Ebook Condon The Theory of Atomic Spectra

KANKAK
26 $92
<!I1
Y.
DDD1
MO PURUC
UttHAMV
DHD070
fi
Condon,
The theory of atomic
CY
J*
*3
Condon
|l!) au
The theory of ai;oiiilc
w
THE
THEORY OF ATOMIC SPECTRA
THE
THEORY OF ATOMIC SPECTRA
by
E. U. CONDON, PH.D.
WAYMAN CROW FEOFB8SOE OP PHYSIOS, WASHINGTON
8T LOUIS, MISSOUEI
and
G.H.SHORTLEY,PH.D.
DIKECtOK, WASHINGTON OFEKATION8
BOOZ, ALLEN AFFUKD EBSBABCHi INC,
CAMBRIDGE
AT
THE UNIVERSITY PEESS

1959
XTNIVEBSITY
PUBLISHED BY
TUB SYNDICS OF THE GAMBBIDGK UNIVERSITY PRKSS
Bentley House, 200 Euston Road, London, N.W. I
American, Branch: 32 East 57th 8troot Now York 22, N,Y.
printed
Reprinted with
1936
corrections
1051
1953
1057
1050
JPirtt
Reprinted
First printed in Great Britain at the

University
Reprinted by
offtet-litho
by Bradford db
/Vaw, Oambrulge
London, W.CLl
Dic,kem>
To
HENRY NOEBIS RUSSELL
;,.;;
,r
.N
630896*
a
-
cm
mm,
if;
CONTENTS
Preface
...........
CHAPTER I
INTRODUCTION
PAGE
xiii
....
CHAPTER II
THE QUANTUM MECHANICAL METHOD
12
12
SECTION
12
Symbolic algebra of states and observable
22
Representations of states and observables
15
32
Continuous eigenvalues and the Sehrodinger representation
20
42
The
24
52
The laws of quantum mechanics
82
statistical interpretation
Schrodinger
........
equation
25
26
Matrix mechanics
27
Perturbation theory
30
R6sum6 of the perturbation theory

Remarks on the perturbation theory
34
1C
II
Perturbation caused by a single state
37
12
The
43
35
analysis of non-commuting vectors
CHAPTER III
ANGULAR MOMENTUM
1
43
5
73
83
93
10
...
Definition of angular momentum

Allowed values of angular momentum
The matrices of angular momentum
Orbital angular momentum
Spin angular momentum
Vector addition of angular
45
46
48
50
54
momenta
56
The matrix of (Ji+J,)

Matrix of Tin the jm scheme.
Selection rule
Dependence of the matrix of T on m

The matrices of /! and /2 where /i +/2=J
?
45
on j
...
....
58
59
61
64
CONTENTS
viii
CHAPTER
III (continued)
...
SECTION
P which commutes with J
PAGJD
67
1 13
Matrix of a vector
12 3
Matrix of P-Q
70
13 s
Sum rules
71
14 s
Transformation amplitudes for vector addition
...
CHAPTER IV
THE THEORY OF RADIATION
73
79
14
Transition probabilities
79
24
Classical electromagnetic theory
83
34
Expansion of the retarded potential
84
4*
The correspondence
87
54
The
dipole-radiation field
90
64
The quadrupoie-radiation field
93
74
Spectral lines in natural excitation
97
84
Induced emission and absorption
94
Dispersion theory. Scattering.
10 4
principle for emission
Natural shape of absorption
100
Raman effect.
103
109
lines
CHAPTER V
ONE-ELECTRON SPECTRA
5
...
112
Central-force problem
112
25
Radial functions for hydrogen
114
35
The
118
relativity correction
Spin-orbit interaction
120
Sketch of the relativistic theory
125
6s
Intensities in
131
75
Experimental results for hydrogenic spectra
hydrogen
General structure of the alkali spectra
Intensities in alkali spectra
105
Zeeman effect
.137
141
.147
149
CONTENTS
CHAPTER
26
46
56
6
76
VI
THE CENTRAL-FIELD APPROXIMATION
SECTION
16
ix
The Hamiltonian for many-electron atoms
PAGE
158
.158
160
Equivalence degeneracy
The dynamical equivalence of the

The PauH exclusion principle
.
electrons
.
Conventions concerning quantum numbers. Closed shells

Matrix components for S^/(i)
86
Matrix components for S^gf(i,^)

Matrix components of electrostatic interaction
96
.162
.166
.
168
169
171
174
Specialization for closed shells
177
1C 6
One
183
II 6
Odd and even states
electron outside closed shells
185
CHAPTER
VII
THE RUSSELL-SAUNDERS CASE; ENERGY LEVELS

7
27
The JD/S-coupling scheme

Term energies
37
The Land6
188
191
interval rule
193
Absolute term intervals
195
57
Formulas and experimental comparison
67
Terms
77
The
197
in the nlx configurations
207
terms of helium
210
triplet
CHAPTER VIII
THE RUSSELL-SAUNDERS CASE: EIGENFUNCTIONS
18
187
Vector coupling in antisymmetric states
28
Genealogical characterization of i/S-coupling terms
38
Land6
48
Calculation of eigenfunctions
58
Calculation
intervals for terms of coupled groups
of
by
eigenfunctions
direct diagonalization
using
213
.213
216
.219
.
220
angular-momentum
226
operators
6s
Calculation of eigenfunctions from vector-coupling formulas
78
Separation of the
D s of d*
228
233
CONTENTS
S:
CHAPTER
THE RUSSELL-SAUNDERS
IX
STRENGTHS
CASE: LINE
PAGE
236
SECTION
P
2
39
49
59
Configuration selection rules
2 10
237
244
Multiplet strengths obtained from spectroscopic stability
249
252
Quadrupole niultiplets
CHAPTER X
jj COUPLING
1 10
...
Line strengths in Russell-Saunders multiplets

Multiplet strengths in a transition array
236
....
The jj-coupling scheme and the spin-orbit interaction

The addition of a weak electrostatic interaction
257
257
259
3 10
Eigenfunctions
262
410
Line strengths
264
CHAPTER XI
INTERMEDIATE COUPLING
1 11
2n
...
Matrix of spin-orbit interaction for

configurations consisting
of coupled groups
Matrix of spin-orbit interaction obtained from the
266
266
eigen-
functions
3 11
411
5 11
270
from LS to jj coupling
Line strengths in intermediate coupling
The forbidden lines of astrophysical interest
Illustrations of the transition
.277
.
CHAPTER XII
TRANSFORMATIONS IN THE THEORY
OP COMPLEX SPECTRA
...
1 12
212
3 12
412
5 12
Configurations containing almost closed shells

The transformation to LS coupling
The transformation to jj coupling

The transformation between zero-order states
The transformation nlm8m *=? nljm
The transformation jjJM^SLJM
l
6 12
271
282
284
284
285
287
287
290
291
CONTENTS
XI
CHAPTER XIII
CONFIGURATIONS CONTAINING ALMOST
CLOSED SHELLS. X-RAYS
SECTION
1 13
The electrostatic energy in LS coupling
2 13
3
13
The
spin-orbit interaction
Pure almost-closed-shell configurations
413
The rare-gas
5 13
The
configurations p* s
6 13
The
configuration p* pin the rare gases
...
PA GE
295
295
.....
299
300
301
spectra
304
and d*$
306
7 13
The configuration^ ^ in the rare gases
312
8 13
Line strengths
316
9 13
X-ray spectra
316
Line strengths in X-ray spectra
322
10 13
II
13
X-ray
323
satellites
CHAPTER XIV
CENTRAL FIELDS
1 14
The
periodic system
2 14
The
statistical
method of Fermi-Thomas
....
....
....
3 M The Wentzel-Brillouin-Kramers approximation
414
Numerical integration of the radial equation
514
Normal state
6 14
Excited levels in helium
7 14
Normal
states of first-row
Hartree
s self-consistent fields
14
9 14
10 14
327
327
335
339
344
345
of helium
348
atoms
351
354
Survey of consistent-field results
358
Self-consistent fields for
362
oxygen
CHAPTER XV
CONFIGURATION INTERACTION
365
366
1 15
2
Interaction of sd and # in
2 15
Perturbed
3 15
Auto-ionization
369
415
Many-electron jumps
375
5 15
of doublet intensities
Spin-orbit perturbation
magnesium
367
series
376
CONTENTS
Xll
CHAPTER XVI
SEOTION
1 16
The normal Zeeman
2 16
The
effect
weak-field case Eussell-Saunders terms

:
316
Weak fields
416
Intensities in the
5 16
The Paschen-Back
effect
6 16
The Paschen-Back
effect
7 16
...
PAGE
378
....
378
THE ZEEMAN EFFECT
Quadrupole
general case
380
384
Zeeman pattern: weak fields
386
388
:
illustrative
examples
390
395
lines
CHAPTER XVII
THE STARK EFFECT.
397
1 17
Hydrogen
398
2 17
Stark effect at the series limit
404
3 17
General theory for non-hydrogenic atoms
4 17
Helium
413
5 17
Alkali metals
415
....
CHAPTER XVIII
THE NUCLEUS IN ATOMIC SPECTRA
409
418
1 18
Effect of finite mass
418
2 18
Local nuclear
420
fields
3 18
Nuclear spin in one-electron spectra
418
The hyperfine
5 18
Zeeman effect of hyperfine
421
structure of two-electron spectra
structure
424
426
APPENDIX
UNIVERSAL CONSTANTS AND NATURAL
ATOMIC UNITS
....
428
List of Principal Tables
434
Index of Subjects
435
Index of Names
439
PREFACE
In this monograph we have undertaken a survey of the present status of the
problem of interpreting the line spectra due to atoms. This interpretation
seems to us to be in a fairly closed and highly satisfactory state. All known
features of atomic spectra are now at least semi- quantitatively explained
in terms of the quantum-mechanical treatment of the nuclear-atom model.
This does not mean that the period of fruitful research in atomic spectra
an end. Fundamental questions are still outstanding in regard to the

theory of the many-electron problem and also in regard to the
of
the
interaction of radiation and matter. In addition every reader
theory
will see many places where more experimental information and better or
more detailed theoretical calculations are desirable. It is our hope that the
book will be useful in stimulating progress along these lines. With this end
in view we have aimed to give explicitly an example of the use of each of
the general types of calculation involved and an adequate survey of the
literature of the more specialized calculations. The literature is covered
approximately to the summer of 1934 although a few later references
have been incorporated.
is
at
relativistic
There exists confusion in the original literature about two matters which
we have made every effort
to clear
up
in this
book
recognized that a matrix is not

place
the relative phases of the
unless
transformation
theory
fully useful in the
states to which the components refer are in some way specified, since two
In the
it
first
has been too
little
matrices cannot be added or multiplied unless these phases are the same in
both. The phase choice is arbitrary, just like the troublesome sign conven
tions in geometrical optics, but one choice does have to be made and adhered
to throughout a given set of calculations. To facilitate the use of the formulas
and tables of the book in other calculations, we have attempted in every case
make explicit the specification of the phase choice employed.

In the second place there occurs, particularly in the theoretical literature,
a great diversity of spectroscopic terminology. We have attempted to
adhere as closely as possible to the original meanings of the nouns which
denote energy levels and spectral lines, and find that this gives a nomen
to
convenient and unambiguous. Briefly, our usage is the

a radiative transition between
following a component (of a line) results from
two states of an atom; a line results from the totality of transitions between
clature that
is
two
levels
ijx
Bussell-Saunders coupling, a multiplet
is
the totality of lines
PREFACE
XIV
connecting the levels of two terms, and a supermultiplet is the totality of

nmltiplets connecting the terms of two polyads a transition array is the
totality of lines connecting the levels of two configurations.
;
We have defined in
find to give a
7 4 a quantity, called the strength of a line,
which we
more convenient
theoretical specification of the radiation

than
of
the
transition probabilities.
either
Einstein
intensity
hope that
this new usage will find favour among spectroscopists.
We
We
take pleasure in making acknowledgment here of the valued help of

whom we have discussed various parts of the theory, and
friends with
many
who have
criticized portions of the manuscript. In particular we are in

debted to Prof. BL P. Robertson of Princeton for most of the elegant new
treatment of spherical harmonics given in 4 3 to Dr F. Seitz of Princeton
;
and to
to
Mr
W.
Uflf ord
of Allegheny College for special calculations and

B. Napier for assistance in preparing the figures. Completion of the
Prof. C.
work was
greatly facilitated by generous arrangements by the senior

at Princeton to relieve him of most of his teaching duties
for a term. The junior author is appreciative of courtesies shown by the
author
s colleagues
physics faculties of the University of Minnesota, where he spent the summer

of 1933, and of the Massachusetts Institute of Technology, where he
spent
the year 1933-1934 as a National Research Fellow. The greater part of his
work on the book was done while he was a Fellow in Princeton University
in the years 1931-1933.
Finally we wish to record our great enthusiasm for the beautiful typo
once more
graphical work of the Cambridge University Press as
exemplified
in the following pages.

E. U. C.
G. H. S.
March 1935
We are naturally gratified at the reception that our work has had and regret
very much that other duties have prevented our giving it the thorough
revision which it needs. The present
printing is essentially a reprint of the
1935 edition except that a number of errors and
misprints have been
corrected.
E. U. C.
G. H. S.
May
1950
CHAPTER I
INTRODUCTION
"And so the true Cause of the Length of that Image was detected to be no other,
than that Light is not similar or Homogenial, but consists of Difform Rays, some
of which are more Refrangible than others." NEWTON.
Spectroscopy is that branch of physics which is the direct outgrowth of a

experiment of Newton s which, led him to the conclusion which we
classic
have placed at the head of this chapter. Newton s experimental arrangement

1
which is taken from Voltaire s Elemens de la Philois shown in Fig. I
sophie de Newton (Amsterdam, 1738, p. 116). The beam of sunlight enters
a hole in the window shutters and traverses the prism, falling on the
screen P. The image is not round like the hole in the shutter, but long in
the direction perpendicular to the axis of the prism and coloured.
,
Fig.
I 1.
Newton s discovery
of dispersion.
This arrangement is the proto-type of the modern spectroscope. As we are

here interested in the theory of spectra we shall not concern ourselves with,
the technique of spectroscopy as it has developed from Newton s time to the
Our principal interest will be the information concerning the nature

atom which is obtained from a study of the characteristic radiations
emitted by monatomic vapours.
present.
of the
After the discovery of sharp dark lines in the solar spectrum and sharp
emission lines in spectra of flames, arcs, ard sparks, the physicists of the
nineteenth century seized upon spectroscopy as a valuable tool for qualita

tive chemical analysis. At that stage the experimental problem consisted in
correlating the various lines and bands seen in the spectroscope with the
chemical nature of the emitting substance. This task in itself was by no

means simple, for the spectroscope is extraordinarily sensitive to small
impurities and it was difficult to deal with sources pure enough to make
certain the correlation of the observed lines to the substances in the source.
INTRODUCTION
Another thing which made for difficulty is the fact that the spectra do not
depend simply on the chemical elements present but on their state of
chemical combination, which in turn is usually altered by the conditions
which render the substance luminescent. To this period also belongs the
beginning of the great task of setting up accurate standards of wave-length.
It early became clear that the observed spectra are of three general types.
having no line structure in the spectroscopes of greatest

resolving power. These are emitted by incandescent solids, but also under
some circumstances by molecules and even by single atoms. Banded, having
a special form of line structure in which close groups of many lines occur
Continuous, that
is,
so densely packed that in smaller instruments they appear continuous.

These are characteristic of the spectra of molecules and arise from the
many possible changes in the rotational state of the molecule during the
radiation process. Line, where the lines are well separated and generally
show no obvious simple arrangement although
grouped into small related groups of a few lines.
individual atoms.
in
cases they are

spectra are due to
many
Such
The line spectrum due to a single chemical element in the form of a monatomic vapour shows still another complication. It was early learned that
quite different spectra are obtained from the same element according to the
energetic violence with which it is excited to luminescence. In the electric
spark more energy is put into the emitting atoms than in the electric arc and
generally quite different lines result from the same element for these two
modes of excitation. These differences are now known to

degrees of ionization of the same element.
arise
from
different
Evidently the next step is that of trying to understand the structural

nature of an atom which enables it to emit its characteristic radiations. For
the purposes of the classical kinetic theory of gases it is not necessary to
assume anything more about atoms than that they behave something like
As a consequence the range of phenomena which that

theory embraces is not in a position to tell us more about the structure of
hard
elastic balls.
atoms than an estimate of their size. In the spectrum of one element we are
given a vast amount of data which is measurable with great precision.
Evidently it is somehow determined by the structure of the atom, so
spectroscopy stood out clearly in the minds of physicists as an important
means
for studying that structure.
In the latter part of the nineteenth century Maxwell developed his electro
magnetic wave theory which received experimental confirmation in the
experiments of Hertz and Oliver Lodge. Because the velocity of light agreed
with the velocity of electric waves, the theory of electric waves was early
applied as a theory of light. The wave theory of light, hitherto developed as
rNTKODUCTION
a theory of elastic vibrations in the universal medium, the ether, was re

written in terms of the theory of electromagnetic waves. A little later the
electron was discovered through researches on cathode rays and the chemical
theory of ions was developed in connection with the electrolytic dissociation
hypothesis. The view became current that atoms are structures built out of
electrons and positive ions.
branch of physics called the electron theory
came into being whose programme called
for the explanation of

the properties of matter in terms of this picture with the aid of the laws of
the electromagnetic field.
of matter
During this period empirical regularities in line spectra were being found.
The best known of these was Bahner
simple formula (1885) for the wave
lengths of the visible lines of the hydrogen spectrum.

Although experimentalists still prefer to express their measurements in
terms of wave-lengths, Hartley showed (1883) that there are regularities

in the spacing of related doublet or triplet lines which are more simply
expressed in terms of the reciprocal of the wave-length, that is, the wavenumber or number of waves in unit length. This discovery is of the greatest
theoretical importance. There is no logical reason to-day for dealing with
wave-lengths at all, and they are seldom mentioned in this book. But the
custom of thinking in terms of them in the laboratory is probably too firmly
entrenched to be shaken off for a long time. After the work of Balmer comes
the important researches of Rydberg and of Kayser and Runge, who dis
atomic spectra, chiefly those of

alkali and alkaline-earth metals, can be organized into series obeying
formulas similar to the formula of Balmer.
covered that
many spectral lines in various
These empirical discoveries of spectral regularities reach their culmination

in the clear establishment of the Ritz combination principle. This came in
1908 after two decades of important work on the study of spectral series.
According to this result each atom may be characterized by a set of numbers
called terms, dimensionally like wave-numbers, such that the actual wavenumbers of the spectral lines are given by differences between these terms.
Ritz thought that lines were associated with all possible differences between
these terms, and this is in accord with modern theoretical views, except that
the lines associated with some differences are millions of times weaker than
others so that practically there are important selection rules needed to tell
which
differences give strong lines.
The principle received striking confirmation in the same year through

Paschen s discovery of an infra-red series in hydrogen. The wave-numbers
of the Balmer series are represented by the formula
INTRODUCTION
where R 109677 cm-1 an empirical coefficient called the Rydberg con

stant. In Ritz s language this means that this particular set of lines arises as
2
2
The
differences of the terms cru =R/7i and the particular term cr 2 =R/2
,
combination principle suggests the existence of lines given more generally by
r=R
For
m = 3 and = 4,
71
5, 6, ...,
1\
/ 1
\m 2
the lines
I.
nr]
fall
in the infra-red. Paschen found
found in the ultra-violet three lines
them at the predicted places* Lyman

assimilated
=
corresponding to ra= 1 and n 2, 3, 4. The principle was quickly
also
as
an important
rule in the analysis of spectra. It holds in all cases,
even
a simple formula as
is
individual terms cannot be represented by
for
the case with hydrogen. Its applicability is of great generality, holding
when the
molecular as well as atomic spectra.

The first decade of the twentieth century was important as showing,
on the
through the work of Planck on black-body radiation and Einstein
laws of interaction of
photo-electric effect, that there is much more to the
matter and radiation than is given by the nineteenth century electromagnetic

The electron
theory. These developments mark the birth of quantum theory.
atomic
led
to
some
had
concerning
assumptions
simple
theory programme
structure and had had some notable successes, particularly in Lorentz s cal
culation of the effect of a magnetic field on the spectral lines, as observed by
Zeeman. Spectral lines were associated with the electromagnetic radiation
coming from motion of the electrons in an atom, generally regarded simply
as harmonic oscillations about an equilibrium position. Great difficulty
attached to the interpretation of the enormous number of spectral lines
without the introduction of unreasonable complications in the model. On
the experimental side a most important step was the recognition, through
experiments on scattering of alpha particles made by Rutherford, that the
which is now
positive electricity in an atom is confined to a small particle
called the nucleus of the atom. Its linear dimensions are not greater than
about 10~ 4 those of the whole atom.
The stage is now set for the great theoretical developments made by Bohr
from 1913 onward. Rutherford s experiments had given a general picture
of a nuclear atom a positively charged massive nucleus surrounded by
the negatively charged and much less massive electrons. The theoretical
developments had given imperfect and unclear indications of the need of
fundamental changes in the electron theory for the process of emission and
absorption of radiation. Empirical spectroscopy was organized by means
of the Ritz combination principle and the extensive study of spectral series.
In 1913 Bohr s first work on atomic structure gave a theory of the spectrum
of hydrogen which involved several important advances.
INTRODUCTION
Most general was the idea of stationary states and the interpretation of
the Ritz combination principle. It is postulated that the possible states of
atoms and molecules are restricted to certain values of the total energy.
These values are determined by the structure of the atom or molecule and
may be continuous in some ranges, as in the classical theory, or may be
Then the postulate is made that the

emission or absorption of radiation is connected with a process in which.
the atom passes from one energy level to another. This is rendered precise by
the statement that the frequency of the radiation emitted is given by the
restricted to a set of discrete values.
where h is Planck ^s constant, v the frequency of the radiation, E the energy

of the atom before the radiative process and E 2 its energy afterwards. If
their numerical values

2
l the frequencies come out formally negative
are the frequencies of light which can be absorbed by atoms in the state of
This expresses the frequencies as differences of numbers cha
l
energy
>
racteristic of the
levels,
atom and
E, and the terms,
<r,
establishes a coordination
between the energy
through the relation
The minus sign arises since the conventional way of measuring term values
was by counting them as positive when measured from, the series limit.
There is no reason to adhere to this convention, so that we shall always write
1
and regard a simply as a measure of the energy in the auxiliary unit,
its
The hypothesis suggests
own means of experimental verification. If
atoms are excited to radiate by single electron impacts in which the electrons
have known kinetic energy E then the only spectral lines appearing should
be those for which the energy of the initial state E^ is less than E, (This
assumes that the zero of energy is the energy of the lowest state and that
before impact the atoms are all in this lowest state.) This is the idea under
lying the experiments of Franck and Hertz and many others on critical
potentials. Such experiments have fully confirmed the energy level inter
pretation of the spectroscopic terms and have been a valuable tool in experi
cm.""
mental work.
controlled electron impact it is possible to bring out a

bit as the kinetic energy of the impacting electrons is
By
spectrum bit by
gradually increased, thereby simplifying the task of determining the energy

which are associated with the production of the different lines.
levels
The other part
Bohr s early work was the development of a special

for the hydrogen atom and the study of rules for the
determination of the allowed energy levels. The model for this simplest atom
consisted of an electron and proton describing orbits about their centre of
dynamical model
of
INTRODUCTION
mass according to classical mechanics under their mutual attraction as given
by the Coulomb inverse-square law. The allowed circular orbits were
determined simply by the requirement (an additional postulate of the
quantum theory) that the angular momentum of the system be an integral
multiple of H = hjSir. This served to give an energy
^=for the circular orbit
whose angular momentum
vakesare
n
is nti,
whence the term
he
Not only is the variation as ?r~ in accord with the scheme of terms as given
by the Lyman-Balmer-Paschen series in hydrogen, but the numerical
coefficient comes out correctly, so by this means the empirical
Eydberg
2
constant
R was for the first time related to universal
constants.
Naturally this definite accomplishment stimulated other work and in the

next few years gseat advances were made in interpretation of finer details of
the hydrogen spectrum due to relativistic effects (Sommerfeld) and the effect
of an electric field on the hydrogen spectrum (Epstein, Schwarzschild). Also
it gave rise to much Important work in
extending the model and the quantum
principle to other more complicated atomic and molecular structures. These
studies were eminently successful in a semi-quantitative way and gave a great
impetus to the experimental study and analysis of atomic spectra. The theory
called for a study of the model by means of classical mechanics. The so-
had to be sought out and of these the

allowed ones determined by a rule of quantization which was an outgrowth
of Bohr s requirement on the angular momentum for the circular orbits of
hydrogen. We shall not trace in detail the work along these lines: for this
the reader is referred to Sommerfeld s Atombau und SpeJctrallinien and to
called multiply periodic motions
Van Vleck
s Quantum Principles and Line Spectra.

The theory was quite incomplete, however, in regard to the details of the
interaction of the atom with the electromagnetic field. It gave no definite
basis for the calculation of the relative or absolute intensities of the spectral
give satisfactory results when attempts were made to
lines. It also failed to
calculate the energy levels of
atoms containing more than one
Numerous attempts were made to
electron.
calculate the energy of the lowest state of
helium but without securing agreement with the experimental result.

Evidently Bohr s principles had to be regarded as provisional indications
of the direction in which a more satisfactory theory was to be
sought. The
unsatisfactoriness of the theory came more and more into the
foreground
in the early part of the 1920-30 decade after almost ten
years in which
INTRODUCTION"
physicists were busy
making such progress as was possible with the original

Bohr theory.
This was the period just preceding the discovery of the formalism of
quantum mechanics, the discovery which has been so extraordinarily
fruitful for all parts of atomic physics in the past eight years. In this period
much was done in the study of atomic spectra and the formulation of the
results in the language of the Bohr theory. The most important theoretical
development was Bohr s correspondence principle. This emphasizes that the
laws of atomic physics must be of such a character that they agree with the
classical mechanics and electromagnetic theory in the limit of large quantum
numbers. This principle was able to make much more definite some of the
results of the previous theory in the way of special calculations
in particular
gave an approximate method of calculating the relative intensity of

spectral lines. Its successes in such special applications were sufficient to
create confidence in the principle. What was much more important than the
it
special applications was the use of the principle as a broad general guide in
the attempts to formulate a more complete set of laws for atomic physics*
The important problem before theoretical physics was thus the develop
ment of a rational system of quantum mechanics. The earlier work is usually

quantum theory: it consisted of a few quantum postulates
the
classical kinematics and dynamics. By quantum mechanics
on
to
patched
we mean the much more unified theory of atomic physics which we owe to
de Broglie, Heisenberg, Schrodinger, Dirac, and others. We shall give just a
sketch of the origin of the new theory. It developed rapidly from 1925
onwards, at first along two quite different lines which were quickly brought
referred to as
into close relationship.
De Broglie built his work on an analogy, due to Hamilton, between the

laws of mechanics and the laws of geometrical optics. He was led to the
formal conclusion that a wave motion is associated with the motion of a
particle such that if the
direction, the associated
to
p by the equation
momentum
wave
is
of the particle is p in magnitude and

propagated with a wave-number a related
p = JIG.
This was suggested by the equation E = hv together with the fact that in
relativity theory E is the time-like component of a four-vector of which p is
the space-like part, while v is the time-like component of a four-vector of
which is the space-like part. This suggestion was made in 1924. Something
over a year later it was taken up by Schrodinger and developed in his famous
series
of papers on wave mechanics. Very roughly the idea
is
that just as
geometrical optics is adequate for phenomena in which all apertures are

large compared with the wave-length of light otherwise diffraction effects
INTRODUCTION
so also classical mechanics is adequate where the linear

dimensions involved are large compared to h/p otherwise effects akin to
diffraction are important: these are the special effects which we find in the
are important
region of the quantum phenomena of atomic physics.

At the same time Heisenberg took a decisive step in the formulation of
matrix mechanics as a definite realization of the correspondence principle

programme. The initial step was to regard the atom as characterized by &n
ensemble of quantities of the type Pnm e2flTiv ^ in which Pnm gives the
amplitude of a classical harmonic oscillator whose intensity of radiation and
type of polarization is the same as that of the light of frequency vnm from the
i
actual atom. This double array of quantities, for all values of n and m, was
regarded by Heisenberg as a single mathematical entity. By treating it and
other quantities related to the atom s structure by the mathematical rules

of matrix algebra it was found that a definite formulation of the laws of
quantum mechanics could be given in accordance with the correspondence

In a few months the mathematical equivalence of Schrodinger s
wave mechanics and Heisenberg s matrix mechanics was established. The
principle.
work of Jordan and of Dirac led
to a formulation of a single mathematical
system of quantum mechanics. This is the formulation in terms of which

current work of importance in quantum physics is expressed to-day.
all
During this period the study of atomic spectra was being actively pushed
on. After the study of series spectra of the alkali-likp metals in terms of the
Bohr theory the next important steps were the empirical discovery by Land6
of the laws of the Zeeman effect and the discovery and extensive study of
related groups of lines called multiplets in complex spectra. The modern
study of multiplets was begun by Catalan. The multiplet structure and the
problems of the anomalous Zeeman effect called for an essential generaliza
tion of the electron-orbit model which was supplied in 1925 by Uhlenbeck
and Goudsmit by postulating an intrinsic magnetic moment and angular
momentum for the electron. This spin hypothesis quickly

many difficult points, so that it at once gained acceptance.
cleared
up
In the pre-quantum-mechanical period the general method of working on

a detailed question in the theory of atomic spectra was to calculate from an
assumed model by means of classical mechanics, and then to try to alter the
formulas so obtained in such a way that the change was negligible for large
quantum numbers but was of such a nature that it brought about agreement
with experiment for small quantum numbers. It is really remarkable how
much
of the modern, theory of line spectra was developed in this way.
Important contributions were made by Pauli, Heisenberg, Hund, and

Russell. There was developed a vector-coupling model for complex atoms
in which the quantization of the angular momenta of the individual
INTRODUCTION
electron orbits and of their vector resultant played a dominant role. To this
period also belongs the discovery of the important exclusion principle of
Pauli according to which no two electrons in an atom may have the same set
of quantum numbers. This
first found a rational place in the theory with the

advent of quantum mechanics. As an empirical principle, however, it was
of the greatest importance, especially through the work of Hund in predict

ing the general structure of complex spectra and extending the theory of the
periodic system of the elements as
begun by Bohr.
Thus it happened that by ingenious use of the correspondence principle a
great deal of the modern theory of atomic spectra was worked out without
the aid of quantum mechanics. But that does not mean that the new
mechanics is without importance for our understanding of line spectra, since
the results obtained were not part of a closed structure of definite physical
principles,
but were obtained in semi -empirical ways from consideration of
a formulation of the theory that was only true in the limit of large quantum
numbers. Moreover, not all the problems of interest could be handled in
this way, so there were detailed calculations on which to test the quantum-
mechanical method as well as the task of securing from definite calculations

by the new methods those results which had been cleverly guessed with the
aid of the correspondence principle. This application of quantum mechanics
to the atomic model has been the programme of research in the theory of
atomic spectra from 1926 onwards. In spite of the mathematical com

plexities, it became clear in the next five or six years that the quantum
mechanics of Heisenberg, Schrodinger and Dirac when applied systematic
study of the nuclear model of the atom is adequate to give an.
accurate and complete unification of the great amount of empirical data
accumulated through analyses of atomic spectra. Naturally in so short a
time it was not possible to make precise and detailed calculations of the
ally to the
system of energy levels of all atoms, but enough has been done to give rise
to a general conviction that the theory is quite adequate for the interpreta
tion of atomic spectra.
Much remains to be done in the way of precise and detailed calculations
and it may well be that when these are made it will be found that such an
estimate of the power and scope of our present theoretical knowledge is overseems at present, there may well
optimistic. Complete as the general picture
be lurking somewhere important residual effects, like the advance of the
which will necessitate essential
perihelion of Mercury in celestial mechanics,
alterations in the theory. At present the theory exists in a somewhat closed
and complete form, so that it is possible to give a unified deductive treatment

of atomic spectra in terms of the quantum-mechanical theory of the nuclear
atom. That is the programme of this book. Future development in this field
INTRODUCTION
10
may be merely a story of the filling in of details without essential change in

the present views. Or it may be that essential difficulties will appear,
requiring radical revision of the present theory. In either case the time seems
ripe for taking stock of the rapid developments of the past decade in order
to prepare ourselves for the completion of the study of line spectra of atoms
in the experimental and theoretical studies of the future.
At present
spectra and
there are several systematic accounts of the subject of line

their theoretical interpretation. The most important are
SOMMEBFELD, Atombau und Spektrallinien, F. Viewcg, 1931;

HITND, Liniensgektren und periodisches System der Mlemente, J. Springer, 1927;
PATJMNG and GOUDSMIT, The Structure of Line Spectra, McGraw-Hill, 1930;
RTTABK and UBEY, Atoms, Mokcules and Quanta, McGraw-Hill, 1930;
GIBBS, Line Spectra of the Elements, Rev. Mod. Phys. 4, 278 (1932);
WHITE, Introduction to Atomic Spectra, McGraw-Hill, 1934.
All except the last of these are written from the
standpoint of the
correspondence principle. In this book we make no attempt at following
the historical order. We have confined our attention to the historical
developments to the brief sketch of
this introductory chapter.
Likewise
we make no attempt at a complete account of the formulation of the

principles of quantum mechanics, as there are now in existence several
works on this subject. Most important among these are
BIBAC, Quantum Mechanics, Oxford University Press, second edition, 1935;
VON NEUMANN, Mathematische Qrundlagen der Qitantmmachanik, J. Springer, 1932;
EEENKEL, Wave Mechanics, Advanced General Theory, Oxford, 1934.
More elementary, and hence more suited to a first approach to the subject, are
SOMMEBFELD (translation by BEOSE), Wave Mechanics, MetBuen, 1930;
BOBN and JOBDAN, JSlementa-re Quantenmechanik, J. Springer, 1930;
ERENKEL, Wave Mechanics, Elementary Theory, Oxford, 1932;
CONDON and MOBSE, Quantum Mechanics, McGraw-Hill, 1929;
MOTT, An Outline of Wave Mechanics, Cambridge University Press, 1930.
In order to make this book more useful for independent reading, however,
we haTe opened the detailed discussion with a brief account, in the next
chapter, of the principles of quantum mechanics. This account is intended
simply as a review of principles used throughout the rest of the work and as
a repository of the theory in convenient form for reference. The
succeeding
two chapters
are devoted to the presentation of special results from the

general theory which are used throughout the book; the detailed develop
ment of the theory of atomic spectra begins with Chapter v.

We wish finally to make a few remarks concerning the place of the theory
of groups in the study of the quantum mechanics of atomic
spectra. The
reader will have heard that this mathematical
discipline is of great import
ance for the subject. We manage to get
along without it. When Dirac
visited Princeton in 1928 he gave a seminar
report on his paper showing the
connection of exchange energy with the spin variables of the electron. In
INTRODUCTION
the discussion following the report,
Weyl
11
protested that Dirac
had said that
he would derive the results without the use of group theory, but, as Weyl
said, all of Dirac s arguments were really applications of group theory.
Dirac replied,
"I
said I
would obtain the
results
without previous
know
ledge of group theory."

That incident serves to illustrate our attitude on this point. When a
physicist is desirous of learning of new theoretical developments in his
subject, one of the greatest barriers
is
that it generally involves
new mathe
matical techniques with which he is apt to be unfamiliar. Relativity theory

brought the necessity of learning tensor calculus and Riemannian geometry.
forces him to a more careful study of boundary value
of new
and
matrix
problems
algebra. Hence if we can minimize the amount
a real
him
we
do
a
new
field
he
must
learn
in
mathematics
order to penetrate
service. WeyFs protest to Dirac is certainly also applicable to this book. But
so is Dirac s answer. Many things which are done here could be done more
Quantum mechanics
the theory of groups were part of the ordinary mathematical

equipment of physicists. But as it is not, it seems like putting unnecessary
simply
if
to treat the subject by a method which requires this

hand the pure mathematician studying a new
the
other
equipment.
branch of physics is likely to take most delight in the fact that the theory
hitherto rather
exemplifies parts of pure mathematics which have been
devoid of physical applications. To him our plan is not as satisfactory as one
obstacles in the
way
On
which shows how the theory of the structure of the atom
is
related to the
abstract theory of groups.

This does not mean that
we underestimate the value of group theory for

nor that we feel that physicists should omit the study of
atomic physics
that branch of mathematics
new theory.
many new things
tool in the
them so
now that it has been shown to be an important

It is simply that the new developments bring with
to be learned that it seems inadvisable to
add
burden to the load.

For those who wish to regard the theory of atomic spectra from the
this additional
standpoint of the theory of groups there
are three books available at present:
and Quantum Mechanic*, Methuen, 1931;

(translation by ROBERTSON), Group Theory
der Atomspektren,
WIGNEB, Qruppentheorie und ihre Anwendung auf die Quantenmechanik
WEYL
F. Vieweg, 1931,
VAN DEB WAEEBEN, Die

1932.
T
,
T
in der QuantenmechamJc, J. Springer,
gruppenthe&retische Metlioden
CHAPTER II
The formulation
our purpose
is
of the laws of quantum mechanics which
that due to Dirac,
and discussed at length in
shall review in brief such of this formulation as
references to Dirac, where a fuller account
followed
by a
is
we
may
most suited to
We
his book.*
shall need, giving
be found. This
page
be
will
discussion of matrix mechanics, perturbation theory,
and
related subjects.
Symbolic algebra of states and observables.

In this section the principles of the theory will be set down in terms of the
properties of certain abstract symbols in a manner corresponding to the
purely symbolic treatment of vector analysis which is independent of any
1.
coordinate system.
The state of a system (Dirac, p. II) is described by a quantity called
which is analogous to a unit vector in a space of a great many (in general an
infinite number of) dimensions (p. 1S).| 4* has as
many components as the
<]*
system has independent states. In any given representation the components

are ordinary (in general, complex) numbers. The sum of two
ij/s is
another fy whose components are the sums of the components of the two
of
4>
tp s.
The product
of a
^ by an ordinary number c is a ty whose components
are c times those of the original

Two states are not considered to be distinct unless the tys which describe them
are linearly independent.
*\>.
Since the components of

are allowed to be complex, and only real
numbers may occur in the interpretation of the theory, we introduce $, the
symbolic conjugate imaginary of 4.J if is not a vector in the same space as
t|>
but
a vector in the dual space. Its components are the

ordinary com
plex conjugates of the corresponding components of |*. Since t|j and $ are
[*,
is
vectors in different spaces, there
is no place in the
algebra for the addition of
and a $, The distinction between and $ is more fundamental than that
between ordinary complex conjugates; there is no sense in which we can
split 4* into a real and an imaginary part.
*J>
<|*
*
DIRAC, Quantum Mechanics (page references are to the second edition). For a
mathematically
more rigorous formulation see VON NEUMANN, Mathematische
Grundlagm der (faantenmechanik.
f
We
shall in this chapter use bold-faced
type for symbolic quantities, reserving ordinary type
and operators in the Schrodinger scheme and for

ordinary numbers. Thus confusion
between the symbolic
and Schredinger s ^ function will be avoided.
in place of $, but we prefer the latter notation because it is more
J Dirac uses
symmetrical
and more easily translated into the Schrodinger
in
representation.
for functions
t|>
cj>
chapters
we
Furthermore,
shall not restrict ourselves to the letter

<|*
different letters for different states.
subsequent
to represent a state, but shall at times use
SYMBOLIC ALGEBRA OF STATES AITD OBSERVABLES
I2
We now suppose that any 4* &&d any
13
have a product which is defined as

the analogue of the scalar product of two vectors in vector analysis and is
therefore an ordinary number, in general complex (p. 22). This product will
where the sub
by convention always be written with the $ first, e.g.
(j>
4v*K>
scripts denote different states of a system. It follows, then, that 4*?4*s an(^
$s4 r are complex conjugate numbers, and that tp^r is real and positive;
unless otherwise stated we shall take all i]/s to be normalized. Hence, for
any
and
s,
In observing a system experimentally we build an apparatus on a macro

scopic scale which interacts with the system by a certain set of operations,
feature of classical
resulting in a scale- or pointer-reading. The essential
the laws of
physics has been that we have expected to be able to formulate
physics in terms of functional relationships between the pointer-readings
and exact
given by various sets of observing apparatus. All of physics
mechanics
lines
hitherto.
such
has
Quantum
natural science
proceeded along
does not do this. Any set of experimental apparatus and operations does not
which
appear in the theory simply as the source of certain pointer-readings
bear a direct functional relationship to other sets of pointer-readings.
Instead it appears as a quantity of a more complicated sort about to be
described. Thus we are dealing not merely with a new set of laws but with an
new mathematical canvas on which to represent these laws. In this
entirely
respect the
quantum mechanics is a much more far-reaching departure from
than was the theory of relativity.

In an experiment we are generally concerned with determining, directly
or indirectly, the particular number which expresses the value of a dynamical
at a particular time.
variable, e.g. the position or momentum of an electron,
will be repre
Any such dynamical variable will be called an observable, and
in the
An
observable
a
linear
a
24).
this
in
sented
(p.
operator
theory by
it
into
and
on
for
rule
is
a
converting
any
mathematical theory
acting
second
the
of
a
tensor
to
rank, or
it is analogous
In this
another
classical physics

no
will
be written as
aat|>
61>
are ordinary numbers.

(ai
+ a 2 )4 = a1^ + a 2
observables
is
ai|>
4>,
defined
We
where
Linearity of a means that we have

and s are any two states and a and b
.
oij>
a<|
+ bfy s =
is another i|*.
by a^, where
a on $.* The conjugate
of
provision for the operation
<|>
r}>
two observables by
The product, a^, of two
define the sum, 0^ + *%, of
where
fy is arbitrary.
by foo^-aôyl*);
* Dirac does
provide for the operation of
a on $,
in general
writing
=
<fa
*&**&.
at4>;
Since
see equation (2).
14
is
oil>
another

is an ordinary number. The
$ r wfy 8 = $ r
8
|j,
(wl>
(Hermitian-) conjugate to
a,
denoted by a
is
",
defined
observable
by the equation
(2)
where the two states r and 5 are arbitrary (p. 29). The relation between a and
a is the same as that between ordinary complex
ia
"
there
is
no
-a *
1
f
adding a and a
rule against
An
conjugates; (ia)*=
observable a is said to be real
if
>
An observable a is said to be purely imaginary if a = a;

such an observable may be written as ip, where p is a real observable.* It can
for all states r and s.
be shown from
(2)
"
that
hence if c^ and a 2 are two real observables the commutator

(oc^ a 2 a 1 ) is a
observable
if
is
also
real
this commutator
purely imaginary
(p. 29),
oc^
vanishes.
We now make
the important physical postulate (p. 30) that a state for
where a is an ordinary number, is characterized

by the fact that the observ
able a has the value a. That is, that a measurement of the observable a with
the system in the state
a real observable a the
multiplying
by (3),
The
so a
(5)
ij>
will certainly give for the result the
number a will be necessarily real,
by $ r to get
$ r k = a$ rfy = a [from
r
1 )]
number a. For
shown by
as can be
.
Here
r ai|; r
is real
is real.
possible values of
observation of
a,
an observable a, i.e. the possible results of an

numbers a for which the equation in ^,
are the ordinary
a4 = a
(6)
4>,
has solutions
(p. 30). These are called the allowed, proper, characteristic, or

of
eigenvalues
a, and may form a discrete set or a continuous set of
numbers.f
We shall follow Dirac in calling the
which satisfy this equation the
eigen-tys and the states which they denote the eigenstates of the observable a.
We shall speak of an eigen-^ which satisfies this equation for a given a as
<J/s
belonging to the eigenvalue
oc
and denote
at|>(oc
The complex imaginary to
i}>(a
it
oc
4>(a
by
4>(a
That
is,
(y a )
).
we denote by $(a
).
).
This satisfies the relation
* The
operators corresponding to real and purely imaginary observables are said to be Hermitian
and ant^Henmtian
respectively, because of the Hermitian and anti-Herinitian character of their
matrices ( 7). Similarly, at is called the Hermitian
conjugate of oc because its matrix is the
Hermitian conjugate of the matrix of a.

f The whole set of eigenvalues is often called the spectrum of a.
I2
SYMBOLIC ALGEBRA OF STATES AND OBSEKVABLES
There
may
be several, say d a linearly independent
tp s
,,
15
belonging to the
same eigenvalue a In this case the value a will be said to be d a -fold

degenerate, and the different 4K a ) s ^^ require another index which takes
on the values 1, 2, ,.., d a for their complete specification.
Now if one makes observations of a on a system in a state i{* which is not
an eigenstate of a, one does not always obtain the same value, but observes
the various proper values of a with certain probabilities. We now make a
second physical postulate, viz. that the average of the values of a obtained
in this way is the number tpa4 which is characteristic of the state 4* (p. 43).
4 is here assumed normalized in the sense of (1).
For a real observable this average value is seen from (3) to be real for any
state
Moreover, by writing a = ar 4- ia$ where ar and c^ are real observ.
4>*
ables, it is seen that a necessary condition for the reality of
$*]* f
r aU
*s
4>
the vanishing of ai Since the results of physical observation are real for all
states
we conclude that all physically measurable quantities are real
.
4>>
observables.
PROBLEMS
that the allowed values of any function /(a) expressible as a power series in a are
This can be generalized to other functions (p. 38). Hence show that the first physical postulate
above is a special case of the second.
1.
/(a
Show
).
2.
Prove that two
commute
real observables
if
and only
if their
product
is real.
that the observable SaS 1 (where S^S-SS"1 ^ 1) has the same allowed values as a,
and find the relation between the eigen-^ s of a and those of 5aS~x
3.
Show
Representations of states and observables.

We shall first show that the eigen-ij/s of a system belonging t0 two different
2.
allowed values of a real observable are (unitary-) orthogonal in the sense that
$(a
To prove
this,
)4(a")
unless a
= a*.
(1)
observe that
I 2 7a)
(from
).
Therefore (a
-a")$(oc
)<]>(a
= 0;
hence the theorem
(from
1 23
and
I 2 7b)
(p. 33).*
We now assume that the
eigen-tj/s of any real observablef form a com

can make a Fourier expansion of an arbitrary fy
system in which we
This amounts to assuming the whole of a kind of generalized SturmLiouville theory at one step; hence much needs to be filled in here by a study
s can be so expanded.
of exactly what classes of
plete
(p. 34).
*I>
if a is
arises, this may be replaced by oc
a real observable.
these considerations hold only for real observables; the eigen-ij/s of a general
f Note that
observable do not form an orthogonal set. Although we shall have occasion to employ non-real
observables as an aid in calculation, we shall never have occasion to determine their characteristic
values or functions, since these lack physical significance in quantum mechanics.
Note that whenever the combination $(oc )a
^>(yf)
22
16
Consider
value
oc
is
first
da
the case in which a has a purely discrete spectrum. If the

d
degenerate, we shall choose the a linearly independent
,
>~fold
a normalized and mutually orthogonal This we may do

states belonging to oc
by a Schmidt process* since any linear combination of
for r = 1, ..., d^.
a/r
states
these
Denote
)
will also belong to a
by 4K
d^
states belonging to
We then have,
according to these requirements, the relations

r
$(
tâ*
r")
= 8(a
, a")
S(r
(2)
r")
= S a ar
is a function of the discrete set of points

satisfied. Here
)
(a
and unity if
which is zero if a ^
representing the allowed values of a
that any state ty may
oc =
(the Kronecker delta). The assumption is, then,
write the expansion
shall
r
We
states
these
of
in
terms
be expanded
).
a"
a"
4>(oc
coefficient as (a r
),
placing
it after
= StKa
4*
the eigenstate:
)(aV|
(3)
).
a r
$ which is being expanded;
to characterize the
The blank space is reserved

for example we write
aV
which results from the action of the observ
This state can be expanded in terms of the complete set
Consider the state
on
For
*]>(aV).
able p
vl*(oc"r").
P4>(
"
r")
this expansion
we adopt the
pt|>(a"r")
notation
= St|;(aVO(aV
/
|
8|a
(5)
V").
a r
Before proceeding further with the algebra of this expansion theory we shall
significant choice and characterization of the
show how we may obtain a
different states belonging to the degenerate level a
observable p which commutes with
a.
We
Let us choose an
observe that
a"r"
from
(5),
while
If a(3 = pa, the coefficients of
must be
equal; hence
(oc
on the right of these two equations

r ) = 0, from which we draw the
|j8|a
i}>(aV)
//
-a")(oc
important conclusion that

(oc
V ||aV)=:0 unless
oc
a".
(pa-ap-0)
(6)
Let us now set ourselves the problem of finding the eigenstates of p. Denote
let (4) represent its expansion
we must then deter
such a state by
j3 ) and
mine the coefficients (oc r |/3 ). Expanding the allowed values equation
4>
* See for
example, COTJRANT
and HILBEBT, Hetlioden
der MathematiscJien Physik, p. 34.
REPRESENTATIONS OF STATES AND OBSERVABLES
22
17
we obtain
S
a* r*
a r
*l(*V)(aV|/J|oe*r
S (a
whence
TMs
set of
reduces
r |j8|a r
)("r
r|/3
)(
|j3 )
= j8
= Sj8XaV)(V|j8
a
(a r
),
for all
|/5 )
a>
(7)
linear equations for the coefficients (

use the special property (6) of p to the simpler set
homogeneous
when we
Sta
ljSla
OtaVljS
for all
J^jSXotVljSO
,r
(8)
r"
Here all the coefficients in a given equation refer to the same eigenvalue of a.
Hence we may consider independently the d a equations corresponding to
>
the eigenvalue a These d a equations determine the direction* of that part

of
which lies in the c a,-dimensional subspace characterized by a Let
us find the solutions which lie wholly in this subspace, i.e. let us set
ft)
/
*I>(/O
r"
(cc"
= ot In order that
equal to zero unless
solution of this type, the determinant
<x"
we may
find
a non-vanishing
Setting this determinant equal to zero gives an equation of

which are consistent
degree in /T whose roots are the d^ eigenvalues
dtf
with the conditions assumed. To each of these roots belongs an eigenstate of
must vanish.f
th
p lying wholly in the subspace characterized by a To a multiple (d a^-fold)

root @ will belong d^p linearly independent eigenstates lying in this subit is clear that
space. Since all the eigenstates of p are given by (8), and since
have
found
a complete
we
of
we have found a complete set of solutions
(8),
are simultaneously eigenstates of a and of p. If
set of states
j5 ) which
.
4>(a
a degeneracy in these states, we may choose a third observable

y (independent of a and p in the sense that y is not a function of a and p)
which commutes with both a and p. We may by a process similar to the
there
is still
above find a complete set of eigenstates of y which are simultaneously eigen

to introduce in
states of a and p, i.e. of the form
/J y )- We continue
dependent commuting observables until we find no degeneracy in a simul
taneous eigenstate of aE of them. The number of such observables is the
7
ty(<y!
quantum-mechanical analogue of the classical number of degrees of freedom.

Thus a calcium atom with 20 electrons and a fixed nucleus will be found to
*
set of linear homogeneous equations determines only the ratios of the unknowns, ie. the
direction of a vector but not its magnitude. But because of the normalization condition, an eigen
state is determined by its direction except for an arbitraiy phase factor e^ f
f See, e.g., BOOKER, Higher Algebra, p. 47.
18
require 80 quantum
of observables, say
numbers
22
for the complete description of its state.
A set
n erms of whose simultaneous eigenstates

Yfc *
Ya
we can describe the state of a system completely, is said to be a complete set.
We shall denote this set of observables by F for short, and shall write the
as 4(F ).*
general state characterized by th&quantum numbers y[ y^, .., yk
fc
Yi>
>
>
>
We can choose a complete set of observables in various ways since any in

8 of the y s w^ &so form a
dependent commuting functions 8 1? 8 2
,
...,
7j
complete set, which we shall denote by A.

Let us now return to a consideration of the algebra of representations in
terms of the eigenstates of such complete sets of observables. A general state
4*
has the expansion
The expansion
[cf. (3)]
coefficients are
given explicitly by multiplying with
OH
$(T
)=$(r*)4*.
For the complex imaginary equation to (10)
we
(ii)
shall use the notation
$=S(
inftr),
r
where
The
set of eigenstates is
of
)
(P
unit vector
|
If the
),
we
)-(P|
(12)
(13)
$<[(!").
a set of orthogonal unit vectors. The component

in the proper sense, of 4* with the
given by the scalar product,
t|>(r
).
we
are expanding is an eigen-i|* of any set of observables A, say

use the following convenient notation, which is seen to be com
t|>
4>(A
is
4>
|r
):
pletely self-consistent:
r
(12
These equations express the (unitary^) transformation from one system of

are known as transformation
eigenstates to another; hence (F A ) and (A
|
I")
coefficients.
*
The normalization condition satisfied by the eigenstates of such a set of observables is given by
As a matter
S(F,r
f
of convenience
we
shall write this
product of S
as
8(y^..y^; y^yj!)
simply
).
A unitary transformation is a transformation from one set of normalized (unitary-) orthogonal
states to another such set.
22
REPRESENTATIONS OF STATES AND OBSERVABLES

The observable a
defined
(Hall")
characterized in the
is
[cf. (5)]
explicitly
T scheme by the set of numbers
by*
a<KP)
and given
19
*(n(na|P),
(14)
by
The square array formed from these numbers as

and
range over the
allowed values of the observables T is called the inairix of a, and the numbers
themselves are called matrix elements or components (cf. 7 2 ).
In the r scheme the matrix of any one of the s,
y say y^, is diagonal [i.e.
= ]. Since Yi^(P) =
we have from
(P|nl r is zero unless
yi^(P)>
(15) that (P^n|P) = %S rr *. The matrix of an observable a which commutes
P P
")
with Yi is diagonal with respect to y

i3
This follows at once from (6).
From
the definition
are given
1 2 2 it
by
i.e.
Is
(P|a|P )
zero unless yi
= yj.
follows that the elements of the matrix of
(rV|r*)-(E>|r
af
(16)
).
The matrix of a f is the Hermitian conjugate of the matrix of a. The matrix of

a real observable, for which
(P|a|n(r|a|P),
is
(17)
said to be Hermitian.
PROBLEMS
1.
Show
that the relation between the components of
ij*
in
two
different representations
2.
is
(18)
).
Show that the relation between the matrix components of a in two different representations is
(A
|oc|A")
(A |P)
(Flam (r*|A*).
(19)
FT"
3. Derive the following identity, which
is
characteristic of
a unitary transformation
(20)
4. Show that the sum of the diagonal elements (the trace) of the matrix of an observable
independent of representation, i.e. that
|A
(21)
).
and that this sum is just equal to the sum of the characteristic values of the observable
by their degeneracies. (The Diagonal-Sum Rule)
5.
is
In the T-scheme we can represent the action of a on any fy [cf.

)-
10)] in
a, weighted
the following fashion:

(22)
* In this notation one obtains a

complete analogy between the form of a quantum-mechanical
observable and Gibbs dyadic if one writes a as the operator
Here the fact that the

t{>
and the
cation indicates that this
appear in the order opposite to that conventional for multipli

not their scalar product but that they form a unit dyad.
<Jj
is
22
20
6.
Show
that
if
a anticommutes with
y,-,
the matrix element
(aY 8 + Ya=0)
(r la|r*)=0 unless yj= -y*.

7.
Show
that the elements of the matrix of the product
ap
(23)
of
two observables
are given
(Flofln-S (Flam (r lflr*).
states i^rj), iKTj), ...,

8. Consider a set of
derived from these by a unitary transformation:
and a
4>(IT)
set of
4W)
and a set ^(Aj),

Consider also a set of n states ^(Pj), ...,
a unitary transformation. If a is any observable, show that
S|(rt||r!)|
m states
4>(Ai),
....
^(A?) derived from these by
=S
...,
by
(24)
(25)
Sl(At|a|Ai)|*.
(The Principle o/ Spectrottcopic Stability)
For the states of the preceding problem, show that
9.
where
K is a constant independent of
and jf, then
S (AiM Aj)
if
also
<Ai|a|Ai)
=*
8<,.
fcl
Continuous eigenvalues and the SchrSdinger representation.

These formulas must be modified in the case of an observable w whose
but is wholly or partially
spectrum of eigenvalues is not wholly discrete,
3
An eigennj* belonging to a value w lying in the continuous range

normalized
be
to
according to
continuous.
is
$(w
where
8(#) is
)ty(w
= *(v/ -w*)
(1)
an improper even function of a? defined by

f
(2)
which see for a complete discussion of the S function).*

The expansion of an arbitrary in terms of the eigen-^ s of w will now be
of the form
(Dirac, Chapter iv,
t|>
)(
where
for
>\
(w
ranges over the discrete spectrum,
trum of w. Here
all
wm
If
"
is
plication of (3) with
/fl
\
$(w
),
(1)
w"
(w"\
),
(3)
over the continuous spec
= $(uf)fy
in the continuous
m
from
)dw"
spectrum this follows, on multi
and the relation
ff(x)S(x~-a)dx=f(a)
(4)
The 8 function as used in connection with continuous spectra is to be regarded as a convenient

abbreviated notation which permits simplification in the Handling of more complicated mathe
continuous spectrum, although able to lead
matically rigorous expressions. This treatment of the
one into trouble, is certainly the most convenient; the 8 function may be in general treated as an
exact symbol when integrated out according to (4).
*
THE SCHRODINGER REPBESENTATION
32
21
In general, a similar replacement of integration over the

wn )
continuous spectrum for summation over the discrete, and of 8(w
for S a a - must be made in the preceding discussion in the case of a continuous
(Dirac, p. 73).
spectrum.
The most important instance of an observable with a continuous spectrum

the coordinate x, which we assume to have a completely continuous range
of allowed values. For a single particle in three dimensions x, y, and z form
a set of commuting observables (complete except for the peculiarity of
is
we shall later introduce as a fourth commuting

observable a having just two eigenvalues). For n particles we have the set
#1
i
*n yn zn We shall denote this whole set of observables by
*i
electron spin which
>
>
J>
x for short,
>
>
>
writing
$(tf)i\>(x*)
S^-^J^S^-Oa^-^S^-O
where
We
= 8(tf-tf),
shall usually denote

x.
(5)
...8-<).
an allowed value of the above set of observables

arbitrary fy i& terms of the eigen-i|/s of
The expansion of an
merely* by
x is then written
)dx,
$=(
(6)t
\x)$(x)dx,
and ( \x) are complex conjugate functions of x.

)
(x\
This representation in terms of the eigen-ij/s of x (the Schrodinger repre
sentation) is the most useful of all representations, since from it is derived
the Schrodinger equation which has proved to be a most powerful tool in the
where
explicit solution of quantum-mechanical problems. It will be
noted that the
is called the
)
completely determines ty. This function (o?|
(x\
as
written
and
will
be
of
(the
Schrodinger
4
Schrodinger representative
function), or $x if we wish to indicate the independent variable explicitly.
Coordinates will throughout this work be relegated to the position of sub
function
if*
when they need be written, the space on the line being reserved for
the more important quantum numbers. Thus the function (x\T ) which is
the Schrodinger representative of fy(T ) will be written as ^x (T ) and called
scripts
the (Schrodinger) eigen-*/r belonging to F. Since iff determines

equally well speak of a system as being in the state fy or the state
4(r
or the state ^(F). In this
notation the expansion
(6)
t|>,
we can
the state
J/T,
becomes
* There will be no confusion with the

Schrodinger operator x because of the omission of the
prime here, since no distinction need be made between the operator x and the allowed value x in
the Schrodinger scheme.
f dx=dxidyi dzt
...
dz n .
32
22
If a
is
an observable,
which we
wty
shall write as
is
and will have a Schrodinger representative
a
4>
i.e.
we
shall write
oc</>,
*=
(8)
This equation defines the operator a in the Schrodinger sense as that func
which represents ^ and turns
tional operator which acts on the function
</r
wft which represents aty*

We shall now prove two theorems which show the great similarity in
behaviour of 4 and ^ that justifies their being called by the same letter.
it
into the function
which holds between ty$ holds also between their

the representative of the sum
representatives. This follows from the fact that
that the
s is the sum of the representatives of the individual
of two
I.
Any
linear relation
<|/s;
tjj
representative of cfy is c times the representative of

are equal, .their representatives are equal.
II.
The
scalar product of two
i|>
and that
s equals the integral of the
representatives in the sense that if 4(1)
is
if
<|>;
represented
by
two 4
product of their
i/r(l) 3
^(2)
by
</r(2),
tlien
(10)
The proof of this
follows
Now the eigenstates
from
4>(F)
(4), (5),
ift(Y
(7),
are normalized (2 2 9) according to
$(r
hence the
and
)4>(r
HS rr
*
;
have the normalization (theorem
II)
* This notation assumes that the

representative off(x) t{j, where /(x) is any algebraic function
of the set of observables X 9 is just the function /(x) t^. This is easily seen to be the case:
/(*)*&.
(9)
If w is an observable which has a continuous range of eigenvalues, its eigenstates must be

jnormalized according to (1) instead of 1 2 1 and integrals must be substituted for sums in the
f
expansions in terms of */j(w ). In this case we obtain for the normalizing condition
-u?)
in place of (11). This normalization Qfi// x(w ) is accomplished
in such a way that the set of eigendifferentials
)
=-
if we
(12)
choose the normalization factor
f
V AJw
(13)
32
23
CONTINUOUS EIGENVALUES
may be expanded in the T scheme according to

f
Any
4>
r
Hence (theorem
I)
$=?>$(? )
where*
The
(P|
state
Hence the
a 4*
(I")
vLift(T
(PI
(16)
),
= $(? )$=
(17)
)$da:.
has the expansion
of
(8) will
have the expansion

|r
(is)
))
where from 2 2 15 and theorem II

(19)
Finally the allowed values equation
cciK0 = a
<K*
becomes the functional equation

)
a</r(oc
which
is
=a
<A(oc
the general form of Schrodinger
(20)
)>
equation for any observable.
As
normalized to unity in the sense of (11). The ^ s will then be orthogonal if \iv ~
A, and will
w chosen at the small intervals A have all the properties of a set of eigenfunctions
going with discrete eigenvalues. To see this let us evaluate the integral
is
>
w"\
for values of
[
/
XtA W/ ) Xx( w
dx =
")
If
J
I
the integral over
and the value
w"
has the value
for all other values of
wr takes on
\tJw )$
w"
for all values of
w The whole
.
|w
w,
A rici^+A r~
w which
integral then
lie between
becomes
w"
and
w"
-f
A,
-!i>*|=-Â.
/C
in terms of the eigen-^ s of w may then be "written as a sum over

Xx( wT
values spaced by the interval
(this is rigorously true asymptotically as A-> 0)
The expansion of a ^
where
A0*
âsK) dw/ /
w"
if
+A-
4-
/"i/j
=4
&J
Now
/W
I
)>
wr
We
could define, in exact analogy, a symbolic x(w ) which would have properties similar to the
above, and it is in some such way that a mathematically rigorous treatment of the continuous
spectrum must be given.
* The set of functions

^(F ) are to be considered as a set of unit vectors in ordinary function
space (Hilbert space) in terms of which any arbitrary ift may be expanded,
24
32
of the
yet we hare defined the operator a only symbolically; the discovery
of
the
co
functions
exact functional form of oc for observables that are
ordinates and momenta was made by Schrodinger and will be the subject
of
6*.
The
4.
statistical interpretation.
make a measurement of the complete set of commuting observables

of
in the state ijj, what is the probability P(T
Yfc on a system
In
the
answered
o4* ^kis question ma y be
finding the values y 1? y^,
65).
following way (Dirac, p.
Consider any function /(F) of y ls y 2 ..., Y&- The average value obtained
is
for this function when making measurements on
If we
Y2>
Yi>
$>
On
the other hand, this average value must equal
Since the last two expressions are equal for arbitrary functions
equate coefficients of each /(I") to obtain
p(r )=|(r
Hence the probability that the y
/,
we can
(i)
)|.
have the values
I"
in the state
is
i|>
in a state
Similarly the probability that the y s have the values
|(r
)|
in which the 8 s are known to have the values A [i.e. In the state
2
I"
now ask the probability P(x)dx that the set x of observables have
(to at x, we must modify the above computation to suit the case of
is given by
a continuous observable. The mean value off(x) in the state
If we
values in
^|>
from 3 2 9 and
3 2 10.
On the
other hand this
mean value
=
jf(x)P(x)dx.
Since /(x)
is
an arbitrary function of the variables
x,
we obtain from this the
physical interpretation of Schrodinger s ?/r function: that $$dx is the prob

ability that the system be in the coordinate volume element between x and x+dx.
This interpretation
given
by
seen to be consistent with the normalization of
is
as
(3 11).
* Since
f(T
/(F)
observables of Problem 1,
<|jff")
<1>(F
2
.
),
an.
obvious generalization to the case of a set of commuting
THE STATISTICAL INTERPRETATION
42
Show
1.
25
PROBLEMS
=
S
directly that
1.
?(!")
I"
Show that
2,
for all 4*;
if
a real observable
and conversely that

2
Show that a + p
3.
2
,
if
has no negative eigenvalues, then
$o&4^0
ôr a^
4>
ên a nas no negative eigenvalues.
where a and p are any two real observables,
is
real
and has no negative
eigenvalues.
The laws of quantum mechanics.

We have now laid out the mathematical pattern in terms of which the laws
5.
of atomic physics are formulated. The remainder of the theory consists of

the recognition of the properties of the operators which are to represent
various observables. To a particular mode of observation with certain
apparatus
is
to be associated a certain operator.
The laws of nature
are not,
between the numerical values given by

certain experiments, but relations between the operators that stand for
various modes of observation. The recognition of what operator is to be
as before, the functional relations
associated with each set of experimental operations has been carried out thus
by appeal to the correspondence principle (as with coordinate

and
conjugate momentum) and partly by appeal to experiment (as
position
electron
with
spin). Of course the correspondence principle itself is a broad
far partly
generalization from experiment, so the
known relations between
operators
from experiment.
The cartesian coordinates x l9 J 13 zl9 ..., xn y n z n or for convenience
...
#3JIJ of a system of n particles, and the conjugate momenta
3C 15 & 2
for physical quantities all spring
Pi
Pz>
->
Psn satisfy the following quantum-theoretic laws of nature

r
91):
i
(Dirac,
^
*?
p.
v V1
L#i>3tyJ
#i*j
**
x^
range of allowed values. The

,
in3
xzn
as
first
u,
both having a continuous

equation has been presupposed in using
These equations are consistent with K$ and
x x & 25
commuting observables
pi
for a Schrodinger representation
2
.
Analogous to the total energy of the system is a Hamiltonian function H,

which is the same function of the p s and x s as on the classical theory for the
AB
-BA. (Dirac uses [A, B]

shall use the notation [A, B] for the simple commutator
closer analogy to the classical Poisson bracket.) The commutator of a single
of two is given by formulas similar to that for a derivative of a
observable with the
*
for
We
(AB -BA)/i& in
product
P roduct
[AB, C] =01
C]B +A[B, C]
[A,BC]=[A,B]C+B(A,C].
2)
26
52
analogous dynamical system* The importance of the Hamiltonian in the

classical theory lay in the fact that
through Hamilton s equations of motion
it determined the time variation of the state. That continues to be its
importance here, the dependence of the state

(Dirac, p. 115)
by
j*
on the time being given
3
(3)
where
dfy/dt indicates a vector whose components are the time derivatives

of the components of t|* in a fixed representation. In
Schrodinger s notation
this
becomes
(3
The time dependence of the eigenstates of energy

One has for the eigenstate belonging to H
is
particularly simple.
and therefore such a
state merely changes its phase in time:
Corresponding equations obtain for the Schrodinger function $,(# ), In an

any observable a is independent
eigenstate of energy the average value of

of the time.f This is proved as follows:
Since the probability that the set of observablesT have the values F
depends
only on the average values of all functions of the y s ( 4 2 ), this probability
independent of the time in such a state. Because of these properties an

eigenstate of energy is called a stationary state.
is
PROBLEM
Show
that the commutator
f(
x )PiPiJ(x)**iH
(5)
9X|
where /(#)
is
any function of xl9
2>
...,
xsn
expressible as a
power
series.
Schrodmger s equation.
We shall now discuss the question of the functional form of an observable
2
in the Schrodinger sense, as defined
by 3 8:
6.
ai}>=
n^(x)tx.^dx,
However, the exact order of factors and related questions, which are immaterial in the
classical theory, are of great importance in the
quantum theory, and for any particular problem
must be determined, in the last analysis, by
experiment.
t If a does not involve the time
explicitly.
62
SCHEOBINGEB
where a
Pi
is
27
EQUATION
supposed to be an algebraic function F(x, p) of xls
...,
X3n
9p*n-
The Schrodinger operator for /(#)
is
2
merely f(x), as shown by 3 9.
Schrodinger s discovery* consisted of the observation that the proper

operator for the observable p i is
ifidjdxi This is easily shown to satisfy
.
the
quantum
conditions 5 2 I for example

;
hence
We obtain the operator foipjpi by letting p$ operate
Hence the Schrodinger operator
for
p^pi
is
general the operator for a function g(p) is g(

From an extension of these considerations
ifidjdx^
iifid/dXi),
and in
ifid/dx).
it is
seen that the Schrodinger
operator for a general algebraic function

a = F(x,p)
is
Hence the Schrodinger equation
= F(x, - iM/dx).
(I)
(3 20)
becomes
(2)
a standard type of characteristic value differential equation whose

characteristic values and functions are to be determined by the use of
The original equation given by
).
auxiliary boundary conditionsf on
case a=J?, which gives
the
for
this
was
particular
equation
Schrodinger t
This
is
</r(oc
the energy levels and eigenstates of total energy.

7.
Matrix mechanics.
We have seen that by the use of Schrodinger s representation in terms of

we can obtain a
theory entirely independent of symbolic 4 s and a s. In
the continuous set of allowed values of the variables x^
formulation of the
*
SCHBODINGEB, Ann. der Phys. 79, 734 (1926).

conditions. This discussion is not complete;
t See p. 143 of DIEAC for a discussion of boundary
the governing condition must be that one find a complete orthogonal set of solutions of (2). See
VON NEUMANN
J
S book, page 53 et seq.

SCHEODENGEB, Ann. der Phys. 79, 361
(1926).
28
the same manner the theory can be formulated in a way in which we deal
only with the matrices representing states and observables in terms of the
discrete set of allowed values of a complete set of observables T. This corre
sponds to the original matrix mechanics of Heisenberg, Born, and Jordan.*
We
in the expansion
represent $ (or iff) by the array of coefficients (F
)
in terms of the ^(F ) [or ^(T )] written as a matrix of one column. That is,
4* and 4K^ ) ar represented respectively by
|
(P|
and
When a $
represented this way, a $ must be represented by a matrix of

one row, in order that *jy|^ niay be an ordinary number (a matrix of one row
and one column). For example, the product $(A )*KA") becomes
is
(1)
by the usual rule of matrix multiplicationf (cf. 2 2 20). In this form we see
clearly that we cannot add a 4? and a ip, and that we must form their product
in the order
tjujj-
The observable a
is
represented by
second index
means
T"
the
matrix as defined in
(P|alP)
(P|a|P)
(P|a|P)
(P|a|P)
(P|a|P)
where by convention
its
the first index
column. If
(P|oc|P)
..
(P|a|P)
(P|a|P) (P|a|P)
..
in
is real,
(P|a|r")
this
22,
i.e.
by
..
row and the

Hermitian, which
labels the
matrix
is
2 2 17) that the diagonal elements are real and that the corre
sponding elements on opposite sides of the diagonal are complex conjugates.
(cf.
HEISENBBEQ, Zeits. fur Phys. 33, 879 (1925);

BORN and JOKDAN, ibid. 34, 858 (1925);
BOKN-, HEISENBERG, and JORDAN, ibid. 35, 557 (1925).
t The product of a matrix &? with m rows and k columns and a matrix 3S with n columns and
k rows is defined as the matrix of m rows and n columns obtained as follows
:
s are over .7 = 1, 2, ..., fc. For a discussion of matrix

algebra see, for example,
Introduction to Higher Algebra, Chapter vi.
where all the S
BdoHEE,
72
MATRIX MECHANICS
It is
29
now readily seen that all of the preceding equations remain true when
one substitutes for
and a s
and uses the laws
their respective matrices
of matrix algebra. The allowed-values equation becomes

a
(P|a|P)
(P|a|P)
.(2a)
Carrying out the matrix multiplication,

2
(cf. 2 7 et seq.)
we
obtain the set of simultaneous
linear equations
S[(F |a|r )-a 8 r^](F|a ) =

/
The condition that

determinant of the
this set shall
for all
have a non-trivial solution
I".
is
(2b)
that the
coefficients vanish, i.e. that
-a
(r |a|r
=o.
(3)
This determinant, though usually of infinite order, will often be such that
the only non-vanishing elements lie in sub-squares along the diagonal, so
that the infinite determinant will factor into an infinite product of finite
determinants each of which can be treated by ordinary algebraic methods.

An approximation to its solution in the case in which certain of the nondiagonal elements may be neglected will be considered in the next section.
The roots of (3), which is called the secular equation, give the allowed values
of a. For a d a/-fold root, the equations (2) furnish d^ linearly independent
these coefficients
condition
(1):
= 1,
..., d^. The values of

must be chosen in accordance with the orthonormalization
sets of transformation coefficients (F |ocV) for r
S(aV|F)(F|aV)==Srr
,.
(4)
These transformation coefficients determine the eigen-ij/s of a, i.e. they

determine those linear combinations of the eigen-4 s off in terms of which
the matrix of a is diagonal; hence this calculation is known as the diagonalization of the matrix of a.
While the coefficients (F| A ) may be arranged into single rows or columns
there is a sense in which the whole square
to represent the matrix of 4*(A
)>
and A may conveniently be considered as the

array of (F|A ) s for all
matrix which transforms the eigen-^ s of T into the eigen-ij/s of A. Let us
write
2
(F |A
(F |A
(5)
)
72
30
The matrix on the right, which we shall write as
||(r
)||,
is
a unitary matrix
in the usual sense: it has the property that its Hermitian conjugate, which
is easily seen to be the matrix ||(A
|r )|| is also its reciprocal, by (1); i.e.
ll(r|A )iHi(A
where
I is
Now if ||(r
|nii=i
the unit matrix.
represents the transformation from the T scheme to the

scheme to the E
scheme and ||(A |E )|| the transformation from the
jA
)||
scheme, the transformation from the

2 2 18)
T scheme
directly to the
by
||(rqEO|Hi(n A 0!h!( A 1 E )l|.

The transformation of the matrix of a from the r scheme to the
given
(of.
accomplished
scheme
2 2 19)
(cf.
is
6)
A scheme is
by the multiplication of three matrices;
(7)*
Al| = ||(A Dll II (r jal nil
of commut
Suppose now that we are given the matrices of a set a, p,
ing observables in the F scheme and we wish to find the transformation to
||(A
oc
-|| (F"
A")||
, .
the scheme in which these are simultaneously diagonal. We diagonalize a

r )| With this matrix, by (7)
by (3), (4) to obtain the unitary matrix ||(r
.
|<x
we transform the matrix of p to the
scheme. Since p commutes with a

this matrix will be diagonal with respect to a hence we can find a set of
(a r
states ip(a ft s
tion of
||(r
|oc>
in the a
of
4* (a
/?
etc.
as in
)ll
and
/}
2 2 . Using (6)
||(<xV|a
/J
we find
)\\.
||(F |a j8
)!!
by the
multiplica
This enables us to obtain the matrix
scheme, the diagonalization of which gives states

8
See 4 for an example of this procedure carried through in
detail.
8.
Perturbation theory.
we know
the allowed values and statesf of some observable oc, we can

develop formally a successive approximation method for finding those of an
observable which differs but slightly from a, say the observable
If
/3
where
e is
supposed to be a small
perturbation on
<x,
this
= a + eF,
(1)
number. Since eF
approximation method
is
may be considered as a
known
as the perturba
tion theory.
Let us denote the mutually distinct proper values of a by a 1 a 2 .... If now
there is a dn -fold degeneracy in the value a n we shall need another index to
,
We may
distinguish the different states going with this same proper value.
denote these states by ^(an*), where Z== 1, 2, ..., dn Since the set ol
.
^r s
for
determined only to within a unitary transformation, it will be

desirable to choose the members of this set in a way significant to the
a given
* This
is
T=
is the equation which is

usually written in the form
t In this discussion the ^ s and a s may be either in the symbolic or Schrodinger scheme.
PERTURBATION THEORY
82
31
problem in hand. Let us then choose ^(oc ) in such a way that all matrix
wZ
nZ
components of the type (a |F|a ) are zero unless l~l This is possible
since the transformation which diagonalizes that part of the matrix of V
77 1
-
71
which refers to the level a for any given set of ^(oc^ s will transform this set
of s to one satisfying the above requirement.
Now it may be, when we have chosen the states in this way, that the com
i/f
ponents
I, ..., d n
nl
(<x.
71
\
Floe
*)
with a given n are
all different as
runs over the range
It is this special case, practically the only case of importance in
which we
applications,
general case later.

The observable
j8,
shall consider
which
differs
but
now.
We
little
shall return to the
from
a, will
more
have dn proper
71
values which coincide with a in the limit = 0, and which differ but little
n
from oc for small e. We shall call these d n values finl /3712 ..., P nd*, and shall
,
suppose that each of them may be expressed as a power series in
^=^+ # +
of the form
10*+...,
n
$^ = a
where clearly
(2)
for all L
(3)
We shall suppose the eigenfunctions which belong to these values to be also

expressible as a power series in
of the form
We
nZ
np
shall
{^ } is a linear combination of the $o(K ) for^p= 1, ..., dn
72
see that the advantage of our particular choice of ^(a ^) lies in the fact that
where
we may take ^ {^} = ^(a ^).

Our problem is now to determine
7
powers of
in (2)
and
(4)
formally the coefficients of the various

in terms of the matrix components of F and the
properties of the observable a. This we may do

the coefficient of each power of e in the equation
known
by equating
to zero
/^(n = W(
when this
[a
is
(5a)
expanded in the form
+ F - JB{? -
J8*
- *W -
. .
[^ {j3*} +
.]
W^} +
e* h{?*) +...]
= 0.
This gives us the set of equations:

(6a)
(8b)
nr
nJ
If we express ^ K{^ ] in terms of the ^(a ), using the notation
= S^(a^ ){awT |^}, (ic-0,1,2,...)
nl
6c )
(7)
32
82
We
we can
have introduced the curly

solve these equations in succession.
braces because the ^K{^nl} do not form a set of orthonormal states, nor are
the {oc^ l/c* 17} coefficients in a unitary transformation. The condition that the
0(j8*
form an orthonormal set independent of
gives us the following
conditions on these transformation coefficients:
= 8(T, id)
|
Sj;{0?]a**
The
attr } {a
7{^
where
{K^| a
(8a)
wT
}
(8b)
0"*}]
(80)
= {anT |K^}.
equation (6a) is just the allowed-values equation for the

and gives us no new information. It tells us that if we take
nr nZ will be zero unless ra =
=
a*, {a
w, as we have already observed.
j8y
jO }
With these facts, the second equation (6b) becomes, when expanded in
terms of the ^(ocu/r ),
first
observable
^(<x"0
a,
- $f {a^
[(
^ - 8(n, w*)
jSjf
{a"|
0^}
n^Z"
+ S (a
Equating the
coefficients of
nl
9Jj(a.
")
is
!
|
F|
0{
n/
tt
|0
}]
= 0.
(9)
to zero in this expression gives
[(a^|Fia-0-ft ]{^ iO^} = 0.

(rl,...,^) (10)
J
Mjr
for each value din and I a set of d^ equations* for the
{a
|0 }. They
3l
This
are satisfied
by taking
jSj
Ha^FIa*);
{a^ |0^-8(Z,r).
(11)
This makes
(12)
the (anl V\
are different, this is to within a phase the
only non-vanishing set of solutions of (10) which satisfies the condition (Sa).
r for
to zero the coefficient of
in
now shows us
For a given
n, if all
<x^)
Equating
0(oc**
n"
^n
(9)
(13)
The third
S ^(a^
ec[uation (6c) becomes,
*)[(^"
-^
r
){"*
2^ -
j8
when expanded in terms
{r| i^ _
jgjfi
"
of the 0(an
"
),
{nr lorti
n"l"
+S
r
(a"
F|a
nT
wT
){a
1^}]
= 0.
(U)
ril
* These
equations have the above simple form because of the way we have chosen our t//(oL nl ).
had used an arbitrary system of ^(a n ) s, we would have obtained at this
point a set of equa
tions determining the ^r(anZ ) as we have chosen them. This would lead to a further
complication of
notation which it seems desirable to avoid.
If we
PERTURBATION THEORY
82
Equating the
coefficient of
- jSjf
nl
{<x^| 1"*}
- J8* + S (a*] F| a
n 3=n
(fyfill
(g
V\ rt n- l
coefficient of
nr
i/f(a
\lft ril
aJ a
rc/j
Equating the
0{a
nl
fflS
from which
to zero gives
ifj(a.
33
(Z"
^ Z)
nT
|
1^} = 0,
V\nnl\
).
(15)
to zero gives
We
have now determined ^{/J^} completely except for the coefficient

which is not determined by equations (6). Equations (5) and hence
},
{a
determine
aset of orthogonal states, but do not take care of normalization.
(6)
nZ l nZ
|
This means that equations (8) will be automatically satisfied for n l ^nl
nl must be expressly
(orthogonality condition), but these equations for n l
considered to secure normalization. (8a) has been satisfied by the choice (11).
(8b) tells us that
<
\ni\ynii
_j_
anf \m\
i
._
Q^
We
or that the real part of (a *| l ^} is zero.

may choose the imaginary part of
n
w/
the arbitrariness in phase of
with
is
This
connected
|l } arbitrarily.
7
{<x
nz
But
e^<
</r(j3
if 0( j8
a solution of the same phase for = as

2
=
given by (4), we have, if f(e) k^ 4- A 2 c +
).
j/r(j8
7t/
is
This shows that

n
</r(a
O]
is
to set
{<z
nl
nl
arbitrariness of
anzi pzi_ Q^
nl
ifj(p
\l
}.
if/(0)
= 0.
\JjQ
[i.e.
of
The most convenient

n 7)
we may obtain any solution from such a particular one by multiplica

we see that the arbitrariness in
which we shall also set equal to
tion with a phase factor. With this choice

an imaginary multiple of t/r
2 is reduced to
?/r
. .
^ contains an arbitrary imaginary multiple of
which occasions the
procedure
since
is
and
on
for the higher approximations.

n r
to zero for n"^n in (14)
) equal
Finally, setting the coefficient of
*
n/
w
r
be continued as long as
|2 }. This procedure may
gives us the value of {a
one pleases, and it is clear that recursion formulas may be obtained express
one. The value of {x.n r \2 nl}
ing each approximation in terms of the previous
zero,
so
"
/f(<x
"
obtained as above. The coefficient of ^(oc^) in the expansion of the fourth

nr
rtr
The normal
equation (6d) gives $ The coefficient of ^(a ) gives {a |2^}.
71
ization relation (8c) for n l ^nl determines the real part of {a ^^}, the
is
imaginary part being arbitrarily set

a collection of formulas.
zero,
and
so on. See the next section for
82
34
If the (an? |F|a^) for a given
are not all distinct, (13)
is
no longer the
Such a problem should he handled by writing three indices

on the states ^(ot^ where (a^|F a7* ) is independent of j. If we are to
choose the set of states for a given n and I with significance, we do it so that
correct solution.
)>
the expression
g* (***
^Tf)(^Tf V an/r
I
(18)
a*-a
nr?
which is written down from (15), vanishes unless =j In this case we shall
= #(***) and that
will be given by (18) for
find that we have #
be much more
j*=j. The formulas for the coefficients in ^/M) will now
should
complicated than before. If (18) for j=*f still has a degeneracy, we
ff
j"
{W
have to use four

9.
R6sume
indices,
and
so on.
oi the perturbation theory.
We know the characteristic functions and values for an observable a and

are interested in those of
an observable
= +
oc
/3
7.
(1)
Denote the distinct eigenvalues of a by a1 a2 ... and let ocn be dn -fold

= 1, 2, ..., dn in such a way
degenerate. Choose the dn eigenstates 0(oc*0 (Z
*
that (a^lFjoc 71* ) vanishes unless Z = Z .*

such that the dn values of (oc^l V\&nl ) are
The
following applies only to
all different.
n
{*
lf
\^ +
Write
+ ...;
jtT
(2)
^^
(3)
Then
= 2
nT
(a
= 2 (a^lFlaÔia^
Idc-l)
111
}-
* One can determine this set of states

by diagonalizing that part of the matrix of
to any set of d n orthogonal states belonging to
a".
F which refers
RESUME OF THE PERTURBATION THEORY
{a
10.
nl
nz
|l
35
}=0
Remarks on
the perturbation theory.
It may not at first sight seem clear why we have set down these com
plicated formulas for the perturbation of a degenerate system when we
might by a simple device have reduced the degenerate system to one which
non-degenerate. If we add to the matrix of oc the diagonal elements

7
(a^leFja ^) of the perturbation matrix, we have a diagonal matrix of which
we know the characteristic values and functions and which is non-degenerate
is
by hypothesis. The rest of the matrix of eV may now be considered as the

perturbation; and the much simplified formulas to which those of 9 2 reduce
when the system is non-degenerate are applicable. But our perturbation
theory is expected to converge rapidly only for small perturbations a small
perturbation being one which causes shifts in the eigenvalues small compared
to the original distance between eigenvalues.* Now if we take as the un
perturbed levels those for which the degeneracy is removed by the diagonal
* This statement is based on the
7
following rough considerations. Let us represent by T a
of the order of magnitude of an element of the perturbation matrix V, and let us
suppose
r
w*
nl
that the difference between two diagonal elements of the type (ocm P |ocnl ) and
) (i.e. a
number
(<x
F|a
difference j3f -j8? such as occurs in the denominators of 9 24 and 9 2 5) is also of order V. Let a.
nn
be a number of the order of magnitude of the difference n - an between two adjacent
degenerate
2
th
levels of a. Then the /c term in 9 2 is of the order of magnitude
<x
while the K ^ term in (an/I
nr
|)8
is
of the order of magnitude
Since the displacements caused by the perturbation are of the order of eF and the
original distance
3-2
36
1C 2
elements of the perturbation, we have no reason to suppose that the nondiagonal elements will have an effect small compared to the effect of the
diagonal elements,
levels.
Hence
if
i.e.
small compared to the distance between unperturbed

rapid convergence we shall probably need to
we wish
consider the set of states going with an unperturbed level as a unit and the
perturbation as a whole, as we have done in the last sections.* It should
be remarked that
if
we do use
the device suggested above
we obtain
for
the coefficient of e 2 in p71*

***
~r>
-n]\
T/l
7>7\
~>n
-w 7
*I7\
nfl \
which contains in the denominator the distance between the first-order

eigenvalues instead of that between the zero-order eigenvalues. For calcula
tions just to the second order this formula will at times be convenient, and
probably as accurate as the former.
To obtain a slight acquaintance with the perturbation theory, let us
consider in detail a simple finite case for which we can obtain an exact
is
solution.
Let us take the matrices of a and
F to be
(5)
[where we have already chosen the states ^(a 11 ) and t/r(oc 12 ) in such a way that
n
12
(0c
(F|oc )=:0] and consider the characteristic values and states of the
observable
in e
= a 4- eF. From the formulas

no
9 2 4 and 9 2 5
we
find that to
terms
_
90e 2 -180e3
-3690
4
,
between levels is of the order a wn we see that we cannot

expect the theory to converge rapidly
unless these displacements are small compared to this distance.
If the differences p? -jgj* are of smaller order of
magnitude than F, (1) and (2) are too small.
In particular the magnitude of the elements {a (* }
connecting two states which originally
belonged to the same unperturbed level is strongly affected. The above criterion is
<
m m
good provided
the differences in first order shifts are of the same size as the shifts themselves. If this
one might take as a guide the requirement that an
expression of the
is
not true,
type
should be small compared to unity for

rapid convergence.
The question of the convergence of the perturbation
theory has been considered by WILSON
Proc. Roy. Soc. A122, 589; A124, 176
(1929). He shows that for finite matrices the theory con
verges for all values of the parameter e. For infinite matrices no such definite results were ob
tained. [This means that (1) and (2) can be correct
only for small *.]
* In
fact, if two or more unperturbed levels lie very close, it is advisable to consider them as one
degenerate level in the perturbation scheme, adding the differences to the diagonal elements of
the perturbation.
10 2
REMARKS ON THE
while to terms in
THEORY
PERTURBATION"
37
- 5
o-5
+45
-3*
-
3e- 6
U-75e 3},
c }
*- 190-5
- 558
(7)
4-1108-5
- 575-5
The eigenvalues
(6) are plotted in Fig.
for values of e
from
0- 1
to 4-0-1
order in e. The curves marked oo give the

0, 1, 2, 3, and
values obtained by a direct diagonalization of the matrix oc + cF.* Fig. 2 a
shows the coefficients in the corresponding states to 1, 2, and 3 a order and
as given by the exact calculation.
4th
to terms of the
The situation is
as follows. In the zero* 11 approximation there is a doubly

degenerate level at 10, and one state at 0. To the first order the level at 10
splits into 2 states of eigenvalues 10 and 10-h5. In the second order these
2
2
by that at with strengths 90e and lOe which soon
0. A crossing of the true levels
results in a crossing of the two states for e
will be shown in II 2 to be impossible. The third and fourth approximations
include the third and fourth terms in the power-series expansion of the true
eigenvalues (marked oo). The second-order perturbation theory for this case
has little accuracyfor a displacement of over one percent. The crossing of the
levels in this approximation contributes somewhat to this inaccuracy, the
0. The fact that the second
values for
being better than those for
is better than the third or fourth must be regarded
e
for
approximation
12
are over
as accidental. The eigenfunctions $(p u ) and
), which for
11
12
u
for
and ^(a ), rapidly change character
92 per cent, i^(a
^(/J )
12
The approximation
becoming, at = 0-1, 70 per cent. */r(oc ), and vice versa.
states are
repelled
>
e<
>
<
<
<A(
<->0,
to these functions
is
seen to be no better for c
11. Perturbation caused
by a single
<
than for
>
0.
state. |
seen that in general the second-order perturbation theory may be

levels resulting from the
interpreted as a repulsion between each pair of
It
is
first-order theory, of magnitude equal to the absolute square of the matrix

element joining these two levels divided by the unperturbed distance
2
between the levels, or in the form 10 4, divided by the first-order distance
between the levels. It is of interest to see what one can say rigorously about
this
*
repulsive* effect of perturbations.
Where two curves are not separated in Figs
1 2 and 2
t BEILLOTJIK, Joux. de Physique, 3, 379 (1932);
2
,
their ordinates differ at most by 0*0 1 unit.
MACDONALD, Phys. Bev.
43, 830 (1933).
38
._
-.A0
2
Fig. I
\08
-M
:C4
-.02
.04
Successive approximations to the eigenvalues of a +
09
.06
.10
V of (5) as functions of c.
II 2
PEBTURBATION CAUSED BY A SINGLE STATE
X"
39
-0.6 L
2
for the eigenstates of
Fig. 2 . Transformation coefficients in several approximations
a 4- eF of (5) as functions of c.
THE QUANTUM MECHAOTCAL METHOD
40
Suppose that we have diagonalized that portion of a matrix referring to

1
a certain set
1
,
A*,
...,
^(J.
n
,
),
$(A
2
),
$(A
...,
n
)
the inclusion of another state
*I*(X)
in our diagonalization.
we say in general about the eigenvalues

A
A*
where we let (A*\

be different from
X) = o^
(X\
That is, what can
of the matrix
...
aj
...
02
(1)
An
...
matrix
of states, obtaining eigenvalues

effect on this set of levels of
and that we wish to know the
<z
X) = X ?
zero, since a state with
We shall suppose all the
a=0 may
oc
to
be removed from the
entirely.
If the state going with the eigenvalue A of this matrix be
we have, from
(1),
the equations
(of.
7 2 2),
(j=l,2,...,n)
0,
0.
(2a)
(2b)
=l
Substituting the values of (4*[X) from (2a) in (2b) gives the equation in
A,
which is equivalent to the secular equation obtained by setting the deter

minant of (2) equal to zero. The roots of this equation give the allowed
values of
A.
i
is such a
(3), which is seen to become infinite at X = A
shown by the heavy curves of Fig. 3 2 for the special case
noted. The individual summands which are added to
give this curve are
indicated by the light rectangular hyperbolas. With these same values of
the A a and cc s, the value of X may be chosen at will. With the
particular
The
left side
of
function of A as
X X
=
= 13, the right side of (3) is represented, as a function of A,
by
the leftmost straight line of unit slope. The abscissas of the encircled
points
of intersection of this line with the heavy curve
give the five allowed values
of A. The allowed values for two other positions of the
perturbing level
are also shown: one for a position
= (2) = 23 within the group of levels
A i9 and one for a position = JT(3) =36 above the whole group of A s.
It is clear from this picture that the (n-f
1) proper values of the matrix (1)
choice
X Z
A\
assume that 1 A*
..
A n one below A\ one between A 1 and
one between A* and A*, ..., one between A n~l and^ n and one above A n
He, if we
<
<
<
PERTURBATION CAUSED BY A SINGLE STATE
II 2
This
is
X and of the a
independent of the value of
s,
41
provided none of the
vanish.*
It is interesting to speak of this in the following way. The perturbation

and a set of levels A\ ..., n which He entirely above it
between a level
pushes each level of the set up, but not above the, original position of the next
itself is correspondingly pushed down. The perturbation
higher level, while
and the set A\ ..., A n pushes up each level of the set
between any level
X
X
2
l=
15, ^. == 20,
Graphical determination of the roots of (^lÎn this example, A
-4 4 = 35; [^1 = 5, |a
a,2 = 5, ou = 2, a4 = -2. The roots are indicated for
2
=
three different values of X: ^>=ii^
>=23,r
2
Fig. 3
J. 3
= 25,
I
|
X&
X (but not above the original position of the next higher

below X (but not below
set which
level);
the original position of the next lower level); while X itself is pushed up or
which
lies
above
pushes down each level of the
down according
to
whether the
lies
sum
a *l 2
*r=2*
* If two of the
are equal, say
A* =A J, there
is
also a root at A*.
II 2
42
calculated for those levels below

those levels
X is greater or less than this sum calculated for

to the displacement of X one may
above X. Hence with regard
correct direction of
say that the second-order perturbation theory gives the
motion even when it does not give an accurate value of the amount.
There is, of course, no rigorous meaning to the correlation in the pre
levels
ceding paragraph of the eigenvalues of (1) with the unperturbed
5
A\ A 2 , ..., A n X. However, when all the perturbations are small, the
states belonging to these eigenvalues will be very closely
,
The curves of Fig.
3 2 are seen to represent, after a fashion, the potential
to
along the A-axis of the plot dne to infinite lines of doublets perpendicular
2
2
2
1
the paper at points
A\ A* y and A^ of linear moment f lo^l fioc 2 f |oc3
One obtains an eigenvalue when

ever this potential equals A X. The dipoles to the left of an eigenvalue raise
and f |oc4
in the direction of increasing
A.
the potential and hence tend to push the eigenvalue up while those to the
down.
right lower the potential and tend to push the eigenvalue
At any given eigenvalue the relative amounts contributed to the potential

different strings of doublets are rather closely related to the relative
contributions of the corresponding states to the eigenstate in question.
These relative contributions to the eigenstate are shown by (2a)
by the
(4)
to be proportional to the quantity
_ ^ instead of to the quantity
*"
>
jQ/
ft
which measures the contribution to the potential.

One sees in Kg. I 2 an illustration of the forces which keep an eigenvalue
from passing the next unperturbed level in the fact that /J11 could not exceed
the first order value of/?12 even though
,
j8
tried
very hard to push it past this
position.
be mentioned that equation (3) is often very easily solved

the values of A, especially for values which lie close to one of
for
numerically
the .4 s so that one term on the left side of the equation is very sensitive to
It should
small changes in A whereas the other terms are relatively insensitive. In this
u
12
way the roots j8 and /? of the problem in 10 2 were easily obtained to six
significant figures
In such a case
/? being then given by the diagonal sum rule.

10 2 the roots must be obtained very accurately to
the root
as that of
give moderate accuracy in the transformation coefficients because of the

occurrence of A^ A in (4).
THE ANALYSIS OF NOK-COMMUTIKG VECTORS
12 2
12.
The
analysis of
non-commuting
43
vectors.
In quantum mechanics, the components of the vectors of physical in
become linear
position, momentum, and angular momentum
operators in place of pure numbers. The manipulation of such noncommuting vectors is facilitated by a set of formulas modified from those
terest
of the usual vector analysis.

If
and
are two vector observables
B
(A=sA ai+Ayj+ A z k, where
A
scalar
and g are
observables),* the commutators connecting the
components of A and B transform like the components of a dyadic; hence
A
A X9 A y
it is
convenient to write
[A,B} = [A x ,BJii+[A x ,By]y + ....
Two vectors mil be said to commute when this dyadic
(1)
vanishes,
i.e.
when each
component of one commutes with each component of the other. This dyadic
is
not the dyadic
AB
BA, but is related to it as follows:
[A Bl = AB-(BA)
9
where
= AB-BA-(BxA)x%
$ is the unit dyadic ii+jj+ kk and
posed) dyadic.! Hence in general
[B,
A]
indicates the conjugate (trans

\A, B], but from (2) we see that
[B,^]=-[J,B] C
With these
(2)
(3)
conventions, for a single particle having position vector r

the set of commutators 5 2 i may be written as
momentum vector
j>,
[r,r]
= 0,
[J>,]
= 0,
[r,j>]=-tor]=&
In a similar fashion, the commutators of a scalar observable

components of a vector observable A form the vector
(4)
X with the
{X,A}~-{A X]~[X,A x}i+{X>A y}j+[X,A z}k.
(5)
(This is the vector XA AX.) A scalar will be said to commute with a

when this vector vanishes.
The following generalizations of 5 2 2 will be found very useful
AB, C] = A-[B, C] - [C, J]-B,

[^xB,C]=^x[B C]~([C,^]xJ8) c
J
[C,
A x B] = [C, A] x B- A x [B, C\)

(
(7a)
(7b)
[Z,JB].
we
vector
(6)
* In this section
and
(8)
shall use light-faced letters for ordinary observables, reserving bold-faced

Gibbs sense.
letters for vectors in the
a complete discussion of the ordinary

f See GIBBS-WILSON, Vector Analysis, Chapter v, for
the
theory of dyadics. It shotdd be pointed out that in combining non- commutative vectors,
order of the factors must be preserved; thus
,
etc.
12 2
44
With non-commuting
vectors
all associative
and
distributive properties
of ordinary vector analysis are preserved; other elementary operations
must
be performed in accordance with the formulas

(9a)
]x
where s and
and vector obtained by inserting dots and

between the two vectors of each dyad in a dyadic the
indicate the scalar
crosses respectively
scalar
(9b)
and vector of the dyadic
as defined
by Gibbs;
A*BxC=AxB-C,
(10)
[B,A]-C
(Ha)
4-[C,B],
(lib)
(AxB)xC=BA-C-AB-C-[B,A]-C
(12a)
=J-CB-JB-C-J-[C,B].
(12b)
The
application of these formulas to the ordinary vector operator

(V) has been discussed by Shortley and KimbalL*
*
SHQRTLEY and KIMBAJLL, Proc. Nat. Acad.
Sci. 20,
82 (1934).
del
CHAPTER
III
ANGULAR MOMENTUM
Because of the symmetry of the atomic model, the dynamical variables
analogous to the components of angular momentum in classical dynamics
The properties of
play a fundamental role in the theory of atomic spectra.
these observables, and of the electron spin, which we shall develop in this
of atomic spectra.
chapter, have constant application throughout the theory
NOTE. In this chapter and henceforth we shall no longer distinguish by a difference in type face
between quantities of a symbolic nature and quantities in the Schrodinger representation. While
this explicit distinction has been found very useful in the exposition of the theory, because of the
two schemes no confusion will be
complete symmetry which has been shown to exist between the
entailed by using the same type for each in what follows. In this way, bold-faced type is released
for its customary role, that of indicating a vector quantity in the restricted three-dimensional Gibbs
1.
Definition of angular
In
classical
origin
is
is
mechanics
r and
if
its linear
momentum.
the position vector of a particle relative to an
momentum is p,
defined to be
its
angular
momentum about
L = rx p.
(I)
in quantum
tentatively define the angular momentum of a particle
mechanics as a vector given by this same formula in terms of the position
and momentum vector observables.
We
From the basic commutation rules (12 24) we find that the vector L does
not commute with itself, i.e. that the three observables Lx L y) Ls do not
commute with each other. Instead we find, from a double application of
,
12 2 7, that
=r x
[p r]
9
xp- (r x [p
r] x p) c
since twice the antisymmetric part of the dyadic r x (p x $)

this
becomes
[,]=-
(r
we were
x p) x
3 = - ML x &
is (r
x p) x g,
(2)
commutation rule as
this be equivalent
would
the definition of an angular-momentum vector,
is
more general than
to the definition by means of (1) ? It turns out that (2)
extra generality that is needed to fit electron
(1), and that this is just the
shall suppose (1) to hold for a special kind of
we
So
spin into the picture.
The question now
arises: if
to regard this
ANGTTLAB
46
angular
general
I3
MOMENTUM
momentum which we shall call

angular momentum defined by
orbital
momentum. The
angular
or in terms of components
by
wiU be denoted hereafter by J while the

3
letter
I will be reserved for orbital
angular momentum.
2.
Allowed values
of
angular
momentum.
We proceed now to investigate the properties of the observables Jx Jyy

Jz which are consequences of 1 3 3. The squared resultant angular momentum
(1)
J2 = J| + J2 + J2
,
shown by application of 12 2 6 to commute with the vector / and

hence with all three components Jx ,Jy and Jz As a consequence there exist
2
simultaneous eigenstates for J and any one component, say Jz We can
2
find the allowed values J and Jz by the following procedure. Let $(y J* J z )
2
be an eigenstate of/ Jz and a set* F of observables which make up together
with J2 and J, a complete set of independent commuting observables for
is
readily
the system in question. We shall require that F commute not only with Jz
and J2 , but also with Jx and Jy for a reason to be given presently. Then
,
From
these relations it follows that
Since
Jx and Jy are real observables, we see from this that Jg ^ V/2/
___
\
2
the allowed values of (J^+Jy) and of JT are essentially positive
Problem 3). Now using the commutation rules 1 3 3, we find that
since
42
(cf.
Jz (Jx iJy = (Jx iJy

)
(Jz
(4)
).
2/
Jz ) gives
Operating with the two sides of this equation on ^(y
2/
=
j; *) ( J* ijy) *(rJ* JJijy ) ^(r/
(
f
J"
W*
But this equation

that unless
is
of the form of the allowed-values equation

(
^ g ^ jv j/
it
5)
and asserts
(6)
an eigenstate of Jz belonging to the eigenvalue Jz ft. It is

vanishes,
2
also an eigenstate of/ 2 belonging to the eigenvalue J and of T belonging to
2
the eigenvalues y since JF and F commute with Jx and Jy (This is the reason
it is
denote complete sets of commuting observables ;

n ve used capital Greek
our individual observables will in
heresffcer w shall use these letters to fill out complete sets
variably be Eomaa letters of special significance. For convenience we shall denote a set of eigen
or F*.
values of r by y or y in place of the previous
* In
Cliapter
letters to
I"
ALLOWED VALUES OP ANGULAR MOMENTUM
23
47
F commute with Jx and Jr Hence starting with a

2
Jz we in general find a
given pair of allowed simultaneous eigenvalues J
for the requirement that
whole
series of
allowed simultaneous eigenvalues
...;
2
,
J2
/;-;
J.;
J ,J
+n;
....
(7)
But we have already seen that the values of Jz which may occur in simul
taneous eigenstates with J z are bounded. Hence this series must have a
lowest member JF 2 J\ and a highest member JF 2 J\ From (6) we see that
the corresponding states must satisfy the relations
,
for otherwise the left sides of these equations would
be eigenstates belonging
and J\ -f fi, contrary to hypothesis. Operation by ( Jx + iJv on the
to J\
first of these equations leads to
ft
= IT - (J*? +%J%
from which, since
ifr(yj
Q
Z)
by
Jl)
<A(r/
-o
hypothesis,
J2 -(Jo)2 + jo =
operation by (J x iJy ) on the second equation
(8a )
Similarly,
J3 -(Ji)2-%Ji =
From these
equations
(8)
we
gives
(8b)
find that
^ J\ by hypothesis, this shows that J\ = J\ Now the difference

be a positive integer, or zero, times H. This integer we shall, in
must
Jl~-Jz
since J\
accord with the traditional notation, write as

values 0,
1, f, 2, .... Hence
2j,
where j
is
restricted to the
^=-J,
and from
()
<^=^>
j = j (j + 1) #*.
2
(8)
>
(10)
to the numbers j (j -f 1) S 2 With any

one of these values the set of allowed values of /,,. forming simultaneous
This restricts the allowed values of
eigenstates
is
determined by
JK,
(9)
as
(j-l)H,
0""2),
...,
-JK.
For these allowed values of Jz we shall use the notation mfi, where m3 takes
on the values j, j 1, ..., j. For convenience we shall often omit the
-
subscript^, writing
j; = m#.
(m=j,
j- 1, ...,
Since J^., Jy and J enter the commutation rules

allowed values of Jx and Jy are the same as those of Jz
,
solution of the allowed-values problem.
~j)
(11)
symmetrically, the
This completes the
48
3.
The matrices
33
ANGULAR MOMENTUM
of angular momentum.
Further development of the method used in the preceding section enables

us to find the matrix components of Jx and J y in the representation in which
2
J* and Jz are diagonal. It will he convenient to denote the value of J by
is the normalization
giving the value ofj and of Jz by giving m. The first step
of the
^r
3
occurring in 2 5.
That equation
$(yjml)~N
where
is
the factor
normalized in the sense

1 2 2,
tells
us that
iJy )$(rJm),
(Jx
(1)
necessary to normalize $(yjml) when $(yjm) is

of 1 2 1. Since (Jx iJ^f = (Jx + iJ y we find, using
],
that
iJy ) $(yj m)
N
y]
=N
If this
is
to equal unity,
we must
(Jx
iJy ) $(yj m)
(Jx
K*[j(j+l)-m(ml)].
have, in
(1),
gi5
(2)
an arbitrary real number. This is the arbitrary phase which occurs

in any $ because two states are not distinct unless they are linearly in
dependent, Such an indeterminacy of phase can have no effect on any of the
results of physical significance. For this reason it is permissible for con
venience to make an explicit choice of phase. We shall, therefore, mean by
where
8 is
that set of states connected
by the relation
m)
which
is
obtained from
(1)
and
(2)
by
(j
m + l)$(yjm
setting 8
We may now immediately obtain the matrix

The
general element of the matrix of
seen from
(3)
Similarly, the non-vanishing
(3)
components of Jx and
Jy
x>
to be zero unless y = y ,j =j,

vanishing elements have the values
is
I),
= 0.
m ~m
1.
The non-
components of Jy are found to have the values
(yjm+l\Jy \yjm)= -%ift/(j-m)(j + m+l)
m + 1).
(o)
THE MATRICES OF ANGULAR MOMENTUM
33
49
.if
The
relation
(y j m\Jy \yjm
e~*
(yj
Jx yj m
(6)
__
connecting the components of Jy with those of Jx will be found useful.

Formulas (4) and (5) may by the use of vector notation* be combined into
the one formula
+ 1).
(7)
l\J\yjm) = (i + ij) %fiV(j + m)
commute
which
of
a
of
observables
the
set
here
represents
eigenvalues
y
Let us consider in the y j m scheme the matrix of such an observable,
with
which commutes with J. This matrix will of course be diagonal with
say
respect to j and m, but by the following argument we can show it to be
(jm
(yjm
J".
jK",
entirely independent of m.
We write
K$(y jm)îjj(yjm) (yjm\K\y jm).

7
Now, from
(3)
j^
y
Similarly
K (Jx
iJy )
m)(jm+l)(yjml\K\y jml).
y
These two expressions are equal by assumption; on comparing

we have the result
(yjm\K\y jm) = (yjml\K\y

for all values of
m such that the coefficient
jml)
+ m)
coefficients
(|X,J] = 0)
m+ 1
(8)
does not vanish.
This theorem is of importance in cases where the Hamiltonian commutes

with various angular momenta (see I 7 ).
We may at this point show, in a similar fashion, that the unitary trans
formation between two schemes of eigenstates belonging to jm is entirely
independent of m. Let
</f(o?
m) = S xG8? m) (pjm\ajm).
(9)
/?
We need not here sum over j

iff
(a j m) unless j
and
in x because %(/J / in
is
orthogonal to
= j, m = m. Then
l|ajm-l).
But from
(3)
(10)
we have
-l) = (Jx
using the expansion
(9).
- iJy
On comparing this with
(10),
l).
we
see that
(11)
AKGULAB MOMENTUM
50
33
independent of m. For this reason, in

writing the transformation coefficients joining two such systems of states we
Hence
this transformation is entirely
shall ordinarily
omit the m^ writing merely
(jSjjoy).
PROBLEMS
1. Show that if j = | the matrix representing the component of angular
tion specified by direction cosines ?, m, n is
lîm
n
,
-n
l+im
where the rows and columns are labelled by (j=J,
2.
this
momentum in the direc
w=J) and
0*=J,
Prove by direct application of the allowed-values equation

matrix are + Jft.
72
- J).
that the allowed values of
3. Find the eigen-i/r s for the states in which J = J and the component of angular momentum
J&. How does the eigen-^r for the value + Jft in the
along the direction (I, m, n} has the values
direction (?, TW n) compare with that for - J# in the direction ( - Z, - m, -n)?
4. Write out the matrices for

5.
Jx Jy Jz when j = 1 and
,
Find the matrix for Jx+J^ and verify that
4. Orbital
angular
its
verify that
JxJy - Jv Jx =iKJ z
allowed values are those of JT 2 -Jf;
i.e.
momentum.
Let us now consider orbital angular momentum in the Schrodinger repre

some of the properties of the ^r functions which will be
useful in later work. Introducing spherical polar coordinates r, 6, 9, we
sentation and learn
find that
Lz =
The Schrodinger equation
#|#\ dy
*fe~-
2/o-|=
*dx}
for this observable
(1)
v
89
is
ji>>
2)
v\Y
which has the solution

z
where
VZTT
A(&Lz ) is independent of the coordinate 9. The Schrodinger wave

function must be a single-valued function of position, so
z must be an
of
ft:
integral
multiple
T*
*
t
^
\
,*\
r
&
We shall write
where
is
$(* m
t)
L^mfi.
=
m
<J(mz )
A(<x.
(m^ an
t)
= -~~
VTT
>(%),
e"^
integer)
(3)
(4)
(5)
seen to be normalized in the sense
= S(m X).
z
(6)
ORBITAL
43
AlsTaiTLAB
MOMENTUM
51
The requirement (3) rules out fractional values for tlae z component of
momentum, and hence also excludes fractional values of the
number
Z, where
quantum
orbital angular
= Z(Z4-1)# 2
(7)
I.
since for a given value of the associated values of m^ are
1,
.,
let us find the dependence on 8 of the Schrodinger representative
2
and 9 In terms of polar
^(yZwj) of a simultaneous eigenstate of F,
Z, I
. .
Now
L
Lx and Ly are given by
coordinates the operators for
may readily be calculated by transformation from

We shall write
= (y m^ &(l %)
m[}
as
where B(y lmt )
Since
is
(Z)
is
independent of & and
known from
(5),
Cartesian coordinates.
^^
9. Since
I,
(9)
we have from
33
this gives us the differential equation
30 (II)
rrom which
0(Z Z)
= - 1)
sinz e.
The
coefficient of sin fl has here been chosen in such a

normalized in the sense
10)
that Q(ll) be
way
The phase has been taken as
Now having
1
I) for
^(yll)
= B(yll)Q(ll)^(l)
we may, by repeated applications

In particular, since
(8)
of 3 3
and
(8),
(11)
find the
*fj(y
and we may
l
on the coordinates 8 and

Hence the factor B(ylmt of
act only
B(yll) will recur for each
depend on
convenience in later work.
mj) for all l

the factor
<p
(9)
does not
finally write
(12)
^(yZm;) = J%Z)0(Zm,)O(ro,).
for
s
will
be
normalized
has
Since (10)
been normalized, these
any ml
@ s for the same Z but different l will not in general be orthogonal, since the
.
<J>
s take care of this orthogonality. However, @ s of different I and the

will be orthogonal since ^r(yZm^) is orthogonal to *fj(yl mj) for l^l
l
insofar as a state
in (12)
is
^ s belonging to Z and
is
same
and
concerned, the function B(y 1)
quite arbitrary.
4-2
52
Now
we
to obtain these
find
From
(8)
_
By
s explicitly.
3
repeated applications of 3 3
iteration,
tL y )*e^/W^
Using these relations and the known form of 0(Z Z),
and
(13)
).
(i2
from which, by
a
43
MOMENTUM
we have
-i
-*<
ASTGTJLAK
we find the general formula
in particular
The polynomial
in cos#
P (cos0),
polynomial
which follows the
irrational factor is the
this particular expression for it being
Rodrigues formula.
We can also use (14) to find expressions for
Legendre
known
as
(lm) in terms of the deri
vatives of the Legendre polynomials, the result being
0(Zm) = (-l) 0(?-m).
Hence
(18)
which we have made here leads to a rather

1 only for positive odd values of m. If we
had approached the problem through the usual form of the theory of
The natural
choice of phases
curious occurrence of the factor
For Z=Q,
1, 2, 3,
the explicit forms of the
@s
are
0(1 Q)=Vfcos0
9(2 0) =V|(2cos*0
0(3 0)=V|(2cos
- sin2 0)
- 3cos08inz 0)
= + V^cosflsinfl
3
2
0(3 1) = +Vfi(4cos 0sin0 -sin 0)
0(2
1)
0(2
0(3
=Vif
=
2)
2)
fl
in2
ORBITAL ANGULAR
43
MOMENTUM
spherical harmonics the natural tendency

normalizing factors with omission of the (
From the form of the operators
would have been to choose the

l)
m in these formulas.*
3
together with 3 3
(8)
53
we find the relations
Go
m cote 0(Z m) = -
m-f 1) - jV(Z + m) (l-m + l)
(l
m- 1),
(20)
which are useful in some calculations. We quote here for reference three
other relations which may be obtained as special cases of results derived by
matrix methods in 9 3
:
(21)
Of great importance
in subsequent calculations
is
the relation
known
as
the spherical-harmonic addition theorem, which expresses the Legendre

in terms
polynomials of the angle between the directions (9, 9) and
o>
of spherical harmonics of
(9,
(#",
and
9)
(0*,
9")-
9")
The formula is
(22)
= cos&cos0 + sin0sin#
/
where
coso>
This relation
may be derived
P,(coso>)
cos(9
9").
as follows:
(l
0)
(231
(0),
<D
where 6 and 9 express the direction of r as referred to

a coordinate system with pole along k (see Pig. I 3 ).
Since this is an eigen-?/r of L 2 belonging to I, it may be
co-0
r
K
expanded in terms of the eigenfunctions referred to k

as pole:
(ZO)0 (0):=
(24)
~
"
and
where the JS?Zm will of course depend parametrically on the angles
which relate k to j, A. Now ZyO Ô) ^)^, where Lz is the com
ponent of angular momentum in the direction k But from vector analysis
6"
9"
* This
question of phase choice has caused some confusion with regard to the relative phases of
the matrix components of angular momentum and of electrostatic interaction. See the discussion
by UITOBB and SHOETLEY, Phys. Rev. 42, 167 (1932).
MOMENTUM
AHGtTLAK
54
Hence, from
(24),
+ 0(1 m-l)*(m-l) iV(Z+m)(Z-m+l)

Equating the
coefficient of
(lm)Q>(m)
m] = 0.
siR6"&*"Ei
to zero gives the recurrence relation
= 0.
On comparing this with
where /is
(20),
we
see that if
we choose
Eu to be
arbitrary, then
J?te
Now
in the special case
P,(l)
= 1,*
(16)
^(0,9*)
= 0, we
shows that/(0,
P,(eoso,)
where
0"
see from (24) that
=
9")
J -^-^
!?,=
metric function of #9 and
<9>".
I 3 it is clear
Hence
since
We have then
= 5^= S0(Z m) *(m) &"(lm) $"(m)
= !. But from Fig.
!.
that
0(0",
Pj(coso>)
Interchanging #9 and
0"
9"
),
25 )
must be a sym
in (25) gives
).
(26)
= g(6, 9) for aU pairs of directions.

9
equal only if
This can be true only if g is constant and hence unity everywhere. This
But
(25)
and
(26) are
proves the theorem
g(9"
9")
(22).
PROBLEM
Show from
(1)
and
an expression which
5.
(8)
that
is essentially
Spin angular
the angular part of the Laplace operator.
momentum.
In 1925 Uhlenbeck and Goudsmitf discovered that great simplifications

could be made in the formal description of atomic spectra by assuming that
electrons possess an intrinsic angular momentum whose component in any
direction
is
restricted to the values
known as the hypothesis of
An
essential part of the hypothesis is that the electron also

moment whose component is + e#/2ju,c in a
intrinsic
electron spin.
possesses an
%H. This is
magnetic
1
- 1
m
* This follows
immediately upon expanding the Z derivative of (cos0 1) (cos0+ 1} in Rodriformula by Leibnitz theorem for the differentiation of a product, and then setting cos0= 1.
gues
264 (1926).
f UELENTBECK and GOXTDSMIT, Naturwiss. 13, 953 (1925); Nature, 117,
53
ANGULAR MOMENTUM
component of angular momentum
55
SPIN
direction in
which the
is
p. The
view that the magnetic

opposite signs are natural enough on the simple
moment originates in the classical electrodynamic effect of a rotating
of magnetic moment to
negative charge distribution. However, the ratio
has twice the value
angular momentum (the magneto-mechanical ratio)
motion of the
orbital
the
for
which is given by the classical theory and holds
electron.
is
Pauli* has shown that a quantum-mechanical description of the spin that

the spin as an
adequate for most purposes can be obtained by treating
3
2 3 and 3 but
in
angular-momentum observable of the type considered
with j restricted to the value |. Further developments by Dirac f have shown
to be intimately associated with relativistic effects, but we shall
the
,
spin
Pauli.
usually treat spin in the simpler way introduced by
in
a
electron
the
for
then
We postulate
spin angular momentum S,
the
of
values
The
L.
momentum
proper
dependent of the orbital angular
2
z component, Ss of the spin are restricted to
p; hence S has always the
,
2
value of
proper value | (| + 1) & If we write the proper
the proper values of Sg as m/i, we always have
.
From
334
and
33 5
we
find for the
S 2 as s (# -h 1) ft* and
components of S the Pauli spin matrices
(1)
In vector notation these matrices* are given by
(oci,|S|a;)||
=
set
With the introduction of the spin, the set *, y, z is no longer a complete

be added a fourth
of observables for a single electron, but to these must
momentum would
observable, say S,. (Any component of the spin angular
)
then becomes the function (x y z St
serve as well.) The Schrodinger
^
or for S
letter
the
introduced
Jfi
we
have
Here
as
which we shall write
thus to
when used in the sense of a Schrodinger coordinate; it is convenient
of
the
for
special case
introduce a notation different from the previous s
>f>
PAULI, Zeits. fiir Phys. 43, 601 (1927).

xn.
t DJGRAC, Quantum Mechanics, Chapter
MOMENTUM
56
ANGTJLAB
Schrodinger expansions.
values
|.
The coordinate a
53
of course, restricted to the
is,
An eigenstate of Sg with quantum number m8 depends on the coordinate

a through the factor
W
/
The factors
8(<r,
J)
and
is,
\GH7YIS J
8(cr,
==
*r;r7/z\^/
0(0" j
AM
/O\
^s)
- f form an orthonormal set of functions

)
of a in
the sense that
that this set is complete
this
/m
5\/
"V
and represent
that
s ),
8(<r,
is
^s/
"\
obvious.
-wi
&(VYI
yw
We may then write any function
/^^
iff
as
by the matrix
rxyz if
A r
The
result of operation of
When
S on $ is then given by the multiplication of (I
expressed in this form, |$,^l gives the probability

for
the
electron at xyz with spin + \ l^^-Î 2 the probability
density
finding
by
(4 ).*
is
\.
density at this point with spin
The direct product
of two eigenfunctions now implies, in the Schro
dinger scheme, summation over a as well as integration over x, y, z:
fXV2a $xyza dxdydz.
(5)
<r=-!/
PROBLEM
From
the results of Problem 3, 3 3 write out in the form (4

component of spin in the direction whose direction cosines are
6.
the most general ifj for which the

m, n] has the value + Jft.
(I,
Yector addition of angular momenta.
Consider a system in which
JL
and
two commuting angularcommute

when each com
(Two
of
one
commutes
each
with
of
the
ponent
other, i.e. when the
component
Then
the states of the system can be represented
dyadic [JF19 jy vanishes.)
momentum
are
vectors are said to
vectors.
in terms of the scheme

^>(yji
J 2 î^ 2 )
(1)
which jx and mx are the quantum numbers labelling eigenstates of J| and

J-tel Jz and m 2 eigenstates of J\ and J to and y eigenstates of a set of observables F which commute with J and J2 We wish now to
investigate the
in
question of the addition of these two vectors to obtain states character

ized by proper values of the sum and the z component of the sum. If we can
*
By
the use of two
DARWIX
[Proc.
^r functions, which are to be correlated with our \jsxyzl an *

-*
Roy. Soc. A116, 227 (1927)] discovered equations which are equivalent to those
s formulation however does not
bring out the physical
obtained by use of PauB matrices; Darwin

significance of the two functions.
VECTOR ADDITION OF ANatTLAB MOMENTA
63
learn to
add two
vectors,
57
we can add any number by repeating the
process.
These considerations are of fundamental importance for the theory of atomic

spectra, for they give us the basis of the rigorous quantum-mechanical tran
scription of the older vector-coupling characterizations of the energy states
of atoms.
We introduce, then,
the vector resultant
of our two angular momenta.
The
above are eigenstates of the
states
tf>
set
complete
r
ya
rs
?Ji>J25
t/
e/
l3>
2z
of independent commuting observables. But the set
FJUIJ ,^
2
seen to be also a set of independent commuting observables, and since it

contains as many observables as the above, it should also be complete. An
is
eigenstate of this set
we
denote by
shall
Consider now the subspace characterized by the eigenvalues y, jl9 and j 2

From 2 3 11 we see that this subspace is (2^-f 1) (2J 2 + 1) dimensional, corre
sponding to the number of ways of independently assigning l and m.2 to
.
them
T
mlJlyJl _i
s:
*->>
2=J2>J2-
The question which immediately

(2/x + 1)
(2/ 2 + 1)
values of j,
The method of answering
l >~i
_ JlJ
-JV
this
query
is
ô
it
follows that
are the
s of this subspace?
apparent when
we
realize that
not
the quantum number m,
only the quantum numbers y, j l} and J 2 û^

are common to the two modes of representation.
Jz = Jlz + J^
What
arises is the following:
m which characterize the
From
the fact that
m = m + m 2 Now since /is an angular momen

.
3
in this subsatisfying the commutation rules 1 3, any j which occurs
I
values
of m. The
the
+
corresponding 2j
space must do so accompanied by
tum
m which occurs in a is m =j + j when m

and
largest m occurring in a $ must be j^+Jz
largest
the
occurring must
cj>
also
j l and
be j l +j z There are two
</>
w =j
2
therefore the largest j
with
m =j
+j%
1,
=j l m 2 = j 2 - 1 and m l = jt - 1, m a = J 2 Hence there must be two

.
this value of
m. One of these #
the other must belong to a j =j

belonging to 771=^+^2
2,
Hence
namely
i/r
with
we know to be that going with^=j1 -f-j 2

+J 2 - 1. In general there will be three s
corresponding to the three values of (m1}
</
2 ):
This means that we must have in addition to the values j=*j_

j^j^+jz-l also the value j=Ji+j 2 -2. However, if either^ oxj2 were
ANGULAR MOMENTUM
58
63
we should not have had all three of the states (3), and consequently would
have obtained no new value of j. In general, the least value jl +j2 n of j is
obtained when js -n= j8 where je is the smaller of j^ and j2 This gives
n = 2js and the least value of j as {fa J2 \. From this point on, the de
generacy of each m must be proper to account for the occurrence of the values
.
That with these fa go the proper number (2^ -f
1)(2J 2
+ 1)
of values of
m is
easily verified.
we have two commuting angular momenta whose

2
2
the squared
squared magnitudes are JiO*i+*)^ an(i JÔ a + l)^
sum
of
momenta
can
take on the
the
two
of
the
vector
magnitude
angular
To summarize,
if
"khen
values J ( j 4-
1)
2
,
where the allowed values ofj are given by
(4).
These values
are just those given by the empirical rule used in discussing atomic spectra
by the vector-coupling method. One can represent the result pictorially by
thinking of vectors of lengths j\ and j2 added vectorially, starting with the
which gives the resultant (j^ + J 2 ) and taking all possible values
from this by integers down to \jl j2 for the antiparallel case.
parallel case
differing
Two
momenta will commute only
they refer to quite indepen

dent particles, or to quite independent coordinates of the same particle
for example to two different electrons or to the spin and orbital motion of
the same electron. This means, in the last analysis, that we can split any
angular
eigenfunction of the type
(1)
into a
sum
if
of the type
>t(pj
where J^ operates only on

7.
The matrix
of
^J
(Ji+J2 )
only on
<
(5)
a ),
It will be convenient for future reference to calculate the matrix elements
of JP in the scheme
If we write JT2 in the form
(1)
it
follows immediately from 2 3 2
and
3 3 3 that
-lm
The matrix components we seek
are these coefficients of

according to the
These components are seen to be diagonal with respect to
as required by the fact that J2 commutes with J
z
<f>
definition 2 2 14.
m=m
l -f
2,
(2)
73
59
+ /2 2
We shall need particularly the matrix of L*S= S L, where a has the value
2
and S2 the value ^ (| + 1) % 2 The components are seen to have the
4- 1)
THE MATRIX OF
I (1
ft
(/!
values
},
where
8.
(3)
m=
Matrix
T in the j m
of
scheme. Selection rule on j.
We shall return to the question of the transformation amplitudes for vector

14 3
Before discussing that question

methods the properties in the j
matrix
vestigate by
addition in
angular
it
of the matrix of a vector
momentum,
will be fruitful to in
m scheme, where J is any
T which
satisfies
^ T] = [T,J]= -rx&
mutation rule
the
com
(1)
by saying that T
satisfies this commutation rule with respect to J. This means that the com
ponents of T have the following commutators with respect to the com
ponents of J:
in the notation of
12 2
We
shall indicate this relation
(2)
[J2 ,Tz ]=o
From
12 2 9
we
[/.,rj=r
tf
[J.,Tv]=-ftTx
find that
j-r=r-j.
This commutation rule applies to a large class of vectors:
itself
(a) Any angular momentum satisfies it with respect to
(3)
3
(cf. 1 3).
where each addend commutes with all the other

addends, each of the addends satisfies this commutation rule with respect
(b)
to
If Jr =Jri+j2 +
JT.
Both r and p satisfy this commutation rule with respect to X; this

follows from a simple calculation of [r xp, r] and [r x p, p] by means of
(c)
12 2 7a.
J=zL l JrL 2 +... + Sl +
the coordinate r and the

rule with respect
commutation
any
th
the
S
i
the
of
moment
to /. (L t is the orbital
electron,
spin moment.)
t
of vectors which satisfy this rule with respect
(e) Any linear combination
(d)
Hence
if
momentum p
to
of
S2+...>
electron satisfy this
J will also satisfy

to that of
(/) A calculation similar
it.
I3
shows that the cross product of any

of
number
of
hence
two vectors (and
vectors) which satisfy this rule will
any
Hence any vector formed by addition and cross multiplication

from vectors which satisfy this commutation rule will also satisfy it.
also satisfy it.
ANGULAR MOMENTUM
60
83
which follows immediately from 12 2 6 is that the

scalar product of any two vectors* which satisfy this commutation rule with
respect to J will commute with Jx Jy Jz and hence with J*
Another important
fact
L/MV2YHO.
[/.zvrj-o,
This
(4)
independent of whether 1\ commutes with 2 or not.

shall now consider the problem of obtaining the matrices of Tx ,Ty
is
We
T3
Js and a set A of observables which commute

We shall first obtain a selection rule on j, i.e. a condition
in a representation in which J
with JT are diagonal.
on/ j necessary for the non-vanishing of a matrix component connecting

the states j and /. This we may do by a method outlined by Dirac
(p. 158).
2
Using the relation 12 6, we find that
L/
2
,
7>7-L7,
rj- [r,7]-7=
-(/-rx 3- rx 3-7)
Prom tMs we have

2
2
= - 2s[y ( j x r[J [J rj]
= - 2ifi{ - ZifiJ x(JxT-ifiT),
2
using 12 lla to expand Jx
2
,
Hence
(/x
T").
iH(J
T- TJ )}
2
But
IP, T]] =
y4r _ 27^7^+ TJ*=2W(J*T+ TJ - 4&J(J T).

Z
(5)
Take the matrix component of this equation referring to the states a j m

and a / m where/ ^j. Since7-rconunutes with 7 [by (4)], this component
will vanish for the last term in the
equation. From the rest we obtain:
,
2
0"+
1)
- 2j (j + l)f (/ + 1) + j * (j + 1) 2
The bracket on the

2 |j
(j
(cy
m|
TV f m
left is
UU+ 1) -/ (/ + 1)]
while
+ 1) +/
= (j-f? (j +f +
+ 1)1 = C? +/ + 1) + (j -j
I)
0"
2
)
- 1.
Hence
In order to obtain a
non-vanishing matrix component one of the brackets
vanish. The first cannot since
f =j and;", j Js 0. The second vanishes
must
only when j
-j=
1.
Hence
must have
*
for
a non-vanishing matrix component we
/-.? = 0,
.,
Aa important special
..
1.
instance of this is the square of such a vector.
(6)
MATRIX OF T IN THE j
83
m SCHEME.
SELECTION RULE ON j
61
of a vector of the
scheme are those for which (6) is satisfied. This Is of course a
in thej
type
necessary, not a sufficient, condition for a non- vanishing component.
may obtain a relation which will be useful later by taking a matrix
The only non-vanishing matrix elements of any component
We
component of equation
and a j m
diagonal in
(5)
j,
say that joining the states a j
j(j+l)1P(*jm\T\v?jm ) = (KJm\J\KJm )(*j

since J T is diagonal with respect to j and m and since
m
respect to a
9.
by
(7)
J is diagonal with
hypothesis.
Dependence
of the
matrix
T on m.*
of
be convenient to consider, in place of Tx and T y individually, the

we can obtain those of
observable
x -iTy since from the matrix of
+ 3?~ and ZiTy^^-f. These
Tx and Ty by using the relations 2Tx
relations obtain because Tx and Ty are assumed to be real observables.
we can very easily obtain a selection rule on m. We
For the matrix of
It will
^T
~^
3"
have
\J.,^ =
\J.>T
Hence
TA-^
-iTA = \J.
J9 3T - ^7, = -
W
?m W
/ m matrix component:
^(xjm\r\* 3 m )-(*jm\r\*
or
(m-m +l)(ajra|^|a /m )==0.
= unless w = m-f 1,
Then
j m
(ajm[ 9
Take the a J m; a
"]a
and
(ajm|
.ffewce
t
e^"
|oc
unless
m = ml.
Tx
or
Tv
are those for
m = ml.
Tz commutes with Jz
the only non-vanishing

f
now obtain the dependence

Jx iJy by /, we have
shall
denote
(la)
components of Tz are those
m = m.
for which
We
&e only non-vanishing matrix component of
which
Since
/m
of the matrix of
3"
lb)
on m. If we
-^
[/,^=fo-jy ,rx
The matrix oi/ is known. From 3 3 3 the only non-vanishing components are
m
is essentially as in DIBAC, p. 158. The method of the

derivation of the selection rule on
29 of BORN and JOBDAN, Ekmentare Quantenmeehamk.
-j + 1 is necessarily all that leads to non- vanishing matrix com
$ Although the range
is true for the full
ponents, it is important to note that this equation, as well as all the following,
The
rest of the section follows
j^m^
range j&zm^s -j.

range,
We
shall later
wish to take some sums over
m and shall
sum over
this latter
62
ANGTJLAB,
MOMENTUM
93
Take the most general matrix component which
the selection
satisfies
rules of the equation
This
is
\a!
r
j
or,
from
m+l)
m)(!j m\/\*
m+l)
-j = 0,
(j
1,
(2) 3
f-m).
-j = 0,
tf
1)
(3)
If/ = j, we have
and
since this holds for
any value of m we
see
that each ratio must be independent of m.

(y.j\T\a! j) 9
shall denote this ratio
by
we
the
independent of m. Hence
of the elements of
diagonal in j:
this quantity being
dependence on
=ji
Multiply through by V(J
wi)/(J-f- wi)
(4)
l).
= (aj m- 1
yaj
\
and rewrite as
\
Again each ratio is independent of m and hence is

Theii
find, for
3"
m\ST a j m+ 1) = (aji T:K j)V(j-m)(j + m +

(3) becomes
(ocj
If j
from the form of the relation
We
OC /_!.
-
CX
~"~
-1 j
set equal to (a j: T\a! j
j-l m+l) = (a^Tja j-l)^- w) tf- w- 1).
If f=j+-1,
(3)
1).
(5)
becomes
V(j+ m+l)/(J-m-f 1) and rewrite:

(qjml^lay+lm+I) ^(aj m-l\r\* j+l_m) = _
V (j+f+2)(j+m+l) V0*+m+l)(j + m)
Multiply by
f
..^.^
"^^
J?:
Hence
).
(6)
We shall now determine the dependence of the matrix of Tz on m. We have

that
is,
27^
DEPENDENCE OF THE MATRIX OF f ON
9s
Now the only non-vanishing matrix elements of /^

Since
63
are
-j-fl)
of
matrix
the
determine
can
the matrices of ^/ and HT we

selection rule
(7). We have from (Ib) the
we know
directly from
m = m.
Then
2% (&j
- (aj: T:aL
j)
or
(ajm|2
Also, since
>
Tz is real, we see that
=j
we have from
1,
jm)=m(ocj:T :oc
/
(aj:T:a
jf)
(8)
j).
= (a jir:a ;),
t
(7):
(ajm|I 2 |a j-lm)==Vf--m (aj:3 :a j--l).

one obtains in an analogous way from (7)
or
/ =J+ 1
(9)
(10)
Since 2^
is
real
we
see
by comparison of this with
(9)
that
matrix (ajjTja /) as we have defined it is Hermitian. This fact

rt For
instance, from (5) we have
enables us at once to obtain the matrix of J
Hence
the
The other components of ff* are found in a similar way.

table of the non-vanishing
Collecting all our results, we have the following
matrix components of
*(*3\T\* 3+^^
(ajm| r|a j
(ajmj T\a?j-
1)
=
=
=
1)
(aLJ\T\!j)
(v.j\T\u!
r
(a j: T\*
%V(j + m)(jm+l)
(i
ij)
j)mk
j-1) $V(j + m)(j + m-l)
*.
(i
ij)
(11)
ANGULAB MOMENTUM
if TX9 Ty and Tz commute
64
From
its definition,
9*
with.
the matrix
with A it is diagonal in a. These

/) is diagonal in j; if they commute
remarks hold in particular for J Jy and Ja9 for which
(ccjiTia
x>
(aj:J:a /)
= 8^8 aa.,
3
so that in this case the formulas (11) agree with 3 7.
3
From the above formulas and from 3 11, we see that in going from the
scheme
ocjra to
a scheme jSjra, the matrix \\(*j\T\<x, j

the matrix of an observable:
)\\
is
transformed by
like
jj(aj|j3j)H, just
act
the proof
(A?|cy) s (j8jm|ajm) justified by
in every
that
we
see
Hence
m.
of
independent
Here we have used the notation

in
that this element
respect IKccjjTja j
)!!
is
behaves like the matrix of a real observable.
The matrices of Jl and JFa where J + J2 = J.*

If J is the sum of two commuting angular momenta JI and JF2 we can
10.
obtain the complete matrices of J^ and J%. Consider the representation

is the total
where the observables T commute with x
2 J. (If
7Ji J2 J
=
=
and
momentum
2
angular momentum of an atom ; x S, its total spin
its total orbital momentum, some of the y s would naturally represent the
/ J
,
>
Then the matrices ofJi and /2

not depend on y, we shall not
will
Since
and
will be
they
J2
^
and j 2
carry the label y. For brevity we shall later drop also the labels j l
s
9 ! 1 gives us the dependence of these matrices on m; hence we need now
determine only the (j\ j 2 jl
j2 j ) and (jx j2 j: J2 :jx j 2 /). For the case/ =j
these follow easily from 83 7 which becomes for the z component ofjt
5
^s,
and 5
s of the individual electrons.)
diagonal in y,
J^
or,
from 9s ! 1
j(3
But
+ l )&(M*3\ Jti3iM) = (:Ji323^

Jl = (J-
JiJ
or
Using this
relation, (1) gives
and
the corresponding element of
terchanging subscripts 1
*
and
J2 is
(2a)
obtained by in
2.
10 and 11 follow GUTTOTGER and PAULI, Zeits. fur Pliya. 67, 743 (1931),
THE MATRICES OF Jx AND /2 WHERE Ji+J* = J
103
65
The components not diagonal

relation
[Jlx _;jlv ,
or
in j are
We
more complicated.
use the
JJ= -tft/lv -t(
(J-ix-iJiv) J-iz-J-iz (îx
3
Take tOnefajijm; 3iJz3 mJf component of this equation, using 9 4
^-
(omitting the labels j ls j 2 )
to 9 3 10
-Jilj-lW-m* (3- l\Ji\})^(j

=*
( 31
Jil ft/ (3 -m)(j
OVJJ)*- 10 Jil
or
Now, from
(2a)
Hence, with the abbreviation
^ JiO\+ 1 )-JaO*+ 1
SB!
we
)>
obtain the relation
ioviu--i)m^
We get another relation in a similar fashion by taking the j x j j m

2
component of the identity
Jl= ttJ
This becomes
or,
from
(2a),
and with the abbreviation
3)
j J2 j m
66
ANGULAR MOMENTUM
We
can solve
former
we
(3)
and
(4) for
lOViL/-l)I
103
and lOI/Jj-fl)} 2 For the

.
obtain
NOW
hence
:
Iff J11;
jIU:
"
~^ +
2
J*
>|
(Ji
+ J2 + *) 2f - O l ~J2) 2 O i
-
;|
8
Since |OViJj + *)i 2 =iO"+ 1|
Ay)|
by substitution of j 4- i for J.
uantity
tllis q.
may be
obtained from
(5)
We
have now arrived at the point where we should choose the relative
phases of the states of different j. This we do conveniently if we take
i) ^ be real and positive for all j. This means that the non-diagonal
(fi.Jilj
elements of J^ are all real and positive. Since the matrix of
+ 2 must
be diagonal with respect to j, we have
=
(j\JJ 1)
(j\Ji\j- 1). Hence if
the non-diagonal elements of J^ are all positive, those of
are all
J= J J
J&
An alternative
choice of phase would reverse these

signs
the roles of J^ and JF2
therefore write
.
negative.
reverse
and hence
We
~h) (
(2b)*
={?}
Âe tsame ex-pression.
In the following pages we must be consistent in

choosing everywhere either
the top or bottom sign wherever the
symbols { } and { + } occur. We shall
carry both signs as a matter of convenience in the sections immediately
following. However, elsewhere we, shall always use (2b) with the upper sign.
In adding J^ andJ2 we choose the relative phases of states
of different j so that the
non-diagonal elements of J^ are positive, of J^ negative
* With the
abbreviations
which ve
=y
~-
+j2 )
(j
+jt +j2 + 1)
QU) =0\ +J 2 -i)
(i
P(j)
}\
shall use in the nert section, this
formula becomes
(2V)
It
is
to be noted that
P is a function only of
and j +jt while
Q is a function
only of j x and j -j, .
TB[E MATRICES OF /x
10
this point we
phases of the states
At
of 6
1.
may
AND J3J WHERE Ji+ J2 =J
also conveniently
</>(yJi
J2 J
make
a definite choice of the
of 632 rotative to the states
67
These phases are completely specified by
<j>(yji3*
mm
i
for the state of largest j and

phases and the following specification
(namely j=m
= fa +fa)
z)
the above choice of relative
P which commutes with jFla

We shall now consider the matrix of a vector P which commutes with J
but which satisfies 8 3 1 with respect to J, where /= Ji H- J Because of this
11.
Matrix
a vector
of
latter relation
we
see that
P must satisfy
8 s ! with respect to 2 also. The

shall proceed to find its depend
3
dependence of P on m is given by 9 1 1 we
ence on j in the yfa fa j m scheme. (If /j. = S, P may be eSr* the total electric
th
moment, or may be Li9 the orbital momentum of the i electron, or r$ or
th
pi9 the coordinate or momentum of the i electron, etc.) The matrix of P
will necessarily be diagonal in fa, so for convenience we may omit the label
for the moment the y / which will occur in all matrix
fa We shall also omit
in y). Because F
components of P (/,. and JT2 are necessarily diagonal
1 or
3
selection
rulej 2 -fa=
satisfies 8 1 with respect to J2 we have the
;
>
zero.
Consider the yfafaj m] y
A Ja/ m+1 component of the equation
(Jto-iJ^P^Pttfte-MiJ3
3
First for f=j- 1,/=J+ 1 (from 9 6 and 9 10):
t-M)
or
Since the matrix of
(fJ^^^
= (J*J-^
known, this furnishes a recursion formula for
is
Now consider the case / = j, =j+l.

1 0*-)
0*2 jVi&J) O t JiJl/2 J+
Here we obtain
j"
a^
independent of m, the
Since this relation
is
must each vanish,
i.e.
[-
& jVitfi
fl
coefficients of
+ Ofo"+ I îyi J + !)] (Jt Ji^Ji J + J
and of unity
5-2
ANGULAB MOMENTUM
68
Subtraction of the second of these from the
first
gives
3\3\P\h3)
In the same way we obtain from
(1) for
the case j =j,
j"
= 0.
=j-
(3)
the two
equations
M& JSubtraction of the
1)
. j-il-PI/J-i)
- O ^Vib ; J-
of these from the second gives
first
[- (j.jV^JW+OU-ilîi^J- 1 )
0"-
If we can determine (j z j\P\j2 j+l) from
(2), (3)
will give
us (hJ\P\jJ),
and (4), (j2 j\P\JzJ-l).

For the case j2 =jz (2) gives us
,
We see that these ratios must be independent ofj, as implied by the notation
(M^lh)- Henee
where the last factor
From
(3)
we
is
3
given by 10 2b.
obtain, using
(5),
- l(JJ\Ji\hJ) (3 + 1) - OW+ 1 i& J + !)

!
or,
+ 2)]
iiÛ.)
= U*flP\J*J)>
from 103 2a
Similarly,
from
(4),
or
by taking the complex conjugate of (5), we have

(7 )
J- 1 ) = ~ 0*P&) O l jVilJi J-l)-
&&P&
For the case ^2=^2
2 ) becomes, on substitution from 103 2b
- 1 (j, j;P=j a -lj
MATRIX OF A VECTOR P WHICH COMMUTES WITH/!
11 s
THs gives
us
(j 2
j\P\j^-l J-H).
Knowing
this, (3)
and
(4)
69
give us
?) are essentially the complex conjugates of these quan

tities.
Using for conciseness the abbreviations*

)
we find the following complete table

the matrix of a vector
(3
+ Ji +h + 1
>
of formulas for the dependence on j of
P which commutes with Jls where J
4- JT2
=/
(8)
When we seek the corresponding components for a vector Q which com

mutes with /2 since the only essential distinction between J^ and Jz is the
,
3
we find that the dependence onjofQ is
opposite choice of signs in 10 2b,
and the double signs inverted. Since JTa
given by (8) with j^ andj% interchanged
3
of the type Q, the formulas 10 2 are special
is itself of the type P, and
cases of
*
(8).
We note tfcat P(j) and
since Ijj.-fa
Q(j] are essentially positive quantities
takes on negative values in certain instances.
AKGULAB MOMENTUM
70
The question
now arises.
of obtaining further information
11 s
about the factors
P satisfies the commutation rule 83 1 with respect to JF2 and
Since
commutes with J^ we can express directly from 9 3 i 1 the dependence on m 2

,
P in the yJ j m m scheme. This matrix will of course be

in
m and by 33 8 the elements will be entirely independent
and
diagonal
j^
of m Further, a vector having the properties of P will usually, if not always,
of the matrix of
be of such a character that when a resolution of the type 63 5

states in question the vector wiH operate only on 2 not on
,
made of the
is
<
In
this case
the matrix elements will be independent also of the value of jx This will be
true in all cases which we shall experience; hence we shall here treat these
.
elements as independent ofj l
We shall
for example^
have from the
first
formula of 93 11:
(9)
There occur here on the right the factors (yJiJjf.P\ y JiJ$) = (yj&Ply fz) for
which in accordance with 9 3 1 1 we use exactly the same notation as we have
adopted for the factors occurring in (8). That these factors actually are equal
is shown very
simply by a calculation of (yJiJ^^m^Pl^ j^j^m^m^) from
3
li and of (yjj J 2 jmjPiy A j 2 j m ) from 93 11 and (8) for the case
<
in
=sm
which we know from 103 6 that these two matrix components are equal.
Knowing that the factor (yJ2\P\y j*) of (8) Jim the same properties with respect
to the vector
repeat in
factor.
J2 as the
(ajI-Tja
3
/) of 9 11 has
with, respect to
J enables
us
to
10 3 or II 3 with respect to this

can determine this factor completely
cases the considerations of
many
In particular if /2 = P-f P we
,
J2 = J2 -f J
and P commutes with J^ formulas

of the type (8) are again applicable to determine the dependence of the
matrix of P on J2 and the present discussion may be repeated. We shall find
by the formulas
103 2; or if
many
12.
applications for such calculations.
Matrix
of P-Q.*
It will be convenient at this point to obtain, in the

yjifajm scheme, the
matrix of the scalar product of a vector P which commutes with
J^ and a
vector
which commutes with J2 both of which satisfy the commutation
rules 8 3 1 with respect to J. From 8 3 4 we see that this
will be
product
P-Q
JOHNSON, Phys. Bev. 38, 1635 (1931).
MATRIX OF P-Q
12 s
71
diagonal with respect to j and m. The elements of P*Q are easily shown from
9 3 11 to be given by the expression
(yjij 2 jwlPiy"jj^^^
"?m
A j. JIÎ/Ji j; j+l) (/Ji/2 J+11CI//1/2 J) + 1) (*f + 8)

j 2 jjpj/ ji j; j)
+ S(yjlj 2 j:^y^
y
When the formulas
iS Cy Jai-fV J2 (y^ iiCI/ Ji)

)
v*
_
j; j; j) j
dependence on j are substituted here
U + 1) ~h Wi + 1) - Js (A +
js -4-
y"
A- 1 ;J m
S (rs^lfh] (y* Ale!/ A!^!/ J4 -i)
(y* Alfll/ Ji-l)VOi
y*
(yj i A j
\P*Q\y ji+l A-i j m)
o*+
(1)
11 3 8 giving the
we obtain
=
(/ ji j; jiQjy
1)
1)>
0"
+A ^* + 1) (A +ja +j) (h +5* - j) OWi -J - 1)
J2 (yA!P|y*A-i) (y*Ji!!yVi+i)Vc/+A -A - 1) O +A -jJ
0>A
-A+ 1)
+A -A +2).
(2)
The only other non-vanishing elements
are those obtained
by taking the
complex conjugates of the last four above. As in the previous formulas we

may interchange P and Q j1 and j2 if we invert the double signs. (Compare
the second and third elements.)
y
Sum rules.
13.
There are certain sums of scalar products and absolute squares of

matrix elements of T and P which will be of interest to us in connection
with intensities of spectral radiation.
From
9 3 I1
we
easily find (the calculation is similar to that
used in
3
obtaining 12 1) that
2
m*
(ajm|T|aT<>
(*
? m*\T\* j m
=$
)
1
8mm- (ayiS !*
j")
/!2V j) S( j, j
),
(i)
where
(2)
ANGTJLAB
72
13 3
MOMENTUM
In particular, then
mf
These sums are independent of the m value of the state ocj m occurring on
left. They are not symmetric in j and /, but become symmetric if
multiplied by (2j + 1) to accomplish the summation over m.
7 4 be
It will be convenient to obtain now a relation which will in
the
interpreted as indicating that the total radiation from

and unpolarized. Using the relation
m
we readily find from
an atom
is
isotropic
Sm
j
9 11 that
2
TO
771
= S
2
=
|(ajm|r|a /m-l)p |(2j + l)|(^jT:a /)| S(j
For the case of the vector

over the j values of the
F of
final state,
).
should like to go on and

obtaining the values of
II 3 ,
we
(3)
sum
(1)
(4)
1. The straightforward evaluation of these quan

l,J2, andJ2
substitution
11 3 8 is algebraically very complicated. This
from
by
unfriendly complication may be avoided by a transformation to the
tities
scheme as follows:
=
(4)
JT $mm
S
5* m" ?ni
(y Vi
S
mm
t
(y Ji J 2 j
m |y^ J
rVJi
JD
(y
l.
, * t
lootnote, p. 49.
mi m a
A^
The sum over fin" may be performed at once

(4)=S,Smm
mi T
|P|y
to give a factor
13 3
SUM RULES
From the discussion at

sum is given by (1) as
The factor remaining
which
is
the end of
11 s
73
we know that the value of the
last
then
unity since the values of these transformation coefficients are

independent of y. Finally then, (4) becomes
is
and in particular
j
Here/
S Kw i J jl P Ir A J2 j m
m
summed
is
over the values j
pendent not only of the

the
2
)|
l9
=
j
|(y;,iP!y #)| S(J..
and j-\-
1.
(5
These sums are inde
m value, but of the j value of the state occurring on
left.
Transformation amplitudes for vector addition.

To complete the discussion of the addition of two angular-momentum
vectors we need to know how to express the eigenstates (6 3 2)
14.
in terms of the eigenstates (6 3 1)
As indicated, one does not need in this discussion to write the yji j z explicitly,
since they do not occur as summation indices in the transformation. More
over, since
it is
for
(?%
m 7^ m
1 4-
an eigenstate of Js =Jlz + J^ with eigenvalue m-^ 4- m2

any eigenstate of Jz with another eigenvalue, e.g. $(jm)
2 ) is
orthogonal to
Hence, in the relation

ift(jm)
=
</>(?%
2)
(n^rn^jm),
(I)
+m
the coefficients (m-L m^ j m) will contain a factor (?%

2 m).
One way to obtain a transformation between these two schemes
(
diagonalize the matrix obtained in
7 3 of
2
J"
in the
m^m^
is
to
scheme. This
method does not, however, give us a set of states $(j m) with the particular
3 3 and 10 3 to use; hence none of the matrix
phases we have agreed in
elements we have calculated would be valid in such a scheme. We can,
3 3 and 10 3 themselves, obtain a set of
though, by using the formulas of
14 s
ANGULAR MOMENTUM
74
recursion formulas which completely determine the transformation coeffi

cients with the proper phases.
state $(Ji+hJi+Js) of highest j
For the
vamsMng
Knowing the
j
and m,
state $(jm),
i.e. all
3 3 gives us the state
SIT
I
and
there
is
only one non-
S
coefficient, namely (el 10 6)
Jl
the coefficients
$(jm- 1). Expand
(m^m^jm) for the given

3 3 3 in the
mi m 2 scheme:
fa
77ll77%2
Since
/ .-iJ *(J -iJly + (Jto -iJ 2y ), we

a
laf
ir
obtain from a âpplication
3 3 3:
of
itf0
or,
equating coefficients o
j w)
-fV(j2 H-m 2 -hl)(j 2 -m 2 ) (w^ma+llJm)*

This
is
(3)
the desired recursion relation.
From
value of j.
9s
and 10s we obtain the recursion formula which steps down the
From
9a ll
we find
that
"
- m)
j -f m j ( j-l|
This becomes, on expanding and equating coefficients of ^(m^m^ as before.
V (j - m) (J 4- m) j- 1: JJ j) (w^ m a j- 1 m) = ])% - m O J^y )] (Wj. m 2 j m)

(
),
I03
(4)
where the (/i/jyl are given by

2. This relation
gives us the state ^( j- 1 m)
when we know the states $(jm) and ^(j+1 w).
With the initial condition (2), the two recursion formulas (3) and (4)
determine completely all transformation coefficients for the addition of a
given jj and j z A general formula for the transformation coefficients is very
difficult to obtain from these relations. The general solution has however
.
14 3
TBANSFOKMATIOK AMPLITUDES FOB VECTOR ADDITION
75
been given by Wigner* by the use of group-theoretical methods as the

following expression:
(j
+ A + Ja + ! Oi )
In this summation K takes on
all
integral values consistent with the factorial
notation, the factorial of a negative number being meaningless.

This formula is so complex that the calculation of coefficients by its use
as tedious as the direct use of the recursion formulas.
^ = 0,
which show the result of adding any j\ to

convenient
way
trivial case j 2
is
But by making tables
|, 1,
f,
2, ,.,,
we
obtain a
of evaluating these transformation coefficients. For the
= 0, we have
Oi
i-e.
0(ji
Ji wii)
= S (J
=
<(
>
Ji)
(>
%)>
ah 0).
(6)
In Tables I to 43 are given the values forJ2 =f, i, f, and 2, which are
sufficient for most cases in atomic spectra: one seldom wishes to add two
3
angular momenta the lesser of which is greater than two.f

In the addition of two angular momenta ja and jb we shall wish to
know
the effect of the reversal of the association of ja and jb with the

and j2 of
the above formulas. If we associate ja with j I jb with j 2 we obtain a state
$(ja jb j m) as a certain linear combination of the states
a
b ). When.
a jb
we associate jb with Jx ja with J2 we obtain a state $( jb ja jm) as a certain
,
m m
<f>(j
e Assume the basic states

a mb ma)
and $(jb ja mb ma ) to be identical (cf 63 5) The states $(ja jb j m)
and $(jb ja j m) must be essentially equal; but that they will have different
e non-diagonal
phases is seen from the fact that for the ^(JaJbJ^)
linear combination of the states
<f>(jf)j
#0 h ma m
a
b)
elements of the matrix of
J^ will be positive while for the ^(JbJaJ^

these elements will be negative (cf. 10 3 ). Hence, for successive values of j,
these states must change their relative phase However, for j^ja+jb the
two states are equal since for
a jbjajb)>
a +jb they both equal
m~j
WIGNEE, Gmppentheorie,
t These tables
<j>(j
p. 206.
from the recursion formulas alone or from (5) alone. Formulas

determine immediately tlte element
the upper left corner. Repeated application of
(3) gives then the top row; from this (4) determines directly the other rows. The bottom row is
obtained very simply from (5), the summation in which reduces to one term for this case. For the
top row, however, this sum becomes very complicated. The most convenient method of com
putation combines the use of (5) and of the recursion formulas, especially of (4) for the passage
from one row to the next.
(2)
and
(3)
may be
calculated
14s
MOMENTUM
76
TABLE
I3.
TABLE 2 3
rn^ji
%j m)
5712=
(2jl
+ l)(2j1 + 2)
y<
v
V
2^ r2f
2^(2^
TABLES 3
__!
Vî
\ + TO - i)( Ji + TO + iX Jl +
v/
<:
/3( Jl + MI
- ^)( J\ TOT + |)( Jj - W 4- f }
2.7!
+ 3)
(2^ + 1X2^ + 3
Jl-i
yi-ffl-*X./i-+*Xj i--+f)
2/i(2Ji-lX2A + l)
7712=
/3( Jj +
Jl+t
Jl-i
Jl-f
m + f )(/i
-0\-3*-i
-v
*
^^
From Wismer.
+m -
/3Q"i+W"i1)(
A - m - *)0*i - w + 1)
-f
OT
)(
A -m +f
v/^
- m + f)
+fX/i - W (2^-1X2^+1X2^
2^(2^-1X2^
t Calculated bv F.
Seitz.
143
TRANSFORMATION AMPLITUDES FOE VECTOR ADDITION
77
c5
78
AKGTTLAB
MOMENTUM
the states for other values of
m are given directly in terms of this by 3
Hence the
must
states in question
3.
satisfy the relation*

a )b 3
(?)
PROBLEM
Investigate the relation of the preceding formulas to the classical vector-coupling picture.
Show that if the spin of the electron is known to have the value + in a direction making the
angle 6 with the 2 axis, the probability that the z component of spin nave the value +J is cos 2 (0/2)
and that it have the value - J is sin*(0/2). [PAUU, Zeits. fur Phys. 43, 601 (1927).]
(a)
the spin is parallel to I as a vector

(6) According to the vector-coupling picture when/ =Z +
and when j has the component m the angle it makes with the z axis is
cosd~m/(Z + }). Compute
eos 2 (#/2) and sin2 ($/2) for this angle and compare with the square of the coefficients in the first row
of Table P.
For the case j =1 - J one has still to use coa9=m/(l + }) instead of m/(l - J) as suggested
by the
vector-coupling picture. This is an example of the difference between quantum-mechanical
formulas and those given by classical methods, although,
consistently with the correspondence
4
principle ( 4 ), the two agree for large values of I.
* This
agrees with the relation
O o J6 m a
given by Wigner.
b\3a Jb j
=( -
l)
^0
ja
CHAPTER
IV

"The
whole subject of
electrical radiation
seems working
OLIVER LODGE,
itself
out
Phil. Mag.,
splendidly."
August, 1888.
The theory of atomic
spectra divides itself rather sharply into two parts

the theory of the energy levels and the corresponding states, and the
theory
of the radiative process whereby the spectral lines arise
transitions
through
The preceding chapters have provided us with most of the

needed
for an attack on the first part of the
apparatus
problem. In this
between
chapter
states.
we
present the general theory underlying the radiative process.

is to be treated
according to the general scheme of Max
That radiation
well
is
electromagnetic theory is unquestioned by physicists to-day, but it

also recognized that these laws require some kind of
quantum-theoretical
modification, to be embraced in a general theory of quantum electro

dynamics. Such a theory of quantum electrodynamics cannot be said to
exist in definitive
form to-day. When it is developed it will probably produce
alterations in the Hamiltonian
which will change the details of the theory

of atomic energy levels as we present it. But the general structure of
the theory of the energy levels can hardly be affected since it possesses so
points of close contact with the experimental data. Likewise it is
possible at present to develop the theory of radiation along general lines
with considerable assurance that such alterations as are later brought by a
many
revision of the theory of the electromagnetic field will not affect our
conclusions.
1.
main
Transition probabilities.
Emission of light takes place when there is a transition of an atom from

a state of higher to a state of lower energy a quantum jump. Likewise
absorption takes place by an upward transition that is caused by the action
of the fields of the radiation on the atom. We start the consideration of these
processes
by a somewhat phenomenological method due
to Einstein.*
The
independent of special views concerning the electrodynamics of

the transition. Let an atom be in an excited level A, that is, in a level of
argument
is
more than the minimum energy. Einstein then ascribes to it a certain pro
bability per unit time, A(A, JJ), of making a spontaneous transition with
emission of radiation to each level B of lower energy. The light sent out in
the process has the wave-number (EA Es )/hc according to Bohr s rule.
*
EINSTEIN, Phys. Zeits. 18, 121 (1917).
THE THEOBY OF RADIATION
80
Then
if
the
number of atoms
I4
A is N(A),
In the level
due to spontaneous
emission processes
(1)
where the summation

defined
is
is
over
all states
of lower energy.
y^ = B A
by
mean life of the level A,
called the
(4,
The quantity r(A)
B)
(2)
for if the only processes occurring are
those of spontaneous emission
N(A)^N
from which
it
follows that r(A)
is
(A)e-
the
Â\
(3)
mean time which an atom remains
in the excited leveL
Einstein also ascribes to the
atom two probability
senting the effectiveness of the radiation
field in
coefficients repre
causing transitions. The
supposed to be isotropic and unpolarized and to have

spectral energy p(a) du in unit volume in the wave-number range da at a.
If C is an energy level higher than A (where A is now not necessarily an
radiation field
is
excited level), then the field produces transitions from

at the rate
A to C by absorption
where a is the wave number corresponding to the transition, by Bohr s rule.

Also the radiation stimulates or induces emission processes from C to A at
a rate given by
N C) B(Gy A)p(a).
^
The coefficients A and B are fundamental measures of the interaction of the

atom with the radiation field and have the same values whether there is
thermal equilibrium or not.
purely statistical considerations we can find relations between the
and B s, for we know that in thermal equilibrium
By
As
(a) the relative numbers of atoms in

Maxwell-Boltzmann law,
different levels are given
N(A)=g(A)erW,
where g(A)
(b)
is
the statistical weight of the level
the radiation density
is
given by Planck
(4)
A
s
by the
formula,
The argument now uses the principle of detailed balancing, according to

which in statistical equilibrium the rates of each elementary process and its
inverse are equal. The two kinds of emission are regarded together as one
elementary process and the one kind of absorption as its inverse. This
TRANSITION PROBABILITIES
I4
81
seemed strange until Dime s radiation theory* showed that the two kinds
of emission were really related to the same process. Hence in equilibrium
we have
N(A)B(A,C) P (a) = N(C)[A(C,A) + B(C,A) P ()].
(6)
Tn order that the last three equations be consistent we must have
g(A)B(A,C)=g(C)B(C A)
A(C A) = Birhca*B(C,A).
9
Using the latter relation the total rate of the emission processes can be
written in the form
N(C)A(C,A)[n(a) + ll
That
where
=
n(<r)
p(<r)lBirkco*.
(8)
way of writing it is shown by the Dirac radiation

a quantum -mechanical outgrowth of the method of Ray-
this is a significant
theory, which
is
leigh and Jeansf and DebyeJ for handling the statistical mechanics of
the radiation field. The number of degrees of freedom of the field asso
ciated with waves in the wave-number range do- at or is shown to be S-rra 2 da
per unit volume; the size of quantum associated with each such degree of
freedom is kcu, therefore n(a) is the mean number of quanta per degree of
freedom for the field coordinates belonging to waves of wave -number or.
In the Dirac theory the electromagnetic field and the emitting or absorb
ing matter are treated as a single dynamical system. Because of the rather
weak coupling between the two parts, field and matter, it is possible to make
fairly precise statements concerning the
amount of energy
in the field
and
matter. (If we regard two separate systems as

one system, the two Hamiltonians commute with one another and so each
commutes with their sum, therefore eigenvalues of each are constants of the
the
amount of energy in the
motion. But if a small inter action is present, this is only approximately true.)
The emission process consists of a change in the system whereby energy
leaves the matter
and
raises the
quantum number of one of the
field s
degrees of freedom by one. The absorption process is a change in which one

of the field s degrees of freedom has its quantum number diminished by one,
the energy going into the matter.
Without going into
details
we can say
that the probability of these in

proportional to the square of the matrix com
ternal changes in unit time is

ponent of the interaction energy between the field
and the matter. The
matrix component is taken in a representation in which the states of the two

to use is the one
parts are labelled by quantum numbers; the component
the initial state and whose final indices specify
the final state. The whole transition probability will be very small unless the
whose
initial indices specify
* DERAC, Quantum Mechanics,

Chapter xi; see also FERMI, Rev. Mod. Phys, 4, 87 (1932).
f JEANS, Dynamical Theory of Oases, Chapter xvi; Phil. Mag. 20, 953 (1910).
f DEBYE, Ann. der Phys. 33, 1427 (1910).
cs
THE THEORY OF BABIATION
82
initial
and
final states of
I4
the whole system have the same (or almost the
same) energy.
Each of the coordinates for a wave of wave-number a is dynamically
=
equivalent to a harmonic oscillator of the frequency v ccr. Hence the possible
field energies associated with each wave are hv (n -f |), where n is an integer.
The interaction of the field with the matter is a linear function of the
wave
amplitudes, so the matrix components, so far as the change in any

field-wave s quantum number is concerned, contain as a factor (n q\ri),
where n and n are the initial and final quantum numbers of the degree of
freedom in question and q is the numerical value of its amplitude. Since the
field coordinates are dynamically like harmonic oscillators, we may take for
the matrix components of q the well-known values
=
,
(Vn.
(n
_
72,
!)
As increase in n corresponds to increase in field energy,

or emission by the matter, and decrease in n corresponds to absorption, the
lack of symmetry is just what we need to get the spontaneous emission as
All others are zero.
+ l)J 2
is proportional to n+l, so emission by

of
of
this
freedom
can occur even in the absence of
to
passage
energy
degree
radiation (^ = 0) in it. But |(n|g|n l)j 2 is proportional to n, so absorption
well as the induced:
j(7i|gj%
by passage of energy from this wave to the matter is simply proportional

to the amount of energy already present in that wave in excess of the zero
point energy, ffiv. The actual emission in the statistical ensemble will there
fore be proportional to \n(a) -f 1] as in (8), where n(d) is the mean number of
all of the degrees of freedom of the right frequency.
This is just a sketch of the theory, given in the hope that it will make clear
the way in which the two apparently different kinds of emission are really
two parts of a single process. We conclude this outline with the remark that
quanta in
not quite true that the only transitions occurring are those in which the
just opposite to that of the field. This gives rise
to the natural breadth of spectral lines, the quantum-mechanical analogue
of the fact in classical theory that a radiating oscillator has its amplitude
it is
atom s change in energy is
damped out by the loss of energy by radiation. Thus the wave-train emitted
is finite and not monochromatic when
expressed as a sum
in a single process
of monochromatic waves with a Fourier integral. This point has been

by Weisskopf and Wigner.* They find that the relative
probability of emission in the range da at a, where a is now measured from
studied in detail
the wave-number given by Bohr
s rule, is
~7T7/T2>
2
I -f
)
TT
where
o-
(cr/<7
* WEISSKOPF and
WIGNEE, Zeits.
-yrv + -TTrJ
4xTC\r(A)
r(C)J
fiir
Phys. 63, 54 (1930).
(
^
10 /)
TKANSITION PROBABILITIES
83
the distribution of energy over the line and relates the half-width,
This
O to the harmonic mean of the mean lives of the initial and final levels.
This line width has no practical importance for ordinary spectroscopy as it
tells
<T
smaller than the width of image produced in spectroscopic instruments

owing to their finite resolving power.
is
2.
We
Classical electromagnetic theory.
regard radiation as a spatial flow of energy governed essentially by

s equations. In these we have a scalar field p which is the electric
Maxwell
charge density in electrostatic units and a vector field I which is the electric
current density in electromagnetic units. These fields are connected by an
equation of continuity,
-^
c dt
which expresses the conservation of electric substance.

Related to p and I are two vector fields, the electric field S and the
magnetic field ^, which are characterized by MaxwelPs equations
(2a)
0,
(2b)
(2c)
-^.
c
SB 4fljr.
(2d)
ot
These equations express respectively the fundamental laws:

(a)
Non-existence of a magnetic stuff analogous to electric charge.
(c)
Faraday s law of electromagnetic induction.

The Coulomb law in electrostatics.
(d)
The Ampere law
(6)
for the magnetic field
due to
electric currents, to
gether with Maxwell s displacement-current hypothesis.
There are various auxiliary fields which may be conveniently employed

in dealing with these equations, the most important of which are the scalar
and vector potentials cp and A. These have the property that if they are made
to satisfy the equations
0)
6-2
THE THEORY OF
84
24
EABIATIO3ST
and ^, derived from them by the formulas
IM
Maxwell s equations. It is to be noticed that if p s one may set
=
S and 3tf simply from A. This is a usual procedure in dealing
and
derive
9
will satisfy
with electromagnetic waves at points far from their source.

In the theory of the differential equations (3), it is shown that the solution
for the scalar potential can be written as
* y
i )
in which
(5)
J P&y^dxdydz,
=t
R/c,
the distance between the volume element dxdydz and the point
r
x y z This is known as the retarded-potential solution because each volume
element contributes to the potential according to the value of the charge
and
J8 is
f
density there at a time earlier than the instant t by the time it takes light
to travel from the volume element to the point x y z . An exactly similar
f
formula gives
A in terms of J.
To develop the classical theory of radiation in a form suitable for use in

the quantum theory we need to expand this retarded potential formula
under the assumptions that p and J vary harmonically with the time, i.e. we
take p(xyzt) as the real part of the expression p(xyz)eîvi Here p(xyz)
may be complex, as in the case of a rotating unsymmetrical charge distribu
.
tion.
Hence we write
where I is in general a bivector [a vector of the form

and I the equation of continuity becomes
(Jp -f iiy] .
div/+$fep = 0,
where
3.
Jc
Expansion
In terms of p
(7)
= 2?rv/c = 27rcr.
of the retarded potential.
In the theory of atomic spectra p and J are essentially zero outside a sphere
1
of radius small compared to
This suggests an appropriate development
of the retarded-potential formula according to powers of k. In the expression
&"
<?(x
)~
p-ikR
(1)
eW^JpLifo
we may use the expansion*
BATEMAN,
Partial Differential Equations, p. 388,
Example
4.
34
EXPANSION OF THE BETARDED POTENTIAL
where r and r are the radii to the points x y z and x y

between them, and A and
are Bessel functions:
85
is
the angle
This shows us that the potential given by

diverging spherical waves, since
We
(1)
corresponds to a system of
be interested only in the two leading terms in Z,\(kr ). The first

measures the transport of energy, the second that of angular momentum;
the higher terms correspond to local fields near the oscillating charge
r
shall
distribution.
Substituting
(2) in (1),
we find for the amplitude of the A th diverging wave
this point we introduce the classification of terms by powers of Jcr,

instead of by the integer A. We shall be interested only in the first three
terms. Since
At
KT
l*<r
we
^
/9}j_l\l
a. o...(Â-r
l;\
9/ O^_i_Q\
.iÂ-roj
find for the leading terms in the scalar potential

tftrivi
1
i
==
-i}
^~
A corresponding expression holds for A on replacing p by I.
Here the terms
moments of the charge distribution.

the second gives what is called (electric-) dipole
are expressed in terms of the successive
The
first
term vanishes,
radiation, the third gives (electric-) quadrupole radiation. It has not been
necessary thus far in spectroscopy to consider higher terms.
We define the dipole and quadrupole moments of p by the equations
P= fprdv,
91
=jprrdv.
Using the values of the Legendre polynomials, we readily find that
(7)
34
86
where 3t s is written for the invariant sum of the diagonal terms in 91, and
rQ is a unit vector in the direction r
Let us make the corresponding reductions for the vector potential. First
we note that from the equation of continuity and the fact that p and J vanish
.
everywhere outside a
finite
sphere
we may write
=z f
where g
is
any
scalar,
vector or dyadic function of position that
everywhere that p and I are not zero. Putting
=
already asserted. Putting g r and rr,
fldv
= ikP,
we
J(Ir
g= 1, we
is finite
find
Jpcfo
= 0,
as
find the relations
+ rl)dv = %
which enable us to express the vector potential in terms of the moments of

the charge distribution. Because of the occurrence of k in these formulas we
need only the first two terms in A to have terms to the order &2 in the result:
f
/*
/^V +
^M_
\ f
I
The
first term gives the part of the vector potential associated with the
dipole radiation. The second consists of two parts:
=
The
first
term here
is
can be written as
ir^
proportional to the quadrupole

r
-r;xM, where M=ljrxldv

the magnetic
moment; the second
moment
of the electric current distribution.
(8)
The second term

and
therefore represents part of the electric-quadrupole radiation field

in
the whole of the magnetic-dipole radiation field.
To summarize, the scalar and vector potentials up to the second power of

k in the moments of the charge distribution are
9=
^
Since the coordinates of the volume elements of the charge and current
distribution no longer appear, we have here, for convenience in later work,
EXPANSION OF THE RETARDED POTENTIAL
34
used unprimed
tials
letters to
87
denote the position and time for which the poten
are computed.
The correspondence
4.
principle for emission.
Before calculating the radiation field associated with the potentials of the
preceding section let us investigate their connection with the problem of
emission in the
quantum theory. The most fruitful idea in passing from the
atomic theory to the quantum mechanics was the correspondence

principle of Bohr. From the principles pf quantization developed for con
earlier
ditionally periodic dynamical systems it followed that in the asymptotic

between states of large quantum numbers the spectro-
limit of transitions
Bohr approached equality with actual frequencies of

the mechanical motion. In classical mechanics the energy
of such a
scopic frequencies of
Jl3 J2
strict
expressible in terms of a set of constants of the motion,
is
system
Jn3 known as action variables. The quantum conditions re

these J s to values that are integral multiples of h,
J = %&. (&=l,2,...,n; %.,aninteger) (1)
,
...,
fc
Tne
-_-
classical frequencies of the
fc
motion are given by the formula

,
so that the combination frequencies which occur in a Fourier expansion of
the motion are of the form

O
-m
= %r
Er^
k
k kwr
vJh
On
(rk , integers)
(3)
the other hand, the quantum frequency associated with the transition
set of quantum numbers
from the
by Bohr s frequency condition is
Developing this in a power
^
The
first
quantum
series in the r
s,
we have
dE
term is just the classical frequency; the other terms represent

is a linear function of
corrections. In the special case in which
that of a system of harmonic oscillators, the higher terms

s,
vanish so that the correspondence is exact.
the
which
is
great idea was that this approximate connection probably holds

not only for the frequencies, but for other features of the spectrum as well.
Bohr
44
88
According to the frequency relation we have a definite term in the Fourier

analysis of the motion, namely, the combination term with multiples
rla r2
...,
Tn
correlated with a definite
+ Tl9 ...)-^(w
quantum
transition,
namely
(n
...), Although the radiation process in quantum theory
1
occurs in jumps, whereas the classical radiation process is essentially con
tinuous, Bohr postulated that statistically the transition probabilities will
,
be such that the expected rate of radiation for a particular transition will
have a similar asymptotic correlation with the corresponding Fourier com
ponent of the motion. Likewise the polarization of the radiation is postu

lated to be that of the classical radiation due to the corresponding Fourier
component.
Thus
for the electric-dipole radiation the procedure is to develop the
components of
Fourier
for
motion in the state specified by
%, w 2
...,
nn
in a
series, for example,
(6)
where each r runs over all positive and negative integers. The notation for
the coefficient is chpsen to suggest its analogy with the corresponding matrix
component in quantum mechanics. Exploitation of this analogy corre
sponds to the first step in the formulation of quantum mechanics as
done by Heisenberg.* Since P(t) is real,
[*!,. ..JP^K-T!,. ..]
= [%,. .-IPJnj + Ti,...],
it
was
(7)
which
is the
analogue of the Hermitian condition for matrix components.
the
By
correspondence principle, therefore, the rate of radiation of energy
is related to the
by quantum jumps in the transition (% -f TX ,...)-> (%
)
>
by the Fourier components of P whose frequency

corresponds to the quantum frequency in the sense of (5). That is, there is an
classical rate of radiation
asymptotic connection between the rate of radiation of this frequency and

the classical rate of radiation due to the dipole moment,
rl3
where v^TirÊfdJ^ and
...]e
^,
(8)
indicates the real part of the expression in
braces.
The expected rate of radiation in quantum theory is
Â^ +
where
T!,.. .;%,...),
the spontaneous-transition probability for the transition in

It
is this which is
question.
postulated to be asymptotically equal to the
classical rate of radiation.
is
HEISENBEEG,
Zeits. fur
Phys. 33, 879 (1925).
THE CORRESPONDENCE PRINCIPLE FOR EMISSION
44
89
For large quantum numbers and small r s the value of this classical rate
of radiation does not depend much on whether we use the Fourier com
ponents of the motion of the
initial state or
the final state.
But
for smaller
quantum numbers these two sets of Fourier amplitudes may be greatly

different and the correspondence principle does not say which should be used
.
In addition to these two
possibilities one might even consider basing the

Fourier analysis on some intermediate motion, say one with quantum
when considering this jump. All suchschemes
numbers (% -f J^ n 2 -f |r2
are equally good asymptotically.
Bohr never offered the correspondence principle as more than an intuitive
view that it is in this direction that we should seek for a more complete
quantum theory it was in this direction that Heisenberg found it. It is not
,
. . .
our aim here to develop quantum mechanics inductively from the corre
spondence principle. Rather we wish merely to sketch its place in the theory
of atomic spectra. Its importance in leading to the discovery of quantum
mechanics cannot be over-emphasized; naturally, after that discovery, it no
longer occupies a prominent place in the details of our subject. It is really
surprising how many developments were made with the correspondence
principle in its original form. The selection rules were obtained by arguments
concerning the absence of certain Fourier components. If certain of these

were absent in the initial and final state, and in a certain class of inter
mediate states as well, one felt very confident in the prediction that the
associated
quantum jumps had
zero probability. For a complete account
of detailed accomplishments of this kind see

and Line Spectra*
Van Vleck, Quantum Principles
The lack of symmetry between initial and final states which characterizes
the above discussion was remedied, and the verscharfung of the corre
spondence principle accomplished, by Heisenberg s replacement of the
Fourier components by the corresponding matrix components. We shall use
the principle in this form.
Therefore we postulate that the radiation
field accompanying a spon

taneous transition from a state a of higher energy to a state b of lower energy
as to angular intensity distribution and polarization, the same as that

given by the classical theory for a charge distribution whose moments are
is,
where
(a\
|6) is
moment under
the matrix component of the kind of electric or magnetic
consideration. Moreover, the spontaneous-transition pro
bability is to be taken as =- times the classical rate of radiation of this charge

distribution.
* National Research Council
Bulletin No. 54, Washington, 1926.
44
THE THEOBY OF RADIATION
90
For an atom the various moments which we
shall
need are expressed in
terms of the coordinates of the electrons by the formulas
the matrix components of these quantities which are to be used in

calculating the radiation.
We haye preferred to put the theory of emission on the correspondence-
It
is
principle basis merely to avoid the lengthy calculations needed to derive the
results by quantum-mechanical methods. The postulates are justified in
Dime
5.
s radiation theory.
The dipole-radiation
field.
Let us investigate the field associated with the terms in 34 9 which depend
on the dipole moment, P. Using 2 44 we find, on retaining only the r~ l and
r~ 2 terms,
where
$ is the unit dyadic
ii+jj+ kk.
The flow of energy is given by the Poynting vector,
(2)
?r
Since
is
S and ^ are given by the real parts of (1), the time average radiation
given by
__
_
(3)
where S and Jf are as in
1).
Therefore at large distances the flow of energy is
-r.r,}-P\^rQ
(4)
In the theory of atomic spectra P is nearly always given by formulas like

For transitions in which Am = 0, the dipole moment is of the form Pk
9 3 11.
directed along the z-axis, and the radiation field is that of a simple linear
oscillator. In this case, in any direction of observation, the radiation is
linearly polarized with the electric vector in the plane determined
and P. The intensity at an angle 9 with the z-axis is
by r
(5)
THE DIPOLE-BADIATIOK FIELD
54
91
so the entire rate of radiation through a large sphere in all directions

*.
For transitions in which Am =
(6)
the vector Pis of the form VfP (i
1,
Am =
if),
so
Am
rotates in the xy plane, in the clockwise direction for

=4-1
sense
for
the
in
the opposite
viewed from
positive ^-direction, and
1. In either case the intensity in a direction making an angle 6 with
The vector
when
is
the z-axis
is
7,4
(7)
and the whole radiation in
We may now
write
all
down
directions
is
the same as
(6).
the transition probabilities. In view of the
4
postulate of 4 they are
,
(8)
A(a,6)-^p|(a|P|&)|,
for the transition
value of
from the upper state a to the lower state
Am, we may
replace
P2
by 4|(a|P|6)j
6.
in equations (5)
For
and
either
(7) to
find the angular intensity distribution.
Before calculating the polarization in the case Am= + 1, let us consider

27
the general question of the type of wave represented by &{$e
}, where
$=z$r + iffi a constant bivector. The variation with time,
"
^{Seîvl} = Sr
GQs2<rrvt
-#
sin2irtf,
(9)
s end-point describes an ellipse turning from ffi into

on looking at the plane normal to
counter-clockwise
r
as
you would be faced if the light
TQ in the direction opposite to r (that is,
is
<f
such that the real

.
If this direction
<?
is
were to enter your eyes), the polarization is called left-elliptic. In general,

in which case they form a pair of conjugate semir and f i may be oblique,
<f
usually more convenient to shift the phase so

so coincide with the principal axes of the
and
that they are perpendicular
1^ ^
and choose S so that the real and
ellipse. We may write Se^^-Seê^
diameters of the
ellipse. It is
i8
leads to the following
imaginary parts of Se are perpendicular. This readily
equation for
S,
that the location of the principal

Usually the form of the calculations is such
atomic radiation problems, the
in
Thus
from
axes is evident
symmetry.
circles of the polar coordinates.
latitude
and
axes are along the meridians
We shall denote by 6 and the unit vectors at each point in the direction
<p
54
THE THEOBY OF BADIATION
92
of increasing 6 and
<p
respectively so that r
cp
are the unit vectors of a
right-handed system.
Then,
if
P = VfP (i
ex (6 e
The
radiation
is,
ij) 9
<p
9 Hf
tj)
therefore, for the
the direction 5 =
0,
= (0
upper
cos0
icp
)e^
(10)
sign, right circularly polarized in
right elliptically polarized in directions
linearly polarized in the equatorial plane,
and
left elliptically
0<#<77/2,
polarized in
the southern hemisphere. For this case the electric moment s variation in
time is represented by a point rotating in a circle in the &y-plane in the same
sense as the rotation of the electric vector in the light wave. The factor
it?
occurring in $ attends to the obvious detail that the phase in the light
wave sent in the direction 9 has a constant relation to the phase of P relative
to the direction of observation.
Next we consider the transport of angular momentum in the radiation

The volume density of electromagnetic momentum in the field is given
2
by S/c where $ is the Poynting vector. The corresponding density of
2
angular momentum relative to the origin is therefore r x S/c and as this is
field.
moving outward with velocity
c,
the transport of angular
through unit area normal to r is r x Sjc.

Another way to find the angular-momentum transport
to consider the
equation
in which
is
momentum
-S) = 0,
(11)
Z is Maxwell s stress tensor for the electromagnetic field
This shows that the negative of the stress tensor gives the transport of
Z*n gives the amount of
electromagnetic momentum, in the sense that
momentum
crossing unit area normal to
n toward
positive
Hence the momentum transported radially outward

angular-momentum transport corresponding to this is
-rx
to
(>r
= --
is
n in
2>r
unit time.
and the
rx
Since the leading term in S in the dipole-radiation field is perpendicular

r, we need to consider the next term to get the angular -momentum trans
port. Using (1), we find the time

large distances to be
mean angular-momentum
transport at
93
THE DEPOLE-BABIATION FIELD
54
F is real except for a phase factor e^, as when Am = 0,

momentum vanishes. For Am = I, where F is of
value of Tav is
ij), the
per unit solid angle. If

the radiation of angular
the form
V|P (i
(12)
oTT
the axial symmetry we see that the result of integration over

directions must be a vector along the z-axis. Its value is
From
Am =
all
k which is in the right

of the atom.
momentum
direction to compensate
Moreover the rate of radiation of angular momentum bears the correct ratio
to the rate of radiation of energy to insure that the angular momentum is
The radiation
for
-f 1 is
thus in the direction
for the increase in angular
ft
radiated in the same time as the energy hv. This
very satisfactory for

a foundation for the
as
used
formerly
the correspondence principle, and was

selection rule in the angular-momentum
is
quantum numbers.*
be objected that we have proved too much, since we found no

radiation of angular momentum for Am= 0. To be sure there is no change in
the ^-component of the atom s angular momentum in this case, but there
It
may
a change in the resultant angular momentum, j according to 9 11.

The answer is that the classical analogy breaks down here, or better that the
classical analogue of the quantum-mechanical state for an assigned j and m
may be
must be considered to involve an average over
all
orientations of the
com
is zero by symmetry,
ponent ofJTin theo^-plane. This time average of Jx or J^
J
and
Jy are zero. From
of
matrix
and also since the diagonal
x
components
have
an average value
of
this standpoint the perpendicular components
of zero before and after radiation, so there is no change in their value, and
this agrees with the vanishing value of the angular-momentum transport in
this case.
PROBLEM
Show
that the magnetic-dipole radiation field is obtained by writing 34? for

in (1). Hence show that for magnetic-dipole radiation
by
and replacing
!HMW,
6.
and -
&
for 34P
(13)
The quadrupole-radiation field.
The qruadrupole-radiation field may be handled in an analogous fashion.

For brevity we omit the discussion of the transport of angular momentum
* E.UBDJOWICZ, Phys. Zeits. 19, 440 (1918);
BOHR, Quantentheorie. der Linienspektren, Part
I,
p. 48.
64
THE THEORY OF BADIATION
94
in this case.
The terms responsible
tials 3 9,
for the energy flow are,
from the poten
fl3 pe 2iri(vt-or)
*-^
-<3-r<r<t
Y&-ro)
(1)
so that the Poynting vector
is,
from 54 3,
Z*6
3~r
r H91-r )pr
(2)
We
are not ready-armed from Chapter in with the formulas for the
so we must stop to work
dependence of the matrix components of 9 on
m>
a dyadic formed of the sum of dyads

er^ for each electron
4
of
vectors
from
formed
is
a
sum
of
so
and
type T considered in
dyads
(4 9),
8 3 et seq. We can find the dependence on m of any such dyad by calculating
them, out.
its
9i is
T given
matrix components from the matrix components of
by
9 3 11.
Thus,
(ocjm|ri T2 |a7m )=
/
at* j*
\T^ j m
/
((i}m\Tl \ofj m^(^j"m
).
(3)
m"
This enables us to see that the selection rules on j and
for T^T* are Aj or

In this way we may prepare a table of all non-vanishing
matrix components for any dyadic made up of a linear combination of dyads
of vectors of type T, If this be done for a general dyadic the formulas can
Am =
2,
1,
or
0,
be expressed as the sum of the matrix components of the symmetric part

and the antisymmetric part. The latter is equivalent to a vector whose
as a vector of type T.
matrix components show the same dependence on
Since any symmetric dyadic is the sum of two symmetric dyads, in making
to get a symmetric dyad.
the calculations we may as well put 1\ = 2 =
Calculations of this type were first made by Rubinowicz.* It turns out
that each matrix component can be expressed as a numerical coefficient

multiplying basic dyadics, S(Ara), which depend on Am only, except that
(aj m\TT\oLJm) also includes an extra term which is a numerical factor
times the identical dyadic, ^ = ii+jj + kk. These dyadics, normalized so
that
! Am
(
)"j|(
Am =
)
are
1,
1)
ffi(O)
They determine the angular

*
= J[+ ik
i(kj+jk)}
(4)
=V|[M-i-ijj].
intensity distribution
and polarization in
ETJBEJTOWICZ, Zeits. fur Phys. 61, 338 (1930).
its
THE QUADRUPOLE-RADIATIOH FIELD
64
dependence on Am.
95
We may
readily calculate the values of the vectorial

factors occurring in (1) for these basic dyadics:
- for
(3 tVo)-fl(
(S
= |sin# e 2i*(6 cos0

9
(e o cos20
*9
l)fo =
2)*r
<,)$(
(3-Vo)-(0)ô
icpo)
oos0)
-V|sinooBfle
=0;
(3-Vo) 3-r
so that the spherically symmetric part of (<x.jm\TT\afjm) does not con

tribute to the radiation. With these factors it is easy to see that the polariza
and on
tion of the light depends on the direction of observation
Am
according to the scheme
Am=0
Am=-fl
Am=-i-2
Here the senses right and left are for the positive values of Am. They are opposite for negative
and opposite at an angle TT Q in the southern hemisphere.* a and TT mean linear
values of
I the elliptic polarization becomes
respectively. For
along cp and
polarization with
circular at 7r/3 and 2-7T/3.
Am
Am=
<f
First
we consider AJ =
for the dyadic
A, =
r^
Here occur three sums,
(ajfoja* j+v) (a* j+ vi^ja j),

a*
2
2
1. Byrec[tdringthatr i xri =0(cf. 12 4 and 12 9b) hare vanish
forv=l,0,
a relation, between them,
ing matrix components, we find
(2j
If we write
+ 3)1)! - D - (2? D = j + 1 )D^ -jD^
!)!>_!
the matrix components
in question
= 0.
become
+ 2:) ft(
l)=D(2m
=
Similarly for Aj
JS?
for v = 0,
If
we
let
1.
l)V(j + m)(jm+l)(
there occur
= S(aj:ri
1)
two sums,
r
;+v)(a ;+vjrf :a ?-l)
:a"
<
The vanishing of the matrix of rt x r^ here shows that
2)
96
64
these matrix components are
=
r
(<x,jm\r i i \ot
lE(j + m)(j*m-l)(j + m-2)(jm+l) ft(
2),
jlml)
(6b)
Finally for Aj =
2 the matrix components are multiples of the single sum
the complete expressions being

2),
m
The matrix components
for
Aj=
)[(j-
or
-h I
these since the matrices are Hermitian.
2
l)
-f
-m2
t(0).
(6c)
2 can be written
down from
The dyadic 9J is of the form
eSr^rz
so that
D,
JJ,
its
and
matrix components are formally like those given above with new
eS of the D, U, and F for the individual electrons.
F which are
To
we need to know that the integral

of the product of each of the first three expressions in (5) by its
complex
conjugate over all directions is 4-Tr/o. Hence using the postulate of 44 , the
total rate of radiation associated with the transition (oy
oc
/ ) is
calculate the total rate of radiation
m->
where the square of the dyadic indicates the double-dot product of the
dyadic with its complex conjugate.
This gives for the corresponding probability of spontaneous emission
5
A(*jm,* j
m )= ?^ J(ajm|%|ay m
2
)]
(7)
Examining the formulas for the matrix components we see that transitions
between the following values of j:
and 1 ^ 0, have
0->Q,
J->|,
vanishing
intensity. Likewise in the dipole- radiation formulas the transitions J
have vanishing
->j=0
intensity. These are special cases of a general rule,
The 2 n -pole radiation is connected with a change in the resultant
angular
momentum of the electromagnetic field by n units. Since there is con
servation of the total angular momentum, the vector sum of that of the
atom plus that of the field remains constant.
must then have
We
64
THE QUABBUPOLE-BADIATION FIELD
97
where j is the resultant angular momentum of the atom before radiation and
/ its value after the transition.

Brinkman* has developed a group-theoretical method of obtaining the
qiiadrupole -moment matrix components. The next term, or octopole
radiation, has been considered briefly by Huff and Houston.j
7.
Spectral lines in natural excitation.
When an atom is unperturbed by external fields, its Hamiltonian com

mutes with the resultant angular momentum J. Asa, consequence (3 3 8) an
energy level corresponding to a given value of j is (2j+ l)-fold degenerate,
each of the different states being characterized by a different value of m. Such
a set of (2/+1) states we shall call a level. When the atom is
perturbed and
the different states do not
all have the same

energy it is still convenient to
such a set of states by a generic name, so we shall also use level in
this wider sense. Likewise we define the word line to mean the radiation
associated with all possible transition between the states
belonging to two
refer to
The radiation resulting from a transition between a particular pair of

we call a component of the line. All components have the same wave
number unless the atom is perturbed by an external field.
levels.
states
The actual emitted intensity of the component from

is equal
to
. T/
,.
,
ergs/second
&
^
r/
x
state
a to
state b in
),
(1)
where N(a)
is the number of atoms in state a and

A(a, b) is the spontaneous
emission probability for that transition. This equation holds only when the
radiation density present is so small that the induced emission is
negligible.
In actual practice it is very difficult to know N(a) and hence to make
absolute comparisons between theory and experiment for emission inten

sity. Most work is therefore on relative intensities of related groups of lines.
In most light sources the conditions of excitation of the atoms are suffi
ciently isotropic that it is safe to assume that the numbers of atoms in the
different states of the
same level are equal. This is called natural excitation.

some definitely non-isotropic way, as by absorp
If the excitation occurs in
from a unidirectional beam of light or by impacts from a unidirectional

of electrons, large departures from natural excitation may be pro
duced. The study of such effects J raises a whole complex of problems some
what detached from the main body of spectroscopy. We shall always
assume natural excitation unless the contrary is stated.
tion
beam
BRINKMAK,
Dissertation Utrecht, 1932.
f HUFF and HOUSTON, Phys. Rev. 38, 842 (1930).

j For a review of this field see MITCHELL and ZEMANSKY, Resonance Radiation and Excited
Atoms, Cambridge University Press, 1934.
98
74
The total intensity of a line is the sum of the intensities of its components.
Hence in natural excitation for the line from level A to level -B,
(2)
in the case of dipole radiation, by 5 4 8. The sum of the squared matrix com
ponents occurring here we shall define as the strength of the line, B(A B).
9
We shall also
ponent
2
find it convenient to call |(&j.P|6)f the strength of the com
a, b and to introduce the partial sums over a and 6 alone, with the
notation
j?)
=
|
a \P\b)\*,
B(A
b)
= S \(a\P\b)\^
a
S(A, B) = S 8(0, -B) = S B(A

a
=
b)
S(a, 6)
a&
intensity of a particular line is therefore proportional to the number of

atoms in any one of the initial states, to a4 and to S(A, B). This represents
a slight departure from the traditional procedure, where the intensity is
The
written as proportional to the total
NA
(
As N(A)
Here
number of atoms
in the initial level,
I(A, B) = N(A) hv A(A, B).
(2jA 4- 1) N(a), this gives
is
the spontaneous-transition probability of Einstein. It has the

is not symmetrical in the initial and final levels
owing
disadvantage that it
to the factor (2jA -f
I)""
which relates it to the symmetrical
S( A, B).
In later
work we shall find formulas for the strengths of certain classes of lines where
one level belongs to one group and the other to another group (e.g. multiFor the strength of a line it will not matter in such formulas whether
or B is the initial level, whereas if jA ^jB the Einstein A s are different in
plets).
the two cases. Therefore
we
shall regard the strengths as the
more con
venient theoretical measure of line intensities.
Before calculating S(A, B) let us consider what properties we expect

have. First, since the excitation is isotropic and there is no external
it
to
field
we expect
the radiation to be of equal intensity in all directions, and unpolarized. Second, if the initial conditions are isotropic we expect the
physical situation to remain so during the radiation process. This means that
the
mean life
state
must be the same. The mean life of a

a is the reciprocal of the total transition probability from a to all lower
states b,c,d,
reciprocal of
of each state of level
...
belonging to
all
lower levels B, C, D,
....
That
is, it is
the
64-TAr3
This should have the same value for each state
a,
otherwise
if the
excitation
SPECTRAL LIKES IN NATURAL EXCITATION
74
cut
is
off tlie
numbers of atoms in the
different states of level
99
A will be un
equal after a time, and a non-isotropic situation will have grown out of an
isotropic one. Since there is no very simple general relation between the
we expect this to be true for each suramand

because
of
the symmetry of |(#|P|6)I 2 with regard to a
S(a, B). Likewise
and b, it follows that if S(a, B) is independent of a, then B(A, b) is indepen
levels
dent of
C,
...
for all atoms,
6.
We shall now show that the dipole radiation in the transition a j
->
/ is
4
isotropic and unpolarized. By 5 5 the intensity in the direction 6 of radiation
plane-polarized in the direction 8
S
m=-j
There will
also
is
proportional to
(Am = 0)
2
2
I(ajm|P|a /m)| sin 0.
be right- and
(4a)
whose
left-elliptically polarized radiations
intensities (5 4 7) are respectively proportional to
That the sum of these three quantities

13
3;
is
Am= -h 1 and Am=
1 is
(4b)
(Am=-l)
(4c)
independent of 8 follows from
hence the intensity of radiation is the same in
of the terms for
(Am* +1)
all directions.
no longer
The sum
elliptically polarized
because the right-handed radiation from m->m-f 1 is of the same intensity

as the left-handed from
1; it is, however, partially plane
m->
polarized with relative intensities
Sf(ajm|P|aY+l)|
ponent
is
radiation
cos
added to that
is
along 8
for
Am = 0,
2
along
j m-f I)]
54 10 and 54 9). When the
SâjmlPla
(cf.
it
and
cp
com
follows from 13 3 3 that the resultant
unpolarized.
The summations over the three values of mb which go with
ma = m have
31
where
the
value of m
13
their
of
in
been
evaluated
independence
already
was noted. Because of this, summation over m merely introduces the factor
(2j + 1), so for the whole strength of the line a j to a j we have
,
where
S
(^">
* J
X)
K*/5
2
)|
S
0">
()
Everything we have said here about electric-dipole radiation is true for

magnetic-dipole, if we replace P by M, since the dependence on m of
4
(4 9) is the same as that of P.
For quadrupole radiation the analogue of
(2) is
(6)
The strengths for quadrupole lines
are defined exactly as for dipole with
% in
100
74
place of P. Expressions for the strengths of the various quadrapole lines are
easily calculated from the values of the matrix components in the previous
section.
The
results are
j- 3) F*.
The
partial
sums here
equal intensity in
analogous to
also
have the property of giving unpolarized light of

The spontaneous transition probability
all directions.
(3) is
S27rVS(4,S)
A^,^)--
,_.
(8)
^j-^-j-y-.
Induced emission and absorption.

In view of the relation found in I 4 between the probability coefficients
for spontaneous and induced emission and absorption, the formula for
B(A, B) in terms of the matrix components of dipole or quadrapole moment
8.
147
with 7 4 3 or
7 4 8.
That gives the result for the

an atom with an isotropic unpolarized field.
There is a further point on which we wish information, namely the pro
bability for absorption of light from a unidirectional polarized beam. From
is
given by combining
interaction of
the following argument, we expect this to be simply related to the A(a, b)

for a line component. The emission in the transition a~>6 is a
spherical wave
whose intensity and polarization in
different directions we have calculated

Let the angular distribution factor be
for the important special cases.

f(0 9
<p)da>,
where
over
Jfda>=l
bivector (//
1)
all directions,
and
let
/(0,<p)
be the unit
normal to the direction of propagation, n, which gives the

wave emitted in the direction B, 9. We suppose for con
polarization of the
venience that
/ is
normalized so that
its real
and imaginary parts are
4
perpendicular, as discussed in 5 .
If A(a, b) is the whole transition
4
4
probability (5 8 or 6 7), then we can say
is the transition
probability for processes
statistically that A(a, &)/(0,<p)&o

in which the light is sent out in the solid angle
place with the polarization described by /.
da>
at
6,
9.
This emission takes
Hence
for the probability of

b
to
with emission in the solid angle dw at
spontaneous transition from a

6, 9, and with the polarization I we
may write
= A(a,6)/(fl, ?
A(a,6 fl,9,
(1)
There will also be stimulated emission. This and the
emission
spontaneous
constitute the inverse process to
absorption from a beam whose range of
directions is dcu at 0, 9. Of the 87ra2 da waves
per unit volume (we say waves
z
as a short expression for degrees of freedom of the
field) there are 2a dada)
associated with waves which travel in directions contained in
the factor
two arising from the two independent states of polarization associated with
s
/)&!>
)&>.
da>,
INDUCED EMISSION AND ABSORPTION
84
101
each direction of propagation. We are at liberty to choose the independent

states of polarization in any convenient way. The natural way in discussing
the transitions a
^ b of the atom,
of polarization and to introduce
to choose I as one of the canonical states
is
orthogonal to
in the sense
l*m = 0,
to
describe the other.
Let the intensity of the beam be p(a, I) da da*. This is the volume density
of energy in wave-number range da of waves whose directions are in the
cone da} and with their polarization described by I. We define the probability
of absorption (or induced emission) of this radiation so as to cause a transi
tion of the atom from a to b as
4
Then, as in 1 6, if a is the upper state, the balance of emission and absorption
between these degrees of freedom and the atom when in thermal equilibrium
gives
N(a)[A(a, b,
8,
9, 1)
+ P (a, /) B(a, b, 6, 9, 1)] = N(b) B(b, a, d, 9,
to get
/)/>(*,
I).
and the Planck law 1 5 divided by

and
that
statistical weights are each unity,
here
the
p(cr, I)
realizing
Using the Maxwell-Boltzmann law

8?r
1 44
B (a, 6, 8, 9, I) = B(b, a, 6, 9, 1)
we have
A(a,b,6,<t,l)
(2)
= hco*B(a,b,e,<p,l).
The same equations hold
but as we
have chosen
wave sent
for the other state of polarization, m,

I in accordance with the actual polarization of the
out in the emission process, A(a,b,

absorption coefficient vanishes.
5, 9,
m) = 0, and
1 48
Analogous to
so the corresponding
the rate of
we may write
emission processes in the form
N(a) A(a,
where n(a) = p(cr
fyjhca* is
beam in the
&, 9, 9, 1)
[*(<r)
+ 1]
(3)
da>,
the mean number of quanta per degree of freedom
under consideration.
Let us consider the description of the polarization a little more closely.
The bivector I may be written in the form
in the
direction
and polarization
/= 8
since
we have found in
and
<p
labelled
state
(4)
cos|S -F icp sinJS,
6 4 that the elliptic polarization always has

for its principal axes when the transitions are between states
5 4 and
by m. The range
<
^ 2rr covers
all
possible types of polarization
according to the scheme

Left
102
The
84
bfvector orthogonal to I is
m=6
icp cos JS,
sin|S
which represents a wave of the same ellipticity as that for I but with the
major and minor axes interchanged and the sense of turning, right or left,
reversed.
If the incident light is polarized in some other manner than with regard
to one of the canonically chosen forms, I and m, say with its electric vector
a scalar multiple of the unit bivector e, then it may be expressed as the
resultant of two components polarized as in I
and
m by the formula
e = l(!*e) + m(m-e).
In
this case the absorption is
simply
(5)
times the value
|f ^|
it
has
when the
If the incident light is unpolarized, the intensity is

polarization
along
distributed equally "between each of any two orthogonal states of polarization,
is
I.
so the absorption is half what it would be if the polarization were along I.

Finally let us consider the absorption by a group of atoms in a state of
natural excitation.
states b of level
whose polarization
b,
Using
(1)
The
B to
is
total
number of
transitions per second
the states a of level

described
by e
is
flCcr
and the expression
54 8
from the
caused by absorption of light
for A(a, 6)
%f a
we find,
for
example in the case
of dipole radiation, that this equals
We
have seen in
7 4 that the total emission of all
components of a
line in
unpolarized and of equal intensity in all directions.

The sum here is the total strength of emission per unit solid angle with
polarization e\ this is accordingly l/8rr times the total strength of the line
natural excitation
is
A, B. Therefore we can express the rate of absorption for the whole

terms of the strength S(A, B) as defined in 7 4 as
line in
03
(6)
This gives the number of quanta per second absorbed from a unidirectional
of energy density p(<y,e)dadu) As we expect, it depends only on the
energy density and is independent of direction and polarization of the light.
beam
The Einstein B
for isotropic radiation obtained

877*
bears the proper relation to
A(A B)
}
from
S(A,B]
(cf. 1
47
and
7 4 3).
this relation,
INDUCED EMISSION AND ABSORPTION
84
103
The results we have reached by the phenomenologieal method follow from

the Dirac radiation theory. Another derivation which gives correct results
for the stimulated processes but does not give the spontaneous emission is
to treat the electromagnetic field of the light wave as a perturbation, using
the form of the perturbation theory in which the perturbation
is
regarded
as causing transitions.*
9.
Dispersion theory. Scattering,
Raman
effect.
In the preceding section we have considered the absorption of light by

matter. This occurs when the light frequency corresponds to a transition
4
frequency to within a small range given by the natural line breadth (1 10).
We wish now to consider the interaction of light and matter when there is
no such exact agreement. There are
essentially three
phenomena:
(a) The macroscopic light wave is propagated with

instead of c, where n is the index of refraction.
a phase velocity cjn
in the wave(6) There is a scattering of light to all sides without change

number. This is called Rayleigh scattering.
is a scattering to all sides of light in which the frequency is
amount corresponding to one of the atomic transitions. This is
an
by
the Raman scattering. It is unimportant in monatomic gases but we
mention it to complete the picture.
(c)
There
shifted
As subheads under
(a)
there are special effects like the double refraction

in an external electric or magnetic field (the Kerr
produced when the gas

effect and the Faraday effect respectively).
is
All these phenomena belong together in the theory, for they are all con
nected with the oscillating moments induced in the atoms by the perturbing
action of the fields of the incident light wave. When these are calculated by
quantum mechanics it is found that they include some terms which oscillate
with the same frequency as the light wave. In addition there are terms
which oscillate with frequencies vv(A,B), where v(A,B) is the Bohr
frequency associated with transition from level A to level J3. The latter are
cannot be
Raman
scattering; they are of different frequency

coherent with the incident radiation.
responsible for
and
so
The forced oscillations of the same frequency as the incident light are to
a certain extent coherent with it (that is, to a certain extent they have a
definite phase relation to the phase of the incident light), so it is necessary
to consider the phases in summing the scattered waves due to different
atoms to find the resultant field. This dynamical theory of refraction
goes back to Larmor and Lorentz, but the best

*
modern treatment
DIKAC, Quantum Mechanics, p. 173.
is
by
94
THE THEOHY OF RADIATION
104
Darwin.* The net result of such calculations can be simply stated. The
induced dipole moment
P is related to the electric vector
by
(1)
P=a<
where a is the polarizability tensor. Then if N is the number of molecules per

unit volume, the total dipole moment per unit volume is N&*$. Assuming
that
that the principal values of Na are each small compared

may neglect the difference between the effective on an atom
S\ hence the dielectric constant, a tensor, becomes
Na<Q, i.e.
to unity,
we
and the macroscopic
e=3 + 4â.
Using the form of Maxwell
equations for
(2)
a material medium devoid of free
charge and of unit magnetic susceptibility:
IdS
c
(3)
=-,
c ct
57-,
at
we may find plane-wave solutions in which each vector is a constant multiple

of ariv(t-nnrlc) where n is the index of refraction and n a unit vector in the
9
direction of propagation.
tions
so
& and &P
The constant amplitudes must
= o, nn x g
satisfy the
equa
=3tf,
are transverse to the direction of propagation. Since the prin
from unity by small quantities,

we may write n = 1 -f S. Mi irrigating &C between the right-hand pair of (4),
we find that the part of &f which is perpendicular to n satisfies
cipal values of the dielectric constant differ
If a is simply a scalar
index of refraction is
any value of
a,
normal to
n==l + 27rj/ya
n is admissible and the
(6)
(5) shows that 8/277^7" is equal to an eigenvalue of

the tensor a regarded as a two-dimensional tensor in the plane normal to n.
The two eigenvalues lead to two values of the index of refraction, that is,
but if a is not a scalar, then
double refraction for each direction of propagation, n. The corresponding

eigenvectors give the two types of polarization associated with the waves
propagated with the velocities corresponding to the two indices of refraction.

We shall see when we calculate a quantum mechanically that it is always
Hermitian (with regard to the three axes of space), so the two eigenvalues
*
DARWIN, Trans. Cambr. Phil. Soc. 23, 137 (1924);

see also ESMAUCH, Ann. der Phys. 42, 1257 (1913);
OSEEN,
BOTHE,
ibid. 48, 1 (1915);

ibid. 64, 693 (1921);
LTTNDBLAD, Univ. Arskrift, Upsala, 1920.
DISPERSION THEORY.
94
are real
and correspond
real unit vectors / and
SCATTERING.
RAMAK EFFECT
105
two
and
n form a right-handed coordinate
m,
to propagation without absorption. Choosing
m so that
/,
system, a will appear in the form,
a = an ll+ ai2 fo+ 512 m/4- a^mm.
There are two cases to distinguish: if a12 = 5 12 so the components are all real,
then the eigenvectors are real and we have ordinary double refraction,
whereas
if
a12 ^ 5 12 the eigenvectors are bivectors and the two fundamental
states of polarization are elliptic, or in special cases, circular. It is easy to
show that the two eigenvectors 2-L and 2% are orthogonal: St^St^Q.
Therefore the two states of polarization are orthogonal as discussed in 8 4
so they have the same ellipticity but one is right and the other left, and the
major axis of one ellipse is the minor axis of the other.
The foregoing discussion applies to the effect of the waves scattered by the
molecules if the medium is absolutely homogeneous, in which case N is the
number of molecules or atoms in unit volume. Owing to the thermal motions
,
of the molecules in a gas the medium is not homogeneous; the number of

molecules in any volume v at a given instant fluctuates about the mean value
Nv, where
N is calculated by dividing the total number of molecules by the
total volume.
The waves
by the number of atoms in each volume

mean do not interfere in regular fashion;
hence they give rise to scattered light. This, together with some non
coherent parts of the induced moment of unshifted frequency, gives rise to
element which
is
scattered
in excess of the
the Rayleigh scattering.

the scattered radiation.
We
shall
not have occasion to consider in detail
The
dispersion and the related double refractions produced by external

fields are of importance in the theory of spectra in that
they provide other
methods for measuring line strength than those based directly on emission
or absorption. For their theory we need to calculate the polarizability
tensor a introduced in
(1). This is obtained by calculating the perturbation

atom by the field of the light wave. Then we calculate the
matrix components (ajm|jP|a J w ) p of the electric moment of the atom
with regard to these perturbed states. Of these the component (oyra|P|oyra) p
gives rise to an electric moment which is coherent with the incident light
wave and which is responsible for the various dispersion phenomena. When
the atom is unperturbed by other external fields the energy is independent
of m, and the matrix components (aJm|Pjoc^ml) also contain terms
p
which are of the same frequency as the incident light, but these are not
coherent with it and so they contribute an additional amount to the
of the states of the
Rayleigh scattering without contributing to the dispersion.*

* See
BEEIT, Rev. Mod. Phys.
5,
106 (1933), for an adequate discussion of this point.
94
106
The calculation is based on a slight modification of the perturbation theory

of 8 2 required by the fact that the perturbation depends on the time. To a
,
first
approximation the perturbation
is
the electric vector of the perturbing light wave. This

transitions. The state (ocjm)
neglects efiects connected with quadrupole
- im
fi will to the first order
whose unperturbed eigenfunction is $(<y.jm)e,
where ^{?e27rM}
is
become
where ^(ctjm)
is
chosen so that this expression
satisfies (5 3)
(ir~;4Wi=i[J*^^
L
dt]
\
Here we have written
E = hv to represent the frequency of the light, and H
for the unperturbed Hamiltonian.
^(ajm) = f
Assuming the expansion
^(a7mO[(aym>jm)^^^
(8)
we find that this is a
solution if
(9)
The matrix component of F with respect to these perturbed states is,
to the
first order,
__
Using the results just found for ^(ajm) and ^(ajra) this can be written as
^{afcjmKe2
(ocjm|F|ocj w)
where a(ajm)
is
the polarizability tensor of the state
a(ajw)
_ y n a J w p a / m/ / w 1 p
~*
W -W + E
LI
l-
)( a/
QC
l
(10)
*},
ajm:
therefore
f m f \*J m T\
+ (*j\P\* f~^(*
-W-E
\
J"
(11)
m can
be found by using 9 3 11 for the matrix

components of F. When this is done it is found that the tensor can be ex
pressed conveniently as the sum of three parts: an isotropic part which is a
The dependence of a on
a doubly refracting part which is a multiple of S(0) of 6 44, and

a gyrotropic part which is a multiple of i(ji-ij)* When there is natural
excitation and no external fields, the last two parts vanish when the sum is
multiple of
5,
DISPEBSION THEORY. SCATTERING. RAMAN EFFECT
94
taken over
tlie different
107
values, leaving only the isotropic part. This
shows that in natural excitation, in the absence of external fields, the gas is
optically isotropic. The matrix components occurring in the isotropic part
are proportional to the strengths S(aj,oc /) of the lines, so that finally the
isotropic polarizability oc for natural excitation may be written as
- W(y,j)]jh. This is the average, not the sum, of

j )
the polarizabilities for the different values of m. The frequency v(&f, oc j) is
positive for those energy levels W(<x!j ) which are higher than TF(cy ), and
where
v(ct!
f, a j) =
\W(<*.
negative for those lower.
The Kramers-Heisenberg dispersion formula is obtained at once when we

= 1 + S. The term
(6) to obtain the index of refraction, n
use this value of a in
(12) gives the average contribution of atoms in the energy level W(CLJ) so
that the effective average value of a is obtained from the distribution
of atoms:
I
N(<x.j)
(13)
the atoms are in the lowest energy level we need here only
(12). In this all the v(oc /, aj) are positive. The dispersion
formula for this case was discovered before quantum mechanics by Laden-
In case
all
one term like
burg* from a correspondence principle argument. The existence of the

negative dispersion terms for atoms in excited states was recognized by
Kramers f and later Kramers and Heisenbergf reached the complete formula
Raman
from the correspondence principle. The ex

perimental reality of negative dispersion has been very thoroughly demon
strated by Ladenburg and his associates.
In the discussion of the dispersion formula (12) it is convenient to intro
duce a new set of quantities called the oscillator strengths, as distinguished
from line strengths. This is based on a comparison between any one term
of (12) with the term given by classical theory for dispersion due to an
and natural frequency
isotropic harmonic oscillator of charge e, mass
when perturbed by the
an
of
such
oscillator
motion
The
of
v(v! a).
equation
including
scattering
JLC,,
electric vector is
f + [2-jT v(a
where the actual motion
is 0l{r}.
a)]
The
r = ~SeMv{
solution of this equation represents
LADENBTIRG, Zeits. fur Phys. 4, 451 (1921).

t KRAMERS, Nature, 113, 673 (1924).
J KRAMERS and HEISENBERG, Zeits. fur Phys. 31, 681 (1925).
LADENBURG, Zeits. fur Phys. 48, 15 (1928);
KCXPFEBMANN and LADEKBURG, ibid. 48, 26, 51

CARST and LADENBURG, ibid. 48, 192 (1928).
(1928);
108
the sum of free and forced oscillations. For the latter
where
94
we find that r =
so the classical value of the

polarizability of the oscillator is
Comparing with (12) we see that the quantum theoretical formula

be written as
where
>
j)
JJ
- 8~
3e*h
^/ a
?
j)
for oc(oy)
(M)
S(q j? g-/)
2j+l
In other words, each spectral line contributes to the

polarizability a term
whose form is that of the contribution from a classical oscillator,
multiplied
by a dimensionless quantity f(a /, a j) called the oscillator strength. Like the

Einstein A s and B s the f s are not symmetric in the initial and final states
3
owing to the occurrence of the factor (2j+ 1)-1 From the symmetry of S
and the antisymmetry of ^(a /, cy ) we find that
.
Using (13) and (14=) it is possible to express the effective polarizability in

terms of thef s as the sum of contributions associated with each
line
5
spectral
in which a j refers to the lower

energy level of each line and a / to the
and
the
sum
is
over
all
the
lines in the spectrum. The
higher,
negative term
in the brackets is the correction due to the
negative dispersion arising from
atoms in excited states.
The oscillator strengths f( /, aj) satisfy an

interesting sum rule which
was discovered independently by Thomas and
by Kuhn* from consideration
of the dispersion formula from the
standpoint of the correspondence prin
ciple. It is a simple consequence of the law of commutation of
and
r and
is
important historically in that it was used by Heisenbergf to find this

commutation law in his first paper on matrix mechanics. The rule states
that for a system containing a
single electron the sum Sf(a /, aj) is equal to
unity.
The proof is very simple. From the commutation law 5 2 1 we have
THOMAS, Haturwiss. 13, 627 (1925);

KTJHN, Zeits. fiir Phys. 33, 408 (1925).
t HEISENBEEG, Zeits. fiir Pbys. 33, 879 (1925).
DISPERSION THEORY. SCATTERING. RAMAN EFFECT
94
For a
single electron in
an arbitrary potential field the Hamiltonian
109
is
from which by the commutation law we find that

so that the matrix components of p in terms of those of r are
j m )= -27r^v(cc a)(ajm|rloc
Hence a diagonal matrix component of (p*r T*p) is
/
(cf.jm\p\a.
4:7Tjji
S Ka
?
a)I(aj?m|r|oc
/m
2
)I
/m
= 3S.*
Substituting the oscillator strengths f(& }
2/-fl.
definition in (14) this
Sf(a j ,aj) =
is
the required
m by multiplying
,
<xj)
from
their
becomes
a
which
).
m"
This is independent of ra and can therefore be summed over
by
l,
"
sum rule.
10. Natural shape of absorption lines.
Let us now consider the actual shape of an absorption line in a gas, taking
finite width of spectral lines as given by 1 4 10. The finite
account of the
width
arises in the coupling of the
atom and the
radiation field
and so
4
applies to absorption as well as emission. When we take account of I IO we
find that the absorption probability varies with a over the natural width of
the
line.
Measuring a from the centre of the
line
we may
write
for the probability that a quantum between a- and a+du produces a tran
sition in unit time. Then, if p(cr) is essentially constant in the neighbourhood
of the
line,
the total number of transitions produced by radiation in the
neighbourhood of the central frequency
is
b(S,A,
0).
In the ordinary treatment in which we neglect the natural width this total
is written p(a) N(B) B(B, A), hence we have
number of transitions
b(J?,4,0) = B(J3,4)/TOp
(2)
The intensity of a light beam is measured by the energy I (a) crossing unit
area normal to the direction of propagation in unit time in the wave-number
range da at o\ Since p(a) is the volume density of energy, evidently
JT(a)=c/>(cr).
but with v 2 v3 , and v4 written in place of v(a

by Vism, Phys. Rev. 41, 432 (1932).
Sums of
this type,
a),
have been evaluated
110
10 4
The "beam goes through, thickness A# in time AOJ/C, so that its diminution in
intensity can be found by multiplying the number of transitions it produces
in this time
Hence the
material
by the average
size of
quantum,
intensity after going through
kv, to give
finite
__ _
is
the relation
thickness x of absorbing
= V~^<^
I(a)
_.
where
= -C
The coefficient of N(B) x in the
We
(3)
mi>
N(B) x.
7T(7
expression for is an area that is character

shall call it the effective area for absorption
of the line in question.

is the number of atoms in the initial level of the absorp
tion process contained in a cylinder whose base has this area and whose
istic
of that line. Then
height
is
the thickness of the absorbing matter.
The amount of energy removed from the incident beam in da at a is
If we are dealing with a small amount of absorbing matter, we may calculate

the total absorption by expanding the exponential function, saving only the
first
77<7
two terms and integrating with respect to o-. This recovers the result
or hvB(B,A)N(B)x[c which applies when we ignore the natural
line width.
But if there is a larger amount of absorbing matter, so that ~I

we have to calculate
or
>
1,
Vo Jf
[1
e"^w53*]
do/a
=I
*o /(O
0>
The integrand gives the shape of the absorption

For large the absorption
several values of
for
plotted
to get the total absorption.

line.
In Fig.
I 4 it is
becomes essentially complete in the centre of the band and the curve be
comes much broader than the natural breadth. Although the absorption
becomes very small as we go away from the centre of the line, the
amount
of matter present gives rise to appreciable absorption at fairly
large
large distances from the centre. The absorption is 50 per cent, for
coefficient
For large values of we can estimate the integral by supposing the integrand
equal to unity out to this point and zero beyond. This gives for the total
absorption
This
is
V 2V/I^2 =
2-39l a
V.
remarkably close to the true asymptotic value.
NATURAL SHAPE OF ABSORPTION LIKES
10 4
Considerations of this kind were
first
111
developed by Ladenburg and

They evaluated the
Beiche* on the basis of the classical electron theory.

integral in (4)
where the J
we
and found
s are Bessel functions.
have, for
>
Using the known asymptotic expansions
1,
Adopting any convenient definition for the width of the absorption

line, such as the wave-number interval between points where the absorption
amounts to some fraction of the total intensity, we see that the width varies
as the square root of the active number of atoms in the hue of sight. Out on
*
the wings of the
4
Fig. I
line,
where the absorption
The natural shape
is
small,
we may develop
of an absorption line for several values of
exponential function, retain only the

the absorption is given by
first
the
two terms, and thus find that
for
Thus the strength at a given point in the wing is directly proportional to

number of atoms in the line of sight. These results have had important
the
astrophysical application! to the theory of the Fraunhofer lines in the solar

spectrum, but we cannot go into this interesting topic here.
*
LADENBUBG and REICHE, Ann. der Phys. 42, 181 (1913).

f STEWABT, Astrophys. J. 59, 30 (1924);
SCEUTZ, Zeits. fur Astrophys. 1, 300 (1930);
MDTNAEBT and MULDERS, ibid. 2, 165 (1931);
UNSOLD, ibid. 2, 199 (1931).
CHAPTER V
The
simplest atomic model and the only one for which all details of the
spectrum can be accurately calculated is that of hydrogen and the hydrogenlike ions,
Zte
Hell, Li III,
etc.
This model consists of a heavy nucleus of charge
single electron, interacting principally according to the Coulomb

shall see in later chapters that the spectra of the alkali metals may
and a
law.
We
be treated by regarding one electron as moving in a central non-Coulomb

field determined by the nucleus and the average effect of the other electrons.
In this chapter we treat the hydrogen-like spectra and the alkali spectra,
anticipating for the latter the justification of the assumption that these
to a good approximation be regarded as one-electron spectra.
1.
may
Central-force problem.
We consider the motion of an electron of mass
ju
in a central field in which
distance from the centre

potential energy
given by U(r),
of force. Neglecting electron spin and relativity effects the Hamiltonian for
the electron is
its
From
where
is
83 4
r is its
we see that H commutes with Lx Ly

,
and any component,
e.g.
equation (6 2) for this
If
we
Hamiltonian
*2
and therefore
and Lz The Schrodinger

.
is
express the Laplace operator in polar coordinates
becomes
LZ9 may be taken as constants of the motion. We
look therefore for simultaneous eigenstates of H,

2
Lz and
and use 43 27,
this
The requirement that be an eigenstate of I, 2 and Ls determines its depend

ence on 6 and 9 according to 43 12, so we may write
<f>
(3)
In this case the equation
for the radial factor of the eigenfunction
becomes
-*
The allowed
values of the energy
W are the values for which this equation
113
CENTBAL-FOBCE PBOBLEM
and remain finite at infinity.

possesses solutions which vanish at the origin
occur in the equation for W,
of
L
does
not
value
It will be noted that the
z
hence each value of
W associated with the quantum number
will
be
(21 -f 1)-
38
commutes with JL.)

since
fold degenerate. (This result follows also from 3
The equation (4) for the radial eigenfunction is formally that of a onedimensional motion under the effective potential energy
the second term corresponding to the effective potential energy of the centri
determined by this equation
fugal force. Starting with the lowest value of
for a particular value of Z, it is customary to distinguish the states by means
quantum number n which has the value I -f 1

and increases by unity for each higher state, so that n
f
of a total
for the lowest state

I
1 is
number
the
of nodes of the radial eigenfunction, not counting the origin as a node.

Also in accordance with custom among spectroseopists we introduce a
letter
code for the values of I, according to the scheme

Value of Z:
Designation:
10
spdfghiklmn
11
12
13
14
15
...
The energy levels will be labelled by the quantum numbers n and I according
to the
scheme W(ls) W(6d),

9
etc.
This scheme arose in connection with the
interpretation of various series of lines in the alkali spectra, known as the

sharp, principal, diffuse, and fundamental series. The quantum, numbers nl
will be said to specify the configuration of the electron.
The preceding
discussion has been for a fixed centre of force. Let us
now
consider the two-body problem in which the electron moves under the in
fluence of the same field due to a nucleus of mass M. Let X, T, Z be the
coordinates of the centre of mass of the system and x y, z the coordinates of
9
the electron relative to the nucleus. Then, as in classical mechanics, the

kinetic energy of the whole system
where p
is
becomes
the so-called reduced mass. Expressing this in terms of momenta

and substituting the Schrodinger operators for the
as in classical dynamics
momenta, the Schrodinger equation of the two-body problem becomes
where Ax is the Laplacian with respect to the internal coordinates and A^- is
that with respect to the centroid s coordinates. The variables may be
separated by writing $ as a product of a function of the internal coordinates
by another function of the centroid
s coordinates.
The separate Schrodinger
114
I5
equation for the internal coordinates is then found to be just (2) with the
reduced mass in place of the electronic mass, while that for the external
coordinates
is
simply the Schrodinger equation for a free particle of mass
fi + M. The total energy E breaks up into the sum of the internal energy
and the energy of translation. The energy
of translational motion does not
need to be considered farther as it does not change appreciably during a

radiation process and thus does not affect the line frequency. Hence we may
allow for the finite mass of the nucleus simply by using the reduced mass in
place of the electronic mass in equation (2).
2.
Radial functions lor hydrogen.
We shall now consider the radial equation 1 5 4 for the

function U(r)
substitutions
a
*
Z
this equation takes
which
is
function.*
potential
Ze*/r:
jit
By making the
Coulomb
on the more concise form
recognized as the equation of the confluent hypergeometric

From the form assumed by the equation near p = 00 we see that
the solution which remains
finite
there contains the factor e~# 2
An attempt
to solve the equation with a power series in p shows that the solution which
l
is finite at the origin starts off with p +\ This suggests the substitution
which leads to the following equation
for/(p):
-/-O.
/
P
P
If this equation be solved in terms of a power series in
(4)
it will
be found that
/>,
the series breaks o2 as a polynomial of degree n I lifîsan integer

greater than I. The polynomials obtained in this way belong to the set usually
known as associated Laguerre polynomials.

series defines
solutions are defined
by
the requirement that
gives just the well-known
Bohr levels
W
*
If n
is
not an integer the power
a solution which becomes infinite like e+f* as p
WHJTTAKEE and WATSON,
for hydrogen-like
^Z*
A
->
oo,
so the finite
n be integral. In this ease,

atoms:
ue4
Course of Modern Analysis, Chapter xvi.
(2)
RADIAL FUNCTIONS FOB HYDBOGEH
25
115
The ordinary Laguerre polynomials may be defined by the equation
Ln (x) = ex Dn (xne~x
where D
formis
is
the symbol
for differentiation
(6)
),
with regard to
x.
An
equivalent
= (>-!)(*).
(*)
(7)
If the indicated differentiations are actually performed, this becomes
*~1 + M2(M
1}
The associated Laguerre polynomial L(x)

I%(x) =
It
is
DL
V--... + (- l)f.
is
defined
by
n (x).
the polynomials so defined that satisfy
(8)
(9)
(4),
the solution being
= L^(p),
(10)
An equivalent definition for the associated
f(P )
a polynomial of degree n
Laguerre polynomial is
1.
Pp^^).
Applying either of the
definitions
(11)
we find that the polynomials
are given
by
the general formula:
From
this
form they are
easily seen to
have the following relation to the
hypergeometric function:
Another form of the solution for the radial functions which is due to
is useful in some calculations. He showed that to within a norm
Eckart*
alizing factor the radial
wave function
r
R(nl)jr is
-t
n
/d\ ~i( /
-(S=f:nji(sy
in
which
z is set equal to
M-=)
Z]na. after the differentiation
Carrying out the indicated differentiations,
Tlie Laguerre polynomials
Z\ n +
may
also
it is
is
performed.
easily seen that
be studied with the aid of their
generating function, according to the following expansion

xt
ECKAET, Phys. R^v. 28, 927
(1926).
8-2
ONE-ELECTEOK SPECTEA
116
TMs
is
the relation used by Sckrodinger* for calculating certain integrals
involving Laguerre polynomials.
32
36
40
B
Fig. 1 , Radial probability distribution a R\nl) for several of the lowest levels in
hydrogen. (Abscissa is the radius in atomic units.)
In order to normalize the radial functions we need the result

(17)
which
is
readily obtained from the generating function or

*
SCHE6DHJGEK, Ann. der Phys.
80,
485 (1926).
from Eckart
FOB HYDROGEN
B.ADIAL FUNCTIONS
25
117
formula. Therefore the final formula for the normalized radial function
which
is
normaKzed in the sense
R\nl) dr=l. Several of these functions
->o
are given explicitly in Table I 5 The probability of finding the electron in

dr at r is 2 (nT) dr; this distribution function is plotted in Fig. 1 s for some of
.
the lowest states.

TABLE
1s.
=!.*
Normalized radial eigenfunctions for
= --
3V3
81V30
64
2V6
*r
V
5
768V35
The average
values of various powers of r for the hydrogenic

functions are given in Table 2 5 .f
TABLE 2s
a-*,
wave
î[63w*
- 357i 2 (2Z 2
-j-
2Z
- 3) + 51(1 + 1)(3Z 2 + 3Z - 10) + 12]
5
72
-2
-3
* In this table r is measured in atomic units. The

general eigenfunctions for any Z and
arbitrary length unit are obtained by multiplying the functions of this table by Vz/a. and
replacing r by Zr/a*
5
6
f The average values of r~ and r~ may be found in VAN VLECK, Proc. Koy. Soc. A143,
679 (1934).
3.
35
COTE-ELECTRON SPECTRA
118
The
relativity correction.*
We shall now investigate the effect of the relativistic variation of mass with
velocity in the central-force problem, particularly in the case of hydrogen.
To obtain the wave equation we take the relativistic expressions for the
Hamiltonian
(total energy)
H=^j==-l] + U(r)
and
for the
ft
is
V] _ #2
>
Hence
JE?
/ov
p~l, c==137,
may safely be
1 1 -f
-o-o
2 2j
AL
Then
= ac 2
The third and succeeding terms

hence
pi
u,y
the rest mass and v the velocity of the electron.
so that
units,
(1)
momenta
IJLX
where
(jS=^/c)
jLt=l, cf.
of this expression are very small (in atomic

Appendix) compared to the second and
regarded as a perturbation.
We
shall
keep only the
third term.
We shall thus solve the equation

i
= E$
(4)
under the assumption that the ift*s and energies do not differ much from those
of the non-relativistic equation
(5)
which we have considered hitherto. Under
Now from
(5)
this
assumption we
-ip
* See the sketch in
5 of Dirac
more rigorous treatment.
may set
35
THE BELATIVITY COBBEOTION
since
p 2 commutes with
U) we
119
may apply p* again to obtain
or to our approximation
Hence the
relativistic central-force
equation becomes
(7)
I5
Just as in
we may
take this
with the quantum numbers
iff
an eigenstate of angular momentum

and obtain the radial equation
to be
and
t
Z,
In the case of hydrogen and hydrogen-like ions we

rigorously.
We set U-~ Ze*/r to obtain
where a denotes the

c~>oo
fine structure
constant
e 2 /#c= 1/137-3.
this equation reduces to the equation 2 5 1,
may
solve this
We see that for
and that
it
may
be dis
cussed in the same way. For simplicity we rewrite the above equation in the
<,,
Near
= oo
the solutions of this become asymptotically
of which the solution with the minus sign remains finite for the energies in
which we are interested. At the origin if we attempt to find a power series
solution starting as r^,
whence
we
find the indicial equation
r=
Of these it is y+ which gives our previous solutions

Hence we proceed to find a solution of the form
nfXr*.
in the limit
c->oo.
(11)
TMs
The
35
OHE-ELECTRON SPECTRA
120
leads to the following recursion formula for the coefficients b k
series lLb krk
determined in this
E which
way converges
as e l ^~ A r except for the
it to terminate as a finite polynomial.

the
value of E which gives a polynomial
gives
in place of the infinite series. This value of E is found to satisfy
particular values of
Setting the coefficient 6
of degree $
cause
~*
_1.
(13)
If we take the negative sign for the square root in this expression we get a
2
set of energy levels just greater than
and converging to a limit at
2fiC
2
that point, with a continuous spectrum below
Such states of
2ju,c
,
E=
negative kinetic energy occur of necessity in
any
relativistic theory,
we may just ignore them by falling back on our treatment of the
but here
relativistic
variation of mass as a mere perturbation on the ordinary energy levels to

which (13) reduce if we take the positive square root and let c oo.
->
If we
expand
(13) in
or, in
powers of a
we
obtain, to terms in
a4
wave numbers
E
4.
RZ
<x?Z*
Spin-orbit interaction.
We shall now complete the picture of the fine structure in the absence of
by considering the effect of the electron spin. The effect of the
spin in one-electron spectra arises from the interaction of the magnetic
moment of the electron with the effective magnetic field set up by its motion
external fields
around the nucleus. Here, as in
all
considerations involving electron spin,
we must in the last analysis choose the term in the Hamiltonian which repre
sents this interaction to agree with experiment. From the picture of an
electron as a spinning top Thomas* and Frenkelf have obtained a formula
which gives experimental agreement, and is of the sort occurring
in Dirae s
5
theory of the electron, 5 Their result for the spin-orbit interaction energy
of an electron in a central field with potential U(r) is
.
where
L is the orbital and S the spin
angular
momentum.
THOMAS, Nature, 117, 514 (1926).

t FBENKEL, Zeits. fur Phys. 37, 243
(1926).
121
SPm-OBBIT INTERACTION
45
We
must then consider the energy
H +H
levels
and eigenfunctions
for the
Q
is the (relativistic or non-relativistic)
where
Hamiltonian
We shall use a representation in
interaction.
Hamiltonian without spin
2
2
Q
L S LX9 and Sz are diagonal. The eigenvalues of fi are in
which
3
of spin (this follows from 3 8 since
dependent of ms the z-component
commutes with S), so we may form an eigenfunction characterized by
= w nl
jjo
8,
ms) and m
I,
merely by multiplying
by
8(<r,
m8
as in 5 3 2:
m^
The matrix components of H1 in this scheme
(nlm8 m \H
mm
\ril
(2)
are
l)
=
Jo
2
"
Jo Jo
The expression
evaluated in 7
we can write*
component of L*S which was

This matrix component contains a factor S(Z,Z )- Hence
in braces
is
just the matrix
3.
f
j o
(3)
This shows that our matrix is rigorously diagonal with respect to L

to n, the perturbation theory shows
Although it is not diagonal with regard
effect
that the non-diagonal elements in which n s differ have a negligible
on the energies since states of the same I but different n have large energy
differences. This shows that the first approximation of the perturbation
will be a good approximation.
theory in which these elements are neglected
On the other hand if it had turned out that there were matrix components
* This method of calculation is given to show the relation between the Schrodinger and the
matrix type of calculation. We should ordinarily say
(n
m, mA&r) L S\n
m rf) =
Sm
(*
,*Z"7?i*7n
Since r2
73 3 the
commutes with L and
second factor contains

$(l,
which
is
seen to equal
(3).
(n
ilMK V
5, the first factor in the
S(T,
<
OK
"
"
rf\L*S\* V m t
<).
summand contains ^(Im8 m t Z*wmJ); by

sum reduces to
T). Hence the whole
m s wiilf(r)K ms m^(n m a
I
122
OKE-ELECTBOK SPECTBA
45
joining states of different I and the same n this approximation would have
been very poor, at least for hydrogen, where these states lie very close
together.
We thus consider the diagonalization of that portion of the matrix of H

which corresponds to a given configuration
With the notation
nl.
&- f
(4)
this portion of the
matrix of
H may be written as
W^n-^^nlm.m^L Slnlm^ll.
(5)*
To our approximation,
are not only eigenstates

then, the eigenstates of
2
with the values l(l+l)Jfi and
but also eigenstates of
JL*5; hence they are eigenstates of 2 = (S-h) 2 .f From 6 3 4 we see that the
quantum number j [ 2 =j(j + 1) W\ may have either of the values I + \ l - J,
except for l=Q 9 when we obtain only J= J. Since
of
and
S2
(6)
Hence
except 5 configurations split into two levels, corresponding to

= Z -|. Each of these levels is still (7 + 1) -fold
fa andJ
degenerate,
corresponding to the 2/+1 values of m[Js =mfi,]. Since aU reasonable
potential functions are increasing functions of r, the parameters
are
all
j=l+
nl
essentially positive. Therefore the level of the higher j lies
above that of the
lower; the centre of gravity of the two levels remains at

mining the centre of gravity the levels are
nl if
in deter
weighted according to their
degeneracy.
The
(one or) two levels into which each configuration is split are
together said to constitute a doublet term. The terms are designated as
2/S ( doublet S*), 2P, 2D, ...
according to the I value of the configuration
from which they
arise.
The separate
value of j as a subscript, thus *#

t
number n
levels are
2
;
P^
or the configuration label nl
symbols, and to specify an individual
a superscript, thus 2 p| or 2p
2
2
designated by adding the
Pj; D|3 2^.

2
is
state
etc
The quantum
frequently prefixed to these

the value of
is given as
P^.
2
*
represents the matrix whose diagonal elements have the constant value
nl and whose
non-diagonal elements vanish, i.e.
nl times the unit matrix.
2
2
t Not only to this approximation, but in general, are the eigenstates of
eigenstates of I , S ,
and J\ since
commutes with these three ohservables. The only
non-rigorous quantum number
is the quantum number n.
^z
SPIN-ORBIT INTBEACTIOK
4s
The
eigen-i/r s
123
3
going with these states are given by 14 6 and Table
for
3.
(8a)
#*-) = fliw-iO),
and
>
(8b)
where the
nlms mi.
<
on the right are those of (2), labelled by the quantum numbers
This transformation
is of importance in many-electron spectra,

where we shall speak of it as the transformation from the nlms rn scheme to
the nljm scheme. In obtaining (8) we have correlated $ with J 1? I with J 2
l
3
3
altering the phases in Table I according to 14 7;
non-diagonal elements (nl*L^\Sz \nl*L^)
easily verified that the

1C 3 TJiis
are positive as in
it is
L J to Jt J2 J respectively we shall adopt as standard.

For S levels one must not conclude from (7) that the displacement due to
for in this case the integral
spin is zero since Z=
diverges at the origin.
This follows from the fact that near the origin the screening of the orbital
electrons must become zero, so that U(r) must represent the full potential
2
due to a charge Ze, and hence have the form /r) 4- const. With such a
correlation of S,
(Ze"
R starts as r*+1
potential
just as in hydrogen, so that the integrand of (4)
becomes essentially 1/r for small r when 1= 0. Hence the spin displacement
of an S level is essentially indeterminate in this calculation.
shall return
,
We
to this point in the next section.

In the case of the Coulomb potential, U(r) =
5
given at once by Table 2
e*/r,
the value of
t, nl
is
With
this value of
R(
3X>2/
the two energy levels given by the spin-orbit inter
action are
nrrn
nrvn
n
.
H -F
-+
n
"
It will be noticed that the spin-orbit interaction gives a correction of the

as the relativity correction 3 5 15; using the value
same order of magnitude
124
of
ONE-ELECTBOK SPECTBA
nl there
45
found we obtain for the fine-structure levels of the hydrogen-like
atoms the formulas
-J
-tf
3/2
/2
=i
*3
C5
Fig. 2P. Fine structure of hydrogen-like energy levels.
These formulas show that the two levels having the same j value but
I have the same wave number
different values of
71
<n
er=
Q\\
o
This system of levels, which we have obtained by perturbation

theory, is the
same to this order of accuracy as that given by a rigorous solution of Dirac s
relativistic equations
spin displacements
5 5 ) and agrees with experiment ( 7 5 ). The relativistic-
Ra 2 24
4n
SKETCH OF THE RELATIVISTIC THEORY
55
of these levels are shown in Fig. 2 for ?i =
125
2
4
4
using Roc J /4?& as a unit
5
in each. case.
displacements given by 3 15 are indicated by
the broken lines; each of these splits to form a spin doublet as in (10).
Because of the occurrence of n~4 in the unit, the actual splitting decreases
The
1, ..., 5,
relativistic
very rapidly as n increases.

5.
Sketch of the
relativistic theory.
Dirac* has developed a theory of the electron in an electromagnetic field

which satisfies the relativistic requirement of invariance under a Lorentz
transformation.
The
made
characteristic feature of
linear in the
it is
that the Hamiltonian
momenta
in order that the operators d/dx

operator
an
with
the
on
be
equal footing
djdt which occurs linearly in the funda
mental equation 5 2 3.
is
In order to factorize the non-relativistic quadratic Hamiltonian into two

is found necessary to introduce some new
non-commuting
observables. It turns out that this introduces changes in the theory which
are akin to the electron spin. In other words, the spin is not introduced
ad hoc but is a consequence of the relativity requirements. If the energy
linear factors it
W are reckoned to include the rest energy
of the electron, besides

the usual energy levels in the neighbourhood of + /*e 2 the theory also gives
2
This was for some time
negative levels in the neighbourhood of
/z,c
regarded as a serious difficulty for the theory, but now it appears likely that
these are connected with the positive electron or positron recently diglevels
/^c
covered
by Anderson.f
In this section we shall give a brief account of the one-electron atom pro
blem as it appears in Dime s theory. The wave equation for an electron of
charge
e in
an electromagnetic
field
whose potentials are
9,
A is
(1)
which in the case of a central
field,
with no magnetic
field,
becomes
(2)
In these equations a and the three components of the vector $ are real
observables whose squares are equal to unity and which anticommute with
each other} and commute with the positional coordinates and momenta.
*
DIBAC, Quantum Mechanics, Chapter xn. See also a report by RUMEE, Pkys.
(1931).
f A205EBSOH,
}
Piiys.
Rev, 43, 491 (1933);
BLACKETT and OCHIAMNI, Proe. Roy. Soc. A139, 699

Ihat is, aj8 z -f jS x a=0, &&,+&,&=<), etc.
(1933).
Zeits. 32,
601
126
55
1. The quantity a we shall call the

which
relativistic effects are small the
in
of
the
In
electron.
aspect
problems
first term in (2) is the largest; hence the allowed energy levels are near to
Each quantity has the allowed values
2
pc as stated before. The
vector p is the quantum analogue of the ratio of the velocity of the electron
to the velocity of light, as Breit* has shown.
2
the allowed values of
a/ttc
that
is,
near to
The states ^ for this problem are required, in order that we may find four
anticommuting matrices, to have four component functions of position.
3
analogous to our use in 5 of ^ s with two component functions of
position to allow for the spin degeneracy.) Hence a and p are represented
by fourth-order matrices acting on the coordinates which distinguish the
(This
is
four components of ^ and these we write as a matrix of one column and

four rows. We may work with a representation in which oc is diagonal and in
which a and p have the forms
5
It follows directly from the properties of the

satisfies the commutation rule
components of p that the
vector p x p
3
comparison of this with 1 3 we see that p x p is a constant times an
angular momentum. In fact, the components of this vector are proportional
By
to the components of the spin vector of 5 3 written twice over:

(4)
that
is,
-Jt
\i
000
It
(1)
may be shown
-i
-i*
it
(Dirac,
72) that because of the
form of the Hamiltonian
the electron must be considered as having a spin angular
and a magnetic moment
momentum S
--S, exactly the values postulated by Uhlenbeck

&
flC
and Goudsmit. Sz is diagonal in the scheme we are using. Hence we may say
first component of
refers to a = 1,
component to a = l,ms = ~^; the third to a =
that the
to
<x
iff
13
m=
Sz = m/i^ -f JS; the second

= +i; and the fourth
1, ms
BEETT, Proc. Nat. Acad.
Sci. 14,
553 (1928).
55
Dirac
127
SKETCH OF THE BELATIVISTIC THEORY
Hamiltonian (2) commutes with J= 1 + S, so we may introduce a
m there are
by j and m. For a definite value of j and
values of a and two possible values of I, namely j\. The
set of states labelled
two possible
determines the state except for a function of the

Tlie fornls of these states are
radius. We denote such a state by xtfmZa ).
5
a = 1 the third and fourth components of x vanish and for
given by 4 8. For
com
a = - 1 the first and second components vanish. The non- vanishing
labelling
by (jmZa
are
ponents of
^ w- tm-
t)
(5)
where
states
this
<(Z
may
These
written for the spherical harmonic
2
5. They are not suitable in
be regarded as labelled by I or by
m^)
is
pi*
form to be eigen-^
of energy since
JL
does not
commute with the
Hamiltonian. The quantity
whose eigenvalues we denote by k, does commute with the Hamiltonian and

of k
so its eigenvalues may serve as quantum numbers. The allowed values
are
+ i) and the states labelled by (jmfcot ) which we shall denote by
(j
ijt(jmk<x.
),
are related to the x s
It is convenient
=r
*p,
now
by
by introducing the operator

the second term becomes
to transform (2)
where rQ =rjr. Since
/}*== 1
SAL-S\
The
last
form here follows from the fact that the dyad pp = 3--^
On introducing the Schrodinger
operator
-ifi grad) for p, this
becomes
128
ONE-ELECTRON"
Let us now consider the action of /?r on

tion ]3r has the form
*fs(
5s
SPECTRA
j
m&*
In the
initial
representa
sin0e-*?
cos<9
sin#e* ?
cos#
sin0e-*
--COS0
3
with the aid of 4 21
we readily
fir
find that
4t(jmkcx.
= $(jmk-(x,
(9)
).
will
The eigen-^ of the Hamiltonian going with definite values of &, j,
and
which
we
write
in
of
be a linear combination
tff(jmk-l)
ifj(jmk^-l)
the form
iGM
j?/ r \
(10)
I).
Substituting this in (2) and using
and
(8)
(9)
we
find the equations
which
determine F(r) and G(r):
These equations
may
be discussed in the usual
and seeking the conditions
way
for U(r)
for polynomial solutions for f(r)
details are given in Dirac ( 74).*

The allowed discrete energy values
found in this
=
and
Ze 2 [r by
g(r).
The
way are
(ii)
the fine structure constant introduced in 3 5 9. In this equa

tion the allowed values of s are
where
a=e2/Sc,
= 0,1,2,3,...
Q
19z, o,
,
The
relation of the
given by
. .
quantum number k
for
ior
ff\T>
4<1
K&
1>
"*>
1 .
i
to the ordinary doublet notation is
the scheme:
i=
-11-22
The levels which lie close together are those

quantum number
n=s+\k\.
-3
...
for a definite value of the total
* See also
GOBDON, Zeits. fur Phys. 48, 11 (1928);
DABWIN, Proc. Roy. Soc. A118, 654 (1928).
(14)
SKETCH OI THE RELATIVISTIC THEORY
129
The lack of symmetry of (12) with regard to the sign of k is just such that for
a given value of n we have the same doublets occurring as in the nonrelativistic theory, e.g. for n = 3 we may have
$
= 2 k=
I,
giving #j,
I,fc=2,
giving Pf,
0,i = -3,
giving
so that the lack of the allowed value s =
Df
Pj,
Df
is just what is
2
the
to
of
half
the
needed
of
term and so preserve the
prevent
appearance
structural similarity with the non-relativistic theory, (11) gives, to terms in
with positive k
the same system of levels as we found in 4 5 12; (13) and (14) show that
levels of the same n and j have rigorously the same energy.
4
ot ,
the energy levels (11) are close to H~/^c2 In the

s belonging to these levels, the third and fourth components, which
i/r
refer to oc =
1, are large compared to the other two, which refer to
states in which oc = + 1. In other words, the aspect is an almost exact
For small values of
quantum number. Besides the
discrete levels there is
an allowed con
2
2 We shall not have
and for
tinuous spectrum for
-f /zc
~/zc
any use for the exact radial functions and so do not work out the detailed
>
<
expressions.
It is instructive to discuss the Dirac equation for the central-field problem
(2) in a way which exhibits the connection with the relativistic and spin35
orbit effects as discussed in

1
01
and 4 5 Replacing
a
.
c *_ cp .
W by E + ô
2
,
(2)
becomes
+m
j|
cOjl
in which the dependence of a and p on the values of a is shown explicitly in

terms of a vector whose components are (2/#) times the spin matrices of
53 1.
The first equation
of the pair implies

c
which
may be used
to eliminate
obtain as the equation for
ifr
+ from the other equation of the pair to
*//_
--o/(r)
in
which
In actual atoms (E
U)
2
<
2/xc
f_
+ U(r) $_ = .Et/L.
(15)
7-ZT1:
everywhere except very close to the nucleus,
we simply
treat f(r) as exactly equal to 1, ^_ satisfies the

for
the non-relativistic problem without spin-orbit
Schrodinger equation
2
*since
However, if we save the first term in the
interaction,
so f(r)~
1.
If
(P*<*)
cs
=P
OKE-ELECTBOK SPECTRA
130
55
development of /(r) we get both the spin-orbit interaction and the

approximation to the relativity correction.
The
first
5 2 5,
we
term of this becomes simply /(r)
find that
first
is
approximation
Hence the equation

1
rr,
simple application of
the
By a
fif (r)
so the second term
Now to
pîfi.
first
for
is
$_
13 ^
~
A/
*
3r
(16)
The
two terms constitute the simple non-spin
non-relativistic Hamilapproximation to the effect of variable mass as

considered in 3 5 the fourth is the spin-orbit term as considered in 4 5 The
last term is peculiar to the Dirac theory and does not have a simple classical
first
tonian, the third
is
the
first
U=
-Ze z jr
interpretation. This last term, in the case of a Coulomb field,
gives rise to the following change in the energy by a first-order perturbation
calculation:
since B(OO) = 0.
is the whole radial factor R(nl)jr of the wave

term
contributes
function.)
nothing except to S terms, for they
are the only ones for which n(0) is not zero. Using the value of the radial
factor from 2 5 18, the amount contributed to the ns energy by this term is
(Here B(r )
Hence
this
Vc% a
3
~
3
n*
the value of the spin-orbit correction of S terms as given in 4 5 10

if we suppose those formulas to hold without change for S terms On the other
which
is
hand, as already remarked, the term in
terms since the integral
1
f-
U (r)ijt*dv
I>S
becomes indeterminate for S
diverges
and L
S vanishes. We now
see that that is simply a consequence of the approximation procedure which
implies
(E
U)
2
<2ju,c
Near the nucleus we can have
U)
2
>
2/zc
and
SKETCH OF THE BELATIVISTIG THEORY
55
131
S states this is important because of the relatively large probability

of the electron s being close to the nucleus in these states. If we avoid using
the expansion off(r) we see that
for
so near the nucleus the quantity

radial integral
-f
(r)
(r)
approaches a
finite limit,
hence the
B 2 dr is really finite, so this term vanishes on account
of the vanishing of L*S. Hence we have shown that the approximation pro
cedure for Dirac s equations when properly applied leads to 45 1I even for
the energy of the S terms. It is of some interest to note that Darwin had
discovered (16) prior to the work of Dirac but without the term in difj]dr;
as a result his equations gave the
to the
6.
hydrogen levels correctly except in regard
levels.
Intensities in hydrogen.
In order to calculate the transition probabilities or the line strengths in

the hydrogen spectrum we need the matrix components of the electric
P=
93
Since the results of
and
1 13
are applicable to
of the hydrogenic intensities are expressible in terms of the integrals
moment,
er.
r rK(nl)R(n
all
l-~l)dr.
The evaluation of these integrals is rather difficult. Schrodinger* calculated

them by use of the generating function for the Laguerre polynomials. Other
simpler calculations have been made by Epstein, Eckart, and Gordonf
BatemanJ has generalized Epstein s result. According to Gordon the in
.
tegral
above
i)*
-*
-1)!
4(2Z-1)!
is
equal to
l(n +
1)
(n
+ -1
1
V w _Z-l)!(rc
(
n - ri
-Z)!
(n
(1)
in which the
F s are hypergeometric functions and n
and nr are the
radial
I
Z
1, and n
respectively.
have
been
of
the
formula
written out and simplified, and
cases
Special
numerical values of the integrals have been calculated by several persons,
quantum numbers n
* SCHEODINGEE, Ann. der

Phys. 79, 361 (1926).
f EPSTEIN, Proc. Nat. Acad. Sci, 12, 629 (1926);
ECKABT, Phys. Rev. 28, 927 (1926);
GOBDOK, Ann. der Phys. 2, 1031 (1929).
f
BATEMAN,
Partial Differential Equations,
Cambridge University Press,
p. 453.
9-2
O1STE-ELECTEON SPECTBA
132
65
5
by Kupper.* In Table 3 we give some of the most important
formulas for special series and in Table 45 numerical values for the most
important values of the quantum numbers, the latter being based on Kupper s work as revised by Bethe.f The expression (1) is not applicable for
n = n. In this case the integral is easily found to have the value
most
fully
%ann*-l*.
J
(2)
This integral, of course, does not occur in the intensity problem in hydrogen,
but is important in the Stark effect (Chapter xvn).

Let us now consider the strengths of the lines in the fine structure of
hydrogen.} We could go at this directly by calculating the matrix com
5
ponents from the eigenfunctions (4 8): this would provide ample exercise in
calculation of matrix components. But it is more instructive to work them
out as an application of the general results of Chapter m. In applying 1 1 3
we identify J^ with the electron spin, so j^ = f and JF2 with the orbital angular
ll 2 (L l)^/ are
momentum, so jz = The strengths of the lines n 1 2 L^
,
I.
given by
->n
7 45
in terms of the quantities (nl Lj\P\n

3
of these on
and,; is given in 11 8, where
dependence
f
in terms of the (nl\P\ril
(I/
1)^).
The
they are expressed
1). According to the discussion just following

11 3 8, this quantity is really independent of the spin and so, as in 11 3 9, we
may express it in terms of the nlm8 l scheme of states in which the spin
and orbital angular momenta are not coupled. Since z = er cos# we have
(nlm$
m^ Pz \n ll m8 m^
f
= -e \r R(n 1) R(n
Jo
3
Using 4 2I,
we integrate
(nlms mi\Pg \n
1-1} dr foos0
Jo
over 9 to obtain
c
-==
lmg raj)=
(n I
Similarly
V4?
rR(nl)R(rill)dr*
J
m m Pz \n m
s
m^ = 0,
so for a non -vanishing component I must change
by one unit. Using 9
1 1,
we
hence, finally,
(3)
iJo
*
KDPPEK, Arm. der Phys.
86, 511 (1928);

SUGIUEA, Jour, de Phys. 8, 113 (1927);
SLACK, Phjs. Rev. 31, 527 (1928);
ÎAXWELL, Phys. Rev. 38, 1664 (1931).
t BETHE, Handbuch der Physik 24/1,
ed., 442 (1933).
j SOMMEEFELD and UNSOLD, Zeits. fiir Phys. 36, 259; 38, 237 (1926);
BECHERT, Ann. der Phys. 6, 700 (1930);
SAHA and BAÊEJI, Zeits. fur Phys. 68, 704 (1931).
2<*
INTENSITIES IN
HYDROGEN
133
4
quantities occurring in 7 5 are thus completely expressed in terms of
over
the
radial
the integrals
eigenfunetions.
The calculations may be exemplified by a detailed consideration of the
The
line strengths in
the fine structure of

r-
TABLE 3 s
Values of\
np
the ensemble of the
n~
-*
r*
I
L/o
nl
Ha
E(n
1}
R(n I- 1) dr
in atomic units for
n f ^ n.
J-l
Is
2s
3*
4s
5s
2 83
?t
- 27) 2 (n
nd
w -f S)- 4){n - 3) 3n- 9 {

- 4}{%2 - 9)(n - 4) 2n ~ 10 (7i
U
U
U
TABLE 45
Fofees of
(re 1)
R(n I- 1) dr
in atomic units.
3->^
=2
2$->
TMs
made up of three
2
2 P. Using 11 3 8 we have
transitions.
65
OKE-BLECTBON SPECTRA
134
is
(3
doublets
3 2 P->2 2 $, 3 2Z}->2 2
P and
Df:P:2 P|) =
2
Df :P:2 Pf ) =
(3
(3
D|:P:2
2
From Table
hence from
35
and
7 45
we
(ScZjPjSj?)
(3),
J.O
22238
==- ;
find for the strengths of the lines,
S(3
(a)
(b)
2
D>->2
(c)
(d)
(e)
(f )
(g)
1 2 2
24 7
P|) = 9.2 3 5- % a
Pf)=L
=
S(3
P^) 5.
2
2
= 2.2 21 3 5 5~ 12e2 a 2
8(3
j)
2
2
8(3 Pj-^2 $j) = L
2
2
= 1.216 3 7 5-12e 2 a 2
8(3 5j^2 Pj)
2
2
S(3
Pf) = 2.
S(3
(6912)
768)
Df->2
In the same way,
=6 a
D|->2
(3840)
Pf~>2
(1
600)
800)
j->2
75)
150)
(4)
It will be observed that the relative strengths of the lines in the same doublet
bear simple integral ratios, 9:1:5 and 2:1. This is a special case of a general
result for multiplets discussed fully in 2*. The relative strengths of the
different doublets are not so simple since these involve different radial in
in this table give the relative

figures in parentheses at the right
14 2 2
16 6
factor
common
the
Iine
each
of
3
being 2 3 5~ e a .
strength
transition
next consider the theoretical values of the Einstein
tegrals.
The
We
line these are expressed in terms of the

to express a with the Bydberg constant as
convenient
strengths by
e2 a 2 Then the numerical coefficient
in
atomic
the
and
units,
unit,
strengths
For a particular
probabilities.
7*3. It is
in
743
becomes
^4^402
-
where
oc
is
eg a 2
= JaV-1 = 2-662 x
10 9 sec" 1 ,
(5)
the fine-structure constant and r the atomic time unit (see
Appendix). The absolute values of the transition probabilities involved in

the first Balmer line are therefore
(b)
8
2
2
=
A(3
P|) 0-643 x 10
2
2
A(3 D|-^2 P|) = 0- 107
(c)
A(3
(a)
(d)
(e)
IV2
A(3
A(3
(f)
A(3
(g)
A(3
2
Df->2
sec"
P|) = 0-536
#) = 0-223
2
= 0-223
#j)
2
=
P|) 0-021
2
=
j-*2 Pf) 0-042.
2
P|->2
P|->2
/5j->2
(6)
65
INTENSITIES IN HYDBOGEIST
135
The values illustrate several important properties of the A s. The sum of the
two transition probabilities from 2D| is equal to the single transition pro
2
bability from J5f This is a general result true in more complex spectra
9
2 ). The two transition probabilities in 2 P- 2 $ are equal, but the intensity
(
ratio of the two lines is 2: 1 because the A s have to be multiplied by the
.
give intensities. On the other hand,

in
the
transition
for
appears
probabilities themselves,
so that in both cases the intensity ratio is 2: 1 although in the first case it is
statistical
2
$->
weights of the
P the ratio
initial levels to
2: 1
due to a factor in the
statistical
weights while in the second it is due to a
factor in the transition probabilities. The quantities (2/4- 1)A, which are
proportional to the strengths, are thus symmetrical with regard to initial
and
final states, while the
A s are not.
Suppose we have an experimental arrangement whereby at 2=0 equal

numbers of atoms are put into each of the 18 states belonging to n = 3 and
that the atoms are not subject to any external disturbance. Then the relative
intensity of the different doublets would depend on the time after the
excitation at which the observation is made. The number of atoms in any
level A after time t would be reduced by the factor e~ k^ where JcA = SA(-4, B)
B
the sum being taken over all of the levels B of energy lower than A. In the
states can only go to 2p,
particular case of the first Balmer line, the 3s and
,
but 3p can go to
Is as well as 25.
A(3
for either 3 2 Pf or 3
P~I 2
We may calculate that

j)
= 1-64 x
10 8 sec-1
Pj , and so the mean lives [the reciprocals of the total

transition probabilities kA *=k(nl)] of each level of the three initial con
figurations
have the values

Configuxation
3a
l-o6xlO~ 8 sec
3<2
mean
lives of the
xlO- 8 sec
0-54 x 1G~ 8 sec
3j?
Since the
Mean life
16-0
two component
levels of
P and 2
are equal
from
them
will not
of
lines
the
the
relative
intensity
respectively,
originating
in
of
lines
levels
with
time.
But
the
relative
the
change
intensity
originating
of different
for a
much
L value does change with. time.
The
$-2 2 P
Z>
will
be emitted
longer time than the others on account of the greater lives of
their initial states.
Such an experiment has never been performed. The actual studies of the
made with an electrical discharge where
the excitation and emission reach a steady state. The atoms are subject to
disturbing electric fields and to collisions which tend to produce a mobile
relative intensity of the lines are
136
equilibrium between the states of differing I and the same n. The result is
that when the disturbance is sufficiently great all of the states of the same n
have the same mean life, corresponding to a total transition probability

k(n) which is a weighted average of the individual total transition
probabilities:
TABLE 5s, Transition
* This
is
2 A(y
->
n Zj
),
probabilities
and mean
a quantity which
is
lives for hydrog&n,.
independent of the value of j.
INTENSITIES IN
HYBEOGEN
137
In Table 5 some of the important values of the transition probabilities

for hydrogen are given together with the mean lives of the individual states,
and in Table 6 5 are given the average mean lives
for several values of n.
discussion of the experimental results of intensity studies in

taken up in the next section.
TABLE
7.
6 5 . Average
mean
life
The
hydrogen
is
lf&(n).
Experimental results lor hydrogenic spectra.
Let us consider briefly the relation of the foregoing theoretical results to the
5
experimental data. We have found in 2 that the energy levels in hydrogenie atoms are given
by
In spectroscopy we are concerned directly

with the wave-numbers of the lines which are obtained as the differences of
where
/x,
is
the reduced mass
I 5 ).
the term values
p__#2
RH
R TT
p
^HT^^
(}\
A
is known as the
obtained
The
by using the
quantity
Eydberg constant for hydrogen.
reduced mass of the electron with tespect to any other nucleus is known
as the Rydberg constant for that atom and we write R^ (or R) for
The
universal constant
occurring in this expression
the Rydberg constant obtained

formula.
by using the
electronic
mass in
this
The spectrum of hydrogen divides itself naturally into a number of series

*
of lines
each
series
being the set of lines having a
common final n
value:
Wave-lengths (.Angstroms)
Name
Lyman
2
3
Pascken
Brackett
Final
The Lyman
series lies
minent feature of the

infra-red.
Balmer
deep in the ultra-violet, the Balmer series is the pro

hydrogen spectrum, the other two are in the
visible
138
known atomic constants.

data made by Birge* led him to adopt the values
The Rydberg constant

critical
study of all the
7s
one of the best
is
R H = 109677-759
R He = 109722-403
Rco
0-05 em.-1
0-05
=109737-42
0-06
The smallness of the difference between R H and R He means that some Hell
lines are
HI lines.
very close to certain
and
close to
it,
in the
Thus
for the first
Balmer line,
Ha
R H I_-_1
Hell spectrum,
is
By making an accurate measure of the

the ratio of the nuclear masses as
separation of these lines and taking
known from the chemical atomic-weight
determinations, it is possible to measure the mass of the proton in atomic

units, that is, the ratio of its mass to that of the electron. The most accurate
study of this kind has been made by Houstonf who finds

JfH //t= 1838-2 1-8.
Taking this in combination with the Faraday constant in electrolysis, the

value of e/jLc for the electron becomes
e//Lt
This
= (1-7602
7
0-0018) x 10 EMIT.
good agreement with the recent precision measurements on free

electrons in deflecting fields by Perry and Chaffee ? and Kirchner.J Thus the
theoretical effect of the finite mass of the nucleus is fully verified. This effect
was the means used in the discovery of the hydrogen, isotope, deuterium 5 by
is
in
Urey.
The
relativity-spin structure for
hydrogen and Hell
is
very
fine,
the
distance of the extreme components of the nth energy level being
5
according to 4 12. For the final states of the Balmer series in hydrogen
Aa= 0-365 cm-1,
while the states of n = 4 in ionized helium have a splitting three times as great.
Since the higher levels show a much smaller splitting, the observed pattern
in a Balmer line is that of two close groups of lines with this separation.
*
BIBGE, Rev. Mod. Phys. 1, 1 (1929).

f HOUSTON, Phys. Rev. 3D, 608 (1927).
j PEHRY and CHASYEE, Phys. Rev. 36, 904 (1930);
KracHNER, Ann. der Phys. 8, 975 (1931); 12, 503 (1932).
UBEY, BEICKWEDDE and MUBPHY, Phys. Rev. 40, 1, 464 (1932);
J.
Chem. Phys.
1,
512 (1933).
EXPERIMENTAL RESULTS FOR HYDROGEKIC SPECTRA
139
Sommerfeld and Unsold* have discussed the modern interpretation of the

fine structure for several lines.
We confine
our attention to
Ha
The theo
5
pattern for this line is shown in Fig. 3 where the ordinates are
proportional to the relative intensities of the lines as calculated in the
retical
preceding section.
The broadening of the

velocities of the
lines
by the Boppler
atoms in the source
is
due to the thermal
effect
considerable, so the experiments are

air. According to the
performed with the discharge tube immersed in liquid
s
Pig. 3
Theoretical fine structure of
Ha
theory of the Doppler effect, an atom emitting wave-*mmber o- if moving

toward the observer with speed u will appear to emit radiation of wavenumber a where
is according to Maxwell s law for absolute
atoms whose speed toward any direction
the
fraction
of
all
temperature T,
If the distribution of velocities
is
between u and
u+du is given by
2-rrkT
hence the intensity of radiation between a and
mc
a+da is proportional to
-^dr-,
dcr.
^fcroIX
r~i^7
* SOMMEBFELD and UNSOLD, Zeits. fur

Phys. 36, 259; 38, 237 (1926);
see also SLATER, Proc. Nat. Acad. Sci. 11, 732 (1925).
140
Numerically, if T is expressed in thousands of degrees and

atomic weight units,
a/(7o== -958 x 10-V5>.
75
m is expressed in
The
liquid-air bath does not prevent some warming of the gas during the
discharge and the observed widths correspond to a temperature of about
200 K., that is, a~0*07 cm*1 , so the width of the lines is comparable with
some of the intervals in the pattern. With this in mind we expect that (g)
will be too weak to show, that (a) and (b) will be lumped together and that
(e) 4- (f ) will
produce an asymmetry on the curve produced by
(c) -f (d).
exactly what was found experimentally.* Especially important is

the occurrence of (e) + (f), since on the original fine-structure theory of
This
is
Sommerfeld this component is forbidden. Its presence therefore lends support

to the present interpretation of the fine-structure levels. The relative
intensities are found to depend somewhat on the discharge conditions.!
The relative intensities of the absorption of the first three Balmer lines
has been studied by Snoek. J He found the relative values to be, for the two
fine-structure groups arising from the two initial levels,
H a :H
j3
:H y = 100:
18-8: 7-4
100:20-2:8-5
The theoretical ratios
for absorption
(from j = f )
(fromj==|).
from the J = f or from 2p are
100:17-6:6-3,
while those for absorption from the
2si
are
100:24:9-9.
The experimental values from J=f are in good agreement with the theory
and those from j = | correspond quite closely to a mean between the theo
retical values for 2s and 2pi
The same question was studied by Carst and Ladenburg by measuring
the amount of the anomalous dispersion around H a and
H^. The uncertain
element here is the amount of correction for negative dispersion due to
atoms in the n = 3 and n = 4 states. They conclude that the oscillator strength
ratio is between 4-66 and 5*91 for H a
are
:H0. Since the oscillator strengths
t
proportional to the absorption,
we
see
from the foregoing
figures that the
* The
principal papers are;
EANSEN, Ann. der Phys. 78, 558 (1925);
KENT, TAYLOR and PEARSON, Phys. Rev. 30, 266 (1927);
HOUSTON and HSIEH, Phys. Rev. 45, 263 (1934);
WILLIAMS and GEBBS, Phys. Rev. 45, 475 (1934);
SPEEDING, SHAKE and GRACE, Phys. Rev. 47, 38 (1935).
t See BETHE, Handbuch der Physik, 24/1,
ed., 452 (1933).
j SNOEK, Diss. Utrecht (1929); Archives Neerlandaises, 12, 164 (1929); Zeits. fur Phys. 50, 600
(1928); 52,654(1928).
CABST and LADENBTJBG, Zeits. fiir Phys. 48, 192 (1928).
2<*
EXPEBIME3TTAL BESULTS FOB HYBBOGEOTC SPECTRA
75
141
theory gives 4-16 if all atoms are in the 2s states and 5-68 if all are in the
2p states, while the average according to statistical weights is 5-37.
Ornstein and Burger* studied the relative intensity in emission of Balmer
and Paschen lines having the same initial states so as to eliminate the un
certainty as to the number of atoms in the initial states. The results were
Obs.
H^/Pa
Hy/Pp
5 /Py
Theory
(4->2):(4->3)
2-6
3-55
(5->2):(5-^3)
2-5
3-4
(6->2):(6-3)
2-0
3-2
theoretical values are based on assumption that each initial state has
number of atoms in it. The departures indicate that this is not the
same
the
case and that there must be an extra supply of atoms in/ states since these
contribute to the Paschen lines but not to the Balmer lines.
An interesting absolute measurement of the life time of the excited states
of He + was made by MaxwelLf He excited the helium by electron impact
occurring in a narrow electron beam. The excited ions produced were drawn
out by a transverse electric field which did not sensibly affect the original
The
electron
field.
beam because of the presence of a controlling longitudinal magnetic
The ions move various distances in the transverse field before radiating,
corresponding to the probability distribution of life times. By studying the

spatial distribution of intensity of light he was able to infer the mean lives
of the excited atoms. For the n = 6 states of He + he found a mean average
8
8
life of (1* 1
0-2) x 10~ sec, while the theoretical value is ! 17 x 10~ sec. The
mean life is -fa that of hydrogen since the frequency cubed varies as ZQ and the
radius squared varies as 2~ 2 inthe expression for the transition probability.
8.
General structure
of the alkali spectra.
In the next chapter we
electrons in the field of a nucleus
number of electrons of each

a given value of nl the
problem of the motion of

see that there is a limit to the
shall formulate the
and
shall
nl value that can occur in the atom.
maximum number is
present,
we speak
When for
of these as
6
forming a closed shell ( 5 ). If we have an atom in which all the electrons
but one are in closed shells, the mutual interaction of the electrons is greatly
and the energy-level scheme is, to a good approximation, just that

of a single electron moving in a central field ( 10 6 ). This effective central
field for the extra electron outside closed shells is the resultant field of the
simplified
nucleus and of the other electrons, so for neutral atoms it is of the form
2
e2/r at large distances from the nucleus and is equal to
(Ze /r) + C at
small distances, where G is the constant potential at the origin due to the
electrons in closed shells.
* OKSTSTEIN and BUBGER, Zeits. fur
Phys. 62, 636 (1930).
f MAXWELL, Pfcys. Rev. 38, 1664 (1931).
8s
ONE-ELECTEON SPECTRA
142
I5
we may apply
and classify the

will
be those corre
factors
n
I. The
and
numbers
angular
by quantum
Since this
states
is
a central-field problem,
sponding to precise Tallies of angular momentum. Since the potential energy

at each distance is less than in the simple Coulomb law, the energy corre
sponding to each nl is lower than for the corresponding hydrogen state.
Moreover, since the difference between the potential energy and the Coulomb
law is greater at small values of r, the difference is greater for smaller values
ofZ.
It
is
convenient to discuss the empirical data

quantum number n* defined by
by
introduction of the
effective
-R/7i*
(I)
HYDROei/t-
UK
65-
Energy level diagram for sodium. The S series is known to n= 14, the P to
n = 59, the D to n = 15, and the F to n = 5. The only doublet terms which have
s
Fig. 4
"been
resolved are the
first
seven of the
P series.
where R is the Rydberg constant for the atom in question. The dependence
of ri* on n is almost linear, so that the difference (n n*) = A, which is called
the quantum defect, is almost constant in a particular series (constant I,
5
varying n) of terms. These facts are illustrated in Fig. 5 where the empirical
quantum defects are plotted for the various series of the alkali
Each
spectra.
plotted point on this scale is big enough to cover the variation
of A with n along a series. This shows clearly that the/ terms even in Cs are
approximately hydrogen-like although all others in Cs show large displace
ments from the hydrogenic values.
values of the
143
GENERAL STRUCTURE OF THE ALKALI SPECTRA
The much smaller dependence of A on. n in a given series is shown in
s
series. This
Fig. 6 for several typical
A for the first member of the series, as
increases with
n in the d series and
shows the difference, A(wZ) minus the

a function of n. It will be noticed that
decreases in the s
and p
series.
The various empirical formulas for the terms in a spectral series which
are in use among spectroscopists may all be regarded as formulas for the
dependence of n* or A on n. The simplest
the Rydberg formula which simply
regards A as a constant. Other special
is
formulas which are in use are

(Ritz),
(Ritz),
(Hicks).
In the
first
of these the deviation of
from constancy is proportional to the

term itself which means that the quantity
represented
is
<
contained implicitly in the
formula which represents
it.
In considering the variation of the

one-electron spectra along an iso-electronic sequence, it is desirable to take
account of the large dependence on the
degree of ionization by writing
s
of
Fig. 5 . Variation with I and
quantum defect in alkali spectra.
(2)
as the equation defining n*, where ZQ is the net charge of the Ion around which
the single electron moves (one for neutral atoms, two for singly-charged ions,
etc.). For higher values of the nuclear charge in an iso-electronic sequence
the departure from the Coulomb law caused by the inner group of electrons
is smaller in proportion to the whole potential energy. Therefore we expect
the terms to become more hydrogen-like as Z increases in such a sequence.
This means that the &(nl) tend to zero as Z Q increases. This is illustrated in
5
Fig. 7
where
A for the s and p
Mgll, A1III, and
terms of the eleven-electron sequence,
Si IV is plotted against
Na I,
In the orbital form of the atomic theory there was a sharp distinction
between penetrating and non-penetrating orbits, the former being those
which extend into small enough values of r that the field has a non-Coulomb
character. This distinction is not so sharp in
quantum mechanics
as all of
the eigenfunctions have non-vanishing values near the origin. Nevertheless

the states of the higher I values penetrate less, as in the old theory. For the
non-penetrating orbits the old theory gave an explanation of the slight
OKE-ELECTEOK SPECTRA
144
85
departures from hydrogenic values by assuming that the core of the atom,
composed of the inner groups of electrons, is polarized by the field of the
outer electron, so that the electron
- cte2
motion was perturbed by an interaction
the polarizability of the core. Using values of a

inferred from ion refractivity data, this gives a fairly satisfactory account
term,
/2r*>
where a
is
of the deviations.
The
alkali spectra have, of course,
spin-orbit interaction discussed in

2
I7 _
2
.
>
z_|, i.e.
a doublet structure caused by the
45
Since
nl is
the level with higher j value
is
essentially positive,
higher in energy. This

,0
O"
-O
in alkali spectra.
tween the quantum defect of

to which it belongs is plotted
numbers of these terms.
called normal.
.at
The
difference
of the lowest term of the series
For fields not deviating too greatly from Coulomb
order
is
fields
we may expect the value of the doublet
by
interval to be given roughly
4 5 11, which can be written in the form

:
If this form has
5-822 -
any kind of approximate validity
-cm~
for the alkalis,
(3)
we expect
G-EKBBAL STRUCTURE OF THE
SPECTRA
145
that the doublet Interval will vary about as 1/n The absolute value of the
Intervals we do not expect to be given at all by the formula. The approximate
.
variation as i/% 3 is shown in Fig. 8 5 in which Is plotted, on logarithmic

scales, the doublet interval as ordinate and n
,
as abscissa for
alkalis.
From
the variation
aH the known doublets of the

main
3
like 7i~ or ^~ 4 but the diagram
this it is clear that in the
is
shows some striking exceptions, thus far

unexplained. For example the D series of both
Rb and Cs show a remarkable discontinuity
also
<
in their values so that hi the case of

tt
=11 the
interval takes a
Rb
for
jump from an ex
1
up to 6 cm"1 which
pected value of about 1
is twice as big as the interval for n = 5, the first
member of the series. In the case of Cs the
cm""
intervals for
= 13 and
14 take a similar sudden
jump upward and become nearly equal

P intervals for the same T& S.
5
Fig. 8
4-
to the
j^
t
7 s,
(0
il
12
&
A in
iso-electronic se-
Variation of
he Na-like
14
Variation of doublet interval with n in alkali spectra.
ONE-ELEcmoisr SPECTRA
146
Data on doublet
interval is
doublet
85
series are scarce, indicating that the

case
the
of Cs it is definitely known that the
in
usually small. But
terms are inverted, that is, the level with the higher j value has
intervals for
the lower energy. The separations, however, are very small:
The
first
order perturbation theory
this inversion since
is
is
quite inadequate for explanation of
essentially positive.
We turn now to the relative intervals of corresponding
terms in
D and P
According to the hydrogen formula these intervals should be as 1:3.

expect them to be less than this in the alkalis, since the p eigenfunctions
series.
We
have appreciable values for smaller values of r than the d eigenfunctions. Since
the deviation from the Coulomb law is in the sense of stronger fields than
Coulomb at smaller distances, this tends to make for a p state greater than
for a d state in a larger ratio than for hydrogen. That those expectations
are borne out is shown in a table for Cs:
10
0-078
0-115
0-135
0-182
0-164
12
11
0-192
0-096
13
0-964
14
1-02
The anomaly of the
last two terms, already mentioned, is here striking.

In the case of the iso-electronic sequence Li I, Bell, etc. the interval for
2p is known and agrees quite well in absolute value with the value given by
the theory for hydrogen:
Various somewhat empirical modifications of the doublet interval formula
have been proposed. Landef from an analysis of the classical penetratingorbit picture proposed replacing Z* by Z\Z\^ where Q is the net ion charge
as in (2) and Zi is an effective nuclear charge in the inner region to which the
(3)
and replacing nz by n**. Another modification that has

been used a great deal by Millikan and Bowen J regards the entire orbit as
determined by a Coulomb field of effective nuclear charge (Z S), where 8
orbit penetrates,
f LAOT>&, Zeits. for Phys. 25, 46 (1924).

j MTT.TJgAK and BOWEN, Phys. Bev. 23, 764 (1924); 24, 209, 233 (1924).
GENERAL STRUCTURE OF THE ALKALI SPECTRA
85
147
represents the amount by which the full nuclear charge is diminished by

screening of the field by the inner electrons. If one chooses values of S to fit
the observed 2p intervals in the Li I to OVI sequence, they are found

(third row in the above table) to be nearly equal to two, which is a reasonable
value. In the older theory the formula (3) was derived as a relativity effect
and therefore should have applied to the interval between s and p terms
P interval. This puzzle was solved by the interpretation

as due to the spin-orbit interaction.
formula
doublet
of the
rather than to the
Intensities in alkali spectra.
9.
we have found a central field which gives a satisfactory representa

of the energy levels of a one-electron spectrum, we may use the eigen-
After
tion
functions of this field to calculate the radial integrals
r r(nl)R(n
l-l)dr
J o
which are needed for the intensity problem. The theory goes exactly as for
5
hydrogen ( 6 ) with these integrals replacing the ones based on hydrogenic
eigenfunctions.
The first point to be noticed is the 2: 1 strength ratio in the principal

n 2 P 2 $, which was already calculated in our discussion of hydrogen.
This value* is observed for the first doublet in Na and K. For Rb the ex
series
~>
perimental ratio for the second doublet is 2*5 while in Cs the ratios for the
second and third doublets are 4-0 and 47, although it is 2-0 for the first
doublet. The explanation of these departures, due to Fermi, is considered
in515
Next we consider the relative strengths of the doublets in the principal

series. Most of the experimental data is expressed in terms of effective
f values for the unresolved doublets, defined in
way exactly
analogous to
For the doublet of wave

number a connecting the low-energy configuration nl and the higher con
the definition in 9*14 of the f values for lines.
figuration
where S
is
nZ
~~
l*\$(n l
,nl)
the total strength of the doublet. Calculations of the f values

np of Li have been made by Trumpy f using
on a Hartree field ( 814 ) The ratios of the f values
for the principal series 25

radial eigenfonctions based
of successive doublets were measured by FilipovJ by the anomalousdispersion method. The striking thing is the very large ratio, 136, of the
first
*
5
to the second doublet. In Table 7 the experimental values of Filipov
The experimental data are summarized by KORFF and BREIT, Key. Mod.
TEUMPY, Zeits. fur Piiys. 61, 54 (1930); 66, 720 (1930).
t FILIPOV, Zeits. fur Piiys. 69, 526 (1931).
Pliys. 4,
471 (1932).
OKE-ELECTEON SPECTRA
148
95
are given as calculated from the measured ratios assuming the first one to
theoretical value calculated by Trumpy. To show how much the
have the
values are affected by the departure of the Li eigenfunctions from the

functions, we have given in the last column the values obtained by using
hydrogenic radial integrals together with the Li frequencies in the theoretical

formula.
TABLE
7 s.
f values in the Li 2s - np
Similar calculations for the 3s
np
series in
series.
Na have
"been
dependently by Prokofjew, Trumpy and Sugiura.* ProkofJew
made
in
method of
14
obtaining the central field is discussed in 3 ; Sugiura s method is similar.
Trumpy used a Hartree field. The experimental ratios were measured by
Filipov and Prokofjewf and as to. the table for Li have been reduced to
absolute values by assuming agreement with Trumpy s calculations for the
These are given in Table 8 5 The small inset table gives the values
by Prokofjew for several other doublets.
Values for the principal series of Cs (for which no theoretical calculations
have been made) and of the ratios of the first two doublets of the principal
first line.
as calculated
and Eb are given in the report by Korff and Breit

has
calculated the spontaneous transition probabilities
Zwaan|
series in
8
1
A(4p-*3d) = 0-13 x 1C sec-
Ga II using
The values are
for
and
A(4p-*4a)
= 1-55 x
(loc. cit.).
10 8 sec"1
14
eigenfunctions of a field determined by the method of 3
of interest in connection with Milne s theory of the support
.
of calcium in the solar chromosphere by radiation pressure.

*
PBOKOFJEW, Zeits. fur Phys. 58, 255 (1929);

TRCMPY, ibid. 61, 54 (1929);
SUGTOKA, Phil. Mag. 4S 495 (1927).
t FILIPOV and PROKOFJEW, Zeits. fur Phys. 56, 458
(1929).
j ZwAAtf, Biss. Utrecht, 1929.

HIKN~E, Monthly Notices B.A.S. 84, 354 (1924); 85, 111 (1924); 86, 8 (1925).
AT,TTATT
INTENSITIES IN
95
TABLE 8 5 f values
.
in
tUe,
149
SPECTBA
NaZs-np
series.
10. Zeeman. effect.*
When an atom with one valence electron is placed in a magnetic field, the
characteristic
Zeeman
energy levels are split into several components, giving

which produces these displacements con
patterns. The interaction energy
the spin of the electron and that arising
from
sists of two parts, that arising
motion.
3
the spin hypothesis ( 5 ) an electron has a component of magnetic
moment of magnitude + eS/2/xc in the direction in which the component of
of magnetic
a
is
%K. Since the energy of particle
spin angular momentum
of
the
interaction
the
is
magnetic
in a field
energy
moment
from
its orbital
On
electron
may
M%^
be written
(1)
The effect of the magnetic field on the orbital motion of the electron is best
=
Haniilobtained by using the vector potential A, where Jtf euxlA. The
obtained
is
A
of
field
by
e in the
potential
tonian for a particle of charge
the momentum p in the Hamiltonian for the
writing p -f (ejc)A in place of
2
/2^ we have
case without the magnetic field, t Then, in place of the term
>
Now since p=
-fftgrad,
i
we
obtain, using a formula of vector analysis,
= - ifi drr(Aifr) = - iffy div-4 - ift^-grad^ = A*pff
* HEISENBEKG and JORDAN, Zeits. fur Phys. 37, 263 (1926);

C. Q. DABWEN-, Proc. Roy. Soc. A115, 1 (1927);
K. DARWIK, ibid. A118, 264 (1928).
26-27. (Note that their development
f See CONDON and MORSE, Quantum Mechanics, pp.
particle of charge +e.)
is for
OKE-ELECTBOK SPECTRA
150
if
we clioose A
"be
a solenoidal vector
as
(i.e.
so that
10 5
divA = 0:
done). Hence the energy due to the magnetic field
this
may always
may be written as
field 3%?, we may take A = \(^f x r).*

We shall now make a rough calculation of the order of magnitude of the
For a uniform magnetic

terms in
which
(2)
will
show that the second
is
entirely negligible for all
thus far used in the laboratory. In atomic units (see Ap

= 1, c= 137, r and p for an electron are of the order of 1.
/z
fields
magnetic
pendix), e= 1,
The usual magnitude of ffi in Zeeman-effect measurements
is from 10,000
to 30,000 gauss, i.e. from 0-6 to 1-8 x 10~3 atomic units. The larger of these
values gives .4~0-9 x 10~ 3 atomic units, and the first term of (2) ~0-65 x 10~ 5
atomic units of energy
1-5
cm"
1
.
visible is of the order of 0-001
200,000 gauss (which

in using), the second
is an energy which may easily be

limit of spectroscopic accuracy in the
This
measured spectroscopically, since the
cm-1 On the other hand, even if we take^ as

.
is about the highest field anyone has ever succeeded

term represents an energy of only 0-0002 cm-1 This
.
magnitude certainly at present undetectable.

Hence weshaEconsideronly the firstterm of (2).
is
this interaction
A-p
*=
\ic
When
field
this is
Whenweset-4 =
becomes
~^f x r-p* = ^^*r x *p =
added to
-&!*.
(3)}
v
2/iC
2jjLC
(1)
we
obtain as the whole effect of the magnetic
the energy
(4)
In particular if we take the z axis in the direction of

energy becomes
where o
is
coefficient proportional to the
magnetic
field
this interaction
strength
Thus the energy
(5) arising from the interaction of the atom and the mag
diagonal in a representation in which Lz and Sz are diagonal,
in the representation 4 5 2.
netic field
i.e.
To show
is
that this represents a uniform magnetic
field
^, we have
r) =
&f is constant, the first two terms are zero. The third term is and the fourth 3^, hence
curL4 =3if. Similarly div-4 =0 as required.
Note
that
this
is 2?r times the o as
f
usually defined (the classical Larmor frequency).
Since
ZEEMAN EFFECT
105
151
Weak-field case.
In the case in which the Zeeman splitting is small compared to the spin
on each of the two levels
doubling we can consider (5) as a perturbation
obtained by the spin interaction separately. We must thus calculate the

M
matrix components (nl*Lf\H \nl*Lf) for the different degenerate states
5
of the level nl ^Lj These are easily obtained from 4 8:
.
2Z-M4-1
,.
"
of the different
already diagonal so that the displacements
2
are
level
Li
given by
components of the
This matrix
Zeeman
is
(8)
2Z
The
factor
+ 11
is
Ztb
~T"
known
Lande
as the
This derivation holds
g-factor.
for 5 levels as well as for those with
0.
>
4 8 represents the eigenThus, for weak fields, to the accuracy with which
of the spin doublets, each level is split symmetrically
s of the
components
to the
into 2j-f 1 equally spaced states, the splitting being proportional
total
the
of
value
the
of
quantum
field, and being independent
magnetic
number n and of the atom in question. The interval between the states after
which depends only on I and j. For
splitting is determined completely by g,
a given I, the level of higher j is more widely split than the level of lower j
.
A table of y-values is given below:

g-values for doublet spectra.
n+i
*S
*P
Hî -
*F
f
*
In terms of wave-numbers the displacement
*B
H
*
(8) is
given
by
(9)
where
is
in gauss.*
The accuracy with which
* Not the accuracy of our estimate of 1-5

of 30,000 gauss.
cm"
this
formula holds for
as the magnetic perforation caused
by a
all
field
10 5
OKE-ELECTBOISr SFEOTEA
152
was early discovered experi

one of the methods
mentally, and is expressed by Preston s Rule* Indeed,
of obtaining e//i is by the accurate measurement of this Zeeman effect.
doublets of
all
and
one-electron atoms
The strong-field case
PascJien-Back
ions
effect.
We have been considering the case where the Zeeman effect is small com
the other extreme,
pared to the spin interaction. We shall now consider
the spin-orbit interaction so small,
that the spin splitting is negligible compared to the magnetic splitting. In
this case we start with the degenerate configuration nl and the system of eigen
where the magnetic
field is so strong, or
functions ^(nlm/n^ in which
HM is diagonal. The magnetic energy
is
then
(10)
(^^m^|o(J,+^)|^msm = ô(m+ms ).
Hence each configuration is split symmetrically into 21 + 3 equally spaced
(Z -f 1),
components corresponding to the 2Z -f 3 possible values, Z + 1, Z,
are
lowest
two
of m + m 5 = Wj-h2ms ). Of these the two highest and the
given
by
z)
. .
.,
to the
non-degenerate, while the others are doubly degenerate corresponding
- 1)
1
+
+
-f
of
value
-f
a
(
2)
),
(m,
two ways of obtaining given
(m^ ( 2m,) (mz ) (
.
If we now superpose a small spin-orbit interaction we get a displacement for
the non-degenerate components represented by fr- (wyraj|L-S| wynj) = mzras

3
we note that because of the
(cf. 7 3). For the degenerate components
in 7 3 3 there are no matrix components of L*S joining the two
2
S(m
states, so that
may
not
split
each state
is
displaced
by the amount m#n
which
may or
the component.
This state of
affairs is
known
as the Paschen-Back effect, the eigen-
functions in this case being those labelled
The
tf
by
transition case.
We have been able to determine very simply the Zeeman splitting in two
extreme
which the magnetic interaction is small compared to

in which case the eigenfunctions are labelled by j and m; and
cases, that in
the spin-orbit,
that in which the magnetic interaction is large compared to the spin-orbit,
mz In order to
in which case the eigenfunctions are labelled by
s and
must apply the
we
cases
extreme
two
these
between
obtain the transition
spin-orbit
and magnetic perturbations simultaneously, treating
H!+Hv^(r)L*S+o(Jz -}-S3
(11)
as a perturbation.
We use the fundamental system
of eigenfunctions labelled
by nlmsm
which the matrix of HM is
t ,
in
We
5
1
is given by 4 3.
diagonal and the matrix of
3
shall again consider just the matrix for a given nl. From 7 3 we see that this
= I + J and (1 4- J) are
matrix will split up according to m. The values
* This rule
applies also to atoms with
more than one electron outside of closed
shells.
See
2 16 .
10 5
realizable in just one
Hence the
from
ZEEMAN E1TECT
way, namely with m =
153
m=
| respectively.
s
energies corresponding to these states are obtained immediately
7^3:
1,
For the other values of m we obtain the second-order secular equation:
which has the two solutions
- t*l (I + 1)}.
(12b)
Here
represents the magnitude of the spin perturbation while Ho

represents the magnitude of the magnetic perturbation. If we neglect
2
($o/) compared to I, we obtain
which agrees with the weak-field case as given by
we
obtain
(8)
and 4 5 7. If we neglect
which agrees with the results obtained in the strong-field case.

The complete transition (12) is plotted in Fig. 95 for the case of a
P term.
the ratio of magnetic

The splitting is plotted in units of as a function of
left
are
the
at
to spin interaction. The broken lines
given by (13) while those
&>/,
at the right are given by (14).

The best experimental data
on the complete Zeeman effect are furnished

Li
doublet
the
the
work
of
Kent*
on
A6708, which arises from the transition
by
2p
2s*S. For most doublet spectra the doublet interval is so large, i.e.
so large, that the highest utilizable magnetic fields will correspond to only
2p_>
is
small values offiojî hence a good Paschen-Back effect is not obtainable.

But the doublet splitting of lithium is very small; the 2p 2 P has a splitting
of only 0-338 cm.- 1 ( = f)- Kent used a maximum field of 44,200 gauss, and
since
fio
= 4-670 x 10~ Jf
beyond the
cm"
limits of Fig. 9
1
,
this corresponds to
fto/
= 9-16, a value even
5
.
compare our curves with Kent s data, which is not of

sufficient accuracy to be fairly reduced to term values, is by plotting the line
the 2 P of Fig. 9 5 and
pattern which should result from a transition between
The best way
to
* KENT,
Astroptys. J. 40, 337 (1914).
$.
From
we
(12a)
and
energies
fio
by two
lines
m=
10 5
154
see that a
is split
uniformly into two states with

would be represented
5
2
So; hence to the scale of Fig. 9 a $
through the origin of slopes
4- 1
and
corresponding to
|.
2
In Fig. 10 5 we have plotted the line pattern for transitions between a P
and a 2 S, using fall lines for perpendicular (a) components (Am= 1) and
broken lines for parallel (77) components (Am = 0). Kent s observed patterns
same figure, using circles for a components and crosses for

The
abscissas are fixed by the relation &o/ = 2O8 x
components.
are plotted on the

TT
&/>+*
oft/c,
Fig. 9
s.
Paschen-Back transformation of a
P term.
For most of the patterns, the absolute

which
values of the ordinates were obtained by Kent by comparison with another
line in the same exposure, but the relative values are certainly to be taken as
of more value than the absolute. For example, the pattern at abscissa 1-12
is certainly to be shifted bodily toward the red.
The agreement of these experimental values with the theory is excellent
in view of the difficulty of accurate measurement of such fine patterns. It is
pleasing that the theory of the Zeeman effect gives, purely from symmetry
holds in this particular case.
considerations, the absolute values of the perturbations, so that the pattern

of Fig. 10 5 is completely determined in absolute value if one knows the
doublet splitting.
ZEEMAN EFFECT
OKE-ELECTRON SPECTRA
156
105
Intensities.
The
relative strengths of the allowed transitions in the weak-field case
may be obtained by use of the matrix components calculated in 9 3 and 1 1 3

The details are given in 4 16 Fig. 1 1 5 shows the expected line pattern for
.
with the relative strengths for observation perpendicular to the

magnetic field indicated by the lengths of the lines.
P->
(a)
.,
(si
(9)(JO)
(7) (8)
I
TT
cr
I
&)(*)
5
Fig. II . Allowed transitions and
strengtiis in the weak-field Zeeman
5
Fig. 12 .
Allowed transitions
and strengths in the PaschenBack hmit.
effect.
In the Paschen-Back limit we have the selection ride

to the rule
Ams =
in addition
Am = Am,) =1,0. The allowed components for 2P--2

(
/S
are in
dicated in Fig. 12 The strengths in this case are readily obtained from
formulas like II 3 9. For large fields lines 2, 5 and 3, 4 respectively become
asymptotically parallel to and equidistant on either side from the outer com
.
ponents of a normal Lorentz triplet of interval fio. Hence Kent s pattern in

which these lines are unresolved forms almost exactly a Lorentz triplet.
For intermediate
eigenfunctions.
fields,
we may
Since $(np-*Pl)
eigenfunctions for these states are
of component
(5) is
at once obtain the strengths from the
<f>(np%l)
and
iff(ri$
s^Q), the
hence the strength
S^)
<j>(n
independent of field;
independent of field. Similarly the strength of
(4) is
157
ZEEMAN EFFECT
10 5
|, we
independent of field. For m
2
in
the
coefficients
7 2 the
expansion
From
these
we
may find from a
set of equations like
find
2
2
So the strength of the component from one of the states to $^ is a from
the other, /J 2 times the strength from np -f m+| in the Paschen-Back limit.
,
2
2
Since a 4- ]3 =
I,
5
Fig. 13 .
the
sum
of these strengths
is
a constant.
2
The variation in strength of the components of P
transition. (The numbers are values of
a Paschen-Back
in
->
&>/
.)
taken from the paper by K. Darwin, are indicated

intensities of the transitions of Fig.
roughly by the widths of the lines the
5
of Fig. 10 5 , lines (4) and (5) are of constant
to the
10
In Fig. 13 5 which
,
is
numbering
Referring
the strengths of (6) and
sums
of
the
strength;
.
components, of
(7)
independent of ffl.
(2)
and of (3) and
(1),
for the a
and (9),
components, are
(8),
have
been made with
While no intensity measurements
and of (10) and
for the
-n
these changes in intensity, we

accuracy to check quantitatively
which are forbidden in the
lines
the
of
may note that Kent observes none
sufficient
Paschen-Back limit at values of fiojt, greater than
2-1.
CHAPTER
VI
THE CENTBAL-FIELD APPROXIMATION*

We shall now proceed to lay the foundation for the theory of spectra of atoms
containing more than one electron.
The mode of describing
spectra that
is
used by working spectroscopists everywhere is based on the idea that the

atoms can be treated to a fairly good approximation by regarding the elec
trons as moving in a central field and not
interacting with each other. This
made the starting-point for a calculation in which the interactions actually
shall see that the relative import
occurring are treated as perturbations.
is
We
ance of different kinds of interactions varies a

great deal from element to
and
that
this is in a sense the origin of the different
element,
spectroscopic
and chemical behaviour of different elements. In all atoms there will be

terms in the Hamiltonian which represent the
magnetic interactions of the
electronic orbits
and spins and terms which represent the Coulomb repulsion
of the several electrons.
At present it is not known how to give a satisfactory

theory of atoms with, more than one electron which takes into account
relativity effects: in fact, even the exact relativistic treatment of hydrogen
which allows for the finite mass of the proton is not known. But
relativity
effects are usually small so that
one can give a fairly satisfactory treatment
of the subject in spite of this defect.
In this chapter we shall also study some of the features of the atomic
problem which have their origin in the fact that all electrons are believed to
be dynamically equivalent. This fact provides a natural
place in the theory
for an empirical rule known as PauWs exchmon
and makes
principle
possible
a fairly complete understanding of the

periodic table of the elements.
1.
The Hamiltoaian
for many-electron atoms.
According to the nuclear model, we regard an atom as made up of a central

massive positively charged nucleus, surrounded
a number of electrons.
by
This dynamical model
described in the theory by a Hamiltonian function

whose proper values give the allowed energy levels and whose proper func
tions are of use in
calculating various properties of the atom. Since for all
is
atoms the nuclear mass is more than 1800 times that of the electrons, we
may
approximately regard the nucleus as a fixed centre of force instead of treating
coordinates as dynamical variables. This amounts to
treating the nucleus
as of infinite mass: the correction to finite nuclear mass is treated in I 18
its
The
*
principal interactions between the particles are
The method used
in this
and the next chapter
is
due to the Coulomb
due to SLATES, Phys. Rev. 34, 1293 (1929).
THE
HAMTLTONIAlSr
FOB MANY-ELECTRON ATOMS
159
For most purposes we can neglect the relativistic

electrons moving
variation of mass with velocity; thus for a system of
about a nucleus of charge Ze we are led to the approximate Hamiltonian
electrostatic forces.
-pi
Here subscripts
1, 2, ...,
the distance of the i
of the
i ih
t31
-W
(i)
N serve to distinguish the different electrons, r
electron from the nucleus
and r
is
is
the mutual distance
th
and^ electrons.
In addition to the terms written down it is essential to treat the magnetic

interactions of the electronic orbits and spins. The exact way of doing this
of the electron
s
along the lines of a generalization of Dirac relativistic theory
to a system of several electrons is not known; some work in this direction and
7
to the helium spectrum we shall report in 7 An approximate
its relation
allowance for the spin-orbit interaction may be made by including for each
5
electron a term of the form gfa) L^St such as we have introduced in 4 for
as
our
the doublet structure of one-electron spectra. We shall adopt this
of atomic spectra
working hypothesis and to a large extent base the theory
the
of
on the quantum-mechanical properties
approximate Hamiltonian
(2)
In this Hamiltonian, the terms expressing the mutual repulsion of the

electrons prevent a separation of variables. These terms cannot very well
5
be neglected as small and treated later by perturbation theory, for

2
is small compared to a Ze /r
although if Z is fairly great any one of them
with the
term, there are so many of them that their total effect is comparable
The
nucleus.
procedure generally
interaction between the electrons and the
used is based on the idea of screening, according to which the greater part
of the mutual repulsion terms is taken into account in the approximate
solution on which a perturbation theory treatment is based.
The mutual repulsion terms, being all positive, tend to cancel the negative
terms which represent the attraction of the nucleus. If r+ is large compared

2
2
1 values
to all the other r^ then r^r^ and e /r^~e /?v Since there are
of j associated with a particular value of i, one
N-
sees that at large distances the electron
in a field that approximates
We say
that the other
screened
off
the force
to(Z-N+ I)e2 /r^
(NI)
field
moves
*
electrons have
of the nucleus. Now
.".
Fi
as ri diminishes to a value comparable with the

values of the other r s, this compensation becomes less exact. From poten
we know that if the other electrons have a distribution with
tial
theory
THE CENTRAL-MELD APPROXIMATION
160
I6
f
symmetry, then inside a shell of radius a and total charge e

e /a; hence the
the potential due to the shell is constant and equal to
potential energy of an electron inside such a shell is ee ja. For an electron
spherical
near the nucleus, ri is small compared to the other r^ so

as a function of ri will approach the form
,
its
potential energy
Ze\ (N-l)e*
a
rt
which a is the harmonic mean of the radii of the shells in which the other
electrons are distributed. From these general remarks we see that a large
part of the effect of the mutual repulsion of the electrons can be allowed for
by starting with a Hamiltonian in which the potential energy of each elec
tron at distance r from the centre is such a function U(r) that
in
---
i-
for r small,
,
f
r large.
for
Later we shall see that in this characterization of U(r), r small and r large
mean respectively râ/Z and r>a/(,Z JV-f 1) where a is the Bohr
5
5
hydrogen radius of 2 2.
Our programme
be to build up a systematic theory of the spectra of

atoms based on the use of such a screened potential-energy function as the
starting point of the perturbation theory. We shall see that a large amount
will
of general information about spectra can be obtained with hardly any more
detailed assumption about U(r) than this. In Chapter xiv the theory will be
extended by detailed consideration of the question of the best assumptions

and the actual evaluation of quantities that depend on the choice
made.
The approximate Hamiltonian E which we make the starting point of our
for U(r)
calculation takes the
form
(3)
while the perturbation potential will consist of the difference
S
2.
(4)
Equivalence degeneracy.
The Schrodinger equation for the allowed values of E (1 6 3) takes the form
= ]B t.
^-^+U(rj^4
J
i=lL
Since the
left side
of this is the
(1)
-*/*
sum of a number of parts each of which refers
EQUIVALENCE DEGENERACY
26
to the coordinates of only one electron,

be separated if we write
it is
161
evident that the variables can
N
= n UiWûâû^
;
Since the Hamiltonian

different electrons, the
E is a symmetric function of the coordinates of the
N equations for the N factors of the wave function are
all of the same form so that it is not necessary to carry a subscript to show to
which electron s coordinates we refer; each u is the same function of its
coordinates for the same set of quantum numbers at. Physically this means
simply that in the absence of interaction each electron moves in the central
field U(r) exactly as it would move if the other electrons were not present.
The equation for each u is
For convenience, we denote by at an individual set of four quantum num

bers which specifies the state of motion of a single electron in a central field.
Thus at represents a set of values (n I s l ) or (n Ij m) according to the way
mm
in
which we wish to
treat the one-electron problem.
E possesses a high degree of degen

be
as
kinds. First, there is that which
of
two
may
regarded
arises simply from the fact that the Hamiltonian for the individual electrons
does not depend on spin or on spatial orientation the energy E (at) is
It is clear
now
that the Hamiltonian
eracy, which
and s . This kind of degeneracy we have already

really independent of
t
studied in connection with one-electron spectra. Second, there is the
individual sets of quantum
degeneracy that goes with the fact that the
numbers of the complete set of

necessary to specify the state $ can be
associated in any way with the electrons 1, 2, 3, ... to give a which belongs
to the same total energy. This degeneracy arises because the Hamiltonian
4JV"
<
a symmetric function of the coordinates of the several electrons, i.e.

because of the dynamical equivalence of the electrons, and is therefore
is
known
as equivalence degeneracy.
convenient notation for the different
set of quantum
numbers
is
<
L z
N
plete ordered set a a ...a of one-electron
understand by
<$>(A)
associated with a particular
now desirable. We shall let A stand for the com

quantum numbers, and
shall
the particular function
which the first electron is associated with a1 the second with a 2 the third
with a3, and so on. Then let the operator P stand for a particular permutation
in
26
THE CEKTBAL-EIELD APPBOXIMATIOlSr
162
of the electron indices relative to the quantum numbers in
<j>(A).
A particular
P can be specified by the scheme

/12345...\
-
\25314..../
whicli means that after the permutation has been carried out the coordinates
of the second electron appear where those of the first were, the fifth replaces
as in (5), the result of
the second, the third remains, and so on. Thus with
operation by P on
is
<j>(A)
PftA) -^(a
1
)
3(a) %(a*) %(a
u,(a*)
N electrons, there are 2V! such permutations.*
With
with the quantum numbers a1 a 2 a? ...
is
5
)
....
Any
(5
thus of the form
rtA^ZCtPftA),
P
CP
state associated
(6)
an arbitrary set of coefficients.

The next problem is that of determining the particular values of CP
which correspond to the states of the atom as observed in nature. Before
considering this question we may note that unless a number of the C^s
vanish, the occurrence of the summation in (6) makes it impossible to speak
where the
s are
of a definite electron, say the first, as having a particular set of quantum

sets of quantum numbers, but there
electrons and
numbers. There are
is not a definite correlation of electron identities with quantum numbers.
3.
The dynamical equivalence
of the electrons.
We
recognize immediately that all physical observables are symmetric

functions of the coordinates of all electrons in the system (in fact, of all
electrons in the world). If this were not so, the particular electron
occupied an unsymmetrical position in the form of some observable
which
would
be observably distinguishable from the rest, contrary to experiment.

There is, however, another phase to the complete dynamical equivalence
of the electrons. If a particular electron occupied an essentially unsym
metrical position in a $ function representing the state of a system, that
electron would behave differently from the rest, even under the action of a
symmetrical disturbance. We therefore postulate that a physical ^r function
be such that
all
electrons have exactly the
same properties
the same
momenta and
The idea of the last sentence is formulated by the requirement that
the mean value of any algebraic or differential function / of the electronic
coordinates be the same as that of the function P/, where P represents a
of having various
probabilities of being at various places,
spins, etc.
permutation of the electrons.

* If two or more sets of quantum numbers are identical, these do not all lead to distinct eigenfunctions; the permutations are considered distinct nevertheless.
THE BYNAMICAL EQUIVALENCE OF THE ELECTBONS
36
163
The form of an observable has physical significance as such; hence a

physical observable must be a really symmetric function of the electrons.
A $ function has physical significance only through the probabilities which

Hence we do not postulate that
symmetric, but only that $ have sufficient symmetry that
may be
calculated from
it.
if/
be actually
(i)
for all permutations

and operators /. Applying the inverse permutation
1
P"
to all factors in the integral on the right, which does not change its value,
we obtain
__
(2)
Now it
seems obvious that
where S is a
if lff*fs=
real constant; for if
some operator
and
<f>
*jt
then ^ = e*s ^,
for all operators/,
were otherwise related we could find
to bring out the distinction.*

Pe/r
for
ff(f>
Hence we
require that
= e%
(3)
any permutation P.
Let us write our eigenfunction as ^-fc..., the function which is obtained

from this by interchanging the coordinates of electrons i and j as ^&... etc.
= et V#fc... By interchanging i and j again we
(3) then requires that ^&...
=
=
see that iikmm
and hence that ^ lfc ... = ^.... In the
e*V#fc- e2
same way, the interchange of any two electrons in merely multiplies the
,
*[r
*V#A>..
>
i/r
* An immediate
proof of this may be given for the case in which the coordinates take on only
discrete values so that the integral becomes a summation, as for the case of the spin coordinates.
formal generalization of this proof to the case of continuous coordinates may be made by use of
the 8 function.
Denote the coordinates by x19 x2
arSJ
....
Then
if
/=S(a:1
-a1
S(a:2
-a2 )...,
the
equation
becomes
#01,0*,
fe,
2,
or
If
=^1,03,
...
<f>a
iy a*, ...
at
tjnff=f<j)
we take /=(! -oj Sfo -%).. .9/3x

#01,01,
we
^-^,02,
=^z,a4
...
^-^.o.,
(a)
...
= $?.
j&
dr
oXf
or
all points,
find that
at all points,
(b)
Pividing (b) by
where ^
or since
a function of a^,
Then
is
constant.
^r
is
(a) gives
2^,
...
log
^=z
log
at all points,
c^
which does not involve x i
(for
any
i)
and which
a continuous function which vanishes at
infinity,
to satisfy
(a)
is
therefore a
36
164
1. Since the functional form of $ is completely determined by

function by
the position of the indices, let the interchange of the first two indices multiply
^ by
c 12 (
of the
first
1),
and
Hence, since
the interchange of the second and the third indices by
third
c| 3
=1
by
etc.
c 23 ,
Then
13>
c12
= c 13
In the same way all the c s may be shown to be equal, so that a $ function for
a -physical state must be either completely symmetric or completely antisym
metric in all the electrons
i.e.
the interchange of any two electrons must leave
1 in the other.*
the function unchanged in the one case, or multiply it by
In the particular case of eigenfunctions 26 6 of the Hamiltonian JJf, the
choice
for all
<=!
is
(4)
(5)
seen to give us a symmetric state, while
Q*p = ( -l)p
for
aH
Here^hasthe parity of the permutation P.f

The antisymmetric eigenfunction (5) can be written as a determinant
gives us an antisymmetric state.
VFJp
-^"
vm
(6)
in this form the antisymmetry follows from a

minants.
From any eigenfunction
ifi,
known property
of deter
not necessarily of the form 2 6 4, we may build a
symmetric eigenfunction and an antisymmetric eigenfunction by the use of

the operators
..
(7)
* See
PAUU, Handbuch der Physlk,
d
24/1, 2 ed., p. 191, for
a.
group-theoretical discussion of this
restriction.
| Any permutation can be regarded as made up of a number of simple interchanges, i.e. of

permutations in which (2V 2} indices are left unaltered and the other two simply interchanged.
The number of such interchanges for any given permutation is uniquely even or odd, and the per
mutation is accordingly said to be even or odd. p is an even number for an even permutation, an
odd number for an odd permutation.
THE DYKA3HCAL EQUIVALENCE OF THE ELECTRONS
36
165
which we shall call the symmetrizer and the antisymrnetrizer, respectively,

These operators may be studied by purely operational methods. From the
definitions we have the relations
(8)
From the first two it follows that the eigenvalues of Sf and sf are VN and 0.
I
From the third it follows that if $ is an eigen-z/r for VN

the zero eigenvalue of &, and
zero eigenvalue of s.
Now a symmetric
is
if
an eigen-^ for VN of
I
one for which
P$ = fy for all P.
of
<$&
ff* it
If
it
belongs to
belongs to the
S^^^VWl^, $ is
^P = P^=^.
But &(&ilt)=VWl&fa hence

is
That
f$ symmetric.
any
essentially the only symmetric
f$
7
linear combination of the A
P^ s is shown by the following argument,
seen to be symmetric, since
for
is
</r,
S CP P$ is symmetric, i.e. that PS P Pijs = S CP P$ for all P.

Then y^ CP Pj= VWl E CP P$ by (7). But ^S(7P P^= (S P )^. Hence
Suppose that
p
if
2 (7P P^ is symmetric.
j/P= PJ/ = - l)^j/ we see

unique antisymmetric linear combination of the Nl P^ s.
Similarly,
Hence
is
from the fact that
that
is
*$/
the
& produces a symmetric,
possible to do so. If
it is
stf an antisymmetric eigenfunction if it

not possible to do so the functions 5^^ or j/$
vanish.
The
factor
1/VFI
is
placed in
& and sf so that
if all
Nl
of the
Pf s are
and j/$ will be normalized if it is. Thus the

mutually orthogonal, then
sets a* of quantum numbers are
state (6) is normalized provided that the
all distinct.
Now we may
if an atom is at any time in an antisymmetric

in an antisymmetric state. For according to
remain
always
an atom has at a certain moment the antisymmetric eigenfunction
show that
state, it will
5 2 3, if
the rate of change of its eigenfunction with time is given by J?$a which
itself is antisymmetric. Similarly, if the atom is at a certain moment in a
always remain in a symmetric state. A symmetric.

observable operating on a symmetric state gives a symmetric state; oper
symmetric state
ating
it will
on an antisymmetric
state
it
gives
an antisymmetric
state.
THE OENTBAL-ITELD APPEOXIMATION
166
Tlie Paul! exclusion principle.
4.
of
46
In the empirical study of atomic spectra which, preceded the development

quantum mechanics, Pauli* discovered a simple generalization applying
to all atomic spectra. He noticed that if four quantum numbers be assigned

to each electron, then states of atoms in which two electrons have all four
quantum numbers the same do not occur. This was a rather perplexing
discovery at the time it was made because it preceded the hypothesis of
electron spin, so that an interpretation of the fourth quantum number was
lacking. Although the interpretation of this quantum number in terms of
electron spin was soon given, the curious fact remains that the two individual
sets of four quantum numbers of each of two electrons cannot be alike. This
is known as Pauli s exclusion principle.
9
We see at once that the quantum mechanics provides a natural place for
the introduction of this principle. We found in the last section two systems
of functions appropriate to equivalent particles such that which ever one
actually occurs in nature must be the one that always occurs, since transi
tions between states of different type are impossible. It is now evident that
the antisymmetric system of states satisfies the exclusion principle, since if
6
any two individual sets in 3 6 are identical the determinant vanishes.
Kothing of the sort occurs for symmetric
states.
Hence Pauli
s empirical
that
the
the
introduced
into
the
$ function,
principle
requirement
theory by
de-scribing the state of a system must be antisymmetric in all electrons. This
requirement is substantiated in other ways for example by its equivalence
is
for free particles to the experimentally verified Fermi electron-gas theory of

metals. For this reason particles having antisymmetric eigenfunctions are
said to obey Fermi statistics. Particles having symmetric eigenfunctions
are said to obey Bose statistics photons are particles of this nature.
We
shall,
as a matter of convenience, reserve capital Greek letters

i.e. for
) for states satisfying the exclusion principle,
completely
X, Y,
(Y,
antisymmetric states.
. . .
The simplest case in which the Pauli principle makes a restriction is in the
lowest energy levels of helium. One would expect that the lowest levels
would be given by states in which each electron is in a la state. This gives
four possible sets of quantum numbers, according to the four choices of the
values of l and ms
A: (0+0*);
sign of ms
B: (0+0-);
C: (0-0+);
D:
(0-0-);
written as a superscript to the value of 7%. When one

forms antisymmetric functions 3 6 6 from these sets one sees that and
are
ruled out by the exclusion principle, while
and C do not give independent
where the
is
PAUXI, Zeits. fur Phys. 31, 765 (1925).
THE PADXI EXCLUSION PKINCIPLE
46
Thus there
functions.
sum of the
values of
is
l
left
but one
set,
and of ms are both
167
say S, and since for it the

zero, it is a term of the type
%(seeF).
It is to be observed that, although we have found a natural place for the
Pauli principle in the theory, we have not a theoretical reason for the
particular choice of the antisymmetric system. This is one of the unsolved
problems of quantum mechanics. Presumably a more fundamental theory

of the interaction of two equivalent particles will provide a better
understanding of the matter, but such a theory has not been given
as yet.
In the zero-order scheme* which we have been considering, certain other

wise acceptable states, namely those in which two electrons have the same
individual set of
quantum numbers, are prohibited by the exclusion prin

on
the
ciple
ground that one cannot find a non-vanishing antisymmetric
function characterized by those quantum numbers. In the same way in
other schemes, in which one has no criterion so simple as the identity of the
individual sets, one wiH find certain otherwise acceptable states prohibited.
There is however one important theorem concerning this exclusion which we
can prove here, namely:
jfea
symmetric angular momentum, and if in any scheme]

yjm is allowed for one m, it is allowed for all m. j
the state
if we can find a non-vanishing antisymmetric function

by the quantum numbers yjrn, say
w), then we can
find non-vanishing antisymmetric functions ^(yjm) for all m. The proof is
This means that
characterized
"*F(yj
3
very simple; consider the equation 3 3:
Here
Jx iJy T(yj m) =
)
Jx + iJy is assumed
to be a symmetric operator, and Y(yjm) an anti

function.
Then
symmetric
Y(yjm-f 1) will also be an antisymmetric func
tion which will not vanish unless Jx -f iJy has a zero eigenvalue for the state
Jx +iJy is easily shown not to

if Jx + iJ has a zero eigenvalue,
y
yj m. But
m=j,
for
have a zero eigenvalue unless

so also has (Jx
But
m(m+ 1)]P, which is zero only if m=j

[j (j-f 1)
A similar consideration holds for the operator Jx iJyi which
has the eigenvalue

(or
steps the
1).
m value down.
* We shall call
any scheme in which the state of a system is specified by giving a complete set
of one-electron quantum numbers as in 3 6 6 a zero-order scheme, since such states are eigenstates only
of the central-field problem, but furnish a useful starting point for perturbation calculations.
THE CENTRAL-HELD APPROXIMATION
168
56
Conventions concerning cpiantum numbers. Closed shells.

In the central-field approximation the state of an AT -electron atom is
specified by a complete set of quantum numbers, which consists of a list of
individual sets of quantum numbers. Each individual set contains four
quantum numbers of the one-electron problem, usually the quantum num
bers nlms ml or n Ijm. Now the Pauli principle requires that all individual
sets be distinct, and it is this requirement which gives rise to the occurrence
5.
of closed shells of electrons. In a complete set there can be only two in

dividual sets
the same n, l and l since s is restricted to the two values
"with,
f For a given
.
ril
there can be but 2-h
individual sets
altogether only 2(21 -f 1)

complete set contains the
we say
different values of
may
have the same
so that
fi
nl.*
When
maximum number of individual sets of a given nl,
contains a dosed shell or amiplete shell of that type. It is to be

noticed that for a closed shell a negative value of l and s occurs for every
it
and of the s values are zero.

positive value so that the sums of the
l
We have seen that the antisymmetry of the eigenfunction makes it im
possible to speak of a definite electron as having a particular individual set of
quantum numbers. However, before the theory was fully developed, spectro-
formed the habit of speaking of a state of the atom as involving,

two Is electrons and one 2p electron; and there is no harm in con
scopists
say,
tinuing this convenient mode of expression if it is understood that in such

a statement the term *nl electron refers to the occurrence, in the complete
of an individual set having quantum numbers nl. In the same way we

may say that an atom has two electrons in the Is shell and one electron in the
set,
2p
shell.
Here, as was done early in spectroscopic theory,

shells, complete or incomplete.
we
let
the nl
values label the different
Since the energy in the one-electron problem depends only on nl, the
energy in an unperturbed state of the central-field approximation depends
only on the distribution of the electrons among the shells, i.e. on the list of
nl values of the individual sets. This list of nl values is said to specify the
Then the first-order perturbation theory will need to

consider only that part of the perturbation matrix which joins states
belonging to the same configuration, and we may to the first order treat the
electron configuration.
by configuration. In specifying a con

atom a notation of the following type is used:
energy-level problem configuration
figuration of an
Till
Is 2 2s 2
6
2jt>
3s 2 3p 6 3d 2 42?
(1)
which indicates an atom (really the ion Ti+) with Z = 21, N = 20, closed
2s, 2p, 35, and 3p shells, two electrons in the 3d shell, and one in the 4p.
* If we calculate this
using the jm scheme: j may be
values of m, the second 2
total 41 + 2 sets as above.
+\
or
J,
the
first
Is,
of which has 2Z -f 2
CONVENTIONS CONCERNING QUANTUM NUMBERS
56
169
Now the
order of listing of the individual sets in a complete set does not

affect the state in an essential way, but two different orders may lead to
6
eigenfunctions (3
6)
with opposite signs,
i.e.
with different choices of the
arbitrary phase. For this reason we must for definiteness list the quantum
numbers in a specified way. In the first place, the sets will be listed shell by
shell, all the sets of a given shell being placed together. The order of listing
the shells will in each case be specified by the order of nl values in the con
figuration designation of the type (1). Within a shell we shall for con
9
venience adopt a standard order of listing the quantum numbers. In the

nlmgrii scheme, the sets will be listed in decreasing order of l values, the set
with ms = + i being placed before that with ms = J, in case both have the
same m^ In the nljm scheme, the sets will be listed in decreasing order of
m values, first all sets with j -
1 +
f then those with j = I \. Having agreed
on this standard order, the sets for a closed shell need not be listed explicitly,
and only the m/n^ oxjm values need be listed for the incomplete shells. In
the m sm scheme we shall use a notation of the type
,
Nel
l5 2 2^ 2 2j) 5
3^(!+I~0-~-l -~i-2-),
where the numbers specify the ml values and the superscripts the
first of the five 2p electrons, then of the 3cL
6.
(2)
m3 values
Matrix components for S$/().*
We need to know the matrix components, in terms of eigenfunctions based

on the central-field approximation, of certain symmetric functions of the
coordinates of the N electrons. The symmetric functions which actually
occur in atomic theory are fortunately of two very simple types. The first is
that in which there exists a function of the coordinates of a single electron
which is written down once with each electron s coordinates taken in turn as
the argument and the results added. Let F be such a quantity. Then we may
&
write
a)
where /(i) operates only on the coordinates of the ii-h electron. It is easy to
reduce the matrix components of this type of quantity to those for the
motion of a single electron in the central field.
The matrix component (A\F\B) connecting states of the antisymmetric
6
type whose eigenfunctions are given by 3 6 is given by
P
where the integral means integration and summation over the
3
and the
spin coordinates respectively, as in 5
<t>(B),
(2)
3N positional
CONDOK, Phys. Rev. 36, 1121 (1930). These matrix elements follow also from the considera
tions of JQBDAN and WIGNER, Zeits. fiir Phys. 47, 631 (1928) on second quantization.
THE CBNTBAL-MELD APPROXIMATION
170
66
When we
are considering only antisymmetric states, since all the in

dividual sets of quantum numbers are distinct, it is immaterial whether we
regard P as a permutation of electron indices relative to quantum numbers,

or of quantum numbers relative to electron indices. The latter point of
view will usually be the most convenient. We may think of P as an operator

acting on the quantum number of each electron and turning it into a different
quantum number.
electron in
P<j>(A)
The summand
We
can thus denote the quantum number of the
i th
by Pa*.
in (2),
...
*APcf)fuAP
b<)...
Jt
(
J
(3)
contains integrals simply of products of eigenfonctions of single electrons in

th Therefore it vanishes unless
of the in
every place except the i
(Nl)
dividual sets in
summand
agree with those in B, since the non-vanishing of the
requires that
P b = Pa\ P b* = Pa\
l
Pb
...,
P b N = Pa*,
(4)
need not equal Pa It follows therefore that all matrix

components (A\ F\B) of a quantity of the type F vanish if A and B differ in
regard to more than one individual set of quantum numbers.
except that
Continuing the calculation, we may suppose that

k
except for one individual set, which is the set a of
conventional order. Denote
of
A is the same as B
A when arranged in
by B the set obtained when the individual sets
B are arranged in the order

a
...
This will not in general be the conventional order for the set
pass to the set
B by an
O(jB){}*(
number of interchanges.
and hence
),
We may calculate
B but we can
We then have
(A\F\) = {}(A\F\B
).
(5)
In order that the summand

) quite simply.
(-4) F\
to
not
should
must
be the same permutation
vanish,
(3)
corresponding
as P, and must be such as to put the unmatched individual quantum num
bers with the
iih
of each term in
electron s coordinates. Since

(2) will
sum over P and
is
be positive. For a given
satisfying this condition, so
for the
we
in
the same as
there are
P the sign
(Nl)l P s
obtain
(2).
The same argument applies for each value
MATRIX COMPONENTS FOR
66
</(*
171
I to
and gives the same result, since a definite integral does not
the
on
name
of the integrated argument. The summation over i thus
depend
multiplies the above result by N, giving a total of Nl, which cancels the
l/Nl of (2). Hence we obtain finally
of i from.
(6)
In the case of the diagonal element (^tjjF|x4) the argument is similar.

P f must equal P or the summand corresponding to (3) vanishes. Hence
3
Here
N\
Now of the permutations P there will be (N

...,
N). Hence on summing over
~XX
A
i
f
J
the
for
which
Pa
=a>
(J
= 1,
t*
S
*,(oO/^)- 5=1
J
(7)
since the value of the integral
To summarize
1)
P we obtain
is
independent of i.
matrix component (A\F\B) of the quantity
B differs from A by more than one individual set;
(a)
vanishes if
(b)
has the value
(A\I\B) =
fc
(a |/|6*)
(8)
agree with all those in

except that a ^ b*. The sign is positive
or negative according to the parity of the permutation which changes the con
all stand
ventional order of
into one in which sets in
which match those in
if all the sets in
at the
(c)
same places in
and
if
E=A,
the lists ;
the diagonal element is
(A\F\A)=%(af\f\a?).
(9)
i~ I
This completes the reduction of matrix elements of F in the .^-electron

problem to those of / in the one-electron problem. From the form of the
results it is evident that if/ is a diagonal matrix in the one-electron problem,
then F is diagonal in the ^-electron problem. Thus the sums of the z com
ponents of spin and of orbital angular momentum are each represented by
diagonal matrices in the JV-electron problem if the representation is based
on the rilmri scheme.
Matrix components for 2^ f g(i, j)

The second type of symmetric function which occurs in atomic theory is
7.
that in which one has a symmetric function of the coordinates of two elec
trons, g(i j) = g(j, i) which is summed over all pairs of electrons. We suppose
}
172
76
that i^}, for terms of the type S^(f i), if present, constitute a quantity of
the type F whose matrix components have already been calculated. It will
5
now be shown how the reduction from the Nbe made in this case.
6
Corresponding to 6 i, a quantity
6 is
G= i
to the 2-electron problem can
defined
by
g(ij).
(I)
i>j=i
The most important quantity of this type is the Coulomb-repulsion term

which occurs in the Hamiltonian 1 6 1, in which ?(^J) = 2 /%. The matrix
component of G joining the states A and B becomes
f
JiJj
& (Pa*) fy
;
where the continued product of delta functions contains a term for each
index except i and j. In order that there be some non-vanishing terms in the
expression one sees that B can differ from A by at most two of the individual
sets.
by
We shall have three cases, those in which B differs from A by two sets,
one,
and by none.
Considering first the case in which A and B differ by two individual sets, let
us deal with the eigenstate B in which (N^2) of the sets are matched with
those in
with the latter in conventional order:
a1 & 2
A:
B
Then B
a* a
...
ak
...
of
...
a*
...
bk
...
If
...
a*.
(3)
generally not in its conventional order, so we shall have to multiply

plus or minus one according to the parity of the permutation from the
conventional order for
to this order to obtain the correct sign for the
is
by
matrix component.
now the expression (2) for a definite i and j.

a double integral we see that for each P there are two
Consider
we deal with
which contribute
such that Pa i = ak
Since
P"s
non-vanishing terms to the surnmand. We must have

or of and Pa = a 1 or a k Then P must agree with P with regard to the (N 2)
electron indices other than i and^*. There are (N 2) of the P s for which
"
Pa = a* and Pa = at* For each of these we get two terms first we may have
P b = & P W = W and P otherwise the same as P; then P ss P, hence this
term is positive second, we can have P b* = V and P W = b k but P otherwise
"
the same as P; here P and P differ by one interchange, so this is a negative

term. Similarly there are (N2)l of the P s for which Pa* = a1 and Pa? = ak
For each of these we can have P == P, giving a positive term, and P differing
.
MATRIX COMPONENTS FOR
173
S^p(t, j)
by one interchange, giving a negative term. Altogether then the

summation over P and P gives
from
Since these terms are definite integrals, the first and third are equal and the
second and fourth are equal. Moreover, the sum over i, j merely multiplies
by the number of terms in the sum, which is N(N
I)
/2 so altogether we have
5
(A\G\B)
=
(4)
sign being determined by the parity of the permutation necessary to

into that of (3).
change the conventional order of
the
Next
if
B differs from A by but one set, we can take a permutation on the
conventional order in
B which will match
up corresponding individual
sets
with the conventional order in A,
A:
B
By
a1 a 2
a1 a2
...
ak
.,,
...
bk
...
arguments similar to those already
used>
ay ,
aN
it is
then easy to show that
(A\Cf\B)
=
(6)
in
which a runs over the
($"
1)
Finally, the diagonal element
given
individual sets that are common to
may
A and B.
be shown by similar arguments to be
by
(A\0\A)
In (7) we shall call the integrals with positive sign direct integrals and those
with negative sign exchange integrals.
If g(i,j) is independent of spin, then in the nlm^n^ scheme the sum over
the spin coordinates a i 0$ implied in the
,
/ 1
sign in (7)
may be
carried out at
once. This gives a factor unity on the direct integral and a factor S(m% mj)
on the exchange integral, so we have exchange integrals only for electrons
,
THE CEKTRAi-FIELD
174
78
APPROXIMATION"
of like spins. Inspection of (4) and (6) shows tliat if g(i,j) is independent of
spin there are no matrix components connecting states A and B if these
differ in
2m
This is consistent with the fact that in this case

regard to their
s
, the z component of resultant spin.
.
G commutes with Ss
Matrix components of electrostatic interaction.

The most important quantity of the type considered in the preceding
section is the Coulomb-repulsion term
8.
Q= I
q(i,j)=
i>j=l
e/r<,
i>j=l
which occurs in the Hamiltonian of the many-electron atom. We wish now

to carry out a computation of the integrals involved in this problem in the
nlm/Yil scheme. The most general integral occurring is one in which four
different individual sets, a, b s
OLlf II IS
i"f^
c, d,
/*
are involved.
%(a)
(ab\q\cd)=-
2 (&)
J J
The
Hence we
shall
have to
/*
r I2
single-electron eigenfunctions are given
u^(c)u^(d).
by
(I)
4 5 2, which in the present
notation becomes
a
Thus
(1)
can be factored into the product of a sextuple integral over the six
and a double sum over the two spin coordinates
positional coordinates
SS(al3 m)S(al5 mpS(a^^

<7icr
(2)
so the spin components of a

integral vanishes.
and c, and of b and
For farther progress we make use of the

!
must be the same or the
relation
1
\
where
d,
2^2 COSO)
the angle between the radii vectores of electrons 1 and 2 from the
This
can be developed in a series of Legendre polynomials
origin.
is
o>
in which r
(1)
the lesser and r
the greater of rx and r2 Therefore the part of

of
spins can be written
independent
is
<
>
.
\
J
f)
sin^
sin<9
dd^^ dp
(4)
MATRIX COMPONENTS OE ELECTROSTATIC INTERACTION
86
175
depending on the spherical harmonics, can be further

3
expanded by making use of the addition theorem (4 22). This expansion
th
permits us to write the angle factor occurring in the k term of (4) as
The second
factor,
-^- S ri(k m) O^Z* mf) 9^1 m
2&-{-l m =_fcjo
t )
sm91 dd^ f
Jo
^(
ff
.
Joo
The 9
integrals can be evaluated at once.
2?T
Hence
S(m,
mf
summation over
in the
They
give
- mf S(m mf - m\)
)
m everything vanishes unless
there are no matrix components connecting states which differ in

the value of the z component of the total orbital angular momentum. This
For states which do not
is consistent with the commutation of Q with Lz
that
is 3
only one term of the sum over
differ in this value,
m=mfm =mf
m\.
When
this condition
is satisfied,
c k (l a mf ,
where
c is
m remains, namely that for

(5)
the angle factor reduces to
mf ) c
d
(l
mf
V>
mf ),
defined by the equation

(6)
This expression
is
not symmetric in
lml and V m^ but from
^(ImtJm^^l^r^c^l
If we
now introduce the
43 18,
m^lmt).
abbreviation
^nW) R 2 (nW)
we
(7)
dr^
(8)
obtain as the final value of our matrix component
c\l
k
m,lc m$c k (ld mf,lb m$R (nW
(9)
k=\m\
The range of k may equally well be written from

identically for
Jc<
to oo, since the c k vanish
|m|.
The values of the Mk here depend on our initial choice of central field and
hence must be calculated independently for each different choice. But the
176
c s are essentially definite integrals of three associated
nomials and
may
be calculated once for
all.
Gaunt*
lias
86
Legendre poly
given a general
formula for the integral of the product of the three functions Q (I m), Q(l
Qty m+m ), for m^Q, m ^0. Our integral is seen to be reducible to
),
this
form by means of 43 18. In the first place ck vanishes unless 1% I, and l satisfy
the so-called triangular condition, which requires that Jc, I, and V equal the
sides of a triangle of even perimeter. Expressed analytically in order that
C A have a value different from zero, k must satisfy the conditions
f
(g integral)
l
These conditions limit the range of Jc to a few values in any practical case
and serve to limit the seemingly infinite series occurring in (9) to the sum of
a very few terms.
Gaunt
formula gives, for
m ^ 0, m
>
2g = I -f V
0,
1",
r
Jo
-f
-m-m--
+m
-m-
m+
where, in the summation, t takes on all integral values consistent with the
factorial notation, the factorial of a negative number being meaningless.
This formula cannot be summed except in certain special cases. Because of
its
cumbersome nature, a table of values is needed for actual computations.
Table
gives the values of the c s for
s, p, d, and/ electrons.
particular matrix elements which occur in the calculation of the
The
diagonal element of Q recur so frequently that

the special notation:
a b\q\ab) = J(a,b)
(
According to
(9),
convenient to introduce
we may write
J(a 6)=
3
it is
ak (la mf,l b mb )Fk (na la ,nblb )
(13)
and
K(a,
6)
= S(m,m*)
k
Here ak and
b k are defined in
ak (la mf
b k (la
*
mf
V>
terms of our previous ck by
m\ ) = c
mb
mf
= [ck (la mf
(l"
V>
GAUNT, Trans. Roy. Soc. A228, 151 (1929)
mf) c
m*)]
b
(l
m\
b
)
2
,
in particular, equation
(9), p.
194.
86
MATRIX COMPONENTS OF ELECTROSTATIC INTERACTION
wMle
(n
and
a la
Gk are special cases of the

b b
l )
= Rk (na a nb
l
b
,
J
,
of
(8),
177
namely
a a
l
y*
/*
^j
n blb ) = E k (nala n bl b n bl b n al a )
,
r
o J o
/*QO
= 2e
Jo
r>
/vi
$rJ
y*
-jSTiîfa
Jor^
(15b)
Tables
I6
and 2 6 give the values of b k and a
fo
for
p, d,
s,
and/
electrons.
avoid the occurrence of fractional coefficients for the JPs and

evaluation of J and
matrix components, we define
<?*s
To
in the
where
Dk is the denominator occurring in Tables I
and
26
The formulas of
the succeeding chapters will usually be expressed in terms of Fk and Gk

It is seen that the R*s which are obtained in taking a matrix element
.
between two states of the same configuration will always reduce to these
F s and &*s. For a pair of equivalent electrons, F*s and G s of the same k
y
are equal:
Fk (na a na a = Gk (na a na a
from its definition, that Fk is essentially positive
l
We
l )
l ,
also note,
and a de
creasing function of k. Although from its definition we cannot make such a
k
statement, we shall find G to share these properties when we calculate its
values in special cases.

shall use later the fact that
We
b
mb ) = I(na la m^3
m^mf,n
O* b lb mOt
where I equals / or K.
a la
I(n
\
9.
tr
if
(16)
/
\
Specialization for closed shells.
If we are dealing with configurations involving closed shells of electrons,

certain simplifications in the matrix components follow which are of im
portance for atomic theory. This case will now be investigated.
First, let us consider a diagonal element for a quantity of the type F.
In 6 S 9 the summation extends over all the individual sets of quantum
numbers. Consider separately the sum over the 2(2Z-h

associated with the nl shell. This has the value
+* ^ f 00 r^f 2
+i
m=
1)
S S
|mj=
Jo Jo Jo
3
individual sets
THE CENTRAL-HELD APPBOXIMATIOK
178
If /
is
96
a function of the coordinates r #9 a only and not an operator, tMs
becomes
/?r
/2
Jco
oJoJo
Now the sum in parentheses has the value (21+ l)/4rr; this follows from the
= 0, and Pj( 1 = 1 Hence
addition theorem 43 22 by writing l = 2 ^ =
,
our

final result is
irp^
that
(1)
the matrix component of the function/ averaged for all directions in

space, calculated as if the eigenfunctions of the states in the nl shell were
is,
TABLE
l e.
c k(lm lf
mi) and &*(Zm z ,
mi).
We write ck = Vx[D Je , where k depends only on I and I . In the table are listed only the sign
preceding the radical and the value of a;,
k being given at the head of each column. Since
2
&*=(c*) , 6*= +z/D k . Note that c*(J j|, lm l )=( - l
c k{lm lt
ml) for l+l odd
SPECIALIZATION FOB CLOSED SHELLS

TABLE
I
spherically symmetric,
the
I6
179
continued.
ml) for l+l even
and multiplied by the
total
number of electrons
shell.
Let us consider next the diagonal elements of the quantity
in
180
T4JJIJ32 6 . a*(Zm z
The value of this
we
coefficient is
,ZX
)-
independent of the signs of m l and mi. As in the preceding table,
common denominator of several related values but once at the beginning of each
For I = 0, a a (0 0, V ml) - S(, 0) for all V m{; for k-Q 9 a(l m tt I mi) - 1 for all values of the
print the
group.
arguments; in the table we give values only for
I, I ,
&
>0.
Note that
which represents the mutual Coulomb repulsion of the

element is given by the expression (cf. 7 6 7 and 8 6 12)
(A\Q\A)=
electrons. This
\\v1 (ak )^z (af )q(l 2)v l (a k )v z (a )drIdT2

i
(2)
N
V
4~A
r 7Y7 /\
It/ ( A, j ll 1
where dr=sin8drd8d and
Consider
fixed value
i?(<z)
W/Jjt 4\1
JjL
is
A/j f JJj
the spin-free eigenfunction
that part of the sum in (2) which results when ak has the
nlm^ and a* runs over all the individual sets of the nT closed
now
We do not assume nl to be different from ril
JInl ril there occurs

one term in this sum, that for which af=zak which does not occur in (2); but
shell.
>
SPECIALIZATION FOE CLOSED SHELLS
96
the
summand
of
(2) is
seen to vanish, for ak == a1 so
a vanishing term. This sum
S S
J2(i,
m^m^-Z LJ J
f"
is,
\(lm^
l,
jj
r)
2)
we have introduced only
in detail,
2) Rl(nl) Rl(ril
181
R^nl) R^(n
(B^)
2 (Z
;)
l )
<X>
R 2 (n
2 (nl)
2 (m,)
(Z^dr^
mJ)
i(Z
2 (Z
l )
O^Zm,)
wj) $ 2 (mJ)
^r^rJ
(3)
The summation over ms gives a factor two in the first term and a factor one
in the second. The summation over mj may be carried out by the use of the
addition theorem (4 3 22). This gives for our sum
I |g(l,
2)
R^nl) R^nT) R^nl) R 2 (n
l )
Q^lm^^)
-f
~^
(2r
Now writing
g( 1, 2)
e2
i\
-
Pr
"
1
)
(Zrjdrg
(4)
r*
r !2
= e2 S -^~
""
each of these integrals can be written as a sum over

62 (r^/rf
(coso>)
/c
of integrals involving
Pôosto) in place of g(l, 2).
In the direct integral one

variables # 2 92
>
may
choose a
^^ ê direction $
new
coordinate system for the
which case = #2 Then

x 9 X as pole, in
in which case the integration over $ 2 , 9 2
.
a>
terms vanish except for /c =

gives a factor 4?7. The integration over 6^ 9 X may then be performed directly
all
giving for the value of the direct integral

/
2(2Z
The exchange
ff r
integral,
+ l)!
(cf.
8 6 15)
(wZ,ftT).
(5)
on using the same expansion, becomes
<
1M
1%
"icjjr*"
i
1\
I
Pr (coso) PJcosoj) (Zr^Tg
(6)
(2Z
the product Pr
P^COSOJ) in a series of
the P^(coso>) form a
since
do
we
This
may
Legendre polynomials PX(COSO>).
IT.
The
coefficient
complete set of functions in the interval to
To evaluate
this
we expand
(coso>)
in this expansion
is
given by
*1 rP^coMPHeosco) 2
Jo
.-
2A+1
182
where we introduce
96
for the integral of the product of three
Legendre
polynomials the notation

J
Pw (cos0) PJcos0) Pw (cos6) smd d9 = Cuvw
(7)
Cuvw is expressible in terms of the c s of 8 6 6 by the relation
or the corresponding forms obtained by permutation of uvw, since

symmetric in the three indices. This integral has the value*
is
_
~
where 2g = u + v + w. The integral vanishes unless the triangular conditions

are satisfied
(cf.
8 6 10).
When we substitute the relation
Pr (cosco) PK (cos co = 2 2A + 1 CXl K PA

,
the exchange integral
(6)
becomes a double sum over K and
mand the dependence on
#2 9 2
,
Expanding
(coso>),
is
(Zm
A.
In each sum-
through the factor
)
2 (mj)
PA (cosco).
PA (COSOJ) by the addition theorem, we find that the integral of
this over the
whole range of
9 2 vanishes unless A=Z, in which case
it
has
the value
The
integration over 6^ 9 X can now be performed directly to give simply a

we are left with (cf. 8 6 15)
,
constant factor. Thus
(10)
for the value of the
exchange integral.
Hence the contribution
(3)
to the diagonal element of

(2P
of the nlm/n; electron and the

is
closed shell
given by
+ 1) [2 F*(nl, n
V)
- J S CIUK G*(ril
ril
)].
The most striking and important property of this result is that it is

independent of the values of ml and ms
(11)
entirely
We may now obtain the contribution to (2) of all pairs of electrons in the
closed shell by
multiplying (11) for nT = nl by the number 2 (2Z-f 1)
of electrons in the shell, and
dividing by two since each pair of electrons is to
be counted only once. This gives for the contribution to the
diagonal
same nl
HOBSOK, SpTierical Harmonics, p. 87;

GAUNT, Trans. Boy. Soc. A228, 195 (1929).
96
SPECIALIZATION FOB CLOSED SHELLS
183
element (A\ Q\ A) of all pairs of individual sets belonging to the same closed
shell the value
2
(21 -hi)
[2
F(nl
nl)
- J S CUK FK (nl, nl)].
12)
there
Similarly if there is more than one closed shell in the complete set
be terms in the sum in which k refers to the closed shell nl and t refers to
will
the closed shell n
The contribution of such terms
2(21+1)
(21
+ 1) [2 F(nl,
ril
-JS
is
Vx G(nl, *
)].
(13)
10.
One
electron outside closed shells.
We are now in a position to investigate the nature of the approximation

which permits us to treat the spectra of the alkali-like atoms as arising from
the motion of one electron in an effective central field as was done in 8 5 We
consider for this purpose the configurations in which all the electrons but
one are in closed shells. The one electron not in a closed shell will be referred
.
to as the valence electron.
The
first approximation to the energy of any configuration,

according to
given by the roots of a secular equation in which appear the matrix
components of the perturbation energy taken with regard to all the states
82
is
of the configuration. We label the states by a set of quantum numbers

for the closed shells and an individual set ?zJms z for the valence electron.
The degeneracy
is
m
m and ms Since for each closed shell Sm, and
and Sma for the complete set is just the m and
only in
Sms vanish, the Sm^

ms of the valence electron.
But we have seen that the
Sm
electrostatic inter
Sm
and
action energy is diagonal with regard to
r
g , so there are no nondiagonal matrix components due to this term in the Hamiltonian.
As to the spin-orbit interaction, it is a quantity of the type F, so the results

of 6 6 apply. The conventional order of listing quantum numbers will make
the valence electron in set B come at the same place in the list of individual
sets as it does in set A, so the sign in 6 6 8 is positive. For the configurations
under consideration it is possible for
-B
to differ from
A only in regard to the
and ms of the valence electron, so the non-diagonal element that occurs is

l
just what would occur for the spin-orbit interaction if there were but one
electron in the whole atom. The diagonal elements of the spin-orbit inter
action are given by 6 6 9 where, as in 4 5 / is of the form ftrJLfSi The
diagonal matrix components of this quantity are of the form ttni msm f r a
single electron in the state nlms m The sum of such quantities for all the
6
applying 6 9
is
obviously zero. Therefore the only term left in

that contributed by the valence electron, which is exactly
electrons in a closed shell
is
what it would have been if the

orbit interaction
is
others
had not been
as in the one-electron problem.
there, so the
whole spin-
184
10 e
Turning now to the diagonal elements of the electrostatic interaction we

use the results of the preceding section. For each closed shell there will be a
contribution of the form 9 6 12 and for each pair of closed shells there will be
the contributions of the form 9 6 I3. These two parts refer to the interactions
of the electrons contained in the closed shells and so are the same for
all
states of the atom under consideration. Therefore they do not affect the
relative positions of the different energy levels.
The remaining part of the diagonal element of Q is the part which

depends on the quantum numbers of the valence electron. For this part k in
7 6 7 refers to the valence electron while the
index
runs over
all
the other
individual sets. This gives a sum of contributions of the form 9 6 11 for the
interaction of the valence electron with each closed shell. The sum is in
dependent of the
ms m
of the valence electron.
From
the definition of F(nl,nT) the direct integral representing the

interaction of the valence electron with the nT shell is
* a
2 (2V H- 1)
f
The
El(nl) S%(n V) dr^r*
Jo Jo
f>
integration with respect to r 2 can be carried out to give

rl
(1)
fi
so the direct integral can be written in the form
In other words the direct integral is just the interaction of the valence
electron with the classical potential field due to the spherically symmetrical
charge density 2 (2V +
1)
R*(n
l )/r*.
The exchange
integral has no such simple interpretation,

sistent with the fact that it has no classical analogue.
which
is
con
The situation is thus as follows. To treat an atom with one electron outside
we may assume all electrons to move in an effective central
field U(r). The one-electron central-field problem has to be solved for this
field to find the unperturbed energy levels and radial eigenfunctions. Then
closed shells
the ^-electron problem will be like this one-electron problem so far as the
spin-orbit interaction and doublet splitting is concerned. It will be unlike
it in that states will occur only for which the valence-electron quantum
numbers are not those used up by the closed
shells.
The
first-order
per
turbation (except the spin-orbit terms) will then be, when the valence
electron is in an nl state, a sum of terms which it is convenient to regard in
three groups.
First, there are the terms which refer only to the closed shells and which
are therefore common to all of the states of the atom and so do not affect
relative energy levels.
These include not only a term
like 9 6 12 for
each closed
10 6
ONE ELECTRON OUTSIDE CLOSED SHELLS
shell
shell
185
and one like 9 13 for each pair of closed shells, but also for each closed
a term like 9 6 1 which, according to 1 6 4, has the special form
6
2 (2V
for the
nT
4- 1)
T- U(r) - ^1 R*(nT)dr
(3)
shell.
Second, there are the terms which are integrals of the form of a function
of r multiplied by the square of the radial eigenfunction for the valence
electron. Altogether they are
focr-
Jo
where the quantities

the
nV
f72
p2
f7
U(r)
Vn r (r)
of all the closed
>
+ X FnT (r)
are defined
by
-|
(4)
(1)
and the summation
is
over
shells.
Third, there are the terms of the exchange integrals between the valence
electron and the electrons in the closed shells. In this category there is one
sum like the second part of 9 6 11 for each of the closed shells.
The doublet structure has been shown to be the same as
in the one-
simply the combined effects of terms like (4)

and the exchange integrals which determine the relative positions of the
different terms. These results give us a little more insight into the way the
electron problem, so that
it is
approximation depends on the choice of the function U(r). The value of (4)
a functional of U(r) of a rather complicated sort; it depends on U(r) not
only explicitly but also in that the S(nl) depends on 17 (r) and so also do the
Vn
through involving the R(ril ). A good choice of U(r) would be one
that makes (4) have a small value for as many of the values ofnl as possible.
Now (4) is of the form which would arise if one were trying to approximate
is
>i>(r)
to the solutions for the potential energy

jj&
+SF
BT
nT
ir-t
TT
t ^
(o)
(r)
having started with the known exact solutions for the field U(r). This
suggests that one might take (5) as a new U(r) for a second trial and then
determine the eigenfunctions which belong to it and in turn calculate new
from these as a means of approximating to the best potential-energy
Vn
>i>(r)
function. Detailed consideration of methods of obtaining U(r)

based on special central fields will be given in Chapter xiv.
11.
and
results
Odd and even states.
In the course of the developments which follow we shall see that of all the
ordinary quantum numbers, only those referring to the total angular
momentum (/ and M) are rigorously constants of the motion for a free
atom. There is, however, another classification of states of the atom which
rigorously a constant of the motion. Each state may be characterized
5
rigorously as even or odd/ a fact of great importance in spectroscopy
is
THE CENTBAIrMELD APPBOXIMATION
186
in that
the dipole radiation of the

between states of opposite parity.
all
atom
is
II 6
concerned with transitions
Let 3* be an observable* defined by the result of its operation on a

Schrodinger $ function of the coordinates xi yi zi ai (i=l,...,N) of the
electrons as follows
From
this definition it is easily seen that
with an observable which
is
function of the position vectors ris the momentum vectors i9 and the
angular momenta JL4 and Si3 8? has the following commutation property
PftrapttLitSj^ft-rt, -pi ,Li9 SJ0

(2)
Now the Hamiltonian for the free atom is a function only of rf tyr 9p* Li9
and Si and therefore commutes with
.
0>:
[0>,H]
Hence stationary
From
states of the
its definition it is clear
= 0.
(3)
atom may be taken

that
&
is
also as eigenstates of 0*.

the identical operator and so the
1 (cf. Problem 1,
are
I 2 ). States
eigenvalues of
belonging to the eigen
value 4- 1 are called even states, to the eigenvalue
1, odd.
We shall now see that the eigenfunctions we have built up for the centralfield
approximation are also eigenstates of
0*.
It is evident geometrically
that in polar coordinates
Hence
if the function is of
the form of a radial function multiplied by a

is affected
by &. On O(m) the operation
spherical harmonic only the latter
produces
^<D(m)
while on 0(2m) the result
= eim^^(m) = - l)O(m),
(
is
so that the whole effect on any spherical harmonic of order I is

multiplication
z
by ( - l) Then if is applied to an eigenfunction like 3 6 6, the eigenfunction
2Z
multiplied by ( 1) . Hence the state is even or odd according to whether
22 is even or odd. This is the origin of the even-odd
for the
is
terminology
1
eigenstates of 8P belonging to
respectively.
Since
anticommutes with the electric moment
&
P defined
in 44 9 3 it
follows from 2 23 that the electric
moment can have non-vanishing matrix

components only between states of opposite parity. Similarly %l and
commute with
and so will have non-vanishing matrix components
only
between states of the same parity.
&
* Real because we shall

shortly find that tlie eigenvalues are real and the eigenstates orthogonal,
or from the definition because clearly
(notation of
3 s ).
CHAPTER
VII
THE RUSSELL-SAUNDERS CASE: ENERGY LEVELS

We
now
are
prepared to build on the results of preceding chapters by

atomic
studying
configurations having more than one electron outside
closed shells. For such configurations there are generally a large number of
energy levels giving great complexity to the spectrum. In the
first
approxi\
(Ze*
ri
J7(ft.)L
1
contribute quantities which have the same values for all levels belonging to
a configuration. Therefore they do not affect the relative position of the
levels of
any one
configuration. Because of this fact their consideration
is
incorporated with the general study of the central-field problem which

take up in Chapter xiv. The Coulomb interaction of the electrons,
"best
we
E e^fr^
is different for different states
of the same configuration
and thus
i>j
serves partly to
remove the degeneracy of the
states belonging to the
same
configuration that exists in the simple central-field model of the preceding

chapter. This is also true of the spin-orbit interaction terms, Slfo) LjSi
.
Experience shows that in many atoms the spin-orbit interaction is small

compared with the Coulomb interaction. For this reason it is useful to
develop an intermediate approximation in which the Coulomb interaction
is taken into account but the spin-orbit interaction neglected, or treated as
small compared with the Coulomb interaction. This approximation is called
the Russell-Saunders case in recognition of the pioneer work of Russell and
Saunders* in which the main features of the theory of complex spectra were
recognized. In this chapter we study the scheme of levels belonging to
first
a particular electron configuration. The next chapter is concerned with

methods of finding the eigenfunctions for this approximation in terms of
the zero-order functions for the central-field problem. Chapter ix com
pletes the general study of the Russell-Saunders case
theory of line strengths in this approximation.
by developing the
The theory covered
in these three chapters, while quite general, will be

suitable for actual calculation only for configurations
form which is
involving a very few electrons outside of closed shells. Chapter xm will show
in a
how the theory may be adapted to the calculation of configurations involving

almost closed
shells, i.e. shells

*
with just a few electrons missing.
EUSSELL and SATTNDEBS, Astrophys.
J., 61,
38 (1925).
THE EUSSELL-SATJNDEES CASE: EKEBGY LEVELS
188
1.
I7
The L^-coupling scheme.*
The Hamiltonian without spin-orbit interaction terms commutes with all
+ LN
components ofthe resultant orbital angular momentum L = L 4- JL 2 +
and of the resultant spin angular momentum S= SI + S% +
4- SN this is
most directly seen from, the considerations in the first part of 8 3 Hence it
. . .
. .
JS+L.
components of the total angular momentum

Therefore this Hamiltonian will have no matrix components connecting
2
2
states labelled by two different precise values of S JL J* y Sz Lzi or Jz
We introduce quantum numbers S, L, J, B
according to the
L
2
scheme
2/ =
S
commutes with
all
M M M
,
S^M/i
S(^+1)^
(I)
which correspond to precise values of

these observables. Only four of these six observables are independent; one
usually takes either the set SLMS L or the set SLJM to label the energy
for convenience in labelling states
8LM M
M
scheme the energy
SLJM a linear combination of states SLM ML with various M ML but
states.
From
33 8
we
is
obtain the important result that in the

S
L
s and
L Since a definite state
independent of
is
the same SL, this has as a consequence that in the

energy is independent of J and M.
SLJM
scheme the
representation of states developed in the preceding chapter is based

on a set of states specified by a particular set of 42V quantum numbers.
The
We
have now to transform to a new set of states in which the energy is diagonal
to a certain degree of approximation. In whatever manner this is done each
of the new states will likewise need to be specified by 4JV quantum numbers,
for this is the number of degrees of freedom of the system. In the first
approximation of the perturbation scheme we neglect matrix components
of the Hamiltonian which connect states belonging to different configura
approximation the energy states can be labelled
tions. Therefore to this
precisely
configuration. If there are -2V electrons not in closed shells

in closed shells, the configuration label amounts to a specification
by the
and
NN
of
(NN
-f
2N quantum
numbers, for
all
the
quantum numbers occur
individual sets
ring in a closed shell are prescribed and the nl values of the
in
addition
Therefore
to the con
are
also
out of closed shells
prescribed.
figuration label
we need 2N more quantum numbers with which
to label
states.
* In accordance with the recommendations of RUSSELL, SHENSTONE, and TUK;NER fPhys. Rev.
for quantum numbers which refer
8 , MZ,
33, 900 (1929)], we shall use capital letters S, L, J,
to tlie resultant momenta of several or all of the electrons in an atom, or ion; the corresponding
small letters will be used only for quantum numbers which refer to a single electron.
189
THE LS COUPIANG SCHEME
I7
In case
= 2 the four quantum numbers needed are provided by SL and
M or M ML
2 these are not in general sufficient and we

shall need to find additional quantum numbers in order to give a unique
8
in many cases ( 2 ) by
specification of the states. This can be accomplished
either
For
>
introducing the concept of parent terms.

2
2
Any set of states in which JL and *S are diagonal will be referred to as an
LS-coupling scheme. On the older vector model of the atom the orbital
angular
momenta
of the individual electrons were regarded as coupled
individual
together to give a resultant orbital angular momentum, and the
resulted
which
and
L
Then
the
a
S
resultant
to
spin.
give
spins were coupled
might be coupled under certain circumstances to give a resultant angular

momentum J for the whole atom.
set of (2j8+l)(2i-hl) states belonging to a definite configuration

and 8 will be called a term. If we are working with an
definite
and to a
SLMS ML scheme, then \ML
<
L and \MS ^ S, giving(2S4- l)(2i+ 1) states

\
SLJM scheme, we have $
^ J ^ $ 4- and
\M ^ J for each J; this counts up to the same total number of states in the
term. The quantity 28+ 1 is called the multiplicity: it Is the number of J
values and hence the number of levels that occur in the term jL^S. Even
if L
S we ascribe to a term this multiplicity and say that its multiplicity is
in the term.
Working in an
LJ
<
not fully developed. The notation for a term, which has become standard,
is to write $, P,D, F, ... fori=0, 1, 2, 3, ... and to write the numerical value
of the multiplicity as a superscript at the left of the symbol for the L value.
Thus 5 P C quintet P ) indicates a term in which = 1 and 8=2. This is in
45 for the doublets in one-electron spectra. If
more than one term of the same kind occurs in a given configuration, we
5
shall for the present distinguish them by an arbitrary extra letter as a P,
accord with the usage of
6 5 P,
.. ;
in
28
we
shall
examine the
possibilities of
a more
significant
way
of characterizing them.
We have now to consider the problem of determining which terms occur in

a given configuration.* By the methods of the preceding chapter we can
individual sets comprising each complete set which occurs
write down the
Each complete set will consist of the individual sets
In the
configuration.
for electrons both in closed shells
and
out.
As it will shortly appear that the
without effect on the following

discussion, we shall in what follows use the word configuration to refer to the
number and nature
of the closed shells
class of all configurations
not in closed
is
which have the same
ril
values for those electrons
shells.
Having made a
list
of the complete sets which belong to a configuration

* PAULI, Zeits. fur Hiys. 31, 765 (1925);
GOUDSMIT, ibid. 32, 794 (1925).
190
THE RTJSSELL-SAUNDEES CASE: ENERGY LEVELS
i7
we may classify them at once by the values of Emj= L and Ems = a To

make the argument concrete let us consider np n p (=pp), the configuration
in which two p electrons with differing n values occur outside closed shells.
6
It will be convenient to use the notation of 5 2 for the
and
m
m
values
of the two electrons, the superscript

s value
sign indicating the
to
which
the
it is
is
value
whose
set
for the individual
integer
given by
l
classified
for
attached. The resulting table of complete sets
pp
by L and
s values is
(2)
=
= 1 there must be a
Because of the occurrence of a state for
L 2 and
8
3
will have a state
is
and
this
that
2
and
L
D,
8=1,
term having the values
=
= 0, only
2
cell
and
in each cell of the table. Of the two states in the
L
8
1
3
term to account
one is accounted for by the D term, so there must be a
ML =
from the
=
=
1
the
1
infer
and
we
occurrence
cell
the
in
states
two
8
L
presence of
1
3
1
of a 3 P term. Continuing this process we see that the terms are /S, $, P,
3
3
P, W, and D.
These terms are the same as one finds on the vector-coupling picture
according to which 8 as the resultant of two spins each of magnitude onehalf can have a value zero or unity, and L as the resultant of two orbital
can have the value zero, one,
angular momenta each of magnitude unity
or two. However, the vector-coupling picture does not tell us which of the
for the other. This accounts for all states with
2. Similarly
terms are eliminated by the exclusion principle in case the configuration

involves equivalent electrons.
Let us see what effect the exclusion principle has in the case just con
sidered if the n s of the two p electrons become equal, that is, for the con
We
2
have to rule out all complete sets in which identical in
figuration p .
dividual sets occur, and must count but once sets which are now the same
6
for different order of listing of the individual sets. From 4 1 we see
except
that the exclusion principle excludes or permits whole terms so that
its
191
THE L^-COOTLING SCHEME
operation does not spoil the possibility of setting up an i/S-coupling scheme.

The S L table now is
MM
(3)
so the terms are *$,
P and 1D. A total of 36 states for pp is reduced by the
exclusion principle to 15 for p z

Similar tables for other configurations
.
may
be made to find the terms
occurring, although for several electrons outside closed shells the tables
become quite extensive. Because of the symmetry which obtains it is
necessary only to construct that part of the table which corresponds to
ML ^0 and Jf
^0.
Term energies.
Now let us calculate the first-order perturbation energy for the terms of a
2.
given configuration in the approximation in which we neglect spin-orbit

interaction. There are no matrix components connecting states of different
ML and M&
therefore the secular equation for the whole configuration

which is associated with
factors into a chain of secular equations, one of
each value of
M ML
states that the
sum
We make
use of the diagonal-sum rule (2 2 21) which

of the roots of a secular equation is equal to the sum of
.
the diagonal matrix components occurring in it. Since all the levels of a
term have the same energy, this rule allows us to write a set of linear equa
MSML
equating the sum of the energies of the terms

sum of the diagonal matrix elements of the
perturbation energy taken with regard to the complete sets in that cell. If,
3
for brevity, we write 3
for the first-order energy of the
term, and write
tions,
one for each
cell,
having states in that cell to the
(1* 1+) for the diagonal matrix element of the

this complete set, we obtain, for pp,
Hamiltonian associated with
-1+1+)
192
THE
RTJSSEU>SAUNDEBS
ENERGY LEVELS
CASE:
and additional equations from the remaining
cells
27
winch are not needed
except as checks in actual calculation.*
From these equations we readily solve for the term energies in terms of
the diagonal matrix components:
l+)-(l+l+)
l+)-(l+l+) ?
(1)
etc.
Corresponding calculations for the configuration p* give the formulas,
analogous to
i
(1),
= (l+l-)
*P=(1+0+)
i/S= s (l +
-l-) + (l--l + ) + (0+0-)-(l+0+)-(l+l-).

If a particular kind of term occurs more than once in the configuration,
the only way in which the energy of these terms occurs in the set of linear
(
as the sum of the energies of all like terms; hence this method
the
sum. In order to find the individual energies one needs then
gives only
to
solve
a secular equation involving the non-diagonal matrix
actually
elements.
case of such a calculation is taken up in 7 8
equations
is
After the calculations for a particular configuration have given formulas

like (1) and (2), these may be still further simplified by making use of some
special properties of the diagonal matrix components from the preceding
chapter. From the way in which these formulas are obtained it always turns
out that the number of matrix components occurring with a plus sign is one
greater than the number with a minus sign. Therefore any quantity which
occurs as a constant added to all the diagonal matrix elements of a con
figuration will come up simply as that same constant in the expression for
the energy of each of the terms. It is the electrostatic repulsion of the
electrons which gives rise to the separation of the terms. From 7 6 7 we see
that a particular diagonal element of
(A\Q\A) =
Q = Sea/r# is given by
Ka6|e/ria la6)-(a6|e/r18 |6a)],
(3)
a>d=l
where the summation extends over
all pairs
of individual sets in the
com
plete set. The sum in (3) breaks up into (a) pairs between individual sets in
closed shells, (b) pairs between an individual set not in a closed shell with
those in closed shells, both discussed in 9 6 and (c) pairs between in

dividual sets neither of which belongs to a closed shell. The parts (a) and (6)
will be constant for the whole configuration. Part (c) is what gives structure
,
We
*
should like to call attention here to a slightly different formulation by VAX VLECK
[Phys. Eev. 45, 405 (1934)], In terms of a vector model proposed by Dirae, of the problem of
obtaining the first-order electrostatic energies. This formulation furnishes a procedure for
obtaining the energies of the terms of any given multiplicity without first determining the
energies of the terms of higher multiplicity.
TEBM ENERGIES
27
to the configuration apart from
Its
193
absolute location on the energy scale.
JV
It
is
simply the sum
K(a
[/(a, b)
b)]
over the
electrons outside of
a>5=l
6
closed shells, in the notation of 8 6 12. From formulas 8 I3 and the tables of
a s and 6 s ? it is a piece of straightforward computation to express the energy
of each term with the F and G integrals. So many coefficients vanish that
the summations in 86 I3 seldom involve more than two or three terms in
cases of spectroscopic interest.
Let us illustrate the method by considering p* in detail. On writing E$ for

the constant part which arises from pairs (a) and (b) we have
1+ 0+)
= EQ +
F<>(np,
np)
- 5F2 (np, np)
and similar expressions for the other diagonal elements occurring in (2).
Here we have used the fact that G-jJ(np,np) = F^p,np) = F k (np,np)IDk
where Dk is the denominator occurring in Tables I 6 and 2 6 Hence finally the
electrostatic contribution to the energies becomes
,
F2
J2
(4)
.
Although we cannot know the numerical values of the integrals F^(np, np)
and F%(np, np) without basing the calculation on a definite assumption for
U(r) we see from the form of this result that, whatever the particular U(r)
chosen, the first-order theory predicts an interval ratio
(i#
- ID)
(iD
- P) = 3; 2,
(5)
with the S as the highest term.

This form of calculation thus provides relations between the term energies
which can be compared with experiment independently of the more difficult
I
problem of finding a good central field. Such predictions and

7
parison with experimental data are given in 5
their
com
3.
Hie Lande interval rule.*
The next problem in the theory of the energy levels in the RussellSaunders case is that of allowing for the spin-orbit interaction, regarding
this as small compared with the electrostatic interaction. The electrostatic
interaction breaks the configuration up into separate terms, each character
ized by a given L and 8, but leaves these terms degenerate with regard to J.
The spin-orbit interaction has the form
t
LANDE,
Zeits. fur
(1)
Phys. 15, 189; 19, 112 (1923).

13
THE EUSSELL-SAFISTDEBS CASE: ENERGY LEVELS
194
Front 8 34 we see
ttiat this
commutes with J;
37
this permits the levels to
be
labelled by J values even when the spin-orbit interaction is not neglected.
does not
But
commute with S and L so that these no longer furnish accurate
quantum numbers. In case the spin-orbit interaction is small compared to

the electrostatic, however, we need only consider the effect of the diagonal
matrix components in a scheme of states labelled by SLJM. In this way,
a given term will be split by the spin-orbit interaction into a close group of
levels characterized still
L to a good approximation and distin

For the individual levels we use the standard
by S and
guished by different values of /.

+l
.
If we wish to
notation,
specify a state
Lj
2S+ 1
right superscript,
state by 2S+1
S9
Z$
A M ML
f>
In the
SLMSML
we
write the value of
scheme we
M as a
shall designate
The complete matrix of H 1 for
all the states in the configuration includes

that
are
components
non-diagonal with respect to the terms. These are re
sponsible for the breakdown of Russell-Saunders coupling, i.e. for the loss
of validity of the characterization
by S and L. The details of such effects are
considered in Chapter xi; here we confine attention to the diagonal elements.

From the fact that r| commutes with Lit we see that (r f )jL 4 is a vector of
type T with respect to L i to L, and to JF. Hence the diagonal element of the
term of (1) is given by the first equation of 12 3 2 if we make the correlation
,
i th
as
(2,
In this expression everything is independent of / and

except the factor
in braces. Hence the whole dependence on J of each term ofH 1 is given by
L S=j(J* L 2 S2 this factor is recognized

(SLJM \L*S\SLJM so that we may write
r(ySLJM) = (ySLJM\H 7 SLJM)
this factor.
Since
matrix element
as the
),
- l,(ySL)(SLJM\L-S\SLJM\
where (y$L)
is
a factor depending only on S, L, and
configuration, etc.)
and which
is
y (which specifies the

entirely independent of J. The spin-orbit
perturbation of the level ySLJ is thus given in the first approximation by (3)
From this result we see that the energy interval between levels differing
.
by unity
in their
J value is
(4)
37
THE
LAOT>:6
INTERVAL RULE
195
that is, the interval between two adjacent levels of a term is proportional to
the higher J value of the pair. This is known as the Lande interval rule. It
has a quite general validity and is of great use in empirical analysis of a
5
term of the configuration
spectrum. As an example we may take the
3d G 4s 2 which is the normal term of Fe I
Energy
1
Level
(cm""
0-000
Da
415-934
D3
704-001
888-126
!>!
*D
*
Here
observed
constant
if
I4
978-068
is
the observed interval divided
by
J, which
would be
the rule were exactly obeyed.
The spin-orbit perturbation given by (3) has the property that the mean
perturbation of all the states of the term vanishes, that is, the mean per
turbation of the levels when weighted by the factor (2 J-f 1) is zero. This
weighted mean of the energies of the levels of a term is called its centre of
gravity, which is thus equal to the energy of the term as it would be in the
absence of spin-orbit interaction. To prove this statement we merely multiply
(3) by (2/-f
formulas
1)
and sum J from |L-$| to
S/ = JX(X+1), SJ
J=0
.7=0
L + S,
using the summation
= p:(Z+l)(2X+l), SJj:P(X+l)a.
J-=0
(5)
Here, as in the case of one-electron spectra studied in Chapter v, there is

a possibility of a small finite displacement of a term of a configuration con
taining s electrons through the occurrence of an infinite value of multiplied
by a factor that is formally zero. In such a case it is necessary to have re
course to the relativistic theory as was done in connection with the 2 /8 levels
in hydrogen.
4.
Absolute terra intervals.*
We
have obtained the Lande interval rule by a simple use of the first
3
equation of 12 2. In order to obtain the absolute term intervals in terms of
the one-electron parameters
^^H^M),
(i)
we do not need
Chapter xi is
to evaluate 3 7 2 in greater detail (which as will be seen in

in general a rather complicated procedure) but can obtain all
*
GOUDSMIT,
Ptiya.
Eev. 31, 946 (1928).

13-2
THE BTJSSELLrSATJNDEKS CASE: ENERGY LEVELS
196
47
from the diagonal-sum rule except when two or

more terms of the same kind occur in a coirfiguration. The intervals in this
the information
we
desire
latter case are considered in
78 .
In order to use the diagonal-sum rule we
shall
need to know the diagonal
SLMS ML scheme and in the zero-order
elements of the matrix ofH1 in the

scheme. In the 8LM8 L scheme
t
L \Hi\ 7 SLMS ML
= S (y SLMSML 7
\
JM
SLJM)(<y
= (y SL) (SLMS ML \L
= MLMs t(ySL),
S\
SLJM^
SLM8 ML
1
\
7 SLJM)( 7 SLJM\ySLMs
(2)
3
7 3. In the zero-order scheme, since
the last form being obtained from

a quantity of type F, the diagonal element for the state
given by
ML
(A \H*\A) =
is
is
A = a1 a2 ... aN
(<#\Hl\tf)
The manner of using the diagonal sum
= S tefrnfa*
rule
The
is this:
(3)
zero-order states
a configuration are classified according to the values of L and s and

7
the terms which have components in each cell noted as in I
By the
the
elements
of
the
sum
and
of
value
a
for
8
L
given
diagonal-sum rule,
for
sum
(3) for the zero-order states is equal to the
of the elements
(2) for
the
terms which have such components. This gives for each cell an equation
expressing certain of the t,(SL) in terms of the nl and there will be a
,
of equations to determine all the (SL) unless two or

more terms of a kind occur, in which case only the sum of the (SL) for these
sufficient
terms
number
be obtained.
may
As a simple example
consider the configuration np rip which has been

3
is the state*
1, 2 hence we
)
From 1 7 2 we see that (I + 1 +
I7.
discussed in
obtain the equation
D>
^D] ^ ^ + ^g
(
or
The
and
or
states
(0+ 1+)
D, 1 1 and
hence
3
3
P>
1, 1
are linear combinations of the states (1+0+)
^D) + ^P)
8
Z(
i(^p +
P) = 1
j|
W + U)-
If we attempt to write a similar equation for a cell with either

0,
s
L or
we obtain of course only the result = 0, but the new terms occurring in
those cells are either S terms or singlets, for which has no real significance.
We thus have obtained the result that

)
(np rip P),
* See
preceding section for notation.
ABSOLUTE TEEM INTERVALS
47
which
197
and says that
directly amenable to comparison with experiment,
is
the total
It will
width of the D should be J that of the

be noted that in this case the lowest J value of each term has the
3 P.
are all essentially positive (4 5 ). This positive

characteristic of all configurations which contain few
lowest energy, since the
nl
sign for (SL) is

electrons the resulting order with the lowest / value lowest is said to be
normal.
term for which ($) is negative so that the lowest J value is
highest is said to be inverted, and the intervals are specified as negative.

Since the situation mentioned at the end of the preceding section in which
a single level has a spin-orbit displacement cannot occur unless the con
figuration contains an s electron, none of the singlet or S levels of np n p can
possibly have a spin-orbit displacement in the Russell-Saunders case.
5.
Formulas and experimental comparison.*
We now turn to a systematic comparison with experimental data of

the formulas for the electrostatic energies and Lande term intervals in
the Russell-Saunders approximation. For the electrostatic energies in the
7
simplest cases the theory predicts certain interval ratios as in 2 5. In other
3
cases where there are more terms and more JTs and 6r s we treat the latter
s
as adjustable parameters. This enables us to see if the formulas agree with

the data in any sense: if they do, then it still has to be remembered that the
F*s and
6? *s
are not really independent.
to be obtained from
The actual values of F s and 6r s have

y
some particular choice of the U(r) underlying the
central-field approximation. This part of the
problem
is
considered in
Chapter xiv.
Departures from these first-order formulas we may regard as due to two
causes. First, the spin-orbit interaction may be so large that the L8 coupling
broken down. Second, even though the spin-orbit interaction be small,

the neglected matrix components of the electrostatic interaction which
connect different configurations may be important. The first point is con
sidered in Chapter xi, the second in Chapter xv. From 1 1 6 we know already
is
that there are no matrix components of the Hamiltonian between con

figurations of opposite parity. Hence all perturbations between E/ussell-
Saunders configurations occur between terms of the same
and of the same
S and L
values
parity.
Is
Here
.F
=F
(ris, nl),
<%
COJO>OK
= G (n s, nl); these arguments need not be specified

t
and SHOBTLET,
Pitys.
Rev. 37, 1025 (1931).
THE RTJSSELLrSAtTHDERS CASE: ENERGY LEVELS
198
unless there
3 11
in
is
57
ambiguity. These configurations are considered in detail
As F%
3
is
essentially positive, the terms in order of increasing energy are
7
P, W, S, with intervals in the ratio given by 2 5. This
with the experimental values*
Atom
Configuration
Theory
np
I
II
2p
Ge
Sn
M3
in case of CI,
compared
1-14
1-14
1-48
1-50
1-39
18-43
3p
2
4#
2
5p
2
Gp
2
6p
La H
Pbl
The departures
1-50
to be
8 - 1D)/(W - 8P)
2p*
2p*
Si I
is
0-62
Nil, OIII are probably due to action of the
configuration 2p %p above. The same remark holds for Si I, so the good

agreement here is probably accidental. In Gel, SnI and Pbl there is
11
appreciable breakdown of LS coupling (see 3 ). In La II the configuration
is high and lies in the midst of many possible perturbing configurations.
2p = 3^
2
(
P)=0
Here again we can eliminate the unknown F% and compare observed with
theoretical ratios:
Configuration
np*
0-667
S II
As I
Sb I
2p*
2p*
3p*
4#*
5p*
0-500
0-509
0-651
0-715
0-908
Bi I
6p
1-121
N
O
The
deviations in
P-2
Atom
Theory
NI
and
OH
2
(
I>
>)/(
- 4S)
are probably connected with disturbance
by 2p*3p the lowest term of which in I is only four times the over-all
3
separation of 2p above this configuration. Similar remarks hold for S II.
}
In the
last three
examples departure from
LS
coupling
is
important
(3").
We wish
here to acknowledge our indebtedness to the excellent compilation of energy levels
by BACKER and GOUDSMIT (Atomic Energy States, McGraw-Hill, 1932) for most of the experimental
data used in tiiis book in energy- value comparisons.
57
FORMULAS AND EXPERIMENTAL COMPARISON

199
The first-order formulas give 2D) = and 2 P) = 0, so the actual term
(
intervals
must be
referred to perturbations or
breakdown of Russell-
Saunders coupling.
The formulas
same
are the
data are
as for
Atom
p2 with 6P
Configuration
2p
4p
Te I
01
there
departure
shown
is
(*
The
- *D)I(*D - S P)
1-50
np
Theory
In
written in place of _F
1-14
1-50
8
probably perturbation by 2p 3p and in Tel the coupling
13
where Te I is
great, so the agreement is fortuitous (see 3
is
to agree well with the theory for intermediate coupling).
where of course the

is 0^(np,n s). Evidently ( 2 S- 2D)/( 2D-P) = i
4
2
where P = |(2 P + P). There are only two cases, Sblll 5s 5p 2 and Oil
2s 2p 2 here the order of terms is right, but the ratios are badly off: 3-58 and
0-481 respectively. In the case of SbHI the 5s 2 6s *S and the &? 2 &PD are
2
2
2D interval. In
2
Oil, 2s 3s $
nearby and located so as to increase the $
;
is
located so as to decrease the
D interval.
In Sb III 55
we find 2140 and 2400 (cm-1 ) for the two values of 4P)
inferred from the two intervals; this is roughly half of ( 2 P) = 4500, as it
should be, while ( 2D) is 507 which, although not zero, is quite small
relative to the others. Similarly in
the ( 4P) is 14-3 and 1 1-4 from the
2
2
intervals, roughly half of ( P) = 27-4, while ( D)=-0-8 which is very
2
52>
OH
much
smaller.
+0, -20J
where
JFO
=3 Fj( n
P>
nP) + 3
This configuration for
ô( w *
P);
Fz- F
O I is discussed in
(G
*(
P>
28 .
nP)
>
&i=
THE RTTSSELL-SAtHSTDERS CASE: ENERGY LEVELS
200
We expect that 6?
2 is
always small enough that
(?
57
determines the sign of the
expressions in the G*B. Thus the theory predicts an alternation: singlet above
triplet for S, then below for P and then above for D. This phenomenon
We
occurs quite generally in two-electron configurations.*

may eliminate
G s by dealing with the arithmetic means of corresponding singlets and
triplets, and find (S
D)I(D P) = f for the interval ratio of the means.
the
The means do not even lie in this order in the known
N II and
cases which are in
C I,
2
probably because of perturbation by p
3
3
Here the ( D) and ( P) should be equal; this is fairly well borne out by
the data:
III,
2p 3p
11-2
2p3p
32-1
10-6
NH
30-4
2p $p
734
68-2
(ip)
10-2
12-5
29-2
35-2
65-3
82-0
The
inferred from eaoh of the intervals in the

table gives the value of
triplet, so comparison vertically tests the Lande interval rule and com
3
3
parison horizontally tests the theoretical equality of ( D) and ( P). The
2p4p configuration in these same elements shows the same degree of
agreement with the theory*
in^. Using means eliminates

ratio
theoretical
a
(P-F)/(.F~D) = |. The means do not
giving
come in this order in 01, Nil and OIII, nor in Ytll, La II, Gel 4p5d,
where the configurations are high and surrounded by perturbing configura
tions, In Gel 4p 4d and Zrlll 5p 4d the means are in the theoretical order
These show the
the
singlet-triplet alternation as
s}
ratios are 0-28 and 3-58 respectively instead of 0-555.

The configuration 4p 3d can be foUowedf in the long isoelectronic sequence
Gal to MIX. The arrangement of the levels is shown in Kg. I 7 from which
but the
evident that the configuration is strongly perturbed in the low stages of

3
1
ionization, as is shown especially by the occurrence of the P below the P.
it is
P mean enough that the ratio
(P F)j(FD) is negative
the
ionization
the behaviour is more
and Sc II. In
higher stages of
in accord with the formulas, but here the intervals inside terms are great
enough to indicate appreciable breakdown of Bussell-Saunders coupling.
This lowers the

for Ca I
RUSSELL and MEGGEKS, Sci. Papers Bur. Standards, 22, 364

CADY, Phys. Rev. 43, 322 (1933).
(1927).

3
The
partially inverted so a detailed study of the
is
would be without meaning. The theoretical

relation
it
larger than
3
(
Similarly in
the
is
Lande
intervals
given in the table imply the
which makes
201
clear
D) to
3
why the P intervals are so
make
3
(
CI 2p M the relation between the
inverted, although the conclusion
large deviations
small since
(*F)
is
enough
P) almost zero.
from the interval
is
s indicates a reason
why
uncertain because there are
rule.
-P
Jp-
isr
?p__
5zI
ag
<=>
Ca
3
5
g
o
3
2
"
55
Tim
Sc n
viv
CrV
FeW
CoVl
NIK
The
Fig. I
configurations in the isoelectronic sequence from Cal to MIX,
The calculated terms are obtained by choosing values of FQ , 2
, and G z which give
approximately the best fit of the six observed levels. (Scales in thousands of wave
7
ipM
numbers.)
Kt
Only one example is known: La II 4/6^. The order of the singlet-triplet

means agrees with theory but the ratio (D - ?)/(? - F) is 1*05 instead of the
theoretical 0-35
inverted.
= 7/20). The interval rule is not satisfied,
jD being partly
THE BUSSEIX-SAUE-DEBS CASE: ENERGY LEVELS
202
57
Here we have the first instance of a configuration in which a term of a

particular kind occurs more than once so that the diagonal sum rule does
not suffice to give complete formulas. The terms are 2 2 4PD.F. The formulas
-
give the energies of the
means of the
(m) is the mean of the
in which
pairs of doublets, indicated
s for the
Of = Q^na, n p}
GF? = Gi(n p, ri d)
two
J7s,
Gff = GJna,
by ^L.
and
n"d)
Gf^Gz (n p,
The predicted order of the doublet means
is
n"d)
expressed in
DPF in Sc I 3d 4s 4p which is probably strongly perturbed by
The order
is
3da 4jp. In
Ytl and Zrll 4e5s 5p
the order
is
correct but with large depar
tures from the ratio, probably mainly because of perturbation
Regarding the
J^2
and
jP4 as
adjustable parameters
we have
z
by 4d 5p.
here four term
intervals to be expressed in terms of two parameters.

convenient graphical
way of doing this is to plot the observed term values as ordinates against the
coefficient of
2 in the formulas as abscissas. If the F& coefficients were all
zero, the points should

is
As they
lie
on a straight
are not
line
whose slope
is
F% and whose
we must draw a line on the graph so that the
intercept
Q
ordinate differences between the points and the line are as closely pro
as possible. Of course the parameters can
portional to the coefficients of
be chosen by some analytic process like least squares, but the graphical
.
method is accurate enough and gives a better idea of the nature of the fit.
The values of the parameters chosen so as to give the best agreement with
the data for 2d* in the isoelectronic sequence Sc II to Ni IX, as investigated
FOBMULAS AND EXPERIMENTAL COMPABISON

by Cady,* are shown in
Fig. 2
as
ionization
should.
of
they
203
They vary roughly linearly with the degree

The values are something like 80 per cent,
.
larger than what is given by using hydrogenic radial functions for the oneelectron 3d states in 8 6 15a, the hydrogenic values being J 2 (3d 2 ) = 203 and
T
= 14-72 in cm" The S term is much too low in every case, so it was
)
out of account in determining the parameters. The accuracy with which
the other terms of these configurations are represented by the theoretical
jP4 (3d
left
formulas in terms of these parameter values is shown in Kg. 3 7 There is no

obvious reason for the discrepancies in Yt II. The Zr III 4d2 configuration
.
cannot be represented at all with possible values of F% and F^, probably

because of strong perturbation by 4d5s. Similarly La II 5d 2 seems to be
strongly perturbed
0*TII
AAJrl
by 5d 6s.
Sell
7
Fig. 2 .
The
interval rule
fulfilled
Tim
IV
CrV
Fe VII
CoVffi Ni IX
Parameter values in the configuration d 2
and the
relation
3
(
P) =
3
(
jF)
are quite accurately
in the long isoelectronic sequence studied by Cady.
Here
Mn VI
YII
F =F (nd,7id)+2Fo(7id
Q
n s);
*
G2
Ft^F^Jnd,
CADT,
loc. cit.
n ^)?
G z =G 2(nd,n s).
THE RUSSELL-SAUISDEBS CASE: ENERGY LEVELS
204
We
2
(?
Neglecting the %S which

Ytl 4d 2 55 with the values
2
,
much
is
^4
as
observe that the separations of
are the same as those of *#, 3 P,
|(
P+ 2 4 P),
P,
too low
we may
TiH
Ytl
1015
59
434
J+2*F), and
Till 3d 2 4$ and
fit
32
configurations overlap
4d* respectively. In Zrll 4d 2 5s the
J(
G in d\
shown in Kg. 4 These
*D
is
and interact with 3d3 and
not so badly
although the
off,
-i?G
20-
-fc-
<?-
<D
15
\~
//>
,,
0U
ScD
VF
Til
CrV
FeW
Mn\[
3d 2
7
Fig. 3 .
CoTO
NiH
agreement in general
J4 =55.
is
first
P- 4 P
4d
J-
Tin
Ytl
Zrll
6620
3964
-1877
(*F-fF)
4558
4357
5404
Further reference to these cases
J = 905 and
4
should equal 2
JP
This
two cases and not at all true in the third:
P- 4P)
2
(
Y3I
{Scales in thousands of wave numbers.)
very poor. The best parameters are
According to the formulas

roughly true in the
3d
F^.
Zrl
Til
3d 2
The configuration d2
_/_
is
made
configuration interaction are discussed.
in Chapter xv,
where
is
only
effects of
FORMULAS
57
AISTD
EXPEBIMENTAL COMPABISOK
205
dd
8(7,
The best example of this configuration is Sell 3d4dl. As before the 1 /S is

much too low and was left out of account. The parameters used are F% = 107,
f
I9r
18-
17
IB
15
14-
13
4d* 5s
7
Fig. 4 .
F = 6; Fig.
2
Configurations d s and d d. (Scales in thousands of wave numbers.)
4 7 shows the relation of observed and calculated singlet-triplet

means. The five singlet-triplet differences are represented by (? = 2230,
THE RUSSELL-SAUNDERS CASE: ENERGY Y LEVELS
206
G 2 = 36-7 and
<?
= 3-5 in the manner shown in the table:
Observed
Calculated
iS~*S
tP-tP
T-D-*D
3872
6368
-4275
-4385
4384
4492
iJ-ajF
-3935
-3810
Other known instances are Yt II 4d 5d, Zr III 4d 5d and

the agreement
57
is
La II
*G-*G
4864
4760
5d d where
not as good; the singlet-triplet differences are entirely in
disagreement in La II.
D terms, so we can obtain only their mean
This configuration gives two

by the diagonal sum rule:
It is interesting to note that
This configuration
iJGT,
and
H have the same first-order energy.
^=1^0 + 10^+ 3^^(15^ + 10^3+
The only known

retical formulas
case
is
La II
P
Calculated
5<i
{(*#)=
6?5 )
4/ a case in which knowledge of the theo

analysis.* The singlet-triplet means are
s
was of use in the
well represented on taking .F

Observed
78
discussed in detail in
is
20534
20506
25214
25216
The G parameters were
= 21400, F%= 115,

D
I
fitted to the
- 357-6,
I= 16:
21467
21191
19038
19323
observed separations by least squares,

(?5 = 3-78 with the following repre
(? 3 = 29-7,
the values being 1

sentation of the data for the singlet-triplet differences
Observed
Calculated
H
23680
22598
4421
4638
-3279
-3397
6112
5837
-4879
-4987
H
9690
11330
This gives a beautiful example of the alternation in sign of the singlettriplet difference with increasing L. The triplet intervals do not follow the
Lande rule sufficiently well for a comparison of the s.
*
CONDON and SHOBTLEY,
Zoc. cit. p.
1042.
57
207
/3 and//
The formulas for// are

F4 +
m6J
6:
- 286.^
286<?
G^ +
3
The formulas
sions in the
3
!/S,
P, iD,
No
-F,
/ are obtained by omitting the expresand noting that the allowed terms are:
3
1
1 /. The
intervals are given by
G,
T,
for
(? s
6)
78(? 6 )
case of //is
fo._
known at present. But one/ 2 is known:
Using least squares the F parameters were

= 21-58, ^6 = 0-262. The
determined as .F 2 = 93-33,
with
is
shown in Fig. 5 7 The
observation
comparison
LaII4/
intervals in the triplets of La II 4/ are badly perturbed.

The upshot of all these comparisons is clearly that,
2
although the first-order formulas agree with the data in
many cases and explain some qualitative facts like alterna

tion in sign of singlet-triplet differences, the second-order
perturbations are usually quite important.
6.
Terms
in the nlx configurations.
It is of interest to consider the configurations in
which
of the electrons outside of closed shells belong to the

same shell. These are commonly the type associated with
all
the lowest levels of an atom (compare I 14 ). The number

of different individual sets possible in the nl shell is
JV/= 2
(2Z-I-1),
so without the Pauli principle the
number
of different complete sets possible with x electrons in the

nl shell would be Nf The exclusion principle however
.
effects a strong reduction in this number, since it requires

that all individual sets in a complete set be different and
that two complete sets containing the same individual

sets in a different order not count as different.
configuration of
Let II.
Therefore the number of states associated with the configuration nlx
is
THE BUSSELL-SAUNBEBS CASE: ENERGY LEVELS
208
67
x
ff
the binomial coefficient which gives the coefficient of z in (H-z)
number
is
symmetric with regard to
maximum
x=N /2
l
since
l
I
^.
*.
This
and has
value for x=*NJ2. Therefore the complexity of these con

a? to the middle of the shell and then
in
figurations increases with increase
for
decreases symmetrically
larger
xuptox^N^
The Russell-Saunders terms corresponding to the

x
p have been
x
configurations s
and
considered already in the course of the preceding section.
For the configurations dx and/x the calculation of the terms by the method
of 2 becomes rather lengthy. Formulas for the term energies as expressions
*
in the F and G integrals have not been worked out, but the kind of terms to
x
7
be expected are given in Table I which, for completeness, includes the s
and p* configurations already discussed.f The values of (SL) as given by
Goudsmit (loc. cit 4 7 ) for p and d electrons are also included.
TABLE I 7 RusseU-Saunders terms and Lmd& factors for nl* configurations.

The small numerals tinder the term designation show the number of terms of the kind occurring
in the configuration. The fractions in parentheses give the values of a in the equations ($!/) = a z
21 -f 1 Where two terms of a kind occur, the mean value of
21 -f 1 and (SL) = - af for x
for x
a is given.
.
<
>
The term energies for d* are given by OSTROFSKY, Phys. Rev. 46, 604 (1934).
f RTJSSEUD, Phys. Rev. 29, 782 (1927);
GIBBS, WILBER, and WHITE, ibid. 29, 790 (1927).
TEEMS IN THE
67
nl*
209
CONFIGURATIONS
Tlae configurations 3dx occur in the elements from

called iron group; 4dx and 5dx occur in the elements
Sol
to
ZnI
YI to Cdl
the so-
(the palla
platinum group) respectively. The con

x occur in the rare
earths, whose spectra are not yet analysed
figurations 4/
in any detail. Even in the case of the iron group where quite elaborate
dium group) and Lu I
Hg I
to
(the
analyses have been made the results are as yet far from complete. For
6
example, in Pel, the normal configuration 3cZ gives rise to sixteen terms
according to the table, but reference to Bacher and Goudsmit s tables shows
that only the
3
?
Z
F, and G terms are known. Similarly of the sixteen terms
6
expected from 3d only one, $,
is
known in Mai.
value of the terms of highest multiplicity.

It is easy to find the largest
For x 21 -f 1, it will be possible to have an individual set in which all s are
<
positive, satisfying the Pauli principle by taking different

the terms of highest multiplicity have s = xl2. The largest
by taking
ma = m
I,
l2
= 11,
...
m values. Hence
m is clearly given
l
in which case
a d shell this gives *$, 2D Z F, *F, 5D, and Q S in agreement with the
table as x ranges from to 5. For x ^ 21 -f 1 we can have ms positive for the
first 21+ 1 individual sets and for these Smz == 0, for the others we must take
Thus
for
ms negative and take the remaining w/s as large as possible. This means that
from the middle of the shell on the highest multiplicity diminishes by unity
for each added electron and the associated L values run through the same
sequence as in the
first
half of the shell.
Hund
* that the
term of largest
S and
among these the term of largest L is lowest in energy.

It is easy to find an expression for the (y SL) of the term of largest
S and
It is
an empirical
fact,
noted by
L in these configurations by the method of

of largest
sole
occupant of the
The
result is that for this
cell
and
47 since this term occurs as the

,
ML in the table of complete
sets.
term
the plus sign applying for x
<
(21
+ 1) and the minus sign for x
>
(21 -f 1).
This
an
interpretation of the trend of the observed (y SL) in the

elements of the iron group in which the 3dx configurations occur. These are
result affords
We
7
see that they remain fairly constant in the
plotted against x in Fig. 6
first half of the shell, then reverse sign and increase rapidly in magnitude in
.
the second half of the
shell.
Using the formula just derived

>,
we can
in each
infer the value of

3<J
Linienspekiren, p. 124.
14
THE BTTSSELL-SATnSTDE&S CASE: ENEEGY LEVELS
210
element from the observed
(y$). The
67
values so obtained are also
smooth curve showing a steady

what we should expect because
with increased atomic number the radial function B(3d) will be drawn in
plotted in the same figure. They lie on a

increase with atomic number. This is just
toward the
origin, giving
more weight
to the regions
where
-=-
is
large
and consequently increasing the value of 3d A similar situation holds in the

x
palladium group where the 4d configurations occur, but the result is
rougher because the data are not as complete and the departures from
t,
Russell-Saunders coupling are greater.
Mg.
6 7 . Term-interval parameters in the 3d* configurations

of the iron-group elements.
PROBLEM
Show that the term
a doublet with
7.
The
Our
of Mgiiest
L value for
L=xl-(x- 1) 2 if x is odd
triplet
terms
of
nl x is a singlet with
L =xl - Ja;
(a:
- 2)
if x is
even, or
(x^ 21 + 1).
helium.
discussion of the relativistic theory of the one-electron problem in

how intimately connected is the spin-orbit interaction energy
5 5 showed
with other relativistic
mately
thus far
effects.
by introducing the terms

with
2fc )LfSt
{
in the Hamiltonian, and in

is
We have allowed for this interaction approxi
57
useful in correlating a large
&<)=-
we have seen that this simple approximation
amount of empirical data on intervals between
THE TRIPLET TEEMS OF HELIUM
77
211
a term. In Chapter xi we shall see that is also quite suitable for

discussion of a wider class of cases in which the spin-orbit interaction energy
becomes comparable with the electrostatic interaction of electrons not in
levels of
closed shells.
But there are cases, particularly for light atoms, in which a more refined
discussion of the spin terms is necessary. The classical instance is that of the
3
3
3 3 and 4 3 D. These
triplet terms of helium. Data are known for 2 P, 3
terms are narrow and inverted, and the departures from the Lande interval
rule are so great that they were at first thought to be doublet terms. Thus
for
2p*P the
interval ratio is 1:14
the Lande value of 2:
1.
These facts
1
1
(- 0-07 cm- -0-99 cm- ) instead of
have been the subject of a number of
:
papers* and are definitely related to the inadequacy of the usual approxima
tion for spin-orbit interaction. The question has been approached in two
ways. One is to adopt the model of the electron as a small magnet and to set
magnetic interaction with the nucleus and

mechanics. The other is to take the
relativistic classical formula for the interaction of two moving charges and
try to adapt it to quantum mechanics by methods suggested by Dirac s
up the additional terms for

other moving electrons by
its
classical
theory for one electron. The

Heisenberg, his
work being
first
earlier
approach is that originally adopted by

than the relativistic theory of Dirac.
The spin contributions to the Hamiltonian, according to Heisenberg, may

be written for the two-electron problem as
((ri-
e*
o.
fSi
S.^-SSi-fo-rJS^fo-r
^12
to be calculated simply for the Coulomb field of the

terms which we have already considered in
contains
nucleus, so the first line
the theory for hydrogen. The next two lines represent the interaction of each
In this expression
electron
|(r) is
magnet with the
field
produced by the other
electron,
v and v%
being the velocities of the two electrons. Finally the last line is the direct
interaction of the magnetic dipoles of the two electrons. The terms of the
last three lines are given
by purely
classical considerations
by
associating
* HEISENBEKG, Zeits. fiir

Phys. 39, 499 (1926);
GAUNT, Proe. Roy. Soc. A122, 153 (1929); Trans. Roy. Soc. 228, 151 (1929);
OPPEKHEIMEB, Phys. Rev. 35, 461 (1930);
BRETT, Phys. Rer. 38, 383 (1930).

14-2
THE BTJSSBLL-SAIJNDEBS CASE: ENERGY LEVELS
212
77
with each electron a magnetic moment

eS[jJiC. The problem of spin-orbit
interaction in a particular spectrum is thus reduced to the calculation of
the perturbation energy associated with these terms.
rearrange the terms in the second two lines we see that they can be
expressed in terms of the products L^Sly L^S^, L^Si and L^S^ together
If
we
with terms in which mixed expressions like {r x x v 2 )

in L*S is found to be
Sl occur. The whole term
on using the definition of ^(r^ in terms of the Coulomb field. This shows in a
rough way how part of the mutual interaction of electrons comes in to reduce
the full effect of the nuclear field in the terms of the form L^Si and affords
a slight justification for the use of U(r) in place of the Coulomb field for
calculating spin-orbit interaction in complex spectra. But the whole
situation is at present quite unsatisfactory and calls for more accurate
treatment.
The approach through Dirac

netic
moment
interaction
s equations
in the model. It starts
by the
does not
make
use of the
mag
by replacing the ordinary Coulomb
interaction law of classical electron theory
which includes the magnetic interaction of the moving electrons as well as

The velocities vjc and v2 /c are replaced by
the matrix vectors p x and (3 2 of the Dirac theory and this term incorporated
into a Hamiltonian which also includes Dirac s relativistic Hamiltonian
written down for each of the two electrons. This procedure is entirely pro
their electrostatic interaction.
visional
and is
details.
When
beset with
numerous
difficulties,
the Hamiltonian so formed
so
is
we shall not
give further
reduced by approximate
elimination of the small Dirac functions that go with negative values of the
rest-mass term it is in close correspondence -with the one obtained from the
classical
model. The differences are akin to those discussed in
to the $~levels of the one-electron problem.

The most careful computation of the triplet structure
Breit,
who
cases of 2 3
is
5 5 in relation
that
made by
obtains a quite satisfactory agreement with experiment in the

of He I and Li II.
CHAPTER VIII
THE RUSSELL-SAUNDERS CASE: EIGENFUNCTIONS
In this chapter we discuss not only the calculation of eigenfunctions in LS
coupling in terms of the zero-order functions, but the question of the mean
quantum numbers referring to less than the whole atom for anti
symmetric states. In 1 we find that operators may be rigorously defined to
characterize the resultant momenta of all the electrons in a given shell in the
ing of
configuration. This enables us in
2 to introduce a characterization of
LS
coupling terms by parents and grandparents and in 3 to calculate the Lande

intervals of a term from those of its ancestors.
4, 5, and 6 are devoted to
calculations of eigenfunctions proper; while 7 discusses the calculation
from these eigenfunctions of the separate energies and Lande splittings in
cases where
two terms of the same character occur in a

two 2D s of d?.
configuration, with
detailed results for the

1.
Vector coupling in antisymmetric states.*
In this section we consider the meaning of vector coupling in antisym

metric states, and the extent to which the matrix methods of Chapter
are available for use in connection with such states. The situation is roughly
as follows. In the antisymmetric state characterized by the quantum

numbers riWm^w^, n 2 l 2 m*m$ ...; of what operator is m] the eigenvalue?
all the m^s are
s -of-the-first-electron (unless
Clearly not the operator
l
in
the
nl*&
all
other
the
differs
from
configuration, it is an
equal). But if n P1 ^-electron. If n 2 P also differs from
eigenstate of the operator Z^-of-the-%
all the other nl s, we may add the two J7s to obtain a resultant L and
L
and the two /S s to obtain a resultant 8 and 8 for the n 1 1 1 and n 2 1 2 electrons
3
by the formulas of 14 Then in terms of these states the matrices of 17 s
l
and $ s of the n P- and n 2 l z electrons would be given by 9 3 11 and 10 3 2. But
2 2
we can no longer define an operator i-of-the-^P-electron,
if T^PsTi ?
since no operator will distinguish between the two electrons in an anti
symmetric coupled state. It is, however, still sensible to define a resultant L
for the two n 1 ?1 electrons, but this operator will not be the sum of two com
muting angular momenta, and will not have the allowed values determined
1
1
by the addition of a vector Z to a vector Z Hence if a group of equivalent
electrons occurs in a configuration, we must be content to work with the
,
whole group as a unit in our vector coupling, not trying to define the angular
momentum of less than the whole group. These ideas are given precision by
*
SHQRTLEY, Phys. Rev. 40, 197 (1932). For a discussion from the point of view of second
quantization see JOHNSON, Phys. Rev. 43, 627 (1933).
THE RTTSSELL-SATOTOERS CASE: EIGEKFUJSTCTIONS
214
I8
the following consideration of the coupling of two inequivalent groups of

electrons.
We consider a group of I electrons (1 ...NT ) in a given configuration I.

Let us denote the allowed terms for this configuration by the quantum
numbers y1 S1 L1 and the corresponding antisymmetric
,
states
by
X^&UMIMD^W.
(i)
Such a state will be a linear combination* of products of the JVj one-electron

functions of the type
SamX)P <k^zxK)<^
xI (Ai)=
m*raf
The phases in
zXf)----
(2)
for the states of a given
term are supposed to be chosen so

that the matrices (AÎÂ St and (A^jA *) of L I = L l + L 2 + ... + L
and
Nl
3
=
S
..
have
+
-f
S
the
values
7.
4Sj
:
5*2
given by 3
Nl
(1 )
Consider also a second group, ofNTI electrons
(JVj
+1
+ 2,...,NI + N u
X (Ai) Xn (An
The products
We
a configuration II whose nl values are all different from any in I.f

denote the antisymmetric states of this group by
in
(4)
of functions of type (1) and functions of type (3) are functions of all
N
I+
U electrons for which the matrix elements of j have the values
=N N
= (APIA S 8(An A n).

5)
(AiAglljIA jA g)
(
1
Let us now define a new operator L to represent, not the sum of the angular
momenta of electrons 1 ... I9 but the sum of the angular momenta of the
electrons in configuration I. Explicitly, we define L1 as that operator which

when acting on a state
of configuration I -f II, expressed as a product
of one-electron functions as in 2 6 5 has the same effect as the
operator
cf>
La 4- Lb +
where a,b ... are the indices in of the electrons in configura

tion I. Here II is any configuration inequivalent to I. When
acting on a state
of a configuration which cannot be represented as I + 11, L1 vanishes. We
make corresponding definitions of n S and Sn
For states formed by the exchange of electrons between groups in (4), the
matrix elements of L1 have the values
.
. .
<f>
3L
Xgi(Ai)
where
XOII (AH)
iXoa (A i) XOTI (A H) =
S(<3,
Q ) (X An|i|Ai A*),
(A* A*| U\Xi A n)
Here $1
is
(Af Agl^jA J A g).

the set of electrons which the permutation Q of electrons
(6)
(7)
1 ...
puts in place of electrons 1 ... N^ Q is restricted to be such a permutation

that the electrons in QI and QI1 are ordered. We shall for convenience call
*
We shall see in
f The symbol A
J
Two
4s, 58 , 6s how to determine these linear combinations.

have quantum numbers A
indicates that group T of electrons (1 ...

Nj)
configurations related in this way will he said to be
J
inequivalent.
1s
VECTOR COUPLING IN ANTISYMMETRIC STATES
such a permutation an exchange. Formula

resultant of operation
from the fact that the
(6) follows
by L1 on its operand in
(6) is
215
the
same as the result of
operation by LQ^ on this operand. Matrix components of L1 between states

of two different configurations are of course all zero.
The
XÂ
state
X$n (An) s Q XT (Ai) Xn (An)
(8)
thus an eigenstate of (S ) (L1 ) 2 S* L* with the quantum, numbers

1
S 1 M\, M\. A similar statement holds for the corresponding operators
referring to configuration II. We may easily construct an antisymmetric
1 2
is
linear combination of these eigenstates:

J
(9)
where the summation runs over all the Nl/NjlN-^l exchanges Q, and g has
the parity of Q. Since the states in the summand of (9) are all orthogonal,
Y is normalized, and is hence an allowed antisymmetric state of configura
tion I -f II. The matrix components of L1 between states of the type (9) are
seen from (6) to be given by
Hence, the states T of (9) have all the properties of the states
of 6 3 1 with respect to the angular momenta L1 L11 and again with respect
to S1 Sn Therefore we can use the formulas of 14 3 to add L1 and JLn or
,
iS1
and
where
*S
n or both,
,
to obtain states
VW&IfilPJSPIPtSLMsMi),
I
II
= JL I + n =
J
5=S + 5II =S + S
The expressions of 1C 3
for the
(11)
11 .
-|-
matrix components of
S 1 L1 Su
,
n
,
will
accordingly hold for this system of states.
For any quantity F = FI + F 2 +

4- FN let us similarly define F1 as that
which
behaves
like
F
F
+
operator
a
b + ... when acting on a state of con
I
and
which
vanishes
otherwise. Then the action of F1 on a
figuration + II,
. . .
state of type (8) gives
(12)
where
(A*iAn| J^|AiAH)
= (AAg| J^|AjAg),
and a is a state of a configuration which cannot be written

III + IL* Similarly the action of F11 on this state gives
(13)
in the form
(U)
P
*
For example, if configuration I is d* and II p 2, a typical III is

while a typical a is
P 1 has of course no components between
configurations differing in more than one electron.
</,
rf
THE BUSSELL-SAinSTDEBS CASE EiaElSTFUNOTIOKS
216
I8
where
ft
a state of a configuration which cannot be written in the form
Is
I-j-IV.
From
these formulas
F= F +F
11
we
see that the general matrix component of

+ II and III 4- IV is
connecting states of configurations I
mX
XQi(A
Hence the matrix components of

the type
(9)
From this,
F between two antisymmetric
states of
are
we see that in evaluating these matrix components

joining corresponding coupled states of type (11), we may treat
F1 -h F* where F1 commutes with Sn n and F^ with S\ Ll If F is
(13),
and
(10)
and those
F as
a vector of type T ( 83 ) with respect to either L or S, and commutes with

the other, the matrix components of 1 and 11 between states of
type (11)
be given by the formulas of 1 1 3

To summarize, we have found that the matrix components of a
quantity
of type F between an antisymmetric state of
configuration 1 4- II and an
will
be reduced to those of FI between

and of Fu between antisymmetric
way we should if the uncoupled states
of I + II and III -f IV had been taken of
type (4), without exchange
antisymmetric state of III 4- IV
may
antisymmetric states of I and III

states of II and IV in just the
of electrons in the process of

antisymmetrizing as in (9). Here con
I
and
II
must
be
figurations
êquivalent, and III and IV must be
inequivalent.
2.
Genealogical characterization of LS-couplxag terms.
As pointed out in I 7 a term is not in general

completely characterized by
its configuration and
L value since a given configuration may contain
several terms of the same sort. We shall now
investigate the allowed terms
,
from a vector-coupling standpoint with a view to

obtaining where possible
a more general characterization and an introduction of the
important
concept of parentage.
If we divide the electrons of a
configuration into two inequivalent groups
as in the preceding
we
see that the allowed terms of the con
paragraph,
figuration
and
may be obtained as in
181 1
vectors of the groups. This
by coupling in all possible ways the S

means that we can obtain the allowed
7
terms for any configuration by
using Table I of allowed terms for groups of
equivalent electrons, and that these terms will be characterized
the
by
terms of the constituent groups. Where a

group contains three or more
equivalent d or / electrons, we cannot distinguish between the terms
2s
GENEALOGICAL CHABACTEEIZATIOlSr OF i^-COTJPLTNG TERMS 217
occurring
more than once
functions, since it
Thus
in this group except by giving the actual eigenimpossible to divide such a group.
is
for the hypothetical configuration p 2 d 2 the allowed terms are

,
3
tf>
3
(
P)
P) d*
S)
S
(
P)
P
3
i.*.5Pi>
P)
*PDF
(iD)iSPDIG
3
(
Although there are
six each of
individually characterized
3
B,
s,
and
3
J"s
P)
(1)
*PDFGH
occurring here, each

2
by giving the terms of p and d from which
is
it
arises.
When one electron is added to an
terms of the ion are called the
ion, the
parents of the terms of the atom. Thus if we consider the addition of nonp, and d electrons respectively to the sd configuration of an
equivalent
,
we obtain the terms
ion,
sd
sdp
sd(T-D}s*D
sd(iD)p*PDF
3 d ( 3D) s
*, *D
3
(
D) p PDF,
*>PDF
dd
d C~D) d ZJSPDFG
d ( SD) d
*SPFG
(2)
f
*SPDFG.
The groups of terms of the same multiplicity which arise from the addition
of s, p, and d electrons are called monads, triads, and pentads respectively*
(in general potyads). Thus, in the example above, the configuration sdp
contains a doublet triad having as parent sd( lD) and a doublet and a
quartet triad having sd(*D) as parent.

The occurrence of these close groups of three or five terms was noticed
empirically soon after the beginning of the analysis of complex spectra.
Hence the eigenfunctions which arise in the convenient theoretical designa
tion by parentage must to a certain approximation be the actual eigen-
fanctions which diagonalize the electrostatic interaction, even when more

than one term of a kind occurs. The reason for this is illustrated by the plot,
in Fig. I 8 of the energy-level spectrum of
1. This spectrum appears essen
tially as the sum of three one-electron spectra converging to the three terms
,
of the low 2^ 3 configuration of II as limits. The terms built on each of the

parent levels have essentially the energies of a single valence electron moving
in a central field due to the ion in a particular state. For this spectrum the
central-field approximation which regards the electrostatic interaction as a
small perturbation on each configuration is quite invalid; the electrostatic
interaction of the core is larger than the separation between configurations
in the central-field problem. Large interactions between configurations are
*
in this way even when the number of terms

P) d gives a pentad containing only three terms 2 PDF.
They are designated

X
e.g.
is
not as large as indicated,
218
THE BUSSELL-SAUNDERS CASE
EIGENF1JN"CTIONS
28
expected, not only
among the low configurations plotted, but between these

and configurations high above the ionization potential which have other
terms of II as parents for the agreement with the formulas of 5 7 of 2p 3 Bs
3
is no better or worse than the
agreement of the II 2p parent, and the per
turbations of this parent are caused by higher states of II. But whatever
;
-40
-30 2
7s
20
3p
4:
&0
&
30-
$0L.
Fig.
1s.
2p fS)
The energy
2p
(*D)
2p
(*P)
levels of
oxygen I. This shows aU observed levels except 2p 4

and 2s2p 5 *P (-13,000 cm- 1 ). Hydrogen levels are
(76,000-110,000 cm- 1 )
indicated by broken lines.
the perturbation of OII# 3 it is certain that the actual

eigenstates of the
p 3 levels are to a good approximation the parents of the plotted levels of O I.
The electrostatic interaction of the valence electron with the ion causes
only
the small separation of the
singlet and triplet, or triplet and quintet polyads.
,
28
GENEALOGICAL CHARACTERIZATION OF LS- COUPLING TEE-MS 219

In
many of the complex spectra, such as those of Ti,
V, and Zr, where the

of
1, the levels of a
overlapping
configurations is not so exaggerated as in
given configuration will still be found to occur in groups (polyads) having
approximately the relative locations of the terms of the parent configura
The parentage is not exact, however, for as soon as the intervals

within the polyads due to the electrostatic interaction of the valence
electron with the ion becomes comparable with the separation of the
tion.
polyads
from
3.
we may
expect an interaction between the similar terms arising
different parents.
Lande intervals for terms
With the
results of
we may
of
coupled groups.*
easily obtain a formula for the intervals
in the terms of a configuration I -f II in terms of the intervals in the in7

equivalent configurations I and II separately. From 3 2 we find that
Y SL) = S (yLg(rt)L&L) (yS\S&8).
(I)
1=1
We have omitted the sum over y
because each quantum number in y will
in general either refer to the spins, in which case f(^)J^ will be diagonal
with respect to it, or will not refer to the spins, in which case S^ will be
1
1
diagonal. If now the term ySL is derived by the addition of the term S L
of configuration I and the term
two parts, one of which refers to I, the other to

two parts
spin-orbit interaction into
This breaks the
sum
S11 U1 of configuration II, we may split the

II.
(1) into
lV:^
(2)
.)
II
In accordance with the discussion at the end of
I8,
we may
evaluate
each of these parts independently without troubling about the fact that the
eigenfunction is antisymmetric. In the first sum, |(rz-)L and S^ may then
be assumed to commute with

given by
11
38
we
if
j->L
11
,
in
which case the factors are each
correlate
Q-^fa)
This reduces the
U and S
first
or
L<
or S,
sum
jj+If or
3r*IP
to a
known multiple
&
or SP.
of
GOUDSMIT and HUMPHBEYS, Phys. Rev.
31, 960 (1928).
THE RUSSELL-SAlHtfDEBS CASE EIGENFUNCTIONS
220
38
The second sum is
similarly reduced, to obtain finally
2L(L+
D-ffi(ffl+i)+s^(s^+i) rtanr
(4)
By this formula, the values of for any configuration may be expressed in

terms of those for configurations of the type nlx which were given in Table I 7
In particular, if a single ril electron is added to an ion, QS&IP)**^.
.
Formula
(4) gives the values of f immediately for two-electron configura

tions nlril if we take
^
we
as
shall
show in
8
,
this gives the correct
and
answer even
if
the electrons
are equivalent.
Calculation of eigenfimctions by direct diagonalization.*

We wish now to show how to determine the eigenfunctions of states in
LS coupling in terms of states in the zero-order nlms scheme.
4.
States characterized
characterized
SLMSML
by
by
SLJM
are given directly in terms of states

the formulas of 143 States characterized
SLM8ML by
for configuration I-fII are given in terms of those of con

8
figurations I and II by the general procedure of I this is discussed in detail
by
68
These considerations reduce the problem essentially to the deter

mination of eigenfunctions in the
8
L scheme for configurations con
3
sisting wholly of equivalent electrons. In order to use the formulas of 14
in the way we have sketched, it is however necessary that the relative
in
8LM M
phases of the states of a given term be such that the matrices of

have the values given by 3 3 7.
L and S
The direct method of obtaining eigenfunctions characterized by SLM S L

2
is to diagonalize the matrices of
and S2 in the zero-order nlmsml scheme.
Since such a diagonalization does not give states with phases chosen in any
particular way, it is necessary, if we wish states characterized by SLJM,
to diagonalize also the matrix of JT2 (or, for greater convenience, of L*S).
Because the process

little
is
mination of the matrices of

*
and the phases
laborious
practical application.
We
Z,
2
,
JOHNSON,
arbitrary, this
shall, nevertheless,
S2 and L-S in
,
Pfcys.
Rev. 39, 197
the nlms
(1932).
method has
consider the deter
scheme because
4s
CALCULATION BY DIRECT BIAGONAMZATION
221
our discussion in Chapter xn of transformations for

configurations con
taining almost closed shells will depend on the properties of these matrices.
Because the simultaneous diagonalization of these three matrices is an
excellent illustration of the process of matrix
diagonalization described in
72,
we shall give detailed results for the configuration p 3

The matrix components of L 2 S*, and L*S in the nlm/n; scheme may be
easily expressed in terms of the one-electron components (cf 3 3 4 and 7 3 3)
.
(a\Lx \b)
(n
a la
m%mf\l
(mf
1,
m\) ffiV(l
b
m*>)(l
+ m\ +
(la)*
1),
ft
=8(*Zmf,w Pmf)8(mJl,mJ)i,
(a\Sx \b)
(lb)*
(lc)*
6
6
by the formulas of 6 7 and the relation 3 3 6. In this way the following
are
results
obtained. (The sign is in all cases to be chosen positive or
negative
according to the even- or odd-ness of the permutation which changes the set
A from its standard order to the order in which sets in 1 which match
,
those in
a,
to
all
stand at the same places in the
lists,
with
a!
corresponding to
b).
Matrix of L 2 We get non-diagonal elements of JL 2 only if A differs from

A in two individual sets, say that A has the sets a, b while A has a & This
component has the value
r
(A\L*\A )=
4{(a
The diagonal element
Lx \a
of JL
)(b\Lx \b
)-(a\Lx \b )(b\Lx \a )}8(M,M
Matrix of
run over
all
(3a)
is
(3b)
where a and
).
a<6
the individual sets of A.
Sz We get non-diagonal elements of S2 only if A

.
differs
from
A in two individual sets, say a and b. This component has the value
(A\S*\A )~
4{(a\Sf \af)(b\8x \b )-(a\8x \b )(b\8f \^)}^Jf,M
).
(4a)
The diagonal element of S2 is

2
(^15 |^)=^(Jfl-f^~JV
where
is
* It
is
A is the number of pairs of

same nlml9 each pair counted only once.
the number of electrons and
dividual sets with the
(4b)
),
T/
in
to be noted that for any of these quantities

a
(n
wij|
\n* I*
mj mf) =(n*
This property will be of use in Chapter xn.
-m -m|
\n
V>
-mj
- wf).
(2)
THE RUSSELL-SAUNDEES CASE EIGENFUNOTIONS
222
Matrix ofL*S. IfA

6,
we
differs
from
A in regard to two individual sets, a and
obtain
-(a \Lx \b
If
48
differs
from
A in
)-(b[Lx \a )(a\Sx \b )}S(M,M

regard to one individual set, a,
)(b\Sx \a
(A\L-S\A )={(a\l-S\a )-2 Z(a\La.\b)(b\8ie a

"-2
).
(5a)
),
(5b)
I, (a \L,\e)(e\Sa \a)}8(M,
where 6 and c run over all the individual sets common to A and A It is
worth noting in connection with this formula that one obtains no value
+ The second term in the braces
unless a, a are of the form (m/)", (mz
1)
vanishes except for one value of b namely (m^ 1)~; the third term vanishes
+
except for one value of c, namely (%) When either of the last two terms
.
does not vanish

just the value
when
(i.e.
(a\L*S\a
the appropriate 6 or
).
c exists in
the set A),
it
has
The diagonal element of *Sis given by
With these formulas it is rather easy to calculate the matrices of i 2 S2

and L Sin. the zero-order scheme. Only those sets of the configuration which
,
are outside of closed shells need be considered in calculating these matrices.
The non-diagonal elements are obviously not affected by the presence of

closed shells. The part of a diagonal element which results from a closed
shell is easily seen to
Illustration.
be zero [using 3 7 5 in the case of (A \L?\A)].
The configuration p*. This method of calculation
will
be illustrated in detail for the
2
case of the configuration ^J3 . This configuration leads to the five levels 4/S |, 2Pj, 2P|, 2D|, D|.
We shall consider only states characterized by = J this will give us one state of each level. The
zero-order states for
M=
J-
are given
by
(1-0+
0-)
(!+,!-, -1-)
(6a)
D
E
If we
make a plot such as
while
has a state only in
we see that the terms 2P and 2J9 have states in both box a and box /J,
Hence we may infer that the eigenftmctions for 4
will not involve
|
]3,
In this scheme, the matrices of S2 2 , and L S have the form

the lines in the matrices indicate the separation between a and j3)
states
A and JB.
JB
(cf.**
7 2 for notation;
D E
(6b)
CALCULATION BY DIRECT DIAGONAIJZATION
223
ABODE
(60)
ABODE
l\(A\L-S\B)\\
p*.
(6d)
So far as the diagonalization of Z.2 and *S2 are concerned a and j3 are quite independent. Consider
the matrix of 2 For a. the eigenvalues of S 2 which may occur are those corresponding to the two
we have in addition to these the quartet eigenvalue ^f-. trans
doublets, namely f and f . For
formation which diagonalizes this matrix is found by the method of 7 2 :
(6e)
\\(A\*M)\\--
(Because of the degeneracy this transformation
is
not uniquely determined for either pair of
doublets.)
When the matrix (6c) of I, 2

the relation (cf. 7 2 7)
is
transformed to this scheme in which S 2 ,
11!,
Sz Lz
,
are diagonal
(6f)
by
THE BUSSELL-SAUNDEES CASE EIGENFTTNCTIONS
224
it
takes the form
*,
2
is diagonal with respect to $, as required. The eigenvalues of Z, for both the stepmatrices
to P and D, namely 2 and 6. The
| and , /S=| are now known to be those corresponding
transformation which diagonalizes (6g) is given by
whicli
a,
(6h)
The new
a.
or
states are
now
in the
SLMsAfz
scheme, the values of
and
MI being indicated by
j8.
TA
transformation from the zero-order scheme,

(of. 7=6) of (6e) by (6h):
SLMSM.L scheme may be
to the
obtained by the
nmltiplieation
We
have chosen the phases

so that
A
PW
oc
atiu
fl
p
,
*!>..,
cc
jina
Q
p
as
have the proper
relative phases
*
*
*
2
determined by the method of the next section, and so that 2
Pf î* -^| an(l | are the proper
linear combinations of these states as given by 143 In this way we can make full use of these
matrices without having to rewrite them, with more useful phases.
.
CALCTJLATION
When
the matrix (6d) of L
BY DIRECT DIAGONAUZATION
S is transformed to this
scheme,
it
225
becomes
(6k)
which is diagonal with respect to L and 8, as required. The eigenvalues of this matrix are calculated
from the formula L-S=MJZ -L 2 -S*), which gives -1, J, -f, 1, and for 2Pj, 2P|, a
l>j,*jZ>|,
and
^s respectively. The transformation to these
states is given
by
(61)
TAe transformation from

(63) and (61):
the zero-order
scheme
to the
SLJM scheme, is
then given by the product of
^
-&
(6m)
D
E
* See footnote to
(6j), p. 224.
11
THE BUSSELL-SAFNDERS CASE EIGENEFNCTIONS
226
7
It is Interesting now to verify" directly the electrostatic energies which were obtained in 5 by
7 6 and 8 6 we find for the matrix of electrostatic
the diagonal-sum role. From the results of
interaction in the zero-order scheme
CDS
(7)
When
this
matrix
is
transformed by (6m) to the
SLJM scheme, it becomes
(8)
which agrees with the results of 5 7 The same eigenvalues would have been obtained by trans
forming (7) to the SLM s M_i scheme by (6j).
.
5.
Calculation of eigenfunctions using
angular-momentum opera
tors.*
We shall now give a convenient method of computing SL eigenfunctions

proposed by Gray and Wills. Having found the eigenfunction *(SLMS ML
ML = L,
this
SL
term
method makes use of the symmetric operator

-iL
to
the
^(SLMS L 1) by 3 3 3, and the
eigenfunction
x
give
y
operator ?~Sx iSy to give ^(SLM8 lML ) and so forth. In this way
for
&=L
~S,
the whole of an
found with the proper relative phases for all

3
states, so that the results of 14 may be used directly to obtain the SLJM
states. The first state is found purely from orthogonality considerations;
hence the states of different terms have no particular phase relation. We
shall need to discuss only the transformation to the SLMS
L states.
In order to carry out the operations mentioned we shall need to know the
and f on a state expressed in terms of the anti
result of operating with
zero-order
functions
Since J5? = ? l +
(mJ m] m| mf
)
symmetric
2 +
we have from the general relation
is
<=$?
<
>
, . . .
. .
(1)
*
GRAY and WELLS, Phys. Rev.
38, 248 (1931).
USE OF ANGTJLAIt-MOMEKTIJM OPERATORS
227
the result
Since S?
is
When we
With
a symmetric operator,
apply s to
(2)
we
it
commutes with the operator $$ of 3 6 6.
find
by J? on any
this formula the result of operation
state
is
easily
calculated.
result of operation by f is very similar, the

s values being suc
cessively decreased by 1 in place of the m^ values. The coefficients which
The
occur are limited to
and
0, 1
for
m = J,
for
m=
J,
so that operation
by P
gives just the sum of the states resulting when successive m^ are
if one has occasion to apply it, is just the
changed to mj~ The operator
siirn
of
and f\
d3 As an illustration of this method of

the
of
some
calculation, let us obtain
eigenfunctions for the configuration d*;
in particular let us obtain one eigenfunction for each of the two 2D s to use
Illustration
the configuration
classification of the
7 8 for the separation of the energies of these terms.
2
down
to
zero-order states of d? according to
,
|,
together with the
s
L
in
M M
7
(cf. 1 2) :
terms which have components in each partition is given below

jjfefff
Jfcf
jfrfff
(2+2-
B
C
(2+2(2+1+
E
F
(2+2- -1+)
H
i
K
L
(2+ 1+
(2+ 1-
j\f^
1+ )
0+)
.
(2+1+
(2+1+-1+)
0+)
4
[
.F]
0-)
0+)
0+)
(2+2- -2+)
(2+1+ -1-)
(2+1- -1+)
- 1+)
(2- 1+
(2+0+
(1+ 1-
0-)
0+)
then, to obtain two (orthogonal) eigenfunctions for

D, J, 2. Before beginning the calculation proper, let us calculate the action
and f on certain of the above states [omitting the factor 1 , cf. (3)]:
of
Our problem
is,
&
ft"
(5)
Sfl
=
15-2
58
THE RUSSELL-SAUNDEBS CASE EIGENFUNCTIONS
228
Now the state *H,
|,
is just
A. From
(1)
we
find that
But
from
(5).
#, J, 4= 10-[V&B- 20].
Hence
The other 2H
s are obtained in succession:
(6a)
2
The
state
4.
G,
The
\>
\,
#,
is
i 2 =30~V - K + 3
*,
J, 3
G, J, 2
a linear combination of
choice of its phase
2H]
- 2M - 3xV + V 60].
=20" *[Vftff
is
and C orthogonal to
purely arbitrary.
4-3JT
- G - 2Q
(6b)
= V^[2J + 3^ +i - 4M + 4^T + V|O].
We have now 4J, f 3 = D. By operation with ^ on this we obtain 4J^ |, 3:

,
i 3) = V3
3 is determined
by
its
4
(
J, J,
3)
orthogonality to the above three states for
=
of the partition
L =2. The other states
8
of this partition are both
s; any state orthogonal to the four above is a
2
and there are two linearly independent such states. Let us choose one
We have now four states
\>
at
random and a second orthogonal to it

[- J- K+L
:=i
l
=84~ [-5J + 3JT+L-4Jf-3zY-2V60].
This procedure may be continued to obtain any desired state. Any other
A check is continuously furnished
configuration is handled in the same way.
by the fact that the states must come out properly orthonormal.
6.
Calculation of eigenfunctions
from vector-coupling formulas.
Let us first consider how we may obtain the eigenfunctions resulting

from the coupling of the terms of two inequivalent groups of electrons as
inl 8
XI (/S i
I
Jf|Jf) represent one of the antisymmetric eigenfunctions

of a given term of group I. This is supposed to be known in terms of the zeroN for the
1
=
in
a2
order functions $
Let
I (J.)
3>
I (a
...,a
*)
configuration
question.
229
USE OF VECTOB-COUPLTNG FOBMULAS
68
Similarly let n (/S^ 1^ Jf |* Jf *) be one of the antisymmetric eigenfunctions

of a term of the second group. The product of these two states will be a
known
linear combination of the
form
A,B
This
is
antisymmetrized as in
easily seen that
is
rf
by
application of the operator
*5/
It
ââ^...^*!)^^
an antisymmetric zero-order function
for configuration I
-I-
II;
hence the
antisymmetric function
(2)
In other words, these antisymmetric functions are obtained by simply

s for I -f II.
s for I and II by
replacing the products of
3
14
to the coupling of $* and
of
the
how
formulas
us
see
let
to
Now
apply
<D
<l>
Let us write the function Y purely symbolically as a

product of a function referring to the S s by a function referring to the L s:
1
T(# If M\ Ml ,S*IPMY n) = T (# /S11 Jf1 Jf I1 ) I? in Ml If* }. (3)
S11 and
of U- and
IP-.
Let T(S1
given by
the linear combination of the r s which is

) designate
14 3 for the coupling of S 1 and S 11 ; let Q(ZJ i11 i Jf L ) similarly refer
to the coupling of Dare given by
and
^(fflI^ SÎfl
y
where the right
Two
o>(
SII SMS
>
D1 *
Then
SLM ML
s
symbolically, the -k$-eoupling states
= T(Si S^SMs )a(VL^LML \
(4)
side of this equation is to be interpreted according to (3).
non-equivalent electrons.^
eigenfunctions for a configuration consisting of just two nonequivalent electrons are given by these considerations if group I is one of the
The
electrons,
and
group II the other.
Y of (2)
just
which we write as in
Xx Xn of (1) is then just
(3)
as
T(mjm|)o>(mjmf).
1,
T{
1,
0)=2"*[T(+
T(
1,
-I)=r{--).
correlation
+)
T(
l)-^
Making the
^
>
-)+T(-f +)]
Note that we are not here following our convention of capital letters for antisymmetric
functions, small letters for all others.
f BABTLETT, Phys. Rev. 38, 1623 (1931).
THE KUSSELL-SAUNDEHS CASE
230
If
we
consider in particular the configuration np rip, with the correlation
=0, If j=0)
The
68
EIGEOTTOTOTIOlSrS
H(
I,
Q{
1,
JQ(
1,
O(
2,
&L eigenfonctions
1, 0)
=:3"*[cp(l,
1)
0) +cw(
-a>(0,
1, 1)]
l)=2^[a>(l,0)-a>(0,l)]
0)=2~*[o>(l,
1)
-!}-( -1,1)]
-
1)
2~*[<o(0,
o>(
- 1,
(6)
0)]
etc.
(4)
are then the symbolic products of
=*3~*[$(1+ - 1+) $(0* 0*)
+$( - 1+
T and fi:
1+)]
etc.
For the important configurations

Il
Hence from
(5)
,,
L ,Q).
a>(M
0,
1,
Jfi)=>(Jft,0+)
2~*[<(
=Z
OTUT
J2_j,)
\
M1
(for allJft)
(8)
0-)
<
/7
(^""a
(7)
we have
Is,
t(L,M ) =
(pp)
r/Fk/ H,r4-
/f\tKT
{v
{\4-\~\
//vx
ifS]
Addition of an electron
to
an
ion.
Let us consider the addition of an s electron to the configuration pp to

obtain pps. In particular, let us find the eigenfunctions for the two 2 $ s
3
2
3
Z
namely pp (*$) s S and pp $) s $. The states pp^S and $
are given by (7). The first of the 2 $ s is given merely by adding + or 0~ to
the single state of p p 1 /S and antisymmetrizing:
which
result,
^
For the other
i, 0)
=6~*[^(l+ - 1-
0) -0(1-
-1
we must add S1 = 1 to S11 = ( to obtain 8 =

T(5=i,Jf s = i)=3^[V2r(l,-)-T(0,+)]
i
T(flf=J, Jtfj,=
-i)=3""
[r(0,-)
-V2 r( - l
^
f
+)].
Hence
)
-!,
This state
etc.
-( - 1+ 1- 0+) ~$( - 1(10b)
2
orthogonal to the other $, |, 0, as required.
characterize the two 2 $ s of this same configuration in another
is
We may
considering the addition of a non-equivalent p electron to p s as

l
2
parent. Let us obtain the eigenfunctions for ps (*P)p *S and ps ( P)p S,
way by
6s
For the
obtain
USE OF VECTOR-COUPLING FORMULAS

=1
first wo must add # = I to Sn = | as In (II), and
&
231
to Ifl=
to
L=Q as follows:
Q(X=0, Jfi =0)=3~
[w(i, -l)
The symbolic product of this with
-0(0,0) +(-!,
1)].
(12)
(11) gives, when we use the functions
(9),
,0)=rl8~*[2$(l+0+ -l-)-2^(0+Ori-0-)+2<&(-l+0+l-)-O(l+Or- -1+)

4
-$(1- 0+ - 1+) +0(0+ 0-0+) +$((T0 -0+) -O( - 1+ 0- !+)-( - 1-0+ 1+)]
J,
The
etc.
(13a)
states of the second
$ are given by
(12)
and
(9):
+${0- 0+ 0) +$( - 1+ 0- 1)
The quantum numbers
ff
np
-<&(-!-
0+ 1)].
(13b)
in these zero-order functions refer to the electrons
np 9 n p, n"s. This
difference of arrangement must be taken into account in a comparison of
the two sets of states.
see that the states (13) are quite different from the
np,
s}
respectively, while those of (10) refer to
We
but since there are only two possible states 2 S, J, 0, they must
be obtainable from (10) by a unitary transformation. This transformation
states (10);
is
in fact the following
-}\^
(
Two
equivalent electrons.
As noted in 46 we may look at the exclusion principle from the following

point of view: We build up non-antisymmetric states characterized by
quantum numbers n^m^m^, ri^m^m^, S12 L12 n^m^m^, ..., SLMS L
,
(The subscripts here refer as usual to the definite electron which has the
indicated by the superscripts.) Then, by application of
quantum numbers
the operator
characterized
cS/ ?
by
these states are antisymmetrized to obtain states still

but not necessarily by any other definite
8
SLM M
L>
quantum numbers. If certain of the electrons are equivalent, two things

happen. First, the antisymmetrized states are no longer normalized and
some of them vanish these are forbidden; second, some of the anti
linearly dependent this reduces the number of
distinct eigenfonctions. In this way we may obtain aE the eigenfunctions
for any configuration by coupling vectors and then antisymmetrizing. In
symmetrized states are
general, however, for configurations containing more than two equivalent

8
electrons, this is less easy than the method of 5 which takes full cognizance
of the exclusion principle from the start.

For the case of two equivalent electrons, however, it is easy to see the
effect of the exclusion principle and to obtain the eigenfunctions by this
method. Let
us, according to the formulas of
143 , couple the s s and VB of
68
THE BFSSELL-SAUKDEBS CASE EIGENFUlSrCTIONS
232
two equivalent electrons to obtain the state ^( nih n SLMS L ). This state
is
literally the product of an orbital function ^ r (n 1 l l n 2 l^LMI) and a spin
function tff^s^SMg). By 14 3 7 interchange of electrons 1 and 2 merely
- l) 2 = ( - 1)* and &, by ( - 1) 28 -8 = ( - 1) 5+1 Hence
multiplies $ r by (
"
-f S
already either antisymmetric or symmetric according to whether
is even or odd. The
the
are
forbidden;
antisymmetric
symmetric states
states are all distinct since they all refer to different
L and hence
S
is
are
all
allowed.
SLM M
TU allowed terms for nP are hence *S, 3 P, 1D Z F,
3
we have found in special

normalized, we may write
as
..., *(2Z),
cases in Chapter vn. Since the states
Y(nZ wJ SLMsMJ^n&nJi SLMS
ML
iff
are
(15)
SL is an allowed term.
Since the operator jtf has the eigenvalue V2l when

6
= 2~*j/^f. It is
applied to an antisymmetric two-electron function (3 )
= n in the eigenfunctions for the con
jtfiff which we obtain if we set n
if
>
nln L Hence we can obtain the eigenfunctions for nl 2 from those

for nlril by setting n = n and dividing by V2 to normalize. Thus we obtain
from (7)
figuration
T(^
while if we set
*8, 0, 0)
n~n
in
=3~*[O(1+, - 1-)
s
/Sf,
these states
(0+ 0-)
all
- $(1~ - 1+)],
1 6)
vanish.
we have
Because of the relation
the important result that we can

(15),
obtain the matrix component of a quantity
l
2 between two allowed
F=F +F
nP
by giving quantum numbers n^s^ to electron 1 and nzs2 l%

This permits us to use formulas of Chapter
for such
8
in
the
same
I
as
we
found
in
that
we
could
use
them
configurations
way
states of
to electron 2.
for non-equivalent electrons. For the matrix

allowed state of ril* and a state ofriln l we have
V8
L M8lfL
component between an
SLMSML F ^Wn^n^ S L MsML

r
M ML F ^(n^ $ S L MSML
- 2-*^^^ SLM ML
FiWn&nHfc S L M8ML
n 2Z2 SL
on interchanging electrons
and 2 in the second term,
= V2$(n&n& SLMS ML
this
y,
becomes
F^n^n^ S L M8ML
r
).
(17)
similar result holds for states characterized by J and M. Hence if we use

the formulas of Chapter
to find matrix components between configura
tions nl 2 and riln V by assigning definite quantum numbers to the
electrons,
we must
set those
multiply the rest
components from forbidden

by
<\/%.
states of
nl2 equal to zero, and
OF VECTOR-COUPLING POBMUIAS
233
No
considerations of this sort have been

given for more than two equi
valent electrons, so that it is probably not
possible to apply matrix methods
to calculations within groups of more than two
equivalent electrons.
Separation of the 2D s of d3 .*
When two terms of a kind occur in a given configuration, we have seen
7.
that the diagonal-sum rule will not determine their

separate energies or
Lande splittings. These terms may be separated, however, by finding the
complete matrix of electrostatic interaction Q by the formulas of 8 if we
know
a set of i/S-coupling eigenfunctions. Since Q is

diagonal with respect
of
it
will
independent
S
L,
only be necessary to use
that part of the matrix of Q which refers to a given
L.
S
8LM8ML and
to
MM
SLM
Thus
we wish
to separate the two 2 s of c 3 we need only use the

second-order matrix connecting the two 2D, |, 2 s of 5 8 6e. This matrix is
found to be
if
3V2l( F2 - 5F4
- 57 J4
2
)
which has for
its
(1)
eigenvalues
(2)
This formula then gives the separate energies of the two 2 terms. The eigen
= J, L =2, may now be obtained. If we
functions for these terms, for
s
write the eigenfonction for either of them as
we
*
A
find
3J?
ot= r/
I
L\
Since
1)
+ 7^22 + 63*4=
.
3^21(^-5^)
and (2) are independent of Ifs and
\*
I
+1
,"H
ML
the same linear combination
D and b 2D obtains for any state of the term.
of a 2
now consider the absolute Lande

7
2
2
T( D, 4, 2) =T( D|), we have from 3 3
Let us
(*D,l,2\H*\*D,l
2)
intervals for these terms. Since
= t(*D).
(4)
Equation (3) gives the eigenfunction for D, \, 2 in terms of zero-order states

= J, JfL = 2. Since these states all belong to the
each characterized by
8
same configuration and have the same s and L no two of them can differ
in regard to just one individual set.
Hence
H1
a quantity of type
* UITOBD and
SHOETLEY, Phys. Rev. 42, 167 (1932).
Jf,
can
78
THE BUSSELL-SATJNBEBS CASE EIGENFUNCTIONS
234
have no matrix components connecting two of these states, so that we can

7
calculate (4) entirely in terms of the diagonal components 4 3. In this way
we
find that
59^-435^4
4 + 8325.F|
V193J1-
We
shall
now see how these
calcula
tions
compare with the experimental
data.
The
Till 3d3
first
()
31450
42700
27300-
instance of d? occurs in
which although completely

mixed up with d 2s gives a fairly good
2
fit for the terms exclusive of the D s
with ^2=845, J4 =54 (3^=17,750);
20-
tn
CC
id
CD
?I5^
see Fig. 2 s If we put these values in (2),

we find for the energies of the two 2 s
.
12,820
served
and 31,450 cm- 1 The only ob

2D is
at 12,710 cm-1 which
.
2
agrees excellently with our lower D.
2
is
The position of the second
pre
dicted 10,000 cm"1 higher than any
tn
Q
Z
G~
c/)
5 -
_
OQ
03
<
<
other level of the configuration, which

may account for its not being found.
The
fit
TH
intervals in this configuration also
the formulas
well.
This
is
zrns
vnr3
3d 3
4d
3d
seen
fairly
s
The d* configuration.
Fig. 2
from the following list of values of d
calculated from the observed intervals (beginning with the widest) in the
.
terms:
*J
85-6,88-6,81*0
*G
89-2
*P
146-7,644
88-9
81-0
*F
102-7
125-0
If the formulas were accurately followed, these values would be all equal.
2
is pleasing to find the
inverted, as the theory predicts. If we take
It
89 cm- as the correct value of

as 142
cm-
1
,
di
we calculate from (5) the lower 2
while the observed value
D is calculated as
677
cm"
J>
is 129-4.
The
interval
interval for the higher
1
.
In the 3$ of VIII, all the terms except *F and one 2

are known.
The configuration d3 has the peculiarity that the calculated electrostatic
2
and 2 are equal. In this instance these energies are not at all
energies for
2
equal, so that we cannot depend much on *H and P. However, if we choose
SEPABATION OF THE 2D S OF
78
235
d*
fit exactly, the *H

while
of the constants
is
The
values
that
of
the
not.
good
energy
=
=
=
which are obtained in this way are 1^ 1171, 4 83 (3
23,891), These
1
constants give for the values of the two 2
and
42,700 cm" The
17,300
the
make
jF s to
the other three terms *F,

2
is fairly
first
of these
the second
is
is
agreeably close to the
predicted 25,800
cm"
s>
D found by White
at 16,317
cm""
higher than any other level of the con
which has a spread of only 16,900 cror1 as analysed Hence it is

not surprising that this second 2D was not found. The intervals as usual
figuration,
agree only roughly;
given
by
fF fits the interval rule
calculated
fairly well,
and if we use the
about an average for the configuration), the

intervals are 248 and
110 cm" 1 the first of these is to be
this level (which
is
1
compared with the observed interval of 147 cm"
Tor the 4d3 of Zrll, in which two 2D s are reported, we obtain an ap
2
36 (3.F = 16,000).
proximate fit of aH terms except P with P2 = 683,
.
F=
With
these values the calculated
D s lie at
11,750 and 27,300
1
cm"
with
309 cm" , respectively. These separations are

separations of 593 and
2
calculated using the J?, which gives an average value for , as standard.
The observed 2 s lie at 13,869 and 14,559 cmr 1 with separations of 734 and
435
1
cm*"
Hence we must infer
that, if these are correctly classified,
one of
3
very strongly perturbed. In this configuration again, d is com
pletely mixed up with
These are all the instances of c 3 which are sufficiently analysed for com
them
is
d?<$.*
2
parison with the theory. In general we have seen that the lower J) corre
2
is predicted
sponds well with the one usually observed; the second
extremely high and inverted.

*
2
interaction between these configurations and the
assignments have "been considered
that he can account for the discrep
who
finds
Rev.
732
UFFOBD,
44,
(1933),
Phys.
by
2
in the above approximation
ancies of Fig. 2 8 satisfactorily. The better agreement of *H over
2
2
s in neighbouring configurations,
is strongly perturbed by
is explained by the fact that
The
in detail
whereas there
is
no nearby configuration which contains a
H term to perturb the *H of d*.
CHAPTER IX
THE RUSSELL-SAUNDERS CASE: LINE STRENGTHS
we shall develop the theory of radiation as outlined in
rv
to
obtain
formulas for the strengths of the various spectral lines
Chapter
for atoms obeying Russell-Saunders coupling. In 1 we obtain some general
results concerning the possible configuration changes in radiative processes
In
this chapter
which are true for any coupling. In 2 formulas for the relative strengths of
lines in a multiplet are obtained as a special application of the results of
il 3 The problem of the relative strengths of the different multiplets arising
.
from a transition between a particular pair of configurations is treated in

the next two sections. In 5 we develop for quadrupole multiplets the for
mulas giving relative strengths of the lines analogous to those for dipole
multiplets in
2.
Configuration selection rules.
1.
From
the results of previous sections we may immediately obtain two

is a quantity
important selection rules. Since the electric-dipole moment
&
which anticommutes with the parity operator

of 1 1 6 it follows that P
has no matrix components between states of the same parity. Hence all
spectral lines
due
to electric-dipole
radiation arise from transitions between
This rule was discovered by Laporte* and is usually

as the Laporte rule. Its importance lies in the fact that it remains
states of opposite parity.
known
valid even in complicated cases where it is no longer possible to assign

configurations uniquely to the energy levels, and enables the spectroscopist
to characterize such levels as uniquely even or odd. The symbol is attached

quantum numbers of odd terms when it is desired to express the
parity explicitly.
is a quantity of type
Since
considered in 6 6 it will have non-
to the
vanishing matrix components only between states which differ in regard to

at most one individual set of quantum numbers. The non-vanishing matrix
component connecting
dividual set
states
and
which
differ in
regard to one in
just the corresponding one-electron matrix element con

the
non-identical
individual sets of
and B. Prom the theory of
necting
is
one-electron spectra we know that this matrix component of

vanishes
unless AZ =
1, where AZ is the difference of the Z s in the two non-identical
individual sets. Hence if the energy levels are accurately characterized

by
configuration assignments, the configuration change occurring in dipole
*
LAPOBTE,
Zeits. fur
PKys. 23, 135 (1924); see also RUSSELL, Science, 51, 512 (1924).
I9
CONMGURATICm SELECTION BULES
radiation
237
1.
by just the nl of one electron, the change in I being restricted to
Such, transitions are called one-electron jumps. Transitions are observed in
is
which there
is an apparent
change in two of the nl values (two-electron
but
this
is
to
be
connected
with a breakdown in the association of
jumps),
with
levels.
configurations
With regard
to quadrupole and magnetic -dipole radiation we observe

from 9 that %t and
are also quantities of type F. The restriction to oneelectron jumps also applies to them. But these quantities commute with the
44
&, so that they have non-vanishing matrix components

states
of the same parity, which is just the opposite of the
between
only
Laporte rule for dipole radiation.
parity operator
Line strengths in Russell-Saunders mtxltiplets.

The ensemble of all lines connected with transitions from one term
r
characterized by L and S to another characterized by L and S we shall call a
multiplet. We wish now to consider the theory of the strengths of the lines of
a multiplet. We confine our attention at present to electric-dipole radiation.
Since P commutes with S, it follows at once that the matrix of P can
contain no components connecting states of different 8. Because of this,
spectroscopists find that they can conveniently divide the terms of most
atoms into systems of different multiplicities, e.g., a singlet system, triplet
2.
system, and quintet system. The lines connecting terms in different systems
are in general missing or weak. For example, only one line connecting the
terms of helium has ever been observed, and

that with considerable difficulty. This is then the first important selection
singlet terms
with the
triplet
rule of the Rmssell-Saunders case: intersystem combinations are forbidden.

In passing we note that this is also true of the electric-quadrupole and the
magnetic-dipole radiations, since the corresponding moments also

with the spin.
commute
We
are given by formulas 7 4 5 if we write ySLJ for oy

have the selection rule AJ = 1,0. Since commutes with S, the depend
J ) occurring on the right of 7 4 5 is
ence on J of the factors (ySLJlPly S
The line strengths
3
given by formulas 11 8 if we correlate^ with
from these formulas that the selection rule on
namely AI/= 1, 0.
The factors (ySL\P\y
SL
S and j 2
L is
with L. It follows
the same
as that
on
7,
3
occurring on the right of 11 8 will be in
since the states W(ySLM s L ) may be expressed as a sum
f
dependent of S
of products of a function of the electronic spin coordinates only and a
function of the positional coordinates only, and the operator
3
positional coordinates only (cf. the discussion of 1 1 ). Hence
in general
S(ySLJ, y
SL J ) =/(
J,
SL J
\(yL\P\y
P acts on the
we may write
L )\\
r
(I)
THE RUSSELL-SATJNDERS CASE
238
LINE STRENGTHS
29
where the whole dependence on S, J, and J is contained in the function

SL J ), which may be evaluated from 7 4 5 and 11 3 8.
First let us consider the multipiets for which L = L - 1 For these we have
f
f(SLJ,
S( Y
SL-U
SLJ, y
axr
4*/(J-f-l)
^-^-+x-,.,^
The sum of these
expressions
is
(2a)
given by 13 5
f
:
that
level
is,
r L-l)^
(3)
the total strength of the lines of the

multiplet originating in the
is proportional to the statistical
weight (2J+ 1) of the initial
ySLJ
Since the formulas for strengths of lines are

symmetric between initial
we may regard ySLJ as the final state in the
preceding
formulas; in which case they give at once the strengths for a multiplet
L = L + 1 and the sum rule says that the sum of the strengths of the lines
ending in the same final level ySL J is proportional to the statistical weight
of that level.
level.
and
final states,
For the multiplets
in
which
L = Lwe have:
(2b )
From
13 3 5
we learn,
as before, that the
ZS( 7 SLJ y
9
of these three quantities

for the case
sum,
SLJ = (2J+l)L(L+l)\(yL\P\y
)
is
proportional to
2J+ 1. The
L = L + 1 has the value
L)\*,
corresponding
SS(y5iJ /5i+l/0 = (2J+l)(L+l)(2i+3)I(yi:P:/i+l)|2.

J
Thus
for
(4)
sum
(5)
any Eussell-Saunders multiplet we have derived the general

sum of the strengths of the lines having a given initial level is
result that the
proportional
to the statistical
sum of the strengths

statistical
by
weight
(2J+1) of that initial level, and that

a given final level is proportional to
of the lines. having
weight of that final level This
Ornstein, Burger,
sum rule was
discovered empirically
and Dorgelo.*
* BITEGEB and
DORGELO, Zeits.
OBKSTEIN and BUBGEB,
ibid.
fiir PJiys.
23, 258 (1924);
24, 41 (1924).
the
the
29
LINE STRENGTHS IN BUSSELL-SAUNDEBS MULTIPLETS
239
Finally we may sum these formulas over the various values of J associated
initial term. Since the statistical
weight of the whole term is
with the
(25+l)(2L+l), we
is
multiplet
given
find that the total strength of all the lines in the
by
= (2S+l)(2L+l)L(L+l)\(yL:P:y L)*
S(ySL,y 8L)
for the three
initial
and
Mnds
(6)
of change in L. These formulas are symmetric in the

4
They bear a striking resemblance to formulas 7 5
final terms.
which express the total strength of all components of a line.

Now that we have seen how the relative intensities of the lines in a
multiplet follow from the general formulas of Chapter in, let us consider the
application to spectra and something of the historical development in this
field.
The simplest result concerns the relative strengths of the lines Z

S^*P%
and 2 5^- 2
Since there is but one level in the S state, the sum rule tells us
P|.
.
that these should have strengths in the ratio 1:2. This is actually the case for
the yellow D lines of Nal and is quite generally true for the S-P doublets
of the alkalis. We consider in 5 15 the explanation of some departures from
this result.
In the 3 P- 3 5 triplets the sum rule predicts that the strengths of the lines
are as 5: 3 1 and this was found to be experimentally the case in Zn and Cd.
M+1 P
Generally the ratios of the strengths of the three lines
5+1 -, ^^Pg-,
25 * 1
are
as
This
was studied* for
^Ps-r
(28 + 3): (25+1): (25-1).
quintets, sextets and octets in Or I and Mnl, with the results shown in the
:
following table (the calculated relative intensities are obtained

4
4
plying the above strengths by the a factors of 7 ) :
Wave-lengths in
Multiplet
Cr:
.W-y
Relative intensity
5208,5206,5204
(
Mn:
,V-i.a
6021,6016,6013
Mn:
**P-.a
4823,4783,4754
>S
by multi
7:5-04:3-15 Observed
,
8:6-15:4-32 Observed
8-28: 6-35 Calculated
|10:
(10:8:6-15 Observed
Doublets and combinations with 8 terms are the only multiplets in which
the strengths are fully determined by the sum rules alone. Experimentally
it
was found however, that the sum

j
rules are valid quite generally.
The
complete intensity formulas (2) were derived by the aid of the corre
spondence principle before quantum mechanics simultaneously by Kronig,
Sommerfeld and Honl, and Russell, and quantum mechanically first by
*
DOEGELO,
Zeits. fiir
Phys. 22, 170 (1924).
THE
240
ETTSSELL-SAUJSTDEBS CASE: LINE
STRENGTHS
29
Dirae.* Before the complete formulas were obtained, Sommerfeld and

Heisenberg found by the correspondence principle an important qualitative
j"
characteristic of them, namely, that the strongest lines are those in which the
change in
/ is the same as that in
L, These are called the principal lines of
the multiplet. Among the principal lines the strongest is that with the largest
J of the initial level, the strengths decreasing with decreasing J. The other
For multiplets in which L changes the satellites

which J does not change; and
of second order, in which the J change is opposite to the L change. The
satellites of second order are, as a class, weaker than those of first order.
lines are called satellites.
may
be
classified into those of first order, in
Among the satellites the strength is a maximum for intermediate values of J,

These qualitative characteristics of the multiplet intensity formulas are
naturally of great importance in helping to recognize the character of a
multiplet when a spectrum is being analysed.
For practical work
it is
convenient to have tables showing the relative
strengths of multiplet lines in the most important cases. These have been
calculated by Russell (loc. cit.) and by White and Eliason.J The tables of
White and Eliason
are given here in Table I 9 . These are essentially tables

of relative values off(SLJ y SL J ) [cf. (1)] for aU J and J consistent with
L and S. The largest value is in each case set equal to 100 the
others are sharply rounded off. Since in the theory of hyperfine structure
and in jj coupling we need tables of this same function with half-integral
a given L,
L and L such tables are also included.

We have taken care to speak of the strengths of spectral lines rather than
values of the arguments
intensities in order to place on the experimentalist the burden of

inferring
the values of the strengths from his observed intensities. Among the
difficulties that arise is the lack of a statistical equilibrium in the atoms of
the source and a partial reabsorption of the light emitted in one part of the
source by the atoms in another part of the source. The latter difficulty is
||
diminished by use of low-pressure sources, but is hard to get rid of entirely.

Many of the experimental data on intensities are not suitable for comparison
with the theory because data are lacking with which to calculate strengths
from observed
*
intensities.
Phys. 31, 885 (1925);

Sitz, der Preuss. Akad. 1925, p. 141;
RUSSELL, Proc. Nat. Acad. Sci. 11, 314 (1925);
DIRAC, Proc. Roy. Soc. Alll, 302 (1926).
KRONIG,
Zeits. fiir
SOMMERFELD and HONL,
f SOMMERFELD and HEISENBERG, Zeits. fiir Phys, 11, 131 (1922).

t WHITE and ELIASON, Phys. Rev. 44, 753 (1933).
Important references are:

DORGELO, Zeits. fiir Phys. 13, 206 (1923);
HARRISON, J. Opt. Soc. Am. 19, 267 (1929).
On this point see an interesting study of conditions in the neon glow discharge by LADENBTJRG,
Rev. Mod. Phys. 5, 243 (1933); see also SCHTJTZ, Ann. der Phys. 18, 746 (1933).
jj
29
LIXB STRENGTHS 1% ETJSSELL-SAT7NDEES MULTIPLETS
241
TABLE P.
tn
16
242
THE BUSSBI&-SAUNDEBS CASE: LINE STBENGTHS

TABLE
continued.
29
LINE STRENGTHS IN RUSSELL-SAUNDERS MULTIPLETS

TABLE
243
continued.
1.6-3
THE BUSSELL-SAUM3ERS CASE:
244
LKE STRENGTHS
For further comparisons between theory and experiment we
29
may
refer
to the work of Freriehs, of van Milaan, and of Harrison.*
Multiplet strengths in a transition array, f

We now consider the problem of finding the relative strengths of the
iniiltiplets which result from transitions occurring between two configura
3.
Formulas 2 9 6 express the total strength of a multiplet in terms of

2
j(y.L:P:y Z/)J so our problem is reduced to the evaluation of these quan
tions.
tities*
configurations are expressible in the form I-f II and I-f IV

the results of I 8 , in particular 1 8 16, to reduce our
the
relative
problem to that of
strengths of the multiplets in the transitions
from configuration II to configuration IV. If the two configurations are not
If the
two
respectively, we may use
expressible in this form, we cannot proceed by matrix methods (except for

two-electron configurations, ef. 6 8 17) but may use the method of 4 9 .J
The quantity we must evaluate is
\tf&&^SlFL\P\y*S L^S*IPL
l
According to
1 8 18 5
this vanishes unless
^ 1 L l sy 1 ^
!/ 1
(I)
)\*.
i.e.
transitions
are forbidden between terms built on different terms of configuration

If the terms of configuration I are the same, (1) reduces to
1
1
From
(y
S*
S 1 Ity* S 11 L 11 L )\*.
we see that we can evaluate this
Ify&8lF LIP
the considerations of
8
,
11
I.
:?
(2)
expression
11
3
1
u
by means of I1 8, treating F as a vector which commutes with L and S
Hence this vanishes unless S IV = S 11 in which case it is the same function of
L*L IV L, If IP L as \( Y 8LJ\P\y SL J )\* is of SLJ 8L J On com
.
paring
745
with 2 9 6, we then see that the total strengths of the multiplets are
given by
= (25+1) f(&IFL, L 1 IPL ((y^l^PJV1 ^ 11)! 2

)
(3)
where the function /is defined by 2 9 1. If we sum this expression over the
various values of L and L which are consistent with the given values of
L 1 L 11 and i iv we obtain, in analogy with 2 9 6,
,
FBERICHS, Ann. der Phys. 81, 807 (1926);

VAN MILAAN, Zeits. fur Phys. 34, 921 (1925); 38, 427 (1926);
HARBISON, J. Opt. Soc. Am. 17, 389 (1928).
f JOHNSON, Proc. Nat. Acad. Sci. 19, 916 (1933); SHOKTLEY, ibid. 20, 591 (1934). Following
Harrison, we shall call the totality of lines resulting from the transitions between two configura
tions a transition array.
2
2
J For example, these considerations are applicable to the transitions between d $ p and d s\
but not to the transitions between d z s z and d* s.
MULTIPLET STRENGTHS IN A TEANSITIOH AERAY
39
The last S here is the
total strength of the multiplet
iy
245
S 1* LW -^YU &1 IfL
between configuration IV and configuration II; it of course

contains a factor S($ IV S11 ). We shall now discuss these formulas by means
of illustrative examples.
in transitions
Addition of an electron to an ion.

The set of all lines arising in transitions between two polyads ( 2 8 ) having
the same parent configuration is known as a supermultiplet. The relative
strengths of the multiplets in a supermultiplet are given by
1^
1 1
1
y ^ ^ represent the parent term of the ion, y^ L = nl and
(3) if
we
let
y&IP^n
the nl values ofthe jumping electron in the initial and final states, and
&11 = i.
In the first place we see that transitions between polyads based on different
parent terms are forbidden. For an allowed supermultiplet,
S(yiSiI?7rfS,)J#!?ttTLO^
(5)
The
which arises in connection with the

5
intensities in one-electron spectra and which was evaluated in 6 3. Itvanishes unless V = I 1, in which case it has the value
last factor here is just the factor
(nl\P\n l-~l}^$(nl,n l~-l)~

*
^=rrE(nl)R(n
v4P-lJo
The functions /are given by 2 9 2 according to the definition

strengths of
l-l}dr.
2 9 I.
(6)
The relative
multiplets in a supermultiplet are thus the same functions of

asthe relative strengths of the lines in a multiplet are of S L /,
the,
U-IL, IZl L
SL J These relative strengths are thus given by Table
.
I9
with this
correlation.
For example, the
relative strengths of the multiplets in the
Ti I d* 4s (*F) 5s 5 F-*d* 4s (*F) 4$ 5DFG

7
supermultiplet are the same as the relative strengths of the lines in a
7
$~>
transition, namely 50:70:90. Several carefully determined measurements

of relative strengths of multiplets in supermultiplets have been made by
Harrison* with good theoretical agreement. For example, for the above
supermultiplet of titanium he finds the relative strengths 53:70:89. The
maximum discrepancy of 6 per cent, between these and the above values is
probably little more than the experimental error.

This relation for the relative strengths of the multiplets of a super
multiplet was first obtained by Kronigf from the correspondence principle
in 1925. It was apparently independently inferred from the experimental
data by Russell and Meggers J in 1927 when they called attention to the
* HABEISON, J.
Opt. Soc.
Am. 19, 109 (1929);

HABEISOK and ENGWICHT, ibid. 18, 287 (1929).
f KBOOTG, Zeits. fur Phys. 33, 261 (1925).
J RUSSELL and MEGGEES, Sci. Papers Bur. of Standards,
22, 329 (1927).
THE RTJSSELL-SAOTDEBS CASE: LINE STRENGTHS
246
39
striking similarity between the visually estimated relative multiplet in

tensities in a dd-*dp triplet supermultiplet of Sell and the theoretical
relative strengths of the lines in a 5 D-> 5

multiplet.
let us consider the relative total strengths of the various super-
Now
multiplets in a transition array between two configurations formed by

adding an electron to the same parent configuration. The total strength of
a supermultiplet is obtained from (4) if we make the same correlation as we
did in
(3)
to obtain the relative strengths of the multiplets. This immediately

the strengths of the different supermultiplets in a transition array
shows that
are proportional to (2$-f l)(2Ll + 1), where If is the L value of the parent
ion and 8 the multiplicity of the supermultiplet. For example, in the tran
2
of the four supermultiplets
sition airayjp 2
2?, the relative total strengths
<s->3)
are 2: 6: 12: 10, as seen from the following table:

p
2
-.
*
l4r>
2.1
WD
2.3
4^3
4.<7z>n
2.5
Experimental data on relative strengths of supermultiplets are almost non

existent. A comparison of the intensities of a quartet and a sextet super
furnished by Seward* for the case of the quartet and sextet
6 5
6 5
supermultiplets of Mill 3d ( D)4p->3d ( D)5s. The relative strengths of
multiplet
is
D are the same as the relative strengths of the lines in

$,
D should be as strong as these,
7:5:3.
The multiplets J 6D,
namely
4
jF,
D,
P->
10|:7|:4|-.
When
i.e.
P~>
P->
-|
these values are corrected
by the actual
4
<r
factors of
Mn I, the theoretical relative intensities of these six multiplets for excitation

at infinite temperature are given by Seward as
382:354:259:1380:777:754.
The measured intensities, when corrected for the
actual distribution
among
by an assumption of a Boltzmann distribution at the most

favourable temperature, are given by Seward as
initial levels
381 400: 278: 1328: 740: 819.

:
The agreement is
rection
and
quite good, considering the difficulty in temperature cor

intensity measurements for lines of rather different frequencies
originating in levels of different energy.

have thus solved completely the problem of reducing to the radial
integral (6) the strengths in a transition array in which the jumping electron
We
not equivalent to any in the ion. It should be pointed out that this solves
in particular the problem of transitions between two-electron configurations
is
SEWABD,
Pfcys.
Rev. 37, 344
(1931).
MULTIPLET STRENGTHS IN A TRANSITION ARRAY
247
neither of which
is composed of equivalent electrons, since such transition

contain
arrays
only two supermultiplets a singlet and a triplet. For ex
ample, the relative strengths of the multiplets in either superniultiplet for
3
3
P but the
ps-^pp are the same as the relative strengths of the lines in
have
thrice
The
the
of
the
line
triplets
strengths in this
strength
singlets.
$->
9
array are given in Table 2 In transitions between two-electron configura
8
tions, at most one can be composed of equivalent electrons; from 6 17 we
.
see that the
above considerations are true for the unexcluded multiplets of
this case also, except that all absolute strengths
must be multiplied by
2.
Thus, e.g., the unexcluded multiplets of ps-*p* have the same relative
9
strengths as the corresponding multiplets of p$-*ppl see Table 2
.
TABLE 2 9
Relative strength of the lines in
LS coupling.
2
ps-*pp and ps-*p in
Invariant sums are shown in parentheses. The signs of S* are indicated.
PP
Pn
(10)
(10)
(30)
(30)
(30)
(30)
(50)
(50)
(50)
(70)
^n
(20)
120
1(180)
60)
(100)
(20)
The
(20)
(60)
(100)
(100)
sum rule.
now derive a sum
J-file
We
sidering in
rule which applies in the case we have been con

which the jumping electron is equivalent to none in the ion. For
of lines originating in a given level

is proportional to its statistical weight 2 J + 1 .
may by formulas analogous
each multiplet the
sum of the strengths
We
THE RUSSELL-SATODERS CASE: LIKE STRENGTHS
248
to 2
3, 4,
sum
(5)
over
39
to find the total strength, of the lines of the

array
which originate in the term y1 U-^nlS L. This sum has the value
which depends on the

(2S+ l)(2i-f 1). Hence
term only through
initial
its
statistical
weight
the total strength of the lines of the transition
originating in (or terminating on)
array
proportional to the statistical

This statement, which we shall see in Chapter xi
weight 2J+lqf that level.

holds for any type of coupling,
level is
any
we
shall call the J-file
sum
rate .(we call
any row or column of a transition array such as that of Table 2 9 a J file).

We shall show in 411 that even if the jumping electron is equivalent to
other electrons in one of the configurations, this sum rule
J files in one direction in the transition array.
still
holds for
the
More
general configurations.
As an
in
illustration of the use of formulas (3)
which the jumping electron
is
and
(4)
for transition arrays
equivalent to one of the others,
we may
consider the strengths in the case d 2 sp~*d*p*.

By (3) and (4) any group of
2
electrons which remains intact
the
transition, such as d here, may
during
be eliminated, and the problem reduced to transitions between the

electrons, as to
2
sp-^p The allowed
.
Is
(1?)
IP
transitions for
8=1
-+yfi (IS)
(sppp _^ (ap)3p
p PP) *SP-*p z (i) 3 P
8
(
P) 1SPD-+
P)
i^PD
is
convenient to consider together
all
(P-+Z)
3
27
45
It
^P-^S)
45
3
(
remaining
d 2 sp->d 2 p 2 are
3
P->
3
(
(8)
P-^ 3P)
S
P->
multiplets of the
P)
P)
same
multiplicity
which are based on the same terms of d 2 sp, and# 2 and which are
analogous
to supermultiplets. Each arrow in the above
represents one such group. If
we let
S1 represent the term of d! 2 L IV S IV the term of sp, and Lu S11 the
2
term of p (3) shows that the relative strengths of the
multiplets in a group
,
&
are the
where
same
as the relative strengths of the lines in the

multiplet
S = If L = L^ and L = IP.
right of
(8).
The
obtained from
SL->
SL
r
,
These multiplets are indicated at the
relative total strengths of the different

groups are to be
(4) in terms of the known relative strengths of the multiplets
in the transition
MULTIPLEX STRENGTHS IN A TRA^SITICOT ARRAY
39
In
this
S(sp
way,
e.g.,
249
for the last entry of (8), 2S -f 1 = 5, 2U- + 1 = 3, 2S11 +1 = 3,

Substitution in (4) gives 45 for the total strength of the
P-j9 P) = 9.
23
group. Thus the relative strengths of the multiplets in a group are the same
as the relative strengths of the lines in the multiplet in parentheses at the
right of
by
(8),
and the relative total strengths of the different groups are given
the numbers in italics.*
4.
Multiplet strengths obtained from spectroscopic stability, f
We
have obtained very simply the strengths of the multiplets for con
figurations in which the jumping electron is not equivalent to any other
electron in either configuration or is equivalent to one other electron in
have also been able in many cases to
only one of the configurations.
We
reduce the problem to a simpler one. But we have not been able to
handle such transitions as p 2 s
& 2 or p 2 s~*p% to name two of the simplest.
It is possible to determine these
strengths from the principle of spectro
3
->_p
much the same way as we determined the electrostatic

from
the
energies
diagonal-sum rule.
As shown in I 9 the only transitions possible are between configurations
which differ in regard to but one electron, which has quantum numbers n I
in one configuration, and n I 1 in the other. We calculate first, in the zeroscopic stability in
order nlm/rit scheme, the absolute squares of the matrix components of F

connecting these two configurations. There are no components connecting
states
which
electron.
differ in regard to the
By the results of
quantum numbers of more than one
6 6 a non- vanishing matrix
component is simply
the corresponding matrix component of the one-electron problem

is given by the formula
6 s ) This
.
l)
(1)
s is given by 3 6. Thus all the squared matrix

components of P may
be expressed as multiples of s 2
Now let us consider the squared matrix components of connecting
where
8LM
these configurations when expressed in the

S
L scheme. All of these
2
are expressed in terms of the quantities j(y$ L\P\y 8
which we are
Z/)|
3
seeking, by means of the formulas 9 11. The states of the zero-order scheme
are related to those of the
8
L scheme by a unitary transformation
,
SLM M
which has non-vanishing components only between

* if we had substituted
we should have obtained
same
Ma
in (4) the absolute strengths of the multiplets in the array sp~+p z ,

the total absolute strengths of these groups in terms of the radial
integral (6).
states of the
CONDON and UFFOED, Phys. Rev.
44-,
740 (1933).
THE BUSSEIX-SAUNDERS CASE: LINE STBENGTHS
250
and
L Therefore we may apply the principle of spectroscopic stability

2 2 25 to the squared matrix
components occurring in any box of which the
.
rows and columns are labelled by the same values

schemes. This enables us to equate the
sum
oM
8 and
L in the two
of the squared matrix com
ponents in a particular box in one scheme to the corresponding sum of the

squares in the other scheme. In this way a set of equations is obtained
which is usually sufficient to determine the quantities
\(yS L\P\y SL )\*
9
from which we obtain the multiplet strengths by 2 9 6. Analogous to the

limitation of the diagonal-sum rule for
energy, the method gives complete
when
there is but one multiplet of each kind occurring.
specification only
If there are several multiplets of the same kind, the

sum of the strengths.
method determines only
the
We shall illustrate the method by consideration of the configuration array

p*-+p*s |n Russell-Saunders coupling. Only one component occurs with
lfg =Jf^ = f and -^=0,
L =l. In the zero-order scheme this is the
0*), for which the squared matrix component

the
formulas.
In
the
by
S L scheme this is the transition
4
3
P
the
matrix
f 0)-(
f, 1);
squared
component being, by 9 11,
transition (1+0+
of
P is
(*S,
~l+)->(l+0+
SLM
s2
Hence we
infer that
This result
may
ML =M^0.
be checked from the component having
M ~M
s
=%
In the same
M =M
8
way we may pass to the part of the matrix associated with

= J and find in the zero-order scheme that the coefficients of s 2 in
the squares of the non-vanishing matrix elements of
P are given by
The table need not be extended to

negative values of
L unless this
The corresponding portion of the table in the SLM
desired for checking.
is
ML
scheme, in which the numbers are the factors which multiply the corre
sponding quantity of the type l(p**D\Plp*s*P)\* } is
MULTIPLET STRENGTHS FEOM SPECTROSCOPIO STABILITY 251
49
Applying the principle to each of the boxes, we obtain a

which lead to results
set of equations
\(p*SL\P\p*8SL )\*. The total

terms of this quantitj7 by formidas 2 9 6 are
for the coefficient of s 2 in the value of
multiplet strengths, given in
TABLE 3 9
Relative multiplet strengths.
The terms of the first configuration are placed on the left, of the second, at the top, of the boxes.
Only allowed transitions are listed, and where two or more nraltiplets of a kind occur, the sum of
the strengths is given. In the same transition array, values in boxes of different multiplicities are
to be directly compared.
4
>p
s2
or
p$
P
2 4Z)
2
Z>
33
i>
120
45
15
THE EUSSELL-SAUNB BBS CASE: LINE STBENGTHS
252
49
9
given in Table 3 together with the relative strengths of the multiplets in
other simple transition arrays which cannot be solved by the formulas of 3 9 .*
,
The method of this
section has the disadvantage, not shared
by that of
3 9 , of being capable of giving us only the amplitude of the electric-moment

matrix. This is of little use for cases of departures from Russell-Saunders
coupling as discussed in Chapter xr, since to effect a transformation of this

matrix from LS coupling to an intermediate coupling we need to know the
correct phases of the matrix elements in a definite scheme of states. Such a
calculation of the electric-moment matrix
may
be
made
if
we know
the
eigenfunctions of the LS -coupling states in terms of the zero-order states. In
order to separate two like multiplets in cases not amenable to characteriza

tion by parentage we must also go to the eigenfunctions, just as we did in
7 8 to separate the energies of the two 2

s of d 3 Several of the arrays of
9
Table 3 are taken from a paper by Ufford,f who calculated them by first
*
obtaining the eigenfunctions.

5.
Quadrupole multiplets.
Let us now obtain the strengths of the
lines in
a quadrupole multiplet,
did for the dipole-multiplet in 2 9 . In 6 4 we were able to find the

of the quadrupole-moment matrix components by building
dependence on
as
we
them up from the dipole matrix components of 9 3 1 1 here we may repeat the
:
3
by applying II to find the dependence of the matrix components
and 8 in the SLJ scheme.
process
on
After the matrix components are found the strengths of the quadrupole
lines may be found by 7 4 7. This gives us a set of formulas (Table 4 9 ) for
9
quadrupole multiplets analogous to 2 2 for dipole multiplets. In the process
of squaring, the sign of the matrix component is lost, so in Table 4 9 we give
the sign in a separate column. If one wants to know one of the matrix
F occurring in 7 4 7, he can find it from
the strengths by
4
dividing by the appropriate factor in 7 7 and attaching the sign here given
to the positive square root of the quotient. G H, I are related to the matrix
components D, E,
components of r t where 92 =
eSrt-ri?
= - e S J S [(a 8 Ljr,:** S
i)(a"
as follows:
S J&ftja S L)
of
A very complete set of tables of relative multiplet strengths is being prepared by Mr Goldberg,
of the
Harvard College Observatory;
in 1935.
this will
probably be published in the Astrophysical Journal

f UJTFOBD, Phys. Rev. 40, 974 (1932).
253
QUADEUPOLE MTJI/HPLBTS
The strengths of Table 4 were calculated by Rnbinowicz*, who showed

that they obey the sum rule of Ornstein, Burger and Dorgelo, namely, that
9
the
sum
of the strengths of the lines in a multiplet from a given initial (or

proportional to the statistical weight, 2J+1, of that level.
final) level is
TABLE 49
Here P(J), Q(J) and R(J)
9
There
formulas.
is
very
The
Strengths of guadrupole-multiplet lines.
are the factors used in
little
IPS with
the correlation j-^J y j l ~^S,j^L.
experimental material available for testing these

2
5 quadrupole
some of the 2
relative line strengths in

*
RTJBINOWICZ, Zefts. fur Phys. 65, 662 (1930).
D->
THE BUSSELL-SAOTTDERS CASE LINE STRENGTHS
254
doublets in the alkalis have been measured
by Prokofjew* and found to be
3 2 in accordance with theory. The difficulty is that multiplets

violating the
Laporte rule may arise because the atoms are disturbed by electric fields in
:
the source, in which case the line strengths are given by other formulas
(3 17 especially 3 17 5). The place where quadrupole lines are of importance
,
in the spectra of nebulae or rarefied stellar atmospheres where the mutual

disturbance of neighbouring atoms is negligible. In fact the most interesting
is
developments in
Hum
lines as
this field
quadrupole
grew out of Bowen
lines.
This
we
s interpretation of the nebuconsider in 5 11 because these lines
are combinations between terms of different multiplicity and so

existence to the departure from Russell-Saunders coupling.
owe
their
There are two good examples of quadrupole multiplets in the Fell

spectrum found by Merrillf in the spectrum of the star ^-Carinae. They are
5
4s 26 $->3d 6 4s 6.D
and $d*4s*G-+3dF. The intensity data are merely

estimates
from
plates taken in Chile by Moore and Sanford. For the
eye
former there are five lines having the same initial level and ending on the
3rf
five levels of the
D term. By the sum. rule the strengths
1) for the final level. If
to
(2J+
numbers proportional
values are
to
62
.
(2J+
?
we take
1),
by
the quotient should be constant. The
6 6
.
are proportional
Merrill s intensities divided
6 g
.
being due to disturbance by another line. This is surprisingly good

agreement, probably accidental in view of the lack of photometric measure
the
ments.
For the other multiplet the agreement is not nearly
as good. In the table

give for each line the observed relative strengths (intensity divided by
a6 ), the theoretical relative strengths, and the ratio, observed to theoretical.
we
The
latter should be constant in

*
any one row
as these lines
PBOKOFJEW, Zeits. fur Pliys. 57, 387 (1929).

t MJERBTLL, Astrophys. J. 67, 405 (1928).
have a common
5s
QFADRUPOLE MULTIPLETS
255
The sum of the observed strengths from ^Gfij. divided by 12

which would be equal to the sum from 4 (?| divided by 8, which is 2-75,
if the temperature were infinite. The observed ratio, if we assume a MaxwellBoltzmann distribution of the atoms, corresponds, jbo a temperature of
1600 K.
initial level.
is 4-4,
We now
consider the absolute values of the quadrupole strengths. The

methods developed in the preceding two sections could be taken over with
proper generalizations to treat the corresponding problems for quadrupole

Such developments have not been carried out in detail because the
lines.
seems too remote.
possibility of experimental verification
We confine atten
tion to the case of one-electron spectra, which is directly analogous to the

work of 6 5 and 9 s The problem is to express the coefficients, G, 3 and I,
of the formulas for line strengths in terms of the radial integral

>=
which obviously
tribution. This
will
2
R(nl)R(n l )dr,
-efr
Jo
(I)
measure the second-order moment of the charge dis
may be done by calculating, by direct integration, one of the
matrix components (nlms l\ \n

for ms = H- J and m^l we have
8l
in the
nlmsmi scheme. Then,
since
we know that this is also equal to (?i?Z+|?-f|| \nf I I* +%V +%) in the nljm
scheme. The same matrix component can be expressed in terms of
E,
or P from 6 4 6, so by appropriate choices D, U, and F and the coefficient of
4
4
|Q in 6 6a can all be found in terms of s 2 Then by combining 7 7 with the
and I in terms of s 2
table of strengths of this section we can find G,
which is all we need. The work is a little tedious but straightforward. It is
clear that for a one-electron spectrum
rule
would be
violated.
H = 0, otherwise the parity selection
The values of G and I found in
g-
this
way are
4 (21 - l)V(2Z+l)(2Z-3)
Combining these with the table of strengths we have the means of expressing
in absolute measure all the strengths of one-electron doublet lines.
It is convenient, as in 6 5 5, to express the numerical coefficient in the
4
quadrupole transition probability in atomic units. In 7 8, if we measure a
with the Rydberg constant as unit, and S(^l, J3) in atomic units, e2 a 4 we
,
may
write
THE BTTSSELL-SAHNDERS CASE
256
LESTE
STRENGTHS
59
The numerical coefficient is about one-millionth, of that for dipole radiation

5
(6 o). To exemplify the formulas just presented, and also to gain a better
appreciation of the order of magnitude of quadrupole transition pro
babilities, we shall calculate A(A,) for the 3^->Is doublet in hydrogen.
First
we
find that the radial integral

2
Jo
hence I 2 = 3 7 /2 19 from which, using the table of strengths,

,
S(3c%->
l*j)
= 3.3 8 /2 13
S(3dj->
= 2.3 8 /2 13
l*j)
These are in the ratio of the statistical weights as the sum rule demands.
this line we have cr = f so the spontaneous emission transition pro
For
bability of either line
From Table
is
5 5 the closely analogous dipole transition probability

,
is
A(3p-*la)= 1-64 x 10 secwhich is about 0-27 x 10 6 times as great.

2
In hydrogenic ions of nuclear charge Z we have
and r oc-Zr 1 so the
dipole transition probability varies as Z* while the quadrapole transition
Q
In all actual atoms the quadrupole transition
probability varies as Z
than the dipole, although for 2=92 the ratio
remains
smaller
probability
,
<rccZ
As
a consequence of this trend quadrupole transitions are

sometimes of appreciable intensity in X-ray spectra. For the higher values
is
only
of
of
32.
the calculation should be
made
using the relativistic radial functions
5 5 but our result shows the trend in order of magnitude.

In view of the great ratio of dipole to quadrupole intensity it
,
is
out of the
question to observe quadrupole lines in the hydrogen spectrum. In the

alkalis, where the terms are well separated, this is possible and the s -*d series
obtained in absorption. Stevenson* has calculated the transition

doublet for the alkali
probabilities from the normal 8 level to the lowest
is easily
metals using Hartree wave functions. Experimental values were obtained

by the anomalous dispersion method by Prokofjew.f His values have to be
divided by four because of a difference in the theoretical formulas for quad
6
rupole and dipole dispersion 4 The final values of 10 times the ratio of the
spontaneous transition probability for the

that for the
first
P->S
first
D~*S
quadrupole
Na
Rb
Cs
Observed
1-1
1-5
2-7
0-6
Calculated
3-5
2-5
2-9
line to
dipoie line are
STEVENSON, Proc. Roy. Soc. A128, 591 (1930).

f PROKOFJEW, Zeits. fur Phys. 57, 387 (1929).
j This was pointed out by KUBINOWICZ and BIATOST, Ergebnisse der exakten Naturwissenschaften, 11, 216 (1932).
CHAPTER X
jj COUPLING
The most
direct
method of solution of the
first-order perturbation
problem
would
consist of the diagonalization of the matrix of electrostatic plus

spin-orbit energy for a given configuration in the zero-order scheme. That
procedure one might call impossible for all but the very simplest con
figurations because of the high order of the resulting secular equations. The
general solution
SLJM
is
possible
for a great
scheme, but in order to
many more
utilize this
cases if one uses the

scheme one must, except in
Because of the complexity of
special cases, obtain the transformation to it.

the problem and the impossibility of obtaining the general solution for
complicated configurations, it is desirable to obtain as much information
as possible of an elementary, although approximate, character. In the pre
ceding three chapters we have considered the important case in which the
spin-orbit interaction is weak compared to the electrostatic in this chapter it
;
be interesting to consider the less important case in which the electro

static interaction is weak compared to the spin-orbit. In this way we shall
know the character of the general solution at both extremes.
will
1.
The ^"-coupling scheme and

1
the spin-orbit interaction.
We
wish to determine the first-order energies which result from the spinorbit interaction, when the electrostatic interaction is absent. Since the
i=
HI=
f (r^L^S,
(I)
^H
i
a quantity of type F, it is diagonal in the zero-order scheme in which

riljm are taken as the electronic quantum numbers. This, then, is the
is
natural scheme to use
when
element for the state A = a1
. . .
considering this interaction.
aN
is
given
The diagonal
by
(A \H*\A) = S (n*Fj*m*\H*\n? I*? mf)
45 ), which depends only on the values of 71*1* j*.

1 is so
highly degenerate as to be diagonal in any scheme
Although
i i i
l
n
the
set
j it is most convenient to consider the particular
specified by
(cf.
>
scheme in which the fs have been combined to form resultant J and M.

The scheme characterized by the set of nH^ and J is known as the jj-
be here noted that although this designation

utilizes more quantum numbers than the i~coupling scheme, it still does
not in general furnish a complete set for more than two electrons.
coupling scheme. It
CS
may
17
258
fj
COUPLING
The process of finding the allowed levels iajj coupling is similar to that in
6
coupling, depending in the same way on 4 1. For example, consider
the configuration npn p. Here j and / may take on the values J and f
If ; = i / = i (m,m ) may have the values (J,J), (|, -J), (-$,J), and
(
i). For the first of these states M=l, and since this is the highest
=
occurring for j | j = |, there must be a level J= 1. There are two states
with Jfef = 0; the level J = 1 has a state Jf =0 and there must in addition
= 1 is taken care of by
be a level with J=0. The one state with
J"=I.
Hence we infer the existence of the levels wpin pi, J=l, 0. In a
LS
1>
similar
way
the levels for the other values of j,
following table. In the body of the table are given

of each column are given the resulting J values.
are determined in the
(ra,
),
while at the foot
(3)
J values check with those obtained in

2
1
3
3
^gj A,2,3- They
namely SQ Si, P1? P
These
,i,2>
I7
for this configuration,
could equally well have
been obtained by the usual vector-coupling picture: j = f and j = f can

combine to give a resultant J = 3, 2, 1, 0, etc. Just as in L$ coupling, the
resultant levels cannot be obtained in this
electrons.
For the configuration np\
(3)
way
if
there are equivalent
becomes modified as follows:
(4)
THE jy-COtTPUNG SCHEME ASTD SPIN-ORBIT INTERACTION 259

These J values agree with those previously found, namely 1 SQ 3 P ,i,2? *A2I 10
Because of the symmetry which obtains, the half of these tables for
will usually give one all the information desired.
The
lf>0
spin-orbit energies for these states as calculated from (2) are
(16)
it,*- k-p
(8)
(5)
(8)
J = !,/ =
a
:
(6)1
-iU
(8)}
-2C W
(1)1
(6)
(The numbers in parentheses at the right denote the degeneracies of the

energy levels.) Thus we see that p* is split by the spin-orbit interaction into
three equally spaced energy levels, while pp is split into four
energy levels
whose spacing depends on the relative values of
and
n
np
p In each case
the centre of gravity of the configuration remains at zero. This is a general
property of the spin-orbit interaction, which follows from the theorem
7
7
given in 3 about the centre of gravity of a term. In 3 we calculated the
1
elements
in
of
the
LSJM
scheme
and
found that their sum for
diagonal
each term, and hence for the whole configuration, was zero; by the diagonalsum rule the sum of the eigenvalues of 1 must have this same value.
.
2.
The addition
of a
weak
electrostatic interaction.*
We shall now determine the effect of the addition of an electrostatic inter

action which is sufficiently weak to be considered as a perturbation upon the
must then, in 1 10 3 and 1 10 4, calculate that
degenerate spin-orbit levels.
We
part of the matrix of electrostatic energy which refers to each column, and
determine its eigenvalues. But the electrostatic energy is diagonal with
respect to
J and independent of M
for each J, with
so that there will occur one eigenvalue

1 Hence unless a given J occurs more
a degeneracy of 2 J +
than once per column, these eigenvalues can be determined by the diagonalsum rule from the diagonal matrix elements.
We therefore need the diagonal elements of the electrostatic energy

)-e/r<,
i>j
LSTGLIS,
(1)
i>3
Phys. Rev. 38, 862 (1931).

17-2
260
jj
in the nljm scheme.
From
COUPUKO
2 10
7 S 7 this element becomes
(A\Q\A)
^
= S r/fe*Z^*)&^^^^
l, 2)
= S
T(n*V^*,n*B?nf).
(2)
&>*-!
Each of the ^ s of (2) must now be expressed by 45 8 in terms of the ^(nlm/n^

in which case we obtain a sum of terms of the type 8 6 1 which have already
We
with giving Table I 10 as

calculated by Inglis, of the quantities T(riljm n l j
in terms of the
)
Fk (ril,nT) and Gk (ril,n l of 8 6 15, for s, p, and d electrons.
been evaluated.
shall content ourselves
For example, for the configuration nprip the

found from 1 10 3 and Table I 10 to be
npn pi
;=|,
=f;J=3: F +
-F2
-<?
electrostatic energies are
- Gz
0:
2:
- 0,
(3)
0:
In the approximation we are considering FQ need not be small since it occurs

uniformly in all diagonal elements and all energies. It serves merely to
raise the whole configuration and may be considered as a constant inde
2
pendent of coupling. In the case of np we have
0:
In Fig. I 10 these levels are compared schematically with those of

Chapter vn for this configuration in LS coupling. The / values occur in
the same order on each side, and as the parameters are continuously varied
we shall see
order.
in
3 11 that the
That two levels
two sides of Fig. I 10 grow into each other in this

of the same J value cannot in general cross when one
TABLE
ll
>.
T(nljm,
n lj
).
Che numbers given are to be used as coefficients for the

integrals at the top of the column. To eack
of these T s is to be added F (nl, nT), which is omitted from the table for convenience.
262
ii
parameter
21
COUPLING
shown by Ton Neumann and Wigner.* The

limits (weighted by 2J-fl)
energies In the two
varied has been
is
sums of the
electrostatic
are equal as required
,Q
by the diagonal-sum
rule.
- o
-&
-5-
*
(A)
(D)
(B)
(O
2
10
Fig, i . Limiting eases for j? ; (A) No spin -orbit interactioxx. (B) Weak spin-orbit
interaction. (C) Xo electrostatic interaction. (D) Weak electrostatic interaction.
3.
Eigenfonctions.
The energy
which result from the spin-orbit interaction are

characterized for a given configuration by a set of j values, and are de
If a weak electrostatic interaction is
generate with respect to J and
levels
applied, these energy levels are split into levels characterized by J values.
Let us for convenience call the set of states characterized by a given set
of J values a (jj-coupling) term/ in analogy with the terms in LS coupling.
We shaE use
a notation in which we represent the three possible terms of

and 2p|. The subscripts denote
the configuration 2p 3 by
2p|2p|, 2p|2yj,
the j values of the electrons, the superscripts the number of electrons of the
same nlj values.

The exclusion
fashion than in
principle operates in jj coupling in a slightly different

coupling. All terms for which not more than 2J+1
LS
electrons have the
same nlj value are allowed. However, not all the levels of
a given term as given by coupling thej
s vectorially are allowed unless all the
nlfs are different. The eigenfimctions for Jj coupling may be found by any
of the methods used in Chapter vin, but we can in general obtain more of
them by coupling vectors than we could there. For example if, as in I 8 we
,
couple the terms of two configurations to get terms of a

*
VOF XEUMAI^ and WIQXEB,
new
Phys. Zeits. 30, 467 (1929).
configuration,
3 10
ECGB^FUNOTIONS
263
Js
obtained by coupling the allowed J1 ^ and J^ s will be allowed

and the eigenfunctions may be found as in 6 8 provided that no nlj Yalue
all
the
occurs in both I and II. This
allowed
is less
stringent than the L5-coupliBg require
value occurs in both configurations. Thus we may find the

values for any term if we know the allowed / values for terms of
ment that no
ril
the type nfj These are given in Table 2 10 Thus

are
levels of
.
np\np^
w
.p|(0)ttpj(i), I
we see that the three allowed
and wp|
(2)^(4), f and
f.
Here in parentheses are given the J values of the groups of nZJ-equivalent

electrons, and at the end the resultant J values.
TABLE 2 W . Allowed J
values /or groups of nlj-equivalent electrons.
In coupling two equivalent electrons, all J values are allowed if j1 ^^2

8
i=j* the states resulting from the coupling are as in 6 either anti
symmetric or symmetric according to whether J is even or odd. Hence only
the even J values are allowed ifj=f. The use of vector-coupling formulas
2
for obtaining matrix elements between two states of the configuration i is
12
Let us here consider, for use in the nest section
considered in detail in 6
.
on
line strengths, the
matrix components
(nljnW JM\I\nlj*n
of J = .Fj -f 1*2
1
lT J JT) = (jj \F\ff)

2
connecting the states of nl and

a consideration similar to that which gave 6 8 17 that
nlriV. If j= /,
(1)
we
see
from
is between states in which the

and is hence amenable
numbers,
quantum
to the use of the formulas of Chapter m. If j /, we see from the relations
where the matrix component at the right
individual electrons hare definite
T(Z?nZf JJf) = 2-^KZ^ 7i2y; JM)

= HtflnAJj n J J
)
Y(nZf ril J
f M
that
j
f) = i
JM
^W^M
JM
)l
j(nj
n&ft
)]
264
In tlie last term here, the order of the quantum numbers
must be reversed by the use of the relation 143 7:
Mh J
= - l)^-J
(
In
way we
this
^<
IFiff) = StfJ Ch/s!* ,!^-?)

)
- so",/) - iX-Ô i tt*

(
3)
!)
0V/)
4.
and
find
1
(jj
is significant,
(*)
Line strengths.*
Line strengtlis in jj coupling may be found by methods similar to those of

Chapter ix for L8 coupling. Their main application is in the rare-gas spectra
are
(Chapter xm) where only the strengths for two-electron configurations

needed. We shall first briefly consider transitions between the two twoelectron configurations n^P-nl and riWnT. If the jumping electron has
different nlj from the other electron in both configurations, we may use the
3
formulas
11
88
to evaluate the quantities
Kn^PfnljJlPln^nTj J
which are needed in
this
way we
7 4 5 to give
the strengths of the corresponding
obtain, as in 2*1,
s(fj*W J
(1)
)|
jW H K^XZ
lines.
In
(2)
=/(j
f)!
Hence the relative strengths of the lines in the *multiplet* connecting the
terms f-j and j1/ are the same as those in L8 coupling for the multiplet
SL~*SL f with the correlation 5-^j 1 L^j, L -*j except that j is always,
L never, half-integral. They are thus given by Table I 9 for half-integral L.
}
The total strengths of the multiplets are given as in
S^V) = (2f+ !)/(&
8l j
395
by
(3)
\(^:P:nT)\^
9
the intensities to the single matrix element 3 6. If one of

the configurations is composed of equivalent electrons, say that wi^n1?,
which reduces
we see from
all
3 1@ 2 that the allowed lines
when j =j 1 have twice the strength
of the corresponding lines for non-equivalent electrons, and from 3 104 that
when j^j1 either
J ) or Sfjj 1 JJ1// ), whichever occurs, has
S^jJ^j
S^jJ^fJ
the same intensity as
calculations are illustrated

lines in
ps-^pp
2
andp#->p
for non-equivalent electrons. These
10
by Table 3 of the
relative strengths of the
in jj coupling.
1
1
generally, formula (2) is applicable, if we write J for j , to
transitions between the levels resulting from the addition of electrons of
More
quantum numbers
momentum J1
nlj or
n lj
if
nlj
and n
Vj* to
any ion
in a state of resultant angular
none of the electrons in the ion has quantum numbers
BABTLETT, Phy. Rev. 35 f 229 (1930).
LESTE
TABOJ
3 10
265
STRENGTHS
Relative strengths of the lines in
-*jp
p and p s~*pz in jj coupling.
Invariant sums are shown in parentheses. The signs of S
are indicated.
PP
i,i
1,1
i,i
if
I,!
i,i
f,i
I,!
(30)
.0
4.4
(90)
I,*
(90)
(150)
(10)
(10)
(30)
(30)
(30)
(50)
(30)
I.*
(30)
(50)
(70)
4,4
(20)
kl80)
ps
100
(100)
(20)
(20)
(60)
(100) (100)
CHAPTER XI
We
have calculated tie
first-order
energy levels and line strengths in two

Hmit in which the
limiting cases: the Russell-Saunders, or JtS-coupIing

electrostatic energy predominates,
and the jj-coupling Emit,
in which the
spin-orbit energy predominates. Practically, the actual levels of all atoms

lie between these limits, although many are very close to Russell-Saunders
coupling, and a feWj particularly heavy atoms

closed shelly are close to jj-coupling.
In
this chapter
in the general
and those containing almost
we shall obtain the first-order energies and

in
line strengths
which the two interactions may be of any relative order
of magnitude by calculating the complete energy matrix for states in the

SiJJl scheme. In this scheme the electrostatic energy is completely
known from Chapter vn) and the

The total energy
spin-orbit interaction is diagonal with respect to J and
diagonal {the diagonal elements being
independent of Jf, hence the least value of JJ/J which occurs for the con
figuration is usually the most convenient to consider, since every level will
is
In Chapter xn we shall see how to calculate the same

energy matrix in the jj-coupliiig scheme which is particularly suitable for
the considerations of the rare-gas spectra in Chapter xm.
have such a
state.
These calculations
which
1.
will
all
neglect tie interactions between configurations,
be considered in Chapter xv.
Matrix ol spin-orbit interaction for configurations consisting
of
coupled groups.*
Our
first
task
is
to obtain the complete matrix of the spin-orbit inter
action
A~
(1)
for the
SLJM scheme of a configuration.
In 3 8 we saw how we could obtain
the diagonal elements of this matrix for configuration I -f II in terms of the

diagonal elements for the inequivalent configurations I and II separately.
Such a calculation
we make
is
however not
restricted to the diagonal elements. If
the correlation
i
Jr*S
J-+J
V
the dependence on
(and If) of the matrix elements

(3)
*
JOHNSOST, Phys. Rev. 38, 1628 (1931).
I 11
is
SPIN-OBBIT INTEBAOHON MATRIX FOE COUPLED GROUPS 267
given
by
12 3 2 in terms of the quantities
S S (yl&IfyZSZIflLigfa) L
"
yiyn
1
"lS
L I y ^S-L L
f
J1
(yi&Li y^^nz/n S^.yWjyi y^Sf njyn s

As
in 3 8 1, this
).
(4)
be broken into two sums, over group I and II inde

3
pendently. Formulas 1I 8 may then, with the correlation 3 8 3, be used to
reduce the first sum to a known multiple of
may
S S tflfl&t) L&iLi)(<yi&lSfriSi).
yi
(5)
The second sum may be correspondingly reduced. Since the matrix elements
of spin-orbit interaction for configuration I are known
multiples of (5), this
effectively expresses the matrix elements for configuration I -f II in terms of
those for I and II separately.*
Two-electron configurations,
These considerations completely solve the problem of finding the matrix

of spin-orbit interaction for a two-electron configuration, since the matrix
for a one-electron configuration is known from 4 5 6:
A comparison of this with
12 3 2 shows that
(nK(r) LM)(ns:S:ns) = tnl %\

(6)
which gives the value of the quantity (5) for a single electron. Thus for a twoelectron configuration the element
S (nWn*P SLJlftffa) LfSilnWvW S L JM)

=g(8LJ, S
where g(8LJ, S
correlation
(2),
L J) S ln if (Ul*\L
)(Si&\S
),
(1)
L J) is the coefficient of (4) in (3) as given by 123 2 with the

and
2L(L+1)
V.IXL-1)}
/(T
iixL-nr***!1
S
(S\S \S) = (SIS \S) = (8:8^-18-1)= -(SiS :S-l) =
2
* For
a given
configuration, say I, to -write that

1
V- J1 JyP|rf ) LfSt \y* S*1 i
(y
&
equals the factor given by 12
S2
J1 if1 )
(8)
(3
1
)
times
E S (/ & &\&rd L^/ & L
^y"
& Li\Sg/i S i
*
/J
(41)
f*i I
nas in general a sense only symbolic, since no state characterized by S* L 1 may be allowed,
1
although (3 ) has a value different from zero. The above procedure is justified by the fact, which
1
follows from I 8 , that the matrix component (3) is a linear combination of the components (3 )
and (3 13 ) which depends only on S1 L1 S11 17*- and S 1 L 1 S fU JD n and not on the structure of
the inequivalent configurations I and II.
INTEBMEDIATE COUPLING
268
TABLE l u . Natrk-m
of spin-orbit interadim.
-i
V5
20
5
8jD
-at
-VlOa
-V 3a
-VlOa -V6^
2V2a
4^
-2a
-S
>.
m
o
V3
V2
o
2
V3 V2
0020
o
-5
-V
.^2
-4
Pi
V5 -V 10
-2
V5
-Vio
-2
2V2
V2
SPIN-OBBIT INTERACTION MATBIX FOE COUPLED GROUPS

TABLE
11
continued.
SV^
2V5/3
2oc
Vsjs
2a
-V7a
3a
-6a
A
-0.
-Voa
4V2/S
-4V5a
V5
**
-2a
*J.
1
j,
V,
*F>
-2V
21 8
J
-6oc
-2V21jS
269
I 11
INTEBMBDIATE COUPLING
270
are the factors given by 10*2 with jS 1 ^ | 5 2 = f 3 $== I. It follows from 6 8 15

that these formulas hold for the allowed elements even if the two electrons
5
are equivalent.
In
this
way the
matrices of spin-orbit interaction for the two-electron

I 11 may be readily calculated. In these matrices the
configurations in Table
l l
phases of the eigenstates in the configuration n l n*P are such that
2
2
1
=
L
-f
5= S -f S*,
S-f L, with the vectors added in this order by
the formulas of I -4 s We shall always use this system of phases for two-electron
J=
configurations.
of n^PnH1 are
According to this convention, the phases of a given term

2
{i)21^ -^-^! times those of the corresponding terms of
More complex configurations may be calculated in a similar fashion.

Thus in Table I 11 is included the configuration p*s. This matrix was obtained
in the way we have sketched from that given for 3 which is calculated in
the next section. In this matrix the phases of the eigenstates are such that an
3
z
8
eigenstate of p s is obtained from those of p with phases given by 4 6j by
S
s
l^+L =l, S-rL=Jin this order. We shall always
adding S^-f S
use such a system of phases when adding an electron to an ion or in general
in coupling two groups.
j>
=S>
2.
Matrix of spin-orbit interaction obtained from the eigenfunctions,
For configurations containing more than two equivalent electrons we

cannot obtain completely the matrix of spin-orbit interaction by the method
of the preceding section. But
if
we know
SLJM
the eigenstates for the

in the zero-
scheme (Chapter vm) and the matrix of spin-orbit interaction

order scheme, this matrix
in the usual way.
scheme
may readily be transformed to the SLJM scheme
The matrix of the
spin-orbit interaction in the zero-order
66
s
(cf. 4 1 for notation and sign convention). One
obtains a non-diagonal element only between states
differing in regard to one
individual set, say that a occurs in A while a! occurs in A The value of this
is
obtained from
component
is
(.4
H*\A )=
H-*t>w(*\L-S\a
).
(la)
The diagonal element has the value
where a runs over aE the
sets (outside of closed shells) of A.
From these formulas we may easily calculate the matrix of Hl in the zeroorder scheme, where it
the
SLJM scheme
pendent of If.
it
is diagonal with
respect to if. When transformed to
becomes diagonal also with respect to J and inde
MATRIX OF SPIN-ORBIT INTERACTION
911
Illustration:
in 4
6m
for
271
The configuration p*. The SLJM states for j? 3 were obtained

In the zero-order scheme 48 6a the matrix of 1 has the
M = J.
value
CDS
A
A
B
C
D
E
When
this is transformed to the
matrix given in Table

3.
ln
SLJM
scheme by 48 6m, we obtain the
Illustrations of the transition
from LS
to jj coupling.
if*
7
By adding the electrostatic energies of 5 to the spin-orbit matrices of
Table 1 U and solving the resulting secular equations, we find for the energy
5
levels of si in
Here
any coupling the
>.
The complete
values:
transition
from
LS
coupling
(fl 1
>2)
FQ
11 for
Fig, I
to jp
an
the configuration sp.
in
coupling (
ffj is plotted
additive constant which does not influence the intervals between the levels
>
and hence does not need here
to be considered further.
Apart from
is
.F
we
see that the ratio e/6^ of the energy value to G^ is a function only of x = f/ Gl .
For = the energies in these units are + 1 and 1, while for QÔ, they
3
Hence we cannot show

*/#! approaches infinity.
fx- As
the whole transition by making a plot of JGl against x- In order to keep
are f x
and
x~>oQ,
the ordinates from going to infinity
we plot instead -===. as
ordinate,
and
in order to confine the abscissas to a finite range, we plot this against

11 In this
way we show the true interval ratios for all x
^j(1 4. x) in Fig. I
from to oo. At the left end, e/G^ is effectively plotted against }/<?! while
.
HOUSTON, Phys. Eev. 33, 297 (1929);

CONDON and SHOBTLEY, Phys. Eev. 35, 1342
(1950).
272
311
at the right end /f is effectively plotted against 1 - (<?!/)

chosen to make the total splitting the same at the two ends.
-5
11
Fig. I
./
-I
-2
-7
-8 -3
I -S -S
-S
-5
-4
-2
-3
The configuration sp in intermediate coupling.
-3
-2
-6
T^ factor | is
-4
-3
-5
-6
-4
-7
-3
-3
-7
-6
-5
-6
-5
-#
-7
-2
-3
&
11
=
Fig. 2 . Interval ratios for sp. (x K/#i *)
In order to see how well the observed si configurations fit these

11 and
3 11 for sp and sd respectively
formulas, we have drawn in Figs. 2
curves which show the theoretical value of ( 3 L|
as ordinate against
(î-^-i)/^-
- 3 L|_1 )/( 3 il4 1

.
LJ) plotted
^-!) as abscissa. These curves
3 11
ILLTTSTBATIOXS OF TRANSITION
are in fact straight lines: if

abscissa by a,
FROM LS TO
we represent
0=
jj
273
COIJTIX&G
the ordinate
by o and the
The plotted points
represent these ratios for the observed si configurations.

It is seen that in genera! they follow the theoretical formulas rather well.
While there is no essential requirement that Gl be positive, and there are
many sd configurations in which the singlet is
"Q
./
-2
-3
-5
-4
below the triplet as
-7
-6
U
Fig. 3 . Interval ratios for ad.
(x =
K/<?,
if
were
-8
.)
negative, the interval ratios for these cases do not obey the above formulas
at all accurately; they are probably all to be explained as in I 15 by a strong
perturbation of the singlet by another configuration. No plot is made for sf

configurations, since there are just a few known, all close to the Russeli-
Saunders limit.*
* Formulas which make an
approximate allowance for the terms in the Hamiltoiiian
which express the interaction of the spin of one electron with the orbit of the other have
calculated for * I by use of group-theoretical methods by WOLFE [Phys. Rev, 41, 443
(1032)]. We may derive Wolfe s formulas very simply in the following way. For a two-electron
7
configuration this interaction is assumed to have the form. (of. 7 I)
"been
(1)
The second
factor here has no elements connecting different configurations.
L-S-ELt
Si,
By
writing
(2)
- JL2 - S2 }] minus that of the spin-orbit interaction

we see that its matrix is that of L*S [= J
with ^(r ) and hence & set equal to unity: both of these are known in LS coupling. This
matrix must be multiplied by that of w(rl9 rx ) for the configuration under consideration. If in
we replace the mutual distance r lt by r we have a function p(rlf rt ) merely of the
*K**i*
magnitudes of the radii. The matrix of this for a configuration in which the I values of the two
2
(J"
*"t)
cs
>t
18
274
ESTERMEDIATE COUPLING
i?*
V, -J-.-W.-tf
1^
SP 7.
1
"!
XGJ
- J-^z - It
V aai ps2 _ jus *2b
=A
zrj- j. rs
^0 ^.F*
f S 44b
s>
-C
4.
>j
If /JF 2 is expanded as a power series in /jP 2 keeping Just the first power,
these results agree with those obtained in Chapter vn for LS coupling. If
s
Is
j
expanded in terms of -?V, the
results agree
with those of Chapter x
for jj coupling. See Fig. I 10 The complete transition is plotted in Fig. 4 11 as

a function of x =
e superpose on this figure the observed levels of
Ge 1 4p 2 3 Sn 1 5j? 2 and Pb 1 6p 2 placed at such values of x that 3 X the mean
.
K ^V
"^
of
IK and SP!,,
and the mean of
ând
Pfit the theory
exactly. This
corresponds to the parameter values

Ge
Sn I
F2
1016-9
918-0
880-1
2097-3
0-173
X
elections are different
constant value
Hence
in
any scheme
is
matrix
this
(2).
Hs
with
t ,
We
TJ
written for
1-583
"
TI
0,J,
7294
0457
diagonal in the zero-order scheme, the diagonal elements having the
times the matrix of

Hcssell-Saunders ease.
For si, the matrix of
- 1, -I - 1, for lL 3
just
Pb I
921-5
To
v
^ times the unit matrix, and the whole matrix of JEf- is
Lande interval rule still holds in the
is
this approximation, the
times the matrix of L

(which is diagonal with elements
TJ
3
lj ^ _i) minus the matrix of spin-orbit interaction of Table I 11
we include this matrix, we obtain the energies
is
s
Tl
When
cannot compare these formulas satisfactorily with experiment because we have now as many
parameters as energy levels.
These considerations can readily be extended to other configurations.
* The secular
equations for this configuration were first calculated by GorBSMJT, Phvs. Rev.
3
35, 1325 (1930), and those for p later by INGLIS, Phys. Rev. 38, 862 (1931), by the following pro
cedure: Since all the matrix elements of the Hamiitonian are linear functions of the JP s, G\
and f s, the secular equation for a J value occurring n times will be homogeneous of ?i til degree in
the J^s, GTs, t?s, and , the energy variable. For small J*s, the roots of these equations are linear
functions of the F*s, G^s, and f s, known from the electrostatic energies and Lande splittings of
Chapter vn. IFor smaE electrostatic interaction the roots are linear functions of these same para
meters known from the considerations of Chapter x. Knowing these roots for limiting values of
the parameters serves in simple cases to determine all the coefficients in the secular equations.
Since this procedure does not give the energy matrix, it is not possible in this way to determine
the eigenstates in intermediate coupling in terms of those for pure coupling for use in Zeeman effect
or intensity calculations. In simple cases, the weak-field Zeeman effect is given by addition of
terms in Jf* to the secular equations.
3n
ILLUSTRATIONS OF TKANSITIO^ FEOM LS TO
jj
COUPUKG
It is seen that the coupling gets progressivelj closer to

jj as the
heavier. The separation between the two levels of J = 2 and the
between the two levels of

are seen to agree
One
%FQ
The
j?
is
gets
separation
the observed data.
configuration ^p
eigenlevel of
atom
= 0, which are absolutely predicted by the theory,
J"
weE with
X.
11
Fig. 4
275
Pi
I/K
in intermediate coupling. (^ = J/JFa .)
with energy 3jF
one
is
6jF 23 the other three levels are linear combinations of
with energy
P f 2D
$|
with energies given by the roots of the secular equation

3
4-
2LF2
-f
(90 J|
- |C 2 - ^jP2 C 2 = 0.
)
These energies are plotted in Fig. 5 11 as a function of x = /jF 2 together

3
which are seen
with the observed levels for As I 4p 3 Sb I 5p 3 and Bi I
to depart- progressively from Russell-Saunders coupling. The values of x
for these configurations were determined by making all the levels fit as
,
well as possible.
The parameter values
6j>
are
The configuration jp 4 is similarly discussed in Chapter sm.

1
8-2
ZN TEEMEBIATE
276
311
For more complex configurations the formulas become more complicated

and the number of parameters necessary to determine the intervals greater
than two, so that it is no longer possible to plot the complete transition in
two dimensions. Xo comparisons with experiment have been made for such
configurations. The secular equations are readily obtained from the electro
11
static energies of Chapter vn and the spin-orbit matrices of Table I .*
-3
-4
-5
-6
-8
-7
x
11
Fig. 5 .
The configuration p
-3
-8
-7
-6
*S
-4
-3
i/x
in intermediate coupling*
(%= JJ/â
At this point we can merely call attention to a recent paper by Bacher

and Goudsmit f which makes a good start on the problem of calculating
the absolute energy of a state of an atom from the experimentally known
5
energy states of
its ions.
It
is
unfortunately too late for us to treat this
important paper in detail.

*
VAX VLECK
[Phys. Rev. 45, 412 (1934)] lias shown that in the addition of an * electron to
tie atom may be easily calculated from those of the ion by
vector mode], and that in tJiis case the spin
of the parent term is always an exact
quantum number. By the procedure of l u , the spin-orbit interaction for such an atom is easily
obtained in terms of that of the parent ion. MEBBELL
[Phys. Rev. 46, 487 (1934)], combining
these calculations, has made an
interesting comparison with experiment of the energy levels
based on certain definite terms of the d k ion in a number of
configurations of the type d*n*
*
an
ion, the electrostatic energies for
the
Dime
f BACKER and GOTIDSMIT, Phys, Rev. 48, 948 (1934).
LIKE STRENGTHS
4 11
ENT
ESTER-MEDIATE COUPLING
277
Line strengths in intermediate coupling.

The strength of a line from a level a J to a level b J is given in any coupling
4
by formula 7 5 in terms of the quantity (aJ\P\bJ }. If the expansion of the
state aJ is known in terms of a set of states T(aJ) and the
expansion of b J
is known in terms of the set Y( J ), we need
to
know
the values of
only
4.
(aJ:P:/8J
(
P:
to obtain the strength of the line in question. Since the factors

like the components of a matrix of an observable ( 9 3 1 2)
transform
-/
(1)
).
The matrix components on the right of this equation must be taken between
states having the same phases as those for which the transformation co
efficients are calculated. Hence we cannot
immediately replace these matrix
components by the square root of the
line strengths as given by Chapter IX.

However it is convenient to replace these components by quantities which
we shall write as S*(oc J, /?*/ ), and which are understood to be the square root
of the line strength taken
lpJ
).
or
-f
according
to the sign
of the matrix component
Hence we write
,
bJ
= S (aJ\*J) S*(oc J, @J
)(fiJ
\bJ
(2)
).
a/?
In
this
way we can
magnitude of the S
use directly the results obtained in Chapter ix for the

and need only learn to choose the proper phases.
s?
Phases in the matrix
S*.
3 9 , the transitions from configuration I -f II to

and ask the phase of
Let us consider, as in
configuration I -f IV
(
This element
W! yIVIVIV SLJ:P:^SÛ
is
expressed
11 8 with
by
>y^$L
SLJ
).
S+L=J in terms of
in turn expressed by 1 1 3 8 with JLx -f IV or Ll -f JLn = L in terms of

(yivjîV|p|ynjrjij jj^ eac |1 case jfaQ proper sign of the coefficient is given by
which
is
3
1 1 8.
In the same way, we find the relative strengths of the lines in a multi2 9 2 and then the relative strengths of the multiplets in a group
from
plet
9
having the same parent terms by a reapplication of 2 2 with SL J, SL J*
?
by U-U^L, I^L^L This expresses the strengths of all lines in the

in
terms of S(yIv v LIV y^S^-U1 }.
group
get the proper phases for the
S^ s if on each application of 2 9 2 we carry the factor
where ( I) 2 is
replaced
We
(1)
written in these formulas and also in the second formula of 2 9 2b, when
J(J+l) S(S+l)i-L(L i- 1) is negative. This amounts to carrying in
Table
I9
satellites
when
a minus sign for second order satellites when L-+L 1, and! for all
and those principal lines for which S(S Jr l)>J(J+l) + L(L4-l)
411
OTEBMEDIATE COUPLING
278
In this way we find the phase of S^ for a given line in terms of the phase of
rv I
S&(y S *IF y^S^L^) correctly for eigenstates with the phases we agreed
9
to use in
For
I 11
energy matrix.
which the relative multiplet
in calculating the
transitions in
intensities
cannot be
obtained by the method sketched above, we must calculate the matrix

(ySLlPvy SL ) directly from the eigenfimctions used for the expression of
the energy matrix. We assume that the states of a term have the proper
relative phases as given
8
by the calculation of 5 Then (ySL\P\y
.
SL may
)
be obtained by finding the matrix element (ySLMs 3IL \P\y SL J}Is 3IL ) or
(ySL J3r\P\y SL J JI ) for any one component of the multiplet, since the
f
values for
all
formulas of Chapter ra be readily

This matrix element is obtained
).
may by the
(ySL\Ply SL
components
expressed in terms of
from the zero-order eigenfimctions by the formulas of
electron components (cf. 4 9 1 and 3 9 6)
(n
m 8 rnêrln
I- 1
66
and the one-
m 8 wj) = s
-l)(iij)}.
The procedure
(3)
for finding intensities in intermediate coupling is thus the
following: Put the calculated eigenvalues of the energies into the energy
matrix to find the transformations from the i$-coupling states in the usual
way. Find the matrix of Si for states with the same phases as those used in
obtaining the energy matrix. Multiply the two transformation matrices and
the S^ matrix according to (2).
There are at present practically no data suitable for comparison with the
theory, so we shall not carry through a detailed calculation.
Sum
rules.
Concerning the line strengths in a transition array, we can make certain

general predictions in the form of sum rules which are independent of the
coupling of the configurations.
The most general of these is the J-group sum rule** We call the array
of lines connecting all levels of a given J in one configuration with all
f
levels of a given J in another
configuration a J group. If neither of the
configurations is perturbed by outside configurations, it follows immediately
from 4 U 2 and the principle of spectroscopic stability (2 2 25) that in any
transition array the sum of the strengths of the lines in a J group is independent
the
of coupling. If several configurations are perturbing, we must
enlarge
/ group
to include all perturbing initial levels

levels before the above statement becomes true,
*
and
HARBISON and Joiosoy, Phys. Rev. 38, 757
all
(1931).
perturbing final
4U
LINE STRENGTHS IN INTERMEDIATE COUPLING
The next sum
rule of importance
the
is
J-file
sum
279
rule* which
an
is
extension of that proved in 3 9 for LS coupling. In a transition array,

the set of lines connecting a single given level of one configuration with
all levels of the other configuration we shall call the J file referring to
We
saw
3 9 that for
a transition array like d^p^zdfid, the

the
different
of
files
strengths
referring to the levels of d*p were propor
tional to 2J+ 1, and the strengths of the different files referring to d*d
were proportional to 2J-f 1, in LS coupling. We shall here show that this
that level.
is
true for
in
any intermediate
In
coupling.
this case the
jumping electron
not equivalent to any of the electrons in the ion in either configuration,

and the J-file sum rule holds for both rows and columns. For such an
is
5
4
array as d p*=?d in which the jumping electron is equivalent to other
electrons in one of the configurations, the strengths of the different files
t
referring to
d5
true of the
files
are.,
in
any coupling, proportional
referring to
In general, we
shall
c?
to 2*7+1, but this
is
not
p.
prove that for any coupling,
the strengths of the Jfile$
>.
referring to the level of the
\
i
jincLl
provided that the jumping electron

configuration.
In
to others in the
is
not equivalent to any other in the
this statement, the

\
{
another in
configuration are proportional to 2
final
jumping electron
may be equivalent
In case
not equivalent
to
^
configuration.
&
it is
either configuration,
the
sum
rule holds for the files referring
to both configurations.
That this sum rule holds in any coupling for the files referring to ps and
to pp in ps^=?pp and for the files referring to 2 in ps^p* follows from
j>
an examination of Table 2 9 or Table 3 10 In these tables, we see that the

sums of the S*s in all columns headed by the same J value are the same,
while the sum of the products of the S^s in any two columns headed by
the same J value is zero. Therefore, since the S^ s transform like the matrix
elements of an observable, it follows from 2226 that these sums are invariant
under any transformations of initial and final states which do not mix up
.
the
A similar statement applies to the rows

but not to the rows of ps^p2
values.
tables,
of p s *^pp in these
In order to prove in general the sum rule as we have stated it, we shall
apply considerations similar to these to the whole electric-moment matrix
for the transition array in the zero-order nlm^ scheme. Let us consider
the transition array connecting the configurations A and /z. We suppose
Z= 1 or 0), A contains nlk and
that in the nl and nT shells (where Z
* SHOBTLEY,
Phys. Rev. 47, 295, 419 (1935).
4n
INTERMEDIATE COUPLING-
280
In addition A and ft have the same numbers of

electrons in ail other shells. For a non-vanishing matrix element of P in
the zero-order scheme, the quantum numbers in these other shells cannot
When they do not differ the element has the same value
differ in A and
iJ^n V.
contains
ft
p.
^nlk ~ n
V which we now consider.

First we calculate the sum of the squares of the elements of P connecting
k~ l n l
k
The
the state O(wijw& wzf m?,...,m*?w?) of nl to all states of nl
must be all different if the state is to be
numbers in
k sets of
as in the
amy
nlk
quantum
>
allowed. This state will combine with the states

0)Y(tn|TOf,
...,
W(mlm}, rojmf
0*X),
...
CWT(i}mJ,
and
For combinations between
of w-Z^nT.
>
...
mjwif,
rojrW
1
*
mX),
...
mX)
WT we must
have
(4)
m8 = m*
the matrix
ml -ml or m| 1. From 6 8 we see that the square of
2
cal
is
cpand
element of P connecting
just |(nZwi*wif|P|wTma7nJ)|
The
numbers.
its
culated for the one electron which changes
quantum
is given by
to (i) Y for all
total strength of transitions from
6
<*>Y
mX
oo
031
s is just k times
The total strength of all transitions from O to all the
tMs and is independent of the quantum numbers in
We shall now show that for two states a and & of nlk the sum of
J
<t.
>
<1>
in the transition array is zero,

the scalar products of the elements of
6 are
a and
referring to
i.e., that the two files of the matrix of
5 do not combine with the same states
and
orthogonal. The states
<t>
ff
>
of nlk
^nT unless
they agree in k
a
of their sets of
where t & has-mf mf
If
quantum numbers
and 6 do not agree
has mfm\
say that
1 sets, the files referring to them are obviously orthogonal. If they
in k
do agree in all but the f^ set, they combine simultaneously only with the
of (4). The sum of the scalar products of the elements con
states (i)
*
<D
<!>
necting them with the
states of
(i)
T is
E (nZmfmpjPlnTmX) (nTmXJPNmf **?

and
<!>
this,
= tj2&
by
13 3 1 3
is
seen to vanish unless
mf
mi
),
mfmf-,
i.e.,
unless
Hence
in the
nlmsm scheme, the sum of the absolute squares of the

l
k
having a common state of nl is independent of
that state, and the files of matrix elements from two states of nlk have
2
vanishing scalar product. A simple extension of 2 26 shows these properties
matrix elements of
to be invariant under unitary transformations. Hence if we return to the

complete configurations A and \L in a scheme characterized by / and
4n
LINE STRENGTHS IN INTERMEDIATE COUPLING
281
we may say that the lines haying the initial state XJM have the
2
strength sum &{(nZ:P:^T)J H(l r) and hence that the J file referring to
the level AJ has the strength
values,
k(W+l)\(n&PWl
)\*E(l,l
(6)
).
This proves the /-file sum rule as stated, and gives the absolute values,
in terms of the one-electron component (nl\P\n r) of the strengths of the files.
f
we may say that if

a pure configuration A combines with two perturbing configurations \L and
v, then if the sum rule holds with respect to the J files referring to the
With regard
and v separately, it holds with respect

/z
in the perturbed case.
the still more special case of transitions a s *=?& p in which the valence
levels of
to these
In.
to cases of configuration interaction,
electron
A in the transitions to
files
Jumps from
s to
or
to s
and
not equivalent to any electron

can obtain yet another rule which
is
in the ion in either configuration, we

may be called the J -group-file sum rule. This rule says that for these arrays
the individual J groups may be broken up in one direction into invariant
The direction is such that all the Ines in the file have a level of xs
in common. These /-group files in Tables 2 9 and 3 10 are set off by solid
and broken limes, and their strengths are shown in parentheses. The
J files
strength of each allowed J-group file is the same as that of the
files.
referring to the levels of a 3? which cross it in the array. These statements

are proved (using 2 226), in jj coupling, by noting that two files of the
same J group are orthogonal since they refer to different quantum numbers
of the ion (the valence electron must have quantum numbers

3
by showing directly from 1I 8 that
J = (2J + 1)
)
\(p\P\8)\*E(p,
*),
ns^),
and
(7)
/ = /+ 1, J, and J- 1.
The best experimental data with which to compare these sum rules are
the anomalous-dispersion determinations by Ladenburg and Levy* of the
13
5
5
2p 3s. (We shall see in S that the
strengths in the neon array 2y 3-p
those in pp->p&.) Such a
strengths in this array should be the same as
made
has
been
by Shortley (foe. cit.) with agreement within
comparison
the same paper is reported an attempt to
In
the experimental error.
calculate the detailed strength pattern for this array by the method we
for
->
13
This calculation
have outlined, using the parameters of Table I for 2p
was not satisfactory because a small change in the parameters has a very
and the parameters used were
large effect on the calculated strengths,
3j>.
necessarily inaccurate.
* LADE*-BUBG and LEVY,
Zeits. fur
Phys. 88, 461 (1934).
5.
o11
282
The forbidden Maes
of astrophysical interest,
For a long time several prominent lines in the spectra of some nebulae
remained unclassified. They were often called nebulium lines because it
was thought that they might be due to a new element not yet known on the
earth a repetition of the history of helium s discovery on the sun before
it was found terrestrially, Bowen* finally showed that they are due to
transitions in oxygen and nitrogen in various stages of ionization. The atoms
involved have p* p* and p^ for their normal configuration and the lines arise
from transitions between different levels belonging to the normal con
j
radiation. We shall see

figuration. They cannot, therefore, be electric-dipole
that they are partly quadrapole and partly magnetic dipole in character.
The most systematic study of astrophysical data for identifications of
such lines following Bowen is that of Boyce, Menzel and Payne.f Calcula
have been made

by Stevenson and by Condon.* An earlier calculation by Bartlett is based on approximations not
tions of the transition probabilities
sufficiently exact to
ourselves to the
We
be of interest.
shall
&o
ô
Emit
configuration.
The most famous of the nebulium lines, called Nl

T
and J\ 2 were identified by Bowen as the transitions
A^
III. These intersystem

indicated in Fig. 6 11 in
transitions are made possible by the slight de~
partures from Eussell-Saunders coupling. As we

have seen in the preceding sections, the eigenfunc- ,,
,
,.
n called
ii j 10
v
tion of the level
is a hnear
^
P The
-
usually
bination of those for
5 and
com-
actual levels
g
3^
t
6U Bowen
-
tffe-
tification of the
Hllm>
^^ N
and
iden-
^-
will be designated
by the letters on the left in the

the approximate Russell-Saunders labels being given at the
right.
breakdown of coupling is expressed by the transformations:
figure,
The
The matrix components of the
00
spin-orbit interaction are given in Table I 11 .

evaluate the parameter f from the intervals in the 3 term,
obtaining
1
III. To the first order of the perturbation
C = 210cm- for
b and d
We
theory
are given
by
These values though small, are what make the nebular lines
possible.
s
BOWEN, Astrophys. J. 67, 1 (1928).

f BOYCE, MJENZEL and PAYSTS, Proc. Nat. Acad. Sci. 19, 581 (1933).
t STEVENSON, Proc. Roy. Soc. A137, 298 (1932);
CONDON, Astrophys. J. 79, 217 (1934).
BABTLETT, Phys. Rev. 34, 1247 (1929).
5U
THE FORBIDDEN LINES OF ASTEOPHYSICAL INTEBBST
283
By means of tie results of Chapter vni the Rnssell-Saunders eigenfunctions are expressible in terms of those of the zero-order scheme. From
the latter the matrix components of quadrupole moment and
magneticdipole moment may be computed. In this "way the following results were
obtained for the line strengths:
Line
Here
s 2 is
Magnetic dipole
the integral
= -e
s2
|*3C
Jo
Quadmpole
r 2 E*(2p)dr.
The value of this
in atomic units, using the eigenfunction of Hartree and

Black,* is 1-24 so using the formulas of Chapter iv connecting line strength
with transition probability we find for the absolute value of the transition
:
probability for the two lines
B2 = 0-018 sec-1
)
A2
The
lines are
A(^4 a spj)
,
= 0-006 sec-1
almost entirely due to magnetic-dipole radiation, the quad

transition probability giving only 0-1 per cent, of the
rupole term in the

whole amount.
The
line (7
->-4
2 is
called
an auroral line
since in
OI
this transition gives
the green line of the aurora and the night sky. Being a line for
which A J = 2 it is entirely of quadrupole character, but not being an intersystem combination (A&= 0) it does not depend on the partial breakdown
of coupling. The calculated transition probability is
rise to
The
triplet
3
->
is
The
interesting.
line
CQ-*DQ
is
forbidden in
all
approximations by the general exclusion of 0-^0 transitions in J. Likewise

3
P1 cannot be a quadrupole line as 0-1 change in J is of vanishing
strength for this type of radiation. On the other hand G Q-+B2 cannot occur
<7
->
with magnetic-dipole radiation because of the dipole selection rule on J.

We therefore have two lines close together, one of which is due to purely
quadrupole and the other to purely magnetic-dipole radiation. The calcula

tions give
A(O
A(C
J? 2 )
3
,
= 1-5 X
P = 0-102
1)
lO-^see- 1
1
sec"
indicating that the quadrupole line
magnetic-dipole
*
is
(Quadrupole)
(Magnetic dipole)
3
only about 10~ as strong as the
line.
HABTEEE and BLACK,
Proe. Roy. Soc. A139, 311 (1933).
CHAPTER XII
TRANSFORMATIONS IN THE THEORY
OF COMPLEX SPECTRA*
We
wish In this chapter to consider, more in detail than heretofore, the

transformations between the various schemes of states of interest in the
of importance the
theory of atomic spectra. There are five representations
scheme and
as
the
shall
two zero-order schemes, which we
l
designate
the
8
L
scheme, tie two LS-Goiipling schemes characterized by
3
mm
i
8LM
jm
SLJM respectively, and the ^"-coupling scheme of I 10 characterized

by the electronic j values and resultant JM. We shall consider in particular
and
the four transformations
nljm *-?
\\
33
JM *=? SLJM
and incidentally the transformation nlmsmi 2=^ SLMS L In preparation

for Chapter xnr on configurations containing almost closed shells we shall
establish a correlation in each scheme between states of a given configuration
containing an almost closed shell and those of a corresponding simpler
configuration, and determine the relation between the corresponding*
.
transformation matrices^
Configurations containing almost closed shells.
1.
Let us consider a configuration which contains, outside of closed shells, a

shell &* which is complete except for e missing electrons, and in addition
other electrons. Here e and -q are considered to be small integers, although
TJ
formally there is no restriction placed on their magnitude. This configura
tion we shall designate as configuration ^, since such configurations occur
5
mainly for elements near the

tion
we
same
right of the periodic table.
shall correlate a simpler configuration
closed shells,
electrons present in shell &*,
To
this configura
which contains the

and the same 77 other
J$?
electrons.
In the nZwywj scheme we shall correlate to a given state of
numbers
listed in the
S? S
and whose
i\
missing
and ml values of the e electrons in shell

s
other electrons have the same quantum numbers as
electrons have the negatives of the
f of
& with quantum
standard order of 56 that state of 3% whose
SiiOETLEY, Pnys. Bev, 40, 185 (1932), 5; ibid. 43, 451 (1933).
considerations of this chapter can all be readily extended to configurations containing
more than one almost closed shell. This case is, however, of little interest.
t
The
I 12
CONFIGURATIONS CO^TAIOTNG ALMOST CLOSBB SHELLS
taken when the quantum numbers are listed in standard order

w where w is the sum of the m values of the electrons in
those of oSf
with the phase

shell
&
285
l)
It is clear that this gives a one-to-one correlation
between all states
two
s and
configurations, the correlated states having the same
12
2
of
values.
we
in
that
matrices
With
show
the
this
correlation
shaU
L
S2 L2 and L*S are the same for the two configurations, and hence that the
of the
same matrix
will transform
coupling thus have the same

tions,
them to
The allowed
states in
8LM8ML or SLJM values for the two
and corresponding states
binations of the correlated
LS coupling.
in
mm
s
LS coupling will
"be
LS
configura
the same linear
com
states.
In the nljm scheme we shall correlate to a given state of JS? with quantum
numbers listed in the standard order of 5 6 that state of 3t whose missing
,
have the same j values but the negatives of the m values of the
e electrons in shell
of 3?, and whose rj other electrons have the same
as
of J? taken when in standard order with the
those
quantum numbers
Here I is the azimuthal quantum number for the shell y
phase (
which in
contains p electrons with j=l + % and q electrons with
=
values. With
|+
(p q=* ). These correlated states have the same
j
2
this correlation we shall show in 3 12 that the matrix of/ is the same for the
two configurations. Hence the allowed states in jj coupling have the same
electrons
<$?
i)<e-KO<+*).
i.
JM
values, but the electronic j values of 3? are those missing in the corre
lated states of ^. Corresponding states in jj coupling will be the same linear
combinations of the correlated jm states.
We
show in 5 12 that with these correlations between zero-order

states of JS? and 0k the transformation from the msm scheme to the Jm scheme
is the same for the two configurations. Hence the transformation from LS
J? and St.
coupling to jj coupling is the same for the correlated states of
This transformation will be given explicitly for a number of two-electron
shall
configurations in
2.
6 12
The transformation to LS
coupling.
In order to show that the same matrix

it
rilmfMi states of J? and 9k to L8 coupling,
transform the correlated

only necessary to show that
will
is
for the two configurations.

,
8
Formulas for the elements of these matrices were given in 4 Let us denote
f
by A# 9 A ^ 3 ... zero-order states of JSf and by A M A m ... the correlated
the matrices of
L 2 S2 and L-S
,
are the
same
states of
^.
2
is given by
of
general non-diagonal element
8
same for
the
4 3a. If neither a nor b is in shell &*5 the value is obviously
3? and 91. If a, but not 6, is in shell SP* the second term vanishes. In order
Matrix of
IA The
to obtain a value for the
first
term a and
a!
must be of the form
(w,|)
>
286 TRANSFORMATIONS IN THE THEOBY OF COMPLEX SPEGTEA

(
M|
_ i).*
2 12
and 01
we must consider the phases in greater detail. If thein-
8
Equation 4 2 shows that in
are essentially equal

dividua! set lying between
this case the
elements for
and (wi - 1 ) is present in JSP so far as shell

A m and A m
is Concerned A s and^ difier by an odd permutation, while
is
element
this
missing. This
do not differ in order. The converse is true if
2
Z
introduces one difference of sign between (A&\L \A&) and (A a\L \A a). How
one in A m and A m This
ever, the sum of the m values in shell Sf differs by
=*=
(wi,)
&
from the different choices of phase for

two elements just equal. In a similar
these
makes
which
^(Agt) and ^(A m ),
a and 6 are both in shell ?.
in
case
wav one demonstrates the equality
results in a second difference in sign
The diagonal element of 2 is given by 48 3b. The first term is the same for
A m and A # as also are the parts of the second and third terms arising from
shells other than SP. The parts of those terms which arise from shell SP are
the same if calculated for A s as for the group $ missing electrons of A M
The equivalence of the calculation for the missing electrons to the calculation
for the electrons present In _4 m is a direct consequence of the following
,
half-odd integers) -Z, ...,

interesting relation: If one takes the integers (or
I and arranges them into two groups a and /?, then the following sum has the
same value for groups a and
(the individual integers are denoted
)8
S[Z(Z+l)-mf] + S[-^Z+l)HX-z]>
by
L ):
(1)
mi
where the second sum
is
taken over only those m/s for which
I is
also
in the group.f
TMs
The
2
completes the proof that the matrix of L
calculation for
Jlatrix of L*S. If
is
very similar and
is
8
given by 4 5b.
* For
example shell
A in two electrons, the calculation is

and A differ in one electron, the matrix
and for j? have the
If a and a! are in shell
.
If
SP>
& of A s and A
Ay
? and 91.
is the same for
not be discussed in detail.
from
differs
much the same as for L 2 and S2

element
will
%.
...
might have the following electrons :
Here x indicates tie presence, - the absence, of a given individual set; all sets except those
noted explicitly are the same for Ay and A#. In this case A& and A& will nave the forms:
...
...
(-mjf (-,)(- TOl + l)x
x
A & ...
x
A a ...
x
f This may be proved as follows: The division into groups a and ft may he made by splitting the
I, ... I into r sections, putting the first section into group a 9 the second into /?, the third
into a, etc. Consider the sections as defined by section points p^, pz , ..., r+1 , such that the
section contains the numbers p t -f- -| Pt -f-f , ..., Pt+i i- The value of the sum (1) when z runs over
the numbers of thissectionis /( pt ) -rf(p t ^.j), where /(p}=|(Z-f ^) 2 - Jj52 . Then the sum for group a has
tlie siam for
.
has tlie value/(^2 ) +/(p 8 ) -f/(p 4 ) +
the value /(ft) -*-/( ft) +/(ps ) +/<P4) +
*
ft
One of these expressions ends with the term f(pr ) 9 the other with/(pr+1 ). But /(ft) =/( - i - J) =0
and/fp^j) ~/(l -r }) =0. Hence the sums for a and f3 are equal. This proof was suggested to us by
series
<**
j>
wMe
Professor Bennett of
Brown
University.
2 12
form
may
(wij)-,
THE TRANSFORMATION TO LS COUPLING

287
(m 1)+, for 31 they will have the form - m + I)-, - m^. One
z
easily verify that
A# and
^ Am ^
and
then
differ
by the same per
mutation; however, one difference in sign of the matrix elements is intro

duced by the different l values for A M and A m For J$? the second, third
terms vanish unless b = (m -l)-,c = (wij)+ respectively are present. For the
c = (-fln; +l)+ are present.

second, third terms vanish unless b = (
l )~,
"Now if neither b nor c is in $?,
must
both
be
in 3$. Hence in this case
they
(A#\L-S\A #)=(a\L-S\a )
(A *\L-S\A a) = - {(a\L*S a ) - 2(a|-S|a )} = (Ajf\L*S A ^}
8
8
(cf. 4 2 and the discussion under 4 5b). Similar considerations hold
other two possibilities.
f
for the
The diagonal element (2 8 oc) of I>Sis obviously the same for J5? and 9t with
our correlation. This completes the proof of the equality of the -Decoupling
transformations for J? and 5?.
3.
The transformation to jj
coupling.
The transformation from the nljm scheme to the jjJM scheme may be
obtained by any of the methods sketched for LS coupling in Chapter vxn;
the diagonalization method is however much simpler than for LS coupling,
since only one matrix, that of J 2 need be diagonalized, and the trans
formation is diagonal with respect to the j values of aE electrons in addition
to If. The elements of the matrix ofjF2 in the jm scheme may be written down
,
2 in
S
the
sml scheme (cf. 4 3). It has nonelements
between
which
differ
in
states
diagonal
only
regard to two in
dividual sets of quantum numbers; say that A contains a and 6 where A
contains a! and b This element has the value
in exact analogy to those of
where (a\Jx \b) = S(nala ,nbl ) )(ja ma \Jx \j bmb )

element of J
is
3
given by 3 4. The diagonal
2 is
(A J*\A] = &M* +
\
W S {/(.f + 1) - (m)
Now with the correlation given in
- 4 S (a| Jx \b)*.
a<b
I 12 it is easily seen,
by an argument
the direct analogue of that given above for JL2 in the 6?% scheme
2
(noting that J is diagonal with respect to the number of electrons in shell ff*
which
is
with j = l -|
matrix of J^
to the q occurring in the correlation factor ), that the

the same for the two configurations and hence that the
i.e. 5
is
transformation to jj coupling is the same.

4,
The transformation between zero-order
states.
We shall in this section consider in general the transformation connecting

two systems of zero-order states, and the relation between these trans
formations for J? and ? with certain general correlations. In the next
section
we shall specialize to the particular transformation of interest to us.
THEORY OE COMPLEX SPECTRA
288 TRANSFORMATIONS IS
4 12
A -eleetr0n configuration for which there are n possible
Let us consider an
quantum numbers.^ In the first zero-order scheme let

individual sets, in some standard order, by the numbers
sets of one-electron
us denote these
and the corresponding states by ^(I),^(2)
1, 2, ...,n,
scheme
let
us use the notation
^(/l)* ^(/^)j --9
l, ,2, ...,
f r t&6 states.
^00
one-electron systems
...
iff(n).
In the second
n for the quantum numbers, and
The transformation between these two
given by
is
The
a2
zero-order antisymmetric function belonging to the complete set a1 ,

v
... a* of
quantum numbers is defined by
?
= j/&(ai)fc(a) ... MO?)

*)
y
a axe JVmimbeisof theset 1, 2,
(2)
<*...
(cf.
6
).
Here a
the order a1
1
,
2
.,
w, arranged in
2
<a
...,
<...
<ot-
The transformation of
following calculation.
this state to the
From
(2)
and
primed scheme
is
given
by the
(1)
Wa... *).* S ^(^) CjPlai)
...
S #(^)(^|^)
= S
two of the
jB s are equal the antisymmetiic factor $/... vanishes, while

a given set of /fs this factor is the same, irrespective of the order of the
/?X to within a sign which is just correct to give us a determinant of the
If
for
transformation coefficients;
i.e.
T( ai*>...a*-)=
S
ff
...<
Q{^-^(^^...^),
*
\ *
/
(3)
]&</?*<
where we use the notation
The transformation
y
... a
state a1 a 3
2
1
J?
connectmg the state ^ / jS ...^ with the
(quantum numbers in standard order) is given, then, by just
coefficient
this determinant:
CJ^JP...^!*
* JOT
1
example, for the configuration 2s Sj) , A =3 and
=8, the possible sets of quantum
numbers being given, for the M/n z scheme, by ,1 =2s C^, ,2 =2 0~,,3=3pl+, /4=3pl-, 5=
/
for the jm scheme by l=2*i, 2=2*&
k 3=
THE TBANSFOBMATION BETWEEN ZERO-ORDER STATES
4 12
289
Since the transformation (I) lias no components connecting sets belonging

to different shells, the determinant (4) and hence the transformation co
efficient (5) splits up into a product of factors each referring to a single shell.
M electrons
For example, if the configuration consists of

in other shells^ we may split off a
and N
on shell ?:
(I
...
M) in shell SP
factor giving the dependence

1
at\
am.
(6)
Because of this property, in considering the relation between these coefficients

for
JS?
and
9t
we may restrict ourselves purely to the
shell
SP so long as the
other quantum numbers are the same for the correlated states of
J2?
and St.
We
states in the shell &*, and con

shall suppose that there are altogether
sider the relation of these components for the configuration consisting of e
electrons in this shell.

electrons to that consisting of
Let us seek an invariant correlation between an e-electron state
(m
e) -electron state
whose
missing electrons have the
and the
same quantum
numbers. This correlation is to be independent of the system of zero-order

states used in the description. We shall denote a given e-electron state by
2
1
am ~ )
a ) and the corresponding (m e)-eleetron state by T(ax a2
YCoc a
where the a s and a s together make up the set 1, 2, ... m.
Let us define a linear operator Z by the relations
. . ,
. . .
(7)
This operator is seen to be unitary, since the states on the right form a
normalized orthogonal system. In the primed scheme we shall define a
corresponding operator
Z by
(8)
).
The relation between Z and Z is given by the following

t
calculation
...a
-D
since the terms for c 1 = a1 ,

cs
. .
.,
cm
~e = am~~ furnish
just the Laplace expansion
*9
290 TBANSFOEMATIOKS Es THE THEQBY OF COMPLEX SPECTRA 412

*""
<yf3\
wMle the otter terms
vanish. Since
\I.,.m/
pressed in terms of the
TMs shows
l 2
T(x a ... a
),
any
state
may
we see by comparison with
(7)
be exthat
to be invariant to within a constant factor,

an
-electron state and then take the complex
If we apply this operator to
conjugate, we obtain an invariantively correlated (m e)-eleetron state.
We take the complex conjugate because by (7) Z acting on a state gives a
the operator
state in the dual space., whereas we want a state in the same space.*
may now calculate the relation between the transformation co
We
efficients:
since a transformation coefficient

tion.
TMs shows
electron states
is
is invariant under a unitary transforma

that the transformation coefficient connecting two eessentially the complex conjugate of that connecting the
whose missing electrons have the same quantum

and
the
exact phase relation between these coefficients.
numbers;
gives
(m
5.
}-eleetron states
Hie transformation nlm/n^nljm.
Of transformations of the type considered in the previous
section,
we
are
most interested in that from the m/rii scheme to the jm scheme. We shall
now show that this transformation is the same for configurations Sf and ^,
with the correlations of
I 12 .
Eecause the transformation factors according

(cf. 4^6), and since in our correlations all electrons except those
in she! ff have the same quantum numbers for 3? and
we need only
to the shells
&
show that this transformation is the same for the

and the m - e electrons in shell 5^ in St.
electrons in shell
? in
a linear correlation, since taking the complex conjugate is an invariant, but not
(c^F^c^F unless c is real), operation. A linear correlation such as is obtained directly by
omitting the bar on the right side of (7) is invariant only under real orthogonal transformations
of the zero-order states*
linear
5 12
THE TBAKSFOBMA.TION
If we denote the
msmj states by
291
nlmsm^nljm
<j>(nlm/rii)
and the jm states by
ifs(nljm),
correlating these schemes respectively with the primed and unprimed

schemes above, the transformation 412 1 is given by 45 8, and is entirely real.
The determinant of this transformation

J
l) .TMsvalueistx>beinsertedin4
Yfa
oc
for
an
shown
shell is easily
to be
Moreover, from the standard order
10.
6
3* *
quantum numbers (5 it is seen that ( 1) 2 = ( I) 5 *,
v
iJ-^-w-etf-i), where q is the number of electrons in
l)2 =(
of listing
and
12
. .
(JPJP ...
0"
ae ) with j = Z - 1. Since
^"-â* ...
am~*) = ( -
M = M + ML
B
4 12 IO becomes
l^^C^jS
1
,
...
j^x1 *2 ... at*).
(1)
The states on the right side of this equation have just the quantum
numbers which are missing on the left. Our correlation requires that they
have the negatives of the m8 ,ml} and m values missing on the left. But this
is easily accomplished. Let the coefficient on the right of (1) be
(wJw,mfm?,...,m;wf|Z4-J^^
where p + q = e. The coefficient we would like to compare with this
(2)
is
(-TOj-^V.^-wiJ-mHJ + i:-^^
3)
(2) is in standard order, (3) is seen to be also. Now from a comparison

of the two determinants of type 4125 which give the values of these two
5
coefficients, and by using the relations (cf. 4 8)
If
(ms mz |Z+|m) = (
|m)=
- 1}
(3) is
(ms mj[l
it is
seen that the element
ms
m^Z+J
7Wj|Z
m),
J
m)
times the element
(2).
Hence, returning to (1), it is seen that the ratio of the transformation

coefficient for the almost closed shell to that of the -electron configuration
whose electrons have the negatives of the s ,m^m values of those missing
from the closed
shell is
__ ^ \jfjt-f (+ix4fl w
When we make the correlations of

for the
mm
that the
G.
scheme and a factor
m/nrjm transformation is
I 12
(4)
which include a phase factor
l)
M*
l)(*HX+0 for the jm scheme, we may say

the same for the two configurations.
The transforinationjj/Jf^îJJf.
We have found a correlation between the zero-order states
of
and
3$
such that the transformations
jjJM^nljm,
nljm^nlm/fi^
and
nlmsm^SLJM
is seen to imply that the trans

the same. More explicitly, the situation is as
are the same for the two configurations. This
formation
jjJM^SLJM
is
19-2
292 TRANSFORMATIONS IN
THEORY OF COMPLEX SPECTRA
8 12
A given state of the configuration J2? characterized, by the quantum

numbers SLJM (this state is not uniquely determined for two reasons
follows:
first,
there
is
an arbitrary phase; second, the set SLJM of quantum numbers
not in general complete for more than two electrons so there may be two
independent states characterized by the same SLJM) is a certain linear
is
combination of the wyn/ states; thesaine linear combination of the correlated
mm
s
states of
a state of 31 characterized
is
statement holds for a jj
way
SLJM. A similar
s
of 5f* The transformation between the
obtained in this
by this same
JM state except that the missing j s of M are those
is
LS- and
^"-coupling
states of 3k
the same as the corresponding transformation
farJS?.
Xow to obtain this SL-jj transformation by the combination of the other

is in general a very tedious process
involving the multiplication of
three matrices of high order, for no two of the three transformations are
diagonal in common with respect to more than If. On the other hand the
three
and
resulting transformation is diagonal with respect to both J and
splits in general into steps of a relatively low order. For example, no twoelectron configuration or its equivalent can have more than four states
of the same J and Jf, while pp or p 5 p has altogether 10 states of Jf =0,
ddoi ds dhas
19.
It seems, then, to be desirable to find a direct method which will enable

this transformation to be calculated with
greater ease. The following
accomplishes this result for two-electron configurations, this being in general
the only case in which the quantum numbers

unique state (to within a phase).
SLJM
or
jjJM
define
The procedure we shall follow is that of calculating the matrices of X2 and

S in the jj-coupling scheme and then diagonalizing them to obtain the
2
L8 coupling. Since these matrices will split into steps of

at most fourth order, and since their eigenvalues are known, this
diagonalization is a very simple procedure.
The matrices of JL2 and S* in a non-antisymmetric scheme in which the
transformation to
quantum numbers n 1, j, and the second electron the

n
numbers
I
are given by formulas
quantum
j with resultant J y
12 3 2 and 103 2. We shall define our phases by
correlating, in the process of
first
electron has the
vector addition,
Xow
s,
IJ; s
f
,
,j ;j,f,
J; with the jl9 j29 j respectively of
10 3
=I|-f f-f 2i1-JL 2 The first two terms have known diagonal
matrices. The matrix of the third term is diagonal with
respect to nl, n l JM.
jL
If
we omit
these
writing simply the values of j

the elements of this matrix are given with the abbreviations
quantum numbers,
to
= +
and j
293
THE TBANSFORMATIOS- jjJM^SLJM.
6 12
--J--l)f
-2wJ^"
(I)
either the upper or lower undotted sign is to be taken

(In the first formula
with either the upper or lower dotted sign.)
The matrices of 2S1-S2 and of 2 1 2 and 2S1 JL 2 are given by formulas
-<S
LT by 5X if we replace
and - Ifiw by + l/2w; similarly we may replace 2 by
(2w + l)/2w by Ij2w
by l/2w and 1/2^ by + l/2ti/.
52 if we replace (2^ +
In (1)
closely related to the above:
we may
replace
l)/2u>
In terms of these non-antisymmetric matrix components, the components

as
between antisymmetric states may be obtained by considerations such
those of
3 10 .
Let a = id,
JS
= n V and let all differences in quantum numbers
Then
be indicated by primes, subscripts merely denoting the electrons.
unless ot = /? and j=/ the antisymmetric
5
&
Now
Jlf
)
(
)!
*(*.&&
Y(oy jSJ JM) 2-^(*i Ji J2
and
to
with
Wgig,
njh
respect
if t; is a symmetric observable, diagonal
it is easily
^")
seen that for non-equivalent electrons

(
(ajjSjVIfHaj^
a ^^)
3)
4)
Consideration of the matrix element

(ajoc/
Jif |v|a7* aj* /If) = (jj
\v\
JT)
cases. First, if j
for equivalent electrons must be divided into three
tt
m
^j (2) applies to both states, and
/ =J
a^^J
either antisymmetric or sym

^(JiJD is
and excluded states. For an allowed state
metric, corresponding to allowed
jV j
Second,
if
we have
(cf.
3*2)
Third, if j = /, / =
OT
(jj \v\j
j"
j>
)=V2tfi
jj)-
0V/)
(5b)
and both states are allowed, we have simply
2
These formulas, then, enable us to calculate the matrices of L and
antisymmetric states in
S2 for
$ coupling using the non-antisymmetric values (1).
294 TRANSFORMATIONS IN THE THEORY OF COMPLEX SPECTRA
612
two-electron configurations up
12
On the left of
to dd gives the set of transformation matrices in Table I
the matrices are given the values of j and/ for the jj -coupling states the
J value is shown at the top in the RusseU-Saundeis notation. The trans
The diagonalization
of these matrices for
all
formations are independent of M.

TABLE
I 12 .
f| a
a
G.%
-S
77?e Transformation from jj Coupling
to
L5
Coupling
The jj-coupMng states here have phases in accord with those used in
deriving (I). The phases in LS coupling are arbitrarily chosen. These trans
formations could be used to obtain the matrix of spin-orbit interaction in
LS coupling, since this matrix has a simple diagonal form in jj coupling.

The transformation to jj coupling can alternatively be obtained by diagonal-
LS
11
izing the matrix (Table I } of spin-orbit interaction in
coupling; in this
case the phases of the jj-coupling states would not be determined. This would
possibly be useful for a configuration containing more than two equivalent

electrons if the matrix of spin-orbit interaction is non-degenerate.
CHAPTER
XIII
CONFIGURATIONS CONTAINING ALMOST

CLOSED SHELLS. X-RAYS
While most of the considerations up to this point have held in general for
configurations containing any number of electrons, actual calculations by
the methods given have usually become very laborious for configurations
containing more than a very few electrons. There is little hope of improving
this situation for multi-electron configurations which actually are very
complicated, such as those which occur for atoms near the middle of the
periodic table.
But there is a class of multi-electron configurations occurring
near the right side of the table which are given unexpectedly simple pro
perties by the Pauli principle. These are the configurations containing
almost closed
shells
which were shown in
I 12
to have essentially the
same
allowed states as the simpler configurations obtained by replacing the

holes in the almost closed shell by electrons. The transformations for such
*
configurations were shown in Chapter xn to be immediately obtainable from

the transformations for the correlated simpler configurations. In this
chapter
we
calculations
shall give
which
an arrangement of the energy-level and intensity

be either obtained directly
will enable the matrices to
from those for the corresponding simpler configurations, or calculated with

no more labour than for those configurations.
1.
The
The
electrostatic energy in
LS coupling.*
electrostatic energy matrix for configuration
M is in general quite
from that for 3? and must be separately calculated. Configuration

3t consists of a closed shell minus c electrons, ij other electrons, and any
number of completely closed shells. For a given state A in a zero-order
2
1
scheme, denote the individual sets of the closed shell SP by a a ...., a
a ~\ ..., em where a6* 1 ..., am are the sets occurring in A and a\ ..., a* are
the "missing* sets. Denote the rest of the sets occurring in A outside of
closed shells by a 1 a 2 .,, a ?. It will usually be most convenient to cha
racterize such a state by giving the
missing sets and the y others. The
missing sets must satisfy the Pauli principle no two of them can be
different
identical.
By writing down
all
fying the exclusion principle,
possible combinations of missing sets satis

all possible zero-order states of the
we obtain
configuration .
*
SHOKTLEY, Phys. Bev. 40, 185 (1932);

JOHSSOS, PfayB. Rev. 43, 632 (1933).
CONFIGURATIONS CONTATSTNG ALMOST CLOSED SHELLS
296
I 13
The electrostatic energy matrix has non-diagonal elements between states

of the same configuration only when these differ in regard to two individual
6
sets. Since these elements have the simple form given by 7 4, involving no
summation over the electrons of shell
they cause no particular difficulty
.9",
for configuration M. Since most of the electrostatic energies of interest are

of the zero-order diagonal elements,
given by the diagonal-sum rule in terms
to a discussion of the diagonal element for the state A.
we
proceed directly
diagonal element
TMs
is
6
96 2
given by 9 2. The summation in
may
con
veniently be divided into several parts.

both of the individual
First, there are the terms in the summation in which
of these terms are
sums
the
sets are contained in completely closed shells;
6
6
in which one set is in a
given by 9 I2 and 9 13. Second, there are terms
closed shell and the other is one of the -q electrons of group a. These terms
6
are given in simplest form by 9 I I. Third, there are terms in which one set
SP. These terms are given, for each closed
is in a closed shell, the other in
terms
already considered are the same for
(in
shell,
e) times 9*11. The
by
remainder of the
diagonal elements of the configuration. Finaly, the
of closed
diagonal element consists of terms in which both sets are outside
all
shells,
and has the value
[J(afa?)-K(afQ?)]
(A)
(1)
The inconvenience here is caused by the long summations over the range
e -f I to wi; these we wish to reduce to summations over the more convenient
[For example,, if shell P contains nine d electrons, the sum in
range 1 to
terms, while if the range were 1 to
(A) runs over 36 terms, that in (JJL) over
e the first sum would contain no, the second TJ, terms.]
.
Consider
first
the
sum
(A).
S - S
We may write
t- S
S
+
[J(oa*)-Jf(tfa*)l,
(2)
THE ELECTROSTATIC ENERGY IX LS COUPLING-
297
{
where we have used the relations J(a i ai )^K(a>ai ], J(ai a^)=^J(aâ ), and
Kiata^^K^tfaf). Of the three sums to which we have reduced (A), the
6
first is a sum over the closed shell SP and is given by 9 12. The second is
6
11 with nl = n l
Only the
given by e times an expression of the form
f
third term depends on the particular state A of the configuration,

has a simple sum over pairs of electrons chosen from e.
The term
(ft)
of
TO
this
becomes similarly
7?
>=E
of which the
(1)
and
]-
=1
first
term
is
S S
[J(a*aO-
6
given by 9 I1 and depends only on the con
figuration.
Since the term
(A\Q\A) which
(v) is
already in
its
simplest form,
we
find for that part of
may vary within the configuration the value

"V
F T//j irti\
/TY/fi
i\1
f [J(oaO- ^(a*^)]
O(\
(fO
w
Hence, aside from common terms, we may calculate the diagonal elements
of the missing electrons of shell /* exactly as
for a simple configuration except that we reverse the sign of the interac
tion of a missing electron with an electron in another shell (i.e. in group a).
In using the diagonal-sum rule to obtain from these the iS-coupling energies
and 8 values of the state A are given by
we must remember that the
using the
quantum numbers
the negatives of the values for the missing electrons plus the values for
group
a.
As an
illustration
we
shall sketch the calculation of the electrostatic
in which the
np n p. We shall use a notation of the type (0+ 1~),
from the np
first entry gives the quantum numbers of the electron missing
electron n p. The double
shell, the second the quantum numbers of the
zero-order states then has
B values of the
L and
entry table giving the
7
the form (cf. 1 2)
energies for
(5)
298
COXFIGBATIOSS CONTAINING ALMOST CLOSED SHELLS
From
this
we
obtain, using the
above
results,
13
by the usual diagonal-sum
procedure
TMs is one of the important almost-elosed-stieli configurations which is

correlated to a (non-equivalent) two-electron configuration. With regard to
such two-electron-like configurations we may make seTeral general state
ments.
m ~l s Jim the same

singlet-triplet separaplace the configuration l
as Is; this means, e.g., that the separations for p*s and d g s are given by
ds. TMs may be inferred, if we like, from a
the same integral as for p $
In the
first
scheme to include two or more almost-closed

The procedure is essentially the same as the above with the additional
generalization of the above

shells.
direction that- one assigns the usual (not the reversed) sign to the integral
connecting electrons missing from two different shells. If we consider the
8
shel of !OT ~% as having one missing electron the result follows immediately
3
from 8 s 16.
m ~i
r, and
Second, the electrostatic energies for the tw-j configurations, l
-1
6
18
the
toll
if
we
that
am-dafed
are
the
same.
TMs
8
note
follows
from
?,
**
double entry table of the type (5) for lm l V may be obtained from that for
r -1 ! by replacing the state (wi^mj
ms l ms m^.
a mz ) by the state (
~~
5
p d
d 9 p have the same electrostatic energies.

Third, since J(i,j) depends on only the absolute values of
3
-1
the terms in the electrostatic
Hence,
e.g.,
energies involving Fk s for Z
II|
m\,m*8 m*l9
will
be just
the negatives of those for IF. Iûthermore K(i,j) vanishes unless m^=m^ 3
~ 0;
hence in a table such as (5) we obtain Gk *s only for states having
8
none of the triplets have G*& in their electrostatic energy expressions. These
facts reduce our calculation for such cases to a determination of the co

efficients of
the
G%s
for the singlets.
The following formulas show certain

&k *s. For each singlet, at
striking characteristics with reference to these
most only one

Since for even
Gk has
namely that with k=L.

all vanish, and for odd
f
I + l coefficients of
with
even
Jc vanish,
other
(?/s
every
singlet has the same
as
the
TMs
energy
corresponding triplet.
peculiarity was not noticed experi
l-f
a non-vanisMng
coefficients
of
Gk s
coefficient,
with odd k
mentally because for such configurations the spin-orbit interaction

usually as large or larger than the electrostatic.
is
THE ELECTKOSTATIC ENERGY IN LS COUPLING
I 13
299
The
electrostatic energies for several configurations correlated to twonon-equivalent-electron configurations follow:
*F=-F -2P2
*P=-F ~2F2 +9CK?3
D= -P + 7P2
D = -F -f 7F2
- 4Fa -
?=~F
- 4F2 -
P=-F
- 5F3
- SPa
P = - F + 15F2
ip = ~F@ -hl5F2
3
D= ~F -I2F2
=-F
*<?=-P
*<7=-P
P4
P4
tfs-Po-lOPj- 3F4
- 7Pa + 84F 4
ip=-p 7Pf + 84F4

s^= -F -14F2 -126F4
*S= -P -14Pa - 126^4 + 10^0
D=-F
6F2 -f99F4
JD=-P
- 6F2 -f99F4
P= -F
P= -P
-24FS -66F4
-24Pa -66^ + 700!
m~~ e
containing -purely an almost closed shell

we may say at once that the electrostatic energies are the same as for the correlated
.For configurations of the type
from 86 16 if we note that the double entry

may be obtained from that for l merely by
7
4
7
and
all
the
of
m/s. Hence d j? and/ have exactly the
m/s
signs
reversing
2
and / 7 respectively. We may also show that
electrostatic energies of d3
~
m
as I s for any e.
l
s has the same electrostatic
configuration l This follows

table of the type (5) for l m
.
"
j>
energies
2.
Tlie spin-orbit interaction.
We shall show here that the matrix of spin-orbit interaction in any coupling
scheme with the correlations of Chapter xri
is the
same for configurations
and ^, except for reversal- of the sign of the spin-orbit coupling parameter
nl re
with the correlations of Chapter xn,

ferring to the almost dosed shdl P. Since
and 3t, it will suffice to show that
Jg?
for
same
the
are
all transformations
The simplest scheme to consider
scheme.
one
this relation is satisfied for
any
the nljm scheme in which the spin-orbit interaction is diagonal and in

10
of
s. Its value is given by 1 2, in which the coefficient
dependent of the
electrons of shell &. If this is written
the (m tni is given as a sum over
sum over the e missing electrons our
a
minus
shell
as a sum over the whole
theorem is proved, since the sum over the whole shell vanishes. (This
is
<)
CONFIGURATIONS CONTAINING ALMOST CLOSED SHELLS

7
L = J1S = Q in 4 2
vanisMng may be inferred, e.g., from the fact that
300
a whole closed
2 13
for
shell.)
Hence if the iiS-coiiplmg
states of 3t
and
<?
arfe
taken as the same linear
combinations of the correlated zero-order states, the spin-orbit matrices in

LS coupling are the same except for the sign of the spin-orbit parameter
referring to shell SP
Pure almost-ciosed-sheli configurations.

We have shown in I 13 that the electrostatic energies of the two corre
which contain electrons in only one shell
sponding configurations I and
outside of closed sheik, are given by the same formulas. This was first shown
by Heisenberg,* who gave two examples of related configurations of this
type which show an almost constant ratio between the term separations.
3.
1^""*,
These are repeated here, with the positions of the centres of gravity of the
1
terms given in
above the lowest term of the configuration:
cm""
The constancy of the ratios is striking in view of the fact that the individual
the first-order theory particularly well (see 5 7 ).
This constancy would imply that the electrostatic parameters all increased
in constant ratio. Such, for example, is the case for the F*s if the two fields
configurations do not
fit
are both hydrogen-like with different effective nuclear charges.

The matrix of spin-orbit interaction for IOT ~ is just the negative of that
for the correlated states of I . This means that in Riissell-Saimders coupling,
~
the multiplets which are normal in l will be inverted in lm
while those
m
inverted in I will be normal in J ~^. This is a fact well known to empirical
5
7
spectroscopists (cf. Fig. 6 ). For example, the lowest configuration (p ) of
all
the rare-gas ions Sell, An, and KrII consists of an inverted doublet, with
the level of lowest j value highest. Since there is only one spin-orbit para
meter if only one shell is involved, the ratios of the term splittings are the
same for the configurations I and Im~~ , except for this inversion.
From
argument we see that to the first order the terms of the con
have no splitting. This is illustrated by Fig. 5 11 for the case
figuration
of j?3 . We cannot infer from this argument that the whole matrix of spinthis
IOT/2
orbit interaction
zero because for example in
is
LS
coupling, while the
HEDSENBESG, Ann. der Piys. 10, 888 (1931).

f Not observed.
PUBE AIiMOST-CLOSED-SHELL CONMGUBATIONS

correlated states have the
same
SLJM
different phases since they are the

correlated zero-order states.
same
or
SLM8ML
301
values, they
have
linear combinations of different
secular equations for the configuration jp 4 are immediately obtained

The energies are
those given in 3 11 for p 2 by reversing the sign of
The
from
plotted in Fig.
Fig.
4 11
I 13
as a function of x =
The only
plotted at
mean of 1
in
is
Tel 5p
3
determined by making P 1; the
(J 2 =1227
and the mean of *SQ and 3 P^ fit exactly.* The separa
available experimental data are for
x = 0-685
D2 and
between the
K/^2 ^ ê same manner as for p

,
which
= 4203),
J= and between those of J = 2 are absolutely

This
the
theory.
configuration shows slightly the partial,
predicted by
instead of complete, inversion of the triplet which occurs for x I
tions
levels of
>
13
Fig. I .
4.
Theeoiifigurationj>
The rare-gas
ininteimediatecoupliiig.
(x=
spectra.
The simplest and most completely analyzed
of the spectra of atoms con
and isoelectronic ions.

taining almost closed shells are those of the rare gases
The best known and most typical of these spectra is that of Ne I, the observed
1S All
levels of which are plotted in Fig. 2
known levels of the rare
gases are
for
neon, argon, ...; this
3, ...
included in the systems n yfiril,
makes these spectra greatly resemble one-electron spectra so far as the
.
where n = 2,
distribution of configurations is concerned.
The configurations of each series rapidly divide, with increasing n into

2
two groups of levels. That the upper group approaches as limit the P^ level
of the ion and the lower the 2 Pj level shows that the interaction which causes
* TMs
Rev. 35, 1325 (1930). We have
comparison was originally made by GQUDSMIT, Phys.
9
corrected a slight numerical error in his comparison.
302
COKJ1GUBATIONS CONTAINING- ALMOST CLOSED SHELLS
413
this splitting is the spin-orbit interaction of the almost closed shell. This is
a large and practically constant interaction for all configurations, while the
interaction of the valence electron and the electrostatic inter
spin-orbit
action between this electron and the p core rapidly diminish with increasing
13
n. This gives the higher configuration members to the resolution of Fig. 2
the appearance of doublets.
J"
IS
_ n
&*
^
4$
20-
3f
>
Alt
known
/evete included
in the
13
Fig. 2 .
The energy
neon I. (Scale in thousands of cm- 1 Each line

a group of observed levels not resolved* on this scale.)
levels of
represents in general
The quantum number which distinguishes levels belonging rigorously to

one parent level from those belonging to the other is the J value of the p 5
group, which is also the j value of the missing electron. Adding an s electron
THE BABE-GAS SPECTRA
4 1S
(of j
the
= f) to the J = |
J = f level gives
level of
two
schematically as follows
5
j)
states,
gives
J=l,
$?
|
f
1,
8^
2, i.
->
The numbers here indicate /

for p 5 p is given by
>
1;
1,0
->
2.
J=0,
adding it to
This parentage may be shown
states, of
while for pH with
two
303
values. Similarly the jj -coupling parentage
1.0
2, l
2.1
3,2,1,0,
we have
->
2,1-1
1-fM
->
2-1,1,2-1,1-2
J+2,Z + l f
U-l.
in Fig. 2 13 3 the upper group of levels (those having the limit

p^)
while the lower
contains two levels for p^ s and four levels for 5 p, p^ d, .
Hence
. .
$>
5
5
group contains two levels for p s six for p p, and eight for p*d
Xel, the configuration p^s is completely known up to 11$, the
9
J>
For
....
5
j3
are
Wd. Only
cf s are completely analysed up to

complete up to Ip, while the
two levels of the lowest p 5 f have been found.
For those configurations which appear in Fig. 2 13 as doublets without
j>
by parentage is almost exact, i.e. the /

a
value of the p group is very good quantum number. Since this is the case,
12
a glance at the transformations of Table I will show that assignment of L
structure, the characterization

5
and S values is impossible. The states are not, however, in pure jj coupling.
for the j value of the valence electron is in general not a good quantum
shall consider the exact characterization of these states in
number.*
We
the succeeding sections.

A plot for the heavier rare gases would present essentially the same
13
appearance as to the absolute location of configurations as Fig. 2 but with
all n values increased by 1 for argon, 2 for krypton, S for xenon, etc. How
3
1
ever, the parent doublet splitting increases rapidly. It is about 800 cm" for
neon, 1400 for argon, 5400 for krypton, and 10,000 for xenon. The fine
structure of the parts of the doublet increases less rapidly than this, so for
corresponding configurations the division into two groups becomes sharper
down the series of gases. For krypton and xenon even the lowest
as we
go
THs is typical of the coupling near the limit of any series in any atom. The quantum numbers
of the parent levels, in particular the J values of these levels, become asymptotically exact quantum
numbers, because the interactions of the valence electron eventually become very small compared
with the separations of the levels of the ion.
*
413
occur In two well-separated groups, so that the parentage
may
be
s if
we
304
5
p"s
quite accurately assigned.

5.
The
The
9
configurations /*$ and d s.*
secular equations for
reverse the sign of
Hence
we have
this category
The transition from LS
in
l- l s are
for
>
the
same as those of
and
all
3 11 for
observed cases come under
LS coupling a singlet
above an inverted
triplet.
13
to jj coupling for p*s is plotted in Fig. 3 as a fimc-
tionofx-K/01-
\
1S
Fig. 3
The
We see in Fig.
/A
5
configuration |? ^ in intermediate coupling.
2 13 that the
pVs
of neon are
all close
(x=
to the jj -coupling
limit, where they appear as two double levels. This is true of all the rare
gases and rare-gas-like ions. For configurations very close to jj coupling,
the theory predicts that the electrostatic splitting of pis/ be twice that of
p\8, as seen at the right

series
members
of Fig. 3 13 The experimental ratios for the higher

.
of neon are
10s
2-18
1
2-16
1-95
Us
1-43
LUPORTE Mid INGOS, Phys. Rev. 35, 1337 (1930);

CosDoy and SHOETLEY, ibid. 35, 1342 (1930).
THE COXFIGUBATIONS
513
p^s
AND
305
d*s
The agreement Is good up to 10$, which has a zero spitting for 2p| 10* j in
1
TMs term lies very close to
place of an expected splitting of about 8
13
7d on which we shall see in 7 that there is indication of perturbation
2j?|
cm"
1
but a glance at Fig. 8 13 shows a perturbation as large as 8cm" to be quite
is not
unexpected. For argon, the agreement of the Mgher series members
are
ratios
observed
the
good;
5
2p fe
7s
8s
9#
1-19
4-24
-M3
1-29
All other observed p*s configurations have larger values of 1/x and are
13
=
and the mean of the
plotted in Fig. 3 at that x for which the levels J 0, 2,
two
levels of
J= 1 fit the theory exactly. The agreement is In general good.

I/I
7
-8
-3
13
Fig. 4 . Interval ratios for d*8.
we can
-8
-9
-7
-6
-5
-4-
-3
-2
(x=fW7a .)
1
^ a pair of linearly related

interval ratios to compare with the theory. If we take as ordinate
8
1
8
8
and as abscissa (3i|- 3i )/(3^-i- 3i ), we find
( iJ/+1 )/( iJ- J!^1 )
Just as for I s
(cf.
3 11 },
find for
1 "
I"
tliat
2Z+1
0==
The ordinate here
is
14-1
+
J-G -P.
the ratio of the splitting of
1
^"j
to the splitting of
^, which is (1+ l)jl for jj coupling (abscissa 0). The intervals in c? s are
13
compared with the theory in this way in Fig. 4 . These configurations tend
to lie closer to jj coupling, which is on the left in this plot, than to L8
Zj^J
coupling.
The agreement is again seen
to be in general good.
306
COSrMGTJRATIOXS CONTACTING ALMOST CLOSED SHELLS
Unfortunately the spectra of the latter rare earths
is
5 13
not known, so we
have no observed instances of/ 13 &.
The configuration j?5 p in the rare gases.*

The configuration pp contains 1 level with / = 3 3 with J = 2, 4 with J = I
6.
and 2 with J = 0. Hence the diagonalization of the Hamiltonian
for this
configuration involves the solution of a linear, a quadratic, a cubic, and a

quartic equation. These equations contain as parameters six radial integrals.
Given the values of these parameters, the solution of these equations is

rather complicated; to obtain the values of these parameters from the
observed energies and then check the self-consistency of the equations, as
we propose to do, is much more complicated. A simplification valid in many

cases can, however, be made. Just as our consideration of one configuration
at a time is an approximation in which we neglect the interaction between
configurations, so we may use an approximation in which we also neglect
interaction between the groups of levels of the same configuration which
have different levels of the ionic doublet as parents. This latter approximation
13
Fig. 2 ) to be increasingly
configurations in the rare gases.
is
expected
(cf.
good as we go up the
series of
quantum number which distinguishes levels belonging to one

from
those belonging to the other is the J value of the jp 5 group (i.e.,
parent
Since the
the j value of the missing electron), in order to split the secular equation
according to parentage, it is necessary to obtain the matrix of the Hamil
tonian in the jj -coupling scheme.
The matrix of
diagonal
mjj coupling, with the

m
o
&
ff*
is
seen from
<l.i):
and
2 13 to be
elements
n pP np
np
(|,f):
I 10
-jr + K
(fcf):
r+K
-W-
(i,:
T~
n J\
l
C-
In the parentheses are given the j values first of the j? 5 core and then of the
added p electron. The letters a, 6, c, d are introduced as a convenient abbre
viation for the four possible combinations of j values. States labelled by a
and b belong to the lower doublet level, states labelled by c and d to the
higher doublet level of the ion. These elements are independent of the
values of the states.
The matrix
of electrostatic interaction
diagonal elements are given in

this matrix
is
I 13 .
With
diagonal in LS coupling; the

the transformation of Table I 12
is
found injj coupling to have the value

*
SHOBTUEY,
Pfcys.
J and
Rev. 44, 666 (1933).
THE CONFIGFBATION P & p IN THE BABE GASES
6 13
307
3*
-Jt
3a
j
2o
2d
2c
26
Ic
(2)
la
Ic
It*
Oa
(ki
In the notation here used the J value is given as an arable numeral followed
by a letter which specifies the electronic j values according to the scheme ( 1 )
}
The matrices are independent of the value of If.

Adding the spin-orbit interaction (1) and spitting these matrices accord
indicated by the broken lines, we obtain the
ing to the parent j values as
following equations for the energy levels:
Upper l&mis
-J-
(2c)
fl
=s
-j +
Lower Umis
The
six
r-
c
(3)
unprimed levels of this group belong in our approximation rigor

quantum numbers assigned to them. The eigenstates for the
ously to the
primed
levels are linear combinations of those for the corresponding unarbitrarily denote the higher of the two levels by of and
prinied levels.
We
the lower by 6 since the level a would lie above the level b if the electrostatic
interaction vanished.
308
CONMGUBATIONS COOTAINESTG ALMOST CLOSED SHELLS
6 13
should in all cases in which

of one of the parameters,
our approximation is good be obtainable from the splitting of the parent
But this splitting is known from the spectra of
doublet, which is
Xow the value
just f
the ions only to an accuracy of one cm" 1 . Hence we prefer to determine this
parameter, along with the rest, from the data for each configuration, and
then to check it against the doublet splitting of the ion later.
The
six linear expressions in (3) should determine the six parameters.
From the values of the parameters we could then predict absolute positions
Od
On
=g=
-*/c
aocl-
p.-,,
=f=
Oa
-WOrL
;[nujr-
._.
00
^b
30
Zb"
Sp
13
Fig. 5 .
3
\
6p
The configuration p 5 p
in neon.
However, in order to make the values of the para

the small perturbations which must exist both
between the two parts of one configuration and between neighbouring
of the other four
meters
levels.
less sensitive to
configurations, we prefer to fit by least squares the eight quantities obtained

by adding to the above six levels the means of 2a* and 26 and of la and 16 .
This leaves
the separations
2n
26
and la
16 to be absolutely predicted,
THE CONFIGURATION p*p IS THE BABE GASES
6 13
The
results of such a calculation are
shown
309
13
in Fig. 5 13 for neon, Fig. 6
7 13
for argon, and Fig.

for krypton and xenon. In these figures the empirical
value for 16 has been taken as the zero for each configuration; a break in
the wave-number scale indicates the separation between the two groups of
levels.
The
scales are in
1
cm""
inter
question one asks concerns the validity of the neglect of
action between the two groups of levels. If the parameters we have obtained
The
first
AffGON
m/o-
too
5p
13
Fig. 6 .
6p
Tp
8p
The configuration p*p in argon.
are approximately correct, the magnitude of this interaction may be found

by using the interaction elements ofthe matrices (2) according to the secondorder perturbation theory. In this way the largest interaction was found to
have approximately the value represented by the length of the short black
bar drawn underneath each configuration. This gives an idea of the agree
ment
to be expected in the comparison.
310
6 13
CO^FIGUBATIO^S CONTACTING- ALMOST CLOSED SHELLS
Having obtained the values of the parameters, one can readily

coefficients in the
find the
expansion
f
(JaiJa }-f^ (^){J61/a
JalJb
We may call
this
work
is
|(
Ja\Ja
2
)\
j+^JbKJblJb
)
(
).
the purity of the levels Jo! and Jb
r
.
It
is,
so far as
concerned, purely accidental that the electrostatic interaction
- -
-/c_
/c
- /c
-"COQOr-
- la
26*
Iff
ib
S
3
O
woo -
Ib
3
^
Ib
*l
KttrPTQM
1S
Fig. 7 .
6p
I
s
4p np
The configuration j3 5 y
XENON
in
(v.ld)
Sp 6p
krypton and xenon.
between Ic and Id Tanishes identically and hence that these levels are to
our approximation 100 per cent. pure. To the same approximation the
separation between these two levels represents just the doublet splitting of
the np electron in the central field due to the core.
Neon I (c/. Fig. 5 13 ).

The 2j9 5 3p is taken from the work of Inglis and Ginsburg,* who solved the
complete secular equation in LS coupling using the levels of J = 0, 2, 3, and
* iNGiisaud
GIXSBUBG, PBTE. Rev. 43, 194 (1933); corrected by SHOBTLEY, ibid. 47, 295 (1935).
In regard to this configuration see also ROZEXTAL, Zeits. fur Phys. 83, 534 (1933).
THE CONFIGURATION p s p IN THE BARE GASES
6 13
the
mean
/= 1 to obtain the
J= 1 levels to be absolutely predicted.
of the four levels with.
otherwise the
311
parameters, leaving
op is the first of the series to which our approximation is at all

inter
applicable. These levels agree to within the rather large maximum
action between the groups.
For 2p s 6p, the interaction is much smaller and the agreement corre
The
2j?
spondingly better.
TABLE
I 13 .
Xeon 1
2j?
np.
5p
3p
lp
6p
found to agree not at all; the reason is that the lower 8p group
interact strongly
exactly overlaps the upper 7p and is therefore expected to
with it.
1
and the coefficients in (4) are given
The values of the parameters (in
)
I 13 The significance to be attached to these values must be judged
Table
by
13
the value
comparison with Fig. 5 The value of is to be compared with
5
2p Ip
is
cm."
by
1
521 obtained from the parent-doublet splitting of 782 cm"
Argon I
(c/.
13
Fig. 6 ).
The pp configurations of argon remain rather well separated from each

other up to 3p 5 Sp, the last of the series which is completely known. Bp (Oa)
1
falls about 200cm- below 7p (Id) and the lower group of lip falls close to
the upper group of Sp as indicated in the figure. There is definite evidence of
The 5/group overlapping Bp does not have a pronounced
perturbation on
8j>.
effect.
TABLE 2 13 Argon I Sp 5 np.

.
312
CONFIGUEATIONS C0XTAIOTKG ALMOST CLOSED SHELLS
The constants
13
are given in Table 2
The value of
is
to be compared
with the 954 obtained from the ionic doublet splitting of 1431
Krypton I and Xenon I
(c/.
G 13
cm-1
13
Fig. 7 ).
the lowest p*p s of krypton and xenon are amenable to treatment

5
with our approximation. For krypton the 2p 5p and 6p are completely
known. The 2p 5 5jp agrees to within the error of the approximation. The 6p
shows definite evidence of outside perturbation which is undoubtedly
of 7p which lies at only 6200 on the Bp scale,
due to the lower
Even
group
mainly
rather than to the 5f levels plotted.
In xenon only the &p 5 6p is complete. This in the first plot agrees very
the lower
poorly with the calculations. It is definitely perturbed by
a rigorous
make
cannot
one
indicated.
as
which
Although
overlaps
group
one
to
suspects on
configurations,
assignment of these overlapping levels
7j>
this figure with the others that the highest J=l of 7p might
make this re
more
exactly to 6p than the Id assigned to it. If we
belong
a decidedly
which
shows
on
the
the
we
obtain
right,
plot
arrangement,
comparing
better agreement.
TABLE 3 13
Xenon I 5p 5
Jn/p$on 1 4j) np
ftp
The constants for these spectra have the values given in Table 3 13 % for
krypton is to be compared with the 3581 obtained from the ion. The doublet
.
splitting for the
7.
The
xenon ion
is
not known.
5
configuration p d in the rare gases.
The diagonal elements of spin-orbit interaction for p 5 d injj coupling

Ti
p nd
-**
d
are
(i)
(j!i):
where we have introduced a notation a, b, e, d similar to that used for p 5 p.

Here the parameter p which is the same as the former
accomplishes the
,
THE CONFIGITBATION
7 1S
p*d IS
THE BABE GASES
313
into two groups, states characterized by a and

spitting of the configuration
c and d in the upper.
b lying in the lower group, those characterized by
13 to
the
I
of
electrostatic
jj coupling gives
energies
Transforming the
matrices
4a
-2F,
3a
35
- 1 Vio Js -
-1-V21F,
(2)
2d
la,
Id
Justasinthecaseof^wefi
the following formulas to express the twelve
belonging to different parents,
levels in terms of six parameters:
Upper
levels
(3c)=-F + J,+
+
(2c) =-F -f
Lower lends
36
&
L
-^^G s
713
V
CONMCUBATIONS CO^TAIKKG ALMOST CLOSED SHELLS
314
take the six levels given by linear expressions and the means
f
of 3a 36 2a 2b ; and la 16 fitting the six parameters to these nine
2a -2b\ la -lfr separa
the 3a -36
quantities by least squares. This leaves
We
tions to be absolutely predicted.
From
5
the p d
13
Fig. 2
we
see that our approximation should
be applicable to
all
these are known completely from 3d through
of neon
lOi,
I. Since
13 These all
the long series of configurations plotted in Fig. 8
6d and Id, which are
agree within the accuracy of the calculation except
clearlv perturbed. The most of this perturbation is due to the fact that the
1
Id. That there is
upper group of Bd lies only 26 cm- below the lower of
s
we obtain
further perturbation on the upper group of Id is shown by the abnormally

the 7^-electron doublet splitting.
large 2c-M separation, which should be just
This further perturbation may be attributed to the fact that the lower part
5
of 9d lies 100 cm- 1 above the upper of Id and perhaps to the 2_|? 10s which
1
lies about 20 cm" above Id.
13
The constants used in these calculations are given by Table 4 . The value
of t p is to be compared with the 521 obtained from the ionic doublet splitting.
should from the values here given predict a doublet splitting of
We
1
comparison with the observed 782 cm"
Thus the present theory accounts very satisfactorily for the observed
structure of these rare-gas configurations. The perturbations which occur
780-4
0-2
cm"
in
seem to be small except when two configurations of the same series overlap.
Since there can be no spin-orbit perturbation corresponding to the inter
action which splits the configuration into two groups, in estimating the
mutual perturbation of two configurations we should compare the distance
of the nearest levels not to the overall configuration size but to the much
smaller spread of each group of levels. This requires that
Me very close in order appreciably to affect each other.
two configurations
TABLE 413 . Neon I 2p*nd.

3d
4d
5d
6d
Id
Sd
$d
lOd
THE CONFIGURATION
P*d IN
THE RARE GASES
315
316
8 13
Line strengths.
We may say immediately that the line strengths in any coupling scheme,
for transitions between two configurations 3t and 3$ having the same almost
7
closed shell, are the same as for the corresponding configurations J? and J2
8.
To
see this
we note that
in the
nlm^ scheme the quantum numbers in the
almost closed shell cannot change since one of the other electrons definitely
jumps. Since all quantum numbers except those of the almost closed shell
and %, the strengths in this
are the same for the correlated states of
scheme must be the same. The transformation to any other scheme is the
same for the two cases, hence the transition arrays have the same strengths
in
any coupling scheme.

Thus,
5
5
p has the relative strengths given by Table 2
M
3 for p$-+pp* We may obtain the relative strengths in any of
e.g.,
the array
j>
->p
by Table
the rare -gas transition arrays by using the matrices of S^ injj coupling and
the transformations of the eigenfunctions to the jj -coupling scheme as cal
or
culated from the observed energy levels.

discussed briefly at the end of 4 11
The neon array 2p5 3p-+2p 5
3s
is
9.
X-ray spectra.
The characteristic X-ray spectrum of an element is obtained usually by
making that element, or a compound containing it the target in an electron
tube so that it is struck by a beam of electrons which have energies corre
sponding to some tens of kilo volts. Under this excitation the atoms emit
?
radiation of wave-lengths in the general range one to ten Angstrom units.

Owing to their great wave-number, the spectroscopy of these radiations
requires a special experimental technique. This is fully treated for example
in Siegbahn s Spektroskopie der EontgenstraMen.*
confine ourselves
to the theoretical problems involved.
We
These high-frequency radiations arise from transitions between highly

excited states of an atom, corresponding to configurations in which an elec
tron is missing from one of the inner closed shells of the normal atom.
Practically
all
X-ray spectroscopy is done with the emitting matter in the
solid state. It turns out that the interaction energy between the atoms in a
solid is of the order of a Bydberg unit or less, whereas in ordinary X-ray
spectra the energies are a thousand times greater. Thus to a first approxima
tion such interactions may be neglected and the spectra interpreted as due
*
Of course the term X-ray spectra is a practical category

a
from
classification by experimental technique. From the theo
arising
retical standpoint there is no sharp distinction between optical spectra and
X-ray spectra, the two merge in a natural way. But experimentally they
have been quite effectively separated because of the great experimental
to isolated atoms.
* Second
edition, Julius Springer, 1931.
9 1S
X-RAY SPECTBA
317
of working In the soft X-ray region and the extreme ultra-violet,

where the merging takes place.
difficulties
Let us consider one of the heavier elements. Its normal configuration

be l 2 2s 2 2p 6 3s 2 3p 6 3d1 4^ 2
Here the shells have been written down
in order of the firmness with which their electrons are bound to the nucleus,
will
. . ..
that is in the order of the energy levels of the central

3
start-ing
Is shell
field
best suited for the
point of the perturbation theory. If we remove an electron from the

(just where, on the outside, does not matter
and put it on the outside
in the first approximation), the energy of the resulting configuration will be

very high. This level (really a group of levels quite close together if the
structure due to the outer electrons not in closed shells could be observed)
is
called the
K level.
Similarly if an electron is
obtain a level not as high as the
removed from the 2# shell we
K level which
is
called the LI level.
removed from the 2p shell we are left with a p 5 configura

tion plus closed shells. This gives rise to an inverted 2 P term. Evidently
removal of just one electron in this way from any one closed shell will give
rise to a doublet term, the nature of the term being the same as in alkali
spectra but with the doublets inverted. The relation of the conventional
If an electron
X-ray
The
is
level notation to the usual spectroscopic notation is given here;
table ends with
in the
known
Y and
Pm since the 5f and 6d shells do not get filled
elements.
scheme of levels as here presented is only present in

contain
those atoms which
normally all of these closed shells. If we start
with uranium^ where this scheme of levels is folly developed, and consider
in turn elements of lower atomic number, the scheme of X-ray levels becomes
Of course the
foil
for one cannot

simpler through the absence of the later entries in the table,
in
the atom. For
exist
not
does
that
shell
closed
a
from
an
electron
remove
T
the 4d shell is
as
A
to
extends
table
nl
example in Mo (42) the scheme of the
only
half-filled
with this element
(see
I 14 ).
To ignore the structure of the X-ray levels due to the outer electrons not
in closed shells is formally the same as supposing that all electrons present
are in closed shells except for the one shell whose openness is indicated by
the above terminology.
318
COXFIGUKATIO^S CONTAINING ALMOST CLOSED SHELLS
The X-ray Hues are produced by
transitions
level scheme, together with the selection roles
between these
AZ=
I,
9 13
levels.
AJ = 0,
1,
This
gives
an
adequate description of the main features of theobserved spectra. Inaddition

there are observed weaker lines some of which receive their
interpretation
as quadrupole transitions in this level scheme. Other lines, sometimes called
non -diagram lines* because there is no place for them in this
simple level
scheme, are supposed to arise from transitions in a level scheme associated
c
with removal of two electrons from the inner closed
shells.
The K series of lines in an X-ray spectrum is produced by transitions from

the JK level to various lower levels. Thus the
line is produced
by the
transition from the
K^
K level to the m level.
In this transition an electron

jumps from the 2p shell to the la shell, that is, in a transition from the
level to the L level an electron
shell. It is
jumps from the L shell to the
sometimes convenient to focus attention on the
electron or hole in
missing
and to say that in the transition from the

level to the
level
the hole jumps from the K shell to the L shell. The behaviour of a hole is
just
opposite to that of an electron the normal state of the atom corresponds to
the
the hole being in the outermost

far as possible in inner shells.
The
shift
shell,
that
is,
to all of the electrons

being as
of emphasis from electron to hole
is also
useful in discussing
X-ray absorption spectra. We cannot have an absorption line in normal

atoms corresponding to a jump of the hole from the L shell to the
shell
since in normal atoms there is no hole in the L shell. In
normal atoms the
holes are oa the outside so the
X-ray absorption spectrum is produced by
the transition of one of these outer holes to an inside shell.
The absorption
spectrum is thus directly affected by the external structure of the atoms
and the way in which this is influenced
and chemical state
by the
in the
absorbing substance.
Such
details
physical
treat as being outside
we do not
the scope of the theory of atomic

spectra.
The theory of the X-ray energy levels,
neglecting structure due to outer

thus that of a one-electron
spectrum. Suppose our central field
approximation is based on the potential function U(r) and that we define
the effective nuclear
charge Z(r) at distance r by the equation
electrons,
is
TJ/
Z7(r)-
Z(r)e*
LL_.
(1)
The force on an electron at distance r is then
_3U__[Z(r)-rZ
fo
(r)]e*
The quantity in brackets is the effective nuclear

charge for the
electron at distance
r,
which we denote by ZJr):
force
2)
on the
X-RAY SPECTRA
913
319
We know that Zf (r)-+Zas r-^0 and Z (r)-*l for r greater than the
f
the atom.
If
radius of
we
are dealing with a state of the atom in which the radial

probability distribution has a strong maximum at r = 6, it is natural to
suppose, as a rough approximation, that the
of the electron will be
energy
close to that of an electron in the
r (r)
ft
Coulomb
field
M-Z
(4)
(6)
same value and the same slope at r=6 as has the effective
Tor such a field the energy levels are given by 5 5 1 1 including the
and spin effects.
as this has the

field Z7(r).
relativity
The development
Sommerfeid.* It
fits
of such a theory of the X-ray levels is largely due to

the facts surprisingly well. The account of it we have
Just given is incomplete in that it does not say how b is to be chosen. Let us
approach this question through an examination of the empirical data. If
we subtract the rest-energy ^c2 from 5 s 1 1 and measure energy with the Ryd~
berg unit, then eaoh electron has an energy given by that formula on writing
quantum numbers, writing Zf (b) for Z and subtracting the
constant term e 2 Z (b). (Note that BA Z f (b) is
negative this adds to the energy
and hence decreases the absolute value of the negative numerical value of
the energy of each electron. It represents the fact that the inner electrons
in the proper
move in a region where their average potential energy is increased by the

presence of the outer electrons; this is sometimes called external screening.*)
Consequently the atom s energy is increased by the amount of one of these
*
energy levels when the corresponding electron is removed. Accordingly the

X-ray levels should be given by this formula with the opposite sign, which
level highest as it corresponds to removal of the most tightly
puts the
bound
electron.
Let us consider first the

levels. Their values are given for most of the
elements in Table 5 13 as summarized in Siegbahn s book. The energies in
crease roughly as the square of Z., corresponding to the first known empirical
regularity in X-ray spectra, as discovered by Moseley.

oseley found that
the square roots of the line frequencies are approximately a linear function
of Z. If the square root of the frequencies in Rydberg units be plotted against

Z, the curve is strikingly linear, with slope close to I . The departures from the
Moseley law are put strongly in evidence by plotting Vp Z against Z to
remove the main linear trend. Such a plot is given in Fig.

The formula for the K level is, in Rydberg units,
SOMMXEFEID, Atomban nnd SpeidraUinien, 5th
9 13 .
edition.
320
9 13
Expanding the second term in powers of a and taking the square root we find
,
(6)
TABLE 5 13 . The
levels of the
Element
Element
The quantity
(ZZ$
is
v(K)
often called the internal screening constant. It

it is some kind of
average of the
might be expected to be equal to one or two
amount of charge
atoms in Sydberg units.
of the other electrons in the atoms that
is closer to the
nucleus than the 1$ electron. It is harder to say how the term e 2 Z (b)j2RZ hc
f
will depend on Z. If we had n electrons on a sphere of radius the potential
/>
energy of an electron inside the sphere would be ne jp and this term measures
such a potential. As Z increases we have more shells so this term becomes
Swe 2/p
for the shells present.
increases
more rapidly with
On
the whole
it
seems reasonable that
Z than the first power; hence in spite
this
of the
occurring in the denominator this is perhaps a slowly increasing function

of Z, As it appears here with reversed sign it will contribute a slowly decreas
level. The third term is quite definite as soon as we

ing function of Z to the
know the screening constant Z Zf Using Z - Zf = 2, the dotted curve in
.
Fig.
9 1S is
obtained by subtracting off the third term from the experimental

Z. The dotted curve therefore shows the combined effect of
values of Vv^
the external and internal screening. We see that it decreases rapidly except
at the large values of Z, where higher terms in the expansion become
appreciable so that the curve becomes meaningless. The value of Z Z*
obtained
screening
is
is
between one and two, so we see that the

as surmised above.
effect of the internal
321
X-RAY SPECTRA
9 13
In view of the eradeness of the model we do not believe there is much to be

f
gained by attempting to choose Zf and Zi (b) in such a way as to represent the
data best. The formula evidently is in accord with the main facts and that is
all that can be expected.
This same model has also been used to discuss the doublet intervals in
X-ray spectra. Let us consider the L levels. L^ corresponds to removal of
a 2s electron while
L u and LITI correspond to the inverted 2 P given by the
5
2p configuration. For the same n the hydrogenic s states penetrate more
closely to the nucleus than the -p states. Therefore the 2s electron is more
L
tightly bound than the 2p so the L^ level will be higher than Lu or IU
The difference will be partly due to the difference in screening and partly to
the spin-relativity effects. For a reason that has since become meaningless,
13
Fig. 9 . Plot of Vv(K)
the
Z against Z to show departure from the Moseley law for the

K. levels of the elements.
LT and Z*n levels^ and in general the pairs of levels of the same n and J
3
but different L, are commonly called irregular doublets.

If we calculate the doublet interval between LU and ijjj by means of the
for the two
hydrogenic formula, assuming the same values ofZ (b) and Zf (b)
Hence we
Z*
as
should
interval
the
that
vary
levels, we find to the first order
.
can conveniently test the formula by calculating empirical values of Z Zf

from the doublet intervals and seeing if they have reasonable magnitudes.
The quantity
calculated in this
way
is
sometimes called the screening
constant for doublet separation. Such comparisons have been
made by
COSJIGUBA HONS CG3TTAIKOG AIMOST CLOSED SHELLS
322
9 13
Sommerfeld and others,* and have shown that the simple screened hydrogenie model gives a good account of the main facts.
10.
Mme
strengths in X-ray spectra.
As emphasized in the preceding section, the X-ray energy-level scheme of

the atoms has the structure of a one-electron spectrum. From 8 13 we see
that the theory of the relative line strengths in a doublet spectrum is applic
2
able here. For example, the relative strengths of Ka 2 (1 2
P^) and
2
2
5
1
2
will
be
as
the
to
results of 6 This is in
K&i (1 5|->2 Pj)
according
S^2
accord with the experimental data:
(2)
DTASE and STEKSTBOM,

DUASE and PATTEKSQN,
(3)
SIEGBAHN and Â&SL, Ann. der Phys. 71, 187
(1)
Proe. Xat. Acad. Sci 6 S 477 (1920).

ibid. 8,
Likewise in the
K series
doublet. Allison
and Armstrongf found
Mo and Cu
the lines
respectively for this
85 (1922),
K$i and
(1923).
form the same kind of

intensity ratios of 2-1 and 2-0 in
doublet. They also measured the relative
j83
intensities of the doublets arising in the hole transitions
and
ls-*3p
(JT j81? Kf}%)
(Kfts ). The ratio of the intensity of the first of these doublets

to that of the second increases with decreasing atomic number:
l*->-4p
Element
Intensity ratio
Strength ratio
(74)
2-3
2-6
Mo
(42)
7-7
8-3
Cu
(29)
41-0
42-0
The third column gives the ratio of the strengths as obtained by
dividing the
intensity ratio by the fourth power of the frequency ratio. The great ratio
obtained in the lighter elements is due to the greater difference in the relative
on the 3p and 4p radial function for small Z. No calcula

have been published.
In the L series we have the possibility of studying the relative intensities
in a 2 P-^ 2D multiple! For the hole transition 2p
3d! the lines and measured
effect of screening
tions of these relative strengths
->
relative intensities are
SOMMEBFELD, Atombau und SpekiTaMinim, 5th edition, p. 297;

PATMHG and GOUDSMIT, TJk Structure of Line Spectra, Chapter s;
PAUUHG and SmESMAy, Zeits. fur Kristallographie, 81, I (1932).
f AXUSOH and AEMSTEOKG, Froc. Nat. Acad. Sci 11, 563 (1925) Phys. Rev. 26, 701, 714 (1925),
;
USTE STRENGTHS IN X-RAY SPECTRA

L a*
Element
Mo
(42)
Eli (46)
Pd
Ag
As the
two
first
(46)
(47)
lines
0-130
0-133
0-119
0-122
have the same
0-33
0-24
0-20
0-17
323
(0-28)
(0-20)
(0-16)
(0-14)
initial state, their relative intensities
should be the ratio of the relative strengths multiplied by the fourth power
of the frequency ratio. The relative strengths from 2 9 are as 9: 1 or 1: 0*111.
The 2D interval in the
M levels
is
so small here that the frequency correction
negligible so the relative intensities are the same as relative strengths and
are seen to be in good agreement with the theoretical ratio Assuming natural
is
excitation the relative strengths of the first and third line are obtained from
the relative intensities by dividing by the fourth power of the frequency
ratio.
less
The
results are given in parentheses. All the experimental values are

is probably due to the lack of
;
than half of the theoretical value f this
natural excitation, the higher level 2 P| being considerably less excited than
the lower 2 Ps level. Jonsson* has estimated the correction for excitation and
has found values which check
much
better with the theoretical 9:1:5
strength ratio.
Wentzelf has given a theoretical calculation of the relative strengths of

the different multiplets originating in the L levels based on use of approxi
mate eigenfunctions. Experimental measurements bearing on this question

have been made by Allison. J
it.
X-ray
satellites.
In addition to the
lines
which are accounted
for
by
transitions in the
doublet energy-level diagram of the preceding sections there are observed

a number of other lines known as satellites or non-diagram lines. These
are close to the
more prominent lines which are accounted for by the discus
sion just given. Such satellites were first studied by Siegbahn when observ
then they have been the object
ing the La. lines of Sn 5 Ag, and Mo. Since
of considerable study.
The doublet structure
so simple that it is evidently necessary to con

sider the energy diagram corresponding to more than one non-closed shell
to get any more detail. Several possibilities are open. WentzelH suggested
double ionization (two holes) of the Is shell as an initial state of the satellites
of the
is
K lines. Then he supposed the satellites to be emitted when one of the

* JOSTSSON, Zeits. fur Fkys. 46, 383 (1928).
f WBHTZEI*, Naturwiss. 14, 621 (1928).
J AMJSON, Phys. Rev. 34, 7, 176 (1929).
SIEGBAHH, Ark. 1 Mat. Astrom. Fys. 14, No. 9 (1920).
WiarrzEL, Ann. der Pkys. 66, 437 (1921); 73, 647 (1924).
fl
C02STFIGUBATI05S COOTAIXIXG
324
ALMOST CLOSED SHELLS
II 13
For initial states of some of the lines lie

L
with two holes in the K shell and one in the
suggested triple ionization
13
from Fig. 10
sheE. The relations proposed by Wentzel are best considered
holes
jumped
to
an outer
shell.
which
is
a schematic level diagram.
column
The first column shows the ordinary level diagram. In the second
The
energy of
we have this diagram extended for the case of two holes.
that for the
removal K\ of the second Is electron will be greater than K,
not as much as the K of the element whose
but
to
screening
owing
to L as
atomic number is one greater. Similar remarks hold with regard
screened
is
because 2p
compared with L. Xow L will be greater than
first,
L"
-Is 22p
-Is 2
-IsZp*
-Is2p
0*3
-Is
-2p*
-2p
-Normal
1S
Fig. 10 .
by the
more by
is
than by another 2p, and so on. Hence one readily concludes
(, - %)
that
where x n
The
is
satellites. The levels are labelled

configuration of the missing electrons.
WentzeFs interpretation of the
<
(*5
- *s)
<
~ **)
wave number of the corresponding

found by Wetterblad* are given below:
written briefly for the
values, in
Rydberg
units,
Na(ll)
Mg(12)
Al(13)
Si (14)
0-52
0-57
0-65
0-64
0-67
0-76
0-71
0-83
0-91
0-94
0-76
0*83
line.
Moreover, since Is is so much closer to the nucleus than 2p absence of a Is

electron is almost like a full unit increase in the nuclear charge in its effect
on the others, so (a e a4 ) of one element should be nearly the (a3 oc^ of the
next element. This is the case here.
y
WBTTEBBLAD,
Zeite. fur Phys. 42, 611 (1927).
X-EAY SATELLITES
325
The main
objection to WentzePs interpretation concerns the question of

these
producing
multiply-ionized states. The mean life is so short that
excitation by two successive singly -ionizing impacts cannot be effective;
therefore the electron impacts
more than twice the usual
must produce double excitation. TMs requires
oc
critical potential to produce the line
4
Backlin* studied this point in Al and found that
a4 appeared at definitely
lower voltages than expected on this hypothesis. It does appear however
that the satellites require higher voltages (by about 25 per cent.) on the tube
.
for their appearance than the related
main lines.
?
,s
fs2$
IP
-Jp Is2p
Is
ip
lip
2s2p
4V
-2s
0%
~2p
(Wolft)
Normal
satellites.
The Langer-Wolfe interpretation of the
(Configoration labels are for the missing electrons.)
1S
Fig. I1 .
In WentzePs work no account was taken of the complex structure arising

from electrostatic interaction of the holes, although he mentioned it as a
possibility in later papers.f The importance of this structure was empha
sized
by Langer. J On account of the
existence of such structure the
satellites can be provided for without using the triple ionization process or
shell. Langer s level scheme and assignment
the double ionization of the
of the satellites is shown in Pig. 1 1 13 This scheme is consistent with the small
r, Zeits. fur Fhys. 27, 30 (1924).

f WEOTSEL, Zeits. fur Piiys. Si, 445 (1925); 34, 730 (1925).
$ LASTGEB (abstract only), Hiys. Rev. 37, 457 (1931).
COKMGUBATIO^S COKTAIKma ALMOST CLOSED SHELLS
326
excess in the critical potentials of the satellites over those of the main lines.
Similar views have been put forward by Druyvesteyn.^ Some calculations
of the terms arising from these configurations have been made by Wolfef
using a Hartree field for potassium. His assignments of the lines differs
from Langer s and is shown in the lower row of the figure. The calculations
show that the
structural energies are of the right order of
magnitude to
but the detailed assignments are quite uncertain.
It seems quite likely that the X-ray satellites are to be accounted for by
double excitation levels, but there still remains much to be done before the
account for the
satellites
interpretations are satisfactory.
MehtmyerJ has developed a somewhat different view of the origin of

the satellites in which two-electron jumps as well as double-hole levels are
employed. As, generally speaking, two-electron jumps are weaker than
single-electron transitions it seems that such processes should not be con
s
sidered unless
it is
found that the one-electron jump picture
is
inadequate.
Another possible source of more detailed structure is the open shells in

the outer structure of the atom, as was noted by Coster and Bruyvesteyn.
The
relation of the filling of electron shells to the periodic system of the

13
the doublet structure of
I 14 As mentioned in 9
is discussed in
elements
the X-ray levels depends on a model in which there

shells. For elements in wMch an outer shell is being
incomplete^ so there should be a
is
but one hole in the
filled
the outer shell is
complex structure due to the interaction of
shell. Such interactions are pre

somewhat
smaller
than
one
sumably
Bydberg unit and in case of ordinary
resolution give rise to broadened lines more than to observable structure.
the outer electrons with the open inner
Structure of this kind, if appreciable, we should expect to be sensitive to the

state of chemical combination of the element in the source. These ideas
have had practically no detailed development thus

*
far.
BErnneSTEYN, BIss. Groningen, 1928; Zeits. fur Phys. 43, 707 (1927).
WOLFE, Phys. Rev. 43, 221 (1933); see also KEKXABD sad BAMBEEG,
Pihys.
Bey. 46, 1040
(1934).
f RICHTMYBE, PhiL Mag.
0,
64 (1028); Phys. Bev. 34, 574 (1929); J. franklin
(1929).
COSTEE and DBUYVJSHTKXK,

COSTEE,
tirid.
Zeits. fir Pliys. 40,
45, 797 (1927).
765 (1927);
Insfc.
208, 325
CHAPTER XIV
CENTRAL FIELDS
1.
We
Tlae periodic system.
have now to consider some of the
relations
between the spectra of

5
different atoms. Chemists long ago discovered the great powerof llendelejeff s
periodic system of elements as a coordinator of their empirical knowledge of

the chemical properties of atoms. As these properties are all, in the last
analysis, related to the energy states of the atoms,
we
expect the
periodic system to play an important part in coordinating the different
atomic spectra. More than that, as Bohr showed, the theory of atomic
shall
its picture of the arrangement of electrons in closed shells

a
clear
understanding of how it is that element with similar
provides
properties recur periodically in the list of elements. In this section we shall,
then, consider the empirical data on the gross structure of the atomic spectra
structure through
in its relation to the chemist s periodic table.

In Fig. i 14 are shown the energies of the highest and lowest observed levels
of the principal configurations of the first eighteen elements, plotted up

from the lowest level of the atom, the even configurations on the left and
the odd on the right. This takes us past the initial hydrogen and helium
pair through the second Mendelejefi period of eight. We have included,
in addition to all the low-lying configurations, at least the lowest con
The lowest level of the ion of each

element is indicated by cross-hatching, and configurations of the ion
which may have importance as parents are shown in broken lines.
figuration of each observed series.
We see that in normal helium the second electron has gone into a la state
and that the next excited state is very high much higher than the whole
ionization energy of hydrogen. Moreover we see that Is 2s is definitely
lower in energy than Is 2p. This we shall connect with the fact that the 2s
comes nearer to the nucleus than the 2p. At any rate, the fact of 2s
being lower than 2p in helium leds us to expect that 2s will be the normal
state of lithium. In lithium we see that 2p is considerably higher than 2s;
this points to 2$ 2 as the normal state of beryllium.* Now the 2s shell is
closed. The fact that in beryllium the 2p energy is considerably less than the
3& leads us to expect boron to have 2p for its lowest configuration.f In the
next six elements the normal configuration is in each ease obtained by
* That He
is, and Be is not, a rare gas, although both have closed 4 shells for their ground states,
state
is
connected with the extreme difference in energy necessary to excite one of the elections in this
closed shell.
f For brevity we omit mention of the closed shells already present

course the complete designation is I*8 2* 2p.
and speak of 2p when of
-|
*
<
~+<t*
3?
"+**
Q?
c/5
\g *
J01 .?
tso
TO
Ki
sr-
-*
<
&
L-.....-.J
03
! !
J *! i
o
>-
.1
"A
h
fo
U**
1111
I
<D
OQ
Ui
^5
s^^W
ssr-
dp-
tfi-j?^
i
L.
J^
ri
iJJ
î
^LJ
g^^^PS
^ 8
-^^
^^
.1
of 4=
CENTRAL MELDS
330
I 14
successive addition of a 2p electron to the normal configuration of the pre

ceding atom. At neon the process is brought to a stop by the operation of the
is the maximum number of electrons in any p shell.
As we go from boron to neon we observe that the interval between the normal
Pauli principle, as sis
configuration and the lowest excited configuration progressively increases.

is
little ground for ambiguity here as to prediction of the order
There
of the low configurations in the element (Z+l) from a knowledge of the

spectrum of element Z.
Thus we see that the first period of eight is connected with the sequence
of normal configurations obtained by successive addition of two 2s electrons
followed by six 2p electrons. Looking at the spectrum of neon we see that
5
2p %s lies below
2j?
3p,
with the other configurations definitely higher. This
us expect a repetition of the previous development in the next eight

by successive addition of two 3$ electrons followed by six 3p
what happens. The
electrons. This is in fact
from Li to Xe and
from
of the second period
close parallelism of the period
Xa to A is very striking,
is
although our know-
far from complete, especially in regard to
phosphorus and sulphur.

What next ? We notice in argon that 4s is lower than 3d!, the departure of
the effective field from the Coulomb form being sufficiently great to upset
the order of n values as it occurs in hydrogen. Looking back we see that the
as far back as neon where
inversion occurred at
p 5 4s
is
definitely
below j) 3d. Even in nitrogen the known terms of p* 4s are below p* M, but
not so much but that the complete configurations probably overlap. This
by the time we reach sulphur,

and
chlorine,
argon it is quite definite. Therefore we expect the next element,
inversion has not occurred in boron; but
potassium, to have 4a for its lowest configuration.

The eighteen elements from potassium to krypton are shown in Fig. 2 14
Here we have a situation which is quite different from that in the first two
.
periods. There is a close competition going on between configurations of the

n 2
n ^ l s and d n
for the distinction of contributing the normal
type d $ 3 d
As aE of these are even configurations their proximity

undoubtedly gives rise to strong interactions so that the assignment of a
definite term to a definite configuration has much less
meaning than in the
state of the atom.
2
2
cases heretofore considered. In scandium, the
configurations ds d $ and P
2 2
3
are rather widely separated, but in titanium the
configurations d s and d s
,
overlap considerably; similarly for succeeding elements. The competition

especially close in nickel where, although the lowest level is d 8 * 28 ^, the
term intervals are such that ds s tz !3 is considerably higher than d9 s 3 3
is
which
lies
1
only 205cm- above the normal
spectra, while extensive
level. Our knowledge of these

owing to their great complexity, is by no means
I 14
THE PERIODIC SYSTEM
331
+
complete, but generally it may be said that the configurations d n s 2 and dn l $
so
that the question of which configuration gives the
overlap completely
normal
state, as far as it has significance, is as
much dependent on
the
internal interactions in a configuration as on the central field.

striking
property of these atoms is their almost constant ionization energy as con
trasted with the
upward trend in the ionization energy during the filling of

The
lowest
levels of the doubly ionized atom are indicated in
p
parentheses for some of these elements.
Copper is somewhat akin to the alkalis in that its level system is mainly
a
shell.
due to various states occupied by one electron outside the closed $d shell.
9 2
configurations like d # and cPsp are still in evidence relatively low,
whereas the corresponding jp 5 2 and p^sp are not observed in alkali spectra.
In copper and zinc we note that 4p is definitely below 5s; this leads us to
But
<s
expect that the 4p shell will be filled in the next six elements. TMs is the
case, although our knowledge of some of these spectra is quite fragmentary.
We thus arrive at another inert gas, krypton, the eighteenth element after
the last preceding inert gas, argon. In these eighteen elements the 4s, 3d and
4p
shells
have been
built in.
TMs
is
known
as the first long period in the
periodic table.
There are
many variants of the periodic system employed by chemists: a

14
Let us note how their treat
fairly common form is reproduced in Table I
.
ment of the
long period differs from that of the short periods and the
relation of this to the facts of spectroscopy. In Groups I and II, potassium
first
and calcium are clearly alkali and alkaline earth, but from Group III to
Group VII the elements do not have such a close relationship to the ele
ments of the first two periods: the metal, manganese, lias certainly little
in common with the halogens, fluorine and chlorine. That is correlated here
with the fact that a d shell rather than a p shell is being filled. As the d shell
is longer than a p shell, they have to introduce a Group VIII and lump three
elements into it. Then they omit an entry from Group and put the remain
ing elements into the second or odd* series of the long period. With these
an a and a p shell are being filled, so we expect them to show more chemical
similarity with elements of the first and second short periods than did the
4
part of the long period. TMs is certainly the case for the elements
bromine and krypton at the end, but not for copper and zinc at the
first
beginning of this group. In the table the rare gases are counted in at the
beginning of new periods, whereas we prefer to count them in at the end:
one example of the kind of arbitrariness present in the empirical

led some to suggest that the elements should be distri
a
buted on spiral curve drawn on a cylinder to emphasize the arbitrariness
in the choice of beginning and end of a period.
this is just
system, wMch has
2
!
>
JD
co
o
;
S
S
lr
i
S
c.s
o S
S*
i
period
"
hort
"8
1a^
E,i1j
131
?,
fl
ill
1.1
*
123
s
035
** 115
S S
-Ii
j~i*
^*
Ae-3O
"^
r"
id!
Ill
s S
si
!&
.s
Illj
I
I
s
Is
U
2
c-^
2
-5
S s
-=)
.>
"
c?
r*T
u lUfrf,
LJ<J
Lll
U*
--------
.r
-I
335
THE PEBIODIO SYSTEM
14
It is not necessary to discuss the remainder of the system in such detail.

The second long period builds in the 5# 5 4d, and
shells, ending with xenon.
5j>
This is illustrated in Fig. 3 M where the similarities with the
first
long period
are evident.
With the third long period greater complications set in (Fig. 414 ). It starts
off like the others with the alkali, caesium, and the alkaline earth, barium,
closing the 6s shell. After this the 65 and 5d electrons start a long-period
competition as before, but now the 4f states enter the competition. The com
has
plexity that presents itself is so great that thus far very little progress
electrons
been made in disentangling the spectra of the elements which have/
the chemists do not
have such an easy time with them either. There are fourteen places, from
cerium (Z = 5B) to lutecium (2=71), where the 4f shell is being filled.
From Hf 72 to Pt 78 the filling of tie &d shell, interrupted by the rare earths,
continues. In the process the 6s electrons have been displaced so that they
must come in again at gold (79) and mercury (80) to give elements chemically
resembling copper and zinc, having respectively one and two s electrons
built on just completed d shells. Finally in the elements thallium to radon
in their normal configuration.* These are the rare earths
the 6p shell is built in. In these spectra, due to large spin-orbit interactions,
there is a pronounced tendency tojj coupling. The radioactive metals follow
almost nothing is known about their energy-level schemes.
For many purposes it is convenient to have a periodic table of the elements
:
which shows the electron configuration of the normal state of the atomTable 2 14 is arranged in this way, and provides a convenient summary of
the main facts which have been discussed in this section.
2.
The
The
consider
is
method of Fermi-Thomas. f
method of getting an approximate central field which we shall
that of Fermi and Thomas, who treated the cluster of electrons
statistical
first
mechanics as modified to
include the Paul exclusion principle, that is, by the Fermi-Dirac statistics.
In this method the phase space associated with the positional coordinates
around a nucleus by the methods of
statistical
3
and momenta of each electron is divided into elements of volume A and the
exclusion principle is brought in by the restriction of the number of electrons
in each cell to two (corresponding to the two possible spin orientations).
IS pis the resultant momentum of an electron, the volume of phase space
and
corresponding to electrons having less than this resultant momentum
3
number
the
us
Let
dv.
is
located in volume dv of physical space
suppose
fTrp
* See the recent

iper by IIJBEBTSON, Phys. Rev. 47, 370 (1935).
Zelts. fur Pfcys. 48, 73; 49, 550 (192S);
f FBBIO, Bend. Lincei 6, 602 (1027); 7, 342, 726 (1928);
THOMAS, Proc. Camb. Phil. Soc. 23, 542 (1927).
2 14
CENTRAL FIELDS
336
We
assume that tke kinetic energy of

electrons at each point is as small as possible consistent with the density
at each point. This means that j? the maximum value of p at each point,
of electrons in unit Tolume
is n.
47TD 3
is
connected with n through the relation
==, so that
2-^
of the electrons occurring at

gives the maximum value of the kinetic energy
a place where the density is n. If 9 is the electrostatic potential at the point
in question, counted from zero at infinity, the potential energy of the elec
tron is -9. If jpg; 2/*<e9 the kinetic energy of the electron is insufficient
to permit the escape of that electron to infinity. As this must be the case in
9
the atom 9
we
write
2/t
where O Q is a positive constant. ^ Since for a neutral atom 9 tends to zero at

= for regions of space where (2) leads to a
infinity we also assume that n
maximum
kinetic
energy. Using (1) and (2) we see that the electron
negative
is
a
related
to the potential o by the equation
at
point
density
[2^16(9
90)3
(3)
op-
Treating the charge distribution as approximately continuous we may

to obtain the basic
use the Poisson equation of electrostatics,
47r/>,
equation of the statistical method:

i_(4)
JftCTj]*,
*j/fr~
where
==
7
?~9o*
This equation has been used for an extension of the statistical method to
molecules! and to metals. J We confine our attention to its application to
the atomic problem. For the atom we need a solution of (4) which is such
that
Ze r as r
and
as r->oc and which has spherical symmetry.
o~>
->0,
ro->
atom we may put

coordinate form of the Laplacian and writing
It turns out that for the neutral
,_^
op
= 0. Using
the polar
"88534a
* The introduction of o is due to Grra and

f
PEDEELS, Phys. Rev, 37, 217 (1931).
f HCSD, Zeita. fur Phya. 77, 12 (1932).
J LENS ABB-JONES and WOODS, Proc. Roy. Soc.
120, 727 (1928).
STATISTICAL
we
METHOD OF EEBMI-THO&AS
337
find that x( x ) satisfies a differential equation free from universal con
stants:
We need a solution of this such, that x = 1 at r =

solution has been
this
at
= oo. Such a
14
computed numerically* and is given in Table 3
TABLE 3 14
With
and ^ =
The function
%(x).
approximate potential function Fermi was able to make a
striMng calculation of the number of s p d electrons in the atom as a func

tion of atomic number. At a point where the maximum value of p is p Q the
actual momentum vectors are equally likely to be anywhere inside the
Hence the number of elec
centred momentum-space sphere of radius p
}
trons in unit volume for which the component of p perpendicular to the

radius vector has a value between p^ and p^ -f dpA is
we introduce the orbital angular momentum L=pj r. Hence the whole

number for which L lies between L and L+dL at all places in the shell
if
between
and r 4- dr is
Replacing p* by its value 2/*ecp and integrating over those values of r for
which the radical has real values, we obtain for the number of electrons for
which
L lies between L and L -f dL

LdL.
* A short table was

given by Fermi. The values given here are taken from the table calculated
witn the differential analyser by BUSH and GJLLDWELL, Phys. Rev. 38, 1898 (1931).
CENTRAL FIELDS
338
We shall suppose in this semi-classical

L= (1+ 1) and dL^ftm order to find
914
argument that we hare to put
ft
dr,
(7)
T
j(iJ) is
A
the number of electrons whose orbital angular momentum
quantum number is I in the atom of atomic number Z. Written in this form
where
the expression has a simple interpretation h~ l times the integral is the value
of the radial quantum number, calculated classically, for an electron of zero
:
energy and orbital angular
momentum
(l
+ %)fi moving in
a central
field
of
2 (21 +1) is the number of electrons in

potential energy,
). The factor
a closed shell and the other factor is the number of closed shells occurring.
e<?(r
The integral in this expression has to be evaluated numerically. Fermi has

done this: the results are compared with the empirical data in Fig. o 14
where the smooth curves give the values of the theoretical N^Z) and the
M
Fig. 5 .
Numbers of s p d9 and/ electrons in the normal state as a function of Z,

as given by the statistical-field method.
9
irregular curves give the empirical values as discussed in the preceding

section. The agreement is extremely good.
Another question which has been studied by the statistical method is that
of the total iordzation energy, that is the energy necessary to remove all of
the electrons from a neutral atom. This was calculated
by Milne and Baker.*
The method is to calculate the electrostatic energy of the charge distribution.

The total energy is half of this since by the virial theorem the
average
minus half of the average potential energy in a

system of
particles interacting by the Coulomb law. The result is that the total energy
of the noimal state of the neutral atom of atomic number Z is
kinetic energy is
W = 20-83 Z^ electron volte.

An examination of the experimental data has been made by
*
MEUSE, Proe. Camb. PHI. Soc. 23, 794 (1927);

BAKZB, Ptye. Rev. 36, 630 (1930).
t Toroe, Phya. Rev. 34, 1228 (1929).
Youngf for the
THE STATISTICAL METHOD OF FEBMI-THOMAS

339
elements lip to neon. He finds good agreement with the exponent of Z but
2 14
that the data are better represented
by the formula
IF = 15-6 J^ electron volts.
Milne gave the theoretical coefficient the value 17 because of numerical

inaccuracies; this accidentally gave a much better agreement.
The Wentzel-BrHlouin-Kramers approximation.*
3.
An important method of finding approximately the eigenvalues and eigenfdnctions of one-dimensional Schrodinger equations was developed by
Wentzel, Brillouin, and Kramersf independently. It is of interest because
it exhibits the connection with the older quantization rules of Bohr and
Sommerfeld. Let us first consider motion along a single coordinate under the
potential energy U(x). For convenience we suppose this has but one mini
mum and increases monotonically to the right and left of it.

We enter the wave equation
f + ^(E-UW = Q
= ef?/s
with the assumption

so that 9 has to satisfy
If fL were zero this
classical
-f
2/z, (
(I)
(2)
U E) = ij&p*
(3)
would be the same as the Hamilton-Jacobi equation of
mechanics for the action function S(x). This suggests writing
Equating successive powers of H to zero in the equation
for 9,
we
find
8 *+2p(U-E) = 0,
for the first
apJS ^S*
two approximations. These lead to
Hence in this approximation the wave function
a=
where
The approximation
infinite
*
there.
is
In
is
evidently not valid where
this
given by
approximation $aâ ~|2/i(J?
17
as
$a becomes
Z7)|~*;
inside the
A good discussion of this topic iias been given by HIRTREE, Proc. Manchester Lit.
Soc. 77, 91 (1933).
t WXNTZKL, Zeits.
ffip
Phys. 38, 518 (1926);
BEHXOUET, Jour, de Physique, 7, 353 (1926);

KSAMEES, Zeite. ffip Phys. 39, 828 (1926).
and
Phil.
3 14
CESTBAL FIELDS
340
where E>U, this is equal to the classical proof being between x and x + dx, the latter being taken proportional
to the fraction of the whole period spent in going through this interval.
classical range of motion,
us investigate the behaviour near the classical turning points of

the motion, that is near the roots x ^ and x 2 (x^ x 2 ) of the equation
E - U(x) = 0. In the neighbourhood of these two points
Xext
let
<
where x and
x.2
are both positive. Writing

g = x\(x-Xi)
or
a|(a? 2 -a?),
the wave equation near either of the turning points assumes the canonical
form
f-f^=0.
The solution of this which

Kramers took in the form
is finite
everywhere
f
(6)
is
an Airy integral* which
exp(# + #)*,
(7)
TTJ
rJ
comes from infinity along the ray t=re"i7r^ and goes to

3
infinitv along the ray 1= re"^- The function may be expressed in terms of
Bessel functions of order one-third, so that the numerical tables and func
where the path
tional properties given in
Watson are
applicable
../""
>
Gr>0)
(8)
(f<0)|
V3
Applying the asymptotic formulas given in Watson (Chapter vn), we find

that for large
a,(f)~f*oos(|-frr),
(f
>
0)
Comparing this with the approximation valid away from the turning points
we
see that
aj
(a;
-*!))! cos/J
Of
|*a:s
adx +
COS
|
("~J
v$
\a\dx\.
-J-JjexplI
\ Jx*
\cr\*
WATSON, Tkwry of Bessd Functions,
p. 188.
314
THE WBHTZEL-BRILLOIIOr-KBAaiEES APFEOXIMATION
These forms will join up smoothly in the region x I

(4) only if the arguments of the two cosine functions
multiple of IT. This implies that
<x<x%
with the solution
differ
by an
integral
f
J
or, in terms of the classical
341
momentum
at the point x } that

(9)
This is exactly of the form of the Bohr-Sommerfeld quantization rule except

for the occurrence of (% + J) in place of n. It thus appears that the use of
quantum numbers increased by J in the old rule gives an approximation to
the energy levels as determined by quantum mechanics. With the energy

level so determined the approximate eigenftmetion (not normalized) is
given by
-^ exp
das I,
(x~x
(10)
\J
<!x:
adx\
This completes the discussion for a single rectilinear coordinate where the
4-00.
boundary conditions require that $ be finite in the range
For atomic theory we are more interested in the equation for the radial
to -f oo and the boundary
eigenfonctions, where the coordinate range is
conditions .5(0) = J2(oo) = 0. This was treated by Kramers as follows:
oo<^<
The equation
is
Near the nucleus U(r)
is
Ze*]r + G, so the equation has the
form
If we neglect the constant term in the coefficient of R, this equation

solved in terms of Bessel functions to give
B(r)~A
may be
(12)
CENTRAL FIELDS
342
the asymptotic expression for which Is
Let us suppose that we have to take
/2Za
a 2* 2
_ -
as the argument of the complex exponential in
becomes
VSZrja +
2Jfc
(2).
To terms of order r*1 this
- irk,
VBZr/a
so with this accuracy the approximation in the region of classical motion
becomes
C os
Comparing
see that they will agree
J1
V8Zr/a
\2jZa/
this with the
2F
V8Z^+ -7= -- TlA-ffrjL

L
T
\i
JZ~(-1-|
asymptotic expression for the Bessel function we
if
At the outer bound of the classical motion in r one may join the approxi
mation smoothly by means of the function as before. Thus Kramers is led
to the conclusion that one should use the classical quantum condition in
}
o>
the form
^r = (n,+i)A=(-Z-i)A,
(13)
where
That
2
(|-f-
1)
one uses half-integral values of the radial phase integral and uses
instead of 1(1+ 1) in the term representing the effect of the orbital
is,
angular momentum on the radial motion.

This result is the basis of a method for finding an effective central field
U(r) which was developed before quantum mechanics by Fues and Hartree.*
This early work is characterized by the use of integral quantum numbers.

This made the work easier because for the circular orbits the radial integral
was zero, so their empirical energies gave several points on the curve directly.
The first calculations with the half-quantum numbers demanded by the

new theory were made by Sugiura and Urey.f They worked by a graphical
process. Prokofjew J made calculations for sodium by a numerical method.
*
Fmss,
Zeits. fur
Phys. 11, 364; 12, 1; 13, 211 (1922); 21, 265 (1924);
HAETEEE, Proc. Camb. Phil. Soc. 21, 625 (1923).

I SUGIUSA and UBEY, KgL Danske Vid. Selskab, Math.
STCICBA, PHI. Mag. 4, 495 (1927).
J PROKOFJZW, Zeits. fur Phys. 48, 255 (1929).
fys. 7,
No. 13 (1926);
THE WENTZEL-BEILLOtnir-KBAMEBS APPROXIMATION
3U
If we write
the
quantum
-2?=-^,
condition
U(p)= ~Q(p)lp*,
343
p=r/a,
is
where the limits of integration bound the range in which the integrand is
Since the /terms
real. For large p we have Q(p) ^=p and for all values Q(p)
of sodium are essentially hydrogenic, to a good approximation Q(p)p
for p>6-125. For smaller values Prokofjew assumes Q(p) in the form
2
ap -f ftp -f y choosing two coefficients in such a way as to join smoothly to
>p.
PRQKQFJEW
0-5
1-0
3-0
2-5
2-0
1-5
3-5
40
r/a
14
field in sodium.
Fig. 6 . Effective nuclear charge for potential
the analytic expression used for larger p values and the third so as to fit the
With this form for
empirical energy values to the radial quantum condition.
can be evaluated exactly. Prokofjew gives the following
the
Q(p)
integral
table as the result for Q(p) determined this way:
Range of p
0-01
0-15
1-00
1-55
3-30
6*74
to 0-01
0-15
1-00
1-55
3*30
6-74
oo
Q(p}
2
Up
p -f 11-53 p- 0-00264
s
p + 4-46 p-f 0-5275
1-508 p 2 - 4-236 p 4-4-876
0-1196P 2 4- 0-2072p4-l-3I9
2
0-0005p -f 0-9933p+ 0-0222
-24-4
- 2-84
3 14
CEKTBAL FIELDS
344
In Fig. 6 14
is
plotted the effect-ire nuclear charge for potential
as a function of p for this field as determined
Fermi-Thomas potential
an electron
for
by Prokofjew and
in sodium.
We see
also for the

that the latter
does not approach the asymptotic value sufficiently rapidly in this case.
4.
Numerical integration
of the radial equation.
Given an approximate potential function, the eigenvalues and eigenmay be found by a numerical integration of the radial equation.
We shall not attempt a thorough discussion of this matter as it involves
fknctions
varied questions of technique.
A trial eigenvalue is chosen;

of the form
jy, y(x
A sketch of the method follows.

the differential equation to be solved
^ ^
g(
wtere
is
then
g ( x )=f(x)y
and/(x) is a known function (D=*djdx). By assuming a power series for y(x)

near the origin one may find the numerical values of the coefficients and
compute a table of values ofy(x) for small values ofx. In terms of the back
ward difference operator V the finite difference tables
We
for interval h in x.
now try to calculate the next value of
in
terms
of
either
known
or almost known.
have D*y = g(x)
V^y
quantities
so
may be set up
We
Since
V = e~m we may
express the operator
D~ 2V 2
in powers of
V by
the series
Of order
from which follows
2
involves the known g%
Referring to the above scheme, we see that V
and the unknown ? 2g4 and V 4^5 These must be estimated from the
trend in the table and a trial value of V 2
calculated. From this we get
Vy4 and y4 which
and so permits the extension of the table of

2
differences ofg by one more row. The
computed value of V g% serves to verify
2
the estimate already made. If
they do not agree the computed V #4 is used
to get a new value of V^/ and the
n
process repeated. With experience the
first estimate will be
right more than half the time and it ought never to be
gives us
4M
SOBlEBICAIi INTEGRATION OF
RADIAL EQUATION
necessary to make more than one correction to the estimate of

this way the table is successively extended.
345
V 2^.
In
The process
is carried out until the behaviour at large values of a: is

In general the trial value of the parameter will not be an eigen
value so the process must be repeated with a different trial. By comparing
the behaviour of the two trial functions one attempts to interpolate or
extrapolate to a considerably improved trial value. No general rules for this
discernible.
have been given.
Normal state of helium.

We wish now to consider another type of approximate method of obtain
5.
ing atomic energy Ievels namely the variation method, and to illustrate its
use with reference to the very exact calculations of the normal state of
3
helium made by Hylleraas.* The key idea consists in recognizing that the
Schrodinger equation is the differential equation corresponding to a mini
mum problem in the calculus of variations. From the general principles of

Chapter n we know that $Htfr will be greater than the least allowed value of
for all except the eigen-i/r belonging to this least allowed value. Moreover
it is easy to see that ijtH^ is stationary for ^ in the neighbourhood of each of
ift
and that the stationary value is equal to the corresponding

allowed value. In a general way it is a consequence of this stationary pro
perty that one can often calculate a fairly accurate value of the allowed
the eigen-^ s of
energy from an evaluation of fH$ from a relatively inaccurate 0.

From the standpoint of the calculus of variations the Schrodinger equa
tion is simply the Euler equation which expresses the fact that the normal
minimizes $H$, Instead of seeking ^ through its Schrodinger equa

tion, we may use the Ritz method.j In this one tries to represent the
unknown function ^ as belonging to a class of functions, say ^(a,6 Cj ...)
depending on several parameters a b,c } .... Then one chooses the para
is normalized. This
is a minimum when
meters in such a way that
ized
i/t
<f>
$H<j>
implies a set of equations

o,
...
(i)
which determine the minimizing values of the parameters. Evidently if the

b,c ...) happens to contain the exact eigenfunction for some set
family
of values of the parameters, the exact energy and eigenfunction will be found
<f>(a,
in this way.
that
That would be a lucky accident. Generally the method furnishes
member of the family which gives the closest fit to the true eigenfunc
found this way will surely be greater than
Moreover the value of
tion ^.
$H<f>
HYLLERAAS, Zelts. fur Phys. 48, 469 (1928); 54, 347 (1929); Norske VId. A&ads.
Mat. KL 1932, No. 6.
mid angew. Math. 135,
f RITZ, Gesammdte Werle, Paris, 1911, p. 192; or Jour, furreiae
*
Skrifter,
1 (1907).
CENTRAL MELDS
346
514
or equal to the least allowed value of H. This one-sided approach of the

successive approximations to the true value is a useful feature of this
method. It gives us a semi-test of the theory in comparison with experiment,

since if we ever find a value lower than the observed lowest energy level the
theory must be wrong.
Thfe method of dealing approximately with variation problems was
developed for classical problems, especially in the theory of elasticity, and
has proved to be a valuable tool in that field. It was first used in quantum
mechanics by Keilner* to calculate the normal state of helium; the problem
has been later treated with much greater precision by Hyleraas. It is
evident that the success of the method depends principally on a fortunate
choice of the family of functions $(a, b, c, ...) on which the approximation is
based. In the work on helium it has been found convenient to use the dis
tances rj r2 and rls and the three Euler angles specifying the orientation of
,
the plane determined by the two electrons and the nucleus as coordinates.
For the normal state, $ does not depend on these angles.
The simplest possible
choice of eigenftinction
is
obtained by normalizing
the product of two hydrogenic Is eigenfonctions (r is in atomic units).

Here tie integrations are simply performed and give, as the lowest energy
level,
(Z.-^f atomic units, and the eigenfonction (not normalized)
This gives 1*605 Rydberg units instead of the observed value 1-810 for the
ionization potential of helium. For the ionization potential as a function of
Z we find
(in
Rydberg
units)
He
H-{1)
-0-055
Calculated
(2)
1-695
1-810
6-4
Observed
% error
Li+(3)
Be++(4)
5-4A5
5-560
11-195
11-307
2-1
1-0
This shows that the simple screened hydrogenic eigenfunction gives quite
good values as Z increases corresponding to the fact that the interaction of
5
each electron with the nucleus increases as

energy increases as Z.
Using atomic units for length,
let
Z2 while the electron interaction
us write
Then one parameter may be handled at
once, in
any trial function, namely
the one which merely changes the scale of lengths in the trial function. For
the eneigy of a function $(ks,
Jen) we find
KELLXEB,
Zeits. fur Phys. 44, 91 (1927).
FORMAL STATE OF HELIUM
5 14
where,
if
we
rs
/co
N=\Jo
let
9=
<j>(s,
t,
347
u),
ru
ds\ du\ dtu(s*-l

Jo
Jo
The condition 8 Eftk =
leads to k = Z//2J/ 5 so
This fixes the value of k in terms of the other parameters and gives us the
minimum value of E with regard to variation of the scale of lengths in
<f>.
For $ Hylleraas assumes

7i,,m=
Odd powers of cannot occur since must be symmetric in the two electrons.
The results of such calculations are extremely interesting and show that
quantum mechanics is undoubtedly able to give the correct ionization
<f>
the neglected spin-relativity

potential to within quantities of the order of
and nuclear kinetic energy terms. In the third approximation Hylleraas
=- 1-80488 RAc,
while in the sixth approximation
(f>(s,
ty
u)
= e~&(l + 0-0972^ + 0-0097*2 - 0-0277
E=- 1-80648 RA.

An eighth approximation led to 1-80749 RJic = 198322 cm-
for the ionization
value of 198298 6 cm- 1 It

potential as compared with an experimental
will be noticed that the theoretical value exceeds the experimental value by
24 em-1 , which appears to contradict our statement that the Ritz method
is due to the neglect
always gives too high an energy value. The discrepancy
and
to
nucleus
helium
of the finite mass of the
relativity effects. When these
cm- 1 which is in agree
198307
are included the theoretical value becomes
ment with the experimental value. This is an important accomplishment of
mechanics since it is known that the older quantized-orbit theories
-
quantum
led definitely to the wrong value.

In order to calculate the ionization potentials of the ions iso-electronic
modified the details of the use of the Ritz method so as
with He
3
Hylleraas*
HYIXEKAAS,
Zeits. fur Pfays. 65,
209 (1930).
CENTRAL FIELDS
348
514
to obtain the ionization potential as a series in Z~~l The result, in Rydberg

iinitSj neglecting the finite mass of the nucleus and the relativity correction
.
is
given by the formula
J?= Z2 -
The data
for comparison of theory
from which
of the
1-252? + 0-31488
it is
seen that there
known lonization
and experiment are given
in Table 4 14
agreement within the experimental error
potentials.
TABLE 4M.
6.
is
- 0-017522-1 4- CK)0548- 2
lomization potentials of two-electron atoms*
Excited levels in helium.
The
variation method is peculiarly adapted to calculation of the lowest

energy level of a system, for this is the level which corresponds to an
absolute minimion of $H\[t. The next higher level is characterized as the
minimum value f$H$ under the auxiliary condition that

to the normal state, and the third level is that which makes
be orthogonal
fHift a minim-pm
under the two auxiliary conditions that ^ be orthogonal to each of the two
states of lower energy. In approximate work these
auxiliary conditions make
iff
trouble, since in the absence of exact knowledge of the ^r s for the lower
states al one can do is require
orthogonality to the approximately known
lower states. This inaccurately applied

auxiliary condition in general intro
duces a large error in the calculations by
permitting the trial to contain a
component along the true $ of the normal state, with a resultant sagging
of the minimum value
$H$ below the correct value.
Nevertheless the existence of quantum numbers that are exact
permits
the exact fulfilment of the
conditions in some cases. This
e/r
is
auxiliary
exemplified by the calculations of Hylleraas and TJndheimf on the la2s*S
level of helium. In so far as He has exact
Russell-Saunders coupling this ^
must be an antisymmetric fonetion of the
position of the two electrons. This
condition alone serves to make it
exactly orthogonal to the normal
state,
which
is
a symmetrical function of the

position coordinates. Using hydro
v- *** W17
(1930), finds that the electron affinity of the hydrogen atom
-
is
roughly 0-6 electron volt (5000 cm"1 }.

f HYLLERAAS and UXBHMM, Zeits. fur Phys. 65, 759 (1930).
EXCITED LEVELS IN HELKTM
6 14
genie elgenfnnctions with.
,2
= 2 for the
and
2 $ becomes
Is state
Zl
349
for the 2s state, the
eigenfunction (not normalized) for
$~e& [(& - 2}sinhft- It coshf q,
(1)
where s and t are the elliptic coordinates of the preceding section, measured
in atomic units. The value offH^ for this, relative to the normal state of the
He ion, is 0-2469 Rhc which is considerably higher than the experimental
0-35048 R he. Preserving the requirement of antisymmetry, one
value,
may generalize
$~e-
(1)
to
*[(Gi +
C& +
C>
+ CsU8)smhct + t(Cs-r 0coshc]
(2)
and choose the parameters to minimize $H*ft. In this way Hylieraas and
Undheim found 0-35044 Rhc for the energy of this level, agreeing with the
experimental value to about 0-01 per cent.
To illustrate the error involved in failing to apply the auxiliary condition
to (I)
properly, they calculated $Hi(s for the symmetric function analogous
for the 2 1 S level. The value, - 0-3422 Rhc, lies considerably below the
0-29196 because of the impurity of the normal state which is
contained in the trial wave function. They tried a more general form
4
observed
iff
and applied the orthogonality condition by a special device. The minimizing

conditions expressed by 5 14 1 are satisfied by several values of the energy.
Instead of taMng the least root, they chose the parameters in such a way as
to minimize the next to the least value of the energy. This gave a value
0*28980 Rhc, that is, 0-7 per cent, higher than the observed value, A more
the accuracy is to note that it gives the
significant way of estimating
0-25 Rhc to within 5-4 per cent, of
departure from the hydrogenic value
the actual departure from this value.
to the normal state and the 2 S
Orthogonality of the 2 P terms relative
the
obtained
levels may be accurately
proper use of the dependence on
by
3
level
the angular coordinates of these eigenfuactions. Calculation of the 2
Is
for
functions
the
hydrogenie
was made in this way by Breit,* by using
and 2p in the proper combination, with the effective nuclear charges Z&
- 0-2616 Rhc, the
and Z%p as the parameters. The value obtained was
- 0-2664 Rhc, an error of 1-8 per cent, on the term
experimental value being
value or 29 per cent, on the departure from the hydrogenic value. Eckartf
1
term, obtaining -0-245
made the corresponding calculation for the 2
instead of the observed value
- 0-2475. He made calculations with hydxo-
383 (1930).
* BRETT,
Phys. Rev. 35, 569 (1930); 36,
878 (1930).
f ECKABT, Phys. Bev. 36,
CENTRAL FIELDS
350
6 14
P and 2 3P levels in Li II as
3
21
genie functions and variable Z for the 2 $.
well as He I. The results were, in Rydberg units:
He I
Li II
Evidently the method could be applied to find the lowest levels in any
the trial functions
series of configurations because of the orthogonality of

in the angle variables.
Historically the first helium calculations are contained in a paper by

Heisenberg.* It was in this paper that he laid the foundations of the theory
of atomic spectra by showing the importance of the symmetry of the eigenfunctions. Heisenberg s calculations were confined to the Is nl configura
tions with !=0. He assumed that the Is eigenfunction was almost entirely
confined to smaller values of the radius than the nl eigenfunction, so he
could take for the potential energy function of each electron
r
(r<r
An electron inside r moves in the full field while one outside r moves in
screened field of effective nuclear charge,
If rQ is taken rather larger than a/2
Zl.
and at the same time smaller than the
values of r for which R(nl) becomes appreciable, one may use as a good
approximation the hydrogenie eigenfunction with full Z for the Is electron
and the corresponding function with nuclear charge
(Zl)
for the nl
electron. This implies a correction to the energy of

(l*|ff(r) Is)
j
+ (ril\v(r) - e2 /r\nl)
do not correspond to U(r) but to purely

Coulomb fields instead. In calculating the first order perturbation one needs
the direct and exchange integrals of the Coulomb interaction as well as the
to allow for tbe fact that they
perturbation energy corresponding to the fact that the true central field
the field,
Ze^/r . This provides the additional terms
is
(l%Jfe /r12 l*7iZ) (I$nl\e*jrl2 \nl Is) (ls\v(r)[ Is) (nl\v(r)\nl)

with the result that the whole energy becomes independent of the choice of
j
TQJ to the
first
approximation. It
is
The third and fourth terms are nearly equal and opposite: on account of the
spherical symmetry of the Is eigenfunction its field is like that of a charge
*
HEISEJCBEBG, Zeits. fur Pfcys. 39, 499 (1927).
EXCITED LEYELS IN
6 14
HEUUM
351
concentrated at the centre. These two terms fail to compensate exactly

because some of the B(nl) charge distribution penetrates to smaller radii
than some of the Is charge.
The numerical value of the

units) as calculated
third
by Heisenberg
and fourth terms together
2p
Rydberg
3j?
(2)
- 0-0020
- 0-OW70
Li+(3)
-0-0098
-0-0032
He
(in
is
6-7
x 1CT5
These values do not give a good approximation to the location, of the mean
of the singlet and triplet terms because we have neglected polarization.
When the outer electron is present its field acts on the inner electron to
distort its wave function and give an induced dipole moment which gives
rise to an altered interaction between the two electrons. This polarization
effect can, in principle, be calculated with the aid of the second order per
turbation theory. The most careful calculations of the polarization effect
are those given by Bethe.*
The separation of the singlet and triplet states is given by calculating the
exchange integral (lsnl\e

(Rvdberg units):
7.
Normal
]T 12 \nlls).
The values given by Heisenberg
are
states of first-row atoms.
In this section
we shall review the work that has been done on the use of
the variation method for finding the energies of the normal states of the
atoms in the first row of the periodic table. For the Li iso-electronic sequence
He works with an antisymmetric
the most complete work is that of Wilson, f
6
combination of one-electron functions as introduced in 3 6. For the radial
functions he uses the forms
and oc.
so that the trial function contains four variable parameters, $, 17,
out
directly
The minimizing with regard to the scale factor | can be carried
5 14 Because of the complicated way
to
in which a, r) and appear in the expression for fH$ it was not possible
a
so
the
of
values
grapMcalsolve analytically for the best
parameters
numerical method was used.
since
here plays the same role as k in
* BETHE, Handbnch der Physik 24/1, 2d ed., 339 (1933),

f WILSON, J. Chem. Pkys. 1, 210 (1933).
CENTBAi FIELDS
352
7 14
To compare with earlier work by Eckart and by Guiflemin and Zener* it

should be noted that Eckart s M(2s) is obtained from Wilson s by the special
= 7| = a, so lie has just two parameters to vary, while Guillemin
ization.
and Zener s is equivalent to writing = ij in Wilson s, leaving them with
three parameters to vary. Comparing with Hylleraas precise calculations
on helium we note the absence of any dependence of the trial functions on
the mutual electronic distances, r j this would make the numerical work
extremely difficult. The results show that the first term in JR(2$) is much
larger than the second, so Zener and Slaterf have pointed out that good
approximations can be obtained by simply omitting the second term in
;
R(2s).
The results of all such

Table
514
summarized by Wilson are given in

where the energies are expressed in atomic units (twice the Rydcalculations as
berg unit). The calculated values of the ionization potential are the difference
between the calculated normal state of the three-electron problem and that
of the corresponding two-electron problem.
TABLE 514 En&gies
.
The
S(ls) functions used in
of the Li iso-electronic sequence*
work are hydrogenic in character

The R(%s) functions as used in the varia
all this
except for the variable scale factor.
tion problem are not directly comparable as they stand, since they are not
*
ECILLRT, Pnys. Rer. 36, 878 (1930);

GuTLLTwra- and ZESTER, Zeits. fur Pkys. 61, 199 (1930).
f ZIHEB, Phys. Rev. 86* 51 (1930);
SLATES, ibid. 38, 57 (1930).
NORMAL STATES OF FIRST-BOW ATOMS
353
orthogonal to the R(ls) functions and therefore cannot be regarded as

belonging to the same effective central field. One can correct this by con
= E(2s) + ft M(ls) as the proper 2s radial function, choosing

sidering JJ(2s)
that
so
B(2s) is orthogonal to J?(ls). This does not alter the value of the
/?
T in determinant form, since this change in R(2a) corresponds to adding to

one row of the determinant a constant multiple of another row.
0-36
1
Fig. 7 *.
Comparison of different approximations to R*(2s) for Li I.

B GuiHemin-Zener;
D, hydrogen -like.
C, Slater;
A, Wilson;
The four
different R(2s) functions obtained, by the fonr processes are

14
2
(2s) is plotted after the E(2s) has
compared by Wilson in Fig. 7 3 where
been made orthogonal to the E(ls). It will be noticed that this process has
form for B(2s) although it was node-less in

the original form. The figure shows a striking similarity of the R(2s) for the
three forms which gave best agreement with experiment.
Zener (loc. cit.) has applied the variation method to the normal states of
introduced a node into Slater
the other atoms in the
first
row from Be to F.
8.
8 14
CENTRAL FIELDS
354
Hartree s self-consistent
fields.
different applications of the variation method, that of seeking the

be classified according
make $H\f* stationary for normalized $,
^f
Eitz method a trial
the
In
for
admitted
to the type of trial function
0.
The
may
s which
function depending on several parameters is used. This makes the value of

and the stationary values are sought by
$Htfi depend on these parameters
other extreme we have the case in which
At
the
methods.
calculus
ordinary
the variation of $ is wholly unrestricted; then the Euler variation equation

is just the Schrodinger equation of the problem. In between these extremes
we may admit trial functions of special forms and determine their detailed
character
by the variation
principle.
The most
useful of these
is
a method
devised by Hartree* on physical considerations recognition of its connection

with the variation principle is due to Slater and to Fock.f
;
Although we know from Chapter vi that $ must be antisymmetric in
all
us neglect that requirement for simplicity and write for the

the simple form (2 6 4)
to
the
complete set
# belonging
the electrons,
let
instead of the properly antisymmetrized function (3 6). Let us suppose each

of the w s normalized, although as yet they are not specified to be solutions
I 6 and 2 6 If we use
of a particular central-field problem as they were in
6
the approximate Hamiltonian 1 1 ? neglecting spin-orbit interaction, the
value of JrHJt is
*
.
E
i>j**i
6 6 and 7 6
the calculation of the matrix components being exactly as in
only simpler because there are no permutations and hence no exchange
The dependence of E on any

more
shown
explicitly by writing
integrals.
particular factor of #, say ^(a*),
fr
JJ
+ terms
^S
e2
is
independent of %(#*).
I )
This dependence on uâl)

exactly of the form of the dependence for a
one-electron problem in which the electron having the quantum numbers
is
a 1 moves in an
effective field for
which the potential energy function
is
OK) /%2
-
This is true for each of the
u s, since they occur symmetrically in
HAETBEE, Proc. Camb. Phil. Soc. 24, 89

f SLATER, Phys. Rev. 35, 210 (1930);
FOCK,
Zeits. fur
Phys. 61, 126 (1930).
(1028).
*/f.
Hence if
8 14
HABTBEE
SELF-COKSISTENT FIELBS
355
E stationary by varying each, uâ ) independently, each must

a
one-electron
satisfy
Schrodinger equation in which the effective field is
that of the nucleus plus the classical potential energy field due to the other
we make
electrons calculated according to their quantum-mechanical probability

distributions.
In this
way
the variation principle leads us to a set of simultaneous
But they
are very difficult, being non-linear and

the
integro-differential through
appearance of u(a^) u(a?) under the integral
in
Hartree
s
(2).
procedure is to solve this system of equations by
sign
equations for the
-w s.
numerical integration using a successive approximation process.

Physically the set of equations appears to be very reasonable. Each
electron actually does move in a field that is due to the fixed nucleus and to
the action of the other electrons. It was
by such an argument that Hartree
up his equations rather than by way of the variation principle. This

field, or set of fields, Hartree calls sdf-consistent in the sense that their own
eigenfunctions are consistent with the potential field from which they are
set
determined. The method is evidently applicable in principle to the calcula

tion of other problems where several electrons are involved. For example,
Brillouin* has developed it for use in the theory of metals.
Generally the potential in
(2)
due to the other electrons
will not
be
spherically symmetric owing to a departure from spherical symmetry of

J
?
To consider such departures from
the charge distribution
")^(a
S(a
").
symmetry in the atomic problem would be very difficult and

not correspond to an improvement in the final result. For
would
probably
spherical
that reason Hartree does not use the actual non-spherically-symmetric

potential field defined by (2) but the result of averaging this field over all
moving in an effective central field.

To keep this symmetrizing process within the form of a variation problem
we may assume for each ufa*) that it is of the form of the wave function of a
directions. This results in each electron s
central-field problem,
where
S(Z%ftf)
mf),
^(aO = -J2(^S(Z
T
(3)
the normalized spherical harmonic appro

numbers a1 This means in the variation pro
is
(Z%f)(mf)
priate to the set of
quantum
blem that only the
radial function -B/r
*.
is
subject to variation. In {!
the
of the electrons with quantum numbers

integral representing the interaction
of
and
o? is
now of the form

J(a\ of) = f ftij(a*) u^)
JJ
% Ufa?) %(aO dvfa
* BRILLOUEST, Jour, de Physique, 3, 373 (1932),

23-2
CENTRAL MELDS
356
8 14
we have
8 6 There
whose evaluation we have considered in
seen that
J(a\ of) = S a*(V ml V m/) F*(nfl* nW),

9
where the a
are the coefficients in Table 2 6
and the
FB
are the radial
6
integrals defined in S loa. From this expression it is clear that the effective
1
central field for the electron with quantum numbers a becomes
F(a<,
r)
-f
7"*
This
is
a central
field,
feV
c=M)
fa*
S S a*(Fml Vm\) -^ 3*(n?V) dr,.

k
J
j
(4)
?">
to be sure,
but
its
value depends on the
of the
electron in question. This is an undesirable feature since we have to use the

various complete sets
belonging to a configuration to calculate the details
of the level structure of that configuration, as in Chapter vn and later
k
structure involves integrals of the type
anyway, it is
not likely that we gain much in accuracy by dealing with different central
fields for each ml value. For any fixed value of Vw$ 9 the a s of Table 2 6 have
chapters*
As that
the property that the average over m] vanishes for k ^

and is equal to
unity for & = 0. Therefore if we use for each value of m\ not the field given by
(4) but the value on averaging over the values of m\^ we have the same result
as if we had taken the non-central field (2)
of r
t ,
that
is,
F(a*,fi)-This
Ti
(>
+S
1
and averaged it over all directions
f fi 2
\XWV)dr,.
JT
3
>
with which the Hartree method actually works.

By
symmetry, the system of equa
tions becomes a system of ordinary
integro-differential equations for the
is
the
field
this operation of introducing central
radial functions, S(af) 9 instead of a system of

partial integro-differential
4
equations , each in three independent variables for the
functions, ^(a
Let us next see what value this method gives for the
energy of the atom.
").
The characteristic value e(af) of the equation for the iih electron is not the
work necessary to remove it to infinity with no kinetic energy, for in the
actual process of removal its contribution to the effective field in which the
others move is removed which in turn alters the characteristic values of all
the other electrons. Nevertheless it turns out

empirically, as we shall see,
that these characteristic values for the
deep-lying electrons do provide good
approximations to the X-ray term values, so

terms are quite small.
it
must be that the correction
The value of E from (I) is given now by using our results

wave functions. It is evident that we shall have
for the self-
consistent
JS^S *(*)- S
(a*af|e*/r|a
f
a>),
(5)
HABTHEE
SELE-CONSISTE^T FIELDS
since the interaction of each other electron with, the
i th
357
electron
is
counted
once in the equations by which each e(a*) is determined. On summing over

in the first terra therefore the interaction of the electrons is counted twice
whereas in (1) it should be counted but once, so the interaction energy has
to be subtracted from Se(a*) to allow for this.* Suppose we compare this
1 -electron problem in which the electron a* is
for the
with the energy
removed:
where, however, the values of the terms occurring are calculated from the
altered self-consistent field (indicated
removal of this electron is
(W - jB) = -
(oJ)
by
).
+ If V) - (0] +
[
The
increase in energy for
o?|eVr<,|a*
of)
which exhibits the terms responsible for the difference between (af) and the
ionization energy of the i ih electron. Roughly, the removal from the equa
tion for each other electron of the positive potential due to the i th electron
will make each
(af)
(a *) so the first sum is negative. In so far as we can
r
<
calculate
(d)
*(a?)
by taking the average over the state u(aj ) of the change
in potential energy without allowing for the change in u(a?) itself, the term
f
7
= (aW|e2/r^ j^W) so that the first two sums cancel each other
(a *)
(a?)
approximately. In this same approximation the third sum vanishes, for its
whole value arises from the changes in the functions u(a?) and u(a k ). This
shows us roughly why the quantity (a{ ) is equal to the energy of removal of
_the
fr
th
electron^
Before proceeding to discussion of the numerical results of the Hartree

method we observe that each electron^ field depends on the particular
configuration being considered. As a consequence the $ s obtained for the
different configurations are not orthogonal and so cannot be made the basis
for a calculation in which configuration interaction is taken into account by
the perturbation theory. On the other hand the various complete sets
belonging to the same configuration do have orthogonal eigenfunctions
since their orthogonality depends simply on the spherical harmonics and
the radial functions found by the
self-consistent field method to calculate the term structure of any one con
on the basis of
figuration by the methods developed in previous chapters
the spin functions. Therefore
a more
strict central-field
*
we can use
approximation as set up in Chapter
CAUOT, Proc. Camb. PhiL
vi.
Soc. 24, 89, 111, 426 (1928); 25, 225, 310 (1929).
9 14
CEOTRAI, FIELDS
358
9.
Survey
of consistent-field results.
In this section we shall give an account of the numerical results which

have been obtained by the application of Hartree s method to particular
atoms. The computing labour involved for accurate results is considerable
but the results so far obtained indicate that the method is an excellent one
for obtaining good radial functions for a first approximation to the atomic
structure problem and as the starting point for more exact calculations.
Therefore a number of workers are now engaged in a programme of these
calculations, the chief activity being that of Hartree and his students at
Manchester. He is constructing a differential analyser of the Bush* type
which will be largely employed on this work and will soon add considerably
to our knowledge of this type of atomic wave function. Owing to the
rapidity with which these calculations are being made this section of our
book will probably be out of date soon after publication, so the reader who
wants to keep abreast of the field will need to follow the current literature.
According to a summary prepared by Hartree for the summer spectroscopic conference at the Massachusetts Institute of Technology in 1933, the
work may be divided into two classes according to the standard of accuracy
with which the self-consistency condition is fulfilled. In class A the maximum
two
or three units in the fourth decimal place. Class

includes
calculations of a decidedly lower standard of accuracy. Of class
the
error
is
following were complete at that time:

Oxygen
Xeon I
I,
H,
III,
IV
Sodium. II
Copper II and chlorine negative ion

Potassium II and caesium II
Silicon
HABTBEE and BLACK, Proc. Boy. Soc. A139, 311

McDouGAii, unpublished.
McDouGALL, unpublished.
HABTBEE, Proc. Roy. Soc. A141, 282 (1933).
HABTBEE, ibid. A143, 506 (1934).
McDouGALL, ibid. A138, 550 (1932).
(1933).
Mm
In addition to these he reported that various persons associated with

were at work on Be, A1IV, A, and Ebll and that he would shortly under
take the negative fluorine ion and Calll.
In class B the following were complete:
Helium I (and two-electron
ions)
Lithium I
Be
II
and Be
Nel, 3? I, and
Boron I
and
I, also
B III
fluorine negative ion
CALDWELL (Mass. Tech.) and unpublished work by Hartree.

HABGBEAVES, Proc, Camb.
HABTBEE, unpublished.
PML
BBOWIST, Phys. Bey. 44, 214 (1933).
BBOTO, BABTLETT, and Dusra,
in progress were carbon
Soc. 25, 75 (1928).
by Torrance
ibid. 44,
296
(Princeton)!, silicon
(1933).
by Lindsay
(Brown), and titanium V, silver II, and mercury III by Hartree.J

It would take too much space to make a full
presentation of all details.
Therefore we shall tell only of the programme of Hartree, in collaboration
*
J. Franklin Inst. 212, 447

(1931).
completed: TOBBASTCE, Phys. Eev. 46, 388 (1934).
J For mercury see HABTBEE, Phys. Rev. 46, 738 (1934).
BFSH,
Xow
SURVEY OP CONSISTENT-FIELD RESULTS
9 14
359
with Comrie and Sadler, and the work of Hartree and Black on oxygen, as
work.* After
representing some of the most interesting possibilities of the
and
standardizing the computing method, Hartree arranged with Comrie
Sadler to carry out the detailed computations. The procedure, as Hartreef
explains,
was
as follows:
the earlier work, I make revised estimates of the contributions to

the field from the various electron groups, and the computing work carried
out professionally is concerned with the calculation of wave functions in the
"From
field so constructed,
wave
functions.
from these
regarded as given, and of charge distributions
reference I will call these calculations the standard
For
Unless estimates of the contributions to the field have been

calculations are not yet
unusually fortunate, the results of these standard
near enough to the self-consistent field to be quite satisfactory, but they
should be near enough for the effect of any variation of the estimates to be
treated as a first order variation from the results of the standard calculations.
calculations.
further revision of the estimate is made if necessary and the variations

in the wave functions, etc. due to the variations in the estimates from those
used in the standard calculations are calculated, and the variations of wave
"A
functions, etc.
added to the
results of the standard calculations; the varia
shorter and
tions are so small that this variation calculation is very much
of the
easier than the main calculation. If necessary, further revisions
of the
results
the
from
variations
estimates are made and corresponding
a
thoroughly satisfactory
standard calculations are worked out, until
approximation to the self-consistent

We now turn to the results for
field is obtained."
01"
electron in this state lying within r
this being
is
given
by
a special case of a more general set of radial functions
Zk (n^n^;r)^r-k P*
which
R(nl) is the normalized

then the total charge due to an
and On*. If
radial function of the self-consistent field,
i*S(nl*)R(nW)dri,
nuclear charge for field

arise in other parts of the work. The effective
13 is
then, at radius r,
(which we called Zf in 9 )
strength
the configuration under consideration.

being over the nl values of
r from unit
The quantity [1 ZQ (nl, nl r)] is the contribution to Z at radius
the
sum
of the
^^..^S^
terns
an extended discussion of the calculation of the
8oa A138, 5oO
from the Hartree field for Si V, see MoDoraAi*, Proc. Koy.
Soc. A141, 282 (1933).
Proc.
HABTBEE,
Eoy.
f
* For
Si IV starting
ftmction
J Haxtree s notation for this
is
fc
(V, nl
\r).
(1932).
CENTRAL FIELDS
300
It is in
charge on the nucleus as screened by an electron in an nl state.
the
on
Z
that
work
self-consistent
to
field is
contributions
terms of these
usually done, and the extent of the agreement between estimated and
"
calculated contributions usually expressed."

In the standard calculations the contributions to
were evaluated to
by one or two units. The standard of

self-consistency was such that the difference between the estimated and
calculated contributions to Z for each whole group of electrons of the same
three decimals with the last uncertain
value did not exceed 0-02 at any radius. Hartree says:

stable in the sense that
"The contributions to Z may be called
*
if
the
estimated contributions from any group is increased over a range of r, the

effect of this is to decrease the calculated contribution from this group (and
from others
also); for
an increase of Z means that the attractive
electron towards the nucleus
increased, the
is
field on an
wave functions of electrons
become more compact, and the proportion of the electron

tribution lying inside any given radius is increased
in the field
dis
14
Fig. 8 . Illustrating over-stability in calculation of Hartree field for Cu II. Full line
shows change of estimated contribution to Z; broken curve shows the consequent
change in the calculated contribution to Z.
but the groups of the outermost shell, it is usually if not always

the case that the change in the calculated contributions is smaller than the
change in the estimated contributions. If this were so for all groups, an
"For
all
iterative proems, taking the calculated contributions from one approxima

tion as the estimates for the next, would give a series of calculations with
results converging to those for the self-consistent field;
though this process
would be unnecessarily lengthy as with experience It is usually possible to

make revised estimates better than those obtained by simply taking the
,
calculated contributions of the previous approximation. But for the

groups
of the outer shel! and particularly the most loosely bound group, there
3
sometimes occurs a phenomenon which may be termed over-stability, in

which a change of estimated contribution to Z causes a change in the
calculated contribution larger than the change in the estimate. When a
group is over-stable in this sense, an iterative process would not converge,
but, for small variations, would oscillate and diverge, and it is then quite
necessary to choose, as revised estimates of contributions to Z, values
SURVEY OF CONSISTENT-FIELD RESULTS
361
better than the calculated contribu

tions of the previous approximation;
it is also unusually difficult to make
satisfactory estimates and adjustments

to them, so the process of approxima
tion to the self-consistent field
troublesome in such
An example
is
most
cases."
(Fig.$
M
)
of this kind of
over-stability
given in the calcula
tions for the 3d 10 group in Cull. The
is
tables of the radial functions obtained
are given in full in Hartree s paper. To

exemplify the point discussed in the
preceding section that the individual
electronic eigenvalues, (a i ) agree fairly
well with the X-ray terms, Hartree
20
gives this table:
From
the radial functions the total
charge distribution in electrons per

atomic unit of the radius can be calcu
Rbll
KII and
lated.
Using the
RbU,
the alkali ions preceding and

Cull in the periodic system,
results for
40
following
an interesting comparison is made in

14
from Hartree s paper. This
Fig. 9
shows clearly that the charge distribu
tion for Cull is much more compact
than for either of the alkali ions, a fact
which is reflected in the interatomic
distances in the crystal lattices of the
three metals:
Half interatomic distance
KH
Cull
RbH
2-309
1-275
2-43
14
Fig. 9 .
Charge distributions in KII,

Cull, and Bbll as obtained by the
Hartree method.
$ 14
CEXTBAL MELDS
362
distribution curves shows that

Comparison of these rallies with the charge
is
just a little greater than the
in each case half the interatomic distance
curve.
abscissa of the last inflection point in the total charge distribution
10. Seli-consistent fields for oxygen.
While most of the calculations on self-consistent fields are for atoms or

the case in
ions in which all the electrons are in closed shells, this is not
and
Hartree
has been made by
oxygen for which an interesting study
Black.* This affords a more severe test of the method than closed shell
structures in which there is no approximation involved in averaging the
over
all
directions in space.
The
results obtained for the total charge
_...
1S
Fig. 10 .
Charge distributions in O I, O IE, O HI, and

by the Hartree method.
O
O
o
II
IV
ra
IV, as obtained
14 which indicates
clearly the
density distribution are shown in Fig. 10
increasing compactness of the atom as the degree of ionization is increased.
This closing-in is mostly due to the change in the 2p radial function as is
M
brought out nicely in Fig. 11 which shows the radial charge distribution
of the 2p electron for several stages of ionization. The curves are not
accurately repiesentable as a single function plotted to different scales of
abscissas,
however which
s
is
the approximation implied
when
the wave
functions are treated as hydrogenic with appropriate screening constants.

In using these radial functions to calculate the energy levels of the normal
configurations of the oxygen ions, the quantity YIPF is evaluated, where

is the properly antisymmetrized combination of the one-electron functions,
as in 3 S 6. This integral can be evaluated for each state of the zero-order

* HJLETSKB aad
BI^CK, Pfroe. Roy. Soc. A139, 311 (1933).
SELF-CONSISTENT FIELDS FOB OXYGEN
[014
363
Approximation in terms of the F and G integrals as in Chapter vi and then

these may be used as in Chapter vn to find the energies of the RussellSaunders terms. The present calculation goes beyond that of Chapter vn,
for now we have the radial functions and so can get approximate theoretical
values of these integrals instead of treating them as adjustable parameters.
There are two points to be noted in extension of the developments of
Chapter vi:
6 s and 7 6 depend essentially
First, the results of Chapter vr, especially of
on the fact that the one-electron eigenfunctions, w(& ), are orthogonal. If the
,
functions are not orthogonal, analogous results can be developed but they
are much more complicated. The self-consistent functions are orthogonal
01
Oil
o
o
in
rv
r
14
Fig. 1 I .
Change in
E2 (2p) with increasing ionization in O I, O II, O HI, and O IV.
and l because the spherical harmonic and spin

with respect to 1 9
s,
factors have been retained, but they are not orthogonal with regard to n
since different central fields are used for R(nl) and R(n l). This is most simply
f
remedied by using for R(nL) and R(n l) not the fonctions given directly by
the Hartree method but linear combinations of them that are orthogonal.
This does not affect the value of T because it amounts to performing the
same linear transformation on all the functions in the same rows or columns
of a determinant. (Compare
7 14 .)
We pause to note that the fact that this
of ways without affecting the result shows

that the exact distribution of the charge density among different values of n,
may be done in a large variety
without significance. Thus for the Is and 2# electrons any

two orthogonal functions for which J22 ( Is) -I- H?(2s) is the same function of r
as is given by the Hartree field will do equally well. In the actual numerical
for the
same I,
is
CENTRAL FIELDS
364
work
it is
10 14
found convenient to take the foEowing combinations as the
orthogonal functions (denoted by
),
- $ 2 )-i[I?(24 - SR(ls)],
S=
where
R(ls) R(2a) dr.
Jo
Second, in the Hartree method the radial functions are not, as in Chapter
vi, eigenfonctions of the same central-field problem. As a consequence the
calculation of the quantities
becomes an actual calculation, whereas before we could replace
at once
the u(a?) were characteristic functions of this centralproblem. This makes necessary a fair amount of additional computing.
The results of the calculations are very good, with regard to both the
by
<((/)
field
absolute values of the terms and the inter-term separations.

Measuring each
term down from the normal level of the ion with one less electron, the values
in atomic units are:
o in 2/
0fl
01
2p*
sp
/>
The agreement here
is as
good as can be expected. On the standard first
the
interval ratio ( lD^S)j^P^D) should be 1-5 while the
approximation
observed ratio is only I- 14 in 1 and III
5 7 ), so evidently large per
(cf.
turbations of the second order are
here.
important
CHAPTER XV
CONFIGURATION INTERACTION
Hitherto
we have
neglected matrix components of the electrostatic and
spin-orbit interactions which connect different configurations when applying
Now we have to consider what properties of atomic

associated with these matrix components. So long
are
distinctively
spectra
as the inter-configuration components are neglected, the resulting eigenfunctions are precisely associated with definite configurations; this has been
the perturbation theory.
the standpoint of the preceding chapters. If they are no longer neglected,

their effect may be treated as a perturbation which causes interacting energy
levels to
be pushed apart and results in an intermingling of character through
linear combination of the *F s of the interacting levels.

in atomic spectra.
cannot
be
level
a
of
accurately approximated by a single
Generally the
6
definite function of the type 3 6 or of combinations of them belonging to one
in the last
configuration such as we have considered exclusively except
were
needed
forms
seen
more
that
have
xiv
we
In
general
Chapter
chapter.
Such interactions are of quite general importance
an accurate calculation of the normal state

is used, all attempts to recognize the
method
of helium. When the variation
trial functions as the wave functions of a central-field approximation are
wave
given up. The reader may readily satisfy himself, however, that any
function depending explicitly on the distance r12 of the two electrons in
helium corresponds to no definite configuration assignment in any centralfield problem. In such calculations of energy levels the central-field ter
elsewhere
minology is more in the background, however, so we must look
by Hylleraas
in order to obtain
for really distinctive manifestations of the interaction of configurations.

These are of several kinds. So far very little theoretical work has been
done in the way of definite and detailed calculations of these effects, so this
sketch.
chapter must be in the nature of an outline
have alarge electro
First,it may happen that two particular configurations
static interaction which is sufficient to make the order of the terms be other
than that given by the ordinary first-order theory of Chapter vn. The best
2
in Mgl, which we
example is the interaction of sd and p configurations
consider in
1.
entire series of
Second, one term of a configuration may interact with an
terms with a consequent departure of the series from the simple Rydberg
or Ritz formulas which usually hold for series. This we shaE discuss in 2.
of an atom are in
Third, terms lying higher than the ionization potential
continuous
the
in
states
spectrum correa position to interact strongly with
I 15
COXHGUEATION mTEEACTION
366
relative to the ion. As a

spending to unclosed electron orbits of one electron
suet levels lose their sharpness and acquire something of the cha
racteristics of the continuous spectrum. They give rise to broad fuzzy lines,
result
their intensity relationships become quite

sensitive to the partial pressure of free electrons in the source. This pheno
menon, which is analogous to predissociation in molecules, has not been
their lifetime is altered
and
treated very exactly theoretically in any special case.
known facts in
We shall report the
3.
c
9
Fourth, as already mentioned in I , the possibility of so-called twoelectron jumps/ that is line emission in which the apparent configuration
change involves two of the id values, is connected with breakdown hi con
3
figuration assignments. This is discussed in 4.

in the alkali spectra
Finally, in 5 we discuss the intensity anomalies
which are associated with spin-orbit interaction between configurations.

1
Interaction of sd
and -p2
in
magnesium.
The general theory of the Russell-Saunders term energies (Chapter vn)

3D
3D for the
1
simple s d configuration and for the interval *D
gives D
the
ij) -82)
= f fl(na,
<Z),
do not need
where the 6 is defined by 86 15b. From their definition the
to be positive like the JF s, and so at first sight it appears that the singlet
might on occasion He below the triplet on the energy scale. This is actually
the case in the 3^ 3d configuration of Mgl. However, Bacher* noticed
that no reasonable approximation to the radial functions jft(3s) and R(3d)
would make the G integral negative. Using radial functions determined by
a
devised by Slater he calculated iJ9 3D~ 4-4000 cm-1 whereas
<?*s
the actual experimental value is ~ 1600 cm" 1

1 below its
Thus the *D is actually about 5600
position calculated by
the ordinary method. This Bacher ascribes to interaction with the 3p 2 con
.
cm""
figuration,
Xe
which gives *D,
P and I S terms.
will
Xj* the electrostatic interaction
Since
JL
and
S2 commute with
have no matrix components con
necting terms differing in regard to L and S. Likewise since the electrostatic

of II 6 there will only be
interaction commutes with the parity operator
&
interaction between configurations of like parity. The parity condition is

2
1
ofp 2 with
satisfied, so we expect an interaction of p and sd, but only of the
the 1
D of ad. The former is above the latter and as all such interactions make
the levels
move apart the situation is right for an explanation of the effect.
What is surprising is that the interaction is large enough,

above the ionization
level of
Mgl, about 15,000
cm"
for the
D ofp* is
above the mean of
D and 3D of the ad configuration.
By setting up the electrostatic interaction in terms of the zero-order wave

*
BACKER, Ptys. Rev. 43, 264
(1933).
INTERACTION
[is
(MB sd
1
AND
2
3>
367
IN aiAGNESIUM
and transforming to LS eigenfunctions, by methods explained in

conChapter vm, Bacher finds that the non-diagonal matrix component
1
2 1D amounts to
!
Their
interaction
cm"
13,200
energy is
aecting sd D and p
Junctions
thus comparable with their unperturbed separation. Calculating the effect

1
1
in detail he finds that the interaction pushes the sd !) down to 4000 cm"
below the
triplet,
whereas
it is
only observed 1600
cm-1 below. The conclu
sion is that the configuration interaction is certainly adequate to account

for the inversion. As to the inaccuracy of the final result, perhaps no better
agreement can be expected because of the approximate character of the radial

functions used.
1
large interaction, of course, the exact assignment of the
to the configuration sd is quite meaningless. Its true wave
With such a
below sd
function will be a linear combination of those for sd

there
is
a large component of the
character of the
spectrum,
latter.
W and
j?
D in which
Naturally this alteration in the
wave function will bring with it other special features of the
e.g. altered intensities.
Another detailed study of configuration interaction has been made by

2
Ufford,* who calculated the interaction between the configurations nd ris,
3
2
nd 7idn- s and nd^n^s. He compares the results with observed data in
f
Ti II and Zr II, and finds definite evidence that the configuration interaction
have seen in I 14 that in all the
has altered the intervals between terms.
elements where a d-shell is being filled, the energy of binding of an s electron
We
n value is about the same as that of a d electron, so that there

is a large amount of overlapping of the terms arising from such configura
x - 2 s*. Hence
tions as dx d x -*s and d
quite generally we may expect large
of one higher
due to configuration interaction here, although as yet the only detailed

calculations we have are those due to Ufford.
effects
Perturbed series.
is that of
Perhaps the most interesting effect of configuration interaction
formulas. Such
producing strong departures from the usual Bydberg-Ritz
various
and
time
a
for
known
have
been
series
explanations
long
irregular
for them have been advanced, but Shenstone and Russell,! following a
evidence that they are
suggestion of Langer, J have given very convincing
due to configuration interaction.
The Ritz formula is of the form
2.
the absolute energy value of the n th term measured from the

series limit, R the Rydberg constant and ft and a are constants, a being small.
where a n
is
UITOSB, Phys. Rev. 44, 732 (1933).

t SHENSTONE and RUSSELL, Phys. Eev. 39, 415 (1932).
Bev. 35, 649 (1930).
, Ftys.
CONFIGURATION IXTEBACTION
363
The
quantum number n*
2 15
by the relation given in (1).

by plotting (n* n) against
a n In that case the Rydberg formula (a = 0) is represented by a straight Hne
n*~-m = parallel to the axis of a, while the Bitz formula is a straight line
with intercept /i on the axis of ordinates and slope a.
Shenstone and Russell find that they can represent many series which
effective
defined
is
It Is convenient to study the behaviour of series

.
$ji f
depart from
(1)
by the formula
n* = n^jJi-rX(J n
8
-i
(2)
On
Ô
the added term in the formula for the effective quantum number representing
the effect of perturbation of the series members by a foreign level at the term
value GPO Such, a formula plotted with n* n against & n represents a hyper
.
bola with a vertical asymptote at 0"=cj

Before the perturbation of series
levels by an extraneons term was recognized, the extraneous term was often
.
counted in as a member of the series. As a result the value of n* n would

rather rapidly by a whole unit in going past the value of the perturb
ing term "because by counting it in the series all Mgher terms would be
erroneously assigned an n one unit too high.
THs is illustrated in Fig. I 15 wMefa. shows the series of the 3D^ levels of Ca I,
Curve
(1)
shows the n*
n values as usually given and curve la shows
how
this becomes a hyperbola with asymptotes at a = CTO when the Hglier series
members are given the altered n assignments. To indicate haw accurately the
hyperbolic law is obeyed, curve (2) shows a plot of (n* n)(cr n G ) which
should not contain the singularity in the curve of (ri*n) against a. In
curve (3) the values of (*-)-j3 (a
cr ) are
plotted to an exaggerated
s
tt
scale
wMch shows in another way how accurately the formula is satisfied.
Related to the perturbation of the
D levels of the series is the perturba
tion in the intervals between the levels of each term. In this series the
maximum up to the perturbing term and then

rapidly sink to zero as one goes up the series. TMs is due to the fact that
there are three series, 3 3 , 3 2 and 3 1? each of which are separately perturbed
intervals increase to a
D D
by the corresponding
level of the
extraneous term. As the perturbing term

has larger intervals than the series would have if unperturbed, this tends to
increase the other intervals by the differential effect of the
perturbation.
Numerous examples of other perturbed series are given by Shenstone and

The subject has recently attracted the attention of a number of
Russell.
spectroscopists,* Pincherle has given a theoretical calculation of the per

21
turbation of the 3s nd a series in Al II
term and finds order-ofby the
%>
magnittide agreement. Beutler has extended the

*
work of Shenstone and
PKCHEBLE, Nnovo dmento ID, 37 (1933);

BETCTLEB* Zeita. fur Phys. 83, 404 {193%
,
404 (1933);
Boy. Soc. A142, 286
ibid. 83,
,
Free.
(19SS).
PERTURBED SERIES
Lussell on the perturbed P and
ew experimental data on the
1
-369
P series in Hg I. Rasmussen. has obtained

F series in Ball showing perturbations.
angstroth has made quite a detailed study of perturbations in Bal,

icluding the effect on intensities. Some general theoretical work on the
which configuration interaction affects line strengths has been done

y Harrison and Johnson.*
The most thorough calculation by quantum-mechanical methods of series
^regularities are those given by Whitelaw and Van Vleck for Alll.f
-ay in
Mr
-\6M
6060
1S
Fig. 1 .
imOO
8000
ffn
The perturbed *DI
in
12.000
*-
cm"
series in
Gal.
An
explanation involving configuration interaction for the occurrence

of inverted doublets in the alkali spectra ( 8 5) has been
proposed by
White and considered in detail by Phillips. J This inversion is shown to be
capable of production by interaction with configurations of extremely high
6
energy in which one of the electrons of the p shell in the core is excited.
3.
Auto-ionization.
At all energies higher than the minimum necessary to remove one electron
of an atom to Infinity, the spectrum of allowed
energy levels is continuous.
In the one-electron problem the continuous range of positive
energies has
* EJLBBISON and
JOHSSOX, Phys. Rev. 38, 757 (1931).
f WHITELAW, Phys. Rev. 44, 544 (1933)?
VJLBT VLECK and WHITELAW, ibid. 44, 551 (1933).
}
WHITE, Phys. Rev.

PHILLIPS,
ibid.
40, 316 (1932);

44, 644 (1033).
CONFIGUBATION INTERACTION
370
3 35
associated with, it a continuum of states for each value of the orbital angular
momentum. To label these states the integral total quantum number n of
by the continuous variable E which gives the

limiting value of the kinetic energy of the electron when at large distances
from the nucleus. It may happen that a configuration in which two electrons
the discrete states
is
replaced
are excited, but in which both are in discrete levels of the basic one-electron
problem,, gives rise to energy levels lying above the least ionization energy
of the atom,
i.e.
in the midst of the continuous spectrum.
interaction between these states
If there
is
no
and the configurations involved in the

do not exhibit any special properties on
continuous speetrum these levels

account of their location in the continuous spectrum. However, if there is
interaction the
of the quasi-discrete level becomes coupled with the
s
5
of the neighbouring levels in the continuum. As a consequence the state

assumes something of the character of the states of the continuous spectrum.
The most important
feature of the states of the continuum is that they are
unstable in the sense that one of the electrons moves in an orbit which extends
Hence as a result of interaction with them the discrete level

some extent the property of spontaneous ionization through one
of the electrons moving oS to infinity. This property is called auto-ionization.
The direct spectroscopic effect associated with auto-ionization is a broad
ening of lines whose initial levels are subject to the effect and an alteration
to infinity.
acquires to
in the intensity of these lines with variation of the concentration of free

electrons in the source. There are two mathematically equivalent ways in
which we
may regard the
We may say that the
broadening of the
lines.
discrete level, as a level, remains sharp, but that-
owing to coupling with the continuous spectrum there is a probability per

unit time that the atom will make a radiationless change of state over into
a state of the continuum of equal energy. Owing to this possibility of
leaving
the discrete state rather quickly, the interaction of the atom with the radia
is limited to the production of short wave trains of mean duration
tion field
mean life of the atom in the discrete level. In the spectroscope

wave trains give rise to a broadened line because the Fourier
integral representation of the short wave train involves a band of frequencies
in the neighbourhood of the mean frequency.
Another way of regarding the matter is to treat the discrete level as com
pletely assimilated into the continuum. At each energy E the T of the corre
equal to the
these short
sponding state in the continuum will contain a certain component of the

eigenfunction of the assimilated discrete level. This will be larger the nearer
is to the
original position of the discrete level. Assuming that the assimi
lated discrete level, A, is capable of strong radiative combination with a
lower discrete
level,
while the ordinary continuum combines

weakly or
371
AUTO-IOOTZATIO2*
3 15
with B, it is clear that the intensity of radiation from the con

tinuum in combination with the discrete level B will be dependent mainly on
the amount of W(A ) contained in the eigenfunction of the continuum at each
not at
all
level JL.
energy. As this is a maximum at the original location of the discrete
we obtain also by this argument a broadened line at this place.
Auto-ionization is thus a consequence of the presence of matrix components
of the Hamiltonian connecting a discrete level above the ionization energy

with the states of the continuum at the same energy. If these are not to
vanish, these states must be of the same parity and the same J value; more
over in case of Russell-Saunders coupling they must have the same
L and S
The perturbation theory for this case requires some modification to

take into account the fact that the interaction is with a continuous spectrum
of states. This has been discussed by WentzeL* Apart from the general
formulation and some qualitative discussion based on the application of the
selection rules, very little has been done so far in the way of attempted
values.
The theoretical
quantitative calculations of the amount of auto-ionization.
the
of
the
is
related
to
that
predissociation,
theory
underlying
problem
in molecules, whereby a molecule may spon
analogous phenomenon
in which
taneously dissociate if put into a quasi-discrete state
exceeds the energy necessary for dissociation.
In view of the lack of an accurate detailed theory we

content with a review of the main experimental
/$
facts The first evidence for auto -ionization was noil- K
shall
s
its
energy
have to be
?
*
spectroscopic in character. Augerf showed that

when a gas in a Wilson cloud chamber absorbs
X-rays there are frequently several electron tracks

diverging from the same point. One trail is long and
c\j
interpreted as due to the primary photo-electron

shell by the light quantum.
ejected from the
is
Another of the trails was found to be of the correct

length to correspond to a kinetic energy of about
and L are written for the corre
2L, where
excitation
energies of the atom. Other
sponding
when
trails,
present, correspond to electrons of
considerably less energy. The interpretation of the

2L is clear from the diagram
trail of energy
G%t
*-*
The atom is left in the Jf level by removal

TT
11
of an electron from the A shell. Of equal energy
(Fie.
2 15
,
).
"
r.
^<-
15
Fig. 2 . Auto-ionization
or Auger effect in a
ieveL
* WBKTZBL, Zeifcs. fur

Phys. 43, 524 (1927); Phys. Zeits. 29, 321 (1928).
de Physique
f AUGEB, Comptes Rendus, 180, 65 (1925); 182, 773, 1215 (1926); Jour,
484 (1932).
(1925); Asm. de Physique 6, 183 (1926). See also LOOKER, Phys. Rev. 40,
6,
24-2
205
with the
the
3 15
CONFISCATION INTEBACTION
372
E=K2L.
W2sQ
Hence
level is the configuration
Es, where
the
auto-ionization
may pass over into this configuration by
atom
2 L. The
which implies expulsion of an electron with energy
in
detail
in that
rather
in
is
correct
rough,
principle, though
argument
we have supposed the energy of removal of the second 2s electron
also equal to L, but such roughnesses axe acceptable in view of the lack
of precision with which the electron energies are measured. The short
process,
electron trails are supposed to be due to further auto-ionization processes

2
occurring in the ion in tie I* 2$ configuration. Such spontaneous ionization
3
cm"
"
r~
i
r COMPTON
HARMS
^-
Theoretical (WENTZEL)
2
15
Fig. 3
radiation as a function of the atomic

Fraction, p of atoms emitting
level is given by i p.
nrnnber. The probability of anto-ionization of the
9
level may get out either

processes are by no means rare. An atom in the
a
line
of
or
radiation
auto-ionization.
For
by
by
light elements the auto-
ionization processes occur much more frequently than the radiation, while
the reverse is true for heavy elements. TMs is shown in Fig. 3 15 3 where the
K level which emit K radia
5 -p, gives the fraction of all atoms in the

so that 1 p gives the fraction which pass
ordinate
from the K level by autoThe abscissas are the atomic numbers Z* The heavy curve is the
theoretical value as estimated by Wentzel. According to him, the probability
tion.,
ionization.
of auto-ionization
is
nearly independent of
so the whole variation of p
3 15
373
AUTO-IONIZATIO3T
Z Is due to the increase of the radiative transition probability, approxi

4
mately as Z?
in the
Recognition of the importance of auto-ionization in atomic spectra
discussed
Shenstone
to
and
due
is
to
Shenstone
Majorana.*
optical region
the effect in general and especially in connection with the spectrum of Hg I.
Among other things he showed how it provides an explanation of the ultraionization potentials discovered by Lawrence.f Lawrence found that the
with
probability of ionization of Hg atoms by impact of electrons having energies

close to the ionization potential shows discontinuities of slope as if new
ionization are becoming effective with increase in the
modes of
producing
as due to excitation to the

voltage. These were interpreted by Shenstone
levels which are subject to auto-ionization. This view seems adequate to
no detailed correla
explain the observed facts, although at present there is
tion between the values of the ultra-ionization potentials and known spectrelevels. Shenstone thought the levels responsible for the ultrascopic
g
ionization potentials were probably connected with the d s*p configuration
in Hg, but later work by Beutler J on the far-ultraviolet absorption spectrum
shows that this particular identification cannot be the correct one,
2
3
terms due to 4p 2 , 5jp 2 and 6p in the
s paper deals with the
Majorana
and Hg respectively. Here the striking thing is that in all

P and 3 PX levels are known but the 3P2 cannot be found.
spectra of Zn, Cd
three cases the 3
The absence of 3 P 2 is interpreted as due to strong auto-ionization pro

as to why this level should be so much
bability, and this raises the problem
more affected than the other levels of the same term. The auto-ionization
the 4s oci limit in
Zn
by interaction with the continua associated with

6s col in Hg). The 4s Ep continuum
(correspondingly the 5$ ooZ in Cd and the
is due to an odd configuration and hence cannot give rise to auto-ionization
arises
of the even 4p 2 configuration, which can only be unstable through inter

action with even parts of the continuum, that is, with 4s Ea and 4s Ed.
2
*S and 1 terms sub
These give rise directly to interactions making the
breakdown of the LS coupling in
ject to auto-ionization. Through partial
3
2
P ) acquires a component of Y^-Dg) and tilus
the
4j>
configuration,
Y(
becomes subject to auto-ionization through interaction with 45 Ed. As P 1 is

the only level in the configuration with J = 1 it is not subject to breakdown
3
auto-ionization. However, P should become
and so
of LS
Z
coupling
escapes
mixed with *SQ and hence unstable; the fact is that the interaction here
3
seems to be much weaker than for P 2 but Majorana s theoretical discussion
does not show clearly why this should be so.
* SHENSTONE,
Phys. Rev. 38, 873 (1931);
MAJORANA, Nnovo Cimento 8, 107 (1931).

,,,*
See also SMITH, Phys. Rev. 37, 808 (1931).
f LAWBE*CE, Phys. Rev. 28, 947 (1926).
710 (1933).
J BEFTLEB, Zeits. fur Phys. 86,
3 15
CO^MGITBATION INTERACTION
374
arc spectrum of copper is known to have a large number of levels

above the ionization limit, some of which give rise to broad lines. Results
of the experiments of Allen* find satisfactory explanation in terms of the
The
auto-ionization process as was pointed out by Shenstone in a note appended

to Allen s papers. Particularly interesting is the behaviour of lines in multi4
term which lies in the range from 95 to 2164 cm.- 1
plets arising from a
above the d
10 ocl
D arises
4
ionization limit of Cu. This
from an even con
and so can show auto-ionization through interaction

is enough breakdown of LS coupling to permit
violation of the selection rule on S. But according to the J selection rule
4
4
only the levels D| and D^ can show auto-ionization. Therefore in a multi
4
4
while those
ple!, lines arising from the Dj and D| levels should be sharp
be
should
broadened.
the
levels
of
two
from
the
other
quartet
originating
This is in fact the case. Allen measured the line breadths in a Cu arc running
9
figuration. (3d 4:s6s)
with
D of dlQ Ed if there
up to 80 atmospheres. The breadth of all lines

originating from the quartet levels was found to increase linearly with the
pressure, with the same rate of increase for unit pressure increment. But lines
from 4D and 4D| were found to approach finite width at zero pressure whereas
those from the other two levels approached vanishing width at zero pressure.
in air at various pressures
Allen also studied the variation of relative intensity of the lines arising in
a
term with change in the arc current. The arc in these experiments ran
4
at atmospheric pressure. It was found that the lines from the unstable levels
of the quartet term were very sensitive to arc current, approaching zero
relative intensity at currents below one ampere and approaching constant
values for currents above twelve amperes. These intensities are relative to
other lines of the spectrum, auxiliary experiments having shown that lines
not subject to auto-ionization retain constant relative intensity with varia
tion of arc current. It seems likely that this effect is due to the fact that the
atoms in the unstable states normally fall apart by auto-ionization before
they can radiate. But if there is a high concentration of free electrons around
the ions the inverse process, in which free electrons are caught up by an ion,
comes into play. This could nullify the effect of auto-ionization and build up
the strength of the lines from the unstable levels. This view has not been
subjected to a quantitative discussion.
Auto-ionization effects observed in spectra of the alkaline earths and in
shown by White, f provide further examples of the
operation of the selection rules.
rare-gas spectra, as
BeutlerJ has found a long absorption series in

*
Hg vapour corresponding
ALLES-, Ebys. Rev. 39, 42, 55 (1932).

f WHITE, Pfcys, Rev. 38, 2016 (1931).
J BETOOK, Zelts. fur Phys. S6, 710 (1933); 87, 19, 176 (1933).
375
AUTO-IOOTZATICm
3 15
to the transitions
&P Q 6s* S
:L
tf
+5cP6s*np
These are aU subject to
that
broadening by the auto-ionization process and the experiments show
the amount of the broadening decreases as one goes up the series, showing
that terms near the ionization limit are more unstable than those of similar
character at higher energies.
4. Many-electron,
jumps.
As we observed in I 9 it is a consequence of the fact that the various
moments electric-dipole, electric-quadrupole, magnetic-dipole, etc. are
,
that the only non-vanishing matrix components con

quantities of type
nect states differing in regard to one individual set of quantum numbers.
Therefore in the approximation in which an energy level is assigned definitely

to one configuration of a central-field problem, radiative transitions occur
to one of the nl values. Such
only between configurations differing in regard
in many spectra,
one-electron
However,
transitions are called
jumps.
observed
are
corresponding
in the elements of the iron group, lines
especially
to transitions in
which two of the
nl values change. These are
known
as
two-electron jumps.
This is clearly due to the breakdown of precise configuration assignments.*

to
(J5) corresponding
If we know accurately the eigenfonctions Y(-4) and
them
between
transitions
two levels A and B, the existence of radiative
from the non-vanishing of the matrix components (xljajJS) con

where a stands for any type of electric
necting the states of A with those of B,
arises
or magnetic
moment of the atom.
If we try to describe the
atom in terms of
appear as expansions in
a central-field
that central field. In
on
terms of the eigenfunctions of the states built
approximation, T(4) and T(B)
will
involve several different configurations for

general this expansion will
and
(-4)
for Y(jB).
not too great, one configuration in each

with a considerably larger coefficient than the others,
If the configuration interaction
is
expansion will appear

and the experimental spectroscopist
tion.
The one-electron jump
label, for
will assign the level to that configura

rule will not then apply to this approximate
A may combine with configurations excluded by this rule in virtue
In this way
of the other configurations involved in its eigenfunction.
be
permitted.
apparent two- or many-electron jumps may
of the
to be fully adequate for the interpretation
This view
appears
there have not been as yet any
apparent two-electron jumps. However,
of this point. What is needed are estimates of
quantitative investigations
calculation
the amount of configuration interaction either from a precise
observed
the
from
from the fundamental wave equation, or inferentially
* CONDON,
Phys. Rev. 36, 1121 (1930);
GOUBSMIT and GROPPEB,
ibid. 38,
225 (1931).
CO^HGUEATIOX INTERACTION
876
4 15
perturbations of the energy levels. With this one could calculate the relative
transition probabilities for the different configuration transitions to compare
with measurements of these transition probabilities.

5. Spin-orbit perturbation of doublet intensities*
We have
seen in
2 9 that one of the simplest conclusions
from the sum
2
rules for line strengths is that the two components of the doublet 2
$
should have strengths in the ratio 2: 1. This result involves the assumption
P->
that the radial factor of the wave function

the
is
the same for the two levels of
P term. The result is generally in agreement with experimental results,

9 5 3 but in the case of the higher members of the
principal
of Cs the ratio is found to be larger than two. The different experi
as mentioned in
mental results are not in complete accord, but agree in indicating that the
2
ratio for the second member of the principal series, 7 2
S is between
3-5 and 4-5 although the ratio seems to have the normal value for the first
member, 6 2P-^6 2 $. This was explained by Fermi* as an effect of the matrix
P->6
components of the spin-orbit interaction connecting different terms of the

2
P series.
The
from
spin-orbit interaction,
value of L*S is |S
ponents
2
5
while for
connecting different
45
is
given by
For 2P| the
(r)L*S.
P| it is I The non-diagonal matrix com

members of the 2 P series are then
.
r)
R(np) R(n p)dr,
where (L m S) f
is the appropriate value of L*S. Because of this

perturbation,
order eigenfunctions are altered and by different amounts for the
two levels of 2 P. As the perturbation is diagonal in I, J and M, it is only the
the
first
radial factor of the eigenftoaction which is altered. If we write

R(np J) for
the
factor in 2 Pj in the first approximation, and
(np) for the radial
factor before the spin-orbit interaction is considered, we have
30
Jo
fW-
n
n
2
2
2
calculate the strengths of the two lines 2
Pf~ 5 and Pj-> $ the
calculations go through as before
that
we
must
use
these
altered
except
When we
radial functions in
calculating the integral irR(np)
enters as a factor in the

strength of the line.
*
The
R(n $)dr whose square
ratio of the line strengths
FEEMI, Zeite. fur Phys. 59, 680 (1929).
SPIN-ORBIT PEETTTBBATION OF DOUBLET INTENSITIES
5 15
for
2
Pf->
377
$ and Pj-- $ is thus 2 rom the angle and spin factors, multiplied
2
by the square of the
ratio of the
two
radial integrals,
R(np |) R(n*a) dr] [j r E(np |) B(n s) dr].

This is what makes possible the departure of the Hne strength ratio from the
r
[J
9
value given by the simpler considerations of 2
The numerical values of the integrals involved have been estimated
.
by
Fermi, who finds that in the case of Cs the values are probably large enough
to account for the observed intensity ratio. The quantities
which measure the change in the radial function are small compared with
the energy differences En En so the actual change in the radial functions
is quite small. For the first member of the principal series the smal change
does not produce an observable effect because its strength is so great relative
*
to the higher members of the series (cf Tables 7 and 8 ) that the introduction
into it of small components of the radial functions of the higher members
.
series members, however, the

produces no appreciable effect. For the Mgher
situation is reversed. Introduction of a small component of the radial func
2
tion of the lowest P term into their radial functions produces a relatively
the fact that the first series

large effect because of
much more
strongly with the normal
S level.
member combines
so
CHAPTER XVI
THE ZEE1IAN EFFECT
10 5
we have treated the Zeeman effect for one-electron spectra. This

serves as a simple pattern for the
present chapter, which is devoted to the
In
Zeeman
effect for the general case.
The normal Zeeman

9
1.
effect.
The argument of 105 which
leads to 10 5 o is valid for each of the

elec
trons in the atom, so that a
magnetic field of strength jg? in the direction of
the s-axis contributes to the Hamiltonian. the term
(1)
with o = eJfY2/Ltc as in
on the energy
10
levels of
6. The whole
theory of the effect of a magnetic
an atom is therefore given by a study of this
field
per
turbation term.
Before developing the theory from this

standpoint it will be instructive
to consider a little of the
of
the
Zeeman
effect. Prior to the intro
history
duction of the electron-spin hypothesis in 1925
had
physicists
attempted to
give a formal description of atomic spectra in terms of a
purely orbital
scheme of electronic states, so that the entire
momentum
of the
angular
atom was given by the sum of the L vectors for the individual electrons. In
Ls the sum of the ^-components of the orbital

angular
a constant of the motion which is
quantized to integer values
such a scheme
momenta,
is
MLfi*
Likewise the magnetic perturbation

energy is given by (1) without
Therefore the magnetic energy is
simply o times a constant of
the motion, so the effect on an
level
characterized
energy
by the quantum
number L would be to split it into 2L + 1
equally-spaced levels by addition
of the quantity olML to the
unperturbed value, where - L ^
L ^ L.
The ordinary dipole radiation involves transitions
between states for
the
Sz term.
which
AlfÔ,
in
which case the radiation is

linearly polarized with the
plane determined by the z~axis and the direction of
electric vector in the
AJ/X =1, in which case the radiation is

circularly
when viewed along the z-axis. Since,
according to the foregoing,
aH the energy levels are
split by the magnetic field in the same
way, the
observed splitting of all lines is the same on this
view of the matter. The
transitions for which AJ/ =
will all have the same
frequency and will
coincide with the unperturbed line. The
transitions for which
+l
propagation, and
polarized
ML^M
Tim is
actually the case for singlet levels
(5=0)
in KraseH-Saunders
coupling.
THE FORMAL ZEEMAN EFFECT
[is
379
have the same frequency which will be in fact shifted

1
toward lower frequency from the unperturbed line by an amount O/ZTT
1
The transitions for which L-*M L 1 will be shifted toward
O/STTC
same amount. For longitudinal observation, that
bigher frequency by the
:>n
emission will
all
see"
>r
cm"
observation in the direction of the magnetic field, the radiatioD is ciris polarized in the
3ularly polarized: for the high frequency component it
is,
direction of the positive current which in a solenoid would produce the

it is oppositely
applied magnetic field, for the low frequency component
polarized.
The conventional way of exhibiting this result, which we shall adopt, is

shown in Fig. I 16 With longitudinal observation one sees simply the two
.
components which are circularly polarized in the directions indicated if one

in a
imagines the magnetic field to be up from the paper. Observation
direction transverse to the field shows all three components, the undisplaced
one showing linear polarization parallel
(77)
to the
field,
the others showing
linear polarization perpendicular (a, for senJcrecJit) to the field.
Longitudinal
Transverse
rfr*t
left
A 777* +1
-1
fir
\
gr
Fig.
I 16 .
The normal Zeeman
triplet.
spectral lines was discovered by

Zeeman in 1896. Soon after, a simple electron theory of the effect predicting
was given by Lorentz. Although we have described the
the normal
The influence of a magnetic
field
on
triplet
this is by no
theory of the normal triplet in terms of quantum mechanics,
means necessary. The same result follows from a consideration of the action
of a magnetic field on a vibrating electron moving according to classical

mechanics. By comparison of the observed behaviour of some of the zinc
and cadmium lines with the Lorentz theory, it was found that the displace
ment corresponded to negatively charged particles having the same value
beams.
of e}p, as had been found for electrons by deflection of cathode ray
effect
Zeeman
the
atomic
theory,
Thus, in the very beginning of modern
with
is
connected
of
light
provided very strong evidence that the emission
the motions of electrons in the atoms.
and
This agreement between theory and experiment for the lines of zinc
cadmium gave a great impetus to the infant electron theory. Almost a year
effect on other lines showed that the
elapsed before further studies of the new
THE
380
ZEE3IA3T
EFFECT
I 16
normal Lorentz triplet Is by no means the general behaviour of a spectral

line under the influence of the field. Generally speaking the patterns are
much more complicated, so that it was something of a happy accident that
the first lines studied were those to which the simple Lorentz theory was
With characteristic love of the simple, physicists fell into the

habit of referring to the triplet pattern as normal*; all other forms of the
effect were caEed "anomalous. This characterization remains in use even
applicable.
to-day although we are now In possession of a complete theory which gives

a rational account of all the observed effects. In terms of the complete theory
the so-called normal effect
is
simply a special case in which the
effects of
electron spin are absent.

9
anomalous effect remained a great puzzle until the

electron-spin hypothesis was introduced and that was a quarter of a
centuiT later. Perhaps the continued use of the adjective anomalous is
appropriate in view of the long time during which the general Zeeman effect
resisted the attempts of physicists to understand it. Before leaving this
brief review of the historical setting it may be remarked that the essential
feature of the spin theory is the occurrence of (LZ + ZS^ in (1) rather than
(Lz -f Ss ). That is, that the ratio of magnetic moment to angular momentum is
Nevertheless the
twice as great for spin angular momentum as for orbital angular momentum.
If this were not the case introduction of electron spin would make no alter
ation in the theory of the normal triplet as just sketched, for Jz would simply
Lz everywhere with no change in the observable results.
be written for
2.
The
weak-field case: RusseH-Saonders terms.
Most of the work on the Zeeman effect applies to spectra in which RusseliSaunders coupling holds quite accurately, so it will be convenient to begin
the study with this case. As we have already seen in 105 special effects arise
,
If the
magnetic
strong enough to produce energy changes comparable
with the separation of the levels of a term, so it is also convenient at first to
consider the case of weak fields, meaning by this fields whose effects are
field is
small compared with the unperturbed intervals the fields commonly used
are weak in this sense. The name Paschen-Back effect is
given to the special
features of the
Zeeman effect which
arise
when the field is not weak.
To
find the weak-field perturbation of a Russell-Saunders level charac

M (1 16 1)
terized by SLJ, we must calculate that part of the matrix of
which
LS +2SS
can be written as JZ ^-SZ and

since Ss commutes with JS9 this part of the matrix will be
diagonal with
to
if.
The
element
J
of
for
the
state
SLJM
is MH; the
respect
diagonal
z
element
of
S
for
this
state
is
obtained
as a special case of the results
diagonal
s
refers to this level. Since
given in
103 . If
we
identify
Jx with S and J2 with L
the diagonal matrix
THE WEAK-FEBIiD CASE: EUSSELL-SATJ^BBRS TERMS

381
3
3
merit of Sg is given by a combination of 10 2a and 9 I i Altogether we find
.
the diagonal element of the perturbation energy the value

(y 8LJM\H*\ySLJM)
= oKgM
(I)
g were equal to unity, (1) would give the splitting which corresponds to
normal Lorentz triplet.
We
see that the energies of the perturbed states are distributed sym3trically around that of the unperturbed level. These states are 2J-f I in
imber, as before, but the scale of the splitting differs from the simple theory
the factor g. The factor g differs from unity by a term which arises from
r
component of Sz For singlet levels S = and L = J, so the

Lditional term vanishes. In other words, the theory of the normal Lorentz
[plet applies to the Zeeman effect of lines which are combinations of
e matrix
iglet levels.
These results which we have so easily obtained are in good accord with the
apirical data. The winning of the result (2) as a generalization from the
apirical data was not so easy and represents a great amount of study by
Preston s role* which
>ectroscopists. This formula expresses implicitly
ys that all the lines in a spectral series have exactly the same Zeeman
tttern. This is due to the fact that the perturbation energy (1) is indepen*nt of y which stands for all quantum numbers other than those explicitly
ritten. More generally, the Zeeman pattern depends only on the S, L, and
of the initial and final levels, provided Russell-Saunders coupling obtains
3
atom in question.
Runge s rulef says that the displacements
r the
of the Zeeman components

the unperturbed line are rational multiples of the Lorentz splitting
27rc. Since the displacement in a line is actually the difference of the dis[acements of the initial and final states, and since the gr s are rational
actions, it follows that this rule is contained in our results.
The empirical fact that the Zeeman effect of Russell-Saunders terms is
om
by the formulas (1) and (2) was worked out by Lande, and g is usually
Qown as the Lande factor. Lande s formulation was based on the modem
srperimental measurements by Back. This work is admirably summarized
the book by Back and Lande, Zeemaneffekt und Mutiiplettstruktur der
[ven
pektrallinien, which, furnishes a good account of the subject as it stood just

efore the electron spin and quantum mechanics gave the theoretical basis
the experimental material.

In Table I 16 are given the g values for the terms of interest. The table
shibits some interesting properties. For J = the formula for g gives g, but
>r
* PBESTOX* Trans.
Boy. Soc. Dublin 7 S 7 (1899).
f Era<5E, Bays. Zelts. 8, 232 (1907),
THE
382
ZEEMA2S"
EFFECT
2 16
/ = implies 31 = 0. As akeady remarked,

the
normal
singlets give
separation. For L = 8 the g factor equals f for all
values of J. A somewhat surprising feature is the fact that certain terms
the splitting is of course zero since
such as 4D| 5 J\ and 6 (?| have g = Q and so are not split

by the magnetic field
others such as B Fi , 7 {? 1 and 8 G| show
negative g values, which means that the
,
sense of the effect
is turned around, as it would be

and
the behaviour normal.
positive charge
if
the electron were a
I 15.
Land gfaclorsfor Ru&dl-Saund&rt
terms.
In the vector-coupling
theory one regarded 5 and L simply as two classical
angular momenta whose sum is J. The simplest cases sax>J=L + S,ia which
L and 5 have to be parallel vectors, and J- L- in which L
anti-parallel.
For
S,
J = L + S we find
L+2S
L+S
and
5 are
THE WEAK-FIELD CASE: RUSSELL-SAUNDERS TERMS
383
just the ratio to be expected from the fact that the spin angular
momentum counts double in magnetic
- $ s however, Normal
effect. The value for J = L
which
Is
tu,i
"I
ioes not correspond to the expectation
based on this simple

r view, but
S
1
Let us
L-2S+1
J+i~ 3^8+ 1
~_
W4
~
now consider the pattern to be
expected in the transition between two

levels. Using a prime to refer to the
initial
S-*
is
/-*2
and a double prime to refer to the

we have
final Ievel 3
tO)l_
1-2
where AJf = M*
rule
AJf
for the
By the
selection
components show
= I
ing parallel polarization, while Alf
for those showing perpendicular polar
213
GW579U
2-2
The pattern is readily seen to be

symmetrical about a& and is conveniently
ization.
(0)t:2)3S
by reducing the values of

the factor in brackets to a common de
characterized
nominator. Following Back and Lande

the numerical values of the numerator
{2)4
which correspond to parallel polarization

are put in parenthesis. Thus one may
(1)35
readily calculate that the values of the

bracket factor for 2 P| to 2D| are
(t)IFf3
IS
(4)8(12)1624
15
m3) 15 17/321
15
which means that there are TT components

at
~ând ff of o/2?rc from the undisplaced line and o- components at
{f and
f|. In this notation, the
is
pattern
independent of which is the
initial and which the final level.
/5
The different Zeeman components are

by no means of equal intensity, as we
16
Fig. 2 .
Zeeman patterns for several

multiplets. The relative
common
strengths of the components of each

multiplet are indicated by the
lengths of the bars (divided by the
indicated factor in the case of the
weaker satellites).
EFFECT
384
stall see in
4 1S , Experimentally both the
916
Zeeman pattern and the
dis
tribution of intensity provide important aids to the recognition of the

of the energy levels involved in a spectral line. The relative in
tensities are readily indicated graphically
by drawing ordinates at the
16
positions of the lines whose lengths are proportional to the intensity. Fig. 2
indicates clearly the diversity of the patterns
contains such diagrams,
some common transitions.

For the Zeeman pattern it
for
is convenient to make a double-entry table

for
used
multiplets. The method will be illustrated
akeady
of the pattern for d & 4p 4 P|->d6 4s 4Df in Mn as studied by Back.*
similar to those
by
Table 2 16 has a row for each state of the 4P| level and a column for each state
of tie 4D^
level. It is
of the
in zero field.
that the pattern is symmetrical around the position

The upper figure in each cell is the theoretical value
of the displacement, the lower is that observed by Back, the unit being
of
Lorentz splitting.
the
TABUS 2 W
-21; -140
-1*
-65
-31
-15
4-35
j
-15-12
;
3.
Weak fields;
4-34-8
1
getteral case.
In the preceding section we have considered the Mghly important case of

the Zeeman effect for Riissel-Saunders terms. Let us now consider the
nature of the effect for an energy level of any type in the weak-field case. An
energy level of a free atom is always rigorously characterized bv a
number J
quantum
of resultant angular
*
BACK,
momentum. The
Zeits. for Phys. 15,
states resulting
206 (1923).
from the
WEAK
16
pplication of a
FIELDS:
GENERAL CASE
385
weak
S resultant spin
is
field will be characterized by J and M. The vector

of the type considered in 93 and hence
Therefore the magnetic alteration of the energy level
is,
in general,
(aLJM\H*\KJM) = clSgM
LS
in
2 16 1
where we have, in place of
2 16
(1)
2,
0=l + (aJ:S:acJ).
The special result of the preceding section is obtained if
(2)
a = y SL.) In other
energy levels are split into 2 J -f 1 equally spaced levels sym

metrically distributed around the unperturbed level. The amount of the
splitting is governed by a Lande g factor as in Bussell-Satuxders levels, but now
s*ords all
Ms factor is given by the general equation (2). The whole theory of the weak
any term
leld effect for
is
If we express the state
lave
thus reduced to an evaluation of (a J:$:aJ).

) in terms of the LS-coupling scheme,
*F(oc
JM
we
T(a JJf)= SY(ySL JM)(ySL Jla J).
ySL
Eence the matrix component that we need
is
T(a JM) (Jz + S2 ) T(a JM)

ySL
Since
JZ +SS = LZ +2SS
(
is
diagonal with regard to y, S, and
in this
scheme.
The ^(Jg+Si) *? combination occurring after the summation sign is,

however, simply g(SL J) MK, whereas the quantity on the left is ^(a J) 31
ft
hence
we have
the result
(ae/)
= S g(SLJ)\(ySLJ\xJ)\^
(3)
which expresses the g factor
Lande
for an arbitrary level in terms of the ordinary

factors of the Russell-Saunders scheme and the transformation
coefficients (y
SL J\cnJ).
If we regard a as a variable label running over all the levels which have
a particular J value, and sum (3) over oc, we obtain
Eg(a/)-E
a
since the coefficient
(4)
g(SLJ),
ySL
2 \(y SL J |a J)J 2 which multiplies g(SL J) on the right-
hand side of the summed equation is equal to unity because the transforma
tion is unitary. The result contained in (4) is known as the g-sum rule.
Actually of course in an atom there are an infinite number of levels
associated with each value of J that occurs at all, so (4) merely becomes an
uninteresting oc = co
cs
if
applied to
all
the levels of the atom. Its importance

25
lies in
the fact
state in
3 16
EFFECT
THE
388
the summation over y which expresses an actual atomic

of states of the RusseH-Saunders scheme Is finite to a good
where the states Y(x J)
approximation. Thus in
may
be assigned
electron configurations, y Si are restricted to those values associated with

the
electron configuration as x. In this case (4) tells us. if we let a run
only over levels of a given configuration; that although the individual ^r

are
by a breakdown of Russeil-Saunders coupling, the sum
of the g values for all the levels of a given J in the configuration is the same
as in Russeil-Saunders coupling.
This Is exemplified by the measurements of Paschen on the Zeeman effect
in neon.
5
configuration 2p 3* gives rise to one level with
=
one with J 2, The coupling is as we have
The
J= i
f5 13
Russeil-Saunders
in
which the
levels
J=
two with
from the
far
would be labelled
P0s
The level of J = 2 Is 3 P2 regardless of coupling so we do not need the y-sum

S
to tell us that the g value for the state of J = 2 should be 7(
the
I and Paschen found 1-503 which is good agreement. For
with
I
g\
/=!
1)
Paschen found ^ = 1-034 and
- g\ 3 Pi = }
i
=1-464 so
2 ).
two
Sg =2-498,
This
Is
states
whereas
good agreement.
be applied to Paschen s values for the levels

Similarly the g
in the 2j/5 3j? configuration. Here J = 1 is represented by four levels, which in
the
LS
Saunders g values are 3 f

3
levels of
the
J= 1
P 1J
Pj and *D^. The corresponding Bussellsum of 5. The experimental values for

are 1-084, 1-340, 0-999, and 0-699, with sum equal to
1
are *S l9
1.
f with a
sum-rule check although the individual values are

5-022 providing a
far
the Lande values For J = there are three states with experimental
.
1-301, 1-229,
and
LS scheme are 3 P2 JD2

9
"2
1*137;
3
3-667.
The corresponding states in the
B2 which have a g sum of ^- in good agreement
with experiment.
4.
Intensities in the
The
Zeeman pattern: weak fields.
arising from a given spectral line when the source is in

be said to form a Zeeman pattern. The different com
field will
ponents. as
we have J list seen., are connected with the different changes in
The
relative strengths of the lines are therefore given by application of

formulas 9s ! 1 as noted in 7 4 ? where the necessity arose of summing over all
these separate transitions to find the total strengths of the lines of an
unperturbed atom.
In the early work on the Zeeman effect, observation was made of the longi
tudinal effect (radiation along
&f) and of the transverse effect (radiation
emitted In any direction perpendicular to 3ff). This revealed quite generally
INTENSITIES IN THE ZEEMAN PATTERN
WEAK
FIELDS
387
phenomenon of the circular polarization in longitudinal observation of

e lines which appear polarized perpendicular to
in transverse observain
the
absence
also
and
observation
of the lines which are
m,
longitudinal
e
darized parallel to
in transverse observation. Since transverse observa-
gives the complete pattern and is more convenient experimentally, that

the arrangement which is always used in modern work.
>n
For the Zeeman pattern of any line in which there is no change in /, the
4
rengths of the components in transverse observation (9 = ?r/2 in 7 4) are
oportional to
\(*JM\P\aSJM)\*=\(*J\Pi* J)\*M*
(rr}\
)
(
a
J|(a/Jtf|P|aVJlfTl)j
1 has equal and opposite displacement
nee the component
om the original line to that of M->M 1, it is evident that the pattern of
rengths given by these formulas is symmetrical about the position of the
=^
M->
aperturbed
M+
line.
we may
write two more sets of formulas giving the relative

Similarly
the
of
components of the Zeeman pattern for J"~/-f 1 transitions
rengths
J 1 transitions in transverse observation. They are
id
J->
aJ^
(*)}
--
hese also give patterns that are symmetrical around the undisturbed line.
It should be explicitly emphasized that these formulas are valid in any
is, are independent of the nature of the quantum
umbers symbolized by a and a The positions of the components of the
eeman pattern depend, as we have seen, on the other quantum numbers,
)upling scheme, that
the relative strengths do not.

first obtained empirically by Ornstein and Burger*
id later derived in terms of the correspondence principle by Kroiiig and
lit
These formulas were
oudsmit, and Honl.f

As an example of the experimental test of these formulas we give some
Leasurements by van GeelJ on a plate, taken by Back, which gave good
for a 8 $->s 8 Pf (A4852) and a 8 S^z 8 P^ (A4752) in MnL
he theoretical pattern for A4852 is given in Fig. 3 16 . Van Geel used the
ieoretical values of the intensities of the perpendicular components,
eeman patterns
CtosTEiN and BUBGER,
Zeits. fur Pfcys. 29, 241 (1924).
f KEOOTG and GOUDSMTT, Xaturwiss. 13, 90 (1925); Zeits. fur Pfays. 31, 885 (1925);
HONL, Zeita. fiir Phys. 31, 340 (1925).
t VAX GEEL, Zeits. fur Phys. 33, 826 (1925).
25-2
THE ZEEMAN EFFECT
388
4 16
together with the observed blackening of the plate for these lines, as a means
of finding the blackening curve of the plate. Using this blackening curve he
found as the experimental values of the relative intensities of the parallel

components
41 36: 27 5
.
15>
whereas the theoretical values are

40:36:28:16.
As the line A4752 was on the same plate, the same calibration was applicable
to this. The relative intensities of IT- and or-components, independently, were:
7r-components
or-components
57:46:27
54:45:25
70:56:37:22:12:?
73:52:35:21:10:3
Observed
Theorv
These are evidently excellent agreements. The paper of van Geel includes
more examples of this sort, all of which agree well with theory.
several
2
Fig, 3
5.
a a
z$ 38
Theoretical relative intensities in the 8 5r
The Paschen-Back
8
->
Ps Zeeman pattern.
effect.
As already mentioned, the departure from the above theory of weak-field

Zeeman effect which occurs with a magnetic field strong enough to produce
splitting comparable with the interval between levels of a term is called
the Paschen-Back effect. The theory has already been presented for one-
electron spectra in
Id 5
Now we may consider the effect in general.
The magnetic perturbation

scheme of states in which
H*
of
ML and M
1 16 1
evidently diagonal in any

B appear as quantum numbers, the
is
value of the diagonal matrix element being

oB(Jffz +21f ff ).
(1)
Suppose we
set of
start, as in Chapter vi, with a set of states based on a complete

individual sets of quantum numbers.
have seen in
vn
We
Chapter
that the inclusion of the electrostatic interaction of the electrons necessitates
THE PASCHE3T-BACK EFFECT
16
>
389
transformation from this scheme to an
$-coupEng scheme in order to

which this part of the Hamiltonian is in diagonal form. In
I 7 we saw that
s and
L can be retained as quantum numbers in an LS~
3oupling scheme it is only when the spin-orbit interaction is included that
it is necessary to pass to a scheme in which J and 31 are quantum numbers
in order to obtain a scheme of states in which the complete Hamiltonian is in
a,
obtain states in
diagonal form.
On the other
hand HM is not diagonal in a scheme in which J and 31 are

quantum numbers. Hence we see the place of the Paschen-Back effect in the
M
such that the
theory of the Russell-Saunders case. For values of
matrix components are small compared to the spin-orbit -interaction matrix
components the eigenstates of energy will be nearly those in which J and 31

are quantum numbers. This is the weak-field case of the preceding sections.
For values of Jf such that HM is large compared to the energy of spin-orbit
interaction, the magnetic term will dominate and the energy eigenstates
will be nearly those in which
s and
L are quantum numbers. For a
&t? from zero to such strong values there will be a continuous

in
the
character of the eigenstates from one of these limiting types
change
to the other.
variation of
Predominance of the magnetic-field energy therefore draws the eigen

states toward those in which J/s and
L are quantum numbers. In view of
2
M
and S2 and that we can have a scheme
the fact that H commutes with
of states in which SLNS L are quantum numbers, we see that there is no
tendency of the magnetic
field to
orbital or spin angular momenta
break down the coupling of the individual

and SL*
of the electrons into a resultant
In the Russell-Saunders case therefore the Paschen-Back effect results from

a competition between the spin-orbit interaction which works for the
scheme and the magnetic field which works for the
L scheme. Since
S
= s 4- L is a quantum number in both schemes, we see that the groups
of states having different values of If may be treated independently. In
group we may set up the secular equation for
dealing with a particular
scheme. For the
the energy values either in the
S
L scheme or in the JM
former we need to have the non-diagonal matrix components of the spinorbit interaction and for the latter we need the non-diagonal matrix com
L scheme is more
S
ponents of the magnetic energy. The matrix in the
in
the J 31 scheme
matrix
the
at the outset for strong fields,
JM
MM
M M M
MM
MN
nearly diagonal
* This is true in so far as we treat the

magnetic perturbation as of the form of (1). Kigorousiy
however there are other terms proportional to the square of the vector potential as mentioned in
lO5 ^. They do not commute with the orbital angular momentum, and so at fields strong enough to
make them important L and the individual Ts of the electrons are no longer quantum numbers.
than
However, this is of purely theoretical interest because the fields necessary are much greater
any attainable ones.
EFFECT
390
weak
5 16
be obtained from tie calculations

already given in other chapters so that the secular equations can be written
down for any special case and solved by the usual methods. Except for the
for
These matrices
fields.
may
of one-electron spectra, which has been already fully treated in 10 5

this procedure will not be
carried out in detail, for it can lead into very
,
lengthy calculations vhich have very little applicability to spectroscopy

since the fields needed to produce large Paschen-Back effect are seldom
The best experimental illustrations of the Paschen-Back effect

are in connection with hvperfine structure (5 18 ) where attainable fields can
a complete transition from one scheme of states to the other.
obtainable.
The Paschen-Back
was important in the pre-quantum -mechanical
effect
theories of atomic spectra for the information it gave about the coupling
relations,
was studied in this connection by Heisenberg and by Pauli.*
Soxnmerfeldt showed the relation of an old classical coupling theory of

Ybigt to the effect* The quantum-mechanical treatment was first given by
and Jordan and detailed cases were discussed by Darwin. J
Although the preceding remarks have, for definiteness, been made for the
Russell-Saunders case, it is evident that they hold with appropriate minor
modifications ia the case of
any other coupling
for the
unperturbed atom.
PROBLEM.
-T--rr:Ls the sum of the magnetic changes in energy of all states of the
equal to c$ J/2*J where Sf is the sum of the Lande g factors for the corresponding states
witn in a n-eaJt field. TMs is known as Pauli* s f/-permanmce ruh.
ShoTsame Jf
6.
*fcat for all
>i
:s
The Paschea-Back
effect:
Ulustrative examples.
3
3
example we take the 4*4cZ jD->4s4p P multiple! in
Zn 1 which has been studied by Paschen and Back and later by van Geel.
TTne first reference includes a number of other cases, tie
emphasis being on
As an
illustrative
the fact that strong magnetic fields bring out lines which violate the
ordinary
S
rale on J so that in the magnetic field the 3
multiplet is
D~>
completed to
all
nine lines instead of tie usual six.
TMs is, of course, a simple
consequence of tie fact that the states in a strong magnetic field are not
by precise J values. The second reference provides accurate
Intensity measurements on the forbidden lines for various field strengths.
characterized
f
J
HZKOBSTBEBG, Zeits, fur Phys. 8, 273 (1922);

PAUII S ibid. 16, 155 (1923); 31, 765 (1925).
SOMMEBEEID, Zeits. fur Phyg. 89 257 (1922).
HnsENKRBG and JOBDAS, Zeits. for Phys. 37, 263 (1926);
C. G. DABWBT, Proe. Roy. SCM:. A115, 1
(1927);
K, BAEWIN, ibid. At 18, 264 (i928).
PASCHEK and BACK, PJiyriea 1 261 (1921);
VAST GEEL, Zeits. fiir Phyi. 51, 51
(1028).
6 16
PASCHEX-BACK EFFECT: ILLUSTRATIVE EXAMPLES

391
The matrix components of HM in the SLJ3I scheme are obtained from
9 3 1I
and IG 3 2. They are
where g(SL J)
/-^-w
~~
VJ
^
the Lande factor of 2 16
is
The Zn 1 4s 4p 3 P intervals are 190 and 389 cm 1 so the weak-field theory

1
is adequate for them. The 3D intervals,
however, are only 3-23 and 4-94 cm
so the quadratic term in the perturbation theory is
important here. The
interval ratio shows that we are fairly close to Russell-Saunders
coupling.
The 4s 4d 1D is 300 cm- 1 below the SD indicating a strong perturbation of it
,
by 4p
21
D as discussed in
I 15 .
3
Applying the perturbation theory to D we find for If = 3 a linear
3
variation with the field of the energy of Df 3 For Jf =
2 there is an inter
action between the corresponding states of 3D 3 and 3D 2 and for Jf =
1
there is an interaction between these states of all three levels of 3D. For the
fields used the second-order perturbation is
adequate. Using the matrix
.
components
(oK/hc)
(!)
and the unperturbed
we readily
level separations
and writing
TJ
for
find for the quadratic perturbations:

Coefficient of ff in quadratic perturbation of
Jf=2
1
The numerical value of TJ is 4-674 x 10~ 5J^ where Jf is In gauss,

Paschen and Back publish data on the line 3D 2 -^ S P for
Jf"
= 39340
3
does not split this gives us directly the magnetic perturba
gauss. Since
tion of 3 2 . For this field TJ
1*84, so the quadratic displacements of the a
1
components which involve the If = 1 states should be 0-22 cm whereas
i and + 1 components respectively,
they find 0-15 and 0-40 for the Jf =
The TT component involves Jf = for which the shift should be 0*35 and
they report 0-32.
TMs agrees within the accuracy of the data.
Van GeeFs measurements on these

reproduce as Fig.
scale at 3281-927,
4 16
given in his Fig. 5 which we

with the addition of a smaH mark on the wave-length
lines are
which is the position of unperturbed 3D2 3 P given from

the energy values of the levels. The broken lines show the splittings as given
by the weak-field theory, the heavy lines are drawn through the experi
mental points. As drawn the departure for 3D 2 B PQ appears to be linear and
too large. That is because the broken lines have been drawn to converge at
->
->
THE ZEEMAN EFFECT
392
3281-95 instead of 3231*927.
When
the change
is
6 16
made
it is
seen that the
too great for a satisfactory comparison with

theory. That Is probably because this is a line which is forbidden in weak
fields and is weak even at the fields used and so is difficult to measure.
uncertainty of the points
The data on
line.
and
For this
0-334
line
->
he
cm-1
theoretical value
finds a shift
is
are probably
more accurate since this is a stronger

Jf = 30,000 to be - 0-185
finds the quadratic effect at
1, whereas the
components involving 31=
he
0*268; and for the component involving Jf =
compared with a theoretical 0-360.
for the
is
0-297 as
3281-95
Fig.
4".
Paschen-Back
effect in
>
->
3282*18
P multiple* in Znl.
The more interesting part of van Creel s work, however, concerns measure
ment of the intensities of the forbidden lines as a function of field strength.
The theory for this was given by Zwaan.* It is easy to see how the intensity
changes occur. For example, the state which in zero field becomes ZD\ is
represented by a T which is given by the perturbation theory to the first
4
ordering as
(cf.
9 2 5)
where the matrix components, given by (1), are the same as we have
already
used.
The
state that grows out of 3

D| therefore acquires
^, Zeits. for Phys. 51, 62 (1928).
some of the
PASCHEIS-BACE: EFFECT: rLLUSTRATIVE EXAMPLES
16
iharacteristics of
393
J= 1 and J=3 in addition to the J=2 which it has, strictly,
vhen unperturbed. If therefore we calculate the matrix component of
moment for transition to 3 P,
electric
for instance,
we
find
-)I\H^DD
other two terms vanishing by the selection rule on J. The magnetic
2
3
to tf? so the quantity j( Pg ;Pjî)l which
perturbation term is proportional
2
measures the radiation intensity will vary as Jf
of all the ordinarily forbidden components as a function of
The
;he
intensity
was worked out in this way by Zwaan. The components of the

3
P for which A J = 3 were found to vary as 3f4 while the two lines
and 3D 3 S P 1 for which AJ = 2 varied as Jf 2 for weaker fields.
Seld strength
Line
->
->
tan
L8
tan
2.0
290m
iOOOO
5GOO
IOQO
H
Fig.
for 3
5.
D2
Dependence of
3
->
^-V^PO
as
HI
Zn I.
(I is
in
3Q8GQ
gauss
effect
strength on field in Paphen-Ba^k
line
the strength measured with yfo of the strength of the
forbldden-ilne
toait.)
in powers of
a
simply the leading term of development
3
M* he found as well the coefficients of tf for the lines,
3 P and 3
3
1V
The results for the relative strengths of the lines D 2
5
=
3*0 x 10~ ^f ):
as found by Zwaan are (p
This JF 2 variation
is
->
Component:
*D*-^PO
1-0
~1
0~>Q
"^
D2 -^ 3 P
relative to
Geel measured the sum of the intensities of
between 10,000 and 30,000 ganss. The
field
*DI-**PQ as a function of the
16
Geel s Fig. 1) where the log of the
agreement is shown in Fig. 5 (van
Van
THE
394
Intensity ratio
Is
8 16
EFFECT
log Jf .
plotted
The broken line Is that given by the
theory. TMs Is a very satisfactory agreement. He also studied the departure

1-0 components from the p 2 law as
of the intensity of the i->0
3
given by the p terms and found good agreement with theory.
The
by Pasehen and Back contains a number of other examples of
the
in the magnetic field, but van Geel s are
of forbidden
the only quantitative intensity
We
in this field.
some old data on the Zeeman effect

a
Pasehen-Baek interaction between
an
which
illustration
of
provide
Hgl
a
a triplet term made possible because of the breakdown of
Kusseil-Saujiders coupling. The data are due to Gmelin.^ He worked with
low resolving
reported all the Zeem.an patterns as triplets, The
now
to a brief discussion of
in
spmtingofA4916(te^ 5 -^ft?6p
which is witMn 1 per cent, of the
1
If we
it is
given by theory,
of an unresolved
the relative intensities of the pattern components are

easy to calculate that the centre of gravity of intensity t
would come at an apparent specific splitting

the normal, where gD
gp are the g factors for the D 2 and
f#
->
a specific
)hefoiindtobe4-72x iO-Serar gauss,

value of the normal spitting. From
that = 1-01 for this *P. For A5769 (85 6d 3D 2
*Pi) he found
of 5-38, and for
(65 BdW^P^) he found 4-95.
we may
It
levels respectively.
and
of 1-04
1-10 for
In
this
ID2 and
of the Lande values is
way Gmelin^s data leads to observed g values

The sum is 2-14, whereas the
2 respectively.
= 2*17,
a fair agreement for such data,

normal we can get, by the method
of 3 18 a
estimate of the departure from Rnssel-Saunders coupling.
Calling the quasi-singlet level A
quasi-triplet B we have
1-GO-f- 1-17
From the departure of the g values

5
g(A)=g?DJ KÛ)!* -r
g(B) = ff ?D 2 )
?<,)
\(*Dt \A)\*
D \B)\
3
|(U>,
B)|
+g<*D t )
\(
wMch
the observed values of g(A) and g(B) together with the Lande
values ofg(
t)
3(D%) lead to
l
is a rather
departure from LS coupling. This rough result is
out by intensity measurements by Bouma f who found A5769 and
A5790 of equal intensity, which would indicate
which
Ann- der Phys. 28 y 1079 (1909); Phys. Zeits. 9, 212 (1908); 11, 1193 (1910);
and LOBISG, Phys. BCT. 46, 888 (1934), have since made much better measure
a more detailed comparison with the intermediate-oonpling theozy than
9
ments on
that given tere.
and BT.ATB, PML Mag. 8, 765 (1929), have made good use of this method of treating
f
ttnresoled Zeeman patterns.
fur MJB. 88, 658 (1925).
J BOIJMA,
PASCHESr-BAGE: EFFECT: ILLUSTRATIVE EXAMPLES
16
So much
395
Zeeman* found a quadratic displaceaent of A5790 which was later studied by Gmelin. If we had pure LS couptug there could be no such effect because the magnetic perturbation has no
for the first-order effect.
components connecting different singlet states. In view of the fact

hat the qnasi-singlet A has a large component of 3 2 in its eigenstate, it is
3
Me to show a Paschen-Back interaction with 9D^
3 It happens that
1
3
a quadratic effect
1 is only 3*2 cm" below
X hence its states will show
aatrix
lownward equal to
3-2
Vorking out the details we find that the centres of gravity of the a and TT
10 a^f 2
1*33 x
somponents should show respectively displacements of
jad - 3-48 x lO-âJ^ 2 where a is written for |( 3 2 j^)j 2 What Gmeiin
10""
a components and a quadratic effect of

components. This corresponds to a =0-40, in
rood agreement with the value 0-32 derived from the departure of the g*s
rom the normal Lande values.
It thus appears that the existing data on A5790 is in accord with the theory
observed
was no
-1-41 x 10~ 10
shift for the
for the
TT
ind exemplifies a Paschen-Back displacement of a quasi-singlet line arising

rom departure from Bussell-Saunders coupling. More accurate measure-
nents on this line are desirable for a more definite check. This interpretation
Gmelin s data disagrees with that given by Back,f who tries to relate the
>f
matter to hyperfine structure.

7.
Quadmpole
lines.
The theory of the Zeeman effect for quadrupole lines has been developed
ncidentaUy to other work in preceding sections, so brief comment relating
t to the experimental work is all we need here. The splitting of the energy
evels, being a property of the levels and not of the transitions, is given by
ihe discussion of other sections of the chapter. The details of the picture
with different
soncerning angular intensity and polarization, associated
4
have already been worked out in 6
9
shanges in
The first experimental work on this point is that of McLennan and
1
1
n
D3
ShrumJ on the green auroral line. As discussed in 5 this is the jS
4
The
first
neutral
of
transition in the normal 2p configuration
oxygen.
Zeeman effect observations were taken in the longitudinal direction. From
2 and Alf =
bhe results of 6 4 we see that in this direction the All=
->
somponents have vanishing intensity, so the pattern
is like
that of a dipole
* ZEEMAS,
Phys. Zeits. 10, 217 (1909).
t BACK, Handbuch. der Experimentalpliysik 22, i&l (1929).
138 (1924); A108, 501 (1925).
J MCLEHKAH and SHBUM, Proc. Roy. Soc. A106,
McLEKNAN, Proc. Boy. Soc. A120, 327 (1928);
SOMMER, Zeits. for PfcjH. 51, 451 (1928).
THE
396
ZEE3IA3*
7 16
EFFECT
The measured displacements gave gr= 1 from which the singlet cha
racter of the line was inferred. The transverse effect for this line was studied
at twice the normal
by Frerichs and Campbell* who found the o- components
line.
separation, corresponding to
normal separation for
A31-
AJ/= r
1, all
2,
as well as the
TT
components at
of equal intensity, in agreement with
the theory.
of the 2D-* 2 S doublets in
Segre and Bakkert studied the Zeeman effect
sodium and potassium. This work provides a detailed verification of the
2
interval is so small that all
theory in several respects. For sodium the D
fields giving a measurable splitting showed a complete Paschen-Baek effect
For potassium the doublet interval is great enough (2-32 cm- }

that the two lines can be studied separately. This was done in the oblique
direction at 45 degrees, as well as transversely, and the theoretical expecta
1
in this term.
and
verified.
polarization of the components
For higher fields the Paschen-Back effect begins to alter the intensities
The theory is exactly like that for
to
new
and
tions as to relative intensity
permits
dipole Hues
components
appear.
we have to calculate the matrix components of the quadrupole
to the perturbed eigenfunctions. In analogy with the

discussion of the preceding section the perturbed matrix components can
moment with respect
be expressed in terms of the unperturbed matrices by means of the trans

formation matrix from the unperturbed to the perturbed eigenfunctions.
Such calculations hare been made explicitly by Milianczuk J and the results
well with the experimental results of Segre
found to
*
FBEEICHS acd CAMPBELL, Phys. Rev.
f SEGPJE and BAKKER, Zeits.

J
MniA>-czrK,
Zeits. fur
fiir Pfcys.
and Bakker.
38, 151, 1460 (1930;.

72, 724 (1931).
Phys. 74, 825 (1932).
CHAPTER
XVII
THE STAEK EFFECT

The influence of an external electric field on atomic spectra was discovered
by Stark in 1913 and is known as the Stark effect. It was natural to expect
such an effect as soon as the Zeeman effect had been discovered and inter
preted theoretically in terms of the electron theory. Voigt gave attention
to the electric perturbation of atoms as early as 1899 and came to the
conclusion that the effects would be very small. He also made attempts to
discover an effect on the D-lines of sodium but without success.*
Stark s discovery was made on the Balmer series of hydrogen. It came in

the same year as the Bohr theory of the hydrogen atom, so that it was an
important additional success of that theory when, independently, Epstein
effect in terms of quantized orbits and
line
the
observed
predicted
patterns exactly. Later KxamersJ carried the
the
correspondence principle to provide estimates
theory further by using
and Schwarzschild f calculated the
of the relative intensities of the lines in a pattern.

It soon turned out that hydrogen occupies a somewhat exceptional
position. Generally the splitting of a spectral line is much smaller than in
hydrogen although there are special cases in other spectra where the effectis comparable in
magnitude with that of hydrogen. We shall see that this is
connected with the accidental degeneracy in hydrogen whereby terms of the
same n but different I have the same energy, except for the small relativity
fine structure.
We shall also see that the cases of large Stark effect in other
spectra are those in which two terms which
may combine
according to the
optical selection rules lie close together.

The theory of the Stark effect in hydrogen
was the first application of the

mechanics. It turns out that the positions
perturbation theory in quantum

of the components in hydrogen as far as the
first order effect (that pro

concerned is the same as on the quantized orbit
But the effects of second and third order differ in quantum mechanics
portional to field strength)

theory.
is
New
experimental work which carries

the study up to high fields has shown that the effect is in agreement with the
new theory. At very high fields another aspect of the effect becomes im
from the
results of the older theory.
levels by the field. This has

portant, namely, the destruction of quantized
toward the red by allowing
Balmer
series
the
of
limit
the effect of shifting the
*
VOIGT, Ann. der Phys. 69, 297 (1899); 4, 197 (1901).
f EPSTEIN,. Ann. der Phys. 50, 480 (1916>;
.SCHWABZSCHULD, Sitzber. Berliner Akad. 1916, p. 548.
and Zeits. fur Phys. 3, 169 (1920).
J KBAMEKS, Danske Vidensk. Selsk. Skrifter (8), ra, 3, 287,
398
THE STAKE: EFFECT
l 17
the continuous spectrum to encroach on the part of the frequency scale

which is
to the line spectrum in the absence of a field.
1.
Hydrogen.
We consider first the application of the ordinary perturbation theory to

hydrogen to find the effect of weak fields on the energy levels.
in the negative z-diiection so the
Suppose the electric field of strength
electron is
on by SL force in the positive z-direction. Its potential energy
<f
in the field of the nucleus
is
then
U=---ez.
r
(1)
x
/
Because of the fact that the potential energy tends to

x for large positive
z. al
energy levels are allowed. But the spectrum, has a quasi-
values of
discrete structure which

be investigated by treating ~~e$z as a per
turbation in the usual way.
Owing to the degeneracy of the unperturbed states of hydrogen it is
the particular choice of unperturbed eigenfunotions with
necessary to
to which ecfz is a diagonal matrix with
regard to the sets of
initially
degenerate states.
TMs is accomplished,
as Schrodinger
by solving the hydrogen-atom problem

parabolic coordinates. We write
slioini,
and Epstein* have
by separation of variables in
The
and
surfaces so defined are paraboloids of revolution for and

-q
planes for 9. The paraboloids of both families have their foci
at the origin, those with f= constant extend to z=
oc, those with 17 = con
stant to z = -f x. For these coordinates we have
and the Sclnodinger equation for the perturbed atom becomes
TMs
permits a separation of variables
$=F()G(
to the equations for
F and
0:
(4)
SCHBODKGER, Ann. der Phys.
80, 457 (1926);
EESTEE?, Piys. Eev. 28, 695 (1926).
HYDROGEN
399
= In the usual way shows that

can
be
solved
in
terms
of
ey
Laguerre polynomials. The solutions of these
discussion of these equations for the case
nations are
tiere
ugj
td
n = ki
and &2 being positive integers or zero. The energy levels are, of course, the
me as are obtained by solving in spherical coordinates

or
/? 2
in terms of k or
fc
are given
2 5 ).
The values of
by
(6)
A--^".
the normalized wave function of a state characterized

the quantum numbers n km where k is written for k is
ae final result for

7
For a
first-order calculation of the perturbation of the atom, due to an

we need the matrix of 2 with regard to the states of the same n.
3ctric field
ais
may be calculated from known properties of Laguerre polynomials and
und
and m. The diagonal elements have the value

k2 ) a,
(nkm\z\nkm)=^^n>(k I
(8)
to be diagonal in k
that states for which 1^

the electron to be at z
>
>
0.
are those which give a greater probability

this result it follows at once that the
Erom
st-order alteration of the energy of the state
n km
is
-f&fo-jyekfa,
(9)
the states in which the electron is more likely to be in the z

region
e lowered in energy. This calculation neglects the relativity-spin fine
.at is,
>
and
so is applicable only for fields producing perturbations large

with
this fine structure.*
mpared
We shaE not give the details of the calculation of the second- and thirdder effectsf but shall pass now to a comparison of the theory with experi-
nicture
The
calculation including the fine structure was carried out independently by RQJANSKY,
and SCTT,AFP, Proc. Roy. Soc. A119, 313 (1928). It turns out that there
weak fields owing to the degeneracy of the levels 2Zjr,4-| and \L -f !),+
ys. Rev. 33, 1 (1929),

i linear effect for very
the hydrogen fine structure ( 45 ). As the Stark displacement becomes comparable with the fine
Ticture a non-linear effect sets in, which for larger fields goes asymptotically into that given in
3
text.
j The formulas (11) and (12) for the coefficients of ^- and &* in the Stark-effect displacements
in the usual seconde z which are non-diagonal in
re not obtained by using the elements of
d third-order perturbation theory (this would involve a very inconvenient summation over
in
the levels of the discrete and continuous spectrum), but by writing 1?\ G,
9 p l9 and
/22
each as a power series in f and then solving in succession the equations obtained by equating
s coefficient of each power of f to zero.
THE STABK EFFECT
400
ment. If we measure
alteration in
cm"
(9)
in kilovolt/em
I 17
and write v^n k^ 1*2 ) for the
first -order
becomes
v^n I i t2 ) = 0-0642 n
(l\
- k2
10)
where the numerical value of the coefficient is calculated from the theoretical
expression. The original absolute measurements of Stark gave an experi
mental value of 0-068 but recent precision measurements by Sjogren and by
Kassner^ agree -within 1 per cent, with the theoretical value. Xearly all of the
experimental data refer to the Baimer lines, f For fields up to 100 kilo volt cm
the quadratic
From
54
effect is negligible in
we know that the AOT =
comparison with the linear

transitions will
effect.
show parallel polariza
tion in transverse observation and will be absent in longitudinal observation.

I give radiation
Also for transverse observation the transitions \m =
like
the Zeeman effect.
polarized perpendicular to the field. So far, this is just
There is a difference in the longitudinal observation. In the Zeeman effect,
the lines due to transitions where m changes show circular polarization.
Here the line due to a transition (m )->(* 1) is at the same frequency as
m}-*( ?-i). Theoretically each of these gives circularly polarized

but in the source there is an incoherent superposition of the two contributions giving no polarization in longitudinal observation. This is also
true for fields where second- or third-order effects are appreciable. These
predictions concerning polarization have been checked experimentally J so
transverse observation is used to get complete patterns.
Consideration of the range of quantum numbers shows that the pattern
will be symmetric around the position of the unperturbed line. The com
{
light
ponents in a pattern
differ in intensity
because
(a)
individual transition
some components are the sum of several transitions,

excitation
conditions
(c)
may put unequal numbers of atoms into various
states of the initial level. As always, this latter effect is most difficult to
control and is probably the principal source of discrepancies between theory
and observation on relative intensities.
probabilities differ, (b)
made by Schrodinger
These agree when a correction in Epstein s work
Theoretical calculations of relative intensities were
and by Epstein
(foe. cit.).
I 17
made. In Fig.
the theoretical patterns for the first four lines of the
Balmer series are shown, the unit of abscissa being the basic splitting
1
0-0642<f cm- and the ordinates
being proportional to the relative intensities
in the pattern for any one line. The cross-marks indicate measured relative
is
SJOGBEN, Zeits. filr Phys. 77, 290 (1932);

KASRTCB, ibid. 81, 346 (1933).
f FBEKICHS, Ana. der Phys. 19, I (1934), has succeeded in obtaining the Stark effect of the
Lyman series. Here Hie patterns are entirely due to the splitting of the initial levels since the
final Is level is unaffectec!
by an electric field.
| STAKK and WENBT, Ann. der Pbys, 43, 991 (1914).
GORDOX and MESKOWSKI, Xaturwiss. 17, 368 (1929).
HYDROGEN
17
401
The general agreement is good but attempts to make an accurate

the theory encounter many difficulties. When the hydrogen atoms
re moving rapidly as in
canal-ray sources, the relative intensity of the
ifferent lines is disturbed by what
appears to be due to differential destrucon of excited atoms by collision from the different states of the initial level.
itensities.
st of
fc
has been shown* that when the atoms are moving in the direction of the
Hn
T
3 2
/4
tO
8 6
i
Jl
1
32282420
16 t2
8 4
26ID&222638
Stark-effect patterns for the first four Bakner lines. The theoretical relative
strengths of the components of each line are shown by the lengths of the bare,
with very weak components indicated by half circles. The cross -marks indicate
the measured relative intensities (principally from Foster and Chalk) of certain
components of the same polarization.
17
Fig. I .
field
the high frequency components of the Stark pattern are weaker than
when the atoms move in the direction opposite
the others (and are stronger

bo
that of the
field).
Since the high frequency components arise from states
energy in the initial state, the experimental result indicates that

these states (for which, the larger values of ifrf are on the forward, exposed/
3f higher
side
of the atom) are more subject to quenching by collision than those of

shown that the intensity dissymmetry is not
lower energy.f WierlJ has

*
STARK and KIBSCHBATJM, Ann.
cLer PJiys.
43, 1005 (1914);
WILSAB, Gott. Nackr. 1914, p. 85;

LusfELAKD, Ann, der Phys, 45, 517 (1914).
f BOHK, PhiL Mag. 3D, 394 (1915);
SLACK, Ann. der Phys. 82, 576 (1927); Phys. Rev. 35, 1170
t WIEBL, Ann, der Phys. 82, 563 (1927).
cs
(1930).
26
THE STABK EFFECT
402
present when the field is perpendicular to the motion of the atoms, nor
the canal rays go into a high vacuum.
17
when
The intensity measurements by Foster and Chalks Mark and Wieri, and
others* do not agree in detail with the theory, but provisionally we shall
regard this as due to the large number of variations in physical conditions
in the experimental work.
An interesting experiment has been performed by Wient by sending a

so that the atom s
hydrogen canal ray beam through a magnetic field
v
is
From
the
to
$*
velocity
standpoint of a coordinate
perpendicular
with
the
rise
to
an electric field of amount
atom
this
system moving
gives
v x 3f c. which can produce a Stark effect. Of course the magnetic field
to a Zeeman effect, but this can be made smaller than the electrogives
&
dynamic Stark
-irrmc
constant. For
Hy
effect.
The
ratio of the basic
of the Zeeman effect

t*-
energy change f&fa to the

where a is the fine structure
but as the strongest lines in the
is (3;a)(i?-c)
iO 8 cm/sec the ratio
is 1-4,
pattern are displaced 15 to 20 times the basic unit we see that at such
the electric effect is about 20 times the Zeeman effect. Such speeds
were used by Wien. The experiment is extremely difficult and the pictures
obtained are quite unsharp. The observed effects agree with theory to within
the rather low accuracy attained.
We turn now to a discussion of the experimental results for the quadratic

and cubic Stark effect in hydrogen. This has been studied up to fields of 10 s
volt/em. The theoretical energy change in the quadratic effect has been
to be
This differs from the quadratic effect in the classical theory by the ap
pearance of the additive -f 19 in the brackets and by the fact that the m
values are all one unit lower than in the older theory. Similarly the theory
for the third-order effect has been
worked out by Doi who
finds for it the
energy change
n].
* FOOTIEB and
CHALK, Prw. Roy. Soc. A12S, 108 (1929);
StABK and WKBL, Zfelts. fur Pays. 58, 526 (1029); 55, 156 (1929);
Kiun, J&p. Jour, Phys. 4, 13 (1925);
ISBGCDA and HITAMA, Sci Pap. lust-. Piiys. Giiem. Res. Tokyo 9, 1 (1928);
STAEK, Ann. cier Pfeys, 1, 1009 (1920); 48, 193 (1915).
f WXEN, Ann. der Phys. 49, 842 {1916).
J Enwrap Phys. Rev. 28, 695 (1926);
WKNTZEL, Zeifcs. fir Phys. 38, 527 (1926);
MO
WAT.T.TO, ibid. 38,

(1926);
V&s YLSCK, Proc. Nat.
Sci. 12, 662 (1926).
Remit only jHished in ISHTOA Mid HIYAMA (Joe.
AeL
cil.}.
(12)
HYDBOGEN
I 17
403
convenient to express the shift of a particular component of a Balmer

3
2
1
line in the form
(13)
ACcm- ) = off - 6e?
It
is
c<f
in million volt/cm in order that the coefficients be of com

order of magnitude. Ishida and Hiyama have given a table, which
measuring
parable
<?
we reproduce as Table I 17 of the coefficients a, b
c,
for the strong
components
of H a, H f$ and H y There are two columns for b and c corresponding to the

and the old quantized-orbit theory predictions.
quantum-mechanical theory
The column headed transition gives the quantum numbers kj^m of
effect is positive,
initial and final states for the transition in which the linear
that headed label gives the linear effect in terms of the basic splitting, and
.
the polarization of the component.

TABLE
Transition
I 17 .
Coefficients of
Stark displacements in (13).

b
Old
Old
Label
care
The quadratic effect was discovered by Takamine and first studied
cubic
The
and
Gebauer.*
von
Traubenberg
and by Rausch
fully by Kiuti
18 IT and 1577 of Hy in a later paper
effect is studied on the components
We shaE content ourselves with
by Gebauer and Rausch von Traubenberg.f
a report of this work. By reversion of the
series (13)
from the observed shift

strength can be calculated
we
see that the field
A by the formula
* KICTI, Zeits. fur Phys. 57, 658 (1929);
*&. 54, 307 (19M); 56 *

EATJSCH vox T^TTBEKBEEG and GEBi^B,
Zeits. fur Phys. 82, 289 (1930).
TuiraESBEMJ,
vos
RUJSCH
f GEBitJEE and
26-2
404
STABK EFFECT
If the theory is correct this formula gives the same value of the field
strength
on a particular plate when applied, to the various components of the patterns.
This is the case to an accuracy of about 0*1 per cent, in the experimental
work. The procedure was to calculate & from the observed effect on H jS and
then to compare the observed effect on Hy -flow and -f-lSw with the
theoretical values as calculated with the values of g determined from H
/?.
We may
the theory of the linear effect as already adequately tested

in other experiments and subtract its theoretical amount from the observed
shift to obtain the observed amounts of the shift due to the
order
violet
higher
We
have prepared Fig. 2 17 in this way from the data on the 1877
component of Hy. The points give the observed data and, for com
effects.
parison, are given theoretical curves for the
Fig.
2".
amount of the quadratic
The second- and third-order Stark effect on the
effect
18 IT violet component of
Hy
alone, the quadratic plus cubic effect, and the

quadratic plus cubic effect
as calculated on the old
It
is
clear
that
the
data axe best
theory.
represented
by the curve
representing quadratic plus cubic effect on the quantummechanical theory. There appears to be a small
systematic departure at
the higher field strengths which is

perhaps due to the fourth-order
The data are evidently not well represented
by the old theory.
Stark effect at the series

As already mentioned in the
2.
effect.
limit.
preceding section, the potential energy of the

oo as z tends to
large positive values.
electron in the applied field tends to

This has as a consequence that the
spectrum of allowed values
is
really
405
STARK EFFECT AT THE SERIES LIMIT
2 17
continuous, there being admissible wave functions for all values of the
line
energy. Nevertheless we have seen that the spectrum gives sharp
situation
patterns in Stark-effect experiments. We wish now to examine the
more carefully.
The potential energy
=d=
Ve/<f 5
where the
that of the external
U of 1 17 1 has an extremum at the point x = 0, y = 0,

electric force
field.
due to the nucleus
At this point the

Z7 = s
is
just balanced
by
potential energy has the value
-2VeV.
(1)
natural to expect that this energy value and the distance d play an
the
important role in the Stark effect. This is in fact the case. Examining
It
is
formulas
1 17 9
and
1 17 1 1
we see that the coefficients of the quantum numbers
can be written - f J7 (aid) and ^Z7 (a/d) respectively, and similarly for the
cubic term of 1 17 12. In a general way this might lead us to suspect that the
of space
developments have a validity only for small (a/d) and, as the region
th
of
linear
dimensions
has
state
occupied effectively by an electron in the n
th
2
state
n
the order n a we may expect the perturbation theory to fail for the
8
when d~n 2 a*
The first theoretical discussion of the special features of the problem in
this region was given by Robertson and Dewey* from the standpoint of the
can be quantized
quantized orbits. In that theory only those orbits
which are conditionally periodic, so the upper limit of the discrete energy
is the maximum value of the energy for which conditionally
classical
spectrum
periodic orbits exist. They found that

total energy is greater than
there exist no quantizable orbits whose
and quantizable orbits in the neighbourhood of this limit only exist provided
that
ne* \*
*
mechanics the orbits for which L3 = are ruled out because

Hence the least admissible value
they lead to collisions with the nucleus.
of Lz is JL This value of Lz leads to an upper limit of the discrete spectrum
In the
classical
which in wave numbers
is
given
by
whereas the value of the extremum of potential energy
where
g is measured in
*
?7
is
given by
I0 6 volt/cm in both formulas. These two expressions
ROBERTSON and DEWEY, Phys. Rev.
31, 973 (1928).
THE STABK EFFECT
406
217
are plotted as functions of S In Fig. 3 17 From this we see that the discrete
spectrum may extend up to higher energy values than the extremum of the
.
potential energy function. At first sight this may seem a surprising result.
If the particle has enough energy to get to the configuration where the field
we are apt to think that it wiU surely

do so. But
is not the case. A simple mechanical example is that of a ball
rolling under gravity inside of a spherical bowl, one side of which is partly
away (as with the Mad Hatter s tea-cup). The ball may perform
oscillations in a vertical plane which does not pass through the broken wall.
whose total energy exceeds the energy limit where the non-periodic motions
force balances the nuclear attraction
begin, that
is,
at the energy corresponding to the lowest place in the rim.
of the bowl.
2009
17
Fig, 3 .
Classical-dynamical energy limits for quantized orbits in the hydrogen atom

in an electric field.
So much for the classical-mechanical theory of the problem. The quatttmnmechaniea! theory has been treated by Lanczos* and is based on a direct
discussion of equations 1 17 4 for the case
^0. The problem is somewhat
<f
complicated by the peculiar dynamical interaction of the motion in the f

and 7] coordinates, but the essential point is the coupling of discrete and
continuous energy eigeastates of the same energy value. TMs feature of
quantum mechanics was first discussed by Oppenheimer t who applied it to
5
a calculation of the probability of ionization of hydrogen in the normal state

by the application of a steady field and showed how this mechanism could
account for the emission of electrons by cold metals in strong electric fields.
It has become well known through its application by Gurney and Condon
*
LAXOSQS, Zcits. fir Phyi. 62, 518 (1930); S5, 431 (1930); 88, 204 (1931).
j OFFESHEDCXB, Fiiys. Rev. 81, 66 (1928).
ind
407
STABK EFFECT AT THE SERIES LIMIT
jW
by Gamow*
to the problem of oc-particle disintegration of radioactive
dements.
The
effective potential energy functions in the
two equations are
forms indicated in Eig. 417 The wave equations cannot be

solved exactly but have been treated by the Wentzel-Brillouin-Kramers
method. The second of them leads to discrete energy levels which depend
sdiich are of the
parametrieally on j8 2
number of nodes in
<?(TJ).
first
and the quantum number k2 which gives the
We shall call this relation
W = TF (& ^ m
2
)*
Tlie
for a fixed &; instead we
values of
equation does not lead to discrete
with the phenomenon of the leaky potential barrier.
are presented
We are
-1ZTFig. 4
Effective potential energy functions in the

in an electric field.
17
.
and
7]
coordinates for hydrogen
of
interested only in the quasi-discrete levels, associated with values
considerable
with
found
be
can
accuracy
These
of
below the maximum
U^).
such as the dotted one in

by replacing i7 1 (|) for large f by some other curve,
will
the figure, so that the new equation has discrete energy levels. These
the
Then
W^^k^m).
depend parametrically on &, &13 and m, say
are the values of IF for which
levels of the
allowed
F=
problem
energy
Wi(p ki,m) =
t (l-pi,kt,m),
has
already been taken into account.
where the relation
Now let us consider the fact that the true U^g) slopes downward for large
For a
& instead of having a horizontal asymptote as in the dotted curve.
have
and
I\
value of W=Wi(Pi,ki,m) the wave function will oscillate
l9
& + jB2 = I
GUBNEY and CONBOK, Mature 122, 439 (1928);

GAMOW, Zeits. fur Pkys. 31, 204 (1928).
Phys. Rey. 33, 127 (1929);
2 17
THE STAUK EFFECT
408
In the region II it
approach zero in a quasian oscillatory
exponential way and in the region III it will again assume
character but will have a greatly reduced amplitude relative to the amplitude
from 1 (^ I
in I. For a value of IF
m\ the behaviour in
nodes in the region
I.
"will
differing slightly
I iviil be quite similar, but owing to the slight difference in the wave-length
in region II so
of the de
waves it will increase
quasi-exponentially
Broglie
that the oscillatory portion in region III Trill have a much larger amplitude
of the eigenfunetion from
relatively than before, TMs change in character
inside
small
to
inside
small outside
large outside* takes place as
*
a continuous function of IF but occurs in an extremely small range of energy

values near the value W^^k^m). We shall want the wave functions
is
so that the integral of their square out to some large value of f

(TMs suffices to give the correct relative normalization,
equal to unity.
wave function from a large

the same, and thus we avoid dealing
for the behaviour of the oscillatory part of each
f= x
value of
is essentially
out to
with divergent integrals and the special technique for normalizing con
tinuous spectrum wave functions.)

In computing tie transition probability from any energy IF to a lower
state which shows a much smaller perturbation it is only the part of the wave
function in region I which plays an essential role, for in calculating the
electric-moment matrix component the wave function of the less-perturbed
By applying the WentzelBriUouffi-Kramers approximation, Lanczos was able to show that the
intensity of radiation from initial states in the energy range da* at
state wiH be essentially zero outside of this region.
JTjJ/Jj
k2 m) T &
is
proportional to
I
X=
where
The
dojX
e~~
of the spectral line (i.e. the interval between the two

the
radiation intensity has sunk to half value) is evidently
at
which
energies
The
value
to
2JT.
of
equal
depends both on v 9 which is the classical
"
half-width
frequency of vibration in the | coordinate for the state in question, and 1

which measures the penetrability of the barrier between regions I and III.
If at I = we are sure that the atom has an energy about equal to that of
the energy level in question and that the electron is in the region I, we
represent the state by a wave packet built to give constructive interference
of the waves in region I and destructive interference in III. Such a packet
STARK EFFECT AT THE SEBIES LIMIT
17
409
be built by superposition of states in the energy range of the order of X.

& time goes on the constituent waves in the packet will get out of phase, so
le amplitude of the packet will decay in such a way that the probability
f the electron being in I contains a factor e~a where 8 = 23/e- I
This
/7r.
iminution in region I is compensated by an increase in region III, so the
an ionization of the atoms by the steady field, or a

from the bound state of region I to the
rocess corresponds to
idiationless transition of electrons

:ee
state of region III. Suppose a is the probability per unit time of leaving
by radiation. Then the total probability of leaving region I in unit
3gion I
ime
is (a -f S),
/(oc-fS).
The
his fraction
so the fraction of excited atoms which leave
by
radiation
is
from this level is therefore

would have been in the absence of the field
total intensity of the radiation
of the value
it
>nization.
Thus the effect of the field is to give an uasharp line whose width is
and whose total intensity is diminished by the field-ionization
rocess. The actual calculation of the quantities v and I for high fields is
omewhat laborious although the values can be expressed in terms of
Uiptic integrals. Lanczos has made some calculations for the IS^r violet
i/e~ 2J /7r
-nd
red components of Hy which are in general agreement with the experiThe rather surprising result is that the violet component
aental work.
emains sharp and strong to much higher field strengths than the red
sonent which fades out at about 0-72 million volt cm. As the violet
com-
com
an initial state of higher energy than the red component

the
might expect
opposite, for the high state has a more penetrable
carrier to go through. The explanation is that the initial state of high energy
orresponds to the electron being mostly on the opposite side of the nucleus
ponent arises from
TO
rom the low
place in the potential where the leaking takes place. In the

component s initial state the electron is cautious and conservative,
Avoiding the leaky barrier; in the red state the electron is reckless and
dolet
mstrelj hurling itself against a less leaky barrier often enough to rob itself
the chance to shine
!
>f
3.
General theory for non-hydrogenic atoms.
In
this section
we shall sketch the general results obtained by calculating

perturbation of any atom by the perturbation theory. If the
in the negative z direction and P is the s-component of electric
,he electric
ield
<f
is
noment of the atom, then the Hamiltonian of the atom
in the field
H=H* + gP,,
rfiere
is
perturbation
(I)
field. The
by the matrix components
written for the Hamiltonian of the atom in zero

is
therefore determined entirely
is
THE STABK EFFECT
410
3 17
of the z-component of the electric moment, a quantity with
"which
we
are already quite familiar because it measures the strengths of spectral

lines.
If for a particular atom we already have a rather complete knowledge of

the energy levels and corresponding eigenfunctions, we may calculate the
matrix of Ps and proceed to apply the perturbation theory. The details of
the results so obtained naturally will depend a great deal on the mode of
coupling of the angular-momentum vectors and the amount of configuration

But we can, nevertheless, make some remarks which hold quite
interaction.
atoms before treating the separate special cases that arise.

place Ps has matrix components only between states of
opposite parity. Second, for a state of given / value it has matrix com
ponents with other states whose J values are / 1, J and J -f 1. Third, the
order of magnitude of the non-vanishing matrix components involving
electrons in states of total quantum number n will be the same as computed
generally for
In the
all
with hydrogenic wave functions. TMs we shall see is of the order 6-4 n 2 cm" 1
for a field of 100 kilovolt/cm. Because of this the second-order perturbation
whose energy difference is more than 10 (6-4) 2 w 4 ~ 400 n4

5
1
is almost sure to be less than 0-1 cm- even at 1C volt/cm. As most of
the experimental work has 0* 1 cm" 1 as the limit of accuracy and is carried out
at smaller fields, we see that the interaction of two terms several thousand
wave numbers apart is negligible. The combined effect of an entire series of
terms on a given term may be appreciable however. From our previous
"between
two
states
cm"""
2
study 10 ) of the general features of perturbations we know that the effects
will be given quite accurately by the second-order formula provided the
(
separation of two interacting states is large compared with the energy

of interaction between them. So for terms separated by more than about
initial
6*4
?i
2 it
wiP be sufficient to use the second-order formula. This qualitative
restriction coupled with the parity and J selection rules suffices to make
most cases tractable in this way. In such cases the displacement of the terms
small compared to the effect in hydrogen and is strictly proportional to the

square of the field strength.
For a group of states lying close to each other a more accurate calculation
is needed. This can be made by forming a secular
equation from the matrix
of the perturbed Hamiltonian by including just those rows and columns
is
which
refer to the small
community of close,
this finite-matrix transformation
interacting states. Solution of
determine the mutual perproblem

turbation of these states and the transformation to a new set of states with
will
regard to which this part of the Hamiltonian is diagonal. This transforma

tion also enables us to calculate the values of the
perturbation matrix
components connecting these perturbed
states with the distent states
and
GENERAL THEORY FOB NON-HYDRO GEKIC ATOMS
LUS to
know how
the second-order action due to them
is
411
distributed over
te community. TMs procedure is exemplified in the next section.
There is another qualitative remark of importance. Generally the electric

Isplacement of excited states is much greater than that of low-lying energy
vels so that in most cases the displacements of the final state of a line are
3gHgible, the observed line pattern being
on of the initial state.
a direct picture of the perturba-
Let us next consider quite generally the kind of pattern to be observed in

quadratic Stark effect of a state described by the quantum numbers
J M. Using 9 3 1 1 in the second-order formula of 9 2 4 to express the dependof the matrix components, we may write for the displacement
ice on
le
(2)
r
a^S^(aJlPjjSJ-l)!*
11
fere
"
-&QLJ-&PJ-1
= J 1; a
^presents the perturbation by the levels for which J
f
milar expressions for the perturbation by levels of J = J
and a+ are
and J+l
^spectively.
The relative strength of the different
components of the line pattern for

when
these
states combine with an unobservation
ransverse
resulting
3 11
9
erturbed level y J is likewise given by
just as in the Zeeman effect
sf.
4 16 ). Writing
(2) in
the form
A(aJJf)=^l-5J/
(3)
*t us consider the transitions from the states of the level a J to a level y J

f the same J value. The TT component whose displacement from the position
f the
unperturbed
line is
jBlf 2 is produced
by the
transitions Jf ->3f
2
proportional to 23f , assuming
line at the
qual population of the various initial states. The a component
whole
1
its
so
land
T
strength
ame place is produced by If
2
for Jf = the strength is
while
If
for
If
J
J
to
^
3
+
1)
]
[
(
proportional
the relative strengths of the
proportional to J/(J-f-I). In the same way
ines in the patterns for a J-^ J - 1 and a J-* J -M transition may be worked
ut. The results are summarized in the table:
nd
M->
so its whole strength
->11
is
->3f
Relative strength of line component at

S
Polarization
A - BJ/ 2
3 17
THE STABK EFFECT
412
These relative strengths are the analogues of the weak-field strengths of

4 16 In the Zeeman effect since they are calculated with the unperturbed
eigenfunctions.
We
can also obtain some general results with regard to the appearance of
forbidden* lines in the Stark effect. The theory, of course, is analogous to
the corresponding case in the
The
states $(j$J
Zeeman
effect as treated
which perturb the state
M)
16
by Zwaan (6 ).
$(zJM)
(where
J =J
or
are of opposite parity and therefore will not give dipole radiation in
combination with the states which combine with ^(a J"Jf). But in the
1)
become perturbed so that in the perturbed

eigenfonetion arising from (j8 J If) there appears some of the eigenfunction
electric field the eigenfunctions
of the states
(ot
J"
If).
By the results
of
92
we have, to the
first
power in
where the subscript refers to the unperturbed eigenfunctions. Hence, if

is a final state, whose perturbation by the field we may
(y
neglect,
J"J1")
of the same parity as j8 (with which it therefore does not combine in the
absence of a perturbation), the matrix component of electric moment con
necting the two states
is
given by
(5)
This expression can be somewhat reduced by using 93 11 to express the

of the first factor of the numerator in each sumdependence on Jl and
ff
mand. Thesquared magnitude of (5) measures the strength of the component

produced by the transition y J*
the square of the applied field.
M"
-+f}J
M.
It is evidently proportional to
It will often happen that only one term in one of the three sums is appre
ciable. The strength of the forbidden pattern is then
given by the square of
3 17
that
GENERAL THE0BY FOR XON-HYDBOGENIO ATOMS

413
term alone. TMs is the case when the perturbation is principally due
to the interaction of
two neighbouring terms. Erom
(5)
we
see that the
relative intensity distribution in the pattern of the forbidden line is not the
same as in the case of a non-forbidden line.
To make
this more explicit suppose that the

important term in (5) in a
case
is
one
of
those
on
the
first
line
for
a
given
particular value of a. Then the
intensities are proportional to
/}
"
(6)
The pattern to be expected is therefore that which applies for a combination

with the perturbing (not the perturbed) level modified by the factor
which
the ratio of the perturbation of the state in question to the un

perturbed interval between the interacting levels.
This
is
about
one can say without consideration of special cases. It is

of the knowledge concerning the matrix components of
which we have acquired in connection with the intensity problem is available
is
evident that
all
all
for application here.
of
10 3 to write
Thus in the Bussell-Saunders case we can use the results
down formulas for the
electric-field interaction of adjacent

Russell-Saunders terms. Likewise knowledge of the eigenfunctions in con
figurations showing intermediate coupling permits us to make corresponding
predictions concerning the Stark effect. The method is so straightforward

that there is no point in writing down explicitly formulas for the various
shall turn therefore to the application of the theory to
special cases.
We
typical cases in the experimental data.

4.
Helium.
The
first
application of
quantum mechanics
to a non-hydrogenic Stark
due to Foster,* who carried out the calculations and applied them
to his own experimental data on helium. He uses matrix components given
by use of hydrogenic wave functions, for which
effect was
5
(cf. 6 2, 3),
In this spectrum the triplet intervals are so small that they may
be neglected. The lines for which the Stark effect is studied are the com
binations of the states with n = 4 and 5 with the 2 S and 2 P terms.
First let us estimate the Stark displacement of the final terms to see that it
17
1
1
is negligible. 2 /S lies 5857 cm" below 2 *P, so by 3 2 the mutual perturba
2
1
= states amounts to 0*0285^ cm- where ff is measured in
tion of the
f
FOOTEB, Proe. Boy. Soc. A117, 137 (1927).
417
THE STABK EFFECT
414
3
10 5 volt/em. The If = 1 state of 2 1 Is not perturbed by 2 *S. Similarly 2 $
3
lies 9231 cm- 1 below 2 P, so the mutual interaction here is even smaller in
the ratio
in the
These
effects are quite negligible alongside
of the Interactions
n = 4 and n = 5 groups.
l
1
Using the observed intervals between 4 8. 4 *P, 4 !JD, 4 F and the hydroIf and J as
genie matrix components (1) together with their dependence on
s
3
for
the cases
given by 9 11 and 11 8 Foster set up the secular equations
5
If = 0, 1, 2 and calculated the roots for fields up to 10 volt /cm. The results
are shown in Fig. 5 17 where the ordinates are wave numbers measured from
3
100
0-2
0-#
in
17
Fig. 5
OB
0-5
10
volts
/cm
1-0
*
D and 4
Stark-effect interaction of the 4 *#, 4 1 P, 4 1
the unperturbed
D level and the abscissas are ff in 1
JF levels in helium.
volt/cm They form an

.
excellent illustration of the perturbation theory. The S term is so far away

that It is very slightly perturbed, but even so its perturbation is large com
states. The tendency of the terms to repel
pared with that in the 2 S or 2
terms. There
nicely brought out, especially in regard to the
= and 1 start out with larger displacements than
the components with
2 state but are prevented from reaching as large displacement as the
the
each other
is
M
M = 2 state because of repulsion by the M =
0, 1
components of the
P level
417
HELIUM
above. The if = 0,
415
level are really not Identical but

components of the
coincide within the accuracy of the
drawing. The accuracy with which the
data is represented by the theory is shown
by the experimental points for
10 5 volt/cm as given by Foster, the o s
X
- 41
representing data from the 2
1
*
*D 1 group and the x s from the
j?
*P,
group.
S
2^-4
Foster also calculates the variation of the relative

strengths of lines in the
patterns with the field-strength. This is done by
the transformation
finding
to states in the groups which makes the perturbed Hamiltonian
diagonal and
calculating their matrix components of electric moment with 2 1 S and 2 1

as final terms. The observations do not include
intensity measurements, so
it can only be said that the observed intensities
agree qualitatively with the
theory. An interesting feature of the intensity theory is that certain com
ponents
may
fade out at a certain field-strength and reappear

again at a
larger value. This was checked experimentally.
Ishida and Kamijima* have obtained plates of the effect in helium

up to
5 x 10 5 volt/cm and find their data in good agreement with the
theory.
A study of the relative intensities of the lines in the patterns has also been
made by Dewey.f Her
on an application of the
formula
to
the
Stark
effect as suggested by
Kramers-Heisenberg dispersion
Pauli.J This formulation of the problem, reached through the correspondence
principle, agrees with the quantum-mechanical theory.
calculations were based
Foster has studied the perturbation of certain helium lines in the presence
of both electric and magnetic fields.
5.
Alkali metals.
We shall consider next the effect on the one-electron spectra of the alkalis.
Passing over the older theoretical work, the first quantum-mechanical cal
culations are those of Unsold, ^f who simply considered the interaction in a
jj
of the same n but different I, using the hydrogenic

and
the empirical values of the separations as in
matrix components
weak field of the terms
A discussion of this formula in comparison

with experimental data has been given by Ladenburg.** It gives fairly good
values in all cases where it is applicable. This type of calculation has been
extended to stronger fields by Rojansky.f f
Foster s helium calculations.
Consideration of the effect in the alkalis falls naturally into two parts the
:
very small quadratic
effect
on
lines
whose
initial
and
final states are
very
* ISHIDA and KAMT.TTM-A, goi.

Papers, lust. Phys. Ghem. Res. Tokyo 9, 117 (1928).
DEWEY, Pkys. Rev. 28, 1108 (1926).
PAUU, KgL Danske Vid. Selskab, Math.-fys. 7, No. 3 (1925).
FOSTEB, Proc. Roy. Soc. A122, 599 (1928); Nature 123, 414 (192ty.
H BECEBB, Zeits. fur Phys. 9, 332 (1922).
f UNSOLD, Ann. der Pliys. 82, 355 (1927).
** LABENBUBG,
Phys. Zeits. 30, 369 (1929).
ft ROJAHSKY, Phys. Rev. 35, 782 (1930).
f
J
THE STARK EFFECT
416
5 17
sightly perturbed, and the transition from quadratic to linear effect, as in

helium., when the interacting terms lie very close together. As an example of
the
Mad we
first
and Ramsauer* on the

The data were obtained in absorption for the
consider the data of Grotrian
principal series of potassium.
second, third and fourth lines.
According to
As remarked
3 17 the 2
P terms will be perturbed by the 2 S and D terms.

2
at the end of that section
we may
use formulas of
10 3 in the
Russell-Saimders case to find the relative values of the perturbations on

the components of a term. If this be done it will readily be found that the
2
term in a one-electron spectrum can be written
perturbation of a
(I)
in
which A, and A d are coefficients measuring the perturbation by the
S and
D series respectively,
These formulas are based on the supposition that the 2D interval is negligible
and the
P interval large compared with the amount of the effect.
For
52
2
in potassium the experimental values of the A s are
0-44^ 2 ,
0-21<f ,
and 0*37<f 2 respectively (unit of = I0 5 volt/cm) which are represented by
<f
taking
*4 S
=(H3 and J^=OG7 probably within the accuracy of the data.
The kind of pattern to be expected and the relative intensities are readily
worked out from the results of 3 17 The prediction is that the line from
.
P.
is
displaced by the
relative strength 2
amount A(
and 2
in
IT
P|),
and a
the single component appearing of
polarization, while
from 2 Pg we have a
2
component displaced by A( P|) which has strength 4 in TT and I in a polariza
tion and a component displaced by A( 2
of strength 3 purely in a polariza
P|)
with the data except that Grotrian and Ramsauer could
not detect the theoretically weak A( 2P:) component in a polarization.
tion, This accords
hard to say anything definite about the absolute value of the coeffi
cients. They are small both because the nearest perturbing terms are about
3000 cm* 1 away and also because the nearest S and D terms below 5 2P are
It
is
almost as close as the nearest 8 and D terms above, so there is a tendency

toward cancellation of effects. The quadratic effect on the sodium D lines
has been studied by Ladenbur^.f
* OEOTBIAN and
RAMSAUER, Phys. Zeits. 28, 846 (1927).
f LADE^BrBG, .Zeits. fir Phys. 28, 51 (1925).
5 17
ATiKATiT
METALS
417
As an illustration of a more hydrogen-like alkali spectrum we shall

mention the effect in litMnm. This has
discussed by Condon* in a
manner analogous to Foster s work on helium. The 4D and 4 F terms are
close enough that a linear effect sets in at fields of the order 30 000 volt/cm,
while the 5 D 5 F and 5 G are so close together that their interaction gives
a linear effect from the outset. Ishida and PukusMmaf have published
observations on the 2 P 4 PDF and 2 P 5 PDFG patterns which show
in all respects agreement with the qualitative predictions of the theory, and
in which the displacements are quite accurately given by calculating the
secular equation with hydrogenic matrix components.
"been
*
COSDOS-, Phys. Rev. 43, 648 (1933).
t ISHEDA and FUCTSHIMA, Sci. Pap. Inst. Phys. Chem. Res. Tokyo
9, 141 (1928).
CHAPTER XVJII
NUCLEUS IX ATOMIC SPECTRA
Except for brief consideration of the
finite
mass of the proton in Chapter v,
in connection with the theory for atomic hydrogen, we have treated the
nucleus so far not as a dynamical particle but as a fixed centre of Coulomb
force characterized solely by the atomic number Z. In this chapter we shall
the way in which the nucleus affects the structure of atomic spectra.
The very fact that this
topic can be put off so long indicates that the effects

are small. Nevertheless they are of great importance and afford a tool for
studying atomic nuclei. The most obvious feature to be considered is the
of the nucleus, as a consequence of which the nucleus has some

energy. The effect of this on the atomic energy levels we consider
1
But more interesting is the fact that some spectra show a fine struc
ture of the lines, finer than the ordinary multiplet structure (on a scale of 0*1
to 1-0 cm" 1 ). This is known usually as hyperfine structure and, following
Pauli 9 is to be associated with quantum numbers specifying a degree of
in
momentum
freedom for
of the nucleus. The theory of the energy

has received a great deal of attention in the
well understood, although much remains to
levels resulting from this picture
i>e
few years and

done.
is
1. Effect of finite
now fairly
mass,
The theory for a nucleus of finite mass in an AT-eIectron problem has been
by Hughes and Eckart* and also by Bartlett and Gibbons.f The
energy of a system of N electrons^ eaeli of mass p, and a nucleus of
31 is given in terms of the
T
to a
r are the
origin.
the
velocities
by
position vectors of the electrons
and nucleus relative
We Introduce the position vector R of the centre of masSj

vectors gt which give the location of each electron relative
-
to the nucleos s
___
***""""
kinetic energy becomes, when expressed as a function of the

conjugate to %, and P, conjugate to U,
Then the
HUGHES and ECKABT,
Ffeya.
Bev. 36, 694 (1930).
t BIETLETT and GIBBONS, Phys. Rev. 4*, 538 (1933).
18
419
EFFECT OF FXOTTE MASS
where \L is the reduced mass pM\(p -f ). Since the translational energy of

the centre of mass cannot change appreciably in a radiation process and
is a constant of the motion, we may as well confine ourselves to
since
states for which
0. Then we see that the finite nuclear mass has altered
P=
the kinetic energy part of the Hamiltonian in two ways, by the substitution
of ft for p in the first term and by the appearance of the second term which
we
shall call
$:
6
evidently a quantity of type (?, as considered in 7
Suppose we know the energy levels for the case of infinite nuclear mass.
This
is
Then
for finite
mass the
is
effect of the
change in the
to multiply these levels
spin-orbit interaction)
is what we found in Chapter v for hydrogen
by
first
term (neglecting
1
jLt
//i= [1 -h/i/Jf
and hydrogen-like ions.
.
j"
TMs
For any
element the result readily follows from the fact that the potential-energy
function is homogeneous of degree
1 in all the positional coordinates.
This part has been called the normal effect.
The specific effect is that due to the matrix components of 8, which we
treat as a perturbation. In the zero-order scheme, we have from 7 6 7 the
diagonal element
(A 8\A) =
i
-Ê {(a*|
j>K)-(a<|
j>|a<)
- (tfi\p\a?Ha!\p\a*)},
the reduction to individual one-electron matrix components being possible

because pipj factorizes. Now p is a vector which anticommutes with the
&
6
1 1 ), hence all of the diagonal matrix components in the
(
parity operator
1
first term vanish. Moreover the second term will vanish unless a k and a
refer to individual sets of opposite parity. Similar
remarks can be made
about the non-diagonal matrix components given by 7 6 With the main

features of the S matrix known the detailed calculation of the energies due
to this term may be made in any special case by the same methods as for
.
other types of perturbation.

Bartlett and Gibbons have carried out the calculations for the 2p 5 3p
5
^>2p
3s
neon using the Hartree field found by Brown.* They calculate the
matrix components in the zero-order scheme and from these the perturba
lines in
tions of the Russell-Saunders energy levels are obtained with the diagonalrule (exactly as in the calculation of electrostatic energies in Chapter
sum
vn). In the case of the p*p configuration all the electrons outside closed
shells have the same parity, so the diagonal matrix components of S contain
only the contributions due to interaction of the s closed shells with the p
electrons; these are the same for all the states of the configuration. In the
*
BBOWS, Phys. Rev.
44, 214 (1933).

27-2
I 18
THE NUCLEUS IS ATOMIC SPECTRA
420
not the ease, and it tunas out there that the 3 term
1
is displaced
by 31% while the P term s displacement is 1% where Jc is a matrix
component connecting the 2p and 3$ states. Calculating the specific shifts
5
j>
5 configuration this
is
Xe 22 and Xe 20
isotopes and taking the difference (heavier minus

calculate
net shifts of 0-0136 cm" 1 for the p s p levels, of
lighter), they
levels.
cm- 1 for the p s s l level, and 0-0293 cm- 1 for the
for the
psP
5
Hence in the ps p
s array., the singlet lines will be displaced 0-0105 cm*- 1
more than the triplet lines. This is in accord with the observations of
Xagaoka and Mishima,* although the theoretical value of the absolute shift
->
j>
is
much
too small,
Local unclear
2.
fields.
Unquestionably the nucleus is not truly a point centre offeree, so that the
Coulomb potential 2e2 /r, cannot be correct in the limit as

.Experi
ments on scattering of a-particles by nuclei have shown that there are
12
departures from the Coulomb law in the neighbourhood of r~ 10~ cm, and
r~>0.
various theories of nuclear structure agree in assigning to the size of the

nucleus a linear dimension of this order of magnitude. Calculations of the
on atomic spectra of such departures from the Coulomb law have been
Breit, and Rosenthaif
Nothing very definite is known about the nature of the departure from
the Coulomb law. Racah makes the simple assumption that the nucleus is
spherical in shape and that the potential inside the nucleus is constant and
effect
made by Racat
continuous with the value at the boundary of the nucleus. Rosenthal and
Breit work with a model having a discontinuity at the nuclear boundary
corresponding to the potential barrier model used in current theories of
x-particle disintegration. The calculations of Rosenthal and Breit are
carried through using the Birac equations of the relativistic theory (5 5 ).
In the first work it turned outs on assuming that the radius of isotopic
nuclei varies as the cube-root of the atomic weight (constant nuclear den
sity), that the theoretical values of the isotope shifts of spectra of thallium,
mercury were considerably larger than the experimental values.
lead.
One of the most uncertain elements entering
into the calculations is the

value of the atomic eigenfunetions at the nucleus. In the later paper, Breit
shows that much of the discrepancy can be removed if a semi-empirical
formula for
first
(Q)
due to Goudsmit is used, instead of the values used in the
paper.
Data on
*
differences in spectra associated with different isotopes
of heavy
XAUAOKA and MISHXXA, Sci. Pap. last, Phys. Chem. Res. Tokyo 13, 293 (1930).
t RACAH, Xature 129, 723 (1932);
ROSESTTHAX and BBEIT, Fiys. Rev. 41, 459 (1932);
BRER, Fhys. Rev. 42, 343 (1U32).
LOCAL NUCLEAR FIELDS
2 18
421
elements is being accumulated rapidly at present and may in the future prove
an important source of knowledge about the nucleus, but at present not
much more can be
said than that reasonable
assumed departures from the
Coulomb law
magnitude
in the neighbourhood of the nucleus can account in order of

for the observed effects.
Nuclear spin in one-electron spectra.

The effects considered in the two preceding
3.
sections
do not produce a
splitting of the levels of a single atomic species and so can only give rise to
a hyperfine structure when several isotopes are present. But such structure
observed in atoms having no isotopes, notably bismuth, so some addi

tional hypothesis is needed for description of this structure. This was sup
plied in 1924 by Pauli,* who postulated that the nucleus itself may have a
is
spin angular momentum and an associated magnetic moment. It is supposed

that a nucleus of given Z and 31 always has the same spin, denoted by J, but
that different kinds of nuclei have different spins. This hypothesis of nuclear
spin has also proven of great importance in the theory of molecular spectra,
so that it is now an indispensable part of atomic theory.
With nuclear spin postulated, we need in addition to know the term in the
Hamiltonian corresponding to the interaction of the nuclear spin with the
electronic structure. This was first obtained from the classical picture of a
nuclear magnetic moment whose energy in the magnetic field produced by
fche electronic structure depends on the orientation of the nucleus relative
to that field. This brings
up the question of the magnitude of the nuclear
magnetic moment. It is difficult in the present state of knowledge of

nuclear structure to say anything about this that is very definite. For an
3lectron the magnetic moment is e/ftc times the angular momentum, so it
was natural to suppose that for a proton the magneto-mechanical ratio

EFOuld be efMc or ft/Jf = 1/1838 times as great. But this appears not to be
bhe case according to the recent experiments of Stern, Friseh, and Ester-
mann f
3
in
molecular
which the magnetic moment of the proton was measured by a

beam method on molecular hydrogen. The result was a magnetic
noment 2-5 times larger than (e/Jfc)(Jft). That the spin of the proton is \H
a known from the band spectra of E2 This unexpected result for the proton
s made certain by a more direct method, developed by Rabi, Kellogg,
.
rnd Zacharias | which uses atomic instead of molecular hydrogen.

Somewhat surprising is the fact that those nuclei which are supposed to
3
jontam an odd number of electrons also have magnetic moments of this

*
PABU, NatorwiBB. 12, 741 (1924),
f FKESCH and STEIC*, Zeits. fur Phys. 85, 4 (1933);
ESTEBMASW and STEBN, ibid. 85, 17 (1933).
{ KABI, KELLOGO, and ZACHABIAS, Phys. Bev. 46, 157 (1934).
THE xrcLErs
422
rs*
3 18
ATOMIC SPECTKA
size; this makes it appear that a nuclear electron is quite different from
a free electron or one in the extra -nuclear structure. This is just one of the
properties of nuclei wMch lends support to the never view that nuclei are
composed wholly of protons and neutrons and do not contain any
The
electrons.
quantitative theory of the interaction of the nuclear magnetic

and the outer electrons is due to Fermi and Hargreaves.* The
first
moment
magnetic
M of the nucleus gives rise to a field described by the
moment
vector potential
^ = ^Hxr.
(I)
Since the magnitude of the nuclear spin is constant in all states of the system,
not
to be explicitly mentioned but we do need to introduce
it
J/j the z-component of nuclear spin, as a coordinate and quantum number.

is proportional to the nuclear angular momentum I, that is
Then, since
,
(2)
where gs
is
analogous to the Lande ^-factor in the theory of the Zeeman

are represented by three nonthat the components of
effect, it follows
commuting matrices whose rows and columns are labelled by the values of
5
Jlj . Fermi handles the one-electron problem with Dirac s equations ( 5 ),
including the interaction with the nucleus through the vector potential
given by
(1).
We
not reproduce the calculations in detail. It is easy to see that we

here
have
another
of vector coupling. We start with a scheme in which
tie
J of the outer electron and the I of the nucleus is known. The resultant
Fis the vector sum of these two and we have to transfrom a scheme of states in which JP and
are diagonal to one in which
F2 is diagonal. All this is exactly what we have been through in studying the
coupling of
L to form J.
the s $, terms of the one-electron ns configurations are

spit into two levels, the displacements from the unperturbed positions
Fermi
being
(3)
Here /% = e
/2/tf is
magnetic moment
the magnetic moment of the Bohr magneton.

of the nucleus and
{0) is
M the actual
the value of the normalized
value of
5-eigenfunction at the nucleus. The factor i belongs with an
I- J
I-rf and the other with F=I~|, For positive M, the level with J
T
fur Piys. 0, 320 (1930);

HAJBGXEAVBS, Proc. Roy. See. AIM, 568 (1029); A127, 141, 407 (1930).
3 18
is
NTJCLEAB SPIN IX ONE-ELECTRON SPECTRA
lower In energy than the other. Positive
423
M corresponds to the magnetic
moment parallel to the angular momentum as If the magnetic moment were

produced by rotation of the positive charge distribution of the nucleus.
Likewise Fermi finds that the 2 P, level is split into two levels In the same way,
the amount of the perturbation being given by
2
The
I
|,
Pj
level
is split
(3)
with r^ In place of TT ^ 2 (0).
in general into four levels
whose J values are 1
f,
J-f J, and I-f f the displacements In energy being

,
respectively. (For 1=1 there are but three levels, as I f Is Impossible, and
for
| there are only two.) The four levels have the same kind of Interval
1=
rule as with Russell-Saunders terms: the Interval between adjacent levels

is proportional to the
value of the pair.
larger
Let us now consider the application of these results to the alkali metals.
Schiller, and Dobrezov and Terenin* find In 2fal that each of the D-lines
consists of two
components with a separation of G-022A. For potassium the

corresponding structure Is either absent or very much narrower, f In Rbl
the second line of the principal series shows doubling as in sodium with a
component distance of O02QA4 Jackson studied the first three lines of the
principal series of Cs I and found that each component In all three lines
showed a separation of 0-300 cm"1 and that the two hyperfine components
were of about equal Intensity.
These observations are in accord with the view that the 2 $ level has been
,
doubled by interaction with the nucleus and that a splitting of the 2 P levels
is negligible. From the splitting one cannot find the value of I, since In any
case the 2
level would be split into two levels.
S^
That the nuclear spin of Na I is equal to f is shown by the measurements
of Rabi and Cohen by a modification of the Stem-Gerlach experiment, by
and Urey from alternating intensities In the Na2 bands, and by the
recent accurate measurements of the hyperfine structure Intensities by
Joffe
We thus have three different methods leading to

value
in
case
the
of this nucleus. The calculation of the nuclear
the same
Granath and Van Atta.
j|
moment from the observed splittings Is of course a much more

and uncertain matter. The value of / is inferred from hyperfine
magnetic
difficult
SCETOLER, Natorwiss. 16, 512 (1928);
DOBBEZOV and TEBESX^,
ibid.
16, 658 (1928).
f ScHiiLEB and BBUCK, Zeits. ffir Phys. 58, 735 (1929).

f ECLEPFOV and GBOSS, Naturwiss. 17, 121 (1929).
JACKSON, Proc. Boy. Soc. A121, 432 (1928).
H BABI and CoBm, Hrys. Rev. 43, 582 (1933); 46, 707 (1934);
JOFFE and UBEY, ibid. 43, 761 (1933);
GBAJJATH and VAX ATTA,
ibid.
M, 935
(1933).
424
NUCLEUS
ES"
ATOMIC SPECTRA
gis
by an application of the sum rales, for evidently the

9
theory of Russeil-Saunders multiple! intensities (2 ) Is applicable here if
we write F for J, J for S and I for L. Thus in the splitting of a line ending
structure intensities
on the Z S level, the two components have an intensity ratio of /: ( I -fbeing the ratio of the two values of ZF -f 1 for F= I- J and JF = / +
1),
this
We
not undertake a detailed review of the theory by which the

nuclear magnetic moment is calculated from the data. For one-electron
spectra in particular this has been most carefully discussed in a paper by
Fermi and Segre.^ This paper includes a compilation of known values of
nuclear moments, as does an independent paper by Goudsmit.f
Table I 18 is based on these two compilations and gives the values of the
moment in unite el 21fc from each paper,

(
so that the reader
for himself the degree of consistency obtainable

in the
by
may
different workers
of development of the theory.

TABLE
I ls .
Unclear spins and magnetic mommis.
The iiyperfine structure of two-electron spectra.

The most thoroughly worked out case of
hyperfine structure wliere there
4.
is
more than one

*
electron outside closed shells is that of the
and SEsai, Zeits. fur Pfays. 82, 729

(1933).
t GOO>SMITS Fhys. Rev. 43, 636 (1933).
f Rrom GEASATH and TAN ATTA (foe. dL).
ls2s^S and
4 18
THE HYPEEPINE STBUCTUEE OF TWO-ELECTEON SPECTRA 425
3
2p P terms of Li II.* The hyperfine structure of the multlplet due to the
combination of these terms is of particular interest because it is comparable
Is
due to spin-orbit
magnitude with the ordinary level structure of Li
The theory is in good accord with experiment! when the value
of / for Li 7 the most abundant isotope, is taken to be f This value is in
accord with that from the spectrum of the molecule Li 2 J
We shall not give the details of the calculations. These are based on the
use of a law of interaction of the electrons with the nucleus as used by Fermi
in
interaction.
for the one-electron problem, together with use of appropriate two-electron

wave functions. For the 3 Sl level, Breit and Doermann find that the
perturbation due to interaction with the nucleus
i-oel^MôH-i,
F=
is
-f,
(i)
i)
components
f f and f respectively. This counts only the
interaction of the Is electron with the nucleus, 0(0) being here the value at
the origin of the Is wave function.
for the three
The 3 P levels are split according to the laws of vector addition of angular
momenta: 3 P is not split, 3 P1 becomes a group
f
of three levels with
of four levels with
F= |
.F
= |,
f and f and
f , f and
f.
P2 a group
An
energy
diagram (after Giittinger and Pauli) is given

in Fig. I 18 which shows how the magnitude of the
3
hyperfine structure in Pis related to the ordinary
level
The fact that they are comparable

that
the
implies
non-diagonal matrix components
of the nuclear interaction which connect different
level intervals.
levels of
^
3
P are important. These were considered
%
|
in the calculations of Giittinger and Pauli.

The experimental data are ingood general agree
ment with the
theoretical calculations, although

there are discrepancies in the relative intensities.
Pf
_
.
The
detailed interpretation of the observed

18
hyperfine structure for the case of more than one Fig. i Hyperfinestructure
of the ** P terai ** Li]L
electron outside closed shells is complicated by the
.
necessity of considering the state in intermediate coupling (Chapter xi)

*
749 (1930);
743 (1931);
BBEIT and DOEKMAIW, Phys. Rev. 36, 1262, 1732 (1930).
f SCH-ULKB, Zeits. ffip Phys. 42, 487 (1927);
GRADATE, PJiys. Rev. 36, 1018 (1930).
J WUEM, Zeits. fur Phys. 58, 562 (1929);
HABVEY and JUNKETS, Phys. Rev. 34, 1286 (1929); 35, 789 (1930).
GtJmKGEB,
Zeits. fur Phys. 64,

ibid. 67,
GOTTES-QEB and PAULI,
and
NUCLEUS IS ATOMIC SPECTRA
428
4 18
The effect of the

Intermediate coupling has been considered in detail by Breit and Wills,*
while the importance of considering the interaction of configurations in this
also the Influence of configuration interaction (Chapter xv).
connection has been stressed by Fermi and Segre.t At present not enough
is
known about
configuration interaction to
make
possible quantitative
but Fermi and Segre show that qualitatively its

effects account for a number of observed features of the hyperfine structure
Trhich are otherwise without explanation.
allowance for
its effects,
The work of Breit and Wills is in extension of that of Goudsmit, J who
first
gave a general discussion of the interaction with the nucleus in the case of
complex spectra by vector-coupling methods. Their developments agree
experiment but there are many outstanding discrepancies,
the
need
for further refinements of the theory.
indicating
In Hg
A1II
occur in which the hyperfine structure is com
fairly well with,
parable with the ordinary spin-orbit and electrostatic separations. This

necessary more accurate calculations of the nuclear perturbation of
the levels. Empirically
effects and their nature were recognized by
Paschen.f They have been discussed in detail by Goudsmit and Bacher
j
with satisfactory
5.
Zeeman
results.
effect of hyperfine structure.
We shall not give a detailed
account of the results which have been ob

tained for the perturbation by a magnetic field of the structure due to nuclear
Tie work is extremely important as indicating the essential

correctness of the nuclear-magnetic-moment
picture, but presents no new
theoretical problems. In the ordinary
of
the Zeeman effect ( I 16 ) the
theory
interaction.
from the simple Lorentz theory arise because of the different

values of the magneto-mechanical ratio associated with the vectors L and S.
We have a
situation here.
For fields such that the ordinary Paschen-
Back effect is negligible we have a magneto-mechanical ratio of the electronic

represented by g(J) where this is the Lande g factor of the level in
question, Associated with the nucleus is a magneto-mechanical ratio about
I0~3 as great so
the direct interaction of the nucleus with the external
magnetic
field is negigible.
A direct application of the formulas of Chapter in then shows the split

ting of a given hyperfine level to be governed by g(J) of the ordinary level
to which the level
belongs multiplied by the factor (I03 2a) which gives the
*
BBMT aad Wnxs,
Pfcys.
Rev. 44, 470
(1933).
f FXEMI and Src&i, Zeite. f3r Phys. 82, 729 (1933).

t Gorasnr, Piijs. Rev, 37, 663 (1931).
Pj.SH!3rs Sizber. PTCIBBB. Aiad. 1832,
502.
II
GOUBSMZP and BACHES,
p.
Pliys.
Bev. 48, 894 (1933).
18
ZEEMA25"
EFFECT OF HYPEEFIHE STEUCTUEE
427
2
natrix component of Jz in a state labelled
and Fz
by precise values of
Ihat is, the g value for a state of resultant
angular momentum F composed
of electronic J and nuclear / is
2F(F+i)
This result
was
first
obtained by Goudsmit and Bacher* from vector-model
3onsiderations.
Just as in the discussion of 5 16 the difference in the

magneto-mechanical
ratio of the vectors
and /means that strong magnetic fields can produce a
,
breakdown of the coupling ofJ and 1. TMs is the Paschen-Back

tiyperfine structure. Experimentally
has been carefully studied
agreement for
and
effect of the
theoretically this
by Back and Goudsmitf
for Bi.
phenomenon
They find good
at field strengths corresponding to various stages of the

transformation of coupling schemes. The magnetic transformation of the
I=f
hyperfine structure in thallium has been investigated thoroughly by
and Wulff.i
*
f
J
GOUDSMIT and BACHEE, Zeits. fur Phys. 86, 13 (1930).

BACK and GOUDSMIT, Zeits. fur Phys. 47, 174 (1928);
ZEEMAN, BACK, and GTOUDSMIT, ibid. 66, 1 (1930).
BACK and WULFF, Zeits. fur Phys. 66, 31 (1930).
Back
APPENDIX
UNIVERSAL COXSTAXTS AXD
XATURAL ATOMIC UXITS
Measurements in physics are statements of relation of the quantity mea
sured to quantities of like kind which are called units. It is customary to
build up the system in such, a way that the unit of any kind of physical
quantity is defined in terms of three conventional units of mass, length, and
The choice of the basic units for these quantities is wholly arbitrary.
the general order of magnitude in the centimetre-gram-second system being
such that the numerical measure of quantities occurring in ordinary labora
time.
tory experiments is of the general order of unity. Thus the Telocity of light
in the cgs system is 3 x 10 10 cm sec* 1 The centimetre and second being so
.
chosen that
cm
1
sec""
is
of the order of velocities of common experience, the
bigness of the number measuring velocity of light on this system is simply a

statement that velocities of common experience are very small compared
withthat of light.
There
is,
therefore, nothing especially
Its basic units are of convenient
fundamental about the cgs
magnitude
for
common
basis.
laboratory ap
paratus, so ultimately the foundation is anthropomorphic since laboratory

apparatus is built and designed on a scale convenient for manipulation and
observation
by a human observer. To recognize this fact is not to deplore it.
Certainly the egs system is convenient for description of the macroscopic

apparatus which provides the refined sense-data of physics. But the fact
shows us clearly that a metric resting on such a basis will probably not
provide units of convenient size for dealing with another branch of physics
like the theory of atomic structure.
THs
is
in fact the case.
The
basic universal constants of atomic theory
have values which are very large or very small compared with unity. For
example*:
Electron eiiarge
Quantum constant:
:
-e=
4-770
1=
1*043
x 10~ S7 g
9-035
xlO~
Electron mass :
^L=
Light velocity:
10
lO"
gt cmi
cm
1
sec""
see"
ffl
2-&979B x 1C10
g;
cm
sec"
1
.
These great powers often are rather inconvenient in theoretical calculations.

There aretwo ways of avoiding them which might be adopted. One is the way
which tie metric system has already adopted for extending itself to larger
* We use the values of the universal
constants as given by BIRQE after a critical surrey of all the
relevant data: Rev. Mod. Phys. 1, 1 (192% Phys. Rev. 40, 228 (1932). For later modifications see
BIBGE, Phya. Bev. 43, 211 (1933); MICHELSON, PEASE, and PBABSON, Science 81, 100
(1935).
UNIVERSAL CONSTANTS AND NATFBAL ATOMIC UMTS
429
or smaller units, namely to choose another system of fundamental units

related to the metric system by conversion factors which are
powers of 10.
By taking as our units of mass, length and time

10-* em,
lO-^g,
the
and
10-"
new
system
sec,
9
electrostatic unit of charge becomes 10~ giem*sec~ 1 ; so in this

the electronic charge has the value
0*4770. Similarly the new unit
of velocity is 10 9 cmsec-* 1 and of angular momentum 10- 27 gcm 2 sec- 1 3 so

the velocity of light and the quantum constant have the numerical values
,
29-9796 and 1-043 respectively.
For theoretical purposes, however, it is still more convenient to introduce

a system of units in which certain of these universal constants are set equal
to unity. From this standpoint the numerical values of the universal con
stants become the numerical conversion factors which connect this system
with the ordinary c g s units. Quite a variety of such systems have an equal
claim to use so far as general convenience is concerned. Thus it is a matter
of arbitrary choice whether we set h or Ji equal to unity. This kind of arbitrary
choice merely alters the place where the pure number 77 appears in the
calculations and is analogous to the difference between the Heaviside-
Lorentz units and the older set of electromagnetic units. Another arbitrary
element lies in the fact that there are only three fundamental units at our
disposal, so that it is not possible to assign the numerical values of more than
three of the universal constants
choice is
different
systems arise according to which
made in this respect. There is not much point in debating which of
the choices
is
most convenient. As it happens one particular
choice,
recom
mended by Hartree * has already been quite generally employed in theo

retical work, so we adopt that one. Hartree s atomic units are such that e,
and ft have each the numerical value unity. Denoting by a and r the Hartree
units of length and time respectively this means that
3
JJL
From
this standpoint the numerical values of the universal constants
e,
ft
and \L given in the
c g s system are the conversion factors by means of which

quantities expressed in Hartree s system are to be expressed in the egs
system. Thus an angular momentum expressed as xK in the Hartree system,
27
2
1
in the
expressed as (1-043 x 10~ )a:gcm
the
as
la
Therefore
in
the
Hartree
length expressed
cgs system.
system
27 2
2S
1
is expressed as (1-043 x 10~ ) (9-035 x 10~ )- (4-770 x 10~ 10 )- 2 cm, which
_
works out to be
la = 528 x 10 8 cm
where x
is
a pure number,
is
sec""
HA&TBEE, Proc. Camb. PMl. Soc. 24, 89
(1926).
430
APPENDIX
Similarly the value of T in the
cgs system is
xIQ-17 see.
IT = 2-41 9
From
these basic conversion factors the value of the Hartree unit of any
derived quantity in the cgs system

1
1 ai-
is
= 2- 1 8 x
readily found.
10
cm sec*
Thus unit velocity
is
1
;
hence the velocity of light has the value
c= 137-29 ar
1
.
The
reciprocal of the pure number which gives the value of the velocity of
light in these units is known in the theory as the fine structure constant.
This
is
usually defined
by the equation
a = e 2/cl.
It appears,, for example, as a parameter of the relativity-spin structure of

the hydrogen spectrum in Chapter v. Another important parameter of
atomic theory
is
the mass of the proton, which
is
11=1838-3/1.
The fact that the numerical magnitudes of c 2 and of JI in this system of units
are both large
compared with unity is of fundamental importance for the

c 2 is what makes the relativistic corrections small
and the largeness of 31 makes it a good approximation to treat the nucleus
theory.
The
largeness of
as a fixed centre of force.
Aside from the fact that

this
e R and p. have simple numerical

magnitudes in
system of units, the other units have simple physical interpretations in
}
the theory. Thus unit length is the radius of the first orbit in Bohr
hydrogen for an infinitely massive nucleus,

of the electron in this first Bohr orbit. Unit
the electron in the
Hartree unit.
theory of
and unit velocity is the velocity

energy is the potential energy of
Bohr orbit and hence the ionization energy is half a

Numerous other examples are to be found in the table in this
first
appendix.
We shall not discuss the precision with which these quantities are known
in terms of the
cgs system other than to say that it is quite generally that

associated with a probable error somewhat less than 0-1
per cent. The
quantity whose relation to the cgs system is known most precisely is the
combination aa-1 The combination aa-1/^ is known as the
Rydberg
constant, R. Its value is
.
R = ~ a-1 = 109737-42 ~4- 0-06 cm-1
4?r
Its relation to the c

gs
known to an accuracy of about 6 parts

system
in 10 million,
the
although
accuracy with which the values of a and of a
are
known
is
much
less. The theoretical
separately
energy levels of atoms
is
thus
CONSTANTS AND NATURAL ATOMIC UNITS

are given in terms of the Hartree
431
energy unit and the parameters like a, p/M,
the atomic
number Z, and the quantum numbers

labelling the
level in question. Of these Z and the
numbers
are
quantum
particular
integers exactly,
so there is no
uncertainty in their numerical values.
on the energy are usually in the form of small
The effects of oc and ft/If
corrections, so
we
are not
hampered by the uncertainty in their values in making com

parisons between theory and experiment. In
using the energy levels to
predict the wave-numbers of spectral lines, the wave-number
equivalent of
the energy given by Bohr s
relation, E^Jic^m what enters. As the wave
number equivalent to one Hartree unit of
is 2R it follows that the
greatly
energy
principal factor needed for passing from Hartree units to
cgs units
is
the
one known with the

greatest precision. The precision of knowledge in R is
much greater than the accuracy attained in the
calcula
perturbation theory
tions at present, so that the

comparisons with experiment are not at all
hampered by the inaccuracy in the Rydberg constant. For that reason we
may
express the results of the theory at once in terms of

paying attention to the uncertainties in R.
em-1 without
The only other universal constant of interest for

spectroscopic theory is
the Boltzmann constant i, which measures the relation between
thermal
energy and the thermodynamic temperature scale measured in conventional
5
Centigrade degrees. As we nearly always express energy in terms of equi

valent wave-numbers in cm-1 , the
important conversion factor is not k
itself
but
value
which when multiplied by T in Centigrade

degrees gives
wave-number equivalent of the energy JcT. This factor has the
A/Ac,
directly the
j^=
0-698 cm-1 deg- 1 .
This is the reciprocal of the quantity often denoted

theory.
by c2 in classical radiation
APPENDIX
432
Values of some Important Physical Quantities

Length:
Radius of
Bohr
first
orbit,
a=F
Shift in X-ray wave-length.
Value in c g s units
Value in atomic units
Kind of quantity
t
by Compion
scat
la
(0-5282
0-0458 a
21-2x10*"
xlO"
0-0004}
cm
cm
tering through TT, 2, 27roca

(TMs is also the wave-length of y-radiatioa
associated with annihilation of an electron)

2
2
2
Electromagnetic radios of electron, e /j^c =;x a
1
4ira" a H
of
of
limit
series,
Lyman
Wave-length
Atomic
2-80xiO-13 cm
5-31
910xlCT 8 cm
1722 a
1-893
unit,
a"
xir* a-1
Rydbeig constant, Rô
5-81
first Balmer line, jf5 R H

Hydrogen doublet constant, y*R E jlft
Wave-number associated with one electron Tolt
0-807
Wave-number of
of electron,
x!Q-*
1
a-"
1
l-ftSxlO- ar
4-281
xKT5 a
81063cmrl
(9-035
1-661
|i
Mass of proton, if
Time:
Time
for electron to
first
Mean
Bohr
life
orbit,
for
go
(Sir)-
2419 xir17 sec
revolutions in
transition in hydrogen
2p-!
Speed of electron in first Bokr

1
Speed of light, ar^r"
in
first
Basic quantity
A ngidw
Basic
Bohi
orbit*
6-62xi07 r
l-6xlOrfleo
2-18 xlO8 cm sec-1
10
1
(2-99796+ 0*00004) x I0 cm sec-
1
ar"
orbit,
xl(r19 g cm
xlO~17 gem
1-966
/lar"
of relativistic theory
jn
0-010} xlO^g
xlO-g
Velocity:
Momentum:
Of electron
xlO8 cm-1
109737-420-06cmrl
15233cm- 1
0-3636 iO-OOOecnr1
137-29 jua-T
2-70
1
sec"
sec"
quantum
unit,
Planck constant,
1-043
fcs/xaV"
6-547
2d!=&
2
2Rfe, twice the

unit, IOL*T-*= ie*, I
ionization energy of hydrogen with infinite
Atomic
nacSe&r
1
sec"
1
sec"
11
xlO-"
=:219474-84
erg
cm-1
mas
Energy equivalent of electron mass,
4-304
x!0~27 gem2
xlO~27 gem3
18859 jiaV*
ic:27-07 electron volt

7
8-10
erg
xir
Force:
Force of attraction toward nucleus on electron

in first
Bohr orbit,
8-19
xlO~a dyne
e*/a
Ekctrit ctergc:
Atomic
unit,
e,
negative of charge on election
Potential:
Potential of electron s field at atomic unit
J O-O
distance, e/a
[27-2
Eltciricfidd:
Field strength at atomic unit distance from
Ekdrie
sec- 1
(17-1
xlOSgcm-
151-3
x 10s volt cnr 1
electron, e/a
volts
mmmt:
Moment
first
of dipole formed
Bohr
2-52xlO- 18 esTicm
by hydrogen atom in
orbit, ea
Magnetic fdd:
Atomic
unit,
first
Magnetic
Bohr
17-1x10* gauss
^sC^f~"
Field at nncleos doe to motion of electron in
7-283
1-245
xlO5 gauss
orbit,
momeM:
Atomic unit, i^&^

Bohr magneton, ei/2jiC=|
2-52
xlO^18 erg gauss"1
0-9174 x
20
lO"
erg
gauss"
LIST OF PRINCIPAL TABLES
INDEX OF SUBJECTS
INDEX OF NAMES
LIST OF PRINCIPAL TABLES

ABLE
U\Jî2iMJ m
I3
2?
i^lm^m^ljm)
3s
OilaMfL;iiJ)
)
.... ........
............
............
... .........
......
.......
.....
PAGE
76
76
76
77
Xonnalized radial eigenfunctions for nydrogen
117
2s
Mean
117
values of r* for hydrogen
Taliiraof
a^Zntj^X)
r(I}fi(
i-I)drforliydrogeii
.............
Kussell-Samiders t-erms and Lande factors for
Line strengths in RusseH-Sannders mnltiplets
nlx
configurations
133
178
ISO
208
.......251
........
.....
.........
.....
241
3s
Relative multiplet strengtlis in sereral taBnsition arrays
4*
Strength
2 ie
Allowed J values for groups of
1 11
Matrices of spin-orbit interaction
1 12
Tlie transformation from jj coupling to
513
The Xeneigy levels
1 14
Tie periodic table
2 14
Tlie normal electron conlgnrations of the elements
333
15
Lande g
382
1S
Xndear spins and magnetic moments
of
253
qTiadrapole-multlplet lines
^"-equivalent
263
electrons
268
LS coupling
.........
...........
.....
.......
of the
atoms
factors for Russel-Saonders terms

.
Values of some important physical quantities
.....
......
.
294
320
332
424
432
INDEX OF SUBJECTS
line shape, 109
Absorption
Addition theorem for spherical harmonics, 53
Alkali spectra, 141, 183
oscillator strengths, 10$
perturbed intensities, 376
Stark effect, 415
Allowed levels, 167
Allowed levels mjj coupling
for nl\ 263
for nljx, 263
method of obtaining, 258, 263
Allowed levels in LS coupling
for nLx, 208

method of obtaining, 190, 216
Allowed values, 14
of a function of an observable, 15
of angular momentum, 46
Almost closed shells, Chapter xn, Chapter
xm
Angular momentum, Chapter
allowed values, 46
commutators, 46
matrices,
48
where JJ.-F J2 = J, 64
matrix of (Ji+Ja) , 58
number of electrons with various values, by
Fermi-Thomas method, 337
matrices of
orbital, 50,
JL and/2 2
113
selection roles, 61, 67,

spin,
96
54
transformation amplitudes for vector

addition, 73
transport in radiation* 92 t 96
use of operators to get Russel-Saunders
eigenfunctions, 226
vector addition, 56, 73

Anticommuting observables, 20, 186
Anti-Hermitiaa matrix, 14
Antisymmetrizer, 165
Antisymmetry of T, 164
Atomic units, 428, 432
Anto-ionization, 369
Average, see Mean
Balmer series, 3, 6, 137

Bose statistics, 166
Brackett series, 137
Central fields, Chapter xrv
Central-force problem, 112
Characteristic values, 14
dosed shells, ICfc, 177, 183
Commutation
rales (cont.)
parity operator, 186
Commutator,
14, 25,
Configuration, 113, 168

interaction, Chapter xv
normal, of elements, 333
selection rules, 236
Continuous eigenvalues, 20
Conventions concerning quantum numbers,
168
Correspondence principle, 7, 9, 87
Critical potentials, 5
Degeneracy, 15, 16, 160

8 function, 20
Diagonaiization method of obtaining eigenfunctions, 220
222
Diagonalization of a matrix, 29,
Diagonal matrix, 19
Diagonal-sum
Dipole
rule,
19
moment
electric, 85,
90
see Magnetic moment

Dipole radiation, 90
Direct integral, 173

Dispersion theory, 103
Doppler effect, 139

Doublet intervab
alkalis,
144
X-ray levels, 321

Dynamical equation, 26
Eigendifferential, 22
21
Eigen-^r s, eigenfunctions, eigenstates, 14,
almost closed shells, 284
262
Russel-Saunders terms, Chapter vin
jj coupling,
simultaneous, 17
Eigenvalues, 14
continuous, 20
Electromagnetic theory, 83
Electron spin, 8, 54
Electrostatic interaction, 114, 141, 159, 1*4,
179, 259, 295
Equivalence degeneracy, 160
Equivalence, dynamical, of the electrons,
162
Equivalent electrons, see Allowed
levels,
Two-electron configurations
Combination principle, 3
Even states, 185

Exchange
angular momentum, 46, 59. 60

coordinates and momenta, 25, 26, 43
16, 19
Complete set of observables, 18

168
Complete set of quantum, numbers, 161,
Component (of a line), 97
Commutation
rules
43
Commuting observables,
integral, 173
Exclusion principle, 9, 166
see Equivalent electrons

28-2
and
INDEX
436
SUBJECTS
Fermi statistics, 166

Fermi-Thomas central field, 334
Forbidden lines of astrophysieal interest, 282
F-type operator, 169
Lavs of quantum meenanics, 25

Legendre polynomials, 52, 174, 176
/value, 108
Line, 97
in alkali spectra, 147
Level, 97, 122

see Allowed levels
shape in absorption, 109

strength, see Strength
g factor, 151,381,384,426
g-pennanenee
rule*
390
10, II
Group theory,
width, 82, 109

LS-coupling scheme, 188, 216, 285, 291
Lynian series, 4, 6, 137
gr-smn role, 385

Gr-type operator, 171
Magnetic-dipole radiation, 86, 93, 283
345
413
triplet terms, 210
Hermitian conjugate matrices, 19
Hermitian matrix, 14, 19
Hennitlan operator* 14
Hicks series formula, 143
Hilbert space, 23
Magnetic moment, 86, 90

atomic unit, 432
of electron, 54
of proton, 421
of the nucleus, 422, 424
Magneton, 422, 432
Manv-electron jumps, 375
Matrix, 19, 27
Matrix mechanics, 8, 27
Maxwell s equations, 83
Mean life, 80
in hydrogen, 136, 137
Hydrogen and hydrogen-like ions
Mean value of an observable, 15
HamDtonian operator, 25, 158, 211

Helium and helium-like ions
excited levels, 348
normal
Stark
states, 6,
effect,
experimental results* 137

radial functions, 114
relativistie theory, 125
relativity correction, 119
spin-orbit interaction, 123
Stark effect, 398
Hyper&ie structure, 418
of *
in.
hydrogen, 117
Monad, 217
Multiplet, 8, 237, 264
qnadrupole, 252
snpermultiplet, 245
see Inverted, Normal, Strength
Multiplicity, 189
selection rule, 237
Individual set of quantum numbers, 168

Induced emission and absorption, 100
Natural excitation, 97
Inequivalent configurations, 214

intensities, 97, 98, 99
Non-commuting vectors, 43
Normal multiplets, 144, 300
see Strength
Interaction between configurations, Chapter
Nucleus
xv
Intermediate coupling, Cnapter si, 301
Intervals, doublet, 144, 321
Intervals in Rnssel-Saunders terms, 193
absolute, 195
Lande rule, 193
nlx configurations, 208
terms of coupled groups, 219
Inverted multiplets, 146, 197, 300
Isotope effect, 418, 420
due to, Chapter xvm

magnetic moment, 422, 424
spin, 421, 424
Numerical integration of the radial equation,
344
effects
Observable, 13
allowed values, 14, 15
complete set, 18
conjugate, 14
realitv condition, 14
Odd states,
file,
248
sum rale,
247, 278
J group,
278
sum rule, 278
J-group file, 2S1
sum rule, 281
jj coupling,
Chapter x, 287, 291
Lagnerre polynomials, 114
Lande interval
rale, 193
see Intervals in Mnssdl-Saunders terms
Laporte
rule,
236
185
One-electron spectra, Chapter v, 183
Orbital angular momentum, 50, 113
Oscillator strength, 108
in alkali spectra, 147
Parabolic coordinates, 398

Parents of a term, 216, 276, 303
Parity operator, 186
selection rule, 236, 237
Paschen-Back effect, 152, 388, 390
in hyperfine structure, 426
Paschen series, 4, 6, 137
INDEX OF SUBJECTS
Pauli exclusion principle, 9, 166
see Equivalent electrons
437
Russell-Saunders case, Chapters vn,
Pentad, 217
Allowed
see
Rydberg constant,
Perturbation, 30
Satellites, 240,
caused by a single state, 37

theory, 30, 34, 35
Perturbed series, 367
X-ray, 323
Scalar potential, 83
effect,
400
effect, 154,
4, 6, 115, 137,
430, 432
277
Scattering of radiation, 103

Sehrodinger operator, 21, 27
Schrodinger representation, 21
Scbrodinger s equation, 23, 26
Screening, 159
in X-ray levels, 320
Secular equation, 29
Selection rule, 60
momentum
angular
61, 67, 96
configuration, 236
multiplicity, 237
parent terms, 245
parity, 236, 237
Self -consistent fields, 354, 358, 362
386
Polyad, 217, 245

Positron, 125
Potential
atomic unit, 432

retarded, 84
scalar, 83
vector, 83, 149
Shell, 168
almost closed, 295

closed, 177
Principle of spectroscopic stability, 20

Probability interpretation of (F ), 24
Proper values, 14
Spectra, 1
banded, 2
continuous, 2
Proton
magnetic moment, 421
mass, 432
line,
Quaclrupole moment, 85, 90
Quantum mechanics, method and
laws,
Chapter n
Quantum number, 18
effective, 142
radial, 131
113
Radial functions, 112, Chapter

for hydrogen, 114
Radiation theory, Chapter iv
nv
classical electromagnetic theory, 83

correspondence principle for emission, 87
dipole-radiation field, 90
of Dirac, 81
quadnipole-radiation field, 93
Raman effect, 103
Rare-gas spectra, 301, 315
Real observable, 14
Relativistic theory, 125
Relativity correction, 117, 125
and
Spectroscopic stability, 20
applied to multiplet strengths, 249
Spectroscopy, 1
Spectrum of an observable, 14
Quadrupole radiation, 93, 99, 282

line strengths in quadrupole multiplets, 252
Zeeman effect, 395
total,
Spin-orbit
interaction, Strength
Phases, choice of, 48, 52, 66, 67, 123, 270, 292
Phases in the matrix of S^, 277
Planck s constant, 5, 432
Poisson bracket, 25
Polarization, 89
of dipole radiation, 91
of quadrupole radiation, 95
Zeeman
Intervals,
levels,
Periodic system, 327

Periodic table, 332, 333
Permutation operator, 162
Stark
vm, ix
Zeeman effect, 380
observables,
Representations of states
15,21
Ritz combination principle, 3, 5
Ritz series formula, 143, 367
Ritz variation method, 345
Spherical harmonics, 52
Spin, 8, 54
Pauli matrices, 55
Spin-orbit interaction, 120, 159
almost closed shells, 299, 306, 312
in helinm, 211
intermediate coupling, 266, 268
jj coupling, 257
LS coupling, 193, 195, 209
perturbation of doublet intensities, 376
spin of one electron orbit of the other, 273
Stark
effect,
Chapter xvii
State, 12, 97, 122
Stationary states,
5,
26
Statistical interpretation, 24
Statistical method of Fermi-Thomas,
Strength
(of
a spectral
line), 98,
100
376
316
dipole multiplets, 237
forbidden lines, 282
alkali spectra, 147,
almost closed
shells,
hydrogen, 131
intermediate coupling, 275
264
many-electron jumps, 375
quadrupole multiplets, 252
Russell-Saunders case, Chapter ix
Stark effect, 401, 411
jj coupling,
335
INDEX OF SUBJECTS
438
Strength, (conf.)
X-rays, 322
Zeemaa effect, 386
Sum rales,
71, 108, 238, 247

^-permanence rule, 390
g-sam rule, 385
J file, 247, 279
J group, 278
/-group file, 281
Supermultiplet, 245
and observables,
Symbolic algebra of states
12
Symmetrizer, 165
Symmetry of $, 164
System of terms, 237
Term,
122, 189
Term energies, 191

Term intervals, see Intervals
Term system, 237
Term values, 3-6
Trace, 19
Transformation coefficients or amplitudes, 18

for vector addition, 73
Transformations betveen different coupling
scliemes, Chapter xn
Transition array, 244

Transition probabilities, 79, 0$, 100, 109
in Can, 148
in hydrogen, 136
in "nebulram*, 283
Triad, 217
TVo-electron atoms, ionization potentials,
346,348
Two-electron configurations
absolute term intervals, 220
allowed levels, 232, 263
Two-electron configurations (cont.)

229, 231
eigenfunctions,
424
hyperfine structure,
interaction of the spin of one electron with
the orbit of the other, 273
line strengths, 246, 264
matrix of spin-orbit interaction,
267
transformation from jj coupling to
LS
coupling, 294
Ultra-ionization potentials, 373

Unitary transformation, 18
Universal constants, 428, 432
Valence electron, 183

Vector addition of angular momenta, 56, 73
Vector coupling, 56
in antisymmetric states, 213
method for Russell-Saunders eigenfunc
tions, 228
model of atom, 8
Vector potential, 83
Vectors, non-commuting, 43
Wave-number,
3, 5, 7,
432
Wentzel-Brilouin-Krarners approximation,
339
Width of spectral lines, 82, 109
spectra, 316
line strengths, 322
X-ray
satellites,
323
Zeeman effect, 8, Chapter xvi

in hyperfine structure, 426
in one-electron spectra, 149
Zero-order scheme, 167, 287
INDEX OF NAMES
Albertson, 335
Allen, 374
Allison, 322, 323
Anderson, 125
Armstrong, 322
Auger, 371
Bacher, 198, 209, 276, 366, 426, 427
Back, Chapter xvi, 427
Backlin, 325
Baker, 338
Bakker, 396
Balmer, 3
Banerji, 132
Bartlett, 229, 264, 282, 358, 418
Bateman, 84, 131
Bechert, 132
Becker, 415
Bethe, 132, 140, 351
Beutler, 368, 373, 374
Birge, 138,428
Black, 283, 358, 359, 362
Blaekett, 125
Blair, 394
Blaton, 256
Bocher, 17, 28
Bohr, 4, 5, 9, 87 et seq., 327, 401
Born, 10, 28, 61
Bothe, 104
Bouma, 394
Bowen,
146, 254,
282
Boyce, 282
Breit, 105, 126, 147, 211, 349, 420, 425,
426
Brickwedde, 138
Brillouin, 37, 339, 355
Brinkman, 97
Brown, 358, 419
Briick, 423
Burger, 141, 238, 387
Bush, 337, 358
Cady, 200, 203

Caldwell, 337, 358
Campbell, 396
Carst, 107, 140
Catalan, 8
Chaffee, 138
Chalk, 402
Cohen, 423
Compton, 372
,>
149, 169, 197, 206, 249, 271, 282,
304,375,406,417
Coster, 326
Courant, 16
Darwin, C. G., 56, 104, 128, 131, 149,

390
Darwin, K., 149, 157, 390
De BrogHe, 7
Debye, 81
Dewey, 405, 415
Dirac, 7 et seq., Chapter n, 55, 60, 61, 81, 103,
124 et seq., 192,240
Dobrezov, 423
Doermann, 425
Doi,402
Dorgelo, 238 et seq.
Drayvestyn, 326
Dnane, 322
Dunn, 358
Eckart, 115, 131, 349, 352, 418
Einstein, 4, 79
Mason, 240
Engwicht , 245
Epstein, 6, 131, 397 et seq.
Esmarch, 104
Estermann, 421
Fermi, 147, 335
efe
seq., 376,
422 et seq.
Fiiippov, 147, 148, 423
Fock,354
Foster, 402, 413, 415
Franck, 5
Frenkel, 10, 120
Frerichs, 244, 396, 400
Frisch,421
Fues, 342
FukusMma, 417
Gamow, 407
Gaunt, 176, 182, 211, 357
Gebauer, 403
Gibbs, J. W., 43
Gibbs, R. C., 10, 140, 208
Ginsburg, 310
Gmelin, 394
Goldberg, 252
Gordon, 128, 131, 400
Goudsmit, 8, 10, 54, 126, 189, 195, 198, 209,
219, 274, 276, 301, 322, 375, 387,424,426,
427
Grace, 140
Granath, 423, 424, 425
Gray, 226
Green, 394
Gropper, 375
Gross, 423
Grotrian, 416
Guillemin, 352
Gumev, 406
OF
440
Guth, 336
Giittinger, 64,
425
Hamilton, 7
Hansen, 140
Hara-eaves. 35S, 422
Harms, 372
Harrison, 240, 244, 245, 278, 369
Hartley, 3
Hartree, 283, 339, 342, 354 et seq., 429
Harvey, 425
Heisenberg. 7 et seq., 28, 88, 107, 108, 149,
211,240,300,350,390
Lannor, 103
Lawrence, 373
Lennard-Jones, 336
Leiy, 281
Lindsay, 358
Locher^ 371
Lodge, 2, 79
Lorentz, 4, 103, 379
Loring, 394
Lozier, 348
Limdblad, 104
Limeland, 401
Lysnaai, 4
Hertz, G., 5
Hertz, H. R. f 2
Hilbert, 16
Mark, 402
Hiyama, 403
Martin, 372
Majorana, 373
Hobsoii, 1S2
Honl, 239, 3S7
Houston, 97, 138, 140, 271
Hsiek 140
Huff, 97
Hughes, 41 S
Humphreys, 219
Hund, S, 9, 10,
2>9,
336
Hyileraas, 341 et seq. 9 365

Inzils, 259, 274, 304,
310
Isfida, 403, 415, 417
Jackson, 423
Jeans, 81
Jenkins, 425
423
Johnson. 70, 213, 220, 244, 266, 278, 295,
369
Jonsson, 323
Jordan, 8, 10, 28, 61, 149, 169, 390
Joffe,
Sfasên, J. C., 2
llarweH, L. R., 132, 141
MoDougall, 358, 359
McLennan^ 395
Meggezs, 200, 245
Menzd, 282
Menifl, P. W., 254
Merrill, E. A., 276
Micheison, 428
iliiiancziik, 396
3kOflikan, 146
31ilne, 148, 338
MinkovsM, 400
Minnaert, 111
Mishima, 420
Mitchell, 97
Moore, 254
Morse, 10, 149
Moseiey, 319
Mott, 10
Mulders, 111
Murphy, 138
Kamijima, 415
Kassner, 400
Kagaoka, 420
Newton, 1
Kayser, 3
Kellner, 346
Kelloscr, 421
Kent, 140, 153
Kimball, 44
OccMalini, 125
et seq.
Kirchner, 138
Kirschbaum, 401
Kiuti, 402, 403
Kopfermann, 107
Korff, 147
Kramers, 107, 339, 342, 397

Kronis, 239, 245, 387
KufaiClOS
Kiipper, 132
Ladenburg, 107, 111, 140, 240, 281, 415, 416

Lanczos, 406 et seq.
8, 146, 193, 219, 381 et seq.
Lancrer, 325, 367
Land,
Langstroth, 368
Laporte, 236, 304
Oppenheimer, 211, 406

Omstein, 141, 238, 387
Oseen, 104
Ostrofsky, 208
3, 386, 390 et seq., 426
Patterson, 322
Pauli, 8, 9, 55, 64, 78, 164, 166, 189, 390, 415,
421, 425
Pauling, 10, 322
Pasehen,
Payne, 282
Pearson, F., 428
Pearson, H., 140
Pease, 428
Peieris, 336
Perry, 138
Phillips, 369
Pincherle, 368
INDEX OF NAMES
Planck, 4
Preston, 152, 381
441
Takamine, 403
Taylor, 140
Terenin, 423
Thomas, 108, 120, 335
Torrance, 358
Prokofjew, 148, 254, 256, 342

Rabi, 421, 423
Racah, 420
Ramsauer, 416
Rasmussen, 369
Trampy,
Rausch von Traubenberg, 403

Rayleigh, 81
Reiche, 111
Uffoid, 53, 77, 233, 235, 249, 252, 367
Uhlenbeck, 8, 54, 126

Undheim, 348
Unsold, 111, 132,139,415
Trey, 10, 138, 342, 423
Richtmyer, 326
Ritz, 3, 4, 345
Robertson, 405
Van Atta, 423, 424

Van der Waeiden, 11
Van Geei, 387, 390 et seq.
Van Milaan, 244
Van Vleck, 6, 89, 117,
Rojansky, 399, 415

Rosenthal, 420
Rozental, 310
Ruark, 10
Rubinowicz, 93,
Runge, 3, 381
147, 148
Turner, 1S8
94, 253,
256
192,
276,
402
Russell, 8, 187, 188, 200, 208, 236, 239, 245,

367 et seq*
Vinti, 109
Rutherford, 4
Von Neumann,
Voigt, 390, 397

10, 12, 27,
262
Rydberg, 3
Waller, 402
Watson, 114, 340

Weisskopf 82
Wendt, 400
Sadler, 359
Sana, 132
Sanford, 254
Saunders, 187
Schlapp, 399
Wentzel, 323, 325, 339, 371, 372, 402

Wetterblad, 324
Sehrodinger, 7, 9, 27, 116, 131, 398, 400

Schiiler, 423, 425
Schiitz, 111,
240
Sehwarzschild,
6,
WMttaker, 114
Wien, 402
397
Segre, 396, 424, 426
76
Seward, 246
Shane, 140
Shenstone, 188, 367 et
Sherman, 322
Wierl, 401, 402
Seitz,
Wigner,
seq., 373, 374,
394
Shortley, 44, 53, 197, 206, 213, 233, 244,

310
271, 279, 281, 284, 295, 304, 306,
Shrum, 395
Siegbahn, 316, 322, 323
Sjogren, 400
Slack, 132, 401
Slater, 139, 158, 352, 354
Snoek, 140
Sommer, 395
Sommerfeld, 6, 10, 132, 139, 239, 319, 322, 390
Spedding, 140
Stark, 397, 400 et seq.
Stenstrom, 322
Stern, 421
Stevenson, 256, 282
Stewart, 111
Sugiura, 132, 148, 342
Weyl, 11
White, 10, 208, 235, 240, 369, 374
WMtelaw, 369
11, 75, 76, 82, 169,
262
Wilber, 208
Williams, 140
Wills, 226,
426
Wilsar, 401
Wilson, A. H., 36
Wilson, E. Bright, 351 et seq.
Wilson, Edwin B., 43
Wolfe, 273, 326
Woods, 336
Wulff,427
Wurm, 425
Young, 338
k,
322
Zacharias. 421
Zeeman, I, 378, 395, 427

Zemansky, 97
Zener, 352, 353
Zwaan, 148, 392, 393, 412
369,
2
C ^^
fjj
QQ
rn
?5
,P
^3 CO
"TJ
106 091

Ebook Condon The Theory of Atomic Spectra

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Ebook Condon The Theory of Atomic Spectra

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ebook Condon The Theory of Atomic Spectra

Uploaded by

Copyright:

Available Formats

KANKAK

The theory of atomic

THE UNIVERSITY PEESS

First printed in Great Britain at the

HENRY NOEBIS RUSSELL

Symbolic algebra of states and observable

Representations of states and observables

Continuous eigenvalues and the Sehrodinger representation

The laws of quantum mechanics

R6sum6 of the perturbation theory

Perturbation caused by a single state

analysis of non-commuting vectors

Definition of angular momentum

Vector addition of angular

The matrix of (Ji+J,)

Dependence of the matrix of T on m

P which commutes with J

Transformation amplitudes for vector addition

Classical electromagnetic theory

Expansion of the retarded potential

The quadrupoie-radiation field

Spectral lines in natural excitation

Induced emission and absorption

Dispersion theory. Scattering.

principle for emission

Natural shape of absorption

Radial functions for hydrogen

Sketch of the relativistic theory

Experimental results for hydrogenic spectra

General structure of the alkali spectra

Intensities in alkali spectra

THE CENTRAL-FIELD APPROXIMATION

The Hamiltonian for many-electron atoms

The dynamical equivalence of the

Conventions concerning quantum numbers. Closed shells

Matrix components for S^gf(i,^)

Specialization for closed shells

Odd and even states

electron outside closed shells

THE RUSSELL-SAUNDERS CASE; ENERGY LEVELS

The JD/S-coupling scheme

Absolute term intervals

Formulas and experimental comparison

in the nlx configurations

Vector coupling in antisymmetric states

Genealogical characterization of i/S-coupling terms

intervals for terms of coupled groups

Calculation of eigenfunctions from vector-coupling formulas

Configuration selection rules

Multiplet strengths obtained from spectroscopic stability

Line strengths in Russell-Saunders multiplets

The jj-coupling scheme and the spin-orbit interaction

Matrix of spin-orbit interaction for

Configurations containing almost closed shells

The transformation to jj coupling

Pure almost-closed-shell configurations

configuration p* pin the rare gases