On the dynamical symmetry of the quantum

Kepler problem
G.F. Torres del Castillo
Departamento de Fsica Matem
atica, Instituto de Ciencias
Universidad Aut
onoma de Puebla, 72570 Puebla, Pue., Mexico
cal
J.L. Calvario Aco
Instituto de Fsica
Universidad Aut
onoma de Puebla, Apartado postal J48,
72570 Puebla, Pue., Mexico

Abstract. Using the fact that the Schrodinger equation for the two-dimensional
Kepler problem with negative energy is equivalent to an integral equation on the
unit sphere in the three-dimensional space, the eigenfunctions and the generators of
a dynamical symmetry group for this problem are obtained from the usual spherical
harmonics and the angular momentum operators on the sphere. It is shown that if
the spherical harmonics are eigenfunctions of Ly , instead of Lz , the corresponding

Rev. Mex. Fs. 44, 344352 (1998)

eigenfunctions of the Schrodinger equation are separable in parabolic coordinates. It

is also shown that in the case of zero energy, the Schrodinger equation for the Kepler
problem in two or three dimensions is equivalent to an integral equation on the twoor three-dimensional Euclidean space, respectively.
Resumen. Usando el hecho de que la ecuacion de Schrodinger para el problema
de Kepler en dos dimensiones con energa negativa equivale a una ecuacion integral
sobre la esfera de radio 1 en el espacio tridimensional, se obtienen las eigenfunciones
y los generadores de un grupo de simetra dinamica para este problema a partir
de los armonicos esfericos usuales y los operadores de momento angular sobre la
esfera. Se muestra que si los armonicos esfericos son eigenfunciones de Ly , en lugar de
Lz , las eigenfunciones correspondientes de la ecuacion de Schrodinger son separables
en coordenadas parabolicas. Se muestra tambien que en el caso de energa cero,
la ecuacion de Schrodinger para el problema de Kepler en dos o tres dimensiones
equivale a una ecuacion integral sobre el espacio Euclideano de dimension dos o tres,
respectivamente.
PACS: 03.65.-w; 02.20.-a; 02.30.Gp

1. Introduction
It is a well known fact that in the Kepler problem (in classical or quantum mechanics),
besides the angular momentum, there exists another conserved vector, known as the
HermannBernoulliLaplaceRungeLenz (HBLRL) vector; whereas the conservation
of the angular momentum follows from the invariance of the potential under rotations,
the conservation of the HBLRL vector is associated with a hidden symmetry, that
is, with transformations that mix the position and momentum variables, leaving the
2

Hamiltonian invariant (see, e.g., Refs. 15).

In the case where the energy is negative, the commutation relations (or the Poisson brackets) between the components of the angular momentum and of the HBLRL
vector coincide with those of a basis for the generators of rotations in the fourdimensional Euclidean space (or of the rotations in three dimensions, if one considers
the Kepler problem in two dimensions) (see, e.g., Refs. 310). Fock [1] showed that
the corresponding Schrodinger equation possesses, in effect, a symmetry group isomorphic to the group of rotations in four dimensions by transforming this equation
into an equation on the unit sphere in the four-dimensional space, where the symmetry becomes obvious. This transformation allows to find the energy levels and to
express the eigenfunctions in terms of hyperspherical surface harmonics [1,3,4].
When the energy is positive, the commutation relations between the components
of the angular momentum and of the HBLRL vector correspond to those of a set of
generators of the Lorentz group and the Schrodinger equation can be transformed into
an equation on a two-sheeted hyperboloid in Minkowski space where the global action
of the Lorentz group can be seen explicitly [1,4]. Finally, when the energy is equal
to zero, the commutation relations of the angular momentum and the HBLRL vector
coincide with those of a set of generators of the Euclidean group in three dimensions
and in Ref. 4 it is shown that the Schrodinger equation can be transformed into an
equation on a paraboloid. However, since the rotations in four dimensions are the
isometries of the unit sphere and the Lorentz transformations are the isometries of an
hyperboloid in Minkowski space, one would expect that in the case of zero energy the
Schrodinger equation could be transformed into an equation on the Euclidean space,
where the symmetry of the Schrodinger equation would be manifest.
In this paper we use the fact that the the Schrodinger equation for the two-

dimensional Kepler problem with negative energy can be transformed into an equation on the unit sphere in three-dimensional space to find the energy levels and the
eigenfunctions explicitly, obtaining a relationship between the generating functions
of the associated Legendre functions and of the associated Laguerre polynomials. We
show that the HBLRL vector (in two dimensions) can be derived from the expressions for the usual angular momentum operators and that the eigenfunctions of one of
the components of the HBLRL vector are the separable solutions of the Schrodinger
equation for the 1/r potential in parabolic coordinates. The analogue of this result
in classical mechanics is given in the Appendix. We also consider the Schrodinger
equation for the Kepler problem in two and three dimensions with zero energy, showing that this equation can be transformed into one on the two- or three-dimensional
Euclidean space, respectively, whose solutions can be easily obtained.

dinger equation for the bound states of the

2. The Schro
two-dimensional Kepler problem
In this section we give a treatment of the Schrodinger equation for the Kepler problem
with negative energy in two dimensions parallel to that given in Refs. 3 and 4 for the
three-dimensional case, following a simpler procedure.

2.1. Invariance of the Schro dinger equation under the three-dimensional

rotation group
By expressing the solution of the Schrodinger equation for the two-dimensional Kepler
problem,

h
2 2
k
= E,
2M
r

(1)

as a Fourier transform,
(r) =
using the fact that
equation

1 Z
(p)eipr/h d2 p,
2
h

(2)

(1/r)ei(pp )r/h d2 r = 2
h/|p p0 |, one obtains the integral
M k Z (p0 ) 2 0
d p,
(p 2M E)(p) =

h
|p p0 |
2

(3)

where p |p|. Throughout this section, we shall consider bound states only, for
which E < 0. Then, by means of the stereographic projection, the vector p can be
replaced by a unit vector n = (nx , ny , nz ) according to [1,3]
p = (px , py ) = p0
where
p0

(nx , ny )
,
1 nz

2M E,

(4)

(5)

or, equivalently,
n = (nx , ny , nz ) =

(2p0 px , 2p0 py , p2 p20 )

.
p2 + p20

(6)

Under the correspondence between p and n given by Eqs. (4) and (6), the plane is
mapped onto the unit sphere and making use of the spherical coordinates , , of n,
from Eq. (4) we find that
p=

p0
(sin cos , sin sin ) = p0 cot(/2)(cos , sin ),
1 cos
5

(7)

therefore,
p = p0 cot(/2),

d2 p = 14 p20 csc4 (/2)d,

(8)

where d = sin dd is the solid angle element, and

|p p0 | =

p0 |n n0 |
= 12 p0 csc(/2) csc(0 /2)|n n0 |,
(1 nz )1/2 (1 n0z )1/2

(9)

where n0 is the unit vector corresponding to p0 according to Eq. (6). Substituting

Eqs. (5), (8) and (9) into Eq. (3) one gets
M k Z csc3 (0 /2)(n0 ) 0
d ,
csc (/2)(n) =
2hp0
|n n0 |
3

hence, by defining
"

p2 + p20

(n)
23/2 p0 csc3 (/2)(n) = p0
2p20

#3/2

(p),

(10)

one arrives at the integral equation

0)
M k Z (n

(n)
=
d0 .
2
hp0 |n n0 |

(11)

The constant factors included in the definition (10) are chosen in such a way that
is normalized over the sphere if and only if is normalized
is dimensionless and
over the plane [3]. Since the distance between points on the sphere and the solid
angle element d are invariant under rotations of the sphere, Eq. (11) is explicitly
invariant under these transformations, thus showing that the rotation group SO(3) is
a symmetry group of the original equation (1), for E < 0. Substituting Eqs. (7), (8)
and (10) into Eq. (2) one obtains the wave function (r) in terms of the solution of
the integral equation (11)
p0 Z
(, ) csc(/2)eip0 cot(/2)(x cos +y sin )/h d.
(r) =
2 2h
6

(12)

The integral equation (11) can be easily solved using the fact that the spherical
harmonics form a complete set for the functions defined on the sphere, therefore the
can be expanded in the form
function
) =
(,

l
X
X

alm Ylm (, ).

(13)

l=0 m=l

Substituting Eq. (13) into Eq. (11), making use of the expansion
X
l
X
1
4
=
Y (0 , 0 )Ylm (, ),
0
|n n | l=0 m=l 2l + 1 lm

where , and 0 , 0 are the spherical coordinates of n and n0 , respectively, and of

the orthonormality of the spherical harmonics one obtains
X
l
X

"

l=0 m=l

2M k
1
alm Ylm (, ) = 0,
h
p0 (2l + 1)

which implies that, in order to have a nontrivial solution, 2M k = h

p0 (2l + 1), for
some l; hence, according to Eq. (5),
E=

2M k 2
,
h
2 (2l + 1)2

(14)

(cf. Ref. 10). The 2l + 1 coefficients alm , (m = l, l + 1, . . . , l) are arbitrary and

al0 m = 0 for all l0 6= l. Thus, the degeneracy of the energy level (14) is 2l + 1; all the
spherical harmonics of degree l are solutions of Eq. (11), corresponding to the energy
(14) and the solutions of the homogeneous integral equation (11) are precisely the
eigenfunctions of L2 , the square angular momentum operator.

2.2. Explicit form of the eigenfunctions

According to the preceding results, the solutions of the Schrodinger equation (1), for
E < 0, are given by Eq. (12),
p0 Z
(, ) cos(/2)eip0 cot(/2)(x cos +y sin )/h dd,
(x, y) =
2h
7

(15)

 ) is an eigenfunction of L2 . The rotational symmetry of the Hamiltonian

where (,
(1) suggests the use of polar coordinates, in terms of which Eq. (15) takes the form
p0 Z
(r, ) =
2
h
p0 Z
=
2
h

) cos(/2)eip0 cot(/2)r cos()/h dd

(,
)
(,

im eim () Jm0 (p0 r/h) cot(/2) cos(/2)dd,(16)

m0 =

where we have made use of the expansion eix sin =

m=

eim Jm (x) (see, e.g.,

) as a spherical harmonic, Ylm (, ), which are the

Ref. 11, Sec. 4). Taking (,
normalized separable eigenfunctions of L2 in spherical coordinates, from Eq. (16) we
obtain
"

2l + 1 (l m)!
p0 Z
lm (r, ) =
(1)m
4 (l + m)!
2h

p0 2l + 1 (l m)!
=
h

2 (l + m)!

Z
0

Plm (cos )eim

im eim () Jm0 (p0 r/h) cot(/2) cos(/2)dd

m0 =

"

#1/2

#1/2

Plm (cos )Jm

(i)m

(p0 r/
h) cot(/2) cos(/2)d eim ,

(17)

which shows that the separable eigenfunctions of L2 in spherical coordinates correspond to separable eigenfunctions of the Hamiltonian in polar coordinates. Denoting
by Ilm the integral between braces in Eq. (17), and introducing an auxiliary parameter
t we have

(2l + 1)Ilm tl =

l=0

Z X

(2l + 1)Plm (cos )tl Jm (p0 r/

h) cot(/2) cos(/2)d

l=0

m+1
(cos )
therefore, making use of the recurrence relation (2l+1) sin Plm (cos ) = Pl+1
m+1
Pl1
(cos ) and of the generating function of the associated Legendre functions,

X
(2m)!(1 x2 )m/2
=
P m (x)tk
2m m!(1 2tx + t2 )m+1/2 k=0 m+k

(see, e.g., Ref. 12), for m 0, we find

Z
(1 t2 )tm

(2l + 1)Ilm t =

l=0

sin

(2m + 2)!(1 cos2 )(m+1)/2

2m+1 (m + 1)!(1 2t cos + t2 )m+3/2

Jm (p0 r/h) cot(/2) cos(/2)d.

Replacing the variable by cot(/2) one finds that

m+1
Z
2 m

J
(p
r/
h
)
m
0
(1 t )t (2m + 2)!
(2l + 1)Ilm tl =
d.

2m+3
2 m+3/2
(1 t)
(m + 1)!
0

X
l=0

2 +

1+t
1t

The last integral can be evaluated by first differentiating with respect to s the equation
Z
0

exs xm Jm (x)dx =

(2m)!
2m m!(s2 + 1)m+1/2

(see, e.g., Ref. 11, Sec. 15), which yields

Z
0

Using now the fact that

exs xm+1 Jm (x)dx =

Z
0

(2m + 1)!s
.
+ 1)m+3/2

(18)

2m m!(s2

f (x)Jn (xy)xdx = g(y) amounts to

Z
0

g(x)Jn (xy)xdx =

f (y) (see, e.g., Ref. 11, Sec. 13), from Eq. (18) one gets
Z
(2m + 1)! sxm+1 Jm (xy)
0

thus

2m m!

(x2 + s2 )m+3/2

dx = y m eys

(2l + 1)Ilm tl = 2m+1 tm (p0 r/h)m ep0 r/h

l=0

and recalling that

e(2p0 r/h)t/(1t)
(1 t)2m+1

X
exz/(1z)
=
Lkn (x)z n ,
k+1
(1 z)
n=0

where Lkn denote the associated Laguerre polynomials (see, e.g., Ref. 13), it follows
that

l=0

k=0

(2l + 1)Ilm tl = 2m+1 ep0 r/h (p0 r/h)m tm

9

h)tk
L2m
k (2p0 r/

therefore
2m+1 p0 r/h
e
(p0 r/
h)m L2m
h)
lm (2p0 r/
2l + 1
and the normalized eigenfunctions of the Hamiltonian, for m 0, are given by
Ilm =

"

p0 2l + 1 (l m)!
lm (r, ) =
h

2 (l + m)!

#1/2

Using the relations Plm = (1)m

(i)m

2m+1 p0 r/h
h). (19)
e
(p0 r/h)m L2m
lm (2p0 r/
2l + 1

(l m)! m
P and Jm = (1)m Jm , from Eq. (17) it
(l + m)! l

follows that

.
l,m = lm

(20)

2.3. The generators of the symmetry

As we have shown, Eq. (15) gives a correspondence between the solutions of the
integral equation (11) and those of the Schrodinger equation (1). As remarked above,
Eq. (11) is explicitly invariant under the rotations of the sphere and, as is well known,
a set of generators of these rotations are the angular momentum operators
!

x = i
,
L
h sin + cot cos

y = i
L
h cos + cot sin
,

z = ih ,
L

(21)

where the indicates that these operators act on functions defined on the sphere.
Then, by means of the correspondence (12) we can find the operators on the wave
functions that correspond to the generators of rotations (21).
on the sphere correFrom Eqs. (2), (7) and (10) it follows that the function
sponding to a given wave function is
Z
p0
1

(n)
=
(r)eip0 cot(/2)(x cos +y sin )/h d2 r.
3
sin
(/2)
4 2h

10

(22)

 x to both sides of the last equation one obtains

By applying, for example, L

Z
ip0
ip0
1

Lx =
(r) 3 cot(/2) sin +
cot2 (/2)(2x sin cos
3
h

8 2
h sin (/2)

ip0
+ y sin2 y cos2 ) +
y eip0 cot(/2)(x cos +y sin )/h d2 r,
h

which can be rewritten as

"

Z
2
p20
h

(r)
3
+
2x
+
y

y eipr/h d2 r
y
x y
y 2 x2
h
2
8 2 sin3 (/2)
"

!
#
Z
h

2
2
p20
ipr/
h
2 2 y (r)d2 r,
=
e
+x
+y
3
2
y
x y
y
x
h

8 2 sin (/2)

where we have integrated by parts. Now, assuming that satisfies Eq. (1), the last
term can be replaced according to p20 = h
2 2 + (2M k/r), hence,
"

Z
h
2
p0
ipr/
h 1

e
Lx =
p0 2
4 2h sin3 (/2)

2
+ 2x
2y 2
y
x y
x

which is of the form (22), with replaced by

1
p0

2
h
2

x y
y x

2
h
2 x

x y
y x

1
p0

2
h
2

M ky
r

M ky

(r)d2 r,
r
2

+ 2x x
2y x

2
y

M ky
r

. Thus, under the correspondence

x
given by Eq. (22), restricted to the solutions of Eq. (1) for a fixed value of E, L
corresponds to the operator

1
p0

2
h
2

x y
y x

2
h
2 x

x y
y x

M ky
r

, which,

apart from the constant factor (1/p0 ), coincides with Ay , one of the cartesian components of the HBLRL vector
1
M kr
A (p L L p)
,
2
r

(23)

where L = r p and p = ih (cf. Refs. 3 and 14).

y and L
z correspond to (1/p0 )Ax
In a similar manner, one finds that the operators L
and Lz , respectively. Thus, one concludes that the components of the HBLRL vector (23) are associated with the SO(3) symmetry of Eq. (1) and that the operators
x, L
y and L
z.
(1/p0 )Ay , (1/p0 )Ax and Lz obey the same commutation relations as L
11

2.4. Separation of variables in parabolic coordinates

The Schrodinger equation (1) is also known to be separable in parabolic coordinates
(see, e.g., Ref. 6). In fact, in terms of the parabolic coordinates u and v defined by
x = 21 (u2 v 2 ), y = uv, Eq. (1) takes the form
h
2
1

2
2M u + v 2

2 2
+ 2
u2
v

u2

2k
= E.
+ v2

(24)

Substituting = U (u)V (v) into Eq. (24), one obtains the separated equations
h
2 d2 V

Ev 2 V = (k )V,
2M dv 2

h
2 d2 U
Eu2 U = (k + )U,

2M du2

(25)

where is a separation constant. Each of these separated equations, for E < 0, has
the form of the Schrodinger equation for a harmonic oscillator (cf. Refs. 6, 7 and 15);
q

making use of the dimensionless variables

p0 /
h u and

p0 /h v, from Eqs.

(25) we get

2M (k + )
d2 U
+ 2U =
U,
2
d
h
p0

d2 V
2M (k )
+ 2V =
V,
2
d
h
p0

therefore, in order to have well-behaved solutions of Eqs. (25),

M (k + )
= n1 + 12 ,
h
p0

M (k )
= n2 + 21 ,
h
p0

(26)

where n1 and n2 are non-negative integers, and

= Ce

2 /2

Hn1 ()e

2 /2

2 /2
h

Hn2 () = Cep0 u

h u)ep0 v
Hn1 ( p0 /

2 /2
h

Hn2 ( p0 /h v),
(27)

where the Hn are Hermite polynomials and C is a normalization constant [6].

Since Hn (x) = (1)n Hn (x) and the couples (u, v) and (u, v) correspond to
the same point (x, y), in order to have a single-valued wave function it is necessary
12

that n1 and n2 be both odd or even; hence, n1 + n2 must always be an even number,
2l, say, and, for a fixed value of l,
my 12 (n1 n2 ),

(28)

can take the (2l + 1) integral values l, l + 1, . . . , l. Then, adding Eqs. (26) and
recalling Eq. (5) we obtain again E = 2M k 2 /[h2 (2l + 1)2 ] and subtracting Eqs. (26)
we find that the separation constant must be quantized according to
=

h
p0
my .
M

(29)

(Note that the existence of 2l + 1 different values of my for a given value of l means
that the degeneracy of the energy levels is 2l + 1.)
The meaning of the separation constant (and, hence, of my ) can be found by
multiplying the first equation in (25) by v 2 V , the second one by u2 U and subtracting,
which leads to

2
2
1
h
2
k(u2 v 2 )
2
2

u
=

+
2M u2 + v 2
u2
v 2
u2 + v 2

or, equivalently,
"

1
h
2

h
2

x
y

x
y
M
2 y
y
x
2
y
x

M kx

= ,
y
r

(30)

which is just the condition (1/M )Ax = (cf. Eq. (23)). Hence, the separable
solutions of Eq. (1) in parabolic coordinates are eigenfunctions of (1/p0 )Ax (which
y ), with eigenvalue M /p0 =
is the operator on the wave functions corresponding to L
corresponding (by means of
my h
. Thus, whereas the functions on the sphere, ,
Eq. (22)) to the separable solutions (19) of Eq. (1) in polar coordinates are the
z and L2 , those corresponding to the separable solutions (27) of
eigenfunctions of L
y and L2 .
Eq. (1) in parabolic coordinates are the eigenfunctions of L
13

It should be remarked that, as we have seen in the preceding paragraph, the

conservation of the HBLRL vector follows from the separability of the Schrodinger
equation (1) in parabolic coordinates. (As shown in the Appendix, in a similar
manner, the conservation of the classical HBLRL vector follows from the separability
of the HamiltonJacobi equation for the two-dimensional Kepler problem.) In this
context, the HBLRL vector arises in a very natural way (compare with the derivation
of the classical HBLRL vector given, e.g., in Ref. 16).
We close this section pointing out that from the commutation relations of the
i, L
j ] = ihijk L
k , it follows, in the usual way,
angular momentum operators (21), [L
z iL
x on an eigenfunction of L2 and L
y yields another eigenthat the action of L
y with the same eigenvalue of L2 and the eigenvalue of L
y shifted
function of L2 and L
by
h. Therefore, owing to the results of the preceding subsection, the operators
z iL
x ) raise or lower the eigenvalue my of the
Lz (i/p0 )Ay (which correspond to L
wave functions
lmy = Cep0 u

2 /2
h

Hl+my ( p0 /h u)ep0 v

2 /2
h

Hlmy ( p0 /
h v)

(31)

[see Eqs. (27) and (28)] by one unit. In fact, a straightforward computation, using
again the relation p20 = h
2 2 + (2M k/r), gives the simple expression
h

i
Lz Ay =
p0
i

p0
u
2h

h

2p0 u

! r

p0
v
2h

h

.
2p0 v

(32)

The operators in the right-hand side of the last equation can be recognized as raising
or lowering operators corresponding to the linear harmonic oscillators described by
Eqs. (25). Letting a1
a2

Lz +

p0 /2
hv +

p0 /2h u +

h
/2p0 /v, a2

i
Ay = iha1 a2 ,
p0

Lz

h
/2p0 /u, a1

p0 /2h v

i
Ay = i
ha1 a2 ,
p0
14

p0 /2h u

h
/2p0 /u,

h
/2p0 /v, we have

1
h

Ax = (a1 a1 a2 a2 ). (33)
p0
2

These equations correspond to the well-known Schwingers realization of the Lie algebra of the rotation group in terms of creation and annihilation operators.

3. The quantum-mechanical Kepler problem with zero energy

3.1. Explicit SE(2) invariance of the two-dimensional Kepler problem with

zero energy
When E = 0, Eq. (3) gives
p2 (p) =

M k Z (p0 ) 2 0
d p.
h
|p p0 |

(34)

Making now the change of variable

p=

2M k q
,
h
q2

(35)

where q is a dimensionless vector in the plane and q = |q|, we get

p=

2M k
,
h
q

d2 p =

4M 2 k 2 2
d q,
h
2q4

|p p0 | =

2M k |q q0 |
,
h

qq 0

and from Eq. (34) it follows that

where

0)
1 Z (q

(q)
=
d2 q0 ,
2 |q q0 |

(36)

2M k
h
2 p3

(q)

(q)
=
(p).
h
q3
4M 2 k 2

(37)

Equation (36) is explicitly invariant under SE(2), the group of rigid transformations
of the Euclidean plane onto itself that do not change the orientation.

15


The homogeneous integral equation (36) can be easily solved. Substituting (q)
=
R

f (s)eisq d2 s into Eq. (36) we get

Z

0
1 Z Z f (s)eisq d2 sd2 q0 Z f (s)eisq 2
d s=
=
ds
2
|q q0 |
s

isq 2

f (s)e

hence, f (s) 6= 0 only if |s| = 1; thus, (q)

= eisq is a solution of Eq. (36) for any (constant) unit vector s. These solutions are separable in cartesian coordinates: eisq =
eisx x eisy y , with q = (x, y) and s = (sx , sy ). Since eisq =

m=

im eim eim Jm (),

where (, ) are the polar coordinates of q and s = (cos , sin ), it follows that
) = Jm ()eim is a separable solution in polar coordinates of Eq. (36), for any
(,
integral value of m. Hence, the degeneracy of the energy level E = 0 is infinite. It
may be noticed that the solutions to Eq. (36) coincide with those of the Helmholtz
= .
A somewhat similar result holds in the case of the integral
equation 2
= l(l + 1)h2 ,

equation (11), which is equivalent to the eigenvalue equation L2

since
h2 L2 is the Laplace operator of the sphere; however, in the latter case, all
non-negative integral values of l are allowed, while in the case of the Laplace operator
of the plane, only one of its eigenvalues is relevant, which is related to the fact that
E only takes the value zero.
From Eqs. (2), (35) and (37) it follows that the solutions of the Schrodinger
through
equation (1) with E = 0 are related with the functions
Mk Z
2

(, ) =
(x, y)e2iM k(x cos +y sin )/h d2 r,
2 3

(38)

where, again, , are the polar coordinates of q. On the other hand, the generators
of the rigid motions of the plane can be chosen as
!

sin

,
P1 = i
h cos

cos

P2 = i
h sin
+
,

16

(39)


z = i
L
h .

z genP1 and P2 generate translations in the x and y directions, respectively, and L

erates rotations about the origin. The commutation relations of the operators (39)
are
[P1 , P2 ] = 0,

z , P1 ] = ihP2 ,
[L

z , P2 ] = ihP1 .
[L

(40)

The operators on the wave functions corresponding to the generators of the symmetry
transformations (39) can be obtained following the same steps as in Sec. 2.3, making
use of Eq. (38). One finds, for instance,
"

!
#
3
M k Z ipr/h h

2
2

e
P1 =

+x

2y
(x, y)d2 r
2M k
x
y 2 x2
x y
h2 3

hence, recalling that in the present case h

2 2 + (2M k/r) = 0, the operator corresponding to the generator of translations P1 is
"

h
3

2
2M k

+ 2x 2 + 2 x 2y
2M k
x
y
x y
h
r
( 2 "
!

!#
)
h

M kx
h

=
x
x
+
=
y
+
y
Ax
Mk 2
y
x y y
y
x
r
Mk
[see Eq. (23)]. In a similar way, one finds that the operators corresponding to P2 and
z are (
L
h/M k)Ay and Lz , respectively. Thus, Ax , Ay and Lz generate symmetry
transformations of the Schrodinger equation (1), as in the case where E is negative.
z under the correspondence
The fact that Lz is the operator corresponding to L
(38) implies that, under this relationship, the separable solutions in polar coordinates
of the Schrodinger equation with E = 0 correspond to the separable solutions in polar
coordinates of Eq. (36) (which are of the form Jm ()eim ). On the other hand, since
Eq. (30) holds for any value of E, the separable solutions in parabolic coordinates of
the Schrodinger equation with E = 0 are eigenfunctions of Ax and, therefore, they
17

correspond to eigenfunctions of P1 , which are the separable solutions in cartesian

coordinates of Eq. (36).

3.2. Symmetry of the Schro dinger equation for the three-dimensional Kepler
problem with zero energy
Considering now the Schrodinger equation (1) in three dimensions, writing
(r) =

Z
1
(p)eipr/h d3 p,
(2
h)3/2

one obtains the integral equation

(p2 2M E)(p) =

M k Z (p0 ) 3 0
d p,
2h
|p p0 |2

(41)

which takes the place of Eq. (3). Making E = 0 and p = 2M kq/(

hq 2 ), where q is a
dimensionless vector in three dimensions, one finds that Eq. (41) is equivalent to
0)
1 Z (q

(q)
= 2
d3 q0 ,
2
|q q0 |2
where

!3/2

(42)

!5/2

2M k
(q)
h

(q)

p4 (p).
=
4
h

q
2M k
Equation (42) is manifestly invariant under the rigid transformations of the threedimensional Euclidean space, which means that this group of transformations constitute a symmetry group of the three-dimensional Kepler problem with zero energy.
= eisq , where s is a constant vector
It can be easily seen that the functions
= jl ()Ylm (, ) are separable solutions of Eq. (42) in cartesian
with |s| = 1, and
coordinates and spherical coordinates, respectively, and these functions are also solu = ;
the existence of these sets of separable
tions of the Helmholtz equation 2
solutions corresponds to the separability of the Schrodinger equation (1) in parabolic
and spherical coordinates.
18

4. Concluding remarks
The cases considered in this paper, as well as those treated in Refs. 3 and 4, show the
usefulness of exhibiting the underlying symmetry of the quantum Kepler problem,
which allows to change the Schrodinger equation by a simpler condition.
The results of Sec. 2 show, among other things, that the usual spherical harmonics
are related with the associated Laguerre polynomials and the Hermite polynomials,
which form bases for representations of the rotation group; these results also show
one of the many connections between the Kepler problem with negative energy and
the isotropic oscillator (cf. also Ref. 9 and the references cited therein).

Acknowledgments
This work was supported in part by CONACYT. The authors would like to thank
Prof. K.B. Wolf for correspondence relating to the topic of this paper and Profs. H.V.
McIntosh and E.G. Kalnins for useful comments.

Appendix
The separability of the Schrodinger equation and of the HamiltonJacobi equation
for the two-dimensional Kepler problem in polar coordinates is a consequence of the
invariance of the Hamiltonian under rotations about the center of force (as in the
case of any central potential) which, in turn, is equivalent to the conservation of
the angular momentum Lz . The 1/r potential is distinguished by the existence of
another conserved vector the HBLRL vector which turns out to be related with
the separability of the above-mentioned equations in parabolic coordinates.
19

The HamiltonJacobi equation for the two-dimensional Kepler problem written

in the parabolic coordinates u, v, defined by x = 12 (u2 v 2 ), y = uv, is

1
1
S

2
2
2M u + v
u

!2

S
+
v

!2

u2

S
2k
+
= 0.
2
+v
t

(A.1)

Even though both coordinates are non-ignorable, Eq. (A.1) admits separable solutions
of the form
S(u, v, t) = Et + f (u) + g(v).

(A.2)

In fact, substituting Eq. (A.2) into Eq. (A.1) one obtains the separated equations
1
2M

df
du

!2

1
2M

k Eu = ,

dg
dv

!2

k Ev 2 = ,

(A.3)

where is a separation constant. Multiplying the first equation in (A.3) by v 2 and

the second one by u2 and subtracting one finds that

1 2 df
v
2M
du

!2

u2

dg
dv

!2
+ k(u2 v 2 ) = (u2 + v 2 ),

(A.4)

therefore, since u2 + v 2 = 2r, df /du = S/u = pu = px (x/u) + py (y/u) =

upx + vpy and dg/dv = S/v = pv = vpx + upy , from Eq. (A.4) it follows that
"

1
M kx
Ax
=
(xpy ypx )py
= ,
M
r
M
where A p L M kr/r now denotes the classical HBLRL vector (cf. Eq. (30)).
From Eqs. (A.3) it follows that, if E 0, then k k (cf. Eqs. (26)).
In a similar way one finds that Ay corresponds to a separation constant using the
coordinate system y = 12 (v 2 u2 ), x = uv.

References
1.

V. Fock, Z. Physik 98 (1935) 145.

20

2.

V. Bargmann, Z. Physik 99 (1936) 576.

3.

M. Bander and C. Itzykson, Rev. Mod. Phys. 38 (1966) 330.

4.

M. Bander and C. Itzykson, Rev. Mod. Phys. 38 (1966) 346.

5.

H.V. McIntosh, in Group theory and its applications, Vol. 2, E.M. Loebl (ed.),
(Academic Press, N.Y., 1971).

6.

A. Cisneros and H.V. McIntosh, J. Math. Phys. 10 (1969) 277.

7.

A. Cisneros and H.V. McIntosh, J. Math. Phys. 11 (1970) 870.

8.

M.J. Englefield, Group Theory and the Coulomb Problem, (Wiley, N.Y., 1972).

9.

O.L. de Lange and R.E. Raab, Operator Methods in Quantum Mechanics, (Oxford University Press, Oxford, 1991).

10. G.F. Torres del Castillo and J.L. Calvario Acocal, Rev. Mex. Fs. 43 (1997) 649.
11. H. Hochstadt, The Functions of Mathematical Physics, (Wiley-Interscience, New
York, 1971), reprinted by Dover (New York, 1986), Chap. 8.
12. G. Arfken, Mathematical Methods for Physicists, 3rd ed. (Academic Press, San
Diego, 1985), Chap. 12.
13. H. Hochstadt, op. cit., Chap. 2.
14. K.B. Wolf, Supl. Nuovo Cimento 5 (1967) 1041.
15. M. Moshinsky, T.H. Seligman and K.B. Wolf, J. Math. Phys. 13 (1972) 901.
16. H. Goldstein, Classical Mechanics, 2nd ed. (Addison-Wesley, Reading, Mass.,
1980).

21