Stochastic construction of reaction paths: a genetic algorithm-based approach

Abstract

A scheme for the stochastic construction of reaction paths is proposed. The scheme exploits the genetic algorithm (GA) for locating the reactant (R) and product (P) minima as well as the appropriate first-order saddle point (SP) on the potential energy surface (PES) relevant to the chemical transformation. The reaction path is defined to be the trajectory traced out by the string of highest fitness as it evolves genetically toward the appropriate first-order saddle point under a specific constraint, starting either from R or P. The fitness contains information about the gradient norm, the square of an energy differential, and may or may not contain information about local curvature. The workability of the scheme is demonstrated with (i) two-model two-dimensional surfaces and (ii) a cluster of 14 argon atoms undergoing a conformational change. Salient features of the proposed method are discussed.