Two-dimensional (2D) layered materials have received extensive attention due to their unique optical and electronic properties ^[1–10]. Molybdenum disulfide (MoS2), a typical member of the 2D transition-metal dichalcogenide materials family, has been the most investigated because of its potential applications in high-performance next-generation electrical and optoelectronic devices such as wearable flexible devices, image sensors, and communication devices ^[11–14]. Compared to the earliest studied graphene (zero bandgap), MoS2 has a stable band gap associated with the number of material layers varying from 1.2 to 1.8 eV; moreover, the monolayer MoS2 is a direct band gap (Eg=1.8eV) ^[15]. The MoS2-based photodetectors (PDs) have achieved significant progress due to their high carrier mobility (∼500cm2·V−1·s−1), high thermal stability (∼1100°C), and high absorbance of MoS2. Various methods for 2D MoS2 material preparation have been investigated such as mechanical exfoliation ^[16], electrochemical exfoliation ^[17,18], thermal decomposition ^[19], magnetron sputtering ^[20], and chemical vapor deposition (CVD) ^[21–23]. In 2012, an exfoliated single-layer MoS2 phototransistor was obtained for the first time, and the photoresponsivity was ∼7.5mA/W at 750 nm ^[16]. Ultrasensitive exfoliated monolayer MoS2 photodetectors were also demonstrated, showing a distinct photoresponse in the range of 400–680 nm, and the peak photoresponsivity was 880 A/W at a wavelength of 561 nm ^[24]. The 2D MoS2 obtained by mechanical exfoliation has a better material quality, but it is uncontrollable and small in size (micrometer scale). Electrochemical exfoliation is also a method of obtaining 2D MoS2 from top to bottom that can be used to produce larger-scale 2D MoS2, but with poor quality and problems like interfacial defects also occur when transferring to a target substrate. The CVD growth method has been the most widely investigated because of its excellent repeatability and large size.

SiO2/Si substrates were used for MoS2 CVD growth initially because they are perfectly compatible with Si-based optoelectronics ^[25,26]. However, it was found that it is easier to obtain high-quality MoS2 on crystalline substrates like sapphire and GaN ^[27,28]. These crystal substrates having excellent thermal and chemical stability can be utilized for the preparation of transparent devices. GaN is significantly superior to SiO2/Si and sapphire substrates because of its nearly perfect lattice matching ^[29,30] (0.8%, aMoS2=0.316nm ^[31], aGaN=0.319nm ^[32]) and the small discrepancy on the thermal expansion coefficients between MoS2 (aMoS2=4.92×10−6K−1) ^[31] and GaN (αGaN=3.95×10−6K−1 ^[32]). The almost perfect lattice matching and small difference in thermal expansion coefficient make it possible to grow a higher quality single-layer (SL) MoS2 on GaN. In addition, it has been demonstrated that the quasiparticle interaction will happen between the 2D MoS2 and GaN substrate, resulting in the modulation of the intrinsic valley carriers in SL MoS2, or strong electron-phonon coupling between MoS2 and GaN can lead to an increasing absorption coefficient of the MoS2 layer ^[30]. Figure 1 shows the calculated absorption coefficient of SL MoS2 in the visible region, in which the value of SL MoS2/GaN is significantly increased. As a typical representative of the third generation semiconductor, GaN-based materials own the advantages of a wide and direct band gap, good thermal stability, and chemical stability ^[33], making it one of the best choices for UV photodetector fabrication. High-quality 2D MoS2 grown on a lattice-matched GaN substrate has the potential to realize new concept high-performance photodetectors with the typical 2D/3D hybrid structure.

With the increasing demand for miniaturization, high integration, and multi-function photodetection, the investigation of 2D/3D multi-wavelength detectors has become a new focus. A hybrid 2D MoS2/3D GaN heterostructure for dual-functional (405 nm and 532 nm) detection was realized with photoresponsivity on the order of ∼102A/W ^[34]. Similarly, multilayer MoS2 grown on a free-standing GaN substrate also showed a multiple-wavelength photoresponse corresponding to 405, 532, and 638 nm, respectively ^[28]. Two-dimensional MoS2-based dual-wavelength or multi-wavelength photodetectors have promising applications in future optical communication and image sensors. However, previous research has mainly focused on visible light detection, and this work achieved dual-wavelength detection of the MoS2-based visible range and the GaN-based UV range. Monolithic integration of UV and visible dual-wavelength detection in this work could broaden optical communications by reducing the signal error rate.

In this work, large-scale SL MoS2 was grown on a lattice-matched and transparent free-standing (FS) GaN wafer by the CVD method. Common but efficient semiconductor fabrication processes were utilized to obtain monolithic integration of MoS2-based visible and GaN-based UV detectors. Both the materials and devices were fully characterized. Raman spectroscopy, transmission electron microscopy (TEM), and absorption spectroscopy were used to characterize the as-grown MoS2 material, proving that a high-quality SL MoS2 material was obtained. The photoresponsive properties of MoS2-based visible and GaN-based UV detectors under different incident light power conditions have been investigated. The photoresponsivity R, photoconductive gain G, external quantum efficiency (EQE), noise equivalent power (NEP), normalized detectivity D*, and the switching characteristics of the monolithic integration dual-wavelength detector were systematically studied in this work.

A (0001)-oriented Ga face FS-GaN substrate with thickness of ∼350μm was grown by the hydride vapor phase epitaxy (HVPE) method. Ammonia (NH3) and metal Ga/HCl were used as the nitrogen source and gallium source, respectively, and N2/H2 mixed gas was used as the carrier gas. After HVPE growth, it was further polished by chemical mechanical polishing (CMP) to obtain a surface with a surface roughness of 0.2 nm. A two-inch FS-GaN substrate was cut into pieces (1 cm by 1 cm) for the growth process, which is limited by the CVD chamber. The GaN substrate was ultrasonically cleaned with acetone, isopropanol, and deionized water for 10 min each time before being placed in the CVD chamber. 20 mg of molybdenum oxide (MoO3) was located in an alumina crucible with an FS-GaN substrate upside down on the center of the crucible. 8 g of sulfur (S) powder was laid upstream, and the alumina crucible was located near the center in 100 sccm of argon (Ar) gas flow. The tube was thoroughly cleaned by Ar gas flow to remove air before the growth process. The SL MoS2 was grown at 750°C for 10 min. The growth temperature has a significant effect on the thickness and shape of the MoS2 layer. The optimized growth conditions were selected in this study. After the SL MoS2 was grown on the GaN sample, a photolithography process was performed, and then part of the MoS2 was etched by the Ar plasma without the protection of the photoresist. Standard photolithography and electron beam evaporation (EBE) for Au electrode (50 nm) preparation were used to define the electrode.

The incident laser wavelength of the Raman spectra was 514 nm from an Ar ion laser with laser power of 0.025 mW to avoid laser-induced heat. The absorbance spectra were tested by a UV-visible spectrometer (Shimadzu-2450) from 350 to 800 nm. Current-voltage (I-V) characteristics of the devices were measured by a Keithley 4200-SCS semiconductor analyzer. All measurements of materials and devices were carried out at room temperature.

Figure 2(a) schematically illustrates the MoS2 growth process by CVD. S powder and MoO3 were used as precursors, while a Ga-face FS-GaN substrate (1cm×1cm) was placed upside down on the center of the alumina crucible. The MoS2 growth was completed in an Ar atmosphere with a growth temperature of 750°C for 10 min. The detailed growth process was described in the previous work ^[35]. The MoS2 grown on GaN was characterized, and the Raman spectrum was shown in Fig. 2(b). Raman spectrum measurement is based on the Raman scattering effect of the laser passing through the sample, reflecting the interlaminar vibration and intralayer vibration of the material, which is an important means for structural characterization of 2D layered materials ^[36]. For the Raman spectrum of 2D MoS2, only E2g1 and A1g optical modes can be observed, which represent the in-plane vibration modes of S and Mo atoms and the out-of-plane vibration modes of S atoms. In addition, the full width at half-maximum (FWHM) of the E2g1 peak characterizes the crystalline quality of 2D MoS2. The value of FWHM for peak E2g1 here is about 4.6cm−1, which indicates the high quality of MoS2. Although the FWHM here is slightly larger than that of mechanical exfoliated MoS2, large-scale continuous MoS2 on GaN was obtained by the CVD method in this work. The frequency difference (Δk) between the E2g1 and A1g Raman peaks is also important information for analyzing the thickness of MoS2. The positions of the E2g1 and A1g peaks are at 384.5 and 403.5cm−1, respectively, and the frequency difference Δk is about 19cm−1, corresponding to monolayer MoS2 which is consistent with other literatures ^[29,37,38]. An additional strong peak centered at around 418cm−1 was observed, which is in accord with the high-order harmonic frequency of an acoustic phonon of GaN, and is not a Raman active mode ^[39,40]. Figure 2(c) is the cross-sectional transmission electron microscopy (TEM) image of the MoS2 grown on GaN substrate. The measured MoS2 film thickness is ∼0.7nm, indicating a single layer of MoS2, which is consistent with the Raman measurement result. Figure 2(d) shows the absorbance of the SL-MoS2 grown on GaN as a function of incident wavelength. There are several absorption peaks in the visible region for SL MoS2: 430.9, 611.8, and 658.12 nm, respectively, and the absorption at 430.9 nm is the strongest, which is the exciton peak associated with the band edge. The excitonic absorption peaks between 600 and 700 nm observed for MoS2 arise from the K point of the Brillouin zone ^[41,42]. The spin-orbit splitting of the valence band of MoS2 results in an energy difference between the two exciton peaks. Absorbance is the quantification of the absorption intensity, which is defined as A=log10(I0/It), where I0 is the intensity of original light and It is the intensity of transmitted light ^[43]. According to the Beer–Lambert law, the relationship between It and I0 is described as It=I0e−αt, where t is the thickness of material and α is the absorption coefficient. The absorption coefficient α is calculated using the formula: α=ln(10A)/t. A is 0.15 at 405 nm, indicating 29.2% of incident light intensity is absorbed. The thickness of SL MoS2 is about 0.617 nm. The value of absorption coefficient α calculated in this work is 5.6×106cm−1, which is similar to the absorption coefficient of the multilayer MoS2 we studied before ^[28]. The obtained absorption coefficient α of the MoS2 grown on FS-GaN is about one order of magnitude higher than that of the 2D MoS2 grown on SiO2/Si substrate, which may result in a highly responsive MoS2 photodetector. The high absorption coefficient α in this work could be due to growth on a nearly lattice-matched and a small thermal expansion mismatch FS-GaN substrate, resulting in a high quality of MoS2.

The schematic diagram of the experimental process for monolithic integration of GaN-based UV detectors and MoS2-based visible detectors is shown in Fig. 3(a). A large-scale SL MoS2 was grown on the pre-cleaned FS-GaN substrate by CVD. Then the SL MoS2 on the FS-GaN sample was etched by Ar plasma to remove part of the MoS2 that was not protected by photoresist. After the photoresist was removed, an interdigitated Au electrode (50 nm) was prepared on the entire sample to obtain the monolithic integrated GaN PDs and MoS2 PDs by standard photolithography and metal evaporation. Figure 3(b) is the optical microscope image of monolithic integration of MoS2-based and GaN-based detectors, and the reference scale in the image is 100 μm. Each photodetector owns an active area of 47,376μm2. Figure 3(c) is the 3D schematic view of MoS2/GaN monolithic integration device. The top view photo image of the fabricated detector is shown in the inset. In this work, monolithic integration of GaN-based detectors and MoS2-based detectors was achieved in a feasible and low-cost way.

Figure 4(a) shows dark current and light current under 280 nm incident light with different incident power of GaN PDs as a function of voltage, while Fig. 4(b) shows that of MoS2 PDs under 405 nm laser illumination. Our study before has compared the three different laser wavelengths, 405, 532, and 638 nm. It was found that the MoS2 PD under the incident light of 405 nm has the best photodetector performance, which is because of a high-photocurrent gain mechanism and high absorption ratio ^[28]. For 280 nm UV light, it can be absorbed by GaN and then generate photo-generated carriers to achieve UV detection. Thereby, dual-color detection was achieved in this study. The ratio of Ilight/Idark for GaN PDs is about 101–102 at a bias voltage of 20 V. It shows as good response to UV light, which is comparable to the results of previous studies ^[44,45]. As for MoS2 PDs, the ratio of Ilight/Idark here is ∼100–101 at a bias voltage of 3 V.

The photoelectric characteristics of both the GaN-based UV detector and MoS2 visible detector are calculated as shown in Fig. 5. The responsivity (R), the quantification of the photosensitivity of PDs under certain wavelength light, is described as R=IphPA, where photocurrent Iph is calculated by Iph=Ilight−Idark, A is the active area of PD, and P is the power density of incident light. Figures 5(a) and 5(d) are photocurrent and responsivity of the GaN PD and MoS2 PD at 20 V and 3 V, respectively. The peak photoresponsivity of the GaN PD is 172.12 A/W under incident light power of 15 nW, while that of MoS2 PD is 17.5 A/W under 2 mW laser illumination. This rather high photoresponsivity may be associated with longer carrier lifetime and less carrier recombination in the 2D material under weaker illumination. As shown in Fig. 5(d), as the incident laser power decreases, the R of MoS2 PD increases slightly, which is in accord with previous reports in MoS2 and other 2D-based devices ^[46,47]. When the light power density increased enough, the trap states would be saturated, resulting in the photoresponsivity decrease ^[48].

Photoconductive gain (G) and EQE of the GaN-based PDs and MoS2-based PDs were also calculated as shown in Fig. 5(b) and Fig. 5(e), respectively. Photoconductive gain indicates the number of photocurrent carriers generated by absorbing each photon, which can be calculated by the formula G=Iph/eμP/hν, where e is the elementary charge, ν is the frequency of the incident light, h is the Planck constant, and μ is the absorption percentage. Iph has been calculated in Figs. 5(a) and 5(d). In theory, GaN can absorb light with energy higher than its band gap, so for 280 nm light, μ=1. The absorption percentage μ of SL MoS2 at 405 nm is 29.2%, which has been discussed before. The trend of photoconductive gain under different incident power is consistent with that of responsivity. For a GaN UV detector, the maximum photoconductive gain Gmax of 763.71 was obtained at a voltage of 20 V and an incident light power of 15 nW. The MoS2 visible detector achieved a Gmax of 183.88 at 3 V and an incident light power of 2 mW. EQE is the percentage of collected electrons to the total number of incident photons, which can be described as EQE=Iph/eP/hν=Rhνq. The maximum EQE of GaN UV detectors reached 76,370.1%, and for MoS2 visible detectors the maximum EQE is 5369%. Both the GaN-based UV detector and MoS2-based visible detector exhibit high photoelectric conversion efficiency.

To further evaluate the detectivity and detection limit of our devices, normalized detectivity (D*) and noise equivalent power (NEP) were calculated as shown in Figs. 5(c) and 5(f) for the GaN UV PD and the MoS2 visible PD. NEP can evaluate the signal-to-noise ratio more quantitatively, which is defined as the incident optical power at unit signal-to-noise ratio, described as NEP=PAS/N, where S is signal power and N is noise power ^[49]. In this work, we can assume that the signal current is the illumination current and the noise current is the dark current (Id). Therefore, NEP can be calculated by the formula NEP=(2eIdΔf)12/R ^[49], in which e is the elementary charge amount and Δf is the amplifier bandwidth. A small NEP means that the noise is small, indicating the better performance of the device. The normalized detectivity (D*), indicating the detection limit to detect the minimum optical signal, is calculated by the equation D*=A1/2/NEP. The larger D* means the stronger detection capability of the detector. For a GaN UV detector under 280 nm illumination, NEP is 8.09×10−15−4.25×10−14W/Hz1/2, and D* is 5.12×109−2.69×1010 Jones. For a MoS2 visible detector, NEP is 3.10×10−13−7.14×10−13W/Hz1/2, and D* is 7.02×1011−3.05×1011 Jones. The performance of the MoS2-based PD in this work is much lower than that of Si-based PDs (R of ∼300A/W and D* of ∼1013 Jones) ^[50,51], but it is better than the SL MoS2-based PD (R of 7.5 mA/W) ^[16]. This work achieved monolithic integration of GaN-based UV PD and MoS2-based visible PD for the first time to our knowledge in a simple process; however, the performance of the detectors is still not ideal, and further optimization of CVD growth details and device fabrication optimization are necessary.

Photo-switching characteristics have also been investigated for both GaN-based and MoS2-based PDs at a fixed voltage under 280 nm and 405 nm wavelength of incident light illumination, respectively. Figures 6(a) and 6(c) show the photocurrent of the GaN PD and MoS2 PD under periodic changes in dark and illumination conditions. When under-illuminated, the current rises to the on state and then falls to the off state under the dark condition. The switch ratio of devices is described as ONOFF=IlightIdark. For GaN-based UV PD, the ratio of photocurrent versus dark is about 27.4, while that of MoS2-based visible PD is about 13.5. The switching behavior is quite stable through multiple times of changing the illumination conditions. The rise time (trise) of the photodetectors is usually defined as the time it takes for the current to increase from 10% to 90% of the maximum, and the fall time (tfall) is the time for current decreasing from 90% to 10%. For a GaN UV detector, trise is 0.21 s and tfall is 0.22 s as shown in Fig. 6(b); for MoS2 visible detector, trise is 1.36 s and tfall is 8.31 s as shown in Fig. 6(d). trise and tfall obtained in this work are shorter than that of graphene-based devices ^[52], but are comparable to the reported SL MoS2-based PDs ^[24]. For MoS2-based PD, when the laser turns off, the photocurrent decays to the dark level exponentially. It is observed that the sustained photocurrent is because of the long-range Coulomb potential captured, which may be related to the charged impurities in the interface of MoS2/GaN or inside MoS2, therefore greatly affecting the recombination of electron-hole pairs. The proper device passivation or packaging technology can further improve the photocurrent switching performance of the device, which is also the next step for us to improve the performance.

In this work, the CVD growth method was used to obtain a high-quality and large-scale monolayer MoS2 on a nearly lattice matched free-standing GaN substrate. And by a low-cost and feasible process, monolithic integration of MoS2-based visible detectors and GaN-based UV detectors was realized. The responsivity of the UV detector reached 172.12 A/W under 280 nm illumination, while that of the visible detector reached 17.5 A/W under 405 nm laser. Meanwhile, high-photocurrent gain G, high external quantum efficiency, high normalized detectivity, and low noise equivalent power have been achieved for both two kinds of photodetectors. This study provides a low-cost, easy-to-process method for the preparation of dual-wavelength detectors with monolithic integrated UV and visible light, which has promising applications in image sensing and optical communication devices.

Acknowledgment. We thank Prof. Ke Xu from Suzhou Institute of Nano-Tech and Nano-Bionics, Chinese Academy of Sciences, for useful discussion and Prof. Hao-Chung Kuo from Taiwan Chiao Tung University. You Wu and Zhiwen Li contributed equally to this work.