Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

ANNALS OF PHYSICS 81, 567-590 (1973)

Solution of Schrijdinger and Related Equations


for Irregular and Composite Regions

LEONARD EYGES

Air Force Cambridge Research Laboratories, L. G. Hanscom Field, Bedford, Massachusetts 01730
Received April 30, 1973

The Schrijdinger equation with a potential is mathematically equivalent to the


Helmholtz equation with a spatially variable propagation constant. A new method is
presented for solving certain standard problems associated with these equations. In
the Schrodinger language these are the ones in which a potential u(l rl) acts inside an
irregular (aspherical) boundary, where r has its origin inside the boundary. In terms of
the Helmholtz equation, these include problems in which a region of constant index of
refraction, but arbitrary shape, is embedded in a second uniform region with a different
index. It is shown how bound state and scattering problems for such a potential (or
region) can be treated in a way that avoids the usually intractable problem of matching
solutions across the irregular boundary. The method requires, in general, the truncation
of an infinite set of equations for partial wave amplitudes. The special case is discussed
of a potential that becomes infinite throughout a region, so the wave amplitude must
vanish inside the region (and, hence, on its boundary). For a long wave length this
becomes a problem with the Laplace equation, and the general technique is illustrated
by a calculation of the free charge on a perfectly conducting spheroid. The theory is
extended from a single potential to an ensemble of such potentials, and in particular
to an ensemble of potentials with spherical boundaries. In the special case that the poten-
tials are arranged in a periodic lattice the formulas resemble those obtained by the
KKR method, but are simpler in some ways. The method is extended to an ensemble of
irregular potentials, and these results are shown to be. applicable to the special case of an
ensemble of finite range, but overlapping, spherical potentials.

1. INTRODUCTION

The SchrGdinger equation1 for a particle in a potential u(r),

is a paradigm for many other equations of mathematical physics. For example,


it is equivalent to the equation for the propagation of the spatial amplitude 4 of

1 We take units with ti = 2m = 1.


567
Copyright 0 1973 by Academic Press, Inc.
All rights of reproduction in any form reserved.
568 EYGES

a time harmonic wave in a medium with a spatially varying propagation constant


(squared) k2(r). This amplitude satisfies (V2 + k2(r))# = 0, and if k2(r) approaches
the constant value ko2 at infinity this equation can be written

(V2 + k2(r) - k,2)# = -k,2ys. (2)

Equation (2) is equivalent to the Schrodinger Equation (1) if we make the identi-
fication
E--j. ko2, v(r) + ko2 - k2(r). (3)

Moreover, if we consider an infinitely repulsive potential (v(r) -+ co throughout


a bounded region) the wave function must vanish inside, but not, of course,
outside the potential. Since the wave function must be continuous across the
boundary (even for infinite potentials) it follows that the wave function vanishes on
the boundary and the SchrSdinger equation in this limit is equivalent to a Dirichlet
problem with the Helmholtz equation. If, moreover, we let k -+ 0, the Dirichlet
problem becomes one that is frequently encountered for the Laplace equation.
With any of the above equations, it is impossible to handle analytically problems
with completely arbitrary potentials, propagation constants, or boundaries. Hence,
the canonical problems of the literature are ones with great symmetry-usually
spherical or cylindrical. In this paper we go one step beyond these to handle
certain problems with a single aspherical boundary (irregular region) or with an
ensemble of such boundaries (composite region). We shall use the language of the
Schrodinger equation because of its generality. Our results can, however, be
extended in a more or less obvious way to the special cases of that equation
mentioned above. For clarity, we spell out this extension at various points in this
paper.
We consider then bound state and scattering problems with the Schriidinger
equation, for a potential that acts only in a finite region of space, i.e., is zero
outside some bounding surface. A standard way of solving such problems is to
write (if possible) a solution involving unknown coefficients in the interior of such
a region and then determine the coefficients by matching this solution at the
boundary to the known exterior one. In practice, this technique has mainly been
limited to spherically symmetric potentials with spherical boundaries for two
reasons. First, it is only for such potentials that one can hope to find general
analytic interior solutions. Second, it is only for spherical boundaries that one can
apply the matching conditions.
It may appear tautological to speak of spherically symmetric potentials with
spherical boundaries, since the former would seem to imply the latter. But there is
a sense in which a spherically symmetric potential need not necessarily imply a
spherical boundary. Given a volume, and a position vector r defined with respect
SCHR6DINGER AND RELATED EQUATIONS 569

to an origin inside it, a potential u(r) which is a function only of the magnitude
r = 1r 1 may act in the volume. If the boundary of the volume is irregular
(aspherical) we have an example of what is meant by a spherically symmetric
potential in an irregular volume.2 Problems with such potentials can be of consid-
erable physical interest. An important example of a spherically symmetric potential
is one that is constant inside a region. With the identification of Eq. (3), i.e., in
terms of a scalar time-harmonic wave, this potential corresponds to a region of
constant index of refraction. The problem of the scattering of a time harmonic
wave from a region of uniform index, but arbitrary shape, is, thus, a problem with
spherical symmetry inside an irregular volume.
Problems with potentials that are nonzero only inside boundaries may be
untractable, as we have noted, because one cannot find solutions inside the
boundary, or because one cannot match even known solutions across an irregular
boundary. In this paper we remove the latter restriction, by developing a method
which avoids the matching problem, assuming the solutions are known inside. We
treat essentially two cases. First, we consider problems with a potential which
acts inside a single irregular boundary. Second, we consider similar problems when
there is an ensemble of such potentials (and boundaries). Thus, after this intro-
duction and the presentation in Section 2 of the basic idea, Section 3 is devoted
to a single irregular potential. Section 4 extends the treatment to the case when the
potential becomes infinite, which essentially defines a Dirichlet problem for the
potential boundary. In Section 5 the latter results are exemplified by the problem
of the free distribution of charge on a perfectly conducting spheroid. In Sections 6
and 7 we consider ensembles of potentials. In particular, in Section 6 we turn to
the case of an ensemble of spherical boundaries. The formulas we obtain are
simpler in some ways, but equivalent to, the usual formulas that are found by
matching techniques [l, 21. Finally, in Section 7 we discuss the case of an ensemble
of irregular boundaries. The general formulas that are derived can be applied to
yield a good approximation to the solution of the problem of an ensemble of
spherically bounded potentials that are ofjinite range but that overlap.

2. THE BASIC IDEA

There are two kinds of problem with Eq. (1): homogeneous (bound state) and
inhomogeneous (scattering). It is well known that either of them can be formulated
as an integral equation. In terms of a Green function gE(r, r’), satisfying

(Vz + E) g&, r’) = S(r - r’),

z The total or global potential distribution is of course not spherically symmetric.


570 EYGES

the integral equation for the inhomogeneous case is

#(r> = de(r) + / W) W) gdr, r’) du’,


where & is the external (applied) field. For the homogeneous case, #e = 0, and
the energy is taken to be negative.
The differential and integrals Eqs. (1) and (5) are complementary in a certain
sense. Namely, for potentials acting only in a bounded region the differential
equation demands that one consider the space exterior to the region, since the
solutions there must be matched to the solutions in the interior. With the integral
equation, on the other hand, one can consistently confine attention to the interior.
For example, the integral equation can be broken up (in principle) into an
(approximate) set of linear equations for the amplitudes of # at various points in
the interior of the potential. The solution of these equations effectively solves the
problem, since from knowledge of the interior wave function the exterior wave
function can be calculated in a trivial way. Thus, it is the virtue of the integral
equation that it does away with the matching problem. It is the virtue of the
differential equation that it may provide partial information about the intreior
wave function. The present method exploits both these virtues at once; it uses
assumed known solutions of the differential equation inside the potential, in
conjunction with the integral equation, in a way that avoids matching techniques.
Specifically, suppose the differential equation yields the form of + inside the
potential in terms of certain known functions xn , but with unknown coefficients A,, ,
say

W = ntl -%xnW.

If we put this into the integral equation and evaluate the resulting equation at N
suitably chosen points rl , r, ,..., r, it becomes a set of inhomogeneous equations
for the A, in terms of the known quantities x&J and $,(r,),

f AnxR(rS) = h4rs> + s (f A,xdr’)) W) g&r,, r’) do’, s = I, I&..., N.


?Z=l L=l

For the homogeneous case this procedure yields an homogeneous set of equations
for the coefficients, with a corresponding secular determinant for the eigenvalues.
As an illustration of these remarks consider a simple example: The scattering in
one dimension of a plane wave eikx incident from the left on the square well
potential: Y(X) = --aO, 0 < x < d. Inside the potential the solution is of the
above form with two coefficients A, and A,
aj(x) = A,,eik’= + Alecik’“, (6)
SCHRijDINGER AND RELATED EQUATIONS 571

where V2 = E + u0 . The integral equation (5) for the present one-dimensional


case is, with g, = (l/2&) eiklz-x’l,

$(x) = eikx - (u,/2ik) lo’ #(xl) eiklz--s’l dx’. (7)

The two expressions (6) and (7) must be identical inside the potential so we have
for 0 < x Q d,
AOeik’x + Ale-ik’x = eikx _ q,/2ik 1’ (Aoeik’“’ + A,e-ik’“‘) eiklz-x’l dx’.
0

Evaluating this at, say, x = 0 and x = d, we get two equations for the two coeffi-
cients A, and A, and once these are found the problem is essentially solved. The
standard results are, of course, obtained.
As an example of the homogeneous problem consider the ground state of a
particle in the square well potential: v(x) = -u. , I x / < a. The integral
equation (5) is, with +e = 0 and /? + ik,

#(x) = uo/2/3 1-1 #(x’) e-61z-s’~ dx’.

From the differential equation, the symmetric (ground state) solution inside the
potential is # = A cos Xx, with X2 = u. - /I”. This must be the same as the
solution predicted by the integral equation so we have

A cos hx = v,A/2/3 /:a cos hx’ ec61z-“l dx’.

Evaluating this at, say, the convenient point x = 0 and doing the integral we find
the eigenvalue condition
h tan ha = /3.

This is, of course, just the condition that is obtained conventionally by matching
the interior solution to the exterior solution B exp(-/3 1 xl).
The above discussion is readily extended to a single spherically symmetric
potential in three dimensions. We have in fact previously considered the bound
state case [3] so we shall merely sketch the derivation. For this problem the
Green function that enters (5) is [4], with E = k2,
gk(r f) = 1 eiklr-”
47r 1r - r’ ( = -ik
> c r&r, r’) Ylz(P’) Y,,(S), (8)
Z.WZ

where, if r<(r>) is the lesser (greater) of r and r’,


rdr, r’) = jdkr,) AZ&r,). (9)
572 EYGES

Let the incident wave ds be the plane wave eikar; for reference we write the well
known spherical wave decompositions of it

eik” = 4a F ilj,(kr) Y,(C) Y,*(k). (10)

The solution for # inside the potential is

* = 7 AL&k) YLW’), (11)

where Rz satisfies

k=) Rt = 0. (12)

Now we put Eqs. (8), (lo), and (11) into (5) to find

c A,R,(r) Y,(F) = 4n c iy,(kr) YJP) YL*(&)


L L

- ik 1 (; AL,&W YL@'))

x u(r’) (z rc-(r, r’) Y,-(P) Y:*(F)) rt2 dr’ dJ’2’. (13)

We equate coefficients of YL(P) and do the integrals over the solid angle dQ’, and
Eq. (13) becomes

ALRl(r) = 4nilj,(kr) YL*(6) - ikAL low R,(r’) u(r’) TJr, r’) r’2 dr’. (14)

Suppose now that the differential equation (12) is soluble, so that the form of Rt
is known. Conventionally, the coefficients A, would be found by matching the
interior solution ALRl(r) YL(P) for the I-th partial wave to the exterior solution
at the boundary. In the present technique we impose the condition that Eq. (14)
be satisfied at one point, which we take to be the origin. This procedure is super-
ficially complicated by the fact that Rc behaves as rz for small r. This complication
is easily bypassed. We write

&(r) = rzQdr), (15)


where Qr is finite at the origin and can be taken to be unity there

QdO>= 1. (16)

3 We shall henceforth let L stand for the index pair 1, m.


SCHR6DINGER AND RELATED EQUATIONS 573

We put Eqs. (9) and (15) into (13) and have

AglQ,(r) = 4rriljL(kr) YL*(@ - ikAL /h,(kr) jar Nr’) U(r’) jt(kr’) r12 dr’

+ j,(kr) [” R,(r’) u(r’) h,(kr’) rf2 dr’l. (17)

We can now multiply this equation by r-z and go to the limit r + 0. We use the
forms valid for small x: j,(x) R+ x2/(21 + 1) ! ! and h, w --i(21 - 1) ! !/x1$-l, with
(21 + I)!! = 1 . 3 . 5 ..- (21-t I), to find

A = 44Wz YL*(@ _ (2zizj! ! AL lorn R,(r’) v(r’) h,(kr’) r’2 dr’.


L (18)
(21$- I)!!

For the homogeneous case the first term on the right side of Eq. (18) is zero and k
is replaced by i/l so that Eq. (18) becomes

&(r’) u(r’) h,(i@‘) rr2 dr’,


l = (21 + l)!!

a result which has been derived previously [3]. Equations (18) and (19) are equiv-
alent to those obtained (for finite range potentials) by matching interior and
exterior solutions at the boundary. We verify this remark for a special case.
Take the square well: u(r) = -v,, , r < u, and consider Eq. (19) for the bound state
case. The interior solution properly normalized to satisfy (16) is

R,(r) = j&)(21 + 1) ! !/X2,

where A2 = u0 - fi”. The integral in Eq. (19) can be done by adapting a formula
from the theory of Bessel functions to find [4]

a l#?r) j,(b) r2 dr = (a2/f12 + X2)(i/3h,-,(i~u) j&z) - xh,(i&z) h.,(h))


s0

- iXz/(/32 + X2)(i/?))E+1.

With this result and a recursion relation for a spherical Bessel or Hankel functionf, ,
that reads @l-I(e) = (I + I)&(,$ + ,$ dfldf, we find that, for example, the
condition (19) becomes, with a prime denoting differentiation with respect to the
argument,
Aj,‘(Au) h&%2) = i/3jl(Xa) hl’(i/3u).

This is the standard secular equation for square well bound states [5]. Equation (18)
similarly yields the standard result for scattering from square wells.

595/81/2-14
574 EYGFS

3. SINGLE POTENTIAL WITH ASPHERICAL BOUNDARY

In this section the above results are extended to include potentials with aspherical
boundaries. As outlined in Section 1 we assume that a potential o(r) is nonzero
only in some arbitrary irregular (aspherical) region V. It is assumed also that
solutions of Eq. (12) for R,(r) can be found (possibly numerically) so that we can
write, inside V,
#9 = s AL&(r) YL@), (20)

where the AL are unknown. We again first consider the scattering problem. We
assume the plane wave & = eik.r is incident on the volume and put Eqs. (Q-(10)
and (20) into the integral equation (5). We can equate coefficients of Y, as before,
to find

ALR1(r) = 47rPj,(kr) Y,*(k)

- ik Iv (F A/R1’(r’) Y;(F)) u(r’) Tt(r, r’) Y,*(F) rf2 dr’ dL?‘. (21)

We cannot now, however, integrate over dl2’ and use spherical harmonic orthogo-
nality since the integrations over dr’ and dsz’ may be interdependent; for a given r
there may be angles for which the potential vanishes, and others for which it does
not. We can, however, evaluate Eq. (21) near the origin, as before, and so we
rewrite it as

ALR,(r) = 4rrizjl(kr) Y,*(k)

- ik /h,(kr) f dll” IO’ (c AL)Rl’(r’) Y,‘(F)) o(r’)j$(kr’) Y,*(F) r’2 dr’


L’

‘(g ALRl’(r’)
+ j,(kr) 1 dsZ’ /TBoundaryof Y/(P)) u(r’) h,(kr’) YL*(i.‘) t-l2 dr’l.

(22)

For small r we have R, = rlQl as in Section 2. The first term on the right side of
Eq. (22), the inhomogeneous term involving j,(kr), also goes as rt. In the first term
in curly braces, involving the integral from zero to r, the angular and radial
integrations are independent, for small enough r. Therefore, the integral over solid
angle collapses the sum under the integral sign to a single term involving R,(r’). For
small r, then, this whole first term in the braces goes as rzf2 as before. In the
second term in braces, involving integration to the boundary of the volume, the
essential r dependence is contained in the factor j,(kr) that stands in front of the
SCHRijDINGER AND RELATED EQUATIONS 575

integral so that for small r the second term in braces goes as rl. In the limit that
r + 0 Eq. (22), therefore, becomes, on using Eq. (IQ,

AL = 47+W y,*<Q _ ikz+l CL’ &(hY 1Ic)LL’ ,


(2z+ l)!! (23)
(21$- l)!!

where , with dv’ = rf2 dr’ dl.2 ,

@Y
I&L’
=s V
h,(kr’) Y,*(F) u(r’) R,,(r’) Y&+‘) dv’. (24

This is now a coupled set of equations for the AL and to solve it the number of L
values must be limited by truncation of the set. In practice this means that the
wave length cannot be very short by comparison with the size of the volume. But
also in practice this truncation will often be abetted by the fact that if the volume V
is symmetric in some way, not all spherical harmonics, but only certain linear
combinations will be allowed. This aspect of the problem is a group theoretical
one that we do not discuss. Equation (23) becomes a set defining bound states
if we set to zero the inhomogeneous term involving Y,*(k) that derives from the
incident plane wave and let k + ifl. In the special case that the volume Y is
spherical this bound state formula reduces, as it must, to the eigenvalue
equation (19), for the states of a spherically symmetric potential.

4. IMPENETRABLE VOLUME WITH IRREGULAR BOUNDARY

An important scattering problem with the Helmholtz equation is that in which


the wave function is prescribed to vanish on the closed surface enclosing some
volume (impenetrable volume). This is a special case of a general Dirichlet problem;
in the general problem the wave function is assigned arbitrary values on the surface.
This latter problem is usually formulated in terms of Green’s theorem, that relates #
inside the volume to # and a$/& over the bounding surface. As we have remarked
in the introduction it is more useful to view the present problem as one with a
potential that becomes infinite. From this viewpoint the technique for handling
the problem is readily seen as an extension of that for finite potential problems.
Consider first a one-dimensional scattering problem in which a wave eikr is
incident on a potential barrier v(x). The amplitude + satisfies the one-dimensional
version of Eq. (5)

g(x) = eikz + 1 1,4(x’)u(x’) gk(x, x’) dx’, (25)


576 EYGES

where g, = (iik) eiklz-2’I. It also satisfies the one-dimensional Schrodinger equation

-d=t,G/dxa + u(x)+ = kZ#. (26)

Consider now the special case of a step potential: v(x) = 0 for x < 0, v(x) = ZQ,
for x > 0. The solution of the scattering problem with eikz incident is elementary.
If k’2 = kZ - vO, the wave function to the left of the potential step (x < 0) is
eikx + Re-ikx with R = (k - k’)/(k + k’), and the wave function to the right
(x > 0) is Teik’2, where T = 2k/(k’ + k). These forms are easily verified to satisfy
the integral equation (25). Note that this equation holds everywhere, i.e., on both
sides of the boundary x = 0.
Consider now what happens when v,, gets large. The wave function for x > 0
is then approximately T exp(-vi/*x) and in the limit that 2r0-+ co it vanishes.
This does not mean, however, that there is no contribution from the integral in
Eq. (25). There is a contribution, but it becomes a singular one that is generated
only at the point x = 0. To see this, first integrate the general equation (26) from
a point --E just to the left of the origin to a point fc just to the right to find

4Wx 16- 4W 1-c = j-’ 4x) #(x) dx - kZ j-’ #(x) dx. (27)
--E -c

To apply this to evaluate the integral in Eq. (25) for the infinite step potential note
that integral has a contribution only near the origin, so we can write

j ~(x?
4~‘)
gdx,
x’>
dx’=ss,
W’>W)gdx,
x’)dx’
- gk(x, 0) j-;c $&‘) 0(X’) dx’. (28)

Now we can use Eq. (27), noting that d#/dx It. vanishes in the limit of infinite
potential, as does J?S #(x) dx, in the limit that E + 0. In this double limit, then,
Eq. (27) when inserted into Eq. (28) yields

s #(x’> @‘) g& x’) dx’ = -d#/dx I_ g&(x, 0)

so that the integral Eq. (25) assumes the degenerate form

#(x) = eikz - d#/dx I_ &(x9 0).

To the left of the boundary x = 0 this yields the obviously correct solution
1/1= eikx - e-(kx; similarly to the right it yields the correct answer # = 0.
We now consider the scattering in three dimensions of the plane wave eik.r from a
SCHRi)DINGER AND RELATED EQUATIONS 577

constant potential u,, that is assumed to exist throughout some irregular volume,
and ask what happens to the integral equation (5) in the limit that z+,-+ co. The
argument is much as in one dimension. Again the wave function will vanish inside
the volume. Consider then a small element dS of surface that, with respect to an
origin in the interior, defines a conical volume C. We want to calculate the integral
of Eq. (5) over that volume. In the limit that u,, gets large there is a contribution to
that integral only from points near the surface. Let 1 be a coordinate (analogous
to the one-dimensional x) that measures distance into the potential along the
direction of the normal to dS. In this direction Eq. (1) becomes essentially one
dimensional and as in one dimension we can integrate it once to find that

s, vvkdr, r') dv' = s, v#g,(r, r') dS' dl

is approximately replaced by

Here n is the outward drawn normal from the volume and rS is a vector to the sur-
face. Integrating over the whole surface, Eq. (5) becomes

#(r) = eik*r + 1 (~~/hz) &(r, rs) dS. (2%

This is, of course, the standard integral equation for that special Dirichlet problem
in which $ = 0 on the surface. This equation is usually derived from Green’s
theorem, but the point of the present heuristic derivation, aside from making
contact with the potential case, is to show more clearly than the standard derivation
that the integral equation (29) holds both inside and outside the boundary.
We can, therefore, apply Eq. (29) inside the volume. But since we have just seen
that 1,4vanishes there, we have for r inside the volume

0 = eiknr + S (i+b/ih) gk(r, rJ dS. (30)

This is an integral equation in which &&an, a function of the surface vector rs ,


can be considered to be the unknown. If, however, rs is a single valued function of
the angle variables we designate by QS , i.e., if no radius vector cuts the surface
in more than one point, we can consider at,b/an to be a function of Sz, . It is then
convenient to replace the surface integral in Eq. (30) by an integral over Q, . We
write dS = (dS/dS2,) di2, and define x(sZ,) by

)#&) = a#/an dSld& . (31)


578 EYGES

Then Eq. (30) becomes one for the unknown ~$2~)~

0 = eik-+ j x(.Q,)
&, rJ a,. (32)

To solve this equation for x we first expand it in spherical harmonics

We put this into Eq. (32) and evaluate the resulting equation for small r, following
the lines of the derivation of Eq. (23) for the finite potential. The result, of course,
resembles that equation except that the left side is zero, and, in fact, we get

0 = 47rP1YL*(&) + k c BLSLL* , (33)


L'

where

SLL' = Mkr3 YL*CQJ YL@J dQ,. (34)


I

In this integration the equation of the surface, rs = r,(QJ, must, of course, be used.
An important special case of the above result is that in which k + 0, and the
Helmholtz equation becomes the Laplace. This special case applies, say, to a
problem with a perfect conductor, if we consider that the volume V is a perfectly
conducting one, and that $ is the electrostatic potential. There is one slight
generalization required. In the electrostatic problem, the potential tj need not be
zero on the boundary (grounded conductor) but may assume a constant value on
the boundary (ungrounded conductor). Aside from this, an integral equation like
(29) still holds, in which now eakar is replaced by $8 , the applied electrostatic
potential, and gk. is replaced by -l/45- I r - rs I [6]. Since (T, the surface charge
density, is u = -1/47r &//an, the electrostatic analog of Eq. (29) reads

#@I = A(r) + j 4) l/l r - rs I ds. (35)

This equation has the obvious physical interpretation that the total potential is the
sum of the potential & and the potential due to the charges on the surface.
Equation (35) holds everywhere, inside or outside the conducting body. But it is well
known that the potential must be constant inside a perfect conductor (the field is
zero there). We exploit this condition, first by expanding $e in the form valid for
small r,
SCHR6DINGER AND RELATED EQUATIONS 579

We put this into Eq. (35) and expand l/i r - rs / for r < rs in Legendre poly-
nomials of cos y where y is the angle between r and rs . As before, it is convenient
to write dS = (&l&J dQ,, to consider u a function of Sz,, and to define r(s2,) by

Then for small r Eq. (35) becomes, if & is the constant value that #(r) assumes in
the interior,

We use the addition theorem for P,(cos r) and equate coefficients of Y,(Q) to find
the equations for the unknown DL

h,&o = CL + 1 DL'TLL' 2 (37)


L'

where
=-
4Vr ‘,*@S) y,t(Q,) dQ
TLL,
21-t 1 s (38)
(r,(-Q,NZ s*
As for the scattering problem these equations are solved by reducing them to an
(approximate) finite set by truncating the spherical harmonic expansion. For the
homogeneous problem of the distribution of free charge on a conducting surface,
the coefficients CL are, of course, set to zero.

5. EXAMPLE-CHARGE DISTRIBUTION ON SPHEROID

In this section we calculate a practical example to illustrate the methods above.


For two reasons we have chosen an example in electrostatics to find the distribution
of charge on a perfectly conducting spheroid. First, this is one of the few problems
involving nonspherical (or noncylindrical) shapes that can be solved analytically.
It is a special case (two axes equal) of the well known formula for the charge
distribution on an ellipsoid. Second, the relatively simple Green function for
Laplace’s equation is involved in the solution, and this simplifies the integrals that
arise.
For reference, we first record the well known exact results [7]. A spheroid is
defined by the equation
580 EYGF!S

In terms of the ratio h


h = c/a
and spherical coordinates r, 8, this equation takes the form
r s rs = aA/(h2 + (1 - h2) cos2 O)lj2. (39)
If the spheroid carries a total charge Q, the charge density u is given by

We want to compare this result with that derived from Eq. (37). Since these
equations are formulated in terms of r(O), as a first step we calculate from ~(0)
the exact value of r(O). Equation (36) then shows that we must find I/r8 dS/&, .
Let dl be an element of length along the perimeter of the ellipse that is the cross
section (containing the z-axis) through the spheroid. Then dS = rs sin 0 dq dl =
rs sin 0 dq de dl/d8 = rs d.0, dlld8. From the equation of the ellipse, dl/d9 is
easily calculated, and we find
1 dS dl Aa(A4 + (1 - h4) cos2 O)1/2
-rs -dsZ, E -df9 = (~2 + (1 _ ~2) cos2e)3/2 -

With this we have from Eq. (36) the simple exact result for r(O)
r(e) = Qh2/4s7a(h2 + (1 - h2) ~0~2e).
We turn now to the approximate solution. The equations defining the solution
are (37), with Cr set equal to zero. Because of the azimuthal symmetry the Y&2,)
in Eq. (36) are replaced by Pr(cos a), and we can choose #,, at our convenience,
since it is essentially only a normalizing factor. Thus, in effect the expansion (36)
becomes

r(e) = T ~~~~~~~~ e).


b-0
Then Eq. (37) becomes, on using Eq. (39) for r = r, and with x = cos 0

(41)
where
tlf = -2 l Pz(x) Pzs(x)[A2 + (1 - h2) x2Jzi2 dx.
A’ I 0
Before presenting numerical results, it is interesting to show that for an almost
spherical spheroid (A M l), Eq. (41) gives the same lowest order result as the exact
SCHRijDINGER AND RELATED EQUATIONS 581

one. If we write h = 1 + E it is easy to see that for the first order in E the exact r(0)
of Eq. (40) can be written
r(e) w (Q/4nu)(l + z&P~(COS
Q.
But this is just the result that Eq. (41) yields. For from these equations we have,
for I = 2 and truncating the expansion at I’ = 2, that &/D, = -tzo/tzz . On
doing the integrals to lowest order in Eit is found that D,/D, = +, in agreement
with the above result. Beyond this we have solved Eq. (41) numerically for various
values of X by truncating them at 1 = I’ = 12. This leads to a six by six set of
equations to solve. The essential results are contained in Tables I and II.
TABLE I
Charge Distribution u(S) on a Spheroid Having Two Equal Semiaxes of Length a and
One of Length Xa
DZ

w 0.5 0.8 1.5 3.0

0 1.0000 1.oooo 1.0080 1.oooo


2 -0.86496 -0.29334 0.54482 1.41192
4 0.39228 0.04458 0.15034 0.97530
6 -0.15716 -0.00600 0.03654 0.56716
8 0.05844 0.00076 0200838 0.28608
10 -0.01920 -0.00010 0.00182 0.11476
12 0.00432 0.00032 0.02782

The latitude angle 0 is measured from the An axis. The result is expressed in terms of
r(0) = kru(0)(P + (1 - ha) cos2 O)lla/(Xa + (1 - X2) co? 0)s/z. The entries in the table are the
coefficients in the expansion r(0) = x:, D&(cos 0).

TABLE II
Comparison of r(e)

~0~ e\x 0.5 0.8 1.5 3.0

0.0 1.0000 1.0000 1.000 1.000


0.2 0.99654 1.0000 1.0000 1.000
0.4 0.99782 1.0000 1.0000 0.99997
0.6 0.99783 1.0000 1.0000 1.0001
0.8 0.99811 1.0000 1.0000 1.0045
1.0 0.99722 1.0000 1.0002 1.0986

Comparison of r(e) from the series whose coefficients are given in Table I, with the
exact value r,,(e) = l/(P + (1 - P) co@ 0) found from solution of the problem in ellipsoidal
coordinates. The entries in the table are IV)/F,,(e) and QP = 47~2.
582 EYGES

6. ENSEMBLE OF POTENTIALS WITH SPHERICAL BOUNDARIES

The results of Section 3 for single irregular potentials or boundaries can readily
be generalized to an ensemble of such potentials or boundaries. Similarly the
various special cases (potential becoming infinite, Helmholtz’s equation becoming
Laplace’s) can be discussed. The treatment of the special cases is quite straight-
forward, however, so we shall concentrate on the generalization to an ensemble
of potentials with irregular boundaries. An important special case of such an
ensemble is one with spherical boundaries. This is so fundamental in molecular and
solid state theory that we have chosen to discuss it first, as a step on the way to an
ensemble of irregular potentials. There is no real loss of conciseness in so doing,
since the results can be taken over directly to the more general case, and there is a
considerable gain in clarity.
Consider then the N spherically bounded potentials of Fig. 1. The centers of the

.-
I’ \\

FIG. 1. N potentials with spherical boundaries and with centers at d, , d, ,..., dN.

spheres are at dI , d, ,..., d, . A vector r defines a point in space and vectors


rl , r2 ,..., r, with origins at these centers define the same point. We apply the
same idea as for a single potential. Namely, we confine attention to the interior of
the potentials, and use the integral equation to exploit the information on the
interior solution that the differential equation yields. Of course, in the present
case there are N separate interiors, but this does not affect the working of the
method. One can, if one likes, imagine the N potentials joined by filaments of
SCHRiiDINGER AND RELATED EQUATIONS 583

potential that are small enough not to affect the problem but that provide in
principle a single simply connected interior. An important special case of the
N-potential problem is when N goes to infinity and identical potentials are arranged
in a periodic lattice-the problem of band structure. The formula we derive for
this case resembles that of the so-called KKR method*, but is simpler in some
ways.
Consider first the bound state problem, for which the wave function of the
ensemble satisfies the integral equation (5) with 4e = 0 and E = -/3”. We denote
by +s the wave function inside the sth potential; it can be considered a function
of rs . From the differential equation the form of lGsis

(42)

Thus, RI”’ is assumed known, in terms of the (unknown) energy eigenvalue /3.
We now want an equivalent expression for $8 from the integral equation (5), in
which we note that v(r) is the sum of atomic potentials ut

o(r)= 5 %(G)*
t4
We put this into Eq. (5). In doing the integral over the tth potential we express g,
in the local coordinates, writing it as gs(rt , r,l) to find

This equation for # holds everywhere, inside and outside the sth potential. Suppose
now that r refers to a point inside this potential so that 4 becomes #s . We break
the sum over t into the term with t = s and the terms with t # s, and use
rt = ra - dt, to find

~.drJ = 1 MS’) dry) g& , rs’) 4


+ C 1 tk’> G-t’) gdr, - 4, , rt’> &‘.

We put the expression (42) for #s into this equation, and a similar one for I,& _
We also expand the Green functions in the integrands according to Eq. (8). For

4 So named in the literature, after the authors in Refs. [8] and [9]. A recent survey of practical
aspects of the method is found in Ref. [lo].
584 EYGES

integration over dv,’ the expansion in terms of the function r, of this equation is
used. For integration over dvt’ the condition rt’ < rt holds since r1 is considered
to terminate in the sth potential. Hence, we can use the explicit form of Eq. (9)
for r, . Equation (43) becomes, on reverting to writing rt in place of rs - dl, ,

x (F r,*(r, , rl) YL-(is) Yz-(2,‘)) riz dr,’ dQ,’

x (z j,-(i/3r,? h;(ij?rJ YLQt) Y$(P,?) ri” drt’ dQ,‘. (44)

The integrations over dl2,’ and dQt’ yield

To now evaluate the limit rs -+ 0 we must further expand

to make the dependence on rs more explicit. Such an expansion for the present
case of interest, namely rs < d,, , has been derived by Williams, Hu, and
Jepsen [IO]. With some trivial modifications their formula is

hi
h(iS I rs - &, I> Y& - 4,)
= 47r c C iz-z’-z”ZLL~L~jz(i~rs) YL(#J hz-(ij3dtJ YL*(dt8), (46)
L L"

where

ZLL’L” = s r,* Y,* Y;* zi!.Q. (47)


SCHR6DINGER AND RELATED EQUATIONS 585

We put this expansion into Eq. (45) and equate the coefficients of Y,(P,) on the left
and right sides of it to find

A~‘R~“‘(r,) = ,!?A(LS)ja @(r,‘) us(r,‘) Tl(r,, ri) ri2 dr,?’


0

+ 47$3 C C c A~~Z~~,L.iz-z’-zg,(i~rs) hp(ipd,,) YL”(&J


t+a L’ L”

X O”R[f’(r,‘) z+(rt’) jl,(i/3rr;) ri2 drt ’ . (48)


s0

We now go to the limit rs -+ 0. On the left side of Eq. (48) we write


Iti” = r,lQ(ls)(rJ with QT’(O) = 1. In a way that is familiar by now, we break
the first integral on the right side over dr,’ into two parts: from zero to rs and from
rs to the spherical boundary. The term involving the first part goes as r:+2 and the
term involving the second contains a factor rsz from the expansion of the Bessel
function jl(i/3rs) that arises. There is a similar factor multiplying the sum over t.
Cancelling this factor throughout the equation we find that Eq. (48) becomes

Here we have defined the two integrals

(h 1 ifi)?’ = Ia l@‘(r) v,(r,J h,($r,,) rn2 dr, ,


0
(50)
(j ) i@’ = joa @‘(r,J v,(r,) j,(i/?r,J rn2 dr, .

Equation (49) is the final result. For the general case of N arbitrary potentials s
varies from one to N, so there are N independent coefficient sets A?: ... Al”,‘.
The energy eigenvalues of the ensemble of potentials are determined by setting to
zero the (infinite) determinant of these coefficients. In practice this determinant
must, of course, be truncated by keeping only a limited number of L values in
each of the coefficient sets. But also in practice, there will often be, as we have
remarked, symmetries in the arrangement of the potentials that introduce relations
among the sets that effectively reduces their number.
An important specific example of symmetry is, as noted above, when N is infinite
and the potentials are identical and arranged in a periodic lattice. The effect of the
symmetry is comprised in Floquet’s (or Bloch’s) theorem which states essentially
586 EYGES

the following: If the wave function at some point r0 in the lattice is #(I$ and if T
is a lattice translation vector, the wave function at the homologous point r,, + T is
I&, + T) = eik.= #t&J, where k is real. We apply this result to Eq. (49). Consider
some specific point rs in the interior of the sth potential, for which rs = F, 9, = D.
The wave function I$~ at this point is

ij* = T Aja)Rp)(F) Y&q.

Consider now the interior of the tth potential. We assume the coordinate system
defining rt is oriented parallel to that defining rS . The wave function & evaluated
at the homologous point rt defined by rl = 7, sZt = D is

But rt = rs + dSt , hence, by Floquet’s (Bloch’s) theorem

This equation is satisfied by taking


&) = eiWt&), (51)
We put this into Eq. (49) but first we simplify the notation. Since all potentials
are equivalent, it is convenient to take s = 0 and simply write At’ = AL and
dSt z dt . Then the condition (51) can be written Ag’ = eik.dtAL . Moreover, since
all potentials are identical we can drop the superscripts on (h I $3)~’ and (j I $):“I
so that Eq. (49) becomes

AL = !%i/? AAh 1ii%


(21+ l)!!

These equations resemble those obtained in the KKR method [lo]. In particular
the sum over t defines the same structure factor that occurs in that method. The
present equations appear to be more advantageous for the following reason. The
standard KKR method requires a numerical calculation of the radial function R1
that is sufficiently accurate to be differentiated numerically. This contrasts with
the present case where only the integral of the radial wave function is required,
and a coarse radial grid for the numerical calculation of RL that does not suffice
for differentiation may be accurate enough for integration.
With the above results at hand, the equations for the scattering of a wave from
SCHRijDINGER AND RELATED EQUATIONS 587

the N potentials of Fig. 1 are easily written down. For the sake of an example let
there be a plane wave incident described by eiL.‘l. In terms of another system,
say rs , it becomes the plane wave with a phase eik.(rs-dls). The integral equation is
now (5) with & = eik.rl. The additional term & does not change the analysis in
any essential way with respect to the homogeneous case. We expand the interior
solutions as before according to Eq. (42), expand the Green function (with ifl
replaced by k) in each of the integrals that arises from the right side of (5), and go
to the limit as rl , r2 ,..., rN go to zero. In doing this for rs , say, the only change
over Eq. (49) is to add the expansion of the plane wave eik.(rs-dlJ for small r, to its
right side. The equations we finally get are then essentially the set (49) but supple-
mented by a term arising from this expansion. They are, in short,

(2’ + 1)” A’S) _ iZy *(~) e-‘k’dl, - $ AZ”‘@ / k):“’


4nk1 L - L
- ik C C C Ajl:)ZLL,L-iz-z’-z”(j / k$’ hzs(kdt,) YL*(&~). (53)
tfu L’ L”

7. ENSEMBLE OF POTENTIALS WITH IRREGULAR BOUNDARIES

In this section we generalize the results of Section 4 to an ensemble of potentials,


each with an irregular boundary. As a special case of this generalization we derive
what appears to be a good approximation for the problem of an ensemble of
potentials with spherical but overlapping boundaries.
The ensemble of irregular potentials is sketched in Fig. 2. The sth potential acts
throughout the volume V, . Some reference point is chosen inside each potential

FIG. 2. Ensemble of potentials with irregular boundaries.


588 EYGES

and the positions of these points are given by vectors d, , d, ,..., dN . The reference
point serves also as an origin for a coordinate system re and the potential’ U, inside
the sth boundary is taken to be v, = v&r,). We shall begin with the bound state
problem. It can be assumed that the wave function inside the sth region is given
again by Eq. (42), and we can follow the derivation which begins with that equation
to arrive at Eq. (44). However, because of the irregularity of the boundaries, we
cannot now (as we did previously) use the orthogonality of the spherical harmonics.
Instead, we rearrange Eq. (44) slightly into the form

x Yf-(2,‘) r:’ dr,’ dQ,’ + /3 1 C C &h,+j?rt) YL@,)


L’ L” t#s

x l@(r,‘) vt(r;) jzm(i/3r,? Y,,(P;) Y$(F,‘) rd2 drt’ a!!&‘.


s
(54)

We insert the expansion of Eq. (46) into this equation and equate coefficients of
YL(fi8) to find

x I?,(!‘@,? vt(r;) j&3ri) YL,(F,? Yz@;) rb2 drt’ d-Q,‘.


I

Now we can go the limit rs -+ 0 in a familiar way. We define two integrals which
are generalizations of Eq. (29) by

(hY 1ip)$)s = 1” h@r,J YL*(3,J v,(r,) R$“(r,J YL@,) rn2 dr, dQ, ,
I

and the process of taking the limit yields

“‘(+g ! &’ = /g c /@yhYI ip)pJ + 47$? c c c 1 A~~ZLL~Ld-z@-z-


L’ t#s L’ L’ L”

x h,-($dtJ Y,-(ci,J(jY I ij?)$ . (55)


SCHRiiDINGER AND RELATED EQUATIONS 589

Equations (55) are general ones for an arbitrary arrangement of the potentials.
If there is some symmetry in the arrangement the same remarks apply as for an
ensemble of spherically bounded potentials. If, for example, the potentials are
identical and comprise a periodic lattice the condition (30) between AZ’ and AF’
applies, and the equations are simplified accordingly. For the scattering problem,
with a wave eik.rl incident, Eqs. (55) must be modified in the same way that the
bound state Eqs. (49) for an ensemble of spherical potentials were modified to yield
Eqs. (53) for the scattering problem.
As an application of Eqs. (55) we mention the problem of an ensemble of
spherically bounded potentials of finite range for which, however, the potentials
may overlap. We assume that no more than two potentials overlap at a given point
of space. This is, of course, not the most general case; given N,, potentials they
could completely overlie one another so that all N,, act at a given point. However.
the case when only two overlap is important practically, and the more
general case when there are triple, quadruple, etc. overlaps can be treated by an
obvious generalization of the method presented here.
An ensemble of No potentials that may overlap in pairs defines N,, i Nt irregular
regions; the N,, partially spherical regions where only one potential acts and Nt
regions where two potentials act. Then Eqs. (55) can be applied, with N replaced
by N, + Nt, subject to one approximation. Namely, to use these equations,
general solutions to the Schrodinger equation must be known inside both the
above-mentioned regions. These solutions are supposedly known for the non-
overlapping regions but in general they will not be known for the overlapping ones,
since the sum of the two potential tails that act will not necessarily be spherically
symmetric. An obvious approximation is to replace the sum of overlapping tails
by its average value. The solution in the overlapping region is then the well known
one involving spherical Bessel functions appropriate to this average potential
strength. As applied to the present case Eqs. (55) become equations relating
coefficients A:’ that refer to the nonoverlapping regions and coefficients AZ”’ that
refer to the overlapping regions. As before the equations must be truncated and
only a finite number of these coefficients retained. It is important to note that the
degree of this truncation will not necessarily be the same for AZ”’ and A;,‘“‘. For
example, if the overlap is small, only the lowest orders of the latter coefficients
will be substantial. Finally, we note that the approximation of replacing the
overlapping potential by a constant becomes exact for square well potentials.
Hence, Eqs. (55) enable one to calculate, subject to the usual partial wave
truncation, the scattering and bound state problems for an arbitrary ensemble of
doubly overlapping square wells.

595/S+15
590 EYGES

REFERENCES

1. L. EYGES, Ann. Whys. (N.Y.) 2 (1957), 101.


2. L. EYGES, Phys. Rev. 111 (1958), 683,
3. L. EYGES, Phys. Rev. 154 (1967), 1207.
4. P. MORSE AND H. FESHBACH, “Methods of Theoretical Physics,” p. 1574, McGraw Hill,
New York (1953).
5. N. F. MOTT AND H. S. W. MASSEY, “The Theory of Atomic Collisions,” Oxford University
Press, London/New York, 1964.
6. L. EYGES, “The Classical Electromagnetic Field,” Addison Wesley, Reading, MA, 1972.
7. J. STRATTON, “Electromagnetic Theory,” McGraw Hill, New York, 1941.
8. J. KORRINGA, PhyGca 13 (1946), 392.
9. W. KOHN AND N. ROSTOKER, Phys. Rev. 94 (1954), 111.
10. A. R. WILLIAMS, S. M. Hu, AND D. M. JEPSEN,in “Computational Methods in Band Theory”
(Marcus et al., Eds.), New York, Plenum, 1971.

You might also like