Nanoscale electro-thermal interactions in AlGaN/GaN high electron mobility transistors

TLDR: In this article, the authors examined self-heating in GaN-on-Si HEMTs via measurements of channel temperature using above-bandgap UV thermoreflectance imaging in combination with fully coupled electrothermal modeling.

Abstract: Self-heating in AlGaN/GaN high electron mobility transistors (HEMTs) negatively impacts device performance and reliability. Under nominal operating conditions, a hot-spot in the device channel develops under the drain side corner of the gate due to a concentration of volumetric heat generation leading to nonequilibrium carrier interactions and non-Fourier heat conduction. These subcontinuum effects obscure identification of the most salient processes impacting heating. In response, we examine self-heating in GaN-on-Si HEMTs via measurements of channel temperature using above-bandgap UV thermoreflectance imaging in combination with fully coupled electrothermal modeling. The methods together highlight the interplay of heat concentration and subcontinuum thermal transport showing that channel temperature cannot be determined solely by continuum scale heat transfer principles. Under conditions of equal power dissipation (PDISS = VDS × IDS = 250 mW), for example, a higher VDS bias (∼23 V) resulted in an ∼44% larger rise in peak junction temperature compared to that for a lower VDS (∼7.5 V) condition. The difference arises primarily due to reduction in the heat generating volume when operating under partially pinched-off (i.e., high VDS) conditions. Self-heating amplifies with this reduction as heating now takes place primarily over length scales less than the mean free path of the phonons tasked with energy dissipation. Being less efficient, the subcontinuum transport restricts thermal transport away from the device hot-spot causing a net increase in channel temperature. Taken together, even purely thermally driven device mean-time-to-failure is not, therefore, based on power dissipation alone as both bias dependence and subcontinuum thermal transport influence device lifetime.