J. Phys. Chem. Solids Pergamon Press 1959. Vol. 8. pp. 21-25.

Printed in Great Britain

SESSION B:
THEORY I’

B.l BAND THEORY?

J. C. SLATER

Massachusetts Institute of Technology, Cambridge, Mass.

Abstract-The band theory of solids represents a one-electron theory, in which an electron moves
in a periodic potential representing the nucleus and the averaged potential of other electrons, in
the sense of HARTREE'S self-consistent field. It is really broader than a one-electron theory, however, in
that group theory shows that a perfect crystal of a semiconductor, having only one electron in the
conduction band, or one vacancy in the valence band, will have energy levels of the same general
type, so that so long as the interaction of electrons and holes with each other, and the distortion
of the lattice by the charges (polaron formation) can be neglected, it forms an adequate approxi-
mation. Important advances have been made in the last few years in the approximate solution of
SCHR~DINGER'S equation for a periodic potential, in the detailed understanding of the band structure
of particular semiconductors, and in the use of experiments to clarify many important points of
these band structures. Some of these advances will be described.

THIS paper is intended as a general survey, to
open our discussion on the theory of semicon-
ductors. Every worker in semiconductors must be
aware of the debt which the theory owes to
BLOCH, BRILLOUIN, WILSON, and others(r) who,
in the early 1930’s developed the idea of energy
bands in solids, and the relation between the
existence or lack of energy gaps and the distinction
between insulators, semiconductors, and metals.
We are aware also of the fact that our Soviet
colleagues believe that in the last few years the
concept of band theory has been overplayed, and
that they feel that the American workers, in
particular, have taken this theory too literally.(z)
I do not agree with their point of view, but I
believe that these are questions to be considered
on their scientific merits, and I should like to
use this paper to discuss in an elementary manner
the way in which a theorist now regards the band
theory, and fits it in with his broader under-
standing of the theory of solids in general.
We realize in the first place that the theory of
energy bands is a one-electron theory. An electron
* Chairman: W. SHOCKLEY; Co-Chairman: H. EHREN-
REICH.
t Supported in part by the Office of Naval Research
under contract with the Massachusetts Institute of
Technology.
21
moving according to the principles of wave
mechanics in a periodic potential, repeating in
each unit cell of a perfect crystal, will have wave
functions of the form of a plane wave, symbolized
mathematically by the exponential eik*r, where k
is a propagation vector, r the position vector,
multiplied by a periodic function &7t(r). This
periodic function is characterized by k, which acts
as a quantum number, and by another quantum
number which we may denote as II. The corre-
sponding energy of the electron in the periodic
potential is E(kn), again determined by the same
quantum numbers. The quantity k takes on
independent values only within a polyhedron
surrounding the origin in k space, called the central
unit cell of the reciprocal lattice, or the central
Brillouin zone ; a value of k outside this poly-
hedron leads to wave functions repeating those
already found from k’s inside the central Brillouin
zone. We may describe the quantum number n
in a simple way. For a given value of k, the periodic
potential problem will have an infinite number of
energy values. We may label the lowest energy
value n = 1, the next rr = 2, and so on. The
energies in general will not be degenerate; de-
generacy comes only for certain values of k having
particular symmetry properties, as has been very
throughly studied in the last few years. Thus there
22 SESSION B: THEORY I

is no ambiguity in general regarding this labelling. bands with gaps above in cases where it is known
For a given value of n, the energy and wave that we have insulators or semiconductors, and
functions vary as smooth functions of k. We then they indicate partly filled bands in cases known
define the energy levels associated with the same to be metals. This is about as far as the agreement
value of n, and with all values of k within the went until the last few years. Even ten years
central Brillouin zone, as constituting an energy ago, the complicated structure of the bands
band. indicated by such theory as was then available
Pauli’s principle, coupled with band theory, was a purely hypothetical thing, with no ex-
shows that there are just enough independent periments available to check it. Now, however,
stationary states in an energy band to accommodate the situation has completely changed, particularly
one electron of each spin per unit cell in the as concerns the behavior of germanium and the
crystal. The theory of electrical conductivity 3-5 compounds. The beautiful experiments on
shows that if the electrons fill the lowest possible cyclotron resonance, carried out by KIP and his
states, and if these fill certain energy bands, associates in Berkeley, and by LAX and his associ-
leaving an empty gap above between the topmost ates at M.I.T.,c4) have given us a very powerful
filled band and the lowest empty band, we shall tool for determining the details of the energy bands
have an insulator or semiconductor, depending for those materials for which the experiments have
on the width of the gap, while if the electrons been possible. The even more recent experiments
partly fill certain bands, leaving unoccupied on the oscillatory magneto-optic effect verify and
states immediately above, we have a metal. extend these results, and give very elaborate and
These are the elementary and familiar principles convincing confirmations of the energy-band type
of the energy-band theory. No one questions them of theory. Experiments which are now being
as forming a correct deduction from wave mech- planned seem likely to extend these powerful
anics, provided we accept the one-electron picture methods to many types of materials which cannot
on which they are based. The real question is, to be examined by present techniques. It seems very
what extent does this oversimplified model likely that these experimental methods, depending
correspond to reality? We could attempt to answer on the interactions of strong magnetic fields and
this question in either of two ways. First, there is very high-frequency electromagnetic oscillations
the empirical or pragmatic answer: to what extent in solids, will eventually be capable of giving very
do the consequences of such a picture agree with complete information about the energy levels in
experiment? Secondly, there is the more theoretical solids, information comparable with that given
answer: to what extent do they agree with the about individual atoms by atomic spectroscopy
more elaborate theory which we should find by and the Zeeman effect. And the further we go in
rigorously applying the principles of wave mech- this direction, the more it seems that an energy-
anics to the problem of a solid? To me, as a band picture is capable of explaining what we are
theoretical physicist, the second type of answer is finding. In other words, the empirical answer to
more satisfying than the first. If the results of the our question about the legitimacy of energy-
simple theory disagree with experiment, then it is band theory seems to be definite : the more we find
wrong; but if they agree with experiment, it is out about it, the better it works, at least for
not necessarily right: other theories as well might materials similar to germanium.
give the right answer. Nevertheless we must Let us then go to the other type of answer, that
realize that the plausibility of the theory is greatly depending on the analysis of more rigorous
strengthened by the many types of agreement approaches to the theory of solids. There are three
with experiment that have been discovered in major types of approximations made in using the
the last few years, many of which will be reported energy-band model of a solid. First is the assump-
on at this conference. tion of the self-consistent field: the assumption
In the first place, calculations of energy that the action of other electrons on the electron
bands,(rJ) using reasonable postulates regarding in question can be replaced by a periodic potential,
the one-electron potential, have agreed with arising from the averaged behavior of the other
experiment in the sense that they indicate filled electrons. Secondly, there is the closely related
 SESSION B:
approximation that we can disregard anything of
the nature of multiplet structure, treating the
problem like that of one electron outside closed
shells. Thirdly, there is the quite separate postu-
late that we can treat the electronic motion
separately from the nuclear motion, as in the
approximation of BORN and OPPENHEIMERfor a
molecule, solving for the energy bands in an
approximation in which the nuclei are held fixed
at the positions which they would occupy in an
ideal crystal at the absolute zero of temperature,
and bringing in the interaction with the lattice
only as a perturbation at a later stage of the cal-
culation. Let us take up these three approxi-
mations, all of which are very significant, in order.
For the first two approximations it pays to recall
the similar and familiar example of the theory of
isolated atoms, where the same two approxi-
mations must be made in the first stages of our
calculation.
A semiconductor is an inherently simpler prob-
lem in quantum mechanics than a metal, just as an
alkali atom is simpler than one with more outer
electrons. The reason is that the semiconductor
ordinarily has so few electrons in the conduction
band, or so few holes in the valence band, that
they can be treated independently of each other,
so that we are essentially dealing with a single
electron outside closed shells. We know that the
alkali atoms can be handled with great success
by a central field model, of a single electron moving
in a properly chosen potential field, determined
ordinarily by the HARTREE-FOCK method. This is
no accident. By group theory we can show that
the behavior of the orbital and spin angular
momenta of an isolated atom show properties like
those of a single electron. The reason is that the
angular momentum operator and the rotation
operator are closely related to each other, so that
the fact that the potential energy of the atom as a
whole is independent of rotation leads directly to
the angular momentum properties, like those of
single electrons moving in central fields.
Similarly the operation of translation through
one lattice spacing is related to the quantity k,
similar to a linear momentum, and often called a
pseudo-momentum, and the fact that the be-
havior of a semiconducting crystal is unchanged
when we make a translation through a lattice
spacing directly leads to a classification of its
THEORY I 23
stationary states by means of a pseudo-momentum
k like that of a single electron. In other words,
it is necessary that the lower stationary states of a
semiconductor with a single electron in the con-
duction band (or a single hole in the valence
band) show the same type of behavior as those of
a single electron in a periodic potential. This
similarity holds not only as far as the translation
is concerned, but also for the rotational operations
which result in the interesting degeneracy and
symmetry properties of the energy bands. This
point of view has been emphasized by a number
of physicists recently,(s) and probably will be
brought out more clearly in some of the later
papers in this series, by KOHN and others, dealing
with semiconductor problems by means of a many-
electron approach.
The second approximation met in the energy-
band theory, namely that we can disregard any
analogy to multiplet structure, or any interaction
of the electrons outside the closed shells, holds
only to a limited extent, though we are familiar
with phenomena where it does not hold, but
where this has long been recognized. The most
familiar of these examples is that of the exciton,
resulting from the interaction between an electron
and a hole. This has close analogy to the atomic
problem of an inert-gas atom, in which one
electron is removed from the filled shell and placed
in an outer shell. For instance, in neon, where we
have the configuration lss2sz2ps in the ground
state, we could go to the state ls??$2p53s, leaving
a hole in the 2p shell and an extra electron in the
3s state. In such a case, we know from the vector
model of the atom that we have a vector coupling
problem between the 2p shell, which acts like a
single 2p hole, and the 3s electron, resulting in
1P and 3P states. This is a simple example of
multiplet structure, replacing the douhlet spectra
arising in the case of alkali atoms. But this situation
as applied to excitons is now well understood.
The work of OVERHAUSER(~) and others has shown
that the multiplet structure of excitons, as com-
plicated by the fact that we are in a crystalline
field rather than a central field, really exists. Here
is a simple case where we must consider a two-
electron problem, and there are other such cases.
They do not invalidate the picture of energy
bands, but merely show that we must expect to
extend it in appropriate cases.
24 SESSION B:
We may remark at this point that it is here that
the theory of metals, and of magnetic media,
becomes so much more complicated than that of
semiconductors. We know from atomic theory
that a partially filled shell has much more com-
plicated multiplet structure than a shell witb only
two electrons. Thus, for instance, there are five
muitiplets for the configuration ds, eight for 8,
and sixteen each for d4 and d5. In the f shell,
even f4 has 47 multiplets, and f 7, the half-filled
shell, has no fewer than 119. A partially filled
energy band in a metal acts, as far as its multiplct
structure is concerned, somewhat like an atom
with a shell capable of holding two electrons for
each unit cell of the crystal, partially filled with
electrons. This is something that cannot possibly
be handled completeIy and rigorously, and we are
pushed into approximations for treating the multi-
plet structure, which often are far from satis-
factory. In the case of paramagnetism, this situa-
tion is not so serious. Here we can treat each
paramagnetic atom separately, so long as their
wave functions do not overlap or interact by
superexchange mechanisms through intermediate
electrons. Thus the problem is no more difficult
than that of a single paramagnetic atom in a
crystalline field, and there has been widespread
success in treating such cases. But once the orbitals
of the various atoms overlap or show superex-
change phenomena, we must consider all the unit
cells of the crystal as forming a single problem and
consequently in ferromagnetism and antiferro-
magnetism we meet theoretical problems of an
altogether higher order of difficulty. Fortunately,
in an ordinary non-magnetic semiconductor, these
situations do not arise, and it is here that we may
expect the energy-band theory in its simpler forms
to hold with relative accuracy.
Now let us consider the third and final approxi-
mation made in the energy-band theory. We
conventionally start by fixing the nuclei at the
positions which they would have in a perfect
crystal at the absolute zero of temperature, setting
up a s&-consistent potential, and solving the
energy bands in that potential. Then, as a later
stage in the approximation, we try to bring in the
interaction with lattice vibrations as a pertur-
bation. It is this approximation which the Soviet
workers have mainly questioned: they feel that
the interaction is so strong that it cannot be
THEORY I
handled by perturbation methods. This would be
expected to be the case more particularly in ionic
crystals, where an isolated electron can produce
large distortion of the lattice of ions, rather than
in a covalent crystal such as germanium. Here
we are meeting problems which are the hardest
ones in semiconductor theory, and where I believe
that we still do not have enough information to
decide the question completely. There seem to
be at least two methods of approach to its solution.
First, there is the experimental approach. The
electron-lattice interaction in such a covalent
crystal as germanium is not likely to be large, as I
have just mentioned. It can well be that this is a
partial explanation of the fact that it is in ger-
manium that the energy-band approach has
worked best. But fortunatefy now there is more
and more work on the III-V compounds like
InSb, and on the II-VI compounds like ZnS and
ZnSe. Surely these compounds will be increasingly
ionic, and in the I-VII compounds like CuBr we
shall have an almost completely ionic case, though
with the same type of crystal structure. Intensive
study of these groups of compounds, such as is
going on at present, is bound to show if, or where,
the energy-band picture begins to break down as
the crystal becomes more ionic. The evidence is
pretty clear that the situation in the III-V com-
pounds is not very different from that in ger-
manium; but those compounds might well be
nearly covalent, and the ionic behavior may begin
to come in only with the II-VI and I-VII com-
pounds.
The second approach, which still must he
developed much further, though very good steps
toward it have been taken, is the theoretical study
of electron-lattice interactions, in the limit where
they are large. Some theoretical studies of polarons
in alkali halides, which have been carried out so
far,(T) suggest that the lattice polarization by the
conduction electrons, though appreciable, is not
enough to affect the properties of the crystal in a
very profound fashion. But these studies have
hardly gone far enough to be conclusive. The
whole question of interaction of electron waves
with lattice waves, and of lattice waves with each
other, is to my mind the leading theoretical prob-
lem in the study of solids at present. Very good
starts toward its solution have been made, for
example by VANHOVE,@)using methods which
 SESSION B:
were originally developed for the study of field
theory. But these studies must be carried much
further than they have already, before they can
give really conclusive answers to the behavior of
polar crystals. It is almost certain that in such a
study, the simple energy-band theory will appear
as a first approximation for the electronic motion,
as the simple harmonic theory of elastic vibrations
will appear as the first approximation for the
nuclear motion. But it is the higher approxi-
mations that will count, and the encouraging thing
about the recent studies in the field is that rigorous
methods are being found to apply perturbation
theory to such problems.
Let me close by describing a few of the many
problems that await solution along this line, but
whose treatment is in sight, and probably will be
worked out in the next few years. First we have
the problems of ordinary elastic vibrations of
crystals. Here we are beginning, through the work
of LAVAL, WARREN,cg) and their colleagues with
X-ray thermal diffuse scattering, and through the
work of BROCKIIOUSE(~@and others with neutron
scattering, to have good experimental knowledge
of the elastic spectra of a few simple crystals. We
must tie these results in with fundamental cal-
culations of microscopic elastic constants, from
energy-band theory or improvements on it. Then
we must consider the scattering of one thermal
wave by another, leading to thermal conductivity
on the one hand and thermal expansion on the
other. Here VANHOVE@) has carried out a thorough
theoretical analysis, which awaits comparison with
detailed calculations on individual crystals. Next
we must consider the interaction of electron waves
with these elastic waves, resulting on the one
hand in a theory of electron mobility and electrical
resistivity, and on the other in an understanding
of the damping of ultrasonic waves by electron
waves which has been the subject of interesting
experimental investigations recently.(ll)
These are but a few of the aspects of these
interactions between waves, as far as semicon-
ductors are concerned; I have already indicated
THEORY I 25
that the problems of metals, and particularly of
ferromagnetic metals, are far more complicated.
Fortunately, as I have mentioned, the theoretical
means of handling these problems are becoming
more powerful; and I have a great deal of con-
fidence that through such advances in theory, we
shall eventually find to what extent a simple
energy-band approach is adequate for ionic semi-
conductors, as it seems to be from empirical
evidence for the covalent semiconductors. At
least, we seem to have here a fruitful direction of
attack, and I believe we shall find that it is work
along these lines which will make the most useful
advance in semiconductor theory in the coming
years. As far as I know, this point of view is in
agreement with that of the Soviet workers, who
are actively pursuing theoretical work along the
lines I have described, and have been making
good progress toward an understanding of these
difficult problems.
REFERENCES
1. SLATER J. C., The Electronic Structure of Solids
from Encyclopedia of Physics Vol. 19. Springer,
Berlin (1956), for bibliography on the band
theory.
2. DUNLAP Jr. W. C., Phys. Today 11, 14 (1958).
3. HERMAN F., Rev. Mod. Phys. 30, 102 (1958), for
references not given in (1).
4. LAX B., Rev. Mod. Phys., 30, 122 (1958), for refer-
ences on experimental methods of studying
energy bands.
5. KOHN W., Phys. Rev. 105, 509 (1957).
6. OVERHAUSERA. W., Phys. Rev. 101, 1702 (1956).
7. SCHULTZ T. D., Solid-State and Molecular Theory
Group M.I.T. Tech. Rep. No. 9 (1956) (un-
published), including references.
8. VAN HOVE L., Unpublished. To be available as
Solid-State and Molecular Theory Group M.I.T.
Tech. Rep.
9. LAVAL J., Rev. Mod. Phys. 30, 222 (1958); CRIBIER
D., ibid. p. 228; JACOBSENE. H., ibid. p. 234.
including references.
10. BROCKHOUSEB. N. and STEWARTA. T., Rev. Mod.
Phys. 30, 236 (1958); STEWARTA. T. and BROCK-
HOUSEB. N., ibid p. 250.
11. MORSE R. W., Phy~. Rev. 97, 1716 (L) (1955).