High-performance 4H–silicon carbide (4H–SiC) power metal–oxide–semiconductor field-effect transistors (MOSFETs) are central to next-generation power conversion, but their design remains constrained by the fundamental trade-off between breakdown voltage (BV) and on-state resistance. Conventional optimization strategies typically rely on Monte Carlo analysis or brute-force parameter sweeps over a small set of geometric or lumped doping parameters, which limits exploration of the full two-dimensional doping design space and becomes prohibitively expensive as the number of design variables increases. This work presents a large-scale, adjoint-based multi-objective optimization framework for a 1.2 kV vertical 4H–SiC MOSFET, in which the net doping concentration at every finite-element node is treated as an independent design variable and the sensitivities of BV and on-state current (ION) with respect to these variables are computed efficiently within a drift–diffusion and Poisson simulator incorporating wide-bandgap-specific physics. The adjoint formulation yields full spatial maps of the doping sensitivity functions for BV and ION by solving only one additional sparse linear system per figure of merit, making the effective cost of gradient evaluation essentially independent of the number of design variables and enabling gradient-driven updates of thousands of coupled doping degrees of freedom. For the 4H–SiC MOSFET considered, the proposed framework achieves an approximate 11% increase in BV and a 3%–4% increase in ION relative to the original device without any geometric changes, while offering a physically interpretable link between sensitivity maps and process-relevant doping adjustments. The results demonstrate that adjoint-based large-scale optimization provides a computationally efficient and experimentally meaningful alternative to heuristic approaches for systematic co-optimization of breakdown and on-state performance in wide-bandgap power MOSFETs.
Brungi et al. (Sun,) studied this question.