ARTICLE https://doi.org/10.1038/s41467-019-13631-2 OPEN Engineering grain boundaries at the 2D limit for the hydrogen evolution reaction Yongmin He1,2,20, Pengyi Tang 3,4,20, Zhili Hu5,6,20, Qiyuan He 1,20, Chao Zhu1, Luqing Wang6, Qingsheng Zeng1, Prafful Golani1, Guanhui Gao7, Wei Fu1, Zhiqi Huang1, Caitian Gao8, Juan Xia9, Xingli Wang10, Xuewen Wang11, Chao Zhu 1, Quentin M. Ramasse12,13, Ao Zhang 1, Boxing An14, Yongzhe Zhang14, Sara Martí-Sánchez3, Joan Ramon Morante 4, Liang Wang 15, Beng Kang Tay 10, Boris I. Yakobson6, Achim Trampert7, Hua Zhang 1,16, Minghong Wu 15*, Qi Jie Wang 2,17*, Jordi Arbiol 3,18* & 1234567890():,; Zheng Liu 1,8,17,19* Atom-thin transition metal dichalcogenides (TMDs) have emerged as fascinating materials and key structures for electrocatalysis. So far, their edges, dopant heteroatoms and defects have been intensively explored as active sites for the hydrogen evolution reaction (HER) to split water. However, grain boundaries (GBs), a key type of defects in TMDs, have been overlooked due to their low density and large structural variations. Here, we demonstrate the synthesis of wafer-size atom-thin TMD films with an ultra-high-density of GBs, up to ~1012 cm−2. We propose a climb and drive 0D/2D interaction to explain the underlying growth mechanism. The electrocatalytic activity of the nanograin film is comprehensively examined by micro-electrochemical measurements, showing an excellent hydrogen-evolution perfor- mance (onset potential: −25 mV and Tafel slope: 54 mV dec−1), thus indicating an intrin- sically high activation of the TMD GBs. 1 School of Materials Science and Engineering, Nanyang Technological University, Singapore 639798, Singapore. 2 Center for OptoElectronics and Biophotonics, School of Electrical and Electronic Engineering & The Photonics Institute, Nanyang Technological University, Singapore 639798, Singapore. 3 Catalan Institute of Nanoscience and Nanotechnology (ICN2), CSIC and BIST, Campus UAB, Bellaterra, Barcelona 08193 Catalonia, Spain. 4 Catalonia Institute for Energy Research (IREC), Jardins de les Dones de Negre 1, Sant Adrià del Besòs, Barcelona 08930 Catalonia, Spain. 5 College of Aerospace Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China. 6 Department of Materials Science and NanoEngineering, Rice University, Houston, TX 77005, USA. 7 Paul-Drude-Institut für Festkörperelektronik Leibniz-Institut im Forschungsverbund Berlin Hausvogteiplatz, 5-7, 10117 Berlin, Germany. 8 Centre for Micro-/Nano-electronics (NOVITAS), School of Electrical & Electronic Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore. 9 Institute of Fundamental and Frontier Sciences, University of Electronic Science and Technology of China, Chengdu 610054, China. 10 CNRS-International-NTU-THALES Research Alliance, Nanyang Technological University, Singaproe 637553, Singapore. 11 Institute of Flexible Electronics, Northwestern Polytechnical University, Xi’an 710072, China. 12 SuperSTEM Laboratory, SciTech Daresbury Campus, Keckwick Lane, Daresbury WA44AD, UK. 13 School of Chemical and Process Engineering, University of Leeds, Leeds LS29JT, UK. 14 College of Materials Science and Engineering, Beijing University of Technology, Beijing 100124, People’s Republic of China. 15 School of Environmental and Chemical Engineering, Shanghai University, Shanghai 200444, China. 16 Department of Chemistry, City University of Hong Kong, Tat Chee Avenue, Kowloon, Hong Kong, China. 17 CINTRA CNRS/NTU/THALES, UMI 3288, Research Techno Plaza, Singapore 637553, Singapore. 18 ICREA, Pg. Lluís Companys 23, Barcelona 08010 Catalonia, Spain. 19 Environmental Chemistry and Materials Centre, Nanyang Environment and Water Research Institute, Singapore, Singapore. 20These authors contributed equally: Yongmin He, Pengyi Tang, Zhili Hu, Qiyuan He. *email: mhwu@shu.edu.cn; qjwang@ntu.edu.sg; arbiol@icrea.cat; z.liu@ntu.edu.sg NATURE COMMUNICATIONS | (2020)11:57 | https://doi.org/10.1038/s41467-019-13631-2 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13631-2 G rain boundaries (GBs) are commonly found in atom-thin Al2O3) and Au42,43 (Supplementary Fig. 3). The initial deposition or so called two-dimensional (2D) polycrystalline mate- time of the Au films prior to heating determines the density and rials1–3, where they can be described as line defects. They the size of the final Au QD structures: a shorter deposition time play a key role in shaping the properties and performance of 2D results generally in a smaller size and a higher density of Au QDs, materials in applications as varied as mechanical strengthening4, as shown in Fig. 1e and Supplementary Fig. 4. The MoS2 film is photovoltaics5,6, electronics7–9, and catalysis10,11. Engineering the then grown using a vapor-phase growth technique (see further structure and/or the density of GBs in 2D materials could thus experimental details in Method section), after which Au QDs can become a promising way to tailor their performance. One parti- be removed from as-grown films, by using KI/I2 etchant at room cular class of 2D materials, transitional metal dichalcogenides temperature. The successful removal of the Au QDs after the (TMDs) have attracted a great deal of attention for its possible etching treatment was assessed by X-ray photoelectron spectro- uses in electrocatalytic reactions12,13, including the hydrogen scopy and STEM imaging in Supplementary Figs. 5–7, showing evolution reaction (HER)14,15. Due to the low cost, earth abun- only minimal amounts of residual Au present. The resulting dance and good stability of a wide range of TMDs, substantial MoS2 films were examined by Raman spectroscopy, as shown in work has thus been undertaken to improve their electrocatalytic Fig. 1f. The Raman spectroscopy measurements indicate that the activity12,13 by, e.g., exposing their edges16–19, doping with het- MoS2 film is 1–3 layers thick (1–3 L). A weak in-plane mode (E2g1 eroatoms20, and/or creating and straining structural at 385 cm−1) was observed in our 1–3 L film, with an intensity defects11,15,19,21,22. In contrast, less attention has been paid to the ratio with the out-of-plane mode A1g (at 405 cm−1), IE2g1 =IA1g role of GBs due to their typically low number density and large (~0.28), significantly smaller than the ratios typically measured structural variations, even though GBs have been predicted to be on the CVD-grown (~1.89) and exfoliated (~1.57) MoS2 films in highly electrocatalytically active23. The poor control over the our experiment, suggesting a highly polycrystalline structure in density and structure of GBs stems from the fast gaseous kinetic this film (see the details in Supplementary Table 1 and Supple- processes and the multiplicity of chemical phases involved during mentary Note 1). the growth of TMDs2,24,25. To date, the most common methods Au QDs regulate the formation of TMD grains in two ways. employed to synthesize atomically thin polycrystalline TMDs The first way is that Au QDs can facilitate the TMD nucleation at include chemical vapor deposition (CVD)2,3,26–36, physical vapor the initial stage of the TMD growth. The melting temperature of deposition (PVD)37–39, and metal organic chemical vapor Au will dramatically decrease as its size decreases to several deposition (MOCVD)40,41. Films grown using these techniques nanometers, as shown in Supplementary Fig. 8. As a result, at the usually exhibit grain sizes ranging from hundreds of nanometers growth temperature of TMDs (650–800 °C), Au QDs tend to be to several millimeters, resulting in a low GB density, as sum- in the liquid phase due to their small size. Such liquid Au droplets marized in Fig. 1a (See the statistical method in Supplementary can facilitate the formation of the TMD nucleation sites during Fig. 1). the growth process. The second way is that the Au QDs can Here, we fabricate wafer-size atomically thin TMD films with confine the size of the TMD domains below 10 nm. Our sub-10 nm grains by means of Au-quantum-dots (QDs)-assisted deposition method can produce well-dispersed Au QDs with a vapor-phase growth, demonstrating an ultra-high-density of GBs very high number density, with average spacing between QDs (up to ~1012 cm−2). The quality of the films was examined by down to a few nanometers. As a result, the formation of the TMD high-resolution transmission electron microscopy (HRTEM), grains will be constrained at this length scale. aberration-corrected scanning transmission electron microscopy (STEM), scanning electron microscopy (SEM), and Raman spectroscopy. Experimental evidence as well as phase-field Atomic structure of the TMD nanograin films. We investigated simulations demonstrate that the Au QDs regulate the forma- the atomic structure of the TMD nanograin films using trans- tion of the TMD grains. We then investigated the catalytic activity mission electron microscopy (TEM). Figure 2a shows an overview of these nanograin films by a four-electrode micro-electro- TEM image of a uniform and continuous MoS2 film suspended chemical cell for hydrogen evolution. An excellent performance on a Cu-supported lacey carbon TEM grid. This region is (−25 mV and 54 mV dec−1 for the onset potential and the Tafel representative of the tens of samples examined, and the MoS2 film slope) was obtained for our MoS2 nanograin films, indicating a comprises polycrystalline patches with regions of 1–3 L MoS2 good intrinsic activation of the GB-rich 2D basal plane. (consistent with the measured Raman frequencies44). To evaluate the grain distribution, we randomly chose six regions of the film, labeled 1–6 on Fig. 2a, for further HRTEM imaging. The original Results HRTEM images from these six regions, corresponding Fourier Controlled growth of the TMD nanograin film. The main transforms (FFTs) and frequency filtered images (IFFTs)45 are obstacles that hinder the vapor growth of TMDs at the 2D limit to shown in the top (a1–a6), middle (b1–b6), and bottom panels grain sizes typically <10 nm consist of the precise control of the (c1–c6) of Fig. 2a, respectively. To aid the visualization of the nucleation sites and the growing rate of the grains. These can be different grains, black-dashed lines are added along the edges of addressed by using a high-density of Au QD seeds and a low mass the grains in the bottom panel (c1–c6). From these images, we flow rate of the vapor sources. Figure 1b illustrates the wafer-scale can identify 8–10 distinct MoS2 grains in a ~700 nm2 region, growth process (see also a full description of the growth setup in suggesting an ultra-high grain density (up to ~1012 cm−2 in this Supplementary Fig. 2). A wafer-scale Au QD layer was fabricated representative patch, consistent with the initial estimate derived on 2-inch sapphire or SiO2/Si substrates and used to subsequently from the Au QD number density) in our sample. Accordingly, the grow the atom-thin MoS2 films. Figure 1c shows the geometries average diameter of the grains is <10 nm, with some observed of the as-grown wafer-size film from the centimeter to the nan- grains smaller than 5 nm in diameter. To our knowledge, this is ometer scale. The SEM image (Fig. 1d) shows the as-prepared Au the smallest grain size obtained to date in materials at the 2D QDs with an ultra-high density up to ~2 × 1012 cm−2 and average limit17,46,47. diameter down to ~4.8 nm, which was achieved by heating a thin We then employed atomic-resolution high-angle annular dark- Au film on sapphire or SiO2/Si substrates at a high temperature. field (HAADF) imaging in the STEM to examine the atomic We attribute the formation process of these Au QDs to a solid- structure of the GBs on the as-grown nanograin films, state dewetting behavior at the interface between the SiO2 (or representative examples of which are shown in Fig. 2b, c and 2 NATURE COMMUNICATIONS | (2020)11:57 | https://doi.org/10.1038/s41467-019-13631-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13631-2 ARTICLE a Sub-10 nm at 2D limit 1012 Grain boundary density (cm–2) 109 106 103 100 100 101 102 103 104 105 106 107 Grain size (nm) b Nanograin film after Wafer (Si, sapphire) Deposition of Au QDs Growth of TMDs removing Au QDs c ~5 cm (2 inch) 10 μm 200 nm 10 nm d e 35 f 30 s Δ ≈ 20 cm –1 ×2 30 70 s 25 Intensity (.a.u) 100 s % of counts This work 20 300 s Mechanical 15 exfoliation 10 E2g1 A1g 5 CVD-growth 0 0 5 10 15 20 25 360 390 420 450 480 510 540 Au particle size (nm) Shift (cm–1) Fig. 1 Synthesis of wafer-size TMD nanograin films. a Overview of the grain size and density in TMD materials obtained by various fabrication methods, such as CVD2,3,26–36, MOCVD40,41, PVD37–39, top–down syntheses of sulfurization/selenization of metal and metal oxide thin films73–77, and thermal decomposition of thiosalts thin films78. Although small grains (~20 nm in diameter) were observed in TMDs synthesized by the top–down methods of sulfurizing Mo-based thin films73,74, or thermal-decomposition of Mo-thiosalts thin films78, the grain density in the film is not very high owing to their interlayer diffusion synthesis mechanism. b Schematic of the wafer-scale growth of TMD nanograin films. Ultra-high-density Au quantum dots (QDs) were used to grow the MoS2 (as well as WS2 atom-thin films: see Supplementary Information). c TMD nanograin film from wafer scale to nanoscale, including photograph (i), optical image (ii), SEM image (iii), and HRTEM image (iv). d False-colored SEM image of the Au QDs on a SiO2/Si substrate, showing an average diameter of 4.8 nm. e Statistical distributions of Au QDs obtained from Au films with different deposition times. The evaporation rate is ~0.1 Å s−1 in our experiments. f Raman spectra acquired from the MoS2 film. The difference (Δ ~20 cm−1) between the out-of-plane (A1g) and in-plane (E12g) Raman modes indicates that the MoS2 film is 1–3 layers (1–3 L)2,44. Compared to CVD-grown and exfoliated samples, the reduced intensity of the in-plane mode indicates a highly polycrystalline structure for this MoS2 film. Supplementary Fig. 9. The inspection of dozens of boundary found, but in contrast, 12|4 rings were never observed in locations systematically found that nanograins in our films are “pristine” GBs (i.e., before any substantial exposure to the stitched by GBs. The atomic structure of the GBs varied electron beam). Their atomic structures in MoS2 are schemati- significantly (great care was taken to exclude any structure cally illustrated in Supplementary Fig. 10. modification arising from beam damage, see Experimental section Although the nature of atomic-resolution microscopy pre- for details). Structures comprising a combination of 5|7 and 6|8 cludes large-scale statistical studies, we investigated as many rings were however dominant in our nanograin films and are different regions in a MoS2 nanograin film as practically possible. illustrated in Fig. 2c. A small number of 8|4|4 structures were Six further regions over a large range were carefully examined NATURE COMMUNICATIONS | (2020)11:57 | https://doi.org/10.1038/s41467-019-13631-2 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13631-2 a 5 2 1 8 Grains 7 Grains 9 Grains 4 3 6 8 Grains 500 nm 8 Grains 7 Grains a1 a2 a3 a4 a5 a6 5 nm 5 nm 5 nm 5 nm 5 nm 5 nm b1 b2 b3 b4 b5 b6 c1 c2 c3 c4 c5 c6 5 nm 5 nm 5 nm 5 nm 5 nm 5 nm b c d 140 120 100 Grain size (nm) 1.0 80 60 40 20 0 2 nm 2 nm 0.5 nm 3 6 9 12 15 18 –1.0 AU nanoparticles (nm) Fig. 2 Atomic structure of the TMD nanograin film. a TEM image of the 1–3 L MoS2 nanograin film, showing a uniform and continuous atom-thin layer with Au nanoparticles (dark spots). Top panel (a1–a6): HRTEM images from regions 1–6. Middle panel (b1–b6): Fourier transforms (FFTs) of regions 1–6 from the corresponding TEM image. Bottom panel (c1–c6): false-colored frequency filtered (IFFTs) images of the same regions. Grains with different orientations give rise to distinct sets of rotated spot patterns in the FFTs, as indicated by colored circles in the middle panel (b1–b6); in turn, the localization of the grains can be visualized in real space by color coding the corresponding contribution to the IFFT of a given set of diffraction spots, and overlaying the IFFTs into a composite colored image (in bilayer regions, the colors are combined). In order to better visualize the different grains, the grains are highlighted in the IFFT images with black-dashed lines. b HAADF STEM investigation of MoS2 grains. Left: HAADF STEM image showing the GBs between three MoS2 grains. Inset on the left figure: Fourier transform of the image showing three distinct sets of MoS2 diffraction patterns with rotation angles of 11.6° and 25.6°. Middle: composite color-coded IFFT image. Right: dilatation map of b after applying GPA routine to the monolayer MoS2. c Higher magnification atomic resolution HAADF STEM image of the brown-dashed line marked region in b, showing the detailed atomic structure of the GB, which is composed of 5|7 and 4|8 rings. Mo atoms are marked with indigo dotted circles and S atoms marked with yellow dotted circles. d Grain size dependence on Au nanoparticle size (average diameter) in the MoS2 nanograin film. The error bars of Au nanoparticles are extracted from the full-width at half maximum in Fig. 1e. with atomically resolved HAADF–STEM imaging (Supplemen- density). Moreover, tens of samples with different sizes of Au tary Fig. 11), and 3 to 5 GBs can be observed in a 400 nm2 area, QDs were grown: the grain size measured in the resulting MoS2 confirming the ultra-high-density of GBs observed in STEM films exhibits a clear linear relationship with the QD size, as (consistent with the ~1012 cm−2 suggested by the Au QD shown in Fig. 2d. This provides a strong proof that the TMD 4 NATURE COMMUNICATIONS | (2020)11:57 | https://doi.org/10.1038/s41467-019-13631-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13631-2 ARTICLE grain size can be carefully controlled by the Au QD substrate. Mo edge (0.561 eV), and 50% passivated S edge (0.446 eV)53,54. It Remarkably, our method can be also extended to other TMD can be seen that GBs indeed show comparable or even better materials such as WS2 (Supplementary Fig. 12 summarizes activity than the edges in general, suggesting GBs are promising similar results to those described above, using WS2 instead of candidates as high-efficient catalyst sites. In addition, we note that MoS2), making it as a universal approach for the wafer-size some edges are not further considered here owing to their synthesis of atom-thin films with sub-10 nm grains. instability in air; for instance, the otherwise energetically favor- able Mo edge without passivation of S atoms (−0.446 eV). We then developed a micro-electrochemical cell to investigate Growth mechanism of the TMD nanograin film. A climb and the HER activity of our sub-10 nm nanograin films. Furthermore, drive zero-dimensional (0D)/2D interaction is proposed to the HER activity of model single-GBs (mirror GB)3,11, single-edge explain the Au QDs-assisted-growth mechanism of the TMD (Supplementary Fig. 19)53, and basal plane structures in MoS2 nanograins at the atomic limit, which is supported by real-time samples from CVD method1 were also examined for comparison. phase-field simulations and experimental evidence. The phase- Figure 5a illustrates the micro-cell’s structure. Figure 5b shows a field simulation of the climb process shows that, once the Au QDs picture of the micro-cell and electrodes configuration where a encounter a MoS2 edge at the growth front, they migrate swiftly graphite counter electrode and a micro-reference electrode were from the SiO2 surface onto the MoS2 surface (Fig. 3a). This used. In our experiment, a vertical MoS2/graphene heterostruc- behavior is mainly attributed to a smaller wetting angle of the Au ture was designed, in which graphene played two important roles: QDs droplet (Au is in the liquid phase at the growth temperature) one is to efficiently inject electrons to MoS2. Such a strategy has on MoS2 than that on the SiO2 substrate48 (Supplementary already been widely adopted in TMD-based semiconductor Fig. 13). This phenomenon is also experimentally confirmed by devices55,56, while the other is to provide a fair comparison by cross-sectional TEM, as shown in Supplementary Fig. 14, where eliminating the contact resistance variations of the individual nearly all the Au QDs sit on the surface of the MoS2. Subse- MoS2 electrodes onto the substrate. Previous work56,57 suggested quently, the phase-field simulation suggests the formation of a the formation of an inconsistent contact barrier between the second MoS2 layer will drive the Au QD droplets along its growth MoS2 and the Au electrode, owing to their complex direction (Fig. 3b). A number of the Au QDs may thus coalesce metal–semiconductor interface leading to effects such as Fermi and form bigger droplets during the driving process. This is pinning, creation of alloy structures, etc. The formation of such a supported by our experimental evidence: the size of the Au QDs contact barrier was also studied by phase engineering21 and field- on the MoS2 is usually larger than those on SiO2, and some of effect gating58 in the HER process. Considering the important them sit at the edge of the MoS2 sheet, as illustrated in Fig. 3c and role of this graphene supporting layer, we introduced one more Supplementary Fig. 15a–b. This drive process will be immobilized working electrode in the micro-electrochemical cell (four- when the Au QDs reach a critical size (as a result of the Au electrode micro-electrochemical cell59,60, see circuit diagram in returning to the solid phase). This would be consistent with Supplementary Fig. 20) to monitor in situ its conductance. The experimental observations that larger Au particles seen in the device fabrication of the MoS2/graphene heterostructure is minority multilayer regions of the nanograin film appear to be detailed in Supplementary Figs. 21 and 22, and Supplementary encapsulated by the MoS2 film, suggesting they were immobilized Notes 2 and 3. The heterostructure was also characterized by after reaching a critical size, leading the MoS2 to grow over them Raman spectroscopy (Supplementary Fig. 23). Figure 5c–f shows rather than driving them forward (see Supplementary Figs. 15c, d the optical images obtained from the typical devices fabricated and 16). with a MoS2 nanograin film, a single-GB model structure (the Figure 3d shows successive snapshots that illustrate the growth presence of a GB was confirmed by Raman mapping, as of TMD nanograins films via a phase-field simulation (also see illustrated on Fig. 5d, right panel), a single edge, and a basal Supplementary Movie 1). The Au QDs serve as nucleation sites to plane, respectively. In these optical images, only the exposed form the first MoS2 layer (Fig. 3d–i). The growth of a second MoS2 in the reaction window contributes to the electrocatalytic MoS2 layer will then drive the Au QDs. Some of the Au QDs will reaction, as indicated by the white arrows. The rest of the areas be immobilized at the GBs (Fig. 3d ii and iii, number 1–4). More are covered by poly(methylmethacrylate) (PMMA), and the interestingly, we found that neighboring grains could be sutured graphene supporting layer (Supplementary Fig. 24) are also together through a zipper effect (Fig. 3d iii-v, number 5 and 6, see electrochemically inert. Method and Supplementary Fig. 17). It is important to emphasize Figure 5g, h shows the polarization curves and the correspond- that the growth model for the TMD nanograin films we describe ing Tafel slopes in a 0.5 M H2SO4 solution and provide several here is driven by a 0D/2D interaction at the atomic limit, which is interesting observations. First, the resistivity of the graphene distinct from the conventional metal-film-assisted growth of 2D supporting layer is found to be as low as 10−4 Ω mm during the materials (2D/2D)29,35, or from the metal-nanoparticle-assisted HER (see the inset in Fig. 5g), indicating the excellent current growth of nanowires (0D/1D)49. injection performance of the graphene layer. Secondly, the single- GB device delivers a better activity than the single-edge device Hydrogen production of the TMD nanograin films. We first (see also the measurement of the active length per unit area of the performed first-principle calculations50 to examine all of the HER GB in Supplementary Fig. 25), and both are superior to the basal catalytic active sites in model MoS2 geometries (Fig. 4a), plane device, echoing our theoretical calculation results shown including the basal plane sites, edges with different passivation above. More importantly, our MoS2 nanograin film shows a (Mo or S)51, and GBs with different atomic structural config- remarkable HER performance: −25 mV and 54 mV dec−1 for the urations (5|7, 6|8, 4|6, 12|4, and 8|4|4 rings). The calculated onset potential and the Tafel slope, respectively. hydrogen adsorption free energies (ΔGH)52 of various atomic In order to evaluate the HER performance of the TMD structures (Supplementary Fig. 18) are shown in Fig. 4b and nanograin film accurately, we have fabricated hundreds of devices Supplementary Table 2. The MoS2 basal plane has a ΔGH as high and tested them in a size-controlled micro-electrochemical cell. as 1.79 eV15, indicating a HER-inert surface. For edges and GBs, The HER data, including current density, Tafel slope, and onset we compare the activities of most of the experimentally observed potential of the nanograin film and other types of electrostatically structures, e.g., 5|7 (0.132 eV), 6|8 (−0.237 eV), and 8|4|4 (0.52 active sites are presented in Fig. 6a–c and Supplementary Fig. 26. eV, −0.044 for defect) ring combinations,2,53 50% S passivated The window size is controlled from 25 to 150 μm2. Our results NATURE COMMUNICATIONS | (2020)11:57 | https://doi.org/10.1038/s41467-019-13631-2 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13631-2 a Climb b Drive Initial Initial 2 nd layer MOS Low surface tension 2 High surface tension MOS 2 Au nanopa rticle 1 st layer SiO MO S 2 2 Growth of MOS 2 Drive Climb Final Final c d (i) (ii) t1 = 0.1 ms t2 = 0.5 ms (iii) (iv) (v) 0 2 π/3 t3 = 1 ms t4 = 1.5 ms t5 = 1.9 ms Fig. 3 Proposed self-limiting climb and drive growth mechanism of the TMD nanograin films. a Schematic of the climb stage: the Au QDs droplets tend to climb onto the MoS2 monolayer from the SiO2 substrate due to the surface tension difference. b Schematic of the drive stage: the growth of a second MoS2 layer tends to push the Au QDs along the growth direction. Most of the Au QDs will be stabilized at the grain boundaries in order to minimize their surface energy. c SEM image showing the distribution of the Au QDs on the edges of growing MoS2 layers (marked by the red-dashed line). Au QDs are usually larger on the MoS2 layers (indicated by blue arrows) than those observed on the SiO2 substrate. d Real-time phase-field simulation showing the formation of the MoS2 nanograin film. The Au QDs numbered 1–4 show that the growth of the MoS2 layer will push Au QDs toward the grain boundaries. Au QDs numbered 5–6 show a zipper effect to suture the neighboring grains together. The movie can be found in Supplementary Movie 1 (Real-time phase-field simulation of the growth process). 6 NATURE COMMUNICATIONS | (2020)11:57 | https://doi.org/10.1038/s41467-019-13631-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13631-2 ARTICLE a b 2.1 Electrolyte 1.8 1.5 H2 generation 1.2 ΔGH (eV) H2 H2 H2 H2 e– H + H2 H2 H + – e Edges 0.9 GBs 0.6 0.3 Basal plain 0.0 Pla ne Ed ge GB Fig. 4 Theoretical calculations of the catalytic sites in MoS2 for HER. a Schematic of MoS2 with an illustration of the main types of catalytically active sites for the hydrogen evolution reaction (HER) including: basal plane atomic sites, edges, and GBs. b corresponding ΔGH of various types of catalytically active sites in MoS2 catalysts. Some types of GBs (e.g., 844, 6–8, and 5–7) show better electrocatalytic activity than edges. demonstrate the excellent HER performance of the nanograin 2D plane. Beyond electrocatalysis, the nanograin films may pro- film. The HER current density of the nanograin film is up to vide further diverse potential applications, such as in resistance- ~1000 mA cm−2 while the Tafel slope and onset potential are memory devices, flexible devices, or for use as mechanical films down to ~50 mV dec−1 and ~−25 mV, thus exhibiting a and molecular sieving films. performance superior to devices using the basal plane of the CVD film, the single edge and the single GB. This suggests that Methods sites on the otherwise HER-inert basal plane of the MoS2 have Growth of water-size MoS2 and WS2 nanograin films. The first step consists in been activated by the presence of GBs. The obtained HER depositing Au layers of various thicknesses on clean 2-inch sapphire or SiO2/Si wafers by e-beam evaporation at the rate of 0.1 Å s−1; these are used as the growth performance is also comparable to the performance of strained substrate for the nanograin films. Subsequently, Au-coated wafers are introduced sulfur vacancies15 or the metastable 1T-phase61, which are into a CVD apparatus, which is depicted in detail in Supplementary Fig. 2. MoO3 arguably more challenging to realize experimentally. or WO3 powders were loaded in an aluminum oxide boat and used as sources. Finally, as a comparison, we have conducted conventional Sulfur powder is placed in a second aluminum oxide boat upstream of the MoO3 or WO3 in the outer tube. In a third step, Ar (flow 100 sccm) and Ar/water vapor macro-cell measurements on a MoS2 nanograin film on a glassy (flow 50 sccm) are introduced as carrier gases for the growth in the inner and outer carbon electrode, as shown in Supplementary Figs. 27–32 and tubes, respectively. The growth temperature was kept at 780 °C in low-vacuum Supplementary Notes 4 and 5. A remarkable HER performance conditions of 1–10 kPa; the S powder is separately kept at ~160 °C. Adjusting the (−44 mV and 55 mV dec−1 for the onset potential and the Tafel growth time is used to control the thickness of films. For example, Condition (I): slope, respectively) of the MoS2 nanograin film is observed 3–5 min of the growth time for 1–3 layers (1–3L); Condition (II): 5–10 min of the growth time for 4–5 layers (4-5 L). The growth parameters for WS2 nanograin films (Supplementary Fig. 28), which is consistent with the results are similar to those used for MoS2, except for a higher growth temperature (900 ° obtained in the micro-cell. It also shows an excellent long-term C). Finally, Au nanoparticles can be completely removed from as-grown nanograin stability for HER, as shown in Supplementary Fig. 29. It is worth MoS2 or WS2 films by using a KI/I2 etcher for 40–60 min at room temperature (see mentioning that Au single-atom exists in our MoS2 nanograin Supplementary Figs. 5 and 6). Special care needs to be taken on two important points: (i) the Au QDs will film and may contribute to the overall HER performance. form once the temperature is above 300 °C, due to the solid-state dewetting However, based on the structure (Supplementary Fig. 6) and behavior42,43,62 at the interface of SiO2 (Al2O3) and Au (see Supplementary Fig. 3). extremely low content (Supplementary Fig. 7) of Au single atoms Prior to the TMD growth, the substrate therefore stabilizes at this temperature to as well as their low HER-activity (Supplementary Figs. 33 and 34, ensure the presence of Au QDs. (ii) Small amounts of water vapor are used to prevent the Mo or W sources from complete sulfurization before they reach the and Supplementary Note 6), we conclude that the residual Au is growth wafer, thus enabling a steady supply of sources during growth. not the main contributor of the good HER activity in MoS2 nanograin film. The measured performance should be mainly Theoretical calculations of ΔGH. The structural optimizations were carried out by from GB due to its superior activity and ultra-high density. As a adopting the generalized gradient approximation (GGA) with the proof-of-concept practical application, hydrogen production on a Perdew–Burke–Ernzerhof (PBE) exchange-correlation functional, along with wafer-size (2 inch) MoS2 nanograin film was also demonstrated projector-augmented wave (PAW) potentials. The electronic wave functions were expanded in a plane wave basis set with a kinetic energy cutoff of 300 eV. For the (see Supplementary Movie 2). A large amount of H2 bubbles Brillouin zone integration, 1 × 5 × 1 Monkhorst–Pack k-point meshes were used. could be produced during the reaction, demonstrating the The energy convergence criterion for the electronic wavefunction was set at 10−5 tantalizing potential for practical application of GBs-engineered eV. A vacuum distance of about 10 Å, both laterally between MoS2 ribbons and catalysts for hydrogen production. vertically between layers, was chosen when constructing the supercell to minimize possible spurious interaction between ribbons due to the periodic boundary conditions. The hydrogen adsorption energy is defined as: Discussion In summary, we have engineered wafer-size ultra-high-density ΔEH ¼ EðMoS2 þHÞ  EðMoS2 Þ  1 =2 EðH2 Þ ð1Þ GBs at the 2D limit in TMD films. Phase-field simulations reveal where EðMoS2 þHÞ is the energy of the MoS2 system with a H atom adsorbed, EðMoS2 Þ a climb and drive 0D/2D-interaction-based growth mechanism. is that of the MoS2 system without adsorbed H, and EðH2 Þ is that of gas phase H2 As a proof-of-concept application in electrocatalysis, devices molecule. based on our TMD nanograin films deliver a superior hydrogen The hydrogen adsorption free energy ΔGH was defined as: evolution performance (onset potential: −25 mV and Tafel slope: 54 mV dec−1), indicating an intrinsic activation of the GB-rich ΔGH ¼ ΔEH þ ΔEZPE  TΔSH ð2Þ NATURE COMMUNICATIONS | (2020)11:57 | https://doi.org/10.1038/s41467-019-13631-2 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13631-2 a Counter b Reference electrode Vc electrode Reference V electrode Ic 0.5 M H2SO4 e– e– e– Counter Working electrodes electrode Effecient injection (I) PMMA MoS2/graphene (II) Ids Au Au Vds SiO2 Si c d Window on GB Raman mapping of GB SiO2/Si substrate GB Graphene Window 10 μm 1 μm e f Nanograin film 10 μm 10 μm 10 μm Electrodes Window on edge Window on plain g h 0 Plane Edge 0.5 Grain Nano Pt Plane Edge 1–3 L nanograin Grain Nano Pt –100 0.4 4–5 L nanograin Onset potential (V, vs RHE) 1–3 L nanograin Current density (mA cm–2) 4–5 L nanograin 128 mV dec–1 –200 0.3 4.0 × 10–4 111 mV dec–1 0.2 103 mV dec–1 Resistivity (Ω mm) –300 3.0 × 10–4 75 mV dec–1 0.1 2.0 × 10 –4 37 mV dec–1 –400 54 mV dec–1 1.0 × 10–4 0.3 –0.4 –0.2 0.0 0.2 –500 Voltage (V, vs RHE) –0.1 –0.4 –0.2 0.0 0.2 0.4 –1.5 –1.0 –0.5 0.0 0.5 1.0 1.5 2.0 2.5 Potential (V, vs RHE) Log|(mA cm–2geo)| Fig. 5 Micro-electrochemical cell-based HER activity of a MoS2 nanograin film. a Schematic of the micro-electrochemical cell for HER measurements, where graphene serves as a vertical electron injector. b Photograph of the micro-electrochemical cell. c Optical images of the MoS2 nanograin device, consisting of a PMMA reaction window, a MoS2 nanograin film, a graphene supporting layer, and a SiO2/Si substrate from top to bottom. d–f Optical images of the MoS2 devices with a single GB (d), a single edge (e), and basal plane (f), respectively. In these devices, the HER process occurs at the exposed widows on the PMMA passivation film as indicated by the white arrow. g, h Polarization curves of the current density (g) and the corresponding Tafel plots (h) of the MoS2 devices. The window size is ~80 μm2 for each device. where ΔEH is the hydrogen adsorption energy, ΔEZPE is the zero-point energy should bind a key intermediate in a manner that it is strong enough to allow the difference, T is the temperature, and ΔSH is the entropy difference for hydrogen reagent H atoms to stay on the MoS2 system for reaction but weak enough to between the adsorbed state and the gas phase. The entropy of hydrogen adsorption enable the release of any produced H2 molecules. is calculated as ΔSH ¼ 1=2SH2 , where SH2 is the entropy of hydrogen molecule in the gas phase at standard conditions. We have calculated the vibrational frequencies of H adsorbed on different MoS2 systems, using finite differences to Phase-field simulations: simulation of the drive mechanism: MoS2 pushing Au determine the Hessian matrix. According to the Sabatier principle, a good catalyst QDs long the growth direction. The equilibrium status of the Au QDs at atomic 8 NATURE COMMUNICATIONS | (2020)11:57 | https://doi.org/10.1038/s41467-019-13631-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13631-2 ARTICLE a 10,000 b 200 c Plane Edge GB Plane Edge GB Plane Edge GB Onset potential (V, vs RHE) J (mA cm–2) at η = 400 mV 1–3 L nanograin 4–5 L nanograin 1–3 L nanograin 4–5 L nanograin 1–3 L nanograin 4–5 L nanograin 0.3 Tafel slope (mV dec–1) 160 1000 120 0.2 100 80 0.1 10 40 0.0 30 60 90 120 150 30 60 90 120 150 30 60 90 120 150 Window area (μm2) Window area (μm2) Window area (μm2) Fig. 6 Comparison of the HER performances in a size-controlled micro-electrochemical cell. a–c Current density (a), Tafel plots (b), and onset potentials (c) for the MoS2 devices with serial window sizes from 25–150 μm2. The nanograin films show higher current density, lower Tafel slope, and onset potential compared to the basal plane, single edge, and single GB. steps dictates they can be pushed by the MoS2 growth front or not, which is crucial QDs on top of MoS2 sheet, the final number of QDs will be 2/5 of the initial QDs. to dynamic evolution process. We undertook a case study of the equilibrium status A subtle issue involved in the simulation is the ratio between nAu (initial number of of a Au QD at the edge of a MoS2 grain. Au QDs) and nn (the number of nuclei). To address this issue, we take a simulation We adopt a phase-field model63 to simulate how the wetting process influences case study to find a reasonable ratio. By fixing nn = 5, we obtain that when nAu = the equilibrium shape profile. The free energy function is given as, 9.2, the final number of QDs is 2/5 of the initial QD number, consistent with the R experiments. We therefore choose nAu = 10 in further simulations. FðρÞ ¼ V ½2c ρ2 ðρ  1Þ2 þ κ2 ð∇ρÞ2 þ λξðρ3  ρ2 ÞdV 3 2 ð3Þ where ρ is the phase field for the density of the liquid, V is the space volume, C = Device fabrication procedure. First, large scale and high-quality single-layer 10 and κ = 1 are constants of bulk energy and surface tension, respectively. ρ = 1 graphene films were grown on Cu foils by CVD, and then transferred onto the denotes the liquid phase and ρ = 0 denotes the vapor phase. The term λξðρ3  ρ2 Þ 3 2 prepatterned chips through a standard PMMA-assisted transfer method. Second, provides the driving force to the Au particle evolution with a coupling constant electron beam lithography (EBL) and O2 plasma were employed to pattern the λ = 180, the dimensionless supersaturation ξ is set as (v0–vt)/vvapor, where v0 is the graphene film into isolated strips with desired size, shape, and location. Third, as- initial volume of the particle, vt is the real time volume of the particle, and vvapor is grown MoS2 films were transferred on the patterned graphene strips, also using a the volume of vapor. The evolution equation in the Allen–Cahn form is, PMMA-assisted transfer method. For the nanograin films, an additional step carefully positioning an EBL-defined PMMA template was needed. Fourth, an dρ annealing process at 200 °C under high vacuum condition (1 × 10−5 Torr) was ¼ κΔρ  Cρðρ  1Þð2ρ  1Þ þ λξðρ  ρ2 Þ ð4Þ dt conducted to remove trapped residual molecules between the graphene and MoS2 63 This evolution equation differs from that of Borcia et al.’s work , which films to optimize their interfaces. Fifth, EBL followed by thermal evaporation was considers full fluid dynamics by including extra terms from the Navier–Stokes employed to fabricate the electrodes (Cr/Au, 2 nm/60 nm) on graphene to connect equation. In our study, we focus on the equilibrium and do not need to consider the device with the Au contact pads. Finally, the entire device as fabricated on the the exact evolution kinetics, thus leading to a simpler form of the evolution chip was passivated by a 500 nm PMMA film, followed by EBL to remove the equation. PMMA above the region of interest in the MoS2 film, and expose this section and We control the contact angle between the Au QD and the substrate by fixing this section only of the active catalyst to the electrolyte in HER test. Any elec- ρ = ρS at the substrate. Then, the contact angle can be analytically obtained via the trochemical activity only occurs within the exposed window on the nanosheet, equation cos θ ¼ 1 þ 6ρ2S  4ρ3S . We set ρS = 0.5 for the MoS2 substrate as while the rest of regions, including the electrodes and the contact pads, are fully corresponding to a contact angle between the Au and the substrate of 90°. Likewise, passivated by the PMMA to ensure a full electrochemical inertness. we set ρS = 0.21 for the MoS2 substrate as corresponding to a contact angle between A mirror GB in CVD-grown MoS2 is chosen as representative of single GBs, the Au and the substrate of 138°. due to the ease with which it can be distinguished in optical images. It also consists We solve the evolution equation in Matlab by discretization with a time step of mainly of 8|4|4 or defect 8|4|4 rings3,11 and has therefore a similar atomic structure 0.01 and let the simulation run for a long enough time to ensure the system reaches to many of the complex boundaries observed in the nanograin films. The equilibrium. The particle has a diameter of ~11 nm if resting on the MoS2 sheet. length–width ratio of the exposed window for the single-GB device or for the edge The result shows that this Au QD will stay on the lower part of the MoS2 atomic device (exposing the edge of an otherwise single-crystalline patch of CVD-grown step, but keeps in contact with the step edge. This suggests that if in the liquid MoS2) is fixed at a ratio of about 2:1, and the length of the window is comparable to phase, the Au QD would be pushed forward by the growth front of the second the length of the edge or GB in experiments. MoS2 layer during the growth. Four-electrode micro-electrochemical measurements. A micro-electrochemical Phase-field simulations: simulation of the zipper effect suturing the neighbor cell with four electrodes (the circuit diagram for this cell is introduced in Sup- nanograins. If the Au QDs can be pushed by the growing MoS2 grains, they are plementary Fig. 20) was developed in our experiment. Among the four electrodes, likely to coalesce when two growth fronts meet, resulting in a complicated evo- two serve as counter and reference electrode, using pencil graphite and a Ag/AgCl lution. To obtain insights into this process, we perform another phase-field micro reference electrode (Harvard Apparatus), respectively. The remaining two simulation. The phase-field model for the grain evolution and the set of parameters electrodes were connected to the graphene supporting layer to monitor the con- is almost identical to that reported in the literature for a similar system64. We ductance signal of graphene and the electrocatalytic signal of MoS2 during HER. ignore the evolution of the supersaturation and keep it at a constant of 0.1 The device fabrication procedure is shown in Supplementary Figs. 21 and 22, and everywhere. A detailed formulation is omitted here for brevity and can be found in Supplementary Notes 2 and 3. In all experiments, only the exposed region of the other work65,66. In addition to the grains, we pay attention to the evolutions of the MoS2 nanosheets contributes to the recorded HER performance. Au QDs. These QDs are modeled as hemispherical balls sliding on the substrate. The typical four-electrode micro-electrochemical measurements were For simplicity, we ignore the effect of the Au QDs on the growth of MoS2 grains. conducted in a 0.5 M H2SO4 electrolyte solution. The scan rate is 5 mV step−1. Then these QDs move on the substrate at a speed of v = 50 tan−1(t)/π, where t is Both electrocatalytic current (Ic) and conductance current (Ids) are simultaneously the dimensionless time of evolution. When two QDs meet, they merge and form a detected. The leakage current in Ic is about 10−10 A (the electrochemical signal larger QD. The volume of the new QD is the sum of the volumes of the previous without any exposed window, i.e., from a region with PMMA passivation). The two QDs and its center is at the midpoint between the centers of the previous two electrochemical current density is calculated by normalizing the current to the area QDs. We then solve the evolution equation in Matlab by discretization with a time of the exposed window on the MoS2. In 0.5 M H2SO4 solution, we measured: step of 0.01. Initially Au QDs are randomly placed on the substrate and there is in ERHE ¼ EAg=AgCl þ 0:219 V ð5Þ the vicinity of these QDs, a circular nucleus of MoS2 with a radius of r0 and a randomly assigned grain orientation. To make the simulation consistent with the experiment, we extract some Standard macro-electrochemical measurements. The macro-electrochemical parameters from the experimental data. The experiment shows that the QDs are measurements were conducted on a glassy carbon electrode (3 mm in diameter). typically 4–5 nm. It also shows that when observed on top of the MoS2 sheet, the Using a similar PMMA-assisted transfer method as described above, we transferred Au QDs are larger and less dense by around 2.5 times. This suggests by pushing Au CVD-grown graphene and MoS2 nanograin films layer-by-layer on a glassy carbon NATURE COMMUNICATIONS | (2020)11:57 | https://doi.org/10.1038/s41467-019-13631-2 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13631-2 electrode, and the sample area exceeding to glassy carbon electrode was scraped off 5. Son, D.-Y. et al. Self-formed grain boundary healing layer for highly efficient before measuments. A standard three electrode system was used, and a Pt plate and CH3 NH3 PbI3 perovskite solar cells. Nat. Energy 1, 16081 (2016). Ag/AgCl rod served as counter and reference electrode, respectively. The mea- 6. Li, C. et al. Grain-boundary-enhanced carrier collection in CdTe solar cells. surement was performed on a biological electrochemical station in H2-saturated Phys. Rev. Lett. 112, 156103 (2014). 0.5 M H2SO4 solution. Linear sweep voltammetry (LSV) was conducted at a scan 7. Bowman, W. J., Zhu, J., Sharma, R. & Crozier, P. A. Electrical conductivity and rate of 5 mV s−1. The onset potential is defined as the beginning potential of Tafel grain boundary composition of Gd-doped and Gd/Pr co-doped ceria. Solid linear region. The stability test was carried out by taking continuous potential State Ion. 272, 9–17 (2015). cycling in the potential window of −0.181 to ~0.219 V (versus RHE) at a scan rate 8. Guo, C. F., Sun, T., Liu, Q., Suo, Z. & Ren, Z. Highly stretchable and of 100 mV s−1. The presented LSV curves in macro-electrochemical cell were iR- transparent nanomesh electrodes made by grain boundary lithography. Nat. corrected by subtracting the ohmic resistance loss (about 9 Ω), the value of which is Commun. 5, 3121 (2014). obtained from electrochemical impedance spectroscopy measurement (Supple- 9. Sangwan, V. K. et al. Gate-tunable memristive phenomena mediated by grain mentary Fig. 32). To evaluate the electrochemically active surface area of the cat- boundaries in single-layer MoS2. Nat. Nanotechnol. 10, 403 (2015). alysts, the Cu underpotential deposition method67–70 was applied, as shown in 10. Feng, X., Jiang, K., Fan, S. & Kanan, M. W. Grain-boundary-dependent CO2 Supplementary Fig. 30 and Supplementary Table 3. electroreduction activity. J. Am. Chem. Soc. 137, 4606–4609 (2015). To test the faradaic efficiency, the H2 products were analyzed with an online gas 11. Li, G. et al. All the catalytic active sites of MoS2 for hydrogen evolution. J. Am. chromatography (GC) setup (Agilent 7890B) equipped with a thermal conductivity Chem. Soc. 138, 16632–16638 (2016). detector. Argon (≥99.999%) was used as the carrier gas. The Faradaic efficiency 12. Seh, Z. W. et al. Combining theory and experiment in electrocatalysis: insights (FE) was calculated by comparing the measured amount of H2 generated by into materials design. Science 355, eaad4998 (2017). cathodal electrolysis with the calculated amount of H2 (assuming an FE of 100%), 13. Voiry, D., Yang, J. & Chhowalla, M. Recent strategies for improving the and the equation is given by: catalytic activity of 2D TMD nanosheets toward the hydrogen evolution 96485 ´ 2 ´ mol H2 ðGCÞ ´ 100 reaction. Adv. Mater. 28, 6197–6206 (2016). FEð%Þ ¼ ð6Þ Q 14. Liu, Y. et al. Self-optimizing, highly surface-active layered metal where the moles of H2 is measured by GC (the calibration is needed in advance dichalcogenide catalysts for hydrogen evolution. Nat. Energy 2, 17127 (2017). through injecting high-purity H2 to GC, and the H2 volume is linearly dependent 15. Li, H. et al. Activating and optimizing MoS2 basal planes for hydrogen on GC peak area), Q is obtained from the electrochemical measurements. evolution through the formation of strained sulphur vacancies. Nat. Mater. 15, 48–53 (2016). 16. Kibsgaard, J., Chen, Z., Reinecke, B. N. & Jaramillo, T. F. Engineering the Material characterization. The microstructures and morphologies based on MoS2 surface structure of MoS2 to preferentially expose active edge sites for were characterized by optical microscopy, SEM (FEI 4200), and Raman spectro- electrocatalysis. Nat. Mater. 11, 963 (2012). scopy (Renishaw inVia). Raman spectroscopy was performed with a 514.5 nm laser 17. Ji, Q., Zhang, Y., Zhang, Y. & Liu, Z. Chemical vapour deposition of group- (with spot size about 1 μm in diameter) at room temperature. The micro- VIB metal dichalcogenide monolayers: engineered substrates from amorphous electrochemical measurements were performed using two source meters (Keithley to single crystalline. Chem. Soc. Rev. 44, 2587–2602 (2015). 2400 and 2450). STEM measurements were performed at the SuperSTEM 18. Jaramillo, T. F. et al. Identification of active edge sites for electrochemical H2 Laboratory, Daresbury, UK, on a Nion UltraSTEM100 aberration-corrected dedi- evolution from MoS2 nanocatalysts. Science 317, 100–102 (2007). cated STEM. The Nion UltraSTEM has a cold field emission gun with a native 19. Xie, J. et al. Defect-rich MoS2 ultrathin nanosheets with additional active edge energy spread of 0.35 eV and was operated at 60 kV acceleration voltage. The beam sites for enhanced electrocatalytic hydrogen evolution. Adv. Mater. 25, was set up to a convergence semiangle of 30 mrad with an estimated beam current 5807–5813 (2013). of ~100 pA. Under these operating conditions, the estimated probe size is ~1.1 Å, providing the perfect tool for atom-by-atom chemical analysis. HAADF imaging 20. Dai, X. et al. Co-doped MoS2 nanosheets with the dominant CoMoS phase was employed to produce atomically resolved images whose intensity is approxi- coated on carbon as an excellent electrocatalyst for hydrogen evolution. ACS mately proportional to the square of the average atomic number Z of the material Appl. Mater. Interfaces 7, 27242–27253 (2015). under investigation. This chemically sensitive “Z-contrast” mode is ideally suited to 21. Voiry, D. et al. The role of electronic coupling between substrate and 2D MoS2 directly identify the nature of individual atoms. All the HAADF images presented nanosheets in electrocatalytic production of hydrogen. Nat. Mater. 15, here were filtered for visual clarity using a wiener filter with coefficient sigma 1003–1009 (2016). ranging from 2–5 and 50 iterations in MEM (Maximum Entropy Methods) with 22. Ye, G. et al. Defects engineered monolayer MoS2 for improved hydrogen fourth order Gaussian filter, as implemented in the STEM-CELL software71,72. evolution reaction. Nano Lett. 16, 1097–1103 (2016). Geometrical Phase Analysis (GPA) was carried out using a cosine mask of suitable 23. Ouyang, Y. et al. Activating inert basal planes of MoS2 for hydrogen evolution size and 1 binning base. HRTEM and low-resolution annular dark field (ADF)- reaction through the formation of different intrinsic defects. Chem. Mater. 28, STEM images were obtained on a FEI Tecnai F20 field emission gun microscope 4390–4396 (2016). with a 0.19 nm point-to-point resolution at 200 kV, equipped with an embedded 24. Li, S. et al. Vapour–liquid–solid growth of monolayer MoS2 nanoribbons. Nat. Quantum Gatan Image Filter for EELS analyses. Some STEM characterizations Mater. 17, 535–542 (2018). were carried out on a JEOL ARM-200F (S)TEM equipped with CEOS CESCOR 25. Zhou, J. et al. A library of atomically thin metal chalcogenides. Nature 556, aberration corrector, operated at an accelerating voltage of 80 kV. The convergence 355–359 (2018). semiangle and acquisition semiangle were 28–33 and 68–280 mrad for the ADF 26. Gong, Y. et al. Synthesis of millimeter-scale transition metal dichalcogenides imaging. The dwell time per pixel was set to 12–20. single crystals. Adv. Funct. Mater. 26, 2009–2015 (2016). 27. He, Y. et al. Layer engineering of 2D semiconductor junctions. Adv. Mater. 28, 5126–5132 (2016). Data availability 28. Wang, X. et al. Chemical vapor deposition growth of crystalline monolayer The data that support the findings of this study are available from the corresponding MoSe2. ACS Nano 8, 5125–5131 (2014). author upon reasonable request. 29. Gao, Y. et al. Ultrafast growth of high-quality monolayer WSe2 on Au. Adv. Mater. 29, 1700990 (2017). Received: 29 July 2019; Accepted: 1 November 2019; 30. Li, S. et al. Halide-assisted atmospheric pressure growth of large WSe2 and WS2 monolayer crystals. Appl. Mater. Today 1, 60–66 (2015). 31. Dathbun, A. et al. Large-area CVD-grown sub-2 V ReS2 transistors and logic gates. Nano Lett. 17, 2999–3005 (2017). 32. Zhou, J. et al. Large-area and high-quality 2D transition metal telluride. Adv. Mater. 29, 1603471 (2017). 33. Naylor, C. H. et al. Monolayer single-crystal 1T’-MoTe2 grown by chemical References vapor deposition exhibits weak antilocalization effect. Nano Lett. 16, 1. Yu, Q. et al. Control and characterization of individual grains and grain 4297–4304 (2016). boundaries in graphene grown by chemical vapour deposition. Nat. Mater. 10, 34. Cui, F. et al. Epitaxial growth of large-area and highly crystalline anisotropic 443 (2011). ReSe2 atomic layer. Nano Res. 10, 2732–2742 (2017). 2. Najmaei, S. et al. Vapour phase growth and grain boundary structure of 35. Gao, Y. et al. Large-area synthesis of high-quality and uniform monolayer molybdenum disulphide atomic layers. Nat. Mater. 12, 754–759 (2013). WS2 on reusable Au foils. Nat. Commun. 6, 8569 (2015). 3. van der Zande, A. M. et al. Grains and grain boundaries in highly crystalline 36. Empante, T. A. et al. Chemical vapor deposition growth of few-layer MoTe2 in monolayer molybdenum disulphide. Nat. Mater. 12, 554–561 (2013). the 2H, 1T’, and 1T phases: tunable properties of MoTe2 films. ACS Nano 11, 4. Chang, S.-Y. & Chang, T.-K. Grain size effect on nanomechanical properties 900–905 (2017). and deformation behavior of copper under nanoindentation test. J. Appl. Phys. 37. Feng, Q. et al. Growth of large-area 2D MoS2(1-x)Se2x semiconductor alloys. 101, 033507 (2007). Adv. Mater. 26, 2648–2653 (2014). 10 NATURE COMMUNICATIONS | (2020)11:57 | https://doi.org/10.1038/s41467-019-13631-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13631-2 ARTICLE 38. Wu, S. et al. Vapor–solid growth of high optical quality MoS2 monolayers with 70. Colmenares, L., Jusys, Z. & Behm, R. J. Electrochemical surface near-unity valley polarization. ACS Nano 7, 2768–2772 (2013). characterization and O2 reduction kinetics of Se surface-modified 39. Huang, J.-H. et al. Large-area 2D layered MoTe2 by physical vapor deposition Ru nanoparticle-based RuSey/C catalysts. Langmuir 22, 10437–10445 and solid-phase crystallization in a tellurium-free atmosphere. Adv. Mater. (2006). Interfaces 4, 1700157 (2017). 71. Grillo, V. & Rotunno, E. STEM-CELL: A software tool for electron 40. Kang, K. et al. High-mobility three-atom-thick semiconducting films with microscopy: Part I—simulations. Ultramicroscopy 125, 97–111 (2013). wafer-scale homogeneity. Nature 520, 656 (2015). 72. Grillo, V. & Rossi, F. STEM-CELL: A software tool for electron microscopy. 41. Kang, K. et al. Layer-by-layer assembly of two-dimensional materials into Part 2 analysis of crystalline materials. Ultramicroscopy 125, 112–129 wafer-scale heterostructures. Nature 550, 229 (2017). (2013). 42. Lee, S.-H., Kwak, E.-H. & Jeong, G.-H. Dewetting behavior of electron-beam- 73. Zhan, Y., Liu, Z., Najmaei, S., Ajayan, P. M. & Lou, J. Large-area vapor-phase deposited Au thin films on various substrates: graphenes, quartz, and SiO2 growth and characterization of MoS2 atomic layers on a SiO2 substrate. Small wafers. Appl. Phys. A 118, 389–396 (2015). 8, 966–971 (2012). 43. Seguini, G. et al. Solid-state dewetting of ultra-thin Au films on SiO2 and 74. Kong, D. et al. Synthesis of MoS2 and MoSe2 films with vertically aligned HfO2. Nanotechnology 25, 495603 (2014). layers. Nano Lett. 13, 1341–1347 (2013). 44. Li, H. et al. From bulk to monolayer MoS2: evolution of Raman scattering. 75. Lin, Y.-C. et al. Direct synthesis of van der Waals solids. ACS Nano 8, Adv. Funct. Mater. 22, 1385–1390 (2012). 3715–3723 (2014). 45. Huang, P. Y. et al. Grains and grain boundaries in single-layer graphene 76. Song, J.-G. et al. Layer-controlled, wafer-scale, and conformal synthesis of atomic patchwork quilts. Nature 469, 389 (2011). tungsten disulfide nanosheets using atomic layer deposition. ACS Nano 7, 46. Shi, J., Ji, Q., Liu, Z. & Zhang, Y. Recent advances in controlling syntheses and 11333–11340 (2013). energy related applications of MX2 and MX2/graphene heterostructures. Adv. 77. Yim, C. et al. High-performance hybrid electronic devices from layered PtSe2 Energy Mater. 6, 1600459 (2016). films grown at low temperature. ACS Nano 10, 9550–9558 (2016). 47. Shi, Y., Li, H. & Li, L.-J. Recent advances in controlled synthesis of two- 78. Liu, K.-K. et al. Growth of large-area and highly crystalline MoS2 thin layers dimensional transition metal dichalcogenides via vapour deposition on insulating substrates. Nano Lett. 12, 1538–1544 (2012). techniques. Chem. Soc. Rev. 44, 2744–2756 (2015). 48. Ruffino, F. & Grimaldi, M. G. Controlled dewetting as fabrication and patterning strategy for metal nanostructures. Phys. Status Solidi A 212, 1662–1684 (2015). Acknowledgements Z.L. gratefully acknowledges funding supports from Ministry of Education (MOE) under 49. Choi, H.-J. Vapor–Liquid–Solid Growth of Semiconductor Nanowires (Springer AcRF Tier 1 (M4011782.070 RG4/17 and M4011993.070 RG7/18), AcRF Tier 2 (2015- Berlin Heidelberg, 2012). T2-2-007, 2016-T2-1-131, 2016-T2-2-153, and 2017-T2-2-136), AcRF Tier 3 (2018-T3- 50. Kresse, G. & Hafner, J. Ab initio molecular-dynamics simulation of the liquid- 1-002), National Research Foundation – Competitive Research Program (NRF-CRP21- metal-amorphous-semiconductor transition in germanium. Phys. Rev. B 49, 2018-0092), and A*Star QTE programme. P.T., S.M.S., J.R.M., and J.A. acknowledge 14251–14269 (1994). funding from Generalitat de Catalunya 2017 SGR 327 and 1246 and the Spanish 51. Lauritsen, J. V. et al. Size-dependent structure of MoS2 nanocrystals. Nat. MINECO coordinated projects between IREC and ICN2 VALPEC (ENE2017-85087-C2- Nanotechnol. 2, 53–58 (2007). C3). ICN2 acknowledges support from the Severo Ochoa Program (MINECO, Grant 52. Nørskov, J. K. et al. Trends in the exchange current for hydrogen evolution. J. SEV-2017-0706). IREC and ICN2 are funded by the CERCA Programme/Generalitat de Electrochem. Soc. 152, J23–J26 (2005). Catalunya. Part of the present work has been performed in the framework of Universitat 53. Zhou, W. et al. Intrinsic structural defects in monolayer molybdenum Autònoma de Barcelona Materials Science PhD program. S.M.S. acknowledges funding disulfide. Nano Lett. 13, 2615–2622 (2013). from ‘Programa Internacional de Becas “la Caixa”-Severo Ochoa’. J.R.M. recognizes also 54. Hansen, L. P. et al. Atomic-scale edge structures on industrial-style MoS2 its affiliation to University of Barcelona. Part of the electron microscopy aspects of this nanocatalysts. Angew. Chem. Int. Ed. 50, 10153–10156 (2011). work were supported by the EPSRC (UK), as the SuperSTEM Laboratory is the EPSRC 55. Yu, W. J. et al. Vertically stacked multi-heterostructures of layered materials National Research Facility for Advanced Electron Microscopy. Q.J.W. acknowledges the for logic transistors and complementary inverters. Nat. Mater. 12, 246–252 supports from MOE, Singapore grant (MOE2016-T2-2-159, MOE2016-T2-1-128, and (2013). MOE Tier 1 RG164/15) and National Research Foundation, Competitive Research 56. Allain, A., Kang, J., Banerjee, K. & Kis, A. Electrical contacts to two- Program (NRF-CRP18-2017-02) and NSFC (61704082) as well as Natural Science dimensional semiconductors. Nat. Mater. 14, 1195 (2015). Foundation of Jiangsu Province (BK20170851). X.W. and B.K.T. gratefully acknowledge 57. Kang, J., Liu, W., Sarkar, D., Jena, D. & Banerjee, K. Computational study of funding support from MOE, Singapore (grant no. MOE2015-T2-2-043). H.Z. acknowl- metal contacts to monolayer transition-metal dichalcogenide semiconductors. edges the supports from MOE under AcRF Tier 2 (MOE2015-T2-2-057; MOE2016-T2- Phys. Rev. X 4, 031005 (2014). 2-103; and MOE2017-T2-1-162), AcRF Tier 1 (2016-T1-002-051; 2017-T1-001-150; and 58. Wang, J. et al. Field effect enhanced hydrogen evolution reaction of MoS2 2017-T1-002-119), Agency for Science, Technology and Research (A*STAR) under its nanosheets. Adv. Mater. 29, 1604464 (2017). AME IRG (Project No. A1783c0009), and NTU under Start-Up Grant 59. Ding, M. et al. An on-chip electrical transport spectroscopy approach for (M4081296.070.500000) in Singapore. The authors would like to acknowledge the in situ monitoring electrochemical interfaces. Nat. Commun. 6, 7867 (2015). Facility for Analysis, Characterization, Testing, and Simulation, Nanyang Technological 60. He, Y. et al. Self-gating in semiconductor electrocatalysis. Nat. Mater. 18, University, Singapore, for their electron microscopy and X-ray facilities. H.Z. also thanks 1098–1104 (2019). the support from ITC via Hong Kong Branch of National Precious Metals Material 61. Voiry, D. et al. Conducting MoS2 nanosheets as catalysts for hydrogen Engineering Research Center, and the Start-Up Grant from City University of Hong evolution reaction. Nano Lett. 13, 6222–6227 (2013). Kong. Theory and simulations work at Rice university (Z.H., L.W., and B.I.Y.) was 62. Nsimama, P. D., Herz, A., Wang, D. & Schaaf, P. Influence of the substrate on supported by the Office of Naval Research grant N00014-18-1-2182. the morphological evolution of gold thin films during solid-state dewetting. Appl. Surf. Sci. 388A, 475–482 (2016). 63. Borciaa, R., B., I. D. & Bestehorn, M. Static and dynamic contact angles- a phase field modelling. Eur. Phys. J. Spec. Top. 166, 127–131 (2009). Author contributions 64. Artyukhov, V. I., Hu, Z., Zhang, Z. & Yakobson, B. I. Topochemistry of Z.L. and Y.H. conceived and initiated the project. Z.L., Q.W., J.A. and M.W. supervised Bowtie- and star-shaped metal dichalcogenide nanoisland formation. Nano the project and led the collaboration efforts. Y.H. designed the experiments, grew Lett. 16, 3696–3702 (2016). nanograin film, fabricated the electrochemical devices, and performed micro- 65. Steinbach, I. & Pezzolla, F. A generalized field method for multiphase electrochemical HER measurement. P.T., C.Z. (Dr. Chao Zhu), S.M.S., J.A., J.M. and transformations using interface fields. Phys. D Nonlinear Phenom. 134, Q.M.R. performed the TEM and STEM measurements. Z.Hu did the phase-field simu- 385–393 (1999). lation of the growth process. Y.H., Q.H. and C.Z. (Mr. Chao Zhu) made the micro- 66. Ingo, S. Phase-field models in materials science. Modell. Simul. Mater. Sci. electrochemical measurement setup. L.W. (Luqing Wang) and B.Y. performed the first- Eng. 17, 073001 (2009). principles calculations of the catalytic activity. Q.Z. did the SEM measurement. G.G. and 67. Zheng, Y. et al. High electrocatalytic hydrogen evolution activity of A.T. conducted the cross-sectional TEM measurement. Z. Huang and Q.H. performed an anomalous ruthenium catalyst. J. Am. Chem. Soc. 138, 16174–16181 gas chromatography measurement. P.G., X.W. (Xuewen Wang), W.F., C.G., X.W. (Xingli (2016). Wang), B.A., Y.Z., L.W. (Liang Wang), and B.T. assisted the growth of nanograin film. 68. Mahmood, J. et al. An efficient and pH-universal ruthenium-based catalyst for J.X. and A.Z. did the Raman measurements. Y.H., P.T., Z.Hu, Q.W., H.Z., J.A. and Z.L. the hydrogen evolution reaction. Nat. Nanotechnol. 12, 441 (2017). wrote the paper. All authors discussed the results and commented on the manuscript. 69. Green, C. L. & Kucernak, A. Determination of the platinum and ruthenium surface areas in platinum−ruthenium alloy electrocatalysts by underpotential deposition of copper. I. unsupported catalysts. J. Phys. Chem. B 106, Competing interests 1036–1047 (2002). The authors declare no competing interests. NATURE COMMUNICATIONS | (2020)11:57 | https://doi.org/10.1038/s41467-019-13631-2 | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13631-2 Additional information Open Access This article is licensed under a Creative Commons Supplementary information is available for this paper at https://doi.org/10.1038/s41467- Attribution 4.0 International License, which permits use, sharing, 019-13631-2. adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Correspondence and requests for materials should be addressed to M.W., Q.J.W., J.A. or Commons license, and indicate if changes were made. The images or other third party Z.L. material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the Peer review information Nature Communications thanks Jing Gu for their contribution article’s Creative Commons license and your intended use is not permitted by statutory to the peer review of this work. Peer review reports are available. regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ Reprints and permission information is available at http://www.nature.com/reprints licenses/by/4.0/. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. © The Author(s) 2020 12 NATURE COMMUNICATIONS | (2020)11:57 | https://doi.org/10.1038/s41467-019-13631-2 | www.nature.com/naturecommunications