Session: 09-01: Computational Methods in Micro/ Nanoscale Transport

Paper Number: 129692

129692 - Machine-Learning-Based Thermal Conductivity Prediction in Two-Dimensional TiS2/MoS2 Van Der Waals Heterostructures 

Abstract: 

Overheating is a significant issue impacting lithium-ion batteries (LIBs). During battery operation, as the ions flow through the battery's internal resistance during the charging or discharging process, heat is generated. Additional heat is also generated during exothermic chemical reactions occurring inside the battery. If not properly addressed, this increased heat generated in the battery leads to overheating, resulting in thermal runaway and causing catastrophic failure of the battery. Hence, exploring new materials and estimating their thermal conductivity is crucial to achieve performant and safe lithium-ion batteries. Two-dimensional (2D) materials and heterostructures display unique thermal characteristics in comparison to their bulk counterparts. These include the ballistic and hydrodynamic transport of phonons, which result in increased thermal conductivity within the materials. However, the accurate estimation of thermal conductivity in two-dimensional (2D) materials, particularly in 2D heterostructures, presents significant challenges for both computational and experimental methods. In this study, we propose a computationally efficient approach to investigate the thermal conductivity of 2D TiS₂/MoS₂ van der Waals heterostructures. Our approach utilizes machine-learning interatomic potentials (MLIPs) to predict the thermal conductivity of the heterostructure. This approach effectively describes intralayer interactions by utilizing moment tensor potentials (MTP) trained on computationally inexpensive density functional theory (DFT)-based datasets, while the interlayer van der Waals interactions are calibrated using the D3-dispersion correction method. We start by generating the interatomic potential parameters by preparing DFT-datasets at low (50 K and 100 K) and high (300 K and 600 K) temperatures for the TiS₂/MoS₂ heterostructures. We then generate moment tensor potential (MTP) parameters using the machine-learning interatomic potentials (MLIPs) technique. The MTP parameters are combined with the D3-dispersion correction method to account for interlayer interactions between the atoms of the TiS₂ and MoS₂ layers. By exclusively incorporating the missing dispersion contribution into the otherwise accurate MTP, this method will provide greater precision in predicting interlayer interactions compared to the conventional Lennard-Jones (LJ) potential. The potential parameters are then validated by comparing the thermal properties with existing data in the literature and conducting standalone DFT calculations. Finally, molecular dynamics (MD) simulations are conducted to determine the thermal conductivity of the TiS₂/MoS₂ heterostructures using the derived potential parameters. The results presented in this study enhance our understanding of thermal conductivity in van der Waals (vdW) heterostructures. The proposed methodology, which utilizes machine-learning interatomic potentials (MLIPs), offers an effective approach for exploring new electrode materials for lithium-ion batteries (LIBs) with improved thermal conductivity.

Presenting Author: Akhil Kunjikuttan Nair University of Toronto

Presenting Author Biography: Akhil is a PhD student in the Department of Mechanical and Industrial Engineering at the University of Toronto, Canada. His research is focused on applying Density Functional Theory (DFT) calculations and Molecular Dynamics (MD) simulations on 2D materials to describe its various properties. Further, he would be linking DFT and MD methods using Machine Learning algorithms for predicting necessary material properties which could potentially be useful for nanoscale battery modelling techniques in the future.

Akhil received his Master’s degree in Nanoscience and Technology from the Indian Institute of Technology Patna (2020) and his Bachelor’s degree in Mechanical Engineering from the Cochin University of Science and Technology, India (2017). His Master’s thesis was focused on tuning the magnetic properties of 2D materials using DFT calculations and Monte Carlo simulations for spintronic applications. His Bachelor’s project involved an automatic airbrake system using sensor fusion concepts.

Authors:

Akhil Kunjikuttan Nair University of Toronto
Carlos Manuel Da Silva University of Toronto
Cristina H. Amon University of Toronto