AlGaN Ultraviolet Metal–Semiconductor–Metal Photodetectors with Reduced Graphene Oxide Contacts

: AlGaN semiconductors are promising materials in the ﬁeld of ultraviolet (UV) detection. We fabricated AlGaN/GaN UV metal–semiconductor–metal (MSM) photodiodes with two back-to-back interdigitated ﬁnger electrodes comprising reduced graphene oxide (rGO). The rGO showed high transparency below the wavelength of 380 nm, which is necessary for a visible-blind photodetector, and showed outstanding Schottky behavior on AlGaN. As the photocurrent, dark current, photoresponsivity, detectivity, and cut-off wavelength were investigated with the rGO/AlGaN MSM photodiodes with various Al mole fractions, systematic variations in the device characteristics with the Al mole fraction were conﬁrmed, proving the potential utility of the device architecture combining two-dimensional materials, rGO, and nitride semiconductors.


Introduction
The III-nitride semiconductors and alloys thereof have been studied extensively because of their wide and direct bandgaps, high breakdown voltages, high mechanical and chemical stabilities, and low reverse leakage currents. These promising properties have inspired the use of III-nitride semiconductors for many applications such as light-emitting diodes, laser diodes, photodetectors, high electron mobility transistors, field-effect transistors, Schottky diodes, and solar cells [1][2][3][4][5]. Among these applications, ultraviolet (UV) photodetectors that are usable in visible-to solar-blind conditions have drawn significant attention for use in missile plume detection, flame engine sensors, ozone layer monitoring, and secure intersatellite communication systems [6,7]. With a bandgap between those of GaN (3.4 eV) and AlN (6.2 eV) [8], AlGaN can be used in both visible-and solar-blind UV photodetectors without bulky filter systems or heavy cooling systems such as those required for Si-based photodetectors, thus reducing the footprint and the production cost of the devices [7]. Regarding the photodetector device architecture, the easy and cost-effective fabrication of metal-semiconductor-metal (MSM) photodiodes with two interdigitated Schottky electrode contacts on their planar semiconductor surfaces has attracted significant attention. Such devices present low dark currents, low noise, high response speeds, and no doped layers [9,10].
Another important factor of planar UV MSM photodiodes is the transparent conducting electrode [11]. The transparency of existing metals and metal oxides declines sharply with decreasing illumination wavelength [12]. Therefore, high transparency in the UV spectral region, high material conductivity, and linear Ohmic contact properties are essential for electrodes in UV photodetectors [7]. Graphene oxide (GO), a derivative of graphene, possesses significant potential as a conductive electrode material with UV transparency [13]. GO can be fabricated via inexpensive and facile processes while retaining graphene's properties. Furthermore, by adding thermal or chemical reduction processing [14,15], so-called reduced GO (rGO) can be obtained, which presents variable optical and electrical properties depending on the degree of the oxygen functional groups contained on the basal plane of the rGO. Recently, Zhu et al. and Li et al. demonstrated the potential of rGO electrodes in the Si photodiodes [16,17]. Pandit et al. studied rGO Schottky contacts on a GaN layer, presenting the dependency of the Schottky barrier height on the GO reduction temperature [18]. Hence, the combination of AlGaN and GO-based materials for UV-enhanced optical sensors seems to be an interesting and attractive topic for exploration. An rGO/AlGaN-based photodiode may provide an enhanced photo-to-dark current ratio and tunable cut-off wavelength depending on the Al mole fraction. Here, we investigated rGO/Al x Ga 1-x N MSM photodiodes with various Al mole fractions, x.
Measurements of the dark current, photocurrent, and photoresponsivity of the fabricated devices showed systematic correlations of these properties with the Al mole fraction in the AlGaN layer.

Materials and Methods
Three Al x Ga 1-x N/GaN (300 nm/1.5 µm) samples with different Al mole fractions were grown on (0001) sapphire substrates by metalorganic chemical vapor deposition. Before device fabrication, the samples were cleaned by ultrasonication in acetone, methanol, isopropyl alcohol, and deionized (DI) water. Then, the samples were cleaned with buffered oxide etchant (BOE), piranha solution (H 2 SO 4 :H 2 O 2 ), and, lastly, DI water. As shown in Figure 1, for device fabrication, first, the aqueous GO solution prepared by the modified Hummers method [19] is sprayed over the sample, which is laid on a hot plate with a temperature of 180 • C. The GO/AlGaN/GaN samples are quickly transferred to a furnace chamber for the thermal reduction of the GO films by heating at 800 • C for 10 min. Next, a probing metal pad of Ti/Au is deposited by an electron beam (E-beam) evaporation system, assisted by a lift-off process. An etch mask of 100 nm thick SiO 2 for rGO etching is deposited by E-beam evaporation and patterned as an interdigitated shape by conventional photolithographic techniques and BOE treatment. Finally, the rGO film is etched out by O 2 plasma treatment before the remaining SiO 2 is dissolved using BOE.  [18]. Hence, the combination of AlGaN and GO-based materials for UV-enhanced optical sensors seems to be an interesting and attractive topic for exploration. An rGO/AlGaN-based photodiode may provide an enhanced photo-to-dark current ratio and tunable cut-off wavelength depending on the Al mole fraction. Here, we investigated rGO/AlxGa1-xN MSM photodiodes with various Al mole fractions, x. Measurements of the dark current, photocurrent, and photoresponsivity of the fabricated devices showed systematic correlations of these properties with the Al mole fraction in the AlGaN layer.

Materials and Methods
Three AlxGa1-xN/GaN (300 nm/1.5 μm) samples with different Al mole fractions were grown on (0001) sapphire substrates by metalorganic chemical vapor deposition. Before device fabrication, the samples were cleaned by ultrasonication in acetone, methanol, isopropyl alcohol, and deionized (DI) water. Then, the samples were cleaned with buffered oxide etchant (BOE), piranha solution (H2SO4:H2O2), and, lastly, DI water. As shown in Figure 1, for device fabrication, first, the aqueous GO solution prepared by the modified Hummers method [19] is sprayed over the sample, which is laid on a hot plate with a temperature of 180 °C. The GO/AlGaN/GaN samples are quickly transferred to a furnace chamber for the thermal reduction of the GO films by heating at 800 °C for 10 min. Next, a probing metal pad of Ti/Au is deposited by an electron beam (E-beam) evaporation system, assisted by a lift-off process. An etch mask of 100 nm thick SiO2 for rGO etching is deposited by E-beam evaporation and patterned as an interdigitated shape by conventional photolithographic techniques and BOE treatment. Finally, the rGO film is etched out by O2 plasma treatment before the remaining SiO2 is dissolved using BOE. High-resolution X-ray diffraction (XRD) measurement was performed to determine the lattice parameters of the AlGaN films and thus obtain the Al mole fractions. The measurements of the photodetection properties were performed using a dark-room probe station system equipped with a High-resolution X-ray diffraction (XRD) measurement was performed to determine the lattice parameters of the AlGaN films and thus obtain the Al mole fractions. The measurements of the photodetection properties were performed using a dark-room probe station system equipped with a Keithley 2400 source meter, a Bentham SSM150Xe switching monochromator, and a calibrated optical power sensor. Figure 2 shows the high-resolution XRD measurements of the three AlGaN/GaN samples that were used to obtain the accurate Al mole fraction. Peaks corresponding to the lattice parameters of GaN and AlGaN are clearly shown in the figure. An enlarged view of the AlGaN peaks is shown in the inset of Figure 2. The lattice parameters (c) of the samples were calculated by the Bragg relation, c = λl

Results and Discussion
where λ is the wavelength of the X-ray radiation used for the measurement, θ B is the Bragg angle, and l is the Miller index. The Al mole fractions (x) of the samples were then calculated by Vegard's rule: where C AlN (4.9792 Å) and C GaN (5.1855 Å) are the bulk lattice constants of AlN and GaN, respectively, and δ c is the bowing parameter available from previous studies [20][21][22]. The Al mole fractions of the three samples were thus confirmed as 9.5%, 13.4%, and 17.9% (henceforth given in decimal form as 0.095, 0.134, and 0.179, respectively). Keithley 2400 source meter, a Bentham SSM150Xe switching monochromator, and a calibrated optical power sensor.  The GO film shows a very high transmittance over a broad wavelength range, as shown in Figure  3. A slight decrease in transmittance occurs after thermal reduction at 800 °C, and a gradual decrease occurs as the wavelength decreases from 350 nm, which is well matched with the previous report in the literature [23]. Regardless, the transmittances of GO and rGO are >90% even for wavelengths <300 nm. The Raman spectral investigation of the GO film on the AlGaN/GaN substrate showed clearly the primary in-plane vibrational mode of sp 2 -bonded carbon atoms (G peak) and the defect-induced breathing mode (D peak) [23]. The samples showed ID/IG peak intensity ratios of <0.8, indicating that the GO layers were successfully formed on the substrates and that defect generation was well suppressed, even during the reduction process. In addition, the sheet resistance of rGO in our study was in the range of 6 × 10 3 Ω/sq [24].  The GO film shows a very high transmittance over a broad wavelength range, as shown in Figure 3. A slight decrease in transmittance occurs after thermal reduction at 800 • C, and a gradual decrease occurs as the wavelength decreases from 350 nm, which is well matched with the previous report in the literature [23]. Regardless, the transmittances of GO and rGO are >90% even for wavelengths <300 nm. The Raman spectral investigation of the GO film on the AlGaN/GaN substrate showed clearly the primary in-plane vibrational mode of sp 2 -bonded carbon atoms (G peak) and the defect-induced breathing mode (D peak) [23]. The samples showed I D /I G peak intensity ratios of <0.8, indicating that the GO layers were successfully formed on the substrates and that defect generation was well suppressed, even during the reduction process. In addition, the sheet resistance of rGO in our study was in the range of 6 × 10 3 Ω/sq [24].

Results and Discussion
magnitude between the dark current and photocurrent implies a high photo/dark current contrast ratio of the devices. Slight increases in photocurrent with the increase of bias are observed in Figure  4. When defects in the active layer are involved, the trapping of photoexcited carriers might be influenced by the induced bias. The photocurrent under illumination is lower for samples with higher Al mole fractions, namely, in samples with higher energy bandgaps, which is caused by the lower excited carrier concentrations in the samples with higher Al mole fractions.   Figure 5 shows the spectral photoresponses of the samples with three different Al mole fractions from 300 nm to 410 nm. The spectral photoresponsivity (R) is calculated with the ratio of the measured output current for a given input optical power as follows [25]: where is the photocurrent, is the dark current, and is the incident optical power. The peak in photoresponsivity is shifted to lower wavelengths (i.e., higher energies) for the rGO/AlGaN samples with higher Al mole fractions. The samples show the peak photoresponses of 0.176, 0.143, and 0.111 A/W at the wavelengths of 335 nm, 325 nm, and 315 nm, respectively. Note  Noise-like dark currents of less than 10 pA are measured, indicating that the dark current is below the measurement limit of our measurement system. Under the 330-nm illumination and the bias voltage of −5.5 V, the photocurrent increases to 24 nA, 58 nA, and 70 nA for the samples with the Al mole fractions of 0.179, 0.134, and 0.095, respectively. The increase of more than three orders of magnitude between the dark current and photocurrent implies a high photo/dark current contrast ratio of the devices. Slight increases in photocurrent with the increase of bias are observed in Figure 4. When defects in the active layer are involved, the trapping of photoexcited carriers might be influenced by the induced bias. The photocurrent under illumination is lower for samples with higher Al mole fractions, namely, in samples with higher energy bandgaps, which is caused by the lower excited carrier concentrations in the samples with higher Al mole fractions.
Appl. Sci. 2018, 8, x FOR PEER REVIEW 4 of 7 magnitude between the dark current and photocurrent implies a high photo/dark current contrast ratio of the devices. Slight increases in photocurrent with the increase of bias are observed in Figure  4. When defects in the active layer are involved, the trapping of photoexcited carriers might be influenced by the induced bias. The photocurrent under illumination is lower for samples with higher Al mole fractions, namely, in samples with higher energy bandgaps, which is caused by the lower excited carrier concentrations in the samples with higher Al mole fractions.   Figure 5 shows the spectral photoresponses of the samples with three different Al mole fractions from 300 nm to 410 nm. The spectral photoresponsivity (R) is calculated with the ratio of the measured output current for a given input optical power as follows [25]:  Figure 5 shows the spectral photoresponses of the samples with three different Al mole fractions from 300 nm to 410 nm. The spectral photoresponsivity (R) is calculated with the ratio of the measured output current for a given input optical power as follows [25]: where I photo is the photocurrent, I dark is the dark current, and P inc is the incident optical power. The peak in photoresponsivity is shifted to lower wavelengths (i.e., higher energies) for the rGO/AlGaN samples with higher Al mole fractions. The samples show the peak photoresponses of 0.176, 0.143, and 0.111 A/W at the wavelengths of 335 nm, 325 nm, and 315 nm, respectively. Note that the difference in the peak wavelengths can be associated with the difference in the energy bandgaps of Al x Ga 1-x N. The energy bandgap of AlGaN can be calculated by linear interpolation between the theoretical bandgap values of GaN (3.43 eV) and AlN (6.2 eV) [26,27]. Energy bandgaps of 3.69 eV, 3.8 eV, and 3.93 eV are obtained for the samples with x = 0.095, 0.134, and 0.179, respectively. These values show high correspondence with the peak photoresponsivity wavelengths in Figure 5. The photoresponsivity curves show gradual decreases for further decreasing wavelengths, which is because the photoresponsivity is proportional to the wavelength [5]. All of the experimental results in Figure 5 agree with the theoretical prediction that is expected for the AlGaN UV MSM photodiodes with various Al mole fractions. Small kinks (only visible in a logarithmic-scale plot of Figure 5) in the photoresponsivity curves were also observed near the wavelengths corresponding to the energy bandgap of GaN, which is possibly because of absorption by the underlying GaN layer, as in the previous report by Kumar et al. with a graphene/AlGaN Schottky diode [28]. As another figure of merit of photodetectors, the detectivity (D*) shows how much minimum signal can be detected by the photodetector. The detectivity is calculated by using the simplified equation [29]: where A is the junction area and q is the elementary charge. At the peaks in photoresponsivity, the MSM photodiodes with the Al mole fractions of 0.095, 0.134, and 0.179 showed the detectivity of 3.9 × 10 12 , 3.17 × 10 12 , and 2.46 × 10 12 cmHz 1/2 W −1 , respectively, which are values that are comparable to the literature [29].
Appl. Sci. 2018, 8, x FOR PEER REVIEW 5 of 7 that the difference in the peak wavelengths can be associated with the difference in the energy bandgaps of AlxGa1-xN. The energy bandgap of AlGaN can be calculated by linear interpolation between the theoretical bandgap values of GaN (3.43 eV) and AlN (6.2 eV) [26,27]. Energy bandgaps of 3.69 eV, 3.8 eV, and 3.93 eV are obtained for the samples with x = 0.095, 0.134, and 0.179, respectively. These values show high correspondence with the peak photoresponsivity wavelengths in Figure 5. The photoresponsivity curves show gradual decreases for further decreasing wavelengths, which is because the photoresponsivity is proportional to the wavelength [5]. All of the experimental results in Figure 5 agree with the theoretical prediction that is expected for the AlGaN UV MSM photodiodes with various Al mole fractions. Small kinks (only visible in a logarithmic-scale plot of Figure 5) in the photoresponsivity curves were also observed near the wavelengths corresponding to the energy bandgap of GaN, which is possibly because of absorption by the underlying GaN layer, as in the previous report by Kumar et al. with a graphene/AlGaN Schottky diode [28]. As another figure of merit of photodetectors, the detectivity (D*) shows how much minimum signal can be detected by the photodetector. The detectivity is calculated by using the simplified equation [29]: * ~ (2 ) where A is the junction area and q is the elementary charge. At the peaks in photoresponsivity, the MSM photodiodes with the Al mole fractions of 0.095, 0.134, and 0.179 showed the detectivity of 3.9 × 10 12 , 3.17 × 10 12 , and 2.46 × 10 12 cmHz 1/2 W −1 , respectively, which are values that are comparable to the literature [29].

Conclusions
In summary, we successfully fabricated and analyzed rGO/AlGaN MSM photodiodes. After a facile fabrication of highly transparent rGO layers, excellent rectifying properties were obtained from