Ohmic Contact Formation to β-Ga₂O₃ Nanosheet Transistors with Ar-Containing Plasma Treatment

Abstract

Effective Ohmic contact between metals and their conductive channels is a crucial step in developing high-performance Ga₂O₃-based transistors. Distinct from bulk materials, excess thermal energy of the annealing process can destroy the low-dimensional material itself. Given the thermal budget concern, a feasible and moderate solution (i.e., Ar-containing plasma treatment) is proposed to achieve effective Ohmic junctions with (100) β-Ga₂O₃ nanosheets. The impact of four kinds of plasma treatments (i.e., gas mixtures SF₆/Ar, SF₆/O₂/Ar, SF₆/O₂, and Ar) on (100) β-Ga₂O₃ crystals is comparatively studied by X-ray photoemission spectroscopy for the first time. With the optimal plasma pre-treatment (i.e., Ar plasma, 100 W, 60 s), the resulting β-Ga₂O₃ nanosheet field-effect transistors (FETs) show effective Ohmic contact (i.e., contact resistance R_C of 104 Ω·mm) without any post-annealing, which leads to competitive device performance such as a high current on/off ratio (>10⁷), a low subthreshold swing (SS, 249 mV/dec), and acceptable field-effect mobility (${\mu }_{\mathrm{e}\mathrm{f}\mathrm{f}}$, ~21.73 cm² V⁻¹ s⁻¹). By using heavily doped β-Ga₂O₃ crystals (N_e, ~10²⁰ cm⁻³) for Ar plasma treatments, the contact resistance R_C can be further decreased to 5.2 Ω·mm. This work opens up new opportunities to enhance the Ohmic contact performance of low-dimensional Ga₂O₃-based transistors and can further benefit other oxide-based nanodevices.

1. Introduction

β-Ga₂O₃ has recently garnered great amounts of attention owing to its ultrawide energy bandgap (>4.5 eV) [1,2,3,4], large critical breakdown field E_br (6–8 MV/cm) [5,6], high optical transparency [7], desirable thermal stability, and long-term stability [8,9]. Its Baliga’s figure of merit (at 3214) is far higher than those of GaN (at 846) and 4H-SiC (at 317) [10,11], despite its low carrier mobility (at 300 cm² V⁻¹ s⁻¹), making it compelling for use in next-generation power transistors. Moreover, β-Ga₂O₃ displays a large Johnson’s figure of merit (at 2844.4) intended for high-frequency operation, far surpassing those of GaN (at 1089) and 4H-SiC (at 277.8) [12]. Currently, device-quality β-Ga₂O₃ single crystals are commercially available. There have been numerous reports of bulk crystal growth techniques [13,14,15], facilitating the extensive study of β-Ga₂O₃. Furthermore, high-quality Ga₂O₃ epitaxial layers can be achieved by various chemical vapor deposition (CVD) methods, such as metal organic chemical vapor deposition (MOCVD) and mist CVD [16]. Analogous to well-known 2D materials (e.g., h-BN, ReS₂, and graphene), β-Ga₂O₃ nanosheets can be peeled off from a freestanding bulk crystal while it is not a van der Waals crystal, owing to the weak out-of-plane force in the plane direction of (100) resulted from its large lattice constant. Compared to atomic layer deposition (ALD), magnetron sputtering, and other thin-film oxide growth methods, the mechanical cleavage can evidently reduce the impact from strain on the β-Ga₂O₃ device structures [17,18]. The exfoliated β-Ga₂O₃ nanosheet also provides a different surface orientation (100) for the research on β-Ga₂O₃ as compared with the conventional crystalline planes. Many studies have focused on the fabrication of high-performance transistors based on the peeled β-Ga₂O₃ nanosheets [17,18,19,20,21,22,23]. However, the Ohmic contact formation between the conductive channel and electrodes is still challenging due to the ultrawide bandgap of β-Ga₂O₃. To achieve excellent device performance, the use of low-resistance Ohmic electrodes is crucial, which plays important parts in the transport properties, such as on-current, field-effect mobility (${\mu }_{\mathrm{e}\mathrm{f}\mathrm{f}}$), and current on/off ratio (I_ON/I_OFF).

It is known that post-annealing is an effective strategy to form Ohmic junctions with low resistance. The influence of the contact metal on the peeled β-Ga₂O₃ nanosheets has been studied using various metals, annealing temperatures, and atmosphere [23,24,25]. Li et al. [23] utilized Ti/Au electrodes annealing in N₂ to generate high-density interfacial oxygen vacancies (V_O), resulting in the formation of Ohmic junctions. Guo et al. [26] reported the conversion of β-Ga₂O₃ samples from Ohmic contacts to Schottky contacts by tuning the concentration of V_O via annealing in an O₂ environment. Annealing of Ti/Al metal stacks has also been explored to produce more V_O at the β-Ga₂O₃/electrode interface for Ohmic contact formation [25]. Currently, the oxygen vacancies in β-Ga₂O₃ have been deemed as the donors in lots of research [23,25,26,27,28,29,30]. Yamaga et al. [28] theoretically demonstrated that the clusters of V_O are responsible for the electron conduction in unintentionally doped β-Ga₂O₃ at room temperature. Hajnal et al. [29] explained the n-type conduction of intrinsic β-Ga₂O₃ induced by V_O at higher temperatures. Very recently, Narayanan et al. [30] rigorously verified the dominant role of V_O in the electrical conduction of intrinsic β-Ga₂O₃. However, different from the bulk-like samples, excess thermal energy of the annealing process may destroy the low-dimensional material itself, including nanowires, nanoribbons, nanotubes, and nanosheets. Given this concern of the thermal budget, a moderate and feasible alternative is required to modify the surface of the low-dimensional β-Ga₂O₃ and form the decent Ohmic junctions. As a fascinating candidate, the reactive ion etching (RIE) treatment has been mentioned in some reports [17,31]. The β-Ga₂O₃ crystals treated with BCl₃- or SF₆-based RIE exhibit almost Ohmic behaviors. Note that compared with chlorine-based chemistry, SF₆ and Ar plasmas have the advantages of lower corrosiveness, lower toxicity, and better compatibility with the silicon manufacturing process [17,32]. So far, a systematic study regarding the evolution of the surface chemical states of β-Ga₂O₃ induced by SF₆ and Ar plasmas is still lacking. It is critical to analyze the surface treatment effects for a deep understanding of the underlying mechanism of Ohmic junction formation.

In this paper, the direct effects of various plasma treatments on the surface chemical states of (100) β-Ga₂O₃ crystals were systematically studied by X-ray photoelectron spectroscopy (XPS). Four types of mixture gases were used, including SF₆/Ar, SF₆/O₂/Ar, SF₆/O₂, and Ar. The optimal plasma process parameters (i.e., Ar plasma, 100 W, 60 s) were extracted. Furthermore, the quasi-2D β-Ga₂O₃ nanosheet transistor pre-treated with the optimized parameters was fabricated and characterized.

2. Experimental Methods

In this study, Sn-doped (100) β-Ga₂O₃ bulk crystals, which had an effective carrier concentration N_e of ~10¹⁸ cm⁻³, produced by the edged-defined film-fed technique, were employed as starting samples. The surface plasma pre-treatments were conducted in a reactive-ion etching reactor (RIE-10NR, Samco, Tokyo, Japan). The chamber pressure was set to be 4 Pa. For the plasma treatments of SF₆/Ar, SF₆/O₂/Ar, and SF₆/O₂, the RIE power was set as 200 W. The device fabrication began with the exfoliation of β-Ga₂O₃ nanosheets from the parent bulk onto the 110 nm thick SiO₂/Si (p²⁺) substrate with alignment markers. Laser direct writing lithography was employed to pattern the drain and source electrodes of the field-effect transistor (FET) using PMMA, followed by Ti (30 nm)/Au (150 nm) metallization and a lift-off process. Before the contact pattern was obtained, the peeled β-Ga₂O₃ nanosheets together with the underlying SiO₂/Si (p²⁺) substrate were cleaned via a standard solvent degreasing procedure and pre-treated with Ar plasmas.

XPS characterization was conducted via the SPECS XPS system (SPECS, Aachen, Germany) with a monochromatic Al Kα source (hv = 1486.6 eV) for photoelectron excitation. The source power was set as ~150 W (15 mA, 10 kV). Charge neutralization was carried out using an electron flood gun. The analysis region was a 1 mm diameter round spot. The incident angle and takeoff angle were, respectively, 58° and 90°. High-resolution scans were conducted 25 times for the binding energy (BE) of specific elements. To compensate for the charging effect, all XPS spectra were referenced to the C 1s peak located at 284.6 eV. Moreover, spectral deconvolution was performed by Shirley background subtraction using the Voigt function. To evaluate the thickness of the pre-exfoliated β-Ga₂O₃ nanosheets, the tapping-mode atomic force microscope (AFM; Dimension XR Icon, Bruker, Billerica, MA, USA) was employed. The surface morphology of the β-Ga₂O₃ nanosheets was characterized by optical microscopy (Axio Scope, Carl Zeiss, Oberkochen, Germany). All of the electrical measurements for the β-Ga₂O₃ nanosheet FETs were performed with the semiconductor parameter analyzer (Keysight B1500A, Santa Rosa, CA, USA) at room temperature.

3. Results and Discussion

Prior to the fabrication of transistors, four types of (100)-oriented β-Ga₂O₃ bulk samples were separately treated with SF₆/Ar, SF₆/O₂/Ar, SF₆/O₂, and Ar plasmas and comparatively analyzed by XPS. Figure S1a shows the C 1s XPS spectrum of the untreated β-Ga₂O₃ (100) bulk crystal. The C 1s region arising from surface contamination can be resolved into three subpeaks. The main peak positioned at 284.6 eV can be assigned to alkyl type carbon (C-C), whereas the subpeak (at 286.1 eV) with a binding energy separation of 1.5 eV to the main peak is attributed to the ester (C-O-C) or alcohol (C-OH) functionality. Another subpeak at 288.5 eV is ascribed to the carbon participating at the O-C=O group [7,33,34,35]. Normally, for the untreated bare β-Ga₂O₃ single crystal, the Ga 2p_3/2 plot can be fitted with a single peak ascribed to the Ga-O bond, and the Ga 3d is often deconvoluted into two major components, ascribed to Ga 3d_5/2 and Ga 3d_3/2 [3,23]. However, in Figure 1b, a very weak subpeak at 18.6 ± 0.2 eV cannot be excluded from the XPS measurements, which is ascribed to the presence of minor oxygen vacancies (V_O) in β-Ga₂O₃ induced by the initial crystal growth process. Figure 1c presents the XPS spectrum of the O 1s core energy level for the untreated β-Ga₂O₃ bulk sample. Note that the O1s envelope was well fitted by two components. The primary peak positioned at 530.5 ± 0.2 eV corresponds to the lattice oxygen bond (L_O), which is closely related to the formation of O²⁻ ions with Ga³⁺ metal cations [34]. The other subpeak with very weak intensities is located at 531.8 ± 0.2 eV, resulting from the existence of a minor V_O at the β-Ga₂O₃ surface. The V_O/(L_O + V_O) area ratio was extracted to be ~5%, indicating that the content of oxygen vacancy V_O in the near-surface region of untreated β-Ga₂O₃ crystal is very low. It corresponds well with the Ga 3d XPS data above.

After the SF₆/Ar plasma treatment, the total peak shape of the C 1s spectrum changed greatly, as shown in Figure S1b. This C 1s signal was still dominated by three species, which are ascribed to adventitious carbon at 284.6 eV, C-O-C or C-OH groups at 286.15 eV, and carboxylate groups at 288.8 eV [33,34,35]. No evident change in the Ga 2p_3/2 and Ga 3d core-level peaks was observed, as indicated in Figure 1d,e. Figure 1f shows the O 1s spectrum for the SF₆/Ar plasma-treated β-Ga₂O₃ bulk. Interestingly, the intensity of the oxygen vacancy-related subpeak (at 531.87 ± 0.2 eV) increased obviously as compared with the untreated one. The corresponding V_O/(L_O + V_O) area ratio increased to 18.3%, suggesting a marked rise in oxygen vacancies near the β-Ga₂O₃ surface, which is beneficial to reducing the contact resistance [23,26]. As previously reported [32], the SF₆/Ar plasma treatment process includes the chemical action and physical bombardment. In this progress, SF₆ gas can slowly react with the surface of Ga₂O₃, forming nonvolatile GaF_x and volatile SO_y based on the reactions Ga₂O₃ + SF₆ → GaF_x + SO_y. And, the Ar plasma with strong bombardment can eliminate the residual GaF_x on the surface. However, the origin of the newly generated oxygen vacancies V_O is confusing.

In order to identify the source and create more V_O, the SF₆/O₂/Ar plasma treatment was carried out. The SF₆ and O₂ gases were, respectively, introduced with the flow rates of 40 and 5 sccm, while the Ar flow rate was set as 30 sccm. The corresponding RIE power was 200 W. The purpose of reactive gas O₂ in this procedure is to remove S from SF₆ and promote an increase in F concentration based on the reaction SF_x + O → F + SOF_x-1 (x ≤ 5) [36]. Similarly, there was no noticeable change in the Ga 2p_3/2 and Ga 3d core energy levels, as depicted in Figure 2b,c. The addition of O₂ was found to barely affect the generation of oxygen vacancies. As can be seen from the O1s signal in Figure 2a, the V_O/(L_O + V_O) area ratio was ~18.8%, slightly higher than the SF₆/Ar plasma-treated one. This suggests that enhancing the chemical action of SF₆ cannot help to produce more oxygen vacancies V_O.

To obtain further insight, we excluded the argon gas from the mixture and performed the SF₆/O₂ plasma treatment on the β-Ga₂O₃ crystal. Figure 2d shows the corresponding O1s plot for the SF₆/O₂-treated β-Ga₂O₃ bulk sample. Surprisingly, without the assistance of Ar bombardment, the oxygen vacancy-associated subpeak (at 531.95 ± 0.2 eV) was still observable. The ratio of V_O/(L_O + V_O) was up to 17.7%, revealing that appreciable V_O exists near the β-Ga₂O₃ surface. These results are in agreement with the work reported by Jeong et al. [37], suggesting that other ions (e.g., SF_x) can also act as a physical bombardment during the plasma process. Moreover, the F 1s signal centered at 685.3 ± 0.2 eV was clearly detected, as displayed in Figure 2e. This peak belongs to the Ga-F bond [2], indicating the formation of non-volatile GaF_X at the surface. Jeong et al. [37] proposed that the introduced fluorine atoms may serve as shallow donors in the resistive β-Ga₂O₃ film and bring about an increase in conductivity. However, the F atoms can behave differently in β-Ga₂O₃ crystals with different doping concentrations. As can be seen from the two-terminal I–V curves in Figure 2f, the n-type β-Ga₂O₃ nanosheet sample pre-treated with SF₆/O₂ plasmas exhibited a common Schottky-like behavior, although the conductivity increased compared to the untreated one. Yang et al. [38] recently demonstrated that fluorine plasma treatment can markedly increase the barrier height of β-Ga₂O₃ Schottky barrier diodes, especially heavily doped or moderately doped β-Ga₂O₃. This is because the Si donors in β-Ga₂O₃ materials were compensated by F atoms and formed neutral complexes. Thus, the SF₆-based plasma treatments could produce two competing effects. There is a hard tradeoff between ion bombardment-induced vacancies and the F-doping effect. The ion bombardment-caused V_O can effectively improve conductivity and reduce the contact resistance [23,30], whereas the F-doping effect can compensate for the n-type dopants in β-Ga₂O₃ and impair conductivity. In addition, SF₆-based gases can evidently etch the gate dielectric layer SiO₂ beneath the β-Ga₂O₃ nanosheet during the device fabrication process, resulting in severe deterioration of the transistor performance. Note that the etch rate of the SiO₂ layer by the SF₆ plasmas was as high as ~55 nm/min, even faster than that of β-Ga₂O₃ (i.e., 16 nm/min) [17]. Figure S2 shows photographs of the 110 nm thick SiO₂/p-Si substrates with and without SF₆ plasma treatment for merely 90 s. Clearly, the SF₆-treated SiO₂/Si substrate displays the primary color of a silicon wafer, suggesting the absence of a SiO₂ layer. Taken together, SF₆-contaning plasma is not suitable, while the Ar-plasma treatment seems to be a better choice for achieving Ohmic contact with the heavily doped or moderately doped n-type β-Ga₂O₃ crystals.

Since XPS can sensitively monitor the evolution of surface chemical states, we performed a systematic XPS analysis on the treated β-Ga₂O₃ samples with pure Ar-plasmas having different powers and times. Figure S3 shows the corresponding spectra of C 1s among the Ar-treated β-Ga₂O₃ bulk samples. All of them display the same peak shapes, revealing the existence of the same surface contaminants after Ar-plasma bombardments. Figure 3a–f exhibit the evolution of O 1s core-level spectra of the β-Ga₂O₃ (100) bulk treated by Ar plasmas with different powers and times. It can be seen that all of the O 1s data were resolved into two kinds of subpeaks. The major peak at 530.5 ± 0.2 eV was identified as lattice oxygen L_O (O-Ga bonding state), while another peak at higher binding energies can be credited to the presence of V_O. The V_O/(L_O + V_O) ratio continuously evolves with the Ar-plasma power and time. Interestingly, all of the V_O/(L_O + V_O) values surpass the SF₆/O₂ plasma-treated one, which highlights the advantage of Ar-plasma bombardment in the creation of oxygen vacancies. Figure 4 plots the V_O/(L_O + V_O) area ratio as a function of the treatment time and power. Obviously, the optimal Ar-plasma parameters were 100 W, 60 s. The corresponding V_O/(L_O + V_O) ratio was as high as 45%, suggesting the formation of a great number of V_O near the sample surface. Note that the higher power may result in a lower Vo concentration. It is possibly due to the clusters of oxygen vacancy being potentially damaged by a higher power. Yamaga et al. [28] have theoretically proposed that the oxygen vacancy Vo of β-Ga₂O₃ crystals exists in the form of clusters.

To examine the Ohmic junction formation into β-Ga₂O₃ devices, the bottom-gated β-Ga₂O₃ nanosheet transistor was fabricated using the optimized plasma treatment (i.e., Ar-plasma, 100 W, 60 s). Figure 5a presents a schematic of the bottom-gated β-Ga₂O₃ device structure. The peeled β-Ga₂O₃ nanosheet served as the conductive channel. Figure 5b shows an optical micrograph of a completed β-Ga₂O₃ nanosheet transistor on a 110 nm SiO₂ bottom-gate oxide. Compared to the traditional 285 nm thick SiO₂ layer, the 110 nm thick SiO₂ can provide stronger gate modulation in the bottom-gate device configuration. The nanosheet channel length (i.e., L_ch) is around 5 μm, and the width of channel (W_ch) is ~2.2 μm. Figure 5c shows an AFM micrograph (top) of the corresponding β-Ga₂O₃ nanosheet channel. The thickness of the nanosheet was measured around 155 nm (bottom). In the absence of post thermal annealing, the fabricated β-Ga₂O₃ nanosheet FET showed linear Ohmic behavior at a low V_ds regime, as displayed in Figure 5d. Figure 5e gives the output characteristics (I_ds-V_ds) from the β-Ga₂O₃ nanosheet FET. Well-behaved line shapes were observed, which can be attributed to the presence of effective Ohmic contact. The bottom-gate voltage (V_gs) varies from −20 to 5 V with a step of 5 V. The drain current I_ds can be well regulated by the gate bias. Clear pinch-off properties and good current saturation at large V_ds values can be seen, indicating good device performance of this n-type nanosheet FET. Figure 5f shows transfer curves (I_ds-V_gs) from the β-Ga₂O₃ nanosheet FET in the log and linear scales at a V_ds of 0.1 V. Owing to the formation of effective Ohmic contact, well-behaved electrical characteristics were acquired, including a high I_ON/I_OFF ratio (>10⁷), a small off-state current (I_OFF < 0.1 pA), and a small subthreshold slope (SS, 249 mV/dec). These electrical parameters of the fabricated β-Ga₂O₃ nanosheet device are comparable with those of other reports [9,12]. The field-effect electron mobility (${\mu }_{\mathrm{e}\mathrm{f}\mathrm{f}}$) was around 21.73 cm² V⁻¹ s⁻¹, as extracted via

$${\mu }_{\mathrm{e}\mathrm{f}\mathrm{f}}={\displaystyle \frac{\mathrm{d}{I}_{\mathrm{d}\mathrm{s}}}{\mathrm{d}{V}_{\mathrm{g}\mathrm{s}}}}{\displaystyle \frac{L_{\mathrm{c}\mathrm{h}}}{W_{\mathrm{c}\mathrm{h}}\cdot C_{\mathrm{g}}\cdot V_{\mathrm{d}\mathrm{s}}}}$$

(1)

where $W_{\mathrm{c}\mathrm{h}}$, $C_{\mathrm{g}}$, $V_{\mathrm{d}\mathrm{s}}$, and $L_{\mathrm{c}\mathrm{h}}$ are the channel width, oxide capacitance, drain bias, and channel length of the nanosheet transistor, respectively. $V_{\mathrm{g}\mathrm{s}}$ and $I_{\mathrm{d}\mathrm{s}},$ respectively, denote the bottom-gate bias and drain current in the β-Ga₂O₃ transistor.

The Y-function method was employed to calculate the contact resistance (R_C) of β-Ga₂O₃ FET via the following equation:

$$Y={\displaystyle \frac{I_{\mathrm{d}\mathrm{s}}}{\sqrt{g_{\mathrm{m}}}}}=\sqrt{V_{\mathrm{d}\mathrm{s}}{C}_{\mathrm{g}}{W}_{\mathrm{c}\mathrm{h}}{\mu }_0/L_{\mathrm{c}\mathrm{h}}}·(V_{\mathrm{g}\mathrm{s}}-V_{\mathrm{t}\mathrm{h}})$$

(2)

where $g_{\mathrm{m}}$ represents the transconductance. µ₀ and $V_{\mathrm{t}\mathrm{h}}$ are the low-field mobility and threshold voltage, respectively [39,40,41]. The Y-function calculation is capable of extracting the electrical parameters from a transfer curve when drain bias V_ds is quite small. The µ₀ was calculated to be ~76.07 cm² V⁻¹ s⁻¹. From the values of $V_{\mathrm{t}\mathrm{h}}$ and µ₀, the R_C was extracted as ~104 Ω·mm, which is smaller than the value (i.e., 800 Ω·mm) reported by Jeong et al. [37] for CF₄-plasma treated β-Ga₂O₃. For comparison, Figure S4 shows the I–V measurements of the moderately doped β-Ga₂O₃ nanosheet samples (N_e, ~10¹⁸ cm⁻³) pre-treated with nonoptimal Ar-plasma parameters. All of them exhibit different asymmetric Schottky behaviors. Actually, the contact performance can be further improved by using high-doping-level β-Ga₂O₃ crystals (N_e, above 10¹⁹ cm⁻³) for plasma treatments. Figure 6a shows the I–V curves of the heavily doped β-Ga₂O₃ (N_e, ~10²⁰ cm⁻³) channel devices with Ar plasma treatment at various channel sizes (i.e., L_ch/S ratio, where S is the cross-sectional area and L_ch represents the channel length). Similarly, an excellent linear Ohmic behavior was acquired. The transfer length method (TLM) was employed to determine the contact resistance R_C based on the following formula:

R_t = ρ·L_ch/S + 2R_C

(3)

where ρ and R_t are the channel resistivity and total resistance of the heavily doped β-Ga₂O₃ device, respectively [41]. As displayed in Figure 6b, the R_C value obtained from the half of the y-intercept is ~5.2 Ω·mm, which is smaller than that of the moderately doped β-Ga₂O₃ (N_e, ~10¹⁸ cm⁻³) channel. For comparison, Figure S5 also presents the I–V measurements of the heavily doped β-Ga₂O₃ nanosheet samples (N_e, ~10²⁰ cm⁻³) pre-treated with nonoptimal Ar-plasma parameters. In this work, hundreds of β-Ga₂O₃ nanosheet devices were fabricated and analyzed. It was found that the optimized plasma pre-treatment (i.e., Ar plasma, 100 W, 60 s) induces Ohmic contact with great probability, especially for the heavily doped β-Ga₂O₃ crystals (N_e, ~10²⁰ cm⁻³). However, the corresponding drain current I_ds in this heavily doped channel cannot be effectively turned off by the bottom-gate voltage V_gs, as shown in Figure S6. This is due to the limited gate controllability of the bottom-gated device architecture for the heavily doped β-Ga₂O₃ channel. High-doping-level channel layers possess high carrier densities, which cannot be fully depleted by the applied bottom-gate bias. Further studies are needed to construct different transistor structures with stronger gate controllability by using this heavily doped β-Ga₂O₃ channel layer (N_e, ~10²⁰ cm⁻³) with argon plasma treatments.

To comprehensively analyze the surface defects generated by the Ar plasma treatment (100 W, 60 s), the dual-sweep measurement was performed on the β-Ga₂O₃ nanosheet FET at a V_ds of 1 V, as shown in Figure S7. Evident hysteresis (∆V) can be noticed, suggesting that a large amount of interfacial traps was introduced. The Ar-plasma bombardment has marked side effects on the transistor reverse characteristics. The results are in accordance with the work reported by Yang et al. [39] for the energetic proton irradiation. Moreover, micro–Raman measurements were carried out. Normally, Raman spectroscopy is sensitive to the biaxial strain at the surface region. Figure 7a compares the Raman spectra of the peeled β-Ga₂O₃ nanosheet on the SiO₂/Si (p²⁺) substrate before and after Ar plasma pre-treatments. Seven phonon peaks near 149, 172, 202, 349, 419, 627, and 674 cm⁻¹ were detected, corresponding to the ${\mathrm{B}}_{\mathrm{g}}^2$, ${\mathrm{A}}_{\mathrm{g}}^2$, ${\mathrm{A}}_{\mathrm{g}}^3$, ${\mathrm{A}}_{\mathrm{g}}^5$, ${\mathrm{A}}_{\mathrm{g}}^6$, ${\mathrm{B}}_{\mathrm{g}}^5$, and ${\mathrm{A}}_{\mathrm{g}}^9$ Raman-active modes. The peaks of the treated β-Ga₂O₃ nanosheet with Ar plasmas are consistent with that of the untreated one, which reveals that the defects induced by Ar-plasma bombardment did not bring an evident strain-related Raman shift. The Raman results are in agreement with the work reported by Yang et al. and Kwon et al. [17,39]. The potential reason is likely due to the strong particle radiation resistance of β-Ga₂O₃. Wide-band-gap semiconductors possess an intrinsic radiation hardness [42], particularly β-Ga₂O₃, due to its greater formation energy of Ga vacancy defect (53.3 eV), compared with 3C-SiC (3.3 eV for Si vacancy) and GaN (7.0 eV for Ga vacancy) [39,43]. The fluctuation of both base lines in the micro-Raman spectra is mostly due to the impact of the SiO₂/Si (p²⁺) substrate beneath the transferred β-Ga₂O₃ nanosheet. In addition, the micro–X-ray diffraction (micro–XRD) patterns of the β-Ga₂O₃ nanosheet before and after the Ar plasma treatments were also measured to evaluate the stress state of the material, as shown in Figure 7b. In the inset, the pattern exhibits three diffraction peaks located at 30.07°, 45.79°, and 61.69°, which is in good agreement with the monoclinic structure from the standard card (PDF#43-1012). The peaks can be respectively assigned to the (400), (600), and (800) crystal planes. After Ar plasma pre-treatment, the positions of the diffraction peaks nearly remained constant (i.e., 30.07°, 45.81°, and 61.73°), which reveals that the lattice strain induced by Ar-plasma bombardment was negligible.

4. Conclusions

In summary, four types of plasma treatments (i.e., gas mixtures SF₆/Ar, SF₆/O₂/Ar, SF₆/O₂, and Ar) were conducted to investigate their impacts on generating oxygen vacancies. High-resolution XPS was employed to examine the evolution of the surface chemical states of (100)-oriented β-Ga₂O₃ crystals. The optimal treatment process parameters (i.e., argon, 100 W, 60 s) were extracted, which can generate appreciable oxygen vacancies at the (100) β-Ga₂O₃ surface. With the optimized Ar plasma pre-treatment, the resultant (100) β-Ga₂O₃ nanosheet FET exhibits effective Ohmic contacts (i.e., R_C = 104 Ω·mm) without any post-metallization anneal, resulting in well-behaved electrical performance, such as a high I_ON/I_OFF ratio (> 10⁷), a steep subthreshold slope SS (249 mV/dec), and acceptable ${\mu }_{\mathrm{e}\mathrm{f}\mathrm{f}}$ (~21.73 cm² V⁻¹ s⁻¹). Lower contact resistance R_C (i.e., 5.2 Ω·mm) can be obtained by using heavily doped β-Ga₂O₃ crystals (N_e, ~10²⁰ cm⁻³) for the Ar plasma treatments. These results highlight its potential to improve the contact properties of Ga₂O₃-based electronic devices by using a facile, environmental, and controllable argon plasma treatment technique.