Editors' Choice Communication—A (001) β-Ga₂O₃ MOSFET with +2.9 V Threshold Voltage and HfO₂ Gate Dielectric

A thin sliver of Ga₂O₃ was cleaved along the (001) face of an on-axis (−201), non-intentionally n-type doped (∼3 × 10¹⁷ cm⁻³) commercial Ga₂O₃ substrate (Tamura Corporation) produced by the edge-defined film fed (EFF) growth method. The cleaved material was verified to be of the (001) orientation by X-ray diffraction (XRD, Fig. 1), and was further thinned by repeated exfoliation using Nitto Denko Revalpha thermal release tape.¹² The target SiO₂/Si substrate was cleaned in a SAMCO UV-1 ozone system for 10 min. at room temperature. The tape was then pressed onto the substrate and upon a final exfoliation the transferred Ga₂O₃ flake was approximately 300 nm thick and 11.7 μm wide. Ohmic contacts of Ti/Au (20/200 nm) were lifted off on both ends of the sample, followed by atomic layer deposition (ALD) of a 42 nm thick HfO₂ gate dielectric, and a 20/200 nm thick Ni/Au gate lifted off using standard photolithography techniques. No implantation or rapid-thermal annealing were performed, and cleaning prior to metal and dielectric deposition consisted of a 10 min. UV-O₃ treatment. The as-processed transistor, depicted in the inset of Fig. 1, had a gate length of 3 μm, a gate-source spacing of 2 μm, and a gate-drain spacing of 5 μm.

Zoom In Zoom Out Reset image size

Capacitance-voltage and current-voltage measurements were performed using a Keithley 4200 and a HP4145b analyzers, respectively. Hall effect measurements were performed using a closed-circuit He-cooled Lakeshore Hall probe station with a superconducting magnet (±0.5 T magnetic field used in this work). The electron spin resonance (ESR) measurements were done at 300 K with a commercial (E-300) Bruker Biospin 9.5 GHz spectrometer. Estimates of the spin concentrations were made from double numerical integration of fits to the ESR lineshapes and the use of a well-calibrated P-doped Si standard.

Gate capacitance-voltage characteristics measured as a function of frequency and voltage with the source-drain contacts connected together is presented in Fig. 2. The inhomogeneity of the depletion capacitance was indicative of traps present at the HfO₂/Ga₂O₃ interface. The flatband voltage, determined by the flatband capacitance C_FB = 0.8 C_MAX, was approximately V_FB ≅ +2 V for the 300 kHz measurement.¹⁵ To our knowledge, this is the first report of a positive threshold voltage measured for a β-Ga₂O₃ MOS interface, and presently speculate that it originated from a larger conduction band offset between HfO₂ and β-Ga₂O₃, beyond the values of 3.1 ± 0.2 eV and 1.5 ± 0.2 eV reported for the SiO₂ and Al₂O₃ interfaces with Ga₂O₃, respectively.^13,14 Further X-ray photoelectron spectroscopy measurements will be performed to accurately determine the band offsets for this novel dielectric interface.

Zoom In Zoom Out Reset image size

The static dc current-voltage measurements of the as-fabricated MOSFET are presented in Fig. 3. A maximum drain-source current (I_DS,SAT) of about 11.1 mA/mm and an on resistance (R_DS,ON) of about 818 Ω-mm were measured in the on-state at a forward gate bias of +18 V (Fig. 3a). Significant off-state current leakage was present for gate biases of 0 V and below, and could originate from the traps at the HfO₂/Ga₂O₃ or the Ga₂O₃/SiO₂ interfaces, or from drain-induced barrier lowering (DIBL) owing to the absence of n⁺ doped source/drain regions. The positive V_T, observed at 100 mV drain bias when I_DS was measured as a function of V_GS in Fig. 3b, was determined to be +2.9 V using a linear fit of the I_DS. Increasing the drain bias clearly lowered the barrier to electron accumulation and pushed the device back toward depletion mode. The peak transconductance at V_DS = 1 V was 0.18 mS/mm, and a field effect mobility of 0.17 cm²/Vs was calculated using the relationship μ_FE = g_m·(L/W)/(C_OX·V_DS).¹¹ Fig. 3c shows that the gate current remained low (10–100 pA) up to the maximum non-destructive drain bias applied to the device. At 80 V, the off-state drain current significantly increased, however, this non-catastrophic breakdown was gate bias dependent and was thus not determined by the theoretical critical field of Ga₂O₃ (8 MV/cm).

Zoom In Zoom Out Reset image size

Due to the restrictions posed by the mechanical exfoliation process, it was not possible to measure device parameters such as channel electron density and contact resistance, which would necessitate additional test structures (van der Pauw, TLM). In order to mitigate this limitation, temperature-dependent Hall effect measurements were performed on a square 8 × 8 × 0.68 mm sample of (−201) Ga₂O₃ with evaporated Ti/Au contacts in each corner and data is presented in Fig. 4a. To-date, Hall data has been reported by Parisini and Fornari for Czochralski-grown Ga₂O₃, and by Oishi and coworkers for EFF-grown Ga₂O₃.^16,17 Both Parisini and Oishi reported peak mobilities around 85 K with respective values of ∼600 cm²/V·s and 886 cm²/V·s, respectively. However, for the sample with a similar doping level to ours (3 × 10¹⁷ cm⁻³ at 300 K), Parisini measured a peak Hall mobility of about 300 cm²/V·s at 100 K. Our measurements indicated a peak mobility of 348 cm²/V·s at 140 K, which was consistent with the reported limits by optical phonon scattering at high temperature and ionized impurity scattering at low temperature.

Zoom In Zoom Out Reset image size

In order to investigate the origin of electrons in the Ga₂O₃ substrate, electron spin resonance (ESR) was performed at room temperature on a 7.5 × 1 × 0.18 mm flake of (001) Ga₂O₃ cleaved from the (−201) substrate. The data is shown in Fig. 4b for the case when the magnetic field was parallel to the 〈010〉 and the 〈001〉 directions. The line width of about 0.2 G was similar to the one reported by Yamaga et al. for their undoped Ga₂O₃ bulk crystals.¹⁸ Our analysis yielded a density of oxygen vacancies (V_O⁺) of 2.3 × 10¹⁷ (±50%) cm⁻³, indicating that other residual impurities in the substrate (e.g., Si) were not a major contributor to the unintentional doping of the substrate. While we have previously reported the presence of O vacancies in MOCVD-grown β-Ga₂O₃,¹⁹ here we note that their presence enabled transistor operation.

Initial results on an insulated-gate (001) β-Ga₂O₃ MOSFET with an ALD-grown HfO₂ high-k gate dielectric were reported. A positive threshold voltage for a Ga₂O₃-based transistor was reported for the first time. Hall effect and ESR measurements indicated that the conductivity was due primarily to the presence of oxygen vacancies in the substrate. Capacitance-voltage and current-voltage measurements showed a positive threshold voltage at low drain bias. It remains to be confirmed that a high-k dielectric gated transistor fabricated on homoepitaxial Ga₂O₃ would yield an enhancement mode device. Further device performance improvements such as higher breakdown field and output current are expected by transitioning the process to homoepitaxial Sn-doped Ga₂O₃. A thorough understanding of the role of point defects such as oxygen vacancies in the donor activation and device performance remains to be developed.

The authors are sincerely grateful to Dr. Jeremy Robinson and Dr. Joshua Caldwell (NRL) for assistance with Ga₂O₃ exfoliation, Dr. Akito Kuramata (Novel Crystal Technology) for insightful discussions, and the NRL Nanoscience Institute staff ( Walter Spratt, Dean St. Amand) for fabrication equipment maintenance. Research at NRL was supported by the Office of Naval Research (ONR).