Growing Potential Energy Surfaces: The Interface of Grow and GAMESS
Heather Netzloff, Iowa State University
Very accurate calculations of small molecules and molecular systems are now possible with ab initio quantum chemistry. Energy/gradient calculations are usually only performed at critical points (reactants, products, transition states, etc.); thus, the calculations are static in nature. In order to describe the dynamics of a system, information about the potential energy surface (PES) must be available. A main driving force is thus to design a dynamics method that is accurate (based on first principles or ab initio computations), computationally cost-effective, and generic to any molecular system. The Grow program uses a Shepard interpolation method to automatically construct an ab initio (AI) molecular potential energy surface (PES). The converged surface may then be used for classical trajectory simulations and, for molecules with four or fewer atoms, quantum dynamics. The aim of the Grow package is to produce a sufficiently accurate PES, based on the smallest possible number of ab initio calculations, by carefully selecting the location of the data points. The algorithm requires ab initio calculation of the energy, energy gradient, and second derivatives at possibly several hundred molecular configurations. The efficient interface of the GAMESS suite of electronic structure programs with Grow to facilitate these calculations, at any level of theory available in GAMESS will be discussed. This interface will be illustrated with the N+H2, C+H2, and O+H2 abstraction and exchange reactions. The interface and these test systems take advantage of one of GAMESS’ hallmark capabilities, the efficient computation of MCSCF wavefunctions and extends the capabilities of Grow from a purely single-determinant program to one that can handle the vast range of interesting multi-determinant and multi-state problems.
Abstract Author(s): Heather M. Netzloff, Michael A. Collins, and Mark S. Gordon