A simulated annealing based technique for locating first-order saddle points on multidimensional surfaces and constructing reaction paths: several model studies

Abstract

We explore the workability of a simulated annealing based method for locating first-order saddle points (SP) on a potential energy surface, starting from a minimum. The search can proceed sequentially, exhausting all the saddle points of index one. It can also be tuned to trace the lowest first-order SP in one run. Four different model surfaces are investigated and the lowest first-order saddle point lying between two local minima are located in each case. Possible use of the technique in the construction of reaction paths is considered. A more realistic system involving six argon atoms interacting via Lennard-Jones potential (an 18 parameter problem) is investigated and the saddle point on the path of transformation of Ar₆ cluster from its global minimum configuration to a neighbouring local minimum is traced out. The computational labour in the SAM-based search is examined vis-a-vis the traditionally used methods.