The rise of atomically thin materials has the potential to enable a paradigm shift in modern technologies by introducing multi-functional materials in the semiconductor industry. To date the growth of high quality atomically thin semiconductors (e.g. WS2) is one of the most pressing challenges to unleash the potential of these materials and the growth of mono- or bi-layers with high crystal quality is yet to see its full realization. Here, we show that the novel use of molecular precursors in the controlled synthesis of mono- and bi-layer WS2 leads to superior material quality compared to the widely used direct sulfidization of WO3-based precursors. Record high room temperature charge carrier mobility up to 52 cm2/Vs and ultra-sharp photoluminescence linewidth of just 36 meV over submillimeter areas demonstrate that the quality of this material supersedes also that of naturally occurring materials. By exploiting surface diffusion kinetics of W and S species adsorbed onto a substrate, a deterministic layer thickness control has also been achieved promoting the design of scalable synthesis routes.
Atomically thin layers of metal group VI disulfides and diselenides (MoS2, WS2, WSe2, MoSe2,) are being extensively investigated as they present unconventional optoelectronic properties compared to commonly used low-dimensional semiconductors1,2. In the bulk form they are layered compounds formed of covalently bonded chalcogen and metal atoms forming tri-atomic layers, which are held together by van der Waals forces3. An individual tri-atom thick layer presents a direct band gap in the visible-near IR range4 on the contrary to the bulk, which manifests an indirect electronic band gap. Monolayer sulfides and selenides show strong light absorption from the visible to the near IR range1,5,6, valley polarization7,8, second-harmonic generation9, tightly bound excitons10 and strong spin-orbit interaction11,12. These properties arise from their intrinsic two-dimensional nature inherently free from dangling bonds and their particular d-orbitals configuration3,13. Further, given their atomically thin nature they are mechanically flexible and they can sustain tensile strain of 20%14,15.
One of the most promising transition metal dichalcogenides (TMDCs) is WS2 owing to light emission in the monolayer form at ~2 eV and the low level of toxicity of growth processes. Any envisioned application relies on materials with high crystal and optical quality extended over wafer-size areas. Chemical vapour deposition (CVD) is a scalable method for materials synthesis and it is being widely employed for TMDCs16,17. The synthesis of tungsten-based materials has revealed to be challenging and generally leading to isolated flakes of lateral size between 5–40 μm4,18,19,20,21,22,23. Monolayer WS2 films extended over centimeter-sized areas has been demonstrated24,25, however with compromised crystal quality as indicated by the low carrier mobilities. The growth is typically performed by co-evaporating sulfur powder and a W-precursor in a horizontal tubular furnace in presence of a carrier gas. Until now, the synthesis of WS2 has been achieved predominantly using WO3 and S powders4,18,19,20,21,22,23 at temperatures greater than 900 °C. This involves a topotactic-like transformation26 which normally yields to sparsely distributed WS2 domains onto an amorphous4,18,20,22,23 or crystalline substrate19,21,27.
Here we demonstrate the synthesis of high quality monolayer WS2 using carbon-free molecular precursors. The high crystal quality is manifested by the record high charge carrier mobilities of mono- and bi-layer WS2 and ultra-sharp PL linewidth at room temperature, which are superior to those of naturally occurring materials. The growth is enabled by molecular precursors, which lead to a complete sulfidization of W and formation of WS2 with lower number of defects compared to the traditionally used direct sulfidization of WO3.
Results and Discussion
The synthesis of WS2 was performed starting from commercial powders of either H2WO4 (hydrated tungsten oxide) or WO3, sulfur and where indicated, we have introduced NaCl. W and S precursors were placed in two separate crucibles well-spaced in a quartz tubular furnace (Supporting Information, Figure S1a) and heated up independently using different controllers as reported in Figure S1b. The heating profile of S powders has been optimized to ensure maximum supply when the W-precursors start evaporating. WS2 was grown on Si/SiO2 (285 nm) substrates loaded in the downstream zone of the tubular furnace. Altering the chemistry of decomposition of tungsten oxide species, we could achieve synthesis of monolayer WS2 over larger area coverage, at low temperatures and with low amount of defects.
Optical micrographs of WS2 monolayers (Fig. 1) grown using different tungsten oxide precursors systems at different temperatures (950 °C, 850 °C and 750 °C) show distinctively increasing lateral size of the triangles and increasingly facilitated synthesis at low temperatures from WO3 to WO3-NaCl and H2WO4-NaCl. The possibility to grow WS2 from WO3 at temperatures not lower than 950 °C is notoriously attributed to the high sublimation temperature of the oxide17,28,29,30. In specific, the size of the WS2 triangles increases from ~10 μm, to 60 μm, upto 200 μm with areas of continuous polycrystalline monolayered coverage of ~0.8 mm (Figure S2). Further, the WS2 synthesis at temperatures as low as 750 °C was enabled only by the precursors system of H2WO4-NaCl with remarkable triangular size of ~100 μm. In addition, increasing the growth pressure (from ~1 mbar to 13 mbar) at 950 °C, bilayered WS2 flakes are preferentially formed (Figure S3) using the H2WO4-NaCl system.
It is worth noting that larger WS2 domains obtained at 950 °C, as compared to 850 and 750 °C, can be explained in the light of the Robinson & Robin model31. At high temperature the diffusivity of the adsorbed precursors on the SiO2 surface is favourable leading to the expansion of the existing domains. At the same time, desorption of absorbed species is higher than at lower temperatures, limiting the achievement of a supersaturation concentration and thus reducing the nucleation density.
From structural investigation of the reaction products in the different precursors systems, we could elucidate the role played by the water intercalated in the H2WO4 and how this can enable the favorable synthesis at low temperature and with low density of defects as compared to the WO3-based precursors. From X-ray diffraction (XRD) characterization (Supporting Information) it was possible to observe that the main products of the reactions between NaCl and either H2WO4 or WO3 are similar: NaxWyOz and tungsten oxychlorides (WClO4 and WO2Cl2). However the reactions occur at significantly lower temperatures for H2WO4 compared to WO3. While NaxWyOz species possesses a high evaporation temperature, as they are found in the crucible (Figure S5) at the end of the synthesis of WS2, the tungsten oxychlorides are volatile (Figures S4, S5) and they can possibly play a key role in promoting the synthesis of WS2. Indeed the system H2WO4-NaCl is likely to enable the formation of tungsten-oxyhalide species (i.e., WO2Cl2, WOCl4) at lower temperatures than WO3-NaCl, as NaCl dissociation is promoted by the H2O molecules gradually released by H2WO4 upon heating. On the bases of previous studies on the synthesis of WS2 bulk crystals, the formation of tungsten oxychlorides (WO2Cl2 and WOCl4) is indeed likely to occur with higher chances with respect to the formation of metal halides (e.g. WCl6)32,33. Tungsten oxychlorides (WO2Cl2 and WOCl4) can be volatile from 200 °C34 and they can be sulfidized in vapour phase forming a few-atom clusters of W-S which can deposit onto the target substrate as adatom species35 where they can form WS2 via a diffusion-desorption mediated mechanism of nucleation and growth. WOCl4 has been previously used36 as precursor for the synthesis of WS2 bulk films. Despite its strong tungsten-oxygen double bonds, WOCl4 proved to be an effective precursor with a clean decomposition pathway without formation of tungsten oxysulfides. We have verified that using this precursor is indeed possible to obtain WS2 at temperatures as low as 550 °C (Figure S6a). The key role played by the oxyhalide species becomes apparent if we try to grow WS2 by using H2WO4 as single precursor. As this decomposes to form WO3, only small WS2 domains are observed with PL characteristics similar to the WO3 precursor-led growth (Figure S6b).
High-resolution transmission electron microscopy (HRTEM) imaging confirms the high crystalline nature of the material (Fig. 2a). The measured lattice constant is ~0.3 nm consistent with that of 2H-WS2 (a = 0.318 nm). The Raman spectra of WS2 obtained using the different precursors systems are shown in Fig. 2b. All of the spectra exhibit two characteristic peaks located at ~(351 ± 0.53) cm−1 and ~(417.6 ± 1) cm−1, which can be attributed to 2LA-E1 2g and A1g Raman modes of pristine WS2 monolayer37,38. Interestingly, the distribution of the peak positions is increasingly narrower from WO3 precursor, WO3 + NaCl and to H2WO4 + NaCl (Figures S8, S10). The Raman peaks intensities are uniform across the entire triangle area (Figures S7, S9) and the frequency difference (Δν) between 2LA(M) and A1g modes is ~(66.5 ± 0.53) cm−1 (Figure S11), as expected for monolayer WS2 37. The AFM thickness profile analysis of WS2 triangles confirms the monolayer (Fig. 2c,d) and bilayer (Fig. 2e,f) nature of the flakes, showing an edge step height of ~0.8 nm (Fig. 2d) and ~1.6 nm (Fig. 2f), respectively39,40.
A comparison of representative photoluminescence (PL) intensity maps of WS2 monolayers grown under the three precursors systems at different temperatures is reported in Fig. 3. The PL peak intensity appears consistently higher for the NaCl-based precursors system as compared to WO3. The intensity variation pattern across an individual flake is not yet fully understood41, and it is likely to be due to different defect concentrations in the form of sulfur vacancies. The FWHM of the PL peaks significantly decreases from ~75 meV, to ~50 meV, to ~36 meV for the three precursors systems, WO3, WO3 + NaCl, and H2WO4 + NaCl respectively, with a distribution significantly narrower and more uniform across the same flake and different flakes (SI). It is worth noting that 36 meV of FHWM is narrower that mechanically exfoliated material (Fig. 4a) which is ~59 meV. This suggests that WS2 grown by using H2WO4 + NaCl possesses less structural defects compared to the other precursors systems, and specifically in the form of sulfur vacancies. Indeed, it has been reported that S vacancies increase the electron density, thus the trions population and the strength of the PL emission at lower energies than the optical band gap with a consequent increase of the FWHM4,42. The PL peak position progressively blueshifts from 1.94 eV, to 1.96 eV, to 1.98 eV, reaffirming a decreased amount of S vacancies and the formation of a progressively more pristine material.
A molecular conversion based-growth mechanism, were tungsten oxyhalide molecules are sulfidized in vapour phase, versus a topotactic-like conversion of WO3 in WS2 can explain the different defects contents in WS2. The greater efficiency of H2WO4 in inducing a complete sulfidization of the precursors compared with the WO3 + NaCl system has been also confirmed by chemical analysis (X-ray photoelectron spectroscopy). Analysing WS2 grown using H2WO4 + NaCl, the W 4f5/2 and W 4f7/2 core levels (Fig. 5a) present peak position characteristic of W4+ in WS2 43,44 (32.7 and 34.8 eV respectively) and the narrowest achievable FWHM (1 eV) (Fig. 5a), using the Mg Kα as X-ray source. This indicates chemical purity and expected stoichiometric ratio of W and S. This has been also confirmed by calculating the concentration of S and W from the integrated intensity of the W 4 f and S 2p core levels. The S 2p1/2 and 2p3/2 core levels, also appear at the expected position for WS2 (162.3 eV and 163.4 eV respectively, Figure 5b)44 and with a very narrow FWHM (1 eV) (Fig. 5c). A very small amount of W6+ (W 4f5/2 and W 4f7/2 core levels centred at 35.9 eV and 38.1 eV respectively in Fig. 5a) attributable to WO3, which partially overlaps with the W 5p core level (38.5 eV), can be observed which however disappears after transferring the flakes on a new SiO2/Si substrate (Fig. 5c) and thus suggesting that it is related to residual precursors on the substrate considering the XPS spot size is ~1 mm. It is worth noting that after transfer the FWHM of the W 4 f core levels remains unchanged suggesting that the transfer process preserves the crystallinity of the flakes and no additional defects are introduced.
Similarly, chemical purity and expected stoichiometric ratio of 2:1 for S:W have been observed for WS2 grown from WO3 + NaCl (Fig. 5a). Nevertheless, a larger W6+ contribution, attributable to WO3 (W 4f5/2 and W 4f7/2 centred at 35.9 eV and 38.1 eV respectively in Fig. 5a) has been detected in this case suggesting that a conspicuous amount of precursors does not get sulfurized and it is just deposited onto the SiO2 wafer. Upon transfer on a new SiO2/Si substrate, this component entirely disappears (Fig. 5c), thus indicating also in this case that WO3 is mainly distributed on the substrate. The FWHM of the W4+ 4 f core levels is ~1.2 eV in this case, suggesting higher concentration of defects compared to H2WO4 + NaCl-led growth (Fig. 5a). The transferred WS2 present a FWHM even larger ~1.3 eV, suggesting the introduction of atomic defects as a consequence of the mechanical stress underwent by the flakes with preexisting defects (Fig. 5c). To conclude, XPS study confirms the effectiveness of H2WO4 as precursor versus WO3.
The progressive reduction of structural defects from WO3 to H2WO4 has been proven by electrical characterization. The electrical properties of the WS2 flakes were characterised through their performance in bottom-gated field effect transistors (FET) (Fig. 6a,b). The FET transfer curve (Fig. 6c) displays an accumulation-type n-channel transistor, where the current flowing through the channel increases with increased gate bias, after the threshold voltage.
The field-effect mobility was calculated in the linear region of the transport graph (marked with red-dashed line in Fig. 6c), using µ n = C ox −1(dσ/dV gs). Overall, monolayer WS2 grown using H2WO4 + NaCl shows electron mobilities systematically higher compared with the WO3 + NaCl system (Fig. 6c,e,f) corroborating the fact that higher crystal quality is expected by using H2WO4 as precursor. Further, monolayer WS2 presents electron mobility of ~(28 ± 1.4) cm2/Vs (Fig. 6c,f) which is the highest mobility reported so far for CVD grown WS2 deposited onto SiO2 (Fig. 7a)19,21,23,24,25,30,45,46,47,48,49 and comparable to mechanically exfoliated WS2 37,50,51,52. The highest mobilities using either H2WO4 + NaCl or WO3 + NaCl are displayed at 950 °C (Fig. 6f) suggesting that the growth temperature does also play a role in improving the crystal quality of the material. While the role played by the different precursors systems in determining the crystallinity of the synthesis product becomes more prominent at low growth temperatures. Monolayer WS2 grown using H2WO4 + NaCl exhibits electron mobilities of ~(10 ± 1) cm2/Vs at temperatures between 750 °C and 900 °C. While the electron mobilities of monolayer WS2 grown using WO3 + NaCl at 850 °C (Fig. 6f) present lower values of ~(0.4 ± 0.1) cm2/Vs. Bilayer WS2 shows electron mobility systematically higher than monolayer and also systematically higher than mechanically exfoliated bilayered flakes52,53 (Fig. 7b). The electron mobility of ~(52 ± 4) cm2/Vs (Figs 7 b, 8a,b,c) represents a record mobility as compared to CVD grown or mechanically exfoliated bilayer WS2 onto SiO2 reported so far52,53,54. The fact that the highest mobility for bilayer WS2 has been obtained using WO3 + NaCl with no use of H2WO4 suggests that the bilayer system is less affected by the precursos choice, and a bilayered material presents in general crystal quality superior to monolayers.
In conclusion, we have developed a synthesis strategy which enables high crystal quality WS2 as reflected in the high optical quality and in the carrier mobility that overcome naturally occurring materials. The molecular precursors approach leads to effective sulfidization of W, revealing to be highly advantageous with respect to the traditional oxide–based conversion synthesis of WS2. These results can be translated and applied to the synthesis of different TMDCs, and pave the way towards industrially scalable synthesis of monolayer WS2 over large areas.
CVD Synthesis of WS2
Commercial WO3 (0.3 g, 99.9%, Sigma Aldrich), H2WO4 (0.3 g, 99.9%, Sigma Aldrich) and NaCl (0.3 g, ≥99.5%, Sigma Aldrich) powders were loaded in an alumina boat placed in the center of a 2 inch-diameter horizontal quartz tube CVD furnace. While an alumina boat containing sulfur powders (0.6 g, ≥99.5%, Sigma Aldrich) was loaded in the upstream zone of the tube, whose temperature was independently controlled by a different heater. The growth substrates were Si wafers (500 microns) with on top 285 nm of SiO2 thermally deposited. The substrates were sequentially cleaned for 15 min in acetone, isopropanol and deionized water in a sonicator, followed by dipping in H2SO4/H2O2 (3:1) for two hours and rising in deionized water. They were then blow dried with nitrogen gas, cleaned with O2 plasma for 5 min and loaded into the downstream zone of the furnace. The CVD growths were then performed at low pressure (~10−1 mbar) and under flow of high purity Ar gas (flow rate of 100 sccm). The furnace was heated to 750–950 °C with a ramp rate of 25 °C min−1, kept at the growth temperatures for 15 min and then naturally cooled down to room temperature. The sulfur powder was independently heated to 125 °C with a ramp of 5 °C min−1, kept at this temperature for 15 min and naturally cooled down.
The transfer procedure was performed by depositing a PMMA film 350 nm thick onto the target sample, which was subsequently immersed in a KOH solution (0.1 M) until detachment of the PMMA from the SiO2/Si substrate. The PMMA/WS2 films was then scooped out with a new Si/SiO2 substrate, repeatedly washed in deionized water and then immerged in an acetone bath at 45 °C for 20 min to dissolve the PMMA film.
TEM analysis of the WS2 flakes was carried out on a FEI Titan 80–300 S/TEM operated at 80 kV, equipped with a monochromator and a Cs aberration image corrector. Focal series micrographs of the representative flake were acquired at different objective lens focus values (using a spherical aberration coefficient Cs ~ −4 µm) and exit-wave reconstruction was performed using TrueImage software (FEI).
Raman and photoluminescence spectra were collected using a Renishaw inVia spectrometer equipped with a 532 nm laser excitation. All the spatial maps were collected under a 100x objective using grating of 1800 line/mm, which provide a resolution of ~1.5 cm−1.
X-ray Photoelectron Spectroscopy
X-Ray photoemission spectra were acquired in a custom made ultra-high vacuum system (pressure < 10−9 mbar) equipped with a VG Escalab Mk-II electron analyzer and a twin anode (Al/Mg) non-monochromatized x-ray source (Omicron DAR 400). All measurements were taken in quasi-normal emission (5° off) at room temperature using a pass-energy of 20 eV, an energy step of 0.1 eV and the Mg Kα emission line as exciting radiation.
For the field-effect mobility measurements, single and bilayer WS2 field-effect transistors (FET) were fabricated after transfer to a new Si/SiO2 substrate where the Si is highly p-doped and acts as a global gate electrode, and a 285 nm thick thermally grown SiO2 serves as the gate dielectric. The FETs were fabricated on dried Si/SiO2/WS2 samples. Besides the current bearing contacts, termed here “Source” and “Drain”, two additional voltage probes were added to each FETs to allow for an accurate determination of the channel’s conductance by eliminating the contribution of the contacts. Both the current bearing leads and the voltage probes were patterned simultaneously using a standard electron beam lithography process. The source, drain and voltage probes consisted of 50 nm Au, while the electronics leads consisted of 5 nm Ti and 50 nm Au. A two-steps annealing process followed the fabrication. The samples were first annealed for 2 hours at 200 °C under H2/Ar (10/90) flow in atmospheric pressure, to etch residues of PMMA that was used as an electron resist for the lithography step. Then, the samples were put under a high vacuum (~10−6 mbar) and baked at 115 °C for 60 hours, to promote desorption of water molecules from the channel surface.
The FETs were measured inside the vacuum chamber at a constant pressure of ~10−6 mbar, without exposure to ambient conditions after the second annealing step. The drain electrode was biased with a low noise source-meter and the source electrode was grounded throughout the experiment. An additional source-meter was used to bias the global gate electrode, with respect to the source. The transistor current, I ds, was measured using an ammeter and the potential difference across the voltage probes, V A-B, was measured with a voltmeter. The channel conductivity, σ, is then readily obtained using σ = (L I ds)/(W V A-B), where W and L are the channel’s width and length, respectively. The measurement set-up is shown schematically (not to scale) in Fig. 6a. The oxide capacitance was estimated to be 115 µFm−2 from C ox = ε0 ε r/d ox, where d ox is the oxide thickness and ε0 and εr are the vacuum permittivity and SiO2 relative permittivity, respectively.
Data availability statement
All relevant data generated or analysed during this study are included in this article and its Supplementary Information file.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
C.M. would like to acknowledge the EPSRC awards EP/K033840/1, EP/K01658X/1, EP/K016792/1, EP/M022250/1, the EPSRC-Royal Society Fellowship Engagement Grant EP/L003481/1 and the award of a Royal Society University Research Fellowship by the UK Royal Society. N.N. acknowledge the Imperial College Junior Research Fellowship and P.C.S. would like to acknowledge the funding and support from the European Commission (H2020 – Marie Sklodowska Curie European Fellowship–660721). I.A. acknowledges financial support from The European Commission Marie Curie Individual Fellowships (Grant number 701704). J.D.M. acknowledges the financial support from the Engineering and Physical Sciences Research Council (EPSRC) of the United Kingdom, via the EPSRC Centre for Doctoral Training in Metamaterials (Grant No. EP/L015331/1). S.R. and M.F.C acknowledge financial support from EPSRC (Grant no. EP/J000396/1, EP/K017160/1, P/K010050/1, EP/G036101/1, EP/M001024/1, EP/M002438/1), from Royal Society international Exchanges scheme 2016/R1 and from The Leverhulme trust (grant title “Quantum Drums” and “Room temperature quantum electronics”).