Achieving Large, Tunable Strain in Monolayer Transition-Metal Dichalcogenides
A. Dadgar,D. Scullion,K. Kang,D. Esposito,E. Yang,I. Herman,M. Pimenta,E. Santos,A. Pasupathy
DOI: https://doi.org/10.1021/acs.chemmater.8b01672
2017-05-15
Abstract:We describe a facile technique based on polymer encapsulation to apply several percent controllable strains to monolayer and few-layer Transition Metal Dichalcogenides (TMDs). We use this technique to study the lattice response to strain via polarized Raman spectroscopy in monolayer WSe2 and WS2. The application of strain causes mode-dependent redshifts, with larger shift rates observed for in-plane modes. We observe a splitting of the degeneracy of the in-plane E modes in both materials and measure the Grüneisen parameters. At large strain, we observe that the reduction of crystal symmetry can lead to a change in the polarization response of the A mode in WS2. While both WSe2 and WS2 exhibit similar qualitative changes in the phonon structure with strain, we observe much larger changes in mode positions and intensities with strain in WS2. These differences can be explained simply by the degree of iconicity of the metal-chalcogen bond. One of the iconic characteristics of monolayer 2D materials is their incredible stretchability which allows them to be subjected to several percent strain before yielding [1]. The application of moderate (~1%) strains is expected to change the anharmonicity of interatomic potentials [2, 3], phonon frequencies [4, 5] and effective masses [6, 7]. At larger strains, topological electronic[8] [9] and semiconductor-metal structural phase changes have been predicted [10-13]. Important technological applications such as piezoelectricity can be explored by the application of systematic strain [14, 15]. One of the chief problems in achieving reproducible strain is the intrinsic nature of 2D materials as single layer sheets they need to be held to a flexible substrate which is then stretched or compressed. Previous experiments [16-19] have used flexible polymers as substrates and metal or polymer caps in order to constrain the 2D material. Using these techniques, approximate strains up to 4% have been reported so far in the literature, but independent verification of the applied strain has been lacking. Achieving large reproducible strains in engineered geometries will allow us to probe these exciting properties of individual 2D materials and their heterostructures [4, 17, 20-26]. In this work, we develop a new strain platform to apply large range accurate uniaxial tensile strains on monolayer and fewlayer materials. One of our chief innovations is the development of a novel polymer-based encapsulation method to enable the application of large strain to 2D materials. Here, we apply this technique to study the strain-dependent properties of monolayer WSe2 and WS2 grown by Chemical Vapor Deposition (CVD) on SiO2/Si substrates [27-29]. We use cellulose acetate butyrate (CAB) to lift the monolayers from the SiO2/Si substrates and transfer to polycarbonate substrates. The two polymers are then bonded to produce encapsulated monolayers and multilayers. The key to achieving good bonding is perfect control over the temperature, time and pressure during the bonding process. Additionally, polymer layers that are in the amorphous phase cause nonlinear strain-deflection behavior which is not desirable in our experiments. To resolve this issue, we crystallize the polymer stacks by annealing near the glass transition temperature followed by slow cooling. The crystallized polymers are fully flexible, elastic and springy substrates as shown in Fig. 1(a). After all of our processing steps, we find that the polymer stacks enter into the plastic regime at 7% strain. We find that strains up to this value are perfectly transferred to the encapsulated 2D material as described below. Our strain method adopts the extra-neutral axis bending technique – Fig. 1(b) in which areas above the neutral axis undergo tensile strain while those below the axis experience compressive strain. In our method, we use a screw-driven vertical translation stage to apply strain to the polymer stacks. We solve the Euler-Bernoulli equation for our geometry in order to achieve an accurate relation between the vertical displacement δ of the translation stage and the strain ε of the 2D material. For a fully isotropic, linear and elastic material, the strain-displacement relation is derived as: ε = 3tδ a(3b + 2a) ⁄ where t is the substrate thickness, b and a are center support and cantilever lengths respectively. In our experiments, the use of a fine adjustment screw gives us a resolution of 0.05% strain for 0.5 mm substrates, with essentially no limit to the maximum strain that can be applied. More details are provided in the Supplemental Material. Shown in Fig. 1(c) is an optical image of triangular flakes of WSe2 encapsulated by this process. We adjust the CVD process to produce triangular flakes in order to easily identify the crystallographic directions of the grown monolayers. Since the strains achievable in our experiments are large, we can directly verify from optical measurements that the strain being applied to 2D layer is the calculated value. This is illustrated in Fig. 1(d). Each of these images is obtained by overlaying two images, one at zero strain and one at a fixed value of strain (4.2% and 6.5% respectively). Only the edges of the triangles are shown in the images, which are lined up to be at the same vertical height at the top vertex of the triangle. We can directly see by inspection that the length of the triangle along the strain direction is larger when strained as one expects. A pixel-height measurement of the edge-detected images gives us a direct experimental measure of the applied strain, which can be compared to the calculated strain based on the screw displacement. It is found that the two measurements match within 0.1% absolute strain. Thus, our technique allows for the application of uniform, highly repeatable and independently measurable strain on TMD monolayers and heterostructures. In order to probe the effects of strain on our samples, we choose to characterize with Raman spectroscopy a simple yet powerful way to measure lattice properties and their coupling to the electronic degrees of freedom. Strains were applied in both zigzag and armchair directions (Y and X axes in Fig. 1(e) ) in our experiments. Our Raman setup with 532 nm excitation wavelength is shown in Fig. 1(f). The measurements were performed while controlling for the incident light’s polarization (Ei) direction (θ in Fig. 1(e) ). For each experiment, Raman spectra were collected in both the parallel(Es || Ei) and crosspolarized (Es ⊥ Ei) detector geometries, shown with standard notations Z(YY)Z and Z(YX)Z respectively. In our experiments, we found no dependence of the Raman spectra on the angle of incidence relative to the crystallographic axis at zero strain. We therefore fix our incidence angle to the Y direction, and measure the unpolarized, parallel-polarized and cross-polarized Raman spectra at each value of strain which is applied in the X direction. We first discuss the properties of monolayer WSe2. Shown in Fig. 2(a) are a sequence of spectra taken at different values of strain in the unpolarized, parallel and cross polarization geometries. Previous Raman spectroscopy measurements performed on monolayer WSe2 have identified three vibrational modes [30-32] termed A, E and 2LA. A is an out-of-plane phonon mode in which the top and bottom chalcogen atoms vibrate in opposing directions; while E is in-plane mode where the metal atoms vibrate out-of-phase with the chalcogen atoms [33]. The 2LA mode results from a double resonance process involving two phonons from the LA branch. Second order processes can in general give rise to a complex lineshape in the Raman spectrum; yet, in the case of WSe2 we find that a single Lorentzian can be used to model well the 2LA mode lineshape. Although Aand E modes are nearly degenerate, they can be distinguished from each other by polarization dependency of their intensities. The out of plane, symmetric A mode disappears due to its symmetry in the cross polarization geometry, leaving behind only the E mode. Our spectra in the cross-polarization geometry can thus be modeled well as the sum of two Lorentzian peaks corresponding to E and 2LA modes. Information of the E mode position can then be used to fit the spectra seen in the parallel polarization geometry in order to extract the nearlyoverlapping A mode position. Having understood the polarization-dependent Raman spectra of unstrained monolayer WSe2, we apply uniaxial strains and measure the Raman response. The effects of uniaxial strain up to 1% on monolayer WSe2 has previously been experimentally investigated via unpolarized Raman [17] and absorption spectroscopy [34]. Raman spectra under increasing uniaxial strain up to 3% are shown in Fig. 2(a). A close examination of spectral lineshapes in the cross polarization geometry shows that the E mode becomes broader with increasing strain. In general, we expect that the initially doubly degenerate E mode splits on the application of strain into E and E. The displacement eigenvector of the E mode is orthogonal to the direction of strain, while it is parallel for the Emode, as has previously been observed for MoS2 and graphene [3, 16, 21]. While we cannot observe a complete separation of the E and E modes in our data, it is nevertheless straightforward to fit the lineshape to two Lorentzian functions and extract the splitting as a function of strain, as shown in Fig. 2(e). The splitting of the E mode under tensile strain due to the anharmonictiy of molecular potentials can be described by Grüneisen parameter γ = (|∆ωE′+| + |∆ωE′−|) 2ωE′(1 − v) ⁄ and the shear deformation potential β = (||∆ωE′+| − |∆ωE′−||) 2ωE′(1 + v) ⁄ where ωE′ is the frequency of E mode, ∆ωE′+ and ∆ωE′− are the frequency shifts of split modes per unit percent strain and v is Poisson’s ratio which is 0.27 for our substrates. We obtain values of γ = 0.38 , β = 0.10 for WSe2 which are smaller than those reported for graphene [2, 3]. Using t