Inside‐and‐Out Semiconductor Engineering for CO2 Photoreduction: From Recent Advances to New Trends

Photocatalytic CO2 reduction attracts substantial interests for the production of chemical fuels via solar energy conversion, but the activity, stability, and selectivity of products were severely determined by the efficiencies of light harvesting, charge migration, and surface reactions. Structural engineering is a promising tactic to address the aforementioned crucial factors for boosting CO2 photoreduction. Herein, a timely and comprehensive review focusing on the recent advances in photocatalytic CO2 conversion based on the design strategies over nano‐/microstructure, crystalline and band structure, surface structure and interface structure is provided, which covers both the thermodynamic and kinetic challenges in CO2 photoreduction process. The key parameters essential for tailoring the size, morphology, porosity, bandgap, surface, or interfacial properties of photocatalysts are emphasized toward the efficient and selective conversion of CO2 into valuable chemicals. New trends and strategies in the structural design to meet the demands for prominent CO2 photoreduction activity are also introduced. It is expected to furnish a comprehensive guideline for inside‐and‐out design of state‐of‐the‐art photocatalysts with well‐defined structures for CO2 conversion.

CO 2 photoreduction were introduced. Then, the photocatalysts for CO 2 reduction with diverse kinds of structures, including nano-/microstructure, crystalline and band structures, surface structure and interfacial structure, were highlighted and discussed in details. Furthermore, several new structure systems were summarized. Finally, the key challenges and outlook on the future direction in this field were proposed.

Thermodynamic and Kinetic Challenges
Carbon dioxide is a stable molecule, featuring the linear and nonpolar nature of the molecule. [26,27] Despite development for many years, the conversion efficiency and product selectivity of CO 2 photoreduction in mild condition are still quite low. So far, energy conversion efficiency of solar-to-fuel is %1% under AM 1.5G, which is mainly attributed to the following reasons: [28][29][30][31][32][33][34] 1) CO 2 is one of the most chemically inert reactants. The breaking of C─O bonds needs to consume huge energy with appropriate catalysts. The Gibbs free energy changes (ΔG) of specific reactions are shown in Table 1. [35,36] A standard state is positive (ΔG θ > 0), and it means that these reactions are nonspontaneous. In addition, the standard enthalpy changes (ΔH θ > 0) are also shown in Table 1, in which the positive ΔH θ values show that these reaction processes are endothermic under a standard state. It is apparent that these multielectrons reduction reactions from CO 2 to CO or organic small molecules are thermodynamically uphill, which are difficult to occur at ambient temperature and pressure. [36,37] 2) During the CO 2 photoreduction process, CO 2 À as a product of single-electron transfer demands a negative equilibrium potential of %1.9 V versus normal hydrogen electrode (NHE). [27,38] This overpotential is too high to accomplish for most photocatalytic systems, whereas other products from multielectrons transfer with a lower overpotential are easy to be generated. The band structure of some typical photocatalysts and the standard reduction potentials in CO 2 reduction are shown in Figure 2. [12,39] In the CO 2 photoreduction process, it requires that the conduction band (CB) position of photocatalysts is above the standard reduction potential, and the valence band (VB) position of photocatalysts needs potentials below the standard oxidation potential to achieve photooxidation Table 1. Calculated thermodynamic parameters (ΔG θ , ΔH θ , and ΔS θ ) of typical CO 2 conversion reactions. Reproduced with permission. [36] Copyright 2017, Elsevier.  process. [40,41] In the absence of sacrificial agent, the VB position of photocatalysts should be more positive than 1.23 V (vs NHE, in pH ¼ 7), and thus photoinduced holes can contribute to O 2 formation. [42] But some photocatalysts can hardly satisfy both the standard reduction and oxidation potentials requirement. Achieving trial CO 2 reduction in the presence of sacrificial reagents is a feasible way for these photocatalysts. For example, triethanolamine is usually used as an electron donor in aqueous or organic solvent systems for CO 2 reduction. The conversions of CO 2 into valuable products, such as CH 4 , [43][44][45][46][47][48][49][50] CH 3 OH, [51][52][53][54][55] or HCOOH, [56][57][58][59][60][61] have been reported, but the yield and selectivity of these products still need to be improved. [62] As the most common product in gas system, CO formation just requires two protons and two electrons, whereas CH 4 formation needs eight protons and electrons and prefers to be produced on noble metal cocatalysts. HCOOH generation can usually be achieved by sulfides (e.g., ZnS) with suitable metal doping, such as Ni or Cd, in liquid systems, and this process needs the same number of protons and electrons as CO formation, but demanding a higher overpotential. [63] Other products are usually hard to be observed with a high activity and selectivity, because they are always thermodynamically uphill and consume different electrons or protons through multiple reactions. [64] In practice, photocatalytic CO 2 reduction is much more complex due to various competing reactions, such as H 2 evolution in aqueous system. Thus, the mechanisms in diverse solvent system were studied for pursuing a high yield of target products. The mechanisms of CO 2 photoreduction in both gas and liquid phases to obtain desired products are still challenging.
There are also many kinetic challenges in CO 2 photoreduction. Specifically, the photoreduction reaction of CO 2 over photocatalysts includes the following processes: 1) light harvesting and generation of electron-hole pairs; 2) charge separation and migration from bulk to the surface of the semiconductors accompanied by the recombination electrons and holes in the bulk and on the surface of photocatalysts; 3) the reaction of electrons with absorbed CO 2 (normally the holes take part in O 2 evolution reaction in the meantime) and desorption of products from the surface of photocatalyst. [65][66][67] Therefore, light harvesting, charge separation, and surface reactions are the three crucial steps that play decisive roles in CO 2 photoreduction.

Development of Semiconductor Photocatalysts for CO 2 Reduction
High photocatalytic activity with desirable products, including CO, CH 4 , CH 3 OH, or HCOOH, is a persistent pursuit for exploration of high-performance photocatalysts CO 2 reduction. [68][69][70] In other words, the purposes of structural engineering for photocatalysts are to increase the conversion efficiency and to control the selectivity of products in the CO 2 photoreduction reaction. [16,[71][72][73] Generally, there are three factors that can regulate the final products of photocatalytic CO 2 reduction. 1) Efficient photogenerated charge separation can realize high surface electron density, which can regulate the final products by promoting multielectron reduction process. 2) Photocatalytic CO 2 reduction as a multistep reaction, diverse active sites on the surface can greatly affect the products selectivity. For example, regulating surface structure can affect the rate-determining step or key intermediates during the surface reaction, and thus obtaining different final products. 3) A strong adsorption capability of CO 2 molecules usually allows a high catalytic activity, but desorption properties may determine the final products. A decent desorption ability of target products is a key factor for pursuing a high selectivity.

Influence of Structure on Photocatalytic Activity
In the photocatalytic process of CO 2 photoreduction, the total quantum efficiency (η) is considered to be determined by light-harvesting efficiency (η 1 ), charge migration efficiency (η 2 ), and redox reaction (charge utilization) efficiency (η 3 ), as shown in Equation (1). [35] η ¼ η 1 Â η 2 Â η 3 (1) After the formation of photoexcited charge carriers, most of the photoexcited electrons and holes tend to be consumed via recombination before the completion of surface catalytic reactions ( Figure 4). [116][117][118][119][120][121] Because the recombination processes significantly decrease the quantum efficiency of CO 2 photoreduction, the researchers have been trying their best to relieve the recombination rate and degree of electrons and holes. [42,122] It is worth noting that the time scales for different charge movement behaviors is also different. The charge separation and transport process take place within 10 À6 s, and the surface reaction is a process of 10 À4 s. In contrast, the recombination of electrons and holes occurs from 10 À12 to 10 À3 s, and particularly the recombination in the bulk phase of photocatalysts (approximately several picoseconds) is much faster than other processes and it also tends to occur through the entire photocatalytic process. [40,[123][124][125] As each of the aforementioned steps needs to be well optimized to achieve highly active and selective photocatalysts, various structural design methods have been developed for promoting the photocatalytic CO 2 reduction performance, as briefly listed in the following: 1) Construction of nano-/microstructures strategies can control the morphology, size, and shape of materials, which have substantial advantages for enhancing CO 2 photoreduction activity, such as shortening charge migration pathway, [126][127][128] increasing reactive catalytic sites, [129][130][131][132][133] boosting charge separation as well as extending optical transmission  www.advancedsciencenews.com www.small-structures.com length. [134][135][136][137][138] 2) Doping and formation of solid solution are two main approaches to tailor the crystalline and band structures of photocatalysts. The alteration in atomic composition bonding and coordination may result in the changes of light absorption, [139,140] band energy edge level, [141][142][143] and charge separation. [144,145] 3) Diverse reactive sites can be introduced on the surface by adjusting exposing specific facet and facet junction. Reactive exposing facet with amount of unsaturated coordinated atoms can facilitate the catalytic reactions, and facet junction with coexposed anisotropic facets allows spatial separation of photogenerated electrons and holes. [146][147][148] In addition, creating acid-base sites or vacancy and ionic modification on the surface are capable of promoting light absorption, charge separation, and CO 2 activation. [149][150][151] 4) Schottky junction and Z-scheme junction are typical and efficient heterojunctions that can well engineer the interface structure. In Schottky junction, electrons move from semiconductor to metal to obstruct backflow, resulting in a high charge transfer efficiency. [152,153] In Z-scheme junction, the photogenerated charge carriers are efficiently separated through the interface between semiconductor and semiconductor or semiconductor-metal-semiconductor, and meanwhile the strong oxidative and reductive potentials were maintained. [154][155][156][157] 5) A series of new photocatalytic systems are developed for CO 2 photoreduction, such as metal organic frameworks (MOFs), [158][159][160][161][162][163] covalent organic frameworks (COFs), [164][165][166][167][168][169] semiconductor biohybrids systems, [170][171][172] and single atom. [173][174][175][176][177] These systems have different chemical affinities to CO 2 molecules, and thus may change the redox reaction pathways during CO 2 reduction process. Meanwhile, polarity enhancement and creation of spatially separated active sites as new strategies for enhancing charge separation and optimizing reactive catalytic sites attracts intense attentions. [178][179][180] 3. Nano-/Microstructure Engineering

Dimension Regulation
With the development of photocatalysts, many nanostructures have been fabricated including 0D QDs, [181,182] 1D structures (NWs, [107] nanorods, [183][184][185][186][187] nanobelts, [188,189] nanotubes, [190,191] and nanofibers [186,192] ), 2D (nanoplates, [193] nanosheets [194] ), and 3D hierarchical structures. [195][196][197][198][199] These structures have received extensive attentions in CO 2 photoreduction because of improved photoresponsive region, optimized charge transfer path, as well as ample active sites for CO 2 or H 2 O adsorption. [105,[200][201][202] 3.1.1. 0D Quantum Dots QDs are aggregated tiny particles or nanocrystals with diameter smaller than 10 nm. Due to their small size, the quantum confinement effect becomes prominent, which can cause the change of band structure and relative properties in semiconductor. [105,[203][204][205] CsPbBr 3 (CPB) QDs were developed to convert CO 2 into solar fuels under AM 1.5G simulated illumination. [206] As a specific feature in QDs, quenching the emission peak in photoluminescence (PL) spectra was observed in CPB ( Figure 5a). CPB continually donated photogenerated electrons to CO 2 , which achieved a yield of 23.7 μmol g À1 h À1 in electron consumption for reduced products. Over 99.3% of generated electrons were consumed by the formation of CO and CH 4 , with the average rate of 4.1 and 1.9 μmol g À1 h À1 , respectively. Further, CPB QDs were anchored on NHx-rich porous g-C 3 N 4 nanosheets (PCN) by N─Br bonding. [207] CPB QDs with a size of 10 nm are dispersed evenly on PCN (Figure 5b), because the lattice spacing of 0.58 nm is square with (110) planes of orthorhombic CPB. The 20 wt% CPB with PCN as the optimal ratio showed a CO yield of 149 μmol g À1 h À1 , which was 15 and 3 times higher than pure CPB and PCN under visible-light irradiation, respectively. The unique interaction via N─Br bond caused improved charge separation and extended the lifetime of charge carriers between CPB and PCN.
In addition to the single semiconductor system, binary QDs systems were also developed for photocatalytic CO 2 reduction. Lian et al. [182] demonstrated that the combination of CuInS 2 / ZnS QDs as sensitizers with trimethylamine-functionalized iron tetrapheny-porphyrin (FeTMA) for CO evolution under 450 nm light irradiation. The absorption peak is located at 420 nm originating from the QDs the size of 2.5 nm (red solid line in Figure 5c). The light absorption of QDs was clearly improved by combined with FeTMA (red dashed line). Furthermore, the FeTMA molecules tended to connect with QDs partially, because of observation of an extra absorption peak at 415 nm. The turnover number (TON) of binary photocatalysts was 450 for CO evolution after 30 h irradiation, with a selectivity of 99%. The sensitization efficiency of this system was over 11 times as large as the current record for Fe-porphyrin system. This enhancement efficiency was inferred from the formation of active superstructures that each FeTMA can collect photoelectrons from surrounded QDs (inset of Figure 5c). Moreover, Li et al. [106] reported that CQDs/Cu 2 O composites achieved a high photocatalytic CO 2 reduction activity for yielding MeOH (55.7 μmol g À1 h À1 ) under visible light. From the high-resolution transmission electron microscopy (HRTEM) images of the carbon QDs (CQDs)/Cu 2 O composite, it can be seen that the 5 nm CQDs were deposited on the surface of Cu 2 O particles, and the lattice fringes of 0. 25  Due to easily tuned surface states and good absorption of visible light, QDs as catalysts or cocatalysts show many advantages in CO 2 conversion. They usually play a vital role in hybrid structures to realize increased redox active sites and promoted photoabsorption. In addition, QDs are much easier to bond with other components by electrostatic assembly than bulk photocatalysts.

1D Nanorods, NWs, and Nanotubes
Constructing 1D nanorods or nanobelts usually expose active facets with large specific surface areas, which are beneficial to surface redox reactions. [208] The morphology of 1D also leads to a directional charge transfer pathway, and the small diameter of these photocatalysts allows a rapid transport of charge carriers from bulk phase to the surface (Figure 6a-c). [209] Thus, the fabrication of 1D photocatalysts is a feasible way to improve the activity for CO 2 photoreduction. [210] Kar et al. [211] reported the preparation of square-shaped crosssections TiO 2 nanorods by flame annealing method. The light absorption was extended to 620 nm with a dominant absorption peak at 450 nm. In addition, it was deduced that the long and ultrathin nanobelts probably have more active sites for CO 2 reduction reactions. As a result, a high CH 4 yield with a rate of 156.5 μmol g À1 h À1 was achieved over flame-annealed TiO 2 nanotubes in aqueous electrolyte (FANT-aq). NWs are also promising substrates for constructing composite photocatalysts. Wang et al. [212] reported the synthesis of large CuO NWs arrays (10 8 cm À2 ) modified with ZnO nanoparticles (NPs) on their surface (Figure 6a), due to the epitaxial relationship of CuO (111) and ZnO (101). The fabricated composite photocatalysts had a favorable band structure position of CuO and ZnO. Under UV-vis light, a peak CO production rate (1.98 mmolg À1 h À1 ) was obtained for the optimal sample in gas system, which was comparing to no CO generation in bare CuO NWs. It was attributed that the defects on ZnO islands trapped electrons and enhanced their lifetimes by three times. Once the electrons were transferred from CuO NWs to ZnO islands, they would be trapped by the defects on ZnO and took a longer time to be recombined.
When the diameter of the NW is in the range of 1-100 nm, NW array structures tend to represent many interesting optical or electrical properties, [109] and these properties can be improved by elements doping and defects introducing. AlOtaibi et al. [213] demonstrated that the Mg-doped InGaN/GaN NW arrays on Si substrate reduced CO 2 into different products under sunlight ( Figure 7a). Pt NPs with the diameter of %2-3 were evenly distributed on the NWs, which were composed of GaN shell at the InGaN segments (Figure 7b,c). Mg-doped InGaN/GaN is p-type semiconductor, whereas Ge-doped InGaN/GaN is n-type semiconductor. The CO and CH 4 yield of p-type InGaN/GaN NWs were 5 and 0.86 mmolg À1 h À1 , much higher than that of undoped NW arrays (0.4 and 0.18 mmolg À1 h À1 ), respectively. The CO and CH 4 production rates of n-type InGaN/GaN NW arrays were just 0.09 and 0.08 mmolg À1 h À1 , separately. In addition, a high CH 3 OH evolution rate of 0.5 mmolg À1 h À1 was observed only for the p-type InGaN/GaN NWs (Figure 7d,e). The CO production rate of Mg-doped NW arrays was over 50 times higher than Ge-doped NW arrays, because Mg doping Figure 5. a) Steady-state PL spectra with an excitation wavelength of 369.6 nm. Adapted with permission. [206] Copyright 2017, American Chemical Society. b) TEM image of CsPbBr 3 QDs. Adapted with permission. [207] Copyright 2018, Wiley-VCH. c) Ground-state absorption spectra of 7.5 Â10 À6 M FeTMA only (black solid line), 15 Â10 À6 M 3-mercaptopropionic acid (MPA)-capped QDs in water with (red dashed line) and without (red solid line) 7.5 Â10 À6 M FeTMA added, and the difference spectrum: (QDs/FeTMA mixture minus QDs only), i.e., FeTMA in the presence of QDs (black dashed line). All samples are in CO 2 -purged water. Inset: proposed assembly mechanism for a subunit of a QDs/FeTMA complex. Only one MPA ligand per QDs is drawn, for clarity, but there is an average of 150 ligands per QD. Adapted with permission. [182] Copyright 2018, American Chemical Society. d) HRTEM images of the carbon QDs (CQDs)/Cu 2 O composite. Adapted with permission. [106] Copyright 2015, Wiley-VCH.
decreased the surface potential barrier of InGaN/GaN to promote the adsorption and activation of CO 2 .
Combining a surface plasmon resonance with NW array structures can promote both light harvesting and transfer of charge carriers in the semiconductors. Cu 2 O nanorod arrays were modified by graphene and Au-Cu nanoalloys (3D Au-Cu/graphene/ Cu 2 O) for CO 2 reduction into CH 3 OH (Figure 6d). [108] The transfer pathway of photogenerated electrons in this system was proposed. After the capture of photons by Cu 2 O to produce electrons and holes, electrons were transferred to the graphene layer and further moved to the Au-Cu particles. Thus, the CH 3 OH yield of 3D Au-Cu/graphene/Cu 2 O system (18.80 ppm cm À2 h À1 ) is more than six times higher than that of pristine 3D Cu 2 O (2.42 ppm cm À2 h À1 ). AlOtaibi et al. [214] reported the preparation of gallium nitride NW arrays (GaN NWs) on silicon (Figure 6e), which were able to reduce CO 2 into CH 4 and CO under UV-vis irradiation. The CH 4 yield of GaN NWs with Rh/Cr 2 O 3 cocatalyst was increased to %3.5 μmol g À1 h À1 in 24 h, and the CH 4 evolution was further increased to %14.8 μmol g À1 h À1 by Pt NPs modification (Figure 6f ). Figure 6. a) Schematic of CuO NWs with dense ZnO islands, Adapted with permission. [212] Copyright 2015, American Chemical Society. b) TEM, and c) HRTEM images of the Zn 2 GeO 4 nanoribbons. The inset of (c) shows the FFT pattern obtained from the HRTEM image. Adapted with permission. [474] Copyright 2010, American Chemical Society. d) SEM images of AuCu/graphene/Cu2O/Cu mesh. Adapted with permission. [108] Copyright 2015, Wiley-VCH. e) A 45 C tilted scanning electron microscope image of GaN NWs grown on Si (111) substrate. f ) TEM image of Pt NPs deposited on GaN NWs. Inset: schematic of the photoreduction processes of CO 2 Pt-decorated GaN NWs. Adapted with permission. [214] Copyright 2015, American Chemical Society.
In addition, by assembling the NW arrays into 3D architectures, such as branched superstructures, the activity of CO 2 photoreduction could be favorably enhanced.

2D Ultrathin Nanosheets
Recent studies demonstrated that 2D nanosheets have abundant reactive sites, which are in the favor of reactants' diffusion and products' desorption for enhancing photocatalytic activity. Meanwhile, some specific products in CO 2 photoreduction can easily desorb from the surfaces of certain nanosheets, resulting in a high selectivity in reaction. [215][216][217] A large variety of monocomponent and multicomponent hybrid nanosheets were prepared to maximize these advantages in CO 2 reduction.
The SAPO-5 (a classic molecular sieve) nanosheets were fabricated with a thickness of about 3.0 nm. [218] The active sites was increased on the nanosheets, and the ultrathin geometry of the nanosheets promotes the diffusion of CO 2 and H 2 O and desorption of the products. It was also found that the excited state existed in a long-time scale on the nanosheets, contributing to high photocatalytic activity. Thus, the ultrathin SAPO-5 nanosheets showed about 6 times as high CH 4 evolution as SAPO-5 microrods. To take advantage of the complementary features of several kinds of nanosheets, composite nanosheets have been developed to improve charge dynamics and visible-light harvesting. The composite photocatalyst consisting of TiO 2 and carbon nitride nanosheets (CNNS) was synthesized for CO 2 reduction to CO. [219] TiO 2 /CNNS heterostructures with {001} exposing facets of TiO 2 led to an increased CO 2 adsorption capacity and enhanced charge transfer. he TiO 2 /CNNS heterostructure photocatalyst showed a high CO production rate, which %10 times as high as commercial TiO 2 as benchmark photocatalyst (P25). In addition, the well-designed heterojunction 2D-ZnV 2 O 6 / 2D-protonated g-C 3 N 4 (pCN) was constructed with intimate interfacial contact. In a liquid media, CH 3 OH production rate of 2D-ZnV 2 O 6 /pCN (3742 μmol g À1 ) was 1.15 and 5 times as much as that of the pure ZnV 2 O 6 (3254 μmol g À1 ) and protonated g-C 3 N 4 (753 μmol g À1 ) within 4 h, respectively. In gas phase system, the CO was the dominant product of 2D-ZnV 2 O 6 /pCN nanosheets with a yield of 3237 μmol g À1 after 4 h.
With the decrease in thickness, large number of unsaturated atoms emerged on the surface and edges, which endow 2D ultrathin materials with sufficient active sites for molecules adsorption and subsequently redox reactions. [220][221][222][223] The BiOBr atomic layers were fabricated to optimize CO 2 reduction processes. The average thickness of BiOBr nanosheets is about 0.81 nm, which is equal to the thickness of single-unit-cell along [001] direction (Figure 8a-c). Single-unit-layer BiOBr contained abundant oxygen vacancy (OV), which enabled the photoabsorption in visible region and trapped electrons to activate CO 2 molecules for producing COOH* intermediate (Figure 8d-g). Therefore, BiOBr atomic layers demonstrated a greatly enhanced CO 2 reduction activity, and the CO evolution rate (87.4 μmol g À1 h À1 ) is 24 times as high as that of bulk BiOBr. Han et al. reported the synthesis of three-unit-cells single-crystal InVO 4 nanosheets, which is %1.5 nm in thickness [224] (Figure 8h  . The x in In x Ga 1-x N is presented in the color scale shown on the left. d) Measured CO, CH 4 , and CH 3 OH evolution rates on Pt-decorated p-InGaN/GaN NW photocatalysts under the full spectrum of a Xe lamp equipped with an AM1.5G filter. e) CH 3 OH evolution over Pt-decorated p-InGaN/GaN NWs as a function of time under visible-light illumination (>400 nm). The inset shows CO and CH 4 generation rates on Pt-decorated NWs. Adapted with permission. [213] Copyright 2016, American Chemical Society.
higher than that of pristine InVO 4 (0-10 mV) under light illumination (Figure 8k-m), indicating that the ultrathin nanosheets substantially depressed the recombination of electrons and holes. CO as the main product can easily desorb from the {110} surface of InVO 4 ultrathin sheets in the photoreduction process, which resulted in a 6-times increase in CO generation rate (18.28 μmol g À1 h À1 ) compared with the nanocube counterpart (3.21 μmol g À1 h À1 ). In addition, ultrathin CoSe 2 with abundant Co vacancies was also developed by exfoliating CoSe 2 precursor by Xie and coworkers (Figure 9a,b). [225] A large-scale synthesis of ultrathin nanosheets with desirable defects is a great challenge. The gram-scale single-unit-cell vanadium(V)-defective orthorhombic-BiVO 4 layers (V v -rich o-BiVO 4 ) was achieved by Xie and coworkers (Figure 9c,d). [226] SPV measurement demonstrated that V vacancies induced the photogenerated electrons to move toward the surface, while the holes were transferred to the bulk of V v -rich o-BiVO 4 ( Figure 9e). Abundant V vacancy increased the lifetime of photogenerated carriers from 74.5 to 143.6 ns (Figure 9f,g). As a result, V v -rich o-BiVO 4 exhibited a high CH 3 OH formation rate up to 398.3 μmol g À1 h À1 and an apparent quantum efficiency (AQE) of 5.96% at 350 nm, far exceeding that of V v -poor o-BiVO 4 ( Figure 9h).
2D ultrathin nanosheets have ample unsaturated atoms as the defective sites, which are capable of promoting the light absorption, charge separation, and reactants adsorption. Although a number of 2D photocatalysts has been developed for CO 2 reduction, there is still a great difficulty to the synthesis of ultrathin nanosheets on the large scale. Especially, the controllable synthesis of 2D ultrathin photocatalysts with adjustable thickness is also challenging.
arrays, [244][245][246][247][248][249][250][251][252] have been widely developed. Due to their special features, including high specific surface area and high light utilization, they always demonstrate excellent photocatalytic CO 2 reduction performance. [197,[253][254][255][256] Wang et al. [257] reported the preparation of marigold-like SiC@MoS 2 nanoflower (Figure 10a-d). The high electron mobility allows an efficient multielectronic reduction reaction on the surface of SiC, resulting in CH 4 as a main reductive product. The overall conversion of CO 2 with H 2 O was achieved with the CH 4 and O 2 evolution rates of 323 and 621 μLg À1 h À1 in 40 h without sacrificial reagents under visible-light irradiation (λ ≥ 420 nm), respectively. In addition, SiC@MoS 2 also showed high stability during the CO 2 reduction reaction, because of the high chemical stability of MoS 2 nanosheets on SiC surface. In addition, Wang et al. [191] demonstrated the fabrication of In 2 S 3 -CdIn 2 S 4 heterostructure nanotubes for visible-light-driven CO 2 reduction (Figure 10e-i). Due to the nanosized interfacial contacts and unique microstructure, these hierarchical nanotubes showed many advantages, like enhanced charge separation and exposed rich active sites for CO 2 adsorption. The In 2 S 3 -CdIn 2 S 4 nanotubes showed high CO generation rate with a rate of 825 μmol h À1 g À1 under visible-light irradiation without cocatalysts.
In addition to the high specific surface area and rich reactive sites for adsorption, the 3D NWs array with desirable configuration have the benefits of promoted light harvesting and charge separation. Kim et al. [258] developed a ternary composite photocatalyst composed of Cu 2 O NW arrays, carbon layers, and BiVO 4 NPs for photocatalytic CO 2 conversion (Figure 11a).  [225] Copyright 2014, American Chemical Society. c) AFM image and d) the corresponding height profiles; e) SPV spectra and (inset) corresponding phase spectra, and the numbers from 1 to 3 in (c). f ) Positron lifetime spectrum. g) Schematic representation of trapped positrons. h) Photostability test for photocatalytic methanol evolution under 300 W Xe lamp irradiation for V v -rich and V v -poor o-BiVO 4 atomic layers. The error bars in (h) represent the standard deviations of three independent measurements of the same sample. Adapted with permission. [226] Copyright 2017, American Chemical Society.
The BiVO 4 /carbon-coated Cu 2 O NW arrays (BVO/C/Cu 2 O NWAs) showed irregular and rough surfaces with a diameter of %40 nm and length of a few micrometers (Figure 11b (Figure 11f ), in view of that CH 4 production is an eight-electron transfer reaction, much complex than CO formation (two-electron transfer reaction). Therefore, the construction of hierarchical nanostructures has proven to be effective strategies for photocatalytic CO 2 reduction. The nano-/microstructured photocatalysts with different dimensions for CO 2 reduction are shown in Table 2.

Porous Structure
Building a porous structure is a practicable strategy for enhancement of the photocatalytic performance. Microporous (pore diameters of less than 2 nm), mesoporous (pore diameters Figure 10. SiC@MoS 2 nanoflower: a) schematically synthetic process, b) SEM images, c) TEM images, d) EDX elemental mapping images, e,f ) FESEM and g,h) TEM images of hierarchical In 2 S 3 -CdIn 2 S 4-10 nanotubes, and i) elemental mappings of a single In 2 S 3 -CdIn 2 S 4-10 nanotube. a-d) Adapted with permission. [257] Copyright 2018, American Chemical Society. e-i) Adapted with permission. [191] Copyright 2017, American Chemical Society. between 2 and 50 nm in Figure 12a-c), and macroporous (pore diameters of greater than 50 nm) materials have been demonstrated to promote CO 2 adsorption, increase reactive sites, and enhance light absorption. [259] It was also reported that mesopores not only enhanced the light absorption, but also facilitated the charge separation and migration of semiconductors. [260] For example, the ordered mesopores extended the photoresponse of TiO 2 from UV to visible region (Figure 12d,e). [261] To explore detailed parameters in porous structure, porosity properties are usually characterized by nitrogen sorption isotherm analysis (Figure 12f,g). [262] A series of ordered mesoporous materials with high specific surface area (over 100 m 2 g À1 ), including TiO 2 , SnO 2 , ZnS, ZnSe, CdS, and CdSe, were synthesized for converting CO 2 into CO or CH 4 under light irradiation. [263] The CH 4 yield of mesoporous TiO 2 and SnO 2 was %10 times as much as that of P25, which is attributed to their high specific surface area. Mesoporous ZnS showed the highest CH 4 production rate of 3.620 μmol g À1 h À1 , while mesoporous CdSe showed the largest CO yield of 5.884 μmol g À1 h À1 (Figure 12h,i). Tailoring the porosity of photocatalysts is effective to improve the photocatalytic CO 2 reduction activity. [264] Feng et al. [265] first prepared www.advancedsciencenews.com www.small-structures.com porous TiO 2 by pyrolyzing MIL-125, and coated MgO overlayer on porous TiO 2 . Porous TiO 2 with five layers of MgO showed the optimal photocatalytic activity with a CO production rate of 13.5 μmol g À1 h À1 , which is 21 times as high as that of P25. The coating of MgO on TiO 2 reduced the recombination of photogenerated electrons and holes on the surface of TiO 2 . Similarly, mesoporous Cu 2 O covered by carbon layer realized the C 2 H 4 evolution from CO 2 reduction. [266] The thin carbon layer solved the photocorrosion of Cu 2 O, and the building of the mesoporous nanostructure promoted the adsorption of reactant molecules (Figure 12j,k) and charge carrier transfer. The optimized sample (C-2/Cu 2 O) showed an AQE of 2.07% for CH 4 and C 2 H 4 at λ ¼ 400 nm under visible light (Figure 12l,m).
To develop composite structures with wide-ranging pore dimensions, metals or metal oxides were loaded into the mesoporous silica. These new active sites introduced by heteroatoms led to improved adsorption of CO 2 for photoreduction. Tasbihi et al. [267] prepared Pt/TiO 2 -ordered mesoporous silica (TiO 2 -COK-12) for photocatalytic CH 4 and CO evolution. Under UV light in a continuous flow gas-phase photoreactor, CO was the major final product catalyzed by TiO 2 -COK-12. Because Pt as a cocatalyst was favorable for CH 4 formation as final product, loading the Pt/TiO 2 catalysts on COK-12 improved the activity of CH 4 with a selectivity of 100% in the reaction. Li and coworkers [268] synthesized mesoporous silica-supported Cu 2 O/TiO 2 nanocomposites for CO 2 photoreduction to CO and CH 4 . The large specific surface area of mesoporous silica (>300 m 2 g À1 ) not only greatly promoted CO 2 photoreduction and TiO 2 dispersion, but also realized the more feasible adsorption of reactants. In blank TiO 2 -SiO 2 system without Cu 2 O, CO was the primary product generated from CO 2 . The overall CO 2 conversion efficiency and the selectivity for CH 4 production were markedly increased by adding Cu 2 O. Cu species as an addition enhanced multielectrons reactions and suppressed the recombination of electrons and holes. Thus, 0.5%Cu/TiO 2 -SiO 2 displayed the peak production rates of CO and CH 4 reached 60 and 10 μmol g À1 h À1 , respectively. Heteroatom loading, immobilizing redox sites and tailoring hierarchical nanostructure have been also utilized for designing porous g-C 3 N 4 -based photocatalysts. The g-C 3 N 4 with a porous structure can facilitate the diffusion of CO 2 towards redox active sites. Wang et al. [269] prepared the 3D porous g-C 3 N 4 /C nanosheets composites, which achieved prominent CO 2 reduction. The CO and CH 4 yields were 229 and 112 μmol g À1 under Figure 12. Mesoporous single-crystal (MSC) synthesis of TiO 2 . a) Replication of the mesoscale pore structure within the templated region (FFT inset, 47 AE 3 nm sixfold symmetry) with crystal lattice vectors implied from the particle symmetry overlaid (reaction conditions 170 C, 40 mM TiF 4 ). b,c) Fully mesoporous TiO 2 crystals grown by seeded nucleation in the bulk of the silica template. Adapted with permission. [325] Copyright 2013, Nature Research. Representative TEM images along d) (100) and e) (110) planes of the ordered mesoporous black TiO 2 materials after hydrogen gas annealing at 500 C. Adapted with permission. [261] Copyright 2014, American Chemical Society. f ) Nitrogen sorption isotherms of polymers at 77 K. For clarity, the isotherms of CPOP-30-Re, CPOP-30'-Re, CPOP-31, and CPOP-31-Re are shifted vertically by 300, 500, 600, and 1000 cm 3 g À1 , respectively. g) Pore size distribution profiles for polymers calculated by nonlocal density functional theory. Adapted with permission. [262] Copyright 2019, American Chemical Society. h) Yields of CH 4 using various photocatalysts, P25 (commercial TiO 2 , Degussa, anatase þ rutile phases) and mesoporous materials (TiO 2 , SnO 2 , CdS, CdSe, ZnS, and ZnSe). i) Yields of CO using various photocatalysts, P25 (commercial TiO 2 , Degussa, anatase, and rutile phases) and mesoporous materials (TiO 2 , SnO 2 , CdS, CdSe, ZnS, and ZnSe). Adapted with permission. [263] Copyright 2018, Elsevier. j) N 2 adsorption-desorption isotherm curves, k) CO 2 adsorption isotherm curves, for pure Cu 2 O and C-2/Cu 2 O. Time-dependent product evolution over l) pure Cu 2 O and m) C-2/Cu 2 O. Adapted with permission. [266] Copyright 2016, American Chemical Society. 7 h simulated solar irradiation, both yields over 25 times higher than that of bulk g-C 3 N 4 .
Constructing porous structures are widely used in CO 2 photoreduction because of their excellent adsorption capability and abundant structural defects. Rationally tailoring the sharp and pore size of photocatalysts enables the remarkable improvement in light absorption and charge separation. [270][271][272] The coupling of single atom with nanoscaled metal oxides, mesoporous silica [65] or shape-controllable porous nanomaterials are appealing nanostructured photocatalysts for CO 2 reduction.

Crystalline and Band Structures Engineering
At present, various modifications on crystalline structure by tuning the compositions of semiconductors, such as element doping, can also change the electronic band structure, which results in the amelioration of the crucial factors during the photocatalytic process, thereby promoting the enhancement of CO 2 reduction performance. [273,274] For instance, constructing ordered mesoporous structure allowed TiO 2 improved light absorption, leading to enhanced photocatalytic CO 2 reduction activity, [263] and QD structure improved the charge separation efficiency of CsPbBr 3 . [206] Other modifications can either increase the adsorption of CO 2 , or adjust the final reductive products, so that the selectivity of CO 2 reduction be controlled.

Heteroatom Doping
Doping is always accompanied by the presence of vacancy defects in the crystal structure, which is due to the imbalance of charge or coordination in bulk structure. [275][276][277] The proper balance between doping and vacancies can extend visible-light absorption, reduce the charge recombination, and maintain the adsorption of CO 2 , thus contributing to the prominent performance for CO 2 photoreduction. On the contrary, excessive doping amount will throw off this balance and does not usually guarantee an increase in photocatalytic activities. [278,279] The doping of metallic atoms in photocatalysts is widely investigated. Pang et al. reported the monodispersed Ni-doped ZnS (ZnS:Ni) nanocrystals as excellent visible-light responsive photocatalysts for converting CO 2 into HCOOH. [280] The introduction of Ni resulted in the presence of sulfur vacancy, and proper concentration of doped Ni maintained the strong absorption and abundant sulfur vacancy, which can reduce the charge recombination in the bulk phase. Whereas, a heavy doping of Ni resulted in the diminishing of sulfur vacancies, losing active sites for CO 2 reduction. It was disclosed that the 0.1% doping of Ni as optimized ratio kept a good balance between the capacity of light absorption and the amount of sulfur vacancy. Consequently, over 95% selectivity of HCOOH as product and quantum efficiency of 59.1% at 340 nm and 5.6% at 420 nm were observed over Ni (0.1%)-doped ZnS nanocrystals. In addition to single element doping, the dual-elements doping was also conducted. A series of Cu/Zn doped/TiO 2 photocatalysts were developed for CH 4 evolution from CO 2 reduction. [281] Due to the Cu/Zn doping, the bandgap energy was reduced from 3.25 to 2.95 eV, and a strong CO 2 chemisorption was observed for 2%CuO-19% ZnO/TiO 2 , indicating the increased the amount of active sites.
The nonmetal doping also plays an important role in ameliorating the crystalline and band structure, which usually improves the light absorption and charge separation efficiency, contributing to the enhanced photocatalytic activity of CO 2 reduction. The boron carbon nitride (h-BCN) was explored for producing CO from CO 2 under visible-light illumination (Figure 13a). [282] In h-BCN structure, C was incorporated in the h-BN lattice, which affected the bonding energies B-K, N-K, and C-K (Figure 13b). The optical absorption edge of h-BCN showed obvious redshifts with the enriching of carbon (Figure 13c). Under visible-light irradiation for 2 h, the evolution of CO and H 2 increased to 9.3 and 2.9 μmol, respectively. Similarly, a carbon-doped SnS 2 (SnS 2 -C) nanostructured photocatalyst was synthesized by Shown et al. [283] It showed a high activity for CO 2 reduction into hydrocarbons with a quantum efficiency of over 0.7% under visible light. Because doped C induced microstrain in the SnS 2 lattice (Figure 13d,e), the electronic band structures and optical properties were largely altered. The observed maximum cumulative acetaldehyde yields after 13 h were 1256.6 mmol g À1 for SnS 2 -C and 5.5 mmol g À1 for SnS 2 , respectively ( Figure 13f ). The largest hydrocarbons yield of the SnS 2 -C (13.98 μmol/ 100 mg cat.) is almost 250 times as high as that of SnS 2 (0.055 μmol/100 mg cat.) (Figure 13g). g-C 3 N 4 nanotubes with O doping (OCN-tube) for replacing di-coordinated N atoms was developed for methanol evolution from CO 2 photoreduction. [284] The OCN-tubes were allowed a narrow bandgap for improved light harvesting, charge separation, and CO 2 adsorption. Thus, the methanol evolution of OCN-tube was 0.88 μmol g À1 h À1 , which is five times as high as that of bulk g-C 3 N 4 (0.17 μmol g À1 h À1 ). Similarly, g-C 3 N 4 with sulfur doping also showed an increased methanol evolution rate of 1.12 μmol g À1 h À1 , higher than that of pristine g-C 3 N 4 (0.81 μmol g À1 h À1 ). [285] The S doping resulted in the formation of impurity state, which was beneficial to the transfer of photogenerated electrons to CB for CO 2 reduction.
Elements are commonly doped into bulk structure either by replacing the lattice atoms or by occupying the interstitial sites. With optimizing the doping with diverse elements, high performance with different final products in CO 2 reduction can be achieved. Various doping sites and concentrations are closely associated with the light absorption, charge carrier separation, and CO 2 chemisorption, which is vital for high activity and selectively of CO 2 reduction. But it still remains to be a significant challenge in ensuring a balance between the proportion of doping element and vacancy defects in bulks.

Solid Solution
Introducing proper elements or components to construct solid solution semiconductors through precise and wide-range replacement are important route for crystalline structure regulation. [286] Ions radius, chemical valance, and crystallographic parameters are basically three vital keys to design solid solution structure and to minimize formation energy. As the composition of semiconductors is associated with the band structure and electron-hole mobility, the introduced elements or components can generally tune the band structure and optoelectronic properties of semiconductors, which effectively adjust bandgap, band edge levels, and transport of photogenerated charge carriers. [287,288] Halogens are good candidates for composition regulation to form solid solution. Bai et al. [289] [282] Copyright 2015, Nature Research. d) High-resolution XPS Sn 3d spectra of SnS 2 -C and SnS 2 . e) High-resolution XPS S 2p spectra of SnS 2 -C and SnS 2 . Comparative photocatalytic CO 2 reduction activity of SnS 2 -C and SnS 2 . f ) Cumulative acetaldehyde formation yield of SnS 2 -C and SnS 2 . g) Comparative solar fuel formation rate and quantum efficiency of SnS 2 -C, SnS 2 , and commercial SnS 2 under a visible-light source (300 W halogen lamp). Adapted with permission. [283] Copyright 2018, Nature Research.  [290] reported the preparation of the nanoflower structured BiOBr x Cl 1-x solid solution ( Figure 14a). The construction of BiOBr x Cl 1-x solid solution optimized the bandgap and facilitated the photogenerated charges separation. Under simulated solar light irradiation, the CO generation rate of BiOBr 0.6 Cl 0.4 (15.86 μmol g À1 h À1 ) was roughly 7.5 and 10.2 times as high as that of single BiOBr (1.55 μmol g À1 h À1 ) and BiOCl (2.11 μmol g À1 h À1 ), respectively. In addition, the solid solution can also be fabricated by cationic substitution in the whole composition range. Cadmium and zinc sulfide solid solution (Cd 1-x Zn x S x ¼ 0.2, 0.4, 0.6, 0.8, 1) were successfully constructed for CO 2 reduction under visible light (λ ¼ 450 nm) in gas phase. [291] With the increase in cadmium concentration, the light harvesting of Cd x Zn 1-x S samples are orderly extended to visible-light region (Figure 14b,c). The prepared Cd 1-x Zn x S solid solution produced CO as the major product in CO 2 reduction process ( Figure 14d,e), with optimal ratio of Cd 0.94 Zn 0.06 S showing the highest activity of CO evolution with a rate of 2.9 μmol g À1 h À1 and a high selectivity of 95%. The replacement of components in similar element composition was also carried out to form solid solution. Yan et al. [292] synthesized ZnGa 2 O 4 /Zn 2 GeO 4 solid solution. Construction of this solid solutions resulted in not only narrowed bandgap, upshifted VB position (O 2p-Zn 3d) and downshifted CB position (Figure 14f,g), but also reduced hole effective mass, which was beneficial to promoting holes transfer. So, the ZnGa 2 O 4 / Zn 2 GeO 4 solid solution exhibited improved water oxidation and protons mobility for CO 2 photoreduction. As a result, the optimal mole ratio of ZnGa 2 O 4 to Zn 2 GeO 4 is 4.5:1 and the CH 4 yield of this optimal sample (0.5 μmol in the first hour) is about 33 times as high as that of ZnGa 2 O 4 (0.015 μmol h À1 ). Liang and Li [293] synthesized the Zn 1.231 Ge 0.689 N 1.218 O 0.782 solid solution wrapped by ultrathin N-doped graphene. As an electron reservoir, N-graphene accepted electrons from ZnGeON solid solution and thus benefited the separation of electron-hole pairs for enhancing the CH 4 formation.
Other abundant ternary systems, such as InGaN [294,295] and CdSSe, [296,297] and some quaternary systems including ZnCdSSe [297] or GaN-ZnO [298] are also potential photocatalysts in CO 2 reduction. Notably, solid solution semiconductors with delicate microstructures, including QDs, NWs, nanosheets, have more advantages than the corresponding bulk materials. In some solid solution systems, defects can be generated due to the different lattice constant, and more defects will directly impair the optical or electrical properties of semiconductors. Especially for the volatile elements, such like Zn, As, a large amount of vacancies are easy to be formed in their solid solutions. [299] Thus, it is necessary to control the synthetic parameters, such as the reaction temperature and annealing process, to tune the concentration of vacancies.
Solid solution as homogenous phase has a uniform structure and composition distribution. According to Hume-Rothery rules, only if solute and solvent atoms have similar radius and valence of iron, these substitutional solid solutions would be successfully synthesized rather than divided into several phases. But heteroatom doping usually has a boarded span of element choices and only focuses on specific areas. These extra dopants mainly affect the band structure and tune the electronic and surface properties of materials. Recently, it has been reported that the sublattice doping enables the fuzzy boundary of solid solution and heteroatom doping, [300,301] which may provide a novel avenue to tune crystalline and band structures of 2D materials, such as graphene, MoS 2 .

Surface Structure Engineering
After the photogenerated charges migrate from the bulk to the reactive sites on the surface, they are consumed by the adsorbed reactants for completing the potential redox reactions. Surface structure engineering with suitable strategies can boost the CO 2 or H 2 O adsorption on the photocatalysts and finally activate these molecules to efficiently react with electrons or holes on the surface for CO 2 reduction. [302][303][304] In other words, the performance and final products of CO 2 photoreduction are greatly determined by the surface structure that involves surface compositions, acidity, specific surface areas, pores, etc. The strategies for surface structure engineering of semiconductor photocatalysts can be divided into crystal facets manipulation and surface structure modification.

Exposed Reactive Facet
In general, facets have an influence on the photocatalysis in two aspects. First, surface atomic arrangements of specific facet have effect on the molecular adsorption and intermediate products. [305][306][307][308] Namely, the conversion and final products of CO 2 reduction can be tuned by exposed facets. Second, the band bending may occur on the surface, and diverse exposed facets are link to the redox abilities of charge carriers in reactions. [309][310][311] In addition, the charge separation efficiency is also related to crystal orientations, as different facets has diverse atomic composition with varied charge densities.
The integration of exposed facets and noble metal loading are demonstrated as a practicable strategy to enhance photoreduction of CO 2 . For example, the tunable performance of CO 2 to CH 4 on anatase TiO 2 -{001} facets (TiO 2 -001) and TiO 2 -{010} facets (TiO 2 -001) was achieved by loading platinum. Without Pt loading, the pristine TiO 2 -010 had a higher performance than pristine TiO 2 -001, because TiO 2 -010 was in more favorable of CO 2 adsorption and prolonging the lifetime of electrons and holes. After loading 1 wt% Pt, the charge lifetime of TiO 2 -001 showed an obvious increase from 0.65 to 0.84 ns, but that of TiO 2 -010 loaded with 1 wt% Pt only increased by 0.02 ns. The CH 4 production rate of TiO 2 -001 with 1 wt% Pt was as twice as that of TiO 2 -010 with 1 wt% Pt. Therefore, it was proposed to produce a synergy effect between Pt-loading and certain exposing facets of TiO 2 for extending charge lifetime and promoting CH 4 production.
Similarly, the noble metal loading on specific exposing facets of BiVO 4 was also reported to extend the lifetime and transfer route of charge carriers. The Au photodeposition over BiVO 4

Facets Junction
Different facets with diverse atomic compositions show distinct functions for accumulating the photogenerated electrons or holes. Herein, exposure of multiple facets in semiconductors to construct a facet junction can propel the electrons or holes to migrate to respective facets, achieving efficient spatial charge separation on the surface. [312] Meanwhile, it is also desirable that the separated electrons and holes can separately participate in the reduction and oxidation, helpful to the high photocatalytic CO 2 reduction. [313][314][315] Anatase TiO 2 single crystal is an ideal model for exposing multiple facets. It was reported that the {100} and {101} facets have high reducing ability, and the strong oxidizing ability was observed for {001} facet in TiO 2 single crystal. [35] Facets junction in coexposed {001} and {101} facets were developed by Cao et al. in anatase TiO 2 nanocrystals (10-30 nm); the ratio of {001} facets can be adjusted from 5% to 51%. [316] Due to the coexposed facets promoting charge separating, the highest performance for CH 4 producing from CO 2 was observed in octahedral bipyramid TiO 2 nanocrystals (TiO 2 -0.2HF) with 51% {001} and 49% {101} facets, with a CH 4 evolution rate of 1.58 molh À1 g À1 . A facets junction built by {001} and {101} facets was developed by Yu et al. [308] It was theoretically predicted the electrons were more prone to move to {101} facets, and holes migrated to {001} facets under light irradiation. Meanwhile, the yield of the final products was greatly affected by the facets ratio in this facet junction, and the best ratio of {101}/{001}facets was determined to be 45:55 for CH 4 evolution. Similar facet-selective charge separation behaviors were also discovered for BiVO 4 single crystals. Li et al. [317] fabricated the {010}/{110} facets junction over monoclinic BiVO 4 crystal for efficient charge separation. Under solar illumination, the reduction reaction occurred on {010} facets, whereas the oxidation reaction took place on {110} facets. For confirmation, the photodeposition of metals and metal oxides were conducted. As shown in Figure 16a For a thin-layered facets junction photocatalyst, the thickness and ratio of exposing facets should be considered simultaneously.
Our group reported the preparation of thickness-tunable BiOIO 3 nanoplates with {010}/{100} facet junctions by controlling the reaction conditions. [102] Reducing the thickness along [010] direction promoted the charge carriers diffusion from bulk to the surface, and a suitable thickness assured the rational distribution of electrons on {010} facets and holes on {100} facets, realizing the efficient spatial separation of charge carriers (Figure 16g-k). It was confirmed by the observation for facet-selective photodeposition of Pt and MnO x . As a result, BiOIO 3 nanoplates with optimal thickness exhibited the highest CO evolution rate (5.42 μmol g À1 h À1 ), %3 times higher than that of pristine BiOIO 3 (1.77 μmol g À1 h À1 ). Generally, the transfer of charge from semiconductors to metals is much faster (<20 fs) than that from metals to semiconductors (>20 fs). [123] Noble metals can be deposited on the active crystal facets to further improve the charge separation efficiency and enhance CO 2 photoreduction activity. Li et al. [318] developed brookite TiO 2 nanocubes with dominantly exposed four {210} and two {001} facets. The amount of Ag loading and the distribution of Ag NP exerted a significant influence on the performance of CO 2 photoreduction. The 0.5% Ag-TiO 2 showed the highest CO production rate (128.8 ppm h À1 ), %5.41 times as high as that of the pristine TiO 2 (23.8 ppm h À1 ). The 1% Ag-TiO 2 had the largest yield of CH 4 (28.8 ppm h À1 ), about 3.80 times more than that of the pristine TiO 2 (7.56 μmol g À1 h À1 ). With a low Ag mass loading level (≤0.5%), the majority of Ag NPs were dispersed on the {210} facets, which was conductive to producing CO from CO 2 reduction. With a larger Ag-loading amount (>0.5%), Ag NPs tended to aggregated on {210} facets and dispersed on {001} facets, which benefited the CH 4 generation. These results show that the Ag loading on different facets exerts great effects on the activity and final products of CO 2 reaction. In anatase TiO 2 nanocrystals, the photogenerated electrons are likely accumulated on {101} and holes tend to move onto {001} facets. [319] When Pt-Ru NPs were selectively deposited on TiO 2 -{101} facet, it promoted the electron transfer to Pt-Ru from TiO 2 . Thus, the CH 4 generation rate of PtRu/TiO 2 increased to 38.7 μmol g À1 h À1 under simulated solar irradiation, which is about 29 times higher than that of P25. Meanwhile, PtRu/TiO 2 exhibited a high selectivity of CH 4 (93.7%) with an AQE of 0.98%.
The report of synthesis TiO 2 with different exposed facets always precedes the study of their photocatalytic properties in CO 2 reduction. Thus, some new trends in CO 2 reduction can be proposed from the development of TiO 2 with new facets. Basically, anatase-TiO 2 {001} planes showed a high surface energy of 0.90 Jm À2 and the dissociative adsorption of water molecules might be associated with the {001} surface by some theoretical predictions. [320] In 2008, Yang et al. [321] first reported that anatase TiO 2 single crystals exposed with two dominant {001} surfaces and eight {101} surfaces (Figure 17a). Anatase TiO 2 with {110} facets is very difficult to be synthesized because of the much higher surface energy of {110} facets (1.09 J m À2 ) than that of {001} facets. In 2010, Liu et al. [322] first reported the preparation of anatase TiO 2 crystals with major {101} and {001} facets and minor {110} facets (Figure 17b,c). After the successful synthesis of TiO 2 crystals with diverse facets, the major task will www.advancedsciencenews.com www.small-structures.com be tuning the ratio of different facets and combining of two or more facets to form facet junctions. Single crystals with specific facets have advantages of reactants adsorption and long-range electronic connectivity for CO 2 reduction. In addition, mesoporous structure is equally important in CO 2 photoreduction for enhancing light absorption and increasing potential active sites. The integration of constructing single crystalline and mesoporous in TiO 2 crystals may be a solution to fulfill both demands. [51,323,324] Crossland et al. [325] developed specific faceted and mesoporous single crystals of anatase TiO 2 (Figure 17d-g). It was revealed that the specific surface area of the samples was 70 m 2 g À1 , which was very similar to those of 20 nm NPs. The integration of single crystal and mesoporous structure for photocatalytic CO 2 reduction are pending research.

Surface Acid-Base Sites and Vacancy
Acidic sites on the surface of photocatalysts are beneficial for CO 2 reduction as well as for specific molecular bonding, which depends on their adequate separation of photogenerated charge carriers. [326] Proper surface acid sites grown on the Nb 2 O 5 photocatalyst can control the activity and selectivity of CO 2 conversion. [327] It was proved that CO and HCOOH were produced under high surface acidity, while low surface acidity tended to generate CH 4 . In addition, the H 2 SO 4 -modified TiO 2 nanosheets were developed, which showed high photocatalytic performance for CO 2 reduction to CH 4 under visible-light irradiation. [328] TiO 2 nanosheets with 0.5 mol L À1 H 2 SO 4 showed the highest CH 4 yield of 13.20 μmol g À1 in 4 h, and the TON and quantum yield of CH 4 evolution were 83.124 and 0.726‰, respectively. The prominent catalytic activity was attributed to that the surface protonation derived from acidification with H 2 SO 4 extended the lifetime of photogenerated electrons and holes. Due to the acidic character of CO 2 , basic sites are more likely for adsorption and further to react with CO 2 by electron transfer. When acid and base sites are both built on the surface, they can cooperatively promote the CO 2 reduction. Dong et al. [77] demonstrated that the Lewis acidity and basicity sites were simultaneously constructed on the surface of In 2 O 3-x (OH) y ) with extra In 3þ with Bi 3þ . The bismuth ion hybridized with oxygen causing an increased Lewis basicity. Bismuth ion also is an acid site to promote hydrogen splitting for efficient CO 2 reduction. All the samples showed decent stability in 12 h (Figure 18a), and Bi-0.03% exhibited the highest activity for CO evolution with a rate of 1.32 μmol g À1 h À1 (Figure 18b).
A strong interaction between OVs and H 2 O molecules have been revealed by scanning tunnel electron microscope (STEM). Meanwhile, CO 2 molecules adsorbed at surface OVs were in favor of the CO 2 reduction into CO. Di et al. [329] reported Bi 12 O 17 Cl 2 superfine nanotubes with surface oxygen defects, leading to accelerated charge carriers migration and facile CO 2 activation. In liquid system without cocatalyst and sacrificing reagents, Bi 12 O 17 Cl 2 nanotubes delivered high selectivity toward CO with an evolution rate of 48.6 mmol g À1 h À1 (16.8 times than of bulk Bi 12 O 17 Cl 2 ), and meanwhile maintained high stability after 12 h of testing. In addition, OV defects on the surface lowered the energy barrier of the CO 2 reduction reaction by activating CO 2 or H 2 O, and the simultaneous activation of H 2 O and CO 2 molecules for matching two half reactions may be achieved by introducing OV defects on  nanotubes. Lu et al. [330] developed LaTiO 2 N modified by OVs and La 2 O 3 . Electron paramagnetic resonance (EPR) signal at g ¼ 2.003 was found in all the as-prepared samples, assigning to unpaired electrons trapped by OVs (Figure 18c). Two peaks located at 457.4 and 456.3 eV in X-ray photoelectron spectroscopy (XPS) (Figure 18d), in good agreement with Ti 4þ and Ti 3þ species, respectively. Meanwhile, the peak at 531.2 eV belonged to lattice oxygen atoms in the vicinity of OVs (Figure 18e). Due to O 2À in La 2 O 3 as basic sites, the adsorbed CO 2 tended to be transformed into CO 3 2À species, which was greatly conductive to bending O─C─O bond and decreasing LUMO energy of CO 2 molecules. The OVs of LaTiO 2 N were beneficial in activating H 2 O molecules, which promoted the reaction of water oxidation (Figure 18f,i). As the CO 2 and H 2 O were activated at separate active sites (Figure 18f), the CH 4 yield was increased to 0.3 μmol g À1 h À1 (Figure 18g,h).

Surface Modification
The surface modification on semiconductor photocatalysts by complexes or ions can not only extended their light absorption range to visible region or even the infrared (IR) regions, but also promoted the surface charge separation. Consequently, the final products in reactions may vary greatly by surface modification. [331][332][333] Wang et al. [334] reported that the modification of hydroxylated commercial TiO 2 (P25m) with Cu(II) tetra(4-carboxylphenyl)porphyrin (CuTCPP) as sensitizer, which allowed a broader spectrum of light absorption and higher separation efficiency than pristine P25m. In this case, 0.5%CuTCPP/P25m generated almost 46 times as high CH 4 as P25m alone under UV-vis irradiation. Wang et al. [272] reported that porous hypercrosslinked polymer-TiO 2 -functionalized graphene (HCPs-TiO 2 -FG) composite structure was prepared. The integration of HCPs and FG endow TiO 2 with large improvement on photoabsorption, charge separation efficiency and CO 2 adsorption as well as CO 2 diffusion (Figure 19a). Specifically, the visible-light absorption and charge separation efficiency of TiO 2 photocatalyst were improved by the graphene and the HCPs layers. HCPs layers also improved the specific surface area and increased micropore volume to 988 m 2 g À1 and 0.306 cm 3 g À1 , respectively. As a result, this composite showed high photocatalytic CO 2 reduction performance, casting a CH 4 production rate of 27.62 μmol g À1 h À1 and selectivity of 83.7%, without sacrificial reagents and cocatalysts.  Metal complexes have been widely used to achieve the surface modification on different photocatalysts. Mesoporous C 3 N 4 coupled with a Ru(II) binuclear complex (RuRu 0 ) was able to selectively reduce CO 2 into HCOOH under visible light (λ > 400 nm). [335] Integrating C 3 N 4 with RuRu 0 produced a new absorption band located at 460 nm, and an increase in loading amount of metallic Ag was accompanied by a new absorption band at 500-800 nm. Thus, RuRu 0 /Ag/C 3 N 4 photocatalyst exhibited an extremely high TON (>33 000 with respect to the amount of RuRu 0 ), while maintaining high selectivity (87-99%) for HCOOH production (Figure 19b-e). Kuriki et al. [336] explored the surface modification of CNNS rather than bulk C 3 N 4 with RuRu 0 and with Ag NPs for CO 2 photoreduction to produce HCOOH in aqueous media. It also showed high selectivity (%98%) and high TON beyond 2000. With the same Ru(II) binuclear complex, a layered perovskite oxynitride of Li 2 LaTa 2 O 6 N with RuRu 0 was synthesized for visible-light CO 2 reduction. [337] RuRu 0 modified Li 2 LaTa 2 O 6 N had a visible-light absorption (wavelength > 400 nm). The selectivity and TON for CO 2 reduction into HCOOH were over 97% and 50, respectively.  To explore the photoinduced charge transfer mechanism in surface organics (electronic donor)-catalyst-acceptor system, a mononuclear C 5 H 5 -RuH complex oxo-bridged TiO 2 hybrid was developed by Huang et al. [338] (Figure 20a,b). C 5 H 5 -RuH served as the photon harvester and water-oxidation sites, whereas TiO 2 played the role of electron collector and CO 2 reduction sites. The fast electron injection from the excited Ru 2þ cation to TiO 2 (%0.5 ps) and the slow backward charge recombination in half life (%9.8 μs) resulted in a substantially boosted charge separation along the donor-catalyst-acceptor pathway (Figure 20c,d), thus giving rise to a quantum efficiency of 0.56% for CH 4 production under visible-light irradiation. To pursue a nonprecious metal system with low cost, CdS QDs as photocatalysts modified with nickel terpyridine complexes were developed by Kuehnel et al. [91] The anchoring of [Ni(terpy) 2 ] 2þ complexes allowed CdS QDs a high average external quantum efficiency of CO production (with over 90% selectivity) of 0.28 AE 0.04% in aqueous solution under 400 nm light (1.5 mW cm À2 ). In addition, surface decoration with inorganic ions also shows large potential. Our group reported that surface halogenation is a desirable strategy to increase the CO 2 reduction activity of Bi 2 O 2 (OH)(NO 3 ) (BON). The halide ions (Cl À , Br À , I À ) were found to bonded with Bi atoms by replacing surface hydroxyls, which consumedly promoted the local charge separation. Furthermore, the surface grafted halide ions activated the surrounding hydroxyls to boost the adsorption of CO 2 molecules and protons for facilitating the CO 2 conversion. [178] Without any sacrificial agents or cocatalysts, Br À -modified BON showed the highest CO production rate (8.12 μmol g À1 h À1 ) among BON-X (X ¼ Cl, Br, and I) samples, which was %73 times as high as pristine BON (Figure 20e,f ).
Different surface modification methods, including regulation of acid and basic sites, introduction vacancies or grafting metal complexes and inorganic ions, were developed to tune to surface  [338] Copyright 2016, Wiley-VCH. e) apparent rate of CO yields by the photoreduction of CO 2 over BON-Xa (X ¼ Cl, Br, and I; a ¼ 1-4) samples under simulated solar light irradiation. f ) In situ DRIFTS for CO 2 photoreduction on BON-Br 3 for different times. Adapted with permission. [178] Copyright 2019, Wiley-VCH.
www.advancedsciencenews.com www.small-structures.com structure of photocatalysts, which play critical roles in enhancing light absorption and charge separation, increasing active sites, and lowering the activation energy of CO 2 or H 2 O molecules, drastically optimizing the activity of CO 2 reduction and selectivity of products.

Interfacial Structure in Heterojunction
Heterojunctions includes the metal-semiconductor junctions, such as Schottky junction, and semiconductor-semiconductor junctions. In a band alignment of two semiconductors, there are five kinds of heterojunctions that can be built, as shown in Figure 21a-e. Clearly, the type I-1 as a n/n junction has superior holes separation character, and type I-2 as a p/n junction tends to realize the separation of photogenerated carriers. The type II-1 or direct Z-scheme junction is n/n junction, in which the holes with weak oxidizing capacity recombine with electrons from neighboring semiconductor, whereas holes with strong oxidizing capacity and electrons with strong reducing capacity are kept for the following surface reactions. The type II-2 or called p-n junction can also separate electron-hole pairs by hindering their recombination, but holes and electrons with weak oxidation or reduction are saved. [35,[339][340][341] The interface is an important parameter in heterojunction system. Interfaces between two adjacent components play a vital role in charge carriers transfer, because the structure of interface impacts electron-hole pair migration from light-harvesting component to surface active sites. [342][343][344][345] As a typical metal-semiconductor junction and one of the most desirable semiconductor-semiconductor junctions, Schottky junction and Z-scheme junction will be discussed here.

Interface in Schottky Junction
In a combination of the metal and n-type semiconductor, the electrons present the majority of charge carriers, and their work function are ϕ m and ϕ s , respectively. If ϕ m > ϕ s , the electrons in semiconductors with high Fermi level are tend to transfer to the metal with low Fermi level for achieving the equilibrium ( Figure 22). The Helmholtz double layer is created at the interface of metal (negatively charged) and semiconductor (positively charged). Free charge carriers in the semiconductor usually show a low concentration. The band edges are shifted continuously originating from the presence of depletion and accumulation layers, which is called band bending, and thus it takes the energy bands upward bend. On the contrary, the band edges bend downward if ϕ m < ϕ s . The work function of metal minus that of semiconductor equal to the degree of band bending of semiconductor at the interface, as shown by Equation (2) Only if an n-type semiconductor with ϕ m > ϕ s , the Schottky barrier (ϕ SB ) can be established at the metal-semiconductor interface. This barrier transfer electrons from semiconductor to metal and obstruct backflow, causing a high photocatalytic performance. [346][347][348][349][350] Lu et al. [76] established a Schottky junction in KOH-modified Ni/LaTiO 2 N for CO 2 photoreduction under visible light. The yield of KOH-modified Ni/LaTiO 2 N was 9.69 μmol g À1 for CH 4 and 0.31 μmol g À1 for CO, which reached 5 times as high as those of LaTiO 2 N. In the Ni/LaTiO 2 N Schottky junction, nickel has a large work function of 5.2 eV, which enables an upward band bending in LaTiO 2 N close to the Ni/LaTiO 2 N interface. The large Schottky barrier allows the formation of a built-in electric field, greatly boosting the charge separation ( Figure 22). In addition, the OH À from KOH on the surface activated CO 2 to transform into CO 3 2À species, and enhanced the kinetics of CO 2 reduction. Chen et al. [351] constructed a Schottky junction in TiO 2 /Pt on the titanium substrate (Figure 23a,b). Under equilibrium, a Schottky barrier has been established at the TiO 2 and Pt interface. As Pt has a higher work function than TiO 2 , the electrons were prone to migrate from TiO 2 to Pt, and at the Figure 21. Schematic illustration for photogenerated charge transfer process in five types of heterojunctions with internal electric field: a) type I-1 (n/n junction), b) type I-2 (p/n junction), c) type II-1/direct Z-scheme (n/n junction), d) type II-2 (p-n junction), and e) type III-1. Reproduced with permission. [35] Copyright 2019, American Chemical Society.  Figure 22. Energy band diagrams of metal and n-type semiconductor contacts. E vac , vacuum energy; E c , energy of CB minimum; E v , energy of VB maximum; ϕ m , metal work function; ϕ s , semiconductor work function; χ s , electron affinity of the semiconductor. Reproduced with permission. [346] Copyright 2012, American Chemical Society.   (Figure 23d-f ). The encapsulation of metal NPs for Schottky junctions has been developed to improve the transfer of electrons and holes. Interestingly, it was highlighted that the integration of plasmonic effect and Schottky junctions has the potential to enhance visiblelight harvesting and photoinduced charge carries separation for high photocatalytic activity. [352][353][354][355] For example, copper with plasmonic effect was encapsulated in UiO-66 to form Schottky junctions (Cu/Cu@UiO-66). [356] The photocatalytic performance of Cu/Cu@UiO-66 was enhanced by encapsulating a very low concentration of Cu, which attributed to a synergy of plasmonic effect and Schottky junctions to allow the visible-light absorption and electron capture. Also, the plasmonic Au was combined with Pt-MOF Schottky junction to develop a Pt@MOF/Au system. [357] In this ternary photocatalyst, Pt NPs were well dispersed into or on MIL-125, followed by assembly with Au nanorods. As a result, both high visible-light harvesting and efficient charge transfer were achieved in Pt@MOF/Au system, attributing to the Au plasmonic resonance and Pt-MOF Schottky barrier.

Interface in Z-Scheme Junction System
Z-scheme junction integrating dual photocatalysts and interface is an efficient way to mimic photosynthesis in nature, which contains two parts in plants, photosystem (PS) I and II. Electrons are produced in PS I and consumed in the CO 2 reduction, whereas the holes are produced in PS II and tend to generate O 2 from H 2 O oxidation (Figure 24a,b). The fabrication of Z-scheme can prolong the charge carriers lifespan and meanwhile maintain the strong redox capabilities. [358][359][360][361][362] For example, Li 2 LaTa 2 O 6 N has an advantage of proper band structure for water oxidation and CO 2 reduction, but it suffers from low product selectivity Figure 24. Comparison of a) biological photosynthesis and b) dual n-type semiconductor model, representing the electron transport chain and redox potentials. Adapted with permission. [363] Copyright 1979, Elsevier. Schematic charge-transfer mechanism in three kinds of Z-scheme photocatalytic systems: c) liquid-phase, d) all-solid-state, and e) direct Z-scheme. Adapted with permission. [35] Copyright 2019, American Chemical Society. from the various CO 2 reduction pathways. [335,337] The Ru(II) binuclear complexes with excellent visible-light absorption have a strong ability to reduce CO 2 to HCOOH with a high TON. Due to the low oxidation capability of Ru(II) binuclear complexes, an electron donor is necessary for keeping reasonable stability and conversion rate during the CO 2 photoreduction. [335] Thus, Ru(II) binuclear complexes as reduction part couples an oxidation catalyst Li 2 LaTa 2 O 6 N to build a delicate Z-scheme system, which solved the aforementioned problems for a high yield of HCOOH with a high stability. So far, there are mainly three types of Z-scheme systems reported in photocatalysis, namely, liquidphase Z-scheme junction, all-solid-state Z-scheme junction with a mediator, and direct Z-scheme junction (Figure 24c-e). [35] 6.2.1. Liquid-Phase Z-Scheme Junction In 1979, liquid-phase Z-scheme photocatalytic system in semiconductors was first proposed by mimicking photosynthesis in plants (Figure 24b). [363] The semiconductor-solution interface controls the charge transfer processes, which can reduce the recombination photogenerated charges to improve the stability and activity of photocatalysts. In 2001, a liquid-phase Z-scheme photocatalytic system was established with Pt@anatase-TiO 2 and primary rutile-TiO 2 in the existence of IO 3 /I as the shuttle redox mediator for overall water splitting. [364] In addition, liquid-phase Z-scheme system of BiOI/g-C 3 N 4 was constructed to promote charge transfer by the intermediate I 3 À ions for the application of CO 2 photoreduction. [365] The I À ions are oxidized to I 3À ions by the holes from BiOI and subsequently, the I 3À ions are reduced by electrons of g-C 3 N 4 , resulting in a full cycle of the I 3À /I À redox mediator. Note that the liquid-phase Z-scheme systems with suitable solution usually have to tackle a potential reaction in the backward direction. For instance, the reduction of oxidative mediator (I 3À ions) competes with CO 2 reduction. To address this drawback, all-solid-state Z-scheme junction with an electron mediator were developed.

All-Solid-State Z-Scheme Junction with Mediator
In all-solid-state Z-scheme junction with mediator, the photogenerated electrons and holes move through a solid-state mediator (Figure 25a,b), which can separate and collect charge carriers from the semiconductors. Thus, the appropriate mediator with interface modification can optimize the interfacial energy of semiconductor-solid-state mediator and boost the electron transfer from semiconductor. [366][367][368][369][370][371] Developing solid-solid interfaces in Z-scheme system provides special opportunities to create new features by combing the distinctive characteristics of different semiconductors. In 2006, Tada et al. [372] reported the preparation of CdS-Au-TiO 2 composite photocatalyst as an all-solid-state Z-scheme system (Figure 25a), in which CdS and TiO 2 , respectively, served as PS I and PS II, and Au as mediator was grown at the interface between CdS and TiO 2 . Thus, the photogenerated electrons from TiO 2 can cross Au to reach CdS to recombine with the holes of the latter for keeping the highly reductive electrons on the CB of CdS and highly oxidative holes on the VB of TiO 2 , leading to the excellent photocatalytic activity of Au@CdS/TiO 2 (Figure 25e,f ).
Similarly, RGO with super conductivity was reported to use to as an ideal mediator construct Fe 2 V 4 O 13 /RGO/CdS all-solid-state Z-scheme junction for photocatalytic conversion of CO 2 into CH 4 (Figure 25b-d). [373] The RGO interlayer establishes a high-speed pathway with high electronic mobility for charge transfer between Fe 2 V 4 O 13 and CdS in Fe 2 V 4 O 13 /RGO/CdS (Figure 25c), which thereby rendered a 30% increase in CO 2 reduction for CH 4 production in comparison with Fe 2 V 4 O 13 / CdS (Figure 25d).

Direct Z-Scheme Junction
Direct Z-scheme system as a new generation of junction photocatalyst was developed, which exhibits obvious advantages than the previous Z-scheme system, such as low cost and convenient preparation. [374][375][376][377][378][379] For example, porous g-C 3 N 4 /Sn 2 S 3 -diethylenetriamine (Pg-C 3 N 4 / Sn 2 S 3 -DETA) composite as a direct Z-scheme system no electron mediator was constructed. [380] It displayed high production rates of CH 4 (4.84 μmol g À1 h À1 ) and CH 3 OH (1.35 μmol g À1 h À1 ), which was attributed to the formation of direct Z-scheme between Pg-C 3 N 4 /Sn 2 S 3 -DETA, promoting photogenerated charge separation. To improve the interfacial structure in Z-scheme junction, Al-O bridges were introduced in the g-C 3 N 4 /α-Fe 2 O 3 Z-scheme system by Wang et al. [381] to reinforce the electrons transfer. The photocatalytic activities of this Z-scheme junction with Al-O bridges are higher than that without Al-O bridges in CO 2 photoreduction. This performance enhancement is resulted from the promoted charge transfer and separation provided by Al-O bridges in g-C 3 N 4 /α-Fe 2 O 3 . Further, a ternary Z-scheme photocatalyst Ag 2 CrO 4 /g-C 3 N 4 /GO was developed. [382] In this direct Z-scheme junction, improved light absorption, redox ability, and enhanced charge separation were achieved in Ag 2 CrO 4 / g-C 3 N 4 , and meanwhile GO served as a cocatalyst further promote the charge separation and the CO 2 adsorption. Ag 2 CrO 4 / g-C 3 N 4 /GO showed a high CH 3 OH and CH 4 production with a turnover frequency (TOF) of 0.30 h À1 under simulated sunlight irradiation, which was 2.3 times higher than that of pristine g-C 3 N 4 . It may be concluded that fabrication of ternary direct Z-scheme junction may show a larger superiority.
In contrast, the morphology control, such as 1D nanobelts, [383][384][385][386][387] or 2D nanosheets, [222,[388][389][390] was introduced to enhance the photocatalytic activity of Z-scheme junction photocatalysts. To build a unique 3D morphology, Yang et al. [227] assembled 2D ZnIn 2 S 4 nanosheets onto 1D TiO 2 nanobelts to form a direct Z-scheme junction ZnIn 2 S 4 /TiO 2 . This 3D structure showed large surface area and high efficiency of charge separation where the electrons were collected in the CB of ZnIn 2 S 4 , whereas holes were accumulated in the VB of TiO 2 , leading to an enhanced activity for CO 2 photoreduction. The evolution rate of CH 4 of 3D ZnIn 2 S 4 /TiO 2 (1.135 molg À1 h À1 ) was around 39 times as high as that of bare ZnIn 2 S 4 (0.029 molg À1 h À1 ). Hexagonal WO 3 nanosheets with dominant {001} facets (WO 3 -001) are more effective for H 2 O oxidation, but the CB edge potential cannot meet the demand of CO 2 photoreduction. In regard to this issue, Shi et al. [391] assembled Cu 2 O on WO 3 -001 to construct the Cu 2 O/WO 3 Z-scheme junction, which was endowed with enhanced light absorption, charge separation, c) The PL decay spectra of as prepared samples. d) CH 4 and O 2 generation velocity of as-prepared samples. a,e,f ) Adapted with permission. [372] Copyright 2006, Nature Research. b-d) Adapted with permission. [373] Copyright 2014, The Royal Society of Chemistry. e) EELS of the support (1, blue line) and shell layer (2, red line) in (a). f ) Photocatalytic activity of Au@CdS/TiO 2 by time courses for photocatalytic reduction of MV 2þ . g) Time courses of CO evolutions. h) Average CO production rates of g-C 3 N 4 , α-Fe 2 O 3 , and α-Fe 2 O 3 /g-C 3 N 4 hybrid, and i) Z-scheme photocatalytic system. Adapted with permission. [392] Copyright 2018, Wiley-VCH.
www.advancedsciencenews.com www.small-structures.com and redox abilities. The CO and O 2 yields were 11.7 and 5.7 μmol under visible light (λ > 400 nm) in 24 h, respectively. Jiang et al. designed an urchin-like α-Fe 2 O 3 /g-C 3 N 4 Z-scheme junction, which was confirmed by the active radical tests. [392] In the absence of any cocatalysts and sacrifice agents, it demonstrated a much higher production rate of CO (27.2 μmol g À1 h À1 ), which was over 2.2 times higher than single g-C 3 N 4 (10.3 μmol g À1 h À1 ), (Figure 25g,h). There are two main reasons are responsible for the enhanced photocatalytic activity: on one hand, the urchin-like structure was very beneficial to the harvesting light, which contributed to the generation of more photoinduced charge carriers. Consequently, the Z-scheme structure of α-Fe 2 O 3 /g-C 3 N 4 allowed improved charge separation and reductive ability for CO 2 reduction (Figure 25i). More examples of Z-scheme junction for CO 2 reduction are shown in Table 3.
The aforementioned junctions mainly include two types of interface structure, namely metal-semiconductor and semiconductor-semiconductor interfaces. At a metal-semiconductor interface, the charge transfer between metal and semiconductor and band bending is closely related to Schottky barrier. Z-scheme junction as a typical semiconductor-semiconductor junction for interface engineering involves solid-liquid and solid-solid interfaces, and the direction Z-scheme junction displays a larger potential for CO 2 photoreduction due to its many merits. The promotion of electrons from oxidative parts to the reductive parts is important for a doable Z-scheme junction. As a practicable strategy, it has been successfully utilized in photocatalytic CO 2 reduction and other reactions, including H 2 or O 2 evolution for overall water splitting systems.

MOFs and COFs
An ideal support can realize the integration of activity, selectivity, and efficiency. MOFs and COFs are emerging candidates to fulfill these challenges. The large specific surface areas and the porous network adsorption ensure the good adsorption of CO 2 , [168,[393][394][395][396][397][398][399][400][401] and easy tailoring on the molecular structures of MOFs and COFs with different building units allowed adjustable bandgap, optimized charge carrier transfer ability, and tunable selectivity of reductive products (Figure 26a). [28,[402][403][404][405] In 2011, the first MOF for CO 2 photoreduction was developed by incorporating Re complexes into UiO-67. [406] The advantage for this system was to replace original ligands para-biphenyldicarboxylic acid (bpdc) in UiO-67 by a molecular catalyst Re I (bpy)(CO) 3 Cl (bpy ¼ 2,2 0 -bipyridine), due to the matchable lengths between bpdc and Re complexes ligands (L 4 ) in UiO-67 frameworks. Using L 4 as an active linker in UiO-67 framework, the Re catalytic centers were stable in the atmosphere (at most 400 C), and the porous MOF platform was capable of converting CO 2 to CO in an acetonitrile solution with trimethylamine as sacrificial agent under illumination. The TON of functionalized UiO-67 was 10.9 in 20 h, which was almost 3 times as high as that of the molecular catalyst Re I (dcbpy)(CO) 3 Cl (Figure 26b,c). Unfortunately, it was observed that functionalized UiO-67 lost catalytic activity after limited reactions.
Secondary building units (SBUs) in MOFs play a vital role in redox reactions due to their potential coordination sites. In terms of MIL-101(Fe), up to three potential open metal sites were available with the Fe-SBUs, leading to desirable adsorption and efficient CO 2 reduction. The efficient charge carrier mobility was also observed from O 2À to Fe 3þ and Fe 2þ was substantially produced in MIL-101 under visible light. Thus, a TON of CO 2 to HCOOH of 1.2 was observed for MIL-101(Fe) under 24 h visible-light irradiation using TEOA as the sacrificial agent. Furthermore, it was demonstrated that there is a strong relationship between HCOO À production and the wavelength of illumination in MIL-101(Fe) system (Figure 26d-e). [407] Yan et al. [408] demonstrated that the combination of binuclear Eu(III) 2 SBUs and Ru(phen) 3 -derived ligands to build a novel MOF, which realized the conversion of CO 2 to HCOOH under visible-light irradiation, and the HCOOH evolution reached 321.9 μmol h À1 mmol MOF À1 upon visible light illumination (from 420 to 800 nm). The Eu 2 (μ 2 -H 2 O) SBUs have open metal sites and metal clusters are covered by terminal water ligands, which are favorable for the adsorption and reduction of CO 2 . Under irradiation, the electrons tended to transfer from metallic ligands to SBUs to participate in the CO 2 reduction reaction and visible light can motivate binuclear [EuII-H 2 O-EuII]-active sites to finish the conversion CO 2 to HCOOH via a two-electron reduction process.
The adjustment of electronic characteristics of the SBU in MOFs is also a practicable way for promoting the photocatalysis. Lee et al. [409] demonstrated Ti 4þ -modified Zr 6 O 4 (OH) 4 SBU in UiO-66 for photoreduction of CO 2 to HCOOH with matched redox potential energies between the organic linkers and SBU. The doping of Ti 4þ ions in SBU enabled a lowered electron accepting levels of Zr 6-x Ti x (Zr 6-x Ti x O 4 (OH) 4 ) SBUs, thus allowing a photocatalytic activity. Eventually, this Ti 4þ -modified UiO-66 was demonstrated to expand the light responsive range, boost the transfer of photogenerated carriers, and reduce the recombination of electrons and holes. Thus, formic acid was selectively produced from CO 2 under visible light in an acetonitrile with TEOA as the sacrificial base and 1-benzyl-1,4-dihydronicotinamide as the sacrificial reductant. This MOF achieved a TON of 6 for CO 2 reduction within 6 h, which indicated that each Ti 4þ site transferred about 13 electrons to reactants over each catalytic run. Overall, MOFs as efficient supports have many benefits, such as 1) single crystals of MOFs are usually transparent, which are able to use solar light without obvious scattering and adsorption; 2) MOFs have large specific surface area with abundant reactive sites; and 3) the highly porous network of MOFs provides an easy access of the CO 2 to potential catalytically active sites (Figure 26a).
Different from MOF systems, layer-structured COFs always show efficient mobility of charge carriers. [167] COF-366 comprising cobalt porphyrins (COF-366-Co) was reported as an electrocatalyst for CO 2 reduction. With an overpotential of À0.55 V, COF-366-Co achieved the reduction of CO 2 into CO with a yield of 36 mL mg À1 over 24 h with a Faradaic efficiency of 90% in water (Figure 26f,g). With an overpotential of À0.55 V at pH 7, TON was up to 290 000, over 24 h without any obvious activity decay (Figure 26h). Covalently linked cobalt porphyrins as active sites were able to efficiently facilitate the reaction route of CO 2 www.advancedsciencenews.com www.small-structures.com www.advancedsciencenews.com www.small-structures.com reduction. Moreover, COFs present high thermostability and chemical durability, due to the strong covalent bonds. Fu et al. [410] introduced two azine-based COFs for converting CO 2 to CH 3 OH with H 2 O without sacrificial agents under visible light. The activity of CO 2 photoreduction of azine-based COFs is higher than that of g-C 3 N 4 semiconductors with water. MOFs as 3D porous crystals have well-defined structure, and various metal clusters in MOFs lead to multiple advantages in light absorption, electronic structure, and catalytic active sites. Regulating organic linkers of MOFs can not only supply an extra light absorption pathway, but also increase the amount of potential active sites. The stabilities of MOFs, including water stability and photostability, are also vital factors for CO 2 photoreduction. The poor water stability limits the application of photocatalysts in aqueous solution systems. According to the hard and soft acids and bases (HSAB) theory, water stability of MOFs depends on bond strength between the ligands and metal clusters. [411] MIL-100 (Al, Fe, and Cr), MIL-125 (Ti), and ZIF-7 (Zn) have a high water stability. The photostability is another issue for MOFs, especially for hydrocarbon products evolution from CO 2 reduction. Generally, MOFs with a high connectivity number of metal clusters and ligands usually show a high photostability, such as UiO-66 and UiO-67. [412] It is worth to note that the structure collapse may occurs in MOFs during long illumination. And in this condition, MOF derivatives may realize high photocatalytic CO 2 reduction than MOFs itself.
COFs usually show high charge carrier mobility, due to the p conjugation and p-p stacking in their reticular structure. Added nitrogen units or metal irons are usually catalytic centers in COFs, thus enabling a precise manipulation of active sites within a designed COF structure. A desirable photocatalytic MOFs or COFs system requires more photosensitive units, abundant active sites for substrates capture and reaction, as well as a decent match of bandgap between SBUs and linkers. Integrating all these functional groups in SBUs or organic linkers into MOFs or COFs will offer great potential for development of new catalytic systems in CO 2 reduction.

Semiconductor Biohybrids
Photosynthetic organisms develop into metabolic pathways to realize CO 2 reduction and store solar energy in chemicals. [413,414] Biologic carbon fixation has made great achievements that artificial catalysts are not able to make, such as the lifetime of enzymes and specificity of catalytic pathways, [415][416][417][418][419][420][421] But the light absorption capability of semiconductors is often higher than that of natural photosynthesis. [422][423][424][425] Thus, it is a promising way to solve CO 2 reduction issues by integrating high lightabsorbing inorganic catalysts with photosynthetic organisms. This novel way enables to develop value-added chemicals from CO 2 reduction by solar energy. Meanwhile, the solid boundaries can be broken by developing biohybrid semiconductor photocatalysts, which is also an approach to explore their potential and complementary functions.
The pioneering work of semiconductor biohybrids photocatalysts was reported by Sakimoto et al. [426] in which CO2 was converted into acetate by combining acetogenic bacteria with CdS NPs that have intense visible-light absorption. M. thermoacetica was one type of acetogenic bacteria, and CdS NPs can be grown on it by simply adding Cd2þ and sulfur source (such as cysteine) into the cells. By enzymatic process, the sulfur was easily reduced to sulfide and then to produce CdS NPs by reacting with Cd2þ (Figure 27a-c). It was found that CdS was primarily growth on the membrane of mycobacterium thermoacetum. These CdS NPs produced photogenerated electrons and holes under low-intensity simulated sunlight. According to the Wood-Ljungdahl path, a reducing equivalent [H] was induced by electrons and subsequently reacted with CO 2 to produce acetic acid. Figure 26. a) The design of MOFs as CO 2 reduction photocatalysts. Adapted with permission. [405] Copyright 2018, Nature Research. b) Plots of CO evolution turnover number (CO-TON) versus time in the photocatalytic CO 2 reduction with MOF 4 (blue square) and homogeneous H 2 L 4 (red circle). c) FTIR of as-synthesized MOF 4 (blue) and MOF 4 after photocatalysis (red). Adapted with permission. [406]  At the same time, holes reacted with cysteine to produce oxidized disulfide-bonded cystine (CySS) (Figure 27b). This photosynthetic reaction is shown in Equation (3) Several blank experiments and tests were conducted to prove this photocatalytic mechanism, including control groups of M. thermoacetica, CdS, and light condition (Figure 27d). The yield of acetic acid increased with increase the intensity of blue light (435-485 nm), and a 4 times increase in quantum yield (85 AE 12%) was obtained (Figure 27e).
However, CdS NPs are dramatic poison in organisms system, which are harmful to the environments. Developing intracellular particles can further enhance photosynthetic process, which means that NPs are grown on the entire cell instead of mainly being on the membranes. Gold nanoclusters (AuNCs) show benign visible-light absorption and good biocompatibility. Zhang et al. [427] reported that glutathione was used to combine AuNCs as intracellular NPs into a network (Figure 27f,g).  [426] Copyright 2016, American Association for the Advancement of Science. f ) The Au 22 (SG) 18 nanoclusters were delivered into bare M. thermoacetica (gray) during the culture process, forming M. thermoacetica/AuNCs with red emission. The simulated chemical structure of Au 22 (SG) 18 nanoclusters is shown in the inset. Light yellow spheres, Au atoms in the core; dark yellow, Au atoms in the staple motifs; red, S atoms in the shell. All other atoms (carbon, hydrogen) have been omitted from this structure for clarity. Also shown are space-filling models of acetic acid and CO 2 (orange, oxygen; gray, carbon; white, hydrogen). g) Schematic of bacterium. The electrons generated from intracellular AuNCs under illumination are used by enzymatic mediators inside the cytoplasm and are finally  Structured light microscopy (SLM) illustrated the position of AuNCs in bacteria, and also provided the distribution of AuNCs from bottom to top in hybrid systems (Figure 27h). The nonphotosynthetic bacteria can be functionalized by AuNCs penetrating to realize photosynthesis of CO 2 to acetic acid (Figure 27i). AuNCs also reduced the amount of reactive oxygen species (ROS) to keep the bacteria alive. As a result, the cell in AuNC-hybrid system demonstrated a high level of proliferation and viability rate after 12 h, but the CdS-hybrid system as a control group began to decay (Figure 27k). According to the Wood-Ljungdhal route, AuNCs produced photogenerated electrons and moved to cytoplasmic mediators in cells (Figure 27g,j). This biohybrid system was enabled to absorb photons and produce electrons to finalize metabolism in cells, which was highly dependent on the light absorption and biocompatibility of AuNCs. It was also proved that the continuous conversion of CO 2 can be efficiently achieved in several days. With the dual advantages of AuNCs and micro-organisms, a novel pathway using solar energy was achieved to realize carbon fixation to obtain hydrocarbons. The further study of inorganicbiological interface is expected to enable the manipulation of the activity and selective of CO 2 conversion in biohybrid systems. Valuable chemical products are expected to be obtained by photosynthetic CO 2 organisms with semiconductors in the future.

Single-Atom Photocatalysts
Homogeneous and heterogeneous catalysis are two main branches of the catalysis family, both of which are applied in photocatalytic CO 2 conversion to produce hydrocarbons. Several advantages of homogeneous photocatalysts are depending on their atomically dispersed active sites, [428][429][430] tunable light absorption, [431][432][433] as well as high activity and selectivity. [434][435][436][437][438][439] Heterogeneous photocatalysts with relatively low cost are easy to synthesize and facile to be extracted and recycled for long-term run. [440][441][442] Single-site catalysts integrate the advantages from both homogeneous and heterogeneous catalysts, thus attracting great interests for photocatalytic CO 2 reactions. [443][444][445][446][447] Cobalt as a transition metal is efficient single-atom catalysts either on the surface or in the body of substrate materials. Incorporating Co single atoms on the partially oxidized graphene nanosheets (Co 1 -G nanosheets) to realize CO 2 conversion was demonstrated by Gao et al. (Figure 28a). [448] The absence of Co-Co peak in Co 1 -G sample indicated that no Co or CoO clusters existed, and the Co species were separated into individual atoms. The Co single atoms showed many advantages, such as rich unsaturated coordination sites and variable electronic configuration that can ensure high activity and selectivity, simplified catalytic mechanism due to the straight forward metallic coordination situation. A high TON of 374 and TOF of 2.08 min À1 for CO formation was observed in Co 1 -G catalyst (Figure 28b), and they were further increased to 678 and 3.77 min À1 , respectively, by combining [Ru(bpy) 3 ]Cl 2 with Co 1 -G nanosheets. [449][450][451][452] Combining single-atomic sites into 2D ultrathin nanosheets is a promising approach to integrate their advantages. [453] Di et al. [454] reported the successful synthesis of Bi 3 O 4 Br atomic layers with well-dispersed single cobalt atoms (Figure 28c). It was proven that the valence state of isolated and highly dispersed single Co atom is close to þ2 (Figure 28d,e), because the position of Co K-edge absorption edge was close to that of CoO rather than Co foil in Co-Bi 3 O 4 Br-1. Due to Co atoms as single sites (Figure 28f,g), the activation energy barrier of CO 2 was decreased to stabilize the intermediates, such as COOH*.  In summary, single atom plays a vital role for CO 2 photoreduction, like lowering the activation energy barrier of CO 2 , and the single-atom catalysts bridge heterogeneous and homogeneous photocatalysts, showing substantial superiority in terms of highly tunable structure and robust performance.

Polarity Enhancement
In noncentrosymmetric (NCS) materials, the centers of positive and negative charges do not coincide, which can produce a polarization-induced electric field to urge the migration of photogenerated electrons and holes to opposite directions, achieving the efficient separation of electron-hole pairs in semiconductor photocatalysts. [455,456] The polarization enhancement boosts the charge separation, which was uncovered in a latest review by our group. [457] Recently, a series of polar photocatalysts with strong polarization were developed for high photocatalytic activity. [458][459][460][461] Ferroelectric structures have spontaneous dipole moment, and thus tend to generate spontaneous polarization electric field to transfer charge carries from bulk to surface. A layered ferroelectric perovskite SrBi 4 Ti 4 O 15 was prepared by our group for efficient CO 2 reduction, and it was treated by annealing at 350 and 650 C for enhancing the ferroelectricity (Figure 29a-d). [462] SrBi 4 Ti 4 O 15 uncovered a strong ferroelectric spontaneous polarization along [100] direction, which boosted charge separation along a axis direction in bulk (Figure 29a,d). Without cocatalysts and extra sacrificial agents, the annealed SrBi 4 Ti 4 O 15 (SBTO) nanosheets at 350 C showed a prominent photocatalytic activity Figure 28. a) HAADF-STEM images of the Co 1 -G catalyst. The atomically dispersed Co atoms in (a) are highlighted by the yellow circles. b) TONs of CO and H 2 production by Co 1 -G nanosheets in the first 3 h under visible-light (λ > 420 nm) irradiation, in comparison with those by graphene (G), CoCl 2 , graphene with CoCl 2 (CoCl 2 þ G), and graphene oxide with CoCl 2 (CoCl 2 þ GO) under the same condition. [Ru(bpy) 3 ]Cl 2 is used as a light absorber in all the measurements. Adapted with permission. [448] Copyright 2018, Wiley-VCH. c) Surface morphology of Co-Bi 3 O 4 Br-1, atomic-resolution HAADF-STEM images of Co-Bi 3 O 4 Br-1, and Synchrotron radiation X-ray absorption fine structure measurements. d) Co K-edge X-ray absorption near edge structure (XANES) spectra, e) Bi L 3 -edge XANES spectra, extended X-ray absorption fine structure spectra of f ) Co K-edge and g) Bi L 3 -edge. h) Theoretical study. a Schematic representation of mechanism on the Co-Bi 3 O 4 Br. Adapted with permission. [454] Copyright 2019, Nature Research. in CO 2 reduction for CH 4 evolution in gas-solid system, with a production rate of 19.8 μmol g À1 h À1 and apparent quantum yield of 1.33% at 365 nm. In addition, Bi 2 MoO 6 ultrathin nanosheets (BMO-U) with boosted ferroelectricity were developed for CO evolution from CO 2 photoreduction. [463] Due to lattice distortion in ultrathin layered structure, the polarization degree of BMO-U can be further enhanced by corona poling, which causes a boosted polarized electric field along the a and c axes in BMO-U, promoting the charge separation between layers. Thus, the CO yield of polarized BMO-U (14.38 mmol g À1 h À1 ) was over 3 times larger than that of bulk Bi 2 MoO 6 (4.08 mmol g À1 h À1 ), and over as 10 times as higher than that of bulk Bi 2 MoO 6 (1.36 mmol g À1 h À1 ). A directional accumulation of polar units can also enhance the macroscopic spontaneous polarization, inducing the transfer of photoinduced charge carriers from bulk to surface. NCS BiOIO 3 contains aligned IO 3 polar units along the [001] direction. When BiOIO 3 grew along this specific direction (Figure 29e,f ), an enhanced polarization electric field was formed from BiOIO 3 particle (BIO-S) to BiOIO 3 nanorods (BIO-L), which promoted the charge carriers transfer from the bulk to the surface (Figure 29h,i). BIO-L with abundant OVs showed a CO yield of 17.33 μmol g À1 h À1 , which was over 10 times larger than that of BIO-S (1.68 μmol g À1 h À1 ), as shown in Figure 29g. [464] Furthermore, a polarized local surface in photocatalysts facilitates charge migrates and thereby promotes redox reactions on the surface. Zhang et al. [465] reported a successful synthesis of around 1 nm g-C 3 N 4 layers, and further construct a polarized surface to convert CO 2 to CO with showing roughly 80% Faradaic efficiency. g-C 3 N 4 characterized with ultrathin layered structure with polarized surface (2D-pg-C 3 N 4 ) was enabled to enrich electrons. As a result, the current density of 2D-pg-C 3 N 4 at À1.2 V versus Ag/AgCl is 3.05 Ma cm À2 , which is almost 30 times as large as that of the bulk g-C 3 N 4 . An intensified electron density of ultrathin g-C 3 N 4 layers was observed, because the Figure 29. a) 2D phase maps and d) corresponding curve of the piezoelectric response. b) Ferroelectric phase curve SBTO obtained by piezoelectric force microscopy. c) Electric hysteresis loop of SrBi 4 Ti 4 O 15 annealed at different temperatures. Adapted with permission. [462] Copyright 2019, Elsevier. SEM images of e) BIO-S, and f ) BIO-L, g) the piezoelectric coefficient (d 33 ), the surface charge and the corresponding curve of the piezoelectric response of h) BIO-S, and i) BIO-L. Adapted with permission. [464] Copyright 2019, Wiley-VCH.
www.advancedsciencenews.com www.small-structures.com characteristic XPS peaks of sp 2 N and C atoms were shifted to lower binding energy regions. The polarized melem subunits tend to become active centers to enhance CO 2 reduction. The polarized melem subunits can act as active centers to promote CO 2 reduction over metal-free g-C 3 N 4 . The CO 2 desorption peak of 2D-pg-C 3 N 4 was much lower than that of bulk g-C 3 N 4 , thus 2D-pg-C 3 N 4 showed a favorable activation process of CO 2 molecules. Therefore, it is an efficient strategy to use polarization as a useful way for reducing the recombination of electrons and holes in bulk and/or on the surface, but the research in this particular area is still in its infancy. The elemental doping is potential to enhance the polarization; precise tailoring the microstructure of photocatalysts system will also help to some content.

Conclusions
Structural engineering is regarded as an efficient way to solve the crucial problems that restrict the CO 2 photoreduction activity, thus resulting in the high activity, stability, and tunable selectivity of products. In this Review, the progress in this research area has been systematically summarized. First, the efficient strategies for structural design from recent works are reviewed, which are divided into five sections based on nano-/microstructure, crystalline and band structures, surface structure, and interface structure. Then, new trends and strategies for CO 2 photoreduction were introduced. Several new photocatalytic systems for CO 2 photoreduction have been introduced, including MOFs, COFs, semiconductor biohybrid systems, and single atom systems. In addition, polarization as a new strategy to enhance charge separation has been summarized. Although great achievements have been made in recent years, there is still a long way in developing structural engineering for CO 2 photoreduction. For future direction in this particular field, structural engineering will be developed from the following aspects.
First, integrating multiple strategies for structural engineering may be a promising approach to maximizing the light absorption, charge separation, and surface catalytic reaction at the same time. For example, the separated redox actives sites with a fast charge transfer route can be achieved in Z-scheme system. The enhanced visible-light absorption is usually established by nitrogen doping. Therefore, the integration of these strategies can accomplish high charge separation efficiency, strong redox ability, and visible-light absorption. [466] To maximize photocatalytic performance, combinatory application of various strategies in structure engineering is necessary to deal with each component on surfaces or interfaces. So, one of the biggest challenges in photocatalysts design is to appropriately integrate these strategies.
Second, more in-depth study should be conducted on some promising strategies, such as the polarization regulation, to further disclose the mechanism for charge separation and photocatalytic activity enhancement. In addition, there are also many nature inspired methods to expand the new tactics of structural engineering, such as fabrication of semiconductor biohybrid systems, [467][468][469] chlorophyll loading, [470,471] or bonding the chromophore, [472] which may attract considerable research interests in the future.
Third, the mechanism of CO 2 reduction has been illustrated in a large number of works by in situ techniques, such as in situ IR and Raman. [473] Using new strategies to design semiconductors usually realized enhanced catalytic activity with diverse final products. But the mechanism behind these results is still unclear and under debate. More attentions should be paid to the changing of key intermediates and reaction pathways, because diverse active sites are built on semiconductor by various strategies. More in situ techniques are required to identify reaction pathways on the specific structure. The mechanism for exact reaction pathways may differ from the currently recognized one, which can advance the development of CO 2 photoreduction.