Preparation of High-Thickness n⁻-Ga₂O₃ Film by MOCVD

Abstract

The homoepitaxial Si-doped Ga₂O₃ film prepared by metal–organic chemical vapor deposition (MOCVD) was reported in this paper. The film thickness reached 4.5 microns, a relatively high value for MOCVD. The full width at half maxima of the (002) diffraction plane of the film was 26.3 arcsec, thus showing high crystalline quality. The film showed n⁻-type properties with a doping concentration of 3.6 × 10¹⁶ cm⁻³ and electron mobility of 137 cm²/V·s. In addition, the element composition and stress state of the film were characterized and analyzed. This indicates that the MOCVD, supporting high-quality, high-precision epitaxy, is promising for Ga₂O₃ power devices.

1. Introduction

Ga₂O₃ is an ultra-wide bandgap semiconductor material with a bandgap of 4.9 eV, high ultraviolet–visible light transmittance and a theoretical breakdown electric field strength as high as 8 mV/cm, which gives it broad applications in optoelectronic devices and power devices [1,2,3,4,5]. Furthermore, large-sized Ga₂O₃ single crystals can be directly obtained by melt growth, which facilitates the homoepitaxy of high-quality Ga₂O₃ films, thereby greatly improving the device’s performance [6,7,8]. Common vapor deposition, such as halide vapor phase epitaxy (HVPE) [9], magnetron sputtering [10], MOCVD [11,12], mist-CVD [13], molecular beam epitaxy (MBE) [14], etc., can be adopted for the growth of Ga₂O₃ films, also accelerating the application of Ga₂O₃ devices. Currently, Ga₂O₃ power rectifiers are mainly based on Schottky barrier diodes (SBDs) and heterojunction diodes (HJDs) [5,15,16,17,18,19]. The fabrication of Ga₂O₃ SBDs and HJDs is based on drift layers with low doping concentration and high thickness since a higher drift-layer thickness is beneficial to significantly improve the breakdown voltage of the devices and thus improve the power figure of merit. In the preparation technology of Ga₂O₃ films, it is challenging to meet the above-two requirements simultaneously. At present, only a few institutions have achieved the production of high-thickness Ga₂O₃ epitaxial wafers through HVPE due to its fast growth speed [20]. Compared with HVPE, MOCVD has advantages in growth cost, accuracy, and repeatability, which have been demonstrated in previous research [21]. More importantly, Farzana et al. have proved the advantages of diodes made of Ga₂O₃ films grown by MOCVD in structural design flexibility and with resistance [22]. However, the preparation of Ga₂O₃ films with thickness up to several micrometers by MOCVD has not been reported.

In this study, to demonstrate the possibility of preparing high-thickness Ga₂O₃ films by MOCVD, the n-type β-Ga₂O₃ films with high thickness for SBDs and HJDs were grown by MOCVD, and the properties of the films were characterized.

2. Materials and Methods

2.1. Materials

The growth was performed on the double-sided polished (001) β-Ga₂O₃ substrates, purchased from Novel Crystal Technology, Inc. (Saitama, Japan). High-purity O₂ (5N) and trimethylgallium (TMGa, 6N) were used as the oxygen and gallium source for the MOCVD apparatus (Emcore D180), respectively. The doping source for the n-type β-Ga₂O₃ films is a mixed gas of SiH₄ and N₂ (SiH₄, 100 ppm). High-purity N₂ (5N) and Ar (5N) were used as carrier gas for oxygen source and TMGa, respectively.

2.2. Experiment

The substrates were ultrasonically cleaned for 5 min with acetone, ethanol, and deionized water, then dried with N₂ before being put in the MOCVD reaction chamber. The TMGa was pre-cooled to 1 °C in the stainless-steel bubbler. The growth started with pumping five times and inflating the chamber with N₂, and the rotation speed of the graphite tray was adjusted to 60 r/min. Then, the substrate temperature was uniformly raised from room temperature to 750 °C in 30 min. During growth, the substrate temperature and chamber pressure were maintained at 750 °C and 40 mbar, respectively. The flow rates of O₂ and N₂ were 600 sccm and 1700 sccm. The SiH₄ flow rate was 3.5 sccm. The Ar gas used as the carrier gas for SiH₄ and TMGa was set to 40 sccm and 60 sccm, respectively. The growth lasted 4 h. After the growth, the film was annealed for 30 min in an O₂ atmosphere at 900 °C and 20 mbar to partially eliminate defects such as oxygen vacancies. Finally, the chamber was cooled to room temperature using Ar as the purge gas.

2.3. Characterization

The crystal quality of the films was characterized by X-ray diffraction (XRD, Rigaku, Ultima IV, Tokyo, Japan, CuKα radiation, λ = 1.54 Å). The surface and cross-sectional morphology were observed by field emission scanning electron microscopy (FESEM, JOEL, JSM-7610, Tokyo, Japan) and atomic force microscopy (AFM, Veeco, PlainView, NY, USA). Energy dispersive X-ray spectroscopy (EDX, Oxford X-max 80, London, UK) with an energy range of 0.01–20 kV was used to determine the types of elements on the surface of the film. The state and content of elements on the surface of the film were analyzed by X-ray photoelectron spectroscopy (XPS, ESCALAB250, Waltham, MA, USA). The XPS instrument with a pressure lower than 5 × 10⁻⁸ Pa was used to acquire core-level spectra from the film. Monochromatic AlKα radiation (hν = 1486.6 eV) was employed. A Hall device (Accent, HL5500PC, Hertfordshire, UK) was used to measure the electronic properties of the films. Raman spectra were obtained by Raman spectrometry (Horiba, LabRAM HR Evolution, Paris, France, resolution: 0.65 cm⁻¹) with an excitation laser wavelength of 633 nm.

3. Results

3.1. Crystal Structure Analysis

The crystal structure of the films was characterized by XRD. Figure 1a shows the XRD 2theta-omega patterns of the film. The diffraction patterns of the film include only sharp (001), (002), and (004) diffraction peaks, which are almost identical to those of the substrate. It indicates the existence of pure β-Ga₂O₃ and a single preferred growth orientation along the <001> direction. Figure 1b shows the 2theta-omega patterns of (002) diffraction plane. It is worth noting that the diffraction peak of the film’s (002) diffraction plane has moved to a high angle compared to the substrate. The diffraction plane spacings can be calculated by the Bragg diffraction equation:

$$2d_{\left(002\right)}sin\theta =\lambda$$

(1)

In Equation (1), d is the diffraction plane spacing, λ is the X-ray wavelength (1.54 Å), and θ is the diffraction angle. It was calculated that the (002) diffraction plane spacings of the film and the substrate are 2.839 Å and 2.841 Å. The lattice constant c can be obtained by:

$$d_{002}=\frac{1}{\sqrt{\frac{2^2}{c^{2}si{n}^{2}103.8°}}}$$

(2)

The c values of the film and substrate are $c_{sub}=5.852 \AA$ and $c_{film}=5.848 \AA$, respectively, basically consistent with the standard lattice constant of β-Ga₂O₃ ($c=5.80 \AA$). Assuming the volume conservation of the unit cells, the lattice constant a/b of the film will be smaller than that of the substrate, resulting in two-dimensional tension of the film at the interface. The shrinkage of the lattice was caused by the substitution of Si atoms (radius = 0.41 Å) for Ga atoms (radius = 0.62 Å). The rocking curve of the (002) diffraction plane is shown in Figure 1c. The full width at half maximum (FWHM) of the film is 26.3 arcsec, which is slightly lower than the 28.5 arcsec of the substrate, suggesting a smaller divergence angle and a better orientation of crystal planes. In addition, the dislocation density can be roughly calculated by the following relation [23]:

$$D=\frac{B^2}{2ln2\pi b^2}$$

(3)

where B is the diffraction peak FWHM, D is the dislocation density, and b is the Burgers vector. The magnitude of the Burgers vector is about −10, resulting in a dislocation density of 10⁵~10⁶ cm⁻², which was comparable to those of the substrate. The FWHM of β-Ga₂O₃ films prepared by MOCVD is smaller, compared to 27.4 arcsec by HVPE [24], 47.0 arcsec by low-pressure chemical vapor deposition (LPCVD) [25], and 38.9 arcsec by mist-CVD [13], implying a lower dislocation density.

3.2. Raman Analysis

The Ga₂O₃ film was further investigated with Raman spectra. Before the measurement, the silicon substrate was first used for calibration. Figure 2 shows the Raman spectra of the substrate and the substrate/film template, respectively. There are 15 active vibration modes in Ga₂O₃ [26]. In Figure 2, 11 of the 15 Raman peaks can be observed. Among them, the Raman peaks in the range of 100–300 cm⁻¹ correspond to the translation and rotation of the Ga-O chain. The Raman peaks of 300–500 cm⁻¹ are derived from the bending vibration of Ga_Ⅰ(O_Ⅰ)₂ octahedron, whereas the stretching and bending of Ga_ⅠO₄ correlate to the Raman peak in the region of 600–800 cm⁻¹. The intensity of all Raman peaks of the template is increased as compared to the substrate. Among them, the FWHM of the Raman peak of the template at 200 cm⁻¹ is 6.8 cm⁻¹, lower than the 8.3 cm⁻¹ of the substrate, indicating the improvement of crystallinity quality. In addition, the Raman peaks of the template are slightly blue-shifted compared to the substrate, which is a result of the internal stress in the film [27]. However, the degree of blue shift is slight, suggesting an approximately low stress state of the film, which is consistent with the XRD results.

3.3. Surface Morphological Analysis

FESEM and AFM were used to examine the surface morphology of the film. The top-view SEM image of the film is shown in Figure 3a. Three elements—carbon, oxygen, and gallium—are detected on the surface with atomic ratios of 17.9%, 47.5%, and 34.6%, respectively, according to the EDX examination of the region within the dashed frame. It is inferred there are two sources of carbon in the films; one is caused by the decomposition of carbon atoms by methyl radicals generated by TMGa at high temperature, and the other is ambient carbon introduced during film measurement and sample preparation. Therefore, a longer growth time will exacerbate carbon fouling, although carbon contamination can be reduced by optimizing the Ⅴ/Ⅲ ratio and gas flow. The AFM picture of the 2 μm × 2 μm region is shown in Figure 3b. The root mean square (RMS) roughness is ~0.3 nm, lower than 1.5 nm by MBE [28], and 7.0 2nm by LPCVD [29]. The height difference between the steps is less than 2 nm, according to the analysis result of the height profile image displayed in Figure 3c. Step flow growth is one of the common growth modes in homoepitaxy, as shown in Figure 3d. Atoms in the fluid are first adsorbed and then diffuse on the mesa. As a result of the lower nucleation work at the step than at the mesa, the molecules will nucleate at the step. A high molecular diffusion speed can lead to the formation of stable step-flow growth modes. The difference in molecular diffusion speed and the morphology of each step layer may cause step bunching. The cross-sectional SEM is shown in Figure 3e. Both the smooth surface and the clear boundary between the substrate and the film can be observed. The film thickness is about 4.5 μm, and the growth rate is estimated to be 1.0 μm/h. In contrast, the growth rate of Ga₂O₃ films through the mist-CVD is about 3.2 μm/h [13], while the HVPE is as high as 10 μm/h [20].

3.4. Elemental Composition Analysis

After a pre-sputter, using the Ar⁺ with an energy of 3 keV, the film was measured by XPS. Figure 4 shows the XPS spectra of the film. The signal peaks of Ga2s, Ga2p, Ga3s, Ga3p, Ga3d, O1s, O2s, and C1s, as well as the Auger electron signal peaks of O and Ga, can be observed in a range from 0 to 1400 eV in Figure 4a. The binding energy was calibrated with the standard value of C1s of 284.8 eV [30,31]. It is found that the carbon element content is 10.3% by the analysis, much lower than that by EDX. This indicates that the surface treatment is beneficial to reduce the contamination from external impurities. Fine spectra were performed on the peaks of Ga2p and O1s. The binding energy peaks of Ga2p are shown in Figure 4b. The binding energies of Ga2p3 and Ga2p1 are 1117.73 eV and 1144.58 eV, which are close to the binding energy of Ga2p [32]. Figure 4c shows the binding energy of O1s. The asymmetric O1s binding energy signal peak can be divided into two peaks by Gaussian fitting. The stronger peak at 530.78 eV is attributed to the Ga₂O₃ lattice, while the weaker peak at 532.08 eV may be derived from C-O [33,34]. According to the integral area of the two characteristic peaks of Ga2p and O1s and Equation (4), it can be calculated that the ratio of O and Ga atoms is approximately 1.46, which has a slight deviation from the stoichiometric ratio of O and Ga atoms in pure Ga₂O₃.

$$R_{O:Ga}=\frac{A_O/S_O}{A_{Ga}/S_{Ga}}$$

(4)

In Equation (4), R is the atomic ratio, A is the integral area of the characteristic peak, and S is the sensitivity factor of the corresponding characteristic peak. To clarify the state of carbon in the film, the C1s peak was scanned precisely, as shown in Figure 4d. It can be observed that the C1s peak can be fitted to two peaks with the extreme point at 284.7 eV and 286.8 eV, which may be attributed to C-H and C-O, respectively. This indicates that partial carbon elements exist in the form of C-O, consistent with O1s peak analysis, while another part of the carbon elements originates from C-H, which is probably related to other by-reaction products. Therefore, if the film surface is completely cleaned, there are two sources of carbon contamination in the film, C-O and C-H. In addition, the valence band edge was measured to estimate the Fermi level, as shown in Figure 4e. It can be obtained by linear extrapolation that the valence band maximum (VBM) is 3.3 eV higher than the Fermi zero point. Concerning the value of 4.9 eV for the bandgap of Ga₂O₃, the Fermi level is about 1.6 eV below the conduction band minimum (CBM). Therefore, the film exhibits weak n-type properties, with a low electron concentration.

3.5. Electron Characteristics Analysis

A Hall test was performed on the film with a size of 1 cm × 1 cm by the Van der Pauw method to characterize the electrical properties. Through five repeated measurements, the average and standard error of the electron concentration are determined to be 3.6 × 10¹⁶ cm⁻³ and 4.0 × 10¹⁵ cm⁻³, respectively. The film has an average electron mobility of 137 cm²/V·s, comparable to previous reports [13,25,35,36]. Therefore, the electrical characteristics of the Si-doped film prepared using the MOCVD technique essentially fulfill the criteria for power diode fabrication.

Table 1. Parameters of Ga₂O₃ substrates and epitaxial wafers.

	FWHM (arasec)	Nd-Na (cm−3)	Thickness (um)	Growth Rate (um/h)
Substrate [37,38]	<100	1 × 1018~2 × 1019	~650	~
HVPE wafer [39]	80~125	1 × 1016~9 × 1016	5~10	3~4
Our work [11]	~30	3.6 × 1016~2 × 1019	4.5	1

shows partial parameters of the current industrially prepared Ga₂O₃ epitaxial wafer and substrates. Although the growth rate of MOCVD is lower than that of HVPE, it is found that the MOCVD has more advantages in optimizing the quality and electrical properties of films, which is helpful for the preparation of various devices and the improvement of device performance. Increasing the MOCVD growth rate of Ga₂O₃ films is the main direction for our next work.

4. Conclusions

In summary, the homoepitaxial n-Ga₂O₃ film with a thickness of 4.5 μm and an electron concentration of 3.6 × 10¹⁶ cm⁻³ was prepared by MOCVD. The XRD confirmed the existence of pure β-Ga₂O₃ and high crystalline quality of the film. The film surface exhibited a uniform and uniformly oriented step-flow growth mode. The ratio of O atoms to Ga atoms on the film surface was 1.46, close to the stoichiometric ratio of 1.5. The Raman spectra results of the film are consistent with the symmetry of β-Ga₂O₃. Moreover, it was found that the film was approximately in a stress-free state through analysis and calculation. Si-doped n⁻-type homoepitaxial Ga₂O₃ film with a high thickness and high crystal quality was obtained, which offers an effective method to fabricate the Ga₂O₃-based SBDs and HJDs.