Prospects of mist CVD for fabrication of β-Ga2O3 MESFETs on β-Ga2O3 (010) substrates

Mist CVD was applied to grow the β-Ga2O3 channel layer of a MESFET on a semi-insulating β-Ga2O3 (010) substrate. The mobility and carrier concentration of the channel layer were 80 cm2 V–1 s–1 and 6.2 × 1017 cm−3, respectively. The device exhibited a pinch-off characteristic with a threshold gate voltage of −9 V, and the maximum drain current was 240 mA mm−1. The maximum transconductance was 46 mS mm−1 and the on-resistance was 30 Ω mm. This device performance suggests that mist CVD is a potential growth technology capable of providing low-cost devices in the future.

Mist CVD, which has been developed for the growth of oxide thin films, [29][30][31][32][33] is another candidate of the growth technology for β-Ga 2 O 3 devices. This technology uses a simple and non-vacuum apparatus, and a water/alcohol solution of metal compounds as a precursor, which enables safe, cost-effective, energy-saving, and highly scalable growth. There have been some reports on homoepitaxial growth of β-Ga 2 O 3 and coherent growth of β-(Al x Ga 1−x ) 2 O 3 alloys by the mist CVD. [34][35][36][37] Conductivity control by Snand Ge-doping of homoepitaxial β-Ga 2 O 3 was also achieved with carrier concentrations on the order of 10 17 cm -3 . 34,36) Following this, it is important to demonstrate device fabrication by mist CVD method for future mass production and from the viewpoint of sustainable development goals (SDGs). In this letter, we report the fabrication of β-Ga 2 O 3 MESFETs with the use of mist CVD on single-crystal β-Ga 2 O 3 (010) substrates and the demonstration of the device performance to show the prospects of mist CVD for costeffective practical devices.
Before fabricating β-Ga 2 O 3 MESFETs, we examined the basic growth characteristics of β-Ga 2 O 3 homoepitaxial films on β-Ga 2 O 3 substrates. The growth was performed by a homemade hot-wall-type mist CVD system on β-Ga 2 O 3 (010) substrates (Novel Crystal Technology Inc.). Si was adopted as an n-type dopant. Figure 1 shows the schematics of the mist generator. For Ga and Si precursors, we chose tris (acetylacetonato)gallium (III) and chloro-(3-cyanopropyl)dimethylsilane, respectively, which were dissolved in deionized water with the addition of a few amounts of hydrochloric acid. Hydrochloric acid was necessary to completely dissolve the precursors in the aqueous solution. Ga concentration in the aqueous solution was set at 0.05 mol l -1 and [Si]/[Ga] ratio in the solution was changed between 1 × 10 −6 and 2 × 10 −4 to control the carrier (electron) concentration in n-type Ga 2 O 3 layers. The source aqueous solution was installed in the mist generator shown in Fig. 1, and was atomized using ultrasonic transducers. The precursor mist was supplied to a horizontal quartz reactor heated at 700°C-800°C using O 2 carrier and dilution gases. The carrier and dilution gas flow rates were set at 3.0 and 0.5 l min -1 , respectively. For the growth of Ga 2 O 3 , the water and/or the O 2 gas act as an oxygen source. Prior to the homoepitaxy, the β-Ga 2 O 3 substrates were cleaned using acetone, ethanol, and deionized water.
The electrical properties, thicknesses, and structural properties of the epilayers were evaluated by Hall effect measurements, surface profile measurements, and X-ray diffraction (XRD) measurements, respectively. As ohmic contacts for the Hall effect measurement, Ti/Au electrodes with Van der Pauw configuration were prepared by electron beam evaporation. Figure 2(a) indicates the growth rate of the mist CVDgrown β-Ga 2 O 3 homoepitaxial films on β-Ga 2 O 3 (010) substrates. The growth rate decreased at higher growth temperatures, which was consistent with the reports by Lee et al. 34) and Nishinaka et al. 35) This trend can be attributed to the fact that at higher temperatures precursors were preferentially consumed in the gas phase and at the tube wall before being transported onto the substrate. 34,35) In this study, we temporarily adopted 750°C as the growth temperature since a reasonable growth rate was obtained as shown in Fig. 2(a). Nevertheless, the growth rate of 750 nm h -1 at 750°C is much lower than the highest value of 3.2 μm h -1 reported by Nishinaka et al., 35) because the Ga concentration in the source solution was lower. For Sidoping, the electron concentrations were controlled in the order of 10 17 to 10 19 cm -3 . The actual electronic properties of the MESFET channel layer were measured using the test elementary group (TEG) pattern in the device. Figure 2 010) substrate. The mesa structure was formed by using conventional photolithography and inductively coupled plasmareactive ion etching (ICP-RIE) using BCl 3 /Cl 2 . Si implantations were performed to form source and drain regions with a 200 nm deep box profile and total Si dose of 2.0 × 10 15 cm −2 (target Si density = 1 × 10 20 cm −3 ), followed by 30-min activation annealing at 925°C in N 2 gas. On the source and drain regions, a Ti(20 nm)/Au(230 nm) metal stack was deposited and then annealed at 470°C as the source and drain electrodes. We had already established the process conditions for the fabrication of n + -regions and their electrodes using the unintentionally-doped β-Ga 2 O 3 layers grown by molecular beam epitaxy (MBE). The transfer length measurements (TLM) on the highly Si-doped layer (∼1 × 10 20 cm −3 ) had revealed the contact resistance R c of 0.32 Ω mm and the specific contact resistance ρ c of the order of 10 −6 Ω cm 2 . In the present experiments for the mist CVD-grown β-Ga 2 O 3 , we adopted the process conditions ever established.
A 60 nm thick SiN passivation layer was fabricated by PECVD. For the gate region, after the SiN passivation layer was etched off by ICP-RIE using SF 6 , a Ti (3 nm)/Pt (12 nm)/ Au (350 nm) metal stack was deposited. Finally, a SiN passivation layer was etched by ICP-RIE using CF 4 for the contact to source and drain regions. The use of SF 6 for forming the gate, rather than CF 4 , is because we expected less damage to the β-Ga 2 O 3 layer, where the Schottky electrode is subsequently formed (not confirmed by experiments). The gate-to-source length (L GS ), gate-to-drain length (L GD ), gate length (L G ), field-plate extension of the gate toward the source (L FP_S ) and that toward the drain (L FP_D ) were 2, 2, 2, 1, and 1 μm, respectively. The gate width (W G ) was 50 μm.
The Hall effect measurement using the Van der Pauw TEG pattern in the device revealed the mobility, sheet carrier density, and sheet resistance of the channel layer as 80 cm 2 V −1 s −1 , 1.24 × 10 13 cm -2 , and 6.2 kΩ sq -1 ,    respectively. From the thickness of the channel layer (200 nm), the carrier concentration was calculated as 6.2 × 10 17 cm −3 . The Hall mobility is comparable to or only slightly lower than those of β-Ga 2 O 3 homoepitaxial layers at the similar carrier concentration grown by MBE and MOCVD. 38) This fact suggests that our β-Ga 2 O 3 channel layer has sufficient quality for electronic devices. Figure 4 shows the capacitance-voltage (C-V ) characteristics of two Schottky contacts (200 μm in diameter) formed on the channel layer at different positions. From the average slopes of the linear regions of 1/C 2 versus V in this graph, the effective donor concentration N d −N a was revealed as 5.1 or 6.9 × 10 17 cm −3 using the relative permittivity of 10. 39) This supports the carrier concentration of 6.2 × 10 17 cm −3 obtained by the Hall effect measurement within the experimental error or potion-to-potion variation.
The electrical properties of the MESFET are shown in Fig. 5. Figure 5(a) shows the drain current density I D versus the gate-source voltage V GS of the MESFET. The log plot of I D is shown in the figure inset of the linear plot. Here, the drain-source voltage V DS was kept at 5 or 10 V. The modulation of I D with V GS was seen and a pinch-off characteristics was observed at a threshold gate voltage of V GS = −9 V. The maximum transconductance g m was 46 mS mm -1 . However, I D did not drop lower than 0.08 mA mm −1 and several devices did not show clear pinch-off. Therefore, we should say that current leakage at the n-Ga 2 O 3 /substrate interface is not sufficiently suppressed. Since the n-Ga 2 O 3 layer was directly grown on the substrate without a buffer layer, Si may have accumulated at the interface and contributed as a leakage path. 40) It is also possible that Fe diffuses from the substrate to the n-Ga 2 O 3 layer. 41) Efforts to overcome these phenomena will continue to improve device performance in the future.
The DC drain characteristics, that is, the relationships between the drain current density I D versus the drain-source voltage V DS for the different gate-source voltage V GS as parameters are shown in Fig. 5(b). In the measurement range, the maximum I D was 240 mA mm −1 . The on-resistance was found to be R ON = 30 Ω mm. The off-state breakdown under V GS = −10 V occurred at V DS = 195 V. To increase the breakdown voltage as well as to reduce the on-state resistance, we will optimize the device structure and the growth conditions in further study.
Mist CVD is a simple, safe, cost-effective, and energysaving growth technology. Now it is hard for most CVD to achieve the control of low carrier concentrations, such as the order of 10 16 cm -3 (this is pointed out not only in our study but also by Ogawa et al. 36) ), which has been easily attained by MBE and MOCVD. One reason for this may be that the source precursors currently used in mist CVD are commercially available chemicals that have not been purified to meet semiconductor grade, leaving any unintentional impurities in the film. It is also possible that Si in the quartz tube decomposes in the water and hydrogen-containing environment of mist CVD and is incorporated into the grown film, as suggested by Konishi et al. 42) in the halide vapor phase epitaxy of β-Ga 2 O 3 . Efforts for solving these problems will be continued. Nevertheless, it was found that for the β-Ga 2 O 3 layers of the carrier concentration of the order of high 10 17 cm −3 , mist CVD can compete with sophisticated growth technologies such as MBE and MOCVD in the film quality. Applications of the films as devices are investigated in the present work, and the performance of MESFETs was not inferior or might be superior to those grown by MBE or MOCVD.
In summary, in this letter, we showed the performance of β-Ga 2 O 3 MESFETs whose active layer was grown by the simple, safe, cost-effective, and energy-saving growth technology of mist CVD. The device showed the maximum transconductance of g m = 46 mS mm −1 , the drain current of I D = 240 mA mm −1 , and the on-resistance of R ON = 30 Ω mm. FLOSFIA Inc., a Japanese company, has been developing α-Ga 2 O 3 devices using mist CVD for the growth of the active layers. The up-to-date data show the low on-resistance and large current SBDs and high reverse voltage of 1400 V of the pn junction. 43) It is expected that mist CVD contributes to the supply of low-cost α-Ga 2 O 3 and β-Ga 2 O 3 devices in the market.