A SIMPLE METHOD FOR PREDICTING ELECTRON DENSITIES IN COMPLEX AND FLEXIBLE MOLECULES. APPLICATIONS TO CONFORMATIONAL ANALYSIS. Pierre J. Becker, Emmanuel Bec and Jean Michel Gillet. Laboratoire Structure Electronique et Modelisation Ecole Centrale Paris, Grande Voie des Vignes, 92295 Chatenay Malabry Cedex, France

A simple method is introduced to decompose the electronic density of a molecule into the sum of fragments. The method is based on the Hirshfeld's partitioning scheme, which is very general and independent on any basis set.

For flexible molecules, one can verify that the density of the fragments linked to a single bond is invariant (to an accuracy of the order of 0.01eÅ^-3) through rotations around the single bond. The method has been carefully checked for alkanes, acetone, urea, water dimers, alanine and glycylglycine. The invariance of the density of fragments on each side of the peptide bond is well fulfilled.

Moreover, it turns out that some fragment densities can be transferred from one molecule to a bigger one within the same limit of accuracy. This allows for the possibility of predicting electron density of complex molecules that cannot be calculated by quantum mechanical methods.

The method is compared with the Mulliken's partitioning used in recent similar studies by Mezey. Within a given molecule, the two methods lead to similar conclusions for various conformations. However, transferability, which is the key for efficient use of such schemes, is significantly better with our present approach .

It is also possible to estimate the variations of energy with changes of conformational parameters, within a Density Functional approach. This simulated energy surface is thus only dependent on the charge density. Preliminary calculations seem quite encouraging. This may be of interest for Molecular Dynamics, since such methods rely totally on the assumed energy surface. The method we propose is also quite adapted for critically reducing the number of leading conformational parameters in complex molecules.

We also considered polarisation effects in fragments. For exemple, for simple molecules, we study the evolution of fragments when changing interatomic distances. Simple and general trends appear, that can be empirically modelled and related to polarisabilities or vibrational parameters.

The method can be used either with theoretical or experimental densities.