1 Introduction

Two-dimensional (2D) transition-metal dichalcogenides (TMDCs) are a class of layered materials with unique electronic [1,2,3], optical [4,5,6], and catalytic [7] properties, which thus have promising applications in high-performance nanoelectronics, optoelectronics, and catalytic reactions [8, 9]. Structural defects like point defects and one-dimensional (1D) line defects (e.g., grain boundaries, edges, and vacancy lines) are inevitable in these atomically thin TMDC materials, whose existences may significantly affect the properties of materials. As reported previously, the misorientation angles of grain boundaries influence electrical conductivity of MoS2 [10], and edges in WS2 monolayers exhibit extraordinarily photoluminescence intensity due to their unique structures and chemical compositions [11]. Chalcogen vacancy lines, another frequently observed 1D line defects in 2D TMDC materials, can provide additional metallic channels [12,13,14] and thus promote catalytic performance such as hydrogen evolution reaction (HER) activity of the host materials [15]. Furthermore, if arranged in a proper way, these chalcogen vacancy lines can even construct novel 2D materials such as the derivative M2X3 phase [16].

To tailor the defect structures and configurations, a variety of methods were developed, including irradiation (via energetic ions, electrons, plasmons), thermal excitations, charge doping, and strain engineering, which are so-called defect engineering if these processes were conducted in a controlled way [17]. Demands for novel defect tuning routes always exist. Nanoscale mass transport, which conveys objects like atoms, clusters, and molecules through well-defined channels and containers, has important applications in memory devices [18] and nanomotors [19]. The widely adopted mass channels include quasi-1D carbon nanotubes (outer surface or inner cavity) [20, 21], 1D line defects like edges, grain boundaries, sulfur vacancy lines in 2D materials [12, 22], and interlayer spacing between adjacent 2D membranes [23]. To date, very few attentions were paid on investigating the interplay between the mass transportation and the related structural transformation of 1D channels, particularly for those sulfur vacancy lines in monolayer TMDCs.

In this work, we demonstrate that mass transport can also serve as an effective route to tailor the configurations of 1D line defects in monolayer MoS2 via in situ atomic-resolved annular dark-field scanning transmission electron microscopy (ADF-STEM) imaging. The whole manuscript was organized as follows. Firstly, we presented the atomic structures of two kinds of line defects including sulfur vacancy line and Mo chain, which were further verified by density functional theory (DFT) calculations. Secondly, healing of sulfur vacancy lines into locally defect-free MoS2 lattice via Mo cluster transportation was reported. And the evolution from sulfur vacancy lines to Mo atomic chains was resolved at the atomic scale which was triggered by the motion of Mo cluster or Mo atoms. Finally, various structures of Mo chains were revealed and found to be closely related to the sulfur vacancy lines pre-existing in the MoS2 matrix which served as transport channels and templates during the formation of Mo chains. Such a mass transport induced structural transformation provides us additional routes for defect engineering (including the local defect repairing), and thus tailoring the electronic structures and physical–chemical properties of the host 2D materials.

2 Experimental

2.1 Sample preparations and ADF-STEM characterizations

MoS2 monolayers were prepared by the chemical vapor deposition (CVD) method in which the Mo containing oxides precursors were sulfurized at high temperatures in the presence of sulfur environment in a furnace. As-grown MoS2 samples were transferred onto a micro-electro-mechanical system (MEMS) heating chip (Wildfire Nano-Chips XT, DENSsolutions Inc.) based on a polymethyl methacrylate (PMMA) assisted wet chemistry process [24], which was then amounted onto a heating holder (SH70, DENSsolutions Inc.). During the experiment, the sample temperature was raised to ~ 1073 K.

ADF-STEM characterizations were conducted with a probe-corrected STEM (Titan ChemiSTEM, FEI) operated at an accelerating voltage of 200 kV. The probe current of ~ 70 pA and a convergence semi-angle of 0.022 rad were used for imaging, with the pixel dwell time set to 10 μs. The inner collection semi-angle was adjusted to be 0.0434 rad, where camera length is properly for the medium-range ADF (MADDF) image mode and to enhance the contrast of sulfur atoms. Experimental STEM images in the main text and Supplementary Information were slightly processed by the standard Wiener deconvolution [25] to partially remove the background noise. ADF-STEM image simulations were done by the QSTEM [26] software, where the input parameters were set according to the experimental conditions.

2.2 Details of DFT calculations

DFT calculations were performed using the generalized gradient approximation for the exchange–correlation potential, the projector augmented wave method [27, 28] and a plane-wave basis set as implemented in the Vienna ab initio simulation package (VASP) [29]. In all calculations, vdW interactions were considered in the framework of van der Waals density functional (vdW-DF) [30] method with the optB86b functional [31, 32] for the exchange energy which was proved to be accurate in describing the structural properties of layered materials [33,34,35,36,37]. A vacuum region of ~ 2 nm in thickness and 1 × 8√3 supercell was adopted. The energy cutoff for the plane-wave basis was set to 500 eV for all calculations. A uniform of 2 × 10 × 1 Monkhorst–Pack k mesh was adopted for integration over the Brillouin zone (BZ). All atoms in the calculation were fully relaxed until the residual force per atom was less than 0.002 eV·nm−1.

3 Results and discussion

3.1 Atomic structures of sulfur vacancy lines and Mo atomic chains

ADF-STEM images (experimental and simulated) and the associated ball-and-stick models in Fig. 1 illustrate the structures of two different line defects, i.e., sulfur vacancy line (Fig. 1a-d) and Mo atomic chain (Fig. 1e-h), both of which were formed as a result of thermal excitation assisted e-beam irradiation-induced sample damage as reported previously. A sulfur vacancy line actually comprises of two sulfur vacancy rows which are located in different sulfur layers (one in the top and the other in the bottom layer) and in neighboring lattices (rather than stacked right vertically) as indicated by red circles in the side view in Fig. 1d, which was reported before [12, 16] and was demonstrated to be nearly metallic (bandgap of Eg = 0.04 eV, as shown in Fig. S1). The Mo atomic chain shown in Fig. 1e-h contains two parallel rows of Mo atoms in which the Mo atoms are closely packed as compared with the pristine MoS2 region in the top view, resulting in the dense bright contrast in the ADF-STEM image. The additional Mo atoms in this Mo chain line defect leave the basal Mo plane of MoS2 to avoid local compression, leading to the two-layered configuration as indicated from the side view in Fig. 1h. The experimentally determined structures were further verified via DFT calculations, which accord well with the experimental results. For example, the experimentally measured width (w defined in Fig. 1f) was ~ 1.255 nm, close to the calculated one of ~ 1.261 nm. Furthermore, DFT calculations revealed that the Mo atomic chains exhibit metallic behaviors (Fig. S2 for detailed structures and Figs. S3, S4 for electronic structures).

Fig. 1
figure 1

Structures of sulfur vacancy line and Mo chain: ADF-STEM images of a sulfur vacancy line and e Mo chain in MoS2, where orange arrows indicate locations of sulfur vacancy lines; experimental (Exp.) ADF-STEM images of b sulfur vacancy line and f Mo chain; simulated (Sim.) images of c sulfur vacancy line and g Mo chain. DFT calculated atomic structures of d sulfur vacancy line and h Mo chain where red circles in side view in d indicate S vacancies

Full size image

3.2 Mass transport induced local healing of sulfur vacancy lines into defect-free MoS2

Time sequenced ADF-STEM images in Fig. 2 and Movie S1 show the dynamic process of the transport of a Mo cluster induced local healing of a sulfur vacancy line. As shown in Fig. 2a, arrays of sulfur vacancy lines aligned along the zigzag direction ([100] direction) were created in MoS2 lattice under prolonged e-beam irradiation and sample heating. One of them (the left one in the marked region) was decorated with a Mo cluster on its bottom end (represented by a dashed circle on the bottom, not fully resolved). Under such an experimental condition, Mo clusters were frequently observed as a result of the agglomeration of mobile Mo adatoms and/or Mo segregation from the host MoS2 lattice. During t = 0–6.4 s, this Mo cluster moves upward along the vacancy line, and surprisingly and easily judged from the Z-contrast ADF-STEM image, those parts of this sulfur vacancy line where the Mo cluster migrates disappeared. In contrast, the other sulfur vacancy line (upper one in the remarked region) that did not undergo any mass transportation well remained.

Fig. 2
figure 2

Transformation from sulfur vacancy line to defect-free MoS2: ADF-STEM images showing transformation from vacancy line to MoS2 at a 0 s, b 3.2 s and c 6.4 s; d, e enlarged plus rotated images of initial state (two sulfur vacancy lines) and final state (one vacancy line and MoS2) of white dotted rectangles in a and b, respectively, where orange arrows indicate locations of sulfur vacancy lines; f, g corresponding atomic models of vacancy line to MoS2 transformation. Note that, for simplification, we treat the top vacancy line into MoS2 since connection between two vacancy lines is very similar to that of vacancy line and MoS2, as discussed in Fig. S5

Full size image

Atomic structures of the same region on the sulfur vacancy line and the associated ball-and-stick model (top and side view) before and after the Mo cluster transportation are presented in Fig. 2d–g. Careful structural analysis confirms that this section of sulfur vacancy line changes to defect-free MoS2 lattice, indicating that the original sulfur vacancy line was healed during the Mo cluster transportation.

To figure out the defect transformation route, detailed comparisons were further conducted, as shown in Fig. 2f (before mass transport) and Fig. 2g (after mass transport). For brevity, the constitutive unit for sulfur vacancy line can be defined as Mo6S10 (according to the atomic ratio) as highlighted in the golden box in Fig. 2f. Then, by choosing the two unchanged Mo arrows surrounding the sulfur vacancy line (i.e., ‘Mo1' and ‘Mo2' in the side view as the reference), we found that the initial Mo6S10 unit transforms into Mo5S10 (the white box in Fig. 2g) with its Mo:S atomic ratio changing from 6:10 to 5:10 (i.e., 1:2), which is the same as that of the host MoS2 lattice, indicating that the pristine sulfur vacancy line defect was healed after the Mo cluster migration.

To keep the stoichiometric ratio consistent, one row of Mo atoms in the sulfur vacancy defect region should be removed during the defect healing process. Time-sequenced ADF-STEM imaging and more clear corresponding ball-and-stick model ambiguously resolved that one Mo row was missing during this process, i.e. changing from 5 Mo rows to 4 rows (Mo1&Mo2 rows were not included). It was found that ~ 23 and ~ 32 Mo atoms in MoS2 were removed and absorbed into Mo cluster during the first 3.2 s (Fig. 2a, b) and the next 3.2 s (Fig. 2b, c), respectively. Owing to the limited temporal resolution, the exact missing Mo row could not be assigned precisely, hence which was somehow arbitrarily illustrated by the red circle in the side view (Fig. 2f, g). As a consequence, the defect healed region shows a slight in-plane lattice contraction of ~ 0.12 nm along the armchair direction (that is [010] direction of MoS2, the dotted black line in Fig. 2f, g), which is ~ 38% lattice constant of MoS2 lattice. The missing Mo atoms were actually absorbed by the mobile Mo cluster.

Ideally, excessive sulfur supply is needed to fix up a sulfur vacancy line. Such a condition can be easily built for ex-situ treatments like using thiol route to repair the sulfur vacancies [38] but hard to establish for in-situ TEM experiments where a harsh sulfur deficient condition exists due to the volatile nature of sulfur and the preferable beam induced formation of sulfur vacancy. Hence, the mass transportation route demonstrated another possibility to repair the sulfur vacancy via taking away excessive Mo atoms in the sulfur vacancy region, at the cost of slight lattice contraction to the healed MoS2 lattice. During this process, the Mo cluster can be regarded as a zipper to heal the line defect and as a mass container to collect redundant Mo atoms.

3.3 Mass transport induced formation of Mo atomic chains

Figure 3 and Movie S2 show the dynamics of the mass transport induced formation of Mo atomic chains. As seen, a slowly moving Mo cluster was migrating toward the right side along the channel as defined by the sulfur vacancy line, during which two-atom wide Mo atomic chains were formed: one in the front and the other one located on the rear side, and all the constitutive Mo atoms were released from the Mo cluster. Taking the front side atomic chain as an example, from t = 0 s to t = 6 s, it grows by ~ 2.3 nm in length, corresponding to ~ 15 Mo atoms. Eventually, the cluster can be fully sacrificed and transformed into atomic chains.

Fig. 3
figure 3

Mass transport induced transformation from sulfur vacancy line to Mo chain: ADF-STEM images showing transformation from a vacancy line to b Mo chain; c corresponding atomic models of vacancy line and Mo chain. Golden and orange boxes indicate structural unit of line defects, that is, Mo6S10 and Mo8S8 as unit cell of vacancy line and Mo chain, respectively; black dotted lines are fiduciary lines, and the red arrows indicate lattice shrinking along y-axis; lattice shrinking was measured based on average position of Mo atoms in intensity line profile; red dotted circles (in side view) indicate possible positions of removed S atoms while red solid circles indicate possible positions of adsorbed Mo atoms during transformation (more details about shrinking in Fig. S6)

Full size image

Similar structural analysis was conducted to resolve the defect structures before and after the dynamic transformation, and the results are presented in Fig. 3c. As seen, the initial unit cell-Mo6S10 (highlighted in the golden box) changes to Mo8S8 (Mo chain, orange box). Note that, two sulfur rows were removed during the process, while it is hard to directly correlate the formation of sulfur vacancy and Mo atomic chains together, since the employed experimental condition (e-beam irradiation plus sample heating) also favors the sulfur vacancy formation even in defect-free MoS2 lattice. In this respect, it is proposed that the newly formed sulfur vacancies may favor the Mo atoms released from the cluster and the subsequent migration along the well-defined channel provided by the sulfur vacancy, and thus leading to the formation of Mo atomic chains.

3.4 Different structures of Mo atomic chains

In addition to quasi-parallel two-atom-wide Mo atomic chains described above, up to nine different configurations of Mo chain were also observed as various results of mass transport, as shown in Fig. 4, where additional Mo atoms present abnormal bright contrasts in ADF-STEM images. The structure of Mo chains in Fig. 4a–f can be regard as the derivates of 1D line of hexagonal Mo rings. The Mo chain in Fig. 4a exhibits regular hexagonal Mo rings, while others are slightly different: there are Mo atoms right above the hexagonal Mo rings for the Mo chain in Fig. 4b; hexagonal Mo rings are elongated and squashed along the direction perpendicular to Mo chains in Fig. 4c, d, respectively; for the hexagon in Fig. 4e, they are slightly separated by two-interval spots; as for Fig. 4f, there is bright contrast in the center of hexagonal Mo rings, indicating the existence of extra atoms. For the Mo chains in Fig. 4g–i, they can be regarded as the assembly of several quasi-hexagonal Mo chains, which exhibit wider bright features in ADF-STEM images. Although it is not easy to get precise 3D atomic models of these complex Mo chains in Fig. 4 through the projected ADF-STEM images, we tried to determine their compositions and structures of Mo chains in Fig. 4a–d via intensity line profiles, as shown in Figs. S7, S8, the results of which indicated these four types of Mo chains varying in both compositions and structures.

Fig. 4
figure 4

ADF-STEM images of different types of Mo chains: a, b, h Mo chains imbedded in MoS2 matrix; c, d, g Mo chains located at boundaries of two inversion domains; e, f, i Mo chains located near Mo2S3 domains (see details in Figs. S9, S10). White triangles in a, c, d indicate orientations of MoS2 domains

Full size image

The different configurations of Mo chains should be closely related to the line defects in the MoS2 matrix such as sulfur vacancy lines which serve as a template and a trigger for the Mo chain formation. For example, the Mo chains in Fig. 4c, d lie on the inversion domain boundaries [39, 40] whose atomic structures are surely different from Mo chain in Fig. 4a (see more details in Figs. S9, S10). Noted that though Mo chains can appear in different configurations, the Mo-chain in Fig. 1 is the most common one under our experimental condition, as shown in the histograms of various Mo-chains in Fig. S11.

3.5 Discussion

Mass transport of Mo cluster with no doubt is essential to the defect transformation as demonstrated above, and this migration behavior needs to overcome a certain barrier that was calculated to be 1.6–3.1 eV [12] for the migration of a Mo adatom along sulfur vacancy line. Such a prerequisite can be fulfilled under the current experimental condition of 200 kV beam irradiation plus sample heating (1073 K). Furthermore, for the structural transformation where the removal of S atoms gets involved (for example, transforming S vacancy line to Mo chain), the presence of electron beam irradiation and sample heating can accelerate the sulfur removal, and hence makes it easier for Mo atoms to enter into S vacancy lines and transform into Mo chains. Energetically, these Mo cluster migration events occurred likely as a result of the lowest-lying structural competition between the three-dimensional (3D) Mo cluster and sulfur vacancy sitting Mo atomic chains. Direct evidence for this is that these as-formed 1D Mo atomic chains can also transform spontaneously back to cluster, as shown in Figs. S12, S13, which indicates that the formation energy of Mo atoms in the Mo cluster and Mo atomic chains is quite similar under our experimental condition. Thus, the motion of Mo clusters cannot be precisely controlled yet. This challenge may be partly solved via the introduction of other excitation such as electric field, as the resultant electro-migration is expected to drive the mass transport in a more controllable way via changing the polarity and strength of the applied electric field as demonstrated previously [18, 19].

The mass transport induced structural transformation should be a universal phenomenon in thicker MoS2 nanosheets, as long as the condition for the formation of sulfur vacancy lines and motion of Mo can be fulfilled. First, the formation of sulfur vacancy defects should be quite similar as the large interlayer separation could be enough to accommodate the slight out-of-plane distortion accompanying the formation of sulfur vacancy defects. Second, the energy barrier for the motion of Mo clusters in the surface layer should be similar to that in monolayer. To realize such structural transformation in thicker MoS2 and even in other TMDC materials, further defect engineering experiments are required with experimental conditions properly selected and precisely controlled.

This mass transport induced structural transformation provides a method to controllably tailor the properties of semiconductive monolayer MoS2, since vacancy line (nearly metallic) and Mo chain are metallic. For example, the formation of sulfur vacancy lines in MoS2 introduces local metallic channels in MoS2, while the healing of these defects to MoS2 tunes the system back to semiconducting. Thus, the mutual transformation of MoS2 and sulfur vacancy line provides a new way to trigger the local metallic-semiconducting transition, similar to 2H-1T phase transition of MoS2. Large-scale S vacancy lines may be fabricated by laser irradiation (create S vacancies) assisted by thermal heating (to active S vacancies to aggregate into S vacancy lines). Furthermore, Mo chain and sulfur vacancy line may also improve the HER performance of the MoS2 due to the active sites in the defect region [15]. Further HER measurements in electrochemical microcells [41, 42] should be conducted to investigate the defect-dependent HER properties.

4 Conclusion

In conclusion, the atomic-scale healing of sulfur vacancy line was revealed by in situ STEM under high temperature. A new line defect-Mo chain was fabricated in monolayer MoS2, and its atomic structures were experimentally resolved, and their electronic structures were explored by DFT calculation. Furthermore, the atomic-scale transformation among Mo chain, sulfur vacancy lines and MoS2 were revealed (as summarized in Figs. S14, S15). More importantly, it was found that the Mo cluster can sever as the source or container of Mo atoms, and the mass transport of Mo atoms or Mo cluster along line defects can induce the transformation of those line defects. The mass transport induced structural transformation provides a new approach for the rational design of defects and modulation of the properties of MoS2.