A quartz boat loaded with WSe₂ powder (∼1 g) was located at the center heating zone of the furnace, as shown in Figure 1A, and a clean silica (∼300 nm)/Si substrate was located at the terminal of the furnace as the growth substrate. The horizontal distance between the powder and the silicon wafer was kept between 11 and 15 cm. During the process of temperature increase, the growth substrate was placed upstream of the source, and the airflow was set to 300 sccm (along the “step 1” arrow in Figure 1A) for 15 min to remove the undesired oxygen and water from the tube. The Ar airflow was then kept at 50 sccm. When the central heating zone reached the target temperature of 1150°C, the 50 sccm airflow with the transferred direction (along the “step 2” arrow in Figure 1A) ensured the growth substrate was located downstream of the source and then maintained for 10 min for growth. Next, the airflow direction was reversed (along the “step 1” arrow in Figure 1A) to stop the growth process and allow it to naturally cool down to room temperature. The entire process occurred under ambient pressure in an argon atmosphere.

To obtain the WSe₂/WS₂ 2D quantum well, the monolayer WSe₂ was chosen as a seed to further the epitaxial growth of the monolayer WS₂. At the same time, quantum wells started forming at the interfaces of WSe₂/WS₂ lateral heterojunctions. For the epitaxial growth of WS₂, the WS₂ powder and prepared monolayer WSe₂ (substituted for clean substrate and WSe₂ powder) were placed at the same position in another specified CVD system similar to the system used for WSe₂ growth. Likewise, during the period of temperature increase, the growth substrate was placed upstream of the source and the airflow (along the “step 1” arrow in Figure 1A) was set to 300 sccm for 15 min, and then reduced to 100 sccm and kept at that airflow rate. When the central heating zone reached the target temperature of 1120°C, the airflow was set to 100 sccm in the transferred direction (along the “step 2” arrow in Figure 1A), ensuring the growth substrate was located downstream of the source, and then maintained for 5 min for WS₂ epitaxial growth. After that, the airflow direction was reversed (along the “step 1” arrow in Figure 1A) to stop the epitaxial growth process and allow it to naturally cool down to room temperature. The entire process occurred under ambient pressure in an argon atmosphere.

Optical microscopy and STEM were used to characterize the morphologies of the nano-heterojunctions. The Raman spectra were performed using a Renishaw confocal Raman system excited by a 488-nm laser at room temperature. The applied laser power was set to 5%, and the exposure time was approximately 0.1 s. STEM imaging and EELS analysis were carried out using a Thermo Scientific Themis Z 3.2 system equipped with a cold field-emission gun and a mirror lens corrector operating at 300 kV. All STEM experiments were completed at room temperature. The strain distribution of the nanostructure was calculated based on the geometric phase analysis method using the FRWRtools plugin for DigitalMicrograph (www.physik.hu-berlin.de/en/sem/software/software_frwrtools).

The STEM samples were prepared using a poly-methyl-methacrylate (PMMA)–assisted method. PMMA was spin-coated onto the heterojunction sample until the PMMA layer fully covered the heterojunction sample and then baked at 180°C for 3 min. Afterward, the wafer was immersed in a saturated 0.2 M NaOH solution to etch the SiO₂ layer. After the sample coated with PMMA was separated from the silicon wafer, it was transferred into fresh deionized water, repeatedly, to wash away residual contaminants and then fished by a conventional lacey carbon film TEM grid. The as-transferred sample was dried naturally in the ambient environment and then immersed in acetone overnight to clean up the PMMA coating layers. All the STEM samples were baked at 150°C for 6 h to avert unexpected contamination from hydrocarbon.

Figure 1B is the synthesis strategy diagrammatic sketch of the WSe₂/WS₂ 2D quantum well, and Figure 1B① and Figure 1B② show the idealized schematic sketch of the monolayer WSe₂ and the WSe₂/WS₂ 2D quantum well, respectively. The WS₂ quantum well was inserted into the monolayer WSe₂, which started to grow from the WSe₂/WS₂ interface, as depicted in Figure 1B②. The width of the quantum well in Figure 1B② is exaggerated to make the display easy to recognize and understand. Figure 1C is the optical image of the monolayer WSe₂/WS₂ lateral heterojunction synthesized on a silica/silicon substrate; it shows two concentric regions with slightly different optical contrast due to the refractive index and thickness difference between tungsten selenide and tungsten sulfide (Henrie et al., 2004; Blake et al., 2007). Here, the nanoscale WS₂ 2D quantum wells formed at the WSe₂/WS₂ interface but were difficult to recognize in the optical image due to the low enlargement factor of optical microscopy. To further explore the spatial modulation of structural and optical properties in WSe₂/WS₂ lateral heterostructures, the confocal Raman microscope was used to characterize the nanostructure. We picked up three points marked ''1,″ ''2,″ and ''3″ in Figure 1C for the Raman study, where ''1'' (black point), ''2'' (red point), and ''3'' (blue point) are located at the internal tungsten selenium, the WSe₂/WS₂ interface, and the epitaxial tungsten sulfide area, respectively. The Raman spectra at different locations showed distinct differences: the internal region Raman spectrum had a significant peak at 250 cm⁻¹ (black line in Figure 1D), consistent with the A_1g resonance pattern of WSe₂, and the peripheral region Raman spectrum had two significant peaks, at 350 cm⁻¹ and 419 cm⁻¹ (blue line in Figure 1D), corresponding to the E¹ _2g and A_1g resonance patterns of WS₂. Significantly, the Raman spectrum of the WSe₂/WS₂ interface showed both the resonance modes of WSe₂ and the resonance modes of WS₂ (red line in Figure 1D), which indicate that tungsten selenium and tungsten sulfide coexist well within the same triangular domain.

These WS₂ quantum wells extending from the WSe₂/WS₂ interface, mostly distributed in parallel, were generally observed in the lateral WSe₂/WS₂ heterojunction samples, as shown in the STEM image in Figure 2A. Figure 2B is the enlarged view of Figure 2A, and the clear boundaries between WSe₂ and WS₂ in the WSe₂/WS₂ 2D quantum well are apparent. Also, the strain distribution corresponding to Figure 2B is given in Figure 2C. The perfect lattice from the monolayer WSe₂ was used as a reference for calculating the strain. In the perfect monolayer, the lattice constant of WSe₂ was ∼4% larger than that of WS₂, which indicates that the WS₂ quantum well had high and uniform tensile strain along its growth direction, leading to the observed dislocation-free transverse interface. However, the situation was different in the direction perpendicular to the growth direction, where there was a comparatively larger lattice mismatch between WS₂ and WSe₂ (∼4.3 ± 0.5%), as revealed by the calculated strain distribution spectrum.

Figure 3A shows a 2D quantum well in which the monolayer of the WS₂ nanoribbons is laterally sandwiched between two monolayer WSe₂ nanoribbons. As shown in the STEM image in Figure 3A, the Se and S atomic sites are distinguished from the image contrast due to different atomic numbers (Krivanek et al., 2010; Zhou et al., 2012), where the WS₂ quantum well shows a lower image intensity than the surrounding monolayer WSe₂, and the WS₂/WSe₂ interface is marked by red dashed lines.

The high-resolution STEM-annular dark-field (ADF) image of a WS₂ quantum well is shown in Figure 3B; the yellow and red circles mark Se and S atoms, respectively. It was observed that the ∼1.2 nm WS₂ quantum well was embedded in a single layer of WSe₂ and growing along the hexagonal armchair orientation. Correspondingly, Figure 3C is the idealized atomic structural model of the WS₂ quantum well according to Figure 3B. The interface is highlighted by the red and black dashed lines in Figures 3B,C, respectively. It is obvious that there are no mismatches or defects at the interface, which means a fully coherent lateral interface was formed.

Similar to film growth, strain relaxation at the epitaxial interface with intrinsic lattice mismatch produces mismatch dislocation once the critical width is exceeded (Matthews and Blakeslee, 1974; Jain et al., 1997). For the lateral WSe₂/WS₂ heterostructures, the mismatch dislocation array spaced ∼8 nm in the zigzag orientation was expected to alleviate the lattice strain due to the presence of ∼4% lattice mismatch between WSe₂ and WS₂. Through careful observation, it was found that WS₂ quantum wells relied on a dislocation core for growth at the lateral WSe₂/WS₂ heterointerface in general, and further growth was considered driven by misfit dislocation climb. As shown in Figure 4A, the insertion of one W atom and two S atoms from the gas source was considered initiation of the dislocation-driven growth as such dislocation drove the W and S atoms to invade the WSe₂ along the hexagonal zigzag orientation, which is consistent with our experimental results. According to statistical analysis, the width of WS₂ quantum wells is about 1.2 nm, which is equivalent to the width of the four WS₂ unit cells. A four-unit-cell-width and a one-unit-cell-width WS₂ nano-seed, shaded in light red in Figure 4B,are shown penetrating to the WSe₂ monolayer to propagate the quantum well.

Theoretically, the formation of the mismatch dislocation at the interface was required per each ∼8 nm interval to release strain. After repeating the aforementioned growth mechanism, parallel WS₂ quantum wells were formed, and these parallel 2D narrow nano-bands were quantum well superlattices, as shown in Figure 4C.

According to the aforementioned results, there is potential to generate a 2D quantum superlattice, consisting of equally spaced 2D quantum wells, if growth conditions are precisely controlled. The WS₂ quantum well grown by this mechanism formed a type II band arrangement with the surrounding monolayer WSe₂, as shown in Figure 4B.

Meanwhile, the n-type doped ultranarrow WS₂ nanoribbons were expected to be conductors due to the type II band alignment. Nanoribbons of higher carrier concentrations can be obtained by this process of modulation doping as ionizing donors are distributed in the WSe₂ region and thus reduce Coulomb scattering in WS₂. Furthermore, the quantum wells shown in Figure 5A are not evenly spaced, and such a phenomenon could be attributed to the periodic array that was not initially formed at the interface. The bottom area in Figure 5B is an enlarged display of the red dotted frame in Figure 5A. There are different energy band distributions in the region of tungsten sulfide and tungsten selenide. From Figure 5A and Figure 5B, it can be concluded that the length distribution of quantum wells ranges from ten nanometers to a few microns, while the spacing between the quantum wells is about 8 nm, which is basically consistent with the previous theoretical analysis results. High-quality 2D semiconductor quantum well superlattices may be prepared by a misalignment drive mechanism if the growth parameters of the chemical vapor deposition (CVD) system are stably controlled.