Homo-metal-element mediated surface modification of sapphire for robust epitaxy of wafer-scale single-crystalline 2D semiconductors

Abstract

Two-dimensional (2D) transition metal dichalcogenides (TMDCs) single-crystal wafers have been envisioned as promising channel materials to drive semiconductor technology down to the atomic limit. Although several efforts have been made, wafer-scale single crystal growth with high robustness and efficiency remains challenging. Here, we design a homo-metal-element (W/Mo and Al) surface modification strategy of commercial C/M sapphire inducing a low-symmetry modification layer. This design enables the robust epitaxial growth of 2-inch scale monolayer WS₂ or MoS₂ single-crystal wafers via a facile chemical vapor deposition (CVD) method. Since no exogenous elements are introduced on the substrate surface, the derived monolayer WS₂ single crystals present high crystal quality and no impurity introduction, as evidenced by multi-scale characterizations and field effect transistors array fabrications (on/off ratio >10⁸ and mobility ~48.2 cm²V⁻¹ s⁻¹). This work should hereby propel the wafer-scale synthesis and the internal epitaxial mechanism exploration of 2D TMDCs single crystals via substrate surface modulation, and promote the high-performance applications in various fields.

The wafer-scale growth of 2D transition metal dichalcogenides is desired for their electronic applications. Here, the authors report a surface modification strategy of commercial sapphire substrates to grow wafer-scale monolayer WS₂ or MoS₂ single crystals, showing their application for high-performance transistors arrays.

Introduction

Two-dimensional (2D) semiconducting transition metal dichalcogenides (TMDCs), featuring high carrier mobility at nanometer thickness, excellent gate controllability, small dielectric constant, etc., have been envisioned as promising channel materials for transistor scaling beyond sub-1-nm node and monolithic three-dimensional integration^1–5. To fully upgrade 2D TMDCs from laboratory-level research to industrial applications, the primary bottleneck lies in the reproducible and high-efficiency preparation of wafer-scale, high-quality 2D TMDCs single-crystal wafers⁶. This largely hinges on the in-depth understanding of the epitaxial mechanism, which still lacks a general consensus so far⁷.

Our group pioneered the epitaxial growth of inch-scale monolayer TMDCs (e.g., MoS₂, WS₂) single crystals on various Au single-crystal substrates^8–12, and established the probable epitaxial mechanism from the synergy of adlayer-substrate lattice constant/symmetry matching, relatively strong interface coupling, and step-edge-guiding effects. Concurrently, the epitaxial growth of 2D TMDCs has also been executed on commercial sapphire (crystalline α-Al₂O₃), which is the customary-used substrate in the semiconductor industry. A series of substrate-processing and growth-process-optimization strategies, including choosing specific miscut directions and suitable crystal planes^13,14, controlling the time sequence of TMDCs nucleation and substrate step formation¹⁵, tuning the precursor ratios^16,17, forming a single-type oxygen atomic plane¹⁸, introducing metal dopants¹⁹, etc., were put forward successively. Nevertheless, most of these strategies are plagued by poor reproducibility, a narrow experimental window, and low growth efficiency, which greatly impede the robust preparation of 2D TMDCs single-crystal wafers and their future deployments.

According to the established theoretical framework for the epitaxy of 2D materials, the symmetry group of the substrate should be a subgroup of the 2D material²⁰. Whereas, the sapphire surface is generally reconstructed under ultrahigh-vacuum or high-temperature annealing (inducing symmetry change from three-fold to six-fold)²¹, which usually leads to the growth of anti-parallel oriented monolayer TMDCs domains on top.

Surface modification strategy has been adapted to decrease the atomic symmetry of sapphire, for achieving the preparation of 2D TMDCs single-crystal wafers. Several recent efforts have revealed that an atomically thin “buffer layer” can naturally evolve at the 2D TMDCs/sapphire interface during the growth process^{17,18,22–25}, and mediate the epitaxy of 2D TMDCs single crystals. However, these efforts only focused on the retrospective characterizations of the interfacial structures after the growth process, which is insufficient to elucidate the surface modulation effect. Very recently, some work developed distinct substrate surface modification strategies before the growth of 2D TMDCs. Chang et al. constructed an atomically thin Fe₂O₃ decoration layer on sapphire²⁶, and Zou et al. used lanthanum to passivate the sapphire surface²⁷, etc. Whereas, exogenous elements (i.e., Fe and La) were inevitably introduced into the growth processes, which may induce alien atom impurities, increase the synthesis complexity and costs, and pose obstacles to practical industrial productions.

In this Article, we aspire to develop a homo-metal-element sapphire surface modification strategy for the robust and high-efficiency growth of monolayer TMDC single crystals. The “homo-metal-element” refers to the modification layer on the substrate that possesses the same metal element as that of the adlayer. The advantages of this design can be considered as follows: (1) the metal oxide precursor (e.g., WO₃, MoO₃) can react with Al₂O₃ at high-temperature annealing process, forming a homo-metal-element surface structure with low symmetry (one-fold), which should favor the mono-oriented growth of monolayer WS₂ or MoS₂; (2) the growth process can be segmented easily and examined with details, due to the high chemical stability and simple composition of the surface modification layer, which should contribute to achieving a deeper understanding of the epitaxial growth mechanism; (3) the wafer-scale synthesis of monolayer WS₂ single crystals is seldom reported so far, due to the high volatile temperature of W-contained precursors and the difficulty in their large-area uniform supply/deposition. Overall, this work aims to propose a robust strategy for preparing monolayer semiconducting TMDCs single-crystal wafers and deepen our understanding of the epitaxial mechanism, which should promote their batch production and applications in the next-generation very-large-scale-integration (VLSI) technology.

Results

Design of homo-metal-element surface modification strategy

Choosing the single-crystal substrate with a suitable lattice symmetry is pivotal for the robust epitaxy of 2D TMDCs. As is known, the topmost surface of commercial c-plane sapphire (Al₂O₃(0001)) is actually of six-fold symmetry after surface reconstruction (through ultrahigh-vacuum or high-temperature annealing processes) for minimizing the surface energy. This should readily lead to the formation of anti-parallel TMDCs domains (Fig. 1a).

Fig. 1. Design of the homo-metal-element surface modification strategy on C/M sapphire for the mono-oriented growth of monolayer WS₂ domains and single-crystal films.

a Schematic illustration shows that monolayer WS₂ domains tend to show bi-orientation on pristine c-plane sapphire. b, c Schematic illustration of the epitaxial relationship of monolayer WS₂ triangles on pristine c-sapphire, and the calculated binding energies of monolayer WS₂ triangles with different rotation angles. d Schematic illustration shows that after the homo-metal-element surface modification, monolayer WS₂ domains tend to unidirectionally align along the M $<10\overline{1}0>$ direction of sapphire. e, f Schematic illustration of the epitaxial relationship of monolayer WS₂ triangles on sapphire with homo-metal-element surface modification, and the calculated binding energies of monolayer WS₂ triangles with different rotation angles. The binding energy difference (ΔE_b) of 30° and 90° aligned monolayer WS₂ on sapphire with homo-element surface modification is much larger than that on pristine sapphire (~55 times), which should result in uniform orientation. g Photograph of a batch of five pieces of 2-inch uniform monolayer WS₂/sapphire wafers. Inset: Close-up photograph of one wafer. h Color-coded line scan image of the photoluminescence (PL) spectra collected along the indicated arrow direction of the wafer (shown in inset), with the typical spectrum shown as a solid black line. A strong peak at ~1.98 eV with a narrow peak width of ~51 meV over the wafer can be observed, tentatively indicating the high crystal quality of the as-derived monolayer WS₂ wafer.

Density functional theory (DFT) calculations were first performed to visualize the orientation preference. As shown in Fig. 1b, c, the binding energies of monolayer WS₂ on pristine sapphire present two local minima at 30° and 90°, with a slight binding energy difference (ΔE_b) of ~0.001 electron volt per W atom (eV/W). These nearly degenerated binding energies should result in the anti-parallel orientation of monolayer WS₂ domains on pristine sapphire.

In contrast, a low-symmetry homo-metal-element (Al and W or Mo) modification layer pre-constructed on sapphire can make a difference (Fig. 1d). In our DFT calculations, O-W-O-Al atomic arrangement (structurally analog to that of Al₂O₃) was selected as surface modification layer (Fig. 1e, and structural models in Supplementary Fig. 1). Notably, the binding energies of monolayer WS₂ on the modified substrate show only one deep minimum at 30° (Fig. 1f), and the energy difference between the 30° and 90° aligned cases reaches ~0.055 eV/W, which is ~55 times larger than that on pristine sapphire. The huge energy difference should mediate the unidirectional orientation of monolayer WS₂ on sapphire with a pre-constructed modification layer at the initial nucleation period. This surface modification can also be inferred from the previous report that chemical reaction can happen between WO₃ and Al₂O₃ with the formation of a new Al₂(WO₄)₃ phase, featuring the low symmetry (monoclinic P2₁/c space group) and O-W-O-Al atomic structure²⁸.

Notably, the reproducible preparations of 2-inch monolayer WS₂ single-crystal wafers were then realized with ease on the modified sapphire surface, as evidenced by the achievement of uniform monolayer WS₂ wafers in different batches of experiments (Fig. 1g). And the high crystal quality of the CVD-derived monolayer WS₂ wafer is preliminarily proven by photoluminescence (PL) line scan image, showing uniform and strong peak (~1.98 eV) with narrow peak width (~51 meV) across the wafer (Fig. 1h). Accordingly, our designed homo-metal-element surface modification strategy of sapphire should be reliable for preparing monolayer TMDCs single-crystal wafers.

Preparation and structure of homo-metal-element modification layer

To explore the composition and structure of the homo-metal-element modification layer, its pre-constructing process is carefully examined. Initially, WO₃ powders were deposited on C/M sapphire holding at ~1000 °C (Fig. 2a, and experimental setup in Supplementary Fig. 2). Lots of clusters or particles emerge on the sapphire surface, according to the atomic force microscopy (AFM, Fig. 2b) and scanning electron microscopy (SEM) images (Supplementary Fig. 3). These clusters are supposed to be accumulated WO₃-related species, as evidenced by the W⁶⁺ doublets (4f_7/2 ~ 35.8 eV, 4f_5/2 ~ 38.0 eV) and energy loss peak (~42.0 eV) in X-ray photoelectron spectroscopy (XPS) W 4 f spectrum, which corresponds to those of WO₃ (Fig. 2c, and full-range spectra in Supplementary Fig. 4, optical microscopy (OM) image in Supplementary Fig. 5)²⁹.

Fig. 2. Chemical composition and possible atomic structure of homo-metal-element modification layer on sapphire.

a Schematic illustration of depositing WO₃ powders on sapphire at ~1000 °C. b Atomic force microscopy (AFM) height image of the sapphire surface after depositing WO₃ powders, showing lots of clusters and particles. c X-ray photoelectron spectroscopy (XPS) W 4 f spectrum collected from the sapphire after depositing WO₃ powders. d Schematic illustration showing the formation of the W/Al-based modification layer on the sapphire surface after annealing at ~1200 °C. e An AFM height image of the sapphire with the W/Al-based modification layer, showing a flat surface and uniform steps with the average height of ~0.4 nm. f XPS W 4 f spectrum collected from the sapphire with W/Al-based modification layer. g Time of flight secondary ion mass spectrometry (ToF-SIMS) spectrum collected on sapphire with a modification layer, showing the signal of AlWO_x⁻ (x = 3, 4) ions. h Normalized ToF-SIMS depth profile of sapphire with pre-constructed W/Al-based modification layer. Inset: Simulated 3D structural reconstruction of AlWO₄⁻ ion. The W/Al-based modification layer should only exist on the surface region of sapphire. i Cross-sectional high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images of monolayer WS₂/sapphire along the Al₂O₃ $<10\overline{1}0>$ and $<11\overline{2}0>$ directions, respectively, showing the existence of a modification layer at the interface of monolayer WS₂ and sapphire. j Sequential zoom-in AFM height images of unidirectionally aligned monolayer WS₂ domains on sapphire with a modification layer, showing that the edge of the triangular monolayer WS₂ domain (zigzag, $<11\overline{2}0>$ direction) aligns along the Al₂O₃ $<10\overline{1}0>$ direction.

The WO₃-treated sapphire was then annealed at higher temperature of ~1200 °C (Fig. 2d). Intriguingly, the sapphire surface becomes flat and clean, free of any clusters or particles, and uniform steps with the average height of ~0.4 nm evolve (Fig. 2e). It is noteworthy that, the W⁶⁺ doublets shift to higher binding energies (4f_7/2 ~ 36.5 eV, 4f_5/2 ~ 38.7 eV), and the energy loss peak of WO₃ disappears in the XPS spectrum (Fig. 2f). These results agree with previous reports that, the reaction between WO₃ and Al₂O₃ happened at >1100 °C, forming a new Al₂(WO₄)₃ phase, as evidenced by the redshift of W⁶⁺ doublets in XPS spectrum^28,30. Overall, a homo-metal-element modification layer, possessing W, Al, and O elements, can be formed on the sapphire surface, due to the surface chemical reaction through the ~1200 °C annealing process. In a recent work, the metal precursor (i.e., W(CO)₆) was first deposited on sapphire at ~850 °C for the formation of a WO₃ thin layer³¹. However, the ~850 °C processing temperature is not sufficient to trigger the reaction of WO₃ and Al₂O₃ (i.e., sapphire substrate). Bi-orientation growth of monolayer WS₂ was realized via the traditional metal-organic CVD process.

Time of flight secondary ion mass spectrometry (ToF-SIMS), featured with superior mass resolution and sensitivity (~10 ppb level)³², was also used to validate the composition and structure of this new W/Al-based phase. Notably, negative ions with mass/charge ratios (m/z) of ~259 and ~275 are detected (Fig. 2g, and full-range spectra in Supplementary Fig. 6, mapping images in Supplementary Fig. 7), which can be assigned to AlWO₃⁻ and AlWO₄⁻ ions, respectively. Notably, the MWO₄⁻ ions (M denotes the metal element, e.g., NaWO₄⁻ and LiWO₄⁻) were considered as the main fingerprint ions to identify the tungstate-related species^33,34. Hereby, the modification layer should possess a similar atomic structure to that of Al₂(WO₄)₃ with the O-W-O-Al atomic arrangement (Raman spectrum of modified sapphire in Supplementary Fig. 8, and atomic model in Supplementary Fig. 9). Besides, the modification layer only exists on the topmost surface of sapphire, since the normalized intensity of AlWO₄⁻ ion decreases rapidly with the sputtering time (Fig. 2h). As a direct proof, cross-sectional high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images at the interface of monolayer WS₂/modified sapphire were also achieved (with energy dispersive spectroscopy (EDS) data in Supplementary Fig. 10), showing an atomically thin modification layer with a brighter contrast than that of sapphire substrate (Fig. 2i, with the interface of monolayer WS₂/pristine sapphire as reference in Supplementary Fig. 11). The W atoms in the modification layer are periodically arranged with regard to the underlying sapphire lattice in both Al₂O₃ $<10\overline{1}0>$ and $<11\overline{2}0>$ directions. And the experimental HAADF-STEM images agree well with those of simulated images (more details in Supplementary Fig. 12), regarding the Z-contrast intensities, atomic distances along different directions, etc. The AlWO_x⁻ ions are also detected in the ToF-SIMS spectrum (Supplementary Fig. 6). In all, the interfacial modification layer is assigned to be the W/Al-based one with the O-W-O-Al atomic arrangement.

Moreover, the zigzag (zz, $<11\overline{2}0>$ direction) edge of monolayer WS₂ is observed to be parallel with the Al₂O₃ M $<10\overline{1}0>$ direction from the HAADF-STEM images, which is also shown in the zoom-in AFM images (Fig. 2j, and AFM inphase images in Supplementary Fig. 13). Specifically, the monolayer WS₂ zz direction ($<11\overline{2}0>$) should be rotated by 30° rotation (R30°) from the Al₂O₃ A $<11\overline{2}0>$ direction, which also corresponds well with the above DFT calculations, i.e., the 30° aligned monolayer WS₂ is energetically the most stable on sapphire with the modification layer, and the in-plane X-ray diffraction (XRD) results (Supplementary Fig. 14). In this configuration, the 5 × 5 monolayer WS₂ supercell (5 × 3.20 Å = 16.0 Å) nearly commensurate with the Al₂O₃ supercell ($2\surd 3\times 4.76\mathrm{\AA }=16.5\mathrm{\AA }$) with ~3.1% lattice mismatch. This growth behavior should follow the traditional epitaxy theory.

Note that this epitaxial relationship is distinct from previous reports using Fe₂O₃-decorated or lanthanum-passivated sapphire as substrates^26,27, where R0° is the preferential orientation. Additionally, the domain edge direction of monolayer WS₂ is independent of the step edge direction of sapphire. Accordingly, the step-edge-guiding effect may not be the dominating factor for the epitaxial growth mode in our system. This is also confirmed by examining the initial growth stage, where triangular nuclei were rotated by 30° with regard to the step edges (i.e., Al₂O₃ $<11\overline{2}0>$ direction) on C/M sapphire (AFM image in Supplementary Fig. 15). In brief, through the pre-reaction between metal precursor and sapphire substrate, a low-symmetry modification layer with the O-W-O-Al atomic arrangement can be fabricated on sapphire. The modified sapphire is beneficial for the epitaxial growth of monolayer WS₂ domains or single-crystal films.

Epitaxial growth of 2D TMDCs semiconductors

To validate the theoretical predictions regarding the domain orientations with the substrate symmetry, the growth of monolayer WS₂ was first attempted on pristine C/M sapphire, by using WO₃ and S powders as precursors at ~950 °C, via the low-pressure chemical vapor deposition (LPCVD) method (experimental setup in Supplementary Fig. 16). As expected, monolayer WS₂ domains with mirror-symmetric orientations occur on pristine sapphire, with the respective ratio of ~1:1 (OM image and statistical results in Fig. 3a, h). For the chemical composition characterizations, prominent doublets of W⁴⁺ 4f_7/2 (~32.7 eV), 4f_5/2 (~34.9 eV), and S²⁻ 2p_3/2 (~162.3 eV), 2p_1/2 (~163.4 eV) are deconvoluted from the XPS spectra (Fig. 3b). In randomly collected Raman data, almost fixed positions of in-plane ($E_{2\mathrm{g}}^1$, ~356 cm⁻¹) and out-of-plane vibration (A_1g, ~418 cm⁻¹) modes are acquired, with a nearly fixed wavenumber difference of ~62 cm⁻¹ (Fig. 3c). Moreover, uniform PL signals (~1.93 eV) are achieved with a full width at half maximum (FWHM) of ~92 meV (Fig. 3d, and more PL data in Supplementary Fig. 17). The achieved results show similar features to those of the monolayer WS₂/sapphire^35,36. Accordingly, bi-oriented monolayer WS₂ domains or films can be synthesized on pristine C/M sapphire.

Fig. 3. Low-pressure chemical vapor deposition (LPCVD) growth of mono-oriented monolayer WS₂ domains and continuous films on C/M sapphire substrates with homo-metal-element surface modification, featuring high crystal quality and growth efficiency.

a Optical microscopy (OM) image of monolayer WS₂ domains grown on pristine C/M sapphire, showing typical bi-orientation marked by red and blue triangles. b XPS spectra for W 4 f and S 2p, showing the formation of WS₂. c, d Stacked color-coded image of the Raman (regarding $E_{2\mathrm{g}}^1$ and A_1g modes, showing fixed peak positions and a uniform wavenumber difference of ~62 cm⁻¹), and PL spectra (showing a weak peak at ~1.93 eV with broadening peak width of ~92 meV) randomly collected at different 10 positions in (a), with the typical spectra shown as solid black lines, indicating the monolayer characteristics of WS₂. The peak marked by an asterisk in (d) should be ascribed to the sapphire substrate. e–g OM images of the mono-oriented monolayer WS₂ triangular domains, merged domains, and continuous film on sapphire with homo-metal-element surface modification (grown for ~5, ~10, and ~15 min, respectively) under the same growth conditions as those of (a). h Statistical analysis on the orientation distributions of monolayer WS₂ domains on the two types of substrates based on the typical OM images (a, f), showing the capability of growing mono-oriented monolayer WS₂ domains on the modified sapphire substrate. i, j XPS spectra for W 4 f and S 2p, showing the formation of WS₂ (W⁴⁺) on sapphire, and the underneath homo-metal-element (W⁶⁺) modification layer. k Color-coded line scan image of the Raman (with almost the same characteristics as those in (c)) collected along the indicated arrow direction of the wafer (shown in insets), with the typical spectrum shown as a solid black line. l Stacked polarized second harmonic generation (SHG) spectra randomly collected from the wafer, showing the uniform lattice orientation of the continuous film. This indicates the successful achievement of high-quality mono-oriented monolayer WS₂ domains and films on modified sapphire with high growth efficiency.

Homo-metal-element surface modification process was then performed on C/M sapphire. Surprisingly, under the same growth conditions as those in Fig. 2a, mono-oriented monolayer WS₂ triangular domains (~10 μm in edge length) appear on C/M sapphire with surface modification (Fig. 3e, for ~5 min), and they grow laterally (Fig. 3f, for ~10 min) and finally stitch into a continuous film (Fig. 3g, for ~15 min). Statistical analysis demonstrates a high proportion of uniform orientation (~98%, Fig. 3h, and more OM images and second harmonic generation (SHG) results in Supplementary Fig. 18). Notably, a “face-to-face” metal precursor supply route is utilized to decrease the diffusion distance of WO₃ precursor onto sapphire substrate, improve the deposition uniformity of metal precursor, and thus greatly enhance the growth efficiency and homogeneity. Notably, only one quarter of growth time (~15 min) is needed for the full-coverage growth with satisfying inch-scale uniformity, compared to that of monolayer WS₂ growth on a-plane sapphire (~60 min) using the up-streamed WO₃ precursor supply route¹⁴. Besides, consistent mono-orientation growth of monolayer WS₂ can be achieved on such modified substrates, regardless of growth temperatures, precursor ratios, and substrate miscut directions (OM and AFM images in Supplementary Fig. 19). This strongly underscores the robustness of our homo-metal-element modification strategy on the epitaxy of monolayer TMDCs under different growth conditions.

The XPS doublets of W⁴⁺ 4f and S²⁻ 2p (the same as those in Fig. 3b) indicate the formation of WS₂ (Fig. 3i, j, and full-range spectrum in Supplementary Fig. 20). Notably, one W⁶⁺ 4f_7/2 (~36.2 eV, pink peak in Fig. 3i) peak is also deconvoluted. Considering that the as-grown sample is achieved from the S-rich growth condition, and its surface is atomically flat, free of any WO₃ cluster adsorptions, the W⁶⁺ signal should arise from the homo-metal-element modification layer. By comparing the two growth systems, the modification layer should take effect in the mono-orientation of monolayer WS₂ on the pre-modified sapphire surface.

Moreover, Raman line scan across the wafer (Fig. 3k) showing the same characteristics (regarding peak positions, wavenumber difference, etc.), suggests the uniform monolayer thickness of WS₂ (AFM image in Supplementary Fig. 21). The unidirectional orientation is also reconfirmed by the homogeneous polarized SHG spectra (Fig. 3l, and more data in Supplementary Fig. 22). Briefly, by designing the homo-metal-element surface modification strategy of C/M sapphire, robust and high-efficiency growth of unidirectionally oriented monolayer WS₂ domains and single-crystal films can be realized, via the “face-to-face” metal precursor supply route. And the wafer size is limited by the size of the CVD furnace and sapphire substrate, which is expected to be scaled to 300 mm through optimization. Notably, this strategy can be extended to the epitaxial growth of monolayer MoS₂ single crystals, as shown in the corresponding ToF-SIMS, XPS, cross-sectional HAADF-STEM, and EDS results (Supplementary Figs. 23–27). And a similar O-Mo-O-Al atomic arrangement of the Mo/Al-based surface modification layer is proposed. This strongly indicates the universality and versatility of our homo-metal-element surface modification strategy.

Seamless stitching of mono-oriented domains

The seamless stitching of mono-oriented monolayer WS₂ domains should be the key premise for the formation of single-crystal films. To address this, macroscopic spectroscopy characterizations were performed on the as-grown monolayer WS₂ on sapphire with a pre-constructed modification layer. The high thickness uniformity and superior crystallinity of the merged monolayer films are firstly evidenced by the homogeneous color contrasts in OM, Raman (extracted from the intensity of $E_{2\mathrm{g}}^1$ mode), and PL (extracted from the intensity of A-exciton peak) mapping images (Fig. 4a–c, e–g). Remarkably, uniform color contrasts among the merged domains and continuous films can also be observed in SHG mapping images (Fig. 4d, h, as noticed by blue arrows). In contrast, obvious intensity drops of SHG signals occur at the merging area of bi-oriented monolayer WS₂ domains and stitched poly-crystal films (more Raman, PL, and SHG mapping images in Supplementary Figs. 28 and 29). These results tentatively address the seamless stitching behavior of the mono-oriented monolayer domains/films.

Fig. 4. Spectroscopy and atomic-scale electron microscopy characterizations revealing the seamless stitching of mono-oriented monolayer WS₂ domains.

a–d OM image of merged mono-oriented monolayer WS₂ domains and corresponding Raman, PL, and SHG mapping images, showing uniform color contrasts. The blue arrows in (d) indicate the stitching positions. The mono-oriented monolayer WS₂ domains tend to be seamlessly stitched at the merging regions. e–h OM image of the continuous monolayer WS₂ film and corresponding Raman, PL, and SHG mapping images, showing homogeneous color contrasts. This tentatively indicates the grain boundary (GB)-free and the single crystallinity of the monolayer WS₂ film. i Typical high-resolution transmission electron microscopy (HRTEM) image of the merged area of two mono-oriented monolayer WS₂ domains. j Large-area HAADF-STEM image of the merged region marked by a red square in (i), showing the continuous lattice at the merged region. k Atomic-resolution HAADF-STEM images ranging over the stitching area (corresponding to the three marked locations in (i)), showing identical lattice orientation, continuous lattice, and defect-free characteristics. This strongly indicates that aligned monolayer WS₂ domains can stitch seamlessly. l Selected area electron diffraction (SAED) patterns collected randomly from the monolayer WS₂ film merged by mono-oriented domains. The angles (with regard to the horizontal direction) marked by the dashed lines are almost identical, again indicating the uniform lattice orientation of the monolayer WS₂ film.

Atomic-resolution HAADF-STEM characterizations were also performed to provide atomic-scale evidence. Continuous lattices with the identical orientation can be achieved at the monolayer WS₂ domains and the merging regions (successive zoom-in images in Fig. 4i–k and Supplementary Fig. 30). The honeycomb lattice shows the d-spacings of ~0.27 nm and ~0.16 nm, corresponding to those of ($10\overline{1}0$) and ($11\overline{2}0$) planes, respectively. This agrees well with that of the CVD-derived monolayer WS₂ reported previously³⁷. Moreover, the high crystal quality with negligible defects or dopants in the monolayer WS₂ can be inferred by the absence of obvious contrast variation in the atomic lattice. This should be attributed to the homo-element characteristics and high stability of the modification layer.

Besides, only one set of hexagonal diffraction spots is detected with orientation deviation <±0.2° in the selected area electron diffraction (SAED) patterns, reconfirming the uniform lattice orientation of the monolayer WS₂ (Fig. 4l and Supplementary Fig. 31). In short, the seamless stitching of mono-oriented monolayer WS₂ domains and their merging into continuous single-crystal films can be identified by multiscale characterization methods. As reported previously, the electrical and thermal properties of 2D TMDCs (e.g., MoS₂) were severely degraded in the grain boundary (GB) regions³⁸. The wafer-scale uniform 2D TMDCs single crystals should access promising applications in highly integrated electronic devices.

Demonstration of uniform and superior electrical devices performance

Back-gated field-effect transistors (FETs) array was fabricated by the monolayer WS₂ single-crystal film transferred on the ~50 nm thick SiO₂/Si substrate (schematic FET device in Fig. 5a). The channel length (L) and width (W) are both ~1 μm (OM and SEM images in Fig. 5b, and more SEM and Raman mapping images in Supplementary Fig. 32). Note that, composite metal contact electrodes (Au/Ti/Pd) were fabricated to achieve ideal electrical performance, which was proven effective in all-out scaled (L ~ 30 nm) monolayer MoS₂-based FETs array³⁹ and flexible monolayer MoS₂ integrated circuits⁴⁰. The output curves of the typical FET device exhibit a linear current-voltage relation (Fig. 5c). Remarkably, superior on-state current density (~217 μA μm⁻¹) can be achieved at drain/source voltage (V_ds) of ~3 V and gate voltage (V_g) of ~30 V, indicating the good Ohmic contacts at the drain/source electrodes. Moreover, the typical dual-sweep transfer curves present n-type behavior and small hysteresis (Fig. 5d). Thanks to the superior uniformity of the monolayer WS₂ single crystals, the measured FET devices exhibit normal operations and low variations of ~19.2% in on-state current (defined as S_Ion/<I_on> in the previous report⁴¹, where S_Ion and <I_on> denote the standard deviation and mean value of on-state current, respectively, Fig. 5e).

Fig. 5. Uniform and superior electrical performance of field effect transistors (FETs) array fabricated on monolayer WS₂ single-crystal film.

a Schematic illustration of the single FET device. The composite metals Au/Ti/Pd are utilized as drain/source electrodes, and ~50 nm thick SiO₂ is used as a dielectric layer. The channel length (L) and width (W) are both ~1 μm. b OM image of the FETs array fabricated on a monolayer WS₂ single-crystal film. Inset: SEM image of the single FET device. Scale bar: 1 μm. c Output curves of the typical monolayer WS₂-based FET under different gate voltages (V_g) from ~0 V to ~30 V with an internal of ~5 V, showing a linear relation and high on-state current density of ~217 μA μm⁻¹. This suggests good Ohmic contact between metal electrodes and monolayer WS₂ channels. d Dual-sweep transfer curves of the typical monolayer WS₂-based FET under the drain-source voltage (V_ds) of ~1 V. e Stacked transfer curves of 40 randomly selected monolayer WS₂-based FETs (V_ds ~ 1 V), showing good uniformity. f Statistics of the field-effect mobility (μ_FE) and on/off ratio extracted from the FETs array, with the average μ_FE of ~40.2 cm² V⁻¹ s⁻¹ and on/off ratio >10⁸, displaying narrow distribution and good uniformity. These indicate the high crystal qualities of monolayer WS₂ single-crystal films. g Comparison of electrical performance (in terms of on/off ratio and μ_FE) of monolayer WS₂-based FET devices using SiO₂ as dielectric reported previously.

Impressively, the on/off ratios of the measured FET devices all exceed ~10⁸, and the average and maximum field-effect mobilities (μ_FE) reach ~40.2 and ~48.2 cm² V⁻¹ s⁻¹, respectively, and they both show Gaussian distributions (Fig. 5f). This performance surpasses most of the reported monolayer WS₂-based FET devices using SiO₂ as dielectric and measured at room temperature (Fig. 5g and Supplementary Table 1)^36,42–51. This unambiguously proves the high crystal quality of our monolayer WS₂ films and the good electrical contacts between composite metal electrodes and monolayer WS₂ channels. Additionally, the designed growth process excludes the introduction of other exogenous atoms (from precursor selection, substrate surface modification process, etc.), thus avoiding the possible disturbance of carrier scattering and interfacial states. This should also account for the improved electrical performance. And we highlight that this basic device structure is fabricated to only test the quality of monolayer WS₂ single crystals, where advanced device configurations are needed to fully realize their potential in high-performance electronics. In all, the CVD-derived monolayer WS₂ single-crystal wafers should hold great promise to serve as promising n-type channel materials for future applications in 2D complementary metal oxide semiconductor (CMOS) logical electronic devices.

Discussion

In summary, we have developed a homo-metal-element modification strategy of industry-compatible C/M sapphire, and realized the reproducible and high-efficiency growth of wafer-scale monolayer TMDCs single crystals, via the designed “face-to-face” metal precursor supply route. The low-symmetry surface modification layer allows the consistent mono-oriented growth of monolayer TMDCs under a wide range of conditions. The derived 2-inch monolayer WS₂ single-crystal wafers can be utilized to fabricate high-performance integrated FET arrays, affording high on-state current, elevated mobilities and on/off ratios, etc. Moreover, this growth strategy also shows good versatility in the preparations of other kinds of 2D TMDCs (e.g., MoS₂). Overall, this work offers an all-round optimization pathway for the industry-standard production of 2D TMDCs single-crystal wafers, which should inspire more in-depth explorations of the epitaxial growth mechanism and accelerate the integration of 2D atomically thin semiconductors into the next-generation integrated electronic circuits.

Methods

Homo-metal-element modification of sapphire

Commercial 2-inch C/M sapphire substrate (Helios New Materials Limited) was placed upon the excessive WO₃ powder (Alfa Aesar, 99.998%) and put into the tube furnace (Lindberg/Blue). The furnace was then heated to ~1000 °C and maintained for ~1 h under the protection gas of argon (Ar) of ~200 standard cubic centimeters per minute (sccm). This deposition temperature is higher than the vaporizing temperature of WO₃ (~850 °C), and ~1 h deposition time guarantees the complete and uniform deposition of WO₃ on sapphire. The WO₃-treated sapphire substrate was then put into the muffle box furnace (Hefei Kejing) and annealed at ~1200 °C for ~2 h to form the homo-metal-element modification layer. Specifically, according to the DFT models (Supplementary Fig. 9), the theoretically required deposition amount of W atoms of the W/Al-based modification layer is ~1.97 × 10¹⁶ W atoms, i.e., ~3.27 × 10⁻⁸ mol. This equals a small mass of the WO₃ powder as ~7.58 × 10⁻³ mg, and the actual deposition amount of WO₃ onto sapphire should be far in excess, corresponding to the WO₃-rich condition.

Growth of monolayer WS₂ single-crystal domains/films

Monolayer WS₂ was grown on the pre-modified sapphire via the LPCVD process, with the graphite boat containing WO₃ powder placed under the substrate. The single-end-sealed quartz tube containing S powder (Alfa Aesar, 99.5%) was placed upstream and heated at ~200 °C. The furnace was heated to ~950 °C and kept for different time periods to tune the sample coverages, under the protection gas of Ar (~100 sccm). The pressure during the CVD growth process is ~55 Pa. The sample transfer was performed using the common polymethyl methacrylate (PMMA)-assisted method.

Characterizations

Monolayer WS₂ and MoS₂ single-crystal samples were systematically characterized by OM (Nikon ECLIPSE, LV100ND), SEM (Hitachi S-4800, ~1 kV, ~10 μA), XPS (Kratos Analytical AXIS-Ultra with monochromatic Al Kα X-ray), AFM (Bruker, Dimension Icon), Raman and PL spectroscopy (WITec Alpha 300 R, ~532 nm laser), SHG spectroscopy (WITec Alpha 300 R, ~1064 nm laser), HRTEM, HAADF-STEM, and EDS (JEOL-ARM300F, ~300 kV), SAED (FEI Tecnai F20, ~200 kV), and in-plane XRD (Rigaku Smartlab). Cross-sectional samples were prepared by the dual-beam focused ion beam (FIB) microscope (ThermoFisher Helios 5 CX) with a gallium (Ga⁺) ion source. For the ToF-SIMS tests (IONTOF), bismuth (Bi³⁺) ions were used as the analysis source with a scanning area of ~50 × 50 µm², and cesium (Cs⁺) ions were used as the sputter source with the area of ~200 × 200 μm². Negative ion signals were detected to investigate the three-dimensional spatial species distribution.

Device fabrications and measurements

Monolayer WS₂ film was transferred to ~50 nm thick SiO₂/Si substrate, and Au/Ti/Pd (~3/2/15 nm) were deposited as composite metal contact electrodes using an electron beam evaporation system (DETECH DE400). The electrical measurements were carried out with the four-probe system (MPI TS200-SE) using a semiconductor parameter analyzer (Keithley 4200A-SCS) in vacuum (<1.3 mTorr) at room temperature. The μ_FE is calculated by the Eq. (1):

$${\mu }_{\mathrm{FE}}=\frac{L\left(\mathrm{d}{I}_{\mathrm{ds}}/\mathrm{d}{V}_{\mathrm{g}}\right)}{W{C}_{\mathrm{ox}}{V}_{\mathrm{ds}}}$$ 1

where C_ox is the gate capacitance of SiO₂, which is calculated by C_ox = ε₀ε_r/d (ε₀ is the vacuum permittivity, ε_r (3.9) and d (~50 nm) are the dielectric constant and thickness of SiO₂, respectively). The dI_ds/dV_g is the differential value of the transfer curve.

DFT Calculations

The construction of the atomic models and their visualization were performed using the Device Studio program developed by Hongzhiwei Technology (Shanghai) Co., Ltd. The electronic structure calculations were performed using the DFT method implemented in the DS-PAW software package⁵². The Perdew-Burke-Ernzerhof (PBE) functional under the generalized gradient approximation (GGA) was employed to describe the exchange-correlation interaction^53,54. The van der Waals correction was described by Grimme’s DFT-D3 scheme⁵⁵. The plane-wave basis set was truncated at ~520 eV. A vacuum layer of ~50 Å was introduced in the out-of-plane direction to screen the interlayer interactions. The convergence criteria for electronic self-consistent energy were set to ~10⁻⁵ eV, with the ionic step forces converging to ~0.02 eV/Å during the structure optimization.

For the Al₂O₃(0001) substrate with single Al-terminated surface (sAl-TS), a Gamma-centered k point grid of (7 × 7 × 1) was utilized to sample the Brillouin zone. For the substrate with a modification layer surface (MLS), the same k-mesh was used; during the relaxation process, only the surface O-W-O layer was allowed to relax, while all deeper atoms were fixed to reduce computational expense. To determine the optimal orientation of monolayer WS₂ on each substrate, a monolayer W₇S₁₂ cluster representing the WS₂ nucleus was placed on the (4 × 4) supercell of Al₂O₃(0001) with either sAl-TS or MLS. For these interface supercell models, total energies were obtained from self-consistent calculations using a (2 × 2 × 1) k-point mesh. The interfacial binding energy is defined as Eq. (2):

$$E_{\mathrm{b}}=\frac{E_{\mathrm{tot}}-E_{{\mathrm{W}}_7{\mathrm{S}}_{12}}-E_{\mathrm{sub}}}{N_{\mathrm{W}}}$$ 2

where E_tot, $E_{{\mathrm{W}}_7{\mathrm{S}}_{12}}$, and E_sub represent the total energy of the entire system, the isolated W₇S₁₂ cluster, and the bare substrate (sAl-TS or MLS), respectively, while N_W = 7 denotes the number of W atoms in the cluster.

Author contributions

Y. Zhang and J.W. conceived the idea and wrote the paper; J.W. performed the substrate processing, CVD growth, XPS, SEM, AFM, HRTEM, SAED, in-plane XRD, and spectroscopic characterizations with the help from Y.P., Y.C., H.D., and X. Wang; Y. Zheng, C.Z., C.Q., and L.P. performed device fabrications and electrical measurements; Q.Z., X. Wu, and Q.J. performed theoretical calculations; Y.J., T.Z., L.Y., and L.L. helped the transfer processes; W.W. performed ToF-SIMS; J.D. and J.X. performed HAADF-STEM and EDS; all authors discussed the results and commented on the manuscript.

Peer review

Peer review information

Nature Communications thanks Fangyuan Zheng, Yoshiki Sakuma, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

Y. Zhang acknowledges the support by the National Key Research and Development Program of China (No. 2021YFA1202901), and the National Natural Science Foundation of China (NSFC, Nos. 92477204, T2188101). C. Q. acknowledges the support by the NSFC (No. T2588301). J. X. acknowledges the support by Beijing Natural Science Foundation Non-Consensus Innovation Project (No. F261029). Q. J. acknowledges the support by the National Key Research and Development Program of China (No. 2024YFA1410500). H. D. acknowledges the support from the China Postdoctoral Science Foundation (2025M780008).

Data availability

Relevant data supporting the key findings of this study are available within the article and the Supplementary Information file. All raw data generated during the current study are available from the corresponding authors upon request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77131-w.