PH YSI CAL REVIEW VOLUM E 100, NUM B ER 2 OCT OB ER 15, 1955

Spin-Orbit Coupling Effects in Zinc Blende Structures*

G. DRESSELHAUst
Departmesss of Physics, Umioersity of California, Berkeley, Calcforrsia
(Received June 30, 1955)
Character tables for the "group of the wave vector" at certain points of symmetry in the Brillouin zone
are given. The additional degeneracies due to time reversal symmetry are indicated. The form of energy
vz wave vector at these points of
symmetry
is derived. A possible reason for the complications which
may make a simple effective mass concept invalid for some crystals of this type structure will be presented.
HE
e6'ect of symmetry on the energy band
structures of crystals of the zinc blende type
can
be readily derived using the machinery developed by
Bouck.aert, Smoluchowski, and Wigner' and Elliott.
'
Recent extensive studies of the semiconductor proper-
ties' of InSb, which has the zinc blende structure, and
preliminary cyclotron resonance investigations' have
indicated a need for a more thorough understanding of
the possible energy band structures of a zinc blende
type crystal.
A zinc blende structure consists of two interpene-
trating face centered cubic lattices; each f.c.c. may be
considered a sublattice. The two sublattices are dis-
placed by
one quarter of a body diagonal and each
consists entirely of one species of atom. If the two sub-
lattices are identical, one has a diamond structure. The
symmetry properties of diamond are fully discussed in
reference 2.
FxG. 1. The erst Brillouin zone for a face centered cubic,
diamond, and zinc blende structure. Points and lines of symmetry
are indicated using the notation of reference 1.
*This work has been supported in part by the Office of
Naval Research and the U. S. Signal Corps.
t
Now at the Institute for the Study of Metals, University of
Chicago, Chicago, Illinois.
'Bouckaert, Smoluchowski, and Wigner, Phys. Rev. 50, 58
(1936).
s
R. J. Elliott, Phys. Rev. 96, 280 (1954).
e
H. Welker, Z. Naturforsch. 7a, 744 (1952); 8a, 248 (1953);
M. Tanenbautn and H. B. Briggs, Phys. Rev. 91, 1561 (1953);
G. L. Pearson and M. Tanenbaum, Phys. Rev. 90, 153 1953);
M. Tanenbaum and J. P. Maita, Phys. Rev. 91, 1009 1953);
H. Weiss, Z. Naturforsch. 8a, 463 (1953); O. Madelung and
H. Weiss, Z. Naturforsch. 9a, 527 (1954).
4Dresselhaus, Kip, Kittel, and Wagoner, Phys. Rev. 98, 556
(1955).
The zinc blende structure has the space group sym-
metry F43m or T&'. There are no glide planes or screw
axes, so the group of any wave vector k has only simple
operations. The 6rst Brillouin zone is the well-known
truncated octahedron shown in Fig. 1. The character
tables for the group of the wave vector k for certain
points of symmetry in the Brillouin zone are given in
Tables I through VI. When the spin is included in the
problem only the double representations occur (i.e.
,
representations for which a
360'
rotation, E, changes
the sign of the wave function). The effect of including
spin in the problem is to form wave functions of a
spatial function times a spin function which will
transform' as D;. The total wave function will then
transform as the direct product of a single group repre-
sentation with D;. This direct product then can be
decomposed in terms of representations of the double
group. If more than one representation of the double
group occurs in the decomposition of the direct product,
a spin-orbit splitting of the level is indicated. A table
of the direct products of the single group representa-
tions with
Dg
is included with each character table.
The compatibility relations for certain lines of sym-
metry are given in Table VII. These relations give the
splitting of the degeneracies as one proceeds along the
symmetry axes. The extra degeneracies due to time
reversal symmetry can be found using the standard
test due to Herring' and Eliott.
'
The extra degeneracies
are indicated in each table.
The principal diGerence from the diamond structure
is the lack of inversion symmetry for the point groups
in the zinc blende structure. Without inversion sym-
metry one still has the result from Kramers' theorem'
that E(k) =8(

k),
but now the periodic part of the
Bloch functions no longer satisfies the condition
I q(r) =N~(

r),
and hence a twofold degeneracy
throughout the Brillouin zone is not required.
The one electron Schrodinger equation for the
'
E. Wigner, Grlppeatheorie U. W. Edwards, Michigan, 1944),
p. 245.
'
C. Herring, Phys. Rev. 52, 361 (1937).
~
This theorem states that in the absence of magnetic 6elds +lr
and i0.%'1,
*
are solutions of the Hamiltonian for the same energy.
The second solution belongs to wave vector

k, and hence we
have two solutions at k and

k with the same energy.
SPIN

ORB IT EFFECTS IN Zn BLENDE STRUCTURES
TAsLz I. Character table of the double group of r;
k=
(000).
r,
Fs
r4(x, y,z)
F5
r6
rv
rs
1
1
2
3
3

2
6C4B
1
1
2

1
0
0
0
1
1

1
0
0
1
1

1
8CB
1
1

1
0
0

1
1
6I XC4
1

1
0

1
1

W2
0
6I XC4
1

1
0

1
1

v2
v2
0
j.2I XCB
1

1
0
1

1
0
0
0
F;
F;XD)
Fy
F6
r,
Fv
Fs
rs
r,
Fv+Fs
r,
F6+Fs
Selection rules
r;
r;xr
r,
F4
F2
r,
F3
r, +r,
F4
r, +r, +r, +r,
r,
r, +r, +F,yr,
F6
Fv+Fs
r,
F6+Fs
Fs
F6+Fv+2F
problem with spin-orbit coupling is
p'
+V+ (vVXp)
~ p~=Ea+~
2m
4m'c'
The translational symmetry of the lattice requires that
the wave functions be of the Bloch form, i.e.
,
%g

Ng(r)e'~'
Degenerate levels are treated
by solving the customary
secular determinants.
In working out the matrix elements for perturbation
theory, it is helpful to use group theoretical selection
rules. Due to the scalar character of the Hamiltonian,
TABLE II. Character table of the double group of 6;
k= kL100].
where N(r)
is periodic and satisfies the equation
P
+V+ (vVXp)
~

2m
4m'c'
ag(x)
~g (y+z)
a4(y

z)
1 1
1 1
1 1
1 1
2

2
2C4B
1

1
0
2I XC2
1

1
1
0
2I XC2'
1

1
1

1
0
f
p
5
l p
fez/z'l
+
I

+ ~xvv I=I E.

(m
4~'c'
) E 2~)
The equation for k+K is
p2
+V+ (vVXp) o+K
. 2m
4m'c'
f
p
+fzk
I

+
Em
4m'c'
&;XD)=~5, i=
1, 2, 3, 4
53 and A4 are degenerate by time reversal
Selection rules
6;XAy
~sX~B ~3
b.;Xh4 A4
A4
b, 4
TABLE III. Character table of the double group of A or L;
Ir=
(k/v3)(111] or (z-/a)(111).
3I XC2 3I XCB A, L
ply'
+5K
I

+ ~xvv I+K
&m
4~zc' )
A&(x+y+z)
A2
(x+cuy+aPz)
A 3 (x+~'y+cuz)
(~'=1)
I+K (3) ~4
2m ) +s
A6

I
E~+K
Treating the term
L4 and L~ are
(A4 and
A;
AsXDy
Select&on rules
A; AI Ag
A;XAz
A, XA&
t' p
K'=5K
I

+ axvv
4m'c' )
as a perturbation, the energy at k+K for a nonde-
generate level is
52)I|2
Ei+K Eg+
.
+(pgI K I
Ng)+
2m
(4)
E 2CB 2CB
1 1 1
1 1 1
2

1
1 1

1

1

1

1 1 z

1

1 1 z

2 1

1 0 0
degenerate by time reversal
A5 are nondegenerate)
Ag A2 AB
A6 As A4+A5+A6
AB A4
As A4
Ay+Ay+As A6
A5 As
A. 5 A6
As A4+A. g+A6
G. DRESSELHAUS
Z, K
(s) (*+y)
Zs(x

y)
Zs
Z4
IXC&
ZsXD(=Zs+Z4, i=
1, 2
Selection rules
Z; Zg
ZsXZz Zz
ZsXZ2 Zg
Zg Zs Z4
Z2 Zs Z4
Z1 Z4 Zs
I XC2
1

1
2
TABLE V. Character table of the double group of X;
h=
(2s/a)(100).
TABLE IV. Character table of the double group of Z or E;
lt =(k/V2)$110j or (34r/2a)(110).
is the irreducible representation of wave vector k
according to which 4'k'
transforms and FR is the repre-
sentation according to which a vector transforms.
Another way of viewing the selection rules is that the
only representations that mix with F; under the per-
turbation BC'
are those contained in the decomposition
of the direct product F;)&FR. Tables I to VII also give
the decomposition of these direct products for the
points of symmetry in the Brillouin zone. For conveni-
ence the combinations of vector components which
transform as a given irreducible representation are
indicated in the character tables.
In order to give a more complete treatment of the
point F, the bases shown in Table VIII may be selected
for the irreducible representations.
'
In this notation
the spin-orbit splittings at k=0
are, for F41

F4'*,
3)z /' BV BV
AEso zl B,

p
p.
5,
l,
4m'c'
4 Bx
By )
4C4~~ 2C4~1) 2I &(C4)) 2I XC411 4I XC2
1

1
1

1
0
0
0
1 1 1 1
1 1

1

1

1 1

1

1

1 1 1 1
0 -2
0 0
o o m

v2
0 0

v2 V2
Xg
X2
Xz(x)
X4
Xz(y, s)
X6
X7
1
1 1
1 1
1 1
2 2
2

2
2 -2
and for I"5&
F5&,
35 BV BV
++Be=
zl el
pw p~ es
4m'c'
Bx
By Xg Xp Xs X4 Xs
Xs X7 X'7 Xs Xs+X7
X;
KXD)
Selection rules
X; Xg X2 Xs X4
XXX, X, X4 X1 X~
X;XX5 X5 X5 X5 X5
Xs X6 X7
X5 X7 Xs
X1+X2+X3+X4 Xs+X7 Xs+X7
TABLE Vl. Character table of the double group of 5';
h
=
(24r/a) (0
,
'
1).
I XC4'
1

Z
z

i/i
C42 C4 I XC4 I XC4 I XC4'
1

1
i+i-
+i
iv'i
Qi
1

1
z
z
4/i
i/i-

+i
1
1

1
1 1
1
1
1 1
1

1
1

1
1
1

1
W1
Wz(x)
Wz(y+is}
W4(y

js)
8'5
W7
Ws

Z
z
gi
iv'i
z

$
Z

Z
'b
z
z
TABLE VII Compatibility relations
Selection rules suggest that the representations F4
and F5 have first-order matrix elements with
3"
. and
hence finite slopes. Actually all 6rst-order matrix
elements with
p
vanish due to time reversal symmetry.
For example, (Bzl
pl
Bz)
=
(8sl plBt) by a reflection in the
(101) plane, but
by partial integration
(8t
l pl
8s)
=
(Bs
l pl Bz), as the 8,
's
are real if the Hamiltonian has
time-inversion
symmetry; hence all such matrix ele-
ments vanish. This argument only holds if all 5 s are
from the same degenerate F4 level.
When spin-orbit interaction is included in the per-
turbation (i.e.
,
the oXVV term is not neglected), the
energy to erst order in k for a F8 level is given by the
W;
W'XD-;
Selection rules
W;
W'X W2
W XWs
W;XW4
the term
Ws+W6
Wg
W7+Ws
Ws
W6+W7
W4
Ws+ Ws
Ws W7
Ws W5
W5 Ws
W7 Ws
5(
fi
R=

l
p+
4rxv'U
m ( 4mc'
Wg Wp Ws W4 Ws Ws
W2 W1 W4 Ws W7 W6
W4 W~ Wr Ws W7
W4 Ws W1
8'g
W5
F1~Q1
F2

+
Q2
Fs
~
~1+~2
P4 ~411+(413+~4)
pz
~
as+(&4+&4)
F6, F7, Fs~ ~5
F1

+
Ai
F2

+ A.g
Fs

+
As
F4

+
41+As
Fs

+
A~+A. s
FF7~X,
Fs ~
~4+&5+As
X1~A1 (L4+Lz)
~
A4+Az
X2~ h. ~
Xs~h1
X4

+
h2
X, (~,+~,}
Xs, X7~ a5
Time reversal degenerate representations are indicated by paren-
theses.
transforms as a vector. Matrix elements of the
type
(+"I&-le") will vanish unless the direct product
4
See p v& r Laze end H. A. Bethe, Phys. Rev. 71, 612
F,XFaXF; contains the unit representation, where F; (1942}.
SPI N

ORBIT EFFECTS IN Zn BLEN DE STRUCTURES
secular determinant

X (i/2)Ck+

(i/2)Ck

X

Ck, (iV3/2)Ck
(i'/2)Ck+ Ck,

Ck,

(iV3/2)Ck+

(i/2)Ck

(i'/2)Ck
Ck,
(i/2)Ck,
where X=E'

(h'ks/2m), which has solutions

X=
~C(k'yL3(k
'k 'yk 'k '+k 'k ')]&)-*'
'=~C(ks

L3(k
sk syk sk s+k sk s)]-:):
where
1 5s) BU
for r,&,
2~pm'c'E
By )
1ks( BU
2%3m'cs E
By
C=O, for F3.
The first-order energies are shown in Fig. 2 for the
L100], [111],
and 1-110]directions. In polar coordinates,
k, =ksin8cosp, k=ksin8sing, k, =kcos8, Eqs. (8)
become
X=~CkL1~(3)&sin8(cosi8+r~ sin'8 sins2&)&]&.
(10)
Figure 3 shows the four 6rst-order energies given by
Eq. (10)
for constant k plotted against the angle 8 for
wave vectors in a (110) plane (@=sr/4). It is interesting
to note that the two-dimensional A.
6
level is the inter-
section of surfaces which arise from two diferent repre-
sentations in a [100] direction. Any large separation
of the energy levels which transform as F8 at the center
ENERGY
of the zone, as one proceeds toward the edge of the zone,
then would entail a rather complex band structure with
highly deformed energy surfaces and accidental de-
generacies.
'
Second-order perturbation theory for the F8 level is
tractable on the approximation that only first-order
terms in e&VV and second-order terms in
y
be con-
sidered. The fourth-order equation which results from
the 4)&4 secular determinant is
(L
M l'
Es
(L

M)'
y'

2y'
I
3
)('(k sk 2+k
sk
2+k sk 2)+Csks
L

Mq'
/4yCsg(ksksyksks+ksks)+
~ ~
k4
3
E'
(L,

M)'
+ (k.
'kys+k'k, '+k,'k, ')
3
+C'Ek' 3(k'k'+k'k'+k'k')]
L M's-
+2~
~
C'(k, '+k '+k ')
)
/Ll' 2

3~
~
+

S' C'k'(k 'k '+k 'kg'+kg'k-')

3 ) 3
pL
M'
+21~ ~
C'k'k'k'=0,
(11)
)
where I,
,
3f, and S are real numbers and can be ex-
pressed in terms of sums over the squares of the
Fn. 2. Plot of energy
es wave vector showing
the first-order energy for
the spin-orbit split F4
or I's level in $100$,
$110), and (111$direc-
tions. The circled num-
bers indicate the dimen-
sion of the representa-
tion.
P7 Qi
Qi g4
{oooj
(ooo)
k[IIO]
k
goo]
(ooo)
Fxo. 3. Cylindrical cross section around the point F of a plot of
energy ss wave vector for wave vectors in a (110) plane.
C. Herring, Phys. Rev. 52, 365 (1937).
G. D RESSEL HA US
TAsI.E VIII. Basis functions for the irreducible representations'
of the double group for the point l .
r~=1
rs, dS
=
Lx'(ys

z')
+yc(z'

x')
+z'(*'

y')7
Fg, 'y1=x +Glg+GPS
y2
=
x'+any'+coo
cd= 1
rs, n

}
0. +)
r Pl-)
lil+)
1
r.(r),

('v+
~)l+)
v2
(~ 'Ys ~71) I )
V2

(~'vs

~v~) I+&
v2
Pro. 4. A cylindrical cross section around the point
I"
of a plot
of energy vs wave vector for wave vectors in a (110) plane
showing the third order splitting of the F6 or I'7 surface.
14, bg=x
82=y
53=8

(~'vs+~v ~) I

)
v2
1
rv('),

I:

i(si

sos) I+&+Ssl

&7
V3
1
I
~(s~ +ss.) I )

I+
)7
VS
r, (r,&},

(s,

Q, ) I

)
W2
1

-I i(a,

sb,
) I+&+2s3I

&7
6
1
--
E~(s~+sss)
I
&+I
l+&7
6
it=
%I 1007 the energy eigenvalues are
)L+2Mq
'
(L

Mq
'
+k[100l
= u'+'
I I
u'+
I I
fs4+c
2m E 3 J,k 3
where each root is double; and for k=
(k/V3)I 1117
(L+2M
)
~[sy~sl [irtl
=
&
+
I
Ik'
-I~vl
(double)
(12)
r,
,
e, =x(ys

z')
es
=y(z*-x')
es

s(x'

y')
z

(~i+fss) I+&
V2
1
rs(re), p
s(eq

its)

I+)+t3I )7
&3

Ls(ei+ses)l

)

s3I+)7
v3
rs(rat),

(e&

s)
I

)
v2
b(e~

s~s)l+)+2eal

&7
6
Ls(e~+s) I

&+2esl+&7
6

(~~+s) I+)
V2
+

lP+%2Ck (13)
3
O'
V2Ck.
An estimate of the magnitude of the constant C may
be obtained if one considers the zinc blende structure
as a deformed diamond
type
structure.
""
Here we
will consider
U'=
V

V; as a perturbation, where Vi
is a diamond
type potential with inversion symmetry
and V is the actual zinc blende type potential. The
correct 6rst-order spatial wave functions for a zinc
blende structure are then given by the following
equations.
Fol Py)
absolute value of matrix elements,
"
and
(~'+I
1"
l~+) (&' I
1"
l~+)
n=u++
Q n,
++
Q P',
' =I,
+
~~0

~; ' =~2 &O

&'
fi' t'L+2M
~
Iu+y.
3 )
+0 ~i i =~2 +0
(14)
For a general k, Eq. (11) has four real roots; for
io
Dresselhaus, Kip, and Kittel, Phys. Rev. 98, 368 (1955).
'
For a more detailed treatment see the paper by F. Herman
(to be published).
"R.H. Parrnenter, Phys. Rev. 100, 573 (1955}.
SPIN

ORBIT EFFECTS IN ZII BLEND E STRUCTURES
For I'2,
p
i =r2+
(-;-I
l"
I=)
n;
+
g
lattice constant and A~o is the spin-orbit splitting at
k=0. The first-order energy at the zone edge in a
[100]direction is
(-;-I
I"
IW)
e=~'+ Z;-+
E
jap
jv.
i =r2+
For I'3,
(v-; II"lv-
)
vx=vx++
2 QKi
(vx; I
&'Ivx+)
+ 2 +xi )
jvp
jv.
& =r12
(v;
II'lv
-)
vIc vK + VKi
(vx;

I
&'I vx-)
+ 2 %xi
~
(sx;+I
&'I ex+)
&z
Ep 8
For I'~,
(~-;-II"
l~
-)
4c=&x + 4c
;=r15-
EP

E;
(
'+Iv'ls
)
+ 2
ex'+,
i =rg5+ Ep

E.
(&x; I& le++)
5K=+
p
. ~Z4
(17)
2m ~so
E C

2s.
0.02 ev (in InSb).
8 Z
Hence the splitting due to the first-order terms would
be only of the order of 0.02 ev if the slope were linear
all the way to the edge of the zone. Actually the second-
order terms which should be large due to the small
energy gap
will turn the surfaces down very rapidly.
Using this value for C and the values of I.
, M, and X
from cyclotron resonance experiments on Ge,
'
it would
seem from Eq. (11)that near the center of the Brillouin
zone the removal of the twofold Kramers degeneracy
is at most 10
4
ev for holes with thermal energies.
Under these circumstances it seems quite likely that a
perturbation expansion about the extremum should
contain several orders of perturbation theory, and a
simple energy surface with effective mass tensor com-
ponents independent of wave vector would be a very
poor approximation.
For the I"6 and F7 level the energy to third order in k
is given by
E=C k'~C [k'(k
'k '+k 'k '+k 'k ')

9k
'k 'k ']'
(20)
In third order the levels are split in all but the [100]
and [111]directions. In polar coordinates Eq. (20)
reads
E=Cek'+Cik' sin8[1

sin'8
(1+2sin'2p)
+(9/4)
sin'2p sin48]l.
(21)
For F5,
(&x;+I
&'I &x+)
err =~K
+ ~z;+
gp g
4
(;-Iv'l~ +)
&Xi
)
p
(&rr;+I
&'I
ere
)
ex=ex
+
~z'+
i =rIfi jvp
jv.
(ex'
I
I"
I )
where the cubic harmonics are as defined in reference 8
except that Vt+
=Vs+*=x'+4ey'+to's', (oi'=
1),
and the &
superscript is used to denote the parity of the functions.
For the I'4 level arising from a diamond type
F+level
fz'
C=-
&3 tS C i =r&4
(et+I ~l'*/~yl34' )(~i'
I
l"
I
et+)
(19)
(Eo

E')
To a very rough approximation"
V'
P';/Z where Z
is the atomic number. C (1/Z)hsoa, where a is the
Figure 4 shows a plot of energy es 8 for constant k and
&=4r/4 [i.e.
,
k in a (110) plane]. The surfaces have
their maximum separation along &il0) axes. In the
III

V class of semiconducting compounds like InSb,

the high mobility electrons are presumably in a spheri-
cally symmetric
I'6
state.
'
Higher orders in k should
not enter until the thermal or Fermi energy for the
electrons is of the order of the band splittings at k=0.
In impure e-type InSb with electron concentrations
10"/cm' and a Fermi energy" of 0.2 ev one may
be entering into a region where the splitting of the
degeneracy for the electrons should be considered.
Perturbation expansions for other points in the
Brillouin zone are facilitated
by writing the perturbation
in a form such that the vector combinations indicated
in the character tables appear; these combinations are,
for the points I',
(000), and X, (24r/a) (100),
K'=
k~ +kR+k&.
;
(22)
for the point W, (24r/a) (O, s, 1),
3C' =
K+,+p[(K+iK,
)
(RiE,
)
+(KsK,
)
(Z+4m.
)];
(23)
'4
Hrostowslti, Wheatley, and Flood, Phys, Rev. 95, 1683 (1954).
G. DRESSELHAUS
TABLE IX. Possible energy extrema for the representations of the
double group in zinc blende structures.
X6
x
-0
7
f22
Cl
12
I
X
c9
CL
X7 -8
z p-. X6
LLI
0-
r
-0
6
rs
0
r
-0
7
Is
r,
L6
L6
( L4& Lg)
L6
L6
Extrema at
(000)
(k/VS) L111]
h.
Representations
of double group
r,
F7
Fa (from Fq)
Constant energy surfaces
sphere
sphere
2 warped energy surfaces as
for holes in Si and Ge
8 spheroids;
&11'1) axes
8 spheroids; &111)axes
0- I
6
(2s./a)(1,
xs 0) W Ws, Wo, W7, Ws 6 spheroids; (100) axes
2
/,
(IOO) (ooo) (ooo) /p
(I I I)
REDUCF D WAVE VECTQR K
General point 48 general ellipsoids
FIG. 5. A schematic drawing of the energy levels for a zinc
blende type structure modification of boron nitride based on
Herman's calculation for diamond PF. Herman, Phys. Rev. 88,
1210 (1952); thesis, Columbia Vniversity, 1953 (unpublished)].
The spin-orbit splittings are highly exaggerated for the purpose
of illustration. The levels marked with an 0 have zero slope along
that axis. This 6gure should be compared with Elliott's Fig. 2.
L'R.
J. Elliott, Phys. Rev. 96, 266 (1954).
g
for the points A, (k/~3(111), and L, (m/a)(111),
x'= ', $(E' +K+E,
) (R,+R+R,)
+(K,+toK+cesK,
) (R,+aPR, +(oR.
)
+(E,+tosE+toE,
) (
R,+ toR+c,
)i, (24)
and for the points Z, (k/v2) (110),
and 6, k(001)
5C'=
K&,+ts[(K,+-K)(R,+R)
+(K, K)(R,

R
)). (25)
Using these perturbations the Grst-order energies
can be written down at sight and are of the following
form: for hs, k(100),
Z=CsK +Cs(Ks+K,
')'*;
(26)
for A4 and As, (k/~3 (111),
E=
C, (E,+E+E,); (2&)
for h6,
E=
Cs(E,+E+E, )
+CsLEs

(EgK+KE, +E,K.))&) (28)

for Z4 or Zs, (k/~2(110),
Z=Cr(E, +E)+CsK,
;'
(29)
and for Xs or Xr, (2m/a) (100),
Z= &C (Es+Es,
')&.
The second-order energies for the point W,
(2s.
/u) (0,
ts,
1),
are of the form
Z=CtoE '+C (Ett'+E', ). s(31)
If the energy extremum is at the point 8', the constant
energy surfaces will be spheroids with &100&axes.
The irreducible representations A4 and A5 only have
slopes along a &111&axis, hence if the constant C4 is
zero at some point along the axis as it presumably is
g.par the ct:nter of the zone in InSb due to the perturbing
influence of the lowest conduction states (see Fig. 5),
then the constant energy surfaces at these extrema will
be spheroids with &111& axes. However, because of
the smallness of the first-order terms compared to the
second-order terms, it is not expected that these
extrema will be of any significance in the band structure
if they occur near the center of the zone.
The representation
I'8 at the center of the zone can
have zero slope if it arises from the two-dimensional
representation F3 of the single group. In this case the
energy to second order in k is given by
E=Ct2ks&LCtssk'+Ct4s(k, 'ks+k'k,s+k,
'k
')]l, (32)
in the limit that all spin-orbit splittings are negligible
compared with the spacing between levels at k=0, then
C~4'

3C~3'. In atomic Sn it is known that the 4d
levels overlap the Ss atomic levels, hence in grey Sn or
InSb it is conceivable that a
I"3
level, which can be
represented in a tight binding approximation by
d-orbitals, could be the uppermost valence band.
Table IX gives a tabulation of the types of energy
surfaces which might be expected at certain points in
the Brillouin zone. It should be emphasized in all these
considerations that if the spin-orbit interaction is small
or the difference in the crystal potential from the
diamondlike potential is only slight, then the region
of convergence of the energy expressions will be small
compared to kT and one should then consider only the
single group representations in the 6rst case, or the
diamond structure double group representations in the
latter case. If the region of convergence of the per-
turbation expansion is ~kT, then one is not justi6ed
in keeping only the lowest nonvanishing term and a
simple effective mass approximation would seem unjus-
ti6ed. From the order of magnitude estimate of the
spin-orbit splitting, it seems this might be the case for
holes in InSb.
I should like to thank Professor C. Kittel for sug-
gesting this problem and for guidance throughout the
course of the investigation. I am indebted to Dr. F.
Herman and Dr. R. Paramenter of R.C.A. Laboratories
for communicating", their results to me prior to pub-
lication. I also wish to thank Mr. R, K, Qehringer for
checking the manuscript.