Broad luminescence from Zn acceptors in Zn doped β-Ga₂O₃

The ultra-wide bandgap of Ga₂O₃ has sparked great interest with prospects for applications within power electronics, optoelectronics, and gas sensors. Ga₂O₃ has a high breakdown voltage (∼8 MV/cm) and a large Baliga’s figure of merit, which make the material particularly promising for power electronic applications.¹ In contrast to other emerging materials for power electronics, such as SiC and GaN, Ga₂O₃ can be grown from melt and is, therefore, potentially a low-cost alternative.² Ga₂O₃ can crystallize in several different polymorphs, where β-Ga₂O₃ is the most stable with a monoclinic crystal structure and an optical bandgap of 4.7–4.9 eV depending on the crystallographic orientation.^1,3,4

Many common impurities in β-Ga₂O₃ cause an unintentional donor doping, which, in most cases, results in β-Ga₂O₃ displaying an n-type conductivity.⁵ Holes, however, tend to form small polarons due to the low valence band dispersion of β-Ga₂O₃, and combined with the fact that acceptor levels are generally deep, p-type doping in β-Ga₂O₃ is challenging.^6–8 However, acceptor doping can be used to compensate n-type to make semi-insulating layers in Ga₂O₃-based devices.⁹ Commonly used acceptors are Mg and Fe.^7,10 Zn has also been proposed as an acceptor dopant and could be more convenient as it can be introduced to β-Ga₂O₃ either into the melt or during post-growth in-diffusion.^11–15 Furthermore, H can also exist as an unintentional impurity in β-Ga₂O₃ and can act as a shallow donor or fully or partly passivate acceptors, including Zn_Ga.^5,16

Gustafson et al. used electron paramagnetic resonance to identify the $ZnGa0$ acceptor in β-Ga₂O₃ and obtained an activation energy of 0.65 eV for the thermal decay of its signal.¹³ A similar activation energy of 0.77 eV was found by Chikoidze et al. from Hall effect measurements on Zn-doped materials.¹⁷ Hybrid functional calculations show that Zn_Ga is a deep polaronic acceptor displaying thermodynamic (0/−) and (+/0) levels 1.3 and 0.9 eV above the valence band maximum, respectively, where the 0 and 1+ charge states correspond to one and two trapped holes.^11,15,18 The optical transitions between the CBM and the charge-state transition levels of Zn_Ga are predicted to give rise to broad emission in the visible part of the photoluminescence (PL) spectrum.¹⁵ 

Zn can be incorporated during growth or in-diffused, e.g., using Zn vapor as a source.^13,14 In Ref. 14, Zn was diffused in via vapor phase, where secondary ion mass spectrometry (SIMS) revealed a large concentration of in-diffused Zn in a box like profile. The diffusion was explained through an interstitialcy-mediated mechanism, where Zn is mobile in a split-interstitial configuration with Ga. The model suggested that Zn can get trapped in an acceptor configuration Zn_Ga and that a second Zn can get trapped on the same Ga site creating the complex Zn_GaZn_i. Zn_GaZn_i is expected to act as a single shallow donor, and the second Zn is trapped with a binding energy of 1.37 eV.¹⁵ Furthermore, since the deep charge-state transition levels are removed when Zn_Ga is complexed with Zn_i, no broad emission is expected from the complex.

In this study, we combine PL spectroscopy and SIMS measurements of β-Ga₂O₃ samples heat-treated at varying temperatures and ambients to monitor the defect transformation between Zn_GaZn_i and Zn_Ga in an attempt to identify the signature of Zn_Ga. The observed PL features are discussed considering theoretically predicted emission sources and compared with the first-principles calculations of luminescent transitions for the relevant defects.

The β-Ga₂O₃ single crystalline samples used for this experiment were single-side polished and grown with a (001) surface orientation. The samples consisted of an 8 μm layer of β-Ga₂O₃ grown by halide vapor-phase epitaxy (HVPE) on a Sn-doped β-Ga₂O₃ substrate grown by the edge-defined film-fed growth method (Tamura Corporation, Japan). Four 5 × 5 mm² samples were put in sealed evacuated quartz ampules together with ∼0.1 g pieces of 99% pure Zn. To make sure that the two materials were kept separate, the quartz ampules were narrowed in the middle. The ampules were pumped with a rough vacuum before being sealed. Four ampules with different samples were all heated to 1100 °C for 1 h followed by SIMS and PL measurements. The four samples went through three subsequent steps of heat treatments at 300–400–500 °C, 400–700–1000 °C, 500–800–1100 °C, and 600–900–1200 °C for 1 h in O-flow. In addition, one separate sample received only the 1100 °C Zn vapor anneal followed by a 1100 °C oxygen anneal for comparison and reproducibility. In between the heat treatments, we performed SIMS and PL measurements.

SIMS measurements were conducted using a Cameca IMS7f, with a 10 keV $O2+$ primary beam raster scanned in an area of 150 × 150 μm², to obtain concentration vs depth profiles of Zn. The craters were measured using a Dektak 8 stylus profilometer to calibrate the depth, where a constant sputter rate was assumed. ⁶⁴Zn implanted β-Ga₂O₃ reference samples were used to calibrate the concentration.

PL spectroscopy measurements were performed at 10 K by employing a closed-cycle He refrigerator system (CCS-450 Janis Research, Inc.). The photo-excitation at 246 nm wavelength (5.04 eV) and 10 mW average power was provided by a third harmonic of a pulsed Ti:sapphire laser operating at 80 MHz in a femtosecond mode-locked regime (Spectra-Physics, Tsunami HP and GWU-UHG-23). The PL emission was collected using a microscope and analyzed using a fiber-optic spectrometer (Avantes, AvaSpec-Mini3648-UVI25) covering the wavelength range 200–1100 nm.

First-principles calculations were carried out using the generalized Kohn–Sham theory with projector-augmented wave potentials,^19,20 as implemented in the VASP code.²¹ The Ga 3d electrons were included explicitly as valence electrons. The hybrid functional of Heyd, Scuseria, and Ernzerhof (HSE)^22,23 was used with the fraction of screened Hartree–Fock exchange set to 33%. This results in a direct bandgap value of 4.9 eV and lattice parameters a = 12.23, b = 3.03, c = 5.79 Å, and β = 103.8°, which are in good agreement with experimental data.^24,25

Defect calculations were performed using 160-atom supercells, relaxing the ions until forces were smaller than 10 meV/Å. The plane wave energy cutoff was set to 400 eV, and a special k-point at (0.25, 0.25, 0.25) was used. Total-energy corrections due to the finite size of the charged supercell were included by using the schemes described in Refs. 26–28. The defect formation energies were calculated using the well-established formalism²⁹ (see Ref. 14 for details pertaining to the present calculations).

Defect luminescence lines were calculated using the methodology detailed in Ref. 30, where the vibronic transition is described by a one-dimensional configuration coordinate (CC) diagram with single effective vibrational modes for the excited and ground states.³¹ The CC model parameters include the zero-phonon line (ZPL) energy E_ZPL, the change in configuration coordinate ΔQ, and the Franck–Condon (FC) relaxation energy in the excited (e) and ground (g) states $deFC$ and $dgFC$⁠, respectively. In this model, the classical optical emission (absorption) energy is defined as $Eem=EZPL−dgFC$ $(Eabs=EZPL+deFC)$⁠. Full luminescence lines are obtained by calculating the FC overlap integral between the vibrational levels of the excited and ground states.^30,32 This model is appropriate for defects exhibiting strong electron–phonon coupling,³⁰ which is the case for all defects studied here.

Zn concentration profiles as a function of depth for a sample heat-treated in an ampule with Zn vapor at 1100 °C for 1 h and later heat-treated at 500 and 800 °C for 1 h in O-flow is shown in Fig. 1. The profile measured after heat treatment in an ampule shows in-diffused Zn with a steady concentration leveled at $∼1019$ cm⁻³ that drops abruptly at around 2 μm depth below the surface, in accordance with Ref. 14. We see similar features for the two other profiles from heat treatment in O-flow; however, the profiles have slightly different shapes. Note that all samples in this experiment showed slightly different concentration profiles after the same heat treatment in ampules with Zn, which can at least partly be attributed to variations in the distribution of Zn on the surfaces as well as increased roughness after the heat treatment in Zn vapor. Note also that the sample was conductive after this initial treatment, in accordance with previous observations.¹⁴ The sample in Fig. 1 was later heat-treated at 1100 °C for 1 h in O-flow (not shown), where Zn depth profiling with SIMS was not possible due to charging effects, indicating a highly resistive sample. The high resistivity after annealing in oxygen flow observed with SIMS is consistent with the formation of acceptor-like compensating defects. In all samples, heat treatment above 1000 °C in O-flow led to insulating samples. As-grown β-Ga₂O₃ samples heat-treated at 1200 °C in O-flow did not show signs of increased resistivity. However, we cannot conclude that the decrease in conductivity is caused by the formation of Zn_Ga.

Further insights into the origins of the defects as well as defect transformations during heat treatments were attained by PL spectroscopy measurements following each of the processing steps. Figure 2(a) presents the PL spectra obtained at 10 K from β-Ga₂O₃ heat-treated in Zn vapor at 1100 °C and then consecutively annealed at 500, 800, and 1100 °C in O-flow. The spectrum from an as-grown material is also presented, demonstrating the typical luminescence of β-Ga₂O₃ with a broad peak in the UV region centered at $∼3.3$ eV ascribed to the recombination of free electrons and self-trapped holes (STHs).^8,15,34,35 The initial heat-treatment in Zn vapor at 1100 °C generally quenches luminescence intensity (almost twofold), while the spectral features remain similar to those of the as-grown material. Such a quenched PL efficiency points toward the formation of new competing non-radiative recombination pathways. Meanwhile, upon subsequent heat-treatments in oxygen ambient, a broad PL band on the lower-energy side gradually builds up with the increase in annealing temperature and ultimately becomes the dominating feature in the luminescence spectrum.

Figure 2(b) presents a spectral analysis for a representative case of high-temperature treatment based on Gaussian deconvolution using the nonlinear least square fitting method. The key emission components revealed by Gaussian deconvolution [the dashed curves in Fig. 2(b)] are centered at 3.3, 2.9, 2.5, and 2.1 eV. The 2.5 eV emission has an apparent full width half maximum (FWHM) of 0.53 eV, while the FWHM of the corresponding Gaussian component is 0.41 eV.

The evolution of the 2.5 eV emission as a function of annealing temperature is outlined in the inset of Fig. 2(b), where the 2.5 eV intensity is normalized with respect to the total integrated PL at each heat-treatment step. Note the rapid dynamics of defect transformation during the heat treatment with the onset-to-saturation of the 2.5 eV luminescence occurring within the narrow temperature interval of 600–900 °C. The 2.5 eV luminescence overlaps with the so-called green luminescence that has previously been reported.³⁶ A broad green emission is also observed in ZnO, which could have been introduced in the form of precipitates after heat treatment in Zn. In that case, the emission would be accompanied by a sharp donor–acceptor pair (DAP) line at 3.2 eV.³⁷ There is, indeed, a sharp peak at this energy in the PL spectrum. However, this feature is also present in the as-grown sample, making a ZnO origin unlikely.

The emergence of the 2.5 eV emission and the changes in the surface conductivity could be explained by the same defects that we considered previously to model Zn diffusion:¹⁴ (i) annealing in Zn vapor leads predominantly to the formation of Zn_GaZn_i and (ii) subsequent anneals in O-flow lead to the dissociation of the complex, leaving behind compensating Zn_Ga acceptors that can be observed by PL. To further investigate this possibility, we have performed hybrid functional calculations.

A pertinent question is how stable the Zn_GaZn_i complex is to increasing temperature. Figure 3 shows the formation energy as a function of Fermi level for the relevant defects. A binding energy of 1.4 eV can be obtained from the difference in formation energy between Zn_GaZn_i and the sum of the formation energies of Zn_Ga and Zn_i. To estimate the dissociation energy (the activation barrier that must be surmounted in order for the complex to break up), we add the 0.7 eV migration barrier of Zn_i to the binding energy, resulting in 2.1 eV.¹⁴ Based on harmonic transition-state theory, this dissociation energy is consistent with the observation of the 2.5 eV emission appearing above 600 °C.

Upon illumination, Zn_Ga can capture one or two photogenerated holes. The illustrations of the resulting polaronic states for Zn_Ga2 are presented in Fig. 4(a). Subsequent radiative recombination with an electron at the CBM is described by the calculated CC diagrams shown in Fig. 4(b). The emissions from the corresponding transitions $ZnGa10/−$⁠, $ZnGa1+/0$⁠, $ZnGa20/−$⁠, and $ZnGa2+/0$ and from the self-trapped hole transitions for O1 and O2 localized holes¹⁵ are shown in Fig. 4(c). The CC model parameters, peak position (PP), and FWHM for all calculated luminescence lines are provided in Table I.

The calculated STH emission has peak positions at 3.1 and 3.0 eV for holes trapped at O1 and O2 sites, respectively. The calculated values for the peak position are thus ∼0.2 eV below the experimentally observed emission peak. Under n-type conditions, Zn_Ga is in the 1− charge state, and luminescence from the (0/−) transition is, therefore, expected. The emission from the (0/−) transition has a peak position at 2.1 and 2.3 eV for Zn_Ga1 and Zn_Ga2, respectively, while the emission from the (+/0) transition has a calculated peak position at 2.7 and 2.6 eV for Zn_Ga1 and Zn_Ga2, respectively. To observe luminescence from the (+/0) transition, Zn_Ga must capture two photogenerated holes, which might be possible depending on, for example, the excitation intensity and hole-capture cross section.³⁸ However, we consider contributions from the (+/0) transition to be unlikely compared to the (0/−) transition.

Assuming that we can expect the same shift of 0.2 eV in peak position for these defect states as observed for the STH emission, the calculated PP for the $ZnGa20/−$ emission (2.3 eV) is close to the observed 2.5 eV emission. This is also apparent from the consistent redshift of the calculated optical transitions (indicated by vertical marker lines in Fig. 2(a) with regard to other deconvoluted PL components. It is interesting to note, however, that the calculated PP for $ZnGa10/−$ (2.1 eV) might correspond to the observed shoulder on the lower side of the experimentally observed emission in Fig. 2. Zn_Ga2 has only a slightly lower formation energy compared to Zn_Ga1, and thus, both configurations are expected to occur. Indeed, Gustafson et al. reported that the concentration of Zn_Ga2 is slightly higher than that of Zn_Ga1 in melt-grown β-Ga₂O₃ crystals.¹³ We speculate that the asymmetry in the 2.5 eV peak comes from a difference in the emission energy from Zn_Ga1 and Zn_Ga2. It should be noted, however, that the ratio between Zn on the two Ga sites can depend on the growth method due to kinetic limitations.³⁹ 

Complexing Zn_Ga with a Zn_i double donor results in a shallow single donor, as shown in Fig. 4(a), and a hole polaron can no longer be captured at the defect. Therefore, we do not expect to observe luminescence from the Zn_GaZn_i complex. We also considered Zn_GaH as a possible source of the 2.5 eV luminescence. As a single donor, H will not fully passivate Zn_Ga, and optical transitions involving the (+/0) transition of the Zn_GaH complex are possible (see Table I). However, we expect the concentration of Zn to be much higher than H in the studied samples.

It is reasonable to assume that PL experiments, in fact, validate all theoretically predicted transitions once offset is taken into consideration. In particular, Zn_Ga appears as a prime candidate for the observed green luminescence, and consequently, our findings while monitoring the formation and dissociation of Zn_Ga related defects during heat treatments strongly support the hypothesis of the two configurations of Zn occurring during in-diffusion of Zn.

In summary, Zn-doped β-Ga₂O₃ was investigated using PL measurements and first-principles calculations. For β-Ga₂O₃ samples heat-treated in Zn vapor, the PL spectra showed features similar to those of the as-grown samples. By combining results from previous work and theoretical calculations, the peak emerging at 2.5 eV for samples subsequently heat-treated in O-flow was ascribed to the formation of Zn_Ga. Furthermore, the samples turn insulating above temperature 1000 °C, indicating that the samples were dominated by the Zn_Ga acceptors. The results support the interstitialcy trap-limited diffusion mechanism for Zn in β-Ga₂O₃ suggested in previous work.¹⁴ 

The Research Council of Norway is acknowledged for the support to the Norwegian Micro- and Nano-Fabrication Facility, NorFab, Project No. 295864, and GO-POW, Project No. 314017. The computations were performed on resources provided by UNINETT Sigma2, the national infrastructure for high performance computing and data storage in Norway.

The authors have no conflicts to disclose.

Ylva K. Hommedal: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Software (equal); Validation (equal); Visualization (equal); Writing – original draft (lead). Ymir K. Frodason: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Project administration (equal); Software (equal); Supervision (equal); Validation (equal); Visualization (equal); Writing – review & editing (equal). Augustinas Galeckas: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Software (equal); Validation (equal); Visualization (equal); Writing – review & editing (equal). Lasse Vines: Conceptualization (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Project administration (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal). Klaus Magnus H. Johansen: Conceptualization (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Project administration (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.