Switchable Photoresponse Mechanisms Implemented in Single van der Waals Semiconductor/Metal Heterostructure.

van der Waals (vdW) heterostructures based on two-dimensional (2D) semiconducting materials have been extensively studied for functional applications, and most of the reported devices work with sole mechanism. The emerging metallic 2D materials provide us new options for building functional vdW heterostructures via rational band engineering design. Here, we investigate the vdW semiconductor/metal heterostructure built with 2D semiconducting InSe and metallic 1T-phase NbTe2, whose electron affinity χInSe and work function ΦNbTe2 almost exactly align. Electrical characterization verifies exceptional diode-like rectification ratio of >103 for the InSe/NbTe2 heterostructure device. Further photocurrent mappings reveal the switchable photoresponse mechanisms of this heterostructure or, in other words, the alternative roles that metallic NbTe2 plays. Specifically, this heterostructure device works in a photovoltaic manner under reverse bias, whereas it turns to phototransistor with InSe channel and NbTe2 electrode under high forward bias. The switchable photoresponse mechanisms originate from the band alignment at the interface, where the band bending could be readily adjusted by the bias voltage. In addition, a conceptual optoelectronic logic gate is proposed based on the exclusive working mechanisms. Finally, the photodetection performance of this heterostructure is represented by an ultrahigh responsivity of ∼84 A/W to 532 nm laser. Our results demonstrate the valuable application of 2D metals in functional devices, as well as the potential of implementing photovoltaic device and phototransistor with single vdW heterostructure.

Two-dimensional (2D) materials, including graphene,[1] black phosphorus,[2,3] transition metal dichalcogenides (TMDs), and their heterostructures,[4−8] have been extensively studied for optical and optoelectronic applications.[9−13] They possess bandgap ranging from 0 eV to over 2 eV,[14−17] which depends on not only the chemical composition but also the thickness of 2D flakes. The widely tunable bandgap gives rise to functional electronic devices like field effect transistors (FETs)[18,19] and facilitates ultrasensitive detection of the light ranging from visible to near-infrared.[20−22] Beyond single materials, many van der Waals (vdW) heterostructures integrating various materials have been successfully constructed by means of vdW stacking or chemical synthesis,[23−26] as ultraclean surfaces and edges are readily available with 2D materials.[27] In order to accomplish tailored functions and working mechanisms, the heterostructure devices have to be rationally designed based on band engineering.[28−30] Accordingly, it is worth exploiting more candidate materials with distinctive band structures.

Beyond the 2D semiconducting materials with deterministic electron affinity χ and bandgap Eg, the emerging 2D metallic materials, which have solely work function Φ that needs to be considered for band engineering design,[31,32] provide new options for the development of powerful devices. Specifically, group VB metal tellurides (XTe2, X = V, Nb, Ta) have been theoretically calculated to be metallic and experimentally obtained by chemical synthesis.[33−36] Quantitatively, ultrahigh electrical conductivity on the level of 106 S/m was achieved with the synthesized XTe2 flakes.[37] Owing to the high conductivity, NbTe2 was employed as conductive electrodes of WSe2 FET, resulting in lower contact resistance and higher carrier mobility compared with the counterpart using Cr/Au electrodes.[35] In contrast, the chemically synthesized lateral WS2/NbS2 (semiconductor/metal) heterostructure exhibits considerable rectifying effect,[38,39] which is a typical characteristic of diode-like heterojunction devices. The distinct transport behaviors of these vdW semiconductor/metal heterostructures are partly a result of the specific band alignment at the interfaces. Accordingly, it is promising to implement versatile functionalities at certain vdW semiconductor/metal interfaces designed based on rational band engineering.

Here, vertically stacked vdW semiconductor/metal heterostructure is designed and constructed by stacking mechanically exfoliated flakes of semiconducting InSe and metallic 1T-phase NbTe2. The electron affinity of thick InSe flake is extremely close to the work function of 1T-phase NbTe2, leading to exotic band alignment at the interface. The electrical and optoelectronic characterization, especially the photocurrent mappings, indicate that the vdW InSe/NbTe2 interface is switchable between heterojunction and ohmic contact, when the energy barrier at the interface is electrically tuned by bias voltage. Working as a semiconductor/metal heterojunction, it demonstrates exceptional photovoltaic effect with ∼0.41 V open-circuit voltage (VOC) and ∼380 nA short-circuit current (ISC) under 100 μW illumination of 532 nm laser. Switched to InSe phototransistor with ohmic contact to NbTe2 electrode, high photoresponsivity of ∼84 A/W is achieved under 10 nW illumination. These results reveal the versatile roles that 2D metallic materials could play in future optoelectronic applications.

Results and Discussion

In this study, InSe and NbTe2 are selected because of their band structures. According to the previously published experimental results, thick InSe flakes have a direct bandgap of ∼1.25 eV and electron affinity of ∼4.45 eV,[17,40] which is extremely close to the work function of metallic 1T-phase NbTe2.[41] The actual band structures of the two materials are measured with ultraviolet photoelectron spectroscopy (UPS), and the results are demonstrated in Figure S1. Accordingly, the energy bands of semiconducting InSe and metallic NbTe2 align as the illustration in Figure a. Owing to the proximity between the electron affinity of InSe and the work function of NbTe2, the band bending at this semiconductor/metal interface is expected to be highly adjustable. To experimentally investigate the InSe/NbTe2 heterostructure, a FET device was constructed on SiO2/Si substrate, as illustrated in the schematic of Figure b. In addition, an Al2O3 layer was deposited for protection of the device. The 300 nm thick SiO2 layer works as a dielectric layer, and the doped Si (G) on backside is used for applying gate voltage Vgate. Two Ti/Au electrodes deposited on InSe and NbTe2 are the source (S) and drain (D) electrodes, respectively. In all the following measurements, the source electrode was grounded, and bias voltage Vds was applied via the drain electrode.

Figure 1

Design and characterization of the InSe/NbTe2 heterostructure. (a) Band diagram between the bulk materials of semiconducting InSe and metallic 1T-phase NbTe2. The key features are determined with the results of UPS measurements shown in Figure S1. EF denotes the Fermi level of pristine bulk InSe, indicating that it is n-doped. (b) Schematic of the InSe/NbTe2 heterostructure device. The upper panel illustrates the crystal structures of InSe and NbTe2. The Al2O3 layer for protection is not shown in the schematic. (c) Optical microscope image of the stacked InSe/NbTe2 heterostructure. The top diagram illustrates the stacking order. (d) AFM characterization of the InSe/NbTe2 heterostructure. (e) Raman spectra of InSe, NbTe2, and InSe/NbTe2 heterostructure. (f) Absorbance of InSe/NbTe2 heterostructure.

In practice, the fabrication of InSe/NbTe2 heterostructure device started with mechanical exfoliation and vdW stacking of the 2D flakes (the details are described in Methods).[6,42] Briefly, bulk 1T-phase NbTe2 was first exfoliated, and the obtained flakes were transferred to SiO2/Si substrate, followed by transferring exfoliated InSe flakes on top. As InSe has a direct bandgap only for thick flakes and NbTe2 is difficult to be thinned,[17] both InSe and NbTe2 flakes in relatively thick form were used. Optical microscope images of the InSe/NbTe2 heterostructure and fabricated FET device are shown in Figure c and Figure S2, and the overlapping region has an area of ∼8 μm × 10 μm. Thickness of the flakes was measured by atomic force microscopy (AFM), and the line profiles in Figure d indicate that the InSe and NbTe2 flakes are ∼17 nm and ∼55 nm thick, respectively. The two flakes were further identified by Raman spectra. As shown in Figure e, Raman spectra were collected at both the single materials and the heterostructure. In the Raman spectrum of InSe, three peaks are observed at ∼116 cm–1, ∼178 cm–1, and 227 cm–1, attributed to Ag, E2g, and Ag2 phonon modes, respectively.[43,44] The Raman spectrum of NbTe2 flake consists of three prominent peaks (∼84 cm–1, ∼109 cm–1, and ∼158 cm–1) and two weak peaks (∼219 cm–1 and ∼254 cm–1, detailed in Figure S3). These results agree well with the published results of NbTe2 flakes grown by chemical vapor deposition and vapor phase transport.[3645] All the major peaks described above could be found in the Raman spectrum of InSe/NbTe2 heterostructure, indicating the high quality of fabricated vdW heterostructure. Furthermore, photoluminescence (PL) spectra of the flakes were measured, and the results are depicted in Figure S4. The obvious peak centered at photon energy of around 1.25 eV is a typical value of thick InSe flakes,[17] while the metallic NbTe2 does not show any PL peak. The significant darkness of InSe/NbTe2 heterostructure in Figure c is unexpected, and it is generally observed in other samples as well (Figure S5a–d). To quantitatively comprehend this observation, the reflection of these 2D materials on SiO2/Si substrate was measured, and the absorbance was calculated with a silver mirror as reference (Figure S5e). According to the calculated absorbance demonstrated in Figure S5f, the InSe and NbTe2 flakes mainly absorb light in the wavelength ranges of ∼600–700 nm and ∼450–600 nm, respectively. It means that InSe and NbTe2 are complementary for the absorption of visible light. Consequently, the absorbance of InSe/NbTe2 heterostructure is universally enhanced in the whole visible range (up to ∼80% at ∼500–550 nm), as depicted in Figure f. One reasonable explanation for the high absorbance is the joint absorption effect of the two thick flakes, as observed in another case of thick InSe/Te (40 nm/120 nm) heterostructure, whose overlapping area is significantly dark as well.[46] However, to the best of our knowledge, dark vdW heterostructures are rarely observed, even in the ones built with thick flakes. Therefore, this phenomenon is worth investigating by systematic and meticulous spectroscopic characterization. Despite the unknown cause, the high absorbance is expected to benefit the photodetection with InSe/NbTe2 heterostructure.

The working mechanism of the InSe/NbTe2 heterostructure device could be predicted based on the published works,[33,36,38] where 2D semiconductor/metal heterostructure devices with various materials exhibit distinctive transport behaviors. On the basis of computational study,[33] the work function of metallic NbS2 drops in the bandgap of WS2, whereas the work function of metallic NbTe2 is much lower than the valence band maximum of WSe2. As a result, an energy barrier with considerable height emerges because of band bending at the WS2/NbS2 interface, while it is absent at the WSe2/NbTe2 interface. These band diagrams lead to an apparent rectifying effect in heterogeneous WS2/NbS2 device and remarkably low contact resistance of homogeneous WSe2 FET with NbTe2 electrodes.[35,39] On the basis of these results, the InSe/NbTe2 interface is predicted to be switchable between heterojunction and ohmic contact, as the difference between the electron affinity of InSe and the work function of NbTe2 is so little.

Initially, electrical characterization was conducted to reveal the basic working mechanism of InSe/NbTe2 heterostructure, and the results are depicted in Figure . For comparison, the FETs with pure InSe or NbTe2 channel were fabricated and measured as well. The transfer curves shown in Figure a, as well as the gate dependent output Ids–Vds curves in Figure S6, suggest that the exfoliated InSe flake is intrinsically n-doped and highly conductive when it is heavily doped by positive gate voltage, and the NbTe2 flake is a metal with conductivity independent of gate voltage. In addition, the linear output Ids–Vds curves indicate ohmic contact between the flakes and Ti/Au electrodes. These results agree well with the published results of semiconducting InSe and metallic NbTe2, indicating the high quality of exfoliated InSe and NbTe2 flakes. In the following, the InSe/NbTe2 heterostructure FET was measured with the configuration illustrated in Figure b, and two different bias voltages Vds of 2 V and 1 V were applied. As shown in Figure b and Figure S7, the Ids could be considerably modulated by gate voltage, leading to a high current On/Off ratio of nearly 105. The high On current and On/Off ratio were maintained even 2 months after device fabrication (Figure S8). In contrast, the bare device without protective Al2O3 layer shows low On/Off ratio of <103 when it was measured right after fabrication (Figure S9). This comparison clearly indicates the protection effect of Al2O3 layer. Besides, the significantly nonlinear relation between the Ids and Vds at positive gate voltage verifies the existence of Schottky barrier at the InSe/NbTe2 interface. The Schottky barrier could be further confirmed with the output Ids–Vds curves in Figure c, which exhibit a common diode-like rectifying effect. The Ids is significantly suppressed when a negative Vds is applied, as the reverse bias increases the height of Schottky barrier. The rectifying effect could be quantitatively assessed by rectification ratio calculated with the Ids at Vds = 2 V and Vds = −2 V. As depicted in Figure d, the rectification ratio rises to ∼300 when the gate voltage of Vgate = 80 V is applied, and a higher value of >103 is obtained with another device (Figure S10). Furthermore, the output curve measured at Vgate = 80 V is fitted by the Shockley diode function and shown in Figure e,[47] leading to an ideality factor of n = 2.2.

Figure 2

Electrical characterization of the heterostructure device. (a) Transfer curves of the devices with pure InSe or NbTe2 channel. Bias voltages Vds of 2 and 0.1 V were applied in the measurements of InSe and NbTe2 devices, respectively. (b) Transfer curves of the InSe/NbTe2 heterostructure device measured with bias voltage Vds of 2 V and 1 V. (c) Gate voltage dependent output Ids–Vds curves of the InSe/NbTe2 heterostructure device. The inset shows |Ids| on a logarithmic scale. (d) Gate voltage dependent rectification ratio calculated with the results in (c). (e) Fitting of the output Ids–Vds curve measured at Vgate = 80 V by Shockley diode function.[47] Ideality factor of n = 2.2 is extracted from the fitting. (f) Schematic of band bending at the interface between n-doped InSe and metallic NbTe2 under reverse and forward biases when a positive gate voltage is applied. EF denotes the Fermi level of InSe.

The exceptional rectifying effect could be readily explained with the band diagram illustrated in Figure f. When a negative reverse bias is applied to NbTe2, its Fermi level is lifted relative to InSe, resulting in an increased band mismatch according to Figure a. The band mismatch induces significant band bending on InSe side, and a relatively high Schottky barrier is built at the InSe/NbTe2 interface. Therefore, the Ids through InSe/NbTe2 heterojunction is dominated by tunneling current. Once a high positive forward bias (e.g., Vds = 2 V) is applied, the Fermi level of NbTe2 is considerably lowered; thus the band bending on InSe side is eliminated. In this case, the majority carriers in InSe could easily diffuse to NbTe2 and lead to a high conductance of the heterostructure channel. Actually, this band bending conflicts with the band diagram determined with UPS (Figure a), suggesting that significant n-doping of semiconducting InSe is induced in the device fabrication process, especially the deposition of Al2O3 (see Methods), where 2 nm thick aluminum layer was deposited and oxidized as seeding layer. The strongly adjustable band diagram at InSe/NbTe2 interface makes it distinct from the massively investigated vdW heterostructures built with semiconductors. This could be verified with the following optoelectronic measurements.

Photocurrent mappings are performed to achieve a more profound interpretation of the working mechanisms of InSe/NbTe2 heterostructure device. In order to study the intrinsic heterostructure, the gate voltage was fixed at Vgate = 0 V in the measurements. The device was laterally moved with steps of 0.5 μm, and a 532 nm laser beam with 1 μW power was focused on the plane of heterostructure and swept in the area shown in Figure S2. Source-drain current Ids (measured with the same configuration in Figure b) was recorded under various bias voltages Vds of −2 V, 0 V, and 2 V, and the results are depicted in Figure a–c. The green, yellow, and white dashed lines in Figure a–c outline the positions of InSe, NbTe2, and Ti/Au electrodes. An intuitive observation of the mappings is the bias-dependent positions of peak Ids. The peak Ids is obtained at InSe/NbTe2 stacking area under bias voltage Vds of −2 V and 0 V. In contrast, under high forward bias voltage of Vds = 2 V, the peak Ids is obtained at pure InSe area. This observation could be more clearly and quantitatively illustrated by the line scannings extracted at the white arrows in Figure a–c, and the results are shown in Figure d. On the basis of bias-dependent peak Ids positions, a conceptual XOR logic gate with truth table summarized in Figure e is proposed. Two inputs of this logic gate are Y position of the laser beam (e.g., “0” for Y = 5 μm, “1” for Y = 13 μm) and bias voltage Vds (e.g., “0” for Vds = −2 V, “1” for Vds = +2 V) in Figure d, and the output is source−drain current Ids (e.g., “0” for small |Ids| and “1” for large |Ids|). The bias dependent photocurrent generation is also confirmed with the device shown in the inset of Figure S10. The irregular outline of InSe/NbTe2 overlapping area makes it easy to identify the photoresponse area in the whole heterostructure (Figure S11a–c). The mappings of photocurrent under high negative gate voltage (Figure S11d–f) suggest that this phenomenon occurs even when the Fermi level of InSe channel is significantly lowered by negative Vgate, despite the decreased absolute photocurrent. On the basis of the results in Figure S10, this device demonstrates a weak rectifying effect under negative gate voltage of Vgate = −80 V. Therefore, the bias dependent switchable photoresponse mechanisms exist as long as the device shows a rectifying effect or an effective Schottky barrier exists at the heterostructure interface.

Figure 3

Switchable photoresponse mechanisms of InSe/NbTe2 heterostructure. (a–c) Photocurrent mappings in InSe/NbTe2 heterostructure device at various bias voltages Vds of −2, 0, and 2 V. The green, yellow, and white dashed lines illustrate the outlines of InSe, NbTe2, and Ti/Au electrodes, respectively. Scale bars, 5 μm. (d) Line scannings extracted from the mappings at the positions and directions indicated by white arrows in (a)–(c). The blue and brown shades indicate the Y positions of pure InSe and InSe/NbTe2 heterostructure. (e) Truth table of the conceptual XOR logic gate with inputs of laser beam position (Y) and bias voltage (Vds) and output of Ids. (f, g) Schematic of the InSe/NbTe2 heterostructure device switched between the photovoltaic device (f, reverse bias) and phototransistor (g, forward bias). The role of NbTe2 is switched between a heterojunction component and a contact electrode.

The bias dependent peak photocurrent positions imply switchable photodetection mechanisms of InSe/NbTe2 heterostructure, attributed to the band bending tuned by bias voltage Vds. Essentially, the excited photocarriers are mainly generated in InSe. Under reverse bias, a high Schottky barrier is built, and the device turns to heterojunction (Figure f). Thus, the heterostructure responds to light illumination as a photovoltaic device (Figure f). In other words, only the electron–hole pairs generated at InSe/NbTe2 stacking area could be efficiently separated by the built-in electric field and contribute to photocurrent (Figure a). However, the energy barrier is significantly lowered or even vanishes when a large forward bias is applied (Figure f). Consequently, the device works as a phototransistor with homogeneous InSe channel and NbTe2 conductive electrode (Figure g), and the thick InSe channel with direct bandgap could absorb light illumination with high efficiency.[17] Given the above, the InSe/NbTe2 heterostructure device could be readily switched between heterojunction photovoltaic device and InSe phototransistor with NbTe2 contact electrode, indicating the rationality of band engineering design in Figure a.

Finally, photodetection performance of the switchable InSe/NbTe2 heterostructure device is systemically investigated with 532 nm laser illumination. As illustrated in the schematic of Figure a, the photoresponse mechanism is subject to bias-dependent band bending at the interface (Figure f,g). For a phototransistor working under bias voltage Vds = 2 V, Ids is raised when the laser power increases from 10 nW to 100 μW (Figure b). Notably, Ids at Vgate = −80 V increases significantly compared with the values at Vgate = 80 V. This phenomenon could be explained by the change of Fermi level EF of InSe. EF is significantly declined, and the density of carriers is extremely low under gate voltage of Vgate = −80 V. Thus even the photocarriers induced by a low-power illumination could lead EF to rise considerably. Therefore Ids changes a lot as the illumination power is increased. The carrier density is intrinsically high under gate voltage Vgate = 80 V. Thus the photocarriers finitely change the band diagram or contribute to the Ids dominated by the thermionic current. Properties of the InSe/NbTe2 heterostructure working as a photovoltaic device are revealed by the Ids–Vds curves measured under laser illumination with various power, and the results are shown in Figure c and Figure S12. The large ISC (Figure d) and high VOC (Figure e) are characteristics of a photovoltaic device. The highest VOC in Figure e is ∼0.41 V, and large ISC of ∼380 nA (Figure S13) is obtained with the device demonstrated in Figure S10, indicating the excellent performance of InSe/NbTe2 photovoltaic device. As demonstrated in Figure S14, the InSe/NbTe2 heterostructure device works with higher response speed in photovoltaic mode and lower response speed in phototransistor mode, and the ultimate response time should be less than 10 ms. Besides, the photocurrent Iph and corresponding photoresponsivity are calculated (Figure f) based on the Ids–Vds curves measured in dark condition (Figure c), and a high photoresponsivity of ∼12 A/W is achieved with the InSe phototransistor when a bias voltage of Vds = 2 V is applied. All the results in Figure are measured with the laser spot centered at stacking InSe/NbTe2 area. Yet, the highest photocurrent in the phototransistor should be obtained when the pure InSe is illuminated under positive bias, according to Figure c. Therefore, the photoresponse was additionally measured with 10 nW illumination centered at pure InSe, where peak Ids is obtained in Figure c. On the basis of output Ids–Vds curves demonstrated in Figure S15, a remarkably high photoresponsivity of ∼84 A/W is achieved. This photoresponsivity is superior to many of the published works of InSe-based photodetectors, as compared in Table S1. The exceptional photodetection performance affirms the rationality of the band engineering design discussed in Figure , indicating a promising strategy for developing powerful vdW devices. As summarized in Table S2, the performance of this InSe phototransistor with metallic NbTe2 electrode is comparable with or even better than the InSe devices with contact electrodes of other materials.

Figure 4

Overall photodetection performance of InSe/NbTe2 heterostructure device. (a) Schematic of the switchable InSe/NbTe2 heterostructure for the detection of 532 nm laser. Working mechanism of this device depends on bias-dependent band bending at the interface. (b) Transfer curves of InSe/NbTe2 phototransistor illuminated by 532 nm laser beam with gradient power. (c) Ids–Vds curves of the InSe/NbTe2 heterostructure device under light illumination. (d, e) Short-circuit current ISC (d) and open-circuit voltage VOC (e) extracted from the Ids-Vds curves in (c). The results are fitted by ISC ∝ Power and VOC ∝ ln(Power). (f) Photocurrent Iph and photoresponsivity calculated with the data in (c).

Conclusions

Switchable photoresponse mechanisms have been implemented in single van der Waals semiconductor/metal heterostructure. Its working mechanism could be readily switched by the bias voltage that tunes band bending at the semiconductor/metal interface. The electrical characterization demonstrates a significant rectifying effect, indicating an inherent diode-like heterojunction. Further systematic optoelectronic measurements results indicate that reverse bias switches the heterostructure to a photovoltaic device, and alternatively, large forward bias turns it to a phototransistor with InSe channel and NbTe2 electrode. Overall, it exhibits exceptional electronic and optoelectronic performance of >103 rectification ratio and ∼84 A/W photoresponsivity, suggesting that it is a promising candidate for practical applications. The rational band engineering design in this vdW heterostructure and its exceptional device performance indicate a promising strategy for building versatile 2D optoelectronic devices and highlight the functionality of 2D metallic materials.

Methods

UPS Measurements

Bulk materials of InSe and NbTe2 were used for UPS measurements in high vacuum. The materials were sputtered with Ar+ for cleaning, and a bias voltage of −10 V was applied. The photon energy of the UV light source is 21.22 eV, and the work function of the analyzer is 4.66 eV.

Preparation and Characterization of Two-Dimensional Flakes

The InSe and NbTe2 flakes were mechanically exfoliated from bulk materials (2D Semiconductors). In the following, NbTe2 flakes were transferred to silicon substrates covered with 300 nm thick SiO2, and the InSe flakes supported by polydimethylsiloxane (PDMS) were stacked on top. Raman spectra were collected with WITec micro-Raman system, and a 532 nm laser was used for excitation. The reflection of silver mirror, InSe, NbTe2, and InSe/NbTe2 heterostructures was measured with SNOM system (WITec alpha300). The absorbance of the materials is calculated with the reflection of a silver mirror as reference: absorbance = [(RAg – RX)/RAg] × 100%, where RAg and RX are the reflection of silver mirror and 2D materials, respectively. The PL measurements were also conducted with a SNOM system, and a 532 nm laser with power of ∼1 mW was used for excitation. AFM image of the flakes was collected by Dimension Icon system (Bruker).

Device Fabrication

The patterning and deposition of metal electrodes were accomplished through patterning PMMA photoresist with electron beam lithography (EBL, Vistec EBPG 5000), deposition of Ti/Au (5 nm/100 nm) with electron beam evaporation system (MASA IM-9912), and finally the lift-off process in acetone. For optimization, the device was annealed at 180 °C in high vacuum (∼10–5 mbar) for 2 h (AML-AWB wafer bonding machine). Then a 2 nm thick Al seeding layer was deposited and heated at 130 °C in the air for 2 min. Afterward, 20 nm thick Al2O3 was deposited at 120 °C by atomic layer deposition (Beneq TFS-500), followed by second annealing with the same conditions. Finally, the Ti/Au electrodes were connected to a printed circuit board (PCB) by wire bonding. The fabrication of bare device was terminated after the first annealing.

Electrical and Optoelectronic Measurements

Keithley 2401 and Keithley 2400 were used for applying bias voltage Vds and gate voltage Vgate, and the drain-source current Ids was measured with Keithley 2401. Data acquisition was accomplished with a customized LabVIEW program. The PCB with fabricated device was fixed on the lateral movement stage of the SNOM system, which was precisely controlled by the LabVIEW program. In the optoelectronic test, a 532 nm laser with adjustable power was illuminated on the device through a 20× objective (NA = 0.4).