Manybody Interaction Potentials Using Symbolic Regression Method

The results of modeling a system of interacting atoms using classical molecular dynamics (MD) methods directly depend on the form of interaction potential used. There are many approaches to obtaining a tabular potential function for conducting MD simulations, however, for the study of multi-scale systems with ultra-high macroscopic parameters, using the potentials obtained in this way is a difficult task. The symbolic regression method based on a genetic programming algorithm can seriously simplify this problem. Thanks to this, it is possible to construct a mathematical form of potential consisting of analytical functions, which can be used in further studies of the system and accelerate the processes of studying the properties of matter. In this work, analytical forms of interatomic interaction potentials were obtained using the symbolic regression method with an evolutionary search implemented using a genetic programming algorithm for iron systems at ultrahigh temperatures and pressures.