Assignments of vibrational lines to OD-impurity complexes for adventitious impurities in β-Ga₂O₃

β-Ga₂O₃ is a transparent conducting oxide with an ultrawide bandgap (∼4.8 eV) that shows promise for next-generation high-power, deep-UV, and extreme-environment devices.^1–5 The effect that defects and impurities have on the electrical properties of β-Ga₂O₃ is being widely investigated.^6,7

β-Ga₂O₃ has a monoclinic crystal structure with inequivalent Ga(1) and Ga(2) sites that are tetrahedrally and octahedrally coordinated, respectively.^8,9 Vacancies on the Ga(1) site are prevalent in β-Ga₂O₃^10,11 and have been predicted by theory to have split configurations of type “a,” “b,” and “c.”^12–14 Infrared (IR) spectroscopy,¹⁵ electron paramagnetic resonance (EPR),^16,17 scanning-transmission-electron microscopy,¹⁸ and positron annihilation methods^10,11 have revealed the experimental properties of split Ga(1) vacancies in β-Ga₂O₃.

Impurities in β-Ga₂O₃ that are of particular interest in this Letter include Si, Fe, and H (along with its heavy isotope deuterium, ²H or D). Si and Fe are adventitious impurities in β-Ga₂O₃ that are readily introduced during crystal growth.^6,7,19,20 Hydrogen impurities are also readily introduced into β-Ga₂O₃ during crystal growth and device processing.^6,7,21 Interstitial H (H_i) and H at an oxygen vacancy (H_O) have been predicted to be shallow donors.²² H also forms complexes with V_Ga1 deep acceptors^23–25 and dopant impurities, such as Si.²⁶ 

Vibrational spectroscopy has proved to be a powerful probe of H and D centers in semiconductors and transparent conducting oxides.^27–29 The dominant O–H center in β-Ga₂O₃ gives rise to an O–H line at 3437 cm⁻¹ that was assigned to a $VGa1ib−2H$ complex where V_Ga1 takes the “b” split vacancy configuration.^15,24 (A few different notations have been used to describe the different split vacancy configurations. We adopt the notation used recently in Refs. 10–14.) The corresponding O–D center has a vibrational frequency of 2546 cm⁻¹. Additional O–H and O–D centers that include V_Ga1 in its “c” configuration have also been reported.^23,24,30 (O–H and O–D centers involving the “b” and “c” configurations of V_Ga1 have characteristic polarization properties that allow them to be distinguished from each other.^{23,24,30–32}) Presently, vibrational lines for more than a dozen O–H and O–D complexes in β-Ga₂O₃ have been found.^{23,24,33–36} The O–D lines could be detected with a higher signal-to-noise ratio in our experiments than the corresponding O–H lines, so O–D centers are the focus of this Letter.

To explain the many O–D lines that have been observed, sometimes with very similar vibrational properties, these O–D centers have been suggested to be due to $VGa1ib−D$ or $VGa1ic−D$ complexes that also include an additional impurity at a nearby substitutional Ga site to form an OD-impurity complex.^24,32 The assignments for OD-impurity complexes have been based primarily on the dominant impurities present in doped β-Ga₂O₃ samples.^26,32–36 However, truly unambiguous assignments would be based on additional information, such as an impurity isotope effect, for example. However, to our knowledge, no such isotope data exist except for the H vs D isotope effect itself that confirms the role of H in the vibrational spectra for hydrogen centers.

The assignment of O–D vibrational lines to specific additional impurities is especially difficult when the impurities are adventitious in the material system. Such is the case for Si and Fe impurities in β-Ga₂O₃.^6,7,19,20 Understanding the reactions of Si and Fe with hydrogen is of interest because these adventitious impurities are also commonly used as dopants to control the electrical properties of β-Ga₂O₃ where Si substituting on a Ga(1) site is used to dope β-Ga₂O₃ n-type³⁷ and Fe substituting on a Ga(2) is used to make β-Ga₂O₃ semi-insulating.³⁸ 

A previous work that has attempted to assign O–D vibrational lines to specific OD-impurity complexes has led to seemingly contradictory results^23,26 and demonstrates the difficulty of assigning vibrational lines when multiple impurities are present in samples. Figure 1(a) shows the O–D lines observed for an Si-doped epilayer grown by molecular beam epitaxy (MBE) on an Fe-doped substrate that was deuterated by exposure to a D plasma.²⁶ Secondary ion mass spectrometry (SIMS) results in this case showed an Si concentration of 5 × 10¹⁸ cm⁻³ in the 0.5 µm thick epilayer. These results led to the assignment of the 2577 cm⁻¹ line to an OD–Si complex. The line at 2585 cm⁻¹ in the same spectrum was assigned to an OD–Fe complex arising from Fe in the substrate. Figure 1(b) shows the O–D lines for an Fe-doped substrate obtained from the Tamura Corp. The increased relative intensity of the 2585 cm⁻¹ line appears to support its assignment to an OD–Fe complex. However, Fig. 1(c) shows the O-D lines observed for a heavily Fe-doped Ga₂O₃ sample that had been deuterated by annealing at high temperature in a D₂ ambient.²³ This β-Ga₂O₃ boule was grown by using the Czochralski method at Synoptics and was doped with Fe by adding Fe₂O₃ (0.100%) to the melt. SIMS results found an Fe concentration of [Fe] = 4 × 10¹⁸ cm⁻³ and an unintentional Si concentration of [Si] = 1 × 10¹⁷ cm⁻³. In this case, it is tempting to assign the stronger line at 2577 cm⁻¹ shown in Fig. 1(c) to an OD–Fe complex. The results in Fig. 1 show that reliable assignments of the vibrational lines of OD-impurity complexes cannot be made based on their relative intensities and SIMS results alone when the impurities involved can be unintentionally present with high concentration.

To provide additional evidence for the assignment of the O–D vibrational line of the OD–Si complex, we have obtained a set of Si-doped β-Ga₂O₃ epitaxial layers grown by metal-organic vapor deposition (MOCVD) on Fe-doped bulk β-Ga₂O₃ substrates, with layer thicknesses ranging from 1.2 to 12 µm, to investigate the thickness dependence of the OD-impurity lines. These samples were grown to have a nominal Si concentration of [Si] = 2 × 10¹⁷ cm⁻³. Additional growth details are given elsewhere.^39,40 Room temperature Hall measurements made for these samples found free-carrier concentrations of ≈3 × 10¹⁷ cm⁻³, i.e., near the nominal Si doping value.

To introduce D into Si-doped β-Ga₂O₃ epitaxial layers with different thicknesses, samples were annealed at 900 °C for 3 h in sealed quartz ampoules in a D₂ ambient (2/3 atm at room temperature).⁴¹ To quickly terminate the annealing treatments, ampoules were quenched in sand. To investigate the O–D centers produced in the samples, infrared absorbance measurements were made with a Nicolet iS50 Fourier transform infrared spectrometer equipped with a CaF₂ beam splitter and a liquid-N₂-cooled InSb detector. The samples were cooled to 77 K in a Helitran cryostat for IR measurements.

These samples were also characterized by electron paramagnetic resonance (EPR) at 300 K at two laboratories with commercial EMX Bruker spectrometers. EPR for the thinnest epilayer was done employing a microwave cavity resonator with enhanced sensitivity operating at 9.39 GHz (UAB), while EPR for the three thicker epilayer samples was obtained using a cylindrical microwave cavity at 9.77 GHz (NRL). The spectra observed with B⊥〈010〉 for samples with different layer thicknesses and approximate dimensions 10 × 3 mm² are shown in Fig. 2(a). A single sharp signal was found with a Zeeman splitting g-value of 1.9585 and full-width-at-half-maximum linewidth between 0.25 and 0.95 G. These magnetic resonance parameters are very similar to those reported previously for shallow donors and delocalized donor electrons attributed to residual Si_Ga impurities in β-Ga₂O₃ bulk substrates.¹⁹ The number of spins was determined by double integration of the EPR signals and a comparison with a spin-standard sample. Figure 2(b) shows that the total number of spins in our Si-doped, β-Ga₂O₃ samples scales approximately linearly with the epitaxial-layer thickness.

O–D spectra are shown in Fig. 3(a) for the set of Si-doped, β-Ga₂O₃ epitaxial layers of varying thicknesses that had been annealed in a D₂ ambient. An estimate of the concentration of OD–Si complexes could be made from the area of the absorbance lines with a strategy used in Ref. 23, Eq. (2), and a calibration factor called the “effective charge.”²⁷ An effective charge, of q = 0.33e per D atom, was determined for the $VGa1ib−2D$ defect in β-Ga₂O₃.²⁵ This value is typical for the effective charges determined for O–H centers in other semiconducting oxides, such as ZnO⁴² and In₂O₃,⁴³ and is used here. The area of the IR line for the deuterated epitaxial layer with thickness 3.6 µm shown in Fig. 3(a), for example, leads to a defect concentration of [OD-Si] = 2 × 10¹⁷ cm⁻³. This estimate is close to the concentration of Si_Ga in the epitaxial layer and is consistent with the majority of the Si_Ga donors in the epilayer having been deuterated by the annealing treatment in a D₂ ambient.

Figure 4 shows a plot of the integrated absorbance of the 2577 cm⁻¹ line vs the thickness of the Si-doped epitaxial layer. The 2577 cm⁻¹ absorbance line increases in strength as the thickness of the epitaxial layer increases. The 2546 cm⁻¹ line assigned to the $VGa1ib−2D$ complex decreases with the thickness of the epitaxial layer, presumably because of its competition with the 2577 cm⁻¹ center for D in the sample. The dependence of the intensity of the 2577 cm⁻¹ line on Si-doped layer thickness shown in Figs. 3(a) and 4 provides strong evidence for the assignment of the 2577 cm⁻¹ line to an OD–Si complex in β-Ga₂O₃.

To provide further evidence for this assignment, the Si-doped layer was lapped and polished away for the sample that was 3.6 µm thick. Figure 3(b) shows that the intensity of the 2577 cm⁻¹ line was reduced by a factor of 4 for the sample for which the Si-doped layer had been removed. The intensity of the 2546 cm⁻¹ line due to the V_Ga-2D complex in the bulk of the β-Ga₂O₃ was affected little when the Si-doped epilayer was removed from the sample. The results shown in Fig. 3(b) provide further evidence that the 2577 cm⁻¹ line is indeed due to an OD–Si complex in β-Ga₂O₃.

This conclusion that the OD–Si complex with vibrational frequency 2577 cm⁻¹ can be the dominant O-D center, even in a sample where SIMS shows that the Fe concentration exceeds the Si concentration by a factor of 40 [Fig. 1(c)], shows the difficulty of making firm assignments of vibrational lines based on impurity content alone. SIMS is sensitive to all the Fe containing defects in the material, not just isolated Fe on a Ga(2) site that would be present in an OD–Fe complex.

With the assignment of the 2577 cm⁻¹ line to an OD–Si complex being more firmly established, the assignment of the 2585.8 cm⁻¹ (5 K) line to an OD–Fe complex becomes more attractive. This assignment for the OD–Fe complex can be compared to the assignments for other OD-deep-acceptor complexes in the literature, 2586.3 cm⁻¹ for the OD–Mg complex³³ and 2582.9 cm⁻¹ for OD–Zn.³⁶ The vibration frequencies for the different O-D complexes with Ga(2)-site deep-acceptors lie within a few cm⁻¹ of each other.

The previous experimental results for the OD-impurity complexes provide clues about the structures of these defects. Samples that contained both H and D found no O–D line splitting for either the 2577 or 2585 cm⁻¹ centers, consistent with defects that contain a single D atom.³² Furthermore, the polarization properties of the 2577 and 2585 cm⁻¹ centers yield O–D transition moment directions that are consistent with defects that have a $VGa1ic−D$ complex at their core.³² 

A structure for the family of OD-impurity complexes suggested by the theory^24,32 that is consistent with the experiment is shown in the inset of Fig. 4,^44,45 where a $VGa1ic−D$ center is perturbed by the presence of a near-lying impurity. For the OD–Si complex, an Si atom is suggested to lie at one of the Ga(1) sites shown in the inset.³² For the OD-deep acceptor complexes, the Fe, Mg, and Zn impurities are suggested to lie at a distant Ga(2) site.³² The different locations of impurities that perturb $VGa1ic−D$ at the core of the defect, with an Si donor at a Ga(1) site and deep acceptors at a distant Ga(2) site, would explain the similar polarization properties observed for this family of defects, the ∼10 cm⁻¹ difference in vibrational frequencies for the Si donor and deep acceptor impurities, and the very similar vibrational frequencies observed for the complexes containing the different deep-acceptor impurities.

In conclusion, IR results for a set of Si-doped β-Ga₂O₃ epitaxial layers with different layer thicknesses that had been treated in deuterium provide new evidence for the assignment of a line at 2577 cm⁻¹ to an OD–Si complex. This assignment helps solidify the assignment of OD–impurity complexes and is consistent with the related structures (inset of Fig. 4) for this family of defects proposed recently by the theory in Ref. 32.

The work at LU was supported by the NSF Grant No. DMR 1901563. The work at UF was sponsored by the Department of the Defense, Defense Threat Reduction Agency, Grant No. HDTRA1-17-1-011, monitored by J. Calkins and DTRA Interaction of Ionizing Radiation with Matter University Research Alliance, Grant No. HDTRA1-19-S-0004 (Jacob Calkins). Portions of this research were conducted on research computing resources provided by LU supported by the NSF award 2019035. The work at the UAB was supported by the NSF Grant No. DMR-1904325. E. R. Glaser and M. J. Tadjer acknowledge the support from the Office of Naval Research. The authors thank Dr. F. Alema and Dr. A. Osinsky from Agnitron Technology for MOCVD growth and Dr. J. Blevins (AFRL, retired) for providing the Ga₂O₃ substrates for this study.