Transiting the molecular potential energy surface along low energy pathways: the TRREAT algorithm

The Transition Rapidly exploring Random Eigenvector Assisted Tree (TRREAT) algorithm is introduced to perform searches along low curvature pathways on a potential energy surface (PES). The method combines local curvature information about the PES with an iterative Rapidly exploring Random Tree algorithm (LaValle, Computer Science Department, Iowa State University, 1998, TR98-11) that quickly searches high-dimensional spaces for feasible pathways between local minima. Herein, the method is applied to identifying conformational changes of molecular systems using Cartesian coordinates while avoiding a priori definition of collective variables. We analyze the pathway identification problem for alanine dipeptide, cyclohexane and glycine using nonreactive and reactive forcefields. We show how TRREAT-identified pathways can be used as valuable input guesses for double-ended methods such as the Nudged Elastic Band when ascertaining transition state energies. This method can be utilized to improve/extend the reaction databases that lie at the core of automatic chemical reaction mechanism generator software currently developed to build kinetic models of chemical reactions.