Development of a General Purpose Geometry Optimizer for Large-Scale Molecular Systems

C. Gonzalez and T. Allison

Objective: To develop and implement efficient geometry optimization algorithms aimed at the characterization of potential energy surfaces of large-scale molecular systems.

Problem: Recent advances in the efficiency of computational methodologies used in the evaluation of molecular energies and derivatives have created a need for efficient geometry optimization techniques. This need is particularly evident as the computational chemistry community attacks problems of increasing size and complexity such as biomolecules and reactions in condensed phases. While some of these techniques have appeared recently in the literature, they are typically integrated into specific codes and are not portable to other packages, restricting their access to a limited population of the scientific community. In addition, careful evaluation of the current geometry optimization packages has pointed to a marked lack of "user-friendly" tools that can aid in the semi-automated search for stationary points on potential energy surfaces of large molecules. Most of the time scientists need skills comparable to those of an expert in geometry optimization to tackle these problems. Given the increasing popularity of computational chemistry software in the study of a large variety of chemical problems, it is therefore necessary to build the appropriate infrastructure that will allow scientists to characterize the corresponding potential energy surfaces with a minimum of effort.

Approach: To address these needs, a general-purpose program with efficient optimization algorithms tailored for large molecules has been created. The program, called TURBO-OPT, performs geometry optimizations using energy derivative information from a variety of computational chemistry codes through a simple interface that gathers the necessary information from the normal outputs generated by these programs. This feature allows scientists to perform geometry optimizations using a common software platform that interfaces to the different theoretical methodologies available in common quantum chemical and molecular mechanics programs. For the advanced user, TURBO-OPT offers the necessary machinery to test and validate the different geometry optimization algorithms available in the literature.