WA VE EQUATION 669

Only the odd powers of t& will have nonvanishing coe6cients. The coefficient of t& itself is zero, and if we consider values
of t for which t'4(1, then
2'
%2ZI=t& (cos q+ sin p)dq ( —,', cos' p —I";- cos' q sin y — ', cos
—, q sin' y ——,'6 sin' y)
0
2n
dp( y e cos p+ s cos y sin y —4 cos y sin y —s cos y sin3 y ——,'6 sin y)
0
= m-/8t",
(b) As I
t~i, becomes infinite, as may bc seen from inspection. The infinity is due to the behavior of the integrand near
p=0, and it is therefore obvious that I
becomes infinite in the same way as
2'
J= cos gdq (1 — ', z, 2, t).
t& cos p) &=const. t&F(-, (See "G".)
0

If J
t~1, behaves as —t& log (1 —t), which is positive.
I
(c) Numerical integration yields positive values of for values of t between 0 and 1. The following values of cI (c a
constant) show the run of the function:
t =0.09, cI = 0.15; t =0.25, cI =0.80; t =0.64, cI=5.5; and t =0.95 cI= 23.
It is therefore to be concluded that I is positive throughout the interval O~t~1, and that the integral Jp't&dtI(t) cannot
possibly vanish.

APRIL 15, 1937 PHYSICAL REVIEW VOLUM L' 5 1

On the Connection Formulas and the Solutions of the Wave Equation

ICUDOLE'H E. LANGER
Department of Mathematics, University of 8'isconsin, Madison, wisconsin

Part 1 gives a general discussion of asymptotic represen- Part 2 deals with the radial wave equation for motion in
tations of the solutions of the one-dimensional wave a central field of force. Both the attractive and repulsive
equation. The forms ordinarily used in the so-called Coulomb field are considered. It is shown that the applica-
W. K. B. method are multiple valued and consequently tion of the W. K. B. analysis to this equation as it has
necessitate a consideration of the Stokes' phenomenon, generally been made is uncritical and in error. The solution
in any region about a turning point, i.e. , a point in which commonly identified thereby as the wave function is in
the kinetic energy changes sign. Except under restrictive fact not the wave function. The "failure" of the W. K. B.
formulas, and the apparent necessity for modifying them
hypotheses they give no description of the solutions near
the turning points. The author's method for representing by replacing the number l(l+1) by (l+-', )', has been noted
by many investigators. This is traced to the misapplication
the solutions of such differential equations by means of
of the theory. When correctly applied the theory naturally
single valued functions is discussed, and the formulas
yields the formulas which have been found to be called for
applicable to the wave equation are given. These formulas on other grounds.
are usable over the whole of an interval which includes a Finally the case is discussed in which a turning point
turning point. The Stokes' phenomenon is not involved. lies too near the point r =0 for the W. K. B. method to be
It need be considered only if expressions of the older type effectively applicable. It is shown how the solutions are
are desired, and then the connection formulas of the describable in this case, the formulas given specializing,
W. K. B. method are immediately evolved. An appropriate when the field is an attractive field and the energy is zero,
formal development of the solutions of the wave equation to formulas which were given for that special case by
as power series in h is given. Kramers.

PART 1. THE ONE-DIMENSIONAL with

WAVE EQUATION
=
2' —l'(x) I
Q'(x) I&
HE wave equation for a mechanical system h'
of one degree of freedom is familiarly of the
form m, E and V(x) representing respectively the
d'u/dx'+ Q'(x) u = 0, (1) mass, the total energy and the potential energy,
 RU DOLPH E. LAN GER

and f(i standing for I(i/2z-, where fi is Planck's association much more is necessary than, for
constant. Kith the exception of a few special instance, the mere substitution of i~Q~ in the
functions U(x) the equation is not explicitly place of Q and the transcription. of (2) into the
solvable. In general, therefore, its solutions must form (3). For, in the first place, each of the forms
be studied through the medium of representative (2) or (3) becomes infinite at any point where
functions which in a specific sense approximate Q'(x) is zero, and hence neither form can be
them. A well-known procedure for obtaining such retained during the transit from one to another
representations' consists in effect of substituting interval between which Q'(x) changes sign. In the
into the diA'erential equation a series in powers of second place the differential equation of which
k. With proper adjustments the coef6cients of the functions (2) and (3) are solutions is
such a series may be successively determined, at d' W/dx'+ —(a(x) I W= 0
I Q'(x)
least theoretically, whenever x is restricted to an
with
interval upon which the function Q'(x) is bounded
from zero. The series so obtained, still with x ~(x) =3LQ'(x) 3'/4Q'(x) —Q" (x)/2Q(x).
restricted as noted, represent the solutions in the
For this equation any point where Q'(x) vanishes
sense that they are approximations to them
is obviously a singular point. The functions (2)
which take on an asymptotic character with
and (3) are, therefore, multiple valued in the
respect to k when this symbol is regarded as an
region about such a point (i.e. , a so-called turning
arbitrary small parameter. Retaining only the
point) whereas the solutions of the equation (1)
first terms of the series one finds in this way, if
are single valued. Since the approximation of a
Q'(x) is positive the real functions
single valued function by a multiple valued one
can maintain only in a restricted region, it is
W, (x; p, S) = SQ-*'(x) cos Jf
&1
Qdx+ 7, (2) clear, that the mere existence of a pair of
relations
and if Q'(x) is negative the real functions u(x) W+(x; y, ff),
u(x) W (x; n, p),
W (x; n, P) = ~Q(x) ~
l n exp I (Q(dx
&1 valid, respectively, in intervals on opposite sides
of a turning point, cannot be used as a basis for
+i(e v
f
—
&1
IQI&* (3) inferring that the right-hand members are one
and the same solution of the equation (4) simply
because the left-hand members are the same
The letters n, p, y, 8 stand for entirely arbitrary solution of the equation (1). The contrary is in
constants, while xI may, in each case, be any fact the case; the so-called Stokes' phenomenon.
constant for which the integral concerned is The possib'ility of inferring either of the rela-
convergent. tions (5) from the other depends, therefore, upon
Upon any interval on which Q'(x) is positive, the so-called "connection formulas, "
which as-
the form (2) with each choice of its constants sociated the values n, I8 with the y, 8, and which
represents some solution of the equation (1). If are, therefore, the quantitative analysis of the
there is also an interval upon which Q'(x) is Stokes' phenomenon. Such formulas have been
negative that same solution is represented deduced from various considerations. Thus they
thereon by a form (3), with specific constants. have been obtained for such cases as permit the
The property that they represent one and the assumption that in a suitable interval including
same solution thus correla, tes each form (2) with the turning point the function Q'(x) is repre-
a form (3) and vice versa, the correlation being sentable with a sufficient degree of accuracy by a
fixed by the association of their respective linear function. ' They have also been found for
constants; It is easily seen that to determine this
'H. A. Kramers, Zeits. f. Physik 39, 829 (1926); also
' J. Horn, Math. Annalen 52, 271 (1899); G. D. Birk- H. Jeffreys, Proc. London Math. Soc. (2) 23, 428 (1923);
hoff, Trans. Amer. Math. Soc. 9, 219 (1908); O. Blumen- also H. A. Kramers and G. P. Ittmann, Zeits. f. Physik
thal, Archiv. d. Math. u. Physik 19, 136 (1912); G. 58, 217 (1929); and for an extension of the method, S.
Kentzel, Zeits. f. Physik 38, 518 (1926). Goldstein, Proc. London Math. Soc. (2) 28, 81 (1929).
 WA VE EQUAT I ON 67i

such cases as permit the assumption of a suitable the turning point, but are adapted to do so in the
mode of passage around the turning point whole interval including the turning point. The
through the complex plane. ' This latter method term "representation" is used to signify, as
unfortunately leads to no description of the heretofore, the leading term of an expression
solutions in the immediate neighborhood of the which is asymptotic with respect to k.
turning point. The problem can, however, be As is evident the representations (6) are
dealt with, and with greater effectiveness, by a explicitly relative to the point xI. At any other
method which eliminates completely the whole turning point they are to be replaced by those
source of the difficulty, ' namely the approxima- relative to it, which are again obtainable from
tion of a single valued function by means of the formulas (6) with only the obvious modifi-
multiple valued ones. cations. It is a necessary restriction, then, that
If in an equation of the form (1) x& is a turning the interval may not extend across or even
point at which the function Q'(x) has a zero, completely up to another turning point. Except
which may be of any order, say v, the functions for this it may extend even to infinity, provided in
such case that the integral
U(x; n, P) =S(x) I~PJ „(()+PPJ„(&)I (6)
are solutions of the differential equation ~I'I~( )rQ( )Id (10)
—+
d'U
dx
I Q'(x) —0(x) } U= 0
converges, when extended over the interval, with
the permissible exception of a neighborhood of xI.
with 8(x) = S"(x)/S(x). Since the representation of a solution of Eq. (1)
by a form (6), if it holds at all thus holds to the
The symbols Jin (6) stand for the Bessel func-
left, to the right and at the turning point, no
tions usually so denoted, while
question of connection formulas arises in associ-
1 z ation with it. The Stokes' phenomenon is there-
Qdx, S(x) =Q '*(x)&'* &. (9) fore obviated, unless, to be sure, representations
v+2 ~
in terms of the more familiar functions (2) or
The differential equation (7) resembles the (3) are desired. In that case it is merely necessary
equation (1) on any range of values for which the to replace the Bessel functions in (6) by their
function 0(x) is small relatively to Q'(x). It can appropriate asymptotic expressions which are
be shown that this includes any range on which available from the literature of the Bessel func-
representation of the solutions of (1) by forms tions. If that is done the connection formulas
(2) or (3) is possible. Beyond that, however, the associating the forms (2) and (3) appear auto-
range need not exclude the turning point xI, for in matically and without resort to additional
the neighborhood of this point, as may easily be restrictions.
verified, the function S(x) is bounded from zero. In the usual case the wave equation (1) in-
Since the turning point is, therefore, an ordinary, volves a function Q'(x) which has a simple zero
nonsingular point for the equation (7), the at xI. With the positive direction of the x axis
functions (6) are single valued, and are thus chosen so that (x — xq) has the same sign as Q'(x),
evidently suited with a distinct advantage over the form (6) becomes more explicit, the general
such functions as (2) or (3), for instance, for the solution of the equation (1) being represented by
role of approximations to the single valued the formula
solutions of the equation (1). They are not
(a
restricted to yield representations of the solutions cos +ri J&&3(8—
of (1) in intervals on the one or the other side of 3Q
I
i I

'A. Zwaan, Dissertation (Utrecht, 1929), see also E. C. (vr

Kemble, Phys. Rev. 48, 549 (1935).
icos I
&3
ri l~ &(:&(6)—
(11a)
4 R. E.
Lankier, Trans. Am. Math. Soc. 33, 29 (1931),
also 34, 447 (1932), also Bull. Am. Math. Soc. 3'7, 397
(1935). with $= J;, Qdx, and g an arbitrary constant on
 RUDOLPH E. LANGER
the range —x/2 (g= vr/2. This form is real when (12a) evidently leads to the connection formula
(x —x&) is positive or zero. When (x — x~) is
negative it may be rewritten so as again to avoid &1

the use of imaginaries in the form IQ(x)I lexp — IQIdx

8{&I
s(x) 7r sin Y/
' Iy/3(
I ( I ) ~2Q l(x) cos Qdx ——, (14)
&1

+2 cos ——rl }X,p(I(I)

I
which is central in the so-called W. K.B. method
of dealing with the equation (1). Near the
turning point the descriptions of the solutions to
wit I&I = I;* IQIdx. These formulas describe which either (13) or (14) apply are obtainable
u(x) over a whole interval including xq. In from the formulas (11a) or (11b) with appro-
subintervals lying completely on one or the other priate values of g.
side of xI the Bessel functions admit of asymptotic It must be born in mind that the formulas
representations, the substitution of which results here in question are all in a sense approximate,
in the formulas and not exact. On that account it is essential that
inference of the one representation from the
(a) u(x) -2Q &(x) cos {$ —(x/4) + g }, other in the relations (13) or (14) be made only in
(12) the directions indicated by the arrows. A com-
(b) u(x) IQ(x) I-l{2sing e'&~ +cosy e-'&~}, parison of formula (12a) with experimental or
other data can suffice only to determine the value
which are appropriate, respectively, when (x — x&) of g approximately. This fact is of dominating
is positive or negative. It is clear that in (12b) the importance when the distinction between "q
term in the negative exponential is negligible equals zero" or "q nearly equals zero" is in
relatively to the other, except when q is numeri- question, for in the former case the first term in
cally very small. (12b) is absent, while in the latter it is present
The formulas (12) clearly involve the con- and then necessarily of dominant importance
nection formulas appropriate to (2) and (3), for if when x is sufficiently remote from xI. The arrow
the phase of the solution u(x) dictates for g in in (14), therefore, may not be reversed. On the
(12a) a value differing appreciably from zero, other hand, since an arbitrary multiplicative
this value substituted in (12b) leads to the constant factor may be inserted in or removed
connection foi mula from u(x), the value of g is identifiable from (12b)
&I only when the second term is not entirely over-
IQ(x)I l 2 sin
g exp IQIdx shadowed by the first, namely when q =0. Hence
one may not reverse the arrow in (13).'
The deduction of power series in 5 which are
p&I
+cos g exp — Jl IQIdx
asymptotic to u(x), and of which the forms (2) or
(3) are the leading terms is familiar, as has
already been mentioned. It is to be noted that
S
—2Q l(x) cos 7r such series with the forms (6) as leading terms
&
Qdx —— +q,
(13) may also be formally derived. Let the expression

U'(x)A(x, h')+ U(x)B(x, 5'),

The case in which there is no other turning point 'The case in which two turning points lie too close
to the left of xi requires special mention. For in together for application of the formulas (13) and (14)
requires independent consideration. Cf. S. Goldstein,
that case the wave function, i.e. , the solution Proc. London Math. Soc. (2) 33, 246 (1932); W. Voss,
u(x) which remains bounded for all x can be Zeits. f. Physik 83, 581 (1933); R. E. Langer, Trans. Am.
Math. Soc. 36, 90 (1934), and Bull. Am. Math. Soc. 40,
given by (12b) only with g=0. This value in 574 (1934).
 WA VE EQUATION 673

in which with any chosen constants as the lower limits of

integration. It is to be noted that the formulas
A(x, l4') = Pn;(x)h'&' (16) are convergent even when x is the turning
j=o point.

B(x, h, ') = QP;(x)f4", PART 2. THE RADIAL WAVE EQUATION

j=o
In the case of motion in a central field of force
the general wave equation in spherical polar
be substituted for 44(x) in the equation (1). It is
found then, if coordinates permits a separation of the variables. '
If the'radial component of the motion is desig-
nated by f(r)/r, the wave equation for P(r) is
—
1
0'(x) = e'(x) + e(x), thus found to be
h2

that d V/«'+ Qp'(r) 0 =0 (17)

with
2qP
U(x) {
28 — — 2qo !A'
l4' ) Qo'(r) =-2m—E+
h2
——
Ze'
l(l+1)
r r2
( ql2 — A+B"+8B
+{ 8 ——
h2
qp! With thecustomaryinterpretation of the symbols
the formula (18) implies the field to be an
attractive field, since the potential energy is
+ U'(x) {2B'+A"+8A } =0. given by the term — Ze'/r. The case of a repulsive
If the coeKcient of each power of k2 which field may be included in the discussion by the
multiplies either U(x) or U'(x) is set equal to simple expedient of admitting for the charge Z
zero, the choice no —0, Po —1, is seen to be negative values as well as positive ones. The
permissible, while the subsequent coe%cients variable r representing radial distance is, of
must satisfy the relations course, restricted to be positive.
Structurally, the equation (17) bears evident
2gPn + 2gigg'n; = P; i"+ 8P; 4
features of resemblance to the equation (1).
+(28 2gp)n; y+(8 gp')n; Insofar as the writer is aware, the degree of this
resemblance has hitherto been regarded without
2P = —n;" —8n;, J 1) 2 ) 3) ~
exception as sufficient for assuming the out-and-
Thus, one obtains formally the expansion out applicability to the equation (17) .of the
formulas deduced in the foregoing discussion,
especially of the formula (14). The first point at
44(x) = U(x)+ U'(x) Qn (x)h" which QpP(r) is zero is found directly to be
j=l
l(l+ 1)l'4'
+ U(x)pp (x)h" (15) (a) pi=
j=l 2El(l+1) 5'i '
mZe' 1+{ 1+ mZPe4
in which the coe%cients are given in succession
by the formulas if Z)0, 1/0, E=—0,
2(8 —Qp)n; i'+(8' —9p')n;
n;(x) =
1
— P'
— +P; l"+OP; l —
4

dx,
Ze' f 2El(l+1) l4'y &
(19)

—2E 1+{0 1+
il

2gg(x) ~ gl
(b) p4= !
mZ'e4
(16)
if Z&0, E)0.
P;(x) = —— ~l (n, "+8n;)dx, 'Cf. , for instance, T. Sexi, Zeits. f. Physik 99, 751
2~ (1936).
674 RUDOLPH E. LANGER
When Z(0, E(0, Qp'(r) does not vanish at all, representing the solutions of the latter by means
and when Z&0, E(0, it vanishes a second time, of those of the former, fails, of course, near the
namely at the point which is given precisely by points where Qp'(r) becomes zero. It fails also,
(19b). The case ) =0 for an attractive field though this seems to have been overlooked, near
stands out as exceptional, for the pole of Qp'(r) at r = 0, for at that point the function ~p(r) becomes
r=0 is then of a different sort than otherwise, infinite similarly to Qp (r). The relation (14),
while the zero otherwise given by (19a) is then therefore, though it can represent a solution of
entirely absent. With the point p& used in the the equation (17) in suitable intervals, cannot do
place of x&, and with r and Qp(r) replacing x and so in an interval which actually reaches up to the
Q(x) the relation (14) has been commonly re- point r =0. It follows at once from this that the
garded as applicable to the equation (17), and to vanishing of the left-hand member of (14) at
describe the wave function for that case. ' This r=0, gives no ground for concluding that the
procedure, however, soon leads to a difficulty. solution of (17) which is elsewhere represented
The resulting formulas were found by Kramers by (14) also vanishes at r=0. That is in fact not
to give an incorrect phase for the solution unless so. The solution elsewhere represented by (14)
the number I(I+1) is replaced by (I+ —,')', a becomes infinite at r=0, and so is not the
modification which is tantamount to raising the wave function.
potential barrier. Young and Uhlenbeck found Let the variables in the equation (17) be
the same modification requisite, both if the changed by the substitution
Balmer formula is to be obtained, and if the wave
function is to vanish to the proper degree at r = 0. r=e, P=e*"u. (22)
This "failure" of the W. K.B. method, i.e. , to the
extent that the change in question is requisite, The domain of the variable becomes then
—~ (x( ~, and the resulting equation is found
has been generally verified both in studies of
attractive and repulsive fields. No explanation of to be
it seems to have been given, though that can be d'u/dx'+ Q'(x) u = 0 (23)
done very simply to the following effect. The with
fault lies not in the method but in the application
of it. The commonly accepted assumption that 2m
Q'(x) = IZ e'~+Ze'e'I —(I+ —
', )'. (24)
the equation (17) is of the form (1) as it stands k2
is, namely, incorrect. The formula (14) is at best
only restrictedly applicable to the equation (17), Now this equation is obviously of the form (1).
and when it is applicable it describes a solution Its turning points computed from (24) but
which is not bounded, i.e. , which is not the wave expressed in terms of r, are found to be
function.
The functions which constitute the several (E+ -')'h'
members of the relations (14) and (13) are (a) r, =
2E(I+-')'h'q '*

solutions of the differential equation mZe' 1+] 1+.

E. mZ'e4
d'+/«'+ I Qp'(~) «(r) I + = 0 (2o) if Z)0, E=O,
3LQ'(r) O' Qo"(r)
(25)
with «(r) = (21) Z8
4Q '(r) 2Qo(r) (b) r~=
The resemblance between this equation and the
equation (17), and hence the possibility of if Z(0, E&0,
~
H. A. Kramers, Zeits. f. Physik 39, 836 (1926); L. A.
Young and G. E. Uhlenbeck, Phys. Rev. 36, 1158 (1930);
E. C. Kemble, Phys. Rev. 48, 560 (1935); F. L. Yost, J. A.
there being no turning point if Z(0, E and a (0,
Wheeler and G. Breit, Phys. Rev. 49, 180 (1935); T. Sexi,
second one given by (25b) if Z)0, The Z(0.
Zeits. f. Physik 99, 771 (1936). formulas (22) and (24) are found, moreover, to
 WA VE EQUATION 675

give the identities stances is questionable, and is generally less

satisfactory than otherwise. The formulas are of
e*&'Q —:(x)
=— -''-(r),
Q, course entirely inapplicable when there is no

x r
(26) turning point at all, namely when with a E(0
— repulsive field. In such cases the following
Q(x) dx = Q, (r) dr, analysis of the wave equation may be resorted to.
The change of variables
in which
r =s', /=slav, (28)
2m Ze2 (~+-o-)'
Q~'(r) = g~+ (27) gives to the equation (17) the form
k2 r r2
(d'v/ds') + I X'qP(s) + (-,' —&((')/s'I v = 0 (29)
The formulas (11), (13) and (14) are directly with X' = 8mZe'/5'
applicable to the equation (23). If they are so P=2l+1, (30)
applied, and the original variables are reintro- (s) = 1+I&-s /Zeo
duced, the results are found to be those which are
formally obtainable by the direct substitution of The solutions of an equation of this type are
r, P for u and x, and of the r& and QP(r) of (25) known, however, to be represented by the
and (27), for x& and Q'(x). The difference between functions
this procedure and the incorrect one of substi-
tuting the pq and Qo'(r) of (19) and (18) for x& and
Q'(x), is seen to amount formally to precisely the (7,
replacement of /(/+1) by (I+-,')'. It will be C( ods (31)
- ~(s) E 0 )
noted, moreover, that for the formula (25a) the
case I = 0 is not exceptional as it was for (19a). the symbol C8 signifying any Bessel function of
Since the applicability of the formulas as the order &P. The representations are valid from
described right up to the point r = 0 depends upon s=0 over an interval which does not rearh up to
their applicability to the equation (23) over the a zero of the function &po(s).
infinite interval to x= —~, the validity of the In the case of an attractive field (Z)0) the
procedure is still questionable. It is, however, value )2 is positive. The choice of Cp successively
easily established, for from the formulas (24), as J2~+~ and Y2~+~ yields respectively the repre-
(8) and (9) it is seen that as x —— & &x& sentations, which in terms of the original vari-
—'). ables are
Q(x) = O(1), (I(x) = O(x

Thus the integrand in (10) is of the order of x ',

and for such the integral to x= —~ converges.
The formulas relative to the right-hand turning
point (which may, of course, be the only turning
A(r)-
,
——.
0

Q (r)
~~&+il
4 o
Q«
)~,
I
(32)
r
point) are similarly seen to be applicable even to
f—
2

——
x=+ oo, for from the formulas (24), (8) and (9) Q. dr
it is seen that as x~+ ~ — (
A(r) Qo«
Q(x) = O(e-.), ~(x) =-O(1),
( Q (r) . I'o&+&~
E o ) ~

2m Ze2
whence the integrand in (10) is of the order of with Qo'(r) = 8+ (33)
e, and the integral to x=+
~ is convergent. PL2

Whenever the number / is small the turning Clearly »(&(r) is the solution which vanishes at
point (25a) lies very near to the point r=0, and r=0, while Po(r) is a specific one which is
this is true also in the case of the turning point there. Upon substituting for the
unbounded
(25b) whenever the energy is very large. The use
of the formulas (13) or (14) under such circum- ' R. E. Langer, Trans. Amer. Math. Soc. 37, 397 (1935).
676 RUDOLPH E. LAN GER

Bessel functions their familiar asymptotic repre-

sentations, the formulas become alternatively J I&I«» t'

)2y
41(r) — —
-**

Qo *(r) cos Q2« —(I+ l) ~,

A(r)-
. IQ, (r)l . z„„l E. o
IQ, I« I.
)
I
(~i I
~ p
The interval upon which they are usable extends
(34)
from r = 0 outward, but when E it must not (0
include or extend up to the turning point which
then exists. The substitution of asymptotic
These formulas are usable in intervals which forms gives the alternative representations
extend across and include the first turning point 1 r
(25a). When E(0
there is a second turning point +)(r) — t Qo(r) I I
' «p I Qol«
which limits the intervals of the present formulas, (2~) p

and relative to which the analysis given previ- (36)

ously may be applied. In the special case in which &.(r) —— — I Qo(r)
—
' exp—
Q I«
E= 0 the formulas (32) and (34) for the solution
I I
I I

P)(r) reduce to formulas which were obtained for

Finally, the substitutions
that case by Kramers. '
In the case of a repulsive field (Z(0) the value (2mB)l —Z ('2nze')
) ' is negative and hence the choices
h E. B )
vd' and
( reduce the equation (17) to
12(+)I Il'I Xo«+)I Il'I v«»
I
)I
o

for the Bessel functions in (31) will avcid the

E. o

' I(1+1)
—+1———
rPQ

introduction of imaginaries. The resulting formu- dp p p

las, in terms of the original variables, are then
Analyses of the equation in this form, both with
found to be
1
1 = 0 and l /
0, have been given. By comparison "
T
with them the present analysis and its consequent
I Qol«, formulas may be found to have at least the
A(r)- — —io+ /
(
IQol« I
advantages of simplicity and of generality of
- IQo(r)l- method.

'H.
T. Sexi, Zeits. f. Physik 56, 72 (1929); F. L. Yost, J. A.
A. Kramers, Zeits. f. Physik 39, 837 (1926). Wheeler and G. Breit, Phys, Rev. 49, 174 (1935).