Rle TR 113 04711211 - 2
Rle TR 113 04711211 - 2
Rle TR 113 04711211 - 2
-astitte of Tei'clLology
L~~~,mpl/1< 1
r, ~~~~~~~~~~~~~~~~~~
J. C. Slater
Abstract
I. Introduction
Many of the most important problems in the theory of solids concern
the motion of electrons in perturbed periodic lattices. Examples of such
problems are: the effect of impurities, of the P- or N-type, in semi-
conductors; the behavior of rectifying barriers, either between semiconduc-
tors and metals, or between P- and N-type semiconductors; and the behavior
of optically excited energy levels in crystals. All these problems have
received much discussion in the literature, by methods involving various
approximations to the solution of the wave-mechanical problem of the motion
of electrons in a perturbed periodic lattice. It is the purpose of this
paper to point out that there is a quite-general theorem in wave mechanics,
regarding the motion of an electron in such a perturbed lattice, which
serves to unify the treatment of all these problems. This theorem was
essentially discovered by Wannier,l and used by him in discussing excited
energy levels of crystals. That problem, unfortunately, is one of the most
complicated to which the method can be applied, and the theorem has tended
to lie buried in Wannier's paper, attracting little attention, without gen-
eral realization of the important simpler problems to which it is applicable.
James 2 has independently arrived at many of the qualitative results of the
present paper, but is apparently unaware of the importance of Wannier's
theorem, and bases his conclusions on quite-different and less-powerful
methods of discussing the problem, applicable only in the one-dimensional
case, though he has carried his approximations one step further than we have.
A number of other writers have used similar less-powerful methods for special
problems. 3
In the present paper, we shall first state the general theorem, and
prove it by a method similar to that used by Wannier for his special problem.
We shall then apply it to discussion of the important problems, such as
impurity semiconductors and excited energy levels, which result in discrete
energy levels. For problems of rectifying barriers and surface states, a
statistical approach is more appropriate, and we point out the relation of
3. For instance, S. C. Tibbs, Trans. Far. Soc. 35, 1471 (1939), who discussed excited
energy levels in NaCl, and S. Peckar, J. Phys. U.S.S.R. 10, 431 (1946).
-1-
our theoremto self-consistent field methods, and thence to statistical
treatment by the Fermi-Dirac statistics. This leads us to a discussion of
rectifying barriers and surface states, not essentially different from that
appearing in the literature, but somewhat more general and unified.
-2-
lack of orthogonality. Wannier I showed that both difficulties can be over-
come by setting up a new set of atomic functions a(q-Qk) (see Appendix I
for their definition), similar to the u's near each atom, but oscillating
and falling off in amplitude like the function (sin x)/x at a distance from
the atom. These functions have the properties that they are normalized, are
exactly orthogonal to each other, and when they are substituted in an ex-
pression like (1) they form an exact solution of the problem of the periodic
lattice. Thus the real solution 4o0 (p,q) can be written in the form
-3-
n (q) = t(Qk) a(q-Qk) (5)
k
That is, we try to find a functionJ 'q) which we can use to modulate the
atomic functions a(q-Qk) to get the correct solution of the problem, replacing
the exponential function (1/11/2) exp [(i/h)p.q] which is used in the problem
of the unperturbed periodic potential, in the solution (2). Our theorem
now states that n(q) satisfies the following differential equation, provided
H 1 varies slowly with position, so that it does not change its value greatly
from one atom to the next:
[E,(-i ) + H1l(q)]N*n(q) = E *T(q) *(6)
Here the first term Eo(-im a/aq) stands as an abbreviation for the differ-
ential operator in which Eo, regarded as a function of the three rectangular
components of the vector p, is transformed by replacing p, by --ih a/ax,
py by -ih /ay, etc., as in the ordinary kinetic energy operator in
Schrodinger's equation. In (6) we then have a Schrodinger equation for
*,(q), in which the perturbative potential H 1 appears as the potential energy,
while the kinetic energy operator is derived from the energy Eo(p) of the
unperturbed problem by replacing p by a differential operator. It is this
theorem which is proved in Appendix I, and which was applied to the problem
of excited energy levels by Wannier.l By means of it, we effectively reduce
the problem of electrons in periodic lattices and additional perturbing
potentials to a problem much like that of free electrons in the perturbing
potential (as we shall show in the next section) and hence make the problem
of electrons in periodic lattices not much more complicated than free-
electron theory.
-4-
--
dix aE 0 d a Eo dz a Eo
Px PM = py ' E = apz-p ' J (7)
and
dP x aH1 dp aH 1 dpz a H1
-= -ax - ay -az
(8)
Both these theorems are familiar, but they are ordinarily derived by much
more involved methods than are used here.6 In (7) we see the formula for
the velocity of a wave packet in terms of the gradient of the function E o
in momentum space, and in (8) the statement that the momentum p of a packet
is governed by the classical equation of motion in terms of the additional
force resulting from the perturbation H 1 . These two results are the basis
of most of the band theory of electrical conduction in solids, but it has
hardly been realized that they form merely the classical Hamiltonian equa-
tions of the Hamiltonian of Eq. (6).
Our next example will be the behavior of wave packets near the bottom
or top of an energy band, and hence the concept of effective mass. At the
bottom of a band, provided the axes are properly oriented, the energy E o0
may be written in the form
2 2 2
+ a +z
E(p) = E1 + x (9)
dx Px dPx a H1
T='mx dt = - ',
(10)
2
d x aH 1
mx dt ax
with similar equations for the y- and z-components, showing that the packet
obeys an ordinary equation of motion corresponding to a particle of mass mx
for the x-coordinate, my for the y-coordinate, and m z for the z-coordinate.
Similarly, near the top of a band, we have:
6. R. Peierls, Zeits. f. Physik, 53, 255 (1929); F. Bloch, Zeits. f. Physik, 52,
555 (1928); A. Sommerfeld and H. Bethe, "Handbuch der Physik", 2d. Ed., vol. XXIV
(Berlin, 1933) pp. 374-375, 506-509; H. Jones and C. Zener, Proc. Roy. Soc. A144,
101 (1934); C. Zener, Proc. Roy. Soc. A145, 523 (1934); L. Brillouin, "Les electrons
dans les metaux du point de vue ondulatoire" (Hermann, Paris, 1934); J. C. Slater,
Rev. Mod. Phys. 6, 209 (1934), pp. 259-262 and Appendix VI; Mott and Jones,
"Properties of Metals and Alloys" (Oxford, 1936) pp. 92-96; W. V. Houston, Phys. Rev.
.7, 184 (1940); and many other references.
-5-
______ I
2 2 2
1 X x (11)
x y z
where these m 's are different from those in (9), but still positive.
Hence the equations corresponding to (10) are
d2x aH1
dx = ax etc., (12)
'x dt
-6-
-- I __
H1
4
~uf~'~s~s~'~f'/i~',%uf
~ ~ El
E
(al
Fig. l(a) Periodic potential, with Fig. l(b) Energy bands and potential
energy bands (shaded). pushed upward by amount H1 (shown in
upper curve). Line of constant energy
E cuts band, at point A, at same rela-
tive height as energy E' in Fig. l(a).
B and C, reversing points of oscillation
of particle.
Our graphical representation of Figure l(b) thus has many of the character-
istics of an energy diagram in classical mechanics, in which potential
energy and total energy are plotted as functions of x, the difference giving
the kinetic energy. When E, in Figure l(b), lies outside any of the energy
bands, the kinetic energy is negative, the wavelength imaginary, and the
wave function is damped off exponentially. When E is inside one of the
bands, kinetic energy is positive, and the wavelength real. A classical
particle of energy E, moving according to the classical Hamiltonian
Eo(p) + H(q), would then oscillate between points like B and C, Figure l(b),
reversing at each point as its kinetic energy and momentum become zero, and
a quantized particle will obey a quantum condition. Such a picture, as
James has emphasized, allows us to deduce the nature of the stationary
states and wave functions in such a problem. If, for instance, the external
field represented by H 1 is constant, so that the energy bands are tilted at
a constant angle, the electron will oscillate back and forth in coordinate
space with a very large amplitude (for a small external field), at the same
time having its position in the energy band go from bottom to top and back
again, in a way familiar in the theory of electrical conduction.
Let us now consider the application of our theorem to cases actually
met in the theory of solids. First we take P- and N-type impurity atoms.
The N-type is a little easier to understand, and we deal with it first. It
is usually an atom substituted for one of the atoms normally present in the
lattice, and containing more valence electrons than the atom which it
replaces; for definiteness, we may be considering an atom of P or As in a
lattice of Si or Ge. If the atom loses an electron, it has enough remaining
electrons to fit properly into the lattice, but it then carries a positive
-7-
)CCUPIED AT
ICTOR
I VALENCE ELECTRON
BAND (OCCUPIED)
-8-
II.
CONDUCTION BAND
(UNOCCUPIED)
VALENCE-ELECTRON
BAND (OCCUPIED AT
OOK)
functions lying at the bottom of the band. In the neutral crystal, there
must be one less electron per impurity atom than the number necessary to
fill the complete valence-electron band, so that at low temperatures this
electron will be missing from the discrete level lying above the band; but
at higher temperatures it will often be missing from one of the continuous
levels lying in the band, leading therefore to hole conduction.
The case of excited energy levels in crystals, which has often been
discussed (see, for instance, Wannier, loc. cit.), is more involved, but
similar to these cases of impurities. It is, perhaps, easiest to under-
stand in the case where a tightly bound, or x-ray, electron is excited to
the conduction band. 7 If an atom of the crystal has lost one of its inner
electrons, it acts approximately, so far as its valence electrons are con-
cerned, like an atom with a nuclear charge greater by one unit; the missing
electron can act like an additional valence electron. Thus the atom tem-
porarily plays the role of an N-type impurity atom, and will set up discrete
energy levels as in Figure 2. When the corresponding emission spectrum is
observed, resulting from an electron in the valence-electron band falling
down into the empty x-ray level, there is the possibility that the electron
may come from one of the levels in the continuum of the valence-electron
band (the possibility considered by Skinner and O'Bryan, discussed in many
references, some given in Ref. 6), but also the possibility that it may
come from one of the discrete levels lying below the valence-electron band,
resulting therefore in a longer-wave radiation than we should otherwise
find. Such tails are observed in the soft x-ray emission spectra, and
Seitz suggests this interpretation of them.
7. The application of the theory to this case, the soft x-ray problem of Skinner and
O'Bryan and other writers, is discussed by F. Seitz, "Modern Theory of Solids",
McGraw-Hill Book Co. Inc., (New York, 1940) pp. 36-438.
-9-
_I I_ I ____ I I
The really optical case of excitation, where an electron is removed
from the valence-electron band to a conduction band, is more complicated,
in that both electron and hole are readily mobile. We may then best describe
it essentially as Wannier did in Ref. 1, and as Frenkel had done earlier.
Considering it classically, the electron in the conduction band attracts the
hole in the valence-electron band, and, since each has a comparable effective
mass, they each execute hydrogen-like orbits about their center of mass,
resulting in certain discrete states. These discrete states lie below the
continuous states; that is, the electron effectively is at the bottom of the
conduction band, like the discrete states in an N-type semiconductor as in
Figure 2, while the positive hole is at the top of the valence electron band,
like the discrete states in a P-type semiconductor as in Figure 3. The
electron and hole together form a stable structure, however, which, because
of the mobility of both electron and hole, is free to wander through the
crystal, forming what has been called an exciton. (See Ref. 8.) Being a
neutral structure, it carries no current. Less energy is required to set
up such an exciton than to raise an electron from the valence-electron band
to the conduction band, leaving both electron and hole dissociated from each
other and free to move, so that the wavelength for absorption to this exciton
level, which does not result in photoconductivity, is longer than for the
limit of photoconductivity. It is well known that such exciton levels exist,
for instance, in the alkali halides. 9
8. J. Frenkel, Phys. Rev. 37, 17 (1931); 37, 1276 (1931); Phys. Zeits. d. SowJetunion
2, 158 (1936).
9. See, for instance, J. C. Slater and W. Shockley, Phys. Rev. 50, 705 (1936), in which
the theory of the exciton is considered without benefit of Wannier's theorem.
-10-
Hartree: we can demand that the potential energy of an electron in this
potential be the same quantity Hl(q) which is responsible for perturbing
the energy bands.
To set up our self-consistent condition, we must first find the net
charge density as a function of position arising from our assumed energy
bands with the assumed Fermi level. First we shall find this in the case
of the unperturbed periodic potential. In this potential, let the number
of energy levels per unit volume, in the energy range dE, be n(E) dE; this
can be found, as is well known, from the volume of momentum space lying
between surfaces Eo(p) = E and Eo(p) = E + dE, since states are distributed
in momentum space with uniform density in the periodic case. In forbidden
bands of energy, n(E) is, of course, equal to zero. Let No be the number
of electrons per unit volume necessary to render the crystal electrically
neutral. Then the excess number of electrons per unit volume, with an
arbitrary value of EF, is
-11-
-- -
____1_11_ I _I I_ _____ __1_11
of the discrete energy levels, but merely handles them in a statistical
or averaged-out way. Our method differs from the Thomas-Fermi method in
three respects: we are handling our kinetic energy by the energy-band
method, so that it is given by Eo(p) instead of the usual expression; we
are dealing with a modulating function ln(q), instead of with the actual
wave function (q); and we are handling our statistics in a form appropriate
to an arbitrary temperature, rather than for the absolute zero of tempera-
ture as is done in the ordinary Thomas-Fermi method.
When we make the assumption above, we can write Poisson's equation
in the following form:
where we have used the fact that EF must be constant over all space, to
satisfy the condition for thermal equilibrium in the Fermi statistics, so
that its Laplacian is zero. In Equation (14), supplemented by Eq. (13) for
the function N, we have the general formulation of the problem of setting
up the electrostatic potential within a solid in thermal equilibrium. This
equation has, of course, been used and solved in special cases by many
writers. It has essentially been used by Schottky1 0 in an extensive series
of papers, and is similarly used by Mott and Gurney, and by Bethe. 1 1 Fan 1 2
has carried out careful studies of the contact between metals and between a
metal and a semiconductor, which are complete enough so that many of our
results will be merely a restatement of some of Fan's conclusions. Quite
recently, Markham and Miller1 3 have used essentially similar methods in
closely related problems. Many other writers are aware of these methods of
handling the problem. To give a complete picture, we shall state some of
the methods of solving this equation, and some of the applications to well-
known cases, as well as some new aspects of the problem.
The nature of the solution depends on the form of the function N(t).
In Figure 4 we show this function for two familiar cases: the metal and the
intrinsic semiconductor. In the first case, N increases very rapidly as
departs from the value appropriate for no charge, and over a considerable
range it can be treated as proportional to -- o, where to is the value
associated with no charge. In the semiconductor, however, N increases very
slowly with --
f , behaving approximately as a hyperbolic sine, although
10. W. Schottky, Zeits. f. Physik, 118, 539 (1941); other references quoted in this paper.
11. N. F. Mott and R. W. Gurney, "Electronic Processes in Ionic Crystals", (Oxford, 1940),
Chapter V; H. A. Bethe, M. I. T. Radiation Laboratory Report 43-12 (Nov. 23, 1942).
12. H. Y. Fan, Phys. Rev. 62, 388 (1942).
13. J. J. Markham and P. H. Miller, Jr., Phys. Rev. 75, 959 (1949).
-12-
__ I_
NIM)
t f
- - ) r tl P \Nsl s ,s
r. '+ I ) s ULIC .utl uVI . Aouv e
NI meal; De.ow, nLrlns lc semlconaucLor.
I
when becomes so large, negatively or positively, that the Fermi level
penetrates either the lower valence-electron band or the upper conduction
band, N begins to get very large, negatively or positively as the case may
be. If N( ) can be approximated as a(-to) (where a is a constant) as it
can over a considerable range in these two cases, then Eq. (14) takes on
the mathematical form of the wave equation, and solutions can be set up by
familiar methods. Thus, if we are dealing with a one-dimensional problem,
we have solutions - = exp ( x/X), where X = (e/ae2) 1/ 2, and where x is
the coordinate in the direction in which the potential is changing. As
shown by Fan,1 2 this quantity X for a metal is very small - of the order
of magnitude of an Angstrom unit; on the other hand, for an intrinsic semi-
conductor it becomes large and in the limit of zero temperature for this
case it becomes infinite. For a three-dimensional problem of spherical
symmetry, we similarly have --o = exp (+ r/X)/r, where r is the distance
from the center, and X is as given above.
We can now examine several applications of these simple results. First
we consider the metal, and the one-dimensional problem; this is met in
investigating the surface charge at the free surface of a metal in an ex-
ternal electric field. Clearly, we can adjust boundary conditons so that
the electric field at the surface resulting from the gradient of H 1 has any
desired value, and yet the resulting potential will penetrate into the metal
only to a distance of the order of a single atomic layer, with the related
charge density confined to this same small depth. This allows us, in other
words, to assume an arbitrary surface charge on the surface layer of atoms
of a metal, of suitable amount to match any external boundary conditions.
Similarly, if two metals are close to each other and connected electrically,
so that they must bear surface charges enough to produce the difference of
-13-
-14-
impurity semiconductors, however, it is usually more convenient to treat
average behavior over a volume which is small compared to the thickness of
a rectifying barrier, but large compared to the distance between impurity
atoms; or, alternatively, it is better somehow to average the impurity
levels, so that we do not have to consider the fine-grained inhomogeneities
arising from the discrete impurity atoms. When we do this, we have a dif-
ferent distribution of energy levels, n(E), for now we include an appropri-
ate number of levels per unit volume arising from the discrete levels in
N- or P-type impurity atoms, and hence located just below or just above
the continuous bands. Also in computing the net amount of charge per unit
volume, we must take account of the charge furnished by these impurity ions.
When we do this, we may get for N(t) for, say, a material containing N-type
impurity atoms uniformly distributed, a curve of the nature shown in
Figure 5.
OF CONDUCTION
BAND
-15-
_ 1__1_1_1 1.-_-_1-1-1-
---- PI(------11-1^_
____I I __
soon acquires metallic properties.
With as complicated a function N() as is given in Figure 5, it is
clear that the simple approximation used previously (of setting it propor-
tional to -) is inadequate, and we must use the whole form of the func-
tion. Even in this case, in the one-dimensional problem, we can integrate
Poisson's equation, (14), as is done, for instance, by Mott and Gurney ll
and by Fan.12 The equation becomes d2 /dx2 = f(), where f(t) is a function
of . Mathematically, this is similar to a one-dimensional equation of
motion in mechanics, m d2 x/dt2 = f(x), where f(x) is the force; and, as in
the mechanical case, we can integrate by a method entirely equivalent to the
energy integral in mechanics. When we do this, carrying out the integrations
numerically if N(Q) is as complicated as in Figure 5, we can find the rela-
tion between and x. We find that there are solutions approaching o
asymptotically for large values of x, one for greater than to, the other
for less than ,o; all such solutions differ from each other only by uni-
form translation along the x-axis. In Figure 6 we show such a solution for
the case coming into the problem of a rectifying barrier at the surface of
an N-type semiconductor.
t -t
There are sections of the curve of Figure 6 which are associated with
the various parts of Figure 5. As departs only slightly from o, we are
in a linear part of Figure 5. Here we have the Laplacian of -- 0 equal to
a constant times --0, so that we can set up an exponential solution for
--to as a function of x, just as we did with the metals. We find, however,
a much slower exponential variation, extending over much greater distances.
Next we have the region where is considerably less than to, so that we are
in the region of the depletion layer. In this region, as we have mentioned
earlier, N( C) is practically constant, and our solution agrees exactly with
Schottky's,1 0 leading to as a quadratic function of x. This parabolic
-16-
____
function, of course, fits smoothly to the exponential function which holds
as approaches its asymptotic value to. Finally, when gets so small that
electrons of the lower band begin to be removed, begins to decrease very
rapidly with change of x. This is because we are now meeting high positive
volume charges, arising from the emptying of this lower band, and the elec-
tric field can change with position about as rapidly as in a metal (where
we have already seen that we can accommodate a large enough charge in a
layer of atomic thickness to be equivalent to a surface charge). This hap-
pens here as penetrates into the lower band, and the reason for its happen-
ing is essentially the large reservoir of charge available, similar to the
case of a metal. In the nature of things, however, this surface layer of
charge can only be positive, since it arises from holes in the valence-elec-
tron band. If there is a negative surface charge, the curve corresponding
to Figure 6 must rise rather than fall at the surface of the material.
The result shown in Figure 6 can now be used to discuss the boundary
between a metal and an N-type semiconductor, or between such a semiconductor
and a vacuum. If we used this result straightforwardly, we should draw the
following conclusions. We should conclude that at the boundary between a
semiconductor and a vacuum, in the absence of an external electric field,
the potential in the semiconductor would be constant. If an external field
were impressed, then the surface charge to terminate the lines of force
would be actually distributed through the whole depth of a depletion layer,
instead of being located on the surface, as in a metal. The only exception
would come if the field were so strong that the Fermi level dipped down into
the valence-electron band at the surface; then any remaining charge required
to terminate the lines of force would be located almost exactly at the sur-
face. In other words, the total amount of charge which can be distributed
through the interior of the volume in the depletion layer is limited. In
such an extremely large field, the energy bands would look as in Figure 7,
and the external field would be indicated by the slope of the curve to the
-17-
LEVELS
OF VALENCE
BAND
BAND OF METAL I
DISTORTION OF CONDUCTION BAND OF
METAL ON ACCOUNT OF SURFACE CHARGE
the top of the valence-electron band of the semiconductor just at the sur-
face, and the resulting situation would be almost independent of the work
function of the metal, provided only that it was different enough from that
of the semiconductor to require a large enough double layer.
It is well known that the situation we have just described does not
fit the observations, at least in germanium and silicon, two semiconductors
which are very well understood as a result of the large amount of work done
on them during the war and since at Purdue University, the Bell Telephone
1 4
Laboratories, the University of Pennsylvania, and elsewhere. Meyerhof
in a set of measurements on contact difference of potential between silicon
and metals found definite evidence that our simple picture is wrong, and
-18-
I ___
his effect was explained by Bardeen1 5 with his theory of surface states.
Since then, the group at the Bell Telephone Laboratories has arrived at
substantially the following conclusions regarding the surface of germanium,
explainable in terms of surface states.l 6 At a free surface between germa-
nium and air, there is good evidence that in the absence of an external
field there is, nevertheless, a well-formed positively charged depletion
layer below the surface, compensated by an equal negative surface charge.
In an external electric field, the extra surface charge required to termi-
nate the lines of force appears just on the surface, rather than in the
depletion layer in the interior. And when contact is made between the semi-
conductor and a metal, the required double layer adjusting the Fermi levels
of the two to coincidence is made up of a surface charge of the usual sort
at the surface of the metal, and a surface charge of opposite sign on the
surface of the germanium.
It thus appears that the surface layer of atoms on a germanium crystal
must behave differently from the interior and, to explain this, Bardeen
introduces the idea of surface states. The action can be described, roughly,
as if there were part of a monomolecular layer of metal on the surface of
the germanium, whose work function differed from that of a hypothetical
germanium which lacked the surface states by something like the amount con-
sidered in Figure 8. This layer of metal would have to acquire enough nega-
tive charge to raise its Fermi level - normally far below that of the ger-
manium - up to equality with that of the germanium. Having the large
reservoir of electrons characteristic of a metal, it could acquire any other
amount of surface charge necessary to compensate for an applied external
field, or for a double layer arising when another metal made contact with
it. Thus the rectifying barrier would remain something like that of Figure 8,
which is not unlike what is observed (see, for instance, Ref. 16), inde-
pendent of the material of the metal making contact.
In this description of the surface, which is substantially that sug-
gested by the group at the Bell Telephone Laboratories, it is not clear
whether the surface states arise from real impurities on the surface (either
metallic or, at any rate, setting up a distribution of states of the sort
characteristic of a metal), or whether they are inherent in the germanium
itself. The evidence indicates that the surface states depend on surface
-19-
VI. Acknowledgments
Most of the material described in this paper was prepared for presenta-
tion at the M. I. T. Conference on Physical Electronics in April 1949, and
at a preliminary session before it, and the writer is indebted to Prof.
W. B. Nottingham for the stimulus to prepare it, and for valuable discussion.
He is also indebted to Prof. K. Lark-Horovitz, Prof. H. M. James, and
Dr. H. Y. Fan, of Purdue University, for valuable comments and much infor-
mation about the semiconductor program at Purdue; and to Dr. J. Bardeen,
Dr. W. H. Brattain, and Dr. W. Shockley, of the Bell Telephone Laboratories,
for valuable comments and suggestions.
-20-
Appendix I
a(q-Qk) 1
iN-/ 2 exp[()pQk] o(P,q) , (1A)
P
where the sum is over all the quantum states p. These functions are the
normalized atomic functions of Wannier, written in our notation; we refer
to Ref. 1 for their properties. Equation (2) follows at once from (1A) by
multiplying by exp [(i/h)p'-Qk], summing over k, and using the theorem
>1
k,p,p'
Tln(Qk)exp [(~)(P' 'Q-PQk)]} f *(pq)HOW(p,q) dq
-21-
_ _ I _ _ _I _I II_ 1 111 I _ I
We use (3), and the orthogonality of the functions To, for different p's,
to show that f Po*(p',q)Hooo(p,q) dq = E(p)(p',p). When we substitute
this above, we have
Now we expand *n(Qm-Qs) in Taylor's series about the point Qm. We have
4i()(Qs) 2
dCidq (Qm)(Qs )
+
~n( Q~ ) 3 i() A +
Inthis
exp (-written
only the case where q is a scalar quantity,
In this expression we have written only the case where q is a scalar quantity,
but an exactly analogous form holds if it is a vector, Qs(d/dq) being replaced
by the scalar product Qs*V, where the V operator denotes vector differentia-
tion with respect to the components of the vector q. We now use the result
(6A) to modify (5A), and it takes on the form
, A(Qs) [exp (-Qs*V)]n(Qm)
8
-22-
Comparison with Eq. (4A) shows that this is what we should get if we took
Eo(p), replaced p in it by the operator (/i) V, and allowed this to operate
on~n( t). When we combine this with the expressions for the terms in H1
and En in Eq. (2A) which we have already discussed, we see that we are led
to Eq. (6), which we wished to prove.
-23-
_ _ 1_ I