Influence of oxygen partial pressure on properties of monoclinic Ga₂O₃ deposited on sapphire substrates

There has been an increasing interest in Ga₂O₃ due its physical properties such as ultrawide bandgap (E_g∼ 4.9 eV at room temperature), high transparency extending from UV to near-IR, and high estimated breakdown field (∼8 MV/cm). High-quality single crystal monoclinic gallium oxide (β-Ga₂O₃) boules have been synthesized by melt based methods, which has resulted in the recent commercialization of wafers up to four inches in diameter with very good control of the free-electron concentration and resistivity through controlled doping. Despite the low thermal conductivity, relatively low electron mobility, high anisotropy of the physical properties, and the lack of p-type doping, this material provides an opportunity to develop devices that could surpass some devices based on well-established GaN and SiC wide bandgap semiconductors.

Homoepitaxial growths of β-Ga₂O₃ thin films on (100), (010), and (001) native substrates by molecular beam epitaxy (MBE), metalorganic chemical vapor deposition (MOCVD), and halide vapor phase epitaxy (HVPE) have also been reported.^1–10 However, challenges such as gas-phase nucleation in the case of MOCVD have limited growth rate significantly in the past.^11,12 High growth rate MOCVD with up to 10 μm/h has been achieved using a close-coupled showerhead reactor designed to suppress gas-phase suboxide nucleation.^13,14 For halide vapor phase epitaxial β-Ga₂O₃, growth rates of up to about 20 μm/h have been reported; however, it appears that the 5–10 μm/h range results in superior quality material, although it should be noted that HVPE β-Ga₂O₃ homoepitaxial layers were grown on (001) substrates as opposed to the (010) and (100) oriented substrates used by Gogova and Tadjer, respectively.^11–13 Homoepitaxial Ga₂O₃ thin films grown by MBE and HVPE have been employed in lateral^15–17 and vertical¹⁸ transistor devices, as well as Schottky and heterojunction diodes.¹⁹ The growth of β-Ga₂O₃ homoepitaxial layers on (010) β-Ga₂O₃ substrates by ozone MBE and plasma-assisted MBE¹⁹ have also been accomplished. The reported growth rate was relatively slow (<0.7 μm/h) but material scientist have striven to develop approaches to increase film deposition rates for high power device applications.

β-Ga₂O₃ thin films have been previously deposited on a number of foreign substrates such as Al₂O₃, Si, GaAs, ZrO₂:Y, and MgO by MBE^20–27 and radio frequency magnetron sputtering.^28–30 It has been demonstrate that devices with good performance, e.g., gas sensors,³¹ solar-blind photodetectors,³² field effect transistors,³³ among others, can be processed on β-Ga₂O₃/sapphire templates. In addition, it is important to develop approaches to deposit electronic-grade/high-quality epitaxial films on foreign substrates because heterostructured layers will be required to develop various device types. We previously reported the deposition of β-Ga₂O₃ on sapphire substrates with a growth rate of 1.1 μm/h (Ref. 34) by low pressure chemical vapor deposition (LPCVD),^35,36 and more recently, the synthesis of a single phase β-Ga₂O₃ homoepitaxial layer on (010) β-Ga₂O₃ substrates by LPCVD.^37,38 The advantages of the LPCVD technique include low cost, high reproducibility and scalability, good uniformity of film thickness and homogeneity of the deposited thin films. The dependence of β-Ga₂O₃ thin film growth rate on the oxygen volume percentage at two fixed growth temperatures was verified. The morphology, crystal structure, and optical and electrical properties of these films were investigated in the present work.

A single heat zone tube furnace with programmable temperature and high-precision pressure controllers was used to deposit β-Ga₂O₃ thin films on c-plane sapphire (on-axis) substrates. The substrates were cleaned with acetone and isopropanol, rinsed with de-ionized water, and dried with nitrogen flow before being placed on the reactor susceptor. High purity gallium pellets (Alfa Aesar, 99.99999%) and oxygen (O₂–99.9999%) gas, the precursors for gallium and oxygen, respectively, were transported by Argon (Ar) carrier gas: the oxygen/argon flow was passed over the Ga melt and proceeded downstream to the growth substrate. The LPCVD film depositions were carried out using different growth temperatures and oxygen volume percentages. Note that the O₂ flow rate is relatively low as compared to the Ar flow rate but the O₂ flow can affect the gas-phase reaction with the Ga vapor. The sample thicknesses, acquired in multiple cross-sectional locations of cleaved samples with an e-beam lithography scanning electron microscope calibrated on a regular basis, were generally uniform across the entire film.

The structure and crystal quality of the thin films were characterized using x-ray diffraction (XRD) and Raman scattering (RS) spectroscopies. XRD spectra were collected using a Rigaku SmartLab 9 kW ray diffractometer with a rotating anode to produce Cu-K_α monochromatic radiation (1.54 Å). Micro-Raman measurements were performed using a single-mode 488 nm solid state laser, which was made coaxial with the detection axis using the beam-splitter part of a volume-Bragg-grating filter set from Optigrate. These filters allow Stokes and anti-Stokes Raman measurements closer than 10 cm⁻¹ to the laser. A 50× microscope objective (NA = 0.65) was used to both focus the incident light into a small spot (∼0.4 μm) and collect the scattered light, which then was dispersed in a half-meter Acton SP-2500 single-spectrometer and detected using a Princeton Instruments CCD array (Spec-10:400BR back-thinned, deep-depleted). Cathodoluminescence (CL) emission was excited with an electron beam current of 3 μA and a beam accelerating voltage of 3 kV in an ultrahigh vacuum chamber at 5 K and at room temperature (∼297 K). The CL emission, collected with a combination of f-number matching lens and mirrors, was analyzed by a compact fiber optic spectrometer. Under this condition, the maximum electron penetration depth (Bohr–Bethe) is only about 80 Å.

Previous electrical characterization via the Hall method of Ga₂O₃ epitaxial films grown via LPCVD at 900 °C has shown unequivocally their n-type conductivity.³⁵ In particular, the mobility of these films was relatively independent of growth conditions and the carrier concentration varied by about a factor of 2.5 from 4 × 10¹⁷ cm⁻³ to 10¹⁸ cm⁻³. In this work, we explore a much wider range of growth conditions and demonstrate that LPCVD Ga₂O₃ conductivity can substantially decrease when growth temperature is reduced by 100 °C or when the O₂ flow rate is increased, regardless of growth temperature. Because of this wider window of process parameters, not all films were conductive which made Hall measurements difficult to implement across the entire sample set. For this reason, resistance measurements were performed using a Keithley 4200SCS semiconductor parameter analyzer using circular 500 μm diameter Ti/Au contacts deposited simultaneously on all samples via e-beam evaporation through a shadow mask.

Table I lists the thickness and the deposition conditions of eight samples synthesized at 800 and 900 °C. Figure 1(a) shows XRD spectra of the LPCVD Ga₂O₃ films deposited on sapphire substrates at 800 °C with decreasing flow rate ratio (FRR). Figure 1(b) shows the XRD spectra zoomed in around (−603) line of films deposited at 800 °C along with one thicker film synthetized at 900 °C. Figure 1(b) highlights the shift in the XRD peak position of the (−603) line of the β-Ga₂O₃ film with the film increasing thickness_. Based on the XRD peak position and line shape, all films have β-Ga₂O₃ structures with the (−201) orientation. Films deposited with higher FRR and higher substrate temperature have larger thicknesses (note that the deposition time is about six times longer) and improved crystalline quality. The variation of (−603) plane d-spacing with the thickness is plotted in Fig. 1(c). As thickness increases, d-value gets closer to the bulk value.

Feng and co-workers verified and reported that as the growth temperature increased above 900 °C, the growth rate started to saturate.³⁵ This may indicate that the growth at higher temperature is mass-transport limited, where the growth rate is primarily limited by the amount of chemical species reaching the growth surface. Considering the dominating gas-phase reactions in Ga₂O₃ LPCVD setup, the precursors gradually deplete during the gas transport. The gas-phase reaction rate could be significantly increased at higher temperatures, exceeding the availability of Ga vapor evaporation. This could lead to even lower growth rates, which could be also described as “precursor depletion” at higher growth temperatures.³⁹ It was noticed that the oxygen flow rate in this setup is generally lower. The lower flow of oxygen might limit the growth rate at 900 °C; we observe an increased growth rate when the oxygen flow rate is increased from 5 to 30 sccm.

RS is a well stablished noninvasive/nondestructive technique commonly used to identify crystal structure and verify crystalline quality based on the observation of vibrational modes and their polarization selection rules. Micro-Raman allows the evaluation of sample morphology, grain boundary orientation, strain, and impurity distribution. All Raman measurements shown here were acquired in a back-scattering geometry z(yy)Z. The axis of the incident 488 nm light was normal to the sample surfaces, and light back-scattered parallel to the incident beam was collected and measured.

Raman scattering measurements performed for all eight epitaxial growths are represented in Fig. 2. These spectra show the forest of Raman lines^40,–43 characteristic of β-Ga₂O₃. Raman line-widths measured here match those measured for high-quality single crystal β-Ga₂O₃ samples. Raman lines from the underlying sapphire substrate also appear with thicker films having weaker sapphire substrate Raman signal.

Given the inevitable lattice mismatch between epi-films and substrates, and the expected relaxation of epi-film lattice constants toward bulk values as films grow thicker, it is unsurprising that we observe in Fig. 3, the A_g⁽³⁾ Raman line at ∼200 cm⁻¹ showing a clear shift to lower frequencies for thicker films. Using the measured⁴¹ Grüneisen parameter for the 200 cm⁻¹ line, the observed shift in line position implies that the thinner epitaxial β-Ga₂O₃ discussed in this work are under a strain equivalent to a fractional volumetric compression of δV/V = −3.8 × 10⁻³ of the thinnest films relative to the thickest films. The atomic motions of the (−201) A_g⁽³⁾ phonons being probed by Raman measurements are dominantly in the (−201) plane; the shift of this phonon to higher frequencies for thinner films is consistent with the films being under lateral compression. Due to the Poisson-ratio effect, the spacing of the (−201) planes of atoms should be larger for thinner films, decreasing toward bulk values with increasing film thickness. This behavior is observed as seen in Fig. 1(a). Comparison between the peak positions of the bare c-plane sapphire phonon line at 577.1 cm⁻¹ with those detected in the spectra acquired from all films showed no detectable shift with film thicknesses.

Polarized Raman measurements were performed on both thick and thin epitaxial films. The results for the 2.1 μm thick film are shown in Fig. 4(a). To avoid clutter, only four phonons are shown: A_g⁽²⁾, A_g⁽³⁾, A_g⁽⁴⁾, and A_g⁽⁵⁾. These phonons were chosen because they are the most intense Raman peaks that do not have sapphire peaks close to them. The intensities of the β-Ga₂O₃ Raman lines did not change as the polarization of the normally incident light was rotated through 360° in the plane of the film. This is in distinct contrast to measurements made on large crystal β-Ga₂O₃ samples using the same measurement system. Figure 4(b) shows polarized Raman measurements for a high-quality single crystal (−201) β-Ga₂O₃ substrate. Measurements for (010) and (001) crystal samples (not shown) exhibit different but also very anisotropic behaviors. The Raman lines of monoclinic β-Ga₂O₃ are normally extremely anisotropic.^42,43

These observations mean that the grains in these epitaxial films are small compared to the ∼0.5 μm diameter of the incident laser beam and that (at the very least) the grains making up the film are oriented randomly in the plane of the film. The possibility of the grains having the same crystalline orientation in the direction normal to the plane of growth remains. The spectra for these epi-films were compared to 360° polarized Raman measurements made on high-quality single crystal (−201), (001), and (010) β-Ga₂O₃. When the spectra for these three crystal faces were averaged over 360° of polarization rotation, an excellent match was found between the epi-films and the (−201) crystal average spectrum (see Fig. 5); the match was very poor for the (001) and (010) crystals average spectra (not shown). Unless the (−201) crystal average spectrum coincidentally matches the 4π-steradian average spectrum, this means the grains in the epi film are oriented in the (−201) direction normal to the growth plane but are randomly oriented within the plane of growth. This result is in agreement with the XRD measurements, which indicate that all eight heteroepitaxially grown samples consist of β-Ga₂O₃ crystals whose [−201] axes are normal to the growth plane. These observations are in excellent agreement with results recently reported by Rafique and co-workers,⁴⁴ who performed Scherrer analysis on their XRD data to estimate a grain size of 5.1 nm for films deposited on c-plane sapphire. In addition, they verify from the AFM, SEM, and TEM studies that these films are composed of randomly oriented small pseudo-hexagonal rotational domains.^34,44–46

As discussed above, the results of the RS study are in full agreement with the XRD observations. Both indicate that the beta (monoclinic) phase is being grown with its (−201) plane being in the plane of the substrate; the Raman adds that the films are composed of small submicrometer crystallites whose orientations are otherwise randomly rotated relative to each other within the plane of the film. Both techniques show a change in atomic spacing with increasing film thickness due recovery from lattice mismatch, with that relaxation occurring in the opposite directions, as expected due to the Poisson-Ratio effect because Raman measures the lateral/in-plane relaxation, and x ray measures the separation of lattice planes normal to the substrate.

The low-temperature CL spectrum of the eight films, characterized by a broad emission band extending from 1.5 to 4.5 eV (∼825 and 275 nm), are shown in Fig. 6. There is a considerable dependence of the emission band peak intensities on the deposition conditions, namely, the FRR and substrate temperatures. Note that as the CL intensity changes, the spectral line shape changes. Such spectral variations are more conspicuous in the corresponding film spectra acquired at ∼297 K (room temperature). The measured emission band appears to have contributions from multiple and overlapping constituent bands. The breadth of the constituent bands and the lack of knowledge of the recombination processes giving rise to these emission bands prevent us from determining their peak positions and line shapes and from achieving unique and meaningful line shape fits. In an attempt to bypass these deficiencies, we carried out a large number of fitting iterations using four Gaussian functions that being the smallest number of Gaussians that resulted in full converging and excellent chi-squared fitting. Parameter ranges for full width at half maximum (FWHM) and peak position of each constituent emission line were established; the resulting fits were reproducible and characterized by small deviations of the average values within the range. After these parameter ranges values were established, the fitting iterations were repeated using parameter values within the established ranges. The final average parameter values of each Gaussian line, obtained from all film luminescence spectra acquired at 5 and 297 K, are summarized in Table II. The Gaussian peak position and FWHM values stay within the established parameter range for all line fits. The deviation of the peak position of individual lines from that of the average values are within ± ≤ 2% for spectra acquired at both temperatures. However, the deviation of the FWHM of individual lines from that of the average values are less than ±5% and ±7% for the spectra acquired at 5 and 297 K temperatures, respectively. The standard deviation of the four emission band peak positions, obtained from the eight sample spectra acquired at both temperatures, increase from the fifth to the third (<0.005) decimal place with decreasing emission band energy. A slightly larger standard deviation was observed for the room temperature data, consistent with increased lines FWHM. The energies of the individual lines acquired at 5 K are at 3.57, 3.29, 3.04, and 2.77 eV. The energy values of the emission bands detected (by fitting) in the 5 K CL spectra are close to those recently reported by Gao and co-workers⁴⁵ acquired from films submitted to remote oxygen plasma processing, neutron irradiation, and forming gas annealing. Note that, our experiments were carried out on as-deposited heteroepitaxial films. Differences in the individual emission band peaks position may result from film morphology, concentration of defects participating in the recombination processes, and instrumental response. Calibrations measurements carried out with standard mercury pen-lamp yields shifts of 0.02 eV near 250 nm, of 0.01 eV near 350 nm, and of 0.005 eV near 550–600 nm. Although Gao et al.⁴⁷ did not associate the higher energy emission band (∼3.6 eV) with a specific point defect, they tentatively assigned the emissions at ∼2.8 and ∼3.0 eV to gallium vacancies (V_Ga) related defects, and the ∼3.3 eV emission to oxygen vacancies (V_O), based on individual band intensity variation after specific processing treatments.

Luminescence is a well-established, highly sensitive, noninvasive, and nondestructive technique to detect and identify native and impurity related point defects and their complexes in semiconductors. Characterization by luminescence involves the measurement and interpretation of the spectral distribution of recombination radiation emitted by the samples. Generated electrons and holes usually become localized or bound at an impurity or intrinsic defect before recombining, and the identity of the localized center that they were bound at can often be determined from the luminescence spectrum. Qualitative information about the crystal quality can be inferred from the efficiency and line width of near band edge emission spectra, and impurities can sometimes be identified based on the binding energies inferred from the spectral positions and free-to-bound transitions. In general, due to the presence of various radiative and/or nonradiative recombination processes competing for the generated electron-hole pairs, luminescence process cannot be conveniently used as a reliable method to measure defect concentrations. However, one can get quantitative energy differences that help identify involved defects as describe above.

The broad band line shape and peak intensity variations observed for different deposition conditions may indicate changes in the constituent components of the fits and, thus, different concentrations of the corresponding point defects being incorporated. The emission intensity reduction with increasing temperature, a commonly observed phenomena in semiconductors, is generally attributed to increasing competition of nonradiative processes and/or additional radiative processes emitting in different spectral regions for the electron and hole carriers excited by the incident e-beam. The low-temperature results will be discussed first in an attempt minimize the detrimental effect of unknown competing recombination processes.

In an attempt to obtain insights on the correlation between point defect incorporation and deposition conditions we compare the integrated intensities of constituent fitting components of the low-temperature CL spectra of the samples 0219 and 0215, and the samples 0209 and 0306. The samples 0219 and 0215 were deposited with same substrate temperatures, but sample 0215 deposition was carried out with higher FRR, which yielded the highest growth rate of 0.797 nm/s. The 5 K spectra of these two samples, highlighted in Figs. 7(a) and 7(b), show an intensity reduction of the sample 0215 CL spectrum as compared to that of sample 0219. Despite that, the integrated intensity ratios of the most intense line at ∼3.3 eV to the intensity of other fitting components are about the same for both samples. This result indicates that there is not a large variation in the relative concentrations of related point defects if the films are deposited at the same substrate temperature. Therefore, larger FRR seems to affect mostly the growth rate.

The samples 0209 and 0306 were synthetized with same flow rate ratio but sample 0306 deposition was carried out with higher substrate temperature, yielding a growth rate of 0.486 nm/s, which is smaller than that observed for sample 0209. The CL spectra of sample 0209 and 0306 are depicted in Figs. 8(a) and 8(b), respectively, and show an increase of the spectral intensity of the CL spectra of sample 0306. The integrated intensity of all individual fitting components of the CL spectrum of sample 0306 is enhanced relatively to similar fitting components of the CL spectrum of sample 0209. It is important to note that the lines at ∼3.6 and ∼3.3 eV increased by 80% and 40%, respectively, which result in a significant CL line shape change. In general, relatively smaller values of the integrated intensity of the constituent emission line ∼3.6 eV were observed for films deposited at 800 °C under smaller FRR values. We observe near 3:1 relative increase of this line integrated-emission intensity for films deposited with FRR of 5/100 at 900 °C in comparison to that deposited at 800 °C. Similar behavior, but with a 2:1 relative ratio, was observed for the line emission close to 3.3 eV. Note that the integrated intensities of all lines increase with increasing FRR and substrate temperature but the larger increases of the lines at 3.6 and 3.3 eV reflects their stronger dependence on these deposition conditions.

As observed in the low-temperature CL spectrum, the 297 K spectrum of sample 0215 (not shown) shows a spectral intensity reduction, as compared with that of sample 0219 (not shown), and similar integrated intensity ratios of the line at ∼3.3 eV to other fitting components. As a result, a similar CL emission line shape is again observed on the spectra for both samples. It is also observed that the spectral intensity contribution from the fitting components lines at ∼3.6 and 3.3 eV are reduced as compared with than that of the line at ∼3.0 eV, in contrast with their contribution observed in the low-temperature spectra, which may be consistent with the shallow bandgap localization of the levels involved in the recombination processes. The 297 K spectrum of sample 0306 (not shown), as observed on the low-temperature CL spectrum, shows an increased spectral intensity, as compared with the CL spectral intensity of sample 0209 (not shown). It is observed that the lines at ∼3.6 and 3.3 eV increase their contribution to the convoluted CL emission band by over fourfold and twofold, respectively. These increasing relative intensity ratios result in a sharper and different line shape emission band from that observed in the 297 K spectrum of sample 0209. This indicates that the high temperature deposition may reduce the deposition rate but favors the incorporation of defects associated to the emission lines at ∼3.6 and 3.3 eV.

Despite of the reduction of the total intensity of the broad luminescence band acquired at 297 K and increased FWHM, the integrated intensities of individual constituent emission bands show a dependence on deposition conditions similar to that observed for the low-temperature spectra. Fluctuations of the peak intensity and FWHM values yields scattered points in the plot of the integrated-emission band-intensities versus FRR. Not knowing nature line shapes, and capture cross sections of the competing recombination processes, and the potential effects of film morphology differences (namely, grain size, tilting, and rotation), and inter- and intra-grain defects competition, prevents us from acquiring reliable values of the defect concentrations and their activation energies. It should be also mentioned that an about 50% increase of the FWHM of the 2.8 and 3.0 emission bands on measurements acquired at 297 K was observed, as compared with values measured at 5 K, suggesting that these two defects have a strong electron-phonon coupling. In general, recombination processes involving carriers bound to highly localize levels behave differently from those related to emission involving shallow centers in that they result in broad luminescence bands caused by strong electron-phonon coupling. Such processes can be conveniently illustrated with a configuration coordinate model⁴⁸ where the equilibrium position of the ground and excited states are displaced by the strength of the electron-phonon coupling, the Huang-Rhys factor “S.” Larger “S” values result in wider emission bands and less resolved transitions associated with the various involved vibrational modes.⁴⁹ In the present case, the overlapping of broad emission bands introduce additional difficulties in obtaining the effective vibrational frequencies and the “S” values of individual deep levels. One additional caveat is the structural and morphological properties of the present sample set, which may affect the intrinsic values of these parameters.

Figure 9 shows the average of several current-voltage measurements obtained between neighboring contacts on each sample. Data from samples grown at 800 °C are plotted together in panel (a) and those from samples grown at 900 °C are shown in panel (b). Arrows are inlaid to describe the increase in the O₂/Ar flow rate ratio (FRR) during the growth of the samples. Interestingly, the difference in measured current for the three different O₂/Ar flow rate ratios at 900 °C span several orders of magnitude, while the variation of measured current as a function of O₂/Ar flow rate ratio in the samples grown at 800 °C is much lower. For both growth temperatures, we observe the measured resistance, obtained from the slope of each curve at low bias, to be correlated with the O₂/Ar flow rate ratio during growth. For the samples grown at 800 °C, we find that a higher O₂/Ar flow rate ratio correlates to a trend of decreased resistivity. Our previous work⁵⁰ reporting thermal annealing of bulk Ga₂O₃ substrates in N₂ and O₂ atmospheres showed that CL emission intensity and carrier concentration correlate with V_Ga (an intrinsic acceptor defect in Ga₂O₃) concentration measured by positron annihilation spectroscopy. Therefore, if the electrical transport is similar to that of bulk crystals, i.e., dominated by intra-grain conductivity V_Ga could be the dominant defect for the low FRR films grown at 800 °C. The situation is reversed for the samples grown at 900 °C, which were thicker and, thus, could have quite different grain characteristics from that of films deposited at 800 °C. For these samples synthetized at 900 °C, the opposite trend was observed with the increasing O₂/Ar flow rate ratio resulting in more resistive films. These results agree with our previous electrical assessment of LPCVD Ga₂O₃ grown at 900 °C.³⁵ We could hypothesize that the decrease of resistivity with the increased flow rate ratio was due to the incorporation of additional defects formed during the LPCVD process and not detected by any of the characterization techniques in this study. Because of the higher deposition temperature of these films, we can speculate that the grain boundaries have different characteristics, such as the incorporation of impurities or defects which would in turn dominate conductivity in these films. Rafique et al.^44–46 demonstrated that the structural, optical, and electronic properties of Ga₂O₃ films deposited on sapphire substrates have a strong dependence on offcut orientation of the substrates. Films deposited on substrates with off-axis angle of 6° toward $⟨11−20⟩$ of sapphire have larger individual grain size and strong in-plane orientation due to the presence of surface steps, which act as preferred incorporation sites for the Ga adatoms, suppressing the grains random orientation of films deposited on plane. As result, off-axis films have sharper XRD linewidth and improved carrier mobility, which are consistent with a reduction of structural defects. The films investigated in the present work were deposited on c-plane, which rule out grains orientation as the mechanism controlling the films electrical property. Thus, further investigations of a specially designed set of samples will be necessary to fully understand the mechanism behind the switch in trends due to the difference in growth temperature in this sample set.

The properties of monoclinic gallium oxide thin films synthetized using the LPCVD method under various oxygen volume percentages at two sapphire substrate temperatures were systematically investigated. Within the investigated growth windows, a maximum growth rate of ∼2.9 μm/h (0.797 nm/s) was obtained for an oxygen volume percentage of 4.8% with a growth temperature at 800 °C. The film growth rate tends to decrease as growth temperature increases, when other growth parameters are kept the same. X-ray diffraction measurements show that all films have a β-Ga₂O₃ structure with the (−201) orientation, and the thicker films deposited at higher oxygen partial pressure have improved crystalline quality. Polarized micro-Raman scattering spectra show that these films have the monoclinic crystalline structure, in full agreement with the XRD study. However, 360° polarized Raman measurements carried out in various films and reference bulk crystals indicated that these films consist of small grains of (−201) β-Ga₂O₃ having random in-plane orientations.

The low and room temperature CL spectra of these films are characterized by a broad emission band extending between 1.5 and 4.5 eV (∼825 and 275 nm) that can be conveniently fit with four constituent emission bands near 3.6, 3.3, 3.0, and 2.8 eV. Changing deposition conditions gave rise to large changes of the relative intensities of these overlapping constituent emission bands with particular emphasis to the samples deposited at 900 °C. We find that samples synthetized at 800 °C, at higher O₂/Ar flow rate ratio show a trend of decreased resistivity. However, we observed an opposite trend for samples deposited at 900 °C with the increasing O₂/Ar flow rate ratio resulting in more resistive films. We speculate that this decrease in resistivity with the increased flow rate ratio at higher deposition temperature is caused by the incorporation of additional impurities and defects in the inter-grain region. Addressing these issues will require a detailed study of a set of samples designed to isolate inter- and intra-grain contribution to film electrical property.

This work was partially supported by the Naval Research Office. A.L. Mock gratefully acknowledges the National Council for postdoctoral fellowship support.

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