All-Optical Reversible Manipulation of Exciton and Trion Emissions in Monolayer WS2

Monolayer transition metal dichalcogenides (TMDs) are direct gap semiconductors with promising applications in diverse optoelectronic devices. To improve devices’ performance, recent investigations have been systematically focused on the tuning of their optical properties. However, an all-optical approach with the reversible feature is still a challenge. Here we demonstrate the tunability of the photoluminescence (PL) properties of monolayer WS2 via laser irradiation. The broad-range and continuous modulation of PL intensity, as well as the conversion between neutral and charged excitons have been readily and reversibly achieved by only switching the two laser power densities. We attribute the reversible manipulation to the laser-assisted adsorption and desorption of gas molecules, which will deplete or release free electrons from the surface of WS2 and thus modify its PL properties. This all-optical manipulation, with advantages of reversibility, quantitative control, and high spatial resolution, suggests promising applications of TMDs monolayers in optoelectronic and nanophotonic applications, such as erasable optical data storage, micropatterning, and display.


Details about PL trajectory and PL imaging of
The PL trajectory shown in Figure 2a was obtained step by step, rather than a continuous measurement. The experiments were performed as follows: 1) In the first step, we scanned the WS2 sample voxel-by-voxel with a large area (typical 90 μm × 90 μm) by a low excitation power density (typical 20 kW/cm 2 ).
The integration time for each voxel was 10 ms. Considering the low excitation power density and the short excitation time, the change of PL properties can be ignored.
2) Secondly, we selected an interested triangle WS2 monolayer (typical 20 μm × 20 μm) and performed the scanning again to achieve PL image, as shown in Figure 2b.
3) Then, the laser was focused on a selected position. During the low irradiation power (20 kW/cm 2 in the manuscript), PL of this position has no significant change, as t0 to t1 (0 s to 100 s) shown in Figure 2a.
4) Later, we switched the irradiation laser from the low power density to the high power density (900 kW/cm 2 ) and monitored the PL behaviors. During this stage, we switched off the laser at t2 (150 s) and switched on the laser again at t3 (200 s).

5)
In the fourth step, we switched off the high power density laser at t4 (244 s) and performed PL imaging. Typically, PL image consists of an array of 100×100. Thus the scanning time for the full image was about 100 s, considering that the integration time for each voxel was 10 ms. 6) After PL imaging, we re-focused the selected position and recovered the PL intensity by using the low power density (20 kW/cm 2 ). PL images can be obtained at any time of the full process, such as t5 and t6, as performed in step 5.
Comparing with the large time scale for the recovering process (~500 s), the influence on the PL properties arising from the short scanning time (10 ms) can be ignored.

Stability of the PL modification under ambient atmosphere
The stability of PL modification has been confirmed by performing PL imaging of the irradiated locations at different times, as shown in Figure S4. The sample was first modified at 2018.11.1, 9:35 pm. Then we held the sample and experimental setup for overnight and performed PL imaging again at 2018.11.2, 9:58 am. No significant changes can be found between the two images. Further confirmation was taken at 2018.11.4, at 3 pm. Both the sample and optical setup were placed in the ambient condition without any protection. These results illustrate the robust of our approach. To get insight into the changes among PL imaging at different times, we plotted the PL intensities along the marked green lines, as presented in Figure S4d. The slight difference between three-time measurements probably originated from the slight alteration in laser powers. To eliminate this effect, we normalized the PL intensities, as presented in Figure S4e. The consistency for the three measurements proved that the modified PL intensity could be held after switching off the laser irradiation. We farther approved the robust of the PL modification by monitoring PL intensity against laser switching, as shown in Figure S5. We can find that the interrupted PL recovery process could be restarted after switching on the laser again. During laser off, the sample was stored in the ambient atmosphere without any further protection. No significant changes in the PL intensity can be determined. Under the assumption of pseudo-first order model, the change of molecule concentration over time, N(t), is proportional to the difference the difference between equilibrium and current concentration, which can be expressed as: where kad(PLaser) is the rate constant of adsorption in s -1 at a certain laser power. The value of kad(PLaser) depends on the laser power used to irradiate the monolayer WS2, when the power is too low, the activation energy cannot overcome the barrier of the adsorption process, kad(PLaser) is close to zero. According to the Arrhenius law, , the high the irradiation power (and thus the pronounced the laser-induced heat effect), the large the kad(PLaser). The integral form of equation S1 can be expressed as: where N'(0) is the initial concentration of the recovery process (i.e. the molecule concentration at t4 in Figure 2a).
On the other hand, the equilibrium condition of desorption process can be regarded as that all the adsorbed molecules have been lifted from the surface. That's to say, the equilibrium concentration of desorption process is zero. Similar to equation S1, the change of molecule concentration in the desorption process can be described as: where kde(PLaser) is the rate constant of desorption in s -1 at a certain laser power. Its integral form can be given as: When the adsorption and desorption processes are both existing at the same time, the change of molecule concentration can be given by combining equation S1 and S3, as follows: The formula of the combining process can be expressed as: This phenomenon can be attributed to the thermal degradation due to the extremely strong excitation laser, coinciding with previous works. However, these quenching PL cannot be reversed by laser irradiation with any low power densities. bi-exponential fitting curve, the bottle panel is the fitting residual. We can find that the bi-exponential function is more reasonable than single exponential function.