Recent advances in lattice thermal conductivity calculation using machine-learning interatomic potentials

Abstract

The accuracy of the interatomic potential functions employed in molecular dynamics (MD) simulation is one of the most important challenges of this technique. In contrast, the high accuracy ab initio quantum simulation cannot be an alternative to MD due to its high computational cost. In the meantime, the machine learning approach has been able to compromise these two numerical techniques. This work unveils how the MD interatomic potentials have been improved through training over ab initio datasets and are able to well calculate phononic thermal transport of materials. Therefore, this powerful tool allows the quantum computational order accuracy with a timescale in the order of classical computations. Besides, the thermal conductivity of a few 2D and 3D structures, which have been calculated using machine learning interatomic potentials (MLIPs), is presented and compared with experimental and quantum counterparts. Finally, it is discussed that how MLIPs can be developed not only to estimate other properties of pristine materials, such as mechanical properties, but also to predict the properties of defective materials.

A Theoretical Investigation on the Physical Properties of Zirconium Trichalcogenides, ZrS3, ZrSe3 and ZrTe3 Monolayers

“…As is clear, due to the multifaceted structural and bonding effects, the ZrX 3 nanosheets showed highly anisotropic and complex mechanical behavior. Worthy to mention that complex material properties can be explored using MTPs with high accuracy and accelerated computational costs [37][38][39][40][41].…”

Predicting lattice thermal conductivity via machine learning: a mini review

“…In addition to such direct prediction of κ L , ML has been successfully used to build accurate interatomic potentials for MD simulations. Generally speaking, the machine learning potential (MLP) employs regression algorithm to determine the ab-initio potential energy surface (PES), and the atomic configurations are usually adopted as input features 17,18 . Recently, the MLPs have been utilized to accurately predict the κ L of systems with complex crystal structures and chemical compositions, such as the alloys 19 , the heterostructures 20 , and the molten salts 21 .…”

Atomistic modeling of the mechanical properties: the rise of machine learning interatomic potentials

“…In recent years, machine learning methods have been successfully employed to accelerate the evaluation of various materials properties. [9][10][11][12][13] One of the most prominent accomplishments of machine learning methods in materials science is undoubtedly related to machine learning interatomic potentials (MLIPs), introduced originally by Behler and Parrinello 14 in 2007. MLIPs belong to the nonparametric family of interatomic potentials, with the goal of providing DFT-level accuracy and flexibility with the computational efficiency of EIPs.…”

“…17 After the great accomplishments of MLIPs in the acceleration of materials design 19,20 and evaluation of various physical properties, [21][22][23][24][25] recently their application has been extended towards the analysis of mechanical and failure responses of nanomaterials, 4,26 outperforming both DFT and EIP counterparts. In the previous review papers, the successful employment of MLIPs in the analysis of diverse properties of molecular systems [27][28][29][30][31][32][33] and evaluation of thermal conductivity of crystalline structures 13,34 have been highlighted. Nonetheless, with respect to the examination of mechanical and failure properties, to the best of our knowledge there exists currently no literature review that highlights the accuracy, advantages, shortcomings and challenges of MLIPs in comparison with the conventional counterparts.…”