Rigorous Determination of the Subsurface Thermal Boundary Conductance Between Bulk‐Bonded GaN and Diamond Die

Abstract During operation, power electronics generate large heat fluxes, and therefore, controlling temperature is key to improving performance and preventing failure. To lower operation temperature in power electronics, substantial attention has focused on placing the semiconductor material, such as gallium nitride (GaN), in close proximity to a high thermal conductivity solid heat sink. Here, diamond is the intuitive selection, having an exceptionally high thermal conductivity (>2000 W/m · K); however, poor GaN‐diamond thermal boundary conductance (TBC) can ultimately limit performance. Here, we provide comprehensive characterization of a vertically‐integrated GaN‐diamond architecture, consisting of a 260 nm gold bond, using hyperspectral frequency domain thermoreflectance imaging. Thermal model parameter uncertainty is propagated through fitting using a Monte Carlo method, which results in a highly asymmetric distribution for the subsurface GaN‐diamond TBC. A three parameter fit that includes non‐uniformity in GaN's thermal conductivity substantiated by hyperspectral confocal Raman stress mapping, results in a GaN‐diamond TBC greater than 144 MW/m 2 K. Taken together, this work provides a rigorously determined value for the GaN‐diamond TBC in a vertical power device as well as a comprehensive methodology to approach subsurface TBC analysis.