New Approach to Unveiling Individual Atomic ... - ACS Publications

Feb 15, 2018 - SIMS surface mapping because of its extreme surface sensitivity while the individual atomic layers of MLG could be precisely revealed f...
0 downloads 4 Views 13MB Size
Article Cite This: Chem. Mater. 2018, 30, 1718−1728

pubs.acs.org/cm

New Approach to Unveiling Individual Atomic Layers of 2D Materials and Their Heterostructures Irfan Haider Abidi,‡ Lu-Tao Weng,*,‡,† Chi Pui Jeremy Wong,‡ Abhishek Tyagi,‡ Lin Gan,§ Yao Ding,‡ Man Li,§ Zhaoli Gao,# Ruiwen Xue,‡ Md Delowar Hossain,‡ Minghao Zhuang,‡ Xuewu Ou,‡ and Zhengtang Luo*,‡ ‡

Department of Chemical and Biological Engineering, †Materials Characterization and Preparation Facility, The Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong § State Key Laboratory of Material Processing and Die and Mould Technology, School of Materials Science and Engineering, Huazhong University of Science and Technology, Wuhan, 430074, China # Department of Physics and Astronomy, University of Pennsylvania, 209S 33rd Street, Philadelphia, Pennsylvania 19104 6396, United States S Supporting Information *

ABSTRACT: Visualization of the chemical structures of two-dimensional (2D) materials and their interfaces at the virtually atomic scale is an imperative step toward devising highly efficient ultrathin optoelectronic devices. Herein, we demonstrate a universal method featuring time-of-flight secondary ion mass spectrometry (ToF-SIMS), coupled with the structure simplicity of 2D materials, as a versatile tool to reveal the vertical atomic layers of various two-dimensional (2D) materials including graphene, hexagonal boron nitride (h-BN), and transition metal dichalcogenides (TMDs). We demonstrated that the vertical atomic layers of those 2D materials can be unveiled layer-by-layer using a strategy of ToF-SIMS three-dimensional (3D) analysis developed in this work. Moreover, we found that the extreme surface sensitivity and chemical specificity of ToF-SIMS also enables the examination of the lateral uniformity of 2D materials. During this process, we first removed interference of adsorbed organic contamination by annealing, which allows the high quality signals specific to the 2D materials which were accumulated along the sputtering depth. The accumulated signal was found to be linearly proportional to the number of atomic layers in the z-direction, thus providing a chemical intensity contrast that can be directly used to reveal the number of atomic layers in the vertical direction. For the case of CVD-grown graphene, up to six individual adlayers have been resolved. The technique developed in this work is substrate-independent and can be directly applied to the grown substrate, which circumvents the time-consuming transfer process and avoids any potential hazard to the delicate 2D materials. Our approach provides an efficient characterization tool for analyzing the 2D materials and their heterostructures, with extensive implications toward preparation of high-quality 2D crystals for atomically thin optoelectronic devices.



INTRODUCTION The present decade is an era of two-dimensional (2D) layered materials, and starting from the birth of graphene it evolved into the addition of other members to the family of 2D crystals such as hexagonal boron nitride (h-BN) and a variety of transition metal dichalcogenides (TMDs, including MoS2, WS2, and so on).1−4 These atomically thin materials offer exceptional optical and electronic properties owing to their structural build, which boost their potential for a wide spectrum of applications.5−9 Recently, scientists came up with another exciting approach of stacking multiple atomic layers of different 2D crystals, either in-plane or vertical configuration, to build © 2018 American Chemical Society

heterostructures at atomic level precisions, which exhibit inimitable and synergistic properties.10−12 This interesting concept brings an aspiration of building an artificial “lego” of high quality 2D crystals with desired sequences, to accomplish the ultimate goal of large area high-speed ultrathin electronic, photonic, and optoelectronic devices.13−16 Nevertheless, ensuring the quality of 2D crystals in terms of layer uniformity and surface contaminations is the key step to fabricate such Received: December 28, 2017 Revised: February 15, 2018 Published: February 15, 2018 1718

DOI: 10.1021/acs.chemmater.7b05371 Chem. Mater. 2018, 30, 1718−1728

Article

Chemistry of Materials

nonidentical individual atomic layers of a variety of 2D materials both laterally and vertically using purpose-built samples grown by the CVD technique. Our study was first focused on CVD-grown graphene with a particular emphasis on the evaluation of monolayer uniformity and the determination of layer number in multilayer graphene (MLG) domains. As graphene signal in ToF-SIMS was seriously affected by the organic contaminations adsorbed on graphene surface, we developed an in situ thermal annealing protocol to remove these contaminations. Our results showed that the uniformity of monolayer graphene could be easily evaluated with ToFSIMS surface mapping because of its extreme surface sensitivity while the individual atomic layers of MLG could be precisely revealed from a new approach proposed in this study. Although the depth resolution of ToF-SIMS is not enough to resolve the atomic layers in MLG, the graphene signal (C2−) accumulated from the depth profile was instead found to be proportional to the number of graphene layers in MLG and, therefore, allows us to interrogate atomic layers of other 2D materials such as hBN and WS2 or even the buried interfaces of vertical heterostructures of monolayer graphene and MoS2 single crystals. Our approach is independent of the substrate and thus avoids the need for transferring to another substrate for characterization. Integrating our approach with the fundamental research of the 2D materials can boost the ongoing progress toward achieving the superior quality 2D crystals for the fabrication of next generation high-performance nanodevices.

high performance nanodevices based on single or multiple layers of 2D materials.17 Among all the synthesis methods, the chemical vapor deposition (CVD) technique is a widely adopted and scalable method to grow a variety of 2D materials and their heterostructures.18−21 Although enormous research has already been done on CVD growth of graphene,22,23 h-BN,24,25 and most of the TMDs materials,26−28 more efforts are required to efficiently characterize the grown 2D crystals and obtain a deep understanding about their growth mechanism. Thus, it is desirable to have an efficient tool to probe the quality of 2D materials and their heterostructures especially in terms of chemical information at virtually the atomic level, preferably directly on the growth substrate. Conventional characterization techniques, however, possess certain difficulties to provide simultaneously morphological and chemical information. For instance, it is very challenging to reveal adlayers of graphene or h-BN films directly on the Cu substrate, by traditional characterization techniques such as atomic force microscopy (AFM) and Raman spectroscopy due to the surface roughness and strong fluorescence effect of the metal substrate.29,30 Although SEM offers limited success, it provides a poor contrast for the inplane heterostructures and lack of information for the stacking order of the vertical heterostructures of the 2D materials. Very often, the 2D materials have very poor contrast in SEM because the secondary electrons coming from substrates dominate the total signal. Apparently all conventional techniques have a common limitationthe measurements are “substrate-dependent”. This is also the reason why the 2D materials often need to be transferred to another substrate (300 nm SiO2/Si), which adversely complicates the characterization results.29 Therefore, it is imperative to develop a new technique that is able to unveil the homo- or heterostructures built by atomic layers of 2D crystals grown by the CVD method or even stacked manually. It is even more desirable if the technique can be directly applied to the grown substrates. Recently, time-of-flight secondary ion mass spectrometry (ToF-SIMS) has emerged as an alternative tool for characterizing atomically thin 2D materials.30−34 For instance, a few studies employed ToF-SIMS to analyze polymer35,36 and metallic contaminations37 on the transferred graphene surface, while the other work investigated adlayers of graphene assisted with isotope-labeling technique.38 Nonetheless, limited success has been obtained for the use of ToF-SIMS for resolving atomic layers of 2D heterostructures.30 From an analytical point of view, ToF-SIMS bears the following unique capabilities for 2D materials characterization: (1) when operated in static mode, its extreme surface sensitivity ( 2,39 as shown in the inset of Figure 1a. An almost negligible D peak at 1350 cm−1 is seen, evidencing low density of structural defects of the samples.39 The signal from organic contaminants adsorbed on the graphene surface is weak, attesting the effectiveness of the transfer process. Figure 1b shows the ToFSIMS chemical map of the area shown in Figure 1a, constructed by overlaying the map of C2− and that of the ions representing Cu (including Cu−, O2−, CuO−, etc.). The right panel (i) and (ii) illustrates the distribution of C2− and the Cu− ions, individually. The chemical map of C2− clearly replicates the graphene domains present in the optical image, indicating C2− arising exclusively from the graphene domains in this case. Therefore, probing the C2− complemented the uniformity of the graphene film grown on the Cu foil.37 Here, the main advantage of ToF-SIMS is its effectiveness of obtaining an overall image of the surface with high lateral resolution (∼200 nm), as opposed to Raman mapping, which take tens of hours to obtain similar images, along with relatively low spatial resolutions. It must be pointed out that the chemical map shown in Figure 1b was acquired after performing in situ thermal annealing of graphene/Cu samples at 450 °C under ultrahigh vacuum conditions for 1 h, to get rid of all the organic contaminations adsorbed on the graphene surface from the environment.36 Without the in situ thermal annealing step, organic compounds also contribute to the C2− ions with a much higher sensitivity and adversely reduce the resolution, 1719

DOI: 10.1021/acs.chemmater.7b05371 Chem. Mater. 2018, 30, 1718−1728

Article

Chemistry of Materials

Besides providing information about the graphene uniformity, we also attempted to unveil the buried adlayers of MLG stack by ToF-SIMS three-dimensional (3D) analysis. The best depth resolution of ToF-SIMS, even using a very low energy sputter source (e.g., 500 V), is around 1 nm, which is obviously bigger than the atomic layer thickness (0.34 nm) of graphene. Therefore, it is difficult to deduce the layer number of graphene in MLG directly from the depth profile. In practice, the numbers of layers and their thicknesses are indirectly estimated from the simulation of depth profile intensity using models (see below and ref 34). In this work, we propose a new approach to tackle this problem. Our approach was based on the ToF-SIMS 3D analysis with a rationale that the graphene signal accumulated from the 3D volume should be proportional to the number of graphene layers. This approach is illustrated in Figure 2 for MLG grown on the Cu substrate. The MLG samples were thermally annealed at 450 °C for 1 h under high vacuum (10−9 Torr), before the 3D analysis. This annealing step enhances the lateral resolution of the surface mapping for graphene domains as shown in Figure S2. Thereafter, the MLG film on Cu was subjected to ToF-SIMS 3D analysis, which consisted of a series of cycles of high lateral resolution imaging followed by very slow sputtering to remove the graphene materials. The imaging was realized using 25-keV Bi3+ primary ion beam while the sputtering was using 3 keV Cs+ sputter ion beam. As an illustration, we reconstructed the chemical map of C2− after accumulating the signals for 20, 100, and 300 s of sputtering as shown in Figure 2a−c, which corresponds to the signal accumulated from different depths. With 20 s of sputtering time (Figure 2a), we barely observed a uniform graphene layer, while for prolonging sputtering for 100 s, another “david-star” shape started to emerge underneath as shown in the Figure 2b. Further, we found that it takes approximately 100 s of Cs+ sputtering for the removal of the first layer of graphene at our present conditions. We further confirmed this by acquiring the C2− map from the same area after 100 s of sputtering, which shows minimal signals in the specified region, as illustrated in Figure S3. Prolonging the Cs+ sputtering for longer time exposes all the buried adlayers underneath the top layer, such that we found a total of three layers of graphene after accumulating C2− signals for 300 s of sputtering, as shown in Figure 2c. A three-layer structure can be easily seen from the accumulated C2− intensity. If an area line scan is drawn across the three layers (red dotted line in Figure 2c), a stepwise increase of C2− intensity with the number of graphene layers is observed. More importantly, the intensity at each step is linearly proportional to the number of graphene layers (Figure 2d), confirming that the rationale of our approach is correct. This result is remarkable because it shows that the C2− signal depends only on the quantity of graphene but has nothing to do with the substrate. Implementing this approach enables us to successfully distinguish as many as six layers of as-grown graphene directly on the Cu substrate, as shown in Figure 2e. It is important to point out that our approach does not require any prior surface treatment of the Cu foil41,42 or isotope labeling technique38 to differentiate and visualize the individual adlayers of graphene, as reported previously. As a proof of concept, we also implemented our approach to the transferred MLG film on SiO2/Si substrate. Figure 2f displays an optical image of the MLG domain transferred to the SiO2/Si substrate. The optical contrast between the substrate and the graphene layers enables us to identify the number of

Figure 1. ToF-SIMS chemical imaging of single crystal graphene on Cu and SiO2. (a) Optical image of as-grown single crystal graphene domains on the Cu foil. Inset is a Raman spectrum acquired from the marked area, indicating single layer graphene (SLG). (b) Surface high lateral resolution ToF-SIMS image acquired from the same area shown in (a), constructed by chemical map overlayers of C2− and the Cu characteristic ions (accumulation of Cu−, O2−, CuO2−). Panel (i) and (ii) depict the individual chemical map of C2− and the Cu characteristic ions, respectively. (c) Optical image of single crystal graphene domains transferred on 300 nm SiO2/Si substrate. Raman spectrum obtained from the indicated area is shown in the inset. (d) ToF-SIMS image of the area (c), using chemical map overlayers of C2− and SiO2−, with individual panels as (iii) and (iv). The ToF-SIMS images shown in (b) and (d) are obtained after 1 h of in situ thermal annealing at 450 °C, to remove all the organic contaminations adsorbed on the graphene surface. The scale bar is 100 μm in panels (i) and (ii) and 200 μm in panels (iii) and (iv), and the y-axis scale is counts/pixel.

which smears edges of the graphene domains, as shown in the Figure S1 (Supporting Information I). We found the prethermal annealing is even more crucial to investigate the surface of the transferred graphene. Figure 1c shows an optical image of single-crystal graphene domains transferred to the SiO2/Si substrate using the polymer assisted wet transfer technique. Apparently, the optical micrograph along with the Raman spectrum reflect high quality SLG but do not provide information about physisorbed or chemisorbed contaminations on the graphene surface during the transfer process. However, ToF-SIMS chemical mapping of transferred graphene samples reveals the species like C2H3O2−, C2O−, and CH3O− ions, originated from the residues of polymer (PMMA) and acetone (solvent used for PMMA removal),36,40 respectively. In situ ToF-SIMS investigation reveals the gradual removal of these residues as indicated by the decline in the intensity of C2H3O2− C2O− and CH3O− signals during the thermal annealing treatment at 450 °C, as shown in Figure S1. Thermal annealing for about 1 h resulted in almost negligible amount of adsorbed organic residues, consistent with the recent report.35 Figure 1d shows a high lateral resolution ToF-SIMS surface image obtained by overlaying the chemical maps of C2− and SiO2− acquired from the area shown in Figure 1c. The chemical map shown in panel (iii) reflects C2− as a perfect marker for the hexagonal graphene domains, after the removal of organic contaminations via the aforementioned thermal annealing. Hence, the uniformity of graphene domains on Cu or SiO2/Si substrate can easily be probed by ToF-SIMS surface C2− mapping, assisted with in situ thermal annealing. 1720

DOI: 10.1021/acs.chemmater.7b05371 Chem. Mater. 2018, 30, 1718−1728

Article

Chemistry of Materials

Figure 2. Determining the layer numbers of multilayer graphene (MLG) stacks. (a) C2− map reconstructed by accumulating the signals for (a) 20 s, (b) 100 s, and (c) 300 s of Cs+ sputtering, showing the distribution of the first, second, and third layer of as-grown trilayer graphene on the Cu substrate, respectively. (d) Plot of the C2− intensity integrated from the area line scan shown in (c). A linear increment in the C2− signals is observed with the increasing number of graphene layers. (e) ToF-SIMS image (C2− map) of as-grown MLG/Cu, revealing more than 6 adlayers of graphene based on the C2− intensity difference. (f) Optical image of MLG transferred to the SiO2/Si substrate to identify the number of layers by optical contrast. (g) Corresponding ToF-SIMS (C2− map) image of MLG/SiO2, demonstrating equal number of distinguishable graphene adlayers (≥6 layers) analogous to the optical image (f). For (a), (b), (c), (e), and (g), the y-axis scale is counts/pixel.

respectively. Figure 3d illustrates the depth profile (intensity vs sputtering time) of the C2− and Cu− secondary ions reconstructed from the microareas of one, two and three layers of graphene on Cu. The intensity profiles clearly reflect the C2− ions distributed deeper for the three-layer graphene region as compared to that for one-layer graphene, as it requires longer sputtering time to expose the Cu surface (Cu−). However, as pointed out above, the number of graphene layers cannot be resolved from the depth profiles as the depth resolution of ToF-SIMS is not enough. In order to estimate the thickness of individual graphene layers from C2− intensity, we implemented a sputtering rate model proposed by Hofmann,43 to convert the sputtering time into depth (nm) (see Supporting Information II for details). A typical converted profile for trilayer graphene is shown in Figure 3e, indicating a good agreement between the C2− profiles simulated by the model with the experimental data points. Implementing this calculation method, the thickness of the outermost four layers of MLG is calculated and plotted against the number of layers (Figure 3f), together with the theoretical thickness of graphene. The results clearly demonstrate a nice agreement between the calculated thickness from the depth profiling and the theoretical thickness of graphene (see Supporting Information II for details). Such an indirect thickness determination of graphene adlayers also vindicated our approach for MLG identification through lateral highresolution C2− mapping above. Therefore, following few simple steps of in situ thermal annealing and ToF-SIMS depth profiling enable us to visualize and distinguish individual layers of as-grown MLG film directly on the Cu foil. The main advantage of the proposed strategy is that no prior surface treatment of the Cu foil or isotope labeling is required for layer determination, which otherwise makes the visualization process complicated and limited. Consequently, the presented approach

layers. Likewise, we also observed six individual layers of graphene in the C2− map presented in Figure 2g, which was acquired through depth profiling the same MLG region (shown in Figure 2f). Obviously, the ToF-SIMS chemical image is perceived as analogous to its optical image, which vindicates our strategy of identifying the layer number of graphene on Si/ SiO2 as well as on the Cu substrate. The layer number identification was also verified by acquiring Raman spectra of the individual adlayers of MLG, as shown in Figure S4. Consequently, as the ToF-SIMS chemical image is independent of the substrate, this technique provides an efficient and precise way to probe the number of layers of CVD graphene directly on the Cu substrate, without requiring a subsequent tedious transfer process. In order to further validate our approach, a more detailed analysis on the above ToF-SIMS 3D data was performed. Figure 3a presents the ToF-SIMS lateral map of accumulated C2− signals acquired from MLG/Cu, also discussed above as Figure 2e. To show the distribution of C2− ions in the perpendicular direction (z-axis), we took slices such as xz and yz in horizontal and vertical directions, respectively, as shown with the marked lines in Figure 3a. The cross-section map of the C2− distribution within the xz and yz slice is shown in Figure 3b. The shallow distribution of C2− for the xz slice represents the single layer of graphene, while for the yz slice indepth distribution of C2− was observed particularly in the central region, indicating the existence of the multilayer graphene domains in the center. These cross-section maps of C2− also correspond well to the lateral intensity map of C2− (Figure 3a), reflecting the exceptional capability of ToF-SIMS for in-depth analysis of the atomic layers. In addition, Figure 3c exhibits 3D overlay distribution of the secondary ions C2− and Cu− representing MLG layers stacked on the Cu substrate, 1721

DOI: 10.1021/acs.chemmater.7b05371 Chem. Mater. 2018, 30, 1718−1728

Article

Chemistry of Materials

Figure 3. Depth profiling and 3D imaging of multilayer graphene on Cu substrate. (a) ToF-SIMS image of as-grown MLG stack on the Cu substrate, constructed by sequential depth profiling of C2− ions. The horizontal and vertical lines indicate xz and yz slices for cross-section analysis. (b) Crosssection views of C2− (graphene) distribution in the z-axis obtained from mapping xz and yz slices. The yz cross section reveals relatively much deeper distribution of C2− ions, which corresponds to the multilayer graphene stacks. (c) 3D representation of C2− ions distribution on the Cu substrate denoted by Cu− ions. (d) Depth profiles reconstructed from the microarea of one, two, and three layers of graphene on Cu, represented by intensity of C2− and Cu− secondary ions as a function of sputtering time. The deeper interface of C2−/Cu− indicates a thick layer of graphene. (e) Normalized intensity C2− simulated profile against the thickness (nm) for a trilayer graphene, converted using the sputtering rate model. (f) A plot illustrating experimentally calculated thickness vs the number of graphene layers, which shows a great agreement with the theoretical thickness of graphene adlayers.

size with that in the optical image. From the inset of Figure 4b showing enlarged image of B+ mapping of the four domains from the marked region, we can observe the uniform distribution of B+ signals within the h-BN domains indicating isolated islands of single-crystal, predominately monolayer hBN. The B+ map accumulated from depth profiling, however, reveals that some h-BN domains are in fact multilayer, which are invisible under an optical microscope, as shown in Figure 4c. We believe this approach of direct visualization of multilayer h-BN stacks on the Cu foil is remarkable as it provides an efficient route to probe large area CVD-grown h-BN, which can guide us for optimizing growth conditions for uniform high quality h-BN films. We have also employed TMDC materials to evaluate their quality in terms of uniformity using our approach. Figure 4d displays an optical image revealing single-crystal CVD-grown WS2 domains stacked on SiO2/Si substrate. The Raman spectrum acquired from the marked area showed characteristics Raman bands of monolayer WS2 at 352 cm−1 (overlapping of second order mode 2LA(M) and in-plane E12g vibrational mode) and out-of-plane A1g mode at 417 cm−1,45 as shown in the inset of Figure 4d. The surface high lateral resolution chemical mapping was performed on the same area, specifying S− and O− secondary ions as representative of WS2 and SiO2 surfaces, respectively. Figure 4e depicts an overlay image of chemical maps of S− and O− secondary ions, clearly presenting the isolated WS2 domains on SiO2 substrate, consistent with the optical image (Figure 4d). Interestingly, some of the

is a milestone as an effective quality evaluation technique for CVD-driven graphene which can boost up the research in this field. 2D Materials beyond Graphene. We stretched our approach to other members of 2D materials family, such as h-BN and transition metal dichalcogenides (TMDCs, MoS2, WS2, etc.). Figure 4a illustrates an optical micrograph of as grown single-crystal h-BN on the Cu foil. Like the graphene visualization method, h-BN domains were visualized under an optical microscope by slightly oxidizing the Cu surface after the CVD growth. However, for the chemical identification of these domains, some other characterization tools are required. Unfortunately, Raman signals of h-BN on the growth substrate are extremely weak44 or sometimes invisible (Figure S5), and hence it is impractical to map a large area of as-grown h-BN domains using Raman spectroscopy. Therefore, it necessitates polymer-assisted transfer of h-BN flakes to SiO2/Si substrates for further probing the quality of domains. Nevertheless, todate revealing the multilayer h-BN domains directly on the Cu substrate is yet challenging due to the low contrast in SEM and extremely weak Raman signals. Next, we will demonstrate that ToF-SIMS high lateral resolution mapping and depth profiling, enabling us not only to reveal the large-area h-BN domains but also to distinguish the number of adlayers in multilayer h-BN domains. Figure 4b shows a large area surface high lateral resolution ToF-SIMS image acquired from the same area as in Figure 4a. We map the B+ secondary ions as the representative of the h-BN domains, showing consistent domain shape and 1722

DOI: 10.1021/acs.chemmater.7b05371 Chem. Mater. 2018, 30, 1718−1728

Article

Chemistry of Materials

Figure 4. ToF-SIMS analysis of other 2D materials beyond graphene. (a) Optical image of CVD grown single crystal h-BN domains on the Cu substrate. The Cu surface was slightly oxidized to visualize the h-BN flakes. (b) ToF-SIMS high lateral resolution surface map of B+ secondary ions representing isolated h-BN domains distribution, consistent with the optical image. Inset is the 50 μm × 50 μm B+ map of the marked region to show an enlarged image of h-BN domains. (c) B+ map accumulated from depth profiles, revealing the multilayers of h-BN domains. (d) Optical image of single-crystal WS2 domains stacked on the SiO2/Si substrate. Inset is the Raman spectrum acquired from the marked spot reflecting monolayer of WS2. (e) An overlayer image constructed by chemical maps of S− and O− secondary ions, presenting the distribution of WS2 crystals on SiO2 substrate. (f) S− ion map accumulated from depth profiling revealing multilayer domains of WS2, inset is an optical image which corresponds to the SIMS analysis area.

Figure 5. CVD grown lateral heterostructure of graphene and h-BN. (a) Optical image of as-grown graphene/h-BN lateral heterostructure on Cu substrate, obtained after aging the sample for one month in ambient atmosphere. (b) High-lateral resolution SIMS image of overlayer maps of C2− and BN−, representing graphene and h-BN, respectively. Panels showing the individual chemical maps of (i) C2− and (ii) BN-secondary ions. Panel (iii) is a CuO2− map accumulated from all cycles of depth profiles, revealing localized oxidation of Cu underneath the graphene domain after one month of aging. (c) Area line scan intensity profile of C2− and BN acquired from the marked (yellow dotted lines) area shown in (b). The sharp decline in the intensity of BN− or C2− reflects the interface between h-BN and graphene, respectively.

Figure 4f. We verified our assumptions about layer number of WS2 by Raman spectroscopy shown in Figure S7. Therefore, we suggest our presented approach is applicable as a universal methodology to analyze a variety of 2D materials. Lateral and Vertical Heterostructures of 2D Materials. Owing to its ultrahigh chemical selectivity, ToF-SIMS emerges as a powerful tool to study the interfaces and buried layers of atomic heterostructures. Figure 5a shows an optical image of the CVD-grown in-plane heterostructure of monolayer graphene and h-BN, after one month of aging under ambient atmosphere. The dark contrast of the graphene region indicates that the underneath Cu surface was oxidized in an ambient environment,46 whereas the surface covered with h-BN film was protected from the oxidation, hence showing a lighter

nonuniform region inside the WS2 domain can also be observed by ToF-SIMS mapping, which vindicates the subatomic surface sensitivity of this technique. Likewise, we revealed the isolated single-crystal domains of MoS2 on SiO2/Si substrate by mapping the S− secondary ions, as shown in Figure S6. Further, we also investigated the layer uniformity of WS2 stacks using ToF-SIMS 3D analysis. Figure 4(f) depicts an S− map accumulated from depth profiling of WS2 flakes shown in the inset optical image. The multilayers of WS2 stacks can easily be distinguished on the basis of S− signal intensity, as marked in Figure 4f. Indeed comprehensive analysis enable us to differentiate between one, two, or more individual adlayers of WS2 based on the gradual increase in the signal intensity, which is also consistent with the optical image shown in the inset of 1723

DOI: 10.1021/acs.chemmater.7b05371 Chem. Mater. 2018, 30, 1718−1728

Article

Chemistry of Materials

Figure 6. Vertical heterostructure of CVD grown monolayer graphene and MoS2 domains. (a) False-color optical image of stacked monolayer graphene/MoS2 vertical heterostructure onto the SiO2/Si substrate. The MoS2 flakes are partially covered by the graphene overlayer. (b) Raman spectrum acquired from the marked spots of graphene (orange) and MoS2 (blue) reflecting monolayers of each domain. (c) PL spectra of bare MoS2 (blue) and graphene-covered MoS2 (green) demonstrating graphene/MoS2 van der Waals heterostructure. (d) SIMS surface image obtained by using overlayers of C2−, S−, and SiO2− secondary ions revealing the lateral chemical map of the heterostructure, corresponds to the optical image (a). S− map accumulated from depth profiling after (e) 20 s and (f) 50 s of Cs+ sputtering, unveiling the MoS2 domain by removal of the top graphene layer. The scale bar used in (a, d−f) images is 100 μm.

contrast.47 Generally, the as-grown heterostructure on the Cu substrate is required to transfer onto another substrate for further investigation of the quality of such heterostructures. We also transferred graphene/h-BN heterostructure on the SiO2/Si substrate, and the Raman spectra were acquired to validate the existence of graphene domains and h-BN film in the grown heterostructure, as shown in Figure S8. In contrast, ToF-SIMS enables us to directly reveal the as-grown graphene/h-BN heterostructures on the Cu substrate. Figure 5b illustrates an overlay SIMS image of surface high lateral resolution chemical maps of C2− and BN− secondary ions representing graphene and h-BN regions, respectively. As for graphene grown on Cu, the sample was thermally annealed at 450 °C for 1 h in order to remove the interference of adsorbed organic compounds. The distribution of C2− and BN− ions clearly elucidates the isolated graphene domains stitched through a sharp interface with the continuous h-BN layer, which also corresponds well with the optical image shown in Figure 5a. The right panels (i, ii) display the individual maps of C2− and BN− ions, presenting the sharp edges of graphene and h-BN, respectively. The secondary ions distribution can also be examined by plotting an area line scan based on the signal intensity. Figure 5c plots signals intensity of C2− and BN− acquired from the area line scan (yellow dotted lines) marked in Figure 5b. The sharp decline in the intensity of BN− at the onset of graphene domain (vice versa) was identified as the interface of lateral heterostructure between h-BN and graphene. Hence, the surface selectivity of ToF-SIMS enables us to reveal the sharp interfaces between lateral graphene/h-BN heterostructure directly on the growth substrate, which can stimulate the progress about CVD growth of such atomic heterostructures with significant implications for manufacturing of integrated atomically thin electronic circuits.48 In order to verify whether

Cu is more oxidized under graphene in comparison with h-BN, a depth profiling was carried out on the same area. As shown in the panel (iii) of Figure 5b, the CuO2− map accumulated from the depth profile discloses a very much similar distribution to that of C 2 − , which specifies that the oxide layer is predominantly present underneath the graphene domains rather than h-BN film. These ToF-SIMS depth profile analyses also indicated interesting feature of h-BN film as an oxidation protection layer. Besides lateral heterostructures, ToF-SIMS was also applied to unveil the buried atomic layers of vertical or van der Waals heterostructures. Figure 6a illustrates the false-color optical image revealing a vertical heterostructure of CVD-grown monolayers of single-crystal graphene and MoS2 domains, fabricated by stacking the sequential layers onto a SiO2/Si substrate by wet transfer method.49 Although the MoS2 flake (marked with a green circle) is buried underneath the graphene domain, it is visible in the optical microscope due to the high transparency of single layer graphene. The Raman spectra shown in Figure 6b were acquired from the marked regions of graphene and MoS2 domains as indicated in the optical image. The Raman spectrum for graphene (orange line) exhibits a 2D/ G band ratio of >2, thus reflecting a single layer of graphene, hereas the spectrum for MoS2 (blue curve) shows resonance peaks, out-of-plane A1g mode at 403 cm−1 and the in-plane E12g mode at 384 cm−1, which implies a monolayer of MoS2 domain.50 Figure 6c displays the photoluminescence (PL) spectra acquired from a bare MoS2 flake (blue) and graphene-covered MoS2 flake (green). The PL intensity of graphene-covered MoS2 is significantly quenched as compared to that of a bare MoS2, which is attributed to the interaction between two atomic layers and electron quenching ability of the top 1724

DOI: 10.1021/acs.chemmater.7b05371 Chem. Mater. 2018, 30, 1718−1728

Article

Chemistry of Materials graphene layer.51 However, there is a lack of direct evidence to determine the stacking order of graphene and MoS2 layers. Figure 6d depicts the ToF-SIMS surface image attained by overlaying the chemical maps of C2−, S−, and SiO2− secondary ions, which denote graphene, MoS2, and SiO2 substrates, respectively. Again the sample had been thermally annealed at 450 °C for 1 h prior to ToF-SIMS analysis to remove the interference of adsorbed organic contamination with the graphene signal. The ToF-SIMS image clearly corresponds well to the optical image shown in Figure 6a. Interestingly, the graphene-covered MoS2 flake is concealed completely in the surface map demonstrating the stacking configuration, i.e, graphene, atop MoS2. The intrinsic capacity of ToF-SIMS allows us to chemically separate the atomic layers in the lateral as well as the vertical plane. Further, the controlled removal of the graphene overlayer by slow sputtering rate can expose the underlying MoS2 layer and SiO2 substrate. Figure 6e shows the accumulated image of S− after 20 s of Cs+ sputtering for removal of the graphene layer. The triangular marked region was designated as the graphene-covered MoS2 flake and focused for further investigation. After 20 s of sputtering, the S− signals appeared to be very weak indicating that the MoS2 is still buried underneath the graphene. The sequential Cs+ sputtering continued the graphene sublayer removal until it reaches the graphene/MoS2 interface and the MoS2 surface is exposed (the serial depth profile images are shown in Figure S9). Figure 6f reveals that, after 50 s of Cs+ sputtering, the MoS2 layer was unveiled after the removal of atop graphene layer as recognized by strong S− signals in the designated region. Likewise, the progressive elimination of the graphene layer can also be monitored by probing C2− ions, as shown in Figure S10. Consequently, prolonging Cs+ sputtering resulted in a complete removal of graphene as well as MoS2 atomic layers leading to the exposure of the SiO2 surface. It is worth mentioning here that the main focus in this study is to probe the interface between the CVD-driven atomic layers of graphene and MoS2; however, the contaminations caused by the wet transfer can also be studied in detail, as demonstrated in previous reports.30,35 We have also demonstrated revealing more complex MoS2/graphene/h-BN vertical heterostructure by 3D ToF-SIMS analysis, as shown in Figure S11. Moreover, a recent report has demonstrated utilization of ToF-SIMS to deduce the stacking order of SMoSe Janus structure.52 Hence, ToF-SIMS high lateral resolution chemical mapping along with depth profiling provide an efficient platform to obtain detailed chemical analysis of the stacked 2D atomic layers, which are critical toward optimizing their CVD synthesis to the device fabrication. However, it should be pointed out that this technique bears a drawback compared to conventional techniques: the depth profiling analysis is destructive. This may limit its applications to final devices.

graphene/h-BN in-plane heterostructure. As the graphene signal can be affected by the adsorbed organic contaminations, graphene-containing samples had to be subjected to an in situ thermal annealing prior to ToF-SIMS imaging. Furthermore, we have developed a new approach based on ToF-SIMS 3D analysis to unveil the atomic layer structure of 2D materials in the z-direction. Given the limited depth resolution of ToFSIMS, our approach proposed to accumulate the signal specific to 2D materials along the sputtering depth and found that this accumulated signal was linearly proportional to the number of atomic layers in the z-direction. Consequently, this signal provided a chemical intensity contrast, similar to the optical contrast that can be directly utilized to resolve the number of adlayers. In the case of CVD-grown MLG, for example, up to six adlayers were resolved by this method. However, contrary to optical contrast, the chemical intensity contrast does not depend on the substrate. This approach has been proven to be versatile as it has been successfully applied to various 2D materials such as graphene, h-BN, and WS2. Moreover, it could unveil the multilayer structures that were otherwise invisible under an optical microscope. It has also been used to reveal the buried interfaces of the graphene/MoS2 van der Waals heterostructure. The main advantage of our approach is that the measurements are independent of the substrate used, and consequently this technique can be directly applied to the grown substrate. This will avoid the time-consuming transfer process and the potential damage that may be brought in. In summary, our work presented a universal methodology to evaluate the quality of CVD grown 2D crystals and their heterostructures, which can contribute toward steering the progress of 2D materials research to the fast track.



EXPERIMENTAL SECTION

CVD Growth of Single-Crystal and Multilayer Graphene. We grew graphene using our previously reported method.37 Briefly, a 25 μm thick Cu foil (Alfa Aesar, 99.8%) was stacked on a Ni substrate and loaded inside the CVD furnace.33,53 The Cu foil was preoxidized at 200 °C for 5 min followed by ramping up the temperature to 1040 °C under the flow of 250 sccm of Ar. The Cu foil was annealed at growth temperature by an Ar/H2 (10:1) mixture. The graphene growth commenced by introducing CH4 gas (5% to the total flow, 500 ppm diluted) into the system for an hour to obtain single-crystals of graphene. For multilayer graphene, the supporting substrate for Cu is replaced with quartz instead of Ni. Finally, the system cooled down rapidly to the room temperature under the flow of Ar/H2. CVD Growth of h-BN and Graphene/h-BN Heterostructure. A 25 μm thick Cu foil (Alfa Aesar, 99.8%) was cut into a piece of size 2 cm × 4 cm and chemically polished in acetic acid bath followed by rinsing with water. Thereafter, the Cu/quartz substrate put in the CVD furnace. The furnace ramped up to the growth temperature of 1040 °C under the stream of Ar/H2 (30:1) mixture followed by annealing for 30 min. Thereafter h-BN precursor is introduced into the reaction chamber by sublimating the ammonia borane (AB) powder. The growth was sustained for 20 min to attain single crystals of h-BN domains. Finally, the growth was terminated by fast cooling under a flow of Ar. For graphene/h-BN lateral heterostructure, after obtaining the single-crystal domains of graphene the CH4 supply is terminated, and simultaneously the sublimation of ammonia borane (AB) powder is started at 120 °C. The growth was sustained for 30 min to ensure the lateral stitching of the h-BN continuous sheet with graphene domains. CVD Growth of TMDs. Synthesis of WS2. First, we prepared tungstic acid solution by dissolving 1.8 g of Na2WO4·2H2O in 100 mL if water. Then, 3 mol/L HCl solution was added dropwise with stirring until pH ≈ 1. After standing for 24 h, the precipitate was dissolved into a concentrated solution of oxalic acid (1.4 g in 20 mL of water) by



CONCLUSIONS Through a series of purpose-built CVD-grown 2D materials and their heterostructures, we have systematically explored the potential of ToF-SIMS as a tool to gauge the layer uniformity and to reveal individual atomic layers in the z-direction. Our results showed that the layer uniformity could be easily verified using ToF-SIMS high spatial resolution imaging owing to its extreme surface sensitivity (