Atomic and structural modifications of two-dimensional transition metal dichalcogenides for various advanced applications

Two-dimensional (2D) transition metal dichalcogenides (TMDs) and their heterostructures have attracted significant interest in both academia and industry because of their unusual physical and chemical properties. They offer numerous applications, such as electronic, optoelectronic, and spintronic devices, in addition to energy storage and conversion. Atomic and structural modifications of van der Waals layered materials are required to achieve unique and versatile properties for advanced applications. This review presents a discussion on the atomic-scale and structural modifications of 2D TMDs and their heterostructures via post-treatment. Atomic-scale modifications such as vacancy generation, substitutional doping, functionalization and repair of 2D TMDs and structural modifications including phase transitions and construction of heterostructures are discussed. Such modifications on the physical and chemical properties of 2D TMDs enable the development of various advanced applications including electronic and optoelectronic devices, sensing, catalysis, nanogenerators, and memory and neuromorphic devices. Finally, the challenges and prospects of various post-treatment techniques and related future advanced applications are addressed.

For example, monolayer molybdenum disulde akes grown by chemical vapor deposition (CVD) do not exhibit high carrier mobility in eld-effect transistors (FETs) because of intrinsic sulfur vacancies (S V ) (S V density z 1.24 Â 10 13 cm À2 ), which limits electronic device applications. 16 This shortcoming can be easily solved by using post-treatment techniques such as thermal annealing under a sulfur-rich atmosphere. Therefore, post-treatment is a promising method of controlling the physicochemical properties of TMDs and, thus, enables the development of various advanced applications.
Although the post-treatment of TMDs and their heterostructures are important for numerous device applications, only specic topics related to defect engineering, phase engineering, and substitutional doping have been reviewed. [17][18][19] Furthermore, applications have not been comprehensively discussed in conjunction with post-treatment methods. Therefore, posttreatment methods for TMDs and their heterostructures must be investigated by focusing on atomic-scale and structural modications along with their device applications.
In this review, various post-treatment approaches for the atomic-scale and structural modications of TMDs are summarized (Fig. 1). Atomic-scale modication is classied into four categories: (I) vacancy generation, (II) substitutional doping, (III) functionalization, and (IV) repair. Structural modication consists of two categories: (V) phase transition and (VI) heterostructures. These modications modulate the material properties including optical, electronic, catalytic, magnetic properties and so on. Therefore, various advanced applications including electronic and optoelectronic devices, catalysis, energy storage, sensors, piezoelectricity, nanogenerators, and memory and neuromorphic devices according to each post-

Vacancy generation
Vacancy generation in TMDs is highly desirable for controlling the performance of catalytic, electronic, and optoelectronic devices via modulating the carrier concentration and tuning the catalytic activity. The focus in most of the previous studies was chalcogen vacancy generation (hereaer, S and Se vacancies are denoted as S V and Se V , respectively). Vacancy generation can be categorized into (i) dry etching processes, including plasma treatment, electron/ion-beam irradiation, and thermal annealing, and (ii) wet chemical etching processes.
Dry etching process. The dry etching process has the advantage of precisely controlling vacancy density. Plasma consists of electrically charged particles that are produced aer the ionization of gases. 20 Argon plasma, for example, effectively generates defects in TMDs, including MoS 2 , 21-23 WS 2 , 21 MoSe 2 , 24 WSe 2 , 25 and PtSe 2 . 26 To prevent severe damage to TMDs, mild plasma treatment with H 2 or He is employed. 27 The atomic structure of WSe 2 aer plasma treatment shows S V generation (Fig. 2a). The corresponding Raman spectra of one-, two-, and ve-layer WSe 2 before and aer H 2 plasma treatment display no changes in the peak positions of the characteristic phonon modes (E 1 2g and A 1g ) and their intensities (Fig. 2b), indicating that the crystal lattice of WSe 2 remains even aer the selective removal of Se atoms by plasma. The Se/W ratio of WSe 2 aer plasma treatment is slightly decreased from 2.05 to 1.98 (less than $5 at% of Se V ). The photoluminescence (PL) spectra of the WSe 2 monolayer aer H 2 plasma treatment are broadened and quenched (Fig. 2c). Oxygen plasma treatment is used for the formation of oxygen-transition metal bonds, such as Mo-O in MoS 2 , Re-O in ReS 2 , and Te-O in WTe 2 aer the removal of chalcogen atoms. [28][29][30][31] In the case of WTe 2 , oxygen plasma can efficiently eliminate Te or W atoms, resulting in the formation of transition metal and chalcogen vacancies, in addition to W-O and Te-O bonds. 32 Electron beam irradiation can precisely control the generation of chalcogen vacancies via several mechanisms such as the knock-on effect, ionization, beam-induced chemical etching, and ballistic displacement. 33,34 The atomic structure of electronbeam-irradiated WSe 2 presents both single (Se 1 ) and double selenium vacancies (Se 2 ) (Fig. 2d). 35 The neutral (X 0 ) and negatively charged (X À ) exciton peaks of pristine WSe 2 are observed in its PL spectra. The additional broad defect band (D) with attenuated X 0 and X À intensities are observed for irradiated WSe 2 by performing scanning electron microscopy (SEM) with $10 7 electrons per mm 2 (Fig. 2e). The density functional theory (DFT) band structures of Se 1 and Se 2 represent the formation of unoccupied midgap bands (D 1 and D 2 ), in addition to the occupied one below the valence-band maximum (D 3 ) ( Fig. 2f  and g). 35 The number of chalcogen vacancies in MoS 2 and MoSe 2 can be precisely controlled by an electron beam in in situ transmission electron microscopy (TEM). 36,37 In addition to electron beams, ion beams such as argon, 38,39 helium, [40][41][42] manganese, 43 gallium, 44,45 and gold, 46 can be employed to generate vacancies. For example, the S V content in MoS 2 and WS 2 monolayers gradually increases with Ar + ion beam irradiation. [47][48][49][50] Thermal annealing is a simple method for creating chalcogen vacancies via thermal desorption. A H 2 atmosphere is commonly used during thermal annealing to etch away chalcogen atoms via H 2 S and H 2 Se formation (Fig. 2h). The thermal texturization process of MoS 2 akes is also observed. 51 The XPS spectra of the Mo 3d core level for MoS 2 aer annealing at elevated temperatures (700, 800, and 900 C) shows the gradual evolution of peaks at 232 and 229 eV (assigned to the Mo 0 doublet) and 236.5 eV (assigned to Mo 6+ ), in addition to a doublet at 233 and 230 eV (assigned to Mo 4+ 3d 3/2 and 3d 5/2 , Fig. 2 Vacancy generation of 2D TMDs with dry (plasma treatment, electron beam irradiation, and thermal annealing) and wet chemical etching processes. Plasma treatment: (a) atomic structure of WSe 2 layers before and after the H 2 plasma treatment, illustrating the creation of Se vacancies. (b) Raman spectra of monolayer, bilayer, and five-layer WSe 2 before and after H 2 plasma treatment, respectively. (c) PL spectra of the WSe 2 monolayer before and after H 2 plasma treatment at room temperature. Electron beam irradiation: (d) illustration of the atomic structure of WSe 2 with a single selenium vacancy (Se 1 ) and double selenium vacancy (Se 2 ) after electron beam irradiation. (e) PL spectra of the pristine and electron beam irradiated WSe 2 , taken at 5 K. respectively) for pristine MoS 2 (Fig. 2i). Such evolution of a Mo 0 doublet and Mo 6+ peaks is attributed to the formation of Mo metal clusters (by the removal of S atoms) and MoO 3 at S V sites (due to Mo oxidation under air), respectively. In contrast, the S 2p doublet in the S 2p core-level spectra is still present, indicating that some MoS 2 can be maintained, regardless of the annealing temperature.
Wet chemical etching process. Wet chemical etching is a facile and mild strategy for creating atomic chalcogen and transition metal vacancies. Various reagents, such as H 2 O 2 , 52,53 NaBH 4 , 54-56 hydrazine, 57 and HCl, 58 have been utilized. For example, MoS 2 nanosheets are immersed in a H 2 O 2 solution to generate vacancies (Fig. 2j). The Raman spectra of the MoS 2 nanosheets with H 2 O 2 treatment show that both the E 1 2g and A 1g peaks gradually broaden and eventually disappear, indicating that the S and Mo atoms are progressively etched away over the treatment time (Fig. 2k). 52 Phonon soening related to the redshi of the E 1 2g peak is also observed. The signicant intensity changes in the E 1 2g and A 1g peaks are attributed to a high vacancy generation of $15 at% aer chemical etching for 150 s. NaBH 4 is used for solid-phase reduction to create S V in the basal plane of MoS 2 (Fig. 2l). 54 NaBH 4 exclusively reduces Mo 4+ to Mo d+ (d < 4), resulting in the removal of sulfur atoms by forming H 2 S and Na 2 S. The concentration of S V -MoS 2 increases almost linearly with the mass ratio of NaBH 4 : MoS 2 , indicating that the concentration of vacancies can be efficiently controlled by the amount of reagent (Fig. 2m). Hydrazine is used to donate electrons to WS 2 , in addition to creating S V . 57 Although various types of dry etching processes (i.e. plasma treatment, electron/ion beam irradiation, and thermal annealing) are employed for vacancy engineering, the detailed mechanisms of vacancy generation are still not well-understood. Plasma or ion beams are more desirable for patterned device applications, whereas thermal annealing is preferable for catalytic applications. Chemical wet etching frequently induces the formation of undesired vacancies and oxides.
Applications. The vacancies in TMDs can be utilized for various applications, including FETs, electrocatalytic microcells for hydrogen evolution, CO 2 hydrogenation catalysts, and Li-air batteries. Anion vacancies, such as S V and Se V , serve as n-type dopants. Thus, the electrical conductivity of TMDs increases with the vacancy concentration until a certain point and then decreases at high concentrations because of the collapse of the crystal structure. 27,[59][60][61] The optimized concentration of Se V minimized contact resistance by forming edge contacts in a homojunction (Fig. 3a). Mild H 2 plasma was introduced to selectively generate Se V in the contact region in the WSe 2 FET. As a result, the on-current was signicantly increased by a factor of 20, attaining a low subthreshold swing (SS) of 66 mV dec À1 (Fig. 3b). 27 Furthermore, semiconductor-to-metallic transitions in MoS 2 and WS 2 were also observed with low-energy He plasma and He + beam irradiation treatments. [61][62][63] This transition was attributed to the emergence of mid-gap states near the Fermi level at an appropriate vacancy concentration.
The edge of 2H-MoS 2 acts as an active site for hydrogen evolution, whereas the basal plane is inactive. 64 In addition, low electrical conductivity limits the electrocatalytic activity of MoS 2 . Vacancy generation in MoS 2 is very useful for activating its basal plane for the hydrogen evolution reaction (HER) and increasing its electrical conductivity. [65][66][67] For example, the basal plane of MoS 2 was irradiated with a He + beam (Fig. 3c). Consequently, with increasing S V concentration, the linear sweep voltammetry (LSV) curves shied toward the Pt reference up to an optimum concentration of 5.7 cm 14 cm À2 , 68 which is clear evidence of the activation of the basal plane by vacancy generation. When the S V concentration was increased above this threshold, the LSV curves shied backward (Fig. 3d), attributed to the collapse of the MoS 2 structure. Many approaches, including NaClO treatment, 69 ozone treatment, 70 and laser treatment, 71 have been used for vacancy generation to enhance the HER activity of TMDs. In quantum information applications, single-photon emission from the defects of the WSe 2 monolayer induced by e-beam irradiation has recently been reported. 72 The basal plane of S V -MoS 2 is an ideal active site for lowtemperature hydrogenation of CO 2 to selectively produce methanol via the following reaction mechanism: (i) dissociation of CO 2 to surface-bound CO* and O* at S V sites, (ii) hydrogenation of CO* to CH 3 O*, and (iii) synthesis of CH 3 OH (inset of Fig. 3e). As a result, S V -rich MoS 2 nanosheet catalysts showed a high methanol selectivity of 94.3% and CO 2 conversion of 12.5%. Furthermore, they were stable over 3000 h at 180 C (Fig. 3e). 73 S V in MoS 2 nanoowers prepared by thermal annealing in a hydrogen environment also improves solardriven CO 2 photoreduction. 74 The rate of CO production was enhanced approximately 2-fold aer S V generation.
Chalcogen vacancies provide abundant active sites for the intercalation/deintercalation of guest ions (Li + , Na + , and K + ), which leads to enhanced reaction kinetics and improved specic capacities. For example, sulfur vacancies in MoS 2 can intrinsically promote O 2 adsorption, enhancing the electrochemical performance of Li-O 2 batteries. 75 The core-shell MoS 2Àx @CNT composite synthesized by hydrothermal and thermal annealing was treated with NaBH 4 to increase S V concentration (Fig. 3f). The initial discharge/charge prole at 200 mA g À1 was signicantly boosted up to discharge/charge specic capacities of 19 989/17 705 mA h g À1 with an overpotential of 0.99/0.26 V (Fig. 3g). 76 Furthermore, S V improves polysulde conversion kinetics in Li-S batteries, 77 facilitates the absorption of Na + /Zn + , and increases the conductivity of Na/Znion batteries. 78,79 Depending on the relative valency of the dopant atoms, they act as electron donors or acceptors. When a Janus group-VI chalcogenide MXY (top layer X, bottom layer Y ¼ S, Se, and Te; X s Y) is formed, the out-of-plane mirror symmetry is broken. This unique asymmetrical feature of Janus structures arises from different atomic radii and electronegativities of X and Y atoms, thus enabling novel applications such as piezoelectric devices and electrocatalysts. 84,85 Impurity doping. Non-metallic (NM) doping of TMDs has been performed by substituting chalcogen atoms with O, Te, Cl, N, P, and F atoms. [86][87][88][89][90][91][92][93][94] For example, carbon doping was carried out via plasma-induced CH 4 gas exposure of WS 2 monolayers (Fig. 4a). 95 No noticeable damage was observed in the monolayers, and their PL intensity progressively decreased and shied toward lower energy values with increasing carbon content (Fig. 4b). Theoretical simulations predicted that, unlike single C doping, CH doping provided the most stable and lowest local stain on WS 2 . The calculated bandgap of pristine WS 2 was 1.791 eV, whereas that of carbon-doped WS 2 had a direct bandgap of 1.574 eV (Fig. 4c). Aer carbon doping, the energy gap is decreased by raising new acceptor levels above the valence band of WS 2 owing to the hybridization of W d-orbitals and C p-orbitals. The extra holes can move to new levels, implying that CH impurities act as p-type dopants.
Transition metal (TM) doping at the M site of a TMD is conducted in three steps: (i) generation of chalcogen vacancies, (ii) replacement of transition metal-adjacent chalcogen vacancy sites, and (iii) healing of chalcogen vacancies. [96][97][98][99][100] For example, WS 2 monolayers were rst grown with S V sites, followed by subsequent exposure to a Sn-rich atmosphere using SnS as a precursor at 550 C (Fig. 4d). 101 The characteristic PL peak of WS 2 was gradually attenuated with doping time and dopant concentration (Fig. 4e). Moreover, Sn dopants in the WS 2 lattice act as electron donors (i.e., n-type dopants). Another strategy for TM doping is the direct use of a metal ux with the aid of electron beam evaporation (Fig. 4f). 97,102 In this case, dopant beams (such as Nb and Re) with low kinetic energy are generated by thermal evaporation of high-melting-point metals on an exposed TMD. The beam ux is modulated to supply metal dopants during the entire doping process to enhance structural reconstruction and regulate the formation of metal-doped TMDs. Concurrently, a Se beam is continuously supplied to heal possible Se vacancies during the doping process. Accurate and position-selective doping can be achieved when patterned TMD materials are used. 102 Molecular dynamics (MD) simulations showed a gradual structural change with Nb and Se exposure (Fig. 4g). When a Nb atom hit a W atom with substantial energy (simulation time is varied from 0 to 100 fs), the W atom was released from the original position. Simultaneously, the Nb atom replaced the vacancy created by the released W atom. Fig. 4h displays an atomic-resolution highangle annular dark-eld (HAADF)-scanning transmission electron microscopy (STEM) image of monolayer WSe 2 aer exposure to a Nb beam. The cutouts from the white squares in Fig. 4h clearly show the atomic structure of Nb-doped WSe 2 : bright and dark spots at W sites in the hexagonal lattice are assigned to W and Nb atoms (as shown in the structural model). A summary of impurity doping in CVD-grown monolayers and exfoliated TMDs is presented in Table 1.
Janus structures. Janus TMD structures can be obtained using two representative methods: (i) thermal annealing under a chalcogen-rich atmosphere and (ii) H 2 plasma stripping and thermal chalcogenization at low temperature (hereaer denoted as remote plasma-assisted chalcogenization). [103][104][105][106] In the past, Janus structures were constructed by a thermalannealing process under a secondary chalcogen atmosphere. For example, Janus MoSSe and MoSeS structures were produced by annealing MoSe 2 and MoS 2 at temperatures of $800 and $450 C under S and Se atmospheres, respectively. 103,104 However, high temperatures are unfavorable for 2D Janus monolayers because of a high probability of alloy or defect formation, such as chalcogen vacancies and cracks.
Remote plasma-assisted chalcogenization consists of two consecutive steps: (i) remote hydrogen plasma treatment of a CVD-grown MoS 2 monolayer to strip off sulfur atoms from the top layer and replace them with hydrogen atoms and (ii) replacement of H atoms with Se through a thermal selenization process at $450 C to form a structurally stable Janus MoSSe monolayer (Mo atoms are covalently bonded to the underlying S and top-layer Se atoms) (Fig. 5a). 103 Similarly, a roomtemperature atomic layer substitution (RT-ALS) process was recently developed (Fig. 5b-i). 107 This less disruptive technique employs hydrogen radicals produced by a remote plasma to strip chalcogen atoms on the top layer of an as-grown TMD. Concurrently, vaporized chalcogens (Se or S) are supplied to substitute stripped atoms, resulting in an asymmetric Janus structure at room temperature in the form of MXY (M ¼ Mo or W, X ¼ S or Se, and Y ¼ Se or S). Fig. 5b(ii) depicts the formation of a Janus structure: (I) before H and chalcogen adsorption, (II) adsorption and diffusion of two H atoms to the  same S, (III) formation of H 2 S, (IV) desorption of H 2 S, and (V) Se occupation of the S vacancy. The free energy values for each step are shown in Fig. 5b(iii). The tilted ADF-STEM image reveals that Se atoms are located on one side of the monolayer MoSSe, while S atoms are on the opposite side. This is direct evidence of a Janus structure (Fig. 5c). The corresponding intensity prole in Fig. 5d clearly shows individual Mo, Se, and S atoms with peak intensities proportional to their atomic numbers. Furthermore, the MoS 2 A 1g (404 cm À1 ) and E 2g (383 cm À1 ) modes in the Raman spectra shied to 288 cm À1 and 355 cm À1 , respectively, due to disrupted symmetry in the vertical direction caused by the formation of a Janus structure (Fig. 5e). Aer further selenization processing of Janus MoSSe was performed, a sharp peak at $239 cm À1 and a broad peak at $284 cm À1 were formed, which are the signature peaks of the A 1g and E 2g modes in monolayer MoSe 2 . In addition, the PL shis from 1.85 eV (pristine MoS 2 ) to 1.72 eV (Janus MoSSe) and nally to 1.60 eV (converted MoSe 2 ) are clearly observed (Fig. 5f). 103,105 Room-temperature doping with remote plasma-assisted chalcogenization offers the possibility of producing highquality TMDs and 2D Janus structures, which will advance the fabrication techniques for industrial applications as a desirable emerging platform.
Applications. The tunable material properties obtained by substitutional doping of TMDs have been utilized to realize novel applications, including electronics, biosensors, catalysis, optoelectronic applications, magnetization, and photocatalysis. Impurity doping is widely used to modulate the electrical properties of TMD materials. 91 For example, the I-V transfer curve of n-type-doped multilayer MoS 2 obtained by N 2 plasma exposure distinctly shows a positive threshold voltage (V th ) shi, which is consistent with the p-type dopant behavior of nitrogen in MoS 2 (Fig. 6a). Moreover, Mn-doped MoS 2 (Mn-MoS 2 ) was used to selectively detect dopamine (DA) levels in serum and articial sweat. 108 Abnormal levels of dopamine in the body can be symptomatic of several disorders such as Alzheimer's disease, schizophrenia, and Parkinson's disease. 109 Previously, DA detection was achieved by employing highly sophisticated methods, such as mass spectrometry, liquid chromatography, and electrochemical detection measurements. 110 Therefore, a low-cost but accurate diagnostic tool for the detection of DA levels is essential. A wearable DA sensor was fabricated on a exible polyimide (PI) sheet with a Mn-MoS 2 working electrode (WE), a pyrolytic graphite sheet (PGS) counter electrode (CE), and an Ag paste reference electrode (RE) (Fig. 6b). DA concentrations as low as 50 nM were successfully detected in articial sweat containing 5 mM glucose (Fig. 6c). Furthermore, Co-doped defective MoS 2 (Co-MoS 2 ) exhibits superior dinitrogen-to-ammonia conversion activity compared with pristine MoS 2 and CoS 2 (Fig. 6d). 111 Such a high faradaic efficiency and production rate are attributed to the effective activation of the dinitrogen molecule for the dissociation of the N^N triple bond in defective MoS 2Àx .
The electrocatalytic conversion of CO 2 into sustainable fuels is considered the most efficient approach for achieving carbon neutrality. 112 Nb-doped MoS 2 can reduce CO 2 to produce useful hydrocarbon derivatives, such as methane and ethanol, along with H 2 (Fig. 6e). The CO formation turnover frequency (TOF) of Nb-doped MoS 2 in an ionic liquid is one order of magnitude higher than that of Ta-doped MoS 2 or Ag NPs in the overpotential range of 50-150 mV. Furthermore, the current density of Nb-doped MoS 2 in LSV experiments was approximately 2 and 50 times higher than that of pristine MoS 2 and Ag NPs, respectively.
The Janus structure of TMDs has been reported to possess piezoelectric properties due to its non-symmetrical structure, which generates electrical polarization in response to externally applied mechanical stress. 103 The resonance-enhanced piezoresponse force microscopy image of Janus MoSSe shows the presence of piezoelectric properties (Fig. 6f), which are not observed in pristine MoS 2 . In addition, Janus TMD monolayers effectively activate TMD basal planes for the HER. 104 Janus SMoSe (for MoSSe) and its reverse conguration of SeMoS (for MoSeS) were constructed by atomic substitution of pure MoS 2 and MoSe 2 , respectively (Fig. 6g). Both Janus SMoSe and SeMoS monolayers exhibit lower overpotentials in the LSV curves than pure MoS 2 and MoSe 2 . Moreover, the HER activity of SeMoS surpasses that of SMoSe, which is attributed to the greater thermoneutral Gibbs free energy (DG H ) for SeMoS in the presence of Se V (À0.007 eV for SeMoS and 0.060 eV for SMoSe). The unique structure of Janus TMDs enables band edge modulation and electronic transfer when heterostructures are designed. Taking advantage of the type-II band alignment of MoSSe-MoS 2 , a back-gate transistor was fabricated to measure the fourprobe output characteristics (Fig. 6h). 107 A plot of current density (I D /W) versus bias voltage (V M ) at various applied backgate voltages (V G ) shows weak rectication behavior at the lateral junction.
Semiconducting 2H-TMDs, such as MoS 2 , WS 2 , and MoSe 2 , are chemically unreactive because they are free from dangling bonds on their surfaces. However, aer the phase transition from 2H to negatively charged metallic 1T-TMDs via lithiation, they become reactive to covalent functionalizations with alkyl halides and diazonium salts (Fig. 7a). 119 For example, the MoS 2 exciton peak in the PL spectra almost completely disappeared aer the phase transition ( Fig. 7b) and two prominent peaks evolved in functionalized 1T-MoS 2 (Fct-1T-MoS 2 ). The peak at $1.6 eV may be attributed to the band structure modication by covalent functionalization, while the peak at $1.9 eV is related to the up-shi of the MoS 2 exciton emission. 119 The appearance of PL peaks indicates that the metallic 1T phase is converted to the semiconducting 1T one aer functionalization. This was further conrmed by the characteristics of the fabricated eldeffect transistor.
Aryl diazonium functionalization can occur at the nearest S V sites of MoS 2 . S V sites, where charge is accumulated, reduce 4nitrobenzenediazonium tetrauoroborate (4-NBD), resulting in nitrophenyl (NP) radical graing aer N 2 release (Fig. 7c). 120 Localized charge density adjacent to the graing sites is renormalized aer functionalization, further triggering a chainlike growth propagation of subsequent NP molecules over MoS 2 sulfur sites (top panels in Fig. 7d). A higher S V density promotes numerous initiation reaction sites for the propagation of NP molecules over the entire MoS 2 basal plane (bottom panels in Fig. 7d). The semiconducting nature of MoS 2 is still maintained aer functionalization. The NP functionalization of S V -MoS 2 obeys pseudo-second-order (adsorbate-surface and adsorbateadsorbate interactions) reaction kinetics, 121 whereas the functionalization of Gr obeys rst-order (adsorbate-surface interactions) reaction kinetics. 122 On the other hand, S sites in MoS 2 can be directly functionalized with NP radicals via active chemical reduction of 4-NBD ions, using potassium iodide as an electron donor (Fig. 7e). 123 Similarly, alkyl halide-functionalized 1T 0 -MoS 2 is activated using metallocene reducing agents, which facilitate high surface coverage with alkyl halide groups. 124 In addition, n-type doping with KI (reducing agent) itself has been observed, providing advantages for low contact resistance and increased charge carrier mobility for FET devices. 123 One of the advantages of diazonium salts, in terms of functionalization, is the controllability of the terminated molecules, which enables the engineering of the MoS 2 electronic structure. 1T-MoS 2 readily reacts with various diazonium salts, including different functional groups (NO 2 Ph, Cl 2 Ph, BrPh, MeOPh, and Et 2 NPh) (Fig. 7f). 125 Each functional group has a particular Hammett parameter (Fig. 7g). The measured work function of the functionalized MoS 2 increased almost linearly with the Hammett parameter, indicating that the work function of MoS 2 can be controlled by functional groups. This surface energy change strongly inuences HER catalytic activity. In fact, HER performance is signicantly improved by electron-donating Et 2 NPh groups. 125 Furthermore, covalent functionalization of 1T 0 -MoS 2 with Et 2 NPh improves the stability of electronic transport properties for at least two weeks under atmospheric conditions. 126 Another approach is the selective functionalization of TMD chalcogen vacancies or edge sites with thiol derivative molecules via a conjugation reaction. 127-129 MoS 2 can be functionalized with thiol molecules (2-mercaptoethylamine, MEA, and 1H,1H,2H,2H-peruorodecanethiol, FDT) at S V sites by soaking for 72 h (Fig. 7f). The charge density of functionalized MoS 2 was strongly affected by the terminated molecules. NH 2 groups in MEA donate electrons to MoS 2 (charge density from 9.4 Â 10 11 to 1.4 Â 10 12 cm À2 ), whereas F groups in FDT withdraw electrons from MoS 2 (charge density À7.0 Â 10 11 to À1.8 Â 10 11 cm À2 ). 130 Likewise, functionalization with aromatic or alkyl thiols tunes the electronic structure of MoS 2 . For example, in the case of alkyl thiols, the Fermi energy level (E F ) upshis with increasing chain length, such as for 1-propanethiol, 1-nonanethiol, and 1-dodecanethiol. 131 Phase transition requires harsh chemical treatment of TMDs with a highly pyrophoric compound (n-butyllithium), which deteriorates the quality of the material. 132,133 The diazonium salt functionalization approach is commonly used to engineer the electronic structure of TMDs. Thiol-derivative functionalization with TMDs is inherently limited by the number of vacancy sites.
Non-covalent functionalization. Non-covalent functionalization consists of the physisorption of organic molecules on the surface of TMDs/Gr without chemical bond formation. [134][135][136] This enables chemical doping and formation of heterojunctions on TMDs.
Chemical doping of organic molecules on carbon nanotubes (CNTs) and Gr has been extensively studied using two main approaches: (i) modulating the reduction potential of molecules (water bucket model) 137 and (ii) controlling electron-donating (e.g., amine, -NH 2 , and hydroxyl, -OH) and electronwithdrawing groups (e.g., nitro, -NO 2 , and trihalogenated methyl, -CX 3 ). Analogous strategies have been widely adopted for TMDs. The water bucket model describes the charge transfer between molecules and host materials induced by the difference in reduction potentials. Species with a lower reduction potential give electrons to those with a higher one. 137 Material's reduction potential can be calculated by using the following equation: 4/e ¼ V (V vs. SHE) + 4.44 V, where 4, e, V, and SHE denote the work function, electron, reduction potential, and standard hydrogen electrode, respectively. 138 For MoS 2 , the reduction potentials of 0.84 eV (vs. SHE) for 2,3,5,6-tetrauoro-7,7,8,8-tetracyanoquinodimethane (F 4 TCNQ) and 0.46 eV (vs. SHE) for 7,7,8,8-tetracyanoquinodimethane (TCNQ) indicate that they withdraw electrons from MoS 2 (p-type doping), whereas À0.32 eV (vs. SHE) for nicotinamide adenine dinucleotide (NADH) indicates that they donate electrons (n-type doping) (Fig. 8a). 139,140 A drastic enhancement in the PL intensity was observed for MoS 2 aer functionalization with F 4 TCNQ and TCNQ (Fig. 8b). In contrast, aer functionalization with NADH an attenuation of the PL intensity was detected (Fig. 8c). 141 These experimental results demonstrate the E F shi of MoS 2 according to p-type and n-type dopants.
Molecules with distinct functional groups can be n-or p-type dopants of TMDs. For example, mechanically exfoliated WSe 2 akes were exposed to vapors of trichloro(1H,1H,2H,2H-per-uorooctyl)silane (PFS) with -CF 3 functional groups for p-type doping, and N-[3-(trimethoxysilyl)propyl]ethylenediamine (AHAPS) with -NH 2 functional groups for n-type doping (Fig. 8d). Before this, however, WSe 2 akes were treated with ozone (UVO 3 ) to improve the functionalization uniformity and ensure high surface coverage. The I-V transfer curves of the WSe 2 /PFS and WSe 2 /AHAPS FETs, fabricated on Si/SiO 2 with Au contact electrodes, show p-type and n-type characteristics with enhanced hole and electron mobilities of 150 cm 2 V À1 s À1 and 17.9 cm 2 V À1 s À1 , respectively ( Fig. 8e and f). Moreover, asymmetric doping obtained by sandwiching bilayer WSe 2 with AHAPS (bottom) and PFS (top) shows distinct ambipolar transport properties with electron and hole mobilities of 5.7 cm 2 V À1 s À1 and 20 cm 2 V À1 s À1 , respectively (Fig. 8g). 142 To achieve better ambipolar transport characteristics, monolayer WSe 2 was subjected to a hybrid functionalization (covalent and non-covalent functionalizations) using two different dopants, 4-NBD and diethylenetriamine. It drastically enhanced the hole and electron mobilities to 82 cm 2 V À1 s À1 and 25 cm 2 V À1 s À1 , respectively. 143 In addition, a reduction in the energy bandgap ($0.24 eV) was observed for WSe 2 , with asymmetric doping with C 60 F 48 and graphite. This behavior is a consequence of the accumulation of holes and electrons at the bottom and top Se layers. 144 Various heterojunctions can be easily constructed by the deposition of organic layers on TMD surfaces. Fig. 8h shows a typical prototype of a type-II organic/TMD heterojunction (pentacene/MoS). When an electron in MoS 2 is excited by a photon, a hole is transferred to the p-type pentacene layer within a very short time (s 2 ¼ 6.7 ps) (Fig. 8i). This enables the extension of the interlayer exciton lifetime to as long as $5 ns at the pentacene/MoS 2 interface (Fig. 8j). 145 This dynamic process originates from the quenching of the PL intensity of pentacene/ MoS 2 when compared to pristine MoS 2 . Similarly, MoS 2 /PTCDA, WS 2 /PTCDA, and MoS 2 /rubrene also exhibit a type-II p-n heterojunction. 134,146,147 Such an extended charge recombination time can improve the quantum efficiency of optoelectronic devices such as diodes, bipolar transistors, photodiodes, and so on. [148][149][150] Applications. COVID-19 sensors, TENGs, multifunctional optoelectronic devices, and memory and neuromorphic devices are some of the applications discussed here. The development of rapid diagnostic tools for the detection of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is highly needed. Real-time reverse transcription-polymerase chain reaction (RT-PCR) is the only specic diagnostic test based on unique sequences of viral ribonucleic acid (RNA). It is, however, very time-consuming. Hence, having rapid, effective, label-free, point-of-care, and low-cost sensors for detecting viral antigens is crucial. 151-153 WSe 2 monolayers were functionalized with 11mercaptoundecanoic acid (MUA) and activated with N-hydroxysuccinimide (NHS) for the detection of SARS-CoV-2 spike proteins (Fig. 9a). Fig. 9b illustrates the response of the device during real-time detection of SARS-CoV-2 under different concentrations of the spike protein. The number of antibodies with SARS-CoV-2 spike protein increased with the concentration (red line), while antibodies without SARS-CoV-2 spike protein showed a slight decrease (blue line). Moreover, the biosensor based on WSe 2 FETs aer functionalization with MUA can detect SARS-CoV-2 spike protein down to 25 fg mL À1 in 0.01 M phosphate-buffered saline (PBS) solution, 154 which is $10 4 orders of magnitude higher than the detectivity of Gr-based biosensors (1 fg mL À1 ). 155 With the rapid development of the Internet of Things (IoT), over 50 billion IoT sensors already exist and are expected to surpass 200 billion by 2025. 156,157 Among various energy harvesters, TENGs can convert waste mechanical energy into electrical energy under ambient conditions. WS 2 nanosheets were functionalized via thiol conjugation reactions at S V sites with various alkanethiol molecules, such as mercaptopropionic acid (MPA), mercaptohexanoic acid (MHA), mercaptooctanoic acid (MOA), and MUA. A TENG device was fabricated using functionalized WS 2 and poly(ethylene terephthalate) (PET) as the negative and positive active layers, respectively. Indium tin oxide was used as an electrode (Fig. 9c). Thiol-containing ligands with different alkane chain lengths act as tribo-electrication layers in TENGs. A high-precision micromechanical tester was used to press and release a TENG device with a vertical force of 6 N and a frequency of 1 Hz. Pristine  (Fig. 9d). The MUA-WS 2 TENG exhibited a maximum power density of 138 mW m À2 . The persistent output voltage of the MUA-WS 2 TENG device was maintained aer 14, 28, 56, and 98 days, conrming high stability and durability. 158 This indicates that alkanethiol functionalization of defective WS 2 surfaces via ligand conjugation suppresses reactions with reactive oxygen and reduces the number of catalytically active sites. Moreover, a stable triboelectric output voltage was observed for the MUA-WS 2 TENG device even aer 10 000 cycles. Several 2D materials, including MoS 2 , WS 2 , MoSe 2 , Gr, and Gr oxide (GO)-based TENGs, were fabricated to utilize their triboelectric charging nature. The work functions of the materials decreased in the following order: MoS 2 (4.85 eV) > MoSe 2 (4.70 eV) > Gr (4.65 eV) > GO (4.56 eV) > WS 2 (4.54 eV). Hence, MoS 2 is likely to be triboelectrically more negative than the other materials. Furthermore, MoS 2 is functionalized with benzyl viologen (BV, n-type) and gold chloride (AuCl 3 , p-type), resulting in positive and negative values over a triboelectric series (by changing the work function accordingly). These results suggest that triboelectric charging can be tuned through functionalization. 159 Another example of the application is a photomolecular switch device, which was fabricated using thin photochromic azobenzene (AZO) physisorbed on MoS 2 that was a trap-free benzocyclobutene (BCB)/SiO 2 /Si substrate (Fig. 9e). The AZO molecules on MoS 2 undergo reversible isomerization between the cis and trans states upon exposure to white and UV light illumination, which efficiently modulates the charge transfer process between AZO and the underlying 2D MoS 2 . The photoresponse of the MoS 2 FET device on BCB represents a fast saturation under 530 nm light illumination for 20 s, whereas it shows a steady increase for MoS 2 on a SiO 2 /Si substrate (Fig. 9f). This high response originates from the suppression of the persistent photoconductivity (PPC) effect. Furthermore, a high thermal stability of over 15 h was demonstrated for the photomolecular switching of metastable cis/trans states. 160 In addition, a type-II heterojunction phototransistor was developed using an organic phosphonic acid monolayer (12- benzo [4,5]thieno [2,3-d]thiophen-2-yl)dodecyl) (BTBT) stacked over a MoS 2 monolayer. It exhibits an unprecedented responsivity of 475 A W À1 and high external quantum efficiency of 1.45 Â 10 5 %. BTBT provides effective charge transfer via p-p interactions and successfully eliminates recombination and charge scattering. 161 A ternary-responsive multilevel memory device was fabricated with few-layer WSe 2 via asymmetric non-covalent functionalization. A photochromic diarylethene (DAE) layer functionalized WSe 2 at the bottom, while a ferroelectric poly(vinylidene uoride-triuoroethylene) (P(VDF-TrFE)) layer was placed on the top of WSe 2 to form a multi-stimuli-responsive Janus WSe 2 FET device (Fig. 9g). The device successfully generated two states because of the polarization of P(VDF-TrFE) in the downward/upward direction for a cyclic AE60 V pulse bias. One cycle consisted of +60 V for 2 s (blue region), 0 V for 30 s, À60 V for 2 s (violet region) and 0 V for 30 s (Fig. 9h (top)). Switching with a photochromic DAE was also demonstrated under UV-Vis illumination, where one cycle consisted of 5 s under UV illumination (orange region), 20 s in the dark, 20 s under Vis light (green region), and 20 s without illumination (Fig. 9h (bottom)). Moreover, a multistimuli-responsive asymmetrically functionalized Janus WSe 2 device was successfully achieved by modulating the population ratio of polarized P(VDF-TrFE) and photoisomerization DAE. This generated nine unique ferroelectric states and 84 photogenerated states, respectively. The overall maximum of 756 current levels was stored in a single device. The cyclic endurance (10 cycles) and data retention (over 1000 h) conrm the consistency of the device and present high-density non-volatile memory devices. 162 A multifunctional neuromorphic device was fabricated via non-covalent functionalization of perylene-3,4,9,10tetracarboxylic dianhydride (PTCDA) on MoS 2 , which forms a type-II heterostructure (Fig. 9i). The charge transfer phenomenon between MoS 2 and PTCDA mimics neurotransmitter release in biological synapses (right panel of Fig. 9i). The MoS 2 /PTCDA hybrid synaptic transistor was operated under an electrical eld by applying a pair of negative gate voltage (V cg ) pulses (À12 V for 25 ms and À20 V for 25 ms). It exhibited inhibitory post-synaptic current (IPSC) behavior (A 2 < A 1 ) and excitatory post-synaptic current (EPSC) behavior (A 2 > A 1 ), which correspond to synaptic paired-pulse depression (PPD) and paired-pulse facilitation (PPF) behavior, respectively ( Fig. 9j and  k). Similarly, a typical EPSC behavior is observed by applying a pair of 532 nm laser pulses (green irradiation with a pulse width of 400 ms, an interval of 100 ms, and a V DS of 0.1 V) (Fig. 9l). Under both operating modes, the type-II MoS 2 /PTCDA hybrid heterojunction synaptic device performed consistently and successfully mimicked biological synapse functions by an efficient charge transfer process at the hybrid interface. 163

Repair
Atomic chalcogen vacancies in TMDs attenuate their electrical and optical properties (as seen by low PLQY, high Schottky barriers, and limited carrier mobility), which is translated to poor electronic and optoelectronic device performances. [164][165][166][167][168][169] Therefore, the repair of atomic defects is critical for the realization of high-performance devices. Atomic chalcogen vacancies can be repaired using the same chalcogens for healing and other elements for passivation.
Healing. Hydrosulfurization and post-sulfuration/ selenization treatments effectively heal chalcogen vacancies in TMDs. 170 For example, atmospheric oxygen atoms can easily react with S V or Se V and passivate CVD-grown or exfoliated TMDs (Fig. 10a). 171,172 Two possible reactions can occur between oxygen atoms and TMDs: (i) formation of oxygen-substituted TMDs (MoS 2Àx O x ) and (ii) creation of SO 2 or SeO 2 volatile compounds. DFT calculations show that the removal of S atoms is thermodynamically favorable (E ¼ À0.49 eV, negative oxidation enthalpy), whereas it is unfavorable for Se atoms (E ¼ +0.75 eV, positive oxidation enthalpy). Bright triangular spots corresponding to single O atoms were observed in the scanning tunneling microscope (STM) image (Fig. 10b). Such MoS 2Àx O x compounds are completely healed and transformed to defectfree MoS 2 by simple thermal annealing at 200 C under a H 2 S atmosphere (Fig. 10c). 173 It is noted that MoS 2Àx O x shows better HER performance than defect-free MoS 2 because the presence of oxygen atoms in TMDs decreases the DG H for hydrogen adsorption, by bringing it closer to thermoneutral conditions. 173 The post-selenization process is performed on defected monolayer MoSe 2 via pulsed laser vaporization of selenium to repair Se V (Fig. 10d). The temperature plays a crucial role in selenization, where the optimum temperature range is between 600 and 700 C. The I-V transfer curves show major carrier type conversion from p-type characteristics for high Se vacancies ($20% V Se ) to n-type characteristics for healed Se vacancies ($9% V Se ). Moreover, the obtained hole and electron mobilities were $0.011 and 0.021 cm 2 V À1 s À1 , respectively (Fig. 10e). 174 Selective post-sulfuration on patterned MoSe 2 readily forms MoSe 2 /MoS 2 heterojunctions and exhibits a type-I band alignment. 175 Another healing method for S V in MoS 2 (S V -MoS 2 ) involves the use of S-containing organic molecules. 176 For example, S V -MoS 2 was coated with 3-mercaptopropyl trimethoxysilane (MPS) and subsequently annealed at 350 C under a H 2 /Ar atmosphere. The healing process involves the following reaction: HS(CH 2 ) 3 Si(OCH 3 ) 3 + S V -MoS 2 / CH 3 (CH 2 ) 2 Si(OCH 3 ) 3 + MoS 2 . The reaction kinetics of S V -MoS 2 and MPS involve two representative steps of chemical adsorption between S V and thiol groups in MPS and dissociation of the S-C bond with an energy barrier of 0.51 and 0.22 eV, respectively (Fig. 10f). The S V density was dramatically reduced from $6.5 Â 10 13 to $1.6 Â 10 13 cm À2 , for topside MPS treatment. Electrical transport properties were measured for three FET samples on Si/SiO 2 : exfoliated MoS 2 (black), top-side-treated MoS 2 (blue), and double-sidetreated MoS 2 (red) (Fig. 10g). The double-sided MPS treatment further reduced short-range scattering and charge impurities and thus enhanced the carrier mobility of 81 cm 2 V À1 s À1 at room temperature. 177 The carrier mobility of these samples at low temperature (10 K) was further increased to 14 cm 2 V À1 s À1 , 106 cm 2 V À1 s À1 , and 320 cm 2 V À1 s À1 , respectively (Fig. 10h). The catalytic properties of MPS-treated MoS 2 drastically decrease due to the depletion of electrochemically active sites, thereby increasing the overpotential and the Tafel slope. 178 As an alternative, bis(triuoromethane) sulfonamide (TFSI) was employed to heal S V in MoS 2 and WS 2 . 179 The exfoliated MoS 2 that was treated with TFSI exhibited a 190-fold increase in the magnitude of the PL peak intensity (Fig. 10i), consisting of a brighter PL image than that of the pristine one (Fig. 10j). 180 An increased quantum yield QY (>95%) and longer lifetime ($10 ns) were also observed due to the elimination of the nonradiative recombination. TFSI-treated MoSe 2 and WSe 2 exhibited a moderately reduced QY. TEM analysis revealed that Se V sites in WSe 2 were not passivated by S atoms. 181 Passivation. Passivation has been conducted not only to improve the optical and electrical properties of TMDs but also to stabilize their structures. To enhance the PL intensity of MoSe 2 , the passivation of Se with Br atoms was introduced (Fig. 10k). 182 CVD-grown MoSe 2 shows a 30-fold increase in PL intensity aer HBr treatment due to Se V passivation and pdoping effects. 182 For example, the electrical and optical properties of WSe 2 and ReSe 2 can be improved via HCl treatment. The halogen atoms efficiently repair the Se V sites and shi the defective states from the donor level to the acceptor level (i.e. transforming from n-type to p-type). 183 (Fig. 10l-n). 186 Three different conditions were adopted for the preparation of three distinct MoS 2 monolayers by growing them under oxygen (O-MoS 2 ), sulfur-mild (SM-MoS 2 ), and sulfur-excess (SE-MoS 2 ) conditions. The neutral Aexciton peak intensity of the O-MoS 2 is dominant in the PL spectra (Fig. 10o). This is ascribed to suppressing the nonradiative recombination and p-doping effect by new acceptor states. 186 The healing of atomic chalcogen vacancies using organic thiol molecules is still under debate. Three possible reaction mechanisms include repair, functionalization, and dimerization. 187 The energy barrier rate-determining steps for both functionalization and repair mechanisms are almost similar, indicating that these are competing reactions. 188 Moreover, numerous factors, such as the nature of thiol molecules, concentration of S V , reaction temperature, and time, are critical in determining the reaction mechanism. Furthermore, thiophenol molecules (C 6 H 5 SH) can heal and adsorb on Se V in WSe 2 . The adsorbed thiophenol molecule displays a vertical conguration, which is consistent with experimental STM images. 189 The mechanisms of defect passivation and/or healing are still not clearly understood for TFSI-treated TMDs. TFSI treatment increases the PL lifetime but limits the carrier mobility by the charge scattering mechanism. 181,190 In order to overcome this issue, exfoliated MoS 2 and WS 2 were rst treated with thiol molecules (3-n-octylthiophene, dipropyl sulde, or ethanethiol) to control S V , followed by TFSI treatment. The twostep treatment effectively eliminates sub-gap states and lowers the Fermi level, which greatly improves the PL lifetime and enhances carrier mobility. 191 Application. Successfully repaired TMDs can be utilized in various applications, including exible PENGs, superconductors, FET contact resistance, and photodiodes. Monolayer 2D TMDs exhibit strong piezoelectric properties when an external force is applied because of their broken inversion symmetry. 192 A exible PENG was fabricated by using healed MoS 2 , obtained via a thermal annealing treatment at 1000 C for 30 min under H 2 S gas ( Fig. 11a and b). The lateral piezoelectric response was measured by applying an external electric eld between two electrodes. Fig. 11c illustrates the piezoelectric responses of asgrown monolayer MoS 2 , S-treated MoS 2 (healed MoS 2 ), and aquartz as a function of the applied external bias. The piezoelectric output current and voltage were measured under a tensile strain of 0.48% with a strain rate of 70 mm s À1 . The piezoelectric response of the S-treated MoS 2 surpasses that of the pristine monolayer MoS 2 and a-quartz. The inset in Fig. 11c shows that the piezoelectric coefficients (d 11 ) are 3.73 pm V À1 , 3.06 pm V À1 , and 2.3 pm V À1 for S-treated MoS 2 , pristine MoS 2 , and a-quartz, respectively. The output current and voltage generated by the S-treated MoS 2 PENG device (100 pA and 22 mV) are $3 and 2 times higher than those of the pristine MoS 2 PENG device (30 pA and 10 mV), respectively. 193 A photodiode was fabricated using selective healing of patterned MoS 2 . Self-healing is carried out by the hydrogenation of poly (3,4- Fig. 11e. V T shied toward zero aer self-healing, indicating unipolar n-type electrical transport behavior. The MoS 2 homojunction photodiode exhibits a high photoresponsivity of 308 mA W À1 (observed at zero source-drain bias) with excellent air stability (Fig. 11f), as a result of efficient electron-hole separation at the homojunction. 194 Recently, research on ultrathin superconductors has been exclusively conducted aer the advancement in exfoliation (from bulk down to monolayer) and encapsulation techniques of 2D materials. However, monolayer TMDs possess a high density of defects, eventually resulting in the localization of Cooper pairs and decreasing transition temperature (T c ) for metal-to-insulator transitions. [195][196][197] S V sites in ultrathin TaS 2 are passivated by oxygen to form oxygenated TaS 2 in air (Fig. 11g). Theoretical calculations predict that oxygen passivation in monolayer TaS 2 signicantly decreases the carrier density of pure TaS 2 . Transport measurements were performed for three-(3L) and ve-layered (5L) TaS 2 devices, either in a fresh (denoted by new) or oxygenated (denoted by aged) form (Fig. 11h). The T c values of both 3L and 5L TaS 2 increase with oxygen, owing to the increase in electron-phonon coupling. 198 FETs were fabricated using three different types of monolayer MoS 2 : those grown under oxygen conditions (O-MoS 2 ), mild sulfur conditions (SM-MoS 2 ), and excess sulfur conditions (SE-MoS 2 ) (Fig. 11i). The I-V transfer curves of O-MoS 2 display a positive V th (+21.0 AE 4.6 V), whereas a negative V th was observed for SE-MoS 2 (À17.0 AE 9.7 V) and SM-MoS 2 (2.8 AE 4.9 V). The contact resistance (R C ) of the three devices was determined using the Schottky barrier height (SBH) at the interface (Fig. 11j). 186 The O-MoS 2 FET exhibits a low R C value of $1 kU mm, whereas high R C values of 3.9 and 7.8 kU mm are observed for SM-MoS 2 and SE-MoS 2 FETs, respectively (at the same carrier density (n 2D ) of 4 Â 10 12 cm À2 ). This is ascribed to the reduced SBH in O-MoS 2 owing to the absence of donor states and the Fermi level closer to the CBM of MoS 2 .

Structural modification
This section presents a discussion on two structural modications of TMDs for various device applications: (i) phase transition and (ii) heterostructure formation. In the rst sub-section, the structural phase transition from trigonal prismatic 2H to octahedral 1T or distorted 1T 0 is addressed, with potential applications in wireless energy harvesting, solar cells, electrocatalysts, Li-ion batteries, and memory and neuromorphic devices. In the second sub-section, the construction of vertical van der Waals and lateral heterostructures is described, with several applications, such as photocatalysts, photodetectors, solar cells, magnetic applications, and sensors.

Phase transition
TMDs exist in a wide range of crystalline phases, from trigonal prismatic 2H to octahedral 1T or distorted 1T. 199 The stability and the free energy between the phases differ, depending on the material. For example, 2H-MoS 2 is more stable than 1T 0 -MoS 2 under ambient conditions and the free energy difference between the two phases is very large (DE > 0.8 eV). In contrast, MoTe 2 exhibits a small free energy difference (DE < 50 meV) between the 2H and 1T 0 phases. 200,201 Therefore, techniques for phase transition are unique for each material. Metallic phase accelerates the electron transport to obtain low contact resistance of FETs and enhances their performance in electrocatalysis, supercapacitors, and batteries.
Li intercalation effectively reduces the energy barrier of the phase transition from 2H-MoS 2 to 1T-MoS 2 and stabilizes the metallic phase. 202 n-Butyllithium in hexane was employed for Li intercalation in MoS 2 and it induced a phase transition from 2H to 1T. This was followed by sonication of LiMoS 2 to exfoliate 1T MoS 2 ($1-3 nm lm) in water (Fig. 12a). 203 The 1T-MoS 2 thin lms, prepared by drop-casting on a DVD disc, were reverted to the 2H phase by IR laser irradiation, as conrmed by XRD (Fig. 12b). The characteristic (002) peaks for 2H-MoS 2 disappeared aer 60 min, and new diffraction peaks (indicated by stars) appeared, which correspond to the 1T phase. Furthermore, electrochemical lithiation has been proposed to replace the ammable and dangerous n-butyllithium. Li foil and MoS 2 were used as anodes and cathodes, respectively. 204,205 Alternatively, alkali metals (lithium, potassium, or sodium) and naphthalene can be used for phase transitions. [206][207][208][209] A low free energy difference (<50 meV) between the 2H and 1T 0 phases in MoTe 2 enables a reversible phase transition. For example, electrostatic doping with an ionic liquid (N,N-diethyl-N-(2-methoxyethyl)-N-methylammoniumbis(triuoromethyl sulphonyl-imide) and DEME-TFSI) gating in a FET triggers a reversible phase transition (Fig. 12c). 210,211 The Raman spectra of MoTe 2 exhibit the gradual disappearance of the A 0 1 mode (171.5 cm À1 ) and E 0 mode (236 cm À1 ) of 2H MoTe 2 with increasing gate voltage from 0 V to 4.4 V, in addition to the appearance of the A g mode (165.5 cm À1 ) of 1T 0 MoTe 2 (Fig. 12d). A fully reversible phase transition of MoTe 2 can be achieved by increasing or decreasing the gate voltage. Similarly, the electrochemical phase transition of MoTe 2 (from a monolayer to a 73 nm thick sample) was achieved using ionic liquid gating at room temperature in air. 212 The phase transition of T-TaS 2 (from T to H) and NbSe 2 (from 2H to 1T) was achieved by applying a high bias voltage under a scanning tunneling microscope (STM) tip at low temperatures. 213,214 The high density of electron doping from the 2D electrode [Ca 2 N] + $e À drives the phase transition from the 2H to the 1T 0 phase in a long MoTe 2 (up to $100 nm) sample. 215 In addition, the phase diagram of MoTe 2 indicates that the phase stability of 2H and 1T 0 strongly depends on temperature and Te deciency (Fig. 12e). Only the 1T 0 phase was obtained by rapid cooling at 900 C during the ux synthesis process, whereas the 2H phase was obtained by slow cooling to room temperature, as conrmed by XRD patterns (Fig. 12f and g). 216 On the other hand, a phase transition from 1T 0 to 2H MoTe 2 has been demonstrated with thermal annealing at 650 C under a Te rich atmosphere. 217 Recently, single-crystal multilayer 2H MoTe 2 lms were successfully synthesized via a phase transition from a polycrystalline multilayer 1T 0 MoTe 2 lm initiated from single-crystal 2H MoTe 2 seeds. 218 The phase transitions of other TMDs, such as TaS 2 , PtSe 2 , and PdSe 2 , were also demonstrated with thermal annealing processes. [219][220][221] A homojunction between the 2H and 1T 0 phases in the MoTe 2 FET was realized by a laser-driven phase transition (Fig. 12h). The metallic 1T 0 phase was induced by selective laser irradiation of the 2H region to introduce Te vacancies (Fig. 12i). This homojunction in MoTe 2 FET devices signicantly reduces the contact resistance by forming an ohmic contact between the 1T 0 and 2H regions. 222 Similarly, the metastable 1T/1T phase of MoS 2 is changed to a stable 2H phase via continuous-wave laser and femtosecond pulsed laser radiation. 223 The introduction of strain is another strategy for inducing a phase transition. The tensile strain lowered the barrier height for the phase transition from 2H-MoTe 2 to 1T 0 -MoTe 2 (Fig. 12j). The temperature is further required to overcome the energy barrier for the phase transition under tensile strain (Fig. 12k). Experimentally, the reversible phase transition between 2H and 1T 0 of MoTe 2 was conrmed using an atomic force microscope (AFM) tip. 224 Supercritical CO 2 treatment can also induce a phase transition in TMDs via the formation of local strain in MoS 2 . 225,226 A chalcogen vacancy is also used to induce local strain, resulting in the spatial phase transition of MoS 2 and PdSe 2 . 227,228 Phase transitions in TMDs can be achieved by alkali metal intercalation, electrostatic doping, electron transfer, thermal treatment, external irradiation, and strain. A summary of the phase transitions of the TMDs is listed in Table 2. Reversible phase transition between 2H and 1T (1T 0 ) can be induced by Li intercalation/annealing as well as ionic liquid gating. However, Li intercalation, plasma treatment, and supercritical CO 2 can't reach 100% yield of metallic phase TMDs.
Application. The phase-transition of TMDs can be used in various applications, including wireless energy harvesting, solar cells, Li-ion batteries, resistive memory devices, and electrocatalytic applications.
For wireless energy harvesting from the Wi-Fi band (2.45 and 5.9 GHz), a rectier device is fabricated using MoS 2 on a exible Kapton substrate. Fig. 13a illustrates a lateral MoS 2 semiconducting-metallic (2H-1T/1T 0 ) Schottky diode with palladium and gold layers forming Schottky and ohmic contacts, respectively. Nonlinear I-V characteristics were observed for a given input radio frequency (RF) power owing to the presence of a Schottky junction between Pd and 2H-MoS 2 , which is attributed to the rectication behavior. For an input RF power of 5 mW, the device exhibited a cutoff frequency of 10 GHz and an output voltage of 3.5 V, because of the rectied voltage increase (V out ) (Fig. 13b). Hence, a high cut-off frequency is sufficient to cover both Wi-Fi bands. 229 A exible Wi-Fi band antenna integrated with a MoS 2 phase junction (2H-1T/1T 0 ) Schottky diode can achieve wireless energy harvesting of electromagnetic radiation, which is efficiently used for self-powered systems.
A perovskite solar cell was fabricated with 1T-MoS 2 as the hole transport layer (HTL) in the device architecture of FTO/c-TiO 2 /SnO 2 QD/Cs 0.1 FAPbI 3(0.81) MAPbBr 3(0.09) /MoS 2 /PTAA/Au (Fig. 13c). 1T-MoS 2 reduced the mismatch between the energyband alignment and trap density of perovskite, which increased the carrier concentration and improved the ll factor. The power conversion efficiency (PCE) of 1T-MoS 2 increased from 15.05% to 18.54% (Fig. 13d). 230 In other work, a singlelayer MoS 2x Se 2(1Àx) nanosheet with the 1T phase (z66%) was synthesized by electrochemical Li intercalation and exfoliation. The metallic 1T MoS 2x Se 2(1Àx) phase facilitates electron transport to the counter electrode surface in dye-sensitized solar cells (DSSCs). 231 A higher PCE of 8.94% with 1T-MoS 2 @HCS (hollow carbon sphere) was attained compared to those of 2H-MoS2@HCS (8.16%) and Pt (8.87%). 232 Metallic-phase TMDs are promising electrodes for metal-ion batteries. Electrochemical cycling for the phase transition from bulk 2H-MoS 2 to 1T 0 -Li x MoS 2 via Li intercalation/ deintercalation induced the formation of reformable nanodomains (Fig. 13e). Li 1.0 MoS 2 showed a high specic capacity of 636 mA h g À1 at 1 A g À1 with a capacity retention of 80% and coulombic efficiency of 100%, whereas the capacity of bulk MoS 2 collapsed very sharply (Fig. 13f). The domain boundaries between nanodomains facilitate the mass transport of Li + ions and charge/discharge reaction: S + 2Li + + 2e À 4 Li 2 S. 233 Another example is the TiO-1T-MoS 2 nanoower composite for Na-ion batteries, which exhibits a high reversible capacity, attributed to the well-distributed conductive TiO with 1T-MoS 2 . This improves the electrical conductivity and stability. 234,235 In Li-S batteries, metallic 1T-MoS 2 nanodots suppressed the diffusion of polysuldes and accelerated redox kinetics, which was conrmed by in situ XRD and EIS characterization. 236 In supercapacitors, metallic 1T-MoS 2 displays an intrinsic capacitance of 14.9 mF cm À2 , which is 10-fold higher than that of 2H-MoS 2 in an aqueous electrolyte (1 M NaF). 237 Metallic 1T-MoS 2 nanosheets contributed to the efficient absorption/desorption of various aqueous electrolyte ions (H + , Na + , K + , and Li + ), which resulted in a high capacitance ($400 to 700 F cm À3 ). 238 1T-WS 2 /GO shows high performance in supercapacitors owing to the fast reversible reaction of W with proton insertion. 239 A resistive random access memory (RRAM) device was fabricated by sandwiching TMD materials between top (Ti/Ni) and bottom (Ti/Au) electrodes on a SiO 2 isolation layer, which ensured vertical transport (Fig. 13g). 240 The switching behavior of MoTe 2 was attributed to the formation of a conductive lament by the gradual phase transition from a 2H phase to a distorted transient (2H d ; intermediate state) and T d conductive orthorhombic (1T) phase with an applied electrical eld. The I-V curves of a MoTe 2 device ($24 nm thick) demonstrate a resistive switching behavior aer forming the conductive lament over 2.3 V (Fig. 13h). The device exhibited highly reproducible resistive switching within 10 ns between a low and a high resistive state. The thickness of MoTe 2 varied with the set voltage of the device. In the case of the Mo 1Àx W x Te 2 alloy, the set voltage can be reduced by increasing the concentration of W, as a result of the reduction in the energy barrier for the phase transition. The Au/Li x MoS 2 /Au device shows typical memristive behavior due to the reversible phase transition between 2H and 1T 0 -MoS 2 lms, which is controlled by the redistribution of Li + ions under an electric eld. 241 Phase transitions are also useful for electrocatalytic applications. The basal planes of 2H-MoS 2 and 2H-WS 2 are effectively activated for the HER via a phase transition to the 1T phase. [242][243][244] An electrochemical microcell was used to study the HER mechanism (Fig. 13i). 245 The HER activity of 1T 0 (EM-1) and a mixture of 1T 0 -2H (EM-2) and 2H (EM-3) phases was tested in an acidic medium. The LSV curves of EM-1 exhibited the highest HER activity, with a low onset potential of 200 mV and a high current density of 607 mA cm À2 at 400 mV compared to other samples (Fig. 13j). Another study demonstrated that the grain boundary between 2H and 1T phases acted as an active site for the HER. 246 Furthermore, mesoporous 1T-MoS 2 nanosheets with many edge sites and S vacancies provided superior HER activity owing to their high conductivity and abundant catalytically active sites. 247 1T-phase ReS 2x Se 2(1Àx) nanodots outperformed ReS 2 and ReSe 2 in terms of the HER performance. 248 Superior HER performance was also observed for molybdenum dichalcogenides (MoSe 2 ) over their tungsten counterparts (WS 2 and WSe 2 ) aer BuLi exfoliation. 249 In photocatalytic H 2 evolution, the lateral 1T@2H-MoS 2 heterostructure demonstrates an improved photocatalytic efficiency because the 1T phase serves as an electron acceptor and transporter to suppress a charge recombination process. 225 During the HER, the 1T phase in the hybrid TiO 2 @1T-MoS 2 is irreversibly converted into a more active 1T 0 phase, providing more active sites and improving the HER activities. 250

Heterostructures
Heterostructures are essential elements in the modern semiconductor industry and play a crucial role in high-speed electronics and optoelectronic devices. [251][252][253][254][255][256] This wide range of applications stem from the tunable band alignment which enables electron and hole transfer across the heterojunction. TMD-based heterostructures consist of vertical van der Waals and lateral heterostructures. [257][258][259][260][261] The former is constructed by the layer-by-layer stacking of 2D TMD materials, and the latter is formed by the lateral growth of another TMD at the edge of an initial TMD. The construction of a vertical heterostructure by applying the "pick-up" and "drop-down" methods with the mechanical exfoliation approach has been described in another study. [262][263][264][265][266][267] The direct growth of heterostructures on pristine TMDs by employing CVD is focused on in this review.
Vertical heterostructures. To grow vertical heterostructures directly, a nucleation site is required to initiate an overlayer growth on top of a pristine TMD. With the aid of focused laser irradiation combined with raster scanning, a periodic array of defects on pristine WSe 2 is created to serve as nucleation sites (Fig. 14a). 268,269 Direct laser patterning enables the creation of local defects at specic sites without contaminating other areas of the underlying WSe 2 . The pre-patterned WSe 2 was placed in a separate furnace to synthesize an overlayer of metallic VSe 2 . The growth temperature of VSe 2 was 600 C (much lower than that of WSe 2 at 850 C), to prevent thermal damage to WSe 2 . The metallic VSe 2 /WSe 2 heterostructure signicantly reduced the contact resistance between electrodes and WSe 2 by forming an atomically clean vdW interface, resulting in a high on/off ratio of 10 7 and a high "on" current (Fig. 14b). These studies demonstrated that synthetic VSe 2 /WSe 2 vdW contacts offer considerable advantages over typical lithographically developed electrodes with low contact resistance. 268 Furthermore, unstable edges or vertices of the TMD akes can be utilized as nucleation seed sites. As an example, epitaxial growth of the SnS 2 overlayer from a vertex of a triangular WSe 2 via a two-step CVD process ake was observed in the optical image, forming vdW WSe 2 /SnS 2 vertical bilayer heterostructures ( Fig. 14c and d). 270 A type-III heterojunction between WSe 2 and SnS 2 was established (Fig. 14e). Therefore, photo-excited electrons and holes in WSe 2 prefer to transfer to low-energy states in SnS 2 , rather than forming excitons in WSe 2 , resulting in signicant PL quenching of WSe 2 .
Lateral heterostructures. Two-step lateral heterostructure growth of pre-synthesized VS 2 nanosheets, followed by stitching with MoS 2 monolayers at its edges, is shown in Fig. 14f and g. 271 To uncover the nature of the proposed epitaxial growth behavior of lateral VS 2 /MoS 2 stitching, second harmonic generation (SHG) imaging was employed. The orientation distribution of the surrounding MoS 2 domains reveals the polycrystalline nature of MoS 2 domains (Fig. 14h (top)). Nucleation arises from the vertices of VS 2 , instead of lateral epitaxial growth from the VS 2 edges (Fig. 14h (bottom)). In FET devices with lateral heterostructures, I DS with VS 2 contact is six times higher than that of the counterpart with the Ni contact (for the same V G and V DS ), implying that this approach is a promising way to reduce contact resistance (Fig. 14i).
Overall, this section consolidates recent advances in postprocessing techniques for the synthesis of lateral and vertical heterostructures. By considering the respective thermal decomposition temperatures of TMDs, a vast number of vdW heterostructures can be synthesized. A summary of the growth of these heterostructures is given in Table 3. vdW heterostructures grown in a one-step process using mixed precursors have been reported; however, there is a high probability of forming alloys or mixed heterostructures with this method, degrading the unique physical and chemical properties of heterostructures.
Application. Heterostructures are desirable in a broad range of elds, including eld-effect transistors, biosensors, lightemitting diodes, photodetectors, photovoltaic devices, and energy storage.
MoS 2 /WS 2 and WS 2 /MoS 2 vertical heterostructures (type-II heterojunctions) were explored as photocatalysts for hydrogen evolution (Fig. 15a). 272 For MoS 2 /WS 2 /Au, the photoexcited electrons in WS 2 were injected into the conduction band of MoS 2 via stepwise band alignment and contributed to the reduction of H + to evolve H 2 , whereas the holes in the monolayer WS 2 were neutralized by electrons from the electrode. Such effective separation of the photo-excited electron-hole pairs in the MoS 2 /WS 2 stacks greatly promoted H 2 evolution, leading to the highest H 2 evolution content aer 6 h (Fig. 15b).
A vertically stacked vdW GaSe/MoSe 2 heterostructure was constructed to create a p-n junction for photodetector and solarcell applications (Fig. 15c). 273 To test the photovoltaic response of vdW GaSe/MoSe 2 , a FET was fabricated. The I-V transfer characteristics reveal diode behavior, which is ascribed to p-type GaSe and n-type MoSe 2 . Furthermore, the photovoltaic characteristics of the heterostructure under white light illumination were investigated. While negligible photoresponse (black solid curve) is observed in the dark, the output current (red solid curve) under light shows an open-circuit voltage (V oc ) of $0.57 V and a short circuit current density (J sc ) of $0.35 mA cm À2 at V bg ¼ 0 V (Fig. 15d). Ultimately, the resulting solar energy conversion performances, such as photo-to-electron conversion efficiency, ll factor, and photoresponsivity, are estimated to be 0.12%, 0.38 and 5.5 mA W À1 , respectively, at The poor stability of TMDs under ambient conditions is oen a bottleneck for scalable device applications. Recent studies have shown that 2D TMD heterostructures can signicantly improve their air stability through interlayer coupling. 2D CrSe 2 nanosheets, with varying thicknesses (down to a monolayer), were grown on a dangling bond-free WSe 2 monolayer via a two-step CVD process. 274 Atomic force microscopy images of the heterostructures revealed that monolayer CrSe 2 exhibited an outstanding air stability for up to 45 days (Fig. 15e), with no apparent change in the surface roughness or magnetic properties. Theoretical calculations suggested that charge transfer for the WSe 2 substrate and interlayer coupling within CrSe 2 play a critical role in the magnetic order of few-layered CrSe 2 nanosheets. Notably, magnetotransport measurements revealed that the layer thickness of CrSe 2 exhibits differential magnetic properties. A thickness of up to three layers exhibits weak magnetic characteristics, whereas an increase in ferromagnetic (FM) properties was observed for four layers. 274 Fig. 15f shows that the remnant anomalous Hall resistance (R r AHE ) of the layer thickness of CrSe 2 decreases with increasing temperature and vanishes at around the Curie temperature (T C ). The T C values of the 7 and 13 L devices were found to be signicantly higher than those of the 4 L device, indicating that the FM properties increased signicantly with layer thickness.
Finally, the interlayer coupling of post-deposited Bi 2 Se 3 on MoS 2 heterostructures can be modulated by regulating the presence of oxygen with controlled thermal energy. 275 Fig. 15g shows an AFM image of a Bi 2 Se 3 /MoS 2 heterostructure with a height of 3.5 nm. While the PL intensity is signicantly quenched under a N 2 atmosphere, it increases in air (Fig. 15h), indicating that intercalated oxygen interrupts the interlayer coupling between Bi 2 Se 3 and MoS 2 . This characteristic can be applied to gas sensors.

Summary and prospects
We reviewed the recent developments and state-of-the-art atomic and structural modications of TMDs with their unique physical/chemical properties. High performance device applications were discussed for post-treated TMDs, such as electronics, catalysis, energy storage, wearable biosensors, piezoelectricity, CoVID-19 sensors, TENGs, exible PNGs,  superconductor devices, wireless energy harvesting, solar cells, and memory and neuromorphic devices. The outlook for each topic is as follows.

Selectivity and scalability of vacancy generation
Expansion from a laboratory to an industrial scale is critical. Therefore, homogeneous vacancy generation by using a costeffective process over a large area of TMDs should be considered. During post-treatment of TMDs, unintended defects such as the coexistence of metal and chalcogen vacancies, line defects and degradation can be generated under high power or temperature. Selective vacancy engineering plays a critical role in the research of the effect of chalcogen vacancies on the properties of TMDs. More research is needed to improve selective vacancy generation (S V or TM V ).

Unclear catalytic mechanism by spontaneous oxygen passivation at vacancy sites
Several studies have been conducted to enhance the electrocatalytic performance owing to chalcogen vacancies in TMDs. However, STEM analysis of chalcogen vacancies has revealed that oxygen passivation at the vacancies is unavoidable due to spontaneous incorporation of oxygen. The actual active sites for catalytic reactions are still ambiguous. DFT calculations have been successfully used to analyze the Gibbs free energy of the HER and adsorption of metal ions on the vacancies. Most structural models are monolayers, whereas synthesized energy conversion and storage materials are multilayers. Therefore, a comparison of experimental results with theoretical ones shows apparent differences. A precise model that is consistent with multilayer materials is required. Moreover, in situ/operando characterization and advanced techniques are needed to understand the intermediate reactions and electrochemical reactions on vacancies for more diverse applications.

Dopant's homogeneity and its position
Studies on achieving substitutional impurity doping in TMDs have progressed considerably over the past year. Several innovative techniques for incorporating dopants have been studied; however, a few challenges remain. Efforts to modulate and control the homogenous distribution of dopants are still daunting. As the homogeneity of dopants varies from the basal planes to the edges, the physical and chemical properties of the doped TMD tend not to be uniform. Therefore, a critical look is required during the generation of defects to ensure the uniformity of the defect distribution, as they will serve as nucleation sites for dopants. Furthermore, the control of the dopant position is very important for some applications, such as single-photon emission and diluted magnetic 2D semiconductors, but it has not yet been resolved.

Broader research for Janus 2D materials
The progress is still in its infancy. Although several theoretical predictions have been performed for numerous Janus structures, only MoSSe and WSSe have been experimentally investigated. Furthermore, the uniqueness of the Janus structures should trigger new breakthrough applications in the eld of science. Therefore, efforts should be made to improve the fabrication skills and applications of Janus 2D TMDs in the future.

Hybrid and asymmetric functionalization
In covalent functionalization approaches, the high coverage and uniform distribution of functional molecules on TMDs are limited, and hence, they deserve special attention. A special focus on the functionalization of Janus or heterostructure TMDs with organic molecules is needed, which will eventually offer some interesting physical and chemical properties based on the functional molecules. In addition, hybrid and asymmetric/ Janus functionalization of TMDs have recently provided enriched electronic and optical properties of pristine TMDs. 276 Therefore, this will facilitate the fabrication of next-generation multi-storage memory devices.

Optimization of functionalization techniques
Conventional techniques (drop-casting, dip-coating or soaking, spin-coating, and thermal evaporation) are predominantly used for organic functionalization of TMDs. [277][278][279][280] These techniques consist of some disadvantages such as thick coatings, nonuniformity, aggregation, solvent evaporation dynamics, uncontrollable surface dewetting, and material accumulation at drop edges (coffee ring effect). Spin-coating has many adjustable parameters (concentration of organic molecules, boiling point of the solvent, and rotation speed) to optimize. [281][282][283] In the future, direct printing can be applied to overcome many issues found in other techniques. Currently, functional organic molecules are dispersed in solvents to form the desired functional ink. As a result, the integration of organic molecule printing techniques with TMDs can be effectively used for largescale fabrication at low-cost, lightweight wearable biosensors, exible photodetectors, micro-energy storage devices, and memory elements. [284][285][286][287] Therefore, it will provide a pathway for developing next-generation electronic technologies.

Unclear repair mechanism
Currently, few methods are employed to repair atomic chalcogen vacancies in TMD materials. The repair mechanism for healing TMDs with organic molecules (i.e. thiol molecules and bis(triuoromethane) sulfonamide (TFSI)) are still unclear. 187 Therefore, more research, using advanced techniques such as STEM and in situ TEM, is required.

Quantum application for oxygen passivation
Furthermore, passivating TMDs with oxygen atoms showed promising applications in electronic devices. Similarly, oxygen passivation offers some interesting quantum phenomena such as elevated electron-phonon interaction, breaking structural symmetry and anisotropy, Rashba spin-orbit interaction, enhanced piezoelectricity, and improved Ising superconductivity. 198 4.9. Stability and scalability of phase transition Phase engineering has been studied to prepare metallic (1T or 1T 0 ) phases from semiconducting (2H) phases. In the eld of energy applications, the most widely studied technique for phase transition is Li intercalation through Li or organolithium reagents, which are highly corrosive and ammable, hindering industrial scalability. Furthermore, long lithiation times at high temperatures result in excess lithium and organic residues. The thermodynamic instability of the synthesized 1T/1T 0 phase is a major drawback that hinders its further application. Thus, the yield and stability of the 1T/1T 0 phase are important challenges that need to be overcome. Finally, it is necessary to develop a different phase-selective route to prepare metallic TMDs with a high phase purity.

Large scale synthesis of vertical heterostructures
The scalable growth of heterostructures has been a challenge, restricting extensive applications. At present, high-quality vdW heterostructures with sizes of only a few micrometers are achievable. The fabrication of large-area 2D heterostructures has been reported by sequential atomic layer deposition, molecular beam epitaxy, and exfoliation transfer techniques. However, clean and sharp interfaces were not obtained using these fabrication methods. Hence, continuous research on the realization of high-quality vdW heterostructures is necessary for the development of novel optoelectronic, electronic, and solarcell devices. CVD is one of the most promising methods for producing high-quality, large-size TMDs. This is because of the ease in designing the growth parameters and revamping homebuilt CVD systems. 288 In the future, further advances are required to improve the synthesis of wafer-scale high-quality vdW heterostructures for manufacturing lines in the semiconducting industry.

Author contributions
B. K., L. A. A., S. M. K. and K. K. K. designed the scope of the review. B. K., Y. S. W., L. A. A. and S. H. C. collected the references, organized the images and wrote the initial manuscript. Y. S. W. and S. H. C. modied and arranged the gures. S. M. K. and K. K. K. supervised and revised the manuscript. All the authors participated in the revision of the manuscript.