Accelerated Lattice Thermal Conductivity calculations for (\alpha-CX) (X=N, P, As) using a combination of Density Functional Theory, Machine Learning, and Molecular Dynamics
Muhammad Shaquille Hisham (a*), Nisrina Nindya (b), Irfan Dwi Aditya (a), Sparisoma Viridi (a), Nadya Amalia (c), Sasfan Arman Wella (c)

a) Department of Physics, Bandung Institute of Technology, Bandung, Indonesia
*10221055[at]mahasiswa.itb.ac.id
b) Department of Physics, Sepuluh Nopember Institute of Technology, Surabaya, Indonesia
c) Quantum Physics Research Center, BRIN, Serpong, Indonesia

Abstract

The growing potential of carbon pnictide (\alpha-CX) monolayers in optoelectronic and photovoltaic applications has sparked interest in their thermoelectric properties. One of the crucial properties in thermoelectric device performance is lattice thermal conductivity (LTC) which reflects how well the heat is conducted through phonons. However, accurate LTC calculations using conventional method such as density functional theory (DFT) are highly time consuming. To address this, a machine learning (ML) based approach was implemented to generate force fields that can be used in molecular dynamics (MD) simulations, which significantly reduces computational cost and data requirements. LTC was calculated for (\alpha-CX) using partial training data and a dataset consisting of different dimensions which consist of 20 random displacements. The result shows that for (\alpha-CX) ML models produced accurate LTC values while halving the data requirements and achieved up to six times faster calculations. This demonstrates the efficiency and reliability for thermoelectric property predictions in materials using ML based approach.

Keywords: (\alpha-CX) monolayers, density functional theory, machine learning, molecular dynamics, thermoelectric

Topic: Material Physics