Phonon thermal transport in Bi2Te3/Sb2Te3 monolayer superlattices: a neural network potential study

Abstract

Superlattices are significant means to reduce the lattice thermal conductivity of thermoelectric materials and optimize their performance. In this work, using high-precision first-principles based neural network potentials combined with non-equilibrium molecular dynamics simulations and the phonon Boltzmann transport equation, the lattice thermal conductivities of Bi₂Te₃ monolayer and lateral Bi₂Te₃/Sb₂Te₃ monolayer superlattices are thoroughly investigated. As the period length increases, the thermal conductivity shows a trend of an initial decrease followed by an increase, which aligns with conventional observations. However, it is found that the decrease of thermal conductivity in coherent short-period superlattices is attributed to the variation of phonon scattering behaviors, including the increased phonon anharmonicity due to enhanced structural complexity and symmetry breaking, as well as intensified scattering between acoustic and optical phonons caused by the downward shift of low-frequency optical branches. This is contrary to expectations that the phonon velocity is the primary driver of reduced cross-plane lattice thermal conductivity in coherent short-period superlattices. In addition, it is found that the four-phonon (4ph) scattering process plays a significant role in the thermal conductivities of the Bi₂Te₃ monolayer and 1–1 and 3–3 Bi₂Te₃/Sb₂Te₃ monolayer superlattices. After the 4ph scattering process is considered, the a-axis thermal conductivities of these three structures decrease by 27.8%, 26.4%, and 31.1%, respectively. This study not only advances our knowledge of phonon thermal transport behavior in superlattices, but also serves as a good research case for investigating the thermal conductivity of complex superlattice structures using accurate and efficient neural network potentials.