Introduction

Gallium oxide (Ga2O3), one of the ultra-wide bandgap (UWBG) semiconductors with a bandgap of ~ 4.9 eV, attracted much attention in the past several years due to its promising applications in power electronics and deep ultraviolet (UV) photonics1,2,3,4,5,6. Ga2O3 has five polymorphous: \(\alpha\)-phase (rhombohedral), \(\beta\)-phase (monoclinic), \(\gamma\)-phase (spinel), \(\delta\)-phase (body-centered cubic), and \(\varepsilon\)-phase (hexagonal)7. Among these polymorphous, \(\beta\)-phase is the most stable. \(\beta\)-phase Ga2O3 is a direct bandgap semiconductor and has high critical electric field strength and high Baliga’s figures of merit8,9,10. Despite extensive studies of Ga2O3, many challenges are still not resolved11,12,13,14. One of them is to find a suitable p-type dopant that can enhance its potential for practical applications such as deep-UV lasers. When the composition of a p-type dopant is tuned in a wide range, not only electrical and optical properties but also crystal structure will change. For example, excessive incorporation of Mg into Ga2O3 increases the bandgaps of the resultant materials within a range of 4.9 eV and 7.8 eV15, and induces a phase transition from beta to rocksalt phases16.

On the other hand, incorporation of elements such as zinc into Ga2O3 can decrease the bandgap. The bandgap is expected to be tuned within the range of approximately 4.9 eV and 3.3 eV by tuning the Zn composition in zinc gallium oxide (ZnGaO) alloy17,18,19. Meanwhile, the crystal structure is expected to change from beta to spinel, and to Wurtzite20. Many studies showed that Ga-doped Wurtzite phase ZnO is an excellent transparent conducting oxide and Zn-doped \(\beta\)-phase Ga2O3 has great potential in solar-blind photodetectors18,19,21,22,23,24. However, spinel structure ZnGa2O4 having a bandgap of approximately 5.2 eV has not yet been fully studied25,26,27. These materials can be used in deep-UV optoelectronic and high-power electronic applications26,28,29. To fully realize the potential of spinel ZnGa2O4 for various electronic and optoelectronic applications, we must understand its band structure. In the recent years, researchers used photoluminescence (PL) to understand the optical transitions of spinel ZnGaO30,31,32,33,34,35 and doped samples using Ni36, Bi37, Cr38,39,40, Mn41,42,43,44,45, and other dopants46,47,48. Nevertheless, these prior PL studies had used lasers with excitation energies less than the bandgap energy of these materials, leading to impossible pumping of electrons from valence band to conduction band. In addition, self-trapped holes (STHs) are assumed to exist in Ga2O3 and ZnGaO materials49,50,51. Although their binding to certain defects or atoms was predicted by density functional theory51, experimental validation of their binding energies has not been carried out. In this work, ZnGaO samples were grown using plasma-assisted molecular beam epitaxy (MBE). While chemical vapor deposition (CVD), pulsed laser deposition (PLD) and other tools are widely used, MBE is known to also produce high-quality oxide semiconductors including β-Ga2O352,53,54,55, ZnO56,57,58, etc. Here, the effect of different Zn compositions on the crystal structure and optical property of ZnGaO thin films was studied. In particular, PL studies using a 193 nm ArF laser with an excitation energy of 6.42 eV were performed on the ZnGaO thin films to understand their optical transition energy levels within the forbidden gap and STH binding energies.

Results and discussion

Table 1 shows the structural and compositional results of nine ZnGaO samples with Zn composition ranging from 0 to 48.0 at% and film thickness ranging from ~ 280 to ~ 420 nm. Figure 1a shows EDX spectra of all samples. The peaks located at 0.53, 1.01, 1.10, and 1.49 keV correspond to oxygen, zinc, gallium, and aluminum, respectively, and a signal at 0.27 keV is due to adventitious carbon. Here, only the composition ratio between Ga and Zn has been estimated, as shown in Table 1, while oxygen atomic percentage is not included because the O peaks in EDX spectra include contributions from both the film and substrate as a result of large electron beam penetration depth. The EDX 1.01 keV peak intensity increases as Zn composition increases from sample 1–9. SEM images of all samples with different compositions are shown in Fig. S1 in the supporting material. Figure S2 shows XPS survey spectra of all samples. The peaks with binding energies located at ~ 1118.1, ~ 1145.0, ~ 1022.0, ~ 1045.2, and ~ 530.9 eV correspond to Ga 2p3/2, Ga 2p1/2, Zn 2p3/2, Zn 2p1/2, and O 1s, respectively. The peak areas of Zn, Ga, and O were extracted to calculate the atomic percentage of each element for each sample, which is listed in Table 1. The Fermi level locations were also extracted from XPS valence band spectra, which are shown in Fig. S3. All samples show n-type characteristics since the Fermi levels are located in the upper portion of the forbidden gap.

Table 1 Film thickness, Ga, Zn, and O atomic composition, optical bandgap, crystal structure and lattice parameters of ZnGaO samples.
Full size table
Fig. 1
figure 1

(a) EDX spectra of all ZnGaO samples. (b) Transmittance spectra of all ZnGaO samples. (c) Tauc plot absorption spectra of all ZnGaO samples. (d) XRD pattern in θ/2θ scan mode of all ZnGaO samples.

Full size image

Figure 1b shows transmittance spectra of ZnGaO samples by using the equation: \(A = 2 - log(\%T)\), where A is the absorbance and T is the transmittance. All samples show high transmittance in the visible spectral range. Figure 1c shows Tauc plot absorption spectra of all ZnGaO samples. Absorbance spectra were obtained and converted into absorption coefficient using the equation \(\alpha =2.303\times \frac{A}{t}\), where \(\alpha\) is the absorption coefficient, \(A\) is absorbance obtained from the spectrometer, and \(t\) is the film thickness in centimeters. Optical bandgap can be extracted by linear extrapolation of Tauc plot onto the x-axis using the equation \({(\alpha h\upsilon )}^{2}= c(h\nu - Eg)\), where h is Planck’s constant, \(\nu\) is frequency, \(c\) is constant, and \({E}_{g}\) is optical bandgap. Figure S4 shows a plot of bandgap versus Zn composition. The optical bandgaps of sample 1–5 seem to increase although undulate slightly around 5 eV or so as Zn composition increases, which may be due to the change of crystal structure from beta phase to spinel phase as seen in XRD studies later. The bandgap monotonously decreases from ~ 5 to 4.84 eV as Zn continues to increase in pure spinel phase sample 5 to sample 9.

Figure 1d shows XRD \(\theta /2\theta\) scan pattern of ZnGaO samples. Peaks located at ~ 20.6° and ~ 61.1° are background signal, which is confirmed by measuring substrate only. For sample 1, \(2\theta\) peaks located at ~ 19.12°, ~ 38.47°, and ~ 59.19° correspond to \((\overline{2 }01)\), \(\left(\overline{4 }02\right)\), and \((\overline{6 }03)\) orientations of beta phase ZnGaO, respectively, indicating the growth direction of < \(\overline{2 }01\)> along < 0001 > c-sapphire. An additional plane of (113) at ~ 60.36° was also detected. The XRD pattern suggests that sample 1 is pure β-phase Ga2O3. For sample 2–7, similar \(2\theta\) peak positions located at \(\sim 18.6^\circ\), \(\sim 37.8\)°, and \(\sim 57.6\)° were observed, although peaks shift towards a lower angle compared to that of sample 1. This can be due to either Zn2+ (74 pm) having a larger ionic radius than Ga3+ (62 pm), which results in a larger plane distance in beta phase, or the change of crystal structure from beta to spinel phase59,60.

Because beta phase and spinel phase have similar out-of-plane peak positions, in-plane RSM was performed to identify the crystal structure, as shown in Fig. 2. Figure 2a shows in-plane RSM of sample 1, and a strong signal located at \(2{\theta }_{\chi }\sim 61.35^\circ\) indicates the beta phase structure of (020) plane. Figure 2b shows in-plane RSM of sample 2, and two signals located at \(2{\theta }_{\chi }\sim 61.5^\circ\) and \(\sim 64^\circ\) were observed, indicating a mixture of beta phase (020) plane and spinel phase (440) plane, respectively. Figure 2c–i shows in-plane RSM of sample 3–9, and the signal located at \(2{\theta }_{\chi }\sim 63^\circ\) indicates the spinel phase structure of (440) plane. Based on the in-plane RSM results, it can be concluded that for sample 3–9, \(2\theta\) peaks located at \(\sim 18.6^\circ\), \(\sim 37.8^\circ\), and \(\sim 57.6^\circ\) correspond to \((111)\), \(\left(222\right)\), and \((511)\) orientation of spinel phase, respectively, indicating the growth direction of < \(111\)> along < 0001 > c-sapphire. Next, according to Bragg’s law \(2dsin\theta =n\lambda\), where \(d\) is the distance between the two planes, \(\theta\) is the incident beam angle from the scattering plane, \(n\) is the diffraction order, and \(\lambda\) is the wavelength of the x-ray, lattice parameters can be extracted from \(\theta /2\theta\) scan and in-plane RSM. To calculate lattice parameters for beta structure, \(b=2d\) was used for in-plane RSM, \(a=2\times \frac{d}{\cos(39.9^\circ )}\) and \(c=1\times \frac{d}{\text{cos}(36.2^\circ )}\) for \(\theta /2\theta\) scan. For spinel structure, \(a=\frac{d}{\sqrt{{h}^{2}+{k}^{2}+{l}^{2}}}\) was used for in-plane RSM and assuming \(a=b=c\) because of the cubic structure. These calculated lattice parameters are listed in Table 1 and plotted as a function of Zn composition in Fig. S5. These lattice parameters are similar to reported values for beta phase Ga2O361,62 and spinel phase ZnGa2O463,64, respectively. In spinel samples, the lattice parameter increases slightly from 8.25 to 8.35 Å as the Zn composition increases.

Fig. 2
figure 2

XRD in-plane RSM of (a) beta structure (020) plane for sample 1, (b) mixture of beta structure (020) plane and spinel structure (440) plane for sample 2, (c) weak spinel structure (440) plane for sample 3, and (d)-(f) strong spinel structure (440) plane for sample 4–9.

Full size image

Figure 3a–i show room temperature PL spectra and the peak deconvolution of sample 1–9. The spectra of all samples except sample 2 and 3 are deconvoluted into five peaks. Figure 4 shows a schematic diagram of optical transition energy levels of the five deconvoluted peaks. Peak 1–5 corresponds to UV, violet, blue, green, and red emissions. As seen from Fig. 3a, which is PL spectrum of sample 1 with Zn composition of 0.0 at% (β-Ga2O3), UV emission is due to electron–hole recombination from conduction band to gallium-oxygen complex vacancies (VGa-O). By using the equation \({E}_{{V}_{Ga-O}}={E}_{g}-{E}_{UV}\), where \({E}_{UV}\) is the energy of the UV emission, and \({E}_{{V}_{Ga-O}}\) is the energy of VGa-O, the estimated location of VGa-O is 0.91 eV above the valence band edge. Violet emission is due to recombination from the conduction band to STHs. Similar violet emission energy was observed and reported by others65,66. STHs exist in some oxide materials, and here the energy is assumed to be 1.10 eV above the valence band67. By using the equation \({E}_{PL}={E}_{g}-{E}_{STH}-{E}_{STH \cdot binding}\), where \({E}_{STH \cdot binding}\) is the binding energy of STHs, \({E}_{STH \cdot binding}\) can be calculated to be approximately 0.39 eV68,69. Blue, green, and red emissions are due to optical transitions between three oxygen vacancies (VO) and VGa-O, commonly known as donor-to-acceptor pair recombination (DAP). Similar emission peaks in the visible light range were observed in literature65,70,71. By using the equation \({E}_{{V}_{O}}={E}_{g}-{E}_{PL}-{E}_{{V}_{Ga-O}}\), the three oxygen vacancies VOI, VOIII, and VOII levels are estimated at 0.91 eV, 1.26 eV, and 1.78 eV below the conduction band edge, respectively. In the PL spectrum of sample 2 with Zn composition of 0.9 at% as seen in Fig. 3b, only 4 deconvoluted peaks were observed: UV, blue, green, and red emission. The violet emission, which is the transition between conduction band and STHs, cannot be observed. This could be due to the fact that the mixture phase has a lower-quality film, which results in stronger emission from the transition of DAP and the transition of conduction band to acceptor level. The origin of the blue, green, and red emissions are from transitions between three VO levels and VGa-O. By using the equation \({E}_{{V}_{Ga-O}}={E}_{g}-{E}_{UV}\), the VGa-O level energy for sample 2 is estimated to be 0.82 eV above the valance band maximum. Similar to sample 2, sample 3 only has 4 deconvoluted peaks as seen in Fig. 3c, which may be due to the weak spinel crystal structure. The acceptor level energy for sample 3 is calculated to be 0.84 eV above the valence band. Figure 3d–f show PL spectra of sample 4–6 with Zn composition of 7.3, 16.5, and 33.1 at%, respectively. The DAP emissions could originate from optical transitions between three VO and Zn vacancies (VZn) or Ga vacancies (VGa) because these spinel samples have lower Zn compositions compared to the normalized cations of spinel ZnGa2O4. The acceptor level energy for sample 4–6 is estimated to be 0.85, 0.77, and 0.78 eV, respectively. Figure 3g–i show PL spectra of sample 7–9 with Zn composition of 43.9, 45.2, and 48.0 at%, respectively. These samples have higher Zn composition compared to the normalized cations of spinel ZnGa2O4. The calculated STH binding energy is 0.24, 0.18, and 0.15 eV for sample 7–9, respectively. These are similar to the theoretically estimated value of \({E}_{STH.binding}=0.21 \text{eV}\) from Lyons et al.51 The origin of DAP emissions can be attributed to the transitions between three VO levels and acceptor level, which could be Zn replaced Ga site (ZnGa) or VGa. The acceptor level energy for sample 7–9 is 0.88, 0.88, and 0.69 eV, respectively. All calculated values of oxygen vacancy levels, STH binding energies, and acceptor level energies are listed in Table 2. The energies of the three oxygen vacancies are referenced to the conduction band, and the acceptor levels are referenced to the valence band.

Fig. 3
figure 3

Room temperature photoluminescence spectra under 193 nm laser excitation and their peak deconvolutions of ZnGaO samples with Zn composition of (a) 0.0 at%, (b) 0.9 at%, (c) 3.4 at%, (d) 7.3 at%, (e) 16.5 at%, (f) 33.1 at%, (g) 43.9 at%, (h) 45.2 at%, and (i) 48.0 at%.

Full size image
Fig. 4
figure 4

Schematic band diagram with intermediate energy levels and optical transitions corresponding to the deconvoluted PL peaks of sample 1–9.

Full size image
Table 2 Calculated oxygen vacancy energies below conduction band, STH binding energy, and acceptor level energy above valance band for sample 1–9.
Full size table

Figure 5 shows temperature-dependent PL spectra with temperatures ranging from 14 K to room temperature, and the inserted graph shows the integrated PL peak area as a function of temperature. As seen from Fig. 5, all samples exhibit general trend of the decrease of PL intensity as the temperature increases in the higher temperature range. This is known as the positive thermal quenching effect, where non-radiative recombination becomes more dominant under high temperature, which results in the decrease of the PL intensity. In the meantime, all samples, especially sample 1–5, show that the PL intensity increases as the temperature increases in a narrower and lower temperature range, which is ascribed to the negative thermal quenching effect. Negative thermal quenching exists because the rates of optical transitions involving intermediate energy levels in the forbidden gap including donor–acceptor pair transitions increase as the temperature increases and these transitions have dominated radiative recombination within the temperature range. To qualitatively describe the temperature-dependent PL intensity, the following equation is used72,73: \(I\left(T\right)=\frac{1+a{e}^{(-{E}_{NA}/kT)}}{1+b{e}^{(-{E}_{A}/kT)}}\), where \(I\left(T\right)\) is the integrated PL intensity at temperature T, \({E}_{A}\) is the nonradiative recombination activation energy that results in positive thermal quenching process, \({E}_{NA}\) represents the activation energy of optical transitions from intermediate states that results in negative thermal quenching process, \(T\) is the temperature in Kelvin, and \(a\) and \(b\) are rate parameters. To extract nonradiative recombination activation energies at high temperature, the term \(a{e}^{(-{E}_{NA}/kT)}\) is much smaller than 1, and the term \(b{e}^{(-{E}_{A}/kT)}\) is much larger than 1, so the above equation is simplified to \(I\left(T\right)\propto {e}^{({E}_{A}/kT)}\)74,75,76. Figure S6 shows the Arrhenius plot of sample 1–9. The activation energy of the nonradiative recombination is 13 meV for sample 1 with β-phase structure, 43 meV for sample 2 with mixture phase, and between 11 to 32 meV for sample 3–9 with spinel structure. It is noted that Arrhenius plot cannot be fitted in low-temperature range due to negative thermal quenching effect. Within this temperature range, as seen from the equation mentioned above, the numerator dominates the PL intensity compared with the contribution from the denominator, which results in an increase of the PL intensity in the temperature range.

Fig. 5
figure 5

Temperature-dependent PL spectra under 193 nm excitation with Zn composition of (a) 0.0 at%, (b) 0.9 at%, (c) 3.4 at%, (d) 7.3 at%, (e) 16.5 at%, (f) 33.1 at%, (g) 43.9 at%, (h) 45.2 at%, (i) 48.0 at%. The inserted graph shows the integrated PL peak area as a function of temperature.

Full size image

Conclusion

In this work, 9 ZnGaO samples with Zn composition ranging from 0.0 to 48.0 at% were grown using a plasma-assisted MBE system. When Zn composition is increased to 0.9 at%, the sample exhibits a mixture of beta and spinel structure. When Zn composition is 3.4 at%, the sample exhibits weak spinel structure. When Zn composition is larger than 7.3 at%, all samples exhibit spinel structure. PL studies of these samples using a 193 nm laser as excitation were carried out to systematically analyze and understand the optical transition of these samples. The STH binding energy and the position of the three oxygen vacancies and acceptor levels were revealed. This study paves a way towards the application of spinel ZnGa2O4 for deep ultraviolet photodetectors and other optoelectronic applications.

Methods

Film growth

A 2-inch c-plane sapphire wafer was used as substrate in this work and was cleaned by using piranha solution (H2SO4: H2O2 = 5:3) to remove any organic contamination, then rinsed with DI water to remove residual acid from the substrate. After being blow-dried by nitrogen, c-plane sapphire substrate was introduced into the loadlock of an SVT Associates plasma-assisted molecular beam epitaxy (MBE) chamber. 6N Ga and 6N Zn from Alfa Aesar were used for sample growth. The composition ratio of Zn and Ga can be adjusted by tuning the effusion cell temperature of Zn and Ga. Oxygen flow rate was set to 2.5 sccm and the RF power was 400 W. Pre-growth annealing of substrate at 800 °C for 15 min was performed, then substrate temperature was decreased to 450 °C for sample growth. After that, substrate temperature was increased to 700 °C and a post-growth annealing process with oxygen was carried out for 20 min.

Film characterization

The thickness of samples was measured using a Nikon Profilm3D profilometer. The surface morphology of ZnGaO thin films were measured using a TESCAN Vega3 SBH Scanning Electron Microscopy (SEM) system at a voltage of 10 kV, a beam intensity of 7, and a working distance of ~ 7.5 mm. The Zn/Ga atomic ratio was measured using a Quantax energy dispersive x-ray (EDX) system at a voltage of 15 kV, a beam intensity of 15, and a working distance of 14 mm. To analyze the ratio of Zn, Ga, and O composition and valence band spectra for each sample, X-ray photoelectron spectroscopy (XPS) characterizations of different ZnGaO thin films were carried out by a Kratos AXIS ULTRADLD XPS system using an Al Kα X-ray source. During measurement, vacuum pressure was kept below 3 × 10−9 torr. X-Ray diffraction (XRD) in θ/2θ scan mode was measured using a PANalytical Empyrean Series 2 XRD system with Cu Kα x-ray (λ = 0.15405 nm) to understand the out-of-plane crystal orientation. The in-plane distance and crystal structure were evaluated using a Rigaku SmartLab X-ray Diffractometer with Cu x-ray tube (λ = 0.1540 nm) and operational voltage and current of 40 kV and 44 mA, respectively. Absorbance and transmittance spectra of ZnGaO samples were obtained using an Agilent Cary 5000 Double Beam UV/VIS/NIR spectrometer from 200 to 800 nm with a resolution of 0.1 nm. PL spectra were obtained using a COHERENT ExciStar 193 nm laser system equipped with ArF gas source.