Introduction

Gallium oxide is an ultra-wide bandgap semiconductor, with its most thermally and chemically stable phase being the monoclinic structure, β-Ga2O3. It has an indirect, fundamental bandgap near 276 nm (4.5 eV) though optical transition energies range from 282 to 253 nm (4.5 to 4.9 eV) depending on the crystallographic orientation due to the anisotropy of this material1. β-Ga2O3 has a large Baliga’s figure of merit which has increasingly garnered interest in various electronics and optoelectronics applications. Understanding and characterizing the material properties, including its defects, has been a priority, and photoluminescence (PL) has been under intense scrutiny in an attempt to define the mechanisms that generate the emissions in this material.

In general, PL spectra can be used to characterize the defects leading to radiative recombination processes within a specific material. In this regard, the PL spectra for β-Ga2O3 has generally been deconvoluted in three emission peaks: UV, blue, and green. However, it is notable that peak shapes from any point defect are expected to require more than a single Gaussian for a complex crystal structure like β-Ga2O3, possibly having an asymmetric shape and requiring a more complex model such as Huang-Rhys or Franck–Condon, which takes into consideration the vibrational broadening of the PL spectrum caused by electron–phonon coupling2,3,4. The sum of so many phonon replicas makes it challenging to fit the spectra uniquely. Adding energy level broadening from disorder or extended defects makes the deconvolution of the β-Ga2O3 PL spectra even more difficult. As such, in β-Ga2O3 with strong electron phonon coupling, there is little chance of being able to spectroscopically discern different types of defects from luminescence as any defect considered can result in a very wide luminescence band, even near 0 Kelvin. This has resulted in an intense debate in defining the defects and phenomenological explanations of electronic processes that cause these particular emissions5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24. Provided in the Supplementary Information is a comprehensive review of the literature.

Point defects have been the only explored and discussed potential source for the visible PL emission peaks in β-Ga2O3. Generally, there has been recent agreement on an intrinsic origin of UV luminescence and an extrinsic origin of visible emission. But, as far as can be determined, no previous works in the literature have discussed whether extended defects affect PL. Here, by systematically analyzing a series of thin-film samples, it is observed that extended defects substantially affect blue/green emissions and that this is correlated with poor structural quality samples. This analysis expands on previous reports that have so far attributed blue and green emissions to point defects and shows that in the case of thin films, extended defects may have a prominent role in emission properties.

Experiment

Sample preparation

All film samples were grown using metal–organic vapor-phase epitaxy (MOVPE) in an Agnitron Agilis reactor. All samples were grown using triethylgallium (TEGa) and O2 as the gallium and oxygen precursors, respectively, and silane as the Si dopant source when applicable. Growth details are listed in Table 1. Parameters used for all samples include a total molar flow of 15.53 μmol/min, argon flow rate of 1100 sccm, an oxygen flow rate of 500 sccm, and a chamber pressure of 15 Torr with a growth rate of around 6 nm/min. The substrates used for growth were either Fe-doped (-201) or (010) oriented β-Ga2O3 grown by Novel Crystal Technology. Otherwise, C-plane sapphire was used for growth and purchased from Cryscore. An unintentionally doped (UID) single crystal (-201) oriented β-Ga2O3 bulk sample grown using edge-defined film fed (EFG) technology was purchased from Novel Crystal Technology. Lastly, a single crystal (100) oriented 10% bulk aluminum-gallium oxide (AGO) sample, β-Al0.2Ga1.8O3 was grown using the Czochralski method. Details on the bulk AGO sample and growth can be found in25.

Table 1 Growth parameters for all the analyzed samples.
Full size table

Three series of samples were analyzed. The first series includes three film samples with varying Si-do** (used to make β-Ga2O3 more conductive and due to its strong PL signature): UID (~ 1016 cm−3), ~ 1017 cm−3, and ~ 1018 cm−3, with required silane flows determined using silicon secondary ion mass spectrometry (SIMS) calibrations26. All samples in the series are (010) oriented Si-doped β-Ga2O3 film grown on Fe-doped β-Ga2O3. The second series also contains three samples and compares (-201) β-Ga2O3. This series includes a bulk crystal of (-201) UID β-Ga2O3, a (-201) UID β-Ga2O3 film grown on Fe-doped β-Ga2O3, and a (-201) UID β-Ga2O3 film grown on sapphire. The last series compares (010) oriented (AlxGa1−x)2O3 films, i.e., AGO of varying Al concentrations, grown on Fe-doped β-Ga2O3. These are compared to a (100) oriented bulk 10% AGO (β-Al0.2Ga1.8O3) crystal. The second half of the series compares (-201) oriented 0%, 2%, 10%, and 28% AGO films grown on sapphire.

X-ray diffraction (XRD) was performed on all epitaxial AGO films. For AGO films grown on sapphire, the value obtained from XRD is used to define the aluminum concentration. Note that the AGO samples grown on Fe-doped β-Ga2O3 yielded Al composition values that were much larger than expected: 20% for the nominally 10% AGO sample (i.e., grown using a molar ratio that should have yielded ~ 10% Al) and 30% for the nominally 25% AGO sample (i.e., grown using a molar ratio that should have produced ~ 20% Al). This is likely due to strain and relaxation that occurs in the homoepitaxial film, which was found riddled with extended defects using STEM (shown in the results section). Furthermore, EDS measurements verified an aluminum concentration of 25% for the nominal 25% AGO sample. Therefore, for the 10% AGO sample, an Al composition value from the precursor molar ratio (PMR) during growth is used instead of the value obtained from XRD. The bulk AGO sample was measured by Washington State University using X-ray fluorescence (XRF) and found to have an Al composition of 10%25.

Characterization

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 polarization angle of the laser. The laser excited the sample at normal incidence within an integration sphere. 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 235 nm (5.27 eV) (236 nm or 5.25 eV for AGO grown on sapphire), 254 nm (4.88 eV), and 267 nm (4.64 eV). Polarization was altered from 0 to 180 degrees with a 15-degree step size. The collected data was corrected to remove the response caused by the spectrometer and eliminate the grating and detector response, so as to extract the response of the sample itself. The data collected was 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 Shanghai Calibration Laboratory. The calibrated blackbody radiant energy spectra from the laboratory was divided by the spectra collected by our spectrometer to get a correction factor. This correction factor was then applied to the spectra collected for all samples except AGO samples grown on sapphire. The AGO samples grown on sapphire used a quartz-tungsten-halogen (QTH) lamp (Oriel Instruments), whose spectrum is well defined as a blackbody source given by Plank’s radiation law. The measured light source was fitted to this radiation law, and a temperature correction factor was applied for the corrected data. These correction factors were then applied to measured PL data for the AGO samples grown on sapphire. All the corrected spectra were then normalized by the power measured (using a Newport optical power meter 1830-C) and integration time (which was kept at 5 s for all samples).

Results

Si series

Depicted in Fig. 1 is the measured polarized PL for all the samples in the series excited above and below the bandgap. Panels (a-c) correspond to the polarized emission spectrum of each sample: UID (~ 1016 cm−3) β-Ga2O3 grown on Fe-doped (010) oriented β-Ga2O3 substrate, Si-doped (1.5 × 1017 cm−3) β-Ga2O3 grown on Fe-doped (010) oriented β-Ga2O3 substrate, and Si-doped (5 × 1018 cm−3) β-Ga2O3 grown on Fe-doped (010) oriented β-Ga2O3 substrate. For the main plots in each panel, an excitation wavelength of 267 nm (4.64 eV) is employed, whereas the insets correspond to 235 nm (5.27 eV) excitation wavelength. All three samples in the Si series show a dominant UV emission, with peak emission around 385 nm (3.22 eV) for excitation both below (267 nm or 4.64 eV) and above (235 nm or 5.27 eV) bandgap. Polarization dependence of emission is seen below the bandgap for all three samples. The observed red emission starting around 700 nm (1.77 eV) is due to the Fe-doped β-Ga2O3 substrate on which the Si-doped β-Ga2O3 films were grown27,28.

Figure 1
figure 1

Polarized photoluminescence of β-Ga2O3 grown on (010) Fe-doped β-Ga2O3 excited at 267 nm (4.64 eV). Insets show data for the same samples but excited at 235 nm (5.27 eV). (a) UID (~ 1016 cm−3) β-Ga2O3 grown on (010) Fe-doped β-Ga2O3, (b) Si-doped (1.5 × 1017 cm−3) β-Ga2O3 grown on (010) Fe-doped β-Ga2O3, (c) Si-doped (5 × 1018 cm−3) β-Ga2O3 grown on (010) Fe-doped β-Ga2O3.

Full size image

The PL of the varying Si-doped samples grown on Fe-doped β-Ga2O3 is compared in Fig. 2, excited above and below the bandgap. This series of samples revealed that increasing Si do** leads to a decrease in overall PL intensity. This is likely due to an increase in non-radiative recombination7,13,29. The dominant emission within the UV region (385 nm or 3.22 eV) does not change between the samples. This is possibly due to the samples having a homogeneous film with limited extended defects. Transmission electron microscopy (TEM) images in previous reports show that β-Ga2O3 films grown on (010) β-Ga2O3 continue to have the same crystal structure as the substrate, with no stacking faults or other extended defects30,31. However, there is a decrease in the blue luminescence (from 400 to 500 nm or 3.1 to 2.48 eV) as the Si-do** in the samples increases. Other reports have observed and explained this observation as a decrease in donor–acceptor pairs, where the oxygen vacancy (VO) donors are reduced7,13,29,9 and conclusively determine what recombination processes are dominant within the different series of samples.