Orientation-Tunable Interfacial Thermal Conductance in Lateral MoSe₂/MoS₂ Heterostructures: A Deep Potential Molecular Dynamics Study

Managing heat flow across interfaces in lateral 2D transition-metal dichalcogenide (TMD) heterostructures is a key challenge for next-generation nanoelectronic and optoelectronic devices.To tackle this, we trained a deep learning interatomic potential (DLP) using DeePMD-kit on density functional theory (DFT) data, then paired it with non-equilibrium molecular dynamics (NEMD) simulations to study interfacial thermal conductance (ITC) in lateral MoSe₂/MoS₂ heterostructures -looking specifically at both armchair and zigzag interface orientations. The potential we trained closely matches DFT energies, forces, and phonon dispersion relations, which confirms its reliability for thermal transport simulations. At room temperature, the armchair interface shows notably higher ITC (218 MW/m²•K) compared to the zigzag interface (188 MW/m²•K), and both increase further as temperature rises. Looking at the vibrational density of states, we found greater phonon spectral overlap at the armchair interface -this is what drives its better thermal transport performance. Together, these results offer a solid machine-learning framework for studying heat transport in lateral TMD heterostructures, and identify interface orientation as an effective means of controlling heat dissipation at these junctions.