Since the creation of integrated circuits in the 1960s, silicon transistors, following the Moore’s law, have been shrinking to boost performance and reduce costs over the past half a century1. Today, as conventional silicon transistors enter the sub-10 nm technology node and approach their physical limits, new channel materials are urgently required to further scale transistors and extend Moore’s law beyond silicon electronics2,3,4,5. Two-dimensional (2D) semiconductors with atomic thicknesses and dangling-bond-free flat surface have attracted tremendous interest and possess promising prospects for scaling transistors to the end of roadmap3,4,6. The International Roadmap for Devices and Systems (IRDS) has listed 2D semiconductors as the potential channel materials in 2017, and forecasts that 2D electronic circuits will be commercially available by 20345.

Monolayer molybdenum disulfide (MoS2) has been considered as one of the most promising 2D semiconductor candidates for high performance electronic circuits because of its intrinsic high mobility, excellent gate controllability, high on/off current ratio, ultra-low standby current, small dielectric constant, and good stability7,8,9,10,11,12,13,14. Indeed, isolated monolayer MoS2 devices have been successfully demonstrated to perform well at ultra-scaled lengths down sub-1 nm, which are inconceivable in the framework of traditional silicon with scaling gate length limit of ~12 nm5,9,10,11,15,16. Monolayer MoS2 transistors have also been identified by Intel as one of three breakthrough technologies to break the scaling limit of silicon. To realize its full potential and high-end industrial applications, it is of utmost importance and a prerequisite to product wafer-scale monolayer MoS2 single crystals on commercial substrates.

Notably, a general theoretical framework is established recently to guide the growth of wafer-scale single-crystal 2D materials17,18. In the light of such guideline that the symmetry group of a substrate should be a subgroup of 2D material, c-plane sapphire with C3v symmetry offers an industry-compatible substrate for the epitaxial growth of wafer-scale monolayer MoS2 single-crystals with point group D3h (C3v plus a mirror-reflection symmetry σh)17,18,19. Indeed, wafer-scale single-crystal MoS2 as well as other transition metal dichalcogenides (TMDs) have been epitaxially grown on sapphire substrates by surface step engineering (e.g., controlling the surface step orientation and height)19,20,21,22,23,24. It is noteworthy that such a step engineering strategy can also be applied to epitaxially grow various 2D materials on other substrates25,26,27,28,29,30,31. However, specially-designed substrates such as deliberately engineered off-cut angles or annealing at harsh temperatures are typically required for surface step engineering. Meanwhile, recent studies demonstrate that growth conditions (e.g., the S/MoO3 precursor ratio) can control the interface such as formation of a specific atomically thin interfacial buffer layer and thus also have a strong modulation on the unidirectional domain alignment26,32,33,34,35,36. Currently, growth condition controlling is usually coordinated with substrate surface step engineering to achieve the grown of wafer-scale monolayer TMD single-crystals.

Now, a natural question is: can we achieve the epitaxial growth of wafer-scale single-crystal MoS2 monolayers on industry-compatible c-plane sapphire substrates by purely growth condition control without the aid of surface step engineering? In this work, we answer this concern in the affirmative and report the synthesis of 2-inch monolayer MoS2 single crystals on general c-plane sapphire substrates by precisely engineering the formation of a specific buffer layer within the substrate-epilayer gap through the control of S/MoO3 precursor ratio. The unidirectional alignment and seamless stitching of MoS2 domains are comprehensively demonstrated via multi-scale characterizations ranging from atomic- to macroscopic-scale. The high quality of as-grown monolayer MoS2 single crystals is evidenced by the state-of-the-art electron, phonon, and exciton properties, comparable to or even better than that of exfoliated ones. Our results offer a simple strategy to epitaxially grow wafer-scale single-crystal MoS2 monolayer on commercial insulator substrates and can also been applied to a wide variety of other 2D materials, laying a strong foundation for 2D electronic circuits to fit into industrial settings.


Unidirectional domain alignment by buffer layer control

Figure 1a–c presents the optical micrographs of the as-grown MoS2 triangular domains on c-plane sapphire substrates with a major miscut angle (~0.2°) towards M-axis under three representative MoO3/S precursor ratios: 3.9% (Fig. 1a), 4.5% (Fig. 1b) and 5.1% (Fig. 1c). Remarkably, the degree of unidirectional alignment strongly depends on the MoO3/S precursor ratio, which is defined as:


where the nmax (nmin) corresponds to the number of oriented MoS2 domains in the majority (minority). For 4.5% MoO3/S precursor ratio, the degree of unidirectional alignment is more than 99%. Note that this value is extracted based on orientation statistics across the entire 2-inch wafer (please see Supplementary Note 1 for microscopy images over a ~1 mm2 area). By contrast, the degree of unidirectional alignment is essentially zero for 3.9% and 5.1% MoO3/S precursor ratios. Figure 1g shows the degree of unidirectional alignment against the MoO3/S precursor ratios. Clearly, the degree of unidirectional alignment can be continuously tuned from ~0 to ~ 100% by controlling the MoO3/S precursor ratio (please see Supplementary Note 2 for more MoO3/S precursor ratios). This indicates that pure MoO3/S precursor ratio control can endow with the unidirectional domain alignment. In addition to c-plane sapphire substrates with a major miscut angle (~0.2°) towards M-axis, the unidirectional domain alignment has also been achieved by precisely controlling the S/MoO3 precursor ratio on pure c-plane sapphires, and c-plane sapphire substrates with different major miscut angles towards other axes (please see Supplementary Note 3), indicating the universality of our method. It is noteworthy that for all the MoO3/S ratios we have investigated (ranging from ~3.83% to ~5.55%), it always belongs to a S-rich condition. Consequently, the shape of MoS2 crystals remains unchanged (please see Supplementary Note 4), in contrast prior work where the drastic changes in S/Mo ratio lead to the shape evolution37,38.

Fig. 1: Unidirectional domain alignment enabled by buffer layer control.
figure 1

Optical microscopy images of the as-grown MoS2 triangular domains under three representative MoO3/S precursor ratios: 3.9% (a), 4.5% (b) and 5.1% (c). Scale bar, 20 μm. Optical micrographs of MoS2 at different growth stages: nucleation (d), stitching (e) and coalescence of grains (f). Scale bar, 30 μm. g The degree of unidirectional alignment as a function of the MoO3/S precursor ratio. Dashed line is the curve fitted by the Gaussian function. h Schematic illustration of the buffer layer control strategy toward the synthesis of wafer-scale monolayer MoS2 single crystals. i Cross-sectional high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images of a MoS2 grown on the c-plane sapphire substrate along the <10\(\bar{1}\)0> and <11\(\bar{2}\)0> directions. j Photograph of the as-grown full-coverage monolayer MoS2 on 2-inch c-plane sapphire substrate with a major miscut angle towards M axis.

Upon extending the growth time, the unidirectional oriented MoS2 domains start to merge and eventually coalesce into a continuous monolayer film at ~45 min (Fig. 1d–f). Figure 1j shows the photograph of the as-grown 2-inch monolayer MoS2 wafer on c-plane sapphire substrates with a major miscut angle towards M axis. Atomic force microscope (AFM) images taken from different locations across a 2-inch wafer manifest a uniform and wrinkle-free monolayer MoS2 film with a thickness of ~0.7 nm (please see Supplementary Note 5). It is worth stressing that compared to surface engineering techniques employed in previous work that require a special design for the substrate19,20,21,25,26, our precursor ratio control strategy holds unique advantages and can fit the c-plane sapphire substrates with a major miscut angle (~0.2°) towards M axis, which are mainly supplied on the market and industry-compatible (please see Supplementary Note 6). Although surface steps can also be present in c-plane sapphire substrates with a major miscut angle towards M axis, they are perpendicular to the zigzag edge of triangular MoS2 domains and therefore have no effect on the unidirectional domain alignment (please see Supplementary Note 7 for details).

To understand the underlying growth mechanism, we perform the cross-sectional high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) of the as-grown MoS2 with unidirectional domain alignment (Fig. 1i). Importantly, an atomically thin buffer layer is observed between the grown MoS2 layer and the c-plane sapphire substrate. By contrast, no buffer layer is formed below the as-grown MoS2 is observed when the degree of unidirectional alignment is ~0 (Supplementary Fig. 11). This suggests that the buffer layer within the substrate-epilayer gap is the key to facilitate the unidirectional epitaxy of the monolayer MoS2 and has been recently demonstrated for unidirectional MoS2 epitaxy on β-Ga2O3 (Fig. 1h)35. It is noteworthy that for previously reported unidirectional TMDs on sapphire substrates, a buffer layer is also typically observed, but its role is usually ignored19,20,34.

To further confirm the effect of the buffer layer and its role on unidirectional domain alignment, we remove the as-grown unidirectional MoS2 from c-plane sapphire substrate by water-assisted technique39,40. It is noteworthy that such removal process utilizes the water intercalation and would not destroy the buffer layer. Then the sapphire substrates with buffer layers are used to re-grow MoS2. Note that we also put fresh sapphire substrates in the growth chamber at the same time for control samples. For fresh sapphire substrates, two antiparallel domains appear simultaneously (please see Supplementary Fig. 12). By contrast, unidirectional MoS2 domains are reproduced on the sapphire substrates with buffer layers (Supplementary Fig. 12). This result strongly demonstrates the key role of buffer layer on the unidirectional domain alignment. By performing X-ray photoelectron spectroscopy and density functional theory calculation, we infer that one possible configuration of the buffer layer is O−Mo−O−Al, with Mo exhibiting a (+5) oxidation state (Supplementary Note 9).

Seamless stitching of MoS2 domains

To realize the goal of wafer-scale single-crystal MoS2 monolayers, seamless stitching of aligned grains must also be satisfied simultaneously, in addition to the unidirectional domain alignment18,41,42. To verify the seamless stitching and the absence of grain boundaries, partially merged MoS2 domains were characterized by atomic-resolution aberration-corrected HAADF-STEM. Figure 2a shows a low-magnification STEM image from the merged area of the two aligned MoS2 domains. The angle between two merged MoS2 grains is ~60°. Figure 2b displays six representative atomic-resolution STEM images taken at the corresponding locations marked in Fig. 2a. The identical lattice orientation without grain boundary continuously across the merging zone strongly evidences the seamless stitching of the MoS2 domains and therefore single-crystal nature. Additionally, the atomic-resolution STEM images show a honeycomb lattice with d-spacing of 0.158 nm and 0.274 nm, corresponding to the (11\(\bar{2}\)0) and (10\(\bar{1}\)0) planes of monolayer MoS2, consistent with previous work43. More STEM images of merged areas that support the seamless stitching of MoS2 domains can be found in Supplementary Note 10.

Fig. 2: Seamless stitching of unidirectional MoS2 domains.
figure 2

a HAADF-STEM image of the merging area between two unidirectional MoS2 domains. The orange dashed lines outline the edge of the two aligned MoS2 domains. b 1–6: six typically atomic-resolution HAADF-STEM images obtained from the locations marked in a, showing that no boundary was formed, the d-spacings of the (10\(\bar{1}\)0) and (11\(\bar{2}\)0) planes of MoS2 are 0.274 and 0.158 nm, respectively. Scale bars, 0.5 nm. The atomic-resolution HAADF-STEM images are filtered to enhance the contrast. c, e Optical microscopy (upper panel) and polarized second-harmonic generation (SHG) mapping (lower panel) of two aligned (c) and misaligned (e) MoS2 domains. Scale bars, 5 μm. d, f Polarized SHG mapping of continuous film formation of single oriented domains (d) and misaligned domains (f). Scale bars, 20 μm.

Polarized second-harmonic generation (SHG), which is highly sensitive to grain boundaries44,45, is further performed to confirm the seamless stitching of MoS2 domains at the large scale. Figure 2c shows the representative polarized SHG mapping of two unidirectionally merged domains. Uniform signal is observed with no obvious intensity fluctuations across the merging area. This is strongly distinct to the antiparallel domains, where a dark line appears at the grain boundary (Fig. 2e). Figure 2d presents the polarized SHG mapping of a continuous MoS2 film consisting of unidirectional aligned domains, showing a uniform distribution of intensity over the entire area. This provides a strong proof of the absence of grain boundaries and seamless stitching, in contrast to result of polycrystalline MoS2 film (Fig. 2f). Besides, etching experiments with hot water vapors are also performed to verify the seamless stitching of unidirectional aligned MoS2 domains (Supplementary Note 11).

Wafer-scale uniformity and high quality

The quality and uniformity of the as-grown single crystal monolayer MoS2 wafers are illustrated via multi-scale characterizations. Figure 3a presents the low-energy electron diffraction (LEED) patterns measured at twelve random locations across over an area of more than 1 cm2. The essentially identical LEED patterns provide strong evidences of the uniformity of the as-grown single crystal monolayer MoS2. In addition, the LEED patterns show three bright diffraction spots, which unambiguously proves the three-fold rotational symmetry and therefore the single-crystalline nature of the as-grown monolayer MoS2 film (Supplementary Note 12)21. Figure 3b is the stacked nine polarization-resolved SHG patterns across a 1 cm2 sample area (SHG patterns from different locations are indicated by different colors). The nearly overlapped SHG six-petal patterns confirm the coherent lattice orientation and uniformity of the as-grown monolayer MoS2 film. Figure 3c, d shows 25 representative room-temperature Raman and photoluminescence (PL) spectra across a 2-inch MoS2 wafer, respectively (please also refer to Raman and PL mapping in Supplementary Note 13). No apparent changes in the peak position and linewidth of both phonons and excitons are observed, illustrating the wafer-scale uniformity. Figure 3e presents a representative fluorescence microscope image of the as-grown monolayer MoS2. The uniform color contrast confirms the uniformity of the monolayer MoS2 film. The wafer-scale uniformity of the as-grown monolayer MoS2 single crystal is also scrutinized by optical micrographs taken at different locations across a 2-inch range (Supplementary Note 14).

Fig. 3: Wafer-scale uniformity and high quality.
figure 3

a Representative low-energy electron diffraction (LEED) patterns of a continuous MoS2 film indicating the single domain orientation nature of the MoS2 film, taken at 95 eV. b Stacked linearly polarized SHG six-petal patterns. SHG from different locations is indicated by symbols with different colors. Solid line is the fitting result to show the six-petal pattern. c, d Representative photoluminescence (PL) spectra (c) and Raman spectra (d) at 25 different locations on the wafer. Dash-dotted box highlights the PL signals from sapphire substrate. e Fluorescence microscope image of the as-grown film. A scratch has been intentionally created in the upper right corner to endow with a clear contrast between MoS2 film and bare sapphire substrate. Scale bars, 100 μm. f Polarization-resolved Raman spectra for 532 nm excitation. g The low-temperature (10 K) PL spectra of the as-grown MoS2. h Circularly polarized PL spectra of the as-grown MoS2 at 10 K, Excitation light is right-handed circularly polarized (σ+) at 1.96 eV (633 nm). i Circular polarization calculated from the PL spectra in h. The high value about 68% indicates the high quality of the MoS2 film.

The high quality of the grown 2-inch monolayer MoS2 single crystals is first characterized by aberration-corrected STEM. A statistical analysis of the atomic-resolution images shows that the density of sulfur vacancies in epitaxial monolayer MoS2 is \(\sim 5.2\times {10}^{12}{{cm}}^{2}\) (Supplementary Fig. 16), which is an order of magnitude lower than that of the previously reported exfoliated flakes46. The quality of the as-grown single crystal monolayer MoS2 film was further scrutinized by helicity-resolved inelastic Raman scattering and exciton valley polarization. Figure 3f shows the circular polarization-resolved Raman spectra at room temperature, excited by left-hand \({\sigma }^{+}\) radiation at 532 nm (2.33 eV). Strikingly, the characteristic in-plane phonon E12g (out-of-plane vibration A1g) features a strong signal only under cross-circularly (co-circularly) polarized detection, while cannot be detected under the co-circularly (cross-circularly) polarized configuration. This evidences the perfect negative (positive) 100% phonon circular dichroism of E12g (A1g), as expected for the Raman selection rule determined by the perfect structure of monolayer MoS247, which proves the negligible defect-lattice scattering and high crystal quality. Figure 3g is the non-polarized PL spectrum of the as-grown monolayer MoS2 at 10 K. The full width at half maximum (FWHM) is ~23 meV, which is only about half of the values in exfoliated flakes48 (Supplementary Note 15). Figure 3h presents the helicity-resolved PL spectra of the epitaxially grown monolayer MoS2 at 10 K, excited by \({\sigma }^{+}\) radiation on resonance with the A exciton at 633 nm (1.96 eV). Remarkably, the degree of exciton valley polarization can reach ~68% (Fig. 3i), which is quantified as:

$${\rho }_{{{{{{\rm{v}}}}}}}=\frac{I\left({\sigma }^{+}/{\sigma }^{+}\right)-I\left({\sigma }^{+}/{\sigma }^{-}\right)}{I\left({\sigma }^{+}/{\sigma }^{+}\right)+I\left({\sigma }^{+}/{\sigma }^{-}\right)}$$

where \(I\left({\sigma }^{+}/{\sigma }^{+}\right)\) and \(I\left({\sigma }^{+}/{\sigma }^{-}\right)\) denote the exciton emission intensities under the co- and cross-circularly polarized configurations, respectively. Note that the exciton valley polarization of ~ 68% is competitive to or even better than those of the best exfoliated flakes (Supplementary Note 15)48,49,50. Given that exciton valley polarization is highly sensitive to defects as they can introduce intervalley scattering to degrade the circular helicity48,49, the high exciton valley polarization offers strong evidence for the state-of-the-art quality of our epitaxial monolayer MoS2 single crystals. The high-quality of the as-grown monolayer MoS2 single crystals can be understood from several aspects. First, we keep a sulfur-rich condition during the growth, facilitating achieving a low density of sulfur vacancies. Second, our chemical vapor deposition (CVD) setup has a unique multisource design (Supplementary Fig. 27), which facilitates the homogeneous cross-sectional source supply and thus high quality. Besides, our monolayer MoS2 single crystals are stitched from unidirectional domains on mono-step sapphire surfaces and the nucleation and growth of domains do not rely on the surface steps, thus we believe that the high-quality of as-grown films would also benefit from such good stitching.

The state-of-the-art device performances

The high quality of the as-grown single-crystal monolayer MoS2, in principle, would enable the superior device performances. To confirm this, we fabricated field-effect transistors (FETs) for performance benchmark testing. To effectively suppress interfacial scattering and extract intrinsic electronic properties, hexagonal boron nitride (h-BN) encapsulated devices based on the epitaxially grown single-crystal monolayer MoS2 are fabricated utilizing the pick-up technique51 (see Methods for more details). Figure 4a schematically shows the typical structure of h-BN encapsulated devices, with heavily doped silicon (Ti/Au) as the bottom (top) gate. It is noteworthy that monolayer graphene with a narrow electron density distribution around the Fermi level, which can act as a Dirac source and enables the low Schottky barriers11,52, is adopted as the as source and drain contact electrodes. Figure 4b, c shows the room-temperature output/transfer characteristics of a h-BN encapsulated device with channel length/width of 5 μm/1 μm. The thickness of top and bottom h-BN layers is ~27 nm and ~34 nm, respectively (Supplementary Fig. 22). Please refer to Supplementary Fig. 24 for the transfer characteristics of more devices. Linear Ids-Vds output characteristic curves (Fig. 4b) indicate the ohmic contact behavior. As can be seen from the transfer characteristic curve (Fig. 4c), the device features an on/off ratio up to ~108, negligible electrical hysteresis, and a sharp subthreshold swing of ~92.6 mV/dec across over three orders of magnitude. Remarkably, the extracted field-effect mobility (\(\mu={G}_{m}\frac{L}{W{C}_{i}{V}_{ds}}\), where Gm is the transconductance, Ci denotes the gate capacitance, Vds is the source-drain bias, L and W are channel length and width, respectively) can reach ~140 cm2s−1V−1 at room-temperature, which is larger than the results based on polycrystalline films and competitive to highest value of the exfoliated flakes39,53,54,55. For total 10 h-BN encapsulated devices from different locations, the average mobility is ~120 cm2s−1V−1. It is noteworthy that the current electron mobility is underestimated and can be further improved by ultra-low/free contact resistance engineering54,56,57,58. Additionally, batch production of uncapsulated FET arrays over several centimeters is achieved based on the as-grown wafer-scale single-crystal MoS2 monolayer (Supplementary Note 17). Although the uncapsulated FETs show lower mobility than encapsulated ones (which may stem from interfacial scattering and contaminations in the device fabrication processes), a statistical analysis of 150 FETs over an inch scale shows a high yield >97% and an average on/off ratio of devices of ~108 (Supplementary Note 17). Finally, 200 nm short-channel FET devices based on the as-grown monolayer MoS2 were fabricated, as shown in Fig. 4d. Figure 4e, f shows the electrical output and transfer characteristic curves of the short-channel device, respectively. The short-channel device exhibits a saturation on-current density of ~535 μA/μm, a high on/off ratio close to 109 and a sharp subthreshold swing of 107 mV/dec. The overall device performances of our wafer-scale single-crystal MoS2 monolayers (e.g., mobility of 140 cm2s−1V−1, on-current density of ~535 μA/μm and 109 on/off ratio) are quite competitive to those best results among all the monolayer TMDC transistors reported so far (Supplementary Table 1), holding promising prospects for large-scale integrated applications.

Fig. 4: The state-of-the-art device performances.
figure 4

a Schematic diagram of the typical structure of h-BN encapsulated devices. b, c Output and transfer curves of a h-BN encapsulated MoS2 device with channel length/width (L/W) of 5 μm/1 μm at Vbg = 0 V. d Photograph of the short channel device. Scale bar, 20 μm. Inset shows the scanning electron microscopy image of the device. Scale bar, 200 nm. e, f Output and transfer curves of the short channel device with channel length/width (L/W) of 200 nm/5 μm.


In conclusion, we demonstrate a simple strategy of precursor ratio control for the batch production of wafer-scale single-crystal MoS2 monolayer films on industry-compatible substrates of c-plane sapphire. The epitaxial monolayer MoS2 single crystals exhibit wafer-scale uniformity and the state-of-the-art properties, evidenced by the perfect phonon circular dichroism, exciton valley polarization of ~70%, room-temperature mobility of ~140 cm2v−1s−1, on-current density of ~535 μA/μm, and a nearly 109 on/off ratio. Our work offers a novel insight into the synthesis of wafer-scale high-quality 2D semiconductors and lays a solid foundation for industrial applications of 2D integrated electronic circuits.


Growth of single-crystal MoS2 monolayers on c-plane sapphire

The growth of single crystalline monolayer MoS2 was carried out in a three-temperature zone CVD system with multisource design manufactured by Dongguan Join Technology Co., Ltd (please refer to Supplementary Figs. 27 and 28 for the schematic diagram and photographs of our CVD setup, repectively). For a typical growth, the center mini-tube is loaded with sulfur (Alfa Aesar, 99.9%, 6 g) and flowed with 40 sccm Ar. Note that to effectively load sulfur, a large rectangular chamber is designed ~5 cm from the end of the quartz tube. The surface area of melted sulfur determined by the rectangular chamber is around 25 cm2. For the outside six mini-tubes, every-other-tube is loaded with MoO3 (Alfa Aesar, 99.999%) and flowed with Ar/O2 (40/0.5 sccm); and the other three empty tubes are also flowed with Ar/O2 (40/0.5 sccm). In addition, to avoid the powder being blown away by the carrier gas, we use MoO3 thin flack (the thickness is ~0.5 mm) which is obtained by pressing the MoO3 powder (Alfa Aesar, 99.999%) with a hydraulic press. All the quartz tubes have an inner diameter of ~1 cm. The distance between MoO3 and substrate (sulfur) is ~32 cm–40 cm (~17 cm). During the growth, sulfur, MoO3 and sapphire substrate are placed at first, second, and third temperature zones, respectively. The temperatures for the sulfur, MoO3, and sapphire substrate are 120, 560, and 880 °C, respectively. The pressure is kept at about 1 torr during the growth process. The growth time ranged from 15 mins (to obtain single oriented discrete MoS2 domains) to 45 mins (to obtain single oriented continuous MoS2 films), and the pressure was kept at about 1 torr during the growth process, and the LPCVD system was cooled to room temperature in the argon gas stream after the growth.

AFM imaging was performed by Asylum Research Cypher S. The thickness of the grown MoS2 film is measured by scraping the film at the wafer corner. LEED measurements were performed in vacuum environments (<3 × 10−7 Pa) using the Omicron LEED system. A high-resolution four-circle X-ray diffractometer (Smartlab, Rigaku) was used to characterize the in-plane crystalline symmetry of the grown wafer-scale MoS2 film.

Optical characterizations

SHG microscopy measurements was performed in a homemade system. A femtosecond laser with 780 nm central wavelength (~100 fs, 100 MHz, generated by MenloSystems) and power of 1.8 mW was used to excite the sample. The laser beam passes through a linear polarizer and was tightly focused to ~1 µm spot diameter by a 100× objective (NA = 0.9, Nikon). The MoS2 samples were attached to a piezoelectric stage to realize sub-micrometer scanning. The SHG signals were collected by the same objective, then passed through another linear polarizer whose polarization angle is perpendicular to the excitation beam. A band pass filter was used to filter out the 780 nm fundamental beam. The SHG intensities were measured by a spectrometer (Princeton Instrument) with an exposure time of 0.3 s each pixel of the mapping.

Both Raman and PL spectra were collected using a HORIBA spectrometer (LabRAM HR Evolution) in a confocal backscattering configuration. For room-temperature Raman spectra, off-resonance light from 532 nm (2.33 eV) continuous laser with a power of about 900 μw was focused through a Nikon objective (N.A. = 0.5 W.D. = 10.6 F.N. = 26.5) onto the sample with a spot diameter of ~1.5 μm. For PL spectra, on-resonance light from 633 nm (2.33 eV) continuous laser with a power of about 400 μw was focused through the same Nikon objective onto the sample with a spot diameter of ~1 μm. The sample was placed in an optical chamber with a high vacuum and then cooled down to 10 K by a closed cryocooler (CS-204PF-DMX-20B-OM from ARS) for cryogenic PL measurements.

TEM characterizations

The MoS2 samples for TEM characterizations were prepared by transferring MoS2 onto the TEM grids (Zhongjingkeyi) using the polymethyl-methacrylate-based transfer method. STEM was performed by an aberration-corrected JEM ARM200F (JEOL) at 200 kV and an aberration-corrected JEOL Grand ARM 300 CFEG operated 80 kV. The SAED was performed with a TEM (Philips CM200) operating at 200 kV.

Transfer of MoS2 films

Propylene carbonate (PC, Chloroform solution of 6% propylene carbonate) was used to pick up boron nitride (h-BN) that has been mechanically exfoliated onto silicon oxide. The picked h-BN was dropped onto the as-grown MoS2 to pick up the single crystal MoS2. Then, two parallel graphene samples were picked up as contact electrodes. Finally, the sample falls onto bulk h-BN on 300-nm silicon oxide substrate to obtain van der Waals heterojunction with clean interface and free of organic residues.

FET fabrication and electrical measurements

Fabrication of h-BN encapsulated devices. h-BN/MoS2/Graphene/ h-BN/SiO2 samples were fabricated via a propylene-carbonate-based transfer method. Ti(2 nm)/Au(30 nm) metal electrodes were deposited as the top gate of the device using electron beam lithography (EBL), electron beam evaporation and lift-off techniques. EBL was used to reverse expose the Hall shape of the device, and reactive ion etching (RIE, Plasma Lab 80 Plus, Oxford Instruments) was used to etch off the excess h-BN/MoS2/Graphene/h-BN to avoid the conduction between the device electrodes. Finally, the device was completed by depositing Cr (3 nm)/Au(30 nm) using an electron beam evaporation system to form a one-dimensional contact with graphene.

Batch fabrication of devices. The patterned Ti (2 nm)/Au (10 nm) metal layer were deposited on sapphire substrate as gate electrodes by UV lithography (MA6, Karl Suss) and electron beam evaporation system, followed by deposition of 20 nm HfO2 as the dielectric layer by atomic layer deposition (ALD). The as-grown single-crystal MoS2 was transferred onto HfO2/Au/Ti/sapphire substrate and then patterned into channels by UV lithography and RIE. The source-drain electrodes are Au/Ti/Au (2/2/10 nm), which are defined by UV lithography, electron beam evaporation and lift-off processes.

Fabrication of short channel devices. The as-grown single-crystal MoS2 was transferred to a HfO2 (5 nm)/Si substrate, and the dielectric layer was deposited by ALD. EBL and RIE were then used to defined the channel and drain-source, and Au/Ti/Au (2/2/10 nm) were deposited as drain-source electrodes.

Electrical measurements were carried out with an Agilent B1500 semiconductor parameter analyzer in a four-probe vacuum station with a bass pressure of 10−6 mbar at room temperature.