Reduction of the Electric Field Strength by the Thick Undoped Layer in the (001) C-H Vertical Diamond MOSFETs

High-performance p-type power metal-oxide-semiconductor field-effect transistors (MOSFETs) comparable to n-type ones are essential in implementing high switching speed and high-power complementary circuits. With wide bandgap (<i>E</i>_g = 5.47 eV), diamond is suitable for p-type MOSFETs. As an electronic device material, diamond has outstanding potential. In particular, the breakdown electric field of diamond (<i>E</i>_c = 10 MV/cm) surpasses that of other wide-bandgap semiconductor materials, being three times as high as that of GaN (<i>E</i>_c = 3.3 MV/cm) and that of 4H-SiC (<i>E</i>_c = 2.8 MV/cm). For the power MOSFETs, a vertical structure is suitable for a thick drift layer which improves breakdown voltage (<i>V</i>_B). In vertical diamond MOSFETs, a layer stack structure with a trench is basic [1]. On a p⁺ substrate as a vertical current path, an undoped layer for reduction of electric field strength and an N-doped layer for current blocking are deposited. After the trench is formed from the upside surface to the p⁺ substrate, a regrown undoped layer is deposited to induce 2-dimensional hole gas (2DHG) and secure the sub-surface channel by hydrogen-terminated (C-H). Previous studies have reported a high breakdown voltage of 580 V with a p^- drift layer in vertical p-type diamond trench MOSFETs [2]. As another way to improve the <i>V</i>_B, the reduction of the electric field strength by thickening the undoped layer is proposed. In this study, we investigated the effects of an undoped layer on reducing electric field strength in a (001) p⁺ diamond substrate without p^- drift layers using a device simulation and measuring fabricated devices.<br/><br/>To estimate the electric field of the device off-state, we conducted a device simulation using SILVACO Atlas at <i>V</i>_GS= +40V and <i>V</i>_DS = -600 V [3]. The device is a vertical structure with a trench whose fundamental structure resembles the device fabricated in ref [1] (Fig.1 (e)). On the (001) p⁺ diamond substrate, we used a 1 μm thick undoped layer ([N]: ~ 6.0×10¹⁶ cm^-3, [B]: ~ 6.6×10¹⁵ cm^-3) and a 1 μm thick N-doped layer ([N]: ~ 4.0×10¹⁸ cm^-3). Including the trench sidewall and bottom of the trench surface, a 200 nm thick regrown undoped layer ([N]: ~ 1×10¹⁶ cm^-3, [B]: ~ 1×10¹⁵ cm^-3) was used as a conductive layer. As the device dimensions, the source-gate length (<i>L</i>_SG), the gate length (<i>L</i>_G), and the gate-trench length (<i>L</i>_GT) are 3 µm. The trench depth (<i>D</i>_T) is 3 µm and the gate-drain length (<i>L</i>_GD) is 6 µm. In the simulation model, the donor level of nitrogen was 1.7 eV, and the acceptor level of boron was 0.37 eV. The negative charge was set in the Al₂O₃ layer to reproduce the induction of 2DHG in the C-H channel region [4]. In the simulation results, the electric field concentration occurred at the boundary between the p⁺ substrate and the undoped layer, and the corner contacting the undoped layer, N-doped layer and the regrown undoped layer. The maximum electric field (<i>E</i>_max) was 8.8 MV/cm at the corner with the 1 μm thick undoped layer. To reduce the electric field strength, we also conducted a simulation with a model increasing the undoped layer to 4 μm thick. The electric field strength at the corner was reduced to 3.4 MV/cm at the boundary by increasing the undoped layer thickness. With the 4 μm thick undoped layer, the <i>E</i>_max was 4.6 MV/cm at the boundary on the p⁺ substrate. To verify the reduction of the electric field strength with an undoped layer in the simulation, we are now fabricating the devices with the 1 and 4 μm thick undoped layer. We will report the simulation results and the device breakdown voltages of the fabricated ones.<br/><br/>[1] N. Oi, H. K. et al, <i>Sci Rep</i>, 8, 10660. (2018)<br/>[2] J. Tsunoda, H. K. et al., <i>IEEE EDL</i>, 43(1), p88-91. (2022)<br/>[3] ATLAS User’s Manual, Device Simulation Software, SILVACO, Santa Clara, CA, USA. (2015)<br/>[4] R. Alhasani, H. K. et al, <i>IEEE 16th NMDC</i>, pp. 1-6. (2021)