Two-dimensional transition metal dichalcogenides (TMDs), especially the monolayer MoS₂, have attracted great research interest in their fundamental and vital applications as field effect transistors, integrated circuits, photodetectors, light-emitting diodes, and solar cells [1,2,3,4,5,6,7,8,9]. To date, it is still one of the most important research topics to manipulate the electrical and optoelectrical properties of the single-layer MoS₂ through designable patterns [10,11,12,13,14,15,16,17,18], electrical and magnetic fields [19,20,21], defects [22,23,24], and strain [25,26,27,28,29,30,31,32,33] for the novel application as electronic and optoelectronic nanodevices. Strain, among them, is regarded as an effective avenue to modulate the properties of TMD materials, as it directly affects their lattice structures and thus alters energy band structures and optical properties [23,24,25,26,27,28,29,30,31,32,33,34]. Various methods [34] have been reported to introduce strain to a 2D material, mainly including pre-treating the substrates and utilizing the mismatch between TMD materials and their underlying substrates, and so on. Moreover, the method of mismatch-induced a strain in TMD materials is easier and simpler compared to methods of pre-treating the substrates such as deforming and pre-patterned substrates [25, 35]. Very recently, with the fabrication of wafer-scale single crystal WS₂ on vicinal substrate via chemical vapor deposition (CVD) [36], the CVD technique has been a feasible approach to achieve desirable structures and patterns with high performance in TMD-based nanodevices. Previous study has indicated that non-uniform tensile strain can be generated to the CVD-grown MoS₂ on SiO₂ due to the interaction of MoS₂ and SiO₂, therefore providing an effective method for bandgap engineering [37]. Despite holding tunable bandgaps, the photoluminescence (PL) intensity of the strained MoS₂ monolayer on the SiO₂/Si substrate is still relatively weak, which could attenuate the practical applications in optical devices. Therefore, using the interaction between the monolayer TMD materials and the substrate it sited on, especially the dielectric substrate, is indispensable to the design and preparation of strained monolayer TMD materials, where the energy band structures and corresponding optical properties can be effectively manipulated by strain.

In our contribution, for the first time, a monolayer (ML) MoS₂ belt-like single crystal is fabricated on an insulate Rutile-TiO₂(001) substrate through carefully controlling the CVD growth parameters. We find that the optical band gap is broadened and the PL intensity of the as-grown belt-like ML MoS₂ crystal is intensively enhanced by nearly 15 times on the TiO₂ substrate than those samples transferred on SiO₂/Si surface. In order to understand the substrate effect on the optical properties of the belt-like ML MoS₂, we have comparably investigated the as-grown ML MoS₂ ribbons and traditional triangular MoS₂ crystals on TiO₂(001) substrate, and those samples transferred on SiO₂/Si surface through Raman, X-Ray photoelectron spectroscopy (XPS) and PL measurements. Our results demonstrate that the in-plane compressive strain is introduced to the ML MoS₂ belts on TiO₂(001) substrate, resulting in the enlarged optical band gap and thus the realignment of the energy band that inhibits the charge transfer from TiO₂ to MoS₂ in the interface of MoS₂/TiO₂. Therefore, the concentration of charged excitons (trion, A⁻) was suppressed, which helps to enhance the PL emission intensity of the as-grown ML MoS₂ belts. The substrate-confined ML MoS₂ belts provide a new route for tailoring light-matter interactions to upgrade their weak quantum yields and low light absorption and thus can be utilized in optoelectronic and nanophotonic devices.

The crystal structures of the belt-like ML MoS₂ crystal

ML MoS₂ structures with belt-like and triangular shapes were fabricated by design on Rutile-TiO₂(001) via the atmospheric pressure CVD method. The setup of the CVD system is provided in Fig. 1a, which contains two temperature zones, I and II. The precursors, MoO₃ and sulfur power were used to synthesize the ML MoS₂. The well-fabricated belt-like ML MoS₂ single crystals require to precisely regulate CVD growth conditions. Previously, we have demonstrated that the Mo concentration plays a key role on the shape of ML MoS₂ on the TiO₂ surface during the CVD growth [38, 39]. Herein, a close distance of about 2 cm between the MoO₃ source and the substrate was chosen, guaranteeing a relative high Mo concentration, to preferentially obtain the triangular ML MoS₂ single crystal on the TiO₂(001) surface, as shown by a typical scanning electron microscopy (SEM) image in Fig. 1b. At the moment, the substrate temperature, the argon gas flow, and the growth interval are set at 700 °C, 15 min, and 50 sccm, respectively. After a great deal of repeated growth experiments, it was proved that the argon gas flow and the substrate temperature seriously affect the shapes of MoS₂ on TiO₂(001). With the argon gas flow increasing to 100 sccm, the substrate temperature heating up to 800 °C, and other growth conditions remaining, the triangular MoS₂ structures were converted into belt-like MoS₂ pattens evenly distributed on the whole substrate, as shown in a typical SEM image of Fig. 1c. The size of belt-like MoS₂ structures reaches up to the length of about 70 μm and the width of about 7 μm. The temperature of two precursors is shown in Fig. 1d for the growth of the ML MoS₂ belt-like structures on TiO₂(001).

a A set of the CVD system with two temperature zones, I and II. b, c SEM images of ML MoS₂ patterns on rutile-TiO₂(001) substrate with triangular and belt-like shapes, respectively. d The temperature curve of MoO₃ and S are set during the growth of the belt-like ML MoS₂

 Full size image 

Firstly, the crystal structure of the as-grown MoS₂ with triangular and belt-like shapes is characterized by transmission electron microscopy (TEM) and scanning transmission electron microscopy (STEM), respectively. Figure 2a and b show bright-field TEM images of the belt-like and the triangular MoS₂ transferred onto the carbon-coated copper grids, respectively, where a quasi-rectangular and triangular geometry can be distinguished, as labeled by the red contour lines. The left and right panels in Fig. 2c show the corresponding selected area electron diffraction (SAED) patterns taken at any location (except for the edge of the ML MoS₂) of the belt-like and triangular MoS₂, both displaying only one set of hexagonal spots as labeled by red lines in Fig. 2c and suggesting the single crystal nature of the two geometrical MoS₂. This indicates that there are no defects in the entire belt-like and triangular MoS₂ samples. A typical high-resolution TEM (HRTEM) image of the basal plane (labeled by a rectangle in Fig. 2a) of the belt-like MoS₂ is shown in Fig. 2d, which exhibits a hexagonal lattice arrangement with interplanar spacing of 2.7 and 1.6 Å (see a superimposed atomic structure model), corresponding to the (100) and (110) planes, respectively. This further confirms the single crystal nature of the ML MoS₂ flakes on the TiO₂(001) surface.

a, b Bright-field TEM images of the MoS₂ flakes with the belt-like and triangle shape after transferred on a carbon-coated copper grid. The geometrical contour is labeled by the red rectangle and triangle, respectively. c The left and right panels are the SAED patterns from a and b, respectively. d The HRTEM image of the center region (the dashed line rectangle of a) of the MoS₂ belt structure from a, wherein an atomic model of ML MoS₂ structure is superimposed

 Full size image 

The analysis of the elemental composition

The elemental composition of these belt-like and triangular ML MoS₂ patterns was further confirmed with XPS. Figure 3a and b show the deconvoluted XPS spectra, revealing the chemical bonding states of Mo and S of the as-grown ML belt-like MoS₂ patterns on TiO₂ and of those samples transferred onto SiO₂ surface, respectively. For the case of as-prepared ML MoS₂ patterns on TiO₂ in Fig. 3a, two dominant peaks at 230.83 eV and 233.92 eV appear in the core-level spectrum of Mo 3d of Mo⁴⁺, attributed respectively to the doublets Mo 3d_5/2 and Mo 3d_3/2, which is consistent with the literatures which report on the ML MoS₂ [39, 40]. Similarly, two peaks positioned at 230.02 eV and 233.18 eV in the XPS spectrum (Fig. 3b) from transferred ML MoS₂ structures onto SiO₂/Si surface similarly originate from Mo 3d_5/2 and Mo 3d_3/2, respectively. At the binding energy (BE) of 228.02 eV (Fig. 3a) and 227.23 eV (Fig. 3b), a typical dome-shape peak can also be seen in both XPS spectra, indicating the S component of MoS₂ and belonging to S_2s [40]. Besides, tiny amounts of Mo⁵⁺and Mo⁶⁺ from a relative high BE region were identified, which might be owing to the air-induced oxidation of MoS₂ at the defect sites. However, it is worth noting that two Mo 3d peaks and one S 2s peak of the as-grown MoS₂ patterns on TiO₂ substrate exhibit a blue shift of around 0.8 eV compared to those from the samples transferred onto the SiO₂ surface. This might be ascribed to the formation of the semiconductor interface of MoS₂/TiO₂ due to the difference in work functions of the MoS₂ and TiO₂ surface. To check the existence of the MoS₂/TiO₂ interface, the core-level spectra of Ti 2p and O 1s were examined on the blank rutile-TiO₂ (the black curve) and on the MoS₂-covered TiO₂ surface (the red curve) as shown Fig. 3c and d. It is clearly shown that two peaks of Ti 2p and the O 1s peak are synchronously blue-shifted to the higher BE by approximately 0.6 eV for the MoS₂-covered TiO₂ substrate. The BE blue shift in different elements of the XPS spectra evidence the interface of MoS₂/TiO₂, which may result in the charge transfer between MoS₂ and the substrate and affect thus the PL properties of MoS₂ on TiO₂ [41,42,43].

a, b Normalized Mo 3d spectra of as-prepared MoS₂ monolayer belt-like structures on Rutile-TiO₂(001) substrate and of transferred MoS₂ onto SiO₂/Si surface, respectively. c, d XPS spectra of Ti_2p (c) and O_1s (d) collected on as-prepared MoS₂/TiO₂ (red curve) and on the Rutile-TiO₂ (black curve)

 Full size image 

The measurements of Raman and PL spectroscopy

After that, we investigate the effect of the interface of MoS₂/TiO₂ heterostructures on the Raman vibration of the belt-like ML MoS₂ on TiO₂ substrate by comparing the Raman spectroscopy measurement of the belt-like and triangular ML MoS₂ on TiO₂(001) and those samples transferred onto SiO₂/Si surface. Interestingly, the Raman spectrum from the belt-like patterns and the triangular ones are remarkably different, as shown by the red curve and the black one in Fig. 4a. Two prominent peaks in the red Raman spectrum from the belt-like MoS₂ were observed at 385.16 cm⁻¹and 404.14 cm⁻¹, corresponding to the in-plane mode of \({E}_{2g}^1\) and out-of-plane mode of A_1g from the single layer MoS₂, respectively. The single layer thickness of the belt-like MoS₂ structure was confirmed by the atomic force microscopy (AFM) image, as shown in the Supporting Materials as Fig. S1. We can see the distance between the belt-like MoS₂ structure and the TiO₂ substrate is about 0.71 nm from the profile in Fig. S1b, very close to the thickness of single layer MoS₂. Compared to the \({E}_{2g}^1\) mode at 383.16 cm⁻¹ and A_1g mode at 404.14 cm⁻¹ detected from the triangular MoS₂ structure shown by the black curve, the \({E}_{2g}^1\) mode from the ML MoS₂ belt shifts to high energy by about 2 cm⁻¹. However, the Raman spectra of the MoS₂ belt and triangular patterns transferred onto SiO₂/Si simultaneously exhibit two peaks at 383.66 and 404.14 in Fig. 4b. Meanwhile, it is found that the Raman vibration modes for the transferred ML MoS₂ structures are independent of each individual shape. More importantly, the Raman spectrum of the triangular ML MoS₂ on TiO₂ (the black curve on Fig. 4a) is very analogous to those of the transferred samples on SiO₂/Si. This implies that the \({E}_{2g}^1\) peak from the belt structure is blue-shifted compared to that of samples transferred on SiO₂/Si surface. Owing to the former studies, the energy shift in the \({E}_{2g}^1\) peak can be generated by the lattice distortion under the in-plane strain [33, 44,45,46,47]. Moreover, it is reported that the peak of the \({E}_{2g}^1\)mode commences a red-shift and even splits into two sub-peaks with increasing the in-plain tensile strain [33, 44,45,46,47]. In our case, the \({E}_{2g}^1\) peak from the belt structure is blue-shifted compared to that of the triangular structure, opposite with most changes (red-shifts) under the tensile strain [26, 33], which might be resulted from an in-plane compressive strain during the growth of the monolayer belt structure confined by the TiO₂ crystalline orientation as reported other substrates [48, 49]. The detailed growth mechanism of the substrate-confined belt-like ML MoS₂ will be discussed elsewhere [50]. Meanwhile, based on the previous conclusion on the shift value of the Raman vibration mode with strain [26, 33, 37], we can deduce the compressive strain is about 0.3–0.6% in the ML MoS₂ belt structure in our case. In view of the similar Raman vibration modes for triangular ML MoS₂ on TiO₂ and on SiO₂/Si surface, it seems that the triangular ML MoS₂ on TiO₂ surface is in free-strain status. It is easily understandable that the strain can be released for the growth of the triangular ML MoS₂ patterns, which are indpendent on the substrate. Accordingly, the Raman vibration modes of the transferred MoS₂ structures with the belt-like and triangular shapes do not depend on their individual geometry.

a, b Raman spectra of ML MoS₂ belts (the red curve) and triangular ML MoS₂ structures (the black curve) on TiO₂ substrate and of those samples transferred onto a SiO₂/Si substrate, respectively. c, d Photoluminescence (PL) spectra of as-prepared ML MoS₂ with the belt-like (the red curve) and triangular (the black curve) shape on TiO₂ substrate and on SiO₂/Si surface, respectively

 Full size image 

Finally, we deeply explore the optical properties of the ML MoS₂ belt-like structure tuned by the strain and the interface of MoS₂/TiO₂ heterostructure via PL spectroscopy. Fig. 4c shows the PL spectrum detected in the strained ML MoS₂ belt and free-strain triangular structures on TiO₂(001) surface. It is obvious that the PL spectrum of the MoS₂ belt (the red curve) is completely different from that of the triangular structure presented by the black curve. The principal PL peak (A peak) in the strained ML MoS₂ belt at approximately 1.859 eV (the red curve in Fig. 4c) is due to a direct transition at the K point [1, 2]. The B peak originated from a direct transition between the conduction band and a lower lying valence band, which is obscured in our case owing to a strong PL intensity in A peak. However, the A peak of the black curve in the triangular ML MoS₂ structures, being located at 1.821 eV, is relatively weak with an intensity decreased by about ten times than that of the peak A in the strained belt structure. Moreover, the peak A in the red PL spectrum is in a blue shift at about 30 meV compared to that in the black PL spectrum. To visualize the spatial photoluminescent feature, PL mapping in the whole belt structure was executed respectively at the energy of 1.859 eV and 1.821 eV for the as-grown and the corresponding sample transferred on SiO₂/Si surface, as shown in Fig. S2a and 2b. The PL mapping images display a belt shape with a uniform photoemission intensity in the whole structure. Moreover, the integral intensity of the belt-like MoS₂ on TiO₂(001) surface is much higher of ten times than that of belt-structure on SiO₂/Si surface, evidenced by the color bars in Fig. S2a and 2b.

According to the previous studies on the strain engineering the band gap transition of monolayer and bilayer MoS₂ [33, 51,52,53], the blue shift of the A peak is still induced by the in-plain compressive strain, which is in consistence with our Raman results. Based on the relationship of the energy shift value of the peak A with the strain reported before [26, 33, 37, 53], the strain introduced to the ML MoS₂ belt on the TiO₂ substrate is in the range of 0.3–0.6%, similar to what have been obtained from the blue-shift in Raman \({E}_{2g}^1\) mode. PL spectra in Fig. 4d obtained in the transferred ML MoS₂ structures with different shapes on SiO₂/Si further show that common PL traits observed in the red curve and the black one with the same intensity peak A and B, projecting at about 1.821 eV and 1.964 eV, respectively. Moreover, the intensity of peak A is severely dropped down by almost 15 times than that of the strained belt structure on TiO₂(001). In other words, the intensity of the peak A in the strained belt-like ML MoS₂ on TiO₂(001) surface is upgraded by 15 times compared to that of the ML MoS₂ belt on SiO₂/Si surface. The enhanced PL intensity and the blue-shift of the PL peak A in the as-grown ML belt structure can be ascribed to the strain effect and the charge transfer between the MoS₂/TiO₂ interface, which will be deeply clarified in the following.

It is well-known that the PL properties of TMD materials are intrinsic to their electronic structures and excitonic effects [54, 55]. To fully understand the increased PL intensity and the blue-shift in PL peak A position for the strained MoS₂ belts, the PL spectra from the ML MoS₂ belt and the triangular ML MoS₂ structures, respectively shown in Fig. 5a and b, are deconvoluted into the neutral exciton peak (A⁰, the blue peak) and the negative trion peak (A⁻, the green peak) based on the Gaussian function fitting. Sequentially, we can quantitatively extract the peak position and PL intensity of A⁰ exciton and A⁻ trion based on the fitting lines. More intuitively, the relationship of the peak A position of the as-grown ML MoS₂ on TiO₂ and those samples transferred on the SiO₂/Si surface with the strain is summarized in Fig. 5c. With increasing the strain, the blue shift (~ 30 meV) emerges in the peak A position of the as-grown belt structure, suggesting an enlarged optical band gap. This phenomenon can be understood that the compressive stain may induce an up-shift in the conduction band minima (CBM) at the K-point and a down-shift in the local valence band maxima (VBM) at the Γ-point. Accordingly, the optical band gap ban be broadened, which has been predicted in the former theoretical results [33, 37, 51, 56, 57]. Under the influence of the strain, the PL intensity for the belt-like and triangular samples on the TiO₂ surface exhibits an overt dissimilarity. To understand the intensity disparity, we should consider the many-body complexes in 2D TMD materials tuned by the strain and different interfaces [54, 55]. The intensity ratio of (R) of the A⁻ and A⁰ peak of the as-grown ML MoS₂ and the transferred samples with strain is listed in Fig. 5d. Here, R is defined by the function:

a, b PL spectra of the belt-like and triangular ML MoS₂ patterns on TiO₂(001) surface were fitted by Gaussian function. The relationship of the energy position (c) and the intensity ratio of A⁻ and A⁰ (d) of the as-grown and the after-transfer samples with the strain. e, f The structural schematic model of the energy band realignment in the interface between the belt-like MoS₂/TiO₂ and the triangular MoS₂/TiO₂, respectively

 Full size image 

Among of which \({I}_{A^{-}, AG}\), \({I}_{A^0, AG}\), means the PL intensities of the A⁻ trion and A⁰ exciton from the as-grown ML MoS₂ structures on TiO₂, and \({I}_{A^{-},{T}_s}\), and \({I}_{A^0,{T}_s}\) means the A⁻ trion and A⁰ exciton from the transferred triangular structures on SiO₂/Si surface, respectively. This A⁻ and A⁰ intensity ratio from as-grown structures on TiO₂ is normalized by the A⁻ and A⁰ intensity ratio in the free-strain ML triangular MoS₂ structure to compare the strain effect on the A⁻ trion density in different samples. The deconvoluted PL spectra of the transferred samples onto SiO₂/Si surface are provided in Fig. S3. According to the above definition of R, the A⁻ and A⁰ intensity ratio (R) from the transferred samples on SiO₂/Si is itself, that is to say, the value of R should be 1. From Fig. 5d, we find that R is decreased down to 50% for the as-grown belt structure under a large compressive strain compared to the as-grown triangular structure. This implies that the trion density of the as-grown MoS₂ belt is lower than that of the as-grown triangular one and the transferred samples, which absolutely help to upgrade the PL intensity. The reduced trion is generally attributed to the decreased electron carrier concentration, originating from the interface charge transfer. Due to the strained belt-like structure on TiO₂(001) surface, the energy band structure in the interface of MoS₂/TiO₂ will be realigned as shown by the band structural schematic model in Fig. 5e, which is ascertained by the band gaps of TiO₂(001) [58] and the optical band gap tested from our PL measurement. According to Fig. 5e, it is found that the Fermi level of the interface is far from the CBM of the MoS₂. Under these circumstances, the charge transfer from the n-type TiO₂ substrate to the ML MoS₂ belt will be effectively suppressed [58], which results in the reduction of the charged exciton (trion). Conversely, the triangular ML MoS₂ without a strain, the energy band realignment is different from that of the large-strain ML MoS₂ belt, as shown by Fig. 5f. In this case, the Fermi level of the interface is close to the CBM of MoS₂ and the electron carrier transfer easily happen from the TiO₂ substrate to the MoS₂ ML. As a result, the trion concentration in the as-grown triangular structure is larger than that in the belt structure. Consequently, the PL intensity in the as-grown ML MoS₂ belt is increased by 5 times for that in the triangular structure and by 15 times for that in the free-strain transferred samples. Our results are the first time to report that the PL intensity can be extremely upgraded by the compressive strain, opposite with the former results on PL intensity reduction with an increased tensile strain.

In summary, monolayer (ML) MoS₂ belt-like single crystals were directly fabricated on Rutile-TiO₂(001) surface via CVD, and their PL properties strongly tuned by the strain, the interface of MoS₂ and TiO₂ and the charge transfer between them. We observe an increased optical band gap and a stronger PL intensity of the as-grown ML MoS₂ nanostructures than those transferred on SiO₂/Si substrate. The optical band gap is broadened by about 30 meV and the PL emission intensity is enhanced by nearly 15 times for the as-grown ML MoS₂ belt-like crystals on TiO₂ compared to those on SiO₂/Si substrate. This has been explicated by the strain that induces the up-shift in CBM and down-shift in VBM in K-space as well as the energy band structural realignment in the interface of MoS₂/TiO₂. From this, the charge transfer from TiO₂ to MoS₂ is effectively inhibited in the strained ML MoS₂ belts and the concentration of charged excitons (trion, A⁻) is thus suppressed, which finally benefit the enhancement of the PL emission intensity. The substrate-confined ML MoS₂ belts provide a new way for tailoring many-body complexes and light−matter interactions to upgrade their weak quantum yields and low light absorption, which would be useful for the realization in optoelectronic and nanophotonic devices.