Ultrawide bandgap LiGa₅O₈/β-Ga₂O₃ heterojunction p–n diodes

β-phase gallium oxide (β-Ga₂O₃) represents an ultrawide bandgap (UWBG) (E_g ∼ 4.9 eV) semiconductor with a critical electric field strength of 7–8 MV/cm.^1,2 Along with its advantageous fundamental material properties and the availability of high quality native Ga₂O₃ substrates, the material shows great promise in realizing commercially viable power devices with efficiencies far superior to current Si or SiC based technologies.¹ Significant improvement to the Baliga’s figure of merit (BFOM) by 2–3 times when compared to wideband gap GaN or SiC is predicted for Ga₂O₃-based devices.¹ 

While vertical Schottky-based β-Ga₂O₃ devices have been demonstrated with breakdown voltages of up to 6 kV with a BFOM recorded between 7.4 and 10.6 GW/cm² for vertical devices³ and >10 kV for lateral Schottky devices,⁴ a key challenge in adopting β-Ga₂O₃ for power devices is the lack of an effective dopant that can reliably achieve p-type conductivity. Bipolar conduction provides a means to further expand the possible devices achievable with β-Ga₂O₃ and to achieve BFOM values closer to the true potential of β-Ga₂O₃. To fully utilize the key advantage of a high breakdown field provided by β-Ga₂O₃, bipolar devices are a critical requirement with the ability to operate up to, or close to, the true critical electric field strength of β-Ga₂O₃. Since p-type conductivity is not possible in β-Ga₂O₃, previous work has shown the utilization of other p-type materials in combination with n-type β-Ga₂O₃ to yield bipolar heterostructure devices.^5–7 For example, p-NiO/n-Ga₂O₃-based heterostructure diodes have recently been demonstrated.^8,9 While significant advances and improvements have been made in improving the overall device performance and device design for NiO_x/Ga₂O₃-based devices, the extracted electric field strength within β-Ga₂O₃ at device breakdown is still significantly lower and far below what is theoretically predicted.

Recently, an ultrawide bandgap material, LiGa₅O₈, grown via Mist-CVD (M-CVD), has shown p-type conductivity from both room-temperature and temperature-dependent Hall measurements.¹⁰ The bandgap of this single-crystal material has been experimentally measured to be ∼5.36 eV. The p-type conductivity in combination with the ultrawide bandgap is extremely promising for power electronic device applications. In this work, the demonstration of a p–n heterostructure with an ultrawide bandgap p-type LiGa₅O₈ on (010) n-type β-Ga₂O₃ is presented.

The device structure started with metalorganic chemical vapor deposition (MOCVD) growth of the n-type, (010) β-Ga₂O₃ layers.^11,12 Before growth, the (010) Sn-doped conductive substrate was solvent cleaned with acetone, isopropyl alcohol, and de-ionized water. This was followed by the growth of a 200 nm Si-doped n⁺ β-Ga₂O₃ layer with a targeted doping concentration of 5 × 10¹⁸ cm⁻³. Following the n⁺ layer, a 1.2 μm Si-doped n⁻ β-Ga₂O₃ layer was grown with a targeted N_d–N_a concentration of 5 × 10¹⁶ cm⁻³. Both layers were grown utilizing a constant TEGa molar flow rate of 31.06 μmol/min at 60 Torr chamber pressure and a growth temperature of 900 °C. An oxygen flow rate of 800 SCCM was utilized for both β-Ga₂O₃ layers, with an estimated growth rate of ∼750 nm/h. The VI/III ratio of 1150 was utilized for both β-Ga₂O₃ layers (n⁺ and n⁻). Following the MOCVD β-Ga₂O₃ growth on the Sn-doped (010) β-Ga₂O₃ substrate, the p-type LiGa₅O₈ layer was grown via M-CVD. Immediately before M-CVD growth, the sample was cleaned with acetone, isopropyl alcohol, and de-ionized water. The sample was transferred between the chambers in a standard, closed sample holder. The LiGa₅O₈ film was grown at 850 °C for 30 min utilizing 0.05M (where 1M corresponds to 1 mol of solute in 1 l of solvent) of gallium acetylacetonate and 0.03M of lithium acetoacetate as Ga and Li precursors, respectively. Ar was utilized as the carrier gas, while O₂ was flowed during growth as a source for O. The grown LiGa₅O₈ film was ∼60 nm thick. Further details on the M-CVD growth of p-type LiGa₅O₈ can be found in previously published work.^10,13

The device fabrication started with protecting the top surface with photoresist followed by etching (BCl₃/Ar-based inductively coupled plasma reactive ion etching (ICP-RIE)) the bottom of the substrate for ∼100 nm to expose etched β-Ga₂O₃ for metal deposition of a common bottom electrode. After a 30-s diluted HCl dip, 30 nm of Ti and 100 nm of Au (as the n-side contact) were deposited by e-beam evaporation. After metal deposition, the sample was annealed at 470 °C in an N₂ ambient for 1 min. Previous work has shown this metallization stack and annealing condition to be a reliable, repeatable, and effective method to achieve ohmic contacts to n-type β-Ga₂O₃.¹⁴ Device areas were then patterned and defined through an isolation etch. A modified BCl₃/Ar-based ICP-RIE recipe at an etch rate of ∼15.3 nm/min was utilized to etch the LiGa₅O₈ p-type layer, followed by further etching of 140 nm into the n-type β-Ga₂O₃ layer.¹⁵ Device diameter defined through the isolation etch was 210 μm. Samples were patterned for p-contact metal liftoff. Pt/Ni/Au was deposited with 20, 30, and 80 nm thicknesses, respectively, through e-beam evaporation. The sample was annealed in N₂ ambient for 1 minute at 375 °C. Followed by the p-contact annealing, a Ti/Au (20 nm/200 nm) metal overlay was patterned through liftoff on top of the p-type contact. The complete process flow is shown in Fig. 1.

The forward and reverse current–voltage (I–V) measurements were performed using a Keithley Instruments 2400 source measurement unit, while capacitance–voltage (C–V) measurements were conducted on an HP Agilent 4284A LCR-meter. High-angle annular dark field (HAADF) scanning transmission electron microscopy (STEM) was performed on a Thermo Scientific Themis Z S/TEM with an acceleration voltage of 300 kV.

The top-side p-type ohmic contact formation was verified through I–V sweeps on linear transmission line measurement (L-TLM) structures fabricated on the same sample. Figure 2(a) shows the I–V data as measured from −3 to 3 V on the sample as-deposited (without annealing). Almost all contacts show linear I–V relationships, with slight non-linearity observed at a gap spacing of 15 μm. The device sample was then annealed at 375 °C for 1 min in N₂ ambient in a rapid thermal anneal (RTA) system. Figure 2(b) shows the I–V data as measured from −3 to 3 V after annealing. The contacts remain ohmic (linear) in nature, with an improvement in the current levels (thus a reduction in the contact resistance) observed for a given voltage. With a 7 μm contact spacing distance, at 3 V, the current was 1.44 × 10⁻⁵ A before annealing. The current increased to 4.77 × 10⁻⁵ A after annealing for the same contact spacing distance and voltage. This represents an improvement in current conduction of ∼3.3 times at this condition and spacing after annealing as compared to before annealing. As observable from the I–V curves presented in Figs. 2(a) and 2(b), while the contacts achieved were ohmic in nature, the change in resistance with gap spacing does not remain consistent. This is likely due to the local sheet resistance of the p-type LiGa₅O₈ thin film having non-uniformity, preventing accurate contact resistance extraction in both scenarios by typical L-TLM analysis. To provide a range for the specific contact resistance, the I–V values at +2 V with gap spacings from 6 μm up to 12 μm on the L-TLM structures were analyzed. To provide an estimated range of the specific contact resistance, hall measurement results of a p-type LiGa₅O₈ sample grown on an Fe-doped β-Ga₂O₃ substrate at the exact same condition were utilized to provide an estimate of the sheet resistance of the semiconductor layer. Under the growth conditions utilized, for samples grown on Fe-doped β-Ga₂O₃, Hall measurements yielded carrier concentrations between 4.4 × 10¹⁸ and 1.1 × 10¹⁹ cm⁻³ with the corresponding hole mobilities between 0.84 and 0.2 cm²/V s, respectively. The sheet resistance measured from the Hall sample was 4.96 × 10⁵ Ω/sq. Based on this, the contact resistance associated with the two metal contacts for each set of measurements was extracted with the assumption that the average resistance of the semiconductor layer for every spacing should be equal to the value obtained through Hall measurements. Table I(a) provides the summary of specific contact resistance obtained for each gap spacing at +2 V bias after metal deposition (as-deposited) and after annealing. Reductions are noted from the specific contact resistance measurements in all cases after annealing, with up to 10.7× decrease observed for a gap spacing of 6 μm. Table I(b) shows the summary of the specific contact resistance extracted for the cases before and after annealing. Overall, an average decrease by 6.8× is observed for the extracted specific contact resistance after annealing, suggesting a marked improvement under the annealed condition. The standard deviation also reduced, suggesting that better uniformity is obtained at the metal–semiconductor interface post-annealing as compared to before annealing. Similar issues in contact resistance extraction for ohmic p-type contacts to ultrawide bandgap AlN have also been shown in the literature, possibly also suggesting the difficulty in obtaining good ohmic p-type contacts for materials with such large bandgaps.¹⁶ While the p-type contact resistivity cannot be extracted reliably through L-TLM measurements on this device sample, the results do indicate that the p-type contacts deposited were ohmic in nature and that the annealing performed improved the overall metal–semiconductor contact performance.

I–V sweeps performed on devices show rectifying characteristics, typical characteristics of a diode. Figure 3(a) shows the linear scale I–V sweep of a representative 210 μm diameter device, while the same sweep is shown in semi-log scale in Fig. 3(b) from −4 to 10 V. The reverse leakage current at −4 V is at or below the noise level of the measurement system. A knee voltage of 5.46 V and an ideality factor (n) of 2.78 were extracted from the I–V curves. Previously investigated conduction and valence band offset determination through x-ray photoelectron spectroscopy (XPS) confirmed that LiGa₅O₈/β-Ga₂O₃ has a type-II (staggered) heterojunction. From the XPS results, p-type LiGa₅O₈ was measured to have a conduction band offset (ΔE_C) of 1.26 eV and a valence band offset (ΔE_V) of 1.86 eV to β-Ga₂O₃.¹³ Using these band offset values and a bandgap of 5.4 eV for LiGa₅O₈ and 4.8 eV for β-Ga₂O₃, technology computer-aided design (TCAD) simulated knee voltage ranges between 5.8 and 6.2 V. These simulated values may have certain level of errors since not all parameters are currently known for LiGa₅O₈, and values (e.g., electron and hole mobility and hole effective mass) were assumed to best effort (supplementary material, Table T1 and Figs. S1 and S2). Figures S1(a) and S1(b) show the TCAD simulated I–V curves. Figures S2(a)–S2(d) show the band diagrams at different forward bias conditions. From the band diagrams, it is obvious that sufficient voltage bias of ∼6 V is required for the bands to align such that the barrier and band-offset between LiGa₅O₈ and Ga₂O₃ are overcome, and thus, forward current conduction is possible.

C–V measurements were carried out at 1 MHz frequency on a representative 210 μm diameter device, as shown in Fig. 4(a). Both electron concentration in the n-type β-Ga₂O₃ region and hole concentration in the p-type LiGa₅O₈ region were extracted based on the C–V measurement. At low reverse bias (0 V to −2.5 V), a change in the slope of 1/C² against V is observed, suggesting changes in depletion width on both sides of the junction. At higher reverse biases, the slope of 1/C² against V remains relatively constant. This implies that the p-type region is fully depleted at the high bias regime (considering the thickness of 60 nm), while further depletion continues within the n-type region. The electron and hole concentrations of the n-type and p-type regions, respectively, were extracted utilizing the high bias regime (−7 to −10 V) of the C–V measurements. Under the high reverse-bias regime, the p-type region is fully depleted, contributing a constant capacitance. The constant capacitance is in series with the varying capacitance associated with the n-type region as it continues to further deplete with the increasing reverse bias. An equal dielectric constant for both materials was assumed for extraction. For a LiGa₅O₈ layer of 60 nm and an assumed dielectric constant equal to that of β-Ga₂O₃, the capacitance at a high reverse bias of the fully depleted LiGa₅O₈ layer was calculated. An analytical equation was derived for the p–n junction to correlate the C–V measurements with the doping densities on either side of the junction. Fitting the measured data with the analytically derived equation [inset of Fig. 4(a)] yields an extremely close fit (R² ∼ 0.9996), with an extracted hole concentration of 1.27 × 10¹⁸ cm⁻³ for the p-type LiGa₅O₈ layer, while the n-type β-Ga₂O₃ region was extracted to have an electron concentration of 7.18 × 10¹⁶ cm⁻³. Multiple devices measured by C–V and extracted via the same methodology yielded hole concentrations between 9.85 × 10¹⁷ and 1.44 × 10¹⁸ cm⁻³.

The built-in voltage is extracted through linear extrapolation of the region between 0 and −2 V. At this bias range, as previously explained for electron and hole concentration extraction, the depletion occurs on both sides (the p-type region and the n-type region). Above this voltage range, depletion is considered to occur only on the n-type side, while the p-type region, due to its limited thickness, is considered fully depleted. Thus, for the extraction of the built-in voltage, the slope is extrapolated from the bias range between 0 and −2 V. Through this linear extrapolation of the 1/C²–V curve, a built-in voltage of 5.56 V is extracted.

Reverse biased I–V sweeps of devices show complete physical breakdown above 100 V, as shown in Fig. 5. Majority of the devices measured showed physical breakdown originating from edges, suggesting that peak electric field build-up on the edges of the devices caused premature device breakdown while also showing damage within the centers of the devices. For majority of devices, reverse leakage current remained at, or near, the noise level of the measurement system up to ∼−23 V. Reverse leakage current then started to increase with increased voltage. A sudden, sharp increase in leakage current was observed between −60 and −80 V on all devices. Localized, small circular areas on devices showed physical damage when observed under an optical microscope at this voltage range. The resultant damage creates a shunt path that causes the increase and surge in the leakage current within the voltage range. The origin of this shunt path is likely local defects, either within the material or at the heterojunction interface that propagates further at high biases. Based on the previously extracted specific contact resistance associated with the metal–semiconductor contact, majority of the reverse biased voltage is attributed to the voltage drop across the diode with almost negligible series resistance associated with the metal–semiconductor contacts. Under this condition, the critical electric field strength at which the shunt path forms is between 4.76 × 10⁵ and 6.35 × 10⁵ V/cm. Further optimization and testing would be required to determine the exact field strength at which this shunt path begins to form and the origin of the defects. Complete physical breakdown of devices occurs at above −100 V reverse biases, with most devices showing breakdown from the edges. It is possible in this scenario that continued electric field build-up on the edges causes local defects present at the edge of the devices to also reach the critical field strength associated with the defects and cause the device to breakdown prematurely, adding to the shunt paths already present within the center of the devices. Further investigation would be required to determine the exact causes, nature, and field strength associated with the premature breakdown of these devices and this heterojunction.

HAADF-STEM imaging shows the presence of a high-quality interface between p-type LiGa₅O₈ and n-type β-Ga₂O₃ while also highlighting the clear difference in crystal structure between these two materials. Analysis of the metal layers and the metal–semiconductor interface also revealed no significant alloying or diffusion of the metals into the semiconductor layer. This further confirmed that the diode behavior can only be attributed to the presence of the p–n heterojunction between the two materials. Individual layers of the p-contact metal stack were distinguishable in the large area STEM cross-section, as shown in Fig. 6(a). High quality, monocrystalline nature of the p-type LiGa₅O₈ can be seen in high-resolution STEM imaging as evidenced in Fig. 6(b) and further confirmed to be distinct from the underlying crystalline β-Ga₂O₃ layer. Nanodiffractions of the LiGa₅O₈ and β-Ga₂O₃ layers further confirmed the monocrystalline structures of the two layers, as shown in Figs. 6(c) and 6(d).

STEM imaging along the [001] β-Ga₂O₃ zone-axis reveals a rougher surface morphology as shown in large-area imaging in Fig. 7(a). Along this orientation, the presence of extended defects across the bulk of the LiGa₅O₈ film is also evident. Detailed imaging in two separate areas of interest is shown in Figs. 7(b) and 7(c). Figure 7(b) shows extended dislocation defects present within the bulk of the LiGa₅O₈ film that extend toward the top surface of the material. Figure 7(c) shows the dislocation defects that seem to originate at the heterointerface between LiGa₅O₈ and β-Ga₂O₃. This suggests that the defects originate from the interface and continue to propagate throughout the subsequent film growth. In addition, the rougher surface morphology of the film in this zone-axis orientation is in stark contrast to the smooth, uniform profile shown in Fig. 6(a), suggesting that the surface roughness is likely dependent on lattice mismatch in one axial orientation between the two materials. The presence of these extended defects could be a significant factor contributing to the electrical performance and characteristics demonstrated. The extended defects can also be a likely cause of increased diode non-ideality and the possible pathway through which early on-set breakdown occurs. As proposed through the breakdown mechanism explained before, the presence of these defects likely leads to the early breakdown of the devices, with larger localized electric fields possibly being channeled through them. Further studies into the growth factors influencing and affecting the presence of these extended defects and optimizing conditions could possibly help yield better electrical results and help improve the ideality factor of these devices in the future.

A p–n heterojunction diode based on ultrawide bandgap p-type LiGa₅O₈ and n-type β-Ga₂O₃ was demonstrated in this work. Evidence of ohmic p-type contacts to p-LiGa₅O₈ was shown in conjunction with rectifying characteristics of the measured diodes. A turn-on voltage of 5.46 V with an ideality factor of 2.78 was extracted from the UWBG heterojunction diode. With further understanding of the recently discovered p-type UWBG LiGa₅O₈ in conjunction with optimization and improvements in processing conditions and device design, it is promising that the overall performance of the diodes can be significantly improved and provide a foundation for high-power device applications that fully utilize the potential of the β-Ga₂O₃ material platform.

See the supplementary material for the TCAD simulation of the β-Ga2O3 and LiGa5O8 heterojunction diodes and the analytical model for Nd and Na concentration extraction.

The authors acknowledge the funding support from the Air Force Office of Scientific Research (AFOSR, Grant No. FA9550-23-1-0142, Dr. Ali Sayir).

The authors have no conflicts to disclose.

Vijay Gopal Thirupakuzi Vangipuram: Conceptualization (equal); Data curation (equal); Formal analysis (lead); Investigation (lead); Methodology (equal); Validation (equal); Visualization (equal); Writing – original draft (lead); Writing – review & editing (equal). Kaitian Zhang: Conceptualization (supporting); Data curation (supporting); Investigation (equal); Methodology (equal); Validation (supporting); Writing – original draft (supporting). Dong Su Yu: Investigation (supporting); Methodology (supporting); Validation (supporting). Lingyu Meng: Investigation (supporting); Methodology (supporting); Validation (supporting); Visualization (equal). Christopher Chae: Investigation (supporting); Methodology (supporting); Validation (supporting); Visualization (equal). Yibo Xu: Data curation (supporting); Methodology (supporting); Validation (supporting); Visualization (supporting). Jinwoo Hwang: Funding acquisition (equal); Investigation (equal); Methodology (equal); Project administration (equal); Resources (equal); Supervision (equal); Validation (equal); Visualization (equal); Writing – original draft (supporting). Wu Lu: Formal analysis (equal); Funding acquisition (equal); Investigation (supporting); Methodology (equal); Project administration (equal); Resources (equal); Supervision (equal); Validation (equal); Visualization (equal); Writing – review & editing (equal). Hongping Zhao: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Funding acquisition (lead); Investigation (equal); Methodology (equal); Project administration (lead); Resources (equal); Supervision (equal); Validation (equal); Visualization (equal); Writing – original draft (supporting); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.