Nanostructured heterogeneous photocatalyst materials for green synthesis of valuable chemicals

The photocatalytic process employing nanostructured semiconductor materials has attracted great attention in energy production, CO 2 reduction, and water/air purification for decades. Recently, applying heterogeneous photocatalyst for the synthesis of valuable chemicals is gradually emerging and considered as a promising process for the conversion of cheap resources (i.e., biomass derivatives, polyols, and aromatic hydrocarbons). Compared with traditional thermal catalytic approaches, the photocatalytic process provides a mild reaction condition and flexible platform (photocatalyst) that allows precise tweaking of reaction intermediates and reaction pathways, thus resulting in fine control of the selective synthesis of specialized chemicals that are challenging for thermal catalysis. In this review, we summarize recent achievements in photocatalytic synthesis of various industrial important chemicals via photo-oxidative and photo-reductive processes. The selective oxidation of alcohols and aromatics, epoxidation of alkenes, hydrogenation of gaseous molecules and hydrocarbons, and coupling reactions by means of various photocatalysts including metal oxides, supported plasmonic metal nanostructures, conjugated organic polymers, anchored homogeneous catalysts, and dye-sensitized heterostructures are discussed from a material perspective. In addition, fundamental understandings of reaction mechanisms and rational design of nanostructured photocatalysts for enhancing efficiency, selectivity, and stability are discussed in detail.

The synthetic paths via photocatalysis exhibit several intrinsic advances compared with conventional thermal-catalytic approaches. While traditional catalysts accelerate reactions by means of forming instable surface adsorbed intermediate species with lower activation energy, the photocatalytic processes involve the excitation of the catalyst and the reactants to higher energy states that overcome the activation energy barriers. Therefore, corrosion-resistant reactors for harsh reaction conditions (i.e., high temperatures and pressures and strong oxidants and reductants) can be simplified for photocatalytic synthesis (i.e., ammonia synthesis, Fischer-Tropsch synthesis, and selective oxidation and hydrogenation reactions) [43][44][45][46] . However, researchers should be aware that specialized reactors with optimized light transmittance and mass transportation are required for heterogeneous photocatalysis. In fact, the photon energy is ~1.7-3 eV in the visible light region, which is sufficient to drive most catalytic reactions. Owing to the mild reaction conditions and tunable band positions of the semiconductor photocatalysts, some challenging redox reactions that require precise control of the redox potentials can be realized by photocatalytic approaches with high conversion and selectivity (i.e., epoxidation reactions and coupling reactions) [47][48][49] . This results in a significant reduction in energy consumption for separation processes (i.e., distillation, filtration, and sublimation), which is desirable for economic and environmental purposes. Nevertheless, a longer lifetime of the photocatalyst materials is expected due to such mild reaction conditions, which is essential for noble metal-supported systems, delicate surface engineered materials, and anchored homogeneous catalysts.
Several in-depth reviews summarize recent progress in utilizing heterogeneous photocatalyst for the synthesis of valuable chemicals from different perspectives, targeting specific reactions (i.e., partial oxidation, C-C coupling, and biomass products conversion), materials (i.e., TiO 2 , polyoxometalate, and supported plasmonic metal nanoparticles), and reactor design [45,46,[66][67][68][69][70][71][72][73][74] . Miyabe and Kohtani [68] emphasized recent developments of TiO 2 -based photocatalysts in heterogeneous photocatalytic reductive and oxidative organic synthesis. Bloh et al. [69] summarized recent developments in heterogeneous photoredox catalysis with a special focus on materials optimization. This review addresses advances in photocatalytic synthesis from an application-orientated perspective, presenting the development in the design of nanostructured photocatalyst materials by covering up-to-date and classical works. We categorize the photocatalytic reactions into oxidative and reductive reactions, which are further subdivided into alcohol oxidation, epoxidation, oxidation of aromatics oxidative coupling, hydrogenation reactions, and reductive coupling [Scheme 1]. The design strategies, selection rules, advantages, and limitations of various nanostructured photocatalyst materials, including metal oxides, supported plasmonic metal nanostructures, organic polymers, anchored homogeneous catalysts, and dye-sensitized heterostructures, are also summarized. For oxidation reactions, the optimization of oxidation power of the photocatalyst is presented for the controlled synthesis of target products to avoid overoxidation. Correspondingly, the manipulation of photogenerated hydrogen atoms for selective reduction reactions by controlling the surface properties of the photocatalyst is discussed. In addition, fundamental understandings of reaction mechanisms for the rational design of nanostructured photocatalysts are presented. Moreover, this review presents the recent design and development of continuous reactors for scaling-up of photocatalytic processes.

Monohydric alcohols
Monohydric and polyhydric alcohols are widely used as hole scavengers in photocatalytic hydrogen production (i.e., methanol, ethanol, glycerol, and triethanolamine) [75][76][77] . Bahruji et al. [76] showed that monohydric alcohols dissociate via the decarbonylation oxidation process when Pd nanoparticle (NP)modified TiO 2 is used as the photocatalyst under deaerated conditions, resulting in the formation of H 2 , CO 2 , and corresponding hydrocarbons, as demonstrated in Equations (1) and (2): It is proposed that the alcohol decarbonylation process takes place on the Pd NP surface with CO as the reaction intermediate, which is further oxidized to CO 2 by the photogenerated oxygen species on the TiO 2 surface. To further enhance the activity and extend the reaction into visible light region, plasmonic metal NPs have been loaded onto various semiconductor photocatalysts (i.e., TiO 2 and g-C 3 N 4 ) to serve as both promoter and light absorber [78,79] .
Beyond the evolution of gaseous H 2 , the liquid phase products may also be influenced by the identity of the metal promoter and the microstructure of the semiconductor during photocatalytic alcohol oxidation under anaerobic conditions. Ruberu et al. [80] revealed that the semiconductor-metal heterostructures dictate photocatalytic dehydrogenation and hydrogenolysis of benzyl alcohol. As demonstrated in Figure 1A, while CdS-Pt favors dehydrogenation of benzyl alcohol to generate benzaldehyde and H 2 , CdS 0.4 Se 0.6 -Pd favors further hydrogenolysis of the generated benzaldehyde by the photogenerated H 2 , resulting in the formation of toluene. A mechanistic study using labeled benzyl alcohol reveals that the adsorption of photogenerated Scheme 1. Categories of heterogeneous photocatalytic synthetic reactions discussed in this review and their major challenges.
hydrogen atoms is stronger on the Pd surface than on the Pt surface, thus resulting in the consecutive hydrogenolysis benzaldehyde. Such a design strategy is useful for synthetic chemistry via photocatalytic hydrogen transfer reactions.
Lu et al. [81] showed that modulation of TiO 2 microstructure and polymorph can alter the oxidation kinetics and pathway of photocatalytic ethanol reforming under deaerated conditions in water solution. While ethanol is oxidized to acetaldehyde and acetic acid over Pt-modified Degussa P25 ® , the value-added coupling product (2,3-butanediol) is selectively formed on rutile TiO 2 . The conversion of ethanol and selectivity to 2,3-butanediol can be further enhanced by employing Pt-modified rutile nanotubes with filled channels, as shown in Figure 1B. Such configuration prevents the deep oxidation of photogenerated α-hydroxyethyl radicals (CH 3 •CHOH) within the pores and thus facilitates the desorption of CH 3 •CHOH, which undergoes a coupling reaction to form 2,3-butanediol in the solution. Later, the same group reported that absolute ethanol can be directly converted to 1,1-diethoxyethane and H 2 using Pt-TiO 2 nanotubes and nanorods under deaerated conditions following Equations (3) and (4) [82] : Interestingly, although almost the same photocatalyst materials were used in both studies [81,82] , the liquid phase products are completely different. Therefore, we suppose that the different reaction mechanisms originate from the presence of water, which dictates the conversion of ethanol to undergo either dehydrogenation or dehydration process. respectively. This figure is used with permission from the American Chemistry Society [80] . (B) Selective photoconversion of ethanol to butanediol using rutile TiO 2 nanotubes with filled channels. This figure is used with permission from the Royal Society of Chemistry [81] . (C) Schematic demonstration of aerobic alcohol oxidation using a Pt/Ta 2 O 5 photocatalyst under visible light irradiation. (D) Au-Pd NPs functionalized HT-PO 4 3for selective photocatalytic esterification of alcohols. This figure is used with permission from the American Chemistry Society [87,91] . NPs: Nanoparticles.
Recently, more attention has been focused on the oxidation half-reaction, where the alcohols may convert to various value-added products in liquid phase under aerobic conditions. Photocatalytic oxidation of alcohols into aldehydes or ketones has been realized under aerobic conditions using TiO 2 -based photocatalysts [83] ; however, low efficiency and poor selectivity are the main challenges that need to be solved. Generally, loading Au, Ag, or Pt NPs on semiconductor photocatalysts can promote visible light absorption via interband transition of d band electrons and produce hot electrons (e hot -) [84] , which are injected into the conduction band (CB) of the semiconductor. e hot -, together with the positive charges (holes, δ + ) formed on the semiconductor, are consequently consumed by surface adsorbed O 2 and alcohol molecules to complete the aerobic oxidation cycle. e hotshould not be confused with the CB electrons (e CB -) generated by semiconductor photocatalysts, as it refers to the surface localized vibrational electrons generated by plasmonic metal nanoparticles under irradiation. To further enhance the charge separation efficiency, Li et al. [85] deposited Au NPs on BiOCl with oxygen vacancies. The photogenerated e hotcan therefore be trapped at the vacancy states, thus prolonging the lifetime of the δ + on the Au surface. However, researchers should be aware that, since the injection of e hotinto the semiconductor needs to overcome the Schottky barrier at the metal-semiconductor interface [86] , the efficiency of these systems is still low in general. To further increase the efficiency, Sakamoto et al. [87] proposed depositing noble metal NPs on large bandgap semiconductors to create sufficient high work function (ϕ B ) and thus avoid electron transferring across the metal/semiconductor interface. Here, Pt NPs are supported on Ta 2 O 5 (E g ~4.0 eV) to absorb visible light.
The photogenerated e hot and δ + will be confined on the Pt surface to react with molecular O 2 and alcohols, as demonstrated in Figure 1C. A series of high-performance systems have been developed with this design principle (i.e., Au/Al 2 O 3 and Pt/SiO 2 ) for the oxidation of various alcohols (i.e., isopropanol and benzyl alcohols) into corresponding aldehydes or ketones. Similarly, the confinement of e hoton the metal surface can also be realized by employing engineered alloy NPs or simply metal clusters [88,89] . However, researchers should be aware that such a process is a photo-induced thermal catalytic reaction rather than a classical photocatalytic reaction.
Beyond aldehydes and ketones, alcohols can also be converted to other products via photocatalytic oxidation process. Han et al. [90] showed that Au-Ag alloy NPs supported on TiO 2 can efficiently oxidize methanol into methyl formate selectively under low oxygen pressure and UV irradiation. Here, methanol gradually dissociates into methoxy (CH 3 O•) or formaldehyde (CH 2 O) on TiO 2 surface. The generated formaldehyde then undergoes dehydration with another methanol molecule to generate methyl formate. The presence of metal NPs promotes oxygen dissociation to consume the surface-adsorbed H atoms into water. The optimum conversion and selectivity to methyl formate are achieved when the Au-Ag alloy with a 1:1 ratio is loaded onto TiO 2 . However, this reaction takes place in gas phase, and the concentration of O 2 (< 0.5%) needs to be carefully controlled to avoid CO 2 formation. To overcome these drawbacks, Xiao et al. [91] employed Au-Pd alloy NPs supported on phosphate-modified hydrotalcite (HT-PO 4 3-) to convert a series of aliphatic alcohols into corresponding esters in O 2 under neutral conditions. As presented in Figure 1D, phosphate anions are introduced into the calcined HT ([Mg 6 Al 2 (OH) 16 ]CO 3 ·mH 2 O) via ion exchange, which is then loaded onto Au-Pd alloy NPs. Such structure couples the basic sites of the HT support and the photocatalytic properties of the alloy NPs, which promotes the dehydrogenation of alcohols into ketones and thus the consecutive esterification of the generated ketones with alcohols via under dehydration.

Polyhydric alcohols
Catalytic conversion of polyols (i.e., diols, triols, sugars, cyclitols, and celluloses) to high-value chemicals (i.e., polyurethanes, polyesters, and polycarbonates) is of potential importance in the chemical industry and has also attracted increasing attention in photocatalysis. Employing glycerol and other polyols as hole scavengers for photocatalytic hydrogen production under deaerated conditions is the most direct application [92,93] , but the conversion efficiency is relatively low. To help the design of high performance photocatalyst, Lercher's group investigated photocatalytic reforming of a series polyols in water solution over the Rh-TiO 2 system based on global kinetic analysis [94,95] . The polyol conversion takes place via three possible pathways: oxidative C-C bond cleavage from direct hole transfer to chemisorbed oxygenates on terminal Ti(IV)-OH groups, oxidation to corresponding aldoses or ketoses initiated by indirect hole trapping at surface lattice oxygen sites, and light-driven dehydration (minor). Jin et al. [96] further revealed the reaction mechanism of polyol conversion using ethylene glycol (EG) at molecular level on pristine TiO 2 and metal NPs decorated TiO 2 by coupling in situ surface science techniques with vibrational spectroscopies. As shown in Figure 2A and B, in situ scanning tunneling microscopy (STM) images suggest that EG molecules are adsorbed on the five-coordinated Ti (Ti5c) of rutile TiO 2 (110) surface and photocatalytically converted into two parts of formaldehyde and surface-adsorbed hydrogen atoms (H ads ) via oxidative C-C bond cleavage. The active sites remain on the TiO 2 surface even after loading Au nanoclusters on rutile TiO 2 (110), as shown in Figure 2C and D. In situ IR-MS spectroscopy suggests that the desorption of H ads to H 2 is the rate-determining step (RDS), which can be promoted by metal NPs (Au, Pd, and Pt) due to the low free adsorption energy of H ads on metal surface, thus improving the dissociation of EG. Chong et al. [97] confirmed that the selectivity and efficiency in polyol conversion can be tuned by  (110) before and after irradiation. This figure is used with permission from Elsevier [96] . (E-G) TEM images of shaped TiO 2 showing tunable selectivity in glycerol conversion. (H) SEM image of the Bi 2 WO 6 catalyst and proposed mechanism for selective oxidation of glycerol under visible-light irradiation. This figure is used with permission from the Royal Society of Chemistry [97,100] . EG: Ethylene glycol.
controlling the electronic and surface structure of the TiO 2 -based photocatalyst materials. As presented in Figure 2E-G, the selectivity in photocatalytic glycerol conversion can be tuned by varying the shape of anatase and rutile TiO 2 . While Pt-loaded anatase with dominant {001} or {101} facets leads to the formation of formaldehyde and CO 2 with low conversion to glycolaldehyde, shaped rutile with a high percentage of {110} facets results in high selectivity towards glycolaldehyde (> 90%) without the formation of CO 2 .
Photocatalytic conversion of polyols with the presence of oxygen has also been investigated. Augugliaro et al. [98] showed that most commercially available TiO 2 -based photocatalysts will oxidize glycerol into 1,3dihydroxyacetone, glyceraldehyde, formic acid, and carbon dioxide in aqueous phase. Such a process requires no metal cocatalyst, as the photogenerated H ads will be consumed by the molecular oxygen to yield water. However, the selectivity in polyol conversion under oxygen conditions is more complex and difficult to control. We employed EG as the model polyol to investigate the reaction pathways and the role of metal NPs under aerated conditions using in situ IR-MS spectrometry [99] . Ethylene glycol undergoes the C-C bond cleavage process to formaldehyde, which is subsequently oxidized into CO 2 or polymerized to paraformaldehyde with the photogenerated water. Similar to the reaction initiated under deaerated conditions, the removal of photogenerated H ads + by O 2 and photogenerated eis the RDS. Interestingly, loading Ag and Au NPs on TiO 2 increases the selectivity to paraformaldehyde, whereas Pt cocatalyst promotes the complete oxidation of EG to CO 2 . In situ IR-MS spectrometry reveals that Ag-and Aumodified TiO 2 can provide water directly through H ads + interaction with O 2 and e CB -, whereas Pt/TiO 2 supply water indirectly via H 2 and formaldehyde oxidation.
The main drawbacks of TiO 2 -based photocatalysts are poor efficiency under visible light and deep oxidation of polyol into CO 2 . Zhang et al. [100] further developed a flower-like Bi 2 WO 6 photocatalyst to overcome these challenges [ Figure 2H]. The Bi 2 WO 6 photocatalyst is characterized by a mild oxidation power (Eg ≈ 2.8 eV), which can selectively oxidize glycerol to 1,3-dihydroxyacetone under visible light radiation in water solution via the direct hole transfer process. Additionally, the 1,3-dihydroxyacetone molecule weakly adsorbes on Bi 2 WO 6 , which prevents deep oxidation and thus achieves high selectivity.
Although the photocatalytic oxidation of alcohols shows huge potential for the utilization of biomass as a chemical feedstock for the synthesis of value-added products, in-depth investigations are necessary to further gear up the productivity and selectivity under mild conditions. This also requires accelerating the reduction of molecular oxygen in a controlled manner for the rapid generation of ideal oxidative radical species, which is crucial in avoiding the use of extra oxidants [e.g., Hydrogen peroxide (H 2 O 2 ) and O 3 ].

Epoxidation reactions
Synthesis of epoxides via heterogeneous catalytic epoxidation of alkenes is a challenging and important reaction in both industry and fundamental research. Ag-, Au-, and Cu-based catalysts are typically employed for the selective oxidation of a series of alkenes (i.e., ethylene, propylene, and styrene), which are normally performed under elevated pressure (1-3 MPa) and moderate temperature (220 °C -280 °C) for optimized activity and selectivity [101,102] . Transition state analysis reveals that tuning the identity of metal catalyst (Ag and Cu) can promote the conversion of the common oxametallocycle intermediate to either epoxide or acetaldehyde in ethylene oxidation, thus avoiding the full oxidation of alkenes to CO 2 [103] .
Since these metal catalysts also show plasmonic absorption in visible light region, it is reasonable to employ them as photocatalysts for epoxidation reactions under visible light irradiation. Indeed, Christopher et al. [104,105] showed that Ag nanocubes (20 wt%, ~ 75 nm) loaded on α-Al 2 O 3 support can realize selective oxidation of ethylene to ethylene oxide with a selectivity of ~50%. Although the photocatalytic epoxidation reaction still needs to be operated at elevated temperature to achieve reasonable quantum efficiency (~40% at 200 °C and ~4% at 150 °C), the epoxidation rate is significantly higher than that of the pure thermal catalytic process. This suggests that the process is more likely a photo-induced thermal process rather than a photocatalytic approach. Later, the same group further developed Cu NPs (2 wt%, ~ 45 nm) supported on SiO 2 for the selective oxidation of propylene [106] . Interestingly, such a system shows a higher selectivity in propylene epoxidation under visible light irradiation than that of the thermal process; thus, it is a photothermic-catalytic process. It is proposed that the high selectivity under irradiation is accompanied by light-induced redox of the surface Cu 2 O layer to metallic Cu via localized surface plasmon resonance of the Cu core.
Zhang et al. [107] employed SERS coupled with theoretical calculations to understand the active sites of individual Ag NP catalyst for plasmon excitation-assisted ethylene epoxidation. It shows that graphene layers are first generated on the Ag surface under irradiation, which in turn promote the epoxidation of ethylene at the edge defects sites. As shown in Figure 3A, TEM image reveals the activation of Ag/Al 2 O 3 photocatalyst with in situ generated graphene layers on Ag NPs under irradiation with the presence C 2 H 4 (A-Ag/Al 2 O 3 ). Based on this evidence, they further developed a graphene/Ag/α-Al 2 O 3 photocatalyst (G-Ag/Al 2 O 3 ), which shows a similar feature in SERS spectra with distinct D and G bands of graphene compared with that of A-Ag/Al 2 O 3 [ Figure 3B]. Both the composite and the activated photocatalyst exhibit excellent performance in ethylene photo-epoxidation under ambient conditions compared with that of the conventional Ag/α-Al 2 O 3 catalyst [ Figure 3C].
Recently, Huang et al. [47] showed that the selective photo-epoxidation of a series of aliphatic and aromatic alkenes can be realized in liquid phase using Cu NPs supported on titanium nitride (TiN) by O 2 or even air at moderate temperatures. The photocatalyst prepared by impregnation-reduction method exhibits welldispersed metallic Cu NPs with a mean size of 4 nm supported on the TiN surface [ Figure 3D], which shows excellent stability in photocatalytic styrene conversion. According to theoretical calculations, the Cu NPs remain in the metallic state under reaction conditions due to strong charge transfer between Cu NPs and the TiN support. Mechanistic study shows that the whole reaction takes place on the Cu surface. Firstly, the molecular oxygen reacts with cyclic ether solvent on the Cu surface under irradiation to produce oxygen adatoms (O A ). Then, the photogenerated O A reacts with chemisorbed alkenes on Cu, forming a Cu-epoxide complex, and eventually desorbs as final products [ Figure 3E]. Interestingly, they also observed product stereo-selectivity in trans-isomers for the epoxidation of stilbene.
The major advantage of photocatalytic epoxidation is the mild reaction conditions, which is ideal for some niche applications in medications and sterilizations that only require in situ generation of epoxides with relatively low concentration. This requires further investigations in robust, non-toxic photocatalyst materials, which can initiate the epoxidation of some light alkenes (i.e., ethylene and propene) under ambient conditions.

Oxidation of aromatics
Selective oxidation of aromatics to value-added products (i.e., alcohols or aldehydes) is important in the chemical industry as they are precursors for resins, electronics, and drugs [108,109] . The challenge in photocatalytic selective oxidation of aromatics is the development of photocatalysts with moderate oxidation power, which can partially oxidize specific functional groups while avoiding ring-opening and eventually CO 2 formation. For example, photocatalytic direct oxidation of benzene to phenol has attracted great interest, as such a process may overcome the Hock rearrangement that produces an excessive amount of acetone as a byproduct. Chen et al. [110] showed that Fe-doped g-C 3 N 4 is active in the oxidation of benzene to phenol under visible light. Although H 2 O 2 is needed as the oxidant and the conversion is relatively low, the low cost of the photocatalyst material shows promising potential for applications. Yoshida et al. [111] showed that Pt-TiO 2 can generate electrophilic oxygen species photocatalytically from water to initiate the oxidation of benzene to phenol under UV irradiation and deaerated conditions [Equation (5)]: More importantly, a series of aromatics with different functional groups can be converted to corresponding phenolic compounds with relatively high selectivity. Mechanistic studies show that the reaction path is pHdependent. Under neutral or acidic conditions, the photogenerated holes react with TiO 2 to produce surface  [107] . (D,E) TEM image of single crystal Cu NP supported on TiN and the proposed reaction mechanisms in photo-epoxidation reaction. This figure is used with permission from the American Chemistry Society [47] . NPs: Nanoparticles; TOF: turnover frequency; SERS: surface-enhanced Raman scattering.
oxygen radicals, which attack the aromatic ring to form surface adsorbed phenoxy intermediates and eventually yield phenols. Under basic conditions, hydroxyl radical is the initial radical species that attacks the aromatic ring to produce phenols [112] .
Researchers have further engineered the surface of photocatalysts with different strategies to improve the selectivity and conversion efficiency in benzene oxidation under ambient conditions using air or O 2 as the oxidant. Zhang et al. [113] incorporated TiO 2 NPs in hydrophobically modified mesocellular siliceous foam (TiO 2 @MCF) for the synthesis of phenol. While the inner environment of the porous cage is hydrophobic due to surface silylation, the TiO 2 surface is hydrophilic after a post-UV treatment. Such structure has optimized the adsorption of benzene and desorption of phenol, thus avoiding further oxidation of phenol. Su et al. [114] showed that the selectivity for photo-oxidation of benzene to phenol can be enhanced by fine adjustment of the metal NPs that serve as a promoter. It is found that tuning the morphology and composition of Au-Pd metal NPs leads to the control of both polymerization of generated phenolic products and the complete oxidation of benzene. Specifically, Au shell Pd core -TiO 2 promotes the formation of phenol and switch-off undesired side reactions to quinones and CO 2 , thus resulting in optimal performance and selectivity in photocatalytic benzene oxidation. Hosseini et al. [115] further supported Au-Pd NPs on visible light response materials (g-C 3 N 4 ) to perform benzene oxidation and achieve high selectivity. A similar strategy has been employed to develop Cu 2 O NPs supported on defective graphene with a long-chain alkanethiols photocatalyst. [116] However, such systems require the addition of H 2 O 2 and face the challenge of improving conversion of benzene. Ide et al. [117] encapsulated Au nano-discs between titanate interlayers (K 0.66 Ti 1.73 Li 0.27 O 3.93 ) via surface modification with octadecyltrimethoxysilane to oxidize benzene under visible light using water as oxidant. Since the pristine titanate only absorbs UV light, the activity under visible light originates from the plasmonic absorption of the anisotropic-shaped Au NPs. Interestingly, it has been found that a high yield (62%) and selectivity (96%) to phenol can be achieved with the initial addition of phenol, although the promotion mechanism remains unclear.
Beyond the commonly used oxidants (i.e., O 2 and H 2 O 2 ), photo-oxidative conversion of aromatic compounds can also be performed with the presence of other oxidants. Jaynes and Hill [118] showed that selective carbonylation of toluene and cyclic alkanes to corresponding aldehydes can be achieved under 1 bar of CO by employing polyoxotungstate ([W 10 O 32 ] 3or [PW 12 O 40 ] 3-), which is a type of polyoxometalates (POMs). During the reaction, the UV-excited photocatalyst first abstracts a hydrogen atom from the alkane to produce an alkyl radical, which is subsequently trapped by CO to yield an acyl radical. Then, the acyl radical is converted to aldehyde by reductive H • donation from the protonated photocatalyst. To further employ POMs under visible light region, Suzuki et al. [119] synthesized a sandwich-type Ce 3+ -containing silicotungstate (CePOM). An intramolecular charge transfer from Ce 3+ to POM(W 6+ ) is observed upon visible light excitation, which can initiate photocatalytic oxidative dehydrogenation of a series of primary and secondary aromatic amines in the presence of O 2 (1 atm). Yamamoto et al. [120] showed that selectivity in oxidative coupling of benzene with cyclohexane is light sensitive using the Pd-modified TiO 2 photocatalyst. Under UV irradiation, the photogenerated hole species react with benzene and cyclohexane to form corresponding radicals, thus resulting in non-selective formation of all possible coupling products (phenylcyclohexane, bicyclohexyl, and biphenyl). In contrast, high selectivity towards cross-coupling product (phenylcyclohexane) can be achieved under visible light irradiation due to a ligand-to-metal charge transfer of benzene complex adsorbed on TiO 2 , which produces benzene radical cation and selectively activates cyclohexane to form a cyclohexyl radical that attacks benzene molecule to form phenylcyclohexane. In addition, the Pd cocatalyst suppresses the formation of bicyclohexyl, thus resulting in a high selectivity towards the cross-coupling product. However, such a system shows a slow conversion rate due to poor light absorption under visible light.
Very recently, Cao et al. [121] reported a purely Bi-based semiconductor system (p-BWO nanosheets) for the efficient oxidation of toluene under visible light and ambient conditions (1 bar O 2 ). As demonstrated in Figure 4A, the photocatalyst is constructed by crystalline Bi 2 WO 6 (lattice spacings of 2.72 and 2.73 Å for the (002) and (200) planes, respectively) that is surrounded by amorphous BiOCl substance. Such structure provides numerous active sites at the crystalline/amorphous boundaries (orange and blue arrows). These amorphous BiOCl layers crystallize in situ under a continuous bombardment of high-energy electron beams [ Figure 4B], indicating they are very sensitive to excited electrons. Owing to the unique microstructure for efficient charge separation, the lattice oxygen of the p-BWO nanosheets is very active upon visible light irradiation, which can be involved in the oxidation of toluene (leached out) and eventually healed by molecular oxygen. The p-BWO nanosheets show a 166-fold enhancement in photo-oxidation of toluene compared with that of pristine Bi 2 WO 6 [ Figure 4C].
Selective cracking of aromatic biomass by photocatalysis has also drawn significant attention very recently. Wu et al. [122] showed that CdS quantum dots (QDs) are efficient photocatalysts for the selective cracking of branched-chain of various lignin model compounds via β-O-4 bond cleavage under visible light irradiation and deaerated conditions, as demonstrated in Figure 4D. Meanwhile, due to the relatively weak oxidation power of CdS QDs, the functional groups and the aromatic ring structures remain stable during the  [121,122] . NPs: Nanoparticles; QDs: quantum dots.
photocatalytic process. The optimized CdS QDs photocatalyst with an average particle size of 4.4 nm presents a satisfactory conversion of lignin compound, which also exhibits significant enhancement compared with other commonly used photocatalysts [ Figure 4E and F]. The CdS QDs photocatalyst also shows an enhanced performance compared with that of the CdS NPs counterpart in terms of both conversion rate and stability [ Figure 4G]. More importantly, the CdS QDs photocatalyst is capable of converting native lignins and can be recycled via a facile aggregation-colloidization strategy, which shows huge potential for industrial applications.
Photocatalysis for selective oxidation of aromatics to value-added products has experienced a booming development. Further development in catalytic efficiency and selectivity may be achieved by further optimization of the adsorption of reactants and intermediates. The design of functionalized photocatalysts and cocatalysts may be inspired by a series of classic heterogeneous catalysis works, especially those surface science-related fundamental works. In addition, extensive efforts should be paid to tune the redox potential of the photocatalyst to avoid over-oxidation of specific functional groups and unwanted ring-opening side reactions.

Oxidative coupling reactions
Selective formation of bonds between carbon and other heteroatoms (i.e., N, S, and O) via cross-coupling is one of the important reactions in synthetic chemistry. The oxidative coupling reactions generally employ transition-metal catalysts to withdraw electrons from the reactants, which is accompanied by the addition of oxygen or deprotonation to the reactants. Employing metal-based molecules (i.e., Ru[(bpz) 3 ](PF 6 ) 2 and Ru[bpy] 3 Cl 2 ) as homogeneous photocatalysts has shown that a series of oxidative coupling reactions can be realized under visible light irradiation (i.e., phenol-arene coupling [123] and amines-enol silanes coupling [124] ).
To further enhance the efficiency and the reusability of the homogeneous dye complex, Tan et al. [125] immobilized the Ru[bpy] 3 Cl 2 molecule into a radical MOF, FJI-Y2 for the photocatalytic crossdehydrogenative coupling of N-phenyltetrahydroisoquinoline derivatives with phosphite esters to form αaminoquinoline phosphonates. As demonstrated in Figure 5A and B, the [Ru II (bpy) 3 ] cations are encapsulated into the porous FJI-Y2 MOF via intermolecular interaction between pyridine rings of dye cations and phenyl rings of the MOF ligands. The FJI-Y2 MOF is characterized by multiple 1D channels with a window size of ~8 × 12 Å 2 , which offers an ideal platform with a large surface area and accessible active sites. It is also found that the dye-modified FJI-Y2 photocatalyst shows a higher performance than that of the homogeneous [Ru II (bpy) 3 ]Cl 2 alone. The promotional effect originated from an efficient charge transfer due to intermolecular interactions between the 1,4,5,8-naphthalenediimide (NDI) group of the MOF and the Ru II (bpy) 3 catalyst, as demonstrated in Figure 5C. Under visible light irradiation, the NDI groups are excited to NDI*, which is immediately quenched by the excited Ru II* via a single electron transfer (SET) process. Such process generates a Ru I complex and a radical cation NDI +* , which separately react with an electron donor (here, D, N-phenyltetrahydroisoquinoline) and an oxygen [O] via SET process to form the corresponding radical cation (D +* ) and superoxide radical [O] -* , respectively. Meanwhile, the NDI +* and Ru I complex are relaxed to their ground states. The [O] -* further abstracts an H atom from the D +* to form the desired iminium ion, which undergoes nucleophilic addition to the desired product.
The oxidative coupling reaction can also be realized by traditional metal oxide photocatalyst systems. Zhang et al. [126] synthesized defect-rich WO 3 (R-WO 3 ) nanosheets via calcination of hydrothermally prepared defect-deficient WO 3 (D-WO 3 ) in N 2 atmosphere for aerobic coupling of primary amines to corresponding imines at room temperature [RT, Equation (6)]: Such R-WO 3 material contains numerous small pits on surface, creating a large surface area with more under-coordinated sites (CUS) for catalytic reactions. In high-resolution TEM imagery, lattice disorder and dislocation are observed in the R-WO 3 nanosheets, revealing the presence of high-density defects [ Figure 5D]. The massive defects result in a large number of electrons trapped at oxygen vacancies, as indicated by the significant signal of electron spin resonance (ESR) spectroscopy for R-WO 3 (g = 2.002) compared with that of D-WO 3 [ Figure 5E]. Density functional theory (DFT) calculations indicate that molecular O 2 is chemisorbed at the oxygen vacancy of D-WO 3 surface via an end-on configuration, which is accompanied by a slight electron transfer from WO 3 CUS site to O 2 [ Figure 5F]. The strong coupling between R-WO 3 and O 2 promotes oxygen reduction half reaction, thus accelerating the overall oxidative coupling.
Raza et al. [127] showed that the oxidative coupling of various amines to imines [Equation (6)] can also be realized by monolayer WS 2 nanosheets under visible-light irradiation. The WS 2 nanosheets synthesized via a modified liquid exfoliation method exhibit an average size of 120 nm with a height of 1 nm according to atomic force microscopy analysis, indicating most of the nanosheets are monolayer. The WS 2 photocatalyst shows a similar reaction mechanism to that of the WO 3 system; however, the monolayer WS 2 nanosheets  3 ] photocatalyst via a SET pathway. This figure is used with permission from Elsevier [125] . (D) Atomic-resolution TEM image of R-WO 3 nanosheets. (E) ESR spectra of D-WO 3 and R-WO 3 . (F) Simulated differential charge density for O 2 chemisorbed at a CUS W site of defective WO 3 . This figure is used with permission from the American Chemistry Society [126] . MOF: Metal-organic frameworks; NDI: naphthalenediimide; SET: single electron transfer; ESR: electron spin resonance.
need to be operated under 50 °C. Xu and Fu [128] also developed core-shell structured sulfide-g-C 3 N 4 to realize oxidative coupling of amines. A 4 nm thick shell of g-C 3 N 4shell is coated on the CdS core nanowire via a selfassembly process, which constructs a heterojunction interface for enhanced charge separation of photogenerated electrons and holes. Nevertheless, it has also been noticed that a much faster conversion rate can be achieved by using H 2 O 2 as the oxidant, therefore suggesting that the formation of •O 2 -radical is the rate-determining step for the photocatalytic oxidative coupling of various amines.
Our group recently reported that the identity of metal NP-decorated g-C 3 N 4 is essential in controlling the selectively photocatalytic dehydrogenative homocoupling of primary amines into imines and secondary amines [129] . While Pt NPs drive the dehydrogenative homocoupling of primary amines into corresponding imines, Pd NPs promote self-hydrogenation of the photogenerated imines into secondary amines using the H atoms abstracted from the amines [ Figure 6A]. Mechanism studies reveal that the adsorption of photogenerated imine molecule is relatively strong on Pd compared to on Pt NPs according to temperature programmed desorption (TPD, Figure 6B), thus facilitating the hydrogenation of imine into secondary amine. Moreover, a significant evolution of molecular hydrogen is observed for Pt/C 3 N 4 under irradiation when primary amine and secondary amine are present, whereas no detectable H 2 is spotted for Pd/C 3 N 4 [ Figure 6C and D]. Moreover, a distinct delay of H 2 evolution is observed on Pd/C 3 N 4 (~3 h after immense irradiation, red curve) in the photocatalytic hydrogenation of imine (2a) using isopropanol as the hydrogen donor [ Figure 6E]. This further confirms that the abstracted hydrogen atoms from the amine molecules can be temporarily stored on Pd NPs, thus hydrogenating imine molecules into secondary amine without additional hydrogen donors (i.e., alcohols).  [129] . TPD: Temperature-programmed desorption; MS: mass spectrometry.
Very recently, the dehydrogenative coupling of simple aliphatic alcohol (methanol) to the value-added product [ethylene glycol (EG)] was realized by Xie et al. [130] using the metal sulfide composite. While conventional photocatalysts (i.e., TiO 2 , g-C 3 N 4 , and Cu 2 O) only oxidize methanol into formaldehyde, CO, and CO 2 , CdS NPs and nanorods produce a reasonable amount of EG. Interestingly, both the conversion rate of methanol and the selectivity to EG can be significantly enhanced by loading MoS 2 nanofoam cocatalyst onto the CdS nanorod [ Figure 7A and B]. ESR analysis reveals that mainly hydroxymethyl radicals ( • CH 2 OH) instead of methoxy radicals (CH 3 O•) are generated under irradiation using CdS photocatalyst [ Figure 7C], suggesting a preferential activation of C-H bond rather than O-H bond of methanol by photoexcited holes. Eventually, the photogenerated • CH 2 OH radicals readily desorb from catalyst surfaces for subsequent coupling. This work presents unique photocatalytic C-H activation that may offer an alternative process for cheap aliphatic alcohol conversion in the petroleum industry.
Nevertheless, plasmonic Au-Pd alloy NPs supported on ZrO 2 have also been employed as photocatalyst for a series of C-C coupling reactions (i.e., Sonogashira coupling, Stille coupling, and Hiyama coupling) with high yields [131] . It is proposed that the alloy NPs absorb visible light and transfer excited electrons to the Pd sites to initiate the reactions. Beyond conventional metal-based semiconductor and plasmonic metal NPs photocatalyst systems, Ghasimi et al. [132] demonstrated that the conjugated organic polymer can be used as a Pd-free photocatalyst for efficient Stille coupling reaction to construct aromatic C-C bond under visible light irradiation. The photocatalyst polymer, namely P-Az-B, can be obtained via Suzuki cross-coupling reaction of 1,3-dibromoazulene with 1,3,5-phenyltriboronic acid tris(pinacol) ester, where the 1,3dibromoazulene is a co-monomer to obtain high porosity of the material. The P-Az-B photocatalyst exhibits excellent light absorption with an optical band gap of 2.03 eV. In addition, the highest occupied molecular orbital and the lowest unoccupied molecular orbital (LUMO) of P-Az-B (+1.14 and -1.10 V vs. saturated calomel electrode) are comparable to that of the Ru-based complex, suggesting its compatibility for driving  [130] . ESR: Electron spin resonance.
C-C coupling reactions. Upon visible light irradiation, the photogenerated hole and electron react with aryl stannane and the aryl iodide, resulting in the formation of destannylation radical and aryl halide radical, respectively. Then, the radicals desorb from the photocatalyst and react to create a substitutional intermediate, which finally leads to the formation of C-C bond by releasing the iodide anion.
The cross-coupling of aniline and alcohol for the synthesis of asymmetric secondary amines can also be realized by heterogeneous photocatalytic systems. Shiraishi et al. [133] demonstrated that the Pd/TiO 2 photocatalyst promotes the N-monoalkylation of amine with alcohol via tandem photocatalytic and catalytic reactions: (i) photo-oxidation of alcohol on Pd/TiO 2 ; (ii) condensation of photogenerated aldehyde with amine on TiO 2 ; and (iii) hydrogenation of the produced imine by the photogenerated surface H atoms on the Pd particles from alcohol oxidation. The rate-determining step is the imine hydrogenation, which strongly depends on the number of surface Pd atoms on the triangle site of Pd particles [ Figure 8A]. Meanwhile, the strong adsorption of alcohols on larger triangle sites of Pd NPs should be avoided as it will suppress the imine hydrogenation. Therefore, Pd NPs with an optimum size of 2-2.5 nm contain a relatively larger number of triangle site Pd atoms but do not show strong alcohol adsorption, present the highest activity for imine hydrogenation and promote N-monoalkylation of amine with alcohol [ Figure 8B]. Similarly, the rapid photocatalytic N-alkylation of pharmaceutically relevant amines with alcohols can also be realized by employing a mixed heterogeneous Cu-Au/TiO 2 photocatalyst [134] . By carefully controlling the reaction conditions, selective mono-and di-alkylation of primary amines and the non-symmetrical dialkylation of primary amines to hetero-substituted tertiary amines can be achieved. In this scenario, alcohol is typically used as the solvent, which serves as both the reactant to couple with amine and a hydrogen donor to supply H atoms for the hydrogenation of the photogenerated imines. Therefore, selectivity control over imine and secondary amine and the formation of aldehyde byproducts remain challenges.
A very recent work provides an alternative strategy for selective cross-coupling of amines and alcohols without the use of additional hydrogen donors [135] . This is realized by controlling the identity of the metal cocatalysts supported on TiO 2 to mediate the photocatalytic activation of alcohols, thus separating the stepwise dehydrogenative-condensation process to imine (Rh) and the direct condensation process to secondary amine (Fe, Figure 8C). TPD and ESR analysis reveals that Rh exhibits strong adsorption of alcohol with fast generation kinetics of the • OH radicals [ Figures 8D and E], thus leading to a rapid photooxidation of alcohol into aldehyde and the consecutive condensation of aldehyde with aniline. In contrast, the Fe/TiO 2 displays a weak interaction with alcohol and slow kinetics in • OH formation, resulting in a mild activation of the alcohol for direct condensation with aniline to yield asymmetric secondary amine.  [133] . (C) Metal cocatalyst tuned photocatalytic cross-coupling of aniline and alcohol. (D) TPD analysis of benzyl alcohol adsorption on Rh/TiO 2 and Fe/TiO 2 , respectively. (E) ESR spectra of Rh/TiO 2 and Fe/TiO 2 in DMSO under irradiation. This figure is used with permission from Elsevier [135] . TPD: Temperature programmed desorption; ESR: electron spin resonance; DMSO: dimethyl sulfoxide.
Up to date, the homo-coupling of some model compounds by photocatalysis has been well established and highly efficient photocatalysts have been developed. Further investigations in cross-coupling reactions are crucial for applications that require careful control over the active sites of the catalyst and reaction conditions. This calls for an in-depth understanding of reaction pathways, intermediates, and radical species. One could learn from traditional homogeneous catalysis of the reaction mechanisms to help the design of heterogeneous photocatalyst. Additionally, the quantification of unwanted byproducts and unknown products should not be neglected, as this may give some useful hints for the design of highly selective photocatalysts.

Hydrogenation of CO 2 and CO
Photocatalytic CO 2 reduction is a very hot topic in photocatalysis and photo-electrocatalysis. Various photocatalysts ranging from simple oxides, sulfides, perovskites, MOFs [136] , and plasmonic metal NPs [137] to composites [138] have been developed for CO 2 reduction with or without sacrificial electron donors. The design of cocatalyst has also been extensively investigated to further boost the photocatalytic CO 2 reduction performance [139][140][141] . Since there are a number of in-depth reviews on this topic [136,142] , we focus on the hydrogenation of CO 2 and CO by molecular H 2 and the incorporation of CO 2 into other reactants for the formation of more complicated chemicals.
Sastre et al. [143] showed that the p-type NiO photocatalyst can reduce CO to CH 4 with the presence of water under visible light irradiation at RT. It is also noticed that CO reduction to CH 4 can be significantly enhanced with the presence of H 2 as a reducing agent. In addition, the deposition of elemental carbon (coking) on the photocatalyst surface is also observed. It is therefore supposed that the NiO photocatalyst not only accelerates the CO reduction by water [Equation (7)] but also promotes a Fischer-Tropsch-like process following Equation (8): Based on these findings, the same group further explored the photocatalytic Fischer-Tropsch process directly from CO 2 and H 2 [144] . It has been found that, although NiO exhibits high conversion of CO 2 to CH 4 , the photocatalyst gradually deactivates upon reuse. Meanwhile, nickel supported on porous silica-alumina (Ni/Si x Al y O z ) is an efficient and reusable photocatalyst for the reduction of CO 2 with comparable performance. As demonstrated in Figure 9A, it is proposed that the reaction is initiated via H 2 activation by the photogenerated charge carriers, where Ni-H species are generated to further activate CO 2 to form surface adsorbed, hydrogenated intermediates and eventually release as CH 4 . Since the silica-alumina is inert support, the reaction is more likely a photo-thermal catalytic process that is driven by the localized hot electrons generated from plasmonic absorption of Ni metal NPs. Researchers should be aware that the Ni/Si x Al y O z photocatalyst only produces CH 4 and CO without the formation of hydrocarbons with longer chains. To further promote the photocatalytic Fischer-Tropsch process, Guo et al. [145] deposited worm-like Ru nanostructures on graphene as photocatalyst for the conversion of CO to more complicated hydrocarbons under H 2 atmospheres. Although the Ru/graphene photocatalyst shows a decent selectivity to liquid fuel products under visible light irradiation (C 5+ , > 70%), such a system requires a relatively high operation temperature (150 °C), and the activity is more dependent on heat rather than light.
Very recently, Zhang's group showed that the synthesis of hydrocarbons from syngas (CO and H 2 ) can be realized by photo-thermal catalytic approach at RT [58,[146][147][148][149] . This type of nanostructured photocatalysts is generally derived from hydrogen reduction of layered double hydroxides (LDH, i.e., ZnCoAl, ZnFeAl, and CoFeAl) at elevated temperatures (300 °C -700 °C), yielding metal/metal oxide composites (i.e., Co/Co 3 O 4 , Fe/FeO x , and CoFe alloy) leached out from the LDH and naturally loaded on the corresponding inert support (ZnO, Al 2 O 3 ). The activity and selectivity of these LDH-derived photocatalysts can be tuned by optimizing the annealing temperature and identity of the metal species in the LDH. While the photocatalyst derived from ZnCoAl LDH shows a reasonable CO conversion (15.4%) and selectivity towards light olefin (C 2 -C 4 = , ~19%), the performance can be further enhanced by employing ZnFeAl LDH as the photocatalyst precursor (Con. > 20%, Sel. > 40%) due to a suppression of CO 2 formation. Furthermore, direct hydrogenation of CO 2 can also be realized by fabricating the CoFe alloy NPs from annealing of CoFeAl LDH nanosheets under H 2 atmosphere, as shown in Figure 9B and C. The CoFe alloy NPs synthesized at 650 °C show remarkable selectivity toward hydrocarbons (60% CH 4 , 35% C 2+ ). It should be clear that the catalytic activity of these photocatalysts originates from the photothermal effect, as the temperature of the catalyst bed increases rapidly upon irradiation with the presence of photocatalyst [ Figure 9D].  [144] . (B,C) EDS line-scan and HRTEM image of the CoFe alloy NPs used for photothermal CO 2 hydrogenation to hydrocarbons with H 2 . (D) Temperature profiles during photothermal CO 2 hydrogenation. This figure is used with permission from Wiley-VCH [146] .
For carboxylation with CO 2 , Ye et al. [150] showed that radical chemistry possesses great potential to bring CO 2 to a new stage using either a transition-metal catalyst or a photocatalyst. Nevertheless, incorporation of CO 2 reduction with oxidation of other reactants to synthesize valuable chemicals has also emerged recently. Yang et al. [151] showed that the photocatalytic reduction of CO 2 can be coupled with the oxidation of amine by using Cu/TiO 2 photocatalyst. While CO 2 is reduced to CH 3 OH by photo-induced electrons, benzylamine reacts with photogenerated holes and undergoes dehydrogenation coupling to produce imine. Such a system provides an economic strategy to fully utilize the photogenerated charge carriers. Chen et al. [152] reported the utilization of Cu 2 O/Cu nanocomposite for photocatalytic coupling of CO 2 with benzyl alcohol. Under visible light irradiation, benzyl acetate rather than C1 products is selectively produced over Cu 2 O/Cu, as described in Equation (9). Upon irradiation, the photogenerated holes react with benzyl alcohol to form benzaldehyde with the release of two protons. The protons along with photogenerated electrons interact with surface-adsorbed CO 2 to generate CH 2 O via hydrogenation of the surface-adsorbed CO* intermediate, during which CO 2 undergoes a facile cleavage of the C-O bond on Cu sites. CH 2 O further couples with a neighboring CO* to form surface adsorbed H 2 C-CO. The H 2 C-CO intermediate converts to surfaceadsorbed H 3 C-CO via H abstraction from benzyl alcohol and subsequently undergoes a coupling reaction with the benzyl oxygen ion to produce benzyl acetate. However, researchers should be aware that the Cu 2 O/Cu material transforms to Cu 2 O/CuO after photocatalytic reaction, which needs to be regenerated by NaBH 4 .
Photocatalytic CO 2 reduction is a challenging and intriguing reaction; however, the value of the products and the inherent slow reaction kinetics render its future questionable from an economical viewpoint.
Nevertheless, fundamental investigations are still important in this field, but they should be focused on reaction mechanisms rather than reporting performance. It is also important to use standard procedures for performance testing and reporting.

Hydrogenation of O 2
H 2 O 2 is widely used in many applications including medication, bleaching, and production of organic chemicals. Previously, industrial synthesis of H 2 O 2 was realized by hydrolysis of ammonium peroxydisulfate ([NH 4 ] 2 S 2 O 8 ) that is obtained by electrolysis of ammonium bisulfate (NH 4 HSO 4 ) in sulfuric acid. Nowadays, H 2 O 2 is mainly manufactured by the anthraquinone process, as demonstrated in Equation (10) [153] : where the anthraquinone (i.e., 2-ethylanthraquinone) is hydrogenated to the corresponding anthrahydroquinone and consequently oxidizes back to anthraquinone and produces H 2 O 2 . The economics of this process depends on the Pd-based hydrogenation catalyst, the effective recycling of the extraction solvents, and the cost of quinones.
Since it has been proven that reactive oxygen species including O 2 •-, OH • , and H 2 O 2 are produced during the photocatalytic decomposition of organic pollutants, it is reasonable to consider the photosynthesis of H 2 O 2 by employing a designed semiconductor photocatalyst under mild conditions with appropriate hydrogen sources. Teranishi et al. [154] showed that H 2 O 2 (10 mM level) can be synthesized by simply using Au NPmodified TiO 2 with ethanol as the hydrogen source under aerobic conditions. Kinetic analysis reveals that, while Au NPs on TiO 2 promote the formation of H 2 O 2 compared with that of pristine TiO 2 or Pt/TiO 2 , the decomposition rate of H 2 O 2 on Au/TiO 2 is also significantly suppressed. It is therefore proposed that the hydrogenation of O 2 and the degradation of H 2 O 2 take place on the metal surface and the TiO 2 surface, respectively, and surface engineering of the photocatalyst material is the key to further optimize the synthesis of H 2 O 2 . The photocatalytic H 2 O 2 production over Au/TiO 2 can be further enhanced by adding borate or phosphate in alkaline water solution [155] , which is ascribed to an anion-mediated hole transfer from Au/TiO 2 to electron donor (methanol). In addition to the anion effect, Moon et al. [156] showed that the metal promoter on TiO 2 can be replaced by reduced graphene oxide (rGO) for the photocatalytic production of H 2 O 2 . Interestingly, by adding cobalt salt to the phosphate-containing catalyst suspension, cobalt phosphate complex (CoPi) that is often served as an electrocatalyst can be formed in situ on the rGO/TiO 2 surface. The loading of CoPi as cocatalyst further reduces the overpotential of water oxidation, therefore enabling photocatalytic production of H 2 O 2 even without organic hydrogen resources (i.e., alcohols).
More efforts have been devoted to realizing the selective synthesis of H 2 O 2 under visible light conditions. Shiraishi et al. [157] employed g-C 3 N 4 as the visible light response photocatalyst to generate H 2 O 2 selectively (> 90%) with the presence of alcohol and O 2 . Mechanistic studies reveal that the high selectivity to H 2 O 2 originates from the efficient formation of surface-adsorbed 1,4-endoperoxide species (C-O-O-N), which suppresses the one-electron reduction path of O 2 to superoxide radicals. He et al. [158] showed that the dimension of carbon support materials is crucial in tuning the recombination rate of charge carriers and light absorption properties, which is essential for boosting the evolution rate of H 2 O 2 . Li et al. [159] demonstrated that the performance of g-C 3 N 4 can be further improved by creating carbon vacancies, which reduce the bandgap and promote the delocalization of excited electrons of the material. The carbon vacancies not only provide more sites for interfacial charge transfer from the photocatalyst to adsorbed O 2 , but they also alternate the H 2 O 2 generation pathway from a two-step single-electron process to a one-step two-electron direct reduction.
Kim et al. [160] further explored the possibility of coupling industrial H 2 O 2 synthetic process with photocatalysis. As demonstrated in Figure 10A, the anthraquinone molecule [anthraquinone-2-carboxylic acid (AQ)] is anchored onto g-C 3 N 4 via a catalyst-free, direct dehydration coupling of the -COOH of AQ and the -NH 2 of g-C 3 N 4 . Upon visible light excitation, isopropanol is oxidized on the g-C 3 N 4 surface to acetone and releases protons. Meanwhile, the photogenerated electrons and abstracted protons transfer to the anchored AQ molecule, reducing it to surface-anchored anthrahydroquinone (AQH 2 ). Eventually, the anchored AQH 2 is oxidized back to AQ in the presence of O 2 , resulting in the production of H 2 O 2 . The efficient anchoring of AQ has been confirmed by X-ray photoelectron spectroscopy (XPS, Figure 10B). While the C1s peaks of g-C 3 N 4 are assigned to the adventitious carbon and defect-derived sp2-carbon (284.8 eV) and the N-C = N bond in the triazine units (288.2 eV), a strong aromatic C-C bond (284.8 eV) is observed when AQ is coupled with g-C 3 N 4 . In addition, Fourier transform infrared spectroscopy (FT-IR) has also spotted increases in absorption at 1278 and 1676 cm -1 , which correspond to C-N and C = O stretching vibrations due to the formation of amide bond. Owing to the efficient interfacial charge transfer and the superior activity of AQ in O 2 reduction, such an anchored system exhibits a high apparent quantum yield in H 2 O 2 production of 19.5% at 380 nm, which is significantly higher than that of pristine g-C 3 N 4 or g-C 3 N 4 modified with metallic co-catalysts.
Very recently, Teng et al. [161] showed that a single Sb atom dispersed on g-C 3 N 4 (Sb-SAPC) can achieve a remarkable apparent quantum yield of 17.6% for H 2 O 2 synthesis under 420 nm irradiation. The solar-tochemical conversion efficiency of Sb-SAPC15 reaches 0.61%, which is higher than most water-splitting photocatalysts [ Figure 10C]. TEM imaging shows that high-density Sb atoms are uniformly dispersed over the entire carbon nitride matrix [ Figure 10D]. Raman spectroscopy has been performed to identify the intermediate in the photocatalytic process with the presence of isopropanol as the electron donor [ Figure 10E]. A new absorption at 855 cm -1 is observed for the Sb-SAPCs, which can be assigned to the O-O stretching of an Sb-OOH species that corresponds to the end-on adsorption configuration of molecular oxygen. Such end-on adsorption notably suppresses the 4eoxygen reduction reaction, thus leading to a high selectivity of the 2eprocess for H 2 O 2 synthesis.
Photocatalysis provides an alternative approach for the on-site synthesis of H 2 O 2 , which is ideal for coupling with some specialized fine chemical syntheses that require a relatively low concentration of H 2 O 2 . This requires leveling up the concentration of photogenerated H 2 O 2 from ppm level to sub wt% level. One possible solution is coupling the oxygen reduction with water oxidation to produce H 2 O 2 via both cathodic and anodic reactions using designed photocatalysts. In addition, the degradation kinetics of photogenerated H 2 O 2 is worthy of investigation. It is also necessary to evaluate the poisoning effect of H 2 O 2 at elevated concentrations for the optimization of photocatalysts.

Hydrogenation of organic molecules
Early investigations showed that fine TiO 2 NPs and ZnS quantum dots photocatalyst can be used to hydrogenate a series of organic molecules (i.e., propene, aromatic ketones, and amines) upon UV excitation [162][163][164] , although the light utilization, quantum efficiency, conversion, and selectivity are far from satisfactory in general. To overcome these challenges, many efforts have been focused on surface engineering of photocatalyst materials. Hao et al. [164] reported surface engineered heterogeneous  [161] . XPS: X-ray photoelectron spectroscopy.
photocatalysts for the hydrogenative fixation of N 2 to NH 3 . Zhang et al. [165] showed that functionalization of TiO 2 with metal NPs can significantly promote the hydrogenative cyclization of levulinic acid (LA) into γvalerolactone (GVL, a potential green fuel) selectively using isopropanol as the hydrogen source, as demonstrated in Equation (11): Here, isopropanol is dehydrogenated to acetone as the main product, donating a proton to LA that is cyclized to form GVL via the formation of acetyl propionyl radical. It is observed that while the polymorph composition of TiO 2 influences the conversion rate of LA, the identity of metal NPs controls the selectivity. Au NPs exhibit the highest LA conversion (79%) and GVL selectivity (85%) compared with that of Pd and Pt promoters, which is compatible with thermo-catalytic hydrogenation approach using Cu/ZrO 2 and Ru/C catalysts.
Lots of efforts have also been made to initiate hydrogenation reactions upon visible light radiation. Kohtani et al. [166] showed that, by anchoring organic dye molecules on pristine TiO 2 NPs, hydrogenation of selected aromatic ketones to corresponding alcohols with the presence of a hydrogen donor (trimethylamine) can be achieved under visible light excitation (> 400 nm). Although the utilization of metal-free dyes (fluorescein and rhodamine B) is an economical and feasible solution, the poor stability limits its applications. Jiao et al. [167] developed a Pd supported on β-SiC photocatalyst for the hydrogenation of furan molecules under visible light excitation at RT [Equation (12)]. Although β-SiC is a narrow bandgap semiconductor (~2.4 eV), it is found that loading of Pd NPs (3 wt%) is essential to initiate the efficient hydrogenation of furan and its derivatives. The authors proposed that a Mott-Schottky contact of SiC and Pd enhances the charge transfer of photogenerated electrons from SiC to Pd NPs, thus promoting the catalytic activity; however, the photothermal catalytic approach induced by the plasmonic absorption of Pd NPs cannot be ruled out.
The same group further employed SiC with Au NPs for the selective hydrogenation of cinnamaldehyde using isopropanol as a hydrogen source under deaerated conditions [Equation (13)] [168] : Figure 11A presents that Au NPs with a particle size ranging from 3-8 nm are evenly deposited on the SiC support. Remarkably, the Au/SiC system exhibits excellent selectivity and reasonable conversion in the whole visible light range, as demonstrated in Figure 11B. A very high turnover frequency (TOF, 487 h -1 ) can be achieved at optimum conditions (30 mg catalyst and 20 mg KOH in 10 mL isopropanol and Xe-lamp with light intensity of 1.0 W•cm -2 ). Nevertheless, the Au/SiC photocatalyst is capable of hydrogenating a series of α,β-unsaturated aldehydes to their corresponding unsaturated alcohols with high conversion and selectivity. It should also be noted that, in contrast to the hydrogenation of furans by Pd/SiC, the hydrogenation of cinnamaldehyde by Au/SiC is definitely a photo-thermal catalytic process that is induced by the plasmonic absorption of Au NPs. While the oxidation of 2-propanol takes place on the positively charged Au NPs surface, α,β-unsaturated aldehydes are adsorbed on the Au/SiC interface and reduced by the photogenerated electrons to unsaturated alcohols.
Shiraishi et al. [169] showed that the photocatalytic selective hydrogenation of a series of nitroaromatics to anilines with isopropanol as the hydrogen donor is sensitive to the polymorph composition of TiO 2 [Equation (14)]: Interestingly, while anatase and P25 TiO 2 show poor activity and selectivity towards aniline, photoactivated rutile TiO 2 samples all show high performance in selective hydrogenation of nitroaromatics. It is proposed that the adsorption of nitroaromatics on rutile is responsible for the enhanced performance. As demonstrated in Figure 11C, the adsorption and hydrogenation of nitroaromatics on the photoexcited TiO 2 surface take place via either asymmetric adsorption on hydroxyl (TiOH) sites or symmetric adsorption on bridging oxygen vacancy (O b vac.) sites. While the symmetric adsorption site can donate two photogenerated electrons to a nitroaromatic molecule, the asymmetric adsorption site only provides one electron at once, thus resulting in a poor performance in reduction. Diffuse reflectance infrared Fourier DRIFT spectra of nitrobenzene adsorbed on selected anatase (A) and rutile (R) TiO 2 at 303 K. This figure is used with permission from the American Chemistry Society [168,169] . NPs: Nanoparticles; DRIFT: diffuse reflectance infrared Fourier transform.
transform (DRIFT) spectroscopy has confirmed a significant density of symmetric adsorption sites on different rutile TiO 2 compared with that of anatase samples [ Figure 11D], therefore leading to a fast aniline formation rate with high quantum yields (> 25%).
New materials have also been developed to drive the hydrogenation of nitroarenes under visible light conditions. Toyao et al. [170] synthesized an amino-functionalized Ti(IV)-MOF modified with Pt NPs (Pt/Ti-MOF-NH 2 ) for this purpose with reasonable stability. In situ ESR analysis reveals that the excited electrons on the organic linker transfer to the deposited Pt NPs via the titanium-oxo clusters, which promote the spatial charge transfer, thus resulting in enhanced performance. At the optimized Pt loading (1.5 wt%), Pt/Ti-MOF-NH 2 shows enhanced performance in both H 2 production and nitrobenzene reduction. However, low quantum efficiencies are observed for this system (~1%-2% under visible light irradiation), probably due to the rapid evolution of molecular H 2 that competes with the hydrogenation of nitrobenzene. Gao et al. [171] employed nanosized Ni 2 P as the cocatalyst to modify CdS as a photocatalyst to realize the hydrogenation of a series of nitroarenes with Na 2 S/Na 2 SO 3 as sacrificial electron donors. In this system, water donates a proton to hydrogenate nitroarenes, as the Na 2 S/Na 2 SO 3 contains no proton. It is also observed that the photocatalytic H 2 evolution competes with the photosynthesis of aniline on the Ni 2 P surface. Therefore, further surface engineering of the photocatalyst to suppress the H 2 evolution is essential to enhance the reduction of nitroarenes.
Very recently, Xiao et al. [172] demonstrated that chemoselective hydrogenation of various nitroarenes to corresponding anilines can be achieved by simply using the pristine g-C 3 N 4 photocatalyst under visible light radiation. The process shows excellent yields (82%-100%) and is even applicable to gram-scale reactions. However, it should be noted that such a process is operated at elevated temperature (70 °C-90 °C) and requires the addition of hydrazine hydrate (N 2 H 4 ·H 2 O) as the hydrogen source and hole scavenger (strong reducing agent with very high toxicity), which hinders its application in large scale due to safety issues. Nevertheless, Huang et al. [173] showed that Au supported on ZrO 2 photocatalyst can initiate the hydrogenation of nitroaromatics under visible light with formic acid and water as the hydrogen source. Since ZrO 2 is a large bandgap semiconductor (> 5 eV), the activity in visible light spectrum originates from the plasmonic excitation of Au NPs. In addition, Au/ZrO 2 is a versatile photocatalyst for the selective hydrogenation of various unsaturated bonds (i.e., C=C, C≡C, C=O, and C=N) under mild reaction conditions. Theoretical calculations reveal that direct photoexcitation of hybridized orbitals that are generated from chemisorbed reactants and metals is the driving force of the hydrogenation reaction.
Heterogeneous photocatalysis provides an ambient platform for hydrogenation reactions that employs hydrogen-containing molecules instead of high-pressure H 2 . Such a mild reaction condition may also benefit semi-hydrogenation of alkynes and selective hydrogenation of complicated pharmaceutical molecules with vulnerable functional groups. In any case, the utilization of abundant and safe hydrogen donors (i.e., alcohols, ammonia) is recommended rather than the use of toxic, expensive, and risky hydrogen sources (e.g., hydrazine). The design of cocatalysts with tunable hydrogen adsorption energy is also vital for efficient and selective hydrogenation, which may be promoted by the experience of heterogeneous hydrogenation investigations.

Reductive coupling reactions
Reductive coupling reactions (i.e., C-C, C-N, N=N) via homogeneous catalysis have been well investigated [174][175][176][177][178][179][180][181][182][183][184] , especially in transition metal-catalyzed cross-coupling reactions promoted under visible light [183] . However, the need for expensive catalysts, catalyst separation, harsh reaction conditions, and toxic reducing agents (i.e., PH 3 , Tl, and NaTeH) limit their applications. Therefore, there is a big demand to develop heterogeneous photocatalyst systems that can realize coupling reactions selectively under ambient conditions. Zhang et al. [184] demonstrated the effectiveness of semi-heterogeneous metallaphotocatalysis using polymeric carbon nitride photocatalyst for C-N cross-coupling reactions. Moreover, Wang et al. [185] showed that, by constructing unique plasmonic Au-Pd nanostructures, Suzuki coupling reaction of a series of aromatic compounds can be realized via harvesting visible-to-near-infrared light [Equation (15)]: Here, the Au-Pd nanostructures consist of Au nanorods for light harvesting via plasmonic excitation, whereas the supported Pd NPs, through seeded growth, serve as catalytic active sites that are frequently used in conventional Suzuki coupling reactions. In addition, cetyltrimethylammonium bromide is required to bring bromobenzene into the aqueous reaction solution under basic conditions. Such a system shows a rapid temperature rise under infrared illumination, revealing a photo-thermal catalytic behavior that originated from the oscillation of localized hot electrons. The higher efficiency observed for the photoinduced coupling compared with that of the thermally induced process indicates that the local temperature may be significantly higher on the plasmonic catalyst surface than in the solution. More importantly, the Au-Pd nanostructures have been further tailored to directly harvest sunlight to drive coupling reactions, which show industrial potential for fine chemical synthesis. This design concept of Au-Pd was also employed by Xiao and coauthors for various C-C coupling reactions [131] . However, the conversion rate in terms of turnover frequency, selectivity, and quantum efficiency still needs to be enhanced for applications. Lanterna et al. [186] later showed that the plasmonic Au NPs can be supported on TiO 2 to realize visible lightinduced reductive C-C coupling of a variety of substituted benzyl bromides [Equation (16)]: Here, the reductive coupling of benzyl bromides requires careful selection of both electron donors and solvents, where the combination of diisopropylethylamine and CH 2 Cl 2 yields the optimum photocatalytic performance. Although the neat Au/TiO 2 photocatalyst system shows decent conversion and selectivity to desired products in general, further engineering of the material is needed to improve the stability.
Li and coworkers recently depicted the rational design of metal cocatalyst for photocatalytic dehalogenative coupling of benzyl bromide based on the Sabatier principle [187] . The optimal activity and selectivity require appropriate adsorption energies of the intermediates (benzyl radical, bromine atom, and surface adsorbed hydrogen atoms (H ads )) on the metal cocatalyst. DFT calculations predict that Cu is the promising candidate for this reaction among a series of transition metals owing to the optimized adsorption of benzyl radical and Br atom [ Figure 12A]. This prediction is confirmed experimentally, where Cu indeed presents high conversion and selectivity in homocoupling of benzyl bromide without the formation of toluene [ Figure 12B]. Remarkably, this design strategy can be generalized for the modification of other photocatalysts. The Cu deposited on g-C 3 N 4 (Cu/C 3 N 4 ) presents a decent performance towards the conversion of benzyl bromide into bibenzyl under visible light irradiation.
Filippini et al. [188] reported the amorphous g-C 3 N 4 (am-CN) material for visible light-driven C-C bond formation towards the synthesis of valuable perfluoroalkylated intermediates from the corresponding iodides. The binding of perfluorobutyl iodide (C 4 F 9 I) occurs via halogen bonding with the N atom of the am-CN. This step affects the rate of the halogen dissociation of C 4 F 9 I to form the •C 4 F 9 radical [ Figure 12C]. The affinity of reagents toward the solid catalyst is further characterized by the T 1 /T 2 ratio in 19 F NMR relaxation measurements. The am-CN presents a higher surface affinity for the fluorinated substrate compared with other CN-based photocatalysts and thus shows the highest activity for this reaction [ Figure 12D].
Beyond C-C coupling, the applications of supported metal NPs as photocatalysts have also been extended into N=N coupling reactions to produce azoxybenzenes and azobenzenes, which are important precursors for the dye, medical, and electronic industries. Zhu and coauthors reported the successful synthesis of azoxybenzenes and azobenzenes using supported Au and Cu NPs photocatalysts from nitrobenzene using isopropanol as the hydrogen source under visible light irradiation and basic conditions, as demonstrated in Equation (17) [189][190][191] :  [187] . (C) Picture of molecular adsorption on CNs. (D) T 1 /T 2 ratio of different moieties of perfluorohexyl iodide in various CN-based materials. This figure is used with permission from American Association for the Advancement of Science [188] .
Here, the nitroaromatics undergo a gradual removal of oxygen by photogenerated protons and electrons via the formation of nitrosobenzene (NBS) and N-phenylhydroxylamine (NPH) radicals, which then generate azoxybenzene via dehydration coupling and eventually to azobenzene via removal of oxygen [192] . The selectivity of azoxybenzene or azobenzene can be tuned by controlling the chemical composition of the plasmonic metal NPs. While Au NPs supported on ZrO 2 exhibit high selectivity towards azobenzene, the Ag-Cu alloy NPs supported on ZrO 2 favor the selective formation of azoxybenzene. In addition, the selectivity can also be controlled by tuning the reaction temperature and employing Cu supported on graphene photocatalyst, though the activity decreases at low temperatures. These findings extend the potential application of the plasmonic metal NPs in photothermal catalysis for fine chemical synthesis; however, the high cost, low efficiency, and poor selectivity of such catalytic systems at RT limit their applications.
N=N coupling reaction can also be realized by using conventional semiconductor photocatalysts. Pal et al. [193] showed that the classical CdS quantum dots can be employed to reduce nitrobenzene to azoxybenzene with relatively high selectivity (68%) under visible light irradiation. It has been found that CdS with a small particle size (~2.8 nm) and the loading of Rh as cocatalyst are necessary to achieve a higher activity and selectivity towards azoxybenzene. Recently, our group showed that a series of azo-and azoxyaromatic compounds can be selectively synthesized from their corresponding nitroaromatics via controlling the irradiation wavelength by simply using the pristine g-C 3 N 4 under basic conditions with isopropanol as the hydrogen source [194] . Firstly, nitrobenzene can be converted into NBS and NPH. NBS and NPH are then further reduced to azoxy-and azo-aromatics. The azo-aromatics may be photoreduced into amines [ Figure 13A]. While azobenzene is gradually formed in a two-step reduction of nitrobenzene under 410 nm irradiation with azoxybenzene as the intermediate [ Figure 13B], azoxybenzene is the sole product throughout the 450 nm irradiation [ Figure 13C]. Notably, high conversion of nitrobenzene (> 95%) and  [191] .  [48] . TPD: Temperature-programmed desorption. selectivity towards both desired products (> 94%) are reached for a series of nitroaromatics with different functional groups, which broaden its applications for the synthesis of complicated azo-and azoxycompounds.
Moreover, the photocatalytic reductive coupling of nitroaromatics can be realized under solar irradiation at a large scale (80 L in volume) with high selectivity to azoxybenzene and a reasonable conversion. The origin of the activity is associated with the adsorption energy of surface adsorbed hydrogen atoms (H ads ) that are generated by photo-oxidation of isopropanol, as confirmed by TPD of the pre-irradiated photocatalysts [ Figure 13D]. While no desorption peak of H 2 over active g-C 3 N 4 indicates weakly bonded H ads on the catalyst surface, two obvious H 2 desorption peaks observed over the inactive amorphous g-C 3 N 4 (A-g-C 3 N 4 ) suggest strong adsorption of the photogenerated H ads on A-g-C 3 N 4 . Therefore, weakly bonded H ads on g-C 3 N 4 promotes the reduction of surface adsorbed nitrobenzene to form azo-or azoxy-benzene.
Although the g-C 3 N 4 system shows great potential in photocatalytic reductive N=N coupling reactions, the low quantum efficiency (1.4% at 410 nm and 0.4% at 450 nm irradiation) and the need for a basic environment limit its application at a large scale. Since the photocatalytic hydrogen abstraction from alcohol is the initial step in reductive coupling, we expect that a photocatalyst coupled with engineered basic sites can improve the photocatalytic performance. Therefore, we developed basic-site-engineered bismuth oxybromide [Bi 24 O 31 Br 10 (OH) δ ] nanosheets to boost nitrobenzene conversion under visible light irradiation and base-free conditions [48] . As presented in the HRTEM image of Figure 13E, the Bi 24 O 31 Br 10 (OH) δ nanosheets exhibit two types of lattice fringes (interplanar spacing of 0.278 and 0.299 nm) that match the (117) and (213) lattice planes of classical crystalline Bi 24 O 31 Br 10 , respectively. CO 2 -TPD reveals that the freshly prepared Bi 24 O 31 Br 10 (OH) δ nanosheets present a broad distribution of active sites with different basicities, whereas the typical BiOBr and g-C 3 N 4 show almost no basicity [ Figure 13F]. Indeed, the basicsite-engineered Bi 24 O 31 Br 10 (OH) δ nanosheets exhibit a robust hydrogen abstraction from isopropanol and fast hydrogen transfer from the catalyst to the adsorbed nitrobenzene according to in situ spectroscopy analysis, thus resulting in remarkable quantum efficiencies under visible light irradiation (42% at 410 nm and 32% at 450 nm). Nevertheless, the Bi 24 O 31 Br 10 (OH) δ nanosheets exhibit excellent performance in upscaling and stability under solar irradiation. As depicted in Figure 13G, a complete conversion of nitrobenzene selectively to azoxybenzene and azobenzene (> 95%) in a sequential fashion is achieved under solar irradiation by employing the Bi 24 O 31 Br 10 (OH) δ photocatalyst, implying that the selective synthesis of two important products can simply be tuned via controlling irradiation time without the need of further separation process. In comparison, the g-C 3 N 4 only converts ~20% of nitrobenzene under similar reaction conditions, which is less efficient than that of the Bi 24 O 31 Br 10 (OH) δ nanosheets.
Besides that, a very recent work shows that shape-engineered AgGaO 2 delafossites with dominant (001) and (012) facets are efficient photocatalysts for efficient and selective photocatalytic redox coupling reactions under visible light [ Figures 14A and B] [195] . Efficient separation of the photogenerated charge carriers is observed over the two facets. The flat AgGaO 2 (F-AGO) with electron-rich (001) facets shows an obvious enhancement over the stretched AgGaO 2 (S-AGO) with hole-rich (012) facets in the reductive homocoupling of benzyl bromide and N-N coupling of nitrobenzene [ Figure 14C]. In contrast, S-AGO performs better than F-AGO in the oxidative homocoupling of aniline and C-N coupling of amines [ Figure 14D]. DFT calculation reveals that the difference in work functions of the (001) and (012) facets is the driving force for the tuned surface redox chemistry, which offers a noble metal-free strategy for the engineering of delafossite-based photocatalysts for synthetic chemistry.
Similar to the photocatalytic oxidative coupling reactions, the development of cross-coupling is far behind the homo-coupling investigations due to the lack of in-depth mechanistic understanding of reaction pathways. Thus, significant efforts should be paid to realize some representative cross-coupling reactions for applications.

CONCLUSION AND OUTLOOK
In this review, we summarize recent advances in photocatalytic synthesis from an application-orientated prospective to illustrate the development in the design of photocatalyst materials. Various reactions catalyzed by nanostructured photocatalysts are categorized into oxidative and reductive reactions, which are further subdivided into alcohol oxidation, epoxidation, oxidation of aromatics, oxidative coupling, hydrogenation reactions, and reductive coupling. The advantages and limitations of these nanostructured photocatalyst materials, including metal oxides, supported plasmonic metal nanostructures, organic polymers, anchored homogeneous catalysts, and dye-sensitized heterostructures, are presented. In addition, fundamental understandings of specific reaction mechanisms are demonstrated to guide the rational design of nanostructured photocatalysts.
Although many challenging reactions can be realized by employing various engineered heterogeneous photocatalyst materials, there are still critical challenges that need to be solved for scaling-up applications. First, a high selectivity, rather than a fast conversion rate, is crucial for all types of reactions from the industrial perspective, as the energy-consuming separation process may be eliminated. For photocatalytic oxidation processes, strong oxidants should be avoided to prevent over-oxidation. For the synthesis of complicated organic molecules, it is also desirable to have a control over the stereo-selectivity, which may be solved by anchoring the designed homogeneous redox catalyst with heterogeneous photocatalyst materials. The selectivity control calls for an in-depth understanding of specific reaction mechanisms to aid the design of engineered nanostructured photocatalyst materials. Besides, it is paramount to avoid using harsh reaction conditions and toxic compounds for reduction reactions (i.e., strong basic or acidic conditions, high temperature, strong reductants, and toxic solvents), as these diminish the advantages of the photocatalytic process. The mild reaction condition is also essential for recycling used photocatalysts and solvents. Nevertheless, the design and development of a continuous reactor (i.e., flow cell) customized for photocatalytic applications are urgently needed to evaluate photocatalyst materials and the scaling-up of photocatalytic processes.

Authors' contributions
Prepared the manuscript: Li Y, Su R Corrected the manuscript: Zhang D, Qiao W, Xiang H, Besenbacher F, Li Y

Availability of data and materials
Not applicable.

Financial support and sponsorship
R.S. would like to thank the NSFC (projects number: 21972100) and the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD, project number: NH10800120) for financial support. R. S. acknowledged Significant funding from Synfuels China Technology Co. Ltd.

Conflicts of interest
All authors declared that there are no conflicts of interest.