1:. r. V.

C TR0M I6N I I IC S l-I I I 'l' 339

than one-fortieth of a lifetime. This, together which a, =0.219. The frequency was then deter-
with the fact that the scattered radiation became mined for which the photoelectric current was
immeasurable at approximately the same tem- zero with the quartz cell at 300'C, at which
perature as did persistence of radiation in the temperature there was no measurable persistence
quartz cell, leads us to assume that reflection in the quartz cell. Using this Frequency in the
introduces no delay at any temperature. modified circuit, measurements of the photo-
As an additional check on the behavior of the current for different vapor pressures were made.
scattered radiation from the quartz cell in the The results are shown in Fig. 3 by curve D, the
region 150'C to 300'C, the alternating current observed points being given by crosses. The
method was employed in a somewhat different scale of ordinates of D, shown at the right of
manner. The ratio, S, of the photo-current for the figure, is of course arbitrary and was adjusted
negative and positive voltages on the grid of to coincide with C at 1.50'C. The two curves are
the photo-cell was approximately 0.28, and the coincident within the precision of measurement.
persistence of the radiation reflected from the This result agrees with the conclusions that
quartz-mercury vapor surface (persistence due reflection was instantaneous for all pressures,
only to B) resulted in more energy arriving at the and that the relation between the intensity of
photo-cell in the positive half-cycles than in the the absorbed and re-emitted (scattered) energy
negative. Hence the reflected radiation a = a, &-'. . and the time which gave the limiting R fcurve-
In this test the phase of the voltage applied to of Fig. 7 remained unchanged at temperatures
the photo-cell was reversed with respect to that above 150'C.
on the excitation cell, so that in the formula for The authors take this opportunity to thank
R, S was replaced by 1/S, i. e. , R = 1 —a, (1+ 1/S) Professor L. J. Hayner for her assistance in this
= 1 —4, 57c, . Hence R =0, for the frequency for investigation.

PH VS ICAL REVIEW VOLUM E 72, NUMBER 4 AUGUST 15, 1947

The Electromagnetic Shift of Energy Levels

H. A. BETHE
Cornell University, Ithaca, Rem York
(Received June 27, 1947)

Y very beautiful experiments, Lamb and explained by a nuclear interaction of reasonable

Retherford' have shown that the fine struc- magnitude, and Uehling' has investigated the
ture of the second quantum state of hydrogen effect of the "polarization of the vacuum" in the
does not agree with the prediction of the Dirac Dirac hole theory, and has found that this eff'ect
theory. The 2s level, which according to Dirac's also is much too small and has, in addition, the
theory should coincide with the 2p» level, is wrong sign.
actually higher than the latter by an amount of Schwinger and Weisskopf, and Oppenheimer
about 0.033 cm ' or 1000 -megacycles. This have suggested that a possible explanation might

troscopic measurements. "

discrepancy had long been suspected from spec-
However, so far no
satisfactory theoretical explanation has been
be the shift of energy levels by the interaction of
the electron with the radiation field. This shift
comes out infinite in all existing theories, and has
given. Kemble and Present, and Pasternack4 have therefore always been ignored. However, it is
shown that the shift of the 2s level cannot be possible to identify the most strongly (linearly)
divergent term in the level shift with an electro-
Phys. Rev. '72, 241 (1947).
~
W. V. Houston, Phys. Rev. 51, 446 (1937). magnetic nzcss effect which must exist for a bound
' R. C. Williams, Phys. Rev. 54, 558 (1938). as well as for a free electron. This effect should
' E. C. Kemble and R. D. Present, Phys. Rev, 44, 1031
(1932); S. Pasternack, Phys. Rev. 54, 1113 (1938). 5 E. A. Uehling, Phys. Rev. 48, 55 (1935).
properly be regarded as already included in the mentum, due to the fact that electr omagnetic
observed mass of the electron, and we must mass is added to the mass of the electron. This
therefore subtract from the theoretical expres- electromagnetic mass is already contained in the
sion, the corresponding expression for a free experimental electron mass; the contribution (3)
electron of the same average kinetic energy. The to the energy should therefore be disregarded.
result then diverges only logarithmically (instead For a bound electron, v' should be replaced by its
of linearly) in non-relativistic theory: Ac- expectation value, (v') . But the matrix ele-
cordingly, it may be expected that in the hole ments of v satisfy the sum rule
theory, in which the main term (self-energy of the
electron) diverges only logarithmically, the result Zn v-I'=(v')- (4)
will be convergent after subtraction of the free
electron expression. ' This would set an effective Therefore the relevant part of the self-energy
upper limit of the order of mc' to the frequencies becomes
28
of light which effectively contribute to the shift 8" = 8' —8"p —+
of the level of a bound electron. I have not 3m-hc"
carried out the relativistic calculations, but . I
Iv-I'(»:-»—.
P—
shall assume that such an effective relativistic
)
X '

dk (5)
limit exists. "p & —& +&
The ordinary radiation theory gives the follow-
ing result for the self-energy of an electron in a
This we shall consider as a true shift of the levels
quantum state m, due to its interaction with due to radiation interaction.
transverse electromagnetic waves: It is convenient to integrate (5) first over k.
Assuming X to be large compared with all energy
W= —(2e'/3mkc') differences E„— E& in the atom,

28 X
X P ''/(8„E+k), (1)—W' = —P v„„ '(8„„—
" E„) ln
0
kdk
n
Iv, „„~,
3.k, . I I

Iz. -Z. I
. (6)

where k =hco is the energy of the quantum and v (If 8„Z is negative, —
easily seen that the
it is
is the velocity of the electron which, in non- principal value of the integral must be taken, as
relativistic theory, is given by was done in (6).) Since we expect that relativity
theory will provide a natural cut-off for the
v=p/m = (k/im)V (2) frequency k, we shall assume that in (6)
Relativistically, v should be replaced by co. where X SPIC .
0. is the Dirac operator. Retardation has been
neglected and can actually be shown to make no (This does not imply the same limit in Eqs. (2)
substantial difference. The sum in (1) goes over and (3).) The argument in the logarithm in (6) is
all atomic states n, the integral over all quantum therefore very large; accordingly, it seems per-
energies k up to some maximum K to be discussed missible to consider the logarithm as constant
later. (independent of n) in first approximation.
For a free electron, v has only d iagonal We therefore should calculate
elements and (1) is replaced by

——(2e'/3m. kc')
Wo ~I kdkv'/k. This sum is well known; it is
~ =2 p-I'(&. —».)
This expression represents the change of the
kinetic energy of the electron for fixed mo- = —k' I ik *V V Vgdr
' It was first suggested by Schwinger and Weisskopf that
hole theory must be used to obtain convergence in this =~k~ t V~yy ~d, =2vk~, 2ZP ~(P)
problem.
 ALKALI HAI. I DE CRYS I AI. S

for a nuclear charge Z. For any electron with This is in excellent agreement with the observed
angular momentum l /0, the wave function value of 1000 megacycles.
vanishes at the nucleus; therefore, the sum A relativistic calculation to establish the limit
A =0. For example, for the 2p level the negative K is in progress. Even without exact knowledge
contribution 218 2g balances the positive contri- of E, however, the agreement is sufficiently good
butions from all other transitions. For a state to give confidence in the basic theory. This shows
with /=0, however. ,
(1) that the level shift due to interaction with radiation
is a real effect and is of finite magnitude,
(10) (2) that the effect of the infinite electromagnetic mass of
a point electron can be eliminated by proper identification
where n is the principal quantum number and a is of terms in the Dirac radiation theory,
the Bohr radius. (3) that an accurate experimental and theoretical in-
Inserting (10) and (9) into (6) and using vestigation of the level shift may establish relativistic effects
relations between atomic constants, we get for an (e.g. , Dirac hole theory). These effects will be of the order
of unity in comparison with the logarithm in Eq. (11).
5 state
W„,' = —te'q'
.8

3m
— Ry Z'
)
—
n'&kc)
~
ln
X
—,
(E„E„), „— If the present theory is correct, the level shift
(11) should increase
roughly as Z4 but not quite so
rapidly, because of the variation of (E„— E )A„ in
where Ry is the ionization energy of the ground the logarithm. For example, for He+, the shift of
state of hydrogen. The shift for the 2p state is the 2s level should be about 13 times its value for
negligible; the logarithm in (11) is replaced by a hydrogen, giving 0.43 cm ', and that of the 3.s
value of about —0.04. The average excitation level about 0.1. 3 cm '. For the x-ray levels LI and
energy (E„— E )A, for the 2s state of hydrogen has LII, this effect should be superposed upon the
been calculated numerically' and found to be effect of screening which it partly compensates.
17.8 Ry, an amazingly high value. Using this An accurate theoretical calculation of the screen-
figure and K=mc', the logarithm has the value ing is being undertaken to establish this point.
7.63, and we find This paper grew out of extensive discussions at
the Theoretical Physics Conference on Shelter
W„,'=136 InLX/(E„—E,„)] Island, June 2 to 4, 1947. The author wishes to
1040 megacycles. (12) express his appreciation to the National Academy
7I am indebted to Dr. Stehn and Miss Steward for the of Science which sponsored this stimulating
numerical calculations. conference.

PH YS ICAL REVI EW VOLUME 72, NUMBER 4 AUGUST 15, 1947

Spectral Location of the Absorption Due to Color Centers in Alkali Halide Crystals
HENRY F. IVEY
Research Jaboratory, Westf'nghouse Electric Corporation, Bloomffeld, New Jersey
(Received June 2, 1947)

SEITZ' has recently

marizing the properties
published a paper sum-
of the absorption
this information. The terminology
for the bands
is that used by Seitz. In this respect it is of
bands arising from electrons trapped in the interest to point out that the band designated by
lattice of alkali halide crystals. The present Seitz as the M-band (after Molnar) has been
writer is not aware of the existence of a compila- called the 0-band (after Ottmer) by Kalabuchov. '
tion of data on the spectral location of these In addition to the F-, R1-, R2-, and M-bands
various bands. Table I was prepared to collect occurring in the pure alkali halides, Table [ also
' F. Seitz, Rev. Mod. Phys. 18, 384 (1946&, ' N. Kalabuchov, J. Phys. t&. S, S.R. 9, 41 (1.945),