Machine learning boosted <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mrow><mml:mi>a</mml:mi><mml:mi>b</mml:mi></mml:mrow></mml:math> <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mrow><mml:mi>i</mml:mi><mml:mi>n</mml:mi><mml:mi>i</mml:mi><mml:mi>t</mml:mi><mml:mi>i</mml:mi><mml:mi>o</mml:mi></mml:mrow></mml:math> study of the thermal conductivity of Janus PtSTe van der Waals heterostructures
Lijun Pan, Jesús Carrete, Zhao Wang, Georg K. H. Madsen
Abstract
Calculating the thermal conductivity of heterostructures with multiple layers presents a significant challenge for state-of-the-art ab initio methods. In this study we introduce an efficient neural-network force field (NNFF) to explore the thermal transport characteristics of van der Waals heterostructures based on PtSTe, using both the phonon Boltzmann transport equation and molecular dynamics (MD) simulations. Besides demonstrating a remarkable level of agreement with both theoretical and experimental data, our predictions reveal that heterogeneous combinations like $\mathrm{PtSTe}\text{\ensuremath{-}}{\mathrm{PtTe}}_{2}$ display a notable reduction in thermal conductivity at room temperature, primarily due to broken out-of-plane symmetries and the presence of weak van der Waals interactions. Furthermore, our study highlights the superiority of MD simulations with NNFFs in capturing higher-order anharmonic phonon properties. This is demonstrated through the analysis of the temperature-dependent thermal conductivity curves of PtSTe-based van der Waals heterostructures and advances our understanding of phonon transport in those materials.