A Response Embedded Atom Method of Interatomic Potentials

TL;DR

This paper introduces a response embedded atom method (R-EAM) that improves interatomic potentials for metals by addressing elastic constant constraints and surface relaxation inaccuracies, aligning better with quantum mechanics at a modest computational cost.

Contribution

The R-EAM explicitly incorporates environment-dependent electron distribution, overcoming key limitations of traditional EAM potentials in modeling elastic and surface properties.

Findings

R-EAM does not impose the elastic constant constraints of EAM.

R-EAM accurately predicts surface relaxations and reconstructions.

R-EAM requires about twice the computational power of EAM.

Abstract

The embedded atom method (EAM) potentials are probably the most widely used interatomic potentials for metals and alloys. However, the EAM potentials impose three constraints on elastic constants that are inconsistent with experiments. At a more subtle (but more important) level, the EAM potentials often incorrectly describe the outward/inward relaxation of surface layers, and therefore will not reliably describe nanostructures. This Letter reports a response EAM (R-EAM) that addresses both issues. Conceptually, the electron distribution from each atom does not respond to the atoms environment within the EAM. In reality, the electron distribution from each atom depends on the atoms environment, and this dependence is explicitly incorporated in the R-EAM. Analytical derivation shows that the R-EAM potentials do not impose these three constraints on elastic constants that EAM potentials do. Further, taking hexagonal close packed (HCP) metals Ti, Mg, and Zn as the prototypes, the authors show that the R-EAM potentials correctly describe surfaces, in terms of interlayer spacing and surface reconstruction, in agreement with quantum mechanics calculations, while EAM potentials are not in agreement. In comparison to EAM potentials, the R-EAM potentials require only approximately twice the computational power.