Disassembling 2D van der Waals crystals into macroscopic monolayers and reassembling into artificial lattices

See allHide authors and affiliations

Science  21 Feb 2020:
Vol. 367, Issue 6480, pp. 903-906
DOI: 10.1126/science.aba1416

Larger monolayers with gold tapes

Although the exfoliation of monolayers of materials such as transition metal dichalcogenides produces high-quality electronic materials with low defect densities, the size of the monolayers is limited to the micrometer scale. Liu et al. modified this method by creating atomically flat gold layers on polymer supports. The strong van der Waals adhesion of the gold layer allowed monolayers to be exfoliated on the centimeter scale. Multilayers could be reassembled to artificial structures, such as a MoSe2/WSe2 single-crystal bilayer with a twist angle chosen to quench intralayer exciton formation.

Science, this issue p. 903


Two-dimensional materials from layered van der Waals (vdW) crystals hold great promise for electronic, optoelectronic, and quantum devices, but technological implementation will be hampered by the lack of high-throughput techniques for exfoliating single-crystal monolayers with sufficient size and high quality. Here, we report a facile method to disassemble vdW single crystals layer by layer into monolayers with near-unity yield and with dimensions limited only by bulk crystal sizes. The macroscopic monolayers are comparable in quality to microscopic monolayers from conventional Scotch tape exfoliation. The monolayers can be assembled into macroscopic artificial structures, including transition metal dichalcogenide multilayers with broken inversion symmetry and substantially enhanced nonlinear optical response. This approach takes us one step closer to mass production of macroscopic monolayers and bulk-like artificial materials with controllable properties.

Since the first report of monolayer graphene in 2004 (1), studies of 2D materials have grown into one of the most active research fields (2). Monolayers, and especially their homo- or heterostructures have been excellent subjects for the exploration of new physical phenomena and hold great promise for next-generation devices, particularly quantum devices (35). Many of these new quantum phenomena, such as Moiré excitons in transition metal dichalcogenide (TMDC) heterobilayers and superconductivity in twisted bilayer graphene, rely on monolayers with single-crystal lattices (312). However, despite extensive advances in research, the absence of high-throughput methods to produce high-quality 2D single crystals with macroscopic sizes remains a major challenge for their mass production and potential commercialization. Past attempts at producing 2D monolayers were often limited by problems such as material quality, scalability, and size. Liquid phase exfoliation is scalable but generates small sizes (only a few micrometers) and poor quality (13). Chemical vapor deposition (CVD) can grow continuous monolayers on wafer scales, albeit in the polycrystalline form with high defect density, thereby limiting their performance in electronic devices (14). Mechanical exfoliation using the famous Scotch tape method (1) has produced the highest-quality monolayers thus far; however, the typical lateral dimensions are <100 μm, with very low yield. Larger flakes can be obtained using metals with stronger van der Waals (vdW) adhesion to 2D materials than interlayer vdW force, as demonstrated for TMDCs. Exfoliation of bulk TMDC crystals on gold substrates has been reported to yield monolayers up to the centimeter scale (15, 16), but it is difficult to remove the TMDC monolayers from the gold substrate. Another strategy involves evaporation of metal onto the surface of a bulk TMDC crystal, followed by exfoliation, transferring onto a desired substrate, and chemical etching (17, 18). This method has been demonstrated to yield single-crystal TMDC monolayers up to 500 μm in lateral dimensions (17) and transfer CVD films up to wafer scale (18). However, direct deposition of metal onto TMDC bulk crystal introduces considerable defects in the TMDC monolayer (19).

Here, we demonstrate a nondestructive, high-throughput, and widely applicable technique to disassemble 2D vdW crystals layer by layer into single-crystal monolayers, with near-unity yield. This method works for a range of vdW materials and generates monolayers with macroscopic dimensions limited only by the sizes of the bulk crystals. As schematically shown in Fig. 1A, a critical step is obtaining an atomically flat gold tape, i.e., an ultraflat gold film on a polymer substrate, which is achieved using a template-stripping technique (20, 21). After evaporation of a thin gold film on an ultraflat surface of highly polished silicon wafer, the gold film is stripped off the substrate with thermal release tape and a polyvinylpyrrolidone (PVP) interfacial layer. The exposed gold surface is templated by the flat substrate with roughness on the Å scale (20, 21). The ultraflat gold tape allows intimate and uniform vdW contact between the gold and the 2D vdW crystal surface, exfoliating a complete monolayer that can be transferred onto the desired substrate. After removing the thermal release tape, washing off the PVP layer, and etching the gold with a mild etchant solution (I2/I), a monolayer with macroscopic dimension (typically millimeters to centimeters) is obtained. In addition to increasing the lateral sizes and monolayer yields by two to three orders of magnitude, our gold tape exfoliation turns the stochastic Scotch tape method into a deterministic and quantitative process. Although gold is used in the present study, one can extend the method to other metals provided the etching chemistry does not damage the 2D monolayer of interest.

Fig. 1 Schematic illustration of the layer-by-layer exfoliation procedure of bulk vdW single crystals.

(A) Method: (1) Depositing gold on an ultraflat silicon wafer; (2) spin-coating the surface with a layer of PVP; (3) using thermal release tape (TRT) to pick up the PVP and gold; (4) pressing the ultraflat gold onto the surface of a bulk vdW crystal; (5) peeling off a monolayer and transferring onto a substrate; (6) removing the TRT with heat; (7) dissolving PVP in water; (8) dissolving gold in an I2/I etchant solution; and (9) obtaining the single-crystal monolayer with macroscopic dimensions. (B) Schematic of the layer-by-layer exfoliation technique to yield even and odd layers from an AB-stacked vdW crystal. (C) Optical images of six monolayer samples (on SiO2/Si substrate) sequentially exfoliated from a centimeter-size WSe2 single crystal shown at the upper left corner.

The strong adhesion of gold to TMDCs can repeatedly generate complete single-crystal monolayers, each adopting the shape of the entire surface of a bulk crystal (Fig. 1B), as illustrated for six WSe2 monolayers in Fig. 1C. The yield of exfoliation, expressed as the percentage of monolayer area picked up from the contacted bulk crystal surface, is close to unity. This technique can be applied to a broad range of vdW crystals, as we demonstrate here for single-crystal TMDC monolayers (WS2, MoS2, WSe2, MoSe2, and ReS2) on various substrates such as SiO2/Si, fused silica, and sapphire. Adhesion of the ultraflat gold to graphene and hexagonal boron nitride (h-BN) is weaker and the monolayers are smaller than those for TMDCs, but still much larger than those obtained by conventional Scotch tape exfoliation. Optical images of the macroscopic TMDC monolayers are shown in fig. S1. Each TMDC sample is dominated by the single-crystal monolayer, sometimes with small regions of multilayers near the edges resulting from the contact of the gold tape with imperfect or nonflat regions on the edges of TMDC bulk crystal.

The quality of the large single-crystal monolayer is comparable to, or slightly better than, that of microscopic single-crystal monolayers produced using the traditional Scotch tape method, as evidenced in the clean surface characterized by atomic force microscopy (AFM) and the crystal quality characterized by photoluminescence (PL) spectroscopy. An AFM image of the TMD monolayer flakes reveals atomic scale flatness, as shown in fig. S2 for an MoS2 monolayer. PL spectroscopy is particularly sensitive to defects and disorder because increased peak width is evidence of inhomogeneous broadening (disorder) and reduced intensity is indicative of defect-mediated nonradiative recombination. We compare in Fig. 2 the low-temperature PL spectra of the macroscopic TMDC monolayers from our gold tape method and microscopic monolayers from the commonly used Scotch tape method, with all monolayers encapsulated in h-BN to eliminate the effect from substrate defects and inhomogeneity. The MoSe2 and WSe2 crystals are from flux growth (22), MoS2 from natural crystals (SPI Supplies), and WS2 from chemical vapor transport growth (HQ Graphene). Note that the highest-quality MoSe2 and WSe2 crystals from flux growth (22) are all of small sizes (~100 μm), and we found that defect density in macroscopic-sized crystals (≥ 1 mm) from flux growth typically is higher than those in the smaller sizes. The PL spectra of MoS2 and MoSe2 from the two methods are nearly identical. The former is characterized by the dominant A exciton and the latter shows the A exciton and the trion peaks, both nearly independent of the two exfoliation methods. PL image histogram analysis (Fig. 2) shows that the PL intensities from our macroscopic monolayers are consistent with or slightly higher than those from the Scotch tape method. For WS2 and WSe2 monolayers, the PL spectra from the two methods are similarly complex because of the presence of dark excitons and many-body states (23, 24); however, the overall PL intensity in our macroscopic monolayers is still comparable to or slightly higher than those from the Scotch tape method. These results confirm the high quality of the macroscopic TMDC monolayers.

Fig. 2 Comparison of PL spectra and intensity distributions in TMDC monolayers exfoliated using the gold tape and Scotch tape methods.

PL spectra (top) and intensity distributions (bottom) for monolayers encapsulated in BN. Black indicates gold tape–exfoliated monolayers; red is traditional Scotch tape–exfoliated monolayers. MoSe2 and WSe2 monolayers from both methods are exfoliated from the same low–defect-density bulk crystals grown from the flux methods. The MoS2 and WS2 monolayers in both methods are exfoliated from the same chemical vapor transport–grown bulk crystals. All PL measurements were at 4 K on monolayer samples with BN encapsulation. arb, arbitrary units.

Obtaining macroscopic single-crystal monolayers of 2D vdW crystals with high throughput opens the door to a broad range of applications spanning from spectroscopy to scalable devices and to the easy assembly of artificial lattices. For example, the macroscopic size of the TMDC monolayers prepared on transparent substrates allows us to use a conventional ultraviolet-visible spectrometer to easily obtain optical absorption spectra of TMDC monolayers (fig. S3), each featuring the well-known A and B excitons (25). Furthermore, the effective disassembly of the bulk crystal into individual single-crystal monolayers with defined crystal orientation allows us to reassemble them into artificial vdW crystals with desired properties. Specifically, we show that the ultraflat gold tape can be used as an effective pickup tool to reassemble higher-order vdW lattices from the macroscopic single crystals. We demonstrate two examples: (i) the reassembly of macroscopic TMDC monolayers into an artificial crystal lattice with AA stacking for greatly enhanced nonlinear optical response and (ii) the formation of a macroscopic heterobilayer from two distinct TMDC monolayers.

The first example targets effective engineering of nonlinear optical properties in an ultrathin material. TMDC monolayers, with intrinsically broken inversion symmetry, are known to have ultrastrong, nonlinear susceptibilities, as reflected in their intense second harmonic generation (SHG) (2628). However, TMDC bulk crystals exhibit 2H centrosymmetry; the crystal orientations of neighboring layers are 180° counteraligned with each other in the so-called AB stacking. Because of cancellation between counteraligned layers, the SHG response is smaller in few-layer samples, becoming asymptotically negligible for the bulk crystal (2628). Indeed, interference in SHG from individual monolayers in stacked homo- or heterobilayers is very sensitive to the alignment angle between the two layers (29). We disassembled the TMDC bulk crystal into individual monolayers and reassembled the even (or odd) monolayers into AA artificial crystals. Figure 3A shows angle-resolved SHG responses from one to five layers of an AA-stacked MoSe2 artificial lattice, with the integrated SHG intensities (ISHG; circles) plotted against the number (n) of the monolayers in Fig. 3B. Similar results for AA-stacked MoS2 are shown in fig. S4. The dashed line in Fig. 3B corresponds to a fit that takes into account reabsorption and interference of SHG from different layers. At ultrathin thickness, the SHG response is dominated by constructive interference from adjacent layers, close to the ideal limit of a quadratic optical response, ISHGn2 (solid curve in Fig. 3B). The close to perfect coherent enhancement benefits from the negligible phase mismatch over nanometer distances. With increasing thickness of the artificial lattice, the increase of coherent response is expected to continue until ~20 nm, when phase mismatch and reabsorption of SHG light become noticeable, as shown in fig. S5. Although synthesis of 3R phases of TMDC with broken inversion symmetry is possible (30), our method is unrestricted by synthesis and can be used to construct any macroscopic vdW multilayer structures with control in interlayer twist angles and chemical identities. As one indication of the consistency of the artificial lattices from our macroscopic monolayers, we measured SHG responses from the MoSe2/WSe2 at five randomly picked spots (fig. S6). The orientation of the artificial lattice is unchanged from spot to spot within the experimental angular resolution of ±0.5°. The SHG intensity variation is within ±15%, likely because of the small changes in optical alignment as the sample was moved macroscopically under the microscope.

Fig. 3 Artificial AA-stacked TMDC lattices from macroscopic monolayers.

(A) Angle-resolved SHG intensity of AA-stacked MoSe2 artificial lattices as a function of the rotation angle of crystal with respect to light polarization. (B) Integrated SHG intensity (circles) for different number of layers in the AA stacks. The solid line is a quadratic fit, and the red dashed line is a fit that takes into account both coherent interference and reabsorption, as discussed in the supplementary materials. We fabricated each multilayer sample by picking up the monolayers on SiO2 substrates sequentially with the ultraflat gold tape and measuring the SHG on the tape after each step.

The second example is the creation of heterobilayers with controlled twist angle from two macroscopic single-crystal monolayers. By precisely engineering angular and/or lattice mismatch, vdW bilayers constructed from the same or different single-crystal 2D monolayers have been shown to exhibit a range of quantum phenomena, but all on micrometer-scale samples from the Scotch tape method (412). An intriguing development for their future technological applications is the generation of such bilayer structures at macroscopic dimensions. Using two monolayers of MoSe2 and WSe2, we fabricated a MoSe2/WSe2 single-crystal heterobilayer, as shown in the optical image in Fig. 4A, with lateral dimensions of ~4 mm and a twist angle of Δθ = 3.0 ± 0.5°. AFM imaging on part of the macroscopic structure identifies the heterobilayer with a high degree of flatness (fig. S7). The large size of the single-crystal heterobilayer on a dielectric substrate (SiO2) allows us to map out the band structure with angle-resolved photoemission spectroscopy (ARPES) using a conventional setup with a hemispherical electron energy analyzer without microscopic capabilities. The ARPES spectrum determined in the Γ–K direction is in excellent agreement with theoretical calculations (Fig. 4B) (31). Low-temperature PL measurement of the BN-encapsulated heterobilayer stack revealed the dominant radiative recombination from the charge-separated interlayer exciton (Fig. 4C), in excellent agreement with previous reports of interlayer excitons in MoSe2/WSe2 heterobilayers fabricated from the Scotch tape method (3234). Moreover, the intralayer excitons from constituent monolayers in our heterobilayer sample are completely quenched, verifying the high quality of the MoSe2/WSe2 interface. Note that the detailed peak shape and intensity of interlayer exciton PL vary from location to location on the sample (figs. S8 to S10), which is well known from recent reports on the same system (8, 9), and is likely the result of sensitivity of PL emission to local variation in electrostatic environment and strain (35).

Fig. 4 Macroscopic MoSe2/WSe2 heterobilayer.

(A) Optical image of the millimeter-scale MoSe2/WSe2 heterostructure on SiO2/Si substrate aligned at Δθ = 3.0 ± 0.5°, as determined in SHG. (B) ARPES of the millimeter-scale MoSe2/WSe2 heterobilayer measured along the Γ–K direction. The dotted curves are theoretical calculations from (31). The sample was at 295 K in ARPES measurement. (C) Low-temperature (4 K) PL spectra of interlayer exciton in h-BN–encapsulated MoSe2/WSe2 heterobilayer (HB, black). For comparison, we also show PL spectra from intralayer excitons in h-BN–encapsulated MoSe2 (blue) and WSe2 (red) monolayers. All monolayers were exfoliated using the gold tape method. For the macroscopic heterobilayer on oxide-terminated silicon in (A) and (B), we used commercial chemical vapor transport–grown MoSe2 and WSe2 crystals. For the h-BN–encapsulated samples in (C), we used higher-quality, but smaller, flux-grown MoSe2 and WSe2 crystals.

To summarize, we show a general method for the facile disassembly of vdW single crystals layer by layer into monolayers with macroscopic dimensions. The quality of the macroscopic single-crystal monolayers is comparable to that of the microscopic dimensions obtained from the state-of-the-art Scotch tape method, as confirmed in AFM imaging and PL characterization. We demonstrate this method for the controlled reassembly of these macroscopic monolayers into artificial lattices, including AA-stacked TMDC multilayers for substantially enhanced nonlinear optical response and heterobilayers for interlayer excitons at macroscopic dimensions. This approach may allow us to extend the exciting discoveries in the so-called “twistronics,” i.e., magic angle or Moiré landscapes in 2D bilayers (412), into the multilayer or bulk region. The latter is a formidable challenge for microscopic monolayers from the Scotch tape method, but achievable with our macroscopic monolayers. With techniques for the high-throughput production of macroscopic monolayers, 2D quantum devices on a large scale may become a reality.

Supplementary Materials

Materials and Methods

Figs. S1 to S10

References (3638)

References and Notes

Acknowledgments: Funding: The development of the ultraflat gold tape exfoliation method was supported by the Center for Precision Assembly of Superstratic and Superatomic Solids, a Materials Science and Engineering Research Center (MRSEC), through NSF grant no. DMR-1420634. All TMDC sample preparation, imaging, and spectroscopy experiments were supported by National Science Foundation (NSF) grant no. DMR-1809680. X.-Y.Z. acknowledges support by a Vannevar Bush Faculty Fellowship through Office of Naval Research grant no. N00014-18-1-2080 for the purchase of the laser system used in the ARPES measurements. F.L. acknowledges support by a Department of Energy (DOE) Office of Energy Efficiency and Renewable Energy (EERE) Postdoctoral Research Award under the EERE Solar Energy Technologies Office administered by the Oak Ridge Institute for Science and Education (ORISE). ORISE is managed by Oak Ridge Associated Universities (ORAU) under DOE contract no. DE-SC00014664. All opinions expressed in this paper are the authors’ and do not necessarily reflect the policies and views of DOE, ORAU, or ORISE. Author contributions: X.-Y.Z. and F.L. conceived this work. F.L. performed all experiments on gold tape exfoliation and characterization. W.W. performed experiments on Scotch tape exfoliation and BN encapsulation. Y.B. and F.L. measured the low-temperature PL. J.H. and S.H.C. conceived the idea of using a PVP protection layer. Q.L. assisted with gold tape exfoliation and SHG measurement. J.W. and Y.B. constructed the low-temperature PL-mapping setup. F.L. and X.-Y.Z. wrote the manuscript with input from all authors. X.-Y.Z. supervised the project. All authors participated in the discussion and interpretation of the results. Competing interests: Part of this work is included in a US provisional patent application. Data and materials availability: All data needed to evaluate the conclusions in the paper are available in the main text or the supplementary materials.

Stay Connected to Science

Navigate This Article