Thermal interface doping strategies based on Bayesian optimization

Abstract

Enhancing interfacial thermal conductance (ITC) plays a crucial role in microelectronics cooling because the continued reduction of the device size and increased power intensity demand more efficient heat dissipation. Although light atom doping and interface roughness were reported to enhance ITC in certain cases, they could also reduce ITC depending on the specific interface configurations. By leveraging Bayesian optimization, we provide the practical guidelines to achieve enhancement of ITC with high confidence via a proper selection of dopant mass and/or concentration, regardless of actual interfacial configurations. We demonstrate that the data-driven approach could be a helpful tool for interface design with desired ITCs.