Interface-Induced WSe2 In-plane Homojunction for High-Performance Photodetection.

2D transition metal dichalcogenides (TMDCs) have been extensively attractive for nano-electronics and nano-optoelectronics due to their unique properties. Especially, WSe2, having bipolar carrier transport ability and sizable bandgap, is a promising candidate for future photodetectors. Here, we report an in-plane WSe2 homojunction formed by the interface gate of the substrate. In this architecture, an insulated h-BN flake was used to make only part of WSe2 flake contact substrate directly. Finally, the structures of WSe2/substrate and WSe2/h-BN/substrate construct an in-plane homojunction. Interestingly, the device can operate in both photovoltaic and photoconductive modes at different biases. As a result, a responsivity of 1.07 A W-1 with a superior detectivity of over 1012 jones and a fast response time of 106 μs are obtained simultaneously. Compared with previously reported methods adopted by chemical doping or electrostatic gating with extra bias voltages, our design provides a more facile and efficient way for the development of high-performance WSe2-based photodetectors.

Introduction

In recent decade, pan class="Chemical">2D transition metal dichalcogenides (TMDCs) have drawn great attention owing to their particular properties. High in-plane mobility, tunable bandgap, mechanical flexibility, strong light-matter interaction, and easy processing make them very competitive for future nano-optoelectronics devices [1-20]. Especially, tungsten diselenide (WSe2), a bipolar semiconductor with facile carrier-type manipulation, allows remarkably potential applications in the junction-based photodetectors [21-28]. So far, the main strategies of constructing junction solely in WSe2 include chemical doping and electrostatic gating. For example, recently, an intramolecular WSe2 p-n junction was reported [26]. The n region and p region within WSe2 were formed by polyethyleneimine chemical doping and back gate control, respectively. The p-n junction presented a responsivity of 80 mA W−1 and 200 μs response time. Sun et al. doped WSe2 by using cetyltrimethyl ammonium bromide to form intramolecular p-n junction, in which the responsivity and the response time are 30 A W−1 and ~ 7 ms, respectively [27]. Baugher et al. demonstrate a lateral WSe2 p-n junction achieved by electrostatic gating through applying two gate biases with opposite polarity. The responsivity of 210 mA W−1 has been obtained [28]. However, due to the inevitable chemical impurities and the necessary multiple bias settings, these methods make the fabrication and application of junction-based devices complex and difficult. Assembling various 2D materials to build vertical van der Waals heterostructures like WSe2/MoS2 junction [29] has become popular for the development of novel photodetectors. But, in this configuration, the process of carrier transport between different layered materials suffers from the interface defects, which restricts the device response speed. For the Schottky junction formed between metals and 2D materials, the Schottky barrier height is usually determined by Fermi-level pinning, which is uncontrollable and has a great impact on the responsivity of the devices. Additionally, the reported works cannot seem to possess both high responsivity and fast response speed.

Here, we demonstrate a facile and more efficient way to realize an in-plane pan class="Chemical">pan class="Chemical">WSe2pan> homojuncn class="Chemical">papan>n class="Chemical">tion. In the architecture, part of WSe2 channel is on the Si/SiO2 substrate and the other part is on the h-BN flake. This scheme is common in floating/semi-floating gate memories, in which the h-BN is adopted as gate dielectric layer [30, 31]. The charges stored on one side of h-BN layer can regulate the conductivity of the material on the other side. In our work, however, the h-BN flake as a perfect isolator is used to eliminate the interface gating effect on the WSe2 channel. The polarity of WSe2, which part is only on the Si/SiO2 substrate, can be modulated by interface gate. As a result, the devices operate in photovoltaic (PV) mode well at zero bias. Meanwhile, it exhibits photoconductive (PC) characteristics at high bias. A responsivity of 1.07 A W−1 with a superior detectivity of over 1012 jones and a fast response time of 106 μs are obtained simultaneously without the intricate device design and the risk of introducing additional chemical impurities.

Results and Discussion

Figure 1a shows a schemapan class="Chemical">pan class="Chemical">tipan>c of the in-plane n class="Chemical">papan>n class="Chemical">WSe2 homojunction. It can be seen that part of WSe2 flake is placed on h-BN flake (WSe2-h) and the other part contacts the Si/SiO2 substrate directly (WSe2-S). The function of h-BN is to isolate the interface gate (IG) of the Si/SiO2 substrate on the WSe2-h. So, the formation of homojunction between WSe2-h and WSe2-S mainly relies on the IG modulating the polarity of WSe2-S. The IG is produced by the trapped charges at the SiO2 surface. This will be discussed below in detail. Figure 1b presents the optical picture of the device. Four electrodes (E1-E4, Ti/Au) were prepared by electron-beam lithography, metallization, and the lift-off process. The thickness of materials is characterized by atomic force microscope (AFM) (see Fig. 1c). The height of WSe2 (h-BN) flake in direct contact with the Si/SiO2 substrate (white dotted lines) was measured as 65 (23) nm (see Fig. 1d, e). It can be seen that there is a slope instead of sharp step in the height profile between the WSe2 (h-BN) and the Si/SiO2 substrate. This may be due to the residual photoresist at the edge of the material. Figure 1f shows the Raman spectra of WSe2 and h-BN flakes. For the WSe2, the first order E2g and A1g Raman modes are clearly distinguished ~ 250 cm−1, suggesting that the WSe2 has a multilayer morphology [32, 33]. For the h-BN, the Raman peak of E2g mode at ~ 1370 cm−1 is observed. Due to the large bandgap of h-BN, the Raman signal is weak compared with that in WSe2 [34].

Fig. 1

Schematic of an in-plane WSe2 homojunction. a Structure of the device. b Optical image of the device. Part of WSe2 contacts h-BN flake while the other part contacts Si/SiO2 substrate. c AFM image of the device. The white dotted lines indicate the positions where the thickness of h-BN (left) and WSe2 (right) are extracted. For the channel between E1 and E2, the average width (length) is ~ 19.15 (~ 6.33) μm. For the channel between E2 and E3, the average width (length) is ~ 23.15 (~ 5) μm. For the channel between E3 and E4, the average width (length) is ~ 22 (~ 5.38) μm. d, e Height profiles of WSe2 and h-BN flakes. f Raman spectra of WSe2 and h-BN flakes with 532 nm laser excitation

Schemapan class="Chemical">pan class="Chemical">tipan>c of an in-plane n class="Chemical">papan>n class="Chemical">WSe2 homojunction. a Structure of the device. b Optical image of the device. Part of WSe2 contacts h-BN flake while the other part contacts Si/SiO2 substrate. c AFM image of the device. The white dotted lines indicate the positions where the thickness of h-BN (left) and WSe2 (right) are extracted. For the channel between E1 and E2, the average width (length) is ~ 19.15 (~ 6.33) μm. For the channel between E2 and E3, the average width (length) is ~ 23.15 (~ 5) μm. For the channel between E3 and E4, the average width (length) is ~ 22 (~ 5.38) μm. d, e Height profiles of WSe2 and h-BN flakes. f Raman spectra of WSe2 and h-BN flakes with 532 nm laser excitation

To explore the effect of substrate on pan class="Chemical">pan class="Chemical">WSe2pan>, transfer characterisn class="Chemical">papan>n class="Chemical">tics of WSe2-S and WSe2-h were studied separately. As shown in Fig. 2a, both transfer curves exhibit bipolar behavior and an obvious hysteresis can be observed in the curve of WSe2-S (black) compared with that of WSe2-h (red). The current of WSe2-h is higher than that of WSe2-S. The steep slope in the curve of WSe2-h indicates a relatively large transconductance, which is proportional to carrier mobility. For WSe2-S, the hysteresis is attributed to the charge trapping at the SiO2 surface [35-38]. When Vg was swept from − 30 to 0 V, the negative Vg makes the WSe2 populated with holes and drives some holes into the SiO2 (see Fig. 2b). The trapped holes in SiO2 generate a positive local gate, i.e., IG, to modulate the WSe2 conductance in return (weak depletion effect). Therefore, the charge neutrality point of Vg appears around − 5 V. Similarly, when Vg was swept from 30 to 0 V, the positive Vg makes the WSe2 populated with electrons and also drives some electrons into the SiO2 (see Fig. 2c). The trapped electrons in SiO2 generate a negative IG to modulate the WSe2 conductance in return (the same weak depletion effect). So, the charge neutrality point of Vg appears around 5 V. For WSe2-h, the h-BN flake inhibits the carrier transfer between WSe2 and SiO2 under Vg modulation. This is the reason for the non-obvious hysteresis in the WSe2-h curve. Therefore, an in-plane homojunction can be formed simply by taking advantage of the IG.

Fig. 2

Transfer characteristics. aId-Vg curves of WSe2-S (black line) and WSe2-h (red line). The sweep direction of Vg is indicated by the arrows. b, c Physical explanation for the hysteresis phenomenon. The arrows indicate the direction of electric field induced by Vg. The red and blue spheres represent holes and electrons, respectively

Transfer characterispan class="Chemical">pan class="Chemical">tipan>cs. n class="Chemical">papan>n class="Gene">aId-Vg curves of WSe2-S (black line) and WSe2-h (red line). The sweep direction of Vg is indicated by the arrows. b, c Physical explanation for the hysteresis phenomenon. The arrows indicate the direction of electric field induced by Vg. The red and blue spheres represent holes and electrons, respectively

Figure 3a shows the Id-Vd curves of the device under dark and light condipan class="Chemical">pan class="Chemical">tipan>ons at Vg = 0 V. The source-drain voltage is applied on electrodes E2 and E3 (see the inset). It can be seen that the short-circuit currents (at Vd = 0 V) increase with the incident power, indican class="Chemical">papan>n class="Chemical">ting a PV effect. Interestingly, the curves also present PC characteristics at Vd = ± 1 V. For the former, the photocurrents are attributed to the homojunction. As shown in Fig. 3b, although Vd and Vg were set at 0 V, a few already trapped holes in SiO2 form a small positive IG to modulate the WSe2-S. So, the n--type WSe2-S and intrinsic WSe2-h (without the effect of IG due to the isolation by h-BN flake) constitute an in-plane homojunction. Under illumination, the photoexcited electron-hole pairs will be separated by built-in field of the homojunction. Although Id-Vd curves present PV characteristic well at zero bias, the homojunction did not show a rectifying behavior maybe due to the relatively weak built-in field compared with the externally applied Vd. For the latter, the whole WSe2 flake as a photoconductor responds the light signal at high bias. The photoexcited carriers will be driven to the electrodes by Vd. Therefore, the photoresponse in Fig. 3a is the result of synergistic effect of PV and PC modes. The responsivities as a function of the light power for different Vd are summarized in Fig. 3c, given by R = Iph/PA, where Iph is the photocurrent, P is the power intensity, and A is the effective photosensitive area of the detector [39, 40]. During the calculation, the effective photosensitive area, i.e., the WSe2 part between E2 and E3, is 115.75 μm2. The responsivities of 1.07 A W−1 and 2.96 A W−1 are obtained for Vd of 0 V and 1 V, respectively. The specific detectivity (D*) as an important parameter determines the capability of a photodetector to response a weak light signal. Assuming that the shot noise from the dark current is the major contribution, D* can be defined as D∗ = RA1/2/(2eIdark)1/2, where R is the responsivity, A is the effective photosensitive area, e is the electron charge, and Idark is the dark current [41, 42]. Benefitting from the extremely low Idark, D* of 3.3 × 1012 jones (1 jones = 1 cm Hz1/2 W−1) and 1.78 × 1011 jones are achieved for Vd of 0 V and 1 V, respectively. Moreover, response time as a key figure of merit has been studied. As shown in Fig. 3d, a high and a low current state acquired at Vd = 0 V have been obtained with the light modulation. The transient photoresponse exhibits highly stable and reproducible characteristics. Figure 3e gives a single modulation cycle of temporal response. The rising time (tr), defined as the time necessary for the current to increase from 10% Ipeak to 90% Ipeak, was found to be ~ 106 μs, and the falling time (tf), defined analogously, was found to be ~91 μs. Figure S1 shows temporal response of the device acquired at Vd = 1 V. tr and tf were found to be ~105 μs and ~ 101 μs, respectively. Table 1 summarizes the reported WSe2 homojunction formed by different methods. Obviously, the device in our work has high D*, comparable R, and relatively fast response speed. Moreover, Figure S2 presents the photoresponse characteristics of the other three devices. Distinct PV and PC currents can be observed at zero and high bias, respectively. The detectivity of all the WSe2 homojunctions is higher than 1012 jones, and the response time is a little more than 100 μs, proving that our devices can repeat the high-performance photodetection very well.

Fig. 3

Photoresponse performance of the homojunction acquired between E2 and E3. a Drain current as a function of source-drain voltage applied on electrodes E2 and E3 (see the inset) with variable light power intensity (637 nm). b Formation mechanism of the homojunction at Vg = 0 V and Vd = 0 V. c Responsivity as a function of light power. d, e Temporal response of the device acquired at Vd = 0 V for 637 nm illumination. An oscilloscope was used to monitor the time dependence of the current

Table 1

Optoelectronic characteristics of WSe2 homojunction formed by different methods

WSe2 homojunction formed by	Wavelength (nm)	Responsivity (A W−1)	Detectivity (jones)	Time	References
h-BN/two gates	532	7 × 10−4	-	10 ms	24
SiN/two gates	500–900	0.016	-	-	25
Polyethylene imine chemical doping	520	0.08	1011	200 μs	26
Cetyltrimethyl ammonium bromide chemical doping	450	30	1011	7.8 ms	27
HfO2/two gates	532	0.21	-	-	28
h-BN/interface gate	637	1.07	1012	106 μs	This work

Photoresponse performance of the homojuncpan class="Chemical">pan class="Chemical">tipan>on acquired between E2 and E3. a Drain current as a funcn class="Chemical">papan>n class="Chemical">tion of source-drain voltage applied on electrodes E2 and E3 (see the inset) with variable light power intensity (637 nm). b Formation mechanism of the homojunction at Vg = 0 V and Vd = 0 V. c Responsivity as a function of light power. d, e Temporal response of the device acquired at Vd = 0 V for 637 nm illumination. An oscilloscope was used to monitor the time dependence of the current

Optoelectronic characterispan class="Chemical">pan class="Chemical">tipan>cs of n class="Chemical">papan>n class="Chemical">WSe2 homojuncpan class="Chemical">tion formed by different methods

Figure 4a and b present the Id-Vd characterispan class="Chemical">pan class="Chemical">tipan>cs of n class="Chemical">papan>n class="Chemical">WSe2-h and WSe2-S separately. The curves of both WSe2-h and WSe2-S exhibit PC property, and there is no photocurrent at zero bias. In fact, Ti/WSe2/Ti should be supposed to form a metal/semiconductor/metal structure which contains two Schottky junctions with opposite built-in field. So, the Id-Vd curves should be cross the zero-point and exhibit PC behavior. In our case, due to the different work functions of WSe2-h and WSe2-S, there are two asymmetric Schottky contacts, i.e., E2/WSe2-S and E3/WSe2-h, as shown in Fig. 4c. At zero bias, the direction of net photocurrents originated from the Schottky junctions is opposite to that in the homojunction, and the experiment result shown in Fig. 3a is consistent with the latter. Therefore, the homojunction formed between WSe2-h and WSe2-S is the reason for the short-circuit photocurrents.

Fig. 4

Effect of Schottky junction on photoresponse. aId-Vd curves of WSe2-h with source-drain voltage applied on electrodes E3 and E4 (see the inset) under light illumination (637 nm). bId-Vd curves of WSe2-S with source-drain voltage applied on electrodes E1 and E2 (see the inset) under light illumination (637 nm). c Schematic band diagram of the homojunction device with asymmetric Schottky contacts, i.e., E2/WSe2-S and E3/WSe2-h, at zero bias

Effect of Schottky juncpan class="Chemical">pan class="Chemical">tipan>on on photoresponse. n class="Chemical">papan>n class="Gene">aId-Vd curves of WSe2-h with source-drain voltage applied on electrodes E3 and E4 (see the inset) under light illumination (637 nm). bId-Vd curves of WSe2-S with source-drain voltage applied on electrodes E1 and E2 (see the inset) under light illumination (637 nm). c Schematic band diagram of the homojunction device with asymmetric Schottky contacts, i.e., E2/WSe2-S and E3/WSe2-h, at zero bias

To further demonstrate that the photoresponse at zero bias is attributed to the homojuncpan class="Chemical">pan class="Chemical">tipan>on, the output propern class="Chemical">papan>n class="Chemical">ties were investigated through measuring the Id-Vd curves of the device with the source-drain voltage applied on electrodes E1 and E4. As shown in Figure S3a, the curves, same as the situation in Fig. 3a, also exhibit the PV and PC characteristics. As discussed above, for the former, the photocurrents are attributed to the built-in field of in-plane homojunction formed between WSe2-S and WSe2-h. For the latter, the photocurrents are attributed to the collection of photoexcited carriers by the externally applied Vd. The responsivities as a function of the light power for different Vd are summarized in Figure S3b. The responsivities (detectivities) of 0.51 A W−1 (2.21 × 1012 jones) and 3.55 A W−1 (5.54 × 1012 jones) are obtained for Vd of 0 V and 1 V, respectively. During the calculation, the effective photosensitive area, i.e., the WSe2 part between E1 and E4, is 519.4 μm2. The response time measured at zero bias is shown in Figure S3c and 3d, in which the rising time is 289 μs and the falling time is 281 μs. For the Vd of 1 V (Figure S3e and 3f), the rising and falling time are 278 μs and 250 μs, respectively. The response speed is a little slower than that measured between electrodes E2 and E3, because the relatively long conductive channel increases the photocarrier transmission distance and the probability for the interaction between photocarriers and defects.

Conclusion

In summary, we have demonstrated an in-plane pan class="Chemical">pan class="Chemical">WSe2pan> homojuncn class="Chemical">papan>n class="Chemical">tion by electrically tuning partial WSe2 flake through interface gate. Compared with existing approaches like chemical doping and electrostatic gating by taking advantage of two gate biases, this design gives a more facile rout to realize WSe2 homojunction. With light illumination, the device produces distinct short-circuit photocurrents with a detectivity of 3.3 × 1012 jones. At high bias, the device presents photoconductive characteristic and generates photocurrents with a detectivity of 1.78 × 1011 jones. A response time as fast as 106 μs is also obtained simultaneously. Our study provides an efficient and reliable way for the development of high-performance WSe2-based photodetectors.

Methods

Both pan class="Chemical">pan class="Chemical">WSe2pan> and h-BN bulk materials were purchased from Shanghai Onway Technology Co., Ltd. First, the h-BN and n class="Chemical">papan>n class="Chemical">WSe2 flakes were mechanically exfoliated onto a p+-Si/SiO2 (300 nm) substrate and a poly-dimethyl siloxane (PDMS) layer, respectively. Then, a micromanipulator was used to put the WSe2 flake, which is adhered to PDMS, onto the target h-BN flake through the microscope to locate the position. Part of WSe2 flake overlaps the h-BN flake. Finally, the WSe2 flake was released from PDMS through heating the substrate. The electrodes (Ti/Au) were prepared by electron-beam lithography, metallization, and the lift-off process. Photoresponse measurements were conducted using Agilent B1500 semiconductor parameter analyzer and laser diode with the wavelength of 637 nm.

Addipan class="Chemical">pan class="Chemical">tipan>onal file 1: Figure S1. Temporal response of the device acquired at Vd = 1 V for 637 nm illuminan class="Chemical">papan>n class="Chemical">tion. Figure S2. Photoresponse of the other three devices under 637 nm illumination. Figure S3 Photoresponse performance of the homojunction acquired between E1 and E4.