2D materials advances: from large scale synthesis and controlled heterostructures to improved characterization techniques, defects and applications

The rise of two-dimensional (2D) materials research took place following the isolation of graphene in 2004. These new 2D materials include transition metal dichalcogenides, mono-elemental 2D sheets, and several carbide- and nitride-based materials. The number of publications related to these emerging materials has been drastically increasing over the last five years. Thus, through this comprehensive review, we aim to discuss the most recent groundbreaking discoveries as well as emerging opportunities and remaining challenges. This review starts out by delving into the improved methods of producing these new 2D materials via controlled exfoliation, metal organic chemical vapor deposition, and wet chemical means. We look into recent studies of doping as well as the optical properties of 2D materials and their heterostructures. Recent advances towards applications of these materials in 2D electronics are also reviewed, and include the tunnel MOSFET and ways to reduce the contact resistance for fabricating high-quality devices. Finally, several unique and innovative applications recently explored are discussed as well as perspectives of this exciting and fast moving field.


Introduction
The exfoliation of two-dimensional (2D) graphite, now known as graphene [1], has triggered world-wide research interest concerning atomically thin materials over the last decade. Although graphene was the first to be realized, many other materials exist in layered form, with interlayer interactions governed by van der Waals (vdW) forces. With appropriate methods, these materials can be thinned to a few layers and even monolayers, and are extensively studied today [2]. One group of such materials is the transition metal dichalcogenides (TMDs). Unlike graphene, which consists of an individual atom thick layer, TMDs follow a MX 2 structure, in which M is a transition metal (Mo, W, Re, Nb, etc) and X is a chalcogen atom (S, Se, or Te). Depending on the combination of M and X, the resulting material can vary from semiconducting to metallic and even superconducting [3]. Additionally, as the number of layers is reduced from the bulk crystal, the band structures of many TMDs distinctively change, resulting in a unique sensitivity of the TMD properties to thickness [4,5]. Moreover, the semiconducting TMDs (STMDs) exhibit complementary properties to insulating and semi-metallic layered materials such as hBN and graphene [6]. Other layered materials besides graphene, TMDs, and hBN include mono-elemental 2D semiconductors (phosphorene, silicene, germanene, borophene), mono-chalcogenides (i.e. GaSe, SnSe), and MXenes (M=Ti, Nb, V, Ta, etc; X=C and N). Heterostructures of two or more of these atomically thin materials with a large library of bandgaps and electron affinities can be created to modulate and achieve a wide range of properties suitable for electronic and photonic applications [7][8][9]. Additional and unique degrees of freedom come from the ability to create stacked vdW heterostructures of different compositions, layer ordering and thickness, and interlayer orientation. However, these possibilities are in their infancy and the field of controlled heterostructures is just arising. The excellent properties shown by 2D TMDs and their heterostructures, such as efficient light harvesting [10], sensitive photo-detection [11], and lowthreshold lasing [12] make them promising materials for the next generation electronics and optoelectronics [2,[10][11][12][13][14][15]. The versatile properties of 2D TMDs can be tuned by band structure engineering to achieve, for example, band gap values matching the solar spectrum [16][17][18][19], suppress deep level defects [20], or to modulate the free carrier type and density [21][22][23][24] for field effect transistors (FETs) and p−n junctions-all being the essential elements for integrated electronics and optoelectronics. In addition, these materials can withstand large amounts of strain without degradation, and thus are excellent candidates for their use in applications such as flexible, wearable electronics, and flexible smart phones or tablets [25][26][27][28].
Due to the fast pace of the 2D materials field, this review aims to highlight only the most recent, significant breakthroughs in 2D materials beyond graphene, being mindful of its potential significant emissions due to the global nature of this research. Specifically, we present advances in the areas of material synthesis, optical properties, doping, defect engineering, heterostructures, and applications of these materials such as 2D electronics and optoelectronics. We also provide perspectives and future outlooks. Summaries of earlier work are not provided herein, but can be found in other reviews [29][30][31][32][33][34][35][36][37][38][39].
2. Improved methods of material production 2.1. Large area exfoliation of TMD monolayers Mechanical exfoliation (or 'Scotch Tape' exfoliation) previously used for graphene [1] is one of the main techniques used to obtain TMD monolayers. The process relies on the probability of cleaving the bulk TMD crystal so that all layers except one remain adhered to the tape and only a monolayer is transferred onto the desired substrate. The size of monolayers obtained by standard tape-exfoliation is relatively small (∼5 μm), their yield is poor, and the process is not selective towards the number of layers. The recently developed gold (Au)-mediated exfoliation method by the Javey group (figure 1(a)) enables selective exfoliation of ultra-large TMD monolayers (∼500 μm) [40]. The Au-exfoliation process relies on the bonding of gold (thermally evaporated gold film; 100-150 nm thick) to the chalcogen atoms [41], which is stronger than that of the vdW bonds between the layers, thus allowing for the selective exfoliation of the topmost layer of a TMD source. Au-exfoliation achieves monolayers several orders larger in area when compared to tape-exfoliation ( figure 1(b)), which is a major facilitator for research of these materials. This large size allows for the use of standard characterization techniques like x-ray diffraction (XRD), x-ray photoelectron spectroscopy (XPS), and angle-resolved photoemission spectroscopy (ARPES). The ultra-large Au-exfoliated monolayers possess similar optoelectronic quality as tape-exfoliated flakes as demonstrated by the photoluminescence (PL) maps (figure 1(c)).

Synthesis of large area TMD films via metalorganic chemical vapor deposition (MOCVD)
While chemical vapor deposition (CVD) provides a relatively simple route for synthesizing 2D materials, these methods do not always provide precise control over the grown materials and can often result in inhomogeneous films. Recently, however, homogeneous, wafer-scale films of MoS 2 and WS 2 have been demonstrated by Park's group through the use of MOCVD (figure 2(a)) [42]. Layer by layer growth of four-inch films of MoS 2 and WS 2 were achieved by maintaining low partial pressures (∼10 −4 Torr) of the metalorganic precursors Mo(CO) 6 , and W(CO) 6 . The resulting films also displayed electrical properties comparable to those of mechanically exfoliated layers, with MoS 2 and WS 2 mobilities as high as 30 cm 2 V −1 s −1 and 18 cm 2 V −1 s −1 at room temperature, respectively (see figure 2(b)). Therefore, these studies have demonstrated the scalability when synthesizing high-quality, large-area 2D TMDs, thus opening the door to their industrial fabrication (wafersize scale) and device assembly.

Colloidal synthesis of 2D metal chalcogenides
Solution (wet chemical) methods offer synthesis procedures that occur at much lower temperatures when compared to CVD or MOCVD. They are being developed for the synthesis of colloidal quantum dots, noble metal nanoparticles, and other classes of nanostructured materials [43]. Experimental parameters such as temperature, chemical reagents, and concentrations can often be modified to tune the shape, size, and uniformity of the solution-grown nanostructures. However, methods to produce colloidal TMDs in high yield with rigorous control over shape, size, and uniformity, as well as layer thickness that defines their characteristic size-dependent properties, do not yet exist. Ideal solution routes to TMD nanostructure formation should be scalable, high-yielding, and phase-controllable, thus permitting the formation of TMDs and TMD superlattices with controlled morphologies and thicknesses. Towards that goal, recent progress has been made in understanding how 2D TMD systems nucleate and grow in solution, identifying empirical guidelines for controlling their morphology and crystal structure, and setting the stage for the formation of colloidal nanosheet superstructures.
2D TMDs that nucleate and grow directly from soluble reagents in solution offer material characteristics that are complementary to those of their substrate-grown analogs discussed above [44,45]. Over the past few years, MoSe 2 and MoS 2 have been among the most studied colloidal TMD systems. Solution-grown TMDs have a tendency to form agglomerated, and/or flower-like nanostructures that minimize the overall surface energy [46]. For the colloidal synthesis of MoSe 2 nanoflowers, it has been shown that as the reaction progresses, small MoSe 2 nanoflowershaped particles first form, and continue growing by an outward radial growth of few-layer MoSe 2 nanosheets (figures 3(a)-(d)) [47]. The diameters of the uniform colloidal MoSe 2 nanoflowers could be tuned from 50-250 nm. Raman spectra of the MoSe 2 nanoflowers exhibited a reversible and dynamic variation of the out-of-plane vibration mode position (A 1g ) when employing different laser powers, which was attributed to laser-induced interlayer decoupling by thermal modulation.
The key to synthesizing nanosheets of monolayer and few-layer TMDs in colloidal solutions is to identify conditions that facilitate anisotropic growth while preventing agglomeration. Working towards this goal, the formations of ZrS 2 , TiS 2 , and HfS 2 have recently been realized, where it was found that continuous influx of H 2 S from the slow in situ decomposition of 1-dodecanethiol promoted lateral growth of monolayer nanosheets while a burst influx of H 2 S from the rapid decomposition of CS 2 produced multi-layer nanodisks [48]. For MoSe 2 and WSe 2 , it has been shown that judicious choice of capping ligands is needed to modulate edge-site binding affinities leading to tunability of nanosheet thickness [49].
Previous studies of the formation of SnSe nanosheets, similar to those described above for analyzing the formation of MoSe 2 nanoflowers, revealed that these metal monochalcogenides first grow laterally in solution and then vertically, with thicknesses tunable based on the reagent concentration [50]. Such growth processes can be considered as lateral and vertical homo-growth modes, since the same material is involved in both steps. By exploiting these distinct lateral and vertical growth modes, it may be possible to achieve analogous heterostructures, such as SnSe nanosheets growing directly on the surface of crystallographically related GeSe nanosheets.
In addition to controlling morphology, thickness, and vertical stacking, solution methods can also facilitate the formation of metastable TMD polymorphs that are typically challenging. In this context, colloidal nanostructures of the 2H-and 1T-polymorphs of WS 2 have been obtained in solution by including or excluding hexamethyldisilazane (HMDS), a putative coordinating ligand, respectively [51]. Interestingly, these subtle synthetic differences that produced different polymorphs also led to different morphologies (nanosheets and nanoflowers), which was attributed to differences in electrostatic stabilization of the two types of colloidal particles.
Among TMDs, MoTe 2 has emerged as a particularly interesting system for studying and achieving phase engineering. The stable 2H-MoTe 2 polymorph is semiconducting while the metastable 1T′-MoTe 2 polymorph is semimetallic; the 2H-and 1T′-phases are separated by only a small energy difference [52]. Calculations predict that the 2H-1T′ phase transition of MoTe 2 is more accessible under lattice strain than disulfide and diselenide systems, and adsorption of molecules and atoms on the surfaces of MoTe 2 monolayers can potentially induce the 2H-1T′ structural transition [53][54][55]. While defect creation via annealing or laser irradiation has been employed to access the metastable 1T′ polymorph, the small energy difference between 1T′-and 2H-MoTe 2 makes it challenging to selectively target one polymorph [56,57]. Direct low-temperature solution synthesis of the metastable 1T′-MoTe 2 polymorph was achieved by injecting an oleic acid solution of MoCl 5 dropwise into trioctylphosphine, trioctylphosphine-telluride, oleylamine, and HMDS at 300°C [58]. The 1T′-MoTe 2 nanostructures appeared as uniform nanoflowers (figure 3(e)) comprised of polycrystalline, few-layer nanosheets that exhibited a ∼1% lateral compression relative to bulk 1T′-MoTe 2 . Density functional theory (DFT) calculations suggested that grain boundary pinning in the polycrystalline structure, which contains ∼10 nm 1T′-MoTe 2 domains (figure 3(f)), effectively suppressed its transformation back to the more stable 2H-polymorph and stabilized the 1T′ phase in the MoTe 2 nanostructures.
3. Doping and heterostructures of 2D materials 3.1. Substrate impacts and importance for doping 2D materials The synthesis of any film, regardless of its dimensions, typically requires a substrate. In the case of 2D materials, insulating substrates such as sapphire and SiO 2 are often preferred [59,60]. Conducting substrates such as Au and graphene are used for electrocatalysts [61], stacking of vdW heterostructures [62], and advanced atomic characterization [63]. Semiconducting substrates such as GaN [64], VO 2 [65], and other functional oxide substrates [66] are also becoming important due to their unusual properties that can be realized by combining novel substrates with 2D materials. Therefore, the substrate becomes a new tuning parameter in the CVD synthesis of 2D TMDs. Both two-and three-dimensional (3D) substrates and their relation to doping of 2D materials is discussed in this section.
The concept of 2D substrates has existed for decades [67,68] and has been revitalized in recent years. Most recently, graphene has been utilized as a substrate for the growth of MoS 2 [69], WS 2 [70], WSe 2 [71], and GaSe [72], while sulfide-based chalcogenides can serve as substrates for nitride epitaxy [73]. For graphene-based substrates, MoS 2 and WSe 2 monolayers are well aligned in the symmetry equivalent orientation (figure 4(a)) even though MoS 2 /graphene and WSe 2 /graphene exhibit >20% lattice mismatch [74][75][76]. The low energy electron diffraction (LEED) pattern also confirms the close azimuthal alignment between WSe 2 and graphene (figure 4(a) inset). Conducting AFM characterization of graphene-based vdW heterostructures reveals sharp negative differential resistance due to the presence of pristine interfaces that can be achieved in such structures ( figure 4(b)) [77]. The direct growth of TMDs on suspended graphene also provides a good template for high-resolution transmission electron microscopy (HRTEM) to visualize the atomic structure of TMDs with minimal residues from the TEM sample preparation (figure 4(c)). Additionally, the first successful manganese doping of monolayer MoS 2 via powder vaporization was achieved only when graphene was used as the substrate [78]. In this context, the XPS spectra (figure 4(d)) and scanning TEM (STEM) image (figure 4(e)) clearly indicate the substitutional doping of Mn at the Mo sites. It is believed that the inert surface of graphene provides better doping efficacy. It is noteworthy that 2D substrates, such as graphene, possess unique advantages for chalcogenide-based growth in structural, electronic, and electrical aspects. Therefore, further experiments should be carried out along this direction with the aim of increasing the crystalline domain sizes for device fabrication of different vdW solids.
It has been proposed that a defective 3D surface (sapphire, SiO 2 ) is responsible for the polycrystalline nature of CVD-grown TMDs [42,79]. On the other hand, 3D substrates with lattices similar to TMDs can be good candidates for epitaxial growth of TMDs, regardless of the lattice mismatch. Considering the lattice constant and band gap of various 2D and 3D materials [73,80], one would not expect to use sapphire as a preferred substrate given the >30% lattice mismatch. However, epitaxial growth of MoS 2 has been demonstrated on sapphire with >90% of the MoS 2 domains being aligned to symmetry-equivalent orientations [59]. Similarly, growth of WSe 2 on annealed sapphire also yielded a preferred orientation [81]. To probe the relationship between TMD orientation and sapphire, DFT calculations were conducted. It has been suggested that the adsorption energy at 0°and 60°exhibits a well-defined minima and maxima while it is flat for other orientations [59]. In addition, other theoretical evidence showed that the MoS 2 /sapphire vdW binding energy is the lowest at 0°and 60°due to strain minimization of MoS 2 in these orientations [79]. While sapphire can provide an excellent substrate for single crystal 2D growth, 3D substrates in the nitride family have also been considered. Gallium nitride exhibits a similar lattice constant to MoS 2 (<1%), which lends itself to being an excellent 'active' substrate for epitaxy of 2D layered materials. MoS 2 growth on GaN was recently reported where it was demonstrated that ∼100% of the MoS 2 domains are aligned on GaN substrates, indicating a lattice-matched substrate can be critical to the epitaxial growth of TMDs on 3D substrates (figure 4(f)) [64].

Isoelectronic doping and alloying of 2D monolayers and heterostructures
Due to the reduced dimension of 2D TMDs, isoelectronic doping is one of the most effective ways to engineer their band structure and tailor their properties for desired applications. For isoelectronic doping, the dopant atoms are electronically similar to those of the host. Isoelectronic dopants tend to more easily form alloys, impede the generation and multiplication of dislocations, and reduce the formation of amphoteric native defects [82]. Although isoelectronic dopants provide no extra electrons or holes, it is possible that their different electronegativity can introduce trapping potentials in the lattice to modulate the free carrier density and net conductivity behavior. So far, two types of 2D TMD isoelectronic alloys have been synthesized by substituting either isoelectronic chalcogens, e.g., MoS 2(1−x) Se 2x [16][17][18], or transition metals, e.g., Mo x W 1−x S 2 , Mo 1−x W x Se 2 [83][84][85][86][87]. These isoelectronic alloys show a continuous variation of band gap energy, resulting in PL with tunable emission wavelength in a wide spectral range [16-18, 83, 84]. It is well known that the electronic properties of TMDs are determined by the localization behavior of the d-orbitals of the transition metals [88]. The strong spin-orbit coupling of the d-orbitals can lead to large valence band splitting of several hundreds of meV [89], which can be exploited in the metal alloys (e.g., Mo 1−x W x Se 2 ) to tune the degree of spin or valley polarization. This makes 2D TMD metal alloys especially attractive for band structure engineering and tuning of properties.
Theoretical calculations provide valuable rules for the choice of metal atoms and the design of 2D TMD metal alloys [90], which are specifically described as: where a 1 and a 2 are the lattice constants of the two TMD materials, E g1 and E g2 are their band gaps, and Δd M-X is the difference in the metal−chalcogen bond). Rules (1) and (2) require good matching between the lattice constants and metal−chalcogen bond distances, while rule (3) requires that at least one compound should be a semiconductor because the mixture of two metallic TMDs is not expected to display a finite band gap. The compound pairs selected theoretically based on the above three rules are plotted in figure 5(a), in which the pairs in the blue and red shaded areas are the most likely to be realized due to their smallest lattice mismatch. The pairs in the blue shaded region show the largest difference in band gap energies and are populated with Mo −V and W−V, representing TMD-based alloys between semiconducting MoX 2 /WX 2 and metallic VX 2 (X=S, Se, Te). Alloying between MoX 2 /WX 2 with electron deficient group-V metals like vanadium and niobium generally leads to degenerate p-doping with metallic behavior.
In contrast, the red shaded region in figure 5(a) features isoelectronic Mo−W pair, resulting in semiconductor−semiconductor MX 2 −WX 2 alloys with good lattice matching and moderate band gap differences. In addition to abiding by these selection rules, the formation energies for Mo 1−x W x X 2 alloys are negative due to similar electronic structures and increased attractive Coulomb interaction (caused by different atomic orbital energy and electronegativity of Mo and W) between MoX 2 and WX 2 [90,91]. Thus, Mo 1−x W x X 2 alloys are energetically favorable to be formed. Indeed, 2D Mo x W 1−x X 2 alloys with tunable band gaps have been widely fabricated recently by mechanical exfoliation from bulk alloy crystals or direct growth using CVD [32].
The tuning of optical and electrical properties (such as the carrier type) in isoelectronic 2D TMD alloys was first demonstrated in a CVD-grown monolayer Mo x W 1−x Se 2 system [20,92]. As has been widely reported, monolayer MoSe 2 is natively n-doped [93] and WSe 2 is intrinsically p-doped [94,95], although the characteristics may vary with substrates or contact metals [96]. Since Mo and W have different d-orbital band energies and the band edges of MoSe 2 and WSe 2 are mainly determined by the contributions from these d-orbitals, it is reasonable to expect that without changing the fundamental band structure, 2D TMD alloys and homo-junctions based on isoelectronic substitution in the MoSe 2 −WSe 2 system should provide uniform alloys with tunable band edge position as well as a modulation between p-or n-type doping. In these CVD-grown Mo x W 1−x Se 2 monolayers, the W atoms uniformly incorporate into the monolayer MoSe 2 lattice by substitution of Mo, forming ideal random alloys (figures 5(b) and (c)). The random, isoelectronic substitution of W atoms for Mo atoms in the monolayers of Mo 1−x W x Se 2 effectively suppressed the Se vacancy concentration by 50% and deep levels compared to those in pristine MoSe 2 monolayers [20]. The resultant decrease of defect-mediated non-radiative recombination in the Mo 1−x W x Se 2 monolayers yielded much enhanced PL (figure 5(d)) and extended the carrier lifetime by a factor of three compared to pristine MoSe 2 monolayers grown under similar conditions [20]. In addition to the enhancement of optical properties, the isoelectronic doping of W into MoSe 2 switches the dominant conduction type of the monolayer flakes, suppressing n-type conduction in monolayer MoSe 2 and enhancing p-type conduction with increasing W concentration. The p-type conduction becomes dominant with only ∼18% of W substitution (figure 5(e)). In contrast to the above mentioned metal −semiconductor alloying that results in degenerate doping, the p-type doping in the monolayer Mo x W 1−x Se 2 is non-degenerate. Monolayers of p-type Mo 1−x W x Se 2 and n-type MoSe 2 were vertically stacked to form vdW p−n homo-junctions (figure 5(f)), showing gate-tunable current rectification (figure 5(g)) as well as long-persistent photoconductivity [92,97]. Such tunable n-and p-type conduction realized within an isoelectronic monolayer alloy is highly encouraging as an important strategy to compensate doping with electron or hole donors to synthetically adjust the functionality of 2D TMD systems for many electronic and optoelectronic applications.

Electrolyte gating of 2D materials
Electrolyte gating has been used for nearly a decade to explore transport through 2D materials in the regime of high carrier density. Sheet carrier densities in the range of 10 13 -10 14 cm −2 have been induced by electrolytes in a variety of 2D materials, including but not limited to graphene [98][99][100] [108,109], and SnSe 2 [110]. These large doping densities arise from electrostatic interactions between ions in an electrolyte and induced image charge in a 2D crystal. In an electrolytegated 2D crystal FET, the gate dielectric is replaced by an electrolyte-a material with mobile cations and anions. Two electric double layers (EDLs), or Debye layers, form: one at the electrolyte-semiconductor interface, and the other at the electrolyte-gate interface. This approach is equivalent to moving the gate to within ∼1 nm of the semiconductor interface, providing excellent gate control and ultrahigh capacitance density (2-20 μF cm −2 ) [98-100, 102, 103, 109].
Electrolyte gating has been used to control transport in a variety of semiconducting materials [111,112], but it is especially useful for 2D materials for several reasons. First, substitutional doping of 2D materials is less developed than electrostatic doping, although recent progress has been demonstrated for nitrogen doping of MoS 2 with little effect on the electronic properties of the 2D material [113]. Second, ion doping is a straightforward technique that does not require special equipment for deposition. Third, unlike substitutional doping, which is permanent, ion doping is reconfigurable and therefore the device can be configured as p-or n-type simply by reversing the polarity of the bias. Fourth, ion doping lowers contact resistance at the source and drain by thinning the Schottky barriers to charge injection [103,105,106]. Ionic liquids (IL) and solid polymer electrolytes (SPEs) are the most commonly used electrolytes for gating 2D materials. ILs are liquid salts while SPEs contain a salt dissolved in a polymer [114], typically polyethylene oxide (PEO).
Recently, electrolyte gating has been used as a tool to unlock exciting new physics in 2D materials. For example, an IL, DEME-TFSI, has been used to demonstrate spin polarization [21] and photogalvanic current [104] in WSe 2 FETs. This IL has also been used to create p-i-n junctions in monolayer and multi-layer WSe 2 that gives rise to current-induced circularly polarized electroluminescence (EL) [115]. The ability to reversibly create p-i-n and n-i-p junctions using electrolytes has been demonstrated on MoS 2 [116] and MoTe 2 [109]. In the four terminal MoTe 2 FET shown in figures 6(a)-(d), the inner two contacts are biased with opposite polarities, thereby accumulating anions and cations near the contacts, creating a p-i-n junction. When the temperature of the device is decreased below the glass transition temperature of the electrolyte (T g =243 K for PEO:CsClO 4 ), ion mobility is arrested and the ions are effectively 'locked' into the pi-n configuration. This approach exploits the temperature-dependent ionic conductivity of the electrolyte. The bias can then be removed and the output characteristics show rectifying behavior between contacts 2-3 and 1-4 with an ideality factor of 2.3 (figure 6(f)).
In addition, there have been recent reports of superconductivity in electrolyte-gated TMDs such as MoS 2 [108,117], 2H-MoTe 2 [108], 2H-MoSe 2 [108] and 2H-WS 2 [108]. However, as discussed by Shi and co-workers [108], a crossover from purely electrostatic gating to electrochemical gating is required to observe superconductivity in MoTe 2 , MoSe 2 , and WS 2 . At an electrolyte V G of 12 V-well exceeding the electrochemical window of the electrolyte-charge will be exchanged between the electrolyte and the semiconductor channel and metal contacts. Moreover, ions can be expected to intercalate between the layers of the 2D crystals [110]-a mechanism that is well studied by the battery community. It should be noted that intercalation and electrochemistry will alter the chemical identity of the TMD (i.e., ions are not simply inducing charge in the 2D crystal, they are bonding with the 2D crystal).

Imaging and spectroscopy of vdW heterostructures
In addition to the synthesis of individual and multilayers of 2D materials, various methods have been developed to create their heterostructures. Successful fabrication techniques of vdW heterostructures have been carried out by sequentially assembling the 2D materials. A wide variety of layered materials have been used to create such heterostructures: TMDs, graphene, h-BN, and even topological insulators. While the basic techniques needed for this vdW assembly have been used for several years, recent advances allow for further control of the resulting structures, including the rotational degree of freedom [118]. Another degree of freedom that is beginning to be explored is the interlayer separation, which can significantly affect the interaction between the layers in these heterostructures [119].
Beyond the developments in creating these vdW heterostructures, measurement techniques at the atomic scale are necessary to characterize them, considering their electronic properties sensitively depend on parameters such as the rotation angle between layers. In general, determining this rotation angle is only possible with an atomically resolved measurement technique. Scanning probe microscopy (SPM) is especially well suited for determining angle rotations because it provides not only topographic information about the devices, but also electronic information. For example, consider a heterostructure consisting of two graphene layers. If the two layers are stacked with zero-degree rotation, then standard bilayer Bernal-stacked graphene is recovered. The density of states (DOS) of such a device is shown in figure 7(a) where the charge neutrality point (white dashed line) moves linearly with gate voltage because of the constant DOS. Beyond the DOS, the lack of a moiré pattern and the presence of a triangular lattice in the topography also confirm Bernal stacking. However, if a small twist angle is introduced between the layers, the linear graphene dispersion is preserved at low energy, but new van Hove singularities are created at an energy determined by the twist angle of the device [120]. Figure 7(b) shows such a bilayer graphene device with a ∼2°rotation between the layers. For a 2°rotation, the charge neutrality point moves with the square root of the gate voltage due to the linearly increasing DOS at higher energy. There are also new van Hove singularities created by the interaction between the graphene layers. When this structure is stacked on hBN, the competing interactions between the layers can cause many novel effects to emerge [121].
In these vdW heterostructures, STM can also provide information about quantities beyond the local DOS. Using Fourier transform scanning tunneling spectroscopy (FT-STS), momentum-resolved information about the heterostructure can be obtained by studying the energy dependence of the local DOS and Fourier transforming the result to obtain information in momentum space. This technique gives the allowed scattering wave vectors for the heterostructure. For example, in bilayer graphene, FT-STS can be used to map the changes in the band structure as a function of increasing electric field, which causes the bands to become broader and a band gap to open [122]. For heterostructures with TMDs, FT-STS also provides information about the spin texture of the different bands. Figures 7(c)-(f) shows typical results for WSe 2based heterostructures [123]. In monolayer WSe 2 , the spin-valley coupling forbids transitions from the K to K′ point without a spin flip. Therefore, these resonances are absent in the momentum resolved FT-STS images (figure 7(e)). In bilayer WSe 2 , the spin degeneracy of the bands is restored at the K and K′ point. However, the layer polarization causes spin up electrons at K to reside on one layer and spin down on the other. The situation is reversed at the K′ point and therefore transitions between K and K′ are forbidden due to the spin-valley-layer coupling (figure 7(f)).

Optical properties of 2D materials
4.1. Impact of defects on photoluminescence quantum yield Before 2D materials can be fully realized in devices, defects must be minimized. Defects are particularly important in a monolayer semiconductor where the carrier transport and generation/recombination is confined to a 2D plane. There are various techniques for characterizing defects in a semiconductor. One attractive technique is the PL quantum yield (QY), which is a fast and non-destructive method to quantify the performance of a semiconductor. QY is a quantity describing the ratio of emitted photons over absorbed photons. In a perfect semiconductor without any trap states, the only recombination mechanism at the low carrier concentration regime is radiative recombination, thus the internal QY approaches 100%. As defects and traps states are introduced, the non-radiative recombination rate increases, resulting in the reduction of the QY. It is also important to note the distinction between internal and external quantum yield. Environmental factors, such as the refractive index of the medium and multiple reflections of incident or emitted photons can strongly modulate the external QY by changing the fraction of light which can escape the sample, and it is not related to the material quality [124,125].
Nominally, the internal QY of many 2D materials is quite poor, with typical values for MoS 2 in the range of 0.1%-1% and WS 2 showing the highest reported values (5%-20%) [4,126]. Several recent studies have demonstrated that through the use of chemical surface treatment, defects in 2D material systems can be effectively passivated [127,128]. In one of these studies [127], treatment in the organic superacid bis(trifluoromethane)sulfonimide (TFSI) was used to passivate the surface of MoS 2 (figure 8). Samples treated by TFSI show two orders of magnitude enhancement in the luminescence efficiency, with a final value near 100% [127]. This demonstrates that optoelectronically perfect monolayers can be obtained through chemical treatment of the surface. The treatment was shown to result in no change of the spectral shape and can result in uniform enhancement of the monolayer, as shown in figure 8(a). Due to the large exciton binding energies in TMDs, the recombination behavior differs substantially from III to V semiconductors. Specifically, Shockley-Reed-Hall recombination (non-radiative recombination of a free carrier with a defect site) is not observed under low-pumping conditions (corresponding to low concentration of optically generated carriers). The dramatic enhancement of the PL intensity also results in a corresponding increase in the minority carrier lifetime. As-exfoliated 2D materials have typically been characterized by extremely short lifetimes on the order of tens to hundreds of picoseconds, as shown in figure 8(f). After superacid treatment, a radiative lifetime of 10.8 ns was measured at low pump-fluence (figure 8(g)).
This superacid treatment technique has more recently been shown to also passivate WS 2 [129] as well as ultra-large area MoS 2 samples prepared by goldmediated exfoliation [40], resulting in a QY near 100% in both cases. However, the TFSI treatment technique is ineffective for selenide based TMDs, including MoSe 2 and WSe 2 , and other passivation schemes need to be identified [129]. Superacid treatment of synthetic MoS 2 monolayers prepared by CVD has also been demonstrated ( figure 8(b)), with a final QY value of ∼30% for samples grown under optimal conditions [130].
These results present an important practical advance by demonstrating a route towards realization of high quality TMD monolayers processed over large areas through a simple chemical treatment. It is well known that defects can significantly degrade the performance of minority carrier devices. In addition to this however, defects can also contribute to the interface trap density (D it ) in MOSFETs which results in a reduced subthreshold swing as well as limit off-state current in TFETs via trap-assisted tunneling. This is particularly crucial in monolayer TMDs, where there is no distinction between a 'surface' and a 'bulk' defect. Reducing the number of defects, including S vacancies, is thus vital to obtaining high performance TMD devices. Although the results of this work show a dramatic enhancement in the PL QY, the impact on device performance still needs to be evaluated.

Tuning many-body interactions in 2D semiconductors
Properties of 2D STMDs are strongly influenced by the efficient Coulomb interaction between charge carriers [131]. As a result, the optical response of 2D TMDs is dominated by pronounced resonances from Coulomb-bound electron-hole pairs, or excitons [132,133]. In addition, the quasiparticle band gap, corresponding to the continuum of unbound electrons and holes and thus particularly relevant for carrier transport, is offset to higher energies in comparison to the bulk by the increase of the so-called self-energy contribution [134,135], which scales with the strength of the Coulomb interaction. Because of the weak dielectric screening and strong confinement in monolayer TMDs, both the exciton binding energy and the self-energy shift of the quasiparticle band gap are found to be on the order of many 100s of meV [136][137][138][139][140]. In close analogy to the physics of quantum well systems, it is thus possible to modify the properties of 2D materials by deliberate tuning of the manybody interactions [141].
Both excitons and the quasiparticle band gap are highly sensitive to changes in the free carrier concentration ( figure 9(a)). These changes can be introduced by optical or electrical injection as well as by chemical doping or local engineering of the band structure. Optical techniques, in particular, typically allow for a large transient increase in the excited carrier population by using intense laser pulses, combined with the ability to monitor the subsequent response of the system on ultra-short timescales. Large photo-induced renormalization of the band gap in TMD mono-and few-layers has been recently demonstrated using both optical pump-probe [142] and time-and angle-resolved photoemission spectroscopy [143]. The latter technique, in particular, allows for direct imaging of the energy-and momentumresolved electronic band structure combined with very high temporal resolution. A typical result of this experiment [143] is presented in figure 9(b), where differential photoemission spectrum of a graphenesupported monolayer MoS 2 is plotted in color-scale (red: positive, blue: negative) along Γ -K direction in momentum space after the excitation of the material by a 1.3 mJ cm −2 laser pulse with a photon energy of 3.0 eV, corresponding to above band gap excitation conditions. The spectrum shows a photo-induced shift in the electronic band structure, corresponding to the overall decrease of the quasiparticle band gap on the order of 200 meV (calculated zero-density band structure is shown by yellow lines for comparison). The maximum transient band gap renormalization is found to be as large as 400 meV, depending on the excitation conditions, and is followed by a fast decay within a few picoseconds [143].
Similar conclusions are drawn by monitoring the optical response of WS 2 mono-and few-layers on fused silica substrates after strong photoexcitation [142]. The upper panel of figure 9(c) shows predicted optical absorption spectra of a 2D semiconductor for low and high free carrier densities, adopted from calculations in [141] for quantum well systems. Optical response of a WS 2 bilayer after the excitation with 0.8 mJ cm −2 optical pulse and 2.4 eV photon energy is presented in the lower panel of figure 9(c) (from [142], after subtraction of a linear background). In the lowdensity regime, the absorption is dominated by the exciton ground state with weaker signatures from excited states merging into the onset of the quasiparticle continuum at higher energies. In contrast to that, the exciton resonance disappears from the spectrum completely at sufficiently high carrier densities due to the exciton dissociation. In addition, a spectrally broad region with negative differential absorption emerges at lower energies. The latter is a signature of the population inversion, allowing for the extraction of the absolute band gap position from the onset of the feature. Thus, an overall band gap renormalization on the order of 500 meV is obtained from the experiment. The subsequent decay of the photoinduced changes and the initial recovery of the exciton resonance occurs on a time-scale of a few picoseconds, similar to the observations from the time-resolved photoemission. It is attributed to the ultra-fast recombination of the charge carries via radiative decay and Auger recombination.
The reported values for band gap renormalization as large as many 100s of meV are understood as the direct consequence of a strong Coulomb interaction, scaling with the zero-density exciton binding energy E B,0 ( figure 9(c)). Experimentally obtained renormalization energies reported in [142,143] also compare well with quantitative theoretical predictions for TMD monolayers [144]. The possibility to induce transient changes of the band gap is further combined with the ability to modulate the exciton binding energy both optically and electrically [145] from the initial values of 100s of meV down to zero.

Nano-optics of local excitonic phenomena in 2D layered semiconductors
The tightly bound exciton complexes of monolayer STMDs have made a rich collection of 2D excitonic phenomena accessible for nanoscale optoelectronic applications. While the excitonic phenomena in these materials is reminiscent of that in traditional 2D semiconductor quantum wells, reduced dielectric screening of the Coulombic interactions between excited state electrons and holes yields much larger binding energies that are sufficient to stabilize the bound states at room temperature and ambient conditions. With Bohr radii that are on order of a nanometer [146], these excitations are nanoscale entities and are locally influenced on nanometer length scales by strain [147], electrostatic fields [148], molecular adsorbates [149,150], and structural defects such as grain boundaries [60]. In addition to being potential sources of disorder, these effects can be utilized to control and manipulate local optoelectronic functionality. Thus, developing a thorough understanding of nanoscale excitonic properties is key to enabling unprecedented exciton-based devices in these materials.
Conventional optical techniques are inherently limited by the diffraction limit (a few hundred nanometers across the visible spectrum) and are unable to directly resolve excitonic phenomena at critical optoelectronic length scales (i.e., the exciton diffusion length, inter-defect spacing, etc). Nanoscale plasmonic antennas offer the means to bridge this crucial resolution gap by confining optical fields and amplifying light-matter interactions in sub-diffraction nanoscale volumes that are commonly referred to as 'hot spots.' Combining such antennas with SPM techniques is the basis of nano-optical imaging (and spectroscopy), where such hot spots are scanned over a sample and mediate optical excitation and/or collection at spatial resolutions that can be over 10× better than that of the conventional diffraction limit [151]. Emphasis has thus far been placed on unraveling the local effects of grain boundaries [152][153][154], strain [154], nanoscale variations of the interplay between various exciton states [152,[154][155][156], and possible charge transfer between a metallic nano-optical probe and the 2D semiconductor [155]. Strikingly, a few of these studies have also managed to perform nanoscale tip-enhanced Raman scattering (TERS) to correlate local strain and structure to observed optoelectronic variations [154,155].
One nano-optical imaging technique utilizes the adaptation of the so-called 'Campanile' nano-optical probe [157] to image exciton relaxation in monolayer MoS 2 [152,158], as summarized in figure 10. The unique structure of the Campanile probe simultaneously confines optical excitation and collection of the luminescence to a nanoscale gap at the apex of an optical fiber probe, effectively eliminating any far-field background while retaining the spectral information of the collected emission. In addition, the confined optical fields of the Campanile probe are polarized predominantly in the plane of the sample, parallel to the in-plane transition dipole moments of the direct band gap excitons. Using the Campanile probe to image local variations of the intensity and energy of the excitonic PL, it was found that CVD-grown monolayer MoS 2 has a peripheral 300 nm wide edge ( figure 10(b)) that is distinct from the interior and exhibits spectral signatures analogous to those of disordered semiconductor systems such as quantum dot solids and conjugated polymers ( figure 10(c)). Notably, a similar edge region was found in CVD-grown monolayer WSe 2 [154]. The synthetic origins of such edge regions may stem from the precise nature in how the growth process is terminated and their existence provides new insight into possible origins of poor carrier mobility in some samples as well as edge-related photophysics and catalytic activity. Within the interior region, in contrast to the disordered edge, nanoscale variations were observed between the relative populations of excitons and trions (singly charged excitons), which provide initial evidence of spatial fluctuations of the local carrier density in analogy to 'charge puddles' in graphene [159]. It was also found that the quenching extends to distances that average ∼130 nm from the grain boundary, as shown in figures 10(d) and (e). These measured distances are not resolution-limited and thus offer an initial quantification of the spatial extent of excited state quenching by the grain boundary, which is significantly larger than its corresponding structural size [60] and likely enhanced by a combination of factors, such as exciton diffusion [160,161] and local strain. However, the precise origins and nature behind this quenching effect are still active topics of research. In CVD-grown monolayer WSe 2 samples, the spatial extent of exciton quenching was found to be smaller (∼25 nm) but commensurate with initial estimates of the exciton diffusion length derived from carrier mobility [154]. Furthermore, in other CVDgrown MoS 2 monolayers, line defects and grain boundaries were found to exhibit less PL due to physical damage and missing material [153]. Thus far, only exciton quenching has been investigated with nanooptical imaging techniques, yet grain boundaries (and edges) can also locally enhance the PL by surprising amounts via mechanisms that are also not yet fully understood [60].

Generation and electric control of spin-coupled valley current in WSe 2
Generation and manipulation of a spin current is one of the most critical steps in developing semiconductor spintronic applications [162][163][164]. Due to the large separation of the two inequivalent valleys in k-space and the resulting suppression of intervalley scattering, the valley index can be used in analogy to the spin in spintronics, opening a new research direction on spinvalley locking and 'valleytronics' [165][166][167]. Such a valley polarization achieved via valley-selective circular dichroism has been demonstrated theoretically and experimentally in TMDs that lack inversion symmetry (in monolayer cases or under an electric field) [21,165,166,[168][169][170][171][172][173]. However, a spin/valley current caused by such a valley polarization has yet to be observed and its electric-field control has not been studied.
Normally, in a 2D electronic system (2DES) with spin degeneracy lifted, illumination with circularly polarized light can result in a non-uniform distribution of photo-excited carriers in k-space following optical selection rules and energy/momentum conservation, leading to a spin current [174][175][176]. Referred to as the circular photogalvanic effect (CPGE) [177][178][179][180], the fingerprint of such a spin photocurrent is the dependence on the helicity of the light. The absorption of circularly polarized light results in optical spin orientation by transferring the angular momentum of photons to electrons. Thus, the nonequilibrium spin polarization of electrons forms a spin current with the electron motion in the 2DES plane, which is of practical significance for spin current control in TMDs.
Recently, a spin-coupled valley photocurrent was demonstrated in WSe 2 , where the direction and magnitude depended on the degree of circular polarization of the incident light and could be continuously controlled by an external electric field [181]. Single crystal flakes of WSe 2 were fabricated into EDLTs which have the capability to generate a large interfacial electric field to control electronic phases of solids [99,[182][183][184][185][186][187] and modulate the spin texture in 2DESs [21,188,189]. Figure 11(a) is a schematic diagram of a WSe 2 EDLT gated with ionic gel [190]. Owing to the band bending caused by the chemical potential realignment between the gel and the WSe 2 , there is an electron accumulation with low carrier density at the gel/WSe 2 interface ( figure 11(a)). This large local interface electric field applied perpendicularly to the 2D plane can effectively modify interfacial band bending and the degree of inversion asymmetry. CPGE measurements induced by circularly polarized light on WSe 2 were used to detect a non-uniform distribution of photo-excited carriers and the generated spin current in WSe 2 , with the configuration shown in figure 11(a). The energy of the laser used (1.17 eV, 1064 nm) is below the indirect band gap of WSe 2 ; thus, the photocurrent generated originated from the surface accumulation layer and not from bulk electronhole excitations.
The CPGE phenomenon and the spin photocurrent are highly sensitive to subtle details of the electronic band structure in the 2D systems with lifted spin degeneracy, and thus even a small band splitting may result in measurable effects. Therefore, the modification of the degree of inversion asymmetry with an external perpendicular electric field provides a simple way to control the CPGE photocurrent. Figure 11(b) shows the light polarization dependent photocurrent obtained in a WSe 2 EDLT at various external biases (V G =0-1.1 V). The magnitude of the electric current for all biases was related to the radiation helicity and the photogalvanic current j CPGE (red curves) followed j ( ) sin 2 . Importantly, j CPGE dramatically increased with V G from tens of pA to thousands of pA, unambiguously indicating an electric modulation of the CPGE photocurrent.
The photocurrents generated by optical pumping discussed above were realized at room temperature with the CPGE in accumulation layers on bulk WSe 2 surfaces [181]. Considering that all of the phenomena happens at the surface accumulation layer of WSe 2 , which is where the surface inversion symmetry occurs, similar CPGE phenomenon has also been observed in monolayer MoS 2 [191].

Ultrafast structural response in 2D materials
While the optoelectronic properties of mono-to fewlayer TMDs have received extensive interest over the last decade, these optoelectronic techniques can also be used to probe structural properties and resulting functionality of 2D materials, which is far less studied. This functionality intrinsically involves the material's dynamic structural evolution, extending across a range of length-and time-scales from local vibrations (both in-plane and interlayer) as probed in the frequency domain by Raman scattering approaches [192,193] or in the time-domain by pump-probe spectroscopy [194][195][196], to longer length-scale structural deformations. From the earliest days of research into graphene, speculation arose regarding the role of the Mermin-Wagner theorem [197,198] in determining its mechanical properties, potentially associated with dynamic rippling and buckling responses [199][200][201]. These mechanical deformations in turn can strongly influence corresponding transport properties [202,203]. From a more general perspective, many important electronic and optoelectronic properties are closely intertwined with structural processes through scattering mechanisms arising from electronphonon interactions. Numerous optical studies have investigated the ultrafast time-scale relaxation processes exhibited by photo-excited hot carriers as they couple to the lattice on picosecond and sub-picosecond time-scales [204][205][206][207].
In contrast, direct studies probing the atomic-scale response, e.g., what the atoms are doing, are just beginning to be developed. Nonlinear optical measurements probing the time-dependence of the second order susceptibility can provide information into changes in the crystallographic symmetry of monolayer TMDs, taking advantage of the breaking of inversion symmetry in the monolayer limit [208]. However, both electronic and structural degrees of freedom contribute to this susceptibility, thus making the extraction of the ultrafast atomic-scale response difficult to obtain [209]. Therefore, time-resolved x-ray or electron scattering has been utilized to directly probe the ultrafast structural response of monolayer and fewlayer 2D materials. These measurements are typically carried out in a pump-probe scheme in which an optical trigger initiates some structural response with the dynamic structure visualized through measurement of a time-dependent scattering pattern. In comparison to x-rays, the electron scattering cross-section is many orders of magnitude larger. Recent studies have therefore focused on electron diffraction techniques when probing monolayer samples [210,211]. The first pump-probe transmission electron diffraction study of the ultrafast structural response of monolayer (2H-phase) MoS 2 single domains has recently been reported [212]. In these studies, 400 nm above gap, femtosecond optical excitation was used to generate photo-excited carriers at high carrier densities (∼10 21 /cc) with the atomic-scale response probed by recording changes in the intensities, position in reciprocal space, and size of different diffraction orders. Figure 12(a) shows typical diffraction patterns recorded, corresponding to an average of approximately 1000 shots, using MeV-scale electron bunches with 15 fC charge and pulse duration ∼300 fs at 120 Hz repetition rate [213]. These measurements showed rapid decreases in the intensities of Bragg peaks ( figure 12(b)) associated with an increase in the mean square displacements of the Mo and S atoms within the monolayer, as reflected in a time-dependent Debye-Waller response [214]. This corresponds to a direct measurement of electron-phonon coupling time-scales, with observed response occurring with a time-constant of ∼1.7 ps. More interestingly, ultrafast changes in the position of the diffraction peaks (as measured by the momentum transfer, Q), and in the width of the diffraction peaks are also observed. In analogy with recent TEM studies of static rippling in graphene as reflected in the width of the diffraction spot [201], this response was interpreted as reflective of a dynamic rippling response in the out-of-plane direction. This effect arises from the tilting of the crystal truncation rods associated with the monolayer as it deforms, thus forming an effective cone in reciprocal space. Where this cone intersects the Ewald sphere then determines the magnitude of the broadening response. Associated with this rippling are corresponding strain fields that are also directly measured as subtle shifts in the position of the diffraction peaks, occurring on few picosecond time-scales. The ultrafast turn-on of these effects are indicative of short lengthscale motions, with an upper limit determined by the product of the in-plane sound velocity and the observed time-scale, corresponding to length-scales <10 nm in size. Also recorded in these measurements are the time-scales for cooling of the sample into the substrate from which the thermal interfacial resistances may be directly extracted.

Ultrafast charge transfer in vdW heterostructures
Interlayer charge transfer in vdW heterostructures is an important process for implementing these materials into electronic and optoelectronic devices. To study charge transfer between the two monolayer TMDs in a heterostructure, a laser pulse with a duration of ∼100 fs (pump) can be tuned to the optical band gap of the first TMD (TMD1). This pulse selectively excites the TMD1 layer, which is assumed to have a smaller band gap than TMD2. Most TMD heterostructures form type-II band alignments, where the conduction band minimum (CBM) and the valence band maximum (VBM) are located in different layers [215]. If the CBM and VBM are in TMD2 and TMD1 layers, respectively, the excited electrons are expected to transfer to TMD2 while the holes remain in TMD1. If the band alignment is reversed, holes, instead of electrons, will transfer to TMD2. The charge transfer was monitored by using a probe pulse that is instead tuned to the optical band gap of TMD2 [216][217][218][219]. Either differential transmission [216,218] or differential reflection [217,219] of the probe was measured. Under typical conditions, these quantities are proportional to the density of carriers in TMD2. Hence, by measuring these quantities as a function of the probe delay, defined as the delay of the arriving time of the probe at the sample with respect to the pump pulse, one can monitor the build-up of the TMD2 carrier density as a result of the charge transfer process.
Time-resolved studies of interlayer charge transfer between WS 2 and MoS 2 have shown that the hole density in a WS 2 layer after excitation into MoS 2 reaches a peak on a time scale limited by the instrumental response time. A careful analysis of the time-resolved data showed that the charge transfer time is shorter than 50 fs [216]. Similarly, fast transfer times of electrons in various TMD heterostructures were reported for MoSe 2 to MoS 2 [217], WSe 2 to MoS 2 [218], and WSe 2 to WS 2 [219]. The charge transfer from a WS 2 monolayer to graphene was also found to be sub-100 fs [220]. These results are consistent despite different materials used to form the heterostructures, different sample temperatures (77 K [216] or room temperature [217][218][219]), and different sample fabrication techniques (mechanical exfoliation followed by transfer [216][217][218] or CVD [219]).
Although this efficient and ultrafast charge transfer appears to be experimentally established, a full understanding on its microscopic mechanism is still under development. There are two proposed mechanisms that could be responsible for the efficient charge transfer [221]. First, resonant electron transfer from CBM of TMD1 to higher energy states in the conduction band of TMD2 would allow the sampling of a large range of parallel momentum vectors. Second, the localization of the electrons in real space due to the large binding energy of the charge-transfer excitons results in a large uncertainty of the parallel momentum vectors of the electron, thus facilitating momentum matching in the transfer process. Theoretical studies have also revealed that quantum coherence at the interface enables efficient charge transfer [222]. The quantum coherence is facilitated by the delocalization of the charge-transfer exciton states, and is sufficient to overcome the Coulomb potential. Calculations performed using MoSe 2 -MoS 2 as an example are in excellent agreement with experimental results [217], and indicate that long coherence leads to fast charge transfer. Additionally, it has been shown that the coupling between the hole states in MoS 2 and WS 2 can be significantly enhanced by the Coulomb potential resulting from initially transferred holes [223].

Tunneling devices from TMD heterostructures
The dangling-bond-free surfaces of 2D TMDs (although practically defects exist) and sharp interfaces enable strain-free vdW heterostructures with reduced lattice matching constraints and sharp band edges. These properties of TMD heterostructures are essential for fabricating novel devices such as tunnel diodes and steep switching of tunneling FETs (TFETs).
Several works have investigated graphene-based vdW heterostructures, but the OFF-state leakage current is typically large due to the lack of an intrinsic band gap in graphene. TMDs exhibit a large band gap and have ignited significant research in the design of 2D-2D TMD heterojunction devices. The demonstration of a 2D-2D MoS 2 -WSe 2 tunable gated tunnel diode (figures 13(a)-(g)) [224] highlights the merit of using TMDs. The cross-sectional TEM image (figures 13(b) and (c)) of the device shows the 2D-2D heterojunction with an atomically sharp interface between MoS 2 and WSe 2 . Using the two gates, the MoS 2 and WSe 2 layers are electrostatically doped n-type and p-type respectively, thus allowing for a tunable diode. Figure 13(f) shows the electrical characteristics of the tunable diode as a function of the applied gate voltages. The same device can operate as a tunnel diode, a Zener diode with negative differential resistance, or a normal forward diode by simply changing the two gate voltages, which effectively tune the band alignment at the hetero-interface. Other recent works have also demonstrated 2D-2D heterostructure-based tunnel diodes with promising performance using CVD-grown MoS 2 -WSe 2 [77] and black phosphorus-SnSe 2 [225]. More recently, a SnSe 2 -WSe 2 heterostructure was used to demonstrate a 2D-2D heterojunction TFET [226]. The I D -V G characteristics ( figure 13(h)) show a minimum subthreshold slope of ∼100 mV/decade ( figure 13(i)). In another TFET device ( figure 13(j)), the I D -V G characteristics show that it can be operated as a TFET or a MOSFET based on the gate biases. The temperature independence of the p-branch is indicative of band-to-band-tunneling mechanism for the TFET, whereas the temperature dependence of the n-branch depicts thermionic emission, characteristic of a MOSFET.
In addition to 2D-2D heterostructures, TMDs offer the ability to make versatile 2D-3D heterostructures. Previous works involving InAs-WSe 2 diodes with an ideality factor of ∼1 demonstrate the ability to achieve high-quality interfaces between materials with distinct crystal structures [227]. Most recently, a sub-60 mV/decade TFET was demonstrated using Ge-MoS 2 heterostructures [228], further showing the uniqueness of vertical heterostructures using TMDs for novel device applications.

Passivation of phosphorene MOSFETs
Since 2D materials are sensitive to the environment, their protection from degradation must be studied before high performance devices can be realized. Significant progress has recently been made regarding the passivation of 2D materials, particularly with respect to the material phosphorene. Since 2014, there has been an explosive interest in phosphorene, a monolayer of black phosphorus, as a unique 2D material for electronics, optoelectronics, and other applications. This is mainly because phosphorene has both an intrinsic and sizable band gap (unlike graphene), as well as a high carrier mobility (unlike most STMDs) [229]. In less than two years, more than 700 papers have been published on phosphorene. Due to phosphorene's air sensitivity, the most significant breakthrough has been the successful passivation against environmental degradation such that the unique performance of phosphorene MOSFETs could be demonstrated [230].
Due to a permanent out-of-plane dipole moment, the phosphorene surface is hydrophilic and easily oxidized, especially in humid environments and under illumination. It was found that within approximately an hour of ambient exposure, the phosphorene surface became measurably rougher by AFM and water droplets were observable by optical microscopy. Thus, although phosphorene is sufficiently stable to allow preliminary MOSFETs to be quickly fabricated and characterized, effective surface passivation is of paramount importance in order to achieve long-term stability. Nevertheless, unpassivated MOSFETs typically exhibit carrier mobilities on the order of 10 2 cm 2 V −1 s −1 , current capacities on the order of 10 2 mA mm −1 , and ON/OFF current ratios on the order of 10 4 . Their subthreshold slope was on the order of 1 V/decade with a back gate and an insulator thickness on the order of 100 nm.
To passivate the phosphorene surface, atomiclayer deposited aluminum oxide (Al 2 O 3 ) [230][231][232][233][234] and exfoliated hBN [235][236][237][238][239] have been used to demonstrate stable phosphorene FETs under ambient conditions up to several months (see table 1). Thermal stability of top-gated phosphorene FETs was also demonstrated from −263°C to 150°C [230,236], thus covering the temperature range of most applications. The wide range of temperature stability implied that the source and drain contacts were ohmic and the charge carriers did not freeze out. The ON current of an Al 2 O 3 -passivated phosphorene FET was stable between −50°C and 150°C with an activation energy on the order of 0.01 eV. The mechanical stability of Al 2 O 3 -passivated phosphorene MOSFETs was demonstrated with 5000 strain cycles [240], thus showing their potential in flexible electronics. In addition to Al 2 O 3 , phosphorene passivation with hafnium oxide (HfO 2 ) has been reported [241,242]. It should be noted that besides the success stories, there have also been reports of passivation failures with Al 2 O 3 and SiO 2 [243,244]. Therefore, well established and reproducible protocols need to be developed in the near future.
The hBN/phosphorene/hBN sandwich, although uncommon in MOSFET fabrication, appeared to be of very high quality when carefully assembled [239]. The stack could then be annealed up to 500°C (black phosphorus would transform to red phosphorus at 550°C) for stable carrier mobility on the order of 10 3 cm 2 V −1 s −1 and ON/OFF ratio on the order of 10 5 [237]. The high mobility, approaching that of bulk black phosphorus and exceeding that of single crystalline silicon, could be attributed to the reduction of interface roughness scattering resulted by atomically flat hBN and the reduction of interface state density by the high-temperature annealing, as shown by the elimination of hysteresis. With proper passivation, topgated phosphorene MOSFETs on a high-resistivity substrate have been shown to operate at microwave frequencies [241,245,246].

Phase engineering for low resistance contacts
Phase engineering is another approach that can be utilized to study 2D materials for integration to device technologies. Phase engineering, or accessing targeted polymorphs (e.g., 1T-versus 2H-MoS 2 ), is important because of the significant impact that differences in coordination geometry and layer-stacking sequence can have on the properties of these materials [247]. For example, 2H-MoS 2 is semiconducting while the 1Tpolymorph is metallic [88,247,248]. Although the perfect 1T-phase is less stable than the 2H-phase, a slightly distorted 1T structure-1T′-phase is a more stable structure. Nanosheets of MoS 2 produced via chemical exfoliation are typically a mixture of 2H-and 1T′-phases [249], and these can coexist seamlessly in the same atomic plane ( figure 14(a)). Considering the drastically different properties of the two phases, phase engineering has become an effective way to enhance the performance of MoS 2 nanosheets in electronics and catalytic applications. For example, a Schottky barrier formed between metal contacts and MoS 2 channels has long been a limiting factor for the transport characteristics of FETs [36,[250][251][252]. Recently, it was found that the 1T′-phase formed by local conversion of a 2H-nanosheet could serve as a contact superior to other metals. Owing to the sharpness of the 2H-1T′ interface, the contact resistance is dramatically reduced ( figure 14(b)). Utilizing the metallic 1T′-phase as contacts also benefits the devices' reproducibility [253]. Contact resistance also plays a key role in MoS 2based catalytic applications. Previously, the basal plane of 2H-MoS 2 has been viewed as largely inert towards hydrogen evolution reactions whereas edges and the 1T′-phase were thought to be more catalytically active [254]. In a recent work by Chhowalla's group, it was demonstrated that the 2H basal plane does contain active sites, possibly at defects such as sulfur vacancies, but does not act as an effective catalyst unless the contact resistance can be sufficiently reduced to facilitate charge injection. When phaseengineered 1T′-2H low-resistance contacts are used, the intrinsic catalytic performance of the 2H-phase can be accessed and it is found to be comparable to that of the 1T′-phase or the domain edges (figures 14(c) and (d)) [255].

Impacts of dielectrics and contacts on 2D electronics
Modulation of properties due to environmental effects must be considered when it comes to nanoelectronic devices with atomically thin channels, where the entire material is simply a surface. Mobility enhancement is observed for ultra-thin channels encapsulated in high-k dielectrics [256,257] while mobility degradation is observed in graphene FETs due to the presence of process residues [258,259]. This section focuses on reviewing two key integration challenges for 2D materials: (1) depositing ultra-thin, uniform high-k dielectric gate insulators and (2) realizing low-resistance contacts.
Atomic layer deposition (ALD) allows a precise thickness control of ultra-thin metal oxide films via self-limiting growth cycles achieved by alternating metal and oxygen precursor exposures. Since the precursor reaction chemistry is well established, the initial nucleation step, which requires reactions with the substrate, is key to achieving uniform and homogeneous films. Unfortunately, the lack of dangling bonds on the basal plane of vdW solids limits these reactions with the surface. In recent years, there has been considerable progress made to achieve scalable and uniform high-k dielectrics on TMDs. Recent studies have demonstrated 3 nm pin-hole free ALD of both Al 2 O 3 and HfO 2 carried out at 200°C on MoS 2 , achieved by pretreating the sample with UV-ozone [260,261]. The MoS 2 surface was exposed to atmospheric pressure of O 2 while being illuminated by a mercury lamp positioned <1 cm from the sample surface. This resulted in the adsorption of oxygen species on the MoS 2 surface via S-O bonding without any evidence of Mo-S bond scission. The oxygen remained stable on the surface up to 200°C, which determines the maximum ALD temperature. While the UV-ozone process shows promise for MoS 2 -based applications, it does not appear to be readily transferable to other TMD materials. Both MoSe 2 and WSe 2 substrates showed evidence of metal oxidation during the UV-ozone pretreatment [261]. An alternative approach to achieving conformal Al 2 O 3 utilizes a titanyl phthalocyanine (TiOPc) seed-layer prior to ALD [262]. In this context, TiOPc was deposited on WSe 2 using organic molecular beam epitaxy, which permitted the conformal deposition of 10 nm Al 2 O 3 at 120°C. Afterwards, FETs were demonstrated with an equivalent oxide thickness of 3.0 nm and a leakage current of 0.046 pA μm −2 . The HRTEM of the resultant structure is shown in figures 15(a) and (b). Similar to the UV-ozone method, the maximum ALD temperature is predicted to be 200°C due to the desorption of TiOPc from WSe 2 at that temperature. This process has the advantage of potentially being transferable to all TMDs; however, the authors noted that the desorption temperature from WSe 2 was 100°C lower than from highly ordered pyrolytic graphite, suggesting that the maximum ALD temperature may be strongly dependent on the TMD substrate.
The realization of low resistance contacts is the second integration challenge for 2D materials. As discussed in section 5.3, a measurable Schottky barrier is observed for all elemental metals in contact with MoS 2 and the Fermi level appears to be pinned below the conduction band of MoS 2 [263]. It has been proposed that, upon metal interactions with the S, gap states are induced in the Mo d-orbital by the weakening of the S-Mo interlayer bonding [264]. Additionally, a metal work function modification could be induced by an interface dipole formed during charge redistribution and S vacancies have been predicted to introduce transition states in the upper half of the gap [265]. A practical consideration in the formation of contacts to TMDs is the details of the interface reactions. It has been known for some time that many metals react with MoS 2 to form metal sulfides, MoS x , and even Mo metal at the interface [266][267][268]. A recent review of the thermodynamic predictions for metal-MoS 2 interfaces shows that a large number of metals may react with MoS 2 [269]. Complicating the matter further, it has also been noted that the processing conditions can play a role in determining the contact resistance. The contact resistance of Au contacts deposited in ultrahigh vacuum(UHV, 10 −9 Torr) has been shown to be lower than that of Au contacts deposited in high vacuum (HV, 10 −6 Torr) [270]. However, Ti contacts deposited in HV have been shown to result in higher contact resistance than Ni contacts deposited in UHV.
Additionally, it has been demonstrated that in the case of Ti depositions, the presence of oxygen during HV deposition can result in the deposition of TiO 2 rather than Ti metal (figure 15(c)) [271]. It can be inferred that this interface yields a lower contact resistance than Ti deposited in UHV. This is supported by the demonstration that Ti deposited in UHV reacts with the MoS 2 , thus resulting in the formation of TiS x , MoS x , and possible Mo metal at the interface. Such an interface will have markedly different charge transport characteristics than those predicted assuming a vdW or weak Ti-S bonding with no Mo-S bond scission [272,273].
Beyond elemental metallic contacts, several novel approaches are being adopted such as using patterned graphene ribbons to provide edge contacts to TMDs grown between the contacts by MOCVD [274] as well as phase transitions from the semiconducting 2H to the metallic 1T-polytypes as discussed in the previous section [253]. It has also been recently demonstrated that encapsulating TMDs in hBN is a successful method to significantly reduce contact resistance [275,276].
6. Further applications 6.1. Scalable processing and applications of 2D nanomaterial heterostructures Advances in scalable nano-manufacturing of 2D materials are required to exploit their potential in realworld technology. Towards this end, researchers have begun exploring methods for improving the uniformity of solution-processed graphene and related 2D materials with an eye towards realizing dispersions and inks that can be deposited into large area thinfilms [277]. In particular, as shown in figure 16, density gradient ultracentrifugation allows for the solutionbased isolation of graphene [278], hBN [279], montmorillonite [280], and TMDs [281] with homogeneous thickness down to the atomically thin limit. Similarly, phosphorene has been isolated in organic solvents [282] or deoxygenated aqueous surfactant solutions [283,284] with the resulting phosphorene nanosheets showing FET mobilities and ON/OFF ratios that are comparable to micromechanically exfoliated flakes. By adding cellulosic polymer stabilizers to these dispersions, the rheological properties can be tuned by orders of magnitude, thereby enabling 2D material inks that are compatible with a range of additive manufacturing methods including inkjet [285], gravure [286], screen [287], 3D printing [288], and photonic annealing [289].
Through sequential controlled deposition, these 2D materials can be stacked into heterostructures that have shown promise in several applications, including photodiodes [290], anti-ambipolar transistors [291], gate-tunable memristors [292], and heterojunction photovoltaics [293]. Unlike bulk semiconductors, the atomically thin nature of 2D materials implies incomplete screening of perpendicularly applied electric fields. Consequently, a gate potential can modulate the carrier concentration in multiple 2D nanomaterials concurrently when they are stacked into vertical heterostructures. In this manner, nearly any two-terminal, passive electronic device can be converted into a three-terminal, gate-tunable electronic device when it is constructed from a 2D nanomaterial heterostructure. For example, figures 16(c)-(e) shows a gatetunable p-n heterojunction diode fabricated from a MoS 2 /pentacene heterostructure [293]. The rectification ratio in the resulting diode characteristic can be tuned by orders of magnitude by varying the gate potential. Furthermore, the transfer characteristics of this device show strong anti-ambipolarity since the p-side and n-side of the heterojunction can be fully depleted at positive and negative gate voltage, respectively. Since MoS 2 /pentacene forms a type II heterojunction, it further shows a strong photovoltaic response, which can also be tuned via the applied gate potential. In another example, figures 16(f)-(h) introduce a gate-tunable memristor based on polycrystalline CVD monolayer MoS 2 [292]. In this case, a lateral electric field leads to bistable switching that has been attributed to defect motion in non-stoichiometric CVD MoS 2 . Since this memristor is fabricated from monolayer MoS 2 , its charge transport characteristics are strongly influenced by an applied gate potential. Specifically, the gate voltage controls the switching characteristics, thereby allowing gate tunability over the memristive response with possible implications for non-volatile memory and/or neuromorphic computing applications.
6.2. Nanosculpting and bio-applications of 2D materials 6.2.1. Utilizing electron beams for nanosculpting 2D materials The significant interest in 2D materials for applications also goes beyond electronic devices. Since the construction of the electron microscope in the first half of the twentieth century, electron beams have constituted unique and powerful tools for probing and shaping materials with atomic resolution [39,294]. In parallel, electron beams have been used well beyond material investigation and towards making atomically controlled devices on a chip ( figure 17). Some examples include the early realization of solid-state nanopores for DNA analysis [295] and more recently, nanoribbons with desired lattice orientations [296][297][298] for one-dimensional conductor and interconnect applications, and nanoribbon-nanopore devices for biomolecular detection ( figure 17(d)) [299][300][301][302].
In conjunction with recent advances in aberration corrected electron microscopy to study structure and defects [32,294], many recent experiments have pushed the limits of device sizes to atomic scale in 2D materials beyond graphene (MoS 2 , WS 2 , MoTe 2 , black phosphorous, etc) to expand their function and precision while addressing fundamental questions about structure and properties at nanometer and atomic scales. In situ TEM experiments have featured the fabrication of nanoribbons from novel 2D materials down to sub-nm widths. Notable examples include the TEM-beam carving of graphene nanoribbons [296,303], graphene transistors (figure 17(a)) [304], phosphorene nanoribbons (figures 17(b) and (c)) [298], carbon chains [305], and simultaneous in situ electrical and structural investigations of such structures.

Applications of nanopores and nanoribbons in 2D materials for single biomolecule analysis
Ex situ TEM experiments using TEM-made structures include the ultrafast, all-electronic detection and analysis of biomolecules by driving them through tiny holes-or nanopores (figures 17(e) and (f))-in thin membranes. These experiments include the efforts towards mapping a human genome under 10 min. As molecules are driven through the nanopores in solution, they block the ion current flow, thus typically resulting in current reductions from which a molecule's physical and chemical properties are inferred. DNA, proteins, microRNA, and other biomolecules can be analyzed in this way. The temporal and spatial resolution, as well as sensitivity in these experiments, have been gradually improved over the last few years thanks to advanced materials, device designs, and new lowering the chip capacitance (and therefore the electronic noise) below 1 pF [306] as well as operating nanopores at bandwidths as high as 10 MHz towards realizing extremely fast detection ( figure 17(g)) [307].
2D nanopores have also been envisioned as promising candidates to sequence DNA molecules by recording the variations in ionic current as the molecules thread through the nanopore. Thinner nanopores lead to larger ionic currents due to decreased nanopore resistance. In addition, when nanopores are thin, only a small section of the molecule resides in the nanopore at a given time, thus making thin pores sensitive to small sections of the translocating molecule, as needed, for example, for DNA sequencing applications to be able to read a small number or individual DNA bases. Single-or few-atoms-thick nanopores in freestanding 2D atomically thin membranes are therefore expected to yield optimal signals that will fingerprint specific nanoparticles as they move through the nanopore.
Graphene, MoS 2 , and boron nitride nanopores have been made using highly focused electron beams in the TEM and used for DNA detection [308][309][310][311][312][313]. Additionally, membrane electroporation [314,315] has been successfully used to make nanopores in graphene [316] and MoS 2 [317] and DNA translocation have been demonstrated, opening possibilities towards inexpensive nanopore fabrication. Very recently, nanopores have been also made in other 2D materials such as phosphorene [298]. Interestingly, phosphorene nanopores were recently found to expand under electron illumination from circular to elliptical shape, preferentially along the zigzag direction.
7. Perspectives and future work 7.1. Colloidal synthesis As can be seen from the examples given in section 2.3, significant progress has been made recently in the solution chemical synthesis of 2D metal chalcogenide nanostructures with desired structures and morphologies, offering an increasingly interesting complement to traditional gas-phase, exfoliation, and substratebound synthetic platforms for accessing single-and few-layer TMD materials. While TMDs naturally accommodate 2D nanosheet growth, controlling this growth and overcoming the tendency toward aggregation and nanoflower formation-e.g., forming discrete nanosheets versus multi-pronged cores that lead to multi-site nanosheet growth-requires a deeper understanding of the reaction chemistry and of the early stages of nucleation and growth. Harnessing this solution chemistry and using it to produce morphology-and size-controlled nanostructures, as well as vertically stacked heterostructures, of a diverse group of TMD systems will enable the rational construction of new 2D materials that can be made directly in high yield and in solution-dispersible form.

Doping
The various doping strategies that have already been employed to engineer the band structure and tune the properties of 2D materials have proven effective and versatile. As the need for multifunctional 2D electronics and optoelectronics keeps increasing, the further development and understanding/controlling doping by electromagnetic fields, substitutional doping and alloying within the 2D crystal lattice, and surface doping by physisorption or chemisorption will only increase as a rapidly emerging area for 2D materials. Moreover, the current doping strategies for 2D materials have already shown both advantages and disadvantages, and therefore combinations of these present potential pathways to compensate the drawbacks to further broaden the functionality of 2D materials. For substitutional lattice doping, many dopant atoms that are theoretically predicted to introduce promising electrical and magnetic properties have not been fully studied experimentally [113]. Novel synthetic pathways to overcome energy barriers must be explored to confront this challenge. Furthermore, understanding how to harness the beneficial opportunity provided by the heterogeneity introduced into pristine 2D lattices by substitutional doping or alloy formation is another research focus for the future. For example, how substitutional doping or alloying reduces or increases the levels and types of intrinsic defects in 2D lattices is still unclear. Can the energetics of alloy formation be understood to induce a switch from random alloy formation to long-range ordered superstructures and to heterojunctions with total phase separation? Such issues that couple theoretical predictions to guide experimental synthesis in realizable 2D systems, along with developments in characterization and assessment of functionality, provide a rich and fertile platform for future studies for many years to come.
Although straightforward and effective, ion doping has some critical pitfalls to avoid. For example, to avoid electrochemistry, where the ions exchange charge with the 2D crystal or the metal contacts, the gate bias must be restricted to the electrochemical window of the electrolyte. Within this window, the interactions between the ions and the induced image charge are purely electrostatic. Another example is water absorption, which will decrease the size of the electrochemical window and decrease the gate coupling [319]. Lastly, ion diffusion is orders of magnitude slower than electrons/holes and therefore slow sweep rates (e.g., 20 mV s −1 must be used to minimize hysteresis when using the ions as an electrolytic gate [109]. In less than a decade, the 2D crystal community has progressed from using ions for reconfigurable, unipolar doping of 2D crystals with ultra-high doping densities, to creating p-i-n junctions, to using these junctions to demonstrate current-induced EL. One of the major questions moving forward is whether electrolyte gating is simply an experimental tool for exploring new regimes of transport in 2D crystals or if it can be implemented in beyond complementary metal-oxide-semiconductor(CMOS) architectures.
Ions are typically avoided in device manufacturing; however, the emergence of ion-based resistive random access memories and the diffusion barriers used to contain Cu in CMOS manufacturing suggest that ion containment can be managed. Very large scale integration will require air-stable, solid-state, ultra-thin ion conductors that can be deposited uniformly and can withstand back-end-of-line processing temperatures. These requirements present an exciting challenge for the 2D crystal community moving forward.

Heterostructures
Regarding heterostructures, the possibilities are practically endless for what can be created by combining different 2D materials. In particular, there are many opportunities for enabling new functionality in graphene-based devices by controlling the coupling to other substrates. For example, using a substrate with strong spin-orbit coupling may be able to introduce a quantum spin-Hall state in graphene without magnetic fields [320]. The challenge will be identifying the proper combinations of materials and developing the measurement techniques to access the unique states that will be created. The substrate is critical to tune the properties of CVD-synthesized 2D TMDs, but many challenges to understanding the substrate-2D interaction and its impact on the device performance still remain unsolved.

Optical properties of STMDs
2D STMDs are likely to play significant roles in nextgeneration and model nanoscale optoelectronic devices. With the ability to directly probe light-matter interactions at critical characteristic length scales, nano-optical imaging and spectroscopy can provide valuable insight into underlying excitonic properties that can be used to both improve synthetic techniques and deterministically control nanoscale optoelectronic phenomena. From imaging and quantifying the effects of dopants, defects, edges, and heterostructure interfaces, to continuing to unravel the vexing local photophysics surrounding grain boundaries, to exploring complex many-body physics at nanoscale dimensions, and to developing chiral nano-optical probes for nanoscale valleytronics, many exciting possible nano-optical studies of these 2D materials exist. Aside from electron energy loss spectroscopy [321], which necessitates electron transparency and irradiation with an electron beam, nano-optics currently provides the only means to directly probe nanoscale excitonic phenomena and could ultimately prove as valuable to understanding excitons in 2D transition metal semiconductors as STM has been for understanding the nanoscale electronic properties of graphene.
With respect to the open questions regarding many-body interactions in 2D TMDs, obtaining the phase-diagram for the photoexcited free carrier ratio to Coulomb-bound excitons remains highly relevant. Of particular importance is the related issue of exciton formation, with recently reported ultra-fast sub-picosecond timescales in TMD monolayers [322]. In addition, the predicted strong renormalization of the bands across the Brillouin zone [144,323], e.g., at the Σ and Q points, is intriguing and promising with respect to the possibility of an excitation-induced direct-to-indirect band gap transition, with consequences for optics and transport. Finally, tuning of the electronic band structure via static screening of the environment [324], using strong light-matter interaction in 2D TMDs for photonics and optoelectronics [325], as well as exploring correlated electronic phenomena, such as the Bose-Einstein condensation [326], in 2D layers and heterostructures provide a variety of venues, among many others, for future research of many-body physics in atomically thin materials.
Ultrafast structural response measurements open up new means of probing dynamic structural responses in monolayer and few-layer samples. Novel future opportunities exist here for applying single-shot studies to extract the mechanistic pathways associated with switching dynamics (e.g., 2H-1T or vice versa [327]) and for investigation of exciton-phonon interactions and associated charge transfer dynamics in monolayer TMDs, multi-layers, and more complex vertical heterostructures under a variety of triggers. Time-resolved x-ray studies complementary to the electron scattering measurements described above may be further capable of providing novel information concerning the structural response, probing both inplane and out-of-plane degrees of freedom and their coupling.

Nanoelectronics
While significant progress has been made over the past several years in the design and fabrication of various TMD and heterostructure devices, challenges still remain and further studies are required for simultaneous demonstration of high ON-currents and low subthreshold swings. Further work on the two key integration challenges for 2D materials still remain: depositing ultra-thin, uniform high-k dielectric gate insulators and realizing low-resistance contacts. Moving forward, it is clear that scalable functionalization processes for TMDs are still required in order to achieve high-quality high-k dielectric layer via ALD. Since the 2D materials will likely be in contact with insulators, metals, and other semiconductors, it is vital to develop an understanding of how the layer interactions modify the properties of the materials when compared to those in an idealized isolated state. In terms of understanding the fundamental physics that would explain charge transport in metal-TMD contacts, the realities of the interface chemistry must be considered when comparing experimental results with theoretical models.
Despite the encouraging progress on understanding interlayer charge transfer, the biggest challenge in developing a sophisticated model is probably the lack of experimental techniques to accurately determine the charge transfer rate. Time resolved measurements reported thus far were performed with a time resolution of longer than 100 fs, which only allowed subtraction of the upper limit of the transfer time and hindered systematic investigations on how the transfer time is influenced by the experimental conditions and sample structures. Using shorter pulses could improve the time resolution; however, their inevitably wider bandwidth could compromise the layer selectivity of the pump and probe processes. In this regard, it might be beneficial to design vertical heterostructures to intentionally slow down the charge transfer process so that it can be time resolved. Such studies, combined with theoretical efforts, could help develop and refine models on interlayer charge transfer.
The performance and stability of phosphorene MOSFETs demonstrated to date imply that they may enable thin-film and flexible electronics to operate in the gigahertz range instead of the present megahertz range [245]. To this end, stability under both normal and accelerated operation conditions needs to be demonstrated [246], not just in storage. Additionally, reduction of the interface state density down to the 10 11 cm −2 range needs to be quantified. The development of more effective doping and contacting techniques are also extremely crucial. In the long run, the biggest challenge that needs to be overcome is material synthesis for depositing phosphorene over a large area with high uniformity and reproducibility. Considering the intrinsic merits of phosphorene and the intense interest in it, rapid advancement on all fronts can be expected to continue. Eventually, phosphorene may compete with other 2D and 3D semiconductor materials in atomically thin MOSFETs for ultra-high-speed and ultra-low-power operations.

Applications and integration of 2D materials
While significant progress has been made in the scalable processing and applications of 2D nanomaterial heterostructures, several challenges remain for this field. In particular, at the atomically thin limit, most (if not all) of the atoms in the nanomaterial are at the surface, and thus the properties of 2D nanomaterials are highly sensitive to surface chemistry, underlying substrate, neighboring materials, and interfaces [328,329]. With reduced degrees of freedom compared to bulk materials, 2D nanomaterials are also more strongly influenced by point and extended defects (e.g., vacancies, substitutional impurities, and grain boundaries) [330]. Consequently, growth and assembly methods that minimize or control defect structures [69,331] are likely to have significant impact on the performance, reproducibility, and reliability of 2D nanomaterial heterostructure applications, particularly for vertical heterostructures where defects introduce leakage current pathways that limit the thinness of the constituent 2D nanomaterials [332]. Another emerging frontier is to directly synthesize novel 2D nanomaterials, especially in cases for which there is no bulk layered material available for exfoliation. A recent example is the growth of 2D boron (e.g., borophene) on Ag (111) substrates in ultrahigh vacuum [333]. Since these synthetic 2D nanomaterials are typically stabilized by the growth substrate, careful consideration will be required when developing transfer schemes and heterostructures with other materials. The broad and growing phase space for 2D nanomaterial heterostructures also presents significant design challenges, which may be addressable with computational materials science approaches including those being developed by the Materials Genome Initiative. By addressing these challenges and developing successful integration strategies, 2D nanomaterial heterostructures will not only yield interesting prototype devices, but will also be well-positioned for incorporation into sophisticated circuits and systems.
There are several notable challenges going forward before 2D solid-state nanopores are to be developed into a commercial platform. While the field of traditional silicon-based nanopores has matured over more than ten years, when it comes to new 2D materials, each one of them may present some materials-specific challenge to be surmounted. Such challenges may include the material transferring process from the growth substrate to suitable nanopore chips, or alternatively, the development of methods for materials growth over apertures for membrane applications [334]. Other examples may include issues related to nanopore wetting properties and the surface chemistry of the new material.