Three-step field-plated β-Ga₂O₃ Schottky barrier diodes and heterojunction diodes with sub-1 V turn-on and kilovolt-class breakdown

Over the past decade, β-Ga₂O₃ has emerged as a highly promising ultra-wide bandgap semiconductor for applications in high-power electronics, owing to its high critical breakdown field (∼8 MV cm⁻¹) and ease of n-type doping. ^1–5) Additionally, large-area single crystal substrates are available, which can be grown using bulk crystal growth techniques such as Czochralski and edge-defined film-fed growth. ^3,6) The availability of bulk substrates has been leveraged to grow high-quality β-Ga₂O₃ epilayers with low defect density, ^6–8) enabling rapid progress in the performance of both vertical and lateral high-voltage devices. In particular, β-Ga₂O₃ two-terminal devices have demonstrated remarkable improvements in performance, with recent reports exceeding the figure of merit of GaN. ^9,10) Among two-terminal devices, p-NiO/β-Ga₂O₃ heterojunction devices showcase the best reported performance, enabling the demonstration of ampere-class devices. ¹¹⁾

To enhance the breakdown voltage characteristics of β-Ga₂O₃ vertical devices, field termination structures are critical for reducing electric-field crowding at the edge of the device. Several field termination structures have been demonstrated in β-Ga₂O₃, including field plates, ion-implanted termination, mesa etch termination, guard rings (p-NiO), and junction termination extensions. ^12–18) Among these field termination structures, field plates are the easiest to fabricate and are compatible with most device structures, requiring only low-temperature processing conditions. Excellent performance in terms of breakdown voltage has been achieved by implementing field plates in vertical β-Ga₂O₃ devices with demonstrations of breakdown voltage exceeding 2 kV. ^12,13) However, the reverse breakdown characteristics of field-plated devices could be further improved by combining field plates with other compatible field termination structures.

In this work, we investigate the integration of field plates with ion-implant termination for improving the breakdown performance of vertical Ga₂O₃ diodes [see Figs. 1(a), 1(b)]. A three-field plate structure is chosen to lower the peak electric field at the edge of the Schottky metal as well as the field-plate dielectrics. We used low-stress SiO_x N_{1- x} (SiON) as the field-plate dielectric, which allows deposition of thick layers (∼2 μm total) without delamination or cracking. To fully reduce peak electric fields at the edge of the Schottky metal, it is critical that the first field-plate (FP1) dielectric is thin and of high quality. To achieve this, high-quality Al₂O₃, deposited using plasma-assisted atomic layer deposition (PEALD), is used to form the FP1 dielectric as well as to encapsulate the SiON dielectric [Figs. 1(a), 1(b)]. To further reduce field crowding, nitrogen (N₂) ion implantation is used to create implant damage below the field plates as shown in the device schematic [Figs. 1(a), 1(b)]. In addition to Schottky barrier diodes (SBDs), p-NiO-based heterojunction diodes (HJDs) were also fabricated using the same field termination.

Zoom In Zoom Out Reset image size

The design of field plates was optimized using a combination of TCAD device simulations and by considering processing constraints. For the simulations, a drift region thickness of 10 μm and doping of 1 × 10¹⁶ cm⁻³ was assumed. Furthermore, a dielectric constant of 5 [in between that of SiO₂ (3.9) and SiN_x (7.5)] was assumed for SiON and a value of 9 was assumed for Al₂O₃. The thickness of the Al₂O₃ field plate (t₁) was chosen to be 50 nm, as TCAD simulations showed complete transfer of the peak electric field from the edge of the Schottky metal to the corner of FP1 for t₁ = 50 nm (see Supplementary data, Fig. S1). The total thickness of the SiON dielectric (t₂) was limited to 2 μm or less to avoid the formation of cracks or delamination. The simulated electric-field profiles for the three-field plated structure at reverse bias of 1.5 kV are shown in Fig. 2(a). We varied the FP2 dielectric thickness t₂, while keeping the FP3 dielectric thickness (t₃) fixed. The peak electric fields at the field-plate corners (locations 1 and 2) are shown as a function of t₂ in Fig. 2(b). Based on the simulated field profiles, the optimal value of t₂ was found to be between 0.7 and 0.8 μm, assuming a similar breakdown field for both Al₂O₃ and SiON. On increasing the value of t₂ beyond the optimal value, the peak electric field at the corner of FP1 becomes higher, and for lower values of t₂, peak field at the corner of FP2 becomes higher. The effect of field-plate length (L_FP,2 and L_FP,3) was also simulated to obtain optimal values. Field-plate lengths above 10 μm were found to have no significant effect on the peak electric fields (see Supplementary data, Fig. S2).

Zoom In Zoom Out Reset image size

The final device structure of the fabricated SBDs and HJDs, and the detailed process flow, are shown in Figs. 1(a)–1(c). A ​10-μm-thick HVPE-grown (001) β-Ga₂O₃ epilayer procured from NCT was used for device fabrication. The samples were first cleaned using solvents followed by a 15 min dip in concentrated HF to remove surface contaminants. The samples were subsequently patterned using a combination of SiON dielectric (∼1 μm) and AZ 4330 photoresist (∼3 μm thick) for nitrogen ion implantation. To ensure that the active region was not impacted by the implant, the implant layer was spatially separated by 7.5 μm from the edge of the Schottky contact (or HJD; see Fig. 1). The nitrogen implant layer was simulated using SRIM to form a 0.6-μm-deep box profile with a concentration of 5 × 10¹⁸ cm⁻³ (see Supplementary data, Fig. S3). Post implantation and removal of the hard mask, a Ti/Au/Pt (50/50/50 nm) metal stack was deposited using e-beam evaporation followed by rapid thermal annealing at 470 °C for 1 min to form the backside contact. Subsequently, to form the field-plate dielectric, about 2-μm-thick SiON was deposited using plasma enhanced chemical vapor deposition at 350 °C. The SiON dielectric layer was etched twice using dilute buffered oxide etch (BOE) to form a staircase structure for FP2 and FP3. Complete removal of SiON was ensured in the active region and the FP1 region by tracking the etch depth using a Dektak profilometer. The final thicknesses of the SiON dielectric in FP2 (t₂) and FP3 (t₃) were about 1 μm and 2.1 μm, respectively. After the SiON field plates (FP2 and FP3) were etched, a 50 nm Al₂O₃ dielectric for FP1 was deposited using PEALD. The sample was cleaned using a piranha solution prior to PEALD. Further details regarding the PEALD of Al₂O₃ is provided in Ref. 19. The Al₂O₃ layer is removed from the active region of the diodes using dilute BOE. The Al₂O₃ layer was intentionally over-etched to ensure complete removal of Al₂O₃ from the active region. Half of the devices were further patterned to deposit p-NiO in the active region to form p-NiO/Ga₂O₃ HJDs. To ensure complete coverage of NiO in the active region, the NiO layer is extended over the first field-plate dielectric as shown in Fig. 1. The NiO layer was deposited by RF magnetron sputtering at room temperature in an 8' Kurt J. Lesker sputter tool. The NiO layer was sputtered at an RF power of 300 W and chamber pressure of 4 mT in pure Ar atmosphere. Post deposition, the NiO was annealed in a rapid thermal annealer for 5 min at 300 °C under N₂ gas. The estimated carrier concentration and mobility of the p-NiO is close to 10¹⁸ cm⁻³ and <0.01 cm² V⁻¹ s⁻¹, respectively. Afterwards, a Ni/Au (50/100 nm metal stack was evaporated and annealed for 5 min under the same conditions to form ohmic contacts to the NiO layer. Finally, processing for the SBD devices was completed by depositing a Pt/Au (50/100 nm) Schottky metal stack using e-beam evaporation. The devices were characterized using a Keysight B1500A parameter analyzer and a Keysight B1505A high-voltage analyzer. Samples for HR-TEM imaging were prepared using focused ion beam milling with a Thermo Fisher Helios 5 UX, and the samples were imaged using an image-corrected FEI Titan 80–300 operated at 300 kV.

Forward current–voltage (IV) characteristics for SBDs and HJDs with different field-plate lengths are shown in linear and semi-logarithmic scales in Figs. 3(a) and 3(b), respectively. The differential on-resistance (R_on,sp) for both SBDs and HJDs are also shown in Fig. 3(a) with solid and dotted red curves, respectively. We notice a slight increase in the on-state resistance of HJDs when compared to SBDs due to the additional resistance of the 50-nm-thick p-NiO layer. The SBDs show a R_on,sp of approximately 6.2–6.5 mΩ cm². HJDs, on the other hand, show a slightly higher R_on,sp of 6.8–7.1 mΩ cm² for 100-μm-diameter devices. The resistance contribution of Sn-doped n⁺⁺ Ga₂O₃ substrate to R_on,sp, including the spreading resistance, is estimated to be about 0.11 mΩ cm² assuming μ = 50 cm² V⁻¹ s⁻¹ and n = 5 × 10¹⁸ cm⁻³. Figure 3(c) shows the charge concentration (N_D–N_A) extracted from the capacitance–voltage (CV) measurements using 250-μm-wide SBD test structures. The built-in potentials (V_bi) extracted for both the SBDs and HJDs are about 1.08V and 1.03V, respectively. The Schottky barrier height for the Pt/Ga₂O₃ contact from extrapolation of the 1/C² plot is found to be ≈1.23 eV (see Supplementary data, Figs. S4 and S5).

Zoom In Zoom Out Reset image size

When comparing the forward characteristics of SBDs and HJDs, we observe similar turn-on characteristics for both the fabricated HJDs and SBDs. The turn-on voltages calculated for SBDs and HJDs (assuming an on-state current density of 1 A cm⁻²) are 0.85 V and 0.8 V, respectively. The obtained turn-on voltage for the p-NiO/Ga₂O₃ HJDs is significantly lower than those reported in the literature. ^9,20,21) In addition, the calculated ideality factors (n) for SBDs and HJDs (over six magnitudes of current) are 1.09 and 1.11, respectively. The ideality factor for SBDs is close to unity, suggesting that the transport is dominated by thermionic emission. ^14,22) However, the ideality factor calculated for HJDs is also quite similar to that for SBDs and deviates from what is reported in the literature for p-NiO/Ga₂O₃ heterojunctions. ^20,23) One plausible explanation for these deviations observed in the ideality factor and turn-on voltage could be due to different band alignment for the p-NiO/Ga₂O₃ heterojunction in this work. ²⁾ It has been reported that the band alignment in the p-NiO/Ga₂O₃ heterojunction varies due to sputtering conditions and is determined by several factors, including defects, interfacial contamination and crystal orientation. ²⁾ Further experiments are necessary to fully elucidate the processing/deposition conditions responsible for obtaining near unity ideality factors and sub-1 V turn-on voltage in HJDs.

The breakdown performance of the diodes was characterized using a current compliance setting of 1 A cm⁻². Semi-logarithmic reverse IV characteristics of the fabricated HJDs and SBDs are shown in Fig. 4(a). Both the HJDs and SBDs showed similar breakdown voltages of ∼1400 V. The observed breakdown is lower than that expected from the simulations, due to a slightly higher value of t₂ (∼1 μm instead of 0.7–0.8 μm). The parallel-plane electric field for the measured breakdown voltage 1400 V and doping 1 × 10¹⁶ cm⁻³ was calculated to be ∼2.3 MV cm⁻¹. We also observe lower reverse leakage current in p-NiO/Ga₂O₃ HJDs when compared to the Pt/Ga₂O₃ SBDs. The reverse leakage current in HJDs is lower than 10⁻⁴ A cm⁻² up to a reverse bias of 1.25 kV. Figure 4(b) represents the distribution of breakdown voltage for different field-plate lengths for SBDs and HJDs. Breakdown voltage is found to not vary significantly with the field length (L_FP,2, L_FP,3 = 10 μm, 20 μm, 30 μm) for NiO/Ga₂O₃ HJDs, while the heterojunction p–n diodes also show a slightly narrower variation range and have higher average breakdown voltage compared to SBDs. Figure 4(c) shows the optical micrograph of the Schottky diode before [Fig. 4(c), (1)] and after [Fig. 4(c), (2)] breakdown. The breakdown is found to occur at the corner of FP1, which is also consistent with the TCAD simulations [see Fig. 2(b)]. Figure 4(d) shows HR-TEM images of SBD and HJD samples. These clearly show the interfaces between Pt/Ga₂O₃ on SBD and Ni/p-NiO/Ga₂O₃ on HJD, respectively. Additionally, roughly linear striations parallel to the growth direction were observed in the p-NiO, which could be due to the presence of strain causing slight local variations in crystal orientation. Furthermore, we also fabricated Al₂O₃ MOSCAP devices with the same field termination, which also showed similar breakdown voltage to SBDs and HJDs (∼1.5 kV) (see Supplementary data, Fig. S6). This suggests that the breakdown voltage of the diodes is limited by the FP1 dielectric and not by the intrinsic breakdown of the Pt/Ga₂O₃ junction or p-NiO/Ga₂O₃ junction.

Zoom In Zoom Out Reset image size

The power figure-of-merit (PFOM) for the diodes was calculated using the equation PFOM = $\tfrac{{V}_{\mathrm{br}}^{2}}{{R}_{\mathrm{on},\mathrm{sp}}}.$ The best-performing SBDs and HJDs showed a PFOM of 316 MW cm⁻² and 288 MW cm⁻², respectively. The PFOM of HJDs is slightly lower due to the higher differential on-state resistance. Figure 5(a) shows the benchmark comparison of R_on,sp–V_br obtained in this work with prior reports of vertical Ga₂O₃ diodes. ^{9,13,16,17,20,21,23–30)} To also include the effect of turn-on voltage, we estimated the effective on-state resistance (R_on ^eff = V/I_on) at a current density of 100 A cm⁻². Figure 5(b) shows the benchmark comparison of R_on ^eff–V_br obtained in this work with prior reports of p-NiO/Ga₂O₃ HJDs. ^{9,15,16,20,21,29,31,32)} Due to the lower turn-on voltage, we obtain one of the lowest effective on-resistances reported for a p-NiO/Ga₂O₃ HJD using a standard 10-μm-thick epilayer from NCT.

Zoom In Zoom Out Reset image size

In conclusion, we report vertical Ga₂O₃ Schottky barrier diodes and p-NiO-based heterojunction diodes integrated with a combination of three field plates and nitrogen ion implant for field termination. The best-performing diodes exhibit a low R_on,sp of 6.2 mΩ cm² and a breakdown voltage of about 1400 V, with reverse leakage current lower than 10⁻⁴ A cm⁻² up to a voltage of 1.25 kV. The heterojunction diodes also demonstrate a low turn-on voltage of 0.8 V and a near-unity ideality factor, resulting in a low R_on (effective) of ∼18 mΩ cm², which is among the lowest reported for p-NiO/Ga₂O₃ heterojunction diodes.

Acknowledgments