The synthesis of large-area molybdenum disulfide (MoS2) films is desired for many emerging electronic and optical applications. A typical approach is to deposit films from vapor-phase precursors either by thermal- or plasma-activated chemistry. Here, we show that MoS2 films exhibiting excellent crystallinity and smooth surfaces are produced by plasma conversion of solution-deposited precursor films. Films containing a single-source precursor for MoS2, ammonium tetrathiomolybdate, were spin coated and converted by an atmospheric-pressure dielectric barrier discharge with substrate heating at 500 °C. X-ray diffraction and micro-Raman spectroscopy show that the plasma is necessary to successfully convert the precursor to MoS2. Further enhancements in the crystallinity and surface roughness were obtained by annealing the films, as revealed by high-resolution transmission electron microscopy and atomic force microscopy (AFM). Insights into the role of the plasma were provided by AFM characterization and a “plasma-trigger” experiment, which suggest that a nanocone surface morphology facilitates the growth of the films.
One of Peter’s research interests was surface analysis of compound semiconductors including zinc oxide (ZnO), which, like molybdenum disulfide (MoS2), is a layered material. Our study of ultrathin, dimensionally confined layered materials serves as a natural extension of Peter’s work and legacy.
The synthesis of large-area, continuous, and uniformly thick ultrathin films of layered materials such as MoS2 is of interest for emerging electronics,1 optoelectronics,2 energy harvesting,3 and energy storage applications.4 A general strategy that has shown promise is vapor deposition in which either a single precursor or a mixture of precursors are deposited from the gas phase onto a substrate.5 The most well-studied example is chemical vapor deposition (CVD), where typically solid or liquid-source precursors are heated and carried by a stream of gas through a flow furnace to be thermally decomposed at the surface of the substrate. CVD requires relatively high temperatures, >800 °C, and specific substrates to control film nucleation and growth, but when optimized has led to the fabrication of highly crystalline MoS2 films with thicknesses ranging from multilayer to monolayer films and areas in the range of a few millimeters-squared.6 Recently, liquid precursors have also been introduced by atomization, and by controlling the substrate temperature and ratio of precursors, produced continuous and large-area crystalline MoS2 films.7 Alternatively, atomic layer deposition (ALD) has been explored to more carefully control the thickness of MoS2 films but has not yet achieved the crystalline quality of CVD.8,9 To potentially lower the growth temperature and improve the film crystallinity, these methods have been modified by incorporating a plasma to decompose the precursor in the gas phase by electron impact and enhance film nucleation and growth by ion-surface interactions. Both plasma-enhanced CVD (PECVD)10 and plasma-enhanced ALD (PEALD)11 have demonstrated that the growth temperature for MoS2 films can be lowered to around 150–250 °C.
An alternative approach to vapor deposition is film conversion, where a precursor is initially deposited at ambient conditions by solution processing and subsequently converted into crystalline MoS2 by heating. The potential of converting ammonium tetrathiomolybdate (ATM), a single-source precursor for MoS2, was first demonstrated by Putz and Aegerter.12 More recently, Liu et al. reported a two-step conversion process consisting of reductive heating at low pressures followed by high-temperature annealing of dip-coated ATM films.13 Yang et al. showed, that by adding a polymer, linear polyethylenimine (L-PEI), wafer-scale, thickness-controlled MoS2 films could be produced in one heating step without needing to add sulfur vapor.14 Other additives have also been explored such as ethylenediaminetetraacetic acid (EDTA) which similarly acts as a chelating agent to improve the wetting of ATM.15 Overall, in comparison to vapor deposition, the potential advantages of precursor film conversion are that: (1) the adsorption step is circumvented which can lower the temperature required for nucleation/growth and allow arbitrary substrates,16 (2) the film thickness/layer number and uniformity is easily controlled by the initial precursor film,14 and (3) the approach bears similarity to additive manufacturing methods in that the material is produced only where it is desired with minimal wasteage.17
Our group recently introduced a modified version of precursor film conversion termed plasma-enhanced chemical film conversion (PECFC), which is analogous to PECVD and PEALD and incorporates a plasma with thermal treatment. The synthesis of large-area hexagonal boron-nitride (h-BN) films directly on silicon (Si) substrates from ammonia borane was demonstrated, and the addition of a plasma was found to lower the minimum temperature for film formation and improve the film crystallinity compared to thermal conversion alone.16 The same method was extended to MoS2, with the plasma allowing crystalline films to be produced from ATM, to serve as a hydrogen evolution reaction electrocatalyst.18 Here, we present a more in-depth study of the nucleation/growth process and properties of MoS2 films fabricated by PECFC. Thin films of MoS2 were synthesized by spin-coating solutions of ATM onto Si substrates with 300 nm oxide and converting with an atmospheric-pressure dielectric barrier discharge (DBD) while heating the substrate. As a point-of-comparison, precursor film conversion was also carried out by only heating. Characterization by x-ray diffraction and micro-Raman spectroscopy showed that the plasma is necessary to produce crystalline MoS2 films. The plasma-converted films were found to be approximately six layers in thickness and exhibit a nanocrystalline morphology characterized by ∼10 nm-sized conical-shaped grains on the surface. To potentially further improve the film structure, the MoS2 films were annealed postconversion by heat treatment. Micro-Raman spectroscopy showed that annealing improves the film crystallinity substantially, and atomic force microscopy revealed that during annealing, the film evolves by first, growth of the nanocones into sharper features, followed by coalescence into a few large islands, and finally a smooth film with an RMS of <0.5 nm. A similar evolution in film growth was not observed for thermally converted films. The potential role of the plasma was further clarified by treating films after thermal conversion with a plasma for a short time. These “plasma-triggered” films were found to grow into highly crystalline films similar to the plasma-converted films by annealing. These results suggest that the plasma uniquely induces surface structuring during precursor film conversion, thereby enabling the growth of a smooth, crystalline film.
To synthesize MoS2 films by PECFC, there were two overall steps: first, the preparation of a precursor film by solution processing and second, the conversion of the precursor film to crystalline MoS2 by plasma treatment with substrate heating. Solutions containing the precursor were prepared by mixing 5 ml of 200 mM ATM (Sigma Aldrich) in dimethylformamide (DMF) with 3 ml of ∼3.4 wt. % 10 000 MW linear polyethylenimine (Sigma Aldrich) in DMF. To completely dissolve the resultant reddish-brown precipitate, 2 ml of ethanolamine (Sigma Aldrich) was added and the mixture was sonicated at 70 °C for 1 h, which finally produced a homogeneous dark red solution. Precursor films were deposited from solution by spin coating at 3000 rpm for 30 s on single-crystal  Si substrates with a 300 nm thick oxide. The substrates were initially cleaned prior to film deposition by ultrasonication in acetone, followed by isopropanol, and then finally de-ionized (DI) water. To remove organic contaminants, the substrates were further cleaned by immersing in a piranha solution (3:1 H2SO4/H2O2 v/v) for 2 h and then rinsing repeatedly with isopropanol and DI water. Finally, the substrates were treated by a low-pressure RF plasma in oxygen (O2) at 150 W for 3 min to enhance wettability of the oxide layer. Following deposition, the precursor films were dried in a vacuum chamber overnight to remove any remaining moisture.
The films of ATM precursor were converted into crystalline MoS2 in a homemade atmospheric-pressure, planar DBD reactor, consisting of a ceramic covering 2 in. in diameter that served as the power electrode and a substrate heater (MHI, Inc.) that served as the grounded electrode, in a parallel plate geometry. The plasma was driven by a high-voltage AC power source at 25 kHz and was found to cover the entire ceramic covering and substrate at 8 kV peak-to-peak voltage. At atmospheric pressure and in Ar, the DBD is expected to be filamentary, but filament formation was minimized by decreasing the gap between the electrodes to ∼3–4 mm and to the naked eye, appeared highly uniform. In addition, adding H2 was found to decrease filament formation. All conversion experiments were carried out in an Ar/H2 (80:20) background at a substrate temperature of 500 °C for 1 h. Control experiments were conducted without applying power to ignite the plasma and converting the precursor films by only thermal treatment in the exact same background gas environment. Following precursor conversion, the films were annealed by heating in an Ar background for 30 min.
Converted MoS2 films were characterized ex situ by several materials analysis techniques. X-ray diffraction was performed using a Bruker D8 Discover Series II X-ray diffractometer with a Cu-Kα radiation source. Micro-Raman spectroscopy was conducted at room temperature using a Horiba LabRAM HR800 confocal Raman microscope system equipped with a thermoelectric-cooled CCD camera. The excitation wavelength of the laser was 532 nm and the power was kept below 1 mW to avoid heating of the sample. The laser was focused by a 100× objective lens down to a spot size of ∼1 μm, and using an 1800 groove/mm grating, a spectral resolution of ∼0.5 cm−1 was achieved. X-ray photoelectron spectroscopy (XPS) was carried out with a PHI Versaprobe 5000 using an Al-Kα radiation source. The atomic ratios of Mo and S in the films were calculated by integrating the areas under the Mo 3d and S 2p peaks after correcting for the background and calibrating the peak positions using the C-1 s peak at 284.8 eV, which most likely originated from adventitious carbon. High resolution transmission electron microscopy (HRTEM) images were obtained using a FEI Tecnai F30 at an accelerating voltage of 300 kV. For HRTEM imaging, MoS2 films were synthesized on Si substrates without the 300 nm thick oxide to allow removal by etching in 30 wt. % KOH. The resulting free-standing MoS2 films were scooped and transferred onto ultrathin-carbon coated TEM grids. Atomic force microscopy (AFM) images were obtained using an Agilent 5500 scanning probe microscope in a tapping mode to avoid damage to the films.
III. RESULTS AND DISCUSSION
A process flow diagram for the synthesis of MoS2 films by PECFC is shown in Fig. 1(a). The first step was to spin coat a solution containing the precursor, ATM, along with a complexing anionic polymer, L-PEI [Fig. 1(a), left]. The ATM precursor-containing film was converted into MoS2 by plasma treatment, specifically using an atmospheric-pressure DBD operated in a gas mixture of 80:20 Ar:H2 with the substrate heated to 500 °C for 1 h [Fig. 1(a), middle]. To further improve the properties of the MoS2 film including surface roughness and crystallinity, the films were thermally annealed in a background of Ar for 30 min [Fig. 1(a), right].
Conversion of the ATM precursor to MoS2 could be visibly observed by the color change from blue-green to deep blue as shown in Figs. 1(b) and 1(c), respectively. The formation of MoS2 was more carefully assessed by micro-Raman spectroscopy. Figure 1(d) shows micro-Raman spectra for the as-converted film, as well as films after annealing postconversion at 500, 800, and 1000 °C. In general, all films exhibit peaks at ∼381.5 and 407.1 cm−1 corresponding to the E12g and A1g vibrational modes of 2H-MoS2, which confirms conversion.19 Increasing the temperature for annealing is found to increase the intensities and decrease the full-width-half-maximums (FWHMs) of both peaks, indicating an increase in the degree of crystallinity. Prior studies have shown that larger crystalline domain sizes lead to stronger Raman mode intensities and lower peak FWHMs.16
The crystalline structure of the converted films was further characterized by XRD. Figure 2(a) shows XRD patterns of the plasma-converted film and the same film after thermally annealing at 1000 °C. In addition, the diffractogram for a film converted only by heating (no plasma) is shown for comparison. Plasma conversion is found to produce a diffraction peak at ∼17° that, based on the Cu-Kα source, can be indexed to the (002) crystal plane of 2H-MoS2 (JCPDS 24-0513). The peak intensity increases and the FWHM decreases after annealing, consistent with an increase in the crystallinity of the film, which is in agreement with micro-Raman analysis. In comparison, no diffraction peaks are observed for the thermally treated film, either for MoS2 or the ATM precursor (Fig. S1, in the supplementary material37), suggesting that without the plasma, there is conversion but only to amorphous MoS2 or MoS3, the latter of which is known to be an amorphous intermediate formed during the thermal decomposition of ATM.20
Figure 2(b) shows a comparison of the micro-Raman spectra of plasma-converted, annealed, and thermally treated films. As in the case of XRD, the E12g and A1g peaks corresponding to 2H-MoS2 are absent in the thermally treated film. In few-layer MoS2, the difference in frequencies between the E12g and A1g modes, Δk, depends on the number of layers.19 We measured Δk to be ∼25 cm−1 for the plasma-converted film before and after annealing, which corresponds to a thickness of approximately six layers. The spectrum for the annealed plasma-converted sample shows low frequency Raman modes below 50 cm−1 that are characteristic of interlayer shear and breathing modes21 [Fig. 2(b) inset]. A comparison with exfoliated 2H-MoS2 films of varying layer number indicates that the layer number is ∼4-8, in agreement with that obtained from the E12g and A1g frequency difference. The appearance of low-frequency interlayer Raman modes supports the formation of a well-stacked, layered crystal structure by the combination of PECFC and annealing.22
The stoichiometry of the MoS2 films was determined by XPS. Figures 2(c) and 2(d) show representative high-resolution XPS spectra of the energy regions corresponding to Mo 3d and S 2p binding, respectively, for the plasma-converted and annealed films. The peaks at ∼233.0 and ∼229.8 eV correspond to Mo 3d3/2 and Mo 3d5/2, respectively, and the peaks at ∼163.3 and ∼162.3 eV correspond to S 2p1/2 and S 2p3/2, respectively.23 We estimated an average S:Mo ratio in the films by correcting the background of the spectra using the Shirley fitting method and integrating the areas under the Mo 3d and S 2p peaks measured from multiple spots across a film (Fig. S2 and Tables S1 and S2 in the supplementary material37). The S:Mo ratio for the plasma-converted sample is 1.96 ± 0.03 and after annealing, 1.97 ± 0.04, confirming that the conversion produces stoichiometric MoS2 with a very small variation of ∼2% across the film. We note that substantial sulfur vacancies which have been reported for postsynthesis plasma treatment of MoS2 films were not observed, most probably because of the processing at atmospheric pressure which reduces the energy of bombarding species such as ions.24–26
HRTEM was carried out to characterize the nanoscale morphology of the MoS2 films. The representative lower and higher magnification images of the plasma-converted films in Figs. 3(a) and 3(b) show crystalline regions, and the SAED pattern [inset of Fig. 3(d)] shows concentric rings, consistent with a nanocrystalline structure.27 Following annealing, relatively large single-crystalline domains >30 nm could be found as shown in Fig. 3(e). High-resolution images revealed well-ordered stacks of atoms in these domains corresponding to an ABAB-type stacking sequence characteristic of the hexagonal close-packed atomic structure of 2H-MoS2 [Figs. 3(d) and Figs. S3(a) and S3(b), in the supplementary material37]. The SAED pattern of these films exhibits discrete spots which can be indexed to crystalline MoS2 [inset of Fig. 3(d) and Fig. S3c, in the supplementary material37] and indicates an increase in crystallinity after annealing, which agrees with XRD and micro-Raman analysis. The HRTEM results clarify that annealing enhances crystallinity via the growth of larger-sized grains in the film. In comparison, representative lower and higher magnification images in Figs. 3(e) and 3(f), respectively, of the thermally treated films show no evidence of a crystalline structure. In combination with the lack of any diffraction in the SAED pattern [Fig. 3(f) inset], the HRTEM results are in agreement with XRD and micro-Raman that thermal conversion alone without a plasma does not produce crystalline MoS2.
Our results indicate that the plasma is critical for the conversion of the ATM precursor to crystalline MoS2. Previous studies have demonstrated successful conversion by only thermal treatment (no plasma).13–15 The thermal decomposition of ATM is believed to proceed through a two-step reaction:20
In an inert atmosphere such as Ar, the initial conversion of ATM shown in Eq. (1) initiates at 120–360 °C, leading to the formation of the solid intermediate amorphous MoS3 (a-MoS3). The subsequent conversion of a-MoS3 to MoS2 shown in Eq. (2) occurs by heating to temperatures ∼800 °C. In the presence of H2, the temperature for conversion of a-MoS3 to MoS2 is reduced to ∼450 °C. The exact temperature shows some variability and is sensitive to reaction conditions such as heating rate.20,28 However, MoS2 decomposes at temperatures >500 °C in the presence of H2 gas.13 Thus, there is a relatively narrow thermal window for the conversion of the a-MoS3 intermediate to MoS2 without any loss of the film, which is why all conversions in our study were carried out with H2 at 500 °C, and higher temperature annealing was carried out in a separate step with only pure Ar. We note that a key difference between our experiments and previous studies was the pressure. As the thermal decomposition of ATM to MoS2 over the two steps occurs through a gradual loss of sulfur to the gas phase,14 the volatility of sulfur atoms is important. Liu et al. carried out conversion of ATM to MoS2 at 1 Torr, while other groups also performed conversion at subatmospheric pressures.14,15 Our process was conducted at atmospheric pressure, where the vapor pressure of sulfur is lower. The resulting lower volatility of sulfur atoms could have hindered the sulfur-loss reaction and impeded the conversion of a-MoS3 to MoS2. This hypothesis is supported by our XRD, Raman, and HRTEM analysis, which suggest that our thermal treatment of ATM at 500 °C produces an amorphous intermediate phase. In comparison, a plasma may help remove sulfur through different surface processes such as reactions involving atomic hydrogen, which have been previously found to be important in the conversion of ammonia borane films to h-BN,16 or etching/sputtering by ion bombardment, leading to the formation of MoS2.
In addition to conversion of the ATM precursor to crystalline MoS2, another critical part of the process is the growth of the film, particularly during the annealing step. To provide further insight into film growth, we performed AFM characterization of the films after annealing at 1000 °C for different times. The initial plasma-converted film is found to exhibit a nanostructured surface morphology with an RMS roughness of ∼1.9 nm [Fig. 4(a)] which is consistent across the film [Fig. S4(a), in the supplementary material37], in agreement with HRTEM analysis. Scanning at higher resolutions reveals that the surface is composed of nanocones with a height of approximately 10 nm [Fig. S5(a), in the supplementary material37]. We suggest that these nanocones are individual MoS2 crystals as has been previously reported by other growth methods.29–31 After annealing for only 5 min, these nanocones grow, mostly in the vertical direction, producing a serrated surface with an RMS roughness of ∼6.3 nm [Fig. 4(b)]. Between 5 and 10 min, we find another change as the nanostructured surface disappears and is replaced by a low density of large islands of ∼100 nm2 in size. After 30 min of annealing, the film surface becomes very smooth with an RMS roughness of <0.5 nm [Figs. 4(d) and S4b, in the supplementary material37] that compares favorably with the smoothest films reported to date.32,33 These features are noticeably unique to plasma conversion; the initial-thermally treated films showed much larger surface features and a corresponding RMS roughness of ∼7.6 nm (Fig. S5b, in the supplementary material37). Annealing did not lead to growth in the vertical direction or large islands, but more gradual changes that eventually produced a rough film with an RMS roughness of 8.2 nm (Fig. S6, in the supplementary material37).
We note that the initial nanocone surface morphology after plasma conversion, as well as the subsequent transition into a smooth film after annealing, is observed only in the plasma-converted MoS2 film. It is likely that the plasma is responsible for the formation of these features, which is important in producing a smooth final film in our process. Gonzalez-Gonzalez et al. reported similar cone-shaped structures during the pulsed laser deposition of  ZnO films and developed a mechanism for their formation termed “slope-selection-driven Ostwald ripening,” in which initially mound-shaped grains grow asymmetrically because of preferential coarsening along the facets.29 The initial growth trajectory observed in our films during the first 5 min of annealing, where the nanocones grow larger, may be explained by a similar slope-selection-driven Ostwald ripening mechanism. Between 5 and 10 min, continued Ostwald ripening could lead to the formation of the large islands. The final transition of the surface from these islands to an ultrasmooth film between 10 and 30 min is a more abrupt change in the growth trajectory and must be explained by a mechanism unrelated to Ostwald ripening. Previous reports have suggested that larger features in multilayer films can coarsen by alternate mechanisms such as island relaxation or grain boundary migration.34–36 One such possible mechanism to explain this late-stage morphological transition in our annealed plasma-converted films is a kinetically driven decay of islands by downward material diffusion from the top layer as outlined by Evans et al.34 This mechanism occurs when the preferred equilibrium state for a rough film is a smooth, well-stacked morphology, such as in the case of layered materials. Material diffusion should be enhanced at higher temperatures, which supports the increase in crystallinity found by micro-Raman analysis with increasing annealing temperatures [see Fig. 1(d)].
We attempted to further show the key role of the plasma in the growth by designing a test experiment. As illustrated in Fig. 5(a), two identical ATM precursor films were first thermally treated at 500 °C (denoted by the red arrow). One of these thermally treated films was then exposed to an inert atmospheric-pressure DBD plasma operated with Ar for 30 min at room temperature and subsequently annealed in Ar at 1000 °C (denoted by black arrows). We refer to this step as “plasma-triggered.” The second thermally treated film was annealed at 1000 °C without the plasma treatment step (denoted by the orange arrow). The characterization of the films by micro-Raman spectroscopy in Fig. 5(b) shows that both films exhibit the E12g and A1g peaks characteristic of MoS2 after the annealing step, but the intensity of the peaks is much higher and the linewidth is narrower with plasma-triggering, similar to plasma conversion (Fig. S7 in the supplementary material37). Assuming that much like the plasma-converted films, plasma-triggering leads to the growth of larger crystal grains, we suggest that the plasma creates either surface species or morphology that facilitates Ostwald ripening. As discussed previously, Ostwald ripening requires a high density of nanoscale structures. It is plausible that the main role of the plasma is to produce these structures through physical processes such as sputtering, or chemical processes such as etching. Future studies should be aimed at identifying the plasma species and determining its energy and flux that are required in order to further optimize the growth of continuous, uniform, and large-area MoS2 films.
IV. SUMMARY AND CONCLUSIONS
In summary, we have introduced a novel approach to synthesize large-area MoS2 thin films by the plasma conversion of solution-deposited precursor films. The key role of the plasma in converting ATM precursor films is demonstrated by comparing with thermal treatment, which did not produce MoS2 films under our process conditions. In addition, annealing of the plasma-converted films was found to produce films with relatively large crystalline grains, a well-stacked layer structure, and a smooth surface. AFM characterization of film growth during annealing and a “plasma-trigger” experiment suggest that the plasma creates unique surface morphologies consisting of nanocones that facilitate the evolution of a smooth, crystalline film. The synthesis approach is generic and should be extendable to the growth of other layered materials from solution-deposited precursor films.
S.B., T.L., and R.M.S. acknowledge support from the National Science Foundation (NSF) under Grant No. DMR-1708742. Z.Y. and R.H. acknowledge support from the NSF under CAREER Grant No. DMR-1760668. The authors thank Xun Zhan for help with TEM which was carried out at the Frederick Seitz Materials Research Laboratory at the University of Illinois Urbana-Champaign.
The data that support the findings of this study are available within the article, its supplementary material, or from the corresponding author upon reasonable request.
R. Mohan Sankaran was awarded the Peter Mark Memorial Award in 2011 for the development of a tandem plasma synthesis method to grow carbon nanotubes with unprecedented control over the nanotube properties and chirality. He received his B.S. in Chemical Engineering from the University of California Los Angeles in 1998 and his Ph.D. in Chemical Engineering from the California Institute of Technology. He began his independent academic career in the Department of Chemical and Biomolecular Engineering at the Case Western Reserve University (CWRU) as an Assistant Professor in 2005, was promoted to Associate Professor in 2010, then promoted to Professor in 2014. In 2020, he moved to the Department of Nuclear, Plasma, and Radiological Engineering at the University of Illinois at Urbana-Champaign and is currently the Donald Biggar Willet Professor in Engineering. His research program primarily focuses on developing atmospheric-pressure plasmas as a chemical platform for the synthesis of novel materials and small molecules with applications in emerging electronics, medicine, and energy conversion. He currently serves as an Associate Editor of the Journal of Vacuum Science and Technology.