Deterministic Light-to-Voltage Conversion with a Tunable Two-Dimensional Diode.

Heterojunctions accompanied by energy barriers are of significant importance in two-dimensional materials-based electronics and optoelectronics. They provide more functional device performance, compared with their counterparts with uniform channels. Multimodal optoelectronic devices could be accomplished by elaborately designing band diagrams and architectures of the two-dimensional junctions. Here, we demonstrate deterministic light-to-voltage conversion based on strong dielectric screening effect in a tunable two-dimensional Schottky diode based on semiconductor/metal heterostructure, where the resultant photovoltage is dependent on the intensity of light input but independent of gate voltage. The converted photovoltage across the diode is independent of gate voltage under both monochromatic laser and white light illumination. In addition, the Fermi level of two-dimensional semiconductor area on dielectric SiO2 is highly gate-dependent, leading to the tunable rectifying effect of this heterostructure, which corporates a vertical Schottky junction and a lateral homojunction. As a result, a constant open-circuit voltage of ∼0.44 V and a hybrid "photovoltaic + photoconduction" photoresponse behavior are observed under 1 μW illumination of 403 nm laser, in addition to an electrical rectification ratio up to nearly 104. The scanning photocurrent mappings under different bias voltages indicate that the switchable operation mode (photovoltaic, photoconduction, or hybrid) depends on the bias-dependent effective energy barrier at the two-dimensional semiconductor-metal interface. This approach provides a facile and reliable solution for deterministic on-chip light-to-voltage conversion and optical-to-electrical interconnects.

Introduction

Two-dimensional (2D) semiconductors have been extensively studied in the past two decades, especially for electronics and optoelectronics.[1−6] The alignment between band gaps of different 2D semiconductors varies a lot,[7,8] thus heterojunctions with effective rectifying effect and photovoltaic effect could be readily built at the interfaces between different 2D semiconductors.[9,10] However, the construction of heterojunctions increases the complexity of the mass production of functional devices, and the inherent relatively high energy barriers at the interfaces limit the electrical conduction of device channels. An alternative to achieve ideal junctions without sacrificing electrical conduction is a homojunction formed with areas of the same material but in different states (doping, dielectric environment, etc.). For a certain 2D semiconductor material, the size and transition type (direct or indirect) of its band gap are highly dependent on its thickness,[11−13] giving us opportunities for building homojunctions with the same material of different thicknesses.[14] The 2D homojunctions based on thickness tuning demonstrate a weaker rectifying effect than heterojunctions, as the energy barrier formed at the interface of different thicknesses is finite, and the width of depletion region is always as short as nanometer scale.[15,16] To increase the energy band mismatch in 2D homojunctions and enhance rectifying effect, the uniform 2D semiconductors are locally doped through gate voltage or chemical doping or placed on ununiform dielectric substrates.[17−21] Another approach to locally modulate the doping level of a 2D semiconductor is charge redistribution, which happens when a 2D semiconductor contacts with a metallic material.[22,23]

Two-dimensional metallic materials have been successfully synthesized by chemical methods and employed to lower the Schottky barrier height (SBH) at the contact interfaces between 2D semiconductors and metallic electrodes.[24−29] In contrast, SBH at the contact interfaces formed with conventional metals (e.g., Au, Ni, Pt) that are deposited by evaporation and result in metal-induced gap states (MIGS) is always quite high.[8,30,31] In addition to facilitating charge carrier transport, the 2D metal–semiconductor interfaces could also induce Schottky junctions through charge redistribution,[22,32] demonstrating a high rectification ratio or photovoltaic effect with an external quantum efficiency (EQE) up to 55%.[32,33] The previously published works of 2D metal–semiconductor interfaces mostly involve lateral junctions at atomic scale or vertical junctions formed with ultrathin flakes.[25−29] Therefore, the electrical performance is highly tunable with gate voltage.[34] In a published computational work, the authors show that the stacking sequence of a vertical 2D graphene–MoS2 junction could significantly affect the electrical transport behavior[35] because the stacking sequence determines electronic density distribution in the device channel under a gate voltage. Furthermore, the thickness of graphene in a graphene–silicon heterojunction could heavily affect the tuning efficiency of charge density of silicon with gate voltage,[36] and ultimately, the graphene layer could almost totally screen the gating effect when it is thicker than 10 layers. Accordingly, the doping level of a 2D semiconductor could be locally modulated by charge redistribution when it interfaces with a 2D metallic material, and the formed energy barrier would be independent of gate voltage if the 2D metallic material is thick enough.

Here, gate-independent light-to-voltage conversion is achieved with a 2D semiconductor/metal heterostructure, demonstrating a gate-tunable rectifying effect. The deliberately stacked 2D semiconductor/metal heterostructure actually incorporates a Schottky junction and a homojunction. First, the results of gate voltage-dependent output I–V curves measurements and scanning photocurrent mappings (SPMs) indicate gate-independent photovoltage generation, resulting from effective dielectric screening at the 2D metal–semiconductor interface. Second, a highly gate-tunable diode is revealed with I–V curves, indicating that a homojunction diode is successfully built with the 2D semiconductor that is locally doped through charge redistribution after contacting the 2D metallic material. Finally, the bias-dependent transition of the photoresponse mechanism is illustrated with SPMs.

Results and Discussion

In this study, the device is fabricated with mechanically exfoliated 2D flakes. Briefly, metallic 1T-phase NbTe2 and semiconducting InSe are exfoliated, and thick flakes are transferred to a substrate of p-doped silicon covered with 300 nm thick SiO2. Thus, a heterostructure with NbTe2 on the bottom is formed (Figure a).[37] Followed by patterning and depositing Ti/Au electrodes, an Al2O3 layer is deposited to prevent the whole device from being damaged in the air, and potentially induces n-doping to the InSe flake.[38] As illustrated in the schematic of Figure a, the Ti/Au electrode deposited on InSe is grounded, and drain-source bias voltage VDS and gate voltage VGS are applied at the Ti/Au electrode on NbTe2 and the conductive silicon substrate on the back side, respectively. In the SPMs measurements, a 403 or 532 nm laser beam with a power of ∼1 μW is illuminated through a 20× objective above the device and scanned in the whole device area. An optical microscope image of the device is shown in Figure S1. The 2D materials involved in the device are identified with Raman and photoluminescence spectra, and the results are demonstrated in Figure S2. A cross-sectional schematic of the device is shown in Figure b, and the thicknesses of InSe and NbTe2 are ∼31.75 and ∼33.43 nm based on the measurement with an atomic force microscope (AFM, Figure S3). The special stacking sequence leads to strong electrostatic screening from thick metallic NbTe2 at the overlapping area,[35,36] but it is rarely observed in extremely thin 2D devices.[33,39] As a result, the InSe area on SiO2 is effectively doped by the charges (circles in Figure b) accumulated by VGS applied via conductive Si substrate, whereas the InSe area on metallic NbTe2 (dashed box in Figure b) is independent of VGS because the gating effect is screened by metallic NbTe2. The band alignment between InSe and NbTe2 is shown in Figure c. The Fermi level of n-doped InSe is higher than that of metallic NbTe2. Consequently, charge redistribution and band bending happen at the InSe/NbTe2 overlapping area after contact, leading to the formation of a Schottky junction (Figure d). A built-in electric field is formed in the Schottky junction. Thus, the photoexcited electron–hole pairs are efficiently separated at the interface. The separated electrons and holes drift to InSe and NbTe2, respectively. Due to dielectric screening, the built-in electric field and photocarriers separation are almost independent of VGS. In addition, the Fermi level (black dashed line in Figure d) of the InSe area on NbTe2 is lowered compared with its intrinsic Fermi level (red dashed line in Figure d) because of charge redistribution, and a significant Fermi-level mismatch (Φ in Figure e) is induced between the InSe areas on SiO2 and NbTe2 after stacking. This Fermi-level mismatch is readily tunable with VGS, which only tunes the Fermi level of InSe on SiO2. Based on this Fermi-level mismatch, a homojunction is assumed to form between the InSe areas on SiO2 and NbTe2 (Figure f). Since the Fermi-level mismatch is dependent on VGS, the energy barrier in this homojunction is tunable with VGS. Overall, this heterostructure incorporates a gate-independent InSe/NbTe2 Schottky junction and a gate-tunable InSe homojunction.

Figure 1

Device structure and band diagram. (a) Schematic of the InSe/NbTe2 heterostructure device. Metallic NbTe2 is placed on the bottom and overlaps with part of InSe, and an Al2O3 layer is deposited for protection. VGS: gate voltage. VDS: drain-source bias voltage. (b) Cross-sectional schematic of InSe/NbTe2 heterostructure. The circles with gradient colors indicate the opposite charges induced by VGS in 2D materials and the silicon substrate. Because of dielectric screening, the InSe area (dashed box) on NbTe2 is not doped by VGS. (c) Relative band alignment between InSe and NbTe2. CBM: conduction band minimum. VBM: valence band maximum. The dashed line indicates the intrinsic Fermi level (EF) of InSe. (d) Band diagram at the interface between InSe and NbTe2. Red and black dashed lines indicate the Fermi level of InSe before and after contact. Photoexcited electron–hole pairs are separated by the built-in electric field. VGS is assumed to have no effect on this Schottky junction, as the field effect is totally screened by the mobile charges in metallic NbTe2. (e) Fermi-level mismatch Φ between the InSe areas on SiO2 and NbTe2 at different VGS. (f) Band diagram of the homojunction formed between InSe areas on SiO2 and NbTe2. The height of the energy barrier in this homojunction is increased when VGS is switched from negative (ΦNeg) to positive (ΦPos).

To verify the dielectric screening effect illustrated in Figure b,d, output IDS–VDS curves are measured with the focused 403 nm laser spot of 1 μW illuminating the InSe/NbTe2 heterostructure area and are depicted with the absolute value of IDS (abs(IDS)) shown in log-scale in Figure a. These results are in stark contrast to the previously published works,[10,40−42] where both short-circuit current (ISC) and open-circuit voltage (VOC) are highly dependent on VGS, suggesting a unique photoresponse mechanism. VOC is almost fixed at 0.44 V and independent of VGS, while the ISC is significantly increased when VGS ranges from −80 to 0 V and almost saturates at positive VGS. The VGS-independent VOC is quite valuable for the application of on-chip light-to-voltage conversion,[43] where a stable photovoltage is desired under a certain light illumination condition. The photovoltaic effect could be optimized by carefully selecting a 2D metal with proper work function.[44] The gate-dependent ISC (Figure a) could be explained with the band diagram in Figure b, indicating that the excited electrons generated by the photovoltaic effect in the vertical InSe/NbTe2 Schottky junction (Figure d) drift in horizontal direction to the InSe channel on SiO2. The drifted electrons have a shorter mean free path when negative VGS is applied and the Fermi level of the InSe channel on SiO2 is pushed downward, as there are plenty of gap states (black bars in Figure b) over the Fermi level of InSe channel and many of the drifted electrons are trapped (gray arrows in Figure b) by the gap states. However, the trapping effect is significantly decreased, and ISC is retained at a high VGS when most of the gap states are filled with the accumulated charges. The assumption of gate-dependent ISC could be confirmed with the photoresponse to wide-field white light (Figure c) that illuminates the whole device area, which is the common illumination condition in practical photodetection and photovoltaics applications. The VGS-independent VOC is also verified with light illumination of different intensities (Figure S4). The major difference between focused laser illumination and white light illumination is the excitation of the InSe area on SiO2. Most of the gap states in the InSe area on SiO2 are filled because of the photogating effect under white light illumination, even if high negative VGS is applied. It is reasonable to conclude that the ISC is retained at a relatively high level when the InSe area on SiO2 is illuminated as well. Therefore, the low ISC at negative VGS under focused laser illumination results from the gap states, and an appreciable level of n-doping from the Al2O3 layer is beneficial for the photovoltaic effect in this device.[38] The dielectric screening effect is further investigated with SPMs at VDS = 0 measured under the illumination of a 532 nm laser. The results are demonstrated in Figure d–f. The mappings confirm the strong photovoltaic effect at the InSe/NbTe2 heterostructure, when the VGS is widely changed from −60 to +80 V. All of the above results in Figure indicate that the InSe/NbTe2 Schottky junction is gate-independent, as a result of strong dielectric screening effect. Thus, the photovoltage generated with this device is determined by the illumination condition and its inherent band alignment, and is highly resistant to electrical disturbance like gate voltage. The photovoltage generated under the illumination of modulated laser is shown in Figure S5. Based on the results, the response time is ∼0.33 ms, and this value can be further improved by optimizing the interfaces in the device.

Figure 2

Gate-independent light-to-voltage conversion. (a) Output IDS–VDS curves measured with the 403 nm laser spot of 1 μW focusing at InSe/NbTe2 heterostructure (inset). ISC: short-circuit current. VOC: open-circuit voltage. (b) Effect of VGS on ISC. Only the gap states (black bars) above the Fermi level (EF) of InSe, which is highly tunable with VGS (dashed lines in the left panel), could trap (gray arrows) photocarriers and lower the ISC. (c) Output IDS–VDS curves measured at different VGS under white light illumination (inset). (d–f) Scanning photocurrent mappings measured with 532 nm laser illumination of 1 μW and VDS = 0, when different gate voltages of VGS = −60 V (d), 0 V (e), and +80 V (f) are applied. Green, blue, and orange dashed lines outline the positions of InSe, NbTe2, and Ti/Au electrodes, respectively. Scale bars, 5 μm.

Subsequently, gate tunability of the heterostructure diode is investigated. The output IDS–VDS curves measured in the dark condition (Figure a) are highly dependent on VGS, as assumed before. Based on these curves, this device is in reverse and forward bias at negative and positive VDS, respectively. This trend agrees well with the band diagram in Figure d,f, indicating that a negative VDS applied on NbTe2 tends to increase the energy barrier height in both the InSe/NbTe2 Schottky junction and the InSe homojunction and switch the junctions to high resistance state, whereas a positive VDS decreases the energy barrier height and switch the junctions to low resistance state. The measured IDS at a reverse bias of VDS = −1 V is lower than 10–10 A, which is an extremely low value compared with other published 2D diodes,[20,40] indicating that effective energy barrier height and width are obtained in this heterostructure.[32] Furthermore, the IDS at positive VDS is highly tunable with V, as indicated by the rectification ratio (absolute value of the ratio between IDS at VDS = +2 V and VDS = −2 V) and IDS at VDS = +2 V demonstrated in Figure b. As the VGS shifts from −80 to +80 V, the rectification ratio is incrementally increased to nearly 104, and IDS at VDS = +2 V is amplified with a factor of >104. As discussed in the above section, the effect of VGS is effectively screened by NbTe2 (Figure b,d). Thus, the gate tunability of this device mainly arises from the InSe area on SiO2. As illustrated in Figure e,f, the Fermi level of the InSe area on SiO2 rises and the energy barrier height in InSe homojunction is increased, when VGS changes from −80 to +80 V. Simultaneously, the conductance of the InSe area on SiO2 is increased by high VGS that increases the density of mobile carriers in InSe. The synergistic modulation of energy barrier height and channel conductance leads to the modulation of on-state current in a wide range. The gate tunability is also investigated with SPM measurement at the same VDS of 0.4 V but different VGS, as shown in Figure c,d. The significant negative photocurrent is obtained at the InSe/NbTe2 heterostructure under the two different VGS, indicating the robustness of photovoltaic effect in the InSe/NbTe2 Schottky junction. Nevertheless, notable photocurrent when the laser beam is focused at the InSe area on SiO2 is only obtained at VGS = −60 V, while it is ignorable at VGS = +80 V. Overall, the gate-tunable photoconduction effect at the InSe area on SiO2 competes with the gate-independent photovoltaic effect in InSe/NbTe2 Schottky junction, leading to the distinct SPM patterns in Figure c,d. Obviously, this device works in “photoconduction + photovoltaic” hybrid mode and photovoltaic mode at VGS = −60 and +80 V, respectively. The dependence of photoresponse at the InSe area on SiO2 on VGS could be explained with the band diagram in Figure f. The energy barrier height in InSe homojunction is decreased at negative VGS; thus, the photocarriers generated by the photoconduction effect in the InSe area on SiO2 could be extracted more efficiently, leading to a more significant photocurrent at VGS = −60 V. Therefore, the InSe homojunction is highly gate-tunable in terms of rectification ratio, field effect current On/Off ratio, and photoresponse, in contrast to the gate-independent InSe/NbTe2 Schottky junction.

Figure 3

Gate tunability of the diode. (a) Gate-dependent IDS–VDS curves measured under dark condition. (b) Gate-tunable rectification ratio and IDS at VDS = 2 V. (c, d) Scanning photocurrent mappings measured with 403 nm laser illumination of 1 μW and VDS = 0.4 V, when gate voltages of VGS = −60 V (c) and +80 V (d) are applied. Green, blue, and orange dashed lines outline the positions of InSe, NbTe2, and Ti/Au electrodes, respectively. Scale bars, 5 μm.

As illustrated in the band diagrams of Figure d,f, the height of energy barriers in the InSe/NbTe2 Schottky junction and InSe homojunction is tunable with VDS. To reveal the dependence of the photoresponse mechanism on VDS, SPMs at gate voltage VGS = 0 are carried out and the results are demonstrated in Figure . At extreme bias voltage of VDS = 0–2 V, significant photoresponse is only observed at InSe/NbTe2 overlapping area (photovoltaic effect, Figure a) and InSe area on SiO2 (photoconduction effect, Figure d), respectively. This dependence of photoresponse mechanism on VDS is consistent with the previously published work.[37] In another published work, photovoltaic and photogating effects are also observed in a homogeneous MoS2 transistor with asymmetric metal electrodes.[45] Under a bias voltage of VDS = 0.4 V, which is quite close to the VOC = 0.44 V in Figure a, this device exhibits a hybrid photoresponse as shown in Figure b. In this case, positive and negative photocurrents are obtained when the laser beam illuminates the InSe area on SiO2 in proximity to NbTe2 and the InSe/NbTe2 overlapping area, respectively, indicating a hybrid photoresponse of coexisting photovoltaic and photoconduction effect at VGS = 0 V and VDS = 0.4 V. Once the bias voltage is much higher than VOC = 0.44 V in Figure a, such as VDS = 1.3 V in Figure c, positive photocurrent is observed in a broad InSe area on the substrate spanning SiO2 and NbTe2. Based on the band diagrams in Figure d,f, a substantially high positive VDS applied via NbTe2 could “flatten” the band bending in both the Schottky junction and homojunction. Consequently, the photovoltaic effect in InSe/NbTe2 Schottky junction almost disappears, and the InSe area on NbTe2 responds to light illumination in a photoconduction manner instead. At a higher bias voltage of VDS = 2 V, the Schottky barrier at the InSe/NbTe2 interface vanishes and Ohmic contact with low contact resistance is accomplished;[25,29] thus, the resistance of the whole device channel is determined by the resistance of InSe area on SiO2, as well as the contact resistance at Au–InSe interface results from MIGS.[30] Consequently, considerable photocurrent is generated only when the laser beam illuminates the InSe area on SiO2, especially the area close to the “source” electrode, which is a common case for 2D phototransistors with uniform channel. These results indicate that the InSe/NbTe2 heterostructure device transitions from a vertical Schottky diode to a lateral phototransistor when the VDS is increased from 0 V to a high forward bias of 2 V and works in “photovoltaic + photoconduction” hybrid mode when VDS is close to VOC under a certain illumination condition. In practical applications, the device could be configured to operate in photovoltaic or photoconduction mode, subject to the specific requirement (e.g., short response time, low dark current, high responsivity).[4]

Figure 4

Bias-dependent transition of photoresponse mechanism. (a–d) Scanning photocurrent mappings measured with 403 nm laser illumination of 1 μW and VGS = 0, when different bias voltages of VDS = 0 V (a), 0.4 V (b), 1.3 V (c), and 2 V (d) are applied. Green, blue, and orange dashed lines outline the positions of InSe, NbTe2, and Ti/Au electrodes, respectively. Scale bars, 5 μm.

Conclusions

In conclusion, deterministic light-to-voltage conversion is accomplished in a two-dimensional diode, which incorporates a gate-independent Schottky junction and a gate-tunable homojunction. The rectification ratio and field effect current on/off ratio of this diode are close to and over 104, respectively. The deliberate stacking sequence, where the semiconducting InSe is placed on metallic NbTe2, leads to so strong dielectric screening effect that the photovoltaic effect in this heterojunction is independent of gate voltage. This exceptional function is quite useful for light-to-voltage conversion in integrated photonic chips, and optical-to-electrical interconnects, as the output photovoltage is directly determined by the light input. In addition, the transition between photoconduction and photovoltaic effects in this device is revealed by scanning photocurrent mappings at various bias voltage VDS. In the future, the stable light-to-voltage conversion could be extended to infrared light with narrow-gap 2D semiconductors,[46,47] and this device could be further optimized with other promising 2D metals with optional work function.[22,48] The robust light-to-voltage conversion and flexible switching of operation mode demonstrated here are quite promising for the future on-chip optical-to-electrical conversion,[43,49] as well as multimodal photodetection and analogue integrated electronics.

Materials and Methods

Device Fabrication

The device is fabricated with the same method employed in our recently published article.[37] Thick flakes of NbTe2 and InSe are exfoliated from bulk materials and transferred to a Si wafer covered with 300 nm thick SiO2, assisted by polydimethylsiloxane (PDMS) stamps. The Ti/Au (5:100 nm) are patterned through electron beam lithography (EBPG 5000, Vistec Electron Beam, Germany), electron beam evaporation (MASA IM-9912), and lift-off in acetone. After annealing at 180 °C in high vacuum (∼10–5 mbar) for 2 h (AML-AWB wafer bonding machine), a 2 nm thick aluminum layer is deposited by electron beam evaporation, followed by atomic layer deposition (ALD, Beneq TFS-500) of 20 nm thick Al2O3. Finally, the second annealing process with identical conditions is conducted.

Optoelectronic Characterization

The fabricated device is connected to a printed circuit board (PCB) by wire bonding, and the PCB is fixed on a translation stage accompanied by a scanning near-field optical microscope (SNOM) (WITec, Germany). Gate and drain-source voltages are applied with Keithley 2400 and 2401 systems, respectively. A customized LabVIEW program controls the translation stage and Keithley systems. Light illumination is provided with a 403 nm laser (Toptica Photonics, Germany), a 532 nm laser (WITec, Germany), or the built-in white light source of the SNOM system, through a 20× objective. All of the measurements are conducted in the air.