Some new techniques for computing intermolecular interaction energies

University of Illinois, Urbana and Laboratoire Dynamique des Interactions moléculaires, Paris 6, France.

Abstract

Within the framework of symmetry-adapted perturbation theories the intermolecular interaction energy is generally evaluated by using the following truncated expansion

ΔE ≈ ΔE⁽¹⁾_elec + E⁽¹⁾_exch + E⁽²⁾_Disp+Ind (1)

where each quantity (electrostatic, first-order exchange and complete second-order Rayleigh-Schrödinger interaction energies respectively) is usually computed at the SCF level. Although these contributions are generally the most important ones in the region of equilibrium of the complex, it is known that missing contributions such as second-order exchange, intramolecular correlation and higher-order contributions are in general far from being negligible. A commonly used approach to improve (1) consists in mixing both perturbational and supermolecular approaches

ΔE ≈ ΔE^SCF + E⁽²⁾_Disp (2)

where Δ^ESCF denotes the SCF interaction energy. By using this formula some part of the second-order exchange induction contribution, and of the third- and higher- order "induction" energies as well as some intramonomer electron correlation contributions due to the self-consistency at the dimer level are recovered. However such an approach is still incomplete. Here we shall describe two very different techniques with a common objective: computing within the pure perturbational approach some of the missing contributions. First, a recently developed method of computing second-order exchange contributions within the framework of ab initio calculations will be presented(1). Numerical calculations for the interaction of some molecules ((H₂O)₂, (C₂H₄)₂, (NH₃)₂, etc...) will illustrate the importance of the second-order exchange dispersion contribution (not present in (2)). Second, we shall describe a quite different formalism based on a Monte Carlo technique to evaluate perturbational quantities(2). This new formalism has attractive features (no basis sets used, no need of approximate excited states wavefunctions, no problem of memory storage, etc...) and in particular enable us to compute intramolecular correlation and higher-order contributions in a quite natural way. Some practical calculations on small interacting dimers will be presented. Results illustrate the importance of intramolecular correlation contributions.