You are on page 1of 6

Discrete variable method for nonintegrable quantum systems

W. Schweizera! and P. Fassbinder


Institut fur Astronomie und Astrophysik, Abteilung Theoretische Astrophysik, Universitat Tubingen,
72076 Tubingen, Germany
(Received 11 April 1997; accepted 8 August 1997)

We present an effective numerical algorithm based on discrete variable techniques combined with
finite elements for solving the non-separable three-dimensional Schrodinger equation. As an
example we study the hydrogen atom in strong external magnetic and electric fields, a doorway
system in the quest of quantum chaology and for studying atomic properties relevant for
astrophysical questions. 1997 American Institute of Physics. @S0894-1866~97!01806-3#

INTRODUCTION
The availability of high-speed computers has opened the
door for studying nonintegrable quantum systems. One of
the fascinating aspects of nonintegrable, low-dimensional
systems is given by their relation to quantum chaology.
Such systems are also fundamental to research in all areas
of physics, e.g., in astrophysics, theoretical atomic physics,
or solid-state physics.
In Sec. I, we will discuss the phrase nonintegrability and describe the computational method. In Sec. II, we
will illustrate this approach by studying the hydrogen atom
in strong external fields. Recent years have seen tremendous progress in studies of quantum chaology and modern
atomic physics. A decisive stimulus came from the discovery that even simple systems like the hydrogen atom in
strong magnetic fields exhibit chaotic behavior in both experimental observations and theoretical calculations. The
hydrogen atom in strong magnetic fields cannot be solved
analytically, as the spherical symmetry of the Coulomb potential is broken by the cylindrical symmetry of the diamagnetic potential due to the external magnetic field.
Therefore, this simple system becomes nonintegrable and
hence a model system for studying quantum chaology. The
problem of hydrogenlike systems in external fields is of
fundamental interest in the atomic, astrophysical, and solidstate context.
In Sec. II we will show some results in the nonintegrable region of the hydrogen atom in external magnetic
and electric fields given by the discrete variable method
combined with finite elements. In Sec. III we will discuss
some advantages of this method compared to other computational methods.

\2
D1V ~ r! c E ~ r! 5E c E ~ r! ,
2m

where the wavefunction in a stationary state with energy E


is denoted by c E (r). In this connection the question arises
as to into which curvilinear coordinates the Schrodinger
equation ~1.1! can be separated. By methods of group
theory1 it can be shown that in three-dimensions the
LaplaceBeltrami operator D can be separated into exactly
11 different curvilinear coordinates. For each of these coordinates the potential V(r) has to fulfill certain properties,
so that the Schrodinger equation can be separated. For
spherical coordinates
x5r sin u cos f ,

0<r,`,

y5r sin u sin f ,

0< u , p ,

z5r cos u ,

0< f ,2 p ,

V ~ r, u , f ! 5V r ~ r ! 1

V u~ u !
r

a!

E-mail: schweizer@tat.physik.uni-tuebingen.de

1997 AMERICAN INSTITUTE OF PHYSICS 0894-1866/97/11~6!/641/6/$10.00

V f~ f !
r 2 sin2 u

~1.3!

and for cylindrical coordinates


x5 r cos f ,
y5 r sin f ,

0< r ,`,
0< f ,2 p ,

~1.4!

2`,z,`,

~1.1! can be separated if the potential can be written as

A. Separability versus nonintegrability


We begin with the time-independent Schrodinger equation
for a three-dimensional one-particle system,

~1.2!

~1.1! can be separated into three ordinary differential equations if

z5z,
I. THE METHOD

~1.1!

V ~ r , f ,z ! 5V r ~ r ! 1

V f~ f !

r2

1V z ~ z ! .

~1.5!

For all orthogonal curvilinear coordinates q j in three


dimensions the following conditions hold:
COMPUTERS IN PHYSICS, VOL. 11, NO. 6, NOV/DEC 1997

641

ds 2 5dx 2 1dy 2 1dz 2 ,

from use of the globally defined basis, which is only correct


for systems of certain symmetry, whereas a locally defined
basis can be optimized with respect to the true system
under consideration.

with ~ x,y,z ! Cartesian coordinates,


3

ds 5
2

i, j51

g i j dq i dq j
B. The computational method

and
g i j 5 d i j g ii ,
D q1,q2,q3 5

( ]qi
Ag i51

SA D
g ]
g ii ] q i

~ LaplaceBeltrami operator! ,

dV5 Agdq 1 dq 2 dq 3

~ volume element! ,

with g5g 11g 22g 33 , and the quantum system becomes separable if the potential can be written as
3

V ~ q 1 ,q 2 ,q 3 ! 5

( V~ qi!.
i51 g ii

~1.7!

For separable systems the eigenvalues ~separation constants! of each of the three unidimensional ~in general
d-dimensional! differential equations can be used to label
the eigenfunctions c (r) and hence serve as quantum numbers. Integrability of a d-dimensional Hamiltonian system
requires the existence of d commuting observables O i ,
1<i<d. Each separable system is integrable, but not vice
versa. Observables are given by Hermitian operators, and
their eigenvalues can serve as quantum labels for the wave
functions. Details can be found in most quantum-mechanics
textbooks ~see, for example, Ref. 2!. Commuting Hermitian
operators require certain corresponding symmetry properties for the system under consideration. For spherical symmetric systems the Hamiltonian H commutes with each
component L i , i5x, y, or z of the angular momentum
operator. As the operators L i do not commute with each
other, a complete set of commuting operators is given, for
example, by $ H,L z ,L 2 % , and the wavefunctions of those
systems are labeled by the corresponding eigenvalues
$ m,l % of $ L z ,L 2 % , and an additional quantum number, for
example, the radial quantum number n r . ~For the field-free
hydrogen atom: n r 5n212l, with n the principal quantum
number.! Systems with cylindrical symmetry are invariant
under rotation along the cylindrical axis, which we choose
as z; hence the Hamiltonian commutes with L z and m is a
good quantum number, as it is for spherical systems.
For nonintegrable systems there are mainly two textbook methods by which to compute eigensolutions. By rewriting the potential into an integrable V int and into a nonintegrable part V non , with V non sufficiently weak, solutions
can be obtained either by perturbation theory or by solving
the eigenvalue problem of the Hamiltonian matrix calculated on a suitable basis. If V non becomes comparable with
V int , both methods hardly converge and become very
costly in terms of computational time. Discretization methods are more efficient to compute eigensolutions of nonintegrable systems. Methods based on basis expansions, such
as perturbational or direct diagonalization treatments, suffer
642

COMPUTERS IN PHYSICS, VOL. 11, NO. 6, NOV/DEC 1997

The motivation for using the discrete variable method3 in


our approach is to transform the three-dimensional Schrodinger equation into a system of coupled unidimensional
differential equations. To derive the differential system, the
Hamiltonian is expanded with respect to the discretized
eigensolution of a suitable operator, which we choose as
the angular momentum operator. But note that this expansion is not justified by a certain approximate symmetry of
the system under consideration; it is a numerical way to
optimize the computation. The resulting coupled ordinary
differential equations will then be solved by finite elements.
By using appropriate units (\51, m e51) and spherical coordinates, the one-particle Hamiltonian becomes
H ~ r, u , f ! 52

1
] 2 ]
]
r
1 2 L 2 ~ u , f ! 2i b
]f
2r ] r ] r 2r
1

1V ~ r, u , f ! ,
where
L 2 ~ u , f ! 52

~1.8a!

1
1
]
]
]2
sin u 1
,
sin u ] u
] u sin u ] f 2

~1.8b!

and the derivative operator i( ] / ] f )5L z is already included to describe additional external magnetic fields with
field strength b . A complete set of orthonormal eigenfunctions of the angular momentum operator L 2 ~1.8b! is given
by

x n~ V ! 5

cos m f ,
1 m
P l ~ cos u !
Cn
sinu m u f ,

where V5 $ u , f % ,
tions,
C n5

Pm
l

for m>0
for m,0

,
~1.9!

are the associated Legendre func-

2p

11 d 0m ~ l1 u m u ! !
2l11 ~ l2 u m u ! !

and
L 2 x n 5l n x n ,
l n 5l ~ l11 ! ,
with

n 5 $ l,m % .
By discretizing the two-dimensional subspace V, the
three-dimensional Schrodinger equation is mapped onto a
system of a unidimensional differential equation. Therefore, we introduce on the subspace V a difference grid with
N nodal points V k :
VV k 5 $ u k , f k % ,

~1.10a!

c ~ r,V ! c ~ r,V k ! 5 c k ~ r ! 1i
c k~ r ! ,

~1.10b!

where c k (r) and


c k (r) are real scalar functions. Note that
for vanishing coefficients b in the Hamiltonian ~1.8a! the
radial expansion will be real and hence
c k (r)50 @see also
Eq. ~1.13e!#. For systems with rotational symmetry around
the z axis, the angular momentum L z is conserved; hence m
is a good quantum number. In this case the original threedimensional
system
can
be
separated
in
a
(2 % 1)-dimensional system, with coordinates (r, u ) for the
nonseparable part, and therefore only the coordinate u will
be discretized to derive the system of the ordinary differential equations.
One important computational point is the optimized
choice of nodal points. Let us first consider the twodimensional case, m conserved. Up to normalization constants the eigenfunctions ~1.9! are then reduced to the associated Legendre polynomials, and only the angular
coordinate u has to be discretized. Taking the values of the
first N eigenfunctions P m
l (cos uk) at the N nodal points u k ,
we construct a square matrix x of maximum rank. The
associated Legendre polynomials are orthogonal
polynomials.4 By taking the N nodal points u k at the zeros
m
of the Legendre polynomial P N1m
(x) in the interval x
P @ 21,1# , a set of orthogonal vectors $ P m
l (cos uk),
m<l<(N211m) % is defined and with suitable normalization constants the matrix x becomes orthogonal and its inverse is simply given by transposing. Note that the corresponding GaussLegendre quadratures are exact. This is no
longer true in general. For three dimensions we design a N 2
grid by choosing N nodal points with respect to u as described above and, for the variable f , N equidistant nodal
points in the interval (0,2p ). Taking the value of the first
N 2 eigenfunctions x n (V) ~1.9! at the N 2 nodal points V k as
defined above, the square matrix x becomes

x kn5 x n~ V k ! ,
and its inverse x 21 can be constructed. To simplify the
computation with respect to all those parts of the Hamiltonian, which are independent of angular derivative operators, we expand the wavefunction in terms of x n (V) and
the radial functions ~1.10b!:
N2

c ~ r,V ! 5 (

k51

S(

N2

n 51

x n ~ V ! x 21
n k @ c k ~ r ! 1i c k ~ r !# .
~1.11!

By normalizing this three-dimensional wavefunction


N2

15

N2

21
( n(51 x 21
nk x n j
k, j51

1
c k~ r !
c j ~ r !# ,

DE

drr 2 @ c k ~ r ! c j ~ r !
~1.12!

we get a normalization condition for the radial functions


~1.10b! to keep the problem symmetric. Inserting the expansion ~1.11! together with the normalization condition
~1.12! into the Schrodinger equation ~1.8a! at the nodal

points V k leads to a system of 2N 2 coupled unidimensional


differential equations for the radial functions c k (r) and

c k (r):
N2

05

j51

c j~ r !% ,
v k j~ r !
$ @ d k j ~ r ! 1 v k j ~ r !# c j ~ r ! 1
~1.13a!

N2

05

j51

c j~ r !2
v k j~ r !c j~ r !% ,
$ @ d k j ~ r ! 1 v k j ~ r !#
~1.13b!

where

d k j~ r !5 2

d 2 d
r
1V ~ r,V j ! 2E g k j ,
2r 2 dr dr
~1.13c!
1

and
v k j~ r !5

N2

2r 2 n 51

21
l n x 21
nk x n j ,

~1.13d!

N2

v k j~ r !5 b

n 51

21
m x 21
nk x 2n j,

2 n 5 $ l,2m % .
~1.13e!
5

Applying the finite-element method, this system of


differential equations is transformed into a generalized eigenvalue problem with banded symmetric matrices that can
be solved, for example, by a spectral transformation Lanczos method.6
As the length of the single finite element must not
necessarily be constant, the finite-element calculation can
be optimized by using single elements of appropriate
length. For Coulomb-like potentials the wavefunction is
more structured close to the origin and the distance of
neighboring nodes of the bound-state wavefunctions increases approximately quadratically. Therefore, the finiteelement calculations are optimized by quadratically widening the length of the single element. The left border of the
ith element is then given by
r i5

~ i21 ! 2

M2

r max ,

i51...M ,

~1.14!

with M the number of elements and r max the maximum


integration distance. But note that such elements would not
be a good choice, for example, for harmonic oscillator potentials. For details concerning the finite-element calculations see Ref. 5.
By the numerical method described above we have a
very accurate, quick, and flexible approach to calculate eigenvalues and wavefunctions of nonintegrable one-particle
systems.
II. THE HYDROGEN ATOM IN EXTERNAL FIELDS
The Hamiltonian of the hydrogen atom in external magnetic and electric fields in atomic units is given by Eq.
COMPUTERS IN PHYSICS, VOL. 11, NO. 6, NOV/DEC 1997

643

Table I. Energy of the 3p 08 state as a function of various external field strengths and relative orientations d between the electric and the magnetic
field axis. The numerical grid parameters used are as follows: number of nodal points for the discrete variable approximation for parallel fields
N55 and for nonparallel fields N 2 525; number of finite elements M 570 and integration limit for the radial coordinates in units of the Bohr
radius r max548a0. For the finite-element calculations we used Hermite interpolation polynomials with two nodes per element.
d 50
F50

10 V/m

5310 V/m

10 V/m

0
0.0001
0.0002
0.0003
0.0004
0.0005
0.0006
0.0007
0.0008
0.0009
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
0.009
0.01

20.111111
20.111110
20.111108
20.111105
20.111100
20.111093
20.111085
20.111076
20.111065
20.111053
20.111039
20.110825
20.110472
20.109986
20.109376
20.108649
20.107814
20.106881
20.105859
20.104757

20.111111
20.111110
20.111106
20.111100
20.111091
20.111081
20.111071
20.111060
20.111049
20.111037
20.111024
20.110819
20.110469
20.109985
20.109375
20.108648
20.107813
20.106881
20.105859
20.104757

20.111111
20.111110
20.111106
20.111100
20.111091
20.111080
20.111066
20.111050
20.111032
20.111012
20.110990
20.110745
20.110415
20.109948
20.109349
20.108629
20.107799
20.106870
20.105850
20.104749

20.111112
20.111110
20.111107
20.111100
20.111092
20.111080
20.111066
20.111050
20.111031
20.111010
20.110987
20.110673
20.110315
20.109862
20.109281
20.108576
20.107757
20.106835
20.105821
20.104725

~1.8a!, where the magnetic induction b 5 B/B 0 ,


B 0 54.73105 T and the magnetic field axis points to the z
direction, and
1 1
V ~ r, u , f ! 52 1 b 2 r 2 sin2 u 1Fr ~ cos u cos d
r 2
1sin u cos f sin d ! ,

~2.1!

with F5 F/F 0 the electric field strength measured in units


of F 0 55.1431011 V/m and d representing the angle between the two external fields. The energy is measured in
hartrees ~27.21 eV! and the lengths in units of the Bohr
radius (0.53310210 m!.
For vanishing electric field strength, F50, the spherical symmetry of the hydrogen atom is broken by the diamagnetic contribution of the strong magnetic field
b 2 r 2 sin2 u to the total potential. The quantum number m
and the z parity, p z , are conserved. A detailed discussion
of this system can be found in Ref. 7 and some calculations
with finite elements in Ref. 5. For parallel magnetic and
electric fields, d 50, the quantum number m is still conserved, but not the parity p z . Therefore, in this two cases
we have to solve a (2 % 1)-dimensional system. For perpendicular fields, d 5 p /2, the rotational symmetry with respect to the z axis is broken; hence, m is no longer conserved, which results in a three-dimensional system. In this
case the z parity, p z , is conserved, but for arbitrary relative
angles d ~unequal 0 or p /2) between the magnetic and
electric field axis no discrete symmetry is left and hence
neither m nor p z is a conserved quantity.
Some calculations for low-lying bound states of the
hydrogen atom in parallel and perpendicular strong mag644

d 5 p /4

COMPUTERS IN PHYSICS, VOL. 11, NO. 6, NOV/DEC 1997

d 5 p /2
6

5310 V/m
20.111111
20.111110
20.111106
20.111100
20.111091
20.111080
20.111067
20.111051
20.111033
20.111014
20.110994
20.110774
20.110440
20.109967
20.109362
20.108639
20.107806
20.106875
20.105854
20.104753

20.111111
20.111110
20.111108
20.111105
20.111100
20.111093
20.111085
20.111076
20.111065
20.111053
20.111039
20.110825
20.110472
20.109986
20.109376
20.108649
20.107814
20.106881
20.105859
20.104757

netic and electric fields using the discrete variable representation, but Newtons method for the radial part instead of
finite elements can be found in Ref. 8, which is a systematic study of spectroscopic data with parameters relevant
for the magnetic white dwarf stars in Ref. 9. As an example, we show in Table I some values for the energy of
the 3 p 80 state. Note that only in the field-free limit can this
state be labeled by the principal quantum number n, the
angular value l51, and m50. Due to the external fields
more and more states of different principal quantum numbers and angular momentum will mix, and if the electric
field and magnetic field axis become nonparallel, even the
magnetic quantum number m will no longer be conserved.
Therefore, we mark the states by an additional prime. ~Further values and oscillator strengths for transitions can be
obtained on request.!
In Sec. II A we will discuss some fascinating aspects
due to the nonintegrability of the system.
A. Avoided crossings
The single electron of the hydrogen atom in external fields
is, of course, a fermion and the Pauli exclusion principle
holds. Hence, there are no atomic states that share the same
quantum number, and therefore only those states can be
degenerate ~have the same energy! that can be labeled by a
quantum observable with different values. As mentioned
above, without an electric field the magnetic quantum number m and the parity p z are conserved. This system can be
reduced to a two-dimensional system because of the rotational symmetry around the magnetic field axis. In this rotational frame the energy E rot is given by
E rot5 ^ H2 b (L z 12s z ) & , where H is the full Hamiltonian,

Figure 1. On the top we show the energy as a function of the magnetic


field strength, in units of 4.73105 T, for states 3d 08 (left, top) and
3 p 08 . The primes indicate that this labeling would only be correct in the
field-free limit. At a certain magnetic field strength the two lines cross
each otherthe two states are degenerate. At the bottom we show the
corresponding probability of their presence in cylindrical coordinates.
The magnetic field B points in the z direction. The wavefunctions remain
undistorted in the region where the two lines cross each other.

L z 52i ( ] / ] f ), and s z is the z component of the spin.


Hence, the linear part in the magnetic induction b is the
energy shift given by the well-known Zeeman effect. ^ &
denotes the expectation values. For fixed magnetic quantum
number m and spin s z , the only quantum number left in
this nonintegrable two-degree-of-freedom system is the energy; the total angular momentum is not conserved and
hence l is no longer a quantum number for labeling eigenstates. States with different parity but all other quantum
numbers equal can be degenerate with respect to the energy
in the rotational frame E rot .
For the hydrogen atom in parallel electric and magnetic fields the symmetry with respect to reflection on the
xy plane is broken by the electric field and hence the z
parity p z is no longer conserved. Eigenstates degenerated
with respect to the energy in the rotational frame are forbidden and can be observed as avoided crossings. This effect is shown in Figs. 1 and 2. In Fig. 1 the electric field
strength is vanishing and the parity conserved, and hence
crossings of eigenstates with different parity are allowed. In
Fig. 2 some results for a nonvanishing electric field strength
are shown. The degeneracy is broken by a small avoided
crossing. One important aspect is the behavior of the eigenstates. In case of an allowed crossing ~Fig. 1!, the wavefunctions run without any distortion through this crossing.
In Fig. 2 the states are interacting with each other. Far away

Figure 2. The states are the same as in Fig. 1 and their corresponding
energies are plotted against the magnetic field strength, but now with an
additional parallel electric field of magnitude F5106 V/m. The parity is
no longer a good quantum number, and hence degeneracy is, due to the
Pauli principle, forbidden. Therefore, we observe an avoided crossing: the
states interact with each other and the wavefunctions are clearly distorted
close to the point of the forbidden crossing.

from the crossing the wavefunctions are similar to the ones


without an additional electric field. Close to the energy
value at which the avoided crossing occurs, both wave
functions are distorted and are mixed with each other.
By the simple hydrogen atom in external fields we
have a very graphic example showing the interaction of
neighboring eigenstates for nonintegrable quantum systems, an effect that holds in general, and which results also
in a completely different statistical behavior of the energy
eigenvalues ~see, for example, Ref. 10!.
III. CONCLUSION
Discrete variable techniques combined with finite elements
provide a useful numerical tool in investigating nonintegrable, low-dimensional quantum systems. Before using this method we calculated eigensolutions of the hydrogen and alkali atoms in external fields by Sturmian basic
type expansions7 on huge supercomputers. For example, for
the Cray-2 supercomputer we could use basis sizes of up to
325 000 ~a limit given by the available memory!. Even with
such a large basis we were unable for some field parameters
to obtain converged results such as wavelengths and oscillator strengths of dipole transitions. For those same field
parameters we have been able to obtain converged results
even on workstations, by using the discretization method
described above.
COMPUTERS IN PHYSICS, VOL. 11, NO. 6, NOV/DEC 1997

645

Recently a computational method based on powerseries expansions of the radial and angular coordinate u
~Ref. 11! was used to compute the bound states of the hydrogen atom in a strong magnetic field. By this method
highly accurate eigenvalues for the hydrogen atom can be
obtained. Depending on the magnetic field strength the
computational time for the ground state is of the order of a
second up to a minute.11 Note that by our method we compute not only single eigenvalues but always a series of eigenvalues and eigenfunctions. The necessary computational time for both together is of the same order, and the
accuracy of the eigenvalues is comparable. But the main
advantage of discretization methods like the one described
above, or for finite-element methods, is the possibility not
only to compute eigensolutions for the hydrogen atom in
magnetic fields, but also in additional electric or van der
Waals fields, and the computations are not restricted to the
hydrogen atom.12 We are free to choose any single-particle
potential, whereas series-expansion methods necessitate the
computation of an entirely new series expansion, which
should be taken into account in benchmark calculations.
Of course, even discretization methods like the one
described above are limited in their area of application. We
extended this technique to time-dependent quantum systems ~the details of which will be published elsewhere!, but
for extremely high-lying quantum states simple discretization techniques will fail. In this case using optimized discrete elements and adaptive techniques might be helpful,
but globally defined basis expansions will become more
and more efficient and each method will necessitate huge
computer facilities to obtain converged results.
Recently a HartreeFock study of two-electron
systems13 ~energy eigenvalues of some ground states and
excited states! in strong magnetic fields was presented.
Finite-element techniques in addition provide a useful
tool14 to solve two-electron systems. The discretization
method described above is limited to three-degree-offreedom systems, hence with respect to atomic physics to
effective one-electron systems. This includes the oneelectron excitations of alkali atoms, like the Rydberg states,
and only for those systems does the discrete variable technique combined with a finite-element method become more
powerful than the two methods mentioned above.
The discretization variable technique combined with
finite elements opened the door for the huge parameter
studies necessary to understand, for example, the observed
spectra of magnetic white dwarfs or to study quantumchaological aspects of Rydberg states in external fields. Not
only under scientific, but also under pedagogical, aspects
are discretizations techniques of interest. With the availability of powerful low-cost computers students have the

646

COMPUTERS IN PHYSICS, VOL. 11, NO. 6, NOV/DEC 1997

possibility of using these more advanced numerical techniques to study quantum systems, and hence, courses on
quantum mechanics and the necessary exercises are no
longer restricted to the perturbative and ground state pencil regime of quantum dynamical systems. ~The computer
codes for the method described above are written in Fortran
77 and are available on request.!
ACKNOWLEDGMENTS
The authors are grateful to Dr. I. Seipp for many helpful
discussions. They would like to thank Professor H. Ruder
for his encouragement. This work was supported by the
Deutsche Forschungsgemeinschaft ~DFG! and by the Deutsche Akademische Austauschdienst ~DAAD!.
REFERENCES
1. W. Miller, Lie Theory and Special Functions ~Academic, New York, 1968!.
2. E. Merzbacher, Quantum Mechanics, 2nd ed. ~Wiley,
New York, 1970!.
3. J. C. Light, I. P. Hamilton, and J. V. Lill, J. Chem.
Phys. 82, 1400 ~1985!.
4. G. Szego, Orthogonal Polynomials, College Publication rev. ed. ~American Mathematical Society, Providence, RI, 1959!.
5. L. R. Ram-Mohan, S. Saigal, D. Dossa, and J. Shertzer,
Comput. Phys. 4, 50 ~1990!.
6. T. Ericsson and A. Ruhe, Math. Comput. 35, 1251
~1980!.
7. H. Ruder, G. Wunner, H. Herold, and F. Geyer, Atoms
in Strong Magnetic Fields ~Springer, Berlin, 1994!.
8. V. S. Melezhik, Phys. Rev. A 48, 4528 ~1993!.
9. P. Fassbinder and W. Schweizer, Phys. Rev. A 53,
2135 ~1996!; Astron. Astrophys. 314, 700 ~1996!.
10. O. Bohigas, M.-J. Giannoni, and C. Schmit, Lecture
Notes in Physics, edited by S. Albeverio, G. Casati, D.
Merlini ~Springer, Berlin, 1986!, Vol. 262.
11. Yu. P. Kravchenko, M. A. Liberman, and B. Johannson, Phys. Rev. A 54, 287 ~1996!; Phys. Rev. Lett.
77, 619 ~1996!.
12. W. Schweizer, R. Gonzalez, and P. Fassbinder, preprint 1997.
13. M. D. Jones, G. Ortiz, and D. M. Ceperley, Phys. Rev.
A 54, 219 ~1996!.
14. M. Braun, W. Schweizer, and H. Herold, Phys. Rev. A
48, 1916 ~1993!; M. Braun, W. Schweizer, and H. Elster, in Atoms and Molecules in Strong External Fields,
edited by P. Schmelcher and W. Schweizer ~Plenum,
New York, 1997!, p. 25.

You might also like