Existing growth methods for large-scale monolayer TMDs have so far produced materials with limited spatial uniformity and electrical performance. For instance, the sulphurization of metal or metal compounds only provides control over the average layer number, producing spatially inhomogeneous mixtures of monolayer, multi-layer and no-growth regions16,17. Although chemical vapour deposition (CVD) based on solid-phase precursors (such as MoO 3, MoCl 5 or WO 3 )18,19,20,21,22,23 has shown better thickness control on a large scale, the electrical performance of the resulting material, which is often reported from a small number of devices in selected areas, fails to show spatially uniform high carrier mobility.

Here we report the growth of semiconducting monolayer films of MoS 2 and tungsten disulphide (WS 2 ) on silicon oxide on a 4-inch wafer scale, with both excellent electrical performance and structural continuity, maintained uniformly over the entire films. Figure 1 presents our continuous TMD monolayer films and shows their wafer-scale homogeneity and intrinsic optical properties. The colour photos of MoS 2 (Fig. 1a; greenish yellow) and WS 2 (Fig. 1b; yellow) films grown on a transparent 4-inch fused silica wafer show that the TMD grown region (right half) is uniform over the whole substrate and clearly distinguishable from the bare silica substrate (left half). The optical absorption, photoluminescence and Raman spectra measured from our films show characteristics unique to monolayer MoS 2 and WS 2 , respectively (Fig. 1d–f). All of these measured spectra have the same peak positions as in exfoliated monolayer samples (denoted by diamonds)13,24,25,26, regardless of the location of the measurements within our films (Supplementary Fig. 1). The X-ray photoelectron spectra taken from our monolayer MoS 2 film show almost identical features to those of bulk single crystal with a low level of defects, further confirming the precise chemical composition and the high quality of our MoS 2 film (Supplementary Fig. 2).

Figure 1: Wafer-scale monolayer TMD films. a, b, Photographs of monolayer MoS 2 (a) and WS 2 (b) films grown on 4-inch fused silica substrates, with diagrams of their respective atomic structures. The left halves show the bare fused silica substrate for comparison. c, Photograph of patterned monolayer MoS 2 film on a 4-inch SiO 2 /Si wafer (the darker areas are covered by MoS 2 ). d, Optical absorption spectra of MOCVD-grown monolayer MoS 2 (red line) and WS 2 (orange line) films in the photon energy range from 1.6 to 2.7 eV. e, Raman spectra of as-grown monolayer MoS 2 and WS 2 , normalized to the silicon peak intensity. f, Normalized photoluminescence spectra of as-grown monolayer MoS 2 and WS 2 . The peak positions in d–f are consistent with those seen from exfoliated samples (diamonds). g, SEM image and photoluminescence (PL) image (bottom inset, at 1.9 eV) of monolayer (ML) MoS 2 membranes suspended over a SiN TEM grid with 2 μm holes (a diagram of the suspended film is shown in the top inset). Scale bar, 10 μm. h, i, Optical images (normalized to the bare substrate region) of patterned monolayer MoS 2 (h) and WS 2 (i) on SiO 2 , taken from the wafer-scale patterned films. The insets show photoluminescence images at 1.9 eV (h) and 2.0 eV (i). Scale bars, 10 μm. Full size image Download PowerPoint slide

Figure 1c shows a photo of a MoS 2 film grown on a 4-inch SiO 2 /Si wafer. The monolayer film was patterned using standard photolithography and oxygen plasma etching to form MoS 2 -covered squares (dark, 6 mm wide) with an array of 3 μm holes. An enlarged, normalized optical reflection image (Fig. 1h) displays a homogeneous reflection contrast for the entire MoS 2 -covered region, confirming uniform monolayer growth everywhere with no gaps. In addition, Fig. 1g shows a scanning electron microscope (SEM) image of an array of fully suspended monolayer MoS 2 membranes (2 μm in diameter) fabricated by transferring our metal–organic chemical vapour deposition (MOCVD)-grown film onto a SiN grid with holes. Its high fabrication yield (>99.5%) suggests mechanical strength and continuity of the film. The widefield photoluminescence images of these films (insets to Fig. 1g, h) show strong, spatially uniform photoluminescence signals, further confirming that they are continuous monolayer MoS 2 , with its high quality maintained even after patterning or transfer. The same spatial uniformity was seen in the optical reflection and photoluminescence images of a monolayer WS 2 film that was similarly grown and patterned (Fig. 1i). Together, the data in Fig. 1 confirm that our MoS 2 and WS 2 films are continuous monolayers, spatially uniform over the entire 4-inch growth substrates with intrinsic optical properties. Below, using MoS 2 as the main example, we discuss the growth (Fig. 2) and the excellent electrical properties (Fig. 3) of these MOCVD-grown films.

Figure 2: MOCVD growth of continuous monolayer MoS 2 film. a, Diagram of our MOCVD growth setup. Precursors were introduced to the growth setup with individual mass flow controllers (MFCs). Red, Mo or W atom; yellow, S atom; white, carbonyl or ethyl ligand. b, Optical images of MOCVD-grown MoS 2 at the indicated growth times, where t 0 was the optimal growth time for full monolayer coverage. Scale bar, 10 μm. c, Coverage ratio for monolayer (θ 1L , green) and multi-layer (θ ≥2L , purple) regions as a function of growth time. d, Grain size variation of monolayer MoS 2 depending on the hydrogen flow rate; from left to right, 5 standard cm3 min−1 (sccm) (SEM image shown), 20 sccm (SEM) and 200 sccm (TEM). e, False-colour DF-TEM image showing a continuous monolayer MoS 2 film. Scale bar, 1 μm. f, ADF-STEM image of a laterally stitched grain boundary in a monolayer MoS 2 film, with red and yellow dots representing the Mo and S atoms, respectively. Scale bar, 1 nm. Full size image Download PowerPoint slide

Figure 3: Electrical characterization and batch fabrication of monolayer TMD FETs. a, Gate-dependent sheet conductance (σ □ ) of monolayer MoS 2 FETs measured at different channel lengths L (curves displaced from the bottom for clarity). Inset: optical image of the device; scale bar, 10 μm. b, Field-effect mobility (μ FE ) measured from five MoS 2 FETs (ellipses) fabricated at random locations with different L. Data from previous results for CVD-grown samples21 (squares) and exfoliated samples7 (triangles) are shown for comparison (purple stars indicate their medians). c, Temperature dependence of μ FE measured from the device in Fig. 3a at L = 1.6 μm (orange) and 8.6 μm (red), and from a previous report on exfoliated samples6 (grey), both showing phonon-limited intrinsic transport. d, Top gate (V TG )-dependent σ □ for dual-gate monolayer MoS 2 FET (device shown in the upper inset). Lower inset: V TG -dependent I SD –V SD curves showing current saturation and ohmic electrode contact. Scale bar, 10 μm. e, Gate-dependent σ □ of a monolayer WS 2 FET showing μ FE = 18 cm2 V−1 s−1. Inset: V TG -dependent I SD –V SD curves showing current saturation and ohmic electrode contact. f, Batch-fabricated 8,100 MoS 2 FET devices on a 4-inch SiO 2 /Si wafer. Top inset: enlarged image of one square containing 100 devices. Middle and bottom insets: corresponding colour maps of σ □ at gate bias V BG = 50 V and −50 V, respectively, with the black block in the middle inset representing the only non-conducting device. g, Histogram of on-state and off-state σ □ of 100 dual-gate FETs showing a median on–off ratio of 106 and a high on-state conductivity. Dark blue, V TG = −5 V; pale blue, V TG = +5 V. All measurements were performed at room temperatures except those in c. Full size image Download PowerPoint slide

Figure 2a schematically explains our MOCVD growth, where we only use gas-phase precursors of Mo(CO) 6 , W(CO) 6 , (C 2 H 5 ) 2 S and H 2 , all diluted in argon as a carrier gas (see Supplementary Methods). The concentration of each reactant can be precisely controlled during the entire growth time by regulating the partial pressure (P X ) of each reactant X. Thus our setup offers an ideal environment for maximizing the areal coverage of the monolayer and for engineering the film structure by controlling the nucleation density and intergrain stitching. Figure 2 summarizes our key findings.

First, our MoS 2 film is grown in the layer-by-layer growth mode, which is ideal for uniform layer control over the large scale. Figure 2c plots the areal coverage of monolayer (θ 1L ) and multilayer (θ ≥2L ; mostly bilayer) regions measured from our MoS 2 grown on SiO 2 /Si; Fig. 2b shows optical images at different growth times, revealing the initial nucleation on the SiO 2 surface (t = 0.5t 0 ), subsequent monolayer growth near (0.8t 0 ) and at the maximum monolayer coverage (t 0 ), followed by nucleation mainly at grain boundaries (1.2t 0 ) and bilayer growth (2t 0 ). We observed no nucleation of a second layer while the first layer was forming (θ ≥2L ≈ 0 when t < t 0 ), producing an optimal growth time t 0 near full monolayer coverage (θ 1L ≈ 1). Additional photoluminescence and electron microscope images taken after different growth times further suggested that the edge attachment was the main mechanism for the monolayer growth after nucleation and that the neighbouring monolayer grains were uniformly connected by tilt grain boundaries with enhanced photoluminescence19 at t = t 0 (see Supplementary Notes and Supplementary Figs 3 and 4). The standard thin-film growth model27 suggests that this growth mode is effective below a certain deposition rate of the growth species, above which it suggests a different mode that forms thicker islands. Indeed, the layer-by-layer growth of MoS 2 film was observed only when we applied a low partial pressure (P Mo ≈ 10−4 Torr in Fig. 2b, c) of Mo vapour (produced by the thermal decomposition of Mo(CO) 6 ; see Supplementary Notes and Supplementary Fig. 5) in the presence of excess (C 2 H 5 ) 2 S. In contrast, the growth at a higher P Mo was no longer in the layer-by-layer growth mode, instead simultaneously producing a mixture of monolayer, multilayer and no-growth regions (Supplementary Fig. 6). For the uniform monolayer growth over a large substrate, it is thus important to maintain a low P Mo constantly over the entire growth region and over time, the key technical capability provided by our MOCVD setup (see Supplementary Fig. 7 for the spatially homogeneous monolayer nucleation on a multi-inch scale).

Second, the grain structure of our MoS 2 film, including the average grain size and the intergrain connection, depends sensitively on the concentrations of H 2 , (C 2 H 5 ) 2 S as well as residual water. As a representative example, Fig. 2d shows the two main effects of H 2 , whose presence is necessary for removing carbonaceous species generated during the MOCVD growth: the average grain size increases from hundreds of nanometres to more than 10 μm with decreasing H 2 flow, and the MoS 2 grains grown under higher H 2 flow (Fig. 2d, right image) have mostly perfect triangular shapes without merging with neighbouring grains, a trend that disappears with lower H 2 flow (left and middle images). These observations are consistent with the H 2 -induced decomposition of (C 2 H 5 ) 2 S (increasing nucleation due to hydrogenolysis)28 and the etching of the MoS 2 (preventing intergrain connection)29 as reported previously. (For further discussion on the effects of (C 2 H 5 ) 2 S and water, see Supplementary Notes and Supplementary Fig. 6.) To grow continuous monolayer MoS 2 with a large grain size and high-quality intergrain stitching, we thus flowed optimal amounts of H 2 and (C 2 H 5 ) 2 S and dehydrated the growth environment.

The darkfield transmission electron microscope (DF-TEM) and annular darkfield scanning TEM (ADF-STEM) images shown in Fig. 2e, f confirm the structural continuity of our MoS 2 film grown under those conditions on the nanometre and atomic length scales. The DF-TEM image shows a continuous polycrystalline monolayer film with no visible gaps and a bilayer area of less than 0.5%. Further analysis of the DF-TEM and electron diffraction data (see Supplementary Fig. 8) confirms a uniform angular distribution of crystal orientations with no preferred intergrain tilt angle for grain boundaries. The ADF-STEM data (Fig. 2f; more images are shown in Supplementary Fig. 9) further confirm that adjacent grains are likely to be connected by a high-quality lateral connection with structures similar to those seen in previous reports18,19. The MoS 2 films shown in Fig. 1 as well as those whose electrical properties we show in Figs 3 and 4 were grown under the conditions described in Supplementary Methods, producing an average grain size of ∼1 μm (see Fig. 2b, e). Almost identical growth parameters with P W ≈ 10−4 Torr produced monolayer WS 2 films as shown in Fig. 1b, i, indicating the same layer-by-layer growth for WS 2 with a similar t 0 .

Figure 4: Multi-stacking of MoS 2 /SiO 2 structure. a, Diagrams (left) and optical images (right) of single, double and triple stacking of monolayer MoS 2 /SiO 2 . b, Optical absorption spectra for single, double and triple stacks, respectively (normalized spectra shown in the inset). c, Diagram for fabrication of MoS 2 device/SiO 2 stacking by alternating MOCVD growth, device fabrication with photolithography, and SiO 2 deposition. See the text for details. d, False-colour SEM image of MoS 2 FET arrays on first (bottom) and second (top) layers (the inset shows an enlarged image of a pair of devices in the same relative positions as in the main panel; scale bar, 50 μm). e, I SD –V SD curves measured from two neighbouring devices on the first (left) and second (right) layers, both showing n-type conductance switching. For the first layer, V BG = 50 V (green) and −50 V (grey); for the second layer, V BG = 100 V (yellow) and −100 V (grey). Full size image Download PowerPoint slide

The electrical properties of our monolayer MoS 2 films have two important characteristics: the spatial uniformity over a large scale and excellent transport properties similar to those seen in exfoliated samples. All our electrical measurements in Figs 3 and 4 (except those in Fig. 3c) were performed at room temperature. Figure 3a shows a plot of sheet conductance (σ □ ) against backgate voltage (V BG ) measured from a monolayer MoS 2 field-effect transistor (FET; optical image shown in the inset) with multiple electrodes for the four-probe measurements (except for a channel length L of 34 μm). It includes several curves for different L ranging between 1.6 and 34 μm (shifted from the bottom curve for clarity), all of which show nearly identical behaviours, including the n-type conductance, carrier concentration (∼4 × 1012 cm−2 at V BG = 0 V) and high field-effect mobility (μ FE ). Figure 3b further plots μ FE measured from five such devices, fabricated at random locations and separated by up to 3.3 mm on a single chip. All the devices show similar μ FE near 30 cm2 V−1 s−1, independent of L and device location, with similarly uniform σ □ –V BG curves (shown in Supplementary Fig. 10), suggesting the spatial homogeneity of the electrical properties of the MoS 2 film at length scales ranging from micrometres to millimetres.

The distribution of μ FE of our devices is compared with the results of multiple devices from two previous reports, each measured from individual grains of exfoliated7 or CVD-grown21 MoS 2 samples. We find that μ FE measured from our MOCVD film is similar to the median μ FE (denoted by a star) of exfoliated samples (and several times higher than the CVD results) and has a much narrower distribution. In addition, the temperature dependence of μ FE (Fig. 3c) measured from the same device in Fig. 3a shows higher μ FE at lower temperatures (92 cm2 V−1 s−1 at 100 K) and intrinsic, phonon-limited electron transport similar to the behaviours previously observed in exfoliated samples (data from ref. 6 shown in Fig. 3c) but different from those observed from a CVD sample with stronger effects from defects30. Specifically, our data show that the temperature dependence of mobility follows a power law of μ FE ≈ T−γ with exponent γ = 1.6 for temperatures between 150 and 300 K, close to the value (1.69) predicted by theory5 and consistent with results from previous experiments (average value ranging between 0.6 and 1.7)6,7,8,9 for a similar temperature range. Finally, Fig. 3d shows a high-performance MoS 2 FET fabricated with an individual top-gate electrode (V TG ). It has a high on/off conductance ratio (∼106), current saturation at relatively low bias V SD (lower inset to Fig. 3d), high field-effect mobility (∼29 cm2 V−1 s−1) and large transconductance (∼2 μS μm−1), all of which are comparable to the best reported results6,7,8. We note that our devices studied in Fig. 3a–d were fabricated at random locations using a polycrystalline monolayer MoS 2 film, unlike the devices with single-grain samples used for comparison. In addition, the electrical properties measured from a separate monolayer MoS 2 film with a larger average grain size of 3 μm (instead of 1 μm in Fig. 3) have almost identical characteristics, including the channel-length independence of μ FE and the phonon-limited transport at T > 150 K (see Supplementary Fig. 11; with the low-temperature mobility as high as 114 cm2 V−1 s−1 at 90 K). Taken together, our data confirm the spatial uniformity and high electrical performance of our MoS 2 FETs independent of the average grain size, which suggests that the intergrain boundaries in our film do not significantly degrade their electrical transport properties. This is probably due to the formation of well-stitched intergrain boundaries with a low level of defects, an explanation also supported by the ADF-STEM (Fig. 2f) and X-ray photoelectron spectrum data (Supplementary Fig. 2) discussed above. Our data therefore lead us to conclude that our optimized MOCVD growth provides an electrically homogeneous monolayer MoS 2 film. Moreover, we successfully fabricated and measured 60 FETs by using a monolayer WS 2 film. Even though the growth of monolayer WS 2 was not carefully optimized, these devices showed excellent electrical properties, with their μ FE as high as 18 cm2 V−1 s−1 at room temperature (Fig. 3e) and a median μ FE close to 5 cm2 V−1 s−1. In addition, the WS 2 device in Fig. 3e showed a high on/off ratio of 106 and the current saturation behaviour (inset to Fig. 3e) as in our MoS 2 devices. (See Supplementary Fig. 12 for data from additional monolayer WS 2 FET devices).

The structural and electrical uniformity of our MoS 2 film enables the wafer-scale batch fabrication of high performance FETs as demonstrated in Fig. 3f, g. Figure 3f shows a photo of 8,100 MoS 2 FETs with a global back gate, which were fabricated on a 4-inch SiO 2 /Si wafer with a standard photolithography process. The middle and bottom insets to Figure 3f show colour-scale maps of σ □ measured from 100 MoS 2 FETs in one square region at V BG = +50 V and −50 V, respectively; the top inset to Fig. 3f shows an enlarged optical image of the devices. We observed an almost perfect device yield of 99%; only two out of 200 FETs that we characterized (including data from an adjacent region) did not conduct. Our data also confirm the spatially uniform n-type transistor operation (larger σ □ for positive V BG ) with similar V BG dependence for all our devices and high on-state device conductance. We further observed similarly uniform V BG dependence from FET devices fabricated using monolayer MoS 2 films with different average grain sizes, as characterized by the histograms of the threshold voltages (Supplementary Fig. 13). Similarly, we fabricated 100 individually addressable dual-gate MoS 2 FETs (similar to the device in Fig. 3d) on another wafer piece. The histogram of the on-state σ □ (V TG = 5 V; median carrier concentration ∼7 × 1012 cm−2) and off-state σ □ (V TG = -5 V) collected from all such FETs (Fig. 3g) shows strong peaks above 10−5 S and near 10−11 S, respectively, confirming a uniform conductance switching behaviour with high on-state σ □ (>10 μS) and on–off ratio (∼106). In addition, most of these batch-fabricated FETs had a high μ FE (>10 cm2 V−1 s−1; see Supplementary Fig. 14).

The data presented in Figs 1, 2, 3 confirm the structural and electrical uniformity of the wafer-scale monolayer MoS 2 film grown by our MOCVD method. This makes our film compatible with batch device fabrication processes on a technologically relevant scale. Moreover, because SiO 2 provides a substrate for its growth, one can produce high-quality monolayer films on a variety of substrates by depositing SiO 2 before the growth. This versatility would permit the fabrication of high-performance FETs directly on non-conventional substrates, such as metal and thermally stable plastic. In addition, one can integrate multiple layers of MoS 2 devices by repeating the TMD film growth, device fabrication and SiO 2 deposition, which could enable novel three-dimensional circuitry.

In Fig. 4 we demonstrate this unique potential by producing multi-stacked monolayer MoS 2 films as well as electronic devices fabricated at different vertical levels. Figure 4a shows diagrams and photographs of three substrates each with single, double or triple monolayer MoS 2 films grown at different levels. The first (bottom) monolayer film was grown on a fused silica substrate; the additional layers were grown on SiO 2 (100 nm thick) deposited on the previously grown MoS 2 monolayer by using plasma-enhanced CVD. The colour of the substrate, which remains uniform for each substrate, becomes darker as the number of layers increases. Their absorption spectra, shown in Fig. 4b, present almost identical absorption at all measured wavelengths, once normalized by the number of stacks grown (see the inset), suggesting little degradation of the optical properties of the monolayer MoS 2 films after subsequent oxide deposition and MoS 2 growth.

Figure 4c shows successive diagrams of our multi-stacked device fabrication process: growth of the first MoS 2 monolayer on a SiO 2 /Si wafer, FET fabrication, deposition of SiO 2 (thickness 500 nm; see Supplementary Methods for details), and growth of the second MoS 2 monolayer and FET fabrication. A false-colour SEM image in Fig. 4d shows an array of MoS 2 FETs successfully fabricated with this process. It includes functioning MoS 2 FETs located at two different vertical levels whose conductance can be simultaneously modulated with a global back gate. The I SD –V SD curves measured from two FETs, adjacent both laterally and vertically (see the inset to Fig. 4d), are shown in Fig. 4e. Both devices show a V BG -dependent conductance change (notice the smaller change for the second layer) with an on-state σ □ of 2.5 μS (first layer) and 1.5 μS (second layer), respectively. Furthermore, we measured similar μ FE values (11.5 and 8.8 cm2 V−1 s−1) from the two devices (Supplementary Fig. 15). The two monolayer MoS 2 films were grown on SiO 2 substrates prepared differently, and the first-layer device had gone through additional steps, including the second MoS 2 growth. Our data in Fig. 4 thus confirm the compatibility of our MOCVD-grown MoS 2 films with conventional thin-film deposition and multi-stacking, which could be used to develop a three-dimensional device architecture based on TMD.