Hydrogen atom in strong magnetic field revisited
ABSTRACT We derive in a straightforward way the spectrum of a hydrogen atom in a strong magnetic field.
arXiv:quant-ph/0309014v2 3 Dec 2003
Hydrogen atom in strong magnetic field revisited
I.B. Khriplovich and G.Yu. Ruban
Budker Institute of Nuclear Physics,
630090 Novosibirsk, Russia,
and Novosibirsk University
We derive in a straightforward way the spectrum of a hydrogen atom in a strong
The spectrum of hydrogen atom in a strong magnetic field was found long ago ,
and is presented now in textbooks (see, e.g.,  (§112, Problem 3)). The approach used
therein is as follows. At first, using the wave functions of electron in a magnetic field, one
constructs an effective potential for the motion along the field, and then the spectrum in
this potential is found. We present here a somewhat different solution of the problem,
with the results coinciding in fact with those of [1, 2]. We hope however that our approach,
physically straightforward and transparent, is of some interest. At least, the present note
may be considered as a sort of mini-review on the subject.
Our starting point is the obvious observation that in a sufficiently strong magnetic
field H (the exact criteria are discussed below) the motion of an atomic electron becomes
almost one-dimensional, along the magnetic field, in the Coulomb potential −e2/z.
The corresponding one-dimensional wave equation is
u = 0.(1)
We have introduced in it the usual dimensionless variable:
here a = ¯ h2/mee2is the Bohr radius, meis the electron mass, ν is the effective quantum
number, related to the electron energy as
Equation (1) coincides exactly with the radial equation for the s-wave in the three-
dimensional Coulomb potential −e2/r, and has therefore the common hydrogen spectrum
n = 1,2,3...,(3)
and the set of solutions
n(z) = exp(−z/2)z F(−n,2;z),ν = n,(4)
where F is the confluent hypergeometric function (here and below we are not interested in
the normalization factors). These solutions vanish at the origin and are trivially continued
to z < 0. Thus obtained solutions on the whole z axis are odd under z → −z (as reflected
by the superscripts “minus” in (3), (4)).
There is however an essential difference between the present problem and the s-wave
Coulomb one. In the last case (4) is the only solution. The reason is well-known. Naively
the radial wave equation for R(r)(= u(r)/r) has two independent solutions, which behave
for r → 0 as R ∼ const (u ∼ r) and R ∼ 1/r (u ∼ const), respectively. However, in
fact R ∼ 1/r is no solution at all for the homogeneous wave equation if the point r = 0
is included, since △(1/r) = −4πδ(r). As to our problem, equation (1) does not describe
really the vicinity of z = 0 since therein we have to consider seriously the magnetic field
itself. Therefore, there are no reasons to discard those solutions of (1) which tend to a
constant for small z (and of course decrease exponentially for z → ∞).
Such solutions are presented in a convenient form in  (Mathematical Appendices,
§ d, (d.17)). To our purpose they can be written for z > 0 as
ν(z) = exp(−z/2)1 − νz
lnz F(1 − ν,2;z)
Γ(1 − ν + k)[ψ(1 − ν + k) − ψ(k + 2) − ψ(k + 1)]
Γ(1 − ν)k!(k + 1)!
here ψ(α) denotes the logarithmic derivative of the gamma function: ψ(α) = Γ′(α)/Γ(α).
Being trivially continued to z < 0, thus obtained solutions on the whole z axis are even
under z → −z (as reflected by the superscript “plus” in (5) and in the corresponding
Under any reasonable regularization of the logarithmic singularity at z → 0, the even
solutions should have vanishing first derivative at the origin. In this way we obtain the
following equation for the eigenvalues of ν:
2ν+ ψ(1 − ν);(6)
field H. We are working in the logarithmic approximation, i.e. assume that
¯ hc/eH is the typical scale for the radius of electron orbits in the magnetic
λ = lna/aH≫ 1.(7)
This allows us to use a crude cut-off at aHfor the formal logarithmic divergence at z → 0,
as well as to simplify somewhat this equation.
The smallest root of equation (6) is
which gives the ground state energy
Other roots of equation (6) are
n= n +1
n = 1,2,3..., (10)
with the corresponding energies
Let us mention that the one-dimensional Coulomb problem was considered in [3 – 5]
with various regularizations of the singularity at z → 0, but without any relation to the
problem of the hydrogen atom in a strong magnetic field.
But let us come back to our problem. The resulting spectrum of the hydrogen atom
in a strong magnetic field looks as follows. Each Landau level in this field serves as an
upper limit to the sequence of discrete levels of the Coulomb problem in the z direction.
This discrete spectrum consists of a singlet ground state with the energy given by formula
(9), and close doublets of odd and even states of energies given by formulae (3) and (11).
There is also a continuous spectrum of the motion along z above each Landau level.
This picture is valid for sufficiently low Landau levels, as long as the radius of a
magnetic orbit is much less than the Bohr radius. Obviously, in a strong magnetic field
this description fails for large magnetic quantum numbers, i.e., in the semiclassical region.
Here we can estimate the orbit radius directly from the well-known spectrum of electron
in a magnetic field (see, e.g.,  (§112, Problem 1)):
E = ¯ heH
mec(N + 1/2),N = nρ+m + |m|
here nρis the radial quantum number in the xy plane, and m is the angular momentum
projection onto the z axis. Now the semiclassical estimate for the magnetic radius is
eHN = aH
Thus the present picture of levels holds as long as
λ = ln
≫ lnN. (13)
At last, let us consider the correspondence between the obtained system of levels in
a strong magnetic field and the hydrogen spectrum in a vanishing field. A beautiful
solution of this problem was given in  (and is quoted in ). We would like to present
the solution here as well (having in mind in particular that our note can be considered as
The crucial observation made in  is as follows. While changing the magnetic field
from vanishingly small to a very strong one, the number of nodal surfaces of a given
wave function remains the same. A hydrogen wave function (in zero magnetic field) with
quantum numbers n, l, m has nr= n − l − 1 nodal spheres and l − |m| nodal cones with
a z axis. With the increase of the magnetic field, the nodal spheres become ellipsoids of
rotation, more and more prolate, tending to cylinders in the limit of infinite field. The
correspondence between nrand nρ(nρbeing the radial quantum number in the xy plane
in a strong magnetic field) gets obvious from this picture:
nρ= nr. (14)
The evolution of the hydrogen nodal cones is less obvious. However, due both to equation
(14) and to the conservation of the total number of nodal surfaces, the number nz of
the nodes of a solution of equation (1) should coincide with the number of nodes of the
corresponding spherical function,
nz= l − |m|.
In other words, l − |m| nodal cones of a hydrogen wave function evolve into nzplanes of
constant z corresponding to the nodes of an eigenfunction of equation (14).
And at last, let us note that during the whole evolution of the magnetic field, m
Let us consider now, for instance, the ground state in the magnetic field, with N = 0
(see (12)). Obviously, it is degenerate, and its corresponding magnetic wave functions have
nρ= 0 and m = 0, −1, −2 .... Let us confine further to its lowest sublevel corresponding
to the ground state solution of equation (1), with nz = 0.
arguments, the hydrogen ancestors of those wave functions should have nr = 0 and
l = |m|. In other words, these ancestors are:
1s;2p, m = −1;
According to the above
3d, m = −2;and so on.
We are grateful to V.A. Novikov, V.V. Sokolov, M.I. Vysotsky, and A.V. Zolotaryuk
for useful discussions. We acknowledge the support by the Russian Foundation for Basic
Research through grant No. 03-02-17612.
 R.J. Elliott and R. Loudon, J. Phys. Phys. Chem. Solids. 15 (1960) 196
 L.D. Landau and E.M. Lifshitz, Quantum Mechanics (1989, Nauka, Moscow)
 R. Loudon, Amer. J. Phys. 27 (1959) 649
 M. Andrews, Amer. J. Phys. 34 (1966) 1194
 L.K. Haines and D.H. Roberts, Amer. J. Phys. 37 (1969) 1145
 W.H. Kleiner, Lincoln Lab. Progr. Rep. (Feb. 1960)