Optimization and analysis of HF-FOM and gate oxide reliability in 1200V split-gate 4H-SiC MOSFETs

The split-gate (SG) MOSFET architecture offers advantages over conventional MOSFETs with respect to gate charge and high-frequency figures of merit (HF-FOM). However, challenges remain concerning the gate oxide reliability of SG MOSFETs. This study introduces a multi-objective design methodology for SG MOSFETs by integrating Artificial Neural Networks (ANN) with the Non-dominated Sorting Genetic Algorithm II (NSGA-II). The proposed ANN model significantly reduces the computational time typically required by technology computer-aided design (TCAD) simulations for multi-objective optimization (MOO) of SG 4H-SiC MOSFET devices, which conventionally depend on resource-intensive three-dimensional numerical simulations. NSGA-II efficiently identifies optimal design variables. SG power MOSFETs were fabricated on a 6-inch wafer using a commercial foundry process, followed by experimental characterization. The performance of the optimized SG MOSFETs was compared with that of non-optimized SG planar power MOSFETs to quantify performance improvements and validate the effectiveness of the proposed ANN and NSGA-II optimization approach. Both experimental measurements and TCAD simulations demonstrate that the optimized SG MOSFETs exhibit a 2.3x enhancement in HF-FOM and a 21% improvement in gate oxide reliability without compromising specific on-resistance.