Correlative imaging of 2D material heterostructures: Raman, Photoluminescence (PL) and Second Harmonic Generation (SHG)

Two-dimensional (2D) materials with remarkable electronic, optical, mechanical, and chemical properties attract attention in fundamental scientific research due to their great potential for industrial application. In the past decade, numerous papers have shown that Raman spectroscopy and imaging can reveal unique characteristics of 2D materials such as electronic and optical properties that differ significantly from those of their respective bulk precursors [1-5]. For example, bandgaps of semiconducting transition metal dichalcogenides (TMDs) can be tuned by changing the number of layers, the chemical composition, or the strain of the materials.

The focus nowadays is on exploring and designing new materials by employing heterostructures consisting of stacks of layered materials. Such heterostructures can be tuned based on their composition and stacking, thus combining the advantages of different 2D material properties to designable quantum phenomena. Interlayer effects within the heterostructure can also give rise to new quantum phenomena that are not present within the individual layers, creating new fields of study for fundamental materials science research [2].

Raman spectroscopic imaging remains a fast, non-destructive and convenient technique for characterising the properties of 2D materials at the laboratory level and mass-production scales. Raman imaging is used to check the quality of 2D TMD materials in terms of number of layers, uniformity, and defects [1-5]. Resonant Raman effects provide information about the optoelectronic properties and doping/vacancies of the TMD-materials. To excite resonant Raman effects in TMDs, specific excitation wavelengths are required. Furthermore, the properties of the electronic band structure can be analysed using luminescence measurements. Grain boundaries can be identified by second harmonic generation. Ideally, such measurements are conducted with full polarisation control, as the properties of 2D materials naturally are anisotropic.

The witec360 confocal Raman imaging microscope with its fundamentally modular design not only allows for the acquisition of single Raman or photoluminescence (PL) spectra, but also confocal images of the highest resolution. These lead to a comprehensive understanding of the phenomena occurring in 2D TMD materials. The edges of 2D crystals often show different optoelectronic properties than the crystal and here the combination of Raman, PL, and second harmonic generation (SHG) imaging can provide detailed information such as chirality, orientation, and symmetry. To enable all of these different Raman, PL, and SHG measurements to be performed with one instrument, a spectrograph with a large variety of gratings, such as the Hexalight spectrometer, is beneficial.

The aim of this application note is to highlight the benefits of the witec360 microscope with the Hexalight spectrometer by presenting a series of Raman/PL/SHG measurements performed on a heterostructure deposited on a Si/SiO₂ substrate. The sample consists of triangular crystals of MoS₂ and WS₂ covered with a graphene sheet.

Results and Discussions

Fig. 1a shows a white light image of the analysed heterostructure (graphene/WS₂/MoS₂). The frame marks the area examined with Raman, PL, and SHG. The crosses mark the positions where Raman spectra were acquired with different excitation wavelengths for MoS₂ and WS₂ (Fig. 1b, c respectively). For both TMDs the in-plane (E_2g) and out-of-plane (A_1g) vibrational modes are highlighted. These Raman bands change slightly to lower wavenumbers as the excitation wavelength is increased. For MoS₂ (Fig. 1b) additional Raman bands arise at 451-460 cm^-1 when excited with 633 nm. These bands are related to strong electron-phonon coupling, leading to resonance enhancement involving longitudinal acoustic phonons at the M point of the Brillouin zone (2LA(M) modes) when the laser energy is close to the material´s exciton states. For WS₂ the 2LA(M) mode arises at 350 cm^-1 when excited at 532 nm and 633 nm.

Fig. 1: (a) Bright-field image of the analysed heterostructure. The frame marks the position of the Raman/PL/SHG images. The Raman spectra of MoS₂ (b) were acquired from the red and the WS₂ spectra (c) from the yellow marked positions, respectively, using different excitation wavelengths. Spectra acquisition parameters: integration time: 0.5 s, 10 accumulations, spectrograph setting: 1200 g/mm.

Characterisation of the graphene top-layer

Differences are observed in the Raman spectra of the top-layer graphene due to the underlying heterostructure. Fig. 2 shows the colour-coded Raman image of the graphene layer covering the analysed heterostructure (a) and the corresponding Raman spectra (b). To acquire the Raman spectra of the graphene, the 300 g/mm grating is required in combination with the 600 mm Hexalight spectrometer and TrueComponent Analysis (TCA) was used to identify the different regions.

Fig. 2: False colour-coded Raman image of the graphene layer (a) and Raman spectra evaluated using TCA (b). The smaller images at the bottom show the corresponding distributions used for the overlay image. Measurement parameters: λ_ex: 488 nm, power: 1.0 mW, objective: 100x/0.9, area: 50x40 µm², 150x120 pixel, integration time: 100 ms/pixel.

The graphene-specific Raman bands are the D-band at 1356 cm^-1, the G-band at 1590 cm^-1, the D´-band at 1623 cm^-1, and the 2D-band at 2704 cm^-1. All spectra in Fig. 2 show an intense D-band and the presence of the D´-band, indicating a defect-rich graphene layer on top of the TMDs. Lorentzian fits were performed for all four characteristic Raman bands of graphene, and characteristic intensity ratios (areas under the peaks) were calculated for the analysed sample area.

These are summarised in Table 1:

• The I_D/I_G ratio provides information about the defect density in the graphene layer and is inversely proportional to the crystallite size [6].
• The I_D/I_D´ ratio is sensitive to the type of disorder present, making it particularly relevant for graphene-based composites that often exhibit diverse defect chemistries due to functionalisation, edge exposure, or matrix interactions. According to [7] a value of 3.4 corresponds to boundary-like defects (e.g., grain edges).

With a value of 3.3 for the I_2D/I_G ratio, the graphene layer can be considered a monolayer with a high defect density.

Table 1: Summary of the results of the Lorentzian fit.

Raman charactersation of the TMDs

Using a high-resolution grating (2400 g/mm), the properties of the WS₂/MoS₂ triangles of the heterostructure are analysed in more detail. With this grating, Raman images of the highest spectral resolution are obtained. Fig. 3a shows a Raman image of the same sample area as in Fig. 2a, measured with 488 nm excitation. The characteristic Raman bands of MoS₂ and WS₂ are shown in red and blue in Fig. 3a, b. For both TMDs the in-plane (E_2g) and out-of-plane (A_1g) vibrational modes are highlighted. The Raman image shows that the heterostructure consists of two triangles of MoS₂ (blue) and an irregular flake of WS₂ (red). The purple colour reveals overlaying MoS₂ and WS₂ monolayers, whereas cyan indicates WS₂ bilayer. This result was obtained from TCA.

The spectra were also analysed using Lorentzian peak fitting of the in-plane (E_2g) and out-of-plane (A_1g) vibrational modes of MoS₂ (Fig. 3c, d). The images and histograms show blue shifts of the MoS₂ vibrational modes from pristine to overlayed with WS₂. The (E_2g) Raman band provides information on the in-plane strain, while the out-of-plane mode (A_1g) gives evidence of the quality of the interfacial contact in the heterostructure. Furthermore, these images show that the two MoS₂ triangles interact differently with the WS₂ layer, leading to a higher blue shift of the A_1g and E_2g Raman bands in the upper left triangle compared to the lower MoS₂ triangle. This shows that the orientation of the two TMD crystals strongly influences the Raman signal, and thus the optoelectronic properties. This finding is more evident using advanced peak fitting.

Fig. 3: False colour-coded Raman image of the TMD heterostructure (a) and Raman spectra evaluated using TCA (b). Raman images and histograms of MoS₂ triangles using Lorentzian fits for the A_1g (c) and E_2g (d) Raman bands. Measurement parameters: λ_ex: 488 nm, power: 1.0 mW, 100x/0.9 objective, 50x40 µm², 150x120 pixels, 100 ms/pixel.

Fig. 4a shows the PL spectra of graphene/MoS₂, graphene/WS₂ and graphene/WS₂/MoS₂ heterostructures. The PL in TMDs is dominated by the recombination of electrons in the conduction band with holes in the spin-orbit split valence bands. There are two distinct emission features: the ground state A exciton (A-exciton) between the conduction and upper valence band, and the higher energy B exciton between the conduction and lower valence band. The strongest PL is measured for the graphene/WS₂ heterostructure with a peak at 1.98 eV, characteristic for A-excitons. The graphene/MoS₂ heterostructure shows two broad PL peaks at 1.87 eV and 2.01 eV corresponding to the A and B excitons. The PL spectrum of the graphene/WS₂/MoS₂ heterostructure (blue) shows a shift towards lower energies of the A-excitons of the studied TMDs. The PL image of the graphene/WS₂/MoS₂ heterostructure shows a different exciton energy of 1.94 eV, thus indicating a structure with different optoelectronic properties. This PL image was obtained using a pseudo-Voigt function to fit the complete array of PL spectra.

Fig. 4: PL spectra (a) and PL image of the heterostructure (b) colours represent PL peak positions found from pseudo-Voigt fit of each PL spectra. Measurement parameters: λ_ex: 488 nm, power: 1.0 mW, 100x/0.9 objective, 50x40 µm², 150x120 pixels, 100 ms/pixel.

Further information regarding properties of the heterostructure can be obtained by using laser energies which generate additional Raman resonances. The 532 nm and 633 nm laser excitation generate a second-order Raman resonance (R) in WS₂ involving the longitudinal acoustic mode LA(M). This resonance results from a coupling between the electronic band structure and lattice vibrations and is used to study the phonon-exciton interaction [8]. Fig. 5 shows the Raman image and changes in the Raman spectra which are unique to the 532 nm excitation wavelength. As expected, the monolayer of WS₂ shows the highest LA(M) intensity [8]. In the regions of the image, where WS₂ and MoS₂ are overlapping, the LA(M) mode of WS₂ reveals small differences which can be visualised by applying the TCA to the array of spectra. This resonant mode is extremely sensitive to the intralayer coupling, revealing also the defects in graphene covering the WS₂ structure.

Fig. 5: False colour-coded Raman image of the TMD at 532 nm excitation (a) and Raman spectra evaluated using TCA (b). Measurement parameters: λ_ex: 532 nm, power: 2.0 mW, 100x/0.9 objective, 50x45 µm², 150x135 pixels, 100 ms/pixel.

At an excitation energy of 633 nm, both MoS₂ and WS₂ show the presence of the LA(M) modes. Fig. 6 shows the Raman image and Raman spectra acquired at this excitation wavelength. For MoS₂, the LA(M) modes appear as a broad Raman band in the range 420-470 rel. cm^-1. The strongest signal is observed for the MoS₂ monolayer (blue colour). In the areas where MoS₂ and WS₂ are overlapping, this resonant mode for MoS₂ decreases, but reveals defects in the triple-graphene/WS₂/MoS₂ heterostructure region (magenta component). The LA(M) modes of WS₂ show a shift towards higher wavenumbers at the edge of the crystal (green component), revealing defects in the WS₂ crystal or the overlaying graphene sheet.

Fig. 6: False colour-coded Raman image of the TMD at 633 nm excitation (a) and Raman spectra evaluated using TCA (b). Measurement parameters: λ_ex: 633 nm, power: 30 mW, 100x/0.9 objective, 50x45 µm², 150x135 pixels, 200 ms/pixel.

Finally, SHG measurements were performed to examine the heterogeneity of the WS₂ monolayer. This imaging technique uses a non-linear optical effect in which two photons with the same frequency interact with a material, are combined, and generate a photon with twice the energy of the initial photons. SHG is sensitive to changes in crystal orientation and symmetry. It can visualise grain boundaries and reveal strain fields by rotating the polarisation direction of the excitation light. Fig. 7 shows SHG images acquired at 0^o (a), 30^o (b) and 60^o (c) rotation of the polarisation of the excitation light with respect to the x-axis of the images. The images show that the irregular WS₂ sheet consists of several triangular crystals with different crystallographic orientation. The polarisation curves (Fig. 7d) measured at the marked positions show the rotation angle of the irregular WS₂ crystals. Fig. 7e is an overlay of the defect Raman image (from Fig. 6 green colour) with the SHG image (Fig. 7a), highlighting the sensitivity of the resonant Raman band of WS₂ to grain boundaries and defects. Fig. 7f is an overlay of the colour-coded Raman image from Fig. 6 measured at 633 nm excitation with the SHG image (Fig. 7c). It shows the sensitivity of the Raman image to crystal orientation, strain and defects.

Fig. 7: SHG images of the heterostructure at different polarisation angles of the incoming light: 0^o (a), 30^o (b) 60^o (c). Polarisation curves (d) were measured at the colour-coded marked positions in (c). Overlay of defects (e) and Raman (f) with SHG images. Measurement parameters: λ_ex: 1064 nm, power: 30 mW, 100x/0.9 objective, 50x45 µm², 150x135 pixels, 100 ms/pixel.

Summary

A heterostructure consisting of triangles of WS₂/MoS₂ covered with graphene was analysed with the witec360 confocal Raman imaging microscope. Correlative measurements were performed using different excitation lasers and spectral gratings to highlight the influence of the excitation laser on the Raman measurements. To characterise the graphene layer a grating covering the entire spectral range was used, whereas for the characterisation of the TMDs a high-resolution grating with the proper blaze angle was selected. PL measurements on the other hand require a low-resolution grating to cover the broad luminescence range. The Hexalight spectrometer with its six-grating capability enables optimised spectroscopic conditions for all the laser wavelengths used. The excitation laser was selected to encompass resonant Raman scattering across both TMDs. The measurements presented show that resonant Raman scattering amplified first order and multi-phonon lattice vibrations under excitation energies tuned to exciton transitions. Resonant confocal Raman imaging offers a high-resolution, non-destructive method to probe phonon dispersion, electron-phonon coupling, and structural characteristics of heterostructures of 2D TMD materials. Polarisation-dependent SHG measurements complete the correlative measurements, revealing that the WS₂ sheet consists of several triangular grains rotated in relation to each other.

Experimental Details

The witec360 confocal Raman imaging microscope was equipped with 4 lasers with excitation wavelengths of 488 nm, 532 nm, and 633 nm for Raman measurements, and a 1064 nm pulsed laser for SHG experiments. A Hexalight spectrometer (600 mm focal length) with the optimal grating combination was used for the measurements presented in this application note. Further experimental details can be found in the figure captions.

References:

1. Xin Cong, Xue-Lu Liu, Miao-Ling Lin, and Ping-Heng Tan, npj 2D Materials and Applications (2020) 13.
2. S. Novoselov, A. Mishchenko, A. Carvalho, and A. H. Castro Neto, Science 353 (2016) No 6298.
3. Gatensby, N. McEvoy, K. Lee, T. Hallam, N. Berner, E. Rezvani, S. Winters, M. O´Brien, G. S. Duesberg, Applied Surface Science 297(2014) 139.
4. Kaplan, Y. Gong, K. Mills, V. Swaminathan, P. M. Ajayan, S. Shirodkar, and E. Kaxiras, 2D Mater. 3 (2016) 015005.
5. Schmidt, C. S. Bailey, J. Englert, E. Yalom, G. Ankonina, E. Pop, O. Hollricher, and T. Dieing, Spectroscopy 36 (2021) 23.
6. Tunistra and L. J. Koenig, J. Chem. Phys. 53 (1970) 1126.
7. Eckmann, A. Felten, A. Mishchenko, L. Britnell, R. Krupke, K.S. Novoselov, and C. Casiraghi, Nano Letters 12 (2012) 3925.
8. O´Brian, N. McEvoy, D. Hanlon, J. Coleman, and G. Duesberg, Scientific Reports 6 (2015) 1.