Design Optimization of Ultra-Wide Bandgap Power Diodes:
A Co-Design Approach to Understanding the Impact of UWBG Semiconductor Materials Properties on Power Device Performance

Ultra-Wide Bandgap (UWBG) semiconductors constitute the fourth generation of semiconductor science and technology, and as such are at the forefront of semiconductor research today. This class of materials is roughly defined as those semiconductors having a bandgap greater than that of GaN ([endif]--&gt; eV). The unipolar figure of merit (UFOM) of [endif]--&gt; is a useful metric for comparing between semiconductor materials for power device application, but tThe incorporation of UWBG devices into systems is not determined on a UFOM b UFOM basis. It i, but is instead driven by the system-level benefits derived by incorporation of a device at given operational parameters (e.g. switching frequency, hold-off voltage, current-carrying ability). The various UWBG materials are at different levels stages of development for devices such as PIiN vs Schottky diodes, so a direct comparison with state-of-the-art wide bandgap materials based on device I-V curves is misleading.<br/>In order to determine the system-level operational regimes for which various combinations of UWBG materials and diode architecture are preferred over more developed WBG devices, we have developed the optimization tool first reported in [1]. This optimizer takes various system operational parameters as input (reverse operating bias, forward current density, system frequency, etc.) and optimizes the width of the device drift region and the doping level to minimize system power dissipation. Effects particular to UWBG materials, such as incomplete ionization due to deep dopant energy depth, space-charge limited current, and doping- and device thickness-dependent breakdown fields, have been incorporated into the optimizer to more accurately model the performance of UWBG materials.<br/>For a given device type and material, a mobility model is used to determine carrier mobilities at a given temperature, doping level, and electric field. For some materials, this mobility model is computed from various scattering contributions, while for the rest of the materials empirical models available in the literature are used instead. The operational parameters of the diode in the system are also defined. These are the reverse operating bias, forward current density, switching frequency, duty cycle, and device operational temperature. The critical fields are tabulated for various values of doping concentration and device thickness using the method described in section IIB. The optimization program then minimizes power dissipation in the system by altering device drift region thickness and doping concentration subject to the constraints described in [1]. By iterating across all available system operating parameters, it is possible to develop a map in system operating space and determine which devices/materials produce the lowest power loss for any given system operational point based on materials properties.<br/>The resulting preferred device color-maps show a strong dependence on many of the material parameters used as input, including impact ionization coefficients, dopant activation energies, and carrier mobilities. We compare preferred device color-maps across a range of temperatures and forward current densities to show the effects of UWBG-specific effects incomplete ionization and space-charge limited conduction. Then we vary the material parameter inputs to show the sensitivity of the final color-maps on each material input. While these color-maps are always being updated with newly published results on UWBG materials, these sensitivity analyses can nonetheless give researchers a more educated insight on which materials show the greatest promise in any given region of operating space.<br/><br/>1. Flicker, J. and R. Kaplar, <i>Design Optimization of GaN Vertical Power Diodes and Comparison to Si and SiC</i>, in <i>2017 IEEE 5th Workshop on Wide Bandgap Power Devices and Applications (WiPDA)</i>. 2017, IEEE. p. 31-38.