Polarized photoluminescence from Sn, Fe, and unintentionally doped β-Ga₂O₃

Gallium oxide (β-Ga₂O₃) is a material of interest in both electrical and optical applications due to its unique properties. It is a wide bandgap material with an indirect, fundamental bandgap near 276 nm (4.5 eV), though optical transition energies range from 282 to 253 nm (4.5–4.9 eV), depending on the crystallographic orientation with E||c < E||a < E||b.^1–3 Its beta phase, which has a monoclinic structure as seen in Fig. 1, is thermally and chemically stable and can be synthesized through inexpensive melt-growth methods, which allow for large, high-quality bulk substrates. Furthermore, β-Ga₂O₃ has a large Baliga's figure of merit estimated to be >3000 times that of Si.² That said, to control the material's electrical and optical properties, understanding and characterizing its defects is a priority.

One method to do this is through photoluminescence (PL). For β-Ga₂O₃, the PL spectrum is complex due to the multiple potential defects that could constitute recombination paths as well as vibrational broadening, electron-phonon coupling, etc. This causes deconvolving the spectrum and defect identification to be quite challenging.⁴ In this context, the literature generally describes the PL spectrum by separating it into regions. The highest energy region is the UV region (centered around 380–390 nm or 3.26–3.18 eV), which is generally described as being caused by the relaxation from the conduction band to self-trapped holes (STHs).^5–11 The next region is called the blue region (centered around 410–480 nm or 3.02–2.58 eV) and is believed to be caused by the recombination of donor-acceptor pairs or donor to STH.^5,9,11–14 In general, gallium and oxygen vacancies or interstitials are typically identified as the cause of the blue emission, though it has also been found that extended defects can cause blue emission in poor-quality samples.^4,15 The next region is the green region (around ∼500–600 nm or 2.48–2.07 eV), which is not as well understood as the UV and blue regions but is typically attributed to defects, recombination of donor-acceptor pairs caused by gallium and oxygen vacancies or interstitials using O_II or O_I to Ga_II, though dopants may also cause this.^9,11,16,17 Last, there is also a red region (within 650–850 nm or 1.91–1.46 eV) that is typically attributed to Cr³⁺, which is an unintentional dopant and comes from the iridium crucible during boule growth.^18–20 It is theorized that Cr³⁺ causes a red emission only when the sample is insulating.²¹ 

However, previous literature for β-Ga₂O₃ typically only discusses results for unpolarized and excitation-polarized PL. In this context, there have been a handful of prior works analyzing polarization properties of the PL emission. From our understanding, Yamaga et al. first measured polarized emission using unintentionally doped (UID) and Si-doped β-Ga₂O_3, whose results showed weak emission from b orientation.⁶ Later, Cho et al. characterized the polarized emission in (−201) bulk UID β-Ga₂O₃ and showed dominant emission from [102] as well as weak emission from b. However, this measurement was a transmission measurement, meaning the emitted PL spectra would transmit through the bulk of the sample, potentially affecting it.¹⁰ Here, we verify their results and explore the polarization of emission on a broader set of bulk crystals containing a wide set of dopants and orientations, demonstrating that the PL emission in β-Ga₂O₃ is polarization-dependent due to anisotropy of the crystal structure. As such, our results give a comprehensive insight into gallium oxide's PL spectral properties.

Several samples of bulk β-Ga₂O₃ substrates purchased from Novel Crystal Technology were used in this study. The samples include (−201) UID β-Ga₂O₃, (−201) Sn-doped β-Ga₂O₃, (010) Sn-doped β-Ga₂O₃, (−201) Fe-doped β-Ga₂O₃, and (010) Fe-doped β-Ga₂O₃. Details on the substrate properties can be found on the Novel Crystal Technology (NCT) website.²² Additionally, elemental analysis for Ga₂O₃ single crystal growths can be found in previous literature^19,23–28 and PL characterization of (−201) UID β-Ga₂O₃ was reported in our previous work.⁴ 

Photoluminescence was performed using ultrafast (fs) pulses from a wavelength-tunable (690–1040 nm or 1.8–1.2 eV) Ti:sapphire (Coherent Chameleon Vision Ultra) laser, which passed through a third-harmonic generator (Coherent Harmonics). The laser was then polarized using a linear polarizer (Glan-Laser alpha-BBO polarizer prism, 210–450 nm or 5.9–2.76 eV) followed by a zero-order half-wave plate to control the excitation polarization angle of the laser.⁴ The polarized laser then reflected off a beam splitter (Semrock 266 nm beam splitter) and through a lens followed by a pinhole hitting the sample and exciting it. This was followed by emission from the sample, which was focused back through the pinhole and lens and transmitted through the beam splitter into a polarizer (Thorlabs). The emission then went through another lens, focusing the emission into an optical fiber. The PL spectra were collected at room temperature using an optical fiber connected to a broadband spectrometer in the range of 300–800 nm or 4.13–1.55 eV (Avasoft AvaSpec dual-channel spectrometer).

Samples were measured using an excitation of 240 nm (5.16 eV) and 267 nm (4.64 eV). Excitation polarization was altered from 0° to 180° with a 15° step size. The emission polarization was measured from 0° to 180° with a 15° step size as well (or left out of the setup for unpolarized emission measurements). The collected data were corrected to remove the response caused by the spectrometer and eliminate the grating and detector response to extract the response of the sample itself. The data collected were corrected for the spectrometer spectra using the Ocean Insight HL-3 plus visible-near infrared (VIS-NIR) light source. The spectrum for this light source was calibrated by the Shanghai Calibration Laboratory. The calibrated blackbody radiant energy spectrum from the laboratory was divided by the spectrum collected by our spectrometer to get a correction factor. This correction factor was then applied to the spectra collected for all samples. All the corrected spectra were then normalized for the integration time.

Photoluminescence (PL) was first done on (−201) UID bulk β-Ga₂O₃. Initially, polarized excitation along b and [102] was done with total, unpolarized emission detected [Fig. 2(a)]. It was found that the total, unpolarized emission depends on the excitation's polarization. Furthermore, the maximum intensity between orientations changes depending on the excitation wavelength. This is because of the anisotropy of the absorption edge for β-Ga₂O₃, with E || b > E || [102] (energy-wise).^3,29 As such, the PL spectral intensity will depend on how the excitation is polarized to β-Ga₂O₃ orientation and how close this is to the absorption edge for that orientation. This is seen in Fig. 2(b), which was made by integrating the PL spectra of E || b and subtracting it from the integrated PL spectra of E || [102]. The difference is then divided by the sum of the two integrated orientations. It is seen that the PL spectrum has maximum intensity along E || b below 260 nm (4.77 eV) excitation. The maximum intensity flips to E || [102] when the excitation is above 260 nm (4.77 eV) excitation.

That said, when a polarizer was placed in front of the spectrometer to determine the orientation of the light emission with respect to the sample, emission polarization was observed to be always most intense along [102], no matter what excitation is used or excitation polarization is used. This is in agreement with the results seen in previous literature.^6,10 In fact, there are very little, if not any, PL spectra coming from b as seen in Figs. 2(c) and 2(d). Another interesting trait is that the overall PL spectral shape does not change, no matter what polarization (excitation or emission) is used. This points out likely to the same radiative-recombination paths occurring along different orientations.

While we are uncertain of the reason for this polarized emission dependence, we theorize that it is a phenomenon likely arising from the intrinsic crystal structure. The conduction band is made up of Gallium orbitals, while the valence band is made up of oxygen 2p orbitals. Self-trapped holes (STHs) are also connected to the oxygen atom in the crystal structure.^30–32 The radiative recombination of an electron from the conduction band to STH could be an electron moving from a gallium site (atom or vacancy) to an oxygen site (atom or vacancy).^10,33 The path the electron takes during recombination lies along [102], which could be along the a, c, or a combination of both. So, the lack of PL emission along b means the electron's path is not along that vector. Previous literature on STHs discus that the threefold O(I) and O(II) sites, which run along a and c, respectively, are more favorable for STHs. While the fourfold O(III) site along b is less favorable energy-wise for STHs.^30–32 This means that, energy-wise, the relaxation path along the O(I) and O(II) are more favorable than O(III),^30–32 or more favorable along a and c while less favorable along b.

(−201) Sn-doped bulk β-Ga₂O₃ has similar results to the (−201) UID bulk β-Ga₂O₃ sample. Figure 3(a) shows that the excitation polarization dependence is weaker for Sn-doped β-Ga₂O₃ than with the UID β-Ga₂O₃. Doping will cause a shift in the absorption edge energy along different orientations.^34,35 This is likely why, when excited at 240 nm (5.16 eV), the excitation polarization dependence is weaker due to the absorption edge for b-orientation redshifting. That is also why the excitation polarization dependence is stronger when excited at 267 nm (4.64 eV) in Fig. 3(c) Also, similar to the UID β-Ga₂O₃ sample, the excitation polarization dependence flips depending on excitation energy, with E || b more intense at lower wavelengths (higher energies) and E || [102] more intense at higher wavelengths (lower energies). This is again due to the location of the absorption edge for different orientations; it is just redshifted due to the Sn-doping.

Further similarities between UID β-Ga₂O₃ and Sn-doped β-Ga₂O₃ can be seen by looking at where the emission originates from. This is observed in Figs. 3(b) and 3(d) where strong emission along [102] occurs and there is a very weak emission along b. This arises at any excitation energy and any excitation polarization used, just like with UID β-Ga₂O₃.

Taking a closer look by plotting both the (−201) UID bulk β-Ga₂O₃ sample and (−201) Sn-doped bulk β-Ga₂O₃ together in Fig. 4 shows that the Sn-doped sample has a weaker emission compared to the UID sample. However, the actual shape of the overall PL emission does not change between the two samples. This is in agreement with the PL signature for pure and Si-doped Ga₂O₃ samples measured by Yamaga et al.⁶ This means that the doping did not generate any new radiative-recombination paths nor cause a change in the probability of the radiative-recombination paths. Instead, Sn appears to have quenched the overall radiative-recombination paths in the sample. Specifically, the ratio of emission along [102] between exciting along the E || [102] and E || b for UID is much larger than for the Sn-doped sample by a factor of around 2.

Looking at a different crystal orientation, Fig. 5 shows the PL emission results of a (010) Sn-doped β-Ga₂O₃ sample. Figure 5(a) shows the results for the total emission when the excitation at 240 nm (5.16 eV) is polarized. As seen, the PL spectrum does not change between excitation polarizations. This is again due to the Sn-doping, which causes a red shift in the absorption edge, allowing the emission along a to match c. To determine the orientation of the sample, transmission was done to find the bandgap and, therefore, sample orientation as seen in Fig. S1.³⁶ 

Interestingly, there is emission polarization dependence between the sample orientations, as seen in Fig. 5(b). c emits more strongly than a. Furthermore, in this sample, contrary to the (−201), we observe differences in PL spectra for different emission directions. As seen in the Fig. 5(b) inset, there is a blueshift when comparing the emission from a to the emission from c. This suggests that the radiative recombination that occurs along a generates a more significant fraction of higher energy photons. Furthermore, the low energy tail along a is less pronounced. That said, the overall PL shape remains very similar in shape with dominant emission in the UV, which is characteristic of samples with low density of structural defects.⁴ So, when comparing light being emitted along different polarizations, these differences in emission spectra point to radiative-recombination paths being different for emission between a and c.

Figure 5(c) compares a (−201) Sn-doped β-Ga₂O₃ sample with a (010) Sn-doped β-Ga₂O₃ sample. When comparing the overall intensity, we see that the emission along c > [102] > a > b, which is in the same order as the wavelength absorption edge for each orientation. These results match with previous literature on Si-doped β-Ga₂O₃ samples.⁶ 

Whenever a β-Ga₂O₃ sample is insulating, it causes a quenching of the UV, blue, and green PL emission. Instead, unintentionally doped Cr³⁺ which originates from the crucible during growth causes the presence of a red emission within the Ga₂O₃ sample. The Fe-doping pins the Fermi level causing Cr³⁺, which has orbital transitions that generate red emission when excited.^18,19,21 This is seen in Fig. 6 where the PL emission for Fe-doped β-Ga₂O₃ is plotted and shows a red emission.

Figures 6(a) and 6(c) are the plots of the polarized excitation and total, un-polarized emission at 240 nm (5.16 eV) and 267 nm (4.64 eV), respectively. They both show polarization dependence, which is dependent on excitation. Furthermore, the intensity of the emission generated by the polarized excitation is also dependent on the excitation energy. Polarization transmission measurements were performed to determine the sample orientation as seen in Fig. S2.³⁶ Exciting along [102] will generate a stronger emission at 240 nm (5.16 eV) excitation, while exciting along b will generate a stronger emission at 267 nm (4.64 eV) excitation. This is the opposite of what was observed for UID and Sn-doped β-Ga₂O_3, though the recombination process differs for Fe-doped β-Ga₂O₃. There might be a correlation between the emission from Cr³⁺ and its position within the crystal structure versus the emission from band transitions. For example, the Cr³⁺ is more likely to occur along orientations where emission from band transitions is less likely.

In Figs. 6(b) and 6(d) are the plots of the polarized emission at 240 nm (5.16 eV) and 267 nm (4.64 eV), respectively. As seen, there is weak polarization dependence from the emission with E || b generating a slightly stronger intensity. Again, this is the opposite of what was observed for UID and Sn-doped β-Ga₂O₃. Furthermore, there was no change within the PL shape when the emission plots were normalized, suggesting no change in radiative-recombination paths between orientations.

(010) Fe-doped β-Ga₂O₃ shows very different results. As seen in Figs. 7(a) and 7(c), which are the plots of the polarized excitation and total, un-polarized emission at 240 nm (5.16 eV) and 267 nm (4.64 eV), respectively, there is no or very weak polarization dependence due to excitation polarization. But, as seen in Figs. 7(b) and 7(d), which are the plots of the polarized emission at 240 nm (5.16 eV) and 267 nm (4.64 eV), respectively, there is polarization dependence depending on the emission's orientation from the sample. Again, polarization transmission measurements were performed to determine the sample orientation as seen in Fig. S3.³⁶ The emission from c has a dominant emission around 690 nm (1.8 eV). But that peak is quenched when looking at the emission from a. That said, it is worth mentioning that the loss of the 690 nm (1.8 eV) peak might not always occur. When a lamp and monochromator were used for excitation rather than a pulsed laser, the polarized emission between E_out || a and E_out|| c both showed a 690 nm (1.8 eV) peak. The causes behind this observation are currently unknown but are being investigated.

We characterized bulk, single crystal samples of (−201) and (010) β-Ga₂O₃ of various dopants using polarized excitation and looking at the polarization of the emission. UID and Sn-doped β-Ga₂O₃ both have a strong emission along c, which is slightly weaker along a and very weak along b. There is no change in the PL shape between orientations for (−201) UID and Sn-doped β-Ga₂O₃ samples. That said, (010) Sn-doped β-Ga₂O₃ shows a blueshift for the emission along a compared to c, suggesting a difference in radiative-recombination paths.

Fe-doped β-Ga₂O₃ exhibits a red emission instead of a UV, blue, or green emission. This red emission is caused by Cr³⁺, which is unintentionally doped during the boule growth. For the (−201) Fe-doped β-Ga₂O₃ sample, there is weak polarization dependence for the emission, with b generating a slightly stronger emission intensity. For (010) Fe-doped β-Ga₂O₃, under our experimental conditions, we observe a strong polarization dependence with a loss of the peak around 690 nm for the emission along a.

This material is based upon work supported by the Air Force Office of Scientific Research under Award No. FA9550-21-1-0507 (program manager: Dr. Ali Sayir). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the United States Air Force.

The authors have no conflicts to disclose

Jacqueline Cooke: Data curation (equal); Formal analysis (equal); Investigation (equal); Writing – original draft (equal); Writing – review & editing (equal). Minhan Lou: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal). Michael A. Scarpulla: Conceptualization (equal); Funding acquisition (equal); Methodology (equal); Project administration (equal); Supervision (equal); Visualization (equal); Writing – review & editing (equal). Berardi Sensale-Rodriguez: Conceptualization editing (equal); Funding acquisition editing (equal); Project administration editing (equal); Resources editing (equal); Supervision editing (equal); Visualization editing (equal); Writing – review & editing editing (equal).

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