Direct Observation of 2D Electrostatics and Ohmic Contacts in

Feb 21, 2017 - Both p–n and n–n junctions are observed, and a flat-band condition (zero Schottky barrier height) is found for lightly n-doped WS2,...
0 downloads 9 Views 5MB Size
Download PDF
Direct Observation of 2D Electrostatics and Ohmic Contacts in Template-Grown Graphene/ WS2 Heterostructures Changxi Zheng,*,†,‡,∥ Qianhui Zhang,‡,∥ Bent Weber,†,⊥,○ Hesameddin Ilatikhameneh,§ Fan Chen,§ Harshad Sahasrabudhe,§ Rajib Rahman,§ Shiqiang Li,¶ Zhen Chen,⊥ Jack Hellerstedt,†,⊥ Yupeng Zhang,# Wen Hui Duan,‡ Qiaoliang Bao,#,□,○ and Michael S. Fuhrer*,†,⊥,○ †

Monash Center for Atomically Thin Materials, ‡Department of Civil Engineering, ⊥School of Physics and Astronomy, #Department of Materials Science and Engineering, and ○ARC Centre of Excellence in Future Low Energy Electronics Technologies, Monash University, Victoria 3800, Australia § Network for Computational Nanotechnology (NCN), Purdue University, West Lafayette, Indiana 47906, United States ¶ Department of Electrical and Electronic Engineering, University of Melbourne, Victoria 3010, Australia □ Institute of Functional Nano and Soft Materials (FUNSOM), Jiangsu Key Laboratory for Carbon-Based Functional Materials and Devices, and Collaborative Innovation Center of Suzhou Nano Science and Technology, Soochow University, Suzhou 215123, People’s Republic of China S Supporting Information *

ABSTRACT: Large-area two-dimensional (2D) heterojunctions are promising building blocks of 2D circuits. Understanding their intriguing electrostatics is pivotal but largely hindered by the lack of direct observations. Here graphene−WS2 heterojunctions are prepared over large areas using a seedless ambient-pressure chemical vapor deposition technique. Kelvin probe force microscopy, photoluminescence spectroscopy, and scanning tunneling microscopy characterize the doping in graphene−WS2 heterojunctions asgrown on sapphire and transferred to SiO2 with and without thermal annealing. Both p−n and n−n junctions are observed, and a flat-band condition (zero Schottky barrier height) is found for lightly n-doped WS2, promising low-resistance ohmic contacts. This indicates a more favorable band alignment for graphene−WS2 than has been predicted, likely explaining the low barriers observed in transport experiments on similar heterojunctions. Electrostatic modeling demonstrates that the large depletion width of the graphene−WS2 junction reflects the electrostatics of the onedimensional junction between two-dimensional materials. KEYWORDS: electrostatics, heterostructure, transition-metal dichalcogenides, graphene, ohmic contacts devices by transfer of graphene onto MoS2.7 Their devices showed record-high low-temperature Hall mobility up to 1020 cm2 V−1 s−1 in monolayer MoS2. Although an impressive demonstration of the potential of graphene−TMD contacts for low contact resistance, the challenging fabrication technique is unlikely to allow processing of devices on a large scale. Consequently, a few groups very recently explored alternative methods of fabricating large-scale edge-contacted MoS2− graphene22−25 and WS2−graphene25 heterojunctions using chemical vapor deposition (CVD)22−24 and metal−organic chemical vapor deposition (MOCVD).25 Electronic devices with graphene edge contacts have been fabricated based on

S

emiconducting transition-metal dichalcogenides (TMDs) are a class of emerging materials exhibiting large (∼1−2 eV) direct band gaps in the visible spectrum at monolayer thickness.1 These materials hold huge potential for optoelectronics and valleytronics owing to their intriguing material properties such as strong light−matter interaction,2 near-unity photoluminescence (PL) quantum yield,3 large exciton and trion binding energy,4,5 high carrier mobility,6−8 large spin−orbital coupling,9 and degenerate valleys.10 So far, a wide spectrum of applications including transistors,11 photodetectors,12 light-emitting diodes,13−16 lasers,17,18 and biological sensors19 have been demonstrated on TMDs. However, the further development of these electronics is hindered by the challenge of making reliable electrical contacts to TMDs with low-ohmic contact resistance.20,21 Recently, Cui et al. demonstrated a technique to make high-performance MoS2 © 2017 American Chemical Society

Received: November 21, 2016 Accepted: February 21, 2017 Published: February 21, 2017 2785

DOI: 10.1021/acsnano.6b07832 ACS Nano 2017, 11, 2785−2793

Article

www.acsnano.org

Article

ACS Nano

Figure 1. Schematic illustration of the steps of preparing a graphene−WS2 heterojunction: (a) wet transfer of CVD graphene on a sapphire substrate, (b) patterning the graphene by lithography and O2 plasma etching, and (c) selective growth of WS2 monolayer connecting to graphene. (d−f) Optical images corresponding to the three steps. The scale bar is 20 μm and can be applied to all the images.

Figure 2. (a) Optical image showing a boundary of the graphene−WS2 heterojunction. (b) Raman and (c) photoluminescence (PL) spectra of the monolayer WS2. (d) Raman spectrum of the monolayer graphene after the process of WS2 growth. Maps of the graphene−WS2 region showing (e) integrated PL intensity, (f) energy of PL peak, (g) PL peak width, and (h) integrated Raman E12g peak intensity. The scale bars are 4 μm.

scanning Raman microscopy, atomic force microscopy (AFM), Kelvin probe force microscopy (KPFM), and scanning tunneling microscopy (STM). The graphene retains high quality even after the high-temperature heating during the CVD growth. The WS2 film is atomically flat and shows strong photoluminescence and uniform optical properties. Our approach demonstrates a facile technique to produce lateral graphene−WS2 heterojunctions that can be transferred to varying substrates. Indeed, we find different doping concentrations by transferring the heterojunctions from sapphire to a SiO2/Si substrate and subsequent thermal annealing of the heterojunctions on SiO2/Si. Interestingly, the KPFM measurements reveal that the built-in potential between graphene and WS2 is small (−200 to +100 meV) and can even be negative (electron transfer from graphene to WS2) for lowdoped WS2. The results imply a much more favorable band line

these techniques, showing low-resistance contacts with small Schottky barriers.22,24,25 Nevertheless the nature of the barrier and the mechanism governing the low contact resistance in these heterojunctions remain unclear. In particular, a large Schottky barrier is expected for graphene−WS2 heterojunctions due to the small electron affinity of WS2.26−29 Herein, we demonstrate a CVD technique to prepare largescale graphene−WS2 heterojunctions at ambient pressure without using seeding. Graphene is patterned to form a template for growth of WS2 on sapphire [Al2O3(0001)]. Monolayer WS2 selectively grows on the bare sapphire starting at the graphene edges and spreading out on the bare region of sapphire to form continuous monolayer films intimately edgeconnected to graphene. The optical, structural, and electrical properties of our graphene−WS2 heterojunctions are investigated by multiple characterization techniques including 2786

DOI: 10.1021/acsnano.6b07832 ACS Nano 2017, 11, 2785−2793

Article

ACS Nano

Figure 3. (a) Optical image of a graphene−WS2 heterojunction on a SiO2/Si substrate. Corresponding photoluminescence (PL) and Raman mapping images of the integrated intensity of the (b) graphene G band and the (c) PL and (d) Raman of WS2. The grain boundary between the two WS2 islands is highlighted by the dashed lines. The scale bar in (d) is 10 μm.

up for graphene/WS2 than previously predicted27−30 and indicate that ohmic graphene/WS2 contact (zero or negative Schottky barrier height) is possible for a range of realistic doping conditions in graphene and WS2. Theoretical modeling of the potential profile in graphene/WS2 junctions demonstrates that the large depletion widths observed in the KPFM experiments are a direct result of the electrostatics of the 1D junction between 2D materials.

and could signify damage to the graphene from the CVD growth process. However, since the peak is intense and very broad, we speculate the signal is not contributed by the defects of graphene but rather due to amorphous carbon generated from the PMMA residue during the CVD growth at 780 °C (see STM measurement below).34 Figure 2e−g present PL mapping of the WS2 film showing integrated intensity (Figure 2e), peak energy (Figure 2f), and peak width (Figure 2g), respectively. The images indicate uniform optical properties across the WS2 film, i.e., low variation in PL intensity and peak energy, and narrow peak width, comparable to single-crystal WS2 grown on sapphire reported elsewhere.36,37 Figure 2h shows a map of the integrated E12g Raman intensity of WS2. The higher intensity observed in narrow regions near the graphene edge likely indicates some multilayer WS2, which implies the graphene edges are the preferential nucleation sites. We use Raman and PL mapping to further explore the nucleation and growth in a sample where WS2 growth is arrested before completion of a monolayer. Since the Raman signal of graphene on sapphire is weak, we transferred a graphene−WS2 heterojunction onto a SiO2/Si substrate using the wet chemical technique; see Methods. Figure 3a shows the optical image of the transferred graphene−WS2 heterojunction. The maps shown in Figure 3b−d present the integrated intensities of the graphene G band, WS2 PL, and E12g Raman mode, respectively. As highlighted by the dashed straight white lines, sharp boundaries form between graphene and WS2. Comparing with the van der Waals surface of graphene, the edges of patterned graphene possess substantial dangling bonds opened up by oxygen plasma during the patterning. These dangling bonds can act as nucleation sites for the growth of WS2 crystals.38 Therefore, during the CVD process, WS2 preferentially nucleates at the edge of graphene and extends on the sapphire surface but not on graphene due to the larger surface energy of sapphire. As shown in Figure 3a, two WS2 islands nucleate separately on the two graphene edges. As the CVD process goes further, the two islands grow on sapphire and meet each other, leading to the formation of a grain boundary. The grain boundary, which can be identified by the PL intensity, is highlighted by the dashed curves in Figure 3. Nevertheless, there is no WS2 growing on the graphene surface due to its van der Waals surface. Therefore, the patterned graphene can be used as the CVD growth template for WS2. We can see that a clean and sharp boundary between graphene and WS2 can be achieved using optimized growth parameters; see Figure S2c−e. We employ low-temperature (∼4.5 K) STM to investigate the atomic structure of graphene on sapphire postgrowth.

RESULTS AND DISCUSSION Figure 1a,b,c present schematically the main process steps to prepare graphene−WS2 heterojunctions, and Figure 1d,e,f show the corresponding optical images of the samples at each step in Figure 1a,b,c, respectively. First, large-area CVD-grown graphene coated with poly(methyl methacrylate) (PMMA) is transferred onto a sapphire substrate after etching away the copper foil using ammonium persulfate (APS) solution (Figures 1a and d). After the removal of PMMA, the transferred graphene is patterned using conventional photolithography and etched by oxygen plasma to form the graphene patterns (Figure 1b,e). Lastly, monolayer WS2 film is selectively grown by CVD on the bare sapphire surface until forming a continuous film (Figure 1c,f). Briefly, WO3 and sulfur precursors are used for the CVD growth under Ar/H2. The CVD setup for the growth of WS2 is schematically illustrated in Figure S1a, and further description is given in the Methods section and Figure S1. The flow of H2 is found to be an important parameter determining the growth rate; see Figure S1b−d, which shows the crystal size and layer number of WS2 crystals increasing as the flow of H2 increases. Optimization of the H2 flow and the substrate location in the quartz tube allows continuous monolayer WS2 film growth to form large-area graphene−WS2 hesterostructures; see Figure S1e. We first discuss the optical properties of our graphene−WS2 heterostructures, as determined by photoluminescence and Raman spectroscopy. Figure 2a shows an optical micrograph of a typical heterojunction. Figure 2b shows the point PL spectrum taken from the WS2 film, showing a sharp single Lorentzian peak at 2.0 eV,31 and the Raman spectrum (Figure 2c) shows the E12g (354 cm−1) and A1g (415 cm−1) Raman modes with a 61 cm−1 difference.32,33 These spectral characteristics indicate that the thickness of the WS2 film is monolayer. Figure 2d shows the point Raman spectrum of graphene. In addition to the G band (1596 cm−1) and 2D band (2704 cm−1), there appears a substantial broad peak at the position of the graphene D band (∼1350 cm−1); this component is much more intense and broader after the CVD growth of WS2. The D band is typically associated with point defects in graphene34,35 2787

DOI: 10.1021/acsnano.6b07832 ACS Nano 2017, 11, 2785−2793

Article

ACS Nano

Figure 4. (a) Scanning tunneling microscopy (STM) image of monolayer graphene on atomically flat sapphire. Inset: Schematic diagram for the STM measurement. (b, c) High-resolution STM images of graphene. A point defect is shown in (c). (d) Scanning tunneling spectrum of graphene. The location of the Dirac point is indicated by the red arrow.

Figure 5. (a) Atomic force microscopy (AFM) topographic image of a graphene−WS2 heterojunction. (b) Correlated contact potential difference measured by Kelvin probe force microscopy over the same area as in (a). (c) Correlated profiles of height and contact potential difference. Inset: Schematic diagram for measuring the surface potential of the heterojunction using the scanning Kelvin probe microscopy (SKPM). The height is taken along the white dashed line in (a), while the contact potential difference is averaged over the vertical direction in (b).

and clean, conforming to the atomic steps of the sapphire substrate. The stitching boundary between graphene and WS2 is a sharp line due to the well-confined growth of WS2 by the graphene template. A higher resolution image of the junction is provided in Figure S2. We can see that a clean and sharp boundary between graphene and WS2 can be achieved using optimized growth parameters; see Figure S2c−e. Figure 5b shows the contact potential difference (CPD) VCPD of the graphene−WS2 heterojunction obtained by KPFM. Figure 5c presents the VCPD profile perpendicular to the graphene−WS2 heterojunction obtained by averaging Figure 5b along the vertical direction. The overlapped height profile taken along the white dashed line (see Figure 5a) is plotted to indicate the boundary between graphene and WS2. The change in VCPD across the junction, ∼0.19 eV, reflects the built-in potential of the graphene/WS2 junction. The CPD is the difference between the KPFM tip and sample workfunctions, VCPD = ϕtip − ϕsample. The workfunction of the Pt/Ir tip is estimated to be 4.89 ± 0.1 eV, verified by calibration with graphite and a Au standard; see Figures S3 and S4. Therefore, we estimate the workfunctions of graphene and WS2 are 4.85 and 5.04 eV, respectively. This is consistent with the workfunction of graphene of 4.9 eV inferred from STS. The graphene is heavily p-type doped, while the WS2 is n-type doped based on the workfunction values, leading to the formation of a p−n junction. However, electron transfer is from graphene to WS2, enhancing the p-type carrier concentration in graphene and n-type concentration in WS2. Therefore, the built-in potential is negative (−0.19 eV) compared to the typical p−n junction.

Figure 4a−c show atomic resolution images of graphene on a terrace of saphhire (0001). We find that the graphene layer is nearly atomically flat (root-mean-square height = 0.124 nm). A small number of point defects are seen as shown in Figure 4c. We infer a point defect density of (8.9 ± 4) × 1012 cm−2. However, such a defect density cannot explain the very broad Raman D peak (Figure 2d) typical of highly disordered sp2 carbon.39 Thus, we conclude that the D peak arises from amorphous carbon, not observed in the areas probed by STM, and the value calculated from the STM image is overestimated. In Figure 4d, the scanning tunneling spectroscopy (STS) data show the Dirac point of the hole-doped graphene at approximately +300 mV. The position of the Dirac point indicates a p-type carrier density of ∼6.6 × 1012 cm−2. The workfunction of the graphene is estimated to be 4.9 eV, assuming the workfunction of neutral graphene is 4.6 eV.40 The slight reduction of the differential conductance around zero bias in the STS spectrum has been observed previously41 and is most likely due to the phonon gap where the STM bias is too low to allow inelastic tunneling mediated by excitation of an optical phonon within the graphene monolayer. We note that no tunneling contact could be obtained on the WS2 areas of the sample, suggesting that the WS2 on sapphire was not sufficiently doped to allow current to flow at 4.5 K. We now turn to the characterization of the electrostatic profile of the graphene−WS2 heterojunction. We employ AFM and KPFM to investigate the morphology and the correlated surface potential of the heterojunction, respectively. Figure 5a presents the AFM topography of a graphene−WS2 heterojunction. As shown, the CVD-grown WS2 film is atomically flat 2788

DOI: 10.1021/acsnano.6b07832 ACS Nano 2017, 11, 2785−2793

Article

ACS Nano

Figure 6. (a) Workfunction of graphene−WS2 heterojunctions on a sapphire substrate (red curve), annealed on SiO2/Si (pink curve), and unannealed on SiO2/Si (blue curve). Also shown are the calculated workfunction profiles for the heterojunctions on sapphire and unannealed on SiO2/Si (dashed black curves) and the calculated workfunction profiles taking into account the AFM tip transfer function (solid black curves). (b) Photoluminescence spectra for graphene−WS2 heterojunctions on a sapphire substrate (red curve), annealed on SiO2/Si (pink curve), and unannealed on SiO2/Si (blue curve). Also shown are the Lorentzian components corresponding to exciton emission (dotted black curves) and trion emission (dashed black curves). (c) Schematic diagrams of doping of the heterojunctions modulated by substrates and thermal annealing. (d, e) Calculated band diagrams for graphene−WS2 heterojunctions.

1012 cm−2); hence the sample has changed from a p−n junction on sapphire to an n−n junction on SiO2/Si. Subsequently we annealed the transferred heterojunction on SiO2/Si under Ar and sulfur vapor at 200 °C for 2 h to expel the moisture and chemical residues.43 Surprisingly, the workfunction profile of the annealed heterojunction (pink curve in Figure 6a) is unchanged across the junction, indicating nearly identical workfunctions of 4.76 eV for graphene and WS2. The built-in potential across the annealed heterojunction thus is zero. The PL spectrum of the annealed WS2 shows both peaks of exciton and trion (pink curve in Figure 6b), indicating light n-doping, i.e., a Fermi energy near the WS2 band edge. The graphene is slightly p-doped (∼2 × 1012 cm−2). The simultaneous observation of the flat-band condition (zero built-in potential, pink curve in Figure 6a) and a significant free electron concentration (comparable trion and exciton PL, pink curve in Figure 6b) indicates a zero-barrier height ohmic contact between graphene and the conduction band electrons of WS2. This is consistent with electrical measurements of a very small barrier between graphene and WS2.25 Our results indicate that the small barrier arises because of similar workfunctions of graphene and lightly n-doped WS2. This is surprising, given that theoretical predictions for the electron affinity of WS2 are in the range of 3.5−4.1 eV,27−29 much smaller than the 4.8 eV workfunction we observe for lightly n-doped WS2, where the Fermi energy is close to the conduction band edge and the workfunction should be similar to the electron affinity. We do not understand this surprising discrepancy. However, we note that due to the small dielectric

Since the doping concentration in 2D materials is very sensitive to their environment such as substrate and ambient,6 we therefore transferred the graphene−WS2 heterojunction from sapphire onto a SiO2/Si substrate (see Methods) and carried out thermal annealing to generate different doping concentrations in the heterojunction. For better comparison, we plot the workfunction profiles of the heterojunctions on sapphire and on SiO2/Si before and after annealing together in Figure 6a. The workfunction profile of the heterojunction on sapphire (red line, Figure 6a) shows the largest workfunctions among the three samples due to the heavy hole doping in graphene and slight electron doping in WS2. The depletion width of WS2, which is the width of the spatially varying part of the workfunction curve of WS2, reaches 4 μm. The PL spectrum of WS2 on sapphire (red curve, Figure 6b) shows a single Lorentizian peak at 2.0 eV, consistent with small doping in WS2 and hence negligible trion formation.5,37 Upon transfer to SiO2/Si, the workfunctions of graphene and WS2 (blue curve, Figure 6a) shift to 4.5 and 4.4 eV with 0.39 and 0.49 eV drops, respectively. The large drops in the workfunctions indicate increased electron concentration in both materials. This likely reflects the electron-withdrawing character of the postgrowth sapphire,42 compared to the relatively less reactive SiO2. The PL spectrum of WS2 on SiO2 now consists of only the single Lorentzian peak at 1.96 eV, corresponding to trion emission, also indicating a large free electron concentration (blue curve in Figure 6b). The graphene workfunction of 4.5 eV corresponds to slight n-doping (∼1 × 2789

DOI: 10.1021/acsnano.6b07832 ACS Nano 2017, 11, 2785−2793

Article

ACS Nano

electrostatics of a 1D junction between 2D materials. Figure 7 illustrates this, comparing the simulated depletion profile of the

constant of isolated monolayer WS2, the interaction corrections to the bandstructure are large, on the order of 1 eV.29 Hence we expect that modification of the screening environment through introduction of a dielectric substrate or through metallic screening by carriers in WS2 will strongly increase the electron affinity. A strong interaction renormalization of the electron affinity, leading to negative compressibility (increase of workfunction with increasing carrier density), has recently been observed in multilayer MoS2/graphene structures,43 and the effect should be larger in monolayer WS2 due to the smaller dielectric constant. More work, both theoretical and experimental, is needed to understand the electron affinity of highly doped WS2 on substrates. Based on the experimental observations, schematics are presented in Figure 6c to interpret the doping of the heterojunctions, which is mainly due to the substrate surfaces. After the high-temperature process of CVD growth, the sapphire surface, Al2O3(0001), is terminated by a single Al plane.44 This surface has a Lewis acid character and can accept electrons to reduce the native electron doping of WS2.45 As a result, we observe a very low level of negative doping in WS2 and a large hole doping in graphene. The SiO2/Si substrate exposed to ambient is passivated by hydroxyl groups and adsorbed water and has less electron-accepting character. After thermal annealing, we expect that some of the hydroxyl groups are removed, creating an electron-accepting surface. Consequently, the electron-doping concentrations of WS2 and graphene decrease. To further understand the characteristics of our 2D heterojunctions, we modeled the workfunction profiles of the heterojunctions on sapphire and on SiO2/Si before annealing using a self-consistent semiclassical Poisson solver (see Supplementary Discussion). The black dashed curves shown in Figure 6a are the calculated workfunction profiles without considering the finite resolution of the AFM tip. The model predicts a very small depletion width in semimetallic graphene; hence we expect that the observed potential profile in the graphene reflects the finite experimental resolution. We therefore treat the potential profile in the graphene as a transfer function46 reflecting the instrumental resolution (see Supplementary Figure 7 and Discussion). Convolution of the calculated workfunction (black dashed curve) with the transfer function produces the black solid curve in Figure 6a, which well fits the experimental curve. We plot the calculated band diagrams of the heterojunctions on sapphire and on SiO2/Si before annealing in Figure 6d. The Fermi levels of the graphene and WS2 are directly read from our KPFM measurements. The electron affinity of WS2, EC, is adjusted within the model to obtain the best fit to the experimental depletion width in Figure 6a. We find the best fit for EC = −4.81 eV (Figure 6d). This implies a low doping concentration of 3 × 107 cm−2 for WS2 on sapphire, consistent with the observation of pure exciton PL. For the sample transferred to SiO2/Si the WS2 is degenerately doped with Fermi energy around 400 meV above the conduction band edge, and for the annealed sample on SiO2/Si, the Fermi energy lies at the band edge (Figure 6e). These Fermi energy positions are consistent with the PL observation of exciton emission, trion emission, and mixed exciton/trion emission in samples on sapphire, unannealed on SiO2/Si, and annealed on SiO2/Si, respectively. Finally, we note that the very long depletion width for the low-doped WS2 on sapphire is a direct consequence of the

Figure 7. Calculated workfunction profiles for a two-dimensional graphene−WS2 heterojunction (blue curve) and a hypothetical three-dimensional graphite−WS2 heterojunction with similar band structure and doping (black curve).

2D graphene/WS2 heterojunction (blue curve) with the profile of a hypothetical bulk graphite−WS2 planar heterojunction with similar band structure and doping level (black curve) calculated using conventional analytical equations (see Supplementary Discussion). The depletion width of WS2 in the 2D case is around 4.5 μm, which is more than twice as wide as would be expected for the bulk heterojunction. The result is consistent with the recent theoretical prediction by Yu et al.23 This confirms that our KPFM measurements have provided the direct experimental image of the electrostatic profile of a 1D junction with a 2D semiconductor.

CONCLUSIONS In summary, a large-scale graphene−WS2 heterostructure is synthesized by CVD without seeding. Owing to the lack of a nucleation site on the van der Waals surface of graphene, monolayer WS2 selectively grows on the bare region of sapphire to form large-scale films connecting to the edges of graphene. The PL, Raman, and AFM measurements indicate the high quality of the CVD-grown WS2 film. Even after the hightemperature process, only a small amount of point defects are observed in graphene by our STM study. Using KPFM, the band alignment between graphene and WS2 as well as their workfunctions are measured when the heterojunctions are on sapphire or SiO2/Si. The substrate and thermal annealing affect the doping concentration in graphene and WS2, leading to the formation of p−n and n−n junctions and positive, zero, and negative built-in potentials. Together with a self-consistent semiclassical Poisson solver, we fit the workfunction profiles across the heterojunctions and plot their band diagrams. The results directly indicate the formation of an ohmic contact (zero Schottky barrier) between graphene and lightly n-doped WS2 for the case of heterojunctions on annealed SiO2/Si. Our results directly demonstrate that the low Schottky barrier between graphene and WS2 is the consequence of a favorable band lineup and have important implications for the design of electrical contacts to WS2. Our observations imply an electron affinity of WS2 much larger than anticipated,27−29 possibly due to increased dielectric or metallic screening in doped WS2 on dielectric substrates. Combining the experiments with theoretical calculations, we also demonstrate that the depletion width in 2D systems is much larger than that of 3D systems due to their electrostatics. 2790

DOI: 10.1021/acsnano.6b07832 ACS Nano 2017, 11, 2785−2793

Article

ACS Nano

sample at a current set point of 3 nA. The bias spectroscopy shows an average of five single spectra with the differential conductance recorded simultaneously using a digital lock-in (amplitude: 50 mV at frequency 1.8 kHz) and with the feedback switched off during the measurement. For both the KPFM and STM measurements, the samples are grounded and biased through a Au electrode deposited by a stencil method, respectively.

METHODS Sample Preparation. 1L/2L graphene film (Graphene Supermarket) was transferred to sapphire(0001) substrates by spinning PMMA (A4) on graphene and etching the copper foil in 0.1 M/L ammonium persulfate (APS, Sigma-Aldrich) solution. The PMMA on graphene was removed by soaking in PG remover overnight and acetone and IPA (isopropyl alcohol) for 1 min. Thereafter, a layer of AZ MIR 701 photoresist was coated on the substrate by spin coating at 4000 rpms for 1 and 3 min of soft baking at 110 °C. To define the pattern, the photoresist was exposed to UV light and developed by AZ 726 metal ion free developer. Oxygen plasma was applied to etch away the exposed graphene to form patterns. Finally, the photoresist was removed by sequentially soaking in AZ 100 K remover (3 min), deionized water (5 min), AZ PG remover (70 °C, 30 min), acetone (5 min), and IPA (1 min). To grow WS2 in CVD, we used sulfur (SigmaAldrich, 1 g) and WO3 powders (Sigma-Aldrich, 2.0 g) as the precursors. Sapphire substrates were cleaned by 10 min acetone and 2propanol sonication. The atmospheric pressure growth was performed in a 1 in. diameter quartz tube flowing ultrahigh purity argon (200 sccm) and hydrogen (15 sccm was found to be optimal for large-area WS2 growth). WO3 and sapphire substrates were loaded into separated quartz crucibles and located at the center and downstream of the furnace, respectively. The sulfur powder was located upstream of the WO3 and substrates and independently heated. The furnace and the sulfur were heated to 780 and 180 °C, respectively, in 30 min and remained at the temperatures for 15 min, followed by fast cooling with furnace opening. Optical Characterization. Raman/PL measurements were carried out on a confocal microscope system (WITec alpha 300R) with a 100× objective with a numerical aperture of 0.90. A 532 nm laser with a spot diameter of ∼500 nm was used to excite samples placed on a piezocrystal-controlled scanning stage. The Raman/PL spectra were collected using a 600 line/mm grating. Short integration time (0.2 s) and low laser power (50 μW) were applied during PL/Raman mapping. A much higher laser power (2 mW) and longer integration time (1 s) were applied to take the single Raman spectrum of graphene on sapphire. The PL intensity images were obtained by summing the PL intensity from 590 to 660 nm, and the PL peak position and full width at half-maximum were estimated from fits to a single-Lorentzian line shape. Heterojunction Transfer. The transfer of graphene−WS 2 heterojunctions from sapphire to SiO2/Si was achieved by a PMMAassisted wet transfer technique. First, PMMA (A6) was spin coated on the surface of an as-grown heterojunction-sapphire sample, followed by a soft bake at 80 °C for 3 min. Thereafter, the PMMA-coated sample was soaked into 3 M potassium hydroxide solution at 100 °C for 2 h. After soaking, the PMMA film with heterojunction was separated from the sapphire by dipping the sample into distilled water and then floating on top. The separated WS2/PMMA film was fished out by a SiO2/Si substrate. The water trapped between the PMMA film and the substrate was evaporated by a 15 min baking at 60 °C. Finally, the PMMA film was removed by acetone and IPA rinsing. Transmission Electron Microscopy (TEM) Characterization. The heterojunctions were transferred to a TEM copper grid and cleaned using the techniques described above. The high-angle annular dark-field experiments were carried out using a double aberration corrected FEI TITAN3 at 300 kV, which gave the best performance of the TEM to image the heterojunction. Scanning Probe Characterization. The AFM and KPFM measurements were carried out on a Bruker Dimension Icon SPM through LiftMode, a dual-pass technique. The topography line profile of the sample was taken in amplitude-modulated tapping mode at the first pass, and the corresponding surface potential was taken at the second pass, tracing the acquired topography line at a set lift height (100 nm) and the ac voltage amplitude of 500 mV, respectively. A Pt/ Ir-coated silicon cantilever was used as the (conductive) probe. STM images were obtained with a CreaTec LT-STM/AFM under ultrahigh vacuum (base pressure