Two-dimensional (2D) materials have been widely studied because of their unique electrical [1–3] and optical [4–8] properties. With high carrier mobility and good mechanical flexibility, graphene has been attracting much attention. However, the practical applications of graphene are limited by its zero bandgap [9–11]. Recently, transition metal dichalcogenides (TMDs) with the chemical formula of MX₂ (M being a transition metal and X being a chalcogen) have emerged as promising materials for electronics and photoelectric applications owing to their sizeable bandgaps and other excellent properties. With a moderate bandgap (while maintaining a direct bandgap from monolayer to the bulk) and intrinsic anisotropy, Rhenium disulfide (ReS₂) has received considerable attention among 40 kinds of TMD materials in the field of transistors and photodetectors. For instance, a ReS₂ based transistor achieved a high current on/off ratio of ∼10⁴ and an ultra-low electron mobility of 0.44 cm²V⁻¹s⁻¹ using graphene as the source/drain electrodes [12]. Furthermore, photodetectors based on ReS₂ [9,13] possess a high responsivity, outstanding external quantum efficiency, and the ability to detect polarized light. However, the slow response time (>100s) of ReS₂ originating from its low electron mobility hinders its application. Therefore, to obtain excellent photodetection performance, it is critical to design appropriate ReS₂ structures with high electron mobility while maintaining a high current on-off ratio with good polarization sensitivity.

In this article, we report a high-performance phototransistor based on suspended ReS₂. The electrical and optoelectronic properties of the suspended ReS₂ device were characterized in detail. Specifically, the obtained devices were found to exhibit an excellent current on-off ratio of 10⁵, a high responsivity of 0.22 A/W, and a fast response time of 83.5/325.3µs. Additionally, the devices maintained the capability of ReS₂ in detecting polarized light, achieving a linearly dichroic ratio of 1.4. These high performances demonstrate the suspending of ReS₂ as a unique and simple way for developing high-performance optoelectronic devices.

Monolayer ReS₂ can be mechanically exfoliated through scotch-tape due to its weak van der Waals interlayer interactions [14]. The inset in Fig. 1(a) shows the optical image of monolayer ReS₂. The schematic of the complete monolayer ReS₂ device is displayed in Fig. 1(b), and the corresponding process of device fabrication is illustrated in Fig. 1(c), including the following steps: (i) The Au film (40 nm) was deposited onto the SiO₂/Si substrate by thermal evaporation and standard photolithography, and few-layer ReS₂ was mechanically exfoliated from bulk ReS₂ onto another SiO₂/Si substrate. (ii) With the help of a microscope, the ReS₂ nanosheet was transferred onto the two portions of Au film through wet transfer (PPC as dielectrics) [15–18]. (iii) The PPC was removed in acetone finally. With such a process, a dual-electrode back-gate field-effect transistor based on suspended ReS₂ was completed. It should be emphasized that short channel (<5µm) electrodes were applied to reduce the stress from PPC. In Fig. 1(a), it is clear that there is no obvious damage in the monolayer ReS₂ channel. Furthermore, we monitored the few-layer ReS₂ quality by acquiring a Raman spectrum continuously throughout the entire fabrication process. Figure 1(d) shows the Raman spectrum of the few-layer ReS₂ before (red) and after (blue) the transfer process. Owing to the asymmetric crystal structure of ReS₂, more than 12 Raman modes were observed in Fig. 1(d), the same as the reported results on ReS₂ [9,14,19,20]. These Raman modes were resulted from the out-of-plane vibrations of Re atoms (A_g-like) and the in-plane vibrations of Re atoms (E_g-like) [16]. PL measurements were also carried out on the few-layer ReS₂ before and after the transfer process and were plotted in the inset of Fig. 1(d), respectively. It is noteworthy that the PL spectra of the ReS₂ nanosheet reveal a tunable emission peak from 1.49eV to 1.51eV, a feature that might be attributed to the strained ReS₂ caused during the device manufacturing process.

 

Fig. 1. (a) Optical microscope image of a suspended ReS₂ device. The inset shows a typical optical image of the transistor. (b) Schematic of the complete device. (c) Device-fabrication process. (d) The Raman spectra and PL spectra of ReS₂ flakes measured before and after the transfer process. (e) Measured peak intensities of the 210 cm⁻¹ modes of ReS₂ after the transfer process as a function of the excitation polarization direction. 0° is defined as the angle at which the polarization direction of the incident light is parallel to the Re-chain.

Download Full Size | PDF

Importantly, for identification of the Re-direction and the convenience of the subsequent measurements [21–23], the Mode V (210cm⁻¹) in ReS₂ was considered and mentioned in Fig. 1(d). Firstly, Re-chain tends to coincide with one of the cleaved edges of ReS_2, and one of them can be defined as 0°. Secondly, the Raman intensity of Mode V reaches its maximum when the laser polarization is parallel to the Re-chain. Therefore, if the maximum occurs at 0°, it means that the Re-chain is parallel to this edge; otherwise, it is the other edge. Here, a half-wave plate was used to manipulate the polarization direction of the incident light. P=0 means that the polarization direction of the incident light is along the Re-chain of ReS₂. The corresponding polar plots in Fig. 1(e) show the polarization dependence of the intensity of Mode V measured in the parallel configuration, from which the Re-chain direction was determined.

To investigate the I-V characteristics of suspended ReS₂ transistors, the I_DS-V_DS outputs and I_DS-V_g transfer characteristics were determined. Here, two devices were fabricated and named device #1 (suspended ReS₂) and #2 (supported ReS₂). Evidently, the I_DS of device #1 had an increase of one order of magnitude than that of device #2 as shown in Fig. 2(a), attributing to the suspension of ReS₂. First, the physical transfer of the ReS₂ channel was a gentle and low-energy manufacturing technique that did not require to go through the traditional fabrication processes (such as thermal evaporation and standard photolithography) to prevent defects, residues, and impurities on the 2D ReS₂ surface as shown in Fig. 2(b). In other words, the ReS₂ channel of device #1 was unbroken due to its gentle preparation method. Second, the ReS₂ was not in contact with the substrate, thus leading to the reduction of carrier scattering losses from the SiO₂/ReS₂. In short, a suspended ReS₂ device could eliminate the effects of Fermi-level pinning making the interface to approach an ideal physical model. In addition, the curvature of device #1 was symmetric and linear, indicating the ohmic contact in device #1, while it was Schottky contact in device #2.

 

Fig. 2. (a) Current-voltage (I-V) characteristics of the device#1 (red) and #2 (blue). (b) Cross-sectional schematics of the transferred ReS₂ on top of Au electrode and conventional Au electrodes on top of ReS₂. (c) Output characteristics of the device at P = 0 V under different back gate changes from −40 to 40V. (d) Transfer characteristics of the device at P = 0 V under different bias, from 0.1 to 1V.

Download Full Size | PDF

To further understand the contact between the suspended ReS₂ and Au, the I-V curves under different V_g values, varied from −40V to 40 V, were plotted in Fig. 2(c). Apparently, the linear curves indicate the well-developed contact between ReS₂ and Au in device #1. Figure 2(d) presents the transfer characteristics of the device under different source-drain voltages. The threshold voltage V_th was around −33 V, indicating that the majority carriers were electrons, as observed in most of the TMDs-based transistors [24,25]. At the gate voltages from −40 V to 35 V, the source-drain current was suppressed (∼20 pA) at the source-drain voltage of 0.5 V, and it increased in accordance with the increase in the source-drain voltage from 0.1 to 1 V. Generally, with an increase in the source-drain voltage, the source-drain current would increase, resulting from its proportional relationship with the electric field between two the electrodes. Additionally, under a source-drain bias of 1V, a high on/off ratio of 10⁵ was realized, similar to the reported result [9]. Then the carrier mobility of ReS₂ can be extracted from the following formula:

To investigate the photoresponse of device#1, the measuremnts were acquired in the dark and under laser illumination with a diameter of 0.5mm, separately. As shown in Fig. 3(a), the device displayed linear I−V characteristics under the bias voltages from −1 to 1V. The I_DS increased linearly with the source-drain bias, which can be attributed to the increase in carrier drift velocity and reduction of the carrier transit time [24], leading to a more significant channel current. Furthermore, as shown in the inset of Fig. 3(a), the photocurrent I_ph (I_ph = I_light − I_dark) also displayed a linear relationship with the bias voltage varying from 0 to 1 V, confirming that a significant photocurrent can be obtained through a stronger bias voltage. In Fig. 3(b), under illumination with a light density of 10.68 mW/cm², the V_CNP (voltage of charge-neutral point) shifted to a higher negative voltage (≈−39 V) compared with that in dark (≈−34 V), suggesting the existence of trap states within ReS₂. These trap states can trap photo-generated electrons and generate a gate electric field to adjust the channel conductance.

 

Fig. 3. (a) Output characteristics of the device at V_g = 0V under different illumination powers from 0 to 15.96 mW/cm². (b) Transfer characteristics of the device at V_g = 0 V under different illumination powers from 0 to 10.68 mW/cm². (c) Photocurrent response of the ReS₂ photodetector under various light powers, fixing V_DS at 1 V and V_g at 0 V. (d) The photocurrent response of device#1 and #2 revealing the rising time and falling time. The photocurrent response of device #2 was scaled for comparison. (e) Rise and fall times of the high-resolution photocurrent response of the suspended device. (f) Photoswitching before and after 500 cycles demonstrating stability of the suspended ReS₂.

Download Full Size | PDF

The transient photoresponse was obtained by measuring the change of I_D during on and off switching of illumination, realized through an optical chopper. As displayed in Fig. 3(c), when the frequency and source-drain voltage were set as 2Hz and 1V, respectively, the photoresponse curves of suspended ReS₂ under illumination with different laser intensities, demonstrated excellent repeatability and periodicity of photoswitching. It is clear that the photocurrent increased gradually with the increasing of light intensity under a fixed V_DS. For instance, the I_light of the device under a light density of 20.2 mW/cm² was approximately three times larger than that under 6.23 mW/cm². Here, the rising (falling) time is defined as the time interval required for the response to rise (decay) from 10% (90%) to 90% (10%). As the light was turned on (off), I_D increased (dropped) rapidly to the maximum (minimum) with a rising (falling) time of only ≈ 15 ms (11 ms). An oscilloscope was introduced into the detection system to obtain the fastest response speed of the suspended ReS₂ device. The rising and falling times were measured to be 83.5 and 325.3µs, respectively (Fig. 3(e)), which are over three orders of magnitude compared to that of the supported device#2 (4 and 13s) and those obtained in previous work. Furthermore, the curve in Fig. 3(f) can keep nearly unchanged after 500 cycles of illumination in ambient conditions indicating the superior stability of the suspended ReS₂. An essential performance metric of photodetectors is the responsivity, $R = \frac{{{I_{ph}}}}{{P\frac{{{S_s}}}{{{S_L}}}}}$, where I_ph, P, S_L, and S_S represent photocurrent, the power density of incident light, the area of the laser beam, and the area of ReS₂ channel, respectively. As shown in Fig. 3(d), the responsivity for device #1 was extracted to be 0.22 A/W [21,28,29], which is superior to device#2 (0.05 A/W) and previously reported photodetectors based on other 2D materials. It is worth mentioning that optical absorption was anticipated to play a certain role for the enhanced photo-responsivity. From the Fig. 4(a), we can see that an excitonic peak located at ≈826nm is clearly identified in the spectrum, which is consistent with the PL peak at ≈1.501 eV (1.51 eV in Fig. 1(d)). The difference in absorption spectrum between the suspended ReS₂ and the supported ReS₂, originating from the reflection of photons (the inset in Fig. 4(a)), indicates the significance of optical absorption.

 

Fig. 4. (a) Absorption spectrum of suspended ReS₂ and supported ReS₂. Inset: the reflection process of photons. (b) Photocurrent mapping of the suspended ReS₂ device under 0° (top) and 90° (bottom) polarization, showing prominent linear dichroic photodetection. Schematic band diagrams of the Au-ReS₂-Au structure in a thermal equilibrium state (c) and under a source−drain bias (d).

Download Full Size | PDF

The fast response time and superior stability under the illumination of suspended ReS₂ offered us an opportunity to study polarization-sensitive photocurrent mapping of suspended ReS₂. Here, the polarized direction of the incident light at 532 nm was adjusted by a half-wave plate. We performed scanning photocurrent measurements, achieved by scanning the surface of the photodetector with a focused 300nm light spot on the suspended device at V_DS= 1V under 0° (top) polarization and 90° (bottom) polarization. It was clear intuitively from Fig. 4(b) that there was a remarkable difference in the current intensity between 0° polarization and 90° polarization, originating from the more significant absorption of ReS₂ channel under the 0° polarization [21,28–30]. The anisotropic photocurrent ratio, defined as I_max/I_min, was calculated to be 1.4, which is comparable with that of other photodetectors based on traditional 2D materials [31–35]. The results indicate that suspended ReS₂ still has enormous potential in the polarization-sensitive photodetector. Note that the presence of the strongest photoresponse area was in the boundary of ReS₂ and Au. The observed phenomenon of suspended ReS₂ photodetector can be explained by the energy band diagrams in Fig. 4(b) and 4(c). When the first Schottky junction was exposed to the light spot, electron-hole pairs generated in the ReS₂ and photo-generated carriers were separated, resulting from the built-in field between ReS₂­ and Au, however, the holes were trapped in ReS₂. Under a source-drain voltage, the holes continued to accumulate, and the electrons could recirculate along the ReS₂ channel, lowering the Schottky barrier. Therefore, a strong photoresponsivity was observed. When the light spot was shed on the other Schottky junction, the reverse built-in electric field caused the electrons and holes to drift in opposite directions as compared to the first junction. Under the same bias, the separation of the electron-hole pairs was suppressed, thus causing a weak photoresponsivity.

In summary, we successfully fabricated high-performance photodetector based on suspended ReS₂, with a high current on/off ratio (10⁵) and a high responsivity (0.22 A/W) at room temperature. By suspending the ReS₂ on the Au electrodes, the response times (83.5/325.3µs) of our device are three orders of magnitude faster than that of ReS₂ reported earlier due to the reduced interface scattering and ohmic contact. Furthermore, the photodetector based on suspended ReS₂ exhibited a strong polarization-dependent photoresponse with a linearly dichroic ratio of 1.4, which is comparable with those of other reported 2D materials. Our results provide the means for increasing the response speed of photodetectors by using suspended 2D materials as channels that offer insights into their potential applications in optical and optoelectronic devices.