Temperature-dependent loss physics and thermal fatigue of GaN HEMTs and SiC MOSFETs under realistic energy-storage mission profiles
Year: 2026
Type: article
Abstract: The temperature-dependent loss physics and thermal fatigue of two 650 V wide-bandgap power switches of equal rating, a GaN HEMT and a SiC power MOSFET—are compared under realistic energy-storage operation. Datasheet-calibrated models of the on-resistance, switching energies, and output capacitance energy E oss = ∫ 0 V C oss ( v ) v d v were traced to their physical origins—phonon-limited two-dimensional electron-gas transport in the GaN heterostructure; channel, JFET, and drift resistances in the SiC device—and embedded in a bidirectional buck–boost converter driven by three annual (8760 h) mission profiles built entirely from measured data: hourly satellite irradiance (PVGIS-SARAH3, 2023), the real hourly load of the Turkish transmission system (2023), and 10 617 real electric-vehicle charging sessions (2019). The hourly junction-temperature trajectories were processed by rainflow counting, and a Coffin–Manson–Arrhenius fatigue model was calibrated to the published power-cycling data. Three results emerge. First, the efficiency-optimal switching frequency varies by more than a factor of four (20–93 kHz) for the same GaN device across the three duty cycles, while under near-idle grid duty, the efficiency gap between the technologies collapses to below 0.03 percentage points, leaving the ranking to secondary design constraints. Second, the lower losses, lower junction-to-case thermal resistance, and shallower on-resistance temperature coefficient of the GaN device compress its junction-temperature swings by up to 27 K at the spectrum extremes. Third, this compression yields 26-fold (EV) and 3-fold (solar) annual fatigue-damage advantages for GaN, whereas grid duty stresses both technologies equally. The framework uses only datasheet characteristics and open mission data.
Source: Physica Scripta
Author: Muhammet Arucu
Institution: Bandırma Onyedi Eylül University
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $2,400
OpenAlex ID: W7223336499