A Practical Approach for Searching Stable Molecular Structures by Introducing Repulsive Interactions among Walkers

We propose a new algorithm that can be practically applied for searching molecular conformations without specifying the explicit reaction coordinates in advance. The basic idea is to introduce repulsive interactions among all the sampling structures (walkers) to spread the walker distributions. By increasing the strength of the repulsive interactions, walkers trapped in a local potential minimum overcome the potential barrier and can flow into other potential areas. Using this method, called the GLobal Area Search (GLAS) approach, all the walkers are optimized in parallel like a conventional geometrical optimization procedure, but in which only the repulsive force is added to the total gradient like an extra restraint term. We tested the GLAS method using simple molecules, i.e., ethane and glycine, and we could show that their conformers are properly generated. As the GLAS method has the following advantages: (i) computational time is the order of the geometrical optimizations, (ii) parallel exclusion is ...