Atomically Thin 2D Transition Metal Oxides: Structural Reconstruction, Interaction with Substrates, and Potential Applications

The discovery of graphene has stimulated dramatic research interest on other 2D materials including transition metal oxide (TMO) monolayers in order to realize novel functionalization and applications. Due to reduced bonding coordination and strong surface polarization, the structures of most TMOs in the monolayer limit are very different from their bulk counterparts, as well as their physical and chemical properties. In this brief review, the authors summarize recent research progress on atomically thin TMO layers. The focus is on the structural properties of the TMOs and their interaction with the substrates from the computational point of view. The authors also introduce the potential applications of the TMO 2D materials on supercapacitors, photocatalysts, batteries, and sensors.


Introduction
The discovery of graphene opened a door of a completely new playground of low dimensional materials. [1][2][3] This atomically thin carbon layer with honeycomb lattices, which was realized by exfoliating graphite crystal at the beginning, has attracted tremendous research interest due to its exceptional physical and chemical properties. [4][5][6] In order to pursue novel electronic, optical, or energy applications, extensive efforts have been made to explore other 2D materials such as hBN, [7] transition metal very diverse properties due to their capability of adopting dif ferent binding configurations. [34][35][36][37] Perovskite oxide thin film based heterostructures form another interesting class of oxide materials, which exhibit electronic and magnetic properties that are very different from their parent bulk materials, and thus have attracted remarkable attention recently. [38][39][40] As an example, an unexpected insulatormetal transition has been observed in LaAlO 3 /SrTiO 3 when the thickness of the depos ited LaAlO 3 layer is larger than 4 unit cells. All these spur wide interest to investigate fundamental properties of 2D metal oxides and explore their potential applications in super capacitors, [41] rechargeable batteries, [42,43] photocatalysis, [44,45] electronics, [16,46,47] piezoelectronics, [48][49][50] superconductivity, [51] etc. We note that there are some excellent reviews published in the past ten years. [46,[52][53][54][55][56][57] However, significant advances in this field have been spotted recently as signatured by the successful synthesis of various oxide monolayers using the novel liquid metal-based reaction route in 2017. [31] Herein, we review recent research progress of oxide monolayers, especially on their structural reconstruction, their interaction with substrates, and their potential applications.

Layered TMOs
In nature, many layered transition metal oxides are com posed of negatively charged slabs with alkaline cations (e.g., K + , Rb + , Cs + , etc.) filling the interlayer spacing. As illustrated in Figure 1, these slabs are commonly made up of corner or edgeshared octahedral units of MO 6 (M = Ti, Nb, Mn, W, Ta, Ru, Mo, etc.), [16,52,[58][59][60][61][62][63] which form ionic bonds with the sur rounding alkaline cations. The cation exchange-assisted liquid exfoliation has been developed to reduce these materials into 2D nanosheets (Figure 1g). By treatment with an acid solu tion, the interlayer alkaline cations can be exchanged with H + cations to form hydrated protonic compounds. The interlayer protons can be further replaced with organoammonium ions in an aqueous base solution, tetrabutylammonium hydroxide (TBA + OH − ; (C 4 H 9 ) 4 N + OH − ) due to their Brønsted solid acidity. Such a reaction often introduces a massive volume of water, which leads to a drastic decrease in the interlayer electro static interaction and the interlayer expansion. Subsequent mechanical shaking or sonication treatments can easily exfo liate the expanded compounds into metal oxide nanosheets, i.e., 0.4 , [59] − Nb O 6 17 4 , [60] − TaO 3 , [61] or − Ca Nb O 2 3 10 . [62,64] Some layered transition metal oxides are also found to be bonded by the weak van der Waals force, such as αMoO 3 [17] and V 2 O 5 . [65] Each layer of MoO 3 is similarly composed of edgeshared distorted MoO 6 octahedra (see Figure 1e), while for V 2 O 5 , the lay ered anisotropic structure is formed by linking distorted trigonal bipyramidal polyhedral O atoms enclosing V atoms (Figure 1f). Micromechanical cleavage or liquid exfoliation techniques can be applied to produce the 2D monolayers in these oxides. [17,65,66] In addition to topdown synthesis routes, bottomup syn thesis methods have also been explored to pursue a high yield, high production rate and precise control of layer numbers. Zhao et al. synthesized MnO 2 nanosheets with a large surface area via in situ replacement of carbon atoms on the graphene oxide framework. [67] By deploying a wetchemical onepot synthesis method, Tae et al. could produce lepidocrocite type titanate nanosheets, , on a large scale. [68] As another example, Xu et al. recently utilized the atomic layer deposition technique to grow WO 3 films in a large area with a controllable thickness, [69] whereas Chen et al. fabricated singlecrystal WO 3 nanosheets with a thickness of 4-5 nm and lateral size up to micrometer via laterally oriented attachment of tiny WO 3 nanocrystals. [70] Table 1 tabulates the layered 2D TMOs and associated synthesis methods.

Nonlayered TMOs
For nonlayered transition metal oxides, topdown synthesis methods are not applicable due to the presence of strong Tong Yang is a PhD candidate in the Department of Physics, National University of Singapore. His research focuses on electronic and catalytic properties of 2D materials and surfaces, and their applications in energy harvesting and storage based on density functional theory.
Shi Jie Wang is a senior scientist III at the Institute of Materials Research and Engineering. His current research focuses on materials surface treatment, surface and interfaces engineering, and advanced ceramic thin film growth and coatings.

Ming Yang is a scientist at the Institute of Materials
Research and Engineering, Singapore. His current research focuses on understanding of electronic, magnetic, excitonic, and catalytic reconstruction in materials at nanoscale for advanced technologies.
www.advmatinterfaces.de interlayer chemical bonds. Nevertheless, some bottomup strategies have been proposed and successfully applied to the synthesis of either supported or freestanding nonlayer struc tured 2D TMOs as summarized in Table 2.
Nonlayered 2D TMOs are often obtained by surface oxida tion. For most metals, thin oxide layers can be naturally formed at the metalair interface. [71] Using in situ atomicresolution electron microscopy, Zhou et al. observed oxide growth during the oxidation of Cu surfaces. [72] They found that oxidation occurs via direct growth of Cu 2 O on flat terraces, where Cu adatoms detach from steps and diffuse across the terraces. Subsequently, they reported that the presence of surface steps inhibits oxide film growth and leads to oxide decompo sition, thereby resulting in an oscillatory oxide film growth (Figure 2a,b). [73] Besides, reoxidation of metal oxide surfaces was identified as another way to form 2D TMOs. Tao et al. observed a 2D phase of TiO 2 on the rutile TiO 2 (011) surface when the surface is slightly vacuumreduced and subsequently annealed in a lowpressure O 2 atmosphere. [27] This is ascribed to the reoxidation of titanium interstitials. For those 2D TMOs formed on surfaces, further isolation from substrates is needed for practical applications, which is usually very challenging.
Physical vapor deposition (PVD) has been widely utilized to deposit 2D nonlayered TMOs layers on metal single crystal sur faces. [74][75][76] Up to date, various 2D binary oxides of transition metals (e.g., Fe, W, Co, V, Ni, Mn, Ti, etc.) have been deposited on single crystal metal substrates (e.g., Pt, Pd, Rh, Ag, etc.) and intensively investigated. [77][78][79][80][81][82][83][84][85][86] For instance, iron oxide has been deposited on Pt (111) and identified as a bilayer FeO (111) with the outmost oxygen layer in a close packed structure. [78,79]   − nanosheets using the cation exchange-assisted liquid exfoliation method. g) Reproduced with permission. [64] Copyright 2014, Springer Nature. Laterally oriented attachment WO 3 [70] Hydrothermal method WO 3 [205] www.   [26] Subsequent investigations suggest that two oxygenrich phases among them can be described in terms of MnO (111)like OMnO trilayers, whereas the other two with lower oxygen content are based on a MnO(100)like monolayer structure. [88] These observed multiple MnO x phases might be ascribed to kinetic effects which lead to stabilized metastable metal oxide structures, and the flexibility of transition metals for adopting different binding configura tions. Besides, some complex 2D TMOs can also be obtained by using PVD, such as NiO(6 × 1) structure on Rh(111), [81,82] Ti 2 O 3 kagomé phase on Pt(111) [86] and ( × 7 7)R 19.1° V 3 O 9 phase on Rh (111). [85,89] Beyond 2D binary TMOs, 2D ternary TMOs have recently been epitaxially fabricated via the onsurface solidstate chemical reaction between two 2D binary metal oxides. [74,[90][91][92] Martin et al. reported the synthesis of a 2D copper tungstate (CuWO 4 ) which is composed of three sublayers with stacking OWO/Cu from the interface. [90] The single crystal line Cu (110) surface oxide is first covered with a monolayer of (WO 3 ) 3 and the surface chemical reaction is subsequently initi ated by increasing the surface temperature, giving rise to the formation of 2D CuWO 4 . Analogously, Pomp et al. observed 2D iron tungstate (FeWO 3 ) with honeycomb geometry on a Pt (111) surface. [91] In addition to the choice of metal substrate and the tuning of oxygen partial pressure during the growth, the use of an atomic oxygen source might promote the formation of the crystalline oxide monolayer, as evidenced by the formation of a NiO(001) monolayer on the Ag(001) substrate. [93] Similar to those obtained by surface oxidation, these PVD grown 2D TMOs usually strongly bind to the metal substrates. Therefore, it may be very challenging to peel off and transfer the epitaxial 2D TMOs. Nevertheless, it provides us with an ideal platform to investigate the strongly interacting oxidemetal interfaces, and explore their potential applications.
As an alternative, selfassembly has been proven to be a promising technique for the synthesis of freestanding non layer structured 2D TMOs. Using 0D and 1D nanocrystals as building blocks, Liu et al. selfassembled CuO monocrystal line materials with controlled dimensionality. [37] The 2D CuO nanosheets were formed by tuning the pH value to 8.5. Analo gously, Yao and coworkers reported the synthesis of Eu 2 O 3 nanosheets from the assembly of Eu 2 O 3 nanowires. [94] In addi tion, Sun et al. generalized the selfassembly method for the synthesis of 2D transition metal oxide nanosheets by strate gical and collaborative selfassembly of metal oxide precursor oligomers into lamellar structures with polymer surfactant molecules, before they are condensed, polymerized and crystal lized into 2D metal oxide nanosheets (Figure 2c). [28] After the removal of the surfactant templates, 2D TiO 2 , ZnO, Co 3 [95] Templateassisted synthesis emerges as another effec tive bottomup approach to fabricate nonlayered 2D TMOs, where the growth of specific nanostructures is either confined or directed by the templates. For example, Using CuO as a insets show schematically the different growth stages of the oxide film with respect to the propagation of the surface step of the Cu substrate. Reproduced with permission. [73] Copyright 2014, American Physical Society. c) Schematic drawing of self-assembly of 2D metal oxide nanosheets, where metal oxide precursor oligomers are strategically and collaboratively self-assembled into lamellar structures with polymer surfactant molecules, before they are condensed, polymerized and crystallized into 2D metal oxide nanosheets with atomic thickness. Reproduced with permission. [28] Copyright 2014, Springer Nature. d) Schematic representation of the salt-templated synthesis of 2D transition metal oxides. Reproduced with permission. [97] Copyright 2016, Nature Publishing Group. e,f) Liquid metal-based reaction route to create 2D TMOs at room temperature according to Gibbs free energy of formation. f) Oxides to the right of the red dashed line are expected to dominate the interface. g) A cross-sectional diagram of a liquid metal droplet, with possible crystal structures of thin layers of HfO 2 , Al 2 O 3 and Gd 2 O 3 . Reproduced with permission. [31] Copyright 2017, American Association for the Advancement of Science.
template, Cheng et al. synthesized 2D ferromagnetic αFe 2 O 3 semiconductor nanosheets with only halfunit cell thickness. [96] Recently, this synthetic strategy (Figure 2d) was further con firmed by Gogotsi and coworkers. [97] They reported the growth of 2D hexagonalMoO 3 , MoO 2 , MnO and hexagonalWO 3 with the aid of watersoluble salt crystals as growth templates. They proposed that both the salt crystal geometry and lat tice matching could guide and promote the lateral growth of 2D oxides, while the thickness could be restrained by the raw material supply.
More recently, Zavabeti et al. proposed a liquid metalbased reaction route to create 2D TMOs at room temperature. [31] They showed that a selflimiting interfacial oxide could be formed thermodynamically, when nontoxic eutectic galliumbased alloys are used as solvent and desired metals are coalloyed into the melt, as shown in Figure 2e,f. Taking roomtemperature liquid gallium alloys as the solvent, this route potentially ena bles access to the 2D nanostructures of all lanthanide oxides and a sizable portion of the transition metal and posttransition metal oxides. Moreover, the accessibility to 2D metal oxides can be further extended when eutectic bismuthtin alloy (melting point: 138 °C) is used as the reaction medium at an elevated temperature. This strategy was confirmed by successful synthesis of 2D HfO 2 , Al 2 O 3 and Gd 2 O 3 .
Recently, epitaxial growth of perovskite oxide ultrathin films has been realized by using the pulsed laser deposition (PLD) technique, molecular beam epitaxy (MBE), or metalorganic chemical vapor deposition (MOCVD), in which PLD is the most commonly used method to grow perovskite ABO 3 thin films. [38][39][40][98][99][100] The model example is the growth of LaAlO 3 (001) on SrTiO 3 (001) substrates, where the layerby layer epitaxial growth can be achieved by tuning the deposition parameters and the substrate temperature as in situ moni tored by the related reflection highenergy electron diffraction pattern. Based on this PLD growth process, a large variety of binary or ternary complex oxide films have been grown epitaxi ally with various different crystal structures such as rocksalt, wurtzite, or fluorite structures for binary oxides, and ilmenite, spinel, or perovskite structures for ternary complex oxides. [100] Some other synthetic methods have also been developed for the synthesis of nonlayered 2D TMOs. For example, atomically thin CeO 2 sheets and ultrathin In 2 O 3 porous sheets have been successfully grown via the hydrothermal method. [ [104] or fast heating of presynthesized CoO nanosheets. [105] In addition, Addou et al. reported the forma tion of 2D Y 2 O 3 by evaporating yttrium in the low pressure of oxygen environment. [106]

Theoretical Aspects of 2DTMOs
Firstprinciples calculations based on density functional theory (DFT) emerge as another attractive and effective technique to investigate 2D TMOs. [107] It can be utilized not only to elucidate experimental observations, but also to theoretically predict new 2D TMOs and study the associated physical or chemical proper ties. The two aspects are further boosted by the recent develop ment of advanced algorithms for structure prediction, such as CALYPSO [32] and USPEX. [33] CALYPSO is based on the particle swarm optimization algorithm, whereas USPEX implements the evolutionary algorithm.

Prediction of new 2D TMOs
Theoretical calculations can allow us to explore new 2D TMOs without having to synthesize them first. In particular, the availability of unprecedented computational resources nowa days enables highthroughput calculations to be carried out to explore material space and to screen materials for specialized applications on a large scale, which is more costeffective than the traditional trialanderror method in experiments. [108][109][110] Recently, Mounet et al. computationally analyzed and screened experimental structures from the Inorganic Crystal Structure Database [111] (ICSD) and the Crystallographic Open Database [112] (COD) to search for potential 2D materials. [113] Among the identified exfoliable 2D materials, there are var ious potential 2D TMOs, such as FeO 2 , CoO 2 , PbO, and PtO 2 . More interestingly, the theoretical calculations show that 2D CoO 2 is a ferromagnetic metal and 2D FeO 2 an antiferromag netic metal. Likewise, Zhang et al. investigated more than 60 000 inorganic compounds in the Materials Project [114,115] (MP) Database to search for potential insertiontype electrode materials for sodiumion batteries (SIBs). [116] Based on massive calculations, they identified layered Na(CuO) 2 and Na 3 Co 2 SbO 6 as promising cathode materials for SIBs.
In contrast to the abovementioned topdown approach, the computational bottomup methods have also been employed to explore 2D TMOs. By means of elemental substitution, Lu et al. proposed a metastable phase of 2D TiO 2 which bears the same lattice as MoS 2 . [117] Compared to rutile and anatase phases of TiO 2 , the proposed TiO 2 monolayer has a relatively small bandgap (2.1 eV), which makes it a good photocata lyst candidate. Similarly, Ataca et al. applied the elemental substitution strategy to systematically investigate singlelayer honeycombstructured transition metal oxides in both 2H and 1Tphases. [118] They focused on structural, mechanical, elec tronic as well as magnetic properties of these 2H and 1TTMOs. Rasmussen and Thygesen later expanded the scope of investi gation of 2Hand 1Tphases of 2D TMOs by taking into account more transition metals. [119] Wherein, electronic structures and optical properties of these honeycombstructured 2D TMOs are comprehensively studied. They found that bandgaps of honey combstructured 2D TMOs vary in a wide range (1.64-7.98 eV at the G 0 W 0 level of theory). Among them, 2D TMOs with wide bandgaps like 1THfO 2 (7.98 eV) and 1TGeO 2 (7.07 eV) might be promising highk materials, while those with small band gaps like 2HMoO 2 (2.20 eV) and 1TNiO 2 (2.15 eV) might be potential photocatalysts.
Structure prediction software also plays an important role in exploring new 2D TMOs. Recently, Song et al. employed CALYPSO to predict a planar Al 2 O 3 monolayer. [120] It is found that this Al 2 O 3 monolayer is energetically and thermodynami cally stable up to 1100 K. It has a direct wide bandgap of 5.99 eV www.advmatinterfaces.de and the calculated static dielectric constant is comparable to that of SiO 2 bulk although smaller than that of αAl 2 O 3 bulk. In addition, they investigated the interface between graphene and Al 2 O 3 monolayer, as shown in Figure 4d. It is found that the interlayer interaction between them is dominated by the weak van der Waals forces and the presence of graphene could fur ther stabilize the Al 2 O 3 monolayer. But different from the Y 2 O 3 (111) monolayer (see detailed description in Section 3.2), the Al 2 O 3 monolayer still favors the planar structure energetically in this case.

Understanding Experimental Observations
Due to inherent limitations of experimental techniques, some information about the materials of interest, e.g., detailed atom istic structures, might not always be obtained at ease. Theo retical investigations as a complement to the experiment may facilitate understanding experimental observations. [36,[121][122][123][124] Since Tao et al. reported the synthesis of 2D phase of TiO 2 with a reduced bandgap, [27] a large amount of attention has been paid to resolve its atomistic structure, owing to its potential application as a photocatalyst in the visible light range (Figure 3a-c). Zhou et al. used the USPEX package to predict a Ti 4 O 4 2 × 1 surface model. [125] They claimed that both the simulated STM images and electronic structures are in agreement with those of experimental observations, as shown in Figure 3d,e. Meanwhile, by means of CALYPSO, Xu et al. proposed an anatase (101)like structural model ( Figure 3f). [123] It is found that this anatase (101)like struc ture could explain not only the observed STM images and the electronic bandgap, but also the measured oxidation state of Ti 4+ . Furthermore, they demonstrated that water and formic acid molecules could spontaneously dissociate on the anatase (101)like structure, indicating its high photocatalytic activity (Figure 3g). But so far, it is still under debate on the exact atomistic structure of the novel 2D TiO 2 phase, and further attempts should be carried out to understand its structural and electronic properties.
In 2012, Addou et al. successfully deposited monolayer yttria (Y 2 O 3 ) on platinumsupported graphene. [106] The scan ning tunneling microscopy reveals that Y 2 O 3 mono layer fea tures a hexagonal lattice and the XPS measurements show that graphene is changed from ptype doping on pure plat inum to ntype doping after the deposition of yttria ultrathin films on it (Figure 4a). Via firstprinciples calculations, Song et al. predicted a stable Y 2 O 3 (111) monolayer with a hexagonal lattice. [124] They found that due to the suppression of surface polarization, Y and O atoms of the freestanding Y 2 O 3 layer prefer to lie in the same plane. However, when the Y 2 O 3 (111) Adv. Mater. Interfaces 2019, 6, 1801160  . The bottom panel of (a) shows C 1s XPS spectra before and after yttria growth. The yttria growth shifts the peak position by ≈0.6 eV to higher binding energy. Reproduced with permission. [106] Copyright 2013, Springer Nature. b) Top and side views of the planar Y 2 O 3 (111) monolayer on the (3 × 3 × 1) graphene (top and middle panels), and the orbital projected band structure (bottom panel). The charge density difference (Δρ), which is defined by , is visualized in the side view. c) Top and side views of the atomic structure for buckled Y 2 O 3 (111) monolayer on (2 × 2 × 1) graphene (top and middle panels) and the corresponding band structure (bottom panel). b,c) Reproduced with permission. [124] Copyright 2015, Royal Society of Chemistry. d) Top view of the Al 2 O 3 monolayer on graphene with contour plot charge density difference projected on the graphene plane and the associated band structure. The blue/red color denotes the excess/depleted charge in the top panel. Reproduced with permission. [120] Copyright 2016, Springer Nature. e) Side view of MoS 2 monolayer on the Hf-terminated HfO 2 (111) surface with imposed differential charge density (the isosurface value: 2.0 × 1.0 −3 eV/Å 3 ) (left panel) and the total and projected DOS (right panel). The Fermi energy is shifted to 0 eV. f) The total DOS and partial DOS for the O-terminated HfO 2 (111)/MoS2 monolayer hybrid structure with the Fermi level shifted to 0 eV. g) Contour plot for charge density difference of MoS 2 monolayer on the O-terminated HfO 2 (111) surface without and with a surface oxygen vacancy, and the total and projected DOS of the hybrid structure with a surface oxygen vacancy. The contour plot is visualized through the Mo plane with an iso-value of 0.002 e Å −2 , where the red/blue color denotes increased/ depleted charge density. e-g) Reproduced with permission. [132] Copyright 2016, American Chemical Society. www.advmatinterfaces.de

Interactions between 2D TMOs and Substrates/2D Materials
For fundamental science and practical applications, a sub strate is indispensable for supporting the 2D TMOs, because the interaction between 2D TMOs and substrates might influ ence structural, electronic, and chemical properties of 2D TMOs due to their atomically thin thickness. On the one hand, the interfacial interactions may bring in intriguing phenomena, such as superconductivity [126] and quantum Hall effects. [127] On the other hand, the hybridization of 2D TMOs with substrates or 2D materials may tailor 2D TMOs or compensate shortcom ings of both parent materials for real applications. [50,128,129] Herein, we focus on the interfacial interactions of 2D TMOs with channel materials, cocatalysts, metals and TMOs as well.

Interactions between 2D TMOs and Channel Materials
Integrating highk dielectrics into microelectronic devices is very challenging but highly desirable because highk dielec trics can significantly increase the gate capacitance of transistors and also effectively suppress electron scattering from interfacial charged impurities. [130] 2D TMOs with wide bandgaps might be potential highk dielectrics to replace conventional gate dielec tric SiO 2 . In this regard, theoretical calculations might shed light on the integration of potential highk TMOs into 2D mate rials based electronic devices.
As discussed above, we predicted stable 2D highk dielectric Y 2 O 3 and Al 2 O 3 monolayers and investigated their interac tion with graphene. [120,124] We found that both Y 2 O 3 and Al 2 O 3 weakly bind to graphene via van der Waals forces (see Figure 4b-d). The presence of graphene further stabilizes Y 2 O 3 and Al 2 O 3 . The weak interaction indicates that graphene is an excellent substrate to grow highk dielectric Y 2 O 3 and Al 2 O 3 . It is worth noting that the growth of Y 2 O 3 monolayer on graphene has been demonstrated. [106] Interestingly, for the interfacial interaction between Y 2 O 3 monolayer and graphene, although the calculated adsorption energy is weak, in the typical vdW interaction range, noticeable structural distortion is found in the Y 2 O 3 monolayer with the presence of the graphene. Meanwhile, the electronic structures of graphene are altered significantly. As Figure 4b shows, the graphene monolayer becomes ndoping with a bandgap about 0.2 eV. Detailed DFT calculations further reveal that this sizable bandgap in graphene is mainly due to the interfacial hybridi zation between O p z and C p z orbitals. Besides, electrostatic potential interaction between graphene and Y 2 O 3 monolayer also contributes to the bandgap because the electrostatic poten tial difference is large in graphene AB lattices. This is similar to that of graphene on Ni, [131] where the adsorption is a typical vdW interaction, but the electronic properties are determined by the interfacial orbital hybridization. In contrast, the elec tronic structure of graphene is nearly intact with the presence of Al 2 O 3 monolayer as the orbital hybridization near the Fermi level is weak.
Meanwhile, we comprehensively evaluated the growth kinetics of highk dielectric HfO 2 on MoS 2 using firstprinciples calculations. [132] For the interface between Hfterminated HfO 2 (111) and MoS 2 monolayer, a strong interfacial interac tion is found because of the formation of HfS bonds and pronounced charge transfer, giving rise to inferior electronic properties (Figure 4e). In contrast, the weak van der Waals force is dominant between oxygenterminated HfO 2 (111) and the MoS 2 monolayer. The weak interfacial interaction leads to symmetric band offsets of more than 1 eV (Figure 4f). How ever, the interlayer interaction will be significantly strengthened upon the formation of oxygen vacancies in HfO 2 , which incurs unfavorable electron-hole puddles, larger effective masses and localized midgap states in MoS 2 (Figure 4g). In addition, it is found that with the increase in thickness of MoS 2 , the forma tion of an oxygenterminated HfO 2 thin film on MoS 2 becomes endothermic and the band offsets increase asymmetrically.

Interactions between 2D TMOs and Cocatalysts
Since the discovery of water splitting on TiO 2 , [133] metal oxides have been intensively studied for photocatalytic applications. Among them, 2D semiconducting TMOs appear to be prom ising photocatalysts. However, the photocatalytic performance of 2D TMOs is usually hindered by many factors, such as wide bandgaps, fast electron-hole recombination and aggrega tions. Hybridization with cocatalysts can enhance photocata lytic activities of 2D TMOs in several ways, [45,[134][135][136][137][138][139] where we mainly focus on the interfacial interactions. Taking gC 3 N 4 TiO 2 as an example, Ma et al. compared the photocatalytic oxidation capability of two kinds of gC 3 N 4 TiO 2 nanocomposites which were obtained by a calcination routine and a simple mechan ical mixing, respectively. [140] The former was found to have a superior oxidation capability over the latter. This superior per formance is partially ascribed to the interfacial interaction, where the typeII band alignment between gC 3 N 4 and TiO 2 is achieved and recombination of photogenerated electrons and holes is efficiently suppressed. When both gC 3 N 4 and TiO 2 are reduced to nanosheets, the intimate and larger contact area between them further improves the rate of electron-hole sepa ration. Moreover, as Figure 5a shows, the interfacial interaction also reduces the bandgap of TiO 2 in the surfacetosurface het erojunction from 3.3 to 2.91 eV, extending the light absorption range of TiO 2 nanosheets. [141] Noting that p-n junctions can potentially act to suppress the electron-hole recombination, Ida et al. demonstrated that the induced potential gradient in p-n junctions could affect the photocatalytic activity (Figure 5b). [142] They prepared a p-n junction made up of 2D ptype NiO and 2D ntype calcium niobate (CNO). It was found that the CNO surface potential of the CNO/NiO junction was lower than that of the CNO crys tals in the same crystal face. The induced potential gradient separates reaction sites, i.e., the CNO/NiO junction parts for photooxidation and the nonjunction parts or their edges for photoreduction. As a consequence, the recombination reac tion was effectively suppressed and an enhanced photocatalytic performance was obtained.
To alleviate the impact of poor conductivity of most 2D TMOs, graphene has been considered as an ideal cocatalyst for 2D TMOs due to its high specific surface area and excel lent electrical conductivity. Xiang et al. reported an enhanced www.advmatinterfaces.de photocatalytic H 2 production activity of TiO 2 nanosheets when graphene is used as cocatalysts. [136] It is reported that the conflu ence of the good conductivity of graphene and the appropriate potential of graphene/graphene − (−0.08 V vs SHE, pH = 0), which is lower than the conduction band of TiO 2 (−0.24 V) but higher than the reduction potential of H + /H 2 (0 V), not only promotes the transfer of photogenerated electrons, but also makes the reduction of protons more efficient (Figure 5c). Moreover, Yuan et al. found that the hybridization of SnNb 2 O 6 nanosheets with graphene exhibits improved photocatalytic activity toward degradation of organic dye in water in the vis ible light range, where the presence of graphene facilitates the transfer and separation of photogenerated charge carriers. [134] It is worth noting that hybridization of 2D TMOs with conducting materials also benefits 2D TMOs in other fields, such as super capacitor applications (see detailed description in Section 5.1)

Metal Substrate Effects on Properties of 2D TMOs
When exposed to an oxidizing environment, nearly all metals will develop an ultrathin surface oxide layer. This oxidation pro cess is spontaneous even under an ultrahigh vacuum. [143,144] In turn, the spontaneously formed ultrathin oxide layers might protect metals from undergoing further environmental degra dation or change mechanical properties of nanostructures.
Fatih et al. showed that the formation of oxide layer could enhance the aluminum nanowire ductility by means of reaction molecular dynamics simulations (Figure 6a,b). [145] The oxide layer exhibits a superplastic behavior, which is due to viscous flow as a result of healing of the broken aluminumoxygen bonds by oxygen diffusion. This theoretical result is elaborated by the recent experimental observations of Yang et al. [146] Using the environmental transmission electron microscopy (ETEM) technique, they discovered that aluminum oxide formed on Al indeed deforms like liquid and can match the deforma tion of Al without any cracks/spallation at moderate strain rate (see Figure 6c). On the other hand, when the strain rate is so high that fresh metal surface is exposed, they observed the selfhealing process of aluminum oxide and subsequent seamless coalescence between new oxide islands and new/old oxide nearby without forming any surface grooves (Figure 6d). These discoveries indicate the greatly accelerated glass kinetics of oxide layers on metal surfaces. The triggered selfhealing function also suggests that a bilayer coating containing both aluminum and aluminum oxide may have better mechanical performance in an oxidative environment than a standalone layer of aluminum oxide.

Metal Substrate Effects on the Stabilization and Growth of 2D TMOs
As discussed in Section 2.2, a number of 2D nonlayered TMOs have been deposited on metal single crystal substrates via PVD. The strong interfacial interaction between 2D TMOs and metal substrates plays a significant role in growing and stabilizing 2D TMOs, which can be attributed to multiple aspects, such as   [142] Copyright 2014, American Chemical Society. c) The photocatalytic H 2 -production in the graphene-modified TiO 2 nanosheets system under UV light irradiation. Reproduced with permission. [136] Copyright 2011, Royal Society of Chemistry.
www.advmatinterfaces.de polarity compensation, strain release, interlayer charge transfer, and interlayer bond formation. [78][79][80]83,147] For the epitaxial FeO(111) bilayer on Pt (111), the FeO interlayer distance is observed to be dramatically compressed by about 50% compared to bulk FeO, which reduces the large surface dipole and therefore stabilizes the FeO(111) bilayer. [78] Meanwhile, this polar surface is further stabilized by an image dipole in the platinum substrate. Another example is the growth of a NiO(100)(1 × 1) island on a Ag(100) substrate. [83,147] Inter estingly, it turns out that at submonolayer coverages, the as grown NiO (100)  showing that the Al core region was fractured and that the oxide deformed without any indication of a fracture or necking. b) A comparison of the stress-strain curves for Al and OC-Al NWs deformed under vacuum and in O 2 at a strain rate of 0.05%ps −1 . A,b) Reproduced with permission. [145] Copyright 2014, Springer Nature. c) Sequential TEM images showing the superelongation and self-healing process of aluminum oxide when stretched in 2 × 10 −6 Torr oxygen environment; oxide between the two white triangular marks in the images are the segments being stretched. The green arrow in the first image represents the stretching direction. An aluminum grain boundary (GB) is pointed out by a light-blue arrow in the first image. All scale bars, 5 nm. d) Sequential HRTEM images (left panels) showing the oxidation process at oxygen pressure of 3.6 × 10 −6 Torr. The seamless coalescence of new oxide islands without forming any glass-glass interface or surface grooves is observed. Right panels are processed images of the left panels to indicate different phase distribution: the aluminum lattice planes are shown by black lines; the oxide is shown by red color. All scale bars, 2 nm. c,d)Reproduced with permission. [146] Copyright 2018, American Chemical Society. e) The STM image (left panel) and corresponding schematic model (right panel) of a NiO(100)(1 × 1) island embedded in Ag(100) with a bias voltage of −0.85 V (blue lattices denote Ag atoms and red lattices denote NiO). Reproduced with permission. [83] Copyright 2012, Elsevier B.V. f) The high-resolution STM image (left panel) of Mn oxide (2 × 1) stripes with V S = −1.9 V, I T = 100 pA, and f) corresponding structural models (right panel) of a MnO stripe grown on the border of a silver step on Ag(100). The structure and energetics of the (100) 1 × 1-like pattern (left) against the (2 × 1) pattern (right) are compared. Reproduced with permission. [80] Copyright 2013, American Physical Society.

www.advmatinterfaces.de
in Figure 6e, instead of being attached to the Ag(100) surface directly. [83] Theoretical calculations imply that the embedded NiO island is stabilized by not only the formation of NiOAg bonds at the boundaries, but also the release of the residual interfacial strain through the NiO bilayer patches due to the fact that compared to the 2% compressed embedded NiO mono layer, the NiO(1 × 1) bilayer is slightly expanded (Figure 6e). [147] It is also found that the structure of the deposited 2D TMOs may vary with the substrate structures. [74,80,148] Taking the 2D MnO(2 × 1) structure as an example, it is known that the polar (2 × 1) structure is a metastable phase, and the MnO(1 × 1) structure is the thermodynamically most stable. However, Ober müller et al. found that the MnO(2 × 1) phase is kinetically stable compared with the MnO(1 × 1) phase and the (2 × 1) phase is highly asymmetrically grown on the Ag(100) surface with very high aspect ratios (>20) (Figure 6f). [80] DFT calcula tions show that the (2 × 1) phase is much more stable than the (1 × 1) phase when grown at the Ag step edges, which is mainly ascribed to polarity cancelation of the (2 × 1) structure around the MnOAg boundary line (Figure 6f). [80] Further calculations indicate that an asymmetry in edge diffusion and attachment energies and the difference in adsorption energetics between Mn and Ag adatoms account for the experimentally observed kinetic asymmetrical growth of the MnO(2 × 1) structure along the highsymmetry <110> directions of the Ag(100) surface.

Interaction between Metal Oxides
Various degrees of coupling among charge, spin, and lattice in oxide materials give rise to a rich playground for novel physics and exotic physical and chemical properties. [38][39][40] The combina tion of these complex oxides leads to very different electronic and magnetic properties in the heterostructures compared with their parent bulks, which gives us another degree of freedom to tailor the properties. Taking the heterostructures of LaAlO 3 and SrTiO 3 as an example, both LaAlO 3 and SrTiO 3 bulks are insu lators. [38][39][40] When a polar LaAlO 3 thin film is deposited on the nonpolar SrTiO 3 (001) substrate with a TiO 2 terminated surface, an insulatormetal transition is found when the thickness of the deposited LaAlO 3 layers is larger than 4 unit cells. [38][39][40][149][150][151] It has been suggested that the transition is due to charge transfer from the LaAlO 3 surface to the LaAlO 3 /SrTiO 3 interface in order to compensate the divergent polarization potential in LaAlO 3 films when the grown LaAlO 3 film approaches the crit ical thickness. [38][39][40][149][150][151][152] Along with the interfacial electronic reconstruction, magnetic ordering reconstruction is also found at the interface, although the LaAlO 3 and SrTiO 3 bulks them selves are nonmagnetic. [153][154][155][156][157][158][159][160][161] Further studies suggest that the interplay between charge transfer, lattice distortion, and the formation of defects might be responsible for the interfacial electronic and magnetic recon struction in LaAlO 3 /SrTiO 3 heterostructures. [161][162][163][164][165] Zhou et al. reported that below the critical thickness, it is the lattice dis tortion in LaAlO 3 layers that partially compensates the polar divergence. While above the critical thickness, the formation of surface oxygen vacancies in LaAlO 3 becomes energetically favorable and triggers the charge transfer from the surface into the interfacial TiO 2 sublayers. This charge transfer fully compensates the polarization potential, resulting in a quasi2D electron gas at the interface as well. [162] Similarly, Yu and Zunger also proposed that the formation of surface oxygen vacancies leads to a conducting interface, and the formation of TionAl antisite defects might be responsible for the interfacial magnetic ordering. [163] More recently, Yang et al. further elabo rated the relation between interfacial defects and interfacial conductivity with the interfacial magnetic ordering, in which they suggested that the strong ferromagnetism in the LaAlO 3 / SrTiO 3 heterostructures prepared at high oxygen partial pres sure is due to the coexistence of surface oxygen vacancies and interfacial antisite defects. [161] In contrast, the weak magnetism found in more conductive LaAlO 3 /SrTiO 3 heterostructures pre pared at low oxygen partial pressure is due to the formation of surface oxygen vacancies alone. Apart from the above efforts made to understand the electronic and magnetic reconstruction in the LaAlO 3 /SrTiO 3 heterostructures, an extensive number of attempts have been made to tune the multidegree coupling at the interface, which includes interfacial strain, doping, fer roelectric substrates, or gate voltage. [166][167][168][169][170][171][172][173] All these not only enrich the understanding of the complex oxide interfaces, but also shed light on further functionalization and potential appli cations of these strategic oxide heterostructures. [174][175][176]

Supercapacitors
Supercapacitors are among the candidates for future major energy storage devices with the typical characteristics of a high power density and reliability. As electrode materials for super capacitors up to now metal oxides/hydroxides, carbon materials and conducting polymers have been employed. [177] Among those, transition metal oxides have the highest theoretical specific capacitance. [178] Compared to their bulk counterparts, 2D TMOs in addition benefit from large surface areas for redox reactions. Therefore, nowadays various 2D TMOs have been studied as supercapacitor electrode materials, [41,179] including binary (e.g., RuO 2 , MnO 2 , Co 3 O 4 , MoO 3 , V 2 O 5 , etc.) [104,129,180,181] and ternary (e.g., NiCo 2 O 4 , LiCoO 2 , MnFe 2 O 4 , Ni 3 V 2 O 8 , Co 3 V 2 O 8 , etc.) [182][183][184] 2D TMOs, which will be detailed in the following. RuO 2 is a typical material with a high theoretical capacitance (2000 F g −1 ). [180] Its hydrous counterpart has a higher electronic conductivity than most other oxides. [185] With 2D RuO 2 as the supercapacitor electrode material, a high specific capacitance of 658 F g −1 at 2 mV s −1 has been experimentally measured by Sugimoto et al. [186] It is about ten times larger than that of its bulk counterpart. However, the cost and need for using acidic electrolytes limit the applicability of 2D RuO 2 . In contrast to RuO 2 , MnO 2 and Co 3 O 4 are inexpensive while still featuring a relatively high theoretical capacitance (1370 F g −1 for MnO 2 ; 3560 F g −1 for Co 3 O 4 ). [184,187] Zhao et al. used graphene oxide as a template to synthesize δtype MnO 2 nanosheets and observed a prominent capacitance (≈1017 F g −1 at 3 mV s −1 ). [67] For Co 3 O 4 nanosheets, Jiang et al. reported a remarkable specific capacitance of 1500 F g −1 at 1 A g −1 and a high energy density of 15.4 Wh kg −1 in the asymmetric supercapacitor cell device configuration (Figure 7a). [104] www.advmatinterfaces.de Poor electronic conductivity is a main shortcoming of most TMOs due to their insulating or semiconducting nature. Hybrid ization with electrically conducting nanosheets or other nano materials has been attempted to surmount this shortcoming. Peng et al. developed a planar supercapacitor by hybridizing MnO 2 and graphene, which showed high specific capacitances of 208 F g −1 at 10 A g −1 as well as excellent cycling stability. [128] Nagaraju et al. demonstrated that the 2D heterostructure of V 2 O 5 and reduced graphene oxide exhibits a specific capacitance of 635 F g −1 at 1 A g −1 , which is about 2.5 times higher than that of the 2D V 2 O 5 nanosheets alone. [129] Besides, Liu et al. reported the performance enhancement via hybridization of 2D TMOs with 0D materials. [182] They grew Co 3 V 2 O 8 nanoparticles on the surface of Ni 3 V 2 O 8 nanoflakes. It was found that this nano composite inherits advantages of both Ni 3 V 2 O 8 and Co 3 V 2 O 8 , showing higher specific capacitance than Co 3 V 2 O 8 and superior rate capability and better cycling stability to Ni 3 V 2 O 8 .

Batteries
2D TMOs have also been extensively investigated as poten tial electrode materials for rechargeable lithium and sodium batteries owing to their ample active sites and shortened ion diffusion lengths. [42,43,54,103,188] As mentioned, one of the major drawbacks of 2D TMOs, especially for binary compounds, is the relatively low conductivity. As a consequence, they suffer from a drastic volume change during repeated charging/dis charging processes.  (Figure 7b). [189] They found that these holey 2D TMOs exhibit strong mechanical stability and display minimal structural changes during charging/discharging processes, and hence might be potential anode materials for both lithium and sodium ion storage. With regard to cathode materials, layered TMOs, such as MnO 2 and V 2 O 5 , have been reported as promising candidates. [65,190,191] For example, the exfoliated V 2 O 5 nanosheets showed stable cycling stability (117 mAh g −1 after 200 cycles at 50 C) and large reversible capacity with the first discharge capacity of 292 mAh g −1 at 59 mA g −1 . [65] The 2D structure wherein enables surface lithium storage with extremely short lithium insertion paths during Li insertion.
Among various rechargeable batteries, LiO 2 batteries attract intensive attention because of their high theoretical energy density. [192] 2D TMOs have been reported to act as cathode  . Reproduced with permission. [63] Copyright 2015, Royal Society of Chemistry. d) The structure of pristine monolayer Bi 2 WO 6 and schematic illustration of photocatalytic mechanism over the monolayer Bi 2 WO 6 . Reproduced with permission. [95] Copyright 2015, Springer Nature. e) Gas response to 100 ppm NO 2 for various kinds of WO 3 thin films at 300 °C operating temperature (left panel). PTA x denotes the WO 3 thin films obtained via deposition with the PTA sol of x mL. The right panel of e) is dynamic sensing transients of PTA6 thin film for 10-100 ppm of NO 2 . Reproduced with permission. [205] Copyright 2018, Springer Nature.
www.advmatinterfaces.de materials as well. [63,[193][194][195] RuO 2 , which is metallic and a bi catalyst for both oxygen reduction reaction (ORR) and oxygen evolution reaction (OER), is a promising candidate. Liao et al. discovered that with the RuO 2 nanosheet, the LiO 2 battery is operable under full dischargecharge condition with a high spe cific capacity (≈900 mAh g −1 ) and stable dischargecharge over potentials (0.15/0.59 V) over 50 cycles (Figure 7c). [63] Recently, Zhang et al. investigated MoO 3 nanosheets for LiO 2 batteries. [196] It is found that the MoO 3 nanosheets with high density of oxygen vacancies can significantly lower the overpotential to ≈0.5 V while still maintaining a high cycling stability (over 60 cycles).

Photocatalysis
TMOs have been extensively studied as photocatalyst candi dates in the past several decades due to their high chemical stability, low cost and nontoxicity. In particular, the large sur facetovolume ratio and the ability to obtain sufficient photon flux density for multipleelectron reactions (e.g., water splitting: 2H 2 O + 4e − → 2H 2 + O 2 ) make 2D TMOs superior to 0D, 1D, and 3D TMOs. [44] A typical example of TMO photocatalysts is the TiO 2 nanosheet. Han et al. reported that (001)terminated TiO 2 nanosheet exhibits an excellent photocatalytic activity, which is even better than commercial Degussa P25 TiO 2 . [197] It is noteworthy that apart from widely investigated photocatalytic water splitting, 2D TMOs are emerging as promising photo catalysts for other valueadded chemical reactions in recent years. [70,198] For instance, Chen et al. synthesized ultrathin, singlecrystal WO 3 nanosheets by 2D oriented attachment. [70] They found that the dimensionality reduction alters the bandgap of WO 3 such that the WO 3 nanosheet becomes active in the photocatalytic reduction of CO 2 into CH 4 in the pres ence of water. More recently, Li et al. discovered that on anatase (101)terminated TiO 2 , CO can be produced via photocatalytic oxidation of CH 4 with a remarkably high yield of 81.9%. [198] They proposed that such high selectivity for CO formation stems from the existence of surface Ti 3+ sites.
Despite of all the remarkable performance that have been experimentally observed, the photocatalytic application of 2D TMOs is still hindered by several shortcomings, i.e., wide optical bandgaps, poor electronic conductivity (namely high electron-hole recombination rate), or improper band edge positions for either reduction or oxidation reactions. To date, many strategies have been proposed to overcome these short comings. [57] Metal doping, nonmetal doping and introducing oxygen vacancies are reported for narrowing the wide band gaps of 2D TMOs toward visible light harvesting by inducing gap states. [44,45,102,[199][200][201] As discussed before, hybridizing 2D TMOs with other materials could alleviate the impact of the poor electronic conductivity and meet the harsh requirement on the band edge positions by forming type II band alignment. Additionally, Zhou et al. recently proposed a bottomup route to prepare the monolayer materials with ultrafast charge sepa ration and highly active surface. [95] (Figure 7d). The presence of unsaturated Bi atoms leads to a highly active surface and a reduced bandgap is observed compared to Bi 2 WO 6 nanocrystals. Upon irradiation, they found ultrafast charge transfer and separation in this het erostructurelike Bi 2 WO 6 (photogenerated holes on the surface and photogenerated electrons in the middle layer), leading to an improved photocatalytic performance for visiblelightdriven degradation of Rhodamine B and hydrogen production.

Gas Sensing and Biosensing
Metal oxides are wellknown gas sensing materials, i.e., WO 3 , TiO 2 , ZnO and SnO 2 . [202] For 2D TMOs, the large surface area provides tremendous reactive adsorption sites for gaseous mol ecules. Therefore, a high response, a short recovery time and a high selectivity toward a certain target gas could be expected for 2D TMOs. [203] For example, Shendage et al. fabricated WO 3 nanoplates with a high selectivity toward NO 2 gas and an excel lent sensing response (the ratio of the electrical resistance when WO 3 is exposed to NO 2 gas and in ambient air) at low operating temperature (10 for 5 ppm NO 2 and about 131.75 for 100 ppm NO 2 at 100 °C). [204] The fast responserecovery capability of WO 3 toward NO 2 gas was recently reported by Harale et al., i.e., gas response time of 16 s and recovery time of 260 s at 100 ppm (Figure 7e). [205] As another example, Alsaif et al. found that the αMoO 3 nanoflakes with a thickness of ≈1.4 nm show large responses to H 2 gas. [206] The response and recovery time toward 1% of H 2 gas are 7 and 24 s at 200 °C, respectively.
Similarly, further improvement of sensing response could be made by hybridizing 2D TMOs with other materials. Fu et al. designed a porous 2D netlike SnO 2 /ZnO heterostructures. [207] It was found that the SnO 2 /ZnO could detect 10 ppb H 2 S gas at 100 °C, which is more sensitive than netlike SnO 2 and ZnO homostructures. Besides, Liu et al. investigated the sensing properties of ultrathin ZnO nanosheets toward ethanol. [208] Although the pristine ZnO nanosheets already show a high ethanol sensing response, decorating the nanosheets with CuO nanoparticles leads to further up to twofold enhanced response to ethanol vapor.
In addition to the detection of target gases, some reports suggest that 2D TMOs could be potential biosensing materials, especially 2D αMoO 3 and MnO 2 . Balendhran et al. fabricated a field effect transistor based biosensing platform with 2D αMoO 3 nanoflakes as the conduction channel. [209] They found that this biosensor can respond to bovine serum albumin within 10 s and the detection limit is as small as 250 μg mL −1 , where the excellent response properties are ascribed to the high permittivity of 2D αMoO 3 . It is noteworthy that the detection limit of αMoO 3 toward bovine serum albumin recently has been pushed to 1 pg mL −1 , where heavy ndoping makes α MoO 3 a plasmonic biosensing platform. [210] For 2D MnO 2 , Yuan et al. used MnO 2 nanosheets to construct two biosensors with respective favorable performance for ochratoxin A and cath epsin D, indicating that MnO 2 nanosheets could act as labelfree nanoplatform for homogeneous biosensing. [211] On the other hand, Zhai et al. investigated the mechanism for singlelayer MnO 2 nanosheets suppressing 7hydroxycoumarin. [212] Based on it, they developed a new method with MnO 2 nanosheets for sensing ascorbic acid in vivo with improved efficiency. www.advmatinterfaces.de

Other Potential Applications
Besides the potential applications as discussed above, there are still other fields, where employing 2D TMOs would be benefi cial. For example, without any guarantee for completeness, lay ered MoO 3 has also been extensively studied for electrochromic based systems (e.g., smart windows and optical displays). [213] Metal oxide thin films could protect metals from corro sion. [53,145,146] 2D MnO 2 nanosheets can promote pHtriggered theranostics of cancer. [214] Apart from photocatalysts, CeO 2 nanosheets with 20% pit occupancy exhibit 50% CO conversion at 131 °C, whereas porous Co 3 O 4 nanosheets are found as a promising electrocatalyst toward oxygen evolution reac tion. [101,105] Besides, 2D ZnO nanosheets and their derivatives are reported for efficient direct current power generation. [48][49][50]

Conclusions and Future Perspectives
In this paper, we have summarized recent advances on the state oftheart synthetic and theoretical aspects of 2D TMOs, the interfacial interactions between 2D TMOs and substrates/2D materials, and potential applications. In the past decades, the emergence of multiple synthetic methods (e.g., PVD, self assembly, templateassisted synthesis, and liquid metal-based reaction route) have made more and more 2D TMOs achievable experimentally, especially those 2D TMOs whose bulk coun terparts are not naturally layered. Besides, the use of atomic oxygen source might be beneficial for the formation of crystal line oxide layers. We may expect that the family of 2D TMOs will be further expanded with the development of novel cost effective and scalable synthetic approaches. Following synthetic improvement, 2D TMOs have been studied and demonstrated as candidate materials for a wide range of applications, such as energy storage devices and photocatalysis. Although great progress on 2D TMOs has been made, there are still some remaining challenges.
For layered metal oxides, most share similar structures: negatively charged slabs with intercalated alkaline cations. Therefore, cation exchangeassisted liquid exfoliation is appli cable to produce most layered 2D metal oxides. However, when thinned to atomic thickness, nonlayered 2D metal oxides likely undergo pronounced structural reconstruction due to polarity compensation and other substrateinduced effects. PVD turns out to be a powerful technique for depositing varying amounts of nonlayered 2D oxides, in particular on metal substrates. But the strong oxidemetal interaction makes it challenging to peel off and transfer the asgrown nonlayered 2D oxides. Therefore, various alternative methods have been proposed to directly fabricate freestanding nonlayered 2D oxides. In consequence, unlike the synthesis of layered 2D metal oxides, the growth of nonlayered 2D oxides so far relies on a rather large number of methods. Attention should be paid to the development of cost effective and scalable synthetic techniques for 2D metal oxides, especially the nonlayered ones.
Identification of atomistic structures is more challenging for 2D TMOs that are not naturally layered. Many studies have indicated prominent structural reconstructions of 2D TMOs in comparison to corresponding bulk materials. [22][23][24][25][26]106,215] For instance, Lundgren et al. reported that the 2D Pd 5 O 4 overlayer grown on Pd (111) exhibits no resemblance to bulk oxides of Pd. [22] Tusche et al. observed that when the thickness of ZnO is less than three or four layers, the assynthesized ZnO exhibits a graphenelike structure in contrast to the bulk wurtzite struc ture. [215] A decrease in the number of dangling bonds and the suppression of surface polarization are responsible for the distinct difference between 2D TMOs and their bulk mate rials. [124] Such difference might make it very difficult to figure out the atomistic structures of 2D TMOs. Applications of 2D TMOs will be definitely impeded without the detailed atom istic structures. Therefore, in addition to synthesis of new 2D TMOs, more efforts should be made to identify atomistic structures of 2D TMOs both experimentally and theoretically. We note that recent studies have shown improved efficiency and accuracy on the structural prediction and identification of bulk materials by employing machine learning or optimizing theoretical and experimental data simultaneously. [216][217][218] Simi larly, we can expect that those methods will be extended for 2D materials in the near future, and their combination with the highthroughput calculations might further result in more effective and efficient structural prediction and identification of 2D TMOs.
It will be also interesting to study the exotic physical and chemical properties of the oxide layers in the 2D limit. The dimensionality reduction and strong structural reconstruction might lead to very different properties compared with their parent bulks. For example, the Coulomb screening is much reduced due to the lower dimension in the 2D TMOs, which might give us more strongly bound excitons. [119] Moreover, lat tice vibrations or electron-phonon interactions in 2D TMOs may also be different from that of the bulk oxides. So ferroelec tricity, polaronic effect, or superconductivity might also emerge in the 2D TMOs, as inferred by recent investigations in other 2D materials, i.e., MoS 2 and CrBr 3 monolayers. [219][220][221] With respect to applications, the nature of layered structures makes layered metal oxides promising electrode materials for supercapactiors or batteries. For nonlayered metal oxides, the dimensionality reduction leads to a large number of dangling bonds and chemically active sites, based on which we may expect that nonlayered 2D metal oxides probably have excellent catalytic or sensing applications. However, the potential appli cations of layered and nonlayered 2D metal oxides definitely go beyond those based on such intuitive structural analysis. For example, Peng et al. reported that holey nonlayered metal oxides (e.g., Fe 2 O 3 , Co 3 O 4 , Mn 2 O 3 , etc.) are promising anode materials for batteries, [189] whereas αMoO 3 has been identi fied to be a potential gas or biosensing material. [206,209] In particular for nonlayered 2D metal oxides, their potential appli cations might be dramatically extended by considering their diverse metastable phases. It is worth investigating properties and potential applications of these metastable phases of 2D metal oxides.
As discussed, the interfacial interaction of 2D functional materials with substrates or 2D materials is of importance for applications. [44,106,132,141,174,176,207,208,222,223] Zavabeti et al. synthe sized 2D HfO 2 sheets with a minimum thickness of ≈0.5 nm and found that the HfO 2 nanosheets have a large dielectric constant of ≈39 and a wide bandgap of ≈6 eV. [31] To integrate www.advmatinterfaces.de the highk HfO 2 nanosheet into fieldeffect transistors, the interfacial interaction between 2D HfO 2 and channel mate rials must be taken into account. Additionally, hybridization of 2D TMOs with other materials has been proven to effectively overcome shortcomings and enhance performance of 2D TMOs for a variety of other applications, such as extension of the light absorption range and improvement of charge trans port for photocatalysis and enhancement of sensing response for gas sensors. [44,207,223] Despite the fact that theoretical cal culations are a more costeffective approach to understand the interfacial interactions and efficiently screen undesirable hybridization between 2D TMOs and substrates/2D materials, theoretical studies are severely lacking in this regard. Intensive collaboration between experimental and theoretical researchers should be made and might speed up the realization of practical applications.