Vertical-type two-dimensional hole gas diamond metal oxide semiconductor field-effect transistors

Power semiconductor devices require low on-resistivity and high breakdown voltages simultaneously. Vertical-type metal-oxide-semiconductor field-effect transistors (MOSFETs) meet these requirements, but have been incompleteness in diamond. Here we show vertical-type p-channel diamond MOSFETs with trench structures and drain current densities equivalent to those of n-channel wide bandgap devices for complementary inverters. We use two-dimensional hole gases induced by atomic layer deposited Al2O3 for the channel and drift layers, irrespective of their crystal orientations. The source and gate are on the planar surface, the drift layer is mainly on the sidewall and the drain is the p+ substrate. The maximum drain current density exceeds 200 mA mm−1 at a 12 µm source-drain distance. On/off ratios of over eight orders of magnitude are demonstrated and the drain current reaches the lower measurement limit in the off-state at room temperature using a nitrogen-doped n-type blocking layer formed using ion implantation and epitaxial growth.

mobility and has been used to form high-electron-mobility transistors (HEMTs), it is difficult to form a 2DEG layer in the vertical direction because the 2DEG appears only on the (0001) GaN surface. Therefore, vertical-type GaN MOSFETs use bulk conduction without using the high mobility of the 2DEG drift layer.
2DHG layers on diamond surfaces are advantageous for fabrication of vertical-type devices with trench structures. 2DHG layers can be formed on hydrogenated diamond surfaces independently of their crystal orientations. Because of the high trench coverage of the ALD process, ALD-Al 2 O 3 is used for gate insulation and passivation layers for 2DHG diamond MOSFETs in 3D structures. Diamond lateral and vertical-type 2DHG diamond MOSFETs with trench structures 17 and triple-gate field-effect transistors 18 were fabricated using inductively coupled plasma reactive ion etching (ICP-RIE) to form their trench structures and fin patterns, respectively, and both devices have 3D channel regions. In this work, we fabricate vertical-type 2DHG diamond MOSFETs with trench structures on highly boron-doped (p + ) single crystal diamond substrates and obtain device operation equivalent to that of devices with lateral structures with the same gate-drain distance.

Results
To fabricate vertical p-channel FET structures with low hole current leakage in the vertical direction, highly resistive n-type layers are required. A nitrogen-doped layer satisfies this requirement because nitrogen is a deep donor with activation energy of 1.7 eV and keeps the Fermi level E F above the intrinsic Fermi level Ei located at almost middle point of energy gap. Simple enlarged view of the trench portion is shown in Fig. 1(a) and band diagram at nitrogen-doped layer/undoped layer/Al 2 O 3 (A-A' in Fig. 1(a)) and p+-type substrate/undoped layer/Al 2 O 3 (B-B' in Fig. 1(a)) placed trench structure is shown in Fig. 1(b,c). Valence band offset is 2.9-3.9 eV 19,20 .
The device structure has a major influence on the device characteristics. Two vertical-type 2DHG diamond MOSFET structures are shown in cross-section views ( Fig. 1(d,e)) and from above ( Fig. 1(f)). The common aspect of these two vertical MOSFETs is the 2DHG layer that is formed on the C-H diamond surface, with the trench and the sidewall acting as the channel and drift regions. The main difference is that two different n-type blocking layers were fabricated by nitrogen hot-implantation with high-temperature post-annealing after growth ( Fig. 1(d)) and nitrogen incorporation during growth ( Fig. 1(e)). We used p + diamond substrates synthesized by the high-pressure high-temperature (HPHT) method for vertical-type devices with N-implanted layer, as shown in Fig. 1(d), and a p + conduction layer deposited by microwave plasma chemical vapour deposition (MPCVD) on an n-type diamond substrate (where the layer includes part of the back side of the substrate) for devices with N-doped epitaxial growth, as shown in Fig. 1(e). (Detail of substrate covered by boron doped layer is shown in Supplementary 1).
We then successfully formed high-concentration (~10 19 cm −3 ) nitrogen-doped layers either by nitrogen ion implantation into pure diamond at high energy (1.7 MeV) or nitrogen incorporation during epitaxial growth using MPCVD.
The p-type conducting region was limited to the sub-surface of the homoepitaxial layer (200 nm thick), which is shown as the light blue layer in Fig. 1(d,e), and was regrown after trench formation. The gate length was fixed at 4 µm. The trench depths of the two devices in Fig. 1(d,e) were set at ~4 and ~4.5 µm, respectively. Here, we define the effective drift region as the sub-surface of the regrown layer ranging from the gate edge to the top corner of the p + region near the trench sidewall, as shown in Fig. 1(d). The lengths from the gate edge to the trench entrance and from the entrance to the p + top corner were 3 µm and 2 µm, respectively. Therefore, the total drift layer length between the gate and the drain was ~5 µm. The distance from the p + top corner to the trench bottom is the overlap depth (~2 µm) of the 2DHG layer with the p + region and is sufficient for hole conduction from the 2DHG to the p + region. The 2DHG channel width (W G in Fig. 1(f)) was fixed at 25 µm. Current-voltage characteristics. Both kinds of the vertical-type 2DHG diamond MOSFETs fabricated in this study exhibited current-voltage characteristics that are comparable to those of lateral-type diamond MOSFETs. Figure 2(a,b) show two typical drain-source current (I DS ) versus drain-source voltage (V DS ) characteristics for the two vertical-type 2DHG diamond MOSFETs over a gate-source voltage (V GS ) range from −4 to 26 V with voltage steps of 2 V. Their drain current densities at a V DS of −10 V are 48.9 mA mm −1 (with N-implanted layer) and 36.7 mA mm −1 (with N-doped epitaxial layer), respectively. The maximum field-effect mobility and specific on-resistance values of these MOSFETs are 98 cm 2 V −1 s −1 and 31 mΩ cm 2 and 74 cm 2 V −1 s −1 and 41 mΩ cm 2 , respectively. Because we only used the left side source electrode for the I-V measurements, we adopted a device horizontal area equivalent to half-length between the two source electrodes composed of the surface channel and the drift area (L SG + L G + L GT + L T /2 = 15 µm in Fig. 1(f)) with the channel width (W G = 25 µm) to calculate the specific on-resistance. The source-to-source length (L SS ) will be reduced in the near future by downscaling of the trench by a factor of 10. Figure 2(c) shows further I DS − V DS characteristics for V DS ranging from 0 to −50 V and V GS ranging from −20 to 28 V with voltage steps of 4 V. The I DS of the vertical-type device with N-implanted layer increases with rising V DS and the measured drain current density is 234 mA mm −1 at a V DS of −50 V. The maximum current density is the highest reported in vertical-type diamond MOSFETs and is comparable to that of lateral-type 2DHG diamond MOSFETs with a similar gate-drain length (L GD ) and the same ALD-Al 2 O 3  Figure 2(d,e) shows the I DS -V GS characteristics at a V DS of −10 V. The threshold voltage (V th ) values of the two device-types were determined from the I DS -V GS characteristics to be 23.2 V (with N-implanted layer) and 23.0 V (with N-doped epitaxial layer), respectively, and these values are close to that of a lateral-type C-H diamond MOSFET with the same gate oxide thickness. These vertical-type 2DHG diamond MOSFETs have normally-on characteristics, which is common in general 2DHG diamond MOSFETs with Al 2 O 3 gate insulators. The 2DHG layer is induced by the negative charges in the ALD-Al 2 O 3 film 21 near the interface, even at a V GS of 0 V. The drain current was controlled well using the gate bias in a manner similar to that of lateral-type 2DHG diamond MOSFETs. This also means that the 300-nm-thick nitrogen-doped epitaxial layer or 50-nm-thick nitrogen-implanted layer effectively blocked the substrate leakage current that flows directly from the surface under the source electrode towards the drain electrode in the perpendicular direction.  This device operates with a high on/off ratio at temperatures under 200 °C. Figure 3(a) shows the logarithmic I DS -V GS characteristics of two vertical-type 2DHG diamond MOSFETs over the range from RT to 300 °C. The characteristics of the vertical-type devices with N-implanted layer and N-doped epitaxial layers are expressed using open symbols and closed symbols, respectively. In the off-state (over 25 V), I DS was approximately 10 −10 A mm −1 for RT operation, and the on/off ratio of this device was more than eight orders of magnitude in both vertical-type devices; these values are comparable to those of lateral-type devices (shown in Fig. 3(b)). At RT and 150 °C, the drain current reached the lower measurement limit and the on/off ratios were approximately eight and seven orders of magnitude, respectively, in both devices. When compared with RT operation, the V GS required to reach the lower limit shifted to positive values at 100 °C and 150 °C. The increase in the leakage current is remarkable at temperatures above 150 °C. At 300 °C, the drain current's dependence on the gate voltage falls and the on/off ratios of the vertical-type devices with N-implanted and N-doped epitaxial layers were approximately one and three orders of magnitude, respectively, indicating that this device will not work properly at higher temperatures (over 300 °C). At 200 °C, the leakage current increased when compared with that at RT, but the on/off ratio was still high as approximately five orders of magnitude and the device turned off at a V GS of 36 V ( Fig. 2(g)). Figures 2(g) and 3(a) show that this device can operate properly at temperatures under 200 °C and the change in the device characteristics between RT and 150 °C is very small.
The increase in the leakage current for these vertical C-H diamond MOSFETs started at relatively low temperatures when compared with the lateral-type C-H diamond MOSFETs shown in Fig. 3(b). The lateral-type devices could operate correctly at 300 °C and still had a very high on/off ratio of approximately seven orders of magnitude (see Fig. 3(b)). This result indicates that we can even obtain the off-state of the 2DHG layer at 300 °C, and the high leakage current of the vertical-type device over the range from 200-300 °C originates from bulk conduction rather than the surface conduction layer. The small current density dependence on V GS at 300 °C shows that the leakage current flows in the substrate in the perpendicular direction and the nitrogen-doped layer thus did not work as the leakage blocking layer at high temperatures. However, this layer could block the leakage currents at temperatures under 200 °C, despite being very thin (~50 nm). Figure 3(c) shows the maximum current densities of six devices (three devices of each type) over the range from RT-300 °C. The current densities of the devices with the N-implanted layers at a V DS of −50 V ranged from 197 to 235 mA mm −1 at RT and the variation of these values from the average value ranged from approximately 10% to 15%. This variation is caused by the surface condition of the trench sidewall, which is dependent on the plasma etching process, so the effects of the variation on both the hole mobility and the current density are significant. The maximum current density remained very stable from RT to 200 °C and the average maximum current densities for the three devices at RT and 200 °C were 220 and 217 mA mm −1 , respectively. At 300 °C, while the average maximum current density decreased by ~15% to 187 mA mm −1 from that at RT, the deviation of the maximum current density at approximately 7% between devices was smaller than that at RT. This result indicates that the current flows not only through the surface 2DHG layer but also through the substrate as a leakage current. Because the current density deviation is caused by the trench surface conditions and it decreases at higher temperatures, the current must flow via another path that is not influenced by surface conditions, such as the bulk. The current densities of the vertical-type devices with N-doped epitaxial layers ranged from 156 mA mm −1 to 191 mA mm −1 at RT and from 116 mA mm −1 to 132 mA mm −1 at 300 °C. At 300 °C, the average current density decreased by ~30%. The current density decreases with increasing temperature in both vertical-type devices. The rate at which the current density decreases in the vertical-type devices with N-doped epitaxial layers is lower than that of the vertical-type devices with N-implanted layer. However, the current density remains stable over a wide temperature range when compared with that of the junction FET 22 and that of a metal-semiconductor FET (MESFET) 23 with a boron-doped channel. This difference in the current density change between the two vertical-type devices is caused by leakage current differences at high temperatures (i.e., over 300 °C).

Analysis via simulations.
The operational simulation of the lateral 2DHG diamond MOSFETs has been performed based on the two-dimensional negatively charged sheet model upon which the 2DHG was realized and the measured I DS -V DS curve was reproduced well 13 . The charge sheet model is often used to simulate AlGaN/ GaN FETs with 2DEG for channels and can be adapted to simulation of the device characteristics of 2DHG diamond MOSFETs. For the vertical-type devices under consideration here, the main difference is the hole transport that occurs from the 2DHG drift region to the p + drain region. Figure 4(a) shows the I DS -V DS characteristics from both the measurement results and the simulated results for V DS ranging from 0 V to 10 V. In Fig. 4(a), the measured I DS -V DS characteristics are shown as open plots and the simulation results are shown as solid lines. It is proposed that the surface conditions inside the trench, such as those of the trench sidewall and bottom, are not the same as those of the surface outside the trench, so we varied the hole carrier mobility in each region. The optimal negative charge areal density is fixed at −6.7 × 10 12 cm −2 and the hole channel mobility based on low-field mobility model in the lateral channel and the vertical channel in the trench structure are 95 cm 2 V −1 s −1 and 43 cm 2 V −1 s −1 , respectively. The measured I D -V DS characteristics were reproduced well when the mobility in the vertical channel in the trench structure was reduced when compared with that of the lateral channel region. These results show that the carrier mobility of the vertical channel was lower than that of the lateral channel as a result of the plasma etching, but that was still a sufficient level for the channel to act as a 2DHG layer. The damage caused by the trench fabrication process using ICP-RIE was covered by regrowth of a 200-nm-thick undoped layer.
The current flow in the vertical-type 2DHG diamond MOSFETs was also simulated. Figure 4(b) shows a cross-section view of the vertical-type device along with the total current density and the hole concentration on the left side of the trench (indicated by the dotted line) at a V DS of −10 V and a V GS of −4 V in the simulations. The broken line on the right side of the figure indicates the boundary between the p + -type diamond substrate and the undoped layer. The left and the bottom side of the broken line indicate the p + -type diamond substrate. The square boxed part that is laterally located in the undoped layer corresponds to the region in which nitrogen ions were implanted at high temperature with an acceleration energy of 1.7 MeV. The trench structure of this device entered the p + -type diamond substrate to a depth of approximately 2 µm and the current flows in the 200-nm-thick regrown undoped layer on the p + -type substrate. Holes are induced in the 200-nm-thick regrown undoped layer on the p+ substrate and the current flow from the 2DHG layer to the p + region is higher in the upper part than at the bottom of the trench. Figure 4(b) shows that the trench structure fabricated in the p + -type diamond substrate is affected by the improvement in the current density and the current can flow through the thin nominally undoped layer, which is thick enough to form a 2DHG layer. In the layer, nitrogen density is low (<10 16 cm −3 ), but exceeds boron density (~10 14 cm −3 ). Thus, the obtained 2DHG is an inversion layer. From Fig. 4(b), current does not flow in most of the undoped layer that was deposited on the p + -type diamond substrate without the regrowth layer and the 2DHG layer. Results of simulation of the current density indicate that the trench bottom length does not influence the drain current density, which is low, and we can downscale the trench length to be as short as 1 µm without reducing the drain current density. Device simulation of vertical-type device with N doped epitaxial layer and without overlapping region are shown in Supplementary 4.
To confirm the importance of the action of the nitrogen-implanted layer as a leakage blocking layer, the I DS -V DS characteristics of a vertical-type device without the nitrogen-doped layer were simulated as shown in Fig. 4(c). While the maximum current density is 72.6 mA mm −1 at a V DS of −10 V and a V GS of −4 V, this value is approximately 50% higher than that of the device with the nitrogen-doped layer produced by ion implantation; however, the device without the nitrogen-implanted layer does not turn off, even at a V GS of 26 V, which is a sufficient gate voltage level to turn off the devices with the nitrogen-doped blocking layers. The left side of Fig. 4(d) shows a cross-section view of the vertical-type device without the nitrogen-doped layer that was used for the simulation. The total current density and the hole concentration for this type of device (surrounded by the dotted line) are shown on the right side of Fig. 4(d). The current flows directly from the source electrode to the drain through the undoped layer in the direction perpendicular to the substrate. These two patterns in the simulation results show that the nitrogen-doped n-type layer plays a significant role in blocking leakage currents. Breakdown characteristics. Figure 5(a,b) show the breakdown characteristics and the I DS -V DS characteristics of both kinds of vertical-type C-H diamond MOSFETs. The breakdown voltage (V B ) of the device with N-implanted layer, which was determined from Fig. 5(a), was 249 V and the drain current density was 193 mA mm −1 at a V DS of −50 V. The breakdown characteristics were measured at a V GS of 35 V, which is a sufficient voltage to turn this device off. The breakdown voltage (V B ) of the device with the N-doped epitaxial layers as determined from Fig. 5(b) was 359 V and the drain current density was 170 mA mm −1 at a V DS of −50 V. This breakdown voltage is more than 100 V higher than that of the previous device with the same effective drift region (~5 μm).  layer, the drain current density is maintained at a value close to the lower measurement limit at a V DS of less than −200 V, which is close to the breakdown voltage of the ion-implanted device. The leakage current increases with increasing drain voltage, but the rate at which it increases is lower than that in the device with N-implanted layer. These results show that the thicker nitrogen-doped layer reduces the leakage current in the high-drain-voltage region.
The total effective drift layer length was ~5 µm in both type devices. The average electric field strengths that were calculated using these values were 0.5 and 0.7 MV cm −1 for the both type devices and these values are lower than that of lateral-type C-H diamond MOSFETs 13 . The reduced breakdown voltage is caused by the individual structure of the vertical type C-H diamond MOSFETs. Figure 5(d) shows the electric field strengths of the device with N-implanted layer and the device with the N-doped epitaxial layer at the undoped layer and at the thin nitrogen-doped layer/regrown undoped layer interface at V DS of −250 V and V GS of 26 V. Results of electric field simulation of both kinds of devices are shown in Supplementary 5. In the device with N-implanted layer, the peak of electric field strength is located 1.2 µm from the surface at the 50-nm-thick nitrogen-doped layer/regrown undoped layer interface and the localized electric field strength is higher than 10 MV cm −1 at V DS of −250 V. In the device with the N-doped epitaxial layer, the peak of electric field is also located at the nitrogen-doped layer/ regrown undoped layer interface. However, the localized electric field is ~8 MV cm −1 , which is lower than that of the former device with thin nitrogen-doped layer. The difference in the electric field and the breakdown voltage caused by the thickness and placement of the nitrogen-doped layer is consistent. A thinner nitrogen-doped layer causes a high electric field concentration and a 2DHG layer between the gate edge and the trench, and the trench sidewall does not play a role as a drift layer. In lateral-type devices, the peak electric field is located near the gate edge, but in vertical-type devices, the highest electric field spot is not at the gate edge but at the nitrogen-doped layer/regrown undoped layer interface. The nitrogen-doped layer is required to block the leakage current that flows in the direction perpendicular to the diamond substrate and the electric field distribution is dependent on the condition of the nitrogen-doped layer. Therefore, the nitrogen-doped layer is very important not only in blocking the leakage current but also in terms of the device breakdown characteristics. There are two source electrodes located on both sides of the trench structure in the fabricated vertical-type C-H diamond MOSFETs. Figure 6(a,b,c) show the I DS -V DS characteristics of vertical-type C-H diamond MOSFETs with N-implanted layer when using only the left-side electrode, only the right-side electrode and using both side electrodes, respectively. The drain current density is normalized by channel width (W G ) of 25 µm. The maximum drain current density of the vertical-type device is 245 mA mm −1 when using only the left-side electrode and 220 mA mm −1 when using only the right-side electrode. The difference in the drain current density is caused by the condition of the trench sidewall and the slight displacement of the source electrode position with respect to the trench structure. The maximum drain current density of the vertical-type device when both side source electrodes are used is 444 mA mm −1 . In this case, the maximum current density is slightly lower than the total current density when using either side source electrode alone. The overlapping of the channel region at the trench bottom causes the reduction in the current density. However, the drain current of the vertical-type device when using both side electrodes is close to the total drain current when using the left-or right-side source electrode only, because the drain current mainly flows through the regrown undoped layer at the trench sidewall. The sidewall length for both vertical-type C-H diamond MOSFETs is 10 µm and the current density when standardized with respect to the channel area is ~1500 A cm −2 and this value is higher than diamond junction FET (JFET) 20 . I DS -V DS characteristics standardized by channel area is shown in Supplementary 6. In the vertical-type C-H diamond MOSFET presented here, both the wide trench structure and the drift region at the substrate surface are unnecessary because all the drift current passes through the sidewall and not the bottom of the trench, as shown in Fig. 4(b). The horizontal area of the device can be reduced to approximately one quarter of the existing size by fabricating the gate electrode in the trench structure and reducing the sidewall distance, and the standardized current density will be more than 5000 A cm −2 .

Summary
We have fabricated vertical-type 2DHG diamond MOSFETs with nitrogen-doped blocking layers formed by either ion implantation or epitaxial growth with maximum current densities (234 mA mm −1 and 191 mA mm −1 , respectively, at V DS of −50 V and V GS of −20 V) and an on/off ratio (~10 8 ) that were comparable to those of lateral-type devices at room temperature. In the trench structure, the current flows in the 2DHG layer of the thin undoped layer and the p + -type diamond substrate. From the I DS -V DS characteristics of these two device types, the specific on-resistance of the vertical-type device is comparable to that of a lateral-type device. The results reported here therefore show that the condition of the trench sidewall surface is close to that of the substrate surface and the approximately 200-nm-thick undoped layer is not highly resistive. The I DS -V GS characteristics and the on/ off ratio show that current flows in the 2DHG layer and the nitrogen-doped layer blocks the substrate leakage current well.

Methods
In this report, we used a p + -type HPHT (001) single crystalline diamond substrate to fabricate vertical-type hydrogen-terminated diamond MOSFETs with N-implanted layer. For the vertical device with the N-doped epitaxial layer, we used n-type diamond covered with a boron-doped layer. The boron concentration of the p + -type diamond was 10 19 -10 20 cm −3 , so we regard this substrate as being part of the drain. Initially, for device with N-implanted layer, we deposited a 2.0-µm-thick undoped epitaxial layer to develop a drift layer by microwave plasma chemical vapor deposition (MPCVD) on the p + -type diamond substrate. The growth temperature, time and chamber pressure for the undoped epitaxial layer were 600 °C, 15 h 20 min and 35 Torr, respectively. The H 2 , CH 4 and CO 2 flow rates were 394, 3 and 3 sccm, respectively. An additional n-type diamond layer was fabricated by 1.7 MeV hot implantation of nitrogen ions. The temperature during implantation and fluence were 800 °C and 2 × 10 14 cm −2 , respectively. From Stopping and Range of Ions in Matter (SRIM) simulations, a nitrogen-doped n-type diamond layer was fabricated at a depth of approximately 1 µm from the substrate surface and the nitrogen concentration was ~10 19 cm −3 . The nitrogen-ion implanted layer thick was approximately 50 nm. (doping profile by SRIM is shown in Supplementary 2) After nitrogen ion implantation, the diamond substrate was annealed at 1500 °C for 2 h. The chamber pressure was 1 atm and the atmosphere was Ar. For device with N-doped epitaxial layer, we deposited undoped and nitrogen-doped epitaxial layer on n-type substrate covered by boron-doped layer. The growth temperature, time and chamber pressure for the undoped epitaxial layer were 600 °C, 13 h and 35 Torr, respectively. The H 2 , CH 4 and CO 2 flow rates were 394, 3 and 3 sccm, respectively. The growth temperature, time and chamber pressure for the nitrogen-doped epitaxial layer were 600 °C, 6 h and 35 Torr, respectively. The H 2 , CH 4 , CO 2 and N 2 flow rates were 393, 3, 3 and 1 sccm, respectively.
To fabricate the trench structure, the PMGI-SF5/TSMR-V90-CP27 positive photoresist bilayers were coated sequentially on the diamond substrate using a spin coater and exposed using a photolithography system. The baking temperatures and times for the PMGI-SF5 and TSMR-V90-CP27 layers were 180 °C and 5 min and 90 °C and 1.5 min, respectively. MgO was deposited using a dual ion beam sputtering system (DIBS) as a mask for the ICP-RIE process. The MgO layer thickness was 150 nm. The trench structure was fabricated by ICP-RIE using O 2 gas plasma. The gas flow, etching power and gas pressure were 90 sccm, 400 W and 0.5 Pa, respectively. The trench depth and length (L T ) were 4 µm and 10 µm, respectively. The MgO was removed using a mixed acid solution (HNO 3 and H 2 SO 4 with a volume ratio of 1:3).
An additional undoped layer was deposited by MPCVD to cover the entire surface, including the trench structure, and form the 2DHG channel. The growth temperature and the chamber pressure were the same as the values given above. The H 2 , CH 4 and CO 2 flow rates were 400, 0.4 and 0.4 sccm, respectively. The growth time was 5 h and the undoped layer thickness was 200 nm. Ti/Au structures were deposited on the diamond substrate on both sides of the trench to act as source electrodes by electron beam (EB) evaporation. The thicknesses of the Ti and Au layers were 30 nm and 100 nm, respectively. The source electrode pattern was formed by photolithography and we used the TSMR-V90-CP27 positive photoresist. To form TiC, the sample was annealed at 500 °C for 30 min. To induce the 2DHG layer, hydrogen termination was performed over the whole substrate surface using remote hydrogen plasma. The plasma power, plasma temperature, chamber pressure and distance between sample and probe were 70 W, 620 °C, 20 Torr, and 50 mm, respectively. The H 2 gas flow was 100 sccm. The substrate surface without the channel region but including the trench structure was oxygen-terminated using an oxygen plasma treatment for isolation. The plasma power and treatment time were 300 W and 2 min, respectively. We used the TSMR-V90-CP27 positive type resist as a mask for isolation. A 200-nm-thick Al 2 O 3 layer was deposited by ALD to act as a gate insulation and passivation layer for hydrogen termination at 450 °C where the oxidant was H 2 O. To expose the source electrodes, the substrate was coated with a resist layer along with the isolation process mask and the 200-nm-thick Al 2 O 3 layers without resist coating were wet-etched using a 2.38% tetramethylammonium hydroxide (TMAH) solution. Ti/Au layers were deposited on the back side of the diamond substrate to act as drain electrodes by Ion Beam Sputtering system. The thicknesses of the Ti and Au layers were 10 nm and 250 nm, respectively. The gate electrode pattern was fabricated by photolithography under the same conditions that were used to form the MgO mask pattern. Al was deposited using a thermal evaporation system and the Al layer thickness was 100 nm. In this work, the gate electrodes were formed between the trench structure and the two sides of the source electrodes.