Volume 126, number 2 CHEMICAL PHYSICS LETTERS 2 May 1986

THEORETICAL EVIDENCE FOR “BENT BONDS” IN THE CO2 MOLECULE

R.P. MESSMER
General Electnc Corporate Research and Development, Schenectady, NY 12301, USA
and Deparimenr of Physics, Unrversrty of Pennsylvanra, Phrladeiphia, PA 19104, USA

P.A. SCHULTZ, R.C. TATAR

Department of Physics, University of Pennsylvania, Philadelphia, PA 19104, USA

and

H.-J. FREUND ’

General Electric Corporaie Research and Development, Scheneeiady, NY 12301, USA

Received 27 November 1985; in final form 11 February 1986

Ab initio correlated wavefunction results for the CO, molecule are presented, which strongly suggest that the carbon-oxygen
double bonds in the molecule are better represented in terms of a pair of “bent bonds” than by a combination of a a-bond and
a n-bond.

1. Introduction N spin-orbitals, pi, from a complete set. In the treat-

ment of molecular problems it is common to consider
The physical interpretation of many-electron wave- only one term of the infinite expansion given in eq.
functions is an important, yet sometimes difficult and (l), which is the well-known “molecular orbital” ap-
often controversial, aspect of computational quantum proximation. Within this approximation, a great deal
mechanical studies. Thus, there is the frequent tempta- of effort has been expended on the interpretation of
tion to avoid the discussion of the wavefunction and the wavefunction. However, the outstanding success
restrict one’s attention to the values of computed ob- of a variety of resulting concepts (e.g., the orbital
servables. Although this course may avert some con- symmetry rules) in enhancing our understanding of
tentious discussion, it certainly also evades a signifl- molecules and their interaction attest to the fact that
cant responsibility of the theorist. the pursuit of an interpretation for the wavefunction
A general Nelectron wavefunction can always be is not a futile exercise. Fortunately for molecular or-
written in the form of a linear combination of deter- bital theory the general character of the canonical
minantal functions, molecular orbitals,$, arrived at by rather diverse com-
putational techniques (from ab initio to semiempiri-
\k= cc&! )
0) cal methods) is very much the same. This has allowed
i
a general consensus to form regarding the molecular
where each determinant is made up of a collection of orbital interpretation of manyelectron wavefunc-
’ Permanent address: Institut fur Physikalische und Theore-
tions *
tische Chemie der Friedrich-Alexander-Universitat Erlangen-
Niirnberg, D-8520 Erlangen, Federal Republic of Germany. * For footnote see next page.

176 0 009-2614/86/$ 03.50 0 Elsevier Science Publishers B.V.

(North-Holland Physics Publishing Division)
Volume 126, number 2 CHEMICAL PHYSICS LETTERS 2 May 1986

By contrast, when one improves the wavefunction Table 1

by taking more than one term in eq. (I), one is faced Results for the COs molecule
with both an embarrassing number of possible options Method Total energy (hartree)
on how to proceed and the concomitant choice of pos-
sible ~terpretations. The routine computations treat- Hartree-Fock -187.674466
ment of multi~eterm~~t wavefunct~ons is rather re- GVB PP a) -187.765041
cent as compared to the routine computation of mo- GVB PP b) -187.776321
lecular orbitals. These two facts indicate why it should a) With c,sorbital symmetry restrictions.
not be surprising that there is, as yet, no consensus re- b, With no restrictions.
garding the interpretation of multi-determinant wave-
functions. A few criteria which might be applied in the
choice of an interpretive approach could include the the above aspects are identical. The calculations differ
following: (1) the simplicity of the interpretation; only in the symmetry restrictions imposed upon the
(2) the coincidence of the interpretive approach with orbitals $I?~. The canonical molecular orbitals, @*,
familiar physical and chemical concepts derived from are symm\try adapted and for a linear molecule (skch
experiment; and (3) the economy of the approach in as CO,) are of 0, n, 6, ... character. These symmetry
devising compact and accurate approbations to orbitals are also frequently used in GVB c~culations
wavefunctions. Two examples of such interpretive f24J . However, such symmetry restrictions in the
schemes are the valence-bond (VB) approach [12-l 61 GVB method are not necessary. Some time ago Hay
and the atoms-in-molecules (AIM) approach [ 17-211. et al. [25] tested the effects of removing symmetry
In the present work we adopt the valence-bond inter- restrictions for ethylene and acetylene, and found the
pretive framework. variational ground state to be described by a;R-bonds
rather than bent bonds. Our results are in agreement
with their conclusion although the two descriptions
2. Computational methods for acetylene are nearly degenerate. In the case of
CO,, however, when the symmetry restrictions are re-
The calculations we describe have been carried out moved, a bent bond description is obtained which
within the generalized v~ence-bond perfect pairing yields a lower energy for the molecule,
(GVB PP) method 1221 (see also refs. [ 15 ,I 61). We Table 1 contains the computed total energies for the
use a valence double-zeta plus polarization basis set ground state of CO2 at the experimental equilibrium
[23] and correlate all of the valence electron pairs. geometry obtained from the following calculations:
Thus there are 256 determinants in the expansion of molecular orbital (Hartree-Fock), GVB PP with o-n
eq. (1) and there are sixteen $pvB describing the va- orbital symmetry restrictions and GVB PP without or-
lence electrons of CO,, with each deter~~t in the bital symmet~ restrictions.
expansion containing eight doubly occupied #Fvt,.
For comparison we also present the results of a
Hartree-Fock molecular orbital calculation using the 3. Results and discussion
same basis set. In this calculation there is only one de-
terminant with a set of eight doubly occupied $y” In fig. 1, we show that two nY molecular orbitals
describing the valence electrons. responsible for the n-bond and the oxygen lone pair in
In both GVB PP calculations we describe below, the yz plane according to the molecular orbital inter-
pretation (each orbital is occupied by two electrons).
We note that the chemist’s usual picture of a CO A-
* There has been considerable discussion regarding equivalent bond and an 0 lone pair is not entirely apparent from
orbit& and localized orbit&s ] 1 -I I] within the molecular
these orbitals. The GVB PP results for CO, in which
orbital framework. However, the fact that they are all re-
lated by a unitary transformation which leaves the energy
the orbitals are constrained to have o and R symmetry
invariant, has not allowed a consensus to evolve on the give the n-like PP orbitals (in the yz plane) shown in
most appropriate local viewpoint. fig. 2. Although the u-like PP orbitals are found to be

177
Volume 126, number 2 CHEMICAL PHYSICS LETTERS 2 May 1986

a b

Fig, I. Tke two sy molecular orbitals of CO2 in the yz plane. Each orbital contains two electrons. There is an equivalent set in the

highly localized (not shown) into two CO bond pairs bitals is removed, the orbitals change dramatically, as
and two oxygen lone pairs, the n-like orbitals in fig. 2 shown in fig. 3. Each orbital contains one electron and
are not easily identifiable as an oxygen lone pair and a the eight PP orbitals in the yz plane are displayed. This
C-O a-bond. description of the manyelectron wavefunction for
If the u, rr symmetry restriction on the GVB PP or- CO2 has an energy which is 0.3 eV lower than the
GVB description in terms of u and n orbitals (fig. 2).
In panels A and B of fig. 3, one can observe the “bent”
bonds which constitute one of the two equivalent sets
of CO double bonds in CO,. Panel A shows an oxygen
hybrid orbital on the left and a carbon hybrid orbital
on the right. These two orbitals overlap to form one
bond component of a symmetric pair which yield a
C-O double bond. The other component is shown in
panel B. The bond formed by the overlapping of the
two hybrids (panel A or panel B) has a shape reminis-
cent of the greek letter R, hence we refer to them as
Q-bonds when they occur in pairs or triplets in the de-
scription of double or triple bonds.
Panels C and D in fig. 3 show the oxygen lone pairs
in the yz plane, Now, contrary to the results of molec-
ular orbitai theory and GVB PP with u, n constraints
on the orbitals, one obtains a set of orbitals clearly
identifiable as lone pairs. In panel C are shown two PP
orbitals which constitute one of the lone pairs. Again,
there is one electron in each orbital and the “m-out”
correlation in the pair is clearly seen, with the orbital
to the right in the panel being closer to the oxygen nu-
Fig, 2. The four GVB PP orbitals (in the yz plane) of ny sym-
cleus and the orbital to the left in the panel being far-
metry from a calculation with a,n-symmetry constraints. Each
orbital contains one electron and the two orbit& in each pan- ther from the nucleus. Panel D shows the other equiv-
el {A and B) are singletcoupled into an electron pair. There is alent lone pair in the yz plane.
an equivalent set of orbit& in the xz plane. A schematic repre~ntation of the manyelectron

178
Volume 126, number 2 CHEMICAL PHYSICS LETTERS 2 May 1986

Fig. 3. The eight GVB PP orbitals in the yz plane of COa from a calculation with no symmetry constraints on the orbit&. Each or-
bital contains one electron and the two orbitals in each panel (A, B, C, D) are shrgletcoupled into an electron pair. Panels A and B
are two equivalent “bent bonds” forming a CO double bond. Panels C and D are two equivalent lone pairs on the oxygen atom at
the right. There is an equivalent set of eight PP orbitals in the xz plane.

wavefunction in terms of the computed GVB PP or-

bitals which give the lowest variationally determined
total energy is shown in fig. 4. We note that this is one
of two equivalent “resonance” structures which are
needed to describe the correct symmetry of the overall
wavefunction. When this resonance effect is taken into
(a)
account via a resonating GVB calculation [26] , it is
found that the energy of the S2-bond description is

lo=C=CJ
even lower relative to that of the corresponding u--n
description. It is interesting to observe that many fea-
tures of the description we find here from ab initio
generalized valence bond calculations are similar to a -=
IO c =a
qualitative discussion of bonding in CO, given by
Pauling [ 141 over twenty-five years ago. The ascen-
(bl
dancy of molecular orbital theory as the paradigm for Fig. 4. (a) Schematic representation of the manyelectron
describing molecular electronic structure in the inter- GVB PP wavefunction. The letters A-D denote the electron
vening years has tended to obscure the fact that the pairs whose orbitals are shown in the corresponding panels of
valence bond language provides a simple framework fg. 3. This is one of two equivalent ‘kesonance” structures;
the other is obtained by rotating 90“ about the internuclear
for describing important electronic correlation effects
axis. (b) Usual chemical representation of the two resonance
which are ignored in a molecular orbital treatment. structures.

179
Volume 126, number 2 CHEMICAL PHYSICS LETTERS 2 May 1986

4. Summary [S] J-E. LennardJones, Proc. Roy. Sot. Al98 (1949) 1.

[6] G.G. Hall and J.E. LennardJones, hoc. Roy, Sot. A202
(1950) 155.
We have presented the results of ab initio calcula-
[7] J.E. Lennard-Jones and JA. Pople, Proc. Roy. Sot.
tions which show that it is energetically f;iorable to A202 (1950) 166.
form “bent bonds” in the CO, molecule . To our [81 J.M. Foster and S.F. Boys, Rev. Mod. Phys. 35 (1960)
knowledge, there has been no published discussion of 300.
such a situation for multiple bonds in any molecule. tg1 C. Edmiston and K. Ruedenberg, Rev. Mod. Phys. 35
(1960) 457.
Our results are in many ways similar to a qualitative dis-
IlO1 M.D. Newton, E. Switkes and W.N. Lipscomb, J. Chem.
cussion presented by Pauling many years ago. A more Phys. 53 (1970) 2645.
detailed discussion of these aspects of CO2 and of oth- 1111M.D. Newton, in: Modern theoretical chemistry, Vol. 4.
er molecules for which we have found a-bonds (bent Applications in electronic structure theory,
multiple bonds) will be presented elsewhere. ed. H.F. Schaefer III (Plenum Press, New York, 1977)
p. 223.
WI W. Heitler and F. London, Z. Physik 44 (1927) 455.
1131 A.C. Hurley, J-E. Lennard-Jones and J.A. Pople, Proc.
Acknowledgement Roy. Sot. A220 (1953) 446.
iI41 L. Pauling, The nature of the chemical bond, 3rd Ed.
This work was supported in part by the National (Cornell Univ. Press, Ithaca, 1960).
Science Foundation MRL Program under Grant No. [W W.J. Hunt, P.J. Hay and W.A. Goddard III, J. Chem.
Phys. 57 (1972) 738.
DMR-82.I 67 18 at the Laboratory for Research on the 1161 F.W. Bobrowicz and WA. Goddard III, in: Modem
Structure of Matter, University of Pennsylv~a. HJF theoretical chemistry, Vol. 3. Methods of electronic
would like to express his gratitude for the financial structure theory, ed. H.F. Schaefer III (Plenum Press,
support and hospitality accorded him by the General New York, 1977) p. 79.
Electric Research and Development Center during his [I71 W. Moffitt, Proc. Roy. Sot. A210 (1951) 224.
[I81 W. Moffitt, Rept. Progr. Phys. 17 (1954) 173.
tenure as a Visiting Research Fellow. I191 A.C. HurIey,Proc. Phys. Sot. (London) A68 (1955) 149.
WI A.C. Hurley, Rev. Mod. Phys. 35 (1963) 448.
** Although this conclusion is based on resonating GVB PP WI B. Lam, M.W. Schmidt and K. Ruedenberg, J. Phys.
calculations presented herein, it is also supported by CI cal- Chem. 89 (1985) 2221.
culations with resonance. Configuration interaction calcula- PI RA. Bair, W.A. Goddard III, A.F. Voter, A.K. Rappe,
tions designed to ease the most severe restrictions of the LG. Yaffee, F.W. Bobrowicz, W.R. Wadt, P.J. Hay and
GVB PP wavefunction (by including alternate spin couplings, W.J. Hunt, GVB2P5 Program (unpublished);
relaxing strong orthogonality between pairs and incorporat. R.A. Bair, Ph.D. Thesis, Caltech (1980).
hrg re~n~ce) have been carried out for both the u,n- and 1231 T-H, Dunning Jr. and PJ. Hay, in: Modem theoretical
n-bond wavefunction descriptions. The R-bond description chemistry, Vol. 3. Methods of electronic structure
is unambiguously favored. theory, ed. H.F. Schaefer III (Plenum Press, New York,
1977) p. 1.
u41 BJ. Moss, F.W. Bobrowicz and W.A. Goddard III, J.
References Chem. Phys. 63 (1975) 4632.
[25] P.J. Hay, W.J. Hunt and W.A. Goddard III, J. Am.
[l] L. PauIing, J. Am. Chem. Sot. 53 (1931) 1367. Chem. Sot. 94 (1972) 8293.
121J.C. SIater, Phys. Rev. 37 (1931) 481. 1261 A.F. Voter and W.A. Goddard III, Chem. Phys. 57
[3] F. Hund, 2. Physik 73 (1931) 1. (1981) 253.
[4] C.A. Coulson, Trans. Faraday Sot. 38 (1942) 433.

180