Wide range tunable bandgap and composition β-phase (AlGa)₂O₃ thin film by thermal annealing

III-oxide semiconductors such as Ga₂O₃ have attracted increasing attention recently because of the wide bandgap, excellent stability, and availability of bulk substrates for high-performance deep-ultraviolet (DUV) solar-blind photodetectors (PDs), transparent electronics, and power electronics.^1–15 The monoclinic β-phase is widely recognized as the most stable Ga₂O₃ crystalline structure,¹⁶ amid bulk β-Ga₂O₃ crystal growth^17–19 and thin film deposition.^20–26 For further development of III-oxide optical and power devices, the progress would benefit greatly from the alloy formation, tunability of the alloy composition, and, thereby, the associated properties such as bandgap, absorption spectrum, and breakdown field.²⁷ In particular, (Al_xGa_1−x)₂O₃ alloys are promising candidates for shorter-wavelength and higher-power devices due to large tunable bandgaps from about 4.9 (β-Ga₂O₃) to above 8.8 eV (α-Al₂O₃).

To form the (Al_xGa_1−x)₂O₃ alloys, researchers have employed RF sputtering,²⁸ PLD,^29,30 LPCVD,³¹ mist-CVD,^32,33 laser MBE,^34,35 MOCVD,^36–39 and MBE.⁴⁰ These “direct-growth” techniques offer the tunability of material composition and, thus, bandgap. Since most of those studies concern the β phase, the bandgap and composition ranges of the β-(Al_xGa_1−x)₂O₃ alloys are summarized in Table I. For instance, the bandgap range reported by Shi et al. for the β-(AlGa)₂O₃ alloys is 6.10 − 6.22 eV by LPCVD.³¹ Zhang et al. have demonstrated β-(AlGa)₂O₃ films deposited by PLD with the wide-range bandgap tunability up to 6.1 eV, above which the X-ray diffraction (XRD) 2θ (−201) and higher-order peaks were almost invisible.³⁰ Overall, the associated costs of the direct-growth techniques are not trivial because of various requirements: targets with different compositions, high-purity precursors/sources, and demanding environments such as an ultrahigh vacuum.

In the 1990s, Fleischer and Battiston et al. discovered that the thermal annealing of Ga₂O₃ films on sapphire could lead to Al diffusion into the films.^41,42 Later, Kokubun et al. found that increased temperatures from 600 to 1200 °C during thermal treatment of Ga₂O₃ sol–gel can boost the polycrystalline Ga₂O₃ bandgap from 4.95 to 5.53 eV which the authors attributed to the Al diffusion.²⁰ Recently, Goyal et al. varied annealing temperatures of polycrystalline Ga₂O₃ films on sapphire from 600 to 1000 °C to obtain a changing bandgap from 4.63 to 5.15 eV with an annealing time of 36 h.⁴³ They performed secondary ion mass spectrometry (SIMS) analysis, which shows Al diffusion into the films, though being largely inhomogeneous. Those studies hint at a viable pathway of forming (AlGa)₂O₃ ternary alloys based on Ga₂O₃/sapphire templates by thermal diffusion, which is low cost and straightforward as opposed to the direct-growth techniques. However, the Al compositions were still low and inhomogeneous in those studies, which greatly hindered the application prospects. Besides, the III-oxide/sapphire interface has not been systematically examined, preventing a good understanding of the diffusion process amid the thermal annealing.

In this work, we report on a wide range bandgap and composition tuning technique of using common β-Ga₂O₃/sapphire templates to obtain ternary β-(AlGa)₂O₃ alloys by thermal annealing at various elevated temperatures. We found that controlling annealing temperature from 1000 to 1500 °C can tune the diffusion rate, leading to a large optical bandgap change from 4.88 to 6.38 eV for β-(Al_xGa_1−x)₂O₃. The corresponding material bandgap and film thickness were characterized systematically by multiple techniques. Besides, the thin-film/sapphire interface was investigated by high-resolution transmission electron microscopy (HR-TEM) to understand the interdiffusion process.

The 50-nm thick β-Ga₂O₃ thin-films were deposited on c-plane sapphire substrates by the PLD. Although the PLD was utilized, it is important to note that the same studies can be performed for the binary β-Ga₂O₃ films on sapphire grown by other techniques such as CVD, MBE, and MOCVD because we expect that the thermal diffusion could occur regardless of how the β-Ga₂O₃/sapphire templates are prepared. The PLD conditions include a heater temperature of 800 °C, a chamber pressure of 4.5 mTorr, and a KrF (248 nm) excimer laser power of 400 mJ with a repetition rate of 5 Hz. An undoped Ga₂O₃ target of 99.99% purity provided by Sigma-Aldrich was employed. Thermal annealing was conducted in an MTI KSL-1700X-A4-DC furnace at different temperatures of 1000 − 1500 °C with the step of 100 °C in air at atmospheric pressure for three hours. The ramp rate for the annealing process is +4 °C/min below 800 °C and increased to +5 °C/min (above 800 °C); the cooling rate is –5 °C/min down to 200 °C. The crystal structures, material quality, material composition, optical properties, thin film thickness, and interface were characterized by using XRD, SIMS, optical UV-Vis transmission, and TEM. The XRD 2θ-ω scans were carried out using a Bruker D2 PHASER system with a wavelength of λ = 1.5406 Å. The SIMS experiments for elemental depth profiling were performed using a Dynamic SIMS system from Hiden Analytical. The optical transmission was measured using a Shimadzu UV-3600 spectrophotometer. The TEM specimens were obtained using a Helios focused ion beam (FIB) system. The cross-sectional TEM images and EDX mapping were acquired using an FEI Titan ST system with a Gatan EDX module.

Figure 1 shows the XRD 2θ-ω scan spectra of the as-deposited and annealed Ga₂O₃ thin films on sapphire with different annealing temperatures. For the as-deposited Ga₂O₃ sample, there are three dominating peaks from (−201) and higher-order diffraction planes of β-Ga₂O₃ film. The samples annealed at 1000 and 1100 °C manifest only the main (−201) and higher order (−402) and (−603) peaks for the β phase. When the temperature increases to 1200–1400 °C, the (−402) peak splits into two peaks: one remains the (−402) peak and the other might be the (401) peak. For the 1400 °C annealed sample, several extra peaks such as (002) and (400) and the higher order peaks were observed, which could be attributed to the lattice rotation or tilting due to thermal strain induced by high temperature annealing.⁴⁴ However, for the 1500 °C sample, those peaks disappeared, indicating good crystallinity. Figure S1 in the supplementary material shows XRD phi-scan of the (−401) plane of the as-deposited β-Ga₂O₃ sample and three β-(AlGa)₂O₃ samples annealed at 1000, 1200, and 1400 °C. All exhibit sixfold symmetry diffraction peaks, which demonstrate the single β phase.

Table II summarizes the positions of the (−201) and higher-order (−402) and (−603) peaks in Fig. 1. They shift toward higher angles as the annealing temperature increases from 1000 to 1500 °C, suggesting reduced lattice constants and, thus, increased Al contents. Based on Bragg's law and the positions of the (−603) peaks, the d spacings were calculated for the samples. Then, using the lattice parameter formulas of β-(Al_xGa_1−x)₂O₃ from the study by Kranert et al.,⁴⁵ a = (12.21 − 0.42x) Å, b = (3.04 − 0.13x) Å, c = (5.81 − 0.17x) Å, β = (103.87 + 0.31x)^o, and Eq. (1),³⁴ the Al content x values were calculated, as shown in Table II. They manifest a wide range from 0 to 0.81 for the β-(Al_xGa_1−x)₂O₃ samples thanks to the thermal annealing,

Table S1 shows the crystal quality and the surface roughness of the samples. The crystal quality of β-(AlGa)₂O₃ improves proportional to the annealing temperature and, thus, the Al content. The XRD (−201) rocking curve (RC) FWHM drops from 2.51° for as-deposited Ga₂O₃ to 0.15° for the β-(AlGa)₂O₃ sample annealed at 1500 °C. This is in marked contrast to the previous works by direct-growth techniques where the crystallinity and crystal quality generally deteriorated when the Al content in the β-(AlGa)₂O₃ alloys increased.^30,31,38 The surface roughness probed by 5 × 5 μm² atomic force microscopy (AFM) scan remains less than 2 nm for the annealing temperatures from 1000 to 1400 °C. But for the 1500 °C annealed sample, the surface starts to decompose due to high temperature, making the surface roughness non-uniform across the sample surface.

The SIMS measurements were carried out by depth profiling of Al, Ga, and O elements for the as-deposited and annealed samples at 1000, 1200, and 1400 °C shown in Fig. 2. Figure 2(a) shows that the Al concentration of as-deposited Ga₂O₃ is at or below the detection limit level, indicating minimal Al diffusion at the PLD temperature of 800 °C. Also, the figure shows that the Ga₂O₃ film thickness is 50 nm. With the annealing at 1000 °C, significant Al diffusion occurred unambiguously, as shown in Fig. 2(b). The annealing also increased the film thickness to 110 nm, indicating that the interdiffusion process of Ga and Al atoms occurred at the III-oxide/sapphire interface. However, there is a small gradient of Ga and Al concentrations throughout the film. Thus, the alloy material composition is not strictly homogeneous for this sample. With higher temperatures of 1200 and 1400 °C, the film thickness increased further to 190 and 250 nm, respectively, suggesting stronger interdiffusion shown in Figs. 2(c) and 2(d). More importantly, the Al concentrations are highly homogeneous throughout the films for both samples, indicating that the samples with annealing temperatures of 1200 °C or higher would possess a homogeneous composition. The carbon signals in Figs. 2(a)–2(d) appear to be below the detection limit, indicating the low impurity level and, thus, low contamination amid the processes.

To measure the optical bandgap and, thus, the composition of the samples, the UV-Vis transmission measurement was conducted, as shown in Fig. 3(a). The blueshift in the absorption edge with higher annealing temperature can be observed, which suggests the increased Al composition in good agreement with the XRD experiments. Because the minimum detection wavelength is 190 nm, the transmission curves of the samples annealed at 1300 and 1400 °C are less complete. Also, it was not possible to obtain the absorption edge of the sample annealed at 1500 °C due to an even larger bandgap, which is, thus, not included in Fig. 3. Nevertheless, the bandgaps of the annealed samples at 1400 °C and below were first deduced from the Tauc plot shown in Fig. 3(b) calculated from the transmission spectra. Then, the bandgaps were determined by extrapolating linear regions of these curves to the horizontal axis. An increase in the bandgap with annealing temperature was observed. The optical bandgaps for as-deposited β-Ga₂O₃ and annealed β-(Al_xGa_1−x)₂O₃ at 1000, 1100, 1200, 1300, and 1400 °C are 4.88, 5.31, 5.82, 6.06, 6.21, and 6.38 eV, respectively.

Based on Figs. 2 and 3, the relationships of the film thickness vs temperature and the optical bandgap vs temperature show the quasi-linear behavior in Fig. 4(a). This suggests that the annealing temperature is an excellent knob to control the two important thin film quantities. Another potential knob is the annealing time, which is worth further investigation. Subsequently, the Al contents of β-(Al_xGa_1−x)₂O₃ were estimated by the equation E_g(x) = (1–x) E_g(β-Ga₂O₃) + x E_g(θ-Al₂O₃) – bx(1–x) eV. The bandgap E_g values of monoclinic β-Ga₂O₃ and θ-Al₂O₃ are 4.88 and 7.24 eV, and the bowing parameter b is 0.93 eV.^46,47 The Al contents of the annealed β-(AlGa)₂O₃ films at 1000, 1100, 1200, 1300, and 1400 °C are 0.26, 0.50, 0.60, 0.65, and 0.72, respectively. These Al contents estimated from the UV-Vis measurements are included in Table II for comparison with the ones estimated from the XRD experiments. It appears that the two sets have considerable discrepancy at lower annealing temperatures of 1000–1200 °C but show good agreement at higher annealing temperatures of 1300 and 1400 °C. Figure 4(b) compares the optical bandgap vs the Al content estimated from the optical bandgap of this work with those reported by others using various direct-growth techniques.^{28,31,34–36} The result shows that the proposed method possesses the largest optical bandgap tuning ranging from 4.88 to 6.38 eV for the β-(AlGa)₂O₃ alloys.

To investigate the III-oxide/sapphire interface, we performed cross-sectional TEM experiments of the samples annealed at 1000 and 1400 °C shown in Fig. 5. Figures 5(a) and 5(b) show the thicknesses of the samples annealed at 1000 and 1400 °C, which are 122 and 242 nm, consistent with the SIMS experiments in Figs. 2(b) and 2(d), respectively. According to the insets of Figs. 5(a) and 5(b), there is a thin transition layer between β-(AlGa)₂O₃ and sapphire.

To further study the interface of the two samples in Fig. 5, the cross-sectional HR-TEM images were obtained in the vicinity of the interface shown in Figs. 6(a) and 6(c). The transition layers around 3.6 nm and 4.5 nm thick at the β-AlGa₂O₃/sapphire interface can be identified for the 1000 and 1400 °C annealed samples, respectively. The fast Fourier transform (FFT) diffraction patterns below and above the transition layers of both samples confirm the α-phase corundum crystal structure for the sapphire substrate and the monoclinic β-phase crystal structure for the β-(AlGa)₂O₃ films. The FFT diffraction patterns of the transition layers of both samples, however, show the mixed α-phase and β-phase.

The STEM HAADF images and EDX maps of the two samples in the vicinity of the interface are shown in Figs. 6(b) and 6(d). The Al and Ga gradients in the EDX maps show that the Al atoms diffused from the sapphire substrate to the (AlGa)₂O₃ layer, which is the opposite of the Ga diffusion. Meanwhile, the O distribution was homogeneous. In Figs. 6(b) and 6(d), the EDX signal overlap maps show a tiny orange region comprising Al red and Ga green signals at the interface of β-(AlGa)₂O₃/sapphire, reaffirming the existence of the transition layer. For the EDX spectra in Figs. 6(b) and 6(d), the signal intensity of Al is weaker than that of Ga for the sample annealed at 1000 °C in Fig. 6(b) amid the β-(AlGa)₂O₃ layer. However, it is stronger than that of Ga for the sample annealed at 1400 °C in Fig. 6(d). The results reiterate that with higher annealing temperatures, more Al atoms could diffuse into the AlGa₂O₃ layer, leading to higher Al contents. It is noted that because the film thickness increased amid the annealing due to the interdiffusion process, one could design experiments with different initial thicknesses and contents to reach the desired thicknesses and contents by utilizing the knobs of the annealing temperature and perhaps the annealing time.

In summary, based on the binary β-Ga₂O₃/sapphire templates, we have employed the thermal annealing method at 1000–1500 °C to demonstrate ternary β-(Al_xGa_1−x)₂O₃ films with wide ranges of optical bandgaps and compositions. Higher annealing temperatures resulted in larger Al contents for the β-(Al_xGa_1−x)₂O₃ films. The optical bandgap range of the samples was 4.88–6.38 eV translating to the Al content range of 0–0.72 for the β-Ga₂O₃ templates without annealing and with annealing up to 1400 °C. Evaluated by XRD, the Al content range of the samples was 0–0.81 for the templates without annealing and with annealing up to 1500 °C. The ternary alloy formation was caused by the interdiffusion of Ga and Al atoms from the thin film and substrate, respectively, which was proportional to the annealing temperatures. The interdiffusion was confirmed by SIMS and TEM. The Al contents of β-(Al_xGa_1−x)₂O₃ ternary thin films were highly homogeneous with higher temperatures (≥1200 °C). The Ga atoms diffused concurrently into the sapphire substrate during the annealing, resulting in an increased β-(AlGa)₂O₃ film thickness compared to the β-Ga₂O₃ templates. Besides, the crystal quality of β-(Al_xGa_1−x)₂O₃ could improve significantly amid higher temperature annealing. Since the annealing temperature is an excellent knob to control the bandgap and composition, one could design a proper initial thickness and composition of binary β-Ga₂O₃ and β-(AlGa)₂O₃ to reach the desired thickness and composition of ternary β-(AlGa)₂O₃ films by utilizing the annealing technique. The proposed method does not involve the direct-growth techniques for alloys and, therefore, is promising for straightforward and low-cost production of the high-quality β-(AlGa)₂O₃ alloys.

See the supplementary material for XRD phi-scan of the (−401) plane and a table summary of XRD (−201) rocking curve FWHMs and 5 × 5 μm² AFM RMS surface roughness of the samples.

The authors would like to acknowledge the support of KAUST Baseline BAS/1/1664-01-01, GCC Research Council REP/1/3189-01-01, and Competitive Research Grant Nos. URF/1/3437-01-01 and URF/1/3771-01-01.