Rapid-throughput solution-based production of wafer-scale 2D MoS₂

Two-dimensional (2D) van der Waals materials such as graphene¹ and MoS₂^2,3 can be thinned down to atomic thicknesses while retaining or enhancing their bulk electrical and optical properties.³ Because of this, 2D materials have significant potential to miniaturize future electronic and optoelectronic devices beyond the limits of traditional semiconductors, toward the goal of atomic-scale devices. In particular, 2D semiconductor transition metal dichalcogenides (TMDCs), such as MoS₂ and MoSe₂, have already been used to demonstrate outstanding capabilities such as a photoluminescence quantum yield of 99%,⁶ an optical absorption > 90%,⁷ a photovoltaic external quantum efficiency of 50%,⁸ and a Hall mobility of 34 000 cm²V⁻¹s⁻¹,⁹ all with nanometer-scale thicknesses. With these properties, 2D TMDCs are leading candidates to revolutionize optoelectronic applications such as ultrathin photodetectors,¹⁰ photovoltaics,¹¹ and LEDs.¹² 

Despite great potential, commercial development of 2D TMDCs has been hindered due to the challenges of large-scale and high-quality material growth. One such challenge is a limitation in 2D TMDC growth throughput caused by long total thermal processing times. The total thermal processing time is here defined as the time a sample spends in its furnace chamber during TMDC synthesis, which includes vacuum pumping, temperature ramp-up, soak, and cooling. The most common MoS₂ growth technique used in research labs is currently chemical vapor deposition (CVD) with a MoO₃ or MoCl₅ precursor, which generally requires a total thermal processing time of 4 to 6 h.^13–16 Large-area MoS₂ can also be grown by flowing hot sulfur vapor over molybdenum precursors within a vacuum tube furnace, a process that also requires 4–6 h, in addition to an external metal deposition step.^17–19 MOCVD can deliver higher quality MoS₂ but can require even longer total thermal processing times in excess of 24 h.²⁰ 

An alternative to CVD growth is the thermal reduction of (NH₄)₂MoS₄ to MoS₂. Precursor solutions containing (NH₄)₂MoS₄ can be dip-coated,²¹ spin-coated,²² or bar-coated²³ onto substrates to form thin precursor films, which are then reduced to 2D MoS₂ when heated in the presence of H₂ or sulfur vapor. Variations of this technique show great promise by demonstrating few-layer MoS₂ growth with the field-effect mobility up to 4.7 cm²V⁻¹s⁻¹.²¹ Because the reaction is solution-processed and requires only one chemical precursor, it is an appealing option for high-yield and rapid TMDC growth, such as roll-to-roll processing.²³ However, roll-to-roll systems are not commonly found in most research labs and are better suited to large-scale manufacturing. To make 2D MoS₂ more obtainable for developers, a simpler, faster, and more cost-effective growth technique is needed that can rapidly iterate through samples. Therefore, we have designed a synthesis technique that uses a rapid thermal processing (RTP) system to synthesize 2D MoS₂ from the (NH₄)₂MoS₄ precursor within 15 min of total thermal processing time, an order of magnitude faster than any reported large-area growth method. The RTP uses high-intensity visible irradiation to heat samples to 1000 °C in an isolated quartz chamber and utilizes convective cooling to rapidly cool samples back to room temperature. While traditional furnaces require 1 hour to heat up and 3–4 h to cool down, the RTP heats to process temperatures within 30 s and cools back to <100 °C within 3 min. This RTP growth is capable of performing dozens of growth trials per day, increasing MoS₂ synthesis throughput by a factor of ten. This speed will accelerate 2D MoS₂ availability and development into commercially viable technologies.^4,5,29–31

Prior to rapid thermal processing, a 0.75% (by weight) solution of (NH₄)₂MoS₄ is prepared by combining 60 mg of (NH₄)₂MoS₄ with 7940 mg ethylene glycol in a small vial and sonicating for 20 min. Substrates, either Si/SiO₂ or sapphire, are first cleaned in acetone and isopropanol and then O₂ plasma treated to promote surface activation. Immediately after, the (NH₄)₂MoS₄ solution is spin-coated onto the substrate at 3000 RPM for 30 s. The sample is then annealed on a hot plate at 100 °C to remove residual solvent. Solutions of 0.5% and 1.25% (NH₄)₂MoS₄ in ethylene glycol were also prepared using the corresponding precursor ratios. Once the solution is mixed, which may be done in large batches, the total time to prepare the precursor film for thermal processing, in a nonautomated lab process, is approximately 10 min.

After spin-coating, samples are placed in an RTP system (Allwin21 Corp. AccuThermo AW 610), as shown in Fig. 1(a). The samples sit in an ambient-pressure transparent quartz isolation chamber atop a silicon carrier wafer. The samples are surrounded by 21 1.2 kW tungsten halogen lamps external to the quartz chamber. The lamps directly heat the sample at a rate of 22 °C/s via absorption of visible light, with minimal heating of the optically transparent chamber walls. The temperature of the sample is monitored using a k-type thermocouple in physical contact with the silicon carrier wafer. To rapidly form MoS₂, the samples undergo a two-step heating process, as shown in Fig. 1(b). Prior to the first step, 10 SLM N₂ with 5% H₂ is allowed to flow through the chamber at room temperature for 2 min to purge the chamber of residual gases. During the next step, called T₁, the sample is heated to 350 °C with a 30 s ramp and held for 300 s while under 10 SLM N₂ with 5% H₂. It is primarily during T₁ that (NH₄)₂MoS₄ is reduced to MoS₂ according to the following two-step reaction:

Ammonium and H₂S molecules evaporate and are carried out of the system via the process gas, leaving behind solid MoS₂ on the substrates. During the following step, called T₂, samples are heated to 1000 °C within 30 s and held for 300 s under 10 SLM N₂ without hydrogen. This high-temperature step induces MoS₂ crystallization into 2D sheets. Afterwards, the samples are convectively cooled to room temperature under 10 SLM N₂ for 180 s.

This method produces uniform MoS₂ across 4-in. wafers, as shown in Fig. 1(c) (inset), and has capacity to process 6-in. wafers. Some nonuniformity near the edges of the samples is observed as an artifact of the spin-coating process, but all other regions contain uniform MoS₂. Figure 1(c) shows an optical microscopy image of an MoS₂ step edge formed by scratching, contrasting the uniform MoS₂ with the SiO₂ substrate. MoS₂ grown from the 0.75% precursor is measured via AFM to be 2.0 nm thick, corresponding to trilayer MoS₂, as shown in Fig. 1(d). This value is in good agreement with the MoS₂ thickness from furnace-based synthesis methods, and the ability to control the thickness of MoS₂ using the precursor concentration is a known benefit of the thermal reduction technique.^21,24 In addition, Fig. 1(e) shows an AFM map of the same scratch region, revealing a surface roughness of 0.68 nm. The surface morphology is shown to be mostly smooth, with an occurrence of voids near the intentionally made scratch. Farther from the scratch, MoS₂ is continuous.

To probe the large-scale growth uniformity achieved using RTP, Raman spectroscopy is employed. Figure 2(a) shows Raman scans taken at nine different locations, normalized to the silicon reference peak. The scans all show a 26 cm⁻¹ peak spacing between the E¹_2g and A_1g peaks, indicating trilayer MoS₂ coverage across the entire sampled surface spanning several centimeters, in agreement with our AFM-measured findings.²⁵ To probe thickness control, Fig. 2(b) shows Raman scans of MoS₂ samples grown from different (NH₄)₂MoS₄ precursor concentrations. As expected, increasing the precursor concentration from 0.75% to 1.25% increases Raman peak spacing to 27.5 cm⁻¹, indicating thicker MoS₂ to >5-layers. Decreasing the concentration to 0.5% lowers the peak intensity but still shows a 26 cm⁻¹ Raman peak spacing, indicating trilayer-growth similar to the 0.75% sample. This is due to mass aggregation at lower thicknesses, as also seen in other reports.^23,24 The difficulty in achieving 1 and 2-layer MoS₂ via solution-processed growth has been observed in several reports, and achieving monolayer MoS₂ using solution-processed precursors is still an active research area.²⁶ However, these results confirm the ability to control the MoS₂ thickness above the trilayer by controlling the precursor concentration.

The crystal structure of the grown MoS₂ is analyzed using in-plane x-ray diffraction (XRD), as shown in Fig. 3(a). The prominent peaks at 2θ = 33.1° and 2θ = 58.9° correspond to the (100) and (110) planes of 2H MoS₂, respectively. A smaller peak at 2θ = 41.1° is identified as the (103) peak of 2H MoS₂, further validating the presence of the expected 2H semiconducting MoS₂ phase. These peaks are consistent with XRD peaks expected from the MoS₂ lattice from the symmetry group P63/mmc and also with card number 77–1716 of the hexagonal lattice parameters of the Joint Committee on Powder Diffraction Standards.²⁷ 

Photoluminescence spectroscopy provides another key indication of 2D material quality and identity. Photoluminescence reveals the 2D 1.85 eV MoS₂ bandgap, as shown in Fig. 3(b). The presence of this peak confirms that MoS₂ grown via RTP is 2D in nature and consequently exhibits strong light-matter interactions.^2,3 The large A exciton peak is seen at λ = 665 nm, while the smaller B peak is located at λ = 618 nm. The splitting of the A and B peaks occurs as a result of spin-splitting from spin-orbit coupling.²⁸ 

X-ray photoelectron spectroscopy (XPS) analysis is shown in Figs. 3(c) and 3(d). The characteristic Mo 3d_1/2 and 3d_5/2 peaks are seen at 234.1 eV and 231.0 eV, while S 2p_1/2 and 2p_3/2 peaks are observed at 164.8 eV and 163.7 eV. In addition, the S 2s peak is also observed at 228.4 eV. From these values, the stoichiometric ratio of S/Mo is calculated to be 1.80, indicating a relatively high density of sulfur vacancies. This is likely caused during the high-temperature annealing phase in which S atoms evaporated from the MoS₂ lattice cannot be replenished due to a lack of sulfur partial pressure in the chamber. This deficiency will cause a high defect density of sulfur vacancies but may be mitigated in future trials by introducing sulfur vapor to the processing chamber.²¹ 

To investigate the electronic quality, field-effect transistors were fabricated onto as-grown MoS₂ on SiO₂/Si substrates. The transistors contained 40 nm thick Au contacts with a 15 nm Ti adhesion layer as shown in the inset of Fig. 4(a). The devices were made with a channel width of 30 μm and a channel length of 6 μm. Figure 4(a) displays gate sweeps taken at source-drain voltages V_sd = 0 V, 2 V, 4 V, and 6 V. The transistor demonstrates n-type transistor behavior and an I_on/I_off ratio of 100. The fabricated devices exhibit a field-effect mobility of up to 0.022 cm² V⁻¹ s⁻¹, a relative mobility value but comparable to many other large-area growth techniques which require much longer MoS₂ synthesis time than the results shown here.^13,15 The low mobility is partially due to sub-stoichiometric MoS₂, which may be addressed in future trials by introducing sulfur vapor to the chamber. This mobility result is therefore an encouraging starting point with ample potential for future improvements.

To probe the critical application-relevant optoelectronic quality of RTP-grown MoS₂, optical responsivity was measured as a function of input power as shown in Fig. 4(b). A maximum responsivity of 300 mA/W was measured at low incident power, with an excitation wavelength of 620 nm. The trend of increasing responsivity with decreasing incident illumination power is expected due to the presence of trap states in the film, as has been seen for exfoliated MoS₂.¹⁰ The responsivity is comparable to or higher than that of untreated exfoliated MoS₂ under similar conditions.^10,24,32 Optoelectronic quality is also quantified via spectral external quantum efficiency (EQE), as shown in Fig. 3(c). The EQE shows that the A and B excitons are areas of increased photocurrent production, with maximum EQE reaching 0.016% at λ = 610 nm. These results indicate that RTP-grown MoS₂ is of sufficient quality to be made into optoelectronic devices and that this quality meets that of benchmark exfoliated samples.

The ability to obtain large-area synthetic MoS₂ within minutes is impactful for several applications. Although the electronic quality of MoS₂ grown via RTP is still relatively low, certain device types can immediately benefit from the growth presented here. For example, 2D-material based photovoltaics will benefit from an abundant source of large-area 2D MoS₂. Several recent reports have demonstrated breakthroughs in photovoltaic performance of 2D materials.^7,8 However, these demonstrated photovoltaic devices thus far have had device dimensions on the order of tens of microns laterally. To the first order, photovoltaic power output scales linearly with the device area, and 2D photovoltaic devices will soon need to expand to wafer-scale sizes to demonstrate practical power conversion capability. For example, 2D photovoltaic top and bottom contact metals, P-N junction structure, doping characteristics, and active layer absorption of 2D photovoltaics must be validated and optimized for large-area devices. Such an effort will benefit greatly from the high-throughput MoS₂ fabrication demonstrated here, in which dozens of wafer-scale samples can be made per day to keep pace with photovoltaic testing. Similar benefits also exist for 2D light emitters, which require large-area samples and device apertures for most applications.³³ In addition, MoS₂-based hydrogen evolution reaction concepts are currently limited by MoS₂ synthesis constraints and have been shown to have similar, or even improved, performance when defect-rich MoS₂ is utilized.³⁴ Finally, purely optical applications may be suitable for the 2D materials fabricated here, either as thin films or nanophotonic structures fabricated over large areas, where high electronic quality is not required.³⁵ 

In any case, the current quality of RTP-grown MoS₂, especially carrier mobility, lags behind that of other synthesis methods.^20,36,37 These first results provide significant room for improvement with some straightforward modifications. The most obvious source of electrical degradation is the measured low sulfur-to-molybdenum ratio of 1.8:1, which indicates a prevalence of sulfur vacancy-type defects in the crystal lattice. A very similar situation was observed by Liu and coworkers, in which MoS₂ films produced by furnace annealing of (NH₄)₂MoS₄ led to MoS₂ with a low carrier mobility of 0.02 cm²V⁻¹s⁻¹ due to sulfur vacancies and defects.²¹ However, they discovered that by annealing the grown films in sulfur and argon, the mobility increased to 4.7 cm²V⁻¹s⁻¹ as the crystal lattice obtained the correct stoichiometry and reduced its defect density. It is expected that RTP-grown MoS₂ will experience a similar improvement in stoichiometric and electrical quality if gaseous sulfur is provided in the reaction chamber. This could be done by modifying the RTP process to include evaporation of solid sulfur precursor in the RTP chamber or by introducing a gaseous precursor such as H₂S. We see this as a next step to improve this RTP growth technique, one which will expand the usefulness of the grown MoS₂ to a broader set of applications that require high-electronic-quality samples.

In summary, few-layer MoS₂ has been synthesized in one-tenth the standard time using a (NH₄)₂MoS₄ precursor and a visible light based rapid thermal processing system. MoS₂ is shown to be crystalline and of the 2H phase, showing characteristic MoS₂ in-plane XRD peaks, Raman signal, and XPS peaks. The 1.85 eV bandgap yields a prominent photoluminescent peak at 665 nm, indicative of a 1.85 eV bandgap. Transistors show a field-effect mobility of 0.022 cm²V⁻¹s⁻¹, and biased photodetectors exhibit a responsivity of 300 mA/W. These results indicate that MoS₂ grown via this rapid thermal processing technique is of good quality and can be used to fabricate a variety of optoelectronic devices. The rapid production technique shown here will lead to more frequent growth iterations, which will in-turn produce more data and discoveries regarding 2D MoS₂ and other 2D TMDCs. This synthesis technique is promising as a high-throughput method of producing MoS₂, assisting with the development of commercially viable 2D materials and technologies.

See supplementary material for additional growth and analysis details.

The information, data, or work presented herein was funded in part by the National Science Foundation under Award No. DMR-1460637.

We acknowledge J. N. L. Albert, N. S. Pesika, and D. F. Shantz of the Chemical and Biomolecular Engineering Department at Tulane University for use of their Bruker Dimension Icon atomic force microscope.

We acknowledge the Coordinated Instrument Facility (CIF) of Tulane University for providing instrumentation and support, especially the use of XRD and micro/nanofabrication equipment.

We acknowledge D. F. Shantz and the Shantz Research Lab of Tulane University for use of their Raman spectroscopy system.