Model selection in atomistic simulation

TL;DR

This paper applies statistical model selection to compare diverse atomistic simulation methods, aiding in choosing optimal approaches and guiding the development of new models that balance computational cost and accuracy.

Contribution

It introduces a framework for fair comparison of atomistic simulation methods using statistical model selection, demonstrated through a semiempirical hydrogen cluster model.

Findings

Statistical model selection effectively compares different simulation methods.

The semiempirical hydrogen cluster model balances cost and accuracy.

Framework guides development of improved atomistic models.

Abstract

There are many atomistic simulation methods with very different costs, accuracies, transferabilities, and numbers of empirical parameters. I show how statistical model selection can compare these methods fairly, even when they are very different. These comparisons are also useful for developing new methods that balance cost and accuracy. As an example, I build a semiempirical model for hydrogen clusters.