Traditional bulk Si or III-V semiconductors exhibit superior electronic and optoelectronic characteristics but have limited flexibility and transparency [1]. Consequently, the use of these materials in flexible displays and transparent solar cells is challenging. In contrast, two-dimensional (2D) transition metal dichalcogenides (TMDCs) exhibit outstanding optoelectronic properties while maintaining favorable flexibility and transparency [2]. A TMDC consists of an MX₂ structure, where ‘M’ represents a transition metal (e.g., Mo, W, and Re) and ‘X’ denotes a chalcogenide element (S, Se, and Te) [3]. TMDCs have band gaps ranging from 1 to 2 eV, making them suitable for various applications, such as field-effect transistors, photodetectors, and solar cells [4–6]; thus, they are expected to play a crucial role in the future of optoelectronic devices as emerging semiconductors [7]. Furthermore, their intrinsic weak van der Waals (vdW) interlayer interactions allow for facile stacking of 2D TMDC layers. Recent research has focused on enhancing the optical characteristics or implementing new devices through strategic stacking of TMDC heterostructures such as MoS₂/WSe₂, MoS₂/WS₂, WSe₂/MoSe₂/WSe₂, and WSe₂/WS₂/WSe₂ [8–11]. These heterostructures have active band gaps with bound electrons and holes localized within individual monolayers (MLs). Notably, TMDC heterostructures can form p-n junctions based on their constituent materials, thereby facilitating type-I, type-II, and type-III band alignments [12]. The distinctive properties of these junctions render them exceptionally promising for optoelectronic device applications. Additionally, their energy and luminescence intensity can be finely tuned by adjusting the vertical gate voltage, laser intensity, and thermal treatment [11,13,14].

In this study, we investigated the effect of annealing on the interlayer coupling in a WS₂/WSe₂/WS₂ vdW heterostructure. The WS₂/ WSe₂/WS₂ trilayer was fabricated using a conventional wet transfer process, and it exhibited a type-II band alignment, resulting in the spatial separation and charge transfer of photogenerated electron-hole pairs between the TMDC layers. The interlayer interactions within the WS₂/WSe₂/WS₂ vdW heterostructure were probed using photoluminescence (PL) and Raman spectroscopy. Interestingly, we observed an abnormal PL enhancement in the stacked WS₂/WSe₂/WS₂ layers before annealing, which significantly decreased after annealing. Furthermore, various Raman peak shifts, including those of $E_{2g}^1$+A_1g and A_1g, provided clear evidence of interlayer coupling between WS₂ and WSe₂ after annealing. These findings emphasize the critical role of annealing in precisely controlling the interlayer carrier dynamics within vdW heterostructures.

ML WSe₂ and WS₂ films were synthesized through chemical vapor deposition (CVD), and the stacked WS₂/WSe₂/WS₂ was fabricated on a SiO₂/Si substrate using a wet transfer method. An additional annealing process was performed to control the vdW interactions between the TMDC layers. The furnace was purged with 1,000 sccm Ar for 15 min, and the as-prepared WS₂/WSe₂/WS₂ sample was annealed at 300 °C for 3 h while being exposed to 200 sccm Ar and 50 sccm H₂ to eliminate polymethyl methacrylate (PMMA) residue and enhance the vdW interaction between the layers. Confocal microscopy systems [Nanobase (NANOBASE) and NT-MDT (NTEGRA SPECTRA)] were used to acquire Raman and PL spectra at the nanoscale. A 532 nm excitation laser (2.33 eV) with a laser power of 0.3 mW was used for excitation, and 150- and 1,800-groove gratings were used for PL and Raman analyses, respectively.

First, we confirmed the optical properties of the CVD-grown ML WSe₂ and ML WS₂ using PL spectroscopy. Figure 1(a) shows the PL spectrum and optical image of ML WS₂, with the main peak of ML WS₂ appearing at 616 nm (2.01 eV), corresponding to neutral A exciton states [15,16]. Figure 1(b) presents the PL spectrum of ML WSe₂, where the A exciton peak appears at approximately 745 nm (1.66 eV) [17]. The sizes of the triangular WS₂ and irregular shaped WSe₂ flakes are approximately 60 and 100 µm, respectively. We fabricated the WS₂/WSe₂/WS₂ structure using these TMDCs, following the wet transfer process illustrated in Fig. 2(a). To extend the etching area, we attached tape to all edges of the as-grown TMDC sample and spin-coated PMMA C4 at 3,000 rpm for 1 min. For SiO₂ etching, we detached the tape and floated the sample on a 1 M potassium hydroxide (KOH) solution heated at 80 °C. After 1 min, we removed the PMMA/TMDC film from the KOH solution using a polyester film and rinsed it several times in deionized (DI) water. We then removed the film from the DI water using a cleaned SiO₂/Si substrate and dried it by blowing N2 gas. Finally, we removed the PMMA by rinsing the samples with acetone for 30 s. This sequence was repeated three times to fabricate the WS₂/WSe₂/WS₂ structure. Optical images of each step are shown in Fig. 2(b).

Figure 1. PL spectra of (a) ML WS₂ and (b) ML WSe₂ (inset: optical images of each TMDC).

Figure 2. (a) Schematic illustration of wet transfer process. (b) Optical images of WS₂, WSe₂/WS₂, and WS₂/WSe₂/WS₂ samples following transfer repetition from the left.

Figure 3(a) shows the structure and band alignment of the WS₂/W Se2/WS₂ heterostructure. An optical image of the WS₂/WSe₂/WS₂ trilayer is shown in Fig. 3(b). The triangular samples marked with white and black dotted lines correspond to the upper and lower WS₂ flakes, respectively. The yellow dotted line represents the WSe₂ layer in the middle of the WS₂/WSe₂/WS₂ structure. Figure 4 presents the PL spectra obtained from ML WS₂, ML WSe₂, WS₂/WSe₂, and stacked WS₂/WSe₂/WS₂ before and after annealing. As shown in Fig. 4(a), we observed significant PL quenching in the stacked WS₂/WSe₂ region compared to the case of the ML WS₂. This is attributed to the charge transfer processes associated with the formation of a type-II band alignment in WS₂/WSe₂, as depicted in Fig. 3(a) [18,19]. However, the PL intensity of the stacked WS₂/WSe₂/WS₂ region increased, even compared to that of ML WS₂. The strong PL peak position of the stacked WS₂/WSe₂/WS₂ was consistent with that of ML WS₂; however, an additional shoulder peak was clearly observed at 630 nm, whereas the PL peak associated with WSe₂ was significantly attenuated. Interestingly, after annealing, a substantial decrease in PL was observed for the stacked WS₂/WSe₂/WS₂ as shown in Fig. 4(b). The abnormal increase in PL before annealing is attributed to the decoupling between WS₂ and WSe₂ as well as the transfer of electrons from the WSe₂ middle layer to each WS₂ layer. As a result, the WS₂ trion peak at 630 nm also significantly increased while maintaining the WS₂ neutral A exciton peak [20]. After the annealing process, the interlayer coupling between WS₂ and WSe₂ was strengthened, leading to charge separation from WS₂ to WSe₂, which eventually resulted in PL suppression [19,21,22].

Figure 3. (a) Schematic representation of the WS₂/WSe₂/WS₂ structure and its band alignment. (b) Optical image of the WS₂/WSe₂/WS₂ sample.

Figure 4. PL spectra of ML WS₂, ML WSe₂, WS₂/WSe₂, and WS₂/WSe₂/WS₂ (a) before and (b) after annealing.

Figure 5 shows the Raman spectra of ML WSe₂, ML WS₂, and the stacked WS₂/WSe₂/WS₂, both before and after annealing. The Si Raman peak at 520.1 cm⁻¹ was used for equipment calibration. In the case of ML WSe₂, two primary peaks were observed at 248.3 and 260.7 cm⁻¹, corresponding to the $E_{2g}^1$+A_1g and 2LA(M) modes, respectively [23]. For ML WS₂, the prominent Raman peak was located at 354.6 cm⁻¹, representing the $E_{2g}^1$ mode [24,25]. The Raman peaks of the stacked WS₂/WSe₂/WS₂ incorporated all the individual peaks of WSe₂ and WS₂, as shown in the top graph in Fig. 5. This again suggests that the WS₂/WSe₂/WS₂ vdW heterostructure was well constructed [11].

Figure 5. Raman spectra of ML WSe₂, ML WS₂, and WS₂/WSe₂/WS₂ before (black lines) and after (red lines) annealing, respectively.

For a more detailed analysis, we focused on the portions of the graph ranging from 220−280 and 260−440 cm⁻¹, which display the dominant peaks of WSe₂ and WS₂, respectively. Upon stacking the WS₂/WSe₂/WS₂ layers, we observed a decrease in the Raman intensity of WSe₂, accompanied by an increase in the WS₂ peak. The stacking order and number of layers may influence the initial Raman intensity, and the overall intensity was found to increase after annealing owing to the enhancement of interlayer coupling. Interestingly, an additional peak was clearly observed at 415.9 cm⁻¹, representing the A_1g mode of WS₂. Furthermore, the $E_{2g}^1$+A_1g mode of WSe₂ exhibited a slight blue shift after annealing. Generally, the A_1g mode is highly affected by the carrier doping effect induced by charge transfer [22,26]; therefore, the changes in Raman peaks can be interpreted as being results of the movement of photogenerated carriers in the WS₂/WSe₂/WS₂ heterostructure. As depicted in Fig. 3(a), electrons moved from WSe₂ to WS₂ and holes flowed in the opposite direction, from WS₂ to WSe₂, which is consistent with the PL results.

We further compared the Raman changes for an iso-bilayer (WS₂/ WS₂) and a hetero-bilayer (WS₂/WSe₂ and WSe₂/WS₂), as shown in Fig. 6. Generally, the frequency difference between the $E_{2g}^1$ and A_1g modes increased with the number of layers, and was accompanied by an $E_{2g}^1$ peak increase. The WS₂/WS₂ iso-bilayer exhibited a shift in the $E_{2g}^1$ mode from 354.6 to 352.5 cm⁻¹ and in the A_1g mode from 415.9 to 417.9 cm⁻¹, without a WSe₂ peak. For the hetero-bilayer, instead of an increase in $E_{2g}^1$, a similar blue-shifted $E_{2g}^1$+A_1g mode and an A_1g mode were observed after annealing; these can be attributed to the charge transfer process. However, the changes in the Raman spectra of the hetero-bilayer were comparatively weaker than those of the WS₂/WSe₂/WS₂ heterostructure owing to the absence of a top layer.

Figure 6. Raman spectra of WS₂/WS₂, WS₂/WSe₂, and WSe₂/WS₂ before (black lines) and after (red lines) annealing, respectively.

We successfully fabricated a trilayer configuration of stacked WS₂/ WSe₂/WS₂ using the wet transfer method and explored the effects of annealing on interlayer coupling. The WS₂/WSe₂/WS₂ heterostructure was characterized using PL and Raman spectroscopy. Significant PL enhancement and Raman mode changes were observed, indicating photogenerated carrier movement between the TMDC layers, facilitated by the type-II band alignment. Strengthening the interlayer coupling through thermal treatment resulted in PL quenching, underscoring the critical role of the annealing process in controlling carrier dynamics. Our findings highlight the potential for tuning carrier movement in vdW heterostructures through a straightforward annealing process, opening up various avenues for future applications.

This work was supported by a National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (Nos. 2022R1A4A1033358 and RS-2023-00214318).

The authors declare no conflicts of interest.