Vacuum level dependent photoluminescence in chemical vapor deposition-grown monolayer MoS2

The stronger photoluminescence (PL) in chemical vapor deposition (CVD) grown monolayer MoS2 has been attributed to its high crystal quality compared with that in mechanically exfoliated (ME) crystal, which is contrary to the cognition that the ME crystal usually have better crystal quality than that of CVD grown one and it is expected with a better optical quality. In this report, the reason of abnormally strong PL spectra in CVD grown monolayer crystal is systematically investigated by studying the in-situ opto-electrical exploration at various environments for both of CVD and ME samples. High resolution transmission electron microscopy is used to investigate their crystal qualities. The stronger PL in CVD grown crystal is due to the high p-doping effect of adsorbates induced rebalance of exciton/trion emission. The first principle calculations are carried out to explore the interaction between adsorbates in ambient and defects sites in MoS2, which is consistent to the experimental phenomenon and further confirm our proposed mechanisms.

spectra of monolayer MoS 2 samples with different amounts of defects, as well as their electrical performance, we uncover the emission mechanism of abnormally strong PL observed in CVD grown samples and attribute it to the p doping effect from the larger amounts of adsorbates on the defect sites of MoS 2 , which rebalance the radiative emission intensities of excitons. Figure 1a shows the PL spectra of typical CVD grown and ME monolayer MoS 2 . Their corresponding PL mappings show the uniform intensities. Clearly, the PL intensity of the former is stronger than that of the latter, which was previously attributed to the high crystal quality of CVD grown crystal 16,17 . The X-ray photoelectron spectroscopies for these two groups of samples are carried out to show the atomic ratio (S/Mo), as shown in Fig. S1. For CVD grown sample, the peaks at 162.3 and 163.4 eV are assigned to the S 2− 2p 3/2 and 2p 1/2 , respectively. The peaks at 229.5 and 232.7 eV are attributed to the Mo 4+ 3d 5/2 and 3d 3/2 . While for ME sample, the peaks at 162.6 and 163.6 eV are assigned to the 2p3/2 and 2p1/2, of divalent sulphide ions (S 2− ) and the peaks at 229.67 and 232.8 eV are from the 3d 5/2 and 3d 3/2 of the core levels of Mo 4+ . The S/Mo ratio for CVD sample is 1.66 and that for ME sample is 1.86, which is consistent to the results of EDS and more defects are introduced in CVD grown samples. Figure 1b,c show typical high resolution atomic structure generated by high angle annular dark field (HAADF) scanning transmission electron microscopy (STEM), and they clearly display the defects in CVD grown monolayer MoS 2 , as indicated in the lines. The lattice spaces of hexagonal lattice structure for MoS 2 is 2.7 Å ((100) plane) and 1.6 Å ((110) plane). The corresponding intensity line profiles shows in Fig. S2 display most of defects are single S vacancy. Fig. S3a gives a directly visual sense of the effect of defects on the PL intensities of ME MoS 2 . The marked circle area with pre-exposed by Argon (Ar) plasma shows stronger PL intensity. Fig. S3b shows the PL spectra of area A (exposed by Ar plasma) and B (without Ar plasma treatment), respectively. Therefore, it Left column: optical image of a typical monolayer sample fabricated by ME method and its PL mapping. Right column: optical image of a typical CVDgrown monolayer sample and its PL mapping. Both of two PL mappings are normalized to the intensity of A exciton. (b) and (c) are the HAADF image of ME and CVD samples, respectively. The lines indicate the defects observed in this image. seems that the more defects contribute to the stronger PL emission in MoS 2 . Nevertheless, whether the PL spectra observed in MoS 2 is its intrinsic emission or defect-induced emission remains unclear. The previous reports have revealed the energy of defect-induced PL is around 1.78 eV, at the low energy side of the reported A exciton 22 . However, as shown in Fig. 1, there is no extra emission peak except a broadened shoulder, which has been identified as the trion emission, marked as A − exciton emission. Moreover, the temperature dependent PL spectra (shown in Fig. S4) indicates the PL peak observed is an intrinsic emission, and the match among the electron spin, the layer pseudospin and the valley pseudospin reported in the literature display the features of intrinsic emission [6][7][8][9]24 . Thus the observed PL peak in Fig. 1 is not the defect-induced emission in MoS 2 . Another potential factor that may lead to the difference of optical properties between CVD grown and ME samples is the strain effect from the substrate since the contact behaviours between the substrate and the sample are different for these two cases. The CVD grown samples are inevitably affected by the strain effect because of the different thermal expansion coefficients between deposited layer and substrate during the cooling process [25][26][27] . Figure S5 compares the Raman spectra of both of CVD grown and ME sample. The E 2g 1 and A 1g modes of CVD grown samples show a red-shift and blue-shift, respectively, when compared with those of ME 1 L MoS 2 sample. This phenomenon differs from the variation trend when strain is applied 28 . Also, if the strain effect is the main factor in the PL spectra of MoS 2 , the energies of PL would shift clearly towards higher energy (compressive strain) and lower energy (tensile strain), respectively 29,30 , but there is no obvious shift of peak position shown in Fig. 1. Thus, the strain effect could be excluded in this study. So far, defects are most likely to be responsible for the enhancement of photon emission in CVD grown monolayer MoS 2 .

Results and Discussion
Since defects in atomically layered materials usually could act as adsorption centres due to their higher active adsorption energy, the PL spectra of CVD and ME MoS 2 sample were measured in vacuum to exclude the role of adsorbates. Surprisingly, as shown in Fig. 2a, the PL intensity of CVD grown MoS 2 decreased dramatically in vacuum. At the same time, the emission intensities are reversible when the CVD grown samples exposed in air again. However, for the PL spectra of ME sample (Fig. 2b), its intensity does not show obvious variations with the ambient varies. The reduction of PL intensity in vacuum for CVD grown sample implies that the adsorbates adsorbed on the surface of MoS 2 greatly affect its photon emission. According to the vacuum-level dependent PL intensities, it seems that more adsorbates contribute to stronger PL intensity. Moreover, it is found that in Fig. 2c, the ratio of PL intensities between CVD grown and ME MoS 2 samples is even less than 1 when the adsorbates level is at the lowest vacuum level (1 × 10 −5 mbar). This result shows that the PL intensity of CVD grown sample is comparable or even weaker than that of ME sample, which verifies that the room-temperature optical performance in vacuum could reflect the crystal quality. Figure 2d,e show the PL images of a typical CVD grown sample with grain boundary in air and vacuum, respectively. The sites at the grain boundary are expected with more defects and its decreased PL intensity measured in vacuum provides clear evidence that PL intensities at defect sites could be decreased when measured in vacuum.
Meanwhile, the in-situ Raman spectra have been obtained for these two groups of MoS 2 samples, as shown in Fig. S6. There is no obvious change for the E 1 2g mode (almost fixed at ~385.7 cm −1 for ME sample and ~384.4 for CVD grown sample), but a clear blue-shift of A 1g mode (from 404.94 cm −1 to 403.35 cm −1 ) for the CVD sample in vacuum compared with that in air, shown in Fig. S6a, which is larger than that of ME sample with a smaller shift from 404.21 cm −1 to 403.71 cm −1 (Fig. S6b). This result is different from the reported strain effect on the Raman spectra for monolayer MoS 2 sample 28 , which could be further exclude the strain effect. The similar variation trends for A 1g and E 1 2g mode have been reported by the gated Raman spectra from a 1 L MoS 2

31
. The results show the softening of the A 1g mode with electron doping, while the frequency of E 1 2g mode remains unchanged, which are due to the stronger electron-phonon coupling of the A 1g mode than that of the E 1 2g mode 32,33 . So whether the shift of A 1g modeobserved in this work is due to the doping effect will be discussed later.
The vacuum-level dependent electrical performances of CVD grown and ME MoS 2 are investigated. Figure 3a shows the schematic of a monolayer MoS 2 based transistor fabricated in this work. The fabrication process is described in the experimental section. The optical images of FETs based on ME and CVD grown samples are shown in Fig. S7. Figure 3b,c show the electrical performance of single layer ME and CVD grown MoS 2 based transistors, respectively. Both of these two groups transistors display n type behaviour, which mean excess electrons introduced in MoS 2 . Moreover, with the vacuum level decreasing, the threshold voltage shifts towards the positive direction, which means that the adsorbates in air play a p doping effect on MoS 2 itself. Additionally, comparing the current at V g = 0 V, the current is larger under vacuum when compared with that in air, which represents a stronger n doped effect in vacuum. While in air, due to the p doping effect from absorbates, the current is decreased. Both of CVD grown and ME samples show this variation trend. The mobility for CVD grown samples in vacuum is 0.69 cm 2 /VS while it is 44.2 cm 2 /VS for ME samples in vacuum, which means a poor crystal quality for CVD grown samples without doping effect from adsorbates. While this p-doping effect due to the adsorbates in air reduced the amounts of the excess electron in MoS 2 and thus increased the radiative decay of exciton emission, which could be described by a three-level model including the excitation, exciton emission and trion emission processes 4,34 . As shown in Fig. S8a, G represents the generation rate of optical excitons. The radiative decay rates of the exciton and trion are marked as Γ ex and tr Γ respectively. When the mass action law is considered with trions together to evaluate the doped electron density in MoS 2 samples, the integrated intensity ratio of trion emission (I tr ) to electron emission (I ex ) can be expressed as 34-37 : tr ex e b I tr and I ex represent the trion and exciton emission intensities, respectively. n e is the carrier concentration, k b is the Boltzmann constant, and T represents temperature.
From equation (1), it is confirmed that the ratio of emission intensities from I tr and I ex indeed highly depends on the carrier concentration in MoS 2 , which corresponds to the experimental results shown in Fig. S8b. The fitting results for CVD grown and ME samples in air and vacuum are shown in Fig. S9, respectively, which directly reflect the vacuum level dependent intensity ratio of exciton and trion. Moreover, it is concluded that the binding energy of trion is ~24 meV. So their emission spectra overlapped with that of exciton emission since such negatively charged excitons usually have finite binding energies, ~20 meV 4 .
Moreover, such a p doing effect also explain well the Raman frequency shift of E 1 2g and A 1g mode observed in Fig. S6. According to the symmetrical group theory, the A 1g mode has symmetrical lattice variation and would have a nonzero expectation value for the matrix element of the electron-phonon coupling, which leads to a larger electron-phonon coupling 31 . While for the E 1 2g mode, the matric element of electron phonon coupling vanishes and its coupling with electrons is weaker on doping in ambient compared with A 1g modes [31][32][33] . Moreover, we have calculated the phonon energies of E 1 2g mode and A 1g mode with and without considering the adsorbates, as shown in Fig. S10, the results are consistent to our experimental results and explanation above.
So far, the physical picture for the reason why the CVD grown sample shows a strong PL intensity is clear. The greater the amounts of adsorbates, the stronger the intensity of the PL emission in MoS 2 . Supposing there is no adsorption on the surface of MoS 2 , the PL intensity from CVD grown MoS 2 should be weaker than that of ME sample due to its poorer crystal quality, which is corresponding to the experimental results on the intensity ratio (0.74) at higher vacuum shown in Fig. 2c. Furthermore, the CVD grown monolayer MoS 2 under different growth conditions marked as S1, S2, S3, S4, and S5, respectively, are investigated to confirm the vacuum level dependent PL spectra. Figure 4a shows the optical images of selected CVD samples. Their PL intensities are different in air (Fig. 4b), and their emission intensities are reduced measured in vacuum (1 × 10 −5 mbar) by different extent, as shown in Fig. 4c. Figure 4d presents the ratios of PL intensities measured in air and vacuum for these samples. The different ratios when the vacuum level varies during measurement are due to their different amounts of adsorbates on the samples. The ratios of exciton and trion emissions for these five samples are also shown in Fig. 4d and the variation tread is almost the same as their vacuum-level dependent PL intensities. Additionally, compared with that in air, the ratios of exciton and trion emissions are decreased in vacuum which further confirms that the amounts of adsorbates are highly related to the PL intensity and the ratios of exciton and trion emission intensities in MoS 2 and the optical performance of MoS 2 in vacuum are positively related to its crystal quality.
To clarify the most likely components of adsorbates in air and how the adsorbates contribute to the p doping effect on MoS 2 , we have evaluated the charge transfer between adsorbates (O 2 , N 2 , OH − and H 2 O molecules) and MoS 2 by performing first-principle calculations as these three kinds of molecules are bountiful in air. We first consider the defects in MoS 2 . As shown in Fig. S11, we consider four typical types of defects by using a 4 × 4 supercell of MoS 2 as the prototype: (1) a single S vacancy; (2) a single Mo vacancy; and (3) and (4)

Conclusion
In summary, we clarify the origin of the abnormally strong PL in CVD grown samples and attribute it to the p doping effects of adsorbates at the defect-sites. HRTEM and EDS are used to confirm more defects in CVD grown samples than that in ME samples, and the in-situ vacuum level dependent PL and Raman spectra, as well as the electrical performance confirms our proposal. The CVD samples under different grown conditions are expected with different crystal qualities, which show similar variation trends by different extent which further confirm the observed phenomenon. First principle calculations were carried out to investigate the charge transfer process between MoS 2 and adsorbates and clarify the p-doping effect of adsorbates due to the strong electronegativity of adsorbates. Such a p-doping effect from adsorbates reduces the concentration of excess electron in MoS 2 and contributes to the radiative recombination process of exciton. This work proposes a new vacuum technique to . The growth temperature is at 650 °C and this temperature is kept with 15 min. Raman and PL spectra were conducted on Witec CRM 300 confocal Raman microscopy. The excitation laser was 532 nm laser with a spot size about 500 nm. The output power measured from the objective lens was controlled below 0.5 mW to avoid damaging and heating on samples. The accumulation time for PL and Raman spectra was 10 s. The Raman spectra were collected by 1800 g/mm grating while PL spectra were collected by 300 g/mm grating. The size of the probe X-ray beam is 200 μm. For sample viewing, firstly, an optical image is taken by an external camera. Based on that optical image, we zoom in and position to roughly the area that we want. Then, similar to a SEM/EDX, a finely focused x-ray beam is used to create a secondary electron image for sample viewing, and actual positioning for interested analysis area.
Fabrication of Back-Gated MoS 2 Transistors. Both of these two groups of MoS 2 samples were deposited on 270 nm SiO 2 /Si substrates. The devices were fabricated by standard electron beam lithography, followed by the thermal evaporation of Cr/Au electrode, and lift-off process. The electric measurement system was carried out using Agilent B1500A semiconductor analyzer. The vacuum level is set at 1 × 10 −5 Torr.
First Principle Calculations. The first-principles calculations were realized based on the density functional theory (DFT), as implemented in the Vienna Ab initio Simulation Package (VASP) 41 . The exchange-correlation potential is chosen as generalized gradient approximation (GGA), formulated by Perdew-Burke-Ernzerhof (PBE) functional 42,43 . The cut off energy of 400 eV is used for the plane-wave expansion of valence electron-wave functions. To avoid artificial interaction between layers, a vacuum spacing of >15 Å is built. In the calculations, DFT-D2 method is employed to describe the long-range van der Waals interactions 44 . For both defects  identifications and gas molecules adsorptions on the monolayer MoS 2 , the calculations were performed on a 4 × 4 × 1 size of MoS 2 supercell, with Monkhorst-Pack k-point meshes of 7 × 7 × 1 45 . The convergence criteria for energy and force are set to be 10 −5 eV and 0.01 eV Å −1 , respectively. The amount of charge transfers between gas molecules and MoS 2 monolayer is estimated by using the Bader charge method.