Introduction

The monoclinic, thermally stable semiconductor β-Ga2O3 with a bandgap of 4.8 eV is attracting interest recently for applications in high-power rectifiers, Field Effect Transistors (FETs) and photodetectors for the far-UV spectral range1,2,3,4. The material can be grown from melt in the form of large diameter high crystalline quality bulk crystals that can be further used as substrates for epitaxial deposition of Ga2O3 films that are then processed into power rectifiers, FETs or solar-blind photodetectors. The bulk β-Ga2O3 substrates prepared by Edge-defined Film-fed (EFG) or Czochralski (CZ) techniques, undoped, semi-insulating due to Fe doping or n-type doped with Sn in a wide range of electron concentration are commercially available currently with dislocation density below 106 cm−2 and diameters up to 4 in.1,5. This sets β-Ga2O3 apart from most other wide-bandgap semiconductors, such as SiC, GaN, AlN or diamond for which high quality bulk crystals have to be grown by difficult low-yield methods that render them expensive and either severely limit the crystalline quality, as for GaN or AlN, or make the production of large diameter substrates difficult, as in the case of diamond. Good crystalline quality epitaxial films of Ga2O3, semi-insulating or n-type doped can now be deposited by epitaxial growth methods, such as Halide Vapor Phase Epitaxy (HVPE), Metalorganic Chemical Vapor Deposition (MOCVD) or Molecular beam epitaxy (MBE). Heterojunctions with films of (AlxGa1-x)2O3 with wider bandgap for use in Modulation Doped FET structures (MODFETS) can be grown by MOCVD or MBE2. The wide bandgap of β-Ga2O3 allows to fabricate power rectifiers capable of handling electric fields strengths of about 8 MV/cm, with a unique combination of high breakdown voltage and low ON-state resistance1,2.

However, the lack of defects or impurities suitable for obtaining p-type conductivity necessary for fabrication of bipolar devices is a serious limitation for Ga2O36. To circumvent this difficulty, the use of various heterojunctions with wide-bandgap materials that show natural p-type conductivity have been tried1,2. Perhaps the most successful approach is growth of n-type β-Ga2O3 heterojunctions (HJs) with p-type NiO films deposited on β-Ga2O3 by a variety of techniques, such as magnetron sputtering, Ion Beam Sputtering (IBS) or Pulsed Laser deposition (PLD) (see e.g.1,7,8. NiO is a cubic semiconductor with lattice parameter not much different from β-Ga2O39. When deposited at moderate temperatures on Ga2O3 the films have been shown to be nanocrystalline, with a sharp interface with epitaxial layers of β-Ga2O310. The bandgap of NiO varies between 3.5 and 4 eV, depending on the deposition method9. The p-type conductivity in the films is due to Ni vacancy acceptors forming small polarons and giving rise to p-type conductivity via hopping9,11,12. The density of these acceptors can be greatly enhanced by growth under oxygen-rich conditions and can promote effective hole concentrations in the impurity band of 1017–1019 cm−3 depending on growth conditions, albeit with a very low hole mobility9,11,12. The concentration can be further enhanced by additional Li doping9,11. The NiO conductivity is strongly temperature dependent, with an activation energy of 0.1–0.2 eV9,11. The heterojunctions of n-type β-Ga2O3 and p-type NiO are of staggered type II kind, with conduction band offsets ΔEc and valence band offsets ΔEv varying from 1.2 to 1.9 eV in the case of ΔEc and from 2.1 to 2.7 eV in the case of ΔEv, the variation being due to the difference in the orientation of Ga2O313.

These heterojunctions have been shown to be extremely useful in Ga2O3-based power device technology, owing to the decrease of reverse currents and increase of breakdown voltages of HJs caused by the high band offset in the conduction band13. Having the p-NiO layer extend beyond the metal contact helps to mitigate the effects of electric field crowding at the edges of the diodes10. Several papers have reported the decrease of the series resistance of the NiO/Ga2O3 HJs compared to straight Schottky diodes, which was explained by injection of holes from the p-NiO layer contributing to the conductivity modulation in the n-Ga2O3 portion of the diode14,15. The bipolar nature of the p-NiO/n-Ga2O3 diodes was used to explain the decreasing the switching times and increasing the stability of such HJs to power surges compared to Schottky diodes16. The ongoing studies have resulted in impressive increases of breakdown voltages of NiO/Ga2O3 HJ rectifiers to 8.2 kV17 or recently, to 13.5 kV18 with low specific series resistances in the ON state.

Currently, Ga2O3-based devices compete in power electronics against more established wide-bandgap devices based on III-Nitrides and SiC1,2. Because of the high bonding energy of wide-bandgap materials power devices based on Ga2O3, III-Nitrides, SiC provide serious advantages over devices based on Si or GaAs when operating in harsh radiation environment characteristic of space and avionic communications, high energy particles accelerators or nuclear reactors monitoring, etc. For space systems the main damaging factors come from protons of various energies, electrons, cosmic ray particles encountered in Van Allen radiation belts around Earth and open space galactic cosmic rays19. Assessment of reliability of different devices types is a complicated technical problem in which the type of device, the type of radiation environment and the characteristics of the system in which the devices are to operate should be taken into account. However, in many cases the radiation tolerance of different materials can be compared based on the carrier removal rates observed for several types of particles with several typical energies and for materials with comparable starting doping concentrations. The general approach is described in some detail in19. For simplified analysis it is rather popular to use comparisons of carrier removal rates observed as a result of proton irradiation with protons of 1–2 MeV energy constituting an important part of radiation environment in the near-Earth orbit. Such experiments have been performed for III-Nitrides, SiC, Ga2O3 and the results are summarized in19. Roughly, for the n-type doping on the order of some 1016 cm−3 such carrier removal rates are on the order of 100–200 cm−1 for all three types of wide-bandgap materials. Given a much high projected electric breakdown field in Ga2O31,2 the Ga2O3-based devices should potentially provide serious advantages over their more established competitors. As the NiO/Ga2O3 HJs now show actual breakdown fields approaching theoretical estimates (see above) such devices are expected to be very promising.

However, rather unexpectedly, the radiation hardness of NiO/Ga2O3 HJs to proton or neutron irradiations was found to be lower than for reference Schottky diodes in terms of stronger changes in the breakdown voltage20,21 or higher carrier removal rate in the n-Ga2O3 portion of the HJ, particularly when irradiation is carried out under high reverse voltage22. The authors of Ref.22 attributed that to the enhanced Ga vacancy acceptors being drawn deeper inside the Ga2O3 film and to the more effective formation of deep electron traps E2* with levels near EC-0.7 eV22,23,24. In a recent paper, we have drawn attention to the observed strong frequency dispersion of the capacitance of NiO/Ga2O3 HJs, the prominent temperature dependence of capacitance and AC conductance showing the presence of two activation energies of 0.1 and 0.2 eV that we attributed to the effects in the NiO layer and at the interface of NiO and Ga2O325. We also noted the differences between the deep electron trap spectra in the near-surface regions of the Schottky diodes and HJs.

In this paper, we present detailed results of the studies of these effects in Schottky diodes and HJ irradiated with 2 × 1013 p/cm2 of 1.1 MeV protons. The protons energy used has been chosen because, for Ga2O3 Schottky diodes, the data on carrier removal rates and deep traps spectra has been documented in multiple papers and can be compared with the results obtained for NiO/Ga2O3 HJs24,26,27,28. It is seen the proton irradiation leads to stronger decrease of the net donor concentration in the HJs compared to Schottky diodes, which cannot be explained by the enhanced formation of the E2* electron traps. The observed changes are, at least in part, linked to the changes of electrical properties of the NiO films upon proton irradiation and the presence of layer with a higher density of defects at the interface between NiO and Ga2O3. This is important in the context of the relative lifetimes HJDs versus SDs in space-borne applications where high proton fluences may be encountered.

Results and discussion

Schottky diodes

The schematic design of the Schottky diode structure is presented in Fig. S1a of the Supplementary Material. Before irradiation the Schottky diodes displayed capacitance versus frequency (C-f) characteristics with capacitance plateau extending up to at least 1 MHz at all temperatures (see Fig. S1c for room temperature), fairly linear 1/C2 versus voltage (V) characteristics with net donor concentration of Nd = 2.7 × 1016 cm−3 and built-in voltage Vbi = 1 V, both only slightly changing with temperature and with probing frequency in C-V measurements (respective Nd values were Nd = 2.73 × 1017 cm−3 and Vbi = 0.93 V at 420 K and Nd = 2.65 × 1016 cm−3 and Vbi = 1.16 V at 90 K). No prominent steps in capacitance or peaks in AC conductance were detected in Admittance Spectra (AS) measurements29,30. In current density–voltage (J-V) characteristics, we observed low reverse current density with slight voltage dependence of reverse current density. At − 5 V the current density, J(− 5 V), was at 300 K equal to J(− 5 V) = 5.6 × 10–9 A/cm2, with ideality factor of η = 1, saturation current density Jo = 7.5 × 10–14 A/cm2, and series resistance Rs = 6.2 Ω (specific series resistivity Rsp = 43 mΩ⋅cm2). At 400 K, respective values changed to J(− 5 V) = 5.6 × 10–9 A/cm2, η = 1, Jo = 2.7 × 10–9 A/cm2, Rs = 6.8 Ω (Rsp = 47.6 mΩ⋅cm2) (see Fig. 1a).

Fig. 1
figure 1

(a) J-V characteristics measured for the Schottky diode before (solid lines) and after (dashed lines) irradiation for 300 K (black lines) and 411 K (red) lines; (b) room temperature J-V characteristics measured for the Schottky diode and the NiO/Ga2O3 HJ diode; (c) J-V characteristics measured for the HJ diode at room temperature (black line), 412 K (red line), 113 K (blue line) after irradiation; (d) comparison of room temperature J-V characteristics of irradiated Schottky diode and irradiated HJ diode.

Full size image

In the deep trap spectra measured by DLTS, two major deep electron traps with levels near EC-0.73 eV (so called E2* defect often associated with divacancies VGa-VO23,31), and EC-0.76 eV (so called E2 trap belonging to Fe acceptors in the Ga site32) were present at low concentration (Fig. 2a displays the spectra measured with quiescent biases of − 1 V and − 20 V and pulse voltages of 1 V (50 ms long); Fig. 2b presents the spectra measured with bias − 5 V and pulse of − 1 V excluding the near-surface region of the diode (these regime will be useful for comparison with the NiO/Ga2O3 HJs, as discussed further). Figure 2c shows the spectra re-built in coordinates of 2Nd × (ΔC/C) × F−1 versus temperature for these three sets of voltages/pulses (here ΔC is the DLTS signal, i.e. the difference of capacitances measured at time windows t1 and t2, C is the steady state capacitance, F−1 is the DLTS spectrometer function29. In such coordinates, the magnitude of the peaks represents the concentrations of respective traps when the so-called λ-correction is not made29. This λ-correction is lower for higher applied voltages, so that the spectrum for the applied voltage of − 20 V gives a reasonable idea of the real trap concentration. The convention in Fig. 2, as in all other DLTS figures in the paper, is that the positive sign of the peak corresponds to majority electron traps for which the capacitance after the pulse increases with time29. The concentration of the minority traps due to deep centers in the lower half of the bandgap can be estimated from C-V profiling with monochromatic illumination that allows obtaining the densities of such traps, their optical ionization energy and the existence for them of a barrier for capture of electrons33,34,35. This is the LCV technique33,34,35 that can also be complemented by Deep Level Optical Spectroscopy (DLOS)35 and Deep Level Transient Spectroscopy with optical injection (ODLTS)33,36. Figure 3a presents the LCV spectrum of the Schottky diode before irradiation. The presence of deep acceptors in the lower half of the bandgap with optical ionization thresholds near 2.3 eV and 3.1 eV can be clearly seen. The concentrations of these traps obtained from the plateau of the LCV concentration for the 2.3 eV and the difference between the LCV concentrations for the plateaus corresponding to the 3.1 eV trap and the 2.3 eV trap33,34,35 were low, close to 1.5 × 1014 cm−3 for the 2.3 eV trap and 2 × 1014 cm−3 for the 3.1 eV trap. The trap with optical ionization threshold near 2.3 eV has been associated with the split gallium vacancy acceptors, possessing a high barrier for capture of electrons33,34,35, while the trap with optical ionization energy near 3.1 eV has been attributed to centers associated with gallium vacancy acceptors33,34.

Fig. 2
figure 2

(a) DLTS spectra measured at 100 kHz for the Schottky diode with bias − 20 V and pulse 1 V, with bias − 1 V and pulse 1 V; (b) DLTS spectra measured with bias − 5 V and pulse − 1 V before and after irradiation, the spectra shown for two time windows settings of 0.15 s/1.5 s and 1.5 s/15 s; (c) the spectra rebuilt so as to show traps concentrations for three bias/pulse settings; (d) changes in trap types and concentrations after irradiation.

Full size image
Fig. 3
figure 3

(a) LCV spectrum of the Schottky diode after irradiation; (b) ODLTS spectrum measured for the Schottky diode after irradiation.

Full size image

Figure 3b shows ODLTS spectrum obtained for the Schottky diode before irradiation at reverse bias of − 1 V and 5 s-long optical pulse injection with the Light Emitting Diode (LED) with the peak wavelength of 277 nm and optical power density 15 mW/cm2. The negative sign peaks near 100 K and 150 K corresponding to the processes with activation energy 0.14 eV and 0.2 eV are similar to the features in ODLTS reported by us previously33 and attributed to the Self-trapped Holes (STH) states linked to different O sites33,37. The peak with activation energy near 1.1 eV is probably related to the same center giving rise to the 3.1 eV transition in LCV spectra and associated with some form of Ga vavancies33.

After irradiation with a fluence of 2 × 1013 p/cm2 of 1.1 MeV protons, the capacitance became considerably lower, as shown in Fig. S1c of the Supplementary Material for room temperature. The net donor concentration at room temperature decreased according to C-V profiling from 2.7 × 1016 cm−3 to 1.9 × 1016 cm−3 (see Fig. S2a,b of the Supplementary Material). The reverse current density of the Schottky diode at 300 K was virtually the same as before irradiation, as was the saturation current, ideality factor and the series resistance. The changes in J-V characteristics measured at 420 K were also minor.

In DLTS spectra measured after irradiation we observed a slight increase of the E2* trap signal, the emergence of three new minor traps with levels near EC-0.21 eV (so called E8 traps26,31,33,34), EC-0.6 eV (so called E1 traps31,38,39), and EC-1.1 eV (so called E3 traps26) (Fig. 2b shows the changes in DLTS signal ΔC/C and Fig. 2c displays the changes in the traps concentrations (not taking into account the λ-correction) for the steady-state bias of -5 V and the bias pulse of − 1 V, i.e. when probing the concentration changes in the bulk portion of the space charge region). All of these levels have been reported to be introduced by irradiation with protons, α-particles, fast reactor neutrons26. The E1 trap has been recently associated with shallow donor complexes with hydrogen39, while the E3 traps have been attributed to deep oxygen donors on the grounds of them being introduced by all types of irradiations26 and their level being close to that of one of the oxygen vacancy deep donors predicted by theory24. It has also been suggested that the centers could be due to Ti donors31,40, however, this Ti donors assignment is not likely to be true in the case of centers introduced by irradiation. It can be seen that the density of the E2* trap has increased, although the changes are not large, about 40%. Unexpectedly, we also observe an increase in the concentration of the E2 trap due to Fe acceptors-the reasons have yet to be understood.

The changes in the deep acceptor states revealed by LCV spectra and ODLTS spectra measurements as a result of proton irradiation were quite low and the density increase did not exceed ~ 20%. Thus, the aggregate changes of the densities of all deep states occurring throughout the bandgap of our n-Ga2O3 sample as a result of proton irradiation was fairly low and did not exceed something like some 1014 cm−3, i.e. was much lower than the observed decrease of the net shallow donors density by 8 × 1015 cm−3 after irradiation. Hence, the said decrease cannot be attributed to simple compensation of shallow donors by deep defects introduced by irradiation even if all of them are acceptors and the observed carrier removal has to be due to other factors. We have previously reported similar observations for n-Ga2O3 films irradiated with high energy protons of different energies, α-particles, fast reactor neutrons26,34 and have proposed that the decrease of the net donor concentration could be caused by the formation of complexes between the shallow donors and triply charged Ga vacancy acceptors26,34. We assume that the carrier removal rate of dN/dF = 400 cm−1 observed in the present study could have the same origin (N here is the change in measured net shallow donors concentration, F is the fluence of 1.1 MeV protons). Our data do not support the hypothesis of Ref. 21 where the carrier removal rate in proton irradiated NiO/Ga2O3 HJ has been attributed to the introduction of the E2* states believed to be deep acceptors.

NiO/Ga2O3 HJs

The schematic design of the NiO/Ga2O3 HJ structure is given in Fig. 4a. The designs for the Ga2O3 Schottky diode and the NiO/Ga2O3 HJ diodes can be compared side-by-side in Fig. S1a,b of the Supplementary Material. In both structures the samples were cut from the same commercial epitaxial Ga2O3 wafer, the characteristics of 6 diodes on the 1 × 1 cm2 piece were measured and compared and the starting C-f, C-V, J-V, AS, DLTS characteristics were shown to be very close to each other. Figure 4b)shows the C-f dependences of one of the studied diodes, d6, at three temperatures 300 K (solid black line), 415 K (solid red line), and 115 K (solid blue line) before irradiation. The most obvious difference with the Schottky diode sample is the marked frequency dispersion of capacitance, with a prominent low frequency step in capacitance. Admittance spectra measurements, i.e. measurements of the temperature dependence of capacitance C and AC conductance G/ω at different temperatures, revealed the presence of two sets of steps in capacitance and corresponding peaks in AC conductance, as shown in Fig. 5a,b. The AC conductance G in Fig. 5b is, as usual29, normalized by the angular frequency ω = 2πf, G/ω, the spectra are shown for − 0.2 V bias. All samples showed similar behavior with the steps in capacitance/peaks in G/ω shifting to higher temperatures with increasing measurement frequency. Two sets of steps/peaks were detected. The activation energy Ea and capture cross σ for the processes corresponding to the step in capacitance/peak in conductance were determined from the shift in temperature corresponding to the step/peak with probing frequency29. For the major step/peak the activation energy Ea1 and capture cross section σ1 were Ea1 = 0.17 eV, σ1 = 1.2 × 10–20 cm2. For the minor low temperature step/peak the activation energy and capture cross section were respectively Ea2 = 0.12 eV, σ2 = 3.1 × 10–16 cm2 (Fig. 5a,b). All studied diodes demonstrated similar behavior, with a small spread in the activation energies between 0.16 and 0.18 eV for the main peak and 0.1–0.12 eV for the minor peak. This has already been reported in our earlier paper devoted to studies of unirradiated NiO/Ga2O3 HJs25.

Fig. 4
figure 4

(a) Structure of NiO/Ga2O3 heterojunction (HJ) sample, (b) C-f characteristics of the NiO/Ga2O3 HJ sample measured before irradiation (solid lines) and after irradiation (dashed lines) at 300 K (black), 400 K (red), 115 K(blue) lines. Concentration profiles calculated from C-V measurements on NiO/Ga2O3 HJ diode at different temperatures and different frequencies for (c) pristine and (d) irradiated HJ diode.

Full size image
Fig. 5
figure 5

(a) the capacitance at − 0.2 V dependence on temperature for capacitances measured at frequencies from 20 Hz (leftmost curve) to 1 MHz; (b) respective G/ω dependences in the same frequency range; two peaks could be clearly seen for frequencies in the range 20 Hz to 3 kHz (marked by blue lines) and 20 kHz to 1 MHz (marked by olive lines); (c) the temperature dependence of current of the Ni/NiO/FTO diode at 0.3 V before and after irradiation; (d) room temperature 1/C2 versus V plots of the Ni/NiO/FTO diode measured at 10 kHz before and after irradiation.

Full size image

The strong frequency dependence of capacitance results in the 1/C2-V plots varying significantly with frequency, as illustrated in Fig. S2c of the Supplementary Material. The plots at all temperatures are not linear, indicating considerable spatial non-uniformity of the centers concentration probed at different frequencies. Thus, C-V profiling at different temperatures and frequencies is more informative. Such profiling done at different temperatures and frequencies yields the dependence of concentration of traps on the distance from the HJ interface25,30 as shown for the sample before irradiation in Fig. 4c for three different temperatures and three different frequencies.

These results clearly indicate that the major step/peak in AS is due to the 0.17 eV states with concentration is very high, ~ 1.4 × 1017 cm−3 at about 0.3 μm from the interface with the NiO layer with Ga2O3 (see the 411 K plot obtained at 100 Hz in Fig. 4c). Closer to the interface the concentration of the 0.17 eV centers decreases, indicating their higher compensation. For frequency of 100 kHz where the step in capacitance is virtually over at all temperatures the concentration is not different from the one given by the 300 K profile obtained at 10 kHz where the C-f dependence at 300 K shows a plateau. Deeper inside the sample, the concentrations measured at all frequencies become close to each other and to the net donor density obtained in C-V profiling of the Schottky diode. Hence, the region 0.2–0.3 μm from the HJ interface contains a high concentration of 1.4 × 1017 cm−3 of centers with energy level ~ 0.17 eV. This concentration is an order of magnitude higher than the net density of uncompensated shallow donors in the bulk of the n-Ga2O3 film as determined from C-V profiling in the Schottky diode. Therefore, it seems reasonable to assume that the states are due to some yet unidentified interfacial states distributed in the region 0.2–0.4 μm from the interface. With further approach to the interface with NiO, the concentration of traps capable of responding to the probing frequency in C-V profiling decreases, suggesting an increased compensation in that region adjacent to the interface. This region contributes to the band bending near the interface and results in high voltage offsets in 1/C2-V of about 2 V, much higher than predicted by X-ray Photoelectron Spectroscopy (XPS) for the (001) orientation with conduction band offset ΔEc = (1.5–1.6) eV13.

As for the low temperature set of steps/peaks in AS spectra in Fig. 5a,b, their origin needs better understanding. If one measures the temperature dependence of the forward current of the Ni/p-NiO/n+-FTO heterojunction diodes (see the “Methods” section below) one observes the temperature dependence of the current at forward bias (proportional to the layer resistivity of p-NiO layers)11 to have activation energy ~ 0.12 eV (Fig. 5c). The hole concentration at 300 K determined from the 1/C2-V plot of the Ni/p-NiO/n+-FTO structure11 is 9.9 × 1018 cm−3 (Fig. 5d), while at 140 K it decreases to 3.3 × 1017 cm−3. The results are like those reported in Ref.11. These observations strongly hint that the low temperature step in AS spectra can be related to the freeze out of the conductivity of the p-NiO film, as the activation energy detected in AS is close to the activation energy of p-NiO conductivity freeze-out. The low temperature 115 K profile in the concentration versus depth plot in Fig. 4c shows an apparent shift by 0.2 μm deeper inside the sample compared to the room temperature profile. Such shift in apparent position of the heterointerface is often observed in p-n heterojunctions in which the p-type layer freezes out, such as e.g. in GaN/InGaN multiquantum well (MQW) Light Emitting Diode (LED) structures when the Mg acceptors in the top p-type GaN(Mg) layer freeze out41.

In J-V characteristics of the NiO/Ga2O3, the reverse current density at -10 V was 2.3 × 10–8 A/cm2, the saturation current density calculated from the intercept of the exponential region of the forward characteristics with the y-axis was 8.3 × 10–20 A/cm2 and the ideality factor was η = 1.6 The series resistances was 10.9 Ω and showed an approximately exponential decrease with increasing temperature with activation energy ~ 0.1 eV similar to the activation energy of the conductivity freeze-out in NiO25.

The high ΔEc of the NiO/Ga2O3 HJ and the lowering of the edge electric field strength due to the p-NiO extending beyond the metallic contact10 (Fig. 4a) contribute to the lower reverse current density of the HJ compared to the Schottky diode (Fig. 1b), albeit at the expense of a slightly higher series resistance because of the additional contribution of the NiO film resistance.

DLTS measurements performed for the NiO/Ga2O3 HJ structures also suggest that the deep trap spectra of the region adjacent to the HJ boundary are changed compared to the spectra in the film with Schottky diodes. From Fig. 6a showing the DLTS spectrum obtained with the quiescent voltage of − 10 V with pulsing to − 2 V, i.e. some distance removed from the HJ boundary, in contrast to DLTS spectra of the Schottky diode, the spectrum of the HJ demonstrates a conspicuous absence of the E2* peak. The spectrum in Fig. 6a corresponds to probing of the region of ~ 0.7 μm from the HJ boundary. In the Schottky diode, the E2* trap is prominent for all probed depths, as can be seen from the spectra measured with different quiescent biases in Figs. 2c and 6a. Only the E2 trap is seen in the HJ sample spectra. As the E2* center is believed to be a complex of native defects, presumably the VGa-VO divacancy23 one concludes that a transformation of native defect spectra should occur in the vicinity of the NiO/Ga2O3 HJ. Some researchers14,15 report observation in DLTS spectra of NiO/Ga2O3 HJs hole-trap-like signals at low temperatures that are attributed to the STH signals of the type observed in ODLTS spectra of n-Ga2O333 and seen in our Fig. 3b. The explanation offered in Ref.14,15 is that the feature is due to STH states coming from holes injected from p-NiO side of the HJ into the n-Ga2O3 side. We observe such hole-traps-like signals in the spectra measured with high positive bias pulsing. An example is given in Fig. 6b for the spectra measured with reverse bias of − 1 V and forward bias pulsing to 2 V for short time windows. However, (a) these spectra are invariably accompanied by the electron-trap-like features with ionization energy of 0.16 eV, similar to the one observed in AS spectra; (b) the hole-trap-like features are also seen with no forward bias pulsing, as in Fig. 6b where they are observed with bias − 10 V and pulsing to − 2 V. This could be the artifact of the strong capacitance decrease at low temperature that can give rise to the inversion of the sign of the DLTS peak of the traps observed under such conditions25.

Fig. 6
figure 6

(a) comparison of DLTS spectra of the Schottky diode and of the NiO/Ga2O3 diode; (b) comparison of the two spectra obtained for short time windows for the HJ diode with bias − 1 V and pulse 2 V and bias − 10 V and pulse − 2 V.

Full size image

Unfortunately, the presence of strongly absorbing for photons with energy higher than 3.5 eV NiO layer and the very low lifetime of charge carriers in NiO25 did not allow accurate measurements of the LCV spectra and ODLTS spectra in the HJ samples to directly compare the spectra and density of hole traps with the Schottky diodes.

Irradiation of NiO/Ga2O3 HJs with 2 × 1013 cm−2 fluence of 1.1 MeV protons led to changes of C-f characteristics, J-V characteristics and concentration profiles illustrated by the data obtained for one of the diodes, d6. The capacitance in C-f dependences measured for different temperatures (shown in dashed lines) considerably decreased, as shown in Fig. 4b. The admittance spectra showed the same two major centers with activation energies 0.17 eV and 0.12 eV. C-V profiling presented in Fig. 4d shows the concentration of the 0.17 eV centers as greatly reduced in the upper 0.3–1 μm part of the space charge region and the rate of decrease with proton fluence of the concentration of these centers as high, about 6.5 × 103 cm−1. Deeper inside the sample the shallow donors are also compensated in the 0.7–3 μm region from the interface, as seen in the 100 kHz concentration profiles in Fig. 4d. The rate of decrease in the concentration of these shallow centers is also high, 850 cm−1, i.e. much higher than in the Schottky diode, where it is 400 cm−1.

These changes in centers concentrations are reflected in the changes of the J-V characteristics of the irradiated HJ diodes. The main effect was in the increase of the series resistance of the diode to 43 Ω at room temperature, 16.7 Ω at 400 K, and 2 × 104 Ω at 129 K compared to the values before irradiation (Fig. 1c).

The advantages of the NiO/Ga2O3 HJ diodes in terms of reverse current remain after proton irradiation, which, however, increases the series resistance of heterojunctions in the forward direction. The increased series resistance still shows activation energy close to 0.1 eV, coming from the impact of the temperature dependence of the resistivity of NiO layer showing a similar temperature dependence of resistivity in the forward direction (Fig. 5c). This increase in resistivity comes from the decrease of the hole concentration in C-V measurements after irradiation to 7.5 × 1018 cm−3 at room temperature, compared to 9.9 × 1018 cm−3 before irradiation (Fig. 5d). The hole concentration in satellite Ni/p-NiO/n+-FTO measured by C-V profiling at 140 K also decreased from 3.3 × 1017 cm−3 to 6.7 × 1016 cm−3.

The results of DLTS measurements after proton irradiation of the HJ diodes are compared to the spectra measured on irradiated Schottky diode in Fig. 7a. The spectrum on the HJ d6 diode (red line) was taken with bias of − 10 V and pulsing to − 5 V, i.e. probing the region deeper inside the Ga2O3 film. The spectrum in the Schottky diode (blue line) is shown for the bias of − 5 V and pulsing to − 1 V. The deep traps after irradiation are similar in the Schottky diode and HJ diode, although the peaks amplitudes are considerably lower in the case of heterojunction. Figure 7b depicts the spectra of the HJ diode measured with bias − 2 V and pulsing to 1 V, thus probing the region nearer to the NiO/Ga2O3 interface. The same centers are visible in the spectrum, but the signal is greatly enhanced, suggesting the increase in the trap densities, although accurate quantifying the traps concentration profiles in the HJ is problematic because of the highly nonuniform distribution of charge in the diode (Fig. 4d).

Fig. 7
figure 7

(a) comparison of DLTS spectra of irradiated Schottky diode measured with bias − 5 V and bias pulse − 1 V of the HJ diode measured with bias − 10 V and bias pulse − 5 V; (b) the spectrum of the irradiated HJ diode measured with the bias − 2 V and bias pulse 1 V.

Full size image

The reasons for the observed changes in the properties of the region adjacent to the HJ boundary need further study. From the way our NiO film was deposited on top of the Ga2O3 layer, two processes seem to be suspect in giving rise to the defect structure changes: a) the deposition itself carried out in oxygen-rich conditions that could facilitate the introduction of defects promoted by O-rich conditions, e.g. gallium vacancies, b) subsequent annealing in air at 300 °C that could result in further stoichiometry changes and, possibly, diffusion of Ni into the material, thus contributing to the increased compensation, as Ni in Ga2O3 is a deep acceptor with several charge states and possible formation of complexes with donor defects42. This can contribute to the enhanced compensation in the near-interface region, but it is not obvious if it could give rise to the formation of deep states near EC-0.17 eV visible in AS spectra. Additional experiments with NiO films produced by different means would be useful in gaining a better understanding.

The effects of irradiation include a strong decrease of the hole concentration in p-NiO responsible in part for the increase in the series resistance of the irradiated diodes. The rate of this hole concentration decrease is high, about 2.2 × 105 cm−1, and the effect is presumably due to compensation of acceptor Ni vacancies by interstitial Ni donors9,11,12. Ni vacancies in NiO have been credibly shown to form small polaronic states and give rise to hopping hole conductivity very sensitive to the compensation of Ni vacancy acceptors by Ni interstitials9,11,12 that changes the position of the Fermi level in the hopping conduction band and the effective hole concentration. For higher fluences of irradiating particles, as in Ref.20,21,22, this could explain the observed strong damage in I-V characteristics of irradiated HJs and the strong changes in net donor concentration. Additionally, we see a high decrease rate of the density of the 0.17 eV centers probed at low frequencies and of shallow donors probed at high frequency as a result of irradiation (compare the profiles in Fig. 4c,d).

In Ref.22,43 the compensation has been attributed to gallium vacancy acceptors. It is argued that the application of reverse bias to the top contact strongly facilitates movement of Ga vacancies inside the sample and, as stated in Ref.22, also facilitates the formation of the E2* centers that are believed to be deep acceptors and are assumed to be the main agents of compensation. As already discussed above, in our NiO/Ga2O3 HJs we do observe a much higher carrier removal rate in the bulk of the Ga2O3 films compared to the Schottky diodes and the emergence of the E2* traps absent prior to irradiation of HJs as a result of proton bombardment. However, we have shown above that although the E2* traps indeed emerge as a result of HJs irradiation, which would be expected if these centers comprise complexes of divacancies Ga-O, as suggested in Ref.23 and assumed to be happening in Ref.22, the observed changes in net donor concentration cannot be explained merely due to compensation by E2*, since the density of the E2* formed is too low. It looks more likely that the high electron removal rate could be caused by more effective complexing of VGa acceptors with shallow donors. In principle, this could be a consequence of a higher built-in voltage in HJs compared to Schottky diodes causing pulling the Ga vacancies inside the sample similar to Ref.22, but the difference in built-in voltages between the HJ and the Schottky diode seems to be too low (2 V versus 1 V) to account for the effect (in Ref.22 the effect was detected when applying a reverse bias of − 100 V). As the annealing results discussed below demonstrate, one needs the application of higher voltages around 10 V to strongly influence the movement of deep center defects. It looks more probable that the increased carrier removal rate in HJs is a consequence of the existence of the region of higher defect density near the NiO/Ga2O3 interface and easier creation of point defects in this region.

The observed in our experiments higher carrier removal rate in NiO/Ga2O3 HJs compared to Schottky diodes as a result of proton irradiation would cause a concern in applications in which such irradiation is expected to be a factor, for example in space communications. However, the main areas where Ga2O3 is expected to make its mark lie currently in other directions not necessarily involving working in high radiation environment1,2, where lower reverse current and higher breakdown voltage of the NiO/Ga2O3 HJs16,17,18 outweigh possible disadvantages related to higher carrier removal rate.

As for the possible radiation induced decrease of the breakdown voltage reported in some of the earlier papers20,21, the samples used in the present work with their thin semi-transparent Ni contacts with no thick Au coating, no field plates, no sensible packaging for high voltage measurements were not suitable for accurate measurements of that sort. Preliminary probing with spring contacts resulted in obtaining breakdown voltages close 750 V not strongly affected by irradiation. However, in serious radiation sensitive applications the observed behavior could be a concern and might suggest additional studies. One of the possible directions could be trying other p-type layers in HJs. For example, recently it has been reported that Mg doped p-type Cr2O3 layers show good results for HJs capable of reliable operation at very high temperatures up to 600 °C and could be worth looking into44.

Annealing

Figure 8a displays the evolution of room temperature C-f characteristics of the NiO/Ga2O3 HJ diode before irradiation, after irradiation, and after anneals at 600 K and 673 K. Figure 8b demonstrates the changes in room temperature concentration profiles for the sample before irradiation, the sample after irradiation and the profiles measured after irradiation and annealing from 450 to 673 K for 0.5 h with the temperature step of 30 K. From Fig. 8b, substantial changes in concentration occur starting from the annealing temperature of 570 K, although the full recovery has not been attained even after annealing at 673 K. The activation energy determined from this isochronal annealing experiment was ~ 0.4 eV, if the ratio of concentration at the given anneal temperature to the starting concentration is taken as an approximate measure of the fraction of annealed defects.

Fig. 8
figure 8

(a) Room temperature C-f characteristics measured for the HJ diode before and after irradiation, and after annealing at 600 K or 673 K; (b) concentration profiles observed in the HJ diode after irradiation and different stages of isochronous annealing; (c) DLTS spectra measured for the HJ diode after irradiation and after annealing at 600 K; (d) the spectrum in (d) re-built in coordinates giving the concentrations of the traps in the peaks (without the λ-correction).

Full size image

Figure 8c compares DLTS spectra measured after irradiation and after annealing at 600 K, while Fig. 8d gives the approximate concentrations of the centers observed. Annealing to 600 K produces a strong increase of net donor concentration and is accompanied by a strong increase of the E2* and E3 center concentrations, while the concentrations of the centers themselves are too low both before and after annealing to have significant effect on the density of uncompensated donors, as seen in Fig. 8d in which the DLTS signals were converted to the centers concentrations. This is a rather crude approximation as a) the carrier concentration changes with depth (in calculations, the concentration nearest to the interface was taken) and b) the λ-correction29 is neglected, but this doesn’t change the general conclusion. Thus, none of the major centers seen in DLTS can be the main compensation agents causing the overall changes of the donor concentration, confirming indirectly the assumption made by us earlier that compensation is provided by complexing Ga vacancy acceptors with shallow donors33,34.

Some further insight can be gleaned from the results of annealing experiments on the two diodes, one annealed under the reverse bias of − 10 V (diode d2), the other annealed with no bias (diode d4). The changes in the concentration profiles of the two samples are compared in Fig. 9a,b (we could do annealing with applied bias only when annealing the samples in the Linkam heating stage, which was done in vacuum and the temperature could be made only 600 K in these experiments; for higher temperature annealing, a furnace annealing in flowing air was performed, which makes the comparison of the results for below 600 K and above 600 K not exactly straightforward). As seen in Fig. 9a,b, for annealing at temperatures below 600 K the application of reverse bias during annealing causes the net donor concentration close to the HJ interface to increase faster and the concentration shows a clear-cut decrease with depth, which is not observed when annealing without bias. This is in line with the expected behavior of the postulated complexes of shallow donors and VGa acceptors leading to compensation. Once the complexes are broken and the VGa acceptors released from the complexes, the strong electric field will draw the VGa defects deeper into the sample where the concentration remains low.

Fig. 9
figure 9

(a) changes in concentration profiles of the irradiated HJ diode annealed at temperatures from 450 to 600 K with a 30 K step with the applied reverse voltage of − 10 V, also shown the results of furnace annealing without bias at 673 K and 723 K; (b) the data for another HJ diode annealed without bias. (c) comparison of DLTS spectra measured on two irradiated HJ diodes annealed to 600 K, one with − 10 V applied, the other without bias; (d) comparison of annealing with bias at 600 K and without bias at 673 K.

Full size image

It is also interesting that the application of reverse bias during annealing for diode 2 causes the concentration of the E2* trap to become considerably higher than for diode d4 annealed without bias. We observe the concentration of the E3 trap to behave in the opposite way (Fig. 9c). If we assume that the E3 trap is due to oxygen vacancy donors that would be expected to be driven by the electric field in the opposite direction to the VGa acceptors, one might expect the divacancy complexes, presumably forming the basis of the E2* centers23,31, to increase in concentration towards the HJ interface where the concentration of both types of vacancies will be at maximum.

Annealing at temperatures above 600 K that was done without bias has little effect on the E2* concentration but decreases the concentration of the E3 centers (Fig. 9d). The latter fact confirms the E3 centers in our experiments are radiation defects including native defects, possibly deep VO donors, not the Ti donors described for as-grown Ga2O3 in Ref.31.

At the highest annealing temperature of 723 K we observe a measurable increase of the leakage current, possibly due to the Ni contacts oxidation. This made the high temperature DLTS measurements used to detect the E2* and E3 traps less reliable, but their estimated concentrations were close to the ones obtained after the 673 K annealing.

The results of annealing of the irradiated Schottky diodes with and without applied reverse bias were qualitatively similar to the above-described HJ results, but the degree of changes after the irradiation and during the annealing was much lower.

Methods

The studied epitaxial structures of n-Ga2O3 were commercial samples fabricated by Tamura/Novel Crystals, Japan5. They consisted of n-type unintentionally doped films with (001) orientation and a thickness of 10 μm grown by HVPE on n + Sn-doped, 600 μm-thick substrates prepared by EFG. The net donor density in the films was 2.6 × 1016 cm−3, and the electron concentration in the substrate was 3 × 1018 cm−3. From the 2-in. diameter substrate two pieces of 1 × 1 cm2 were cut from adjacent places close to the center of the wafer. One piece was used to fabricate on the top n-Ga2O3 surface a set of Schottky diodes with semi-transparent (20 nm) Ni and the diode diameter 1 mm, using e -beam evaporation through a shadow mask. Before the Ni Schottky preparation the full-area Ti/Au (20 nm/80 nm) Ohmic contact stack was deposited on the back surface of the sample previously treated in Ar plasma and annealed at 300 °C in forming gas.

NiO growth was performed using the Ion Beam Sputtering (IBS) process8 commonly used to prepare NiO hole transport layers in our perovskite solar cell structures45. The advantage of the IBS technique over more commonly used magnetron sputtering is the alleviation of the problems related to the direct interaction of samples with plasmas containing relatively high-energy particles9. We used an Ni metal target (200 × 150 mm2) with a purity of 99.999% for the reactive conversion to NiO. The oxidation of the sputtered Ni was provided by O2 flow (2.2 cm3/min) directed to the substrate holder. Argon (99.998% purity) was used as a working gas for the generation of the ion beam.

The ion beam source used constant parameters of voltage of 1000 V and current of 100 mA, which enabled a deposition rate of 1.2 nm per minute. Prior to the process, the Ni target was etched with an ion beam for 10 min to remove the oxidized surface layer and possible impurities. The uniformity of deposition thickness was better than 1.5%. NiO films were deposited through a shadow mask with circular openings 1.2 mm in diameter onto the β-Ga2O3 sample with the back Ohmic contact already prepared in the same way as the Schottky diode. After deposition, the Ga2O3 sample with NiO dots was subjected to 1 h long annealing in air at 300 °C to improve the structural, electrical, and optical properties of NiO. The thickness of the deposited NiO film was 300 nm. The film was nanocrystalline, with an average grain size of 14 nm from Atomic Force Microscopy mapping. From transmittance spectrum measurements of the NiO films on satellite sapphire substrates, the bandgap of the deposited NiO film was 3.5 eV.

To independently assess the electrical properties of NiO films they were deposited and annealed under the same conditions through a shadow mask of 1.2 mm diameter on highly conducting commercial Fluoride Tin oxide (FTO) films on glass that are heavily n-type with the donor concentration over 1020 cm−3. The p-type NiO layer forms a p-n heterojunction with n+ FTO. Ohmic contacts were prepared in both cases by deposition of thin Ni contacts. These p+/n+ heterojunction diodes were used to measure capacitance–voltage characteristics and to extract the hole concentration from the 1/C2-V plots11. The temperature dependence of the NiO film conductivity was estimated from the temperature dependence of the NiO/FTO HJ at forward current11. As discussed in Ref.11 such characterization technique provides serious advantages over simple Hall/van der Pauw measurements of satellite NiO films deposited on say sapphire because of the acceptors in undoped NiO are in fact small polarons due to Ni vacancies9,11,12 and the conduction is of the hopping conductivity type with the actual density of holes difficult to establish at room temperature9,11.

The Ni Schottky diodes and the Ni/NiO/Ga2O3 HJs were characterized by capacitance–voltage (C-V) measurements in the frequency range 20 Hz–20 MHz, current–voltage (I-V) measurements in the dark and under monochromatic illumination, admittance spectroscopy (AS) (i.e. measurements of the temperature dependence of capacitance and AC conductance at different measurement frequencies in the frequency range 20 Hz–20 MHz30) and Deep Levels Transient Spectroscopy with electric injection ((DLTS)29 or optical injection (ODLTS)29,36. These measurements were complemented by C-V proofing measurements with monochromatic illumination, the LCV technique33,35. For monochromatic illumination, a set of AlGaN/GaN Light Emitting Diodes (LEDs) with peak wavelengths in the range 259 nm–970 nm was used. The optical power density was 250 mW/cm2 for LEDs with wavelengths 365 nm–970 nm, 15 mW/cm2 for LEDs with the wavelength 277 nm, 1.2 mW/cm2 for LEDs with the wavelength of 259 nm. ODLTS spectra in this work were measured with near-bandedge illumination with the 277 nm LED. Experimental setups are described in our earlier papers33,46,47. Irradiation was performed with protons with energy 1.1 MeV with fluence of 2 × 1013 cm−227. The protons flux was 1011 p/cm2s, the temperature during irradiation was close to room temperature. The samples irradiation was done in a separate facility from that at which characterization was done. Usually, the interval between the irradiation and the measurements was 1–2 days. We have not noticed any changes of the samples characteristics at room temperature without exposing them to elevated temperatures treatments.

Annealing of the samples was performed for temperatures up to 600 K in the HFS350EV-PB4 temperature controller stage (Linkam Scientific, Great Britain) in vacuum either without bias or with no bias, the temperature was changed from 450 to 600 K with the 30 K temperature step, keeping the sample at each temperature for 0.5 h. For higher temperatures up to 723 K the samples were annealed in the tube furnace in flowing air to prevent the loss of oxygen from the NiO layer. Higher temperatures of annealing were not used for fear of damaging and oxidizing the Ni contacts to NiO.

Conclusions

We have shown that heterojunctions of p-NiO/n-Ga2O3 prepared by deposition of p-NiO on HVPE grown epitaxial n−Ga2O3 layer on bulk n+Ga2O3 substrate show a considerable difference in the electrical properties and deep trap spectra as compared to Schottky diodes fabricated on the same material. These changes include a higher series resistance showing a temperature dependence close to that of the NiO, the presence of a strongly compensated region in the immediate vicinity of the NiO/Ga2O3 interface followed by a region of some tenths of a micron thickness containing a high density of deep centers with levels ~ EC-0.17 eV not observed in Schottky diodes. This region of changed electrical properties is also characterized by the absence of the E2* deep traps commonly associated with VGa-VO divacancies and very dominant in Schottky diodes. Irradiation with 1.1 MeV protons with fluence of 2 × 1013 cm−2 leads to much stronger increase of the series resistance of the NiO/Ga2O3 heterojunctions than observed for Schottky diodes. This is accompanied by the strong increase of the resistance of the NiO films caused by a fast decrease of the density of hole concentration. The carrier removal rate observed in NiO is high, on the order of 2 × 105 cm−1 and in a large part explains the observed stronger increase of the series resistance of the heterojunction compared to the Schottky diode. We also saw a very high rate of decrease with irradiation of the net density of the 0.17 eV centers on the order of 6.5 × 103 cm−1. These centers dominate the electrical properties of NiO/Ga2O3 HJs in the region close to the NiO/Ga2O3 interface. In addition, we detected a much higher than in Schottky diodes rate of decrease of the shallow donors with irradiation (850 cm−1 versus 400 cm−1). An introduction of the E2* and E3 deep electron traps was observed after irradiation both in the Schottky diodes and NiO/Ga2O3 HJs. The concentration of the E2* centers is enhanced by annealing at 600 K with reverse bias of − 10 V and is accompanied by the strong increase of the net shallow donor concentration. The model proposed to explain the observed phenomena is that there is a region adjacent to the HJ interface in which the density of compensating acceptors due to the increased concentration of VGa acceptors, and decreased VO-related deep donors contribute to increased compensation of the interfacial region and decreased density of the E2* centers attributed to the VGa-VO divacancies. This region of changed initial defect concentration is the source of strong changes of the concentration of shallow and deep traps after irradiation causing, together with strong changes of the hole concentration in the NiO layer, much more prominent changes of electrical properties of the NiO/Ga2O3 heterojunctions compared to Schottky diodes. This lower radiation tolerance of the NiO/Ga2O3 HJs compared to Schottky diodes could present problems in applications in which irradiation with high energy particles is involved, despite the great advantages of the HJs observed before irradiation.

Annealing up to 723 K almost restores the starting properties of the HJ diodes and Schottky diodes.