Nanoscale metal oxides – 2D materials heterostructures for photoelectrochemical water splitting — a review 2022 Journal of Materials Chemistry A

Photoelectrochemical (PEC) water splitting is an interesting approach to harness clean and renewable solar energy to generate green hydrogen. To this end, metal oxides (MOs) have been investigated as photoelectrodes (photoanode and photocathode) due to their tunable optoelectronic properties and abundance, but achieving an e ﬃ cient overall performance based on single MOs is very challenging due to their narrow visible light absorption, unfavorable band position, low charge mobilities, and limited stability. Heterostructuring MOs with other materials has therefore been proposed in the literature. In this review, we speci ﬁ cally highlight photoelectrodes based on the heterostructure of MOs and an emerging group of 2D materials consisting of mono elemental (Xenes), carbides/nitrides/carbonitrides (MXenes), boron carbon nitride (BCNs), transition metal dichalcogenides (TMDs), metal – organic frameworks (MOFs), and bismuth oxyhalides (BiOX). The bene ﬁ ts of the formation of MOs/2D materials heterostructures are outlined, and the state-of-the-art of MOs/2D materials heterostructures as photoelectrodes with various architectures are extensively discussed. Finally, a critical outlook on fundamental challenges and potential future directions in the development of MOs/2D materials heterostructures is presented. of Physics, Universitas Indonesia. Her research interest currently focuses on 2D materials for photodetection, a Unclear charge transfer mechanism arises from heterostructures that use non-semiconductor 2D materials. works, the materials form other lower dimensionality structures, e.g. , quantum dots (0D). Various onset potentials and reference electrodes are converted into values vs. RHE using Nernst equation.

Amsterdam, the Netherlands. He holds a Bachelor degree from Nanyang Technological University, Singapore, a Master degree in Nanoscience and Nanotechnology from KU Leuven, Belgium and Chalmers University of Technology, Sweden, and a PhD in Materials Science also from Chalmers University of Technology. His works on plasmonic alloy hydrogen sensors won him Graduate Student Awards from European Materials Research Society and the Arne Sjögren's Prize for Most Innovative Thesis in Nanoscience and Nanotechnology. His research interest includes materials science, nanofabrication, nanoplasmonics and optics, and sensor technologies.

Introduction
The global energy demands have been constantly increasing and are predicted to reach 30-50 terawatts by 2050. Fullling this demand with the status-quo condition is worrying since nearly 83% of today's global energy supply is still based on fossil fuels, such as natural gas, petroleum, and coal. 1 The primary byproduct from burning these fuels is CO 2 , the gas that makes up almost 76% of the annual greenhouse gas emissions. 2 Over the past decades, these emissions have signicantly increased and severely threatened the environment and public health, resulting in e.g., diminishing coral reefs due to ocean warming, ooded coastal cities, deadly heatwaves, strong hurricanes, intensied droughts, wildres, and air pollution. [3][4][5][6] These situations call for urgent exploration of alternative fuels that are environmentally sustainable and commercially viable.
Hydrogen is one promising alternative to satisfy current and future energy demands due to its high gravimetric energy density compared to other fuels ( Fig. 1) 7 , making it attractive for various applications. For example, for applications as transport fuels (i.e., in vehicles), hydrogen can be reacted with ambient oxygen using fuel cells to produce electricity with water as the only byproduct. In another context, hydrogen is also an important feedstock in the production of many value-added chemicals (e.g., ammonia, hydrocarbon, pharmaceuticals). However, hydrogen production is not without challenges. Apart from its low volumetric energy density (Fig. 1), hydrogen is not readily available in nature (despite being the most abundant element in the universe) and thus requires additional steps to produce. Nearly 96% of the hydrogen produced worldwide is derived from fossil fuels, 8,9 namely steam methane reforming, naphtha reforming, and coal gasication, while only 4% originates from renewable sources, such as photo/dark fermentation, 10,11 thermochemical pyrolysis, 12 and water electrolysis. 8 In the effort to increase the latter portion, extensive research is done to further advance the sustainable routes to become cost-competitive with the fossil-fuel based ones. To this end, water electrolysis requires a thermodynamic potential of 1.23 V and some kinetic overpotential (0.4-0.6 V). To minimize the dependency on electricity in such a system, new approaches have been proposed. Among several options, photoelectrochemical (PEC) water splitting stands out as an attractive method [13][14][15][16][17] because it relies on an identical mechanism, but with the thermodynamic and kinetic potential being supplied by sunlight, and it can be constructed using inexpensive semiconductors. On average, the sun illuminates the earth's surface with $23 000 TWy annually; 18 a tremendous value compared to the annual worldwide energy demand of $20 TWy in 2019. 19 Together with the maturing development in fuel cells, the large-scale implementation of green hydrogen energy in households, transportations, and industrial manufacturing is expected in the near future. Fatwa Abdi is a group leader and the deputy head of the Institute for Solar Fuels, Helmholtz-Zentrum Berlin. He obtained his Bachelor degree from NTU, Singapore, and Master's degree from NUS, Singapore, and MIT, USA, all in Materials Science. Aer a short stint in the semiconductor industry, he completed his PhD in Chemical Engineering (cum laude) at TU Del, the Netherlands, in 2013. He was the recipient of the Martinus van Marum prize from the Royal Dutch Society of Sciences and Humanities. His research focuses on the development of novel materials and engineering of devices for solar energy conversion applications.
Vivi Fauzia received her Bachelor's and Master's degrees in Physics (1995Physics ( & 1997 from Institut Teknologi Bandung, Indonesia and she completed her Ph.D. from Institute of Microengineering and Nanoelectronics (IMEN) Universiti Kebangsaan Malaysia in 2013. Currently, she is an Assistant Professor at the Department of Physics, Universitas Indonesia. Her research interest currently focuses on 2D materials for photodetection, photothermal water evaporation and clean energy generation applications.
In the quest for high solar-to-hydrogen conversion, decades of research have devised several PEC architectures from single [20][21][22] to tandem photoelectrodes, 13,23 including hybridization with buried photovoltaics system, with the current record of 19.3% demonstrated STH efficiency. 24 Complementarily, the exploration of materials for improved PEC water splitting is an active research endeavor. To this end, several classes of materials have been studied, namely III-V group compounds, 25 organic semiconductors, 26 and metal oxides (MOs). 27 Apart from its low-cost, 28 MOs particularly display tunable optoelectronic properties through facet-, 29 morphology-, 30 and defects engineering, 31 which are favorable for PEC water splitting. MOs are also easy to synthesize through, e.g., solvo-/hydrothermal, 32 electrodeposition, 33 spray pyrolysis, 34 sol-gel, 35 radio frequency sputtering, 36 chemical vapor deposition, 37 pulsed laser deposition, 38 atomic layer deposition, 39 and molecular beam epitaxy. 40 Despite the rising number of PEC water splitting works based on single MOs (Fig. 2), major problems related to limited visible light absorption, unfavorable band position, low charge mobilities, and self-degradation are inherent, resulting in low overall STH efficiency. Additionally, the generated photovoltage in a single MOs is rather limited; even high-grade semiconductors have only achieved photovoltages <70% of the band gap (E g ). 41 Employing a single MO as a photoelectrode in a PEC water splitting has thus been considered unlikely. To address these problems, one can establish a heterostructure device with other materials of complementary properties. Among them, hybridizing MOs with two-dimensional (2D) materials presents several advantages. 2D materials are a group of novel materials that recently emerged in energy storage and conversion applications due to their distinct thickness-dependent optoelectronic properties, 42,43 large surface areas, 44 and enormous experimental charge mobilities. 45,46 Thus, MOs/2D materials heterostructures offer a promising alternative for durable implementation toward affordable and sustainable hydrogen generation. Consequently, in the past decade, the number of published research articles related to MOs-and 2D materialsbased PEC water splitting in the literature is steadily increasing (Fig. 2).
Acknowledging this rising attention and emerging research direction, in this review we summarize the state-of-the-art of PEC water splitting employing nanoscale MOs/2D materials heterostructures. To provide a solid background into the topic, we rst focus on the fundamentals of PEC water splitting, including a discussion on its common performance characterizations. Second, we highlight the materials requirements for PEC water splitting, which emphasize the need to establish a heterostructure device, specically using MOs/2D materials composites. We then in detail recapitulate the latest progress in PEC water splitting using different types of MOs/2D materials heterostructures. Finally, we identify related future challenges and provide our outlook for further development in the eld.

Fundamentals of PEC water splitting
PEC water splitting comprises two chemical surface reactions: (i) water oxidation or oxygen evolution reaction (OER), and (ii) water reduction or hydrogen evolution reaction (HER). PEC water splitting is a thermodynamically uphill reaction with a Gibbs free energy (DG ) of 237.2 kJ mol À1 , 17 with the following net reaction A PEC water splitting device is made of a semiconductor as the working electrode paired with a metal electrode as the counter electrode (Fig. 3a). The semiconductor serves as either a photoanode for OER (Fig. 3b) or a photocathode for HER (Fig. 3c). Due to the respective position of the Fermi level aer equilibrium, an n-type material is generally required as a photoanode, while a p-type as a photocathode. These electrodes are immersed in an electrolyte containing a redox species and are illuminated by sunlight. Mechanistically, PEC water splitting comprises three subsequent steps, namely band gap excitation, charge separation and transfer, and surface reaction.
When a semiconductor is irradiated with photons of energies (hn) higher than the semiconductor E g , electrons are excited from the valence band (VB) to the conduction band (CB), leaving holes in the VB. 16 Conceptually, E g denes the number of charges that can be excited by incoming photons with targeted wavelength range. E g , VB, and CB are optoelectronic properties that depend on the electronic structure of the semiconductors. Downsizing a semiconductor into the nanoscale would alter its band structure due to the quantum connement effect, where decreasing particle size increases the E g and vice versa. [47][48][49] Furthermore, optical scattering is likely to be increased by nanoscale semiconductors; if the nanoscale structure is properly engineered, the optical pathway in the semiconductor and the corresponding absorption can thus be increased.
Next, the photogenerated electrons and holes oppositely migrate to the surface of the photocathode and the photoanode, respectively. 50 In an n-type photoanode, holes are the minority charge carriers, which move to the surface of the photoanode to Fig. 2 The number of published research articles indexed at Scopus related to PEC water splitting based on metal oxides and 2D materials. The search strings used were 'metal oxides photo* water splitting' and '2D materials photo* water splitting.' Accessed on 25 September 2021. perform OER. 51 Conversely, electrons, as the majority charge carriers in an n-type photoanode, migrate to the p-type photocathode through an external wire. 51 Similarly, accumulated electrons in the p-type photocathode ow into the surface of the photocathode to execute HER. The separate movements of the charge are determined by the electric eld in the depletion region and the selectivity at the semiconductor and electrolyte interface, with high mobility signifying a fast charge transport. 50 Since excited electrons have a short lifetime, it is crucial for them to migrate toward the surface of the photoelectrode to further participate in surface reactions. Thus, a swi charge transport is preferable. Unfortunately, several semiconductors suffer from low charge mobility. One technique to tackle low charge mobilities is via nanostructuring, where the shortened diffusion length facilitates a rapid charge transport from the interior to the surface of the photoelectrode, hence minimizing charge recombination rate. An efficient charge transport resulting from establishing a heterostructure can also mitigate self-oxidation/reduction due to excessive charge carriers. 52 However, nanostructuring also poses a negative impact by shrinking the depletion region that is benecial for charge separation, as the thickness of a depletion region is limited by the size of the semiconductor. 49 Finally, the surface reaction OER occurs on the photoanode, while HER does on the photocathode. The reaction rate is largely dependent on the active catalytic surface area. One strategy to increase the reaction rate is therefore by engineering the photoelectrode's architecture into the nanoscale. 53 At the same time, mass transport of ions from the nanoscale surface to the bulk of the electrolyte (and vice versa) may be limited. Diffusion in porous media (e.g., nanostructure) is expected to be slower than that in the bulk electrolyte. Accumulation and/or depletion of ions could therefore be an issue with nanoscale photoelectrodes and result in additional overpotential. On the other hand, it has been reported that the produced gas bubbles are transported away more efficiently with rough electrodes, as shown experimentally in the case of Si microwire photoelectrodes. 54 Nanoscale electrodes are therefore expected to show this effect, which can be further improved by increasing the surface wettability, including the formation of superhydrophilic and/or superaerophobic surface 55,56 and the introduction of material porosity. 55 In a PEC water splitting, when referring to the normal hydrogen electrode (NHE) potential, the values of water reduction potential (E re ) and water oxidation potential (E ox ) are 0 V and 1.23 V, respectively. 17,57 Consequently, a theoretical E g of 1.23 eV in a semiconductor is necessary to thermodynamically split water into oxygen and hydrogen. In addition, kinetic overpotentials are needed to sustain the reactions. The OER overpotential in an n-type semiconductor photoanode is dened as the gap between the quasi-Fermi level of holes (E f,p ) to the E ox , while the HER overpotential in a p-type semiconductor photocathode is referred to as the gap between the quasi-Fermi level of electrons (E f,n ) to the E re (Fig. 3). 58 As a result, to achieve a low additional electrical bias, the OER and HER overpotentials should be kept minimum. Considering the overpotential (0.4-0.6 eV) and thermodynamic energy losses during the interfacial charge transfer due to possible mismatch (0.3-0.4 eV), the actual driving voltage for water splitting reaction is higher than 1.23 V, known as the photovoltage. 58 Therefore, it has been identied that the ideal E g for PEC water splitting is $2.0 eV (equivalent to light absorption at $620 nm). 16,58,59 This nding is also in mutual agreement with the fact that the majority (43%) of the solar radiation energy is within the visible range (380-740 nm). 60 Under Air Mass 1.5 Global (AM 1.5G) illumination at 100 mW cm À2 , a semiconductor with a 2.0 eV E g can reach a maximum photocurrent density of 14.5 mA cm À2 and theoretical solar-to-hydrogen efficiency of 17.9%. 59 As novel materials for PEC water splitting emerge, standardized efficiency measurements and parameters need to be strictly regulated. This guidance ensures that efficiency measurements coming from a broad scientic community is properly catalogued and comparable. Typically, the illumination source is standardized to simulate the intensity of 1 sun (100 mW cm À2 ). Meanwhile, the selection of the electrolyte is rather loosely stipulated due to the disparate stability of various materials within a particular range of electrolyte's pH. For more details, the readers are directed to excellent reviews and book chapters on standardized protocols for PEC measurements. 58,61,62 Efficiency calculations in PEC water splitting Efficiency calculations in a PEC water splitting device provide a quantitative information about the system's capability to split water into oxygen and hydrogen in the presence of sunlight. There are three types of efficiencies based on how they are measured, solar-to-hydrogen (STH) efficiency, applied bias photon-to-current efficiency (ABPE), and incident photon-tocurrent efficiency (IPCE). Ideally, all of these values should be disclosed when reporting a PEC water splitting device (unless they are unattainable due to certain conditions).
The STH efficiency is dened as the ratio of the chemical energy of the produced hydrogen to the solar energy input 58 and serves as a comparative value from one device to another since it reects the true ability of semiconductors when used in a PEC water splitting device at 1 sun or 100 mW cm À2 . An extra care has to be done when calibrating the lamps used as solar simulators in the laboratories to prevent over-or underestimations of the efficiency values. 63,64 The equation to calculate the STH efficiency is where DG , r H 2 , P total , and A* are the Gibbs free energy (237.2 kJ mol À1 ), the production rate of hydrogen (mmol s À1 ), the intensity of the AM 1.5G source (100 mW cm À2 ), and the illuminated photoelectrode area (cm À2 ), respectively. Due to several conditions (e.g., unfavored band position or extensive charge recombination), some semiconductors fail to generate photocurrent without an applied bias. Applying an external bias between two electrodes requires a new type of efficiency since the situation no longer translates the true ability of materials in performing PEC water splitting: the ABPE. 16 This efficiency is calculated using where J ph (mA cm À2 ) is the photocurrent density under an applied bias E b (V). It should be noted that when three-electrode congurations are used (i.e., a reference electrode is included in the measurement), the term "half-cell" should be clearly mentioned.
Last, the ratio of photogenerated charge carriers relative to the measured photocurrent per incident photon as a function of wavelength or energy of monochromatic light is dened as the IPCE. It measures how efficient the semiconductors are in generating charge carriers under a single wavelength, including optical loss factors in transmission and reection. IPCE measurement is typically performed at a particular applied potential and can be determined as IPCE ðlÞ ¼ Total energy of converted electrons Total energy of incident photons where J ph (l) is the photocurrent density at a specic wavelength (mA cm À2 ), e is the elementary charge (1.602 Â 10 À19 C), h is the Planck constant (6.626 Â 10 À34 J), c is the speed of light (3 Â 10 8 ms À1 ), l is the wavelength used in the measurement (nm), and P(l) is the intensity of that specic wavelength (mW cm À2 ). The IPCE can also be depicted as the product of the efficiency of photogeneration of electrons and holes per incident photon ux (h e À /h + ), the efficiency of charge separation through the semiconductor (h separation ), and the efficiency of the semiconductor and electrolyte interfacial charge transfer, (h transfer ), 65 expressed as Here h e À /h + is assumed to be equivalent to the number of photogenerated electrons and holes over the number of absorbed photons and can be estimated by Beer's Law 58 where A is the absorbance with a logarithmic relationship to the transmitted intensity (I) over the incident intensity (I 0 ). Having discussed all the efficiencies above, STH is the most appropriate value in correctly benchmarking the PEC water splitting devices under strict conditions. However, other efficiencies of ABPE, and IPCE can be considered diagnostic efficiencies and are also benecial in the development and characterization of materials for PEC water splitting.

Materials selection for photoelectrochemical water splitting
As a reminder, PEC water splitting can be achieved by accomplishing three steps: charge photogeneration, the separation and transfer of the photogenerated charge, and the participation of electrons and holes to the surface reactions of HER and OER, respectively. Consequently, each of these three steps need to be considered carefully when selecting effective materials for PEC water splitting to maximize the overall performance. Below, we discuss the different material aspects that are taken into account and have been implemented using individual MOs and 2D materials, and their composites.

Metal oxides
MOs is a group of inorganic semiconductors comprising metal atoms that bind to oxygen atoms. They have distinct physicochemical properties, including structure, morphology, and optoelectronics. Furthermore, MOs are robust, easy-tosynthesize, and cost-effective materials, which are attractive for low-cost applications. For instance, some of the most commonly used metal oxides are composed of Fe, Zn and Ti, which were globally produced as much as $3000, 12 and 5 million tonnes per year, respectively, according to the World Mineral Production. 28 Thus, MOs have become one of the most prevalent materials in photogenerated charge separation and conversion applications, such as in hydrogen production via PEC water splitting, 66 CO 2 reduction, 67 pollutant degradation, 68 and anti-microbial properties. 69 Over the past decades, numerous binary MOs have been studied extensively for PEC water splitting. These include ZnO, 70 90,91 have also emerged as viable options. However, nding the best performing MOs for PEC water splitting is not without challenges. Many MOs typically exhibit a wide E g , resulting in a narrow light absorption range that limits the generation of charge carriers. For example, ZnO, TiO 2 , SnO 2 , and NiO have E g in the range of 3.2-3.6 eV. Although the wide E g may provide sufficient photovoltage for PEC water splitting, it limits light absorption within the UV range, which only contributes $5% to solar spectrum. 60 To this end, several MOs also exhibit visible-sensitive E g , such as a- In general, the VB of MOs lies well below E ox (since it is typically dominated by O-2p orbitals), but only a few that display higher CB than E re (e.g., ZnO, NiO, Cu 2 O, and TiO 2 ). This limitation means that many MOs are only sufficient for the OER half reaction, but not for the HER due to unfavorable CB position. The CB of MOs is typically governed by the s-and d-metal orbitals with a strong overlap of electrons clouds, while the VB comprises largely localized p-oxygen orbitals. 92 In general, MOs have low charge mobilities caused by low crystallinity and high defects density, such as oxygen vacancy, 93 that act as impurity scattering centers, resulting in high charge recombination. 94 Besides that, due to the ionic nature of the lattice, charge carriers in MOs create polarons by displacing atoms in the vicinity. This phenomenon lowers the energy of the carriers and traps them within the potential well. 95 Hence, MOs yield low-to-moderate charge mobilities from 0.05 to 300 cm 2 V À1 s À1 depending on the measurement conditions. 83,85,96,97 The redox ability of MOs is also determined by their metal activity, which is dened as the ability of metals to be oxidized by releasing electrons during a redox reaction. Typically, metal activity decreases as the element goes to the bottom le portion of the periodic table. For example, Zn is more reactive compared to Fe. OER kinetics in MOs is also generally sluggish owing to faster holes migration toward the solid/liquid interfacial area compared to holes injection rate into the electrolyte. 16,98 This condition results in holes accumulation and further intensifying charge recombination rate. MOs also have a tendency to be self-reduced and/or self-oxidized by accumulative charge depending on their respective self-reduction and self-oxidation potentials compared to the E re and E ox . 52 On top of that, photogenerated carriers of MOs have a short lifetime of pico-to nanoseconds to participate in OER and HER, 99 that also contributes to high charge recombination. It is also noteworthy that electrolyte pH affects the stability of MOs during PEC measurements. For instance, TiO 2 displays high stability in harsh conditions in both acidic and alkaline environments, 100 while NiO shows high stability only within high pH. 101

2D materials
2D materials have become the center of interest in the past decade due to their exceptional characteristics, 44,102-105 following the rst successful attempt to experimentally synthesize graphene in 2004. 106 This group of materials can be further divided based on their E g : metals or semi-metals ($0 eV), semiconductors (<4 eV), and insulators (>4 eV). Interestingly, a number of 2D materials exhibit thickness-dependent E g modulation. 42,43 In particular, their bulk form has an indirect E g , while its monolayer counterpart yields a slightly wider and direct E g . 107 2D materials also offer large interface areas to hold surface reactions while being extremely thin, thus establishing shorter charge transfer pathways. 50 This is in contrast with 1D materials, where they depend on the anisotropic longitudinal shape to transfer the photogenerated charge efficiently. These wide ranges of physicochemical properties offer high versatility for widespread and practical applications, including energy conversions and energy storages, 108,109 environmental remediations, 110 optoelectronics, 111 and electronics. 112 To this end, there are several subgroups of 2D materials based on their compositions: mono elemental (Xenes), carbides/nitrides/ carbonitrides (MXenes), boron carbon nitrides (BCNs), transition metal dichalcogenides (TMDs), metal-organic frameworks (MOFs), and bismuth oxyhalides (BiOX), as shown in Fig. 4.
Finally, it is noteworthy that 2D materials should not exist above the absolute zero (the lowest limit of the thermodynamic temperature scale, 0 K) based on the theoretical framework. 113 Thus, 2D materials are not perfectly at and exhibit intrinsic ripples with the bending factor of the plane proportional to the temperature. 114 At room temperature, 2D materials may experience lattice distortion and make them appear like crumpled 3D materials. For instance, a typical graphene sheet at room temperature forms wrinkles with a height displacement of 0.7 A. 114 Throughout this review, several authors report 2D materials with the appearance or morphology of non-conventional layered structures (e.g., 0D or 3D). Nevertheless, they are still considered as 2D materials as per the denition, and this is also the classication we use here.
A common feature of 2D materials is that multiple repeating layers are held together by weak van der Waals (vdW) forces to establish bulk crystals. One can exfoliate 2D materials to obtain mono-to-few layers by, for example, mechanical cleavage, 115 strain engineering, 116 and sonication, 117 that lead to higher exposed catalytically active sites. 104 The high surface areas of 2D materials provide substantial photoreactive sites for both light absorption and surface reactions. However, several studies have suggested that the edges are chemically more active than the basal planes. [118][119][120] This is rather unfortunate because the ratio between basal planes to edges is typically high (i.e., basal planes are more exposed than the edges). The difference in catalytic activity between these two sites accommodates carrier delocalization and diffusive transport along the interface areas as an effect of non-hybridized atomic orbitals of the 2D materials. These effects result in higher charge mobilities in the in-plane (edges) in contrast to the out-of-plane (basal planes) surface. 121 However, the activation of basal planes to be catalytically active is also possible by covalent functionalization, 122 defects engineering, 123 hydrogenation, 124 or electron beam direct surface patterning. 125 This activation is advantageous since charge carriers do not have to travel far to the edge sites to participate in the reaction, and thus the probability for charge recombination can be minimized. Furthermore, 2D materials also exhibit enormous experimental charge mobilities, e.g., graphene ($350 000 cm 2 V À1 s À1 ), 45 phosphorene ($4000 cm 2 V À1 s À1 ), 46 and MoS 2 ($200 cm 2 V À1 s À1 ), 126 that is benecial for fast charge transfer across the material.
2D materials, however, also possess signicant drawbacks related to restacking and layer overlapping because of the lack of interfacial interactions. 127,128 Therefore, the experimental surface areas, specically the active sites, are typically lower than the calculated ones. One way to avoid such problem is by forming a heterostructure system with, e.g., MOs. In this case, 2D materials can serve as co-catalysts by enhancing MOs merit for PEC water splitting. To this end, co-catalysts based on 2D materials have shown promising potentials in such heterostructure architecture to improve light absorption and electrons and holes separation, 129,130 to provide additional active sites for surface reactions, 131,132 and to increase durability against photocorrosion. 133 As a result, one major research direction in PEC water splitting is routed toward establishing composite heterostructure devices to mitigate the above problems. On the one hand, 2D materials with narrow E g , such as phosphorene (1.5 eV), MoS 2 (1.3 eV), MoSe 2 (1.4 eV), offer excellent visible light absorption. In addition, various types of heterostructures can be formed owing to diverse CB and VB positions as summarized in Fig. 5. For example, PbS is one emerging 2D material for photocathode due to its narrow E g and advantageous CB position, 134 whereas BiOI can exibly act as photoanode or photocathode owing to its favorable CB and VB positions for HER and OER, respectively. 66,135 Metal oxides/2D materials heterostructures Forming a heterostructure by integrating 2D materials and MOs can be an elegant solution in tackling the weaknesses of MOs for PEC water splitting discussed above (i.e., narrow light absorption, short charge lifetime, high charge recombination rate, and poor structural stability under specic pH). Specically, nanoscale MOs-2D materials heterostructures provide advantageous aspects for PEC water splitting in ve ways. First, although nanoscale MOs provide signicantly higher surface area compared to their bulk counterparts, they tend to agglomerate into larger clusters with decreased surface area. 136,137 Here 2D materials can be used as an immobilization platform where MOs are dispersed throughout their surfaces. Agglomeration can thus be reduced with the increasing surface interactions between MOs and 2D materials. Second, the energetic misalignment (CB, VB positions, Fermi level E F ) between MOs and 2D materials can be engineered to enhance photogenerated charge separation. Hence, the lifetime of charge carriers can be prolonged as electrons and holes are spatially isolated. Third, the lattice mismatch, formed at the interface between MOs and 2D materials, 121,138,139 may introduce interfacial strains that could contribute to E g tuning. 140,141 This provides an alternative route to optimize the light absorption range. Fourth, 2D materials that possess high aqueous stability can be used as a protective layer for unstable MOs, thus preventing the structural decomposition of MOs. 133 Finally, these two materials can play different roles that synergistically work to enhance the performance of a PEC water splitting. Notably, the majority of MOs act as photoactive materials, while 2D materials can serve as photoactive and/or catalytic-active materials. Photoactive 2D materials can absorb light, which in turn increase the amount of electrons and holes pair. Furthermore, they may boost the surface reaction as catalysts by reducing the corresponding activation energy. On the other hand, non-photoactive 2D materials do not interact with light and thus only serve as catalysts. Combining these two materials together into a heterostructure may result in a photoelectrode with good light absorption while maintaining high catalytic activity compared to their isolated counterparts. 76 Another consideration when forming a composite is related to the heterostructure type. To this end, there are four types of heterostructures that are categorized based on their charge transfer mechanism: type I, type II, p-n junction, and Z-scheme (Fig. 6). In a type I heterostructure, the CB Min and VB Max of material 1 (M1) straddle those of material 2 (M2), such that photogenerated electrons and holes move simultaneously from M1 into M2 and potentially recombine in M2 (Fig. 6a). 142 Conversely, the CB and VB of M1 are both higher than those of M2 in type II heterostructures; photogenerated electrons transfer from M1 to M2, while the holes move oppositely from M2 to M1 (Fig. 6b). 142 This spatial separation of electrons and holes typically is considered to be benecial in reducing charge recombination. When the type II heterostructures are composed of p-type and n-type semiconductors, a p-n junction is formed. Electrons in the n-type semiconductor diffuse into the p-type and holes in the p-type diffuse oppositely into the ntype and further to form a built-in electric eld for improving charge separation (Fig. 6c). 143 Finally, a Z-scheme heterostructure has the same band alignment as type II, although the charge transfer mechanism is slightly different because of the different types of interface (i.e., Schottky instead of ohmic). The photogenerated electrons in M1 and holes in M2 are spatially conned and do not move into the other semiconductor, while the photogenerated electrons in M2 recombine with the holes from M1 (Fig. 6d). 143 These conned charge carriers can reduce the steric hindrance that contributes to the detrimental effect on the performance. 144 To this end, one can validate the formation of Z-scheme instead of type II using several methods, as discussed elsewhere. 145 It is worth noting that type II, p-n junction, and Z-scheme heterostructures are favorable for PEC water splitting since all of them provide better charge separation and transfer compared to that in type I. 142,146 Besides the charge transfer mechanism, heterostructures can also be divided based on their respective dimensionality. The main focus here is 2D materials, which exhibit no chemical bonding between each stacking layer and no dangling bonds on their surfaces. 139 Therefore, they offer exibility when combining with MOs without the constraint in lattice mismatch. 121,138 Heterostructures based on 2D materials are termed vdW heterostructures. There are four types of vdW heterostructures: 0D-2D (Fig. 6e), 1D-2D (Fig. 6f), 2D-2D (Fig. 6g), and 3D-2D (Fig. 6h) heterostructures. It is also noteworthy that 2D materials can be synthesized into lower dimensionality forms, e.g., quantum dots (0D) of Ti 3 C 2 T x , 147 h-BN, 82 and MoS 2 . 148 These novel heterostructures of mixed dimensionality have unfolded numerous possibilities in widespread applications, such as light-emitting diodes, 149 photovoltaics, 150 photodetectors, 151 eld-effect transistors, 152 and solar energy conversion. 72,153 There are wide options of 2D materials to establish MOs/2D materials heterostructures: Xenes, MXenes, BCNs, TMDs, MOFs, and BiOX. In the following sections, we will discuss each group of these heterostructures thoroughly, where an exhaustive summary of their performances as photoanode and photocathode is presented in Tables 1 and 2, respectively.

General synthesis strategies
To this end, there are two classications of MOs/2D materials heterostructure fabrication route, namely top-down and bottomup, as summarized in Fig. 7a. Top-down fabrication methods include mechanical exfoliation, 106 liquid phase exfoliation, 205 and intercalation followed by exfoliation. 206 Aerwards, these mono-tofew-layer 2D materials are immersed in a solvent to form a colloidal suspension. This suspension can then uniformly coat MOs using various techniques, including spray-coat, dip-coat, and spin-coat. These methods are compatible with exible substrates and can be done at room temperature. Recently, Ma et al. fabricated Bi 2 MoO 6 /g-C 3 N 4 heterostructure by dip-coating exfoliated g-C 3 N 4 nanosheets colloid onto the surface of Bi 2 MoO 6 nanoarrays for 1 h and annealing at 350 C for 1 h, as shown in Fig. 7b. 86 On the other hand, the bottom-up route stacks atom-by-atom in form of precursors to build material with desired size and dimension. Typically, 2D materials precursor is placed in a reactor alongside MOs on top of a substrate. Bottom-up method includes solvo-/hydrothermal deposition, 189 electrodeposition, 170 chemical bath deposition (CBD), 167 chemical vapor deposition (CVD), 192 and successive ionic layer adsorption and reaction (SILAR). 89 The reaction takes place directly on the surface of the MOs to produce MOs/2D materials heterostructure. This route offers low cost and scalability, although it is time-consuming. Fig. 7c

Metal oxides/Xenes heterostructures
Xenes is a class of elemental monolayer made up of materials in group IVA-VA in the periodic table. Generally, Xenes exhibit high surface areas, high exibility, good stability during surface reactions, and high tunability of physicochemical properties. 207,208 Moreover, the distinctive atomic conguration of each Xenes bestows an adjustable electronic band structure. For instance, group IVA Xenes like graphene, silicene, and germanene possess a semi-metallic characteristic, while VA group Xenes (phosphorene, arsenene, antimonene, and bismuthene) are semiconductors. 207 Except for graphene, Xenes do not form the ideal planar hexagonal structure. The large gap between each atom causes Xenes to preferably exist in a buckling structure, rather than a perfect planar hexagonal lattice, with a particular buckling angle and height. 208 Recently, Tayebi et al. integrated ZnO nanorods with graphene nanosheets, creating a 1D-2D heterostructure with spatial charge separation, as shown in Fig. 8a. 180 While ZnO nanorods show a good light absorption within the UV range (that drastically drops in the visible because of a wide E g of 3.25 eV), graphene exhibits no E g due to the existence of Dirac cone. Integrating graphene onto ZnO nanorods shis the E g to 2.90 eV, i.e., within the visible range. Under 150 mW cm À2 illumination, a photocurrent density of 0.30 mA cm À2 at 1.23 V vs. RHE was demonstrated, nearly three-fold higher than that of bare ZnO (Fig. 8b). Interestingly, the ZnO nanoparticles/ graphene sample only yielded 0.20 mA cm À2 at the identical potential due to a fast charge recombination generally found in nanoparticles. They also reported the graphene's ability to mitigate ZnO photocorrosion under UV illumination. Specically, the addition of graphene signicantly reduced the photocorrosion from 27.0% on ZnO to 0.5% on ZnO nanorods/ graphene heterostructure. Finally, using impedance spectroscopy (see Fig. 8c for the Nyquist plot), lower charge transfer resistance (R ct ) was revealed in ZnO nanorods/graphene   178 In PEC measurements using a 300 W Xe lamp, the heterostructure demonstrated an AM 1.5G photocurrent density of 4.5 mA cm À2 at 1.23 V vs. RHE using KP i as the electrolyte (pH ¼ 7.1). This photocurrent value is fourfold higher than that of BiVO 4 . Additionally, the addition of black phosphorus nanosheets also improved the stability of BiVO 4 ; $99% of the photocurrent was retained over 200 min of chronoamperometry measurement. This enhancement was attributed to the phosphorene layer that provides an exceptional hole transfer channel and minimizes the recombination. Another approach by Dubale et al., who fabricated a synergetic 1D-2D photocathode of Cu 2 O/graphene for solar-tohydrogen production, is depicted in Fig. 8d. 75 The addition of graphene onto the Cu 2 O nanowires drastically improved the AM 1.5G photocurrent density from À2.3 mA cm À2 to À4.8 mA cm À2 (Fig. 8e). However, an excessive amount of graphene was found to reduce the photocurrent due to parasitic absorption. At the highest graphene concentration studied of 3.0 mg mL À1 , the light absorption of Cu 2 O dropped (Fig. 8f), resulting in a photocurrent density of only À1.5 mA cm À2 . It is also important to note that the presence of graphene did not alter the absorption edge of the heterostructure, which was 2.03 eV.  The best performing sample with 1.0 mg mL À1 of graphene demonstrated a higher visible light absorption and an enhanced photogenerated electrons transfer from the CB of Cu 2 O to graphene, resulting in lower charge recombination. 75 The system also achieved 3.3% of ABPE at À0.6 V vs. Pt and enhanced stability with 83.3% of the photocurrent retained over the course of 20 minutes. Moreover, the ability of the heterostructure in producing hydrogen was demonstrated by the direct measurement of up to 0.02 mmol of hydrogen using a gas chromatograph.

Metal oxides/MXenes heterostructures
Recently, a new class of few-atoms-thick materials has emerged, termed MXenes. 209 MXenes can be derived from MAX bulk crystals (M is the transition metal, A is the group A element, and X is carbon and/or nitrogen) by selective etching to remove the A element, leading to the formation of 2D MX phase or MXenes.
MXenes have a hexagonal crystal structure, where M layers are located in between X planes. 210 Furthermore, MXenes can be divided into two groups based on its structural formula of M n+1 X n and M n+1 X n T x , where T x represents functional groups of -OH, -O, or -F. 210 As the consequence of the etching process, the formation of those functional groups is unavoidable, thus generating disparate properties from metal to insulator. MXenes therefore exhibit a high versatility in their mechanical, optoelectronic, and transport properties for widespread applications such as photothermal, catalysis, eld-effect transistors, and energy storage. 210,211 One example of MXenes, Ti 3 C 2 T x , has shown a weak thickness dependency in its resistivity, contrary to other 2D materials. 212 Ti 3 C 2 T x is obtained by selective etching of Al layers from Ti 3 AlC 2 . Fig. 9a shows Ti 3 C 2 T x nanosheets coupled with BiVO 4 nanowalls to form a 2D-2D heterostructure photoanode. 188 Here, Ti 3 C 2 T x acted as the visible light sensitizer. The heterostructure displayed increased light absorption above 485 nm. However, the E g remained at around 2.55 eV (Fig. 9b). Due to its energy level, the Ti 3 C 2 T x served as a hole accumulation layer in the heterostructure. Therefore, electrons from BiVO 4 were able to ow efficiently to the counter electrode (Pt rod) as the holes were accumulated in Ti 3 C 2 T x , thus extending the lifetime of the charge carriers. Due to this efficient charge separation, the AM 1.5G photocurrent density was improved by >50%, from 2.1 mA cm À2 to 3.5 mA cm À2 (1.23 V vs. RHE). They also reported reannealing treatment in Ar environment at 300 C for 30 min (RA-BiVO 4 ) to produce better contact between bare BiVO 4 and the conductive substrate, which resulted in a higher photocurrent density of 3.0 mA cm À2 . Furthermore, the heterostructure generated 0.8% ABPE at 0.8 V vs. RHE and 56.0% IPCE at 400 nm compared to 0.4% and 29.0% for a bare BiVO 4 , respectively. 188 Fig. 9c displays the Mott-Schottky plot of the bare and modied BiVO 4 , revealing the n-type characteristic of all samples indicated by the positive gradient. The at-band potential (V  ) also disclosed a negative onset in reference to the bare BiVO 4 , and the decreasing slope suggested increasing donor density as compared to the bare BiVO 4 .
As a photocathode, Ti 3 C 2 T x nanosheets also showed promising results when combined with Cu 2 O nanorods superstructure, creating a 3D-2D heterostructure as represented in Fig. 9d. 74 Ti 3 C 2 T x loading on top of Cu 2 O extended the light absorption; the absorption edge is red-shied from $640 nm (1.93 eV) to $680 nm (1.82 eV). Oxygen vacancies were introduced to the nanosheets by performing a H 2 /Ar post-annealing (H : Ti 3 C 2 T x ), resulting in further improvement of light absorption and increased conductivity (from 200 to 1500 mS cm À1 , see Fig. 9e). 74 Overall, the Cu 2 O/H : Ti 3 C 2 T x heterostructure exhibited AM 1.5G photocurrent density of À4.5 mA cm À2 at 0 V vs. RHE, which is 10% higher than that of bare Cu 2 O (Fig. 9f); the ABPE and IPCE were also improved similarly.

Metal oxides/boron carbon nitrides heterostructures
There are several 2D materials in the boron carbon nitride (BCN) family that have shown excellent PEC water splitting performance. To date, this class of materials is categorized based on its elemental composition: boron nitride (BN), boron carbon nitride (BCN) itself, and carbon nitride (C 3 N 4 ). BN is a 2D layered material where boron and nitrogen atoms are bonded to form a hexagonal lattice. Similar to other 2D materials, multiple layers of B-N planes are stacked together by vdW forces. BN comprises several crystal structures, hexagonal (h-BN), rhombohedral (r-BN), wurtzite (w-BN), cubic (c-BN), and wurtzite (w-BN). 213 The hexagonal phase, h-BN, has been reported to be the most stable crystal among them. 214 It also exhibits outstanding thermal stability, electrical conductivity, adsorption capacity, and large surface areas. 213 However, because of its wide E g ($5.0 eV), a stand-alone h-BN cannot be used to efficiently absorb sunlight in PEC water splitting. Hence, h-BN is usually coupled with other semiconductors with lower E g , e.g., Bi 2 WO 6 (ref. 88) or WO 3 , 82 to expand its light harvesting ability. Alternatively, several researchers synthesized the ternary version of BCN, which possess an intermediate property between insulating h-BN and semi-metallic graphene. 215,216 The BCN may exhibit semiconducting properties and its E g can be engineered via the carbon content in the h-BN lattice. 215 Another rising material in the BCN family is the polymericbased material of carbon nitride (C 3 N 4 ). Because of its ability to absorb visible light (E g ¼ 2.7 eV), suitable band positions for water reduction and oxidation, high biocompatibility, and low cost, C 3 N 4 is a potential material for PEC water splitting. C 3 N 4 has a number of different structures: alpha (a-C 3 N 4 ), beta (b-C 3 N 4 ), graphitic (g-C 3 N 4 ), cubic (c-C 3 N 4 ), and pseudo-cubic (pc-C 3 N 4 ); all with diverse characteristics, in which g-C 3 N 4 is the most thermodynamically stable one. 217 To this end, g-C 3 N 4 can be derived from either s-triazine or tri-s-triazine rings. 217 Similar to the h-BN counterpart, g-C 3 N 4 has a hexagonal lattice structure with vdW forces holding the layers.
One example where BCN is integrated into a heterostructure for PEC water splitting was demonstrated by Li et al. 88 In this work, they integrated one of the Aurivillius MOs of Bi 2 WO 6 quantum dots with h-BN nanosheets to create a 0D-2D heterostructure for water oxidation (Fig. 10a). In its isolated form, Bi 2 WO 6 quantum dots yield a strong UV absorption due to the large E g of 2.90 eV. The loading of 0.01 M h-BN into Bi 2 WO 6 altered the latter optical properties in that it became sensitive to the visible light, as indicated by the red-shi of the absorption edge to 462 nm (2.68 eV). This widened absorption range also materialized into a higher photocurrent density of 0.9 mA cm À2 ; 4.5 times higher than that of Bi 2 WO 6 quantum dots (Fig. 10b). The authors argued that the improvements were triggered by the presence of boron vacancies in the h-BN, originating from the thermal substitution method. This defect then led to a negative charge h-BN at approximately À0.25e, compared to pristine h-BN at +0.67e. Due to electrostatic interactions, the negative charge introduced attractive force on the defect site (edge N atoms) toward holes originated from Bi 2 WO 6 , thus facilitating hole transfer from Bi 2 WO 6 toward h-BN. 88 Fig. 10c shows that the smallest diameter of R ct (5.9 kU) was demonstrated in 5BN/BW quantum dots (12.6% lower than pristine h-BN), indicating the highest charge transfer across the heterostructures.
Furthermore, Arzaee et al. recently have developed a a-Fe 2 O 3 / g-C 3 N 4 -based photoanode immersed in 0.5 M Na 2 SO 4 (pH ¼ 7). 159 The heterostructure exhibited an extended visible light absorption and higher absorption below 470 nm due to the g-C 3 N 4 shell on top of the a-Fe 2 O 3 core. However, a sharp spike in the phototransient curve of a-Fe 2 O 3 /g-C 3 N 4 suggests that charge recombination is prevalent in the heterostructure, and an improved interface may be needed to avoid the undesired reaction between electrons and O 2 molecules in the electrolyte.
In another work, Dong et al. successfully created a type II hybrid of 3D nanoporous NiO and 2D nanosheets g-C 3 N 4 as a photocathode (Fig. 10d). 199 With a E g of $2.85 eV, a bare polymeric g-C 3 N 4 can only absorb light up to 440 nm. By hybridizing this 2D material with a wide E g semiconductor of NiO, the absorption is extended up to $500 nm. As a result, a 20fold improvement of the cathodic photocurrent (À0.2 mA cm À2 at 0 V vs. RHE) was observed (Fig. 10e). The heterostructure also maintained a steady photocurrent density over 10 h of photostability testing at 0 V vs. RHE. Finally, it exhibited a hydrogen production rate of 0.16 mmol h À1 ; a four-fold higher rate compared to bare NiO at 0.04 mmol h À1 (À0.2 V vs. RHE under light illumination), as shown in Fig. 10f.

Metal oxides/transition metal dichalcogenides heterostructures
Transition metal dichalcogenides (TMDs) is a class of layered materials with the structure of MX 2 , where M represents the transition metal such as molybdenum (Mo) or tungsten (W), and X denotes sulphur (S) or selenide (Se). In this arrangement, M atoms are sandwiched by X atoms. TMDs display three phases depending on their stacking order: semiconducting phase of 2H (hexagonal) and 3R (rhombohedral) with trigonal prismatic coordination, and metallic phase of 1T (trigonal) with octahedral coordination. 107 Among them, 2H is the most thermodynamically stable phase, while 1T is metastable. In terms of catalytic activity, the 1T phase offers higher activity due to its active basal planes compared to those inert ones in 2H. 218 According to their electronic band structures, TMDs has a thickness-dependent E g . MoS 2 , the most studied TMDs to date, 50,126,219,220 exhibits an indirect E g of 1.29 eV in the bulk form. 107 Because of the connement effect, monolayer MoS 2 has a direct E g of 1.90 eV. 107 WSe 2 also shares the same phenomenon with the bulk and monolayer E g being an indirect 1.20 eV and a direct 1.61 eV, respectively. 221 Another class of chalcogenide-based 2D materials is several nanometers-thick CdX (X ¼ S or Se) with buckled hexagonal structure. 222,223 CdS and CdSe are visible active semiconductors with E g of 2.2 and 1.7 eV, respectively. They also exhibit favorable CB and VB positions for both HER and OER, rendering them as promising semiconductors for PEC water splitting.
Several authors have reported their efforts in utilizing MOs with TMDs for PEC water splitting. Depicted in Fig. 11a, Li et al. realized a vdW heterostructure of 3D TiO 2 nanoowers and 2D MoSe 2 nanosheets with a type II band alignment through a dipcoating process. 77 In this case, the amount of MoSe 2 in the dipping solution dened the photocurrent density of the heterostructure. Since TiO 2 is a wide E g semiconductor, it is unable to absorb visible light above 400 nm. Complementarily, MoSe 2 displays high absorption in the 400-700 nm range. Combining the two, Li et al. showed a heterostructure with smaller E g of 2.77 eV compared to the bare TiO 2 of 2.94 eV. 77 They also found that the optimum dipping concentration of MoSe 2 is 6 mM, which produced a heterostructure with the lowest photoluminescence signal (Fig. 11b), indicating a low charge recombination rate due to the fast photogenerated electron-hole pairs separation across the heterojunction. Consequently, the same heterostructure also demonstrated the highest AM 1.5G photocurrent density of 0.9 mA cm À2 at 1.23 V vs. RHE (Fig. 11c). Some degree of stability was also demonstrated, as only measurements for up to 400 s were shown; no degradation was detected within this limited period.
Another example is by Tekalgne et al., who fabricated a 0D-2D heterostructure of SnO 2 nanoparticles/WS 2 nanoowers as a photocathode as illustrated in Fig. 11d. 78 The effect of WS 2 loading on SnO 2 is evident from the optical absorption spectra. The SnO 2 /WS 2 heterostructure underwent a shi in the absorption edge at 565 nm (2.10 eV), as shown in Fig. 11e. The charge separation was also considerably enhanced, which is indeed expected based on the type II band alignment. The study also demonstrated the detrimental effect of excessive co-catalyst loading, where it lowered the photocurrent density due to parasitic absorption in the WS 2 layer. WS 2 loading of 0.06 M generated À3.5 mA cm À2 at 0 V vs. RHE, while higher WS 2 loading (0.08 M) only generated À2.4 mA cm À2 . 78 Because of the lower charge recombination with the addition of 0.06 M WS 2 (Fig. 11f), the heterostructure had lower R ct compared to pure SnO 2 . Other works by Zheng et al. and Pathak et al. used a similar approach to sensitize TiO 2 nanotubes with a novel TMD of PbS quantum dots for water oxidation. 163,164 EIS analysis under simulated illumination revealed that the heterostructure of TiO 2 /PbS had a minuscule R ct in the low frequency range, signifying low charge recombination due to the improved charge separation.

Metal oxides/metal-organic frameworks heterostructures
Metal-organic frameworks (MOFs) are a class of polymer-based materials with high porosity, comprising metal ions as nodes and organic linkers. 224 The high porosity feature translates to a massive surface area, with the record experimental value of 7140 m 2 g À1 . 225 Beyond surface area, MOFs also offer other benets, such as thermal and mechanical stabilities, large pore volumes, excellent host-guest chemistry, and highly adaptable functionalities. As a result, MOFs have been demonstrated for diverse applications such as gas adsorption, 226 energy storage, 227 solar energy conversion, 228 and sensors. 229 For PEC water splitting, MOFs with high stability in an aqueous medium can be achieved with either high or low coordination numbers. High valence metals (Zr 4+ , Fe 3+ , and Cr 3+ with carboxylate bridging units) can establish a rigid structure so that the MOFs interact less with water molecules, thus preventing structural degradation. 230 Low valence metals (Zn 2+ , Co 2+ , Cu 2+ , and Ni 2+ with azolate linkers), on the other hand, can also form a stable framework in an aqueous medium owing to the formation of a tetrahedral structure that binds to the azolate linkers with angles resembling highly stable zeolite minerals. 231 Recently, Dou et al. demonstrated the effect of MOFs incorporation for PEC water splitting using low valence metal group as depicted in Fig. 12a. 190 They established a ZnO nanorods/ cobalt-derived MOF (ZIF-67) photoanode in a core-shell heterostructure system. Pure ZnO displays an absorption edge at around 410 nm, suggesting a large E g of 3.18 eV. The hybridization with ZIF-67 slightly shied this edge, hence the E g . However, according to the UV-visible absorption spectrum, ZIF-67 provided a substantial improvement in light trapping ability for the heterostructure within 450-700 nm due to its porous nature. The effect of porous ZIF-67 was also prominent for the charge transfer efficiency. Here, ZIF-67 improved the transfer of photogenerated electrons by forming multiple transfer channels from the ZIF-67 shell toward the ZnO core, resulting in a lower probability for charge recombination. Fig. 12b shows the resulting ABPE and IPCE of 0.1% (0.2 V vs. SCE) and 3.1% at 400 nm, respectively. ZnO had a low AM 1.5G photocurrent density of 0.4 mA cm À2 due to intensive charge recombination, which was greatly increased to 1.1 mA cm À2 when ZIF-67 was introduced (0.6 V vs. SCE), as shown in Fig. 12c. This was further improved to $2.0 mA cm À2 when Au is also introduced due to the plasmonic effect. The heterostructure generated a higher oxygen evolution rate of 30 mmol h À1 , nearly three times higher compared to bare ZnO.
Other  Fig. 12d. 233 The presence of Cu 3 (BTC) 2 on top of Cu 2 O only had a little impact on the light harvesting ability indicated by the absorption edge. Nevertheless, the porous nature of Cu 3 (BTC) 2 improved the photocurrent density by providing a larger catalytically active area and by forming a preferential electron transfer channel from Cu 2 O to Cu 3 (BTC) 2 , affecting the IPCE as the heterostructure registered 44.0% IPCE at 400 nm, compared to 18.0% of pristine Cu 2 O under the same condition (Fig. 12e). The J-t curves of the heterostructure displayed a constant photocurrent density of À0.7 mA cm À2 throughout the testing period (3500 s), signifying a great photostability (Fig. 12f).

Metal oxides/bismuth oxyhalides heterostructures
Bismuth oxyhalides are semiconducting layered materials with the ternary composition of BiOX, where X is the halide element of Cl, Br, or I. They have a tetragonal matlockite crystal structure with each layer comprising an oxygen plane in between two halide and bismuth planes with the covalent arrangement of X-Bi-O-Bi-X. Typically, BiOX shows exceptional physicochemical properties and has been investigated for several applications, such as reective pigments, 234 water purication, 235 gas sensors, 236 storage materials, 237 and catalysis. 238 In PEC water splitting, it has been identied that BiOX has an increased photogenerated charge separation in the [001] direction due to the perpendicular electric eld within a single layer of BiOX. 198,239 Additionally, it has been argued that their indirect E g is benecial for PEC water splitting as the photoexcited electrons need to move through the denite k-space and it considerably lowers the direct charge recombination rate. 240 Efforts in integrating BiOX with other MO semiconductors also recently received considerable attention. Fig. 13a     Due to the corresponding Fermi level of each semiconductor, electrons in n-type BiVO 4 diffused into p-type BiOI, resulting in a positively charged BiVO 4 , and vice versa. This further enhanced the charge carrier separation and reduced the charge recombination. Such a 2D-2D heterostructure of BiVO 4 /BiOI could reach 3.3 mA cm À2 AM 1.5G photocurrent density at 1.23 V vs. RHE, far higher compared to the pristine BiVO 4 at 1.2 mA cm À2 under identical circumstances. Furthermore, the ABPE reached as high as 1.0% at 0.7 V vs. RHE. Correspondingly, the heterostructure's IPCE also showed an improvement about three times higher of 44.0% at 400 nm compared to the pure BiVO 4 (Fig. 13b). The separation efficiency was also studied using 15% H 2 O 2 as hole scavenger alongside 0.5 M KH 2 PO 4 electrolyte to eliminate any catalytic limitation or surface recombination. Under an applied potential of 1.23 V vs. RHE, the heterostructure's separation efficiency reached $70%, which is more than double that of BiVO 4 at $30%, signifying suppressed charge recombination originated from the presence of a p-n junction (Fig. 13c).
Several groups also reported other integration of MOs and BiOX. For example, Zhou et al. fabricated TiO 2 /BiOCl, 177 while Zhou et al. synthesized CuWO 4 /BiOI 89 to create a p-n junction heterostructures for OER. The creation of p-n junction in both cases signicantly lowered the probability of carriers to recombine by providing an efficient built-in electric eld induced by the depletion region on the surface of the heterostructures.
As a photocathode, BiOX of 2D non-stoichiometric BiOBr 0.9 I 0.1 nanoakes has also been coupled with 2D ZnO nanosheets that have reversible conductivity type modulation from n-type to p-type, or vice versa, based on the ZnO's mass fraction. 198 The radiative recombination could be determined by PL spectra, where the increased emissions at 374 nm (nearband-edge), along with 400 and 581 nm (defects) from 40ZB sample (i.e., 40 wt% ZnO) compared to pure ZnO counterpart can be attributed to intensied radiative recombination. However, 10ZB and 20ZB samples register lower PL emissions, suggesting an improved charge separation (Fig. 13d). According to the reectance spectra, the pristine ZnO and BiOBr 0.9 I 0.1 showed an absorption edge at 412 and 537 nm, revealing their E g of 3.00 and 2.42 eV, respectively. In their heterostructure form, a negligible E g shi was detected. However, the heterostructure showed an increased light absorption within the visible range as a function of the wt% of ZnO. Moreover, increasing the wt% of ZnO modied the nature of the conductivity from n-to p-type. The pristine BiOBr 0.9 I 0.1 exhibited an anodic photocurrent density, signifying an n-type conductivity. When the ZnO content reaches 10 wt%, the heterostructure showed a cathodic photocurrent instead, suggesting a switch to a p-type conductivity (Fig. 13e). This conductivity tuning was attributed to the presence of defects along the interfaces. At this stage, i.e., when the ZnO content is still not that high, Zn 2+ substituted Bi 3+ without forming sufficient oxygen vacancies. As a consequence, additional holes were introduced as the charge compensation, altering the properties of BiOBr 0.9 I 0.1 from ntype to p-type. 198 Increasing the mass fraction of ZnO further to 40 wt%, however, also introduced more oxygen vacancies, which is oen suggested as electron donors. Therefore, these excessive donors reverted the conductivity type to n-type. Under chopped visible light, the heterostructure with 20 wt% of ZnO registered a cathodic photocurrent density of 0.8 mA cm À2 at À0.2 V vs. SCE. This reversible conductivity type also affected the carrier mobility shown by the Nyquist plot with smaller R ct , where the p-type ZnO/BiOBr 0.9 I 0.1 (20ZB) was more efficient in suppressing the charge recombination than the n-type counterpart (40ZB) (Fig. 13f).

Conclusion and outlook
In summary, we have thoroughly reviewed the various nanoscale heterostructures that are composed of metal oxides and 2D materials. This exercise is motivated by the many limitations of metal oxides as one of the most investigated materials for PEC water splitting. In particular, many metal oxide photoelectrodes suffer from having a narrow visible light absorption, unfavorable band position, and poor charge carrier mobilities. Some oxides also suffer from instability in aqueous solutions due to self-oxidation or reduction. The formation of nanoscale heterostructure with 2D materials simultaneously addresses these limitations. From the literature, it is clear that the efforts in developing MOs/2D materials heterostructure as photoanodes are more prominent than as photocathodes. This nding is in mutual agreement with the research direction in PEC water splitting, where OER poses more challenges compared to HER (e.g., sluggish OER kinetics). Correspondingly, the building block of photoanodes is an n-type semiconductor, which is fullled by most MOs (e.g., ZnO, TiO 2 , a-Fe 2 O 3 , and BiVO 4 ) due to the low formation energy of electron donor defects, such as oxygen vacancies. 241 On the other hand, photocathodes require p-type semiconductor, which is a much less reported for MOs (a couple of examples include NiO and Cu 2 O) because of the high energy barrier of hole acceptor defects (i.e., cation vacancies) and low formation energy of hole nullier defects (i.e., anion vacancies). 241 We also surveyed and categorized the heterostructures based on the nanoscale dimensionality (i.e., 0D, 1D, 2D, 3D). In most cases, MOs are morphed into a 1D morphology (e.g., nanorods, nanowires, and nanotubes) to capitalize on their high surface areas for light trapping and catalysis owing to their high aspect ratio (length vs. diameter). On the other hand, 2D materials have also been employed in their lower dimensionality forms of 0D (quantum dots) or 1D (nanotubes; by folding the 2D plane into a cylindrical shape, similar to the case of carbon nanotubes). For the most part, 2D materials are interfaced as is, to fully take the advantage of the abundant basal planes of the 2D plane that can be engineered into active sites for catalysis. We also reviewed the possible interfacial electronic structure alignment in the heterostructures (i.e., type II, p-n junction, and Z-scheme). We found that type II is more frequently compared to other mechanisms due to its simplicity (assuming that the energy levels from the two materials allow charge transfer). Z-scheme, on the other hand, is a less established system, but shows notable architecture developments in charge transfer channels (i.e., traditional shuttle redox mediator, all-solid-state, and direct Z-scheme). 143 Thus, in our perspectives, deeper exploration in Zscheme MOs/2D materials heterostructure is to be expected in the future.
The combination of metal oxides with specic material classes within the various 2D materials family is also highlighted. In many of these cases, the introduction of 2D materials successfully (i) increases the visible light absorption range, (ii) enhances the charge carrier separation due to the formation of a favorable interface, (iii) induces manifold improvement of the photocurrent, and (iv) extends the stability of the metal oxide photoelectrodes. Among these reports, graphene and its derivatives (e.g., graphene oxide and reduced graphene oxide) still dominate as the most commonly used 2D materials in PEC water splitting, but growing numbers of work are available that employ more abundant and simpler-to-synthesize 2D materials, e.g., transition metal dichalcogenides and boron carbon nitride. Fig. 14 visualizes the performance (in terms of the reported photocurrent densities at the water redox potentials) of these vast combinations of MOs/2D materials heterostructures over the last decade (see the complete list in Tables 1 and 2).
Despite the progress and growing amount of interest, it is clear that the development of MOs/2D materials heterostructures for PEC water splitting is still some steps away from the required target. Photoelectrodes that deliver high performance (e.g., photocurrent density >5 mA cm À2 ) have been reported over the past decade, but the number is limited, and they represent more the exception rather than the norm. This is of course disappointing given the premise of various advantages from adopting such a heterostructure for PEC water splitting. Such a limitation, however, does not seem to be specic to the heterostructure architecture itself; instead, the limited performance seems to be caused by the intrinsic limitations of the investigated MOs so far. In addition, efforts in the eld in coupling the appropriate MOs and 2D materials are still mostly based on trial and error, instead of driven by guidelines. It is therefore clear that deeper fundamental studies are still needed to fully unravel and exploit the expected theoretical advantages of such heterostructures. The interface between the two materials is obviously a critical element; we have to rst understand the key parameters that dene the interfacial properties between the metal oxides and the 2D materials. The combination of advanced spectroscopic experiments with material computations will most likely be needed. At the same time, synthesis approaches and fabrication methodologies that allow full control of the interfacial properties will also be needed. From the materials perspective, exploration of novel MOs and 2D materials with appropriate crystal structure and optoelectronic properties aided by advanced calculation methods, e.g., density functional theory (DFT) calculations, accompanied by machine learning may hold the key for future advancement. 242 For example, recent advances make it possible to directly synthesize 2D MOs using scalable methods, such as exfoliation by weakening the strong covalent bond of MOs 243 and selfassembly by introducing amphiphilic block copolymer and alcohol co-surfactant. 244 Therefore, the establishment of heterostructure of these 2D MOs with other 2D materials, specically for PEC water splitting, is not far from realization. Finally, more general beyond the MOs/2D materials heterostructure, employing principles from other interdisciplinary research may provide novel approaches to accomplish more efficient PEC water splitting devices, such as utilizing solid oxide

Conflicts of interest
The authors declare no conicts of interest.