Normally-Off Vertical Diamond MOSFETs with Drain Current Density over 200 mA/mm using Oxidized Si Terminated Diamond Channel Formed by Si Molecular Beam Deposition Approaches

For realization of high-power high-speed miniaturized complementary inverters [1], development of p-type field-effect transistors (p-FETs) corresponding to SiC or GaN n-FETs is essential. Normally-off operation with threshold voltages (<i>V</i>_th) more negative than −3 V are required for power devices with operating voltage above 500 V, but the conventional vertical diamond FETs are normally-on operation (<i>V</i>_th: +15 ~ +30 V) using hydrogen-terminated diamond (C-H) channel [2,3]. In contrast, we have fabricated (001) vertical diamond MOSFETs with oxidized Si terminated diamond (C-Si-O) channel by the reaction of chemical vapor deposition (CVD) SiO₂ with diamond and achieved the first normally-off operation (<i>V</i>_th: −9.0 V) in vertical diamond FETs with a maximum drain current density (<i>I</i>_D,max) of 108 mA/mm[4]. In this study, by Si molecular beam deposition (MBD) method, we fabricated C-Si-O side-wall channel in vertical diamond MOSFETs for the first time. We obtained the <i>V</i>_th of −4.5 V and the <i>I</i>_D,max of 214 mA/mm, which is the highest <i>I</i>_D,maxamong the normally-off vertical diamond FETs.<br/>The C-Si-O channel formation method is as follows. After 200 nm regrown undoped layer was deposited by microwave plasma CVD, SiO₂ was deposited 400 nm as a mask by tetraethyl orthosilicate based plasma-enhanced CVD. A p⁺⁺ layer ([B]: 1×10²¹ cm^-3) was selectively deposited by MPCVD after etching 400 nm of the SiO₂ mask and 40 nm of the regrown undoped layer by inductively coupled plasma reactive ion etching. After SiO₂ mask removal, the diamond surface was treated with hydrogen termination. 0.5 nm-silicon was deposited and followed by in-situ annealing at 1193 K for 15 min in the MBD apparatus. During the annealing in high vacuum, the reaction between C-H and Si occurred and C-Si- happened [5,6]. The C-Si- surface was naturally oxidized and transformed to C-Si-O after the C-Si- channel was exposed to air [7].<br/>The vertical device dimensions for source-source length (<i>L</i>_SS), effective channel length (<i>L</i>_ch), trench width (<i>W</i>_T), gate-drain length (<i>L</i>_GD), trench depth (<i>D</i>_T) and gate width (<i>W</i>_G) are 8 µm, 5.2 µm, 4 µm, 0.4 µm, 3 μm and 25 µm, respectively. The <i>I</i>_D,max normalized by the <i>W</i>_G was 214 mA/mm, and the<i> I</i>_D,max normalized by the active area (<i>L</i>_SS ×<i>W</i>_G: 2.0 × 10^-6 cm²) was 5350 A/cm² at <i>V</i>_DS: −45 V and <i>V</i>_GS: −40 V, and the on-resistance (<i>R</i>_on) was 8.1 mΩcm². The on/off ratio was 10⁷and gate leakage current was less than 10^-5 mA/mm at <i>V</i>_DS: −10 V. The maximum transconductance (<i>g</i>_m.max) was 15 mS/mm, and the maximum field-effect mobility (<i>μ</i>_FE.max) was 68 cm²/Vs. The <i>V</i>_th was −4.5 V at <i>V</i>_DS: −10 V. As a result, we achieved the normally-off operation of |<i>V</i>_th| &gt; 3 V, and the <i>I</i>_D.max of over 200 mA/mm. The <i>V</i>_thshifted more than 20 V in the negative direction compared to conventional C-H channel vertical diamond FETs [2,3]. The negative electron affinity of C-Si-O surface (−0.25 eV) [7] is weaker than that of C-H surface (−1.3 eV) [8] and the negative fixed charge of Al₂O₃ on the C-Si-O interface is less than that on the C-H interface. The large negative shift of<i> V_th</i> is due to the smaller band offset and the smaller band bending between diamond and Al₂O₃ in the C-Si-O channel than in the C-H channel. In the near future, by introducing a p⁻-drift layer, and C-Si-O channel only on the sidewalls, we will realize vertical diamond devices with high breakdown voltage over 1000 V and normally-off operation.<br/>[1] K. Okuda, N. Iwamuro, et al., EPE'16 ECCE Europe, 1-10. (2016)<br/>[2] N. Oi, H. K, et al., Sci. Rep., 8, 1-10. (2018)<br/>[3] J. Tsunoda, H. K, et al., Carbon 176, 349-357. (2021)<br/>[4] K. Ota, K. Narita, H. K, et al., Carbon, in press. (2023)<br/>[5] A. Shenk, C. Pakes, et al., Appl. Phys. Lett., 106(19), 191603. (2015)<br/>[6] Y. Fu, H. K, et al., IEEE TED, 69(5), 2236-2242. (2022)<br/>[7] A. Schenk, C. Pakes, et al., J. Phys., Conds. Matt, 29, 025003. (2017)<br/>[8] J. B. Cui, J. Ristein, and L. Ley, Phys. Rev. Lett. 81, 429. (1998)