A Highly Transparent β-Ga₂O₃ Thin Film-Based Photodetector for Solar-Blind Imaging

Abstract

Ultra-wide bandgap Ga₂O₃-based optoelectronic devices have attracted considerable attention owing to their special significance in military and commercial applications. Using RF magnetron sputtering and post-annealing, monoclinic Ga₂O₃ films of various thicknesses were created on a c-plane sapphire substrate (0001). The structural and optical properties of β-Ga₂O₃ films were then investigated. The results show that all β-Ga₂O₃ films have a single preferred orientation ($2(\_)$01) and an average transmittance of more than 96% in the visible wavelength range (380–780 nm). Among them, the sample with a 90-minute sputtering time has the best crystal quality. This sample was subsequently used to construct a metal-semiconductor-metal (MSM), solar-blind, ultraviolet photodetector. The resulting photodetector not only exhibits excellent stability and sunblind characteristics but also has an ultra-high responsivity (46.3 A/W) and superb detectivity (1.83 × 10¹³ Jones). Finally, the application potential of the device in solar-blind ultraviolet imaging was verified.

1. Introduction

Solar-blind photodetectors (SBPDs) with ultra-high sensitivity and light-to-dark ratios in natural backgrounds are an indispensable component of spectral photodetection. They play a unique role in various critical applications, including missile tracking, flame warning, secure communication, high-voltage corona detection, and environmental monitoring [1,2,3]. However, Si-based SBPDs based on low-cost proven technology exhibit low responsivity due to filter dependence and the low bandwidth of high-energy UV photon penetration [4]. Si-based SBPDs are unsuitable for high-temperature applications due to their narrow bandgap and severe thermal instability. Meanwhile, the newly emerged ultra-wide bandgap semiconductors, such as AlGaN [5], SiC [6], MgZnO [7], diamond [8], and Ga₂O₃ [9,10,11], have attracted considerable research attention as potential alternatives to Si-based SBPDs due to their high thermal and chemical stability and efficient absorption in the solar-blind region. However, the concentration of Al in AlGaN must exceed 40% to achieve a high energy band corresponding to that of solar-blind UV. This high Al content induces the generation of high-density defects, including disorder, dislocation, and grain boundaries. Moreover, the slow atomic migration of Al can produce strong parasitic reactions [1]. Similarly, Mg doping leads to phase separation between ZnO and ZnMgO, making achieving energy bands > 4.5 eV difficult. Meanwhile, the detection range for diamond is limited to a narrow radiation region generally below 225 nm, corresponding to a bandgap of 5.5 eV [8]. Hence, among the ultra-wide bandgap materials, Ga₂O₃ is the most ideal natural candidate for SBPD applications. Its ultra-wide bandgap (4.5–5.3 eV) corresponds directly to the solar-blind region and does not require doping or alloying for bandgap engineering. Moreover, its high absorption of high-energy UV photons facilitates increased detection and response in the solar-blind region [12,13,14]. Importantly, Ga₂O₃ exhibits structural stability at various temperatures, radiation sources, and electric fields. Among its five main phases: α, β, γ, ε, and δ [15,16,17,18], monoclinic gallium oxide (β-Ga₂O₃) is the most stable and belongs to the C2/m space with a bandgap of ~4.9 eV. Indeed, β-Ga₂O₃-based solar-blind UV detectors have been evaluated by many researchers [19,20,21].

Many β-Ga₂O₃ thin film growth techniques have been recently described, including molecular beam epitaxy (MBE) [22], metal organic chemical vapor deposition (MOCVD) [23], halide vapor phase epitaxy (HVPE) [24], atomic layer deposition (ALD), pulsed laser deposition (PLD) [25], low-pressure chemical vapor deposition (LPCVD), and radio frequency magnetron sputtering (RFMS) [26]. Among these, RFMS is a simple and affordable fabrication device with lower deposition temperatures and faster growth rates. Notably, RFMS can also grow large-area thin film materials, making it a mainstream deposition method. Jang et al. deposited β-Ga₂O₃ thin films by RFMS at different sputtering powers and found that the film growth rate increased linearly with increasing sputtering power, and the films at different sputtering powers had high crystalline quality and flat surface structures. They also prepared β-Ga₂O₃ thin film-based SBPDs with high light-to-dark ratios (>10³); however, the responsivity only reached the mA/W scale [13]. Nevertheless, numerous types of β-Ga₂O₃-based SBPDs have been developed, including metal-semiconductor-metal (MSM) [27,28], Schottky junction [12], p-n junction, and heterojunction devices [29]. MSM-type devices are widely used as they are simple to fabricate and compatible with conventional microfabrication processes. However, most MSM-type β-Ga₂O₃ photodetectors do not provide high responsivity and fast responsivity due to the intrinsic defects of β-Ga₂O₃. However, reducing the device area and regulating the crystal quality can effectively improve the device’s performance [2,27].

Herein, the response speed and responsivity of MSM-type devices are enhanced by preparing small interfinger pitch shrinkage and regulating the crystal quality. The crystalline structure and optical properties of β-Ga₂O₃ thin films at different sputtering times are then investigated. Subsequently, MSM-type photodetectors are constructed using β-Ga₂O₃ thin films with optimal crystalline quality optical properties, and systematic optoelectronic characterization is performed. The device was found to not only have a rapid response time but also demonstrate high responsiveness.

2. Materials and Methods

A commercially available GaO target was selected as the sputtering source for depositing amorphous Ga₂O₃ films onto c-plane sapphire (0001) substrates by RF magnetron sputtering. Before deposition, the c-plane sapphire substrates were ultrasonically cleaned in ethanol, acetone, and deionized water for 5 min each and then blown dry using high-purity nitrogen gas. During deposition, the sputtering chamber air pressure was pumped to below 2 × 10⁻⁴ Pa. The argon flow rate, working pressure, and sputtering power were set to 40 sccm, 2 Pa, and 150 W, respectively. The sputtering times were 30, 60, and 90 min for the three samples, designated S1, S2, and S3, respectively. The deposited amorphous Ga₂O₃ films were annealed in argon (Ar) at 700 °C for 2 h in a tube furnace to change the phase to β-Ga₂O₃. Finally, three pairs of inserted finger Ti/Au electrodes with a length of 500 μm and width of 40 μm were deposited on the surface of the β-Ga₂O₃ film to construct an MSM-type, solar-blind UV photodetector by magnetron sputtering. The spacing between the fingers was also 40 μm.

The crystal structures of the β-Ga₂O₃ films were determined using a Bruker D8 Advance X-ray diffractometer (XRD) with a Cu Kα line (λ = 0.1540598 nm), Bruker AXS, Germany. Raman scattering spectra were collected by a Raman spectrometer (LAB RAM HR Evolution) with a 532 nm laser excitation source at 50 mW power. Ultraviolet-visible (UV-Vis) absorption and transmission spectra were obtained using a Hitachi U-3900 spectrophotometer, Hitachi, Japan. The surface optical structures of the three samples were tested by Apreo2C electron scanning microscopy (SEM). The atomic force microscope (AFM) image was assessed by a Park NX10 SICM, Park, South Korea. The optoelectronic properties of the devices were measured using a Keithley 4200 source meter, Keithley, the United States. All characterization was carried out at room temperature.

3. Results

Figure 1a shows the XRD patterns of three annealed films with different sputtering times of 30, 60, and 90 min, denoted as samples S1, S2, and S3. All samples had distinct diffraction peaks throughout the plot, excluding the substrate peak. The peaks around 18°, 38°, and 58° corresponded to the ($2(\_)$01), ($4(\_)$02), and ($6(\_)$03) crystal planes of the β-Ga₂O₃ crystal [30]. Figure 1b plots the Raman patterns of the three samples. Low frequency (<200 cm⁻¹), medium frequency (300–500 cm⁻¹), and high frequency (500–800 cm⁻¹) peaks were observed for β-Ga₂O₃, respectively. In addition, the vibrational mode of 230.7 cm⁻¹ corresponded to the infrared vibrational mode Eu (TO/LO) (TO: mode of the transverse optical (TO) phonons; LO: mode of the longitudinal optical (LO) phonons) [13]. The three thin samples were all β-Ga₂O₃ films obtained by two steps: RF magnetron sputtering and post-annealing. To compare the effect of different sputtering times on the growth quality of the thin film material, a magnified comparison was performed for the peak position at 58° in the XRD pattern and the Raman characteristic peak at 202.1 cm⁻¹ in the Raman pattern (Figure 1c, d). The XRD diffraction peak and Raman feature peak intensities gradually increased with increasing RF sputtering time. Meanwhile, the full width at half maximum (FWHM) for both peaks decreased with increasing sputtering time (Figure 1e). The change in FWHM of the ($6(\_)$03) crystal surface also reflected the change in grain size of the films. According to the Scherrer formula, the grain size increases when the FWHM decreases:

$$D=\frac{K\lambda }{\beta cos\theta }$$

where D is the grain size, K is the constant (0.89), λ is the X-ray wavelength, β is the FWHM of the diffraction peak, and θ is the diffraction angle. The average grain sizes of the ($6(\_)$03) crystal plane of S1, S2, and S3 were 4.45 nm, 5.99 nm, and 7.4 nm, respectively. Hence, the RF sputtering time had a certain influence on the crystalline quality of β-Ga₂O₃ films; the sample with a sputtering time of 90 min (S3) had the best crystalline quality.

To further demonstrate the optical properties of the films, the absorption and transmission spectra of the samples with different sputtering times were evaluated. Figure 2a presents the transmittance spectra of the three β-Ga₂O₃ films in the UV−visible band region (240–800 nm). All three samples had an average transmittance of >96% in the 380–780 nm band, where spectral fluctuation was due to film interference. However, below 280 nm, the transmittance of the samples dropped sharply, indicating a strong absorption of deep UV light with an absorption energy corresponding to −4.8 eV (Figure 2b). For direct bandgap semiconductor materials, the optical bandgap (Eg) and absorption coefficient satisfy the Tauc formula [31]:

$${\mathit{\left(}\alpha h\nu \mathit{\right)}}^2=A\mathit{(}h\nu -Eg\mathit{)}$$

where α is the absorption coefficient, hν is the incident photon energy, and A is a constant. The S1, S2, and S3 bandgaps were estimated by extrapolating the formulae to yield 4.81 eV, 4.75 eV, and 4.73 eV, respectively, corresponding to the solar-blind UV band. In addition, the bandgap decreased with increasing sputtering time, which may be due to the combined effect of the Al elements diffusing in the substrate and the film thickness [13].

Figure 3a shows the cross-sectional SEM images of the S3 film. The thickness of the β-Ga₂O₃ thin film with a 90-minute sputtering time was 407 nm, with a film growth rate of ~4.5 nm/min at a 150-watt sputtering power. The thicknesses of S1 and S2 were estimated to be 135 nm and 270 nm, respectively. Figure 3b,c present SEM images of the surface and AFM with quantitative surface roughness characterization of the S3 film. The sample surface was flat (Rq = 1.3 nm) with a uniform and relatively large grain size, which effectively reduced the grain boundary defects and promoted carrier migration [32]. To investigate the optoelectronic properties of the β-Ga₂O₃ thin films, S3 was selected to prepare MSM-type photoconductive photodetectors. Photodetectors with three pairs of cross-finger Ti/Au electrodes were constructed using a simple mask process and sputtering technique (Figure 3d). The thickness of the Ti and Au layers was ~5 nm and ~100 nm, respectively. The inserted finger electrodes were 100 μm in width and 500 μm in length with 40 μm spacing. The effective irradiation area was 0.0012 cm².

Subsequently, we conducted a preliminary exploration of the optoelectronic characteristics of a β-Ga₂O₃-based SBPD. Figure 4a shows the semilogarithmic coordinate current-voltage (I–V) characteristic curves of the device under 254 nm and 365 nm illumination and dark conditions. The dark current of the device reached only 4 nA at a bias voltage of 5 V, which is lower than previously reported data [33]. The photocurrents under 365 nm light and dark conditions were relatively within the same order of magnitude; however, the rapid rise of the photocurrent reached 24 μA under 254 nm illumination, demonstrating the excellent solar-blind UV selectivity of the device. Figure 4b displays the time-dependent photoresponse curves of the device under 254 nm and 365 nm illumination. When the UV light was turned on, the photogenerated carriers accumulated instantaneously by non-radiative transition. Compared with the photoresponse at 254 nm illumination, that under 365 nm was negligible, which is consistent with the I–V characteristic curves in semilogarithmic coordinates. Figure 4c presents an enlarged view of the individual transient photoresponses of the device under 254 nm illumination. The 10–90% photoresponse time calculation law was used to determine the rise and fall times, calculated to be 2.82 s and 0.32 s, respectively. The I–V characteristic curves of the device were then investigated under various light intensities (Figure 4d). The photocurrent and voltage exhibited an obvious nonlinear relationship, indicating a back-to-back Schottky contact between the Ti/Au electrode and β-Ga₂O₃ film [23]. Figure 4e,f present the curves of the photocurrent and photo-to-dark current ratio (PDCR) of the device with a light intensity at a 5-volt bias voltage. The PDCR approximates the signal-to-noise ratio of the device [32] as follows:

$$PDCR=\frac{I_{photo}-I_{dark}}{I_{dark}}$$

where I_photo and I_dark denote the photocurrent and dark current, respectively. As the higher light intensity promotes the generation of electron-hole pairs, the photocurrent and PDCR of the device increase with increasing light intensity.

Next, the performance of the photodetectors was quantified. The responsivity (R) was defined as the photocurrent generated per unit power of incident light over the effective area of the photodetector and can be expressed as follows [17]:

$$R=\frac{I_{photo}-I_{dark}}{S·P}$$

where P is the light intensity and S is the effective illumination area. The curve of R versus light intensity is shown in Figure 5a. The maximum responsivity of the device was 46.3 A/W under 254 nm with a light intensity of 100 μW/cm². R decreased with increasing light intensity, caused by the increase in photogenerated carrier recombination activity at high light intensity [27]. In addition to PDCR and R, the external quantum efficiency (EQE) and detectivity (D*) were the main performance parameters of the optoelectronic device. EQE measures the ability of the device to convert an optical signal into an electrical signal; the magnitude of the EQE is the ratio of the electron-hole pairs to the incident photon, which can be expressed as follows [10]:

$$EQE=\frac{hc{R}_{\lambda }}{e\lambda }$$

where h is Planck’s constant, e is the charge, and λ is the wavelength of the incident light. D* denotes the ability of the device to detect weak signals and is described by the following equation [34]:

$$D^{*}=\frac{R{S}^{\frac{1}{2}}}{{\left(2q{I}_d\right)}^{\frac{1}{2}}}$$

where S is the effective illumination area and q is the charge. Figure 5b,c demonstrate the dependence of the EQE and D* of the device at different light intensities and a 5 V bias voltage. The EQE and D* of the device versus the light intensity were consistent with the relationship between R and light intensity, which both decreased with an increase in light intensity. The EQE and D* reached maximum values of 22,603 and 1.83 × 10¹³ Jones, respectively, under 254 nm light and 100 μW/cm² light intensity. To better understand the optoelectronic response characteristics of the device, Figure 5d shows the energy band schematic of the solar-blind UV photodetector. The energy band near β-Ga₂O₃ was found to bend downward to maintain the same Fermi energy level at the interface of Ti/Au and β-Ga₂O₃, creating a high-potential barrier region. The electrons flip over the potential barrier region in the form of hot electron emission, resulting in a symmetric nonlinearity in the I–V curves. Under 254 nm illumination, the conduction band converges, rapidly generating many electrons and photocurrents. Most electrons are obtained from the valence band directly jumping to the conduction band (process 4), while others are provided by the oxygen vacancies inside β-Ga₂O₃ and the related sizing defects (process 1), which together determine the photocurrent value. Certainly, large numbers of defects do not only provide electrons but also act as trap centers (process 3). When the 254 nm light is switched off, most conduction band electrons and valence band holes become rapidly and directly complexed, while a small number of electron-hole pairs undergo indirect complexation at the trap centers [17,32].

The performances of β-Ga₂O₃ thin film-based MSM-type SBPDs over the past two years are summarized and compared in

Table 1. Comparison of MSM-type β-Ga₂O₃ solar-blind UV photodetector performance in the past two years.

Sample	Growth	Idark (A/cm2)	R(A/W)	Τr(s)	Τd(s)	Ref
β-Ga2O3	PAD	~1.7 × 10−8	0.14	0.4	0.45	[25]
β-Ga2O3	MBE	2 × 10−7	1.18	4	24	[35]
β-Ga2O3: Ta	MOCVD	1.36 × 10−7	8.23	0.37	0.41	[36]
β-Ga2O3/Pt NPs	PLD	/	29.08	0.41	1.06	[37]
β-Ga2O3	RFMS	3.3 × 10−6	46.3	2.82	0.32	This work

. Due to the vacancy defect of β-Ga₂O₃, most β-Ga₂O₃-based photodetectors are not capable of providing high responsivity and a rapid photoresponse speed. Numerous studies have demonstrated that enhancing and equalizing device performance can be achieved by doping processes to equalize intrinsic defects and modify the membrane surface with nanoparticles. However, the production process for the device is relatively complicated, and using precious metals drives up the product’s price. Hence, the performance of solar-blind UV photodetectors can be markedly improved by implementing the simple magnetron sputtering technique and mask process described in the current study.

Finally, the imaging function of the solar-blind UV photodetector was investigated. Figure 6a shows a schematic diagram of the imaging system’s operation. A mask was placed between the laser source and photodetector, which was moved by a motorized positioning system. The current data from the source meter was collected in real time by a computer with each movement. The resulting “T” target at the computer terminal exhibited improved clarity (Figure 6b), suggesting the potential of the photodetector for solar-blind UV imaging and machine vision.

4. Conclusions

In summary, β-Ga₂O₃ thin films with different thicknesses were successfully prepared on c-plane (0001) sapphire substrates using RFMS and post-annealing techniques. Meanwhile, the best crystalline-quality film with a sputtering time of 90 min was selected to construct the MSM-type solar-blind UV photodetector. The results show that the β-Ga₂O₃ thin film-based solar-blind UV photodetectors not only exhibit superior solar-blind characteristics but also demonstrate high responsivity and relatively fast photoresponse times. Meanwhile, the application potential of the device in solar-blind ultraviolet imaging has been verified. This paper provides a favorable solution for the next generation of high-performance β-Ga₂O₃ thin film-based solar-blind UV photodetectors and their application.