On the Quantum Theory of Molecules

M. Born
a
, J.R. Oppenheimer
b
a
Institute of Theoretical Physics, G ottingen
b
Institute of Theoretical Physics, G ottingen
Abstract
It will be shown that the familiar components of the terms of a molecule; the energy of electronic motion, of the nuclear
vibration and of the rotation, correspond systematically to the terms of a power series in the fourth root of the ratio of
electron mass to (average) nuclear mass. The treatment yields among other things an equation for the rotation, which
represents a generalization of the treatment of Kramers and Pauli (top with built-in y-wheel). Furthermore, there
appears a justication of the considerations of Franck and Condon on the intensity of band lines. The relationships
are illustrated for the diatomic molecule.
Introduction
The terms of molecular spectra are usually made up of parts of various orders of magnitude; the largest con-
tribution comes from the electronic motion about the nuclei, then follows the contribution of the nuclear vibration,
and nally that from the nuclear rotation. The basis for the possibility of such a classication obviously rests in the
comparative magnitudes of nuclear and electronic masses. From the standpoint of the old quantum theory, which
computed stationary states with the aid of classical mechanics, this is the concept applied by Born and Heisenberg
[1]; it was shown that the energy terms appear as the terms of increasing order with respect to the ratio

m/M, where
m is the electronic mass and M an average nuclear mass. Thereby, however, nuclear rotation and vibration both appear
in the second order, which contradicts empirical ndings (for small rotational quantum numbers).
Here the problem will be approached anew from the standpoint of quantum mechanics.
1
It then becomes necessary
to make our development with respect to (m/M)
1/4
rather than with respect to

m/M, so as to obtain the natural order
of energy terms. The considerations also become much simpler and more transparent than in the old theory. The
nuclear vibrations correspond to terms of second order and the rotations to fourth order in the energy, while the rst
and third order terms vanish. The absence of the rst order terms is related to the existence of an equilibrium position
of the nuclei, in which the electronic energy for stationary nuclei is at a minimum. The fourth order terms for the
rotational motion illustrate the generalization of the treatment of Kramers and Pauli [2] in which the the behaviour of
a molecule is compared to that of a top with a built-in y-wheel. In order to determine the eigenfunctions and thereby
the transition probabilities only to the zeroth approximation, the energy calculation must be carried out to terms of
fourth order (rotations). One obtains expressions for the probabilities of simultaneous jumps of electronic, vibrational
and rotational quantum number through which the representations developed by Franck [3] and elaborated by Condon
[4] may be given precise interpretation.
$
Published originally as Annalen der Physik, 84, 457-484 (1927)
$$
Translated by S M Blinder with emendations by Brian Sutclie and Wolf Geppert
1
Through the discussion of the basis of this work with us, Dr. P. Jordan has helped us with valuable comments, for which we express our thanks.
The approximations to higher than fourth order will not be treated in this work; they correspond to coupling among
the three basic types of motion. A calculation of this eect is only meaningful for simultaneous consideration of all
degeneracies of electronic motion for stationary nuclei, especially the Heisenberg resonance degeneracy which arises
from the equivalence of electrons (also possibly of some nuclei) and in diatomic molecules, from the degeneracy of
the eigenrotation about the internuclear axis; these complicated considerations will be forgone here.
As an example we will consider diatomic molecules in detail, using not only the general method but also another
utilising the separation of variables in which the rotation becomes signicant even in the zeroth order approximation,
as Born and H uckel [5] have done it in the older quantum theory.
Part I. Notation and Denitions
We denote the mass and rectangular coordinates of the electrons by m, x
k
, y
k
, z
k
and of the nuclei by M
l
, X
l
, Y
l
, Z
l
.
Letting M be any average value of the M
l
, we set
=
_
m
M
_
1/4
(1)
and
M
l
=
M

l
=
m

l
; (2)
the
l
being dimensionless numbers of order of magnitude 1. Let the potential energy of the system be
U(x
1
, y
1
, z
1
, x
2
, y
2
, z
2
, . . . ; X
1
, Y
1
, Z
1
, X
2
, Y
2
, Z
2
, . . . ) = U(x, X) (3)
where we denote by x the totality of electronic coordinates and by X, that of the nuclear coordinates. The function
U depends only on the relative positions of the particles; however, we make no use of its particular form (Coulombs
law). The kinetic energy of the electrons is represented by the operator
T
E
=
h
2
8
2
m

2
x
2
k
(4)
where the symbol

x
denotes the sum which arises from the above expresion by cyclic permutation of x, y and z.
The kinetic energy of the nuclei is
T
K
=
4
h
2
8
2
m

2
X
2
l
. (5)
The total energy is represented by the operator
H = H
0
+
4
H
1
(6)
where
T
E
+ U = H
0
_
x,

x
; X
_
T
K
=
4
H
1
_

X
_
. (7)
We introduce now, in place of the rectangular coordinates of the nuclei, 3N 6 functions

i
=
i
(X) (8)
2
which denote the relative positions of the nuclei with respect to one another, and 6 functions

i
=
i
(X) (9)
which determine the position and orientation of the nuclear conguration in space. One can in a symmetrical fashion
introduce the rectangular coordinates

X
l
,

Y
l
,

Z
l
of the nuclei relative to the instantaneous principal axes of inertia;
between these there are 6 relations:

l
M
l

X
l
= 0

l
M
l

Y
l

Z
l
= 0 . . .
One can thus express the

X
l
by the 3N 6 independent parameters
1
,
2
, . . . :

X
l
=

X
l
(), . . .
There then exist transformations between the original and the new coordinates, of the form
X
l
= X
0
+

xy
(, , )

Y
l
(); (10)
X
0
, Y
0
, Z
0
are the coordinates of the center of mass and the
xy
are the coecients of the orthogonal rotation matrix,
and are thus known functions of the Eulerian angles , , . The quantities X
0
, Y
0
, Z
0
, , , are the functions denoted
by
i
in (9). By (10), the X
l
are determined as functions of the
i
and
i
; by solving, one obtains the expressions (8)
and (9).
2
This transformation does not, of course, separate the energy H into parts corresponding to translation, rotation and
relative motion of the nuclei. However one can separate H
1
into three parts:
H
1
= H

+ H

+ H

; (11)
H

is linear homogeneous in the

2

j
; H

contains the

i
; H

is independent of all derivatives with respect to the

i
. One can make further generalizations about these operators. If we apply the entire operator H
1
to an arbitrary
function f () of the relative nuclear coordinates
i
, the resulting quantity H
1
f () must be independent of the position
in space, hence of the
i
. In particular, in H

the coecients of the

2

j
cannot depend on the
i
. In contrast, these
do appear in H

, associated with the

i
, the
i
,
i
and

i
; in H

associated with

2

j
the

i
,
i
and
i
.
We will consider these operators explicitly for
diatomic molecules.
The mechanical problem we must solve is
(H
0
+
4
H
1
W) = 0. (12)
We will show that any arbitrary solution which corresponds to a combination of nuclei and electrons forming a stable
molecule can be found by a development in a power series in .
Part II. Electronic Motion for Stationary Nuclei
If one sets = 0 in (12) one obtains a dierential equation in the x
k
alone, the X
l
appearing as parameters:
_
H
0
_
x,

x
; X
_
W
_
= 0. (13)
2
It is of physical signicance that this solution is in general made using ambiguous functions; compare [6].
3
This represents the electronic motion for stationary nuclei. We assume this eigenvalue problem is solved. The eigen-
values depend only on the functions
i
of the X
i
; then one can use the coordinate system dened by the principal axes
of inertia, i.e. let X
l
=

X
l
(). In this system of axes, the eigenfunctions depend, besides on x
k
, only on the
i
; however,
if one transforms back to the arbitrary space-xed axes, the
i
again become involved.
We designate the nth eigenvalue and the corresponding normalized eigenfunction as
W = V
n
() =
n
(x; , ) (14)
so that the identity
_
H
0
_
x,

x
; ,
_
V
n
()
_

n
(x; , ) = 0 (15)
is valid. Here we assume that V
n
is a nondegenerate eigenvalue. As a matter of fact, this is never the case since
the indistinguishibility of the electrons introduces the resonance degeneracy, discovered by Heisenberg and Dirac;
for diatomic molecules there is an additional degeneracy of the angular momentum about the axis. But since we are
concerned here only with the systematics of the approximation procedure, we will not consider these degeneracies.
Their consideration would result in secular equations in the higher approximation.
The most important goal of our investigation is the proof that the function V
n
() plays the role of a potential for the
nuclear motion. For this we must have several auxilliary formulas which will be derived now. It is necessary to show
that the matrix corresponding to the derivative of the operator H
0
(x,

x
; , ) with respect to
i
, (for constant x,

x
) can
be related to the derivative of the function V
n
().
Instead of taking the derivative with respect to the
i
directly, we replace the
i
by
i
+
i
and dierentiate with
respect to ; the coecient of a power of is then a homogeneous polynomial in
i
, these coecients being derivatives
with respect to
i
. Thus we write
V
n
( + ) = V
(0)
n
+ V
(1)
n
+
2
V
(2)
n
+ . . . , (16)
where
a) V
(0)
n
= V
n
()
b) V
(1)
n
=

i
V
n

i
c) V
(2)
n
=
1
2

i j

2
V
n

j
,
. . . . . . . . . . . .
(17)
and correspondingly
H
0
= H
(0)
0
+ H
(1)
0
+
2
H
(2)
0
+ . . .

n
=
(0)
n
+
(1)
n
+
2

(2)
n
+ . . . (18)
. . . . . . . . . . . .
One can now develop the quantities
(1)
n
,
(2)
n
in the eigenfunctions
(0)
n
(x; , ), setting
a)
(1)
n
=

u
(1)
nn

(0)
n

,
b)
(2)
n
=

u
(2)
nn

(0)
n

.
(19)
4
Thus u
(r)
nn

is a homogeneous polynomial of the rth order in

i
, for instance
u
(1)
nn

i
_

(0)
n

(0)
n

i
dx
u
(2)
nn

i j

j
_

(0)
n

(0)
n

j
dx. (20)
These integrals, in which dx denotes the volume element in conguration space, are independent of the orientation of
the nuclear system in space, hence independent of the
i
; one can thus evaluate them in the principal axis system.
If now, F denotes any operator on the x
i
, we dene the rth order matrix element of F
_

(0)
n

F
(r)
n
dx = F
(r)
nn

. (21)
For r = 0 this becomes the usual matrix element
F
(0)
nn

= F
nn
=
_

(0)
n

F
(0)
n
dx. (22)
In general, by (19),
F
(r)
nn

u
(r)
nn

F
n

n
. (23)
Using (15) for = 0
(H
(0)
0
V
(0)
n
)
(r)
nn

= u
(r)
nn

(V
(0)
n

V
(0)
n
). (24)
Furthermore, we obtain by substituting (16) and (18) in (15), the following identities:
a) (H
(0)
0
V
(0)
n
)
(1)
n
+ (H
(1)
0
V
(1)
n
)
(0)
n
= 0
b) (H
(0)
0
V
(0)
n
)
(2)
n
+ (H
(1)
0
V
(1)
n
)
(1)
n
+ (H
(2)
0
V
(2)
n
)
(0)
n
= 0
. . . . . . . . . . . .
(25)
Multiplying these by
(0)
n

and integrating over the x

i
, by virtue of (24) we nd:
a) u
(1)
nn

(V
(0)
n

V
(0)
n
) + (H
(1)
0
)
nn
V
(1)
n

nn
= 0
b) u
(2)
nn

(V
(0)
n

V
(0)
n
) + (H
(1)
0
V
(1)
n
)
nn
+ (H
(2)
0
)
nn
V
(2)
n

nn
= 0
. . . . . . . . . . . .
(26)
From these one can compute the (H
(1)
0
)
nn
, (H
(2)
0
)
nn
, . . . , ie the matrix elements
_
H
0

i
_
nn

,
_

2
H
0

j
_
nn

, . . . . We will later
apply these formulas.
3
Part III. Setting-up the Approximate Equations
An arbitrary conguration of electrons and nuclei cannot always be treated by a general approximation procedure.
We will here consider only states which correspond to a stable molecule. We will begin with the following question:
3
The classical analogue to the simplest deduction from these formulae, namely the identity (H
(1)
0
)
nn
= V
(1)
n
which follows from (26a) for n = n

,
is found in [7]; compare especially with 4, formula (11).
5
Is there a system of values of the relative nuclear coordinates
i
such that the eigenfunctions
n
of the energy oper-
ator (6), in so far as they depend on the
i
, have values signicantly dierent from zero only in a small neighbourhood
of this set?
This wave-mechanical requirement corresponds obviously to the classical condition that the nuclei undergo only
small oscillations about the equilibrium conguration; the |
n
|
2
is the probability of nding a certain conguration of
given energy.
We consider, as the unperturbed system, the electronic motion for an arbitrary but henceforth xed nuclear con-
guration,
i
. We then develop all quantities with respect to small changes of the
i
, which we designate by
i
; we
presume then that the domain of oscillation is such that is close to zero, an assumption which is only justied by
its success.
We have then as in (18), part II, the development
H
0
(x,

x
; + , ) = H
(0)
0
+ H
(1)
0
+
2
H
(2)
0
+ . . . , (27)
where
a) H
(0)
0
= H
0
(x,

x
; ),
b) H
(1)
0
=
_
i

i
H
0

i
,
c) H
(2)
0
=
1
2
_
i j

i

j

2
H
0

j
,
. . . . . . . . . . . .
(28)
and from (11) since

=
1

4
H
1
(X,

X
) =
4
_
1

2
H

+
1

+ H

_
(29)
=
2
H
(0)

+
3
_
H
(0)

+ H
(1)

_
+
4
_
H
(0)

+ H
(1)

+ H
(2)

_
+ . . .
where
a) H
(0)

= H
(0)

(,

2

j
)
b) H
(1)

=
_
i

i
H
(0)

i
. . . . . . . . . . . .
(30)
a) H
(0)

= H
(0)

(, ,

)
b) H
(1)

=
_
i

i
H
(0)

i
. . . . . . . . . . . .
(31)
a) H
(0)

= H
(0)

(, ,

2

j
)
b) H
(1)

=
_
i

i
H
(0)

i
. . . . . . . . . . . .
(32)
6
The arguments
i
are hereafter to be considered constants.
The total energy operator is then
H = H
0
+ H
(1)
0
+
2
_
H
(2)
0
+ H
(0)

_
+
3
_
H
(3)
0
+ H
(0)

+ H
(1)

_
+
4
_
H
(4)
0
+ H
(0)

+ H
(1)

+ H
(2)

_
+ . . . (33)
The succeeding terms all have the same form and can be formed from the term in
4
by increasing the superscript by
1.
We also develop the desired eigenfunction and energy parameter with respect to :
=
(0)
+
(1)
+
2

(2)
+ . . .
W = W
(0)
+ W
(1)
+
2
W
(2)
+ . . . (34)
We then obtain the following approximation equations:
a) (H
(0)
0
W
(0)
)
(0)
= 0
b) (H
(0)
0
W
(0)
)
(1)
= (W
(1)
H
(1)
0
)
(0)
c) (H
(0)
0
W
(0)
)
(2)
= (W
(2)
H
(2)
0
H
(0)

)
(0)
+ (W
(1)
H
(1)
0
)
(1)
d) (H
(0)
0
W
(0)
)
(3)
= (W
(3)
H
(3)
0
H
(0)

H
(1)

)
(0)
+(W
(2)
H
(2)
0
H
(0)

)
(1)
+ (W
(1)
H
(1)
0
)
(2)
e) (H
(0)
0
W
(0)
)
(4)
= (W
(4)
H
(4)
0
H
(0)

H
(1)

H
(2)

)
(0)
+(W
(3)
H
(3)
0
H
(0)

H
(1)

)
(1)
+(W
(2)
H
(2)
0
H
(0)

)
(2)
+ (W
(1)
H
(1)
0
)
(3)
. . . . . . . . . . . .
(35)
Part IV. Solution of the Approximate Equations of zeroth and rst Order: Equilibrium of the Nuclei
The zeroth order equation (35a) describes the electronic motion for stationary nuclei as discussed in Part II. From
the normalized eigensolution
(0)
n
(x; , ) belonging to the eigenvalue V
(0)
n
= V
n
(), we nd the general solution in the
form:

(0)
n
=
(0)
n
(, )
(0)
n
(x; , ) (36)
where
(0)
n
is an, as yet, arbitrary function of the arguments
i
,
j
; this must be included in order to enable solutions of
the following approximation equations.
The following approximation equation (35b)
(H
(0)
0
W
(0)
)
(1)
= (W
(1)
H
(1)
0
)
(0)
(37)
is soluble only when the right-hand side is orthogonal to
(0)
n
(relative to the electronic coordinates x
i
).
4
This gives the condition
__
H
(1)
0
_
nn
W
(1)
_

(0)
n
(, ) = 0 (38)
4
We dene the orthogonality of two functions f (x) and g(x) by
_
f (x)g(x)dx = 0.
7
where
_
H
(1)
0
_
nn
is the diagonal matrix element of the operator H
(1)
0
relative to the x
i
, thus by (28b) a homogeneous
linear function of
i
. This must however, by (38), be constant, since
(0)
n
(, ) cannot vanish identically without the
same being true for
(0)
n
.
Thus it follows that
W
(1)
= 0,
_
H
(1)
0
_
nn
= 0. (39)
From (26a) and (17) we have however
_
H
(1)
0
_
nn
= V
(1)
n
=

i
V
n

i
.
Thus:
V
n

i
= 0. (40)
The validity of continuing our approximation procedure requires that the relative nuclear coordinates
i
must not
be arbitrarily chosen, but must correspond to an extremum of the electronic energy V
n
(). The existence of this is
therefore the condition for the existence of the molecule, a law which is usually assumed to be self-evident. We will
show later that it must necessarily be a minimum as well.
The function
(0)
n
(, ) remains, as yet, undetermined. Setting in (37) W
(0)
n
= V
n
(), W
(1)
n
= 0 and
(0)
n
=
(0)
n

(0)
n
we nd the equation which determines
(1)
n
_
H
(0)
0
V
(0)
n
_

(1)
n
= H
(1)
0

(0)
n

(0)
n
. (41)
A solution of this by (25a) is
(1)
n
=
(0)
n

(1)
n
where
(1)
n
is the function (19a) dened by (18). The general solution is
obtained by adding a solution
(0)
n
of the homogeneous equation with the yet undetermined factor
(1)
n
(, ):

(1)
n
=
(0)
n

(1)
n
+
(1)
n

(0)
n
. (42)
Part V. Solution of the Approximate Equations of second and third Order:Nuclear Vibration
We now reach the approximation equation (35c), which after substitution of the solutions for the lower order
approximations is
_
H
(0)
0
V
(0)
n
_

(2)
n
=
_
W
(2)
n
H
(2)
0
H
(0)

(0)
n

(0)
n
H
(1)
0
_

(0)
n

(1)
n
+
(1)
n

(0)
n
_
. (43)
In order for this to be solvable, the right-hand side must again be orthogonal to
(0)
n
; using the notation of part II this
yields, because of (39):
__
H
(2)
0
+ H
(0)

_
nn
+
_
H
(1)
0
_
nn
W
(2)
n
_

(0)
n
= 0.
It follows from (26b) with V
(1)
n
= 0 that
_
H
(2)
0
_
nn
+
_
H
(1)
0
_
(1)
nn
= V
(2)
n
. (44)
Since H
(0)

by (30a) is seen to be independent of the x

k
we nd:
_
H
(0)

+ V
(2)
n
W
(2)
n
_

(0)
n
= 0. (45)
Noting the meanings of H
(0)

and V
(2)
n
given by (17c) and (30a) we see that (45) represents the equation for harmonic
nuclear vibration:
_

_
H
(0)

_
,

2

j
_
+
1
2

i j

2
V
n

j
W
(2)
n
_

(0)
n
= 0. (46)
8
This equation shows that the function V
n
() plays the role of a potential energy for the nuclei, up to terms of 2nd order.
For the existence of a stable molecule there is a further condition that the extremum of V
n
() determined by (40) must
be a minimum; then the quadratic form V
(2)
n
must be positive denite, thereby all degrees of freedom
i
stable and
oscillating about the equilibrium conguration are possible. It is known that the equation for the vibration (46) is
separable through a linear transformation of the
i
to normal coordinates
i
. If
(0)
ns
() be the normalized eigensolution
of (46) belonging to the eigenvalue W
(2)
ns
, the general solution is
a) W
(2)
= W
(2)
ns
,
(0)
n
=
(0)
ns
, where
b)
(0)
ns
=
(0)
ns
()
(0)
ns
().
(47)
The index s thus represents the set of vibrational quantum numbers.
(0)
ns
is an, as yet, undetermined function of the
i
,
the introduction of which is necessary for the continuation of the procedure.
It is known that
(0)
ns
() is a linear combination of products of orthogonal Hermite functions for the individual
normal coordinates
i
; these functions have the property that they approach zero very rapidly (exponentially) outside
the limit of classical vibration. So our substitution of ( +) is justied since it indeed leads to a solution, with regard
to the -oscillation within the limit, which vanish with . We apply the further property of the orthogonal Hermite
functions that they are either even or odd functions of their argument.
Let be any operator on the
i
. We can then construct the corresponding matrix

nn

ss

=
_

(0)
n

(0)
ns
d (48)
where d is the volume element in the space of the
i
.
In order to solve equation (43) we substitute on the right-hand side, using (45),
_
W
(2)
ns
H
(0)

(0)
ns
= V
(2)
n

(0)
ns
;
(43) then becomes:
_
H
(0)
0
V
(0)
n
_

(2)
n
=
_
V
(2)
n
H
(2)
0
_

(0)
ns

(0)
n
H
(1)
0
_

(0)
ns

(1)
n
+
(1)
ns

(0)
n
_
. (49)
The general solution is

(2)
n
=
(0)
ns

(2)
n
+
(1)
ns

(1)
n
+
(2)
ns

(0)
n
, (50)
where
(2)
ns
denotes a new, undetermined function of the
i
,
j
; this is easily veried using the identities (25).
We now consider the approximation equation of 3rd order (35d); after substitution of the already determined
quantities, this becomes:
_
H
(0)
0
V
(0)
n
_

(3)
n
=
_
W
(3)
H
(3)
0
H
(0)

H
(1)

(0)
ns

(0)
n
+
_
W
(2)
ns
H
(2)
0
H
(0)

_ _

(0)
ns

(1)
n
+
(1)
ns

(0)
n
_
H
(1)
0
_

(0)
ns

(2)
n
+
(1)
ns

(1)
n
+
(2)
ns

(0)
n
_
. (51)
We may consider the right-hand side as a development in the
(0)
n
; we write
_
H
(0)
0
V
(0)
n
_

(3)
n
= W
(3)

(0)
ns

(0)
n

F
(3)
nn

(0)
n

, (52)
where
F
(3)
nn

= F
(3,1)
nn

(2)
ns
+ F
(3,2)
nn

(1)
ns
+ F
(3,3)
nn

(0)
ns
; (53)
9
where the F are operators on and , and
a) F
(3,1)
nn

=
_
H
(1)
0
_
nn

b) F
(3,2)
nn

=
_
H
(0)

H
(2)
0
W
(2)
ns
_
nn

+
_
H
(1)
0
_
(1)
nn

,
(54)
we can say about F
(3,3)
nn

only that it is a homogeneous function of 3rd degree in the

i
and the /
i
.
If (52) is solvable, we must have
W
(3)

(0)
ns
F
(3)
nn
= 0
because of (53) and (54a)
F
(3,2)
nn

(1)
ns
=
_
W
(3)
F
(3,3)
nn
_

(0)
ns
, (55)
where, by (54b) and (44)
F
(3,2)
nn
= H
(0)

V
(2)
n
W
(2)
ns
.
Thus (55) is the inhomogeneous equation corresponding to the vibration equation (45); since (45) has the normalized
solution
(0)
ns
belonging to the eigenvalue W
(2)
ns
, (55) is solvable only when the right-hand side multiplied by
(0)
ns
has a
vanishing integral over -space. This gives, using (47b), a dierential equation for
(0)
ns
():
_
F
(3,3)
nn
ss
W
(3)
_

(0)
ns
= 0.
However, F
(3,3)
nn
is odd in the
i
and /
i
so the diagonal element of the -matrix must vanish. When one transforms
to the normal coordinates
i
,
(0)
ns
becomes a sum of products of orthogonal Hermite functions, F
(3,3)
nn
, a polynomial of
odd order in the
i
and /
i
, so that every term contains at least one of
i
or /
i
in an odd power; therefore every
term in the -matrix vanishes. It follows therefore
W
(3)
= 0 (56)
and
(0)
ns
remains, as before, undetermined.
Now we may solve:

(1)
ns
= S
(1)
ns

(0)
ns
(57)
where S
(1)
ns
is the following operator with respect to the
i
:
S
(1)
ns
=

F
(3,3)
nn
ss

(0)
ns

W
(2)
ns
W
(2)
ns

. (58)
Finally the solution of (52):

(3)
n
=

F
(3)
nn

(0)
n

V
(0)
n
V
(0)
n

(59)
and by (53), this has the form:

(3)
n
=

_
G
(3,1)
nn

(2)
ns

(0)
n

+ G
(3,2)
nn

(1)
ns

(0)
n

+ G
(3,3)
nn

(0)
ns

(0)
n

_
, (60)
where
G
(3,2)
nn

=
F
(3,2)
nn

V
(0)
n
V
(0)
n

. (61)
10
Noting (54) we see that G
(3,1)
nn

is a number, G
(3,2)
nn

a dierential operator with respect to the

i
and G
(3,3)
nn

an operator
with respect to the
i
and
i
.
By (26a), Part II

G
(3,1)
nn

(2)
ns

(0)
n

_
H
(1)
0
_
nn

(0)
n

(2)
ns
V
(0)
n
V
(0)
n

u
(1)
nn

(0)
n

(2)
ns
=
(1)
n

(2)
ns
thus

(3)
n
=
(1)
n

(2)
ns
+

_
G
(3,2)
nn

(1)
ns

(0)
n

+ G
(3,3)
nn

(0)
ns

(0)
n

_
. (62)
Part VI. Solution of the Approximate Equations of fourth and higher Order: Rotation and Coupling Eects
After substitution of the quantities already determined, the 4th order approximation equation (35e) becomes:
_
H
(0)
0
V
(0)
n
_

(4)
n
=
_
W
(4)
H
(4)
0
H
(0)

H
(1)

H
(2)

(0)
ns

(0)
n
(63)

_
H
(3)
0
+ H
(0)

+ H
(1)

_ _

(1)
ns

(0)
n
+
(0)
ns

(1)
n
_
+
_
W
(2)
ns
H
(2)
0
H
(0)

_ _

(2)
ns

(0)
n
+
(1)
ns

(1)
n
+
(0)
ns

(2)
n
_
H
(1)
0
_

(1)
n

(2)
ns
+

_
G
(3,2)
nn

(1)
ns

(0)
n

+ G
(3,3)
nn

(0)
ns

(0)
n

_
_

_
.
We develop again the right-hand side in the
(0)
n
:
_
H
(0)
0
V
(0)
n
_

(4)
n
= W
(4)

(0)
ns

(0)
n

F
(4)
nn

(0)
n

, (64)
where
F
(4)
nn

= F
(4,2)
nn

(2)
ns
+ F
(4,3)
nn

(1)
ns
+ F
(4,4)
nn

(0)
ns
; (65)
here we have
F
(4,2)
nn

=
_
H
(0)

H
(2)
0
W
(2)
ns
_
nn

+
_
H
(1)
0
_
(1)
nn

(66)
and is identical with F
(3,2)
nn

(54b). While F
(4,3)
nn

is of odd order in the

i
, /
i
, F
(4,3)
nn

is of even order. The integrability

of (64) requires:
W
(4)

(0)
ns
F
(4)
nn
= 0;
this means that by (65)
F
(4,2)
nn

(2)
ns
=
_
W
(4)
F
(4,4)
nn
_

(0)
ns
F
(4,3)
nn

(1)
ns
. (67)
The left side agrees again with the vibration equation (45) because of (66). The right-hand side must also be
orthogonal to
(0)
ns
. Substituting the expressions for
(0)
ns
and
(1)
ns
from (47b) and (57), and using the symbol
()
(1)
ss

=
_

(0)
ns
S
(1)
ns
d =

_
F
(3,3)
nn
ss

_
ss

W
(2)
ns
W
(2)
ns

, (68)
11
we nd
_
F
(4,4)
nn
ss
+
_
F
(4,3)
nn
_
(1)
ss
W
(4)
_

(0)
ns
= 0. (69)
This equation determines nally the function
(0)
ns
(), hence the motion of the principal axes of inertia: the translations
and rotations. The principal term of the operator in (69) is the one which contains the second derivative with respect
to the
i
; a glance at (63) shows that it arises from H
(0)

(0)
ns

(0)
n
, the term corresponds in F
(4,4)
nn
to
_
H
(0)

_
n
=
_

(0)
n
H
(0)

(
(0)
n
. . . )dx, (70)
where in the place of the dots we have to put in the function which is operated upon. Since the operator (70) is
independent of the
i
, the diagonal elements of the corresponding s-matrix are identical with it. Physically the fact
that the complicated operators
_
H
(0)

_
n
appear instead of the simple operators H
(0)

indicates a coupling between the

top motion of the nuclei and the electronic motion.
These are, as we will later see for the case of the diatomic molecule, the same eects that Kramers and Pauli [2]
have tried to demonstrate using the assumption of a y-wheel built in to the top. Thus there are terms in (69) that
contribute to the operator H

; these correspond to a coupling of the top motion with angular momenta which are a
consequence of nuclear vibration. Finally, there are terms which do not concern the
I
; these are the additions to the
vibrational energy of order
4
.
Since the translations can always be separated in a trivial fashion, we consider only the rotations. If r be the
rotational quantum number, we have for the solution of (70)
W
(4)
= W
(4)
nsr
;
(0)
ns
=
(0)
nsr
(). (71)
Then one can solve (67) and nally also (64). It is of no use to write out the formulae explicitly.
Clearly, the procedure may be continued; however nothing new of signicance will appear. The higher approx-
imations describe couplings among rotations, vibrations and electronic motions. Quantum numbers other than the
ones already introduced do not enter.
We summarize now the consequences of our solutions. The most obvious result is that in order to determine
completely the eigenfunctions to 0th order it is necessary to solve the approximation dierential equations to 4th
order; we have

nsr
(x, , ) =
(0)
n
(x, , )
(0)
ns
()
(0)
nsr
() + . . . (72)
where
(0)
n
is the eigenfunction for electronic motion for stationary nuclei,
(0)
ns
that for nuclear vibration, and
(0)
nsr
that
for rotation. Thus are dened the vibrational coordinates
i
from an equilibrium conguration
i
which is dened by
the requirement that in this conguration the electronic energy V
n
() is a minimum. The determination of the three
functions
(0)
n
,
(0)
ns
and
(0)
nsr
yield the energy to 4th order:
W
nsr
= V
(0)
n
+
2
W
(2)
ns
+
4
W
(4)
nsr
+ . . . ; (73)
where V
(0)
n
is the minimum value of the electronic energy which characterizes the molecule at rest, W
(2)
ns
is the energy
of nuclear vibration, and W
(4)
nsr
contains (along with additional terms for the vibrational energy) the rotational energy.
In this approximation (to
4
) the three basic types of motion are separated; the coupling among them involves terms
of higher powers of .
Given (72) we can now calculate transition probabilities (intensities of bands).
The electrical moment of a molecule Mconsists of a nuclear part P and an electronic part p; the x-component is:
M
x
= P
x
+ p
x
, where
_

_
P
x
=
_
l
e
l
X
l
p
x
= e
_
k
x
k
. (74)
12
Hence from the set of matrix elements with respect to x
k
,
i
and
j
;
(p
x
) n
n

=
_
p
x

(0)
n

(0)
n
dx (75)
is a function of the
i
and
j
, then the
(p
x
) ns
n

=
_
(p
x
) n
n

(0)
ns

(0)
ns

d
(P
x
) ns
n

=
_
(P
x
)
(0)
ns

(0)
n

d (76)
are functions of the
j
, nally
(p
x
) nsr
n

=
_
(p
x
) ns
n

(0)
nsr

(0)
n

d
(P
x
) nsr
n

=
_
(P
x
) ns
n

(0)
nsr

(0)
n

d (77)
are numerical constants which determine the radiation and the transition probability for nsr n

. We can interpret
this step by step procedure as follows: for every electronic transition n n

, there corresponds a virtual oscillator

with moment (p
x
) n
n

; from this one obtains the matrix (p

x
) ns
n

which corresponds to a system of vibrational bands

(transitions from s s

), by a rule (somewhat dierent from the ordinary one) in which one uses one eigenfunction
of the lower and one of the upper electronic level (equation (76)). We repeat the procedure for the line of the band
corresponding to the transition r r

. The method of evaluation of the intensity of vibrational bands contained here

is rst given by Franck [3] and further developed by Condon [4].
These are determined by variation of the functions V
n
() and V
n
(); only in the neighbourhood of their minima
are the corresponding eigenfunctions
(0)
ns
and
(0)
n

signicantly dierent from zero; their product is so only when

these regions overlap. When the function V
n
() changes only slightly in an electronic transition n n

, the bands
corresponding to a small change of s will be intense; however if V
n
() changes greatly in the transition, an overlap of
the intervals in which
(0)
ns
and
(0)
n

do not vanish becomes possible only when the dierence s s

is large. These
relations are quantitatively discussed by Condon. Similar considerations apply for the rotations mutatis mutandis.
Part VII. Special Case of the Diatomic Molecule
As an example we will briey treat the diatomic molecule. Besides the resonance degeneracy, which is a con-
sequence of the indistinguishibility of the electrons, there is an additional degeneracy since corresponding to every
energy value there are two possible modes of motion in which the angular momentum about the internuclear axis
is oppositely directed. Since we are not concerned here with the ne structure of bands, we will not consider this
degeneracy; we limit our consideration to cases in which the angular momentum about the axis vanishes or when the
electronic energy is independent or only slightly dependent on the angular momentum component.
For two nuclei we have only one coordinate, the nuclear separation, and ve coordinates: the coordinates of
the center of mass X
0
, Y
0
, Z
0
, and the polar coordinates of the internuclear axis , .
The kinetic energy of the nuclei becomes
T
K
=
4
h
2
8
2
m
_

0
+

_
+

_
(78)
where
=
_
m
M
1
+ M
2
_
1/4
and =
(M
1
+ M
2
)
2
M
1
M
2
(79)
13
and

0
=

2
X
0
2
+

2
Y
0
2
+

2
Z
0
2
,

=
1
sin
2

2
+
1
sin

_
sin

_
. (80)
Thus:
H

=
h
2
8
2
m

2
,
H

=
h
2
8
2
m
2

, (81)
H

=
h
2
8
2
m
_

0
+

_
.
Substituting + for and developing in , we nd:
H
(0)

=
h
2
8
2
m

2
,
H
(p)

= 0, p = 1, 2, . . . (82)
H
(0)

=
h
2
8
2
m
2

,
H
(1)

=
h
2
8
2
m

, (83)
. . . . . . . . . . . . . . .
H
(0)

=
h
2
8
2
m
_

0
+

_
,
H
(1)

=
h
2
8
2
m
2

. (84)
. . . . . . . . . . . . . . .
The nuclear separation is determined by the equation
V

n
=
V
n

= 0. (85)
The equation for nuclear vibration is
_

h
2
8
2
m

2
+
1
2

2
V

n
()W
(2)
n
_

(0)
n
= 0. (86)
If we set
a =
8
2
m
h
2

W
(2)
n
b =
8
2
m
h
2

n
= b
1/4
(87)
we have [8]
_

2

2
+
_
a

b

2
__

(0)
n
= 0.
14
The eigenvalues are
a/b
1/2
= 2s + 1 (s = 0, 1, 2, . . . ),
with eigenfunctions

(0)
ns
= exp (
2
/2)H
s
(),
where H
s
is the sth Hermite polynomial.
The energy of the vibrations is thus:

2
W
(2)
ns
= a
h
2
8
2

m
= (2s + 1)b
1/2
h
2
8
2

m
=
_
s +
1
2
_
h
4
_

4

m
V

n
or

2
W
(2)
ns
=
_
s +
1
2
_
h
0
(88)
with
1
4
_

4

m
V

n
=
1
4
_
_
1
M
1
+
1
M
2
_
V

n
=
0
(89)
the frequency of the oscillator.
We set up now the equation (69) for the rotation, neglecting any detailed estimation of the correction to the
vibrational energy. Since H

by (81) does not contain derivatives with respect to the

j
, we need consider only the
term
_
H
(0)

_
n
in (69); all remaining terms we include in the constant C
ns
. The rotational equation (69) is then:
_
_
H
(0)

_
n
+ C
ns
W
(4)
_

(0)
ns
= 0. (90)
Since we have dropped the translational part from H

(0) , we have by (70) and (84) for an arbitrary function f ():

_
H
(0)

_
n
f () =
h
2

8
2
m
2
_

(0)
n

(
(0)
n
f )dx
and by (80)

(
(0)
n
f ) =
(0)
n

f + f

(0)
n
+ 2
_

_
1
sin
2

(0)
n

(0)
n

_
.
Thus
_
H
(0)

_
n
f =
h
2

8
2
m
2
_

f + f
_

(0)
n

(0)
n
dx
+ .
2
sin
2

_

(0)
n

(0)
n

dx + 2
f

_

(0)
n

(0)
n

dx
_
.
If we write

in the form:

2
+ ctg

+
1
sin
2

2
we see that it is convenient to introduce the following notation:

n
=
_

(0)
n

(0)
n

dx,
n
=
_

(0)
n

(0)
n

dx,

(2)
n
=
_

(0)
n

(0)
n

2
dx,
(2)
n
=
_

(0)
n

(0)
n

2
dx.
(91)
15
These quantities are the diagonal matrix elements of p

, p

, p
2

and p
2

(aside from a factor

h
2
2
,
h
2
4
2
respectively);
the rst two denote the average value of the electronic angular momentum about their corresponding Eulerian angle;
the second two, the average of the square of the angular momentum of electronic motion. We write then for (90)
explicitly:
__

2

2
+ 2
n

+
(2)
n
_
+ ctg
_

+
n
_
1
sin
2

_

2

2
+ 2
n

+
(2)
n
_
+
8
2
m
2
h
2

_
W
(4)
C
ns
_
_

(0)
ns
= 0 (92)
This is very similar to the equation of Kramers and Pauli for a rotor with a built-in y-wheel; the dierence is
essentially that they use the squares of the average values
n
2
and
n
2
, instead of the average of the squares
2
n
and

2
n
.
The dependence of the quantities in (91) on the angles and may be established by elementary considerations
if it is assumed that for this purpose the diagonal elements of the quantum mechanical matrix may be replaced by the
corresponding classical averages. One may decompose the motion of the electronic angular momentum vector into
an irregular variation without average rotations and a superimposed uniform rotation about the molecular axis. We
represent the variation in the average by a constant vector; this rotates uniformly about the axis. This exhibits the same
behaviour as a symmetric top with angular momentum components with respect to the top-xed coordinate system
having values L, M and N. From this we may express the components of the angular momentum in the , direction
as follows:
= L cos M sin
= L sin sin + M sin cos + N cos ,
where is the angle of the eigenrotation about the axis. Averaging over , we nd:
5
= 0 = N cos

2
=
1
2
(L
2
+ M
2
)
2
=
1
2
(L
2
+ M
2
) sin
2
+ N
2
cos
2
.
We identify N with the quantum number which gives the angular momentum about the axis, and
1
2
(L
2
+ M
2
) and
1
2
N
2
with the averages p
2

and p
2

of the total electronic angular momentum perpendicular and parallel to the axis;
since N is constant, p
2

= p
2
. We have nally:

n
= 0

= p cos

2
n
= p
2

2
= p
2

sin
2
+ p
2
cos
2
.
(93)
This result requires naturally a rigorous quantum mechanical verication; presumably p
2
is replaced by p(p + 1).
In the eigenvalue problem (92), the quantity
8
2
m
2
h
2

W
(4)
is equal to a numerical function of the rotational quantum
number r, say g
ns
(r); the rotational energy is thus:

4
W
(4)
nsr
=
h
2

4
8
2
m
2
g
ns
(r) =
h
2
8
2
J
g
ns
(r), (94)
where
J =
m

2
=
M
1
M
2
M
1
+ M
2

2
, (95)
5
Compare, for instance [9]
16
the moment of inertia of the nuclei at equilibrium.
A discussion of the higher approximation is meaningless unless we consider the degeneracies; we will not attempt
this here.
We will now show briey that one can treat the diatomic by a completely dierent perturbation procedure; the
classical analogue of this treatment was carried out by Born and H uckel [5]. The motion of the electronic system is
considered to be unperturbed not for stationary nuclei but rather for uniform rotation of the nuclei.
Part VIII. Independent Treatment of the Diatomic Molecule.
We go back to equation (12), and rewrite, substituting (11):
_
H
0
+
4
_
H

+ H

+ H

_
W
_
= 0.
Diatomic molecules have the peculiarity that H

is generally independent of the . In this case, the method enables

separation from the translations and rotations. From (81), dropping the translational terms:
_

_
H
0

h
2

8
2
m

4
_

2
2
+
2

+
1

_
W
_

_
= 0. (96)
We set
= Y
r
(, )
r
(x; ), (97)
where Y
r
is a spherical function of rth order which satises the equation:

Y
r
+ r(r + 1)Y
r
= 0;
thus we nd for
r
the condition
_

_
H
0

h
2

8
2
m

4
_

2
2
+
2

r(r + 1)

2
_

_
W
_

r
= 0. (98)
We again substitute + for ; thus considering vibrations about the state of uniform rotation. Denoting the energy
of this state as:
R =
h
2

2
8
2
m
r(r + 1)

2
=
h
2
8
2
J
r(r + 1) (99)
and setting
W = E + R, (100)
we nd for (98)
_
H
(0)
+ H
(1)
+
2
H
(2)
+ E
_

r
= 0 (101)
where
H
(0)
= H
(0)
0
H
(1)
= H
(1)
0
+ R

H
(2)
= H
(2)
0
+
1
2

2
R

h
2

8
2
m

2
(102)
H
(3)
= H
(3)
0
+
1
6

3
R

h
2

8
2
m
2

. . . ;
17
H
(0)
0
, H
(1)
0
, . . . are the operators given earlier. All the formulas of Part II are valid without modication. The approxi-
mation equations are:
a)
_
H
(0)
E
(0)
_

(0)
r
= 0
b)
_
H
(0)
E
(0)
_

(1)
r
=
_
E
(1)
H
(1)
_

(0)
r
c)
_
H
(0)
E
(0)
_

(2)
r
=
_
E
(2)
H
(2)
_

(0)
r
+
_
E
(1)
H
(1)
_

(1)
r
. . .
(103)
The rst has the solution:
E
(0)
= V
n
(),
(0)
r
=
(0)
rn
=
(0)
rn
()
(0)
n
(x; ), (104)
where V
n
() and
(0)
n
(x; ) are the previously introduced functions and
(0)
rn
() is, to begin with, arbitrary. The condition
for integrability of (103b) is
_
E
(1)
H
(1)
_

(0)
rn
() = 0.
Now, by (26a) (Part II):
H
(1)
nn
=
_
H
(1)
0
_
nn
+ R

= V
(1)
n
+ R

(V
n
+ R).
Hence, as before (Part IV),
E
(1)
= 0,

(V
n
+ R) = 0. (105)
This condition obviously states that for the unperturbed rotation, equilibrium must prevail between the centrifugal
force and the quasi-electric force, which, as a consequence of the electronic motion, resists a displacement of the
nuclei. The centrifugal force is:

_
1
M
1
+
1
M
2
_
p
2
r

3
=
_
1
M
1
+
1
M
2
_
h
2
4
2
r(r + 1)

3
,
where the quantum mechanical value
h
2

r(r + 1) for the angular momentum is substituted for p

r
; by (99) and (95),
this agrees with R

.
From relation (105), how to calculate the equilibrium separation
r
; depends on the rotational quantum number r.
For small values of the rotational energy R, one can develop
r
in powers of , where:
=
4

m
h
2
4
2
r(r + 1) =
_
1
M
1
+
1
M
2
_
h
2
4
2
r(r + 1); (106)
we nd:
6

r
= +
1

3
V

n

3

7
V

n
2
_
1 +

6
V

n
V

n
_

2
+ . . . (107)
Since is of order
4
, we will use by systematic procedure only as many terms of this set as correspond to the
order of the approximation in the perturbation method.
Since we consider this again, we will shortly see that this is the same method as before, only simplied by the
previous consideration of the rotation. The solution of (103b) is:

(1)
rn
=
(0)
rn

(1)
n
+
(1)
rn

(0)
n
(108)
this corresponds to (42); and the condition for integrability of (102c):
_
H
(2)
nn
+
_
H
(1)
nn
_
(1)
nn
E
(2)
n
_

(0)
rn
= 0.
6
One can easily deduce this formula from the cited work of Born and H uckel.
18
This is, however, the vibration equation
_

h
2

8
2
m

2
+
1
2

2
(V

n
+ R

) E
(2)
n
_

rn
= 0. (109)
Thus, as in Part VII:

2
E
(2)
rns
=
_
s +
1
2
_
h
r
, (110)
where the frequency,

r
=
1
4
_
_
1
M
1
+
1
M
2
_
(V

n
+ R

) (111)
still depends on the rotational quantum number r, from the R.
Further, as in Part VII,

(0)
rns
= exp(
2
/2)H
s
() (112)
with
= b
1/4
, b =
8
2
h
2
m
2
(V

n
+ R

).
The procedure may be continued in the usual fashion. We nd E
(3)
= 0, while E
(4)
, besides the deviation from
the harmonic vibration law, contains a coupling with the electronic motion. A thorough consideration of the formulae
would, however, be beyond the scope of this work, which demonstrates only the principle of the development; also
the calculation of the higher approximations is meaningful only when the degeneracies are taken into account.
[1] M. Born and W. Heisenberg Ann. d. Phys. 74 1 (1924)
[2] H. A. Kramers Zeitschr. f. Phys. 13 343 (1923); H. A. Kramers and W. Pauli jr. ibid. 13 351 (1923)
[3] J. Franck Trans. Faraday. Soc. (1925)
[4] E. Condon, Phys. Rev. 28 1182 (1926); Proc. Nat. Acad. 13 462 (1927)
[5] M. Born and E. H uckel Phys. Ztschr. 24 1 (1923)
[6] F. Hund Ztschr. f. Phys. 43 805 (1927)
[7] W. Pauli Ann. d. Phys. 68 177 (1922)
[8] S. E. Schr odinger Ann. d. Phys. 79 361, 3 (1926).
[9] F. Klein and A. Sommerfeld Theorie des Kreisels 1 p 108
19