Machine-learning-based interatomic potential for phonon transport in perfect crystalline Si and crystalline Si with vacancies

Hasan Babaei, Ruiqiang Guo, Amirreza Hashemi, and Sangyeop Lee

Phys. Rev. Materials 3, 074603 – Published 26 July 2019

Abstract

We report that the single interatomic potential, developed using Gaussian regression of data from density functional theory calculations, has high accuracy and flexibility to describe phonon transport with ab initio accuracy in two different atomistic configurations: perfect crystalline Si and crystalline Si with vacancies. The high accuracies of second- and third-order force constants from the Gaussian approximation potential (GAP) are demonstrated with phonon dispersion, Grüneisen parameter, three-phonon scattering rate, phonon-vacancy scattering rate, and thermal conductivity, all of which are very close to the results from density functional theory calculations. We also show that the widely used empirical potentials (Stillinger-Weber and Tersoff) produce much larger errors compared to the GAP. The computational cost of GAP is higher than the two empirical potentials, but five orders of magnitude lower than density functional theory calculations. Our work shows that GAP can provide a new opportunity for studying phonon transport in partially disordered crystalline phases with the high predictive power of ab initio calculation but at a feasible computational cost.

Received 23 May 2019

DOI:https://doi.org/10.1103/PhysRevMaterials.3.074603

Physics Subject Headings (PhySH)

Anharmonic lattice dynamics Heat transfer Interatomic & molecular potentials Lattice dynamics Lattice thermal conductivity Nonequilibrium statistical mechanics Phonons Thermal conductivity Thermal properties Thermodynamics Transport phenomena

Boltzmann theory Density functional theory First-principles calculations Green's function methods Lattice-Boltzmann methods Machine learning Molecular dynamics

Atomic, Molecular & OpticalCondensed Matter, Materials & Applied PhysicsStatistical Physics & Thermodynamics

Authors & Affiliations

Hasan Babaei ^1,*, Ruiqiang Guo¹, Amirreza Hashemi², and Sangyeop Lee^1,3,†

¹Department of Mechanical Engineering and Materials Science, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, USA
²Department of Computational Modeling and Simulation, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, USA
³Department of Physics and Astronomy, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, USA

^*Present address: Department of Chemistry, University of California, Berkeley, California 94720, USA; hasan.babaei@berkeley.edu
^†sylee@pitt.edu