Volume 62A, number 4
PHYSICS LETTERS
22 August 1977
THE ZEEMAN EFFECT REVISITED
J. AVRON~and I. HERBST2
Joseph Henry Laboratories of Physics, Princeton University, Princeton, New Jersey,08540, USA
B. SIMON3
Belfer Graduate School of Science, Yeshiva University, New York, New York 10033, USA
Received 7 June 1977
We announce three new rigorous results for the quantum mechanical hydrogen atom in constant magnetic field:
(i) Borel summability of the small field perturbation series, (ii) detailed large field asymptotics, and (iii) non-degeneracy
of the ground state ti~and a proof that it has L~t~o
= 0 for all values of the field.
The weak field Zeeman effect [1] in simple atoms
was one of the earliest problems studied [2] in quantum mechanics. More recently, Ruderman [3] and
then others [4] discussed the analogous problem in
super-strong magnetic fields of the type encountered
in neutron stars. It is perhaps surprising that any problems remain open for such a well studied theory but
there are some unresolved theoretical questions of
interest: (i) The Rayleigh-Schrodinger perturbation
coefficients for the energy levels almost surely [5] diverge as n! Do they nevertheless determine the answer
in some way? (ii) Is there a systematic large B expansion for the ground state energy beyond the cB and
d ln2B terms of ref. [4]? (iii) There are central potentials [6] where the ground state fails to bern = 0 for
B in a suitable interval away from zero and is therefore degenerate for at least one value of B by continuity [7]. Is the attractive Coulomb potential one of
these or not?
We wish to describe here solutions of these three
questions; full details of our results and methods will
appearelsewhere [8].Someoftheresultsextendto
more general atoms and we have studied the corrections due to finite nuclear mass [9] but we will state
our results for the simple model:
H(B) =
2
(—iv
2
—
—
r’
(1)
A)
On leave from NRCN, Beer-Sheva, Israel. Research supported
by NSF Grant MPS 74-22844 and MPS 75-22514.
Research supported by NSF Grant MPS 74-22844.
On leave from Departments of Mathematics and Physics,
Princeton U. Research supported by NSF Grant MPS 75-11864.
214
A
=
-4(r X B);
B = (0,0, B)
(2)
Theorem 1. Let E~(O)be any negative eigenvalue of
the Hydrogen Hamiltonian H(0). Then there is an eigenvalue [10] En(B) of H(B) for B small which is the Borel
sum [11,12] of the Rayleigh-Schrödinger perturbation
coefficients for En [10].
Theorem 2. The ground state [13] energy Ern(B),
of H(B) restricted to the subspace with fixed azimuthal
angular momentum m is asymptotic for large B and
fixed m to
Ern(B) = B(ImI
—
m
+
1)
r
‘ln~lnB~~12
ln B
—[~
—
ln(ln B)
+
+
O~ln B
where
rn—i
q
rn
=q
1
0
—~
i=O
________
~—
—
i!(m
q =11n2+4C+ ~ ~÷
0
1
m~l
r e_X —1 dx~00S796,
x
Cis Euler’s constant and ~
1(x) is the exponential integral.Theorem 3. The ground state of H(B) for any B is
non-degenerate and has L Z = 0 [14].
The complete proofs of these results are too lengthy
to give here but we can say something about the methods.
To prove Theorem 1 [12] one needs to prove stability of
the eigenvalues [15] for complex values of B in some sec-
Volume 62A, number 4
PHYSICS LETTERS
tor. This turns out to be somewhat more subtle than
the corresponding stability for the anharmonic oscillator [12]. By scaling (see eq. (3) below) and standard
perturbation theoretic arguments, one shows that the
domain of analyticity contains the cut plane intersected
with a disc. The last element of the proof is then!
bound on the series expansion in a suitable region of
the complex B plane. Here we exploit a technique of
Combes and Thomas [16] developed originally to prove
the exponential falloff of bound state wave functions.
There is one very interesting aspect of our study of
22 August 1977
References
9Gauss so that con[lj The natural unit for the field is .-‘~i0
ventional laboratory fields are “small”.
(21 W. Heisenberg, P. Jordan, Zeit. für Phys. 37 (1926) 263.
[3] R. Cohen, L. Lodenquai, M. Ruderman, Phys. Rev. Lett.
25 (1970) 467;
B.B. Kadomtsev, Soy. Phys. JETP 31(1970)945;
M. Ruderman, Phys. Rev. Lett. 27 (1971) 1306.
[4] R.O. Muller, R.P. Rau, L. Spruch, Phys. Rev. Lett. 26
(1971) 1136, Phys. Rev. A 11(1975)789, 1865, Astrop.
J. 207 (1976) 671.
[51 Then!
comes from the following: The leading term in
Rayleigh-Schr~idingertheory is ~ = (t2~,V(S
0 V)’~no),
Theorem I we should mention:
is the ground
1ln(i,1i If ~1i~
for 11(0),
t~ perturbation
0, e tH(B)series
~p0)o~~(B)
a~.(B)is
astate
function
withthen
a formal
= ~ a~,(t)B’~
obeying a
0(t)I ~<A(t)B(ty~ (n/2)! des1
pite
the
fact
that
the
perturbation
series Z an(00)B’
for —E
0(B) 1im~~0.
t~(B)undoubtedly has an(°°)I
n! This example [17] is relevant to the Lipatçv
theory [18] of the asymptotics of the perturbation
series for anharmonic oscillators and 0 field theories
where similar ~ 00 and n oo limits are interchanged
with abandon; we believe this interchange is correct in
that case but it is clearly more subtle than previously
believed,
To prove Theorem 2, one introduces a coupling constant X in front of the r~ term in (1) and notes that
E(B, X), the ground state of H(B, X) obeys:
—~
E(B, 1) = B E(l, B
where S0 = (Ho —E0)~[1 — (~io, )H0]. c~o~— e~ 1ate~
infinity, S0 has no falloff in x and V ‘-~ ixi so a -~ f xI’
[61
M.ofO’Carroll,
to be
published.
[7] R.
TheLavine
usualonand
proof
the
nature
of the ground
depends
the fact
thatnodeless
the Green’s
function
kernelstate
of
(H —El’, (E < E0) is strictly positive. This falls for
Hamiltonians in magnetic fields where this Green’s function is no longer even real.
[81 magnetic
J. Avron, I.
Herbst
and B. Simon, Schr~idingeroperators in
fields,
in preparation.
[9] The reduction of the center of mass is actually subtle; for
a free particle in a magnetic field, the momentum perpen-
-~
(3)
2x.
Eq.
(3)
reduces
the
large
B
since H(B, 1) and BH(1, B 1/2) are unitarily equivalent
under the scaling x BU
behavior to a small coupling problem for H(1, 0) ~r
Because the magnetic field in H(1, 0) discretizes the
dicular to the field is not conserved but the gauge invariant quantity C p + eA (with e the charge of the particle)
related to the position of the center of rotation, is conserved
and*plays
a role
similar to the
momentum.
[Cx,Cy]
0. For
a multiparticle
system
with pairBut
interactions C = E,C1 is conserved. C~and C~will commute
1/2)
if and onlykind
if the
total charge
is zero. In this case a conventional
of reduction
is possible.
-~
—
~,
spectrum in two dimensions, this is essentially a small
coupling problem in one dimension where systematic
expansions have been recently developed [19].
Theorem 3 depends on certain monotonicity results
obtained by developing a Wiener path integral for the
Hamiltonian reduced to a fixed m subspace, discretizing
the corresponding “time” and using correlation inequalities [20] for the corresponding Ising-like
system [21].
1 is distinguished
The Coulomb
potential
V(r)
—rthe conditions
from
the potentials
of ref.
[6]= by
V’ ~a0, V” ~ 0. The critical input is a proof that under
certain circumstances the ground state wave function
of a quantum mechanical particle must collapse to
wards the origin as the potential becomes more attractive.
suits
hold for
states of are
maximal
and
minimal m
[101 Of
thethe
eigenvalues
highly
degenerate.
Ourwhich
arecourse,
non-degenerate
on the fixed
m subspace.
[11] That is, if ~ a~B°
is the formal perturbation series, then
g(x) ~ a
0x°/n! converges for x small, has an analytic
continuation to a neighborhood of the positive reals and
forB smallE(B) = fo°°
g(xB) e~~
[12] Borel summability methods for Rayleigh-Schrödinger
series go back to S. Graffi, V. Grecchi and B. Simon, Phys.
Lett. 32B (1970) 631, who prove this summabiity for the
anharmonic oscillator. It has since been extensively developed in a variety of situations, including certain field
theories; see e.g. J.P. Eckmann, J. Magnen and R. Sén~or,
Commun.
Math.
Phys.
39 are
(1975)
251.
+
asB
-~ these
[13] All0(1)
states
but
states
of orderB(Irn~—rn
+ 1)
[141This is true in any gauge for which L
2 is a constant of the
motion.
[15] This is a technical condition which says that the eigenvalue remains isolated and its multiplicity constant for
small complex B. An example of a non-stable perturbation, even for real coupling, is the Stark prob1em~see e.g.
J. Avron and I. Herbst, Commun. Math. Phys. 52 (1977)
239.
215
Volume 62A, number 4
PHYSICS LETTERS
[16] J.M. Combes, L. Thomas, Commun. Math. Phys. 34
(1973) 250.
[17] A more artificial but even stronger example is to take
an(t) = (n/2)! + j~yne~’dy, where an(t) (n/2)! as
n —~ °° for any fixed t <= but an(oo) -~ n! as n [181 L.N. Lipatov, Pisma JETP 24 (1976) 179;
E. Brezin, J.C. LeGuillou and J. Zinn-Justin, Phys. Rev.,
in press.
[19] B. Simon, Ann. Phys. 97 (1976) 279;
R. Blankenbecler, M.L. Goldberger and B. Simon, Ann.
Phys., to appear. Their method must be extended to
treat Coulomb tails.
[20] There is an extensive literature on correlation inequalities beginning with R. Griffiths, J. Math. Phys. 8 (1967)
478. We actually must prove a new kind of FKG inequality for a very special class of multicomponent
models.
[21] The idea of using correlation inequalities for discretized
path integrals is due to F. Guerra, L. Rosen and B. Simon,
Ann. Math 101 (1975) 111, and has been extensively de-
veloped.
216
22 August 1977