Evaluation of 4H-SiC MOSFET transfer characteristics using machine-learning techniques

The performance of 4H silicon carbide (SiC) MOSFETs critically depends on the quality of the SiC/silicon oxide interface, which typically contains a high density of interface traps. To solve this problem, fast and reliable characterization methods are required. The commonly used evaluation schemes for 3-terminal transfer characteristics, however, neglect the presence of interface traps. Here, a method based on machine-learning techniques is presented which extracts reliable performance parameters from transfer characteristics of 4H-SiC MOSFETs including a quantitative estimate of the density of interface traps. This method is successfully validated by comparison with Hall-effect measurements and applied to various MOSFET types.

Automated summary

The performance of 4H silicon carbide (SiC) MOSFETs relies on the quality of the SiC/silicon oxide interface, which often has a high density of interface traps. To address this issue, a fast and reliable characterization method is introduced using machine-learning techniques. This method extracts accurate performance parameters, including a quantitative estimate of the interface trap density, from the transfer characteristics of 4H-SiC MOSFETs. It has been successfully validated against Hall-effect measurements and applied to different types of MOSFETs.