Non-genetic global optimization methods in molecular science: An overview

Global optimization methods that are alternative to evolutionary approaches are reviewed. Improvements over the well-known simulated annealing algorithm are discussed. Several finite-temperature sampling methods, including parallel tempering Monte Carlo and flat-histogram approaches, can also be used efficiently to locate low-energy structures. Methods based on deformation of the potential energy surface, as well as dimensional strategies and other tricks, are briefly presented. Some illustrations are provided for different molecular clusters, a simple glass-forming system, as well as a polypeptide.