Ultra-low power threshold for laser induced changes in optical properties of 2D Molybdenum dichalcogenides

The optical response of traditional semiconductors depends on the laser excitation power used in experiments. For two-dimensional (2D) semiconductors, laser excitation effects are anticipated to be vastly different due to complexity added by their ultimate thinness, high surface to volume ratio, and laser-membrane interaction effects. We show in this article that under laser excitation the optical properties of 2D materials undergo irreversible changes. Most surprisingly these effects take place even at low steady state excitation, which is commonly thought to be non-intrusive. In low temperature photoluminescence (PL) we show for monolayer (ML) MoSe2 samples grown by different techniques that laser treatment increases significantly the trion (i.e. charged exciton) contribution to the emission compared to the neutral exciton emission. Comparison between samples exfoliated onto different substrates shows that laser induced doping is more efficient for ML MoSe2 on SiO2/Si compared to h-BN and gold. For ML MoS2 we show that exposure to laser radiation with an average power in the $\mu$W/$\mu$m$^2$ range does not just increase the trion-to-exciton PL emission ratio, but may result in the irreversible disappearance of the neutral exciton PL emission and a shift of the main PL peak to lower energy.

commercial samples. In ML MoSe 2 the sharp exciton emission lines at cryogenic temperatures serve as a sensitive probe of the trion-to-neutral exciton PL emission ratio, which indicates presence/absence of additional carriers. As the Ti-doped sample exfoliated from VPT grown bulk is presumably p-type (confirmed in Hall measurements on bulk) an increase of the trion contribution is consistent with adding holes. We then show that in theses MoSe 2 sample systems the effect of laser doping is clearest on SiO 2 substrates and much weaker in samples exfoliated onto h-BN and Au. Laser induced doping is also studied in ML MoS 2 . Here exposure to pulsed lasers leads to irreversible changes in the emission spectrum: the neutral exciton signature is lost in PL and the main PL emission is shifted in energy below the initial trion emission. This laser engineering of optical properties for ML MoS 2 is demonstrated at T=4 K and at room temperature. (99.9999%), Selenium shots (99.9999%) and I 2 , which acted as a transport agent, were sealed in a quartz tube at 5×10 −5 Torr vacuum. The precursors (hot end) were kept at 1085 • C in a 3-zone horizontal furnace while maintaining a 55 • C temperature difference on the cold end (≈1030 • C) to initiate nucleation and growth [22], see schematic in Fig. 1a. To incorporate Ti into MoSe 2 , the temperature difference between the two ends of the tube varied from 55 to 65 • C. For comparison bulk MoSe 2 crystals were purchased from 2D semiconductors.
The MoS 2 bulk was also supplied by 2D semiconductors. Using a dry-stamping technique [23] MLs from different bulk sources were deposited on either SiO 2 /Si, h-BN [24] or gold substrates. Monolayer thickness was confirmed by several techniques: in optical contrast measurements (Fig. 1b), in Raman spectroscopy [21] (Fig. 1c) and atomic force microscopy ( Fig. 1d). The thickness of MoSe 2 MLs measures ∼0.7 nm in height, and the out-of-plane (A 1g ) peak of MoSe 2 softens from bulk (blue dashed line) to ML (red dashed line) as shown in Fig. 1c due to much reduced restoring forces acting on individual ML sheets.
Optical spectroscopy techniques.-A purpose-build micro-PL set-up is used to record the PL spectra in the temperature range T = 4 − 300 K [25]. The sample is placed on 3-axis stepper motors to control the sample position with nm precision inside the lowvibration closed cycle He cryostat. MLs were excited with a linearly polarized cw laser (532 nm or 633 nm wavelength) or with 1.5 ps pulses at 400 nm generated by a tunable modelocked frequency-doubled Ti:Sa laser with a repetition rate of 80 MHz [26]. In all cases, the 3 FIG. 2: Low temperature PL of monolayer MoSe 2 (a) PL spectrum of MoSe 2 ML (exfoliated from a VPT-grown bulk) at T=8 K for a 40 nW cw excitation (633 nm) before and after being exposed to a power of 40 µW at 633 nm during 4 minutes. After exposure, the trion intensity increases. The inset shows that the trion's dissociation energy increases as a function of the T/X ratio, which is a signature of laser-induced doping of the ML.  [27][28][29][30][31][32][33][34][35]. On the other hand, laser excitation can physically induce non-reversible changes of the ML sample and therefore its optical emission [16,[36][37][38][39][40][41]. Time-evolution of the PL spectrum at T = 4 K while being exposed to a cw excitation (633 nm) at 40 µW during 4 minutes. Only a very slight increase of the T emission is observed. (d) Timeevolution of the PL spectrum at T = 4 K of a commercial MoSe 2 ML exfoliated onto a gold layer (different colours for different times as in panel (a)). The ML is exposed to a cw excitation (633 nm) at 50 µW during 10 minutes. No significant change of the spectrum is observed.
Our target is to distinguish between these two scenarios. A simple approach is to perform power dependent measurements several times to verify if the laser radiation induced irreversible changes to the spectra. For this we start each experiment with pristine, as-exfoliated flakes that we probe at very low laser power (nW/µm 2 ). We start with low temperature measurements at T=10 K in vacuum (10 −6 mbar) on MoSe 2 samples grown under controlled conditions and exfoliated onto different substrates for comparison. Fig.2 shows the PL spectrum using an extremely low excitation power of 40 nW (black curve) of ML MoSe 2 exfoliated onto SiO 2 /Si from a VPT grown bulk sample. The PL spectrum of the pristine sample is dominated by strong and spectrally narrow (FWHM≈ 8 meV) neutral exciton emission (X) at 1.66 eV. A much smaller peak attributed to the trion (T) is detected at 1.63 eV. The X and T PL energies are in agreement with standard MoSe 2 ML samples [42][43][44]. The PL FWHM for both transitions of just a few meV are among the best reported in the literature and confirm the excellent sample quality of the pristine flakes. After this initial low power measurement, the laser excitation power at this sample position is increased to 40 µW and kept constant during 4 minutes (no measurable evolution is detected for longer times). Directly afterwards, the laser power is lowered to 40 nW, to compare with the black curve before laser treatment, and the PL response is measured (red curve). Remarkably, the trion-to-neutral exciton PL emission intensity ratio T /X has significantly changed, indicating strong doping as a result of the laser treatment.

Panel (a) of
Shown in the inset is the trion's dissociation energy, defined as the difference between the emission energy of the neutral exciton (E X ) and the trion (E T ), as a function of the T /X PL intensity ratio. In a simple picture, the trion's dissociation energy can be written as [45] : is the trion binding energy (typically ∼ 25 meV) and E F is the Fermi level with respect to the bottom of the conduction band for electrons, with respect to the top of the valence band for holes, respectively. The observed linear increase of E X − E T when T /X increases is a signature of an increase of the Fermi level, i.e., of the doping of the ML. Panels , where it can be seen that the X intensity presents a first rapid decrease (within seconds) followed by a second slower decrease in a timescale of several minutes. The trion evolution, in contrast, is marked by an increase on similar timescales. Laser treatment for longer than 4 minutes did not result in any further, measurable evolution of the PL spectrum. The total PL intensity (Trion + exciton) is decreasing in Fig. 3b as more carriers are added, which might be due to the complex interplay between optically bright and dark state of the trion and exciton [46][47][48].  Fig. 4 shows the time-evolution of the PL signal from MoS 2 MLs at T=8 K when exposed to a cw excitation at 60 µW at 532 nm. Three distinct PL emission peaks are observed, associated to the X (1.96 eV),T (1.93 eV), and spectrally broad localized exciton emission [49], which peaks approximately at 1.85 eV. Please note that under laser illumination, both the X and the localized emission decrease, whereas the trion increases (panel (b) of Fig.4).
Again, the dynamics is characterized by a fast and a slow component, of several seconds and several minutes, respectively. Also in this case the total PL intensity decrease as for ML shown for MoSe 2 MLs in Fig. 2a, an increase of the trion's spectral weight is accompanied by an increase of the trion's dissociation energy. In panel (d) of Fig.4, the PL of a MoS 2 ML after laser treatment probed subsequently with 0.4 µW excitation reveals high doping as inferred from the strong T emission (black curve). By illuminating the same region of the ML with an halogen lamp focused into a spot of ∼ 1 µm, the differential reflectivity of this doped region is obtained. The spectrum plotted in Fig. 4d (red curve) indicates that the X transition is still the dominant absorption mechanism and that no significant shift of the X transition is observed with respect to undoped regions. This suggests that no signifi-cant band-gap renormalization is induced by the laser treatment. The same conclusion can be drawn for the MoSe 2 MLs studied, as the X peak PL energy (the optical band gap) in  [29,37]. Please note that in many studies, the broad peak at ∼ 1.9 eV has been attributed to the neutral exciton emission of MoS 2 , and that valleypolarization experiments have been often performed with HeNe laser excitation at 1.96 eV, which corresponds to a perfectly-resonant excitation of the neutral exciton transition. These findings demonstrate that for pulsed excitations, a time-averaged power of only a few µW is enough to change the MoS 2 MLs optical properties in a non-reversible way. This has to be taken into account when analysing the complex physics probed in time-resolved PL measurements [26,50], pump-probe [51][52][53][54] and Kerr rotations experiments [55,56], which are often carried out in this excitation power regime.
Many experiments on MoS 2 with pulsed or cw excitation are carried out at room temperature [57]. We demonstrate the effect of the excitation laser on the optical spectrum also at Discussion.-There are several physical processes that can contribute to the modification of the Trion-to-neutral exciton PL emission ratio due to laser treatment. Behind this lie the different physical origins of excess carriers coming from doping of the TMD material, charges trapped at the ML-substrate interface and molecules on the ML surface.
One possibility is that the laser induced changes are purely electronic i.e. due to optical ionization of impurities. These effects can have lifetimes from fractions of second to days [58]. These physical processes were initially uncovered by Staebler and Wronski in hydrogenated amorphous SiO 2 [59]. One possibility is that the additional charges are optically created from defects in SiO 2 and are subsequently transferred to the TMD ML. In addition, charges trapped at defects could be optically activated in the TMD ML itself, as suggested by photoconductivity measurements [60]. A charge transfer from SiO 2 to the TMD monolayer could be suppressed by insertion of an h-BN layer [61], which might explain the absence of optical doping for this particular sample in Fig. 3c.
Another possibility is that local heating results in defect formation/modification, as suggested by our substrate dependent studies. In Fig 4 we see that the laser treatment does not modify the emission for ML MoSe 2 exfoliated onto h-BN or gold. Their thermal conductivities are much larger than that of SiO 2 , as κ(SiO2-amorphous) ≈ 1 − 2 W/m.K, κ(h-BN) ≈ 300 W/m.K [62] and κ(Au) ≈ 300 W/m.K [63]. This may indicate that the laser-induced doping of TMD MLs is thermally driven. In this case, the contact between the ML and a good thermal conductor may facilitate heat dissipation and as a consequence prevent the laser-induced doping of the ML. As the neutral exciton transition energy did not shift measurably for ML MoSe 2 (Fig. 3a) and ML MoS 2 (Fig. 4a) during laser treatment at T=4 K, strong heating effects are not detected in our experiments. Thermal conductivity is not the only difference between the substrate materials. For example, the atomic flatness of h-BN can also influence charge trapping processes and defect creation/propagation in the ML [64,65].
Although the exact microscopic mechanisms still need to be understood, there are several important practical implications coming from our experiments on the modification of the optical properties of 2D materials following exposure to laser radiation. For thorough optical studies MoX 2 ML samples should be investigated at very low laser power i.e. nW/µm 2 .
When investigating power dependence, hysteresis effects are very likely to occur if the maximum laser power used is too high. Our laser treatment technique could be used to locally pattern doped regions in the 2D crystal, possibly by using patterned substrates (SiO 2 /Si versus h-BN). Our work shows that the neutral exciton emission of ML MoS 2 exfoliated on SiO 2 is at 1.96 eV, i.e. excitation with a HeNe laser is resonant with the neutral exciton transition. This will results in sharp, intense Stokes-lines from resonant Raman scattering [66,67] superimposed on the PL signal. The interplay between laser treatment and super-acid treatment might help in the future to identify how these two techniques influence different type of defects [49,57]. Experiments probing pristine samples that need high laser power to generate enough signal will be difficult to compare with low power measurements such as white light reflectivity, as the optical properties will be modified.