Problems 53

metal, H. A. Lorentz s howed, using the classical Maxwell-Boltzmann velocity dis-

tribution, that an energy-dependent relaxation time would lead to temperature depen-
dence in the DC and AC conductivities, as well as to a nonvanishing magnetoresis-
taoce and a field- and temperature-dependent Hall coefficient. As one might now
expect from the inappropriateness of the classical velocity distribution, none of these
corrections were in any way able to bring the discrepancies of the Drude model into
better alignment with the observed facts about metals. 31 Furthermore, we shall see
(Chapter 13) that when the correct Fermi-Dirac velocity distribution is used, adding
an energy dependence to the relaxation time bas little significant effect on most of
the quantities of interest in a metal. 32 Tf one calculates the DC or AC conductivities,
the magnetoresistance, or the Hall coefficient assuming an energy-dependent •(E), the
results one finds are the same as those one would have calculated assuming an energy-
independent T, equal to T(EF).ln metals these quantities are determined almost entirely
by the way in which electrons near the Fermi level are scattered. 33 This is another
very important consequence of the Pauli exclusion principle, the justification of which
will be given in Chapter 13.

PROBLEMS

• 1. T!Je Free and Independent Electron Gas in Tf<•o Dimensions

(a) What is the relation between n and kF in two dimensions?
(b) What is the relation between kF and '• in two dimensions?
(c) Prove that in two dimensions the free electron density of levels g(B) is a constant indepen-
dent of 8 for 8 > 0, and 0 for & < 0. What is the constant?
(d) Show that becauseg(&) is constant, every term in the Sommerfeld expansion for n vanishes
except the T = 0 term. Dedm:e that J1 = SF at any temperature.
(e) Deduce from (267) that when g(E) is as in (c), then

(2.95)

(f) Estimate from (2.95) the amount by which J1 differs from BF. Comment on the numerical
signitlcance of this of the Sommerfeld expansion, and on the mathematical reason for
the •failure."'

2. Thermodynamics of the Free and Independent Electron Gas

(a) Deduce from the thermodynamic identities

c,, =
cT n
=
c1 n
, (2.96)

" The Lorentz model however. of considet"a"ble imponaoce m the description of semiconductors
(Chapter 29).
•• The thermopower is a notable exa!ption.
J• These assertions are correct to leading order in k6 Tftr. but in metals lhis is always a good expansion
parameter.
54 Chapter 2 The Sommerfeld Theory of 1\letals

from Eqs. (2.56) and (2.57), and from the third law of thermodynamics (s - 0 as T- 0) that the
entropy density, s = Sf Vis g.tven by:

f
dk
s = - ks n [flnf + (1 -f) In (1 - /)], (2.97)
4 3
where /(8(k)) is rbe Fermi function (Eq. (2.56)).
(b) Srnce rbe pressure P satisfies Eq. (B.5) in Appendix B, P = - (11 - Ts - Jln), deduce from
(2.97) that

P = kBT f:: 3 1n ( 1 + exp [ - ]). (2.98)

Show chat (2.98) implies that P ss a homogeneous function of Jl and Tof 5/2; that is,
P(J..J.L, ).T) = ). S t l P(J.L, T) (2.99)
for any constant )..
(c) Deduce from the thermodynamic relations in Appendix B that

= n, = s. (2.l00)
GJ.l T c.:T
"
(d) By differentiating (2.99) with respect to;_ show that the ground-state relation (2.34) holds
at any temperature, in the form
P = ju. (2.101)
(e) Show that when k8 T « CF, rbe ratio of the constant-pressu re to constant-volume specific
heats satisfies ·
(k 8 2+
(c.e
C.,
) _ 1= !\
3 SF-)
0
&F-)
(f) Show, by retaining further terms in the Sommerfeld expansions of 11 and n, that correct
to order T 3 the electronic heat capacity is given by
nz
c., = ) k 82Tg(EF)

-
90 B
k 4T3
g
(E )
F
[ts (g'(<.F))
g(&F)
2
- 21 g"(EF)J
g(&F) •
(2.102)

3. The Classical Limit of Fermi-Dirac Statistics

The Fenni-Di.rnc distribution reduces to the Maxv;eU-Bolt:zmann distribution, provided that the
Fermi function (256) is much less than unity for every posiuve &, for in that case we must have
J(e) .:::: (2.103)
The necessary and sufficient condition for (2103) to bold for all positive f, is
e - "·"'-BT » J. (2.104)
(a) Assuming lhat (2.104) holds, show that
r. = e - ,. /3t8T 31 '3nl/6f1( 2 mksT)- t' '-. (2.105)
 Problems 55

In conjunction with {2.104) this requires that

(2.106)
'• »
which can also be taken as the condition for the validity of classical statistics.
(b) What is the significance of the length rs must exceed?
(c) Show that (2.106) leads to the numerical condition

r,.
a0
»(lOs K)l/2.
T
(2.107)

(d) Show that the normalization constant m3 /4n3 n3 appearingin the Fermi-Dirac velocity dis-
tribution (2.2) can also be written as (3.JicJ4)n(nlj2nk8 T F ) 312 so that f s(O)//(0) = (4/ 3.Jir)( Tr j T)3 12 •
4. Inseni>itivity of the Distribution Function to Small 01anges in the Total Number
of Electrons
ln deriving the Fermi distribution (page 41) we argued t.hat the probability of a given level being
occupied should not change appreciably when the total number of electrons is changed by one.
Verify that the Fermi function (2.56) is compatible with this assumption as follows:
(a) Show, when k 8 T «SF, that when the number of electrons changes by one at fixed tem-
perauue. the chemical potential changes by

(2.108)
where: g(S) is the density of levels.
(b) Show, as a consequence o f this. that the most the probability of any level being occupied
can change by is

(2.109)

[Use the free electron c:v<iluation (2.65) of g(&r).] Although temperatures of millidegrees Kelvin
can be reached, at which Srfk8 T::::: 108 , when N is of order 10 22 then llfis still negligibly small

•
82 Chapter 4 Crystal Lauices

OTHER ASPECTS OF CRYSTAL LATfiCES

This chapter bas concentrated on the descnptioo of the cranslatioual symmetry of
crystal lattices in real physical space. Two other aspect s of periodic arrays will be
deall with in subsequent chapters: in Chapter 5 we examine the consequences of
translational symmetry not in real space, but in the so-called reciprocal (or wave
L'ector) space, and in Chapter 7 we describe some features of the rotational symmetry
of crys tal lattices.

PROBLEMS

1. In each of the following cases indicate whether the structure is a Bravais latti<.-e. If it is, give
three primitive vectors; if it is not, describe it as a Bravais lattice with as small as possible a basis.
(a) Base-centered cub1c (simple cub1c with additional points in the centers of the horizontal
faces of the cubic cell).
(b) Side-centered cubic (simple cubic with additional points in the centers of the vertical
faces of the cubic cell).
(c) Edge-centered cubic {simple cubic with additional points at the midpoints of the Jines
joining nearest neighbors).
•
2. What is the Bravais lattice formed by all points with Cartesian coordinates {n 1 , 11 1 , n3 ) if:
(a) are either all e-.-en or all odd?
•
(b) The sum of then, is reqUired to be e\'ell?

3. Show that the angle between any two of the lines (bonds) joining a Site of the diamond lattice
1
to its four nearest neighbors 1s cos- ( - 1/3) = 109"28'.

4. (a) Prove that the Wigner..Seitz ce.ll for any tw<Hlimensional Bravais lattice is either a
hexagon or a rectangle.
(b) Show that the ratio of the lengths of the diagonals of each parallelogram face of the
Wigner-Seitz cell for the fa<;e-centcrcd cubic lattice {Figure 4.16) is J2:1.
(c) Show that every edge of the polyhedron bounding the Wigner-Seitz ceU of the body-
centered cubic lat6ce (Figure 4.15) is fi/4 times the length of the conventional cubic ceiL
(d) Prove that the hex agonal races of the bee Wigner-Seitz cell are all regular hexagons.
(Note that the axis perpendicular to a hexagonal race passing through its center bas only threefold
symmetry, so this symmetry alone IS not !!llOugh.)

S. (a) Prove that the ideal cfa ratio for the hexagonal dose-packed StruCture is .j8i3 = 1.633.
(b) Sodium transforms from bee to hcp at about 23K (the ··martensitic" transformation).
Assuming that the density remains fixed through this transition, find the constant a of
the hexagonal phase, given that a = 4.23 A in the cubic phase and that the c/a ratio is indistin-
guishable from its value.
6. The facoxentered cubiC-is the most dense and the simple cubic is the least dense of the three
cubic Bravais lattices. The diamond structure is Jess dense than any of these. One measure of
this is that the coordination numbers are: fcc. 12; boc, 8; sc. 6;diamond, 4. Another 1S the follo\\'ing:
Suppose identical solid spheres are distcibuted through space in Sl. way that their centers
 Problems 83

lie on the points of each of these four s truCtures, and spheres on neighboring points just touch.
v.ithom ovcrlappmg. (Such an arrangement of spheres IS called a close-packing arrangement.)
Assuming that the spheres have umt density, show that the density of a set of close-packed spheres
on each of the four structures (the -packing fraction-) is:

fcc : .Ji.nf6 = 0.74

bee: .J3nj8 = 0.68
sc: nj 6 = 0.52
diamond: .J3nj l6 = 0.34.
7. Let N. be the numtxr of nth nea rest neighbors of a given Bravais lattice poin t (e.g.. in a simple
cubic Bravais lattice N 1 6, N 1 = J 2. etc.). Let r. be the distance to the nth nearest neighbor
expressed as a multiple of the nearest neighbor distance in a simple cubic Bra•ais lauice
r 1 = I, r 1 = .j'i = 1.414). Make a table of N. and r. for 11 = I, ., 6 for the fcc, bee. and sc
Brava1s lattices.
8. (a) Given a Bravais lattice, let a 1 be a vector joining a particular point P to one of1ts nearest
neighbors. Let P' be a lauice point not on the line through P in the direction of a, that is as
close to the line as any other lauice point, and let a 2 join P to P'. l et P" be a lattice point not
on the plane through P detennined by a, and a 2 that is as close to the plane as any other lattice
point, and let a 3 join P to P". Prove that a 1, a 2 • and a 3 are a set of primitive vectors for the
Bravais laHice.
(b) Prove that a Brava1s Iau ice can be defined as a discrete set of vectors, not all in a plane.
closed under addition and subtraction (as on page 70).
 Problems 93

(010) ( 110) (Ill)

Figure 5.5
Three planes and their Miller indices in a simple cubic Bravais lattice.

the (100), (010), and (001) plane.s are all equivalent in a cubic crystdl. One refers to
them collectively as the {100} planes, and in general one uses {hkl} to refer to the
{ltkl) planes and all those that are equivalent to them by virtue of the crystal symmetry.
A similar convention is used with directions: the [100], [010], [001], [IOO], [OIO],
and [OOI] directions in a cubic crystal are referred to, collectively, as the (100)
directions.

This concludes our general geometrical discussion of the reciprocal lattice. ln

•
Chapter 6 we shall see an important example of the utility and the power of the
concept in the theory of the diffraction of X rays by a crystal.

PROBLEMS

1. (a) Prove that the reciprocallallice primitive vectors defined in (5.3) satisfy

(5.15)

(Hint : Write b 1 (but not b 2 or b3 ) in tenns of the and usc the orthogonality relations (5.4).)
(b) Suppose primitive vectors are constructed from the bi in the same manner (Eq. (5.3)) as
the bi are constructed from the a;. Prove that these vectors are just the ai themselves; i.e., show that

b2 x b 3
2n b, . (bz x bl) = a" etc. (5.16)

(Hirzt : Write b 3 in the numerator {but not b 2 ) in terms of the· ah use the vector identity A x
(B x C) = B{A • C) - QA • B), and appeal to the orthogonality relations (5.4) and the result
(5.15) above.)
(c) Prove that the volwne of a Bravais lattice primitive cell is

(5.17)
where are three primitive vectors. (In conjunction with (5.15) this establishes that the volume
of the reciprocal lattice primitive cell is (2n) 3j v.)
94 Chapter 5 The Reciprocal Lattice

2. (a) Using the primitive vectors given in Eq. (4.9) und the construction (5.3) (or by any other
method) show that the reciprocal of the simple hexagonal Bravais lauice is also simple hexagonal,
with lattice constants 2nfc and 4n!.j3a, rotated through Jo• about the c-axis with respect to the
direct lattice.
(b) For what value of c'/a does the ratio have the same value in both direct and reciprocal
lattices? If cfa is ideal in the direct lattice, what ts its value in the reciprocal lattice?
(c) T he Bravais lattice generated by three primitive vectors of equal length a, making equal
angles 0 with one another, is known as the trigonal Bravais lauice (see Chapter 7). Show that the
reciprocal of a trigonal Bravais lattice is also trigonal, with an angle 0* given by - cos =e•
e
cos 0/ [ 1 + cos 8], and a primitive vector length a*. given by a* = (2rt/ a)(l + 2 cos cos 0*)- •
111

3. (a) Show that the density oflattice points (per unit area) in a lattice plane is d/v. where vis
the primitive cell volume and d the spacing between neighboring planes in the family to which the
given plane belongs.
(b) Prove that the lattice planes with the greatest densities of points are the {111} planes it1
a face-centered cubic Bravais lattice and the { 110} planes in a body-centered cubic Bravais lattice.
(Hint: This is most easily done by exploiting the relation between families oflatticc planes and
reciprocal lattice vectors.)
4. Prove rhar any reciprocal lattice vector K is an integral multiple of the shortest parallel
reciprocal lattice vector Ko. (Hint: Assume the contrary, and deduce that since the reciprocal
lilllice is a Bravais lattice, there must be a reciprocal lattice vector parallel to K shorter than K0 .)
146 Chapter 8 Electron Levels in a Periodic Potential

PROBLEMS

1. Pe•·iodic Potentials in One Dimensio11

The general analys.is of electronic levels in a periodic potential, independent oft he detailed features
of that potential, can be carried considerably further in one dimension. Although the one-dimen-
sional case is in many respects atypical (there is no need for a concept of a Fermi surface) or
misleading (the possibility-indeed, in two and three dimensions the likelihood-of band overlap
disappears). it is nevertheless reassuring to sec some of the features of three-dimensional band
structure we shall describe through approximate calculations, in Chapters 9, 10, and 11, emerging
from an exact treatment in one dimens ion.
Consider, then, a one-dimensional periodic potential U(x) (Figure 8.4}. 1t is convenient to view
the ions as residing at the minima of U, which we take to define the ze ro of energy. We choose
to view the periodic po tential as a superposition of potential barriers t;{x) of width a. centered at
the points x = + na (Figure 8.5):
00

U(x) = L v(x - na). (8.64)

, := - CXl

liigure 8.4
A one-dimensional periodic
U(x) potential U(x). Note that the
ions occupy the positions of
a Bravais lattice of lallice
constant a. lt is convenient
to take these points as having
X coordinates (n + and to
-3a/2 -a/2 0 311/2 choose the zero of potential
to occur at the position of
the ion.

Figure 8.5
Jllustrating particles incident
u(x) v(x)
from the Jell (a) and right (b)
on a single one of the bar-
riers separating neighboring
ions in the periodic potential

-, -I

-,
I'
of Figure 8.4. The incident.
transmitted, and reflected
-a(1. a{2 -a/2 a/2 waves are indicated by ar-
(a) (b) rows along the direction of
propagation. proportional to
the corresponding ampli-
tudes.

The term t'{x - na) represents the potential barrier against an electron tunneling between the
ions on opposite sides of the point simplicity we assume that v(x) ,. v(- x) (the one-
dimen sional analogue of the inversion symmetry we assumed above), but we make no other
assumptions about v, so the form of the periodic potential U is quite general
 Problems 147

The band structure of the one-dimensional solid can be expressed quJte Stmply in teems o f the
properties of an electron in the presence of a single-barrier potential t{.>.). Consider therefore an
electron incident from the left on the potential barrier with energyn & = 11 1 K 1 :2nL Since
r( "<) = 0 when J."<J u/ 2. in these regions the wave function 1/11(x) will have the form

a
2"

(8 .65)

This is tllustrated schematically in Figure 8.Sa.

The transmission and reflection coefficients t and r give the probability amplitude that the
electron will tunnel through or be reflected from the barrier; they depend on the incident wave
\'ectOr K in a manner determined by the detailed features of the barrier potential v. However,
one can deduce many properties of the band structure of the periodic potential U by appealing
only to very general properties of l and r. Because v is 1/l,(x) = 1/11( - x) is also a solution
to the Schrodinger equation with energy{;. From (8.65) it follows thatl/l,(x) has the form

a
2'
a
2' (8.66)

Evidently this describes a particle incident on the barrier from the right , as depicted in Figure 8.5b.
Since 1/11 and 1/1. are two independent solutions to tbe single-barrier SchrOdinger equation with
the same energy, any other solution wtth that energy will be a linear combinatioo 34 of these two:
1/J = Al/11 + B!/1,.. ln particular, since the crystal Hamiltonian is identical to that for a single
ion in the region - af2 x af'l, any solution to the crystal SchrOdinger equation with energy
& must be a linear combination of r)J 1 and rJ!, in that region :

(8.67)
1/f(x) = Al/11(x) + Bl/f,(x),

Now Bloch's theorem asserts that 1/1 can be chosen to satisfy

1/f(x + a) = fil"'l/f(x), (8.68)

for suitable k. Dilferentiaung (8.68) we also find that 1/1 = dl/lf dx satisfies

1/J'(x + a) = e''""'/l'(x). (8.69)

(a) By imposing the conditions (8 . 68) and (8.69) at x = - o/ 2, and using (8.65) to (8.6 7),
show that the energy of the Bloch electron is related to its wave vector k by:
2 - r 2 .•. 1
cos ka =t 2r
e•no +-
2t
• v,.

'
(8.70)

Verify that this gives the right answer 111 the free electron case (v !!!!! 0).

u Nme · in this problem K is a continuous variable and has nothing to do with the reciprocal tau ice.
•• A speoal orthe general theorem that there are n independent solutions tO ao nth-order linear
dafferential equation.
148 Cllapter 8 Electron Lenis in u Periodic P o tentia l

Equation (8.70) is more informative when one :supplies a lillie more information about the
transmission and reflection coeffic•cots. \Vc write the complex number tan terms of its magnitude
and p hase:
t = tl".ltl (8.71)
The real number lJ is known as the phase !>hllt, since it specifies the change in phase of t he trans-
milled wave relative to the incident one. Electron conservati on requires tha t the probability of
transmission plus the probability of reflection be unity:

(8.72)
This, and some other useful information. can be proved as follows. Let </> 1 and 4>2 be any two
solutions to the one-barrier Schrodingcr equation "ith the same energy:
hz r12K2
-2m <Pt + t·<!J, = -2n-;- </J;, i = I, 2. (8.73)

• • <Pzl (the -wronskian" ) by

(8.74)
(b) Prove lhat n· is independent of x by deducing from (8. 73) that its dcnvativc vanishes.
(c) Prove (8.721 by evaluating ""1.;,,. .;,,•) for x -a/ 2 and x a/2. noting that because v(x)
is real .;,,• will be a solut ion to the same SchrOdingcr equation as.;,,.
(d) By evaluating "1r/la. .;,, •) prove that rt• is pure imaginary, so r must have the form
(8.75)
where c is the same as in (8. 7 n
(e) Show as a consequence of(8.70). (8.72}. and (8.75) that the energy and wave vector of the
Bloch e lectron are related by

+ b) = cos ka, (8.76)

Since It! is a lways less than one, but unity for large K (the barrier becomes in-
creasingly less effective as the incident energy grows). the left side of (8.76} plotted against K has
the structure depicted in Figure 8.6. For a given k. the allowed values of K (and hence the allowed
energies E(k) = lt 2 K 2 '2m) are given by the intersect ton of the curve in Figure 8.6 with the hori-
zon tal line of height cos (ka). N o te that va lues o f Kin the neighbo rhood o f th ose satisfying

Ka + {) = mr (8.77)
give Ieos (Ka + ltl > I . and are therefo re not allowed for any k . TI1e corresponding regions
of energy arc the energy gaps. If 6 is a bounded function o f K (as is generally the case). then
there will be infinitely many regions of forbidden energy, and also intiniteJ} many regions of
a llowed energies for each va lue of k.
(f) Suppose the barrier ts very weak (so that lcl l. lrl : : : 0, lJ 0). Show that t he energy
gaps arc then very narrow. the width of the gap containing K = nnto being
,,2
s•• p 2nn
ma
2 jrj. (8.78)

(g) Suppose the barrier is very strong, so tha t lrl 0. ::::: l. Show that • lowed bands
 Problems 149

Figure 8.6
Characteristic form or the
function cos (Ka + o)/jtj. Be-
cause jt(K)j is always less
than unity the function will
exceed unity in magnitude
in the neighborhood of solu-
tions to Ka + o(k1 = rm.
Equation (8.76) can be satis-
fied for real k if and only
if the function is less than
unity in magnitude. Conse-
quently there will be allowed
(unshaded) and forbidden
(shaded) regions of K (and
therefore of e = h 2 K 1 j2m).
Note that when jtj is very
near unity (weak potential)
tJ1e forbidden regions will be
narrow, but if jtj is very small
(strong potential) the allowed
regions ·will be narrow.
of energies are then very narrow, with widths
•

e.,.,. - t::m;n = O(jti). (8.79)

(h) As a concrete example, one often considers the case in whicl't t{x) = go(x), where o(x)
is the Dirac delta function {a special case of the .. Kronig-Penney model"). Show that in this case
tz 2 K
cot fJ = - mg , ltl = cos 5. (8.80)

Tltis model is a common textbook example of a one-dimensional periodic potential Note, how-
ever, that most of the structure we have established is, to a considerable degree, independent of
the particular functional dependence of jtj and oon K.
2. Density of
{a) In the free electron case the density of levels at the Fermi energy can be written in the
form (Eq. (2.64)) g(EF) = mkF/h1n 2• Show that the general form {8.63) reduces to this when
&"(k) = 1r2 k 2 j2m and the (spherical) Fermi surfa<."C lies entirely within a primitive cell.
(b) Consider a band in which, for sufficiently small k,e.,(kl = E0 + (tiJ./2)(k/j m, + k/f m,. +
k, /m..) (as might be the case in a crystal of orthorhombic symmetry) where m_., m1 , and m, are
2

positive constants. SllOW that if e is close enough to eo

that this form is valid. then g.(&) is pro-
portional to (E - to) 111, so its derivative becomes infinite (van Hove singularity) as & approaches
the band minimum. (Him : Use the form (8.57) for the density of levels.) Dcduc.-e li:om tllis that if
the quadratic form for &.(k) remains valid up to eF, then g.(SF) c-c1n be written in the obvtous
generalization of the free electron form (265);
3 ll
9n(EF) = - (8.81)
2 SF- Eo
where n is the contri' ··)n of the electrons in the band to the total clect[()nic density.
150 Olaptcr 8 Electron Len!ls in a Periodic Potential

(c) Cons1der the density oflevels in the neighborhood of a saddle point, where f,.(k) = f.o +
(ll f2)(k/fm" + li} fm} - k,
1
where m", mn and m, are positive constants. Show that when
& to, the dcrivat;vc of the density of levels has the form

Bn'(e) constant,
e )- 112, (8.82)

-
 Spin-Orbit Coupling 169

IMPORTANCE OF SPIN-ORBIT COUPUNG AT POINTS OF HIGH

SYMMETRY
Until now we have regarded the electron spin as being completely inert dynamically.
In fact, however, an electron moving through an electric field such as that of the
periodic potential U(r), experiences a potential proportional to the scalar product of
its spin magnetic moment with the vector product of its velocity and the electric field.
This additional interaction is referred to as the spin-orbit coupling, and is of great
importance in atomic physics (see Chapter 31). Spin-orbit coupling is important in
calculating the almost free electron levels at points in k-space of high symmetry,
since it often happens that levels that are rigorously degenerate when it is ignored
are split by the spin-orbit coupling.
For example, the splitting of the electronic levels on the hexagonal faces of the
first zone in hcp metals is entirely due to spin-orbit coupling. Since the strength of
spin-orbit coupling increases with atomic number, this splitting is appreciable in the
heavy hexagonal metals, but can be small enough to be ignored in the light ones.
Correspondingly, there are two different schemes for constructing free electronlike
Fermi surfaces in hexagonal metals. These are illustrated in Figures 9.11 and 9. 12.

Figure 9.12
A representation of the Fermi surface of a divalem hcp metal
obtained by reassembling those pieces in Figure 9.11 that were
severed from each other by the horizontal hexagonal faces of the
first Brillouin zone. The first and second zones together make up
the structure on the left, and the many pieces in the third and
fourth zones lead to the structure on the right. This representa-
tion ignores the spin orbit splitting across the hexagonal face.
(After W. Harrison. Phyl. Rev. 118, 1190 (1960).)

PROBLEMS

I. Nearly Free Electron Fermi Suiface Near a Single Bragg Pla11e

To investigate the nearly free electron band structure given by (9.26) near a Bragg plane, it is
convenient to measure the wave vector q with respect t O the point JK on the Bragg plane. If
we write q = !K + k, and resolve k into its components parallel (k 11 ) and perpendicular
170 Chapter 9 Electtons in a Weak Periodic P otetltial

to K, then (926) becomes

hl kl +- (4&0.--z 2m
2m
hl k
H
2 + IV ·12)1/l
"
(9.36)

II is also convenielll to measure the fermi energy CF with respect to the lowest value assumed
by either of the bands given by (9.36) in the Bragg plane, writing:

(9.37)
so that when A < 0, no Fermi surface intersects the Bragg plane.
(a) Show that when 0 < tJ. < 2jUKI• the Fermi surface lies entirely in the lower band and
intersects the Bragg plane in a circle of radius

P=vv· (9.38)

(b) Show that if A > j2U"I' the Fermi surface lies in both bands, cutting the Bragg plane
in two circles of radii p 1 and (Jz (Figure 9.6). and that the difTcrcnce in the areas of the two
circles is

(9.39)

(fhe area of these circles can be measured directly in some metals through the de Haas-van
Alphen effect (Chapter 14), and therefore jU1 1can be determined directly from experiment for •

such nearly free c!cctron metals.)

2. Density of Lfl•els for a Twt1-Band Model

To some extent this problem is artificial in that the effi::cts of neglected Bragg planes can lead
to corrections comparable to the deviations we shall find here from the free electron result. On
the other hand. the problem is iustructive in that the qualitative features are general.
If we resolve q into its components parallel (q 11 ) and perpendicular (tU.) to K, then (9.26)
becomes

(9.40)
where

(9.41)

IS only a function of q 0• The density of levels can be evaluated from (8.57) by performing the
integral in an appropriate primitive cell over wave vectors q in cylindrical ooordinates with the
z-axis along K.
(a) Show that when the mtegral over 11 is performed, the result for each band is

g(&) -
_4nl1(2m)
h2 ( mo
qu - qll
min)
• (9.42)

\\here. for each band. qi" and qjj1" arc the solutions to & = hi:(q1). Verify that the familiar free
electron result is obtained 10 the limit 0.
 Problems 171

g(Q Figure 9.13

Oects1ty of levels in the two-band approxima-
1(7 I tion. The dashed line 1s the free electron result
I
Eq. (2.63). Note that in contrast to earlier
I
I I figures in this chapter, th1s one explicitly shows
II I second-order corrections to Lhe free electron
I result far from the Brdgg plane.
7 I
rr I I
'/
'I L 21UI(I_J
&
&k/ 2

(b) Show that

qf" = - n + O(U,. 2 ), (& > 0), ,..,... -

'1 11 -
I
2
K (9.43)

for the lower band, if the constant energy surface (at energy E) cuts the zone plane (thai is,
Ei;1- IUKI E ,.;; ei;1 + IU.J).
(c) Show that for the upper band (9.42) should be interpreted as giving a dens1ty of levels

g+ (e)= 4rr21 (2m) (qu ft 2

max
- -zK),
1
(9.44)

(d) Show that dgfde is singular ate = l'"' 2 ± IV..!, so that the density oflevels has the form
shown in Figure 9.13. (These singularities are not peculiar either to the weak potecttial or two-
band approximations. See page 145.)
3. Effect of Weak Periodic Potential at Places ink-Space Where Bragg Planes 1\1eet
Consrder the point W (kw = (2nf a)(l, f, 0)) in the Brillouin zone of the fcc structure shoWTI (see
Figure 9.14). Here three Bragg planes ((200), (lll), (lll)) meet, and according.ly the free electron
energies

eo = fz:z k2
1
2m '
1
0t 12 ( 2n )
&2 = 2m k --;; (1, 1, 1) '

= ( k - (1, 1,1) y,
2m (k - 2na (2, 0. 0))1
= ft1 (9.45)

are degenerate when k = kw. and equal to ew = !Jlkw2/ 2m.

Figure 9.14
First Brillouin zone For a face-<:cn tcrcd cubic crystal
(Ill)
u
172 Chapter 9 £1ectrons in a Weak Periodic Potential

(a) Show that in a region of k-space near \V, the first-order energies are given by solutions to15

ey-e v. u. Uz
u. e Vz u. - 0
u. Uz eg - 8 u.
u2 UJ u. e2- e
U 117, and that at W the roots arc

(9A6)
(b) Using a similar method, show that the energies at the point U (ku = (2nj a)(l , {, !;) ) are

e= Su- Vz. S = Su + -!U 2 + !(U 2 2 + 8U/) 112 , (9.47)

where Bv = IJ 2 ku 2 /2m.
4. Alternative Definition of Brillouin Zones
Let k be a point in reciprocal space. Suppose spheres of radius k are drawn about every reciprocal
lattice point K except the origin. Show that if k is in the interior of n - 1 spheres, 211d on the
surface of none. then it lies in the imerior of the nth Brillouin zone. Show that if k is in the
interior of t1 - 1 spheres and on the surface of m additional spheres, then it is a point common
to the boundaries of the (n -t- J)th, ... , (n + m)th Brillouin ?.Ones.

5. Brillouin Zones in a Two-Dimensional Square Latrice

Consider a two-dimensional square lattice with lattice constant a.
(a) Write down . in units of 2n1 u. the radius of a circle that can ilccommodate m free elec-
trons per primitive cell. Construct a table listing which of the first seven 7.ones of t he square
lattice (Figure 9.15a) arc completely full. which arc partially empty, and which are completely
empty form = I, 2, ... , 12. Verify that if m 12. the occupied levels Iic entirely within the
first seven zones, and that when m ;;. 13. levels in the eighth and higher zones become occupied.
(b) Draw pictures in suitable primitive cells of all branches of the Fermi surface lor the cases
m = 1, 2, ... , 7. The third zone surface for m = 4, for example, CaJ1 be displayed as in Figure
9 . 1Sb.

lS
Assume tJ1at the periodic po tential U has inversio n symmetry so that the UK are real.
 Probii?IDS 173

•I

I
/ F.gure 9.15
(a) The firsl seven Bril!omn
zones for the lwo-d amensional
I
I
square lauicc. Because or lhe
I Ialli<:\! symmclry it is necessary
10 display only one eighth of
the figure. The rcsl can be
rcconslructcd by reflection in
the dashed lines (which arc not

/I
(a)
--- . zone (b) Fermi
surface in the third zone for a
square lauice w1th four clec-
rrons per lUll! cell (The scale
in (b) has been expanded con-
siderably.)

• (b)
 Problems 241

5. Smce the correction to the hlg)J-field result ( 12.51) is of order H 2 , the general form for the
current indut:ed in an electron band with closed orbits is

nee - 1
j = - - (Ex II)+ CJ1l·E, (12.75)
f1
where
lim < cc. (12.76)
11-m
Show tl1at the high-field magnetoresistance 1s given by

- 1 ,. 2 (I)
( 12.77)
p - ( )2 1m H Gyy>
nee H- <>=

where the y-axis is perpendkular to the magnetic field and the direc.1ion of current flow. Note
that Eq. (12.76) requires the magnetorcsistance to sawrare.
6. The validity of the semiclassical result k(t) = k(O) -eEtfh for an electron in a uniform
etectrit: field is strongly supported by the following theorem (which also provides a useful starting
point for a rigorous theory oJ electric breakdown):
Consider the time-dependent SchrOdinger equation for an electron in a periodic potential
U(r) and a uniform electric field:

(12.78)

• Suppose that at timet = 0. 1/J(r. 0) is a linear combination of Bloch levels, all of which have the
same wave \'ector k. Then at time c. 1/t(r, t) will be a linear combination of Bloch all of
which have the wave vector k - eEt/1•.
Prove this theorem by nocing that the formal solution to the SchrOdinger equation is
(12.79)
and by expressing the assumed property of the imtiallevel and the property to be proved of the
final level in terms of the effect on the wave function of translations through Bravais lattice
vectors.
7. !a) Does the orbit in Figure 12.7 enclose occupied or unoccupied levels?
(b) Do the closed orbits in Figure 12.10 enclose occupied or unoccupied levels?

.... The semiclassical theory of an electron in a uniform clccuic field is not exact in spite ofUtis theorem,
because tbc coefficients in the linear combination of Bloch levels wiU in general depend on lime; th us
interband transitions may occur.
 Problems 259

SE:\!ICLASSICAL CONDUCTIVITY IN A UNIFORM MAGNETIC

FJELD
The DC electrical conductjvity at uniform temperature in a uniform magnetic field
H can be cast in a fonn qwte similar to the H = 0 result (13.25). In a magnetic field
v(k{t')) does depend on t ', and the integral occurring in the nonequilibrium distribu-
tion function ( 13.21) can no longer be explicitly evaluated in the general case. Instead,
the zero-field result (13.25) must be replaced by

(13.69)

where vn(k) is a weighted average of the velocity over the past history of the electron
orbit 42 passing through k :
o dt
v,(k) =
I - oo -r,(k)
e•!•,.lklv,(k,(t) ). (13.70)

In the low-field limit the orbit is traversed very slowly, only points in the immediate
vicinity ofk contribute appreciably to the average in (13.70), and the zero field result
is recovered. In the general case, and even in the high field limit, one must resort to
some ra ther elaborate analysis, even to extract the information we arrived at in
Chapter 12 from a direct examination of the semiclassical equations of motion. We
shall not pursue such computations further here, but some of the applications of
(13.70) are illustrated in Problem 6.

PROBLEMS

1. On page 250 we argued that in a metal with cubic symmetry the tensor is a
constant times the unit matrix, i.e., that j is always parallel to E. Construct an analogous argu-
ment for a hexagonal close-packed metal. showing that the conductivity tensor is diagonal in a
rectangular coordjnate system with z taken along the c-axis, ·with CTxx = cr,Y, so that the current
induced by a field parallel or perpendicular to the c-axis is parallel l.o the field.

2. Deduce from (13.25) that at T = 0 (and hence to an excellent approximation at any T « TF}
the conductivity of a band with cubic symmetry is given by

(13.71)

41
Here k.(r) is the solution to the semiclassical equations of motion (12.6) in a uniform magnetic
field that passes through the point kat time zero (k.(O) = k). (We have used tl1e fact tl1at the distribution
function is independent of time when the 6clds are time-independent, and wriuen the integral in (13.::! l)
in the fonn it has at l = 0.)
260 Chapter 13 The SemklassicaJ Theory or Conductioo in Metals

where Sis the area of Fermi surface in the band, and li the electronic speed averaged over the
Fermi surface:

v = f dS jv(k)j . (13.72)

(Note that this contain!\, as a special case. the fact that filled or empty bands (neither of which
have any Fermi surface) carry no t:urrent. Jt also provides an ah ernative way of viewing the fact
that almost empty (few electrons) and almost fiUed (few boles) bands have low condut:tivity,
since they will have very small amounts of Fermi surface.)
Verify that (13.7 I) reduces to the Drude result in the free electron limit.
3. Show that the equations desc.ribiog the electric and thermal currents, (1 3.45) a tld (13.50)
to (13.53), continue to hold in the presence of a uoifo rm magnetic field, provided that Eq. (1 3.48)
for o(E) is generalued to include the effects of the magnetic field by replacing the second \ "(k)
defined in E.q ( 13.70).
4. T he response of the condlH--tion electrons to an electric field
E(r, t) = Re [E(q, (13.73)
which depends oo positioo as well as time, requi res some special consideration. Such a field will
in general induce a spatially varying charge density

p(r , r) = - e bn(r , l),

b11(r, t) = Re [ b1(q, w)erlq· • t.Jfl). (13.74)
Since electrons a re conserved in collisions, the local equiljbrium distribution appearing in t he
relaxation-time approximation (13.3) must correspond to a density equal to the actllll.l inStanta-
local density n(r. c). Thus even at umform temperature one must a llow for a local chemical
poteolial of the form
p.(r, r) = p. + t),
bJl(r , t) = Rc [ bp.(q, w)e'1q · ' "'''] , (13.75)
where bp(q, co) is chosen to satisfy (to linear o rder in E ) the condition

t5n(q, co) = -ong11(Jl) bp.(q, w ). (13.76)

(a) Show, as a result, that at uniform temperature E.q. (13.22) must be replaced by"l
g(r, k, t) = f(C(k)) -1 Re (bg(q. k, w)e''q · •- ""'] ,
( w) = ( - c[\ w)/ r) - ev(k ) • E(q, w) (13.77)
g q. iJe) (1 /'r) - i(c.o - q · ,·(k}) ·
(b) By const ructing the induced current and charge deosities from the distribuuon function
( 13.77). show that the chojce (13.75) of CJI(q . w) is precisely what 1s required to insu re that the
equation o f continuity (local charge conservation)

q · j(q . w) = wp(q. w) (13.78)

IS satisfied

.., See foolnOIC 7

 Prohlems 261

(c) Show that if no charge density is induced, then l.:be current is

j(r. t) = Re [ a{q. w) • E(q, w)e'!q · r-u••l] .

cr(q w) = el
.
k. (-_of\-:---,.::--'
f -d_
4n3
-'''- ---=
ae) (1 /T) - i[w - q . v(k)J
(13.79)

Show that a sufficient condition for (13. 79) to be valid is that the electric field E be perpendicular
to a plane of mirror symmetry in which the wave vector q lies.

5. Consider a metal in which thermal and electric currents flow The rate at
which heal is generated in a unit volume is related to the local energy and number densities by -
(cr. (13.J9J ):
_dq __du _, _du (13.80)
dt - dt r dt'
where JL is the local chemical potential Using the equation of continuity,

dll .
- = - V•j" (13.81)
dt '
and the fact that the rate of change of the local energy density is determined by the rate at which
electrons catry energy imo the volume plus the rate at which the electric field does work,

du
dt = -V . y + E . j, (13.82)
•
show that ( 13.80) can be written in the form

(13.83)

where jq is the thermal cunent (given by (13.38) and (13.40) ), and t = E + (I /e)V Jl. Assuming
cubic symmetry, so that the tensors Lij are diago nal, show that under conditions of uniform
current flow (V · j = 0) and uniform temperature gradient (V 2 T = 0) that

dq .
- = pJ 2 + -dK (VT)
. 2
-
dQ
T-(VT) · J
.
(13.84)
dt dT dT
where p is the resistivity, K is the thermal conductivity, and Q is the thermopower. By measuring
the change in bulk heating as the cUJ·rent direction is reversed for fixed temperature gradient
(known as the Thomson effect) one can therefore determine tbe temperature derivative of the
thermopower, and thereby compute the value of Qat high temperatures. given its low-temperature
value.
Compare the numerical coefficient of VT· j with that of the crude estimate in Problem 3,
Chapter 1.

6. The average velocity v (Eq. (13.70)) appearing in the expression ( 13.69) for the conductivity
in a uniform magnetic field takes on a li!.irly simple form in the high-field limit
(a) Show that for a closed orbit. the projec:tion ofV in a plane perpendicular to H is

IJc H
-v... = - eH-r - x (k - ( k.)L + 0 ( H2'
I ) (13.85)
262 Chapter- 13 The Seruidasskal Theory or Conduction in l\letals

where ( k ) is the time average of the wave vector over the or-bit:

(k) = dt. (13.86)

(b) Show that for an open orbit the high-field limit of Vis just the average velocity of motion
along the orbit (and hence parallel to the direction or the orbit).
(c) Show.... that in the lligh"field limit, when E · H = 0,

h = -e f (- k · w, (13.87)

where w = c(E x H)/H 1 is the dnfl velocity defined in (12.46). Deduce the forms (12.51) or
{12.52) from (13.87), depending on whether the band is particle- or bole-like. (Noce: Because k
is 11ot a periodic function in k-space, one cannot automatiL"lllly integrate by parts in (13.1!7).)
(d) Deduce, from the result of(b), the limiting form {12.56) for the conductivity in the presence
of open orbits. (Hint : Observe that vis independent of the component of k parallel to the k-space
direction of the open orbiL)
(el Show from the general form of the semiclassical equation of motion in a magnetic field
(12.6) that the c.:oodUl.1ivity tensor (1 3.69) for a given band in a uniform magnetic field bas the
functional dependence on R and 'C of the form :
a = tF(Hr). (13.88)
Dedut:e from (13.88) that when the current is carried by electrons in a single band (or if the
relaxation time is the same for all bands), then

P:u(H) - Px.x(O)
(13.89)
px.x(O)
depends on Hand • only through the product H< (Kohler's rule), for any diagonal componeut of
the resistivity perpendicular to H.
(f) Deduce from the properties of the semiclassical equations of motion in a magnetic field
that
= a.,( - H). (13.90)
This is known as a11 Onsager relation.45 (Hirzt: Make the change of variables k(r) = k', and
appeal to Liouville's theorem in replat:ing the k-space integrals in (13.69) by it1tegrals over k'.)

44
Argue that the tem1 in ( k ) in (13.85) makes no contribution because it depends only one and/<..
•• Such relationships between transport coefficients were first formulated in very great generality by
L. Onsager. The first equality io Eq. (13.51) is another example of an Onsager relatJon.
586 Chapter 28 Homogeneous SemicondudDrs

(b) Verify that the positions of the electron resonances in Figure 2!l.9b are consistent with
the electron efft'Crive masst'S given for silicon on page:: 569 and the formulas, (28.6) and (28.8),
for the resonance frequency.
(c) Repeat (a) for the resonance in germanium (Figure 28.9a). noting that Figure 28.7 shows
four electron pockets.
(d) Verify that the positions of the electron resonances in Figure 28.9a are consistc:nt with
the electron effective masses given for gennanium on page 569.
3. Level Density for Ellipsoidal Pockets
(a) Show that the contribution o f an ellipsoidal pocket ofclectwns to the conduction band
density of levels g<(G), is given by (d/df.)h(t ), where lr(S) is the number of levels per unit volume in
the pocket with energies less than e.
(b) Show, similarly, that the contribution of an ellipsoidal pocket of holes to the valence
band density of levels g,je ) is given by (dj df.)ll(e), where fl(t ) is the number of electronic lc\•els per
unit volume in the pocket with energies greater than e.
(c) Using the fact that a volume Q of k-space contains Of4n 3 electronic levels per cubic centi-
meter and the formula V = (4nf3)abc for the volume of the ellipsoid x 2 /a 2 + y 2 f b2 + z 2f c2 = 1,
show that formulas (28.14) follow directly from (a) and (b), when the conduction (or valence) band
has a single ellipsoidal pocket.
4. S tati.stics of Donor Le1•els
(a) Show that if the energy of a doubly occupied donor level is taken to be 2fv + 11. then
Eq. (28.321 must be replaced by

(28.43)

(b) Verify tha t Eq. (28.43) reduces to (28.32) as 6 -. oo, and that it reduces to the expected
result for independent electrons as 6 -. 0.
(c) Consider a donor impurity with many bound electronic orbital levels, with energies &1•
Assuming that the electron-electron Coulomb repulsion prohibits more than a single electron
from being bound to the impurity, show that the appropriate generalization of (28.32) is
Nd (28.44)

Indicate how (if at all) this alters the results described on pagt'S 582-584.
5. Constraint on Carrier Densities i11 p-Type Semico11ductol·s
Describe the electronic configuration of a doped semiconductor as T-+ 0, when N. > N 4•
Explain why (28.35) (derived in the tex t when N 4 N.) con tinues to give a correct constraint on
the electron and hole densities at nonzero temperatures. when Nc > N.,.
6. Can·ier S tatistics in Doped Semiconductors at Low Tempe•·atures
Consider a doped semjconductor with N 4 > N •. Assume that the nondegeneracy condition
(28.10) holds, but that (N4 - N.)/n1 is so large that (28.39) does not necessarily yield a value of p.
compatible with (28.36).
(a) Show under these conditions that p, is negligible compared with n., and p. is negligible
compared with N •• so that the chemical potential is given by the quadratic equation

Nce - - PI = Nd - Nu - ---=-__,-=-N
,...,.,.tl_ -,-- (28.45)
!elltr.r 111 + 1