Experimental study of Ni/TiO₂/β-Ga₂O₃ metal–dielectric–semiconductor diodes using p-NiO junction termination extension

β-Ga₂O₃ is an ultrawide bandgap semiconductor (∼4.8 eV) with numerous merits that potentially surpass the material limits of other semiconductors for power electronic applications, namely, a high predicted critical field strength of 8 MV/cm². Such a large theoretical critical field strength leads to a very high Baliga figure of merit (BFOM),^2,3 which predicts a shorter and more highly doped drift region compared to devices with the same breakdown voltage (V_bk) on narrower bandgap materials. Thus, lower conduction losses than what many other semiconductors demonstrate can be achieved.² The availability of melt-grown substrates further reduces the projected cost of chips and provides a promising outlook for scalability to larger substrates.⁴ 

Currently, β-Ga₂O₃-based lateral devices have been extensively studied by numerous groups to understand the potential of β-Ga₂O₃ as the fundamental material for the next generation of power electronics.² Vertical device architectures have also been explored to fully utilize the advantages that β-Ga₂O₃ can offer.^1,5–26 The demand for high operating voltages over 1 kV and high current over 100 A with better form factor of overall circuits necessitates vertical device structures, as conventional lateral device structures are not ideal to achieve compact chip area utilization. In addition, vertical devices are favorable for maintaining high fields in the bulk and minimizing parasitic capacitances. This paper represents the latest results of vertical diodes from our group as part of efforts to advance the research in vertical devices.

Diodes are a fundamental application in which β-Ga₂O₃ can demonstrate its power handling capabilities. Researchers have implemented various means of improving the breakdown of diodes including heterogeneous integration with p-type materials⁶ and interface engineering with ultrathin dielectrics.^27,28 However, these devices resulted in higher turn-on voltage (V_on) and turn-on resistance, leading to increased conduction loss in order to achieve higher breakdown. The diode in this paper is a continuation of a novel vertical β-Ga₂O₃ diode that uses the high-permittivity (high-κ) dielectric TiO₂ in a metal–dielectric–semiconductor (MDS) structure to reduce the turn-on voltage and improve breakdown.²⁹ 

The MDS diode in this paper utilizes an ultrathin, high-κ TiO₂ layer to engineer the band diagram in the diode junction to achieve improved on and off state characteristics. Figure 1 shows the band diagram illustrations that depict the approximate effect of the TiO₂ dielectric interlayer during generalized zero bias and high reverse bias conditions. TiO₂ has a dielectric constant of ∼43 $ε 0$ and a conduction band edge that is approximately 0.3 eV lower than the conduction band edge of β-Ga₂O₃,³⁰ illustrated in the band diagram near the anode junction in Fig. 1. This negative conduction band offset does not impede forward current. However, the electric field in the dielectric and semiconductor are partitioned according to the inverse of the materials’ relative permittivities. Compared to the relative permittivity of β-Ga₂O₃ (12.4 $ε 0$⁠),³¹ the high relative permittivity of TiO₂ results in a much lower electric field in the dielectric relative to the β-Ga₂O₃. A wider tunneling barrier is, thus, formed due to the presence of high κ TiO₂, consequently reducing the transmission coefficient for the tunneling electrons responsible for large reverse leakage in SBDs. In addition, for a sufficiently thin dielectric layer, the MDS diode did not increase the R_on of the device.²⁹ 

However, breakdown in β-Ga₂O₃ typically occurs at electric fields lower than 8 MV/cm due to electric field crowding at the contact edge and high tunneling current under large reverse bias.³² A number of methods have been proposed and studied to improve breakdown voltage through field management, such as field plates, field rings, mesa-etch edge termination, junction termination extension, etc.³³ To further improve the breakdown voltage, we are reporting a novel integration of vertical heterojunction diode based on Ni/TiO₂/β-Ga₂O₃ with p-type NiO as the guard ring (GR). Since p-type doping in β-Ga₂O₃ is very challenging,³⁴ some low mobility p-type oxides have been recently integrated to create heterojunction diodes. Among those, p-type NiO demonstrates considerable improvement in device integration and performance.⁶ By using p-type NiO as the guard ring (GR) to the anode edge, the high electric field crowding at this region is alleviated and V_bk is increased.

Figure 2 illustrates the vertical β-Ga₂O₃ diodes that were fabricated in parallel on sister samples for this study. Those devices are SBD, TiO₂ MDS, SBD with GR (GR SBD), and MDS with GR (GR MDS). The β-Ga₂O₃ drift region grown by halide vapor phase epitaxy (HVPE) on a heavily Sn-doped (001) substrate was obtained from Novel Crystal Technology, Inc. (NCT). The manufacturer supplied a map of nine epitaxial thickness measurements across the substrate and reported 8.5-μm-thick epitaxy at the location used to fabricate these devices. The manufacturer also reports a nominal carrier concentration of 4.3 × 10¹⁸ cm⁻³ and a thickness of 623 μm for the substrate. A back-side cathode ohmic contact was formed by evaporated Ti/Au (20/380 nm) followed by 470 °C rapid thermal annealing for 1 min in nitrogen ambient. The sample was submerged in 100% reacted piranha etchant at room temperature for 90 s to clean the surface of the anode side before further deposition. Devices with 100-nm-thick NiO GR were fabricated by radio frequency (RF) sputtering the NiO at room temperature and lithographically patterned with lift-off. The target was 99.99% NiO ceramics with a 3 in. diameter, and the growth pressure was maintained at 7.4 mTorr in an Ar/O₂ mixed ambient with 20% O₂. RF power was fixed at 150 W/cm². The NiO growth rate was approximately 10.6 Å/min, and no postdeposition treatment was used. The hole concentration and mobility of the p-NiO were 1.1 × 10¹⁹cm⁻³ and ∼0.26 cm²/V s measured by Hall measurement performed on a Al₂O₃ witness sample that received the same deposition as the Ga₂O₃ devices. DC Hall measurements were performed at room temperature using a Nanometrics HL 5550 instrument with a magnetic field strength of 0.49 T. AC Hall measurement was also used to verify the p-type and the mobility. AC Hall measurements were performed at room temperature at 100 MHz using a Lakeshore 8407 instrument with a magnetic field intensity of 1.2 T. Once the GR was created, a thin TiO₂ layer (42 Å) was deposited on the MDS sample by atomic layer deposition (ALD) at 250 °C with O₂ plasma source. The thickness of the TiO₂ was measured on a silicon witness sample using thin-film ellipsometry. The TiO₂ was then etched back using CF₄ plasma in reactive ion etch (RIE) to define the anode contact pattern for the GR MDS and the TiO₂ MDS. The Ni/Au (20/300 nm) anode was deposited via electron beam evaporation and patterned using a liftoff process for all devices. Devices were isolated through mesa etch using Ar/BCl₃ gases in inductively couple plasma (ICP) etcher to provide edge termination and isolation to all SBDs and MDS diodes. The devices have circular contacts (D = 100 μm) with an additional 5 μm GR.

All SBD and MDS diodes were fabricated from the same wafer as highlighted above for evaluation. Current density-voltage (J-V) and capacitance-voltage (C-V) characteristics (1 MHz) were measured using an Agilent B1505A. Reverse J-V was measured with the sample submerged in Fluorinert FC-77 liquid dielectric, also using an Agilent B1505A.

Understanding the doping concentration of the material is important in diode studies to understand the electric field at breakdown. Figure 3 shows the C-V and 1/C²-V plots of an SBD measured at 1 MHz. The depth profile of N_D-N_A in the drift layer was also extracted from the slope of 1/C² in C-V measurements.³⁵ The average net donor concentration of the HVPE epilayer was extracted as 2.6 × 10¹⁶ cm⁻³ from the slope of 1/C². The Schottky barrier height (SBH) extracted from the C-V measurement was 1.2 eV.

Among all four diodes, result for the TiO₂ MDS with GR yields 514 MW/cm² in Baliga’s power figure of merit (BFOM), exceeding the previous, unterminated MDS diode on β-Ga₂O₃ by our group.²⁹  Figure 6 shows BFOM benchmarking plot of all diodes presented in this work with literature.^6–26 However, MDS diodes in this paper have demonstrated low V_on. To understand the important contribution of improved V_on to conduction losses, an effective R_on,sp is calculated from the ratio of V_F/J_F at the forward power density of 100 W/cm². With lower V_on, a diode can reach a higher current density at lower voltage under this power limit. The diodes in this study are benchmarked for a power dissipation of 100 W/cm² with minimum possible effective R_on,sp along with their V_bk in Fig. 7. At higher power density levels, the losses from R_on,sp will dominate compared to the V_on losses, and the material limits will more closely resemble the traditional BFOM values. The 100 W/cm² power density is considered as a potential benchmark for industrial applications and to highlight the operating regime where V_on has a significant impact on effective resistance. We use β-Ga₂O₃ and SiC with the same V_on as the diodes in this paper to illustrate the theoretical limits of device operation. In Fig. 7, solid lines represent β-Ga₂O₃ devices and dashed lines represent 4H-SiC devices. Between SBD and TiO₂ MDS, TiO₂ MDS marks itself with low V_on and high V_bk. The GR MDS diode (V_on = 0.8 V) exceeds the limits of β-Ga₂O₃ with a V_on of 1.1 V and 4H-SiC with a V_on of 0.8 V under this power limit. While TiO₂ MDS consistently exhibits low V_on, continued advances in field management integration with TiO₂ MDS will be important to push the device performance toward these limits by achieving higher breakdown.

This work demonstrates the continuous improvement of vertical TiO₂ MDS β-Ga₂O₃ diodes with the field management technique by incorporating p-NiO junction termination extension. Forward J-V characteristics confirm the lower turn on voltage from TiO₂ MDS versus a regular SBD, which contributes to a 30% lower effective R_on,sp at a power density limit of 100 W/cm² in the TiO₂ MDS than a regular SBD. Reverse J-V characteristics reveal that with p-NiO as the guard ring to reduce the field crowding around the anode, the breakdown voltage can improve across all devices. The breakdown voltage of 1776V (3.7 MV/cm) was recorded for the GR MDS versus only 826 V (2.5 MV/cm) for the GR SBD. Such improvement has placed the diodes in advantageous positions in practical power operations, while demonstrating the importance of field management research in devices based on ultrawide bandgap materials for power electronics. Additionally, the 5 μm GR length presented here was not thoroughly optimized and selected from an initial experiment based on prior work from literature. Optimizing the GR length with electric field simulation offers a path to maximize the field management benefits we demonstrate. The proposed benchmark provides a new way to assess power devices practically. By considering the specific requirement of the application, we could evaluate devices accordingly to make the best selection. Among all diodes presented in this paper, the TiO₂ MDS with guard ring possesses the best opportunity to deliver optimal performance. It surpasses the limit of 4H-SiC devices and approaches the theoretical limit of β-Ga₂O₃ when accounting for V_on and a power density limit of 100 W/cm². This work provides a starting point for researchers to realize the true benefits of β-Ga₂O₃ in practical applications.

The authors would like to acknowledge Kevin Leedy, Andrew Browning, and Jason Edwards for expert operation and assistance with the AFRL Sensors Directorate cleanroom equipment and Timothy Cooper and Stefan Nikodemski for assistance with Hall measurements.

The authors have no conflicts to disclose.

Jeremiah Williams: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Project administration (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Weisong Wang: Conceptualization (equal); Investigation (equal); Writing – original draft (equal); Writing – review & editing (equal). Nolan S. Hendricks: Conceptualization (lead); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Project administration (equal); Writing – review & editing (equal). Aaron Adams: Investigation (supporting). Joshua Piel: Investigation (supporting); Writing – review & editing (supporting). Daniel M. Dryden: Validation (supporting); Writing – review & editing (equal). Kyle Liddy: Investigation (supporting); Methodology (supporting); Resources (supporting). Nicholas Sepelak: Investigation (supporting); Resources (supporting). Bradley Morell: Investigation (supporting). Ahmad Islam: Funding acquisition (equal); Project administration (lead); Supervision (equal); Validation (supporting); Writing – review & editing (supporting). Andrew Green: Funding acquisition (equal); Project administration (equal); Resources (equal); Supervision (equal); Writing – review & editing (supporting).

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