Photoluminescence spectroscopy of Cr³⁺ in β-Ga₂O₃ and (Al_0.1Ga_0.9)₂O₃

Alloying Ga₂O₃ with Al₂O₃ to create (Al_xGa_1−x)₂O₃ (AGO) is a promising way to increase the bandgap. For power electronics applications, bandgap widening could increase the breakdown field.^1–5 This alloy may also enable high electron mobility transistors for radio frequency (RF) power devices.^1,6

β-Ga₂O₃ is monoclinic (space group C2/m) while α-Al₂O₃ has the corundum structure (⁠$R3¯c$⁠).^1,3 β-Ga₂O₃ has a bandgap of 4.8 eV and α-Al₂O₃ has one of 8.8 eV.¹ In the monoclinic phase, there are two cation environments, the octahedral and tetrahedral gallium sites. For the corundum phase, only octahedrally coordinated cation sites are available.³ Thermodynamically, Al has an energetic preference for occupying octahedrally coordinated Ga(II) sites, although experimentally was shown to incorporate on the tetrahedral Ga(I) site up to 35%.^1,7,8 It was found theoretically and experimentally that the monoclinic phase is energetically preferable up to 70 atomic% of Al on the cation sites.^1,3

Cr³⁺ is a common unintentional dopant in β-Ga₂O_3.⁹ Cr³⁺ preferentially occupies the Ga(II) site and has the d³ electronic configuration.^10,11 Octahedral crystal field splitting results in multiple nonequivalent energy levels, which give rise to red luminescence.^12,13 The Cr³⁺ emission in β-Ga₂O₃ has potential applications in medical sciences, due to its long lasting near infrared afterglow. This is particularly useful for in vivo luminescent imaging for drug delivery or to monitor drug temperatures.^10,14–18

Cr³⁺ in β-Ga₂O₃ has defect absorption bands at 430 nm (2.88 eV) and 600 nm (2.07 eV), corresponding to ⁴A₂ →  ⁴T₁ and ⁴A₂ → ⁴T₂ transitions, respectively.^12,19–23 Both the absorption and emission spectra are similar to those of ruby (Al₂O₃:Cr³⁺),^12,20,21 which has R₁ and R₂ lines at 694 and 693 nm (1.787 and 1.789 eV).^12,24–26 The β-Ga₂O₃ Cr³⁺ emission spectrum consists of two sharp peaks at 689 nm (1.80 eV) and 696 nm (1.78 eV) with a broad red peak around 715 nm (∼1.7 eV). The ²E → ⁴A₂ and ⁴T₂ → ⁴A₂ internal Cr³⁺ transitions are responsible for the sharp and broad peaks, respectively.^{19,21–23,27–29} As compared to ruby, β-Ga₂O₃ has a larger R₁-R₂ splitting due to spin–orbit coupling and lower symmetry fields.^10,12 Changes in temperature may also lead to changes in crystal field strength and phonon frequencies.^10,20,27

Variable temperature photoluminescence (PL) spectroscopy of the R₁ and R₂ lines, and other related Cr³⁺ emissions, provides information about the thermal population, temperature, and pressure dependence of the energy levels. Changes in peak parameters (energy shift, peak height, and width) with temperature can also yield insight into the temperature and pressure response of the material,^30–32 as studied previously for ruby.^20,33–37 Similar studies have been performed for β-Ga₂O₃,^{10–12,14–16,19,21–23,27,29,38–42} but not for AGO. Additional emission features in the UV and blue have been identified in the x = 0.1 (Al_xGa_1−x)₂O₃ bulk crystal and were attributed to other types of defects such as self-trapped holes (STHs), other dopants, or extended defects.⁴³ 

In the present work, we measured the temperature dependence of the R₁ and R₂ lines in AGO and compared the results to β-Ga₂O₃. We find that the temperature-dependent ratio of the two peak intensities is dominated by nonradiative recombination effects and is significantly different for AGO versus β-Ga₂O₃.

Monoclinic single crystals were grown at the Washington State University Institute of Materials Research via the Czochralski (CZ) technique, similar to methods previously published.^7,44–48 Cr-doped Ga₂O₃ was grown with Cr concentration batched at 0.05 at.  % from a 3N7 (99.97%) purity Cr₂O₃ precursor and 5N (99.999%) purity Ga₂O₃. Glow discharge mass spectrometry (GDMS) indicated that the incorporation of Cr was 7 × 10¹⁹ atoms/cm³ (incorrectly stated in Ref. 48). High purity precursor powders, 5N (99.999%) purity Ga₂O₃, and 4N7 (99.997%) purity Al₂O₃, were batched and mixed with 10 mol. % Al₂O₃ in Ga₂O₃. From x-ray fluorescence (XRF), Al incorporation is 11.7 mol. % Al₂O₃ in β-Ga₂O₃ (x = 0.1).⁷ (100)-oriented single crystals were obtained from the CZ growths. GDMS indicated that the unintentional incorporation of Cr in AGO was 3 × 10¹⁶ atoms/cm³.⁷ Sample were cleaved to a thickness of 0.76 mm. β-Ga₂O₃:Cr had a dark green color due to heavy Cr doping, while the β-(Al_0.1Ga_0.9)₂O₃ crystals were colorless. We refer to these samples as β-Ga₂O₃ and AGO, respectively.

Photoluminescence (PL) and photoluminescence excitation (PLE) spectra were measured by a Horiba Jobin-Yvon FluoroLog-3 spectrometer with double-grating monochromators and a photomultiplier tube. The excitation source is a broadband 450 W xenon continuous-wave gas discharge lamp. Excitation wavelengths of 240–600 nm were selected by passing light through a monochromator. While PL spectra show emission for a specific excitation, the PLE measurement collects light emitted at a specified wavelength and plots the intensity of this emission versus excitation wavelength. Low-temperature spectroscopic measurements were carried out using a Janis continuous-flow liquid-helium cryostat with a Lake Shore 325 temperature controller. The light was incident at a ∼30° angle to the (100) plane of the crystal. All measurements except β-Ga₂O₃ at room temperature (298 K) were taken with the sample in a cryostat.

R-line positions, linewidths (FWHM), peak heights, and areas were obtained by fitting the measured PL spectrum to two gaussian peaks. The background was assumed to be a linear function of wavelength and was added, as an adjustable fitting parameter, to the gaussian fits. Due to alloy broadening, AGO required three gaussian peaks. The long wavelength tail of the R₁ line was significant at low temperatures and was subtracted from the R₂ line to improve the peak-area accuracy. Peak areas, positions, and full width half maximum (FWHM) occasionally needed manual adjustment for improved fitting. Fitting was performed in units of nanometers.

Room temperature PL and PLE spectra of β-Ga₂O₃ and AGO are shown in Fig. 1. The R-lines are similar, but for AGO the R-lines exhibit inhomogeneous broadening due to the addition of Al. Generally, the Cr³⁺ emissions for AGO were weaker than those of β-Ga₂O₃. To facilitate a comparison, β-Ga₂O₃ with a relatively weak emission intensity, comparable to that of the AGO sample, was selected. Cr³⁺ spectra from multiple AGO samples are presented in Fig. S2 (see the supplementary material⁴⁹ for additional spectra and fitting results).

The Cr³⁺ emission is strongest when excited by energies near the band edge but is also strong when excited with energy within the absorption bands [Fig. 1(a)]. This is further illustrated by examining the PL and PLE simultaneously with a 2D PL-PLE contour map (Fig. 2). In these maps, the emission intensity is represented as a false color. The contour maps show the absorption bands and that the emission peak energy does not change with excitation energy for either material. β-Ga₂O₃ has sharper peaks than AGO for all excitation energies.

A blue shift in peak energies is observed with decreasing temperature (Fig. 3). The blue shift is explained by interactions between Cr³⁺ electronic levels and lattice vibrations.²⁶ At lower temperatures, new peaks emerge, at 709, 727, 735, and 746 nm for AGO and similar wavelengths for β-Ga₂O₃.

For both materials, the lower-energy R₁ line becomes stronger at lower temperatures, indicative of a thermal depopulation of the higher energy ²E excited state component (which results in the R₂ line).^11,19 AGO peaks were broader at all temperatures (Fig. 4).

The trend with the line positions as seen in Fig. 5 generally agrees with the trend seen previously in the literature,^15,19 with some minor differences (Figs. S3–S5 in the supplementary material).⁴⁹ Fittings for a select few temperatures are shown in Fig. 5; fittings at all temperatures are reported in the supplementary material⁴⁹ (Figs. S6–S11).

Peak fitting was performed for the R lines for temperatures from 20 to 295 K (Fig. 5). It was found that for AGO a third peak between the R₁ and R₂ lines improved the quality of the fit. This peak is denoted the “middle peak” and was smaller than the R-lines. The area of the middle peak tracks closely with the R₁, as a function of temperature (Fig. 6). We, therefore, attribute this speak to a specific center [for example, Cr³⁺ with one neighboring Al on a Ga(I) site]. Additional centers contribute to the inhomogeneous broadening of AGO as compared to β-Ga₂O₃.

For AGO at low temperatures, the R₂ peak overlapped with the short-wavelength tail of the R₁ peak. The R₂ peak area at 20 K was used as a correction factor that was scaled, based on relative intensity of R₁, and subtracted from all other spectra,

This correction was minor and predominantly impacted results at low temperatures.

The R₁ and R₂ peak intensities do not follow a simple Arrhenius behavior. Instead, they are strongly affected by nonradiative recombination. In general, the electron recombination rate is the sum of the radiative and nonradiative recombination rates,

Radiative recombination, which is measured in our PL experiments, is assumed to be temperature-independent.³¹ The nonradiative recombination frequency is taken to be thermally activated and is represented by an Arrhenius function. With these assumptions, the total recombination rate is given by

The measured intensity (peak area) of the R₁ line is proportional to the fraction of all decays that are radiative,

where E₁ is the activation energy of the nonradiative channel and k is the Boltzmann constant.³¹ The R₂ line intensity is described by a similar function.

The ratio of the intensity of R₂ to the intensity of R₁ is given by

where the e^−E/kT term represents the thermal population; E₁ and E₂ are the nonradiative recombination activation energies for R₁ and R₂, respectively. Taking a ratio highlights the difference between the two lines and eliminates overall changes in intensity. The experimental intensity ratio is plotted against temperature (Fig. 7) fit using least-squares fitting to Eq. (5). The fit parameters are the activation energies (E₁ and E₂) and the scaling factors (A, A₁, A₂). E is held constant and is equal to the R₁-R₂ splitting (19.82 meV), similar to Tokida and Adachi.¹⁹ Table I shows the fit parameters.

Dotted lines in Fig. 7 represent the thermal population fit only, using only the Arrhenius Ae^−E/kT term. The trends of these lines demonstrate that nonradiative recombination dominates the peak intensities for temperatures >50 K. The curves approach zero as temperature approaches zero, due to the depopulation of the higher energy state of the ²E level which is the upper level for R₂ emissions. The fitting parameters (Table I) show significant differences for the two materials. Nonradiative recombination in β-Ga₂O₃ is thermally activated. At low temperatures, as the nonradiative effect goes to zero, the R₁ line becomes very strong [Fig. 6(b)]. In AGO, however, the nonradiative recombination for the R₁ line is temperature independent. This means that the R₁ line in AGO does not experience the large intensity increase at low temperatures [Fig. 6(a)] that β-Ga₂O₃ experiences.

In β-Ga₂O₃ and AGO, the Cr³⁺ R₁ and R₂ lines were observed at 696 nm (1.78 eV) and 690 nm (1.80 eV), respectively, along with a broad emission band at 709 nm (1.75 eV). The R-lines for AGO show inhomogeneous broadening, especially the R₁ line, which had features on the short-wavelength side of the main peak. All peaks show a redshift with increasing temperature. At low temperatures, the R₁ line is dominant due to thermal depopulation of the R₂ line. At temperatures above ∼50 K, the R₂/R₁ intensity ratio is strongly affected by nonradiative recombination. In β-Ga₂O₃, the nonradiative contribution is thermally activated, which results in strong emission at low temperatures. AGO, in contrast, has smaller thermal activation barriers, perhaps due to the randomness in the alloy, providing a large number of nonradiative recombination pathways.

The authors gratefully acknowledge fruitful collaborations with Mike Scarpulla from the University of Utah. This work was supported by the Air Force Office of Scientific Research under Award No. FA9550-21-1-0078 monitored by Ali Sayir. Additional support was provided by the U.S. Department of Energy/National Nuclear Security Administration under Award No. DE-NA0003957 (L.M.B.) and U.S. Department of Energy, Office of Basic Energy Sciences, Division of Materials Science and Engineering under Award No. DE-FG02-07ER46386 (M.D.M.)

The authors have no conflicts to disclose.

Cassandra Remple: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Writing – original draft (equal); Writing – review & editing (equal). Lauren M. Barmore: Formal analysis (equal); Writing – review & editing (supporting). Jani Jesenovec: Investigation (equal); Writing – review & editing (equal). John S. McCloy: Project administration (equal); Writing – review & editing (equal). Matthew D. McCluskey: Conceptualization (equal); Writing – original draft (equal); Writing – review & editing (equal).

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