Analysis of accuracy in ab initio calculations of the static dipole polarizability components. Examples of the water molecule and hydroxide ion

Abstract

A comparison of ab initio calculations of the static dipole polarizability components for the water molecule and hydroxide anion is reported. The discussion is focused on the difference between the computational needs for an accurate treatment of these two systems. A particular attention is paid to the danger of extrapolating to the anion some approximate techniques that are valid for the water molecule. Four important points are considered in this analysis: the analytical/numerical calculation of the components, the effect of the electronic correlation, the characteristics of the basis set, and the vibrational effects. In the finite-field approach, the interdependence of some control parameters, such as the energy convergence threshold and the associated field strength, are recalled within the frame of parabolic or quartic polynomial fits. The correlation effect is examined within the coupled-cluster method restricted to single, double, and approximate triple excitations, CCSD(T), and the nth-order Møller-Plesset perturbation theory, MPn. The oscillating behaviour of the perturbation series is very pronounced for the anion case, and its convergence is doubtful. It is crucial to include much more diffuse functions (valence and polarization) in the basis set for the anion than for the water molecule. A comparison of many basis sets is presented. The role of the vibrational effect is briefly addressed. Because ab initio calculations on supersystems generally cannot be performed at the best level reachable for the monomer polarizabilities, lower (but reasonnable) accuracy of the polarizability components is also discussed.