Influence of neutron irradiation on deep levels in Ge-doped (010) -Ga2O3 layers grown by plasma-assisted molecular beam epitaxy

The impact of high energy neutron irradiation on the creation of specific radiation-induced deep level defect states and the ensuing influence of these defects on the electronic properties of (010) β -Ga 2 O 3 , doped with Ge and grown by plasma-assisted molecular beam epitaxy, were explored. A significant amount of carrier removal was observed in the irradiated samples exposed to 1 MeV equivalent neutron fluences of 8.5 × 10 14 cm − 2 and 1.7 × 10 15 cm − 2 , which suggests the formation of compensating defects by neutron irradiation. Using a combination of deep level transient/optical spectroscopy (DLTS/DLOS) techniques to probe the entire ∼ 4.8 eV bandgap with high energy resolution, three specific trap states were introduced by neutron irradiation at E C -1.22 eV, E C -2.00 eV, and E C -0.78 eV. Of these, the former two states, observed by DLOS, were also present prior to irradiation, whereas the trap at E C -0.78 eV, observed by DLTS, was not evident prior to neutron irradiation. The radiation dependence suggests that intrinsic point defects are the likely physical sources for these states. Subsequent lighted capacitance-voltage measurements further revealed that three states are the source for the observed strong carrier compensation, with the trap at E C -2.00 eV appearing as the strongest compensating defect for the neutron-irradiated β -Ga 2 O 3


ARTICLE scitation.org/journal/apm
Filling this need is particularly important now since the β-Ga 2 O 3 epitaxial growth field is rapidly evolving; understanding growthrelated defects pre-and postradiation for different epitaxial device structures will be needed. Already, β-Ga 2 O 3 materials and device structures are being grown by molecular beam epitaxy (MBE), [2][3][4] low-pressure chemical vapor deposition (LPCVD), 5 metal-organic chemical vapor deposition (MOCVD), 6 and halide vapor phase epitaxy (HVPE). 7 To date, there have been reports on electron irradiation of β-Ga 2 O 3 revealing carrier compensation 18 and on gamma radiation-induced degradation of gate insulators and surface passivation for β-Ga 2 O 3 MOSFETs. 19 Defect generation due to proton irradiation has been reported in edge-defined film-fed growth (EFG)-grown β-Ga 2 O 3 substrates, [20][21][22] MBE-grown homoepitaxial materials, 20,21 and HVPE-grown materials (including alpha particles) 20,21,23 in studies that employed deep level transient spectroscopy (DLTS) to probe defect states within approximately 1-1.5 eV of the conduction band edge or lighted capacitance-voltage (LCV) measurements, 24 which were used up to an illumination energy of 3.4 eV for the 4.8 eV bandgap material. Neutron irradiation effects have been recently reported using depth-resolved cathodoluminescence and surface photovoltage spectroscopy that were used to optically detect defects that exist with emission energies greater than 1.1 eV in the β-Ga 2 O 3 bandgap. 25 Our previous work on neutron irradiation-induced defects involved both DLTS and deep level optical spectroscopy (DLOS), the combination of which allows quantitative defect spectroscopy across the entire β-Ga 2 O 3 bandgap, which was applied to EFG (010) β-Ga 2 O 3 . 26 However, a comprehensive report on irradiation-induced defects throughout the entire ∼4.8 eV bandgap of epitaxially grown β-Ga 2 O 3 materials is still lacking. Given that the grown-in deep level defect spectra of epitaxial and bulk substrate gallium oxide are known to be different, 14,20,31 understanding how high energy neutron irradiation might affect epitaxial materials differently is of interest, and this work focuses on that question. Here, both DLTS and DLOS measurements are used before and after irradiation at several doses to detect and quantitatively characterize irradiationinduced defect states and defect-specific introduction rates throughout the 4.8 eV bandgap of Ge-doped β-Ga 2 O 3 homoepitaxial layers grown by plasma-assisted molecular beam epitaxy (PAMBE). The epitaxial structures were grown by PAMBE on commercially available (Tamura) EFG-grown β-Ga 2 O 3 Sn-doped (010) substrates. 27 The test structure consisted of, from bottom up, highly Sn-doped (010) substrates (approximate doping ∼5 × 10 18 cm −3 ), ∼200 nm of n+ β-Ga 2 O 3 :Ge to support a subsequently formed n-Ohmic contact, followed by ∼450 nm of lightly Ge-doped β-Ga 2 O 3 to serve as the test layer for the high sensitivity trap spectroscopy measurements. Note that n-type doping using Ge is of interest since Ge 4+ is a favorable donor for β-Ga 2 O 3 due to its close size match with the Ga 3+ cation, and the Ge oxides have a higher vapor pressure compared to that of Si oxides, suggesting that problematic oxidation of the group IV source in MBE growth could be mitigated. 11,28,29 Secondary ion mass spectrometry (SIMS) measurements revealed the n− and n+ Ge concentrations are ∼5-6 × 10 16 cm −3 and ∼1 × 10 18 cm −3 (Fig. 1), making this ideal for DLTS and DLOS studies. The PAMBE growth for the n− and n+ layers was performed within a slightly Ga-rich regime using a substrate temperature of 600 ○ C and a Ga flux (beam-equivalent pressure, BEP) of 1.0 × 10 −7 Torr. The respectively. The black line shows an average Ge-concentration of ∼5-6 × 10 16 cm −3 in the test layer. Inset shows a schematic diagram (not to scale) of our standard Schottky diode structure, with the nominal layer thicknesses. The SIMS profile for Sn shows that there is no out-diffusion of Sn from the Sn-doped substrate into the lightly Ge-doped epitaxial test layer above the detection limits (Sn detection limit ∼2 × 10 16 cm −3 ). There is nominal Si incorporation of ∼2 × 10 16 cm −3 in the epitaxial layer. n− and n+ layers were grown at Ge effusion cell temperatures of 530 and 625 ○ C, respectively. Further information regarding the PAMBE growth methods and conditions of β-Ga 2 O 3 have been previously published. 2,4 After growth, the sample was processed into arrays of 290 μm × 290 μm diodes by depositing an 8 nm semitransparent Ni-Schottky contact that also serves to facilitate optical penetration for DLOS measurements. To form the Ohmic contact, a ∼430 nm Ti/Al/Ni/Au Ohmic metal stack was deposited on the underlying n+ contact layer that was exposed by BCl 3 -based inductively coupled plasma-reactive ion etching (ICP-RIE). 30 This standard Schottky structure has been reported in our previous work. 31 Prior to irradiation, devices were screened using currentvoltage (I-V) and capacitance-voltage (C-V) measurements to ensure that high quality diodes were used for this study and internal photoemission (IPE) was used to assess the Ni/β-Ga 2 O 3 Schottky barrier height (SBH). Deep level spectroscopy measurements and the electrical measurements were performed in the same test diodes before and after irradiation to make reliable comparisons before and after irradiation. The IPE measurements were performed at 300 K 32 using multiple diodes to obtain reliable statistics where the error magnitude was determined from the standard deviation of the dataset as described in earlier work. 15 DLTS measurements were used to quantitatively detect defect states present within ∼1 eV from the conduction band edge, which is based on controlling thermally stimulated emission of electrons to the conduction band from individual defect states in the bandgap. DLOS measurements were used to explore individual optical defect transitions (deep level photoemission) to the conduction band from states within the remainder of the ∼4.8 eV bandgap that is not probed by DLTS due to its thermal emission limits. Both DLTS and DLOS measurements were performed using the same filling bias of 0.00 V and a reverse bias of −1.00 V on the Schottky diodes but with different filling pulse durations of 10 ms and 10 s, respectively. Both DLTS and DLOS monitor the effect of trap emission on the transient and steady state junction capacitance. The DLTS data were analyzed using the conventional double boxcar method for multiple windows between 0.8 s −1 and 2000 s −1 to ensure a high degree of accuracy throughout the scanned temperature range of 77 K-430 K. The magnitude of the individual capacitance peak heights were used to calculate DLTS trap concentrations, and here, we have accounted for the modulated depletion region volume for each particular trap state, which can change for different trap energy levels as a function of the applied bias values (the so-called lambda correction). 33 DLOS measurements were performed at 300 K and used a tunable, monochromatized light source to provide photoexcitation of individual trap states. A 600 W quartz halogen (Q TH ) lamp was used to provide incident photons with energies in the range from 0.50 eV to 2.00 eV, and a 1000 W Xe-lamp was used to provide incident photons with energies in the range from 1.20 eV to 5.00 eV. Two monochromators were used to step the photon energies from 0.50 eV to 2.00 eV and 1.20 eV-5.00 eV in 0.02 eV for the Q TH and Xe-lamp, respectively. The concentration of each DLOS-detected state is obtained from the individual step-heights in steady state photocapacitance (SSPC) spectrum when an incident photon energy resonates with the emission of a trapped electron from a state to the conduction band. However, the precise energy level for each SSPC onset is obtained by analyzing the photocapacitance transient behavior and fitting the optical cross section associated with the onset using the well-known Pässler model, which provides the Franck-Condon energy (D FC ) that accounts for the degree of lattice relaxation associated with each of the trap states. 34 Full details on the trap spectroscopy measurements and analysis methods by DLTS and DLOS can be found in earlier reports. 31,35 Prior to irradiation, the β-Ga 2 O 3 devices were screened using I-V and C-V measurements to ensure the suitability of Schottky diode quality used for this study. The 1 MHz C-V extracted doping for these epitaxial β-Ga 2 O 3 samples yielded uniform doping values of ∼8-9 × 10 16 cm −3 across the test layer. This is higher than the intentional Ge concentration from SIMS in Fig. 1, which additionally shows no evidence of out-diffusion of Sn into the epitaxial test layer from the highly Sn-doped (010) substrate during the growth from the lack of a Sn signal above the SIMS detection limit of 2 × 10 16 cm −3 . Some unintentional background dopants, such as Si, might be contributing to the additional doping as shown from SIMS in Fig. 1.
To perform the irradiation exposure at multiple doses, the sample was subsequently diced into two pieces, and each piece was exposed separately with fast neutrons (energy > 0.5 eV) to a different dose achieved by irradiating for different time durations, 30 min and 60 min, at 450 kW power in the Ohio State University Research Reactor (OSURR) Rabbit facility. The displacement damage dose, D disp , was calculated to be 2.5 × 10 11 MeV g −1 and 5.0 × 10 11 MeV g −1 for the 30 min and 60 min of exposure, respectively. 26,36 The 1 MeV neutron equivalent fluences were obtained as 8.5 × 10 14 cm −2 and 1.7 × 10 15 cm −2 for these respective lower and higher displacement damage doses. It is to be noted that for this fast neutron energy range (1 eV-20 MeV), the mean free path for scattering interaction (λs) is estimated to be greater than 1 cm in β-Ga 2 O 3 using the Monte Carlo N-Particle Transport Code (MCNP) software, 37 which is well beyond the entire epitaxial film of ∼450 nm thickness. Details of this neutron irradiation for β-Ga 2 O 3 are available in previous reports. 26 After the neutron irradiation, the same diodes were screened again by I-V, C-V, and IPE measurements to assess the device quality. The Ni/β-Ga 2 O 3 SBH values extracted using IPE were unaffected with a value of 1.40 ± 0.02 V for both doses (Fig. 2), which ensured that the neutron irradiation did not degrade the Schottky barrier quality. The reverse leakage currents of the Schottky diodes were fairly similar, on the order of ∼1 μA/cm 2 , before and after irradiation for both doses as shown in Fig. 3. However, strong carrier removal effects occurred from the irradiation effects as indicated from the C-V measurements shown in Fig. 4. The C-V extracted net carrier concentration profiles (n = N D Fig. 5(a). A reduction in overall carrier concentration of ∼1.5 × 10 16 cm −3 and ∼3.2 × 10 16 cm −3 is observed compared to the as-grown net doping concentrations for the lower and higher neutron fluences, respectively. This translates to a carrier removal rate of ∼19 ± 0.3 cm −1 estimated with respect to 1 MeV equivalent neutron fluences [ Fig. 5(b)]. However, the carrier removal rate, if calculated using the more simplistic approach of integrated neutron fluences as used in some previous studies with GaN, 38 gives a carrier removal rate of ∼7.9 ± 0.1 cm −1 for PAMBE-grown β-Ga 2 O 3 , which is similar to GaN values of 1-10 cm −1 found using the same method as reported in Ref. 38. Given that this latter, simplistic approach does not account for the damage cross section for a particular neutron energy, we have considered the 1 MeV equivalent neutron fluences throughout as prescribed in our prior work, 26 which considers both neutron flux and damage cross section as a function of neutron energy. This method gives a higher carrier reduction rate compared to the previously used simplistic method of involving only integrated neutron fluences 38 but provides a more accurate and meaningful approach since both the neutron flux and damage cross section are important parameters to evaluate the displacement damages performed by a particular neutron energy present in the fast neutron spectra. Regardless of the calculation method, the significant carrier reduction implies the creation of compensating traps in the bandgap for which DLTS and DLOS measurements were employed to qualitatively and quantitatively detect the compensative trap centers, as discussed next.
DLTS measurements were employed first to detect defect states in the upper ∼1 eV part of the bandgap for which the spectra of before and after irradiation are shown Fig. 6(a), and the trap concentrations are provided in Table I. Before irradiation, the DLTS spectra consisted of shallow states at E C -0.21 eV, E C -0.42 eV, E C -0.60 eV, and E C -0.96 eV with capture cross sections of 6.2 × 10 −16 cm 2 , 1.1 × 10 −15 cm 2 , 1.3 × 10 −14 cm 2 , and 2.0 × 10 −13 cm 2 , respectively. The level near E C -0.21 eV was previously observed in Gedoped epitaxial samples and has an energy, and may be associated with localized states for Ge donor dopants incorporated into octahedral GaII sites as opposed to the tetrahedral GaI sites. 39 After neutron irradiation, two new traps not seen in this material in the as-grown state have emerged, one at ∼E C -0.78 eV with a concentration on the order of ∼10 15 cm −3 and a capture cross section of 7.0 × 10 −14 cm 2 , and a state at ∼E C -0.33 eV with a concentration on the order of ∼10 13 cm −3 and a capture cross section of 2.5 × 10 −15 cm 2 . The ∼E C -0.33 eV trap had not been previously reported. However, there have been several previous reports 20, 23 of the ∼E C -0.78 eV state, labeled E 2 * , seen after proton irradiation of β-Ga 2 O 3 substrates and also in nonirradiated MESFETs 8 grown by PAMBE. The Arrhenius data of these reports (including the capture cross section) match what we see here after neutron irradiation, suggesting that the source of this level is probably related to an intrinsic  It is important to note that this region of the β-Ga 2 O 3 bandgap is crowded with trap data reported from a number of groups, which can confuse the interpretation of data in the literature. 8,13,14,20,21,23,26 As an example, an entirely distinct trap with an energy of ∼E C -0.8 eV, labeled E2, has also been reported by several groups, including ours, but the Arrhenius data over a wide temperature for this trap are entirely distinct from the one we are reporting here. 8,13,14,20,26 Indeed, the ∼E C -0.8 eV (E2) trap has been attributed to Fe impurities that can be present in various materials, and this discernment has been noted by several groups. 8,20 Getting back to the Ec-0.78 eV (E2 * ) trap, theoretical calculations have predicted that gallium vacancies (V Ga ), gallium vacancy-interstitial complexes (V Ga i ), or antisites (Ga O ) have energies near this experimentally determined value, 20,21 and it is reasonable that these intrinsic defects can form upon irradiation with high energy particles due to displacement events from Ga-sites. Finally, for completeness, there is a broad feature in the DLTS scan that appears after radiation whose peak is beyond our DLTS temperature range of 430 K.
We now turn to DLOS to explore the remaining majority of the bandgap. DLOS detected two trap states, one at E C -2.00 eV and another at E C -4.49 eV for which trap energies and D FC values are extracted from the optical cross section analysis using the Pässler model within <5% error of the fitting (Fig. 7). For the E C -4.49 eV state, a sharp optical cross section spectrum is observed with a small extracted D FC = 0.04 eV. The E C -1.22 eV state is associated with a high degree of lattice-coupling with D FC = 0.48 eV, and the near midgap state E C -2.00 eV reveals D FC = 0.56 eV (Fig. 7), typical for deeper states in wide-bandgap materials. 26,40,41 Figure 7 also demonstrates the magnitudes of optical cross sections of the detected states, which reveals a significantly smaller cross section for E C -1.22 eV and E C -2.00 eV states compared to the near valence band state of E C -4.49 eV. A small optical cross section could result from a variety of physical properties, including the spatial size of the defect, its charge state, and large lattice relaxation causing substantial localization of wavefunctions at defect states with strong phononcoupling. 42 Any of these can lead to relatively slow optical emission processes and partial emptying of the trap states as reported earlier in previous work, 43 and this affects the ability to achieve full saturation by stimulated optical emission processes, such as steady state photocapacitance (SSPC) and LCV. 24,26 The SSPC data correlating to the DLOS-detected traps before and after both neutron irradiation fluences are plotted in Fig. 8. Considering the afore-mentioned small optical cross sections of the DLOS-detected deeper states, we turned to LCV measurements to obtain more accurate trap concentrations since LCV measurements can facilitate a longer illumination time (2.5 h used here) compared to DLOS (5 min at each energy increment) in an attempt to mitigate the small optical cross sections and achieve more complete trap saturation and more accurate total trap concentrations. The details of the LCV measurements can be found in previous reports. 24,26 The LCV measurements were performed to sequentially photoionize each defect state at 300 K under sub-bandgap monochromatic illumination for 2.5 h before implementation of C-V sweep.
The photon energies were chosen such that they can emit electrons from all energy levels less than the photon energy, and thus, the difference in net doping between each sequentially applied illumination energy will provide the net carrier removal performed by that state whose energy level exists between the sequential energy steps. With this in mind and noting the DLOS-detected energy levels, the LCV measurements were performed for each of the following incident photon energies: in dark (no illumination) and then under illumination with single energies of 1.80 eV, 4.00 eV, and 4.70 eV photons, with each illumination for 2.5 h. The LCV results are shown in Fig. 9, and the resultant LCV-measured trap concentrations are summarized in Table I. First, the C-V was applied in the dark, holding at reverse bias with a delay of 2.5 h to thermally ionize traps that are closer to the conduction band edge, i.e., those detected by DLTS. However, this "dark-hold" step itself resulted in a carrier concentration recovery of ∼7 × 10 14 cm −3 and ∼1.5 × 10 15 cm −3 for the samples receiving 1 MeV neutron-equivalent fluences of 8.5 × 10 14 cm −2 and 1.7 × 10 15 cm −2 , respectively. In contrast, no such change was observed after LCV measurements were made on the preirradiated samples. These concentrations are comparable to the concentrations of the irradiation-induced DLTS-detected states (E C -0.78 eV and the deeper states near ∼1 eV) for these respective fluences, identifying a correlation that may suggest that these two states contribute to irradiation-induced carrier compensation. Next, for the 1.80 eV illumination for which the excitation energy is sufficient to depopulate the E C -1.22 eV trap but keep the deeper states at E C -2.00 eV and E C -4.49 eV unaffected (filled), there is a slight increase in net doping in the irradiated samples of ∼7.5 × 10 14 cm −3 and ∼1.5 × 10 15 cm −3 in the lower and higher neutron fluences, respectively. This indicates that the E C -1.22 eV trap is also a compensating center and comparable to the traps at E C -0.78 eV and ∼E C -1 eV. For the 4.00 eV illumination, the E C -2.00 eV and E C -1.22 eV trap levels are emitted while keeping the E C -4.49 eV level unaffected, a significant increase in the net doping concentration is observed for the irradiated samples. This implies that the state at E C -2.00 eV is a strong source of carrier compensation. Finally, with the 4.70 eV illumination that would affect the ∼E C -4.4 eV, the same increase in net doping was observed both before and after irradiation, suggesting that while this state appears to contribute to carrier removal, it shows no dependence on neutron irradiation. What this might mean in terms of identifying its physical source is currently under investigation. Figure 10 shows the change in irradiation-introduced trap concentrations for the states at E C -0.33 eV, E C -0.78 eV, E C -1.22 eV, and E C -2.00 eV as a function of neutron fluence, where the introduction rates were found to be 0.03 ± 0.004 cm −1 , 0.85 ± 0.05 cm −1 , 0.89 ± 0.02 cm −1 , and 2.31 ± 0.15 cm −1 , respectively. The summation of these individual introduction rates is lower than the carrier removal rate of ∼19 ± 0.3 cm −1 , which we attribute to the afore-mentioned small optical cross sections and difficulty in achieving their optical saturation, resulting in underestimation of trap concentrations for the E C -1.22 eV and E C -2.00 eV states. The large differences between the carrier removal rates and the summation of individual introduction rates in trap states detected by LCV or photocapacitance measurements have also been reported earlier in HVPE-grown β-Ga 2 O 3 irradiated with proton and alpha particles. 23 A diagrammatic representation of the distribution of bandgap states in the PAMBE sample for the higher fluence case is shown in Fig. 11 in order to more easily visualize the different radiation sensitivities for each defect state. Regarding physical sources for the radiation-sensitive states, we here overview current density functional theory (DFT) calculation results to observe possible correlations but noting that theoretical calculations are continuing to be refined at the time of this writing. 21,44-51 Density functional  , and ∼E C -1.7 eV (D FC of 1.13 eV), which have even larger relaxation energies than the V Ga and related complexes. This shows that there appears to be a reasonable agreement between the trap activation energies of the radiation-sensitive states measured here and a number of intrinsic defects predicted from DFT calculations that may be created upon the displacement damage of Ga and O-sites upon radiation. Regarding the ∼E C -4.4 eV state, the source is currently controversial. For example, here, we find its concentration to be unaffected by irradiation for both doses, implying that its physical source is unlikely to be an intrinsic point defect. However, this is in apparent conflict with DFT studies that suggest the presence of an energy level near this bandgap position attributing to vacancies. 48 Another, a recent DFT study has reported the presence of an extrinsic state due to nitrogen substituting on an oxygen site (N O II ) near this energy, 49 whereas a different work has suggested a tentative association of this state with a self-trapped hole process 50 with a correlation to theoretical predictions. 51 Due to these uncertainties and the ubiquitous presence of this DLOS feature, the physical source for this state remains under active investigation.
To summarize, an investigation of neutron irradiation-induced effects has been performed on Schottky diodes fabricated on PAMBE-grown Ge-doped β-Ga 2 O 3 Schottky diodes. Strong carrier compensation was observed, at a rate of ∼19 ± 0.3 cm −1 , based on 1 MeV equivalent neutrons. The sources for the radiation-induced compensation were shown to result from three specific traps at E C -0.78 eV, E C -1.22 eV, and E C -2.00 eV. Of the nine trap states seen in the preradiated material, only these three showed sensitivities to neutron irradiation. This work and other works tentatively suggest that the source of the E C -0.78 eV trap, which was not present in the preirradiated samples, is related to vacancyrelated complexes or Ga O antisites. For the E C -1.22 eV trap and the E C -2.00 eV traps, there are different acceptorlike intrinsic origins predicted such as gallium vacancies, gallium vacancy-interstitial complexes, and hydrogenated gallium vacancy complexes. LCV measurements identify these irradiation-induced trap states to be the most strongly correlated with the observed carrier compensation mechanism. The ubiquitous state at ∼E C -4.4 eV from the DLOS measurement was insensitive to radiation, inviting additional studies to clarify its origin.