Uniform Vapor-Pressure-Based Chemical Vapor Deposition Growth of MoS2 Using MoO3 Thin Film as a Precursor for Coevaporation.

Chemical vapor deposition (CVD) is a powerful method employed for high-quality monolayer crystal growth of 2D transition metal dichalcogenides with much effort invested toward improving the growth process. Here, we report a novel method for CVD-based growth of monolayer molybdenum disulfide (MoS2) by using thermally evaporated thin films of molybdenum trioxide (MoO3) as the molybdenum (Mo) source for coevaporation. Uniform evaporation rate of MoO3 thin films provides uniform Mo vapors which promote highly reproducible single-crystal growth of MoS2 throughout the substrate. These high-quality crystals are as large as 95 μm and are characterized by scanning electron microscopy, Raman spectroscopy, photoluminescence spectroscopy, atomic force microscopy, and transmission electron microscopy. The bottom-gated field-effect transistors fabricated using the as-grown single crystals show n-type transistor behavior with a good on/off ratio of 106 under ambient conditions. Our results presented here address the precursor vapor control during the CVD process and is a major step forward toward reproducible growth of MoS2 for future semiconductor device applications.

Introduction

Layered two-dimensional (2D) semiconducting transition metal dichalcogenides (TMDCs) (MX2: M = Mo, W, X = S, Se, Te) have attracted a great deal of attention because of their exciting properties emerging from quantum confinement effects including intrinsic direct band gap in a single layer, high photoconductivity,[1,2] appreciable charge mobility, and high on/off ratio.[3] In particular, 2D molybdenum disulfide (MoS2) has been studied extensively over the past few years because of its natural abundance and making expedited progress in its applications in the fields of electronics and optoelectronics with photodetectors,[4] light-emitting devices,[5] solar cells,[6] and biosensors.[7−9] Most of these studies were carried out in exfoliated samples which limit their practical application because of low yield, small lateral size, and uncontrolled layer number. For scalable device fabrication, large-scale synthesis of MoS2 is highly desirable. Chemical vapor deposition (CVD)-based synthesis methods including thermal vapor sulfurization of Mo/MoO3 films,[10−12] thermal decomposition of thiosalts,[13] and gas phase reaction of MoCl5 and H2S gas[14] have been employed for large-scale synthesis; however, obtaining MoS2 samples with controlled layer number and high electrical and optical quality remain elusive with these methods. The coevaporation of Mo and S precursor-based CVD has received much attention for the past few years because of its ability to produce high-quality MoS2 single crystals with excellent electrical and optical properties.[15−18] The challenging nature of CVD growth of MoS2 continues to stimulate intense research effort in this field. If the CVD is going to be used for large-scale production of MoS2 samples, it is important to understand how small variations of CVD process parameters could lead to different results.[19,20]

In the coevaporation-based CVD technique, the MoO3 powder and the S powder are commonly used to produce Mo and S vapors which then react to produce MoS2 crystal, following the two-step chemical reactions.[21]

Because most of the growth-related reactions occur in the vapor phase, the uniform melting/sublimation of the precursors and maintaining uniform vapor pressure of the gas-phase precursors are highly desirable for clean and reproducible growth of MoS2. However, such a control of uniform vapor pressure turned out to be very challenging, resulting in a growth of oxide/oxysulfide (MoO2/MoOS2) species along with MoS2.[19,22] To our knowledge, the importance of controlling sulfur vapor pressure is discussed in detail in the literature,[23,24] but the discussion of controlling Mo vapor pressure is rare. In particular, having a uniform vapor pressure of Mo in the reaction region when sulfur enters can help to maintain proper Mo/S ratio for the successful and reproducible growth of MoS2. However, when small amount of MoO3 powder is used and because of the different particle sizes in the powder, MoO3 cannot be uniformly distributed in the crucible which might hinder clean and reproducible growth of MoS2.

Here, we show that by replacing MoO3 powder with a MoO3 thin film for Mo precursors, clean and reproducible growth of monolayer MoS2 can be obtained. MoO3 thin film of 5–20 nm was thermally evaporated from MoO3 powder which was then used as the Mo precursor for coevaporation. Uniform evaporation rate of our thin-film precursor facilitates better control of Mo vapor in the vapor phase, resulting in clean triangular crystals of monolayer MoS2 on Si/SiO2 substrates with a high crystal quality and uniformity proved by scanning electron microscopy (SEM), Raman spectroscopy, photoluminescence (PL) spectroscopy, atomic force microscopy (AFM), and transmission electron microscopy (TEM). Electric characterization of these pristine crystals with bottom-gated field-effect transistors (FETs) showed an n-type transistor behavior with a good on/off ratio of 106 under ambient conditions, verifying the ability of these materials in modern semiconductor electronics.

Results

We synthesized MoS2 single crystals in a home-built atmospheric pressure CVD system. Figure a shows our experimental setup for this study. MoO3 thin films were deposited on Si/SiO2 substrate using thermal evaporation and used as the Mo source for coevaporation synthesis of MoS2 crystals in a single-zone tube furnace. Substrates were placed faced down toward the thin film source. Sulfur was placed upstream at a low-temperature zone. The temperature profiles for MoO3 and S are shown in Figure c. Further information on the growth process can be found in the Materials and Methods section.

Figure 1

Experimental setup. (a) Schematic representation of the atmospheric pressure CVD setup and the relative S, MoO3, and substrate positioning. (b) Cross-sectional view of the substrate boat: target substrate was placed face down toward the film at a small distance d. (c) Temperature profile of MoO3 and S at a setpoint value of 750 °C.

Figure shows a representative result for our optimized CVD growth of MoS2. For this, we used 20 nm thin film of MoO3 as a precursor for Mo vapor. S was placed at the edge of the furnace, 23.5 cm upstream from the center of the furnace, and the growth temperature was 750 °C. Figure a shows an optical image of the whole Si/SiO2 substrate after MoS2 growth as well as a bare substrate for comparison. The darker region of the growth substrate is where the crystals are grown. This region of the substrate is directly facing toward the thin-film MoO3 source. To show the growth on this substrate in detail, we divided this region into nine sections, as shown in Figure b, and presented optical micrographs of each of these region (labeled 1–9) in Figure d. From here, we see that dense crystal grains of self-seeded MoS2 are grown throughout the growth substrate without any traces of MoO2/MoOS2. This is in clear contrast to what we have observed when we used a small amount of (∼1 mg) MoO3 powder as a Mo precursor, where we saw isolated regions of MoS2 growth along with rhomboidal MoO2/MoOS2 on MoS2 in other regions (Figure S1). Increasing amount of MoO3 powder only increased the regions of MoO2/MoOS2. Because the difference between the two scenarios is the uniform sublimation of MoO3 thin film which causes uniform evaporation of Mo in the growth region, we conclude that uniform evaporation of Mo has an important role in dense singe-crystal growth of MoS2. We repeated the MoO3 thin film growth over 50 times, and we always obtained clean MoS2 growth without any oxysulfides. This was not the case when we used MoO3 powder, where clean and reproducible MoS2 crystals were hard to achieve. Another important observation is that all of the single crystals are triangular in shape. Given the conditions of the growth and the rough edges of these triangular crystals (Figure b) which are S-zz terminations, we conclude that they are grown at a sulfur-sufficient environment.[24,25] The shape and size variation of MoS2 crystals has been discussed in the literature using both experiment and theory.[25−27] Depending upon the Mo vapor concentration with respect to S, MoS2 crystals can show different shapes (hexagon, truncated triangle, and triangle), sizes, and density variation as well as formation of MoO2/MoOS2 structures. Wang et al.[26] observed variations in the MoS2 crystal shape from triangular to hexagonal geometries with respect to the distance from MoO3 powdered precursor location. This variation in crystal shape was attributed to the variation of Mo vapor concentration along the growth substrate. Wu et al.[27] also observed shape and size variations of MoS2 with respect to MoO3 concentration. These studies suggest that there is a small window of Mo/S vapor ratio (both in Mo-rich and S-rich regions) during which the MoS2 crystal can maintain triangular shape although their size could vary slightly because of a slight variation of Mo/S vapor ratio. With a relatively larger variation of vapor ratio, both shape and size can vary. Whereas too much of the Mo vapor variation could lead to MoS2 in some places and MoO2/MoOS2 in other places. When MoO3 powder was used as a precursor in our experiment (Figure S1), there appears to be a large variation of Mo/S vapor pressure throughout the growth region. On the other hand, when MoO3 thin film was used as precursor for Mo vapor, we observe the same triangular crystal shape throughout the growth region (no shape variation, neither any oxysulfide species), suggesting that we are able to maintain a tight control on Mo vapor concentration in the gas phase by replacing the powdered precursor by the thin-film precursor.

Figure 2

Growth results. (a) Contrast between bare Si/SiO2 substrate and substrate after the growth of MoS2. (b) Sketch of the sample with the corresponding image positions. The blue regions marked here are outside of the boat. (c) Grain-size histogram with the surface coverage. (d) Optical images of the substrate at the positions specified in (b). Scale bar is 100 μm in each image.

Figure 3

Characterization of the MoS2 single crystals. (a) Optical image of one of the largest monolayer crystals observed. (b) SEM image of a single crystal. (c) AFM topography and height profile (inset) taken at an edge of a monolayer domain. (d) E2g1 and A1g vibrational modes of atoms and Raman single spectra for a single crystal.

Figure d also shows a size and density variation of the MoS2 grains in different regions. A statistical view to the grain size variation is provided in Figure c. We observed crystal sizes up to 95 μm which is the largest we measured for this growth, and most crystals favored to grow in 20–40 μm size that covered around 40% of the growth region. The size variation and increased nucleation density are consistent with the picture of maintaining a tight control on the Mo vapor concentration discussed above. A slight spatial variation of the Mo/S vapor ratio because of a small variation of either Mo or S or both dictated by the flow, temperature gradients, and diffusion patterns within the gas can give rise to different sizes of MoS2 crystals while maintaining triangular shape, even though Mo evaporation is homogenous. In some regions, triangular domains are merged to form polycrystalline aggregates[28] that can be seen because of the random orientations of the crystals and higher density.[28−31] Collectively, the absence of any oxysulfide phase along with the ability of maintaining triangular shape validates our capability to maintain a controlled Mo vapor condition for the clean growth of MoS2.

In addition to MoS2 grown on the growth substrate, we have occasionally observed scattered MoS2 growth on source substrate as well (Figure S7), although the coverage density is much lower in comparison with the face-down growth substrate. This is an interesting observation, given the unique geometry of our growth setup, where Mo and S vapors are trapped between two parallel plates and the gas flow patterns also differ to the conventional face-down technique, where the growth substrate sits on top of a crucible with MoO3 powder precursor. We note that a continuous coverage of monolayer MoS2 was not obtained using our growth method neither on the growth substrate nor on the source substrate. The MoS2 single crystal growth reported here using MoO3 thin film as an evaporation source is different than the sulfurization of Mo thin films previously reported by us and other groups.[32,33] In the present work, the reaction between Mo and S occurs in the vapor phase unlike the sulfurization of Mo thin films, where layer by layer sulfurization occurs at the thin film surface. In the present work, S was introduced to the reaction zone when the center of the furnace was at 650 °C. This is to ensure that all MoO3 thin film was sublimated before S enters the reaction zone. In a control experiment (no S, all other parameters are identical to the optimized growth process), we found that 10 nm thin MoO3 thin films were completely gone from the source substrate at a temperature of 550 °C (in atmospheric pressure). The resulting MoS2 crystal qualities are also different as the grain size in the present work is tens of microns, whereas in the case of thin-film sulfurization, the grain size is tens of nanometers. As a result, the MoS2 crystals grown using this method shows a very high crystal quality, uniformity, and better device properties discussed below, compared to the polycrystalline films obtained by thin-film sulfurization methods.

Figure a shows an optical micrograph of a single crystal grown with the size of ∼95 μm. These big crystals are mostly composed of monolayer regions along with some isolated dendritic bilayers[34−36] region which can be clearly seen in the SEM image as well (Figure b, spot 2). We only observe this bilayer formation in large size crystals which could be because of the added nucleation sites by the defects in the first layer.[36] The self-seeded nucleation center is also visible in the SEM image (Figure b, spot 1). The tapping-mode AFM image (Figure c) taken at a single-crystal edge was analyzed to identify the number of layers of the grown crystals. The step height at this edge was measured to be 0.78 nm which corresponds to the monolayer MoS2 thickness, and the topography of the crystal shows a highly uniform layer. Raman single spectra (Figure d) was acquired to confirm the chemical composition of the material grown as well as the layer number. As shown in Figure c, Raman spectrum has two prominent peaks, E2g1 and A1g, which correspond to in-plane and out-of-plane vibrations of Mo and S atoms with respective wave numbers of 390.76 and 410.76 cm–1, respectively. Spacing between these two peaks (Δ) is 19.82 cm–1 consistent with monolayer MoS2.[37,38]

Figure a shows a PL spectrum from a representative single crystal with two signature peaks at 1.85 and 1.99 eV which corresponds to A (direct band gap excitations) and B (excitations resulted from valence band splitting because of strong spin–orbit coupling[39]) direct excitronic transitions.[37,40,41] The full width at half maximum of the A peak is measured to be ∼80 meV which is narrower than the exfoliated monolayer MoS2 samples on Si/SiO2 substrates[39,42] showing better optical quality of our samples. The even-color contrast of the PL intensity mapping of A and B peaks (Figure b) confirms the uniform layer thickness of the crystal.

Figure 4

PL characterization. (a) PL single spectra of a MoS2 crystal grown with the optimized growth recipe. (b) PL intensity mapping of A peak for the same crystal. Inset shows the mapping of B peak.

Scanning TEM (STEM) was employed to further verify the crystal quality of the MoS2 crystals. The as-grown MoS2 crystals were transferred to a TEM grid with a carbon mesh using buffered oxide etchant (BOE; details in Materials and Methods). Figure shows dark-field STEM (DF-STEM) micrographs for the monolayer MoS2 grown using the thin-film-based coevaporation. High-resolution STEM (HRSTEM) image (Figure b) clearly shows the high crystalline quality of the monolayers with a periodic hexagonal atomic arrangement. The lattice constant is measured as 0.27 nm which agrees well with previously reported results.[29,43] The monolayer can also be identified from the folding edge of the transferred film (Figure c). Evidently in the SEM image, these big crystals show some multilayer growth at some regions (nucleation center and adlayers), and accordingly, the STEM data show Moire patterns caused by the diffraction because of the misalignment of these layers (Figure S5).

Figure 5

TEM characterization. (a) Low magnification DF-STEM image of the transferred thin film on the copper grid. (b) HRSTEM image of the monolayer MoS2 showing hexagonal atom arrangement. (c) DF-STEM image for the folded edge of the monolayer.

We also experimented this growth at higher temperatures. Figure S3 shows the optical micrographs of the crystals grown at 750, 800, and 850 °C by using 20 nm MoO3 thin films. We observe triangular-shaped crystals at all growth temperature, suggesting that uniform Mo vapor pressure can be maintained at elevated temperature. We also found that the coverage density generally decreases at higher growth temperatures. This could be linked to the reduced nucleation at higher temperatures. The sticking coefficient which defines the nucleation ability of the surface significantly reduces because of the enhanced desorption rate at higher temperatures.[44] This phenomenon also explains the multilayer growth at the nucleation center at higher temperatures. There are less nucleation sites to facilitate the gas-phase MoS2 to be adsorbed on to the substrate surface which results in the growth of multiple MoS2 layers at the same nucleation site. The multilayer growth at high temperatures is discussed as an effect of a high precursor supply in the previous reports[31,45,46] when the MoO3 powdered precursor is used which is not applicable for our thin-film-based growth. In our experiments even at 750 °C, MoO3 thin film precursor is completely evaporated and reacted to grow MoS2, therefore more MoO3 supply is not possible. The AFM, Raman, and PL characterizations for the crystals grown at higher temperature are shown in Figure S4. The AFM topography shows a clear stepwise height variation, and each step height measured is similar to monolayer MoS2 thickness. Raman single spectra taken near the center of nucleation show the E2g1 and A1g signature peaks for MoS2 with the spacing of Δ = 23.43 cm–1 which corresponds to a few layers of MoS2.[37,46] PL mapping of the crystals grown at 850 °C shows a clear PL intensity drop at these multilayers, and very thick layers at the center show no PL response. The effect of MoO3 precursor amount is tested by varying the thickness of the film from 5 to 20 nm with the same amount of S (Figure S5). All of the crystals grown have a similar shape, suggesting that the uniform Mo pressure control is possible even with 5 nm thin film. However, the crystal sizes are reduced at lower thicknesses of MoO3 because of the low-precursor supply.

We have also measured electrical transport properties of MoS2 samples in a bottom-gated FET geometry. FET devices were fabricated on the as-grown MoS2 single crystals by using standard electron beam lithography (EBL; see Materials and Methods). The nucleation center was excluded at the device fabrication to minimize the reflection of multiple layers in the transport measurements. We used nickel for source and drain electrodes. Linear ID–VD confirms that Ohmic contacts are formed at the two electrodes, and all of the measurements were performed under ambient conditions for the as-grown samples. Figure shows electrical characteristics of a sample device (optical image is shown in the inset of Figure b). The field-effect mobility was extracted based on the slope dID/dVG fitted to the linear region of the ID–VG using the equation μ = (L/WCGVD)(dID/dVG), where L, W, and CG are the channel length, channel width, and the gate capacitance. For the sample device shown in the figure, 5 μm channel length and 6 μm channel width are used in mobility calculations. This device shows an n-type conduction with an on/off ratio of 106 and peak mobility ∼4.5 cm2/V s, which is comparable with the mobilities for pristine devices fabricated on CVD-grown monolayers in previous reports.[28]

Figure 6

Electrical characteristics of the as-grown MoS2 single crystals. (a) Output characteristics of the transistor device by sweeping the gate voltage (VG) from −10 to 80 V. (b) Transfer characteristics. Plots are provided for the device shown in inset of (b).

Conclusions

In conclusion, we introduced a new CVD-based method to synthesize monolayer MoS2 crystals by using MoO3 thin films as a precursor for coevaporation. This new method facilitates uniform vapor pressure of Mo, resulting in a clean single-crystal growth of MoS2 throughout the substrate, and is highly reproducible. The as-grown MoS2 crystals show excellent Raman and PL response corresponding to high-quality single crystals. TEM characterization further confirms the crystalline quality of the sample. Room-temperature electrical transport measurements show a mobility of ∼5 cm2/V s with a current on/off ratio of 106. We believe this new CVD process can be used for the synthesis of other TMDC materials, and the growth can be extended to a wide range of substrate types.

Materials and Methods

Growth Process

Si substrates with 250 nm thick oxide layer were used as growth substrates as well as the substrates to deposit the MoO3 thin films which were used as a precursor for coevaporation. The substrate was cleaned via sonication in acetone for 5 min, followed by sonication in isopropyl alcohol (IPA) for 5 min, deionized water rinse, and 10 min mild oxygen plasma treatment. MoO3 thin films with the thickness of 5–20 nm were deposited using thermal evaporation of MoO3 powder (99.5%, Sigma-Aldrich) in a vacuum chamber at low evaporation rates of 0.2–0.5 A/s. The thickness of these thin films was confirmed via AFM topography (Figure S6). The thin-film-deposited substrates were cut into 1 cm × 1 cm size and placed in a ceramic boat; bare target substrates were placed on the same boat facing down with the separation (Figure b) of 3–4 mm toward the source and placed at the center of the furnace. Although the front end of the boat was left open for sulfur vapors to enter the reaction zone easily, the rear end was covered by the substrate to trap the vapor phase precursors near the substrate, but a small spacing (less than a mm) was left at the end to ensure that the sulfur vapor pressure does not overstep the balance between precursors. Ceramic crucible containing 600 mg of S powder (99.5%, Sigma-Aldrich) was loaded in to a 1 in. quartz tube placed at the edge of the tube furnace (Barnstead International F79300 Tube Furnace) upstream. For different temperature setpoints, the MoO3 and S precursor distance (r) is adjusted accordingly in such a way that S starts melting below 100° from the setpoint value. Argon gas (99.995% purity) was used as the carrier gas. Temperature of the furnace was raised to the growth temperature of 750–850 °C at the rate of 20 °C/min and held for 10 min (dwell time). After the dwell time, the furnace could cool down naturally until the temperature dropped to 500 °C at which point, the furnace hood was opened for rapid cooling. The Ar flow (200 sccm) was passed initially before heating up the furnace for 10 min to saturate the environment with argon, after which the flow rate was kept constant at 10 sccm. For the parameters used in this process, we found that 10 sccm gas flow with 10 min dwell time is optimum for our growth. When the temperature of the furnace dropped to 350 °C, it was purged with 200 sccm Ar again to flush off the excess reactants.

Characterization of the Materials

The surface imaging and the grain size measurements were carried out by an Olympus BX51M microscope equipped with a Jenoptic Progres Gryphax camera, and the SEM images were taken with a Zeiss ULTRA-55 FEG SEM. A tapping-mode AFM (Veeco instruments, Dimension 3100) topography was used to determine the height of MoS2 crystals. The samples were scanned at slow rates in small scan areas at the edge of MoS2 crystal to achieve more accurate height profiles. Raman and PL spectroscopy measurements were carried out with confocal Raman microscope (WITec Alpha 300 RA) at an excitation wavelength of 532 nm and with laser power of 0.293 mW at ambient conditions. For TEM characterizations, a droplet of BOE was placed on the Si/SiO2 substrate where MoS2 crystals were grown. After a few minutes, the oxide layer was etched and the crystals were transferred to the liquid. The liquid was then picked up by a clean syringe and released on a TEM grid with a carbon mesh. All TEM/STEM measurements were performed with a JEOL ARM200F FEG-TEM/STEM with a Cs-corrector.

Device Fabrication and Transport Measurements

The bottom-gated FETs for the transport measurements were fabricated using standard EBL using a Nabity Pattern Generator System connected to a Zeiss ultra 55 SEM for selected single crystals. Poly(methyl methacrylate) was spin-coated on to the substrate containing MoS2 samples and baked at 130 °C for 3 min. The source and drain electrodes were then defined using EBL and developed in a solution of methyl isobutyl ketone and IPA. A 30 nm thick Ni film was deposited for the metal contacts using E-beam evaporation. The electron transport measurements of the final MoS2 devices were carried out in a room temperature probe station using a Keysight Semiconductor Device Analyzer.