Genetic Programming for Multi-Timescale Modeling

TL;DR

This paper introduces a genetic programming approach to efficiently model multi-timescale dynamics by predicting potential energy barriers from limited data, enabling faster kinetic simulations without explicit barrier calculations.

Contribution

It presents a novel use of genetic programming to regress potential energy barriers from sparse data, reducing computational costs in multi-timescale modeling.

Findings

GP regression predicts barriers within 0.1-1% error using less than 3% of data.

Significant reduction in CPU time for kinetic Monte Carlo simulations.

Method successfully applied to vacancy-assisted migration in binary alloys.

Abstract

A bottleneck for multi-timescale dynamics is the computation of the potential energy surface (PES). We explore the use of genetic programming (GP) to symbolically regress a mapping of the saddle-point barriers from only a few calculated points via molecular dynamics, thereby avoiding explicit calculation of all the barriers. The GP-regressed barrier function enables use of kinetic Monte Carlo (KMC) to simulate real-time kinetics (seconds to hours) using realistic interactions. To illustrate, we apply a GP regression to vacancy-assisted migration on a surface of a binary alloy and predict the diffusion barriers within 0.1--1% error using 3% (or less) of the barriers, and discuss the significant reduction in CPU time.