ON THE USE OF QUANTUM ENERGY SURFACES IN THE DERIVATION OF MOLECULAR-FORCE FIELDS

This review summarizes the work in our laboratory dealing with the dev elopment of a new class of force fields for simulating molecular prope rties for a wide variety of practical applications. This new generatio n of force fields, termed Class II, is based on the derivation and par ameterization of analytic representations of quantum mechanical energy surfaces. The energy surface is derived from a set of representative species by systemically sampling distorted molecular structures and co mputing the corresponding quantum mechanical energy, energy gradient a nd Hessian matrices, i.e. the energy and its first and second derivati ves with respect to atomic positions. This development has a number of key implications for the development, refining and testing of molecul ar force fields. It has been found that molecular energy surfaces are characterized by significant anharmonic and coupling interactons. Anal ysis of hte quantum mechanical results indicates that these effects mu st be included in analytical representations of the force fields if ac curate structures, energies, and dynamic properties such as vibrationa l frequencies are to be obtained from molecular mechanics and dynamics calculations. In addition the inclusion of such terms greatly increas es the transferability of the force field. Such transferability is cri tical in predicting the properties of new species not included in the derivation of the force field, as is invariably the requirement in pra ctical research applications. In additon it has been shown that this m ethodology also has the pragmatic advantage of allowing for the deriva tion of a reasonable force field based on quantum mechanics for molecu les where little or no experimental data exist. In this report we revi ew the methodology of developing and testing the quantum mechanically based force fields, showing the derivation of the functional form of t he energy expession, sampling the quantum energy surface, testing agai nst explicit quantum results, scaling the force field to account for s ystematic errors, and testing against experiment. We focus simultaneou sly on criteria for establishing force field accuracy and transferabil ity. Where possible we describe parallel comparisons with the earlier diagonal-quadratic of Class I force fields. Finally we discuss deficie ncies remaining in the existing level of force field development and o utline how they may be addressed in future work.