Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Journal of Molecular Structure: THEOCHEM 861 (2008) 18–26

Contents lists available at ScienceDirect

Journal of Molecular Structure: THEOCHEM


journal homepage: www.elsevier.com/locate/theochem

Electronic structure of Li3


Fabio E. Penotti *
Consiglio Nazionale delle Ricerche, Istituto di Scienze e Tecnologie Molecolari, Via Golgi 19, I-20133 Milano, MI, Italy

a r t i c l e i n f o a b s t r a c t

Article history: Non-orthogonal single- and multi-configuration ab initio calculations have been carried out on ground-
Received 26 February 2008 state Li3 in its minimum-energy C2v geometry. Their results have been compared with published work
Received in revised form 1 April 2008 and with those of SCF, frozen-core SDCI and full-valence CI calculations.
Accepted 2 April 2008
The calculations confirm the feasibility of explicit basis-set optimisation, by second-order analytical
Available online 10 April 2008
methods, for correlated wavefunctions in non-linear molecules, and compare use of optimised STO basis
sets with standard, high-accuracy, GTO basis sets.
Keywords:
The results are used as basis for a discussion of the molecule’s electronic structure.
Electron density
Electron density difference
The molecule’s electron density is shown to exhibit a non-nuclear maximum, both at the SCF and frozen-
Non-nuclear attractors core full-CI levels, and with core-correlated non-orthogonal wavefunctions.
Non-orthogonal multiconfiguration An ‘Atoms-in-Molecules’ topological analysis of the electron density shows features that may be viewed
methods as related to the occurrence of Interstitial Orbitals in non-orthogonal electronic wavefunctions for this
Basis-set optimization system.
The electron density difference map (molecule minus atoms) exhibits a non-nuclear maximum at roughly
the same location as the molecule’s total electron density, plus three 3p-type enrichment-depletion pat-
terns, one centred on each nucleus. Corresponding patterns are found in Li2.
Li3’s electric dipole and electric field gradient at the nuclei are also computed. Unexpectedly, the electric
dipole value is found to exhibit significant dependence on the inclusion of inner-shell out-of-plane cor-
relation in the wavefunction.
Ó 2008 Elsevier B.V. All rights reserved.

1. Introduction which the three nuclei are at the corners of ‘obtuse’ isosceles trian-
gles, separated by interconversion barriers culminating in saddle
In its equilateral-triangle (D3h) nuclear geometry, the Li3 mole- points that correspond to ‘acute’ isosceles triangles [1,2]. (In this
cule has a degenerate electronic ground state (X 2E0 ). This is a point context, an isosceles triangle is usually said to be ‘obtuse’ if its un-
of conical intersection, and as the symmetry is lowered to C2v, the ique angle exceeds the equilateral value of 60°, ‘acute’ otherwise).
degenerate state’s two components split in energy, giving rise to The X state interconversion barriers are low enough to give rise
what are usually described as the two sheets of the X 2E0 adiabatic to the dynamic Jahn–Teller effect, making Li3 a ‘floppy’ molecule.
potential energy surface. A descent-in-symmetry table immedi- Its rovibrational motion cannot be properly characterized in terms
ately shows that the two sheets must correspond to 2B2 and 2A1 of a C2v asymmetric rotor and its normal modes of vibration [2]. In
electronic states. The B2 sheet turns out to be the one that goes fact, by numerical solution, in hyperspherical coordinates, of the
down in energy as one moves away from D3h geometry, while Schrödinger equation for nuclear motion with the exact rotation–
the A1 sheet goes up. The two sheets extend of course to cover Cs vibration Hamiltonian, it has been shown that the height of the
geometries as well, where the two states are both A0 , but obviously interconversion barriers above the minima is less than twice the
cannot split any further as there is no residual electronic degener- zero-point rovibrational energy [2].
acy, other than that from the non-zero value of the spin. On the other hand, the X state’s lowest conical intersection, i.e. the
The adiabatic potential surfaces for the X 2E0 state and the A2E00 D3h geometry of lowest energy, lies at a height, above the state’s low-
state, which also exhibits a conical intersection, have been er sheet’s C2v minima, equal to more than twelve times the zero-
accurately characterized in relatively recent work which joined point rovibrational energy, while the vertical excitation energy in
high-resolution spectroscopy with high-accuracy theoretical calcu- the region of the minima is more than 40 times that same zero-point
lations [1–3]. This has shown that, for both the X 2E0 and the A2E00 energy [2]. This implies that, in the system’s lowest few rovibrational
states, the lower sheet exhibits three equivalent C2v minima, in states, the D3h geometry, and the upper sheet, are essentially inac-
cessible. Therefore, the adiabatic approximation can be safely ap-
plied to obtain low-lying rovibrational states, provided only the
* Tel.: +39 2 50314294; fax: +39 2 700507208.
E-mail address: f.penotti@istm.cnr.it ‘geometric phase’ (Berry phase) factor [4] is properly dealt with [1,2].

0166-1280/$ - see front matter Ó 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.theochem.2008.04.003
F.E. Penotti / Journal of Molecular Structure: THEOCHEM 861 (2008) 18–26 19

All this leads to a simple conclusion that is of direct significance requirements, though selective such constraints may be imposed
to the present work: the relative inaccessibility of the D3h geome- on arbitrarily chosen configurations [21,23].
try, and of the whole 2A1 sheet, imply that, with a few caveats that In the ‘‘Optimised Basis Set” variant of the GMCSC approach
will be raised in this paper’s final section, it makes sense to study (OBS–GMCSC) [20,21], basis function exponential parameters are
Li3’s electronic structure and density at (any) one of its three equiv- also treated as adjustable parameters, to be determined by energy
alent C2v 2B2 minimum-energy geometries, as representative of the minimization simultaneously with all other variational parameters
molecule’s actual properties, despite the presence of a conical in the wavefunction, i.e. configuration, spin-coupling, and orbital
intersection right in their midst. coefficients. Aside from often significant variational improvement,
To this end, the present work will take as its starting point a study this ensures rigorous compliance with the virial theorem; of
[5] of the orbital structure of the single-configuration Spin-Coupled course, optimisation of a single scale factor, common to all expo-
[6–10] electronic wavefunction for this system (among other Li clus- nential parameters, would suffice for the latter purpose.
ters), and build on its results by considering a series of electronic Energy minimization is carried out by use of the so-called Sta-
wavefunctions that include an increasing number of variationally- bilized Newton–Raphson method, originally introduced in eco-
optimised non-orthogonal orbital configurations, in the process nomic theory as ‘‘quadratic hill climbing” [24], a maximization
bringing to the forefront Interstitial Orbitals that enter the single- method, and long used as the workhorse of the single-configura-
configuration SC wavefunction, in part, in a somewhat disguised tion SC approach [7]. This second-order iterative minimization
manner [6–10]. Although this will be a series of nine-electron calcu- method requires the availability of the first and second derivatives
lations, leading up to core-correlated wavefunctions, an effort to as- of the energy with respect to all variational parameters, cross
sess the accuracy of numerical results will be made by comparing terms included. These are computed analytically at each iteration.
those of a shorter series of corresponding frozen-core non-orthogo- Convergence on an energy minimum, as appropriate for electronic
nal wavefunctions with those of a frozen-core full-CI calculation. ground states, is assumed when the second-derivative matrix is
The molecule’s electron density will be shown to exhibit a non- positive definite, and the gradient is essentially zero. In GMCSC cal-
nuclear maximum, located rather close to the centre of mass. This culations, the convergence threshold for gradient components is
is in agreement with previous RHF results [11], and is of course usually set at 108 atomic units.
hardly a surprise, given that such non-nuclear maxima, though ab-
sent from most molecules [12], appear in higher lithium clusters 2.1. Geometry, basis sets, integrals
[13], while Li2 has long been known to exhibit one at the SCF level
[14], even though its persistence at correlated levels of approxima- Except when otherwise indicated, calculations were carried out
tion has only been proved in more recent times [13,12,15–18]. at the equilibrium C2v geometry, with the two equal Li–Li distances
Moreover, an ‘Atoms-in-Molecules’ [19] topological analysis R set to 5.225 a0 and the intervening angle a set equal to 71.6° [1],
will be used to characterize the molecule’s bonding structure, so the third Li–Li distance was 6.113 a0 (here a0 denotes the atomic
highlighting Bond Path features that may be viewed as related to unit of distance, or Bohr: 1 a0 = 0.5291772  1010 m [25]).
the occurrence of Interstitial Orbitals in the electronic This minimum-energy geometry, determined in a high-accu-
wavefunction. racy theoretical–experimental study [1–3], is quite close to that
Finally, Li3’s electron density difference map (molecule minus found in high-quality CCSD(T) calculations [26], i.e. R = 5.218 a0
atoms) will be shown to exhibit 3p-type enrichment-depletion pat- and a = 71.8°.
terns, one centred at each nucleus, on top of the expected non-nu- In the present work, the two symmetry-equivalent nuclei were
clear maximum located at about the same position as in the total placed at (x = 0.0, y = ±3.056404 a0, z = +4.237808 a0), with the
density. Corresponding patterns will be shown to exist for Li2. third nucleus at the origin.
Most calculations were carried out with even-tempered
2. Method [8s,4p,1d,1f] or [8s,4p,1d] Slater-type basis sets, even-tempered
meaning that the exponential parameters of the 1s STOs were
Calculations were carried out in the framework of the Generalized assumed to be given by
Multiconfiguration Spin-Coupled (GMCSC) approach [20,21]. The fi ¼ a1s bi1
1s
method is essentially a multiconfiguration generalization of the
Spin-Coupled (SC) approach ([6–8]; see e.g. [9] or [10] for a review), for i = 1, 2,. . ., 8, and similarly for the four 2p STOs.
though it can also be viewed as a non-orthogonal generalization of STO-based energy integrals were computed using M2CNEW
the Multiconfiguration Self-Consistent Field (MCSCF) approach. [27,28], a program that in its newer version, SMILES [29], is being
More specifically, in the GMCSC method the wavefunction is ex- included in the MOLPRO package [30]. The use of M2CNEW in con-
pressed as a linear combination of configurations, each the anti- nection with basis-set optimisation has been discussed elsewhere
symmetrized product of an orbital ‘string’ and a spin function. [31]. STO-based integrals of the electric field gradient at the nuclei
Each orbital ‘string’ is just the product of as many orbitals as there were computed with the highly accurate and efficient INTEGARM-
are electrons, while its associated spin function is a linear combi- IRREG program [32].
nation of all possible modes of spin coupling, embodied by the Exponential parameters for the larger [8s,4p,1d,1f] basis set
Yamanouchi-Kotani (YK) spin functions [22], with adjustable coef- were as optimised for an eight-configuration CASSCF wavefunction
ficients termed spin-coupling coefficients. The theory was initially for Li2 [23]. Those for the smaller [8s,4p,1d] basis set were either
formulated under the assumption of a single set of spin-coupling equal to the corresponding ones in the larger set, or as directly
coefficients, common to all configurations, as the MCSC approach optimised for the smaller set, either for the same Li2 CASSCF wave-
[20], and later generalized to allow for a separate set of spin-cou- function, or specifically for Li3 at various correlation levels. In all
pling coefficients for each configuration [21]. these optimisations, the parameters (a and b) of each of the two
Just as in SCF, SC or MCSCF theory, each orbital is in turn ex- geometric progressions were determined by effectively treating
pressed as a linear combination, with adjustable coefficients (orbi- them as variational parameters (details available, on request, from
tal coefficients), of basis functions, usually of Gaussian or Slater the present author), together with the exponential parameters of
type. And, just as in the single-configuration SC method, the orbi- the 3d and, when present, 4f STOs, and all other variational param-
tals are not restricted by double-occupancy or orthogonality eters in the wavefunction. Optimised values are given in Table 1:
20 F.E. Penotti / Journal of Molecular Structure: THEOCHEM 861 (2008) 18–26

no simple trend with increasing correlation can be easily the two equal sides (the occurrence of IOs was a recurrent theme
discerned. for higher Li clusters as well [5]).
RHF calculations were carried out to establish the zero of corre- The valence SC orbitals are reproduced here mostly for compar-
lation energy, both with these STO basis sets and, for comparison, ison with those from multiconfiguration wavefunctions to be dis-
using the freely available GAMESS-US package [33], with standard cussed later.
high-accuracy CGTO basis sets. These consisted of (11s,5p,2d,1f) As for the spin part of the SC wavefunction, it is mostly deter-
and (12s,6p,3d,2f,1g) cartesian Gaussian primitives contracted to mined by the observation that there are 42 possible YK spin func-
[4s,3p,2d,1f] and [5s,4p,3d,2f,1g] CGTOs, respectively, (cc-pVTZ tions for nine electrons coupled to a doublet, but only two for three
and cc-pVQZ [34,35]). Numerical results are given in Table 2. As electrons thus coupled. Thus, if the inner orbitals are taken as the
can be seen, at the RHF level specific optimisation of the basis- first six orbitals, and ordered pair after pair, only the 28th and
set parameters more than makes up for omission of the 4f STO, the 42nd (perfect pairing) standard-order YK functions correspond
and leads to an energy that is even lower than that given by the to singlet coupling of each of the three inner pairs, and not surpris-
much larger cc-pVQZ Gaussian basis set, albeit not by much ingly the coefficients of the other 40 spin functions turn out to be
(0.001 eV). of order 103 or less. If furthermore the a1 valence orbitals are ta-
The same two Gaussian basis sets were also used for the single- ken as orbitals 7 and 8, the coefficients of YK functions 28 and 42
configuration Spin-Coupled wavefunction, to provide a correlated- turn out to be 0.44 and 0.90, respectively, so that perfect pairing
level comparison with the STO results. In this case, all three STO alone accounts for 81% of the wavefunction.
basis sets explored (Table 2) led to lower energies than even the Correspondingly, omission of the 40 ‘secondary’ spin functions
larger Gaussian basis set, by 0.031, 0.034 and 0.040 eV. leads to a small energy penalty, of the order of microhartrees, both
for the SC wavefunction and for the multiconfiguration wavefunc-
3. Results tions to be described later. Nevertheless, the results given in this
paper always refer to calculations including all 42 spin functions,
3.1. The SC description of Li3 subject only to those constraints that are required by symmetry
[38,31] and/or the presence of any doubly-occupied orbitals (the
For the single-configuration SC wavefunction, variational opti- spin part of the SC wavefunction spontaneously satisfies the
misation at the equilibrium C2v geometry spontaneously yields a requirements for overall 2B2 symmetry, as expected, in the sin-
‘clean-symmetry’ 2B2 solution, with no need for any symmetry con- gle-configuration case, when orbital symmetry allows [6,38]).
straints. The SC orbitals are all even under reflection in the plane of At the SC level specific optimisation of the basis-set parameters
the nuclei, and stable with respect to mixing-in of basis functions does not quite make up for omission of the 4f STO (see Table 2).
that are odd under such reflection. They consist of three pairs of Still, it does lead to an energy which is 0.03 eV lower than that
‘inner’ orbitals (in-pair overlap 93%), and three ‘valence’ orbitals. yielded by the cc-pVQZ CGTO basis set, even though the latter in-
Unsurprisingly, one of the inner pairs consists of a1 orbitals that cludes almost three times as many basis functions (249 cartesian
are essentially atomic 1s orbitals centred on the ‘apical’ nucleus, Gaussian primitives contracted to 210 CGTOs, vs. 75 STOs).
a second pair also consists of essentially 1s orbitals, but centred
on one of the two C2-related nuclei, and the third pair can be ob- 3.2. The GMCSC description
tained by applying C2 to the second pair. The ‘valence’ orbitals con-
sist of two a1 orbitals, with a 62% mutual overlap and a 5–14% It is also possible to obtain, at the same equilibrium C2v geom-
overlap with inner orbitals, and a b2 orbital with ±7% and ±12% etry, a solution consisting of a pair of configurations that are con-
overlap with the inner orbitals in the two symmetry-related pairs. strained to be connected by a C2 rotation, have equal and opposite
The valence orbitals, as obtained with a specifically optimised configuration coefficients, and share the same set of spin-coupling
[8s,4p,1d] STO basis set, are depicted in Fig. 1, and are qualitatively coefficients (so this is, strictly speaking, a MCSC rather than a
much the same as those obtained in previous work by Tornaghi et GMCSC wavefunction). All orbitals turn out to be even under
al. [5]. This despite their use of a (9s 5p) ? [3s 2p] Gaussian basis reflection in the plane of the nuclei, and the solution thus has clean
set, and of a geometry chosen as appropriate for a cut through a bcc 2
B2 symmetry, even though each individual configuration does not,
lattice, rather than the isolated molecule, and thus somewhat dif- and in fact the two configurations’ 44% overlap implies their sym-
ferent from that adopted here. And despite the fact that those cal- metry distortion is significant. The wavefunction can of course be
culations adopted a ‘frozen core’ consisting of the lowest three RHF seen as a single symmetry-adapted linear combination (SALC) of
orbitals (doubly-occupied), to which the valence orbitals could be two configurations. The six ‘inner’ orbitals in either of the two ‘res-
taken orthogonal, with the presumable introduction of extra nodal onating’ configurations are quite similar to their SC counterparts.
surfaces close to the nuclei. On the other hand, two of the valence orbitals appear to be essen-
In that work [5], one of the valence a1 orbitals was described as tially IOs associated with the two equal sides of the triangle. Their
an Interstitial Orbital (IO) associated with the unique side of the overlap is only about 5%. The third valence orbital looks like a rota-
isosceles triangle, and the other two were viewed as the in-phase tion and distortion of SC orbital 9, i.e. the essentially unpaired b2
(a1) and out-of-phase (b2) combinations of IOs associated with orbital. The common spin part can be made to be heavily domi-

Table 1
Optimised parameters for even-tempered [8s,4p,1d,1f] and [8s,4p,1d] STO basis sets

Molecule Method Number of configurations 1s (a/a01, b) 2p (a/a01, b) 3d f 4f f


Total Indep.
Li2 CASSCF 8 7 (0.573899, 1.455226) (0.636270, 1.424770) 0.788438 0.948885
Li2 CASSCF 8 7 (0.578254, 1.451966) (0.720272, 1.390743) 0.863570 ––
Li3 SCF 1 1 (0.565501, 1.454940) (0.523926, 1.648522) 0.789639 ––
Li3 SC 1 1 (0.602985, 1.499617) (0.583532, 1.498391) 0.800953 ––
Li3 GMCSC 6 4 (0.517043, 1.420801) (0.708769, 1.703738) 0.858543 ––
Li3 GMCSC 15 9 (0.520651, 1.420011) (0.704527, 1.745813) 0.866966 ––
F.E. Penotti / Journal of Molecular Structure: THEOCHEM 861 (2008) 18–26 21

Table 2
Main numerical results at MR–CI equilibrium geometry (R = 5.225 a0, h = 71.6° [1])

Method Configurations E/Eha Ecorrb/eV lc/D Electric field gradient d/(103 e a03)
Total Indep. Nucleus 1 Nuclei 2 and 3
|| || || ||
RHFe 1 1 22.308131 0.67 9.9 +1.4 +8.5 4.7 2.9 +7.6
RHFf 1 1 22.308196 0.65 10.1 +1.5 +8.6 4.7 2.8 +7.5
RHFg 1 1 22.308229 0.65 10.4 +1.8 +8.6 4.9 2.7 +7.6
RHFh 1 1 22.308277 0.68 10.3 +1.9 +8.4 4.9 2.7 +7.6
OBS–RHFi 1 1 22.308325 0.00 0.68 10.4 +1.8 +8.6 4.9 2.9 +7.8
SCh 1 1 22.372502 1.75 1.48
SCi 1 1 22.373652 1.79 1.46 14.1 6.2 +20.3 9.6 3.5 +13.1
OBS–SCj 1 1 22.373739 1.79 1.44 13.9 6.3 +20.3 9.5 3.6 +13.0
SCg 1 1 22.373976 1.79 1.46 14.3 5.9 +20.2 9.8 3.1 +12.9
MCSCg 2 1 22.380238 1.96 0.12 8.1 0.2 +8.3 13.9 3.3 +17.2
VB–CIg 3 2 22.380994 1.98 0.51 13.9 3.8 +17.7 8.3 0.8 +9.1
GMCSCg 3 2 22.399022 2.47 0.36 17.5 +5.6 +11.9 11.8 +0.2 +11.6
GMCSCg 4 3 22.403465 2.59 0.53 16.2 +5.7 +10.5 10.7 1.2 +11.9
GMCSCg 6 4 22.407860 2.71 0.49 15.2 +5.3 +10.0 8.9 1.6 +10.5
GMCSCg 8 5 22.408013 2.72 0.49 15.1 +5.4 -9.8 8.9 1.7 +10.6
GMCSCf 6 4 22.406898 2.68 0.48 14.4 +4.0 +10.4 8.1 1.4 +9.5
GMCSCk 6 4 22.408555 2.73 0.42 15.3 +5.9 +9.4 9.0 2.0 +11.0
OBS–GMCSCl 6 4 22.412503 2.84 0.34 19.5 +5.2 +14.3 13.1 2.9 +16.0
GMCSCl 9 6 22.418797 3.36 0.37 4.0 2.4 +6.5 13.3 2.5 +15.8
GMCSCl 15 9 22.444839 3.72 1.01 10.8 3.8 +14.6 5.2 3.5 +8.7
OBS-GMCSCm 15 9 22.444953 3.72 1.00 10.8 3.9 +14.7 5.2 3.5 +8.7
CCSD(T)n 22.48830 4.90
D–QMCo 22.4900 ± 0.0023 4.94 ± 0.06
Experiment 22.502 ± 0.006p 5.3 ± 0.2
a
Eh denotes the atomic unit of energy, or Hartree: 1 Eh = 4.35975  1018 J [25].
b
Correlation energy, in electron-volts: 1 eV = 1.602177  1019 J [25].
c
Electric dipole moment, in Debye: 1 D = 3.335641  1030 C m [25].
d
Eigenvalues of the electronic distribution’s electric field gradient at the nuclei (traceless part only), in atomic units: 1 ea03 = 1.081203 1022 C m3 [25]. Symbols label
the direction of the corresponding eigenvectors (perpendicular to the plane of the nuclei: ; in the plane of the nuclei: ||).
e
(11s,5p,2d,1f) ? [4s,3p,2d,1f] cartesian GTO basis (cc-pVTZ [34]).
f
[8s,4p,1d] even-tempered STO basis, exp. parameters as optimised for Li2 [8s,4p,1d,1f] basis (CASSCF).
g
[8s,4p,1d,1f] even-tempered STO basis, exponential parameters as optimised for Li2 (CASSCF).
h
(12s,6p,3d,2f,1g) ? [5s,4p,3d,2f,1g] cartesian GTO basis (cc-pVQZ [34]).
i
[8s,4p,1d] even-tempered STO basis, exp. parameters as optimised for SCF wavefunction.
j
[8s,4p,1d] even-tempered STO basis, exp. parameters as optimised for SC wavefunction.
k
[8s,4p,1d] even-tempered STO basis, exp. parameters as optimised for Li2 (CASSCF).
l
[8s,4p,1d] even-tempered STO basis, exponential parameters optimised for six-configuration GMCSC wavefunction.
m
[8s,4p,1d] even-tempered STO basis, exponential parameters optimised for 15-configuration GMCSC wavefunction.
n
CCSD(T) calculation [26], with (15s,9p,5d,3f,1g) ? [8s,7p,5d,3f,1g] dunning cc-pwCVQZ GTO basis.
o
Diffusion Quantum Monte Carlo calculation, with a geometry optimised at CASSCF level (R = 5.31 a0, h = 71.2°) [11].
p
Three times the total energy of Li [36], plus the experimental dissociation energy of Li3 [37], plus its zero-point energy [1,2], assuming errors to be dominated by that on
the dissociation energy [37].

nated by perfect pairing, with a 0.9998 coefficient, provided this In such a calculation, the re-optimised SC-like configuration
latter orbital is taken as # 7, i.e. spin-paired with one of the two ends up having a limited overlap (±33%) with each of the other
IOs, with which it has a 66% overlap, while the other IO, which two configurations, whose mutual overlap rises to 72%.
has a 31% overlap with orbital 7, is made orbital 9, i.e. the un- Perfect pairing remains dominant for all three configurations.
paired orbital. The SC-like configuration assumes a weight (Chirgwin-Coulson
As can be seen from Table 2, the two-configuration solution’s occupation number [39]) equal to 0.36, with the two symmetry-
correlation energy is 1.96 eV, to be compared with the SC solution’s equivalent configurations making up the balance to 1.0. This sug-
1.79 eV. gests all three configurations are equally important, and should
The relative importance of the two descriptions can be better be taken as reference functions for the introduction of double exci-
assessed by putting these two equivalent configurations together tations, to provide ‘‘out-of-plane” correlation.
with the SC configuration, in a three-configuration wavefunction. The re-optimised orbitals in the SC-like configuration are rea-
If this is done simply in a non-orthogonal CI calculation (VBCI), sonably similar to their counterparts in the SC wavefunction itself,
equivalent to varying only configuration and spin-coupling coeffi- but the three ‘valence’ ones display extra nodal surfaces, a struc-
cients while keeping all orbitals fixed, the correlation energy in- ture that they will retain when doubly-excited configurations are
creases a paltry 0.02 eV to 1.98 eV. However, if all orbitals are added. In fact, they are quite similar to their counterparts in a
simultaneously re-optimised, in a three-configuration GMCSC cal- 15-configuration GMCSC wavefunction to be described later (see
culation that includes an SC-like configuration, identified as such Fig. 2), though in the latter the a1 orbital that was viewed as an
by appropriate orbital and spin symmetry constraints [38,30], IO associated with the triangle’s unique side moves towards the
and a pair of configurations that are constrained to be related by molecule’s central region, its amplitude showing a maximum at
C2 and to have identical spin-coupling coefficients, with all orbitals y = 0.0 a0, z = 2.48 a0, and could be described as the in-phase com-
constrained to be invariant under reflection in the plane of the nu- bination of either all three IOs, or of three distorted sp2-type hy-
clei, the correlation energy increases to 2.47 eV. brids, each centred on one of the nuclei. The two descriptions
22 F.E. Penotti / Journal of Molecular Structure: THEOCHEM 861 (2008) 18–26

Fig. 2. Contour plots of the amplitudes, in the plane of the nuclei, of the SC-like
Fig. 1. Contour plots of the amplitudes, in the plane of the nuclei, of the single- configuration’s ‘valence’ orbitals in the 15-configuration GMCSC wavefunction.
configuration SC ‘valence’ orbitals. Contours have been drawn every 0.02 a03. Contours have been drawn every 0.02 a03. Dashed lines mark negative contours.
Dashed lines mark negative contours.

can perhaps be reconciled, if the IOs themselves are assumed to


arise from the superposition of the sp2-type hybrids.
The appearance of the orbitals in the two C2-related configura-
tions also remains more or less the same as in the two-configura-
tion MCSC wavefunction. And while the two that are essentially
IOs associated with either of the equal sides of the triangle will re-
tain their essence in the final 15-configuration wavefunction (see
Fig. 3), the one that in the three-configuration wavefunction still
looks like a distortion of the b2 SC orbital will there take on its ex-
pected role as the third IO, the one associated with the triangle’s
unique side, undergoing substantial changes that will make it look
like a very recognizable distortion of its a1 counterpart in the SC
wavefunction (Fig. 3). It thus appears that it takes the inclusion
of three reference configurations and of ‘out-of-plane’ double exci-
tations out of them to see all three IOs make their direct appear-
ance in the wavefunction.
To gain some sense of the convergence behaviour, double exci-
tations out of the three reference configurations were added in
stages. First, one consisting in the replacement of the ‘valence’ pair
of a1 orbitals in the SC-like configuration with a doubly-occupied
b1 orbital, recovering an extra 0.12 eV of correlation energy. Then,
a pair of C2-related configurations obtained by replacing the first
two of the a0 valence orbitals in either of the other two reference
configurations with one of a pair of C2-related doubly-occupied
a00 orbitals (a0 and a00 are used here to denote orbitals that are only
constrained to be even or odd, respectively, under reflection in the Fig. 3. Contour plots of the amplitudes, in the plane of the nuclei, of the ‘valence’
plane of the nuclei, using Cs-group notation). This GMCSC calcula- orbitals in one of the pair of C2-related reference configurations, as they appear in
tion, whose configurations are the first six listed in Table 3, recov- the 15-configuration GMCSC wavefunction. Contours have been drawn every
0.02 a03. Dashed lines mark negative contours.
ered a further 0.12 eV of correlation energy. Adding a pair of C2-
related configurations obtained by replacing the first and third of
the a0 valence orbitals in either of the two C2-related reference con- two configurations, not listed in Table 3, were omitted in subse-
figurations (orbitals 7 and 9) with one of another pair of C2-related quent calculations.
doubly-occupied a00 orbitals, in an eight-configuration GMCSC cal- As can be seen from Table 2, at the six-configuration level spe-
culation, yielded only an extra 0.01 eV of energy (Table 2). These cific optimisation of the basis-set parameters more than made up
F.E. Penotti / Journal of Molecular Structure: THEOCHEM 861 (2008) 18–26 23

for omission of the 4f STO, as had been the case at the RHF level In an effort to assess the accuracy of the present results, at least
(but not at the SC level). with respect to the treatment of valence correlation, frozen-core
Subsequent calculations added inner-shell ‘out-of-plane’ corre- calculations, using SCF core orbitals, were carried out at various
lation, first for the SC-like configuration: one excitation consisted correlation levels, all the way up to full-valence CI (FVCI). To make
in the replacement of the pair of ‘inner’ a1 orbitals with a dou- the latter calculation possible, with the freely-available GAMESS
bly-occupied b1 orbital, while two other symmetry-related excita- [33] program, on the present author’s PC, the cc-pVTZ CGTO basis
tions were obtained by replacing either of the two C2-related set was used. The results, given in Table 4, show that the frozen-
‘inner’ orbital pairs with one of two doubly-occupied (and C2-re- core six-configuration GMCSC wavefunction recovers 86% of the
lated) a00 orbitals. This nine-configuration GMCSC calculation full-valence correlation energy, and provides a value of the dipole
yielded an extra 0.52 eV of correlation energy (Table 2). moment that is in reasonably good agreement with the FVCI value.
Finally, six more configurations were added, similarly providing In fact, the six-configuration GMCSC dipole value is better than
inner-shell ‘out-of-plane’ correlation for the pair of C2-related ref- that obtained in a CI calculation supplementing the SCF configura-
erence configurations, yielding an extra 0.36 eV of correlation en- tion with all (valence) single and double excitations out of it (SDCI),
ergy. As can be seen from Table 2, this final GMCSC wavefunction even though the latter, with its 5713 configurations, recovers 95%
recovers about 70% of the total correlation energy. The list of all of the full-valence correlation energy. On the other hand, the six-
15 configurations can be found in Table 3. configuration GMCSC dipole moment is still little more than half
The final wavefunction remained dominated by the first three that yielded, upon inclusion of core correlation, by the 15-configu-
configurations (weights: 0.28, 0.35, 0.35), the other 12 configura- ration GMCSC wavefunction. Since ‘in-plane’ core correlation does
tions having weights that, while positive, were all below 0.01. Per- not increase dipole moment values, as can be seen in a comparison
fect pairing still dominated all configurations. The re-optimised SC- of Table 4 results with their Table 2 counterparts, the accuracy of
like configuration ended up having a ±88% overlap with each of the the 15-configuration GMCSC value hinges on its last nine configu-
other two reference configurations, whose mutual overlap became rations providing an adequate description of ‘out-of-plane’ core
large and negative (95%). correlation.
As can be seen from Table 2, inclusion of inner-shell ‘out-of-
plane’ correlation, aside from lowering the total energy, unexpect- 3.3. Electron density
edly almost tripled the molecule’s electric dipole moment, to 1.0 D,
a value that falls roughly halfway between the RHF and SC values. At the C2v minimum-energy geometry, the molecule’s electron
One might suspect that this may be due to the fact that some core density was searched for any maxima or minima. Density values
correlation was allowed in, already at the single-configuration SC were first computed on a three-dimensional grid covering the
level, by lifting the double-occupancy constraint on inner orbitals. range 0.0 a0 6 x 6 +4.0 a0, 6.0 a0 6 y 6 +6.0 a0, 2.0 a0 6 z 6
However, when the SC calculation was repeated, using the even- +6.0 a0. Then, each grid point at which the density was higher
tempered STO [8s,4p,1d,1f] basis set, with the six inner orbitals (lower) than at all nearest neighbours and next-nearest neighbours
constrained to be identical in pairs, thus restoring the double-occu- was taken as a starting point for a stabilized Newton–Raphson
pancy constraint on inner orbitals, the overestimated dipole mo- search, using analytical values of the density’s first and second
ment fell by less than 0.01 D (while the energy rose by as much derivatives with respect to the three cartesian coordinates. As
as 1.21 eV). mentioned above, one of the nuclei was placed at the origin, the
The electric field gradient at the nuclei also appears to be signif- other two at (x = 0.0, y = ±3.056404 a0, z = +4.237808 a0).
icantly affected by inner-shell ‘out-of-plane’ correlation (Table 2), The molecule’s electron density was found to exhibit a non-nu-
but this is perhaps to be expected. More generally, both quantities clear maximum; no minima were found. The contour plot in Fig. 4
appear to be quite sensitive to the level of correlation included. Of has been obtained from the 15-configuration OBS–GMCSC
course, the best values should be those obtained with the 15-con- wavefunction.
figurations OBS–GMCSC wavefunction. Here the non-nuclear maximum has about the same value as
that exhibited by Li2: 0.0121 a03 compared with 0.0132 a03 for
the diatomic (see e.g. Ref. 40). In Fig. 4 the triatomic’s maximum
Table 3
Orbital configurationsa used in GMCSC calculations on Li3 lies at z = 3.095 a0, just 0.27 a0 from the molecular centre of mass,
which is at z = 2.825 a0.
1 1a1 2a1 1a0 2a0 C21a0 C22a0 3a1 4a1 1b2
As originally shown by R.K. Owen at a similar geometry
2 3a0 4a0 5a0 6a0 7a0 8a0 9a0 10a0 11a0
3 C23a0 C24a0 C25a0 C26a0 C27a0 C28a0 C29a0 C210a0 C211a0 (R = 5.31 a0, h = 71.2°) [11], a corresponding non-nuclear maximum
4 1a1 2a1 1a0 2a0 C21a0 C22a0 1b1 1b1 1b2 is found at the RHF level. In the present work, its existence has also
5 3a0 4a0 5a0 6a0 7a0 8a0 1a00 1a00 11a0
6 C23a0 C24a0 C25a0 C26a0 C27a0 C28a0 C21a00 C21a00 C211a0
7 2b1 2b1 1a0 2a0 C21a0 C22a0 3a1 4a1 1b2 Table 4
8 1a1 2a1 2a00 2a00 C21a0 C22a0 3a1 4a1 1b2 Main frozen-core results (cc-pVTZ basis set) at MR–CI equilibrium geometry
9 1a1 2a1 1a0 2a0 C22a00 C22a00 3a1 4a1 1b2 (R = 5.225 a0, h = 71.6° [1])
10 3a00 3a00 5a0 6a0 7a0 8a0 9a0 10a0 11a0
11 3a0 4a0 4a00 4a00 7a0 8a0 9a0 10a0 11a’ Method Configurations E/Eha Ecorrb/eV lc/D
12 3a0 4a0 5a0 6a0 5a00 5a00 9a0 10a0 11a0
Total Indep.
13 C23a00 C23a00 C25a0 C26a0 C27a0 C28a0 C29a0 C210a0 C211a0
14 C23a0 C24a0 C24a00 C24a00 C27a0 C28a0 C29a0 C210a0 C211a0 RHF 1 1 22.308131 0.00 0.670
15 C23a0 C24a0 C25a0 C26a0 C25a00 C25a00 C29a0 C210a0 C211a0 SC 1 1 22.329182 0.573 (42%) 1.449
a
MCSC 2 1 22.335155 0.735 (54%) 0.288
Orbitals are identified by a symmetry label preceded by a number, which is only GMCSC 3 2 22.342923 0.947 (70%) 0.379
meant to distinguish different orbitals of the same symmetry. Thus, 1a1 is the first GMCSC 6 4 22.351403 1.177 (86%) 0.508
orbital of a1 symmetry, in the full C2v group, appearing in the Table, while 2a1 is the SDCI 5713 5713 22.355417 1.287 (95%) 0.437
second orbital of the same symmetry. The a’/a00 labels are used to denote orbitals FVCI 92,167 92,167 22.358161 1.361 (100%) 0.587
which are even/odd under reflection in the plane of the nuclei, but are neither even
a
nor odd under the other operations of the C2v group. Note different orbitals are not Eh denotes the atomic unit of energy, or Hartree: 1 Eh = 4.35975  1018 J [25].
b
orthogonal, unless this is implied by symmetry. In fact, the first three pairs of every Correlation energy, in eV, and as a percentage of the full-valence correlation
configuration are essentially ‘inner’ orbitals, and the two orbitals of each such pair, energy.
c
even when different, overlap strongly (and are almost exactly singlet-coupled). Electric dipole moment, in Debye: 1 D = 3.335641  1030 C  m [25].
24 F.E. Penotti / Journal of Molecular Structure: THEOCHEM 861 (2008) 18–26

been verified at the FVCI level (cc-pVTZ basis set) and, at the RHF y = ±1.8 bohr, z = +2.1 bohr. However, the cc-pVQZ single-configu-
level, both with CGTO (cc-pVTZ and cc-pVQZ [34,35]) and STO ba- ration SC electron density exhibits a single maximum at
sis sets (the [8s,4p,1d,1f] optimised for Li2, and with the specifically z = +3.69 bohr, and two equivalent saddle points at y = ±1.87 bohr,
optimised [8s,4p,1d]). Of course, quantitative details do depend on z = +2.05 bohr. These two saddle points are characterized by a sin-
basis set and correlation level (see Table 5). gle positive (and rather small) eigenvalue of the second-derivative
Fig. 4 also shows, superimposed on the contour plot, so-called matrix. Because of this, and of their positions, they are obviously
Bond Critical points (BCPs) and Bond Paths, defined in accordance closely related to the two symmetry-equivalent non-nuclear max-
with Baden’s ‘‘Atoms-in-Molecules” theory [19]. The Bond Paths’ ima observed, also at the SC level, with the cc-pVTZ and STO basis
trajectories support the contention that the three nuclei are indi- sets.
rectly bound together, through the electron density’s non-nuclear For the six-configuration GMCSC wavefunction, an admittedly
maximum, in accordance with similar observations for Li2, Li4 minor splitting of the electron density non-nuclear maximum, into
and Li5 [13]. It is interesting to note that two of the three Bond a pair of closely-spaced and symmetrically placed maxima with
Paths, as they lead away from the non-nuclear maximum, initially barely a dip between them, also appears to be basis-set dependent
bend in the general direction of the midpoints of the two equiva- (see Table 5).
lent sides of the ‘nuclear triangle’, before turning towards the Though at this minimum-energy geometry such ‘splittings’ are
two equivalent nuclei. This feature, which is even more obvious probably an artefact of specific correlation levels and basis sets,
in the shape of one of the two 0.011 a.u. contours, may be con- similar behaviour would actually not be unexpected over some
nected with the appearance of Interstitial Orbitals in the electronic range of geometries, in analogy with what has been well docu-
wavefunction. In any case, this ‘bending’ is barely discernible in a mented for Li2 (see e.g. Ref. [40], and references therein).
similar plot based on the SCF wavefunction, but well developed At the D3h geometry of lowest energy (R = 5.428 a0 [1,2]), the
for the full-valence CI density. It thus appears to be mostly a prod- electron density also appears to show a single non-nuclear maxi-
uct of electron correlation, whether or not one wishes to ascribe it mum which, for the 2E0 component correlating with 2B2, lies at
to Interstitial Orbitals. 4.1, 2.8 and 2.8 a0 from the three nuclei (1.0 a0 from the molecular
Values of the laplacian of the electron density at the three BCPs centre of mass), and has a value of 0.0131 a03. For the orthogonal
are approximately +0.01 a.u. They are therefore of comparable component, it lies on the opposite side of the molecular centre of
magnitude to those found in Li2, Li4 and Li5 [13], and much smaller mass, at 0.6 a0 from it and at 2.5, 3.5 and 3.5 a0 from the three nu-
than typical molecular values [19], implying a low value of kinetic clei, and has a value of 0.0128 a03.
energy per electron, a trait that has been associated with good con- At the C2v interconversion saddle points (R = 5.715 a0, a = 52.2°
duction properties [12]. [1,2]), an electron density maximum is found close to the midpoint
In general, positive values of the laplacian imply the electron between the two equivalent nuclei, whose separation is only
density is locally depleted and, when found at BCPs, are usually 5.029 a0, close to the experimental equilibrium internuclear dis-
associated with non-covalent interactions dominated by kinetic- tance in Li2 (5.051 a0 [41]).
energy contributions [19]. However, the significance, in this re-
spect, of laplacian values at BCPs has been downplayed for Li2, 3.4. Electron density difference
Li4 and Li5 [13], in favour of the dominant negative sign of the
laplacian in the internuclear regions, taken to imply covalent inter- If the electron density difference (molecule minus atoms) is
actions dominated by potential-energy contributions [19,13]. This plotted in the nuclear plane, one also observes a non-nuclear max-
behaviour of the laplacian can be seen in Li3 as well (Fig. 5). imum not far from its location in the total molecular density, plus
At intermediate correlation levels, the non-nuclear maximum three 3p-type enrichment-depletion patterns, with their own posi-
can actually split into two or three maxima. Notably, at the sin- tive-lobe maxima and negative lobe minima, one such pattern cen-
gle-configuration Spin-Coupled level, with STO basis sets (both tred at each nucleus. The contour plot in Fig. 6 was obtained using
the [8s,4p,1d,1f] optimised for Li2, and the specifically optimised the 15-configuration OBS–MCSC wavefunction for the molecule,
[8s,4p,1d]), and with the CGTO cc-pVTZ [34,35] basis, there are and, for each Li atom, a nine-configuration OBS–GMCSC wavefunc-
three non-nuclear maxima. For these basis sets, the highest of tion that can be described as
the three non-nuclear maxima lies on the z axis, at z = +3.7 bohr.
The other two, symmetry-equivalent, non-nuclear maxima are at X
3 X
5
2
ðsI sII sIII Þ þ k ðp2a sIV Þ þ l ðdb sV Þ;
a¼1 b¼1

with both spin functions included for the first configuration. The
atomic wavefunction used a specifically-optimised even-tempered
[8s,4p,1d] STO basis set.
In this contour plot, the axes of the 3p-type patterns appear to
meet at approximately z = 1.0 a0. On the other hand, the non-nu-
clear maximum closest to the one in the total density lies at
z = 3.40 a0, and has a value of 0.006 a03. Qualitatively similar pat-
terns have been observed at all correlation levels and with all basis
sets (Gaussian and Slater) explored. They also occur at the D3h
geometry of lowest energy (R = 5.428 a0 [1,2]), with details
depending on which of the two spatial components of the 2E0 state
is being considered.
Such patterns have been previously published for higher lith-
ium clusters [42], and in the present case their orientation does
Fig. 4. Contour plot of Li3’s electron density, in the plane of the nuclei, as computed not appear incompatible with the contention that this molecule’s
from the 15-configuration GMCSC wavefunction. Contours have been drawn every electron density binds the three nuclei indirectly, through its
0.001 a03, up to 0.012 a03 (the nuclear cusps rise to 13.862, 13.862 and 13.819 a03).
non-nuclear maximum. They also occur in the diatomic:
Crosses mark the positions of the nuclei, black dots those of the three Bond Critical
Points. Bond Paths have also been marked. Fig. 7 shows a plot, at the diatomic’s experimental equilibrium
F.E. Penotti / Journal of Molecular Structure: THEOCHEM 861 (2008) 18–26 25

Table 5
Electron-density non-nuclear maxima at MR–CI equilibrium geometry (R = 5.225 a0, h = 71.6°; all values in a.u.) [1]

Method Configurations Position Electron density Hessian eigenvaluesa/103


Total Indep. y z xb yb zb
c
SCF 1 1 0.00 +2.73 0.01339 4.57 0.53 3.31
SCFd 1 1 0.00 +2.69 0.01336 4.62 0.31 3.19
OBS–SCFe 1 1 0.00 +2.70 0.01321 4.37 0.32 3.14
FVCIc 92,167 92,167 0.00 +2.79 0.01280 4.24 0.43 2.77
GMCSCf 3 2 0.00 +2.76 0.01279 4.88 0.45 2.38
GMCSCf 6 4 0.00 +2.74 0.01239 3.92 0.32 2.53
OBS-GMCSCg 6 4 ±0.15 +2.68 0.01206 3.80 0.02h 2.32i
OBS-GMCSCj 15 9 0.00 +3.10 0.01208 3.78 0.56 1.87
a
Eigenvalues of the matrix of the electronic density’s second derivatives wrt cartesian coordinates.
b
Direction of the corresponding eigenvectors. The nuclei are in the yz plane; the C2 axis lies along z (see text).
c
(11s,5p,2d,1f) ? [4s,3p,2d,1f] cartesian GTO basis (cc-pVTZ [34]).
d
(12s,6p,3d,2f,1g) ? [5s,4p,3d,2f,1g] cartesian GTO basis (cc-pVQZ [34]).
e
[8s,4p,1d] even-tempered STO basis, exponential parameters as optimised for SCF wavefunction.
f
[8s,4p,1d,1f] even-tempered STO basis, exponential parameters as optimised for Li2 (eight-configuration CASSCF).
g
[8s,4p,1d] even-tempered STO basis, exponential parameters as optimised for six-configuration GMCSC wavefunction.
h
The corresponding eigenvectors actually lie at ±3.75° from the y axis.
i
The corresponding eigenvectors actually lie at ±3.75° from the z axis.
j
[8s,4p,1d] even-tempered STO basis, exponential parameters optimised for 15-configuration GMCSC wavefunction.

4. Conclusions

A compact non-orthogonal multiconfiguration wavefunction


has been obtained for Li3 in its ground-state C2v minimum-energy
geometry. This wavefunction is based on just three reference con-
figurations, two of which symmetry-equivalent, and 12 ‘excita-
tions’ out of them, for a grand total of 15 configurations, which
can be grouped in just nine SALCs. Yet it is able to recover, for this
nine-electron open-shell system, about 70% of the total correlation
energy.
The description of the electronic structure that emerges can be
viewed as a relatively straightforward generalization and refine-
ment of that provided by the single-configuration SC wavefunction,
and in fact brings to the forefront the Interstitial Orbitals invoked
in the interpretation of the latter [5]. Given that, at the SC level,
Fig. 5. Contour plot of Li3’s electron density laplacian, in the plane of the nuclei, as descriptions in terms of Interstitial Orbitals apply to higher Li clus-
computed from the 15-configuration GMCSC wavefunction. Contours have been ters as well [5], broad aspects of its GMCSC generalization, as ob-
drawn at ±2  10n a05, ±410n a05, ±810n a05, for n = 3, 2, 1, and 0, as in [13].
tained for Li3, can be reasonably expected to also extend to
Dashed lines mark negative contours. The laplacian reaches arbitrarily large nega-
tive values sufficiently close to the nuclear cusps, at which it is of course undefined. higher clusters.
Li3’s electron density has been found to exhibit, in agreement
with previous RHF results [11], a non-nuclear attractor, coupled
with features that may be viewed as connected with the appear-
ance of Interstitial Orbitals in the electronic wavefunction. A topo-
logical ‘‘Atoms-in-Molecules” analysis of the electron density
supports the contention that the three nuclei are indirectly bound
to one another through the non-nuclear attractor, in agreement
with findings for higher clusters [13].

Fig. 6. Contour plot of Li3’s electron density difference (molecule minus atoms, see
text), in the plane of the nuclei. Contours have been drawn every 0.001 a03. Dashed
lines mark negative contours.

Fig. 7. Contour plot of Li2’s electron density difference (molecule minus atoms, see
internuclear distance of 5.051 a0 [41], as obtained from a nine-con- text), in a plane containing the internuclear axis. The nuclei are at z = ±2.5255 a0.
figuration GMCSC molecular wavefunction [23] and single-config- Contours have been drawn every 0.002 a03 (note this is twice the interval of Fig. 5).
uration SC atomic wavefunctions. Dashed lines mark negative contours.
26 F.E. Penotti / Journal of Molecular Structure: THEOCHEM 861 (2008) 18–26

While this triatomic belongs to the very select group of mole- [6] J. Gerratt, Adv. Atomic Mol. Phys. 7 (1971) 141.
[7] N.C. Pyper, J. Gerratt, Proc. R. Soc. Lond. A 355 (1977) 407.
cules displaying non-nuclear electron-density maxima [12], it is
[8] J. Gerratt, M. Raimondi, Proc. R. Soc. Lond. A 371 (1980) 525.
perhaps not quite on a par with Li2 and most of the higher Li clus- [9] D.L. Cooper, J. Gerratt, M. Raimondi, Chem. Rev. 91 (1991) 929.
ters. This is because the triatomic’s ‘floppy’ nature implies the non- [10] J. Gerratt, D.L. Cooper, P.B. Karadakov, M. Raimondi, in: S. Wilson (Ed.),
nuclear maximum would presumably be ‘smeared out’ by the Handbook of Molecular Physics and Quantum Chemistry, John Wiley,
Chichester, 2003 (Part 2, Chapter 12).
pseudo-rotational motion that, even in the rovibrational ground [11] R. K. Owen, Ph.D. Thesis, University of California, 1990 (available at http://
state, makes it quickly interconvert between all three equivalent owen.sj.ca.us/rk/rkowen/qmc/).
C2v minimum-energy geometries [1–3]. It is not immediately clear [12] W.L. Cao, C. Gatti, P.J. MacDougall, R.F.W. Bader, Chem. Phys. Lett. 141 (1987) 380.
[13] C. Gatti, P. Fantucci, G. Pacchioni, Theor. Chim. Acta 72 (1987) 433.
whether such motion would erase any experimentally observable [14] S. Besnainou, M. Roux, R. Daudel, Compt. Rend. Acad. Sci. 241 (1955) 311.
trace of the non-nuclear maximum, or perhaps leave a threefold- [15] J. Cioslowski, J. Phys. Chem. 94 (1990) 5496.
symmetry ‘ridge’ surrounding the molecular centre of mass. Ascer- [16] R. Glaser, R.F. Waldron, K.B. Wiberg, J. Phys. Chem. 94 (1990) 7357.
[17] K.E. Edgecombe, R.O. Esquivel, V.H. Smith Jr., F. Müller-Plathe, J. Chem. Phys.
taining this would presumably require computing the electron 97 (1992) 2593.
density over the energy-accessible range of nuclear geometries, [18] G.I. Bersuker, C. Peng, J.E. Boggs, J. Phys. Chem. 97 (1993) 9323.
and then carrying out a partial averaging, at a given (arbitrary) ori- [19] R.F.W. Bader, H. Essén, J. Chem. Phys. 80 (1984) 1943.
[20] F.E. Penotti, Int. J. Quant. Chem. 46 (1993) 535.
entation of the molecule-fixed reference frame, over the (squared [21] F.E. Penotti, Int. J. Quant. Chem. 59 (1996) 349.
modulus of) the ground-state rovibrational wavefunction. [22] M. Kotani, A. Amemiya, E. Ishiguro, T. Kimura, Tables of Molecular Integrals,
Unsurprisingly, the molecule’s electron density difference map Maruzen, Tokyo, 1955 (Chapter 1).
[23] F.E. Penotti, J. Comput. Chem. 27 (2006) 762.
(molecule minus atoms) also shows a non-nuclear maximum not
[24] S.M. Goldfeld, R.E. Quandt, H.F. Trotter, Econometrica 34 (1966) 541.
far from that in the molecular density proper. Perhaps more inter- [25] E.R. Cohen, B.N. Taylor, J. Phys. Chem. Ref. Data 17 (1988) 1795.
estingly, it shows 3p-type enrichment-depletion patterns, whose [26] S.E. Wheeler, K.W. Sattelmeyer, P.v.R. Schleyer, H.F. Schaefer III, J. Chem. Phys.
orientation does not appear incompatible with the contention of 120 (2004) 4683.
[27] J. Fernandez Rico, R. López, A. Aguado, I. Ema, G. Ramírez, J. Comp. Chem. 19
indirect ‘‘four-centre” bonding. Similar patterns also occur in the (1998) 1284.
diatomic, and have been reported for higher Li clusters [42]. [28] J. Fernandez Rico, R. López, A. Aguado, I. Ema, G. Ramírez, Int. J. Quant. Chem.
At the C2v minimum-energy geometry, the molecule’s electric 81 (2000) 148.
[29] J. Fernandez Rico, R. López, A. Aguado, I. Ema, G. Ramírez, J. Comp. Chem 25
dipole moment is predicted to be 1.00 D, although with a very (2004) 1987.
uncertain error bracket. It would be very interesting to see whether [30] H.-J. Werner, P.J. Knowles, M. Schütz, R. Lindh, P. Celani, T. Korona, G. Rauhut,
the dipole moment’s apparent sensitivity to the inclusion of inner- R.D. Amos, A. Bernhardsson, A. Berning, D.L. Cooper, M.J.O. Deegan, A.J.
Dobbyn, F. Eckert, C. Hampel, G. Hetzer, A.W. Lloyd, S.J. Mc Nicholas, F.R.
shell ‘out-of-plane’ correlation could be confirmed, or disputed, by Manby, W. Meyer, M.E. Mura, A. Nicklaß, P. Palmieri, R. Pitzer, U. Schumann, H.
independent high-accuracy calculations, perhaps using orthogo- Stoll. A.J. Stone, R. Tarroni, T. Thorsteinsson, MOLPRO, a Package of Ab initio
nal-orbital methods. Programs Designed by H.-J. Werner and P.J. Knowles, Available online at:
<http://www.molpro.net>.
[31] F.E. Penotti, D.L. Cooper (Eds.), Valence Bond Theory, Theoretical and
Acknowledgments Computational Chemistry Series, vol. 10, Elsevier, Amsterdam, 2002 (Chapter
10).
[32] I. Ema, R. López, J.J. Fernández, A. Aguado, G. Ramírez, J.F. Rico, Int. J. Quant.
The present author wishes to express his sincerest thanks to J.F.
Chem. 108 (2008) 25.
Rico, R. López, A. Aguado, I. Ema, G. Ramírez and J.J. Fernández for [33] M.W. Schmidt, K.K. Baldridge, J.A. Boatz, S.T. Elbert, M.S. Gordon, J.J. Jensen, S.
making available to him their highly accurate and efficient programs Koseki, N. Matsunaga, N.A. Nguyen, S. Su, T.L. Windus, M. Dupuis, J.A.J.
[27–29,32] for the calculation of STO-based molecular integrals. Montgomery, Comput. Chem. 14 (1993) 1347.
[34] T.H. Dunning Jr, J. Chem. Phys. 90 (1989) 1007.
[35] Extensible Computational Chemistry Environment Basis Set Database, Version
References 02/25/04, Molecular Science Computing Facility, Environmental and Molecular
Sciences Laboratory, Pacific Northwest Laboratory (www.emsl.pnl.gov/forms/
[1] H.-G. Krämer, M. Keil, C.B. Suarez, W. Demtröder, W. Meyer, Chem. Phys. Lett. basisform.html).
299 (1999) 212. [36] A. Lüchow, H. Kleindienst, Int. J. Quant. Chem. 51 (1994) 211.
[2] M. Keil, H.-G. Krämer, A. Kudell, M.A. Baig, J. Zhu, W. Demtröder, W. Meyer, J. [37] C.H. Wu, J. Chem. Phys. 65 (1976) 3181.
Chem. Phys. 113 (2000) 7414. [38] F.E. Penotti, Int. J. Quant. Chem. 78 (2000) 24.
[3] W. Meyer, M. Keil, A. Kudell, M.A. Baig, J. Zhu, W. Demtröder, J. Chem. Phys. [39] B.A. Chirgwin, C.A. Coulson, Proc. R. Soc. Lond. A 201 (1950) 196.
115 (2001) 2590. [40] F.E. Penotti, Int. J. Quant. Chem. 78 (2000) 378.
[4] M.V. Berry, Proc. R. Soc. Lond. A 392 (1984) 45. [41] M.M. Hessel, C.R. Vidal, J. Chem. Phys. 70 (1979) 4439.
[5] E. Tornaghi, D.L. Cooper, J. Gerratt, M. Raimondi, M. Sironi, Theochem – J. Mol. [42] I. Boustani, W. Pewestorf, P. Fantucci, V. Bonačić-Koutecký, J. Koutecký, Phys.
Struc. 91 (1992) 383. Rev. B 35 (1987) 9437.

You might also like