622

P. R. WALLACE
may exist which could be completely masked by the eigenvalue dependence upon wave vector direction.
ACKNOWI EDGMENTS

of high resolving power for potassium, which is expected to show wider departure from free electrons, have been carried out by Platt. The X edge investigated shows quite close agreement with his theoretically predicted absorption which was based on the assumption that the electrons are free. No evidence existed to show an energy gap. This, of course, is at best supporting evidence of the ! non-existence of a gap in potassium since gaps

X-ray absorption measurements

"

"J.B. Platt,

Phys. Rev. 69, 337 (1946).

The authors are indebted to Dr. %. A. Bowers for permission to publish his results on the five additional states shown in Table IV and also wish to thank Dr. Bowers for continuing the application of the method when the war interrupted the work.

PHYSICAL REVIEW

VOLUME

71, NUMBER

MAY

1, 1947

The Band Theory of Graphite

P. R. War. AcE* National Research Council of Canada, Chalk River I.aboratory, (Received December 19, 1946)
usirig the

Chalk River, Ontario

The structure of the electronic energy bands and Brillouin zones for graphite is developed "tight binding" approximation. Graphite is found to be a semi-conductor with zero activation energy, i.e. , there are no free electrons at zero temperature, but they are created at higher temperatures by excitation to a band contiguous to the highest one which is normally filled. The electrical conductivity is treated with assumptions about the mean free path. It is found to be about 100 times as great parallel to as across crystal planes. A large and anisotropic diamagnetic susceptibility is predicted for the conduction electrons', this is greatest for fields across the layers. The volume optical absorption is accounted for.

1. INTRODUCTION

HE purpose of this paper is to develop a

basis for the explanation of some of the physical properties of graphite through the band theory of solids. We shall be concerned primarily with a discussion of its electrical conductivity, but the treatment given makes possible the explanation not .only of the electrical conductivity and its anisotropy but also the thermal conductivity, diamagnetic susceptibility, and optical absorption. The electrical resistivity of single crystals of graphite is about 4 to 6 X 10 5 ohm-cm. ' This corresponds to a conductivity of the order of that of a poor metal. The temperature coefficient of the conductivity is negative, as in the case of
by E. Ryschewitsch, Zeits. f. Elektrochem. ang. 6)&10 ~ ohm-cm. physik. Chemic 29, 474 (1923), as 3.9

a metal. Polycrystalline graphite, on the other hand, has a much higher resistivity which varies very strongly according to the type of graphite used, and has a positive temperature coefficient of conductivity' to about 1400'C, and negative thereafter. Since the crystals of commercial graphites tend to be of the order of 10 ' cm, and 1.6 as against 2.25 it is quite porous (density for single crystals), it seems reasonable to attribute the high resistivity of polycrystalline graphite to the crystal boundaries, on which may be lodged impurity atoms. The latter would tend to be driven off on heating, thus accounting for the observed temperature dependence. We shall show, however, that the band theory would seem to make possible the explanation of the conductivity properties of single crystals.
~ C. A. Hansen, Trans. Am. Electrochem. Soc. 16, 329 (1909) gives 137.5)(10 ~ at O'C 82.5X10 ' at 1400'C.

' Given

* Now at McGill University.

BAND
2. ZONE STRUCTURE OF
HEXAGONAL
A

THEORY OF GRAPHITE.

623

SINGLE

LAYER

Since the spacing of the lattice planes of graphite is large (3.37A) compared with the hexagonal spacing in the layer (1.42A), a first approximation in the treatment of graphite may be obtained by neglecting the interactions between planes, and supposing that conduction takes place only in layers. Graphite possesses four valence electrons. Three of these form tight bonds with neighboring atoms in the plane. Their wave functions are of the form

FiG. 2.

(2s)+&2/, (a.';2p)) Q,
v3

(i = 1, 2, 3),

where P, (2s) is the (2s) wave function for carbon and P, (0;2P) are the (2P) wave functions whose axes are in the directions o.; joining the graphite atom to its three neighbors in the plane. The fourth electron is considered to be in the 2p. state, its nodal plane being the lattice plane and its axis of symmetry perpendicular to it. The three electrons forming co-planar bonds will not play a part in the conductivity; we shall therefore treat graphite as having one conduction electron, in the 2p, state. For the hexagonal layer the unit cell, which is designated by 8'XYZ in Fig. 1, contains two atoms, A and B. The fundamental lattice disAA' and a2 AA"; their placements are a~ magnitude is 1.42 g V3 = 2.46A = a. The reciprocal lattice vectors have magnitude 2/V3a, and are in the direction AB and AZ, respectively. From this it follows that the first Brillouin zone is a

hexagon (Fig. 2) whose sides are distant 1/&3a from its center. It is easily shown that this zone contains one electron per atom, for the density of electron states in 0-space is 2A, where A is the area of the crystal. The zone in question therefore contains 2A X2/(&3a') electron states. But the atomic area (area per atom in the layer) is VBu'/4, and the number of atoms is 4A/V3a, ', which is exactly equal to the number of electrons in the zorie. The zone containing two states per atom is obtained by extending the sides of the hexagon to form a six pointed star. Let us now consider the problem from the viewpoint of the "tight binding approximation. If X(r) is the normalized orbital 2p, wave function for an isolated atom, then the wave function in the tight binding approximation has the form:
where

"

0= v'~+&v'2
r~) p, = P~ expL2~ik r~]X(r
rQ) Q B exp L2 s ik rs]X (r

(2.1)
(2.2)

and

Fro. 1.

The first sum is taken over A and all the lattice points generated from it by primitive lattice translations; the second sum is similarly over the points generated from B. From variational principles it is known that to get the best value of 8 for this approximation we substitute in the wave equation from (2. 1), multiply by q and y2, respectively, integrate, and then eliminate X from the resulting two equations. Let us neglect the overlap of the p, wave functions centered on different atoms, i.e. , let us

624

P. R. %ALLA
+
+

C F.

i
I

I I
I

I I

I
I I

+
I
I

I
I
I I

I
I I

)
I I

i
I

I
I
I

I
I

I I I

I I
I

+
FrG.

FIG. 4.

3.

crystal, and by symmetry H

1

=H.
1

Introducing

assume

X r X r r~ r~
Then, substituting

d7. =0.

Hl 1

H22

Hl 1

(2.3)
1

H22I

(2.6)

(2. 1) in
(2 4)

H12

H12

we have, finally

and proceeding as outlined, we get

Hg $+ XH] Q

E =Hn'a

Hgg' ~.

(2.7)

ESI

where
H~~

H21+ ~H22
r,

The positive sign will apply to the outside of the hexagonal zone, the negative sign to the inside. The discontinuity of energy across the zone
boundary
@~*H&gd v,

y~*Hy~d

Hgp

Hgg*
r

is then

(2.8)
Let us now calculate H~~'=H~~' and H~~'.

Hgg

~pg*H&gd

and

Hg~'

1 Q

exp/

2~ik

(rg

rg

)]

S=
Eliminating

Qx*pjdr

@g*ypd v.

X*(r rg rg)HX(r )d v-.

Omitting the exchange integrals over atoms more distant than nearest neighbors amongst the atoms A, and writing
Zo

we obtain the secular equation

Hgg

BS
Hgp

Hgg

BS

from which
1

it is found that

X*(r)HX(r)d~,

2S

j Hl 1+H22

+ ((H '+4 IH ')'l H)

I

(2.5)
where
g

go'

= X*(r
J.

y') HX (r) d r,

Now by virtue of the. neglect of the overlap integrals, S=X, the number of unit cells in the

=a& (say) is a vector joining nearest

BAND
neighbors among the atoms A, we get

THEORY OF GRAPHITE
which is greatest

625

at the center and decreases to corners. zero at the ~ii' =&o 2yo'(cos2mka+2 cosnk, a% cosya). The degeneracy at C and similar points, and the zero-energy gap at these points, are preIf we write sumably consequences of the symmetry of the Id = ~o+ (If IIO), lattice, and are independent of the approximafor an isolated where II0 is the Hamiltonian tions considered. carbon atom, and put It is easy to determine the form of the wave functions corresponding to different k. Consider, V H IIp U(0, for example, the point F. Inside the zone where U is the periodic potential of the lattice boundary ) = and consequently the nodes 1, and U' is the potential field for an isolated atom, of the wave functions bisect the lines joining A we may write, since HpX =E0X, and 8 atoms: the sign of the wave function at or signs in lattice points is designated by tX*(r) U (2.9) Fig. 3. It is readily seen how the expression for Zo = E V)X(r)dr, ( J the energy arises. Since an atom has two nearest neighbors of the same sign, and one opposite, and tX"(r y')(U U) V)X(r)dr) 0. (2.10) since (V is negative, there will be a contribution to the energy from exchange between nearest neighbors of 2pp+pp= Of second pp. Ep is the energy of an electron on the 2P, state in neighbors, we have two of the same sign and four carbon. us 2+0'. Let us next calculate H~2. For this, we shall opposite, giving In Fig. 4 are represented the nodes of the wave nearest consider only interactions between function corresponding to a point outside the neighbors in the lattice, the nearest neighbors of zone boundary at A. Since here ) =1, the maxima atoms of type A being always atoms of type 8 and minima correspond to the nodes of the and vice versa. Writing previous case. The expression for the energy can be verified in the same way. Similar arguments V)X( 0, (2.11) may be carried out for other points in k-space. r)dr) yo ~l X*(r p) (U Consider next the energy contours. Near the bottom of the band they are circular: where g =AB, we obtain

Hie'

= 2~ik yo(exp[

(a/K3)

]
v3u).

E =Ep 3yo 6yo'

+2 cosmka
so that

expL2vrik, (a/2%3)]),

'(v. +6vo')(k*'+k. ')"

i I'm i' = yP(1+4 cos'~ka+4 cos~ka cosxk

The energies at various points may now be written down:

The hexagon whose sides are distant 1/2a from the center of the Brillouin zone is a surface of constant energy. (This conclusion is, of course, true only in the approximation which we are

at 0: at D: at C: at F, inside:
outside:

8 =Eo 3yp 6yp, 8 = So+3+0 6+0

~0+3+0
+0 +0+2+0 & = ~o+vo+2vo'.
y

Over a side of the zone there is across the boundary at any point a discontinuity of energy of amount
2yo(2 cosxke 1)

(2. 12)

FIG. 5.

P. R.
here considering. ) Curves of constant energy are in, dicated in Fig. 5. Near the corner C

M7ALLACE

E=EO+3yo'a~3m'yolk

k la

3&'va'Ik

k I'a' (2. 14)

and the surfaces of constant energy are again circular. It should be noted that, when the reduced zone scheme is used, the energy contours have the same shape for the second zone as for the first.

3. NUMBER OF FREE ELECTRONS

AND CONDUCTIVITY OF A SINGLE HEXAGONAL LAYER

Fro. 6.

In this section we shall neglect yp' relative to yp. We then have

E. k, IE = v3w y, a k
I I

near the corners of the Brillouin zone. At the absolute zero of temperature, the hexagonal zone will be completely filled, and the next zone will be completely empty. At higher temperatures, there will be some thermal "overflow" into the outer zone. The extent of this overflow will be (in energy) of amount comparable to kT. The energy contour E = 8, +k T is approximately provided circular with radius k T/(xv3yoa) which is certainly the case at ordinary kT&&yp, temperatures. Now if N(E)dE is the number of electronic energy states in the energy interval dB,

the positive holes (equal in number) created in the lower band. To calculate the number of these, we must know the Fermi distribution. For moderate temthe peratures, N(E) is even in e = E E, over whole range in which the Fermi distribution is different from its value at absolute zero. Therefore, in the Fermi distribution function
I I

f(E) =
we shall have

expl

(E )/&Tj+ l f
other words, we may

g=Z. in

write

f(E) -f(e) = l/(e

"+l).
~
6%3

(3.5)

It follows that the number of free electrons plus positive holes per atom is
t

"N(E)

fkT
E.

N(E) =2A

dr
~

(3.6)

lgd, EI

(3.2)

&p

yp

where A is the area of the lattice, and the integral is taken over the curve on which the is grad&E energy is E. This gives, since
I I

constant,

N(E) =

4A

IE E,
8

37rpp

l~V3yp'

Now C. A. Coulson' has estimated that yp is about 20 kcal. /mole, or about 0.9 ev. At room temperature kT = 0.025 ev. Therefore the "effective number of free electrons" ff, per atom, is

I,

n, sr=2. 3X1G

4.

(3 3)

or, since the area per atom is &3a'/4,

N(E)/N =
mV3yp'

E. IE

=,

To determine the conductivity of our graphite layer, let us calculate directly the current in the presence of a given external field. Explicitly, this is given by

(3.4)

j =2~~evf(k)drl,
where f(k) is the Fermi presence of the field, and
,

(2/c),

(3.7)

where N is the number of atoms in the lattice. The'form of N(E) near E, is illustrated in Fig. 6. Conduction will take place through the electrons excited into the upper band, and through

'c=3.37A

in the distribution is the dis-

' Private

communication

to the author.

BAND

THEORY OF GRAPHITE
We will not attempt to calculate this quantity in the present paper. Let us note, however, that experimental evidence requires that the mean free path decrease with temperature faster than T '. In this connection, one might call attention to the sharply increased mobility found in polar salts at low temperatures for thermal electrons. The mean free path has the form
1,

tance between graphite layers. f(k) is given by

er
E

(3.8)

where fo is the undisturbed Fermi function (3.5), and 7 is the inverse of the probability of scattering per unit time. For moderate fields we may expand this

f(lr) =fo(I )+g(&)

where

(3 9) (3 10)

=Cl

(T& (e"r 4 1), ( ll)

l

'*

(3.13)

g(k)
Writing
v

= F gradfo.
h

e~

where

= gradqE,
h

gradfo

dfo

4E

grad qE,

87'y

=do g
I

&E

g~~dE

and C is a 8 is the Debye temperature constant of the order of one atomic distance. The electrons in graphite are also conductivity essentially thermal. If the mean free path is assumed to be of the same order of magnitude as given by (3.13), with C=2X10 ' cm and ll estimated from specific heat evidence as about 2000'C we will get

l'=3X10 'cm.
on

element of length E=constant, we have

d0~

being

the

curve

(3.14)

j=

4e'r
h2C

l
~'

dfo

l'

dZ. ~

(F gradE)

gradE
da&

This gives rise to a resistivity ' =5 X10 ohm-cm.

p

dE,

lgradE

where the inner integral is over the surface of constant energy K The component of current j p in the direction of the field is

seems therefore, that (3.14) gives at least the right order of magnitude for l. This point, and the temperature dependence of l, seem worth further investigation. From the formula
0 v ef f~ 2 &/ jeff& c

It

Jp= ae

4e'r
~

dfo

(F gradE)'
lgradEI

dE

r~

80 je

SB.

(3.15)

Averaging over all directions of the field in the plane of the layer we get
0II

we can calculate the "mean effective of our conductivity electrons. This is found to be

" mass, m, f f,
(3.16)

h2kT
1Spff

287 f'Gp
k'c

36 log2

g2yp2

J dE

lgradEldo.

dE.

(3.11)

which yields, on numerical

evaluation,

at room

By virtue of (3.1) this is

8xe2r
h2C

temperature
7B ff
p

1/18 electron mass.

(3.17)

4. THE BRILLOUIN ZONES OF GRAPHITE

Be

(3.12)

16+e2V

kT log2.
h2C

The graphite lattice is built up of hexagonal layers whose relation to each other is indicated in Fig. 7. Dotted lines indicate one layer, and
4 N. F. Mott and R. A. Gurney, Electronic Processes in Ionic Crystals (Oxford University Press, Oxford, 1940}, p. 107. Note also the increase of mobility with decreasing temperature in the semiconductor Cu20 shown in Fig. 66, p. 168.

To get a numerical value of the conductivity

it
is- necessary to know something about 7, or alternatively about the mean free path l=vv-.

P. R. WALLACE

/
)
I

~
~l

1+~ J

I
r
/

FiG. 7.
,

FrG. 8.

the full lines the one immediately above it. A set of basic displacement vectors of the lattice is
ag

= [-, 'a, 'v3a, 0],

,

a2

a, 0], [0, a, = [0, 0, c], (4. 1)

c being twice the distance between layers of the lattice. In Fig. 8 the relation between the layers is shown in perspective. The points A, 8, t", D are the four atoms of a unit cell. The volume of the unit cell is

&. u=ai (a~Xaa)

so that the atomic volume is

= a'c,
2

V3

of the lowest zone is 1/c. This is; however, not the case, for the plane at height 1/2c corresponds to Bragg reflection of waves whose wave-length is 2c. Thus reflections from successive planes. (distance c/2) differ in phase by ~, and therefore destroy each other. The structure factor therefore vanishes on the planes b, = &(1/2c), and there is no energy discontinuity over them. The horizontal boundaries of the first zone are therefore at b. =&(1/e). The lowest Brillouin zone so described has the volume 4/v3a'c. But the density of states is twice the volume of the crystal, that is,

2X(number N of cells) X (volume of cell) =&3a'cN.

v3a'c
I atomic=

(4.2)

The intercepts of a possible unit cell on adjacent planes are shown in Fig. 8. The cell would extend
beyond these two planes on each side halfway to the next planes, which are not shown. Reciprocal lattice vectors corresponding to (4. 1) are

bl

=,
V3a

0,
1

b2=, v3a
1

0,
c

Thus there are 4 X electron states in the lowest Brillouin zone, or one per atom. Since as in the case of a single layer, there is just one "conductivity" electron per atom, the lowest Brillouin zone will be just filled at the absolute zero of temperature. In Fig. 9 we indicate the upper balf of the zone in question. Let us now consider the problem from the " viewpoint, of the "tight binding approximation. As in the case of a single layer, let X(r) be the 2P, wave function of an isolated C atom. Then we assume for the collective wave function of the crystal
j
e
~

b3=

0,

0,

X;(Q exp[2sik r;
i=1
~ ~
~

r; ]X(r )), (4.3)

It follows that the lowest Brillouin zone is bounded by six vertical planes forming a right hexagonal cylinder whose sides are distant 1/V3a from its axis. It appears at first that the height

where i=1, 2, 3, 4 correspond to atoms of the type A, 8, C, D, respectively, and n enumerates the different atoms of a given type in the crystal.

BAN

THEO R Y OF GRAPH

I TE

629
Hp,

Proceeding exactly as in the case of the hexagonal layer, we are led to a secular equation for the energy

2 cosmk, c = I' and

If we put the diagonal elements equal to

Zvrik, (a/v3)]+2 cosska exp[

~

Now the diagonal elements are

exp[sik, (a/43) = 5, (4.5)

Htl =Hen =Has

=.

H44

=So

Zpo

(cosZs'k&a

the secular equation becomes

+2 coss&3ka cosska) =Ho. (4.5)

Bp' is not quite the same as the Ep in the previous section, due to a difference in the potential; we shall, however, subsequently drop the prime for convenience. In .calculating the non-diagonal terms, we shall at first consider first and second neighbors in the planes, and also first and second neighbors between planes. The elements may now be calculated: H. This is equal to
yo(e xp[
Zsik(a/v3)

E Hp

' y1I. ypS S* E ~,'rS II, ~, y1'I'S B H

y1'FS

y1'FS

y1'FS*

ypS

= 0. ypS* E IIp

~,'rS

(4.6)

The exact solution of this equation presents some difficulty. We shall neglect y1' at first, and later treat it as a perturbation. In the lowest approximation, then, we get on solving,

]
V)X (r)dr) 0

Z=H, ~-,y)ra[-'y, 'r'+yg'~ '

S~']&.

where

+2 cosnka exp[sik, (a/K3)])

go=~I X*(r
pAB) (U

H. This is equal to
where
y~

2y1 cosxk, c

J/X*(r ( V U)X(r)d~ g~c)

X*(r yea),

y~)0

because (V is negative, and U)

With regard to the sign in front of the 21y1F term, the choice here is immaterial, a change of sign merely corresponding to reHection in the plane k, = i. /Zc of the two halves of the upper half-zone, and similarly in the lower half-zone. By choosing the negative sign, we adhere to the convention that the bottom of the zone (point of lowest energy) should be at the center. The signs in front of the square root cor'respond to the outside and inside of the lowest zone, respectively, i.e. inside

-', Z =H, ~,r '~, 'r'+~, 2I Sl [-,

']',

(4.)

X(r) have opposite signs between the planes, where the contribution to the integral is greatest. H14. This is equal to
Zy ~' cosm. k.c(exp [Zs ik, (a/v3)

]
erik, (a/ ]). , v3)

where
yg'

+2 cosvrk, a exp[
J

X*(r (V U)X(r)dr, ygD)

and is almost certainly positive, though it has not been evaluated. The other elements may be written in terms of the above. It may easily be verified that
II23
H14

H24

H23

H34

H12

FiG. 9.

630

P. R. WALLACE

Fre. 10. Energy contours on the

inside of the zone.

Fj.G. 11. Energy contours on the outside of the zone.

and outside

E =ap ~,r+P~grP+ypP
-',

sl 'g&.

(4.7b)

The energy inside is always &~HO=BO, it has the latter value for lsl'=0 and (i.e. , )'4&1/2c). Now

r(0

lSl'=(1 cosprka)' 2

(4 8)

on the sides of the zone, and is zero at the vertical edges. Thus the maximum energy inside (Ep) is atta, ined on the upper (and lower) quarter of the vertical edges of the zone. By a similar argument, the energy outside has its mpmmlm value, (which is also Ep) on the middle half of the vertical edges of the zone. The energy gap is therefore zero only at twelve points; two on each vertical edge. In Fig. 10 we indicate the form of the sections of surfaces of constant energy by a plane such as that through OS V'Q in Fig. 9. In Fig. 11 we represent the cross sections of the energy contours by a plane such as the extension of I'QRS beyond QS. W is at a tip of the six-pointed star which forms the upper boundary of the second zone, this zone being obtained by extending the vertical sides and the top and bottom of the 6rst zone. The energy contours in planes k, =constant will have a form similar to those shown in Fig. 5. The hexagonal lines of constant energy will appear in each plane, but will correspond to di8'erent energies in different planes. The zero energy gap will of course only appear in the planes k, = &(1/2c). The strong anisotropy in some of the proper-

ties of graphite (electrical and thermal conductivity, magnetic susceptibility) is caused by the very anisotropic form. of the energy surfaces near the corner of the zone. These anisotropies will become greater as the temperature is lowered, and the surfaces of energy (Ep&AT) approach segments of lines. Let us now see how this picture is altered if we take account of y~', the exchange integral between atoms such as A and D in Fig. 8. If we expand (4.6) to terms of the 6rst order in y&' and solve, using the relation (4.7), we 6nd that the energy is increased by an amount
p

(s+s*) (s+s*) 2E. = vpvi'r ~, lsl 2Ey,r a4y, ' S '

l l

E is the "unperturbed" energy given by Since in the vertical edges S=O, p=0 there (4.7). both inside and out, and consequently there is no overlap on these edges. Also, ~ vanishes identically in the planes k, =~1/2c. Thus, at the only points at which the energy is continuous on crossing the boundary of the zone, e =0, and it is also identically zero in two directions at right angles through this point. This alone makes it seem likely that the addition of e will not give rise to any overlapping of the zones. More detailed consideration shows that such is in fact the case. It should be noted 6nally that the two energy bands corresponding to the zones considered above arise from the same energy level (corresponding to the 2p-state for C). The "splitting" of the energy level into two bands when the C
where

,
(4.9)

BAND

THEORY OF GRAPH

TE

63i

atoms are brought together to form the crystal is caused by the fact that the graphite crystallizes in a form having more than one atom per unit cell. Thus, the general properties of the zones (and in particular their touching) does not depend on the details of the potential field, but rather on the geometrical form of the lattice.
S. NUMBER OF FREE ELECTRONS IN GRAPHITE

which yields, on evaluation

N(E) =

SV
3x'6 cpp

(4yi'

&+ e')

e ir
~

+2s sin '

2p]

(5.4)

The first problem is to find N(E), the distribution of energy levels. The well-known formula for this is

N(E) =2U
~

do

gradiE

where V is the volume of the crystal, and the integral is taken over the surface on which the energy has the constant value Z. The formula may be readily transformed to

N(E) =2U
&

dkgdky

(BE/Bk, )., i

..

(5

&)

the result has been extended to the case &&0 by observing that it is, for eyp, even in ~. The result (5.4) holds only for ~(2yi. Coulson' has estimated that y] yppp=0. 09 ev. This corresponds to a temperature of nearly 1000'C, so that, insofar as calculations of the properties of graphite near room temperature are concerned, the expression (5.4) will be adequate. [t should be noticed that N(E) does not go to zero when e but approaches a finite value +0, proportional to yi (see Fig. 12). Nevertheless, as can easily be seen, at the absolute zero of temperature the number of free electrons is zero. The number of free electrons is in fact approximately

the integral now being taken over the projection of the surface E' = constant on the (k plane. k) Throughout this and the subsequent sections we shall consider only the exchange integrals pp and yi (nearest neighbors in and between planes). Let us write (4.7)
&

N. ii=2)
p

N(E) f(E)dE,

f(E) = eeIk1'+

=@ = cos~k, c +p pr

as in the case of the plane layer. Now with small error we may expand N(E):

+ (yi' cos'irk, c+y0'i 5

Calculating Be/Bk, and eliminating 2yy& e)0,

') ~.

(5.2)
N(E) =

SUA,

k, we get, for

3ira'cy0'

2+

y,

~+- +0~
4 y'

. I+"
~.
Doing

and extend

the integral

from 0 to

X (4y

i/2

(y

i g~
i

l. gi)')

(5.3)
&n

Now we shall be interested this case we may write

i
~

in &=kgbyp.

g i = 3s'(k 2+ k ') @i = u

The integral (5.1) may be transformed integral over

I:

to an

N(E) =
377'8 CE

(e~+2ey~) I&0~

(y0'u+ e')du

(Vi'~'-

4 (V0'u

-~') ') '

FIg. I2,

632
this we obtain
&elf =

P. R. WALLACE
32 U kTyi,
3xG2c pp2

si+
2

kT
$2

where n is the normal to the surface E. = constant and dS~ is the element of area in this surface. Averaging over-all directions in the plane and putting dSo 8 ~d/nwhere ' = ,P+ oo,
oq

+-I
where

3 pkTq
I

'

4&i

s. +O~ p
L

(kTq
I

~y,

) )

(5.5)

e=.

e'r
k

dfo
i

(aE/a)

'
SxddE.

dE

J~s& BE/Bk~

(6.2)
perpen-

s=g( 1)
m=o

(m+ 1) "

(5.6)

On the other hand, the conductivity dicular to the layers is

sz ~'/12, log2, so

etc. , the s's being tabulated

Op=-

2e'r
h2
&

dfo

BE SwddE.

in Dale's Mathematical Tables. Now the number of free electrons per atom is n, ff zf/XX, being the number of atoms. X, Since the atomic volume is v& o/ 8, c

dE' ~ (g) Bk,

(6.3)

On the surface

= constant
2oyo'(3xza'y}'*
+0

ae
BK

= 8 V/v3a'c.
m

g+C

(6.4)

So

st =K3~

keg

go'
3

sy+
2

kT
$2

kTq

'

kTq

op

(5

&)

where 3wz(k~'+k ')a'=y. Now in the Eq. (5.2), if we remain within the limit o~ (yz, $ ' may be replaced by y with an error of order (yz/yp) 1 percent. In terms of y, the integral in the square bracket of (6.2} may be written
~ ~
~

= 2.25 X 10~
at room temperature. This is not at all in agree' ment with the value of 2. 3 X 10 obtained by
treating a single hexagonal layer; in fact, the dependence of e, on T is quite different in the two cases. On the other hand, as we shall see in the next section, the conductivities in the plane agree to the first order. Thus the discrepancy is merely caused by our definition of n, f', and is compensated by a corresponding change in m, f f, the effective mass.
6. CO NDUCTIVITY IN THE DIRECTIO N OF THE
LAYERS AND PERPENDICULAR TO THEM

16oyop

)(,~+z.v, )/v,
~o

(yooy+ o') [4o'yzz

(yo'y

o')' j*

On evaluation

this yields

8~ m+2 sin ' '

2v

"j' [v
[vz
o

log-

o+71+ [Pl'+ 'J'* "j' hz' o'3'* vz [4vz' *

X log

o+Vi

j*

[4v

' oo3'+ [yz' "j' -

(6.4a)

By a procedure closely parallelling that used in Section 3 for the single hexagonal layer, we obtain for the conductivity in the direction of the vector u, if we assume that 7 does not depend significantly on k,
2e'r e(u) =&'
p dfp

Let us note first that if we take only the leading term in this expression, we get

See'r
h'c
16me'v
~

dfo

dE

edE

(6.5)

dE

kT log2
k
C

E=cpnsg

w hich is precisely what we obtained in (3.12) for the case of a single plane layer. It follows from

this and (5.7) that to the first order the eRective mass for electrons in motion parallel to the layers is independent of temperature. Taking account of the higher terms does not change the result drastically for moderate temperatures; at room temperature (T = 293'C) we get, using the numerical values already indicated (including the value for f assumed in Section 3),
&ir

for the "mean e6'ective mass" of electrons moving parallel to the graphite planes. This is valid so long as kT&&yl, and ls equal to =-, Similarly, from (5.5) and (6.6) we obtain for the mean effective mass for motion across the graphite

'.

planes

241og2 k'
8$gf f

' = pl I = 5 3 )& 10 ' ohm-cm.

~c'kT logyi/kT

Turning now to 0~, we get for the integral in the square bracket in (6.3)
8 typic

3 yo2e' yi

e
~

fn'

+sin '

&2

) 1e:
e

2y, &

4yi2

( e+y, + 1 ilog'-

' e'j' Ly, ~'3' yim- & ~+yi+[yi' "j: Ly "3' y L4y ' ~ "3' y L4yi' "j'+Lyi' &e'-'

With decreasing temperature this increases as 1/Tl logTI, becoming infinite as T-+0. At room temperature its value is =25 to 30 electron masses. More accurate calculation of o~ by numerical integration Rt loom temperature gives 0~ ' = p~ = 4.8 & 10 ' ohm-cm,
and therefore

(6, 6)

1y, 10-2. = =1.

pit
fT I I

PL

valid up to e=yl. We shall neglect the contribution to (6.3) from beyond this point. To determine first the order of magnitude of the result, consider the leading term only under the bracket; this is log yi/e. Then we get
'16 e'r

yic
&

3 Ii' yg'u'

~e~
log
yp'

yi Bfo
Be

16m'

e'r c yi(kT)'
h' a'

It is of interest to note that the anisotropy the ratio p~/pbedepends on temperature, coming larger, i.e. , the anisotropy becoming greater, with decreasing temperature. An attempt to confirm this prediction experimentally would be interesting; the measurement of o.~ would, however, necessarily be difficult. In Fig. 13 is plotted a graph of the ratio &r~/o for temperatures up to room temperature.
REMARKS ON OTHER PROPERTIES

kr

(6 6)
Several qualitative remarks can be made about the magnetic susceptibility and the volume absorption properties of graphite. Graphite is strongly diamagnetic, and the diamagnetic susceptibility of single crystals can be expected to show marked anisotropy. For a field in the s-direction, the susceptibility depends on the value of the quantity

This expression will give a good approximation only at temperatures somewhat below room temperature. We will use it, however, to indicate orders of magnitude. It gives for the anisotropy

factor
o.~
o-If

c' ylkT
log

9log2 a' yp'

kT

=0.01 at

room temperature.

(6.7)

Since this does not involve v it is completely calculable without assumptions about mean free path. A comparison of the formulas (5.5) and (6.5)
gives

- ( '8 , Bk,' Bk' (Bk,Bk)

B

B 6

over energy surfaces in the neighborhood of the boundary of the Fermi distribution, that is to say, near the zone boundary. ' This is a large
5

m. ii" = 2k'yi/3m'a'yo'

See Seitz, Modere Theory of Solids, pp. 594-595.

P. R. WALLACE

ts

fL/r

Fzo. i3.

lOO

Qoo

T'K

quantity, of order of magnitude 9s'a'&0'/pP, or about 6X 10' times its value, k'/m', for free electrons. On the other hand, if the 6eld is in the k, -direction, say, the susceptibility depends on

Unlike the preceding case, this depends on r. It is, in fact, =3m'a'c'kTy02/y~, or =kTyi/yo' times the value for the field across the graphite planes. The difference may be ascribed to the difference in effective masses in the quantized orbits perpendicular to the 6eld. The anisotropy is therefore of the same order of magnitude as

that of the conductivity.

Our other remark concerns the "optical" absorption spectrum. It is well known that electrons may under the influence of radiation make only "vertical" transitions in the reduced-zone scheme, that is, transitions in which the wave vector changes by one of the vectors of the inverse lattice. Let us then consider extreme cases. A transition from a point such as U (Fig. 9) involves zero-energy change. The greatest energy jump is from the center of the zone
6

(0 in Fig. 9) to a point at the middle of a vertical edge defining a "corner" of the star-shaped second zone. This is of amount 2LyP+9yo')' = 6yo. Transitions involving all intermediate energies are possible. An energy. jump of 6po corresponds to a frequency of 1.31X10"or a wave-length of about 2300A, fairly deep in the ultraviolet. Thus, there is absorption from the longest wave-lengths through the visible spectrum and a substantial part of the ultraviolet. At the extreme. limits absorption will be weak, however, due to the small number of states involved. The greatest number of states will correspond to energies such as those at the mid-points of the sides of the zone, which will give rise to transitions with energy jumps of the order of 2yo. Thus we should expect the strongest absorption to be in the neighborhood of 6900A, i.e. , in the red end of the spectrum.
ACKNOWLEDGMENTS

Reference 5, Section 71, pp. 326 8.

I wish to thank Professor N. F. Mott of Bristol University for valuable discussions, and Professor C. A, Coulson of Oxford University to me some preliminary for communicating results of as yet unpublished work.