2020/02/03 by Martín Leandro Paleico, Jörg Behler, Paleico, Martín Leandro +1
Materials Science · Computer Science · #Machine Learning in Materials Science #Optimization and Search Problems #Catalytic Processes in Materials Science
paper · pdf · doi:10.48550/arxiv.2002.00716
Global optimization is an active area of research in atomistic simulations,\nand many algorithms have been proposed to date. A prominent example is basin\nhopping Monte Carlo, which performs a modified Metropolis Monte Carlo search to\nexplore the potential energy surface of the system of interest. These\nsimulations can be very demanding due to the high-dimensional configurational\nsearch space. The effective search space can be reduced by utilizing grids for\nthe atomic positions, but at the cost of possibly biasing the results if fixed\ngrids are employed. In this paper, we present a flexible grid algorithm for\nglobal optimization that allows to exploit the efficiency of grids without\nbiasing the simulation outcome. The method is general and applicable to very\nheterogeneous systems, such as interfaces between two materials of different\ncrystal structure or large clusters supported at surfaces. As a benchmark case,\nwe demonstrate its performance for the well-known global optimization problem\nof Lennard-Jones clusters containing up to 100 particles. In spite of the\nsimplicity of this model potential, Lennard-Jones clusters represent a\nchallenging test case, since the global minima for some "magic" numbers of\nparticles exhibit geometries that are very different from those of clusters\nwith only a slightly different size.\n