Strategies for Controlled Growth of Transition Metal Dichalcogenides by Chemical Vapor Deposition for Integrated Electronics

Abstract

In recent years, transition metal dichalcogenide (TMD)-based electronics have experienced a prosperous stage of development, and some considerable applications include field-effect transistors, photodetectors, and light-emitting diodes. Chemical vapor deposition (CVD), a typical bottom-up approach for preparing 2D materials, is widely used to synthesize large-area 2D TMD films and is a promising method for mass production to implement them for practical applications. In this review, we investigate recent progress in controlled CVD growth of 2D TMDs, aiming for controlled nucleation and orientation, using various CVD strategies such as choice of precursors or substrates, process optimization, and system engineering. We then survey different patterning methods, such as surface patterning, metal precursor patterning, and postgrowth sulfurization/selenization/tellurization, to mass produce heterostructures for device applications. With these strategies, various well-designed architectures, such as wafer-scale single crystals, vertical and lateral heterostructures, patterned structures, and arrays, are achieved. In addition, we further discuss various electronics made from CVD-grown TMDs to demonstrate the diverse application scenarios. Finally, perspectives regarding the current challenges of controlled CVD growth of 2D TMDs are also suggested.

1. Introduction

The study of two-dimensional (2D) van der Waals (vdW) materials represented by graphene is one of the most interesting research topics because of their extraordinary solid-state physics¹⁻⁴ and application potential for next-generation nanoelectronics.^5,6 However, graphene as a zero band gap semimetal fails to play a role as a semiconducting active channel in electronics. Transition metal dichalcogenides (TMDs), with a layered structure similar to that of graphene, complement this disadvantage. Previous research has extensively investigated group VIB TMDs such as MoS₂, MoSe₂, WS₂, MoTe₂, etc., in both fundamental research⁷⁻¹¹ and industry.¹²⁻¹⁴ The majority of these group VIB TMDs are thermodynamically stable semiconductors with a sizable band gap and are commonly used as active channels in optoelectronics. In recent years, due to the development of producing technology, much research is also devoted to other novel TMDs. For instance, 2D ferromagnetism^15,16 and 2D ferroelectricity¹⁷ have been found on VSe₂ and twisted TMDs, respectively. Due to its large family and diverse nature, TMDs are undoubtedly one of the most promising branches of two-dimensional vdW to realize versatile applications in various scenarios like optoelectronics,^18,19 electrocatalysis,²⁰ membrane filtration,²¹ memories,²² etc.

Many synthesis routes for 2D TMDs have been widely explored, including mechanical exfoliation, chemical exfoliation, molecular beam epitaxy (MBE), and chemical vapor deposition (CVD). Mechanical exfoliation operated with Scotch tape obtains samples with high crystallinity, cleanliness, low defects, and controllable thickness.²³ However, this manual operation obviously cannot meet the needs of large-scale production, thus it is rarely used outside of cutting-edge fundamental research. Chemical exfoliation, such as liquid exfoliation²⁴ and intercalation exfoliation,²⁵ weakens the interlayer vdW force of TMDs by polar molecules or ions to exfoliate bulk TMDs to few-layer equivalents using a top-down method. Although mass production is achievable by chemical exfoliation, limited surface area, poor controllability, and toxicity remain challenges of this approach. CVD, the most representative “bottom-up” method, is perhaps the most promising route among these for efficient production. The CVD growth of TMDs still obeys classical nucleation theory (CNT)²⁶ and is mainly divided into two stages, nucleation and growth. Therefore, by reasonably controlling the growth conditions of the crystals, the synthesis of large-sized high-quality TMD thin films is completely possible by CVD. However, the controllable CVD growth of 2D TMDs is still a huge challenge, despite nearly 10 years of development. The main reason is that the CVD process involves a series of complex chain chemical reactions involving vapors and also involves complicated interactions between the vapor precursors, formed crystals, and the substrate. Generally, the CVD process is extremely sensitive to the local environment in the reaction chamber, and thus most studies related to the controllable CVD growth are based on experimental results, supplemented by an explanation of the mechanisms, although computational simulations related to it have only been preliminarily explored so far.²⁷⁻²⁹ In this regard, we review the recent works reporting the synthesis of 2D TMDs and 2D TMD heterostructures, with a focus on the strategies made to achieve a high degree of growth controllability via the CVD method. In addition, mechanistic insights are also provided to understand the chemistry in the controllable CVD growth. To better understand the necessity and significance of synthesizing 2D TMDs, the structure and properties of 2D TMDs are introduced first. Then several of the most common types of CVD are discussed, with a focus on how to achieve controllable growth of pure TMDs by choosing and designing precursors or substrates. To further illustrate the advantage of CVD for controlled synthesis, the growth of more complex products such as TMD heterostructures and patterned TMD arrays are presented, involving various technologies, smart strategies, and design routes. At the end of this review, a brief summary of the CVD-grown TMD applications is presented to better understand what practical use can be achieved by controlled CVD growth.

2. Structure and Properties of 2D TMDs

Transition metal dichalcogenides have structures adopting the MX₂ formula, where M and X refer to transition metal atoms (M = Mo, W, Nb, V, Re, Ta, etc.) and chalcogen atoms (X = S, Se, Te) separately. Unlike graphene, which is composed of only carbon atoms in a plane, TMDs have a sandwich structure with chalcogen atoms on top and bottom and are covalently bonded to transition metal atoms in the middle. According to the coordination, TMD monolayers are classified into 1H (trigonal prismatic) phase and 1T (octahedral) phase. Beyond basic 1H and 1T phases, polymorphs including 1T′ (distorted 1T), T_d (orthorhombic), 2H, and 3R are discovered. 1T′ and T_d are derivatives of 1T, and they differ from 1T in intralayer coordination between chalcogen atoms and transition metal atoms. 2H and 3R originate from multilayers stacked with 1H monolayers by ABA and ABC sequences,^30,31 respectively. Generally, phase transition leads to the change but is not the decisive factor in the electrical properties of TMDs. The most typical example is the phase transition of MoTe₂, from the metallic 1T′ phase to the semiconducting 2H phase.³² The difference in electronic properties comes from the filling level of nonbonding d orbitals of transition metals. When no partially occupied nonbonding orbital exists, the TMD is semiconducting, e.g., 2H-WS₂ and 1T-PtS₂; otherwise, it is metallic like 1T-ReS₂ and 2H-TaS₂.³³ The phase of TMDs could be engineered by diverse approaches such as intercalation,³⁴ strain,³⁵ and thermal treatment.³²

3. Controllable CVD Growth of TMDs

TMDs with wafer-scale size, uniform thickness, and high crystallinity are the main considerations for future material development, aiming for optimizing the fabrication and performance of electrical and optoelectronic devices. The morphology of the TMD films will be easily modified by the growth parameters, such as the type and structure of the substrate,³⁶ reaction temperature,³⁷ flow rate,³⁷ type of precursors,³⁸ etc. To reach the controllable CVD growth of wafer-scale, grain-boundary-free, and highly uniform TMDs, precise research should focus on atmospheric pressure CVD (APCVD), low-pressure CVD (LPCVD), and metal-organic CVD (MOCVD).

3.1. Thermodynamics and Kinetics in CVD Growth of TMDs

In CVD synthesis, very small change in parameters may largely vary the kinetics and thermodynamics of TMD growth, leading to drastic changes in the products. It is necessary to explore a deeper understanding of the kinetic and thermodynamic factors in the CVD growth mechanism of such 2D materials. The impact of thermodynamics and kinetics on the vapor deposition process is mainly reflected in two local parameters: (1) reaction temperature and (2) partial pressure around the substrate.

To date, thermodynamic and kinetic studies based on the gradient temperature growth of TMDs via the CVD mechanism have been explored.^26,39 Detailed thermodynamic and kinetic studies based on the CVD mechanism were performed and found that the crystal nucleation density significantly depends on the temperature.⁴⁰ This phenomenon can be explained by the classical nucleation theory.

where N is the nucleation rate, T is the zone temperature, P is the partial pressure of the Mo species, E_des is the desorption energy for molecules back to the vapor phase, E_s is the activation energy for surface diffusion, and ΔG* is the CNT free energy. The nucleation growth of monolayer MoS₂ in the low-temperature range, in which crystal nucleation density increased, is in contrast to the high-temperature range. For a high-temperature reaction, the partial pressure of the precursor is regarded as a constant due to almost complete sublimation, thus the nucleation rate is exclusively determined by temperature (Figure 1g).⁴¹ The nucleation density undergoes inhibition as substrate temperature increases, whereas the supersaturation is satisfied.⁴² In contrast, both partial pressure and temperature contribute to the nucleation rate and density and thus the grain size of MoS₂ at a relatively low temperature, due to the unsaturated local concentration of the precursor (Figure 1a–f).

Figure 1.

Thermodynamics and kinetics in CVD growth of TMDs. (a–e) Optical images of monolayer MoS₂ with growth temperature from 290 to 430 °C and growth time of 6 h. (f) Evolution of nucleation density of CVD-grown MoS₂ with temperatures. Reprinted with permission from ref (40). Copyright 2021 Wiley-VCH. (g) Relationship between average flake size and growth temperature of monolayer WSe₂, showing suppressed nucleation density with high temperature. Reprinted from ref (41). Copyright 2015 American Chemical Society. (h) Plot of edge length of WSe₂ against time, showing the growth rate is affected by temperature-dependent partial pressure of the precursors. Reprinted with permission from ref (45). Copyright 2017 Wiley-VCH. The temperature and partial pressure as parameters jointly affect the nucleation and growth in the CVD process of TMDs.

The partial pressures of the vaporized solid precursors can be controlled by different approaches, such as gas flow rate adjustment,⁴² zone temperature modification,⁴⁰ precursor concentration modulation,⁴³ etc., and different kinetic regimes exist, leading to the production of the desired TMD monolayers. For instance, the stable TMD crystal growth under thermodynamic and kinetic control can be balanced by an increasing gas flow rate, thus modifying the partial pressure of the transition metal source along the gas flow, and the result shows a strong dependence of TMDs’ crystal growth rate on the precursors’ partial pressure.⁴⁴ A study investigated the effect of precursor partial pressure on the MoS₂ growth rate and crystal domain shape and found that increasing the MoO₃ mass transfer under a high gas flow rate has an impact on the faster crystal growth rate and showed the influence on the kinetic growth dynamics of edges.⁴² Moreover, the partial pressures of precursors depend on the temperature, which determines the decomposition and sublimation rates of the precursors. As shown in Figure 1h, faster W source supply can be achieved by increasing the growth temperature while ensuring sufficient Se feeding, leading to a higher growth rate of WSe₂.⁴⁵

3.2. Atmospheric Pressure CVD

APCVD is a vapor deposition process under a normal pressure environment. The TMD materials were grown by delivering evaporated precursors, usually inorganic and less toxic, to the substrate surface.⁴⁶ The control of nucleation density, oriented growth, numbers of layers, crystal shapes, domain size, etc. is considered to be the major direction for desired TMD growth. In order to control the nucleation and oriented growth in APCVD, in addition to the parameters, such as temperature, airflow, etc., the main consideration is to select a proper substrate and its placement.

Recent research discussed the distribution of a suitable substrate in TMD growth by controlling desired nucleation density, orientation, and domain size. As shown in Figure 2a, the highly uniform, epitaxial wafer-scale, single-crystalline growth of a MoS₂ monolayer was obtained on a Au(111) substrate.⁴⁷ Gold foil was melted and resolidified on tungsten foil to obtain the single-crystalline Au(111) film and to produce unidirectional nucleation–growth of the single-crystalline MoS₂ wafer. Similarly, recent research achieved the production of millimeter-sized MoSe₂ crystals with high uniformity on a molten glass substrate.⁴⁸ An ultrasmooth and liquid-like isotropic surface was generated by increasing the temperature to 750 °C. The successful growth of a large domain crystalline MoSe₂ monolayer on the molten glass substrate is due to the suppression of nucleation (Figure 2b). As a result, the nucleation density reduction led to a large-sized crystalline TMD, illustrating the importance of a suitable substrate for the APCVD system.

Figure 2.

Controllable nucleation and epitaxial growth of TMDs via APCVD. (a) Evolution of wafer-scale unidirectional MoS₂ monolayer on a Au(111) substrate. Reprinted from ref (47). Copyright 2020 American Chemical Society. (b) Growth scheme of large-scale single-crystalline MoSe₂ on a molten glass substrate. Reprinted from ref (48). Copyright 2017 American Chemical Society. (c) Growth manner of MoS₂ film via substrate control. Reprinted with permission from ref (49). Copyright 2016 IOP Publishing Ltd. (d) Schematic illustration of the growth of MoSe₂ with the use of PTAS seed promoter. Reprinted with permission from ref (50). Copyright 2018 The Royal Society of Chemistry. Using a suitable substrate and adjusting the substrate locations and orientations, the morphology of the TMD grains can be effectively controlled.

Other than the substrate selection, the location, position, and orientations of the substrate also play an important role. The production of large-scale monolayer molybdenum disulfide film, up to centimeter scale, was achieved on the surface of a vertically positioned silicon substrate.⁴⁹ Unlike the horizontal substrate, the vertical substrate enables uniform seeding of the MoO₃ precursor on the substrate (Figure 2c). The position height between the substrate and the precursor is another key factor in the APCVD system that determines nucleation density and shape. The growth of different types of MoSe₂ nanosheets, including two-dimensional triangle domains, hexagons, three-dimensional pyramids, and vertically aligned grains, was achieved. The CVD growth of MoSe₂ was processed at 750 °C with the help of a seed promoter (tetrapotassium perylene-3,4,9,10-tetracarboxylate). The tailored morphologies were achieved by adjusting the height between the silicon substrate and solid precursor, resulting in the production of differently shaped TMDs (Figure 2d).⁵⁰ Moreover, the role of distance between the precursor source and substrate in CVD morphology was also shown to be vital. Another TMD shape-controlling method, by varying the spatial locations of the silicon substrates in the furnace tube, was reported,⁴² and it found that the shape of the MoS₂ domain was highly associated with the locations of the SiO₂/Si substrate. The shape of MoS₂ changed from triangular to hexagonal when the distance from the substrate away from the MoO₃ powder was varied. Compared to the silicon substrate, recent research showed a significant development in domain size and uniformity of TMD films using other types of substrates. Further development of the APCVD engineering system to fabricate the TMD materials with better uniformity and quality is still highly desired. APCVD is the most widely explored CVD technique and has potential for the development of high efficiency, simple setup, and low-cost 2D TMD film fabrication with large-scale and high uniformity. However, the sulfurization and selenization process requires a high growth temperature of greater than 750 °C, where the high processing temperature limited the use of some specific substrates,⁵¹ like polymeric substrates. The melting temperature of many polymeric substrates, such as polyamide, was below the CVD processing temperature of TMD synthesis.⁵² The additional transfer is necessary and thus limits the applications of polymeric films in TMD flexible devices.

3.3. Low-Pressure CVD

LPCVD growth of TMD films is processed under a low-pressure environment. Lower pressure can reduce unnecessary gas-phase reactions, leading to the increase of the uniformity of TMD films on the wafer.⁵³ As the domain size of TMDs is highly related to the nucleation density, the concentration of precursors and the type of substrates are the noteworthy factors to control the nucleation, to enable epitaxial growth, and to eliminate the grain boundaries of CVD-grown TMDs. Based on the recent development of LPCVD synthesis technology, the growth of large domain size single-crystalline TMDs with limited grain boundary films can be obtained.

In LPCVD, the selection of a suitable pretreated substrate still plays a significant role in the synthesis of TMDs with controlled domain orientation and alignment. Many research groups discovered that the use of suitable single-crystalline substrates, such as a-plane sapphire,⁵⁴ c-plane sapphire,⁵⁵ GaN,⁵⁶ etc., is the key factor in the growth of satisfactory oriented and aligned TMD films. A dual-coupling-guided epitaxy growth of WS₂ on a step-edged a-plane sapphire under a 300 Pa Ar atmosphere demonstrated the formation of large-sized single-crystalline WS₂ grains with unidirectional alignment and seamless stitching.⁵⁴ They found that the epitaxial stitching of WS₂ on the substrate of a-plane Al₂O₃ depends on the dual-coupling-guided mechanism, and the theoretical calculation was analyzed by density functional theory (DFT). Due to the coupling between WS₂ grains and a-plane Al₂O₃ and the coupling between WS₂ and a step-edged substrate, unidirectional alignment of WS₂ with a 2 in. single-crystalline monolayer was achieved (Figure 3a). Similarly, another report demonstrated the TMDs’ epitaxial growth of unidirectionally aligned MoS₂ single crystals on step-edged c-plane sapphire (0001) under the LPCVD processing system.⁵⁵ The orientation of step-edging was perpendicular to the standard sapphire substrate, and MoS₂ was grown on the C/A α_A = 0.89° sapphire substrate (Figure 3b–d). Using another single-crystalline substrate, β-Ga₂O₃ film, a controlled MoS₂ growth via ledge-directed epitaxy-assisted chemical vapor deposition processing under 30 Torr was achieved.⁵⁷ The single-crystalline monolayer MoS₂ was formed as a high-quality nanoribbon with continuous arrays and aligned orientation. It is demonstrated that the nucleation of MoS₂ occurred at the ledges of the β-gallium(III) oxide substrate, followed by the merging of an aligned MoS₂ domain, and the synthesis of self-aligned continuous nanoribbons was achieved ultimately (Figure 3e).

Figure 3.

Controllable nucleation and oriented growth of TMDs via LPCVD. (a) Schematic of the CVD processing of large-size unidirectionally aligned WS₂ on step-edged a-plane sapphire. Reprinted with permission from ref (54). Copyright 2022 Springer Nature. Schematic of MoS₂ domains growth on (b) C/M and (c) C/A sapphire and (d) optical image of MoS₂ monolayer growth in step-edged C/A sapphire. Reprinted with permission from ref (55). Copyright 2021 Springer Nature. (e) Schematic of the self-aligned growth of continuous MoS₂ nanoribbons on β-gallium(III) oxide substrate. Reprinted with permission from ref (57). Copyright 2020 Springer Nature. (f) Schematic diagram of the face-to-face precursors feeding LPCVD growth of MoS₂ on the soda-lime glass substrate. Reprinted with permission from ref (58). Copyright 2018 Springer Nature, open access. (g) Schematic diagram of multisource CVD system for large-size MoS₂ kinetic controllable LPCVD growth. (h) Optical image of wafer-scale (4 in.) monolayer MoS₂. Reprinted from ref (60). Copyright 2020 American Chemical Society. The controllable nucleation and oriented LPCVD growth were developed by the use of suitable substrates and reaction programming, leading to the improvement in uniformity and quality of TMD films.

Apart from the oriented growth, a suitable substrate efficiently determines nucleation density, similar to APCVD growth. Batch production of MoS₂ is demonstrated by an efficient growth of MoS₂ by face-to-face precursor feeding.⁵⁸ Using soda-lime glass as the substrate and oversaturated S precursor, the evaporation of Mo source can be limited and fabricate large domain size 6 in. MoS₂ (Figure 3f). It is concluded that selecting a suitable substrate controls the nucleation density, and the single-crystalline substrate with the pretreated surface can promote the oriented and aligned growth of high uniformity TMDs.⁵⁹

Moreover, the growth of TMDs with high uniformity and large size can be obtained by modulation of the reaction system, e.g., process optimization, and system engineering. The production of high-quality and large domain size monolayer MoS₂ film on the sapphire wafers by multisource system design of LPCVD was reported.⁶⁰ To ensure steady evaporation, the multipocket sources were designed to achieve facile evaporation of precursors and growth of TMD films. As shown in Figure 3g,h, S and MoO₃ sources were loaded in several inner-miniature quartz tubes to achieve the precise kinetic control for the growth of single-crystalline MoS₂. To control the nucleation density by adjusting the precursor concentration,⁶¹ a partial pressure programming LPCVD system was reported.⁶² The concentration and distribution were controlled by changing the amount of solid MoO₃ and the height between substrate and precursor. By controlling the flux of the Mo species, a reaction transition for MoS₂ growth was achieved, and the MoS₂ crystal was changed from a gradient-distributed triangular shape to a large-sized continuous monolayer. The vitality of metal precursors in the LPCVD-growth system was well-proven due to its role in controlling the layer number and thickness uniformity.

The low processing pressure leads to the improvement in the uniformity and purity of the produced TMD films. In contrast to APCVD, LPCVD is able to reduce the rate of gas-phase prereaction and film deposition and thus facilitates the TMD synthesis with better quality and uniformity and has the potential to produce a large-scale continuous film.⁶² Nevertheless, for growth control of wafer-sized TMD films, from isolated domains to a large continuous film, the systematic understanding of the growth mechanism is still not clear.⁶³ The commercial purposes of the semiconductor industry prefer the high uniformity and continuous films up to wafer-scale, but most LPCVD-grown TMDs still cannot meet the requirement. The controllable LPCVD fabrication of high uniformity and wafer-scale TMD films is still a key challenge and prospective direction in further research. In addition, the high processing temperature of the LPCVD system also limits the use of low melting temperature substrates.

3.4. Metal-Organic CVD

MOCVD is the CVD technology that uses organic metals, such as metallocene complexes, metal carbonyls, cyclopentadienyl ligands, etc., as the precursors.⁶⁴ The nucleation, epitaxial growth, and control of film thickness are the major prospective directions in recent MOCVD research that contribute to the improvement of coverage and quality of TMD films.⁶⁵ To control the lateral epitaxial growth in the MOCVD system, the precursor and the process must be designed elaborately.

The variety of precursors directly correlates with the concentration of metal source and chalcogen source and results in the varying of morphology, lateral growth, and quality of the films. A recent report demonstrated the production of few-layered MoS₂ growth by a pulsed MOCVD system with the use of bis(tert-butylimido)bis(dimethylamindo)molybdenum [(NtBu)₂(NMe₂)₂Mo] and diethyl disulfide (Et₂S₂) as the precursors.⁶⁶ To achieve thickness control of the MoS₂ film, the deposition process was conducted by several precursor injections (n = 0, 15, 50, 100) and resulted in the achievement of 1–25 nm MoS₂ film in a very short deposition time (Figure 4a). Apart from the selection of precursors, the precursor flow ratio also plays a role in the MOCVD system. The synthesis of WSe₂ films with outstanding improvement in surface coverage, thickness uniformity, and crystalline quality was reported.⁶⁷ The extreme precursor ratio of (CH₃)₂Se/W(CO)₆ was set at 14000, and the lateral growth and nucleation of the TMD were controlled, resulting in the fabrication of high uniformity films (Figure 4b–d).

Figure 4.

Controllable lateral epitaxial growth of TMDs via MOCVD. (a) Overview of the pulsed MOCVD processing design for thickness-controlled MoS₂ growth. Reprinted from ref (66). Copyright 2017 American Chemical Society. (b) AFM scan of WSe₂ film on the scale of 1–2 μm without a nucleation step (low density and coverage). (c,d) AFM scan of WSe₂ film with nucleation and growth steps, showing an increase in density and coverage with different temperatures. Reprinted with permission from ref (67). Copyright 2016 IOP Publishing Ltd. (e) Schematic of the top view of the substrate with unidirectional WS₂ domains. (f) Composite dark-field TEM image and the inset with higher resolution confirms well-aligned crystal orientation. Reprinted from ref (69). Copyright 2021 American Chemical Society. (g) Schematic of the MOCVD reactors shows the sapphire in a hot-wall tube reactor and hypothesized mechanism of NaCl-assisted growth of early stage and late stage. Reprinted from ref (68). Copyright 2018 American Chemical Society. In order to obtain the wafer-size and high-uniformity TMD films, the MOCVD growth was developed by the use of suitable precursors and well-designed growth programming and processing steps.

The important role of alkali-assisted precursors in CVD processing is also demonstrated, which leads to the impacts in growth morphology, including growth rate, loss of epitaxy, nucleation density, etc. Production of two-dimensional TMD films with a large domain size and fully coalesced polycrystalline structure was successfully achieved.⁶⁸ The MoS₂ film was synthesized in the MOCVD hot-wall tube reactor using Mo(CO)₆ and DES as the metal and chalcogen precursors with the utilization of alkali salts, such as sodium chloride and potassium iodide. Comparing the epitaxial growth of MoS₂ under alkali-free growth and alkali-assisted growth, the dramatic increase in the domain size of the MoS₂ monolayer by ∼20-fold was achieved by NaCl alkali-assisted growth. However, the alkali-assisted growth led to the saturation of Na–O bonding on the substrate surface and resulted in the reduction of substrate surface energy, weakening of the epitaxial growth, and the occurrence of multilayered MoS₂ film (Figure 4g).

In order to control the epitaxial growth of TMD films in the MOCVD system, modifying the reaction process is also important to promote large-scale TMD films. A production of large-scale single-crystalline WS₂ continuous monolayer films with control of epitaxial growth and stitching of the oriented domains was realized.⁶⁹ With the use of a variable-temperature multistep process, the coalesced unidirectional WS₂ film grew on the 2 in. diameter c-plane sapphire by MOCVD. With the help of multistep programming, control of the surface diffusion, epitaxial domain growth, and self-stitching of domain growth were achieved (Figure 4e,f).

In addition to the controlled epitaxial growth, the TMD growth on the selective area can also be achieved using the well-designed processing treatment. A recent report demonstrated a method for controlling the TMD lateral growth on desired regions by a seed-free, site-specific nucleation MOCVD.⁷⁰ In order to achieve the selective area growth of TMDs, a multistep surface functionalization approach was developed. The polymer functional layer was formed by the conventional lithographic process, and the sapphire surface is used as the original photoresist mask, followed by MOCVD growth at 825 °C. A route for the controlled TMD growth with exact patterns was demonstrated, and the device performance of the TMD film is comparable to that of the film produced by mechanical exfoliation.

The MOCVD method provides a new approach for manufacturing TMD films with a faster deposition rate. The significant advantages of MOCVD are its capability for large-scale TMD film fabrication with excellent conformal coverage and its ability to synthesize new metastable materials, including oxide,⁷¹ carbides,⁷² silicide,⁷³ etc. MOCVD also showed the potential for versatile TMD growth under relatively low temperatures^40,74 and its ability to promote the use of a wide range of low melting temperature substrates, including polymeric film. However, it is worth noting that the organic metal precursors and other reaction sources are usually highly toxic. Apart from the toxic precursors, the deposition process will also lead to the emission of highly hazardous exhaust gas, such as arsine, phosphine, etc. The low production rate, high production cost, and toxic organic precursors limit its application and prohibit its development.

4. Controllable CVD Growth of TMD Heterostructures

The conventional transfer method cannot satisfy the scalable production of TMD heterostructures, limiting its potential in practical technologies. In order to solve this problem, the current approach is to achieve the direct growth of heterostructures via the CVD method. In terms of the way of stacking, the CVD method not only fabricates vertical heterobilayers or multilayers like mechanical exfoliation but also prepares the in-plane (lateral) heterojunctions. To date, obtaining these two structures by rational design of the synthesis steps and parameters is an important issue in controllable synthesis because of the huge differences in performances of the devices.

4.1. Vertical TMD Heterostructures

2D materials with a dangling-bond-free van der Waals structure break the limitation of the lattice-matching condition of heteroepitaxial growth and thus are promising candidates for vertical heterostructures.^75,76 The growth of TMD-based vertical heterostructures has been intensively developed in the past few years, including TMD/graphene,⁷⁷⁻⁷⁹ TMD/h-BN,^80,81 TMD/perovskite,^82,83 TMD/iodide,^84,85 TMD/TMD, etc. Among them, TMD/TMD is worthy of great attention, as TMDs with a wide range of band gaps⁸⁶ enable heterojunctions with distinct interfaces to be obtained, including metal–metal, semiconductor–semiconductor, and semiconductor–metal, adapting the requirements of diverse electronics and optoelectronics. The growth of the TMD heterostructure is mainly classified into one-step and two-step (or multistep) routes, which were first implemented in 2014⁸⁷ and 2015,⁸⁸ respectively. One-step growth is more suitable for constructing heterojunctions with two TMDs of similar growth chemistry, including MoS₂–WS₂/WS₂,⁸⁹ NbS₂/MoS₂,⁹⁰ ReS₂/WS₂,⁹¹ and MoTe₂/MoS₂.⁹²

Considering the huge differences in the synthesis conditions of the various TMDs, the two-step route is more universal for vertical TMD heterostructures, although its procedures and growth strategies are more complex. The bottom layer should resist high temperatures and complex chemical changes in the second growth process, thus a growth temperature not too high in the second step is crucial to protect the bottom layer. However, the temperature of the second growth process is not as low as possible. Some attempts⁹³ have been made to change the growth sequence of WS₂ and MoS₂ in a two-step approach, and they found, when following the WS₂–MoS₂ growth sequence, that MoS₂ tended toward in-plane epitaxial growth at the edges of the existing WS₂ triangular single crystal. It was found that while using the MoS₂–WS₂ sequence, WS₂ nucleated from the edge to the crystal center of the MoS₂ triangle to form a vertical bilayer (Figure 5a,b). In the second growth, when MoS₂ is produced under a relatively lower temperature (660 °C), the nucleation rate of MoS₂ on silicon wafers is greater than that on WS₂, so the formation of in-plane heterojunctions is energetically favored. Meanwhile, the second growth temperature is set to a higher growth temperature for WS₂ (800 °C), and the nucleation of MoS₂ is promoted faster, leading to the formation of a rotation misfit-free heterobilayer. In addition to temperature, hydrogen flow is another key factor as well in determining whether a vertical bilayer can be formed, and the absence of hydrogen during the growth process is more conducive to vertical heteroepitaxy⁹⁴ in a two-step route.

Figure 5.

Controllable growth path of vertical TMD heterostructures. Growth mode of (a) WS₂–MoS₂ using WS₂–MoS₂ growth sequence and (b) WS₂/MoS₂ using MoS₂–WS₂ growth sequence, with the optical images. Reprinted with permission from ref (93). Copyright 2015 Wiley-VCH. Optical images of (c) partially covered and (d) fully covered VSe₂/WSe₂. Reprinted from ref (96). Copyright 2019 American Chemical Society. (e) Schematic illustration of substitution of sulfur to multilayer 2H-MoTe₂. (f) Cross-sectional STEM image of symmetrical MoS₂/MoTe_2(1–x)S_2x/MoS₂. Reprinted from ref (99). Copyright 2021 American Chemical Society. The two-step method is the main CVD method to obtain vertical heterobilayer and the top layer usually starts their growth from the edge of the bottom. Other ways such as chalcogen substitution also enable the preparation of vertical heterostructures.

The two-step route is more versatile and flexible in terms of the product in preparing vertical TMD heterostructures. First, compared to the one-step method, the two-step method avoids cross-contamination and alloying during deposition. Second, the two-step method has realized the tuning of staking modes, providing an optimized platform to study twist-angle-dependent optical physics.^9,95 Third, it also promotes the stacking of TMDs with distinct lattices and phases, implying that metal–semiconductor contact is possible. Several groups have used the two-step approach to obtain a metallic semiconducting vertical heterostructure. A direct synthesis of metallic 1T-VSe₂ onto semiconducting 1H-WSe₂ has been reported and found an inward growth mode of VSe₂ from the edge of WSe₂ to the center (Figure 5c,d).⁹⁶ Similar work has been carried out extensively in recent years, including the preparation of VSe₂/MoSe₂,⁹⁶ NbS₂/WS₂,⁹⁷ and ReS₂/MoS₂.⁹⁸ According to their band alignments, these heterojunctions either with Schottky or Ohmic contact are valuable for different electronics applications.

In addition to directly depositing TMDs layer by layer with one- or two-step routes, partial substitution is a reliable way to synthesize vertical heterojunctions, as well. In our recent work,⁹⁹ Te atoms of crystal 2H-MoTe₂ multilayer were replaced inward from the top and bottom layers sequentially, and a sub-centimeter-scale single MoS₂/MoTe_2(1–x)S_2x/MoS₂ sandwich structure was obtained, as presented in Figure 5e,f. This provides an efficient way to obtain large-scale vertical TMD heterostructures with scant grain boundaries.

4.2. Lateral TMD Heterostructures

The lateral TMD heterostructures arouse great scientific interest for their unique atomically sharp interface, where the two sides connect covalently and share one crystal orientation showing a clear and regular boundary without much atomic diffusion.^100−102 Therefore, they are selected candidates in building blocks for diodes and other junction devices utilized in photovoltaics and photodetection.^103,104

For the preparation of lateral TMD heterostructures, the one-step approach is widely accepted for its high efficiency and controllability. The growth of MoSe₂–WSe₂ with lateral heteroepitaxy was reported by physical vapor transport.¹⁰¹ Later, the preparation of MoS₂–MoSe₂ and WS₂–WSe₂ lateral heterostructures via an in situ modulation of the vapor-phase precursors was systematically reported.¹⁰⁵ MoS₂ (WS₂) was synthesized at the first stage to expose peripheral edges where fresh dangling bonds form and serve as the active sites of the following growth of MoSe₂ (WSe₂) (Figure 6a). The authors concluded that there are two key factors for the successful synthesis of lateral TMD heterostructures. The first factor is the in situ modulation, which means a rapid switch of the precursor vapor after the first-stage growth and prevents the termination and passivation of edges of TMD formed first. The second factor is optimized growth conditions for both sides of the lateral heterostructures, as already formed 2D TMDs are very sensitive to the changes in the chamber environment. Almost at the same time, MoS₂/WS₂ in-plane heterojunctions were successfully synthesized by the CVD method, as well.¹⁰² It is interesting that they found MoS₂/WS₂ growth can be tuned by adjusting the deposition temperature (Figure 6b). A higher temperature (850 °C) is preferable to obtain a vertical bilayer, while a lower temperature (650 °C) enables the formation of a lateral monolayer. The temperature-selective growth may be explained thermodynamically and kinetically. While ensuring sufficient nucleation and growth rates of both MoS₂ and WS₂, the vertical stacking is more thermodynamically favored, thus it is more stable at a high enough temperature. However, the low temperature significantly prevents the nucleation and growth of WS₂, which causes the epitaxial domain growth from the edge of MoS₂ with less nucleation energy.

Figure 6.

Controllable growth path of lateral TMDs heterostructures. (a) Illustration of in-plane epitaxy of MS₂–MSe₂ (M = Mo or W) in one-pot growth. Reprinted with permission from ref (105). Copyright 2014 Springer Nature. (b) Growth-temperature-dependent architecture of heterostructure based on WS₂ and MoS₂. Reprinted with permission from ref (102). Copyright 2014 Springer Nature. (c) Reconstruction of edge epitaxy in WS₂–MoSe₂, influenced by temperature. Reprinted from ref (106). Copyright 2016 American Chemical Society. PL mapping of MoSe₂–WSe₂ (d), MoSe₂–WS₂ (e), and MoS_2(1–x)Se_2x–WS_2(1–y)Se_2y (f,g). Reprinted with permission from ref (110). Copyright 2018 Springer Nature. To synthesize in-plane TMD heterojunctions with aimed core-ring structure, the control of temperature and vapor sources is significant.

The growth of in-plane TMDs heterojunctions is achievable via two-step CVD method, as well. In 2015, in-plane MoS₂–WSe₂ heterojunctions were successfully synthesized via a two-step method.⁸⁸ First, the WSe₂ triangle monolayer was deposited at a higher temperature, and it could serve as a seed for the lateral epitaxy growth of MoS₂ at a lower temperature in the second step. This method requires the TMD to grow at a higher temperature as a seed to avoid alloying caused by substitution and decomposition caused by high temperature in the second step. The most typical feature of two-step growth is that it avoids the alloying of the junction at the interface since the growth of inner part and outer part is completed in two different stages separately. Moreover, compared with one-step growth in which similar growth conditions of the two sides are required, it is more universal for controllable in-plane epitaxial growth. It seems that the chronology determines the composition distribution of the heterojunction; that is, the first grown TMD constitutes the core part, and the subsequent in-plane epitaxial growth starts from the edges of the core. However, it has been found that this process is also thermodynamically driven.¹⁰⁶ As illustrated in Figure 6c, the two-step route was used to grow a WS₂ seed and a MoS₂ outer ring, but the product was not WS₂–MoS₂. Instead, the composition distribution was strongly determined by the growth temperature. When the growth temperature was 650 °C, the core was MoS₂ and the ring was Mo_1–xW_xS₂. When the growth temperature was varied from 680 to 710 °C, core-ring structures disappeared, followed by the formation of a uniformly distributed Mo_1–xW_xS₂ alloy. This phenomenon can be explained by the interdiffusion of Mo from the edge to the core and W from the core to the edge to reduce its interfacial energy to a minimum, followed by Mo’s replacement for W at the core. At higher temperature, the entropic part has a greater effect on the Gibbs free energy, so the products tend to mix more. In contrast, the enthalpic part has more influence on the Gibbs free energy at a relatively lower temperature, and phase segregation is preferable and thus the MoS₂ core surrounded by an alloy ring was synthesized. However, the interdiffusion and alloying could be avoided by system design. MoS₂–WS₂ and MoSe₂–WSe₂ in-plane heterojunctions were also fabricated with 1D interfaces clearly along the zigzag direction using a homemade quartz reactor to separate the Mo vapor and the W vapor.¹⁰⁷

On this basis, more elaborate in-plane heterojunctions have been synthesized in recent years. The composition-tunable WSe₂–WS_2(1–x)Se_2x and WS₂–WS_2(1–x)Se_2x was obtained via a two-step route,^108,109 and such a precise modulation of composition of the heterojunction can be exploited for band alignment engineering. The sequential edge-epitaxy realizing a multijunction lateral heterostructures was reported.¹¹⁰ Both MoX₂ and WX₂ (X = S or Se) powders were sublimated into oxides and hydroxides at the flow of N₂ and H₂O(g), but the most gaseous W source was hydroxides, thus only the deposition of MoX₂ occurred. When the gas flow was Ar + H₂, the reduction reaction consumed the gaseous Mo oxides, there were relatively more W oxides, and the chemical deposition started due to its low reduction rate.By switching the flow gas back and forth, controllable growth of a striped TMD pattern without complex design was achieved, as shown in Figure 6d–g. WS₂ with WSe₂ nanodots were synthesized by partially damaging the monolayer WS₂ to create growth sites of WSe₂ via O₂ plasma.¹¹¹ Moreover, precisely patterning the in-plane TMD heterojunctions is successfully achieved by a lithography-like strategy, which is discussed in detail in the next section.

5. Design and Growth of TMD-Based Arrays

In order to implement large-scale production of TMD-based heterojunctions and put them into practice, highly ordered TMD arrays need to be realized and integrated. TMD arrays are divided into two types: (1) homogeneous TMD arrays¹¹² and (2) heterogeneous TMD arrays.¹¹³ To prepare TMD arrays, there are mainly four approaches: (1) patterned etching of wafer-scale 2D TMDs,^74,114 (2) in situ sulfurization/selenization/tellurization,¹¹⁵ (3) selective sulfurization/selenization,¹¹⁶ and (4) selective area growth.^117,118

5.1. Patterned Etching of Wafer-Scale 2D TMDs

The most straightforward method for preparing TMD arrays is to etch wafer-scale 2D TMDs by plasma as required for array patterning.^12,119 The biggest challenge of this route is the preparation of wafer-scale TMDs. In terms of wafer-scale polycrystalline 2D TMD preparation, as shown in Figure 7a–c, molecular beam epitaxy (MBE) was used to thermally combine evaporated Se sources with electron-beam-evaporated Hf sources at a pressure lower than 10^–8 Torr to prepare 2 in. HfSe₂ for arrayed memristors.¹²⁰ Compared with MBE, MOCVD does not require a high vacuum. As shown in Figure 7d–f, MOCVD was used with Mo(CO)₆ and (C₂H₅)₂S₂ as precursors for the preparation of 6 in. monolayer MoS₂ at 5.0 Torr,¹²¹ and an array of MoS₂ devices was obtained, with an average carrier mobility of 3.4 cm² V^–1 s^–1 by testing the electrical properties of 900 MoS₂ field-effect transistors (FETs). Compared with the MOCVD, LPCVD with a higher reaction temperature for the preparation of TMDs generally obtains high-quality crystals. The 4 in. monolayer MoS₂ films were prepared using precursor multisource distributed vapor deposition, and the monolayers’ average room temperature device mobility was ∼70 cm² V^–1 s^–1.⁶⁰ The wafer-scale TMDs synthesized above are all polycrystalline materials with more grain boundaries. The grain boundaries of polycrystalline MoS₂ were utilized to fabricate multiterminal memtransistors.¹²²

Figure 7.

Engineerable fabrication of wafer-scale TMDs and TMD arrays. (a) Schematic of setups and configuration of MBE system for growing wafer-scale HfSe₂ thin film. Optical images of (b) MBE-grown HfSe₂ on a 2 in. SiO₂/Si substrate and (c) crossbar array as memristors. Reprinted with permission from ref (120). Copyright 2022 Wiley-VCH. (d) Schematic of the MOCVD system with a cyclical supply of gaseous precursors for growing wafer-scale monolayer MoS₂ and WS₂. Optical images of (e) monolayer MoS₂ and WS₂ on 6 in. quart and (f) MoS₂ FET array fabricated by photolithography with Al₂O₃ as the gate dielectric. Reprinted with permission from ref (121). Copyright 2020 Wiley-VCH. Various ways to synthesize wafer-scale TMDs has been explored, and devices fabricated based on these samples are easily arrayed by conventional lithography.

The more grain boundaries there are, the more electron scattering occurs. The small crystal defects and high carrier mobility of single-crystal TMDs make them suitable for a wide range of applications in optoelectronic devices. The controllability of single-crystal 2D TMDs at the wafer-scale plays a pivotal role in moving TMDs toward practical applications. In the preparation of wafer-scale single-crystal 2D TMDs, the phase transition of MoTe₂ was used to introduce the crystalline species 2H-MoTe₂ in the center of a wafer-scale 1T′-MoTe₂ film to transform metallic 1T′-MoTe₂ into wafer-scale single-crystal semiconducting 2H-MoTe₂ under a Te atmosphere.³² The FET arrays, based on 1T′/2H/1T′ MoTe₂ obtained from seed-induced phase transition, have effectively reduced contact resistance at the 1T′/2H interface. However, this method is highly limited to TMDs which can easily achieve phase transition.

Array etching of wafer-scale TMDs, which can be achieved directly by a laser beam,^123,124 should minimize the impacts of traditional photolithography glue residue on device performance. Moreover, the high growth temperature destroys the marked substrate, making the transfer of material inevitable before the fabrication of devices. However, the transfer without folding and defects put forward many requirements, such as the use of small molecule polymer transfer¹²⁵ or gold film-assisted transfer.^126,127

5.2. In Situ Sulfurization/Selenization/Tellurization

In situ sulfurization refers to patterning the metal precursor source, followed by sulfurizing the patterned metal precursor.¹²⁸ As shown in Figure 8a,b, well-aligned vertical Mo/W layers were deposited by E-beam lithography and sputtering, respectively,¹²⁹ and sulfurized into MoS₂/WS₂ vertical heterojunctions. To avoid the phase mixing in the interface, a two-step method was used to in situ sulfide WO₃ and Mo metal precursors sequentially to obtain vertical MoS₂/WS₂ arrays with a photoresponsivity of 2.3 A W^–1.¹²⁸ It is likely that precursors prepatterned by common mask or lithography possess a poorly defined profile or organic residual adhesive. For this reason, precursor arrays were written with (NH₄)₂MoS₄ ink with an atomic force microscopy (AFM) tip on graphene-covered SiO₂/Si substrates¹³⁰ and annealed in a CVD chamber with sulfur flow to obtain arrays of MoS₂ (Figure 8c,d). In situ tellurization/selenization is suitable for synthesizing TMDs that are chemically unstable. The patterned W/Mo was used as the metal source with a Ni_xTe_y alloy as the Te source to synthesize patterned WTe₂ on a 4 in. SiO₂/Si substrate, and the Schottky–Mott limit of the metal–semiconductor was achieved (Figure 8e,f).¹³¹ Moreover, as shown in Figure 8g,h, to fabricate NbSe₂–WSe₂ metal–semiconductor heterojunction arrays, the Nb₂O₅ pattern and WO₃ thin films were selenized for the formation of NbSe₂ arrays and WSe₂ films, respectively.¹³²

Figure 8.

TMD arrays synthesized by in situ sulfurization, selenization, or tellurization. (a) Schematic of patterned W and Mo seeds and their sulfurization to form WS₂/MoS₂ vertical heterojunctions. (b) Optical image of a micron-scale WS₂/MoS₂ vdW heterostructure array. Reprinted from ref (129). Copyright 2016 American Chemical Society. (c) Drawing shows the pattern writing using liquid–metal precursor ink by AFM tips. (d) AFM images of MoS₂ dot arrays with submicron accuracy. Reprinted with permission from ref (130). Copyright 2016 Wiley-VCH. (e) Schematic diagram of the CVD setup to convert W or Mo patterned beforehand by photolithography to WTe₂ or MoTe₂ arrays. (f) Optical images of diverse WTe₂ patterns. Scale bar, 100 μm. Reprinted with permission from ref (131). Copyright 2020 Springer Nature. (g) Schematic of Nb₂Se₃ pattern and WSe₂ film through selenization of Nb₂O₃ pattern and WO₃ film. (h) Side view and optical images of NbSe₂–WSe₂ patterned heterojunctions. Scale bar, 1 mm. Reprinted from ref (132). Copyright 2017 American Chemical Society. In situ sulfurization/selenization/tellurization is suitable for metal precursors that are easily patterned in a regular way.

5.3. Selective Sulfurization/Selenization

Selective sulfurization is essentially the localized substitution of chalcogen atoms of TMDs in the atmosphere of the desired gas.¹³³ Mimicking the lithography process, TMDs are masked with an inert material pattern and converted to other TMDs in a specific gas flow, thus an artificially designed pattern is realized. As present in Figure 9a,b, striped SiO₂ was deposited as a mask on the MoSe₂ nanosheet, and the sulfur source was evaporated to transform exposed MoSe₂ to MoS₂ at 700 °C, constructing the MoSe₂–MoS₂ lateral array with a type I alignment.¹³⁴ Such a method involves a covered layer, inevitably introducing impurities that affect the performance of the device, so a high energy laser is used to achieve selective conversion of the TMD. Laser-assisted selective conversion of MoSe₂ to MoS₂ in H₂S/H₂ mixed gas was completed,¹³⁵ suggesting that the laser is able to provide a new platform to extend the controllability of TMD array growth. Figure 9c–e demonstrates the preparation of MoS₂/MoS_2–xO_δ heterojunctions by creating multiple S vacancies on MoS₂ and slightly oxidizing it to MoS_2–xO_δ by high-energy laser treatment.¹¹⁶ The MoS₂/MoS_2–xO_δ heterojunction was used to fabricate synaptic devices, which exhibited both short- and long-term plasticity in response to electricity and light-based stimulation simultaneously. On this basis, as presented in Figure 9f,g, the pattern on MoS₂ was directly written and oxidized to MoO_X, followed by the selenization to MoSe₂ to form the MoSe₂/MoS₂ in-plane heterostructure,¹³⁶ implying that a nanoscale precise patterning is available.

Figure 9.

TMD patterns by selective sulfurization. (a) Schematic diagram showing the formation of striped MoSe₂/MoS₂ in-plane heterostructures. The SiO₂ mask patterned onto MoSe₂ by E-beam lithography protects partial MoSe₂ from being converted to MoS₂ by a laser-vaporized sulfur plume. (b) Optical images and Raman mapping of two predefined MoS₂–MoSe₂ patterns. Scale bar, 5 μm. Reprinted with permission from ref (134). Copyright 2015 Springer Nature, open access. (c) Schematic view of boundary-editable technology via laser treatment. (d) Optical images of MoS₂/MoS_2–xO_δ later heterojunctions with various types of boundaries. Scale bar, 10 μm. (e) Elemental mapping at the boundary between MoS₂ and MoS_2–xO_δ. Reprinted with permission from ref (116). Copyright 2021 Wiley-VCH. (f) Schematic diagram of irradiation-induced selective oxidation of MoS₂ followed by selenization and (g) optical image of junction device based on as-prepared in plane MoS₂–MoSe₂. Reprinted from ref (136). Copyright 2021 American Chemical Society. Selective sulfurization finely edits TMDs, enabling the preparation of elaborate in-plane heterojunctions at a small size.

5.4. Selective Area Array Growth

Selective area growth is mainly achieved by surface treatment of the growth substrate¹³⁷ to create differentiated local growth environment for TMDs^138−140 to selectively grow TMD arrays directly. As demonstrated in Figure 10a,b, SiO₂/Si substrates were treated with oxygen plasma to enhance the surface energy, followed by the preparation of MoS₂ using the CVD method, where MoS₂ would selectively grow in the arrayed regions with high surface energy, and polycrystalline MoS₂ arrays were achieved.¹⁴¹ The oxygen plasma was used to etch SiO₂/Si substrates partially covered with graphene to enhance the hydrophilicity of the SiO₂ surface so that MoS₂ was selectively grown on the plasma-treated SiO₂ surface, and MoS₂-graphene lateral heterojunction arrays were obtained consequently.¹⁴² Based on this, graphene could serve as source-drain electrodes of the FET, and the measured carrier mobility was about 17 cm² V^–1 s^–1. This method is universal for the preparation of TMD-based lateral interfacial heterojunctions. Moreover, surfactant-mediated concept of using Na ions as surfactants was proposed to prepare monolayer TMD materials.¹⁴³ The 6 × 6 arrays of NaBr were deposited on a 4 × 1 cm^–2 SiO₂/Si substrate, and subsequently MoS₂ would selectively grow in the NaBr region to realize patterning of MoS₂. As shown in Figure 10c, a rubbing tool was utilized to introduce triboelectric charge of the surface of SiO₂/Si substrate, and MoS₂ would selectively grow on the areas with frictional charges to produce arrayed MoS₂ in CVD growth.¹⁴⁴ The arrays, with an average carrier mobility of 0.18 ± 0.17 cm² V^–1 s^–1 and high switching ratio of ∼8, could be cost-efficient for MoS₂ memristors.

Figure 10.

Selective array growth of TMDs arrays via surface treatments. (a) Schematic of selective growth of MoS₂ on energy-differentiated SiO₂/Si surface after O₂ plasma treatment. (b) Optical images of FET fabricated on the MoS₂ array. Reprinted with permission from ref (141). Copyright 2017 Wiley-VCH. (c) Schematic of the formation of triboelectric charge line pattern caused by rubbing, which serves as the nucleation sites for growing MoS₂ in the CVD process. Reprinted from ref (144). Copyright 2018 American Chemical Society. (d) Schematic of inkjet printing of DI water pattern as designated growth sites of MoS₂. (e) Transfer curve of MoS₂-based FET at V_ds = 3 V with optical images of FET arrays in the inset. Reprinted with permission from ref (145). Copyright 2020 The Royal Society of Chemistry. (f) Schematic flow of a general way to prepare m-TMD/s-TMD arrays, where irradiation creates defective surfaces to form well-ordered nucleation sites of heteroepitaxy in a two-step CVD method. (g,h) Optical images of VSe₂/WSe₂ vertical heterostructure arrays. Scale bar, 10 μm. Reprinted with permission from ref (146). Copyright 2020 Springer Nature. Diverse treatments can be implemented to create differentiated surfaces for region-selective growth of TMDs.

Through preparation of nucleation sites for arrays, Figure 10d,e illustrates the inkjet printing technique¹⁴⁵ to define an array of microscale deionized (DI) water droplets on substrates, leaving arrays of water traces when the water evaporates. Then MoS₂ will selectively grow on the array’s water traces and FETs based on the array was tested with a carrier mobility of approximately 0.15 cm²V^–1s^–1. The ink printing technique greatly improves the efficiency of preparation of array precursor, but the crystal quality of MoS₂ needs to be further improved. In 2020, a general approach was proposed to prepare a series TMD arrays grown on WSe₂ including VSe₂ (Figure 10f–h),¹⁴⁶ NiTe₂, CoTe₂, NbTe₂, and VS₂. The matrix of defect spots was created on s-TMD by laser beam to form nucleation sites for the second TMD. The nucleation sites etched by laser avoid the influence of photolithographic glue residue, and high-quality TMD single crystals were prepared directly by CVD to obtain arrays of TMD heterojunctions.

6. Applications of CVD-Grown TMDs

2D TMDs are particularly attractive in the field of micro/nanoelectronics due to their superior mechanical, transport, and optoelectronic properties. For mechanical properties, the 2D nature overcomes the shortcomings of rigidness and brittleness, making 2D TMDs outstanding candidates for flexible electronics including FETs,¹⁴⁷ supercapacitors,¹⁴⁸ logic gates, and oscillators.¹⁴⁹ In terms of transport performance, theoretical mobility of TMDs ranges from 10 to 1000 cm² V^–1 s^–1, an impressive upper limit.¹⁵⁰ 2D TMDs are especially competitive in optoelectronics among low-dimensional materials for several reasons: (1) 2D TMDs realize photoresponse to a broad spectrum. On one hand, due to the wide variety of TMDs, wavelengths from visible to near-infrared can be realized.¹⁵¹ On the other hand, methods such as modulation by alloying^152,153 and external force^154−158 are highly applicable to 2D TMDs, enabling the artificial band gap with tunability. Moreover, the photodetection range of TMDs can be efficiently expanded by means of heterojunctions. The recently reported interlayer excitons detected in various TMD heterostructures have revealed great potential of TMDs for long-wavelength infrared photodetection.^159,160 (2) A considerable part of the 2D TMDs is materials with strong light–matter coupling.^161,162 When the TMD film is a monolayer, there is a conversion from an indirect band gap to a direct band gap for TMD materials, such as MoS₂, WSe₂, or MoTe₂. (3) 2D TMDs enable an ultrafast photodetection with response times of low to hundreds of nanoseconds.^19,163 In addition, the strong spin–orbit coupling and broken inversion symmetry of TMDs reveal non-negligible potential for applications in spintronics and valleytronics.⁵³ To illustrate the application scenarios of 2D TMDs more concretely, Tables 1–3 summarize TMD materials included in this review and their representative electronics reported.

Table 1. Summary of CVD-Grown Large-Scale TMDs.

materials	architecturesa	methods	characteristics	applications	ref
MoS2	ML	APCVD	wafer-scale thin film	flexible FETs	(164)
	ML	LPCVD	wafer-scale thin film	FETs, logic inverters	(60)
	ML	APCVD	centimeter-scale thin film, single crystallinity	FETs	(47)
	ML	LPCVD	wafer-scale thin film, single crystallinity	FET array	(55)
	multilayer	MOCVD	thin film on glass	low-power photodetectors	(170)
	ML	MOCVD	wafer-scale thin film		(66)
	ML	MOCVD	centimeter-scale thin film	FETs	(68)
MoS2 or MoSe2	ML	APCVD	millimeter-scale single crystals	FETs	(48)
WS2	ML	MOCVD	wafer-scale thin film, single crystallinity	FETs	(69)
	ML	LPCVD	wafer-scale thin film, single crystallinity		(54)
WSe2	ML or BL	APCVD	millimeter-scale single crystals	FETs	(166)
MoSe2	ML	APCVD	centimeter-scale thin film,		(49)
MoTe2	multilayer	APCVD	centimeter-scale single crystals	FETs	(32,99,167)

^aML, monolayer; BL, bilayer.

Table 3. Summary of CVD-Patterned TMDs.

materials	architectures	methods	applications	ref
MoS2	array	selective-area growth	FET array	(145)
	array	selective-area growth	FETs	(141)
	patten	selective sulfurization	FETs, memristors	(144)
	array	patterned etching	FET array	(122)
MoS2 or WS2	array	patterned etching	FET array	(121)
MoS2/WS2	array	in situ sulfurization	transistor, hydrogen evolution reaction catalysts	(128,129)
NbSe2–WSe2	array	in situ selenization	FETs	(132)
MoS2–MoSe2	patten	selective sulfurization/selenization	FETs	(134,136)
WTe2 or MoTe2	array	in situ tellurization	FETs	(131)
MoS2/MoS2–xOδ	patten	selective sulfurization	synaptic devices	(116)
VSe2 (NiTe2, CoTe2, NbTe2,VS2)/WSe2	array	selective sulfurization	FET array	(146)

Table 1 focuses on examples of large-scale highly crystalline semiconducting TMDs, which shows great potential for monolithic integration of chips. For the fabrication of TMD-based electronics, one of the most pressing tasks is to realize the synthesis of boundary-free wafer-scale TMDs with uniform and low thickness. In fact, it is not difficult to meet any of these requirements individually, but the key is to satisfy them simultaneously. It is noted that recent progress has been made in preparing single-crystal wafer-scale monolayer MS₂ (M = Mo, W),^47,54,55,69 while no other TMD with such features has been synthesized successfully. These wafer-scale TMDs with high quality are used as semiconductor channels in FETs and FET arrays on the whole. Moreover, large-scale integrated arrays can be transferred onto flexible polymer substrates to form flexible FET devices.¹⁶⁴ It is believed that, in the next few years, similar work and applications based on other TMD materials will significantly increase, as millimeter-scale monolayer single crystals such as MoSe₂¹⁶⁵ and WSe₂¹⁶⁶ have been successfully obtained. Specifically, wafer-scale multilayer MoTe₂ with centimeter-scale single-crystal domains can be synthesized by phase transition.^32,99,167

Table 2 lists CVD-grown TMD heterostructures, in which it is found that one-step or one-pot methods are preferable for the synthesis of lateral heterojunctions, whereas the two-step route satisfies the demands of preparation of both lateral and vertical heterostructures. Devices based on bare TMDs are extremely limited in both functionality and performance. For instance, there is no effect rectification found in devices made of pure individually CVD-grown monolayer MoSe₂ or WSe₂, but the in-plane MoSe₂–WSe₂ heterostructure exhibits I–V characteristics of a typical p–n junction and can be applied as both self-powered photodetectors and electroluminescent photon emitters.¹⁶⁸ Moreover, a heterostructure composed of two types of TMDs is commonly a type II band alignment, which is highly efficient to separate photoexcited electrons and holes and suitable for photodetectors.¹⁶³ In addition to being used as active channels, TMDs can act as metal electrodes, as well. For example, in the lateral VS₂–MoS₂ heterostructure, the metallic VS₂ serves as the contact between the Ni/Au electrode and MoS₂, decreasing the Schottky barrier to ∼30 meV.¹⁶⁹

Table 2. Summary of CVD-Grown TMD Heterostructures.

materials	architecturesa	methods (step)	characteristics	applications	ref
MoS2/WS2	LHS or VHS	APCVD (1)	clean interface	FETs	(102,171)
	LHS or VHS	LPCVD (2)	clean interface		(93,94)
	LHS	APCVD (1)	photovoltaic effect	self-powered photodetector	(90)
MoSe2/WSe2	LHS	LPCVD (1)	p–n junction photovoltaic effect	photodetectors, light emitting diodes	(168)
	VHS	APCVD (2)	p–n junction, strongly coupled vdW heterostructures	excitonic devices	(88,172)
WSe2/MoS2	LHS	LPCVD (2)	p–n junction photovoltaic effect	photodetectors	(88)
WSe2/WS2	LHS	APCVD (2)	WSe2 puddle in WS2	photogating photodetectors	(111)
WSe2/WS2(1–x)Se2	VHS	APCVD (2)	strongly coupled vdW heterostructures		(96)
WS2/WS2(1–x)Se2	VHS	APCVD (2)			(97)
MS2-MSe2 (M= Mo, W)	LHS	APCVD (2)	clean interface	FETs, complementary metal–oxide semiconductor	(105)
MoX2-WX2 (X= S, Se)	LHS	APCVD (1)	clean interface		(107)
	VHS	APCVD (2)	strongly coupled vdW heterostructures		(95,173)
MoS2–WS2/WS2	LHS and VHS	APCVD (1)	LHS and VHS achieved in one structure		(89)
VS2/WS2	LHS	APCVD (2)	1T-2H metal–semiconductor contact	FETs	(169)
VSe2/MX2 (M = Mo, W; X = S, Se)	VHS	APCVD (2)	1T-1H metal–semiconductor contact		(96)
MoTe2/MoS2	VHS	APCVD (2)	p–n junction	UV–vis–IR photodetectors	(92)
MoTe2/MoTe2	LHS	APCVD (2)	1T′-2H-1T′ metal–semiconductor–metal contact	FET array	(32)
MTe2 (M = V, Nb, Ta)/ WSe2 (WS2)	VHS	APCVD (2)	1T-2H metal–semiconductor contact	FETs	(174)
CrSe2/WSe2	VHS	APCVD (2)	thickness-dependent magnetic order	Hall devices	(175)
NbS2/MoS2	VHS	APCVD (1)	R-type stacking metal–semiconductor contact		(90)
NbS2/WS2	VHS	APCVD (2)	1T-1H metal–semiconductor contact		(97)
ReS2/WS2	VHS	APCVD (1)	fully covered twinned bilayer		(91)
ReS2/MoS2	LHS	APCVD (2)	2H-1T′ type I band alignment		(98)
MoS2/MoTe2(1–x)S2x/MoS2	VHS	APCVD (2)	centimeter-scale heterostructure		(99)

^aVHS, vertical heterostructure; LHS, lateral vertical heterostructure.

Table 3 shows TMD-based arrays and patterns obtained via the four approaches mentioned. Patterned etching relies on proven synthesis technology for large-scale TMDs, thus it is mostly used in the fabrication of well-aligned active channels on wafers.^121,122 To achieve more complex and wide-ranging application scenarios, heterostructures with a wafer size as well as a well-defined boundary are required as discussed above, which is expected to be achieved by in situ sulfurization/selenization/tellurization,^{128,129,131,132} selective sulfurization/selenization,^{116,134,144,146} and selective area growth.^141,145

7. Perspectives for CVD Growth of TMDs

In general, CVD is currently one of most promising approaches to meet the practical demand for production of 2D TMDs, due to its ability to produce large-area thin film and high controllability. The related research concentrates on the synthesis of wafer-scale and boundary-free TMD thin films, TMD heterostructures with designed architectures, and pattered TMD arrays. Although the recent decade has seen a boom in engineering of TMDs and TMD heterostructures, there remains a lot of tough problems to be solved. The primary issue is the scarcity of a general method to mass produce wafer-scale single-crystal TMDs, with limited varieties and substrates failing to satisfy the demands in various applications. To date, the success made in the synthesis of high-quality wafer-scale TMDs mainly comes from common species like MoS₂ and WS₂, and the research on most other TMDs has not progressed. Although highly oriented TMD domains could be achieved by the selection of growth substrates, few of these substrates support a transfer-free route to fabricate electronics, and it may bring transfer-induced imperfections including defects, wrinkles, folds, impurities, and bubbles.

Moreover, further improvement of controllability of the CVD process is urgently required, as well. Although many studies in recent years have emphasized that defects can be effectively improved, defects are a prominent shortcoming of CVD-grown materials which are not competitive with those exfoliated equivalents. Further thermodynamic and kinetic studies related to CVD growth mechanisms need to be developed to assist the refinement of CVD-grown crystal structures. In addition, further improvement of controllability is urgently required, as well. Twist angle, as one of the hottest trending topics in the field of 2D materials due to unconventional physical phenomena originating from it, is extensively studied using artificially stacked TMDs. However, the transfer process inevitably destroys the interlayer interface, and the size of such samples is usually less than 100 μm, resulting in deviations in related measurements. Therefore, a twist-angle-tunable synthesis approach of bilayer or multilayer TMDs is pursued. Moreover, the poor reproducibility of experiments is still a great challenge, and this undoubtedly impedes their further commercialization. Thus, how to create a stable local growth environment under vapor flow by improving equipment setup is a key consideration in future work.

Author Contributions

^§ T.K., T.W.T., and B.P. contributed equally.

The authors declare no competing financial interest.