E.
must he positive referred to zero on the Fenni sphere. '
The conservation of ellergy rC(luires that lEd < EI, for otherwise
E3 + E.t :::: EI + E2 could not be positive. This means that collisions are possible
only j£the orhital 2 lies within a shell of thickness EI within the Fermi sUiface,
as in Fig. ISa. Thus the fraction -EI/E~' of the electrons in filled orhitals pm-
vides a suitable target for ekelron 1. But even if the target e lectron 2 is in the
suitable energy shell, only a small fraction of the final orbitals compatible with
coilservatiOll of encrgy and momentum are allowed by the exclusion principle.
11tis gives a set:ond factor of EI/EI"
In Fig. 1& we show a small sphere on which all pairs of orbitals 3, .4 at
opposite ends of a diameter satisfy the conservation laws, but collisions can
occuronJy if both orbitals 3, 4 lie outside the Fermi sea. 11te product of the two
fractions is (EI/EF'f. If EI corresponds to 1 K and E~' to 5 x Jat K, we have
(EI/E~''f - .4 X 10- 10, the factor by which the exclusion principle reduces the
collision rat('.
The argument is not changed for a thermal distribution of electrons at a
low temperature s1Ieh that kilT <C EF . We replace EI by the thermal energy
- ku'f, and now the rate at wh ich e1ectron·electron collisions take place is re-
duccd below the dassical value by (k nT/Efl, so that the effective collision cross
section 0' is
(63)
where 0'0 is the cross section for the electron-clectron interaction.
The interaction of one electron with anolher has a range of the order of the
screening lellb>th Uk~ as in (34). Numerical calculatiOlls give the effect ive cross
section with screen ing for collisions between electrons as of the order of
IO- IS em 2 or 10 A2 in typical metals, The effect of the electron b'3S background
in elcctron-clcclron collisions is to reduce 0'0 below the value expected from the
Rutherford scattering equation for the unSCfccned coulomb potential . How-
ever, much the greatest reduction in the cross scctiOll is caused hy the Pauli
factor (kB'fIEF'f .
At room temperature in a typical metal ka'fIEf' is _10- 2 , so that
4 19
0' - 10- 0'0 - 10- em 2 . The mean free path for electron-eh."Ctron collisions
4
is f .. 1I11u - 10- em at room temperature. This is longer than the mean free
path due to electron-phonon collisions by at least a ractOl' of 10" so that at room
tempcrature collisions with phonons are likely to be dominalli. At liquid he-
lium temperatures a contribution proportional to 'f2 has l)('Cn found in the
 (
resistivity of a number of metals, consistent with the form orthe electron-elec-
tron scattering cross section (63). The mean free path of electrons in indium at
2 K is of the oroer of 30 em, as expected from (63). Thus the Pauli principle
c,,"plains one of the central questions of the theory of metals: how do the elec-
trons travel long distances without colliding with each other?

ELECfRON-PHONON INTERACIlON: POI.ARONS

111c most common effect of the electron-phonon interaction is seen in the

or
temperature dependence the electrical resistivity. which for pure copper is
1.55 microhm-em at we and 2.28 microhm-cm at 1()(f'C. The electrons arc
scatteretl by the phOIlOIlS, and the higher the temperature, the more phollons
there arc and hencc more scattering. Above the Ochye temperature the num-
ber of thermal phonolls is roughly proportional to the absolute temperature,
and we find that thc resistivity increases apprmcimately as the absolute te mper-
ature in any reasonably pure metal in this temperature region.
A more subtle effect of the electron-phonon intcmction is the apparent
incrcase in electron mass that occurs because the electron drags the heavy ion
cores along with it. In an insulator the combination of the electron and its strain
field is known as a polaron,6 Fig. 19. The effed is large in ionic crystals because
of the strong coulomh interaction between ions and e1cclrons. In covalent crys-
tals the effect is weak because neutral atoms have only a weak interaction with
electrons.
The strength of the electron-lattice interaction is measured by the dimen-
sionless coupling (."Ollstant a given by

1 deformation energy
-a= (64)
2 hwt.

where WL is the longitudinal optical phonon fre(Juency near zero wavevector.

We vicw ia as "the number of phonons which surround a slowmoving electron
in a crystal. "
Values of a deduced from diverse eJl:periments and theory are given in
Table 4, after F. C. Brown. The values of a arc high in ionic crystals and low
in covalent crystals. The valucs of the effective lIlasS mf><>l of the polaron
are from cyclotron resonance eJl:periments. The values give n for the band
cffective mass m* were cruculated frolll mi",l' The last row in the table gives
the fact or mj>ol/m* by which the band mass is increased by the deformation of
the lattice.

G&-e QTS. Chapler 7.

 ..
.
,
Table 4 Polaron coupling COlUtanu 0', masses m;.,a, and band manes m'" for
electrons in the conduction band

Crystal KCI KB, AgCI AgIl, ZnQ PL' InSb CaAs

a 3.97 3.52 200 1.69 0.85 0.16 0.014 0.06

m~m 1. 25 0.93 0.51 0.33 0.014
m·/m 0.5O 0.43 0.35 0.24 0 .014
m;.Jm* 2.5 2.2 1.5 IA 1.0

Theory relates the effective mass orthe polaron mf,ol10 the effective brul(l
ma~s m" of the electron in the undcformed lattice hy the relation

mf,ol eo: mt' ( 1 'a

I - O.OOO8a' )
+ O.OO34a2 ; (65)

for a <$; 1 this is approximately m"(l + in}. Because the coupling constant a is
always positive. the polaron mass is greater than the hare mass, as we expect
from the inertia of the ions.
It is common to speak oflarge and small polarons. The electron associated
with a large polaron moves in a band, but the mass is slightly e nhanccd; these
arc the polarons we have discussoo above. 'The electron associated with a small
polaron spends most of its time trapped Oil a Single ion. At high temperatures
the electron moves from site to site by thermally activated hopping; at low
temperatures tbe electron tunnels slowly tbrough the crystal, as if in a hand of
large effective mass.
Holes or electrons can become self-trapped by inducing an asymmetric
local defonnation of the lattice. This is most likely to occur when the band edge
is degenerate and the crystal is polar (such as an alkali halide or silver halide),
with strong coupling of the particle to the lattke. The valence band edge is
more often degenerate than the conduction band edge. so that holes al'e mOre
likely to be self-trapped than are electrons. Holes appear to be self-trapped in
all the alkali and silver halides.
Ionic solids at room temperature generally have very low conductivities for
the motion of ions through the crystal, less tha n 10- 6 (ohm-cmr', but a family
of c.'ompounds has heen rcported1 with conductivities of 0.2 (ohm-cmr' at
2(f'C. The compounds have the composition MA~Is. where M denotes K, Rb.
or NH 4 • The Ag+ ions occupy only a fraction of the equivalent lattice sites
available, and the ionic conductivity proceeds by the hopping of a silver ion
from one site to a nearby vacant site. The crystal structures also have parallel
open channels.
7See B. B. Owcn, and G. R. Argue , Science 157, 308 (1967); S. Geller. Sciencc 157, 310
(1967).
 W p/Il&nJOM. Po/ariloN, ond polDroov

,.
0
eEIc<:l:...,
(0 (0 ,
0 8 0 (0

(0 (0 (0 G
'"

o 8. (0 G
•0
0 G (0

0 8 Ib,
0 G
~

Figure 19 The formati ..... of a polaron. (al A comlu«ion eI.,dron is shu....n in a rigkllanicc of An
ionic crystal, KCL 'me I<m:u on the ions adjacent tu the electron ~ shown. (b) The electron i,
shown II. an o:I...lie or deformable Janice, The electron plus the auociated strain field is called.
polaron. The displacement oIthc iuns inc:n::ases the cffed:i~'C inertia and, hence, the dlt.'d h'e rna"
of the cI('Ctron; in KC I the mlln is ine reased bya f.aclor uf2.5 with rcspc« tu the hand thl-Ury m:l.<'
in a rigid 1.11;«. In c~treme situations, oilen wilh holes, the particle can l>erorne self-trapped
00caIi.wd) in the-lattice. In CQ\-aknt crystlk the IOrca OIl the atoms flOm the c le«ron are weaker
than ;n iunic e~tab. SO that polaron deformations ~ smlln in ('Ow]cn! cry.tals.
JOO

PEIERLS INSTABlUTV OF LINEAR METALS

Consider a one-d.imcnsionru metal with an electron gas fillin g all conduc-

tion hm}(] orbitals out to the wnvcvector k,.-, at absolute wro of temperature.
Peicrls suggested that such a linear fllctai is unstable with respect to a static
lattice deformation of wavevedor G = 2k,.-. Such a deform ation creates an en-
ergy gap at tile Fenni surface, thereby lowering the energy of electrons helow
the energy gap, Fig. 20. The deformation proceeds until Iimitetl by the in-
crease of elastic energy: the equilibrium deformat ion A is given by the root of
<I
dl1 (E"Iec:t ....."" + EdullJ = 0 . (66)

Consider the cla~tic strain A cos 2k,.'X. The spatial-average clastic energy
per unit Jenb>1h is E,,\titlc :: iC.o,2(cos2:2k,.'X) = iC.6. 2 , where C is the force con-
slant of the linear metal. We flext calculate £.,1....1ro<1"'. Suppose that the ion
contribution to the lattice pote ntial seen by a conductioll electron is propor-
tional to the de formation; Uex) "" Ma cos 2kF%. From (7.51) we have
E" = (h2/2m)(k~ + K~ ~ [4(h2k~/2m){!J 2K2/2m) + A 2 A2 J112 • (67)
It is convenient to define
x" . !J2K 2/m ;
We retain the - sign in (67) and form
dE" - A 2A
-d-~- : O(r-'-~-K-+:"cA~'~~-;~"'"~
whence. with rlKhr as the number of orbitals per unit le ngth .

We put it all together. The equilibrium deformation is the root of

iCA - (2A 2mai'lTh 2 k,,) sinh- I(!J!kj/mAA) = 0 .
The root A that cotTesponds to the minimum energy is given by
!J2kJ·/mAA = sillh(-!J 2 krnCI4mA2) I (68)
whence
IAIA c. (2h2k';'./m) exp(- h 2 k,.-rl.'/4mA 2 ) , (69)
if the arb'Ument of the sinh in (68) is :J> 1. We assume k,. O!IO. t k -.'
 10 PlaomQnl, f'olaritont, and Polaron~ 301
(

E"""lt)· Pi' ;"lrod"..... l

by p,;"rt,. \n11.hiI,,)

101""'1"'" CDt~
011<-, Iolrodu<t"",
"~'W

Fi,,", 2() P~ierh inSbhility. Electrons "'ith ",.. YCYff-

lor~ nt"ar the: Fermi surf.'Ia: have their e nergy lowered o
II)' a Iatt~ dd"~ion.

The result is of the form of the energy gap equation in the DCS theory
or superconductivity, Chapter 12. The clefonnation b. is a collective effect
of all the electrons. If we set W = h2,q.·/2m = conduction band width;
N(O)::: 2mhTfl 2k f ; = density of orbitals at Fermi level; V "" 2A 2/C = effective
electron-electron interaction energy, then we can write (69) as
/Alb. ~ 4W exp[ -I/N(O)\'l , (70)
which is analogous to the DCS enerb'Y gap equation. II An example of a Peierls
insulator is TaS3'

SUMMARY

• The dic1et-·tric function may be dell ned as

<( K) = p_,(w,K)
w, p~><I(w, K) + PInd(w,K)

ill terms of the applied and induced charge density components at w,K.
• The plasma frequcncy w,.
= 141Tlllflt{oo)mJlI2 is the fn.'quency of the uniform
collective longitudinal oscillation or the electron gas against a background of
fixed positive ions. It is also the 10\\1 cutoff for propagation of transverse elec-
tromagnetic waves in the plasma.
• The poles of the dielectric function define Wr and the zeroes dellne WL'

eA revle>\o tJ charge density wavell and ~"pcrlalticcs in melallk byercd ~Iructurc~ b given hy
J A. Wil$Ofl, F. J. Di Sa]\IOand S. Mahaj.:l.n. Advances in Ph)'Sici 24, 117 (1975); Ih is artideconlainl
50Ine .:llhe most heautir"l dilTraction photographs in fOIid slale pI\)~icI;. P. B. Littkwood and
V. I~ci n", J. Ph~'s C. 14, 2493 (1961) ha~ sh",,'11 hu", the Peie rls elTect;s modified by theelcctron-
elt'ctron interllCtion; see also J. E. H inch, Pb)'$. Rev, Lett. 51, 296 (1983).
'"
• In a plasma the coulomh interaction is screened; it becomes (qlr ) e~p(-k.r).
where the screening length 11k. = (1f~.f61T'1~ 1 12.
• A metal-insulator transition may occur when the nearest-neighbor separation
a becomes of the order of 40o. where (10 is the radius of the firs t Bohr orbit in
the insulator. The metallic phase exists at smaller values of a.
• A polariton is a quantum of the coupled TO phollon-photon ficld~. The cou-
pling is ass ured by the Maxwell equations. Thc sp(..-clrnl rcSioll w-r < W < wI.
is forbidden to c1ectroma~nct ic wave prop,agatioll.
• l1ic Lyddilllc-S.lchs-Tcller relation is WfJW? == E(O)lE(ce).

Probletll3
I. Sur/ace p'a.smo~. Consider a semi·infinilf' plasma on the pruilh'c side the or
plane z .. O. A solution of Laplace's equation V'ip - 0 in the pl.nma ili
<Pb:,::) "'" A cus k.r; e - h , whence £". - kA cos Ia II As; E.. - kA sm kr e - h .
(a) Show that in the vocuum 'PfI..:X,z) - A cos fu ~ for z < 0 satisfies the boundary
condition th.!t the tangential component ofE be COlltinuOU$ al lhe Ixmnc:\ary, th.11 is,
I1ml £,,,. (b) Note thai D j = "(w)~; D" .. E.,. Show that the boundaryeondition th."lt
the normal component of D be eontilluous a.tlhe boun<Lv)' retluires that E{w} .. - I,
wht:nce from (10) we ha\'e the Stem-Ferrell rC'm lt :
(71)

for the frequelle)' w. of a surface plasma oscill;Ition.

2. Itlterfoce l!iasmon:J. We consi<\er the plalle in terfact! :t - 0 between metal I a.t

:t > 0 antI metal 2 at :t < O. Me tal I has bulk plasmon frequenL"}' ""pl; metal 2 has
wl4' The {Iielectric constants in both melllls are those of free electron gases. Show
thai surface plasmons associated wi lh Ihe inter£dc.c have Ihe frequency
...... (l{w~ 1 + .....~PI2 .
3. AlfutrllC~. Consider a solid ""ith all e<lual concelltrntlon n of e1cdnms of mass
m~ and holes of mass m .... 111is sItuation ma), :lrisc ill a sClllimetai or in a compensated
semi<'Onductor. Place the solid in a uniform rn."lgIlCIic l1eltl B - 1I~. Introduce the
coonlinate f "" x + iy appropriate for c..iTCularly polari:tetl motion, with f Ila\'ing
time <Iependence e- - . Ld w.. - eB/m"c alKI w" - ellhnJ,C. (a) In CCS units, show
that t.. - e£+/I11.. 14.."" + w.}; 6. - -e £+/11II.""w - w,,) .ue the dispbcemenb of tile
electrons amI holes in the electric Reid £1" e "'" - (E" + IE.) e -""'. (b) Show th.lt
the dielectric polari7.ation p+ - rlr{6. - £,) in the regime w cC w,., w" may be ....rit-
tell as p + = rlc'(m~ + m.)£+f1I2, amI the <Iidectric function <'(w) ·
f/ + ·hrP+f£+ =" ~ + 4-m!-plII , ... here "/ is the <Iidedric constant uf the hust lat·
2

tice and p _ .. (m~ + mh) ii the mass dem:;t)' of the carriers. If ~ ma), be nq;lected,
 the dispersio n re l.l tion cJlf(l<» .. c!~4 becomes, for electromagnetic wa\'e5 I,rol>.l-
gating in the Z tlin.'Ction, cJl ,. (8 2/41rp)K'. Such w:n es li re known as Alf"co wm .'\
they propagate with the constant \'e:OOty 8 /(41rp)ll2. If 8 :.. 10 kC , n .. IOle em :>,
m .. 10 I'J g. the ",elocit y is - lOS on s I . Alf"e ll w:wes h.we been observetl in
scmimetals anti in clectrol1-llOle (Irops in germaniullI (Chapter II).

4. lie/icon IOOVU. (a) Employ the method orProblem 3 to t reat a specimen with ooly
one carrie r tytle, SIlY hole s in concentr.ttion p • .and in the li mit I<> <I """ .. e8JlII"c.
Show that e(w) !;II 4 ~/m~ • ..... here D +(w) '"" E(W)£·(I<». '1'1le ten n f l in e has
W n neglrctell. (b) Show furthe r thOl t the d ispersion relation becomes
w a (lk/4npe)K ', the helicon dispenion relation, in CCS. For K .. 1 em I :uul
B ... 1000 C, eJ ti mOlte tile helicon freque ncy in sodium me tal. (The frcqu('O(:y is
ncgath'e; \li th drcular-pol-lri:red modes the sign d the frequency refers to the lense
of the rot,llion.)

5. PkJ.nnon mode of D . phere. Th e frequeu e)' of the uniform plasmon mode of a sphe re
is de te rmined by the depolarization field E '" - 41TP/3 of a $jlhcre, where the polar-
or
iu t;ulI P .. -riff, with r as the a",erage disjJlacemen t of the elcc::trons conCt;:ntra-
{ion n . Show from F .. rna that the relOnance frequcncy of the electron gas is
u.r3 .. 41mCl3m. bause all electrons Ilarticillate ill the oscillation, such all cxcita-
tion is called a oollecth'e excitalKm or collecth'e mode of the electron gas.

6. Mrrg ndopl.a.mw frequency. Use the mctlwxl or Probl em 5 to fin(1 the rrequency of
the uniform plas mon mod e or a sphere placed ill a constan t uni form mllgnetic field
8 . Let S be .along the z axis. The solution should go to the cyclotron frequency
-we" - e8Jmc ill one li mit and to Wo - (41mc l3m)1I2 ill anot her li mit. Take the
mo tio n in the :t - !l lllane.

7. Photon brtmch at low toovetJ«Cor . (a) Find what (56) becomes when ~oo) is taken
int o account. (b) Show that there i.5 a solution of (55) which at low wavcvectnr is
w .. cI<l'\/f(ij), which is whOlt you expect for a Ilholon ill a cryst:ll of refracti"'e in(1ex
/I' ... f.

8. PW ma frequency OM ek ctricnl condu('tivity. An .,rganic cond uctor has recently

been foulltlby optical studies 10 have wI' =' 1.80 X 101.5 5- 1 for the plasma fre-
quCllC)', anll 'T == 2.83 X 10- 15 s ror the electron relaxation titne at room tempem-
ture. (a) Calculate the electrical co nducti"'ity from these (lata. TIle c~rrier mass is
not known and is nOt IJeetieJ here . Take ~",,) .. l. Corwert the result to un its
(n cm)- I. (b) f rom the crystal aod chemical structllre, tile contluction electron
CQucent ration is 4.7 x 1()21 clll- 3. Calculate the elcctnm efl"ectillC mass m·.

9. Bulk modulus of the Fermi gas, Show that Ihe (:oll trib ution of the kinetic energy 10
the bulk modu lus of the clectron gas at absolute zero is 8 - l,..n.t!,·. It is con\'enient
to use (6.60). We can use our result for R 10 fi ll(1 the velocity of sOllnt!, which ill a
compreu ible fluid IS v .. (81p)V2, whenre v '" (mllM) II2""., in agreemen t wtth (46).
These es timates ne glect atlracti~'e interactions.
 10. Rerponu of eltctro .. ga •. It is sometimes dated errom.'Ously in books on ek'Ctro-.
magnetism that the static conductivit y 0', which in gaussian units has the dilnens;ons
of a frt,'Quency, measures the response frequency of a metal to an e lectric Bekl
suddenly applied. C riticize this statement as it might apply to copper at room
temperature. 1fie resistivity is - 1 ~hm-cm; the electron concentration is
8 x IOZZ cm-'3, the mean free path is - 400 J..: the Fermi "elocity is
1.6 x lOS em s- I. You will not necessarily need all these data. Give the order of
magnitude of the three frequencies u, Wp , and 11., that might be relevant in the
Pfoblem. Set up and sol,'e the problem of the respQflse ~t) of the system to an
electric field E(I < 0) .. 0; E(' > 0) = I . 1fie system is a shee t of copper; the field
is applied normal to the sheet. Include the damping. Sol,'e the differential t,'Quation
by elementary methods.
-II. Cop pill.7mOrl$lmd tile carl del" Wooh inleracliorl. Consider twosemi-inflnite media
with pla ne surfaces.t '" O,d. 'nle dielectric function of the identical media is E(W).
Show that for surface plasmons symmetrkal with respect to the gap the frt,'Quency
n)ust satisfy t:{w) = -tanh (Kd/2), where K2 - Po + ~. The electric potential will
have the form
'P - f (%) exp(ik,x + ik.}J - lu) .
Look lOr nonretarded solutions-that is, solutions of the Laplace t,'Quation rather
than of the wave equatioo. The sum of the zero-point energy of all gap modes is the
nonretllrded ~t of the van der Waals attraction Letween the two specimens- see
N. C. van Kampen , B. R. A. Nijboer, and K. Schram, Physics Letters 26A , 3ffl
(l968).
References
J. N. 1I0dp0n, OptictJ aWorpl iotlllnd d~ In 1Ol;d,s, Chal'm"l1 "lid Hall . 1970. Good ;ntro-
<luctory sur.C)'.
F . Stem, HElcmcntary thwry of the optic'd1 properties of metals," Solid state physic-s 15, 3OO{196J).
II. G. Schuster, cd., One-di",~,,"r"'al conductaf3, Sprin~t, 1975.
N. F. Ma tt, Melal-insuwla! Iramillon$, Taylor & Fral1ciJ, 1990.
L V. ):cklysh, D . A. Kirthnit,.~ an<l A. A. "'"ra,looi", DldllCf.'fc fUrICtion of CI.ImlenJtd $V$le"'$,
E;1iC\'/cr, 1989.

souo STATE PL4.SMAS

A. C. Baynham lind A. D. Uoan]man, PIanntJ effUU In umicofIdudof:f; mlicon IIlId AlfL'in U'ava,
T"ylOI" & Franris, 1971.
J. Danicb. C. v. Fcstcnbcrg, II . Raether, and K. Zcppcnfcld, HOpt/cal constants of solids by
electron spcctroKOpy," Spring/l! l rae" in modem ,JhVsfc:s 54, 7l:I ( 1910).
E . N. E~'OnQm()lJ and K L . Ngai , "Surface pla$n,a oscillations Hnd rcl3lcd surface effects in solids,"
Ad~'>IIJlCe6 in cJ...,,,,il;oI l/hllSia 21. 265 (1974).
P. M. I'bt:t.man and P. A. Wolff, wo~~~ o"d inlera(:tiom In flQ{id stille "w...1Ul8, ACIOdcmic Press,
1973. Good general rcfe I"Cl1CX' : HI,;o contai~ OlCWIlllt of Fermi liquid theory and experiments.
II. Raether, &citlllion ofpl(WJl()m , Springer, 1960; Sur/ot:e pllUmons, Springer, 1968.

POI.A.RONS
J. J. Dev",'CSC, cd .. l'olaroM olld neilON, Plenum, 1984.
J. Appel, " Polarons," Sohd ~tlltc physia 21, 193 ( 1968).
J. Singh, HDynlllJlic-s of exdtolls," Solid .tate physia 38, 2SlS (1U&4).

··I'llis problem is somewh". dUneult.

 (

11
Optical Processes and Excitons

OPTICAL REFLEcrAr-;CE 307

Krarners-Kronig relations 308
Example: conductn.ity of oollisionless elearon gas 311
Electronic interband t rnn~ilions 312

EXCITONS 312
FrenJ.:el excitOlls 316
Alkali halides 318
Molecular crystals 318
Weald)' bound (Moll-Woo nier) eJ:cihms 3 18
Exciton condensation inlo eledron-hole drops (Ell D) 3 19

RAMAN EFFECT IN CRYSTALS 322

E lectron speclrQSWpy with !f-rnys 324

ENERCY LOSS OF FAST PARTICLES IN A SOUD 326

SU M~IARY 329

PROBLEMS 329

1. Causality and the response fU!lc tion 329

2. Dissipation sum rule 329
3. Reflection at norma] incidence 330
4. COJldUCli\'ily sum rule and supercollducti\'ily 330
5. Dielectric constant and the semiconductor energy gap 330
6. Hagen-Rubens relation for infrared renecl i~;ty of metals 33 1
7. Dnvytlt.l\' splitting of exciton li nes 331

REFERENCES 332

--
 ., .
I.
.. . ... !. n
b """""'" -
~ • • ' :!: II:
SliM _ f<w p"""- obIorpt .....(SWln fJI"'<ft.)
(anti-Stolrs)

-, .
n'.,,·

.' igure I Tncre are m... y 1»- of exper!mcn~ io which lighl intcract~ with wan,lIke exci tations ,n
a crystal . .se. .. n1 procascs arc ,Uusl"'tcd here

-
 CHAPTER 11: OPTICAL PROCESSES AND EXCITONS

The diclect.-ic function E(w,K) was introduced in the p.-eceding chapter to

describe the response of a crystal to an eit-ctmmagnetic field. The dielectric
function depends sensitively on the electronic band st.-ucture of a crystal, and
studies of the dielectric function by optical spectroscopy arc very usdul in the
determination of the overall band structure of a crystal. Indeed, in the last
decade optical spectroscopy devc\op(."tl into the single most important experi-
mental tool for band stwctu re detenninalion.
In the infrared, visibk, and ultraviolet spectral regions the \\".lvevector of
the radiation is very small compared with the shortest rt."Ciprocallattice vector,
and then." fore it may usually be taken as zero. \Ve are conC<:rned then with the
real E' and imaginary i' parts of the dielectric function E(W) -= E'(W) + U'(w),
also written as f'1(W) + iC"z(w).
Howe\'er, the diclechic fUllction is not din,."ctiy accessible experimentally
from optical measurements: the directly accessible fUllctions are the .-eflectan(.'C
R(w), the refractive index n(w), and the extinction coefficient K(w). Our first
objective is to relate the e xperimentally observable quantities to the real and
imaginary parts of the dielectric function.

OPTICAL REFJ.ECfANCE

The optical measurements that give the fullest information on the elec-
tronic system are measurements of the reflectivity of light at normal incidence
on single crystals. The reflectivity coefficient r(w) is a complex fUllction defined
at the crystal surface as the ratio of the reflected electric fleld E(refl) to the
incident electric field E(inc):
E(rdl)/E(inc) !II r{w) ill p(w) exp[i8(w)1 , (I)
where we have separated the amplitude p(w) and phase 6{w) components of the
reflectivity coefficient.
The rerractive index n(w) and the extinction coefficient K(w) in the crystal
are related to the reflectivity at normal incidence by
li+iK - l (2)
r{w) - ~:-C:7-:-~
- 1i+iK+l'

as derived in Prohlem 3 by elementary consideration of the continuitv of the

componenl~ of E and B parallel to the crystal surface. By definition II(W) and
K{w) are related to the dielectric functi on E(W) by

V E(W) - n(w) + iK(w) .. N(w) , (3)

3'·7
""
where N(w) is the complex refractive index. Do not confuse K(w) as used here
with a wavevector.
If the incident traveling wave has the wavevector k. then the !I component
of a wave traveling in the :t direction is
E~inc) = E.-o exp[i(kt - wt)] . (4)
The transmitted wave in the medium is attenuated because. by the dispersion
relation for electromagnetic waves, the wavcvcctor in the medium is related to
the incident k in vacuum by (n + iK)k:
EJtrans) oc cxp{i[(n + iK)kx - wtn = cxp( - Kkx) cxp{i(nkx - wt)), (5)
One quantit y measured in experiments is the reflectance R. defined as the
ratio of the rencctcd intensity to the incident intensity:
R = £"(rcn)E(rcn)/£"(inc)E(inc) = r.r = p'l . (6)
It is difficult to measure the phase 6(w) of the reflected wave. but we show
below that it can be calculated from the measured reflectance H(w) if this is
known at all frequencies .
Once we know both H(w) and 8(w), we can proceed by (2) to obtain n(w) and
K(w). We use these in (3) to obtain E(W) = E'(W) + iE'(w), where E'(W) and E'(w)
are the real and imaginary pru1:s of the dielectric function. The inversion of (3)
gives
E'(w) = 2nK • (1)
We now show how to find the phase 6(w) as an integral over the reflectance
H(w) ; by a similar method we relate the real and imaginary parts of the dielec-
tric function. In this way we can flnd everything from the experimental H(w).

Kramen-Kronig Relations
The Kramers-Kronig relations enable us to flnd the real part of the re-
sponse of a linear passive system if we know the imaginary part of the response
at all frequ encies, and vice versa. They are central to the analysis of optical
experiments on solids.
The response of any linear passive system can be represented as the super-
position of the responses of a collection of damped harmonic oscillators. Let the
response function a(w) = a'(w) + ia"(w) of the collection of oscillators be
defined by
%... = a(w) F... , (8)

where the applied force field is the real part ofF... exp(-iwt) and the tota1 displace-
ment % = LJ% is the real part of %... exp(- iwt). From the equation of motion,
J
•
" 0phaJ Pr~ ..a and Ucil..... 309

(
w(; have the complex response function of the oscillator system:

_~ jj _~ wj- ~ + iWPI
o(wl-L. , , . --L.h ('
J "f- w- - IWfJJ Uf- ."'+ UTpj
W I
' ... (9)

where the constants /j = 11M) and relaxation frequencies PJ are all positive for a
passive system. nle masses are Mr
If a(w) is the dielectric polarizability of atoms in concentration rl , then f
has the fonn of an oscillator strength times nfillm; such a dielectric response
function is said to be of the Kramers-Heisenberg form. The relations we dev-
elop also apply to the electrical conductivity o{"') in Ohm's law, j",:= o(w)E.....
We need not assume the specific form (9), but ....-e make usc of three prop-
erties of the response function viC\\Ie(\ as a function of the complex variable w.
Any function with the following propertit..>$ will satisfy the Kramcrs-Kronig rela-
tions (11):

(a) The poles of a(w) are all below the real axis.
(b) The integral of a(w)lw vanishes when taken around an infinite semicir·
de in the upper half of the complex w-plane. It suffices that a(w) _ 0 unifonnly
as IwI-o:>.
(c) The function a'(w) is even and o"(w) is odd with res(X."Ct to real w.

Consider the Cauchy integrnl in the fonn

1 P
a(w)===-.
111
I- _ ...
-0(,)
-d,
s- w
(10)

where P denotes the principal part of the integral, as discllssed in the mathe-
maticalnote that follows . 111e right-hand side is to be completed by an integral
over the semicircle at infinity in the upper half-plane, but we have seen in (b)
that this integral vanishes.
We cquate thc real parts of (10) to obtain

1 I- 0"('1 1
a'(w) = -; p _ ... -;::-;:; ds.= 11 r [1- u
0"(,)
~ ,h, +
I'_.., 0"(,,)
P _ w dp
1
In the last integral we substitute s for -1) and use properly (el that o"(-s) =
-a"(s); this integral then becomes

i-
o s+w
0"(,) d,

and we have. with

1 J 2.s
-,---w- + -,-+-w- ; -,.,=",.
310

the result

2 LO< sa"(s)
a'(w) = -; P ,,1 J- ds. (l1a)

This is one of the Kmmcrs-Kronig relations. '111C other relation follows on

equating the imaginary parts of Eq. (10):

1 p
a"(w) => - -
11"
I- .'(.) 1
- - ds:: --P
_,. S - W 11"
[I" sa'(s)
- w
ds - L
" s+w
a'(s) dS]

whence

. "(w) = - z:
"
p L.. . a'(s)
sl-cJl
(Is (lib)

These relations arc applit!t\ below to the analysis of optical re flectance data; this
is their most important application.
Let us apply the Kramcrs-Kronig relations to r(w) viewed as a response
function between the incident and reflected \VdVCS in (1) and (6), We apply
( 11) to

In r(w) = In R1/1(41) + iO(w) (12)

to obtain the phase in lenns of the rdlectance:

8(w)=- -;P
w L" s:ZIn _R(s)J- ds (13)

We integrate by parts to obtain a form that gives insight into the l'Ontribu-
lions to the phase angle:

e(w ) = - - 1
21f
L-
0
10 l<+wldloR(')d
--
s- w
, .
ds
(14)

Spl,:ctral regions in which the reflectance is constant do not contribute to the

integral; further. spectral regions & J> wand & « w do not contribute mILch
because the fUllction In 1(& + w)J(& - w~ is small in these regions.
/IIlltlwma fical Note. To obtain the Cauchy integral (10) we take the integral
f a (s)(s - wt l ds over the contom l in Fig. 2. 111e function a(s) is analytic in the
upper half·phme, so that the ,·a]ue of the integral is zero. The contribution of
segment 4 to the integral ,'anishcs if the integrand a(s)J& _ 0 faster than 1sl- 1 as
181- w. For the response function (9) the integrand_ 0 as 181- 3 ; and for the

'&-e E. T. \\1'ittaker ~nt.l C . N. Watson, Modern lInnlyril. C 'ridge, 1935, p. 117,

\
 JJ Oll/iall I'r'Oa!ue, ,,,,d &cit0nl 31 ,

figure 2 Contuur Xl' the Caochy principal v.alue • ,

integral.

conductivity o(s) the integrand _ 0 as 1.~ -2. 111C sl."gmcnt 2 contributes, in the
limit as II _ 0,
o(s) (0 iu elf! dO
J- - ds_ cr(w) ).
(2) s - w .. lie
If! '" -7Ticr{w)

to thc intl."gral, wherc s = w + II e 19• 11,c segmcnts 1 and 3 are by definition thc
principal part of the intl.ogral between --:r;I atHI 00. Because the integral over
1 + 2 + 3 + 4 must vanish.
-
JJ J + e p
0(,)
- _ - d.~ = 7Tia{w) (15)
_ ", S W
III (3)
as in (10).

EXAMPLE: Cortduc,ivity ufCollisiortless Ekclnm Cas. Consider a ps offrec e1e<:tron~

in the Umit as Ihe collision rrequency goes to:terO. From (9) Ihe response function is,
withf '2 Urn.

a{w) .. - -1- lim - -1. - " - -1- [1

- - hrc5(w) ] , (16)
1111£1 p-oO W+ lp mw w

by the Dirac identity. We confirm that Ihe delta function in (16) satisfies the Kromerso
Kronig relation (lla). by which

,2L-a:~)
a(w) -- ---ds ~ - -I - , (17)
Tn r - J- md'
in agreement with (I6).
We obtain the electrical conductivity 0(1£1) from the dielc..'Clric function

t:(w) - I ,. 47rP.,IE.. = - 4mlltt.,lE.,. .. 47r1Ieicr(w) • (i8)

where a(w) '" x.,l(-e}£., is the response fundion. We use the equi,'8.Ience

(ees) o(w) .. (- ir.J47r1l~w) - Ii . (19)

fOr the M:uwel\ equa tion can be wrilten either as \, curl H "" 47ro(w)E - IwE or as
c cud H - -fwt:(w)E. We combine (16). (18). and (19) to find tbe conductivity of a
eollisionless electron gas, I
m
i
a'(w) + io"(w) _ ne ["7TIi(w) +
1£1
i..] . (20)

For collbionless electrons the real part of the conducth ily has a delta fund ion at w - O.
'"
Electronic lntcrlxmd Transitions
Jt came as a surprise that optical spectroscopy developed as an important
experi mental tool for the Uctermination of b and structure. First, the absOI-ptioll
and reflection hands of crystals arc broad and apparently featureless funct ions
of the pholon energy when this is greate,- than the band gap. Second, din.'ct
interhancl absorption of a photon fl(AJ will occur at all points in the Brillouin wile
for which ellcrgy is conserved :
(21)
where c is an empty band and v is a fi lled hand. The total absorption at given w
is an integral Over all transitions in the zone thaI satisfy (2 1).
Thn.:c factors unraveled the spectra:
• The broad bunds arc not like a sp(:ctrallinc greatly broadened by damping.
but the bands convey much intelligence which e merges whe n derivatives are
takell ofthe rcllcctam.'C (Fig. 3); derivatives with respect to wavclenb>th , dec.
tric fi eld, te mperature, pressure, or unJaxial stress, for exam ple. The spec-
troscopy of de ri\'utivcs is called modulation spcctrosoop)'.
• The rellition (21) docs not exclude spectral structure in a crystal, because
transitions accumulate at frequcncies for whidl thc bands c, v are parallcl-
that is, at frequenCies where
V' ['dI<) - .,.(1<)) • 0 . (22)
At these critical points in k spa<''C the joint density of states DJ.EL + flW)D,.(E,.)
is sing ular, according to the same argument we us(.'d in (5.37) to show that the
density of phonon modes D(w) is singular when V"w is zero.
• Thc p seudopotcntial method for calculating energy bands helps identify the
I>ositions in the Brillouin wne of the cri tical points found in modulation spec...
tm. Band-band energy differenccs can be calculated with an accuracy as good
as 0.1 eV. The experimental results can the n be fed back to give impTOn.....
ments in the pscudopotential calculatiOIlS.

EXClTONS

Refl ectance and absorption Spectn\ often show structure for photon ener-
gies just below the ellergy gap, where we might expect the crystal to be tra ns-
parent. This stnldure is caused by the absorption of 11 photon with the creation
of a hound electron-hole pair. An electron anda hole may be bound together by
their attracti ve coulomb interaction , just as an elec tron is hou nd to a proton to
Con n a neutrn1 hydrogen atom,
The bound electron-hole pair is called an cxciton, fi g. 4. An exdtoll <."iln
mOve through the CT)'stal and transport energy; it does not transport charge
 (

3.4 3.6

«)

t'sur" 3 Coonpari_, ~(..) .dlc<'tJfI<!C, (b) " ......,length Ocri,,.t;.-,, rdlft'131K.e (lint dcr',.,.ti\"t). and
(,,) cle<.trorcfiN.tancc (third dem'llli--e), of the spt:clnl region in gennllniun. between 3.0 and
3.6 c\'. (Afler &Ot .. I»' D. D. Sell, E. O. "MOe, and D. E. A$pncs.)

becnuse it is ckctrically neutral. It is similar to positron in Ill , which is [ofl11{.'(1

from an ekdroll and a positron.
Excitons can be formed in eyery insulating crystal. When the haml gap is
indirect, cxcitons ncar the direct gap may be unstable with respect to d<.,'c ay
into a frcc electron and free hole. All cxcitons arc unstable with respect to the
ultimate rt..-combination process in which the electron drops into the hole. Ex-
citons can also form complexes, such as a bicxciton from two cxcitons.
\\le have seen that a free cit,.ctron and free hole arc created whenever a
photon of energy greater than the energy gap is absorbed in a crystal. The
threshold for this process is fiCu > Ej( in a tlin:ct process. In the intlirt.'CI
phonon-assis ted process of Chapter 8 the threshold is lower by the phonon
energy flO . But ill the formation of cxcitons the e nergy i 5 Iowcrt~:I with respect
to these thresholds by the billding energy of the exciton , which may be in the
range I meV to 1 eV. as in Table 1.
Excitons can be formed hy photon absorption at any critical point (22), for
ifV",-E. = V~E" the group velocities of electron and hole a re equal and the parti-
cles may be bound by their coulomb attraction . Transitions leading to the for-
mation of excitollS below the energy gap ru-e indicated in Figs. 5 and 6 .
31<

• • • •
• • • • •
•
• • •
• • • • • • •
• • • • •
• • • ·, • •
• • • • • •
Figure 4. An l'oeil"" i$. bound dedron-hole pair, un"," ), rreo: 10
,00\'" togetht-r through the crystlll In som" rcspccls it i~ ~i '1\ilar to
110 alOffl of positron;u"" fonnro from a positron and an electron_
The e~<:IIOf' ~hown is 11 Mott-Wanllicr (,u:;lon: it is .....,:I1J)' bouoKl,
"ith an '>'Cr.o.gc ekctron·h,,1c disl3r>ee large in com lJRO!oOll "'"'Ih a
lalt;,oe cons tant.

•
•
• •
•

F;~re 4b A tighdy bound os ,,'re.,kcl exciton show" 1ocaIO«d on

onc alom in an alkali halide CT)'>"lal. All i.kal ,"....,,,kel cllCiloll ",11
t"wc! as a ''1'3.>" t1,roughoullhc crystal, bulthc clcdron IS alwlI)'S
ck>Je 10 doc hole.

Si
C.
CoA,
14.7
<1 .15
42
Table I Binding energy or excitons, in mcV

BaO
I, F
InSb
56.
'.0
(OA)
RbCl
LiF
AgBr
...,.
(1000)
ZO.
C,P 3.5 K! 480. AgCl 30·
CdS 29. KCI .wo. 11CI II .
CdS. 15. KO, .wo. 11o, 0
•
Data .w<:mbled by Fn:dc,;d.:: C . 80"0\\TI and Arnold Schmodt.
- II Oplkal PN,Jt:f:un arid E:r:Cj,Of1~ 31 ...

COlldul'tior, baOO ./'

(cffed ive ",,"$0 m,.)

~ ---------T-
£Miton lewis
~1""~I:"Il. Ec

Vale""" band
(" Oro;,., MIa'S "'!J ____,- ----------L

.-
Figure 5 Exdtoll Ic''els in relation to the rond....,tiu" band ....1gc, for a ')n'ple hand .t"",lun: wIth
both conduction and vak.....,.. b""d ooges alk = O. An cxciloll ~;\" h ..,c trnnsbtiolMI k;tl<'liccncrg¥.
Excitons arc ,,,,st able with respt.'Ct 10 nul;'lI;,'c Il....~)rnb;nali()n ill which the cll.'Ctron tlrops 'n lo 'he
hole in the valence b",I<I, acrompanied by the e mission of a photon or ph,,"()n ~ .

CooUu.:tion band con tinu um

•

Valence l.oand "<IIlIlauom

Figme (I Energy levelsof"n .. xcilon erc.!!cd ill" d~1 prooC5S. Oplirnilrallsilion. from the topuf
the valence band arc shown by Ih.., arrows; tnc
longest arrow COITCsp"",ls to the cn('~ gap. n.,.,
binding energy oftl,e exciton is E#" rcfcrrc<! to aff'C<) ck-ctron and f,,* hole. The Iowe.t f'<'fluency
absorption line oltll.. (:ryslal al aboolute zero is not E~, but is E~ - E~_

The binding energy of the exciton can be measured in three ways:

• In optical tmnsitions from the v.llenee band, uy the difference betwccn the
energy require<1 to create an exciton and the energy to credte a free e led.ron
and free hole, Fig. 7.
• In rec.mnbin.ttion lumines<.:em:;e, by comparison of the energy of the free
e1ectnm-hoie recombination line with the energy of the exciton recombina-
tion line.
• By photo-ioni7~\tion of excitons, to form frcc carriers. This experiment re-
quires a high c:oncentration of excitons.
'"
(
" ~[,.cil"" ..... r. ,."
" •
••
• \
. ,..~.'"
...,...... •

", "
•
0 0.11 -
\..,..,"\
"1""..,11",,, ""to ...... "..&t~'"
• "fr....,,· eJ.....c"",·I"'k ",,"I
~
'•" ,, -
~ •
~ 07 -
•
~ ,
II'.

,,,
•
,,
" ,,, ,, ,,
, , II , , , ,
" '''' ." ."
I'Ik>Iuo,
."
._ , ~ In ~\ '" I S-l ."

Figure i Effect of an emtoll kl\'c1 on II,.. uph ...,,1 alll>OfllllOn of a SoI"micondurtor Ii,.. ~<MlS of
<'''''~ near the 1)3,><1 J¥P E. in fta,lh"'" ;inClude at 21 K Th., 'crheM "",I.e ;. till' ,,,t e n,ih' aborp-
'1010 (''Of'ffi<:lC' nt. a. in ' (x ) "" In cap( - ax). The energy !\lit) and c~C'llon bindi"~ CII," '1<:Y alC' tk-duo;d
from Ih C' sh~pc of the ahsorption <;UlV<;: • • h(' 1(31' £~ is I 52 1 c\' lIud the cx<:ilon bmo.li.ijI cne'll)' Is
0003-1 t \ -, (After M D_ Sturgc )

\VC diS<.l.Iss excitons in two different limiting approximations, one by

Frenkel in which the exciton is small and tightly bound, and the other by Mott
and Wannicr in which the exciton is \\lCakly bound, with an electron-hole
separation large ill comparison with a lattice const:U1t. Intermediate examples
are kno"'l .

Frenkel Excitons
I n a tightly bound exciton (Fig. 4b) the excit:llion is locali7.cd on or neM a
single atom: the hole is usually on the So.1.me atom as the electron ahhough the
pair may be anywhere in the crystal . A Frenkel exciton is essentially an exdted
state of a single atom, but the excitation can hop from one atom to another by
virtue of the coupling between neighbors. 111e exdtation wave travels through
the crystal much as the reversed spin of a maJ(non tr.wels through the crystal.
The crystalline inert gases have excitons wh ich in their ground states
correspond somewh.1t to the Frenkel moc.lel. Atomic krypton has its Iov.est
strong atomic transitio" at 9.99 eV. The corresponding transition in the crystal
is dosely equal and is at 10. 17 eV, Fig. 8. 111e ene rgy gap in the cryst:11 is 11.7
eV, so the exciton ground state energy is 11 .7 - 10. 17 = 1.5 eV, referred to a
frcc e lectron and frcc hole separated and at rest in the crystal.
Th e tr.mslational states of Frenkel exeitons have the form or propag;lting
waves, like all other excitations in a periodic structure. Consider a (:rystal of N
 (

,
I~
,, ~
..........,.
, v ['.,..
13.0 11.0
'" '"
Figurc= 8 AlJiOr"pli"n spoc1rum of JO!K[ krypton at 20 K. (After C . Baldini.)

atoms 011 a line or ring. If II) is the ground state of atom j, the ground slate of
the crystal is
rJI,:, = Ilrtl z ' . • U." _ IU", , (23)

if interactions between the atoms arc neglected. If a single atom j is in an

excited state oJ' the S) stem is clCSl:tibed by
(24)
TIlis func.:tion has the same energy as the function /PI with lUlY other atoUl I
cxcitcd. However, the function s I{J that describe a single excited atom and
N - 1 aloms in their ground state arc not the stationary ([uantum states of the
prohlem. If there is any intemction between an excited alom and a nearby ~\tom
in its ground state, tIle excitation cne!l.'Y willl>c passed fmm atom to atom . The
cigcnstatcs will have a wavelike form, l\S we now ~how.
When the hamiltonian of the system operates 011 the runction 1;1) with the
jth atom excited, we obtain
(25)
where" is the rree alOI11 e"citation energy; the inte raction T measures the rate
ortr:.msrcr ofthe cxcitation rromj to its nearest neighbors, j - 1 andj + L 111c
solutions or (25) are wa"es or the Dloch rorm :

1/11: "" L, e~p(Uka) I;IJ (26)

3,.
To see this we let Jf operate on 1/1,,:

!}{.pk"" L elika:Jl<Pj "" L elil"(ECf'} + 1'(1/')- 1 + ~+I)l

) )
(27)

from (25). We rearrange the right-hand side to obtain

JfI/Jk "" 2: dfo"[E + T(e1ku + e- "'IJ)lCf') "" (e + 21' cos ka)",,,

j
(28)

so that the cncrb'Y eigenvalues of the problem are

E,,=e+2T<:oska . (29)
as in Fig. 9. The application of periodic boundary wml itions dctcnnines the
allowed. values of the waveveclor k:
k = 2m/lila; s = -iN, - iN + I , ... , iN - 1 . (30)
Alkali lIalides. In alkali halide crystals the lowcst-cnct'b'Y cxdtons are 10-
cali7.ed on the negative halogen ions, as in Fig. 4b. The negative ions have
Iov.'cr eledronic excitation levels than do the positive ions. Pure alkali halide
(.'T)'Stals are transparent in the visible spectral region . which means that the
exciton energies do not lie in the visible. but the crystals show considerable
excitonic absorption structure in the v:.lcuum ultr.lviolet.
A doublet structure is p:lrticularly evident in sodium bromide, 1I structure
simillir to that of the lowest excited state of the krypton atom- which is isoelec-
tronic with the Br- iOIl ofKBr. The spl itting is causoo by the spin-orbit intcrac-
tion. 111esc exdtons are Frenkel exdtons.
Molecular Crystab. In molecular crystals the (''Ilvalcnt binding within a
molccule is strong in (''Ilmparisoll with the van der Waals binding between
mok>cules, so that the excitons arc Frenkel ex(.;tons. Electronic excitation lines
of:1Il individual molecule appear in the crystalline solid as an exciton, onen
with little shift in frequency . At low temperatures the lines in the solid are
quite shllrp, although there may be more strudure to the lines in the solid th:m
in the molecule because of the D:wydov splitting, as discussed in Problem 7.

Weakly Bound (Moll-Wormier) E1Ocitot18

Conside r an elcctron in the conduction band and a hole in the valence
blllld. 111e electron and hole aUr;l(.1 each other by the coulomb potential
(eGS) U(r) == - e2/Er , (31)
where r is the distance between the p:lrticles and £ is the appropriate dielectric
col1.5t:mt. (ll ie Ilittice polari7.ation contribution tu the dieledric (.'IlnsL1.nt should
not be included if the frCtjuency of motion of the exciton is higher than the
optical phonon fre< juencies.) There will be bound states of the exciton systcm
haVing total cnergics lower than the bottom of the (.'Ilmluclion blllld.
\
 - ---------"'.,..

___ -t-~-"'
-- - - ( ... iT

----~/----,r_--~----

• - 2T ----

Figure 9 Em.'TIO' """';IlS ''''''1!n'rtor for a

t'rc"kt'l (,,",,,Inn, fIIkulatc:d wit h po5;I"1! - vi.. o
l>e;1n.. I-,It-;ghlx:..- ..,srn interactioll
t .. 'r.

'Ille problem is the hydrogen atom problem if the energy surf.u::cs for the
electron and hole are spheric,ll and nondegenemtc. 11\c energy levels referred
to the top of the vale nce band arc given by a m<KliIlcd Rydberg equation
MC'
(ees) En = Eg - 'll,2e-n2 (32)

Here II is the principal quantum number Md l' is the reduced mass:

1 1 1
-=-+- (33)
l' me m,,'
funned from the effective masses 1IIr. m" of the elcctron ;md hole.
111c exciton ground state cncl'b"Y is obtained on setting n "" 1 in (32); this is
the ionization energy of the c!{dton . Studies of the optical absorption lines in
I..'u pmus oxide, CuzO, at low tempcmtures give results for the exciton level
spacing in good agreement with the Rydbe rg equation (32) except for transi-
tions to the s t~'te n "" 1. An empirical fit to the lines of Fig. 10 is obtained with
the relation t{cm - I ) "" 17,508 - (800-'11 2). Taking £ "" 10, we find p. =>I 0.7 III
from the cocffident of 11n2 • '111c oonstant term 17,508 em - l corresponds to an
energy g:.IP Eg "" 2, 17 eV.

Exciton Coriderlllotion iuto Electron-Hole Drop, (ElID)

A (:ondens<.><.i ph,tSe of an c1cdron·holc plasJTI."I fonns in Ce and Si whcn
maintained at a low temperature and irrndintcd by light. The following sc-
qucll(."c of c,'ents takes place when an electron-hole drop (EHD) is formed in
Gc: "[lie absorptiof' of a photon of energy flw > E/: produces a free electron and
'20
(

Photoo eoc"fg)l in ~\'

'" '" • "

-,

1-,

1 ~.~"""--~"C.200""---C"".300",,---""C...,,,,,--'
A.oIon mcrg) in .... - .

Figure 10 ~Ithm of the optical transmbsion "l.'nUS pooton energy in cuprous odd.., ;<117 K,
Soo....,", a ICriCS of cxitoo lines. Nole that on tht "crlkai a..is the Ioga.ithm is plo.u(.~ dccreuing
up....ard . thus a pcilk rorroponds to absorption 11K' hnnt! gap E. is 2.17 c\', (Al'tcr P. W.
U.,,,,ncistcr.)

free hole. with high efficiency. These combine r'apidly, perhaps in 1 ns, to form
an exciton . The exciton may decay with :umihilation of the e-h pair will1 a
lifetime of 8 jJ.S .
But if the exciton conc.'cntralion is suffidcntly high- over 10,3 cm - 3 at
2 K- most of tile excilons will (:ondcnsc into a drop. The drop lifetime is 40 IJ.S.
but in strained Ge may be as long as 600 JLS. Within the drop the exciton!>
dissolve into a degencl'iltc Fermi gas of electrons and holes, with metallic prop-
erties: this s lal e \VdS pn.,aided by L. v . Ke1dysh. 111C binding energy in Ce is
1.8 meV with respect to free cxciton5, and the wm::entration II = P "" 2.57 X
10 17 cm - J .
Experimental studies of the condensed EIID ph:u;e have utili7.cd the
me thods of l'C(.'Ombinntion (luminescence) radiation, light scattering, plasma
rcson:Ul(:e, Alfvcn wave reson:mcc, and p - n junction noise.
Figure 11 shows the rccombin:1.Iion radiation in Ge from free excitOfls
(714 meV) and from the EIID phase (709 meV). The width orthe 1 14 meY line
is accounted for by Doppler broadening, and the width of the 709 meV line is
wmp,atible with the kinetic energy distribution of electrons and holes in 11
Fermi gas of concentration 2 x 10 17 em- 3 • Figure 12 is a photograph of a large
E I ID.
 '"
~"' IM\'
'"
Figu re II R"«Knbinatioll radiation of fn:<: clectrol15 and of cI<rtron-hdc d"l'S in Co: at 3.04 K.
The Fermi "'lIerg)' In the drOll is ~I' aoothc eOOc.'S;'·c CtlCrg)' of the drop "'ith fCipcct lo a free
exti lon ill <fI,- (Arlcr T. K. La.)

Figure 12 Photograph of an electron-hole d rq) in a 4 mm .lIlk of Jlu", germani"m. 11lc oJroll Is the
;ntense spot adjacent 10 the l(!1 SC,'C\V 00 the left oft" c t.!jsk. Th., Ilhntogr;oph is the image of the
dro(1 OOlajn<.-d by I'ocu~ing lis e k ctron-OOIc re<.~!ml>im,ti()n lum,ncSCUlce onto the surfact: of an
j"rrnrcd ·sem,lj,'e videe"" im~g" '"be. (After J. r. Wolfe ct at)
'l22

Table 2 Electron-hole liquid parameters

Cr)slal BinJing energy relative Con<."Cnlralion, Critical

(u,,~t""'sM:«I) to free ",.dtoll. in me\' "or 1'. in "", - 3 temperature, K

Go 18 2.57 x 1017 6.7

Si 93 3.5 x IO U\ 23.
CaP 17.5 8.6 x 10 16 45.
3C-SiC 17. 10. x 10"1 41

Comtcsy of D. Bimbcrg, rot'" rurther Jata see IMK.IoI I· BOrn~lc"', Vol. 111117, Springer, 1984.

The exciton phase diagram for silkxln is plotted ill the temperature-
con(;cntration plane in Fig. 13. l11c exciton gas is insulating at low pressures At
high pressurCS (al the right of the diagram) the exciton gas breaks up into a
(:olluucling plas ma of unpaired electrons and holes. The transition from ex-
cilons 10 the pl,lSma is ,ill example of the Mott transitioll , Chapter to. Further
data are given in Table 2.

RAMAN EFFECT IN CRYSTALS

Raman scattering involves two photons- one in, one out- and is one step
more complex th.1O the one photon processes treated e:ulier in this ch.lpter. In
the Raman effect a photon is 1cattered ineidStically by a cl"),stal, with creation or
ann ihilation of a phonon or magnon (Fig. 14). The process is identical to the
inelastic scattering of x-rays and it is similar to the inelastic scattering of ncu·
trons by a crystal.
The selection rules for the first-oruer R.lInan effect are
w=w' ±n; k = k' ±K, (34)
where w, k refer to the incident photon; w', k' refer to the scaltert-.(I photon;
and n.
K refer to the phonon created or destroyed in the scattering event. In
the sccolld-order Ibman effect. two phollons arc involved in the inelastic scat-
tering of the photon.
The Raman effect is made possible by the strain-dcpendenee of the eI{'(.~
tronic polari7.ability. 10 show this, we suppose that the polari:wbility a :lSsoci-
ated with a phonon mode may be written as a power series in tbe phonon
amplitude u:
(35)
If u(t) = Ito cos Ot and the incident electric field is E(t) = Eo COl> wt. then the
induccd electric dipole moment has a component
a.Eolio cos wI cos Ot = ia.EoUolcos(w + O)f + (:os(w - OJI] . (36)
 .

wr------------------,r-------------,
I
I
I 0

I
I
I
, I 0

,•
~

• I 0

\,,~, <:nk..- _ _ II a

,I, " I

"
"
,m· ,,- ,,.. 10" ,.' ,.'
A,<.-.,:c <>M ........ tr.Uon. a111~""'''' ."",lInn'
r";llR! 13 rha:se .I~r"tn 1'0." phoioe..nlo;-tl .,1",rtf1')n5 and hoI..-s in "'>lt~cssed sillCOll n..: t!~r.lrl1
~bows. ror cump\c. thai with an a,·c~ COOC't.~ltmtlon ncar 1(l'7 (,,,- 3 "I l!i K. a fl'l'e-cU.,lon /(3.'l
"ilh salor~ted-g;u: concentration a lO,e en. 3 coc.ists ....,th" (v-.... iablc:) wi"",., or liquid ti.q,\cls,
l'llldl ".th a dc,>!.!), of 3 X 10'· cm- 3. The liquid enlkal temperature is about 2J K. lbc.'Ord,.:aI
and ,,~pcrim"'"tal v"iu.,J "r lire rnctal-inll..lator Iransiti"" "'" cu:itons arc ~ho (From J. P sno--n.
Wolfe.)

.....
. .
'
0/, .'

., -,
'" RK

Figure 14 ltunan scaHerlng of. lll>OtlHl ",Ib .. miss;on or at-.pIlon of a phonon The pro,:. ,ss;s
~ned Bnllouin $(.'altering whcn an IICQUshc phollon is i,wohed and IlOiar'lon iiC'attcring "hen ."
optK.t1 phonon is ,tl\oh,."tl. S;m il~r pro.:t.'i5Cli OC'Wr with magnans (~lin waves).

Thus photons at frequencies w + nand w - flcan be emitted, ac'('f'.",,,anie<l by

absorption or emission Or.l phonon or rre<luel'\(;Y n.
·111e photoll OIl W - n is called the Stokes line and thut at w + n is the
anti-Stokes line. The intensityohhe Stokes line involves the matrix clement ror
phonon creation, which is ju ... t the matrix clement for the hilfmonic oscilbtnr,
as in Appendix C:
I(w - fl) 0: !(nt.,. + IjulllJ,:llil 0: "I( +I (31)
where rll( is the initial population or phonon nMXle K.
'" (
The anti-Stokes line involves phonon annihilation, \\,HII a photon intensity
proportionaJ to
(38)
If the phonon population is initially in thermal equilibrium at temperature
T, the intensity ratio of the two lines is

l ew+0)
(39)
I{w - 0 )
with (nK) given by the Planck distribution function lI[cxp(fIIlIk R I) - 1]. We
see tllat the relath'C intensity of the anti-Stokes lines vanishes as T _ 0, be-
cause here there arc no thermal phonons available to be annihilated.
Observations on tlle K = 0 optical phonon in silicon aTC shown in Figs. 15
aodl6. Silicon has two identical atoms in the primitive cell , and t here is no
electric dipole moment associated with the primitive cell in the absence of
defomlation by phonons. Hut alII does not vanish for silicon at K = 0, so that
we can obscr-vc the mode by first-order Raman scattering of light.
The second-order Raman cIfed arises from the term 02112 in the polari7.3-
bility. Inelastic scattering of light in this order is accompanied by the creation of
two phonons. or the absorption of two phonons, or the creation of one and the
absorption of anot her phonon. The phoilons may have diITerent frequencies.
TIle intensity distribution in the scattered photon spectrum may be quite com-
plicated if there are several atoms in the primitive cdl because of the colTe-
sponding number of optical phonon modes.
Second-order Haman spectra have been observed and analyzed in numer-
ous crystals. Measurements on gallium phosphide (CaP) are shown in Fig. 17.

Electron Spectroscopy willi X-Rays

The next degree of complexity in optical processes involves a photon in
and an electron out of the solid, a~ in Fig. 1. TIle important techniques of x-ray
photoclnission fmm solids (XPS) and ultraviolet photoemission (UPS) have re-
cently been developed. In solid state physics they are lISed in band structure
studies and surface physics, including catalysis and adsorption.
XPS and UPS spectra can be c<lmpared directly with valence band densi-
ties of states D(£). The specimen is irradiated with highly monochmmatizoo
x-rays or ultraviolet photons. The photon is absorbed, with the emission of a
pllotoelectron whose kinetic energy is equal to the energy of the inciden t pho-
ton minus the binding energy of the electron in the solid. The electrons come
from a thin laye r near the surface, typically 50 A in dep th.
The resolution of ESCA spectrometer systems is of the order of 0.6 eV,
whidl permits refined studies of band structure, particularly of core electrons.
ESCA stands for elcctron spectroscopy for chemical analysis.
 ,
~
i·
"
""s,,,", •
j
,.,. !
•" '
3

!
!lam"" ",,,rt. ('I'll ,

Figure 16 Int ensi ty nlt;o anll·Stoke< to Stokes

figure 15 Fint-onlcr Raman spc«r.I oi"lhe K - Oopti· lines u II. fundian oftcn.pcnllurc. for the oIm,.--

nol mode of" silicon CfY' lai ohlcn'cd at th ree lempcr~­
lurel. l 1loe incident pKllon has a ",..,-elent:,h o(5145),.
The optia.1 phonon rrcqucl'lC)' i~ equal to the freq ucll(.')'
dlif' , it depends slighll yon the temperature. (AftcrT R.
lI art, R L. Aggarwal, and B. i....ax. )
vat ion! or Fig. 15o.. the optir.""i mode of ,ili<.vn.
'The observed IcmJ1ellll ure depeoderK:e is in
good Igr.-emenl ....~Ih the prediction off:q_ (39):
the solid curve is a plot of the funct ion
Cltp(- lIrllkIlT ).

.ft n..
."
II II

""
""

,.
J

-600 - 700 -6JO - SIXl - ~ oo 300 - 200 - 100 ,

&rn.o .. ,t"rt ......

Figure 17 Raman spectrum clCaP at 20 K The two h'~1 peak. are the lint-order fullna" I".."
associRlcd with the ('sellalion of an to phonon at 40-1 em - I and II TO pll(lnon at 366 on - ' , All the
oeher peaks in.vl,'C two phonons (After M. V. lI obdcn and J P. Russell )
326

TIle \'alence band ~truct\lre of silver is shown by Fig. 18, with the zero of
encrgy set at the Fermi level. Ekctrons in the first 3 eV bclO\v the Fermi level
come from tIle 5s comlut1 ion hand. TIle strong [.lC'olk with structure belov.· 3 cV
is frorn the 4(1 \'alplicc electrons.
Excitations arc also seen from deeper levcls, olhm au;ompaniecl by excita-
tion of pl;\l;ITIolls. For example, in silicon the 2p electron with a binding energy
close to 99.2 eV is obscrvt>(1 in replica at 117 eV witli si ngle plasmon excitation
and at 134.7 eV with two pla.~moll exdtatton. Tile plasmon energy is 18 tN.

ENt:RGl' LOSS 01" FAST PARllCLES IN A SOUD

So far we have ust'll photons as probes of the electronic structure of solitk

\Ve can also use electron beams for the same purpose. 'n le results also involve
the dielectric function . nov.' through the imaginary part of U~w). The (liekdric
funeli(lIl enters as InJ{E{W)} into the energy loss by an electromagnetic .....ave in a
~olid . but as - lm{IIE(w)} into the energy loss by a dlargecl particle that pene-
troltes a solid.
Consider this diffcrent'C. Tile gencrnl result from ck'CIromagnetic theory
for the pov.'cr dissipation densily by diclectric losses is

(ees)
,
'!j = 41T E' (iJD/ilt) , (40)

per IIllil volume. \Vith a transverse clCt1romagnelic wave Ee- f ..., in the (:ry~tal,
we huve dDtdt "" - iw E(w)Ec- I ",', whence the Iill1t'-uveragc pov.'Cr is

~ ~ -1-{ne{£e- I""IHe{ - iw~w)Ee- i""'})

4w
(41)
~ _1_ wJ:.~ {(f" cos wi - E' sin wt)cos wt) := - '- WE"(w)J:.-2
41T R1T

proport ional to E"(w). TllC tangen tial component of £ is C(mtilluotJs aCTOS$ the
boundary of the solid.
If a purtidc of charge It and velocity venters a crystul. the dielectric dis-
placement is
,
(eGS) D(r,f) = -grad Ir vt
I (42)

bcetHtse by the Poisson equation it is 0, und not E, that is rclatt-'(I to the free
churge. In un isotropic mediulll the Fouricr componcnt £ (w.k) is ."Ciatt"(llo the
Fourier componcnt D(w,k) of V(r.t) by E(w.k) = D(w,k)lE(W,k).
\
 II Optkal Procnaa and £:Irt."ilo... 3:=:

"
~< ~

."
~

j 10

o,
8,,,,,,"1 .... "'W'. cV

Figure 18 Vnlence hand .,l~roo emlision from silver. after Sil'ghahn lind co-wQrkers .

The time-average 1)O\.\'cr dissipation associat(.'(l with this Fourier cornpo-

nell! is

whence

'(')" 1
qp(w,k) = - 81T W - ; 0 2(w,k) = 81T W -W-
t "(CoJ,k)
lJ2(w,k) . (43)

The result is the motivAtion for the illtroou(;tioll of the energy loss function
- lm{lIt{w,k)} and it is also a motivation for (',"xperimCT,ts 0fI cnergy losses by
fast ck"(.1rolls in thin Ilims.
If the die\(.'(.1 ric function is imk'))emlent of k, the power loss is
2 C
@I(w)= - - - lm{UE(w)}ln(kot;Jw) , (44)
1T 'w
where liko is the maximum possible momentum transfer from the primal)' parti-
cle to an electron of the cry!otal. Figure 19 shows the cx(.'el\ent C').:pcrimclltal
agreement between values of E"(w) dCtlueed from optical reflectivity mea~tlre­
ments with values deduced from electron energy loss meas\lremcnts.
Table 31isls the :\LTonymS for some of the prillcipal experimelltalmcthods
used in studies of energy band structure, particularly of the gross aspects of the
structure. These methods differ from those IJsed in Fermi surface studies,
Cll.'lpter 9.
32S

, ,
, ,
,
~
c. , ,i '"
, , I
,
", ,
, \ , ,,-?',..~
c

- .~u i"
"'--'-. .-
'0 '0
" .,"
30
" .,-'"
Figure 19 <!"{w) for Cu and Au; Ihe bold lines are from enCTgY lou meru;uremenls by J. l)anicls,
and lhe other I",es were cakulaloo from optical measurements II)' D. ll.eagIchok and L R. Can·
field ct .1.

Table 3 Acronyms of curn,.·nt expe rim ental methods for bMd slruclure studi C$

AES Auger electron spedroscopy

ELS Electrou (energy) loss spectroscopy
E"~ 1 Evaporation rate monitoring
ESCA Electro'l spectroscopy for "llell'i~1 :h1alysis
EXAFS EAtended )I.-ray absorption fi ne structure
FOMS Flash (iesorption mass spectrometry
FI~ 1 Field iooization microscopy
lAP JmaS:i llg atom probe
INS Inelastic neutron scattering
IH~I Ion rate moni toring
ISS Ion scattering spectrometry
lTD Isothermal Oewrption spectrometry
LEED Low encrb'Y electron diffraction
RIIEED Renection high eneTb'Y electron uiffract iOlI
SA}.I Scanning Auger microscopy
SEM Scanning el~tron microscopy
SES Seconoor;' elect ron spectroscopy
SI~ I S Secondary ion maliS spcch ollleiry
SLEED Spin-polarizeJ 10\11 energy electlOll tliff.-a<.1ion
TOMS 11lcrmal (\esorption mass spectrometry
TEM TnulSIIlis)ion electron microscopy
THEED Transmission high energy electron tliffrnctioll
UPS UltrJ";olct photoelectron siX!droscopy
XI'S X-my photoele(.1l'On )peclroscopy
 SUMMARY

.. 111C Krnmcrs-Kronig relations conned the real and imaginary parIs of a re-
spomc functi on:

• 111e complex refractive illdex N(w) = n(w) + iK(w), where n is the refractive
index and K is the extinction coefficicnt; furth cr. feW) = ~w) . wlll.'l'K-'e
f'(W) = n 2 - J,."l. and t"(w) = 2nK.

• Thc renectancc at normal incidcnt"C is

(n - 1)2 + I(l
R = ;::-:-~-'--3;­
(n + 1)2 + J,."l
• The cncrgy loss fUllction - Im{lIt{w)} giv{."S the cncrgy loss by a c1l;1rgt.'(1
particlc moving in a solid.

Problems
I. Ca usalit!! and ,he rUllOttse Junelio", TIle Kramers·Kronig relations are l-'Q nsistcnt
wi th the principle that an cffe<:t not precede its ca use. Consider a delta-function force
applied at time I = 0 :

I
F(I) = 6(1) = -2. I·
lwI
_ e- dw ,

whence f .. - 112'IT. (8) Show by direct intcgrntkm or by use of the KK relations that
the oscillatOl' l't'Sponse fun ction

a(w) - (~- ~ - iwp)- I

gives zero displacement, x(t) e 0, for I < 0 under the above force. For I < 0 the
contour integral may he comple ted h). a scmicircle in the upper half-planf'.
(b) Evaluate x(I) for t > 0, Note that a(w) has poles at .:t:(~ - tp')lIt - lip. both in
the Iowl-'I" half-plane.

2. Di3Biptltiorl ,urn rule. By comparison of cw'(w) from (9) and from (I ta) in the limit
w-+!:e, show thai the following sum rule for the 05dllator strengths must hold.

~ J; - -;:-
2 L·
80"(8) ds
>10

3. Rljlecticm (II normal j,jciflcJlCf). Con~i(ler 'Ill d ...'CtromJgnetic wave in WICUUnI, with
fi('1l\ ('Oll1poncnb uf the fonn
£,(inc)'" B~( inc) ~ A~L-- ~ .

Let the \va,'c be inciUent UI)oII a medium of ,1.e1ectric constant f: anll permeability
IJ. - 1 tl!.lt fllls the Iwlf-sp.~ x> O. Show tbat tile rl,nectivity coemdcnt r{w) as
deBnt:d by £(rell) = r(w)";(juc) Is givclJ by
II + il( - 1
dw)= .. +iK+ I

where n + il( - tY2, with " anc.! K real. Show further til:" the rdlC('lance is

R(w) = (n - If + 1..>'1
(n+ lf+Kl

*4. Con(fllcticity '"m rille Ilml ~" ,)(Jroo/lfrllctidty. \Ve write the elcdrirnl ("OlIdlldivity
as u{w) .. u(w) + jol'(w). wlllTC u', U' are real. (a) Show by.t Klmucrs.Kronig rela-
tion tklt

--
lim w d'(w) ~ -2
~
L- 0'(,) d, .

l11is restllt is used ill th e theory of superconducti\·ity. If at ''ery IJigh fTt."(llIencics

(sudl :u; :>.·r.. y frequencies) <I'(w) is ilit-ntiC31 for the superconducting and lIorrnal
stlltes, then we must have

r v.l.wl dw ~ r vJwl dw
But s t rn.:quencieli 0 <w< w~ within tbe superconducting energy gap the real part
of the conductivity of a superconductor \"alli~hes, so that in this regiOLl the illiegnli on
the left-Illlnd side is lower hy .. (T"w~. There must be lin udditioual COIltribution to <T.
to halance this deficiency. (1)) Sllow that ir u'.(w < wJ < u~(w < w~), as is observed
expetimelltall)'. then u'.(w) can Ilave a delta runcti on contribution 011 W" 0, and from
the dells function there i) II COIllribution ~(w) .. u'"w,jw. The delta function oorre'
spon J~ 10 Infinite conducth,ily at zero frequ('fl{.·Y. (c) By e\emenwl")' consideration of
the daMical motion of couduction elcctrOfl~ .II ,'ery high frequencies, show that

(CCSI L" u'(w) dw = 7mc/2t11

a result found by Ferrell ant! Clo,"er.

5. Dieleclrlt; oom'lanl and "Ie 3emicomluclor energy gap. The effect on t"{w) of an
energ)' ~p w~ in a ~mkonJuctor rna)' Oc approKimateti vCI")' roug!.Jy by )ub~t illliing
t6(w - w,,) for 6(w) in the re§}>OlIse function (16); that is, \vc take "'(w) ..
(27T11 ~IIIW)7TB(w - w.). Tlli) b crude becau)e it puis all the absorption at the gap

\
 frequency. The factor 112 enters as soon as Wf' move the delta function away from the
origin, because the integrJI in the sum rule of Problem 2 starts at the origin. SI,ow
that the real (ldrt of the dieLectric constant on this model is

t'(w) .. 1 + w!/(cJ,: - wil,

It follows that the staticdielcctric comtant is (f'(O) - 1 + ~w:, "'idt'll' used as a rille
of thumb.

6. Ilogen.Hube,.. rd(llion for in/rarccl rejlectidty of mclalr. 11,e complex nfracti\ e

hu\ex II + iK of a metal for WT <C I is given hy

(CCS) t (w) - (II + iKjl = 1 + 47rlur/w

where 170 is the conductivity for static Heidi. \ Ve assume I,ere that intrabaml CUTrents
are dominant; interband Iran~i lions a,'e neglected . Using the result of Problem 3 for
the reflectkm oodHcienl at nonnal incidence. show that

(CCS)

pfovide(llhat Un » w. Thh i~ the II WIl· Rube lls relation. For sodium at room tern-
perature, Uo =- 2.1 X 1011 s • in CCS and T "" 3 . 1 X IO- H s, as deduced from T =
uoIII/HC I • Kadiation of 10 p.m has w - 1. 88 X w" 5 - 1, so that the Hagell. Rubens
result )iloul<1 appl y: R .. 0.976. -nl(" rcsuh calculated from experimental \':Ilues of II
<lnd K Is 0 .987. I-li nt: If (70 » W, then 112 "" ~ . 111is )impliHcs the algebl';l,

·7. Davyt'111tJ split/illg of exciton lilltll. The Frcllkcl exciton hand of Fig. 9 is doubled
when there art two atoms A, 8 ill a pri miti ve cell. Extend the theory of E(lS. (25) 10
(29) to a linear (:ry~tal A8. AB.AB.AI). wit h transfer intcgrals 1', \x.·tween AB ,md '[ 2
between I).A . Find an Ctluation tOr the two hnnds ~ functions of the ....'a\'e\·wor. TIle
splitting between the bands at k = 0 is rolled the Da"ydov splitting.

on,. I'robkm is IOm_11:I1 cllme"I •.

""
Refer-ences
M. Balkans)";, aI. , Opticol PI'OI>erlie, of umkt>,ufuclorW, NQrth. HoUa.nd, 1980.
N. Bloembergen, Nonlinear eplin, Beujamin, 1965.
D. A. Shirley. ed., Elcctn'" Rpt'dnnoopy . North · Holiand, 1972.
M. Cardona, Moelllla/ion spectrwrol''J. Al'ad~'mic, 196'1.
C . D. Jeffries and L. v . Keldysh , 005., Electron-Iwle dro/lldr In UmfC:oIlifuc/Orl, North-lIol13nd,

'''''.
J. C. Phillips, ".'undame"lal optical s~ro Of 50Ild$,'- Solid slnt., ph ysics III, 56 (1 966).
R. Loudon . QUlin/lim theol-Y of light , 2nd ed., o~rord, 1983.
E. I. Rashb,a and M. D. Sturgt:. eds., E~'dlornr. North _llo1Iand. 1982.
D. C. Reynolds and T. C . CoII;n., £xcilOru, 1/1d.- propU1/u and ...su, Academic, 1981 .
K. 0 '0, EXcilom, Springer, 1979.
W. M. Yen and P. M. Selzer, fils., Laser rpectro&c:'l>W of sol/dr. Springer, 1966; ''01. II . 1989.
Y. R. 5]"".), Principia of nonli'l£{lr opla, Wile)". ]984.
Jo'. C. Brown, ~U1tra"i<:>le1 sl'ectroK'Op)' of solid>; with lise of synd, rotron radiation," Selid slale
ph)'5ics 29. 1 (19'74).
P. O. f'iluon. ~Optical properties or n>dali and .I!or$, ~ Solid 11:IIr. phYIic'S 29. 139 (19'74).
T. M. Rice. ~E1eclron-ho1e liquid in semirondUC"lon: thMreticalasptttli. ~ SolId Jlalr. pll)'5ics. 32. I
( 19'Tl).
M . Cardona. ed .• Liy)Il _ ' teri"fl In -elida. 3 vo .... Springe>". 1 982~83.
 12
Superconductivity

EXPERIMENTAL SURVE\' 33:S

Oa:urrcnce of superwoouctivity 337
Destruction of superrondUCCivity by nlagnetic faelds 338
Meissner effect 338
Heal cap3city 342
Encl"gy gllp 344
Microv.'lwc and infra red properties 3415
Isotope effect 346

THEORl-:11CAL SURVEY 346

Tllcnnocl),namics of the ~upen:Ollducting t l'1lnsirion 347
London equulioll 349
Coherence lengd! 3152-
BCS thoory of supercollducth'ily 3M
BCS ground stale 3S5
Flu.l quantization in a superconducting ring 3.56
l>..,,"lIlioo of persistent currclllS 3!S9
Trpc II superconductors 360
Vortel stale 361
l<:slimation of Il .. and Ilc'l 362
Single particle tunneling 364
Josephson superconductor tunneling 366
Dc Josephson effect 366
Ac Josephson effect 368
Macroscopic: quantum interrerence 369

HIGH ..TEMPl::RATURE SUPERCONDUCTORS 371

Criticul fields lind critical currents 372
Hall nwnber 373
Fullercncs 373

SUMMAR\, 374

NOTATION: In this chaptc.-r B. clcnotcs the applie<.l magnetic field. In the CC5 sy~tcm
the critical value BatO or the applied field willlle denoted by the symhol H~ in accordance
.....lth the custom of \VOfkers in supen::on(llK1ivity. Val11($ or R"" are given in gaun in CC5
units and ill te~las in 5 1 units, with 1 T = lOt C . In 51 we ha'·e UO<" .... fl.f//c<
 PROBLEMS 375

I. ~I agncl ic field pcnclratioo in Ii plate 37'

2. Critical field of ,"in nlms 37'
3. Two-nuid mood of Ii superconductor 376
4. Structure of a ,'ortex 376
5. L(lndUfJ penetration depth 317
6. Diffraction "ffe<:t of J osepliw n junction 377
7. Meiss ncr effed in Sllllcrtl 377

REt"EREl"CES 377

APP ..: NDlCES Rf<: LEVANT TO SUPEUCONDUCnVrn'

II Cooper I'airs 656
I Cinwurg·l.andau ECluatioll 658
, ElectrOll· Phonc)II Collisions 'w.

,u

----
.. ,
/'

: ~:i
"
,
'" ,
"
',~
"\
• • • •
Fifture I Reo..i,ra.~ on oh ms of. Specimen of mercury versus a!Jso], 'emp<' rature. Thi s pk>t by
K~me.lingh On ..... . mark..d tloe dl5cO'~ry c:l $ulM'r<'"oudUCI i,·ily.
---- CHAPTER 12: SUPERCONDUCTIVITY

The ek-'dncill resistivity of lTIilny metals and oIl1oys drops suddenl y to zero
when the spedmen is cook>d to a sufficientl y low temperature. often a te mpera-
tu re in the liquid helium range. This phenomenOlI, culled superrond ucth-ity,
was observed first by Kamerlingh Onnes' in Lciden in 19l L, three years after
he first liquified helium . At a critical temperahlre Tc the siX'"dmen undergoes a
pha!>e transition from a state of normal electrical resistivit y to a super<.."omluct-
hlg state, Fig. 1.
Super<."onductivily is now very well understood. It is a fiel d w ith ma ny
practical and theoretical aspects. The length of this chapter .I!ld the relevant
appendices relleet the rich ness and subtleties of the fi eld .

EXPERIMENTAL SURvlnr

In the sllpercondllcting state the de ek-clrical resistivity is zero. or so close

to zero that persistent c1t..-ctril.-a1 currents have been observed to flow without
attelHlation in superconducting rings for more than a yea r, until ut last the
experi men talist wearied of the experi ment.
The dl.'C3Y of supercurrents in a solenoid was studied by File and Mills2
using precision nuclear magnetic resonanee methods to meaSllre the magnetie
field associall.'ll with the supercurrent . They concluded that the del.-ay time of
the supercurrent is not less than lOO,OOO yeu rs. We estimate the decay time
below.
In some superconducting materials. particularly those used for super<."on -
ducting magnets, finite decay times are observed because of an irreversible
redistribution of magnetic !lux in the material.
The magne tic properties exhibited by superconductors are as dramatic us
their elcdrical properties. The magnetic properties cannot be al.'COlm ted for by
the a~sumpt ion that a superconduetor is a normal eonductor with zero electrical
resistivity.

>Ii. Kamerlillgh O,,"es, Abd. \lllI1 Welellsehappt:n (Alllsteroam) 14. 113, 818 (1911). ·-n 'e
viluc orlhe mem>ry n:s;slanre used was t 72. 7 ohms in tl,e liquid ~-ondII'On at O"c; e.lrnpolation
rrom the melting poillt to O"C hy meanS " r the tempernture coefficienl or solid me~ury Il:".~ a
relisiance conespondillg to Ihis or 39.7 ohms in the solid slate. At 4.3 K tlus had sunk to 0 0&1
oI,ms, thai il. 10 0 0021 times the res.islance which the wild mcrrury \\ouW have al We. At 3 K Ihe
msistance was round 10 have rallcn below 3 x 10 8 ohms, tt,at is to One ten-m,lIionth of the .-alue
which ,t wouW have at O"C. As tI,e temper-liure san k rurther to I .S K Ihis value remah,oo tI,e upper
limit of the resistance." Hislorical rcre..,nces are give" by e. J. Cortcr. Rev. Mod Ploys. J6, 1
09&1).
zJ. "ile and R. .hils, Phys. Rev. Lell. 10,93 (t963).

3:35
 Table 1 SUllerconuuclivily purumct('(. oi til(' ci('m('nls
~
U B, B C N a F N,

- .- -
An ast('risk denotes an el('ment sup('roonducting only in thin flIms or under
high pres~ure in a crystal modification not normally stable. Data courtesy of
0.026 B. T. Matthias, revised hy T, Gebllll('.
" _lr ~
-

N, M.
• AI 5i- P' 5' CI A,

,- Transition temperature in K
Critical magnetic field at absolute zero in gauss (10 'tesla)
l.f40
i05-
- -- - --
K

Rb
C,

5,
5,

V'
Ti
0.39
100

Z,
V

5.38
1420

Nb
-
C,'

Mo
M,

-
~

T,
F,

R,
Co

Rh
Ni

- -
Pd
C,

A.
Z,
0.875
53

Cd
G,

_1.09r
51

I,
G,'

--- - - -
As'

5n (.. ) Sb'
5,'

T,'
B,

-
K,

X,

---- - --
1

" 0.546
47
9.SO
1980
-0.92·- 1.77
95
- 0.51
l4l 0 70
.0003
.049
-- - 0.""-
30
340> ·'rt22
293 309 - ~

- -
C,' B,' la Icc HI To W R, 0, I, Pt A, Hg (~) TI Pb Bit Po At R,
~

- - _.,
~.-~
6~OO
1100
0.12 4.483
830
0.012
1.07 - 1.4
198
- -'
- .- --
0.655
65
0.14
19 -
...-:"". eo ~
4.15J
412
21]' J.193
171 803
~

F, R, A,
"
- - - -- - - - --- -- --
- -- - - ---- ,
C,' p, Nd Pm Sm E, Gd Tb D, Ho E, Tm Vb L,
"
- --- - --
p,
~- -
p,
~

E,
,- §;.C

-- -- - -
Th Ut (a) Np Am Cm Cf Fm Md No L,

- --
Bk

J'W
1.62
:rc. -~ ~

-- --,
~

-_ . .
~

- - -
Figure 2 M~I55ner cffect in a 5uperconducting spht'l"e rooled In a constant applied -magn~~ field,
on passing below the transition temperature: the hne l d induction B an!: ejected from the .phe re.

It is an eKperimental fact that a bulk superconductor in a v.'eak magnetic

field will act as a perfect diamaJ,(net, with zero magnetic induction in the inte-
rior. When a specimcn is placed in a magne tic field and is then cooled through
the transition temperature for superconductivity, the magnetic flux originally
present is ejected from the specimen. 111is is called the Meissner effect. 111e
Setluence of events is shown in Fig. 2. 'nle unique magnetic properties of
superconductors are central to the characterization of thc superconducting
state.
'me superconducting state is an ordered state of the conduction electrons
of the metal . 11le oruer is in the iOnnatioll of loosely associated pairs of elec.-
trons. The electrons are ordered at temperatures below the transition tempera-
ture, and the)' arc disorder<.-d above the tmnsition temperature.
The nature and origin of the ordering was explained b), Banieen, Cooper,
and Schrieffer.3 In the present chapter we develop as far as we can in an
elementary ,\13)' the physics of the supcrconducting state. \Ve shall also discuss
the hasic ph)'sidl of the materials used for superconducting ma&'llets, but not
their technology. Appendi<;es II and I brive deeper treatments of the superoon-
dueling state.

Occurrence of SII1JcrcorlJllctivify
Superconductivity occurs in many metallic elements of the periodic sys-
tem and also ill alloys, intermctallic compounds, and doped semiconductors.
The range of transition temperatures best confinned at present extends from
00.0 K for the compound YB3:tCU;,o6.Wto below 0.001 K (or the clement nit.
Several f-band superconductors, also known as "exotic superconductors," are
li~h.'() in Ch.lpter 6. Scveral materia1s become sliperconductin~ only under high
pressure; for example, Si has a sllpert"Ondllcling form at 165 khar, with
T~ .. 8.3 K. The elements kno,vtl to be supcrconducting arc displayed in Table
I, for zero pressurc.

3) BardL"c:Il, I. K Coopcr, and J n Schne ffer, Phys 11<...,. 106, 162 (1957), lOS. 1175 (ISSi).
 will every nonmagnctic metaHic c lement become a superconductor at suf-
ficiently low temperatures? \\le do not know. In experimental searches for
superconductors with ultralow transition temperatures it is important to elimi-
nate from the specimen even trJ,(,'C quantities of foreign parnmagnetic ele-
ments, because they eaulower the tran.~iti on temperature severel y. One part of
Fe in 104 will de:.troy the superconductivity of Mo, which when pure has Tc ::
0.92 K; and 1 at. perccnt of gadolini um lowers the transition tcmpcrature of
lanthanum from 5.6 K to 0.6 K. Nonmagnctic impurities have no very marked
effect on the transition temperature. Thc transition temperaturcs of a numher
of interesting snperconduetingcompollnds are Iistcd in Table 2. Several organic
colllpOunds show superconductivity at fai rl y low temperatures.

Destruction of Superconductivity by Magfledc Fieldt

A suffiCientl y stro ng magnetic fie ld will destroy superconductivity. TIle
threshold or critical value of tile ap plied mnguetie fi ekl for the destruction of
supercond uctivity is denoted by Hd.'f} and is a function of the temperature. At
th e critical temperature thc eritic-.i..I field is zero: HAT,,) = O. 111e variation of
the (;ri tical field ,,,;th temperature for several slJperconducting e lements is
shown in F ig. 3.
TIle thresh old curves separate tile superconducling state in the lower left
of lhe figure from the no n nal state in the uppe r right. Note: We should denOte
the critical value of the applied magnetic fi eld as BjU, but this is not common
practice among workers in superconductivity. In the CGS system we shaIl
always understand that II" a Boc. and in the SI we have H" - BoJlJ.<). l1le
symbol Bll d cnotes the ap plicd magnctic field .

Meissner Effecf
Me issner and Ochscnfeld (1933) found that if a superconductor is cooled in
a magnetic field to be low the transition temperature. then at the transition the
lines of induction B are pushed ou t (Fig. 2). The Meissne r effect shows that a
bulk superconductor bellaves a~ if inside the specimen B = O.

Table 2 Superronductivit)' of selected compounds

T" T"
Compollnd in K Compoutld ~K

Nb,Sn 18.05 V3Ga 16.5

N~Ge 23.2 V3Si 17.1
Nb,/Il 17.5 yHatCu3D&, 90.
NbN 16.0 HhtC'40 31.3
K,c;,O 19.2 ~Jl

-
lOA
 ",

, ..1-- -+ ----'''<:---+---\
I.

, , • ,
Tc,~p.'nIu,,·. In )(

}o'i,ure 3 Exl'crilllcnlnl 1I"".hold C1Jn~'S olthe coUiNl Ilcld 1/..(1) "cnlls Il'mllCfUlufe lOr >c,-end
supe«.'Ot>ductun. A .pcdnll'll ;s ~\lprreo"dlldmg I>f'klw the L'tll'"e nod normal ..Lo.c the l'U .... C.

We obtain a particularly useful forlll of this result if we limit ourselves to

long thin specimens with long axis parallel to B,,; now the tlcmagnetizing field
contribution (see Chapter 13) to B will be negligible. whence ;~
M I
(ees) B=B.+4nM::O; 0. - - = (I)
Bo 4w
M I
(Sl) H = I1d + 1J.oI\1 :::: 0 ; - 0. = -- = - .<If
B. ""
The result H = 0 C"olnnot be derived from the characterization of a ~lIpcr­
ronductor as a medium of zero resistivity. From Ohm's law, E = pj, we see
thai if the resistivity p gocs to .tern while j is held {illite, then E must be zero.
Bya Maxwell equation dB/cit is proportional to curl E. so that zero resisli\;ly
implies dB/tit = O. TIlis argument is not entirely tl~lnsparcnt, but the result
predicts that the flux through the metal cannot change on cooling through the
transition. -Ole Meissner effect contmdicts this result and suggests that perfect
diamagnetism is an essen tial property of the superoonducting lolate.
We expect another difference between a superconductor and a perfe<:t
conductor, defined as a conductor in whkh the electrons h."We an infinite mean
free path. When the problem is solved in dctail, it turns out that a perfc(.1
conductor when placed in a magnetic field cannot PrOOu('"e a permanent eddy
current screen: the field will penetrate ahout 1 cm in an hour.5

• Dbmagnctism, the magnetiutioo l', and the ~etic iU~pt,I.>iUty an: definro in Chapter
14. TI,e Il1lIgnitude of tile a.,pllrcnt dllmagneUc su~phbilily ofhulk ,upcrrooouctonl, ~-ery much
beRet" than In typtcaI diamllfl:l1dic sum~. In (I), M Is the magtM':tlZiltioo equl\IIlent '" the
s"Jlercollduct".g currents In thc spcchncn.
a A. 8 PlpP:lI"'d. I' 'Itl"oja of oondOCfIon d«,ron., Cordon nnd Breact., 1965.
(
,,0

"f).1t 1/

••,• •:,

. \ '001<'>

II"
"All....! .) fit'.'
I,
Ruprl lC B.-
II.~

}o'igure.. (<II) MliJ:ndil',alion '-erMIi IIpplird n~g;IIc1ic Ilekl r....

blJlk sup'.'l"conductor- e~hiLilj,lfC"
complete f-Idsoucr efT..-ct (pcrk'd diama,gnt'tlsm). A supc ,cooductur wilh this behavior u; coa/lcd a
type I SUII<:rt'Qt1titJ(.1or. AI)(We the cril iclll ncld II,. the spcdmCfl Is a n.. nnal conductor and 100
m~etiu:llon is 100 small 10 he seen On th is..,..1o: Note tM! minus "11M is plotted 00 the ,..,rtb!
!mk . the negative Yll}uc of M l"OtTC$ponds to d ,nmllgJ"lclism. (h) Supercoooucting mllflncli;aotion
"Il ......, cla In'" II .upf'r(.'(Jr1<luctor. lhc Illl' starlS to IX"><.iratc the spcdmen at II Ileld II" lower
dUll) the thermodynamic crItical lidd II ... l1lc specimen iI in a vor."" stat~· belW!X'J1 II" and 1/<1;,
and it has supeK'(lnducting d,..-hlea! PNJI)Crtlci up 10 111"1' 1.110\,,- 11<"1 IIII' speci men is II normal
rollductur in '''TfY . C5pcd, cJtCqlt fo,- pmsi,,Ic ~.~ cIT....,b. •"or g"Tn II, the area unde r the
tna~l1ct i~li<m curve is the $lime fur II t)'lle II sUJ>l! rronductor as for II type I (CCS unil s In all po.m
of Ih" frs:ur .... )

TIle magnetization curve expeded for a )upercondudor under the condi-

tions of the Meissner-Ochsenfcld expt'rimcnt j ) ske tched in Fig. 4a. TIlis ap-
plies quantitatively to a spL'(:imcn in the form of a long solid cylinder plrlCed in
a longitudinal magncli<' ficld . Pure specimens of many materials exhioit this
behavior; they arc called type I supercond\IClors or, forme rly, soft supercon-
ductor5. The vruues of fi r arc alw<I)'s too low filr type 1 sllpcn.'Onductors to have
any useful technical applkatioll in coi ls for super<:onduding magnets.
Other materials CAhiLlt a Illab'lletization cun e of the fOl1n of Fig. 40 and
arc known .IS type 11 superconductors. 'nleY tend to be alloys (as in Fig. Sa) or
transition metals with high values of the clC(."hicai resistl\'i!)' in the normal
state : thai is, the electronic mean free path in the normal state is sllOr!. We shall
sec later wll}' the mean free path is i]H olved in the "magncfi7.ation" of )Ilpcr-
cond.uctors.
Type II superumductors have sllpcrt:ond ucting ele<:'i]'iw.i properties up 10
a fi eld denoted by Ncz. Between the lower lTitical field Hd and the upper
critical field lid the flux density n "" 0 and the Meissner elfcd is said to be
incomplete, The value of Ild may be 100 times or more hi~er (Fig, 5h) than
the value of the critical field IIr eilculated from the thermodynamics of the
transition . In the region hetween IIc ] and Hc'l the superconductor is threaded
by flux iinestl.nd is said to be in the \orto state, A field Hc'lof 410 kC (41 teslas)
has been atl.lincd ill an allo)' of Nb, AI, and Ge ilt the boiling point of Ilelium,
and 540 kG (54 teslas) has been reported for I'L"to(;ss>
,
•, ,.,
Ar>I~;.,d "'''PM!'''' r",k] H. In I:>U»
Figure 5a Supcrwndtlding magnctiulion a,o-es of annealed poI)'CI}'$.t.llline lead and 1".1<1.
indium alloys at " 2 Ie (A ) lead, (8) lead- 2.08 "1. pcrttnl IndIum; {C} 1cad-8.23 wt. pt:tce'n'
indium; (D) kad-20.4 wt. pm:t'nl Indium. (Aflu U\'ingnQn )

'"

"

--- .... ,
....

Nb·T;

, ,
" "
TC"'1>",~ru.c. K " " "
I;ignrte 5b Stronger mag""'i,, fields than any now l'OII tcrnpiated III po-actil.-al supcrronducting
dC\~s an: "Hhln the callaLilily of cerlain Type II materials. 'Illes" material. ~W1not he elplo, lw.
how.,vcr. untU their "'; Iical current density can I.e •• ,sed and ulllill!tt.oy can I>t- f.. b",,3Iro as finely
divided ronductun . (M.gncl., fickh of ",<lrC than 111>001 20 Icsla.~ can L.- gencr.too (01)' in I'ulw-s.
and.so porUons oC the cuo't.'S ,11UYm as !wok.." Ii~. were ~u red in dwI way )
""
"

"
,".
l
E 10

05 -

"~~----",~,----..~,~.•::~---.,c•.------.,c•.-------c"o-------,,,,----
T~"'ptt1IIu, ... t::
Figure 6 Entropy S uf al"",lnunl in th .. IlOnn.d and ,up"K'(Inducting stales as a function of the
tempernlun,', "ll1e entropy i. lower in d,c $u l"-'rcond,,cting ~Illtt.' Lc.·(:~u,,, Ihe eI ...'CtI'OflS art: more
ordcrt.'tI hen: than in Ihe normal . Inle. At lilly '<,mpc"""' " hdow the ",ifical temperature T~ Ih ..
sp<:dmcn cat) he put in the norma. stoIC II)' application or a magnetic field ~Irongcr than Ihe critical
field.

Commercial solenoids wound with a hard superconductor produce high

steady fields over 100 kG. A "mlnl superconductor" is a t)'PC II supcrcondlK:'fOr
with a large magnetic hysteresis, usually induced by mechanical treatment.
Such materials have an important medical application in magnetic resonance
imaging (MHI).

Heat Capacity
In all supcrronducton the entropy decreases markedly on cooling below
the CTiti~11 temperatllre Te:. Mea.~uremcnts for <lluminum arc plotted in Fig. 6.
The decrease in entropy between tJle normal state and the supert:onduding
statc tclls us that the superconducting state is more ordered than thc normal
stnte, for the cntropy is a measure of the disonler of a syste m. Somc or all of the
e lectrons thermally excited ill the normal statc are ordered in the supcr«ln-
ducting state. l1le change in cntropy is small, in aluminum of the order of
10- 4 ks per atom. The sn1<l1l entropy change must mcan that only a sm..'111
fraction (of the order of 10- 4 ) of tJle conduction electrons participate in the
transition to thc oruered sllperconducting state. The free energies of nurmal
and superronducting states arc compared in Fig. 7.
Thc heat capadty of galliu m is plotted in Fig. 8: (a) compares the normal
and superconducting statc s ~ (b) shows that the electrt' .<;: ("Ontribution to the
he-dt capacity in the superconducting state is an cJCpOnl_ .al fOflll with an argo-
 ',---
- Ol '.
.......,
1 -OA I-____"'---____
,
;:; - 0.5 SUI"'r<OIKlu<'tor

.5 - 06

f - 0,7

l - 08

- ,"

- 10
- 11
- 12
,l---"----"---"----'--.,<,..--"'---"----"---"---.,."C--"----'---"----"---c!•. 5'

T'''''I''-'fature. K
Figun 7 E)Jl't'rinlcntnl v:llues ofth.. free o>nCll:)' ~ a fundion oI" le rnpo:rature for aluminum in the
sup'-'fC.'O ndu.:ting slate _....1 in the normal state. Below the IroUlsilion temperat"re T, - 1.160 K th..
fro:e cnc~ is km~r In the St'IICrconducting Itatc, 'I'll., Iwo CUI"'C'II mc~c;u the transition tempera-
t ure, $0 thai the ph.sc t ran$il ion is ~>Tld order (there Is nO IOlenl heal "f l ran.ition al Tel. The
rurve F~ Is rlll-ru;uret! ,,, "ero mllgnctic lidd, and F", is me.s"r..~ t ,n a nl,'gneli<' fidd s"ffic1enllo pu l
Ih .. SIX',:;", .. n in th .. "onun! ~lolc. (Courtesy of N. E. PI'ilUp, .)

1.:; -
0 """'.
8.-200 (;
• B. _ O I
~ "'I
t
,. Co1'~

$~"" "II
, /.- I -
"
;
<,,<>".
" I
'1.,. " t ~,
I" t
, ~"-"-.
u

•"-
u :- crr .. 0596.0~T"
• O.OJ

i
•
"
"

,
/
" " 10
"
0.00 1
, , , , ,
~. K'
I.
',"
Ib)
Figure 8 (al n.e
heal capA,,,ly of gallium In the nomw and suvereooduCli ng nates. Th ... n()rmal
Slate (...hleb il restorcd by 1200 G rlCld) has eit.>ctronic, latt la-, and {at low to~ml,...raturcil nuclc-ar
qUlld, upolc oonlrihlltion~. In (b) the elect ronic pa,t C .. of lhe h(>lll cllj)a(ity III the silperconduelmg
~Iatc II 1'1011...-<1 on a "'1: $l-ak: vcrsus 'r J r: the CXl'on('tll;nl dcpcndc I>l'e on l iT is e' ,denl. Ucrc
l' - 0 60 onJ IOQI- ' i, (Arter N. E. I'hi lli pl.)
,.,
men! proportional to - lrr. suggestive of c\cit~tion of clt.x . IIS i\C.TOSS an en_
ergy gap. An cncll,'Y I}\p (Fig. 9) is a c!Jar..K.'tcrist ic, bill not uni\'cr~al , feature of
the slipert.'o m]ucting st-.ite. 111C gap is ;l(x:ountro for Ily the Bardeen-Cooper-
SchlicfTcr (ReS) theory of superconductivity (see Appendix H),

Energy Cap
The energr gap of 5npcrm TlductoTs is of entirely different origin and
nature IllOu] thc CllCr},'Y gap of insulators. In lIlI insulrlto .. the energy gap is
caused by the electron-lattice inlcractio:l, Chapler 7 . This intcmction tics the
clcctrous to the lattice. In a supcn:.'Onductor the important illtcmclion is the
electron-electron interaction wh ich orders the e lcctl'Ons in k space with rCspl."Ct
to thc Fermi gas of electrons.
The argument or tIle CX[XlnCnliOlI f"dor in the clc<:lronic heal capacity of a
slIperconduetor i.s found to be - E,,l2ks T ltnd not - E,:/knT. This h,LS beef} learnt
from comparison with optical and clC(:tron hmneling dcterminlttions of the gdp
Ell. Values of thc gap in sC\"cral sllpercoodudors arc given in Table 3.
llte tr.msition in zero magnetic field from the Sllperconducting state to the
nOnllal state is observed to be a sL-"CQnd-oruer phase transition. At a sccond-
order transition therc is no latent heat. bllt there is a d iscontinuity in the heat
t:dpacHy. evident in Fig. Sa. Further, the energy gap decreases L'Ontin uously to
zero as the temperature is increased to the transition tempcrnture T~. as in Fig.
lO. A first-ordcr tmnsiti(lll w(luld be characterized hy a latent heat and by a
dismntinllity in the energy gap.

ToLlc 3 Energy g."lpS in supcrconduclon, III 1" = 0

AI Si

f , {O) In 10 4eV.
f~O )lk ... r~. ,.,
'.4

S< Ti V C, M" f, Co N' Cu Z" G. G,

15.
3.4
2.4
' .2 ,.
'3
Y b Nb Mo To Ru Rh Pd Ag Cd I" So (..)

30.'
' .eo
2.7
'4
I.'
' .2
10.5
' .5 ,.,
11.5

La "" HI fa W Re 0, I, PI Au Hg {_j TI Pb

19. 14. 16.5 7.35 27.3

' .7 ' .60 4." 3.57 4.38
 "

ngu~ 9 Cal Conduction 1.00 in the normal sbtc; (hl ent'Q," y g;ap at the . "cr ml level in Ih('
sup<'l'C'OPdllCting stat ... Eke-h om in ....xcil.,d .tai L'S .Lm.·c the gap behave as normal e!co.1 rons in rf
f,ehk: they cause resi~talll.~, It de they are shorll..t ou t by the Rlpc:r<:ood ucti ng cledrons. The gap
E. is e.uggcraICd in the Ilgure: typically E, - 10- < ~~"
.,
I
, ""- ...,
•.'1- •
, ~.
BCS <'III",. ~

•o TTaobl.un
• N
, o NIOI .....,n
C-
,•
,
o
l;igure 10 Hcduced "nlue" of Ih .. oiJserV(.-d
energ)' gap E'/'TYE,!.O) as " fund ion of the re-
,,
duced temperature Tffc • aflcr TOWlIK'fld I ud
Sullo". The ~ id Ctlrvc Is clra" '" for th .. 8CS • 0, 0 ,
" "
., .., ••, " o• ,.,
....". Trr.

Microwave arid l1ifrarccll' roperties

' n )C existence of an e nergy gllp mCllns that photons of ellerg)-' less than the
gap cnerb'Y arc nol absorbed . Nearly all tile photons incide nt are reflected as for
any metal because of the impedance.- mismatch a l the boundary between \'a~­
UIIIIl and metal. bllt for a very thin (-2.0 A) mm m Ore photons ar c tmnsmilted
ill the supcl'conducting state than in the normal state.
For photon energies less than the energy gaP. the resisth'ity n super- or
conductor vanishes at absolute zero. AI T <c '1 -.. the resistance in the supcrcon-
ducting state has a sharp threshold at the gap energy. Photons of lower energy
see a resistmccless surface. Photons of higher energy see 11 resistance that
approaches that of the nonnal state lx.'Cause such photons cause transitions to
unoccupil.'ti nonnal e nerb'Y le\'c1s above the gap.
As the temperature is increased 1101 only docs the gap dl.'crease in ellcrgy.
but the resisth'ity for photons with energy l>ctow the gllp energy no longer
\~nishes, exCt."{l1 at zero frequency. At zero frequency the superconducting
electrons short-circuit any normal electrons that have beel! thennally excited
above the gap. At finite fn..'tluellcil.>S the incrtill of the supcrconducting elec-
trons p revents them from completely scrCl.'fli ng the ek'Ctric field , so that ther-
mally exl'itl.'ti nonnal eit.'Ctrons now can absorb energy (Problem 3).

l~otope Effect
It tUl.~ bl.'Cll obser\'l.'ti that the critical temperature of superconductors
\'aries with isotopic mass. III mercury l'c valies from 4. 185 K to 4.146 K as the
8\'ernge atomic mass M valies from ]99.5 10 203.4 atomic mass units. The
transition tcmpcratUl'C cll<lnges smoothly when we mix (lilfefl."nt isotopes of the
same cle ment. The experimental results within each series of isotopes may be
fitted by a relation of the fon n
M"Tc: = ('Onstant (2)

Obsen'cci values of a arc given in Table 4.

From :he dependence of Tc: on the isotopic mass we h.-a1"l1 that lattice
vibrat ions and hencc ck'CIron-tatlicc intcractions arc dl.'Cply il1\'olvcd in super-
conductivity. This was a fundaml."Iltal diSCO\'cry: there is no other rl.-ason for the
supcrconducting transition tempernture to depend on the number of neutrons
in the nucleus.
l11C original BCS Illoclcl gave the res ult Tt:; IX 6o,.l»~ IX AI - I12• so that a = j-
in (2), but the inclusio n of coulomb illtemctions l>ctwl.'Cn the ek'Ctrons changes
the relation. Nothing is Sa(.Tl.'ti about a = i. 'l1lC absence of an isotope cffect in
Ru and Zr has been accounted for in terms of the electron band stmcture of
these metals.

THEORETICAL SURVEY

A theorctical understanding of the phenomena associatl.'(l with sUIX.'rcon-

duetivity has been reached in SC\'eral ways. Certain results follow directly from
thermodynamics. Many important results can be described by phenomenologi-
cal eq\lations: the London '-'(Iuations and the Landau-Cinzburg equation s (Ap-
pendix 1). A successful quantum theory of supcrconducth'ity was given by Har-
d een , Cooper, and Schrielfer, and h:lS pro\ idl.>d the basis for subsC<luent work.
Josephson and Anderson discovered tll C importancc of the phase of the SUI)Cr-
conducting wavcfuTlctioll.
 Table 4 Isotope effet:t ;n superconduo:::tors
Eq>crimcntl.1 ''aluel or Q in M"Tc = CQI15ta nt, where !If is the isotopic mass.

Sulnlanc.: • Sub..!anc:e •
Zn 0.-1.5 ~ 0.05
Cd 0.32 ~ 0.07 "" 0.00 ~ O.OS

So
H/o(
0.47 ~
0.50 ~
0.02
0.03
'"".
N~n
0.15 ~
0.33
0.08 :!:
0.05

0.02
Pb 0.49 :!: 0.02 U O.OO :!: 0.05

]'lumlIOtlYlwmicll of the Supercolltiliciing Truwritioll

l11e tmnsition bctweelJ the norm,ll and supercollducting states is thermo-
dYllamk-ally fC\'ersible, just as the transition betwccn liquid and vapor phases
of a substancc is reve rsible. Thus we m,IY apply thel"lnlxl)'nalllics to the transi-
tion, lind we the reby ohtain an expression for the entropy differencc between
normal and slLpe rcondllcting states in terms of the critical field curve lie versllS
'1'. -nis is amdogolls to the v,lpor pressure (.'(luatioll for the liqUid-gas coexist-
ellce eu ....'& ("fP, C hapte r lo).
We trcat a type I superconductor with a complete ~ I eissner cfft.'c:i, 50 that
B = 0 inside the superconductor. We shall Sl.'C that the critical field I1~ is a
quantitative measu re of the free energy difference hetween the superconduct-
ing and normal st.ltes at constant tempera ture. TIle symbolll r will alwa)"s refer
to II bulk specimen, ne\oer to a thin mill. For type II supcrconductors, fir is
undentood to be the the rmodynam ic critic.\J field related to the stnbilizatioll
frt.'C clle'l,'Y.
'111c stabilization free energy of the superconducting state with res]wct to
the normal state can be (!etermined by calorime tric or magnetic mellSUI"e-
ments. III the l'alorimctric method the heat capacity is mt."'3Sun.."(l as a function
of temperature for the superconductor and for the normal conductor, \\hich
rneans the superconductor in a magnetic field larger than He. From the diO'er-
encc of the ht.>at capacities we can compute the frcc energy differcl1C'e, which is
thc stabili7.ation free energy of the supcrconducting state.
In the magnctic method the stabili7.ation fr(.'C cncrgy is found froUl the
valuc of the applied magnetic field that will dcstroy the sUl>crcond ucting state,
at constant temperature. The argument foIlO\<ls. Consider the work done (Fig.
ll ) on a superconductor when it is brought reversibly at constant temper,ltu re
from a position at infinity (wherc the applied field is "lero) to a position r in the
field of a pcnnanent magnet:

W = - i"" 1\1. ,fU.. (3)

...
-~
c==::» Supcl't<ln<lu<1o.-

.)
Figure Il Ia) A sUPf'«lOI'Icluctor in ....·hid, Ih~ Meissner dftct is complete 1_ B - 0, as if the
rnagneti~tion wcrt /of .. - 8.14"" in CCS units. (b) \\'hl,n the applied rICk! read," the ,..due- B~
IIIf' nonnal slate carl coexist in equilihrium ""Ih the supr.roDl1dud"'g slatC'. In l~ulence the &ee
energy de~itiM are equal: I' ..~T, 8 _ ) - Fsrr, 8 ..,).

per unit volume of specimen. This work appears in the cnCl"b'Y of the magnetic
field. 111c thermodynamic identity for the process is
(IF '= -l\-J . dB" • (4)
a.. in TP , Chapter 8.
For a superconductor with 1\1 rclawd to B" by (I) we have
. I (5)
(CGS) elf's = - 8" dB"
4"

I
(S I) dFs "" - B" d8~
""
The increase in the free cllcrb'Y density of the supcrconduclOl" is
(CGS) (6)

(51) Fs(B,,) "'" Fs(O) = B~I-'o ,

being brought from a position where the appli(.,(\ field is zero to a position
011
where the applied field is B".
 J2 SUfII'rconduct;" i/y 34!

Now collsi a nonnal nonmagnetic metal. If we ncglect the small susccp-

tibili t ~ of a metal in the normal state, then M = 0 and the energy of the nonnal
mctal is indepcuclellt of field . At the critical field we have
(7)
The results (6) and (7) arc all we nt..>ed to detcrrnine the stabilization energy
of the supercondueting state at absolute zero. At the critical ...alile Dae of the
apptk<\ magnetic field the energies arc l..'"<luru in the normal and supt!rconduct-
ing stati.'S;
(GGS) (8)

(SI)

In SI units li e - B"dlJ-ih wherea<; in CCS units H e - D"c'

'J1lC specimen is stable in either state wilen the applit.'·d field is l.'"<lual to the
critical field. Now by (7) it follows that
(GGS) t::.F - F,..{O) - Fs{O) = D!.I811" • (9)

where t::.f' is the stabilization fH.oc cncrgy density of the supercondllcting state.
For aluminum, Bae at absolute "Zero is 105 gauss, so that at absolute "Zero tJ.F =
(105)2/811" = 439 erg cm - 3 , in excellcnt agTt..'elllcnt with the H.'Sult of thennru
mC"olsurements. 430 erg cm - 3 .
At a finite temperat ure the normal and supcn:onducting phases arc in
equilibrium when the magnetic field is such that their free energies f' = U -
TS are t..'"<lual. 'Ille fwe cnergil..'S of thc two ph.\.<;es arc skctche(l in Fig. 12 as a
function of the magnetic ficld . Experimental curvcs of the free encrgil.'S of the
two phast..'S for alum inum arc shuwn in Fig. 7. lll."'Cause the slopes <iF/efT arc
equal at the tra nsition te mperature, there is no latent heat at 1'.,.

London EqllfJtioll
We saw that the Meissner efft..'Ct i1T\plil..'S a magnetic SU~1}tibili ty X =
-\/411" in CCS in the supcrconducting state or, in SI, X "" - I . This sweeping
assumption tends to cut off further discussion, and it docs not account for the
Ilu" penetration ohserved in thin filllls. Can we modify a constitutivc equation
of electrodynamics (such as O hm's law) in sollle wa}' to obtain the Meissner

6"fhi$ is an adeqllate as.lllllpiiOll for I)1IC I Sllpo!'r("Qnullcturs. In typo!' II su pl"rronullctOls in

higll liekls thl" dll\nge in spin pilramagllCtbm of the t"OlldoeliOn elf.d:.-ons 10....l"" tile energy clthe
normal ph.asoo signirlCanli)·. In $OffiI'", bUI nul all, t)-pe II superronducton the "pp"r critical fleW is
limited b y this elTect. Clogston has suggested that H."(m,,,,} - 18,400 T" "tlCf"C H4 ;,; in gaucu and
Tc ill K. S..e A, M . Clogiton. I'h~s. l\ev. L..:1t. 9, 266 ( 1962); B. Owndn.sd,har. ApP1 I'h)~. Lett.
1.7 (l962l.
 "
'\oi,."..,j >lat~

'-
"1'\.. ,,-..1 "'«"'~'" f,dd ll. -

Figuu 12 11.., r,ce CflerJ:Y den.it)' F, r:A .Inonm...:ndic nClrmalll1d~l is ~pprmir""rd)' indCI'Cn-
<knl of the iI1IM"I)' of Ihf. applied n.agM'tlc fM'ld B•. At .I temperature T < T, thO!' n .... t.d " a
sl1percm"luctor in n:ro magnet;., fH-,kl, SO lloat F~(T, 0) is 10 .....>.:. thAI. F",(T, 0} An "pl'[ied magnetic
Geld j""f'C.IS<.-S F. by B! I8w-, in CCS units, 50 that ,.·stT, 8 .. ) - F,(T, 0) + sa) 8'11' Ir 8. 1$1arget" Ihan
,hc<'-IlI ....,.! fK'ld lI .... lh.· rr~ enc' ~ density,s !o.>·t.T in II .... normal.latc th..... in the supcn."Onduclm)(
state. a.,<1 ."-",, the nonnaisLoIl1" is the stable ~t.ll l!' . TIle origin of Ihf. ,,,,rt;"-aI ...Je III ti,e dnwinK is
at ,.', (T, 0). The figure equally applies to U~ and UN at T - O.

L>{fect? We do no t wallt to modify the ).1axwc ll cllualiolls themsek es. Ek·ctrical

conduction in the nonnal state ofa lllCt[l1 is dCSCrilx.'(\ by Ohm's law j = uE. \ Vc
need to modify this d ras tically to describe (.'QllUuction and thc l\lcissncr effect
in thc supcrconducting state. Lt:t us make a llOlItllitJte and SL'C what happens.
\Vc postulate that in thc supcrconducting state thc current dcn~ity is di-
rectly proportional to the vcctor potential A of thc loc-al magnctic ficld, where
B "" curl A. The gauge of A will bc spL'Cified. In CGS units we write the con-
stallt of proportionality as -d47TAi. for rcasons that" ill oc'OOrne clear. IIcre e
is the speed of light and AL is a constant with thc dimensions of Icngth. In 51
units wc writc -lflJ.oAf.. Thus
I
(SI) j --- A (10)
I-!-tv,f.
nlis is the Lonoon etluation. \ Vc c'\;press it anothcr way by taking the eml of
both sides to obtain

(CGS) curl j "" -~ B (51) curl j -= (II)

471"A,_

The London equation (1O) is undCrStood to be writtcn with the \"ector

potcn tial in the London gaugc in wllich div A "" 0 , and A" "" 0 on any extcmal

c."
 •

'. •

I-'gure 13 PMW:t .... tilll\ vi an applic-d mago'''''k' flt"kllllto,. ~mi·inr;nile superconductor. n.e pen ..•
,,,,lion depth ,I. IS d .. r.n~·c1 <IS the «htmlCt: in which the 6 ..1<1 d«n:ases by the r.'lctor . - ' . Typically,
It ... 'iOO A. in a PUrf' ,upcn.vnduct()l".

surface thl'Ough whiclL no external current is fed . 111c subscri pt n denotes the
compo nent normal to the surface. 11\115 div j ::: 0 ,I.,d j~ '" 0, the actual physical
boundary conditions. 111c fonn (10) appliL-s to a simply connected Supt~ rcoll(ll1 c­
tor; additionru terms may be present in.1 ring or cylinder , but (11) holds true
independe nt of gl..'OlIlctry.
First we sllow that the London t.'quation l ead~ to the l\leissncr effect. Bya
l\hl'\;wcll equation we know that
4n
(CCS) curl B : : : - j
, (5) curl B ::: f.4Jj (12)

unde r static conditiolls. \Vc take the ClIft of botl. sidl..'S to oblaill

(CGS) cllrl curl B = - V'28:: - 'r-, cli rl j

(51) curl curl 8 = -V28 = 1'-0 curl j ,

which may be combilJed with the London I;..'q uation (II) to give for a supe r-
cofl(luctor
(13)

111is e<luation is sccn to account for the Meissne r e lTect because it docs not
allow a solution uniform in space, so that a uniform magnetic field cannot exist
in a supercomluctor. 111at is, B(r) "" 50 = constant is not a solution of (13) un-
less the constant fi eld Bo is identically zcm. 111e result follows because V2 Bo is
always zero, hut BoIAl is not zero unless 50 is zero. Note further th.d (12)
=
ensures that j 0 in a region where B "" O.
In the pure superconducting state the only field allowl,-cl is expone ntially
<lampl,><:! as we go in fmlll an external surface. Let a sem i-infinite s\lpert:onduc-
tor OCCUP)' the space on the positive si<Ic of the x ;Illis, as in Fig. 13. If l.l(O) is
the field at the plane boundary, the n the field inside is
B{x) "" B(O) cxp( -;t/ Ad • (14)

\
'"
for this is a solution oHI3). In this example the magnetic fl is assumed to bc
parallel to the boundary. Thus we Sl.'C AI. measures the depth of penetration of
the magne tic field ; it is known as the London penetration depth . Actua l pene-
tratio n depths arc not dc.:scrilx.,(\ pn..>cisely b y A,,. a lone, for the London l."<luation
is now knO\\l1l to be SOIlll..'whal o\'crsimplifk-d. It is shown by comparison of (22)
with (II) that
(eGS) At. = (mc'l147Ttlq~1I2 ; SI) A,. a:: (£{)mdllJl(r)112 (14a)

for particles of charge q and mass III in concentration II . Values are given in
l a ble 5.
An applied magnetic field B" will penetratc a thin fil m fairly unifon nly if
the thicb H.'sS is much less thu n AI.; thus in a thin film the hlciss ner cm...d is not
complete. In a thin film the inrlucL'<i field is IIll1ch !Loss than Bt" and there is
little elTec t of On on the energy density of the sllp( ·rconducting state, so that (6)
docs not apply. It follows that the critical field 11~ of thin films in 1)'1rol1c1
magnetic fields will be very high .

Collerellce Lellgth
The London pene tration dep th At_ is a fundmnentallength that character-
izes a superconductor. An independent length is the coherence le ngth {. 11lC
coherence length is a measure of the d istance within which the sllperconduct-
ing electrOIl concentration cannot change drastically ill a spatially-varying mag-
ne tic field .
The London equation is a local equation ; it relates the c utTe nt density at a
point r to the vcctor potential at the same pOint. So long as j(r ) is given as a
constant tinu."S A(r), the current is n ..'quiroo to follow exactly any variation in t he
vector potential. But the coherence le ngth ~ is a 1l1t-"3S\1l'e of the range over
which we should average A to obtain j . It is also a measure of the minimum
spatial extt-'n t of a transition layl..'1· between normal and superconductor. The
coherence length is best int roduced into the tht-'Ory through the Landau-
Ginzburg t-'{ Iuations, Appendix 1. Now we give a plausibility 3J'gument for the
energy rl..'{Juired to modulate the supcrconducting ek>ctron conccntration .
Any spatial variation in the state of an elect ronic system !"l..'(luires ex.tra
kinetic e ne rgy. A modulation of an eigenfunction increast-"S the kinetic energy
bt-'Cause the modulation will increase the intt-"grJ.1 of d2f{Jfdl"~. It is reasonable to
restrict the spatial \'ariation of j(r) in such a way that the extra energy is k'Ss than
the stabilization energy of the super(;onducting state.
We compare tile plrule wave ~x) .:: eib with the strongly m()(luinted wave-
function;
(l 5a)
 12 Supntond "ctkily

l ilr J Calculated intrinsic rohercnce length and

LOndon penetration de pth, at absolute 'a!ro

Inlrinol.... Pippar,j Loodoo

,,""'~ penetratIon
It'n!!th 61. ',jt'pth A,••
Metal in 1O- 6 cm in 1O- 6 an 'df.
S.. 23. 3.4 0. 16
AI ,GO. 1.6 0.010
l'b 8.3 3.7 0.45
Cd 76 11.0 0. 14
Nb 3.8 3.' 1. fr2

•
Ant', n. ~Ics('rv('y and Il. U. Schwartz.

The probability density associated w ith the plane wavc is unifon TI in space:
4I'oJI = e- 1k , Jib: = I . whereas ip*cp is modulated with the \va,vcvector q:
ip*ip = *(e->!i+q)l- + e- ;I;'Xel(h'l)l- + c'h")
(l 5b)
::: *(2 + e~~ + e- 1q ,) = I + cos qx
1'ne kinetic energy of the wave ~x) is t" = 11 2 1(212111; the ki nctic cncrgy of
the modulate<1 dcnsity distribution is higher. for

I (Ix ip* ( -~ ~) ip = ~(~) I(k + q)2 + kiJ &II ~k2 +.if...- kq ,

2111 dil 2 2111 2m 2m
wherc we negk "Ct q2 for (I q; k.
111e incrL'fIse of energy re<juire(\ to modulate is h2kq/2m. If this increase
exceeds the cncf!,'Y gap E/I.' superconducti vity will be dcstroye<1. TIle critical
value qo of the modulation wa\'C\'L'({OI" is given by

We defi ne an intrinsic L'Q herencc lenb'lh

.'
- - k"{lo ::: Et!. .
2m
~
(100)

rclate<\ to the Cl·itical lTl{Xlula-

{ion by {o ::: IIqo. We have
(161,)
where v,.. is the ek"Ctron vc!ocity at the Fe rmi surface. On the BeS theory a
similar result is fou nd (QTS . p. 173):

(17)

or
CnlculatL"(1 \'alucs {o from (17) arc given in Table 5. The in trinsic m herence
length {o is characteristic of a pure superconductor.
 In impure materials aud in alloys the coherencc length {is ~horter thall &J.
This may be understood qualitatively: in impure 1I1.Itc riai the e\t."Ctron eigcn~
functions already ha\c wiggles ill them: we (:an construct a given localized
\'ariation of current density \,itl! less cncq,'Y from w3vefullctiollS ,vith wiggles
than from smooth wavcfullctions.
The coherence length first <lppcarro in the Lnndau-Cinzhmg equaliolls;
these equations also follow from the lies theory. -n ley describe the structure of
the transition layer betwL"Cn normal and sllpercondllcting phases ill contact.
111c coherence length and the actual pellt'tration depth A depclld on tile lnL>un
free pat ll f of the c1L"Ctrons measured in the normal lotate; the relatioll~hips arc
illdicated in Fig. 14. Whell thc slIpercondl1(:tor is very impure, witll a very
small f, then t ~ ({uf)I12 and A ... A,_ ( ~ ,1t') I I2, so that Aft - A,jt. "I1lis is the
"dirty superconductor" limit. TIle ratio Aff is denotLoU by 1(.

Res Tlleory of sIIJJeJ"Cotuillctir;ily

111e lJasis of a quantulIl theory flf superconductivity was laid by the classic
1957 papers of Uanlccn, Cooper, and Schrieffer. 7 There is a hues theory of
superconductivity '\1th a "cry wide range of applicahilily, from lIe3 atoms in
M

their condensed phase, to type I and tn>c II mctallie slIl>CfCOmluctors, and to

high-temperature superconductors based on planL'S of ellprate ions. Further,
t
there is a "ues w;Jvefunction" compo!>Cd of particlc pairs Ie and -Ie! . which,
when treated by the Bes theory, givL'S tile tuniliar electronic sllperconduclh'-
ity observed in metals and exllibits the ene'l,'Y gaps of Table 3. This pairing is
known as S-W3\'e pairing. lhere arc other fonns of particle pailing possible with
the Bes theory, but we do 1I0t have to consider otlie.- than the DeS wavcfunc-
tion hcre at this time. In tlJis chapter we trL-at the specific accomplishments of
BeS theory with a BeS wavcfunction, \\hich inchule;

I. An attractivc intemction betwL"Cn electrons can lead to a ground state

separated from excited states by an e ncrgy g.1p. '111e critical ficld, the thermal
prol>crties. and most of the eiL"CtTOmagnctic prOt>crties arc conSL,<]uem:es of the
energy g."lp.
2. '!lIe electron-lattice-electron intcl'U(.1.ion leads to an c nerb'Y gap of the
observed magnitude. TIIC indirect interaction prOCCL'ds when one elcctron in-
te.-acts with the lattice and r1cfonns it; a second e\t.'Ctroll SL'CS the deformed
lattice and adjusts itself to take advantage of the deformation to luwer its en-
crb'Y· l1lUS the second electron interacts with the first electron via tile lattice
defonnation.
3, lne pe netration dL'Pth and the coherencc length emerge as natural
conscquences of the DeS theory. The LoIl(lon L"<)IL3tion is obtainL>d for mag-
netic fields that vary slowl)' in space. 'JlIUS the central phenomenon in super-
conductiVIty, the Meissncr effect, is obtained in a natural way.
1J. ~n , L N. Coo~r, and J. It &hnt-ffn, Ph)'S. Roev. tOG ., (1957). 108, 1175 (1957)

,. -
 12 S"~rc(lnd"cli.:;t!l $~

....gure 14 Pe'lClntiun depth A

~nd Ihl" cnhenmr.'C len!,'!!! f ,\I
.., A
r""chOl'$ of tire ITKl:m r~ patl. l
..rlhe conouctlOlll"la1rons in Ihe O• ;0
"".. lUll 51.11.,. A ll I,-nglhs are til
"nils of ". thl" Inlrimlo.: coh~­
0.3
l"nee I",,~th. Thl" l'Urvt'S arc
0.2
sklch.:d for £:, - lOA,. For dlOtl
meml free I"'tm II", cohc:renct:
0.1 -
len~1I1 1X"..'(:Imes ~hortl"r antl the
pclll1r.l1ion (\t:pth bc."COfI>r.1i
Io"';l"r. 11.c irrc ...~ in the "diu
0
0 ,
d/f fa,...,1'$ Iype II sup.,«'orn!"c-
ti\'il)'.
£,--

4. The criterion for Illc transition temperature uf an c1emt.'Ilt or alloy in-

volves tile c1t.'c tron density of orbitals D(E~.) of one spin at the Fermi level and
the electron-lattice interaction U. which can be estimated from the electrical
resistivity ix..>(;3use the re.~ isti\'ity at room temperature is 3 measure of the ek'O-
tron-pluJIlon interaction. For UIJ(E,,) ~ 1 the BCS theory prt.'tlicts
T.,= 1.14gexpl -IIUD(E~')} • (18)

wherc 0 is tile Deb)'c temperature and U is an attractive intcraction . l11e result

for 'I'., is satisfi(..'<l at least qmditativcly by the eKperirncntal data. 11lerc is an
interesting apparent p.1.r:tdoxl the higher the resisth'ity at room temperature
the higher is U. and thus the more likely it is that a metal will bc a Super<.'On-
tluctor when cooled.
5. Maglletic flu){ through a supcrconducting ring is quantized and the ef-
fective Illlit of charge is 2e mther than e. 111c ues ground state ilwolvt.'S pairs of
ch..'Ctrons; thus nux ljuanti7.ation in terms of tile pair chargc 2e is 3 consequence
of the theory.

BCS Ground Siule

The filk-'(\ Fermi sea is the ground state of a Fermi gas of non interacting
e\t.'Ctrons. 'nlis state allows arbitrarily small excit"'tions-wc can form an ex-
dt(..'(1 stale by taking an elt.'Ct ..on from the Fermi surface and ..",ising it jUl>t above
the Fenni Slilface. 'Ole BCS tlK.'Ory shows that with an appropriate atlmcth'e
interaction betwccn elC(:trons the new grollll{1 stale is sllpcroonducting and is
separatc(1 by a finite energy Ell from its lowest e){cih.-'(\ stale.
The fonnation of the BCS growul state is suggestl.'<l· by Fig. 15. 111e DCS
state in (b) contaillS admixtllrt.'S of one-e\t.'Ctron orbitals from above the Fermi
ellcrgy €"" AI first sight the BCS state appt.'ars to have a higher energy than the
Fenni state; the comparison of (h) with (3) shows that the kinetic energy of the
BCS state is hirL .. than tlMI of the Fenni state. But the attractive potential
\
energy of the Bes state, although not represented in tl gure, acts to lower
the total energy of the BeS state witll rcspt.-'Ct to the Fenni state.
When the BCS ground state of a many-clcctron system is described in
terms of the occupancy of one-particle orbitaJs, those near EF arc filled some-
what like a Fermi-Di,?c distribution for some fi nite temperature.
'Ine central feature of the BeS state is that the one-particle orbitaJs are
occupied in p airs: if an orbital willi wavevL>ctor k and l>l lin up is occupied, then
the orbital with wavevector - k and spin down js also occupied. lfklt -i! vacant,
then - kit is also vacant. The pairs are ealiL'(] Cooper pairs, treated in Appen-
dix J-I. They have spin zero and have many attributes of basons.
'Nle boson condensation temperature (TP, Chapte r i) calculated for metal-
lic concentrations is of the order.of the Fermi temperature (J(yt - l dl K). 'nle
superco nducting transition temperature is much lower and takes place when
the electron pairs break up into two fen nions. "n le model of a supercond uctor
as composed of noninte racting bosons cannot be taken absolutely literally , for
there are about }(f electrons in the volume occupiL'(] by a single Cooper pair.

Flux Quantization in a Superconducliflg Ring

We prove that the total magnetic Ilux that pa55(''5 th rough a superconduct-
ing ri ng may assume only quanti:a.>d \'alu(.'S, int(."gral multiples of the Ilu.\ quan-
tum 2Trlrd q , where by experi ment q = 2e, the charge of an electron pair. Flux
quantization is a beautiful example of a long-range quantum df(."(:t in which the
coh erence of the supcrc;ond ucting state extends over a ring or solenoid.
Let us first consider the eiL'Ctromagnetic field as an example of a similar
boson fiel d. The electric field intensity ";(r) acts qualitatively as a probability
fiel d amplitude. When the total number of photons is large, the energy density
may be written as
E*(r)E( r)/47T a n(r)/tw ,
where n(r) is the number de nsity of photons offrcquency w. Then we may write
the electric field in a sem iclassical app roximation as

where 6(r) is the phase of the field . A similar probability amplitude describes
Cooper pairs.
"111e arguments that follow appl)' to a boson gas with a large number of
bosons in the same orbital. We then can trcat the boson probability amplitude
as a classical quantity, j ust as the electromagnetic field is used for photons. Both
amplitude and phase are then meaningful and observable. The arguments do
not apply to a metal in the normal state because an electron in the nonnal state
acts as a single unpaired fennion that cannot be treatL.od class ically.
We first show that a eharged boson gas obeys the London equation. Let
 (

" "
.J
l'igu~ IS (a) Probability P tlull an orbital cJl.:;nc.:tic e nergy If is ot'O.:I.Ipicd In t he ground st~te oft},e
" oninteracting t-"e nni gas; (b) the BCS ground state differs rrum lill,! f."enn; state in a "--gilIn of width
of the order 0( the energy gap Et 80th "'''Vel are ror abscl.ule urn.

.p(r) be the particle probability amplitude. We suppose that the pair concentra-
tion n = ",*!JI = constant. AI absolute zero II is one-half of the concentration or
electrons ill the conduction band, for II refers to pairs. '1l1cn we may write
(19)
111c phase 9{r) is important for what follows. (0 SI units, sd e e l in the
equations that follow.
"l'lIc velocity of a particle is, from the Il amilion equations of mechanics,

(CGS)

nlC particle nux is given by

(20)

so that the electric current density is

j = q4t'*"'IjI = - q)
n<l( ,.ve - -A (2 1)
no c
We may take the curl of both sidl..'S to obtain the london equation:
nq'
(CGS) curl j = - - - 8 , (22)
me
with use of the fact that the curl of the gradient of a scalar is identically zero.
The constant that multiplies B agrees with (14a). We recall tha t the r...leissner
effect is a consequence of the london equation, which ,",,'C ha\'e here derived.
Quanti7.ation of the magnetic !lux through a r ing is a dramatic consequence
of Eq. (21). Let.us take a closed path C through the interior of the supcrcon-
I' igure 16 "alh of integrnlion C t1uv"lth lhe inle rioc" of III
supcttv,ulul'ling ring. TIle flm th ruugh tI,e ring is tl, e SUm
of Ihe flux 4><., from e>;!.'n'" SOlIrtt'li and ti,e flux <1>... from
Ih.., s"p.'n:ondtl<..1 i n~ cu rrcnt. "hieh flow in the ."rf.-= vi"
Ihe ring; <1> - 4>..... 4>... TI,e flllt 4> is 'Juanliu'(l. There u
norm.llly no quantizalion rondit ion 011 the flux fm ln e"'cmal
"""1'1.'('5. 50 that <1>". 'IIllSllIdjlls l ilsl'1f "ppTl,priatcl)' in ordl'r
Ih.. t <1> UJurJlC a qllMl.i:te<1 , ·alue.

ducling material, well away from the surface (Fig. 16), The Meissner efft.'Ct tells
us that 8 and j arc zero in the interior. Now (21) is zero if
flcVO = qA (23)
We ronn

c
f
VO' dl :: ~ - 8,
for the change of phase on going once around the ring.
The probability amplitude 1/1 is measurable in the classical approximation,
so that .p must be single·."aluc<1 and
(24)
where s is an integer. By the Stokes theorem,

fA' dl =
c c
1
(curl A) . da ""
c
B· da :: $ 1 (25)

where da is an clement of area on a surface boulll.k-d by the curve C, and ¢ is

the magnetic flux through C. From (23), (24), and (25) we have 2r.f!cs ". q¢, or
(26)
Thus the nux through the ring is quantized in integra] multiples of 2rr1u:Jq.
By experiment (I = -2e as appropriate for e lectron pairs, so that the quan-
tum of flu x in a supe rconductor is

(CCS) (flo "" 27T1ic/2e ill 2.0678 x 10- 1 gauss cm2

(5 1) ~o = 2rr11/2ea 2.00'78 X IO':1:s tesla m"2

ThiS unit of nux is called a flWloid or fluxon .

'!lle flux through the ring is the sum of the Ilux ¢"'XI from exlernal soun::cS
and the nux 11>.., from the superconducting currents which now in the surface of
the ring: 11> "" ¢~U + 11>..,. The Ilux 11> is quantized. 'illere is normally no quanti-
zation condition on the nm from edemal sources, so that Ill,., must adjust itself
appropriately in order that III assume a quantized value.
Suppose a magnetic monopole of strength g is situat<.->d just below the
center of a supcrcomlucting ring. The magnetic flux through the ring is
(g/,-!ZX21U·2 ) = 2ng, and by (27) this Illust cqual an integral multiple of ,",Ide.
Thus the minimum permissible value of g in CCS is hd2£, the mmous Dirac
result.

Duration of Persisterlt Curren's

Consider a persistent curTellt that flows in a ring of a type I superconduc-
tor of wire of len~,'th L and cross-sectional area A. -111c persistcnt CUrl"Cut main-
tains a flux through the ring of some integral number offluxoids (27). A nuxoid
cannot lea k out of the ring and the reby rl,<lu(."C the persistent current unless by
a thermal fluctuation a minimum \"Olume of the superconducting riug is mo·
mentarily in the normal stale.
'111e probahility peT unit time thai a Iluxoitl will leak out is the pr<xluct
P = (attempt frcquency)(activ3tion 1);.lrricr factor) ("")
'1111..' activation barrier factor is exp{ - ~Flk/J'f), where the free energy of the
barrier is
I1F - (m inimum volume)(exCl-'Ss free energy density of normal slate) .
The minimum \'olume of the ring that must tum normal to allow a fluxoicl to
(..'Scapc is of the order of uf. where f is the coherence length of the super-
conductor and II the wi re thickncss. The excess free enerh'Y density of Ihe
nonnal state is H'!!81r, whence thc barrier frce energy is
(29)

Let the WiTC thickness be 10- 4 cm. thc coherence length = 10- 4 em, and
He = J(tI C; thcn I1F - 10- 7 erg. As wc approach thc transition tcmperature
from below, 6.f will d<.'Creasc toward zero, but thc value gi\'cn is a fair cstimatc
between absolute 7..cro and 0.8 Teo Thus the activation barrier factor is
exp( - 6.FlklJT) - exp( - Jff) ... 10 (43-1><10')

'11lecharacteristic frequency \o,.ith which the minimum "olume can attempt

to change its state must be of order of £J'I. If Eg = 10- 15 crg, the attempt
frcquency is - 1O- 1 ~1J0 -27 - LO I2 5 - 1. The leakage probability (28) becomes
P = 10' 2 10- 4 . 3011 ><10' S-I = 10- 434 >< 10' s- I

The Tl.'Cillrocal of this is a m(.>asure of thc lime required for a fluxoid to leak
out, T "" liP = ut 34>< 10' s.
The agc of the universe is only 10 1'i s, so that a Iluxoid will ncver lea).. Ollt
in the age of the universe, under our assumed conditionl ccordingly, the
current is maintaine<1.
'J11ere are two eircumstanccs in which the activation energy is much lower
ilnd a fluxoicl can be obscrved to leak out of a ring- either very close to the
critical temperature, where He is very smaU, or when the material of the ri ng is
a type II superconductor and already has fluxoids embedded in it. '111ese special
situations are discussed in the literature under the subject of lIuctuations in
superconductors.

Type II Superconductors
'111ere is no difierence in the mechanism of superconductivit y in type J and
type II sup(~rcond uctors . Both types ha\'e similar thelmal properties at the
superconductor-nonnal transition in zero magnetic field. nut the Meissner
eIT<."Ct is entirely different (Fig. 5).
A good type 1 superconductor excludes a magnetic field u ntil supercond uc-
tivity is destmy<.>d suddenly, and then the field penetrates completely. A good
type II superconductor excludes the field completely up to a field Hc1 . Abo"e
H ~ l the field is partially excluded, but the specimen remains c1<.'Ctr1(:ally supcr-
conducting. At a much higher field , H c2 , the flux penetra tes rompletely and
superconductivity vanishes. (An outer surface layer of the spt.'Cimen may r<."-
main supercondueting up to a still higher field H~3.)
An important difference in a type 1 and a type II superconductor is in the
mean free palh of the cond uction ek'Clrons in the normal state. If the coherence
length ~ is longer than the penetration depth A, the superconductor will be
type I. Most pure metals are type I, with K < I (see Table 5 on p . 35:3).
But , when the mean frt:c path is short, the coherence length is short and
the penetration depth is great (Fig. 14). 'C11is is the situation when K = Alf> I ,
and the supercondudor will be type II .
We can change somc metals from type 1 to type II by a mod cst addition of
an alloying clement. In Figure 5 the addition of 2 wt. (lCrt-cnt of indium
changes k'1ld from type 1 to type II , although the transition tcmper.lture is
scarcely chang(.>d at all. Nothing fundamental has bCt."ll done to the e1ectmnic
stn-ctUTe of lead by this amount of alloying. but the magnclie behavior as a
supe rconductor has changed drastically.
The theory of type II SupcT<:onductors was de\"e1opt:d by Cinzburg,
Landau, Ahrikoso\', and Gorko\,. Later Kunzler and co,workers obsCIVed that
Nb:JSn wiTL'S can carry large supercurrents in fie lds approachiug ]00 kC; this
kd to the commercial de\dopment of strong field sllpcroonducting magnets.
Consider the interface hctween a region in the superconducting stllte and
a region in the normal state. 'I11e interface has a surfacc energy that may be
pOSitive or negative and that d(."Creascs as the applk>d magnetic field is in-
 12 S"~r('Ooufuctit'ily 361

Ill! 111·· '·1111111111111··

I
7.
I
7.

'" (bJ
Filure l7 (II) Magl'll.:tic field pmctrntion into a thin film of lhiI."'JIal equal to the pctlCtration
tleptll A. "Ole arrows indiCRte the intCll$ily of the rnagntlic flCld. (h) Magnetic flCld.penetMllion in
a homogeneOlls bulk strocture in Ihe mixed or vorlex slate. with ahemale laye" in nor mal and
supen.'Qrldooing st,ltt."$. 'me SUllerronduding byers are thill in C(lfnpari.otl with A. "Ole laminar
.trocIure is ~hO'wn fur ro.wcnicnce; the .auaI structure l'OIlliisls of rods of the 001'11131 state "" r_
rounded by t he su perron,Iucting slate. (TIre 11' regions in the vortex slate are not exactly normal,
but are dClCribetl by low ~'3lues of the stabil ization energy ,\crull)'.)

creased. A superconductor is t)1lC I if tile surface energy is ahvays posilh'e as

the magnetic field is increased, and type II if the surface energy lx .'c omes
n(.>gative as the magnetic field is increased . 'fhe sign of the surface energy has
no importance lOr the transition temperature.
'111c free e nergy of a bulk supcrconductor is incre:tscd wh cn the magnetic
field is expelled. lIov.'C\·er, a parnUel field can pcnetratc a vcry thin film nearly
uniformly (Fig. 17), only a part of the lIux is expe lled , and the energy of thc
supcrronducling film will increase only slowly as the external magnetic field is
increased . This causes a large incn.':lSC in the field intensity required for the
destruction of superconductivity. 111e (jIm has the usual cnergy gap and will be
rcsistanccless. A thin film is not a type II superconductor, but the fil m res ults
show that under suitable conditions superconductivity can exist in high mag-
netic fields.
Vortex Stale. The results for thin mms suggest the qu(."Stion: Arc there
stable configurations of a superconductor in a magnetic field with regions (in
the form of thin rods or plates) in the nonnal state, each normal region sur-
rounded hy a superconducting rt.->gion? In such 11 mix(.>rl state, called the vortex
state, the external magnclic field will penetrate the thin normal regions uni·
fonnly, and the field will also penetrate somewhat into the surrounding super-
conduding material, as in Fig. lB.
The term vortex state descri bes the circulation or supcrronducting cur·
rents in \'orticcs throughout the bulk sp(.'Cimen, as in Fig. 19 below. There is no
chemical or crystallographic dificrencc between the normal and the supercon-
ducling regions ill the vortex statc. The vortex state is stahle when the pcnetra-
 ,~

'.
,-

~"
'" - ,
,

,.,
1') .... II ",,,,'K<IOod,ll1n,.

""
-, Fi~re 18 Varialion oflbe magnctlcfid(l.nd energy
g21l panlrnctcr 6(:\") at the intcrf~ of supcr<.'()nduct-
in!; .mll normal regioo$, for type I nnd t )'J)C II super-
eonduc:fon. The energy gap panuneter is a mcll.'iurc
, of the stabilization energy densit)' of the $UPCfUIn-
o dueting ~tatc.

tion of the appli<.>d field inlo the supcrconducling material causes the surface
energy to berome negative. A type II superconductor is characterizL.u by a
vortex state stable ovcr a certain range of magne tic field strength: namely,
between lf~1 and H d .
E3limation oj'H", and lid ' \ \llIa! is the condition fur the onset oflhe , 'ortc,;
state as the applied magne tic field is increased? We estimate lid from the
penetration depth A. 'I1lC field in the normal core of the fluxoid will be lid
when the applil.>d field is Hel .
The field will extend out from the nonnal core a distance A into the super-
conducting environment. 'J11C nux thus associated with a singlcoore is rrA211cl.
and this must be equal to the flux quantum ¢o dellm.-'f1 by (27). ' J1lUS
He! ... <Pc/rrA2 . (30)
"Ibis is the field for nucleation of a single fluxoid .
\
Figun.' 19 Flux bUice in NbSe. at 1,000 glIlIH at O.2K, 115 "iev.w "~Ih a K'lUIning Inflllelins
'nicro<cope. l11e photo 5bl1"'ll the <kniity of ~tllln aI the '''crm! Io-'el, as in "ig.,Y(' 23. The ....,rt"",
coreli ha,'c It hlltl' density of 51"tes alld are sllNJetl while; the supc rcontlUding ...woo. are .lali., w ,th
110 st"l"s 3t the Fermi le...,1. 'Ille amplitude "nd SP.1tit~ e~t cnt of th ese .tatc~ Is .Jctcrmin cd by a
poten tial wdllimned by 6(:r) as in Figore 18 for a Tn'" II sopercoodUdor. 11.e potential weU
ronfines the core stale w,,",·e(unctions in the image ....""'. The s.lar shape is ... fI,}('r r"..lure, "reli,,]t
spoe;a\ to "bSc,; of the sixfold disturbance of the doa'19' tlensi!)" at tbe J'ermi 5urfol(.'('. Photo
wurtesy of II . F. lI ess, 1\1'&1' BelJ Labol1l1orie!1.

At lid the fluxoids are pack(.'(\ together as tightly as possible. l'OllSistent

with the preservation of the supcn::onducting state. This means as densely as
the coherence lellbrth {will allow. 111e external field pencll':1tcs the sl)(.'Cilllen
almost unifonnly. with small rippk'S on the scale of the Iluxoid lattice, Each
COI"C is responsible for carrying a (lux of the order of 1f{2 11<'2. which also is

quantiu.>d to ~ 11lUS
(3 1)
gi"es the upper critical field. The larger the r.ltio )"f{. the larger is the ratio of
11<:2 to H~,.
It rmnains to find a relation betwt"Cn these critical fields and the the rmody<
namic ctitical field He that measurt.'S the stabilization e ncrgy densit)' of the
superconducting state, which is known by (9) to be H';!8r.. In a type II super-
conductor we can <letennine He only indirectly by calori metric measurement of
the stabilization energy. To estimate He. in terms of H.:, ( _'Onsider the stabil-
ity oftlie vortex state at absolutezcro in the impure limit ~ < A; here I( > 1 and
the coherence length is short in comparison with the penetration depth.
We estimate in the vortex state the stabilization energy of a fluxoid core
viewed as a normal metal cylinder which carries an avcrage magnetic field Btl.
'I11C radius is of the order of tile coherence length. the thickness of tile bou nd-
at')' between Nand S phases. The energy of the normal core referred to the
energy of a pure superconductor Is given by the proclud of the stabilization
<'''11efl:,'Y times the area of the core:

(ees) , _ - 1 Ill- X 7T/:2 (32)

Jrore 87T c ."

per unit length. But there is also a decreasc in magnetic energ)' lx."C3use of the
penetration of the appli<.>cl field B" into the su percond ucting material arou nd
the core:

(ees) (33)

For a single fluxoid we add these hw contributions to obtain

(ees) (34)
'J11e core is stable if1< O. The threshold field for a stable fluxoid is atl = 0 , or,
with He. written for B..,
HellHe "" {fA (35)
-111e threshold field divides the region of positi\'c surface cnerg)' from the
rcgion of negative surface energy.
We can combine (30) and (35) to obtain a relatioll for He:
7T~AHe "'" <1>0 . (36)
We can combine (30), (3 1), and (35) to obtain
(37a)

(37b)

Single Particle Tumwling

Conside r Iwo metals separa ted by an insulator, as in Fig. 20. 'f11e insulator
normally acts as a barrier to the flow of conduction electrons from one metal to
the other. If the barrier is sufficiently thin (less than 10 or 20 A) there is a
Significant probability that an ck'Ctron which im pinges on the barrier will pass
from one metal to the other: this is called tunneling. In many experiment s the
 (

Figu~ 20 Two mctals, A and B. K'paraloo by • thin layer ~ an insulator C.

"!gllre 21 PrcparatiQn r:i 1\11 AVAl~S., sandwich. (I) C\a.s!; slide with indium contacts. (hl An
aluminum ~rip I mm ",ide 111<\1000 1o:tOOO A thid.: has been ~itC(1 across the contacts. (c)1be
aluminum s.trlp has heen v~idi_1 11I form an 1.12 0 3 bya- 10 to 2(l A In th,,,knc:ss. {<II A tin m,n
h""
becn dcposih.,<IICTOSS the aluminum ram , formIng an AVAI~O:JSn slInt\wid,. 111e c~len,a1 lcacb ,'re
connectoo 10 the indium l~n t"'h ; two conh,ds are IISC!' for the cum-I'll mca.m rcment amI twu 10.-
the \-dt:Jge mCl.5U""llCIll . (Aflcr Gioc'\"1:r and Mcgcrle.)

".)"
\'ul~.

ng,,", 22 {a} Linear cum·nl ...."ltdgc relation ro. junction of f\Of',,~...1 metals ~pnr.. tcd b) QIIide
larer; (hl cunenl-mltagc relation "ilh one metal oormal and the other metal $uJ"It'n'ond"ctillg.

insulating layer is simpl)' a thin oxide layer formed on one of two evaporated
ITlCtal films , as in Fig. 21.
When both metals are normal concluctors, the current-\'ohage relation of
the sandwich Of tunncling junction is ohmic at low voltagcs, with the curfent
directly proportional to the applicd voltage. Giacvcr (1960) diS(.'O\'cn.'d that if
one of the metals becomes supcrconducting the current-voltage characteristic
changes from thc straight line of Fig. 22a to the curve shown in Fig. 22b.
Fib'Ure 2.3a contrasts the ck"Ctron density of orbitals in the superconductor
with that in the normal metal. 1n the superconduclor there is an energy gap
eenten..od at the Fermi le\'c1. At absolute zero no current can Oow until the
applil.>d voltage is V =- f~J2e = Me.
 I ', .. ,. ",

.',..-,,,,
2lo, - - - - - - - """''IlY

,. N

"'I
Figul"(' 23 The Ill""'''' of "rhit.ds and the 111rr"nl-volt:lJ:." d~"'al.11'mlie /Ora tunnclinlti,,"dion. In
(iI) the "ne'J.!Y,5 plott .....! "" th.., ....,rlieal M:~lc ~Ild th.., d"'o,ilyof orhit.tls ",.the horiwutal <I.;/.Ie. 0"..,
1I..,!allS III Ih" nom",l.!.II.., aTld.,.", in Ih" 'UI'lCn..'OII<!ud '''1t ~Ialc (b) I 'enu, I', dl<'d;)~hcs iud'mt..,
the ex,><-'eIL..! bn",k al T '" O. (An". Gwe"cr and ~1L,.".1c. J

11le gap E~ ..-orresponds to Ihe hreak-up!lr a pair or clct-lrons in the super-

c()lIllucting state, with the rormaliolt of two cicci roilS, o r an cJt.'Cl rou ;Illd a hole,
in the normal s ta te. 1'lc current shu'ls when eV"" fl.. At finite temperatures
Ihere is a small curren t flow cven .11 low voltages, i.lcctuse o r clc.!ctrons ill the
supen:onductor Ih.1t arc thermally excilcll 3CI"OSS the energy gap.

j O$C/)/, wn Superconductor TUllllelilig

Under suitable conditions we flh~ervc rcmarkahle crfec:ts a.~sociall....! wilh
the tunneling of supen:onducti ng ek'Ctron pain rl'OUi a supen:onductor through
it layer o r a n insulalor into another supcrl'Oliductor. Such ajunclioll is (:allnl a

weak link. The clfl.'Cts of pai r tunneling include:

Dc Josephson effect. A dc <'1Jrrellt flows across the junction in the abscill-c
uf allY clt'dric: o r magne lic field.
Ac Josephson e ffect. A dc voltage applil.-u across the junction causes If
current oscill.ltiolls across the junction. This clfcct ill.ls beC'n ulilized in a preci-
sioll determination of the vdlue of file. Further, ,I\) If voltage applied with the
de voltage can then caus,e a dc eUITent <lCfUSS the jUllctiun .
Macroscopic long-range quantum illterferel"lre. A de magnetic: field
<lpplied t hrough a supen.-ondueting drl'uit l'Olltaitling Iwo junct io ns C\USl.':S thc
maximum supercurrcnt to show intelferencc cffc(:ts as a rUnction of magnetic:
field intensity. 11lis effect (:an be utiliZl.'(1 in sensiti.... c magnetometers.

Dc JO$C,)I~-oll Effect. Our disc...'u~sioll o r Joscphsml junction phf..'lIolllClia

follows the discussion of flux tjuan tjzation . Lei 1/11 be the probability amplitude
of electron pairs Oil one sidc of ajllllctioll, aJw
k-t I/Jt be tIle amplilude Oil the
other s ide. For simplK::ity, lei both SlIllCn.:onductors be itlc nli<.-al. For the pres-
cnt \\e suppose that they arc both a t zero potl.'11tial.
"l11C lime-delx::nc.!t..·nt SdtriXlinger l.'fIIJdtion i"iI~;!l ::::Hif! applied to the
1\\'0 amplitudes gives

.
.~-.~----- -
 (38)

Ilere /,.,. rcprcs~n t s the dft!~t of the cleclron-p.lir t.'tl ullfing or transfer i lll crJ.l~
lion UlTUSS the insui.tlllr; T Ius the (limcnsiolls of a ru le o r frcqucnq'. It is a
mc~urc of the leakage of 4/1, into the region 2, .lTld uf ~2 into the r(.'gion 1. Irthe
ill~ ulal o r is vc r y thick, T is /'(''1'0 and tllcre is 110 IMir tunnding.
Let VII :: III'l e'60 ;m(\ 1/12 "" 1I~'2 eHJ,. TIICn

'.p,
--=
I 11 -\ I ll
c .0. - ".r, ao,
nil,- + ,., --""- ''T'""'2 (39)
at ilt ilt

- = I 112- 112 erD. -

iI!J!l
- z + ,.,
all- 302
' ,I, --::::; - ,'I '.....,., (40)
ilt at ill

We multiply (39) Ly ,11 '2 c- · fh 10 olJt.lin, with 8 III 8..! - 0"

lilli,
---+ ao,
ill, --:::: - i'f(1I I 11,0 , e'
12 ~
. (41)
2 ill ilt
We multiply (40) by 1I~12 e- III, to obtai n

-I -ihl-Z + iIl 2 -a(k- "" .

- ;1 (111110 112 e ,/I (. 2)
2 at at
Now C(lualc the real and imaginary l>arts of (<I I) and silll illlrly of (12):

(.13)

(·11)

If II , 03 1/2 as fm idcntk"J.\ supcnxmdllclol"'S 1 and 2, we haH: from (-101) tha t

-ilO-I - -ilO!- (45)

ilt I1t
From (43) wc scc tha t
iln2 ilII I
--~
at
---
at (46)

11lC l'Ufrent now from (1) to (2) is proportinll.alto aI/Jilt Of, the S.IIlIe thin g,
- alillilt . \Vc tl w rcforc l'Ondudc fWln (43) that the l1.JlTcnt 1 of superl"oll(Juctol'
1);.lirs al1.OSS the junction depend" nn the phase (lirrerclIl"C 8 .IS

lZ: : 1n ..ill li :: 1n sin (01 - O.l ,:] (.17)

whe re 10 is proportio nal to the transrer inlcntclioll T. 'Ibc l1.Jrrc nt 10 is the

 Cu,..,,'

' 1--,,(
,,,
,
,,,
,
--------jl'----;:----- \"uI.~.
,
,, ,
V.

,, 1".,,", 24 Cu~" l-\'oltage characteristic of a

/ Josephson jundiou. De rofTCnlS f\ow under zo:ro
applietl voltage up 10 aerilical <.'IlfT<,n' 1<: this is lhe
tic Josephson effect. At voltages abo\'0: V. the junc-
tion has a finite I"<lSi~'allCC, but the CUlT<:nt has an
mcillatory component off""lu{'J1C)' '" - \!.eVIl>: thili
is the ac JlIW:'(lI)!;o.. efliect_

maximum zero-voltage (.'urrt..'nt thai can be p.'lSSt..'t1 by the junction. \Vith nu

applk.o voltage a dc cU'Tent will flow across the junction (Fig. 24), with a value
betweeni!) and - io according to thc value oftltc pll<L~e diflC,·cn(.'C O2 - 01 , lbis
is the dc Josephson effcct.
AcjoselJI.son EfJccf. Let a de voltage V be applied across the junct ion. We
can do this be(.-ausc the junction is an imlJlator. An elt..'dron pair experiences a
potential energy difference qV 0'11 1)'1.ssing across the junetK>n. when: q '" - 2e.
\ Ve ('':I n say that a pair on one side is at potenti'!'l energy - cVand a pilir on the
other side is at cV. The equation.~ of motion that replac.:c (38) arc
(48)
We pnx:eed as above tu find in place of (41) the equation

-I -.1111- + il1 , -.10,- = icVntf~ - ' - iT(lllnti l rI

n. 15
(49)
2 at iJl
This L'(lu:ltinll breaks up into the real pat·t
illl,lat:: 2T(n lnti"~ sin 8 (50)
exactly as withuu t the voltage V, and the imaginary part
aO,l ill = (INlh) - T(ntlrlt )tll (.'()S 8 (5 1)
wllieh d iffers from (4'1) by the term eVIIt.
Further, hy cxtcnsioll of (42),

I~
-I -(J1l2 il02
- + iI12-- = - i CVI12f1 - 1 - iT('llll2.l
1/2
c- (52)
2 ill ill
 (

tigure is The arrangement fOl" eaperimelll on mac-

rO$ropic quantu m intcrfcrence. A magne~ flu" <l>
pa55CS through the interior of the loop.

whcm:c
itll.jat "" -2'r(n \n0 1t2 sin /j ; (53)
iI(J,j11t ::. - (eVil.) - T(II ," 12)112 '-'OS {j . (54)
From (5 1) and (54) wilh I'll ;;;.: 112. we have
cl(02 - o, yat = aOlat ::: - 2eVlh (55)
We scc by integration of (55) that with a tit· voltage across the junction the
relative phase of the probnhi!ity amplituut.'s V;lr)' a.~
0(,) - 0(0) - ('leVUI,) _ (56)
11lC s upcn.:o nuudillg l:urrcnl is given by (47) wilh (56) for the phase:

1 i - i , 'in (0(0) - ('leVtIt.) ) -I (57)

Tnc current oscillates wilh frCfjllcncy

w = 2eVlIJ (58)
111is is the l\C J osephson cO"ect. A de voltage of IILV produu :s a fre qllcl1(.,), of
483.6 ~f liz. The relation (58) says 111(11 a photon of cnc rg:. fzw ::: 2cV is emittet!
or absorbed when an cl l.."C lroll pnir '-1'OSSCS the barrier. B~ Illcasuling Ihe \olt-
age and the frequelll')' it is possible to obtai.. a very precise valuell of cll!.
M{lcrVSCO/)ic Quwl/UIrI Illterference. We saw in (24) Md (26) that the phase
differe nce Oz - 0 1 al"Ound a ciOSt."<1 dn.:uit which ellC;:OlnI"KISSCS a total magnetic
Il ux ¢ is given b y
~ - 01 = (2c1f1c)¢ . (5'J)
11lc flul( is li lt' sUln of Il ial due 10 exlcmal flelli s and that due to cuncnls in the
circuit itself.
We <.:ousider two Josephsoll jUIIl1ions in p.'II':dlcl. as ifl Fig. 25. No voltage
is applil.,j. Let the phase diffcn:1l<.."c LetwcclI pOinls I and 2 takell on a p.1.th
tluOllgh junc tion (I Le f>". \"hen taken on a path through j UJ"lCltoll b. the phase
 310

Figure 26 J-:~pcri"lC:nta1lr.w:e of J~ .. '~n"~ "'~fo:Jl('tic ".,10.1 showing illl('rf('rcn~'e and d,lfrnortion

dft"ts ror h.O ;,,,,chon. A B",1 B. TI", field rcrioll«-ilv " 39.5and 16 ",C lOr A anti B, n .... IJoc....·h ...,I)·.
AI.pnni",ate '":t~imum ~.... r...,nl\ are I ",A (lI)and 0.5 ",A fBI. 11M: jo"" ' tioll scpar.tion" 3 ,"In and
iurll1ion "ithh 0.5 n'UI fur hoth ca54:~. TI.e "'m ofISel 01"11 is tl"c \0 a Il.lckl9"ouml magnetic flcld.
lACier R. C. J d~!t:.ic, J. Lambe, j . E Mcm,rcau and A. II !li!- .., .)

dificrclll"c is li". In the absence of a magnetic field these two phases must be
equal.
Now lei the flux $ pass through the il.lcrior of the cil'C:uil. \\le do tllis willI
a ~Iraighl solenoid lIorll1al lo the plane of tile I,};IPC I" olnd lying ilisidc the t'in:ui! .
By (59). 5j, - lin "" (~tflc)q,. or
c
• ~.ln - -<I> (00)
" /'c '
TIle total current is the sum of}" OIndl b_Thccurrcnt through each junction
is of the fonn (47), so thilt

hob.! = ) o{sin (50+ :c tl» + sin (~ - :c ¢)} = 2(Josin 80) (:OS :

l11C Clirrent \la rit..'S with ¢ .Ilid h.1S maxima when

c<l>l flc = S1T , $ = integer (61)

'n.c periodi(;it}' of the current is shown ill Fi~. 26. The ~hort l)Criod varia-
tion is (lrtK.\lICl..o by inlcnere llcc from the two junctions, as pre dicted hy (61).
"I11C longer period variation is II diffrllCtion effl.'Ct alld ariscs (mill the finite
dimc nsiol"ls of cach junction - tl.is causes (I> to dCI)clld on the p<lrtil.-uhu· path of
iull.-gralion (Pmblem 6). The di/Ji-aclion eflCct w.tS not antidpJ.ted in carly work
Ih.ll rcsoh ed only single elcctroll tunncling.

'0
 12 S"po:~colld"cti";ly

HlGII-U:MPt:nATUHE SUPEHCoNDucrons

High Tc or HTS, dennles sllpen:onduclhil}' in materials, chiefly oxides,

with high transition temperatures, a(;(;ompaniro by high crilic..-al currenls lind
magnetic fields. By 1988 Ihc long-standing 23 K ceiling of Tc in inkrmetallic
(;ompounds had been elevah.:t1 to 125 K in bulk supercllndllcltng o\ides; these
pa5SIX.I the st,lIltlanl tt..'sts for sur~:n:ol\(lm;th ily- the ~1ciss!l<; r eITect, ac Jo-
sephson ciTed, persistent l'lUTents of 10llg t1uratioll, and substantially zero de
resistivity. D)' 1994 HTS showed promise in pre-<;ommercial applkations, as in
thm 6Im devicl..'S, and wires were wing fahricated. Levitated vehicles allil long
distance power transmission have not arrived.
Memorable steps in Ille advance indutle:
BaPbo 75Bio 2503 Tc = 12 K lBPllO]
L.1.. b5Bao 15CU04 Tc = 36 K lLBCO]
YDa2Cu307 Tc = 90 K lYBCO]
1·lzBazCa zCu:)0 .0 Tc= 120K ]TIlCO ]
A relatc<1 lIlen;UI)' l"Ompolllld may have T" = 140 K. 111C material wilh most
rescart:h behintl it is YBCO.
111e CI),slal stmdllres of Ihese materials, except for BrBO, arc oxygen-
deft..>(.'t modifications of the perovskite structure of Figure 13. 10, with about
onl..~thin:l of the oxygell positions vacanl. The primitive cell is dl..'Vclopeti from
that of a tetragonal \x:rovskitc tripled along the c a.\js. Per formula unit of
YDlI:2CU307, the IXJsitive ion valencies based on y+:\ Da+2, Cu+Z arc 3 x 1 +
2 x 2 + 2 X 3 = 13 lUld the negative (oxygen) valency is - 2 x 7 = -14. The
qllasitetragonal or OIthorhOlllbic cell dimensions of YlICO are, in nm, a =
0.388, b = 0.38-', c := 1.163, with cell volume 0.173 nm3; if there is one excess
charge (;arrier in this volume the can-ier concentration in 110 = 5.77 X 1O~!.l
cm - 3. '11\C excess carriers typically arc holes embedded in doped ulltiferromHg-
netic insulators.
Parallel shl.."ets ofCIlOz arc a structural feHlure of all HTS with Tc > 40 K;
these are known as the cuprate I-ITS. In rBCO sllOwn in Figure 27 the CuOz
planes arc parallel to the plane of tbe ab :l"(es in tile orthorhombic primitive
cell. Two of these planes, those through the interior of the lOCI!, arc the domi-
nant l..'ollducting pathways. 'l1lC oxygell ltmlent per cell can be changed rcvers-
ihly from 7.0 to 6.0 atoms, simply by pumping oxygen in aud Olll of the pJ.r.lllcl
cimins of CuO that nm along the b axis of Ihe figure. AI the l'Olllposition
YBa2Cu306 the crystal is all insulator, with antiferromagnetie ortler of the
Cu spins. An im.'I"ease in ox}'gen ahove 0 6 ~ makes the crystalmetJ.lIic and 110n-
magnctk; tile crystal is supermnducting al)()\ e Ob 1).1 . A dmrgC-lleutral formula
for YBa ZCu:)01 can be written ll.'i YBad C u2-t-MCu3+) (0 2- ), or as
YBaiCuz+h(Oz- M O - ).
 (
r

(?~hV03
(2 ) 0 Ba I ().I

? 1w'~)~C"I_"
~~Ol
b
/ ' &
Figure 27 ll.e tt}"J1aI structure of )13a~,~07. sl'l)\<.i.", a primiti,'C COlU. 1bc planO's containing
.it"" marked Cu2. 02, ."d 03 form tI .., conducting pIIthways.

III the llBaCnCu (.'Ompoumls, Tr; increases with tile number of slll.'Cts of
CU02 stach.'li l1:mliIlUOusl), in a crystal cell: thus for the sclics
(CuOt.lDaOiJ10 2• (CuOt.l2CaBa021102, (Cu02hCa2Ba202T1202 tiLe arcT.:5
82, 109, 122 K. lsoslrudural (.'(Jlupoullds where Oi I'cplaccs Tl ;lnd Sr I'cpbccs
8.1. have Tc'" 10,90, 110 K for the one, 1\.\-00, and three la)'cr compounds.
All tht:sc layer l'Ompounds arc strungly anisotropic in clct:trical propcrti(:s
mC:L~ufL·.:1 parallel or pcrpcllllicular to the layers. It is unrealistic in ;\II)' approx-
imation to n(."gk'd the anislrop)' of the normal or supcn:onducling propcrtit..'s.

Critical Fieltls Utlrl Cri,ict,l Cu,.,.rnts

Critit'3l field and cUlTcnl vaillt.'5 in I-ITS may be qllite Ulll'Crlaill, be(:ausc
in anisolropic m(.'(lia thc measurcments arc dimetl lt and perhaps subject to
largc reevaluation for giant flux en.:ep lime effe<.:ts. We give rea.~unabl y ae-
(.'Cplcd vaitJ(.'S of the critical }Xlmmcters for ('-<Jmparison wit h conventional mate-
rials.
High T~ sllgg(.'Sts a high stabiliZOltion cncrgy and hi~h t..'11Cl'gy gap flW thc
sllper(.unducting statc, from whi(:h follow high Hi; (from Eq. (9» and short
(:ohererK'C k'11gths ~(from Eq. (17» . 'I:'lkell ill (.umhinatioll with a normal peoc-
tration depth from Eq. (14;\), thcse results lead to extreme type II beh."Ivior;
high I( ill Eq. (37b) and very high lId in Eqs. (3 1) a1ld (1.25). At the high
tcmpcraturt..'5 1«.'(.'l.'Ssiblc with IITS materials, thermall)' activatcd crCt..1) of
fl\lxons mlly limit L Jble cllnent vruucs; thus the achievcmcnt of high T" dOt:s
not a~sllrc that a giVl.'11 material will hlwc useful high CUITl.'tlt applications.
III the low tempcrature limit the 1..olldoll penctmtion depths are A." .... 140
nm and A., "" 700 nm, ftum magnetization measurements; the (;ohercncc length
f..b "'" 1.5 om from (;ritk."lll field measurl.'mcnts. 111C Giozhurg-Land'lu paramc-
ter Kat. = A.J/€.t, .... 100, which amply s,1.tisfics the eritel-]on K> 1 for a type II
supercolllluctor. TI!C(;oherence length normal to the CuOz planes is f., .... 0.2 -
0.6 nm, of the order of the interplanar spacing itself. 111 thin films the critical
rurrent density at 4 K is}oh = 1.2 X 101 Afcm 2 in the ub plane, and)" "" 4.2 x
lW Afcm2 in the C din.dion, suggestive of a quasi-2D sllperconductor with
plancs (."Onnccted by Josephson hlllncling. Critkal curren ts ill ceramic powders
are mudl smaller, typi<.-ally 10"' Afcm 2. From }.b wc may infcr Ht;2(uiJ) ~ 107
G := 1000 T; values up to 100 T h.1.YC been measured in YOCO.
In narrow bridges (SO nm) mudl higher critkal cUlTent dcnsitil."s have
bl.'Cn observed, being limited Ly thc depairillg of Cooper pairs. A hrillge is
narrow if no flwmns form within the volume of the Lridge, and thc CL para me-
(l.'f 4/1 may be taken as (."'(Jnstant within the Lridgc. TIlen (1.10) 1JC(."'(Jmes
(62)

where thc last tcnn is the kinetic cnergy of the Cooper pairs, each of mll.'1S ffl.

The (."Urrent oensity is

(63)

whkh is a ma.ximum with rcspc(:t to/= 1!JIl2~4/Io12 when I" = \1(213). Here !/to is
the GL parameter a.t zero (:UlTcnt. Thus the maximulIl (o(.-pairing) eUITcnt den-
sity is

the eritictl dcpliring velocity v" - Klmf For A - 100 11m and { .... 1 nm, we
have ie, - 1.3 x 109 Afun 2 in good agreemcnt with experiment.

Hull Number
TIle Iiall number is dt:finoo a.~ the inverse Jlall constrult Iln/le nonnaliZl.-.d
to the fonnula volume. If only holes or cledrons condud:, as in simple metals,
thc Hallnumbcr gives thc carrier l1J11l'Cllh-ution PCI' f(JI'mula volume. In yeCa
the Iiall number is p-Iype and prop0l1ionaito temperature for Hlle, for 'f some-
what above Tc> and is ortllt' ordcr of unity at 300 K, giving a Ulmer (''Olleentra-
tion '" 7 x HI' cm - 3 .

Fuller-ene,
TI1(.'se arc stable, (."fIgelike molecules that (."'(Jnstitutc the third fonn of pure
Ulrbon; the other two IIrc diamond and graphite. The archetype fullerene is
,,.

figure 28 G.o fullerene molca,ICII tT))otalliu in a f:>ce..eo:nlcru.l cubic ~tnlCtur~. CourtCII)' tI

Ste,-en Louie.

coo; cach mok'CUlc has the fonn of a trull(;atcd icosahedron with 20 hc,.;agonal
f.II..'Cs and 12 pentagonal r.1.<.....-s. like a socccrb.1.II. 111is is one of the 13 Arcllim(....
dean solids dlaraclcriz(."tl by having all their angles equal ami all their f.1.(:cs
regular polygons. Coo cryslalliZl."S in a facc-cclltcrl."(1 cubic structure, as in Fig-
ure 28. Alkali-fullcrene (:ompounds such a.~ ~C60 arc slIpcn.'o mlucting; tllis
onc has T c = 19.2 K. The K atoms nt'cup>, the octah"tlral silL'S ill the (;ubk cell.
HhCs 2 COO has T", = 33 K.

SUMMAl\V
(In CCS Units)

.. A superconductor cl\hibits infinite COIKlulth ity .

• A bulk spt..'d mCll of metal in the supcrt'Omlucting state exhibits perfect dia-
magllctism. with the magnetic induction 8 = O. 111i5 is the l\lcisslier effee! _
The exterlla\magnetic field will pt:nclrate Ihc Smi1(;C of the spt..·t:imcn O\-cr a
distan<."C dctcnnint...u by the pCllctratiuo depth A_
• There are two types of s uper<.'Onductors, I ,u,d II _In a bulk s pt."(.imcn of type I
SUllCl-<.'Ondue!or the supcn:onducting state is dt.'Stro}'<''(\ and the 1I000mai state
is restored by applicali0l1 of an external magnetic field in cx<."Css of a critk'lll
value He_ A I)'IX! II supcn.'Omlu<.1o r has two <'Tilk-al fiel lIe! < lIe < H,;2; a
 J2 S"pucondw:,;dt!l 'l~5

vortex state exists in the rangc l>ctWC(''(1 lIel {Uld lid' TIle stabiliz.1tion cncrgy
dcnsity of thc illlre sllperconduding statc is J1~R1r' in both type I and II
sUI>Cl'COndlldors.
• In the superconduding state an energy gap, E~ "" 4kB1 ~ sep..1rat(."S supcl'(;on-
ducting ek'Ctl'Ons below from norlllal doctl'UnS above the gap. Thc gap is
detccted in experiments on heat l'3I};.l(:ity, infrared absorption , and tunnclillg.
• 'nlC Lomlon equat ion
. c . c
, =---A
41r'A~
0' curlJ = - - - B
41r'A2 ..
lealls to thc ~Ieissller clfed through thc penetration equation V 2B = BIAi.
whefc AI.. .... (mCI41r'nC)' 12 is the London 1)(''1lctraUolI lkllth.
• l1uce imlXlrtant lclIgtlL~ enter the tht!ory of sllpcrcondudivity: the LonOnn
pcnetrlltion d(.1lth At,.; the intrinsic coherenl'C length ~ and the normal clCt..'-
tron mc.1n free path e.
• In the l..omloll equation A or B should be a weighted avcrage ove]' the (;ollcr-
ellcc length ~. 1'he intrinsic coherenl'C le ngth 6J = 2/iv,J1r'Eg.
• '111t~
BCS tht:ory aa:ou nts for a sUI>cn:onducting state fonned frulll pairs of
electrons k t and - k t. 'nICSC pairs ali as i.losons.
• Type II supefCOndudors h..·we ~ < A. Thc lTitkill ficlds are rclatctl b)' Hel ....
(§lA)H" and lId "" (AJ{)HC' 111e Cinzburg-L.1Illlau parameter K is defin(.'tl as
Alf, Values of lld arc as high as 5(X) kC = 50 T.

Problems
L Magl1dic field pctJCir(ltiol1 in (I plate, loc pcnctration IXjuation may be written il$
A2V2n ... n, whcre A i5 tllC penctmtion depth. (a) Show that n(r) in5ide a supercol\-
dueliog plate pcrpt.'l'ldicular to the r axis nod of thieko(''$$ [; is giVClI by

B(x) "" 8 cosh (VA)

• cmh (mA)
where n. is thc field outsidc the plate a'ld parallel to it; here x == 0 is:lt the ccnter of
thc plate. (b) The effectjvc magncti:t.atioo M(x) in the plate i~ defined by B(f) - B• ...
4'11'I\I(x). show that. in CCS, 4'17M(r) _ -B.,( II8Atx~ - ~. for [; 4 A. In Sl we
(('place the 4'17 by ~.

2. Critical field of tliin fiim&. (a) Using the result of Problem lb, show that the frt'C
energy demity at T ,., 0 K within a supcr<:Ondoctiog 111m of liJiekncss [; in 11.0 e"temal
magnL'Iic field n" is gin..,.. by, for [; 4 A.

(ees)
376

In SI the factor '" is replacOO by ~J.Io. We nt.'gII,.'d a kinetic energy c:ontribotioo to the
problem. (b) Show that the magnt!tic contribution to Fs when aH:ragt"<i over the
thickness orthe film is 8!<&A)'I961T. (c) Show that the critical field of the thin film is
proportional to (JJ6}H", where If~ is the bulk cri tkal field , if we consider only the
magnetic contribution to Us.

3. Two.Jluid model of a superconductor. On the two-fluid model of a sUJlerconductor

we assume that allernpcraturcs 0 < T < T~ the cUrTcnl density rn(Ly be written as the
sum o f the l.'Qntributions of nannal and supl.'T"COnducling cil..-clrons: j - j N + is. where
i ... - uuE and js is giY(.'fl by the London L'<juation . lien: 0"0 is an ordinary normal
conductivity, <k>crcased by the reduction in the numlx:r of normal ck'Ctrons at tern-
pconturc T as compared to the nonnal state. NI.'Glt.'Ct inertial cffl..'els on both)... andis.
(a) Sllow &om the Maxwell c(luations that the dispersion relation ronnt.'Cting
v,oa\'evt."CI.or k and frequency w for eh.octromagnetie waves in the ~upcrconductor is
(CCS) .<
(51)

where A~ is gh'el1 by (14a) with n replaced by ns. Reca11 that curl curl B '" _-V28.
(0) If T is the relaxation time of the nonnaJ electrons and nN is their concentration,
show by usc of tl.e expression (To - III\£Tim that al frequencies w «
lIT the disper-
sion relation does not involve the normal eh.'Clrons in an important way, so that the
motion of the electrons is deSCTibed by the London t.'1llmtion alone. TI,e supcrcurrcnt
short-circuits the nonnaJ electrons. The London equation itself only holds tme if liw
is small in romparison with the energy gap. Note: TI,e fre(Juellcies of inteR'St are
such tha t w « w,.. where w,. is the plasma fl"(!"(lucncy.

· 4. S'ruclure of a vor'ex. (a) Find a solution to the London equation that has erlilldrical
symmetry and applies ou tside a line l.'Ore . In crli ndrlcal polar coordinates, we wan t a
solution of

that is Singular at the origin and for wllieh the total nu~ is the flux ljuanlum:

2;; r dp pB(p) - 41 0 •

TI,e C<juation is in fact valid on l ~' outside tIle normal core of radius E. (0) Show that
the solution has the limits

B{p) "" (<I>oI2 1TA~ln(Alp) •

B{p) "" (¢J271"A~ ;;)J2p)1Ii exp( - piA)

°1"hi5 problem ;S $OmCwhat difficult.

 .- " :'

I2 S"puco"duc/idl!l 377

5. Lolldoll pell{tfratiOl • .u:pth. (a) Take the time derivative of the London (.'(I'":.1ion (10)
to show that aj/at "" (c"/4'11",\1JE. (b) If mdv/dt - qE, as fot free carri(.·u or
charge (/
and mass m , show that '\1- rntJI4'11"nr(.

'"fi. Diffraction 4Jm of jOKplw)n jllllction. Consider a jullCtion cf rorumguiar CTO$5

St.'Ction "itl1 a magnetic Held B applit-d in the plan e of the junction, normal to an (.odge
awidth w. Lei the tl1ickness a the junction be T. Assume for conveniellce that the
phase difference of the two superconductors is '11"12 when B ., O. Show that the de
curr(''1lt in the presence of the 1lIab'lletic Held is

sin(wTBellic )
1 = 10 (wTBcl1ic)

7. Meiuncr efftcl in rphflre. Consider a sphere of a type I 5uperconductot with critical

Held H". (a) Show that in the Meissner n:gime the cff(.octive magnetu.ation U witilin
the sphere is given by -SwUf3 " B". the uniform applied magn(.>l.ic fldd. (b) Shew.'
that tile magnetic field at tlll! surface of the sphere in the cquatoru.1 plane is 38.12. (It
follows that thc applic..-d field at whidl the Meiss ner l.ifc..'Ct starts to br(.'"3.k dew."li is
211/3.) Reminder: The demagnetu.ation field of a un iformly magn etized sphl."1"e is
-4'11"MI3.

References
R. D . Parks, cd., SupermnducIlL"'Y. Deld«...... 1969. Good collection of revir.:w artic!cs.
J. R. Sdlriclfcr. ThoorJI 0/'"1Ieramtiudiuilll, ...,vised cd .• BcnjamlniCummlngs. 1983.
C. !llI..byi.Cn. Theory <Jj $Ullf'rconduclluily. interscie nct'. 1965.
M. TInkham, inlrotludion 10 '"1If'rco,uiucllllllll. Krieger, 1960.
M. Cruan. Theory of 6lIperwnd"diLiII/. World Scicntil1c. 1969.

TITE II SUI'ERCONDUCf[vrrV
P. C. dOl C<..~ltIe5, S"IJt!rcorl(fucljv'I!I of me/ab ,wd aI/OilS. Addisou ·WcsIOl)'. 1989.
T. Luhman and D . Dew· Hughes, ..-cis., Melall"rgy of $UJl<!rco.ufud'.'f: "lIIlerwb. Ac-adcmi<'. 1979.
M. N. Wihoo, S''Ile roonc/uctlng magnel •• OxIOro, 1963.
D. Saint·James. E. D. ""lOm~. alld C Sanna. Type 11 , "pnronduditul/. Pergamon. 1969.
R. P. lleubc"er , Ma{ywlicjlwt ""'C/UI""U III $Upc<rcorll/uc/ors. Sprll1~,'e r, 1979.

HIGII TEMI'ERATURE SUPERCONDUCrOnS

C . Bums, lfigl,.,n"lIf't""6/lIr't fI"'Jl<!ramducth..'UII. Acadcmk-. 1991 .
M. 'I1nkham and C. J. l.obh. Mrh~'sical p~rti<.'S of the new supcn...·" 'dllctors:· Solid state ph~'sia
42, 90 (1989).
R. t'. Curl and R. E. Smalley. " t'Ullercll~," Sci. Am "'0. 4. p. 54 (t99I ).
D. R. lI uffmali . "Solid Cc.o. Physics Too.)" No. II . 22 (1991). Summary or phys;c.J propert ic~ or
P

thl.· bud:m;n$lOlrful1erc"<..".

JOSEPHSON EFFECl'S
A Bartlne and C . Paterno), P}'IPiu lJlld oW/ko/ to .", 0/ tile J~I.IO" efftcl. . W,I<..""),. 1962.
E. 1_ Wolf. Prind,ies of electron 'u,mdirrg ' IIf'Ct.wcollll. Oxford, 1965.
j . C. CarloI', SQUiDS , t/,e Jo$ephron efftcl.'. ond S"IJe rc:onducling electronics. Uilge" 1991.
J Clark... and R. /I . Koch. "Impact ofhigh· tc mpcrnlurc ~upcrwnductl\'ily on SQUII) mllgne tomc·
krs, Seit,ncc 242, 217 (1 968).
M
 (
(-- .
 13
Dielectrics and F erroelectrics
_ a
Maxwe ll etluations
J>olariz.ation ""
381

MACROSCOI'IC ELECTRIC FIEI.D 38'

Oel)Olarization field , El 38<

LOCAl . ELECfRIC FIELD AT AN ATOM 3S6

Lorentz field, E,; '88
Field of dipoles inside ca,·ity. ~ 388

DlEI.ECTRIC CONSTANT AND l'OLARJZABlLITY 388

E lectronic pobrizability 390

Sfn UCTURAL I' BASE TRANSITIONS 393

FERROELECfRIC CHYSTAI.S
Classification of ft:TTOcil.."Ctric cr)'stals ,..
393

DISPLAC fVE TRANSITIONS 398

Sort optical phooom .00
Landau theory of the phase transition .00

...
Second-order transition '02
F irst-order transition .02
Anli fen-ociedricity
Ferroel«tric domains ,06
I' >eweleclricity '08
Ferroelasticity 410
Optical ceramics 410

SUMMAI\Y .. 0

.........
)'nOBLEMS
I. Atomic hydrogen
2. Conducting sphere
3. Erred of air gap
4. Interfacial polarization ......
5. I'olarization of spilere
6. Ferroelectric criterion
7. Saturation polarization
...."
...
8. Dielectric constant below transition 413
9. Soft modes 413
10. Ferroelectric linea r arra y

NOTATION, ~ _ 107/4wC ;
(ee5) D - r; + 477f' ,.. ef - (1 + 4'iTx)f (l = III flccol ;

(51) (l "" pI f lon] ;

41TXC,-S - XSI ;
 Figu.... I The pcrmalle.,1 dipole momen t Q( II moIocule fA
""~ter has the magnitude 1.9)( 10- 11 el;u~m and iii di_
reded rrom 0 -- lOll to"aro the midpoillt of the line C.'Oll'"
n....ctillg the II · ion5. (To rolwerl to 51 units, multiply by
t )( 10".)

~.
"' .. p--;:t
.
£ .. ~
£
"-~
par'
,- I)
, "' II ,..

, ,,
•
• ~C- __________________.-,

Figure 2 Ekctrostatic potential aud field rotnpollents in CCS at po:$itiOt' r, 8 for a dipole p
e-
diredc...:! a10ug the :t lUi$. For O. we have E, - E~ - 0 and E. - 2pJ,fi; for 8 - wI2 we have
E, - E~ - 0 alld E." -pl,fi. To couvert to SI, replace l' by ,!/4mo. (After E M. PUlcell.)
 CHAPTER 13: DIELECTRICS AND FERROELECI'ruCS

First we relate the applied electric field to the in ternal electric field in a
dielectric crystal. TIle study of the electric field within dielectric mailer arises
when we ask
• What is the relation in the material between the dielectric polarization P and
the macroscopic electric field E in the Maxwe1l equations?
• What is the relation between the dielectric polarization and the iocru ekdric
field which acts at the site of an atom in the lattice? TIle local field determines
the dipole moment of the atom.

MaxweU Equafions

(CCS) (51)

curl H = -
41T
j
1 a
+ - - (E + 41TP)
.
.curl H :: J +
,
~«(oE + :P)
c c at

curl E =
1 '8
curl E =
'8 "
c at •
div E = 41TP div(oEF p; "
div B = 0 ; div B >=. 0 .

I'olarizotion
The poitwization P is defined as the dipole moment pel' unit volume,
averaged O\'er the volume of a cell . TIle total dipole moment is dt-fined as
( 1)

where r " is the position veclor of the charge q". TIle value of the sum will be
independent of the origin chosen for the position vectors, provided that the
system is neutral. TIle dipole moment of a water molecule is shown in Fig. L
The electric fi eld at a point r from a dipole moment p is given by a standard
result of elcmelltary electrostatics:

(CCS) (2)

The lines of force of a dipole point ing along the z axis arc shown in Fig. 2.

36'
 MAcnOSCOPIC EU·:crRIC FIELD

One contl'ibution to the e lectric field inside a body is that of the applied
ele<:tric field . defined as

Eo - fi eld produced by fixed charges extemal to the bOOy (3)

The other contribution to the e lectric field is the sum of the. fields of all charges
that constitute the body. If the body is neutral. the contribution to the average
field may be expressed in terms of the sum of the fields of atomic dipoles.
We deOne the average electric field E(ro) as the averllge field over Ihe
volume of the crystal cell that contains the lattice point ro:

E(ro) = _ 1_
V<
JdV c(r) , (4)

where c(r} is the microscopic electric field at the point r. The field E is a much
smoother quantity than the microscopic field c. We could well have written the
dipole field (2) as c(r) because it is a microscopic uuslnoothed field.
We call E the macroscopic electric field . It is adequate for all problcms ill
the e lectrodynamics of crystals pro\'ided that we "-now the connection between
E, the polarinltion P, and the ClLrrent density j, and provided that the WaVe-
lengths of interest are long in comparison with the lattice spaCing. I
To find Ihe. contribution of the pllari ....ation to the macroscopic field, we. can
simplify the sum over all the dipoles in the specimen . By a famous theorem of
clectrostati~ the macroscopic electric fleld caused by a unifonn pllari;taliOIl is
cquaJ to the electric field ill Vacuum of a fictitious surface charge density

'11 detaikd dcrh-ation of tbe MaWo'f:'1l equatious for the macrosmpic fields E ,'nd B. starting
from the ~lax",eU equatKx.s in terms ol the mir.ToK'opic fields e aI,d h , is gi'en by E. M. p"rcdl,
Elfflricily Gild ""'4l"dinn. 2nd L'(\ .. McC ... _II ,II. 1965.
IThe cltdrOlitatie pote"tial In CGS ""its ol a dipole p is <p(r) - p' gr.u:l(lfr). For A volume
distn!;ution or polarization P we hd"f:

¥(r) - f dV(P'b'T1lll~),
which by a "ector iuentity bt.-romes

!p(r) - f dV(--;dIV P+diV!)

If P is constant, tben div P - 0 .nd b)' the Cau~. tlleoTt.,m we J,a"e

!p(r) - f liS ~" - f dS: '

where dS is Dn clement of area on the surface or the bod)'. 'This comple tes the proof.
 <I '" +1'

~. ~: ,v-)
<r " - I'

r;" gure;} (~) A uniformly polarired didectric , lah, wit h the poIari1.(1tlon n!Ctor P normal to the
plane of the dab. (h) A p"ir of unirormly charged p.1ra llel plateo wh leh gi\c rise to the identical
dcdri<: fleld E, AS In (a). llJe upp"r plate has the surf","", charge dcnsoty u - +1', And the low<.'T
I,late IW$ Q' - ~ I'.

u =:: n' P on the surf., cc of the lxldy. Here Ii is the unit normal to the surfact',
drawn outward from the polarized matter.
\Ve apply the result to a thin d ielectric slab (Fig. 33) with a uniform volume
polarh,;tlioll p, "nle electric field EI(r ) produced by the po lari~t i on is equal to
the field proouced by the fictitious surface charge dcnsity u "" n' P on the
surfoc'C orthe slab, On the IIpper boundary the unit vcctor n is directed upwaru
and on the l()\1,'Cr boundary n is directed dO\\'JIward. The upper boundary bears
the fictitiotlscharge u =:: n' P "" P pcr unit area, and the lower boundary bears
- P per unit area.
The eiOOnc fieJd E. due to these charges has a simple form at any point
belween Ihe, plates. but comfortably re moved from their edges. Dy Causs's law

(CCS) E. = - 4TTjul = - 4r.P ; (SI) E, = lui

__ c

" ..
P
·1 I',)
J

We add E t to the applied field Eo to obtain the lotal macroscopic field

inside the slab, with z the unit vector normal to the plane of the slab:
ICeS) (5)

(51) - -,
p

"
We define

E I _ field of the surface charge density n' P on the boundary (6)

1ltis field is smoothly varying in space inside and outside the body and satisfies
the Maxwell equations as written for the macroscopic field E. 1lte reason E, is
a smooth fun ction when viewed on an atomic scale is that we have replaced the
discrete lattice of dipoles Pj with the smoothed polari7.ation P.
Depolari:wtion Field, EL
11le geometry in many of our problems is such that the polari7.ation is
uniform within the body, and then the only coulributions to the macroscopic
field arc from Eo and E 1 :

I E = EO+E 1 (7)

lIere Eo is the applied field and EJ is the field due 10 the uniform polarization.
TIle field EI is called the depolari:.r..ation field , for within the body it tends
to oppose the applied ficld Eo as in Fig. 4. Specimens in the slJape of ellipsoids,
a class that includes spheres, cylinders. and discs as limiting form s, have an
advantageous property: a uniform polari~.ation produces a unifonn depolari..a.-
lion field inside the body. This is a famous ~athematiatl result demonstrated in
classic texts on electricity and magnetislll. 3
If P~ , PilI P". are the components of the polari7.ation P referred to the
principal axes of an ellipsoid, then the components of the depolari7.ation field
are written
(eeS) Eh = - N,P. ; EIII = - Nllf ,l ; E 1z = - N".P,," ; (8)

(S I) E.
_ NJ'"
EIII =
_!!ti Eh =
_ N".P&

'" '0

Here N", Nil ' N,," an> the depolari7.ation factors ; their values depend on the
'0

ratios of the principal axes of the ellipsoid. TIle N's arc positive and satisfy the
sum mle N. + Nil + N: = 4"IT in CC5, and N. + Nil + N: = I in 51.
Values of N parallcl to the figure axis of ellipsoids of revolution are plotted
in Fig. 5; additional cases have been ca1culalcd by Osborn"' and by Stoner. In
limiting cases N has the values:

N N
Shape:- Ad~ (CGS) (51)

~phere '"I 41T13 1/3

TIlill slab normal <. 1
TILin slab in plane 0 0
I...olg circular c),liooer longitudinal 0 0
Long circular crlinder tranS~"eI"Se 2w 112

We can reduce the depolarization Reid to 7.em in two ways, either by working
with a long fine specimen or by making an electrical connection between elec-
trodes deposited on the opposite surfaces or a thin slab.
~R Becker. Electro.nlll(ndic {.elm lind intcrm;tion&, Bla.i)dell. 19&1, W . 10-2_ 107.
'J. A. o.bom. Phrs· Ita. 67 , 351 (t 945), E . C . StOller, PhUosophll'lli MagD.irte 36, 8<XJ{l9-I5)
 + + +
• +
• +
. Eo + E~--r
• +
+
+ + +
~re" 'J1}e depolaliution !'tcld E, is Olll)()l>ite 10 P. 'l1'e r.ct;tiou~ surface charges a..., indicated ,
the field fltl1(.'\Al chargc~ Is E, "ithin the ellipsoid .
ce,
"
" '"
" "
•• <
;;• "
" '"
"
.,
,, , , + ,
ri. "
Figure 5 I)epob.riutioc, &ctor Nparnllcl to the figure ads of c llipso;J s of rt:\-oIuhoo, a. a limct.1on
d tI,e axioll ratio cia.

A uniform applied field Eo will induce uniform polarization in an ellipsoid.

We introduce the dielectric susceptibility X such that the relation s

(CCS) P = XE , (SI) I) ,'::::i. £oXE (9)

connect the macroscopic field E inside the ellipsoid with the polariz.1tion P.
Here XSI "" 47rXccs·
If ~) is uniform and parallel to a principal axis of the e llipsoid, then

(CCS) f; = Eo + "~t "" Eo - NP ; (S I) E=Eo - - - ,

NP
'0
( 10)

by (8), whence

X
(CCS) P = x(f.:o - .v P) P= Eo; (II)
1 + r.;X

X<o
(SI) P "" x(EoEo - NPl P= Eo .
1 + -NX

The ",due of the polari zation depends On the depolarization factor N.

 3IJ6

LOCAL ELECTRIC FIEU) AT AN ATOM

The value of the local electric fleld that acts at the site of an atom is
5ignificantly different from the value of the macroscopic electric field. We can
convince ourselves of this by consideration of the local field at a sile wjth a cubic
arrangement of neighbor~ in a crystal of spherical shape. The macroscopic
electric field in a sphere is

(ces) (12)

(Sl)

by (to).
But consider the field thai aels on the atom at the center ohhe sphere (Ihis
atom is not unrepresentative). If a1l dipoles are parallel to the z axis and have
magnitude P. the z component or tile field at the center due to all other dipoles
is, from (2),

(ces) . _ '" 3~-,f _ '" 'l4-r'/ - y'f

Edil'ole-P£.J !I. -P£.J :s (13)
I rr rf j

In 5 1 we replace p by p/41reo. TIle :t, y, Z dir(.'Ction~ arc equivalent because of

the symmetry of the lattice and of the sphere; thus

whence E<li[>O'" = O.
The correct local field is just equal to the applied field, E"-I = Eo. for an
atom site with a cubic environment in a spherical specimen. nlU S the local field
is not the same as the macroscopic average fie ld E-
We now develop an expression for the local field at a gene ral lattice site,
not necessarily of cubic symmetry. The local field at an atom is the sum of the
electric field Eo from extemal sources and of the field from the dipoles w ithin
the sp{:cime n. Jt is convenient to decompose the dipole field so that part of the
summation over dipoles may be replaced by integration.
\ Ve write
(14)

~A tom site, in acubicCl)1ta1 do not n~-ccssarily have cubic symmetry: th"s the 0"- sites in the
barium titanate ' tructure of Fig. 10 do not have a cubic environment. Howev<: r, the Na+ and CI-
sites 10 the lIIaCl structure and the Cs· and 0 - sites in the C.CI $Iructurc ha"e <:ubic symmdt)'·

•
 - •

- ..... E,rnM"

--
",,'t"
b..""d,,~

1::, I"",, d't.,I.>,

,,,,,,1"'1,1,.·,,·
f ';Kure I)The ill teo,al ek:ctric field 011 an atom in a cr)'~tal is the sum oflhc c~lcnl~1 yp(lli~1 flcld
Eo and of the field title 10 th" other atoms ,n the cry~ta\. The stanllar<1 mdhod of ,umming the
.lipoic fieltls of the other atoms ;s fint to Sum individually 0\-.,,. a modcl'1ltc "umber of ncighOOring
alolns I",ide an imaginary sphere COIl<:<'ntric "ith the r"f"'Tenee alom: this dc:fines the Ilcld ~
which ..ani5~ .t a ",ICrenee Site with ""hie symmetry. 1llc alOlns o.,l.side the sphere ('all be
tn,alcd as a uniform!)' polamed dielectric_ TIleir cunlriblltion to the field 81 IIIC' n..-fc:rcnc-e point IS
E, + E.. where E, iii the depolarization field auociatC<! "ith the outer oo,,"cbry and Ea is the Held
alOSOCialcd with the surface of the spht'rical cavity.

Ilere
Eo = field produced by fixed charges external to the bcxly;
E, = depolari7..ation field. from a surface charge density i\' P on the outer
surface of the spt.'Cimen;
Ell "" LorentL cavity field: field from polarization charges on inside of a
spherical cavity cut (as a mathematical fiction) out of the SIx.'Cimen with the
reference atom as center, as in Fig. 6; EI + Ell is the fkld due to uniform
polarization of the body in which a hole 1m... been created;
E3 "" fie ld of atoms inside cavity.
111e contribution EI + ~ + E3 to the local field is Ihe lolal field alone
atom caused by the dipole moments of all the other atoms in the specimen:

(ces) (15)

and in 51 we replace Pi by p/47T£o.

~ipoles at distances greater than perhaps ten lattice l'Ol\Stants from the
reference site make a smoothly varying contribution to this sum, a contribution
which may be replaced by two surface integrals. One surface integral is taken
over the outer surface of the ellipsoidal specimen and defines E\. as in Eq. (6).
TIle second surface integral defines ~ and may be taken over any interior
surface that is a suitable distance (say 50 A) from the reference site. We count in
~ any dipoles not included in the volume bound<.-d by the inner and outer
surfaces. It is convenient to let the interior surface be spherical.
 (
wrent1. Field. ~

The field E2 due 10 the polarization charges on the 5urface of the fictitious
cavity was calculated by Lorentz. Ir (J is the polar angle (Fig. 7) referred to the
polarization direction, the surface charge density on the sunace orthe cavity is
- P oos 8. The electric field al the center of the spherical cavity of radius a is

(CCS) E2 =
r
o
(u - 2)(21TCI sin 8)(a d8 )(P cos fJ)(cos 0) = -
4~
3
p (16)

(SI)

This is the negative of the depolarization field E\ in a polarized sphere. so that

E I + E2 = 0 for a sphere.

Field of Dipoles Insule Cuvity. E3

The field E3 due to the di poles within the spherical cavity is the only term
that depends on the crystal structure. We showed for a reference site with cubic
surroundings in a sphere that £3 = 0 if all the atoms may be replaced by point
dipoles parallel to each other.
TIlt~ total local field at a cubic site is, from (14) and (16),

471 471
(CCS) EJocaI = Eo + E , + ""3 p:; E +""3 P (17)

(SI )

This is the Lorent-L relation: the field acting at an atom in a cubic site i5 the
macroscopic field E of Eq. (7) plus 471P/3 or P/3Eo from the polarization of the
other atoms in the specimen . Experimental data for cubic ionic crystals support
the Lorentz relation.

DlELECfRlC COXSTANT AND POLARlZABIUTY

The dielectric constant E of an isotropic or cubic medium relative to vac-

uum is defined in terms of the macroscopic field E:

(ces) E + 471P
= 1+ 471X (18)
E

(SI)
('of + P
"""' Ii-X _
<0£

Remember that X~I = 471XCCS . by definition, but ESI e ECC5'

 .....

•
•
fQ:u,e 7 Calculation dille fleld in . ,pl'~'fi<.';I1 ca~ily in • UII' - ('I..,,,,. ,.. ,in!,:
foon l)' polariU.Q ml-dium. 21T N . 1" 6 · ~ II~· I' "" 6

The susceptibility (9) is related to the dielectric constant by

P E - 1 P
(ces) X "" - == - - (Sl) X - - - =i - l (19)
E 4"lT ,0£

In a noncubic crystal the dielectric respOJlse is described b)' the components of

the susceptibility tensor or of the dielectric constant tensor:
(ces) P,. ::: X,. •.E~ ; (20)

(Sl)

TIle polari"tability a of an atom is defined in te rms ofthe loc-a.1 e1l."c tlic field
at the atom:
p "" aE .........1 , (2 1)
where ' ) is the dipole moment. This definition applies in CCS a\1d in SI, but
aSJ "'" 4"lTEtPCCS' 'nlC polarizability is an atomic property, but the dielectric
constant will depend Oil the manner in which the atoms arc assembk-d to form a
Cl)'stal. For a nOll-spherical atom a will be a tensor.
The polarization of a crystal may be expressed approximately as the prod-
uct of the polarizabilities of the atoms times the 10l"3.1 eit..-chic field :

p = 2: Njpj == L NpJI-:iuJ.j ) •
j ,
(22)

where Nj is the concentration and aJ the polarii'.ability of atoms j , and I-:"..{J) is

the local field at alom sites j .
We want to relate the diek'Ctric constant to the poia.ri.-.abilities; the result
will depend on the relation that holds between the macroscopic electric field
and the local electric field . We give the de rivation in CCS units and state the
resuh in both systems of units.
 If the local field is given by the Lorentz re lation (Ii), then

(CGS) p: (INpA (E+ 4; p)

and we soke for P to find the susceptibility

(CGS) X = 1'- == 'f.NJo'j (23)

E 41T
1 - 3INJoj

By definition € = 1 + 41TX in CCS; we may rearrange (23) 10 obtain

E - 141!'
(CGS) -- ~- ,£N -aj (SO (24)
E +2 3J

the Clausius-Mossetti relation. '£bis relates the dielectric constant to the ek-'C-
Ironic polari~bility, but only for crystal structures for which the Lorcnt:t: local
field (17) obtains.

Ekctronic Poillri::.aiJility
The lolal polarizability may usually be separated into three parts: e lec-
tronic, ionic, and dipolar, as in Fig. 8. The electronic contribution arises from
the displacement of the electron shell relative to a nucleus. The ionic contribu-
tion comes from the displacement of a charged ion with respect to other ions.
ihe dipolar polarizabilily arises from molecules wilh a permanent electric <li-
pole moment that can change orientation in an applied electric field.
In heterogeneous materials there is usually also an interfacial polariz..'l.tion
arising from the accumulation of charge at structural interfaces. This is of little
fun<lamental interest , but it is of considerable prnctical interest because com-
mercial Insulating materials are usually heterogeneous. 6
111e <lielectric constant at optical frequencies arises almost entirel)' from
the electronic polarizability. TIle <lipolar an<l ionic contributions are small at
high frequencies because of the inertia of the mok'CUles an<l ions. In the optical
rnnge (24) re<luces to
,,2 _ I 41T
(CGS) , 2 ~ - INpfielectronic) (25)
n + 3
here we have use<l the relation ,,2 = E, where" is the refrnctive index .
By applying (25) to large numbers of crystals we detcnnine in Table 1
e mpirical values of the e lectronic polarizabilities that are reasonably consistent
with the observed values of the refractive index. 111e scheme is not entirely

81'0r r..fercl"lC"es s.,e D. E Aspnt."1i. Am. J. l'hys. SO, 704 (1982).

 1.1 Dieltttm' lind f'ur oelutrie.

"'''" I
I",;.,."'.,"'\
"d'IM>I...

-Lr ----"--~
--~r~---------

Figure 8 J' ,t'<I<1t' lIcy tlqlCLltlcn('C of Ihc SC-·...·.,.1 C{I1llr.1J.,lio"s to the pol,";7-'lb;\ih·

Table I Electronic poiariz.abilitici or ions, in IO-~ crnl

110 U' nc" n" C..

I'auling 0.201 0.029 0.008 0.003 0.0013

IS O.fr29
0'- F- N. Na ~ ~It .. AIH Si H
Pauling 3."" 1.04 0.300 0.179 0.094 0.052 0.0165
jS.{TKS) (2,4) 0.658 0.290
S'- e l- M K' c." So" TiH
Pauling 10.2 3.00 1.62 0.83 0,47 0.286 0.185
jS·n'KS) (55) 2.947 1.133 (1.1) (0. 19)
Se' - Br- Kr Rb ' S~' y" Z..'"
Pauling 10.5 4.n ~ 46 1.40 0.86 055 0.37
j S-cr KS) (7. )
Tel - ,-
4.091
X,
1.679 (1.6)
13aZ, ..." Co"
I'auling
IS{fKS)
14 .0
(9.)
7, 10
6. 116
3.99 ""
2.42-
2.743
1.55
(2.5)
UJ.I 0.73

Val,," from L . Poul.n", 1'"",. R. Soc. London A114 . 181 (1927), S. S. J'svm\ and T. P. Shalln.,
I'hp. C hcIII . Solids 34. 509 {l973\; and J. T,'Ssman. A. Kahn , and W. Shockk'}·. l'by5. Rc'·. 92.
8OO{l953). The TKS pol. llzabilities arc at thc frt'q""' ''c), of lhe D Iincs of so.li"III. 110e val"..., are in
CCS ; 10 <"OIwert to 5 1. multiply by' )( \0 - ''',
self-consistent. oc'Cause the ci.:.'Ctronic polarizability of an I" " depends some-
what on the environment in which it is IJI:u.'e d. TIle negative ions are highly
polari7.able because they are large.

Classical Tllcory of Electronic PoIuriwiJility. An electron bound hannoni-

cally to an atom will show resonance absorption at a frequency Wo = (ftlm)II2,
where {3 is the force constant. The displacement 1" of the electron occasioned by
the application of a field Eloc is given by
-eEk>c = (31" = II)~ (26)
so that the static ck'Chonic polari7.ability is
a(e1ectronic) = 1J/Eloc :: - exlE Ioc = ff/lll~ . (27)

The electronic polarizability will depend on fr(!quency, and it is shown in

the following example that for frequency w
,rIm
(ces) a{ek'Ctronic) = ,;:.:::::" (28)
wS - WI

but in the visible region the frequ ency dependence (dispersion) is not usua1ly
very important in most transparent materials.

EXAMPLE: frequerlClJ dqwnderlce. Find tlK! frC(luency depcndcl'\('"t! of thc electronic

polarizability of an electron ha\';ng the resonance fl"C<lucncy Wo. treating the system as a
silllple harmonic oscillator.
'nle e(luation of motion in tile local ell.'Clric field Ek>c: sin wI is

!i() that, for;( = :to sin wI,

m( - w2 + ~o - - et:"", .
TIle dipole moment ha.~ the amplitude

from which (28) follows.

In quantum theory the expression corresponding to (28) is

(ces) a (electronic) = !:.... L ~ii foj-

111 ) W
2 • (29)

where JiJ is calk.od the oscillator strength of the ek'Ctric dipole transition be-
tween the atomic states j and j.
 · .

~1\UcruRAL PIIASE TRAi'\SJTIOJlo;S

It is not um.'ommon for crystals to transform from one cl)'5tal stnlcture to

another a'i the temperature or p ressure is varied. l bc stable structu re A at
absolute zero generally ha.c; the lowest accessible intemal energy of all the
possible structu res. Even this selection of a structure A can be varied with
application of p ressure, because a low atomic volume will favor c1osest-packl.'ll
or even meta1lic structures. Hyd rogen and xenon, for example, become metal-
lic under extreme pressu re.
Some other structure B may have a sorter or lower frequency phonon
spectrum than A. As the temperature is increased the phonons in B will be
more highly excited (higher thermal average occupancies) than the p!lOllons in
A. Because the e ntropy increa.<;es with the occupancy, the entropy of B will
become higher than the entropy of A as the temperature is increased.
It is thereby possible for the stable structure to transform from A to B as
the temperature is increased. l 11e stable structure at a temperature T is deter-
mined by the minimum ofthe frt.'e energy F "" U - TS. '[bere will be a transi-
tion from A to B if a temperature T" exists (below the melting point) such that
f,,(TJ = F8(Tc)'
Often several structures have nearly the same internal energy at absolute
zero. The phonon dispersion relations for the structures may, however. be
rather different. The phonon energies are sensitive 10 the number and arrange-
ment of nearby atoms; these are the quantities that change as the structure is
changed.
Some structural phase transitions have only small effects on thc macro-
scopic physical p roperties of the material. llowever. if the transition is influ-
enced by an applicd stress, the crystal may yield mechanka1ly quite easily ncar
the transition temperature bet.'fIuse the relative proportions in the two phases
will change under stress. Some other structural phase transitions may have
spectacular effects on the macroscopic electrical properties.
Ferroelectric transitions are a subgroup of structural phase transitions. a
subgroup marked by the appearance of a spontaneous dielectric polarization in
the crystal. FerroelectriC'S are of theoretical and technical interest because they
often have unusually high and unusually lemperature-dependl:nt "alues of thc
dielectric constant, the pie-ax:lectric effect, the pyroelectric effect, and electro-
optical eftt.'c ts. including optical frequency doubling.

FERROELEc....... R1 C CRYSTALS

A ferroelectric crystal exhibits an electric dipole moment C\'cn in the ab-

sence of an external electric field . In the ferroelectric state the center of posi-
tive charge of the crystal does not coincide with the center of negative charge.
The plot of polari1.atiollS versus eI(:ctric field for the ferroeil.-ctric state
shows a hysteresis loop. A crystal in a normal dielectric state usually does not
39.

show significant hysteresis when the electric field is increased and then re·
versed. both slowly.
In SOUle cr)'s tal~ the fe rroelectric dipole moment is not changt..-d by an
e lectric field of the ma:<im ulIl intensity which it is possible to apply before
causing electrical breakdown . In these crystals we are orten able to observe a
change in the spontaneous moment when the temperJ.lure is changed (Fig. 9).
Such (·rysta1s are called pyroelectric.
Lithium niooote. LiNb0 3• is pyroelectric 011 room temperature. It has a
high transi tion temperature (1'" = 1480 K) and a high saturation polarization
(SO ~Clcm7. 11 can be "poled," which means given a remanent polarization, by
an electric fie ld applied over 1400 K.
Fermclectricity usually disappears above a certain lempcfollure C;\Ued the
transition temperature. Above the transition the cr~ stal is said to be in a
paracled ric state. The term paraclectric suggests an analogy with paramagne t-
ism; similarly, there is ul>ually a rapid drop in the dielectric constant as the
te mpel ature increases.

Cwsiflrotion of FcrTOfIlu tric Crystal&

We li ~ t in Table 2 Mlllle of the crystals commonly <."Onsidercd to be fer-
roek'Ctric, along with the tranSition temperature or Curie point Tr at which the
crystal changes from the low temperature polarized state to the high tempera-
ture unpolarized state. 111ermal Illotion tends to destroy the ferroelectric
order. Some ferroelectric crystals have no Curie point because they melt before
leaving the ferroelectric phase. The table also includes values of the spontane-
ous polarization p•.
Ferroelectric crystals may be classified into 1">\0 main groups, orde r-
disorder or displacive. Jf in the paraelectric phase the atomic displacements Me
oscillations about a nOlllX'lar site, the n after a displacive tra nsition the oscilla-
tions are about a polar site.
Jf in the paraelcctric phase the displacements are abou t some double-well
or multi-well <.'O nfiguration of sites, then in an order-disorder transition the
displacements are aoout an ordered subset of these "ells.
111ere ha.~ been a te ndency recently to define the characte r of the tnmsi-
tion in te rms of the d ynamics of the lowest frequency C"MlIl') optical phonon
modes. Jf a soft mode can propagate in the crystal at the transition, then the
tran ~ition is displach·e. If the soft mode is only diffusive (non-propagating)
there is really not a phonon at all, but only a large am plitude hopping motion
between the wells of the order-disorder system. Many ferroeleclrics have soft
llIodes that fall between these two extreme5.
The orde r-disorder class of ferroelcctrics incl udes crystals with hydrogen
Londs in which the motioll of the protons is related to the ferroelectric proper-
ties, as ill pota.s.~ium dihydrogell Ilhosphate (KH 2 PO.J and isomorphous 5al ts.
 13 Vif'/edr;n "nd "'erro"'«tric~

' OM

... ,
,
•
•

...., ".
~
8 7
•

~~ 100
T• • 4-I9'C
T, - -I92"C

(h)

". 300 '00 ..

Figure 9 TIle tempc ... ture iation or (a) the dielectric constant e, (b) the p)roelectrie ~"QCffidcnt
<I",,", and (e) th., specirtc heat
V"....

c~, of PbTi~. (After Remeika and CIa ... )

'"
Table 2 }<' crroe\ectrK: crystals·

To obtain p. in the CCS unit or e~u em-I, multiply lite w lue given in poe cm -~ by
3 x 103.

T" ill K P" in ,..e <,' ru- 2 , at T II:

KDP type KH 2PO,J 12:3 4.75 [96]

KD 2PO. 213 4.63 [180)
5 .• [90]
HbIl 21'O.
KI-I 2AsO.
GeT,
'"'"
fiTO
5.0 [78J

TGS type Tri~gl )'cinc

sulfate 322 2.8 [29]
Tti-gl)'cine
selenate 295 3.' [2B3]
Perov$ldtcs BaTiOa «l6 26.0 [296]
KNbO, 708 30.0 [523]
PbTiD, 765 >50 [296]
50
LiTaD,
Li NbO:J ""
1480 71 [296]

- " compilation of data on ferroelectric and antiICrroclcctric: materiab is given by E. C. Sub.

arao, I'crroek'CIric< 5. ffi1 (1973).

"111C behavior of crystals in which the hydrogen has becn rcpiaL'CtI lJy de ute rium
is inlcrcsUng:
KDzPO,J
Curie temperature 213 K
The substitution of deuterons for Pl"Ot OIlS nearly doubles '1'". although the [raL'-
tionai dlangc in the molecular weight mlhe l'(lmpound is less than 2 p en:cnL
111;5 cxtmort!inaril y large isotope shift is believed 10 be a quantum effect involv-
ing the mass-depcndence of the de Broglie wavelength. Neutron diffract jon
data show that above the Curie lemper-olture the proton distribution along the
hydrogen bond is symme trically clong;lted . Below the Curie temperature the
distribution is more concentrated and asymmetric with respc<.'1 to neighboring
ions, so tha t one end of the hydrogen bond is preferred b)' the proton over the
other end.
-111e displacivc class of ferroclcctrks includes ionic crystal structures
closely related to the perovskile and ilmenite structures. -111e simplest fer-
rocledric crystal is CeTe with the sodium chloride strudure. We sh,,11 dcvote
ourselves primarily 10 <:ryslals with thc pcrovskHe structure, Fig. 10.
 ""

I~
o. o n"
On."
o
0-- C;
0"'

., ,.,
Figure 10 (a) The crystal structure of barium tit anate. The prototype cr)'Sta\ Is calcium titana te
(pc~k;te). n.e structure i. cubic. with. ija"' ions at the cube C01TIcn. 0"- ions at the faoc
C'Cnters. and a"'~ ion at the body center. (1)) Below the Curie temperature the structure is slightly
deformed. v.ith Sa·· and n" ions displaced relative to theO"- ions. thereby de~'c1oping a dipole
moment. n.e upper and lower mergen ioOli may mo,'e downward slightly.

2(1 x 10 •
//
//
,7-
V gJ "~I

, It j""o.,
, I I

\I"""'"~ill ...
, 1lI"""buI........ Tr1r~l

- -'00 ,~ - . '"
T"II" "".,III«{"q
• ., ., ,.,
Filun: II Spontaneous powiution projttkd OIl cube edge ofN.rium titanate. as _ function of
temperature. (Mter w. J. "len)

Consider the order of magnitude of the fcrroclc<.1.ric efTeds in barium

titanale: the observed saturation polarization p. at room tcmpcraturc (Fig. 11)
is8 x lQ4 esu ('TIl - 2 . Thcvolumcofal'CIl is(4 x IO- Bfl == 64 X 10- 24 cm 3, so
that the dipule moment of a cell is
(CCS) p := (8 X 10"' esu lm -~(64 X 10- 24 cm3) iii! 5 X 10- 18 csu em

(Sl) p £!!'(3 X 10- 1 C m-2J(64 X ~O-30 ml) B: 2 x 10-2'.) em.

If the po~ iti ve iom Ba2-+- and TIH wcn' movcd by I> CI 0. 1 A with rcspect
to the negative 0 2 - ions, the dipole mOlllent of a l'ell v>'Ould bc
Gel> a 3 X 10- 111 esu em. In LiNbOa thc displacements are l."Oilsider8Lly
larger, be ing 0.9 Aand 0.5 Afor the lithium and niobum ions rl.'~pectivcly.

DISPIACIVE TnA:"ISITIUNS

Two viewpoints contribute to an undcrstanding of a ferroelectric displ.lcive

transition and by extension to displadve transitions in general. We may ~veak
of :1 polariw.tion catastrophe in which for some critical (;ondilion the polari.la-
lion or some Fourier component of the polari7.ation becomes very large.
Equally, we may speak of the conde nsation of a transverse optical phonon.
lI ere the word condensation is to be umlcr..tooc! in thc Bmc-E in~t ein sense
(TP. I). 199) of .1 time-independent displal."Cmcnt of finite am plitude. 1"is can
occur when the corresponding TO phonon fr()(luency vanishes at sollle point in
the Brillouin zone. LO phonons always havc bigher frC(IUencies than thc TO
phonons of the same wave\"cctor, so \\"C arc not l."Ollcern(:d with LO phonon
<."OllIlcnsation.
In a polarization catastrophe the loll!.l ekx;tric field caused hy the i<mic
displace lllent is larger than the eia,tic restoring force, therchy giving an ,Isym-
metril-al shift in thc ~i tions of the ions. lIigllf'r order restoring fOrl"Cs will
limit the shift to a finit e displal"Cment.
The oc"(;urrencc of fcrrodrttrility (and antiferroelcctricity) in man)'
pc roHkite-structure crystals suggests that this ~ tm ctme is f,lvor.lhl y di~IXlscd
to a <Iisplaci\'c transition. Loc·... 1 field calcu lations make d ear the reason for the
favored jXlsition of this structure: the 0 2 ions do not have cubk surroundings,
and the loo:.-al field factors turn out to he unusually large.
We give first the simple form of the catastrophe theory, su ppo~i n g that the
100:.'31 field at all ato ms is equal to E + 41TP/3 in CCS or E + PI3Eo in Sf. l11c
tlll..'01")' given now leads to a SCl."Ond-orde r transition; the physical ideas can he
carried over to a first-order transition. In a seu>nd-ordcr transition there is no
latent heat; the order parame ter (in this instan(."C, the polari.l.atioll) is not dil,-
continuous at the tnl.nsition tempcmturc. In a fir~t-order transition there is
a \.Itent heat; thc order parameter (,'hanges discontinuousl)' "t the transition
temperature.
We rewrite (24) for the dielcctrk (.'Onstant in the form
Bn
+3 l: NP•
ICCS) I~)
4" ,
1 - - :rNa
3 ' ,
 ,

_&.,5<. ,T.o,
. r.... ....... ,T'O, SrT.o,

OLo--t--+--1,--~~~~--t--t--~-'~-+'--r
jj 1 ,j 6 7 Ii Ii 10 11 I!
11100 ..1 '
~ ... !

figure 12 Di.,I.'dnc ronstant VCr.;Uii 11(1' - l " ) ;n the "",racl."Ctn(: Itale (T > T.l nC peronk.t""
aU('r C . RUPllrecht ~nd R. O. Bdl.

where 0:, is the electronic plus ionic polarizabilily of an iOIl of t)<Ve i and N/ is
tl\(" number of iOlls i per unit volume. The dielectric eonstant becomes infinite
and pcnnits a finite polarization in zero applied field when
(eGS) (3 1)
TIlis is the (;ondilion for II polarization l'atastropile.
The value of E in (30) is sensitive to small departures of L Npi from the
critieal value 3/411". If we write
(eGS) (4m3) L N,lXl = I - 3s (32)
where 8 <C I , the dielectric l'Onstant in (30) be<.-omes
E "" 11$ . (33)
Suppose ncar the critical temperature oS varies linearly with temperature:
(34)
where f is a constant. Such tl variation of s or L N,o, might come from nornl.l\
thermal cxp.msion of the lattiC\' . The diel~1ric constant has the fonn

(35)
E =- T _ T" '

dose to the observed temperature variation ill the panleioctric state, Fig. 12.
Soft Opticm l'IU)nons
The Lyddane-Sadls-Teller relation (Clmpter 10) is
w}lwL = E(C:O)/I!(O) (36)
-llie static dielectric constant in{:rea.'Ies when the transve rse optit'al phonon
frequent')' decreases. When the statk' diclectrit· constant E(O) has a high value.
sudl as 100 to 10,000, we find that f.l}-J' has a low value.
When Wr = 0 the crystal is unslilbl e beuHlse thcre is no effective restor-
ing force. The ferroelectric BaTi03 at 24°C has a TO mode at 12 cm- I , <I low
frC<juency for an optiml mode.
If the transition to a fe rroelectric stale is first order. we do not find w-r ,., 0
or E(O) = c;:c at the transition . The LST relation suggests onl)' that E(O) ex trapo-
lates to a singularity at a temperaturc '1'0 below Te . In disordered Cdirt)",
ferrocl(x:trics the static dielectric l'Onstallt can be larger than suggested b)' the
LST relations. 7
The association of a high static diclectric l'Onstant with a low frequenc)'
optit'al mode is supported by experiments 011 strontium titanate, SrTi03. A{,-
l'Ording to the LST relation, if the reciprocal ofthc staticdielcctrie constant has
a temperature depende nce 11E(0) a: rJ' - 'IOJ, the n the ~luare of the optical
mode fre(IUenl,), will have a similar tem perature dependence: wl a: ('J' - '( 0), if
Wt. is independent of temper.lture. The result for w} is very well confirmed by
Fig. 13. Measurements of cur versus 'I' for another ferroele<:tric crystal, SbSl,
arc shown in Fig. 14.

Landtlu Theory of the Phase Transition

A ferroelectric with a first-order phase transi tion betwccn the fcrroelectric
and the paraelcctric state is distinguished by a diSl'Ontinuous change of the
saturation polarization at the transition temjX!raturc. 'nIl' transition betv.'Ccn
the normal and supen.'Onducting statcs is a scl'OlId·order tr'.Insition, as is the
transition bctv.'Cen the ferromagnetic and panunagnelic slates. In these transi-
tions the degree of order gocs to zero without a diSl'Ontinuous change as the
temperature is increased.
We can obtain a consistcnt form al tilennoo)'namic theory of the behavior
of a ferroelectric crystal by considering the form of thc expansion of the e nergy
as a function of the polariwtion P. Wc assume that the LandHu8 frcc energy
density F in one dimension rna)' be expanded formall)' as
F(P;T,E) = -EP + go + i~p2 + ig4f>4 + ,~6 +... (37)
whe re the coefficients g" depend on the temperature .

7G Bums and E. Burste,n, F'"rroe\cctrics 7. 297 (i9i<l).

'In TP, see PI). 69 and 2:9S for a di~"'lon of the Landau (unction.
 13 Vid«,ri.,. "nd Ferrod"fdri<'. 40

, oro
I
I
;! "
I
,
- I
•
I
/ "
,

, I
I
I
I
I

l/ ,
,

j/' V I
I
,

'" , ..,
00
T.·"'"..·••lul~ . .. '"'
l'igure 13 I'JoI (If the: );Quare: of 1m: fll:(luenqo of the uro waVl!\'eclo r h'llrtSVene optic mode Ig;linst
te1npcnotuK. for S.Tt03• IU obse.nd in ne"tron d,rr...
.tclion experi ments by Cowl'!)·. The broken
]nll' is th~' redpmcil of the dicll,<:tric constant from the measuKnICuts of MilSu; and Westphal.

"
"
'f
s "
,f "
l
"
, ,., . ,
"" '" ".,IT - T.l." in K " "
f igure 14 Dt.>qL-asc: of ~ tr.mst'e:rsc: 111.0000 frell"""CY as the Curie terllperat ure Is app~hed
from bL·)o..... in tht- 1i.'fTOClectric crystal antimony ~u]pholodide . ShS I. (After Rama" scatt"rlng Cl(-
pcrime"h by c. II !'erry ""d D. K. Agrawal.)

nle series does not contain terms in odd powers of P if the unpohlrized
crystal has a center of inversion symmetry, but crystals arc known in whid} odd
powers arc impOItant. Power series expansions of the frec energy do nol always
exist, for non analytic tcrms arc known to OC"(.'\.Jr , especially when vcr)' ncar a
transition. For example. the transition in KII 2 PO" appears to have a logarith-
mic singularity in the heat C'dlx\City at the transition, wh ich is not claSSifiable as
either first or s<x'{)nd order.
""
The v<I)ue of P in the rmal e<luilibrium is give n by the mmimUffi of f: as a
function of P; the value of f. at this minimum defines the Ilellllholtz free e nergy
F(1',E ). The equilihri u m po larization in an applied clcc;trk: field E satisfi es the
extremu m cund ition
at
(38)

In this section we ass ume that the specime n is 11 long rod with the external
ilpplicd field E pumlle l to the long axis.
10 obtain a fc rrocicc:lric slate we must sup pose that the coefficie nt of the
ter m in p2 in (37) passes th ro ugh zero at some temperature 10:
(39)
where l' is take n as a positive constant and 10 may be equal to or lo\\c r than the
tmnsition te mperature. A s mall positive value of g2 means that the lanice is
"soft" and is ('Iusc to instability. A negative value of g2 means that the unpolar-
ized lattke is unstable. The variation of ~ with tempemture is ao:;o unted for by
the rlllal expan ~ ion and othc r e ITcc;ts of anharmonic lattice inte ractions.

Se<-"Qnd-Ortlc r Tra n sition

If g.. in (37) is positive, nothing new is added by thc term in l!J" and this
may then be neglected. The polaril:ltioll for zero applied electric; field is found
from (38):
(40)
so that either p. "" 0 or P; "" (yI~)( To - 'f). For '1' ~ 1'0 the only real rout of
(40) is at p. "" 0, bcc,luse l' and g , are positive. Thus '1'0 is the C urie te mpera-
ture. For T < 1'0 the minilllum ol"the I.andau free energy in zero applied field
is at
(41)
as plotted in F ig. 15. The phase transition is a second-orde r transition bcc;aosc
the polarization goes oontinuously to zero at the tnlnsition temp('r:l.ture. The
tnlllsi tion in LiTa0 3 is an example (Fig. 16) of a scconcl-order transition.

Jo'irst-Order Transition
The tra nsition is fi rst order if g., in (37) is ncgative. \'Ve must now retain gt;
and take it positive in ord er to rcs troin t from going to min us infinity (Fig. Ii).
The equilib rium cond ition for E "" 0 is given by (38):
,{1' - 'r ,)p, - 1",11') + g,;P~ ; 0 (42)
so tlJaI eithe r P~ = 0 or
(431

l
 -- ---------_-..,
",," '

13 Didec:tri.,. <lnd Fnrocln:lri...

r~ T)
P,{O)
"
"
"
" "
., m.- _ " n,
"
Figure 15 Spu<lI:lIll'OUS polarization Vt: rSus temJkralllrc, for a Sf.·(·oml- onlcr phase transItion .

"
•
'.,,
,
•,
,
,"
" •, ..,,,
•
'"
, •

Figure 16 l hnpt'..alure variation of the p<Jlar-axis .lallc dielectric w"st.Lnl of L..Ta~. (After
C lass .)
 (
T :> T,

Fiaure 11 Landau free energy f"nClio<! "ersus (pobrizalionJl in II r\l"lOt-ordcr I .."",ilion. at repre-
senlalhe temperatura. At T. the l.:orodau (mod;"" Ita,; <-'(I'lal miniona al P - 0 and at a finite P &!
wown . For Tbdow T, the .iliwIule minimum is at I;un.Tvalul'5 ofP; as Tpw.es Ihu."gh Tr lhLTt: is
II discontinUOUli change in the position of the at.olulL" mmimum. The arro""" mark the minima.

At the transition temperature T~ the free energies of the paraclectric and

fcrroclcdric phases will be equal. '£llat is, the value of f' for p. = 0 will be
C(IUal to the value of F at the minimum given b)' (43). In Fig. 18 we show the
chamclcristic variation with temperature ofP, for a first -order phase transition;
(:ontrast this with the variation shown in Fig. 15 for a SC<.'Ond-ordcr phase transi-
tion . Thc Imnsitio ll in BaTi03 is first order.
The dielectric constant is calcu lated from thc C!<lu ilibrium pobri7.ation in
an applied elcdlic field £ and is found from (38). In cquiliblium at tempera-
tures over the transition the terms in p4 and p6 may be negleeted; thus
E ::;: y(T - Jo)P, or
(eeS) t:(1' > J'J = I + 4'1TPI£ "" J + 4rr1){T - 'JO) . (44)
of the form of (36), 111e result applics whcther thc transition is of the first or
sceond order, but if second order we have To ::;: '/~~ iffirst ordcr, thcn To < Te.
Equation (39) defines 'fo, but Te is thc transition temperature.

Antiferroelectricity
A fcrroelectric displacemcnt is not thc only t)1>C of instabilily that may
develop in a dielectric Crystal. Othcr deformations oa."ur, as in Fig. 19. "mcse
dcrormations, evcn if thc)' do not give a spontancous polarization, may be
accompanied by changes in the dielectric (."onstant . Onc type of defOrmation is
callcd antiferroelectric and bas neighboring lines of ions displM."cd in opposite
scnses, Thc pcrovskite structure appears to be su!>(."cptiblc to many types of
deformation , often with littlc differcn<."C in e nergy bc,twecn them . The phase
 -
'"
, OJ t----
v

00
Figure 18 Calculated values of !l,e spontaneous ..:
polarization as a runct;on or
temperature , with
par-uncten as ror barium titanate. (After
w. Cochran .)
0
-., - 00 -., -ro
T - T, .nd.: g C -
.
T > T, T < T, T < T,

EBEEJ -
EBEn
I @I @1'1@ 1@ I 1@1@ I
I @I @I 1@ 1@ I I @1 @1
I @I @I 1@ I @
11@I @1
I @I @I 1@ I @1 I @1 @1

Figme In Schemat ic representation of fu nd.o.me ota\ trIleS (If ~tnJCtura' pha5e tr,m,; ;hons from a
cenlJ"OSrmmdric prototypp. (ARer Uncs and C'as~.)
 ". Table 3 Antifcn'OCll:ctric cl"},slals

Tran"(1Qn I~mpel':ll"r" ID
Crysloll anlif"rrocl<'d.r;" slate, in K

\V03 1010
NaNbO a 793, 9 11
PbZrO,l 506
PbHf03 488
NII~H 2 PO. 148
ND t D2PO. 242
NII.I ' 2AIiO, 21.
N04DI!A~Ot 304
(N1I,hi4I0e 254

FTOm a compilation by Walter J. Men.

diagrams of mixed pero~kite systems, such as the PbZr03- PbTi~ system,

show transitions between para·, felT()o, and antiferroelcctric states (Fig. 20),
Ordered antiferroclectric arrangements of permanent electric dipole
moments occur at low temperatures i n amm on iUIll salts and in hydrogen ha-
lides. Several crystals believed to have an ordered nonpolar state are lisled in
Table 3.

Ferroeleclric Domains
Consider a ferroelectric crystaJ (such as barium titanate in the tetragonal
pbase) in which the spontaneous polaril'..ation may be eithe r up or down the c
axis of the crystal. A ferroelectric crystal generally <.'Onsists of regions called
domains within each of which the polarization is in the same direction, but in
adjacent domains the polari:tA'Ition is in different directions. In Fig. 21 the polar-
ization is in opposite directions. The net polari:tlltion depends on the difference
in the volumes of the upv.>ard- and downward-directed domains. TIle crystal as
a whole will appear to be un polarized. as measured by the cbarge on electrodes
covering the e nd s, when the volumes of domains in opposite senses are eCJual.
The total dipole moment of the crystal may be changed by the movement of the
walls between domains or by the nucleation of new domains.
Figure 22 is a series of photomicrographs of a single crystal of barium
titanate in an electric field normal to the plane of the photographs and p.'1rallcl
to the tetmgonal a'l:is. 111e dosed curves are boundaries bet"..ecn dOlllains po-
iaril'.Cd into :md out of the plane of the photographs.
The domain boundaries change size and shape when the intensity of the
electric fleld is altered. The motion of domain walls in fe rroelcctrics is not

\
/ .. -
 ..
""
~

,w
E, ~

~ 'w
~ ,., n :,
"
" ,~
~."'''''

11,1.00,
w
"" ".11>1"10,

Figt're 20 F'cuock'dnc F, KntifcrroclcctricA, anclllanIC'~'C.·tr;" P phases ofthc 1t.'"MI 7.ir<:Qnat~~ Ic;Jd

titanate solid $OI"lion S}S ICffi. The subscript T dcnotn II Icl",gon~1 phas<:. C a "ubic 1l1WlSt: ; R II
rhomOOh ...dral ,.haw, of "hid. there are high temperat" ' e (111) and low temperature (U ) IOrms.
N..ar the thombohl>dral-tctragOlla ' pl~ boun&.ric."S one finds ''''ry high piewclcclrk COI.pl ing
rodJ'. dcnl$. (After Jafft,.)

• + • • • • +
+ +

• +
• • • • +
+ +

• + • • • • +
+ + I ""
• •I, • •
• (>,

Figure 21 (a) $dlc:malic dnowing of atomic dl.placements on ",ther ~idc or

a boundary Ix:twt.'t!'n
domains pobri:r.t.-d in QIl~itc dirertions in I k rrock:dric crystal; (h) "jew of a domain SIru('b'""
~hll....ing 1110" m'lld... ico between domains polari7.cd in O(Ipt..sllc directions.
Figure 22 F~..roel~'ctric doona'ns On th" rae:.: ri a ~ngl" crystal or !Jar."," "Iaolal". 11." rare is
normal 10 Ih" Idragmal or C Ill<i$. 111e nel pola.iulion ri Iht crystal as judged by domain ,"Olume.s is
illC.,...'iISCd markedly IS the electric neld Intensity ~t1ello the axis is ina"easc.-d (rom 550 ,-o!W"m
to 980 Vlcm . '11C domain IlOUndark""S arc made ,;siLlc by eldling Ih" "')'Stili in a weak acid wlu-
tioo (R. C . Mllk... .)

simple: it is knQ\.1l11 9 that in all ele<.1ric fleld a 180" wall in BaTi03 appears to
move by the repeated nucleation of steps by thermal fiuduations along the
-parent wall. A "lSOO wall" is the boundary between regions having opposite
poL-triz..1.tion dircdions.

Piezoelectricity
All crystals in a ferroelectric state arc also piezoelectric: a stress Z applied
to the crystal will change the cledric polarinttion (Fig. 23). SimiL"lriy, an elec-
tric fleld E npplied to the crystal will cause the crystal to l:>Cl..'Ome stmined. III
schematic one-dimellsionrunotntion , the piezoelectric equations are
(CGS) P = Zd+EX; e = Zs+Ed. (45)
where P is the polarirntion, Z the stress, d the pie-.welectric strain constant, E
the electric field, X the dieledric susceptibility, e· the clastic strain, and s the
elastic compliance constant. To obtain (45) ill Sl, replace X by E"uX. These rela-
tions exhibit the development of polari7Alion by an applied strcss and the
development of clastic strain by an applied electric field .
A crystal may be piczoelectric without being ferroelectric: a schematic
examplc of such a stmc.1ure is given in Fig. 24. Quartz is piczoclectric. but no t
ferroelectric; bnrium titannte is both . For order of mngnitude. in quartz

Uw. J. Merl., Phys. Re\.. 95, 690 (1954); R L..nd~llCr, J. Appl Ph)·s. 28, 227 (1957); R C.
Millet and A. S.\"~ge, Ph)"S. Rt....•. I IS, 1176 (1959). For Ih" th<X>ry ofth" thidrnt."$~ of domain wall.
in fcrroclt.'<;"t,;"' , It."(.' C . Killel , Solid Slate Comm. 10, 1)9 (1972) and rcfercno.."el cilt...t th t.TC.
 ,t • • 1,
j
• •
Figure 23 (a) Unstn,sscd r"nock-ct"" 0')"'513\ amI (h) slrcs ...'tl rerroclc.;,tric crystal. The sirCH
d,M,.gCS the l'ol~ri7J1tion hy l!.P. the ind,,(.'(.') pie;wck.'<.1ric pol~~lion.

w (I.,
Figure 24 (a) -n,e ul1strcs.s<.,) cr~'slal has II thrcl·fold .yn nnciry meis. 11,., arrow> rcpr ncnt dipole
'TIOfr>ents, cooch sci ~Ihree arrow5 rept'C1i<-'tlh "planar group of ion. denoted by N 8 3 , " 'itll" B~­
ion at each \'Crtc". The MIIIl oflhe thn-cdipole moments ~t cao:h "crtc~ is zero. (h) Thcl'f)'sl..! when
strt.~....-d t1t.,\·c!op$ II pol"rizalion In the d,rt.oction indk-~h:d. llu: sum of the dipole moments about
~':ICh "ctlCX i.s no longer zero.

d - 10- 7 cm/statvolt and in barium titanate cf "'" 10- 5 cm/statvolt. 111C general
dcfinition lO of tbe piezoclectric slmin COllstauts is
(46)
where i - x, y, z and k - xx, yy, 2:7;, !/Z, :tX, xy. To convert to em/stat V from
values of d;Io; given in mIV. multiply by 3 X 10".
llle lead zir<:Ollate-lcad titanate system (rolled the P'l:T system), Fig. 20,
is widely used in polycrystalline (ceramic) ronn with compositions or vcry high
piezoelectric <.:Oupling. The synthetic polymer polyvinylidcnfluoride ( PVF~ is
five times more strongly piezoelectric than crystalline quart"L. Thin stretched
films or PVF2 arc fle>Lib le and arc easy to handle as ultrasonic tmnsdu(:ers.
Being flexible. the film s arc easy to handle in mcdirol applications to monitor
blood pressure and respiration.

'''111e pkzocl<:etric 'train constants form a thi"']·nUlk ten..,... It. ~isc treatlllent of the effc:ct
of symmetries on the ten..,.. propertit.-$ of Cl')stak is gi\=;n the A./ P Handbook,3rd ed .. pp-
9-10 to 9-15, Sl.'Cl also II. J. JUfct:;c:kc, Cryslal physiu -macro.rcop;c 11hl/i":J 0/ anisotropic ,oli,b .
llcnjamin. 1974
""
The response ofpie:r..oeleclric crystals in transduecr applications is charac-
terize{! by the eledromcchanical ooupling factor k, whose square is defined by

".2 = mechanical energy stored . (47)

electric energy stored
The result can be expressed in terms of tile material (''Onstanls used in (45). The
relation is given in works on the applit:ations of pie7..0electric crystals.

Ferroelasticity
A crystal is fcrroclastic if it has two or morc stable orientation states in
the absence of mechanical stress and if it can be reproducibly transfonncd from
one to another of these states by application of mcchanica1 stress. Healialtions
are few: gadolinium molybdate CdiMo0 4 h has a combinc(\ fcrrock'clne-
ferroelastic trnnsition; lend phosphate Pb:J(PO' I)2 has a pure fcrroelastic transi-
tion; severnl crystals have combined ferroelC<..iric-ferroelastic transitions.

Opticlll Cerllmics
The addition ofkmthanum to the PZT system gives a reasonably tmnspar-
ent product with useful eledro-optical properties. The PLZT system lias been
used as the basis of optical memories.

SUMMARY
(In CGS Units)

• The electriC field average(l over tile volume of the specimen defines the
macroscopic e1cc1ne field E of the Maxwell equations.
• The eledric field that ac.1s at the site rJ of an atom j is the loca1 electric
field, E loc• It is a sum over all dlurges, grouped in terms as
EIoc(rj) = Eo + EI + E2 + E:J(rj), where only E3 varies rapidly within a c.-ell.
Here;
Eo = external electric Geld;
EI = depolari~"ltion Geld associated with the boundary of the specimen;
~ = Geld from polarization outside a sphere centered about Ij;
E3(rA = Geld at rj due to all atoms inside the sphere.
• The macr05COpie field E of the Maxwell equations is equal to Eo + EJ>
which, in gener.lJ, is not equal to E..",(rj).
• The depolarization field in an ellipsoid is E I ,.. = -N,.."P", where N,.." is the
depolari,..ation tensor; the polarization P is the dipole moment per unit vol-
ume. In a sphere N = 477"/3.
(
....
,
-
• TIle Lorentz fleld is Ez = 411"1'13.
• '111e polarizability 0' of an atom is deflned in terms of the local eiC(:tric field as
p = aE\nc.

• TIle dielectric susceptibility X and diclectric {:onstant E arc defined in terms of

the macroscopic elct.1ric field E as 0 "" E + 4nP = EE =
(1 + 411"x)E, or
X = PIE. III S I, we have X = P!¥.
• At atom at a site with cubic symmetry has Eloc E + (4m3)P and satisfies
the Clausius-MossotH relation (24).

Problems
I. Polariwbility of atomic I,ydrogen. Consider a semiclassic-.t1 model of the ground
state of th e hydrogtm alom in an electric field nonnal to the plane of the orbit
(Fig. 25), and Sl10W th;d for this I1'lOlIci a "" at,.
wllere UII b the radius of the unper-
turbed orbit. Note; If the applied field is in the JC dirL-eiion, then the JC compom::nl of
the field of the nucleus at the displac.,(:d position of the ei{'Ctrol1 orbit mus t be C<luaJ
to tile applied field . The CQrrect quantum. mechanical result is larger than this by
the factor I. (We are speaking ci au il1 the exp;lnsioll a .. 0'0 + ali:: + . .) We
"*
assume x fill ' One can also m1cuL1tc 0'1 on this model ,

Figure 25 An electron in 1\ dra, lar omit of m·

dius 01/ is displa..'(.'(t a distance JC on applkatiOll

of an electric nekl '" in the -x direcl1on. n le

fon:-e on the dcaron due to the nudeus;s " 1":,
in CCS ~ rI4'Jtf!oI1f, ;n SI The problem as-
sumes x <Ii a".
 (
2. Polariwbility of conducting 6JJllere. Show thai the polnrixal y of a C'Oflducting
metallic sphere of r:Wius 11 is 0" = cPoThis result is most easily ohtaill(.'f.! by noting
that E - 0 insiile Ihe sphere and then using the depolarization &etor 4m3 fur a
$phcrc (Fig. 26). TIle res"lt gives \".llues of a of the order of magnitude of the
observed polarizabilitics of atoms. A lattice ofN conducting spheres per unit volume
ha$ i1icit..'('lric: coostanl ( _ I + 4'ITNa 3 , for Nrr
<c I. TIle sU&gt.'s tcd proportionali ty
of a to the wlx: of the ionic radius is !k'\lisfk-d (Iuite wel l fOl' alkali :\,1(! halogen iol1s.
To do thc problem in 51, usc! as tlie depolarization f,lClol'.

3. Effect ofair gap. Oiscuss the effect of an air gap (Fig. 27) bct Wt.'CI1 condenser plates
and dielectric on the measurement of high didcc:tric C()n~tan l s. What is the highest
upparcnt dicicl;:tric ct)llslanl possible if the air gnp thickness is 10- 3 of th e Iota!
Ihickncss? The pn.>SC1'lCe of air gaps can seriously dislort Ihe mt.'llSurcment of high
dielectric comlants.

4. InterfaciDl pvlarization. Show lhal a jXlrallcl-platc capacilor mooe up ofn..u paral-

tclla)'Cr! rL malcrial---()fle layer with diclt.-clric constant t , 7.cTO conductivity, and
thickncss d, and the other la)'cr with £ . , 0 lOr convcnience, flnite conductivity u,
and thickness qd- behavcs as if the space between the condenser plates wcre fllled
wilh a homogeneou$ dicle<:lric with dicledric con$tant

(eGS)
, £"ff =
t(l + q)
-;-l-,c~(ic,wq
=I;;'=.::;
.)

where w is the angular frequency. Values of fdFa5 high as 10" or 1(iI cauS(.-d largely
by this Max .....ell-Wagncr mCdlanism arc sometimes round, but the high values arc
alWdYS accompollit..(! by large ac losses.

5. Polarization of tlp1wre. A sphcre of dieit.'Ctric conSlant t is plact.'\! in a uniform

eX lemai dt.'Ctric field Eo. (a) What is ttIC volume aver.lgc electric field I:: in the
spherei' (b) Show that the polari~1lion in the sphere i~ P - xM I + (4wXI3)),
where X - (f - 1)1411. Hint: You do not nt.«.'\llo calculate t::k><: in this problem; in
fact it is CQIlfuSillg to do so, lJ.e.cause f and X are defined so that I' - Xl::. 'Vc require
Eo 10 be unchanged by insertion ofthe sphere. \Ve CIIn produce a fi~ed Eo by placing
positive ('harges on OTIC thin plate of an insulator and nt.'glItl\'c charges on an oppo-
~ite plate. If the plales arc always far from the sphere. the field of the plates will
remain unchanged when the sphere is inserted bctw(.'(.'fI them . TIle results above
are in CGS.

6. Fer,.odedriceriterionfo,. a'onlf. Consider a system ofl"o'O neutral atoms S(.-parated

by a fixed distance a, each atom ha\ ing a po!arizabilit), a . Find the rclalion between
a and a lOr such a system to be felTOClcdric. Hint : 'The dipolar Reid is stl'OTlg~st
along the axis of the dijXlle.

7. Scturation poiari=tioll at Curie point. In a firsl-()rder tr,lOsition the C<luilibrium

condition (-'13) with T sd C<luai Tc gives one <.'<juation for the jXllarization P.(Tc). A
 . " ""
13 lJiell'Clrin ancl Fl'nod«lria 413

, , ,", , , I
I,
I

Figure 26 "Ole total field inside Ii conduding sphere is n:ro. If ill flCld F~ is applied extcrTllIlly . then
thl' flC)d E, duc to lUrfa..'C d,arg<.'S 00 tile sphere Hlust just L;I'X'CI Y.... SO that E o + E, - 0 within
the SillIer ... But E, CWl he simulatt:d hy ti le dqlOl .... i1.>t1lon fleld - 4.".P13 or II. ullirormiy polarin..!
sphere of poiari7..alioo P. Relate P to E,:, and cak:ulatc the dipolc mon,..,,! p of the ~phcrc. In 51th"
depolarization field is - PI~

,-
r ,
"
,. ,
"A. -\ir

"
Figure 27 An air gIIll ofthkkncss qd is in ""ri~ in a OIpacilor wilh Ii diek...1rie sial" ofthicl:ness d.

further condition at the enrie point is that f.·(P., Te) = 1'(0, Te). (1\) ComLining
these two conJili~lIs , show IlulI r:<Tc) = 31g.V4g. (b) Using this rC51111, show that
Tc"" To + 3~fI6~

8. Dielectric emulant beL:no .ramilian temperature. In ter ms of the parllmctcrs in the

Landau frt.'C energy cxpnnsioo, show that lOr a second-order phase transition the
dielectric ('Onstant Ix:low the tr.lru;ition temperature is

t" = I + 41'1'6P1l:: '"' I + 2Trl;(:fc - 'I) .

11.is result may oc oomp.lreU with (44) above the tranSition.
9. Soft moot'S and intlice 'ransjornwlioM. Sketch a monatomi<: linear lattice oflalticc
constant u. (a) Add to l.-ach of six atoms a vector to indicate the llirt'<:tion of the
displacement at a given time causcO uy
a Iongi}ooinal phonon with \\'a\'l.....·cctor at
the zone boundary, (b) Sketch the crystal stn,clure that results if this zone boundary
...
phonon oc'Comt'S unstable (w--+ 0) a.~ the crystal is (,'OOlcu through 'fe> (e) Sketch on
one gnlph the essential aspt.ds orlhe longltmlinal phonon dispersion relation for the
monatomic lattice at T ....-ell aoovc Tc and at T - Te. Atld to the gro{lh the salTlC
iuformation ror phooons in the new ~truct"rc at T well below Te>

10. Fern~led,.ic linftlr array. ConsiuCf" a Hoe of atoms of poIariUlbility a and separa·
lion u. Show that the array can jXllarizc spontant'Ously if a ~ il/4I,,-3. where the
sum is over all po.o;ilivc inlt:gcrs and is given in tables as 1.202.

References
A. D . BnlCe ~nd R. A. Cowley. St.Ilcrllmi pi""", tru.uitiO'I1. ';')'10< & Francis, 1961 .
H. Frohlich. 1'hc<ny "I dieit'drlc4, O~for'lI95&. 19S6.
E. Fal"~.7.o ali(I W. J. t.h.'17., F.,-roc!tctrk;ify. North· tlolland. 1967.
K. A. Mioner I nd H . llIomrul, cds., Sln,cturai ph ... tran.ti/io"", Springer, 1981.
M. E. Unes Mnd A. M. Glass. f'crrorlrctriQ a.,J rdoted IIl11leru./" Oxford. 1977.
B. Jaffe, W. R. Cook, Jr.. and II . JaIIi:. l'ir...Df:IKtric uremic-, AUldcmic, 1971.
F"rrodectriu. A Journal.
C. A. S;lUlUOI.and p. S. Peercy, .. Study or sort·mod" transitions at Iligh pressure, " Solid lOMe physics
36, I (1 98 1).
1.-C. Toledano and P. Tolcdano , Lanrluu IJ.roryo!pI_ tranntion,: U/)plicul/orl lOs/rue/urol, Inrom.
mCnSUN/I", mog.u'tic, unrlliqultf crv~lol ""Js/ellts. World Scientific, 1987.

£. ,
 14
Diamagnetism and Paramagnetism

LANGEVIN DIAMAGNETISM EQUATION 417

QUANTUM THEORY OF DIAMAGNETISM OF

MONONUCLEAR SYSTEMS 419
PARAMAGNETISM 420

QUANTUM THEORY OF PARAMAGNETISM 420

Rarc carth ions 423
Hund 1'Ulcs 424
lrlm group ions 425
Crysbl field splitting 420
Quenching of the orbital angular momentum 426
SpectrOl;copic splitting factor 429
Van Vleck temperature-independent paramagnetism 4JO
COQUNG BY ISENTROPIC DEMAGNETIZATION 431
Nuclear demagnetization 432

PAHAMAGNt.- nc SUSCEYflBILITY OF
CONDucnON ELECfnONS 433
SUMMARY 436

PROBl,EMS 436
1. Diamagnetic $UliC'eptibility of atomic hydn)gen 436
2. Hund l'Ules 431
3. Tliplet e:o;citcd states 437
4. Ileal capacity f."Om internal degrees of freedom 438
5. Pauli spin susc:eptibility 438
6. Conduction elech'On ferromagne tism 438
7. Two-Ie"el s)'Slcm 440
8. Paramagnetism of S = 1 system 440

REFERENCES 440

NOTATION. In th c problcms trcatc(l in tliis cliaptcr the magnetic field B is always

closely equal to the applied field Ba. so that we write B ror Ba in most instances.
(
 (

La.W'vin (r.C'e 'lllnj

panmagnctum

..I ------------------------------_.
-------y~~~!!"----
,I-----~~'-=====~~--~===
I ___________""_~_,.....
_____._t:m__(_~_"'______ T~~~
DtanugnHlOm

Figure I Charac1.eristic magrK'lJc susccpl ibilihes uf di~lnl.gnctjc am) paramagnct~ s"hstances.

• 16
 CIiAPTEI. 4: DIAMAGNETISM AND I'ARAMAGNETISM

Magnetism is inseparable from Quantum mechanics, for a strictl y c1assic.11

S)'stem in thermal equilibrium can display 110 magnetic moment, e\cn in a
Inagnetic field . 111C magnetic moment of a free atom has thrcc Plint:ipal
sources: the spin wilh which ck'CI:rons are endowed; their orbital angular mo-
mentum abo ut the Ilucleus, and the change in the orbital moment induced by
an applied m agnetic field .
'111C fi rst two elTects give p.1ramaguetic contributions to the magnetization,
and the third gives a diamagnetic contribution. In the ground Is state of the
hydrogen atpm the orbital moment is zero, and the magnetic moment is that of
the electron spin along with a small induced diamagnctic moment. In the 1s2
state of helium the sp in and orbital moments are both zero, and there is onl y an
indul.-'Cd moment. Atoms with filled elcctron shells have zero spin and zero
orbital momcnt : these moments are associated with unfilled shells.
'111e magnetization M is defined a.~ tllc magnetic moment per unit volume.
111e magnetic susceptibility per unit volume is defined as
,
M
X~B' (51) (I)
ICGS) X~--

""'"
B

where B is the macroscopic magnetic field intensity. In both systems of units X

is dimen sionless. \ \le shall somct im !~s for convenience rclcr to MIB as the sus-
c(.'ptibility without specifyi ng the syste m of ullits.
Quite frequently a susceptibility is defined referred to unit mass or to a
mole of the suhstance. 11lC molar susceptibility is written as XM; the magnetic
moment per gram is sometim es written as (T. Silbstanl.-'CS with a negative mag-
netic SUSCl.-l 1tibility are called diamagnetic. Substanl.-'CS w ith a positive suscepti-
bility are called paramagnetic. as in Fig. 1.
Ordered arrays of maRnetic moments arc discussed in Chapter 15, the
arrays may be ferromagnetic, ferrimagnetic, an tife rromagnetic, helical, or
more colllplex in fonn. Nuclear magnetic moments give rise to nudear
para magnetism . Magnetic moments of nuclei arc of the order of 10- 3 times
smaller Ih.1Il the magnetic moment of the electron.

LANGEVIN DIAMAGNETISM EQUATION

Diamagnetism is associated with the tendency of clectlical charges par-

tially to shield the intcJior of a body from an applied magnetic ficld . In electro-
magnetism we arc liul1i1iar with Lcnz's law: when the !lux through an electrical
circuit is changed , an IIlduced Cllrrent is set up in such a uircction as to oppose
the nux change.

m
 In a supel't.'Onductor or in an electron orbit within an atom , thc induC(.od
CUO'Cllt pcrsists as long as the field is prescnt. Thc magnctic field of tlte induced
current is opposite to the applied field, and the magnetic moment associated
with the current is a diamagnetic moment. Even in a normal metal there is a
diamagnetic contribution from thc conduction electrons, and this diamag-
netism is not destroyed by collisions of the ek"CIrons.
'111c usual trcatment of the diamagnetism of atoms and ions employs the
Larmor thcorcm: in a magnetic lIeld the motion of the clectrons around a
ccntral nllclcus is, to tIle first order in B, tbe same as a pos."ible motion in the
absencc of B except for the superposition of a precession of the electrons with
angular frequency
(CGS) w :: eB/2111c (SI) w = eB/2m (2)

If thc field is applied slowly, the 1lI0tion in the rotating rcference system will be
the same as the original motion in the rest system before the applk.'tltion of the
field.
If the average electron current around the nucleus is zero initially, the
appliCfltion of the magnetic field will causc a finite cu rrent around the nu-
cleus. "J1IC currcn t is equivalent to a magnetic moment oppositc to thc applied
field . It is assumed that thc Lamlor frCt]lIcn<.), (2) is much lower than thc fl"c-
qucn<.)' of the original motion in the cen tral field . °l11is condition is not satisfied
in free Cflrrier cyclotron resonance, and the cyclotron frequency is twice the
frequency (2).
ollIe Larmor precession ofz electrons is C(luivaien t to an electric current

(SI) 1 = (charge){revolutions per unit lime) "" (-Ze{--~-' en) , (3)

.,.;.00 2"11" 2m ...J

lllC magnetic moment po of a current loop is given by the product

(curreot) X (area of the loop). 'Ille area of the loop of radius p is 1T,r. We have

Z<'B Z<'B
(SI) " . - .-Ip') (CGS) ~ = - -- Ip') (4)
'on 4m~

Here vr) ;; (r) + <if) is thc nll..'tln square of the perpendicular distance of the
electron from the field axis through the nucleus. 'I1le mean square distancc of
thc ele<..1rons from the oucleus is (,.2) ;; (,?) + (y2) + (r). For a spherically
symmetrical distribu tion of charge we have (.r) "" (if) "" (z2), so lilal (,-2) :=
ilp').
From (4) the diamagnetic sllsl,;eplibilily per unit volume is, if N is the
number of atoms per unil volume,

(CCS) (5)

.~.~,--~-- •.
(51)

This is the classical Ulngevin result.

111C problem of calculating the diamagnetic Sllsccptibility of fin isolated
alom is reduced 10 the (.'l\\culalion or(r~ for the electron distrihution within the
atom. The distribution am be calculated by quantum mechanics.
Experimental values for neutral atoms arc most easily obtained for the
inert gases. Typical cKpcriment,,1 vflllIes of the molar susceptibilities arc the
following:

/I, N, K, x,
X.II in CGS in 10- 11 {.",3/nlOlc: - 1.9 - 1.2 - 19.4 - 28.0 - 43.0

In dielectric solids the diamagnetic contribution of the ion cores is de-

scribed roughly by the Langevin rcsult. 'Inc contribution of conduction elec-
trons is more oomplicatct!. as is evident from the de I-Iaas-van Alphen effect
disl:tlsscd in Chapler 9.

QUANTUM TI-IEOHY OF DIAMAGNETISM OF MONONUCLEAR SYSTEMS

From (G. 18) the effect of a magnetic field is to add to the ham iltonian the
tcnns
ieh rr
,r = -2mc
( V · A + A . V) + - -A'
2mC'
(6)

for an atom ic c1ed:roll these terms may usually be treated as a small perturba-
tion. If the magnetic field is uniform and in the:;:: direction, we rna)' write
A~:;; -iyB (7)
and (6) becomes

J{'
iehB (a
= - - x- - y-
a) + -2B'
-(r + y~ (8)
2mc Otj in BmCZ
·nle first term on the right is proportional to the orbital angular mOlllen-
tum component L~ if r is measured from the nucleus. In mononuclear s)'stems
this term gives rise only to paramagnetism. 111e second term gives for a spheri-
caJly symmetric system a contribution

{r~
2
E' = rrB (9)
12m<f
by first-order perturbation theory. 'n1e assodatoo magnet~ ,ment is diamag_
netic:
aE' <,<,~
p. = - - = - - -8 (10)
iJB 6mc2
ill agreement with the classical result.

PARAMAGNETISM

Electronic paramagnetism (positive contribution to X) is found in:

L Atoms, molecules, and lattice defects possessing Rn odd number of
electrons, as here the total spill of the system cannot be 7.cro. EKsmplcs: free
sodium atoms; gaseous nitric oxide (NO); OI'gRnic frcc radk'8.ls such as
triphenylmethyl, C(4H~h; ,; ccnters in alkali halides.
2. Frcc atoms and ions with a partly filled inner shell: transition clements;
ions isoclectronic with transition elements; rare carth and actinide clements.
Examples: Mn2 +, Gd 3 +, U H . Paramagnetism is exhibited by many ofthesc
ions even when incorporat<.-d into solids, but not illvariallly.
3. A few compounds with an even number of electrons, including molecu-
lar oxygen and organic biradicals.
4. Metals.

QUANTUM THEORY OF PARAMAGNETISM

The magnetic moment of all atom or ion in free spat.'C is given by

(ll)

where the total angular momentum fl.J is the slim of the orbital ilL and spin liS
angular momenta.
The constant yis the ratio of the magnetic moment to the angular momen-
tum; y is called the gyromagnetic ratio or magnetogyrie ratio. For electronic
systems a quantity g called the g f."lCtor or the spectroscopic splitting factor is
defined by
fJl.1l • -yt, . (12)

For an electron spin g = 2.0023, usun1ly taken as 2.00. For a free atom the g
factor is given by the Lande equation

g=l+ -,,-](~J~+~I )~+_S?(~

S ~+~I~)_-~~~L~+
~ I) (13)
2J(] + 1)--
 t 'igure 2 Enetl.'Y lewl spl itt illg for one electron
in R magnet;., field B dircdcd along the l)(J§iti\'C .r;
axis, For an e\cetron the " "'I;rK,t k IIIOIIlCflt ~ is
opposite III sign 10 the spin S, SO that ~ ­
- W"S. In Ihe low energy state tire magnetic
moment is parallcl to the magnetic field .
F~'llrc J Fradionai poJlO.,lations of a two,Ie.'Ci
s~ . t ~.,..
in t1 ...Tmai equilihrium at tClopel"llture T
iu a masnc hc field H. Th .. magnetic moment is
proportional to the diITen.-nre betw~en the tv."O
C\' ....'CIi.

Tne Boltr magneton J.l.H is defined as efJlr11(; in CGS and dil2m in SI. It is
closely equal to the spin magnetic moment of a free eloc1ron.
l11e energy levels of the system in a magnetic field are
(14)
where 711} is the azil11uthalquantum numbel" and has the values J, J - I, '
- J. For a Single spin with no orbital moment we have mj :: ::!: i and g = 2,
whence U = :!: J.l. H13. 'n lis splitting is shm\n in F ig. 2.
If a system has only two levels the equilibrium populations are, with
"T - kfjT,
N, cJ(pVL13h)
: (IS)
N eXpVLBI"T) + exp( J.l.1J17)
N, exp(- J.l.Bh-)
: (16)
N eXIl{J.i.W"T) + cxp( J.l.BI"T)
here N ., N2 arc the populations of the lower and upper levels, and
N :: NI + Nt. is the total number of atoms. TIle fractional populations arc p lot-
ted in Fig. 3.
The projection of the magnetic moment of the upper state along the field
dil"edioo is - J.I. and of the Im\'el" slate is J.I.. TIle resultant magne timUon for N
atoms per unit volume is, with x - J.l.Bl kHT,
e" - e- X
M = (N ) - N2)J.I. "" NJ.I. . = NJ.I. tanh x . (17)
e' +e %

l~or x ...: 1, tanh x "ill x, and we have

(18)
In a magnetic field An atom with angular momentum qumltum numbel" J
has 2J + 1 equally spaced energy levels. lllC maglletization (Fig. 4) is given by
(19)

-
 , ,
"" /~ ,--,, (~")
'" -
,.,
l
/ $ .. t {fc-" ,
"
~ } v'"
I-

~ I s- ! , (Cr" ) ,
J V :
...-

,.,
,,1 _
0

•
•
0
'''''
tOOK
3., ,
4 21 K
1\"IIoo;n .........11'.... -
=

."
I I I
" "
BIT In kC.,.. I
30

Figure... PIoI: of magnetic moment vt'r.I'lI$ 8ff lOr sphcnOlI samples o£ (I) potas<ium chn.mnurn
alum, (II) rerric ammoo;um alum. IW1d (III) gadoImium ,u[r"le octal'ydratc. O\'Cr 99.5% rnagllctic
saturation is achin'oo al 1.3 K and about 50,000 gauss. (51). After W. E. lIenry.

wllc rt! Ihe Brillouin funt'tiOll 13J is defined by

BAx) :::: ~ clnh

2J+1 (2J+I)X
2J
)- I
2J clnh 2)
(X) (20)

EqUAtion (17) is a special (.'OlSC of (20) for J = •.

For x: <:t I , we have
I x i'
ctnhx :::: - +- - - + (21)
:t 3 45
and the susceptibility is
M C
- ~ (22)
B T
Here p is the effective number of Bohr magnctons, defined as
') - gfj{J + l)Jlf2 . (23)
(
....
 .. !

""l-
I
II
, '"
V
I
"
,
,,
/
Figure S Plot oil/x u T (01" a g:odoI inimn salt, Cd(c..1I,.50J, · 911.0. nit' ltraighl line i. the
Curie la..... (Aftt'T L. C. Jackson and H. Ka~rlint:h Onncs.)

The constant C is koown as the Curie conshlllt. '111c form (19) is known M the
CUlie-Brlllouin law. and (22) is known as the Curie law. Results for the para-
magnetic ions in a gadolinium sa1t are shown in Fig. 5.

Rare Earth Ions

The ions of the rare-earth elements have closely similar chemical proper-
ties, and their dlcmicaJ separation in tolerably pure form was fI<:complishcd
onl y long afte r the ir discovery. 111cir magnetic properties arc fascinating: the
ions CKhibit n systematic variety and intelligihle complexity. '111c chemical
properties of lhc trivalent ions arc similar because the outcrmml electron shells
arc identically in the S,ilsp6 configuration, like ncutml xenon. III lanthanum.
just before the rare earth group begins. the 4/shell is e mpty; at cerium there is
one 4/ electron. and the number of 4/ electrons increases steadily through the
group until we have 4/,3 al ytterbium and the filled shell4p"' at lutecium. The
radii of the trivruent ions contract fairly smoothly as we go through the group
from L 11 A. "t ceri um to 0.94 A. at rtterhium. This is the ramous " Ianthanide
cont"'ction." What distinguishes the magnetic behavior of one ion species from
anothcr is the numhcrof 4f electrons comp3<..1ed in the inllcr shell with a ",dius

\
"..
{
of perhaps 0.3 A. Even in the metals the 4/ core re\ s its integrity fllld
its atomic properties: no other group of elements in the pcliodic table is as
interesting.
The pre<-'Cding discussion of paramagnetism applies to aloms that have a
(2J + I).fold degenerate ground state, the dcgcnctlIc), being lined by a mag_
ndk field . The influence of all higher ellergy slates of the system is Ilcglct.'l:ed.
111csc as.~tmlplioIiS appear 10 Le satisfied by a !lumber of rare-earth ions,
Table 1. 111C <''8.lculatcd 1118&IIC101l numbers arc obtained with g valucs from the
Lande result (13) and the ground-state Icvel llSsignmcnt predicted by the J lund
theory of spct..1.ra1terms. The discrepancy between the cxpcriml.'1ltaJ magnclon
numbers and those calculated on these assumptions is quite marked for Eu3 +
and 5m3 + ions. For these ions it is necessary to consider the influence of the
bigh states of the L-S multiplet , as the intervals between successive states of
the multiplet arc not large compared to kilT at room temperature. A multiplet
is the set of levcJs of different} values arising out of A given Land S. The levels
of n multiplet are split by the spin-orbit interm.otion.

IItJnd Rilles
'111e Hund nIles as appJi<.'<i to electrons in a given shell of~n atom affiml
that electrons will occupy orbitals in such a way that the ground state is c11arnc-
terized by the follOWing:

1. '111e maximum value of the total spin S allowed by the exclusion

principle;
2. 111e maximum value of the orbital angular momentum Lconsistent with
this value of S;
3. The value of the total angular momentum} is equaJ to IL - 51 whcn the
shell is less than half full and to L + S when the shell is more than half full .
'\'hen the shell is just half full, the application of the first rule gives L "" 0, so
that] "" S.
The first Hund rule has its origin in the exclusion principle and thc (.'o u-
10mL repulsion between electrons. The exclusion principle prevents two elec-
trons of the s.'1me spin from being at the same place at the same time. Thus
cicctl'01I5 of the same spin are kept apart, further apart than e](..ctrolls of oppo-
site spin . Becausc of the coulomb interaction the energy of electrons of the
same spill is lower- the avcrage potential enel'gy is less positive for parallel
spin thall for antipanlllel spin. A good example is the ion /I.In 2 +. This ion has
five eleCh"OflS in the 3d shell, whieh is thercf~re half-filled . nle spins can all be
parallel if each electron enters a different orbital, and there are cxACtly five
different orbitals available, characterized by the orbital quantum numbers
1111. = 2, 1, 0, - 1, - 2. '!llese will be occupied by one electron each. \Vc expe<.1

S :: t. and because ~ml_ = 0 the only possible vallie of L is 0, as observed .

 .. ....
Tllble I EfTL e magnelO(j numben v (or triva]ent lanthanide group ions

- --
(NeaT room tcmpcraturt:)

p(caJc) -

--
,J(exp),

-"""
100

p~ '
eonflgurlllioJ,
-
4j'srV6
'{'5ry'
Basic \evel

ffM
'H,
g[jU + Ill'"

2.54
3.58
approximate

~,

3.5
l\'d 3 + '/,5,.". 'I~ 3.62 3.5
Pm3• 4psrp6 'I, 2.68
Sm3+ 4fssrp6 'H~ 0.84 1.5
Eu 3+ 'f'5?'P' 7fQ 0 3.'
ed 3+ '/,sr,,' s5 71:1 7.94 8.0
Th" 4rsr,I' 7,...S 9. 72 9.5
0," 4r-~ps SHIM 10.63 10.6
1-103+ 4f1Oss'-p~ 'I, 10,60 10,4
E~' 4f'ISs',l' "1 1m 9.59 9.5
Tm3+ 4PYjI aIls 7.57 7.3
Yb3 + 4p~r; 2F71l ' .54 '.5

'Inc second Hund role is best approached by model calculations. Pauling

and \Vilson. I forcxample, give acalculatioll of the spcctr.l.l teons that arise from
the configuration JJ 2 . The third Bund rule is a consequence of the sign of the
spin-orbit inll..'rnCtion: For a single ek"Ctron the energy is lowest when the spin
is antiparallel to the orbital angular momentum. But the low enel'gy pairs ml ...
ms are pl'ogn..'Ssh'cly used lip as we add electrons to the shell; by the exclusion
principle wbcn the shell is more than halrrull the state orJowest energy neces-
sarily has the spin parallel to the orbit.
Consider t\\'() examples of the Bund roles: The ion Ce3 + has a single f
electron; an f ek'Ch'oo ha~ I = 3 and s = i. Bet,:ause the f shell is less than half
full . the J value by the preceding rule is IL - 51 = L - i = t. 1'le ion p~ + has
twof electrons: olle of the I'llics tells us that the spins add to give 5 = I. Bothf
elCt."trons cannot have ffl/ C 3 withoul violating the Pauli exclusion principle, so
Ib.lt the maximum L consistellt with the Pauli ptindplc is IIOt 6, but 5. The J
value is IL - 51 := 5 - I := 4,

Iron Croup Ions

Table 2 shows that the cxpclimental magneton numbers for salts of the iron
transitiOll group of the pCliodic table are ill poor agreement with (18), 'I11C
valucs orten agree quite well witl, magnetoll numbers V = 2[5(5 + 1)Pt2 calcu-

IL Pallhng . ud E. 8. Wilson, 'n,roductiOfl 10 quant"m " ,td"",I<:$, M~l'2w- lI il1 . 1935,

pp 239-2~6.
 ."
Table 2 ElTcclh'c magneton numbers for iron group ions

•• Conng- Basic Jl(calc) - p{calc) ..

I~ uration level dJU + 1)]'11 ~S(S + 11]" p(exp)"
~

Ti H
V"
• yt+ "",
3d'
'D~
3"'2
1.55
1.63
1.i3
2.83
1.8
2.S
C~ ... y'2+ 3d' ~F3.'2 •.n 3.87 3.8
"l n3 ~ . Cr·
"'",p"
liDo
• ' .90 ' .9
Fc3 + , MnH
Fc H
"'~
·s",
'v,
'F.,
a,.
5.92 5.92
•. 90 ...
5.9

COH
NiH
Cu H
"''
3d"
3d'
'F,
'v",
6.63
5.59
3.55
3.87
.83
1.73
• .S
3.2
1.9

lated as if the orbital momcnt were not there at all. We Say that the orbital
moments arc CJuenched.

CnJldal Fiehl Splitting

'Inc difference in behavior of the rarc cooth and the iron group salts is thai
the 4f shell responsible for paramagnetism in the rare earth ions lies deep
inside the ions, within the 5s and 51> shells. whereas in the iron group ions the
3d shell responsible for paramagnetism is the outcnnost shell. nlc 3d shell
experienccs the intense inhomogeneous electric field produced by neighboring
ions. This inhomogeneous electric field is called the crystal field . 111e interac-
tion or the paramagnetic ions with the crystal field has two major clTccts: the
L'O upling orL and S vectors is largely broken up, so that the stateS arc no longer
specified by their J values: rurther, the 2L + 1 sublevels belonging to a given L
which are degenerate in the rree ion may nO\ ... be split by the crystal field, as in
Fig. 6. 111is splittillg diminishes the contribution or the orbital motion to the
magnetic moment.

Quenching of tile OrbitLll Angular Mom erltwn

In an electric field directed toward a fixed nucleus, the plane or a classical

orbit is fixed in space, so that all the orbital angular momentum components Lx,
L". L,. are constant. In quantum theory onc angular momentum component,
usually takc n a.~ L,., and the square orthe tota] orbital angular momentum L'2 arc
constant in a central field . In a nonccntr.l.! field the plane or the m·bit will move
about; the angu lar momentum components arc no longer const.lIlt and may
average to zero. In a crystal L~ will no longer be a constan t or the Illotion,
although to a good approximation L2 may continue to be constant. When Lz
an:rages to zero, the orbital angular momentum i~{ d to be quenched. The

.....
 14 Diamagnetism and Paramagnetism ~.

===-P"P.

, , , - - ,.

(., (,' ("

Figu",6 ConsidL", an atom with orhitalallguJar mIl"",nlum L '"' I placl'tl iulhe uniaxial crystalline
elt:dric lidd of the two posit",e ions along the z axis. In the free alom the states m.. .. :to l, 0 have
identical ellergies- th~"y are degenerate. In the cryslal the atom has a lower energy "'m"TI the
electr<m cloud is close to positive ions lU in (a) than whell it is oriented midway between them, as
in <b) and (c). l1le w,,,",,functions that g",e rise 10 t~ charge Jensities are of the fonn "-f<r), Ilf(r)
and vlCr) and arc called the P.. P.. ,,.orbitals, respecti~-el y. In an axiaUy symmetric liekl, lU shown,
the p, and Pw orbitals are degenerate. The energy levels referred 10 the free atom (dot ted line) are
shown m (d). If the electric fi"kl does nol ha~'e axial symmetry, all three statL"5 \\,i ll have d,ITercnt
energies.

magnetic moment of a state is given by the average V'".Ilue of the magnetic

moment operator J.l.fj(L + 2S). In a magnetic field along the:;; direction the
orhital contribution to the magnetic moment is proportional to the quantum
expectation value of L,.; the orbital magnetic moment is quenched if the me-.
chanical moment L,. is qucnched.
When the spin-orbit interaction ellergy is illtroduced, the spin may dmg
some orbital moment along with it. If the sign of the intcmction favors parallel
orientation of the spin and orbital magnetic moments, the total magnetic mo-
ment will be larger than for the spin alone, and the g value will be larger than 2.
The experimental results are in agreement with the known variation of sign of
the spin-orbit interaction: g > 2 when the 3d shell is more than half full, g "" 2
when the shell is half full, and g < 2 when the shell is less than half full.
Wc consider a single electron with orbital quantum number L = I moving
about a nueleus, the whole being placed in an inhomogeneous crystallille elec-
tric field. We omit electron spin.
In a crystal of orthorhombic symmetry the charges on neighboring ions
will produce an electrostatic potential q; about the nucleus of the form
(24)
where A and B are constants. This expression is the lowest degree polynomial
in x, y, :;; which is a solution of the Laplace equation V2 q; = 0 and compatible
with the syml '"Y of the crystal.
\
'"
In free space the ground state is thrl:.'C·fold def; ..:rate, with magnetic
quantum numbers filL = 1, 0, -1. In a magnetic field these levels arc splil by
energies proportional to the field B, and it is this field-proportional splitting
which is responsible for the normal paramagnetic susceptibility of the ion. In
the crystal the picture may be different. We take as the three wllvefunctions
associated with the unperturbed ground state of the ion
v.. = xf{r) ; VII :: yf(r) ; (25)
-I11CSC wavcfunctiolls are orthogonal, and we assume that they arc normalized.
,
Each of the U's can be shown to have the property
!£lUj = L(L + l )V, :: 2U, • (26)
where ~ is the operator for the square of the orbital angular momentum, in
units of Ii . 111c result (26) confirms that the selected wavcfunctions arc in f.'lCt p
[unctions, having L = 1.
We observe now that the u's are diagonal with respect to the perturbation,
as by symmetry the nondiagonal elements \'anish:
(U,k<lu,) ~ (V.le<lVJ ~ (UJe<lUJ ~ 0 . (27)
Consider for example
(U.le<lU,) ~ I nAf(,)F{Ax' + .,;' - (A + .»') <Ix dy dx (28)
tIle integrand is an odd function of x (and also of !I) and therefore the integral
must be zero_ The energy levels are then given by the diagonal matrix
elements:
(u.le~IU.-> = f if{r)12{Ax· + ByY - {A + B)H} ell' (ly dz
(29)
A(II - I i) ,
where

In addition ,
(VJc<lV,) ~ .(1, - I,) (v,lc<lvJ ~ -(A + .)(/, - I,) .
'l1lC three eigenstates in the crystal field are" functions with their angular lobes
directed along each of the x, y, z axes, respectively.
111e orbital moment of each of the Ic\'e1s is zero, because
(U,IL.lu,) ~ <UJLJU') ~ (UJLJUJ ~ 0.
l'llis effect is kno"'TL as quenching, The level still has a definite total angular
momentum, since ~ is diagon.ll and gives L = 1, but the spatial compQnents
of the angular momentum are not constants of the motion WId their time aver-
age is zero in the first approximation. Tbereforc the components of the orbital
magnetiC moment also vanish in the same approximation. 111e role of the crystal
 .

field in the qu • ...:hing process is to split the originally degenerate levels into
nonmagnetic levels separated by energies ~ J.LH, so that the magnetic field is a
small perturbation in comparison with the crystal field.
At a lattice site of cubic symmetry there is no term ill the potential of the
foml (24), tha.t is, quadratic in the electron coordinates. Now the ground slate
of an ion wi th onc p electron (or with one hole in a " shell) will be triply
degenerate. However, the energy of the ion will be lowered if the ion displaces
itself with respect to t1le surroundings, thereby creating a lIollCtlbic l'lOtcnti<l1
such a.~ (24), Such a spolltaneous displacement is known as a )aho-Teller effect
and is often large and hnport."l.nt, varticularly with the Mn3 + and C\l2+ ions2
and with holes in alkali and silver halides.

Spec::frosoopic Splitting Fador

\Vc suppose for conveuience that the crystal field constants, A, B nrc such
that U~ = if(r) is the orbital wave function of the ground state of the atom in
the cryst.'1I.. For a spin S '"' i there arc tvro possible spin states S~ ... ± I repre-
sented by the spin functions a, p. which in the absence of a magnetic field are
degenerate in the zeroth apl)roximation. The problem is 10 take illto account
the Sl)in-orbit interaction energy AL· S.
Jfthe ground state functiou is ifJo = Up :z: xf(r)a in the zeroth approxima-
tion, then in the first approximation, considering the AL' S interaction by
standard perturbation theory, we have
(30)
where AI is the energy difference between the Va and VII states, and ~ is the
difference between the U~ and U~ states. The term in VJ3 actually has only a
second order effect on the result and may be discarded. The expectation value
of the orbital angular momentum to the first order is givcn directly by

and the magnetic moment of the state as measured in the :t direction is

~,(<IIL,. + 2S.I~) = [-(AId,) + l]~ •.

As the separation betwecn the levels S" = ± j ill a field H is

ll£ = g)J.nH = 2[1 - (AlA I)]J.l.nH •
the g "~dluc or spcctfOSC(lpic splitting factor (12) in the :t direction is
g = 2[ 1 - (AId,)] . (31)

'See L. Orgel. l nt ..oducllDn to trontit ion mt'tc/ d'tmill'1I . 2",.1 ed., Wiley. 1966; exlells;,'e
refet'enee5 af'e g;''en by M. D. Sturge. I'hys. Rev. )40. A880 (1965).
 Van Vleck Temperature-tnclepemlelll Paramagneti&m
\\'e consider an atomic or molecular system which has no magnetic ,no-
ment in the ground state, by which we mean that the diagonal matrix clement
of the magnetic moment operator ~ is zero.
Suppose that thcre is a nondittgonal matrix clement ($I~JO) of the magnetic
mOlllent operator, connccting tile ground state 0 with the excited state s of
energy 6. ::: E~ - Eo above the grollnd state. 11len by standard perturbation
theory the wavefunction ofthe ground state in a weak field f.J..t,.B ~ A) bccomt.'S
(32)
and the wavcfunction of the excited state becomes
(33)
The perturbed ground state now has a moment
(34)
and the upper state has a 'Iloment
(35)
'I1lCre are two interesting cases to consider:
Case (a). II <f ksT. TIle surplus population in the ground state over the
excited state is approximately equal to NA12k Il T, so that the resultant magneti-
:I..ation is

At ~ 2BI('I~.IO)I' NA (36)
A 21BT
which gives for the susceptibility
X ~ NII'I~. IO)I'lk,T . (37)
Ilcre N is the numbcrof moleculcs per unit volume. 111is contribution is oCthe
usual Curie form, although the mechanism of magnetization here is by polariza-
tion of the states of the system, whereas with free spins the mechanism of
magnetization is the redist ribution of ions amollg the spin st.'ltes. We notc that
the splitting A docs not cnter in (37).
Case (b). 6. ". kilT. Ilere the population is nearly all in the ground state, so
that

(38)

TIlc susceptibility is

(39)

{
•
 J4 DianWlgndum /!I,d Pll .... m<lgnefll'" .:Jl

independent of temtlCrature. 'Ibis type of contribution is known as Van Vleck

pm-amagnetism.

COOUr\C BY ISENTROPIC DE!\(ACNETlZATION

111e first method for attaining temperatures much below 1 K was that of
iscntropic, or adiabatic, demagneti:w.tion.3 By its usc temperatures of 10- 3 K
and lower have been reached.. The meth<xl rest.<; on the fact that at a fixed.
temperature the entropy of a system of magnetic moments is lowered by the
application of a magnetic field.
The entropy is a measure of the disordcr of a system: the greater the
disorder, the higher is the entropy. In the magnetic field the moments will be
partly lined up (partly ordercd.), so that the entropy is lowered. by the field. The
entropy is also lowered if the temperature is lowered, as more of the moments
line up.
Ifthe magnetic field can then be removed. without changing the entropy of
the spin system, the order of the spin system will look like a lower temperature
than the same degree of order in tile presence of the field. \\'hen the specimen
is demagneti;.r..cd at constant entropy, entropy can flow into the spin system only
from the system oflatticc vibrations, as in Fig. 7. At the temperatures of inter-
est the entropy ofthe lattice vibrations is usually negligible; thus the entropy of
the spin system will bc essentially constant during adiabatic de magneti;.r.aHon of
the specimen . Magnetic cooling is a one-shot ollcrntion, not cyclic.
We first find an expression fur the spin CIltropy of a system of N ions. each
of spin S. at a temperature sufficiently high that the spin system is entirel),
disordered. l1mt i~. T is supposed to be much higher than some temperature A
which characterizes the energy of the interactions (E tnt !!I k8A) tending to ori-
ent the spins prefcreotially. Some of tllese interactions are discussed in Chap-
ter 16. l11e definiHoo of the entropy U of a s}'stem of G accessible states is
u = ks 10 G. At a temperature so high that all of the 2S + 1 states of each ion
are nearly equally poPlllated. G is the number of ways of arranging N spins in
2S + I states. Thus G = (2S + l)N, whence the spin entropy Us is:
Us = ka In (2S + It' = Nks In (25 + 1) . (40)
This spin entropy is reduced by a magnetic field if the lower levels gain io
population when the field separates the 25 + 1 states in energy.

3-fh.e method was 5l1ggCSted by P. Debye. Ann. Physik 81. 11s.l (1926): aOO W. F. Gtauque.
1. Am . Chrm. Soc. 49, 1864 (\927). For many pu~ the method has been 5upplantt-d by the
lie-lie' dilution refrigeratO<' ....ilich opentCl '" • continuous C}de The lie atorm in solution in
liquid lie' play the role of atomti in a gas, and cooling is effcdell by ~''IIporiu.tion~ of He; 5ee Tl',
Chapter 12.

\
'"

,I
I
I
I
f :
"
~'~.~.~-,-
1';"
~.,--------
I
...
:
I

L,.>IIiff> I
I r-; ..... '"'1"- , N"",' ~",/Jbriurn
Ti ..... ~I..-h;ch Tome II "hod!
'''''I:I'''t", ~dd "'.t:n~·tioofl..k1
i> ",,,,,",'.'Il .. rein.",,,.!

Figurt' 7 Durlng adiabatic J emagnc til.ation the total ..."tropy or

tht: specimen is <'OI'SI~llt For
.,rr.,<.'1;,.., C()Olillg tl, ... i"itial entropy dlhe latti<:e should be small III comp.mson with.he entropy of
the spin system.

The Ji ICpS carried out in the cooling process are shown in Fig. 8. 111c fiel d
is applied at temperature TJ with the specimen in good thermal contact with
the s urrounding.~, giving the isothermal path lib. The specimen is then ins u-
lated (au"" 0) and the field removed; the specimen follows the constant e n-
trop), path be, ending up at temperature T:z. The thermal contact at TJ is pro--
vidcc.l by helium gas, and the thenna! <:oulacl is Lrokcn by removing the gas
with a pump.

Nuclear Demagnetization

' I1I C pop"lalion of a magndic sublevel is a function only of J-tB1k8T, hencc

of Brr. The spin-system entropy is a function only of the population di~tribu­
tion; hence the spin entropy is a function onl), of BIT. If Bt:. is the effective fleld
that (.'Orresponds to the loca1 interactions, the flnal temperatufC T2 reached in
an adiabatic demagneti~tion experiment is

1 T, : T, (B"IB) ·1 (41)

where B is the initial field and T. the initial temperature.

Because nuclear magnetic moments arc weak, nuclear magnetic interac-
tions arc much weaker than similar electronic interactions. We expect to reach
a temperah.lfe 100 times I()\yer with a nuclear paramagnet than with an electron
paramagnet . The initial temperature TI of the nuclear stage in a nuclear spin-
cooling experi ment must be lower th.1Il in an c1(.'Ctron spin-coolillg experiment.
If we start at 8 = 50 kG and TI "" 0.01 K, then f.dJlk 8 1'1 .... 0.5, and the en-
 _ "7

O
.'r-----------_....""-,""
0.6
i., O.l!
§
S 0.4

..~ 0.'

~ O'l~~~::_..:=~~~[t1
" '.00 l! )0 15 20 ~ 30
T, mt.:
Figure 8 Entropy for. ~pin I $~'5tem as . fur.ctioo of temperature. aJ.Suming all in!<-.. nat random
mlgJ>etk 6",.:1 B.. of 100 gauss. 11>c specimen is maj.,'1><' UlflI isothemla1!~· along Db. an,] is then
in ~u\;, too thl'rnU\U)·. TIlle extenla! magJletic fidel ;,; tunued olf along be. In order to ~eep the figure
Oil ~ reasonable iK'a1e the initi.l t(' mperature T, is low"r than ",..,.,IJ be liSt'<! in praclic:e , and so is
the external m.W'rtic:: roeld

tropy decrease on magneti:tation is over 10 percent of the maximum spin c u-

trop)'. 111is is sufl'icient to oven\'hclm the lattice and from (41 ) we estimate a
final temperature Tz - 10- 7 K. The first' nuclear cooling el[pcrimcnt was car-
ried out on Cu nuclei in the metal, stal1.ing fmm a first stage at about 0.02 K
as attained by electronic cooling. 'fhe lowest temperature rcached was
1.2 x 10- 6 K.
The results in Fig. 9 fit a line of the fonn of (41): T! '" T 1{3.l/B) with B in
gauss, so thai 8 6 '" 3.1 gauss. This is the efTccLive inlcraction field of the mag-
netic moments of the Cu nuclei. 'nlC motivation for using nuclei in a metal is
that conduction e1ectronl! help ensure rapid thcrmal contact of lalli~ and nu-
dd at the temperature of the first stage. The present recordS for a spin Icmper-
ature is 280 pK. in rhodium,

PAHAMACNt:TIC SUSCEPTIIHlJTI' OF CONDUcnON ELECTRONS

We arc going to try to show how 011 the 1>.15ls of these slatistics Ihe faclthat many
metals are diamagnetic, o~ on ly weakly pararnagnetic, can be broughl inlo agree-
ment with the existence of a magolelic rnOllwnt of the electrons.
W. Pauli. 1927

Classical free electron theory gives an unsatisfactory account of the para-

magnetic sllsccplibilit)' of the conduction ek'Ctrons. An electron has associated
with it a magnetic moment of one Bohr maglleton, /An . One might expect that

4N. Kurti , F. N. II. Robin. OI1 , F. E. Simon, lind D. A Spohr, f"aturt' 178, 450 (1956); ror
WV;C"" S SCf: N. Kurti, Cryogen;c>; I, 2 (1960). A,]v. in Cryogenk EngJnL-e~;nS\ 8, I (l963).
~P. 1. Hakoot'n rt aI. , PhflS. 1Iftl . I..dl 70,2818 (1993).
 ""
""""lei", /kid ,n kG
, 1,,;h.1
ro
",, - " '"
,
"< 6, -
~
!• •
,,, ,
,,
,,••
(!;

, , ,
" " 1,,,1,,,1 /JfT ill lif CJJ(
Figure 9 N"de", d.·mdgnct;""tlo"$ of <'I.lpper "uclei in the melsl , Stdrt;Il8 fro", 0.012 K d,Kl
y" n OOS FIelds. (After \1 . V Ih;>h<len and N. Kurti.)

the conduction electrons would make a Curie-type p..'lramagnctic contlihution

(22) to the magnetization of the metal: AI = Np-tBlkBT. Instead it is obsclved
that the magllctiza.lion of most normal nonfcrromagnetic metals is independent
of tempcrahll"c.
Pauli showed that the application of the Fermi-Dirac distribution (Clulp-
ter 6) would correct the theory as required. We fin;t give a qualitative explana-
tion of the situation . The result (18) tells us that tIte probability an atom will be
lined up par,lUd to the field B exceeds the probability of the antipnra1lc1 OriCIl-
!.llion by roughly p.BlkBT. For N atoms per unit volume, this gives a net mag-
netization -Np.2BlkBT, the standard result.
Most conduction electrons in a metal, however, have no possibility of
turning over when a field is applied, lX!Cause most orbitals in the Fermi sea
with parallcl spin arc already m;cupicd. Only the electrons within a range k,!'f
ufthe top ufthe Fermi distribution have a chance to turn ovcr in the field; thus
only the fraction TIT~· of the total nu!nbcr of cledrons contrihute to the suscep-
tibility. Hence
Np.'lB T Np.'l
M-~~-- - =~-B
kBT TF kBh
which is independent of temperatuTC and of the observed order of magnitude.
We now calculate the expression for the paramagnetic susceptibility of a
free electron gJ.S at T .g Tf ... We follow the method of calculation sugge~tcd by
Fig. 10. An altcnl.lte derivation is the subjed of Problem 5.
(
."'., ,. .
 Toy! Ctl"~. ~""-'ht +
,""'W"-11t. <i ~k<:t""'"

,.J Ib'
Figure: 10 Pau li paramagnetism at ahoolute zero; the o rbitals in the u,lldeti regiont In (a) are
oocul*d. The numhen of eledr...s III the HUpHand - dov....- bam.! ,,;\1 Adjust to moo).e tbe energies
ellual at the Fermi level. lhe cbcmLCII\ potential (F<.:rmi level) ofille II1UIl1~" up dectrons is C'IU:>!
10 lhat uf the Illomcut do..T1 elcclrom. III (h) we show the eJlCCU of mOlllent up declrolll III tbe
magnetIC field.

The concen tration of electrons with magnetic moments llamllcl to the

magnetic field is

N+ = - 'f"
2 H"B
lrE D(E + IlB) a - Ii"
2
dE D(E) + -1
2
IlB D(E,.·)

written for absolute zero. Here iD(E + pB) is the density of orbitals of one
spin orientation, with allow,mce for the downward shift of energy by - p.B.
The approximation is written for kilT <C E".
The concentration of electrons with magnetic moments antipamllcl to the
magnetic field is

N_ = - , J"
2 "II
dE D(E - IlB) ill - , i" dE D(E) -
2
-1 IlB D(E~')
2

1ne magnetization is given by M = p,{N+ - N _), so that

3Np.2
~-- B (42)
2k nT,.,
with D(E,,) :::: 3N/2€,.. = 3N12k. 8 T,.· from Chapter 6. 1ne result (42) gives the
Pauli spin magnetization of the conduction electrons, for ksT <t E".
in deriving the paramagnetic susceptibility, we have supposed that the
spatial motion of the electrons is not afTcdcd by the magnetic field. But the
wavefunctions arc modified by the magnctic field; Landau has shown that for
(
."

fn.'C electrons this causes a diamagnetic moment equal tv -i of the paramag_

netic moment. l1lUs the total Il)llb'Tlelization of a free electron gas is

_r_'_ B
/II == N'" (43)
kaT"
Bt:forc comparing (43) wit h the experiment we must take al'COlJut of the
diamagnetism of the ionic cores, ofballd efft.'Cts. and of electron-electron inter_
actions. In sodium the interaction efft."Cts increase the spin susceptibility by
perhaps 75 percen t .
The maguclic s us(:eptibility is l'Ollsiderably higher for most trallsitioll me t-
als (with unfll k--d inner ck'Ctron shells) than fol' the alkali metals (Fig. 11). TIle
high \'<liucs suggest that the de nsity of o rbitals is ullusually high for transition
meta1s, in agreement with mcasurcmcuts of the ciectrouic heal capacity. \Vc
saw in C k l.plcr 9 hm\l this arises from band Ih(.'ol")'.

SUMMARY
(In CCS Units)

• The diamaglletic sus<:cptibility of N a tom s of atomic number Z is X =

-Z(?N(,-2)/6m~. where (r') is the mean square atomic radius. (Langcvin)

• Atoms wi th a permanent magne tic moment p. have a p.1ramagnctic suscepti-

bility X = NI.83kB'r. for J.tB <C kB'f. (Curie-Langevin)
• For a system of spins S = t the cxact magnctization is M =
Np, tan h(p,BlkBn. whe re p, "" hlP,B' (Brillouin)
• 11.e gro und state of e lectrons in the same shell have the ma.ximum value ofS
allowed by the Pauli principle and the nlaximum Lconsis tc nt wilh this S. 11le
) value is L + 5 if the she ll is mo re tha n half full and IL - 51 if the shell is less
than half full .
• A (.'OOling process opemtes b y demagnetization of a p.1r.l.lllagnctic salt at ( .'0 11 -
stant entropy. The final tempera tu re reached is of the order of (8."IB)Tkllhalo
where 8." is the effective loc.11 field and 8 is the initial applied magnetic field .
• The paramagnetic sll~ptibili t y of a Fermi gas of collductioll electrons is
X = 3Np, 2/2€,:. inde pclldent of temperature for kB'f <C ~/:' (Pau li)

Problems
1. Diamagndic 3usceptibility 0/ a'omic hydrogel!. 'Ole wave fun('tion of the h)·Jrogen
awm in its ground state (b) is .p '" (~ - 1IiI C):p(- riao). "here aD = "2/mil,,.
0529 x to "cm. TIle dlarge m'llsity is p(x:. y. z) - - ell/II~. according to d ie statisti-
cal interpretation of the wave function. Silow that for this slate (r) "" 3a~
ani! caktllate tile molar i!iamagnetic susceptibility of atomic hyi!rogen
(- 2.36 x 10 "cIll3/mole).
 * . .,

14 Du"nasnt'li.-.IO ""d "Dramas"..,;'." 437

"
, 0
1'\
, 0

To
0

\
0
V -
,...-C, "
1\ ~\\'

VI:
0

\
, 0

/ "
" m,
\10
.0 "
~
• l::::- "r "
lib
0
o 1'00 400 600 IiOO 1000 1200 1400 JGOO JijOO 2000 2200
T . jll K
Figure II 1;,mpcrnturc dCllC.u]cnct' of the magnetic sU$Ct.1It il>ility of ",clals. (Cou rtesy of C. J.
Kri~'SSma... )

2. Hund rules. Applr the ll uncl ml<·§ to find the gruu n{1 slate (the basic level in the
notatioll of Table I) of (a) Eu~", in thet·i.mfiguratioIl4.f 5?p6; (b) YIP~ ; (C)Tb3 +. The
rt'§ ult!S for (b) and (c) arc in Table I, but )'uu should give the separah~ stellS in appl) ing
the niles.

3. Triplel clOCilf!Cl sloles. Some organic mulecules have a triplet (S = I) excited state at
an energy k/lfl above a Singlet (S = 0) ground state. (a) Find an expressiun fur the
magnetic moment (p.) in a field B. (b) Shuw tllat the susceptibility for T '" fl is
approximately indepl'lldent of fl. (c) With tile help of a diagram of energy le"els
versus field and a rough ~ketch uf entrupy versus field , explain how Ihis system might
be cooled br adiauatic magnetization (not demagneIi7.1tion).
4. Hoot copacityfrom internal (iegrcn offrccdom. (a) Cunsider a two-level system with
an energy splitting kllfl between upper ,md lo....'('r states, tl.e splitting may arise fru m
a magnetic field or in other ways. Show that tl.e heat capacit)· pe r system is

c = (aT
'U) 1\
(!lIT)".'"
= kH(I +~'f '
11.e functiun is plutted in Fig. 12. Peaks of tMs type in tl.e he-olt cap.lCit y are often
knuwn as Schott!.." anuma lies. The ma'l:imum heat capacity is q uite high , LlU t for
T 4 t::. and for T ~ t::. the heat capacity is luw. (b) Shuw that for T ~ fl we I,ave
C ... k8(MT)2 + . ... TI.e hypcrfine in teraction between lIudear and electrunic
magnetic muments in paramagnetic salts (and m systems havmg electron spin order)
causes splittings with t::. "" I tu 100 mK. These splittings are ufte n dek:ctw experi-
mentally by the presence uf a term in In'2 in the heat capaci ty in the r~ion T ~ fl.
Nudear elcctTic quadrupole in terolctions (s.o:e Chapte r 16) with crystal fleld~ also
cause splittings, as in Fig. l 3.

5. l'(mli 8p,n susceptibility. The spin slI.;ccptiui lity of a l'Ullduction electron gas at alISO-
lute zero may Le discussw by anotll{~r method. Let

be the CQIlccntratinns of spin-up and spin.(lo"n electrons. (a) Show that in a magtletic
field B the total energy of the spin·up Land in a free dl'Ctron gas is

where Eo = /oNI!"~" in terms of the Fenni encrgv f':F in zero magnetic field. Find a
similar expression fur £ - . (b) Minimize £,,,,a1 = £+ + £- with respect tu {and sulve
for the C(luilihrium value of, in the approximatiun {4 l. Go un tu sI.ow that the
mdgnetizatinn is At - 3f',l P-2 BI2f':F. in agreement with Eq. (42).

6, CUf1(luction electron ferronwgnef.ism. We approx ima te the effect of exchange illter·

actions among the conduction electrons if we assume th at electrons with parallel
spins interact witl. eadl uther wi th encrgy - V, and \' is poSitive, while electrons with
antiparallel spins do not interact .....ith each other. (a) Show witl. the help of PrublfOtn 5
that the total energy uf the spin·u p band is
£+ = £,,(1 +~- jVN~l + (f - INP-B ( 1 + () ;
And a similar expression for E - . (b) Minimize the tutal energy <IJId solve for { in tl.e
limit {~ l. Show that the magneti7.(ltion is
3fo,'p-~
AI -
2f'F
i VN B ,

tI.e total energy is unstable at ,=

so that the exchange illteraction enhances the susceptibility. (c) Shuw tl.at .....itl. IJ = 0
0 when V > 4f;,J3.'IJ. If this is satisfied a ferromag-
netic state ({ '" 0) wi ll have a lower ene'1,.'Y tl.an the paramagnetic state. Because of
tl.e assumption {..:g I . this is a sufficient condition for ferromagnetism, but it I1ldY
not be a necessary conditiun. It is knuwn as the Stoner condition.
 ",---.----.---,----r---.---,

"
,,, "
.J: I
•I
l<~d

Lc>ell
2

,•J"
,.,

, • , 6
~ . TI!J.
.-'Kun> 12 Ilcdl ~"I),M;ily <l ~ h..,..1c""'\ sy.tcn\ >\I " fundi .... , ,,(Tll1. where II i, the it'VcI ~ilh ui ng.

..
'n,,; SchottkY .nlol1l~ly Is a vcry IIsdul toul for d"lennininJ: encr),,), level splillin&, ('I ions in IlIre-
t~"lh and trun>ition-.l:TOup llIetal., coltll'o..."b. ~nd >tlk,>"s.

," ''''" C~T' - 431( OSElEiTl/ r •

,,
1
0001 -
,>./
10 ....

'-U•
0.002 -

" .. ." ,,_''''"

t
, -', III
,
FiRure 13 11](, nurm~l -stale heat C<lI)Jl(ity of Ilallium lit 'r < 0.21 K. The "lid...... quadrupole
.. '"
(e '" T -~) and cond ultion ,,1..""1011 (e« f) rol1 I.I\)lI lioos dumb""c the hc.Il CIIlooty at '~ry low
temperatures. (After N. E. Ptullips_)

,
\
4·10

(
7. Two-level Bystem. The result ofl'roLiem <I is often seen in anOl"., r furm. (a) If the two
energy levels are at 6. and - 6., shU\\t that the energy and h eat capacity are
u= - 6. tan h(6.fk aT) ;

(1) If the sys tem has a random composition such tllat all \talues of 6. are equally likely
up to some limit ~, shuw that the heat capacity is linearly propurtionalto the tem_
per.ltU.-e. pruvided kaT.o::g 6.0- This resu lt was applied to the Ileat capacity uf dilute
magnetic alloys by W. Marshall. PllYs. Rev. UB, 1519 (1960). It is also used in the
theUl"}, of glasses. Chapte r 17.

B. Pa,.amagnetism oIS = 1 By~tem. (a) Find the magnetization as a function of magnetic

Reid and temperat ure for a system uf spins with S = I , mument p., and concen tration
1"1. (b) Show that in the limit ~ .o::g k1' tile result is M ;i2 (21"1J.( f3kT)B.
i

References
A. Abragam and B. Blcancy, EI""tron parm1l<"lgnetic rt8Qnana (Jj tran$i!ion ions, Dover, 1986..
H. B. C. Casimir, MagnetISm and ury /ow tempertJturu, Don'r, 1961. A cJassic.
M. I . Darby and K. N. R. Taylor, Pllysia of rare oorlh solids, Halsk d. 1972.
A. J. Freeman, rIle actinides. electronic structure and related properlics. Academic, 1974 .
R. D. II udso", Principies and applicat ions of mlIgnetic ccoling, Elsevier, 1972.
H. Koocllfel , Pv/.red "Igh mlIgrwtic!t£lds. North- Holland. 1970.
O. V. Louna$maa, E%}"M'rimental principles "nd methods below 1 K, Academic p, ~. 1974.
L. Orgel, Introduction to trtmsitioOi "'elal chemistn;, 2nd l.J., W,lt:y. 1966.
!. I I. Van Vleck, 1'''e ,lleory of electric and magnetic susceptibilit ies, Oxfurd, 1932. Supe,b derivlI-
tions of !.>asic theorems.
C. K. White, l-:l'perlmenl"l techniqUes in low tOOlpcmtufll physics, 3rd cd. , Oxford, 1987.
R M. White, Quantum thecry of magnetism, Springer, 1982.
A. J. Frt:cma" and C. H . Lander, Il andbook on tIle physl~ and clleml$try of the actU.wes. North.
Ho lland, 1984_1993.
M. D. Stu rge, "Jahn-Tel!t:r effect in solids," Solid state physics 20, 9 1 {1967}.
M.e. M. O'Brien lind C. C. C hancey, "rhe Jah,,-Te ller eff~'CI; An introduction lind current re-
vk'W," Amer. J Physics 61, 688 (1993).
 ~ .

15
Ferromagnetism and Antiferromagnetism

FERI\OMAGNJ:o:TIC ORDEI\ "'43

Curie point llncl tl-Lc exchange integral 443
Temperature depenoence of tile saturation magncti:w.tion 446
Sah,ration magnetization al absolute zero 448

~~CKONS 4~
Quanti'l.IItion of spin waves 454
Thermal excitation of magnons 455

N"EUTRON MAGI'ETIC SCATn::IUXG 4S6

FERRJMAGNETIC ORDEII 458
Curie tcmpel1lture and 5uSCt!plibilily of ferrimagnels 460
Iron gIl.rncts 461

ANTIFEIIROMACNEllC ORDER 462

Susceptibility below the N~l tempernturc 465
Antiferromngnclic magnons 466

FERROMAGNETIC DOMAINS 468

Anisotropy CneTgy 470
TnlOsition region between domajns 472
SoIilulls 473
Origin of doJlU\ins 474
Cocrcivity and hysteresis 475

SINGlE DOMAIN I'ARTICIES 471

Geomagnetism Dnd biomagnctism 478
Magnetic force microsc:o(>y 478
PltACr-;ETlC BUnBLE DOMAINS 480

SUMMARY 481

I'ROBLEMS 481
1. Magnon dispersion relation 481
2. Heat capacity of magnons 482
3. Nl!cl temperature 482
4. Magnctoeillostic coupling 402
5. Coercive ror<.'C or a small particle 482
6. Saturation magnetization ncar T~ 483
7. Nl!el wall 483

REFERENCES 483

KQTATlON: (CCS) B = II + 4'11111 ; (SI) B - J.4I.. H + /11 ). We call B. the H)plied mag·
netic Geld in both systems or units: in CCS we Ilave llG - H. And in SI we have B. -
p,oIf•. The susceptibility is X "" Mill. in CCS and X - Mill. - lJoAflB. in SI. One
tcsla '"" lot gauss.
 111111 1 I 1I 1 I I I
S;,nple r~rt",nagnel Simple anllrc .to,~
1 1
Fcn i ...."'-1
1

/\/\/
r....,ltd_Ur.:~
D~~~~~ IIdl<'>l.pm .....,
- - ~---

F~",a-zybo..d
, ,.,
--- - !.-

t'igure I Ord ..... L"Il arnongernc;nls of ",k..,lron ' pillS.

.,
 CHAPTER 15: FERROMAGNETISM AND ANTIFERROMAGNETI5M

FERRO~IAGNEnC onOEH

A ferromagnet has a r;pontan cous magnetic momcllt-a magnetic moment

even in 7.cro applied magnetic field , The e'l:istence of a spont.llleous mOlllent
suggests that eIL-'ctron spins and maglletic moml'n ts are arranged in a regular
malineI', The order net:<! not he simple: all orthe spin arrangemen ts sketched in
Fig, 1 except the simple an tiferrolilagne t haVe a spontaneous magnetiC mu-
ment, called the saturation momeut ,

Ct,rie Point (nit! the Eu/longe hltegmJ

Comidcr a pammagnet with a concentration of N ions of spin S, C ivcn ,In
inte m alinterJ.etioll tellding to line lip the magnetic moments parallel to each
other, we shall have a rerroma1,'11et. Le t liS postulate s uch an inteTat,tion and call
it the excllllllge field. I The oricllting effect of the exchange fi eld is opposed by
thermal agit:\tioll, and at elcv:\h.:d temperatures the spill order is destroyed.
We treat the exchange field :\s eljuivalen t to a magnetic field H,.-. l11e
magnitucle or the exchange field may be as high as 107 gauss (10l te~la). We
ass ume that B t; is proportional to the magncU;(A1tioll M .
111e magneti7..1tion is defined as the magnetic moment per unit volume;
unless otherwise specified it is understood to be the value in thermal equilib-
rium at the temperature T. If domains (regions magnc ti7.ed in different direc-
tium) al'e prescllt , the magne tization re fers to the value within a domain,
III the meun field approximation we assu me each magnetic atom experi-
ences a fie ld proportional to the maglle t i7~1t io ll :
BE = AM , (I)
where A is a con~tan t , independent of temperature. Ac(.:ol'ding to (I), cac.:h spi n
sees the average magnetization of all the other spins. In truth, it lIIay see olll},
near ne ighbors, but our simplification is good for a fi rst look at the problem.
'!lIe Curie temperature T" is the tempcrature above which the spontane-
ous magnetization vanishes. it scparate.. the disordered paramagnetic pll.'\Se at
'1' > 'I'.: frnlll the orden.'<i fcrrom 3b'llctic phase at T < Te. \Ve e;\n fi nd Te in
terms or A,

'Abo calk-d thc molecular Ildd or thf Weiss Ildd, ~fter Pierre Weiss "nO was 'he Ilut to
It,.
iTllllgine such a r:eld. 11", C~C"Rnl:c Reid Br .i,nuilltes II r(:al m8J{l.t:tic fI.,ld in Inc expressIOns Ihe
eneT),')' - IJ. ' Br ..,d the lor'l"""')( Bo: "".11 m;aJ:nrhc mOll","t,... But B. Ii lIot re,<lly. n'4;,,,,he
f",1d .,.,d thererore do..s not ~'fllfr mto Ih.. \lu",'cli ttjUA11On5. for eumple, th<Te is no curre,,'
densily I rdalt,d 10 Bf! by curl H '"' 41rj1c. The maguitude of 8" is 1)'IlIcally 10' 1.org<."I' than the
....·cnge ,nagnclic rIC'" of ,h., magtICI", dipoles or the r.,rromagr>el

' ..
 Consider the paramagnetic phase: an applied fiela v" \viil cause a fin ite
magnetization and this in turn will cause a finite exchallge field Bf ;. If X" is the
paramagnetic susceptibility,
(CGS) M = X,.(B.. + Bd ; lSI) """': X~B. + Bd . (2)

111c maglletization is equal to a col/stant susceptibility times a field only if the

fractional alignmcnt is small: this is where thc assumptiOIl elltcrs that the speci-
Incn is ill the paramaguc tic phase.
The paramagnetic susceptibility is given b y tllc Curie law X" :: CIT,
whe re C is the Curie collslallt. Substitute (I) in (2); we fin d M T = C(B" + " M)
'nd

(CCS) X " = -=c--"-=,,-

= -8. C (3)
(T - CAl
The susceptibility (3) has a singulmi ty at 'f = CA. At th is tcmpc.-ature (and
below) there exists a spolltancous magnetization. because if X is infi uile we can
have a finite !II for zero R". From (3) we have the C urie-Weiss law

(CCS)
c (4)
x := T - Tc '

111is expressioll describes fai rl y well the observed susceptibility variation in the
paramagnetic region above the Curie point.
Detailed calculalionr predict

at temperatures vcry close 10 'I"", in gencral agreemcnt with the experime ntal
data summarized in Table I . The reciprocal susceptibility of uickel is plottl.--d ill
Fig. 2.
Fro m (4) and the definition (14.22) of the Cllrie constant C we may deter-
mine the valuc of the mean fi eld constant A in (I):

(CGS) (5)

For iron Tc - 1000 K, g - 2, and S - I; from (5) we have" .... 5000. \ Vith
M. - 1700 we have BE = AM - (SOOOXI700) "'" 107 C = IfrI T. The exch angc
field in iron is very much stronger than the real magllctic fi eld d ue to the o ther
magllelie ions in thc crystal: a magnetic ion produces a field ... ~1J1t? or about
Ifrl G = 0. 1 T at a ne ighboring lallil.'e point.

IExpl'rilllCIl lally the SUSC'rllliloility fur 1· .. ·f. is ~"l'" qllit., ACau11td)' by Clrf - UJ. whero: 8
i, allllro:cillLly ycaler than the, ftCtual transition Icmpcl1\turo: T•. S ..... the rC\l;"-w by C. Doonb in
MOptiJDI, \'01. 2A, C. T. Rado lind II . Suhl, lods .. ~n'ic PA.-U, 1965
 .. ..

TA... .: 1 Critical point c:<ponents for ferromagncts

"As T - T~ from above. the susceptibility X becomes jJl"OpOrtiooai to (T - T~) - 7; as

T- Tc from below, tile magnetization M. becomes proportional to (Tr - T)I!. In the
mean field approximation, ")' "" 1 and fJ "" 1.

- , ~ T,. in K

F. 1.33 ~ 0.015 0.34 ~ 0.04 10<3

Co 1. 21 ~ 0.04 1388
Ni 1.35 ~ 0.02 0.42 ~ 0.07 627.2
G<l 1.3 ~ 0.1 292.5
CrOll 1. 63 ± 0.02 "'.i5
Cr Ur3 1.215 ± 0.02 0.368 ± 0.005 "~56
EuS 0.33 ± 0.015 16.50

Experimental data roneetcd by II . E . Stanley.

,
~
, /

, /
/
V
/
"" - ""
Temp"nll''''' ;n "(;

Hgur" 2 fu....il'UIC..J of U,e w$tt1JliLilit y PCI' Rnlm cl nide! in the ""ighborhood cl the Curie
'"'
t"ml~l1Itul't! (35S"C). 'nle d..",,~ity i~ 1'. 1be daJ>cd JillC i~ IIli,,~'III' Cl(trapot...tion flllm high le.llp'.ft.-
ture~. (Arter P. Weiss And R. Forrer.)
...
The c'\ch.mge fidJ gives an ,lpproxim,lte representation of the CJuanlum-
mech.lllicai excllange inte raction. On certain assumptions it C'"dll be shown:! that
the energy of inter.tetion of almns i, j bearing electron spins S,. 5j contains ;1
term
(6)
where J is the exchange integral and is rel.lled to the overlap of the charge
distributiolls of the atoms i, j. Equation (6) is clilled the Heisenberg model.
11le charge distribution of a ~ys tem of two spins depends on whethe r the
spins arc parallel or antiparallel,4 for the Pauli plincipJc c:oocludes two eledrons
of the same spin fTOm being at the same place at the SimlC time. It does not
cxd'lde two electrons of opposite spin . Thus the clcct rostlltic e nergy of a sys-
tem will depend on the relative oricnt:ltion of the spins: the difference in en-
erg)' deflncs the c);change e ne rg)'.
TIle e);changc c nergy of t",.o elcctrons may be written in the form - 2JSI • ~
as in (6), just as if there were a di rect coupling between the directions of tile ro\lO
spins. For IIl .UlY purposes in ferromagnetism it is a good ;\pproximation to trc;;lt
the spins as classical anb,\llar momentum vectors.
We (';.Ul establish an approximate collnection between the exchange inte-
b"-") J and the Curie temper"ture Te. Suppose that the 'Itom under cOflSiderd-
lion has :t nearest neighbors, each connected w ith the cent!'",,1 atom by lhe
inter.JCtion]. For more dist.mt ne ighbors we take] .1S zero. The mean field
theory result is

] = 2zS(S + 1)
Better statistic'll <Ipproximations give somewlmt diffe rent results. For
the se, bee, and fcc structures with S =!. Husllbrooke and ' Yood give
kn7"J:t] ; 0.28; 0.325; and 0.346, respecti vely, as comp;\red with 0.500 from (1)
for all th ree structures. If iron is re prescnted by the H eisenberg model with
S = I , then the observed Curie temperature coll'esponds to] = 11.9 meY.

Temperature Dependence of tile Suturation Magnetir.otioll

'Ve can also usc the mean field approxi m;\tion below the Curie tempera-
ture to flnd the magnetiz;.ltion as a functi on of tem perdture. 'Ye proceed as

"See mo<t te~ts on quantu.,., theory: 31so j . II . Van \Ikdc. Ilcv. Mod . Pll)'S. 11, 'Z1 (1945). The
origin of ndwJSr in insulators is n)viewe..1 by P W Anderson in Rado lind Suhl, \'01 . 1.2.5 {1963}.
III metals by C. lierring in Vol. IV ( 1966).
· Ift-.."() spillS are anliparaUei. the _"('(unctions o(l ho: t-.."() e~trooll must be symmctri<', as in
the combination M{r,}t;{r ') + u{r~r,). If tI~ two spins ~ ~lIel. the Pauli principle n:qUII"eS
that the orb;L,] part o(tl-.e ...."vefuncloon be ,,,,tl1ymlTl('tric, as;n u(r,}v{r') - lI{r .)o:(r,), ror l>ne If
we inlerehange the l'OO!'d inalH r io r , the wavefim<'liDn changes sign. If we WI the positIons equid
>0 that r, - r., then d-.eantisymmelri<' function \"IUlishcs: ror paraDel spins there is:zero probabhl)'
of findinK tI-.e tv.u elrdmm at the SlIme pa lltion .
 15 l'erromagndiJ'" o,otl Antiferro"",g.",ti4.11 l-/,

"
"

,.
."- 08
"
.,
.' iKure 3 Craphil·,,1 solution of MI. (9) for til(! reduced mllJ{l1f:hution III lIS. fUlldioo"l of ternpen-
ture. 1be r..-dUl'CI1.n;tgnetiutioo Is tkfioed IIS.II ... /1m,'",. TIIC left·hand m\c ofEq. (9) Isiliotted;r.s
a ~'r~ight line'" with uni! ~lnl)C TIK! right-hand side is , .." I,{",/I) a'MI i> plotko \'s. III for three
<li\Ter... ,,! \'3iuc$ .,(the r ....:Iu('l~1 teml)!,ratu.e I - k/lTI1o.'p.2A ... TI1'c. TIle three eun'1:soom'spono to
the temperatures 2T~, T" ant! O.STc' TIle n'n'e lOr I '"' 2 lnt ....eeu the str.ught I;n" III ollly at
'" .. O. a5 approprutc IOc II>.- fIoU""m.ognetic relQc)n (t""re is noedl'rnal a&l\,IK" m.agnetie Ildd). TIle
run'e for' .. I (nr T = T,) i. "m,.,.·"t to the >-Inii,:lIt line "' at till" origin ; this , ..",.,.,ralul'l' murkslh"
onset of rerrom'~lIetism. TIll'rurve for t ... 0 5 11 ;11 thc ferrom.lgJlt:tie region and ;nterse<11 the
str,,;ght li ne '" Jt dbout III ... 0.9410.'",. As t - 0 till': intcn1:p' move. up to '" .. I , so Ihal ..J I
magJletic nMlmCllts .In: lined ,,\, :11 absolute um.

before, but instedd of the Cltrie \;IW we usc the complete Drillo\lin expression
fur the magncli7.ation. For spin t this is Ai :; NI-' t;mh(p.BlkoT).
If we omit the applicd magnetic field and rcplace B by tile moleC\llar field
8 e :: AM, then

M ::0 Np. hmh(p.AM1koT) . (8)

Wc shall see that solutions of this equation with nonzero Ai cxist in the tcmpcr-
"ture range between 0 lind T e-
To solve (8) we write it in terms of the reduccd magnctization III _ fllfNI-'
~lIld the redu<"Cd temperature t - k uT1N1L2A, whcnce
m = tanh(mlt) (9)

We then plot thc right and left sides of this equOItion selxlmtely us functions of
III, us in Fig. 3. The intcrccpt of the \\\'o curves gives thc value of III at the

tempcrdture orinteresl . The critiC".Jllemperahlre is t "" 1, or Tc "" NI-''lAlk o .

The curves of Ai versus T obtained in this way reproduce roughly the
features of the cxperimental rcsults, as shown in Fig. 4 for nickel. As T in-
creases the !mlgnetiZi.ltion decreases smoothly to zcro at T = 1'c. This behavior
classifies the usual ferromagnetidparolmagnetic tr.msition as a second order
trnnsition.
4-15

, o 0~
, ""- "-
, \0
.. \0
,,
,, ..,
"'.
Figure 4 Saturation rJ1;!l;netization uf nickel as a function oftemIJC.... ture, together with the theo-
retical eurve for S _ Ii on thc oncn" field theory. Experimcntal values by P. Weiss and R. .Forrer.

The mean field theory does nol give a good description of the variation of
M ,It low temperatures. For T <!I: Tc the argument of tanh in (9) is larl,'C, and
tanh ~ S!! 1 - 2e- 2t, ..
To lowest order the magnetizalion deviation l:J.M • M(O) - M(T) is
(10)
The argument of the exponentiaJ is equal to - 21".,ff. For l ' = O. lTc we havc
l:J.MINp. S!! 4 x 10- 9 .
The experimental results show a milch more t<lpid dependence of flM on
tcmpentturc at low tempet<ltures. At T == O.lTc wc have l:J.MIM 5!!: 2 x 10- 3
from the data of Fig. 5. The leading term in h.M is observed from experiment to
have the form
h.MIM(O) = A1.:l1'l • (Il)
6
"1lere the constant A has thc experimental v,due (7.5 ± 0.2) x 10- dcg-JI\!.
for Ni and (3.4 ± 0.2) x 10- 6 deg- 312 for Fc. The rcsult (11) finds a natural
explanation in terms of spin wave theory, us discussed below.

Saturation Magnetization at Absolute Zero

Tablc 2 gives representative values of the saturation mal,'lletization M~, the
ferromagnetic Cmie tcmpcrdtllre, and thc effective magncton 1)11mbcr defined
by MJ..O) =: IIUNP.B , where N is the number offormula units per unit volume.
Do not confuse 118 wilh the par.tJnagnetic effective magneton number p dclined
by (14.23).
 - .. ,

( ,
"''''1,1' .•
"
" ,,-
r
- !
," "
T(I\) <:lLt''' r."
'He "
'I-
,--
."...
Modell" ",e
- ",
He ",
..
I .,
'

,
T. 10." "
Fig ure 5 I),.:crease ill magllt.'ti1.ation of ,, ~kcl with tempe ..... lure , aRc,' Argyle. Charap, and Pugh.
In the plot t./If • 0 at 4.2 K.

Tablc 2: Fcrromagnctic crydals

Cur;':
It.fagnetWtion 101 .. in g;.ws
.,,,(0 "'1. temperature.
Sub>tar.cc: Jloom tempcmluTl" OK per ronllula unit I" K

F, 1707 1740 2. 22 HH3

Co 1400 1446 1.72 1388
Ni 4&5 510 0.606 627
Cd 2000 7.63 292
0, 2926 10.2 88
"l nAs 670 87' 3.4 318
MnUi 620 680 3.52 630
MnSb 710 3.5 587
C.-o, 515 2.03 386
MnOFcA 410 5.0 573
I'cOF~Oa
N iOl'c203
,,'27.I 4.1
2.4
858
858
CnOFc~3 135 1.3 728
M gOF~O:J 110 1.1 713
EIIO 1920 6.' 6!l
Y:J<'Q\O,2 130 200 5.0 580
 Obscrvc<1 values of nu ure often nonintcgrai. There are many possible
causes. One is the spin-orbit inter.Jction \\hich adds or sllbtracts some orbital
magnetic moment. Another cause in fen"Omagnetic metals is the conduction
electron magnetiz.1tion induced locally about a paramagnetic ion core. A third
u lUse is suggested by the drawing in Fig. 1 of the spin arrangement in a fcr-
rimagnet: if there is one atom ofspin projection - S for every two atoms +S, the
average spin is is.
Are there in fact .my simple ferromagnetic insul.lIors, with ,III ionic spins
parallel in the ground st<lte? The few simple ferromagnets known at pre~ent
include CrBra, EuO, ,md EuS.
A bum! or itinerant electron modelS accounts for the ferrom agnetism of the
transition metals Fe, Co, Ni. The approach is indialted in Figs. 6 and 7. The
rcbtionship of 45 and 3d bands is shown in Fig. 6 for copper, which is not
ferromab'lletic. If we remove one electron from copper, we obtain nickel which
has the possibility of a hole in the 3d hll1d. In the bum! structure of nickel
shown in Fig. 7a for T > T" we have taken 2 x 0.27 = 0.54 of an electron away
from the 3d band and 0.<16 away from the 4s bam!, as compared v.ith copper.
The band stroelure of nickel at absolute zero is shown in Fig. 7b. Nickel is
ferrom<lgnetic, and at ab~olute zero nu = 0.60 Bohr lTIagnetons per 'Itom. After
,,!10WiU1ce for the mab'lletic moment contribution of orbital electronic motion,
nickel has lin excess of 0.54 electron per atom having spin prefe rentially ori·
ented in one direction. The exch,mge enhancement of the susceptibility of
met"ls was the subject of Problem 14.6. The properties of amorphollS fe rromag-
ne ts arc treated in Chupter 17.

MAGNONS

A magnon is a quantized spin wave. \Ve usc a classical aTglllnent. jllst as we

did for phonons, we find the magnon dispersion relation for w versus k. We
the n quanti7.c the m<lgnon energy and interpret the quantization in terms of
spin revers,u.
The ground state of a simple ferromagnet has all spins pamllcl, as ill
Fig. &1. Consider N spins each of lnagnitude S on u line or a ring. with nearest
ne ighbor spins coupled by the Heisenberg interaction;
N

U '=' -2} 2: Sp' Sp+l

,-. (12)

Here } is the exchimge integrcll and ftS p is the angular momentum of the spin at
site p. If we treut the spins S" as dassiall ,ectors, then in the ground state
Sp'Sp+1 = S2 and the exchange energy of the system is Uo "" -2N}S 2.
5C. II crring in Rado and Suhl, \'01. IV (J966).
 r! Fe"", ....w.,
2l~v
7 I ~V ~

L v
1+ t I

'" ."
""
I.
~',Ikd- JO ehi"" .. 5ck..:,,,,,,,

'"
5ckd,,,,,.

Figure 6a Schematic rd.. tionsh,p of 4.t and 3d hands in '1lC!;ollic (.'OIlpe1"- 11.e; 3d band holds 10
electrons pcr Mom IIntl is filled ill c"l'per. 11,11'; 4s hand CIi" hold two declroos per atom; it i. ~1)()Wn
Iwlf·llllW. "" copper has une ''liM-nee <'1edron oubiJ" the filled 3d ~hdl.

Figure 6b The mIN 3t! hnlld dropper .ho....n as two ""--par~It: ruh... b;uKk of ul"l'0.lIe elcdroo flliu
orientation, each 0000 holding fhe eI('(.1rons. W,tI. hoch Mlb-h"mls filk...:! .l> \hown, the "et spin
(and hence the net m.lg" d iu tioll) of the ,/ bmd IS zero.

* ~.
-GS-m
4.73 ..,.,,""'"

"" "' I
"" Ib) '"
Figure 7a B.md relntionshil'5 in nickel abm<e the Curie tcmpcmture. 'The net lTI<ognt:lic moment is
>:ero, a' th(,TC are «I"al numbers ofiJOlC5 in both 3d 1 and 3d T b::mds.

Figure 7b s.:hcmatic rcLtionshipofwnds in nickel at ,obsolutcttrO. nH~ em-rp:ies of the 3d f and

3d 1 sub-bands arC separJted bY;In och,mgr: mll'ra.1 ion. The 3d f ban<! is fllk"ll; the 3d 1 bond
<.'Ontains 4.46 elt.'t1rons and 0.54 I~. n", 4, ha",] " usually thought to eonlaln aPl'roxl"~tely
e<:1"a1 numbers of ..!t.-drons in both spin di~k:o's, und W) w"
h.t~ not lroublrd 10 di,'w .. 'I inlo
sub-b.u,t1s_ Th e 'WI mnl:nctic "'OInenl ufO 54 1'-8 per alom ari>cs from Ihe ex<."C:1S pOllu lal1Of) oflhe
3d T band o,er Ihe 3d t b... w It 11 often eon,"enient to 1jlC".Ik oCthe magnctimtlOn as ansing from
Ihe 0.5-1 hole in tI", 3d 1 ba,W.
452

,
What is the energy of the first e,;cited state? Consider. _excited state with
one pal"ticuhlr spin reversed, as in Fig. 8b. We see from (12) that this increases
the eller~'Y by 8}S2, so that VI ;: Uo + 8}S 2.
\Ve can form ,In e,;citation of much lower energy if we let ,111 the spins share
the reversa]' as in Fig. Be. The elementary excitations of a spin system have a
w;welike form ,lIId aTe called magnons (Fig. 9). These are analogous to lattice
vibrations or phonons. Spill w;wes are oscillations in the relative orientations of
spins on a lattice; lattice vibrations are oscilJ,ltiollS in the rehltive positions of
,ltOlns on a lattice.
We now give a classical derivation of the magnon dispersion relation. TIle
terms in (12) which involve the pth spin are
- 2}Sp-(Sp_1 + Sp+l) (13)
We write the magnetic moment at site p as PI' ;: -g}.t8Sp' Then (13) becomes
(14)
which is of the form -PI" HI" where the effective magnetic field or exchange
field that acts on the pth spin is
(15)
From mechanics the rate of change of the angular momentum "Sp is equal
'1
to the torque Ill' x HI' which acts on the spill: dSpldt = PI' x HI" or
dSplcll = (- l!JtalM SI' )< 81' = (2}1"XS,>X S,>_ I + Sp)< Sp+ I)' (16)
In Cartesian components
dS:J(u = (2}"I)(S~(S~_1 + S:, ~ I) - S;,cS~ _ 1 + S~ ' I )]' (J7)
and similarly for dS~hll and dS~hll. These equations involve products of spin
components and are nonlinear.
If the amplitude of the excitation is small (if S;, S~ <t,; S), we may obtain an
approximate set of line,11" eqllations by taking all S~ = S and by neglecting
terms in the product of S~ and SII which appear in the equation for dszlcll . The
lineali7.ed equations are
dS;hll = (2}S/" )(2SX - S~_ I - S~+ I ) ; (1&1)
dSr/dt"" -(2}SIfI)(2Sr, - S~ _ I - S~ + I) (ISb)
dS;Jdt = 0 . (19)
By ;uwlogy with phonon problems we look for traveling 'w,IVe solutions of
(18) of the fonn

S; = u exp[i(pku - wt)) ; Sf, = v e,;p(i(pku - wt)) , (20)

where u, v ,Ire constants, p is all integer, and u is the lattice constant. On
substitution into (18) we have
 ....

.....111111 111111
.... . >-
,-
~

0' ,<,
Figure 8 M C~ical picture of" the gJOIUld state of ~ simille ferromagnet ; all silins are Ilarallel.
(hl A IlOUiblc C>titatioo, one spin is re\'U1oed, (el nil' "".~h'ing ele mentary exeltations a~ ~Il;n
"'lI,·es. n,e ends nf thc spin vt.'dOB pra'Qs on the S,"f.1('C1i of C(lncs. with ~ll«'C$si\'c .pins ad,
vant~ in phaj,e by " w nslan! angle.

CD
I

Figure 9 A S&lin "-."'con a lined lillinli. (a) n,e silins ,,~cd in periilll'Cll"e, (hl Spins "icwed from
ai.JO>-e, showing One wa,·elength. TIle wa''t! is drawn through till' ends ur th" spin ''CCI.on.

-/W/I = (2}Slfl)(2 - e- ;.\:" - eiJ,,,)v ::: (4}SIfI)(l - cos ka)v ;

- iwv = -(2}SMX2 - e- 1b - e"4u ::: -(oi}S/h)(1 - cos ku)u
These equations h.we a solution for It and 0 if the determinant of the
coefficients is equal to zero:

~4}SIIi)(1 - cos ka)1 = 0 ,

I iw
-(4} SlftXl - cos ka) 'w
(2 1)

whence
Ilw = 4} S( 1 - cos ka) . (22)
TIlis result is plotted in Fig. 10. With this solution we find thllt v = - ill,
co~sponding to circular precession of each spin about the z axis. \\le sec this
on t<lldng Telll p.'1rts of (20), with v set equal to - ill . Then
S~ = u cos{pku - wt) ; S~ = u sin(pka - wI)
Equation (22) is the'dispersion re hltion for spin w... ves in one dimcnsion
with nearest-neighbor intenlctions. Precisely the same result is obtained from
the quantum-meclmniCi.,1 solution; see QTS, C h,'pter 4. At long wavelengths
ka « 1, so that (I - cos kul a i(ku)1!. .lI1d
trw Eli: (2}Su 2lk 2 • (23)
The frCl:}lLency is proportional to k 2 ; in the S<lme limit the frequency of a phonon
is proportional to k.
·..
• l

~~~---------------i

Figure 10 Dis l>ersion reL1!;"" for ma)(nOn) In Hr"1ro""'gncl In onc dime"'''''' Mlh ncare,t-neigh-
bor ;nlernction$.

The dispersion rel.llion for a fC ITOmagnetic cubic lattice wit h n('.lres!·

neighbor interact ions

(24)

where the summation is over the z vectors denola.1 by S which join the ccntral
atom to its nearest neigh~ors . For ka <C 1,
I,w = (2J5..,'" (2.5)
for all three cubic lattices, where a is the lattice conS!;ln!.
The coclJicient of ~ often may be determined accurately by neutron sat!-
tering or by spin \,"owe resonancc in thin films, Ch,lpter 16. By neutron scatter-
ing C. Shirullc .lIld coworkers find, in the equulion liw = Df.!l, the values 281,
500, ;\Ild 364 meV ito?- lor D :It 29S K in fe. Co, Ni, respectively.
Quutlti::ation of Spin Wave$. The quantizat ion of spin waves procectls eX-
actly as for photons ,1I\e! phonons . The energy ofa modc offrequcncy WI< with nl;
m..gnons is givcn by
{;J. = (nk +~ )!JWk • (26)
nlC cxcit-<ltion of a magnon corresponds to the reversal of one spin i.
 15 Fe~romagJJCtis", "nd AfltijerTomagnctism

Thermal Excitation of !\IagRom

In thermal equilibri um the ave rage value of the 1U1Inber of m<lgnons ex-
cited in the mode k is given by the Planck distribution:6

(27)

The total number of mHgnons excited at ~I temperature T is

.t J nk = dw D(w)(n(w) , (28)

where D(w) is the number of magnon modes per unit freque ncy nmge. The
integr;J is taken over the allowed r..mge ofk. which is the first BrillOllin zone. At
sulTIciently low temperatu res we may carry the integral between 0 <llId X> be-
cause (n(w) ~ 0 exponenti;Jly ~IS w~ (Xl .
Magnon.s Iwve a single poi<U'i7.ation for each value ofk. In three dimensions
the number of modes of w~lVevector less than k is (l/21T)l(41TP/3) per lmit vol-
ume, whence the number of magnons D(w}dw with frequency in (lw at w is
(1I2 1Tf'(41Tk 2)(clkldw) dw. In the approximl\tion (25),
2
dw = 4]Sa k =
2
2(
2]Su )112 w l12
dk ft ft
Thus the density of modes for magnons is

D(w) = _1_(_'_,_)312 w l12

(29)
4~ 2]Sa 2
SO that the lohll number of Imignons is, from (28).

.t k - 4~
n - -1- -
( -- I, )" L"" dw elY>'" -
2]St?
w
l12

1
=-
4~
1 ----
2
(knT)'"
2]Sa
1"" dx e' -
0
Xl 12
1

l11e definite integnll is found in tables lmd has the v;Jue (O.05B7)(4 ~.
The numher N of atOins per unit volume is Qla l , where Q = 1,2,4 for sc,
bec, fcc lattices, respect ively. Now (I,nJlNS is eqmli to the fmctioll<li c111luge of
magneli7_<'\tion tJ.MIM(O), w he nce

MI
M(O) ~
0.0587
SQ .
(kDT)'"
2JS
(30)

"111c argu ment is t'l«lct ly as for phonolls or photons. The PI;iJld di~lnbulion foll""", rOT ally
problem wh('re the energy t...... els are Identical with tll.OSC <:i n h"rmollic OSCIllator OT collf:ction of
harmOl1 ;c oscillators.
.,.
This I'C!)ult is the Bloch 1-312 law a nd has OC'C1l collfinncu ¥Apcrimcnlally. In
neutron scattering experiments spin waves have been obsclVl,.'(l up to tClnJX.'l'3-
lures ncar the Cmie temperature and CVCll above the Curie tCIllpl.uturc .

NEUTI10N MAGNETIC SCA1TElUNG

An ;>,. -ra)' photon sees the spatial distribution of electronic charge, whether
the chal'gc de nsity is magncli7.cd or unmagnctiz(.'1l. A Jleutron sees two aSJX'Cts
of a crystal: the distribution of nucle i and the distribution of eleclrollic magneti-
7-3li0I1 . The neutron diffraction pattern for iron is shown in Fig. 11.
The magnetic moment of the neutron intl.'facts with the magnetic moment
of the electron. The cross section for the neutron-electron interaction is of tlle
s.'lmc order of magnitude as for the neut ron-I1Uc\(.'8 r inl(.'ractioll. Diffraction of
neutrons by a magnetic el'Ystal allows the determinatioll of the dislIi bution,
direction, and order of lhe magnetic momellts.
The magnctic momcllts associated with particular compollenls of alloys
may bc invt.'stigdted by neutron diffraction. Results for the Fe-Co binary alloy
S}'l'itCIll (\\ hich is fcrromagnetic) arc shown in Fig. 12. nlC magnetic 1Il0ment on
the cobalt atom doc'S not apJX>3r to be affected by alloying, but that 00 the iron
atom increases to abou t 31-' B as the cobalt concent rat ion increases. TIlc magnet-
i:l..ations arc shown in Fig. 13.
A neutron can be inelastically scatte rcd by the magnetic structure, with
thc creatioll or an njhilatiOll of a magnon (Fig. 14); such events make possible

I
(lICij

Scot'~ri"l! ""gIe
"-igu." II Neutron dilfra~l ioll pattern for iro". (Aft"r C C. Shull. E O. WolI;",. ~"d W. c.
Kot:hler.)
 - ....,

(
I

I ~o
I
, ,
i
J': 1,0

t1gure 12 Moments Bttdblllabl~ to 3d ~I«­

Irons In F ..... CU alloys &$ g f,,!lClion of cornpoo:J-
lion, After M. f . Collins and J. B. Fonyth.
"
I'~'
.,
o:t.oJl - "'"
,
..
e fe-V
, .. toe<:.

, ~
Lt.:: ... o F~ ·NI

0'=
~

o N!-CoI
IO~
, W •
~
....,
• Nl·z"
• N~V
"=
/ 'b Ni-Mn

,
.,
/ I
//
•
~ .-......
. ~~

• Co-Mn

,c, ,.,
I IV 1" 1
, ,
M"
•
!:led..... o:oncmtrati<:wl '"
I'igur e 13 Average lltomic ~nll of binary aUoys of the elements in the iron group. (Afler
Bowrlh .J

}' igure 14 So:atlering of. neulron by an ordcno:d nl~lic •

siruct\,re, "ith creation of a magoon. M....

-
".
"
0 (100)
A (IJO)
a ( III)
"
>"
,
!
.i 0.'1,
Mnl'\,

"
,,
''''' "'" 1;". in}.,
...
I
."" 0010

Figure 15 MllgllOIl enel)l;Y III B function of the MInar., of th.. Wllv~or, for the ferromagnet
Mil l\.. (Arkr B. Anlonini and V. J Mi,,~iewicz.)

the experimental dclcn ninatioll ofmagnoll spectra. Utile incidcll lncutron has
wavcVL'Clor k" and is scattcTt.'d to k~ with the c reation of a magnon of W3\'Cl'CC-
tor k, then lJy conservation of crystal mome ntum k" ::I k:, + k + G, where G is
a reciprocal laltice vector. Ry conservation of energy
h2~ fi2k~2
-- = -- + "Wi. (31)
2M" 2M"
where It ...... is the energy of the magnOIl created in the process. TIle observed
magnoll sfx:clrulIl for MnPt 3 is shown in Fig. 15.

fE HIUMAGNEllC OnDER

III many ferromagnetic crystals the saturation magneli7.3tion at '[ :::: 0 K

does not correspond to parallel alignment of the magnetic moments of the
constituent paramagnetic ions, even in crystals for whk:h the individ ual pam-
magnetic ions have their normal magnetic moments.
TIle most familiar cxample is magnetite, F e304 or FeO· FC1/93' From
Table 14.2 we see that ferric (Fc 3+) ions are in a stale with spin S = t and zero
orbital moment. Thus each ion should contribut e 5MB to the saturation mo-
ment. TIle ferrous (Fe 2 +) ions haye a spin of 2 and should contribute 4MB. apart
from all)' residual orbitaJmOll1ent contribution. Thus the effective number d
Bohr magnclons per Fe304 formula unit should be about 2 x 5 + 4 "" 14 iraU
spins were parallel.
The observed value (Tab le 2) is 4. 1. The discrepancy is accounted for if the
moments of the Fe3+ ions are antiparallel to each othe r: the n the observed

--.- -
 ")

~~~ 1111111III !J 111,-,·,

f'igu~ 16 Spin ur.1ngemt'nl.$ in magn~tite, FeO . FetD:.. st..;,wi"g how the moments of lhc Fe'"
ions can«:l out, I~,·jng only the m~nts ol the Feu ;0111.

moment arise.~ only from the Fe 2 + ion, as in Fig. 16. Neutron diffraction results
agree wilh this model.
A systematic discussion of the consequences of this type of spin ord er was
given by L. Nee! ,'lith reference to an important class of magnetic oxides known
as fcrrites. TIle usual chemical fonnula of a ferrite is MO . Fei 03. whe re M is
a divalent cation, often Zn, Cd, Fe, Ni, Cu, Co, or Mg. The term fcrrim agnctic
was coined originall), to describe the ferrite-type ferromagnetic spin order such
a'i Fig. 16, and by extension the tenn covers almost any compound ill which
some ions have a moment antiparallel to other ions. Many ferrimagncLs are
poor conductors of electricity, a quality exploited in applications such as rf
transformer cores.
TIle cubic ferrites have the spinel crySlal structure shown in Fig. 17. There
are eight occupied te trahedral (or A ) sites and 16 occupied octah edral (or B)
sites in a unit cube. TIle laUice constant is about 8 A. A remarkable feature of
the spinels is tlmt all exchange integrals J/I.A, lA/J' and ) /Ju are negative and
favor antipnl"Clllel alignment of the spins connected by the interaction.
But the AB interaction is the strongest, so that the A spins are pal'alle! to each
other and the B spins arc parallel to each other, just in order that the A spins
may be anliparallcl to the B spins. If) in U := -2)S," 5J is positive, we say
that the exchange integral is ferromagnetic; if] is negative, the exchange inte·
gml is antiferroma.g.etic"
\Ve now prove that three antiferromagnetic interactions can result in fer-
rimagnetism . TIle mean exchange fields acting on the A and B spin lattices may
be written
(32)
taking all mean field constants A, p., II to be positive. The minus sign then
corresponds to an antiparallel interaction. TIle interaction eneJl,'Y density is
U:: -1(B.-\·MA + BB"l\f B) = V.Ml + P.MA"Mn + !JlM~ ; (33)
·'"

• AI"

Figure 17 Crystal $Irud~ of the mint'rnl spinrl MgA\,O.; the Mg" ions OC'aJpY letruh.,dral
lik'S. eac-h su rrounded by fou r Ctt)'gen iOlls; the A'" OC'C'U",," Odaho.-dral .ites. e.ro IUrmundP.d by
sbe QJCyge" iOll5. This Q; a nonnal.pinel arrang<-TI>enl: the di...aJoo,nt metal ions occupy lhe tetr.ahMnl
)itft. In the in'~rse spinel arr41lgC1Tlm1 the tctraItcdn!J )ites are OttUpied by trh'lll~t metal ions,
whiJ.: the 0<.1aho:dnol site'i ~ ocmpicd half b) dival""1 IIKI half by triVAlen t metal Ions.

Ihis is lower whcll M" is antiparallel to MB Ihan when M" is parallel to M/.I. 1be
(..'I lcrgy when antiparallel should be compared with zero, because a pos.<;ible
solution is M" = AlB = O. 11lUS when
(34)
the ground state will have M... directed oppositely to MlJ. (U mlcr ccrtain condi-
tions there may be nOllcollincar spin arrays or
still lower e nergy.)

Curie 'l e mpemhlre and Susce,ltibility vf "e rrimagtlets

\ Vc d efine s(!parate C urie constants C... and CD for the iollS on the A and B
sites. For simplicity, let all interactions be zero except fo r an anti parallel inter-
action between the A and B sites: BA "" - ,..,M Il ; 8 0 = - ,..,!\fA, where IJ. is
positive. TIle same constant IJ. is involved in ooth expressions because o f the
form of (33).
\ \'e have in the mean field approximation
(CCS) (35)
where E" is the applied field. 'I11ese equations have a nonze ro solution for M"
and M8 in zero applied field if

~CAI ~ 0 (36)
'[ ,

so that the ferrimagnetie Curie 1t..'Illpcralul'C is given by "re; = IJ.{CA C B )I12.

 · ~-

".
,,
,,
, /'
.,/' -- f.-"'"

, /
•, 1/
, '-, / .. Tcrnpenu~re rei

Figure 18 Rmprocal sulOCptibility of lnagr>elit~. FcC)· FeA

We solve (35) for M" and AlB to obtain the susceptihility at 'f > Tc:
M... + Mn (CA + GB)T - 2p.C"Cn
(eGS) x= B"
=
'P - 'r:
(37)

a result more complicatt.'d than (4). Experimental values fOr Fe.:J04 arc plotted
in Fig. 18. The curvature of the plot of lfX versus T is a characteristic feature of
a [crrimagnet. \Ve consider below the antifcrromagnctic limit C... "" CD.
Iron Cortlees. The iron gamc..'U arc cubic fenimagnclic insulalors with the
general fonnula M3FcsO I2. where M is a trivalent metal ion and the Fe is the
trivalent ferric ion (5 = t. L = 0). An example is yttrium iron garnet Y3FcSOI2.
known as YIC. He re y :H i~ diamagnetic.
The net magncti7.atioll of YIC is due to the resultant of two oppositely
magncti7.ed lattices of Fc 3+ ions. At absolute zero each ferric ion contributes
±51-'8 to the magneti7.ation, but in each fonnula unit the tllH,'e Fe3+ ions on
sitl.'S denohxJ. as d sites are magneti7.oo in one SL'flSC and the 1....'0 Fe3+ ions on a
sites are magnetized in the opposite sense, giving a resultant of51-'8 per formula
unit in good agn..-ement with the mcasuremt.'1lts of Geller et at
The mean field al an a site due to the ions on the d sites is Ba =
-{1.5 X l(J')Mdo 11le observed Curie temperature 559 K of fi G is due to
the (I-d interaction. The only magnetic ions in TIG are the ferric ions. Because
these are in an L = 0 state wilh a spherical charge distribution their interaction
with lattice defom)ations and phonons is wt.--ak. As a result fiG is characterized
by very narrow lint.' widths in ferromagnetic resonance experiments.
In the rare-carth iron garnets the ions M3 + are paramagnetic trivalent
rare-earth ions. Magneti7.atioll curVeS arc given in Fig. 19. nle rare-earth ions
OCCUP)' sites labeled c; the magncti7.ation Me of the ions on the c lattice is
opposite to the IIet magnetization of the ferric ions on the a + d sites. At low
temperatures the combined moments of the three nlH.'-C3l"th ions in a fonnula
46.

"~-----------------;-----I

1i 101+'- - -- - - -- - - --(<1) 1·1

(c) - --I
,:
1
i
>
5

T, in Ii:
Figun! 19 E~I)[;Timent.al values cil\lr. ""turation magl1<-1i .... tion venul It'm~rnlure of various iron
garnets, .ft~ It I'authmet. The i:lrm,,1a unil is MJFe,O I'. ",hen M is a In>'a1ent metru ion . n,,~
It'mperatu~ al which I~ maguetOOllion ~ zero.is can~ I~ compensat ion temllcrnlure; here
the magnetiZ<llion of!he M sublau ice i$ l'ilual atw::! 01"1'(..;10: to the net magnetization of the fffric ion
subL>Uices. I'nformula unil there are3 Fe3 ' ions On l etraJM!dralsid.-s d ; 2 Ft'~· jon. on OCIahed~1
sites (I ; ~nd 3 Mh ioo$ on sil" denoted h)' c. Tbe ferric jotlS coolribule (3 - 2)5~8 .. 5".." per
funnula unil. , ,,.,,, ferric ion coupli ng Is liltong and determines Ih.. Curie Icmpernture. Iflhe MJ+
iOlls are ,..~arth iUII5 Ihe)' are IIlllo$olncti~>d opposite 10 the re~uh;t1\1 of the Fc3 • ious. The M3 •
w ntrihutioll dropli rapidly with increasing tClIlpemlu re bCCllu$fl the !\.I.Fe coupling is weak- Meal;·
uremcnlS On singie Cry.llai sl'ccimc.'ll III'C reported by Gdler d ..J .• Phyo;. Rev. 131. 1034 ( 1965).

unit may dominate the net momcut of the F~~ ions, but because of the weak
c-a and c-d coupling the rare-t:arth lattice loses its magneli7.ation rapidl)' with
increasing temperature. TIle total moment can pass through 7.cro and the n
increase again a~ the Fe3 + moment starts to be dominant.

AJ'I.""TIFElutOMACI'IETIC onDEn

A classical example of magnetic stmcture dcll.'rminaHon b), neutrons is

shown in Fig. 20 for 1\·1 00. which has the NaCI structure. At 80 K thc~e are
extra neutron reflections not present at 293 K. TIle reflections at 80 K may be
classified ill terms of a cubic unit cell of lattice constant 8.85 A.. At 293 K the
reflections correspond to an fcc unit cell of lattice constant 4.43 A.

\
i

' 00
I (I II) "',I) ,1) (S1l)
I
... - 8M}.
-
"
", - .OK

, - -~ ~
.J(311)~
'''I' ( II~
I
(Ill)
"'i'" I
.. - U3}.

" M"" ,."

"
-

, --~ \
.,. ,
'" """""".d<
Figut"\:: 20 N,.,ull'()ll diffraction pattaos 1m ~lnO be low RIld allOV~ the 'pin>{)fdering tempernture
of 120 K, after C. C. Shull , W. A. StrJUsH. and E, O. Wollan. The reflection ind ica are bast.'<1 On
an S.SS }. cell at 80 K and on. 4.43 J.. ~1I .... 293 K. At the higheT telllpf'nl"re the Mn u i01l1 arc
still m~netic. but they an: nO looser ordered.

But the laUice constant determined by x-ray reOection is 4.43 A at both

temperatures. 80 K and 293 K. We conclude that the chemical unit cell has the
4.43 J.. lattice parameter, but that at 80 K the e lectronic magnelic moments of
the Mn\!+ ions are ordered in some nonferromagnelic arrangement. If the Or·
dering weTe ferromagnetic. the chemical and magnetic cells would give the
same reOections.
l1le spin arrangemen t shown in Fig. 21 is consistent with the neutron
diffraction results and with magnetic measurements. The spins in a Single 1111J
plane are parallel. but adjacent (Ill] planes are antip..'lrallel. Thus MnD is an
anlifcrromagne t, a'i in Fig. 22.
1n an antiferromagnct the spins are ordered in an antiparallel 3lT3nge-
ment with zero net moment at temperatures Lelow the ordering or N~el tem-
perature (Table 3). The susceptibility of an antifermmagnet is not infinite at
T =< "f"" but has a weak (.'u sp, as in Fig. 23.
An antiferromagnet is a special case of a ferrimagnel for which both sublat-
tices A and 8 have equal saturation magncti7.ations. l1ms C/o = Cn ill (31), and
'"

,
I

:G ""il.,,1

Figure 11 Orden.'<I ar~e'llls oC ~pin~ of'lhe Mn t • Knll I" fIl3.l'P'JeS<' oxKIe, MilO, as deter-
mined by neuiron diffraction . ...}'" 0 '- ,0'" are not Sho:w.~l •

..~" & 1'-

FIgUre 22 Spin ordenng in fem:Jm~gnets 0 > 0) and autiferromagnets U < 0)

1'><""!Iaj;f'Icti.m F~rn'ru~.bm Anti"""""""",..t.,.,,

S""""",Lili'~ /I{ X
.,, X

~ ~=i\
,
,,
I
I
" C
T
" T.
C
-, , 1 '~
C
T
X -- X ----
T T - T, X'"' T+"7
Oooie ..... c..n.·\\'e'" .... (T> T~ )
(T > T,)

Figure 23 T(IllIlt'rature dependeoo: orthc mOJgilelir liwuptiuility ill paruill3f;l1eU, femllnagllets,

alld al\tifcrrom3"nel~. Below Inc N~II"'mJll'ralure of an ""('ferrumagnet the spms ha"", ""l'paral-
lei u.i('n I3Iio"5, the suseel'lib,lity attainl its m,,~jmum v:tlue lit TN where there is II ....-ell·dellned
kink in the CU I'\'e of X ,'e,-sus l ' 'J1M' Imn.ition ;s also m~rked I», peaks ill the heat C1l[ll1Clt)' Ilna the
thennal eKp:msioll l."OCff>cient.
 ',," "y

Fv.........gndw.. "nd ""'if~rOO>Jaf:""""m <6,

(
"
Table 3 Anlir~rromagnetic: cryslals

I'arnmagll<-'tic
Tramition
telllJlftl'turc, Curie-Weiu
, -<I1"L
Sumtan«: iOtl lanice TN. in K 6, in K T., xlTN)

116 610 5.3

M"O
MuS
M"Te
""
r~
h~x, la)'cr
160
307
528
600
3.3
2,25
!
0.82

MnF! be letr fiT 82 1.24 0,76

FeFI
FeCI1
be Ictr
he., layer
7'24 117
48
1,48
2.0
0,72
<0,2
FoO 570
CoCl, ""
hex, layer "'"
25
291
38.1
2.
1.53
O.S

CoO
NiCl 2
NiO
""
he.. , la)'cr
r~
50
525
330
68.2
- 2000
1.14
1,37
-,
-
C, boo 308

the N6c1 temperature in the mt..>an field approximation is given 1.,)'

-
(38)
where C refers to a single sublaltice, TIle susceptilJilit)' in the paramagnetic
region T > TN is obtained from (37);
2CT - Zj.l.C 2 2C 2C
(39)
x= f2 - (p.C)1 = 1'+j.l.C

TIle experimental results at T > 'fN are of the form

2C
(CCS) (40)
X = T+O '

orr
Experimental values of N listed in Table 3 often differ substantially from the
value unity e:l:pected from (39). Values of orfN of the olJserved magnitude ma)'
be ohtaint..'(\ when next-nearest-neighbor interactions are provided for. and
when p()ssible sublattice arrangements are considered. lf a mean field con-
stant -E is introduced to describe interactions within a sublattice, then
OITN = (p. + E)/(P. - E),

SU8Ce1Jfihility Belo1V the Neel Temperature

There are two situations: with the applied magnetic field perpendicular to
the ruds of the spins; and with the field parallel to Ihe axis of the spins. At and
aLove the Nb!1 temperature the susc<:ptibilily is nearly independent or the
direction of the field relative 10 the spin axis,
 For Dol perpendicul,lr to the nxis or the ~pins we can c.lk11late the SUM:Cpti-
bility by elementary (:'Un~ideratio"s . 'Ine (''1wrgy density in the presencc or tile
field is, with AI = IM. . I = IMsl,
U ::: ,uM ....·MB - D..-(M .... + Mil) iii - ,uMZ[l - f(2rpfl - 2B"Mrp. (41)
where 2rp is the anglc the spins 1ll,Ike with each other (Fig. 24a). The energy is
a minimum when
dUltfrp = 0 = 4,uM zrp - 28".U , if' = 8 ,/2,uM (42)

so that
(ces) (43)
hl the parallel orientation (Fig. 24b) the m,lgllctic energy is not changed ir
the spin systems A and B make <:(Illal angles with the field . Thus the slisc.:epti-
bilit~ nt T = 0 K is 7.ero:

XI(O)=O. (44)

The pnrallel susceptibility incrcases smoothly with tempe ri\tul'c up to TI\'.

Meas ure ments on MIlF~ arc shown in Fig. 25. In vel)' stron~ fields the ~pill
system s willturu disc.:ontinllollsly from the lk'1I11Ud orientation t(l the 11Cll>cn-
diclllar o ricntation where the e nergy is lower.

Antlferrornogneric Mugnonll
\Ve obtain the disl>crsion rc!lltion of 111.'1gnolls in a ol1c-dimensional.mtifer-
romagllet by making the appropri.lle substitutiol1s in the treatment (16)- (22) of
the ferromagnetic line. Let spillS with even indi<:cs 2p C'Oll1pose sublatticc A,
that with spillS III' (S~ = S); and lei spins with odd indices 211 + 1 compose
sublattice B. that with spins dowu (S~ .= -5).
\Ve consider only nearcst-neighbor internetions, with} negativc. "nen (18)
written for A becomes. with a careful look at (17),
dS;'/<Ir = (2JSIhK-251,. - 51,._, - SI,..,) , (450)
dSVdt = - (2} Slh)( -2 S~ - S~ _ l - S~,.+I) (4Sb)
The corresponding equations ror a spin 011 8 are
dS~+Ilc1t "" (2JSlh)(2S~+I + S~,) + S~,)+2l ; (46.'1)
dStJ ~ ,/dt = - (2JS/~ )(2 S~"+I + S~ + S~p +2) (41ib)
\Ve rorm 5+ "" 5~ + is\!; then
dS;,Jdt = (2 iJSlh)(2S;" + S;p_I + S;;,+I) ; (47)

dS;P+lldt = -(2i}Slh)(2S;"+I + S;" + S~~2l (45)

\
-

"
~t5.-
~t2¥
------
Figure 24 Calculation of (a) p<:rpcndicu-
lar and (b) panolle! ~" >ccptibilitics at 0 K.
in the mean field approximation.

,
,'i. ",..,
•",,
,
'00
"

'"
" 40 80 120 160
T. In K
200 240 2Ml 320

F~re is Magnetic suS«ptibility ci mang.lln CIie fluoride, MnF,. parallcl and IlCTJlCOdlCUbrtu the
le' .... gon.I"" ... (Afi.,r S. t'oncr)

\Vc look for solutions of the fclTm

5;,. = u exp(i2pkt/ - iwtJ; 5;,,+1 =vcxp(i(2p + Ilku - iwtl , (49)
so that (47) and (48) become, with W"X - - 4}Slft = 4lJISlft,
wu = IW",.(2u+ vc- u.:" + ce'~... ) (SOn)
- wtJ = IWex (2v + ue- ;.I:o + Uc'M) (SOb)
Equations (SO) have a solution if

We. COS kal = 0; (51)

We. + W

thus Wi = w!.{l - cos 2 ku) ; W = wnlsin ka] (52)

The dis persion relation for magnons in an antifclTomagnct is quite differ·
ent from (22) for magnons in a ferromagnet. For ka <c I we see that (52) is
linear' in k : w !ili we.lkal. The magoon spectrum of RhMnF3 is shown in
Fig. 26, as determined by inelastic neutron scattering experiments. Thcre is a
large region in which the magnon frequency is IineaT in the wavc"cctor.
' A IIh ~ical discuulon of lhe dIfference between the disl'crsion ...,btions lOr rcrrml\a8nctJC and
."ltrerromagnC:lie magnons is gwen by F. Keffer. II. K",,,,,,,. and V. Varel. Am. J. Ph),s. 21 . 2.50
( I ~l.
'"
"

,•, "'
M

f" "
,•!
'"
Figure 26 Magnon di5pCf!;ion relation ill the ';ml'le
cubic antifcrro,nagnct IIbMnf'3 as dctcnnincd at
" "
\\'It' e,,,,,lor k, ,,' A- I
04 4.2 K b)' inelast ic '\(.'\llron scallc';ng. (Aficr C. C.
Windsor Mnd II. W. II. Stevenson.)

\Vell-resolved magnons have been observed in MnF 2 at specimen temper-

at ures up to 0.93 of the N&I temperature. Thus c\'cn at high temperatures the
magnon appro~ima ti on is useful . Further details conceming anlifcrromagnclic
magnons a TC gh'cn in Qrs, Chapter 4.

}' J:o:RROMAGNETIC DOMAINS

At temperatures well below the Curie point the electronic magnetic mo--
menls of a ferrolllaguel arc essentially parallel when r cg.1Tdcd 011 a microscopic
scale. Yet . looking at a specimen as a whole. the magnetic moment ma)' be very
mILch less than the saturation moment, and the application of all external mag-
ne tic field may be requi red to s,'lturate the spccimell . The behavior observed iu
polyerystalliue specimens is similar to that in single crys tals.
Actual specimens arc composed of sillall regions called domnins, withill
each of which the local magncti7.ation is saturated. The directions of magnctiz.'l-
tion of d ifferel1 t domail1S need not be parallel. An arrangement of domaills with
approximately zero resultant magnetic moment is shown in Fig. 27. Domains
fonn also in :mliferromaglletics, fcrroelectrics, antiferroclect rics, fcrroclastics.
supcrcollductors , alld s()metimes in metals under conditions of a strong
de H aas-van Alphen cflcct. The increase in the gross magnetic momcut
of a fe rromagl1e tic spt!cimen in an applied magl1etie field takes place hy two
independcnt processes:

• 111 weak applied fields the volume of domail1S (Fig. 28) fa"orahl )· Oril.'lltoo
with respect to the field increases at the expense of ullfa\'orably orietltoo
domains;
• In strong a pplk>d fields the domain l11agueti7.atiou rotates lo\\o.... rd the direc-
tion of the field .
 .

Figure 27 F(."nt)fllagnClic domam pauem on a single n ),lb11l1at"kt of niel d . " ,., domain bound·
arks IIrc mllde v~iblc by the IIIUe,' magn"u., pov.-dcr pattern teclm iquc . The din'CIi<)n ol magld;·
zalion y,i lhin a .lUlnain 15 detenninf'<l b,' obs......inl! gro.."th or C(lnb acCion of the domain in II
magn el;c Ikld. (Arter II. W. I)., llIois.)

ro,. 1Io..

11'1"e\,..... i!k
bou,..tory
di<tllore"",nl>

t' is"'., 21:1 Ilcp rc~nt3ti'e m:>gnl:til:OltJOfl cur,'e, Sh(l\<i ng the OOnnM.nt magllCll:calion procCSM'5 in
tiM! diffe.cnt rcgion~ of the CUI"\'C .
 M~t ... ",11".. h.,., (~)
B _ '1 ~4".\ f

_ _---/~-- 8, + 1/

'.

Figure: 29 The trt·lonirni r~n"(izalIO" ll.ln'" . 11,e rocrci>"it)' ''< is the fC\.'fne rICk! that .-...l,,(.'~.-,; 8
to zero. II. n:b.tl>d coef'l'i'l it) II", reduces M or 8 - II to lI!CIO. n~ remanence 8 r ;s the \lIluc of B
",II ... 0 'noc saturation indudioll 8, .~ Ih" hUll! or R - /I HII~rg<l If, and the ~turalion magnc,hu-
lioll M. - /J ) 4,". \n 51110., . '"rtit·..! ....;. i. B = I'J)f + ,\fl.

Technical tenus dcflnt..'il by the h~ ... t('rcsis loop lIrc showlI in Fig. 29. '111C
(''OCrdv;ty is us ually defined as the re\'e rse field He Ih.11 reduces the inductinn B
to zero. slarling from saturation. III high ('oercivil), materials thc cocrcivity If.:.
is defined as thc reverse field that reduces the magncl i'l.atioll .11,1 to zero.
'(1IC domai n s tniciurc of rerromagnetic matcri31s en ters their applications.
In a transforme r core we wanllo\\l losses in taki ng th~ core aro und a cycle: this
means that we want low coercivity, \\hich goes wilh a high pcrmedhility. By
making pure, homogeneous, .md well-oricnted matcrifll we facilitate domuin
boundary displacement ,md thereby attai n high permeability. Relative pennea-
bility values a.~ high ,IS 4 x HfJ have been rcported.
In a permanent maglUet we want a high coercidty, which we may achieve
by s11ppressing boundary displacement, which is best done by eliminating
domain bollildaries. This is accomplbhcd ill very fine particles or crystallites
thai consist of .1 ~ ingIc domain, \lithoul OOundaries. Bycontrolled precipitation
of a sC<.'Ond metallurgical phase the specilll(.'n may be heterogeneous IlIl a vcry
fine 5(:3Ie. Characteristically there is a critical r.ldius bdowwhich single domain
attributes are found .

Ani,wtroP!I Energy
There is an CI1Cq!;Y in a ferroma gnetic crystal which directs Ihe magnctiu.-
lion .11011g certain cryst.lllographie axes called directions of .... L'y magnctizatlou.
\

---
 .." 11001
~ ~111 --
soo 11111I
!V \ V
......
1
1200
t/'
'00
11 /
3OO ~.tf!OOI

.001....,,-
IU"'~I
- +_1
..,,1--+ -.-1''-+-

0 , '". "
"',---d
.oo.--.,oovo-,>oo",
B. (gwl>'
Fi.,'u n· 30 :\Ia~etlul illn ('I'n e ~ for single el)~tals ofll'on . lIIeld. Jlld rob.t)1 F,'OIl' Ihe t lln'e~ rot
;ru" "" noe 11J.lt tnc !100] dm.."CI'ons ale (•..,y din"'tio,,' " f magnc t,:-", tion and the [Ill] oIirL'Ct iQ,,,
Mro' lwnI direchons The awl'L...! r",k! .. 8 •. (Anc. Iloodil alld Kay...)

,.

G- ,,'
Figure 31 AS)'mmch), of the O\'erlap of ek'Ctrou (li stnbutions on nl'ighbori" s: ions pro\ides olle
",cChaO'~01 of mMgnetocl)'SlaJline anisotropy. lk(,ao~ of silin-orbit iolcl1loClioll the ch,1rKe oI;\I.i],,, ..
tioll IS spncmidal ""d IIot sphc ri<"ll. The as), m,nel.y IS tit·d to the diredion uf Ihe ~pl" , so th~1 II
roWtion of the spin d'n-ct'ons relalive 10 Ihe <:ry.IMl :ues ch~"g"l the e"",h~go: ene rgy 111\01 ~ lsu
d".n~e~ the clcctfO'\l..li<' inler.lction encrJo()' of the d,argc di~trib"tlo" s on P"'<5 of al""'5. lloth
o...ffL'<:ls gi\l: rosc 10 an anisotropy energy. The energ)' of (It) ;s 110t the WIllie as Ihe eIHlIlW of (b).

This e nergy is called the magnetocr),stalline or anisotropy cnergy. It docs not

come about from the pure isotropic c;t{change interaction considcn.'ll thus far.
Cobalt is a hcxagonal crystal. Thc he;t{agonal ruo::is is thc direclion of ('asy
ma~netization at room temperature, as shown in Fig. 30. One origin of the
anisotropy energy is illustrated by Fig. 31. '"l11e magncli.t.1tion of the crystal
sees the crystal lattice through orbital o\"crlap of the electronS: the spin inter-
nets wilh the orbitru motion by !lIeans of the spin-orbit couplillg.
In cob.,]t the anisotropy cnergy density is given by
UK = K; sin 2 9 + K; Sill~ 9 • (53)
whe re 8 is the angle the magnetization makcs with the he;t{agon.ll ruo::is. At room
temperature Ki = 4.1 x HJ'i erglcm3 ; K; = 1.0 x lOS erg/etn 3 ..
Iron is a cnbic crystal, and Ihe ('"be edges are the directions of ea.~y mag-
netization. To rep rescnt Ihe anisotropy cnergy of iron magnetil.ed in an arbi-
ti'al'Ydirection wit' lin_"Ction cosines at. Ql . a3 rc ferred.to the cube edges. we
 (
arc guided by cubic symmetry. lne c)(pressioll for the anisotropy energy must
be an even power of each a" prodded opposite ends of a crystal a.~i s arc equiva_
lent ITUlgnetically, and it mllst be iu\'ari.mi under intcrch:Ulgcs of the 0, amung
themselves . The lowest order comhinatiol1 s..... tisfying the symmetr y rC(jllirc_
men !s is of + ~ + 03. bllt this is identically cqu(ll to unity and docs not de-
scribe anisot ropy effects. The nc);1 comhinalioll is or the iClIIrth degree:
afa~ + af~ + ~~, and then of the si;>,lh degree: a~a~a5. Thus
(54)
At room temperature K. == 4.2 x lW erg/emu and K2 ::: 1.5 x l(f> crglcm3 •

Transition Ref,'ion Between Domains

A Bloch wall in a crystal is the transitioll layer that separates adjacent
regions (domains) magncti:£oo in different directions . 'Inc entire change ill spin
direction bchveell domains does nol occur in one discontinuous jump across a
singlc alomic planc, but takes place in a gradual way over many atomic planes
(Fig. 32). 111C exchange ellergy is 100ver wheu the change is distributed over
mAny spins.
This Ix:havior may be 111lderstood by intell1reting the Heisenberg eqt13tion
(6) classically. We replace cos (/J by I - i~ ; thcll U)e~ = )S'1.Ip'1. is the e){Challge
energy between two spins making a small angle Ipwith each nthct'. Here) is the
eJiehange integral and S is the spin quan tu m 11ttmber; U).,~ is referred to the
energy for parallel spins .
[f a total change of ."occurs in N equaJ steps, the angle between ncighbor-
inR spins is ."IN, and the cxchange energy per pair of neighboring atoms is
W~. = )S 'l(."IN'f. Tlle total exchange energy of a Iiue of N ·j 1 atoms is

(55)
The wall would thicken without limit were it \lot for the anisotropy energy ,
which acts to limit the width of the transition layer. '111e spins contained within
the wall arc largely directed away from the axes of easy magneli7.ation . so there
is an anisotropy enefb')' associated with till! wall , roughly proportional to the
wall thickness.
Consider a ,vall parallel to the ctlbc lare of a simple ctlbie lattice and
separating domains magnetized in opposite din.octions. \\'e wish to de termine
the number N of atomic planes contaiued withll1 the wall . The energy per uuit
area of ,vall is the sum of contributions from ~challge and anisotropy energies:
0'"" == (1.,. + O'. ni•.
The e){change energy is given approximatel) by (55) for each line of atoms
normal to the plane of the \\',tli. lncre are 11(,2 such lines per unit area , where (l
is the latt;Cf! constant. Thus u~~ = u 'l)S2/Ntr.
 . . ..
N

Figu,'e 3i 11\<.: , tructutc of the llIoch "'all !I(:var~ tiog domaio~ In irou tl,c th;cI<HC~S ofthc t ransi-
tion rcgiorl is about 300 laUil"C <.'OustantJ.

-111c anisotropy energy is of the order of the anisotropy umstant times the
thid<,III.."Ss Na , or u,,,.. ... KNa ; therefore
(56)
"111is is a minimllill with respect to N when
(57)

0'
(56)
For order of magnitude, N ... 300 iu iron.
TIle total wall ellcrb'Y per unit area on our model is
if" = 2n(K)S 2/U) 1I2 ; (59)
in iron if" . ... 1 crglcm 2. ACClirate calculation lor a 180" wall ill a (tOO) plane
gives i f", = 2(2KJJS 2/a) tl2.

Solitons

Bloch walls are an ex.'1mple of a (.'1 sdualing class of physical and mathemat i-
cal problems lmown as solitons. A rough deSCription of a classical soliton is a
sol ita!")' wave thai shows great siabilil)' in collision with othcr solitary wa'·es.
Bloch walls in the absencc of applied magnetic fields are slatic solitary wa\'cs,
bnt they movc uniformly in au applied field aud their t'Ollisions eml be stndied,
at least theoretk~ny. It is sumewlhlt surprising that this cml be done because
the equations of motion arc mmlinc<lr. 8

Orih,in of Domains
Lmdau and Lifshitz showed that dOlll.lin structm'C is a natural conse-
qut'nce of the various contrihutiOllS to the energy- exchange, anisotl1lpy, and
magnetic-of a ICrromagnctic hody.
Direct evidence of damaill structure is furnished by photomicrographs of
domain boundaries obtained by the technique of m'lgnctic powder p..1tterns and
by optical studies usmg F.lraday rotation. TIIC powder p.1ttcrn mcthod de\c1-
oped by F. Bitter consisb in placing a drop of .1 colloidal suspension of finely
divided ferromagnetic materi.ll, sudl as magllctite, on the surfolCC of the ferro-
magnetic crystal. TIle wllnid particles in thc suspension concentrate strongly
about the oounddries between domains wherc strong local magnetic fields exist
which attract the magnetic p,uticics. The discovery of transparent ferromag-
netic eOlnlxlIIllds has ellcouragcd the use also of optic.11 rotation for domain
studies.
\Vc may understand the origin of domains by considering the structures
shown in Fig. 33, cach representing a crO'iS ~ection through a ferromagn etic
single cl")'s tal. III (a) we have a sillglc domain; as a c()llsequence of the magnetic
"poles" formed on the surfaces of the crystal this configuration will have a high
valuc of the magnctic energy (1I81T) f 8 2 dV. The magnetic energy density for
the configuration shown will he of the order of M; = 10'; e rglcm 3; here .\1$
denotes the saturation magnetization, and the lin its are CCS.
In (b) the IIIHgnetie energy is reduced by roughly one-hali" by dividiug the
crystal into two domains m.lgnetizoo in op()()site directions. In (c) with N do-
lllolins the magnetic cllergy is reduced to approximately liN of the magnetic
energy oi" (a), because of the reduced sp..1tial extellsion of the field.
In damaill arrangemcnts such as (d) and (e) the magnetic energy is ;r.ero.
Hcre thc I)()undarics of the triangul.lf prism domains IIcar the cnd faces of the
crystal 1l1.J.e equal angles (45") with the maglll:ti7~1tion in the rectangular do-
mains and with the magneti7.atioll in the domains of closure. "lIe component of
magnctil.atioll normal to thc boulldary is euntinuotls across the bound.IlY and
there is 110 mugnetie field associated with the maglleti;r.atioll. TIle Ilux circuit is
completed within the crystal-thus giving rise to the term domains of closure
fur smfacc dom.lins that complete the flux circuit, as in Fig. 34.
Domain structures arc uften morc (.'Omplieated than our simple cxmnplcs,
but ({omaill stl"llcture alu;(jys lias its origin ill tlie possihility of lowerillg tIle
energy of a system by gOingfrom a .wturatcll confiWlration with liiglilllllgnetic
ellergy to a domain configllratioll with a lower ellergy.
blWc<:1l1 monO)!raphs on solilonlbeo,", tndllUC Il. K. Bllllough, SoJil()n~. Springer, 1980. G L.
LJ.Il'(,. Hcmc,," of rolilon IlIre ....' . WIley. 1980. For" ,llOr, illtnxlu(1ory arllel<: , ><:<: Il . K But-
lou~ , 1'I1), i<;. Bullctlll, J'l'hruar)" [978, PI'· 78-82.
 {,:::\
.'1 N S S

-
I t I III t t I till
s ~ .'1 .'1

~
,.

nJ!urc 34 Duma", of closu re~! the end or a .mgle l'f')'Stai iron wtuSkel . l11C r...::e ;s • (100) plane;
the .... biJokcr a.\is is 10011. (Courtesy eM II. V. Coleman , C . G. Scott, anti A. Isin .)

Cocrdvity unlillyllterellill
The cocrcivily is the magnclic field He re quired to reduce the mng.. cti:t. ...l-
lion or the induction B to ....ero (Fig. 29), 'l1'e vnltlc of the cocrdvity ranges mer
scven orders of magnitude; it is the most sensith'C property of fe rromagnetic
materials which is subject to control. The col'rcivity may vary from 600 C ill 11
loudspeaker pcnnancnt magnet (Alnico V) and 10,000 C in a ~ r>ccial high stabil-
ity magnet (SmCOs) to 0.5 C in a commercial ()()\\lcr transformer (J'e-Si 4 wi.
peL ) and 0.002 C in a pulse transformer (Snpermalloy). l ow coercivity is de-
sired in a transformer. for this means low hysteresis loss per cyde of operation .
Materials with 10\.... cocrci\·ity are called soft; those with hi~h cocrd\"ity are
called hard, although there is not nccess.'\rily a I: 1 relationship of magnetic
hmuncss with mechanical ha rdll CS~ (Chapter 20).
TIle cocfeivit y decreascs as the impurity collleni decreases ami a1so as
internal strains arc rcmon.'d by annealing (slow cooling). Amorphous fcrromag~
netic alloys may h th'C klwcoclci\"ity, low hystercsis Imscs, and high pcrmcabil-
ity, as treated in Chapter 17. Alloys that conl.lin a pl"(.'Cipitateo phase may ha\'c
a high (.'OCrci\"ity. as III Alnico V (Fig. 35).
176

Figure 3S Mlcrostnll.iure or Alllico V in its optimum .t.'te~. ~ perm!U1ent magnet. The compos i_
tion or Alnico V is, by weight , 8 AI, 14 NI, 24 Co, 3 Cu, 51 Fe. As a pcrTTlllJlen t magnet it i~ a
h>o-phase system , wi th fine partick'l or one phlOSC cml,eddcd in th.~ otht.'T phll$C. The prcdpJtlltion
is csrried out in a mag.It..... ;.; fw:kl, alld the poIrtides are ori ented " i tl, their long axis parallel to the
fiekl direction . TIoe ",idt}, shown is 1.1 ,..In. (Courtesy d F. E . Luharsky.)

Sort magnetic materials are used. to concentrate and shape magnetie flux .
as in motors. generators, transformers, and sensors. Classes of sort materials
include c:Ieetricai steels (usually alloyed with several percent of silicon to in-
crease c1C(.1.ril'al resistivity and to decrense anisotropy); various alloys of Fe-Co-
Mn. starting with permalloys of composition ncar Ni78Fe22. whieh have near
zero anisotropy enel"b'Y and ncar zero magnetostriction ; NiZn and ~lnZn fer-
rites; and metallie glasses produced by rapid solidification methods (Ch:lpter
17). A commerdal metallic glass (METCLAS 26055-2) with composition
FelllB I3Sig has a hysteresis loss per t.)'de much lower than the best brrain ori-
entccl silicon steel.
11le high coercivity of materials composed of very smal l gntins or fine
powders is ....-ell understoOO. A sufficiently small pnrtide. with diameter less
than 10- 5 or 10- 6 cm. is ahvays magnetized to saturation as a single domain
because the fomlation of a flu:o:-<:Iosure configllllltion is energetically unfavor-
able. In a single domain particle it is not p:v.;sible for magneti;r~'tion reversal to
take plat.'C by means of the prcx.'Css of boundary dispinccment, which usually
requires relatively weak GeMs. Instead the magnetization of the particle must
rotate as a whole. a process tllat may require large fields depending on the
anisotropy energy of the material and the anisotropy of the shape of the pm1i-
cle.
The coerdvity of fine iron pmticlcs is cJ(pcctClI theoretically to be about
500 gauss on the basis of rotation opposed by the crystalline anisotropy cnergy.
ami this is of the onler of the ohserved value. lligher t.'Ocrcivities have been
reported for elongated iron particlcs, the rotation here being opposetl by the
shape anisotropy of the demagnetization energy.
 Rare earth flU/.
in allo~ with Mn, Fe, Co, and Ni Ilave very large crystal
Misotropics and correspondingly large cocrcivities, of the order of2KJ/Il . These
alloys arc exccptionall), good permanent magnets. For e);''\mple, the hexngonru
compound SmCos has no anisotropy energy 1.1 x uP erg em - :I, cquivruent to
acocrcivity 2KJM 01'290 kG (29 1). Magnets ofNd 2 Fe ,.,B hm'e energy products
as high as 50 MGOe, eXcccOing all other commercially available magnets.
Friedberg and Paull< developed a theory of the pinning of domain wnIls in
single phase polycrystnIline materials that gives a good account of experimental
results over a wide range of materials. In particular their theory gives a good
account of the high cocrcivity ofSmCo:;, predicting He = 9 kG where 10 kC is
observed.

S I ~CLE DOMAI~ PAlI'nCLES

TIle dominant industrial nod commercial applications of ferromagnetism

are in magnetic recording devices, where the magnetic material is in the form
of single domain particles or regions. 'Ibe total value of the production of mag-
netic devices for recording may be comparable with the total vnJue of semicon-
ductor device production and greatly exceeds the value of superconducting
device production, the latter being held back by low critical temperatures, as
compared \\;th magnetic eune temperatures. ·111e magnetic recording devices
or memories 1)'picaUy are in the form of hard disks in computers and tape
cassettes in "ideo and audio recorders.
An ideal single dorn.."lin particle is a fine particle, usu.'\lIy elongated , that
has its magnetic moment directed toward one end or the other of the particle.
Tile altemalive orientations may be labeled as N or S; + or - ; in digital record-
ing, as 0 or 1. To have digital properties a ferromagnetic particle should be fine
enough, typically 10-100 nm, so tllat onl y one domain is within the particle, lO
If the fine p..-utide is elongated (acicular) or has unia-.;ial c rystal symmetry, only
two values of the magnetic moment of the single domain arc permitted, which
is what one wants for digital p roperties, 111e first successful recording materinJ
was adcular "T-F~03 with Icnb>th-to-width ratio of about 5: 1, coerch'ity ncar
200 Oc nod a length < Lf-tm; c1uomium dioxide CrO\! is the b.."\Sis of a better
material, II in a form highly acicular (20: 1) with cocrcivity near 500 Oe.
Effective e1ong.'\tion CM be attained with spheres by making a chain. like a
string of beAds. An ensemble of such dmins or of elongated Single domain
particles is said to exhibil superparamagnetism if the magnetic moment of a

QR. Fricclhcl)I; alld I) I 1'11111. I'h)'$. !kv. I..e ll. 34, 1£3.1 (1975).
10c. Kitt~1. ·Ihrory of the stnld:ure of rCl'l"OtYlllgn{'tic .Iurnrun~ in films 1111<1 small psrlidc$, ,.
I'h\S Rev. 70, 965 (19-16), · PhYSICSl theory of fcmmlllgnctic oomalns:' Ren. Mod "h >'li. 21,
541-SS3 (19-19).
"R M. \\~u lc. ~ \Ialcriab aspect: or magnetic r<'COrdong. ~ J. Mag. and Magnetic Mal. 88.
165- 1;6 (1990).
unit is constant. If p- is the magnetic moment ill a magnetic field Ii , tllen the IIt't
magneti7..'Ition of the ensemble "ill follow the Curie·Bnllouin-Langevin law of
(1-1. 19), (14.20) if the particles arc embedded in a liquid so that they are each
free to rot.'Ite as a whole. If the particles are fr07..cn in a solid, there will be a
remanent magncti7..'Ition (Figure 29) after removru of aJ l applied field.

Geomaglletism and Biomaglletis",

Single domain ferromagnetic properties art: ofspcdal geologic;11 intcrclot in
sedimentary rocks because the rocks through their remanent mngneti"..'Ition
carry a memnry of the directioll of the earth's maglletie field at the time that
they were laid down, and thus of the geographical lociltion of the rocks at that
epoch. The magnetic reoml is perhaps the mo~t importilllt bnsis of the theory
of the drift of contincnts. Annually. layers of ~Ctlimen t arc deposited in stream
beds, layers that mny benr some magnetic particles in ~ingle domain form. This
record persists ovcr at lci.lSt SOO million years of gcologieil! time and can tell us
wherc on the surface of the eru-th the deposit was laid dO\\11 at a b'iven lime.
Lava nows also record m ..gnetk field directions.
The eh.ange in magnctizatioll from bycr to layer givcs ,'I sllperb historical
record of the drift of the continental plntcs 0 11 the earth's surface. TIle p.-tlco-
magnetic rerord is one basis of the branch of geology ca1lcJ plate tectonics. 111e
original interpretation of the rceo"l was made more difficult, or more excitifl g,
by the associated discovery (Brunhes, 19(6) tlwt the magnetic field of the earth
itself can show fevc rsrus in direction, an efft'Ct contained within the standard
d}nnmo thenry of the earth's mllgnetism. Reversals hnve wken place once
e\ ery I X 1ij4 tn 25 X loG years. When a revcrsal OC{.'lrs, it is relatively sud-
den.
Fine single domain p ..rticlcs, on en of magnetite Fe304, nrc even of imp or-
lnnce in biology. A direction-seeking effect known as magnclotllxis II!. oftell con-
tmls. possibly sometimes along with an astJollomical gUide system, the molion
of b.'lCteria, the migratiou of birds, and the movements of homing pigeons aJld
bees. The effect is due to the internetion of a si ngle domain particle (or cluster
of such particles, Figure 36) in the organism with the external ffi.'IgJletie field of
the earth.

Magnetic Force MiCT08COPfJ

The success of the scanning tunneling microscope (STM) stimulated the
de\elopment of relnted scanning probe devices, of which the scanning mag-
netic force microscope is olle of the mo~1 crfecth'c. A sharp tip of a tTJ.'\SIlctie
material, sudl as nickel. is mounted on a cantilever lever (Fib'Ure 37). Ideally.

l~T"ri.r in t>Lology lIleanS the IJlO\·(.-mcnt of a ttll or orgamsm loward or :tw.ly from wme
c~temal fidd of /Or<:e.

--
 or aeen
Figu re J6 Thtn SoI..'Ction ul.l .l'I.lgJ>ctutActic IlIiCtcrtum ~h"winga~'hmn tloo
nm p,.rti<:ies ul
F~O •. Or.ming by ~larta l'uebld from a photogr.oph by Il H. Frankd .lilt.! ullw:n

Flgure:.rr Hasic COliC..!'! of magnet;e furre microscopy. A nldW,ctic tip ,\ltach.~1 10 a IIc~tblc
QnliK:ver;. used 10 c\ch'Ct the magnetic r.eld produced II)' the: n,gions or Mltcmalmll m3J:tlcti"l.lllion
til the pluue of the 5lllUl'ic. (Afte r Groette •• Mamtn. and Rug;.r. \992. )

but not yet. the tip is a single domain pnrticle. Forecs from the md~netic
sample oct on the tip and rouse a change, such as n deflection, in the cantilever
st.'\tus, and an image is formed by scanning the sample relative to the til), '111e
magnetic force microscope (~'IFM ) is the only magnetic imaging technique that
c:ln proVide hi~h resolution (10- 100 11m) with little surt'1CC prepamtion . One
ron, for example, observe and image the m"'l8"etic flux that exits from the
surface at the intersection of a Bloch wnll with thc surface (Fib'Ure 32). An
import-mt application is to the study of magnetic reeording media- Figure 38
shows the magnetiC sibTJIal from a tcst pattern of 2 ,...m bils magnelll.w in the
plane of a Co-alloy disk; tlJ(~ parallel component of tllc fil'id sensed by the
sensor tip is what the photo shows.
,0

Figure 38 Test d rip mal,'tl('l i23tion ill the plane or II Co-alloy disk in 2 pm hils. 8$ <Icleded hy
M FM close alxJ\C the plane of the di~k. (After f1ugar ct at )

MACNETIC BUBBLE DOMAINS

Magnetic recording discs and tapes ofTcr the lowest cost form ofmnss data
storage or memory. 'l1w ir dmwback is that in order to write or read data they
requi re mechanical motion of the head with respect to the mab'llctic material.
'l1lC mechanical motion involves an electrical power dlive. slow access times,
and periodic maintenancc.
E lectrical methods of entry and read-o ut using semicondllctor memories
a m avoid these difficulties. However, semk-ondllctor memories arc not truly
permanent- they lose their data if all power is removed.
Magnetic bubble domains in a ferrite or garnet film can be used to record
data. and a sequence of bubbles can be shilled by electromagnetic signals.
Buhble memories are nonvolatile, which means tlmt they retain data without
need for electrical pm\ er.
Small cylindrical magnetic domains (bubble domains) can be stabilized in a
uniaxial magnetic material by a bias magnetic field normal to the plane of the
specimen and parallel to the easy axis of magnetization. 11lC formation ofbub-
bles is illustrated in Fig. 39. Magnetic bubbles arc used for high density mem-
ory storage, for example in portable personal compu ters.
Below a critical value of the bias field a circular domain as in (b) is unstable
with respect to a serpentine domain p<"lttem , as in (a). At still higher fields the
bubble is unstable with respect to a saturated configuration. The energetk"S arc
treated by Bobeck. 13

13A II. Boh<.><:k, Bell S)sl. Te<:h. J. 46 , 1001 (19(7)

 Figure 39 Fonnnlion of IlUhhlc Oom~in. in a garnet
film 6 ,.... thick: (.) unmagnl.'t in:d, (hi tlO C. (After
A. II. BOOcck amI E. Dell.. Torre.)

SUMMARY
(Tn CGS Unil5)

• -nle susceptibi lity ofa ferromagnet above the Curie temperature has the form
X ""' C/{T - ToJ in the mean field approximation.
• In the menn field approximation the effective magnetic field seen by a mag-
netic moment in a ferromagnet is B" + AP-t , when A = Tde and B" is the
applied magnetie field.
• 'nle elementary excitations in a fe rromagnet arc magnons. Their dispersion
relation for ka ~ 1 has the form liw "" J ~at in zero external magnetic field.
11le thermal excitation ofmagnolls leads at low temperatures to a heat cap.."\C-
ity and to a frnctional magnetization change both proportional to ~.
In an antiferromagnet tv."O spin laUices are equal, but :mtiparallcl. In a ferri-
magnet two lattices are antip.."lraJleI , but the magnetic moment of one is larger
thall the magnetic moment of the other.
• In an antiferromagnet the susceptibility above the Neel temperature has the
fon n X = 2C/(T + 8).
• l11c magnOIl dispersion relation in an antiferromagnet has thc form liw - Jka .
111e thermal excitation of magnons leads at low temperatures to a term in P
in the heat capacity. in addition to the phonon term in r.
A Bloch waJl separates domains magneti7..cd in different directions. The thick-
ness of a wnl l is = UIKD~ tl2 lattice constants, and the energy per unit area
is .... (K}/a)llt, where K is the anisotropy energy density.

Problems
L MagllOn disllcrgicm relatiun. Deriw th e magnan dispersiun re lation (24) for a spin S
on a simple olbic lattice, :t - 6. Hint; Show first that (IBa) is rcplllCl->tI by

dS;ldt - (21SIIi)(6S~ - L S~ ~iJ

•

..
 ....
.,
where the Ct~ntral atom ,s at I' and the ~ilr uc:r.rcst nt'ighho rs me connected to it t.y ~ix
\ccton: 8. Look for solutions of tl,e L'quatiol1~ for dS;;lfl and dS:'1tI1 of the fon,.
('xp(,k . I' - iwt).

2. 1I('{lf ctl,)QCify of magllol'S. U~C the approx imate magiloll displ'rsion rdation w = Al?-
to find UIC leaJing term in the heat C'.J Il..'tCity of a three-uimcllSlOllaJ ferromagne t :I t
low tClllperatlll"("S k8T 4, j. The resu lt is 0. 113 k~k/jTfIiA 'f1, (Jer UIlll vo lume. The
zet.'l function tlltlt e nters the resu lt may I>c c~t ill1alt"-' nunwri<;all y, it is t.'1uulatro ill
Jah,,!..c-ErnJ (,.

3. Ntel temperature. T'lling r..he effe<:tivc fields on the Iwo-sub!aUire 1I10J('\ of all anti-
fcrro rnagnclic as

-(J - -I-'-+-E
T... II. - E

4. Mogllctoelastic coupling. In a cubic cryst.l1 the clastic ellell()' JellSity in h."rms of

the lI~lIal shain components l''l is

U'" "" iC II(e~. + e~ + e":J + ! C~ie~w + e:" + c..) + CI ~ewt''-' + e... e... + e..e..,) ,
3m] tile 1C'J.oill~ telln III the ma!{IIe tic anisottopy ffieill:Y density is, from (54),

Coupling between cl3.')tie st'-dil, J.nUmaglletization dil cctiOIl may IJ.c take n fon nall y
into tl(.'count by indlluing in tliC totul energy dells;ty a tl'l1ll

V c "" 8 1(~eo + .}:#w + ~e.J + B ~('t l('t:#,v + u2u;jt:W' + ~u,e ...)

arismg from tl ,c sh~in d epenoencc of U,,; herc B, and 8 2 arc cilled magndol'iastlc
coupling cvns tJ.nU. Show tlmt th e total e ncrgy is a 1II;lIillIlIlII whcn

(I '" J)

n, is cKpiailis the origin of magnetostriction. the ehnngc of Icn),111 Of\ magl'll.-lizatioll,

5, Coe1"(;iceforee of a IlIJUlIl parl icl~. (a) Consider a small spl,erical sillgle-uomain IJalli~
de o j a IIIlwial fcrromagnet . shO\\ th ut the l"C'\«' rse fidu nlong thc axis required to
rL"\'Cf'Se tile magneti7.ation is 8 . - 2.KiM•• 111 CGS unit5, nO('
OOl'rdve IOrcc or single·
domain particles is observed to be of ti,ls magnitude, Take U" - K sin~ 8 115 the
'lIli~otropy energy dcn~ty and V\l '"' -B.Ai COS lias the intcrnd:km energy density
w,tll the external field ; here (J is !lIe angle bct->"t'Cn B. and 1\1 lIint, EKpand the
energil'~ for small angles about 8 - 'IT, and fin.1 thc value of 8. for \\hic\o V" + V II
dot's not have a minimum near (J - 'IT, (IJ) SllOw that the magn etic energy fir a satu-
J'i",n: 40 A Bloch ";Ill and It Ntd ",all in a tlun film The magnetization In the BlCI(h wall IS
nor",,,IIO Ihe plane "r the film and adtl~ to Ihe ",,,II c, .... n!'> II dcnll'gne!il'1lh"" ,'nj~rgy M:&! per
um! 1e1lwh of ....,.!I, "here 6 is lite walllltlclmcss and d the fil m Ilti"kllt'S,_ In lite :':l'd ,, ~1I II ..,
"'_'W,,-'i17.ah,," is pardJld to th,) ~",facc , tbe ",1,lition to Ih e "all cne,!otY is .-..:gloglhle "hen d 46.
The adtI'hon 10 Ihe Ne.,1 wall energy " I,,'n
,k·Ul(lt.o\; )
d'" 6 IS Ih .. sob;ect or I'roblc", 7 (Ancr S ~1Id.

rotctl Spllel"C of, Ii.tme ll'r ,/ is - M ~(f, An arrong" II1cl lt .,ith app.-t!cidhJy Je~s m.!gllc!ic
,,"CI'gY has a sillgjc wall in an Ctll.lalorial plane. 'n,c <10"14;11 wall energy "ill be
TrfT..JiI4, where iT", is the wall Cl'1l"'lo'Y pe r unil .lrea. Estimafe for (:01",1.11 the criti<:al

radius helow which the p..lrtickos are stdhle as single domains, taking th e ",,1ue of
JS% M for irOn.

6. Sail/ration mag"etiz" tiQII "'XI! T~, SllOW tlmt in the l1)(':1.n fie ld .lll pro,il1l.ttiun the
sal uration nlagl,eliwtion just below the C urie telllpcratun' has the £lolII;lIa llt t(::1\1-
pcralure ucpendentt (T~ - T)'~. Assume the spill is • . 'nle result ili the s,-une as t[lolt
for II ~ntl-onll.·r tral15 itiull in II ferrock't1rk CT\ st.t1 , dS tliiICusse.1 in C[lllpter 13. TIIC
experimental unta for fe rromagn e ts (fable 1) suggest th ~lt the expollelll i~ closer to
0.33.

7. Nhlwall. Thc tli"-'Clkl11 of mBg:netizalioll change in II domain wall goes rrom th.lt IIf
the Bloch wall to that 01' II !'\cd wall (Fig. 40) in t[lin films 01' mateti<l[ 01' nrgligible
crystalline anisotropy enellO', such us Permalloy. 11le intercept of Ihc Bio(:h \\,,111 wilh
lhe sutfacc of ti, e film CrI.'>It..,s II SlJrF"cc n'l!I0n of h igh demllgneti7.ation eilerII.)'. ·11,c
Neel wall avoids Ihis intercept contribution . bllt .. t tile t"xpense of II tlel1laglwti7~'l.tion
~'OntriblltiOl1lhrOllglIO\Jllhe volume of the ....-all_ l11C Ned wal l becomes energl.'tj(;ally
(d,'or"l;le whell the film becomes sumcientl)' thin . COllsiclcr, how~'vcr, Ille e nergetics
oCthe Nee1 wal[ in bulk malerial of n~gligi ble cr;. st.111ine an isotropy t'ner~y , ' nlerc is
now a clemagnctiLo1tion conlribu tion til the wall cller~y J..,,,sity. By a qu,llitative
argUIl)(.'fl t sirnil.1r to (5t;). show that iT.., .. (~JS ll.\'al) + (21TM~Na). Find ,\ for "hk-II
u" is II minim llm. Esti mate the o rtl 'T of magni tu de of iT", fm' typic;11 \ol,les of} . AI••
and a,

References
E. Dena Torre and A II , Bob.:... k. Magmfic bubbler, North· Jlollal ..l, 1974
A II Esch~"rdd..,r. Magnnlc buWe IW",olOt;Jl, Sp";"gt"r. 1981
A Il erptn. Thtorie tl .. moC"fflmW!, P..."._ uni,,,-'n,tar.rt'S. 1968.
f." KeITer, ~Spin \Va_s." Enq,da. of "/'!I'i~ 11112 (I966).
C . K,lIeI anti J K Galt, ~Fcrrollla.l""'I'" donwns,~ solKI JI~tl' I'I,)~i 437 (1956).
G . T. R"do 800 II. SuM. ed., Alagl>eII.!IIl, humic Pre~s . n~K: encyt'iull('thc wool;
K J. SI.,,,,dky, O~id" magn~tic ml1l~ril1k 2"d cd., O~f<I<1I . 1U72.
S V \'of)J()Vl;kii. Magnetism. llaisk'·d , 1975, 2 vol
II t.l Wh,te. QllontW» '!.ro.!! of magnet ISm. 2nd (.'(\ , Sp'''',,'''. 1963.
I\. M \\'I"te Ilnd T. II. Gcballc. Vmg nmq onlc. III 1011,/.,. Academic, 1971l.
K. r. Sinha and N. Kuma.-. 1"leroctwns 111 magnetical/v or,l"rc-d IOIfdr, O~rord. 1900.
I) C Matti •. Theon; of magnetism, Spring'~I', 1965.
P. G Ono~in. Sollto.lS. Ounuridge. 1983
E. r . Wohlfarth , cd., "mulboo~ all ",agllc/ic Ilwl"rhIls , Narth-lIolillml, scl c".1 ,·aluma.
Jou rnal of Magndisoll and Magnetic Ma.ter~'lis
Y I. lnkaw;l and N Mnora, cds., Physics Grill cnglneen", GI",/kal/ons of maglwlis"'. Springer,
1991.
J. Jenscn and A. It Md<:kmtosh , R G,"" flQrl/, nUlg,."tl:m , O~ford . 1991.
J. Cranglc, Sol,d II<IIe magnetmn. Arnold , 1991.
C. O. ~lce Ilnd E. D. Daniel, MagnclK" rtoordinl! hanll/}OtJl. , :\I<:C':h.~lIill , 19'JO.
It S. Wadas. Bio."..gnelirnl, Ellis 1101"\'0011, 1991 eoo.l g(.~ 'cral ."it.w alii" """,ocl;o:: 1L'ipc."Cts or
argauiuns.
II. N. Be,1nm and JiaJl.G:Ul)( ;£to", "Fundament",1 magnt."iU:,Uilllllll"OCl:Slo';S in !luIl-mm recording
.ncd'a. ~ SoI,,1 sWe l,nysics 46, 271 (1992)
P. Gruetlcr-, H . J Mamill, and D. Ruga.-, ~ :\Iagnetic IOroc microscopy ( ),fF\lJ,~ on Sro,,,,ing Wn·
""'I." m/cJwoop!l lI. Sllringcr- 099-2)
 .- - .. '

16
Magnetic Resonance

r\'UCLEAn MACNE'nC n..:SOr\'ANCE 487

E(IUalions of moti{)Jl 400

Ur\'E WIDTII 494

Motional narr<ming 494

1IYl'1~IIFlNE SPLrrll!'l;C 497

I~XIlmples,paramagnetic point defects 499
Knigl,l slMt 500

NUCI..EAU QUADJWPOLE HESONA!'I;CE 503

FEIIHOMACNI<:nC nESOfliAr\'CE 503

Shape effects in FMH 504
Spin wave reson31'1Ce 506
A!'I;'fIFEnROl\tAC)'I;EllC RESONA!'I;CE 50G

[LECmON PAUAMACr\ETlC IIESONAr\'CE 510

E)<change narrowing 510
Zcro-£ield splitting 510

PIUNCIPI, E OF !\IASEn AcnON 510

TI,rcc-level maser 512
nuby laser 512
SUMl\tAnV 514

FIlOBLEMS 515

I. Equivalent electrical circuit 515

2. notating coordinate system 515
3. H yper£inc effects on I-: 5n ill metals SIS
4. Fl\IH in the anisotropy £icld 515
5. I~)<change frequency n.'sommce 516
6. HI" satlll'alion 516

r\'OTA'1l0r\': In this chapter the symbols Dd anu Du refer to the al'lllictl (jeld, and HilS
the appi,t.'(\ Geld plus tile dem~cti'Zing Geld In p:lrticular we writ ... B. = Hoi. f or CCS
readers il may lx· simpler to rcail H for 8 "ht.'I1cvt.... ,I occurs in ti,ls chapter
 1-

" .., Vi
i

"" "'"
"t- ,
Ek!cl...,,, $pill n ,$On3ntt absorption ill /I.I"S04 al 29S K M 2 75 C II:z, after Za>'Oisl;v.

\
~ .. .
 CHAPTER 16: MACNETIC RESONAl\"CE

In this chapter we discuss dynamical magnetic effects associated with the

spill angular momcntum of nuclei and of elcctrons. The principal phcnomena
arc often identified in the litcrature by their initial letters, sudl as
NM H: nuclear mllgnctic resonlUlCC
NQR nuclear quadrupole resonance
EPR or E5R: electron paramagnetic resonance (Fig. 1)
FMR: ferromagnetic resonance
5\\,H: spin wave resonance (ferromagnetic films)
AFMH: antiferromagnctic resonance
CESH: colld uetion e lectron spin rcsonanee
The information that can be obtained about solids by resonance studies
may be catcgonzcfl:
• Electronic structure of single defects, as rcvealed by the fine structure of the
absorpt ion.
• ~totion of the spin or of the surrounding.~, as revealed by changcs ill the line
width.
• Internal mllgnetie fields sampled by the spin, as revealed by the position of
the resonance line (chemical shift; Knight shift).
• Collective spin excitations.
It is best to discuss NMR as a ba.~is for a brief account of the other reso-
nance experiments. A great impact of N~IR has been in orgnnie chcm istry nnd
biochemistry, where N~IH proVides a pO\vcrful tool for the identification and
the structure determination of complex molecules. This Sllc<.'CSS is due to thc
extremely high resolution attainable in diamagnetic liquids. A major medical
appliC'J.tion of NMH is magnetic resonancc imaging (MRO. which aIl0\\'5 tile
resolution ill 3D of abnormal growths, configurotions. and reactions ill the
wholc body.

NUCLEAR MACNETIC RESONANCE

\Ve consider a nucleus that possesses a magnetic moment po and an angular

momentum Ill. 11le two quantities are parallel, and we may write
po=')'fII; (l)
the magnetogyrie ratio')' is constant. By COll\'cntion I denotes the nuclear angu-
lur momentum m..asured in units of II.
."
II _ B•

,, •• ,•
r
A.... .. fp.H. ..
'"
"Its.,
, I "" •
.'1-

Figure 2 EI\~' Jco.eI splItting of 1\ nucicus of spin I .. l j" a static m.agnetic field Be..

The encrg)' of inler.octron with the applied magnetic field is

U= - p.·B,,; (2)
if B" = Boi, thell
(3)
111e allowed values of ,~ are 111, = I , I - I, .. • - I , and U '" - mahBn.
In a magnetic field a nucleus with I :::: i has two energy le\'els correspond-
ing to In, = :t+. as in Fig. 2. Iffuq, denotes the energy difference bch."~11 the
Iwo levels, then hWo = 1'(180 or
W(l = yBo. (4)
11lis is the funnamental (;oudition for magnetic resonance absorption.
For the proton' y:::: 2.675 x Hr' 5- 1 ga u ss - I::: 2.675 x 10" 5- 1 lesla- I ,
so that

"'M Hz):::: 4.258 Bo(kiloglluss) :::: 42.58 Bollesla) ,

where v is the frequency . One tcsla is precisely 10" gauss. Magnetic data for
selected !lucid arc given in Table l. For the electron spin,

~Cllz} = 2.80 BoOcilogauss) :::: 28.0 Bo(lcsla) (4b)

'"fht. nugrM'l lC moment p.,. of the pro/OIl iii 1.4106 x iO- Q erg C -' Or 1 4106 X 10 '" J T - '.
Ilnd y . '4<,1" The nudear magncton p.. is ddl"ed as cftJ2].,~ and;s equal to 5.0509 X 10 U erg
C 'or 5.()509 X JO-t'I J T ' ; thus p. .. .. 2 7Itl n"dear nUlgucto"s.
r r"blc 1 Nudear magnctlt reSOnanCe data ,----
H' H,'
1/2 1/2
99.98 For every element the most abund:mt m3gnetic isotope is shown. Afte r 10-"
2.792 Varian Associates NMR Table. -2127
LP B,' B" C" N" 0" F" Nell
3/2 j;2~ I 5/2 1) 2 .111 3/2
-'/2
9257 100. 81.17
.1/2
!.lOS 99.64 0.04 100. 0.257
3.256 -1171
NaZl
3!2
100.
Mg1S
5/ 2
10.05
Most abundant isotope with
Nuclear spin; il'1 units of K
nonzero

Natural abundance of IsotOpe. In pert~ent

nuclear spm
- - .......... . ~ .
2.68$
Ail'
5/ 2
100.
0.702

SF'
1/ 2
4.70
6.404 -1.89l
P"
1/2
100.
S"
3/2
0.74
3/2
75.4
.
2.627 =Q.662
Cps

' 2.216 0.855 Nuclear magnetic moment. in units of eX/2M.c 3.639 0.555 1.131 0.643 0.821

K" Ca') Sc' s Ti" V" CrH Mn ss Fe H Cos, Ni u CUbl Zn 61 Ga n Gel l As's Sen Br" Kr u
7/2" 7/2 5/ 2 7/ 2 5/ 2 1/ 2 7/ 2 3/ 2 3/2 5/2 3/2- 1/2 ... 3/2
'/2
93.08
0.391
0.\3 100
-1.315 4.749
7.75
0.787
':'100
5.139
'/2
'54
b.474
100.
3.461
2 .245
0.0'0 4.639
100 1.25
0.746
69.09
2.221
4.12
0.874
3!2
60.2
2.01 1
'/2
7.61
0.877
100.
1.435
7.50
0.533" 2.099
SO.S7 '"
11.55
- 0.9£7
RbiS
's/ 2
Sr'l
9/ 2-
y .. Zr91
1/ 2..... 5/7
Nb 9l M09S
9)2'""' 572 -
T, Ru lol R" lol Pd loS Agl OI
5/2 172 572'- 1/2.....
Cd lll Inlls Sn ll9
"172-- '/2 'I7'2- "f5/2'''''
Sb lll Xe l19
"572'=: Tj2' "
Tells
~ r72 ~
jill

72.8
1.""
7.02
r~"
100 11.23
0.137 1.298
Sa ll7 la ll '
100. 15.78
6.144 0.9""10

Ta lll
- 16.98 100.
-0.69

Re U1 Oslt 9 Ir l91
22.23 51.35
-0-:--088 "':tl:S1 ":o.m:
Au 191
12.86 95.84 8.68 57.25 7.03 100. 26.24
- O.59f 5.507 "'IOlr 3.342 ":b.882 2.794- -O.m
Hg l 99 Til OS .,-
CSl ll
'172 '3/2 '"71'2""""
HfiH
--"/2:- 7/2"""
WI'l
17'~ 572 ~/2 372
Ptl9S
f72~ '"3/2-- rn~ 172'
PblOl
-17'-
SP09
-.n- -Po At
~

--
- - •
. c

.,
100. 11.32 99' 18.39 100. 14.28 62.93 16.1 61 .5 33.7 100. 16.86 70.48 21.11 100. ~
2.564- 0.931 2. 761 0.61 2.340 -0.115 3.176 0.651 0.17 Q.6oo tUM 0.498 1.612- 0.584 4.039 ~

- -- "
F, A,
""'.... Ce 141 Prl'l Ndl' l Pm Sml<) EU lSl Gdl!l Tb ls' Dy l61 HOI U Er161 Tm 16' Yb lll lUllS
1-
- - 7/ 2 ... ' 5/ 2_
-
0.16
Th
100.
3.92
p,
7/2
12,20
- 1.25
U Np
" 7/2
15.Q7
-0.&8
p,
5/2
52.23
1.521
Am
3/2
1564
-o.3f
Cm
312
100.
1.52
Bk
5/2 - 7/2 -
24.97
-0.53
Cf
100.
1/'
22.82
3.31'"F 0.48

E, Fm
1/2
100.
57? 1)t'
16.08 97.40
-020- :'0517 2.9

Md No L,
i
- -
~

-- - - - - -
~ .... ''!""'''II''" ~ ~
EqllatimlB of Motion
The rate of change of angular momentum of a system is cCJualto the torque
that acts on the system. The torque on a magnetic moment /oJ. in a magnetic field
B is po. It n, so that we have the gyroscopic equatiOll
ftdIldt "'" po. If D" ; (5)

(6)
'Ole nuclear magnetization M is the sum ~"'. over all the nuclei in a unit
volume. if only a single isotope is important, we consider only a single value of
'Y, so that
d'fl-udt = yM x Da (7)
We place the nuclei in a static field D,, :: Boi. In thermal equilibrium at
temperature T the magnctization will be aleng z:
M. = O; (8)

where tht' Curie COlllitant C = NJL213k 8 .

The magnetizatien of a system ef spins wilh I "" i is related to the popula-
tien difference NI - N2 ef the lewer and upper le\'cls in Fig. 2: ~J= "'"
(N 1 - NlJp., where the N's refer to a unit volume. 111e popoi<1tien ratio in
thcrmal equilibrium is just givcn by tile Boltzmann factor for the energy differ-
ence 2p.Bo:
(9)
111e eqUilibrium magnetizatien is Mil = Np. tanh(JL8Ik 8 1"l-
\-\'hen the magnetization l.'Omponcnt M~ is not in thermal equilibrium,. we
suppose that it approaches equilibrium at a rate proportiellal to the departure
from the equilibrium "alue Mo:
dMz Mo - M=
-- ~ (10)
dt TI
In the standard notatiOn TI is called the longitudinal relaxation time or the
spin-Ianice relaxatien time.
If at t = 0 an unmagnetized specimen is placed in a magllclic field 86l, the
magnetizatien will increase from the initial value M. = 0 to a final value I\fz ::=
Mo. Before andjllst after the specimen is placed in the field, the populatien N .
will be equal to N2 , as appropriate to thennal equilibrium in zero magnetiC
field. II is necessary Ie reverse some spins to establish the new C<juilibrium
distribution in the field 80- On integr.lting (10):

(11)
 I --------- -----

,
'",- -
Fif;ure 3 At time I .. 000 ullmagoclizcd specimell M.(O) '"" 0 is pl3C<.'d m a sl ~t ie 1l.agnt'11<: rock! Bo.
'n\C: nlllglw: hmtio" increases wilh time >Irnl app~ the ncw .....:.uilibt"'''' val"e M" - xoBo. This
t'XpCTimcnt ddlncs the lonl:itoolnal ...,I:wo.t;oo time T, . 'nw., magnetic energy dellsit)' - M ' 8
dccncascs IU l)l8Jt oft"'" "Pill populahoo mQYC5 into the lower IcYd . The a5)'mll«l11<: value al I .. T,
Is - MoBo. T he ~ flows frol11 Ihe SI"" 5)",;1"'" to Ihe s),.lem of laltke ";I",,hons; thus T, is al50
c.dlOO lhe SIKn-I.ltlice rcbxatiotl tiruc .

0<
(12)
T,
as tn Fig. 3. 111e magnetic energy - M' Bd decreases as M1 apprQaches its new
~Jui librium , 'alue.
Typical processes whereby the magnelizalion approaches equilibrium are
indicated in Fig. 4. The dominant spin-lattice interaction of paramagnetic ions
in <-'I)'stals is by the phonon moduilition of the cryst:llIine elect ric fleld . Helaxa-
lion proceeds h~' three principal processes (Fig. 4b); direct (emission or absorp-
tion of a phonon); Haman (scattering of a phonon); ~lI\d O.-i)."lch (intervention of
a third state). A thorough experimental anul)'sis of spin-lattice relax.1tion in
several rare-earth salts at helium temperatures has been given by Scott and
Jeffries;2 they discuss the evidence for the three processes.
Taking account of (10), the z component of the equation of motion (7)
becomes
ddM1 "'" ;{l\.·1 )( Ba)z + Mo - Mz (133)
t T,
where (!\to - MJI1'I is an e):tra term in the equation of motion, arising from the
spin-lattice interactions not included in the magnetic fie ld B.. _l111t is, besides
precessing about the magnetic field, M will rela~ to the equiill'r " m value Mo-
If in a static fie ld Boi the tranS\'erse magnetization compol, III M~ is not
zero, then M~ will decay to zero, and similarly for M,. The dee:!y occurs he-
t P. L. Scott and C. O. Jclfncs, Ph ys. Hev. 127,32 (1962).
,
'"

10,." ,,"10, Metal

Figure 4a Some important procc~ses that LXln lri butc to iongitu<Iinw magnc\i7.3lion rela.ation in an
insulator and ill a m<.1aJ . )<' or the insulator we show a phonon .caUe,..,d ineJaslio:..J ly by the spin
System. T he spin syste m 010\"(."5 In a lower energy sLl!c, antl thc cmitk..! phonon has hight'r energy
by 11"'0 than the absorb<.... t phonon . •-or the metal we show a s;", ilar inclash" SOlttcring proc.,ss in
which a colI.luct ion elect ron is scattered.

Oi,"""
lIT,'" T
f'-igu.-c 4b Spin rciaxation from 2 --+ I hy phonoo emission. phonon scatteri ng. and .. two-slag"
phQllon proc.,ss. "111e temperature <lcpcndcncc of the longitudiu al rd:lX>ll.ion IlI1IC To is shown fur
the SI,.'\"Cra\ procesS<.-"S.

cause ill thermal equil ibrium the transverse components are :£ero. \Vc can p ro-
vide for transverse rcla~a tiOIl :
tlMjdt = y(M x B.J.:r - M) T2 (131,)
dM,Jdt = y{M II Ba)~ - M,j'f'J. , (13<)
where T2 is called the transverse relaxation time.
111e magnetic energy - M' Ba does 1.lot change as M~ or M" changes. prO-
vided that Ba is along i. No energy need flow out of the spin system during
relaxation of M~ or My, so that the COnditions that determine 12 may be less
strict than fo r TI . Sometimes the two times are nearly equal, and sometimes
TJ » T 2 , depending on local conditions . ~
The time 1'2 is a measure ofthe time during which the individual moments
that contribute to Mx , My remain in phase with each other. Different local
magnetiC fields at the different spins will cause them to precess at different
frequencies. If in itially the spins have a common phase, the phases will bef.."Ome
random in the course of time and the values of Mx . M'1 will become zerO. We
can think of T2 as a dephasing time.
The set of equations (13) are called the Bloch equations. They are not
symmetrical in :t, y. and z because we have biased the system with a static
magnetic field along i . In experimen ts an rf magne tic field is usually applied
 --"

:Llong the x or 5' a.xes. OUT main interest is in the behavior of the magnetization
in the combined rf and static fi elds. as in Fig. 5. The Bloch equations lire
pbusiblc. but not exact; they do not describe all spin phenomena, particularly
110t those in sol irls.
We determine the frequency of free precession of the spin system in a
static field H" =:: 13d}.und witll Ma = )10_ The Bloch e(]uations reduce 10

dM At -tlAL_. = 0
-liM"
- = ..,noAf -M
-.. ..:::..:..:.t..=
dt
- "'8 - ~1 -....:.:.:..I!..
0" ~ T2 (14)
dt ' 1/ T2 1 <It

We look for damped OScillatory solutions of the form

AIr = rn exp(- tl1") cos wi ; M .. = - m exp( - tn"') sin wt (15)
On substit ution ill (14) we have for the left-hand equnlioll
1 1
- w sin wi - -
T'
cos wi = - ",n, sin wt -
1
-
T2
cos wI (16)

so tlllL! the free precession is characterizerl by

(17)
The motion (15) is similur to thai of a damped harmonic oscillator in two
nimcnsions. The analogy suggests corrcctly that thc spin system will show reso-
nance absorption of encrgy from a dridng field ncar the frequcncy Wo = ')'Bo•
and the fre<1ueney winth of the response of the system to the driving field will
he .6.w '" In'S!. Figure 6 shows the resonance of protons in water.
111e Bloch equations may be solved to give the power absorption from a
rotatillg magnetic field of amplitude lh:
Bv :: - HI sin wt . (18)
After a routine eulculation one finds that tht' poo.ver absorption is
~( \_ wy,\1.TSI 2
(ees) '3"WI - BI (19)
] + (Wo w)211
The half-width of the resonance at half-maximum poo.\ler is
(20)

"
..
,

Fill:urc: 6 l'rot .." ,,'SOII<lIln' ab!.orp-

lIo n 111 w~I"r. (E I. IIdhn .)

LINE WIDTH

TIIC mUAllc lic dipolar interaction is l1l>ually the most impmtant t:. lU ~e of
linc hroarlcning in a rigirllatllce of magnetic dipolc~. The nmgne tic fidei 68
seen by a m,lgnclicdipolc p-, due to a magnetic clipole P'l at a point fl 2 from the
fir,' dipule is

ICGS)
.0.8 = 3(P !1' r d r ,\! - P.2"~2 (2 1)
r12
by a fundament,.1 result of m'lgneto~t.ltics .
The order of maguitude of the interaction i~. \\ilh B, written for 68,
ICGS) (22)
The slrong dependence on r lIouggcsts that dose neighbor interactions will be
dominant, SO thaI
ICGS) (23)
where 1I is the scparnlioll of nearest neighbors. 'nlis result gives liS J. mcmow c of
the width of the spin reSOnance linc. assuming random orientation of the neigh-
hors. For protons at 2A. sep.mllion.
23
1.4 x 10 - C emu
B, - .:.c::-c,-,=:-:..-=-::;:;,--
8xi0 1l
cm'
- 2 gau!>s = 2 x 10- I tcsla (24)

To express (2 1), (22). and (23) ill 51, multiply the right-hand sides by 1-4/41T.

,U o liOJllll NCirrou:j'lg
The line widt h c.1ccre•• ~es fo r nuclei in rapid relative motioll. The cff'cc,,1 In
solids is IlI u:. trn tcd hy Fig. 7. (hflu:.ion rcse mbl l'~ a r,md~' w,Llk as atoms jllll1P
 "-
["
, '"
,, ""
~

,
"
>0 -

,
"" "'" ""
T~",pcr.>t\,,.., . K "" '"
Figure 7 Effed of (Mfuslon of nudei on tho.· 1.,7 N\II\ li ne wIdth in mcla1!j(' lithIUm. At low
Ic"'p"r.. turc~ the ..... klth dgrt"es ,,;Ih the t11<-·crcll~,.] "aJue ror 1I rigid l,tllK't' . A. Ihe 1Cml'l'rJIlIrc
IIlcrc"scs the d,Ai",Otl r .. le I11CrC,,,,,5 aml the lon e width de<:n·~es. '111(: abrupt dc.... casc III tine
wklth abo..." T "" 230 K OC<.:"f'S when the d,ffusIon hopping IIme,.- becomes shorter than I f)'ll, Thu~
the experiment J.:i\"cs a c\in.'cl measure of lh" hopping lime for an atom toch"""c 1,\lIK'" sIte •. (After
II S. eutaw."y ami II. n McCarvey ')

from one crystal site 10 another. An alom remains in onc site for .Ill av{' rage
time 7' that decreases markedly a.~ the temperature increases.
The motional effects on the linc width ure even more spectacular in normal
liquids, bemuse the molecules are highly mohile. The width of the proton
resommce line In water is only lO- s of the width expectro for water ilIolecules
frozen in positiOn.
The effect of nuclear motion on 1'2 und on the line width is subtle, but can
be undcrstood hy an clementJ.ry urgument. \Ve know frOIll the Bloch eqn:l.tions
tklt T2 is a measure of the time ill which an individual spin becomes dephased
by one radian because of a local pertUJbJ.tion ill the m.lgnetie field intensity.
Let (t..wlo "" r81 denote the local frequency de\ iation due to a perturbation 8,.
TIle loC'al field may be caused by dipolar interactions with other spins.
If the atoms are in rupid relative motion, the local field 8 1 seen by a given
spin will fluctuate rapidly in time. We suppose that the local fidd has a v.llue
+8; for an a\'efllge time ,,~md then chJ.ngcl> to -8" as in Fig. Sa. Su{'h a
random change could be cUllscd by a change of the .mgle between p. umJ r in
(21). In the time T the spin will precess by 1111 extra phase angle &p = ±r8,,,
reiati\'e to tbe phase ungle of the l>teady prccc~sion in the applied field Bo.
The motional narrowing effect urises for short" such that 81j1 ~ 1. After /I
intervuls of duration T tIle mean square dephasing angle in the field Bo \\ill be
<rp2) = ,,(&p)~ = Il'fB~-,2 • (25)
by analogy wIth a rmuJom walk pi Qcess: the mean squ.lre displuccment from the
e
initl.ll position after steps of length in random dilc{·tions is (,.2) = 11("2.
·96

,, (
• -
0
.-
. '- - '- '- -
~ ~I) " ~ on ""'.<Iao~
", local rock! 8 , - -t I

,/
///'

.-
."igurc 8 I'ha!;c of a spill ;u a CQlIs';JlI t lornl fleld. a~cornpllrcd ",i.h dcphas;lIg of a .pill whkh Bfler
flxl-d time ill. crvals,. hops al random arnougsilL'S ha~i"gJocal fields ::1: 1. 11lc dcpilaslng is mcas_,red
,e!ati,'C 10 the pl\aSe ri a spiu ill .he .pplk-d field Bo.

The average number of steps necessary to dcpbase a spin by one mdian is

n = lIYBj!r. (Spins deplmsccl by much more than one radian do nol contrib-
ute to the absorption signal.) l'lis !lumber of ste ps takes place in a time
(26)
quite different from the rigid lattice result T2 !iI 111'11,. From (26) we obtain as
the linc width for rapid motion with a characteristic time r.
(27)
0'
(28)
where (Aw)o is the line ....1{1I1. III the rigid lattice.
111c argument assumes thai (AW}o7 <Iii 1, as otherwise &p will not be -c I.
Thus 6w <C (l!.w}o. The shorte r is 'T, the narrower is the resonance line! TIlis
remarkable effeel is known as motional narrowing.3 1he rotationru relaxation
3fhe physicltJ idcas are d"e to N. B1ocmbergcn, E ~f . F\,ra.-II. and R. V. Po"nd. PI,}'S. Rev.
73. 679 (1948) TI-.c result di!Teu from the t hl"Ory of opticall i"c wklth caU5t.... by $!nln~ <"Olli sion5
!x,lweCH atom. (a. in a ga~ d iS<'"hargl'), ... here a short ., gi\"c~ a broad lill". I n the ""rlcar spill
prohlem 11.., rollisions are weak. In moSI oplical problem. lhe roll'~iom of atoms W·" lil rong eHough
10 illierrupi I..... phase of Ihe OKillation In nuclear r~wnallce the pll<lSC mOly "al)" smooth lr in a
mU;sion , although the frcq"en~ may vary sudden])· from one ,...J,..,
to another nearLy ,·al"c.
 ...
16 Mngnelie Jlno...,nu 49

time of water mOlecules at room temperature is known from dielectric constant

measuremcnt.. to be of the order of 10- 10 s; if (6.wlo ""' 1()5 S- I . then (.6wlo'T ""
1O-~ amlllw - (llw~'T - 1 5 - 1 . l1lUs the motion narrows the prolon resonance
line to about 1O-~ of the static width.

I1 YFEUFINE SPUTnNG

The hypcrfine internction is the magnetic interaction be'\Oo'CCn the mag-

netic momc nt of a nucleus and the magnetic moment of an electron. To an
observer stationC(1 on the uucleus. the iuteraction is cause<! by the magnetic
field pro<luccd by the magnet ic moment of the ek-ctrOIl and by the motion of
the electron about thc nucleus. Tllerc is an electron CUtTCnt about the nucleus
if the electron is in a state with orbital :U!I:,'ular momcntum about the nucleus.
Bllt evell if the elect rOil is in a state of zero orbital angular morne ntllm, there is
an electron !>-pin current about the Ilucleus. and this current gives rise to the
contact h YJ>crfille interaction, of particular importrulcc in solids. We aU! un-
derstand the origin of the contact interaction by a qualitative physical argu-
ment, given in CCS.
The rcsu h s of the Dime theory of tile elect ron suggest that the magnetic
moment of IJ-fl = efll2me of the electron arises from the circulation of ru1 elec-
tron with velOCity c ill a current loop of radius approximately the electron
Compton wavelength . ~. = fume - 10- 11 cm. The electric current associated
with the circulation is
1 - e x (tums per unit time) - edK., , (29)
<U1d the magnetic field (Fig. 9) prOOuccd by the current is
(eGS) n- I1~c - elK.! . (30)
lnc observer Oil the nucleus has the probability

p - III>(O)I"~ . (31)

of finding himself inside the ell..'ctron, that is, within a spllcre of volume K.~
about the electron. He re 1/1(0) is the value of the electron wavcfunctioll at the
Ilucleus. l 11Us the a\'el1\ge value of the magnetic field SCCIl by the nucleus is
(32)
where fJ.fl = ehJ2mc = tcK~ is the &hr magnctOIl.
TIle contact part of the hyperfinc interaction cnergy is
U = - p.,' B "'" - p.,' #L8!I/I(O)ll "'" yfilJ-fll¢(O)1 2) . S (33)
where I is the nuclear spin ill units of h .
TIle contact interaction in rul atom has the form
U = al' S (34)
'"

I'igure 9 \/agocl;!: field 8 produced by a charge moving in .. dITIJI"r loop_ TIle ('Oulad pari or lb"
hypcrll ne inlemction ... ;I h a nuclear magnetic 1Il<)l11cn\ arises from the region ,,;Ihin or lIe.Of 10 Ih"
current Ioop_ 111(' fldd a"eraged (Wer a spherical.he n Iloal "ndoses the loop gj, cs 7.cro. Thus foran
$ ck'CIn:>n (L '" 0) only the rontact pari rolltr ibule5 10 the inlcr.lction .

Values of thc hypcrflnc consl:ml (J for the ground sta les of several free atoms
arc;

nucleus II' Li7 Na 23 K~ K"

I i i i i
(! in grlllSS 5ffI 14·1 310 83 85
(I in ~ IJ-I z 112ll 40' 886 Z31 I"
The value of a in gauss as seen 1;)' an electron spin is defined as o/2~fJ.
111 a strong magllclic Held tlle energy level scheme of a free atom or ion is
dominated by the Zeeman encfh'Y splitting of the e lectron levels; tbe hyperfinc
illteraction gives an additional splitting Ilmt in strOng fields is U' i:! amsm,.
wherc fils, 1/1, are the ma!,'llctic quantu m numberS.
For the energy level diagrnm of Fig. 10 the two c1l.'ctTQnic transitions have
the selection rulcs lUlls = ± 1, 11m, = 0; thc frequcncies arc w = yHO ± aiM .
111e nuclear tnmsitions arc not marked; they ha\e I1ms = 0, so that
W"ue = a12!J. The frequcllcy of thc nuclear transition I -.. 2 is cqual to thotl

of 3 ..... 4.

l
...
 •,

.
,
r
I
" .
t'igurc 10 Energy 'c.'cls in Ii magnetic field ofa ~y5l cm ",th 5 - t, I - t. n,e diagnl.lll;5 dr.. wn for
the strong field approxim~llon J.<tlB to n. whe re " i~ the hype ,'flnc roupling CQn~tant . l:,ken to be
p<Xltj,c n,,~~ lour b.-I:!; are WJclcd by the magnetic qua ntum numbers ms. m,. The strong elec-
tronic Iram,lions ha,.. 6U1, .. 0, lin,s - ::t l .

'ne hypcrfinc internetioll in a magnetic atom may split the ground energy
level . 111C splitting in hyd rogen is J.l2O M Hz; and th is is the radio frC<lucllcy
linc of interstellar atomic hydrogen.

Examples: Pu,"omoglletic Point Defeds

11.c hypcrfinc splitting of the ciL'Clron spin r esonlUlCC furnishes valuable
structural information ahou l (lammaguctic point (iercels. such a.'\ the F Cf"ll t c rs
in alknli halide crystals and the donor alollls ill scmiCOlldllctor crystals.
I,' Cen ters in Alkufi Halides. All F center is a negative iOIl \faCrulCy \\.jlh one
excess electron bound at the vacancy (Fig. ll). TIlc wavefullction of the trnpped
electron is share<! chiefly among the six alkali iOlls adjacent to the vacant lattice
site, with smaller am plitudes on the 12 halide iOlls that form the shell of second
nearest neighbors. The t'Ountiug applics 1"0 crystals "ith the NaCI st ructure. If
q:(r) is the w",wefunction ofl he valence elcctron on a single alkali iOIl, then in the
first (or LCAO) approximation

I/J(r) = C 2: r,c{r - r p) (35)

••
where in the Nael structure the six valucs of r p mark the alkali ion sites thai
bound the latticc vacancy.
-nlc width of the electron spin rrSOnlUlCC Iinc of lUI F ccnter is dctermined
e~nliall Y by the itX!rfine interaction of the trapped ck'Ct ron with the nuclear
 (

Figure' II An F eenlC>" ;s a nc:g;.th'c ion

Vllroncy with one exceu electron bound at
th" vaC'flocy. The distribution or the CXC'e:lS
"lcctron is largely on the positive ml'lal ions
adjacent 10 the vacant JaU ice "ile.

magnetic moments of the alkali iOllS adjacent to the vacant lattice site. l1le
omcn'L'(J linc width is C\'idcllcc for the simple p icture of the w;\vcfunctiOIl of
the electron. By Iinc wi<lth \\.'C m<.'3n the width of the envelope ohhe possihle
hypcrflnc stnlcturc componeuts.
As rut example, collsiclcr iUJ F cen ter ill KCI. Naluml potassium is 93 per-
ccnt K 3{l with nuclear spin 1 "" J. The total spin of the six potassium nuclei at the
F center is 1mb = 6 X t "" 9, SO that the number of hypcrfinc components is
21,...... + I "" 19: this is the number of possible values of Ihe quantum number
"" . TIlcre arc (2/ + rf "" 46 :c: 4096 im:lepcndclIl amUlgcmcnls of the six spins
distribule<\ into the 1900mponellts, as ill Fig. 12. Oftcn we observe only the
envelope of the absorption line of an F center.
DmlOr Atoms ill Silicon. Phosphorus is a donor whcn present in silicon.
Each donOr atom ha~ five electrons, of which four enter (liamagnetically into
the covalent bond network of the crystal, mid the fiAh bound electron acts as a
paramagnctic center of spin S = i. The experimental hypcrfine splitting in the
strong field limit is shown in Fig. 13.
Whcn the concentration exceeds about I X 10111 donors cm- 3 • the split
line is I'eplaced by a single narrow line. This is a motional narrowing efTt.'ct
(Eq. 28) of the rapid hopping of the donor electrons among many donol' atoms.
The rapid hopping averages out the hyperfine splitting. The hopping nlte in-
creases at tile higller concentmtions as the overlap of the donor electron ""·.Wl....
functions is increased, a view supporlCfI by conductivity measurements (Chap-
ter 10).
The donor electron wavcfunclion extends Ilot only over the cclliral dOllor
atom but significantly over some IlUlldreds of silicoll atoms. The Si211 nuclear
spillS give nddit iorm1 hypcmne splittings first studied by Feher with a powerful
electron nuclear double resonance technique known as ENDOR.

Knigllt S/lift
At a G.~t.-d frequency the resonance or
a nuclear spin is observed at a
slightly diffeI"Cnt magnetic field in a metal than ill a diamagnetic solid. The
effeet is knOwn as the Klli;ht shirt or metallic shirt lUl(1 is valuable as a tool for
the study of conduction c1ectrous.
 - 16 .\lal!.l""~ J1... o",,,,("(' SO l

".
•- I
- 3
9-8 7 6 $ 4 - 2
I- - - • . , . , ,I , • ,
QuMtum nu,nlxY ....

l' igILrc 12 11K: 4096 3mtngcmcnls d the six nuclear spins of ({3\I as distributed in lO 19 hn>erfinc
component.. Earn oornponcnt "ill be split furthe r ;nlo II. vcr)' large Ilumher of romJlUnenUo by
";rlue d tN' residual h)'lX"rfioe intcrx1ion wi llI the 12 ~Shbor CI Ducici, whid . mil)' be CP' (is
pereenl) or Q37 (25 percent). '11C eI>,-d ope of the pattern is Ilpprollirnatcly gauu;an ;01 ronn.

!L-____-=~~~-------
~ I l'hospboruo (6 x 10" "",-~

I'igurc 13 Ek."t.1ron spin resonance hnt.'S ofF donor atOll" in $Ili""". At the higher donor I.'(Ir"lC'Cn-
ITlilion neat the melal.iru;ublor tramillon a dooor electron can hop from site 10 site so rap;dl~ thai
the hyperfioc Itmcture ;s S\lppn'Sw.'<i. (After R. C , Joldchcr, W. A. Yaget. G . I~ Pearson, lind F. R.
Merrilt.)
,,>2

The interaction energy of a Iludells of spin I ancl magnctogyrk ndio 'Y, is

U = (- -rtfiBo + a(S J)/~ , (36)
where the first term is the interaction with the applied magnetic Held Bo and
the second is the average h}1lcrfinc inlcracliOIl or Ill(! lIucleus with the conduc-
lioll ck-Clrons. The average collduct iol1 electron spin (SJ is rdated to the Pauli
spin susceptibility X. of the conductiOIl clccll'On ~; M~ :: gNl-'/J(S=) "'" xJ3o,
wllcncc the inlCIa<;lioll may lie writtCIl as

u ~ (-Ytfl +~) Bol~ = -y,MJo (1 + 6.8) I. (37)

gNIJ.D Bo

The Knight shift is definoo as

(38)

and simulates a fractional dlangc in the m.'lgllclogyric ratio. l3y the definition
(34) ori llc hypcrfinc contact cuc'1,'Y. Illc Knight shift LS ghell appro.1(imalc!y h)'
K - X.I4<O)j2/N; that is, by the Paul. Spill Ml.'oCCpt illllity iflcre.uccl in the ratio of
the conduction electron l'Onccnlration al thc nucleus to the average conJuctioll
electron concentnllion.
Experiment.ll values arc gh'en in Table 2. TIle \'alue of the hypcrfinc cou-
pling constant a is somewhat different ill the Ulelal than ill the free alom be-
cause Ille wave fi.ll1ctiOtls alillc nucleus arc differeili. From the Knight shift of
mctallic Li it is deduced Ihat thc , 'alue ofi1P(O)12 ill the metal is 0.44 of the v:alue
ill the free alom; a calcul.lled valuc of Ihe ratio using Ihl.'OrcliCllI wave functions
is 0,49.
It is on ly in rarc in~tallcCS that the absulutc vruuc of the spill cont ribution
X ~ to the magllelic l>lIsccptibility C.lll be clclcnnilled, usudlly by very careful
conduction electron spin resonance experiments. '111c Klligllt shift has been of
value ill the study of metJ.ls. alloys. soft and intcmwtallic superconductor", mid
unusu.u electronic systems such as Na"WO).

·1:1.ble 2 Knight sllifts in NMR in metllllic clements

(At room tcmper,liurcj

Kni~ll $h;ft Knight sl"n

in l>t rreul tn pc.-c<enl

[.j7 0.0261 CutiJ 0.237

Nan 0. 112 Rb~ 0.653
AI'" 0162 Pd 1o.1 - 30
K~ 0265 pt lllCl - 3.533
\,;;1 0580 Au l97 L4
C,~ 0.69 Pblll7 "47
 16 .\ lag"e' ic II"'Q"OI'<"e

e e
,,,
(!)

0 e
(!) (!)

.,
e
(!)

,,'
( .I.

'"
p;!p<'I)

'"
.'igurc 14 (a) l .•w:est energy orientation of 3 nuclear dCdnc <]u.Kln.po1c momenl (Q > 0) in the
Iocd1 c1cdrie field oflhe Ii".r ion. , how" . '11,e eleclrons of lhe iun Ils.·If art: not shown (Il) Il ighe.1
encrgy uricnl~liun . (e) Thc cnc¢}' Icve] ' phlli"/( for I .. 1.

NUCLEAR QVADRUi'OLE RESO~A NCE

Nuclei of spin J ~ 1 have an elcetric quadrupole moment. The quadrupole

momcn t Q is a measure of the ellipticity or tile distribution of charge in the
lIude us. The quantity or intcrest is de fin ed classically by
(39)
whe re p{r) is the dJarge demity. An egg-shaped nucleus has V positive; a
saucer-shaped nucleus has Q negative. The nucleus whell placed in a crrstal
"">ill see the clectl-ostatic field of its environment, as in Fig. J4. If the symme try
of this field is lowe r than cuhic. then the nuclear 4uudrupole moment willlcad
to a set of energy levels split by the interaction of the quadrupole momcnt with
the local electric field .
The states tbat are split arc the 2l + J states of a spin l. l11e quadrupole
splittings can often be observed directly because lUI rf magnetic field of the
appropriate frequency can causc transitions between the levels. The te nn
nuclear <luatJrupole resonance refers to observations of nuclear quadrupole
splittin g.~ in the absellcc of a static magnetic field . The quadmpole spliUiu&!> are
particuldrl y large in covalently bonded molecules such as Cl z, 81'2, ~Uld 12 ; the
splittings are of the order 107 or l<f Hz.

FERROMAGNETIC RESONANCE

Spin resona ncc at microwave rrequencies in rerromagnets is similar in

principle to nucleiu spin resonance. 111e total electron magllctic moment of tIle
specimen precesses about the direction of the stat ic maglletic field. and encIb'Y
is absorbed strollgly from the rf tl'lUlsverse ficld when its frequellcy is equal to
the precessional frequcncy. We may think of the macroscopic vector S
representillg the total spin of the ferromagnet as quantized in the static mag-
netic field . with 'rgy levels sCpJ.mted by the usual Zeeman frequencics; the
'" (
magnetic selcdioll rule .6.ms = ~ 1 allows transitions omy betwccn adjacent
levels.
TIle unusual features of ferromagnetic resonancc include:
• 111C transverse susceptibility l..'o mponculs X· ,md X' are vcry large because
the magncti:mtion of a ferromagnet in a given slatic fielel is very much larger
than the magnctiza.tioll of electronic or !luclear paramagnets in the same
field.
• The shape of the specimen plays an impol'tallt role. Because the magnetiza-
tion is large, the demagnetization field is large.
• TIle strong exchange coupling betwccn the ferromagnet ic ck'CtrQns tends to
suppress the dipolar contribution to the line Width. so thai the ferromagnetic
resonance lines can be quite sharp «1 G) under favorable conclition.<;.
• Saturat ion effects occur at low If poY.'Cr levels. It is lIot possible, as it is with
nuclear spin systems, to drive a ferromagnetic spin system so hard that the
magnetization M~ is reduet.>d to zero or rcversc<l. TIle ferromagnetic reso-
nrulcc excitation breaks down into spin wave modes before the magnctization
veclor can be rotated apprL"'Ciably from ils initial direction .

Slwpe EffeclB in FMR

Wc treat the effects of specimen shape on thc resonancc frequcncy. Con-
sider a specimen of a cubic ferromagnelic insulator in thc form of an ellipsoid
wilh principal axes parallel 10 x. y, z axes of a cartesirul CQOr<linate s)'stcm . l11C
demagnetization factors Nr , Nil' N% are identical with the dcpolari71ltion factors
dclined in Chapler 13. The componcnts of the inlel1lul magnetic field Hi in the
cllipsoid arc related to thc applied field by
B~ = ~ - N~l~ ; B~ = n:: - Ntl-1" ; B~ = B~ - N,M:
Thc Lorcntz field (4w/3)l\I and the exchange field AM do not contribute to thc
lorque because their vector product with M vanishes idcntic-ally. In 51 we
rcplace the (."Omponcnts of M by I-IoM, with the appropriale r(.'(ldlnition of
thc N·s.
The componcnts of the spin equation of motion ~I = ')'(1\1 )C HI) becomc.
for an applied static field DoZ.
dM. _, . . _, J
----;it = nMvB~ - M~B~) = "Bo + (Nil - NJ M Mil ; (40)

dill
~ ~ >lM(- N.AI.) - M, n" - N,M )] ~ - >lB, + (N. - NJ.II J.I I.
To first ordcr wc may sel dM) dl = 0 ruld M% :z M. Solutions of (40) with
time dcpclldence exp( - iwt) cxist if

'w >In" + (N, - N,)MJI = 0 .

I->lB, + (N. - NJ Al J ow
 • • • 'W

•
,

,
tlgure IS FMR In a polished sphere of
) lIrium iron gamcl at 3.33 GIi1. and 300 K
&:'r Bo II [Il l ]. TIle tolal line "idlh at Mf·power
is 001" 0.2 G. (Afler R. C. LcCraw and
E. Spnocc..... )

so that the ferromagnetic resonance frequency in the applicc] field no is

(CGS) "" = yin, + (N, - NJ'I][n. + (N. - NJM] ; (4 1)

(SI) w:: .: ,y[B, ~ IN, ~ NJ"""I][n, + IN.: - . . J!'oM]

1lle frequency !til) is callcd the frequcncy of the uniform mode. in distinction to
thc frequcnc.i cs of magnon and other nonuniform modes. In the ulliform nl«ie
nil the moments prccess together in plmse with the same nmpl it u<]e.
For a sphere N. ::::;: N,/ ::: Nz • so that !til) "" ')'Do. A "ery sharp resonance
line in this geomctry is shown in Fig. 15. For a nat plate with Do perpendicular
to the plate N. "" N,/ "" 0; Nz. = 471'. whence the ferromagnctic resonance fre-
quency is

(ees) au "" -y(Do - 47TMJ (SI) "'" - ,;,D, - """I) . (42)

If Do is parallel to. tile plane of the . plate. the xz plMe, then N. "" Nz = 0;
N,/ = -I7T, and

(CGS) "" = >fB.,(D, + 4wM)]"' ; (SI) "'" = >fB.,(D, + """I)l"'. (43)

...
The C'-I>Cl imcn ts dctcnnillC y, which is relaled to the .~pccl roscop ic splil-
ting factor g Il\' - y iii lW-BIII. V•• lues of g [or metallic Fe, Co, Ni at room ICIll-
ller:lluTe arc 2. 10.2.18, and 2.2 1, re~pC(:thcly .

S,>in \Vace Resorlllllce

Uniform r[magnetic fields can cxcit(' long wavelength spin W<l\CS in thin
ferromagnetic fi lms if the electron spins on the sm f.1C(!s of the min see diffe rent
anisotropy ileitis tha n the ~pins withi n the film s. In (:0<.'<:1, thc surface spins may
be pinned by "mfacc :lIlisotropy interact ions, as shown in Fig. 16. If the .ffleld
is lIniforl n, it ('an excite W3\C'i with an ooJd number ofhalf-wavclcngths wullin
the thicknes,'i of the film. \\',\\ cs with lin evell number of h.\lf-wavelengths have
no ne t interaction energy "ith the fieili .
The condition for spin wave resonance (SW H) with the applied magnetic
field norlllal to the film is obtainCll from (42) by adding to the right-hruld side
the e)(change contrib ution to thc frequcncy. ,11C e)«;hange contribution molY be
writtcn a.~ D~, where D is the spin wave C)(Chrulge constant. The .lSSuml)tion
ka <C 1 is valid for the SWR e>.:periments. 11111S in an applied field Do the spin
wave resonance fre<lucncies arc;
(CCS) % '= ;(Bo - 471'.\1 ) + DJ.2 :: ;(Bo - 4wM) + D(w;rlL'f • (14)
whcre the wavevcctor for a mode of .. half-wavelcngt hs in a fil m of thickness L
is k = 'miL. An e>.:pelimental spectrum is ... howl} in Fig. 17.

ANTI FERRmlAGNETIC RI~SONANCf:

\ Ve conside r a uniaxial antife rrolllagnet with spins on two sublatliccs, 1

and 2. We suppose that the lIlagncti.tation M. on subl'lttice 1 is direcled along
the +z direction by an .misotropy field B,,2; the anisotropy field (Chaple r 15)
results from an anisolropy energy density U,,{O.) :::I K sin 2 0 •. Here O. is the
angle between M. and the;: a>.:is, whence S" "" 2KJM, wilh AI ;:: 1M II = IM21·
The magneti7.ation Mz is direclcO along the -z direction by an anisolropy field
- B,\i. If +;: is an easy dircctlon of magneti7.ation, so is -z. If one sllblalticc is
dircctCll along +'Z, the other wi ll be dircctCll along - z.
The t!)(ehangc interaction between M. and Mz is trcatClI in the mean flCld
app rmimation. 11le e)(change fields arc
~ex) :c -AM . , (45)

where A is positive. J Jere 8. is the field thai acts on thc spins of subl.lttice I,
and Bl acts Oil ~ubl atticc 2. I II the absence of;1II c)(tcrnal magnclic neld the lot:!l
field "cting on MI is BI = - AM 2 + B,,~; the tutal fiekl on M,! is
B2 :: - AMI - B"z. as in Fil<;:. 18.

\
 16 Mfl(,IIrlic Ikwn"ncoo 5('.

\ t11.,- 4""[

t
-"- I
t
\
t
Figure 16 Spill wlWe n '';Ufm,K.'e in a thin fillO, Th~ plan(' of the film is normal to the applied
",agnet;.; 'fidd B,.. A cross w.:tion of the film is ~hown h .... e . Th" in!crnal 1n.1gn"tic Field is
B., - 411'\/. 1bc spins on the snrf3CC$ ofth<, film arc assume,1 to be hdd find in direction!.>y SlIlf.ll."<,
'IFlI$u[ropy forees. A "niron n .rl" field ",11 excite .pin "'II"" modei 11:I\'ing an odd ntlml)t.'r of half-
",a\dC'l'b<ths. The WolVe shuwn IS fur II .. 3 h.alf-w.., e1c ngth.

, ,
..-ove oro..r numh",

, " ,,
!:it""

,
1358-1 C

13-(74

13219

"'"
12 19 1

11 732.

FiJl;ure 17 Spin waVe rcsonallC'C .pectrum in a Permalloy (80,\'.20Fe) Film at 9 C lb. '11101) order
"n",ber is tI..: number of half.".lvc1cngths in the Ih;.;luess of Ihe film. (Afier R Weber.)
50.
(

I
f
/II,(n) • - AM,

I
f
f
f
f

\
\

.
\
\
\
\
- A .\ lYeI)" - AM,
\ Figure: Iii Effedi,-e lock\!; in 'Ulliferromagnc:lk Tewnance. The
\ magnetization til, of r;ubbttit:e I $Cd a rlekl - "M~ + BAt; the
magru..'1ization Ma sees - A~I, - B"L Both ends of the crystal
:uis are -cali), axcs- of nugndiulion.

In what falla,,,, we set Mt = M; Mi "" -AI. The linearized equations of

motion arc
riMtfdt = 1'IMf(AM + B,.,) - M(- AAI¥))
dMfldt "" nAt{- AMI) - MH.AM + BAl] (46)
dMfj df ~ 1'IMi(- AM - n,) - (- M)(- AMni
dlll'6/dt = 1'I(- M)(- AMn - M ~{ - AAf - BAl] (47)
We define I\J I' = M1 + iMY; 1.12 "" M! + iM~. Then (46) and (47) be-
come, for time dependence cxp(- iwt),
- iwM I' = - iyfM;(n" + AM) + Mt(AM))
-iwMt = iy(Mt(B" + AM) + M t(AM) ]
These equations have a solution if, with BE ': AM,
}'(B;. + B,.;) - w
I
-rEI::
Thus the antifenornagnctic resonance frequency is given by
~ = r 8 ... (8 .... + 28,,) . (48)
~lnF2 is an extensively studied anlifcrromagnct. "Ibc stnldurc is shown in
Fig. 19. lhc obscn.et.l variation or ~ \\ilh temperature is shown in Fig. 20.
Carerul estimatcs wCl'"e made by Keffer or nil and n,.; fol'" MnF2. IIc estimatal
Bt; = 540 kC and BII = 8.8 kC at 0 K, whcnce (lBIIBd l12 = 100 kC. The oh-
scl'"\lcd value is 93 kC.
 · .-,

Figu.... 19 Chemical ,011 magnetic" stnlCture ol MnF.. n.e OM"'"

indicate the din:diQt, and Immgemenl ollhe mag_
IIITO\'-'S
nelic momenl$ lWlignl'<l 10 the mangan..-sc aloms .
,--

£
.
u
..
.'"
•i .00
-

'\
'\"

'"
" " .. '"
.,
Tempenlu ..... 1" "
'" 00
'"
Figure 20 Anlirerrom..gnelic resonance fr"'qn..,lC)' for MnFt \,cu, ,, Icmpcr,U ll re . (After John!iOn
IIrld Ncl hercot. )

Richanls has madc a compilation of AFt-in frequcncics as c)(trapolatCll

to 0 K:

C r ys tal CoF, MnF~ FeFt MnO NiO

F"re<lutmcy in 1010 Hz 85.5 26.0 158. 82.8 109
'"
ELECfRQN PARAMAGNETlC RESONANCE

Electron spin resonance is a vast area of research. We mention two topics

of interest.

Exclwlige NurrOll..'ilig
\Ve consider a paramagnet with an exchange interaction] amOllg nearest·
neighbor electron spins. The temperature is a'iSmncd to be well above any
spin-ordering temperature Tr;. Under these conditions the width of the spin
resonance line is usually much narro\\'er than c1.pcctClI for the d ipole-dipole
interaction. The effect is called exchange nalTowing; there is a close analogy
with motional narrowing. We interpret the exchange frequency WI;"'< ... Jlft as a
hopping frequency lIT. Thcn by generalization of the motional-narrowing re-
sult (28) we have for the width of the cJ(changc-narrowcJ line:
(49)
where (dw~ = r<Bf) is the square of the sta tic dipolar width in the absence of
exchange.
A useful and striking cxample of cxlhllnge narrowing is the paramagnetic
organic crystal known as the g marker or D I)l'II , diphcnyl picryl h)'drazyi,
onen used for magnetic field calibration. This frcc radical ha.~ a 1.35 C half-
width of the resonance line at half-power, on ly a fcw percent of the pure dipole
width.

Zero-"ield SJ,lilting
A number of paramagnetic ions have crystal field splitt ings of their mag-
netic ground state energy levels in the range of 101U - 1011 liz, conveniently
accessible by microwave techniqlles. l\lueh of the pioneer work is due to
H. meaney and co-workers at Oxford. The Mn2+ ion is popular and has heen
studiet.1 in many crystals as an additive impurity. A ground state splitting in the
range 101 - 109 liz is ohscrved , according to the environment.

PRINCIPLE OF MASER ACTION

Crystals can be us~1 as microwave and light amplifiers and as sources of

coherent radiation. A maser amplifies microwaves by the stimulated emission
of radiation; a laser amplifies light by the same method. 'Ille principle, due to
Townes, may be understood from the t\\on-Ievel magnetic system of Fig. 21.
'1l1ere arc flll atoms in the upper state ::.nd IIJ atoms in the lower state. \\'e
immerse the system in radiation at frequency 41; the ampl ltmie of the ma&'Jletic
component of the radiation field is Drf• The probabil ity per atom per unit time
(
. ... " "
 UI'I"'" <tole ..l
"·-----r--T'~

f Oisum 21 "lwo-lc,c1 ~) skm, 10 ....xplain ITI>tS~T opcT-

alioll. The popublions I)f Ihe UP\l<'T and lower ~lales

•-~';--'~";';";"-"'---* ~~
are " Mand >lh r... spttU, .... l) . The rrC<l"Cnc~ ofille emil-
led radlalion is w; the rombined ,with of lhe stalcS is
~ "'~ +t...J, .

of a transition between the upper and lower stales is

1
(SO)
6w
here IJ. is the magnetic moment, allli /:J.w is the combined width of thc two
levels. TIle result (50) is from a standan.1 result of quantum mechanics, called
Fermi's golden rllic.
The net cne rgy emithxl from atoms in both upper and lower states is

,"B'f)' /:J.w1 · flw· (11 .. - '1/) ,

'iI' = (-h- (51)

per unit timc. Ilcre tjJ denotes the powcr out: flw is thc energy per photon; and
n" - II / is the cxccss of thc number of atoms n" initially able to emit a photon
n,
over the number of atoms able to absorb a photon .
In thennal C<luilibrium n" < II" so there is no net emission of radiation,
but in a noncquilibrium condition with II" > II / there will be emission. If we
start with II" > II, amI reflcet the e mitted radiation back onto the system, wc
i ncrea~e nrf and thereby stimulate a higher rate of em ission. The enhanecd
stimulation continues \lntil the population in the upper state decreases and
be<.'OmC5 equal to the population in the lowe r statc.
We can build up thc intensity of the ralliation field by placing thc CT}'Stal in
an electromagnctic cavity. This is like multiplc rcllection fmm thc walls ofth c
cavity. "Inere will be some po\\Cr loss in the walls of the cavity: the ratc of
power loss is

(ces) tjJ "" n~rV ~ 0

(SI) r!i',
n~v
=-- w
. -. (52)
'0 8n Q . 2"" Q
•
where V is thc volumc amI Q is the Q factor of the cavity. \Ve un\lcrstand D~f t o
bc a volumc average.
The condition for maser action is that thc cmitted powcr tjJ cxceed the
power los.~ !!P". Both quantities invoke n~. The ma.rer condition can now be
 I
cxprcsseoJ in terms of the population excess in the uppel ~ ,atc:

vilB
,
(CGS) n -11/> - - - (51) (53)
" 81TJ.tQ
L
where J.I. is the magnetic moment. 11le line width!J.B is defineo.l in terms of the
combined line width !J.w of the upper and lower states as p.lJ.B "" li!J.w. TIle
centra.! problem of the maser or laser is to obtain a suitable excess population in
the upper slate. This is accomplished in variollS ways in variolls devices.

Three-Level Maser
The three-level maser system (Fig. 22) is a clever solution to the excess
population problem. Such a system may derive its energy levcls from magnetic
ions in a crystal, as Hlocmbcrgcn showCl.-l. Rf power is applied at the pump
frequency liwp = E3 - E, in sufficient intensity to maintain the population of
levd3 substantially equal to the population oflevell. Now consider the rate of
change of the population "z of level 2 owing to normal thermal relaxation pro-
cesses. In terms of the indicated transition rates P,
d"zldt "'" - llzP(2- 1) - Il zP(2 - 3) + 1l3P(3 - 2) + 1l 1 P(1 - 2) (54)
In the steady state d",Jdt = 0, and by virtue of the saturation rf power we liave
113 = IIi> whence

"z P(3_2)+P(1_2)
~ (55)
fi t P{2-1) + P(2_3)
The transition rates arc affected by many d etails of the paramagnetic ion
and its environment, but one can hanny fail with this system, for either fiZ > II,
and we get maser action between levels 2 and I , or flz < '" = "3 and we get
maser action between le\·els 3 and 2. The energy levels of the ErH ion are uSCl.l
in communication fiber optics amplifiers, as treated at the end of Chapter 17.
The ion is optically pumped from level 1 to level 3; there is fast nonradiative
decay from level 3 to level 2. The Signal at a ""avelength of 1.55 }Lm is amplililxl
by stimulatccl emission from level 2 to level 1. The wavelength is favorable for
long-distance propagation in the optical fiber. The bandwidth is of the order of
4 x 10 12 Hz.

Ruby Laser
The same crystal, rllby, used in the microwave maser was also the first
crystal to exhibit optical maser act ion, but a different set of energy levels of
Cr3 + arc involvtXl (Fig. 23). About 15,000 cm - 1 above the ground state there
lie a pair of states labelCl.I 2£, spaced 29 em - , apart. Above 2£ lie two broad
bands ofstatcs, labeltXl4F, and 4F2 . Because the bands are broad they can be
 '"
£ .~,. ..
-
".,
..
~ ........
""""

'. ".
•
,~.
.. -
1'I0,,I.;0,
... ~.,

Figure 22 TIlI'I:oe-levd mao;er spleln Two IlOlit ible modes of operation are shown, starting from rf
n lUration of the ~ t3tes 3 and I to obtain "3 - " ' _

'l'.- - -
..:----- 211 "", I

,-
, .
Hgure 23 Energy level diagram ofC.-3' in " ,hy. as used in laser (lpCMIlion. "'e 'nitial c~ci l~I;OIl
ta~C1i place to t .....• broad bands; d,e) dCC'ay to till' intcnnediatc k-vcls by Ihe emission of phonon.,
and the inlcunediatc k>vel. radiltle photons MS the ion mak..'S the transition to the ground le"I'1.

populatctl effiCiently by optical absorption from broaJbanJ light sources such as

xenon flash lamps.
In operation of a rllby laser both of the broaJ. -IF banJs arc populatC\! by
broadb..1.Otl light. Atoms thlls excited will decay in 10- 1 sec b)' radiationlcss
processes with the emission of phor)()Jls to the states 2E. Photon emission from
the lower of the states 2£ 10 the ground state occurs slowly, in abou t
5 x 10- 3 sec, so thal a large excited population call pile up in 2£. For laser
action this population must cxcccO that in the grounJ state.
l be stored energy in nih)' is 10" erg em- 3 if 1020 cr3+ iollS em - 3 are in an
exciteJ state. 11le ruh)' laser can emit at a very high powe r level if all this stored
energy comes out in a short burst. The overal1 efficiency of.conversion of a nlby
laser fmm input electrical energy 10 output lase r light is abou t onc percent.
5,.

P....'r

"igurtl 2-1 F(Mlr·k"VCl la$I'f f~ltcon, as in the

ncoJ)nlluln gI.Iu la.",r.

Another popular solid state laser is the neodymiulll glass laser. made of calcium
tllll~~latc glass dopeJ with Nd'+ ions. 'Inh operates as a fOlif lc\'c1 system
(Fig. 24). Here it i.~ not necessary to empty out the ground stale before laser
action can occur.
Semiconductor junction lasen arc Ircatcll in Chapler 19.

sml~I'\R\,

(In CCS Units)

• "J1 1C resonance frequency of a free spin is WI) :::2 ylJ o• where y '" plM is the
mab'l1ctogyric ratio.
• TIlt' Bloch equations arc
dJ\l.Jdt = ;{M x Bl. - MJT2
dM.jdt = "rlM It 8)1/ - M.jT2 ;
dM.Jdt = )'(I\I)C D)~ + (Mo - "'JlT,
• TIle haU-width of the resonancc al half-power is (6W}112 = UT2.
• Saturation effects al hig,h If power cnte r when rBlTI T2 exceeds unity.
• 'Ole d ipolar lioe width in a rigid lattice is (dBlo - pJcP.
• If the magnetic moments arc ambulatory, with a dlaracteristic time
'T <c l/(6w)u, the line width is f(ltluccd by the factor (dw)o'T. In thi~ limit
1fT. "" lITz - (6w~'T. With cl(changc coupling in d paramagnct the line
willth brtomes =(Aw~ex'
 16 lIfagrWfic Resonance S'

.. The fcrromagnctic rC~Onance frequency in all ellipsoid of dcmagnetiLation

factors N x• Ny, N~ is w~ = y lBo + (Ny - N.JMIl Bo + (N, - N.),\[J .
.. Tllc antiferromagnetic resonance frequcnl')' is wi\ = yBJ\( BJ\ + 2B.d, in a
spherical spe<:imcn with zcro applifXl fie ld . lIere BJ\ is the anisotropy field
and BI:: is thc cxchangc field .
.. Thc condition for ma.~cr action is that II" - III > V!:J.8181Tf.LQ

Problems
1. £ quiwletlt elutrical circuit. Consi,lcr an empty (.'O il of inductan(.'C to ill a Sl,.>rie~ \\ ith
a resistance Ro; show if tllC coil is completely mk... l \\ith a spin s)'stcm cimr.u.1crize<1
by the susceptibility COmpOnL"Tlts x ' (w) and x"(w) that the mductance at frC<juency w
becomes L = (I + 47TX'(w)ILo. in series with an eff(.dive resistance R =
47TWx"(w)4 + Ro. In this problem X = x' + ix" is defined for a linearly polari7.ed rf
field. Hillt; Consider the impL,<lance of the circuit. (eeS uni t ~ . )

2. Rotating coordinate system. We define the \'cctor F(t) "" F ..(t)i + I'Jt)y + I:,(t )z.
Let the coor<lmate system of the unit vectors i, y, i rotate with an instantaneous
angular \"t.'locity n, so that diMt = n~i - nS, etc. (a) Show th at
dFlc1t = (dFldt)R + n I( F, where (dFldl)R is tile time (lcrivati\'e of F as "iCWed in
th e rotating frame R. (b) Show that (7) may be wntl<'n (dMldt )R "" )'M x (Bd + fiI"),).
This is the equation of motion of M in a rotati ng coordinate sy~tem . TIle tr.uufo nna-
lion to a rotating system is extraordinarily useful; it is e~ploited Widely in tl w liter.t-
ture. (c) Let n "" - )'Boi; thus in the rotating frame thel"e is no st.ltic magnetic fiel<1 .
Shil in the rotating frame, we now apply fl. (Ie pulse Blx for a thne t. If the magnetlzn-
lion is initiall)' along Z. find an expression for the pulse length t such that the maj!;neti-
znlion \\ ill be din..>ctcd along -z at the end of the pul~e. (Negkd relaxation elfcds.)
(d) Decribe this Pllise a.~ "iewe(l from the laboratory frame of leference.

3. lIyperjine effects 011 ESn in metab. We suppose that the electron spin of a COnduc-
tion electron in a metal sees an elfective magnetic field from the hn>crfine intemction
of the ek>ctron spin with the nuclear spin. Let the;:; component of the ReM !i£'cn by
the con.luction electron be writtcn

B, - (~) J~I ,; ,

where I; is equally likely to be ±i· (a) Show that (Bf) = (al2N)zN. (b) Show that
(Bt) = 3(af2N j4 N1• for N ". 1.

4. FMR in tl.e allisotropy field. Consider a spherK-.t1 specimen of a unia,ial ferromag-

netic crystal with an anisotropy energy (lensity of the form V" "" K sint 8, where 8 is
the angle bet....-een Ihe nlolgnetization and the;:; axis. \Ve as.~ume that K is positive.
Show tha t the ferromagnetic resonance frequency in an e..dem.ll magnL1:ic fiel,l Boi is
Wo = '}(Bo + BA ), where BA, - 2KlM•.
'"
5. Exclumge/reqllency rcsonanu. Consider a fcrrimagnct with J sublattices A and B
of magncti1A1tions M... ana M8 • wllere Ms is opposite to M" when the spin system is
at rest. Tile gyromagnctic ratios arc ')'.... ')'8 and the moloxular fields arc 8 A '" -).1\18;
BB = - ).1\1,... Sllow thaI there is a reSOflancc at

111is is called the cxcll<lngc frequency resonance.

6. HI saturn/lOll. Given, at equilibrium for temperatllre T, a two-\cvel spin system in II

magnetic fldclll,ji., with populatiOf1~ N •. N2 and transition rates \\'12 . W 21 _ We apply
an If signal that gives a tmnsi!ion ral e W n' (a) Derive the equation for dAl,jdt alld
show that in the steady state

M~ -= Md(l + 2WtfT j )
where 1fT, = Wl2 + W~l' It will be helpful to write N = N j + N 2• n =; N J - N 2; and
no = N(W21 - W\2)1(W21 + Wl:V' We sec that as long as 2WrfT I ~ 1 thc absorptioo
of energy fnlln thc rf field docs !lot substantially alter the population distribution
from its thermal C<jllilibrium value. (b) Using the eKpression for ". write down the
rate at whidl energy is absorbcd from the rf field. Whnt happens as W,f approaches
If2T,? This elTect is c;allcd saturation, and its onset may be used to measure Tj .

References
INTRODUCTION
C P. Slichter, Princip/u of magnetic rcsonana; , 3rd ed., SI>nnger, 1990. An c"cellent intrOOuc-
tion.

NUCLEAR MAGNETIC RESONAr>;CE

D . Wolf, Spin t""'pernture and JIIJC/car spin rr/a:fation , Ch:ford, 1979.
A- Ahragllnl, Nudear magnetISm, Oxford. 1961. Definitive anll C(>mprchen,ivc
A Ahrnham and M. Goldman, Nuclear ",ogfUltisJ,~ onlcr and disonlcr, Odorn, 1962.
T. P. Das and E. L Hal"" "Nndear <lllallrllpole resonance 'pedrv.><.'Opy," Solill state phy~iCj,
Supplement I, 1958.
D. C. Gadian, Nuclear ma&,elic Teso.",nt:(! and iI~ appllctllkm 10 lio;ing syslenu, (h:fon] , 1982-
V. Jaccarino, "Nuclear reson"nce in antiferrom~lleh," MagnctU",- IIA .
A. ~1. I'ortis anll R. H. Lind'lui.)ct, "Nuclear resonance in fefT(>magndic mat<;riah." Mognnl.i"'" II A,
357.
Pr<lf!TCU ill nuc/eo ...nog'l<"tfc res,monee ~pcctrmC<lp!i, I'crgamoll. A ",Tie s.
G A. Wehb. cd , Anmw/ ff'porl$ 011 NMR $p.;ctrosCUp!l, Academic Pre" , 1980.
A. L Horowitz, Mill p"y$iC$ for the phY$ician , Springer, 1969.

ELECTRON Sl'I N RESONANCE

A. Abragam and H. Hlean<:1', Uectr"" poromllgm:tic rosonollCe of tra,,"silion iOll8, o.JOrll, 1910.
W. Low, " Paramagnet><: resonance in solids," Solid.tatephysic-s . Supplemt:nt 2. 1960.
K. , . Standley ami R. A Vaughan, I':lectron SpIll rela:falion In so/Uh, thlger, Londoll , 1969
J. E. Wertz ami J. R. Holton , Elcctnm 8pin reSllnonce, Chapman ami lI an , 1986.
*E<hted hy G. 1'. Rado and II Suhl, Academit: rr...,s.
 -"---~-~~

FERRO. ANn •TlFEftIlOMAC Nl:..IC RESONANCE

M Sperb, Fcrrcmwg' le/ic ndoxflllorl theory. Mt:Cr,m- lhll, 1964,
S Fones", -AnlLff'rromagnclic and f..,rri magoehc resonance,- Magrld ...... I. 384.
C W. ll aas and II . 8 . C,U.,n, - Fcrromagncl,,, relaxation a....I.('S()Oa/K'I!! IiIlC W>dli,s:' M ogrwlISlu I,
450.
It W. Damon, -Ferrurnagnchc rcSOllllllce at j';gh power:' Mag"num I, 552.

QUAJ'Io'11JM ELECTfION1CS
C. J. F. Troup, MIl$Cr $ and /(Jun, 21\.1 etl • lI alst~'(] , 1973.
A YaT;\" QUflnlu", electro.da. Wiley. 3rd.,.j . 1969.
J W Orton, O. II l'axman. and J. C. Wan,ng. Solid 6tale ",ucr, Pergamon. 1!11(l.
II . M. fl,'tU5enzvcig. IlIf roo:/ .. crlo<l / 0 quon/"", optics, Coni"" and Breach, 1973.
 ...

-
 17
Noncrystalline Solids
.
DIFI~RAcnON PATfERN 521
Monatomic amorphous mate rials 522
nadial distributiOIl function 523
Structure of vitreous silica, SiO,; 324

CLASSES 1527
Viscosity and the hopping rate S28

AMOHPliOUS I~ER nOMAGNETS 521)

AMORPHOUS SEMICONDUCTORS 530

LOW ENERCY EXCITATIONS IN AMOHPHQUS SOLIDS 532

Heat Cllllacity calculatioll 533
Thermal ccmductivity 534

FIBER OPTICS 53S

Rayleigh au enuatioll 53(;

ItEFEHENCES 537
 I
(
 .
-- CHAPTER 17: NONCRYSTALLINE SOLIDS

It is generally agreed that the terms amorphous solid, noncrystalline solid,

disordered solid, glass, or liquid have no precise strllctural meaning beyond tile
description that the structure is "not crystalline on any significant scale." The
principal structural order present is imposed by the approximately constant
separation of nearest-neighbor atoms or molecules. \\le exclude from the pres-
ent discussion disordered crystalline alloys (Chapter 21) whcre different atoms
randomly occupy the sites of a regular crystal lattice.

DIFFRACTION PATTERN

The x-ray or neutron diffraction pattern of an amorphous material such a~

a liquid or a glass consists of one or more broad diffuse rings, whcn vie\\1!<.i on
the plane nQl'mal to the incident x-ray beam . The pattern is different from
the diffraction pattern of powdered crystalline material which shows a large
number of fairly sharp rings. The result tells us that a liquid docs not have a
unit of structure that repeats itself identically at periodiC intervals in three
dimensions.
In a simple monatomic liqUid the poSition!'! of the atoms show only a short
range structure referred to an origin on anyone atom. \\le never find the center
of another atom closer than a distance equal to thc atomic diameter, but at
roughly this distance we expect to find about the number of nearest-neighbor
atoms that we find in a crysta1iine form of the material.
Although the x-ray pattern of a typical amorphous material is distinctl),
different from that of a typical crystallinc material, there is no sharp division
between them. For crystalline powder samples of sma1ier and smaller partide
size, the powder pattcm lines broaden continuously, and for sma1i enough
crystalline particles the pattem becomes similar to the amorphous pattern of a
liqUid or a glass.
From a typical liquid or glass diffraction pattern, containing thrce or four
diffuse rings, the only quantity which can be determined directly is the radial
distribution function . This is obtained from a Fourier analysis of the experimcn-
tal x-ray scattering curve, and gives directly the average number of atoms to be
found at any distance from a given atom. 'i11e mcthod of Fourier analysis is
e<lually applicable to a liquid, a glass, or a powdered crystalline material.
It is convenient to begin the analysis of the diffraction pattern with
Eq. (2.43). Instead of writing it for the structure factor of the basis, we ",{rite the
slim for all the atoms in the specimen . Further, instead of specializing the
scattering to the reciprocal lattice vectors C characteristic of a crystal, we con-
sider arbitrary scattering vectors 6.k :: k' - k , as in Fig. 2.6. We do this be·
cause scattcring from amorphous materials is not limited to the recip.rocallat-
tice vectors, which in any event cannot here be defined.

'"
-
 Tllercfore the scdttcreu amplitude from an IImorphous materi.ll is de-
scrihed hy

(I)
"
with!.;, the atomic form factor of atom, as in Eq. (2.50). TIlc sum runs ovcr ali
atoms in the specill\en.
The S('.lttered intensity :It scattcring vcctor 6.k is given by

1= S·S = L LI,J.. exp(i.6.k · (r ... - r ..)J • (2)

'" .
ill units referred to the scattering from a Single electron. If 0 denotes the angle

between 6.k and r '" - r ... then

. "
""''here K is the magnitude of 6.k and r,nn is the magnitude of r,,, - Tn '
(3)

In.ln amorphous specimen the vcctor Tm - r .. rna)' lake on all orientations,

so we average the ph.lsc factor over a sphere:

(cxp(iKr cos a» "" - '- 211"

411" _\
r
d(cosa) cxp(iKr,..n cos 0)
(4)
sin Kr.....
Kr..,..
Thus we havc the Oehye result for the scattered density:

I = L L (/"./.. sill Kr~",)IKr..", (5)

M "

Monalomic AmorphOlU Malcrial&

For atoms of only onc type, wc let 1m = I.. =< I and se parate out from the
summation (5) the terms with 11 = m. For a spctimen of N atoms,

1=Nr[ 1 + 2:' (sin Krm")lKr,,,.. ] . (6)

'"
The sum nms ovcr all atoms 111 cxcept the origin atom m = n .
If p(r) is thc conccnlmtion of atoms at distarlCe r from a rcferel'lCC atom,
\\'c can "Tile (6) as

1= NP[ I + LR dr 47Tr2p(rXsin Kr)/Kr] • (7)

where R is the (very large) radius 01 the SI>CClJIlI:'"II ................. _~. __
concentratioll; thell (7) may I>c written as

I = Nr{ 1 + 1'1 dr 41T,-2(p{r) - Po)(sin Kr)/Kr + (A/K) r dr 41Jr sin Kr}

(8)
The second integral in (8) gi"es the scattering from a uniform concentralion alld
ma), be neglected excel)t in the forward region of very small angles; it reduces
to a delta function at the origin as R _ 0';1.

Radial Distribution Fllnclirm

It is convenient to introduce the liquid structure factor defined by
(9)
Notc that this is not at all the same as S(6k) in (1). From (8) we have. after
dropping the delta function contribution.

S(K) = 1 + Loo dl' 41TT2(p(r) - Pu](sin Kr)/Kr . (10)

We define the radial distribution function g(r) such that

",) = "',)"" . (11)

Then (10) becomcs

S(K) = 1 + 41TPo L'" dr(g(r) - l]r~sin Kr)/Kr

= I + Po f (11' (g(r) - IJ exp(iK' r) .
(12)

because (sin Kr)/Kr is the sphcrica\1y symmetric or s term in the clpansion of

exp(f.K . r).
Oy the Fourier integral theorem ill three dimensions.

b1r) - I :: _ 1 _ JdKIS(K) - I)cxp( -iK'r)

8';>""
(13)
= ! J
27f Por
dK (S(K) - 1) K sin Kr .

This result allows us to calculate tbe radial distribution function g(r) (also called
the two-atom correlation function) from the Illc.tsured structure factor S(K).
One of the simplest liquids well suited to x-ray diffraction study is liquid
sodium. The plot of the radial distribution 41frp{r) vs. r is given in Fig. I.
together with thc distribution of neighbors in crystalline sodium.
'"
~.

, !l
, If u

,
u

,
(~
"
VJ' ' [,
"'" .)
i [,
,, , ,
• • "
Figure I (8) II,.)ia! diSiribulion (:"Irve 4...rp(r) ro.. liquid sodium. (b) A"erogc dcn.ity cur...e
4"'1'2",. k) l)i,lribution of neighl>or.i in crystallir>e sodium. (After T9r1UOV amI Warr('n .)

Structure of Vitreous Silica, SiO!

Vitreous silica (fused quartz) is a simple glass. The )(-f3)' scattering curve is
given in Fig. 2. 111e radial distribution curve 41fr 2p(r) vs. r is given in Fig. 3.
Because there arc two kinds of atolns, p{r) is actually the superposit ion of two
electron concentration curves, one about a silicon atom as origin and one about
an oxygen atom as origin.
lnc first peak is at 1.62 A, close to the a\ICragc 5i-0 distance found in
clystallinc silicates. The )I-ray workers conclude from the intensity of the first
peak that each silicou atom is tetrahedrally surrounded by four oxygen atoms.
The relative proportions ofSi and a lell us thai each a alom is bonded 10 two Si
atoms . From the geometry of a tetrahedron. Ihe 0 -0 distance should be
2.65 A, compatible with the distance suggested b), the shoulder in Fig. 3 .
The x-ra)' re!>uh s are consistent with the standard model of an oxide glass.
due to Zachariasen. Figure 4 illustrates in two dimensions the irregular struc-
ture of a glass and Ihe regularly repeating structure of a crystal of identical
chemical composition. The x-ray results are completely explained by picturing
glass)' silica as a random network in wllidl each silicon is tetrahedrally sur-
rounded by four oxygcns, eaeh ox)'gen bonded 10 1\.\"0 silicons, the two bonds to
an oxygen being rouglll)' diametricall)' opposite. The orientation of one tetrahe-
dral group \.\ilh respect to a neighboring group about the connecting Si-O-Si
 .. .

I\..
,
" -,-, " . "' "
Fig"." 2 Sc<iI lerw x· ra~ jnll"lIsily \'$ SC'allcrin, allgle 8, ~r >ih'eous SiD. (After B. 1:::. W;IO"eIl .)

bond can be practically random. 111cro is a definite structural scheme involved :

each atom has a definite number of nearest neighbors at a definite distance, but
11 0 unit of structure repeats itself identically al regular intelVals in thrcc dimen-
siom, and hence the material is not crystalline.
It is not possible to explain the x-ray rcsuhs by assuming that vitl"CQus silica
consists of vcry small crystals of some crystalline form of quartz, such as cris-
toballite. Small angle x-ray scattering is not obselVed. but would be expected
from discrete partieles with breaks and voids ocr-veen the m. The sche me of
bonding in glass must be essentially continuous. at least for the major part of
the material, although the scheme of coordination about each atom is the same
in vitreous silica and in crystalline cristoballite.
TIle low thermal conductivi!)' of glasses at room te mperature, as discussed
presently, also is consistent \\ith the continuous random network !Il~el.
A comparison of experimental and calculated n.'suits for amorphous germa-
nium is sho\\1l in Fig. 5. TIle calculations arc for a random network model and
for a microcrystallitc model. 111C latter model gives a very poor agreement . The
random network model is supported for amorphous silicon b)' st udies of the
band gap and spectroscopic work on the 2" shell.
52'

/
..I ,.."
/
I ..L- I
go S"S, ••

I(V

f"\
\
"
I / '" '" '" '"
Figure 3 Radial tlistribulion curve lOr ,·jtrCOlJl; S,u.. as the Fourier transform of Fijt. 2. The
[lO'Illions 0[100 !leaks give the tlisl;ux:es ofatoms frOI'll a s,li<IoIl or an n.yw:n. From the areas unJo.....
the peaks it is l)OI;slble tu l'ak,,1atc the numb.: r of neighbors al t hat tlistance . ,,, .. vertical lines
mt.hcal., tht.- fint lew """rage in lcntOll'lic w$l.:I~; the heights of lhe lines arc llroporliooal to the
IWUk afca.s. (Aner 8. E. \Van-en .)

»--«
»- -« »-
-« »- ...(
)0- -« »-
-« »- ...(
»--«
Fig"re 4 Schematic ,,,,"O·,jhnensiow,1an:ll"fo:~ ,Jlmtrating the lIiffcrer.ces between: M the rcgu!.orly
n:p0'3ting ) truclure of a crySlaI and Ib) ronlinuous random "","'urI.: rJ a gIru.s. (Aficr Zachariascn )

-
 ..,
r,
" " -i
0.0
,, , --- i
i
>
.~
- 2.0
'-
1I
I '"
i
l <0

.,
- .0

,ao ."
0
" " '"
Figure S CumlMson of experimental (dashed curve) and rnkulatcd (solid CW'Ve) r«llK'ed inten-
sity function for amorphous germanium. (a) '""orphow germanium comlJllred wit l, ,nlcrocfyst;&lhtc
model. (h) Amorphous germanium C(lmll".red wIth random n<:two,,\; model. (Results by J. Cra~'2~k
Ind I'. Ch:lIlJhn.)

GLASSES

A glass has the random Sin-cluTe of the liquid from which it is dcriveJ by
cooling below the freezing point. without crystallization. Also, a glass has the
clastic properties of an isotropic solid.
By gcner.t.1 agreement, we say that a liquid on being cooled becomes a
glass when the visCOSity equals lOll poise. where a poise is the CCS unit of
viscosity.' This defines the glass trnnsilion tcmpcrnturc T,_
At te mper.llures
above Til we have a liquid; below Til we have a glass. The Imllsiti('n is not a
thermodynamic phase transitioo, only a transition for "pr.tetical purposes. "
Relatively few liquids can be cooled fast enough in the bulk to form a glass
before crysta1lization intervenes. Molecu les of most substanCC'i have high

IThe SI unit or viscosity is 1 N$m -~. so that 1 poJi~ .. O. t NIIIl1 t. It iii quite C'QInmon to lind
viSCOSI ties gI''en in cp Or centipoise. ~nll 10- 2 iXl'~.
''''
enough mobility in the liquid so tJlal on cooling a liquid-w Ho melting transition
occurs a long time before the viscosity increases 10 10 13 poise or 1015 cpo
By controlSt, liquid wate r has a v iscosity I .B cp at the freezing point; the
viscosity increases enormollsly on freezi ng. \ Ve can ofte n make a glass by de-
positing a jet of atoms of a substnlte cooled to a low tcmperaltll"c, a process
which will sometimes produce an amorphotls layer with g1asslikc properties.
Amorphous ribbons of some metal alloys may be prodUl'cd in this "~J.y in indus_
trial quanlities.
The value 10 13 poise used to define Til is arLitra ry, Lut not unreasonable.
If we bond a slab of glass 1 em thick to two plane par.ulel vC ltical surfaces, the
glass will now perceptibly in one year under its own "'eight when the viscosity
drops below 10 13 poise. (For comparison, the viscosity o f the mantle of the
earth is of the order of l cP poise. )

Vucosity arid tile llopping Rate

"Ine viscosity of a liquid is related to the rdte at which molecules undergo
thermal rearrangement on a local scale. a:; by hopping into a vacant neighbor
site or by interchange of two neighbor molecules. The physics of the transport
process is somewhat different from that of vi5(."osity in the gas phase, but the gas
phase res ul t gives a qualitative lower li mit to the viscosity of the liquid phase. a
limit that applies to nearest-neighbor hopping of atoms.
The ga:; result (TP 14 .34) is
"I"j =ipCl , ([4)
where Tj is the viscosity, p the density, (; the mean the rmal velocity, and f the
mean free path. In the liquid l is the order of magnitude of the intc nnolccular
separation a. With "typical" values p " 2 g cm - 3 , C - Hi' em s- \ a =-
5 X lO- ti em, we have
Tj(min) - 0.3 X 10- 2 poise =. 0.3 cp ([5)

as an estimate of the lower lim it of the viscosity of a liquid. (rabies in chemical

handbooks onl)' rarely list values below this .)
We give now a very simple model of the viscosity of a liquid. In order to
hop successfull y, a molecule must sumlOunt the potent ial ene rgy barrier pre-
sented by its neighbors in the liquid. The preceding estimate of the minimum
viscosity applies when this barrier may be neglected. If the barrier is of height
E, the molecule will have sufficient thermal energy to pass over the barrie r only
a frdCtion
f- cxP{ - Elk,/f) (16)

of the time. Here E is an ap propria te free energy and is called the activation
 "."

11 Nonn-yolalIinr SalidI: 529

energy for the process that detcrmines thc rate of hopping. It is related to the
activation energy for self-diffilsion treated in Chapter 18.
11\e viscoSity will be increased as the probability of successful hopping is
decreased. Thus
7J - Ti.min}lf ..., Ti.min) cxp(Elks'f) . (17)
13
If 7J "" 10 poise at the glass transition, the order of magnitude of f mllst be
f "'" 0.3 X lO- 15 (I S)
at the transition, using (IS). The corresponding activation energy is
£Iks'fll = - In f = In(3 x 101~ = 35.6 . (19)
If TII - 2000K, thcn k s Tg =2.7 x lO- 13 erg and £=9.6 X lO- 12 erg _
6 eV. 11lis is a high potcntial energy barrier.
Classes with lower values of Til will have correspondingly lower values of
E. (Activation energies obtained in this way are often labeled as EVi"") Materials
that are glass-formers are chamcterized by activation energies of the ordcr of
J eV or more; non-glass-fonners may have activation energics of the order of
0.01 c\'.
When being pressed into molds or drawn into tubes. glass is used in a
range of tcmperatures at which its viscosity is 103 to 106 poises. The working
range for vitreous silica begins over 2O()(fC, so high that the prdCtical useful-
ness of thc material is sevcrely limited. In common glass about 25 percent of
NazO is added as a network modifier to Si0 2 in order to reduce below lOOO"C
the temperature needed to make the gla.<;s fluid e nough for the forming opera-
tions needed to make elc<:tric lamp bulbs, window glass, and houlcs.

AMORPIIOUS f 'ERROMAC NETS

Amorphous metallic alloys are formed by very rapid quenching (cooling) of

a liquid alloy, commonly by directing a molten stream of the alloy onto thc
sUlface of a rapidly rotating drum. 111is process produces a continuous "melt-
spun" ribbon of amorphous alloy in industrial quantities.
Ferromagnetic amorphous alloys were dcveloped because amorphous
materials have nearly isotropic properties, and isotropic materials should have
essentiall)' "l.ero magnetocl"ystalline anisotropy c ne r!.')'. As discussed in Chap-
ter 15, the absence of directions of hard and casy magnetization should result in
low cocrcivities, low hysteresis losses, and high permeabilities. Bec.lUse amor-
phous alloys are also mndom alloys, their electrical resistivity is high. All these
prope rties have tcchnological value for application as soft magnetic materials.
The trade name Metglas is attached to sevcral of these .
'nle transition metal-metalloid (fM-M) alloys are an important class of
magnetic amorphous alloys. The transition metal component is usually about H()
percent of !<'e, Co, or Ni, with the metalloid component e, C, Si, P, or AI. TIle
p resence of the metalloids 10\\ crs the melting po int, making it possiblc to
que nch the alloy through the glass transition tempe rature rapidly enough to
stahilize the amorphous phase. For e)(.lmple, the composition FewBro (kno\\llJ
as Mciglas 2605) has 'r g = 441 °C, as compun:d with the melting tcmpcr.lture
1538°C of pure irOn.
111e Curie temperature of this composition in the amorphous phase is
647 K, and the value of the magnetization M~ at 300 K is 1257. compared with
Tc = 1043 K and .H$ = 1707 for pure iron (fable 15.2). TIle cocrcivity is
0.04 C. and the ma>;imum value of the permeability is 3 X Ifr'1. Coercivities as
low as 0.006 C have been repo rted for another composition. 2
lligh coercivity llIaterials can be produced by the s'lIne melt-spill procc~s
if the spin rolte or quench rate is decrcdSCd to produce a fine -grained cry~ taIline
phase, which may be of metastolble composition. If the Wolin sil.e is amillged to
match the optimum size for Single domains, the coercivity can be quite high.
J. L. Croat has reported lIn = 7.5 kG for the metast;}ble alloy Ndo~Fe0.6at the
optimum melt-spin velocity 5 III 10 - 1. This lIlaterialmay (:OIne to have pr.lcticaI
application. Figure 6 ~hows the pC;lk in the coerdvity of a Sm-Fe allo)' as the
spin velocity is varied.

AMORPI-IOUS SEMICONDUCI'ORS

Amorphous sem iconductors can be prepared as thin film s by evaporation

or sputtering, or in some matcrirus dS bulk glasses hy supercooling the melt.
What happens to the e lectron e nergy bollld model in a solid without regu-
lar crystal line OJ-der? The Bloch theorem is not applicable when thc stnlcture is
not IlCriodic. so that the electron states cannot be descrihed by well-<lcfint="d k
values. Thus, the momentum selcction nIle for optical transitions is rcla'l:ecl;
hence all infrared and Raman modes contribute to the absorption spect ra. 1ne
optical absorption edge is mther featureless. Allowed bands and e nergy g;tps
still occur because the form of the de nsity of states vs. e ne l'b'Y is determined
mml strongly by local electron bonJing configurations.
Both electrons and holes can carry current in an amorphous semiconduc-
tor. TIle carriers may be scattcred strongly hy thc Ji~ldered structure, so that
the mean free path may sometimes be of the order of the sc-.t!e of the disorder.
Anderson proposed that the states ncar band ed ges may be locaJized and do not
extend through the solid (Fig. 1). ConJuetic.n in :'hese states may take place by
a the rmally-assisted hopping process, for which the Iiall elTect is olllOlllalous
"'Ie "alues are Iak...-. rrom the eKcellenl review by F. E. Lubarsky. ~ AllInl1lhous ferronyg·
nets,~In E. I'. Wnhlrarth , ed. Ferron ... gnelic mol&,,.....I,, \'. I. North. lI O;bnd. 1960.
 I

-
c "
~
,
<

-I-

,
" ,
". (m/t.) "
j
..
l--igure 6 Coerd\'lIyat room tcmpo:roture "5, meh-lopin "clocity V. lOr Sfllo4Feo ... The InlUimum
rocrdvl ty 15 24 kG and <ltturs at 1.65 m s L. which i$ believcd to conespo" d to tingle dOlTlolin
behavior IR each cf)slallilf'. At highI'. Ipm fillet Ih", coe rrivity deocn:a5eS b«.\use the depoSited
m:ucnal becoInes amorphous (more Isotropic). AI lowe r spin niles the crystallites anneal tn sitts
above the ~ingle domain Il':gIlTle; domain boundaries gh"e a lower ooerrivity. (After J. I. Croat.)

and cannot be used to d etermine the carrier concentration. TIle qu'lIltum Hall
f'freet (Chapter 19) provides c1I:pcrimcntal evide nce for localized states and
mobility edges.
AmOl1lhous materials appear to behave almost like intrinsic semiconduc-
tors, with the Fermi levcllying near to the center of the gap. Hov.'ever. the
pinning of the Fermi level is known 10 arise from the presence of ddccts, such
as dangling bonds and other misfits in the structure, wllich produce Ioc-.dizer.i
st.iles in the gap. The presence of these states, along with the tendency in
covalent .tlloys for 10C'.u valence requirements to be satisfied. necessitates the
usc of special techniques for preparation of dopcr.i specimens.
Two distinct classes of amorphous scmiconr.iuctors arc wir.icly studied:
tetrahcdrally-bonr.ied amorphous solir.is such as silicon and gennaniulfl, and
the chalcogenide glasses. The lattcr are multicomponcnt solir.is of which one
major constituent is a "chalcogen" clement- sulfur, selenium, or tellurium.
111e tetraher.imlly-boudcd materials have properties similar to those of
the ir crystalline forms, provirJed the dangling-bond defects are compclIsatcr.i
Figure 1 Density of ek'Clroll state~ as bdie\'l'() to OC<'\Lr ill alllcIIvhous soli..!s, whell st-.ll.., arc
ll on·loc-aliL.l'd ill Ihe <X'IIler of Ihe band. 1.ocaIi7-<.-d ~t;ltes ;'Ire shown sh,Kled. The mobi lity band
ed);t's E,.. E; Sf.1xmUc the ...nges fi energy ",hell' stal'"-\ arc 10000hwd ~nd non·\OCIdiwd. (After
K lI.tott and E . A. D;l\·is.)

with hydrogen. l'ley can be doped with small amounts ofchetnical impurities,
and their condllctivitycan be sharply modifkd by injection offrcc carriers from
a metallic contact. By contrast, the chalcogcnide glasses arc largel}' insensitive
to chemical impurities and to free carrier injection.
Amorphous hydrogenated silicon is a candidate material for solar cells.
Amorphous silicon is a much less expensive material than single crystal silicon.
Attempts at using pure amorphous silicon, however, faiW because of struc-
tural defects (dangling bonds) which were iU'Ipossible to eliminate. Introduction
of hydrogen into amorl>hous silicon appca~ to remove the undesirable stnK:-
ture defects. Relatively large proportions of h)'<lrogen are incorpor.lted, of the
order of 10 percent or more.

LOW ENERGY EXCITA1l0NS IN A~IORI'HOUS SOLIDS

There is a great regularity in the unusuulthermal properties of amorphous

materials at low te mperatures, yet the experimental discovery of these proper-
ties w·..tS entirely unexpected, completely different from crystalline solids, and
not ....'ell unde rstood in terms of amorphous structure.
The low lempemture heat capacity of pure dielectric clystalline solids is
known (Chapter 5) to follow the Oebye .~ law, precisely as el(peCI<!d from the
excitation of long wavelength phonons. The same behavior was Cl.pected in
glasses and other amorphous solids- the point was SO obvious that it did not
e ncourage experimental investigation.
After early indications by a Bcrkelt.')' group, Zeller and Pohl discovered
that many insulating glasses show an uneJ>."p<.oclcd lillear tcnn in the heut capac-
ity below 1 K. Indeed, at 25 mK the obsel""ed heat capacity of vitreous silica
exceeds the Deb),c phonoll contribution by a factor of 1000.
Anomalous linear terms of comparable magnitude are fo und in all, or
nearly all, amorphous solids. 111eir presence is believed to be an intrinsic con-
S('quencc of the alllorphous statt..'S of maliC!", but the details of why this is so
 "

remain unclear. Work by lIunklinger and associates gives strong evidence that
the anomalous properties arise from two-level systems and not from multi-level
oscillator systems; in brier, the evidence is that the systems can be satumted by
intense phonon fields, just as a two-level spin system can be saturated by an
intense rf magnetic field (Problem 16.6).

Ileal Capacity Calculation

Consider an amorphous solid with a concentration N of two-level s)'stems
at low energies ; thai is, with a level splitling 11 much less than the phonon
Deb)"c cutoff kilO. The partilion function of one system is, with T = koT.
Z = cxp(1lI2'T) + cxp( - Al2.,.) = 2 co5h(A121") (20)
The thennal a\'crage enerb'Y is
U = - it.. tanh(a/2T) (2 1)
and the heat capacity of the single system is
C v = ks<aUlaT) = k8(Af2T)2 sech 2{Ar.h) (22)
These results arc given in detail in TP, pp. 62-63.
Now suppose that I::. is distributed with uniform probability in the range
I::. "" 0 to I::. = 1::.0 • 111e average value of Cv is

Cv = (kw'4~) r~
, dl::. (1::.2/l::.o)sech2(612'T) (23)
(-
21<-'-')
"" ,
1""" clx xi sechix

The integral cannot be carried out in closed form.

Two limits are of special interest. For'T <C Ao. the sech2x tenn is roughly I
from x = 0 to x = I , and roughly zero for x > 1. 111e value of the integral is
roughly U3, whence
C" ... 2kiT131::.o , (24)

for 'J' < t::.dk B .

For 'T ~ 6 0 , the value of the integral is rou~ly i(lid'lJ..BT'f, so that in this
limit
(25)
which approacill.>S zero as T increases.
Thus the interesting region is at low temperatu res, for he re the two-level
system contributes to the heal capacity a tenn linear in the temperature. This
term, originally introduced by w. ~1arshall for dilute magnetic impurities in
metals. ha.~ no COnnection with the usual conduction e lectron he:at C'dpacity
which is also proportional to T.
The empirical result appears to be that all disordered solids have
'"
100

I
I
I
I
I
I
I
I
I
I
Co.o.(SiOJ-_/
I
b I
01
Oltl 004 01 0.'
"
.' isuI'1l8 1I1':3t GlpaCit)' or ,,(reouli "I,ea and KKIa sil.ca Kl..,.~ <IS a funch on dlemp"'""u.!':, 1be hcat
capaclt) .5 roughly hneu on T below I K.. The o.bsht-d line rcpr"'iiCnts t he calcu lalL-d Debyc heal
CApaolv of .;I,NUS Sil ica_

N - 10 17 em --3 "new type" low energy excitations uniformly d istributed in the

e nergy interval from 0 to 1 K. TIle anomalous specific heat can now be obtained
from (24). For 'f = 0.1 K and fldka '" 1 K.
C" = jNkll(O. I) .... 1 e rg cm - 3 K- 1 . (26)
For comparison, the phonon conlriblltion at 0.1 K is, from (5.35),
e" .... 234Nk n(TIO't - (234X2.3 X 102~(1.38 x 10- 16)(0.11300)3
(27)
_ 2.8 X 10- 2 e rg cm - 3 K 1 •

much smal ler than (26).

The experimelltal results (Fig. 8) for vitreous Si0 2 arc represented by
C" "" CIT + C3'!'l , (28)
whe re CI = 12 erg g - I K- 2 and C3 "" 18 erg g - I K- 4 .

TI,ermal CO,ldllctiviry
The the rmal conductivity of g1a'iscs is very low. It is limited J.t room te m-
pe ratu re and above by the scale or the disorder or the strllct ure, ror this scale
determines the mean free path or the dominant thermal phonons. At low tem-
perature!>, below L K. the conduct ivity is C'drried by long wavelenp,th phonons
 Ii ;>.' rJllr'1lftalline SoI;./.

I
T
'T
I
H~ure 9 Short Imonon ",..'an free patl, in a dl';On\cn:d .tructur.... A ~h()tt "", C'l ... nglh phonon tl~'1
dblli;Klf'S alOm L, .IS sho"n, wIll .lispl:;...., .. toon R by a much sm"lIcr distann: , !J«"us<' of the ph:;.o::
('anedL~tion of the "pper dnd Iowt:r palhs f, om L 10 It Tho:: displacelllenl of R is t + 1 - 0, So
Ih"l Ihe wave innd""t flOm L is ren ..-ct,'d .. t n.

and is limited by phonon scattering from the TnvstcriOlls two-Ie\ cI systems or

tUllneling states discussed earlier for their colltributioll to the Iwat clIpacity of
amolllholiS solids.
As in Chapter 5, the expres~ion for the thermal colJdodi\ ity K has the
ronn
K=kt;f , (29)
where c is the heat capacity per unit volume, v is an average phonon velocity,
and t is the phonon mean free path. For vitreous silica at room tcrnperd.ture,
Ksl.4xlO - 2 Jcm - ls IK 1

c "S!; 1.6 J em 3 K - 1 ;
(0) "S!; 4.2 X lOS em S-I
l1ms the mean free path f "S!; 6 X lO- 1j cm; b)· reference to Fig. 3 we see that
this is of the order or magnitude of the disorder of the structure.
Thi~ value of the phonon mean free p;'lth is remarkably short. At room
tempe ....tturc and above (that is, above the Debye tempcraturc\ most of the
phollons have half-\v3velengths or the order or the intemtomic spacing. It is
through phase C".lIIccllation processes, as in Fig. 9, that the mean free path is
limited to scvcr.tl intcrJ.tomic spacings. No other structure for fu sed quartz will
give a I) A mean frce path .
The normal modes of vibration of the glass structure are IItterly unlike
plane wavcs, BlIt the modes, as distorted as they may seem, still have quan-
tized amplituJes and may properly be called phonons. The thermal conductiv-
ity below room temperature aud particularly below 1 K is tre:"lted in detail in
the book edited by W. A, Phillips that is cited at the end of this chapter.

FIBER OPTICS

Fibers of silica-based Iightguides carr)' a high proportion of the data and infor-
matiOll transmitted on the surface of the earth and under the M!d.5. and are the
nu.'d ium of ehoice for the commull)cations industry. 111C optical fiben consist of
a thin core (- lO /.Lm) of high-refractive index glass surrounded by a cladding.
536

,-
E
"
;;;3. 0.4
,, ~- 25T1"---I
3 ,,
,
0.2
,,,
,,
"' ,"
12 13 I.' 1.5
Wave1cllgtb (pon)
1.6
"
Figure 10 1be trallsnlissioo. characteristic of cornlllunicat;';"-qualily optical fibers. wowing t}..,
aflClluwl ion In UIl;tJ of decibels per ~m as II (uOlction olthe wav.,length oflight, ill ",m. 1111! Ila)'1e;gh
ICtttering noglrne is domililini on the left of the CUT""', e~cept for II strong impurity absorption Ii"e
associliled with 0 11 ions llIat aceQmpoilllY SiO. ; Ihe Ii"" i. the second harmOllic of II lint: at 2.7 Ilm
)mown as the ~w:lIer I;ne. ~Th~ wlwelenglh mll.ked at 1.31 ,..m is used h. 1994 1<8115ml5.5lo" lilies; it
will be ~Illaced by the wa...,leng th 1.55 /-1m .\1Iilaille from E,,3+ iOIl laser ampliO" f$, "fIlch are
used ,,\"t::ry 100 km in typicallol1g di~tatlCe 'lIIdersea IIpplicatiollS. The power nceded 10 pump th.,
arnpllfw.n I~ liupplied by copper wires. {Court",,}, of Tiugye U , AT&T Bell LabonI0ric5. )

The digital data are carried by light, with a minimum attenuation near 0.20 db
km- ' at wavelengths ncar 1.55 JA.m, which is in the infrared (Figure 1.0). A
range of 100 km corresponds to a loss of 20 db, which is readily supplied by an
Eu 3 + laser amplifier.
The optic window of high -purity glasses near this wavelength is limited o n
the low lrC<Juency side by phonon absorption bands and on the high frequency
side by Rayleigh scattering and, ultimately, by electronic absorption. In the
optie window the losses are determined by the Rayleigh scattering intrinsic to
static Ouclualions in the local dielectric constant of an inhomogeneous mooium.
and the attenuation varies as the fourth power of the frequency.
It is fortunate that an e)(cellent source is available for mdiation at ].55 JA.m .
As shO\vn in Figure 16.24, e)(citoo (pumpoo) erbium E~ ~ ions can amplify in
an erhilllll-doped section of fiber.

Royleig/l attenuation
l1le attenuation oflighl waves in gL'lSS is dominated at short wavelengths in the
infrareU by the same scattering process, called Rayleigh scattering, that is re-
sponsihle for the blue light of the sky. l1le extinction coefficient , or attenuation
coeffICient, II has the dimension of rcciprocallcngth and for light scattered in a
gas is given after hdyleigh hy
11 = (2w4/3m;4N){(n - I )~, (30)
where n is the local relmctive index and N is the number of scattering centers
per unit volume. The energy flux as a function of distance has the form
exp(-hx}.
Derivations of (JO) are found in good texts on electrodynamics; the struc-.
ture of the result may he uuderstood hy a gcueral argument: The radiant en-
ergy scattered from a dipole element l ' is proportional to (dp2/drt. and this
accounts for the factor w4 • '1be local polari7",,'lhility a enters as if; if there are N
random scattering centers per unit volume, the scattered energy averaged over
these random sources will go as N({l:J.a'f), or ({l:J.n'f}IN. 'Thus we have the essen-
tial factors that appear in (30). As applioo to a glass, I:J.n should refer to the
variations in polari7.ation around each group of Si- O bonds, and s.'ltisfactory
numerical estimates of the attenuation may he made in this way.

Problem
I. Metallic Optic Fibrc3? In Nature 319, 124 (1996) it is speculatoo tlmt metallic wires
can actlfS optic fibres, transm itting light witll a long delay appropriale to th e lligh
refractive index characteristic of metals. UlllOrtunalcly the refractive index of a typi-
eal mctalls dom inated by a frce-clcdroll tenn in illl, so that the propagatiOIl of a light
wave is in fact highly damped in a ml.1al. Show that in sodium (p. 331) at room
temperature a ""ave of vacuu m wave!t..·ngl h lO 1J.1l1 will ]Iave a damping lengll) of
0. 1 IJ.m. This may be contmstc<l with the 100 krn damping length fou nd fo r light in
high (Juality glass fibres .

References
S. 1\ Elliolt, PIolPia of ",no'l'/IOW 'ootmau, 2nd ed., Wiley 1990.
C . Keiser, Opticlll f.ber communical1Dm. 211d l.-d. . MeCra...~ III1l . 199 1.
II. -). Cunthl'rodl and II. Llec:k, eds., Mctullic.g!MSeli. 2vok , Springer, 1980, 1983.
N . M. Marcl" UquUl metub, concepts and lhoory, Cambridge, 1!:lOO.
W. A. Pilillips, ed., .... JJJorphcm 1IOlid.. I~Q tlrl..,~rlltun' pr<lpl!rliu, Spri'lgcr, 1981.
N. F. Motl I"d E. A. Davis, £/eclromc ~ in ~Iaalnf! materiDili, ODOnI, 2r.d ed.,
I!m>
J. Ziman, Modell of dl.wrtkr, Cambridge, \979.
C . o. J onc~ . CI<m, 2nd ed., Chipman ,Ind Hall, unl .
I\. Zallen, PlJl/stU of """"7,'"n~ roiidl, Wiley, \963.
F. Woolen and 0 Wea.ire , ~ Mockling let ....hedr:olly bandt'll ...."do'" networks by computer, " Solid
~laIe physk:l 40, 1 (1987).
Journal or Nou-Crystallinc Solids.
C. Vcnkatar, D . Sahoo, a"d v. BaIa"riShl lilll. Beycmd 1M cryflillfinf! stole, Sprillgcr. 19119.
C. S. CargIll, Structure of maul/Ie alloy g/ouu, Solid sl~te ph~io::s 30, 227 (1g75}.
Fum,"" l'o'lJczaw~. ~ CIass I ..nsiliou and .eb.ut iOf' c:f dl~red strudure.,~ Solid state physics
-'5, [19 (1991 ).
II . />1 ......la, Handbook of oplical f.bers and coblu. D o''''''''', 1988.
 18
Point Defects

LA:rnCI!: VACANCIES 541

DIFFUSION 544
Metals 547
COLon CENTEHS 548
f' centers 5-18
Other center) in alkali halides 548
l'ROBLEMS 551

l. Frenkel defects 551

2. Schottky vacancies 551
3. F center 552

nEFEHENCES 552
 t=.
D • •
•
I :J
oe +

Figu~ l A pl'-lie of a pure al kali halide .... rystal , ~how;ng a ''&ClInt positi"e iOt> ~i te. a ,-acant
nq;;.tl,,, ion sit." and a coupled I».;r of '>'1tCIint "tes of opposite sign .

...
 CHAPT£R 18: POINT DEFECTS

TIlC commOfl [Xlint imperfections ill crystals are chemical impurities,

vacant lattice sites, and extra atoms not in re~:,'ular lattice positions. Linear
imperfcctions ace treated under dislocations, Chapter 20. The crystal surface
is a pl.mar imperfection, with surface electron, pilOnon, and magnon states,
Chapter 19 .
Some important propertics of crystals are controlled a<; much by imperfec-
tions a<; by the nature of the host crystal, which may act only a<; a solvent or
matrix or vehicle for the imperfections. The conductivity of somc semiconduC"-
tors is due entirely to trace amounts of chemical impurities. The color and
luminescence of many crystals arise from impurities or imperfections, Atomic
diffusion may he accelerated enormously by impurities or imperfections,
Mechanical and plastic properties are usually controlled by imperfections.

LATIICE VACANCIES

The Simplest imperfcction is a lattice vacancy, which is a mlssmg

atom or ion, also known as a Schottky defect. A lattice vacancy is often indicated
in illustrations and in chemical equations by a square (Fig. I), We create a
Schottk y defect in a perfcct cryst..'lI by tmnsfen'ing an atom from a latticc site in
the interior toa lattice site on the surface of the crystdl, In thermal equilibrium
a l'Crtain numher of latti<.'C vacancies are always present in an othenvise perfect
crystal, because the entropy is increased by the presence of disorder in the
structure,
In metals with dose-packed structures the proportion ofl attice s itc ~ vacant
at temperatures just below the melting point is of tile order of 10- 3 to 10- 4 , But
in some alloys, in particular the very lIard transition metal carbides such as TiC,
the proportion of vacant sites of one component can be as high as 50 percent.
TIle probability that a given site is vacant is proportional to the Boltzmann
factor for thcrnt.'lI equilibrium: P = exp(-Evlk81'), where Ev is the energy
retluirecl to take all atom fmm a lattice site inside the crystal to a lattice site on
the surface. If there are N atoms, the eqUilibrium number 11 of vacancies is
given by the Boltzmann factor
n
- - = exp( - Evlkn'f) (1)
N- n
If 11 -« N, then
nfN ~ e.xp(- El'lk81') . (2)

If E v = 1 eV and r "" 1000 K, then nfN "'" e - 12 = 10- 5 ,

..,
 ....
'"
The equilibriu m concentration of vacancies decreases as the tempernture
decreases. The actual concentration of vaamcies will he higher than the et:lui-
Jihriunl value if the crystal is grown at an elevated temperature and then cooleJ
suddenl y, thereby freezing in the vacancies (sec the discussion of dilfusion
below).
In ionic crystals it is usually favorable energetically to form roughl y equal
numbers of positive and negative iOIl vacancies. The formntion of pairs of vacan-
cies keeps the crystal electroStatk~\ lIy neutm1 on a local scale. From a st.ltistica]
ca1cubtion we obtain
n eN exp(-E,I 2k8 T) (3)
for the number of pairs, where E" is the energy of fonnaliOl\ of a pail".
Another vacancy defect is the Frenkel defect (F ig. 2) in which an atom is
transferred from 01 lattice site to an interstitial position, a position not normally
occupied hy an atom. The calculation of the equilibrium numher of Frenkel
defects proceeds along the lines of Problem 1. If the number 11 of Fl'enkel
defects is m uch smalle r than the number ofL'\ttice sites N ,md the numbe r of
interstitial sites N', the result is
(4)
where E, is the energy necessary to remove an atom fmm a lattice site to an
interstitial position.
In pure alkali halides the most common lattice vacancies are Schottky
defects; in pure silvel' halides the most common vacrulcies are Frenkel defects.
Lattice vacancies are present in alkali halides when these t'Ontain aliditions
of divalent elements . If a crystal of KCl is grown with controlled amounts of
CaCI 2 , the density varies as if a K+ lattice vacancy were fomled for each Ca2 +
ion in the crystal. The Ca2 + enters the lattice in a normal K + site and the two
Cl- ions enter two CI- sites in the KCI crystal (Fig. 3). Demands of chargc
neutrality result in a vacant metal on site.
The experimental results show that the addition ofCaCI 2 to KCllowf'rS the
d ensity of the crystal. The density would increase if no vacancies we re pro-
duced, because Ca2+ is a heavier and smaller ion than K+ .
The mechanism of electrical conductivity in alkali and silver halide crystals
is us ually the motion of ions and not the motion of electrons. Tlib. has been
est".thlished by comparing the transport of charge with the transport of mass as
measured h y thc material plated out on electrodes in contact with the crystal.
The study of ionic conductivity is an important tool in the investigation of
lattice defects. Work on alkali and silver halides containing known additions of
divalent metal ions shows that at not too high temperatures the ionic conductiv-
Ity is d irectly proportional to the amount of divalent addition. This is not he-
t'(luse the divalent ions are intrinsically highly mohile, fol' it is predominantly
the monovalent metal ion which deposits at the cathode. The lattice vacancies
introduced with the divalent ions are responsible fur tho<> enhanced diffusion

-
(
 ,
• I

( -
•
Figure 2 ~dJOt tk)' and Frenkel defects in all ioule Cl')'5ln l. The a..ro.... s l"d,C;\le the displacemt'llt or
the kllls . In a Schottky defect the ion end> till on the ~"riace ,,( the Cl'}'5lal. In a Fn:nkd defect ,I
IS n:mo"ed to an ill ieu lilial ~hon.

o
tlg~ 3 Production of II Iauice \'ac:I.IlC)' by the soIutioll of c.a. in KCI: to ensure e~
o
1\l'utmity II positive ton vacllncy iI; jnrrod,K't'd into the LUke ",th ,:och dr'':Ilt'nt calion Ca". TIle
two CI ions of CaCl~ en ter normal neg:llive ion ,ill".

0000000 0000000 0000000

0000000 0000000 0000000
0000000 OOsgOOO 0000000
OOQ\)OOO 00 000 000 --000
0000000 0000000 0000000
OOOOOGO 0000000 0000000
0000000 0000000 0000000
1<)
figure" Three i>;osk rnO!<'hanism$ of difl""siQn: (a) Inlo:rchallgf! by roMlOn .1.0.... , II midpoint. ~Iore
th,," Iwo alo,",' may rotate tocethcr. (b) Migration through illtentitial sitel. (c) Atoms ellChan&<,
position ""h \'lK'IInt laltice sites. (After Sejt~)

(Fig. 4c). The difrusion of a vacancy in onc direction is equivalent to the diffu-
sion of an atom in the opposite direction . When lattice defects arc gener.tted.
thermally, their cncrgy of formation gives an extra contrihution to the heat
capacity of the cT)'!>tal. as shown in Fig. 5.
An associated pair of vacancies of opposite sign e"hibits an electric dipole
moment, with contrihutions to the dielectric constant and d ielectric loss due to
the motion of pa;"'" of vacancies. The dielectric rclaJIation time is a measure of
 "
iI
"
iI
I
I
I
" -c~~~~~~~,
----c;,------~
- 12" O.OO5T I
I
'~~-----c..
±'------..
~------"..
~-----c.~--J .

Figure 5 Ileal captCity or silver bromide exhibiting the e:<eeU heat capaeily from the rormatioll cl
lattice dtfe<:ts. (After R. W. Christy and A. W. l..a" ·501l.)

the time requiroo for one of the vacant sites to jump by one atom ic positiOl!
with I'CSPCct to the other. TIle dipole moment can change at low frcqllcncie...,
hut not at high. In sodium clllOlic.le the rela.>;:ation frequCflcy is 1000 S- I at
SS"c.
DIFFUSION

\Vhen there is a concentration gradient of impurity atoms or vacancies in a

solid. there will he a flux of these through the solid. In equilibrium the impuri-
ties or vacancies will he distrihutcU uniformly. TIle net flux J N of atoms of olle
species in a solid is related to the gradient of the concentration N of this species
by a phe nome nological relation called Fick's law:
J", :c - D grad N (5)
Here INis the number of atoms crossing unit area in unit time; the constant D
is the diffusion constant or diffusivily and bas the units cm 2fs o r m2fs. The
minus sign means that diffusion occurs away from regions of high concentm-
tion. The form (5) of the law of diffusion is often adequate, but rigorously the
gradient of the chemical potential is the driving force for diffusion and not the
concentration gradient alone (J'P, p. 406).
TIle diffUSion constant is often found to val)' with temperature as
(6)

here E is the activation energy for the process. Experimental results on the
dift·u sion of carbon in alpl13 iron are shown in Fig. 6. The data are representoo
hy E = 0.87 eV. Do "= 0.020 cm 2 /s. Representative values of Do and E are
given in Tahle 1.
 18 Point V tfa:b 545

•
•, . - ,
.
/:
• ./
/"
10 - " /
~ V
•
C) 10' , /
" • ;(;'
/
.. 2 3.8 3.4 3.0 2.6 2.2 1.6 14 10 0.6 0

,""""
):igure 6 Diffusia •• coefficient of <;arl)(ln in Iron, after Wert. The logari llun of 0 is di r~"CIJy propor-
tional to llT.

Table 1 Diffusion constants and activation energies

U(I&t Do E Host ,)" E

cry,'" AO~ cm"s- ' .V CI)"5ul Alom ants' ,V

C" C" 0.20 ' .04 Si AI 8.0 3.47

C,
C"
Ag
Z"
Ag
0.34
0.40
L il8
J.91
Si
51 ,,, 3.•
16.0
3.51
3.90
Ag
Ag
C"
A"
I.'
0.26
'.00
Lil8
Si
Si '"
51>
0.32
5.•
3.:-">6
3.94
A, Pb 0.22 1.65 51 U 2 X 10- 3 0.66
N, N, 0.'" 0.45 Si A" J X 10- 3 1.13
U U 2 X 10- 3 1.20 Co C. 10.0 3. 1

To diffuse, an atom must overcome the potential energy harrier presented

by its nearest neighbors. We treat the dilfusion of impurity atoms between
interstitial sites. The same argument ",i\l apply to the diffusion of vacant lattice
sitcs.
If the harrier is of height E, the atom will have sufficient thennal energy to
pass over the barrier a fraction exp(-Elkn'f) of the time. Quantum tunneling
through the hatTier is another possihle proccss, hut is usually important only
for the lightest nuclei, particularly hydrogen .
 ....
"6

If" i~ a characteristic atomic vibrational frequency, then the probability p

thai .sometime during unit time the atom will have enough thermal encrg)' to
pass over the barrier is
P "" JJ cxp( - Elk BT) .
In UIl!' lime the alom makes" pasSC.'5 al the barrier, with a probability
e\li. - Elk/!T) of surmounting the blm;cr 011 each try. TIlC qU.lntity J1 is called
the jump frequency.
We l'Qllsider two pamllcl planes ofimpurit)' atoms in interstitial sites. The
planes are separated by lattice constant n. There are S impurity atoms on one
plane and (5 + (I dSfcf.r) on the other. The net number of aloms crossing be-
tween the planes in unit lime is - -pa dSfdx. IfN is the total conccntration of
impurity atoms, then S "" aN per unit area of a plane.
The diffusion nu x may now be written as
IN - -l w 2(dN/dx) (8)
On comparison with (5) we have the rewit

I D:: 11(1
2
exp(-Elk n1') (9)

or the rorm (6) wi th Do :: 11(12.

If the impurities are char~ed , we may find the ionic mobility jl and the
conductivity u from the diffusivity by using the Einstein relation kn1/l = lJD
rrom 1'P, p. 406:
jl :: (qll(/2/k nT) exp(- Elk nl ') ; (10)
u = Nqjl = (Nllva 2/k aT) exp( - ElknT) (II)
where N is the concenlnltion or impurity ions or charge l/.
The proportion of vacancies is independent of temperature in the range in
which the numbe r of vacancies is determined by the Iwmber of divalent metal
ions. Then the slope of a plot of In u versus lIk nT gives E+, the barrier activa-
tion energy for the jumping of positive ion vacancies (Table 2). Diffusion is very
slow at low temperatures. At room temperature the jump frequency is of the
order or 1 S- I , and at 100 K it is of the orde r of 1O- :z:; 5 - 1,
'Ibe proportion of vacancies in the temperature mnge in which the (':oll<..'Cn-
h-ation of defects is determined by thermal ge neration is given by

(12)

\~he re Ef is the energy offormation of a vacancy pair, aet.:oruing to the theory of

Schottky or Frenkel dek'Cts. Ilere the slope ofa plot of In u versus l lknT will
be E + + iE" aet.:o rding to (10) and (12). From measurements in different t("m'

--
r TQ.ble 2
J8

Activation energy E+ ror motion of a l)()5ithe ion >aCllnt')'

"oin' lk/ed. Sfj

Vailles oflhe energy orrormatiOll ofa vacancy pair, E/. arc :llso given TIle IlUIllix-rs
given in parentheseS" for the silver salts Icfer to in ler~l ili.tl sil'-cr ions.

ery,ul E.(eV) EJ<eV) Wmcrs

Nael 0.86 2.02 Et7.c1 and ~1aurcr

LiF 0.65 2.68 II:wen
LiCI 0041 2. 12 llavcn
LiBT 0.31 1.80 lIa\'e n
Lil 0.38 1.J.1 lIa,-en
KCI 0.89 2.1-2.4 Wagner; Kelting 1111(1 \Vilt
AgCl 0.39(0.10) 1.4- Tchow
AgHr 0.25(0, II) 1.1- Complon

• ...or Frenkt-' d...f~>ct.

pernlu .'c mnges we determine the energy offormation of a vacancy pair EJand
the jump activation energy E+.
The diffusion constant can be measured by radioactive tracer tcchuiCltlCS.
The diffusion of II known initial distribution of radioactive ions is followed as II
function of time or distnnce. Values of the diffUSion (;onstant thus determined
may be (.'ompared with values from ionic conductivities. TIle two sets of values
do not usually agree within the experimental accuracy. suggesting the presence
of a diffusion mechanism that does not involve the transport of charge. For
example. the diffusion of pairs of positive and negative ion vacancies does not
involve the transport of charge.

Metals
Self-diffusion in monatomic melals most commonly pruceetls by lattice
vacancies. Self-difT\lsion means Ille diffusion of aloms of Ihe melal itself. and
not of impurities. The activation energy for self-cliffusion in copper is expected
to be in the range 2.4 to 2.7 eV for diffusion through vacancies and 5. 1 to
6.4 eV for diffusion through inlerstitial sites. Observed values of the activation
energy are 1.7 to 2.1 eV.
Activation energies for diffusion in Li and Na ('an be determined from
me-.tsuremcnls of the temperature dependence of the nuclear resonan("e line
width. As discussed in Chapter 16, the resonance line width narrows when the
jump frequcncy of an atom between sites becomes rapid in comparison with the
frequenc)' corresponding to the static line width. '111e values 0.57 eV and
0.45 eV were determined by NM R for Li and Na. Self-diffusion measurements
for sodium also give 0.4 eY.

(
'18
COLOR CEl'inms

PUTC alkali halide crystals arc transpa.rcnt throughout the visible rcRioll of
the spectrum. Thc crystals may be colored in a number of ways:

• b y the introduction of chemiCftI impurities;

• by the introduction of an excess of the metal ion (we Inay heat the crystal ill
the vapor of the alkali metal and then cool it qUickly-an NaCI crystal heated
in the prCSClll'e of sodium vapor ba'Omes yellow; a KC] crystal healed in
potassium vapor becomes magenta);
• by x-ray, y-my, neutron, and ek-ctroll bombanlment ; and
• by electrolysis. A color center is a lattice defect that absorbs visible light. An
ordinary lattice vacancy docs not color alkali halide cl"yslals, alt hough it af-
fects the absorption in the ultraviolet.

F Centers
11le name F center comes from the CCrTn.11l word for color, Ftlrbe . \Ve
usually prOOure F centers by heating the crystal in excess alkali vaJXW or by
x· irrodiation. The central absorption band (F wnd) associated with F centers in
several alkali halides arc shown in Fig. 7, and the qunntum energies are listed
in Table 3. Experimental propertit.'s of F centers have been investigatt.iI in
detail, originally by Poili.
The F center has been identified by elcdron spin resonance as an electron
bound at n negative ion vacancy (Fig. 8), in agreement with a model suggeslt.iI
by de Boer. When excess alkali atoms are addt.'lI to an alkali halide crystal, a
corresponding number of negative ion vllcancies nrc crented. The valence elec-
tron of the ;llbli atom is not bound to the atom; the electron mignltes in the
crystnl and bc<'omes bound to a vacant negative ion site. A negative ion vacancy
in a perfect periodic lattice has the cfTl,.'ct or an isolated po~itive charge; it
attracts and binds an electron . We can simulate the electrostatic effect of a
negative ion va(.'3ncy by adding a positive <.h'lrgc q to the normal charge - (J of
a.n occupied negative ion site.
'l'he F center is the simplest trappcd-clcctron center in nlkali halide crys-
tals. The optil.'3i absorption of an F center arises from an e lectric dipole trans i-
tion to a bound excited state of the center.

Other Centen in Alkali lIalidCll

In the FII center one of the six nearest neighbors of an F cen ter has been
replaced by a differ-ent alkali ion, Fig. 9. More complex troppcd-clcdron cen-
ters are formed by groups of F centers, Fig. 10 and 11. Thus two adjacent F
centers form an AI center. I hrec adjacent F centers form an El cel1 t('r. Different
centers are distinguished by their optical absorption frequencies .
 J8 I'oInt lkff!'(" 5-19

Ta ble 3 xperimentlll ,.. center lIilsOl"ption energies. in eV

Lie l 3.1 NaBr 2.3

NaCI 2.7 KBr 2.0
KCI 2.2 RbBr 1. 8
RLeI 2.0 LiF 5.0
C,C1 2.0 N,F 3.S
LiBr 2.7 KF 2.7

-
... "'" ... w~InA

• ... "'"' ... "'"' ,

ro<l c.a
, ocr
"'" II '0
. I
. ,\ .!"
, • , , ,
r
,I , ,
• • t:n"'IY ;n "v

Figure 7 The F bar,cl, for .cvernl alkal i halides: optical absorsJtIon .~rso.lS " lI.dength for crystal,
that contain F Ct.'I1ten.

"'igure 8 An '" cclltcr is a nega tive K.lI1 vacancy with one exce5ll de<'tron bound at the v.lCarlC)'. lhc
t1b1riburion of the ace" electron is la' gel)' on the fJO'itivc mda! io'lS .djan'll! to the WCl\Iltlaltice
'ite ,
 cr t d- @
0 e C1-

, f igure 9 An FA renter in KC]' (InC of the six K ' ion.

® I C1- "hieh bind all F renter is replaced Iw another a1bli ;vn,

here :,,\,, '

0 <-
•
0 { - (2) t f igure 10 An M <-'ell ie r ron.ists oflwoadjaccnl
F renters .

r "'" F

0
, •
0 0
,,(P
,!

.';glll"<) II An R <:enter consists of thrt:c .ulja.;-enl F celll..,TS; thai is. a group of three negative ion
"",--ancjcs in It [Il I J plane o£lh" NaC] structu re, wilh three aS$OCial~..! ck·,:frOtl$.

Holes may be trapped to form color centers, but hole centers arc not
us ually as simple as electron centers. For example, a hole in the filled 1)6 shell of
a halogen ion leaves the ion in a 115 configuration, whereas an electron added to
the filled p6 shell of an alkali ion lea\'cs the ion in a jl6S conflgllration.
nle chcmiStly of the two centers is different: p6S acts as a spherically
symmetric ion, but If> acls as an aspnmetric ion and. by virtue of the Jahll-
Teller efICct, will distort its immediate surroulldillgs ill the crystal.

(
 /8 "rn'" , 0,,/«1. ,
CI"
@ a
0 "' t·

® a . ' CI >. a

CI

€:)
0
CI"
a
CI

(Q-'
'"
(" " a-
- VI( «nlc.

CI
® a

• a ,
tlgun.' 12 A V" center IOrmcd wllell a Il<lk is trtlpl.lcd by II paIr of lIegatIVe io.,5 r~lllblcs 1\
ncgatl...., h.llogcn m..,k'Cult'- ion, wh ich j . CI. in Kel. No latt ice \'lICalldcs or C'\1r:l atoms I.I'C in-
\ 'Olv«! in • VI( renter. 'Ille <'1.Tltcr :II: the left of the figure probo.bly is I)(ll Slal,1c: 'he hl'Xagun
rcpre~nts II hole ITltI'PCd ncar a poIoil"~ ion \'aCalIl:)'. s1J(.h. amII" would be the antimorph to at>
F "'-'Ille r , lIoles ha,·c a iov.'Cr energ)' trappt'f'l In a VO( ccnter Ih&n in an anti-F cellt< r.

The anthnorph to the F ccnter is a hole trapped at a positivc ion V3('3IlCy,

but no such center has been identified experimentally in alkali halides; in
instllating oxides the 0 - (called V-l tlefcet is known. The best-known trapped-
hole center is the VI( center, Fig. 12. T11c VKccntcr is forme<.! when a hole is
trappal by a halogen ion in an alkali halide crystal. Electron spin resonance
shows that the cente r is like a negative halogen molecular ion, such as Cli in
KC!. 11le Jahn-Teller tropping of h 'C holes is the most effective form of self-
tropping of charge carriers in perfect LT),stals.

Problems
L ,.·rcnh11 defee'll. Show t llat thc number II of interstiti,li atoms in L'quilibrillm willi II
JaUice ."3ClIOcles in a crystal ha'ing;\! b it ice points and ,\ 1' possihle intclStlti,,\ posi-
tions i~ given by the cqlJatioll
f., .. 1.81' In I(N - II'KN' - nYnl \ ,

""lienee, for n « N, N', we ha" e 11 i!I (NN' )"! c~p( - £,J2k1!n. Ilere E, is the energy
necessary 10 r('move an atom from a lattice site to au inlentitial positkm.

2. Schottky WClincie•. Suppose that th e ell(' rgy lequi red to removc a sodi um atom
from the imide of a sodium crysl,1.I to the houn<l<uy is 1 eV. C!k1Jlate the oonccnlra·
lion of Schottky vaca ndcs at 300 K.
-.,
~-

3. ,.. unter. (a) Treat an F center as a fret:! clC('Imn of mass JU mo_ '" in the frcld of a
point c1mrgc e in a mediuUl of diek"Ctric constanl € - ,,2; ,o;Jltlt is the h -2' J energy
difTcrcII<.'C of f' centers in NaCi? (b) Compare from Table 3 the ,; center excitat iou
cn('rgy in NaCi with the & -3" ('lIers), difference of th e free sod ill m alom.

References
C. I'. Fl rnn. Pm", df'ic cls and diffusion. O~rOfd. 1972.
W. Il . Fowle r , ro .. P/upics of eclaT CCli ters, AC'adcmic !'ren. 1968.
N. B. lI annay. Solids/ale wmf~ll1J, l'Ie""Ill. 1973-1976.
F. A. Hu n/It'}' ...>d. , l..<'ll li~ dc/teCls ,n ."m;conducto .... 1n51il ulo;: or PiI)'Sics, London, 1975.
J. J Markham , F """,,, ... in alkali halide,. Academic I'rt'Sl, 1966.
A. S. Nowid, and J. J- BurtOJl, eds .• 'Diffupl,lf' if! loIitU- r~1I1 dcvdvlml(ml., Academic Press,
11m;.
K K. Rcb;",c, I ",pu"'" speclrn of ""id" 1'I~lu lU . u no.
J H. Sd,ulmall a nd W. D. Comp ton. Colv ~ «"'err i'l ~vlid.r. I'c.pmon. 1002.
A. M Stoneham. Theory &/ dq"Ck in svloJ~, o..rord. 1975.
W. II B)"eS alld A. M. Stoneham. fhf"Cf~ ,,00 dtll!d p~J#J In "cmmelallic Jolltb, Wile)', 11l65.
R Balian, M. Kkmall , and ,.-P. Poirin, eds., Plt"d" al dtff(lJ, t>:orth-liolland. 1981.
M Lannoo aoo }.lk>oJrgoin, P"'n' dtfl!d.s in IeUllronductOr.f. 2 ,00s., Sptillgl..... 1961-83.
Y. Fa~ aoo M . Fontana, Ekdronlc and I!ib""'ia.~l p~r'ks of point d4«fJ j .. janie cr'J'faU,
NQI"th-lIolla oo. 1m.
fkfocfS in wlidJ. r-. orth-lIo1land. A series.
Crystal wUice d .. ft.' ds. A journal.
 (

19
Surface and Interface Physics

HccomtructiOIl and relalWtion 5SS

SURFACE CRYSTALLOGRAPHY 55G

n cflection high -energy electron diffractiOll 500
SVRio'ACE ELECTRONIC STRUCfUl\E 500
Work function 560
Thermionic emission 561
Surface slates 562-
Tangential surface transport 5G4

MAC NETORESISTAf'IOCE I N A lWO-DlMENSJONAL CHANNEL S64

Integral QUlllltizcd Hall effect (IQIIE) 500
IQIIE in real sy~tcms S68
Fractional quantized Hall effect (FQIIE) 570
p-n JUNCTIONS 570
Rectification 572
Solar cells and photo\'ohaic dele<:lors 573
Schottky harrier 573
HETEROSTIWCTURES 574
noN hctcrojullction 576

SEMICONDUCTOR LASERS 577

LlCIIT-EMITTIl'iC DIODES 579

SCANNll'iG TUNNEL!r-;C MICROSCOPY 580

PROHLEMS .$81

1. Diffraction from II linear .!IITay and square 8ITay 581

2. SUlfacc subbllods in electric quantum limit 583
3. Shubnikov-dc Haas effect in a surface channel ~

REFElmNCES 584
 J.'igure 1 Dangling Lands from the ( III ) suoacr. of. co>"lently bolw.l~,,1 d,~mond cubic structure.
(Aner \I Prulton, Surfau "h!lslc.f. Clart:ndon, 197~.)

""
 CHAPTER 19: SUIU'-ACE Ai"lJ INTElWACE Pt-IlSICS

Reconstruction and Relaxation

The surface of a crystalline solid in vacuum is gcnerally defined as the few.
approximately thrce, outcrmost atomic layers of the solid that diffcr signifi-
cantly from the bulk. The surface may bc cntircly dc.U) or it llIay have fo reign
atomS deposite.:1 on il or incorporated in it. The bulk of tllC crystdl ls called the
substrate.
If the surface is clean the top layer may be cither reconstructed or, wille.
times, un rcconstructed. In unrcconstruclt..J surfaces the atomic arr..lI1gemcnt is
in rej:!:istry with that of the bulk exccpt for an inte.-lH),er spacinj:!: change (Cl.lllcd
multilayer relaxalion) at the top ~urface.
Tile shrinking of the int erla~'e r distance octwccn the first and second Iaycr
of atomS with respect to sllbscqlU2'IIt layers in tlJ(' bulk is a mtller dominant
phenomenon. The su rfacc may be thought of as an intermediate between thc
diatomic molecule and the bulk sllucture. Bt.>cause the illtcralOll,ic distal Ices in
diatomic molecules are milch \ 1I111ller than in the bulk, there is ,I rationale for
the surface relaxation. This may be contrastcd with reconstruction where the
relaxation of atoms yields new surface primitive cclls. In rclaxation the .ltoms
maintain their structure in the surface plane as it \\1IS according to the projec-
tion of the bulk cell on the surfacc; on ly their distance from the hulk ch;mgcs.
Experimcntal Vl.llues of the top-layer relaxation for lIlctals are given in
Table I. Note the frcqucncy of occurrcncc of contractions in the interla),cr
sp.1t::ing at the surface.
Sometimes in metals, but most oftcn in no nmetals, the atoms in the sur-
face layer form superst luctures in which the atoms in the laycr arc not in
registry with the atoms inl'Orrcsponding layers in the substrate. S"rf.1ce recon-
struction can oc a consequence of a rearrallgement of bro"en covalcnt or ionic
bonds at the surface. Under such conditions the atoms at the surface hUllch into
rowS with altcrnatcly Iargcr and smallcr spacings than in the hulk. That is, fo r
some crystab hc1d together by valence bonds, creation of;\ sUlf.lce ",ould Icave
unsaturated bonds dangling into space (Fig. l). The energy may then be low·
ered if neighboring atoms app roach each other and form bonds with their
othelwise unused valcnce electrons. Atomic displ.lcements can oc as large as
0.5 A.
Reconstruction does not necessarily re(juire formdtion of a Sllpcrstructure.
For examplc, on CaAs (110) surlaccs a Ca-As bond rotalion oc'Curs that leavcs
the point group int.lct. The drivin~ forcc is electron tmmfcr from Cd to As,
which fills the danglin~ bonds on As and depletes thcm on Ca.
Surfaces of planes nominally of high indices may be bUilt up of low illdex
 Tablc I Rcla'l.ation of topmos t inte rla)'cr sl)''lCing at u nrL'ClOnsh udcll dean mctal
s urfaces

Each (' nlry consists of:

Clll'm;C'.d S} mbol (in leriayer s pacing dL:lngc from bulk value).
[Aftl'l' f , Jona Olrnl J', M. Ma l'('llS , ill "The Slrol't"fe of SllliK~'" II ," ('(b. J. F. van der
VI!l'l' and M. A. Van lIove, Spri nge r-Vc rlag (J k illdlx'rg, 19S!:i), p. 001

hcp(OOO I) He (-5'1), Sc (-2%), 1'i (-2%), Zr (- 1%)

fcc{ lll ) ,\I (+ I<;t), Ag (0%), CIl (-0.7%), 1'1 (+ I 'k.), Ri. (0%)
1'1(.'«110) F(' (+0.5%), Na (0%), V (-0 .3%), W (0'1-)
fcc{ IOO) AI (O'k), Co (-1%), lUI (lY.i»
bcc{ I OO) Fe (-5%), Mo (-9.S%), Ta (- II %). V (-7%). W (-8%)
fcc{ llO) AI (-8.5'il). Ag (-8%), CII (-8.5<::t), r>;i (-8.5%), l' b (- 16'.1»,
RI. (-3<::t)
hcp(lQ..-iOJ Rc( - 17%)
1x.' «2 JJ ) Fe (- 10%), W (- 12%)
f<'C(3 JJ ) AI (- 13%), r>;j (- 16%), e ..
(-5%)
bcc(310) Fe (- Iti%)
fcc{33 I ) AI (- 1:2%)
fcc(210) AI (- 15.5%)
bcc{ 11I ) Fe (- 17%)

-
llCT(2 1O) Fe (-:2:2%)

planes separated by steps one (or two) atoms in height Such te rmcc-step ar-
r.mgements are importallt in evaporation and desorp tion because the attach -
ment en erg~ of atoms is often low at the steps amI at kinks in tl.e steps. The
che miC.ll acth'ity of such sites may be high. The presence of periodic arrays of
steps llIay be detected by double m.d triple beams of uilTraction ill LEED (sec
below) experime nts.

SURFA CE C IWSTA U DC n i\ I' I [Y

The 5\lrface structu re i~ ill general d iperiooic. This uot.'S lIot necessarily
mean tl'flt all its atom5 lie in a plane, but rather that the structure is periodic
only ill two u irncusiuns. 111c surface structure l"aU Ix: the structure of foreign
material deposited on the substrate or it can be the sd\"agc of the pure sub-
strate.
III Chapter I we used the term Bravais lattice for the array of equivalent
points in two or in three dimcnsiollS, that is, for uipcnooic or tripcriodic struc-
tures. In the physics of sunaces it is com mOl. to speak of a two-dimellsional
Inuicc, F\lfther. the area unit may be called a mesh.
We sllO\\'ccl in Fig. 1.9 four of the five nets possible for a dipcliodic struc--
 J9 SUrffIU and ' nlerface Ph",i« 5S7

ture; the fifth net b . Ie general oblique net, with no special symmetry relatioo
between the mesh basis vectors Ill> 8Z' Thus the five distinct nets are the
oblique, square, hexagonal, rectanb>uL'lr, and centered rectangular.
The substrate net parallel to the surface is used. as the reference net for the
description of the surface. For example, if the surface of a cubic substrate
crystal is the (I J 1) surface, the substrate net is hexagonal (Fig. 1.9b), and the
s11rfac:e net is referred to these axes.
The vectors Ca. C2 that dcfine the mesh of the surface structure may be
expressed in tenns of the reference net 81> 82 by u mntrix opcratioo P;

t;) ~ p(:) ~ (~:: ~::)(:). (I)

Provided tlmt the included angles ofth c two meshes are equal. the sh0l1-
hand notation due to E. A. Wood may be used. In this notation, which is widdy
ust.'tl, the relation of the mesh CI, ~ to the reference mesll 81> tlz is expressed as

(" ,,)
-X -
tl, 6z
flO', (2)

in terms of the lengths of the mesh basis vectors and the angle a of relative
rotation R of the two meshes. If 0' =: 0, the angle is omitted.. Examples of the
Wood notation are given in Fig. 2.
The reciprocal net vectors of the surface mesh may be writteIJ as ci, c:,
defiued by
(3)
Ilere the 211' (or 1) indicates that two conventions are in use. 11m definitions (3)
used in Fig. 3 may be compared with the definitions (2.10) and (2. 11 ) for the
reciprocal lattice vectors of a tripcriodic lattice.
The reciprocal net points of a diperiodic net may be thougllt of-when we
arc in three dimensions- as rods. The rods are infinite in extent and normruto
the surface plane, where they pass through the reciprocal net points. It may be
helpful to tllink of the rods as gcneratal by a triperiodic lattice which is cx-
pan<.k--d without limit along one of its axes. 111en the reciprocal lattice pOints
ruong this axis are moved closer together and in the limit fOlm a rod.
The usefulness of the ro<l concept comes Ollt with the EWdld sphere COIl-
stf1lction explained ill Fig. 2.8. D iffraction ocellrs everywhere the Ev.md
sphere intercepts a n."Ciprocal net rod. Each diITracted Ix:am is lalx:II(.-'(1 with the
indices hk of the reciprocal net vector

g =: hci+ kC: (4)

fonning the beam.
Low energy electron diffractiOll (LEED) is iUustrated by Fig. 4. The ek'C-
 -

[>( 1 x I) .,(2 x 2) (V3 x v3) 113IJ' I'l l x I) <~2 x 2) 1'(2 x 2)

r."<{ I I I), hq-.(OOO I) 1"«(100). 1)(~{ IOO)
1'1
'"

p(2 x I) .,(2 )( I) ,(2 x 2)

1>0.'.«110) fcc(1 10)
«I (,I)

Fi",'l.Ire 2 Surfa..-e IICIS of adsorbed aloms. TIle drde. rep,·csenl aloms in thc lop layer of the
s"l"trdle. I" (a) the dcsig"~lion fC<:{ lll ) mcan. the 1111) face of ~n fcc structure. This fat.-e deter-
mioe. a reference nct. ·.nc lilies rcprcsclI I onlcred o,·"rlarerJ, "ill.
ad~loms ~t the jnlcrscdion.' of
two li nes. "inc inlcrscelion poinlS rcprclCnt d iperiodi<' nets OaUicc~ in two dimCIlSlOns). TI.c desig-
nation p{ l X I) in (a) i. a primiti,"e mesh Ullil for "hid, thc basi. i. Klellt ical with the OOsi. of the
refcrellet: nel. III (b) tl,c.:(2 X 2) mesl, "nil is" ~-e" tered mesh with basi. ,·cdors twice as Io"g as
tll(l1;C of the reference lIet. Atomic adsorption on metals takes pl..ce most oncn into those su rface
. ites (i.ollow sites) Iha l maximize Ihe nllmber of nc.lTesl-lleighbor ~Ioms on the suhslrate . (Arl~'"
\'~ " lI u'·e.)

I,

-
 19 Sur/ace "tid lrJtO!./acc I'hytlk. ,

,!=:=i ,/
• f r-=-IJ'~/
,It,-------;1
00
0 ". ....-U
"0 o s"r£ott.
~M "",nl>

'I, : l: ",>
/~ / ,I • 0
• Os...... ,.,~
net p"",1!;
Lb- 1--Io S,/o
7/ bI J" 0'
".
I.'
I~• • • •
. " t,'
S,,"'tr.ole I II
n...... I b,: I I
•,,, °ll" ~~ I'
m
"0
, -'.; / ofl 1/ "•
1~-7--*
,,• ,, s,,~ocel "nM t",::,.,>h
lr.? • •

.
, 00
•
. •
@

M
•
"
Figure 3 A (3 x I) wrface structure, (Ill ICIlI-Spal"t', ..mi (b) n'd[lroCl'l.space dia,.;nlms. (After E. A.
WQO(!.)

,
,

---------
';,,, ......' ,rbn" .....

"'" (01 )

."
Figure" E......Kl ~pherc con~tructioo lOr illlrraction of incident W2'"C k by 1Il<luarc net. when k is
p3>'lIlId to !JIlt axu ofthe me.h. TIle back _ ttcrct.l bcam~ ill the pbne ohhe papt'r arc k;. k.;. k:s.
k; Dlfft4L1w ocams ootofthc plancoftbc (lII(lCr mil..!"" QU.'Ur, such as (12) 11..0 ( 12). The \"e!1K',.t
lines lire the rods of the recipnxal nd.

(
"'igure :> LE ED pIIUcm~ from a !'t{lll ) CI)'stal su~ lOr indtllcnt electron enel}:lcs of:>l and
""'w
63 5 "V, The diffnctioll angle is grt'ater at the l'fle'l;Y. (Aller C. A. SortK)ljai, Cllemislry;"
tom Jimr....io..r. rur/ace. .J

tron energy is typically in th e range 1O- 1<XXl cv. 'nlis is the arrangement with
which Davisson and Cenner in 1927 discovcrc:..od the wave nature or the elec-
trOll. An experimcntal pattern is shown in Fig. 5.
lkj1ection lligh-Energy Electron Diffn/clio,... In the RlTEED method a
beam of high-encl"b'Y cleclrollS is directed upon a crystal su rface at gra:tjng
incidencc. B)' adj ustment of the angle or incidencc one can arrange the nonnal
l;o lllpo nent or the incoming \\'3\'e\'t.'CI:or to be very small , which will minimize
the penetration of the e1ectroll b eam and enllancc the role of lIle clJ'stal Sll l"-
r,ICC.
'J1le mdius kor the E\\'3ld sphe re for JOO kc'V electrons will be - liP le i,
which is much longer than the shOJ1est reciprocallatticc vector 2ma - I A- I ,
It rollows that the Ewald sphere will be nearly a n.ll SIln..'lCC in the central
SCtlttc ring rcgion , Thc intercept of the rods or the rc:.."Ciprocal net wjth thc nearly
nat sphere will be nearly a Ijllc whcll the beam is dirc:..'CIed at g m.dng incidcnce.
The cxperimental arrangement is shown in Fig, 6.

SURFACE ELECfRO!l>IC STHucrURE

Work Function
111e work fllllction W of the unuOI'I)) sUlface or a metal is d efi ned as the
difference in potential e nergy of all ek'Ctron betwCt!n the vacuu m Ic\'c1 and the
Fcrmi le\"cl. 'nle \ 'acUlIIll level is the cnel1,'Y of an ck-'CIron at rest at a poin t
sufficielltly far outside the SUlface so that the electrostatic image rorce On the
 19 S ..rfoce olld Intcrfoce Ph'.lm. 561

<00,

."

'"
Figure 6 lhc RimEl) method. In (a) the higll t,ncrgy inciden t electron beam at a glancing wlgte
to tl,e L~m' $una<:e Is MSociated with an Ewald sphere ofbrgc radiu1, w Iargc tl"'t the surrace is
nearly flat in relation to the scpatlltion between ac.Ijacent l'Oll.ofth .. redpruc;o,l nt.' t. Tbeb-mation of
diffract;"n Iinc" OIl a 1'1..e .ereen is SlKM11 In (b). (Afler I'rllttpn.)

electron may be neg!(.'Clctl- more than JOO}. from the surfUce. 111e Fenni
level is the electroch e micnl potelltial of the electrons in the metal.
Typical values of electron y.urk functions arc given inlable 2. The orienta-
tioll of the expost'tl C1 ystal face affects the value of the work rimction became
0

the strength of the e lectric double layer at the surface depends 011 the concen-
tration of surface positive ion cores. The double layer exists because the sUlface
iOlls are in all asymmetrical e nvironment, with vacuum (or an adsorbed foreign
lItom layer) on olle side and the substrate on the other sideo
The work function is t.'tJua\ to the thr(.-shold energy for photoelectric emis-
sion at absolute 7.ero. If liw is the e nergy of a,l incide nt photon, tht.'Il the Ein-
stein equ.1.tion is liw = W + T, where T is the kinetic ellerb'Y of the emitted
electron and \V is the work functioll.

Thermionic Emission
The rate of emissiOiI of thennionic electrons depends exponentially on the
work function . The derivation follows .
We first filld the ekctroll concent ration in vacuum in equilibrium with
electrons in u metal at temperahlre T(=k,/11 and chemical potcntiallL- We treat
the dectrolls ill tbe vacuum as all ideal gas, so tbat theil' chemical potential is
JJ. = JJ..,.., + 'T log(n/nQ} , (5)

by 'I"P, Chapter 5. Here

'"
Table 2 Electron work fllnction sa

(Vailles OI)t;,lillcd by phOlOcmission, except tungsten obtained hy field emiSloioll.)

Element Sur/a."." plane Work funct ion, in c\'

A. (100)
( 110)
4.64
452
(Ill) 4.74
C, polycrystal 2. ].I
C" (lOO) 4.59
(110) 4.48
(JJ J) 4.98
Co (Ill ) 4.80
Ni (100) 5.2.2
(110) 5.£).1
(Ill ) 5.35
w (HJO) 4.63
(110) 525
(111 ) 4.47

- Aflcr II. O. llagltrum

IIQ = 21..111,,(2,",i'13/2 , (6)

for partic!l.'S of spin 112.
Now J4..~t - P, = 'V, by the definition of the work function W. Thus,
from (5),
11 = /lQ exp{ - \Vir) . (7)
'(lIe nux that leaves the metal smface when all electrons are dr.wJO off is
equal to the flllx incident on the surface from outside:
j" =inc = (7121T7Jl)1I2n (8)

by 'l'P(14.95) and (14. 121). Here c is the mean speed of the electrons in the
vaCl.llIm. The ek'Ctric charge flux is ~J" or
(9)
]11i5 is called the Richal'dson-Du.~hman equation for thermionic emission.

Surface Stoles
At the free slJrfa(.'e of a semiconductor there often exist !SurfaCt:-lx )llIld
electronic states with energies in the forbidden gap between the valence and
conduction h:mds of the bulk semiconductor. \Ve can obtain a good impression
of the nature of the surface states hy (.'{msidering the Wf\ve fun ctions ill the \\o~.\k

\
..
 19 Sur/on "".I Int('r/"« f'h~ic.

binuin!!: or two-componcn t appro\imat ion of Ch.lpter 7 , ill (Jill' d illlcmiun. (111e

wave fU llcliollS ill th ree dimcmiow; will have C\tl1l £.(:IO(s c .~ p [i{kuY + k~l;) 1 in
t he Ij,Z plane of the surf.K::c.)
Jfthe vacuum lies in the region x > 0, the potential energy of all clc 'ct.·on
in this region can be sct (.'f lual to zero:
U(x)= O, x>O (10)
In t he cl)Slal the potential enCfb')' has t be ...\u.l1 periodic fo rm:

U(x) = L Ur; exp(IC1) , x<o (Il)

G

In onc di mens ion G :::: ",mo, whe re II is a ny intcger, incl ud ing 7£ro.
In Ihe vaClHlln the wave fUllction of a ho und surfM.<C state must fall olr
expone nti •• II )':

1/1""1 = C'IX-SX) , l>O. (12)

B), the wave equation t he energy I)f t he state rt:fe rred to the vaCllum level is
r
E = - fi2 /2m . (13)
\Vilhin the c r),stal the two-(-'ompone nt wave fu uetioll e,f a bound 1>urlace
state will have the form, lo r x < 0,
I/Ih' "'" cxp(qx + ih)[C(k) + C(k - C)exp(-iCx)] , (14)
by analogy wilh (7.49), but witll the addition oflhe factor ('XI' (q .l) wh kh serves
10 b ind Ihe e k·c t ron to t he Surf.lC'C.
\ \le now come 10 an important consideration that rt'1>tncts t he allowed
values of the \V3.vevecto r k. If the st.lle is bound the re call be no Clin ent now ill
the x di rectio n, no rmal to the surface. 'nlIS c(lfl(lition is J.\Mlred in qU.ln tum
meCha nics if the wa\'e lilllctioll can he Wl·itlc n as [I rCol1 fum·Uon of x, a colJditio n
alread y satisfied by tile exterior ,\,lYe fu nction (12). But (1-1) clIn l.le a real
function o nly if k = ie , so tha t
I/I,n -= exp(q t )(C(iC)exp(iGxl2) + C(-iC)exp(-IGx/2)] . (15)
'n IlS is Teal p rovided C·OC) = C(-{C; ). '11111s k.. for a surface state d~ no t
ha\e a continuum of \illues, but is linu ted to disc re te ~tates associ;;lted wilh
Bli l10llin 7.Dne boundaries.
The state (15) is daml>ed e"pollc nti..'1 J\y in the crystal. 11le constants s, q a re
related by the condi tio n that I/Ialld t/r.{Jdx are COlllilnlOUS at x = O. 111c b illdi ng
c nefb'Y ~ is dete n n ined h y solving the tW<HXIlll pOlll'nt secular equation I [malo-
guus to (7.46). 'o,e plot of Fig. 7. 12 i ~ helpful in tllis COIlll(,"Ctk>n .
""

-----.,
p.#ntwl

---------------.,

---., ------.,
---
- - - - - - - _ _ _ _ _ _ _ _ r,"

(.,
.
,,)
Figure 7 n.Uld bcu.Jinl! 'w:;ll- a . emiconductor ~urfacc Ih" t ('QI' gi,·c Mhighly CQIldt>cting ~urflO<'e
It')l;io,, _(M) I",c r..iun L),<,r oo. 311 n·ty(l(' scmirnnduli.... For tl,e bcOOillgllS . he""" the hole woccn-
Ir~lion lit ti,,, ~urf.«::e i5 far t..l)(Cf than the ekdron ronttntlllt io,n In tile interior. (L) AccumulatIon
Llfcr on an II-type ...."miconduc;tor, witl. an electron n>nt"entn.tion Mt tile ,un...:-., that Is fa,· higher
than'" the IrIt.:rior_

TungcI,tial Surface Transport

\ Ve have seen that there may exist sllIfacc-boulld electronic states with
c llcrbrj(' S in the forbidden gap between the \'alcIlCt: and conduction bands of the
substrate crystal. 111c5c states may be cx:cupil.'(1or varon!; their existence mllst
arrect the .statistical mechanics of the problem. 'Ilt is means that the states mod-
ify the local equilibri um concentration of electrons and holt:s, as express(,>(! as a
sllift of the chemical potelltiHI relative to the band t:dges. Becausc the cllemical
potential is independent of position in an equilibrium system, the e ne rgy bands
mllst be d isplaf..w or bt:nt, as in Fig. 7.
The thkkness and ean i e r conccntra tioH in the SlIrf..\Ce layer may be
chang(.>(! by applying an e1('Ctric field normal 10 till.' surface. The elf(."(:t 01" an
extemru fl e ld is utilizt..--cl in the metal-oxide-semicond uctor fl cld-efft.'C1 trallsistor
( M~F':T). -n is has a metal electrode just olltsille the s(;micond llctor SIH-face and
insulated from it hy a layer of oxide (Fig. 8). A \'01t..1ge, the gate voltage VI;' is
applit.>(1 hetwt."Cli the metal and the bu lk scmiCOlllluclor and modulates tile
oonullctancc betwccn any other e lectrodes pL1Ct.·d in contact with the eit.<trolls
of the sUl-fare spa('"C charge. The physics of tht: surface conductance channel is
trcatt:d ill Problems 2: and 3.

MACNETORESISTANCE IN A TWO-D I ME~S I ONAL C IIM,' NEL

-nlC
static magnetocon{luctivity tensor in 3D was found in Problem 6.9.
Ilcre we translate that result to a 20 surface conductance chanllcl in tIle xI}
plane, with the stHtic magnetic field in the z direction , normal to the MOS layer.
\Ve assume" = ""/1.2 electrons per unit arca. Tilt: surface conductance is de-
Figure 8 A met;LI-mkk~..,mi<.'Qnductor trolnmWr. The gate \'Ohage V" i . applied between the
metal and Lhe ~emIcooKItK.tor, cau~ing all accumulation cl d lll');c carricn at Ihe omle interrace.
Thil; C!itablishes. l.....,..din""",ionaI oonductiyjty d1l\nnel in the plaone J lhe intcrf8C.'C, which Is the
xy plane.

fi ned as tht~ volu me COI1ductivity times the layer thickness. The surface current
d ensity is dcfill(..~ as the current crossing a line of ullit le ngth in the surfuee.
'I1ms, with (6.43) and (6.65), the surface tensor conductance COlllp0nt;nts
become

(J6)

where 0"0 '" 1Ifl--rl1ll and We = eBlmc in CCS and eB/1Il in SI. The follOWing
discussion is written in CCS only, except where ohms are used.
These results apply specificall y in the relaxation time approximation usl.'{1
in Chapter 6. when w.:'l"~ J, as for strong magnetic field and low temperd-
tures, the surfaCX! conducthi ty components appro.1 ch the limits
0"'1/ = wclB (J7)
The limit for u'"\I is a general property of free e lectrons in crossed electric
EI/ aud magnetic fields B z • We establish the resu lt that such electrons drifi in
the x direction with velocity Vv = cE,jB z. Consider the ek:ctrons from a
Lorent-I. frame that moves ill the x direction with this velocity. By electromag-
netic thl."'01"y there is in this frame an electric field E~ = - vv Bjc that will cancel
the applit:d field E" for the above choice of V ,). Viewed in the laboratory frame,
all electrollS drilt ill the 1" direction \\ith velocity VI) ill addition to allY velocity
components tllt..'Y had before E" was applil....1.
11111s j~ ... u~vEl/ ;:: vetil) "" (veclB)EI/' so that
u~1/ = vedB ( 18)

as in (17). 'Ill() cxperiments measure tllC voltage V in the y direction and the
current I in the x direction (Fig. 9). lIere I~ = J)"'JJ = (wdB ){£h) ;::
(vcclB )VI/. 111e Ilall resistancc is

PII = V,jl" "" Bl wc (ISa)

 \Ve see thatj" can flow with zero Ex, so that the effective conductance call
be infinite. Pamdoxically, this lim it occurs only when UJa and U w are zero.
Consider the tensor relations
(19)

rn the' Hall effect geometry j~ = 0, so that E~ = (ur/u~!)Ex, with ux~ = - u"X"

TIm,
j" = (u.... + d';.ju~~)Ex = o{efl)E" , (20)
and in the limit un = u~~ = 0 the effective conductance is infinite.

Integral Quantized HaU Effect (IQUE)

TIle results of the original measurements2 under quantum conditions of
temperature and magnetic field are shown ill Fig. 10. TIle results are remark-
able: at certain values of the gate voltage the voltage drop in the direction of
current flow goes essentially to zero, as if the effective conductance were infi-
nite. Further, there are plateaus of the Hall voltage near these same values of
gate voltage, and the values of the H all resistivity UlllIx at these platt:aus are
accurately equal to (25,8 131in teger) ohms, where 25.813 is the value of hlc
expressed in ohms.
TIle IQHE voltage minima V,,,, may Le explained on a mooel that is, how -
ever, oversim plified. Later we give a general theory. Apply a strong magnetic
field such that the separation liwc P kn'f. rt is meaningful to speak of Umd.;J.u
level.~ that are completely filled or comple tely empty. Let the electron surface
concentmtion (proportional to the gate voltage) Le adj usted to any of tile set of
values that cause the Fer mi level to fall at a Landau level: from (9.33) and
(9.34),
seBJhc = v , (21)

where s is any integer and v is the electron surface concentration.

When the above conditio ns are satisfied. the electron collision time is
greatly en hanced. No elastic collisions are possible from one state to anoth er
state in the same t.... . ndau level because all possible final states of equal energy
are occupied. The Pauli principle prohibits an clastic collision. Inelastic colli-
sions to a vacant Landau level are possible with the absorption of the necessary
energy from a phonon. but there are very few thermal phonons of energy
greater than the interlevel spacing by virtue of the assumption liwc P knT.
TIle quantization of the Hall resistance follows on oombining (18a) and (21):
PII = hlse'1. = 27r1sco: , (22)
2
where a is the fine structure constan t e /lic =0: 11137, and s is an integer.

·K. ' Oil Kht1.in g. G_ Dorda. and M. P"ppeT, Pb~'5 R~". Lett- 45, ~94 (1960).
 19 Surface and lnlerfac" Physic"

"
1 "
I"
- - f., ,I
J
Fi.gure 9 Applied fleld E. and drift current t. in a quantum Iiall elTcct (IQUE) e"penment.

" r-------------------------~--_,

" "

I,
,,
w

>,,
"
.;
" " " v",..

" "
,
,
" "
Fogurc 10 In the original lQl1t:: measu rements a magnetic field uf 180 kG (1ST) pomts out ur tl.e
paper. The tcmper.ltllre i~ 15K. A constan t .:urrent of I jJA IS furc<:d to flow between the source
and the drain Volt"&cs V"" a"d V" arC plotted vcuus tI.e g:tte >ultage Vgo whoch 's proportIonal to
the Fermi lcw\. {Afiet K \on Khtzing, G Dorda, and M. Peppe r }
'" (
lQUE in Relll Systems
TIle measurements (Fig. 10) suggest that the above theory ofthe IQHE is
100 good. -nle HaU resistivit)' is a(.'Cumleiy quantized at 25,8J3I$ ohms.
whethe r or not the semiconductor is of very high purity and perfection. TIle
sharp Landau levels (Fig. Ib.) are broadened in the feal crystal (Fig. llb). but
this doe.~ not affect the Hall resistivity. The occurrence of plateaus in the Hall
resist.'lnce, evident in the UII curve of F ig. to, is nol expected in idenl systems
because pal'tially filled Landau levels will exist for all gate voltages except those
for which the Fenni level exactly coincides with a Landau level. Yet the experi-
ments show that a range of V" values gives the exact Hall re5istance.
Laughlin 3 interpreted the results for real systems as the expression of the
general principle of gauge invarianC'e. TIle argument is subtle and some\\hat
reminiscent of the flux quanti7,,,ltion in a superconductor in Chapter 12.
In Laughlin's tJlought-experime nt the 2D electron system is bent to form a
cylinder (Fig. 12) whose surface is pierced everywhere by a strong magnetic
field 8 normal to the surface. The current 1 (fon ner IJ circles the loop. TIle
magnetic field 8 nets on the charge carrie rs to produce a Hall voltnge VJI (for-
mer VII) perpendiculnr to the current nnd to 8-that is, VII is developed be-
tween one edge of the cylinde r and the other.
The circulating current I is accompanied by a small magnetiC flux Ip that
threads the current loop. The aim of tJle thought-experiment is to find the
relation between I and VII' We start with the electromagnetic relnlion that
relntes 1 to the total energy U of a resistanceless system:
au
- =-V,1 = - _
I alp
at ~' c at '. (23)

The value of 1 can now be found fro m the variation l>U of the electronic energy
thaI nccompanies a small variation Sip of the flux.
TIle carrier states divide into , ....'0 classes:

• Loculil'.£(1 stales, which are not continuous around the loop.

• Extended states, continuous around the loop.

Locali7.ed and extended states cannot coexist at the same energy, according to
our present understarJding of iocalinttion .
The two classes of states respond differently to the application of the r.p. nux
TIle localized states are unaffected to fi rs t order because they do not enclose
any Significant part of Ip. To a localized state n change in Ip looks like n gauge
transformation, which cannot affeet the energy of the state.

3n a La ughlin, Ph)'$ I\c\'. Il 23. 5632 (1981). Ke abo I"s art icle In the r.kGrnw-1li1/ vear-
book 0/ 'cu:'OICC and technology, 19&1. PI). 200-2 14. A rcvW:w i.s gi\'en 0) II. L . Stl)Tm1'1' and D . C.
T sul, Sc'enc-e 220, 1241 ( 1963).
 19 Surf_ lind Interfoce Ph~

,~
j
". ". ". ...
I.,
"' lgun! II Density of states in II 2D ekctron gas in II strong magnet ic field. (a) Idcal2D crystal .
(b) Real 20 tT)'~tal. with imp"rities and impelf,x:tions.

, .
F;,;un! 12 Ceomctry for Laughlin's tho"ght...,al)Criment . The 2D electron syste m iJ "'nIIJped
around to form a cylinder. A strong ITIlIOgI"ICtic foeld R P>Crl:es theC)'linder C'o't'TY"'h ere IKlI"mal to its
surfa~ . A ';UfTC"t I circles the looJ). giving rise to tl,c Hall voltage VII and .. small magnetic nu" <p
through the Ioo!)

11le exte nded states enclose Ip, and their energy may be changed. How-
ever, if the magnetic flu x is varied by a flux quantum, &p = " de, all extended
orbits are identical to tllOse before the nux qunntu m was added. TIle argument
here is identical to that for the flux quallti7,ation in the superconducting ring
treated in Chapter 12, but with the 2e of the Cooper pair replaced by e.
If the Fermi level fall s within t11e localized states of Fig. lIb, all exten[ied
states (~ndau levels) below the Fermi level will be filled with electrons both
bt.-fore and after the fl ux change &jJ. However, during the change an integral
number of states, generally one per Landau le\'el, enter the cylinder at one
edge and leave it at the opposite edge.
The number must be integral because the system is physically i.dentical
hefore and after the nux change. If the tmnsferre d state is tmnsferrt.'tI while
occupied by one elect ron, it contributes an energy change eVil . if N occupied
states are transferred, the energy c1mnge is NeVil '
'111is electron transfer is the onl), way the degenerate 2D electron system
 Q

'"
can change its enert-.'Y. We can understand the elTect by looking at a model
system \\~thout disorder in the umdau gauge for the vector potential :
A = - Byi . (24)

An increase M that corresponds to the flux increase &p is equivalent to a

displacement of an extended state by MIB in the !I direction. By the Stokes
theorem and the definition of the vector potential we have 6", = L,.&\. Thus liq;
causes a motion of the entire ekctron gas in the !I direction.
By /jU = NeVil and {jrp = Ilde, we have
(25)

so that the Hall resistance is

Pu = VI/II = IIfNc (26)
as in (22).
Fraclkmol Quan'ized llall Effect (FQllE). A quantized Hall effect has
been reporterl for similar syslt:ms at fmctional values of the index s, by working
at 1000~'er temperatures and higher magnetic field s. In the extreme quantum
limit the lowc~t Landau level is only partially occupied, and the integral QHE
tre.1ted above should not occur. It has been obscr\'ed. ~ however. that the Hall
resistance PJI is quantized in ullits of 3hJe2 when the o(:cupation of the lowest
l .mdau level is 1/3 and 213. and Pn vanishes for these occupations. Similar
breaks have been reported for occupations of 215. 3/5. 415, and 217.

,,-n JUr-;CfIONS
A p-n junction is made from a single crystal modifled in two separate
regions. Acceptor imlllUity atoms Me incorporated into one part to produce the
1) region in which tile majority carrier.; are holes. Donor impurity atoms in the
other part produce the n region in which the majority C<lrriers are electrons.
TIle interface region may be less than 10--1 cm thick. Away from the junction
region on the 11 side there are (-) ionized acceptor impurity atoms and an equal
concentration of free holes. On the n side there are (+) ioni7.ed donor atoms
am) an equal concentration of free electrons. Thus the majority carrier.; are
holes on the p side and electrons on the n side. Fig. 13.
Holes concentrated on the " side "'{JUld like to diffuse to fill the crystal
uniformly. Electrons would like to diffuse from the 11 side. But diffusion will
upset the local electrical neutrality of the system.

'D. C. n"i.ll. L. Storm('r. and A. C. Gouanl, I'h)$ . Rm.. Lett 48.1562(1982). A M. Chang
~Iai, Ph)"s. Rc\.. Letl. 53. 997 (1984). For a di..,unKm oIll1e theory loCI! R Laughlin In C. Bauer el
a1. eds .. Thv-dnnen$ionai system'. het~rostf1jct"rtI , a"d , upulaU/cr..3, Springer. 1964.
 19 Surfoce and rn" 'r!uu ""!ls;';a 5.

(.J

figure 13 (3) Variation or thc hole amI cJe..1ron ronccntrutions ;:.. - ;'+'--_ ___ _
across a/\ nnhonsed (;,:em applied voltallCl junction. The carriers al'e l - +
in thennal C<lu,librium with the llCI;cptor and donor impurity ~ ...
atoms. so th.lt the product JI/\ of the hole and electron conccnt ra- 8. , __ ____--=+,--____
tlons i$ run.tan t throughoutlhc e')'$ta] ill co/\ror mity \<;th thc law ~ r t
or mass action, (h) ElectrostatIc potcnti.,) rrom .. cceptor (-) and ] +

"'"-----~
d",.or (+ ) ;"IIS n<-'(Or the IU/\<-1;'",_ The potential gradient inh lb.ts
tI lfii'Sioll or hules from the p SIde to tI.e " side, W,,] It in"ibiU - +
+
d,fiusion or dect m... rmm the" side to thc p sid e. 11,c electric field
in the J",lCIion regiun is called the built -in electriC field, .J
A small ciwrge transfer by diffusion leaves behind on the p side an excess
I.
of (- ) Ionized acceptors and on the side an excess of (+) ionized donors. This
charge douhle layer creates an electric field directed fro m n to p tllat inhibits
diffusion and thereby maint<lins the separation of the two carrie r types. Becau~e
of this double l:lyer tile e1edrostatic potential in the crystal takes a jump ill
passing through the regio n of the j unction.
In thcrmru eq uilibrill m the chemical potential of each carrier type is every-
where constant in the crystal. even across the junction. For holes
kuT In p(r) + eq{r) = constant , (27a)
where p is the hole concentration and if! the e lectrostatic potential. 11ms p is
low where if! is high. For electrons
k8T In lI(r) - erp(r) = constant (27b)
and n will be low where if! is low.
TIle total c1lemical potential is constant across the cryst.ll. The effect of the
concentration gradient exactly cancels the ei€(:trostatic poten tial , and the net
particle flow of each carrier type is 7~ro. However, even in thermal eqUilibrium
there is a small flow of electrons from n to 11 where the electrons end their lives
by 1-e<"Qlllhination with holes. The recombination current Jnr is balanced by 11
current Jllg of electrons which at-e generated thermally in the p region and
which are pu.~hed hy the built-in field to the 11 region. TIllIS in zero cxtel-nal
applied electric lIeld
(28)
572

for othe rwise electrons would accumulate indefinitely on om.. ~Jde of the har_
rier.

Rectification
A r -n junction can act as a rectifier. A large current will flow irwe apply a
voltage across the junction in one di rection, but if the voltage is in the opposite
direction only a very small current will flow . If an altemating voltage is applied
across the j unction the current will now chiefly in one direction - the junetion
has rectified the CUrrent (Fig. 14).
For hack volmge bias a negative voltage is applied to the p region and a
positive voltage to the n region, thereby increasing the potential d ifference
between the two regions. Now prncticaJly no electrons can climb the potential
euergy hill rrom the low side of the barrier to the high side. The recombination
current is re(luceO by the Boltzmann metor:
(29)
-nle Boltzmnnn t,etor controls tIle OIlmber of electrons \,,;th enough energy to
get over the barrier.
The thermal generation cu rrent or electrons is not particularly affected by
the back voltage because the generation electrons flow downhill (rrom p to n)
anyway:
(30)
We saw in (28) that 1..,.(0) = - 1..,1,0); thus the generation current dominates the
recombination current for a back bias.
\vllen a fonvard voltage is applied, the recombination current increases
because the po tential energy La.Tie r is lowered, thereLy enabling more elec-
trons to flow from the " side to the lJ side:
(3J)
Again the generation current is unchanged:
I""V ro~.nl) = 1",,0) (32)
111e hole current flowing across the junction behaves similarl y to the elec-
tron current. The applied voltage which lowers the height of the Larner for
electrons also lo"'ers it for holes, SO th..,t large numbers of electrons flow from
the II region under the same voltage conditions that produce large hole currents
in the opposite direction.
11le electric currents or holes and electrons are additive, so thnt the total
fonvard electric current is
(33)
where 1~ is the Sum of the two generation currents. This equation is well s..,tis-
 J9 S"rfan Imd J.. /u!oa Fhysic. S13

(
.00
OE.Jri"",nl<ol
• - "->
,

•
_ ........, .... dt~ ....

••

.~/
vV •

Figt>~ 14 &dification charad.,.;,ti<;mall-.. junction .00•

'-'1'-
.. .., "
in germanium. afier Shod;Iey. The "oItage ~ plotted
"ertialn), ami the eurrent horiwutlllly_
." Curr."t In ,mlliampok..,,1
'00

fled for p." junctions in germanium (Fig. 14), hUI not quite as well in other
semicomluctors.

Solar Cells and p/lOtOlJ(Jltaic Detea()J7j

Let us shine light on a P-lI junction, one without an external bias voltage.
Each absorbed photon creates an electron and a hole. When these eaoiers
diffuse to the junction, the built-in electric field of the junction separates them
at the energy barrier. The separnlion of the carriers produces a forward voltage
across the barrier: forward , be(."ausc the e lectric field of the photoexcited
carriers is opposite 10 the built-in field of the junction .
The appearance of a forward voltage across an illuminated junction is
called the photovoltnic effect. An illuminated junction can deliver j)O\\'er to an
extenml circuit. Large area P-lI junctions of silicon are used to convert solar
photons to electrical energy.

Schottky Barrier
When a semiconductor is brought into contact with a metal, there is
formed in the semiconductor a barrier Inyer from which eharge cnrriers are
severely depleted. The barrier laye r is nlsa calletl a depletion lnyer or t<xhaus-
tion layer.
In Fig. 15 an ,t-type semiconductor is brought into contact with a metal.
111e Fermi levels are coincident after the transfer of electrons to the conduction
band of the metal. Posilively charged donor ions are left behind in this region
,7<

Figu~ 15 Re«iI)'ing barrier bcho.'effi • metal and a n-t)"pc ~mico..dllctol The Fcnni b·cI is
sh<.:M'n 'IIa broken lioe.

that is practically stripped of electrons. Here the Poisson e(Juation is

(CGS) div 0 = 4mle (SI), <Ii. D ".,a", , (34)

where II is the donor concentration. The electrostatic potential is determined

by

(ees) tflfp/~ = -47Tne/E (35)

which has a solution of the fonn

(CCS) If = -(21T1leIE)~ SI (36)

11u;~ origin of x has heen t:lken for convenience at the L'i ght-haml edge of
the l)alTier. I11t~ contact is at -x", and here the potential ent::rgy ,'elative to the
right-hand side is -elPo. whence Ihe thickness of the harrier is

(ees) x" = {ElfAll/2r.ne)112 (37)

\Vitll E - 16; equ = 0.5 eV; PI = WIG em -3, we find Xb = 0.3 ~m. This is a
somewhat simplified vie\Y of the met.ll-semiconductor contact.

HETEROSTRVcrURES

Semiconductor hetenlStructures (lfe layers of two or more different semi-

conductors grown coherently with one common crystal structure. Hetcrostruc-
tures offer extra degrees of rreedom ill the design or semiconductor junction
dC\'ices, because both the impurity doping and the conduction and valence
band offsets at the junctiOfl can be controlled. This &eedom is the basis of the
prediction that most de\ices that utili7.e compound semiconductors \\il1 in the

-
", --,

t'igure 16 -fhr<.'C t~'I)C$ of ""'nd edge of\i;cts ~t hctcro-interfJ«!S. n,e fo,bidden I!llP~ are slim""
shaded. The uIT$<.'t c.-..J led nom,a/ OC'C\Irs, ror e><atnl'le, '" CaA..J(AI,Ca)A" TI,e " brokcn-g:lp" olTs"t
O.X.'IJr$ ill the GaSbIlnAs I'ctcrojllnchon.

future incorporate heterostruchlres. \Ve tre-'ll them here in ordf'f to keep ahead
of the times.
A heterostmctufe may be viewed as a single crYl.taJ in which the OCCll-
pancy of tile atomic sites changes at the interface. As an example one side of the
interface can be Ce and the other side CaA~: both lattice constants are 5.65 A..
One side has the diamond structure and the other side the cubic zi llc sulfide
stmcture. Both stmctm-cs are built up from tetrahedral covalent lJonds and fit
togethe r coherently as if they were a l>ingle crystal. 'J11ere are a few edge dislo-
cations (Chapter 20) to relieve the strain energy near the inter£'lce.
The band gaps, however, are difft'fent, and this difference is the source of
the re'".ll intere~t in the heterostructure. apart from the techniC'.\1 virtuosity in
forming the structure, The band gaps are 0.67 eV fo r Ge and 1.43 eV for GaAs,
at 300 K. The I'dative alignment of the conduction and valence baud edges
offers several pos:.iuilities, as shown in Fig, 16.
CalculatiomS suggest that the top of the valence band Ev in Ce should lie
about 0.42 eV higher than in CaAs. '111e bottom of the conduction band E" in
Ce should lie about 0.35 eV lower than in G.lI\s, so that the offsets are classified
as normal in the scheme of F ig. 16.
Band edge offsets act as potential balTiers in opposite senses on e lectrons
.md holes. Recall th,'lt electrons lower their e lle rgy by" sinking" on an energy
band dia~r-Hn. whcn~as holes lower their energy hy no.'lting on the same dia-
gram, For the nOl'mal alignment both electrons and holes are pushed by the
barrie r from the wide-gap to the narrow-gap side of the Ileterostructme.
Other important semiconductor pairs used in heterostl1Jctures are AlAs!
CaAs, rnAslGaSb, CaP/Si. and ZnSeiCaAs. Good lattice matching in the range

~St.'CTable 10,1 in W. A. Ilanison, Eiecllorric &lnldure allli/he pro,"'rlie~ v/solid&, Fre.:.maOl.

1900. See alw II. Krt)Cmcr, '''1111'''!)' orhctcrosln lct"~$' a cnlic..1 rc,-,ew," Eric.. !)ummcr ScI,OOI,

"','
 - +
___,r-----rt..
"
}' I. - - - - -

" --11
• __
-----------~--------
"
}' 1-

Figure J7 (a) Two semiconductors not in <:Onlact; II.., ~ute band edge energies arc labeled £.
for the conduction band e<1ge and £~ tOt" the valence band edge. An Rabsolute cncrgyR IJIC'aI1:5
referred to Infmitc distance. The Fermi 1c.~1s in the two materials are determined by the donor
too
('OIlrentr&lions.... well u by bOln.J ~tructure. (h) 'Ibc $3mc ~miconductors as a he lerojunctioo,
so that the two parU; are in diffusi\'c equiHbt-ium. 'Olis requires that the Fe rmi le>'Cl (F. L ) be
indepen<lenl of pmit;oo, which is ~mplishcd by transfer of electrons from the N'$il1c to the
n-slde or lhe inlcrf_. A depl.etion layer ofj:Uilh'Cly ioni7.e<i dooOTs is left behind on the N-side.

0.1- 1.0 percent is often accomplished by usc of alloys of different clements,

which may also adjust energy btaps to meet specific d evice needs.

n-N llel.erojtmdion
As a practical example. consider two n.type semiconductors with a large
offset of the two conduction bands, as sketched in Fig. 173 for a semiconductor
pair with a nonnal band line-up. The "-type material Witll the higher conduc-
tion band edge is labeled with a capital leiter as N-type. and the junction shown
is called an n-N junction. The electron tl"3llsport properties across the junction
arc similar to those across a Schottky barrier. Far from the inte rf.ICC the two
sem iconductors must he electriC'.Illy neutral in composition. However, the two
Fermi levels, each determined by the doping, must coincide if there is to he
zero net electron tnmsport in the absencc of an extema1 bias voltage.
'J11cse two oollsidel"3tions fix the "far-off" conduction Land edge encrgies
rdative to the Fenni levcl. as in Fig. 17b. The combination of a specified band
off.set (detennined by the host material composition) at tile interfucc and the
distant band energies (determined by the Fermi level) can be reconciled only if
the bands bcnd ncar the interface, as in the fib'Ure . The neccssary band bending
is created by space charges consequent to the transfer of electrons from the
N-side to the lower n-side. This transfer Icaves behind on tile N-sidc a positive
donor space charge layer, which through the Poisson equation of electrostatics
is the .source of the positive secolld derivative (upward curvature) in the COIl-
duction Land edge energy on that side.
On the ,,-side there is now a negative space charge because of the exccss of
dectrons 011 that side. 'Ifle layer of negative space charge gives a negative
 19 SUr/DCt: Dnd 'uler/OCf' Ph",ica 577

second derivative (t., .. nward curvature) in the conduction band edge energy.
On the l1-side the band as a whole bends down toward the junction. This differs
from the usual p-n junction. The dowmv.\rd bending and tile potential step
foml a potential well for electrons. The weII is the hasis fo r the new physical
phenomena chal1lCteristic of heterostnlcture physics.
If the doping on the f)-side Oow Eel is reduced to a negligible value. there
will be very few ionized donors 011 that side ill the electron-rich layer. The
mobility of these electronS is largely limited only by lattice scattering, which
fulls off sharply as the temperature is lowered . Low temperature mobilities as
high as 2 x 10& cm2y- ' s- ' have heell observed in CaAsI(Al.Ca)As.
If now the thickness of the N-side semioonductor is reduced below the
depletion layer thickness on that side. the N material will be entirely depic ted
of its low mooility electrons. All of the electrical conduction parallel to the
interface will be ron-ied by the high-mobility electrons on the f)-s ide, equal in
number to the numher of iollized N-side donors, but spatially separated from
them by the potential step. Such high mohility structures play a large role in
sol id state studics of 2D electron gases and also in ncw classes of high speed
field effect tl1lnsistors for computer applications at low temperatures.

SEf>IICO ND UCfQR LASERS

Stimulated emissioll of radiatiOn can occur in direct-gap semiconductors

from the radiation emitted when electrons recombine with holes. The electron
and hole conccntrations created hy illumination are larger than their equilib-
rium concelltmtions. 'n lc recombination times for the excess carriers are much
longer than the times for the conduction electrons to reach thermal equilibrium
with eadl other in the conduction band, and for the holes to reach the rmal
Cfjuilibrium witl l each other in the valencc hand . This steady state condition for
the electron and hole populatiolls is described by separate Fermi levels p.., and
JJ.v for the two bands, called quasi-Fermi levels.
Witll IL.: alld p-.. refen·ed to their h3lld edges. tile condition for population
inversion is Illat
P-c > JJ.v + Efl • (38)
For laser action the quasi-Fermi levels must be separated by more thall the
boU\d gap.
Population inversion and laser action can be achieved by fOf\yard bias of an
ordillary eaAs or Inl' junction, but almost all practical illjection lasers employ
tlle double heterostructure proposed by II . Krocmer (Fig. 18). 'Jere thc·lasing
semiconductor is embedded betwccn two \Yider-gap semiconductor regions of
opposite doping. An example is CaA:; emhcd.ded in (A1.Ca)As. In such a struc-
ture there is a potential harrier that prevents the oulllow of electrons to the
57.

~:kctron Ilow

--- \:-c-

./If'''v
'"
-r-
- - - - - - - ; 0 -,
__1
'---,"
t- 1161~ flow •

\.' I
FiJ;1lre 18 Double hctcrosln'cturc injection 1........, •. Elccho'lli flow from the right into the optirolly-
active !a}-cr, where they form aucgl'nelOlte clcclro" gas. TIle potentiallllllTier provided b) thcwlde
energy gap on the 11 side prevent. the eJectmlls fTorn cSClIpiog 10 the left. Holes now fro." the Icrt
ililo the active laycr, bllt canllQl escape to tltt' right .

p-type region, 311d an opposite potelltial barrier that prevellts the outflow of
holes to the TI-type region.
The value of J.I-c in the optically active layer lines up with Jln in the u
C()Iltact; similarly, J.tu lines up with P-p ill the p contact. Inversion C'dll be
achieved if we apply a bias voltage larger than the voltage equivalent of tlle
active layer energy gap. The diode wafer provides its own electromagnetic
C'dvity, for the reflectivity at the crystal-air interface is high. CI),stals are usually
poli~hed to provide two nat parallel surfaces; the radiation is emitted in the
plane of the heterojunction.s.
Crystals with direct band gap.s are required Ilonnally for junction lasers.
Indirect gaps involve phonons as well as photons; carriers recombine less effi-
ciently because of competing processes, and no laser action has been observed
in illdirect brap semiconductors.
Gallium arsenide has been widt:ly studied as the optically active layer. [t
emits in the !lear infmred at 8383 A or 1.48 eV; the exact wavelength depends
on temperature. The gap is direct (Chapter 8). III a heterojunction the .system is
very efficient: the nltio of light energy output to dc electricu\ energy input is
near 50 percent, and the differential efficiency for small changes i.s up to
90 percent.
The wavelength can be adjusted over a wide range in the alloy s~ stem

-
 19 S"rfore Qn,l I",u/au P"yW:.

v
II
(a)

Ec 0

__
"'-0 0 0 0

~, '~~=I }v
Ec
t<
w
~ '"
~
•
"
~
E"
OOOO~
L
(h)

"igure 19 Electron-hole rccom(,indlion into phololU, oICroSS a 11 n JUliction.

Ca..Inl _. I\,As1_II, so that we call match thc laser wavelcngth to the <lbsorption
minimum of optical fibers used as a tnlnsmission medium. '111C combination of
double heterostructure lasers with glass fibers forms the basis of the ncw
lightw<we communication technology thu! is gradually replacing tl1\n~mission of
signals over copper lilies.

UCHT-EMJ"ITI"'C DIODES

The effiCiency of light-emitting diodes is now at the pOint of cxcec<lillg

incandescent lamps. Consider a p-rl junction with a voltage source V splitting
the two chemical potcntials JLn and JLp by eV. as in Figure 19. Electrons from
the" side are injected into the p side. and holes from the p side are injected
illto thc " side. These injected carriers annihilate each other across the junc-
tion, thus gcnerdting photons if the quantum eRkiency is unity.
The gencration or recombination process will be much stronger in a direct-
gap semiconductor (Figure 8.5a) than in an indirect gap semiconductor (Figure
B.Sb). In a direct gap semiconductor such as CaAs, the ban_ .0·banO photons
are absorbed in a distance ... I ,",m, which is stroug absorption . lhe direct gap
tcmary semiconductor CaAs 1 _ . p. givcs light tuned to shorter waveleugtlls as
the composition variable x is iucrease<l. This oompos.itiou was made by
J lolonyak into one of the first INI diode lasers and into the first visible-spec..
trum (red) LE D. B1ue·emitting heterostrlleturcs have now bccn made, such as
IniCal_ .. N - ..\JIIGal _IIN.
The performance of LED's has increased markedly over tile years, from
about O. t lumens/watt in 1962 to about 40 lumens/watt in 1994, compared \\1th
15 lumens/watt for a standard white unfiltered incandescent lamp. To quotc
Craford and Holollyak, "We arc cutering an entirely new era in lighting (illumi-
nation) with an ultimate furm oflamp-a direct gap [I[·V alloy p-tl heterostruc-
ture."

SCANflill':C TUNNELII':C MICROSCOP\'

'J1le original and still the ccntral application of scannhlg tunne ling micros-
copy is to the ohservatioll of the atom ic structure of crystal surfuces. STM is a
dream method that sounds much too simple to permit observutions on tile
atomic scale for wbich it was dcvcloped by Binnig and Rohrer. -C1le develop-
ment was an extraordinary accomplishment. for wllich tllCY received the 1986
Nobel Prizc for Physics. Au account of the history is given in their Kobel
lecture for that year-accompanied by a quotation from the almost traditional
letter of rejection of major work from a scientific publication to which they
submitted an early paper.
One takes a sharp metal ti p, prcferahly a tip tim! has Olle atom that pro-
trudes a bit more than all tim other atoms of the tip, and brings it to within 4 A
to 8 A of a crystal surface. A voltage applied between the tip and the surface
rouses a tunneling current of electrons to flow . By tunneling we mean the
standard quantum tunneling of a particle through a potential barrier. 'Inc trans-
mission of electrons through a harrier is exponentially sensitive to the width of
the barrier. The transmission flux in the experimental situation varies by a
factor of aoout 10 for a l-A. change in the separation ofthc tip from the surface.
'I1lis fact is the key to how such a simple device can have a very fine spatial
resolution. The greatest part of the tunnel current flows from tile surfaec into
the oulermo~t atom of the tip, thus giving resolution 011 an atomic scale.
The most common way to operdte an STM is to apply a voltuge Lctwecll
the tip and the surface. and to adjust tllc height of the tip continuously to keep
the tunnel current constant as the tip is scanned across tile surface. The signal
actually rcrordcd is proportional to the voltage applied to a piezo-ceramic crys-
tal whose displacement moves the tip perpendicular to tile slJIface. -nle theory
shows that for small hias voltage the trdjcctory of the tip measured in this way
Figure 20 A "qllau lum conal"' of mean radius 71 A was forlll.:J by mo.'lng 48 Fe atorm. On a ell
(Ill) ~urfacc, 'n'e I'e atOt Il~ ICIIlte r surface stale electrons, oonflning them to tile interior of the
conal, 1be rings in the corml arc the demil)' di!.hibntion of the electron. ill ti,e Ihn..'1l quan lunl
sl~tt:S of the (.'Orral Ihlll He don' to Ihe Fenn; l"nc,!;,oy. The alom~ were imag.:t1 and rno\~l inlo
~tiol1 by a low lemperature . IIltn.·high \'aClIUm lCIJInillg tunndlng microscope. Phoco court"')'
ol D. M. Eigler, IBM Re~eareh D ivision.

corresponds to a contour of constant local dcnsity of states in the surface. taken

at the Fermi lcvel. What you are seeing is what the STM "fccls"-a map of how
th e conduction electrons in the surface extend out iuto thc vacuum dIal sepa-
rates the surface from the tip.
Witl, STM olle call resolve single atoms, as in F igurc 1.27, and surface
comlgatiollS as small as one thousandth of an atomic diameter. 'I1te tip can be
used to move atoms around 011 the surface. strikingly shown h ere ill Figurc ZO.

Problems
I. DijJradif)11 frcnn (I lil1f!tlr array and 11 square arroy, TI,e diffractiOl) patWI11 of a

linear structure ofJattk'C COnShillt a is explained!! in Fig. 21. Somewhat similar struc-
tures arc important in molecular biology: DNA and many proteins arc linear helices.
(a) A cylindrical film is exposed to the dif'&actioll pattenl of Fig. 21 b; the axis ci the
crlinder is l'Oincidellt \\;th the lUis ci tl,e liuear structuTC or IIbCT. Describe the

- Another \iewpoint i. lISf'r"I ' rOO' a h"f._ lattICe t he d,lfrtlClion pBtt<,m is described by the
single Uolle equation B' 61. .. 21:'(/. "here q h an inleg.". 11}e 1,lIl ice ." no~ which led 10 the olher
La"e equationS do not occu,' for a Imear lathce. Now a . 6L: .. <'Onst . is tile (.'qu4Ilion or a plane, thu,
the rcciproc;ollatt icc I,,'cornes a 'iel of paralld planes norl1\;)\ to the lUlU of alOLlU
 -

, •..-'_ _ ~ 10. e - I,..ch

d4k",1K"<'t - NA
r.,.. "", ...lfUctJ...
",~ .... t-nee

,0) ,,'
Fig,,", 21 1hc dllTracliorl polle.n from il Single linc ofl .. ttkc CQllst.n' II in a monocbrov,atic ". noy
be;lm perpendicular to the lille. <al 1'be CUlKhtioo for oorutroclh'c inlcrfcn:ntt is Il CO$ 6 - "A,
where .. is an integer. (b) For J.(i''(ln R the dilTl1Id ..-d fa)'S of constant A lie On the IUM-.: of a CUfll'.

\
-~ . ,
• •
•
,.'•
•
, ,
j "
't).
..,
• • ;~

"
'.,
...
~
';~' ..
, ~Aa'

.'
,
I,
'. ·
,~

•
.. •

Figure 22 (a} Backwanl scattering pallem 0176 cV electrons incident normally On !heO 10) r_of
• nide! crystal. a model of the surface is , I,own in <bJ- (Cou rtesy of A. U. Mac Hae )

-
 appe-drancc of tile .liffl<K1ion p.lttelll On the 111111, (h) 0\ 1I.lt pl,utO!-'l"II,I,il' pl.,tc L~
placed behind the nllI.'r :md 1101I1ldi to tile mcld~nt h....11I1 . Skdch roul-!hI) tl)(' .LI>~'.IJ ­
alice of the dlffractioll pallcrn On Ihc 1II.Ite , (r J ,\ \111glc l,lallc of lU0111~ fom .s .1 ~' III .lrC
lattlCc of' I,\ttice con~t;Lnt a, 'nlC planc is IlOnl1:11 to the il\c"k'nt \· r. I~' Ix:~m . Skell!.
roughly the apllI.'arallCC of th.~ diffr:u,:tion pnttel1l on Ihe I'hUh')!;r.lpll1~ plat..•. I liul;
' n le difTr.lClio.l fro1l1 .1 planc of atoms c.m be iuferred f l'Hlll the paU"'fl11 «Ir hlO
,Jcr[X'Il<liculnr lines of aloms. (tI) Figurc 22 shows the "'!octrnn ..hn'r.lellon pJit<'''" III
Ihe backward dirl'Clion from Ihe nickd atOlll~ 011 the ( 110) ~urr.\Ce of ,I lllckel l'r.~t,11
E'lIlaill the orientatioll of the ddli-adillll p"litenl in n·l..toull to the .ltOI11I'" pO~iI101I\ or
the ~urfaCt': lltoms shown in the modcL ASMlmc Ikll onl), the hU Ir.~ .ltOI11S ,III.' efTct'"
tile ill the refk'('tion of low-energy dectuJIIs,

2. Surface ~uhbanll.. in decide qltaJllwn lim;I, Cnll~il lt.. r the ront,lt't plane bchlcen an
insulator and 01 semicomlul'lor, as in a lII~t.lI- uXIt!l'· ~t'I"icI;>Il(IIll'I()r tr,msi\tor or
\10!<0fl:T. With 01 ~trollg clcc;tric field ,\llpliw oI(:nllo~ .he SLOr Si mtc.f.ICC. the 1I0lclltiai
enc rgy ,,r a ftlll.llIctlnn e!t:.:tron may I...• approxiluatcd IJ\' \'(x) = cr.,,: klr I ' posil"e
aud by \ 1.1) ... ~ fOr ~ ll1.'~\ijve . where th e ongiu ,I' x I~ nt tl,c j1lk. oo· 'n,t, Will e-
function is 0 forl' neg.J.tive and may be scp.lrntcd.\5 I/(x.y ,:::) '" lI(l) c\Jl[I(kvJ + k~)I .
"here II(X) salbflcs the dilTerenli.d ('<I"alio>,

With the nKXkI potential for V(x) the exact cigt·nrunl·110n~ Me A'r)' flmcti",,~. hnl
we Coin find .1 f.'lirly good ground slale CIICI'l,'Y from thc '.tIi.IIIIIII,1I trIal fUnditll'
l' C'IJ(-IIX), (...) Show thaI (t') = (/i!I23I1}ff + 3ef~I , (b) Show tlml thc I'llcr)!;I' is d
1IIiuimuni wben a "" (3eflll'21i~1!;I. (e) show Illat (t'}'"'6 - l ,b!J(Ii f21J1)lrl r3l' f /2)l.3 In
'
Iheexact solull.,,\ for the ground ~ tal e energy tll~ f.1l·tor I .b\) I~ repldCcd hr \.7lS. ,\ 5
E is 1I\creascd lIte exlent of the \V'Jvefunction in thc r dir ..'etion h (k'Creased , 'n,C'
functionll(r) defines ,\ sllIf~'C cotlliu('lion cIt,lnnc1 <>11 tlLC ~"',"ico'"!l1ct(>r ~ide of the
illierfucc, '111e VJnOllS eigell\·allles of u(x) define wk.t .lie e"lIed dt'Clric ~ul,ban(ls ,
or
flccause the ('igenfl1ndi()n~ arc real functions r the ~t"tes dn II"t ( ',11 Iy ..'urn'llt in
thc l'dirt'C:tion, but th<.'Y do cal 1)' a surf,we lh,lIme! ('IIII"'llt il1 the !/,::. pl,ule. TIle
or
depend..'nce the channel on the elcetric field E ill the l ' di rf'(,tioll m~kes the dCI ICC
a field efT..'Ct transislor.

3. S/lIlbllik<ro-tle lIaca 41ect in a s",.face channel. 11iC clf(:t"I de5t,:rihcs th..• l)Crin<iie
variatioo wilh l IB (o r II ilil clectmn concentralion) of the 10nl,;illl(II";l.1 dl't'tri<;.tl rt'sis·
tance ofa specimen in a strong IJC'l)endieular magnctlC Ik ld at J.low tellllll.'r.llure.
These are the conditions in Wlliell 1I de 11a.u·lnn Alplwll clT<.'CI is ol,sen .....J, .I~ di~­
euSS<.'tl in Cllalltcr 9, ,11C effect Coin lit' disclI'>SCd )impi) Itl 20. (,,1 Show 1)"lt ti,e
dCII~il)' of' free cll'Ctron mllilab in 20 is .,af2w1it , with omlS$M:m of ~pi" ,tnd \aU~y
<1egc.lernC)', 11te surface C(I'lceutration or electrons COI.1! now 00 IHillen ,IS ,, _
mf~~II. (b) s how that the S-dH o.scillatiOlu hn\'e thc IlI.'riod 6(U8l - elihllCfF -
d ZrmcII. In SI sct G"" I. (c) ACCI.lfdillg to the viewofthc in\en:ioclIJ)(' r ,IS a pl:um
t"ap;cilor lhe surface charge conrenlratiOIl "I'aries directly as the WIle lollagc ".e
across the interface. SllOlV tilat thc rt'$istancc osclll~titJns arc ]XTiOOit. Junctions of tile
~te loltnge. as in Fig. 23,
 I

'"
figure 2.3 SOlln':C ...l'>li" resislance of an ,,·cI,anne! MOSI'ET as a functin" of the ~Ie ,'oltagc V~. lit
Imos,'ersc magnetic rocills of 33.3 and 3'i-7 kG , 31 4 K. (Aftcr C . Landwehr clal. )

References
It Van.-.:Iow, cd., Clwmlslry Otld physics of solid surjaC/?s. Springer, 1979-.
C . A Somorjai lind M. A. Van Ho\"e, Ad..orl,cd mo"o/ayers On solid surfatXs. Springer, 1919.
T. N. 11...,Jln and C . Ertl, eds., Natll.e of tl,e 1IlnfoCf! chemical bond, North Ilolland , 1979.
1>1. A. Va" 1I0\'e and S. Y 1b,,8'. Surface cryslallogmpl>lj by LEED, Springer, 1979.
A. A. Maradutlill, II . F. WalH s, "lid L. D<.>b,....lynski, cW., J/a"dbook of surfaces and Interfaces.
Carland STM Press, 1!l8O.
M. Prtlllon, Surface physics, 2nd cd_ . O,,(on1, 1963. Good int roduct ion.
C. A. Somorjai, Chemislry in tlW dinl£""Ums. surfaces. Camell, 1961. Fine experimental intro-
duct,un , with cmpha.~is On II(horptiun of moloculcs OJ) surfaces.
1: Andu, A. 13. Fowler, ami F. Stem, "Electronic pro(l(l rtlesoftwo-dimensional ')"lterm,··I\t.'\'.
Motl. Phy~. S.l, 437 - 672 (H182),
II. C. CaSC) and M. B. Panish. 1I~'fcHJ&l mct"ff lasers. Academ ic, 1978.
II. K. lI eni$CI" Semicrmduclor cnnlacl8, Oxfo ...l, 19&4.
Surface Scicnce. A Jo urnal.
A. z...u~,"·iII , PIIY"ics III rorfaces, Cambridge, 1988.
1. " I. Blakely, ed, Swface pll~;"" ofmaledl!/s , 2 ,'015.. AcademiC, 1975.
II. E. Prange lind S. M. Ci",·in. eds, Qualllll'" 1/ill/ effect, 2nd cd., Springer 1990.
M. Lannoo and P. Friedel, Atomic and C!ecllVlllc $Irucf urf of ~'urfact8, Springer, 1991 .
S. C. Da"lson and J. ]). Levine, "S"rf,ICC states.~ Solid slate physics 25, 1 (1970).
T Chakraborty and P. PidnCII, Frac/iona! quantum lIilll effect. Sp ringer, 19S8.
M. C. Craford alld F M. Stcranb., "Llgh t·emittiug diodes," Encyclope<lw of applied phySiCS 8 ,
485- 5 14 (loo.t).
C Julian Chen. ]"/rrlducWm to SCl"!lUlillg '",,,,el,,,!; microscopy. O~fol'(l , 1993.
 20
Dislocations

SHEAR STREr-;CTH OF SINGLE CR\'STALS 561

Slip 568

DISLOCATIONS 589
Burgers vectors 592
Stress fields of dislOClltioM 593
Low-lingle grain boumlarics 595
Dislocation densities 598
Dislocation multiplication and slip 1300
STRENCTII OF ALLOYS 600
DlSLOCATlO,,"S AND CRYS'Ii\L CROWnl GOO
Wiliskcrs G().j

HARDNESS OF MATERIALS 600

PROm,EMS 606
I. Linc~ of dosest packing 606
2. Dislocation pMirs GOG
3. Foroo on di~lQ('ation 606
REFERENCES 001
 •

t-.-j

G--- &, G-
, G- G-
0:0:0 ~
0 0
,,, ,
: I
''
,, '' ,,,
I I I
' ,I
t ' I I
b I ' I I
. I I I

~r--------------'
<--\'--":---------Co-~--C--
:.
,!
ou...........,.... ~
"

Figu.-e J (a) Rc:blh e ~ht= " rtWQ p ....1eS of 3toms (d.<7'''''' In section) in. umformly ~tr;\incd CT)slal;
(bl ,he". ~I ress as a fUflchon of Ih" rdat(\e lhsplacell'lCnt of the .1IanO's from tl\C.r eqUlhbrlum
l'O'itlnn l1>C ""a~y broJ..c". Ime <1'''''11 .<1 the ini h~1 ~Iopc d..finl'5 the shear mOOulu. C

\
 CHAPTER 20; DlSLOCATlONS

l11is chapter is concerned with the iuterprctrttiou of the pb~tic mech,lIlical

properties of cl;.strtlline solid ~ iu terms of the theory of dislocati()n~ . Plastic
properties rtre irreversible deform,\tious; elrtstie properties rtre reH'IMhlc. lllC
ease with whieh pllre siugJc eryst<lls dcforn] plaslicalJ)' is striking. 111i~ illt .. in~ie
weakness of cryst,lis is exhibited in vrtrions W<I)'S . I>lll e ... iI"er chloride mt·lts ,11
455"C , )'et nt room tcmperatmc it has <I ehecsdikc con... islcncy and c.m be
rolled inlo shccts. Pure rtlnminum cryst'lls rtre elastic (follow IInokc's law) only
to rt strain of about 10- 5 , aft('r whieh they ddiJrm plastic.IIl)'.
Thcon~tical c\timates nf the clastic Iimil of perfect er),slals ,l.\iw vallie ... liP
or 10.1 higher thrtll the lowe... t observCf.1 \'alllcs , althollgh a frtctor lIT is more
usual. There arc few exceplions to the rulc that pure cly\tals an.' pla ... tic ,Ull] <Ire
nol stTOug: cl)'st:lls of germrtnil1l11 aud silicon are not pltl~tic al rOOm tempt: w -
ture rtnd fail or yield onl)' b)' fracture. Class <II rotJI1] templ'l,llure f,lils only b)'
fnlclure, but it is not cryst.tlline. The fraetllTt' of gl 'l"'~ is caused b)' ... trcss COll-
centration at minute cracks.

SHEAR SnmNCTII OF SINCLE Cnl"STALS

Frenkel g.1Ve a simple method of ('stim.lting the theoretical \hcar ... trength
of a perfect cr),stal. We CQusider ill Fig. I the force needed to IlhlJ.,e ,j ~hear
displaeemeut of two plaues of ato lll~ past c,lclL other. For slll ..1I eh~tic strains
the stress u is related 10 the displ.iccmenl .{ hy
u = Grid (I)
Here d is the iulcrplrtmu· spacing, and G dcuoks thc approprirtle sht: n modu-
Ills. When the displ,ll'Cmenl is large (Iud has proccCf.lcd to Ihe point thai atom A
is directly over atom B in the figure , the two planes of alOlllS rue ill a l"Ollfigl1frt-
lion of unstable equilibrillm and Ihe ~tress is :l£ro. As a first llpproximalion we
represent the stress-displacement reJ:ltion by
if = (GaIZ7TCl) sin (21Tx/a) (2)

where a is the interrttomic sp.."\cing in the direction of ~hcar. This rdJ.tion is

constructed to reducc to (I) for small \lrtilles of x/a. l11c critical sill'at" stress u~ ,It
which the lattice bewmes unstable is ghcn by the masinlllill value of cr, or
(3)

If a "" tl, Ihcn UC ' " G/21T: Ihe ideal critical shear slrc...s is of the ordcr of! of
the shear modulus_

5-!l7
 (
The observations in Table I shO\.\.' the experimental values of the elastic
limit are much smaller than (3) would suggest. The thcoretic-.t! estimate may be
impro\oo by Cl)nsidcratioll of the actual form of the inle rmolccllbr forces and
by cOllsidc r:'ltioTl of other configurations of Illcc!J:lIlk'al stability 3vaibble 10 the
lattice as it is sheared. Mackenzie has shown that these two e ffects IlKI)' reduce
the theoretical ideal she:n- strength to :'Ibou! CIaO, (.'Orrcsponding to a critical
shc:w s tmill :lIlglc of about 2 degrees. 111C observed low \'ailles of the shear
strength can be cxpbincd only by the presence of impclfcd iolls that call act as
sources of mcchanical weakness in real crystals. The 1ll0VCIllCIlt of cT)'stai im-
perfcctions called dislocatiOIlS is responsible fOl' slip :'It vcry low applied
stresscs.

Table I Comparison of shear modulus and cI:u:tic limit

Shear 'nooulu, C. E lasliC limit u c'

ill <l>'nIcm t ill d>'nIC"I"a C/u,

SII , Jingle crystal 1.9 X l Oll 1.3 x 107 IS.000

Ag. single crystal 2.8 X 1011 6X lo" 45,000
AI , sillgle cryst.!.I 2.S x 10" 4x 10" 60,000
AI, pure, polycrystal 2.5 X 1011 2.6 x lo5 900
AI, commercial drawn - 2.S X 1011 x 105
Durlliumill
fe , sort, polyel)'Stai
- 2.5 X 1011
7.7 X JO"
9 .9
3.6
1.S
x 1(f
x Idl
'"'"
70
500
lIeat-trclltt.'tl <:arholl steel - 8 X 1011 6.5 x l dl 120
Nickel-d,rome sleel - 8 X 10" 1.2 x to'O 65

Arter I\lntl.

Slip
pb stic dcformation in crystals OOOl rs b)' slip, 8 11 cxam ple of whidl is shown
in Fig. 2. In slip one part of Ihe crystal slidcs as a nnit across an adjacent )1., rl.
Ttl(' su rface On which slip takes place i.~ kIlO\\1. as thc slip plane. 11.e direction
ofmOliOI1 is known as the slip direction. '111e great importance oflattice proper-
tics for pbstk: strain is indicated by the highl y an isotropic nature of slip. Dis-
pl:\<:"Clllent takes pbce along crystallographic planes with a set of small Mille r
indices. slldl :'IS the {Ill} pbncs in fcc metals :lIld the {1I0}, {1l2}, and {l23}
planes in Ix-c melals.
111e slip dirc<"tion is in the line of closest atomic packing, (110) in fcc
metals and (III) in bee metals. To maintain the crystal stmctnrc after -,lip, the
displacemcnt or slip vector must c(l"al a lattice translation vector. The shortcst
lattice translation \·cctor expressed ill tcrms of the lattice constant a in a fcc
structure is of tile fonn (lli2)(x + 5); in a Ix-c structllrc it is (aI2)(~ + y + i) . nllt
 t'jg"rc 2 Tmmlational slip in 7.inc single cryslal,. (E. R. PArker.)

ill fcc crystals one also observes p.,rtial dislllaCCtllCnts which upset the regular
scqueuce ABCABC ... of dosest-packed planes, to produce a slacking fawt
such as ABCABABC. ... The result is then a mixture of fcc and hcp stacking.
Defonm'ltion b~· slip is inhomogeneolls. Large shear displaccments occm
on a few widely separated slip planes. while parts of the crystal lyillg between
slip planes remain cssentially undcformed. A property of slip is the Schmid law
of the critical shear stress: slip takes place along a given slip plane and direction
whcn the L'Orresponding component of shear stress reaches the (;ritical V'dlue.
Slip is OTle mode of plastic ddonnation. Another modc, twinning, is ob-
servcd particularl}' ill hql alld bee strllcluTCS. During slip a considcrable d is-
placement occurs on a few widel}' sep.11<lted slip planes. During twinning a
partial displacemcnt OCCII I'S SlIcccssi\'c1y Oil each of many neighboring crystallo-
graphic planes. Arter twinning, the defonned part of the crystal is a mirror
image of the undefollned part. Although both slip and twinning are caused by
the motion of dislocations, we shall be concerned primarily with slip.

DlSU>CAllQr-;S

'Inc low obscrved \'alues of the el'iticai shear stress are explained in tenl1S
of the motion through the lattice of a line imperfection known as a disIIX":l.tion .
"Inc idea that slip propagates by the motion of dislocations was pllblished ill
1934 indepcndently by Ta)'lor, OTO\V:l.n , and Polanyi; the conccpt of disloca-
tions was introduct,"(1 somcv.hat earlier by Prandtl aud Dehlinger. There are
sc\'eml basic t}l>CS or disloc":l.tions. We f'i rst desclibe :m edge dislocation. Fig-
ure 3 shows a simple albic crystal in whidl slip of one atom distancc has 0c-
curred over the left half of the slip pbne but not o\'er the light half. 111e
 -
".
Slip
d•.-..cI ion

Figure 3 An edge dw.x-:.ticm EF in the gliOc plane ABCD. TIll: ngtlre ~ th., s!ippecl region
ABf;J: ill which the atoms h.lVe been Ui>~ b) man: tI,~ n haIfa Lince comtau! and the un·
JhpjX'd rcl!ioo n ;CD with wspl:tl'emcnl less than h:Ilr ;ll "'llice OOIlslanl.

boundary between the ~lippcd and ullslippcd regions b culled the dislocation .
Its po~ition is Ill.lrked by the ICnllination of an ('>;Ira vcrtical half-plane of atoms
crowuccl into the upper Ihllf of the crystal as sho-.vn in FiK. 4. NCM the dislocd-
lion the cryst.,! is highly strained. 111c simple edge dislocation cKtencls imlefi-
nit(1), ill the slip plane in a direction normal to the slip dil'cctioll . In Fig. 5 we
show a photograph of a dislocation in a two ..<!imcnsiOllal 50:11) bubble raft ob-
tained by the method of Bragg and Nyc.
The Illcchallism responsible for the mobility of a dislOcation is shown in
F ig . 6. The Illotion of an L"tige disloattioll through a crys!.,1 is :l11.I1og01l5 to the
passage of a ruck or wrinkle across a rug: the ruck movcs more easily than the
\\ hole rug. If atoms 011 one side of the slip plane are mo\'ed with respect to
those on the othe r side, atoms at the slip pl:llle wili expeliencc repulsi\'e forces
from some neighbors and attractive forces from others across the slip plane.
These forl.-es cancel to a first approximation . 111C extemal stress rC<juired to
move a dislocation has been ca10llatcd and is quite small, probably bck)w
105 dyn/cm 2 , provided th:lt the bonding fOrres ill the crystal arc 1I0t highly
directional. 'l1ms dislocations may make a <.'f),stal \'ery plastic. Pass:tge of a
disl~tion through a crystal is equivalent to a slip <lisplaccment of olle pall of
the crysta1.
'l1le second simple type of dislOC'8tion is the scrC\v dislocation. sketched in
Figs. 7 and 8. A screw dislocation llIarks the boundary between slipped :md
unslipped parts of the crystal. 111e bo.mdary lKlralfels the slip direction . instead
of I} ing perpelldiollar to it as for the edge (lislocalion. The screw dislocation
may be thought of as produced b)' clltting the crystal partway through with a
 20 D;./0<'6f;.""

Figure... Stmclure of an edge J .. Ioc'.lion.

TIle ddonn.tion m:ow he thought of a~ ruu,;cd
by inserting .n extm rL'nc of .tOlns on the "
Up[l(.... half of the !I i1~is, Atoms in Ihe upper
11I"f..cry~lal Me romllreued by the in>ertion;
",', <'
those in the lower h.lf lire cd"",kd. "

Figure 5 A d,.location In a two.dimen~ional bllbbl" ,.,.£t. TI,e d,ilocatMI j~ "IO~t ".l'i'h' ken I"
turning the page by JO" III Its pla"e lind sighlill~ at ~ low IlIII<it'. CW. .\1. ulUltr, Jfttr n'd~ ."d
N}'".)

Figure 6 Motjon of a dldoo::a-

hon under II shear tend"'g to
IIIO\'C the upprr .• urfJ~ of the
~peClmen to the rIght
(D. lI ull)
"
D C

, / , , , , , ,, , ,
,; ,,, ,• ;,I ,
'c
...~; /'
,• ,,;
; ; ;, ;
D
/,
"I , • I •, , ,• ,
,,
,,
I

..I
,
,
•
.' ;gure 7 A screw dislocation. A pari ABE,.. of tilt) slip plme has ~I;l'ped III the direction P'lr:l!]cI to
the d;~lncat ion I'lle f;F. A sCn'W dldocation ma~' be vis",1lired a$ II hclirnl anangclTll.'TIt oC bUice
1'1 Mlf's, such that we OW.),'" plalle1' Oil ,.;01111; (''Uffiplclcly 8mund the disloc:llion li ne . (Arter Coc-
tre ll. )

Fij.,'Ure II Anol hcr view nfa ,;crew disloation. TIle brok...'1'

\crticallin.. IhM marks the lhsloc:uion is ",ITOU,,,.!..,,I by
stn;ncd mat ...'1i:1,I .

kllifc :lnt! ~h cOl rillg it par:'t llcl to the edge of the cut by one atol11 spacing. A
screw di~'O<."rllioll tr:msfo.-ms suc(;cssi\"c atom phlllcs into the surface of:\ helix;
th is accounts for Ihl! name of the dislocation.

Bllrgers Veelors
Other disl ~ tio ll forms lila)' be constructed from segments of edge and
scre-.v dislocations. Burgers has .~llOwn that the Illost gcnerd! form of a linear
disi()C:'ltion pattcfIl ill a errs!:!1 call be described as shown in Fig. 9. We consider
any closed curvc not nccessarily planar within a crystal, or an open curve te nHi-
Ilating on the surfa(."C at both cnds: (a) t-. lake a ellt along an)' simple swfa(."C
bounded by the line. (b) Disll\aCC the material on one side of this surfi\(.'c by a
vector b re lath'c to the othcr side ~ here b is called the Burgcl"S vC('tor. (c) In
regions ,,1lCrc b is not parallel to the cut surface. this relative dispL1ccment will
either produce a gap Or (",dllX€! the two hah·cs to o\'crlap. In these rnscs we
imagine tll:lt wc eithe r :\dd Ill<lteri:'ll to flll thc gap or suhtract Illatcri:'ll to pre-
vent ovcrklp. (d) Hejoill thc mate rial on both sides. We leave the straill dis-
placement intact at the time of rewcldillg, but llfkrwards we allow the medium
 (

Figure 9 General mcthn<.l orraml;n!; '" di,location ring In a meoium. n,e medium ,s rcprescntl.11
by the n:ct""h"'l~r block. The ling is repren"nh:d by the dus...'<l curve in the interior in the block. A
C\ll Is m:\(I.., along the ~ur&<'(! bounded by the curve and indicated by the contoured area. 'l1lc
m.~tcr\"l 00 one )ioc or Ihccul is displac-ed rdath'Ollo that on the other h), \'<'dordistaJ>C'e b, whICh
may be orient..d .,bilrarilr relalh'C to the surfiocr_ Fora.'S "'ill be required to dTt:ct the displace-
ment . TIlC medium is fill ..'<1 in 0>' ('ttl awa) so 115 to be t"Ol""'U<!US lifter the di~-plllCCment . II is Ih Nl
join..-d in the displ.. CtXi ~I~tc aml til" "ppli..-u fOlws arc rda, ..-d. I lele b is the Burgen vector "fthe
di.location. (After Seitz.)

to come to illtcrn~1 equ ilibrium . The rcsliiting s trllin p:lttem is th::.! of the
dis\(l(:ation ch:'lractcrized jointly by the boundary Cl1r\'C aud the Burgers vC(;tor.
'Ibe Burgers \'cctor Illust be cqu:'ll to:\ lattice vcctor il) order that the rewelding
process will maintain the cr)'stallinit~' of the material. The Burgcrs vcctor of II
screw dislOCltion (Figs. 7 :md 8) is parallel to the dislocation line; that of an
edgc dislocation (figs. 3 aud 4) is perpendicular to the dislOCltion line and lies
in the slip plane.

SlreSl Ficllh of DislocatiOlls

The stress field of a screw dislocation is particularly simple. Figure 10
shows a shell of material snrrounding an axial screw dislocation. The shell of
circumference 27fr has been sheared by an amount b to give a shear strain
e "" iN27Tr. The corresponding shear stress in an elastic con tinuum is
0' = Cc = GiJl27Tr . (4)
This expression ('annot hold in the region immediately around the dislocation
line, as the strains here arc too large for continuum or linear elasl idty theory to
app1)'. 111e clastic energ)' of the shell is ell':. = iGe 2 elV = (G1J2' -I7f) dr/r pCI'
unit length. TIle total elastic energy per Hllit length of a screw disl~tion is
found on integration to be

(5)
'"
"
b

figure to Shell of CL'lSli~,.]J)' mslortrd t'rY~tal surroond"'g

1oC":W di)lnc.-ation wIth nnrgcn Vl'CIor b; Ke al)() Fig. 16.

where R ;md ' 0 aTe appropri<ltc upper ,md lowe r limits for the variable r . A
rca.rolla!,lc vallie of TOis comp.1r:lblc to th e magnilm]c bof thc Dmgc rs \'ccto r nr
10 the 1,\ttil..'C constant ; the "allic of H cannot exceed the dimensions of the
cr}'~t"l. 111C ,,,llle uf the r.l.tio /UfO is not vcr)' important because it enters in a
logari thm term.
\Vc now show the (onn of the t' llcrg~' of an edge dislocntion. Let 0' .... .tn(l
U(JB denote the tellsile stresses in the radial and circumfcrc nti .• i direcliollS, and

let UrfJ denote the shear stress. In .111 hot Top ic e lastic CQnt ill l ll llll , at r .I nd U6fI arc
proportional to (sin (j)/r. we need a fUnction that falls off as lIr and that changes
sign when y is replaced by - yo 'J11C shc.lr strcss u ,e is proportionrll to (cos (f)/r:
collsidering the plrille y = 0 we scc fro m Fig. 4 Ihrlt the shcrtr stress is an (Xlcl
function of x. Thc COllst:mts of proportiollrllit)' ill the stress are proportiollrll to
the shc:... moo\lll1s G and to thc Burgers \'ct:tor h ofthc displrlccmcnl. 11le final
result , which is derived in books citro in the rcferences, is
Gb sin (J Gb (,' 0$ (J
U = u8(I = - Uri) = (6)
..... 211(1 - v) r 211(1 - v) r
where the Pois~oll ratio v - 0 3 for most crystdls. The strdin energy of a unit
length of eoge dl.\localion is

(7)

We want:m expreSSion for the shear stress component uxvon planes PaId!·
lei to the slip plane in Fig. 4. From thc stress componcnts Ur r • Uoe. dml (T~
c\'aluated 011 the plane a distilllce y Joove the slip piallC, we find
cb
(8)
(T,v "" 2n( 1 v)
 ,

Figure II (0) I-","~.,,)(lc grain boo",brv, aner Ilu!,):cn (lI) Flcdmn mlN'l>W~l'h "I ~ 1000\ -a,,)Ik
1;1'1"11 Ooundiry in mol) Ix\o;;num. The tlm..'t! disloC:lIion. 'n the h'IlI!!" I'ad. h ... c II ,,, .".'nC 1I" ~lu
,'«'tor Il5 in the dr.,wlIlg III Figure II". lhc wh.te (,,,,,,Ie. "' ...·k , .... poi',tiQf,s uf atom'" cui""",.
110.11\:1110 the plane of the paper Eaclo 3rr~)' of " "dc ~ dcline s the IJai .h..." of .tdi~IOC'atl"" , ",th rou.
circles on Ihe top ri each IImI)' and !l're<! {"inlcs "dow. ('Io>ure failu re i~ indout ed loy the 3n",.'
"Inch Jeflne the llulgcn ' "ffi Ors. (Court,,,,)' of It CIOIl,k, )

It is shown ill Problem 3 th:'lt the force ('.H1secl b)' a resolved uniform shear
stress u is f' "" ba per Ilnit length of dislocation. 'l1,c force that .," ct!ge clisloc\ l-
tioll :'It the origin exerts upon a similar Olle at the locat ion (y. 0) is
Cbl s iu ·JO
F = buXIJ == 217'( 1 v) 4y (9)

per IInit length. He re F is the compollcnt of fort-'C in the slip dilCltioll .

LoUHIligle Crain Boundaries

Burgers suggested tlldt low-.lIIgle boundaries between adjoining er) stal-
lites or crystal grains COnsist of arrays of dislocations. A simple e"alllple of the
Burgers mOOcI ofa grain boundary is shown in Fig. 11 . TIl(' boul1d.lry oc'ellpies
a (010) plane ill a simple cubic latllcc and dividt:s two parts oCthe C1\·stal that
have" (OOI J axis in (.'01ll1l10n. Such a boundary is called a pure tilt boundary: the
misorientation (.':\n be dcscribed by a small rotation 0 ;,bout the common [001]
axis of Ol1e part or the crystal relative to the other The tilt boundar)" is repre-
sented as an array of edge (hslocations of .~ I)adllg D ::z blS. \\here b is the
Burgers ' ·cetor of the dislocations. E'l:p,·, illlenb ha\ e subsl.mtiated this modd .
FiguTC 12 shows the <listT/bulion of ulslocdtiollS along sIIMII-.lIIglc ~rnin hounJa-
I·ics. as observc\ ith an electron microscope. Flllllu~r , ne.ld ,lnd Sholklcy
Figure 12 Elcc:tron micrograph of dis\ocalion ltfU("l ulU in b~angle lVIIin bou"dan.:-s in an Al-
7 JlC'rccnt Mg solid solution . NI)tict: the lines of smal l dots on Ibe right. M~. x 17,000. {R. Good-
r~h and G. TIlomas_l

(Icrivcd a thcory of the interfacial energy as a function of the angle of tilt . with
rc~u lts in excellent agreement with measurements.
Direct verification of the Burgers model is pro\'ided by Ihe quantitative
x-ray and optical studies of low-angle boundaries in germanium crystals hy
Vogel and co-workers. By counting etch pits lllong the inte rsection of a low-
angle grain boundary with an etched germanium su rf.'lCC (Fig. 13), they de te r-
llIined the dislocation spacing D. They assumed that each etch pit marked the
end of a dislocation. The angle of tilt calculated from the rclalion 0 :: bID
agrees well with the angle Ineasu red directly hy mealls of x-rays.
111e interpretation of low·angle boundaries as arrays of dislocations is fur-
ther supportt.'<l by the fact that pure lill houndaries move normal 10 themselves
on application of a suitable stress. The motion has been demonstrated in a
beautiful experiment, Fig. 14. 111c specimen is a bicrystal of zinc containing a
2" tilt boundary with dislocations about 30 alomicplanes apart. Oloe side of the
crystal was e1amped, and a force was applied at a point on the opposite sidc of
the boundary. t..lolion of the boundary took place by cooperative motion of lhc
dislocations in the array. caeh dislocation moving all equal distance ill its own
slip ]llanc. "Inc motion was produced by stresses of the order of magnilude of
the yield stress for zinc crystals. strong cviderx.'C that ordinary dcfonnation
results from the motion of dislocations.
Grain boundaries and dislocations offcr relatively little resistance to dilTu-
sion or atoms in compa rison with diffusion in pcrfect crystals. A dislocation is an
 (

t"igure 13 l)j~'ocalion etch pits;n low-angle l oumlar)' on ( IOO) foc'e of g(.·,.,oa,,;o n', the "HglC ofth~'
hountlar)' is 'n.S", The bou ndar)' lies in a (OJ I) plan~; the I'nc of the ,'is\or..lic)ns is (100). The
Burgen "edor is tbe shortest bU ice tr.>nslahon 'Ktor, or 11.01 - tJ!V'2 - 4.0 A. (I". L Vogel, Jr.)

,..
, ,,
- -
~~..-•.-.~~ '\

.-igure 14 Motion of a 'ow-a"gle grain \Joumlan und" •• tress. Tlw boundaty is t ilt: straight vertiCIl
lille, amI il b photographed under \'crlidl ,numination, thc:rc~ mal<lllj( e-,.k",
lilt- 2" angular
change in the clc:wal(e su..tac-r of the -une <:r) . tal ailhe- boundar,.. Tk irregubr hOriZ()n lalln.e is
a small step in thf' de.''''gc ~urface whio:n ~nC~ M II nil'lenet' mark '11)(' CT}'stal,s damped al ,h..
lel't. al the right it IS ~lIbject to a foree normal 10 the plant· of the page. 10", original rm,tion of
boundary, boIlono, 1I>O,ed hack 0.4 mm U \\ ashburn and E . It P.,k,.)
59'
-
open pass.\gc for (Iiffu~ i on . DilTII~ioli is greate r in p1..))lic.llIy dcfollnro nMtcr;'l]
than in :'IllIwalcd (Tyst.tJ.~. DIfTmioll alOl,lg gmill ooundaries controls thc ,",lles of
some precipitation readions in solid~ ; the Jl,'ccipilation of tin frOIll lead-tin
solutions ,It rOom te mperature prOCCc<h nOo1i1 lif' times f:lSter th,ln expected
[rom diffusion ill an ide.ll lattice.

Dislocation Densities
The density of disioc,ltions is thc l1umhcr of di ~ l oca tio n linl's that inte rsect
a Itnit area in the crystal. 11)c denslt)' nlllgc~ from well below Hi- dislocatiOns!
cm 2 in the Ucst ~e lmaniulll and silicon crysluls to 1011 or 10 12 dislocalionslcnl'~
in hea\-i!y deformed metal crys tal~ . -11.e methods ,\\'ailablc for estimating dido-
cation densities are compared in Table 2. The ;lclu,11 dislOC.llion l.'Ollfigur,ltions
in cast or anne.lled (slowly cooled) crystals corrcspoml eithe r to a group of
low-angle grain boundaries or to a three-dimension.11 nct\\ork of disioc.ltiol1s
armnged in cells. as shown ill Fig. 15.
lattice V3Callcics that precipitate 'ilong nn existing cd~c dislocation \\ ill cut
away a portion of thc cxtm h.llf-plane of atoms and causc the dislocation 10
climb. \\ hK-h means to ffiO\'C,jt right angles to thc ~ lip din:ciioll. If no disloca-
tions arc preSCllt, Ihe erysl.u will become slIpe rsdtllr.lted \\ ilh latticc \'aci.lIlcic~;
their precipitation in pclHly-sha pcd \acancy pLltes liMy be foUO\\'I,.·d by collapsc
of the plates and form,\liol1 of (Iislocatioll rings that grow with further \"dc.m ey
prCCipltdtioll. as in Fig. 16.

Widlh M...~i", u ,"

"
pr.l<'h (":tl
Techmque Spcc;= tlllckn('SS Image ,tensity, per em!

Ek'(.·tron rnierQlicopy >1000 A - 1Il0 ,\ 101l_ JOI'Z

X-my t ransrnluion 0. 1- 1.0 mm 51-' m IO'- IW
X-my rt'fleclioo <21-'m (m ill.) - 501-'111 (ma.l.) 21-'111 la>- lQ;
nc...'o mtion - 101-1111 (depth offocus) 0.51-'rn 2 x 101
Etch pib no limit 0.51-'rn L ·1 x loS

·W. C . J~tl)n.
bUrm! of .esok!!ioo of etch pib.

{
Hj,'Urt 15 Cell slnoctllre or Ihrec .. h",<",, ;':mal l.mWcs of' .hsloc..li<"" in ,(.fur",,·d ,,1""';111""
(P. It 5",.. u" .)

Figure 16 Elcctl"On mocrograph cJ do~IQ""(Iho" 1001'S IQrnwd h). ~"'"WIlioo .lIld roIl:rll:>e of .;tI;":ln·
1.'1." III AI-S pe .....·nl \I g ' Iutnffied fro.." 55O'C TIll" hd;c..\ didocatio,,~ Art fOO 11lrtl L) , ...."3Ill")
("O(ldensation on • sen.·.... hslocatlOn \I ag. x-U.OOO. (A. EikUf" aud C 111<)(""')
Dislocufiorl M"lti,)/icafion and Slit,
Plastic deformation causes a very great increase in dislocatiOJl density,
typically from l OS to about lOll dislocationslcm 2 during deformation. If a dislo-
cation moves completely acro.~s its slip planc an of[~t of one atom spacing is
p rodu<:ed, but off.~e ts up to 100 to 1000 atom spacing.~ arc obscrved. This means
tlmt dislocations multiply during dcfonndtion.
Considcr a closed circular dislocation loop of radius,. surrounding a
slipped area haVing tile radius of the loop. Such a loop will be partly edge,
partly screw, and mostly of jntcnnediate character. The stmin e ncrgy of tlle
lOOp increases a.~ its circumference. so that the loop will tend to shrink in size.
However. the loop will tend to expand if a shear stress is acting that favors slip.
A <-'OmmOI~ feature of all dislocation sources is thc bowing of dislocations. A
dislocation segment pinned at each end is called a Frank-Head source, and it
can lead (Fig. 17) to the gencration of a large numbc r of l..'o ncentric dislocations
on a single slip plane (Fig. 18). Related types of dislocation multipl ication
mechanisms account for slip and for the incrcased dcnsity of dislocations during
plastic defonnation. Double cro.u-slip is the most c;:ommon source.

STREI'iGTIl OF Au.o1'S
Pure crystals arc vcry plastic and yield at vcry low stresses. 11lerc appcar
to be fou r important ways of increasing the yield slrength of an alloy so that it
\\; 11 withstand shear stresses a.~ high a.~ 10- 2 C. Thcy an:: mechanical blocking of
dislocation motion, pinning of dislocations by solu te atoms, impeding disloca-
tion motion by short-range order, and increasing the dislocat ion de nsity so that
tangling of dislocations results. All fOur stre ngt he ning mcchanisms depcnd for
their success upon impeding dislocation motion . A fiJ\h mechanism , that of
removing a1\ dislocations frOm the crystal, may operate for certain fin e hairlike
crystals (whiske rs) that are discussed in the section on crystal growth.
fo,lechanical blocking of dislocatioll motion can be produced most directly
by introd ucing liny particles of a second pha.~c into a erystal latlice. This pro-
cess is followed in the hardening of steel. whe re particles of iron carbide are
precipitated into iron, and in hardening aluminum, where particles of AI 2Cu
are precipitated. The pinning of a dislocation by particles is shown in Fig. 19.
In strengthening by the addition of small particles there arc two cases to be
considered; either the particle can be defonned with the mattix, which re-
quires that the particle can be tm" e rscd by the dislocation. or thc particle
cannot be tra"ersOO by the dislocation. If the particle cannot be cut, the stress
necessary to force a dislocation between particles spaced L apart on a slip plane
should be appro1.:imately
ulC = blL . (10)
The smalle r the spm.ing L, the higher is thc yield strcss u. Before particles
p recipitate. L is large and the strength is low. Immediately ancr precipitation is
 ,., <
"
•
•
•
Fillure J7 .'rank-Read ,,,,-"Chanum lOr multipkatiol1 of di~IocaI")ns, $I}()Y.inJ:; i"~ssi ..., stages in
the !)'TlCI"lltioft of a doskx-ali()n loot) b\ the 5t'gIDen' He nf a didoc81ion linc:. 'Ih" P'CX'Cu can t.c
"'vcal<-..J indc/ln;leI,

figure 16 A Frank·Rnd d"kx-a11On 500tJK'e in s;hct'ln, d('('<Jfa,,--d ",th roPl~r Pl'Klpilates and
,_f'<I with iufrarc<l ;nUII"JllItiQn T"o (.,..mpklc <hdocahol11uops a.e \ I~,hk, 8'111 ri,e third , jllflt'r-
mo5t loop is nt'a' lOmplctll)lI. (AfIl'T W C Dasll.)
'"

Figure 19 1);~Ioc,d;om pinnord by p••.udcs U1 mag1>l'~lum o, od ... (Electron micrograph by

G Thomas and J. W.uhbum.)

comple te and many slIlali particles arc prescnt, L is a minimum and the
strength is a ma.~im l1l n. If the allo), is ,lie n held nl a high temperature, some
particles grow at the CX)X'rlSC of others, SO tl1.l1 L incrc.\Sc) and the strength
drops. li ard intcrmetall ic phases, such a..~ rcCrador)' o'\idcs, cannot be cut by
disloc.\tions.
111c ~1 rcnglh of dilute solid solutions is IJclic\'cd to result from the pinning
of dislocations by solute atoms. The solubility of a foreign atom will be greater
in the ncighbo l'llOod of a dislocation tlwn elsewhere ill a crystal. An atom Ll I'"
tends to expand the crystal will dissolve preferentially in the expanded region
near an edge dislocation. A small alom will lend todissokc preferentially in tbe
<.'Ontractecl rcgioll ncar the dislocation- a dislocation offen both expanded and
contracted regions.
As a resull or the affinily or solute ,ltoms ror dislocations, each dislocatIon
will collect a cloud or a.~soc iatcd solute atoms during cooling. at a timc when the
mobility or solute atoms is 11igh. At still lower tempemtures diffusion or solute
atoms effccti\'cly cea.~es, ami thc solute atom cloud bet.'Omes fixed in tile crys-
tal. When a dislocation moves, 1€lIVing its solute cloud Lchind , the energy or
the cnst.li must increa.~e_ 'me increase in energy call only be provided by all
increased st ress acting on the dislocation a.~ it pull ~ ,may rlOm tile solute atom
cloud. and so the prClocncc of the cloud ~t re n gtl1ens tbe crystal.

\
 20 ViJ()('Qfiom.

The passage of a dislQC'dtioll across a slip pi..lnc iu pur(' ('r))tals dO<.'~ not
alter tbe binding energy ,\Cross tllc pl,mc ,uter the di~locati()n is gone. The
internal enell-'Y of the crystal remains undfTeded. '111e s,I IllC is tn-IC for random
solid solutions, bct:ausc tile solution is eqnally 1-,ln(\OIll 'ICro~s a slip plane after
slip. ~'lost solid solutions, however, Ii,\\ e ~hort-runge order, I\loms Ilf different
species arc not arranged at random On the lattice sites, bllt tend to h.\\ e an
CAL'eSS or a deficiency of p.1.irs of unlike atOIllS, l1ws in ordered 'lllo} s di~loca­
lions tend to move in pairs: the second dis\ociltion reorders tile IOCil1 disolder
left by the first dislocation.
111e strength of a ery~ tall in e material increases with phLslie dcfonnation .
'111e phenomenon is cllled wOI'k-hardening Or strain-hardening. l1le strength
is belie\.ro to inere.lSe because of the incrc.u.e<! demit)' of di~locations and the
greater dilfieuhy of mo\ing a given disloc.l!ion aeros» a slip pl,me that is
threaded by many dislocations. Straill-hardcning frequcnt ly is cmployed in the
strengthening of materials, but its w.efulness is limited to low enough tempera-
tures so that annealing dues not occllr.
An important factor ill strain-har<lening is the total density of dislocations.
In most metals dislocations tend to form cells (Fig. 15) of dislorntion-free areas
of dimensions of the order of 1 ",Ill . But unless we can get a uniform lugh
density of dislocations we cannot strahdmrden a metal to its theoretic.11
strength, because of slip in the dislocation-free areas. A high total density is
accomplished by explosive deformation or by speCial thermal-mechanical trt'at-
ments, as of martensite in sted.
Each of the mechan isms of strengthening crystals can 1 .li~e the )'icld
strength to the range of 10- 3 C to 10 2. C. All Illeclmnisms begin to brcak
down at temperatu res where diffl~~ion ('an occur at an apprcci.lhle rate. \Vhen
diffusion is rapid, prccipitate<! particles dissolve; ~olut c douds dlifi .llong with
dislocations as they glide; short-range order repairs itsclfbehind slowly mo"ing
dislocations; and dislocation climb and anneilling tend to dt.'Crl';l5e the disloc-.I-
lion densit)'. The resulting time-dependent deformation is called creep. nlis
irre"ersible motiol! precedes the clastic limit. 11le scart'h for 'llloys for usc at
..,ery Iligh temperatures is a search for reduced diffusion rates, so tllat the four
strengthening mechanisms will sUf\'h 'e to high temperatures. But the central
problem of strong II1l0ys is not strength, but ductility. for fai lure is often by
fracture.

DISLQC.,,'nONS AND CRYSTAL CUmVfll

In some ca.~es the prescnL'C of dislocat ions may be the L'Ontrolling factor in
crystal growth. When crystals aTe grown in eonditions oflow supersatumtiou , of
the order of 1 percent, it has been obsened that the growth rate is enormously
faster than th.lt calculated for an ideal crystal. '11IC actual growth rate is ex-
plained in terms of the eRect of dislOC'ations on growth.
 The theory of growtll of ideaJ crystals predicts that in crystnl growth from
vapor a supersaturation (pressure/equilibrium vapor pressure) of the order of 10
is required to nucleate new crystals, of the order of5 to form liquid drops, and
of 1.5 to form a two-dimensional monolayer of molecules on the face of a perfect
crystal. Volmer and Schultze obscrved growth of iodine crystals at vapor super-
saturations down to less than I pcrcent, where the growth rate should have
becn down by the facio." exp(- 3000) from the rate dcfin ed 3.~ the min imum
observable growth.
TIle large disagreement expresses the difficulty of nuclealing a new mono-
layer on a completed surface 'of an ideal crystal. But if a screw dislocation is
present (Fig. 20), it is never necessary to nucleate a new layer: the crystal will
grow in spi ral fashio n at the edge of the discontinuity shown. An atom can be
bound to a stcp more strongly than to a plane. llte calculated growth rates for
this mechanism are in good agreement with observation. We expect that nearly
all crystals in nature grown at low supersaturation will contain dislocations, as
otilC.w isc thcy could not ha'·e grown. Spi ral gro\'1h pattelTls havc been ob-
served on a large nu mber of crystals. A beautiful example oftilC growth pattern
from a single screw dislocation is given in Fig. 21.
If the growth rate is inde]lCndent of direction of the edge in the plane of
the sUlface, the growt\l pattern is an Archimedes spiral, r = 00, where (I is a
constant. TIle limiting minimum radius of eurvaturc ncar the dislocation is
de tennincd by the supersaturation. If the radius of cun'ature is too small,
atoms on the cu n ·ed edge evaporate until the equilibrium curvature is attained.
Away from the origin each part of the step acquires new atoms at a constant
ratc, $0 that drldt = cons!.
WhiJkers
Finc hairlike crystals or whi~kcrs havc been observed 10 grow under (..'O n-
ditions of high supe rsaturation without the necessity fOr more than perhaps one
disloelltion. It may be that these crystals contain a single axial screw dislocation
that aids their essentially one-dimensionai growth . From the absence of dislo-
cations we would expect these crystal whiske rs to ha\'c higll yicld strengths, of
the orde r of the calculated value GOO discussed carlier in this chapter, A single
axial screw dislocation, if present, could not cause yielding, becaUSE' in bending
the crystal the dislocation is not subjected to a shear stress parallel to its Bur-
gers "ector. TIlat is, the stress is not in a (Iircction that call cause slip. Ile rring
and Galt obsen'ed whiskers of ti n of radius _ 10- 4 cm with eI:lstie properties
near thosc cxpected from theoretically perfect crystals. 'n ley ob.~crved yield
strains of the order of 10- 2 , which correspond to shear stresses of order 10- 2 C,
about lOOO times greater than in bulk tin , confirming the early estima tes of the
strength of perfect crystals. l1lcoretical or jdenl elastic properties have been
obscrved for a number of mnteriaJs. A single domain whisker of nickel is shown
in Fig. 22.
 .
20 lJis/l)CalioWt £05

,~_,-
I
, I

<
"I;-'-,--,0-
I I

Figu~e 20 Ikn'iopmcn t of a ~piral st.,p prod ll('t.-d by ill tCr'S('('liOlI of a scre"' dislocation "ilh the
,,,rfM!C of. Cl')'stal as in "' ig. 8. E:ach robe rl'prCS<;lIu a molN:"lc. (fo". C. Fran k. )

l ' ;gure 21 Phase contrast micrograph of a h.,xagonai Spir.. l growth pattenl on a SiC <T)'stal. 'The
step height ;s 165 A (.... R. V" rmL )

. 'igure n A nickel " 'ilisl..l'. of dlallleter 1000 A Ixu! in a loop (II. W De BloD.J
 I1Al\I)NESS OF :\IATEIUALS

The hardness of materials is measured ill several ways, the simplest test for
nonmetals being the scratch test . Substance A is harder than substance B if A
will scrJ.tch B but 8 ,viII not scratch A. A standard scale is used for representa-
tive mincrals, with diamond, the hardest. assigned the value 10 and talc. the
softest. assigned the value 1:
10 diamond C 5 apatite Ca5(PO.hF
9 corundum Al20a 4 lIuorite CaF 2
8 topaz A~SiO.F2 3 calcite Cacoa
7 quarr.l Si02 2 b'Ypsuln CaSO..! . 211 2 0
6 orthoclase KAISiaOs 1 talc 3Mbr() . 4Si0 2 . 1120
TIlerc is great current intcrest in thc development of materials of great
hardness. for cxample as films for use as scmtch-rcsistant coating... on lenses. It
is widely felt that the scale between diamond and corundum is misleading,
because diamond is much. much harder than corundum. It ha... been suggested
that one might assign diamond the hardness 15. ' vith the gap between 9 and 15
to be filled in eventually by synthetic materials. such as compounds ofe :md B.
Modern scales ofhardncss. such a.~ the VIIN scale. arc based 011 indenter
tests in which an indenter is pressed into the surface of the material and the size
of the impression is measured. 'nle Vickers Hardness Numbers of selected
materials are tabulated below. after conversion by E. R. Weber to units of CPa
{CN/m 2 1:

Diamond 45.3 1le0 7.01

SiC 20.0 Steel (quenched) 4.59
SiaN. 18.5 eu (annc:llcd) 0.25
AI 20 a 14.0 AI (anne:lled) 0.12
B 13.5 Pb 0.032
we 11.3
The data are from thc text below by J. c. Anderson and othen.

Problem.s
I. Lillt' of closts' packing. Show that the lines of closest atom ic packing nre (110) ill rcc
structure5 and (lI l) ill bee $tructures.

2. DUlocation pai",. M Find a pair cl dislocatioll5 equivalellt to a row of lattice vacruJ-

des; (b) find a pair of diskx:atj()1IJ eqUivalent to a row ci interstitial atoms.

3. Force on tlislocation. Consicle r a crystal in the form of a cube of side L (.'ontaining all
or
edge dlslocatloll Burgers vector h . If the ery5tall~ subjected to a shenr $tress U 011
tile ujJpcr and lower races In tile directiOIl of slip. show by considering energy bal-
ance that the force acting 011 the dislocation is F - bu per' .~ lellgth.

...
 Refenmces
ELE~I EN"fAR l'

D. Hull and D. J 8aeorl. Introduction to dislocadoo,,- 3rd cd .• ~rga,non, 1984.

C. R. fiMrl'lt , W. D . Ni~. and A. S. Tctclrnan, Prb'c!ples of englneerl,,!,: materiah, Prent ice Ii all.
1973.
M. J. whelan, \l'orbd e:rt:mplts til di.slouti(m~, Institute ol ~Ietal" 1990.
S. Amelinc:b:, Dired obs.".ootil)n of dtslocotillnt, S"ppl. 6 to Solid ~tate (lh)'~IcI, A<:adetnic, 196-1.
J. C. :\ndcl'$OO and otllel")', Malerill/4 ide"ce, 3rd editlcn , Van NQ!;lrand Reinhold (UK). 1983.
ADV,..",crO
F. R. N. ~abarro, Tlwory of Cfl/5lal fii31ocatillor', Oow:' , 1987.
J. P. HIrth and Jem Lothe, Throry of di..JloaJtlotu, 2nd ~., Wiler. 1982.
G. Thoow and ~I. J. C<.>ringe, Tnm' miuiIm elwron m6cro.COPIi of materilll" Wiley. 1979.
 t

21
AUoys

CENEnAL CUNSIDEnA'1l0NS 611

SUBs'nTunONAL SOUD SOLUTlOfliS-
IIUME-R<yrHERY IWUS 614

OHDEI\·DISOROEn 'O\ANSFOm.IATION (i17

ElemcJ1tary theory or ortlcr GIB
PltASE DIAGRAMS 62:2-
Euteclics 622

THANSl'l10N METAL ALWYS (;2.4

Electrical Conduclh'ity li2:4

KONDO EFFECT 626

l'nOBLEMS (;30
1. SUI>crlattice lines in CII:.AI,I 630
2. Configurational heat caJ)acily f030

ItEFEIlENCES 630
 (

0 0 0 0 0 0
>---
0 0 0 0 () 0

0 0 0 e 0 0

FIgUre I Oniered (a) an.:! disordered (b) .mallgemcnt, of AB ions;n . '..JiQ)' AH

'"
------- CHAPTER 21: ALLOYS

CE!,;ERAL CONSIDEHATlONS

The theory of the band structure of solids assumes that the crystal has
tnmslation.ll iuvariance. But suppose that the crystal is composed of two cle-
ments A lind 13 that occupy at random the regular lattice sites of the structure,
in proportions x and 1 - x for the composition A.JJI _." The translational sym-
metr)' is no longer perfect.
Will we lose the consequences of band theory, such as the ('xistence of
Fermi surfaces and of energy gaps? will insu lators become conductors because
the energy gap is gone? We touched on these questions in the discuSSion of
amorphous semiconductors in Cllapter 17.
EI>periment and theor)' Hgree that the conse(IUellCCS of the destmction of
perfect translational symmetry arc much less serious (nearly alw·J.YS) th:m we
expl.."Ct at first sigll!. The \iev.]>oint ofthc elTccth'e screened potential of Chap-
ler 9 is helpful in these matters, first because the effective pote'llt;,lls are rela-
tively weak in comparison with a free ion potential and, S(."Cc)Ild and most im-
portant, the differences between the elfl..'Ctjve potentials of the host and the
additive atoms llIay be very weak in comparison with either alone. Alloys of Si
and Ce or of Cll and Ag are elassic examples of what we may call the relative
illcffl..>divencss of alloying.
In any event, a low concentmtion of impurity atoms cannot Ilave much
effect on the Fourier components Vc of the effective potentiill VCr) that is
responsible for the band gaps and for the form of the Fermi surface. (11lis
statement implies that the C's exist, which implies that a regular Iilttice exists.
This is not an important assumption because we know tholt thermal phonons do
not have dr.lStic effects, so that lattice distorttolls described as fro7.en-in
phonons should 1I0t have drastic elfl.."Cts. If the distortions are more serious, as
with amorphous solids, the electronic changes can be significant.)
It is II tie that an impurity atom will introduce Fourier components of U(r)
at waVe\'cctors that i1re not reciprocal lattice vectors, but at low impurity con-
centration such components are never large in comparison with the Vc • argu-
ing from the statistics of random potentials. The Fourier components at the
rcciprocallattice vectors C will still be large and will give the hilml gilpS, Fermi
SUr£.ICCS. and sharp x-r.ly diffroction lines chamcteristic of a rebrular lattice.
The consC<luences of alloying WIll be p.-l.rticularly small when the impurity
element belongs to the same column of the p'eriooic table as the hmt clement it
replaces, because the atomic cores will make rJ.ther similar contributions to the
effective potent
 "r-------r------,-------,-------,

"

,
c. '" " ".
'"
i"ig u.c 2 Res;.!i,"i!), of a diwrd" ..w hinary alloy of coIlII.-r and b'Old. The "ari;ltion of the residual
rcsi~ti"It )· depends on the <'(>mpvsiliQll Co. AUt _. as r( l - lr), whk:fl is l:nown 11$ Nordheim"s Rule
!'or II d.wnkrcd a1lo}'. l lere x( l - x) is a m('ltSure of ti,e degree of tn~hnu'n diwrdcr possihle fur II
SI,'cn value or r . (JoIlans'>Oll a",l Linde.)

One measure of the effect of alloying is the rcsidlJul electrical resistivity,

dcfinc.."<.! as the low temperature limit of the resistivity. 1{ere we must distin-
guish between disordered and ordered alloys. An alloy is disordered if the A
and B atoms are randomly arranged. ~n ich occurs for a genen!.! value of x in the
composition AABI _r' For special \·alucs of x. such as 1/4, 1/2, and 3/4 for a cubic
stnK:ture, it is possible for ordered phases 10 form, phases in which the A and B
atoms form an ordered array. The distinction betv.'ccn order and disorder is
shown in Fig. 1. The effect of order on the electrical resistivity is shown in
Figs. 2 :Uld 3. TIle residual resistivity increases with disorder. as dis('llssOO for
amorphous materials in Chapter 17. The effect is shown in Fig. 2 for the Cu-Au
alloy system. When the specimen is cooled slowly from a high tempcr.lture.
• 21 AIIO!J8 613

"',, - --,----,------,------,

"
,"
~

,
'I- -.\
I'

' 00

'"
Figure 3 Effect of o rdered phases On the resistivity of a binary alloy Cu.Au' _r '11'e 1I11o~5 I,ere
have been a""<!.Iled. "ilereas those in Fig. 2 have bee ') quenched (<:OOIl:d rapidl y). The <:omposi-
lions orJvw residual !'Csil;livity C()rl't'spond t{) the ordered cornp..>!;itivns Cu.,Au and euAu_ 0 01,.." 5-
son and Linde. )

ordered structures arc fOflm"O .. at ClI:}Au and euAu; these structures have a
lower residual resistivity by vil1ue of their order, as in Fig. 3.
Thus we can use the residual electrical resistivity to measure the clfcct of
alloying ill a disordered structure. One atomic percent of copper dissolved in
silver (which lies in the same column of the periodic table) increases the ,-esid-
ual resistivity by 0.077 p.ohm-cm. 111is colTcsponds to a geometrical scattering
cross section wh ich is only 3 perccnt of the naive "pmjcctl,.-d area" oftbe.impu-
rity atom , so that the scattering effect is vel)' small.
In insulators there is no experimental evidence for a signiflcant reduction
of band gap caused by the mndom potential components. For example. silicon
and germanium form homogen(.'o us solid solutions. known as substitutional
'"
uIl0)'5. over the en lire composition nmgc. bul the band edge energies vmy
continuously with composition from the pure Si gap to the pure Cc gap.
It is widely believed, however, that the de nsity of s ta lL'S neur the buml
edges in amorphous materials is smudged out by the gross ab~encc of transla-
tional symmetry. Some of the Ilew states thus formt..od just inside the gap m<ly
not necessarily be current-carrying stales because they may not extend
throughout the crystal.

SUBSTITUTIONAL SOUD SOLU'flONS-HmIE-ROT HEI\\' RULES

\ Ve now disc uss substitutional solid solutions of one metal A in another

metal B of different valence. where A and B occupy, at random, equivalent
sites in the structure. Hume-Rothery treated the empirical requirements for
the stability of a solid solution of A and B as a single phase system.
One Hx]uirement is that the atomic diameters be compatible, which
means that they shoulduot differ by more than 15 pe rcent. For c."<amplc, the
di.lmeter.; are favorable in the Cli (2.$ A) - Zn (2.65 A.) alloy system: ~I"K'
dissolves in copper as an fcc solid solution up to 38 atomic pcrt'ent 7;inc. The
diamete rs arc less r.worable in the Cu (2.55 ft.) - Cd (2.97 ft.) system, where
onlv l. 7 atomic percent C".u:lmium is soluble in copper. The atomic d iameters
rcfern><i
.. to copper .lre 1. 04 lOr zinc and 1.165 for cadmium.
Although the atomic diameters may be favorable, solid solutions wiJl llOt
form when there is a strong che mical tendency for A and B to form Uintermetal-
lie compounds." which arc com pounds of dl.1rnite chemical proportions. If A is
strongly electronegative and B strongly ek'Ctropositive, compounds such as AB
and A2B may pn'Cipitate
.. from the solid solution. rolls is dilTerent from the
formation of .m orderecl alloy phase only by the greater chemk-.I! bond ing
strength of the intermetalliccompounds.) Although the atomic diameter ratio is
favorable lor As in Gu (1.02), only 6 atomic percent As is soluble. 111e diameter
ratio is also favorable for sb in Mg (l. 00), yet the solubility of Sh in Mg is vcry
small.
The electronic structure of alloys C"an often be described by the average
number of conduction electrons (or vall.'llCe electrons) per atom. denoted by ' 1.
In the alloy CuZn the value of n is 1.50; in euAl, 11 :: 2.00. Ch(Ulges in electron
concentration determine structural changes in many alloy systems.
The phase diagram of the copper-zinc system I is shown in F ig. 4. The fcc
stmcture of pure copper (n = I) persists on the addition ofzinc (1I "" 2) until the
electron concentration rt!uches 1.38. A bee structure occurs at a minimum elcc-

'The phas..., ofintcr~t are usu.al1)·denot<,d b) metanurghts b\ Creek <.....mw;I"'"' ,n the Cu·Zn
!~.Iern "eha,e Q (1«). fJ (ocd. y{ro.npb: l'Ubic n:Ud52atQms).' (bcp) and" (kp). (and 1) d,ffer
OOtlSltknbly III dtl rallo The meamng of the dwacten dqx-OOs on the a11o). s)"'tcrn.
 , '"
"oo~
,." ,,-+- J"od 1~,Jod
~, ~<{"'.,.,'
"".
/ ~
'y +l"l",d -

, , , ",
"
, ,, :\.,\6+3\
,.,. '"
.\
.",.

,...
"'. - - ---
• .".
• 00

/1'pltr+'Y ,,, ., ,
'"' , "
""j 00
W"~I p<.Irent 'm"
J-'igurc'\l E'lUlhbnullI di~g.·,"n of pl~s In the <:Opp"r~z;n<: a1lov ~l'~tem. TIle a pha.;c is f<:c; f1 al,d
(1' me be.:. Y,$ a complex structu,..., . ~ aIld 1] art' both l><-p, but" 1"'5 a d a mHo I1c"r I 56 and 1] [or
pure ZIl) has d a : 1.66 TI,., (1' pll<l!;e i.I ordc""! box, by" hi<:b we mCan that mO'lt orthe C.. "',.ms
OC'CUPY Sltcs On One S<c .ublatll<.'e ami "'<.lSI of the Zn _.toms Ol~'Up~' SJlc~ on .. !;t,.'«llld sc .llblattla!
that mte,'pcn(!tr:tlcs the first .ublatti~e. "l1,e {3 phdSC IS dISordered 00,: all)' .,t ..
is t.~lu"ny M.l'l~ to
be Ottupiro by a Cll ur Zn atom, ,,1nlO:St irresp.:cti\e of "hat atoms are '" the ndghbormg "\t'S.

tron conccntr.llion of about 1.48. The y phase exists for the appro"imnte range
of II between 1.58 and 1.66, and the hql phase € oa:urs ncar 1. 75.
The term electron compound dcnotes all intermediate pha~(;' (such as the/3
phase of CuZn) who~e crystal structure is determined by .. fairlv well denned
electron to atonl ratio. For many alloys the ratio is close to the I [ume-Rothen'
mles: 1.50 for the f3 phase, 1.62 for the "y phd.~e, and 1.75 for the € phase.
Representative cxpcriment.ll values arc collected in Table I, based on the usual
chemical valence of 1 for eu and Ag; 2 for Zn and Cd; 3 for AI and Ga; 4 for 5.,
Gc, and Sn
The lIume-Rothery mlc~ find a simple e:l.pression in terms of the bar ,d
theory of nearly free electrons. The observctllirnlt of the fcc phase occurs close
to the e1ectr-on concentration of 1.36 at which an instribed Fermi ~phere maJ..eS

I contact with the Brillouin zone boundary for the fcc lattice. The observed e lec-
tron concenh.ltion of the bet: phase is close to the concentntiOIl 1.48 at which
an inscribed Fermi sphere nMkes contact with the zone llOllndilT\· for the bec

I
 Table I Electron/atom ..... tios of electron rompoune4

MinimuUl
fcc .,h;l.le bee phase 'l'-.,hase
AI." boundary bouudary bou ndaries

Co-Zn 1.38 1.48 1.58--1.66 1.78-1.87

Cu-'" 1.41 1.48 1.63- 1.77
Cu-Ca 1041
Cu-Si 1.42 1.49
Cu-Ge 1.36
Cu-Sn 1. 27 1.49 1.60- 1.63 1.73-1.75
Ag-Zn 1.38 1.58- 1.63 1.67- 1.00
Ag-Cd 1.42 1.50 1.59- 1.63 1.65-1.82
Ag-Al JAI 1. 55- 1.80

•
laltice. Contact of the Fermi sphere with the zone boundary for the l' phase is
at the concentration 1.54. Conlact for tile hep phase is at the concentration 1.69
for the ideal da ratio.
\Vhy is the re a connection between the e k"Ctl"on concentrations at which a
new phase appears and at which the Fermi surface makes contact Witll the
boundary of the Brillouin zone? We recall that the energy bands split into two
at the region of contact on the zone boundary (Chapter 9). If we add morc
eh.."Ct rons 10 the alloy at this stage, they will have to be accommooah..--d in till;
upper band or in states of high energy near tilt: 7.one l'Orners of tile lower band.
Both options are possible, and both involve an increase of e nergy. 111erefore it
may be energetically El\'orable for the crystal structure to change to one which
can contain a Fenni surface of larger volume (more eit.."Ctrons) before contact is
made with the zone boundalY In this way II . Jones made plausible the st;-
quence of structu reS fcc, bee, 1', hep with increasing electron concentration.
Measurements of the lattice paramcterofli-r-.·lgalloys are shown in Fig. 5.
In the range shown tile structure is 1x.'C. TIlC lattice contracts during the initial
stages of thc addition of r-.lg to Li . WIlen the lithium l'Ontent drops below 50
atomic IlCreent, corresponding to an average electron concentration increasing
abo" e I,SllCr atom, the lattice starts to expand. \Ve have seen that for a spheri-
cal Fermi surface (.'ontact ,,;th the zone boundary is established at fI == 1.48
eit."Ctrons per atom , in a bcc lattice. It ap pears that the expansion of the lattit'e
arist."S from the onset of overlap acrOss the zone boundary.
The transfonnation from fcc to bee is illustrated by Fig. 6; this shows the
number of orbitals IlCr unit energy range as a function of ene rgy, for the fcc and
bee structures. As the number of ek"Ctrons is increast."'<i, a point is reacllt.--d
where it is easier to accommodate additional electrons in the Brillouin zone of
the bee lattice rather than in the Brillouin zone of the ft'C lattice. The figure is
drawn for COJ.)llCr.
 21 Alloy" 617

"
..,,
,
3520 -

, 3510
li
<
~

~ '"'"
i
~
3-490

J•

"-!"'''Ic ",,,,,,,nt lith""

Figure 5 l.attice llardmeter of body-ccnlered cubic magncsium -lilil ium alloy$. (Aflcr
I). W. Levinson. )

.1\'"
/ ~
,/
,/
I
,
•,
,,'igure 6 Number of orbitals per ullil energy range
ror the firsl Brillouin ZOne ofille r~"C and !Jcc lattices. 01234 5 , ,
as a fonctiun of cncrgy. !::'>crgy in "V

ORnER-DISORDER TRANSFORMAll0N

The dash(.>d horizontal line in the beta-ilhase (bee) region of the phase
diagram (Fig. 4) of the Cu-Zn system represents the transition temperature
between the ordered ~ow temperature) and disordered (high temperature)
states of the allOY. In the common ordered arrangement of an AD alloy with a
bee structure all the nearest-neighbor atoms of a B atom are A atoms. and vice
versa. This arrangement results when the dominant interaction among the
atoms is an attraction between A and B atoms. (If the AB interaction is weakly
attractive or repulsive, a two-phase system is formed in which some crystaJlites
are largely A and other crystallites are largely D.)
 c
618

11lC alloy is (.'OmpJctcly ordcrf'd lit abo;olute zero. It becomes If'SS ordered
as the te mperature is )llI,:reased. until a transit jon temperature is rf'aclled above
which the structure b. disordered . nle transition temperature marks the d isap-
pearance oflong-rangc ol'def, whicll is order over many interatomic distances.
but some short-range order or correlation among near neighbors may persist
alxwc'the transition. 11m long-range order in an AB alloy is shown in Fig. 7a.
Long- nno shorl-r.lnge order for an alloy of composition An:! is given in Fig. 7b.
The df'gree of order is defined below.
If an alloy is cooled rapidly from high temperatu res to a temperature below'
the transition, a metastable condit ion may be plooU(;ed ill which a nonequilib-
rium disorder is frozen in the structure. TIle reverse circe! occurs when an
ordered specimen is disorde red at constant te mperature by heavy irradiation
with nuclear particles. The degree of order may be investigated experim(·ntally
by x*ray diffraction. 11le disordered structure in Fig. 8 has diffraction Jines at
the same positions as if the lattice points were all occupied by Ofliyonc type of
atom, because the effective scattering power of each plane is e<Jual to the aver-
age of the A and D S<.'Uttcring powers. 11le on.lered structure has extra diffffil'-
lion lines not posseSSl."<l by the disorde red structure. The extra lines are called
super struct ure lines.
TIle use of the tenus order and disorder in this chapter ah vays refers to
rebrular lattice sites; it is the occupancy that is randomly A or D. Do not l'Onfuse
this usage ,\lith that of Chapter 17 on noncrystalline solids where there are no
regular lattice sHes and the structure itself is random. Both possibi1iti~, hO\v*
ever different, occur in nature.
TIle structure of the ordered CuZn alloy is the cesium chloride structure
(Fig. 1.20). The space ll"ltlice is simple cubic, and the basis has one eu atom at
000 and olle Zn atom at U+ The diffraction structure factor
r
S(hkl) ::: Jell + JZ.
J: n ,'..-(III+k+f) . (1)
This cannot vanish bc<.-ause fe .. '" fzn; therefore all reflections of the simple
cubic space lattice will oc'Cur. In the disordered structu re the situation is d iffer-
ent: the basis is equally likely to have either Zn or Cu at ()(Xl and either Zn or
Cu at Hi. Then tl'le average structure factor is
(S{hkl) ::: (f) + (fk ...r(h + ~ + I ) • (2)
whe re (f) ::: j(fo.. + 17...)· Equation (2) is eAactly the fonn of the result for the
bee latlice; the reflections vanish when 11 + k + 1 is odd. We see that the or*
dered lattice lms reflections (the superstructure lines) not present ill the disor-
dered lattice (Fig. 8).

Eleme,!ta ry rTlcory of Order

\Ve give a simple statistical treatment of the dependence of omCT on tem-
perature for all AD al loy \\jtb a bee structure. TIle case~'" 'iffers from AB, the
 21 AJ/¥

"'r---=:::::----,

"
,. "

1---____ _
"
Figure 7 (a) Long-range order venus tcmpl'rntmc ro.. an A8 alloy. TIle tnl!lSror ..... lion ts scc.-oud
order. (b) WI1Il:-r:!.l1ge and short-range order lOr an J\~ alloy. TIle tr.lIlsronnatiOIl ror thi~ Q)mpml-
lion is firsl order .

.,
.,
I Figure 8 X-ray I'ow&. photog1aphs In AuCuJ alloy. fa) I)iKlrucrcd by (lue"dung from T> T,.
(1.0) ordered loy alll1eahng ..1 T < T.. (C..,urt,~y ofC. \I Cordon.)
"'"
(
fonner having a first-order transition marked b). a latent h(:.... and the latter
having a sC<.'Ond-order transilion mark<.-d by a discontinuity in the heat capacity
(Fig. 9). We introduce a measure of the long-range order. We call one simple
cubic lattil'e a and the other b: the bee structure is composed of the two inter_
penetrating sc lattices, and the nean.-st neighbors of an atom Oil one lattice lie
OIl the other latti(.'C. If there are N atoms A and N atoms B in the alloy, the
long'I'ange order pm'ameter P is defined so thaI the number of A's on the
lattice a is equal to -1(1 + I')N. 11m number of A's on lattice IJ is c(jual to
l(I - l')N. \ Vhcn P = ± l , the order is pcrft.'d and each lattice contains only
one type of atom. When P = 0, caell lattice contains f'Qualnumbcrs of A and B
atoms and there is no long-range order.
We consider that part of the intemal energy asSOCialt.'<l with the bond
energies of AA, AB, and BB nearest-neighbor pairs. The total bond energy is
(3)
where Ny is the number ofnearest-neighborij hoods, and Vb is the energy of an
ij bond.
111e probability that an atom A onlalliCCll \~il1 have an AA bond is e<Jual to
the probability that an A occupies a particular nwrest-neighbor sile on b, times
the number of nearest-neighbor sites, which is 8 for tile bee structure. We
assumc that the probabilities are inoc'pendent. l11US, by the preceding expres-
sions for the number of A's on a and I),
NM = SlI(1 + I')N](I(I- 1')] = 2(1 - P']N ;
N .. = 8[1(1 + P)N ](I(I - 1')] = 2(1 - P")N ; (4)
NAB = 8N{l(1 + p)J2 + BNti(1 - p)J2 = 4(1 + p2)N
"f11e energy (3) bc.-'"COmes
E = Eo + 2NJ>2U , (5)
where
v = 2UAB - V AA - U BB (6)
We now calculate the cn tropy of this distribution of atoms. 'l1\Cre are
i( L + P)N atoms A and. ifl - P)N atoms n on lattice tl; the re arej( l - P)N atoms
A and 4(1 + p)N atoms B on lattice I}. TIle number of arrangements G of these
atoms is

N! ]' (7)
G = [ [1( 1 + l'JNJI[I(1 - I')NJ! .
From the definition of the entropy as S = kll In G, we have, using Stirling's
approximation,
S = 2Nk. In 2 - Nk,[(1 + I') In (I + P) + (I - I') In (I - 1')[ . (6)
This deflnes the entropy of mixing. For P = ±1. S = 0; for I' = 0, S =
2.Nk8 In 2.
 21 A"oy~ 621

(
"
,
"
I
"
- ~II/
~ ~--
-
.w
w- <~
"
~

/
," ./
-- -- -
Figurc 9 Ikat capacity vcrsus tcmpcraluJ"<; of C.,zn
alloy ep-bnl$S).
", 100 200 300 '00
Tem!",r'ttJre In "C
500 600 700

'l1te equilibrium order is determined by the requirement that the free

energy F "'" E - TS be a minimum with respect to the order parameter P. On
differentiating F ,""ith respect to P, we have as the condition for the minimum
I+P
4NPU + NkBT In - - = 0 (9)
I- P
111e transcendental (.>(juation for P may be solved graphically; we find the
smoothly decreasing curve shown in Fig. 7a. Near the transition we mayex-
pand (9), to find 4NPU + 2Nk B TP"'" O. At the transition temperature P = 0, so
that
(10)

For a transition to (K"Cur, the d k-ctive interaction U must be negative.

'l1te shOli-range order parameter,. is a measure of the fraction of the
average number q of nearest-neighbor bonds that are AB bonds. When com -
pletely disordered, an An all o~' ha'! an average of four AB bonds about each
atom A. 111e total possible is eight. We may ddine
r=!(q - 4) , (II)
so that,. = 1 ill comple te order and ,. = 0 in complete disorder. ObserVe thal r
is a measure only of the local Order about an atom, whereru; the long-range
order parameter P refers to the purity of the ent ire population on a gi\'en
sublaltice. Abo\'e the transition temperature T~ the long-range order is rigor-
ously zero, but the short-range order is not.
 PHASE DlAGRAMS
There is a large amouut of infonnation in a phase diagram e\'en for a hinary
syste m, as in Fig, 4. Thc areas cnclosc.>d by cu .... cs relate to the equilibrium
state in that region of (:omposition and telllperniure. TIle cu rves Ill.uk the
course of phase transitions as plott!'" in the T-x plane. where x is tile composi-
tion paramete r.
11le equilibrium state is the state of minimum free energy of the binary
sy~ tem at given T, x, Thus the analysis of a phase diagram is the subject of
thermodynamics. Several extraordinary results (:ome out of this analysis. in
particular the existence of low-mclting-point e utectic compositions. Because
the anal y~is has been treatcd in Clmptcr 11 offf, we only outline the principal
results here.
Two substances will dissok e in each other and form a homogeneous mix-
ture if that is the configuration of 100\lI..'5t free energy accessihle to the compo-
nents. The substances will foml a heterogeneous mi~ture if the combined free
energy of the two separate phases side by side is lower than the free energy of
the homogene<lus mixture. Now we say that the mixture exhibits a solubility
gap. In Fig. 4 we see that compositions Ilear Clio ooZno .ware in a solubility gap
and are mixtures offcc and bee phases of different structures .Old com llOSitions.
11le phase diagram represents the temperature dependence of the solubility
gaps.
when a small fraction of a homogeneous liqUid fl"f'Czes . the composition of
the solid that forms is almost always different from that of the liquid. Consider a
horizontal section ncar the l..'Ompos ition Cuo ~no 2() in Fig. 4. Let x denote the
weight percent of zinc. At a given tempelature. there are three regions:
x > XL. the equilibrium sy.;tem is a homogpneous liquid .
Xs < x < XL. there is a solid phase of composition Xs and a liquid
phase of composition XL.
x< Xs, equilibrium s~tem is a homogeneous solid.
111e pOin t x,. traces a curve~lkd the liquidus cu ....·e. and the pOint Xs traces tile
solidus curve.
Eutectics. r..Uxtures with two liquidus branches in the ir phase diagram are
called eutectics. as in Fig. 10 for the Au-Si system. 111e mini mum solidification
temperature is called the eutectic temperature; here the composition is the
eutectic romposition. The solid at this compos itioo (:ollsists of two separate
phases. as in the microphotograph of Fig. 11.
·nlere are many binary systems in which the liquid phase persists 10 tem-
peratures below the 100\'er melting temperature of the constit uents. Thus
AUo &g5io 31 solidifies at 37ff'C as a two-phase he terogeneous mixture. although
Au and Si solidify at 1063°C and 1404°C. respecti\'cI\'. One phase oflbe eutec-
tic is nearly pure gold; the other nearly pure silicon.
 2 1 Mloys

",.,
..., V l-
t.\(w

",., ..,.
"~ ...,
. /'
V
./'

,, .., ~ /
".
,
• :\
".,
.., II
'- - 3t ""
,

Figure 10 E"tcdlC phase di~\tll J w.>ld-sdiwn alloV$ 1bc cutedicron~isb of two branchcs that
comc together at T~ _ 37ctc and x. " 0 31 atomic pcrL'Cnt Si. (After Kittel ;mel Kroc""'r, Tl'.)

'2
~~1§ ;i?
;;= .:
'P'

:£3~
II
~ : ~,
/I '" 1\\
'II
'(1
V{III
rll'.
III
rh\ ,

I~ ~~~
11111
rlllll'111
...... A.
~

10 J.un
Figure II Microphotograph of the I'l.-Sn eu tedte. (Coortesy of J. D. 110llt and K. A_ J,lCk:w,. )

•
'"
°n.e Au-Si euh.>ctic is important in semkondudor tecilnl>. 6Y because the
cutl..>ctic permits low temperature welding of gold conlat'!. wires to silicon dp..
vices. Lead-tin alloys have a similar eutectic of 1'lJ() 2r;SUo.74 at 183"C. -I11is 0 1-
nearby compositions are used in solder: lH!arby if a range of melting tempera_
tures is desired for ease in handling.

TRANSITION METAL ALWYS

When we add copper to nickel the eff(."Ctivc magneton number per atom
decrease5linearly and goes through zero ncar Cuo.roNio.40. as shown in Fig. 12.
At this composition the extra eJt.>ctron from the copper has filled the 3d band. or
the spin-up and spin-down 3d sub-bands that were shown in Fig. 15.7b. 111C
situation is shO\vn schematicnlly ill Fig. 13.
For simplicity the block drawings represent the density of states as uni-
form in enerb'Y. The actual dell5ity is known to be far fTOm uniform; the r(.'Suit ci
a modem calculation is shown in Fig. 14 for nickel. TIle width of the 3d band is
about 5 eV. At the top, where the magnetic eITt..'Cts arc determined, the density
of states is particularly higll. 111e average density of states is an order of magni-
tude higher in the 3d band than in the 4$ band. This enhanced density of states
ratio gives a rough indication of the expected enhancement of the electronic
heat capacity and of the paramagnetic susceptibility in the nonferromagnetic
transition metals as compared with the simple monO\'aJent metals.
Figure 15 shows the efft..'Ct of the addition of small amounts of other ele-
ments to nickel . On the band model an alloying metal with :t valence eit.>ctrons
olltside a fillt:."<l d shell is expected to decrease the magnetization of nickel by
approximately:t Bohr magnetons per solute atom. TIlis simple relation holds
well for Sn, AI, Zn, and Cu , with:t = 4, 3, 2, and J, respt.'Ctively. For Co, Fe,
and t-,'In the locali7.1.>d moment model of F rit."<lel acco unts for eIT(."Cti\'c z values of
- J, - 2, and -3, resp(.>ctively.
'i1le average atomic magnelic moments of binary alloys of the elements in
the iron group are plotted in Fig. 16 as a function of the concentration or
e1cctrons outside the 3p shell. This is calk>(\ a Slater-Pauling plot. 'J1\C main
setluence of alloys on the right-hand branch follows the ruk:s discussed in con-
nection with Fig. IS. As the e lectron conccntration is decreased a point is
reached at which neither of the 3d sub-bands is entirel y fillt..>(\, and the mag·
netic moment then decreases toward the left-hand side of the plot.
ElecrriCllI Conductivity. It might be thought that in the transition metals
the availability of the 3d band as a path for conduction in parallel with the 4$
band would increase the conductivity, but this is not the \vay it works out. 'nit:
resistivity of the $ electron path is inCTeascd by colli~ions with the d eit.>ctrolls,
this is a powerful extra scattering mcchanism not a\'ailable when the d band is
fHlt..>d.
 21 AlloyJ

,,
,,
,"
, 1"\
, I'"
,
,, '" ,,
,
,, [JJ[O
,, 00
" " <,

Figure 12 Bohr m"gucton "umbeD of IIickcl-ropper Figure 13 DI5Iribu l;on or electrons in the alloy

"'' ' . 6QC,,·10Ni. The utra 0.6 electron prmided by

the copper hllli r.J led the d barxl enti rely and in-
"",,,sed slightly the number of electrons in the,.
band ,,;11, resiled 10 f:ig. 15.7b.

r------------------------------"
0
,.,
0 ° 0

- ,

0 ..
' ,'--.
-, ••••••
-,
•• 0 I

-, -, ,
- 00
"
l-'igure 1. Dens ity or states in nid<el. (\'. L. Moruzzi, J. F. Janak, and A. R. Williams.)
62'
..
,,
,
,, ,,
,
,,
,
,, ,/,;.,"'-'"
, •,
~
/
,i. ~/
£
•
t , Y /

~ V V--
,, • / ~
/
•• , v
<

~
• o. /
~ _'In
/
---- ------ -----
o
~\ C,· "
, ~;~ snX ~ ' ..
0
o
" ", , '"
"
--, ~

".
Ad,k-d de""',," in ah>rn percent
Figure 15 Saturation magnet ization of nickel alloys in Bohr milglletons p<-'r ~Iom as ~ fundioll ~
the atomIc perrent of so lute element.
 21 Alfoyr

,,
® .·c-\'

,, • F...c.
o r.... x,
L,..·~· .':,. ,<"<.
, J!-
".
-~ ,
0 N,·Co

,
j/"
Jji
,
,/ ~
, • N, Co
•V ;-.,-\,
o
'\:,·7...

N,·C,
I
I

I/~tI
"0 Ni·Mn
I
, i\. • c.c.
•• '0"
II
Q,.M"

')..
, / •
"",,,,I.
o

,
<>
, ..., ,.,
,
!
,
Co
•
I WI ' ""
!
Figure 16 A"erase "'-olllic rT\QfnCn(s or biliary ..110)$ nfthe dement s in the irUIl group (Ikm>!th.)

We compare the values of the e1ectric.11 res i~tiv ities of Ni, Pd, and PI in
microhm-cm at 18"C with that of thc noble Uletals Cu, Ag, and All immediately
following them in the pe riodic table:

Nt Pd

7.4 10.8 105

1.7 1.6 2.2

'nlC resistivities of the noblc m(:tals arc lower than those of lhe t ransition met-
als by a factor of the order of 115. This shows thc cffectivcness oftlle s-d scatter-
ing mechanism.
 • '"

" '"

- 200 3 6 9 12
~.,

Figure 17 ~'agnctiz3tion of a frccoeleclron Fenni S"'i lit T - 0 III nt'igI,'-hood of. poi!>! t~OC
moment at tbe origin r _ 0, IlCCOming to the IIKKY theJry n..c hori1.Oflial axis b 2k,r, ....here k.. u
tl'e "",..,vector on the Fenni sphere. (t1e Gennes.)

KONDO EFFECf

In dilute solid solutions of a magnetic ion in a nonmagnetic metal crystal

(such as Mil in Cu) the exchange (,.'Oupling between ti,e ion and the conduction
electrons has important con scqucl1(;cs. 111c (:onduc(ioll electron gas is magne-
tized in the \licinity of the magne tic ion, with the spatial dependence shown in
Fig. 17. This magnetization l,.'aUSCS an indirect exchange interaction\! bc hvccn
two magnetic ions, because a second ion percei\les the magncti7~1.lioll induced
by th e first iOIl. The interaction, known as the Friedel or HKKY interaction,
also plays a role in the magnetic spin orde r of the rare-earth metals, where the
SIlins of the 4fion cores are coupled together by the magnetization induced in
the conduclion elcctron gas.
A consequence of the magnetic ion-conduction electron interaction is the
Kondo effect. A minimum in the elt.'Ctrica\ resistivity-temperature curve of
dilute magnetic alloys at low temperatures has been observed in alluys of Cu,
Ag, Au, Mg. Zn with Cr. Mil, and Fe as impuritic!>. among others.

tA re\~ ol indirect exchange interactionl in metals Is swen b) C K,ttel. Sohd stale ph)'ila
22, I (1968), a re\';ew olthe KondoefTect is given by J Kondo, -n1l'Ql')' of tI,1"le m~nctlC alloys, ~
ScJ1,d state ph)'liic$ 23. 184 (1969) and A J Hceger. -Lcocll],7.ed moments and nonmon.enis ,n
rnet~II thl' Kondo effi.>d. ~ Solid ~Iale p},y.:ics 23, 24!i (t OW). 1'hc notatIOn RKKr stands r.... Ruder-
man. Kmel, Kasu)'lI. and Yosid.o.
 Allo~ OW
"
"'" AuF.,
"00

.. 0198

""\
''''l
.. 0 Ill(;

, 0 02at~ F.,
"" ,
'i "" 'i,,
0 . 190

O.07S no ..
J
0006
"'"'
0.074 .,"
0

, ,
T. t.:
,
Figure 18 A l"OOlparison of e xperimental a lit! th<'Orei,rnl rcsult:s for the i,,(.~ ea~ of .,leo:;tTkai
,",-,sISti .-ily 01 1m, lem\Wrnlures on d,l ute alk>y. of iron In gold. The resistanoc minimum lies 10 the
righ t of th e figure, for the resistl>ilv i"Crt,as<>s at high k mpernlUl es bemusc d scattering of ek e-
tron~ by thermal phonons. 11,,) expc rllnc nt:s are due to D. K. C. MacDonald , W. B. ('e arson, and
I . M. Tem pleton, the th"Q'Y i~ by J. Kondo. An exact suh,tion was given by K Wil ""...

The oc'CU fTc ncc of a resistancc minim um is connected with tbe existence of
localized magnetic mome nts on the impurity atoms. \ Vhcrc a resislan<.'C mini-
mum is found, there is inevitably a local moment. Kondo showed tlmt the
anomalously h igh scattering probability of magne tic ions at low temperatu res is
aconsequence o f the d ynaillic nature of the sc;lttering by the exchange coupling
and of the sharp ness of the Felnli surtace at low temperatures. The tempera-
ture region in which the Kondo e ffect is important is shown in Fig. 18.
The central result is that the spin-dependen t contribution to the resistivity

[ 3'1 _] _
"
P.p,,,=CPM 1+~ l n1 =cPo-cP1 In 1, (12)

whe re ] is the exchange energy; z the nu mber of nearest neighbors, c the

concentration; and PM is a measure of the strength of the exchange scattering.
630

We sec that the spin resist i\"ity increases toward 10\\ te mperatures if] is nega-
ti ve. If the phonon con trih ution to the e lectrical res istivity goes as '1 6 ill the
region of interest and if the resistivities arc addHh"c. then the total resistivity
has the form
p= a1Q + CPo - cp, In 'f (13)
with a minimum at
dp/dT::: 5aT~ - cPllT = 0 , (14)

\\ hence
(15)

'I11C te mperat ure at which the resistivity is a m in imum varies as the o ne-fifth
power orthe concentration orthe magnetic impurity, in agreeme nt with experi-
ment at least for Fe in Cu.

Problems
I. Sliper-iattice lilies ill Cuy\u. CuaAu alloy (75% Cu, 2.'5% Au) lIAS ,In ortlered slale
below 400"C, in Wllich the gu l<! aloms ocellpy the 000 positions 31ullilc Ct.lpper aloms
the B o, JOI. aml OU posi lit.lllS III a filCC·.;cnICre{! cubic lattice. Give tile indices of
the new x-ray rcflections that appear wi,en ti le alloy goes from the diso rtlered to the
un!crt.>t! state. List all uew rdlections with indices ;§:2.

2. Conjigm'(Iliomd heal capacify. Derive an cllpression in tcnllS of P(1) for tllC he.!t
capacity associatetl with unlcrlrlisc nler effects ill an .AB alloy. [The entropy (8) is
callctlthe <Xlnfigumtional c ulropy or e ntropy of mixing.!

References
H. Ehrenreich and L. M Seh",:.. t", "Electronic Stn.cture of Al lo\"~," Solid state ph y~ics 3 1, ISO
(tfITfi).
V. L. Moruni, J. F. Janak, and A. n. Williams. Calculated electronic proper1ies of metab, Per-
W'1ll0l1., 1978.
F. n.;.;enbeTger, Fundamentals of crystal grou;th I, Spring" r, 19S1. Discussion of phase diagrams.
D. de Fontaine, "Configurational Ihennudynalllics of solid solution s.~ Solid .Iate ph~.cs 34, 73
(19ig).
J. Kondo, 'Theory of dil ule magnetic alloys." Solid state phY~K;S 23, 1R3 (1968).
C. Klllel , " Ind ired exchange mteract ion in metals:' Sohd . Iale phy~i<:s 22, I (1968).
 APPENDIX A: TEMPEHATURE DEI)E"'DENCE OF THE
REFLECTIO", LINES

w co

. .. I came to the 0011c1u8ion that the /i/lUrpness of the

interference lines would not liufJer but tilUt tllCk ill-
tensity S/lOulrl (limillisll IVith illcreasillg (mgle of 8cat-
tering. the more 80 the higher the temperature.
P. Dcbyc

As thc tcmperature of the crystal is increased, the intensity of the Bragg-

reflected beams decreases. but the angular width ofth c reflected li ne docs not
change. Experimcntal intensitics for aluminum are shown ill Fig. L It is sur-
prising that we c:m get a sharp x-ray reflcction from atoms undcrgoing largc
amplitude random thermal motion, with instantancous nearest-neighbor spac-
ings differing br to percent at room temperature. Before the Laue expcrimcnt
,
9

,"
..I. - - .. - 110(\\

,
, '~~~. '--. ,
r . . --. (400)

, " '. ~)
.'.,
, ,
'"" "' "
"-
,
, "
" ,
" r--~)
'. "
,
\. (10,00)

\ "-
• .00 - --
"" 200
T.to K
Figure I The depend<-'Ill"e of iolen.it\, on lempernlure rOT Ihe (/,00 ) x-rnr refle<'lions of alu minum.
Refl<-'CIion$ (1,00) wi lh h odd are fOTb,dden fOT an fcc st ruchore. (Mtcr R. .\1. 'icklow ~nd R A.
Young)
.,
was done. but wilen the proposal \\~\S discIIssC(P ill a coffee h".,5(; in Munkh ,
tile objection was made that the instantaneous posit ions of the atoms ill a cl")'5lal
at room tcmpcruhJre are (If from a regular periodic array, because of the large
thermal fluctuation. Therefore, the argument went, OIlC should 1I0 t CKpcct a
well-defined diffracted beam.
But such a beam is found . The rcaSOn was given by Dcbyc. Consider the
rndiatioll amplitude scattered by a crystal: let the IXlsilion of the atom nom i-
nally at fJ contain a Icnn II(!) fluctuating in time: r et) = fJ + net). We suppose
cal;h atom fluctuates independently about its own equilibrium position. 2 'Olen
tbe thennal average of the structure fador (2.43) contains terms
jj cxp{- iC' rJ){cxp(-iC ' u)} , (1)
whe re ( ... ) denotes thermal avemge. The series expansion of the expone ntial
;,

(exp( - iC ' u )} = 1 - i(G' u) - !«C· U)2) + .. (2)

But (C ' u) = 0, because u is a random thermal displacement ulll..'OfTc\ated with
the direction of G. Further,
«C 'ul) = C 2(1I2)(COS 20) = !(II~G2 .
The fad or i arises as the gcometrical average of cos 2 0 over a sphere .
The fUllction
(3)
has the same scries expansion as (2) for the first two terms shown here. For a
harmonic oscillator allte n llS in the series (2) and (3) can be shown to be idcnti-
cal. Then the scatte red intensity, which is the square of the amplitude, is
(4)
where 10 is the scattered intensity from the rigid lattice . The exponential r.\ctor
is the Debye-Waller factor.
lIere (1/ 2) is the mean squarc displacement of all atom. The thcrmal avcr-
age pote ntial e nergy (U) of a classical harmonic OSCillator in three dimensions is
~k/jT, whence

(5)
where C is the force constan t, M is the ma."s of an atom, and w is the frequency
of tile oscillator. \Ve have used the result w2 = CIM. 1111)s the sl;attcred inte n-
sity is

(6)

'I'. I' EWllld . pr",,,I f' t"Omm" "irnl,on .

2·11". is Ihc Elml"'n model of" solid. ,I i, nul a "cry \!:ood mOOel "I low lemperlll"res. L"I it
works w{'lI al h igh tcmperJlures. 1-'"r a gc'l<'l"lIllrca lmcnl of S<."3tlcring by Ih~ rm",l nnc!"ati""" so..,
QTS . ChapleT 20.
 Appendix 633

whcre Ilkl arc the Jiccs of the rcciprocallaUice \'cdor C. 111is classical result
is a good approximation at high temperatures.
For quantum oscillators (1/ 2) docs nol vanisil even at T "" 0; there is zero-
pOint motion. On the independent harmonic oscillator model the ....ern-point
energy is i llw; this is the energy of a three-dimensional quantum harmonic
oscillator in its ground state refcrred to the classical ene rgy of the same oscilla-
tor at rest. ilalf of the oscillator energy is potential encrgy, so that in the
ground state
(U) "" iMw2(1/ 2) """ !hw ; (7)
whencc, by (4),
(8)
at absolute zero. If G ", 109 I,
cm - w = lOH 5 - 1,
and AI = 10- 22
g, the argu-
ment of tile exponential is approximately 0.1, so that IIIo "'" 0.9. At absolute
zero 90 percent of the beam is elastically scattcred and 10 percent is inelastic-
ally scattered.
We see from (6) and from Fig. 1 that the intensity of the diffracted line
decreases, but not catastrophically, as thc tempcrature is increased. Reflection s
of low G are affected less than reflections of high C. Thc intensity we have
ca1culate<1 is that of the w he rent diffraction or the elastic scatb:ring in the
well-defined Bragg directions. The intensity lost from these directions is the
inelastic scattering and appears as a diffuse background. In inelastic scattering
the x-ray photon causes the excitation or de-excitation of a Ifl ttk-e vibration , and
the photon changes direction and energy.
At a given temperature the Debye-Waller factor of a diffraction line de-
creases with an increase in the magnitude of the reciprocal lattk-e vector C
associate<l with the reflection. Thc larger Icl is, thc weaker the reflection at
high temperatures. The te mperature dependence of the reflected intensity for
thc (/,00) reflections of fl luminUin is shown in Fig. 1. n le theory we have
worked out here for x-ray refle<.1:ion applies equflily well to neutron diffraction
and to the Mossbaucr effect. the recoilless c mission of gamma rays by nuclci
bound in c rystals.
X-rays alii be absorbed in a crystfll also by the inelastic processes of photo-
ionization 9f electrons find Compton scattering. In the photoclTect tlu: x-ray
photon is absorbed and an electron is ejected from an atoll1. In the Compton
effect the photon is scattcred inelastically hy an electron; the photon loses
energy and the electron is cjcdcd from an atom. Thc depth of pclH:tration3 of
the x-ray beam depends on the solid and On the photon energy. but- 1 CIl1 is
typical. A diffracted beam in Bragg reflectionillay remOve the cnergy in a Inuch
shorter distance. perllaps 10- 3 cm in an ideal crystal.

"See w. I ldll er, Q"on/"", II,wry of ran;aUQI1. 3rd cd .. O.ford. 1954, p. 223.
 -
Al'I)El'iOIX D: EWALD CAl.CULATION OF LUTICE SUMS

11)(: prubiclil is to calculate the dectrrn.t,ltic potential c\p,.. ricmx'(l by one

iOIlin the presence of all the other ions in the crystal. We ronsidcr a lattic!::
made up of iUlls with positiv\:: 01' nC'/!;1tivc ch'lI'gcs and sh.J11 aSS lIIll C that t ile ions
arc spherical.
We (:olllpute the total potential IP "" (,0 \ + IP2 .It ,Ill ion ,IS the sum tlf two
(Iislinct but rclatctl potentials. The potenti,ll /P I is th.lt of a structure' \.,.i th a
Galls~ i a n distnbuliou of charge situated at each iOll site, with sigll s the same as
those of the real iOlls . Accordiug to the definition of the ~ f adelungconst.l l lt , the
chnrgc dbtl illillion on the reference point is not considered to [,'onlributc 10 the
potential IPI or IP2 (Fig. 1.1). We therefore Citlculatc the potential IjI I as the
dilTcrcll{.'('

'PI = IP.. - fIJI.

...
V ~--v=j . ,- - -
, ,,

.,

1101

Fillun~ 1 (..) Ch.lr~e J"tnh"t.on """ I lOr t'OlIlpubng potl·llh.ll.p, . the poll'ut ...1 .p" is ,,,,,,pul<><1(11
",d",les the d."h,·d ('n ....·•· .. t ti,,, o:f.·,.,.,,'C 1'",,,1). ,,1,,1,· ,.", 's Ihe IlOIenli." or Ihe .b,hecl ,."''''
~lonl' (b ) Ch.ll)(t" d"I.,h•• lloo Il" 1")Il',, liJI <;2 The n·f(o'l'TII'C point IS d,-,,,ot.-.I hy an \
I

of two potentials, CPt! being the potential of a continuous series of Gaussi'lIl

distributions .lI1d l/ib being the potential of tIle single Gaussian distribution on
the rderencc point.
111C potential (/!2 is that of .1 lattic."C of point charges with an additional
Gaussian distribution of opposite sign !>upcrposcd upon the point charges
(Fig. IL).
TIle point of splitting the problem itlto the two parts ip l and 1/>2 is th.d by a
suitable choice of the parameter determining the width of each Gaussian peak
we ~an get very good convergence of both parts J.t the same lime. The Gaussian
distributions drop out completely on taking the SUIll of the separate ch.ugc
distributions giving rise to 'PI and 'P2, so that the value of the total potenti,,) II'
is independent of the width parameter, but the rapidity of cOlwelgence dc-
pcnds on the valuc chosen for the par,lIneter.
\Vc calculatc first thc potcntial 'fIa or a c.:QntiuILOus Gaussi.m di!:;tribution .
We expand f/!a ,md the chargc dcnsity P in Fourier sCTics:

f/!4 = LG Cccxp(iC . r) (I )

p = L Pccxp(iG . r)
G
(2)

where G is 211 timcs a vector in the reciprocallatlic..-e. The Poisson equation is

V2f/!a = -411P •

0'

so that
(3)
We suppose in finding Pc that there is associated with each lattic..-e point of
the Bravais ldtlic.:e a basis contaiuing ions of chargc q, at positions r, Tclative to
the lattic..-e point. Each ion pOint is therefoTe the C€nter of a Gaussian charge
distTiuution of density

where the factor in front of the exponential cnsll r~s that the totdl ch,uge associ-
ated with the ion is q,; the range parameter 1] is to be choscn judiciously to
ensure rapid c.:ollveTgence of the final result (6), which is in valuc independent
of 1].
We would normally e\'aluate Pc by multiplying both sidcs of (2) by
exp( - iG . r) and in tegrating over the volume tJ. of one cell, so that the chdrge
distrib\ltion tQ be considered is that originating on the ion points within the c.:ell
 (
and also thaI of the tails of the distributions origill:ltillg in a" .Aher C'€lls. It is
ca~y to see, however. thai the integral of the total charge density time~
c.xp! -(iG' r») over a single cell is equal to the integra l of the charge density
originating in a single cell times expi - (iG' r)] over all sl""ce.
\Ve have therefore

Pc f
00,·
exp{iG' r)cxp(- iG· r) (Tr '" Pef).
....II

= J~
.11
q,(TJln)312 exp! -~r - T,)2Jexp(- iG ' r) dT

..-
'nlis expression is readily evaluated :

{Jcl:l = ~ q,exp(- iC' T,) {T1.,,)31~

.,J expi -(iG . ( + 'l'jt'~] (It'

"'""
= (2:, q,exp(- iC' T,») exp{-C 2/411) = S(C)exp( - C 2/4Tj)
wlle.'c S(C ) = 2:, q,exp( - iC' T,) is just the stnlcturc f.'lctor (Chapler 2) in up--
propriatc nllils. Using (I ) and (3),

(4)

At the origin r = 0 we have

111c potential IpIJ at the refere nce ion pOint i due to the central Gaussian
distribution is

and so

111e potential IP2 is to be c\'aluated at the refere nce IlOint, and it differs
fmlll zero because other ions have the tails of their Gaussian disllibutions-over-
 App""di;r 637

(
lapping the refen. •. ~e point. 111e potential is due to three mntributions frOI11
each ion point:

q,[!.. - !..
r,rlO
r p(r) elr - loor ,Per)
r
(lr] ,

where the terms are from the point charge, from the part of the Gaussian
distribution lying inside a sphere of radius r, about the lth ion point , and from
that ,}'1rt lying outside the sphere, respectively. On substituting fo r p(r) and
cmTying out e1ementalY manipulations, we have

(5)

where

"'(x) = ('1hrll2) J'" eAp(-il) cis

•
Finally,

!f(i) = 411 2: S(C)C -

tJ. (;
2
eAp( - C 2/4T) - 2q;(T)/1T) If2 + L
1
!l!.... f(Tt2 r,)
n
(6)

is the desired total potential of the reference ion i in the fi eld of all the other
iollS in the crystal. In the application of the Ewald method the Irick is to choose
T) such that both sums in (6) converge rapidl y.

£wtdd.-Kornfeld Method for Lattice Sums for Di}}o/e Arrays

Komfcld extended the Ewald method to dipolar and q uadrupolar anuys.
We discuss here the field of a (lipoic array at a point wh ich is !lot a lattice point.
According to (4) and (5) the potential at a point r in a lattice of positive unit
point charges is

If' = (41T1tJ.) 2: S(G)C - 2

eAp[iC . r - C 2/4111 + 2: F(v:;] r,)lr, [l)
G 1

where rl is the distance from r to the lattice pOint I.

TIle first term on the right gives the potential of the charge distribution
P = (TjI11'f'2 exp(- T)r 2) abou t each lattice pOint. Bya well-known relation in
c1ectrostlltics we obtain the potent ial of an alTay of unit dipoles pointing in the z
direction by taking - tllelz of the above poten tial. The te rm under discussiun
contributes

- (47ri/6.) 2:c S(C)(C j C 2

) expiiC . r - C 2/4111 ,

and the z compon{'nt of the elect ric field f!"Om this term is E~ = .-r-If/.lZ2, or
 ..
- (411fd) L S (GXG ~/G2) ("~p(iC . r - c 2/4111 (8)
c
Tile second term on the right of (7) aft c r olle diffe rentiation gi\'CS

- L, z,{(F{v;jr,l/ri') + ('lJrlXT)l1T) 112 e ..p(-1)rl)]

.\IId the.: componen t of this pM ! of the field is

L: {z?H3F( v;:jr/Y I'I~ + (6Ir/)(TI,")II'1. exp(-l'),l)

(ll)
+ (4hlXrlJw)lf2 cxp(- "1n2JI - !(F(\hjI) lrl)
+ (2Jrl~ r"7T) 112 C\I>( - W,2))}

111C tolal £~ is given by the sum of (8) and (9). 111c effe<'ls of any lI umher of
lattiCf's lIlay he added.

APPENDIX C: QUA f\'TlZATlON OF ELASTIC WAVES: PIIONONS

PhOllons were introduced ill Chapter 4 as Ql4Ultizcd el.~ti(' w.wes. Ilowdo

we quantize an elastic w,1\'c? As a simple modd of phonons in a (,1'Y~t.\1, COn-
side r the \ ib ratio ns of a lincar lattice of particlcs connf..'t.1cd by ~p r jngs . We cun
quantize the particle lIlotion exactly as (01";\ hannon ic o.'.Cillalor or set of ('O U-
pled harmonic oscillators. To do this wc make a transformation from Jl.1rtidc
coordinOltcs to phonon coordinates, also called wavc coordilllltcs beCUIISC they
rcpresent a tnlVcling wave.
Let N partides of mass M be con nected by springs offoree const,Lnt C and
length o. To fix the boundary conditions, let the particles form a circtll,lr ring.
We consider the transverse displacements of the particles out of the plane of
the ri n~. The displacement of particle s is q. and its momentL1m is P.' The
hamiltonian of the syste m is

( I)

TIle Il<Imiitonian of a harmonic oscillator is

1 1
H~ - ,Jl+ - Cx2 (2)
Ut 2
I

and the e nergy e igenvalues are, where n = 0, J. 2, 3, ...

(3)

The ci~cnvalue pro!Jlf'm is also CKat:l ly solvable for a chain with the different
hamiltonian (I) .
To solve (1) we make a Fourier trollsfonnation from the coordinates PA. q.
to the coordinates Ph Qk. which are known as phonon coordinates.

P/101Ion Coordinate.
l11C tmnslormation from the particle coordinates q. 10 the phonon coordi-
nates Qk is used in all periodic lattice problems. We let

q. = N - I12
L, Qk exp(ikstl) (4)

consbtc nl with the inverse tnllll.formalion

(5)

llerc the N values orlbe wavcn.,.clor k allowed by the pe riodic boundary condi-
tioll If. = q.+N are given

k = 2-mllNa ; n = O. :t i, ±2, .. ", ± (+N - 1). ~ N. (6)

\Ve need the transformation from the particle momentum TJ. to the mo-
mentum Pk tlmt is canonically conjugate to the coordinate Qk. The transform,l-
tion is

r. = ",-iii. L, PIr. exp(- iksa) ; PIr. = N- 112

.
L r. exp(iksal . (7)

111is is not quite what one would obtain by the naive substitution of r for (/ and
p for Q in (4) and (5), because k and - k have been intcrchan ~cd betweCit (4)
and (7).
We verify that our choice of PIr. and QIr. sathl'ies the quanlUm commutation
relation for canonical variables. We form the commutator

[Q~,p,,·l = N- 1 [2: q .. cxp(-ikral, 2: 1). eXP{ik'sal]

• •
= N- 1
.
2:, 2: (q""V.] exp( - i(kr - k's)a ) (8)
Because the operator:> q, " are conjugate, they satisfy the commutat ion relation
[q~,,'~J = ihD(r,s) (9)
where 8(,-,8) is the Kronecker delta symbol.
Thus (8) becomes
IQ"p,,·] :: N - ' ill L, exp(-i{k - k')ral :: ih5{k,k') , (10)

so tim! Q~, P" also are conjugr.te variables. Here we have evruuated the smnma·
tion as

L, exp[- i(k - k')ral = L, exp(- i217{n - ,,')rlNI

(II )
= Nfi,n,n') = NO(k,k') ,
where we have used (6) and a standard rcsult for the finite selies ill (I I).
We carry ou t the transfolllH\lions (1) and (4) on the hamiltonian (I), lind
make use of the su mmation (I I):

( 12)

x exp(ik'stl){exp(ik'u) - I ] :c 2 L QJ9- ID - cos ko) . (13)

•
Thlls the hamiltonian (I) becomes, in phonon coordinates,

(14)

If we int roducc the symbol WI: defined by

Wl !!I! (2C/M) 112 (I - cos ka)' 12 (I5)
we have the phonon hamiltoniHIl ill the f01111

(16)
 (
111e e<jualioll Ot molioll of the phonon (;oonlillatc Opcmtor Qt is fount! h)
the standard prescription of {tuan lum lIlecha:lli(~ ;

ifliJ" = (Qk, H I '" ihP _,fAt , ( 17)

with H given by (14). Further, using the commutator (17),
ih(h = [(h,Hj = fl.}" IlP_l,lf j = ihwk 2Qk , (18)
so that

(19)

TIlis is the equation of mo tion of a harmonic oscillator with the fre<luc-ncy Wk _

The energy e igenvalues of a quan tum IHl.lmonic oscillator are

E, = (Ilk + ~ )'UUl . (20)

where the quant.un nlunber Ilk '" 0, J, 2, .... 111(' energy of the entire 5),5-
!elll of Jill phonon ... is

(21)

11115 result demonslr.tle!i the qllanti7..1tion of the enC'I"g)' of elastic \\~I\es on a

line.

Creation 011(1 Allllillilalioli Operators

It is hel pful in "dwnced wo rk 10 tr,msfonl1 Ihf' phonon hal11iltoni.lIl (16)
into the form of a sel of harmonic oscillatOrs;

(22)

Here at .
al arr harmonic Oscillator operators, also ('ailed creat ion and dc-stnl('-
lion operators or boson operators. lne transfo rmation is dcrived lx-Iow .
111e boson creation operator 11 + whieh "creates a phonon" is defined b\' the
propert)·

(23)
when adillg on a harmonic oscillator stale of quantulll number I I, and the boson
annihilation operator (I which "destroys a phonon" is defined by th t- In'operty
(2·1)
 II f" liow!> thJ.1
(25)

~n Ih,11 III) is an ci~cnslalc of Ihe opcrator (1 + a with Ihe intc~r.,1 ei~cll v"lllc II,
(:'. I\lCO t!IC quantum L111mhc r or 0('('UI)'1IlCY of Ih~' o~d ll.llor. \Vh c lI Ihe phonon
IIl(Kle k i.\ ill the Cil!;eml.lh· 1.Ibelcd by 11k, we nM)' S.l }' t hat thNC Me III phono ll!>
III the Iliodl', The e ige nvalues o f (22) Me Ii =- L (IIA + ! )/IWl, in a~rt.'t' llll· llt witll
(21).

(/(/ ~ III) = a{II + 1)112111 + J) = (II + 1)111) • (26)

the lUlllmutalor of Ihc boson w.we operJ.tors at. and ale !>a ti sfil.' ~ the Ielatio n
la . (I + I " (ItI ~- tl · a = l (27)

We ,nil h,l\c In PI"O\'C 1IIoIIIhe h,lInilto\li.1II ( 16) call he cK p rcs~ed ,IS {I 9) ill
1I.:rnh of II II' phOl101I npt..'nLlors at , (I~ , 111i ~ can be d olU: Ii)' the tr,lIisfOflllillio n
(It = {2h)- 112{('\/WA)1 ~ () ,, - i(Mwd ' IIII',,1 (28)
('J. = (2ft) I12!( MW1)"I{h + i( Mwi) If2P_d (29)

'n le inverse rcl.llions arc

Q" = (1J2Mwi.) I {~ak + a+ Ie) (30)

P i "" i(hMwJ.12) ' 'l(.at - fLd (3 1)

By ( I), (5), ,.IIlll (29) Ihe p.U"lidl' l"lO!>ilion o pe r,ltor bccomt.' ~

q~ = L, W2N~Jwdll!l(lk c 'l>(ih) + tit c'CI>( - iks)] (32)

111is (~qll;Jtion rel.ltl'~ lilt! p:\I tide ui'pia<.'t! lI1cnl operator 10 the phonon creatioll
and ,lIlll lhiiatiOIi o pt'r;\tors .
To ohl,lin (29) from (28). wc LIse Ihe pmperlic!>
(33)

\\ lli('h fe llow from (S) ,mel (7) by use o flhe ql Lanlu m nll,1.:hanK.-aII"l"qll ircmcn l
Ih,11 q. IIl1d 11* he he lillitiJ.1l opcr.tlnrs:
If. = tl ~ , 11. = r,' (34)

'niC II(28) fo\lO\\~ fnll1\ the tr.U1 , IOl"1n:ll i()n~ (·t ), (5), and (7). Wt' \e nfv t h.lt the
l'OIlHLlLltati()1l re latio n (33) is s,ltisfic li by thl' opera tors defined b y (2H) lind (29):
[a.t.,fI'l "" (2ftr l ( M w~1QA,Q kl - i[Qk, Pd + i[P ~,Q - d
(.l5)
+ [P k,Pd/.HwJ
By lise oflQI",Pd "" ili&,k,k') r"()I1l (10) we h,lVc
IUI;,(I;.] "" [j(k, k ') (36)
It n. . lI1ains to show th;lt the vcr~ iom of (16) ,\1111 (22) or the phonoll hllmih t)-
ni,1Il lire idl't1t ical. \Ve note thdt Wk = W _ l rrom (15), .ltld \\l;' form

This CJ;hihits the e<luivall'nce of the two l·xpn.·s~ion~ (1-1) ,\lid (22) ror H. We
identify WI: = (2C1/11 )II2(1 - cos ka)lfl in (15) with the ddS~icdlfrl'IJ'll'nl'v ofth(,
oscIllator mode of W,\\'l'VCl·tor k.

APPENDIX D: FERMI-DIRAC DISTRIBUTION FUl'iCfION

111e Fermi-Dimc dbtrilmtion function' may be derived in ~cvera l ~tcps h y

lise of a mo{lcm ,Ipproach to statistical mechanics. \Vc outli ne the argument
he re. The notation is s uch that l'ollventional <'nh'op), S is related to the hllldll-
mental entropy u by S "" k au, and the kelvin tcmperature T is reluted to the
fundamental te rnpe r11turC 'j by 'j = kaT, where ka is the Bohzrnanu CQn~t,mt
with the v11llle J. 3OO6fi X 10- 23 J K- '.
TIle leading quantities are the entropy, the tClllpe r,lIurc, the 1301tzm:1Il1i
f.1ctor, the chemkal po tentidl, the Gi hhs !:Letor, and the distnbution functi om.
The enllopy me,lSurcs the number of quantum state'! lIcces~ible to 01 ~ystc m . A
d osed system miJUlt be in lIny of thcse (JUiUltmn ~t .ltcs and (we assume) wilh
equal prornlbility. Thc fundamcllta l assumption is that {j1l,l\ltlllll 'It,ltes arc
eithe r act:essihle or in acccs~ ible to the Systl'llI, and the S\stCIII is clJllollly likely
to bc ill anyone lIcccssiblc state as in any othe r <Icc(~ssi hl(' stah'. Given ~
;ll"Cessible states, the c llllopy is defined as a = log g. The entropy thus ddined
will be a runct ion of the e nergy U, the number of pllrtjdt~s N, and the vol ume V
of the s)'ste m.
\ Vhen two S\"stem~ , C<Jell or ~ pecified enl' r~v . ale bro~l~ht into tlwrlllal
cont;lct they m:ly tr,lIlsfer ene rgy; their total e ne rgy lemain s COtlstllnt , h ut thc
const raints on their individual energies art! hfte<1 A tr<m~re r of e ne rgy in one
din..'<.'I:ioll, or perhaps in the other, may increase the product gil!.:! that 1Ilc,lsures
the number of IIccessihle ~ I atcs of thc (."omhincd s)Stems. Wildt we e;11I the
fundamental assumption biases the outcome in favor of tll.lt allol"oltion of thc
totlll energy that m,nimi.ws the numbe r of.lcccssihie ~ta l e~ : mort: IS beurr. und
more likely. This statement is the kernel of the ldw of illt:rease of entropy,

" nIlO .'ppt'll. ,,,\1""5 dowl) tt,e ,,,trotl.... t,,,,, to ( . K,ud "nd II """""wr. fllC ...""II'I".. '.~
2,,<l .~I . Fr.·emall . 19HO.
'"'
which is the general expression of the second law of Ihc n mlUrnamics.
\Ve have brought two systems into Ihc l1Jlll1 contact so that they may lrans-
fer energy. \Vha! is the most probable Oillco m e of the eIH.·O\Hlle l"'~ One system
will gain energy at the expense of the other, 311(1 meanwhile the lotal entropy of
the Iwo systems will inc rease. Eventllally the entropy will reach a maxim Ulil for
the given total energy. II is not difficult to show that the maximum is attained
when the vallie of (ilulilU)NY lor one syste m is equal to the vullle of the same
qll<'1ntity fo r the s{.'COnd system . This equality propc I1 y for two systems in ther-
mal COnl act is the property we expect of the temperature. Accordingly, the
fundanl cntai te mperature 7 is defin ed by the relation

I (ik')
~ ... au N.V
(I)

TIle use of l IT assures that energy will flow from high 7 to low 7; 110 more
complicated relation is needed.
Now COnside r a very simple example of the Boltzmann faclo r. Let a small
syste m with only two slates, one at e nergy 0 and olle at energy f:, be placed in
the rmal contacl with a large system that we call the reservoi r. The totlll e ne rgy
of the combi ned systems is Vo; whe n the small system is in the state of encrgy
0, the resen'oir has energy Vo and will have l,iVo) slates accessible to it. \Vhen
the small system is in the state of e ne rgy E, the resen'oir will havc encrgy
Vo - £ and will have g(Vo - E) states accessible to it. By the fundamental
ass umption, the ratio ofthc p robability offin{ling the s mall system with encrgy
E to the p robability of finding it with energy 0 is

PtE) l,iVo - E) ex p(u(Uo - £)1

-- = = (2)
P(O) g(Uo) cxp[u(Uo)]
The rescrvoir entropy u may be expanded in a 1ilylo r se ries;
u(Uo - E) "'" u (Uo) - E(ilafiJUo) = a-(Uo) - EIT , (3)

by the definition ( I ) of the te mperat ure. Higher on::ler te rms in the expansion
may be dropped. CllIlccllation of the te rm exp!u{Uo)], which ()(:("urs in the
Ilume rator and denominator of (2) aftc r the sl.lbstit l.ltion of (3), leaves us with
P{£)/P(O) = exp( - £IT) (4)

11lis is Boltzmann 's result. 10 show its 11SC, we calculate the thennal aver-
age cnergy (E) of the two-state system in thennal contact with a reservoir lit
te mperature 1";

(f:) = 2: £,1'(£,) = o· P{O) + EP(E) = J

€ exp{- €fT)
+ ~p( £17)
(5)
 App""dix 645

where we have imposed the normali7.a.tion condition on the sum of the proba-
bilities:
P(O) +'P(I:) = I (6)
The argument can be genemlized immedilltely to find the average energy of a
hm'mollic oscillator at tempel-:ltme 'T, as in the Planck Ill\\,.
TIle most important extension of the theolY is to systems that can transfer
~1I1ides as well liS e ne rgy with the reselvoir. For two systems in diffusive and
thermal contact, the entropy \\ill be a maximum with respel:t to the trllnsfer of
particles ~ well as to the tmnsfer of energy. Not only must {ilolilU)N. v be equal
for the two systems, but (ila lilX){'.v must also be equal, where N refers to the
number of pal1icles of a given species. 111e new equality condition is the OCCll-
sion for the introduction 2 of the chemiC'll) potential p..

- ~ = (:)u.v (7)

For two systems in thermal and diffusive l'ontact, "11 = "12 and P. I = p.').. The
sign in (7) is chosen to ensure that the direct ion of particle flow is from high
chemical potential to low chemical potential <1S equilibrium is approal.·hed.
The Gibbs facto r is an extension of the Boit-..;mann factor (4) and allows us
to treat systems that call transfl'r particles. The simplest e~amp l e is a system
with two states, one with 0 particles and 0 energy. and one with I pllrticle and
energy E. The system is in con taet with 11 reservoir at tempemture T lind chemi-
cal potentbl p.. We e~tend (3) for the reservoir e ntl"Opy:
a(Uo - £;No - I) :; u {Uo;No) - l:(iJutiJUo) - I . (iJutiJNo)
(8)
= a(Uo;No) - EI'T + ph
By analogy with (4), we hllve the Gibbs factor
P(1,I:)JI'(O,O) = exp{{p. - 1:)1"11 (9)

for the rat iO of the p robability that the system is occupied by 1 particle at
energy £ to the p robabi lity that the system is unoccllpied, with energy O. The
result (9) after nOimalization is readily e~p ressed liS

P( I ,.) ~ - " [-'-:-:--;-1-:-:- (10)

exp {E p.)h + I
This is the Fermi-Dirac disilihu tion fundion .

tl'P Chapter 5 has a. car.. rul tr~'atm~nl or the c" cmi~"3 1 potcnhal
 APPENDLX E: DER IVArION O F TH E dkldt EQUAnON

......
The simple and rigorous de rivation that follows is due to Kraemer. In
q ua ntum mechanics, fo r any operator A Vie have
d(A)ldt = (Vh)( [H ,A I) (1)
whe re If is the hamilto nian . See also C. L. Cook, American J. Phys. 55, 953
( 1987).
We let A he the lattice translation operator T de fined by
l'ff.x ) = f(x + a) • (2)
whe re a is a basis \cctor. hcre in one dimension. For a Bloch function
(3)

This result is us ually written for one band. but it holds even if 4JlJ; is a linear
combination of Bloch slales from any nu mber of bands. b ut having ti le identi"a l
wavevector k in thc reduced ~o ne sche me.
The crystal hamiltonian 110 commutes with the lattice translation operator
T. ~o that [lfo:n = o. If we add a unifo nn external fo rce F, then
11 = lfu - Fx , (4)

[H,T] = FaT . (5)

From (J) and (5),

d('I)/dt = (Uh XFa )('I ) (6)
F ro m (6) we fo nn
m'd('l)i<It = (iFalh)I('I) I' ;
(,/}d('/")ldt = -(iFwh)I('I)I'
O n add ition,
d[('I)I'ldt = 0 ('I)
This is the equation of a circle in the complex plane. The «JQrdinate axes in
the plane are the real and illlaginary parts of the eigenvalue cxp(ika). If (1) is
initially on the unit circle, it will rem,lin on the unit circle.
For 4JI's tlMt satisfy periodic boundary cond it ions, (1) can lie on the unit
circle only if ifiJ; is a single Bloch function Or a s u perpo~ition of Bloch fu nctiOns
from d iffe rent bands, bu t with the same reduced k.
As ('I) moves around the lin it circle , the waveveclor k changes exactly at
thc s.lI11e r,lte for the COirlPOIl Cllts of ifiJ; ill lllll).lllds. \Vith ('I) = exp(ika), we
have rrom (6) that
ia dfJelt = il'allr (8)

dklilt :: f'lfl , (9)

~111 e~ac( r~ult ,
This docs not mcal) thdt mtcrh.md nlixing (sul'h a~ Zencr tUl1llcling) dOt!s
not occur ullder the influenl'e or 'lpplicd electric fields, It just means tklt k
evolves at a constant r.lte ror eve.'y component or <I w.wc pad..ct, 111e result is
easily extended to three dimensions.

APPENDIX F: BOLTZMANN TKAi'\SI>OKT EQUATION

11le classical theory of transport processes is based on thc Boltzmann

tnunport equdtioll. We work ill thc six-dimensIonal spa('c or
G.lrtcsian coordi-
nates I" .lIld velocity v. The c1as~ical di~tribution runction /(r, v) is defined by the
rel.ltion
/(1", v)drJv = number or partides in drclv " ( 1)
The Boltzmann equation is derived by thc rol1owill~ argumcnt. \Ve con-
sider the effeet or a time displacemcnt lit 011 thc distribution runction. The
Liouville theorem or classical IIlCchanics te lls us that ir we rollow a volullIe
elcmellt along a Ilowline the distributioll is conse.ved:
1<t + dt ,r + dr.v + dv) - /(i ,r ,v) (2)

in the absence or collisions. With collisions

I (t + dt ,r + clr. v + elv) - I (t,r,v) = rit(of/ot )c-on"j",,, (3)
"111U S

(It(of/ot) + (Ir" grad r j + dv . grad v 1 = dt(ojlot )<:nll (4)

Let ex denote the acceleration dvlclt; then

I
af/at + v . grdd r j + a' grad ... j "" (ilf/iJt)..",lI I (5)

'n lis is the Boltzmann transport equation.

In many proble ms the collision te rm (iJf/iJl),-on mdY be tre.lted by the inlro-
duction of a relaxation time T,(r, vl, defined by the C(luatiol.l
(6)
Herc/o is the distribution funct ion ill thermal efluilihriulIl . [)< t confuse Tr
for relaxation time wilh T for temperature. Suppose that a nOIlC<juilibrium dis-
tribution of velocities is set up by external forces which are suddenly removed.
'J11C decay of the distribution towards equilibrium is then oblilined from (6) a~
iJ(j- f~ ~ _ f - fo
(7)

"
if we no te that iJii./dt = 0 by definition of the equ il1briuIll dis tribution . 111is
equat ion hlls the solution
(8)

It is1I0t excluded that Tc may be a function of r ;Ind v.

We combine (1), (5), and (6) 10 obtain the Boltzmann transport equation in
the relaxation time approximation:

of
at + a' gradyf + v' gradrf = (9)

In the steady state apal = 0 by defin ition .

Particle Viffll8ion
Consider all isothermal system with a gradient of the particle conccntra-
tion. The steady-state Boltzmann transport equation in the relaxation time ap-
proximation ba:omes
v..fid., ~ -(f - f"iTo • (10)
where the Ilonequilibrium distrib ution fun ction f varies along the ::t direction.
We may write (LO) to first order a~
(ll )
where we have replaced iJflib: by (lfofdx. We can iterllte to obtain higher order
solutions when desired. Thus the sCt,.'Ond order solution is
f2 = fo - V.Tcllf1Id::t = fo - v,T..dfold.r + CiT~d2frldx'l . (12)
The iteration may be used in the treatment of nonlinear elTects.

Classical lA$tribulion
Let fo be the distribu tion function in the classical limit:
fo = (.'Xp{(s.t. - f)fTJ . (13)

We are at liberty to take whatever normalization for the distribution function is

most colwellient 0. lse tile troillsport Ctluation i~ lin('ar in f and fo. \Ve (.."Olll
take the nOrmali1..i\tion as in (13) rather th:m as in (1). 11iCn
(ifo!c/x = (dJ;JdJJ-XtlpJtlx) :: <fo!TXdpJdx) , (14)
and the lint order solution (11) for the none<luilibrium distribution becomes
f :: fa - (v,Tdo/T){(fiJ-Idx) (15)
The particle flux density in the x direction is
j ,,' ::: I VJIJ(E)(/E , (16)

where IJ(E) is the density of electron states per unit volume per unit energy
range:
_ 1
D(E) - 27ft
2M .(~)'"
It'll
E , (17)
as in (6,20), Thus
jft":: I vJoD(E)dE - «(/pJdx) J (V/ 'TcIolr)D(E)cfE (18)
The firs t in tegral vanishes hecause v, is an odd function and fo is an even
functiou of v.' "nl is conlirms that the net particle flul( vauishes fOr the e(luilih..
riu m distribu tion fo. The second integnll will 1I0t vanish.
Bcfore evaluat ing the srt'Ond in tegral , we have an opportunity to Illilke usc
of what we may know abollt the velocity dependence of the n'lalation time '1".. .
Only for the sake of cl(ample we assume tha t 'T... is constallt, independent of
velocity; T... may then be taken out of the integral:
j,,' = -(dpJ(I:cKTdT ) J v. 2folJ(E)tlE ( 19)

The inteb'l1ll may be written a~

i J IffolJ(E)(fE = 3~1 J (;Mv2)J;.,J)(E)dE ::: nTiM (20)

because the integral is just the kinetic e ne rgy density i n'l" of the p articles. H ere
ffoD{E~/E ::: II is the conl'Cntralion. The particle fl ux density is
(21)
becausc
Jl. = 'T log 11 + (''Ollstant . (22)

TI1C result (21) is of the form of the diffusion equation with the diffusivity
(Z3)
Anotl1cr possible assumption abou t the rela.~at i on time is thai it IS Ill "
verscly proportional to tile velocity, a~ in T" = IIv, whe re the mcan free path f
is t;onstant. Inste.u] of (19) we have
j"' ::: - (dpJlh ){flT) J (v,P"/V)JoD(E)dE • (24)
650
-
;Iud now the inlcv;r,ll mav be wriltcll .IS
~ J VJuD(E}f/t: = 1nc , (25)
where C: is the ;l\Cr.tge speed. Thus
J.: = - j(li:lllr)(dJ.llclx) = - VC(llnldt) (26)

and th(,: difru~ivity is

(27)

Fermi-Virac Distribution
-nlc distrihution funchon ;,

(28)
fu = C)(p[(t: ,u)hJ + l
To form ll/oldx as jll ( L.J) we need the deriv,lli\'e (lfuMIA-' We ,uguc hduw IIMI
rl/ulflJL "" .s(E - ",,) • (29)
,\I low temperatures T « ",. H<.'rc 5 is the Din«' delta [unci iOn, which h.tS the

prOperl)' for a general fum-lion f eE) that

!''•' F(t:).s(t: - ILlciE ::: F(p.) . (30)

Now consider the integral Jt f (t:XdJo1d,u}tlE. Allow temperatures (l/uMp.. is \cry

large for E "" JL and is 5111,111 elsewhere. Unless the fUI1CtitJ I1 f eE) is vcr~ r.lpiJly
v,u'ring IlcaT ,IL we m... y t.,ke f eE) oUbkle the illle~ral . with the value F(p.):

I'" P(E)(dfofdJL )tlE '" f(j..t)

n
r"
0
(dJo/dp.)dE ~ - f (JLl i'" (dfolclE)dE
(31)
~ - F(~) If"(')lo ~ FIJ>I/.,O) •
where we have Iised dfoMJL ". - t1foltll;'. We h.we (ilso UM!dfo = 0 lOr l;' "" x. At
low temperiltllres f(0) "" I; thus the riKht-hand ~ id e of (31) is just F(JL), consist-
ent with the delta fun ction IIpproximation. Thlls
(/fulrix = ll{l;' - JL)diJ-lclx . (32)
The particle flux density is, from (16),
],: = -(diJ-ldx)Tc f l'.z8(E JL)D(l;')clf, (3:l)
where Tc is the relaxation time at the surface f "" JL of the FClmi sphere. The
integral has the value
(34)

•..
by use of D(p,) = 3"12(',, at absolute zero, \\here E" D imv~ defrm.'s the velocity
"F on the Fermi ~Ulface" Thus
In'' = - (m.Jm)!IJAldx . (35)
At ahsolute zero fA.(O) = (h 2/2 m)(3~IIF. whence
dJAldx = {lW/2m)(31T 2 F,,, ' fJ }dnldx
(36)
= ~(E,..In)dllldx ,

so that (33) becollies

1,/ = - (2TJ3/11)E" tinldx = ~!Vl-"2T" c/,,/dx " (37)

The diffllsivit y is the coefficient of dnldx:

Dn = ~ViT" (38)
closely simiiOlf in form to the result (23) for the classical distribution of veloci-
ties" In (38) the ."elaxation time is to be taken at the Fermi energy.
\Ve see we can solve transpo rt pTOblems where the Fen ni-LJi rac u istlih u-
tion .Ipplics, as in metals. as easily a~ where the cl a~sical approximation applies"

Electrical Conductivity
Tlle isothermal electrical conductivity u follows from the n !sult for the
particle dilfusivity whe n we multiply the particle flux density by thc particle
charge q and replace the grad ie nt tlJAldx of the clJemical [Xltcntial by thc grud i·
ent qtllf>ldx "" ~qEz of the extemal potential, where E." is the x component of
the electric field intensity" The electric currcnt den~ity follows from (21):
(39)
fOr a classical gas with relaxation time T ... For the FeTmi-Dh":lC distrihution,
from (35),

APPENDIX G: VECfOR POTENTIAL, FIELD MOMENTUM,

AND GAUGE TRANSFORMATIONS

This section is includcd because it is hard to find the m.lgnetic vector

potential A discussed thoroughly in One place, and we need the vector potential
 (

in supcrconducti\"ity. It may set!m mysterious that the hamiltonian of a pari ide

in a magnetic Ileld has the form derived in (18) below:

I ( P -~ A
I f : 2M Q)' +Q'I'. (I)

where Q is the charge; M is the mass; A is the vector potential, and r.p is thl'
electrostatic poll'll lia!. This expression is valid in classical mechanics and ill
quantum mechanics. Because the kinetic energy of a particle is not changed by
a stalic magnetic field, it is perhaps unexpected that the \ ector potential of the
magnetic field enters the hamiltonian. As we shall see, the key is the observa-
tion thai the momentum p is the sum of two paris, the kinetic momentum
P~;n = Mv (2)
which is familiar to us, and the potential momentum or fidd momentum

Pfi,,1d ::: .!l.

c
A (3)

The total momentum is

I p = ..,. + "''''''' = Mv + ~ A (4)

and the kinetic energy is

- I Mv2 I
= -2M (Mvl I
= -2M (p - - Q)'
A (5)
:2 (;
The vector potential l is related to the magnetic fleld by
B = curIA . (6)
Wc assume that we \\1Ork in nonmagnetic material so that Hand 8 arc treated as
idcntical.

Lagrangian Eqllatimls of Motion

To flnd the ham iltonian, the prescription of cla.~sica11l1echanics is clear: we
must flrst flnd the Lagrangiun. The Lagrangian in gcncnllizcd ('tJordinates is

L = ~ Mi,'- - Qip(q) + ~q . A{q) (1)

2 c

'For an elemc nlar:- treatment ofthc ' "ector poIcnliai 5Ct: E M Purcell. ElrcllidlV ont! mal,:·
nctiml. 2nd cd., McC ......· lflll. 1984
This is correct heL. ... se it k.uls to the correct equation of motiou of a charge in
combincd e lectric and magnetic fields , as we now show.
In Cartesian ("oordinatcs the LagrUllgC equation of motion is
d aL aL
-----= 0 (8)
fit ai ax
and similm-I)' for y and ;:;_ From (7) we form

dL= - Q -d~
- + -Q (., --+
aAr . aA"
y ~ +z -­
. aA ~ ) (9)
ax ax c ax ax ax
dL
-= Afx +- Q A . (to)
ax c "
d -
- . =Aix.. +Q
aI~
--dAr ._
- = Mx+-Q (aA r
- +y. -
aA ~
- - +,. - aAr- +,. -
aA,)
- .
dt ax c cit c at ax ay az
(11)
Thus (8) becomes

[d."
M'i:+Q -d~ +Q- - - + y -- - (aA, dAy) +z (dA. dA.)]
--- - -
- =0 (12)
ax c at ay ax az ax
0'
d 2x , Q
.\1 -, 1t 2 = Qf.. + - C [v>< Blr ( 13)

with

(14)

B = curiA (15)

Equation (l3) is the Lorentz force equation. This confi rms thai (7) is corrcct.
We notc in (14) that E has one contribution from the electrostatic potential 'P
and ano ther from the timc derivativ~ of the magnetic vector potential A.

DerivatiOlI of tlw Hamiltonian

'nle momentum p is defined in terms of the Lagnlngian as

P ... aL = Mq' +.Q. A (16)

iJil c'
in agreement with (4). The hamiltonian I-1(p,q) is defined by
l1(p,q) £:; p' II - L , (17)
0'

11 ::: Mil + 2. q' A - ~ Mil" + QIP - 2. q' A =

c 2 c
_ 1_
2M
(p_SLc A)2 + QIP ,
(18)

.IS in (I).

FicicI MOllle1ltmn
TIle momentum in the electromagnetic field thai accomp.tnie" a particle
moving in a magnetic field is given by the volume integral of the Poynting
vector, so thai

Pr",kl = _ 1_
4we
f {IV E )C H (19)

\VC ,."ork in the llonrcJat ivistic appro:<imalion with v <.i; c, where v~" the veloc-
ity of the particle. At low values of vIc we consider B to arise from an cxternal
source alone. but E arises fro m the charge on II Ie part icle . for a charge Q at r',
E = -Vq;; (20)
Thus
P, = _ _ 1_
he
f dV~!p x cu rl A (2 1)

By a standard vector relation we have

f dV VIP x curl A = - f dv (A x curl (VIP) - A lliv Vtp - (VIP) eliv AI (22)
Bul curl (VIP) = 0, and we can always choose the gauge such that eliv A ::: O.
TIlis is the transverse gauge.
lllUs, we have

PI ::: __1_
4we
f dV A V2ip "" ..!...
e
f dv AQ 8(r - r') = SL A
e
(23)

This is the interpretation of the field con tribution to the total momentum
p = Mv + QAle.

GAUCE TR'\NSFOR~IATION

Suppose 1I4fr = fiji. whe re

H = - I ( p - -A
2M e
Q)' (24)
,
Lt:1 us m,lke a gauge 1r.lIlSform;ltioll tQ A'. where
A' =A + 'ilX, (23)

\\herc X is a scalM. t\Ow B = cu rl A "" (' uri A', hccilu!.C curl (VX) .. o. 'nlC
Schrodingcr equation becomes

-
2M
'(p - -Q
c
A' + -Q)'
c
Vx 4b = ~y, . (26)

What 1/1' ~,\tisfles

_,_
2M
(p _2_
c
A')2..,/ :: E4J' • (27)

wHh thc s.lmc E as for .p? Equation (27) is equivalen t to

_,2M_ (p - 2....c A - .£VX)2",. = EIjJ'

c
(28)

We try
1/1' = cxp(iQx' lic)'" . (29)

Now

P4I = cxP(iQXlfrc)prJ! + 2....

c
('VX)e)(p(iQx1hc)1/!

51.1 that

;lud

-
2M
'(p - -Q
c
A - -Q)'
c
Vx tJI = exp(iQxIl1c )'(
- I' - -Q)'
Z.U c
A ~
(30)
= cxp(iQx1Iic)ErJ! .

l1\Us 41 = eAp{iQxIl1clt/! satisfies the Scil rOOlllgn equation [lner the gauge
transformation (25). The ener!!)' E is IIlVariallt under the tr.msform;\lioli. The
gauge tmll5formation on A merely changes the loc.'\ p hase Q(: lll c wavcfunction.
We see 11M!
(3 ' )
1/1'.""
= ""'''' '
so th.lt thc chclrge density is illv,irian! under a gauge trarurormatilln.

Gauge ill tIre London Equafian

RL'CaU!>e of the e<luation of continuity in the flow uf electric charge we
rt.'<} u irc Ihat in a sup(.'I-crmductor
 di\' j "" 0 ,
so that the "ector potential in the London equatiou j = cN41TA~ !nllst satisfy

di\' A '" 0 (32)

Further, there is no curre nt flow through a \ill'Uum/sup('rl'OlIlluctor interfa(.'('.
TIle normal com ponent of tile CllrTlm t across the illterf.'ll'C must wmisll : j" = 0,
so that the ,,(;'Clor potenti,d in the LOII(Iol1 equatiou must satisfy

A" '" O. (33)

111e gmlge of the vector potential in the London equatiOi l of superconductivity

is to be chosell so that (32) and (33) are s,ltisfied .

APPENDIX II: COOPEll PAms

For a complete set of states of a two-electron systcm that satisfy periodic

boundary conditions in a cube of unit volume, we take plane wave product
functions
(1)

\ Ve ass ume that the electrons arc of opposite spin.

\Ve introduce center-of-mass and relath'e coonlinates:
1l = Hr, + rJ (2)

K= kJ + k2 ; ~ ~ I(~, - ~J . (3)

so that
(4)

11ms (1) becomes

q:{K,k;R,r) = exp(iK ' R) exp(ik· r) (5)

and the kinetic energy of the two-electron system is

te l>: + E. = (h 2/m)(fK2 + k') . (6)

\Ve give special atten tion to the product fu nctions for which the ('('nter-<lf·
mass wa\'e\'l.'Ctnr K = 0, so that k, = - k2 . With an interaction HI between the
two electrons, we set up the e igen\'alue problem in tcrms of the expansion
x(r) = LgJ. exp(ik· r) , (7)
 (l1u + 11, - t')x/r) = 0 ::: ~1 H[L - t-:)I!l< + J/1 1!.i. k "p(ik' 'I'I , (~)

wlll're II , h till' inh'ral'l ioll l' IICI'K\ (If till' Iwo dectl1lll ~ , Il ert' f i, 11ll' d~('n ­
v.ll ul'.
\\'e 1,Ikl' tht, Sl:.llar llJ"()(iu(·t with e \l >{ik, r ) to nbt.lin
( E~ - ()I!J. + ~~ ~ (kill ,Ik') = 0 , (U)
till' H'I: nl,!f l·(I\l.IliOIl of the lno hll'm .
Now t r.ll\~rOn ll till' ~nm to an mtegral:
(10)
w he fl' N(E') i s th l' IIIlInl)("f of two ('I('dron ,t.de~ \\ ilh lulal momentum K == 0
uml with !..inel;,· clll'Igy in (f£ ' at E' .
Nnw (:ollloicl('" til(' matrix dcn1('nt~ J/ 1 ( 1~, E '1 = (klll ,l k'). St\lJi l'~ oftlU's('
hy l3an\cI'n sUAAI'st IIMI they art' important when the two dCI·tnm~ arc (1)11 ,
fillet.! to a thin l"1lt'~ shell near the F e n lli ~urfac('- within a ~hcl l of th it:kness
"w" abo"I' E" where w" is the V('II)"(' plwnol1 l"utnff fn'tlu cnl .... . \\"l' a,sullW
that
H .(f, E') ::: - v (II )
for E,E' within the she ll and l.ero otlll'rwis('. Here \' is a.~sl\nwd to he positivI'.
Thus (10) !Jerollles

(t - (If,t(E } ::: V
",
1'" dE' ~(E'},'''(E ') ::: C , (12)

with fm = ( I' + fl w" . He re C is a l"OII.~lanl , illclqx'mk'nt tlf ,..

From (l2) W (' h,\VI'

"(E)- ..,E- ( 13)

E - (
and

1t o,
2' ~ N(E ')
1 = V cI E' ""'
.. "'-''- (14)
f, - E

, Vi th N(E '} apl)roxilllatel y ('(lIlstan t ;\\Id equal (0 NI' o\'(' r till' s \ll;LII (' I ll'r~~ r;:llIgC
between 2Em and 2E,., we take it out of thl' integral tn oht ain
I
12_.
2'~
I = !V,.V clE ' " = l\' ,.\' l(l~ (IS)
E £

Lei the e igl' m . llu(' ( or (15) lx, writll'n as

E = 2£,. - 6, (I 6)
which dcfines the hlnding e nerg\ 6 of thl' e1eclwn I,,'ir. rclatiH' tn tWI) fn'\.'
658

electrons at the fenni surface. The n (IS) be<-omes

2t:-2£,.. +lJ.
1= N~ V log = N,.·V log (1 7)
0'
m

•
I/NFV:: 100(l + 2hwd A) . (lB)
This result for the bimling energy of a Coope r potir may be written as
2hWf)
• : -----;~'!!c-__;_ (19)
CXP(lINFV) - 1
For V lXlsitive (aUradivc interaction) the energy of the system is lowen."{l by
excitation of a pair of electrons above the Fermi It..vel. Therefore the Fermi sea
is unstable in an importallt way. The binding e nergy (19) is closcly related to
the slIpc("(.'Onducling e ne rgy ga.p E. " The BeS calculations show that a high
density of Cooper pairs may form in a me tal.

APPENDIX I: CINZBURC-LANDAU EQUATION

\\'e owe to Ginzburg :Uld I_mdall an eleg.1.nt theory of the phenomenology

of the 5uper<.'O nJu<.'ting slate and of the sp.lIia! variation of the order parameter
in thai slate. An extension oCthe theOl), by Ahrikoso" describes the structure of
the vortex state which is exploited te(:hnologically in sllperconducting magnets .
'nle attractions of the GL theory are the naturallntrlXiuction of the coherence
length and of the wu"cfl1nct ion used in the theory of the Josephson effects in
Chapter 12.
We introduce the order parameter tf!{r) with the property that
(I)
the local concentration of sllpcrcomlllcting electrons. 'nle mathematical fonnu·
lation of the definition of the function IjJ(r) w ill come out of the DeS theory. \Vc
first set lip a fonn for the free energy density Fs{r) in a superconductor as a
fuoction of the order paramete r. \Ve assullle that in the general vkinity of th e
transition temperature

with the phenomenolOgical positivc constants a , fl, amI m, of which more will
be s.."Iid. Here:
1. F,, ' is the free energy density of the normal state .

l
r A"I"!"'Ux

2. - o:l~~ + !{3ltpj"' is a I\'pi(:al Landau form (.is in C!t:lptef 13) for the
c:l.pl'lnsioll of the free energy in terms of an order pa rameter lilal vanishes at a
sl.'t:ond-ordcr phase transitiOIl . This term may be viewed as - ails + 1t~Il~ and
by itself is a minillllllll with respect to 11.\ wllcn ns(T) = al11
3. 111l' t{~rm in Igr,uJ y~2 represents an increase in energy caused hy a
spatial variation of tile order parameter. It has the form of the kine tic cllcrgy in
quantum mccllanics. I 11le kinetic 1Il01llt.'lltUIll - iii V is accompauieo lIy the field
momentum - ClAie to ensure the gauge invarianl'e of the free encq,'Y, as in
Appendix G. Here q = - 2e [or an electron pair.
4. "nll' term - JM . tiB,,', with the fictitious magneUi'.ation
M = (8 - Ba)/41T, represents the increase in the sllperconoucling free energy
(:aused by the expulsion of nl<lbrnetic flux from the sUJx:rconductor.

111e separate terms in (2) will be iIIustratoo by examples as we progress

further. First let us derive the GL equation (6). We minimize the total free
energy fdV Fs(r) witll respect to variations in the function ""r). We have
8Fs{r} = I - o..p + f311/Jl 2 .p + (1I2m)(- ihV - qAlc).p· (ihV - qAlc)J8t/1* + C.c.
(3)
We integrate by parts to obtain
fdV iV")(V8.,,.) ~ - fdV (V'")..,,. (4)
if 8t/1* vanishes 011 the boundaries. It follows that
8fdV Fs = f(IV 81/1*[-o..p + J311/Jl 2 .p + (1/2m)( - iliV - qAlcf.pJ + c.C. (5)

l11is integral is zero if the term in brackets is zero:

I
[(1I2mX-ihV - qA!cr~ - a + J3lrJol 2 1rJ< = 0 . \ (6)

111is is the Ginzburg-Landau equation; it resembles a Schrtklinger equAtion

for .p.
By minimizing (2) with rcsJ)(.'Ct to 8A we obtain a b"Ulgt..... invariant eXjw's-
sion for tllc supercurrent flux:
(l)
At a fn..'C surface of the specimen we must choose the gauge to s.1.tisfy the
boundary condition that no current flows out of the superconductor into the
''al'uum: n' h = 0, wllCre it is the Sllli.1.CC normal.

lAcontrilmtiun of the form IVMI', when: M is the magnetization. """"' i"trtxlll~",J by l..ambn
and Llfslntz t\ resent the ""whangc enl.'r!(y d('nsity in a fen"Olnag'lCt, ~ce QTS , p. r>5.
o

Coherence !.englli. The intrinsic coherence length { may be defined from

(6). Let A = 0 and suppose that 131~2 Inay be neglcctec.l in comp..'trisol1 with a.
In olle dimension the CL equation (6) reduces to
h 2 (/ 2 1/1
- 2m d x2 = at/J (8)

This has a wavelike soll1liol1 of the form exp(ixl(), whe re f is defined by

{ _ (,,2/21110')112 . (9)
A mo re inte resting speci.,'ll solution is obtained if we re tain the nonlinear
term Pl1bl 2 in (6). Let us look for asol utioll with t/J "" 0 at x = 0 and with t/J- %
as:l:_ 00, TIllS situation rep resents a boundary betwt."e11 normal and superCOIl-
d ueling states. Such states can coexist if there is a magnetiC field He in the
nonna1 region. For the moment we neglect the penet ration of the field into the
superconducting region; we take the field pe nctration depth A « f. which de-
fines an extreme type I superconductor.
1lle solution of

(10)

subject to ou r boundary conditiolls, is

~:x) = (alJ3)' t2 tan h(:xN2$ (II)
TIlis may be verifie<1 by din..oct substitut ion. Decp inside the superconductor
we have ~ I = (0//3)112, as lollows from the minimization of the terms
- aI4.f + iJ3ll/ll ~ in the free encrgy. We see from (II) that ~ marks the extent of
the coherence of the supen.:onduding wavefunction into the normal region.
\Ve have s~'en that deep inside the supef<.:olldudor the free ellcrgy is a
minim um whcn 1~~2 :t:: 0//3, so that
(12)

by definition or the thermodynamic critical fiel d 11,. as the stahililA1tion free

energy density of the slIpef(.'Ondllcting state. It folluws that the critical field is
rd.l ted to a ;Ind /3 hy
(13)

Consider the penetrat ion depth of a weak magndic fil,llI (n <C II r) into a
superctlmiuctor. ' Ve assuille that 11/11 1 in the superconductor is equal to l!Por"\ the
value in the ahsen<.'C of a fl eld . Theil the equation for the SlIpe rcurrent flux
reduces to
 (14)
which is just the London equation j s(r) ::: -(d41TA 2)A. with the penelr:l.lioll
depth

(15)

The di mensionless ratio K .... Alf of the two dlUracteristic le ngtlls is an

important parameter in the Illcor), of superconductivity. From (9) and (15) we
find

(16)

We now show thai the value K ::: ltV2 diviclt.'S type I superconductors
(I( < UY2) from type II SUI)(~rrondlictors (I( > 11\1'2).
Calcull1tioll of tile Upper Critical "'ield. Supcrcontiucting regions nucleate
spontaneously within a normal contluctor when the applied magnetic field is
decreased below a value denoted by Hc2 . At the onset of supcrcolHluctivil y ItPl
is small and we lineari ze the C L equntiolT (6) to obtain
1
- (- ihV - qAlc'fi/l = a (17)
2m
TIle magJlelie held in a supercondu<.iing rt.>gion at the onset of supercon·
ductivity is just the applied lIeld, 50 that A ::z: IJ(O,:t.O) and (11) h<.."{'Omes

_ ....'!- (~ + J!-.);d _I_(ih ~ + q8 X)2 rjJ = arjJ . (18)

2m a'? 2
vz. 2m vY c
This is of the same form as the Sehrodinger equation of a free particle in a
magnetic fie ld .
We look for a solution in the form e:l:p(i(kl/Y + k;:z)lrp(x) and find
(1I2m)( -h 2cfld'? + f12k~ + (f1k l/ - q8x1c)2}'P = arp . (19)
this is the equat ion for an harmonic oscillator, if we set E "" a -
(f12121I1X ~ + ~) as thl~ eigenvalue of
(1I2m)[ - h2 d 2!dl.2 + (q2 8 2/c'l),? - (2f1kv4B/c)x ]'P = E'P . (20)

The tennlinear in :lcan be transformed away by a sh ift of the origin frOin 0

to ;Co = fikrll1Jl2mc, so that (20) becomes, with X = x - ;co,

" dx2
- [ 2m d' 1
+ ~m(qBlmcfX2 rp "" (E + f12kV2rn )lP . (2 1)
.,
The largest value of the magnetic fkld n for whit'h solutions of (21) c\:ist is
given by the lowest eigenvalue. which is
!"w = hqB,.--IZmc = a - 1121.:~I2.m • (22)

\\ h,'re w is the oscillator frequency qBlmc. With k~ set ('(Ilia! 10 zcro,

1J",~, • 1("2 = 2Ctllldqll . (23)

This result may be e)(pressed by (13) and (16) in terms ofthc thennody-
namic l'f'iti,:al field He and the GL p.valOetcr I( • Mf.

11
'lame
=---
H"
TtCJ?1m 112 ""
_ r;: me
v2 - y~
T7f
- H<;= v21(J1" .
. t;;.
(24)
2
<; qli (4 /iq 211

when Alf > lI'V2. a sllperCOn(!uc!or has H,-2 > lie and is s.lid to be of type II.
It is h elpful 10 write 11<2 in terllls of the flux quantum 4>0 = 2'1Tllclq and
e = r,2/2ma:

Hd = _2_"_,,,_"_ . _,_,_",_,_ . -;:-chc'-;;;- q" (2'»

qll 2TTflc 2mae = 211"f
In.s tells us thai at the upper criticallleid the (lux density 11<"2 in the U1alcri~ll is
eQual to one flux Qlk1.lltuUl per area 2TTe.
consistent With a fluxoid lauke spa('-
ing of the order of f.

APPENDIX J: ELECrRON-PHONON COLUSlONS

rhonons distOl-t the loc'al crystal structure and hence distort the localll<lI1d
structure. This distortion is sensed by the ('(Induction electrons. The important
effccts of the ('Oupling of ele<:troll5 with phollons arc

• Electrons are scattered from onc state k toanotber state k'. ie..'tding to electri-
cal resistivity.
• phonollli can be absorbed in the scattering event , leading to the attenuation
of ultrasonic waves.
• An ek'"Ctron will carry with it a Cfystal (listortion, and the effective m'lss of the
electron is thereb y increa.~ed.
• A cryst.1.1 d istortion associated with olle ek'<.1ron can be sensed by a second
electron , the reby causing the ek'"Ctron-electron interaction th:lt enters the
thwry of superconductivity.
'nle (leforrnation potenti.1.1 approximation is that the electron energy f:(k) is
coupled to the crystal dilation 6.(r) or fractional volume change by
tOe, r) ::: Eo(k) + C6.(rl • II)
whe_rc C is a constant. The approximation is usef\11 for spherical band edges
f:()(k) at long phonon wavelengths and low electron concentrations. The (lilation
may he expressed in term s of the phonon operators uq , of Appendix C by u;
6.(r) :: i 2: WU"w )lf2!q1(o"exp(iq' r) -
q o~ exp(- iq ' rll • (2)
•
as in Q1·S. p . 23. Here M is the mass of the Cf)'st.1.I. The result (2) also follows
from (C .32) on fonning q, - q~_ 1 in the limit k <c 1.
In the Born app roximation for the scattering we are concerned with the
matrix clements of CA(r) between the one-electron Bloch states Ik) and Ik'l.
with Ik} = exp(ik· r )uk(r). In the wave field representation the I1Mtrix element
;,
H' = f d'r ~+ I,)Cal') ;'(') =
,,'L " ,,,(l<'icaik)
:=0 iC 2: ~C\r. 2: (hI2MwJl12jq](U'l f d3x 1Ii:'Ukeo{lt- ~
k " 'I)' r (3)
k "k "

where
(4)

where c~, C\r. arc the fcn11ion creation and annihil.1.tioll operators. The product
"k·(r)lIk(r) involves the periO(lie 1).1.rts of the Bloch functions and is itself periodic
in the lattice; thlls the integral in (3) vanishes unless

k - k' :!:: q :=o [0

vector in the reciprocal lattice.
In semiconduclors al low temperatures ollly til(' possibility ...ero (•• rocesses)
may be allowed energetically.
Let us limit ourselves 10 N processes , alit! for convenience weapproxilllaic
f cPx 11.(11),; b)' ullit)'. TIlen the deformation pot!'ntial perturb..1.lion is

H' = iC L (hl2Mw~)ll2lql(llqci~<,aC1c - fI;ck-qeV . (5)

"
Rclflxtllion 'rhllt'. l u the presen<-'e of the electron-phonon interact ion the
waVCl'ector k is not a constant of the motion for the electron alone, but the sum
of the waveveclors of the ek-ctron and \<irtual phonon is conserved. Suppose all
electron is initiaily in the state Ik); how long will it stay in the same state?
We calculate first the probability w per unit lin1(' that Ihe electron in k will
emit a phonon q . If n'l is the initial population of the phonon stale,
u(k - q;n~ + l lk;n<,a ) "" (21TlfI)l{k - q;r1'1 + IIH'lk;n.)1 28(t:),; - flWq - t:1<- <,a ) •
(6)

by time-dependent perturbation theory. Here

"ll1e total collision r.lte W of all electron in tilt' statc Ik) with a phonon
system at absolute zero is, with 11'1 "" 0,

w "" ~
41TPC*
I _I
I d(cos 0.) Lq~ dq q3li(t:. - t:1< _q - flw,,) (8)

where p is the mass dellsit)'.

11lc argumen t of the delta flll)<-'lion is
fl 2 h'l
- - (2k' q - q'lJ - hC.Cf "" - - (2k' q - q2 - qq,) (9)
2m· 2111·
where qc 5 2f1m- c., with c. dIe velocity of sound. The minimum value of k for
whid\ the argument can be zero is k...1n "" j(q + qe), which for q :: 0 reduces
to kmm :: jqe :: m·c/h. For this \'aloe of k the elect ron ~roup \-elneily
v, "" k"'l.,Im. is equal to the velocity of sound. nllls the threshold for the
emission of phonons by electrons in a crystal is that the electron group velneity
should exc(.'(!d the acoustic velocity. nlis requirement resembles the Cerenkov
threshol(l for the emission of photons in crystals by fast electrons. The ek-ctron
energy at the threshold is hll·~ - 10- 27 . 1011 - 10- 16 erg - I K. An c1e<.'-
Iron of energy below thi s threshold will not be slowetl down in a pe rfect crystal
at absolute zero, e\'eo by higher order e lectron-phonon interactions, at least in
tile harmo nic approAimation for the phonollS.
 for k » qr W, .1)' neglect the qq,. lerm ill (9). 111c integrals ill (8) bct'Olilc

r- I
d~ J dq q3(2mt<' h 2(J )jj(2kIL - q) = (Bm*' h 2)
~)
[I (/p, /...2Jl. 2 = 8111*/...21311 2

(10)
and the phonon emission rate is
2C l m*P
W (emission) = -';;"--"-;71 ( II )
31Tpc,f1
dira",ly proportion.11 to the ('Icdroll e nergy t~. TIle l os~ of the componl: llt of
wavevedor Il<'\rallcl to the oril':inal direction of tile ele<.'tron when a phonon is
emil led al an angle 6 to II is Kivell b y q cos (). 111e fractional rale of los!> of k~ is
given by the transition rale integral with Ihe extra factor (qtk) <'~IS () in the
integrand. Instead of (10). we have

(2111*lh 21<.) I
I (IlL 8I2p:' = 16m .. ,,2'1511 2 (12)
"
so that tile fraclional rale of de<:rcasc of k~ is
W(k,) = 4C 2m*k2{S7TPC.1!2 ( 13)

This quantity ell ters into the c1ecl ricaJ resistivity.

111(' aoove results apply In ubso lutc zero. At u tCllIpcwtl1rc kill' ~ hc,k thc
integrated phonon emission rate is
C~IIi"kkB'r
W(emission) = ..2 :) (14)
"fTC:,p"
For clcctwns in the mlal eljui lihriullI a t lIot too loW tc mp .... nltiIH' s the rcquired
inequality is easily satisfied for the rms valliC of k. If wc take C == 10- 12 erg:
111* == 10 - 27 g; k = 10' COl - I; c, == 3 X Hi' t·m S- I; p = 5 g CIII - :\ thcll \V =
IOIZ 5 - 1• At absolute ze ro (l3) g ives W :z 5 X 10 1" S I with tll('st, S'1I11e
parametcrs.
Subj ect In<kx
Absolute thermoelectric
powe<. ""
Acceptor ion ization energies,
22S
Acceptor $tates, 224
Acoustieal phonon branch,
10<
Acronyms , 328
Act;iV;ltion ellcrgy, 5«
Adiabatic denw.gnetU.ation.
431
Alfvtn waves, 302
AlkaJi laUde crystals. table,
73
Alloys, 6 11
$trength. 600
tranSition metal, 624
Alnico V, 476
AmQrphous solids, 521
f'erromllgllet s, 529
$emiconduc:tol'$, 530
Anharmonic intenctions, 129
Ani$otropy energy, 470, 471
Annihilation opentors, 641
Antiferroelednc Cl')'JtaU, 404,
..,
406
Anliferromagnetic et)'$tals.
..............
resonance, &16, 508
Anti-S lokes line. 323
Atomic fonn factor, 4.5
Atomic radii. 77
table. 78
BCS theory, 345, 354
energy gap equaijOll, 301
Band bending. $16
Band gap, 175, 199
Band structure, CaAs. 218
germanium, 215
Barium titanate, 397
Basis, 4, 5
Binding, c~talline . 54
8 iomagnetism, 478
Bloch equations, 492
Bloch frequency. 230
Bloch function, 179, 246
Bloch OICillatOl'. 229
Bloch T"- law, 456
Bk>cit theorem. 179, 185
Bloch wall. 472
Bohr rnagneton. 421
Boltzmann transpOrt equation.
64'
Bosoo openton, 641
Boundary condllions.
periodic, 120
Bragg law, 29
Bravab lattice, 10, 556
Brillouin function, 4.2.2
Brillouin SCilttering. 306
Brillouin zone, 37, 217, 266
first, 39, 100, 176
yolume. 51
Bulk moduli, table, 59
Bulk modulus, 86
Bulk modulus, electron
100
Fenni gas. 303
Burger! vector, 592
Cauchy integral, 309
Causality, 329
Cell, pnmitive, 6
unit, 6
.,
WigneI"-Seitz, 8, 38
Centipoise, SZ7
Central e<juatim, 186
Cesium chLonde structure,
186
Chak:ogenide gla.ue$, 531
Charge dell.$ity waves, 301
Chemical potential, 146, 153,
154, l Ee
CWsi.cal di$tribution, 648
Clogston relation. 349
Owsill$-Mossotti relation,
390
Coertive r(ll"Oe, SQall partk:ie,
472
Coerrivity, 470, 475
Coherence length. 352. 660
inb"insic, 353
Cohesive energy, 57, 56. 80
Fermi gas, 170
sodium metal, 251
square well potential, 266
Cohesive energies, table, S1
Colfuionleu eltlCtJ'oo gas, 311
Color centers, S48
gall.
CompreuibiHty, 86
Compcessibilities, table, 59
Concentration, tlbIe, 24
Conduction band edge, 199
Conduction electron
ferromagnetism, 438
Conduction electror:ll.
susceptibility, 433
Conductivity, electrical. lM,
158, 221
im purity, 222
ionic, 298
thermal. 13 1, 166
Conductivity lum rule. 330
... ,""'"
Conductor, perfect, 339
Configurational heat capacity,
630
Contact hypernne inter3Ction,
.97
Continuum wave equation,
112
Cooper pairs. 356, 656
Covalent bond, 72
Covalent crystaU, 72
Creation openton. 641
Creep, 603
Critical currents, 372
Critical field, 339, 372
thin films , 375
Cntical magnetic field, 372
Cntieal point exponents. 445
Cntical points, 312
Critical shear stress, 588
Critical temperature. 399
Crystal Geld, 426
splitting:. 426
Crystal growth, 25
Cry1tal momentum, 109, 186
Crystal structure, elements,
23
Cl'}'j:tallinc binding. 54
Cl'}"tallograph)', surface, 5S6
Cubic lattices. 12
Cubic zinc sulfide structure,
20
Curie constant, 423
law, 423
temperature, 44J
CUJ"Ie-Wew law, 444
Cyclotron frequency, 164
Cyclotron resoclance. 213
spheroidal energy .surface,
23'
Dltngling bonds, 554
Oavydov splitting, 331
Deformation, 80
De Haas-""II Alphen efft:ct.
2SI
period, 266
Debye model density of
states, 122
Debye scattering. 522
Ocbre r
law, 124
Ochre temperature. 122
table. 126
Debye-Waller factor, 632
.Defe<:t.s. paramagnetic, 499
Deficit temiC(lnductors, 222
Degeneracy, 145
Degenerate semiconductor,
2M
Demagneti:tation. isentropic,
43'
DemagnetiUltion factors, 504
Density, table. 24
Density of modes (~e density
of states).
Density of states. 128, 149
Debye model, 120
Einstein mode l, 124
general result, 127
one dimension, 118
singularity, 139
three dimensions, 119
Depolarization factors, 384
Depolariz.:ltion field, 384
Diamagnetism, 417
Diamond nructure, 19, 51,
'95
Dielectnc constant, 388
semiconductors, 223, 330
Dielectric function, 307
electron gas, 271
l..indhard, 282
Thomu-Fermi, 282
Dieleetrics, 381
Dielectric sl,uceptibility, 385
DilITaction, Josephson
junction, 369
Diffraction c;onditions, wave,
34
DiffUSion, 495, 544
particle, 648
DiffusiVlty , 544
Dilation, 82
Direct gap semioonductor,
213
Direct photon absorption,
201, 202
Dislocation densities, 598
Dislocation multiplication,
600
Dislocations, 589
Dispersion relation, phooOnl,
100, 104
electromagnetic waves, 273
Displacive tranSition , 394, 398
D issipation sum rule, 329
Distribution. classical
Fermi-Dirac, 130
Planck, 101
Dog', bone oroit, 263
Domains. origin of, 474
dosure. 474
feTfoelectric, 406
magnetic bubble. 481
Donor ionization energies,
224
Donor $tates. 222
Doping. 222
Dulong and Petit value. 127
Edge di5loc::ation. 589
Effective mass, 209, 210. 212
n~tive. 211
semiconductors, table. 212,
21.
tensor, 210
thermal, 156
Einstein equatioo, 561
Einstein modd. density of
states,
Eilutlc compliance. 83
Elastic limit,
Elrutic stiffness, 83, 91
Elulie strain, BO
Elastic waves, 87
Elastic wave quanti~on,
E lectric field, local, 386
macrosoopic, 382
Electric quadruple moment,
503
Electrical conductivity, 156,
."
table, 160
Electrical resistivity. 159
table, 160
Electromechanical coupling
factor. 410
E lectron affinities, 68
Electron compound. 615, 616
E lectron density distribution.
67
Electron heat capacity. table,
257
Electron orbits. 242
Electron spin resonance, 485
Electron work functions, S60
table, 562
Electron-electron collIsions,
294
Electron-electron inter2Ction,
294
E lectron-hole drops, ~ ,321
Electron-lattice interaction,
297
Electron-phonon collisions,
66'
Electronic pol:lrizabilities,
table, 391
Electronic structu re, surface,
560
Electrostatic screening, 280
Elementary excitations, 98
Empirical pseudopotentiai
method, 253
Empty core model, 253
Empty lattice approximation,
188
EN DOR,500
Energy band calculation, 244
Energy bands, 17J, 175
E nergy gap, 173, 175
superconductors, 344
table, semiconductors, 201
Energy levels, one
d imension, 144
Energy loss, fast particles,
3"
Energy 1055 spe<::tra,
electrons, 279
Entropy. supereonductnn,
342
Entropy of mixing, 620
Equation of moti on, hole , 208
Equations of motion,
electron, 200
ESeA,324
Eutcctics, 622
Ewald construction, 37
sphere, 559
sums, 634
Ewald-Kornfeld method, 637
Exchange energy, 446
field, 443
freq ucncy resonance, 5 16
integral, 443
in teraction, 74
narrowing, 510
Excitons, 312
Exclusion principle, Pauli, 64
Exotic superconductors, 156
Extended zone sche me, 238
Extinction coefficient, 307
E'\tremal orbits, 260
F centers, 499, 548, 552
Factor, atomic rorm, 43, 45
Itructure, 44
Fermi energy, 145
gas, 141 , 146
level, 144
Iiquid,2:£).1
surface, C'Opper, 262
surface, gold, 262
.urface ptlrameters, table,
150
sorfaces, 238
Fermi-Dir3C distribution, 146,
643
Fermions, hcavy, 156, 337
Ferroelast icity, 410
Fel'roelectriCity criterion, 412
crystals, 381, 393
domains, 3S8
linear amly, 414
Ferromagnetism, crystals,
443, 449
amorphous, 529
conduction electron, 43S
domains, 468
Older, 458
resonance, 503
Fib« optics, 535
Fick's law, 544
Fine structure constan t, 566
First Brillouin zone, 39, 100,
17.
First-order transi tion, 402
Flux quantiution, 356
Fluxoid, 355, 363
Fourier analysis, 30. 42, 183
Fractional quantized Hall
effect (FQ HE), 570
Frank-Read )()UJCe, 600
Free-energy,
superconductor3, 343
Frenkel defect, 542. 55 1
}<' rcnkel exciton, 3 14. 316
Friedel oscillat ions. 283
}<' ullerenes. 373
Fused quart'l., 524
Gap, direct, 20'2
Gap. indirect, 202
Gap p lasmon~, 304
Gauge transfOlmation, 654
Geomagnetism. 478
Gln7.burg-Landau equation, 65S
Ginzb urg-Landau paramete r,
354, 661
CLAC theory, 360
C lass, 527
chalcogenide, 531
transition tempe rature, 527
Crain boundary, small-angle,
505
CrQlJp velocity, 102
Criineisen constant, 129
Cyromagnetic ratio, 420
Gyroscopic equation, 490
Hage n-Rube ns relation, 335
Hall coefficient, 164
table, 167
Hall effect, 164, 373
two carrier types, 231
Hall resistance, 165
Hall resistivity, S64
Hardness, 606
Be3 • liqUid, 170
Beat capacity, C'OnHgurational,
630
electron gas, 151
glas~cs, 533
metals, 155
phonon. 1I7
superconductors, 342.
Heavy fennions, 156, 337
Heisenberg model, 446
Helican waves, 303
Hetel'ojunction, 576
HeterOlitructures, 574
Hexagonal dose-packed
structure, 17, 25
H exagonal lattice. 51
H igh temperature
superconductors. 37 1
Hole, equation or motion,
200, 2GS
Hole orbits, 242
Hole trapping, 298
Booke's law, 83
HTS, 37 1
Hume-Rothery rules,
Hund rules. 424
Bydrogen bonds, 76
Hypemne constant, 498
Hyperfine effects, ESR in
metals, 515
Hyperfine splitting, 497
H)'$teresis, 475
Il meni te, 396
670 lnde.,
Impurity conductivity, 222
Impurity orbits, 230
Index s)'stt:m, 12
Indirect gap, 202
Indirect photon absorption,
201
Inelastic scattering, ph~OIlS,
tlO
Inert gas crystals, table, 60
Injection IdSer, 577
Insulators, 193
Interband transitions, 312
Interface plasmon!, 302
Interfacial polarization, 412
Intrinsic carrier
concentration, 200, 216
Intrinsic coherence length,
353
Intrinsic mobility, 220
llltrin~ic semiconductors, 199
Inverse spinel, 460
Ionic bond, G6
Ionic character, table, 76
Ionic conductivity, 298, 542
Ionic crystals, 66
louic radii, 79
table, 78
Ionization, thermal, 226
10nif,3,tiun energies, acceptor,
225
donor, 224
table, 61
Iron garnets, 461
hOIl group, 425
Isentropic demagnetltatioll,
431
Isotope effect,
superconductors, 346, 347
Isotope effect, therm.al
conduction, 138
Jalm-Teller effect, 221, 429
trapping, 551
Josephson tunneling, 367,
369, 377
Jump frequency, 546
Junction, supercondueting,
3"
Junctions, p-n, 570
kp perturbation theory, 'JJj7
Kelvin relation, 225
Knight shift, 500, 5O'l
Kohn anomaly, 113
Kondo effect, 628
Kramers-Hei5>'llbcrg
dispersion, 309
Kramers-Kl'Omg re lations,
308, 310
Kronig-Pennl.'Y model, 180,
195
reciproad space, 187
U modes, lO·1
Lagrangian equatiuns, 652
landau gauge, 570
...
Landau level, 268, 278, 566
L.om&.u theory, Fermi liqUid,
Land:lu theory, phase
transition, 400
Landau-Ginzburg equations,
354
Langevin diamagnetism, 417
umgevin result, 419
Lanthan ide group, 425
Lannor frequcllCY, 418
Larmo r theorem, 41S
Laser, 5lO, 577
injection, SiS
ruby, 512
semiconductor, 5T1
Lattice, 4
Bravais, 10
Lattice constants,
equiliblium, 55
Lattice rrequcncies, table, 292
Lattice momcntum, 205
Lattice sums, dipole arrays,
637
Lattice tr.Inslation operator,
647
Latticf: types, 8 , 10
Lattice ,-acanc~s. 5-11
Lattices, cubic, 12
Laue equation, 36, 581, 631
l.tw of mass actiun, 219
LCAO apPI"OXimatiOll, 245
LEIi:D, 557, 558
Lt:n nard-Joncs potentiui, 63,
64
Lenz's Ia.w, 417
Lindhard dielectric function,
282
Line width, 49-i
Liquid He3 , 170
Liquidus, 622
LO modes, 104
Local electrie fle ld, 3S6
London equatiun, 349, 655
London gauge, 656
London penetration depth,
353, 373, 377
umgitudinal plasma
oscillations, 276
Longitudinal ,clalaljon time,
490
Long-range orde r, 620
Lorentz fleld, 388
Lorenz number, 168
Low-angie grain boundary,
595
LST relation, 291, 29-1
Luminescence, 315
Lrddane-Sachs-TeIler relation,
291
r.bdelung constaIlt, 69
Madelung energy, 68
Magnetic breakd()\\'ll, 268
Magnetic bubble domains,
4SO
Magnetic force microscopy,
478
Magnetic monopole, 358
Magnetic resonance, 487
Magnetic resonance
tomography, 487
Magnetic susceptibility, 417
MagllCtite, 461
Magnetization, 417
saturation, 446
Magnetoconductivity, tensor,
170, 271
Magnctocrystalline energy,
471
Magoetoelastic coupli~ 479
Magnetogyric ratio, 420
Magneton numbers, 42.5
MagnetopIasma frequency, 303
MagnetoreSislance. 564
two carrier types, 231
Magnetotaxis, 478
Magnons, 450
antifclTomagnetic, 466
dispersion relation, 481
Maser action, 510
Matthiessen's rule, 16 1
Maxwell equations, 382
Mean field apprOlumation,
443

Mean free paths. phonon. 233
Meissne r effect, 337. 338
sphere. 371
Melting points, table. 58
Melt-spin velocity, 529
Mesh,556
Metal-insulato r transition, 284
MetaU, 75, 141
Metal spheres, 171
Metglas. 530
Miller indices (.see index
system), 12
Mobility. intrinsic. 220
Mobility edges. 569
Mobilit y gap, 569
Mobilities, table, 221
Molecular crystals, 318
Molecular field, 44J
Molecular hydrogen, 73, 78
Momentum crystal, 109, 186
field , 652. 654
lattice, 205
phollon, 108
Monopole, magnetic, 358
MOSFET, 564, 583, 584
Motional narrowing. 494, 496
Mott exciton, 314
Mott transition, 284. 320
Mott· \Vannier excitons, 318
np product, 219
~anos~u re s, 168
~early free electron model,
276
Neel temperature, 463, 465,
48'
Nee) wall. 483
Negative effeetive mass, 211
...,.
Neutron diffraction, SO, Ill,
$dttenng, 454, 456
spectrometer. 111
NMR tomography, 487
Noncrystalline solid, 519
~on ioea.l structures, 21
Normal mode enumeration.
117
Normal processes, 136
Normal spinel, 460
Nuclear demagnetization, 432
Nuclear magnetic: resonance,
487
table, 489
Nuclear magneton, 488
Nuclear paramagnetbm, 417
Nuclear quadrupole
resonance, 503
Ohm's law, 1:56
Open orbits, 242, 266
magnetoresistance, 267
OptIcal absofptiOn, 203
Optical ceramics, 410
Optical phonon branch, 104
Optical pbOl'lOlUi, soft,
Orbit, dog's bone. 263
Orbits. electron, 242
extremal, 260
hole. 242
open, 266
Order. long.range. 620
short-range, 621
Oroer parameter, 658
O rder-di50rder
transformation, 617
transition, 394
O$cillation5, Friedel, 283
OscillatOl" strength, 392
p-n junctiollS, :570
Paramagnetic defects, 499
Paramagnetism, 417, 420
conduction electrons, 433
nuclear, 417
Van Vleck, 4J()
Particle dilJusion, 648
Pauli exclusion principle , 63
Pauli spin magnetization, 435,
438,502
Peierls instability, 300
Insulator, 301
Peltier coefBcient, 22B
Penetration depth, London,
353
Pe riodic ooundary conditions.
120
Periodic zone scheme. 237
Perovskite, 396
Persistent currents, 359
phase diagrams, 615, 622
Phase tran5itioru, strueturaJ.,
393
Phonons. 107, 108. U7
coordinates, 639
dilpeniOIl relations, 127
gas, thermal resutivity, 133
heat capacity, 117
- - ---
,..... OJ
inelastic scattering. no
mean free path, 133
metals. 2B6
modes, soft, 113
momentum, 108
rhotoYoltaic detectors, 573
Piezoelectricity, oW8
Planck distribution, US
Plasmon frequency, 98, 273,
218
interface, 302
400 mode, sphere, 303
optics, 272
o$Cillation, m
surface, 279, 302
table, 278
Poise, 527
Poisson equation, 272
Poisson's ratio, 94
Polarilons, 'JET
Polarilability, 388, 389
COIlducting sphere, 412
electronic, 390
Polari:w.tion, 381
intertacial, 412
saturation, 412
Polaron, 297, 299
coupling cOMtant, 298
Polytypism, 22
Power a.b$Orption, 493
Primitive cell, 6, 193
PseudopotentiaI, components,
283
metallic !!Odium, 254
method, 252, 312
Pyroelectric, 394
PZT system, 409
Quantiz.ation. elastic wave,
107, 638
Hall effect. 566
orbits In magnetic:: field.
255
spin waves, 454
Quantum cornU, 581
Quantum interference, 366,
369
Quantum solid. 92
Quantum theory.
diamagnetism, 419
paramagnetism, 420 J
QullSlaystals, 48
QuSliI-Fermi leveb, 577
Quaslparticles, 294
·nl.'lIillg. orbital angular Schottky harrier, 573 Stress C()mponcnt, y J.
m"I1lCntum, 426 Schottky defect, 541 Structural phase transitions,
1:::,5&1 Schottky \-acancies, 551 393
Screened Coulomb potential, StnK-'1ure factor, 42, 63:>
282 bee IaUice, 44
ial cli~tribution function, Screening. electrostatic, 2.SO diamond, 51
521 Screw dislocation, 590 fcc IaUice, 45
1all effeet, 322 Second-order transition, 402 Substrate, 555
'atl<'rmg, 306 Self-diffusion, 547 Superconductivity, table, 336
dom network, 526 Seif-trappi.1g. 221 exotic, 156, 337
dom ~taI.'king, 22 Semiconductor, degene''l'Ite, type 1, 340
~ (",,11. ions, 423 2S6 type II, 340, 360
' .,; 'h ~ttenuation, 536 Semk<:mductor crystals, 174, Superlattices, 618, 630
r , allattice points, 38 199 Superparamagnetism, 477
,pnx,>1 lattice ''eCton, 33 SemiC()nductor lasers, 577 Surface crystallography, 556
omuination radiation, 321 SemiC()nductors, deficit, 222. electronic structure, 560
(,nshuction, 554 Semimetals, 174, 2.03, 228 nets, 558
lin~tion, 570 Shear constant, 94 plasmons, 279, 302
.,(.'("t! :rone scheme, 195, Shear strength, single reSistance, 171
233 cryrtab, 557 states, 562
<>d.mce, 293, 3(f1, 308 Shear stress, critiC'"I, 557 subbands, 562
ccUrm, normal incidence, Silort-range order, 621 transport, 564
33" Shubnikov-de I laas effcct, 583 Susceptibility, dielectric, 385
('I.'ti\'ity coefficient, 307 Single domai.l (Xlltides, 477 magnetic, 417
~!cti \-e index, '$)7 Singulm.-itics, Van Hove, 129 Symmetry opemtion, 4
I.Ul' ion, 556 Siugularity, ocnsity of states,
Tco.·:, 492 139
rooch, 492 Slater-Pauliug plot, 627 TA modes. 104
'lmall, 492- Slip, 5S8 Taxis, 478
Ll::3tion time, Sodium chloride structure, 15 Temperature, Deb)"e , 12.2
Ilgitmlinal, 490 Sodium metal, 142 Temperature dependence,
' in-lattice, 490 Soft modes, 400, 413 reflection lines, 633
.aneuI.'C, 470 Sular ('1'115, 573 'Ietrahedral angle, 2.5
~ta.,CC, surface, 171 SolidUS, 622 'f1lermal cxmducti\,ity, 131,
:.t:lI1ce per square, 171 Solitons, 473 166
~tHty, elechical, 155, Solubility gap, 622 glasses, 534
159 SpectroscOJ.lic splitting facto r, isotope effect, 138
sth-ity mtio, 162- .29 metals, 166
Jon~e, electron gas, 304 Spheres, metal, 171 lable, 126
lome function , 309 Spinel, 459 TIlermal dilation, 139
JI"r;l.tion, 516 Spin-lattice intemction, 491 TIlermal effective mass, 156
~ ED, 560 Spin-lattice relaxation time, TIlCrmal excitation of
,vn-Dushman .90 magnons, 455
.....:I"ation, 562 Spin wa\'e (sa tUM) magnon), TIler-mal e~pansion, 130
Y theory, 62S '50 TIlermal ionization, 226
in14 «>ordinate system, quantization, 454 '1nermal resistivity, phonon
3 1::> resonance, 506 gas, 133
I laSCr, 512 Square lattice, 'nlermion;c emission, 561
Stahility criteria, 9.') Thermodyuamics,
Stabilization free enerb'Y, 347 superconducting tranSition, 347
ration magnetization, 446 Stacking fault, 589 'l'hermoelectric effects, 227
ration polarization, 412 STM,20 TIlomas-Fermi approximation,
ration, rf, 516 Stokes line, 32.3 28.
Biug tunneling Strain component, 81 'nlomas-Fenni diclectriC
microscopy, 20 Strength of alloys, 600 function , 282
 fr"I"" 673

Three-level maser, 51~
light bi nding mcthod, 245
TO moclcs, HM
Tomography, magnetic
resonance, 342, 487
Transistor, MOS, 564
Transition, displadvc, 394
Ilrst-order, 40-2-
meW-insu lator, 284
ordcr-disonler, 394
second-onler, 402
Transition metal alloys, 624
Trallsition temperatu re , glass,
527
Transi tions, displacive, 398
Translation operatioll, 4
Translation , 'cctor, 4
Transparency, alkali mctals,
274
Transport, surface, 564
Transport equation,
Boltzmann,
Transverse optiC'<II modes,
plasma, 274
Tranwerse relaxation time,
492
Triplet excitcd states, 437
Tunneling, 364
JOSepIISOn, 366
Tunneling m io -oscopc, 5SO
Twinning, 589
Two-nllid model, 372, 376
Two-level systcm, 440
T)llC I superconductors, 340
Typc II supe.-conductors, 340,
300
Ultraviolet transmiss ion
limits, 275
Umklapp processes, 135, 162
Unit cell, 6
UPS, 324
Upper Clitical field, 661
Valence band edge, 199
Vall der \ Vaals interaction, 54,
56, 304
Van Hove singulari ties, 129
Van Vlcek paramagnetism,
430
Vecto r potential, 651
Vickers hardncss number, 606
Viscosity, 527
Vi treous silica, 524
Vurtex sta te , 340, 361, 376
Wann ier functions, 267
\Vann ier ladeler, 230
\Vave equation, continuu m,
112
periodiC lattice, 184
pe ri odic potential, 183
Wavevectors, energy gap, 196
We iss field , 443
\Vhisken, 604
\Viedemann-Franz law, 166
Wigne r-Seit:r. boundary
- condition, 250, 251
W ig ne r-Sei tz l.'CII, 8, 38, 251
\Ihgner-Seitz method, 241,
~lS
Work function , 560
work-hardcning, 603
XPS, 324
Young's modulus, 94
Yttrium iron garne t, 505
Zener breakdown , 231
Ze ne r tun neling, 230
Zero-puint lattice
disp lacement , 139
Zero-point motion , 55, 62
l.ero-field splitting, 510
ZnS structure , 20

Table of SI Prefixes

Fraction fufi, Symbol

1O- ' ~ femto r

10- 12 pico P
10- 11 nano
10" 6 micro "~
10- 3 milli m
10' kilo k
10' mega M
10' giga C
1012 tera T
 Table of Values
Q U-intity Symbol Value CCS 51
. ·IM :;'h
Veloc ity of light , 2.997925 10'0 em 5- ' U}> m 5- '
PrQton c hm-ge e 1.60219 10-" C
4.80325 ro-1a esu
h 10- 1'1 erg s 10-" JS
Planck's constant
'I = 1i/2r.
6.62620
1.05459 10-" erg 5 10- )0 Js

Avogadro's numbe,r .V 6.02217 x l C)"J 11)01-'

Atomic mass unit amu 1.66053 10- 10 g 1O-1'J kg
Electron re~t mass on 9.10956 10-- g 10- 3 1 kg
Proton rest mass .\1" 1.67261 \0-10 g 10-1'1 kg
Proton massfelectron mass Mp/m 1836.1

Reciprocal fine structure

constant t.del
"0 137.036 •

ElectTo n radiUS et/ mc l '. 2.8 1794 10- 13 em 1O- '~ m

Electron Compton
wavelength t.{mc
'. 3.86 159 10- 11 elll 10- . 3 m

nohr rlldiu s hI/mel

Dohr mngllcton eM'2m c
'.
".
5.29 177
9.274(0
10- 1 em
10- 1 1 erg G- '
10- 11 m
1O-t l J T - '
Rydberg cons tant me·,'2h 1 11 .. or Ry 2. 1799 1 10- " e rg lO- l$ J
13.6058 eV
II
94 '
I electron volt eV 1.602 19 10- 11 erg to-if J
eVIl! 2.4 1.97 x 10" Ih:
eV!I,c 8.06546 10" cm - ' 10' m-'
eV/kt/ 1. 160-18 )( 1Q4 K

Boltzmann constant
Permittivi ty of free space
Permeabil ity of free sp:u:e
..
k.

".
1.38062 10- 11 erg K- '
I
10- 23 1 K
lO'/4r,-cZ
411' )( 10-

s..,..u. B ~ T~,t ..... W H P... ~r •• ..,.j 0 -..; L.a"S.~b.." R~ Mod Pb)' 41. 375 41969) Sot< ....., E. R. CcI>nI.....d B S To,...,...1000".•
clPhy.""'.!t.nd Chem .....1 Rok'-'"r ou. 2 r~ l663 t1~3)