Extrinsic Origin of Persistent Photoconductivity in Monolayer MoS2 Field Effect Transistors

Recent discoveries of the photoresponse of molybdenum disulfide (MoS2) have shown the considerable potential of these two-dimensional transition metal dichalcogenides for optoelectronic applications. Among the various types of photoresponses of MoS2, persistent photoconductivity (PPC) at different levels has been reported. However, a detailed study of the PPC effect and its mechanism in MoS2 is still not available, despite the importance of this effect on the photoresponse of the material. Here, we present a systematic study of the PPC effect in monolayer MoS2 and conclude that the effect can be attributed to random localized potential fluctuations in the devices. Notably, the potential fluctuations originate from extrinsic sources based on the substrate effect of the PPC. Moreover, we point out a correlation between the PPC effect in MoS2 and the percolation transport behavior of MoS2. We demonstrate a unique and efficient means of controlling the PPC effect in monolayer MoS2, which may offer novel functionalities for MoS2-based optoelectronic applications in the future.


S1. Device Fabrication
MoS2 flakes (SPI Supplies) were mechanically exfoliated onto octadecyltrichlorosilane (OTS) self-assembled monolayer (SAM) functionalized SiO2 (300 nm)/Si substrates. The surface of the OTS-functionalized SiO2/Si substrate was hydrophobic with a typical contact angle above 110°. The hydrophobic surface decreased the number of absorbate molecules, thereby decreasing the chargedimpurity scattering in MoS2 and charge traps. 1 First, the MoS2 flakes were identified and characterized under an optical microscope using variations in contrast. Figure S1a is a typical optical microscopy image of the MoS2 sample after deposition of the electrode. Figure S1b is the photoluminescence (PL) spectrum of a typical monolayer and bilayer MoS2 sample that were fabricated on OTS-functionalized substrates, showing the exciton peaks A and B. Due to the difference in the quantum efficiency, the intensity of the exciton peaks A and B in the monolayer MoS2 are larger than those in the bilayer MoS2 samples. Another PL peak (I) at ≃1.6 eV corresponding to the indirect interband transition was observed in bilayer MoS2. Figure S1c shows a Raman spectrum (blue curve) of a monolayer MoS2 sample with two characteristic peaks at 388.7 cm -1 and 407.0 cm -1 that correspond to the E2g and A1g resonance modes. The difference between the two peaks is ≃18.3 cm -1 , which is consistent with that obtained for the monolayer MoS2 from previous reports. For comparison, the Raman spectra of the bilayer and trilayer MoS2 samples are also shown. We adopted resist-free fabrication to prevent contamination of the MoS2 samples from the resist residue of the conventional lithography process. We used nanowire as a shadow mask to deposit metallic contacts (Au, 50 nm) with an electron-beam evaporator at a base pressure of ≃ Keithley 2400 was used to apply the back gate voltage. Figure S1d shows the source-drain current ( DS I ) as a function of the source-drain bias ( DS V ), which is linear over the small bias region.

S2. Gas Adsorbate Effect
We stored the MoS2 samples in vacuum ( Torr) for 12 hours before conducting the measurements to minimize the gas adsorbate effect. Figure  conditions. Following illumination in vacuum, the enhanced mobility, the on/off ratio, and the threshold voltage are observed. Figure S3 shows a schematic of the measurement system that integrated optical microscopy, We characterized the PPC relaxation by choosing the starting point of the PPC ( 0 I in Figure   1b of main text) as the first data point after the photocurrent sharply dropped. Here, we assumed that the band-to-band recombination time was comparable to the lifetime of the PL (≃100 ps). 6,7 Because the time delay of the PC measurement was 200 ms, which was much larger than the photocurrent relaxation due to band-to-band recombination, the photocurrent after 0 I was dominated by PPC. Figure S3. Schematic of the cryostat system with integrated OM, PL, Raman, and photocurrent measurements.

S4. Temperature dependence of PPC relaxation
Before each measurement, the MoS2 samples were warmed up to room temperature to ensure that the dark current had reached equilibrium and then cooled down in the dark to the desired temperature. Detailed data on the temperature dependence of PPC are shown here, in which the PPC is normalized by 0 I for purposes of comparison. The PPC clearly weakened as the temperature decreased for samples A and B. As shown in Figure 2b in the main text, both  and  decreased as the temperature decreased. At low temperatures, the PPC dropped more quickly for t <  but became more persistent for t >  , which could be accurately described by a stretched exponential function with a small  . 8 Figure S4. Temporal evolution of the photoresponse for (a) sample A and (b) sample B at different temperatures. The PPC is normalized by 0 I and offset vertically for purposes of comparison.

S5. Fitting of the PPC relaxation curves
To verify the validity of the stretched exponential decay for describing the observed PPC in monolayer MoS2, we consider other possible relaxation approaches, including single exponential, double exponential, and logarithmic decay. The fitting results of two representative PPC relaxation curves at T = 300 K and 180 K by these different schemes are shown in Figure S5. It can be seen that only the stretched exponential decay yields satisfactory fitting result for the whole temporal range at different temperatures. As mentioned in the main text, the observed stretched exponential decay of the PPC relaxation in our MoS2 devices suggests a disordered system, which is consistent with the percolation transport model discussed in Figure 5 of the main text.  Figure S5. Two representative PPC relaxation curves at T = 300 K and 180 K and the fitting results with various schemes, including stretched exponential, single exponential, double exponential and logarithmic decay.

S6. Gate voltage dependence of the PPC
We discuss the back gate voltage ( ) dependence of the PPC relaxation here. We show the PPC relaxation for = −60 V and = +60 V in Figure S6a (also in Figure 4b of the main text), as well as the dependence of τ in Figure S6b. The position of Fermi level, which is controlled by , can greatly affect the carrier density and consequently the conductance in the MoS2 channel. This dependence of conductance is revealed in Figure 1a of the main text and the related discussion.
Nevertheless, the PPC relaxation is mainly determined by the density of the trap states and the extent of the carrier trapping in the RLPF model. Because there are various possible sources of trap states, including electron traps, hole traps, and mid-gap states in MoS2 samples, 9 it is plausible that the strength of the PPC is correlated to the density of the trap states as Fermi level is tuned with .
However, we note that the correlation may be complicated and further study is required to elucidate detailed dependence of the PPC effect.

S7. Transport characteristics of the MoS2 FET
In addition to characterizing the transport behavior of sample B, as discussed in the main text, we show the transfer curves for sample A at different temperatures in Figure S7a. Figure S7b  K, as shown in Figure S7c. K, which is obtained from the transport measurement. Figure S8 shows a distribution of τ versus mobility for 7 monolayer MoS2 samples that were obtained by same illumination condition. The data suggests that the PPC becomes more persistent for higher mobility samples. It is conceivable that more persistent PC is caused by higher density of the trap states in the MoS2 channel. This could result in shorter average distance between the localized states and greater hopping rate among these states, leading to higher mobility. However, the detail mechanism behind the relation between τ and mobility requires further study.