Tuning the lattice thermal conductivity of Janus SnSSe by interlayer twisting: a machine-learning-based study

Twisted two-dimensional materials have recently attracted tremendous interest owing to their unique structures and fantastic electronic properties. However, the effect of interlayer twisting on the phonon transport properties is less known, especially for the twist-angle-dependent lattice thermal conductivity ( κ L ). Using the emerging Janus SnSSe bilayer as a prototypical example, we develop an accurate machine learning potential, which is adopted to efficiently predict the κ L at a series of twist angles via iterative solution of the Boltzmann transport equation. It is found that the κ L exhibits a distinct non-monotonous dependence on the twist angle, which can be traced back to the bonding heterogeneity between high-symmetry stacking regions inside the moiré unit cell. In contrast to the general belief, the optical phonons make a major contribution toward the κ L of the twisted structures. Moreover, we demonstrate that four-phonon scattering can significantly reduce the κ L of SnSSe bilayer at higher temperatures, which becomes more pronounced by interlayer twisting. Our work not only highlights the strong predictive power of machine learning potential, but also offers new insights into the design of thermal smart materials with tunable κ L .