Atomically thin materials such as semiconducting transition metal diselenide materials, like molybdenum diselenide (MoSe2) and tungsten diselenide (WSe2), have received intensive interests in recent years due to their unique electronic structure, bandgap engineering, ambipolar behavior, and optical properties and have motivated investigations for the next-generation semiconducting electronic devices. In this work, we show a nondestructive method of characterizing the layer number of two-dimensional (2-D) MoSe2 and WSe2 including single- and few-layer materials by Raman spectroscopy. The related photoluminescence properties are also studied as a reference. Although Raman spectroscopy is a powerful tool for determining the layer number of 2-D materials such as graphene and molybdenum disulfide (MoS2), there have been difficulties in precisely characterizing the layer number for MoSe2 and WSe2 by Raman spectroscopy due to the uncertain shifts during the Raman measurement process and the lack of multiple separated Raman peaks in MoSe2 and WSe2 for referencing. We then compared the normalized Si peak with MoSe2 and WSe2 and successfully identified the layer number of MoSe2 and WSe2. Similar to graphene and MoS2, the sample layer number is found to modify their optical properties up to 4 layers.

1. Introduction

Because of their unique structure and exotic physical and mechanical properties, two-dimensional (2-D) materials have drawn tremendous interests since their discovery [15]. The introduction of mechanical exfoliation method has enabled us to tailor the thickness of bulk layered materials down to a single unit cell [1, 2, 6]. One particular group of 2-D materials, i.e., transition metal dichalcogenides (TMDCs) [712], has been extensively studied for their electrical and optical properties. This is because of the tunable electronic band structure with layer numbers [13] and temperature [14, 15] and more recently with heterostructure engineering leading to the discovery of superconductivity and exotic excitons [1619]. As an example, molybdenum disulfide (MoS2), one of the TMD materials, has been widely explored for its properties of Raman spectroscopy [20], photoluminescence (PL) [1012, 14], and magnetic field influence [21] and with applications of field-effect transistors [7, 22] and chemical sensors [23].

Recently, transition metal diselenides MoSe2 and WSe2 have become the new 2-D stars owing to the creation of moiré excitons in their heterostructures [17, 18]. They are both at the transition of direct- and indirect-bandgap electronic structures around few layers, which provides new opportunities to engineer their electrical and optical properties [14, 2427]. By sharing the same chalcogen atoms, their lattice constants also match well, which even allows for epitaxial growth of heterostructures [28, 29]. For these applications, a good understanding of their band structures and accurate thickness determination are in need. Despite the abundant explorations of their electrical conductance properties [30], optical properties [14], and even angle-resolved photoemission spectroscopy [24], a quantitative and nondestructive characterization method of characterizing the layer number of MoSe2 and WSe2 on silicon substrates is still lacking. Especially, due to the infeasibility of extracting multiple Raman peaks of MoSe2, which has only one Raman peak, and WSe2, which has two Raman peaks that are not fully separated, the relative Raman peak position could not be characterized like what has been done with MoS2 [15]. Recent studies have been using reflectance spectroscopy [31, 32] and absorbance spectroscopy [33] to characterize the layer number of 2-D transition metal dichalcogenides, but the limitation with some studies [31, 33] is that these 2-D materials are fabricated on the transparent substrate, and the additional fabrication step is needed to transfer 2-D materials to the target substrate for semiconducting applications. Moreover, a clear trend of reflectance spectroscopy and absorbance spectroscopy could be observed in 2-D materials with different layer numbers [3133], but there lacks a quantitative value to identify each layer number alone. All of this makes it infeasible for the 2-D materials’ semiconductor device applications. Our method of directly characterizing 2-D MoSe2 and WSe2 on the SiO2/Si substrate confirms this previously reported method’s [34] feasibility to a broader range of 2-D materials. It has the advantages of keeping the sample’s physical properties closest to pristine, providing quantitative values for precise layer number characterization, and nondestruction, thus providing direct industrial applications.

In this paper, we characterized the layer number of MoSe2 and WSe2 by Raman spectroscopy with referencing to the substrate silicon’s Raman peaks. Both materials possess increasing Raman peak intensity with increasing layer numbers until reaching 4 layers (L). We also found 1L MoSe2 and WSe2 flakes possess higher PL intensity than the few-layer flakes.

2. Materials and Methods

2.1. Materials and Procedures

MoSe2 and WSe2 bulk crystals (SPI Supplies) were used for mechanical exfoliation to obtain few-layer materials. The MoSe2 or WSe2 crystal was procured and exfoliated by using a scotch tape. The crystal of size 3 mm × 3 mm was used. After exfoliation on the scotch tape for ∼8 times, the crystal with scotch tape was pressed onto a clean SiO2 (285 nm)/Si substrate, and the end of a sharpie was used to abrade for 3 minutes. After removing the scotch tape, the flakes are left on the SiO2 (285 nm)/Si substrate. 285 nm SiO2 was used because it provides the best contrast for identifying the thin flakes under an optical microscope (Nikon Eclipse 150) (Figure 1(a)). The final MoSe2 or WSe2 flakes obtained in this way has a size from 2 μm to 15 μm. Each flake’s thickness was determined by the tapping mode of atomic force microscopy (AFM) (Bruker AFM) (Figure 1(b)). Further confirmation of the layer number is by the optical method described in this article.

2.2. Raman Optical Measurement

A micro-Raman spectrometer (RENISHAW InVia Raman Microscope system) was used to measure the Raman spectra (Figures 2-3) for studying peak shifts and photoluminescence (PL) (Figure 4(a)) of MoSe2 and WSe2. Laser with a wavelength of 633 nm was used. 1 mW laser power was used for Raman and PL measurement, to provide the best result without damaging the sample. Flakes from 1 layer to up to 15 layers were measured. The 100x objective with a numerical aperture of 0.90 of the Raman microscope system was used. For a more accurate characterization, we recharacterized the laser spot size by moving the laser across a sharp edge and fit Raman peak mapping’s intensity curve. The laser spot size obtained in this way is 0.46 μm, which is the smallest in all the objectives and could provide the best resolution from the spot on the sample of measurement interest. All the measurements were performed in air at room temperature.

3. Results and Discussion

3.1. Optical Microscopy and Atomic Force Microscopy

Figure 1(a) shows an optical microscopy image of a mechanically exfoliated 1- to 4-layer (L) MoSe2 flake. By choosing a Si substrate with a 285 nm thermal oxide SiO2, we can obtain a good optical contrast for different thicknesses (Figure 1(a)). The same preparation method is used for WSe2 flakes. Figure 1(b) shows an atomic force microscopy (AFM) landscape of this flake for confirming thickness, with the height profile along the dashed line shown in Figure 1(c). The tapping mode of AFM is used. A thickness of ∼1 nm is found from the measurements, which is greater than the theoretical value of 0.65 nm but agrees with previous reports of AFM contact mode measurements [14, 27, 35]. For other layer numbers, flakes in other locations are used and confirmed by AFM using the same methodology. The AFM method is not practical for most applications due to the difficult and time-consuming device characterization process, which limits its applications in semiconducting industries, and the tip of AFM has a potential of damaging the sample. However, the nondestructive optical method is universal for semiconducting industrial applications, which is a more convenient, straightforward, safe, and accurate method.

3.2. Raman Spectroscopy

Laser used in the Raman measurement was focused on the MoSe2 and WSe2 flakes using the 100x objective lens with a numerical aperture of 0.90 and 1 mW laser power. Laser with a wavelength of 633 nm is used to provide a comprehensive characterization. For MoSe2, A1g Raman peak is the only observable peak and is the most visible and studied peak. Figure 2(a) shows the Raman spectra of MoSe2 flakes measured by 633 nm laser from 1L to 5L. For each flake, two sharp peaks can be observed. The peak around 520 cm−1 (Figure 2(a)) comes from the Si substrate. Si-peak shift remains unchanged while the intensity changes due to the optical shielding from the different flakes that it is supporting. The peak around 243 cm−1 (Figure 2(a)) comes from the A1g mode of MoSe2, which is the most visible Raman mode for the study [35, 36]. The E2g mode around 300 cm−1 is also weakly excited. However, because our goal is to find an easy way to distinguish the layer thickness, we choose to focus on the strongest A1g Raman mode in this study. For both MoSe2 and WSe2, at least 3 samples for 1L–4L were measured. For each sample, we repeated the measurements 3 times. With this information, we calculated and obtained the error bars.

Figures 2(b)2(d) summarize the data for MoSe2 from 1L to up to 12L. We find (Figures 2(a) and 2(b)) that the A1g peak position shows a rapid increase as the layer number increases from 1L to 5L. This A1g mode with a blue shift means the vibration stiffens with the increasing sample layer number. For films of 5L or more layers, the frequencies of A1g mode converge to a constant value. In the meanwhile, the Si peak remains at the same position around 520.9 cm−1. Due to Raman measurement’s nature of drifting, usually another Raman peak is used and the difference between the two Raman peaks is used to characterize the layer number [13]. However, only one Raman peak is significant for MoSe2; thus, the substrate Si’s Raman peak which is measured simultaneously is used as reference to eliminate drifting during the measurements. The peak offsets are plotted in the main panel of Figure 2(b). When the layer number is up to 4, a monotonic decrease of the peak position difference is observed. Thus, our data show the capability of quantifying the layer number by the Raman peak offset between the A1g peak and Si peak, which is facile and nondestructive and can avoid contact contamination and damage from AFM measurements. Also given that AFM sometimes yields different thicknesses for very thin layers (<4L) and is sensitive to sample-substrate impurities [14, 27, 35], Raman spectroscopy is used to characterize the collective dynamics of layered 2-D materials. Finally, when the layer number is greater than 4, the Raman peak shift remains unchanged, which serves as the limit of thickness characterization. As a result, for 1L MoSe2, the Raman peak difference is larger than 278.5 cm−1; for 2L MoSe2 the Raman peak difference is larger than 278 cm−1; and for 3L MoSe2 the Raman peak difference is larger than 277.5 cm−1.

Figure 2(c) shows the Raman peak intensity of MoSe2’s A1g peak and Si peak for different layer numbers and their ratio. Figure 2(d) shows the Raman peak line width of MoSe2’s A1g peak and Si peak. They both show distinctive variations as a function of film thickness and provide additional assistive information for layer thickness identification. The relative peak intensity between A1g peak and Si peak has a monatomic increase with the layer number. Figure 2(d) presents the Raman peak line widths of both A1g peak and Si peak. While peak position and intensity depend on the layer number, the peak line width is largely independent of layer number.

Figure 3(a) shows Raman spectra of WSe2 using the same method by 633 nm laser. Figure 3(b) shows the Raman peak positions of WSe2’s peak and Si peak for different layer numbers and their difference. Figure 3(c) presents the Raman peak intensities of WSe2’s peak and Si peak for different layer numbers and their ratio. And Figure 3(d) shows the Raman peak line widths of WSe2’s peak and Si peak. In Figure 3(b), Si’s Raman peak is also used as reference. For WSe2, although there are two Raman peaks— and A1g, they are not fully separated making peak position identification inaccurate; thus, the reference to the Si peak is needed. Here, the peak and Si peak’s position difference are characterized on the WSe2 samples of 1L–10L and 15L. It is clear that Raman spectroscopy could be used to characterize the layer number of WSe2 when it is equal to or thinner than 3L based on their Raman peak shift, and by referencing peak intensity and line width, WSe2 of 4L could be characterized. There is a limit of characterizing WSe2 with layer number more than 4. As a result, for 1L WSe2 the Raman peak difference is larger than 270 cm−1, and for 2L WSe2, the Raman peak difference is larger than 269.6 cm−1. Again, the information of Raman peak intensities and line widths provides additional assistive information.

3.3. Photoluminescence and Electrical Characterization of the Sample

Figures 4(a) and 4(b) are the photoluminescence (PL) curves of 1L–4L MoSe2 flakes and 1L–3L WSe2 flakes with reference to the bulk sample. It provides more information for layer number and acts as an additional layer number identification engineering tool when there is a need. It has been found that bandgap increases with the decrease in thickness, and the same trend has also been discovered for WSe2, which conforms with the previous study [27]. The energy shift from 1L to 2L MoSe2 is ∼0.04 eV, and there is no PL peak observed for the sample with layer number larger than 3. The energy shift from 1L to 3L WSe2 is ∼0.05 eV. These two values provide information of characterizing layer numbers in addition to Raman spectroscopy using Si-peak analysis. It also conforms with the previous study [27] that 1L MoSe2 has the strongest PL intensity, which indicates that bandgap transforms from indirect to direct. 1L MoSe2’s PL result is in accordance with 1L MoS2, which shows that it is a semiconducting material with direct bandgap and the bandgap decreases with the increase in layer number [11]. These physical properties provide an additional candidate to the semiconducting industry.

4. Conclusions

In this work, two atomically thin transition metal diselenide materials, MoSe2 and WSe2, with layer number from 1L to up to 15L have been studied by Raman spectroscopy, and the Si Raman peak from the substrate has been used as a reference for the precise characterization. By characterizing both the sample and substrate’s Raman spectra, we are able to determine MoSe2 and WSe2’s layer number for up to 4 layers. We believe that the Si-peak analysis remains the most powerful tool for determining the number of layers of MoSe2 and WSe2. Furthermore, this identification method can be exploited for van der Waals heterostructures made of various 2-D materials such as hBN and TMDCs, when substrate-related peaks are found and the relationship with the number of layers is verified. Their PL properties are also studied as additional information. It has been confirmed with the previous study that there is an enhanced PL in single-layer MoSe2 and WSe2 because of the transition from indirect to direct bandgap electronic structures, and the bandgap decreases with an increase in thickness. These results demonstrate more robust measurements of thickness of transition metal diselenide materials and provide potential of new optical and electrical applications of van der Waals semiconducting materials. In addition, the potential of combination of 2-D materials and other emerging materials such as organic materials will attract extensive attention from researchers in the organic, electronic, and nanotechnology communities.

Data Availability

The data used to support the findings of this study are available at https://www.dropbox.com/sh/3c1l292cqxcvz2l/AACuw8dw0_ClvYaTd0czZhdKa?dl=0.

Conflicts of Interest

The author declares no conflicts of interest.


This work is supported by the startup funding of the Stevens Institute of Technology.