NEW PARALLEL CORRELATED ELECTRONIC STRUCTURE ALGORITHMS

DESCRIPTION (provided by applicant): This proposal aims to significantly enhance existing electronic structure methods appropriate for modeling biological systems. Electronic structure methods permit the modeling of molecular systems from first principles without any experimental input or empiricism at very high computational cost. They find wide use in modeling small reaction regions as well as data generation for development of simpler empirical methods. There is strong demand to increase the size of treatable systems and reduce the time to solution for systems at the current limit of feasibility. The opportunity is to leverage the greatly increased power and greatly reduced cost of workstation/PC clusters as platforms for high-performance electronic structure calculations. New parallel algorithms based on new approximations that are suitable for distributed computing will be developed for second order Moller-Plesset theory, which yields accurate descriptions of covalent bonds, torsional potentials, hydrogen bonding and dispersion interactions in biological systems. These algorithms will be directed towards energy evaluation, and analytical derivative methods, for characterizing structural parameters and potentials surfaces. The end-result aim is to increase the size of feasible calculations by a factor of 2 to 4, without loss of accuracy, and reduce time-to-solution by more than a factor of ten for large problems.