Growth of two-dimensional Janus MoSSe by a single in situ process without initial or follow-up treatments

Two-dimensional (2D) Janus transition metal dichalcogenides (TMDCs) are highly attractive as an emerging class of 2D materials, but only a few methods are available for fabricating them. These methods rely on the initial growth of 2D TMDCs in one process, followed by an additional plasma or high-temperature (T) process. To overcome these drawbacks, we employ the new approach of NaCl-assisted single-process chemical vapor deposition, which consists of three steps that proceed only by altering the temperature in situ. In the first step, MoS2 is deposited onto a SiO2/Si substrate with the Mo and S atoms activated in different temperature zones. In the second step, S vacancies are formed in the upper layer of the grown MoS2 by annealing. In the third step, the vacancies are filled with activated Se atoms. Throughout the steps, NaCl lowers the melting point of the constituent atoms, while the T in each zone is properly controlled. The growth mechanism is clarified by a separate annealing experiment that does not involve a supply of activated atoms. These results highlight a simple and cost-effective approach for growing Janus MoSSe, which is more useful for fundamental studies and device applications. (Left) Schematic diagrams describing a single in-situ chemical-vapor-deposition process for growing two-dimensional Janus MoSSe without initial or follow-up treatments and the growth mechanism consisting of three steps; 1) growth of a MoS2 layer, 2) formation of vacancies on the top MoS2 layer by an annealing effect, and filling of Se atoms in the vacancies. (Right) Optical-microscopy and photoluminescence-mapping images for a typical two-dimensional Janus sample showing that the central MoSSe region is distinguished from the peripheral MoSe2 region.


Introduction
Two-dimensional (2D) Janus transition metal dichalcogenides (TMDCs) have recently attracted increasing attention as an emerging class of 2D materials. The transition metal layer of Janus 2D TMDCs is sandwiched by two different chalcogen layers, resulting in broken mirror symmetry along the out-of-plane direction, which distinguishes Janus 2D TMDCs from pristine TMDCs. This unique characteristic of 2D Janus TMDCs leads to their novel phenomena, such as Rashba spin splitting, piezoelectric polarization, second-harmonic generation, catalytic effects, and long carrier recombination [1][2][3][4][5][6][7] , which are very promising for their application in sensors, actuators, piezo/thermal-electric devices, solar cells, ion batteries, and other electromechanical devices [8][9][10][11] . The asymmetry of Janus TMDCs has also been used to tailor the van der Waals (vdW) interlayer coupling and charge transfer in vdW heterostructures [12][13][14][15] ; such tailoring of Janus MoSSe/MoS 2 heterobilayers was performed by tuning the twist angle and interface composition 16 . The energy of graphene plasmons in vdW Janus MoSSe/graphene heterostructures was engineered up to 0.5 eV by varying the number of Janus MoSSe layers 17 .
The synthesis technology of 2D Janus TMDCs has not yet been established, but recently, monolayer (ML) Janus MoSSe was successfully grown by controllably substituting S or Se atoms on the surface of a MoS 2 or MoSe 2 layer with Se or S atoms, respectively, based on two major approaches 18,19 . It is common for both materials, MoS 2 and MoSe 2 , to be grown as a ML in a separate process. In one approach, the top S layer in the pregrown MoS 2 2D sheet was stripped and replaced with H atoms, and the H atoms were then thermally selenized under vacuum 18 . In another approach, the top layer of Se in the pregrown MoSe 2 ML was sulfurized by a controlled substitution reaction with vaporized S 19 . The drawback of these methods is that they rely on follow-up treatments in addition to the initial chemical vapor deposition (CVD) of 2D MoS 2 or MoSe 2 TMDCs, which can result in increased complexity and cost. To overcome these limitations, we developed a new method of Janus MoSSe growth, simplified by employing a single NaCl-assisted process in a CVD apparatus. This approach involves three-step processes in situ to sequentially activate Mo and S for the growth of MoS 2 , form S vacancies in the top MoS 2 layer, and activate Se to fill the vacancies only by altering the temperature (T) without using any other means, such as H 2 plasma 18 . Throughout the steps, the temperature in each chalcogen and transitional-metal zone was properly controlled, and NaCl lowered the melting point of the material. The grown Janus MoSSe was systemically examined by Raman scattering, photoluminescence (PL), atomic/Kelvin-probe force microscopy (AFM and KPFM), and optical microscopy (OM). Based on the experimental results, a possible growth mechanism is proposed. These results are believed to provide a simple, cost-effective, and universal approach for growing 2D Janus TMDCs; this approach is more useful for fundamental studies and device applications.

Materials and methods
Triangle-shaped 2D flakes were produced on 300 nm SiO 2 /Si wafers by the NaCl-assisted vaporization of a mixture of MoO 3 , S (or Se), and NaCl powders in a quartz tube furnace under a controlled gaseous environment. NaCl in the growth process was found to play a vital role in lowering the growth T 20,21 . A typical run of growth for MoS 2 was performed by loading 15 mg MoO 3 + 20 mg NaCl 3 powder in one boat in the high-T zone and 35 mg S powder in the other boat in the low-T zone within the same quartz tube. A piece of the SiO 2 /Si substrate was placed at the downstream location close to the high-T boat. The low-and high-T zones in the furnace were heated to 150 and 750°C, respectively, for 10 min; they were kept at these temperatures for 10 min for growing MoS 2 . The vapor-phase reactants were transported by a mixed gas of H 2 and Ar, thereby facilitating the growth of the 2D crystals in the growth region. The flow rate of H 2 gas was fixed at 20 sccm. All the reactions were carried out at 10 −1 Torr, and their heating rate was 25°C min −1 . At the end of the growth phase, the furnace was left to naturally cool to room temperature. The procedure for synthesizing Janus MoSSe is detailed in the next section. To optimize the growth of Janus MoSSe, the distance between the boats and the rate of Ar gas were varied from 11 to 17 cm and 82 to 88 sccm, respectively.
Raman scattering and PL measurements were carried out using a confocal Raman/PL microscope (HEDA, NOST, Korea) with a 532 nm laser. A laser with a typical incident power of 100 μW was linearly polarized and focused to a spot size of < 1 μm by a 100× objective lens (numerical aperture, 0.9). The sample was mounted on a piezo-stepper table and scanned under the microscope. The spatial resolution of this setup was approximately 500 nm. The PL emission (or Raman scattering) signal was collected with the same objective, dispersed with a monochromator, and detected using a charge-coupled device (Andor Technology). The Raman scattering behaviors of MoSSe and MoSe 2 were also measured as a function of the polarization angle by employing a linear polarizer. The KPFM surface potential profiles and mapping images were obtained in the noncontact mode of an atomic force microscope (Park Systems model XE-100). The samples were transferred to indium tin oxide (ITO)coated glasses for the KPFM measurements to prevent the problem of surface charging. The I-V curves were also obtained on ITO substrates in the conductive mode of the atomic force microscope. Figure 1a and b show the CVD apparatus for growing Janus MoSSe and a schematic diagram explaining the growth mechanism, respectively. The CVD system is divided into two zones, the low-and high-T zones, referred to as zones 1 and 2, respectively. The Se and S powders are located at separate positions in one boat, which is located in zone 1, while the other boat containing MoO 3 + NaCl powder and a SiO 2 /Si substrate is located in zone 2. In the first step, a MoS 2 flake is synthesized on a SiO 2 /Si substrate, where the Mo and S atoms are activated at T = 750 and 150°C, respectively, in each zone (Supplementary Fig. S1). The grown MoS 2 flake showed typical Raman and PL spectra peaked at 382.6/406.4 cm −1 (E 2g /A 1g modes) and 691 nm, respectively ( Fig. 2a and b), as previously reported [22][23][24] . The resulting difference between the Raman peaks was calculated to be 23.8 cm −1 and corresponded to a triple-layer MoS 2 25 . In the second step, S vacancies were formed in the MoS 2 flake (Fig. 1b), while the T in zone 2 was changed to 580°C for 10 min without the addition of any activated atoms. This formation of vacancies was demonstrated in a separate annealing experiment, as described below. In the third step, the S vacancies in the top-surface MoS 2 were filled with Se atoms activated at T = 300°C ( Fig. 1b and Supplementary Fig. S1) to finally form Janus MoSSe. The vacancies were believed to be formed concomitantly in the layer during this last step, as was also demonstrated by the annealing experiment.

Results and discussion
The growth of the Janus MoSSe layers was optimized through variation of the principal deposition parameters.
During this procedure, the shape, size, and uniformity of the Janus MoSSe were shown to be very sensitive to a few major parameters, such as the distance (d) between the boats and the Ar flow rate (R Ar ). Figure 3 shows OM images of the flakes; these images were obtained by varying d at a fixed R Ar of 86 sccm (see Supplementary  Fig. S2 for the optimization through the variation of R Ar at a fixed d of 15 cm) and the corresponding Raman and PL spectra. These results suggest that the Janus MoSSe with the best quality was fabricated under the optimized conditions: d = 15 cm and R Ar = 86 sccm. As shown in the OM image of the optimized flake (Fig. 4a), the inner and outer regions are clearly distinguished by their triangular boundary between the layers. The Raman spectrum in the inner (or the outer) region exhibits major peaks at 287.2/ 350.2 (or 239.6/287.3) cm −1 , corresponding to the A 1g /E 2g modes of MoSSe (or MoSe 2 ), respectively 18,19,26 , as shown in Fig. 3e (or Fig. 2a). The difference between the A 1g and E 2g modes of MoSSe is calculated to be approximately 63 cm −1 , which is very close to the theoretical and experimental values (∼60 cm −1 ) reported for a MoSSe monolayer 19,27,28 . Here, it should be noted that the MoS 2 peaks are also visible in the Raman spectrum of the inner region, indicating that the MoS 2 layer remains at the bottom of the Janus MoSSe even after optimization, resulting in the actual formation of a MoSSe/MoS 2 heterostructure. Figures 2b and 3f show PL spectra of the inner and outer regions of the optimized flake. These spectra have peaks at 726 and 809 nm, which are consistent with those of MoSSe and MoSe 2 , respectively 18,19,26 . As shown in the PL mapping images (Fig. 4b-c), the variation in the peak wavelength is clearly observed as the position moves in the direction from the inner region, the interface, to the outer region. This further demonstrates that the Janus MoSSe inside the flake is surrounded by the peripheral MoSe 2 region with the boundary between the regions. Near the boundary of the inner/outer regions, the intermediate PL and Raman spectra exhibited peaks corresponding to both MoSSe and MoSe 2 ( Supplementary Fig.  S3), resulting from the mixed phases of MoSSe and MoSe 2 near the boundary. As the position moves in the inner-toouter region direction at the boundary, the Raman and PL peak intensities of MoSSe decrease, while those of MoSe 2 increase. Within the inner region, the PL peak redshifts in the center-to-edge direction (Fig. 4c). Figure 4d-g and e-h compare the KPFM mapping images/surface potential profiles and AFM height profiles of MoS 2 and MoSSe. As shown in Fig. 4g, the potential profile of MoS 2 looks very uniform in the full region and is clearly differentiated from that of the substrate. Figure 4h shows a considerable change in the potential due to the formation of Janus MoSSe with respect to the substrate, together with a potential dip in the peripheral MoSe 2 region, consistent with the PL mapping images in Fig. 4b-c. The polarization-angle dependences of the Raman behaviors were measured for the polar plots of the peak intensities in the inner (MoSSe) and outer (MoSe 2 ) regions ( Supplementary   Fig. 1 Growth of Janus MoSSe. a CVD apparatus for Janus MoSSe. The CVD system is divided into two zones, zones 1 and 2. One boat for the Se and S powders is located in zone 1, while the other boat for the MoO 3 + NaCl powder and the SiO 2 /Si substrate is located in zone 2. The Se and S powders are separated within the boat. b Schematic diagram detailing the growth mechanism. The in situ single growth process for MoSSe consists of three steps: synthesis of MoS 2 flakes, formation of S vacancies on the top-surface MoS 2 layer, and growth of Janus MoSSe. Fig. S4). According to the Raman selection rules based on symmetry, the intensity of the E 2g mode was predicted to be independent of the polarization angle, while that of the A 1g mode was predicted to be polarized following the formula I A1g = A cos 2 θ, where I A1g is the intensity, A is a constant, and θ is the angle 29 . This reasoning explains why the line shapes of the A 1g and E 2g modes were typical dumbbell and circular structures in the polar plots. Considerable differences in the polar plots between the MoSSe and MoSe 2 regions were not observed, and this result was consistent with those of previous reports 29,30 .
To analyze the conversion of MoS 2 to MoSSe, a separate annealing process was performed by keeping a grown MoS 2 flake at 580°C (same T as was used for the growth of MoSSe) for 10 min in zone 2 without atomic activation under Ar flow. After annealing, almost no difference in the Raman spectrum of the MoS 2 flake was observed, but its PL spectrum was slightly redshifted ( Supplementary  Information, Fig. S5), which was consistent with a  previous report 31 . Figure 4d-g and f-i compare the KPFM mapping images and surface potential profiles of the grown and annealed MoS 2 flakes. A large change in the potential of MoS 2 , including a sharp drop at the edge, was observed with respect to the substrate during annealing, as shown in Fig. 4i. In other words, the MoS 2 flake was fully changed by annealing, with the inner and peripheral parts being structurally different. No correlations between the KPFM results and the surface morphologies were observed, as confirmed by the AFM images (Supplementary Fig. S6). In addition, several vacant regions were identified in the annealed MoS 2 flake by OM and PLmapping images ( Supplementary Information, Fig. S7) and can be regarded as S vacancies, as previously reported [31][32][33] . These results demonstrate the formation of S vacancies in the MoS 2 flake before and during the addition of activated Se (2nd & 3rd steps) in the fabrication process, as described above.
Mono-and bilayer MoS 2 flakes were also grown in the 1st step of Janus MoSSe fabrication, but after the 2nd and 3rd steps, no Janus MoSSe had formed, possibly due to the excessive thermal budget of these steps. This result further suggests that there is a critical minimum thickness for the successful growth of Janus MoSSe using our approach. Similar behaviors were reported in previous publications 33,34 . In these reports, the mono-layer MoS 2 was cracked when annealed above a critical temperature (this temperature depends on the approach). We also examined the samples after the 2nd step (similar to annealing); these samples included the mono-and bilayer Fig. 4 Mapping images and height/potential profiles. a Optical microscopy and b PL peak-wavelength mapping images showing clear delineation between the inner Janus MoSSe and the outer MoSe 2 regions. c PL peak-wavelength mapping image showing position-dependent variation of the peak wavelength within the inner Janus MoSSe region. The variation in the peak wavelength is clearly observed as the position moves in the direction from the inner region, the interface, to the outer region. Within the inner region, the PL peak redshifts in the center-to-edge direction, possibly resulting from the position-dependent sequential variation of the S vacancies and/or the occupation of the Se atoms in the vacancies. KPFM images and corresponding surface potential profiles of d-g as-grown MoS 2 , e-h Janus MoSSe, and f-i annealed MoS 2 flakes. Here, the surface potential profiles were obtained following the red lines drawn on the KPFM images, and the AFM height profiles of the three flakes are indicated on the KPFM images.
MoS 2 flakes grown in the 1st step. Almost no monolayer material remained, while for the bilayer, some MoS 2 debris was scattered on the SiO 2 /Si surface (Supplementary Information, Fig. S8). To solve this problem, our system should be redesigned to reduce the excessive thermal budget, especially for the cases of the mono-or bilayer MoS 2 . In contrast, the thermal load in thicker samples with ≥ 3 layers seems to be well distributed throughout the whole volume. The thick MoS 2 layer may also protect the Mo-S bond at the bottom. In addition, MoSe 2 formed in the peripheral region of the flake could act as a thermal barrier to reduce the thermal budget of the bottom Mo-S bond. This result explains why the breakage of the Mo-S bond at the surface was described when vacancies formed in the growth mechanism (Fig. 1b). If the annealing time and/or temperature is longer, even the bottom Mo-S bond will break. The optimum time and temperature for the Janus MoSSe were obtained after much trial and error.
The uniformity in the region of the Janus MoSSe was evaluated by measuring the PL and Raman spectra at different positions in the direction from the center to the vertex within the inner region of the triangular flake, as shown in Fig. 5a. Figure 5b, c show the positiondependent Raman spectra of MoSSe and MoS 2 in the inner region and the corresponding shifts of the A 1g /E 2g peak wavenumbers. Here, no considerable changes in the Raman peak energies were observed, as shown in Fig. 5c, but the main Raman peak ratio of MoSSe/MoS 2 decreased from the center to the vertex, as shown in Fig. 5b (and Supplementary Fig. S9a). The PL peak of MoSSe redshifted in the same direction, as shown in Figs. 5d and 4c (and Supplementary Fig. S9b), and this result is very consistent with a previous report 33 . These results are possibly due to the position-dependent sequential variation of the S vacancies and/or the occupation of the Se atoms in the vacancies, as previously reported 31,32 .
It is widely known that the crystalline structure is almost unchanged by the MoS 2 → MoSSe transformation, as shown in various publications 18,19 . The Janus MoSSe remains in a 2H crystalline phase even by the top layer S (or Se)-atom substitution of Se (or S) in MoSe 2 (or MoS 2 ), indicating that there is neither significant lattice distortion nor phase transition. Additionally, according to our results, the crystalline structure of the Janus MoSSe flake remained stable but its shape and size changed depending on the preparation conditions ( Fig. 3 and Supplementary Fig. S2), as confirmed by the PL and Raman spectra. Even though other phases of Janus MoSSe, such as 1T or 1T′ exist, PL cannot be observed due to the metallic properties of the 1T or 1T′ phases, and their Raman spectra differ from that of the 2H phase, as shown in other 2D materials.
The shape of the 2D flakes is known to strongly depend on the preparation conditions [35][36][37] , but most of the 2D flakes are predominantly triangle-shaped. For example, the crystal shape of the MoS 2 at equilibrium is hexagonal under Mo-rich conditions, while it becomes triangular under S-rich conditions 36 . More importantly, constructing larger and more uniform Janus crystals with the smallest possible defect density is a preeminent challenge. We accidentally succeeded in making Janus MoSSe with almost no peripheral MoSe 2 region ( Supplementary Fig. S10), even though its preparation conditions needed to be further optimized for stabilization. In contrast, more S vacancies tended to form in hexagonal MoS 2 shapes, and their distributions gradually varied from the vertices and edges (triangular shape) to the center (hexagonal shape) 37 . Based on these findings, we focused on trying to make the Janus flakes as large and uniform as possible to secure a larger central MoSSe region in each triangular flake.
The I-V curves measured for the inner MoSSe and outer MoSe 2 regions showed rectifying behaviors, with steeper voltage-dependent current variation in the MoSe 2 region ( Supplementary Fig. S11). Both layers formed heterojunctions with the bottom MoS 2 layer. The work function of MoS 2 was measured to be −4.85, −4.96, −5.09, and −5.13 eV for layer numbers 1, 2, 4, and 8, respectively, as was noted in our previous report 38 . Therefore, the work function of the bottom MoS 2 layer was estimated to be approximately −5.0 eV. The work function of monolayer MoSe 2 is known to be 4.70 eV 39 . The heterojunction of Janus MoSSe with MoS 2 is classified into two types: S/S and Se/S heterostructures 16 . The former corresponds to our case because the Se layer is at the top of the MoSSe. The work function of MoSSe in the S/S heterostructure was calculated to be −4.26 eV 16 . Based on these considerations, the barrier height of the MoSe 2 /MoS 2 junction is lower than that of the MoSSe/MoS 2 junction, and this conclusion is consistent with the I-V behavior. These results are preliminary, so further in-depth characterization is needed to clarify the device properties of Janus MoSSe in future work.

Conclusion
We have shown the successful fabrication of Janus MoSSe by a new simplified method that employs a single NaCl-assisted process in a CVD apparatus. This approach involved three-step processes to sequentially activate Mo and S in situ to grow MoS 2 , form S vacancies in the topsurface MoS 2 layer, and activate Se to fill the vacancies by only altering the temperature. The MoS 2 layer was shown to remain at the bottom of the Janus MoSSe even after optimization, resulting in the formation of a MoSSe/MoS 2 heterostructure. Throughout these steps, NaCl helped lower the melting points of the constituent substances, while the T in each zone was properly controlled. A possible growth mechanism was proposed based on the results of another annealing experiment that did not supply activated atoms. This simple approach may open an easy way to fabricate various kinds of 2D Janus materials, which are very promising for their fundamental characteristics and applications in novel devices. b-d Position-dependent Raman spectra, shifts of the Raman peaks, and PL spectra. Here, any considerable position dependence in the Raman shifts is invisible, but as the position moves in the direction from the center to the vertex, the intensity ratio of the MoSSe/MoS 2 Raman A 1g peaks decreases, and the PL peak redshifts (see also Supplementary Fig. S9).