You are on page 1of 9

A simpler solution of the Dirac equation in a Coulomb potential

Bernard Goodman and Siniša R. Ignjatović

Citation: American Journal of Physics 65, 214 (1997); doi: 10.1119/1.18531


View online: http://dx.doi.org/10.1119/1.18531
View Table of Contents: http://scitation.aip.org/content/aapt/journal/ajp/65/3?ver=pdfcov
Published by the American Association of Physics Teachers

Articles you may be interested in


Comment on “A simpler solution of the Dirac equation in a Coulomb potential,” by Bernard Goodman and Siniša
R. Ignjatović [Am. J. Phys. 65 (3), 214–221 (1997)]
Am. J. Phys. 66, 638 (1998); 10.1119/1.18923

Addendum to “A simpler solution of the Dirac equation in a Coulomb potential,” by Bernard Goodman and Siniša
R. Ignjatović [Am. J. Phys. 65 (3), 214–221 (1997)]
Am. J. Phys. 66, 637 (1998); 10.1119/1.18922

Comment on “A simpler solution of the Dirac equation in a Coulomb potential,” by Bernard Goodman and Siniša
R. Ignjatović [Am. J. Phys. 65 (3), 214–221 (1997)]
Am. J. Phys. 66, 636 (1998); 10.1119/1.18921

Comment on “A simpler solution of the Dirac equation in a Coulomb potential,” by Bernard Goodman and Siniša
R. Ignjatović [Am. J. Phys. 65 (3), 214–221 (1997)]
Am. J. Phys. 66, 635 (1998); 10.1119/1.18920

Comment on “A simpler solution of the Dirac equation in a Coulomb potential,” by Bernard Goodman and Siniša
R. Ignjatović [Am. J. Phys. 65 (3), 214–221 (1997)]
Am. J. Phys. 66, 634 (1998); 10.1119/1.18919

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
155.97.178.73 On: Wed, 26 Nov 2014 20:39:57
A simpler solution of the Dirac equation in a Coulomb potential
Bernard Goodman and Siniša R. Ignjatović
Department of Physics, University of Cincinnati, Cincinnati, Ohio 45221
~Received 19 February 1996; accepted 26 June 1996!
We present an alternative to the standard method of solution of the Dirac equation for a particle in
a Coulomb potential, which hews closer to the more familiar solution of the nonrelativistic
Schrödinger equation. The bound-state wave functions so obtained are also closer in form to the
nonrelativistic Schrödinger wave functions and make the transition to this limit more evident. The
equivalence of these and the ‘‘standard’’ solution form of the Dirac wave functions is demonstrated
in the Appendix, which, at the same time, shows the advantage of a new trigonometric
parametrization in the algebraic manipulation of relativistic quantities. © 1997 American Association of
Physics Teachers.

I. INTRODUCTION tion, we work out the normalization of the wavefunctions in


Appendix B. ~Normalization is generally not a simple pro-
The solution of the Dirac equation for hydrogenlike atoms cess.! Here, too, the approach is ‘‘pedestrian,’’ but also
by Darwin1 and Gordon2 two-thirds of a century ago was a hopefully instructive, relying on Ref. 11, whose very useful
celebrated conceptual advance; but it still has the aura of appendix on integrals of products of hypergeometric func-
being much harder and removed from the nonrelativistic tions seems to be insufficiently appreciated.
Schrödinger and relativistic Klein–Gordon solutions. Even
now there are relatively few quantum mechanics texts which
present the solution in detail.3,4 Most other authors content
themselves with discussing the energy level formula, its fine II. SEPARATION OF RADIAL AND ANGULAR
structure, etc., omitting the wave functions altogether5—or VARIABLES
showing the functions without derivation.6,7 The present pa-
per is intended to be a self-contained textbook-style solution In the standard representation of the Dirac equation,
of the Dirac–Coulomb equation which hews closely to how ~ a–p1 b m1V ~ r !! C5EC, ~1!
we solve the Schrödinger equation.
There is an irreducible extra complication, relative to the where the Dirac matrices are

S D S D
nonrelativistic theory, in the separation of the four-
I 0 0 s
component wave functions in a central field into radial and b5 , a5 , ~2!
angular parts because of the spinor eigenfunctions of the to- 0 2I s 0
tal angular momentum L1S, which replace the scalar spheri-
the radial-angular separated wavefunction C5~xw! has the
cal harmonic eigenfunctions of L. The textbook presenta-
form
tions on this point are quite efficient and need only a brief
review ~Sec. II!.
The remainder of the paper is devoted to our ‘‘user
friendly’’ solution of the two-component radial equation.
C jlm ~ r! 5 S g ~ r ! V jlm ~ u , f !
D
i f ~ r ! V j ¯l m ~ u , f !
, j5l6 21 , l1l̄52 j,
~3!
The standard procedure for solving the radial differential
equations has been via coupled power series expansions for where V jlm ( u , f ) are two-component angular momentum
both components, each series eventually decomposing into a ‘‘eigenspinors,’’ with total angular momentum J5j and
pair of power series for confluent hypergeometric functions Jz 5m ~units \51!, constructed by vector addition of s51/2
F(a,b;z). A feature of this solution is that, in the nonrela- and L5l angular momentum states. For given j, there are
tivistic limit, the parameter b becomes 2( j11), where j is two distinct l values, l5 j71/2 ([l 6 ), both appearing in
the total angular momentum, j5l61/2, so that the limiting C jlm , where l̄ 6 [l 7 .
wavefunction remains the sum of two F functions; whereas The notation C j k m ~and V j k m ! is often used instead of
the Schrödinger equation solution has a single hypergeomet- C jlm ~and V jlm ! to designate j5l11/2 ( k 5 k 1 ) and j5l
ric function with b NR→2(l11). Our solution reduces di- 21/2 ( k 5 k 2 ) states. There is a sign convention in the defi-
rectly to the nonrelativistic ~b NR! form and may have advan- nition of the quantum number k, namely,
tages in the calculation of matrix elements. To show the k 6 [7 ~ j11/2! 57 u k u , ~4!
equivalence of the two forms of solution is a bit complicated
and is relegated to Appendix A. The trigonometric param- which seems inverted and deserves comment—but first a re-
etrization introduced there of the various relativistic quanti- mark about the parity of C jl( k )m . The parity of the eigens-
ties considerably facilitates the demonstration. pinor V jlm is determined by l, P5(21) l . Since l̄5l61, the
The Schrödinger-like form of the Dirac wave function is upper and lower spinors in Eq. ~3! have opposite 3-space
not new, but previous derivations, except for Ref. 8, employ parity. The full parity operation, namely, r→2r, a→2a
nonstandard representations, like the chiral representation, of changes C~r! to bC~2r! and restores the parity of l to C jlm .
the Dirac matrices.9,10 In this respect our solution is ‘‘pedes- ~The quantum number k depends on both j and l, and does
trian,’’ not appealing to deux ex machina types of ansatz. not by itself tell the parity of C jlm , except that k and 2k
Finally, in order to make this a self-contained presenta- states have opposite parity.! From the relations

214 Am. J. Phys. 65 ~3!, March 1997 © 1997 American Association of Physics Teachers 214

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
155.97.178.73 On: Wed, 26 Nov 2014 20:39:57
L–s5 H l
2l21
for j5l11/2
for j5l21/2
M2 5 F k2
r 2 1m 2 E1
2
Za
r S DG
2
1,

it follows that k6 are the eigenvalues of the operator


k̂ [11L–s. ~5!
dM 1 k
dr
5 2
r Za S 2Z a
2k
D [
1
r2
B.

The sign convention ~4! carries over to k̂ Equation ~11! decouples into separate second order differen-
tial equations in the basis where the 232 matrix B is diago-
k̂ V j k 6 m 56 ~ j11/2! V j k 6 m [2 k 6 V j k 6 m ~6! nal. The eigenvalues of the matrix B are g1,256g,
in anticipation of the fact that V k̄ ~omitting j, m indices!, not g [ Ak 2 2Z 2 a 2 ~12!
Vk , eventually remains in the Dirac equation for the large-
component spinor w 5gV k and the matrix whose columns are its ~normalized! eigenvec-
tors, namely,

S D
~ s–p! gV k 5 ~ E1m2V ! i f V k̄ ~7!
1 Za k2g
because s–p changes parity. This change is effected through A5 ~13!
the important relation, derived in detail in Refs. 3 and 4, A2 k ~ k 2 g ! k 2 g Za
namely, diagonalizes B

S
~ s–p! g ~ r ! V k 5 p̂ r 2i

r D
g ~ r ! V k̄ , ~8! B8 5A21 BA5 S g1
0
0
g2
DS
5
g
0
0
2g
D . ~14!
where p̂ r is the radial-momentum operator, p̂ r g52ir 21
The functions u 1 , u 2 in the diagonal basis, namely,
3 ] r (rg). This leads to the positive-sign counterpart of Eq.
~6! for k̄ ,
k̂ V k̄ 6 5 k 6 V k̄ 6 ~68!
S D u1
u2
5A21
g
f SD ~15!

satisfy differential equations similar to the nonrelativistic


and V k̄ cancels from both sides of Eq. ~7!. Similarly for the Schrödinger equation for Coulombic wave functions

F G
small-component spinor equation,

S D
d 2 u 1,2 2 du 1,2 1 EZ a g ~ g 61 !
k̂ 2 1 1 2 1 1 u 1,250,
~ s–p! i f V k̄ 5 p̂ r 1i i f V k 5 ~ E2m2V ! g ~ r ! V k , dr r dr 4 r r2
r ~16!
~88!
where
yielding the real radial Dirac equations in the standard rep-
resentation ~2!: r 52lr, l[ Am 2 2E 2 . ~17!
d ~ rg !
dr S
1 k g5 E1m1
Za
r
rf, D ~We restrict ourselves to the bound states u E u <m, but con-
tinuum states may be discussed in the same way.!
The usual substitutions

S D
~9!
d~ r f ! Za u 1 5 r g e 2 r /2y 1 ~ r ! , u 2 5 r g 21 e 2 r /2y 2 ~ r ! ~18!
2 k f 52 E2m1 rg.
dr r
and definition
III. SOLUTION OF THE RADIAL EQUATIONS EZ a
n 8[ 2g ~19!
l
The ‘‘standard’’ solution of the coupled linear differential
equations ~9! starts with expanding g(r) and f (r) in power transform Eqs. ~16! into the form
series. Both series are then recognized to be separable into d2y 1 dy 1
two series each corresponding to a confluent hypergeometric r 2 1 @ 2 ~ g 11 ! 2 r # 2 ~ 12n 8 ! y 1 50, ~20a!
function.3,4 We start instead by differentiating Eq. ~9! as a dr dr
matrix equation, namely, d2y 2 dy 2
d rg
S D S D
5M
rg
,
r
dr2
1~ 2g2r !
dr
1n 8 y 2 50. ~20b!

SS D
dr r f rf
These are so-called Kummer-type equations,12
~10!
k Za d 2w~ r ! dw ~ r !
2 E1m1 r 1 ~ b2 r ! 2aw ~ r ! 50, ~21!
r r dr2 dr

D
M5 .
Za k whose solutions, regular at r50, are confluent hypergeomet-
2 E2m1
r r ric functions:
This gives a r a ~ a11 ! r 2
F ~ a,b; r ! 511 1 1••• . ~22!

dr S DS
d 2 rg
2 rf 5 M 1
2
dM
dr DS Drg
rf
, ~11!
Thus
b 1! b ~ b11 ! 2!

where y 1 5c 1 F 1 ~ r ! 5c 1 F ~ 12n 8 ,2g 12; r ! , ~23a!

215 Am. J. Phys., Vol. 65, No. 3, March 1997 B. Goodman and S. R. Ignjatović 215

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
155.97.178.73 On: Wed, 26 Nov 2014 20:39:57
y 2 5c 2 F 2 ~ r ! 5c 2 F ~ 2n 8 ,2g ; r ! ,
where c 1 and c 2 are normalization constants. The wave func-
~23b!
S D
d c 1F 1
d r c 2F 2
tions defined by Eqs. ~19! and ~23! are not normalizable
unless the series ~22! for F 1 and F 2 terminate as finite poly-
nomials. This requires, as usual, that n 8 (E)[2a 1 and
5 FS 2
2 g 11 1
r
1
2 DS D1
0
0
0
~26!

n 8 (E)21[a 2 be non-negative integers 0,1,2,... . Together


with Eq. ~19!, this gives the bound state energy-level for-
mula:
1
1
2l g S
r~ mg2kE !
ZaE r 21 ~ k E1m g !
ZaE
D GS D
c 1F 1
c 2F 2
.

@To get Eq. ~26!, use the relation

S D S D
m
~24! r 0 21 r 21 0
A S D
E5 .
Za 2 5 , ~27!
11 0 1 0 1
n 81 g
plus the transformation,
The case n 8 50 ~i.e., a 2 521! requires in addition that F 1 ,
which does not terminate and so is not normalizable, be ab-
sent; thus c 1 must be zero. This case will be discussed sepa- A21 MA52
2l g 1
r 0 21
0
S D
S D
rately.
Apart from a common phase, the coefficients c 1 and c 2 are 1 ZaE k E1m g
determined by two conditions. First, their ratio c 1 /c 2 is de- 1 ; ~28!
g
mg2kE ZaE
termined by referring back to the original coupled first order
differential equations ~9!, in which they are connected. @This the latter is facilitated by writing the 232 matrices in terms
connection is broken by the differentiation since the second- of Pauli matrices. For example, A5a11b s x .# The left-hand
order differential equation ~11! has more solutions than has term of Eq. ~26! begins with the power r0, while the right
Eq. ~9!. The diagonalized form of Eq. ~11!, in which c 1 and term has the power r21, whose coefficient must, therefore,

S DS D
c 2 are independent, shows this fact most clearly.# The second vanish; i.e.,
condition is the overall normalization. The coupled equations 1
y c F 2 ~ 2 g 11 ! ~mg1kE! c1
for ( y 1 ) 5 ( c 1 F 1 ) are obtained by inserting 2l g 50 ~29!
2 2 2
c2

SD
g
f
5 r g 21 e 2 r /2A S D
r c 1F 1
c 2F 2 or
0 0

S DS D
c 2 2l g ~ 2 g 11 !
r 0 c 1F 1 5 . ~30!
5 r g 21 e 2 r /2A ~25! c1 k E1m g
0 1 c 2F 2
The normalization of C jlm is carried out in Appendix B.
into Eq. ~9! Combining ~13!, ~30!, and ~B13! gives

S D
g~ r !
f ~r!
5
~ 2l ! 3/2
g G ~ 2 g 12 !
A G ~ n 8 12 g ! E ~ k E2m g !
8m 2 ~ n 8 1 g !~ k 2 g ! n 8 !
~ 2lr ! g 21 e 2lr

3 S Z a ~ k E1m g !
~ k 2 g !~ k E1m g !
~ k 2 g ! l ~ 2 g 11 !
Z a l ~ 2 g 11 !
DS 2lrF ~ 12n 8 ,2g 12;2lr !
2 g F ~ 2n 8 ,2g ;2lr !
. D ~31!

It should be noted that Eq. ~31! represents a twofold degen- The wave function ~25! becomes, in this case, since
erate energy level, with the two different states correspond- F(0,2g ; r )[1,
ing to the sign of k in Eq. ~4!, namely, to l5 j21/2 and
l 8 5 j11/2, despite the fact that the functions F 1 and F 2
contain k through g only and, like E, are independent of the S Dg~ r !
f ~r!
5c 2 r g 21 e 2 r /2A
0
1 SD
S D
sign of k. One needs only to note that Eq. ~9!, the decoupling
matrix A and, consequently, the ratio c 1 /c 2 are different for r g 21 e 2 r /2 2 ~ u k u 1 g !
5c 2 ;
k1 and k2 . We will see that the nonrelativistic wave func- 2uku~ uku1 g ! Za
tions for the k1 and k2 cases are F 2 and F 1 , respectively.
For n 8 50, as mentioned below Eq. ~24!, the normalizable and the normalized form is easily verified to be

S D
state requires that c 1 50, i.e., that the denominator in Eq.
g~ r !
~30! vanishes. There is only one solution, namely,
k5k152uku, for which f ~r!

E5
mg
uku
, ~32! 5
~ 2l ! 3/2
A2 u k u ~ u k u 1 g ! G ~ 2 g 11 ! Za S
2 ~ u k u 1 g ! g 21 2 r /2
r e . D
in agreement with the result of putting n 8 50 in Eq. ~23!. ~33!

216 Am. J. Phys., Vol. 65, No. 3, March 1997 B. Goodman and S. R. Ignjatović 216

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
155.97.178.73 On: Wed, 26 Nov 2014 20:39:57
@The minus sign ensures agreement with the sign of Eq. ~31! mZ 2 a 2
when k2g52~uku1g!.# Equation ~33! may be put in simpler e n 52 .
2n 2
form with the help of the relations ~12! and ~30! in terms of
the unit spinor For g 2 , the corresponding nonrelativistic notation is l̄[ j

S D
11/25 u k u 5l11 and n̄ r [n 8 215n r 21, and gives
U0 [
cos c
2sin c
, tan c [ A m2E
m1E
5 A kk u u2g
u u1g
, gk
NR
2 ,n 8
} r u k u e 2 r /2F ~ 12n 8 ,2u k u 12; r !
namely, ¯
} r l e 2 r /2F ~ 2n̄ r ,2l̄12; r !

S D g~ r !
f ~r!
5
2 ~ 2l ! 3/2 g 21 2 r /2
AG ~ 2 g 11 !
r e U0 . ~34! 5u n,l11 ~ r ! .
The normalization follows directly from that in Eq. ~31!. The
~368!

Equation ~34! may also be obtained from Eq. ~31! by taking normalized form of u n,l is

S D
its n 8 50 limit.

IV. THE NONRELATIVISTIC LIMIT


u n,l ~ r ! 5
2 ~ mZ a ! 3/2
n 2 ~ 2l11 !
A ~ n1l ! ! 2mZ a
n r! n
r
l

The measure of relativistic effects is v /c, where v is a


typical velocity. For hydrogen-like atoms the Bohr formula
S
3e 2 ~ mZ a /n ! r F 2n r ,2l12;
2mZ a
n
r , D ~37!

for the lowest state, n51 is in agreement with, say, Eq. ~36.13! of Ref. 13. The
Z 2a 2 Laguerre-function version of ~37! is not presented here ~but
e Bohr5E NR2m52 5KE1PE52KE, see Ref. 13! because there are different conventions for these
2 functions, e.g., in Ref. 13 vs Ref. 12. Since the scalar states
using the virial theorem, KE52PE/2. This yields the well- u n,l , u n,l11 have the same energy, all l states for given n are
known fact that the measure v /c is equivalent to the more degenerate, which leads to the familiar 2n 2 degeneracy ~in-
convenient parameter, z [Z a . The z-sensitive parts of the cluding spin! as compared with the at most 2(2 j11)-fold
solution ~31! are the terms in the coefficient matrix, namely, degeneracy in the solutions of the Dirac equations. @j appears
first in the energy formula above in the term O ~z2!.#
1 In conclusion, to lowest order in z the four component
A~ k 2 g !~ k E1m g ! wave function Eq. ~3! has

3 S Z a ~ k E1m g ! / ~ 2 g 11 !
~ k 2 g !~ k E1m g ! / ~ 2 g 11 !
~ k2g !l
Zal
D , ~35!
w jlm 5u n,l ~ 11O ~ z 2 !! V jlm ~ u , f ! ,
x j ¯l m 5c z u n, ¯l V j ¯l m ~ u , f ! .
within which To extract the physical positive ~and negative! energy states,
it is necessary to make a final transformation of the Dirac-
l5O ~ m z ! , k 2 2 g 5O ~ z 2 ! , representation state according to the Foldy–Wouthuysen
k 1 E1m g 52 u k u E1m g 5O ~ m z 2 ! . scheme:14
Thus for the cases k1 and k2 , respectively, ~35! is C D→CFW5e iS C D .

S O ~z2! O~1!
D S O~1! O ~z2!
D The lowest-order transformation S 1 is7

S D
and . ~358!
O~z! O~z! O~z! O~z! i ba–P 1 0 s–p
S 1 52 5
To lowest orders in z, and for both degenerate states, ~358! 2m 2m 2 s–p 0
gives g 6 (r)5g 6 (r) NR1O ~z2! and f 6 (r)5O ~z!. Consider
which gives
g 1 first, where l5 j21/25 u k u 21. Inserting the estimate
~35! into Eq. ~31! gives g→g FW5 ~ 12 d ! g,
g k ,n 8 } r u k u 21 e 2 r /2F ~ 2n 8 ,2u k u ; r !
NR
1
f → f FW52 f d ,

} r l e 2 r /2F ~ 2n r ,2l12; r ! where

5u n,l ~ r ! . ~36! E2m2V


d5 5O ~ z 2 ! .
2m
The last two lines of Eq. ~36! are written in the nonrelativ-
istic notation; and u n,l is the usual radial wave function for The physical state to O ~z2! is then

S D
the scalar Schrödinger equation. The radial quantum number
is n r [n 8 ; and the arguments, r 52mr/n and 2 u k u 52l11, u n,l V jlm ~ 11O ~ z 2 !!
C FW5 .
differ by O ~z2! from r 52lr and 2g, respectively. The prin- 0
cipal quantum number is n[n 8 1 u k u 5n r 1l11, and the The O ~z2! corrections to u n,l contain the various effects of
Bohr formula is obtained to the requisite order, namely, the mass change, the Darwin term and the spin–orbit inter-

E2m5 e n 11O F S z2
~ j11/2! n
DG ,
action ~H SO!. Conventionally, these terms are put in ‘‘by
hand’’ into the Schrödinger theory. The Foldy–Wouthuysen
change-of-basis is a systematic refinement of the original
where derivation of the Thomas precession which employed a suc-

217 Am. J. Phys., Vol. 65, No. 3, March 1997 B. Goodman and S. R. Ignjatović 217

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
155.97.178.73 On: Wed, 26 Nov 2014 20:39:57
cession of Lorentz transformations ~‘‘boosts’’! to comoving tally, the same as for the matrix B in Eq. ~11!; the resulting
reference frames of the orbiting electron.15 diagonal-element differential operators in Eq. ~39! are just
those of Eq. ~16!. However, IZ do not find the actual eigen-
V. SUMMARY OF RELATED LITERATURE functions but revert instead to stating the ‘‘standard’’ solu-
tion ~A1!.
The impetus for the presentation here was the treatment of We later found that Auvil and Brown9 had already derived
the Dirac–Coulomb problem in Itzykson and Zuber’s ~IZ! the eigenfunctions, starting from a similar, but different
book ~Ref. 6!, Secs. 2–3, which produces the energy-level separation into two-dimensional subspaces, namely, the ones
formula ~24! in a few lines. The IZ, as well as other defined by the projection operators

S D
treatments,9,10 start with the second-order equation,
I 6I
@~ g m p m ! 2 2m 2 # C50, ~38! P 1,25 21 ~ 16 g 5 ! 5 21 .
6I I
and separate the four-component C-spinor space into
2-component spinor subspaces. Equation ~38! is the familiar These decompose C into

S DS D
result of multiplying the Dirac equation,
c1 c2
~ g m p m 2m ! C50, C5 ~ P 1 1 P 2 ! C5 1 ,
c1 2c2
by g m p m 1m. The g matrices in ~38! are gm5~b,ba! and where c6 are the same as in Eq. ~40!.
p m 5i ] / ] x m 1eA m . With no magnetic field and V52Z a /r, Similar remarks apply to the treatment of Wong and
the energy-eigenvalue equation from Eq. ~38! is Yeh,10 who chose as subspaces eigenspaces of the operator

F p̂ 2r 1
L2
r 2 2 E1 S
Za
r D 2
1m 2 1i a–“ S DGZa
r
C50. ~39!
L, the relativistic generalization of the Runge–Lenz
operator.13 The operator had been introduced much earlier;
Ref. 10 is a continuation of the previous works that used L
The IZ separation results from the use of the chiral represen- in this context.
tation of the Dirac matrices ~see their Appendix A-2!, in A solution of the radial differential equations ~9! directly
which is also contained in the recently translated book by Nikiforov

S D
and Uvarov ~Ref. 8!. They start from the mathematical ques-
s 0 tion of whether it is possible to obtain, from the coupled
achiral5 . linear equations ~9!, equations of the Schrödinger or Klein–
0 2s
Gordon type for the Coulomb problem, namely, of the form
The unitary matrix taking the standard ~Dirac! representation ~16! with at most a second order polynomial in r21 in the
to the chiral representation is square brackets. Differentiation of Eq. ~9! and elimination of

U5
1
A2
~ 12 g 5 g 0 ! 5
A2
1
S D
2I
I I
I
,
one of the functions, for example, f (r), will not achieve this;
so they resort to the ansatz of making a linear transformation,
like Eq. ~15!, of the equation before differentiating it. A de-
tailed argument then produces, in fact, two distinct choices
so that
for the matrix A. One leads to the ‘‘standard’’ sum-of-F’s

C chiral5
c1
c2
5S D
1 w1x
A2 w 2 x
S D ~40!
solution; the other is of the form of Eq. ~13! and gives our
solution. This is because the differentiation of Eq. ~9! and the
transformation A commute.
on the upper and lower spinors of which a56s. The
spinors c6 still depend on r, u, f and are linear combina-
APPENDIX A
tions of V j k 6 m . Thus neither L2 nor s is diagonal. The
diagonalization in each two-dimensional space is, coinciden- The standard solution, taken directly from Ref. 3, is

S D g~ r !
f ~r!
52
~ 2l ! 3/2
G ~ 2 g 11 ! A 4m
G ~ n 8 12 g 11 !

F
~ 8 g !m ~ n 81 g !m
n 1
2 k n 8! G
S D
E E

2 Am1E
EF
~ n 81 g !m
G
2 k Am1En 8
S F ~ 2n 8 ,2g 11;2lr !
D
F G
3 ~ 2lr ! g 21 e 2lr . ~A1!
~ n 81 g !m F ~ 12n 8 ,2g 11;2lr !
Am2E 2 k Am2En 8
E

As remarked in Sec. I, this solution has the feature that both Eq. ~31! and ~A1! is to relate the F functions appearing in
f and g are linear combinations of F functions with b52g11, Eq. ~31!, where b152g12 and b252g, to those in Eq. ~A1!.
which becomes b52j12, instead of b52l12, in the nonrel- The appropriate ‘‘contiguous’’ hypergeometric function rela-
ativistic limit. The first step in showing the equivalence of tions ~13.4.3,4! in Ref. 12 may be written in the form

218 Am. J. Phys., Vol. 65, No. 3, March 1997 B. Goodman and S. R. Ignjatović 218

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
155.97.178.73 On: Wed, 26 Nov 2014 20:39:57
S 2lrF ~ 12n 8 ,2g 12;2lr !
2 g F ~ 2n 8 ,2g ;2lr ! D and
k E1m g 5 k m ~ cos 2c 6cos 2f ! ,
5C S
F ~ 2n 8 ,2g 11;2lr !
F ~ 12n 8 ,2g 11;2lr !
, D ~A2! ~ n 81 g !m
2k5
Zam
2k5uku
sin 2f
71 , S D
~A6!

E l sin 2c
where

C[ S 2 ~ 2 g 11 !
n 8 12 g
2 g 11
2n 8
D .
where the upper ~lower! sign is for k1 ~k2!, respectively.
First put the matrix in Eq. ~A1! into this representation,
where it is proportional to

S D
It remains to compare the matrix prefactor in ~A1! to the
2cos c ~ sin 2f 6sin 2c ! cos c sin 2~ f 2 c !
product of the matrix prefactor in ~31! with the ‘‘contiguity’’ ;
matrix in ~A2!. Since both forms ~31! and ~A1! are normal- sin c ~ sin 2f 6sin 2c ! sin c sin 2~ f 2 c !
ized, only ratios of respective matrix elements need be the ~A7!
same. Comparing Eq. ~31! with the middle expression in Eq.
~25! shows that the matrix in Eq. ~31! is the product of A and by use of the trigonometric identities,
with a matrix formed by the coefficients c 1,2, namely,

S DS D
sin 2f 6sin 2c 52 sin~ f 6 c ! cos~ f 7 c ! ,
c1 0 k E1m g
0 may be rewritten in the form
D[ c 2 } 2 g 11 , ~A3!
0
2g 0 l 2 cos~ f 2 c !
so that the matrix product A3D3C should then be propor-
tional to the matrix in Eq. ~A1!. A direct confirmation of this
is quite complicated, involving repeated use of relations be-
3 S 2cos c sin~ f 1 c !
sin c sin~ f 1 c !
cos c sin~ f 2 c !
sin c sin~ f 2 c !
D , for k 1 ,

tween the various quantities. The manipulations are greatly ~A8!


2 sin~ f 2 c !

S D
facilitated by the following trigonometric representation of
the four basic parameters m, E, Z a , k: 2cos c cos~ f 1 c ! cos c cos~ f 2 c !
3 , for k 2 .
E[m cos 2c , Z a [ u k u sin 2f ~A4! sin c sin~ f 1 c ! sin c cos~ f 2 c !
~c is closely related to the rapidity v, cosh v[E/m, defined Turning now to the ~three-matrix product! form of Eq.
for E.m!, where c and f are restricted to the interval ~31!, the matrix A has the trigonometric form
@0,p/4#; from which follow the identities
l5m sin 2c ,
Am2E5 A2m
g 5 u k u cos 2f ,
sin c , Am1E5 A2m cos c , ~A5!
A1 5 S sin f
2cos f
2cos f
sin f
D , A2 5 S cos f
sin f
sin f
cos f
. D
~A9!
sin 2~ f 2 c ! sin 2~ f 1 c ! After some trigonometric identities, the remaining product
n 85uku , n 81 g 5uku ,
sin 2c sin 2c becomes

S sin~ f 1 c ! sin~ f 2 c ! 2sin~ f 1 c ! sin~ f 2 c !


D
S D 5
2m u k u for k 1
2 ~ k 6 E1m g ! k 6 E1m g sin~ f 1 c ! cos~ f 1 c ! 2sin~ f 2 c ! cos~ f 2 c !

S D
D3C5 5 .
l ~ n 8 12 g ! 2ln 8 2cos~ f 1 c ! cos~ f 2 c ! 2cos~ f 1 c ! cos~ f 2 c !
2m u k u , for k 2
sin~ f 1 c ! cos~ f 1 c ! 2sin~ f 2 c ! cos~ f 2 c !
~A10!

The product of the matrices ~A9! and ~A10! is


2m u k u cos 2f 3 @ k 6 -matrices in~A8 ! ].
A† A511
Za
k
s x 511
Za
k S D 0
1
1
0
,

This completes the proof of equivalence of solutions ~31! in ~B1! gives


and ~A1!.

APPENDIX B
~ 2l ! 3 5 Er S
0
`
2g 2r
e c 21 r 2 F 21 1c 22 F 22

The normalization condition is

ES DS D
` g †
g 2
r d r 5 ~ 2l ! 3 . ~B1!
1
2Z a
k D
c 1 c 2 r F 1 F 2 d r 5c 21 I 111c 22 I 22

0 f f 2Z a
1 c 1 c 2 I 12 , ~B2!
Substituting Eq. ~25! and the matrix product relation, k

219 Am. J. Phys., Vol. 65, No. 3, March 1997 B. Goodman and S. R. Ignjatović 219

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
155.97.178.73 On: Wed, 26 Nov 2014 20:39:57
with the integrals labeled as
J msp ~ a , a 8 ! 5 Er
`
m 211s 2 ~ k1k 8 ! r /2
e F ~ a , m ;k r !

Er` 0
2 g 12 2 r
I 11[ e F 21 d r 3F ~ a 8 , m 2 p;k 8 r ! d r . ~B3!
0
To specialize ~B3! to the integrals above, we must set k5k 8 ,
5 Er
0
`
2 g 12 2 r
e F 2 ~ 12n 8 ,2g 12; r ! d r , which we do by choosing k511 e , k 8 512 e and then tak-
ing the limit e→0. The need for this procedure is made clear
by looking at a particular case of ~B3!, namely, s5p50,

I 22[ Er
0
`
2g 2r
e F 22 d r 5 Er
`

0
2g 2r
e F 2 ~ 2n 8 ,2g ; r ! d r ,
which, according to Eq. ~f.13! of Ref. 11, is
J m00~ a , a 8 ! 5 ~ 21 ! a G ~ m ! e 2 ~ a 1 a 8 ! F ~ a , a 8 , m ;2 e 22 ! ,
~B4!

I 12[ Er
0
`
2 g 11 2 r
e F 1F 2d r
where F is the hypergeometric function
aa 8 z a ~ a 11 ! a 8 ~ a 8 11 ! z 2
F ~ a , a 8 , m ;z ! 511 1

Er` m 1! m ~ m 11 ! 2!
2 g 11 2 r
5 e F ~ 2n 8 ,2g ; r ! F ~ 12n 8 ,2g 12; r ! d r .
0 1••• . ~B5!
The indices a and a8 are negative integers; so that Eq. ~B5!
These integrals are special cases of the integral ~f.12! dis- is a polynomial which, in the limit e→0, is determined by the
cussed in Appendix f of Ref. 11, namely, last term only. This is summarized in the relation

e m F ~ a , a 8 , m ;2 e 22 ! →
e →0 H ~ 21 ! n 2
0
n 1!
m.2n 2
G~ m !
~ n 1 2n 2 ! ! G ~ m 1n 2 !
undefined m,2n 2 ,
m52n 2 , ~B6!

where J 202g 12 ~ 12n 8 ,2n 8 ! 52 g ~ 2 g 11 !@ J 200g ~ 12n 8 ,2n 8 !


n 1 5max~ 2 a ,2 a 8 ! , n 2 5min~ 2 a ,2 a 8 ! . 22J 200g ~ 2n 8 ,2n 8 !
The integrals in ~B2! are 1J 200g ~ 212n 8 ,2n 8 !# . ~B10!
I 115J 210g 12 ~ 12n 8 ,12n 8 ! , According to ~B10!, only the middle term survives, and
I 225J 210g ~ 2n 8 ,2n 8 ! , ~B7! G 2~ 2 g !
I 12522 g ~ 2 g 11 ! n 8 !. ~B11!
G ~ 2 g 1n 8 !
I 125J 202g 12 ~ 12n 8 ,2n 8 ! .
Inserting ~B9! and ~B11! into ~B1! gives
The relationship ~f.16! ~of Ref. 11!, for s50,

J m10~ a , a 8 ! 5 S a 82 a
D
1 m 2 a 8 2 a J m00~ a , a 8 ! ~B8!
2n 8 !G 2 ~ 2 g 12 !~ n 8 1 g !
n 8 G ~ n 8 12 g 11 !
11 F ~ n 8 12 g ! n 8
~ 2 g ! ~ 2 g 11 ! c 1
2 2
c2
S D 2

e
2Z a ~ n 8 12 g ! n 8 c2 2
enables us to evaluate the first two integrals of ~B7!. 2 ]c 5 ~ 2l ! 3 , ~B12!
k 2 g ~ 2 g 11 !~ n 8 1 g ! c 1 1
I 115 ~ 2 g 12n 8 ! J 200g 12 ~ 12n 8 ,12n 8 ! and inserting the ratio c 2 /c 1 from ~30! yields finally
G ~ 2 g 12 !
F
2
52 ~ g 1n 8 ! 2n 8 !G 2 ~ 2 g 12 !~ n 8 1 g ! l 2 ~ n 8 12 g ! n 8
~ n 8 21 ! !;
G ~ 2 g 1n 8 11 ! n 8 G ~ n 8 12 g 11 !
11
~ k E1m g ! 2
~B9!
I 225 ~ 2 g 12n 8 ! J 200g ~ 2n 8 ,2n 8 ! 2Z a l ~ n 8 12 g ! n 8
2 ]c 2 5 ~ 2l ! 3 .
G ~2g!
2 k ~ k E1m g !~ n 8 1 g ! 1
52 ~ g 1n 8 ! n 8 !.
G ~ 2 g 1n 8 ! The trigonometric parametrization ~A4!–~A6! again consid-
erably facilitates the reduction of the bracketed terms, with
A little more work is required for I 12. The recursion formula, the result
Eq. ~f.15!, namely,
J ms p ~ a , a 8 ! 5 ~ m 21 !@ J ms,p21
21 ~ a , a 8 ! 2J m 21 ~ a 21,a 8 !# ,
s,p21
c 15
~ 2l ! 3/2
g G ~ 2 g 12 !
A G ~ n 8 12 g 11 ! n 8 k E ~ k E1m g !
4m 2 ~ n 8 1 g ! n 8 !
.
applied twice, yields ~B13!

220 Am. J. Phys., Vol. 65, No. 3, March 1997 B. Goodman and S. R. Ignjatović 220

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
155.97.178.73 On: Wed, 26 Nov 2014 20:39:57
1 9
C. G. Darwin, ‘‘The wave equations of the electron,’’ Proc. R. Soc. Lon- Paul R. Auvil and Laurie M. Brown, ‘‘The relativistic hydrogen atom: A
don, Ser. A 118, 654–680 ~1928!. simple solution,’’ Am. J. Phys. 46, 679–681 ~1978!.
2 10
W. Gordon, ‘‘Energy levels of the hydrogen atom in the Dirac quantum M. K. F. Wong and Hsin-Yang Yeh, ‘‘Simplified solution of the Dirac
theory of electrons,’’ Z. Phys. 48, 11–14 ~1928!. equation with a Coulomb potential,’’ Phys. Rev. D 25, 3396–3401 ~1982!.
3
W. Greiner, Relativistic Quantum Mechanics ~Springer-Verlag, Berlin, 11
L. D. Landau and E. M. Lifschitz, Quantum Mechanics ~Pergamon, Ox-
1990!, pp. 178–182. ford, 1977!, 3rd ed., pp. 662–665.
4
M. E. Rose, Relativistic Electron Theory ~Wiley, New York, 1961!, pp. 12
Milton Abramowitz and Irene A. Stegun ~Eds.!, Handbook of Mathemati-
169–181. cal Functions ~Dover, New York, 1965!, pp. 504–509.
5
J. J. Sakurai, Advanced Quantum Mechanics ~Addison-Wesley, Reading, 13
Reference 11, pp. 117–128.
14
MA, 1967!, pp. 122–131. The physical meaning of the Foldy–Wouthuysen transformation has been
6
Claude Itzykson and Jean-Bernard Zuber, Quantum Field Theory elegantly described in a very recent issue of this journal: John P. Costella
~McGraw-Hill, New York, 1980!, pp. 72–84. and Bruce H. J. McKellar, ‘‘The Foldy–Wouthuysen transformation,’’
7
James D. Bjo” rken and Sidney D. Drell, Relativistic Quantum Mechanics Am. J. Phys. 63, 1119–1121 ~1995!.
~McGraw-Hill, New York, 1964!, pp. 46–60. 15
L. H. Thomas, ‘‘The kinematics of an electron with an axis,’’ Philos. Mag.
8
A. F. Nikiforov and V. B. Uvarov, Special Functions of Mathematical 3, 1–22 ~1927!; W. H. Furry, ‘‘Lorentz transformation and the Thomas
Physics ~Birkhäuser, Boston, 1988!, pp. 330–341. precession,’’ Am. J. Phys. 23, 517–525 ~1955!.

221 Am. J. Phys., Vol. 65, No. 3, March 1997 B. Goodman and S. R. Ignjatović 221

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
155.97.178.73 On: Wed, 26 Nov 2014 20:39:57

You might also like