A digital twin for predicting the solder degradation lifetime of a GaN eHEMT integrated power module under power cycling conditions

Novel packaging designs eliminate bond wires in GaN eHEMTs power modules for high-frequency applications, making solders another critical packaging failure point due to thermomechanical fatigue. This study presents a digital twin (DT) to predict the power module lifetime by considering solder degradation in power cycling (PC) tests. An in-house PC test with direct temperature measurement followed by an industrial PC test for DT validation was conducted. The power module failed when the junction temperature swing increased by 20%, caused by solder degradation coupled with device self-heating effects. The digital modeling process addressed the multiphysics and multiscale challenges, wherein the electrothermal simulation demonstrates the coupling model effectively captures the power loss increment phenomenon, while also considering different time scales in two domains. By embedding temperature-dependent power loss into the thermomechanical model, replacing the physical device model with a reduced order lumped thermal network, and modeling the thin PCB features as layers, the DT’s computational speed improved over 20 times. A model decomposition method extracted the viscoplastic dissipation density, which is incorporated into Morrow’s Model to predict power module lifetime by transforming failure criteria into a failure threshold for solder degradation ratio. The DT’s applicability and fidelity were validated experimentally, demonstrating a lifetime prediction error of 1.07% and a post-failure junction temperature prediction deviation of 4.4%. The DT offers insights into solder behavior during PC unattainable through direct measurement and servers as a valuable tool for GaN packaging design, facilitating PC performance optimization while minimizing time and cost-intensive prototype fabrication and testing. • Proposing a digital twin for predicting power cycling lifetime via solder degradation. • Considering multiphysics electro-thermo-mechanical coupling and multi-scale issues. • A lumped thermal network and PCB simplification accelerate computation over 20 times. • The solder submodel enables accurate viscoplastic dissipation and fatigue analysis. • The digital twin predicts the power cycling lifetime within 1.1% error.