High-Performance Photodetectors Based on Lateral Monolayer MoS2/WS2 Heterojunctions

Monolayer transition metal dichalcogenides (TMDs) show promising potential for next-generation optoelectronics due to excellent light capturing and photodetection capabilities. Photodetectors, as important components of sensing, imaging and communication systems, are able to perceive and convert optical signals to electrical signals. Herein, the large-area and high-quality lateral monolayer MoS 2 /WS 2 heterojunctions were synthesized via the one-step liquid-phase chemical vapor deposition (CVD) approach. Systematic characterization measurements have verified good uniformity and sharp interfaces of the channel materials. As a result, the photodetectors enhanced by the photogating effect can deliver competitive performance, including responsivity of ~ 567.6 A/W and detectivity of ~ 7.17  10 11 Jones. In addition, the 1/f noise obtained from the current power spectrum is adverse to the development of photodetectors, which is considered as originating from charge carrier trapping/detrapping. Therefore, this work may contribute to efficient optoelectronic devices based on lateral monolayer TMD heterostructures.

Although crystal defects in TMDs giving rise to the carrier trapping effect can result in high photosensitivity, they can unavoidably lead to slow response speed yet [17]. Combining respective superiorities and showing unique electronic transport at the junction, TMDs heterostructures either lateral stitching or vertical stacking are presented [18]. Such heterostructures can tailor intrinsic electronic properties and improve the optical absorption [19], showing emerging and designable features [13,20]. For example, the builtin electrical field [21] or energy level difference [22] induced by TMD heterostructures should accelerate photocarrier separation, suppress photocarrier recombination [17,23] and lower dark current [24] as well, which is beneficial for achieving high-performance photodetection.
Besides, Wang's group [25] has certified suppressed electron-hole (e-h) recombination in lateral heterostructures. As previously reported, the lateral heterostructures showed higher carrier mobility [26] whereas the vertical heterostructures usually increased the photoactive area [23] and/or enhanced current drive per area [27]. Moreover, the in-plane interfaces of lateral heterostructures showed stronger emission intensity than both sides [14]. However, the suppressed photoluminescence (PL) emission could be observed in the vertical hetero-interface because of the reduced direct radiative recombination [28]. Additionally, both lateral and vertical TMDs heterostructures make it possible to create new excitonic transitions [14].
In terms of crystal lattice quality, MoX2/WX2 (X = S, Se or Te) lateral heterojunctions could induce structural defects scarcely due to their similar honeycomb-like [29,30] configuration and lattice parameters [30]. In addition, this kind of heterojunction can form type-II band alignment generally, which is desirable for high-efficiency photodetection [28,30,31]

Device fabrication
The standard electron beam lithography (EBL) was used to define the markers and electrode patterns on the as-grown lateral monolayer MoS2/WS2 heterojunctions. The Ti/Au electrodes (10 nm/100 nm) were evaporated on the channel and lifted off in acetone. The device was thermal annealed at 400 ℃ for 2 hours in vacuum and cooled down to room temperature rapidly.

Material characterization
The optical images were captured with OLYMPUS microscope (LV100ND). The Raman, PL and AFM mapping images were measured with a Raman-AFM confocal spectrometer (Witec, alpha300 RA) with a laser of 532 nm.

Device characterization
The (351.5cm -1 and 416.5 cm -1 ) in Fig. 1(b) [26]. High crystal quality of MoS2 and WS2 are implied because no oxidation peak observed in the corresponding Raman spectra [33]. Especially, the eigen-peaks of MoS2 and WS2 both were observed in the stitched interface marked 3 in Fig. 1 to A1g mode, i.e. I2LA/IA1g, is more accurate to verify the thickness than frequency difference [14]. The ratio was estimated to be ~ 2, in agreement with monolayer WS2 measured by 532 nm laser [14].
The distinct red shift of E2g mode (in-plane vibration) can be observed, resulted from alloying effect [37] in the lateral heterojunctions. Notably, this similar behavior were also observed in the vertical heterojunctions, caused by dielectric screening and interlayer coupling [38].
Furthermore, the Raman mapping result in Fig. 1(c) with the blue region of MoS2 and the red region of WS2 indicates the seamless high-quality in-plane heterostructure [13,39]. Fig. 1 area in (a). The corresponding false-color bar is inserted at the bottom of (c)-(e). The corresponding cross-sectional height profile of the blue (between WS 2 and MoS 2 ) and white (between WS 2 and substrate) lines marked in AFM morphology image. Note that few grain boundaries resulting in charge carrier scatting [41] are observed in material inside but edges indicating better electrical transport performance as shown in Fig. 1(f) [13,18,21,44]. We attribute this to the photogating effect, such as a special case of photoconductive effect [45]. The photogating effect can work as a local photogate modulating channel conductance [46]. The optical image of the device with the effective device area of ~ 40 μm 2 is  described in Fig. 2(b) with E1 and E2 electrodes as the source and drain electrodes. In order to figure out the heterojunction configuration, combined Raman mapping was carried out (Fig. 2(c)), indicating the channel materials of lateral MoS2/WS2 heterojunction between the measured source and drain electrodes (E1 and E2) [24]. The blue, red and dark sections are MoS2, WS2 and metal electrodes, respectively. Fig. 2(d)  the channel and the electrodes [47][48][49][50][51][52]. The linear I-V character is conducive to achieving high responsivity but poor sensitivity of photodetectors due to a high dark current [53]. Additionally, the Iph (i.e. Ilight -Idark) of the photodetector increases to 12.5 times of that before thermal annealing, which maybe ascribe to decreased contact resistance [42,54], removal of defects [55] and improved electrical conductivities [56]. Fig. 2(e) depicts the photoswitching characteristics excited by the above wavelengths. The transient current rises rapidly when the light is on and drops as soon as the light is off, implying this photodetector can serve as a prompt light-activated switch [57].
The semi-logarithmic output characteristics with the same wavelength but varied laser power densities are depicted in Fig. 3(a). As expected, photocurrent is enlarged as the laser power densities increase due to more induced photogenerated carriers [58]. Fig. 3 where P and S are laser power density and effective device area, respectively [58]. Fig. 3 where A is a constant and 0 < α < 1. The value of α, obtained by fitting the curve of Iph versus P in Fig.   4(a), is related to the process of carrier capture, recombination and transfer [65,66]. The sublinear relation between Iph and P suggests the presence of the photogating effect in the device further [61].
The higher value of α (such as ~ 0.73) can be obtained when the lower power densities are applied due to reduced photocarrier recombination and the interactions between carriers [66,67]. In contrast, higher power densities can result in a degraded α value of ~ 0.55 because of stronger recombination losses and more trap states [68]. The precondition of the calculated D* via the equation is that the photodetectors are limited by shot noise as the main noise source [45,62,69]. In order to further evaluate D* more accurately, the noise current obtained in Fig. 4(b) is measured under different frequencies [65]. Fig. 4(b) shows the typical 1/f noise [70] in our photodetectors, which is significant impediment to semiconductor industry from new materials. This kind of noise is mainly resulted from the charged impurities and trapping sites in the conductive channel [53,71]. A higher material quality and small structural defect density are desired for reducing the 1/f noise [72].

Acknowledgements
We acknowledge engineer Wanghua Wu for helping PL data analysis. Authors

Availability of data and materials
The datasets supporting the conclusions of this article are included in the article.