Characterization of deep-level defects in OFZ grown Nb-doped β-Ga₂O₃ single crystals Open Access

Gallium oxide (β-Ga₂O₃) with its ultra-wide bandgap of 4.8 eV enables high breakdown fields in addition to a relatively large Baliga figure of merit (BFOM) compared to materials such as Si, SiC, and GaN.^1,2 The ease of large-area single-crystal growth makes β-Ga₂O₃ a cost-effective alternative to the existing wide bandgap materials.³ However, defects that form in the doped crystals during melt-growth can significantly impact the electrical properties of the material, thereby compromising the device performance. This makes defect characterization in doped β-Ga₂O₃ crucial for optimizing material behavior and improving the performance of β-Ga₂O₃-based power electronics in real-world applications. Melt growth of β-Ga₂O₃ produces large and high-quality, single crystals, ensuring superior electrical properties, and scalability for cost effective industrial production, with the main techniques being Czochralski (CZ) technique,⁴ edge-defined film-fed growth (EFG) technique,⁵ Bridgman technique,⁶ and the optical floating zone (OFZ) technique.⁷ The OFZ technique offers significant advantages, operating in a crucible-free environment that minimizes contamination—an issue common in other methods such as CZ, EFG, and Bridgman. Despite OFZ growth producing smaller crystals than CZ, the growth process is faster and cost-effective, making it an ideal technique for research and development.

With the growing demand for highly conductive substrates, doping β-Ga₂O₃ with transition metal elements such as niobium (Nb) has gained significant attention, with density functional theory studies highlighting Nb as a promising candidate for enhancing n-type conductivity in Ga₂O₃.^8,9 Nb is predicted to be a double donor in β-Ga₂O₃ with small activation energies for ionization of the first electron.^10,11 In one of the studies, the activation energies were calculated as 0.03 and 0.15 eV for a Nb atom at the tetrahedral [Nb_Ga(I)] and octahedral [Nb_Ga(II)] sites, respectively.⁹ Another recent study reported these values to be 36.9 and 36.3 meV for the tetrahedral and octahedral sites, respectively.¹¹ In addition, the ionic radius of Nb⁵⁺ (0.64 Å) closely aligns with that of Ga³⁺ (0.62 Å), offering a significant advantage over commonly used n-type dopants such as Si (0.40 Å), Ge (0.53 Å), and Sn (0.69 Å).¹² Despite its potential, experimental investigations of Nb-doped β-Ga₂O₃ remain limited.^13–15 Thus, further experimental analysis is crucial to fully explore the potential of Nb doping, and comprehensive defect studies are essential in order to understand the electronic properties of the material.

Deep-level transient spectroscopy (DLTS) is a key technique for investigating deep-level defects within the bandgap, offering valuable insights into electronic properties of the traps that can influence device behavior and performance.^16,17 Advanced techniques such as Laplace-DLTS offer significantly higher energy resolution for defects with close deep-levels and are essential to gain a deeper understanding of the defect landscape in wide-bandgap material.^18,19 Throughout the literature, DLTS has been applied extensively to bulk β-Ga₂O₃ crystals grown via OFZ,^13,20,21 CZ,^22,23 and EFG,^24,25 in addition to homoepitaxial layers grown via hydride vapor phase epitaxy,^25–27 plasma-assisted molecular beam epitaxy,^28,29 and metal–organic chemical vapor deposition.^30–32 

In this work, we studied the structural and electrical properties of Nb-doped (0.05 and 0.1 mol. %) β-Ga₂O₃ single crystals grown via the OFZ method and applied conventional DLTS combined with high-resolution Laplace-DLTS (L-DLTS) on 0.05 mol. % of Nb-doped β-Ga₂O_3, to provide a precise view of the defect states present within the bandgap of the material. Compared to conventional DLTS, L-DLTS revealed finer details of the defect landscape by resolving the broad E2 electron emission signal into three distinct sublevels (E2a, E2b, and E3), showcasing its enhanced capability in characterizing closely spaced defect levels. Our findings further suggest that these levels arise due to significant concentrations of Fe and Ti impurities in the grown crystal, potentially originating from the source material. In addition, a surface-related defect (E_s) was observed. These insights into the electrically active defects are crucial for refining doping strategies and improving the quality of β-Ga₂O₃ substrates for power electronic applications.

Nb-doped β-Ga₂O₃ single crystals were grown using the OFZ technique with a CSC FZ-4000 apparatus. Alfa Aesar provided Ga₂O₃ powder of 5N purity and Nb₂O₅ of 4N purity, which were taken as the starting materials and Nb doping concentration of 0.05 and 0.1 mol. % were chosen. Powders of Ga₂O₃ and Nb₂O₅ were mixed according to the desired stoichiometry and ground to enhance dopant uniformity. This mixture was packed into a latex tube to form a feed rod with a 10 mm diameter and a 60 mm width, which was then compressed at 70 MPa using a hydrostatic press. After removing the latex tube, the rod was sintered at 1350 °C for 24 h, yielding a dense, uniform ceramic feed rod with minimal porosity. For crystal growth, the feed rod was positioned inside the furnace on top of a pre-existing β-Ga₂O₃ crystal with (100) orientation, which was used as the seed. The feed rod was melted by focusing light from 4000 W halogen lamps using reflective mirrors within the OFZ setup. Both the seed and feed rods were counter-rotated at 20 rpm. The crystal growth rate was maintained at 10–12 mm/h, and the growth process was conducted in a compressed dry air atmosphere at a flow rate of 2 l/min. After crystal growth, the ingot was sliced into wafers using a diamond wire saw. The wafers were subsequently polished using chemical mechanical polishing (CMP) to achieve a smooth surface for device fabrication. For electrical characterization, vertical Ni-based Schottky barrier diodes (SBDs) were fabricated. The samples were initially cleaned using a standard solvent cleaning using acetone, isopropyl alcohol, and de-ionized water. Using e-beam evaporation, backside Ti/Au (30/150 nm) metallization was deposited, and the samples were annealed at 450 °C in N₂ atmosphere to form ohmic contacts. On the topside, an array of circular features with a diameter of 300 μm were patterned using lithography, and Ni/Au (30/150 nm) Schottky contacts were formed using e-beam evaporation and liftoff. The processed samples were subsequently mounted and wire bonded to ceramic substrates for performing electrical measurements, including DLTS.

Figure 1(a) shows the x-ray diffraction (XRD) pattern of the melt-grown Nb-doped Ga₂O₃ samples with 0.05 and 0.1 mol. % of doping concentration, displaying sharp and distinct reflections from the (100) family of planes. This confirms that the sliced substrate maintains a well-defined (100) orientation. Rocking curves from high-resolution XRD (HR-XRD) scans corresponding to (400) planes in Figs. 1(b) and 1(c) reveal a peak with full width at half maximum (FWHM) value of 150 arcsec for the 0.05 mol. % sample and 170 arcsec for the 0.1 mol. % sample, indicating a good-quality crystal for both the samples. Increasing Nb doping concentration in β-Ga₂O₃ from 0.05 to 0.1 mol. % leads to a rise in FWHM due to the lattice strain caused by the substitution of larger Nb⁵⁺ ions (0.64 Å) for smaller Ga³⁺ ions (0.62 Å) at octahedral sites. Even though the variation is small, this size and valence mismatch introduces local distortions, which result in increasing micro-strain and broadening the rocking curves. In addition, higher doping levels increase the defect concentrations, enhancing the mosaicity and strain, which lead to a higher FWHM in the 0.1 mol. % Nb doped sample.

The slight shift in the HR-XRD peak positions for the samples is due to a minor tilt in the wafers resulting from the wafer slicing process.³³ Atomic force microscopy (AFM) images in Figs. 1(d) and 1(e) show a smooth surface morphology after CMP, with RMS roughness values of 1.2 and 1.9 nm for 0.05 and 0.1 mol. % doped samples, respectively. However, further CMP optimization could minimize some line-shaped features on the surface caused by manual lapping process.³⁴ 

Hall effect measurements at room temperature were conducted on Nb-doped Ga₂O₃ (0.05 and 0.1 mol. %) samples using Ti/Au ohmic contacts sputtered onto the four corners of the sample. The obtained results summarized in Table I, indicate a free electron concentration of 6.1 × 10¹⁷ cm⁻³ and electron mobility of 83 cm²/Vs for the 0.05 mol. % doped sample. Increasing the Nb doping concentration to 0.1 mol. %, resulted in a higher carrier concentration of 1.20 × 10¹⁸ cm⁻³, accompanied by a slight reduction in mobility to 75 cm²/Vs. The increase in carrier concentration is attributed to a greater number of Nb atoms acting as donors and provides convincing evidence that the dominant donor is Nb. The decrease in mobility results from enhanced ionized impurity scattering and slight degradation in the crystal quality at higher doping levels, as indicated by the HR-XRD FWHM values in Figs. 1(b) and 1(c). Overall, the Hall effect measurement data show that Nb doping effectively tunes the conductivity of β-Ga₂O₃, making it a viable doping strategy for optimizing the material’s electrical properties.

Figure 2(a) shows the current density–voltage (J–V) characteristics of a 300 μm diameter Ni-based vertical Schottky barrier diode fabricated on the sample with 0.05 mol. % Nb doping, which is used for all electrical device measurements in this work. The turn-on voltage (V_on) increases from 0.91 to 1.18 V as the temperature decreases from 295 to 40 K, which is likely a consequence of the changes in the barrier height. Figure 2(b) shows the J–V characteristics under reverse bias, with the leakage current ranging between 10⁻⁵–10⁻⁶ A/cm² at −5 V for both 295 and 40 K, which is sufficiently low to conduct DLTS analysis without compromising the signal-to-noise ratio. The junction spectroscopy measurements were carried out on this sample only as the higher carrier concentration in the 0.1 mol. % Nb-doped sample induced an increased leakage current under reverse bias conditions. Figure 2(c) depicts the doping concentration profiles in the sample derived from capacitance–voltage (C–V) measurements at 295 and 40 K, with the bias voltage varying from −10 to 0 V and 1 MHz frequency of the probing AC signal. Effective doping concentration (N_d–N_a) and depletion depth (W) values have been obtained using a static dielectric constant of 10.2.³⁵ The details for extracting the doping concentration and depletion width from the C–V measurements are provided in the supplementary data of this article. Figure S1 presents the 1/C² vs V plot used for this analysis. The doping concentration profiles measured at 295 and 40 K are relatively uniform within the probed regions with the average concentration values of about 8.1 × 10¹⁷ cm⁻³ at 295 K and 6.7 × 10¹⁷ cm⁻³ at 40 K. The effective electron concentration obtained from Hall effect measurements closely matches with the values determined from the local C–V measurements. This suggests a nearly uniform carrier concentration throughout the depth of the samples, indicating the homogeneous incorporation of Nb dopants.

Figure 2(d) shows a capacitance–temperature (C–T) scan recorded from 18 K up to 430 K at 1 MHz, with an applied bias of −5 V. A capacitance drop of about 25 pF between 400 and 200 K can be attributed to ionization of deep-level defects. Below 200 K, the capacitance changes with temperature are not significant down to 18 K, showing no evidence of carrier freeze-out for the free electrons resulting from Nb dopants. In contrast to other studies on Ga₂O₃, which exhibit carrier freeze-out at 20 K,^36,37 the absence of such behavior in this sample is consistent with the predicted small activation energy of Nb dopants in Ga₂O₃. This conclusion is further supported by the effective carrier density calculated from the concentration depth profile in Fig. 2(c), which do not decrease significantly at low temperatures. These results confirm that Nb in β-Ga₂O₃ acts as an effective donor with a shallow level, consistent with theoretical predictions.¹¹ 

The distribution of deep level defects within the bandgap of the sample was analyzed using conventional deep level transient spectroscopy (DLTS). To probe the region near the metal–semiconductor (MS) interface, a reverse bias (U_b) of −3 V with a pulse bias (U_p) of −0.1 V was applied to the Schottky contact with a rate window (e_n) of 50 s⁻¹ and a filling pulse length (t_p) of 1 ms. According to C–V measurements, with the applied bias and pulse voltages, the region from 50 to 70 nm from the MS interface was probed. Figure 3(a) shows the conventional DLTS spectra of the sample on the logarithmic scale with the aforementioned measurement parameters. Two distinct defect peaks, E_s with its peak maximum at 150 K and E2 at 390 K were observed. The strong broad E2 peak overshadowed the E_s peak in the full-spectrum DLTS in the temperature range of 40–430 K. Figure 3(b) shows the DLTS spectra on a linear scale from 120 to 220 K recorded with a rate window of 200 s⁻¹ and a filling pulse length of 1 ms revealing a peak at ∼170 K corresponding to the E_s defect, with a relatively low trap concentration of ∼10¹⁴ cm⁻³. Figure 3(c) shows the DLTS spectra from 40 to 430 K with a rate window of 50 s⁻¹, dominated by the strong broad E2 peak with a trap concentration of ∼10¹⁷ cm⁻³.

The E2 peak in the spectra shown in Fig. 3(c) is related to a well-known electron trap in the β-Ga₂O₃ material, which was consistently observed in samples produced by various growth techniques; this defect is associated with iron (Fe) impurities.^25,39–44 The concentration of the E2 trap in our samples is higher than 10¹⁷ cm⁻³, indicating a significant presence of unintentional Fe incorporation. Fe replaces Ga at the tetrahedral site [Fe_Ga(I)] and the octahedral site (Fe_Ga(II)), with calculated (0/−) levels at E_c −0.68 and E_c −0.78 eV, respectively.⁴³ 

The primary source of Fe impurity is the Ga₂O₃ powder used as the starting material for the single crystal growth. Previous studies, supported by glow discharge mass spectrometry, have confirmed the presence of Fe in both the source powder and the grown crystals.⁴⁵ As detailed Sec. II, the Ga₂O₃ powder used in this work is of 5N purity. Despite its high quality, trace amounts of impurities, including Fe, are inevitably present, making Fe incorporation a persistent challenge in Ga₂O₃ material regardless of the growth technique employed.

Further analysis of the defect levels was conducted using Laplace DLTS (L-DLTS), an advanced technique that enables the separation of closely spaced defect levels with similar energy states.^17–19  Figures 5(a) and 5(b) present the results of L-DLTS analysis for the two peaks identified in conventional DLTS. L-DLTS is an isothermal technique that provides significantly enhanced resolution by analyzing all the data in a capacitance transient recorded at a fixed temperature and plotting an emission rate spectrum, allowing for more accurate differentiation of overlapping defect states. This higher resolution is essential in cases where broad peaks observed in conventional DLTS may consist of multiple discrete defect levels that cannot be individually resolved.

To mitigate the potential influence of the electric field under the junction on the electron emission from deep-levels, a “double” L-DLTS approach was employed,^18,46 wherein, two pulse biases are applied with a fixed reverse bias to test a small window within the depletion region, where the gradient in the electric field can be minimized. U_b of −8 V with pulse biases of U_p1 = −3 V, U_p2 = −4 V and a filling pulse length of 10 ms was used for the analysis. The L-DLTS spectra in Fig. 5(a) illustrate the splitting of the broad E2 into three discrete deep-level defects, namely, E2a, E2b, and E3. Figure 5(b) represents the L-DLTS spectra obtained for the E_s trap. Due to the relatively small signal and proximity of this defect to the sample surface, a single U_p of −0.1 V with a U_b of −2 V was used, along with a t_p of 50 ms. The shift in the E_s peak in the L-DLTS spectra as the temperature is increased from 162 to 170 K is an expected behavior, as the emission rate $eem$ [x axis in Fig. 5(b)] is temperature-dependent. This relationship is governed by Eq. (2); that is, the electron emission rate from corresponding deep level will increase with increasing temperature. Figure 6(a) shows the Arrhenius plot for each of the peaks detected in the L-DLTS; the corresponding defect signatures and concentrations are summarized in Table II for each of the identified traps. From Table II, the activation energy for the electron emission, ΔE_c for the E_s defect, obtained from the Arrhenius plot is 0.28 ± 0.01 eV with an apparent capture cross section of 9.9 × 10⁻¹⁶ cm² and trap concentration of 2 × 10¹⁴ cm⁻³. This trap has been previously identified as a point defect arising specifically from mechanical polishing, as already discussed above.²¹ We note that a trap with an electron activation energy of 0.28 eV has been reported in Ga₂O₃ samples irradiated with 18 MeV α-particles, which could also suggest it may have a possible association with a native lattice defect;²⁶ however, this may not necessarily correspond with E_s observed within our samples. In that study, the Ga₂O₃ sample was irradiated with 18 MeV α-particles, a process that likely causes significant lattice displacement and damage.²⁶ In contrast, our study involves an as-grown crystal. Furthermore, the E_s trap observed in our sample is confined to a very narrow sub-surface region and vanishes at depth exceeding 100 nm, indicating it is not a bulk-related defect. Previous studies have shown that a sub-surface trap with similar activation energy for electron emission can arise due to mechanical polishing, suggesting the interpretation of E_s as a surface-damage related defect.²¹ We do not rule out that the surface-damage related E_s trap observed in our work and in Ref. 21 of this article as well as the radiation-induced defect detected in Ref. 26 of this article could be the same defects; however, we do not have sufficient evidence to prove that these reported traps are equal. There is a possibility that these DLTS emission signals could originate from different defects, which possess similar activation energy for electron emission.

The E2a and E2b traps are attributed to Fe_Ga(I) and Fe_Ga(II), respectively, and E3 is identified as a deep donor linked to a titanium-related defect Ti_Ga(II).⁴¹ For the Fe-related defects, E2a and E2b, the measured ΔE_c values are 0.68 ± 0.01 and 0.71 ± 0.04 eV, respectively. Their corresponding σ_app values are 1.1 × 10⁻¹⁵ and 1.2 × 10⁻¹⁵ cm² with a significantly high trap concentration of the order 10¹⁷ cm⁻³. Figure 6(b) shows the dependence of the electron emission rates e_em on the electric field (E) for the E2a and E2b traps. This analysis was performed with a fixed U_b of −8 V while maintaining a difference in pulse biases, U_p1 − U_p2, of 1 V. The linear dependence of ln(e_em) on the square of the electric field is indicative of emission via the phonon-assisted tunneling mechanism, as expected for an acceptor level in an n-type material,^43,47 indicating that both E2a and E2b are deep acceptor levels. Consequently, these trap levels could play an important role in reducing β-Ga₂O₃ conductivity by compensating and capturing the free electrons in Ga₂O₃.^41,48 The E3 defect, attributable to Ti impurities substituting Ga on the octahedral site, has ΔE_c values of 0.89 ± 0.01 eV, a σ_app of 1.2 × 10⁻¹⁴ cm², and a N_T of 10¹⁶ cm⁻³. Note that Ti is also an unintentional impurity that originates from the source powder.⁴⁵ 

Secondary ion mass spectrometry (SIMS) analysis (Model: CAMECA IMS 3f) was performed on Nb-doped β-Ga₂O₃ samples with Nb concentrations of 0.05 and 0.1 mol. %. As a reference sample for absolute quantification was unavailable, analysis was conducted using relative quantification between both samples. Figure 7 confirms the presence of Nb, Fe, and Ti in the sample through SIMS analysis. From the data, it is evident that the Fe concentration remains similar in both samples, while the Nb concentration is higher in the 0.1 mol. % sample than in the 0.05 mol. % sample, inferring that origin of the Fe impurities is from the Ga₂O₃ source powder rather than Nb₂O₅.

The findings suggest that the E2a, E2b, and E3 defects may play a significant role in influencing the properties of OFZ grown β-Ga₂O₃: Nb. All these defects can be attributed to impurities originating from the source material (Ga₂O₃ powder) used for crystal growth.⁴⁵ Nb doping was utilized to enhance the conductivity of β-Ga₂O₃ by increasing the free electron concentration. However, Nb exhibits a segregation co-efficient of less than one, meaning that Nb tends to remain in the melt during the growth. As a result, higher Nb incorporation occurs in the tail-end of the crystal. In our recent work on Ta-doped β-Ga₂O₃, ICP-OES analysis revealed that less than 10% of the nominal doping concentration (0.05 mol. %) was incorporated into the crystal.³³ Assuming similar behavior for Nb doping, it is evident that a significant portion of the dopant is lost due to segregation during the growth process. Furthermore, the presence of E2a and E2b defect levels act as compensators, reducing Nb dopant activation and hindering conductivity by trapping free electrons.

To maximize the benefits of Nb doping and achieve higher conductivity in β-Ga₂O₃, it is necessary to address these challenges. Apparently, minimizing concentrations of the Fe compensating defects is critical. This can be done by using ultra-high-purity or rigorously purified source powders. In addition, optimization of the growth parameters to control the segregation loss during the growth is needed. Such efforts will help unlock the full potential of Nb doping in β-Ga₂O₃.

This study demonstrates the first application of high-resolution Laplace-DLTS to Nb-doped β-Ga₂O₃ single crystals grown via the OFZ technique. This advanced characterization resolves closely aligned energy levels within the well-known E2 defect band in β-Ga₂O₃, identifying three distinct deep levels (E2a, E2b, and E3) with precise values of their activation energies and capture cross sections. The E2a (0.68 eV) and E2b (0.71 eV) traps originate from Fe atoms at different Ga sites, which act as compensating acceptors, while E3 (0.89 eV) is associated with Ti impurity. In addition, a surface-related defect (E_s) with activation energy of 0.28 eV is identified, underscoring the importance of surface preparation in device fabrication. A key finding highlights that, in addition to segregation losses during the growth, the presence of Fe-related deep acceptor level defects impacts Nb doping levels and reduces the achievable free carrier concentration in β-Ga₂O₃ single crystal substrates. These findings highlight the importance of source material purity, particularly regarding Fe impurities, in achieving optimal electrical properties in β-Ga₂O₃. The insights presented lay the groundwork for future investigations into defect engineering and doping strategies in β-Ga₂O₃.

The supplementary material provides details on the extraction of doping concentration and depletion region width from the capacitance–voltage (C–V) measurements for the Nb-doped β-Ga₂O₃ (0.05 mol. %) Schottky barrier diode. Figure S1 shows the C–V and corresponding 1/C²–V plots used for analysis, recorded at 295 and 40 K at a measurement frequency of 1 MHz.

We acknowledge the Commonwealth Scholarship Commission for funding this research under the Commonwealth Split-Site Ph.D. scholarship (2023–2024) and the SJSGC fellowship by University Grant Commission (UGC), Ministry of Education, Government of India (Grant No. UGCES-22-OB-KER-F-SJSGC-7611) for V. L Ananthu Vijayan. M. Kuball acknowledges financial support from the Royal Academy of Engineering through the Chair in Emerging Technologies Scheme. Also, we acknowledge in part financial support from the Engineering and Physical Science Research Council under the Innovation and Knowledge Centre (IKC) REWIRE under grant number EP/Z531091/1. The authors in Manchester acknowledge support from the EPSRC-UK under Contract Nos. EP/T025131/1 and EP/S024441/1. Christopher A. Dawe acknowledges support from the EPSRC CDT in Compound Semiconductor Manufacturing.

The authors have no conflicts to disclose.

Ananthu Vijayan V L: Conceptualization (lead); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (lead); Methodology (equal); Writing – original draft (lead); Writing – review & editing (equal). Christopher A. Dawe: Data curation (equal); Formal analysis (equal); Methodology (equal); Writing – review & editing (equal). Sai Charan Vanjari: Formal analysis (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal). Vladimir P. Markevich: Supervision (equal); Validation (equal); Writing – review & editing (equal). Matthew P. Halsall: Supervision (equal); Validation (equal); Writing – review & editing (equal). Anthony R. Peaker: Supervision (equal); Validation (equal); Writing – review & editing (equal). Moorthy Babu Sridharan: Funding acquisition (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal). Martin Kuball: Funding acquisition (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal).

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