The never-ending interest for the development of new layered two-dimensional (2D) materials is the legacy of the graphene success. Among various 2D materials, transition-metal dichalcogenides (TMDs) with MX₂ formula (M: transition metal; X: chalcogen) have emerged as the ace of the post-graphene era. TMDs can exist in various crystalline phases including semiconducting and metallic. More importantly, the electronic bandgap of the semiconducting phase can be controlled via changing M or X elements (Fig. 1(a)), offering a digital portfolio of 2D materials for optoelectronic applications over a relatively wide optical window that covers visible and near infrared regimes. In addition, this digital portfolio can be easily turned into an analogue one through alloying of 2D TMDs, a unique feature that enables design of 2D materials with customized properties. The combination of these properties renders 2D TMDs a family of unmatched qualities for different applications including micro/nanoelectronics [1–3], optoelectronics [4–6], sensing [7,8], and energy harvesting [9].

 

Fig. 1 Heterostructures of TMDs. (a) The electronic bandgap and the relative alignment of band edges for several group-VI TMDs. Energies are calculated with respect to the vacuum level (set to 0 eV). (b, c) The schematic illustrations of lateral and vertical HSs, respectively. Panel (a) is adopted from [63].

Download Full Size | PDF

The rich diversity of material phases, electronic bandgaps and band alignments, and chemical compositions has been the motivation for employing 2D TMDs as building blocks for the realization of devices with complex functionalities as well as the demonstration of novel physical phenomena. In a simple analogy, different layered TMDs can be considered as atomically thin Legos that can be vertically stacked or laterally stitched to construct engineered structures. Such a Lego-like vision has been the driving force for the synthesis of lateral (Fig. 1(b)) and vertical (Fig. 1(c)) heterostructures (HSs) with unprecedented properties. Early experimental demonstrations started with the vertical stacking of heterogenous 2D materials either using the deterministic transfer of exfoliated films [10] or the direct growth of different TMDs on top of one another [11,12]. However, the realization of lateral HSs with in-plane heterogeneities was slowed down by the development of advanced growth protocols, but it was eventually demonstrated using the edge epitaxy method [12–14]. Since the unique behaviors of HSs predominantly stem from the junction area, obtaining a high-quality junction is mandatory for putting the potentials of TMD HSs into the practice. Therefore, the realization of ideal interfaces has been the driving force for major innovations in the field of TMD HSs. In a vertical HS, an ideal interface generally refers to a defect-free, contamination-free, rotationally aligned, and strongly coupled junction between two TMDs. In a lateral HS, an ideal interface refers to a seamless, defect-free, and atomically sharp and coherent junction. These parameters define a roadmap for further advancements in the rapidly expanding field of TMD HSs.

In recent years, much work has been dedicated to the development of novel methods for the synthesis of advanced HSs for applications in the next generation of optoelectronic devices. The purpose of this article is to review recent developments in the synthesis techniques used for the realization of lateral and vertical HSs, with a focus on group-VI (e.g., Mo, W) TMDs. As we will discuss throughout this paper, most of HS synthesis methods are modified versions of conventional techniques used for the alloying of 2D TMDs. Thus, in section 1, we will start with a brief review of mainstream alloying strategies. Then, in section 2, we will review the state-of-the-art methods used for the synthesis of lateral HSs. Next, in section 3, the synthesis of vertical HSs will be reviewed. Finally, in section 4, we will briefly review the spectroscopic characterization of lateral and vertical junctions before we offer our concluding remarks.

1. Alloying

Alloying has long served as an irreplaceable approach for tuning optoelectronic properties of atomically thin semiconducting TMDs. In a ternary TMD alloy (i.e., MX’_2xX_2(1-x) or MX’_2xX_2(1-x); M, M’: transition metals and X, X’: chalcogens), changing the alloying ratio (i.e., x) enables tuning the bandgap of 2D materials, offering specifications that binary crystals (i.e., MX₂) fail to provide. Because of identical crystal symmetries and relatively small lattice-constant mismatch among various TMDs, thermodynamic calculations prove that 2D alloys of TMDs are stable and can be synthesized at moderate temperatures [15]. Accordingly, several studies experimentally demonstrated synthesis of ternary alloys such as Mo_xW_1-xSe₂ [16], WS_2xSe_2(1-x) [17], MoS_2xSe_2(1-x) [18,19], and Mo_xW_1-xS₂ [20]. In addition to the rudimentary bandgap tuning, harnessing the kinetics and thermodynamics of alloying reactions is at the heart of state-of-the-art methods that are currently being used for the synthesis of lateral and vertical HSs. Therefore, to systematically trace the role of alloying in the formation of HSs, in the following we divided the alloying methods into two major categories that we refer to as direct and indirect (or post-growth) alloying methods. However, the mechanical exfoliation from alloyed bulk crystals (e.g., Ref [16].) is not covered in our survey of alloying techniques.

1.1. Direct alloying

This approach, which is the primary method for the synthesis of 2D TMD alloys, uses simultaneous reaction of multiple precursors for the synthesis of TMD alloys in a single step. In the most commonly practiced strategy (see Fig. 2(a)), three sources (e.g., two metals and one chalcogen) are evaporated at appropriate temperatures to provide precursors for the chemical vapor deposition (CVD) [21,22] or physical vapor deposition (PVD) [23] synthesis of a ternary alloy on a substrate that is held at the upstream of the reaction chamber. In an alternative approach called the vapor-phase transport (VPT, also called chemical vapor transport (CVT)), the solid phases of two “stoichiometric” precursors such as MoS₂ and WS₂ are simultaneously evaporated at relatively high temperatures, and the generated vapor phase of these crystals are transported via a carrier gas (usually Ar) to the colder zone of the reaction chamber, where a mixed composition (i.e., an alloy) condensates on the substrate. It is worth noting that, because of more stable stoichiometric precursors, the synthesis of TMD alloys via VPT method is often performed at relatively high processing temperatures. The direct alloying is the mainstream method for the synthesis of ternary TMD crystals because of unique flexibilities such as: compatibility with various synthesis protocols (CVD, PVD, VPT, etc.), compatibility with a diverse set of precursors, the possibility of producing high-quality alloys with randomly mixed metals (Fig. 2(b)) or randomly mixed chalcogens (Fig. 2(c)), and compatibility with various substrate materials (e.g., SiO₂ and sapphire (Al₂O₃)). In this approach the ratio of precursors, the vapor pressure of individual precursors, and the reaction temperature are key parameters for controlling the optoelectronic and structural properties of alloys. However, the co-existence of multiple precursors often demands the generation of a complex temperature profile for solid-phase precursors or a complex gas-delivery system for gaseous precursors. The role of the direct alloying method in the formation of lateral and vertical HS will be discussed in the following sections.

 

Fig. 2 Alloying of 2D TMDs. (a) The schematic illustration of a representative direct-alloying approach in which the simultaneous reaction of WCl₆, MoO₃, and S precursors yields metal-mixed W_xMo_1-xS₂ alloys. (b) Elemental analysis using EDS spectroscopy confirms the uniform distribution of all elements (i.e., a random alloy). (c) An atomic-resolution STEM image (left) and the elemental mapping (right) of a chalcogen-mixed MoS_2xSe_2(1-x) alloy synthesized via the direct reaction of MoO₃, S, and Se precursors. Similarly, a randomly mixed alloy is obtained. (d) The schematic representation of the post-growth alloying method. A binary TMD such as MoSe₂ is first grown and then annealed under a sulfur ambient to partially replace Se atoms by S atoms, which yields a MoS_2xSe_2(1-x) alloy. A complete replacement will convert MoSe₂ into MoS₂. (e, f) PL and Raman spectra of MoSe₂ films sulfurized at various temperatures. The gradual blueshift of the PL emission energy as well as the appearance of MoS₂ Raman modes confirm the formation of alloys with 0 < x < 1 using the post-synthesis alloying approach. Panels (a, b) are reprinted with permission from [64]. Panel (c) is reprinted with permission from [19]. Copyright 2014 American Chemical Society. Panels (d-f) are reprinted with permission from [18]. Copyright 2018 American Chemical Society.

Download Full Size | PDF

1.2. Indirect (post-growth) alloying

As schematically shown in Fig. 2(d), the indirect alloying method starts with the growth of a binary MX₂ film. In the second step, using a high-temperature annealing process in the presence of a dissimilar chalcogen vapor (i.e., X’), the X chalcogens of the MX₂ crystal are partially replaced by the X’ chalcogens, which yields a ternary MoX’_2xX_2(1-x) alloy. This approach is called post-growth alloying because the allying step occurs after the growth of a binary crystal. Raman and PL spectroscopies (Fig. 2(e, f)) have shown that this technique enables a precise control over the x value from 0 to 1. However, the use of this technique has been only reported for the formation of mixed-chalcogen alloys (i.e., MoX’_2xX_2(1-x)) [18,24–26], while mixed-metal alloys have not been demonstrated yet. In a recent study [18], we demonstrated that the quality of the starting film (grown in the first step) is very critical for the success of the post-growth alloying. Our study showed that the presence of vacancy-type atomic defects in the starting film promotes the alloying process. In a case study on a MoS_2xSe_2(1-x) alloy, we showed that Se-vacancies in the starting MoSe₂ monolayers (i) facilitate the Se-S exchange and (ii) further assist the random intermixing of Se and S atoms by mediating the diffusion of introduced S atoms within the MoSe₂ host lattice.

Contrary to the direct approach, in the post-growth alloying, a well-established lattice (grown in the first step) with fixed lattice parameters needs to be disturbed and converted into an alloy with different lattice parameters. Thus, the post-growth alloying process may yield strained TMD alloys in which the level of strain can be precisely controlled by the alloying ratio, x [25]. Such a control can potentially introduce a novel approach for practical realization of the emerging straintronics in 2D materials [27,28]. However, the presence of extended defects such as nanoscale pinholes and pre-existing cracks in the starting TMD crystal is very detrimental to the post-growth alloying process because it causes the breakdown of the 2D alloy under the alloying-induced strain [25]. These observations confirm that the accurate control over the synthesis of binary TMDs in the first step is as important as the second step in which the actual alloying takes place. The role of the post-growth alloying method in the formation of lateral and vertical HSs will be discussed in the following sections.

2. Lateral (in-plane) HSs

A lateral HS forms when two heterogenous 2D materials are brought together to form a seamless junction within an atomically thin plane (Fig. 1(b)). Such junctions have been realized between various pairs of 2D materials including graphene-TMDs [29,30], graphene-hBN [31], and TMD-TMD [24,32]. In the following sections, we will mostly focus on the synthesis of junctions made of heterogenous TMDs, and we divide the synthesis of lateral TMD-TMD HSs into two main categories: the edge epitaxy (section 2.1) and the lithographic patterning (section 2.2). Also, in section 2.3, we will review recent developments on the synthesis of lateral “hetero-phase” structures formed between different crystalline phases of a “single” TMD film (e.g., metallic-semiconducting junctions).

2.1. Edge epitaxy

Lateral TMD-TMD HSs were first realized using a method referred to as the “edge epitaxy”. This technique relies on the growth of a TMD crystal on the active edge of a dissimilar TMD crystal. Indeed, because of the relatively small lattice mismatch between various TMDs, the unsaturated edge of a TMD film can serve as an active growth front for the lateral epitaxy of a dissimilar TMD film (Fig. 3(a)). This approach has been implemented in several variations including single-step [12,13], two-step [14], and sequential [32] schemes. In the single step approach, two stoichiometric TMD crystals such as MX₂ and M’X₂ are simultaneously placed in a single boat held at high temperatures, and the vapor phase of these two TMDs are carried to the colder zone of the reaction chamber, where they condense side by side on a substrate to form a MX₂-M’X₂ lateral HS. Using this strategy, several groups demonstrated the synthesis of MoS₂-WS₂ [12,33], MoSe₂-WSe₂ [13], and MoS₂-MoSe₂ [13] lateral HSs. As exemplified in Fig. 3(b), such HSs are composed of a central TMD crystal surrounded by a dissimilar TMD crystal on their periphery.

 

Fig. 3 Synthesis of Lateral HSs via the edge epitaxy method. (a) The schematic illustration of the edge epitaxy for the synthesis of a MX₂-MX’₂ lateral HS. (b) The STEM image of MoSe₂-WSe₂ HSs synthesized through a single-step edge epitaxy reaction. These HSs are synthesized via a single VPT step using the evaporation of MoSe₂ and WSe₂ powders, typically at 950 ºC. (c) STEM images obtained at the MoSe₂-WSe₂ junction, which shows formation of an alloyed interface. (d) A representative STEM image that illustrates formation of an atomically sharp interface obtained via a two-step epitaxy of MoS₂ on the edge of a WSe₂ monolayer. (e) The optical image of a multi-junction MoSe₂-WSe₂ Lateral HS synthesized via a sequential edge epitaxy technique. (f, g) PL maps obtained at 1.6 eV (WSe₂) and 1.52 eV (MoSe₂) emission energies, respectively, which clearly confirms formation of a multi-junction structure. All scale bars represent 10 µm. Panels (b, c), (d), and (e-g) are reprinted with permission from [65], [14], and [32], respectively.

Download Full Size | PDF

The one-step edge epitaxy approach encounters two major issues, mostly routed in the co-existence of multiple precursors in the vapor phase. First, it often yields an alloyed junction with a finite width (Fig. 3(c)) instead of an abrupt junction with an atomically sharp interface [12,13]. Second, the one-step nature of the process necessitates a growth condition that works for both TMD crystals that are intended to form the lateral junction. Because of this limitation, lateral HSs created by the one-step strategy are composed of either common-chalcogen TMDs (e.g., MoS₂-WS₂) or common-metal TMDs (e.g., MoS₂-MoSe₂), which have relatively similar growth conditions. These two limitations are addressed in the two-step synthesis scheme in which a TMD crystal is first grown and then transferred to another chamber for the lateral epitaxy of the second TMD crystal on the edge of the first one. Such an approach offers two major advantages: (i) it eliminates the interference of the precursors needed for the growth of the two TMD films, which allows for the synthesis of atomically sharp junctions, and (ii) it enables changing both transition metal and chalcogen elements on either side of the junction, which expands the variety of lateral junctions that can be achieved via the edge epitaxy. Using this strategy, WSe₂-MoS₂ lateral HSs with atomically sharp interfaces and dissimilar chalcogens and transition metals have been demonstrated (Fig. 3(d)) [14]. However, the similarity of this approach to the post-growth alloying recipe (section 1.2) limits the growth sequence. Indeed, to avoid alloying or destabilizing the first-grown TMD film in the second step, the TMD film with a higher growth temperature (e.g., WSe₂, 925 ºC) must be synthesized in the first step and then followed by the edge epitaxy of the second TMD crystal with a lower growth temperature (e.g., MoS₂, 755 ºC).

The edge epitaxy can be performed in a sequential mode for the synthesis of structures composed of multiple junctions between two heterogenous TMDs (Fig. 3(e-g)). The development of such structures requires the cyclic exchange of active precursors to alternatively perform the epitaxy on the fresh edge of the last-grown TMD film. To realize such structures, selective and cyclic evaporation of “solid” TMD targets via cyclic switching of the carrier gas was proposed [32]. In this approach, solid MoX₂ and WX₂ (X: S or Se) crystals were simultaneously placed in a boat held at a moderate temperature at which neither of the crystals are volatile (thus no growth). Then, by flowing N₂ + H₂O(g) carrier gas into the reaction chamber the growth of only MoX₂ was promoted. Subsequently, switching the carrier gas to Ar + H₂(5%) terminates the growth of MoX₂ and selectively promotes the epitaxy of WX₂ on the edge of MoX₂. Therefore, cyclic switching between the two carrier gases enables the sequential epitaxy of multiple-junction HSs. In this approach, the fast depletion of pre-existing precursors is very important, otherwise the trace of precursors from a previous step leads to the formation of an alloyed junction instead of a sharp one [32]. It is worth noting that the two-step edge epitaxy method is also capable of producing sequential junctions via sequential transferring of the sample from one reaction chamber to another. However, this strategy has a very low yield and it encounters edge contamination issues because of exposing the sample to the ambient during the transfer process.

2.2. Lithographic patterning

The newest development in the synthesis of lateral TMD HSs relies on applying the post-growth alloying technique (explained in section 1.2) to the lithographically patterned TMDs. As shown in Fig. 4(a), using a lithography step followed by the deposition of an impermeable mask, designated parts of a MX₂ film are protected from the elemental exchange during the post-growth alloying process. Thus, only exposed (i.e., unprotected) regions of the MX₂ film will undergo composition change, while protected regions remain intact to form a lateral junction. As shown in a case study in Fig. 4(b, c), this approach was demonstrated via annealing a partially protected MoSe₂ monolayer under a sulfur-rich ambient to replace Se atoms by S atoms and form a lateral MoSe₂-MoS₂ HS [24,25]. The use of the lithography in this technique enables synthesis of lateral HSs with arbitrary shapes and lateral dimensions in pre-defined locations (Fig. 4(d)) [24], a set of unique features that offers an unprecedented flexibility for the fabrication of practical devices. In contrast, in edge epitaxy methods, the shape of the lateral junction is strictly defined by crystal orientations, lattice symmetries, and the chirality of exposed edges. Therefore, via the edge epitaxy, lateral HSs with only certain shapes such as triangles or hexagons can be obtained [32]. In addition, unlike the lithographic patterning, the edge epitaxy yields lateral junctions in random and unpredictable locations across the growth substrate, which is a major setback for the implementation of practical devices.

 

Fig. 4 Fabrication of lateral HSs via lithographic patterning. (a) The schematic representation of the lithographic patterning method used for the fabrication of lateral MX₂-MX’₂ HSs. Red, green, and blue spheres represent X, X’, and M atoms, respectively. (b) The optical image of a MoSe₂-MoS₂ HS synthesized via the sulfurization of a patterned MoSe₂ monolayer. Purple strips represent protecting masks made of a 50 nm-thick SiO₂ layer. (c) Raman spectra obtained across the line shown on panel (b). Alternating appearance of MoSe₂ Raman modes (in protected regions) and those of MoS₂ (in exposed regions) confirms the formation of a lateral MoSe₂-MoS₂ junction. (d) Optical images and Raman maps of different HS geometries synthesized via the lithographic patterning approach. (e) STEM imaging of the MoSe₂-MoS₂ interface shows that this approach produces an alloyed junction instead of an atomically sharp one. The inset shows a Z-contrast image intensity obtained from the highlighted box. This analysis confirms that the composition gradually changes from MoS₂ (lighter chalcogen, lower intensity) to MoSe₂ (heavier chalcogen, higher intensity). The scale bar in panel (e) represents 5 nm. Panels (b, c) and (d, e) are reprinted with permission from [25] and [24], respectively.

Download Full Size | PDF

The fabrication flexibilities offered by the lithographic patterning method, however, comes at a price: achieving atomically sharp junction is extremely difficult in this technique. As shown in Fig. 4(e), this method often produces an extended interface with a gradual composition variation from one side of the junction to the other. This effect is primarily attributed to the impaired protection of the starting TMD crystal at the edge of the protecting mask [24]. In fact, nanoscale textures, gaps, and pores at the edge of the protecting mask and non-ideal adhesion to the starting 2D material increases the permeability of gaseous precursors underneath the mask edges, an effect that gradually decreases as moving away from the edge. Therefore, a short transition region (a few nanometers) forms between fully protected and fully exposed regions, which projects itself as an alloyed interface with a mixed composition (e.g., MoS_2xSe_2(1-x) in Fig. 4(e)). In addition to the sharp edges, an ideal protecting mask should be stable for the operation at relatively high alloying temperatures (up to 900 ºC) [5], and it must be chemically inert to avoid undesired reactions with the underneath 2D material. In the few reports on this technique, a 50-70 nm-thick SiO₂ layer has been shown to be effective [24,25]. However, the formation of an ideal protecting mask needs further explorations. Compared to the edge epitaxy technique, the lithographic patterning approach has been less explored and further developments in multiple fronts are yet to be made. For example, the side effects of the lithography step (e.g., electron-beam damages) and polymer contaminations on the material properties of the as-synthesized lateral HSs are still unclear. Also, to electrically access both sides of the junction, the development of a robust technique for the removal of the protecting mask is also inevitable.

2.3. Phase engineering

TMDs exist in various crystal structures (Fig. 5(a)) including trigonal prismatic, octahedral, and distorted octahedral phases, usually referred to as H (semiconducting), T (metallic), and T’ (semimetallic) phases, respectively. Motivated by the phase diversity, a line of research has been focused on (i) controlling the crystalline phase of TMDs [34–36] and (ii) stitching different crystalline phases together to realize lateral “heterophase” structures [37]. Ab initio studies have shown that the density of electrons in the d orbitals of transition metals is the primary factor defining the stable phase of a TMD film. With a few exceptions, filled d orbitals yields H-phase TMDs, and partially filled d orbitals yields T-phase TMDs. Accordingly, all group-VI TMDs (except WTe₂) are primarily stable in the H phase, while the group-V TMDs may obtain T or T’ phases. Thus, efforts for the phase engineering have been mostly focused on manipulating the density of electrons in d orbitals of TMDs, often via chemical treatments with reducing agents such as alkali metal-based solutions like n-butyllithium (LiC₄H₉) [36–38]. Studies have shown that, alkali metals (often Li) are capable of transferring electrons into TMD films (e.g., MoS₂ and WS₂) to change the filling of d orbitals from d² to d³, which destabilizes the H phase and induces a transition to T or T’ phases [36,38]. Kappera et al. applied such a chemical treatment to lithographically patterned MoS₂ films to locally induce a phase transformation in pre-defined locations of MoS₂ films, demonstrating lateral H-T heterophase junctions (Fig. 5(c)).

 

Fig. 5 Synthesis of lateral heterophase structures. (a) The schematic representation of three crystalline phases of a monolayer MX₂ film (red: X and blue: M) from the c-axis (upper row) and the side view (bottom row). The arrows on the H lattice displays the gliding direction of the top X plane, which induces a H-to-T phase transformation. As highlighted by dashed lines in the side-view schematics, the out-of-plane alignment of chalcogens in top and bottom planes of the H phase is absent in the T phase. H, T, and T’ phases demonstrate semiconducting, metallic, and semi-metallic behaviors, respectively. (b) The DFT calculation of the ground-state energy for six monolayers of group-VI TMDs. For all TMDs (except WTe₂), the semiconducting H phase is the most stable phase, followed by the T’ phase as the second most stable phase. The noticeable energy difference between T’ and T phases may drive a T-to-T’ phase transition. (c) The optical image of a lateral H-T heterophase structure realized via treating a lithographically patterned MoS₂ monolayer by the n-butyllithium (LiC₄H₉) solution. (d) A STEM image of an atomically sharp interface made between H and T phases of a MoS₂ monolayer. (e) XPS spectra showing the co-presence of H and T phases in a chemically treated MoS₂ sample. The bottom spectrum is obtained from a sample composed of 100% H phase. Panel (a) is adopted from [36]. Panels (b) and (c-e) are reprinted with permission from [35] and [37], respectively.

Download Full Size | PDF

Using a different approach, Cho et al. showed that exposure to high-power laser irradiations can also drive a H-to-T’ phase transition in MoTe₂ crystals, offering a laser-writing approach for patterning lateral heterophase structures in an otherwise homogenous TMD film [39]. In situ STEM analyzes and DFT calculations hinted the role of Te monovacancies in driving the phase transition in MoTe₂ films under the laser irradiation. Indeed, in the competition between different phases, the metallic phase becomes the most thermodynamically stable phase as the concentration of Te vacancies increases (>3%). Other alternative approaches such as electrostatic gating [34], strain engineering [35], and exposure to an electron beam [40] have also been proposed or experimented for inducing the phase transformation and the synthesis of lateral heterophase structures in TMD films. However, despite rich scientific demonstrations, scalability issues and practical shortcomings have limited the scope of such alternative methods to fundamental studies so far.

Lateral heterophase structures are particularly important as metallic phases (i.e., T and T’) can serve as a seamless contact to the semiconducting H-phase, enabling a very low contact resistance to atomically thin TMDs. Using such a configuration, a record-low contact resistance of 200-300 Ω.µm has been reported using the phase engineering of MoS₂ monolayers [37]. Such a low contact resistance is attributed to the sharp atomic interface between H and T phases (Fig. 5(d)) as well as the alignment of the work function in the T and T’ phases with the edge of the conduction band in the H phase [37]. Similarly, the phase engineering is shown to be effective in lowering the Schottky-barrier height from 200 meV (between Au and H-MoTe₂) to only 10 meV (between T’ and H-MoTe₂), which enables the formation of virtually Ohmic contacts [39]. The contact resistance values demonstrated via lateral heterophase synthesis have well surpass the contact resistance values (~30 KΩ.µm) demonstrated via graphene-TMD edge contacts [30].

Despite the great promise that the phase engineering holds for the realization of ultimate all-2D devices, the stability of the metallic phases (i.e., T and T’) for the operation under the ambient condition is still uncertain. Thermodynamic calculations (Fig. 5(b)) predict that, in the absence of external stabilizing stimuli, T and T’ phases of group-VI TMDs are unstable under the ambient condition (except WTe₂, which obtains a stable T’ phase) [35]. Therefore, the observation of stable metallic phases in experimental demonstrations is believed to be due to structural defects [39] and residual counterions such as Li + (in chemically treated samples) or other electron-donor surface adsorbates [36–38]. These stabilizing factors are very fragile and subject to ageing effects, posing questions about the long-term reliability of devices fabricated using lateral heterophase structures. The purity of the phase is another concern that needs to be further investigated because the phase patterning of TMD crystals usually yield a mixture of metallic and semiconducting phases. Studies performed on MoS₂ demonstrated that after treatment with n-butyllithium solution, between 50 and 70% of a H-MoS₂ film is transformed into the T phase (Fig. 5(e)), and this portion reduces to only 10% after annealing the sample at 300 ºC [37,38]. Furthermore, under annealing, a transition between T’ and T phases has also been reported in several studies [36,38], which adds to the aging effects and raises serious concerns regarding repeatability of performance measures obtained from devices fabricated based on phase-engineered TMDs. These stability issues call for further studies on the long-term operation of phase-engineered TMD devices.

3. Vertical (van der Waals) HSs

Vertical HSs can be obtained through stacking dissimilar 2D TMDs (Fig. 1(c)). The layer-by-layer transfer of 2D films has been the most straightforward strategy for the assembly of vertical HSs. Indeed, the development of deterministic transfer processes has enabled stacking any arbitrary combination of 2D van der Waals materials including hexagonal boron nitride (hBN)/graphene [41–43], graphene/TMD [44,45], hBN/TMD [46,47], and TMD/TMDs [48,49]. More importantly, the transfer process can be performed sequentially to assemble complex vertical HSs with a prescribed number of layers stacked in a desired order such as graphene/hBN/MoS₂/hBN/MoS₂/hBN [50]. However, the use of polymers such as PMMA and PDMS during the transfer process often leaves organic contaminations on the surface of 2D materials, which prevents effective coupling between stacked layers and disrupts the charge transfer at the interface [51]. The claimed self-cleaning of organic contaminations in Ref [52] also seems to be very localized and hard to control. Therefore, polymer-free techniques were developed for the assembly of vertical HSs with clean interfaces [53]. A comprehensive review of these techniques can be found in Refs [54,55].

Despite the demonstration of novel physics and the state-of-the-art device functionalities, the mechanical transfer strategy encounters several shortcomings. First and foremost, it is not scalable and cannot be integrated with the mainstream fabrication protocols, a drawback that limits the application of mechanical transfer methods to proof-of-concept demonstrations. In addition, it is quiet challenging to control the rotational alignment between stacked layers, a key feature that significantly influences interfacial charge-transfer properties [51,56]. To address such shortcomings, the direct synthesis of vertical heterostructures was pursued. Different variations of CVD have been frequently used for the direct growth of vertical TMD heterostructures. In its simplest form (Fig. 6(a)), metal oxides are vaporized at high temperatures to react with a chalcogen vapor that is provided via the mild annealing of a powder source. Using such a straightforward protocol, vertical TMD HSs with perfect rotational alignments can be obtained (Figs. 6(b-d)). In this approach, changing the temperature or the vapor pressure of precursors may drive the reaction towards the formation of lateral HSs via the edge epitaxy (as in section 2.1) or the formation of alloys via the random intermixing of elements (as in section 1.1). Thus, controlling growth parameters in this approach is very critical.

 

Fig. 6 Direct growth of vertical TMD HSs. (a) The schematic representation of a furnace used for the CVD growth of WS₂/MoS₂ heterostructures using solid-phase metal precursors and S powders. (b, c) The schematic illustration and optical image of a WS₂/MoS₂ bilayer. (d) The STEM image of a WS₂/MoS₂ HS. The two triangles indicate the orientation of the MoS₂ (top part of image) and WS₂ (bottom part) layers. Inset: fast Fourier transform of the Z-contrast image showing only one set of diffraction patterns, which confirms the perfect rotational alignment between top and bottom layers. (e, f) Cross-sectional TEM images of MoS₂/WSe₂/graphene and WSe₂/ MoS₂/graphene, respectively, on a SiC substrate. (g) The schematic illustration of MoS₂/WS₂ HSs synthesized via the sulfurization of a Mo/W stack. Panels (a-d), (e, f), and (g) are reprinted with permission from [12], [11], and [58], respectively.

Download Full Size | PDF

Metal-organic precursors can also be used to grow TMD HSs via the metal-organic CVD (MOCVD) technique, which offers a better scalability as well as a higher control over the morphology [11]. However, metal-organic precursors are not currently available for all required elements. Using a combination of vaporization CVD and MOCVD steps, Lin et al. demonstrated the growth of MoS₂/WSe₂ (Fig. 6(e)) and WSe₂/MoS₂ (Fig. 6(f)) HSs on a three-layer epitaxial graphene [11]. In their study, the MoS₂ film was grown on the three-layer graphene via a vaporization CVD step (precursors: MoO₃ and S) and then the WSe₂ layers were grown onto the MoS₂ layer via a MOCVD step (precursors: W(Co)₆ and (CH₄)₂Se). Changing the growth sequence simply flipped the stacking order of the HS. It is very important to note that in the vertical stacking, 2D materials are separated from one another via a van der Waals gap that enables the low-strain and rotationally commensurate epitaxy of layers with large lattice mismatches. Therefore, despite a 20% lattice mismatch, a low-strain MoS₂ film can be successfully grown on the epitaxial graphene [57].

Another approach for the synthesis of vertical TMD HS is the sulfurization of a bilayer stack of transition metals. As shown in Fig. 6(g), this technique relies on the conversion of a metal stack such as W/Mo to a HS stack such as WS₂/MoS₂ using a sulfurization process. This process offers a straightforward approach for the high yield synthesis of bilayer HSs on a wafer-scale level [58]. However, the small grain size of deposited metals usually projects into the growth of TMDs with crystalline domains as small as a few tens of nanometers (i.e., poly crystalline). In addition, the layer number is hard to control via the sulfurization process, and the obtained junctions are usually made between multilayer TMD films.

4. Spectroscopic characterization of HS junctions

In vertical TMD HSs, the rotational alignment between stacked layers plays a major role in the optoelectronic behaviors that a junction offers. In fact, a small misalignment leads to the generation of spatially periodic structures with a long-range period, called Moiré patterns (Fig. 7(a-c)). Such patterns induce a spatially periodic variation in the electronic bandstructure, which modifies the properties of interlayer and intralayer excitons in vertical HSs, as being observed in the splitting of exciton and trion energies in a MoSe₂/MoS₂ HS [59]. The impact of rotational misalignment on the transport properties of a vertical HS has also been reported [51]. As shown in Fig. 7(a-d), STEM-based characterizations provide the most accurate way of measuring rotational misalignment between the top and bottom layers of a van der Waals stack. For instance, as shown in Fig. 7(d), the angular offset between the diffraction patterns of top and bottom layers provides a direct estimation of the rotational misalignment in a vertical HS.

 

Fig. 7 Characterization of vertical HS interfaces. (a) The schematic representation of a MoS₂/WSe₂ HS from a side view. (b) The schematic illustration of the rotational misalignment between stacked layers. (c) The formation of Moiré pattern in the STEM image of the HS confirms the misalignment between layers. (d) The diffraction pattern obtained from the MoS₂/WSe₂ HS. Two sets of angularly rotated hexagonal spots can be identified. The rotation offset between these two sets reflects the rotational misalignment between the two layers. (e) PL and absorption spectroscopies of the MoS₂/WSe₂ HS, which reveals interlayer excitonic emission (the red spectrum). (f) The schematic illustration of the interfacial charge transfer mechanism that leads to the interlayer excitonic emission observed in (e). (g) Reducing the interlayer coupling via the insertion of hBN layers in the van der Waals gap between the MoS₂ and WSe₂ layers. (h) The intensity of the interlayer excitonic emission drops as the number of hBN layers (i.e., N) increases. (i, j) The optical image and PL map of a WS₂-MoS₂ HS, showing that PL is confined at the junction. (k, l) PL and Raman spectra of a MoS₂-WSe₂ junction. The inset in (k) displays the map of PL peak energy. Panels (a-h), (i, j), and (k, l) are reprinted with permission from [60], [12], and [14], respectively.

Download Full Size | PDF

As we saw in Fig. 1(a), the alignment of bands in group-VI TMDs endorses 2D TMD monolayers for the formation of type-II HSs in which electrons and holes reside on separate sides of the junction. However, the observation of this effect requires a strong coupling between stacked TMDs. PL spectroscopy has been the most extensively employed means for confirming the formation of a type-II band alignment in vertical HSs. As shown in a representative example in Fig. 7(e) [60], in a successfully formed vertical junction between MoS₂ and WSe₂ monolayers, PL spectrum displays an emission peak with an energy smaller than those of isolated MoS₂ and WSe₂ monolayers. However, a complementary absorption spectroscopy (dashed lines, Fig. 7(e)) reveals that the absorption peaks of stacked regions are simply a superposition of isolated MoS₂ and WSe₂ absorption peaks. Thus, PL and absorption spectroscopies collectively conclude that “intralayer” excitons are first generated in top and bottom layers and then “interlayer” excitons form following the charger transfer across the junction (as shown in Fig. 7(f)). This conclusion was further confirmed through inserting a varying number (i.e., N) of hBN layers into the van der Waals gap between MoS₂ and WSe₂ monolayers (Fig. 7(g)). Interestingly, upon increasing N, the emission intensity of interlayer excitons dropped and finally disappeared, suggesting that the charger transfer at the interface is blocked by thicker hBN layers (Fig. 7(h)). This study provides a compelling evidence regarding the charge transfer at the interface of a vertical TMD HS. However, the notion of interlayer excitonic emissions and the complete charge transfer across the junction have been challenged in a new study performed on a WSe₂/MoSe₂ vertical HS [61]. In fact, this study suggests that electrons are strongly hybridized (i.e., shared) between both sides of the junction and a complete transition does not occur. In another study performed on a WS₂/MoS₂ stack [12], interlayer and intralayer excitons were simultaneously observed. This unelaborated observation might be due to the slow charge transfer [62] that allows for the recombination of intralayer excitons prior to the formation of interlayer excitons.

In contrast to vertical HSs, the formation of interlayer excitons in the PL spectra of lateral HSs has not been observed yet. This effect might be due to the extremely small junction area in lateral HSs, which makes the direct observation of this effect very challenging. However, the confinement and enhancement of the PL emission at the vicinity of lateral junctions have been unanimously reported (e.g., Fig. 7(i, j)) [12,13]. such confinement has been attributed to possible effects such as composition variation at the junction and the trapping of excitons by interfacial defects or by the built-in electric field originating from the type-II band alignment.

Unlike the van der Waals interaction of TMDs in vertical HSs, in lateral heterostructures, TMDs are tightly connected via a seamless interface. Thus, the lattice mismatch-induced interfacial strain at lateral junctions cannot be relaxed as effective as in vertical junctions. Li et al. studied this effect via PL (Fig. 7(k)) and Raman (Fig. 7(i)) spectroscopies performed on a WSe₂-MoS₂ lateral junction that was grown via a two-step edge epitaxy method [14]. In this study, the shift of the PL peak position and the splitting of the in-plane resonance mode (i.e., E²_1g) of the Raman signal was employed to map the strain distribution at the vicinity of the junction (the inset of Fig. 7(k)). They found that the MoS₂ side of the junction is under ~1.8% strain of tensile type, which is attributed to the larger lattice constant of WSe₂ compared to that of MoS₂. However, to accommodate the lattice expansion close to the junction, some parts of the MoS₂ layer (mostly close to the periphery) are contracted, which generates a sporadic compressive strain profile.

Interestingly, the development of the interfacial strain has not been reported in lateral junctions made via the one-step edge epitaxy. One possible explanation is that the two-step nature of the MoS₂-WSe₂ growth obstructs the lattice relaxation at the vicinity of the junction. Indeed, after the high-temperature growth of WSe₂ at 925 ºC, the sample is cooled down to the room temperature, which stabilizes the lattice of WSe₂ at the end of the first step. Therefore, the lower growth temperature (~750 ºC) in the second step (i.e., the edge epitaxy step) cannot fully relax the WSe₂ lattice for the formation of a strain-free interface. Interestingly, similar strain generation was reported for the two-step allying of TMDs [25], as we explained in section 1.2. In the one-step edge epitaxy, however, both sides of the junction are created without cooling down the chamber or lowering down the temperature, which allows for an effective relaxation of the lattice at the interface. A similar rational can explain the strain-free synthesis of alloys obtained through one-step alloying methods (section 1.1).

5. Concluding remarks

An overview of efforts made in the field of TMD HSs shows a rapid development in multiple fronts including growth and assembly techniques, characterization and spectroscopy methods, and state-of-the art device and physics demonstrations. These efforts are further excelled by the growing demand for new material platforms for the next-generation of mainstream technologies as well as emerging applications such as flexible and wearable optoelectronics. In addition, TMD HSs have several unique features that enable a set of applications complementary to those offered by conventional III-V HSs. First, the van der Waals interaction of 2D HSs with their surrounding materials lifts the lattice-matching constraints that needs to be satisfied in the growth of conventional III-V HSs. This feature enables the formation of junctions between heterogenous 2D materials with largely mismatched lattice constants on virtually any substrate. For instance, despite a ~20% lattice mismatch, the epitaxial growth of MoS₂ layers on graphene yields a low-strain and rotationally commensurate junction with a long-range lattice matching [57]. Thus, 2D HSs can be synthesized on a wide range of substrates including flexible materials for emerging applications in wearable optoelectronics. Second, the intrinsically passivated surface of 2D materials is another advantage of 2D HSs over the conventional III-V HSs in which dangling bonds and surface defects cause issues such as nonradiative recombination in optical devices [66] or reduced carrier mobility in electronic applications [67]. Third, in 2D materials, reduced dielectric screening and the spatial confinement of carriers in an atomically thin plane enhance the oscillator strength and Coulombic interaction between electrons and holes, leading to the formation of excitonic bonds with binding energies as large as ~800 meV [68], which is almost two orders of magnitude larger than that in GaAs/AlGaAs HS quantum wells (~10 meV) [69]. Thus, TMD HSs offer a significantly larger dynamic range (nominally equivalent to exciton binding energies) for the electro-optical modulation of the excitonic emission/absorption with the possibility of operation at room temperature. In addition, the lifetime of radiative recombination in 2D TMDs (a few picoseconds) [70] is shorter than that in conventional GaAs/AlGaAs HSs (several hundreds of picoseconds) [71], which enables the design of potentially faster and more efficient electro-optical modulators and light emitters based on 2D HS platforms.

Despite great promises shown in the state-of-the-art demonstrations, the field of 2D HSs needs further maturity before practical devices can be mass produced. The quality of demonstrated junctions are still far away from fundamental limits, and more importantly the yield and repeatability of results for large-scale device fabrications need further improvements. Similarity of methods used for the alloying of TMDs (section 1), the edge epitaxy of lateral HSs (section 2), and the direct synthesis of vertical HSs (section 3) show that only slight changes in a synthesis scheme may yield fundamentally different structures. Thus, for a reliable and repeatable growth, a deep understanding and a precise control over the kinetics and thermodynamics of the synthesis processes are mandatory. Further advancements in the synthesis of TMD HSs demand cutting-edge instruments capable of creating complex temperature profiles across the reaction chamber, the rapid evacuation/injection of precursors in a cyclic manner, and the dynamic tuning of the pressure during the process. Also, the development of complex precursors with customized vapor pressures, dissociation temperatures, and chemical reactivities seems crucial for pushing heterojunctions of 2D TMDs to their fundamental limits.