Nanoparticle-Assisted Phonon Transport Modulation in Si/Ge Heterostructures Using Neuroevolution Potential Machine Learning Models

Abstract Reducing the thermal boundary resistance (TBR) is critical to enhance the thermal management efficiency and optimize the performance of electronic and thermoelectric devices. In this study, we employed non-equilibrium molecular dynamics (NEMD) simulations using neuroevolution potential (NEP) machine learning models to investigate the impact of embedding nanoparticles in Si/Ge heterostructures on the TBR. Our results showed a significant reduction in the TBR. This was attributed to the enhanced phonon density of states matching via resonance, which promoted more efficient elastic phonon transport across the interface. However, this approach also led to a substantial increase in the bulk thermal resistance, highlighting a trade-off in which the overall heat dissipation is compromised. To address this, we investigated an alternative strategy in which a nanoparticle was positioned directly at the interface to modulate the interfacial modes, thereby improving the phonon transport efficiency without adversely affecting the bulk thermal properties. NEMD simulations validated this approach, showing a comparable TBR reduction, while mitigating the bulk thermal resistance increase observed with the resonance-based embedding method. This study offers valuable insights into resolving interfacial heat dissipation challenges and provides a balanced strategy for optimizing the thermal transport efficiency of nanoscale material systems.