INTERNATIONAL JOURNAL OF QUANTUM CHEMISTRY, VOL.

VI, 271-279 (1972)

The Non-Crossing Rule in Molecular
Quantum Mechanics
K. RAZI NAQVI *
Atomic and Molecular Physics Group, The Schuster Laboratory, University of Manchester,
Manchester M I 3 9PL, England
AND
W. BYERS BROWN
Department of Chemistry, University of Manchester, Manchester MI3 9PL, England
Abstracts
Previous mathematical proofs of the non-crossing rule concerning the potential energy
curves of a diatomic molecule are examined and criticised. A new treatment is presented
which is simple and mathematically rigorous.
Un traitement 8 la fois simple et avec rigueur mathkmatique est donnC pour la non-
crossing rule concernant les courbes de potentiel dans les molkcules diatomiques. Un
examen minutieux et critique des preuves que jusqu8 present ont ttk donneks de cette
rtgle prtctde notre deduction et en justifie sa nkcessite.
Hier werden bisherige mathematische Beweise der non-crossing Regel in Verbindung
Es mit Potentialenergiekurven eines zwei-atomigen Molekuls untersucht und kritisiert.
wird eine neue Behandlung vorgeschlagen, die einfach und mathematisch rigoros ist.
The non-crossing rule, which is continually invoked in constructing the
correlation diagrams for the electronic states of diatomic molecules, asserts that
potential energy curves of the electronic states of the same symmetry species cannot
cross as the internuclear distance is varied [I ]. The rule is of immense importance
in quantum mechanics and applies, in fact, to any system whose energy depends
on a variable parameter which occurs in the Hamiltonian of the system. There is
massive experimental evidence in spectroscopic literature for the validity of the
non-crossing rule; the doubt which some earlier experiments cast on its ubiquity
has now been dispelled, and in no instance does the rule seem to be violated [2].
I t is our intention to point out in this paper that, of the two attempts that have
been made at proving this indispensable rule, one is based on a non-sequitur, while
the other incurs a defect by considering only two wave functions in a variational
discussion; and to present a new derivation which seems to lack these deficiencies.
The most frequently cited proof, due originally to Teller [3] and quoted with
or without some minor alterations by a number ofother authors [4-91, is reproduced
below for the sake of completeness. We assume that we already know all the correct
* Rutherford Scholar of the Royal Society (1968-71). Present address: Department of Chem-
istry, The University, Sheffield S3 7HF, England.
21 1
0 1972 by J ohn Wiley & Sons, Inc.
272 K. RAZI NAQVI AND W. BYERS BROWN
eigenfunctions of the molecule except those corresponding to the two potential
curves whose intersection we wish to investigate. Further, let us have two wave
functions d1 and 4, which are orthogonal to all the known wave functions and to
each other; the two unknown wave functions may be written as linear combina-
tions of d1 and $2. Let H( R) be the electronic Hamiltonian of the molecule, where
R denotes the internuclear distance. The energy levels can be determined by
solving the secular equation
where
Thus
Hij = ($*I = H;
4 . 2 = Wll + H22) f Wll - H22)2 + 4 IH121211'2
The two values of E will be equal only if the two conditions
and
Hll = H22
H12 = 0
are satisfied simultaneously. If at some value of R ( = Roy say) Hll(Ro) = H,,(R,),
it will be a mere accident if Hl2(R,,) is also zero. Thus it is almost impossible that
by varying one parameter, R, we will achieve the condition El = E,; hence the
non-crossing of the two curves. However, if dl and d2 belong to two different
irreducible representations of the symmetry group of H, H12 will be identically
zero (at all R) and crossing is possible.
We will now demonstrate that the foregoing argument proves the non-crossing
rule only by begging the question at issue, and that it is logically deficient. As this
fallacy seems to have been overlooked so far, a detailed discussion may not be
amiss.
and 4, belong to the same function space as that spanned by the two
unknown eigenfunctions y1 and y,, we can relate the two pairs of functions by the
following orthogonal transformation matrices :
Since
[;;I = [ cos sin 7 [;;I
-sin 8 cos 8
where 8 is some angle which will depend on R; conversely,
cos 8 -sin 8
[ : j = [sin e cos J [j
NON-CROSSING RULE IN QUANTUM MECHANICS
273
We now calculate the relevant matrix elements :
H,, = $(El - E,) sin 28
H,, = El cos2 8 + E, sin2 8
H,, = El sin2 8 + E, cos2 8
whence
H,, - H2, = (El - E2) cos 28
I t can now easily be seen that when HI , = H,, , H,, must either be zero or
&(El - E,). But the former possibility can be ruled out only if we know that
El #E,; thus, the argument that the two conditions
Hll = H,, and H,, = 0
will not be satisfied simultaneously, does not follow unless it is shown that El
cannot equal E, , which is to be proved.
As 4, and 4, must vary with R, they cannot be constructed without a knowledge
of the eigenfunctions yl,, and the eigenvalues El, , , and it seems extremely artificial
that one does not start with the correct wave functions themselves. Further, since
any discussion of the eigenvalues is inherently restricted to the particular value of
R for which 4, and 4, have been defined, it is impossible to examine the intersection
of two potential curves. It will be seen later that only by considering two suitably
chosen neighbouring points is it possible to derive two independent conditions and
prove the theorem stated in the non-crossing rule.
The earliest discussion due to von Neumann and Wigner [lo] was based on a
somewhat different proof, which has also been given in slightly modified forms by
Landau and Lifshitz [ 1 11 and by Heine [ 121. It proceeds as follows. If two energy
curves El (R) and E,(R) are to intersect at some point, they must have values which
are very nearly equal in the vicinity of the crossing point. Let us consider a point
R(O) where the electronic Hamiltonian is H(0) and the two functions El and E,
have values Eio and EP) which are very close but not equal to each other. Let
y? and y$ be the corresponding wave functions. We now change the inter-
nuclear distance to R(O) + 6R and see whether El and E, become equal. Increasing
the internuclear distance by an amount 6R changes the Hamiltonian to H(O) + V,
where V, given by
6 R + * * . V = H(RO + 6R H(Ro + 6R)- H(O) = -
aH0
aR
is a small correction term. E,(R(O) + 6R) and E,(R(O) + 6R) can be regarded as
eigenvalues of the new Hamiltonian; V, as a perturbation to the operator HO).
2 74 K. RAZI NAQVI AND W. BYERS BROWN
We look in the zeroth approximation for a solution of the form *
(1) y = ayy + by?
and evaluate E,(R(O) + 6R) and E2(R(0) +
I t is easy to arrive at the secular equation
EP) + V,, - E
E p
6R) by applying perturbation theory.
v12 L o
+ v22 - El
We see then that the conditions for crossing to occur are:
(2)
Eio) - ELo + V,, - V2, = 0
(3) v,, = 0
I t should be noted that these conditions, unlike those derived by Teller, are not
interdependent, and the lefthand sides of Equations (2) and (3) are not explicit
functions of (El - E2) . The rest of the argument is identical with the concluding
part of Tellers proof and will not be repeated.
At first sight it might seem that writing the zero-order wave function at
(R(O) + SR) as a linear combination of yio) and y r ) , rather than as a linear com-
bination of the complete set of functions is a legitimate use of perturbation
theory. However, the contribution due to the states not included in the truncated
expansion of Equation (1) will vanish in the limit 6R --+ 0 only if EP) = EP) at
R = RO). But then the two curves will intersect by definition and the degeneracy
will persist as V will also vanish at R(0) and there will no longer be a perturbation.
The wave function of Equation (1) is admissible if the molecule has only two
potential curves, for y? and yp) will then form a complete set by themselves;
or if the operator H(0) has two close eigenvalues EY) and EP and all other eigen-
values lie far from these two. However, in the absence of any knowledge about the
other eigenstates of the molecule, employment of Equation (1) is difficult to accept,
and a proof which hinges on the veracity of Equation ( 1 ) loses its rigour when
applied to a diatomic molecule with a plethora of energy curves many of which
may have close values at a particular internuclear distance.
I t is often remarked [ 1, 1 1 , 131 that if, in a certain approximation, two energy
curves belonging to the same symmetry species intersect, then taking into account
the small terms in the Hamiltonian that were originally ignored would obviate
the seeming crossing and the curves would repel each other. The inclusion of
the previously neglected terms in the Hamiltonian is tantamount to the introduc-
tion of a perturbation in the system and would, of course, lift the degeneracy
which was initially present. But this argument, though correct in its logical form
* Von Neumann and Wigner start by expressing the wave function in terms of a complete
set, but drop, in the later stages, all other functions except the two shown in Equation ( 1 ) .
NON-CROSSING RULE IN QUANTUM MECHANICS 275
and physical contents, is not altogether conclusive; what, one may ask, would
happen if one knew the exact eigenfunctions and a Hamiltonian which takes
account of all the terms?
We now proceed to derive an alternative proof of the non-crossing rule by
considering the exact eigenfunctions and the complete Hamiltonian for a diatomic
molecule. Consider the crossing of two energy eigenvalues El(R) and E,(R) at
R = R, ; so that
(4) El@,) = EZ(R0) = Eo
Note that this condition does not by itself guarantee crossing, as the states may be
degenerate for all values of R. The crossing of two eigenvalues as functions of R
means that for R = R, + 6R the energies
and E2(R, + 6R) = E, + SE, E1(Ro + 6R) = E, + 6El
are dzyerent, so that
(5) 6E1 #6E2
This inequality will now be used to derive an additional requirement for crossing.
Let the Schrodinger equation for the first state be
( H - E l h = 0
Consider a small change from R, to R, + 6R and let the perturbed Schrodinger
equation be
( H, - E,) 6w1 + (6H - 6El)y: = 0
Should one choose to argue that, in order to obtain the above equation, we have
immolated mathematical rigour by discarding the terms which are quadratic in 6,
i t will be enough to mention that a completely rigorous treatment, presented in the
Appendix, vindicates the analysis which follows.
Multiply the complex conjugate of the last equation by the second eigenfunc-
tion y: for R = R, and integrate to get
( 6) (w3 6H I w3 = W ( w : I 34)
By repeating the above analysis with the Schrodinger equation for the second state
and multiplying by the complex conjugate of yy it follows that
276 K. RAZI NAQVI AND W. BYERS BROWN
and, knowing that 6El #6E2, we deduce
(8) (Y? 1 Y 3 = 0
I t will be well to pause here momentarily and discuss the implications of
Equation (8). Our demand that the two potential curves intersect at R, forces us
to conclude that the overlap integral must vanish even when El = E2 . I t is
tempting to argue that, since (yl I y 2) = 0 for all R in the vicinity of R = R, ,
it seems likely, on account of continuity, that it would also be true at R, . However,
this argument is not only unnecessary but misleading, for we know that degenerate
eigenfunctions need not be orthogonal; indeed we can choose them at will and
make them to be non-orthogonal, if we so desire. The only exception occurs when
the two functions belong to different irreducible representations of the symmetry
group of H, in which case (yl I y 2) will be identically zero at all R because of the
different symmetries of the two functions, and there is no way of making the overlap
integral different from zero at any R.
Returning to Equation ( 6) , it can now be seen that the inequality (5) implies
the further condition :
(9) (w?l 6f f I W3 = 0
Let us suppose that SH can be expanded as a power series in 6R,
6H = -F6R + * * *
Then, retaining only the first-order term for simplicity, Equation (6) requires
that F,,(R,) = 0, where
Fl2 = (Wl l F Iyz) = -(.lull aH/aR Iy2)
Thus for crossing, two conditions * are required :
E,(R,) = E,(Ru) (degeneracy)
(crossing)
(10)
F12(R,) = 0
The usual argument now applies: there is no reason why these should be satisfied
simultaneously,t unless one of them vanishes automatically for all R; that is,
either the degenerate case E,(R) = E,(R), orF12(R) = 0 for all R due to symmetry
( F =: - aH/aR will have the symmetry of the Hamiltonian).
We emphasize, in conclusion, that Tellers simplified formulation fails to
arrive at the essential condition for crossing because it gives no special considera-
tion to a value of R distinct from the crossing point, R, . This is the underlying
* If the eigenfunctions y 1 and ys cannot be chosen to be real (which is unlikely), then the
t It is, however, not strictly impossible that both are fulfilled for just one value of R (accidental
equation F12(Ro) = 0 is equivalent to two real conditions, and crossing is even less likely.
crossing).
NON-CROSSING RULE I N QUANTUM MECHANICS
277
difference between Tellers approach and that adopted by von Neumann and
Wigner and by us. Our treatment seems more satisfactory than the pioneering
discussion due to von Neumann and Wigner, for the analysis is obviously rigorous
and does not require the use of an approximate form of degenerate perturbation
theory.
Acknowledgements
Some useful discussions with Dr. J . G. Baker and Dr. I. L. Cooper are gratefully
acknowledged.
Appendix
A. Further discussion of crossing
Two curvesf(x) and g( x) which touch each other at a point for which x = a
are said, in the parlance of differential calculus, to cross at the point of contact,
if the differencef(a + h) - g( a + h) has opposite signs for small positive values
of h and for numerically small negative values of h[ 141. We will reserve, in the Appen-
dix, the terms crossing and intersection for the behaviour specified in the
preceding sentence, and use the more general words contact and touch to
describe what we have been indiscriminately calling crossing of curves. I n other
words, we will now discuss the no-contact rule and explore the conditions under
which energy levels of like symmetry can touch each other when the parameter R
is allowed to vary.
Consider two curves E,(R) and E,(R) which have a contact at R = R,, i.e.
(4) El(R0) = EdR,) = Eo
Now, for a contact of nth order to occur at R = R, , we must also have
(1 1)
and
Ei(R,) = &(R0), E;(R,) = El (Ro), * * * , Ep)(Ro) = Ep) ( Ro)
(12) ( R,) # (R,)
where the superscripts denote differentiation with respect to R.
We begin by writing the Schrodinger equation for the first state
( H - E,)W1= 0
(Ho - EO)Wi(R,) + P ( R0) - E;(Ro)M = 0
Differentiation with respect to R, at the point R = R, , yields
If we multiply the complex conjugate of this equation by yi and integrate, we
obtain
278 K. RAZI NAQVI AND W. BYERS BROWN
Differentiation of the Schrodinger equation for the second state, and multiplication
by the complex conjugate of v; gives
(14) (YYl H(R0) l w3 = E; ( d I w 3
P; ( Ro) - E;(Ro))(Y! I Y 3 = 0
Subtraction of Equation (1 4) from Equation (1 3) yields
(15)
If the contact is of zero order, so that E,(R,) = E,(R,) but E;(R,) #E;(R,),
we must have, as we found earlier in the text
(8) s = (y! I y;) = 0
If E,(R,) = E,(R,), and also Ei(R,) = Ei(R,), but E;(R,) #E;(R,), i.e. if the
contact is of first order, we differentiate Equation (15) with respect to R and find
S{E;(R,) - EWO) } + s(R, ){q(R, ) - 4&)} = 0
which again leads to Equation (8) of the text.
I n general, if the contact is of order n, we differentiate Equation (15) n times
and, remembering the set of Equations (1 l ), arrive, in the last step, at Equation
(8). The curves intersect, or do not intersect, at the point of contact according as
n is even or odd.
Substitution of Equation (8) into Equation (13) gives us the equation
which is identical with the second of the pair of Equations (10) of the text.
We have thus established that the requirements which must be met, if two energy
curves are to have a contact at R, , are the same as those expressed by Equations
(10) which were derived earlier in the text.
Bibliography
[I ] G. Herzberg, Molecular Spectra and Molecular Structure, Vol. 1 (D. Van Nostrand, New York,
[2] A. G. Gaydon, Dissociation Energies and Sfectra of Diatomic Molecules 3rd edn. (Chapman and
[3] E. Teller, J . Phys. Chem. 41, 109 (1937).
[4] H. Eyring, J . Walter and G. E. Kimball, Quantum Chemistry (Wiley, New York, 1963), p. 206.
L5] C. A. Coulson and J . T. Lewis, Chemical binding, in Quantum Theory, Vol. 2, Ed. D. R.
[6] C. A. Coulson, Valence (Oxford University Press, London, 1952) pp. 65-67.
[7] M. Kotani, K. Ohno and K. Kayama, Quantum Mechanics of Electronic Structure of
Simple Molecules, in Handbuch der Physik, Vol. 3712, Ed. S. Flugge (Springer-Verlag Berlin,
1961), pp. 63-64.
1950), p. 295.
Hall, London, 1968), pp. 60-68.
Bates, (Academic Press, New York, 1962), p. 204.
NON-CROSSING RULE IN QUANTUM MECHANICS 279
[8] B. Di Bartolo, Optical Interactions in Solids (Wiley, New York, 1968), pp. 163-164.
[9] J . N. Murrell, S. F. A. Kettle and J . M. Tedder, Valence Theory (Wiley, London, 1965) p. 154.
[lo] J . von Neumann and E. Wigner, Z . Physik, 30 467 (1929).
[l l ] L. D. Landau and E. M. Lifshitz, Quantum Mechanics (Pergamod Press, London, 1958),
[I21 V. Heine, Group Theory in Quantum Mechanics (Pergamon Press, London, 1960), pp. 223-224.
[13] J . G. Valatin, Proc. Phys. SOC. 58, 695 (1946).
1141 R. Courant and F. J ohn, Introduction to Calculus and Analysis, Vol. 1 (Wiley-Interscience, hTew
pp. 262-265.
York, 1965), pp. 458-461.
Received August 25, 1970.
Revised March 31, 1971.