Alchemical molecular dynamics for inverse design

Abstract

We present a molecular dynamics (MD) implementation of an extended statistical mechanical ensemble that includes ‘alchemical’ degrees of freedom describing particle attributes as thermodynamic variables. We demonstrate the use of this alchemical MD method in inverse design simulations of particles interacting via the Oscillating Pair Potential (OPP) and the Lennard–Jones–Gauss potential (LJG) – two general, previously studied models for which phase diagrams are known. We show that alchemical MD can quickly and efficiently optimise pair potentials for target structures within a specified design space in the low-temperature regime, where internal energy adequately represents the features of the alchemical free energy landscape. We show that alchemical MD can be also used to inversely design pair potentials to achieve target materials properties (here, bulk modulus) directly, without explicit knowledge of the structure–property relationship. Alchemical MD can easily be generalised and applied to any target materials properties or structures and used with any differentiable interaction potential.

GRAPHICAL ABSTRACT

Keywords:

• Digital alchemy
• materials design
• inverse design
• self-assembly
• molecular dynamics

Acknowledgements

P. Z. thanks Julia Dshemuchadse for helpful discussion on an earlier version of the manuscript and Matthew Spellings for providing the structure identification code and training data set. J. C. P. thanks Joshua A. Anderson for guidance with the alchemostat's implementation in HOOMD-blue. GvA thanks Jens Glaser for discussions on Trotter factorisations.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This research was supported in part by the National Science Foundation, Division of Materials Research under Computational and Data-Enabled Science and Engineering (CDS&E) Award # DMR 1409620 (2014-2018) and DMR 1808342 (2019-2023), and also by a grant from the Simons Foundation (256297, to SCG). Support for computational resources and services were provided by Advanced Research Computing at the University of Michigan, Ann Arbor.