Name
Actinium Aluminum
Ame ri cium
Symbol
Name
Hafnium
Helium
Symbol
Name
Symbol
P,
A,
AI
Hr
Praseodymium
Promethium
Antimon"
Argon Arsenic
Am Sb
M
H. Ho
Ii
Protactinium Radium
A.
At
I" I
h
Radon
Rhen ium
..
Po
Pm R" R, Rh
Rb
Astatme
Iridium
Rhodium
B. Bk
Be
Iron
Kryp ton
F. K,
Lo
Rubidium
Ruthenium
Lanthanum
Lav.'Tcncimn
Bismuth
....00
Bromine
Cadmium Calcium
Californium
Carbon
Cerium
Cesium
Chlorine Chromium
BI B B, Cd C. Cf C C. C. CI C,
Co
L,
Samarium Scandium
Selenium
R.. Sm S. " SI
Lo.J
Lithium
Lutetium
Pb
:'lagnesiurn
\Ianganese \ Ienaele\ jum
\1en:urV
~ \lg
\ 111
Silicon
Silver Soc!ium
Strontium
" N. S,
S T, T, T. Tb
11
\Id
Hg
\10
Sulfur Tantalum
Techne ti um Tellurium
\loIybdenum
Keodymium
Nd
!'\e
Cobalt
Copper
Curium
Kron "'''ptuniurn
:\p
Tl:rbium Thallium
TII(lrium
TIlLl];lJIn Tin
C.. Coo
Dy
Klcke\
r\iobiurn
Dysprosium
Einstein ium
E. E,
Ell
Cold
." ,
Fm F F, Cd C, C.
"'I trogen ;':obelium Osmium Oxygen PalladIUm Phosphorus Platinum Plutonimn Polonium Potassium
No Nb N
:\0
0.
Pd P
PI p" Po
"
Th Tm Sn Tt
W
U
V
X, Vb V Z"
Z,
= . . . .
 ~
.. .
.,...~
... _.
..
H'
I.,
HT
Periodic Table, with the Outer Electron Configurations of Neutral Atoms in Their Ground Siaies
h'
U'
0,
Be'
",'
Na il
Mg "
The nolation u\cd to describe the elc(:lwnic configuration of' ~lt()m' ,md ioo, i' di~cu'~cd in all textbook., of introductory alomH; phy_ie" The letter~ $, fl. d, .. .\ ignify cledron~ having orhital angular momentum D, I, 2, .. . ill unih t\; the numher to the lei! of the letter denute, the pTlncipaJ quantum numher ot Olle urbit, and the ,uper~cript to the right dcnole~ the nl1mllcr IIi' electroll~ in the orhit
B'
2.\.2211
N'
0"
F'
Ne'o
2.~~2pl 2s72p~
AP]
SI"
P"
5"
CJ'1
At "
3,
K"
4,\"
'CaN "
4.s!
3I.'3p 3s'3p'
~"3p' :1.~'31'
3.,'3p' 3.I:'3p'
Brl~
Se"
3d
4,\,!
Tj irl
V"
:JrJl
4,~l
Ct"
3(/5
Mn"
Fe""
3d"
4.'
Co"
Nili'
Cu"
3d'~
Z,'
3el l " 4.' Cd"
4d"
5.~2
Ga"
Ge"
As]1
Se
K,'
3d'
4.,"
3d'
4."
3d'
4.'
Mon
4,'
3d"
4.~2
4,
Ag<l
4s 2 4p 4s'4 p 2 4s14p'l
4.r4,,
Te u
4s 24p' 4s'4p'
Rb 17
5,5,"
Y"
Z,
4cP
Nb"
Tc"
4d'
Au"
Rh'l
Pd'
In"
5,"
Sb"
5,~25p
I"
Xe o
,d
5,
CS~I
4d'
s.'
La l l
s.'
HIll
8a l 
'"
4d' 5.,
5,
Re fS
4d' 5,
OS'I
4d"
5,
Ir71
4d "
Pt"
4d'"
5.,
Au"
58 25p 58 251"
, 5.15,)' 58 2 5p'
Po"
5.~25p l
Ta 7']
W"
Hg"
5d'~
n"'
Pb"'
Bi")
At'S
An'"
4f"
5d
fu
5d'
fu'
Fr"'
"" Ra"
7.\.'1
fu'
Ac"
I'
"'.
S,P
.,'
5d'
5d'
6,'
5d'
""
41'
6.'
41'
Sri"
Sd'
fu
Gd
Sel'
6,
6.'
Ce4f'
6.,.2
Pro'
Nd"
4f'
fu'
Pm" Sm u E,"
4/,
Tb"
5d
Oy'
4fl~
Ha"l
4fll
6<1
7,
4f'
6.~2
7~
5d
Pa'l
41'
4f'
4fIJ
E..
Tm"
Yb7~
LU ll
4/"
4fH
4/"
5d
fu'
Th"
U"
S!", 6d
7 .'
Np
5f'
7,2
Am'!
"" Cm"
Sf' 6d
__ 7,.2
",'
6.\,1
Sk']
CI"
"" Es"

.,'
Fm'"
! Md 'o, NolO
"" ""
Lr '"
 ..:.:.
7~1
6(P 7 .2
Sf' 6d
Sf'
L:C.. . 7.v2

CHARLES KITTEL
John Wiley & Sons, Inc., New York, Chichester, Brisbane, Toronto, Singapore
J 5.1.1
 \Lir
/ '>410 o[)::
C c7
Copyright C 1953, 1956, 1966. 1971, 1976, 1966, 1996 by John Wiley &; Sons, Inc.
All righu resened. 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.
L Title.
0518445 e lP
Preface
fOf
This hook is the seventh edition of an elementary text on solid state physics 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 printingswhich it was time to pull togetherand a num ber ofnew topics besides. Signni<.: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, lightemitting 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 because, 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 possibility 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 clecii,
)
trOlls in onedimensional 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 phonons ("ould be considered as recreational reading. In a more gradual approach. the first eight chapters through the physics of semi<:onductors are read consel.'Ulively as a Onesemester 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 intended 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 CCSCaussian 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 parallei usage Simple. The dna! usagc in this book has been found useful and acceptable. 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 problems 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. Beasley, D. Chemla, T.C. Chiang, M. L. Cohen. M. C. Craford, A. E. Curon, 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 addressed to the autilOr at the Department of Physics, University of California, Berkeley, CA 947207300; by email to kiue l@uclillk4.l3erkelcy.cdu; and by fax to (510) 6439473.
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 101580012. Limited requests for permission to copy figures or other material should be addressed to the Permissions Editor at tllis address.
Contents
,
Cuide to Tables
Ceneral References
i. .i
1 2 3 4 5 6 7 8
9
CRYSTAL STRUCrURE
27 53
...
141 113
91
SEMICONDUCfOR CRYSTALS
FERMI SURFACES AND METAL't PLASMONS, POURITONS. AND POlARONS
OPTICAL PROCESSES AND EXOTONS
191
2"
269
10
11
12 13 14 15 16 17 18
SUPERCONDUCnvITV
"'"
333
POINT DEFFrvrS
.39
"",
viii
"" 58'
609
21 ALWYS
Appendix
A
C
631
634
638
643
646
647
H COOPER PAIRS
I
GINSBURGLANDAU EQUATION ELECTRONPHONON COLUSIONS
...
656
002
667 673
Guide to Tables
I.' 1.3
1.4 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 3. 10
1.I
10
Crystal
Struchll"C$
of the c1emCflts
12 23
24
Cohcsi\'c energies of the elements Mdting temperatures of the elements Hulk modulii and oompressibilities of the clements Properties of inert gas crystals Ionization energies Electron affinities of ncgati,'e k.tns Properties of alkali halide crystals Fractional ionic character of bonds Atomic and ionic radii Standard radii of ions 3. 11 t:lastic stiffness rorutants of cuhic crystals at 0 K and 300 K 3. 12 t:lasHc sti ffness constants of cubic crystals at 300 K
Dchyc temperatuTe and thermal conductivity Phonon mean free paths
57 58 59 60 61 68 73 76 78 79 91 92 126 133 1S<l 157 160 167 168 201 21' 221 223
224
Free electron Fermi ~urface parnmctcn for metals Electron heat capaci ty of metals Electrical oondocthi ly and J"e$istivity of metaL,; !lall coefficients Lorenz numbers
Energy gaps in semiconductors Elfccth'C masses of holes and e lectrons
8.3
8.4 8.5 8.6 8.7
of semkonductOrlii
225
229
;,
10. 1 10,2 10,3 lOA Ultraviolet transmi~sion limib of alkali metals Volume plasmon energies Lattice frequencies I'olaron ma:sscs and coupling COI1~lants
11.1
11.2
11.3
12. 1
1~2
Exciton (,lulling energies Electronhole liquid pm'1lmctcJ'li Acronyms of current expelimcntal methods for Land structure studies Superconductivity parameters of the clements Superconductivity of selected l'Ompounrls Energy gaps in superconductors Isotope effect in supcroonducton Coherence length and penctratjon depth
Ekctronic polari:vlbilitics of ions Ferroelectric crystals Antiferroclcctric crystals
12,3 12.4
34. 347
353
12.5
13.1 13.2
13.3
14. 1 14.2
Magneton numbers of L"1nlhanide group kllls Magneton numbers of iron group ions eritia! point c.>;poncnts of fcrromagncts Ferromagnetic: crystals Antifcrromab'llctk crystals
Nuclear magnetic resonance data Knight shifts in NMR
15. 1
15.2
15.3
Hi. I
Diffusion cons tants and <lCI; ivation cnergies Activation energy for a prn;ith'c ion vac:.\ncy Bpcrimental F center absorption energies 'lop layer atomic TClaxation Eiedn.m work functions Elastic lim it and shear modulus Dislocation dcmities E lectron/atom ratios of electron compounds
545
547 549
558
56.
588
598
20.1 20.2
21.1
616
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
Lattice translation l't.'C:tors
3
4
5
6 8 8 10
INDEX SYSTEM FOR C RYSTAl. PLANES SIMPLE CHYSfAL STRUCrURES SOOium chloride structure Cesium chloride structure He~ooal closepacked ~truclurc Dian)()nd structure Cubic zinc 5\LUidc structure
lZ 15 15 17 17 19 20
DIRECT IMAGING OF ATOMIC STRUCrURE 20 NONlDEAL CRYSTAL STRVefURES Random stacking and poIytypism
CRYSTAL STRUcrURE DATA
21
22 22
SUMMARY
PROBLEMS
25
25
25 25 25 25
REFERENCES UNITS, I
A=
(.)
<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) . o (b), but d;ff"re"nt <T>... ~I &.:cs are Je..~Iopcd. (e) CIea," k!
Inp; a
"'j..... 1
cf ~ I.
Solid state physics is largely concemed with crystals and electrons ill crystals. TIle study of solid state physics began in the early years of this century following the discO\'ery of xray 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 threedimensional 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 presented to the Bavarian Academy of Sciences in Munich. In the first part of the paper, Laue developed an elementary theory of the diffraction of xrays by a periodic array. In tile second part, Friedrich and Knipping reported the first experimental observations of xray 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 condensed matter physics, and it is now the largest and probably the most vigorous area of physics.
An ideal crystal is constnlcted by the infinite repetition of identical stmctural 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 smallest 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 basis; when repeated in spaCe it forms tile crystal structure.
'R. I. ~l a"'Y, .sai .fUM thtone sur Ia $/n,cture des cristoux, Paris. 1784; TroiM de cri$laliographie. Parl~. 1501. "For penonal acroU'l" of the early years ofxray 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
I.
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
where
tll> 1/2.
are arbitrary integers. The set of points r' defined by (1) for all
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 abstraction; 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. However, nonprimitive crystal axes are often used when they have a Simpler relation 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 xray 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 translation and point operations. Textbooks on crystallography are largely devoted to
Cryol<l/ St~!ldure
~
"
.
~.,
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 plimiti'"(" trwoslalion vectors of the tu"Odimt'nsiOOlli iatti("f:.
"
0 0
0
n :don a ..... , and i'....lf. "mis QCC:tln
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 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 indicated b y dots ill Figs. 2 and 3, but in Fig. 4c the dots arc omitted.
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~1lion 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
Crylt(ll
StrlOCturfI
. [] / 7 ., r~: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 parallelograms I, 2. 3 are equal in area and any of tile," wuid be taken lUi the primitive >11. The parallelogram 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.
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 rSeil>: 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 WignerScitz 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.
TwoDimensiolUd Lattice lypes llICre is an unlimitoo number of possible lattices because there is no natural 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.
Cryotal Struclure
Figu.e 1 It f1w:fold "~i$ of ~)'I)lmctl')' cannot exilil in It periodic lattice ~ It is not IlOi!Iibl.. to fill t~ artII of a p~ ,,11 h connected array of pentagons. We urn, how~er. IlII all 100 area or a plane " ith just to.'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 21T1 27T/4. or 21T1 or under m irror renection. 'Ve must impose restrictive 3. 6. 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.
System
T,iclinic
]'.Ionociinic OrtllorllO Lnbic
a""{3
2
4 2
a, #ala3
a = y=fU .F{3
a,;o! a 2;o! a3
a={3=y = OO"
Tctr.lgonal
Cubic
a . = al""a. l
a = {3 = y = 9O"
a L  a ~= a3
a={3:y"9O"
Trigonal
al = a2 = a3 Q = f3 = <
a, = aZ""a3
a = {3 = 9C1'
y =
1 20~
"
!2
~
r~EJ
(rl) c..nlHed red.",... lar bitloe;
UPS
"
,
"1 ~1aItn
'~Lt .. JootI;
.110"
prim;"';'"
~h
are ........" lor hath tIM! ",,11 .I for lhe r<:<.~anguW llbi! celL for
[a,I" laJ;
rp 9(1.
Figure 9
.'
lattit~.
The noll.
~howll
"
TaMe 2 Olaractenslics of cubic
Simp":
1311io::c~
11o<kcenlered
Fatt(.'Cn tcred
VoluHle , oolwcntiouru cell LaUK."C potllts per cdl Volume, primit ive cell Lattice points per unit vlllume
n'
, "
In'
" 4
In'
4/'
12 (l12 i12 6

" Un' 6 ,
21fl!O
V,,
8 3 111012 = 0,86&
0_7ff7o
I'
I.
= 0.524
bYJ
 0.680
hrV2
=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) laltkc, til(' bodyccntcrcd 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 loordinates of the body center ofa cell are iii. and the face <'Cnters include BO O IO , H; i. 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
The orie ntation 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:! .
or
CrydtU Slruclu ..
F"ogure It Bodycenlered cubic Iallice, .bowing 11 primitive lcll. erimith'e a.'Ulihown is 11 rhomoobe<lron edge i 0, and the IUlglc between adjaemt edga is 109"28'.
or
n.e Va
Figure 12 Primiti~'e tranWation vectors the bodyo::ntered cubic lattice; these "cctOrii connect the latlice point al th e origin 10 latlice poinU; al the body cen teno The primiU"e cdl is obtained on completing the rhombohed ron. In lemu clthe cube edge" the prin';tin" tran.Jalion vectors arc
or
.,  IOI(i+y f.) ;
aa  Io( i+y+i} ;
aJ  lo (i  ~+i).
Figure 13 TI,,:, rhombohedral primilr.e cell of lhe facccen Icreel (:ublc cr)'l;lal. TIM' primitive Irnnslalion ve~1ors a" aa. a 3 con .....d the Ialtia.' point at the OI"igin "~th IaUice poinU; al the face l't'llten. As drlMTl. the prim(t;\'e "eclOn a~: r  Ia{i+t) ;
sian unit ' e<."tOIS.
Figure 14 J\t>lalion clille primit ive eel in the be~agonol $y$lem (lleavy lines) 1< a pri.ttl or hexagonal .ymrJlCtry. lien
OIl 0 , " ' 03
aa  Io (t+t) ;
.~  I ,,(i+i)
n.e angles ber....'I'Cfl the a.,,~ure 60". lIere 1, t, t. are the Carte
..
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. " smal lC!>t Ihr~ inlefo(en ha" ing nlC 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 nclosed 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 irection , referred to the axes. The al axis is the (loo] direction; the a2 axis is the
C'1l'lal Struct"re
15
(100)
(J 10)
i
(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.
We discuss simplc crystal structures of gener.d interest: the sodium chloride, cesium chloride. hcxagonal closepacked , diamond, and cubic zinc sulfide structures.
1l0;
III
001;
101 ; 010 ;
01/ 100
16
,
,~~~,,~.. ~ .. . !:~':
j .
,
'.
t;"\ ,.....
j.
. . ..~ ...
l 
'
. ...'..:. '
.
.
. .;.~
; '.
.. '
. .....
. ,.
.::."
.
0 1'0 ; '. ' .. .:'.' . . ;'.. , . ' . '. ' f!'.~ ,.. ,..:. ".'. 4!t ,. ,.. . ....,. ... ~ ..~ . ..':"
.
t'igurc 11 We ",a~ ~'()nstr"t1 the sod ium chloride crys" tallilnoct"n: b) alT'anl(ing and CI ions alternately
N,,
. . ..
..
III tho: lattice pointb' of 11 simpk:: cubic bltice. In the ~"li tal cach ion iii su rrlllmded by . ix nearest neighbors oft!.., opposite d wgc. "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 htt", nrc n...!,,,,,d on rd~tioll 10 th e (.ell In o, dcr to clnrify tht: sp"liai armnt;(."'''nl.
.'
..
"' igun. 20 7hc ccsium d llorirJc CJ}"S1a1 structu ... 1lx 5pat'"e lattke i~ ~inlple rubic. and the ln~is I,.." onl Cs ' lOll at 000 and un" CI ion at
III.
,
Each atom has as nearest neighbors
si~
1 CrtPfol S,ru<:,.....
17
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.
C",,\4\
2.70 A
2.88
Cl)"tal
UHg NH...CI
nBr esCI
3.29 A
3.87 3.97
nl
4"
4.20
IS
"
fo'igure 21 A closeP"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 closcp,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
I','
"
<>:
a
t' igurc 22 lure. Th" atom po$ition~ in Ih~ ~lru<.111rc do not C(ln~titllie a sp:wc lalli,..,. l'11e space lattice
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,
whcre a is in the basa.l'lanc and c is the magnit "de of lhe axis 83 of Fig. 14.

II
II
t'ig,,~c 23 The p~imili"" L..,II has u, ~ "2. wilh an incl"ded angle of 120". The c axiS (or R,) i~ oo~mal 10 the plane of 8, and R. TIlt' 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:;ilion ~ '" la, + Ia. + III;).
'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 closestpacking 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 nearestneighbor 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:::::, :
I , , , ,
 a
~/

,": ,
: : , , ,
, ,
1 ,
i ,
____
"i~'re 24 Atomic po6ltioni in the cubic reO ci the diamond "ructure projed .. on /I cube m ; f'racI:iOIls Ilcnote height d .l.Jo,.e the base in unit, of IIl1,bc t:~lge. "The poinu Rl 0 and I are 00 the fcc Lotl;':e; those at f and f are on II similar IlIttk'C di~placffi along the body diagonal by onelOurll, of it51t:ngth. With ~ It'C space lattice, the IJUSis CQOlsi~ts r;l tv.'o idt!ntical moms at 000; HI.
Hgu",
m'.'Tlt .
Cry~ lal
structure cl diamond,
.
, ,
I
:
','
nearestneighbor honds per atom, there \I,Quld be no difference in energy between the fcc and hcp structures.
Cl)'slal
da
1.633
1.581
C!')"l1ll1l
lie Be 'Ig T;
Diamond Structure
Zn
Cd
Co l'
da 1.861 I S86
I.fi22
Crystal
da
1.594
1.592
Z,
Gd
Lu
1.623 1.586
1,586
t.57IJ
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 0.31. which is 46 pcrccnt ofthc filling'" 'tor for a closestpacked structure such as fce or hcp. TIle
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 structure, ",;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 centerofinversion symmetry operation at the midpoint uf evel)' line between nearestneighbor 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 C uF SiC CllCI ZnS AlP Cal)
"
Crystal ZnSe
C"'"
AlA,
CdS InSb
Agl
6.47
'(be close ccJuaiity of sevcral pairs, notably (AI,CalP and (A I.CalAs, makes possible thc construction uf semiconductor hcterojunctions (Chapler 19).
Direct images of crystal structure have been produced by transmission electron microscopy. Perhaps the most beautiful images arc produced by scanning tunneling microscop); in STM (Chapter 19) one cxploits the large variatiuns 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.
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 .)
The ideal crystal of classical crystaU ographcrs 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 tmesec 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 closepacked planes of atoms. The structures differ in the stacking sequence of the planes. fcc having the sequence ABCABe .. . and hcp having the sequence ABABAB . . Structures are known in which the stacking sequence of closepacked plane5 is random. 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 closepacked 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 mechanism that induces such longrange crystallographic order is not a longrange 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 structUres 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.
\
\
H '~
ho,
3.75 6. 12
T.ble 3 Crystal structures oCthe elements The data given are at room temperature Cor the most common ronn, or I t
the stated temperature in deg K. F or further descriptions or the clements see Wyckoff, Vol. 1, Chap. 2. Structures labeled complez are described there.
hop
H e' 2K
hop .
. u'" B
boo 3.491
~a !oI(
2.27 3.59
....
B
4.05
...... ...
C
3.567
N "" 0
S."
(N:)
4.225
'"
M , hop
3.21 5.21,
, , ,
So
3.31 5.27
_.
Tf
2.95 4.68
, , ,
NI foe
3.52
Pd
AI f"
......
SI
5.430
K"
'" 5.225
C. f"
S."
""
""
C,
M,
tIIb~
F.
2.87
Co
2.51
4.07
WIIplu
'" ""
C. f" A, foe
l'
ho,
2.66 4SS
G.
ClGI1l~lu
G.
dllmond
'hoInb.
3.61
5.658
..
<0,)
3.57 S.83
N. ~
foe
4.46 Ar 'K
f"
5.31
A.
S.
B,
Kr 4K
foe
S.64
Xe 'K
(Brl )
'"
boo
Rb " S, foe
608
Y
hop
l,
ho,
3.23 5.15
Nb
boo 3.30
Mo
3.15
T,
ho,
2.74
R.
hop
5.585
C. "
,04S
B.
...
3.65 5.73
'"
4.0
2.71 4.28
3.80
Rh f"
Cd
ho,
2.98 562
I,
telt. 3.25 4.95
foe
3.89 PI
4.09
.. ....
6.49
Sn (a)
Sb
. ..."", ...
T. I
(11)
f"
6.13
Hf
hop
T.
boo
W
boo 3.16
'"
R.
h".
R hq>
2.76 4.46
O.
ho,
2.74 4.32
I, foe
3.84
foe
3.92
A. fe<
4.08
5.02
3.77 ABAC
3.19 S.OS
3.30
H, ......
TI
ho,
3.46 5.52
Pb
foe
4.95
BI
E<
Po
3.34
At
R,
"
l.
hop
F,
foe '"
I"
5.31
C. f"
5. 16
h".
3.67 ABAC
'"
Nd h".
3.66
Pm
N,
.,
P.
Sm
..
E.
boo
Gd
ho,
3.63 5.78
Tb
hop
Oy
Ho
hop
hop
3.59 5.65
ho,
3.56 S.59
Tm hop
3.54 S.56
Yb foe
S."
4."
Th foe
5.08
Po
tetr. 3.92 3.24
3.60 5.70
3." S.62
3.SO 5.55
"',. 3.64
ABAC
Am
Cm

Bk
Cf
E .
Fm
Md
No
L,
 . .'. 
.......
eJ
,
H"
0.088
He
2~
0.205
Itllltm)
li
78J(
B.
1.82 12.1 2.22
B
2.47 13 .0
C
3.516 17.6 1.54
N ~
F
1.44
1.03
."
0.910 1.402 4.525
, ,
So
2.99 4.27 3.25
Co
8.93 8.45 2.56
C,
1.53 2.30 3.95
TI
4.51 5.66 2.89
C,
7.19 8.33 2.50
M,
7.47 8.18 2 .24
F.
7,87 8.50 2.48
..,
Co Rh I,
51
2.33 5.00 2.35
CI 93K
Ar
4K
2.03
2.02
NI
8.91 9.14 2.49
Z,
7.13 6.55 2.66
G,
5.91 5.10 2.44
G.
5.32 4.42 2.45
As
5.77 4.65 3.16
S.
4.81 3.67 2.32
Br l23K
Kr
41(
8.97 2.50
4.05 2.36
Y
4.48 3.02 3.55
Z,
6.51 4.29 3.1 7
Nb
8.58 5.56 2."
Mo
10.22 6.42 2.72
To
11.50 7.04 2.71
Ro
12.36 7.36 2.65
Pd
12.00 6.80 2.75
pt
Ag
10.50 5.85 2.89
Cd
8.65 4.64 2.98
I,
7.29 3.83 3.25
5,
5.76 2.91 2.81
Sb
6.69 3.31 2.91
T.
6 .25 2.94 2." Po 9.31 2.67 334
I
4.95 2.36 3.54
Xc
41(
B,
3.59 1.60 4.35
l,
6.17 2.70 3.73
Hf
13.20 4.52 3.1 3
T,
16.66 5.55 2."
W
19.25 6.30 2.74
R.
21.03 6.80 2.74
Os
22.58 7.14 2.68
Ao
19.28 5.90 2.88
Hg227 TI
14.26 4.26 3.01 11.87 3.50
Pb
11.34 3.30 3.50
BI
9.80 2.82 3 .07
A t

R,
lo
'.84
3.39 3.43
3A6
F,
R,
Ao
10.07 2.66 3.76
I"
C.
6. 77 2.91 3.65
P,
6.78 2.92 3 .63
Nd
7.00 2.93 3.66
U 19.05 4.80 2.75
Pm
Sm
7.54 3.03 3.59
Eo
5 .25 2.04 3.96
Gd
7.89 3.02 3.58
Tb
8.27 3.22 3 .52
0,
8.53 3.17 3.51
Ho
8.80 3.22 3.49
E,
'.04 3.26 3.47
Tm
9.32 3.32 3.54
Yb
6.97 3.02 3.88
N.
20.45 5.20 2.62
Th
11 .72 3.04 3.60
P,
15.37 4.01 3.21
Po
19.81 4.26 3.1
Am
11.87 2.96 3.61
Cm
Bk

Cf

Es
Fm

Md
No
l'

"
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
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 osepacked 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 cllly packed atoms. the planes being klosely stacked.
References
ELEMENTARY W. 8 . Pear$<,)n. CryslDI chnnisl'1l oOld physics of rmtDb Dnd Dlloys. Wiley, 1912II. I) . MegII..... Crys/lll structures: D workillg (/JIf'f"06Cn. Saundcn. 1913. CRYSTALLOCI\A.PIIY
M. J. Buerger. Introduaion 10 Cf'JIStlU g<!Omet'1l, McGraw mll , 197!. G Bums allil A. M. Claser. Spoce groups for >lid IIDle physldJts. Acatlcmic. 1918. F. C . Phill.t, An introauclion 10 crysIDIlogl"lll"'fJ. 4th .:t.I . Wiley. 1911. A W,."j place 10 begin.
II J
B K. Vainshtein. Modern cn;ltllllogrnphy. Sp""ger. 1961. J. F. ,,)..., Physiall proprl"fon of cn;stDls, Oxbd. 19M.
CRY~TAL
.
Series, Jourrn'l\ ufCrystai Growth, includes procet."<.Iillgsofthc International Conference, OIl Cry~I9.1
Croo.1h.
Springer Seri.,. , Crystals CrO'.''th, P'roprrtie,. and App\l(:aIiollll. CLASSICAl. TABLES AND I lANDBOOKS
InterTllltklno / tablu jor "'ray crJI.It(lUogrophV. K)~,ocIt Press, 4 volumes, Birmin!. m, 1952 19'74. .... J. F. Nye; Phll$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$>U,
Amer. Soc. Met als, 3 Yols., 1985. A. F. Wells, Structurol il'\Org<'nk dremtst'll. 5th ed., 0d0n1 Uni''''mty Preu, 1000, 19S4c:. W. C . 'V)"Ckolf, Crystal "ructura. 2nd ed., K......ger, IQliI.
2
Rec;iprocal Lattice
DlFFRAcnON OF WAVES BY CRYSTALS Bragg Jaw SCAlTERED WAVE AMPUTUDE Fourier analY5is Reciprocal lattice vectors Diffraction conditions
Laue etjuations
29 29
30 30
33
34
36
BRILLOUIN ZONES
Reciprocallallice to
c
31
laUice
4()
FOURIER ANALYSIS OF THE BASIS Structure factor of tile bee lattice Structure factor of the fcc lattice Atomic fonn factor
QUASICRYSTALS
SUMMARY
'" " 42
...
49
4. 45 .5
PlmBLEMS
1. lnterplanar separation 2. Hexagonal Spite<! bttice 3. Volume of Brillouin ;tone Width of difTUClion maximum 5. Structure factor oC diamond 6. Form factor of atomic hydrogen 1. Diatomic line
51 51 51 51 51 51
., .,
5.
REFERENCES
'n ..........
.0
1 1'
.5
[ IedIl)l'lS
i',
."
Figure I \\'~ ..d eogth ~'ersU$ parli. de eIlCf'gy. lOr pho lom, lI.,uITOtl S, and dectroflS.
Figure 2 Dt:rivation of the Bragg C<lualiOll 2d jim 9  nA . here d is the .pacing ri parallel atomic 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 .
n,.,.
28
'JAPTER
2:
RECIPROCAL LATTICE
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 ordinary 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 directions 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 reproduces the corn'Ct result. Suppose that the incident waves are reflected .. spccularly from parallel planes of atoms in the crystal, with eaeh plane refll.'Cting 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""aUering. in wllich the e nergy of the xray 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 interference of the radiation fro m successive planes otX"Urs when the path diflcrence is an integral number 11 of wavelengths A. SO that
(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 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 Braggreflected 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 hery 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 interesting 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
(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:
11(.1'
II
+ a) = "0 + :rIC,. cos(211pXia + 2'7Tp) + S" sin(2'7Tpda + 2'7Tp) = no + :rIC" eos(2'7Tpda) + 5" sin(2npxJa) = fI(X) .
(4)
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.
,
I
31
  'lMA
I'~
u.o.ooo~
' 00
....
~  I
.. A
.. ..
. I 16.l
....
1IlOi . 
..
BruOOfOlt'
'" ..
"""In
figu" J Sketch of a nl<XlOChromator which by Bragg rc flt'ct loll .ei.'d S a narrow sp<.'ctnlnl orxray
nelliron " ....,,,lengths from a broAtl splru m ludden! beam. TIle up.,.,' part oflhe figure shows the analysis (obtained by rc:flt'ctiotl from a SOnd 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 . ~,.)
<X'
Figure of Xray oJiffllldomcter rewnling cl pOW<\ered sili(ml, dMJWing a diffracted beams.. (CclUrte$Y of w. Parrish.)
~unter
R'COroing oflhe
Figure:;: A periodk: function n(l) of period II, Qnd the tenns 2fr{J1a that may appear in the Fourier tram/orm n~) .. I nl' eo:p(I2"F"a). The magnitude. 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' cocmcit' llls II" now arc complex numbers. Tn e nsure th"t nC ) i5 a real fundion. we r require
(6)
11_,..
r and
+ i sin
!p)
+ "  IA IP COS
I sin
\pI
= (II,.
(8)
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) :::
L lIe; exp(iC r ) c
(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 xray 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
(10)
L ",,"f" dx exp{i2'1T(J)'I'
~)
JI ).rJa]
(II)
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
(12)
. (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 orthogonal 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 defint..>d by (1.3) . From (9), nCr
+ T) :::
L nc exp(iC' r) exp(iC" T) C
(J6)
(17)
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 concentration 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 :::
(18)
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.
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 foctor 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 C
6k)' rl
(20)
the argument of the exponential vanishes and f' = Vnc;. It is a simple exercise 'llmcantly from (Problem 4) to show that " is negligibly small when .6.k differs sib 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
+ G ::::
k' ,
50
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 orientationor 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 SULh 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 dimensions (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 the origin of the recil'rocallaUiCC'. The rt'<.iproc."3! lat Ii<..'e ,.,.,IOf C c conn<.'C.1S pointl OC, lind C" con n('(.is 00. Two plant'l I and 2 ~ dra"'" ",Ilk'" .are the p" ....
pendicular biSC'<'tors of ~ and Cn, rt')Jlectf\"(~I )'_ An y ,edor from the origin to the plane I , suchlU k" wiH
Figure 9h Square ~iprocal wtK'e "il" recip1lX:'a! lalliec \'CCIors ~hown as fine black lines. Ole Jines shown in white are perpendicular biSl,.'(io/"$ of the reo
volume about the orqon which is bounded en lin~ly by ,.mile li l\Cli. The square i$ the W;gnerSE'ilz p rimili'e rell of Ihe rec;l'ro<.~1 IaltlCe. It is "'llL1ed the IIrst Boliouin 7.0111'.
satisfy the difTra'1ioo ",cl it ion Ii,' (ICd  (lGrjl. Any ndor from the urigin 10 the plane 2, ~Udl as ~ will sal i~fy the dilTl"lIClion condi lioo lei ' (lel)) 
(lC"t.
A Brillouin zolle is defined as a WignerSeitz plimiti\'e cell in the reciprocal 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 xray beam in the crystal will be diffracted if its \Yavel'ooor k has the magni'CIion requir<.>d by (2:6). The diffracted bt:am will then he in the tude and din_ dirt.'ction k  G, as we see from (19) with Ak =  G. l1lUs the Brillouin construction exhibits all the wav<.,,ectors k which <.'an be Braggrc(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 WignerSeitz 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,
'.
.....
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.
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 COIlsh'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 xray diffraction analysis of crystal structures, but the zones are all eSSential part of the analysis orlhe electronic energyband structure of crystals. The spt.>Cia1 utility of the first Bl"illouin zone is dcvc!opt..(I in Chapter 9.
Reciprocal Lattice to
nlC
3C
Lattice
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 le ge 211'10 and of volume (27Tla'f; this cube is the d first B..iIIouin zone of the 5C crystal lattice.
la fI2:
)(
aJl = 1a3 .
(29)
111t: primitive translaltollS oftne reciprocallaUice are defint.'d by (13). We have, using (28),
h.
(211'Ia)(5' + i ) ;
= (2m'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 =
+ (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 (....,bic buice. The figure Is I. regular rhombic
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 oneeighth of each of eight corner points. In solid state physics we take the centr.d (WignL"TScit"..:) cell ofthe reciprocal 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 12fac<.>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)
~a(y + i); ~
V =
=! a(i + 2);
la, . a
2 )(
(34)
aJ ""
in'
(35)
nle primitive tra nslation vectors of the lattice recip rocal to the fcc laUice hi = (2'ITla)(i
(36)
.,
r 1
"
(2maX2i) ;
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; (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 = C of Eq. (21) is satisfied, the scattering amplitude is determined by (18), which for a crystal of N cells may be written as
.6."
Fe =
NSe
(:39)
l11C quantity Se is called the structure factor and is defined as an integral over
Rtciprocd Lufiu
43
Figure IS BrillOllin zones of the r:lCeC'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. d tion of electron concentration functions nJ associalt.. 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 electron 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)
(42)
We combine (4 1) and (42) 10 obtain the structure factor of the basis in tile form
5c =
L h exp(/C' fJ)
J
(43)
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 +
= 21J{Vllj
z.,a:J
+ '1!YJ + V:JZj)
(45)
(46)
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
x, =
!Il ==
(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 V I
+ 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 contain 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 . bodfoentcred 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.
(48)
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 multiples 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.
~
.u
...
~~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 (. iones "re Il<jual, The scattering ampli1 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 monatomic' ~imple cubic bllke 0( lattice constant 012. Only C"cn integers O(:CU' in the renmiol)lI ;"dkes when these ""' ba5ed on . cubic lattice on att ire coostanl D. In KB. the form factor of Br is (Juite differellt than that of K'. and all ren ..",!ions of thll fcc lalti<.e are present. (COIlI10:5)' of R. "a" No rdstraml. )
"
h """ f
f11e scattered radiation from a single atom takes account of inte rfere nce
effects within the atom. We defined the rorm factor ill (42);
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) r _ c 1C ,. "" 211" f d,. ,.211,;( ,.) ~;.;;'
ec
,C,
after integration over d (cos a) betwee n  1 and L TIlUS the form factor is given by
si.'l' C,"
C,
(50)
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
:<%
(51)
Rcciproool
umia
47
"
Ilartrco; t'oo:\: <;>\.; ..... _
A 0: 109
" ,
f
(111)0
0
""OJ
on
0
"'"
1 m('
{3JI}O 0
(~20) (42'~
(.5 11 )
.0
,
0 .'
"
0,
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 indkel 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 amplitude 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 xray crystallography. Vol. 3. The overall electron distribution in a solid as seen in xray 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 Kray 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 CQworkers find agreement within 1 percent in a comparison of the xray 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 xray 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 nearestneighbor atoms, there is an appreciable increase in electron concentration over what is e"pected from the overlap of the electron densities 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 cannot 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 nucleation 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 longrange orientational order to the structure. It is perhaps surprising 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 electrical COllductors; they are ncarly insulators with a somewhat welldefined 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 periodic 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:
+ 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 longrange nonrandom order to the structure. and the longrange order gives a diffraction pattem, which appears split off from the pattern defined by the shortrange 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< OrJCntationai order 8.l\d the kmgl1ongc nonpo..Tiodic order arc shown.
will appear to be clustered and spread Ollt (broadened). l1le diffraction pattern of a threedimensiollal quasicrystal is quite different. however; the pattern is well defined and tan have the fivefold symmetry by which quasicrystals were first discovered. A computergenerated diffraction pattern with fivefold symmetry is sh~wn in F igure 20.
SUMMARY
.6.1.:"" G.;
Laue conditions:
~ .6.1.: =21Tt1
lIere a ..
A:z, 83
v,. VJ!.
li3
'Ole scattered amplitude in the directK>1l 1.:' "" L: tional to the geometrical structure factor:
+ .6.1.:
== k
+G
is propor
Sc  k
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) ""
n le flrst Brillouin zone is the WignerSeilz 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.
I'irst B"illouill zone Cube Hhombic dodecahed.on (Fig, 13) Truncated octahedron (Fig, 15)
','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
+ (af2)Y
11t
(a) Show that the volume of the primitive cell is (3 I2p c. (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 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)]
"1 ~
.. 0
IFlt
(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 threedimensional crystal.
5. StruC/UTe factor of diamond. The crystal structure of diamond is described in Cilapter I . The basi3( sists of eight atoms if the cell is taken as the conventional cube.
"
~ , '" !. i ..,
(I ' I )
.~
~ .., 
.."
f z
 o.4S'
0.65" 
I
P) H , \0.70"
.,.
Counter .,.,.11000 2f
....=i\::
(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 ind 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 xrays is pcrpendiC\Jlar to the li ne oiatoms. (a) Show that the interference condition Is 1'1" "" a cos 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.
or
e,
References
XHAY 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 structures . M. J. fluergcr. ContmpOrcry crydci/ogrul'h!l , McCr.l.wlIi1l , 1970. A Hne introd"ction. B. E. \\'''lTtln , Xrll!l diffrlldlct1 . Addi.oo\\'Cli]cy. 1900.
C . JallOt. Qt<mNryslIILr: 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, 201 ( 1988).
R
NEUTIION DIFFRACTION
G . E . Baron. Neulre diffraction. 3rd <!d., Od"ord , 1975. m 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
OOVALE"~r
00 72
CRYSTALS
7.
77
ATOMIC RADII
ANALYSIS OF ELASTIC STRAINS
80 83
87
SUP.tMARY
PROBLEMS 1. 2. 3. 4. 6. 7. 8. 9. 11. 12. 13. 14.
Quantum solid Cohesil'c energy of bee and fcc neon Solid molecular hydrogen Possibility of ionic crystals R+R ,
"
.,
!l3
00
., "'
93 93 !l3
U2
Cubic znS structure Divalent ionic cl"}'$tals Youngs modulus and Poisson's ratio Longitudinal wave velocity Effective shear constant Determinantal approach Cenernl propagation diredioll Stability criteria
.3
.3
94
REFERENCES
""
CJ).blb" .. a.g<II1
(..... de.. Waal.)
.j
Fiplrfll priflci~ typr$ of t'1')'1ital1ifl<l bifldi'!& 10 (al neulrallltOfflS with ck>sal ..kctroo 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 10 the h.logen atoms, and the ns re.ulting io<lf ""' heM tflg.,ther by IIttr.>di,"C electrostatic fi:wt.es beI"''fIffIII~ posilhe 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 pruts of lh';;r electron t1i~lnbut;onli . ~
n.e
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 situations: exchange energy, vall der Waals forces. and covalent bonds. TIle observed 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 hemeen different columns of the periodic table. The inert gas crystals are weakly bound. with cohesive 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 summarh.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 closepacked (fcc), except He3 and He"'.
'Zeropoinl ,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!;tIldghbor dista.nce. The heR\"ier the Rlom_ the less import&llt ~re the :teros (>Oint effects. lf we olTlit :wropoinl 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"Opoint 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. WaalsLondon "Iteraction
Consider tv.'O identical inert gas atoms at a separation R large in compariSOil 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 electrostatic 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 moments in each other, and the induced moments cause an attractive interaction betv.'Cen the atoms.
Table 1
LI
Be
320. 3.32 76 .5 145. 1.51 34,7
158. 1.63
37 .7
Energy required to form separated neutral atoms In their ground electronic B e N state from the solid at 0 K at I atm . The data were supplied by Prof. Leo 561 711 474 Brewer in units kcal per mole, revised to May 4, Hm, after LBL Report ~~1 1:3t.: 4.92'; 3720 Rev. 134  170. 113.4 kJ/mol = =';'=====Oe:VlatOm kcallmol
V
0 251
o.s;c
F 81 0
N,
107, 1.113 2567
M.
.~,.,
J43."
A,
I"
K,
oaol
K
901
OJi3~
Ca
S,
T;
C,
M,
21.54
178 . 376 468 . 512 395. 282. 1.84 3.9~ 4.85 sr3f.... 4 . 1 1f ~.92 ..... ~ 425 899 11 1,8 122.4 945 6 7.4
413, 285 ,3 237 118. 11 .2 424 :.... 428 . 336, . .E0__ _ 2Z.L. 372 4.28' 4.39 4.44 .... 3.49 1,35 2.81 3.8~ 235s 2.4F T22 O)ftli 987 1013 102 .4 80.4 31 .04 648 88.8 68.2 56.7 28 .18 2.68
Fe
Co
NI
C,
Z,
G,
G,
S,
B,
Rb
822 0.852 1964
S,
166
Z,
Nb
Mo
T,
R,
Rh
Pd
AS
284.
Cd
I,
243.
S,
Sb
T,
X,
1.72
39 .7
661 422 603. 730 . 658 650. 554. 376 4.37 6.25 7.57 6.82'" 6.85_ o.74~ 5,75_ 3.89 100 ,8 1442 174.5 157.2 158. 155.4 1325 89 .8
303 265. 211 107. 15.9 "2 '2.95"" JJ13 232 3.1. r 27~_ 2:"1 9 rfr" 1).'1"6"' 58 0 26.73 58.1 724 63.4 50.34 ~ 2562 3.80
C,
776 0.804 1854
Ba
183, 1.90 437
T, HI W 431. 621. 782. 859. 4.4.1 644 8l0 8.90_ ..... 103 1 148.4 186.9 205.2
"
R,
775.
0,
I,
PI
A,
HS
TI
Pb
BI
Po
144. 1 . SO~ 345
788 670. 564 . 368. 65. 8.17 6.9~__ ~.84 . 3.81 Q:ij7 ~93 185.2 188.4 160. 1 134 7 8796 15.5
F,
Ra
160 1.66 382
Ao
410 4.25 98
r"
AI
R,
195 0..202 466
Ice
Th
Pr
Nd
328 . 785
Pm
Sm
206
... 14
E,
Gd
Tb
0,
Ho
E,
Tm
Vb
493
154. 428 317 . 233 179 400 391. 294. 302 1.86>0, ':" J tlC A. ~ 3&4: .:U~ ~29;, :2,42. 1.130 4~3"" 4 O _ 70 .2 72 .3 75.8 55.8 37 .1 1022 428 95.5 93.4
"
p,
Np
p,
Am
Cm
385 3:gg 92 ,1
5:$.
128,
8k
CI
E,
Fm
Md
~
No
~':!~":"::
"
I
~
"
Ct
B. 453.7 1562
Table 2
Melting points, in K.
So
1814
C, 1358
Z,
692.7
B
2365
N.
N,
371.0
M9 922 C,
1113
AI
51
933.5 1687
..
K,
K
336.3
TI V 1940 2202
C,
2133
M,
1520
F.
1811
mo
Rh
2236
Co
Nt
1728
G,
G.
302.9 1211
115.8
Rb
312.6
R,
2527
Pd
1827
A9
1235
Cd
I,
S,
Xo
C,
301 .6
Pb
BI
Po
AI
R,
F,
R, 973
Ao
1324
I"
C.
1072
p,
1205
Nd
1290
Pm
5m
E,
1340 1091
p, 913
Ho
1745
E,
1797
Th
2031
Am 1449
Bk '
1562
CI
E,
Fm
Md
No
lw
H ,.
0.002 r~1
U
T.able 3 h:othc nnal bulk mod uli! and oompreuibilitlc5 at room te mperatu re
;=U~ k~t~
~1+1
Na Mg
IBe
Mer K. Gschnl. dner, Jr., Solid state phYSics 16, 275426 (1964); several ... data are frnm F. Birch, in Handbook of phyrical coNlantl, Geological Society of America Memoir 97, 107173 (1966). Origin:u refe rences should be consulted when values are nceded for research purposes. Values in parentheses are estimates. Letters in parentheses refer to the crystal form. Let leTS in brackeu refer to the temperature:
raj  77 K; [bl  273 K:
to miLl
00
Held.
~~ o:~.
B 1.78
C [41 4.43
N 1.1 0 0.012
I, ~
i Ne If)
leI 
1 K: (d ) = 4 K; leI = 81 K.
I~.t~'
Bulk modulus in units 10lJ dynlc~ or 10" Nlm> Compressibility in units 1O" f cm'/dyn or 10' " m'1N
lSi
IS "'
~
fc;=I;':
CI
1.385 l ..ol.2
3.29 ~
5.62
.....
.
1 7.a.:~
At 1,1 0.013
~ 'l"S 6:58
"b
0.032 0.152 0.435 Ti 1.051 1.619 1.901 0.596 1.683 1.914 1.86 " " Sc
~ ~......
IV'''',.IC' 1M IF. ICo IN' 0." IZ" G" " ~G' lAS "0 : C, .2.",... " n.m "..,... .51 == n526 n.,".6;5~ 'O~S22
1.37
...~~ .~ ~
,:>o~ '2~5'
ISe I"d"'"
0.Ql8
~ ~
56 
0.031 0.116 _D.l6.. _D.Bl! 1702 2.725, (2.~7) 3: 208 2.70~. 32~ J!.li,;... 31L ~ ]~ 1).= O~_.~
I~ I~ I~
,1.]) _ O.2~~ :.=' . .. ASSJ 0 ".9. U:C :2;R. ..Q.9tll .2..!ilA ~;35
}.007 0.467 0.411 Cd
1~8 08 _
1 I'"
S' "" Sb
Cs '5'1, :; 
Po 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) At ~w.. 3.l?. O.~?; ,O.M[! .QJ~ .Pi209' (O~ 0.281 Q.,352 O.S7Z, I.~.Q h79 kll... 3,.17,., (3~L ... I Ra l Ac
I"' I'"
Q..
IPb I"'
=:
O.~ll Lu
I""
Fr ~
I'
<l.l1lZ
~
QY ,t2.020)1C l2!\O.J~
l4l.':.
Tm Ho Gd Tb 99 0 .384 0.397 _E, 0.306 0.327 (0.35) 0.294 0.147 0.3B1. 0 .3_ 0.411 0.392 _Vb 0.133
.~zm;
I lit I ..
Cf
Es

2~S4~ .2~,..
I I
No
i
~
"
"k IP' IU INP ~ lAm I_I~_I_ I ~ ' ~I I K:n;m: 1.6'r; tr.st: ~A...~ l I
Cm
Th D.54! (O}.., 0.987 ,0 .8) 0.54
Fm
Md
 I'~I 
~tI
11
IJ
.
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
"0""  PI
1 2 .2 + iC1J 2m
1 2 .2 + P2 + i Cri 2m
(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 gQm~ elry is shown in the fi gure. TIle internuclear coordinate is R. TIlen
(CCS) in the approximation
(2)
2e2XIX2
R3
Table 4
ExperiHlent'll collt'sive
in
.....,
7.<ltJOIl
PllramelcB in LennardJones
potential.
"'r~
RtOHl.
Eo,. 10
1e)lrool
eViatolll
.v
<.
in 101& e rg 14
u.
in
II,
.,
K. X.
N,
(liquid at zero prCSSUf(') 3. 13 1.88 0.02 3 .76 7.74 0.0fJ0 4.01 11. 2 0. 116 4.35 16.0 0. 17
24
84
117
161
50
167
225
320
H
13.595
Table 5
Joni7,ation energies
LI
5.39 81.01
9.32 27.53
. ,
The total enerlO' required to remove the first two electrons is the sum or the flrst and seco nd ionizatiOn potentials. Source: National Bureau of Standarus CilUllar 467.
~
24.58 7898
B 8.30 33.45
N,
S.14 52.43
M,
7.64
22.67
n
6.sT 20.46
Co
,
Z,
AI
SI
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,
62.63
V
6.74 21.39
M.
C,
M,
F.
7.'43":
T.
7.28 22.54
Co
NI
7:90.." 7.~ l.W 7!fT"" 9.39 .... 24.08 24,91 25.78 27.93 27.35 26.51
JOO:
I,
24.64
G,
G.
23,SI
A,
30.0
Las.;..,. 9.C
S,
21.97
.S.
9.75. 31.2
B,
33.4
~ 14.00
38.56
Rb
4.1S 31.7
S,
5,69 16,72
Z,
Nb
6.77 21.22
Mo
7.1S 23.25
Ro
7.36 24.12
Rh
Pd
Ag
7.57 29.05
Cd
25.S9
Sb
25.1
Te
X.
7. ~ 864
C,
B,
La
5.61 17.04
7. ::
22.
HI
T,
7.88
24.1
R.
0,
I,
PI
Ao
9.22 '"
29.7
2244 23.97
At
 ",
10.74_
F,
R,
5.28 15.42
A.
6.9 "!. 19,0
"'
"
C.
6.9f'"
p,
5.76
Nd
~
Pm
Sm
5~
Eo
5.67.....
Gd 6.t6
Tb
Dy
6.74 . ~
    Ho E, E,
Tm
Yb Lo =. 6.2= S.O
L,
1
1
Th
p,
Np
Po
Am
Cm
Bk
CI
4.
1 =

   Fm Md No

"
"
The total hamiltonian with the approximate form (3) for:H1 can be c.Iiago!lal~
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
(1)
TIle two frequencies of the coupled oscillators are found by inspt.'Ct:ioll of (7) to
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
'( 2<1')' :
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 dipoledipole interaction. It is the principal attractive 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 dipoledipole 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.
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 longrange 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 between the h\1) atoms is given by dU/dR. Values of E and ugiven in Table 4 (:an
The mathemaHcai calculation
~ ",'erlap energy naturaU)' depends 00 the radial dhtribulion of charge about."...h _ . is always complicateO a'Cf'l ir the ch;lrge distribution is kflO ... w
O O 0
II II
II ~
Total """I,WI
"'\"'8Y ~.311 ~ \'
.,
hi
"I
.' igure 5 effect 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 oobilal 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).
n.e
, ,
,
U(El)/w
, , , ,
.,
, ,,
FIgUre 6 . 'onn orlhe LennardJone! pote ntial (l O} .. hkh describes the I"ten etio" or lWl,l lncrt gas atom s. '[be ml,,;mum occurs at RIfT "" 1. 12. Notico: how sl CCp the curve i'I" iruide the minimum, and how flat it iii outside the m llumum , The .'IIl ue of U at the m ini mum hi  ~ ; and U  0 at
2''''.
R  u.
3 Cryslal Binding
65
be obtained from gasphase 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.
p!iR
PuR
(ll)
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 .c... P" =
12. 13188 ;
... P!I J
'<"
 , = 14.45392 .
(12)
There arc 12 nearestneighbor 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 corresponding 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:
~ = 0 =  2NE (12)(12.13) RI3  (6XI4.45) R7
dUo.
if'
,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
1.14
I.ll
Xc 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.cropoint 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 presSUfe is obtained by substituting (12) and (14) in (11):
Uro.(R) =
and, at R = flo.
"'V'[(1213)(~)"  (1445)(;),]
(15)
(16)
the same for all inert gases. This is tbe calculated cohesive cnerb'Y when the atoms arc at re.~t. Quantummechanical 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 understand the origin of the Ijuantum correction by considerllion of a simple model in which an atom is confined by fixed boundaries. If the particle has the quantum 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.cropoint (:orrectioll to the energy is inversely proportional to the mass. Thc final calculated cohcsivc energics agrcc with the cxperimcntal 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 oonstants (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 common 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
\
Cry!lal Dinding
61
)"igu,.., 1 electron density distribution in the base plane of NaCl, ..ner xny studies by C. Sehoknecht. 'I1le numbe.. on the contou .. gin' the relat;"e electron COIlCCflt ..... tion.
+
c.
5. 14 cV
GJ +
Figure 8 The energy per molecule unil of a crystal of wdium chloride is (7.95. 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.
a c.,
c..
GJ
3.61 co\'
c..
'''"'"'
affinity
G) +
c.. c.
l,<) .. al
+ 79,\
Co,hn;
rnt.,.~\
an ionic CryStal will have approximately spherical symmetry, with some distortion near the region of contact with neighboring atoms. l1lis picture is confinned by xray studies of electron distributions (Fig. 7). A qUick estimate suggcsts that wc are not misguided in looking to electrostatic 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 p0tential 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 CIions. \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
'>m
s;
P S CI
U C
0 F
a,
I
N.
AI
0.55 0.46
0.50
Eleclrostalic or Madelung EnerWJ The !ongrdnge intcrJction beN'cen ions with charge z.q is the electrostatic interaction t/lr, attmctivc between iOlls of opposite charge and repulsive 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 interactions 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} ,
J
(17)
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/"I47roT; 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 (.'onsttnts 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 gasphase data available to permit the independent 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 nearestneighbor separation in the crystal. Ir we include the repulsive interaction onl)' among nearest neighbors, we have
(ces)
A ex",RJp)
'f  R
(nearest neighbors)
(19)
(otherwise).
(ces)
al)
(20)
a  '" J
""
(~)
PIj
. Madclung constant
(21)
The sum should include the nearestIlcighbor 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
(CCS)
0'
+ = 0
Ri
Nul
(22)
(CCS)
RJ cxp(RJp):::: paq2tzA.
(23)
n.is dete rmines the equilibrium separation 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 separation Ro may be written, using (20) and (23), as
(CCS)
no
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.
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 posithc ions and the minus sign to ncgative ions. An cquivalcnt definition is
~=
R
L:'
J
()
rJ
(25)
whcre TJ is the distance of thc Jlh ion from thc reference ion and R is thc nearestneighbor distance. It must be emphasized that thc valllC given ror a will dcpend on whethcr it is defined in tcrms of the ncarestneighbor 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
..
"
,>
2[~  2~ + ~  4~
+ ... J
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
In(1 +x)=x+ 2
..c
i1
3
  + ...
4
1'4
Thus the Madelung constant for the onedimensional chain is a = 2 In 2In 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 nearestneig1lllOr llistance:
Structure
({
1.747565 1.162675 \,,;;:;n < ;; oo;;;;;,".;;;;I,lc 7.n; :...__ 1. 638 l Zi' ~~b l;; ,d c , ," :.;;..;;;;S
l1)c Madclung and
repuL~ive
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.
7t
"
"
" ,
l\tpul<i.,., rn~<gY
>
, ,
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 Octween 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 carbon. 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 closepacked structure. The tetrahedral bond allows on1), four nearest neighbors, whereas a closepacked 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
All values (except those in brackcts) at room temperature and atmospheric pressure, \\ith no correction for changes in and U from absolute zero. Values in brackets at absolute zero temperature and 1.ero pressure, from private communication by L. Brewer. Nearestneighbor separation Ro. in A Repuhive energy parameter zA, in 1O ~ erg Repuhive range parameter p, in A Lattice energy compared !(I free Ions, in kealJmol Experimental Calculated
no
UF
LiCI LiSr Lil NaF NaCI NaBr Nal
2.014 2.570 2.75 1 3.000 2.317 2.820 2.989 3.237 2.674 3. 147 3.298 3."" 2.815 3.291 3.445 3.671
0.296 0.490 0.591 0.599 0.641 1.05 1.33 1.58 1.31 2.05 2.30 2.85 1.78 3.19 3.ro 3.99
0.291 0.330 0.340 0.366 0.290 0.321 0.328 0.345 0.298 0.326 0.336 0.348 0.301 0.323 0.336 0.348
242.3[246.8J 198.9[201 .8J 189.8 177.7 214.4[217.9) 182.6[185.3) 173.6[174.3) 163.2[162.3] 189.8[194.5] 165.8{I69.5) 158.5[159.3) 149.9[151.1] 181.4 159.3 152.6 144.9
242.2 192.9 181.0 166.1 215.2 178.6 169.2 156.6 189.1 161.6 154.5 144.5 180.4 155.4 148.3 139.6
RbI
i
Ii
"
Data from various tables by M. P. Tosl, Solid nate physil;s 16. 1 (1964).
~
~
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 ). bond, as we e lectrons pcr primit ive ceO l\'ote the high concentnltion midway along the expect lOr oow1ent bonding. (Afie.' J. R. Clte hkowsl<), and M. L. Cohen.)
c...ce
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 spindcpendent 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 accommodated 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 interaction 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
\.
Crystal Binmng
7:;
g
~
:t o
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 .... uve N refers to a dassical OIkulation ....ith free atom dunge densities; A is the result for pBIallci electron spins, takin g the Pauli e~d"sion principle into account, and S (the stable stat,,) lOr antiparallcl spins. 11,edL'Tlsity of dunge 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 number of electrons in a metal arc frce to move about, usually one or two per atom. The electrons available to move about are calkd 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 electrons always makes a large contribution to the binding enerb'Y. but the chamctcristic 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 exhibited by several simnle models treated in Chapters 7 and 9.
"",..J S; SiC
ionic character
Co
ZoO ZnS ZnSe ZnTe edO CdS CdSc CdTc JoP JOM InSh
C"",,
CaSh
0.31 0.26
0.86 0.85 0.77
"sO
!..iF NaCi
MgS MgSe
RbF
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 because the kinetic cncrb'Y of the conduction electrons is lower at large interatomic distances. This leads to weak binding. MctaL~ tend to crystalli7.c in relativc!y close packed structures: hcp, fcc, bee, and some other closely related
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 conditions 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 xray 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 conHguration of the constituent atoms often can be inferred by comparison of measured and predictccl values of tlle lattice constants. To make predictions of lattice constants it is convcnient to assign (Table 9) sets of sclfconsistcllt radii to various types of bonds: one set for ionic crystals with thc constituent iOns 6coordi nated in inert gas closedshell configurations, ano ther sct for the ions in tetruhedrally<.'OOrdinated structures, and anot her set for 12coordinatt.d metals (closepacked). The prcdictt.>d sc1fronSistent 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
Values approximate only. Uni ts arc 1 ).= 10 10 m. For original r{!fcrences see W. B. Pearson, Crystal chemistry Gnd physics of metals and alloys, Wiley, 1972.
H.
Li
B.
0.35 1.06 1.13
0.68
1.56
N' 0.97
0.23 B 0.88 0.98 _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~ Radii of ions in 12coordinate<l metals _
Sc
N.
1.36 IF
1.58
1.40 0.66
0.64
1 M,
Ca
1.91
K 1.33
5 1.84 1.04
CI 1.81 0.99
I"
K,
1,88
0.99
1.98
2.38
Rb 1.48
0.81
1.&4
Ti 0.68
1.46
Zr 0.80
C,
M,
F .
Co
Ni
5,
1.13
2.15
0.93
1.80
2.55
1.60
Hf
I t4: /
Nb 0.67
1.25
C,
1. 35
1.28
"
G,
0.62 1.26 1. 41 0.81 "
G.
0.53 1.22 1.37
A,
2.22
5.
1,98 1.14
B,
1.95
0. 74 1.31 1.39
2.00
1.18 1.39
50
loll,.,.
Mo
T,
R,
1.34
Rh
Pd
A,
1.26 1.52 1. 45
C, 1.67
B,
1.35 2.24
2.73
"
T,
1.47
i40 1.36 w R.
.J..~~
1.38
1.35
Ir
1.38
5,
0.71 J,44.. 1.40 1.66 1.55
X.
2,17
0,
1.35
I Pt
1.15
1.88
1 
1.58
1 1 1 Nd
1.36 ..
1~
Gd
A, 1.37
1.44
H,
1.10 1.48 1.57
TI 0.95
1.72
Pb
I At
R,
0.84
1.75
1 1 1.70 1.76
Vb
.J.~~' '
1 7410
I R, 1.75 I 1.37
A, 1.1 1 " I
Pm
5m
( SO
1.82~ ~i . 8 1
I 1 Np
"
Tb
0,
Ho
"
Fm
Tm
1.75 Md
"
i.77Cf
1.7717:"76
p,
Am
em I Bk

1.80
1.56
"
'1.81 
I  'I ~
 
I No
"
J
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 I 2 3
o..,(A)
0.50 0.31 0.19  O. ll
o.,,{A)
0.05 0 +0.04 +0.08
.v
9
10 II
0.".(A)
5 6 7
12
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; onehalf of this is 0.77 A. In silicon, which IlllS the same crystal structure, onehalf 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 CSi 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
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
J. C. Slater. J. Chern.
"
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 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 distinguish in our notation between isothermal (constant temperature) and adiabatic (constant entropy) deformations. The small diiTerences between the isothermal 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<..>dded 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 defonned ill the same way. 11lC new axes ,,'. y'. z' ma), be writteJl ill terms of the old axes:
x' = (l + E;....)X
y' =
~If.'x
I
.'
+ ~'1Iy +
~w)Y
~~~2 ;
+ (1 +
z' = V
+ =!I)'
+ vz2 ; + (1 + Euri. ,
(26)
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.
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
(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)
"
tI,."
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 stra1ned state (b).
(28)
'n lis rna)' be written ill a more general form by introducing II, v, w such that the displacement is givCll by
(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,
x'" a x  ~ ilx'
OU
ctc.
(30)
It is usual to work with coelHcicnts eop rather than Eol1' " 'e define the strain components e...... ew> e= by the relations
IN
 e
lit'
= 
Ov
iJy
ez.o . u =az .. 
Ow
(31)
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
.
e"Y Iii
,
.. ~
X'Y ;:;:" '"
"
OIlOV
, . yz a1
E___
~
e= . a t:u n =  + .. zx +E
azdx
, .
all
ale
\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
I<
Z'
(33)
by virtue of a wellknown result for the volume of a pwaJlelcpipcd havingcdgcs x', y'. z' . From (26) we have
1+
""y'I<Z'=
IE.,.
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 t;.. Z&. The capitallctarc nine stress components: Xr X~. X&. Yro Y~. Y2 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 applied 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
z...
sn"s Out.'$ nol mean we umnot 1........1 .,ro/.lc",.!n ... h,ch Ihere Is mt.'1lnS th.tl we ron use Ih" st.ltic SIIUM ..", to de"ne the cbstoc con
ElaotM: C_ot.."u
,, ,,
'
" "~ ,
, ' ,"
11
I ~~
I , I , I "I
"
'"
I'
',' ,, ,, , ,
"
"
",
figu~ 15 Stress comporlC.'nt X. IS a bu: II{lplieC I" 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,
1
Figure 16 o.,.monstration that for a body in static a Juilibrium Y . .. X... The sum 0( the fo~. in the :t direction is 7l'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
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 dimensionless.
or
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:
..
)
+ 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~ .. ; 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
e~ = SllX~
(37)
x" =
Cllen
y 1/ = Clne""
z. =
y.
=
Zr =
XII =
+ C1Zew + C1#u + C l4e ll1 + else.... + CU;e,,'I ; + C:ue w + Cd~ + C",ell" + C~ + ewe... ; e3l en: + C3~w + C3#u + C:J,4eF + C~u: + C~''1 ; Cue,." + Cuevv + C43e,.. + C44ep + C~u; + C4#.t}I ; CSIu. + CSzw + C~ + C,54e p + C:sseu. + C~ ... ; C61 e"" + C!>'#w + Croe~ + C64eV<' + C6.u + C6f!!."JI .
(38)
The quantities Sil. S 12 arc called clastic compliance constants or clastic constants; the quantities Gil. C I2 . . arc called the elastic stiffness constants 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 cnCJ"b'Y of a stretched spring). Thus we may write
U ="2
L L CA,.eAe,. A I ,._ 1
(39)
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 p0tential energy. Consider the strcss X. applied to one face of a unit cube, the opposite face being held at rest:
"""'ie eo...,,,,,,.
)
Note that only the combination 1(C"13 + CrsJ enters the stressstrain relations.
It follows that the elastic stiffness constants are symmetrical:
C"13 :: i{C"1l + Cp..) :: CjJl> . Thus the thirtysix clastic stiffness constants are reduced to tv.'entyone,
(42)
'
(43)
do not occur. The minimum symmetry requirement for a cubic structure is the existence offour threefold rotation a... es. The axes arc in the (Ul] and equivalent directions (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;  xyzx,
(45)
according to the axis chosen. Under the first of these schemes, for example,
e;.+e:v+t{,e!v+e~+~" ,
i
, ,, ,,
Figure 17 Rotation by 2m'3 "bo'" the axis mark.,.! 3 changes r ..... y. y ...... ~ and z_r.
,,3
'b ,, ,
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 1 agrees \~;th (38). On further comparison, we sec that e,;x
(47)
(48)
C 61
:;;
C62 = C63
C64
= eM ""
(49)
Thus from (43) we find that the array of values of the clastic stiffness con
X. Y, Z. Y. Z. X,
0 0 0
0 0 0 0
0 0 0 0 0 (SO)
C..
0 0
C..
0
C..
For cubic crystals the stiffness and compliance constants are related by
C 44
= 11544
CII  C I2
= (5 11 
51
v
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
+ 2C I"~
(52)
81
Figure 18 Cu"," of volume lu, 4y fJ.:t acted on by a stress  X.(%) O il the face at %, and
X.{%
face
+ fJ.%.
"
on the parallel
TIle
IIet
force
J
(i1~~
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
". _
' 11e force equals the mll$S of the cube tin'e< the rompollcnt of the accekrntion in the x direction. ' 11e mass is pfJ.x l!.y t1zand the act",leration 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
p, ~+ 
ilu
ax~
aXil
at
ax
ay
aXiL + 
az
(55)
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
!!
(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
p
a2 u
at 2
11, ::
iJ
Gil
a:c +
~
C 44
where
v. w arc
"
Figu", 19 lhpri"WiA and B are stretched equally. the block bel"'Ccn them expcriern.'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:
p iJf = Gil
a2v
2
a2 v
2
ar! +
CH
(ilv + 3 v) + (C ar aZZ
2
12
GH )
;
(57b)
p :2 =
aw or
C II
2 W
(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
(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)
(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)
(62)
(63)
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.
(64)
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)]
= Va
(67)
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
(66)
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
(70)
whence the displacement components satisfy u = v. Thus the particle displacement 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
(71)
...
/L
K
/
IL
T
T : C..
i(CuCIll
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 propag;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 directions than in general directions.
SUMMARY
Crystals of inert gas atoms arc bound by the van der Waals interaction (induced dipoledipole interaction). and this varies with distance as ll RI!. lbe repulsive interaction between atoms arises generally from the electrostatic 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
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.
In lO, a dynclcm l (10" N/m 2)
C I),st;ol
Stif!iwu cons~b,
C"
e"
5,326
W
Th
..,,,
2.015
1.582
Coo
Temperatun', K
Ot.'rl$ity, gfem'"
1.631
5.233
~663
1.607
0.874 0.818
2.609
1.574
1. 249 1.2 14
ell
Ag A" AI
K
1. 762
1. 684
1. 3 15
0.818 0.754
0.511 0.46 1
1.240
2.016 1.923
1.143
0.973 0.937
1.697
l. 631
0.619
0."'"
0.420
19.488
2.733
1.068
0.04 16 0.0370
0.607
0.0341 0.03 14
0.316 0.282
0 300
4 295
0.02116 0.0188
0, 194 0. 149
Ph
Ni
0.555 0.495
2.6 12
0.454 0.423
0
300
11.599
2.500
1.500 1.500
1.761 1.761
1.311
1. 235
0.712 0.7 17
0 300 0 300
8.968
12. 132
Pd
2.34 1 2.271
(ces)
neighbors.
Metals are bound by the reduction in the kinetic energy of the valence electrons in the metal as compared with the free atom.
"
Table 12 Adiabatic elastic 5tiiTneSl oonstanb of several cubic crystals al room temperatun: or 300 K.
Stiffness
UlIl$tllll15,
C"
Diamond
C"
1.25 0.062 O.1l4 0.463 0.639 0."" 0.367 0.87 0.124
c.
5.76 0.042 0 ..... 0.680
0.796 0.433 0.302
1.48
N.
U
Ce
S; CoSb JnSb
"gO
NaCI
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 attraction.
Problems
I. Quantum .did. In a quantum Did the dominant repulsive energy 15 the zcropoint energy of th e atoms. Consider a crude oncdimL>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 zeropoint kinetic ellergy per particle.
2. Cohesive energy of be<: ondfa: neon. Using the LI..'rInardJQlle.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:
3. Solid molecular hydrogen. For lis one finds from measurements on ti le ga:s tllat the Lennardlone.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.
(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)
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 described ill Chapter I. Compare with the value C"'uculat(.'(} fo r KCI in tile NaCI stnlcture.
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 referred 10 separaled neulral aloms. '111e observed neaTestneighbor intem uclear distance 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 electron 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 bh forms, and n(.'gk'CI the repu l!;, '!! t.'T1t."rgY.
8. Y oung', 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 modulus 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).
TensIOn
~+'''1,
I I I I I I I I I I I I I I
L
\)ndetornwd body 7
=i r
, ,
I I I
I

=ll
w_6w w
L.
n.is
L _ _ ________ J
Fogure 21 Young's moJ"lus is defined as streuf straill ror a t<:. \li de .Iress adillg in one directi<m , wi lh Ihe sides a spccim""l k.ft rroc:, Pois..o,,'s mlio is do.'
Ilnc.:d
Illi
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 crystal 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 Rdimensional 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
'1 _ wtp p p
0 ,
I'
'I  ".}P
whe re q . jKZ(c n + 2Cw) 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 components 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~
...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
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 primithe 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, McCrawHili, 1961, Sec. 13.56.)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, .. . "hysical tJCQlIsllcr, in s.'v eral volu mes, Academic Press, Vol I, 1964. d C. z.." ICI", ";la$tidly and anellJ$ticily of fIUi/tJls. UniVl.',."ity of Chicago PrU's, 1918. 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
104
107
108
110
SUMMARY
PROBLEMS
1. Monatomic linear lattice 2. Continuulll wa,,e equation 3. Ba.~is of two unlike atoms 4. Kohn anomaly
112
112
112
112
112 113
5. Diatomic chain
6. Atomic vibrations in a metal
113
113
113
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
./II!I"
I',""1roo
 ~'idtl
1 1,,,,,,,,,
ilt
....moo
, ___~.~':!:n"li""ion "a,,,
""Iamn
.!
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
J.
, , 'f , , ,
,
y ,
, ,
i
U t
'i ,
? '''..
, ,
"
II ,
? i ,
,
Y , , ,
, , , ~, ,
.;; ,
I
'j'
.,,
(
" ~, ,
~. I ,
.. ~,
,
.'f,
... ..
J,
, , ,
 .~
,,
: , ,
Fig"re 2; (Dasheu lines) Planes of atoms when in .'q uilibrium . (Solid lincs) Planes
of atoms whf,.'" di$plaeed a.s for a IOllgit udi "al Wa'"c. The c.)(>r<linatc II measures the di$placcmcn l of the planes .
II.
,'It ..
"
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 wavevcctor that describes the wavc and in tcrms of the clastic constants. 111e mathematical solution is simplest in the [100], [110], and [I ll] propagation 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 tluadratic 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 minimum in Fig. 3.6. Cubie and higher order terms may he negl('cted for sufficiently 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 displacement of the plane $ + P is proportional to the difference u.+"  u.oftheir displacements. For brevity we considcr only nearestneighbor 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 nearestneighbor 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)
100
urll.,
(4)
where tI is the spacing between planes and K is the wavevcor. The value to usc for a will depend on the direction of K.
With (4), we have from (3):
(5)
(6) w'lM .:: C{cXP(iKa) + cxp( iKa)  2J With the identity 2 cos Ka = cxp(iKa) + cxp{ iKa), we have the dispersion
(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
(8)
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
(9)
(10)
 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
o
UJnc
__ '__,1
>oj
F"tgure 4 Plot of", versus K. n.e ..eglon or Ii: C 1/11 or A .. II wrTe'!iponds to the <;OIllinuurn 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 Chapter 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)
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 latlke 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
(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 xrays: 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 xrays 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: =
0'
dcJdK ,
(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.
103
' "r~o:::c''
(C"'f"O'~
"
Figure 6 Croup vel<Jrit)' ". VenU5 K. for modd d Fig. 4. At the zone boundary the group >e Wry 15
tt."ro.
,,~.}'!
.,
iii
(\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.
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 independent values of K:
f"'
 wid
L
p >O
Cp
f"
 wi"
=  27rC,1o .
(17)
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 direction 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 nearestneighbor planes and that the force constants are identical between all pairs of nearestneighbor planes. We refer to Fig. 9 to obtain dZu = qv. + V. _ I  211.) ; (18)
M. d/
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
ph01l011 br.nMl
Figure 7 O"tk:il and aOOOllitical branch of the rewt..., fOl" d iatomic linear l~ltkc, showmg the hmiting frequen..;el at K " 0 and K  K.  ../1f. 1he lattice comta"t il a.
dilJ~ion
.....
j (2CIMI)'"
, , , , ,
,
~ ~~<
"',,
,
6
. , ", J,
:"
'"
TO
'A
J
,.
"
o.~
TA
,.,
1>.'''..
In
\1 111 d"cdion
Figure Sa Phonon dlipt:TSion rcbtionlln the 11 11) direclion 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" Kal te rin~ 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 (oalkil w,.; Iht5t'l.fC diSl""ussed m Chapter 10.
w"
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) ;
(19)
Recall the definition of u in Fig. 9 as the distance bct\\'eE'n nearest identical planes, not nearestneighhor planes.
u.
0,_,
".
.... ,
M,
Il...
M.
L
.
.' 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'\aco.menu of atoms M , are denoted by 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 .
11._,. It,." . .,
 w2 M 2 v = Cu(clql(iKo) + I)  2Cu .
(20)
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{U(o)j
0'
C( I
+ c><p{;Ka)J
2C  M'JlW2
~O
(21)
\\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) (24)
(acoustical branch) .
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
V
M.
(26)
1" 7
figure 10 TransvcnlC optical and tran~ . ....,ne k()ustical "''aVes in a di~tomic linear lattice, ill",tratoo by the particle displacements 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 and {2CIM.JII'l. This is a characteristic feature of ela.~lic w<'!ves in poly)"'2 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 blackbody 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 cigcnvalll(.'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 energy 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
and the square of the amplitude is
HfI + i)hw
+ l)/llpVw .
(28)
ug =
4(11
(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 distinguish 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 .
"
109
(30)
= IlI(duldt}[1 
cxp(iNKa)!1 (1  cxp{iKa)!
(31)
where
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 = :!: 2rrrlNo.. 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 crystal 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 xray 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 uniform moUe momentum is rarely collsiuered explicitly. Equation (34) is an example of the nllc that the total wavevcclor of interacting 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 . If a phonon K is absorbed in the process, we have instead the relation k'= k+K + C . Relations (35) and (36) are the natural extensions of (34).
(36)
(35)
."
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.
Phonon dispersion relations <t.{K) arc most often determined by the inelastic 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
~('W
~2k2
~2k'2
2M..
2M"
(38)
OT
absorbed ( ) in the
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
111
U'"'
",
""
, "
' 0
e' "
IIIOj
,
0
,0
0 0
00
II LlI
.., .
III
,
.00 "
110 0
\\\\,':"'<.10< in "ni" 2n/Q Figure II n.e dis)X'rsion curves of sodium forphonons pro.,agating in rl,c (OO I J. [IIOJ. and (I J IJ dioiioni at 90 I( "Ii detcrmined by inelastic scattering of neutrons , by Wood" Brod.:oousc. March and 1J,o..('JS.
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 alkalihalide 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 isoclccIronic 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
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 per atom is
timeave~e
total cnergy
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
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.
113
4. KoIm anomaly. \ Ve suppose d,al the interplanat fOrce constant C" between planes $ Qnll .f + Jl is of the fOrm
rkoo ,  A~inP' .
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 nearestneighbor atoms are altemately C and IOC. Let the masses be e<lual, 811tl let the nearestneighbor 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 shortrange 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 conslants is C,.c  2( lf rllrlct, where a is the C<lllilibrium nearestneighbor d istance, (h) From (IGa) show that the Jispenion relation may be written as
0'
L ,, ( I )" (I 
cos pKa)p 3 ,
where ~  4yM anll 0'  Clyol, (c) Show that J is negative (unstable mode) at the zone boumlary Ka ,. '11 if 0' > 0.475 or 4n3). 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 lt. u

114
Refertmces
II . llilz and W. Kn..u, Phc"'nn disJlf!nion rdation! in i"$llla/I)r" Sprin~r. 1979. An atlas. C. A. Ilorton and A. A. Mal"'<ludin, als., Dllnam.iooi,lro/lertlt.J til w/i!h, NorthHolland. A contin"jug n'ries 1974 {l900).
W. P.
~Iason. ~"'.,
Ph ysic4l
/lCOII&/iGf,
W. Cochran, Dynamic:! of lIta"', In r.rys/ou, Crane, nuSllak, N. Y., 1973. A good It,);! . J. A. Ik~ [and, Phlj'ic$ vf plotwns, 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 Planck distribution 1'\0.,11111 mode ellumeration Density of stales in one dimension Density of slates in three dimensions Ocbye model for density of slates Ocbre T3 law Einstein model of the dcnsily of stales Ceneral result (01" D(w) ANHARMONIC CRYSTAL INTERACTIONS Thermal expansion THERMAL CONDUCflVITY Thermal resistivity of phonon gas Umklapp processes ImllCrfcdions PROBLEMS
I. 2. 3. 4. 5.
117 117 118 119 120 122 124 124 127 129 130 131 133 135 131 139 139 139 139 139 139
Singuhllity in density of slales RnlS thermal dilation of crystal cell Zero point lattice dislJlaccmcnt and strain Heat c(tIJacity of laycr lattice Grunc;scn constant
nEFERENCES
140
< .>
,/
,,/'
,
,
/
/
'/
1//
,//'/ , / / , //, ,
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
We discuss the heat capacity of a phonon gas and then the e ffects of anharmonic laltke interactions on the phonons and on the crystal. The rmal prope rties of metals are treated in Chapter 6, superconductors in C hapter 12, magnetic mate rials in ChapteTS 14 and 15, and noncrystalline solids in C hapter 17.
By heat capacity we shall usually mean the heat capacity at constant volume, which is more fundamental than the heat capacity at constant pressure, which is what the experiments iletermine. I TIle heat capacity at constant \'01ume 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 polari7.3U p . TIle form of (11K.,) is given by the Planck distribution funcon tion: (0) exp(licd"T)
1
(2)
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 excitation 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 oscillators in the nth quantum state is
N"
. ~
cxp(nliwlT)
L
~mO
N,
L
~_ o
(4)
e<p(,n",h)
(n)
s cxp(s/iwl" ','
L cxp(  sliwh) ,
The summations in (5) are
(5)
L" ~ 1 x
wit h
%
,
, 1
(6)
(7)
wK.p
in the rmal
pose that the crystal has Dp{w)dw modes of a given polari ...alion p in the frequency range w to w + dw. Then the energy is
U = ;
(9)
The lattice heat capacity is found by differentiation with respect to temperature. Let x = liwh = IiwlkBT: then iJUliJT gives
Cia, = kB L. "
(10)
The central problem is to find D(w), the number of modes pe r unit frequency range. This function is called the density of ( 'Ies or, more often.
I.J
F~J
,
n.e
. .. 0
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.
100
,.
100
'W,. 0
or
FIg"re 3 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. 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 alloo.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 olledimensional 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~
(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 fixedend boundary oonditions to the "alues
K:::L '
~ 2~
3w L ... ,
(N  l)w L
(12)
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 $11"; this permits no motion of any atom, because sin 511" ,anishes at each atom. TII US there are N  I allowed inrl ....endent values of K in (12). ntis Ilumber is equal to the
.
number of particles allowed to m()\'c. Each allowed value of K is associated with a standing wave. For the onedimensional line there is one mode for each intervai llK "" m L. 50 that the number of modes per unit range of K is U'lf for K:$ ,"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 polarizations 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 periodic 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 solution u. = u(O) exp[i(sKa  wxt)] the allowed values of K are
K = O,
.. . ,
N~
(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 twodimensional 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 =   .
TT dw
1r cbldK
(15)
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.
+ K~y + KLZ))
exp{i{K.{x
(16) (1 7)
2~ + .
N~
(18)
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 , .... be of I~ ro.m sin .KG 01" 00II IlKG; Ihne are m 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 complcK 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).
.. .,
.,
.
1:
N.
;..
,
..
O'
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
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
::::I
(VK2/2zr)(dKJdw)
(20)
w = oK .
wit h
0
(21)
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 =
6rr'lrlNlV
(23)
(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::::
dw D{wXn(w)liw 
r'"
dw
(2";:3) (ij:~w_ 1)
(26)
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
whcre x
"wi.,
>II
81T .
(27)
This defincs tile Dcbye temperature 0 in terms of WI) dcfined by (23). We m ay express 0 as
o ~ ~. (&n'N)'~
kll V
(28)
123
, ,
rogure 1 lIeat capa<.~ty C,. of a SQlid , according to !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 gioll of the T a la.. is below 0, I e. TIle asymptotic , ..1..., at high val"es of TIe is 24.943 J mol I deg I.
II
/
I I
0.8 1,0
') ,
, , , ,
, " ,. ., .,
"
'A
"
l
C,.m.'1llium
t/
/
/
s~icon anti genna ilium. Note the m'<"l"ea5e at low temperat",..,.. Tocolwert a "at..., in caVmol K to J/nlQlK. m"l tiply by 4. 186.
:.1/ ,
0
/ / //
'00
(29)
SilK.>on
I
I
!
I I
T(~)' LXV dx ~
8 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 expressiOn of (26) with respect to temperature. Then
C
"=
27ilti'kBT2
3Vh
I"b dw (e"w"'....
0
e"wlT
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 approaches the classical \laJue of 3Nk B Measured values for silicon and germaig. 8. nium are plotted i
..,
Debye T3 Law
At very low temperatures we may approximate (29) by letting the upper limit go to infinity. We have
IT 
." 1
15
(3 1)
where the sum over 6 4 is found in standard tables. 1'nu5 U >;II 3rr"NksT4l5rfJ for T <t 9, and
Cv
!!
12;4
(32)
which is the Debye r 3 approximation. Experimental res ults for argon are plotted 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 classical, 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" wavevector defined such that fl vKT = k8 Tand Kv is the Debye cutoff wavevector. Thus the fraction occupied is (TIO of the total volume in K space. There are of the 'jl 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.
(33)
125
17.78
..
13.33
i ,
" .
/
o
"
....
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.)
1/ ...
rI"
V
,."
..,
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 (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'.
or
U
344'
e.
1<40
 ~
e
AI
0.85
2.00
0.27
o .
C
2230
1.29
N
~
N.
75
$ CI A,
N.
158
Mg
1,41
1.56
C.
400 ~
~
$0 TI Y
<l!O ,
0.17
$1
P
~
428
2.37
 645'
1.48
.... $.
90
92
K
91
C,
6300.94
M,
416'" 0,08
F.
Co
NI
4500.91
Co
343 4.01
Z,
G,
320 0,41
G.
374 0.60
A,
282
e,
K,
72
230 "$,
1.02
0.31
327 1.16
0.50
0 02
Rb
0, 58
Z,
Nb
2'15 _ 0.54
Mo ]50
1.38
I
T o
~
Ro
Rh
480 1.50
Pd
27. 0.72 PI
240
Ag
225
Cd
2090.97
I,
lOS
0.82
$,
200 0.67
$b
211
T.
153
56 ;: T4:.
C,
"291'
0.23
"606"
1.17
64
,
X.
0,51
4.29
0,24
0.02
e,
c. p Hf
T,
R.
0,
500
I,
420
Ao
165
Hg
71)
TI
Pb
el
Po
~
At
R,
00 ]~O ~' 110 ,.., 14;: l,l;; ~~ ,[ 0,14 0.23 0.36 0.58 1.74 0.48
0,88
1.47
0.72
3.17
11'
0.08
lo
210
120
F,
R,
Ao
.....
C.
0.11
p,
0.12
Nd
0.16
Pm
 Th
i~
0.54
$m Po
Eo
Gd
200
Tb
0.11
0,
210 0.11
Ho
0.16
E,
~
0.13
0.11
0,14
p,
N,
_ 20'1'00.28 0.06
..0.07
  ~
Am
Cm
ek
Tm
~
Yb
0.35
0.17
0.1 6
Cf
~
E,
' ...... 
Fm
. . 
Md
No
L,
~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 l" <l.n: Ir"Il' II. W ,,!t'
Powell a,,<l Y. S. ToulouK'J.n, ~clence 181, 999 (1973),
,
,
/
 
'
,
, , ,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 calcubled on th e earlle.t quantum (Einstein) model, ".;ng the ~ristic I..'mpcrature 8" .. lib/A:... 1320 K. To convert to Jlmolcleg. 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) (aT v
= Nkll :;
(34)
as plotted in Fig. 11 . This expresses the Einstein (1907) result for the contribution 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(lih), 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 approximate the optical phonon part of the p honon spectrum.
~ (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 between 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
Here dKJ.. is the perpendiCUl.1T distance (Fig. 13) between the surface w constant and the surfuce w + dw constant. The value of dK.,t will vary from one
point to another on the surface,
I..."
d 3K =
JdS...dKJ..
(36)
IV"wI dKJ.
= dw
is the difference in frequ ency between the two surfaces connected by dKJ.. Thus the element of the volume is dS... dK J.. ::: dS...
where VIC we have
""
IV ..w!
dw
dw dS... vol:
D(w) dw
~ (~)'
211'
f 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) ~
(21T)l
dS. VII
(37)
129
Figure 13 The Iluant;ty dK J is the per]lendirular disilltice be1......'Cn Iwo (:0115l:01nl frCllu<o:ncy nufaces in K space, une at frequency '" and the other al freo:luency '" + d ... ,
D{w
D{w)
(.)
(h)
rLgure 14 Denlily of stales II~ II functiw. of fieo: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 deviations 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. Threephonon processes are caused by thirdorder 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 presence of one phonon ClUses a periodic clastic strai n which (through the anharmonic 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 threedimensional grating.
Thermal E:tpansion
We may understand thermal expansion by considering for a classical oscillator the effect of anharmonic terms in the potential energy On the mean separnlion 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
x3
Ix .
(38)
with c. g. and I all pOSitive. lne term in 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 minimum, but for small oscillations the fonn is an ade<lU3te representation of an interatomic potential. We calculate the average displacement by using the Boltzmann distribution function. which weights the possible values of % according to their thermod),namic probability;
(,)
.:.:;;;.
r
L :
dx CKp(  PU(:r))
,
'"
~
'"
1.62
Triple pom!
5.42
, , " ,
~
'0< 
""
Figu", IS Lattice oo,..la,,1 of ~olid argon as a fUllc_ tion of temperature.
/'
//
.,
/
K
./
.I
./
'"' ,
1.70
"
"
T~mper.'lure.
"'
'" ,"
with f3 '" lIk nT. For displacements suclJ that the anharmonic terms in the energyare 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 lowest 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 steadystate flow of heat down a long rod with a tempel".lture gradient ,",<lx,
iu=K d,
dT
(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 proceed directly in a straight path to the other end, but diffuses through the specimen, 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, 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 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 local 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
Ju
=  n(v~ J; = i'I(v2) CT J;
v..'C
dT
<If
(43)
iu =
with
icve dx '
(44)
III!
VT
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
 ,.
Quartz"
0  190
Nael
 190
'Parallel to optic axis.
0.50 0.07
540 23
'"
0. 27
'00
llOum"tal1ic
"..
,..
..
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 to right wllhoul particles and their "neq:y remain unchanged . Thull energy is tr.ulsported from being drivCfl by II temperature gradient. n~rdore the thnmal Tesistivity is zero and t1~ thcnnal conduct;,;!), is infinite.
len
rogun 16b The usual definition of thermal rondUCI;.ity lO .gas rcfen to 1\ situ.lllioo """'"" 00 ma$lI 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\eavcrngc 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
now
~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 scatteroo phonon is equal to the rrcqucm:y WI of the incident phonon . It is rather remarkable also that a threephonon 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 threephonon collisions of the form (45). For sueh (."Ollisions the phonon momentum
(46)
is conserved, because on collision the change in J is nl( is the number of phonons having W3vcvcctor K.
~
 K:z  KJ = O. Here
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 pen;S! down the length nfthc crystal. On arrival c:f phonons .. ,n 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 establishing 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 resistance. 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 threephonon 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 momentum 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
'36
.,
F1gurt: 17 (al Normal 1\, + K.,  ~ and (hl umkJapp 11:, + 1\2  ~ + G phonon coIli~ion pro(It'SK'S In a twodimensional 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 rcomponenl 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;,
tnc
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" < by an umklapp process (C " 0) create a phonon with :alL 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 conserve 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
:;
131
" .3
100
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. McNdly.
"
" Xl
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 imperfections, 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 tempcraturcdependent 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.
l
136
00
~
K  O.06T' /
~n nd\t<..J
c..
f
'" , ,
10
1
/J
If.
r/
Normal
1 \ "c;., ""'"
j\
1/
I'\;
if
, o,
1
rJ
, ,
In 20
I'"
50
100
200
500
r,,,nlK'r.l!lI ,e, K
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. Synthetic 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 sapphire 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 elcctronic (.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 gcnnanium 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.
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 .
"1f
(wt~
. .2\lfl
W /
wl,ere w", is the ma;o: imum frequency. (b) Suppose that an optical phonon hrancll has the knnl w(K)  ~  AK~ , ncar K  0 in thrcc dimcns ions. Show tl I.1t 0 (,,",) = (U2"1ft'(2mA: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 Vfar 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~ = 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 onedimcnsionallattiee, bu t tl,at the mean square strain is fl nite. Consider (aruaxf) = IIKtllS as the mean square 5train, and show that it is Lqual 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.
3Ii~n'pv3, where
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 restricted 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 extrL'rncly low temperatures?
5. Gnmdsen constan'. (a) ShowtlUlt the free cnergyof a phonon modeclfrequcncy w
is ksT In 1 sinh (1Iw12kIlT )j. It is nl'C'Cssary to retai n the zeropoint energy illw to 2 obtain this result. (b) If C. is the fractional volume c1Jange, then ti le fice 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, 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.
M
THERMAL CONDUCflVITY
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, $emlco,ul"cl~, W iley, 1958.
J. E.
6
Free Electron Fermi Gas
ENERGY LEVELS IN ONE DIMENSION EFFECT OF TEMPERATURE ON TIlE FERMIDI.RAC
DISTRIBUTION FREE ELEcrRON GAS IN THREE DIMENSIONS 146 146
144
HEAT CAPACITY OF TIm ELECTRON CAS Experimental heat capacity of metals Heavy fermions ELECfRICAL CONDUCllVrry AND OHM'S LAW Experimelltal electriClll resistivity of metals Umklapp scattering MOTION IN MAGNETIC FIELDS Hall effect
TIIERMAL CONDUcnVlIT OF METAI. S
166
NANosrnucruREs
PROBLEMS
L. Khlelic energy of e ltttron gas 2. Pressure and bulk modull15 of an electron gas 3. Chemical potential in two dimensions
168
IG9
4. 5. 6. 7. 8.
Fermi gases in aslroph}"sics Liquid lIe' Frequency dependence of the clectrieal conductivity Dynamic magnetoconductivity t('llsor for free electrons Coheshe energy of (ree eleclron Fermi gas
9. Static magnelooondudivity tensor 10. Maximum surface rfiistance 11. SmaIl metal spheres 12:. Density of statesnanometric wire 13. QuantiUltion of ronductance
REFERENCES
1' ;&11 .11' I Schematic model of a crystaJ. of sodium metal. 1lIc atomic cores are Na ioll!; they are Immersed in a rea of c:onductlon clcd:rons. TIle (~o",:ludion electrons nrC deri\'ro from the 3s ,'tdcnce " le<..1rons of the free al oms. Tnc alomic t'Or<;$ 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:
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 potelltial of the ion cores. The utility of the free electron model is greatest ror properties 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 metalslithium, 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 electrons 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 onehalf of the ncarestneighbor 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 conducthity. '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 electrons 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 consequence ofthe Pauli exclusion principle. Bya free electron Fermi gas, we mean a gas of free electrons subject to the Pauli principle.
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 solution of the SchrOdinger equation 11/1 = 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 approximate 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 infinite potential energy barriers. They are satisfied if the wavefunction is sinelike with an integral number" of halfwa\'elengths between 0 and L:
"'~ = A sin(~7T.r)
"
d~. _    A   "" (nw) .   r cb: L L '
whence the e nergy
E"
inA" = L ,
(2)
(nr.)
dr
(~l")
L
is given by
E ..
!!:....( nw)' L 2m
(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
Figuri:! Fint three cn<Tgy !c...,.. and wa~c !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 On of the le,u of'i\lanlllm number n i$ <'qu91 to (h"I2mXnI2L)".
;J:i
~
"
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 quantum 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;
,
I
m,
Electron oceupancy
,
3 3
4 4
m,
Electron occlIp:ln<')'
I
j j
I
j
I
2 2
I
j
0 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 FERMJDInAC 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 FermiDirnc 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 7ro (Fig. 3). 111c FenniDirac 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 because in the limit TO 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 Boltlmann or Maxwell distribution.
h' ( if'
if'
if' )
(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
en,
,.,.
, ,.,
yo: /0' /(
JQ<
r:..~ .. ,
i'.
4. ,, ~' 1'
, 'I'['~< ,, ,,
,
':f ,~/(
,
~ t..
" "'\'\
~ t""" f'. 1' , , .~.II("
 50,000 K.
111C results apilly to a J.!." illtllre<: dimensions. '111e total number of p'rtic\~') is constant. inllcpmdent 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,
l/A.,x
z;
+ L.
z) = (x, !J, z)
(8)
and similarly for the y and Z ('()Ordinates. Wavefunctions s.llisfying the freeparticle Schrodinger equation and the periodicity condition are of the fonn of a traveling plane wave:
!J!l,(r) = exp (ik . r) ,
(9)
= 0
+
2~
 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
'48
I
On subslituting (9) in (6) we have the energy vector k:
Ek
(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):
",..=   /&. .
2m
I,'
(14)
From (10) we see that there is one alJov.w wavevectorthat is, one distinct 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
(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
(16)
"
=~ (3~N)"'.
2m
V
(17)
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 clectrtNl$ the O<'C\lpied orbitals of the system flU a sphere of ,adi'l$ Ie,., when: 1",. .. ~lIl.iI2m is the energy of an ekctwll hiving I w..."'".,ctor Ie"
V(o '
~,
\\
FigureS Density of singlepartide 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 7ro. The a"c.rage (",)C,'g)' is increa5Cd when the temperature is illa..a..ctl from 0 to T, fur dectrons are thenna]ly excited from region 1 to rl"gion 2.
\ \
L__...L~'_'''"''~ ___ _
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:
N 2
_ V (2m<)'"
3.r
1i.
(19)
I!!  
(20)
'Strictly, Djf) i. th e dellsity of onepartie1c 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)
~
";deney
Met;d
Electron concentration. in em  3
4.70 X lcP
Radius"
parame ter
Fennl
wa\'evec!or .
~'ennl
Fenni
energy, In eV
Fermi temperature
'.
3.25 3.93 4.86 5.20 5.63
2.67 3.02 3.01
In
em I
velocity, In em ,  I
T,. (,1k".
in dcg K
Li Na K
1.11xldl
1.29 X loB
4.72
Rh
C, Cu
Ag
2.65 1.'" US
0.91
0.92
0.75 0.70
1.01
0.86 0.81 0.75 1.57 1.39 1.39
3.23
2. 12
5.48 x 10 4 3.75
2.46
2.15
0.64
1.36 1.20 1.20 1. 93
1.37
8.45
5.85
Au
2
5.00 24.2
1.63
8. 12
6.36
6.39 16.41 8. 27 5.43
B,
Mg
8.60
4.60 3.56 3.20
13.10
1.88 2.85
3.27
C, S, B, Z. Cd
3
1.11
1.02 0.98 1.51
2.23 1. 58 1.28
U8 1.13 1.82 1. 62 2.02 1.91 1.74 1.82
4.68
3.95
3.56 3.69
2.3 1 2.59 2.07 2.19 2.41
3.65
9.39
4.58 4.24
10.90
9.28
18.06 lS.3O 11 .49 13.20 14.48
~,
1.'"
1.75
7.46
11.63 10.35
8.66
13.49 12.01
Al
G,
In
4
1.65 1.50
1.57
8.60
9.37
9.98
10.87 11.64
Ph
Sn{w)
2.30
'.23
rd~H'
1.62
1.88
10.03
when: OH Is the Rrst Bohr ndlu. and~" II the ndlu. or a sphere that oontalns one deet""n.
151
l1lis result may I~ obtained and expressed more simply by writing (19) as ILl N::. whence
"'2 In E + const;mt
N="2';
dN
dE
dN 3N D(E) a  =  . dE 2E
(21)
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.
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 contribution 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 mobile. while not contributing to the heat capacity? The question was answered only upon the discovcryorthe Pauli exclusion principle and the Fermi distribution runction . Fenni found the l.'Orrect equation. and he wrote, "One ret.'Ognizes that the specific heat vanishes at absolute zero and that at low temperatures 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
f'
i"
dE ED() .
(24)
Here feE) is the FermiDirac function and D(E) is the number of orb{tals pel"
unit energy range. We multiply the identity N= I
(25)
by
E"
to obtain
" L
(26)
We USe (26) to rewrite (24) tiS
l::.U =
J'"
"
dE(E  E,,)f(E)D(E}
L
e ,.
(27)
TIle first intt:gral on the righihand 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 temperaturedependent term in (27) is feE), whence we can group terms to obtain
C~l ::
(1 = 1U ,T
L
m
tr
(28)
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
L
.50
i ', s
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 plot l<.d as k~TIE,.
/ ,, ./'
,.,
v
0.4
/
0.6
"
..,
,
I~ :. ' I ' ~.
r
,
\ !.''
 r..
Jogiofi of <ta..L..
1[01> _
~
~
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 FermiDirac distribution function and replace Jl. by
..
..
I ",
"
.~,'c. c,",~ l,
, 
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 cunes were cakubtcd from oeries "~pAnSions of the integral for the number of particles in the 1)"Slem.
T B
k8T,
(30)
We set
(3 1)
and it follows from (29) and (30) that
f"
~,h
til: r
(~+ 1)'
e'
(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 tempemtures such that EFh  100 or more. 'nlC integnlfZ beromes
f"
e'
3 '
the m(lre &mili:1r result
"..
(33)
.'r,
~\lm
_______. 4....~
....
.. ... .
err  2.ll'I + 2..57 P ~
_ _.  
.~
'~,'.~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.)
(34)
(35)
51
"p.
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 characteristic of the material. The electronic term is linear in T and is dominant at sufficiently low temperatures. It is convenient to exhibit the experimental values 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 the relation y(obscrvcd) ")'(free)
JIl,
(
where m,1> is defined by
(38)
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 inlerdction of the conduction electrons with themselves. A moving elechun causes an inertia] reaction in the surrounding electron gas, thereby increasing 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 compounds may have inertial masses as high as 1000 m, because of the weak o\'erlap 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, S2O 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)
(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
LI
163 0.749
B.
0.17
Table 2 (From
E:lperimenlal and free electron values of electronic: heat capacity constant 'Y of metals kindly fll rni'hcd by N. Philli!,S <lnd N. Pearlma n. The
j,
O.SO<>
0.34
compil a t ion~
2.18
 _.......
~
N
~ ~
N.
M ,
L.
10.
Observed'Y in mJ mol' K '. Ca~tited rree~ecrian: i~':m:rmOP"k'm"lm  (observed y)/(Iree electror'l yl.
C,
.
A,
0.646
C.
2.'
So
10.7
1.668
1,25
flIT
I'
...:."
Rb
2.4 1
S,
3.6
Y 10.2
~
:..
C,
3.20 \.43
ici38 ~
    .   .  V
Mn(y)
~..
TI
C,
F.
Co
NI
Z,
 I.'"  AI SI
~
1.35
(f91~ 
~
G.
G.
3.35
9.26
1.40
9.20
......
4.98
4.73
~.,.
.......Pd
9.42
~
7.02
 0.1 9
(>ol
A,
Z,
Nb
Mo
To
R,
2SO
7,79
20
Rh
Cd'
0.688
I,
1.69
I~33
Sn
Sb
3.3
4,
1.78
1.4.\..Q.,
1.26
1.37
B.
2.7 1.4
HI
~ ."'"T'>
T.
W
1.3
R.
2.3
0,
I,
pt
A,
0 .729
BI
"
0.11
,~,...
1"Jt ~
2.16
 5.'
.....,.,
2.4
3.1
6.8
Pb
'f$69~ 1.97
2.98
":""""1"
0.008
f
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 intcnalt 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 (41) =  eEtlh .
oa.
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 :::: eErim. 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
(43)
The electrical resistivity p is defined as the reciprocal of the conductivity, so that
p = mlncT .
(44)
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 becduse 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 lutics 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; 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 electronelectron collisions. see Eq. (10.63).
:+,
7
(46)
'TL
7,
where TL and 7, are the collision times for scattering by phonons and by imperfections. respectively. The net resistivity is given by
P=PL+fJI
(47)
Table :l
rcsi~ tivity
of metals at 295 K
LI
1.0"9.32
B.
3158 3.25
(Resistivity value~ as given by C. T. Meaden, Electrical resistance of metals, Plenum, 1965; res idual resistivities have been subtracted. )
N,
2. 11.
4.75
Mg
2,3r
 Co
5.8
AI
2'
5.92
Cc!ndtICt&i!Y I n+ unitsof
S<
l O~ (o1iii!!!)1::''''
"
Ii
430 C,
36
3.I~:'
2.74
G.
Sn
(wJ
SI
   ..
~
N .
CI
::::T
A,
 ... b
B, I
K
y",
7.19
TI
43.1
C,
~
M,
139.
F.
9.8
NI 7.0 Pd
10.5
C,
1.70
G,
14.85
.rtJ2 T:~ T~
C ~~ TW' 43::
A g
1.61
0.6t::: ~
19.9
12.9
Rb
080
12 ,5
S,
0:;;["
21.5
2,
0.24
42.4
Nb
14.5
Mo
5.3
To
':"' 0.7 14.
R, 7.4
0,
l.l{)
91
Rh
Cd 727
I,
m~
~
OlL :58.5
66~~ T~89
f.3
20lC
4B I,
1.% 5.1
1l ~ ~~J:: T,j'g";
8.75
   "' ~ 
S.
K,
Sb
T.
x.
~ ~
C,
O SO 200
B,
0.2639
L, 013
79.
Hf
0.33 30 6
T,
13.1
R.
O.Se
!S6
PI
o.96 ~
A,
4.5~
Hg 'IQ.
TI
0.61..
16.4
Pb
o:~
BI
Po
AI
O~7r:;: 1.89
5.3
b E
95.9
o.OSEL
116.
10.4
2.20
21.0
O .n:: 46
Yb
26.4
F,
R,
Ao
I'
C.
6:12 ~
 L,
53.
R,
p,
> 6.(5"
67.
Nd
0.1'59.
Pm
~
Sm
0.10
81
99.
Tm
"O":lr"
62.
0.12 ...
1)'~ 'If[9''"
134.
Ill .
90.0
77.7
81
Th
O.~
15.2
p,
U
6]~
Np
0.085 11 8.
p,
0.070. 143.
25.7
        "  
Am
Cm
Bk
Cf
" 
Fm
Md
 N o L'
161
Figuno II Eic'dricai re.isti,ity in most metal. ariJe$ from coll isions of ele<.1rons with ir...,gularitics and in (b) by bnpuritles and \xant 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 difICrent 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 resistivity 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 thity 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.ohmcm (1 x 10 6 ohmem) 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 .uohmcm, 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 temperaturedependent part of the electrical resistivity is proportional to the rate at which an electron collides with thermal phonolls and thermal 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 e1ectronphonon 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 electronphonon 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 e1ectronphonon 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 collisions 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 geometry 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 .
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 temperdtures (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'dttering, 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 proloeSseS contribute to the resistivity in all metals, but they have not yet been clearly isolated for any me tal beeduse of the large competing effects of imperfection scattering, electronelectron 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)
F=
(49)
(SI)
E + .;. 8)
(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 + :
T
1/ 1/
vv)
(51)
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
Vz
=   E
m
n.
(52)
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 conductor, in the direction j )( n, wben a current j flows across a magnetic field B. Consider a rodshaped specimen in a longitudinal electric field E. anda trunsverse 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 =    E~ c" me
en,
(53)
(51)
j~ =
is called the Hall coefficient. To evaluate it on our simple model we use rW'TE.Jm and obtain
(ee5)
RII =
_ ce,B"'cE~'J,;":::'c'_ = __ 1_
ne'lrE.Blm
(55)
ne<
(51 )
I nJl = ne 
perpo!ndiwlac
...
e
I
+++
rri' I
+
~
lL:.
~J'
ill>!
SUrting liP.
:.
~. ~J'
, ,
++
Figure 14 '111C ~Ialldard geometry for the n ail effect: a rodshaped Ipccimcn of n!dangulal' CI"QIO'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,clycharged 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 concentration. The symbol RII denotes the Hall coefficient (54). but it is sometimes used \\i th a d ifferent meaning. that of Hall rcsi5tance in twodimensional 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 expression becomes somewhat morc complicated if both electrons and holes contribute 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 lIeisenberg 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.
In Chapter 5 \"C found an expression K "" led for tile thermal conductivity 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 temperatures. 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 WiedemannFranz law states that for metals at not too low temperatures the ratio of the thermal conductivity to the electrical conductivity is directly proportional to the temperature, with the value ofthcoonstant ofpl"oportionality independent of the particular metal. This result wa~ important in the
16
Table 4,
~mcients
{The experimental v"lues of RII as obtained by OOIlventional methods are Summarized trom data at room temperature presented in the LandoltBomstein 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 voItcmlampgauss, multiply by 9 X 1011; to con\'('rt fill in CCS to m'/cou:omb, multiply by 9 X
JOI3. J
Metal
Method
uperirr.."'la\
A~umt:d
C..lt:ulaloo
 J/n.
N.
K
2.603  4.9
 6.04
 0.82  1.19
 4.7
Rh
Cu
Ag
 5.6
 0.6
 1.0  O.S +2.7
 0.92
Au
Be
Mg
 US
conv.
~" .
Al
10
+1.136 +1.774
1 holc 1 holc
+1.135 +1 .780
'" Sh
m
+50.
 22.
~" .
 6000.
J..
Table 5
L X 10" WlIuohmldci" Metal
JOO'C
Ag
A, Cd C,
Mo
2.23 2.61
""" pb
PI
<1'C
JOO'C
S"
W
Z"
2.56 2.60
2.49
3.20 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. =
u
(kB)'r .
e
(fiT)
(58)
dcg)2
0:2.45
11\1 remarkable result involves neither tl nor m. It docs not involve 'T if the 5 relaxation times arc identical for c lcctricv.l and thermal processes. Experimental 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 minimum 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 ypically contain between 10 and 1.000 atoms. Scmiconductor technology (see section "supcrlattices in Chapter 8) has made it possible to fabricate small pools of electrons called in various ways: singleelectron transistors. quantum dots, artificial 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:
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 cltd rOrls. for the energy release 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'lllrntion 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 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 dispersion relaUon for th is wave in the medium is
w: 
(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 FernU 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.
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 selfenergy 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
+L.)(~:)
(&I)
= nedB ..,
0'., .
(65)
lIw~7.
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 mtal. 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)
wa\'C
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 eigenvalues ~ 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 lIatesnlJnometric wire. (a) Consider II nanometric wire in the form of
a rectangular parallelepiped, with two sides L. .. Lv " I nm and the 1 0.lg axis L . I em. TIle single particle eigcnstatCli of tIle system may be written as
(n~TrYll..,)
'\.,
exp[i2nN.tl
AN'  'hmui ,
where v is the electron velocity alollg th e .tonxis. Here A  (2ntitl2m and NV((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
..
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 conductance is (~/7rll~ n<>=
References
D. N. l .:tngcnberg. "Resource letter OEPMI on the onlinary elcclronic propertlCl< ofmclals,"
Amoc... J. Ph)'I. 36. TI7 {I968}. An cu'CJlcnt carl)' bibliogral,!')' on
~J.:in
tn.n~port
dfCdl , anomalous
cited, Al.belKaner cyciotnlfl rcso... nce. ma.gnetoplasmll w,wes, liize elfcW, conduction
eicc.1ron spin rcsonar..:e , opli<:al sp:.'Clra H pl'otocm ission, (Iua"tum oscillationl, rnlIgllctic lld 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\edimen.rlonaloonduClcrr, Springer, 19611. J. Ziman, f/eclN"m . and pllUl"Mlrnl, Oxfonl , 1900.
7
Energy Bands
NEARLY FREE ELECfRON MODEL Origin Qf the energy gap Magnitude of the energy gap 116 177 179
179 180
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 KronigPenney model in re<:iprocal space Empty lattice approx imatioll Appmxi rnale soIutkm near a zone boundary NUMBER OF ORBITALS IN A BAND Metals lind insulators SUMMARY 186 187 188 189 111 193 195
PROBLEMS
195
195
195 195 195 196 196
REFERENCES
196

E.,,rgy
D
Cd
D
B
""" Schematic .,Ie<'tron
QO
0
D D
Semiro,,""UQr
D
Stmlrond...._
DI ....bta
. .. D
"
.
~,,~ I O<Ulpaocy of allo ....ed e.,."rgy bands fOr an Insulator. mt.'lal, scmioncl:!l, and semiconductor. 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
n,,!

G
(
U D
'"
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 priodic mOlllllatiOlI.
F. Bloch
Thc rree clectron model or m<.>tals givcs us good insight into thc heat capacity, thcnnal conductivity, electrical condllcthity, 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 ohmcm at a te mperature or I K. apart rrom thc possibility or supe.'Onductivity. The rcsistivity or a good insulator may be as high as 1 ohmcm. This range or l(il2 may <f2 he the widest or any commOn physical property or solids. Evcry solid contains electrons. 111e important question ror electrical conductivity 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 percent 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 crystals. 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 negatile. Electrons in crystals rcspon. ,) applied fields as ir endowcd with ncgath'c or positive
17'
charges, e or +e, and herein lies the explanation 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 essentially continuously from zero to infinity. We saw in Chapter 6 that
E~ = 
(1.; + ~ + 2m
I,'
k!)
(/)
'
~ L
'
.. ..
(2)
!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 propagation in crystals . Bragg reflection of electron waves in crystals is the cause of energy gaps. (At Bragg reflection wavelike solutions of the Schrodinger equation 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 = arc not the traveling waves exp(i'7T:tla) or cxp{  i1TXla) offrce electrons. At thesc special "alues of k the wavcfullctiolls arc
""a
7 E_rgV Hcndr
177
,
+''''"'''!c. , .
figure 2 (:i) Plot < energy ~ ,~rsus wa\l:\< c tor k t.' elt.t:tron. (b) riot of.,.,.,rgy \'Cr5\15 "'lI\~.'t:ctor IN WI electron in a monatomic linear lallict < bottk~ constant a. 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.
ror ar,ee
nle
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 Braggreflected 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 timeindependent 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)
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 leftdircctcd tra\'eling waves.
Origin
The two standing wavc.s .p(+) and .p( ) pilc up electrons at dilTerent regions, 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 ::: 142. 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 valence electrons taken off to ronn the conduction band. The potential energy of an electron in the field of a positive ion is negdtive, 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 energy 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:uison ,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 hy to understanding the angin of the COl gap.
F.nergy Haruh
17!
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 ~ ).
Y2 cos
TIle firstorder energy difference between the two standing wave states is
(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 Schroclinger Ct"\uation for a pcr;ooic potential mllst be of a special fonn;
I!her)
= IIk{l")
exp(ik . r) ,
(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 oneelectron 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 functions 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)
eN .p(r)
because (x) mllst be singleva1l1oo. It follows that C is one of the N roots of unity, or
c=
We sec that
cxp(i2m1N) ;
$=0, 1, 2, .. . ,N I .
(9)
(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) .
KRONlCPENNEY MODEL
A periodic potential for which the wave equation can be solved in terms of elementary functions is the squareweI! army of Fig. 4. ' J11C wave equation is
h2  2m
(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 :::
with
CeQ~'
Dc Q ,
(14)
(l5)
181
. ,_ _
u.
Irod llC!
1, Krooig and
PennL1'.
 !+b)  b
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
(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;
'K(A  B) = Q(C D) .
(17)
(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~
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
(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.
' 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" BkdlJi).;e solutions 10 the wa,~ equatioo. so thai lOrb;dclcn gaps in the enc.rgy spedrum an: 5:>r11'll..i
.,
~ :>
i ,
"
10
., ,
./
 "'
~ /
,,
I
,,
,p _ 3'1T12. Notice
.. '"
'"
Figure 6 Plot of .,nerg)' \'$. w;t'I,... lUmbe. lOr the Kronig.I'cnnL")' polential, with the energy gaps at kit  'IT, 271', 3'IT .
1 nerp &n<h
183
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 .
c
(22)
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 decrca.scs as lIG:it. We w.tnt the potcntial encrgy V(x) to be a real function:
V(x) =
c>o
Vc(e IC %+e IC
') =2 L
Vc cos Gx .
(23)
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 eigenfunctions or orbitals or Bloch functions. Explicitly, the wave equation is
_ )2 ( _1 1 2m
+ V(x)
I/f{x) = (_ I p2 _
2m
+L
c
(24)
~;
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 components of the ion motion outside the first zone. The elcctron problem is like the xray 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
_p2.p(r):::: 
1 2m
1 ( d  ifi 2m dr
2mtlr
2m"
elL: ;
C(k) "'" .
(26)
Each Fourier component must have the same coefficient on both sides of the equation. Thus
E)C(k)
U(;C(k  G ) = II
I
(27)
(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
.E_rgy &nd.r
IllS
.~
. 1.:.+ to
 30
.f rt..... . . t
 20  10
Tn ......
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 particular "lI,.,.:tor k  k., ~  8(2.../L). The shorte' rcdj>rocal bttice ,cdar ~ 2fT/a ~ 2012""')'
rrom
the usual differential equation. nle set appears unpleasant and fonnidable because 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.
L C(k G
G) <",,0).
(29)
L C(k G
G) e  /C~ ,
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 verify this directly by evaluating IIA(r + T);
11*
IIdx
+n
... l: C(k 
G)
e  iG (.<+ T ) ""
Because exp( iCT) :: I by (2. 17), it follows that u,rlr + T) "" lIi{X), thereby establishing 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 function? It bas several properties; Under a crystal lattice translation which carries r to r
+T
we have
(30)
because U"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
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.
(A,  .)C(k)
+L
c
UGC(k  G) ~ 0
(31)
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.
0
E
0 0 0
0 0
U A"  E U
0 0
U Ak +1!  E U
0 0 0
U A..... 2gE
(32)
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.
As an example of the use of the central equation (31) for a problem that is solvable, we use the KronigPenney modcl of a periodic dcltafunction potential:
U(x) = 2
G>O
Uc cos ex =
(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 potential arc
Uc =
LI
dx U(x) cos ex
A(I~
LI
J.
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 (36) I(k) C(k  2mJ.) ,
L
"
(39)
We sum both sides over all n to obtain, using (36) and cancellingf(k) from both sides.
(h'/2mA)  'OIC
L [(k "
(40)
sum can be carried out with the help of the standard relation
ctnx""L
..
1
017"+%
(4 1)
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.
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
(42)
+ cos Ka =
cos ka
(43)
which agrees with the KronigPenney 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
with k in the first zone and G aH ov.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 lowlying 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:
...
I
Cal21T
000 100,100 O1o,oio,OO I ,OOI 11 0,101, 110,101 JIO,Io l ,TIo,101 OIl,OII,OIi,OTI 0
( (000)
t (k,OO)
(2,",0)1
" 1..;+
(k... :t 2mof (2mo)lI (k... + 2maf + (27TJO)' (k.  2ma)1 + (27T1a'f I..~ + 2(21'T/o)t
1l1 free elcctron bands are plotted in Fig. 8. It is a good exercise to plot the ese 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' =
100
L __
=:::,_;,_""'__,!
0 ..
i._
"i"
Jo'jgure 8 LowIying free electron energy I~ the emjlty Ie laUice. "5 translOrmed to the fln! Brillouin 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 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)
= 0
(44)
= 0
(45)
These two equations have nontrivial solutions for the two coefficients if the ener'b'Y E satisfies
A E
I =0
(46)
Energy Bond.
191
whence,
E = A U = _(!G)2 U .
h'
2m
(47)
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 ~ ) C(IG)
_E__,
~ + 1 ,
(48)
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 twocomponent 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)
(Ak_C 
;
0
E
with Ak defined as h 2J(l12m. These equations have a solution if the energy satisfies
Ak _ C
 0
o.
(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) ""' (h2/2m)(lG 2 +
(51)
luI.
'"
,
"
.
~ ,
"
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 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 twoIerm a l'l',.oo;j"",,;o1'l to be IK'CUnlle.
"'Im ..
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.
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
L
Nw
' L
(53)
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 near the boundary of the first Brillouin ZOne. OnecompOOlent dominates "" ...e 1110\." away from the boundary.
I'1g11re 10 Ratio of the roelr.cft,Tlts in~)  C(k) np(ikt)
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 81e 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 primithe cell. the '>and can also be ell:actly filled .
'94
1(0)
,;
1 
,b ,
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 a semimc.1al.
'0'
or
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 thcory.) 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 wavefunctions 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 prerequisite 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
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 problem 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 deltafunction 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 comentional cubic cell. (b) Show that in the usual fh~ton:ler approximation to the solutions of the W""d,e equation in a periodic lattice the energy gap \'alli~hes at tile zone boundary plane normal to the end of the \'cdor A.
'"
0.27
0."
O~
i ., ,
E",,'lY gap
.
0 0."
... ,
/
O~
1
."
Rc.l

" , ,,
0.01
."0."
rlgUn:
0."
0.51
pan or l/G
'"
In the energy gap tht.n a1st solutions ofttw. wave eqll8lion roc romple!< values of the the boundary of tho. fil"$t ZOIlt' the real pari of the WI~tttor Is iG. 1be imaginary part of I: in the gap is plotted in the: apprmimation oflwo plan" """"., lOr U  0.01 ""C~I2m. III an infin ite unbounded <.T)'HaJ the wa'~r musl "" retJ, or ebe the amplitude win increase wlthoullimit. But on wr.,e or at aju1)dion then can nist 5OIuliooli with complex \\IltYe\ect<ll".
wa\'e\~or. At
u.
S. Complex ICtIVI!VI!ctor. in 'he enugy gap. Find an exp.ession (or the imaginary part of the wave\'edor in the energy gap at the boundary of the first I3rillouln rone, in the approximation that led to EQ. (46), Gil'e the result for the Im(k) at the center of the energy gap. lhe result for small Im{k) is
(h~l2rn)!Im(k)J2
_ 2rnU 2/f1 2C t .
The fonn as plotted in Fig. 12 is of importance In the theory of Zener tunneling (rom one oond to another in the presence of a strong electric neld.
6. Sqll6re lemice. Consider a sfluare lattice in tw() dimensIons with the crystal potential
U(,;.,,)   4U cos(21t:tla) cos(2wyla) .
Apply the centraJ equation to find approximately the energy gap al Ihc comer point
(wla. wIll) ofthe Brillouin zone. It will suffICe 10 soh'e a 2 x 2 determinantal equation.
References
J. M. lim.n,
Principlt$ 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
EQUATIO,,S OF MOTION Physk:a1 derivation of lil; == F Holes Effective nlass Physical intCl"pretalion of the effective man Effective masses in semiconductors Silicon and germanium
199
203 20:; 206 209 210 212 214
216 220
221
222
224 226
THERMOELECfflIC EFFECTS
SEMIMETALS
221
228
SUPERLATIlCES
SUMMARY
228
231
PROBLEMS
I. Impurity orbits 2. lonb:alilm (lr donotS 3. Hall effect with tv."\) carner types
231
231 231 231 232 232
232
NOT., 1be discussion of carrier orbits in applied fields is continued in Chapter 9. semiconductors are treated in Chapler 17. Junctions and barrierS are treated in Chapter 19.
Amorphou~
Sb
. ,
<
c"",L' I~ ' ~
HI'
_ 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 ""III. and the range <:an be extended downward to merge eventual ly with the Ins"L1ta" ~ ....
C'OIlCIffil
CHAPTER
8:
SEMICONDUCTOR CRYSTALS
Carrier concentrations representative of metals. semimetals. and semiconductors 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 ohmcm. 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' ohmem. 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 formula AB. where A is a trivalent clement and B is a pentavalent clement. arc called HIV (threefive) compounds. Examples arc indium antimonide and gallium arsenide. Where A is divalent and B is hexavalent. the compound is callt. d a IIVI compound; examples are zinc sulflde and cadmium sulfide. Silicon and germanium are sometimes called diamondtype semiconductors. because they have the crystal structure of diamond. Diamond itself is more an insulator rather than a semiconductor. Silicon carbide SiC is a IVIV compound. A highly purified semiconductor exhibits intrinsic conductivity. as distinguished from the impurity conductivity ofless pure specimcns. In the intrinsic temperature range the c1ectrical properties of a semiconductor arc not essentially 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 conduction 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 conductivity 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
V
"
1 /
" "
V
V

.,
! !
,/ .
J" I
'"
H I'
10
"
V
'" ""
",j
liS
'"""
~
(.j
'"
V..
.'"
4"
,
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 ghen 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
""
Si
Table J
EC",N
E.. .. v
c.",",
c.p
OK
5.4
300 K
Crystal
G.p
OK
 0.30 0.286 0.165 0.190 2.582 1.840 1.607 3.436
3.91 0.3
300K
0.340.37 0.'" 0.29 2.421.74
DialOond
d d d d d d
HgT"
I.ll 0.66 0.00 0.17 0.36 1.27 2.25 1.43 0.68 1.
Ge
aSn InSb
''''', I.P
G,p
C..A,
GaSh
PbS
PbSe
PbTe
d d
,
d
1.42
2.32
1.52
d d d
I ." 3.2 3 .
0.18 3.2 2.'
AISb
SiQl1ex)
T.
'7..nsb
,/
0.56
o~'t'r\ap_
2. 1723.00
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 1 1hol011 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 substantial wavevcctor kc. Here a dire<:t plloton transition at the energy of the minimum gap cannot satisfy the requirement of conservatioll of wavevcctor, because 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
or
\.:(photon) =
k.. +
K e O;
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 ....'
A1""'VIion
GAP
0"....1 of lnoJl,c:d
photon ! . . ...iI;""
.'"
'
11~~.,,,
"K"' I: ft .. _
t::....
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 Ihne. 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 electronhole pair, called au eliCUOII.
.,
,,,
"1.
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.. TIM: Indirect t...ulIition in (b) ""'01",,1 both a photon and a phonon because the band ed~ of the eonduction and ,.'knce 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$ E. /1.'01.: n.e rlgUre..haws only the threshold tnlnJi1 10m . T""",itlonl occur generally between almost all points of the two bands lOr .... hictl the wn.,........tors and t'nerg)' ClIn be COIISCOed.
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
"I.> ..
"n.
S"micond"cfor Crlld"!'
203
'~'r,
,, ,
,~
l!
l<
"p I,
"I,
"
rLgU~e
,.,
~
... ineV_
"
6 Optical absorption in pure indium antimonide. InSb. The transition is direct lK...:auSt' or the Brillouin mne , \(  O. Notice the sharp th~cshold . (Afle~ C. W. Cobeli and H. Y. Fan.)
hoth conduction and valenc.o. barn:! cdgtc'$ are at the cente~
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 (Chapter 6). sometimes supplemented by conductivity measurements. Optical measurements 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 (2) & "" eEv, 8t . We observe that
(3)
Ek
(,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 overthrown 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
(ees)
,II< e h::o:vx B dt ,
dI<
(SI) A  == ev x B
dt
(6)
wbere the righthand 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 crossproduct 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.
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 defined by the intersection of this plane with a surface constant energy.
or
or
fhyrical Derivation of ~k = F
We consider the Bloch eigenfunction 4/Jo.. belonging to the energy eigen. vdlue t:k and wavevector k:
'" =
L C(I< + G) o>p[;(1< + G) d c
(8)
(9)
using L[C(k + G)[2 == 1. We examine the transfer momentum between the electron and the lattice 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 crystal system by the impulse ","'Quid appear in the change momentum of the conduction electron:
or
J=
(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 crystal lattice, we must have
J=
aptat
11:1
aPlat
+ ape!
(11)
+L
c
AG[(V,IC(k
+ G)I~ . "k]
(12)
The change aPIa! in the lattice momentum resulting from the change 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
or
with momentum Ii(" + G), thc lattice acquircs the momentum IiG, as required by momentul11 conservation. The momentum transfer to the lattice when thc state th goes over to th+d. is
6",.. ~
(13)
as the portion
+ 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. (17)
",. = k.,. .
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 IlOffiCS 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\'evectorafier the absorption of the photon is k. + Ie.,  0, so thai Ihe lotal wavcv<."Ctor is UI>changed by the a1JSOrplion 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 dectron, 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 neglected. 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 splnorbit 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,,).
m..,
(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 is positive. 5.
dk,
,,8)
(21)
This comes from the equation of motion (CCS) It dk,. "" ...IE
(It
"\
+ ~v II B) c"
(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 carried 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.
:?.,{>'..,, '.
..4!.'; .
c.
,
,.)
'0
0'
.)
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. dirtdion. The on the electrons Is in the  It, direction and all electrons make transitions together in the  k.t\iredipn, ffiOYing the hole the stale E. (el After a further ,n'ervallhe electrons mO\'c tanher along in k spaee lind the hole ;s now at D.
ron:e
a,
drift "clodlles ~ in opposite direction., h"t their tledric currenb arc ill the ",me dirL'Chon, th.: dircdlOll of the ckctric fleW .
Effective MaSl When we look at the cncrgywavcvcctor 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 curvature 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. 10.01 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 valuesee (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
dk)
d~
dP
(26)
112
0'
dv
dill
1i2 dP F,
F = d 2ddP
(27)
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 ')
dt
m~~'"
(29)
, , ,
I
r COO
, v.,
F"tgllre 11 Explanation fA nt.gal;n efICd;~'e mau<.'S ...hichOOOJr .......... Brillouin zone bounJary. In
I I
'"
(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 increases 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 momentum component /i(IG) cancels the momentum component /i(iG). It is not surprising to find negative values for m* just below a zone boundary. 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 tightbinding approximation d iscusscd in Chaptcr 9 gives quick insight into the formation 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 example, 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.
(eeSI
we :::
eB m*c
<Ilil
(301
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 requiremcnts arc met with thc use of highcr frcqucncy radiation and higher magnctic fields, with high purity crystals in liquid helium. In directgap semiconductors with band edgcs at thc centcr of thc Brillouin 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
S~ironduetor
Coyal<lb
213
Figor" 12 Arnmgemen l of fields in (.,....,Iot ron r..sonallCe "XIl".iment ill i5 oppo$ite f.... electrons ..,d holes.
il
semicondlX1or. 1"b..
liJ:I,t hoi...
figu." 13 Simplilkd
V;"'W
...
Table 2 Effective masses of dectrons and holes in !lired gap semioonduclOJrs
Electron
C~tal
Heavy hole
mJm
0.015
0.0"2"i 0.073
m."lm
0.39
0.41
Splitoff hole
Spin orbil
'"..wIm
(0. 11) 0.08 (0. 15) (0. 14) 0.17 0.69
4 , eV
I,,,,
fli P
InSh
e,""
CaSh
CUiO
0.82
0.43
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 spinorbit splitting A:
~(M)
5!!
 1i'lk'lJ2mhh ;
E~(lh)
ill
 1i21?/2m'h
(32)
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 sphericalsec 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 tightbinding approximation (Chapter 9) to the wavcfunctions. The P3I2. level is fourfold degene rate as in the atom; the four states correspond to 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 spinoruit 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()inorbit interaction. The energy surfaces are not sphericd.!. but warped (QTS , p 27 1);
m,
(33)
.;r,,,~_,
, ,
~""IH "'.V ".
5.4 e\l
~~
F.',  O.9<:V
3.8 c
,
~
3
,
.
,
,
,
w
u
 "L,>.C,,~,,_,JL,_cr.cl
k_
!'G.(.!..!...!.) ! 222
im
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 .... 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
,os
(
The choice of sign di~tinguishcs tlle t\.\'O masses. The splitoff band ha~ {k) =  /1 + AJ....2. The cxpcliments give, in units ,,212m,
5i : A:::  4.29 ;
Ce: A = 13.38 ;
;
;
A = 0.044 eV
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 oriented along a (ill) crystal axis, with a longitudinal mass 1. 59 m and a 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
m, ""
(34)
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 conduction 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 FermiDirac distribution function 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 .... <:$ofthc Brillouin WIll'!i of lhc iCcand bee lattices. TIle wile cenh,.." are r . In (a) the boundary point at (2.,./a)( JOO) is X; the boundarypoinl at (2.,.fll}(l l i) is L; the line!J. runS belwCt.'1l rand X . In (b) the COITCSp:mding srmbo1s are H . P and fl .
0."
'"
03
IT
'" i ,
0.2
II
~
~
~
.~
0.20
1/
Figure 16 Eff..'Ctiv" <)dotron mass of dectrons in gt'1"ma. niurn at " K ror magnetic field diTl.'CbOlls in II. 110 plane. Then: an:: lOur independ<.'1lt mass sph<.'ll)id. ill GO', olle along c....n III a,i~. but vi<,...wed in the 110 plane tWQ spheroids alW3)'S appear "'(Iu;valent. (A ller Oresselha"., Kip, and KilteL )
,..
V
I~
II
t.:
I~ '= 4(} "'WffUm 10011 00 "',. lIO pia ..,
a>;l,'
 10 0 10 20 30 Angle in degree> In
" "=
00
'"
'.
,
,
> " ., , ~ " , , , , ,.
el~
'.
Ib,
Figure 17a Con~tanl L.,>erg' ellipsoods b In)IlS in silicon , d",wn for m/ III, _ 5.
(38)
The problem is solved for n when Ii is known. It Is useful to calcu late the equilibrium concentration of holes p. 111c dishibution function flo for holes is related to tlle e lectron distribution Cunction f~ by f" = 1  f~. because a hole is the absence of an electron. We have
flo = 1 
a CXP{(E  ,u)lknT] ,
If the holes near the top of the valence band Ix:have as particles with effective mass mh, the dcmity of hole states is given by
I DIr(E) = 2"r
(2m"),, (E" 7
)112 ,
(4 1)
,''"r:;:r
,.....~~""'"
I
= __"", 1
( )
Figure 18 Energy i>ClIlc lOr statistical cakula1;0.'$. 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 1 lic ",,,II ""Illin thto 0 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 emII, 2.89 x 1()26 emII, and 6.55 x 1012 emII, 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 blackbody photon radiation at temperature T. The photons gencrate e lectronhol.... pairs at a rate A(D . while 8(J') np is the !'ate of the recombi
nation reaction e
+"
d,Jdt
;;0
;;0
(44)
In equilibrium dnJ(it 0; dpldt = 0, whence lip = A(1)B(1"). Because the product of tile electron and hole concentrations is a constant independent of impurity concentration at a given temperature, the introclucBon of a small proportion of a suitable impurity to increase ", say, must decrease 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 controllcel introduction of suitable impmitics. Sudl a I~uction is called compensation. In an intrinsic semiconductor the number of electrons is equal to the number 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",
(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);
po. = lEI:
(46)
(47)
+ ik8T In (mJm.J
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 contributions:
(T
= (nepo... + pep..,J ,
(49)
where nand p are the concenhations of e lectrons and holes. In Chapter 6 the drift velocity of a charge q was found to be v "" qTEJm. whence (SO)
,
(
Cryslal
~,,w,o.,d""lOr
Cry.fab
"I
Electrons
Hol.,s
Diamond
Si Ce
1800 1350
3600
1200 48"
IBOO
eaAs
8000
5000 550 I"'" 2500 50
300
GaSb
1000
600
PhS
PbSc PbTc Agel
KBr (HJO K)
lusb
I, A>
Inl' AlAs
BOO
30000 4500 280 900
AISb
SiC
100 100
The mobilities depend on temperature as a modest power law. The temperature 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 scattering 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 tllennrulyacti vated hopping from ion to ion. 111e principal cause of this "selftl'llpping" is the lattice distortion a~sociated with the JahnTellel' 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 21Vs in PbTe at 4 K, where tIle gap is 0. 19 eV.
IMPURITY COl"iDUCTJVITY
Certain impurilies and imperfections drastically affect the electrical properties 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 stoichiometric 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 elements 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 impurity 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 motions 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 satisfied 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 constant of the medium and the effective mass of an elcctron in the pcriodiCJX)tential 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~.,'
~4'7Tft,;)r.r
. '
(51)
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~
(O.53f) A m/m
(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 ",Iirons. Thus lOu r electrons 011 the arst.'tIic form telr2hedrall'OV1llcll t bonds ..milar 1 ~ilicon , and the fifth electron is ,1lI,tibble for conduction. n.e 0 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 Q the donor energy because it enters as the square, n 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 mJml "" 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"
Crystal
or semiconductors
CrySlll1
11 .7 15.8
Diamond 51 Cc InSb
I"", I,P
c..sb
17.88
14.55 12.37
.'"
AISb SiC
GaA,
CUtO
Table 5 Donor ionizatm energies Ed of pcnta"alent impurities in germanium and silicon, in mcV
p
Sb
39. 9.6
s;
Co
45. 12.0
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 metalinsulator 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 germanium 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 comparable 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 }
,
Table 6 Acceptor ionization cncrgiC$ E. of trivalent impul'ities in gcnnanium and silicon, in mcV
Si Go
45. lOA
AI
G
65. 10.8
"
157. 11.2
57. 10.2
0 ,
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 SiS 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 RCplor 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 3 intrinsic electron concentration n. at 300 K is 1.7 X 101 cm  3 in germanium and 4.6 x 1<f cm 3 in silicon. The electrical resistivity of intrinsic material is 43 ohmem for germanium and 2..6 x J(fI ohmcm 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 common electrically active impurities the shallow donor and acceptor impuritieshas been reduced below 1 impurity atom in 1011 Ce atoms (Fig. 21). For example, 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 (11, 0,
"E . E. Haller, W. L Hansc:n, and F. S. Gou lding, " l'bysic:s 0 ultl"'l'pun: gcrmanium, Ad, . Phys 30. 93135 ( 19EH). "s M. Kogan and T M. uf~hits. Phys. Statu, Solidi (.) 39. 11 (1977).
".
10"
Figu", iI Tefrlpern.lure dependence ufille r~", camer roncen 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.
In at
(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 complicated 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 temperature. An excess of donors will increase the electron concentration and d ecrease the hole concentration; the sum n + 1) will increase. 11le conductivity will increase as n + p if the mobilities are equa), as in Fir "2.
S"micontiuclor Crystal.
227
10" 'i
1
10" ~
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 G drives through it an electric current density jq. If the current is eld carried only by electrons, the charge flux is
(54)
where fJ.e is the electron mobility. The average e nergy transported by an electron 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 TXILe)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~
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)
l1 =QT.
(60)
This is the famous Kelvin relation of irreversible thermodynamics.S A measurement 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 compositions. 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 , "." , ""']" ,
..
\
iJ"
(\n~)
Q
Q
~ .J.
,
...
""
.., ,.,
.. ....
""
(2. 12 0.01) X JcPO (5.49 a.m) X JOIIl
3.00 X 1017
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
7 2.88 X 101
2.72 X lO'S
2.04
10 18
molecularbeam epitax)' or metaiorganic 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 SUPeTperiodic stmcture and acts on the conduction electrons and holes to create new (small) Brillouin 7.ones and mini energy bands superposed on the band struclures 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) The velocity in kspace (momentum space) is
v = h  Idslfk = (A&J/i) sin kA
(62)
(6/)
and the position of the electron in real space, with the initial cond ition z = 0 at t := 0, is given by
z :=:
(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 adjacent 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 midl993). 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 localization 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 fieldinduced 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
Im1near
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 velocity 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 coefficient? 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)(imation is
'"
(CCS)
1 Ii ,f " "  
where IJ = Prip." u the mobilit)' ratio. In the dcrhation 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) ""
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 W '"" eBl(m(m,)lI2c when the static magnetic field B lies e in the 1",1/ plane. This result agrees wilh (31) whe n '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 velocity 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.and relaxation time T~; an d a conccntration p ofholes of effective mass TIl" and relaxa1. (a) Show in t\ili; tion time 1), . Treat the limit of \'ery strong magnetic fields, lim it Ihat ()'~.  (n  p)edB . (b) Show that Ihe Hall field is gi\'en by. with Q  0,,'1,
'llll5."
"leT"
E~ (np)(;,
which vanishes if I I
...
+;J'E"
:I:
din.'CI.ion is
I f n .. " , ()' 0: 8  2 If n
ind ependent of B
35
B_
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 Periodic zone scheme
CONS'"RUCTIO N OF FERMI SURFACES Nearly free electrons ELECfRON ORBITS, HOLE ORBITS. AND OPEN ORBITS CALCULATION OF ENERCY BANDS Tight binding method for energy bands WigncrScitz mel hod Cohcsi...~ ener gy J\cudopotcntial metho.ts EXPEIIIMENTAL METlIOIJS IN FERMI SU RFAC E STUDIES Quantization of orb i t.~ in II magnetic field De Haas\'aI! Alphen cffect btrcmal orbits Fermi surface of copper Example: Fermi sur ratt of gold Magnetic br eakdown
2.3:!i
'" '" ,,
...
252
Z55
...
200
SUMMARY
PROBLEMS
I. Urillouin zoncs of rectangub.r lattice 2. Brillouin 7.llOC, rcdangular lattice 3. lIeugonal dosepacked structure 4. Brillouin zones of twodime nsional divalent m etal
5. 6. 7. 8. 9. 10. II.
Open orbits Cohesive energy for II square wd l pote ntial I)c HIUlSvan Alphen period of potassium Band edge structure 00 k . p perturbation theor), Wflnnicr runctions Open orbits flnd magncto .. csistancc Landau levels
.,
REFEREKCES
... ...

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 coppeolw been deformed from a ~phere 1 argro:e 0 wilh the exper imen lal results. TIll'! serond zone of aluminum is nc... rly h.lffilled with .,kctrons.
(A. R. MlIICkinlosh.)
CUAPTER
9:
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 racecentered 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 concentration. How do we constnlct these sllrL'l.Ces rrom a sphere? The t"Onstructiolls 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) = e1 k'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 'cu".(r
iil
(I)
where "k(r) .. e IG'uk.(r). Both e .c :md u".(r) are periodic in the crystalla.ttice, 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
(
f
flrst wn.,
Figure 2 First Bril100in zolle d a square lattice or side d . 'l1le ""lI\", ..n . e(or 10:' can be cwried into the 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<~~ ~
FiJl;lIrl' 3 Enf'rg~'wa'c'f!'d.or re lation fl  fi 1k"/2m f ....'C electrons as drawn in the :<lured:w~ OT M:heme. n'i~ construction ofIen gh"es a lIiOCul 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 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.
<'I,""'' '
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 characterizes a different band. Two wnvcfunctions at the same k but of diHerent energies will be independent of each other: the wavefunClions will be m ade up of different combinations or the pLmt: wave components exp[i(k + C). r] in the expansion of(7.29). Because 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
~
L c
ClICk
+ G) cxp{i{k + G) . r]
(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 equation (7.27). Consider for example nn energy band of a simple cubic lattice as calcubted in the tightbinding approximation in (13) below:
Ek
=  a 
2y .'Os kxU
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
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
c.
0 0
2 ', :}
o
G,
0
0
0
0_
"
,,,
0
..,
(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 Gs. (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\vhed in the COIlllrl.ll1ion of the ou , ..... boundary of ,he region.
the area enclosed by the perpendicular bisectors ofC t and orthe three reciprocal lattice vectors equivalent by symmetry to C. in Fig. Sa. TIlese four reciprocal 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 section 3., of the third zone are fonned from the perpendicular bisectors of three Cs. namely (2r.1aik,; (4r.1al~; and (2r.1a)(k" + 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.
kv).
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 ..,... 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.
"'"It
A third 7one 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.
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.
tEt1f~, , , , ,
,
, ,
, ,
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 5Uracc like that o;f thc third lOne is clcctrooI,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 perpendicularly (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 procedure credited to Harrison, Fig. II. TIle reciprocal lattice points are determined, and a freeelectron sphere of radius appropriate to the electron concentration 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 similarly 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 boundaries are distant from the approximatciy spherical Fermi surface that flUs onellalr 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 slllra<.'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. \
2ta
Figure 11 Harrison <'''OlUlruction of frt'e cl1ron 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 eketrollS.
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 $urra"..,llopoIogka!Iy cquhalcnl to tllO$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 eiLdron ~t~ tcs lie inside the Fermi s" r&ce. We ....11 the orbi t in (b) electrunlilc. 1bc $Cnse of lhe motion in II magnetic Ileld is op~ile in (a) to that In (bl, SO that we rder to the orbit in Cal as hoIclike. A holc 1IlO' ..,S as >I pnrtide of pn<ih\"e charga: e. In (e) for a n.'C!angular nme ".., show Ihe motion on an open orbit In the: periodic zone KilClTle. This is topologic.JJy inlenm.diale betv.cen a hole orbn and an elcctron orlHt.
F"'.gure 13 (a) Vacant Sillies at thl, <."(lrnCrs of an al'llOIitfillc.d band, drawn In the reduced zone ...hemc. (b) In the periodic zone Kht.'me the 'arious l)IIrh of Ihe Fermi ~"rfal"C are con n<.'Ctcd Ead! circle ronnl a holelike orbit. The diFl"c.TCnt circles arc entirely t~lui\... IL"rlt to tacb othc., and ti!edens,ty of ~tatC$ is that of a ~nglc arclc. (ll.., orb;b ",oed not be true cirdcs: bthe 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 twodimensional 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 fllkd \Vlume COIltains One ekct,on pcr primitive (b) TIIC same .uri'ac<: exhibil(."Il in the period;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.)
<.'en.
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 wavcfunction. 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 lightbillding method. useful for illterpoiation; the WignerSeitz method. useful for the Visualization and understanding of the alkali metals; and the pseudopotentia! method. utilizing the general theory of Chapter 7. which sho\'''s the simplicity 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.
'"
" ...
0.'
l' I ~
~
 0.2
 1.0
"
2.6
3.4  4.2
h"o:areot ~ d~.
4 in &hr radii
Figure 11 The Is band or ~ ring ofm hydrogen atoms; the ollet'ledron energy olk... laled in the lighlbinding appro>: ;malion ,,~th the nea",stncighhor 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 degenerate (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 J
rj)
(5)
We prove (5) is of the Bloch form . Consider a translation T connecting two lattice points:
241
L exp{ik . r)/pcr + T J
I12
== exp(ik' T) N 
L exp{ik' (rJ J
(6)
= exp(ik . T)I/Jk(r) exactly the Bloch condition. We find the firstorder energy by (.akulating the diagonal matrix clements of the hamiltonian of the crystaJ:
(killik) ~ N  ' L
J
"")] ( ...IHI.)
where 1,0,,,
s:;
(kIHlk)
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: ;
(9)
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 separation. For a simple cubic strudure the nearestneighbor atoms are at Pm = (a,O,O);
SO
(12)
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 
(14)
ik~
+ ('OS
ik~
CQs ik..a
+
cos ik~a ('Os lk~)
(15)
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)
(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.... estndghbor t ~t
hi"dinJl: 3I'Proxi"",t;OIl .
t 
 0
+ 21li.
atn
(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}' uSCful 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 2m
1I(~r) =
EoIIp(r) ,
(19)
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 mllulating " o(r). Figure 19 shows the WignerSe 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 markedly 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
vv
. '.'.
,
.'. '"
" ..,k  O ,
~I
,.,.. alorn
I
,
 """""'"
kla!.
~ oJ
Brillouio
r(8<ior
u~u.)
Figure 19 Rad ial w"\'cfUnctions lin the 31 orb ital offrce sodium atom and & 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 $Odium; their enagy is +2.7 eV. (ARn E. \Vlgr>el" and
"r
F. SdtL.)
.,
~
, , " "
0.45
,
 0.6
1 >
~
b/d
Figu re 20 Cround orbital (k .. 0) en ergy rOo" :.on electron in a periodic ,;qUlI", well potential or depth !UJ  'lIl1/mat . The energy Is lowe red as the ",ells come doser togt:lher. Ikre a is hek' comtan t and b is varied. Large bla cocretpond s to sel>al1ltcd atoms. (Courtesy ore. Y. Fong. )
 ,'; IS,"
,Cro..nd
f "dt,
A,~rage
CoI""i,'~ ~""'IY
FIgure 2 1 Coh('Si~ energy uf sodiu m 01<:1,,1is the {lifl'erenw l>el~n th e Hvcrage energy of an dcc1111 in the mela! ( 6.3 tV} and lhe Jo(rOund slale energy ( 5.15 cV)ufthe valence 3" ek'Ctron in the free: atom, relerr'l!d to Ilfl Na ion .,ILI$ ireI' electron ..r. infinite It'(l'Iralion.
 8.2 )::::::~
_ 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 decrease 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 adjacent atoms. III a spherical approximation to the shape of the smallest WignerSeitz cell we usc the WignerSeitz 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 nearest 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 represented in a rough approximation by wavcfuncfions of the form (18). with
EIc=Eo+ .
h'l<'
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 orthogonal to the wa\'cfunctions of the core electrons. This alll'Omes out of the Schroclinger 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 Singlycharged 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 pseudolJOtential 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 theorem.
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','
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.,;
forr>~ .
(21)
This potential should now be scrccned as described in Chapter 10. Each (.umponent 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 ThomasFermi dielet.1ric function (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 values 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 p0tentials, 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 potentia] U(O) = E,.. In the remarkably successful Empirkal 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 EPMsee Fig. 3. 11. TIle rc.~u hs are in cx(.'Clien t agreement with xray 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 contributions 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 determined 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 pseudopotential calculations the basic inputs arc the crystal !itru(.1ure type and tIle atomic number, along with welltested thcoretkal approximations to exdlange energy
O4+1f'j~3__i.
P'I,"Idopol<'fl.lla l
o  
of
~
~ ,
,
.. .. . i :
 ...........
y~._..I.,wt.'"\t;il
ti ro\.
.._
...\:~..;;'".,
Ionic potentia!
Figure 22.a Pseudopotential fur melallic sod ium, based 00 the empty core model and screened by the ThomasFermi 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 VC are 50.4,  B .6. and 4.6, for r " 0. 15, 0.4, and 0.7, reSIlf!Cr) 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.
O~~c~.~c~.~7'~:;::O,.
c,
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 ' crecoedion 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 experiment in the table that follows.
Lattice
Cohe!iive
constant
(AI
energy (.VI
Silicon Calculated
Experimental Cermanium CaIClliatt:d
E~perimclltal
5.45
5.43 5.66 5.65
3.60
4.84
4.63
4.26
3.85 8.10
3.57
7.35
Powerful experimental methods have been developed for the detennination of Fermi surf.1ceS. "Mle methods include magnetoresistallce. anomalous skin effect. cyclotron resonance. magnetoacoustic geometric effects, the Shubnikowde Hans effect. and the de Haasvan Alphen effect. Further information on the momentum distribution is given by positron annihilation, Compton scattering. 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 Haasvan Alphen effect because it exhibits very well the characteristic 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 momentum or field momentum Pfkkl = qAlc. where q is the charge. The vector potential is related to the magnetic field by 8 "" curl A. The totaimomentum is (CGS)
t
P "" Pki"
+ PIM<l
:z:
~k
+ qAlc
(22)
Following the sem icla~sicaI approach of Onsager ane. .fshitz. we a~sume that the orbits in a magnetic field are quantized by the BohrSommerfeld relation
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)
(250)
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
(25b)
where I is the magnetic flux contained within the orbit in real space. We have used the geometrical result that
I(
("5c)
by the Stokes theorem; here du is the area e leme nt in real space. The momentum path integra! is the sum of (25b) and (25c): (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) .
1
(27)
The flux unit 211'f1de.:: 4.14 X 10 gauss or T In the de Haasvan 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
cm 2
m2 .
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" .
It follows that
(28)
4>"
(29)
2..
(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
(31)
from (30). We have the important result that equal increments of liB reproduce similar orbitsthis periodicity in liB is a striking feature of the magnetooscillatory 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 llaast:4n Alphen Effect The de I Iusvan 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 twodimensional (20) system; in 30 we need only multiply the 20 wavefunction by plane wave factors 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 orbits 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~ 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~, .
'"
"
'<I
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
(33)
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 accommodate 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 degeneracy 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$;"'e cirdcs COrffspond to StiO.'Ceui.. e \";I.lue~ of th., between sua:essi,.., circks II Quantum number n in the energy {n  Ill:...,.. The _
(ceS)
n.e angular positiln of the points has flOlignif.call~e. "The numberrI o.bitab 00 acircle b ('()Clltant
and is t.'Quai to the area between Sll<:USSJ.e circles times the nwnber of orbitab per unit area in (a). or (2 .....BI fu:)(U2.1I"' '" L'"Bf}.Trfrc. neglecting electron .pin.
I t
.CO
%r1J,t,"t1000B
0"
I'1gure 2.5 (a) "The hea\')' line gi"es the number or particles in Ic,"Cl~ which arc completelyoccupicd in a magnetic field B, for a twodimCO$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
..
(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 om e moment he re is an o!iCillatory fUnction of liB , Fig.~27. This oscillatory magnetic moment of the Fe rmi gas at low temperatures is the de Haasvan 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' dominates 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.
261
ToI.I
"""rIP'
"
.5
,
1 0018
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 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 nonspherical: eighl necks make contact with the hexagonal faces of the firlit Drillouin zone of the fcc lattice. The e lectron concentration in a monovalent metal with an fL'C structure is n = 41 the re are four e lectrons in acube of volume cr. t?: 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 hexagonal 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 S
eo:
21relflc .o.(IIB)
iii
9.55 x 10 2 x 10 8
1
ill
8 16 2 4. x 10 em
From Table 6. L wc have k"  1.2 x H em I lOr a fn:."e electron Fermi sphere for gold, . i' 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
"' ..
t'igure 29 Fennl surface copper, after Pipila rd. The Brillouin :woe 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.
or
or
. '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 Haasvan 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
 ( 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 pku'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 importan<.'C. However, we know that at low magnetic fields the motion is d~cribed
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 increased 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 magnetorcsistance that depend sensitively on the connectivity. The condition for magnetic breakdown is that "w~,. > 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 spinorhit 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.
E :.
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 surfate. 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 WignerSeit. The periodicity in the de Ha:lsvan Alphen effect mcasurL"S the extremal crO:'issection 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.pnoIu~ physIcs, 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 twudhncnsiooal luttice with ruces G, b  36.
2. Brillouin ume, rtangular lattice. A twodimensional 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 clorepacked 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 hexagolmldost.... packl'tl structure? 4. Brillouin :one, of twoJimenrional Jit."(llerlt metoi. A twodimensional 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 formld 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 Shubnikowde Haas ('ITet.
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 sccondonkr perturbation
thL'Ory tu Gnd the result
(39)
Slim
f"(O)
fj.O)
(40)
The mass at tll<: l'Onduction l;ond L"tige in a narrow gap semiconductor is oftell duminatLcl 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 definld in terms of the Bloch functions of the same band by w(r  rJ = N  1I2
(<2)
where r" is a lattice point. (a) Prove that Wannier functions abou t different lattice points n,m arc orthogonal
(43)
This orthogonality llroperty makL'S the functiuns often nf grL'IIte r usc than atomic orbitals ccnterLd on differcnt lattice sites, because tile laller are not generally orthogorwl. (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"
c!lBlc]' .
form
(,.if'lm)tPJld,r
 lm~ 1I
 !lJllx" 0
wllcrc ~  eBlme ilI1cl Y dlkJeR (b) Show thai this is the wave (.'<luation of a o"
hanllQllic oscillator with frt:qucnl,), We. where
References
T. E. .'abeT Introduction 10 tile llleOI"JI of f"l,,/d mettlb, Aoldcmic Press. 1974. .
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$d/ll1lionr in I/IcUlb. ea."bridgc. 1981. C. R. St~"Wart, "' kavy fermion s)stcm ~ Revi. Mod. I"')'s. 56, 755 (I984). A C . Hewson, Kondo problem 10 I~JV!I f enuloWl. Cambridge, 1993.
l'SEll DOPOTEf'iTIALS
Solid ~tlltc phpics 24 (I97Ojlf\lains three articles oro Im:udopolcnt ial~ by V. Heine. M. l... CohCll,
and O . Wea;..,.
10
Plasmons, Polaritons, and Polarons
DIELECfRIC FUNCTION OF THE ELEcrRON GAS Plasma optics Dispersion relation for electromagnetic W',1ve5 Transverse optical modes in a plasma Tran.~parcncr of alkali metals in the ultraviolet Longitudinal plasma oscillations 271 212 273 214 214 216
PLASMONS
218
ELECntoSTATlC SCREEI\'lNG Screened coulomb potential Pseudopotential compor'K:nl U(O) Molt memlinsulator trnnsition
Screening and phonons in metals
287 291
294 204
207
300
301
PR08LEMS
I. Surface plasmons 2. Interface plasmons 3. Alfvtn W.lVC5 Helicon waves 5. ptasmQn mode of a sphere
302
302 302
,.
302
303
303
6. Magneloplasma frt.''Qucocy 7. Photon bra nd, at low Wi\vt.'vt.'(.1or Plasma frequl.'1lC)' and ekaTkal conductivity Hulk modulu.s o f the Fenni gas 10. Hesponsc of e\tttron gas II. Cap plasmons ano the van oer Waals interact ion
,.,
303 303 303
".
,.. ""
~
REFERENCES
NOTE, nlC h:~t and problems or this chaptL'r aumc fM:ility in the use or ek'(.i:Tomagndic thL'Ory at the k'Vcl a good senior COUr.iC.
...
,, ~~~~~~0,5 1.5
,
_. L _ __
I I j
, I
I , I
, "
" ".
\
..... hen .. is ~itive.nd real. is ncg;.tive .
Figure L Dielectric function c(...} of a free cl.,ctron gas \.:rs ... frequency In units of 1M pluma
frequency "'... E lectromagr>ctic WilYC$ Iln>pagale without damplngr" t: k(:tromagnct." wavc) arc.: 1()ldlly rdl~'CIed fro,,, the Illediu m "It
___________________
CHAPTER
10:
The dielectric flmction ~w,K) of the electron gas, with its strong dcpcnden<.'C on frequ ency and W3vcvector, has significant consequences for the physical properties of solids. In one limit, E(w,O) describes the collc(.1ivc excitations of the Fenni seathe volume and suruu..e plasmons. In another limit, E(O,K) dcscribes the ek"Ctrostatic screening of the electronelectron. electronlattice, and c!cdronimpurity 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 COIlccm(.>ti with tile electron gas in metals. Dejinition1l of the Dielectric Function. The d ick>ctric constant f' of electrostatics is defined in lenns of the electric field E and the polari7.ation P, the dipole momcnt density;
(CCS)
D = E + 47TP = EE
+ 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 E "" 4np = 4niAxI + Piool div D = cliv EEOE =. Pco. l
rov E
(2) (3)
'w.
Paris oflhis chapter will be written in CCS; toohlain results in SI, write I/Eo for
."
(
\\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
("')
(31))
(&)
Each of the .:qllations must be satisfit.'tI term by h.:rrn ; we divide one by the other to obtain
p(). 1
(3d)
The clt:ctrostatic potcntirulpr,' dt:.fint:d by  Vtp..., .. D satisfies the Puiswn <'qllation 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<)
$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
,.~
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)
JO
273
(
The dielectric function at frcqucm. w is y
D(w) f (w)     l E(w)
(CGS)
(51 )
, (w) =
D(G) lEoE(w)
:=
1 +   .
pew) oE(w)
The dielectric function of the free electron gas foll ows from (6) and (7):
(CGS) <{w) = 1 
47Tflt?'
"..!
iSQ
w) = I
(8)
(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
IW)= I  ~ , I
E(w) == ,,(00)  41Tnt!lmWl = (oo){ l  wVw2] wh{.'re
(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 ....
(ll)
wI' is
defined as
(12)
Nolice that
Ii "'"
0 3t w = wp
We look fur il solution with E <X cxp( iwt) exp(iK ' r) and D we have the dispersion relation for electromagnetic waves:
(CGS)
o(w,K)w' =
= E(w, K) ; the n
(14)
C'K' ;
'"
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 dElit, 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.
(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 > w for here the dielectric f\lIlCp 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 1010 1014 I3 W S1 5.7 x lOiS 5.7 X 1011 5.7 X LO 5.7xlcf p 3.3 X 10 3 3.3 X 10 05 0.33 33 Ap. cm
Transparency of Alkali Metal. in 1M Ultraviolet. From the preceding discussion 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.
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 .
u
1550 15S0
N.
Rb
'090
2100
287.
31SO
3""
3400
276
The zeros of the dielectric fundion determine the frequencies of the longitudinal 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.alion ficld E  41TP, discussed below. Thus 0  E + 41TP = 0 fot'" a longitudi_ nal wave in a plasma or more generally in a clysta). In SI units, 0 == EoE +
CI
P=O.
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 lowfrequency 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 displacement 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 concentration II is
(eGS)
0'
(19)
(CCS)
  + W211=O dt'
d'u
41Tncl) (   '" m
(20)
This is the equation of motion of a simplc harmonic oscillator of frequency w,.. the plasma flequenc),. 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
lO
"b _
277
Figu~"
Ind)rnt~
,
'. : t '
t
...
..
,.:,
........
.. ..
..
..
.. o. :
., .. +"...,
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 dielectrks are included. The calculated plasma energies 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.
Volume plasmon energies, in eV
Calculated Material
Table 2
Ohscl"ved
'".
8.02 5.95 4.29 10.9 15.8 16.0 16.0 12.0
"w,.
7.96 5.58 3.86
MetCl/s
N.
K
r.lg AI
Dielectrics
Si
G.
InSb
 .'"""
!M::>tlt'l'<'<l ded",,,
figure 6 Creation of a plasmon in II metal film by iudas!ic srnlIt,'fing of an ck"drun . The incioc'fll electron lyp;.,.."]I)' has an CIl<'lJ;Y I 1 10 keY, th., plasmon cncrgy may tJe of the order of 0 10 eV. An evt,nt is alo 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.)
.1
" , ," ~ ,
~
j
.~
'I>
'"
E""'troo
.. .,
'0
,
50
j"
" " , , ,
"" '"
" '"
" '"
.,
'"
(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 combinalions 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.)
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 electrostatic field. We start with a uniform gas of electrons of charge concentration noe superimposed on a background of positive charge of concentration nac.
(24)
The electron gas will be deformed by the combined influences of the electrostatic 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(.'Cntration to the electrostatic potential. We develop this connection in what is called the ThomasFermi approximation. The approximation consists in assuming 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
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 2m
2 1i 213  e!f(x) a  [31f nol 2m 2
(27)
at absolute zero, according to (6. 17). In a region where the electrostatic potential is Ip{x), the total chemical potential (Fig. 9) is constant and equal to
I' = ~,.{x)  e<p{x) a [31f n{x}]
Iii
213
2m
(28)
ere E.I{x) is the local value of the Fermi energy. The expression (28) is valid for static electrostatic potentials that vary slowly compared with the wavelength of an electron at the Fermi level; specifically. the approximation is Cf ~ k,.. By a Taylor series expansion of ~F ' Eq. (28) may be written as
\.\o....
dE,..
'no
("')
From (27) we have d~Fldno = 2l':F13no. whence 3 elPCx) n(x)  no iii! 110 .
~F
(30)
111e leftll3nd 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
" K)
(33)
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 ThomasFermi dielectric funct ion, and Uk. is the ThomasFermi 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:
O ,K) = 1+ _ . K' '
,1
W,O) = 1  ~ 2.
W
'
(35)
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 wK 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
(36)
(37)
dK exp(iK' r)
(38)
ruK)fq{K) = K) ,
where fP(K) is the lotal or screened potential. We use E(K) in the ThomasFermi fonn (33) to find
Owq
K2
+ k!
(39)
4wq
1 0
dK
J' _)
q
I'"' dK K&in Kr
0
J(2
+ k.
2.
= 
,.
(40)
I" IJ)Od discussion of t he undhard dielectric runction is given by J. Zimao. Prindpla ofrhe T Ilw:ory 0/ aclids, 2nd 00_, ~mbrldge, 1912. Chaplt. 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 calculationS with all wavf,.vectors indud.:d (,,,hihit spatial oscilblions, called Frit.... d oscillations, in 2Jr. .. r
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 potential 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)
The result (34) for
= ezn!#'CO) = 4'1TZlk~
k; = 6'1TZnrrIEF
(41)
whence
V(O) =
 JEF .
(43)
(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)
With no"" 1/0 3 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 metalinsulator 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 independentelectron model; the insulator phase may suggest important electronelectron interactions. Sites randomly oc;c upk'1l introduce new and interesting 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 metalinsulator 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).
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 <."QIlcen1.... 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. Experimental results for P atoms in silicon are shown in Figure lOb. Here the insulatormetal 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\'ilydoped 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 function 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 ThomasFermi dielectric function
(47)
Provided also that the ions are weUspaced 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 ekc. Ironic polarizability of the iOIl cores, is
47171r I.'; .(wK) = i    + M~ Kt.
(48)
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
and this function has a 7.ero when
I I I =+p.
(51)
AI
(52)
This is the e lectron plasma frequency, but with the reduce<! mass correction fOr the motion of the positi\'e ions.
281
0.050
'/
A"'w
, U..,
; 0.044
>
, , , : , ,
'"
...
U"'"
Figure 11
.' ..
0'
, , ,
,
1: , , , , ,
01 O .!l 0.3 0.4 0.5 0.6 0.1 0.11
~:
A lot of the observed energies .nd ",..,..,,ectors of the polariton .:md oftlu: 1.0 pIoonons ;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 phononphoton 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!Ctors 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 absence of any coupling between them. In reality, however, there al\\.'Sys is coupling implicit in Maxwell's equations and expressed by the dielectric function . TIle quantum of the coupled phononphoton 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)
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 polarization 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 negatht: 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 polarization .
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 coupling with photons. The dielectric function obtained from (54) is: e(w) = I
41TPIE  1
+Wr .
4'tTNq'lIM
w
(51)
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)
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)
, ,
.,
;
  "'''' K
,K
.., ~
,
.. 't~/''_::::=:::::::::::: ========~~~~~~'~~;=~= ,
1.0
"
Figure 12 Coupkd 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: cl1l"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!lctromagllcUc 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('lricmechanicalaspeds 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)
wrW
:::
E{o:o)
(wI.  (2) .
wr w
.2
lI'
(60)
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:(]eed at the bou ndary.
' ,", .. , ,
,
 I
 I0
.,
..
~.~.W
t""!"""")" in lit
. . ..
..
'~~
,~
" Igure 13b I);ekclric function (rea1 INrt) ofSrF. me~uud O'o"er a wide frequency 1"lIlge. exhibiting the dOXTea5e of the ionic polanu.billl), at high rreq...:nc~ {A. 11011 lIippd.}
. ,
L,
ro phc....,
f
............
<I
............... ... + ...... +
.,
, F~'re
,t
'I
L.
WI""""""
,. 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 .. macroKOiJic 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 argument, 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 highfrequency limit of the dielectric function, defined to include the core electron contribution. TIlis result is the LyddaneSachsTeller 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 rattng 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 die1ed:rlc
Opti(:ai
dielectric
3 ""r'. in 10 1 5 '
experimental
CO Il ~lanl
constan t
01
12.9
" I
3.6 1.9 2.7 3.' 1.7 2.25 2.6 1.5 2.1 27 1.9 2.6 2.6 3.0 5. 1
II.
8.9
12.0
13.2
5.1 5.9
6A
5.5 4.65 5.1 6.5 5.5 7.2 5.65
31.9
nbF
Itbl C,CI Cd
SA
),9 3. 1 1.6 3.0 1.9
IA
),9 1.2 1.2 0.81 ),9 1.5 7.5 6.9 5. 1 4.3 5.7 4.1 3.5
14.9
11C! 11B,
AgCl
AgBr MgO
29.8
12.3 13.1
SA
4.0 4.6 2.95 8.5
10.9 14.4
3A
2.5 14. 7.6 5.5 4.6 6.5 4.5 3.7
17.{1
9.8
10.7 12.9 16. 1 12.4 14.9 17.7
C,P CaA,
C.Sb
InP
InA$ inSb SiC
9.6
12.3 15.6
C
Si
9.6 5.5
11.7
15.8
6.7 5.5 1l .7
15.8
ee
25. 1 99 5.7
10
293
(
'OO,,~='_~r_,_.__,__.
..
,I
I I I ,J
r!
!
!
I /
I ,i
"" ,
I:
.
'000 ,.,
..
..
,
'00
Fi8u re 15 Rdl"ctance of aCTyStal ofNaCl at Jeveral ternller:illllres. \o't:nu~ wavelength. n.e oominal 'lIll1el of ...... and...,.. at room temperJlure c:orn:s~ 10 wan,lengths ci 38 and 61 )( IO~ an. respectho:Iy. (After 10.. Mibllishi el aI.)
..
.6
i"I': r
4I I I I I
. ..I,.,. ..
..
,
" P 11U!arir.ot..., ,lOJbri>OlIia,
I I I
,
I I I I
.
,
r=
I."ngil.,.]ina! V opt;...!
I Tr..l'I'~
,hmoo
I , , I
12..5
'"
~. 6
.,
"
Fizure 16 R"flectarK'e ,.,nus wa".,kngth cli UF film bacl<ed by siher, 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.)
,,.
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;
w,}w.(
NaI 1. 44 0.05
KU,
[E(0)/(00)JII2
1.45 0.03
Fermi Liquid. Because of the inte raction of the conduction e lectrons with each other through the ir electrostatic inte raction, the e1cctrons suffer collisions. 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 electronelectron 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 ninteracting fe nnions; the same system with interactions is a Fermi liq uid. lAndau's theory gives a good account of the lowlying single particle excitations of the system of interacting electrons. These Single particle excitations are called quasiparticles; they have a onetoone correspondellce with the siug1e 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 effective mass of the electron; in the alkali metals the increase is roughly of the onler of 25 percent. ElectronElectron 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 e1cctronelectron 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 twobooy 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 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 1 final 0 .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 jthe 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 pmvides 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 e1ectronelectron collisions take place is reduccd 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 electronclectron 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 elcctronclcclron collisions is to reduce 0'0 below the value expected from the Rutherford scattering equation for the unSCfccned coulomb potential . However, 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  10em 2 . The mean free path for electroneh."Ctron collisions is f .. 1I11u  10 4 em at room temperature. This is longer than the mean free path due to electronphonon collisions by at least a ractOl' of 10" so that at room tempcrature collisions with phonons are likely to be dominalli. At liquid helium temperatures a contribution proportional to 'f2 has l)('Cn found in the
(
resistivity of a number of metals, consistent with the form orthe electronelectron 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 electrons travel long distances without colliding with each other?
111c most common effect of the electronphonon interaction is seen in the temperature dependence the electrical resistivity. which for pure copper is 1.55 microhmem at we and 2.28 microhmcm 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 number of thermal phonolls is roughly proportional to the absolute temperature, and we find that thc resistivity increases apprmcimately as the absolute te mperature in any reasonably pure metal in this temperature region. A more subtle effect of the electronphonon 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 crystals the effect is weak because neutral atoms have only a weak interaction with electrons. The strength of the electronlattice interaction is measured by the dimensionless coupling (."Ollstant a given by
or
a=
deformation energy
hwt.
(64)
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
KCI
Crystal
..
KB,
AgCI AgIl,
ZnQ
PL'
0.16
InSb
CaAs
a
m~m
m/m
m;.Jm*
3.52 0.93
0.43
200
0.51 0.35 1.5
0.85
0.014 0.014
0 .014
0.06
2.2
1.0
ma~s
Theory relates the effective mass orthe polaron mf,ol10 the effective brul(l m" of the electron in the undcformed lattice hy the relation
mf,ol eo: mt' ( 1
I  O.OOO8a' ) + O.OO34a2 ;
'a
(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 selftrapped 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 selftrapped than are electrons. Holes appear to be selftrapped 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 (ohmcmr', but a family of compounds has heen rcported1 with conductivities of 0.2 (ohmcmr' 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).
,.
0 eEIc<:l:...,
8
(0
'"
(0
(0
(0
0
(0
0
(0
o 8 .
0
0
(0
G
(0
G
0
Ib,
~
Figure 19 The formati ..... of a polaron. (al A comluion 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 thlUry m:l.<' in a rigid 1.11;. In c~treme situations, oilen wilh holes, the particle can l>erorne selftrapped 00caIi.wd) in thelattice. In CQ\aknt crystlk the IOrca OIl the atoms flOm the c leron are weaker than ;n iunic e~tab. SO that polaron deformations ~ smlln in ('Ow]cn! cry.tals.
JOO
Consider a oned.imcnsionru metal with an electron gas fillin g all conduction 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 energy gap at tile Fenni surface, thereby lowering the energy of electrons helow the energy gap, Fig. 20. The deformation proceeds until Iimitetl by the increase of elastic energy: the equilibrium deformat ion A is given by the root of
dl1
<I
(E"Iec:t .....""
EdullJ = 0 .
(66)
Consider the cla~tic strain A cos 2k,.'X. The spatialaverage clastic energy per unit Jenb>1h is E,,\titlc :: iC.o,2(cos2:2k,.'X) = iC.6. 2 , where C is the force conslant 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 proportional 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 ;
  : O'~K+:"cA~'~~;~"'"~ d~ (r
whence. with rlKhr as the number of orbitals per unit le ngth .
dE"
 A 2A
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) whence
I
(68)
(69)
t k .'
10
Polaron~
301
011<, Iolrodu<t"",
"~'W
101""'1"'" CDt~
Fi,,", 2() P~ierh inSbhility. Electrons "'ith ",.. YCYfflor~ nt"ar the: Fermi surf.'Ia: have their e nergy lowered II)' a Iatt~ dd"~ion.
or superconductivity,
The result is of the form of the energy gap equation in the DCS theory 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 electronelectron interaction energy, then we can write (69) as
(70)
which is analogous to the DCS enerb'Y gap equation. II An example of a Peierls insulator is TaS3'
SUMMARY
p~><I(w, K) +
p_,(w,K)
PInd(w,K)
ill terms of the applied and induced charge density components at w,K. The plasma frequcncy = 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 electromagnetic waves in the plasma. The poles of the dielectric function define Wr and the zeroes dellne
WL'
w,.
eA revle>\o tJ charge density wavell and ~"pcrlalticcs in melallk byercd ~Iructurc~ b given hy
'"
In a plasma the coulomh interaction is screened; it becomes (qlr ) e~p(k.r).
A polariton is a quantum of the coupled TO phollonphoton ficld~. The coupling is ass ured by the Maxwell equations. Thc sp(..clrnl rcSioll wr < W < wI. is forbidden to c1ectroma~nct ic wave prop,agatioll.
l1ic LyddilllcS.lchsTcller relation is
WfJW?
== E(O)lE(ce).
Probletll3
I. Sur/ace p'a.smo~. Consider a semiinfinilf' plasma on the pruilh'c side the 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
or
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 StemFerrell rC'm lt :
(71)
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 interdc.c have Ihe frequency
Consider a solid ""ith all e<lual concelltrntlon n of e1cdnms of mass 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 ....rittell as p + = rlc'(m~ + m.)+f1I2, amI the <Iidectric function <'(w) 2 f/ + hrP+f+ =" ~ + 4m!plII , ... here "/ is the <Iidedric constant uf the hust lat 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>.lgating 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 clectrol1llOle (Irops in germaniullI (Chapter II).
4. lie/icon IOOVU. (a) Employ the method
Problem 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 drcularpollri:red modes the sign d the frequency refers to the lense of the rot,llion.)
or
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 polariu 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 cxcitation is called a oollecth'e excitalKm or collecth'e mode of the electron gas.
or
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 toovetJCor . (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 frequCllC)', anll 'T == 2.83 X 10 15 s ror the electron relaxation titne at room tempemture. (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.
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 ~hmcm; 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 twosemiinflnite 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 solutionsthat is, solutions of the Laplace t,'Quation rather than of the wave equatioo. The sum of the zeropoint 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 .
<luctory sur.C)'.
F . Stem, HElcmcntary thwry of the optic'd1 properties of metals," Solid state physics 15, 3OO{196J). II. G. Schuster, cd., Onedi",~,,"r"'al conductaf3, Sprin~t, 1975. N. F. Ma tt, Melalinsuwla! 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.
J. J. J.
J. Appel,
Dev",'CSC, cd .. l'olaroM olld neilON, Plenum, 1984. " Polarons," Sohd ~tlltc physia 21, 193 ( 1968). Singh, HDynlllJlics of exdtolls," Solid .tate physia 38, 2SlS (1U&4).
11
Optical Processes and Excitons
OPTICAL REFLEcrAr;CE KrarnersKronig relations Example: conductn.ity of oollisionless elearon gas
Electronic interband t rnn~ilions
EXCITONS FrenJ.:el excitOlls Alkali halides Molecular crystals Weald)' bound (MollWoo nier) eJ:cihms Exciton condensation inlo eledronhole drops (Ell D)
322
324
326
329 329
PROBLEMS
1. 2. 3. 4.
Causality and the response fU!lc tion Dissipation sum rule Reflection at norma] incidence
COJldUCli\'ily sum rule and supercollducti\'ily
5. Dielectric constant and the semiconductor energy gap 330 6. HagenRubens relation for infrared renecl i~;ty of metals 33 1 7. Dnvytlt.l\' splitting of exciton li nes 331
REFERENCES
332

.,
.
n
b (SWln fJI"'<ft.) SliM _ f<w """""'" p""" obIorpt ..... (antiStolrs)
I.
,
.
n'.,,
intcract~
.' igure I Tncre are m... y 1 of exper!mcn~ io which lighl a crystal . .se. .. n1 procascs arc ,Uusl"'tcd here
The diclect.ic function E(w,K) was introduced in the p.eceding chapter to describe the response of a crystal to an eitctmmagnetic 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 experimental 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 electronic 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 r{w)  ~:C:7:~  1i+iK+l'
(2)
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) =
(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( 
(5)
One quantit y measured in experiments is the reflectance R. defined as the ratio of the rencctcd intensity to the incident intensity:
(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 dielectric function. In this way we can flnd everything from the experimental H(w).
KramenKronig Relations
The KramersKronig relations enable us to flnd the real part of the response 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 superposition 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 (8) %... = a(w) F... , where the applied force field is the real part ofF... exp(iwt) and the tota1 displacement % = % is the real part of % exp( iwt). From the equation of motion, ...
L J
" 0phaJ Pr~ ..a and Ucil.....
309
(
w(;
_~
jj
_~
wj
W I
+ iWPI
(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 KramersHeisenberg form. The relations we develop 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 properties of the response function viC\\Ie(\ as a function of the complex variable w. Any function with the following propertit..>$ will satisfy the KramcrsKronig relations (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 wplane. It suffices that a(w) _ 0 unifonnly
as
IwIo:>.
(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 a(w)===. P
111
I_ ...
0(,) , d
s w
(10)
where P denotes the principal part of the integral, as discllssed in the mathematicalnote that follows . 111e righthand side is to be completed by an integral over the semicircle at infinity in the upper halfplane, but we have seen in (b) that this integral vanishes. We cquate thc real parts of (10) to obtain 0"('1 1 1 a'(w) = ; p _ ... ;::;:; ds.= 11
I
[1u
0"(,) ~ ,h, +
I' _..,
0"(,,) P _ w dp
In the last integral we substitute s for 1) and use properly (el that o"(s) = a"(s); this integral then becomes
i
0"(,) d,
o s+w
,w + ,+w ;
,.,=",.
2.s
310
the result
2 a'(w) = ; P
(l1a)
This is one of the KmmcrsKronig relations. '111C other relation follows on equating the imaginary parts of Eq. (10):
1 a"(w) =>   p
11"
I .'(.)
_,. S 
1   ds:: P
W 11"
whence
. "(w) = 
z:
"
L.. .
(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 KramcrsKronig 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)
8(w)= ;P
(13)
We integrate by parts to obtain a form that gives insight into the l'Ontribulions 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 halfphme, 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 11 w. For the response function (9) the integrand_ 0 as 181 3 ; and for the 8
'&e E. T. \\1'ittaker
~nt.l C . N. Watson,
Modern lInnlyril. C
JJ
Oll/iall I'r'Oa!ue,
,,,,d &cit0nl
31 ,
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   ds_ cr(w) ). If! '" 7Ticr{w) (2) s  w .. lie
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.
as in (10).
III
J+ J J
e p
(3)
 _  d.~ = 7Tia{w)
W
0(,)
(15)
_ ", S
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.
(16)
by the Dirac identity. We confirm that Ihe delta function in (16) satisfies the Kromerso
r  J
 I  ,
md'
(17)
in agreement with (I6). We obtain the electrical conductivity 0(11) 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 a'(w) + io"(w) _ ne ["7TIi(w) + (20) m 11
i..] .
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 absOIptioll 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 spectroscopy 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 parallclthat is, at frequenCies where
(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. Bandband 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 energies just below the ellergy gap, where we might expect the crystal to be tra nsparent. This stnldure is caused by the absorption of 11 photon with the creation of a hound electronhole 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 C n a neutrn1 hydrogen atom, on The bound electronhole 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'lllie), 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$pncs.)
becnuse it is ckctrically neutral. It is similar to positron in Ill , which is [ofl11{.'(1 from an ekdroll and a positron. Excitons can be formed in eyery insulating crystal. When the haml gap is indirect, cxcitons ncar the direct gap may be unstable with respect to d<.,'c ay
into a frcc electron and free hole. All cxcitons arc unstable with respect to the ultimate rt..combination process in which the electron drops into the hole. Excitons can also form complexes, such as a bicxciton from two cxcitons.
\\le have seen that a free cit,.ctron and free hole arc created whenever a
photon of energy greater than the energy gap is absorbed in a crystal. The threshold for this process is fiCu > Ej( in a tlin:ct process. In the intlirt.'CI phononassis ted process of Chapter 8 the threshold is lower by the phonon energy flO . But ill the formation of cxcitons the e nergy i 5 Iowcrt~:I with respect to these thresholds by the billding energy of the exciton , which may be in the range I meV to 1 eV. as in Table 1. Excitons can be formed hy photon absorption at any critical point (22), for ifV",E. = V~E" the group velocities of electron and hole a re equal and the particles may be bound by their coulomb attraction . Transitions leading to the formation of excitollS below the energy gap rue indicated in Figs. 5 and 6 .
31<
Figure 4. An l'oeil"" i$. bound dedronhole pair, un"," ), rreo: 10 togethtr 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 MottWanllicr (,u:;lon: it is .....,:I1J)' bouoKl, "ith an '>'Cr.o.gc ekctronh,,1c disl3r>ee large in com lJRO!oOll "'" a 'Ih lalt;,oe cons tant.
,00\'"
F;~re 4b A tighdy bound os ,,'re.,kcl exciton show" 1ocaIOd 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.
Table I
Si
C. CoA, C,P
14.7
<1 .15
CdS CdS.
RbCl
LiF
AgBr
(1000)
.
ZO.
..,.
30
II . 0
.wo. .wo.
480.
AgCl
11CI 11o,
(cffed ive
II
E:r:Cj,Of1~
31 ...
Vale""" band
(" Oro;,.,
MIa'S
"'!J
~1""~I:"Il. Ec
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 ~ .
.
Figme (I Energy levelsof"n .. xcilon erc.!!cd ill" d~1 prooC5S. Oplirnilrallsilion. from the topuf
the valence band arc shown by Ih.., arrows; longest arrow COITCsp"",ls to the cn('~ gap. n.,., binding energy oftl,e exciton is E#" rcfcrrc<! to aff'C<) ckctron and f,,* hole. The Iowe.t f'<'fluency absorption line oltll.. (:ryslal al aboolute zero is not E~, but is E~  E~_
tnc
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 e1ectnmhoie recombination line with the energy of the exciton recombination line. By photoioni7~\tion of excitons, to form frcc carriers. This experiment requires a high c:oncentration of excitons.
'"
" " 5 ," "
0
~[,.cil"" .....
\
r.," .
~
~
0.11 
\..,..,"\ "1""..,11",,,
II'.
,..~.'"
...,......
' ,, "
~
07 
", , ,
Figure i
" ''''
<'''''~ near
."
."
, , , , , , II
, ."
._ , ~
,
In
~\
I'Ik>Iuo,
'"
I Sl
, , , ."
, ,
Effect of an emtoll kl\'c1 on II,.. uph ...,,1 alll>OfllllOn of a SoI"micondurtor Ii,.. ~<MlS of the 1)3,><1 JP 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' tkduo;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
00031 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 electronhole 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
'"
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
!}{.pk""
1/1,,:
(27)
(28)
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 lowcstcnct'b'Y cxdtons are 1 0cali7.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 isoelectronic with the Br iOIl ofKBr. The spl itting is causoo by the spinorbit intcraction. 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.
(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!
l>e;1n.. I,It;ghlx:..t ..
..,srn interactioll
 vi..
'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
(ees)
Here
II
MC' En = Eg  'll,2en2
(32)
=+l'
1 me
1 m,,'
(33)
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 transitions 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.
'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 eh 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 properties: 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<rtronhdc 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 electronhole 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 ctronOOIc re<.~!ml>im,ti()n lum,ncSCUlce onto the surfact: of an j"rrnrcd sem,lj,'e videe"" im~g" (After J. r. Wolfe ct at)
'"be.
'l22
Table 2
Cr)slal
(u,,~t""'sM:I)
Critical
temperature, K
Go
Si
18 93
17.5 17.
2.57 x 1017
3.5 x IO U\ 8.6 x 10 16
CaP
3CSiC
10. x 10"1
IMK.IoI I BOrn~lc"',
41
Vol. 111117, Springer, 1984.
The exciton phase diagram for silkxln is plotted ill the temperaturecon(;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 excilons 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 xrays and it is similar to the inelastic scattering of ncu trons by a crystal. The selection rules for the firstoruer 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 K refer to the phonon created or destroyed in the scattering event. In the sccolldorder Ibman effect. two phollons arc involved in the inelastic scattering of the photon. The Raman effect is made possible by the straindcpendenee of the eI{'(.~ tronic polari7.ability. 10 show this, we suppose that the polari:wbility a :lSsociated with a phonon mode may be written as a power series in tbe phonon amplitude u:
n.
(35)
If u(t) = Ito cos Ot and the incident electric field is E(t) induccd electric dipole moment has a component
= Eo COl> wt.
a.Eolio cos
wI
.
wr,r,
I I I
I I
0
, I" , ,
,
I
0 0
I
\,,~,
<:nk.. _ _ II a
" " ,m
~bows.
,,
,,..
10"
."",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 ror cump\c. thai with an a,c~ COOC't.~ltmtlon ncar 1(l'7 (,,, 3 "I l!i K. a fl'l'ecU.,lon /(3.'l "ilh salor~tedg;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 lire rnctalinll..lator Iransiti"" "'" cu:itons arc ~ho (From J. P Wolfe.)
"r
snon.
.,
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 antiStokes 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(
or phonon nMXle K.
'"
The antiStokes line involves phonon annihilation, proportionaJ to
(
\\,HII
a photon intensity
(38)
+0) I{w  0 )
(39)
with (nK) given by the Planck distribution function lI[cxp(fIIlIk R I)  1]. We see tllat the relath'C intensity of the antiStokes 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 obscrvc the mode by firstorder Raman scattering of light. The secondorder Raman cIfed arises from the term 02112 in the polari7.3bility. 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 complicated if there are several atoms in the primitive cdl because of the colTesponding number of optical phonon modes. Secondorder Haman spectra have been observed and analyzed in numerous crystals. Measurements on gallium phosphide (CaP) are shown in Fig. 17.
" i
" '
!
!lam""
",,,rt.
('I'll
figure 15 Fintonlcr Raman spcr.I oi"lhe K  Oopti 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~
lines u
Figure 16 Int ensi ty nlt;o anllStoke< to Stokes II. fundian oftcn.pcnllurc. for the oIm,.vat ion! or Fig. 15o.. the optir.""i mode of ,ili<.vn.
dlif' , it depends slighll yon the temperature. (AftcrT R. lI art, R L. Aggarwal, and B. i....ax. )
'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 ).
""
,.
J
,
600
 700
6JO
 SIXl
 ~ oo
300
 200
 100
Figure 17 Raman spectrum clCaP at 20 K The two h'~1 peak. are the lintorder fullna" I".." associRlcd with the ('sellalion of an to phonon at 401 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 excitation 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.
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 penetroltes a solid. Consider this diffcrent'C. Tile gencrnl result from ck'CIromagnetic theory for the pov.'cr dissipation densily by diclectric losses is
(ees)
(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
(41)
E'
:=  '
R1T
WE"(w)J:.2
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 displacement is
(eGS)
D(r,f) = grad
Ir
,
vt
(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
3:=:
"
~ ~ <
~
j
Figure 18 Vnlence hand
.,l~roo
."
1 0
o,
8,,,,,,"1 .... "'W'. cV
(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 '
(44) @I(w)=    lm{UE(w)}ln(kot;Jw) , 1T where liko is the maximum possible momentum transfer from the primal)' particle 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.
2 C
'w
32S
, , ,
,
",
,
~
c.
, , ,
, I
i ,
'"
,
,
,
'0
"
.~u
" i
,,?',..~
"''.
.,'"
., "
30
'0
.
"
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
ELS
E"~ 1
lAP INS
IH~I
ISS lTD
LEED RIIEED SA}.I SEM SES
SI~ I S
Auger electron spedroscopy Electrou (energy) loss spectroscopy Evaporation rate monitoring Electro'l spectroscopy for "llell'i~1 :h1alysis EAtended )I.ray absorption fi ne structure Flash (iesorption mass spectrometry Field iooization microscopy JmaS:i llg atom probe Inelastic neutron scattering Ion rate moni toring Ion scattering spectrometry Isothermal Oewrption spectrometry Low encrb'Y electron diffraction Renection high eneTb'Y electron uiffract iOlI Scanning Auger microscopy Scanning el~tron microscopy Seconoor;' elect ron spectroscopy Secondary ion maliS spcch ollleiry SpinpolarizeJ 10\11 energy electlOll tliff.a<.1ion 11lcrmal (\esorption mass spectrometry TnulSIIlis)ion electron microscopy Transmission high energy electron tliffrnctioll UltrJ";olct photoelectron siX!droscopy Xmy photoele(.1l'On )peclroscopy
SUMMARY
.. 111C KrnmcrsKronig relations conned the real and imaginary parIs of a respomc 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,. and t"(w) = 2nK. "l. 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 KramersKronig relations are l'Q nsistcnt wi th the principle that an cffe<:t not precede its ca use. Consider a deltafunction force applied at time I = 0 :
I F(I) = 6(1) = 2. _ e lwI 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 halfplanf'. (b) Evaluate x(I) for t > 0, Note that a(w) has poles at .:t:(~  tp')lIt  lip. both in the Iowl'I" halfplane.
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  ;:
80"(8) ds
>10
Con~i(ler 'Ill
d ...'CtromJgnetic wave in
WICUUnI,
with
Let the \va,'c be inciUent UI)oII a medium of ,1.e1ectric constant f: anll permeability IJ.  1 tl!.lt fllls the Iwlfsp.~ x> O. Show tbat tile rl,nectivity coemdcnt r{w) as deBnt:d by (rell) = r(w)";(juc) Is givclJ by
II
+ il( 
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 relation tklt
But s t rn.:quencieli 0
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 ilitntiC31 for the superconducting and lIorrnal stlltes, then we must have
v.l.wl dw
vJwl dw
<w<
of the conductivity of a superconductor \"alli~hes, so that in this regiOLl the illiegnli on the leftIlllnd 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
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
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 H eidi. \ 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 ternperature, 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 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. Ili nt: If (70 W, then 112 "" ~ . 111is )impliHcs the algebl';l,
w"
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.
""
References
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., ElectronIwle dro/lldr In UmfC:oIlifuc/Orl, NorthlIol13nd,
J. C. Phillips, ".'undame"lal optical s~ro Of 50Ild$,' Solid slnt., ph ysics III, 56 (1 966).
R. Loudon . QUlin/lim theolY of light , 2nd ed., o~rord, 1983. E. I. Rashb,a and M. D. Sturgt:. eds., E~'dlornr. North _llo1Iand. 1982.
D. C. Reynolds and T. C . CoII;n., xcilOru, 1/1d. propU1/u and ...su, Academic, 1981 . K. 0 '0, EXcilom, Springer, 1979. W. M. Yen and P. M. Selzer, fils., Laser rpectro&c:'l>W of sol/dr. Springer, 1966; ''01. II . 1989. Y. R. 5]"".), Principia of nonli'l{lr opla, Wile)". ]984. Jo'. C. Brown, ~U1tra"i<:>le1 sl'ectroK'Op)' of solid>; with lise of synd, rotron radiation," Selid slale ph)'5ics 29. 1 (19'74). P. O. f'iluon. ~Optical properties or n>dali and .I!or$, ~ Solid 11:IIr. phYIic'S 29. 139 (19'74). T. M. Rice. ~E1eclronho1e 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 Tllcnnocl),namics of the ~upen:Ollducting t l'1lnsirion London equulioll Coherence lengd! BCS thoory of supercollducth'ily BCS ground stale Flu.l quantization in a superconducting ring l>..,,"lIlioo of persistent currclllS Trpc II superconductors Vortel stale l<:slimation of Il .. and Ilc'l Single particle tunneling Josephson superconductor tunneling Dc Josephson effect Ac Josephson effect Macroscopic: quantum interrerence
HIGH ..TEMPl::RATURE SUPERCONDUCTORS Criticul fields lind critical currents Hall nwnber Fullercncs
SUMMAR\,
371 372
373 373
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
I. ~I agncl ic field pcnclratioo in Ii plate 2. Critical field of ,"in nlms 3. Twonuid mood of Ii superconductor 4. Structure of a ,'ortex 5. L(lndUfJ penetration depth 6. Diffraction "ffe<:t of J osepliw n junction 7. Meiss ncr effed in Sllllcrtl
375
37' 37'
376 376
317
377 377 377
REt"EREl"CES
APP ..: NDlCES Rf<: LEVANT TO SUPEUCONDUCnVrn' II Cooper I'airs 656 I Cinwurgl.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], K~me.lingh On .... . mark..d tloe dl5cO'~ry c:l $ulM'r<'"oudUCI i,ily. .

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 mperatu re in the liquid helium range. This phenomenOlI, culled superrond ucthity, 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<.."omlucthlg 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 ekclrical 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 1 have rallcn below 3 x 10 8 ohms, tt,at is to One tenm,lIionth of the .alue 0 which ,t wouW have at O"C. As tI,e temperliure 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).
3:35
B,
0.026
Table 1
~
B
C 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 " B. T. Matthias, revised hy T, Gebllll('.
_lr
5i
N,
,
M.
Transition temperature in K Critical magnetic field at absolute zero in gauss (10 'tesla)
AI
l.f40 i05
P'
C,
5,
Ti
0.39 100
V
5.38 1420
C,'
M,
F,
Mo
Co
Ni
C,
Z,
0.875 53
G,
G,'
As'
 .    ~
N,
5'
CI
A,
5,'
B,
T,
W
 R, Rh
_ 1.09r
51
Rb
5,
V'
Z,
0.546
47
K,
Nb
9.SO 1980
"
C,'
B,'
~.~
.0003
.049
la
Icc
HI
0.12
_.,
6~OO 1100
  
F,
R,
A,
"
C,'
"
Th
   . '                 , To
 Pt
Pd
A.
Cd
0.""30
I,
293
(~)
340> 'rt22
TI Pb
R,
0,
65
I,
A,
   1
~
T,'
X,
Hg
Bit
Po
At
R,
4.483
830
0.012 1.07
1.4 198
0.655
0.14
19
...:"". eo
4.15J
412
21]' J.193
171
803
p,
Nd
Pm
Sm
E,
Gd
Tb
D,
Ho
E,
Tm
Vb
L,
,Fm
~
;.C
p,
Ut (a)
  Np
~
p,
Am
Cm

,
_ . .
 Bk
Cf
E,
Md
No
L,
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.
field will act as a perfect diamaJ,(net, with zero magnetic induction in the interior. 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 temperature, 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 superoondueling state.
Occurrence of SII1JcrcorlJllctivify
Superconductivity occurs in many metallic elements of the periodic system 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 fband 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
n Schne ffer, Phys 11<...,. 106, 162 (1957), lOS. 1175 (ISSi).
will every nonmagnctic metaHic c lement become a superconductor at sufficiently low temperatures? \\le do not know. In experimental searches for superconductors with ultralow transition temperatures it is important to eliminate from the specimen even trJ,(,'C quantities of foreign parnmagnetic elements, 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.
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
Compollnd
in K
T"
Compoutld
~K
T"
Nb,Sn
N~Ge
18.05
23.2
17.5
Nb,/Il NbN
16.5 17.1
90.
31.3 lOA
K,c;,O
16.0 19.2
HhtC'40
~Jl
",
I.
, ..1 + '''<:+\
,
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 ;~
(ees)
(Sl)
B=B.+4nM::O;
H = I1d
0.
M   = Bo
4w
(I)
M I = =  .<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
+ 1J.oI\1 ::::
0 ;
0.
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....!
Ruprl lC
.) fit'.'
I,
II.~
B.
}o'igure.. (<II) MliJ:ndil',alion 'erMIi IIpplird n~g;IIc1ic Ilekl blJlk sup'.'l"conductor e~hiLilj,lfC" complete fIdsoucr 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 1 small 10 he seen On th is..,..1o: Note tM! minus "11M is plotted 00 the ,..,rtb! 00 !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 . C5pcd, cJtCqlt fo, pmsi,,Ic ~.~ cIT....,b. "or g"Tn II, the area unde r the tna~l1ct i~li<m curve is the $lime fur II t)'lle II sUJ>l! rronductor as for II type I (CCS unil s In all po.m
r....
'''TfY
TIle magnetization curve expeded for a )upercondudor under the conditions of the MeissnerOchsenfcld expt'rimcnt j ) ske tched in Fig. 4a. TIlis applies quantitatively to a spL'(:imcn in the form of a long solid cylinder plrlCed in a longitudinal magncli<' ficld . Pure specimens of many materials exhioit this behavior; they arc called type I supercond\IClors or, forme rly, soft superconductor5. The vruues of fi r arc alw<I)'s too low filr type 1 sllpcn.'Onductors to have any useful technical applkatioll in coi ls for super<:onduding magnets. Other materials CAhiLlt a Illab'lletization cun e of the fOl1n of Fig. 40 and arc known .IS type 11 superconductors. 'nleY tend to be alloys (as in Fig. Sa) or transition metals with high values of the clC(."hicai resistl\'i!)' in the normal state : thai is, the electronic mean free path in the normal state is sllOr!. We shall sec later wll}' the mean free path is i]H olved in the "magncfi7.ation" of )Ilpcrcond.uctors. Type II superumductors have sllpcrt:ond ucting ele<:'i]'iw.i properties up 10 a fi eld denoted by Ncz. Between the lower lTitical field Hd and the upper critical field lid the flux density n "" 0 and the Meissner elfcd is said to be incomplete, The value of Ild may be 100 times or more hi~er (Fig, 5h) than the value of the critical field IIr eilculated from the thermodynamics of the transition . In the region hetween IIc ] and Hc'l the superconductor is threaded by flux iinestl.nd is said to be in the \orto state, A field Hc'lof 410 kC (41 teslas) has been atl.lincd ill an allo)' of Nb, AI, and Ge ilt the boiling point of Ilelium, and 540 kG (54 teslas) has been reported for I'L"to(;ss>
,
, ,.,
In I:>U
Supcrwndtlding magnctiulion a,oes of annealed poI)'CI}'$.t.llline lead and 1".1<1. indium alloys at " 2 Ie (A ) lead, (8) lead 2.08 "1. pcrttnl IndIum; {C} 1cad8.23 wt. pt:tce'n' indium; (D) kad20.4 wt. pm:t'nl Indium. (Aflu U\'ingnQn ) Figure 5a
'" "

.... ,
....
NbT;
,
I;ignrte 5b
"
TC"'1>",~ru.c.
"
"
III
"
"
poactil.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'ulws. )' and.so porUons oC the cuo't.'S ,11UYm as !wok.." Ii~. were ~u red in dwI way )
""
"
. , "
l
E
"
10
05 
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 supcrlnducting 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 hedt capacity in the superconducting state is an cJCpOnl_ .al fOflll with an argo
 Ol
',
'.
.......,
;:;  0.5
.5  06
 0,7  08
 ,"
 10
 11
 12
5'
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 temperat 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 rlllru;uret! ,,, "ero mllgnctic lidd, and F", is me.s"r..~ t ,n a nl,'gneli<' fidd s"ffic1enllo pu l
0 8.200 (; B. _ O
"""'.
,
;
u
'1 .,.
"
<,,<>".
/.
$~""
"'I t " I I
I I
,.
u
Co1'~
"
~"".
" "
I"
" "
t t
~ ,
:
crr .. 0596.0~T"
O .OJ
i "
/
'," Ib)
"
n.e
10
~. K'
0.00 1
"
, , ,
, ,
I.
Figure 8 (al 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 nuclcar 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: $lak: 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 BardeenCooperSchlicfTcr (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 electronlattice inlcractio:l, Chapler 7 . This intcmction tics the clcctrous to the lattice. In a supcn:.'Onductor the important illtcmclion is the electronelectron 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"CQndoruer phase transition. At a sccondorder 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 firstordcr tmnsiti(lll w(luld be characterized hy a latent heat and by a dismntinllity in the energy gap.
AI
Si
10 4eV.
,.,
'.4
S<
Ti
V
15. 3.4
C,
M"
f,
Co
N'
Cu
Z"
2.4 ' .2
G.
G,
'3 ,
I"
.
So (..)
Nb
Mo
2.7 '4
To
Ru
Rh
Pd
Ag
Cd
I.'
' .2
10.5 ' .5
,.,
Pb
11.5
HI
fa
14. ' .60
Re
0,
I,
PI
Au
Hg {_j
16.5 4."
TI
7.35 3.57
' .7
27.3 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 ... Ekeh 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
"" ...,
T o Taobl.un
o NIOI .....,n
, .'1,
BCS
<'III",. ~
~.
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
....".
Microwave arid l1ifrarccll' roperties
,,
0,
" "
Trr.
., .. " o ,., , ,
' n )C existence of an e nergy gllp mCllns that photons of ellerg)' less than the gap cnerb arc nol absorbed . Nearly all tile photons incide nt are reflected as for 'Y 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 superconductor vanishes at absolute zero. AI T <c '1 .. the resistance in the supcrconducting 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
or
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 shortcircuit any normal electrons that have beel! thennally excited above the gap. At finite fn..'tluellcil.>S the incrtill of the supcrconducting electrons p revents them from completely scrCl.'fli ng the ek'Ctric field , so that thermally 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'CIrontatlicc intcractions arc dl.'Cply il1\'olvcd in superconductivity. 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 = jin (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.'rconduetivity has been reached in SC\'eral ways. Certain results follow directly from thermodynamics. Many important results can be described by phenomenological eq\lations: the London ''(Iuations and the LandauCinzburg equation s (Appendix 1). A successful quantum theory of supcrconducth'ity was given by Hard 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)Crconducting wavcfuTlctioll.
Table 4
Eq>crimcntl.1 ''aluel or Q in M"Tc = CQI15ta nt, where !If is the isotopic mass.
Sulnlanc.:
0.1.5 ~ 0.32 ~ 0.47 ~ 0.50 ~ 0.49 :!: 0.05 0.07 0.02 0.03 0.02
Sub..!anc:e
0.00 ~ 0.15 ~ 0.33 0.08 :!: O.OO :!: O.OS 0.05 0.02 0.05
Zn
Cd So H/o(
N~n
Pb
]'lumlIOtlYlwmicll of the Supercolltiliciing Truwritioll l11e tmnsition bctweelJ the norm,ll and supercollducting states is thermodYllamkally 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 transition, 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 liqUidgas coexistellce 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 superconducting 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"ements. III the l'alorimctric method the heat capacity is mt."'3Sun.. as a function "(l of temperature for the superconductor and for the normal conductor, \\hich rneans the superconductor in a magnetic field larger than He. From the diO'erencc 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.~
.)
Ia) A sUPf'lOI'Icluctor in ....hid, Ih~ Meissner dftct is complete 1_ B  0, as if the  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 ..,).
Figure Il
rnagneti~tion wcrt /of ..
per unit volume of specimen. This work appears in the cnCl"b'Y of the magnetic field. 111c thermodynamic identity for the process is
(4)
(5)
(S I)
""
B" d8~
(51)
011
being brought from a position where the appli(.,(\ field is zero to a position where the applied field is B".
J2
SUfII'rconduct;" i/y
34!
Now collsi
tibili t ~ of a metal in the normal state, then M = 0 and the energy of the nonnal
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!rconducting stati.S; '
(GGS)
(8)
(SI)
In SI units li e  B"dlJih 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)
(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~
'
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. Ekctrical 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 directly proportional to the vcctor potential A of thc local magnctic ficld, where B "" curl A. The gauge of A will bc spL'Cified. In CGS units we write the constallt 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
(SI) j
 A
I!tv,f.
(10)
nlis is the Lonoon etluation. \ Vc c'\;press it anothcr way by taking the eml of both sides to obtain (CGS) curl j ""
~ B
471"A,_
(51 )
curl j =
(II)
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."
'.
n.e
I'gure 13 PMW:t .... tilll\ vi an applicd mago'''''k' flt"kllllto,. ~miinr;nile superconductor. pen .. ,,,,lion depth ,I. IS d .. r.n~c1 <IS the htmlCt: in which the 6 ..1<1 dn: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) appliLs 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\hl'\;wcll equation we know that 4n (CCS) curl B : : :  j
l ead~
(12)
(CGS)
'r cli rl ,
(51)
which may be combilJed with the London I;..'q uation (II) to give for a supe rcofl(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) unless 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 iinfinite s\lpert:onductor 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
(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 penetratio 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\'crsimplifkd. It is shown by comparison of (22) with (II) that
SI)
A,.
a::
({)mdllJl(r)1 12
II .
(14a)
in concentration
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 characterizes a superconductor. An independent length is the coherence le ngth {. 11lC coherence length is a measure of the d istance within which the sllperconducting electrOIl concentration cannot change drastically ill a spatiallyvarying magne 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 LandauGinzburg 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 wavefunction; (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 ',jt'pth A,
in
,,""'~
Metal in 1O 6 cm
It'n!!th 61.
1O 6 an
'df.
0. 16 0.010 0.45 0. 14 1. fr2
S..
AI l'b
Cd
23.
,GO.
8.3 76 3.8
Ant',
Nb
n.
~Ics('rv('y
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")
+ e 1q ,) = I + cos qx
(l 5b)
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
(Ix
ip*
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
  k"{lo ::: Et!. .
2m
.'
(100)
rclate<\ to the Clitical lTl{Xlula(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)
CnlculatL"(1 \'alucs {o from (17) arc given in Table 5. The in trinsic m herence length {o is characteristic of a pure superconductor.
or
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 LnndauCinzhmg 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(.
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 electronlatticeelectron intcl'U(.1.ion leads to an c nerb gap of the 'Y observed magnitude. TIIC indirect interaction prOCCL'ds when one elcctron inte.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 encrb'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 magnetic fields that vary slowl)' in space. 'JlIUS the central phenomenon in superconductiVIty, the Meissncr effect, is obtained in a natural way.
1J. ~n , L N. Coo~r, and
J. It &hntffn,
,.
12
S"~rc(lnd"cli.:;t!l
$~
....gure 14 Pe'lClntiun depth A ~nd Ihl" cnhenmr.'C len!,'!!! f ,\I r""chOl'$ of tire ITKl:m r~ patl. l ..rlhe conouctlOlll"la1rons in Ihe "".. lUll 51.11.,. A ll I,nglhs are til "nils of ". thl" Inlrimlo.: coh~ l"nee I",,~th. Thl" l'Urvt'S arc sklch.:d for :,  lOA,. For dlOtl meml free I"'tm II", cohc:renct: len~1I1 1 X"..'(:Imes ~hortl"r antl the
pclll1r.l1ion
..,
0.3 0.2
0.1 0 0
O ;0
(\t:pth
bc."COfI>r.1i
Io"';l"r. 11.c irrc ...~ in the "diu d/f fa,...,1'$ Iype II sup.,'orn!"cti\'il)'.
,
,
4. The criterion for Illc transition temperature uf an c1emt.'Ilt or alloy involves tile c1t.'c tron density of orbitals D(E~.) of one spin at the Fermi level and the electronlattice 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'OtronpluJIlon 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<.'Ontluctor when cooled. 5. Maglletic flu){ through a supcrconducting ring is quantized and the effective 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.
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 manyclcctron system is described in terms of the occupancy of oneparticle orbitaJs, those near EF arc filled somewhat like a FermiDi,?c distribution for some fi nite temperature. 'Ine central feature of the BeS state is that the oneparticle 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 Appendix JI. They have spin zero and have many attributes of basons. 'Nle boson condensation temperature (TP, Chapte r i) calculated for metallic 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.
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 concentration n = ",*!JI = constant. AI absolute zero II is onehalf 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)
n<l( ,.ve  A q)
no
(2 1)
We may take the curl of both sidl..'S to obtain the london equation:
(CGS)
curl j =    8 ,
nq'
me
(22)
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 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.
."rf.=
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
VO' dl :: ~  8, c 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)
(25)
where da is an clement of area on a surface boulll.kd 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 quantum of flu x in a supe rconductor is (CCS)
(5 1)
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/,!ZX21U2 ) = 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.
("")
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
(431><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
SI
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
ccordingly, the in the age of the universe, under our assumed conditionl 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 superconductornonnal 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 uctivity 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 supcrconducting. 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. (lCrtcnt 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 stnctUTe 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
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 magnetic 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 surrounded 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 superconduding 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 superconducling regions ill the vortex statc. The vortex state is stahle when the pcnetra
,~
'.
,
~"
'"  ,
,
,.,
""
,
o
Fi~re
g21l panlrnctcr 6(:\") at the intcrf~ of supcr<.'()nductin!; .mll normal regioo$, for type I nnd t )'J)C II supereonduc:fon. The energy gap panuneter is a mcll.'iurc of the stabilization energy densit)' of the $UPCfUIn
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 superconducting 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)
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 superconductor we can <letennine He only indirectly by calori metric measurement of
the stabilization energy. To estimate He. in terms of H.:, ( _ 'Onsider the stability 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 ndat')' 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)
,
Jrore
_  1 Ill87T
X 7T/:2
."
(32)
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)
For a single fluxoid we add these hw contributions to obtain
(33)
(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)
(37b)
Figu~
~ an
insulator C.
AVAl~S.,
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 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 cumI'll mca.m rcment amI twu 10.the \dt:Jge mCl.5U""llCIll . (Aflcr Gioc'\"1:r and Mcgcrle.)
m,n
\'ul~.
" ".)
ng,,", 22 {a} Linear cumnl ...."ltdgc relation ro. junction of f\Of',,~...1 metals ~pnr.. tcd b) QIIide larer; (hl cunenlmltagc relation "ilh one metal oormal and the other metal $uJ"It'n'ond"ctillg.
insulating layer is simpl)' a thin oxide layer formed on one of two evaporated ITlCtal films , as in Fig. 21. When both metals are normal concluctors, the current\'ohage relation of the sandwich O tunncling junction is ohmic at low voltagcs, with the curfent f directly proportional to the applicd voltage. Giacvcr (1960) diS(.'O\'cn.'d that if one of the metals becomes supcrconducting the currentvoltage characteristic changes from thc straight line of Fig. 22a to the curve shown in Fig. 22b. Fib'Ure 2.3a contrasts the ck"Ctron density of orbitals in the superconductor with that in the normal metal. 1n the superconduclor there is an energy gap eenten..od at the Fermi le\'c1. At absolute zero no current can Oow until the applil.>d voltage is V = f~J2e = Me.
I ', .. ,. ",
.',..,,,,
2lo,       
"""''IlY
Figul"(' 23 The Ill""'''' of "rhit.ds and the 111rr"nlvolt:lJ:." d~"'al.11'mlie /Ora tunnclinlti,,"dion. In (iI) the "ne'J.!Y,5 plott .....! "" th.., ....,rlieal M:~lc ~Ild th.., d"'o,ilyof orhit.tls ",.the horiwutal <I.;/.Ie. 0".., 1I..,!allS III Ih" nom",l.!.II.., aTld.,.", in Ih" 'UI'lCn..'OII<!ud '''1t ~Ialc (b) I 'enu, I', dl<'d;)~hcs iud'mt.., the ex,><'eIL..! bn",k al T '" O. (An". Gwe"cr and ~1L,.".1c. J
,.
"'I
11le gap E~ ..orresponds to Ihe hreakup!lr a pair or clctlrons in the superc()lIllucting state, with the rormaliolt of two cicci roilS, o r an cJt.'Cl rou ;Illd a hole, in the normal s ta te. 1'lc current shu'ls when eV"" fl.. At finite temperatures Ihere is a small curren t flow cven .11 low voltages, i.lcctuse o r clc.!ctrons ill the supen:onductor Ih.1t arc thermally excilcll 3CI"OSS the energy gap.
Dc JO$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, kt I/Jt be tIle amplilude Oil the other s ide. For simplK::ity, lei both SlIllCn.:onductors be itlc nli<.al. For the prescnt \\e suppose that they arc both a t zero potl.'11tial. "l11C limedelx::nc.!t..nt SdtriXlinger l.'fIIJdtion i"iI~;!l ::::Hif! applied to the 1\\'0 amplitudes gives
aJw
.~.~
(38)
Ilere /,.,. rcprcs~n t s the dft!~t of the cleclronp.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 ::
TIICn
'.p,
at
ilt
I 11 \
ilt
(39)
(40)
iI!J!l all z ' ,I, 302 ,.   = I 112 112 erD.   + ,., ::::;  ,'I '..... ,
at
ill
We multiply (39) Ly
,11 '2 c fh
2 ill
8..!  0"
12
lilli, +
ao,
(41)
at
at
 ;1 (111110 112 e
,/I
(. 2)
Now C(lualc the real and imaginary l>arts of (<I I) and silll illlrly of (12):
(.13)
(11)
If II ,
03 1/2
I1t
at
ilII I 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: :
whe re
1n ..ill li :: 1n sin
(01
O.l ,:]
(.17)
l1.Jrrc nt
10 is
T. 'Ibc
10 is
the
Cu,..,,'
jl';: \"uI.~.
' 1,,( ,, ,, , ,, ,
V.
, ,, , ,,
1".,,", 24
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 junction 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 zerovoltage (.'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 O  01 , lbis 2 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)
at
iJl
icVntf~  ' 
iT(lllnti l rI
n.
15
(49)
(5 1)
wllieh d iffers from (4'1) by the term eVIIt. Further, hy cxtcnsioll of (42),
(J1l2 il02 I   + iI12 =
ill
ill
 i
CVI12f1  1 
iT('llll2.l
1/2
c
I~
(52)
tigure is The arrangement fOl" eaperimelll on macrO$ropic quantu m intcrfcrence. A magne~ flu" <l> pa55CS through the interior of the loop.
whcm:c
/j ;
''OS {j .
(53) (54)
we have
 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.~
(56)
('leVtIt.) )
I
(57)
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 \oltage 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
dft"ts ror h.O ;,,,,chon. A B",1 B. TI", field rcriollilv " 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
"
~.ln
(00)
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 :
c<l>l flc =
S1T ,
= integer
(61)
'n.c periodi(;it}' of the current is shown ill Fi~. 26. The ~hort l)Criod variation 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/Jiaclion eflCct w.tS not antidpJ.ted in carly work Ih.ll rcsoh ed only single elcctroll tunncling.
'0
12
S"po:~colld"cti";ly
HlGIIU: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 longstanding 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 Josephson 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 L.1.. b5Bao 15CU04 YDa2Cu307 1lzBazCa zCu:)0 .0
Tc = 12 K Tc = 36 K Tc = 90 K
Tc= 120K
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 oxygendeft..>(.'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 canier concentration in 110 = 5.77 X 1O~!.l cm  3. '11\C excess carriers typically arc holes embedded in doped ulltiferromHgnetic insulators. Parallel shl.."ets ofCIlOz arc a structural feHlure of all HTS with Tc > 40 K; these are known as the cuprate IITS. 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 dominant l..'ollducting pathways. 'l1lC oxygell ltmlent per cell can be changed rcversihly 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 110nmagnctk; tile crystal is supermnducting al)()\ e Ob 1).1 . A dmrgClleutral formula for YBa ZCu:)01 can be written ll.'i YBad C u2tMCu3+) (0 2 ), or as YBaiCuz+h(Oz M O  ).
(?~hV03
(2 ) 0
?
/' b
~~Ol &
1 w'~)~C"I_"
Ba
().I
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 arc 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)' approximation to n(."gk'd the anislrop)' of the normal or supcn:onducling propcrtit..'s.
T.:5
Critical Fieltls
Utlrl
Cri,ict,l Cu,.,.rnts
Critit'3l field and cUlTcnl vaillt.'5 in IITS 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 materials. 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 peoctration 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 GiozhurgLand'lu paramcter 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 quasi2D 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.
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 density 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'Cllhution PCI' f(JI'mula volume. In yeCa the Iiall number is pIype and prop0l1ionaito temperature for Hlle, for 'f somewhat above Tc> and is ortllt' ordcr of unity at 300 K, giving a Ulmer (''Olleentration '" 7 x H cm  3 . I' Fullerene, 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
,,.
~tnlCtur~.
CourtCII)' tI
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 facccclltcrl."(1 cubic structure, as in Figure 28. Alkalifullcrene (: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 bulk spt..'d mCll of metal in the supcrt'Omlucting state exhibits perfect diamagllctism. 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 <'Tilkal 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. 1rat(."S supcl'(;on In the superconduding state an energy gap, E~ "" 4kB1 ~ sep.. 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
, =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 (;ollcrellcc 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.
we ~ < A. Thc lTitkill ficlds are rclatctl b)' Hel .... Type II supefCOndudors h.. (lA)H" and lId "" (AJ{)HC' 111e CinzburgL.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 twofluid 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 ternpconturc 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 dispersion 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 shortcircuits 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
nu~
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)
.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 charge (/ and mass m , show that '\1 rntJI4'11"nr(.
or
cf rorumguiar CTO$5 St.'Ction "itl1 a magnetic Held B applitd 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
1 = 10
sin(wTBellic ) (wTBcl1ic)
References
J.
R. D . Parks, cd., SupermnducIlL"'Y. Deld..... 1969. Good collection of revir.:w artic!cs. . 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. Acadcmi<'. 1979. M. N. Wihoo, S''Ile roonc/uctlng magnel OxIOro, 1963. D. SaintJames. E. D. ""lOm~. alld C Sanna. Type 11 , "pnronduditul/. Pergamon. 1969. R. P. lleubc"er , Ma{ywlicjlwt III $Upc<rcorll/uc/ors. Sprll1~,'e r, 1979.
""'C/U I""U
HIGII TEMI'ERATURE SUPERCONDUCrOnS C . Bums, lfigl,.,n"lIf't""6/lIr't fI"'Jl<!ramducth..'UII. Acadcmk. 1991 . M. 'I1nkham and C. J. l.obh. Mrh~'sical p~rti<.'S of the new supcn..." 'dllctors: Solid state ph~'sia 42, 90 (1989). R. t'. Curl and R. E. Smalley. " t'Ullercll~," Sci. Am 4. p. 54 (t99I ). D. R. lI uffmali . "Solid Cc.o. Physics Too.)" No. II . 22 (1991). Summary or phys;c.J propert ic~ or thl. bud:m;n$lOlrful1erc"<..".
"'0.
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$ephron efftcl.'. ond S"IJe rc:onducling electronics. Uilge" 1991. J Clark... and R. /I . Koch. "Impact ofhigh tc mpcrnlurc ~upcrwnductl\'ily on SQUII) mllgne tomc krs, Seit,ncc 242, 217 (1 968).
M
(
(
13
Dielectrics and F erroelectrics
_
Maxwe ll etluations
J>olariz.ation
MACROSCOI'IC ELECTRIC FIEI.D Oel)Olarization field , El LOCAl . ELECfRIC FIELD AT AN ATOM Lorentz field, E,; Field of dipoles inside ca,ity. ~ DlEI.ECTRIC CONSTANT AND l'OLARJZABlLITY E lectronic pobrizability Sfn UCTURAL I' BASE TRANSITIONS FERROELECfRIC CHYSTAI.S Classification of ft:TTOcil.."Ctric cr)'stals DISPLAC fVE TRANSITIONS Sort optical phooom Landau theory of the phase transition Secondorder transition F irstorder transition Anli fenociedricity Ferroeltric domains I' >eweleclricity Ferroelasticity Optical ceramics
SUMMAI\Y
""
381
38'
38<
388
390
393
,..
393
...
.. 0
)'nOBLEMS
I. Atomic hydrogen 2. Conducting sphere 3. Erred of air gap 4. Interfacial polarization 5. I'olarization of spilere 6. Ferroelectric criterion 7. Saturation polarization 8. Dielectric constant below transition 9. Soft modes 10. Ferroelectric linea r arra y
NOTATION, (ee5)
(51)
~ _ 107/4wC ;
D  r;
+ 477f' ,.. ef 
(1
+ 4'iTx)f
= III flccol ;
(l ""
pI f lon] ;
41TXC,S 
XSI ;
Figu.... I The pcrmalle.,1 dipole momen t Q( II moIocule fA ""~ter has the magnitude 1.9)( 10 11 el;u~m and iii di_ reded rrom 0  lOll to"aro the midpoillt of the line C.'Oll'" n....ctillg the II ion5. (To rolwerl to 51 units, multiply by t )( 10".)
~.
~C
,, ,
__________________.,
Figure 2 Ekctrostatic potential aud field rotnpollents in CCS at po:$itiOt' r, 8 for a dipole p 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.)
e
CHAPTER
13:
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 :: J + curl E =
curl H = curl E =
41T
+   (E + 41TP)
1 a c at
~(oE
1 '8
c
'8
"
+ :P)
at
"
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 dtfined 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'
One contl'ibution to the e lectric field inside a body is that of the applied ele<:tric field . defined as
(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 WaVelengths 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 dcrhation 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:
f dV(P'b'T1lll~),
f dV(;dIV P+diV!) f
liS
~"
f dS: '
~.
I,late IW$
Q' 
<I '"
+1'
~: ,v)
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'.
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,
__ c
lui
"
..
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)
(51)
(5)
,
p
"
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) (S I)
Eh
=  N,P. ;
EIII
=  Nllf ,l =
E 1z =  N".P,," ; Eh = _ N".P&
'0
(8)
E.
_ NJ'"
EIII
_!!ti
'0
Here N", Nil ' N,," an> the depolari7.ation factors ; their values depend on the 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 Shape:~phere
Ad~
'"
(CGS)
N (51)
'"I
normal
41T13
TIlill slab TILin slab I...olg circular c),liooer Long circular crlinder
<.
0 0 2w
1/3 1
in plane longitudinal
tranS~"eI"Se
0 0 112
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 electrodes deposited on the opposite surfaces or a thin slab.
~R Becker. Electro.nlll(ndic {.elm lind intcrm;tion&, Bla.i)dell. 19&1, W . 102_ 107. 'J. A. o.bom. Phrs Ita. 67 , 351 (t 945), E . C . StOller, PhUosophll'lli MagD.irte 36, 8<XJ{l9I5)
. Eo
~re" 'J1}e
+ + E~r + + + + +
r.ct;tiou~
+ + +
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 ,
(9)
connect the macroscopic field E inside the ellipsoid with the polariz.1tion P.
Here XS "" 47rXccs I
"" Eo  NP ;
x(f.:o  .v P)
P= P=
On
(S I) E=Eo    ,
'0
NP
( 10)
P =
1+ r .;X
Eo;
Eo .
(II)
(SI)
X<o
1 + NX
3IJ6
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)
(Sl)
(12)
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)
Edil'oleP.J
(13)
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 EWe 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 CIsites 10 the lIIaCl structure and the Cs and 0  sites in the C.CI $Iructurc ha"e <:ubic symmdt)'
f ';Kure
I)
..... E,rnM"
",,'t"
b..""d,,~
1::,
,,,,,,1"'1,1,.,,
I"",, d't.,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 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 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:rcnce 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.
Eo and
0\.,,.
Ilere
= 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 m polarization of the body in which a hole 1 ... 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:
Eo
(ces)
(15)
and in 51 we replace Pi by p/47To. ~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
4~
(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),
(CCS)
EJocaI = Eo + E ,
+ ""3 p:;
471
+""3 P
471
(17)
(SI ) This is the LorentL 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 vacuum is defined in terms of the macroscopic field E:
(ces)
(SI)
+ 471P
E
= 1+ 471X
(18)
('of + P
<0
"""' IiX _
.....
in
UII' 
(ces)
""  ==  
E 
4"lT
(Sl) X  
,0
 =i  l
(19)
In a noncubic crystal the dielectric respOJlse is described b)' the components of the susceptibility tensor or of the dielectric constant tensor:
(ces)
(Sl)
(20)
TIle polari"tability a of an atom is defined in te rms ofthe loca.1 e1l."c tlic field at the atom: (2 1) p "" aE .........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 assembkd to form a Cl)'stal. For a nOllspherical atom a will be a tensor. The polarization of a crystal may be expressed approximately as the product 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.
p: (INpA (E+ 4; p)
X =
1' ==
E
'f.NJo'j
1  3INJoj
10
41T
(23)
By definition
= 1
+ 41TX
obtain (24)
(CGS)
 ~ ,N aj E J
E  141!' +2 3
(SO
the ClausiusMossetti relation. 'bis relates the dielectric constant to the ek'CIronic 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 lectronic, ionic, and dipolar, as in Fig. 8. The electronic contribution arises from the displacement of the electron shell relative to a nucleus. The ionic contribution comes from the displacement of a charged ion with respect to other ions. ihe dipolar polarizabilily arises from molecules wilh a permanent electric <lipole 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 commercial 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 _
41T
~
(CGS)
INpfielectronic)
(25)
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.
1.1
"'''" I",;.,."'.,"'\I
~r~Lr "~
"d'IM>I...
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
IO~
crnl
U'
0.029 O.fr29
Na ~
nc"
0.008
~It
n"
0.003
AIH
C..
0.0013
Si H
I'auling
0.201
IS
Pauling
jS.{TKS)
0'3."" (2,4)
S'Pauling
jSn'KS)
10.2
(55)
N.
0.300
M
..
0.179 0.290
0.094
0.052
0.0165
TiH
K'
0.83
1.133
c."
0,47
S~'
So"
0.286
y" 055
1.62
0.185
(0. 19)
(1.1)
Se' Pauling
j Scr KS)
Br4.n
4.091
Kr
~ 46
Rb '
1.40 1.679
10.5 (7. )
Tel 
,
0.86 (1.6)
13aZ,
Z..'" 0.37
X,
3.99
I'auling
IS{fKS)
14 .0
(9.)
7, 10 6. 116
""
2.422.743
1.55
(2.5)
..."
UJ.I
Co"
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  ''',
selfconsistent. oc'Cause the ci.:.'Ctronic polarizability of an I" " depends somewhat 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 hannonically 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) = 1 J/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
(ces)
wS 
(28)
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()
m(  w2 + ~o   et:"", .
TIle dipole moment
ha.~
the amplitude
(ces)
a (electronic) =
2
W
(29)
where JiJ is calk.od the oscillator strength of the ek'Ctric dipole transition between 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 c1osestpackl.'ll or even meta1lic structures. Hyd rogen and xenon, for example, become metallic 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 determined by the minimum ofthe frt.'e energy F "" U  TS. '[bere will be a transition 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 arrangement of nearby atoms; these are the quantities that change as the structure is changed. Some structural phase transitions have only small effects on thc macroscopic physical p roperties of the material. llowever. if the transition is influenced 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 lemperaturedependl:nt "alues of thc dielectric constant, the pieax:lectric effect, the pyroelectric effect, and electrooptical eftt.'c ts. including optical frequency doubling.
FERROELEc.. R1 C CRYSTALS .....
A ferroelectric crystal exhibits an electric dipole moment C\'cn in the absence of an external electric field . In the ferroelectric state the center of positive 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 tism; similarly, there is ul>ually a rapid drop in the dielectric constant as the te mpel ature increases.
13
Vif'/edr;n "nd
"'erro"'tric~
' OM
~~
.... " , .
8
7
...
, ,
100
T 4I9'C T,  I92"C
(h)
".
<I",,", and (e) th., specirtc heat
V" ....
300
'00
..
~"QCffidcnt
iation or (a) the dielectric constant e, (b) the p)roelectrie c~, of PbTi~. (After Remeika and CIa ... )
'"
Table 2 }<' crroe\ectrK: crystals
To obtain p. in the CCS unit or e~u emI, multiply lite w lue given in 3 x 103.
T" ill K KDP type
KH 2PO,J P"
in
poe cm ~ by
at
T II:
12:3
213
GeT,
TGS type
Tri~gl )'cinc
fiTO
'" '"
[78J
sulfate
Ttigl)'cine
322 295
l6 708 765
2.8
[29]
selenate
Perov$ldtcs
3.'
26.0 30.0 >50 50 71
[2B3]
[296] [523] [296] [296]
BaTiOa
KNbO, PbTiD, LiTaD,
Li NbO:J
1480
""
 " 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 involving the massdepcndence of the de Broglie wavelength. Neutron diffract jon data show that above the Curie lemperolture 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 ferrocledric crystal is CeTe with the sodium chloride strudure. We sh,,11 dcvote ourselves primarily 10 <:ryslals with thc pcrovskHe structure, Fig. 10.
""
o.
0
o n"
On."
o
C;
0"'
I~
,.,
.,
(pc~k;te).
Figure 10 (a) The crystal structure of barium tit anate. The prototype cr)'Sta\ Is calcium titana 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.
//
,7//
2(1
x 10
V
I
I
gJ
It
\I"""'"~ill ...
"~I
j""o.,
Tr1r~l
, ,
,
1lI"""buI........
'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) (Sl)
p := (8 X 10"' esu
lm ~(64
X 10 24 cm3)
iii!
5 X 10 18 csu em
B: 2 x 102'.) 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.lalion 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 BmcE in~t ein sense (TP. I). 199) of .1 timeindependent 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 ,Isymmetrilal 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 roHkitestructure 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."Ondorde r transition; the physical ideas can he carried over to a firstorder transition. In a seu>ndordcr 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~torder 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
+3
ICCS)
Bn
l: NP
I~)
SrT.o,
OLot+1,~~~~tt~'~+'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)
11$ .
oS
(33)
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
E
= T _ T" '
(35)
dose to the observed temperature variation ill the panleioctric state, Fig. 12.
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 restoring force. The ferroelectric BaTi03 at 24C 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 wr ,., 0 or E(O) = c;:c at the transition . The LST relation suggests onl)' that E(O) ex trapolates 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 1 1E(0) a: '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.
rJ' 
i~p2
+ ig4f>4 +
,~6
+...
(37)
13
40
, oro
I
 I
I I
;!
" "
, ,
I I I
l/
..,
T."'"..lul~ . ..
V j/'
,
00
,
'"'
'"
l'igure 13 I'JoI (If the: );Quare: of 1m: fll:(luenqo of the uro waVl!\'eclo r h'llrtSVene optic mode Ig;linst .tclion experi ments by Cowl'!). The broken te1npcnotuK. for S.Tt03 IU obse.nd in ne"tron ]nll' is th~' redpmcil of the dicll,<:tric constant from the measuKnICuts of MilSu; and Westphal.
d,rr...
"
" 'f s " f ," l " , ,., IT "" '" ".,  T.l. "
in
"
"
f igure 14 Dt.>qLasc: 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 logarithmic 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 unpolarized lattke is unstable. The variation of ~ with tempemture is ao:;o unted for by the rlllal expan ~ ion and othc r e ITcc of anharmonic lattice inte ractions. ;ts
Se<"QndOrtlc 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 mperature. 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 secondorde 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 scconclorder transition.
Jo'irstOrder 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, 
; 0
(42)
so tlJaI eithe r
P~
= 0 or
(431
",," '
 _..,
13 Didec:tri.,. <lnd Fnrocln:lri...
r~ T)
P,{O)
"
" " "
"
.,
m. _
"
n,
"
Figure 15 Spu<lI:lIll'OU polarization Vt: rSus temJkralllrc, for a Sf.(oml onlcr phase transItion . S
" "
'.,,
,
,"
'"
..,,,
Figure 16 C lass .)
(
T :> T,
Fiaure 11 Landau free energy f"nClio<! "ersus (pobrizalionJl in II r\l"lOtordcr I .."",ilion. at represenlalhe 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<.'Ondordcr 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 temperatures over the transition the terms in p4 and p6 may be negleeted; thus E ::;: y(T  Jo)P, or
(eeS)
t:(1' >
J'J =
+ 4'1TPI
"" J
+ 4rr1){T
 'J . O)
(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 v
00
t
Figure 18 Calculated values of !l,e spontaneous polarization as a runct;on temperature , with paruncten as ror barium titanate. (After w. Cochran .)
or
..:
0
.,
 00
.,
ro
T  T, .nd.: g C 
T > T,
T < T,
T < T,
EBEEJ EBEn
I @ @1@ 1@ I 1 1@ I I 1' @ I I I @ @ 1@ 1@ I I @1 @1 I @ @ 1@ I @1 I @1 I I 1@ I @ @ 1@ I @1 I @1 @1 I I
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 \V03
anlif"rrocl<'d.r;" slate, in K
1010
NaNbO a
PbZrO,l PbHf03
NII~H 2 PO.
793, 9 11 506
488
148
242
21.
304
254
(N1I,hi4I0e
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 halides. 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 polarization is in opposite directions. The net polari:tlltion depends on the difference in the volumes of the upv.>ard and downwarddirected 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 poiaril'.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 :,
"
,~
w
11,1.00,
"" ".
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.
+
+
I,
+
+ +
+
+
+
+
+
+
I ""
or
(>,
Figure 21 (a) $dlc:malic dnowing of atomic dl.placements on ",ther ~idc 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.
F~..roel~'ctric doona'ns O th" rae:.: ri a ~ngl" crystal or !Jar."," "Iaolal". 11." rare is n 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 wlutioo (R. C . Mllk... .)
Figure 22
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 poLtriz..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 electric fleld E npplied to the crystal will cause the crystal to l:>Cl..'Ome stmined. III schematic onedimellsionrunotntion , 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 relations 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
amI (h) slrcs ...'tl rerroclc.;,tric crystal. The sirCH
pol~~lion.
r"nockct"" 0')"'513\
d,M,.gCS the
l'ol~ri7J1tion
(I.,
Figure 24 (a) n,e ul1strcs.s<.,) cr~'slal has II thrclfold .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 thncdipole moments ~t cao:h "crtc~ is zero. (h) Thcl'f)'sl..! when strt.~.... t1t.,\c!op$ II pol"rizalion In the d,rt.oction indk~h:d. llu: sum of the dipole moments about d ~':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<:Ollatelcad 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 .. pp910 to 915, 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 characterize{! by the eledromcchanical ooupling factor k, whose square is defined by
".2 =
(47)
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'clneferroelastic trnnsition; lend phosphate Pb:J(PO' I)2 has a pure fcrroelastic transition; severnl crystals have combined ferroelC<..iricferroelastic transitions.
Opticlll Cerllmics
The addition ofkmthanum to the PZT system gives a reasonably tmnsparent product with useful eledrooptical properties. The PLZT system lias been used as the basis of optical memories.
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 volume. In a sphere N = 477"/3.
(
,
....
'111e polarizability
p = aE\nc.
0'
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 the ClausiusMossotH 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 wllere UII b the radius of the unper(Fig. 25), and Sl10W th;d for this I1'lOlIci a "" turbed orbit. Note; If the applied field is in the JC dirLeiion, 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 ,
at,.
"*
dius
Figure 25 An electron in 1\ dra, lar omit of m 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
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 <c I. TIle sU>.'s tcd proportionali ty of a to the wlx: of the ionic radius is !k'\lisfkd (Iuite wel l fOl' alkali :\,1(! halogen iol1s.
Nrr
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 jXlrallclplatc capacilor mooe up ofn..u paraltclla)'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)
where w is the angular frequency. Values of fdFa5 high as 10" or 1(iI cauS(.d largely by this Max .....ellWagncr mCdlanism arc sometimes round, but the high values arc alWdYS accompollit..(! by large ac losses.
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
""
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~
"A.
Figure 27 An air gIIll ofthkkncss qd is in ""ri~ in a OIpacilor wilh
"
,
,. ,
Ii
,\ir
"
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 secondorder phase transition the dielectric ('Onstant Ix:low the tr.lru;ition temperature is
t"
= I
+ 41'1'6P1l::
'"' I
+ 2Trl;(:fc 
'I) .
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 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 sortmod" 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
QUANTUM THEORY OF DIAMAGNETISM OF MONONUCLEAR SYSTEMS
PARAMAGNETISM
417
419
420
QUANTUM THEORY OF PARAMAGNETISM Rarc carth ions Hund 1'Ulcs lrlm group ions Crysbl field splitting Quenching of the orbital angular momentum SpectrOl;copic splitting factor Van Vleck temperatureindependent paramagnetism
431
432
SUMMARY PROBl,EMS
1. 2. 3. 4. 5. 6. 7. 8. Diamagnetic $UliC'eptibility of atomic hydn)gen Hund l'Ules Tliplet e:o;citcd states Ileal capacity f."Om internal degrees of freedom Pauli spin susc:eptibility Conduction elech'On ferromagne tism TwoIe"el s)'Slcm Paramagnetism of S = 1 system
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.
paramagnct~
s"hstances.
CIiAPTEI.
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 momentum 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
ICGS)
M X~B'
(51)
X~
""'"
B
(I)
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 susc(.'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 magnetic SU 1tibility are called diamagnetic. Substanl.'CS w ith a positive susceptiSCl.l 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 partially to shield the intcJior of a body from an applied magnetic ficld . In electromagnetism 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.
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 diamagnetism 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 nucleus. "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"cqucn<.)' 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). o llIe Larmor precession ofz electrons is C(luivaien t to an electric current
(SI) 1 = (charge){revolutions per unit lime) ""
.,.;.00
(Ze{~'
2"11"
2m
en) ,
(3)
...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
(SI)
" .  .Ip')
Z<'B
'on
(CGS)
=   Ip')
4m~
Z<'B
(4)
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,
 1.2  19.4
K,
 28.0
x,
 43.0
 1.9
In dielectric solids the diamagnetic contribution of the ion cores is described roughly by the Langevin rcsult. 'Inc contribution of conduction electrons is more oomplicatct!. as is evident from the de IIaasvan Alphen effect disl:tlsscd in Chapler 9.
QUANTUM TIIEOHY OF DIAMAGNETISM OF MONONUCLEAR SYSTEMS
From (G. 18) the effect of a magnetic field is to add to the ham iltonian the tcnns
(6)
for an atom ic c1ed:roll these terms may usually be treated as a small perturbation. If the magnetic field is uniform and in the:;:: direction, we rna)' write
iyB
(7)
(a
a) +   ( r + 2B'
BmCZ
y~
(8)
nle first term on the right is proportional to the orbital angular mOlllentum 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 sphericaJly symmetric system a contribution
E' = rrB
12m<f
{r~
(9)
magnet~
,ment is diamag_
  =   8
iJB 6mc2
aE'
<,<,~
(10)
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 molecular oxygen and organic biradicals. 4. Metals.
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 momentum; 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
1)
(13)
t 'igure 2 Enetl.Y lewl spl itt illg for one electron ' in R magnet;., field B dircdcd along the l)(Jiti\'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
s~ . t ~.,..
J Fradionai poJlO.,lations of a two,Ie.'Ci 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, N N, N
eXpVLBI"T)
:
(IS)
(16)
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 lotted 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, M = (N )  N2)J.I. "" NJ.I. .
l~or
e"  e X
e'
+e
= NJ.I. tanh x .
(17)
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)
"" ,.,
/~
, ,,
$ ..
(~")
'" "
I
} v'"
J
{fc" ,
~
,.,
."
I
...
s !
(Cr" )
V
0
:
0
,,1
I
''''' tOOK
3., ,
4 21 K
=

1\"IIoo;n .........11'....
"
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
+  x 3
I 2J clnh 2)
(X)
(20)
I
:t
i'
45
(21)
M B
C T
(22)
Here p is the effective number of Bohr magnctons, defined as ')  gfj{J + l)Jlf2 .
(23)
....
..
""l
I
II I
, '"
"
, ,,
/
nit' ltraighl line i. the
Figure S Plot oil/x u T (01" a g:odoI inimn salt, Cd(c..1I,.50J, 911.0. 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 CUlieBrlllouin law. and (22) is known as the Curie law. Results for the para
"..
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 rareearth ions, Table 1. 111C <''8.lculatcd 1118&IIC101l numbers arc obtained with g valucs from the Lande result (13) and the groundstate 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 LS 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 spinorbit 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 c11arncterized 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 u1 0mL repulsion between electrons. The exclusion principle prevents two electrons 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 opposite 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 halffilled . 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
....
"""
Th"
100
eonflgurlllioJ,
p~ '
l\'d 3 +
'{'5ry'
4j'srV6

Basic \evel

g[jU + Ill'"
2.54 3.58 3.62 2.68
0.84 0
p(caJc) 
,J(exp), approximate
~,
ffM
Pm3 Sm3+
'/,5,.".
'H,
4psrp6 4fssrp6
'I,
'I~
3.5 3.5 1.5 3.' 8.0 9.5 10.6 10,4 9.5 7.3 '.5
Eu 3+ ed 3+
1103+ Tm3+
E~'
'f'5?'P'
'/,sr,,'
4rsr,I'
4r~ps
4f1Oss'p~
0,"
4f'ISs',l' 4PYjI
4p~r;
'I,
"1 1 m
Yb3 +
aIls 2F71l
'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 spinorbit 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 necessarily 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 calcuIL Pallhng . ud E. 8. Wilson, 'n,roductiOfl 10 quant"m " ,td"",I<:$, M~l'2w lI il1 . 1935, pp
2392~6.
."
Table 2
I~
~
Conng
uration
dJU +
Jl(calc) 1)]'11
p(exp)"
Ti H
yt+
V"
C~ ...
"", 3d'
3d'
",p "'"
'D~
3"'2
y'2+
"l n3 ~ .
Fc3 + , MnH
Cr
~F3.'2
liDo
Fc
COH
NiH
Cu H
"''
3d" 3d'
"'~
s", 'v,
.n
1.55 1.63
'F., 'F,
a,.
5.92
6.63 5.59 3.55
'v",
...
.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 interaction 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 mbit 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
.....
~.
,
===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 < and (c). l1le w,,,",,functions that g",e rise 10 t~ charge Jensities are of the fonn "f<r), Ilf(r) b) 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 spinorbit 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 moment 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 spinorbit 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 electric 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:.'Cfold 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 fieldproportional 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
(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,)
Consider for example
(V.le<lVJ
(UJe<lUJ
~ 0 .
(27)
(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.> =
if{r)12{Ax +
A(II  I i) ,
(29)
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 average 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 )ahoTeller 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.
\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 represented 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)inorbit interaction energy AL S. Jfthe ground state functiou is ifJo = Up :z: xf(r)a in the zeroth approximation, 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
:t
direction is
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
:t
direction is
(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).
\\'e consider an atomic or molecular system which has no magnetic ,noment 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)
'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:Iation is ..
At ~ 2BI('I~.IO)I'
NA
21BT
A
which gives for the susceptibility
(36)
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 polarization 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
.:Jl
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 interest 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 oneshot 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 orient the spins prefcreotially. Some of tllese interactions are discussed in Chapter 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
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.
1.
3fh.e method was 5l1ggCSted by P. Debye. Ann. Physik 81. 11s.l (1926): aOO W. F. Gtauque. Am . Chrm. Soc. 49, 1864 (\927). For many pu~ the method has been 5upplanttd by the lielie' 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
f
"
L,.>IIiff>
I ~'~ ~.,.~.~,1';" I
:
I
...
N"",' ~",/Jbriurn
I
~I
Ti .....
Tome II "hod!
i> ",,,,,",'.'Il
fl..k1 .. rein.",,,.!
Figurt' 7 Durlng adiabatic J emagnc til.ation the total ..."tropy 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.
or
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 ulated (au"" 0) and the field removed; the specimen follows the constant e ntrop), path be, ending up at temperature T:z. The thermal contact at TJ is providcc.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 JtB1k8T, hencc of Brr. The spinsystem 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)
(41)
where B is the initial field and T. the initial temperature. Because nuclear magnetic moments arc weak, nuclear magnetic interactions 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 spincooling experi ment must be lower th.1Il in an c1(.'Ctron spincoolillg 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.'
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 utrop)'. 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 carried 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 magnetic 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 nudd at the temperature of the first stage. The present recordS for a spin Icmperature 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 agreement with the existence of a magolelic rnOllwnt of the electrons. W. Pauli. 1927
Classical free electron theory gives an unsatisfactory account of the paramagnetic 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 EngJnLe~;nS\ 8, I (l963). ~P. 1. Hakoot'n rt aI. , PhflS. 1Iftl . I..dl 70,2818 (1993).
1,,;h.1
""""lei", /kid ,n kG
"
ro
'"
, "
"
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 Curietype p.. 'lramagnctic contlihution (22) to the magnetization of the metal: AI = NptBlkBT. 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 FermiDirac distribution (Clulpter 6) would correct the theory as required. We fin;t give a qualitative explanation 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 magnetization 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 susceptibility. Hence
M~~  =~B
Np.'lB
Np.'l
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.
(
."'.,
,.
,.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"
lrE D(E
+ IlB)
a 
H"B
Ii"
+
dE D(E)
+ 1
2
IlB D(E,.)
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"
"II
dE D(E  IlB)
ill 
1 IlB D(E~')
2k nT,.,
(42)
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 /II == N'"
_r_'_ B
kaT"
(43)
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 electronelectron 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 tals (with unfll kd 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 susceptibility X = NI.83kB'r. for J.tB <C kB'f. (CurieLangevin) For a system of spins S = t the cxact Np, tan h(p,BlkBn. whe re p, "" hlP,B' (Brillouin) magnctization is
M =
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 statistical 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
"
, ,
1'\
To
\
0
V
1\
,...C,
"
~\\'
VI:
0
\
"
/
m ,
"
\10
.0
"
l::::lib
0
"r
600
"
JijOO 2000 2200
1'00
400
IiOO
1000
1200
1400
JGOO
T . jll K
Figure II 1;,mpcrnturc dCllC.u]cnct' of the magnetic sU$Ct.1It il>ility of ",clals. (Cou rtesy of C. Kri~'SSma... )
J.
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 theti.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 twolevel 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)".'"
11.e functiun is plutted in Fig. 12. Peaks of tMs type in tl.e heolt cap.lCit y are often knuwn as Schott!.." anuma lies. The ma'l:imum heat capacity is q uite high , L t for lU 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 experimentally 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 alISOlute zero may Le discussw by
anotll{~r
method. Let
be the CQIlccntratinns of spinup and spin.(lo"n electrons. (a) Show that in a magtletic field B the total energy of the spinup 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 P2 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 spinu p band is
+ = ,,(1
+~
jVN~l
+ (f
 INPB ( 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 ,
so that the exchange illteraction enhances the susceptibility. (c) Shuw tl.at .....itl. IJ = 0 tI.e total energy is unstable at 0 when V > 4f;,J3.'IJ. If this is satisfied a ferromagnetic 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:
," , ,
,., ,
~ .
l<~d
Lc>ell
J ,"
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 IlIret~"lh and trun>ition.l:TOup llIetal., coltll'o..."b. ~nd >tlk,>"s.
..
, ''''" "
1
C~T'  431(
10 ....
,,
0001 
,>./
OSElEiTl/ r
' U
0.002 
"
..
..
'"
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_)
, \
410
(
7. Twolevel 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=
(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 i 1"1. (b) Show that in the limit ~ .o::g k1' tile result is M ;i2 (21"1J.( f3kT)B.
References
A. Abragam and B. Blcancy, EI""tron parm1l<"lgnetic rt8Qnana (Jj tran$i!ion ions, Dover, 1986.. H. B. C. Casimir, MagnetISm and ury /ow tempertJturu, Don'r, 1961. A cJassic. M. I . Darby and K. N. R. Taylor, Pllysia of rare oorlh solids, Halsk d. 1972. A. J. Freeman, rIle actinides. electronic structure and related properlics. Academic, 1974 . R. D. II udso", Principies and applicat ions of mlIgnetic ccoling, Elsevier, 1972. H. Koocllfel , Pv/.red "Igh mlIgrwtic!tlds. North Holland. 1970. O. V. Louna$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 derivlItions 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, "JahnTel!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 revk'W," Amer. J Physics 61, 688 (1993).
15
Ferromagnetism and Antiferromagnetism
FERI\OMAGNJ:o:TIC ORDEI\ Curie point llncl tlLc exchange integral Temperature depenoence of tile saturation magncti:w.tion Sah,ration magnetization al absolute zero
~~CKONS
454
455
4S6 458
460 461
462 465
466 468 470 472 473
474
475
SINGlE DOMAIN I'ARTICIES Geomagnetism Dnd biomagnctism Magnetic force microsc:o(>y PltACr;ETlC BUnBLE DOMAINS
SUMMARY
I'ROBLEMS
Magnon dispersion relation Heat capacity of magnons Nl!cl temperature Magnctoeillostic coupling 5. Coercive ror<.'C or a small particle 6. Saturation magnetization ncar T~ 7. Nl!el wall
1. 2. 3. 4.
481
481 482 482 402 482 483 483 483
REFERENCES
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
S;,nple
r~rt",nagnel
I 1I
IIdl<'>l.pm .....,
Simple
anllrc .to,~
Fcn i ...."'1
1

/\/\/
r....,ltd_Ur.:~
D~~~~~
  ~

, ,.,
!.
F~",azybo..d
.,
CHAPTER
15:
FERRO~IAGNEnC
onOEH
A ferromagnet has a r;pontan cous magnetic momcllta 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 mument, called the saturation momeut ,
Ct,rie Point
(nit!
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 . 1tion is defined as the magnetic moment per unit volume; 111e magneti7.. unless otherwise specified it is understood to be the value in thermal equilibrium at the temperature T. If domains (regions magnc ti7.ed in different directium) 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 experiences 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 spontaneous 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 calkd thc molecular Ildd or thf Weiss Ildd, ~fter Pierre Weiss "nO was 'he Ilut to 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
It,.
' ..
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 ;
(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 speciIncn 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
(3)
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 urieWeiss law
(CCS)
x :=
c
T  Tc '
(4)
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 determine 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
~n'ic
PA.U, 1965
..
TA... .: 1 Critical point c:<ponents for ferromagncts
..
" TAs
F.
Co Ni
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.
,
1.33 ~ 0.015 1. 21 ~ 0.04 1.35 ~ 0.02 1.3 ~ 0.1 1. 63 0.02 1.215 0.02 0.34 0.42
~
~ ~
T,. in K
0.04 0.07
G<l
CrOll Cr Ur3 EuS
, ,
,
/
/
"" '"'
/ ""
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 CJuanlummech.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 depe