Designing artificial two-dimensional landscapes via atomic-layer substitution

Significance

Manipulating materials with atomic-scale precision is essential for the development of a next-generation material design toolbox. Tremendous efforts have been made to advance the compositional, structural, and spatial accuracy of material deposition and patterning. Here, we presented a new reaction pathway to implement the conversions of two-dimensional materials within the atomic-layer thickness at room temperature for electrical dipole manipulation. Not only could various Janus monolayer transition metal dichalcogenides with vertical dipole be realized, but also some heterostructures, including the dipole-nondipole heterostructures (MoS₂-MoSSe) and multiheterostructures (MoS₂-MoSSe-MoSeS-MoSe₂) within the same monolayer host structure are developed, in which the dipoles can be selectively patterned to be zero (MoS₂, MoSe₂), positive (MoSSe), and negative (MoSeS).

Abstract

Technology advancements in history have often been propelled by material innovations. In recent years, two-dimensional (2D) materials have attracted substantial interest as an ideal platform to construct atomic-level material architectures. In this work, we design a reaction pathway steered in a very different energy landscape, in contrast to typical thermal chemical vapor deposition method in high temperature, to enable room-temperature atomic-layer substitution (RT-ALS). First-principle calculations elucidate how the RT-ALS process is overall exothermic in energy and only has a small reaction barrier, facilitating the reaction to occur at room temperature. As a result, a variety of Janus monolayer transition metal dichalcogenides with vertical dipole could be universally realized. In particular, the RT-ALS strategy can be combined with lithography and flip-transfer to enable programmable in-plane multiheterostructures with different out-of-plane crystal symmetry and electric polarization. Various characterizations have confirmed the fidelity of the precise single atomic layer conversion. Our approach for designing an artificial 2D landscape at selective locations of a single layer of atoms can lead to unique electronic, photonic, and mechanical properties previously not found in nature. This opens a new paradigm for future material design, enabling structures and properties for unexplored territories.

The wide and rich physics of two-dimensional (2D) materials have led to the fabrication of heterostructures (1–5), superlattices (6–8), and twisted structures (9–11) with breakthrough discoveries and applications. During the past 3 y, a new class of atomic layer structures, in which the top layer chalcogen atoms within the three-atom-thick transition-metal dichalcogenides (TMDs) are replaced by a different type of chalcogen atoms, has been developed, resulting in a Janus monolayer with broken out-of-plane symmetry, inherit strain, and a variety of attractive properties (12–16). However, such a transition has been subjected to thermal energy input because the top chalcogen–metal bonds need to be broken in order for the substitution to occur. Previous strategies (12, 15, 16) usually break off the original chalcogen atoms by purely harsh energetic means and then form new bonds with the incoming replacements. Since the displacement energies of these chalcogen atoms in 2D TMDs range from 5 to 7 eV (17, 18), the process requires either knocking off the top chalcogens via the kinetic energy of atoms in a hydrogen (12) or Se plasma (16) or purely high temperature (to form sulfur vacancies) (15) to happen.

It is well known that thermal energy has been widely used to enable chemical reactions. However, high reaction temperatures incur severe limitations (for example, the types of substrate on which materials can be synthesized) and undesirable energy consumptions. There has been a continuing pursuit to reduce material synthesis temperatures while maintaining the quality of as-synthesized material (19–21). Recently, a room-temperature method to synthesis monolayer Janus WSSe and MoSSe by sulfurization from monolayer WSe₂ and MoSe₂ was reported (14). Nevertheless, the mechanism for such conversion was only hypothesized and has not been completely understood.

This paper reports our concurrent work for room-temperature atomic-layer substitution (RT-ALS), including the conversion of WSe₂ and MoSe₂ (2H phase) being sulfurized into Janus WSeS and MoSeS, WS₂ and MoS₂ (2H phase) being selenized into Janus WSSe and MoSSe, and the conversion of 1T’ MoS₂ being selenized into Janus 1T’ MoSSe. Using density functional theory (DFT) calculations, we were able to elucidate the evolution of free energies in each reaction step and present a clear understanding of why such a reaction can occur at room temperature. More significantly, by combining RT-ALS with lithography patterning and flip 2transfer, a completely new class of monolayer lateral heterostructures and multiheterostructures are designed and fabricated.

Fig. 1A illustrates our RT-ALS process. The presence of hydrogen radicals (22) produced by a remote plasma strip the chalcogenide atoms on the top layer in a gentle and chemical way. Meanwhile, the low-pressure system facilitates the supply of vaporized chalcogen substitutes (here illustrated with Se) to take the place of missing atoms, resulting in an asymmetric Janus structure of MXY (M = Mo or W, X = S or Se, and Y = Se or S) at room temperature.

Fig. 1.

RT-ALS with DFT calculation and comparison. (A) Schematic illustration of the RT-ALS process from monolayer MoS₂. (B) Schematics of the five key reaction steps for RT-ALS process (cartons from left to right): before H adsorption, 2 H adsorption and diffusion to the same S, formation of H₂S, desorption of H₂S, and Se occupation of the S vacancy. Purple, yellow, red, and green balls are Mo, S, Se, and H atoms, respectively. (C) Free energy of each step in B, relative to that of the first step. (D) Comparison of activation energy barrier between the RT-ALS strategy (red) and the conventional high-temperature substitution (blue). The axes of the two pathways correspond to distances (d) between the substituted S atom to Mo plane in each case. The vertical dashed lines (blue and red) indicate equilibrium S to Mo plane distances (1.54 Å for high T pathway, 1.64 Å for RT-ALS pathway). In RT-ALS process, the largest energy barrier occurs at a critical distance where H₂S starts to form. In the high-temperature pathway, the largest energy barrier corresponds to when S and Se are both disconnect with Mo.

Results and Discussion

In sharp contrast to previous severe strategies (12, 15, 16), the RT-ALS steers the reaction via a different pathway. We use DFT (within the generalized gradient approximation, including van der Waals corrections [PBE-D2] on a supercell containing 4 MoS₂ unit cells [larger supercells are included in the SI Appendix]) to examine two reaction pathways to illustrate the difference, as shown in Fig. 1 B–D and SI Appendix, Fig. S1D. One pathway is the conventional high-temperature substitution (replacing S by Se via thermal activation), and the other is our RT-ALS process. Due to the active roles of H radicals in the RT-ALS process, the energy barrier becomes much lower, and the process is overall exothermic. Fig. 1B exhibits that H radicals firstly adsorb on the upper surface of monolayer MoS₂ with adsorption energy of ∼0.5 eV/H (which further increases with the H coverage, as shown in SI Appendix, Fig. S1A). As the H coverage increases, two H atoms form bonds with one S atom (SI Appendix, Fig. S1B), and the H₂S desorbs, leaving an S vacancy. Surprisingly, this desorption process only involves a small activation barrier <0.5 eV (i.e., 12 kcal/mol), an order of magnitude smaller than the total bond energies of 3 Mo-S bonds (∼6 eV). In such a scenario, the removal of S is unlikely to be caused by collisions with hydrogen radicals because the kinetic energy of an H radical in the remote hydrogen plasma is about two orders of magnitude smaller than the displacement energy of a chalcogen atom (16), and the masses are very different between H and S atoms. The S vacancy is then occupied by a nearby Se atom. In contrast, the high-temperature pathway (SI Appendix, Fig. S1D) involves Se adsorption on surface, Mo-S bond breaking, and Mo-Se bond reforming, requiring an energy barrier of ∼2.5 eV that occurs when the Mo-S bond breaks (Fig. 1D). This calculated energy barrier matches well with the experimental findings, wherein the high-temperature substitution happens at ∼1,000 K (23). Such a calculation comparison clearly reveals that RT-ALS is made possible via reducing the reaction energy barrier, and the reaction kinetics is controlled by the active H radicals.

For RT-ALS experiments, we start with chemical vapor deposition (CVD)-synthesized monolayer MoS₂ and perform selenization for the top sulfur layer to obtain MoSSe. Afterward, we use flip-transfer to expose the bottom surface and perform RT-ALS a second time to achieve MoSSe conversion to MoSe₂ (SI Appendix, Fig. S2). From the optical microscopy (OM) images (SI Appendix, Fig. S2 B–D), the optical contrast of the MoS₂, MoSSe, and converted MoSe₂ triangular single crystals remain uniform and intact after RT-ALS, suggesting the process makes negligible disruption to the 2D lattice. This is also confirmed by atomic force microscopy (AFM) images (SI Appendix, Fig. S2 E–G), as their surfaces remain pristine and flat after each RT-ALS step. In addition, it is straightforward to extend the single/double-sided conversions to continuous large-area films (SI Appendix, Fig. S2 H–J). From a material synthesis perspective, here MoSe₂ is obtained from a MoS₂ template by room-temperature reaction, without the need of high-temperature conversions as in contrast to earlier approaches (21, 24), and thus avoids subsequent lattice or substrate damage. This provides an alternative route to obtain Se-based TMDs at room temperature. More importantly, this could be used to synthesize TMD materials that are hard to be directly grown by CVD or other methods.

To confirm the substitution of the top chalcogenide layer, measurements on Raman and photoluminescence (PL) spectroscopy are performed due to their sensitivity to the crystal structure. For TMDs, the A_1g and E_2g mode correspond to the out-of-plane and in-plane lattice vibrations, respectively. Therefore, changing the top layer of chalcogens produces different phonon frequencies for the same type of lattice vibrations. As shown in Fig. 2A, the MoS₂ A_1g (404 cm⁻¹) and E_2g (383 cm⁻¹) modes shift to 288 cm⁻¹ and 355 cm⁻¹, respectively, in Janus MoSSe¹² due to the change of atomic mass and the broken symmetry in the vertical direction (12). Further flip-transfer and RT-ALS on the other sulfur layer yield full selenization, with a sharp peak at ∼239 cm⁻¹ and a broad one at ∼284 cm⁻¹, which are consistent with the A_1g and E_2g modes in monolayer MoSe₂ (25). These conversions are further evidenced by the strong PL emission shown in Fig. 2B, in which the PL peak energy shifts from 1.85 eV (pristine MoS₂) to 1.72 eV (Janus MoSSe) and then to 1.60 eV (converted MoSe₂). All the observed PL peak energies are consistent with reported values of previous literature (12, 25). Besides, Raman mappings (SI Appendix, Fig. S3 A–C) and PL intensity mappings (Fig. 2B, Insets) also confirm that crystals produced by RT-ALS have high spatial homogeneity.

Fig. 2.

Characterizations of the monolayer (Janus) TMDs converted using RT-ALS and the conversion process. (A) Raman spectra of starting monolayer MoS₂, Janus MoSSe, and converted MoSe₂. (Insets) Crystal structures of MoS₂, Janus MoSSe, and MoSe₂, respectively. (B) PL spectra of starting monolayer MoS₂, Janus MoSSe, and converted MoSe₂. (Insets) Spatially resolved PL mappings of MoS₂, Janus MoSSe, and converted MoSe₂ flakes at 1.85 eV, 1.72 eV, and 1.60 eV, respectively. (C) Raman spectra of starting monolayer WSe₂, Janus WSeS, and converted WS₂. (D) PL spectra of starting monolayer WS₂, Janus WSSe, and converted WSe₂. (Insets) Crystal structures of WS₂, Janus WSSe, and WSe₂, respectively. (E) Tilted ADF-STEM image of a MoSSe sample to visualize the asymmetric atom structure in the vertical direction. The corresponding Mo, Se, and S atoms are schematically shown with blue, red, and yellow circles, respectively. (F) Intensity profile for the atomic chain highlighted in blue in E shows the intensity of individual Mo, Se, and S atoms. (G) Raman spectra collected at several monolayer Janus MoSSe samples treated with hydrogen plasma for 10 s, 30 s, 1 min, 3 min, 5 min, 10 min, and 15 min, respectively. (H) The relationship between the conversion time (10 s, 30 s, 1 min, 3 min, 5 min, and 10 min, respectively) and the corresponding intensity I_A1g(MoSSe)/I_A1g(MoS2) ratio over the conversion process. (Inset) Spatially resolved Raman mappings of two coalesced Janus MoSSe flakes converted within 30 s and 5 min, respectively, which show uniform Raman intensity over the whole MoSSe regions.

This RT-ALS is found to be a general strategy to obtain various Janus materials (including different transition metal atoms, chalcogen atoms, for converting only top layer or both top and bottom layers, and different crystal phases) through our experimental investigations. The conversions from WS₂ to Janus WSSe and then to WSe₂ were successfully obtained (SI Appendix, Figs. S2 K–M and S3 G–I), with the corresponding PL emissions shift from 2.01 eV (WS₂) to 1.83 eV (WSSe) and to 1.63 eV (WSe₂) (Fig. 2D). Compared with MoSSe, the larger elastic modulus and smaller carrier effective mass of WSSe may give rise to higher carrier mobility, especially for holes, due to the stronger spin-orbit coupling and overall cleaner optical signatures of W (16, 26). In addition, the conversion from monolayer MSe₂ (M = Mo and W) to Janus MSeS and then to MS₂ are realized by sulfurization (Fig. 2C and SI Appendix, Fig. S3 D–F) as well as the conversion from 1T’ MoS₂ to Janus 1T’ MoSSe (SI Appendix, Fig. S3 J and K). The evolutions of Raman peaks are all in consistent with literature (12, 16, 27) and our theoretical predictions, suggesting the universality of RT-ALS strategy.

In order to verify the atomic structure of Janus MoSSe, we used annular dark-field scanning transmission electron microscopy (ADF-STEM) to obtain the tilted images of MoSSe. The image was obtained by rotating the MoSSe sample 10° from the [001] crystallographic orientation to a new orientation that resulted in the Mo, Se, and S atoms projected in a straight line. In that way, the Mo, Se, and S atoms can be individually identified. As shown in Fig. 2E, the tilted ADF-STEM image enables to separate the Se and S atoms. The results show that the Se atoms are located on one side of the monolayer MoSSe, and S atoms on the opposite side, which is direct evidence of the Janus structure. The corresponding intensity profile in Fig. 2F clearly show the individual Mo, Se and S atoms with their total peak intensities proportional to their atomic numbers.

To achieve a better understanding of the conversion process, we investigated the evolution of the Raman peaks during the synthesis. The results indicated in Fig. 2 G and H, and SI Appendix, Fig. S4A show that within 10 s, the A_1g peak of Janus MoSSe begins to emerge [at a frequency ∼10 cm⁻¹ lower than the final point, most likely due to the modulation of phonon frequency from the local phonon vibrations at the reaction onset (26)]. The intensity of the Janus MoSSe A_1g mode increases with longer conversion time; meanwhile, the intensity of the MoS₂ A_1g peak decreases and eventually vanishes in 15 min. Based on the variation of Raman peaks at different time, the conversion yield (roughly defined by Y% = Intensity A_1g of MoSSe/[Intensity A_1g of MoSSe + Intensity A_1g of MoS₂]) from S to Se after RT-ALS treatment are estimated (SI Appendix, Fig. S4B). For material conversion through high-temperature substitution, previous reports (21) showed that the atomic replacement begins at energetically favorable sites (crystal edges, defect sites, and grain boundaries) and gradually progresses to the other regions of the flakes, indicating it is controlled by thermodynamics. In contrast, the room temperature operation here suggests it is a kinetic process, in which the energy needed for the atomic substitution is supplied by the highly reactive hydrogen radicals. This leads to the spatially homogeneous substitution of atoms over the whole flake since the reaction onset (Fig. 2H, Insets and SI Appendix, Fig. S4C) and a fast progression of the reaction with a time scale as short as 30 s (Fig. 2G and SI Appendix, Fig. S4A).

Taking advantage of the room temperature procedure of the whole process, the programmable design in monolayer TMD can be realized using lithography combined with flip-transfer techniques, as illustrated in Fig. 3 A–C. Arbitrary patterns in a 2D plane can be implemented by using a physical mask (here, lithographic patterns is just an example), enabling locally programmed out-of-plane atomic structures (Fig. 3A, where electron-beam lithography [EBL] can define arbitrary patterns using a polymethyl methacrylate [PMMA] mask). RT-ALS will convert the exposed areas to MSSe, allowing the formation of lateral heterostructure between MS₂ and Janus MSSe within the same host monolayer (Fig. 3B)_. Furthermore, the chalcogen on the other side can be replaced by using a flip-over transfer (Methods) and another selective RT-ALS conversion, resulting in lateral multiheterostructures of MS₂-MSSe-MSeS-MSe₂ (Fig. 3C) with tailored functionalities. Fig. 3D shows the scanning electron microscopy (SEM) image of monolayer Janus MoSSe-MoS₂ stripe patterns with different widths. Even though both MoSSe and MoS₂ are semiconductors with comparable bandgaps, MoSSe is brighter than MoS₂, possibly due to the intrinsic vertical dipole in the monolayer. Fig. 3E shows the spatially resolved PL mapping of Janus MoSSe with the pattern of the Massachusetts Institute of Technology mascot “Tim the beaver” made on a continuous monolayer CVD-MoS₂ canvas (RT-ALS converted the exposed regions to Janus MoSSe, while the masked regions remained as pristine MoS₂). The same pattern was characterized using various methods: OM image (SI Appendix, Fig. S5A), SEM image (SI Appendix, Fig. S5B), and Raman and PL mappings (SI Appendix, Fig. S5 C and D), which all clearly verify that MoS₂ and MoSSe are well-located as designed. ADF-STEM is used to characterize the junction, which shows that the MoS₂-MoSSe interface is seamlessly connected at the boundary defined by lithography (Fig. 3F and SI Appendix, Fig. S6A). Elemental diffusion (2, 4) is not anticipated for the RT-ALS strategy, as the substitution occurs at room temperature within a short time. As ADF-STEM intensity scales with the atomic number, the MoS₂ region with weaker intensity from the S-S column on the left side of Fig. 3F is clearly distinguished from the stronger intensity from the S-Se column of MoSSe on the right. Additionally, the different structures of MoS₂ and MoSSe are further verified by the intensity profile of electron scattering, where the atomic positions of Mo, S/Se, and 2S are corresponding to their gradually decreased intensities (Fig. 3 G and H). The energy-dispersive X-ray spectroscopy results (SI Appendix, Fig. S6 C–H) indicate that Se and S atoms are well-confined in the selective regions. As the selected area electron diffraction patterns show nonseparable patterns from these two lattices (SI Appendix, Fig. S7 A–D), real space strain mappings were further performed (SI Appendix, Fig. S7 G and H), in which the slight lattice difference in the MoSSe region can be observed compared to the reference MoS₂ region.

Fig. 3.

RT-ALS with programmable design in dipole/nondipole lateral heterostructures and multiheterostructure. (A) Schematic representation of programmable ALS process by lithography patterning and ALS process. (B) Programmable monolayer MoS₂-Janus MoSSe heterostructure on SiO₂/Si substrate. (C) A multiheterostructure composed of monolayer MoS₂-Janus MoSSe-Janus MoSeS-MoSe₂ regions. (D) SEM images of monolayer Janus MoSSe-MoS₂ stripe patterns with different widths. (E) Spatially resolved PL mapping for a Massachusetts Institute of Technology mascot “Tim the beaver” pattern (Janus MoSSe) on a monolayer MoS₂ canvas. (F) Pseudocolor ADF-STEM image of a MoS₂-MoSSe interface (Left), the zoomed-in views of a MoS₂ region (Right Top, highlighted in yellow) and a Janus MoSSe region (Right Bottom, highlighted in blue), and their respective intensity profiles in G and H. (I) OM image of monolayer multijunction composed with MoS₂-Janus MoSSe-Janus MoSeS-MoSe₂. (J) AFM topography image for the monolayer multiheterostructure of MoS₂-Janus MoSSe-Janus MoSeS-MoSe₂. (K) SEM image of monolayer multiheterostructure composed with MoS₂-Janus MoSSe-Janus MoSeS-MoSe₂. (L) Spatially resolved Raman mapping for A_1g mode intensity of monolayer multiheterostructures based on MoS₂-Janus MoSSe-Janus MoSeS-MoSe₂. The red area represents MoS₂ region, the blue area represents Janus region, and the green area represents MoSe₂ region.

Flipping over the MoS₂-MoSSe heterostructures and repeating selective RT-ALS on the other side yielded lateral multiheterostructures with MoS₂, Janus MoSSe and MoSeS, and MoSe₂ (Fig. 3 I–L and SI Appendix, Fig. S8). From the OM image in Fig. 3I, it can be seen that Janus MoSSe and Janus MoSeS has very similar optical contrast, while the AFM image Fig. 3J indicates the slight height increase after each RT-ALS step, consistent with the observation from SI Appendix, Fig. S2 E–G. Fig. 3K shows the SEM of the multiheterostructure regions. The two types of Janus regions (MoSSe [having vertical dipole pointing up] and MoSeS [having vertical dipole point down]) demonstrates clear intriguing contrast under SEM. These two types of regions should have very similar properties (e.g., material composition, bandgap, etc.) except the dipole direction, such observation indicates the role played by the dipole moment in material functionalities. In Fig. 3L, spatially resolved Raman mapping for A_1g mode intensity of such monolayer multiheterostructures is presented, with the red region being MoS₂, blue region being Janus (either MoSSe or MoSeS), and green region being converted MoSe₂.

In principle, the minimum feature size of RT-ALS should be determined by the resolution of lithography (Fig. 3D and SI Appendix, Fig. S5 E–H). With current EBL, features of hundreds of nanometers can be achieved, with potential for scaling down to tens of nanometers using extreme ultraviolet, helium-beam, or AFM-based lithography (28). The interface revealed by ADF-STEM (Fig. 3F) indicates edge roughness due to EBL limitations. However, sharp edges are anticipated if atomically sharp physical masks such as aligned carbon nanotubes (29, 30) can be used in the future. Nevertheless, even with the current EBL, many new device structures with unusual physical properties and potential applications can already be envisioned. For example, periodic arrays of these planar heterostructures with 100-nm-scale pitch can be defined as an atomically thin optical grating, and metasurfaces for efficient modulation or steering of visible/near-infrared light can be designed. In these structures, the extremely strong light-matter interactions in 2D materials and the optical nonlinearity in Janus TMD (31, 32) could play critical roles in expanding the design space of these atomically thin optical metasurfaces with unprecedented efficiencies or novel functionalities. The ability to create these artificial TMD multiheterostructures within a continuous atomic-layer “canvas” has never been achieved before and presents extremely exciting potentials.

A major application of 2D materials and their heterostructures lies in novel electronic or optoelectronic devices. With RT-ALS, we found that high-quality materials are obtained, and the energy misalignment of band edges exist at both the MoSSe-MoS₂ and MoSe₂-MoSeS lateral heterojunctions with a type-II band alignment. Multiple back-gate transistors with long and short electrodes were fabricated and four-probe measurements were used to eliminate the contact resistance (OM images in Fig. 4 B and C, Insets). Typical transfer characteristics (channel current I_D versus back-gate voltage V_G) are plotted in Fig. 4A and used to extract the threshold voltage V_T, defined as the intersection of the linear fit of the on-current and the x-axis, and the field effect mobility µ_FE, defined as µ_FE = (L/W)g_m/(V_DC_ox), where L and W are the channel length and channel width, g_m is the transconductance, V_D is the source-drain voltage (bias across the channel), and C_ox = 12.1 nF/cm² (capacitance of the 285 nm SiO₂ gate oxide). The V_T and µ_FE of tens of devices were measured for the starting CVD MoS₂, single-side-converted Janus MoSSe, as well as double-side-converted MoSe₂, summarized in Fig. 4A, Inset. Average µ_FE for MoS₂, Janus MoSSe, and ALS MoSe₂ are 8.94 cm²/Vs, 2.11 cm²/Vs, and 0.17 cm²/Vs, respectively. Average V_T for them are −1.9 V, 36.0 V, and 51.6 V, respectively. The gradual degradation of mobility from MoS₂ to MoSe₂ may arise from the introduction of defects through the RT-ALS or transfer processes. Note that the mobility value measured in our Janus MoSSe devices is two orders of magnitude higher than through the route of conventional high-temperature substitution (15), which is indicative of the high quality of materials from RT-ALS. Furthermore, the increasing of V_T values indicates that MoS₂ is the most n-type doped, whereas Janus MoSSe and MoSe₂ are decreasingly n-doped, in agreement with DFT calculated band alignment (Fig. 4F). This band alignment configuration further determines the electrical polarity of Janus MoSSe-MoS₂ and MoSe₂-Janus MoSeS lateral heterojunctions. The Janus MoSSe and the converted MoSe₂ regions are the anodes of these two n-n⁺ junctions, respectively, and confirmed with output characteristics (I_D versus V_M with different V_G, where V_M is the voltage drop across the two inner short electrodes [M1 and M2] as I_D is applied across the two long electrodes [D and S]), as shown in Fig. 4 B and C. The weak rectification behaviors observed on both lateral junctions suggest that the barrier heights (denoted as Φ_B) at the lateral heterojunctions are very low. We performed temperature-dependent four-probe transport measurement and used the thermionic emission model to extract the gate-dependent barrier heights (see Methods for details, and SI Appendix, Figs. S9 and S10), plotted in Fig. 4D, for Φ_B = 30 meV and 50 meV for MoSSe-MoS₂ and MoSe₂-MoSeS lateral heterojunctions, respectively, when the channels are turned on completely (V_G > 50 V).

Fig. 4.

Electrical properties of MoS₂, Janus MoSSe, MoSe₂, and their lateral heterostructures produced by ALS. (A) Transfer characteristics (current density I_D/W versus back-gate voltage V_G, V_D = 1 V) of typical MoS₂, single-side-converted Janus MoSSe, and double-side-converted MoSe₂. (Inset) Field effect mobility µ_FE versus threshold voltage V_T for tens of devices. The average mobility values for MoS₂, Janus MoSSe, and MoSe₂ are 8.94, 2.11, and 0.17 cm²/Vs, and the average V_T for them are −1.9, 36.0, and 51.6 V, respectively. (B and C) Four-probe output characteristics (current density I_D/W versus bias voltage V_M with various back-gate voltage V_G) for (A) MoSSe-MoS₂ and (B) MoSeS-MoSe₂ lateral heterojunctions. (Insets) Diagrams and OM images of these devices. (D) Barrier height Φ_B as a function of V_G for the MoSSe-MoS₂ (red) and the MoSeS-MoSe₂ (orange) lateral heterojunctions. (E) Kelvin probe force microscope image of the MoS₂-MoSSe-MoSeS-MoSe₂ multiheterostructure. The Janus MoSSe region with the Se layer on top and the MoSeS region with the S layer on top display the highest and the lowest surface potential, respectively, whereas the surface potentials for the MoS₂ and MoSe₂ regions are in between. (F) Conduction band minima and valence band maxima energies of MoS₂, Janus MoSSe, and MoSe₂ relative to vacuum, obtained from DFT calculations. For Janus MoSSe, the vacuum energy levels from upper and lower planes are averaged.

More uniquely, our RT-ALS approach brings about a programmable control of the out-of-plane electrical dipole in TMD materials (induced by the electronegativity difference between S and Se atoms in Janus MoSSe). We performed Kelvin probe force microscope measurements on the MoS₂-MoSSe-MoSeS-MoSe₂ grid samples (Fig. 4E) to observe clear surface potential differences among these four distinct materials. As expected, the Janus MoSSe region with the Se layers on top and the flipped MoSeS region with the S layers on top display the highest and the lowest surface potentials, respectively, whereas the surface potentials for the MoS₂ and MoSe₂ regions are in between. The measured potential difference between the top Se layer of the MoSSe region and the top S layer of the MoSeS region is around 100 meV. Such potential difference is even larger than that between the MoSe₂ and MoS₂ regions (which is around 15 meV), suggesting that the Janus MoSSe (or MoSeS) holds a large intrinsic dipole that is locked to the 2D surface normal. Note that this dipole can be selectively patterned to be zero (MoS₂, MoSe₂), positive (MoSSe), and negative (MoSeS) on such a 2D material “canvas,” which would enable many nanostructures and devices with intriguing electrical and optoelectronic properties. For example, with proper choice of physical masks, such as e-beam resist, or carbon nanotubes (28, 29) with atomically sharp edges, it is possible to make in-plane quantum wells, superlattices, or photonic devices by generating periodic arrays of such dipole/nondipole lateral heterostructures with 1- to 100-nm periodicity. Not only can the transport and optical properties of these MoXY multiheterostrutures be altered, any materials in close vicinity would also be modulated by such an electrostatic “canvas.” Unusual physical properties and unprecedented functionality may be possible on such a platform, ranging from nonlinear optics, electronic band engineering, and nanoscale origami to electrochemical catalysis, all of which can be modulated by different electrostatic forces.

In summary, the RT-ALS combined with patterning and flip transfer present a powerful yet universal strategy to program material properties at the atomic-layer limit. It allows for creating diverse artificial low-symmetry 2D materials and their heterostructures. Such a process generates seamless, high-quality interfaces between different structures and can complement existing vdW heterostructure fabrication techniques by adding another intriguing class of materials (Janus and lateral Janus heterostructures and multiheterostructures). Furthermore, by designing these heterostructure patterns and surface charge distributions, great potential is anticipated in new physical discoveries and future applications.

Methods

Synthesis of MoS₂.

A target SiO₂ substrate was suspended between two other SiO₂/Si substrates with predeposited perylene-3, 4, 9, 10-tetracarboxylic acid tetrapotassium solution. All of these substrates were placed facedown on a crucible containing MoO₃ precursor in a 1-inch quartz tube. This crucible was placed in the middle of the heating zone with another sulfur crucible on the upstream. Before heating, the whole CVD system was purged with 1,000 sccm Ar (99.999% purity) for 3 min. Then, 20 sccm Ar was introduced into the system as a carrier gas. The growth system was heated to 625 °C for 15 min. The MoS₂ growth was carried out around 620 to 630 °C for 3 min under atmospheric pressure. After growth, the whole system was naturally cooled down to room temperature.

Plasma-Assisted ALS.

We use a remote commercial inductively coupled plasma system to substitute the top-layer sulfur atoms of monolayer MS₂ (M = Mo, W) with selenium. The CVD grown monolayer MoS₂/WS₂ were placed in the middle of a quartz tube. The plasma coil made by a cylindrical copper tube placed at the upstream of CVD furnace. The distance between the sample and the plasma coil is around 8 to 10 cm, with the selenium powder placed around 5 cm away from the plasma generator on the other side. Whenever masks were needed, the sample was covered with PMMA patterns defined by EBL. At the beginning of the process, the whole system was pumped down to a low pressure to remove air in the chamber. Then, hydrogen was introduced into the system and the plasma generator was ignited. The hydrogen atoms assist the removal of the sulfur atoms on the top layer of MoS₂. At the same time, the vaporized selenium filled in the vacancy of the sulfur atoms, resulting in the asymmetric Janus structure of MoSSe. The whole process was performed at room temperature. After the reaction, the whole system was purged with Ar gas (99.999% purity) to remove the residual reaction gas, and the pressure was recovered to atmospheric.

Transfer.

For normal transfer, the samples were spin-coated with PMMA as a supporting layer. Then, they were put in the KOH solution and the PMMA/2D material was detached from the growth substrate and was floated on the surface of KOH solution. Afterward, it was taken to a deionized (DI) water bath by a glass slide and washed several times and then picked up with the target substrate. After baking on a hot plate at 100 °C for 15 min, the PMMA layer was removed with acetone and isopropanol (IPA). For flip-over transfer, the premade MoSSe sample was picked up by a polydimethylsiloxane stamp and then was released onto an intermediate substrate. Then, the MoSSe was released through etching the SiO₂ in a KOH solution. The film was washed with DI water for several times and then put upside-down on the target SiO₂/Si substrate, with the bottom-layer selenium atoms touching the substrate. At last, the PMMA film was removed by acetone and IPA.

Data Availability

All study data are included in the article and/or SI Appendix. All materials are available upon request to J.K.