Highly efficient gate-tunable photocurrent generation in vertical heterostructures of layered materials

Abstract

Layered materials of graphene and MoS₂, for example, have recently emerged as an exciting material system for future electronics and optoelectronics. Vertical integration of layered materials can enable the design of novel electronic and photonic devices. Here, we report highly efficient photocurrent generation from vertical heterostructures of layered materials. We show that vertically stacked graphene–MoS₂–graphene and graphene–MoS₂–metal junctions can be created with a broad junction area for efficient photon harvesting. The weak electrostatic screening effect of graphene allows the integration of single or dual gates under and/or above the vertical heterostructure to tune the band slope and photocurrent generation. We demonstrate that the amplitude and polarity of the photocurrent in the gated vertical heterostructures can be readily modulated by the electric field of an external gate to achieve a maximum external quantum efficiency of 55% and internal quantum efficiency up to 85%. Our study establishes a method to control photocarrier generation, separation and transport processes using an external electric field.

Layered materials such as graphene have attracted considerable interest for possible applications in diverse electronic and optoelectronic devices^1–6, including transistors^7–13, photodetectors^14–18, ultrafast lasers¹⁹, polarizers²⁰, touch panels²¹ and optical modulators²². With its broad spectral absorption²³, high carrier mobility²⁴ and short carrier lifetime²⁵, graphene exhibits exciting potential for wideband, high-speed photodetection^14–18. However, the intrinsically weak absorption characteristics and small built-in potential in these graphene-based photodetectors have severely limited their external quantum efficiency (EQE) to the range of ~0.1–1% (refs 16,17). Also, to date, the design of graphene-based photodetectors has usually largely relied on a lateral metal–graphene– metal junction with a rather small photoresponsive active area near the graphene–metal contact, which is not ideal for efficient photon harvesting^14–18. Similar lateral metal–MoS₂–metal devices based on few-layer MoS₂ have also been reported with a relatively small photoresponsivity of ~0.1 AW⁻¹ (ref. 26). Here, we report highly efficient photocurrent generation from vertically stacked graphene–MoS₂–graphene and graphene–MoS₂–metal junctions.

Vertical graphene–MoS₂–graphene devices

Figure 1 presents a schematic illustration of a graphene– MoS₂–graphene vertical heterostructure device on a Si/SiO₂ substrate. The device fabrication procedures are described in the Methods (Supplementary Fig. 1). A confocal laser was used to generate electron–hole pairs in the MoS₂ layer, which can be separated by the asymmetric potential in the top graphene (Gr_T)–MoS₂ and bottom graphene (Gr_B)–MoS₂ junctions to produce a measurable photocurrent (Fig. 1a,b). The current–voltage (I_DS–V_DS) data obtained in the dark (blue line, Fig. 1c) and under 514 nm laser irradiation (red line, Fig. 1c) show a clear photoresponse in the vertical heterostructure, with an open-circuit voltage of ~0.3 V and short-circuit current of ~2 µA. In general, the photoresponse exhibits a rapid temporal response beyond our experimental time resolution of 50 µs (Supplementary Fig. 2), demonstrating that the photoresponse originates from photocarrier generation rather than any other extrinsic effects.

Figure 1. Photocurrent generation in vertical heterostructures of graphene–MoS₂–graphene.

a, Schematic illustration of the three-dimensional device layout. b, Schematic illustration of the side view of the device, with the semiconducting multilayer MoS₂ sandwiched between the Gr_T and Gr_B electrodes. Red and blue colours indicate electrons and holes, respectively. The silicon substrate can be used as a back-gate electrode with 300 nm SiO₂ as the dielectric layer. c, Experimental current–voltage characteristic of the vertical device in the dark (blue) and under illumination (red) by a focused laser beam (wavelength, 514 nm; power, 80 µW; spot size, 1 µm).

Single-gated graphene–MoS₂–graphene devices

Figure 2a,b presents an optical image and scanning electron microscope (SEM) image of a typical vertical heterostructure device. The 8 µm strip of Gr_B (red in Fig. 2b) is located under the MoS₂ flake (blue in Fig. 2b), and the Gr_T layer (yellow in Fig. 2b) is located directly above the MoS₂ flake to overlap with Gr_B to ensure vertical charge transport. The thickness of the MoS₂ flake was determined by atomic force microscopy (AFM) to be ~50 nm. The underlying silicon substrate functions as a back gate to modulate charge transport across the graphene–MoS₂–graphene vertical stack. The I_DS–V_DS measurements of the device under dark conditions at various back-gate biases (V_BG) indicate that the device exhibits n-type characteristics (Supplementary Fig. 3), which is consistent with the intrinsically n-type nature of the MoS₂ layer¹¹. Additionally, all the I_DS–V_DS curves in the dark, at variable gate voltages, pass through the origin (dark curve in Fig. 2c and Supplementary Fig. 3), indicating that there is no gate leakage that could interfere with photocurrent measurement. For the measurement of photocurrent, the device was tested under focused laser illumination at variable V_BG. A series of I_DS–V_DS plots show that the short-circuit current I_sc and open-circuit voltage V_oc can be readily modulated by V_BG (Fig. 2c), with both I_sc and V_oc increasing with decreasing V_BG (Fig. 2c, inset).

Figure 2. Field-effect modulated photocurrent generation in single-gated graphene–MoS₂–graphene heterostructures.

a, Optical image of the vertical heterostructure with a multilayer MoS₂ flake (~50 nm thick) sandwiched between the Gr_T and Gr_B electrodes. b, SEM image of the same device with the Gr_T (yellow), Gr_B (red) and intermediate MoS₂ layer (blue) labelled with different false colours. c, I–V characteristics of the device under laser illumination (on white dot in b) at V_BG varying from −60 V to +60 V in steps of 20 V. Inset: variation of I_sc and V_oc with V_BG. d, Scanning photocurrent images taken at gate biases between −60 V and +60 V under a 514 nm laser (excitation power, 80 µW; spot size, 1 µm). Dashed and solid lines indicate the edges of the Gr_B and Gr_T electrodes, respectively. e–h, Schematic band diagrams of the vertical heterostructure with zero (e), negative (f), positive (g) and large positive (h) bias on the silicon back gate. Blue dots, black circles and green arrows indicate electrons, holes and photons, respectively. i, EQE of the vertical heterostructure device as a function of back-gate voltage under 80 µW, 514 nm laser excitation. Inset: EQE of the device as a function of excitation laser power at V_BG = −60 V. j, Excitation laser power-dependent EQE of another graphene-MoS₂ (16 nm)-graphene device under various excitation wavelengths at V_BG = −60 V. k, Wavelength-dependent EQE (black line with squares), absorbance (red line with circles) and IQE (blue line with triangles) of the device at V_BG = −60 V under a focused laser power of 5 µW.

To further probe the photoresponse characteristics of the device, we conducted scanning photocurrent measurements under a confocal optical microscope. A series of scanning photocurrent (I_ph) images of the device were taken under short-circuit conditions at different V_BG between −60 and 60 V (Fig. 2d). The Gr_B and Gr_T electrodes are highlighted in the photocurrent images by dashed and solid lines, respectively. Unlike graphene-based lateral photodetectors, which have a rather small photosensing active area near the graphene– metal contact, the vertical heterostructure device clearly shows broad-area photocurrent generation throughout the entire area of the vertical graphene–MoS₂–graphene stack. The image taken at V_BG = +60 V shows a strong photocurrent in the area where Gr_T and Gr_B overlap, demonstrating that the photoresponsive area covers the entire vertical stack. With the gate voltage stepping in a negative direction, the overall photocurrent increases and the photoresponsive area also appears to expand beyond the area of overlapping Gr_B and Gr_T. In particular, the photocurrent image for V_BG = −60 V shows that the photoresponse can be observed from the entire MoS₂ flake. This observation demonstrates that the photocarriers generated outside the vertical stack can also contribute to the overall photocurrent and suggests that the minority carrier diffusion length in our devices is at least on the micrometre scale.

Overall, the ability to modulate the photocurrent with an external back-gate voltage indicates that it is possible to manipulate the photocarrier generation, separation and transport processes in such vertically stacked devices. These experimental observations can be explained by using the band diagrams of the vertical stack under various back-gate voltages (Fig. 2e–h, Supplementary Fig. 4). Before applying a back gate, the initial built-in potential is determined by the Schottky barrier height difference between the Gr_T–MoS₂ and Gr_B–MoS₂ contacts. Because the channel length (~50 nm) is much shorter than the total depletion length (~140–170 nm) near both graphene–MoS₂ contacts (Supplementary Fig. 5), the top and bottom Schottky barriers merge together to form a monotonic band slope across the entire stack. This band slope determines the separation and transport of the excited electrons and holes. In our device, the Schottky barrier height in the Gr_B–MoS₂ contact is higher than that of the Gr_T–MoS₂ contact, probably due to a p-type doping effect to the graphene by the substrate oxide²⁷. As a result, the band in the MoS₂ layer slopes downward towards the top contact with a built-in field that drives excited electrons to move towards the Gr_T (Fig. 2e).

Importantly, with the finite density of states in graphene and weak electrostatic screening effect^9–12, the Gr_B–MoS₂ Schottky barrier height can be effectively modulated by an external field applied through the back-gate electrode V_BG. With negative V_BG, the workfunction of the Gr_B increases and so does the Gr_B–MoS₂ Schottky barrier height, which increases the band slope in MoS₂ to further promote the electron–hole separation and enhance the photocurrent (Fig. 2f). On the other hand, a positive V_BG decreases the Gr_B–MoS₂ Schottky barrier and reduces the band slope in MoS₂ to suppress the electron–hole separation and reduce the overall photocurrent (Fig. 2g). In principle, when the positive V_BG is large enough, the Schottky barrier height of the Gr_B–MoS₂ junction can be decreased to be lower than that of the Gr_T–MoS₂ junction, thus inverting the band slope and switching the photocurrent into the negative regime (Fig. 2h). However, this switching effect was not observed in this device before dielectric breakdown under increasing gate voltage, but is readily observable in dual-gated devices (Fig. 3).

Figure 3. Field-effect switchable photocurrent generation in dual-gated graphene–MoS₂–graphene heterostructures.

a, Schematic illustration of the side view of the device, with a multilayer MoS₂ flake sandwiched between the Gr_T and Gr_B electrodes, the silicon substrate as a back-gate electrode with a 300 nm SiO₂ dielectric layer, and another graphene layer as the top-gate electrode (Gr_G) with a 60 nm HfO₂ gate dielectric layer. b, Optical image of a dualgated device. c, Transfer characteristics of a dual-gated vertical heterostructure device as a function of V_BG (−60 V to +60 V) at various V_TG (−6 V to +4 V) at V_DS = 100 mV. Blue and red arrows indicate the Dirac points of the Gr_B and Gr_T electrodes, respectively. d, Two-dimensional colour plot of the transconductance (dI_DS/dV_BG) of the dual-gated vertical heterostructure device. The Dirac points of the Gr_T and Gr_B electrodes are shown by solid and dashed lines, respectively. e–g, Schematic band diagrams of the dual-gated vertical heterostructure device with negative, zero and positive top-gate bias at a back-gate bias near the bottom graphene Dirac point (−30 V). The red dashed arrow in f indicates the recombination of electrons and holes. h, Schematic band diagram of the device with negative top-gate bias and more positive back-gate bias than the Gr_B Dirac point −30 V to further enhance the band slope and overall photocurrent. i, Scanning photocurrent images taken at V_BG = −30 V and variable V_TG (−6 V to +6 V) (under 80 µW, 514 nm laser excitation, with a spot size of 1 µm). Dashed and solid lines indicate the edges of the Gr_B and Gr_T electrodes, respectively. j, Scanning photocurrent images taken at fixed V_TG = −6 V and variable V_BG (−30 to +60 V).

Based on the photocurrent response and input laser power, we can determine the EQE of the device. The EQE (η) is defined as the number of carriers produced per photon, or

$$\eta =\frac{I_{\text{ph}}}{q\varphi }=\frac{I_{\text{ph}}}{q}\left(\frac{hv}{p_{\text{opt}}}\right)$$

where I_ph is the photocurrent, ϕ is the photon flux ( = P_opt/hv), h is Planck’s constant, v is the frequency of light, q is the electron charge and P_opt is the optical power²⁸. Our study shows that the EQE can be increased from ~6% at V_BG = 60 V to ~15% at V_BG = −60 V under excitation by 80 µW laser power (Fig. 2i). Furthermore, it is found that the EQE increases with decreasing excitation laser power and saturates at ~27% when the excitation laser power is <5 µW (Fig. 2i, inset). The EQE observed in these vertical graphene–MoS₂–graphene heterostructure devices is much higher than that of lateral metal–graphene–metal photodetectors (EQE ≈ 0.1%)¹⁴. This higher EQE can be largely attributed to more efficient photon absorption in the broad-area vertical heterostructure with multilayer MoS₂, as well as more efficient charge separation resulting from a much larger and tunable band offset. EQE values obtained at various excitation powers and wavelengths from another device (with 16-nm-thick MoS₂) show a similar trend, with the signature of saturation at a power of 10 µW or below (Fig. 2j). The decreasing EQE with increasing excitation power could be attributed partly to absorption saturation in MoS₂ (Supplementary Fig. 6a) and partly to the screening of a built-in electric field by the excited electrons in the conduction band of MoS₂. The wavelength-dependent EQE plot shows two peak EQE values at ~480 nm and ~600 nm (black line in Fig. 2k), which can be largely attributed to the absorbance peaks in MoS₂ (red line in Fig. 2k and Supplementary Fig. 6b) and is consistent with previous studies^{29, 30}. Taking the optical absorption of the MoS₂ layer into account, the maximum internal quantum efficiency (IQE) of this device is estimated to be ~85% (blue line, Fig. 2k).

Dual-gated graphene–MoS₂–graphene devices

To further probe the ability to tune the photocurrent with an external field, we created an additional top-gate electrode on top of the graphene–MoS₂–graphene stack, using 60 nm HfO₂ as the insulating dielectric layer and another layer of graphene as the transparent top-gate electrode (Gr_G)³¹ (Fig. 3a). The transfer characteristics of a typical dual-gated vertical heterostructure device (Fig. 3b) are shown in Fig. 3c. Each curve was obtained by sweeping the back-gate bias V_BG from −60 to +60 V at a fixed drain–source bias V_DS of 100 mV while stepping the top-gate bias V_TG from −6 to +4 V in steps of 1 V. The transfer characteristics exhibit n-type behaviour (dominated by n-type MoS₂) versus V_BG, with two kinks corresponding to the Dirac points of Gr_T (red arrow, Fig. 3c) and Gr_B (blue arrow, Fig. 3c). The Dirac points of the Gr_T (solid line) and Gr_B (dashed line) can also be seen in the two-dimensional image of the transconductance (dI/dV_BG) in Fig. 3d. Overall, the Dirac point of the Gr_B is only slightly shifted from V_BG = −32 V to −27 V by varying V_TG due to the weak coupling between the Gr_B and the top gate, while the Dirac point of the Gr_T can be greatly shifted from −60 V to +40 V by varying V_TG due to the much stronger gate coupling between the Gr_T and the top gate. When V_TG is small (~0 V), the Dirac points of both Gr_T and Gr_B almost match at V_BG ≈−30 V. With positive V_TG, the Gr_T Dirac point shifts to be lower than that of the Gr_B, and with a negative V_TG the Gr_T Dirac point shifts to be higher than that of the Gr_B.

Modulation of the Dirac points of the Gr_T and Gr_B can directly impact the amplitude and direction of the band slope in the MoS₂ layer, as shown in the band diagrams under different top- or back-gate voltages (Fig. 3e–h). Simulated band diagrams are shown in Supplementary Fig. 4. With the back-gate bias set near the Gr_B Dirac point (V_BG = −30 V), a negative V_TG increases the workfunction of Gr_T, resulting in a higher Schottky barrier in the Gr_T–MoS₂ contact than in the Gr_B–MoS₂ contact, with a downward band slope in MoS₂ towards Gr_B, which drives the excited electrons towards Gr_B to produce a negative photocurrent (Fig. 3e). When V_TG is small (around 0 V), the Gr_T and Gr_B have similar Dirac points around V_BG = −30 V; there is little band slope or driving force for the separation and transport of excited electrons and holes (Fig. 3f), resulting in negligible photocurrent. When V_TG is increased to a positive value, the band slope in MoS₂ switches its direction to slope downwards towards Gr_T to drive the excited electrons moving towards the Gr_T and produce a positive photocurrent (Fig. 3g). A sequence of scanning photocurrent images of the device taken under short-circuit conditions at V_BG = −30 V and different values of V_TG between −6 and 6 V clearly shows that the photocurrent can be readily switched from negative values (V_TG<0) to nearly zero (V_TG = 0), followed by positive values (V_TG>0) (Fig. 3i). With a negative V_TG that increases the workfunction of Gr_T, applying V_BG> −30 V reduces the workfunction of Gr_B, which can further increase the band slope in MoS₂ and enhance the negative photocurrent (Fig. 3h,j). Overall, within the dual-gated graphene– MoS₂–graphene heterostructure, the amplitude of the photocurrent can be greatly modulated and the polarity of the photocurrent can be completely switched by the external gate field. For example, the photocurrent obtained at V_BG = 60 V and V_TG = −6 V is more than 500 times larger than that at V_BG = −30 V and V_TG = 0 V.

In an alternative and probably more simplified picture, the generation and polarity of the photocurrent within the dual-gated vertical devices can be easily understood without considering the details within the graphene–MoS₂–graphene stack. In general, the polarity and amplitude of the overall photocurrent is largely determined by the field direction and field strength across the top and bottom gates, with the photogenerated electrons (holes) always moving towards the positively (negatively) biased gate electrode. Unlike conventional photodiodes in which the photocarrier separation and transport are dictated by the internal built-in potential and are barely changeable with any external factors, our study clearly demonstrates that, with the partial electrostatic transparency of graphene, the photocarrier separation, transport and photocurrent generation processes in the vertical graphene–MoS₂–graphene devices can be substantially manipulated by an external field.

Single-gated graphene–MoS₂–metal device

The above studies were based on a nearly symmetric graphene– MoS₂–graphene stack. We also explored an asymmetrical vertical heterostructure of graphene–MoS₂–metal (Ti) for photocurrent generation. With a graphene–MoS₂ Schottky contact and a nearly ohmic MoS₂–Ti contact¹¹, the asymmetric device could allow for further enhancement of the photocurrent. With the opaque metal electrode on the top of the stack, the device was fabricated on the glass substrate with a 100 nm transparent indium tin oxide (ITO) thin film as the back-gate electrode and a 30 nm Al₂O₃ thin film as the gate dielectrics, to allow for excitation laser illumination from the back side of the substrate to reach the graphene–MoS₂– metal stack (Fig. 4a,b, Supplementary Fig. 7).

Figure 4. Field-effect modulated photocurrent generation in a single-gated graphene–MoS₂–metal heterostructure.

a, Schematic three-dimensional illustration of the device layout. b, Schematic illustration of the side view of the device, with a multilayer MoS₂ flake sandwiched between a Gr_B electrode and a top metal electrode (Ti). The device was fabricated on transparent ITO on a glass substrate (as the back gate) with 30-nm-thick Al₂O₃ as the gate dielectric to allow illumination to transmit through the back side of the substrate to reach the graphene–MoS₂ contact. c, Optical image of a graphene–MoS₂–metal heterostructure device. d,e, Schematic band diagrams of the graphene–MoS₂–metal heterostructure at negative and positive back-gate voltages, respectively. f, Scanning photocurrent images taken at V_BG = −1, 0 and +1 V (80µW, 514 nm excitation, with a spot size of 1 µm). Dashed and solid lines indicate the edges of the Gr_B and top metal electrodes, respectively. g, I–V characteristics of the device under laser illumination (on red dot in c) with V_BG varying from −1 V to 1 V in steps of 0.5 V. The I–V curve obtained in the dark at V_BG = −1 V passes through the origin, indicating that gate leakage is negligible. h, Excitation laser power-dependent EQE under various excitation wavelengths at V_BG = −1 V. i, Wavelength-dependent EQE at a laser power of 5 µW.

In the vertical heterostructure of the graphene–MoS₂–metal (Ti) stack (Fig. 4c), the top MoS₂–Ti forms a nearly ohmic contact, while the Gr_B–MoS₂ contact forms a Schottky junction (Supplementary Fig. 4). Accordingly, the energy band in the MoS₂ slopes downwards towards the top contact, and the amplitude of the slope can be modulated by the potential applied through the back-gate electrode. A negative V_BG increases the workfunction of Gr_B and the Schottky barrier height of the Gr_B–MoS₂ contact (Fig. 4d), which induces additional band bending and increases the band slope in MoS₂ to promote electron–hole separation and photocurrent generation. A positive V_BG decreases the Schottky barrier height of the Gr_B–MoS₂ junction and reduces the band slope in MoS₂ to suppress the photocurrent (Fig. 4e). A sequence of scanning photocurrent images of the device taken under short-circuit conditions at variable V_BG are shown in Fig. 4f. The measurement at V_BG = −1 V displays a strong photocurrent in the overlapping area of the Gr_B and Ti electrodes. The photocurrent gradually decreases with increasing gate voltage in the positive direction. A series of I–V plots obtained under laser illumination show that both the open-circuit voltage and short-circuit current increase with negative V_BG and decrease with positive V_BG (Fig. 4g). EQE measurements obtained at various excitation powers and wavelengths show that the EQE gradually increases with decreasing power and saturates at a power of 10 µW or less (Fig. 4h), similar to the results for graphene–MoS₂–graphene devices (Fig. 2j). Importantly, a maximum photocurrent EQE of 55% (corresponding to a photoresponsivity of ~0.22 AW⁻¹) is achieved at an excitation wavelength of 488 nm (Fig. 4i).

Discussion

In summary, we have shown that vertical heterostructures of graphene– MoS₂–graphene and graphene–MoS₂–metal stacks can be created to obtain highly efficient photocurrent generation and photodetection. We further demonstrate that both a top- and bottom-gate can be integrated within the vertical heterostructure to create dual-gated graphene–MoS₂–graphene devices, which allows us to use an external electrical field to modulate the amplitude or even completely reverse the polarity of the photocurrent in the vertical junctions.

Although gate modulation of photocurrent has been reported in lateral metal–graphene–metal junctions^14–17 and top-gate induced lateral p–n junctions in graphene devices³², it is important to note that the modulating field in such lateral devices is perpendicular to the charge separation/transport direction, and the external field modulates the band-bending and built-in potential in the lateral junctions to indirectly tune the charge separation and transport processes. In contrast, the external field direction in the vertical device is parallel to the current direction and can directly modulate the charge separation and transport. This capability is enabled by several unique characteristics of such vertical heterostructures, including (1) the optical transparency of graphene, (2) the partial ‘electrostatic transparency’ of graphene due to its finite density of states and weak electrostatic screening effect, and (3) the ultrathin nature of the layered semiconductor (MoS₂) sandwiched between. Within the unique architecture of the vertical heterostructure, the top and bottom Schottky barriers are merged together in the ultrathin semiconducting layer to form a monotonic band slope. This band slope dictates the separation and transport of the excited electrons and holes and is highly tunable by an external field across the dual-gate electrodes. This is in stark contrast to conventional photodiodes or the recently reported graphene–silicon junctions^{33, 34}, in which the photocurrent is dictated by the internal built-in potential. It should also be noted that the use of the external gate field to modulate the photoresponsivity in the vertical heterostructure is entirely different from directly biased photodiodes (metal–semiconductor or p–n diodes). For example, directly biasing the graphene–MoS₂–graphene stack or any conventional photodiodes would greatly increase the dark current and power consumption, whereas the gated devices in our studies operate at zero bias voltage, and the capacitively coupled external gate field does not generate any additional dark current. These unique features of the vertical heterostructures will open up exciting opportunities for the design of future photodetection and photovoltaic devices.

During the peer-review process of this manuscript, we became aware of another study on a similar topic³⁵. Our study was developed independently.

Methods

Fabrication of the vertical heterostructure devices

To fabricate the vertical graphene–MoS₂–graphene devices, graphene was grown using a chemical vapour deposition process^36–40 and transferred onto a Si/SiO₂ (300 nm SiO₂) substrate, then patterned into 8 µm × 30 µm strips to obtain a bottom electrode using photolithography and oxygen plasma etching processes. The MoS₂ flakes were then exfoliated onto the graphene strips through a micromechanical cleavage approach³⁸. The top graphene electrode was transferred and patterned onto the stacked MoS₂ flake and bottom graphene. The directly overlapping graphene area was patterned and etched away to avoid shorting between the top and bottom graphene. The metal electrodes (for probe contact or wire bonding purposes) were patterned on the bottom and top graphene electrodes by electron-beam lithography and electronbeam deposition of Ti/Au (50/50 nm) thin film. For the dual-gated heterostructure devices, a 60-nm-thick HfO₂ dielectric layer was deposited by electron-beam evaporation. The gate graphene electrode was transferred onto HfO₂ and patterned using electron-beam lithography. For the graphene–MoS₂–metal heterostructure devices, ITO glass with a 30 nm Al₂O₃ dielectric layer was used as the supporting substrate to allow illumination from the back side. The Al₂O₃ dielectric layer was deposited by an atomic layer deposition method.

Microscopic and optoelectrical characterizations

SEM imaging was performed on a JEOL 6700F unit operated at 5 kV. Electrical transport measurements were conducted with a probe station equipped with a computer-controlled analog-todigital converter at room temperature. The scanning photocurrent measurements were conducted with the same electrical measurement system under a SP2 MP Film confocal microscope coupled with an Ar/ArKr laser (wavelengths of 458, 476, 488, 496 and 514 nm) and a HeNe laser (543, 596 and 633 nm).