A Single-Step Route to Single-Crystal Molybdenum Disulphide (MoS2) Monolayer domains

We report a simple, single-cycle synthetic method for forming highly-crystalline, micron-sized monolayer domains of phase-pure MoS2. This method combines liquid chemistry with discrete, layer-by-layer deposition from a novel Mo precursor. Single-crystalline MoS2 with domain sizes up to 100 μm have been obtained and characterised by optical and electron microscopy as well as Raman and photoluminescence spectroscopy.

The quartz tube was placed in the middle of the reaction tube, nearest to the thermocouple. The sulphur flakes were placed in a ceramic boat upstream of the substrate container in the cooler part of the reaction tube. Different annealing temperatures were tested, and some samples were post-annealed.
Regions of interest were identified by optical microscopy using an Olympus DX51 microscope with top-view imaging (DP12 digital camera system). Field-emission SEM (FE-SEM, JEOL FESEM6700x) was used to characterise morphology. Morphological and lateral domain size data were obtained by SEM and were complemented by thickness measurements from AFM (JPK NanoWizard ® NanoOptics) in force modulation mode. Raman microscopy (Renishaw RL532C10 InVia microscope) was used to obtain domain thickness information with a 532 nm 500 mW excitation source and a grating of 1800 lines/mm. Room-temperature PL spectra and maps were collected (WITEC photon scanning tunnelling microscope) with a 532 nm excitation wavelength for a quantitative measure of crystal quality. An STEM (FEI Titan 80-300) at an operating voltage of 200 kV was used to visualise the nanostructure of the MoS 2 monolayers. For SEM, samples were characterised on a 35 mm standard sample holder with carbon tape. The material was not sputter-coated with a conductive layer. Images were acquired in secondary electron imaging mode, at an acceleration voltage of 5.0 kV and working distances of 6.4-8.3 mm. In this mode, charging was low-level and acceptable.
TEM samples were prepared by the following method: The as-synthesised films were cleaned by dipping in 2 M aqueous KOH for 5 min. at room temperature. A thin layer of PMMA-950 was spin-coated onto the samples at 2000 rpm for 30 s. The PMMA-coated films were then etched in 2 M aqueous KOH at 75 °C. The material films detached from the SiO 2 /Si substrate after 20 min. The detached PMMA-coated films were collected on TEM grids, and the organic coating was removed by an acetone dip.

Results and Discussion
Optical images as shown in Fig. 1(a) demonstrate MoS 2 growth on bare dice from 0.13 g mL −1 aqueous solutions of (NH 4 ) 6 Mo 7 O 24 .4H 2 O in as little as 15 minutes which is evidenced by the triangular-faceted morphology well-known in CVD-type syntheses of 2D sulphides. Increasing the concentration up to 0.2 g mL −1 yielded mostly bulk and few-layer films, so dilute precursor solutions are preferable. Water was the best solvent because the precursor had the best solubility therein. Acetone and ethanol-water mixtures were found to be unsuitable solvents for ammonium molybdate as the surface coverage was very poor. Table S1 in the supplementary information provides a summary of dip-coating parameters.
FE-SEM as shown in Fig. 1(c) reveal straight-edged triangular morphologies that represent monolayers as well as multi-layer terraces 4,5,7,37,38 . These morphologies are consistent with previous reports of single crystal growth. A straight-edged morphology indicates that Mo-terminated single domains were synthesised by our method (S-rich conditions result in Mo-limited kinetics), whilst S-terminated crystals would result in curved edges 4,7 . A layer thicknesses of 0.6-2.5 nm could be determined by AFM line profiling, consistent with 1-4L growth 5,12,14 .
A selection of 50 domains was sampled from six regions of growth to give a representative lateral domain size distribution (see Fig. 1(b)) estimated from OM. The size distribution was obtained after sulphurisation of two good samples that were dip-coated at 80 °C, and confirms that large domains were obtained by our method, with a modal range of 15-50 μm and several single crystals exceeding 50 μm. The largest crystals had diameters of ~100 μm (see Fig. 1(a)) and provided evidence of large-scale growth. At deposition temperatures below 50 °C and above 80 °C, the precursor did not sufficiently cover the substrate. This resulted in poor sulphurisation and a lack of good-quality domains. It is unclear why high-temperature dip-coating was ineffective. It is possible that the kinetic energy of precursor molecules was greater than the adsorption energy at temperatures near the water boiling point leading to desorption. There might have also been a complementary effect from re-dissolution of the precursor at elevated temperatures.
A clear relationship exists between dip-coating temperature and lateral grain size: The largest crystals (r max ~ 100 μm) were grown by dip-coating at 80 °C. At 70 °C, crystals were much smaller (r max ~ 20 μm) whereas at 50-60 °C the majority of crystals was found to be sub-10 μm in diameter. This trend may be attributed to poorer precursor adsorption at lower temperatures. Our results suggest that dip-coating temperature was the key parameter in lateral domain size control, with a smaller contribution from dip-coating time, and that layer thickness was dependent on precursor solution concentration.
Sulphurisation parameters can be found in Supplementary Table S2. A previous report indicated a trade-off between the lateral and vertical diffusion rates of sulphur vapour over and into the precursor 15 . At low temperatures, the rate of mass diffusion of sulphur vapour over the sample exceeds the rate of mass diffusion into the precursor, which results in poor coverage. This situation is reversed at high temperatures where coverage is improved but films are thicker. However, other authors have found no correlation between synthesis temperature and sample quality 20 . In our work, optimal sulphurisation parameters were identified as comprising ten minutes Prior art References Improvements with our method Surface modification or seeding refs. 13,18,19 Bare, untreated SiO 2 surface gives good-quality crystals Two-substrate method ref. 33 Single substrate improves efficiency Complex precursors ref. 19 Commercially available precursors reduce cost and complexity Multiple deposition cycles refs. 29 Single cycle improves efficiency Polycrystalline film refs. 4,6 Large single crystals better for optoelectronic applications Long/multistep annealing refs. [13][14][15]18,19,22 Single, short anneal gives good-quality crystals Table 1. Improvements on the literature by our method.
www.nature.com/scientificreports www.nature.com/scientificreports/ of annealing at 800 °C under nitrogen (100 sccm) with 1600 mg sulphur flakes (see Supplementary Fig. 1 for TGA of the precursor, which indicates why 800 °C is a good sulphurisation temperature and Supplementary Scheme S3). In contrast, sulphurisation at 600 °C yielded poorer coverage and morphology.
There are two phases of monolayer MoS 2 : the semiconducting 2H phase, which has a direct bandgap, and the metallic 1T phase. These two phases can be distinguished by Raman spectroscopy and PL. The 2H phase possesses a hexagonal lattice and exhibits two characteristic Raman modes of MoS 2 which are the in-plane E 2g and the out-of-plane A 1g , at ~38 cm −1 and ~40 cm −1 , respectively. The 1T phase possesses a tetragonal lattice and exhibits the same Raman modes, along with an additional intense mode at ~33 cm −1 that is forbidden in the 2H phase. Given that the 2H is the thermodynamically stable phase, whilst the 1T is metastable, it is expected that any 1T-MoS 2 synthesised in the sulphurisation process was converted to 2H-MoS 2 during annealing. This is confirmed by both the absence of a peak at 335 cm −1 in the Raman spectra and the presence of strong PL as in Fig. 2 (the 1T phase, being metallic, does not generate PL).
The Raman spectrum of point iii in Fig. 2(a) suggests a layer thickness of 2-3L. The thickness is corroborated by a drop in PL intensity at that point. Variations between ∆υ of discrete monolayer domains is attributed to strain-related effects 39,40 . Thermal diffusion also has a role in the shifting of vibrational energy 41 , and for this reason laser power was controlled at 10% of maximum output. Generally, characteristic peak separation increases with layer thickness up to the 5L limit; bulk MoS 2 exhibits ∆υ ≥ 25.0 cm −1 .
Important information about material quality can be obtained from the ratio of peak intensities, I A /I E , between the A 1g and E 2g modes 42 . Cases in which (I A /I E ) < 1, as shown in Table 2, indicate doping of the material, and this likely comes from oxide impurities relating to the substrate. This can explain the reduction in PL intensity across different monolayer regions of domains in Fig. 2(a,b). Another possible explanation is the existence of grain boundaries, which are known to quench PL by up to 50%. With the exception of point 2(a) iii, the domains sampled show good uniformity.
PL maps as shown in Fig. 2(a,b) reveal domains that are mono/few-layer and have adopted the 2H phase, while 2H produces strong PL. This is corroborated by the Raman peaks (cf. Table 1), which are consistent with production of the 2H phase 43 . Figure 2(c) shows representative PL spectra from MoS 2 regions of different thicknesses, including points ii and iii. The A-exciton emission in the ML case (blue) represents a nine-fold improvement in intensity over the www.nature.com/scientificreports www.nature.com/scientificreports/ corresponding 3L peak (red), and a threefold improvement over the 2L (green) peak. This compares well to previous reports on PL of mechanically-exfoliated samples. By conducting Raman and PL experiments on different areas of the sample, we could not detect significant variation in the Raman and PL signatures. In addition, we have conducted Raman mapping on different samples as detailed in Supplementary Fig. S2, however PL mapping was preferred over Raman mapping due to the higher sensitivity of PL as a function of number of layers, giving a clear image of the overgrown areas.
Our data also compares well with the CVD results from the literature. For instance, Jeon et al. obtained a roughly fourfold improvement in ML-PL intensity over the 2L condition 1 . Our spectra show the expected redshift in emission energy with increasing layer thickness and the corresponding decrease in intensity 44 . The A-exciton energy peak at ~66 nm (1.86 eV) agrees with previous reports of ML-MoS 2 PL emission [1][2][3]45 . The weaker B-exciton peak is also well-resolved in the monolayer on SiO 2 at ~62 nm (~2.00 eV) and arises from the κ-point band splitting due to the valence band spin-orbit coupling. The splitting between the A-and B-excitons in our monolayers is ~14 meV, which is in excellent agreement with the theoretical value (148 meV) for MoS 2 film 46 . This is strong evidence for the excellent crystallinity of our material. www.nature.com/scientificreports www.nature.com/scientificreports/ TEM provides further confirmation of crystalline monolayer formation (Fig. 3). FFT patterns confirm the hexagonal crystal structure of the material, complementing the Raman and PL data that show the 2H phase was synthesised.
The TEM images presented in Fig. 3(a,b) demonstrate that a defect-free crystalline structure has been achieved, which is comparable to results obtained from CVD 12,13 . The diffraction patterns shown in FFT (insets) correspond to the (100) and (110) lattice planes of hexagonal MoS 2 11,12 with an interplanar d-spacings between neighbouring planes of d 100 = 0.27 nm and d 110 = 0.16 nm. These findings are in agreement with the reported spacings for hexagonal MoS 2 11 (see Supplementary Fig. S3 for details of calculations). FFT was performed across a grain boundary of a 100 μm complex-faceted crystal to determine its uniformity, as shown in Fig. 3(c). Previous reports suggest that the butterfly, hourglass and kite morphologies can result from single crystal growth of 2D materials. FFT taken from small regions on either side of the grain boundary were compared to the FFT of the whole 100 × 100 nm region. The three patterns show the same set of six-fold-symmetrical diffraction points and are consistent with single-crystalline monolayer growth 5,13 . These results combined with OM and SEM images suggest that single crystals with lateral sizes up to 100 μm have been synthesised by our method. Key Raman data of domains in Fig. 2 Table 2. Raman modes of PL-active domains in Fig. 2 and data extracted thereof. www.nature.com/scientificreports www.nature.com/scientificreports/

Conclusions
We have demonstrated a simple, single-cycle and single-substrate route to large single-crystal domains (up to 100 μm) of MoS 2 using a novel Mo precursor that requires no surface pre-treatment. We anticipate that grain size can be further improved by careful control of substrate and sulphurisation parameters. The single-cycle nature of our process and short annealing time could lead to a significant improvement in the cost and volume of large-scale MoS 2 production.

Data Availability
All data generated or analysed during this study are included in this published article (and its Supplementary  Information files).