Geometry-Mediated Current Redistribution as a Testable Path for SiC Trench MOSFET Short-Circuit Robustness

Silicon carbide trench MOSFET design presents a difficult asymmetry. The low resistance and strong current capability desirable during normal operation can become damaging during short-circuit stress, when high current and high drain voltage produce rapid power dissipation, localized heating, electrothermal feedback, and eventual failure. This exploratory investigation considers a bounded possibility: whether modest changes to JFET-region or trench geometry could alter early current-path formation sufficiently to reduce localized electrothermal concentration without imposing an unacceptable normal-operation penalty. The proposed mechanism is not strong current suppression and is not intended to prevent thermal runaway once established. It is more limited: geometry-mediated redistribution may broaden current participation during the early phase of a short-circuit event, reducing the concentration of current density, power density, or accumulated local energy density before nonlinear thermal amplification becomes dominant. The central methodological requirement is to distinguish genuine spatial redistribution from the simpler effect of merely reducing total current. A geometry candidate should therefore be compared not only with the original device, but with a current-matched baseline in which an ordinary operating control is adjusted so that its drain-current waveform closely matches that of the geometry candidate. If the geometry candidate retains a local electrothermal advantage at closely matched electrical loading, that would support the geometry-mediated redistribution interpretation rather than establish a geometry-only causal effect. The exploratory comparison actually implemented was narrower than this proposed protocol and matched the candidate to the baseline only on the reported maximum-current quantity. Subsequent exploratory TCAD analysis provided preliminary support for the narrower redistribution interpretation. With total device width and drain voltage held constant, the candidate gate voltage was adjusted to match its reported maximum current to the baseline, and differences in local peak current density remained at depths selected after inspection for the greatest inter-case separation. The present evidence provides preliminary numerical support at the level of local current-density behavior; electrothermal behavior, short-circuit safe operating area, manufacturability, and experimental device performance remain subjects for subsequent validation. The present contribution is therefore a physically bounded hypothesis, a discriminating test, and preliminary numerical evidence: a local current-density difference remained after maximum-current matching, consistent with the geometry-mediated redistribution hypothesis.