Effect of Ultraviolet-Ozone Treatment on MoS2 Monolayers: Comparison of Chemical-Vapor-Deposited Polycrystalline Thin Films and Mechanically Exfoliated Single Crystal Flakes

We report the different oxidation behavior between polycrystalline chemical-vapor-deposited and mechanically exfoliated single crystal MoS2 monolayers by ultraviolet-ozone treatment. As ultraviolet-ozone treatment time increased from 0 to 5 min, photoluminescence emission and Raman modes of both MoS2 disappeared, suggesting structural degradation by oxidation. Analysis with optical absorbance and X-ray photoelectron spectroscopy suggested the formation of MoO3 in both MoS2 after ultraviolet-ozone treatment. In addition, ultraviolet-ozone treatment possibly led to the formation of oxygen vacancies, molybdenum oxysulfide, or molybdenum sulfates in chemical-vapor-deposited MoS2. The measurement of electrical resistance after ultraviolet-ozone treatment suggested the transformation of chemical-vapor-deposited MoS2 into doped MoO3 and of mechanically exfoliated MoS2 into negligibly doped MoO3. These results demonstrate that the crystallinity of monolayer MoS2 can strongly influence the effect of ultraviolet-ozone treatment, providing important implications on the device integration of MoS2 and other two-dimensional semiconductors. Electronic supplementary material The online version of this article (10.1186/s11671-019-3119-3) contains supplementary material, which is available to authorized users.


Introduction
There is a great interest in transition metal dichalcogenides (TMDs), such as MoS 2 , since they offer an attractive possibility for various device applications including transistors, optoelectronic devices, heterojunction structures, sensors, and electrocatalysis [1,2]. The existence of direct bandgaps in monolayer TMDs makes these twodimensional semiconductors especially promising for optoelectronic devices [3,4]. However, critical challenges to fabricate TMD-based optoelectronic devices such as phototransistors include the deposition of high-k dielectrics on TMDs and the doping of TMDs. Because of the absence of dangling bonds on the surface of TMDs, it is challenging to deposit high-k dielectrics on TMDs [5]. Moreover, the doping of TMDs is also challenging as the substitutional doping used for bulk semiconductors such as silicon modifies the two-dimensional structure and properties of monolayer TMDs [6].
To overcome these difficulties, surface functionalization of TMDs by O 2 plasma [7,8] or ultraviolet-ozone (UV-O 3 ) [9][10][11] has been suggested. While these methods can functionalize the surface of MoS 2 by surface oxidation, they can simultaneously influence the structure and properties of monolayer MoS 2 [12][13][14][15][16]. For example, oxidation by O 2 plasma or UV-O 3 treatment altered the Raman vibration modes and photoluminescence (PL) emission of monolayer MoS 2 [12,16]. However, as most studies were based on micrometer-scale monolayer MoS 2 flakes obtained by mechanical exfoliation from bulk single crystals, little has been known on their interaction with large-area monolayer MoS 2 thin films, which are typically polycrystalline. Grain boundaries in polycrystalline monolayer MoS 2 may allow higher reactivity with UV-O 3 than that of single crystal, resulting in different oxidation behavior. Therefore, in this study, we explore the effect of UV-O 3 treatment on MoS 2 monolayers by directly comparing the oxidation behavior of polycrystalline chemical vapor deposition (CVD) thin films and mechanically exfoliated single crystal flakes. We systematically investigate the PL and Raman spectra of both MoS 2 monolayers for different duration of UV-O 3 exposure. We also investigate the oxidation behavior of both MoS 2 monolayers during UV-O 3 treatment with X-ray photoelectron spectroscopy (XPS). We further measure electrical resistance of pristine and UV-O 3 -treated MoS 2 monolayers to understand the effect of UV-O 3 treatment on MoS 2 monolayers.

Methods
Monolayer MoS 2 thin films were deposited on (0001)oriented sapphire substrates (~1.5 × 1 cm 2 ) by CVD in a two-zone tube furnace. MoO 3 (99.98%, Sigma-Aldrich) and S (99.98%, Sigma-Aldrich) powders in two separate Al 2 O 3 boats were used as precursors. MoO 3 powder (14 mg) was placed upstream at zone 1 (750°C) and S powder (1.4 g) was placed at the upstream entry of the furnace. Substrates were placed downstream at zone 2 (700°C). MoO 3 powder was heated at a rate of 15°C min −1 and substrates were heated at 38°C min −1 . After 30-min deposition, the furnace was slowly cooled down to room temperature. Ar flow of 100 sccm and a pressure of~0.5 Torr were maintained during deposition. Monolayer MoS 2 flakes were obtained by the gold-mediated exfoliation method [17] from bulk MoS 2 crystals (2D Semiconductors) and transferred on highlydoped Si substrates with thermally grown SiO 2 (300 nm). Figure 1 shows schematic structures of both MoS 2 monolayers on substrates. The thickness of monolayer MoS 2 was measured using atomic force microscopy (AFM, Park Systems XE-100). The crystallinity of bulk MoS 2 crystals and CVD MoS 2 thin films was investigated by X-ray diffraction (XRD, Bruker D8 Discover with Cu-Kα radiation) and transmission electron microscopy (TEM, FEI Titan 80-300 at 300 kV), respectively. MoS 2 monolayers were exposed to UV-O 3 (SEN LIGHTS PL16-110, 185 nm and 254 nm) for 0-5 min at the irradiance of 58 mW cm −2 . Optical absorbance was measured by UV-visible spectroscopy (PerkinElmer Lambda 35). Raman/PL spectroscopy (Horiba Jobin-Yvon LabRam Aramis) were measured on pristine and UV-O 3 -treated MoS 2 monolayers with a 532-nm laser and a beam power of 0.5 mW. XPS (Thermo Scientific K-Alpha) was carried out using a monochromatic Al K α x-ray source (hν = 1486.7 eV) with a take-off angle of 45°, a pass energy of 40 eV, and a spot size of 400 μm in diameter. For all samples, C 1s and O 1s were observed presumably because they are exposed to atmosphere before loaded to ultrahigh vacuum chamber for XPS analysis. Adventitious carbon (C 1s at 284.8 eV) was used as a charge correction reference for XPS spectra. The energy resolution is 0.7 eV measured using the full width at half-maximum intensity of the Ag 3d 5/2 peak. MoS 2 samples were exposed to atmosphere while they were brought to XPS equipment. Although in situ XPS analysis could provide more accurate information, it was unavailable in this work. For peak deconvolution and background subtraction, Thermo Scientific Avantage Data System software was used. Gaussian functions were used to fit XPS spectra.
To measure the electrical resistance of MoS 2 monolayers, Au contacts (100 × 100 μm 2 , 70 nm thick) were deposited on top of MoS 2 by electronbeam evaporation. Spin-coated photoresist on top of Au layer was then patterned by conventional photolithography to form opening areas for subsequent etching. After Au in opening areas was removed by wet etching in aqua regia, remaining photoresist was removed in acetone. Then, the devices were annealed at 200°C for 2 h in a tube furnace (100 sccm Ar and 10 sccm H 2 ) to remove photoresist residue and to decrease contact resistance. Electrical resistance was calculated with current-voltage (I-V) measurement (Keithley 4200-SCS) in atmospheric environments.

Results and Discussion
Beside AFM measurement, PL and Raman spectra are measured to confirm the formation of MoS 2 monolayers. Because of its direct bandgap, MoS 2 monolayers allow PL emission at~1.88 eV [3,4]. In addition, the frequency difference between the two characteristic Raman A 1g and E 1 2g modes of MoS 2 monolayers is less than 20 cm −1 [18]. In Fig. 3, the PL emission of pristine MoS 2 at~1.88 eV indicates that both MoS 2 are monolayers. In Fig. 4, pristine MoS 2 exhibits the frequency difference between 19.6 and 19.9 cm −1 implying monolayer MoS 2 . XRD and TEM analysis indicated the single crystal nature of bulk MoS 2 crystals and polycrystalline nature of our monolayer MoS 2 thin films (Additional file 1: Figure S1). The grain size of monolayer MoS 2 thin films is~10 nm [19].
After UV-O 3 treatment, MoS 2 monolayers change their color and become transparent. In Fig. 2a   This slight difference is probably due to the effect of underlying substrates as substrates can strongly influence the Raman and PL emission [21]. The wider width of PL emission peak in CVD monolayers suggests higher defect density. Interestingly, further negative shift of PL emission peak is observed in single crystal MoS 2 flakes (~50 meV) than in CVD thin films (by~10 meV) after UV-O 3 treatment. As the negative shift of PL emission is comparable with trion binding energy (10-40 meV) of MoS 2 [22], this may be due to different concentrations of trion (neutral excitons accepting an electron or a hole) formed by oxidation-induced doping [23,24]. (In this work, single crystal MoS 2 flake is more conductive than CVD MoS 2 , suggesting higher doping levels in single crystal MoS 2 .) The higher doping level in single crystal MoS 2 flakes will allow high concentration of trions, of which recombination will dominate their PL emission. In contrast, the lower doping level in CVD MoS 2 thin films will allow low concentration of trions. Hence, their PL emission will be dominated by the recombination of neutral excitons. However, as the negative shift of PL emission may also be related to the effect of underlying substrates or strains, more systematic investigation is needed in the future.
Next, to investigate the structural degradation by UV-O 3 treatment, we measure the Raman spectra of MoS 2 monolayers after UV-O 3 treatment for 0, 1, 3, and 5 min, respectively (Fig. 4). The intensity of both E 1 2g and A 1g modes decreases as the treatment time increases. While the frequency difference between E 1 2g and A 1g modes remains unchanged for 0-5 min of UV-O 3 treatment time, the two Raman modes almost completely disappear after 5-min treatment, suggesting severe structural distortion and degradation. AFM analysis indicates an increase of surface roughness after UV-O 3 treatment (Additional file 1: Figure S2), which is consistent with the oxidation of MoS 2 [23].
To further investigate the structural degradation of MoS 2 monolayers by UV-O 3 treatment, we measure XPS spectra of MoS 2 . Because the beam size of XPS is much larger than the size of single-layer MoS 2 flakes, XPS spectra for single crystal MoS 2 flakes are obtained from large-area MoS 2 single crystals (~1 cm in size and1 00 μm in thickness). Figure 5 shows the XPS spectra in  suggesting higher n-type doping [25]. The peak shift after the oxidation of MoS 2 in this work (0.41-1.09 eV) is comparable to that in literature (0.6-1.1 eV) [23,24]. (To prevent charging effect, which may induce similar positive shift, we used a flood gun during XPS measurement.) (4) In CVD MoS 2 thin films, the peaks of Mo 5+state also appear with UV-O 3 treatment suggesting possibly the formation of oxygen vacancies [26] or molybdenum oxysulfide MoO x S y [27]. These results can be understood by the oxidation of Mo 4+ -state in MoS 2 into higher oxidation states (Mo 5+ and Mo 6+ ) with UV-O 3 exposure. This is also consistent with the XPS results on polycrystalline multilayer MoS 2 thin films after O 2 plasma or UV-O 3 treatment [26,28,29].
In S 2p region, the existence of S 2− -state can be observed from the binding energy of S 2p 1/2 and S 2p 3/2 orbitals in pristine MoS 2 . The binding energy of S 2 − -state in single crystal MoS 2 shows further positive shift than that in CVD MoS 2 thin films suggesting higher ntype doping [25]. Although S-O bond is observed atF ig. 5 XPS spectra of MoS 2 a bulk single crystal and b CVD thin films on sapphire substrates after UV-O 3 treatment for 0, 1, 3, and 5 min 165 eV in UV-O 3 -treated single crystal MoS 2 , it is below the detection limit in CVD thin films. Instead, a new doublet peak of sulfur oxidation state appears at higher binding energy (~169 eV) in CVD thin films after UV-O 3 treatment for 3 min. This new doublet corresponds to the S 2p peaks of oxidized sulfur S 6+ , suggesting possibly the formation of various molybdenum sulfates Mo (SO 4 ) x [28]. While the intensity of S 2− doublet keeps decreasing with longer UV-O 3 exposure, the intensity of S 6+ doublet further increases after 5-min UV-O 3 treatment, suggesting further conversion of S 2− into higher oxidation state (S 6+ ) by oxidation. Similarly with Mo 4+ peaks, the intensity of S 2− peaks does not change with UV-O 3 treatment time in large MoS 2 single crystals. The existence of S 6+ -state after O 2 plasma or UV-O 3 treatment is inconsistent in literature. Its existence was reported in polycrystalline multilayer MoS 2 thin films after O 2 plasma treatment [28]. However, it was not observed in other polycrystalline multilayer MoS 2 thin films [26,29] or single crystals [9,16,30] after O 2 plasma or UV-O 3 treatment. While this inconsistency may be related to dose-and time-dependence of MoS 2 oxidation [30], more systematic investigation is needed to clarify this in the future.
The different XPS behavior may be related to the difference of composition and crystallinity between single crystals and CVD thin films. The composition of Mo:S is 1:1.97 in bulk single crystals and 1:1.5 in CVD thin films, suggesting higher concentration of S vacancies in CVD thin films. The higher concentration of S vacancies, combined with the existence of grain boundaries in CVD thin films, may allow higher reactivity to oxygen than that in single crystals.
To further understand the oxidation of MoS 2 monolayers by UV-O 3 treatment, we measure the electrical resistance of pristine and UV-O 3 -treated MoS 2 monolayers. Because there is sample-to-sample variation of electrical resistance, we use relative ratio of electrical resistance (R After /R Before ), where R After and R Before are electrical resistance after and before UV-O 3 treatment, respectively. Figure 6 shows R After /R Before as a function of UV-O 3 treatment time. While R After /R Before of exfoliated MoS 2 single crystal flakes significantly increases with longer treatment time, R After /R Before of CVD MoS 2 thin films decreases with longer treatment time. These results suggest that MoO 3 formed by the UV-O 3 treatment of CVD MoS 2 thin films possesses higher doping level than that of MoS 2 single crystal flakes. This is supported by XPS analysis suggesting the possible existence of oxygen vacancies, MoO x S y , or Mo (SO 4 ) x in CVD MoS 2 monolayers. This is seemingly contradicting with the higher doping in single crystal MoS 2 suggested in Fig. 5a. However, as Fig. 5a is based on bulk single crystals, we cannot exclude the possibility that it does not provide accurate information of the top monolayer. Hence, surface oxidation of bulk MoS 2 single crystal may possibly provide doping only to MoS 2 single crystal underneath, transforming top surface region into negligibly doped MoO 3 . Consistent with these results, electrical resistance also increased when monolayer MoS 2 single crystal flakes were oxidized by O 2 plasma [12]. As single crystal MoS 2 without grain boundaries could be more tolerant to oxidation than polycrystalline MoS 2 , the effect of oxidation-induced doping may be stronger in polycrystalline MoS 2 than in single crystal MoS 2 . However, further investigation is needed to understand this difference in the future.

Conclusions
In summary, we investigated the effect of UV-O 3 treatment on polycrystalline CVD thin films and single crystal flakes of monolayer MoS 2 . Monolayer MoS 2 becomes transparent after UV-O 3 treatment suggesting the formation of wide bandgap semiconductor MoO 3 . As UV-O 3 treatment time increases, the intensity of PL and Raman spectra significantly decreased, suggesting the formation of oxides or defects. In both MoS 2 , XPS analysis indicated the formation of Mo-O bonds and MoO 3 . However, in CVD MoS 2 thin films, the conversion of Mo 4+ -and S 2− -states into Mo 5+ -and S 6+ -states was also observed after UV-O 3 treatment, suggesting the possible existence of oxygen vacancies, MoO x S y , or Mo (SO 4 ) x . As the electrical resistance of single crystal MoS 2 monolayers significantly increased with longer UV-O 3 treatment time, the oxidation of single crystal MoS 2 into MoO 3 seems to provide negligible doping. In contrast, the electrical resistance of CVD MoS 2 monolayers decreased with longer UV-O 3 treatment time, suggesting that the oxidation of CVD MoS 2 into MoO 3 provides doping. These results demonstrate the significant impact of crystallinity on the effect of UV-O 3 treatment on MoS 2 monolayers, providing possibly interesting implications on fabricating heterojunction structures based on two-dimensional nanomaterials.