Supramolecular Coordination Cages for Artificial Photosynthesis and Synthetic Photocatalysis

Because sunlight is the most abundant energy source on earth, it has huge potential for practical applications ranging from sustainable energy supply to light driven chemistry. From a chemical perspective, excited states generated by light make thermodynamically uphill reactions possible, which forms the basis for energy storage into fuels. In addition, with light, open-shell species can be generated which open up new reaction pathways in organic synthesis. Crucial are photosensitizers, which absorb light and transfer energy to substrates by various mechanisms, processes that highly depend on the distance between the molecules involved. Supramolecular coordination cages are well studied and synthetically accessible reaction vessels with single cavities for guest binding, ensuring close proximity of different components. Due to high modularity of their size, shape, and the nature of metal centers and ligands, cages are ideal platforms to exploit preorganization in photocatalysis. Herein we focus on the application of supramolecular cages for photocatalysis in artificial photosynthesis and in organic photo(redox) catalysis. Finally, a brief overview of immobilization strategies for supramolecular cages provides tools for implementing cages into devices. This review provides inspiration for future design of photocatalytic supramolecular host–guest systems and their application in producing solar fuels and complex organic molecules.


INTRODUCTION
Chemical photocatalysis in a broad context aims at using light as a green reagent for driving reactions in a sustainable fashion, thus contributing to combat climate change and to a sustainable economy. 1 In this sense, light-driven chemical conversions could be used in two main directions: (1) for synthesizing building blocks that could be used to produce bulk and/or fine chemicals such as pharmaceuticals 2−4 and (2) to produce solar fuels that can contribute to transforming the current global energy supply into one that is sustainable. 5 Although there is strong interest in the application of both directions, a detailed mechanistic understanding of the photochemical processes is often still lacking and their application remains underexplored. Whereas both directions are very different in scale and reaction scope, their elemental (photochemical) steps are similar. In that regard, both photochemical research fields can benefit from each other's insights and detailed understanding. Much of this research has been inspired by natural photosynthesis, which is responsible for the global oxygen content as well as most of Earth's energy stocks in the form of sugars and fossil fuels. Mimicking principles of natural photosynthesis is a key part in photocatalytic research to pursue light-driven reactions efficiently. 6 Several proteins, cofactors, and metal-centers are involved in photosynthesis, all of them purposefully having their own function. 7 Overall efficiency is achieved due to a high degree of spatial organization, cooperativity, and chemical logistics, i.e., providing the right components at the right place and time and matching energy levels. Due to the unique chemical environment that is induced by the protein around active sites, nature is able to achieve high control over the activity and selectivity of photosynthesis and many other chemical transformations. In short, increased local concentration, substrate preorganization, as well as cooperativity of several functions within the protein, make them excellent catalysts. 8 Supramolecular self-assembly is a powerful strategy to build similarly complex and well-organized structures from simple building blocks, thus mimicking nature's principles. 9,10 Using metal ions and organic linkers, discrete coordination cages can be obtained (Figure 1). 11−22 Coordination cages provide a single, defined cavity, which is easily accessible by substrates from bulk solution through dynamic exchange. 23,24 Their homogeneous nature allows processing and mechanistic studies in solution. Coordination cages have extensively been used for enzyme-mimicking catalysis, 25−33 making use of substrate preorganization, stabilization of reactive intermediates, and increased local concentration. For photocatalysis, redox-active moieties 34,35 and light-absorbers such as organic dyes or metalbased photosensitizers need to be integrated, 36 which has been recently reviewed. 37−40 Additionally, other types of supramolecular assembly strategies used in chemical photocatalysis to achieve efficient photo(redox) reactions have been reviewed elsewhere. 41,42 Herein we focus on coordination cages with application in light-driven catalysis. After introducing general principles, we will discuss photocatalysis using coordination cages within in two sections: (1) application of cages in the synthesis of solar fuels with the aim of storing (solar) light energy in molecules, and (2) application of cages in the synthesis of complex organic molecules. While the challenge in the first part mostly lies in managing multi-electron processes, the latter aims at enabling novel and sustainable synthetic strategies in the production of high-value chemicals. The sections will be structured according to the position of the photosensitizer (PS, in this review, the unit that absorbs the light) within the cage to generate the photocatalyst (in the review, the complete system needed for the light driven conversion). At the end, we will analyze which strategies lead to most successful generation of photoproducts and how such cages can potentially be integrated into devices.

Artificial Photosynthesis
Artificial photosynthesis aims at the conversion of solar energy into chemical fuels, often referred to as solar f uels, by mimicking principles of natural photosynthesis. Three main functions are essential in natural photosynthesis, and these are the central features to be mimicked by artificial systems: (1) lightharvesting, (2) charge-separation, and (3) redox catalysis. 43 Natural photosynthesis is a very complex process comprising multiple light and dark reactions, and the reaction centers are fostered by a manifold of cofactors. Water splitting takes place in photosystem II (PSII). Protons and electrons generated through light-driven water splitting are used to produce energy-rich ATP and NADPH, which are the chemical fuels used to drive consecutive dark reactions fixating carbon dioxide. 44 Lightdriven water splitting in PSII is achieved by the excitation of a distant chlorophyll (P680), which transfers the exciton to a quinone derivate about 26 Å away from the photosensitizer (Figure 2a). 7 The oxidized photosensitizer then accepts electrons one-by-one from the catalytically active Mn 4 O 4 cluster via a redox-active tyrosine residue. Spatial organization of the different components involved is crucial, as it minimizes unproductive reaction pathways while maximizing directed charge separation. Artificial photosynthesis simplifies these schemes: protons and electrons generated from water-splitting are directly used for the production of molecular hydrogen, carbon dioxide reduction, or for the reduction of other substrates. Likewise, water oxidation can in principle be replaced by the oxidation of organic matter, e.g., biomass, 45,46 plastic waste, 47 and small organic molecules. 48 While the combination of both an oxidation reaction and a reduction reaction in one device or material is the ultimate goal of artificial photosynthesis, research toward the design and optimization of catalysts for energy conversion is typically focused on one of the half-reactions. This can be done either by chemical oxidation/reduction, by electrochemistry, or by lightdriven catalysis using a photosensitizer (PS) and a sacrificial oxidant/reductant together with the catalyst. For instance, in light-driven proton reduction, a sacrificial reductant such as triethylamine (TEA) is used to supply the excited photosensitizer with electrons and to close the photoredox cycle. Similarly, in light-driven water oxidation, a sacrificial oxidant such as Na 2 S 2 O 8 is applied. 49 In a complete and functional system for solar-driven water splitting, both water oxidation and proton reduction (or alternative oxidation/reduction reactions) are coupled. For this, various designs are possible, including catalyst-PS colloids suspended in aqueous electrolyte. 50 An alternative design more common in research laboratories and conceptionally similar to solar cells is the photoelectrochemical cell (PEC). The first example of such an artificial water splitting system generating molecular oxygen and hydrogen was described already in 1972 by Honda and Fujishima. This system was based on a TiO 2 anode and a platinum cathode conducting water splitting under UV irradiation. 51 Inspired by this pioneering study and by the design of dye-sensitized solar cells, dye-sensitized photoelectrochemical cells (DS-PECs) have been developed. 52 A DS-PEC combines an oxidation reaction with a reduction reaction, possibly both driven by light ( Figure  2b). The photocathode is made of a p-type semiconductor (e.g., NiO), on which molecular reduction catalysts and photosensitizers are immobilized. The photoanode is based on an ntype semiconductor (e.g., TiO 2 ), functionalized with a molecular water oxidation catalyst (WOC) and a photosensitizer. Energy levels of the different components need be balanced carefully in order to obtain an overall driving force for the paired redox reactions. The overall light to fuel efficiency of a DS-PEC is determined by (1) light harvesting followed by charge separation (preventing charge recombination), (2) catalytic rates on each side, and (3) overpotentials. The overpotential is the extra potential added to the thermodynamic potential required to drive each reaction and is dependent on the catalyst used. In order to achieve efficient overall light-to-fuel conversion over a sufficiently long period of time, highly stable and fast catalysts are required. A major issue in DS-PECs is unproductive charge recombination, which can occur via different pathways: (1) decay of the excited electron at photosensitizer to the ground state before injection into the semiconductor, (2) charge recombination from semiconductor to the photosensitizer, and (3) regeneration of catalyst by semiconductor before catalysis occurs (Figure 2c). The difference in time scales of the various processes is often the reason for an inefficient interplay of the different components. Energy-and electron transfer should be guided by directionality, thus preventing unproductive side reactions such as back electron transfer (BET), leading to charge recombination. Ideally, forward electron transfer processes are faster than BET, and once the electron reaches the catalyst, it should be converted rapidly. Thus, ordering the components in a rational way ensures that the processes occur in a desired stepwise fashion. Supramolecular preassembly of different chromophores for instance leads to fast and directional electron transfer and thus long-distance charge separation in the light-harvesting part. 6

Principles of Synthetic Photocatalysis
Nature uses light mainly for the generation of chemical fuels, i.e., NADPH and ATP, which are used to drive dark reactions creating complex organic matter. However, in organic synthesis, light can be used directly to drive multiple (catalytic) reactions relevant for chemical research and industry. The application of visible light is highly promising because it is abundant, clean, cheap, and safe. Furthermore, by using light, one-electron redox processes typically occur, introducing open-shell organic molecules as intermediates, which may lead to new reaction pathways. Crucially, a specific low-energy wavelength (i.e., visible light) should be used to activate only the PS without damaging other (organic) molecules in the reaction mixture. In the past two decades, research in light-driven catalysis for organic synthesis has been very fruitful. 41,53−55 In general, two different productive mechanisms can occur after excitation of the PS (Figure 3a): (i) energy transfer (EnT) Figure 2. (a) Schematic representation of photosystem II, which catalyzes the oxidation of water to oxygen while charge separating protons and electrons. Flow of electrons is shown by the black arrows. (b) Schematic picture of an artificial photosynthetic device with an anode for light-driven water oxidation (PS = chromophore, WOC = catalyst) and a cathode for light-driven proton reduction catalysis (PS = chromophore, PRC = proton reduction catalyst). (c) Schematic representation of the relevant energy levels for light-driven water oxidation (left) and proton reduction (right) pathways unproductive charge recombination indicated in red (CB = conduction band, VB = valence band). from the excited state (PS*) to an energy acceptor ([A]), and (ii) photoinduced electron transfer (PET) to form either an oxidized (PS + ) or reduced (PS − ) species that is subsequently neutralized via single electron transfer (SET) to the original state. 55 Visible light can therefore be used to activate (organic) molecules to overcome energy barriers and to drive redox reactions by enabling one-electron chemistry. 3,56−58 Several catalysts for EnT and photoredox reactions have been developed, ranging from metal complexes to organic dyes. Photoredox catalysis typically proceeds via PET to or from a sacrificial agent such as TEA or persulfate after initial excitation of the photosensitizer. EnT catalysis involves the transfer of the excited-state energy from the PS to an energy acceptor ( Figure  3b). Generally, two types of nonradiative EnT mechanisms can occur: Dexter energy transfer (DET) or Forster resonance energy transfer (FRET). Crucially, both significantly rely on the distance between the energy donor and acceptor, which can be controlled by supramolecular preorganization. 59 Each excited state has a limited lifetime (τ) before emissive and vibrational relaxation pathways result in a return to its ground state. In order to undergo energy or electron transfer from the excited state, the reaction partner (i.e., catalyst or substrate) has to be in proximity to PS*. In solution, these processes are limited by diffusion and are highly dependent on the concentrations of the solutes. Therefore, the excited-state lifetime of the photosensitizer is mainly determining the viability of the overall process. Generally the excited-state lifetime should be τ > ns to ensure energy or electron transfer to a catalyst or substrate. However, if photosensitizer, catalyst, and/or substrate are preorganized, the system is no longer limited by diffusion, and shorter lifetimes could become feasible. In addition, BET processes are often faster than a chemical reaction and thus are challenging the turnover. In natural (photo)systems, electrons are transferred through multiple units far away from the PS by making use of charge transfer linkers, which is an elegant strategy to combat BET processes. 7 In simplified artificial systems like most synthetic supramolecular photocatalysts, it is thus key to consider that the desired forward (electron transfer) process is faster than the backward one. Interestingly, in the area of photoredox catalysis, there is little attention for solar to product efficiency as the focus is typically on getting new products in high yields.

Supramolecular Coordination Cages
Supramolecular architectures based on coordination complexes have been extensively explored and have been demonstrated to display excellent properties for biomimetic catalysis. 60−63 A coordination complex is build up from a metal ion and ligands ( Figure 1). Depending on the type of metal, its oxidation state, and the structure of the ligands, specific structures can be made via self-assembly. 64 Ligands containing two or more donor groups are able to coordinate multiple metals, leading to larger self-assembled objects. Depending on the angle between these donor groups, determined by the ligand backbones, and the coordination geometry of the metal, either discrete coordination cages or coordination polymers like metal−organic frameworks (MOFs) can be made. 11,65 Besides using multidentate bent ligands, 11,66 metal ions and clusters are used as nodes, and these can be capped 67 or protected by bis-dentate ligands 68 in order to selectively obtain discrete coordination cages. In addition to highly symmetric, homoleptic cages with one type of ligand, in recent years, a synthetic toolbox for the self-assembly of heteroleptic cages with increasing complexity has been developed. 69 In contrast to MOFs, coordination cages are soluble and feature a single, well-defined cavity, providing a chemical environment that can be tailored similarly to enzyme pockets. The chemical environment in these cavities can be tuned by the nature of ligands, (endohedral) functional groups attached to the ligands, as well as the charge that is induced by the metal connecting nodes. Additionally, computational work toward the prediction of self-assembly, binding, and catalysis within coordination cages has recently gained attention. 70,71 Small molecules (such as substrates, intermediates, and catalysts) can be selectively bound within the cavity. 72,73 Supramolecular coordination cages have been extensively researched in various kinds of chemical transformations in which their porosity is exploited to preorganize substrates in the accessible cavity. 25−27 Because the cavity is shielded from the bulk solution, two different environments are created that are separated from each other ( Figure 1). This allows for rational design of the cage structure to place desired functionalities outside or inside the cage, which can be used to induce charge separation. In addition, bifunctional cages can incorporate both a PS and the desired excited-state quencher, e.g., substrate or catalyst, to remove diffusion constraints of the PET. In such a system, the use of photosensitizers with shorter lifetimes can be realized. Generally, the PS can be installed in four different places ( Figure 4): in the bulk solution, encapsulated inside of the cavity, in the linkers forming the cage, and in its metal nodes. In summary, coordination cages show a high potential for increasing efficiency and scope of photochemical processes, mainly due to the unique constraint environment they provide.

COORDINATION CAGES FOR ARTIFICIAL PHOTOSYNTHESIS
In the following section, the focus lies on coordination cages that have been used for artificial photosynthesis, e.g., the generation of solar fuels. We will discuss various approaches in which lightharvesting, charge separation, and catalysis have been combined in supramolecular coordination cages. So far, such systems have been mostlty used for either the oxidation or the reduction reaction and examples of the combination are rare. In the following, we first describe the relevant reactions before discussing how coordination cages have been used in this context.

Water Oxidation.
Water oxidation is a four-electron process generating molecular oxygen, protons, and electrons (eq 1). Because this reaction proceeds typically with slow kinetics, it represents one of the bottlenecks in artificial photosynthesis. 74 In the past decades, multiple molecular and heterogeneous catalysts have been developed that partially overcome some of the limitations by lowering the overpotential and displaying fast kinetics. 75 One of the best molecular water oxidation catalysts to date, [Ru(bpa)L 2 ] (1, bpa = 2,2′-bipyridine-6,6′-dicarboxylic acid), has been developed by the group of Sun. 76 The seven-coordinate configuration of this ruthenium-based catalyst allows the formation of higher oxidation states (up to Ru V ) of the metal center via proton-coupled electron transfer (PCET) processes.
In combination with Ce(IV) as chemical oxidant, this catalyst features similar activities to PSII. 76−78 As described in eq 1, two water molecules are required to form molecular oxygen, and the key step in the overall reaction is the O−O bond formation. For this, two mechanistic pathways have been identified, depending on the nature of catalyst and the reaction conditions: (1) water nucleophilic attack (WNA) and (2) oxyl-radical mechanism (I2M), also referred to as an intermolecular face-to-face mechanism ( Figure 5). 79 Both mechanisms initially follow a similar pathway: Ru(II) (I) is oxidized to Ru(III), followed by coordination of a water molecule to form intermediate II. This is followed by two consecutive PCET steps via intermediate III, abstracting in total two electrons and two protons to generate either a Ru(V) oxo (VII) intermediate which will follow the WNA mechanism, or a Ru(IV) oxygen radical (IV) intermediate that will follow the I2M mechanism. In the I2M mechanism, two Ru(IV) oxygen radicals combine (V) to produce intermediate Ru(III)-O-O Finally, PCET leads to the peroxide radical IX, which releases oxygen and regenerates starting complex I. Sun's water oxidation catalyst 1 follows the bimolecular I2M mechanism, while the slightly different catalyst [Ru(pda)(pic) 2 ] (2, pda = phenanthroline-2,9dicarboxylicacid, pic = 4-picoline) follows the WNA mechanism. 77,80 While molecular catalysts following the I2M mechanism typically feature lower overpotentials, the WNA mechanism has so far resulted in the highest catalytic rates. 81,82 Based on its high activity and stability as molecular catalyst, Sun's water oxidation catalyst 1 and its derivatives have become some of the most used water oxidation catalysts. 83−85 3.1.2. Proton Reduction. The second key reaction in artificial photosynthesis is the generation of molecular hydrogen from protons. Per definition, the reduction potential of the proton reduction reaction is 0 V vs NHE at pH = 0 (eq 2). As these thermodynamic potentials scale with pH, water reduction at neutral or basic pH occurs at a lower potential (eq 3), while the water oxidation requires less potential under these conditions. 86,87 While hydrogen can be generated electrochemically, for example on Pt electrodes, numerous homogeneous systems for light-driven proton reduction have been reported (e.g., natural hydrogenase mimics). 88

CO 2 Reduction.
Carbon dioxide is the most common greenhouse gas in the atmosphere that plays a significant role in global warming by the greenhouse effect. 95 In order to close the carbon cycle, novel and sustainable approaches are required to convert CO 2 into other chemicals such as CO and HCOOH, which can be used to generate fuel or as building blocks for the chemical industry. Below the two reactions and reduction potentials are given for CO 2 reduction into CO and HCOOH, eqs 4 and 5, respectively. 96 Lately, substantial progress in light-driven CO 2 reduction to CO and formic acid has been made, 97−99 however homogeneous (photo)catalysts have shown low stabilities because they are readily transformed to nonactive forms during prolonged lightdriven reactions. 100 As a potential solution, these homogeneous (photo)catalysts can be incorporated within cages, resulting in more stable catalysts while featuring high catalytic activity and selectivity. Another challenge is that the reduction potentials required are close to that for proton reduction, thus the reduction of CO 2 competes with the proton reduction reaction. 101

H 2 S Oxidation.
H 2 S is abundant in natural gas and crude oil, which despite the toxicity of the gas has resulted in increased interest as feedstock. 102 In analogy to water splitting, H 2 can be generated from photocatalytic H 2 S splitting, forming elemental sulfur as a side product. 103,104 This overall reaction, sulfide oxidation with proton reduction, is displayed in eqs 6 and 2, respectively. As has been discussed in section 2 of this review, functionalities can be generally introduced into coordination cages in four different ways ( Figure 4): The PS can be (1) external in solution, (2) encapsulated in the cavity, (3) installed as linker, or (4) introduced as metal node. In the following discussion, we divide recent literature into subsections according to the position of the PS and by the investigated redox reaction.

Cage Catalysts with External Photosensitizer
In this part, we discuss literature where the catalyst is part of the supramolecular structure and an external photosensitizer is used.

Water Oxidation.
Wurthner and co-workers incorporated [Ru(bpa)(L) 2 ] complex 1 as metal node in a series of supramolecular rings ( Figure 6). In a first report, ditopic pyridine ligands were used to prepare supramolecular triangle C1a, containing three Ru(bpa) catalytic centers. 106 The catalytic activity of the new supramolecular triangles was initially compared to the parent complex 1 by using cerium ( Embedding the catalyst in the supramolecular triangle showed two effects. Firstly, the stability of the molecular catalyst was increased, leading to higher TON. This is due to the tethering effect of the connecting ligand in the triangle, avoiding decomposition of the catalyst. In the mononuclear control complex 1, dissociation of one pyridine ligand leads to decomposition of the catalyst, and such dissociation is prohibited in the triangle structure. Secondly, the oxidation potentials increase for the triangles by ±100 mV per formal oxidation event from Ru II to Ru V compared to the monomer, possibly due to the close proximity of the catalytic centers.
Embedding the catalyst in the supramolecular triangle leads to a change in mechanism, as revealed by kinetic studies and 18 O labeling experiments. While the molecular reference complex 1 typically follows the I2M mechanism, triangle C1a performs water oxidation via the WNA mechanism. Experiments revealed that the process follows first-order reaction kinetics in the oxygen evolution reaction for both catalyst and CAN concentration. 109,110 Furthermore, multiple water molecules are preorganized within the supramolecular structure. DFT calculations revealed that up to 10 water molecules can fit in the cavity of the triangle, while only three are required to form a hydrogen bonding network between two Ru−OH centers. Due to preorganization of the water molecules, the activation energy for O−O bond formation is lowered and as a result the overall activity is significantly increased when using the triangle as catalyst.
In a follow-up study by Wurthner and co-workers, two derivatives of C1a were synthesized, bearing water-solubilizing groups in the backbone of the supramolecular triangle in order to increase the water solubility of the WOC. 107 C1b was synthesized containing tertiary amines in the backbone structure of the ligand, being able to perform chemical water oxidation with CAN in a solvent mixture of 7:3 MeCN:H 2 O, obtaining a TOF up to 147 s −1 and TON up to 5.2 × 10 3 at pH = 1. C1c, bearing oligo(ethylene glycol) functionalities on the backbone, was also active in fully aqueous solution, however, the photocatalytic activity decreased by ca. 20%. On the other hand, lowering the MeCN content for C1b led to precipitation of the catalyst. The lower activity observed for C1c is likely due to Coulombic repulsion between the charge on the catalyst and the charged Ce IV ions. Overall, the more water-soluble supramolecular triangles displayed relatively high activities yet did not outperform the parent triangle C1a in terms of activity. This was attributed to the new supramolecular triangle being prone to oxidative decomposition.
The same group also investigated the effect on water oxidation by substituting the bpy backbone of the photosensitizer with electron-withdrawing groups (3b and 3c) to increase the PS + / PS oxidation potentials and, therefore, the thermodynamic driving force for activation of the WOC. 108−111 The photosensitizers were compared using catalyst C1a and the more water-soluble catalyst 3c in the presence of sodium persulfate as a sacrificial electron acceptor. Unexpectedly, the highest catalytic activity for C1a was observed in combination with 3a, obtaining a TOF up to 10.9 s −1 and TON up to 430 in a 1:1 MeCN:H 2 O (v/v) solvent mixture, while using 3b and 3c, the TOFs decreased to 2.8 and 0.5 s −1 , respectively. A similar trend was obtained for C1b in combination with the photosensitizers, obtaining TOFs of 9.5, 2.2, and 0.4 s −1 for 3a, 3b, and 3c, respectively. Photocatalytic water oxidation using catalyst C1b was also explored in a 5:95 MeCN:H 2 O solvent mixture, as this WOC is more soluble in water than its C1a derivative. Interestingly, the highest catalytic activity was observed when 3b was used as sensitizer, obtaining a TOF of 10.8 s −1 and a TON of 320, compared to TOFs of 2.9 and 0.7 s −1 for 3a and 3c, respectively. It is notable that a 7.5 times lower PS concentration was used in the 5% MeCN mixture due to limited solubility of the photosensitizer in the respective solvent mixture. In comparison, in the 1:1 MeCN:H 2 O mixture, 3b in combination with C1b decreased the TOF and TON to 1.1 s −1 and 45 at 0.2 mM 3b concentration, respectively, indicating high efficiency in the 5:95 MeCN:H 2 O system for the relevant photosensitizer concentration. Nanosecond flash photolysis was used to study the efficiency of electron transfer from C1b to the photosensitizers. The electron transfer rate k ET was observed to be 1 order of magnitude larger for 3b compared to 3a. In addition, Stern−Volmer quenching studies revealed an emission quenching k q for 3a of one magnitude higher than 3b in both 5:95 and 1:1 MeCN:H 2 O mixtures. This indicates that electron transfer from the WOC to the photosensitizer is not the ratedetermining step in the 1:1 MeCN:H 2 O mixture. The low photocatalytic activity of both catalysts in combination with 3c is explained by electrostatic repulsion of anionic carboxylates on the PS and the negative charge of the persulfate anions, which could improve upon performing the reaction in stronger acidic conditions.
The synthesis of a cyclic dinuclear Ru(bda) complex equipped with oligo(ethylene glycol)-functionalized (OEG) axial calix [4]arene ligands for homogeneous catalytic water oxidation under highly diluted conditions (C2, see Figure 7) was reported by the same group. 112  Wurthner and co-workers improved the system even further by synthesizing di-, and tetranuclear Ru(bda) OEG-functionalized supramolecular rings (C3a and C3b, see Figure 7) based on previously mentioned triangle (C1c). 113 OEG side groups were incorporated in the structure to increase the watersolubility of the catalysts, enabling all of the catalysts to perform photocatalytic water oxidation in a 1:1 MeCN:H 2 O mixture. Photocatalytic experiments revealed that upon increasing the size and number of catalytic centers of the supramolecular ring, the activity of photocatalytic water oxidation increases. Using C3b as catalyst resulted in a TOF of 23 s −1 and TON of 500, while the smaller rings obtained TONs of 400 and 36 for C1b and C3a, respectively. In addition, not only did the total TOF increase but the TOF per Ru II -center also increased, ranging from 5.8 s −1 /Ru for the tetramer, compared to 3.3 s −1 /Ru for the trimer. Because for all catalysts similar oxidation potentials were measured, the increase in photocatalytic performance is suggested to be a result of higher rates by water pre-organization increased stability by the macrocyclic effect (in which case not all the ruthenium in solution is active).
The first supramolecular cage for photocatalytic water oxidation was synthesized by Li and co-workers, [Co 20 (4) 12 (OH) 12  For C4b and C4c, only 50% and 25% of the activity was obtained, respectively, while for C4d, no oxygen evolution was observed at all. The enhanced activity of C4c over C4d could be attributed to more feasible substitution of the NO 3 − counterion acting as ligand with H 2 O, compared to the Br − counterion. The higher activity of C4b over C4c and C4d is attributed to the Co−O active metal centers. Mechanistic studies revealed that water oxidation in the bis-cobalt site is  facilitated by PCET in the rate-determining step, explaining the higher oxygen evolution rates using C4a as a catalyst compared the C4b, where the bridged bis-cobalt is not present.

Proton Reduction.
Hong and co-workers were the first to report cages bearing zirconocene clusters as metal nodes, which gained popularity because of their high stability and the relatively easy synthesis. 118−121 Yuan and co-workers synthesized [(Zr 3 O(OH) 3 Cp 3 ) 4 (5) 6 )]Cl 4 (C5), with proton reduction catalyst 5 incorporated as ligand ( Figure 9). 122 A hydrogen evolution rate of 10.5 mmol g −1 h −1 over the course of 27 h was obtained using TEOA as the electron donor and rhodamine B (6) as photosensitizer. The activity of C5 was compared to the free ligand 5 as reference. Interestingly, for the first 5 h, similar activity was observed as for the cage. However, the molecular reference ligand 5 then lost its activity, while C5 continued to produce H 2 for another 7 cycles of 3 h with minor loss of activity. Overall, the cage analogue is 1.8 times more productive than the reference. The higher stability, and thus overall better productivity of C5, is a result of site isolation of the complex. This causes isolation of the metal complex from other metal complexes and prevents degradation to nanoparticles. 123 In 2017, Boomishankar and co-workers used a fully watersoluble octahedral [Co 6 7 8 ] C6a (7 = tripodal silane ligands MeSi(py) 3 ) cage for photochemical proton reduction, in conjunction with Ru(bpy) 3 2+ 3a as PS and ascorbic acid (AA) as a sacrificial reductant (Figure 10a). 124 As shown in the X-ray structure of octahedral cage C6a, each Co(II) node is coordinated by one chloride ion pointing inside the cavity ( Figure 10b). In addition, each Co(II) center coordinates four pyridine donors and one water molecule pointing outside of the cavity. This structure suggests that the cage cannot bind additional guest molecules in its internal cavity, as it is fully occupied by chloride ligands. Therefore, proton reduction likely occurs outside of the cavity. In photochemical hydrogen evolution, a TON of 43 was obtained after 2 h irradiation of the reaction mixture and a TOF of 21.5 h −1 in an aqueous phosphate buffer solution (pH = 7). Control experiments showed almost no hydrogen production in the absence of the PRC. The cease of photocatalytic H 2 production after 2 h was investigated by adding additional SED, PS, and catalyst one by one, none resuming the H 2 production. Photocatalytic H 2 production was only resumed after adding additional PS and catalyst simultaneously, suggesting rapid degradation of photosensitizer 3a and catalyst C6a. A Pourbaix diagram (plotting the potential of redox events vs pH), obtained from differential pulse voltammetry (DPV) in Briton−Robinson buffer, displays a linear dependence of the reduction Co(II)/Co(I) with a slope of −59 mV/pH in the range of pH 2.8−5.6, indicating the mechanism involves a PCET in this pH regime. This is beneficial, as PCET lowers the thermodynamic barrier by avoiding charge accumulation. 125 Changing the metal node to nickel led to increased activity of the resulting cage [Ni 6 7 8 ] (C6b). 126 As in the previous system, the authors state that the polypyridyl silane ligands 7 stabilize low oxidation states on the nickel ions and act as redox-active functions to assist the PCET process. In contrast to the Co(II) analogue, C6b is significantly more stable and showed to be active for 69 h under irradiation in photocatalytic conditions, with a maximum TOF of 41 h −1 and TON of 2824.
Nitschke and co-workers have previously reported the binding of neutral guest molecules with large π-conjugated systems into cubic porphyrin cages ([Fe 8 8 6 ], C7) formed by subcomponent self-assembly. 127 Following the same principle, Sakai and co-workers introduced molecular Pt(II) PRCs such as Pt II (dmb)Cl 2 (9) bearing π-conjugated ligands as guest into the cavity of C7 ( Figure 11). 128 In addition to 3a as PS and ethylenediaminetetra acetic acid (EDTA) as sacrificial reductant, methyl viologen (10) was used as electron relay in the photochemical system with [9⊂C7] as catalyst in aqueous acetate buffer at pH 5. The role of the redox mediator is to act as shuttle between PS and catalyst. 10 oxidatively quenches the excited PS* quickly and thus reduces unproductive energy transfer from PS*. Reduced 10 then diffuses to the catalyst and transfers one electron, thus leading to unidirectional electron flow. It has been shown previously that electron relays can facilitate both charge separation and efficient electron transfer between PS and catalyst. 129,130 A TOF = 23 h −1 and TON = 58 after 200 min of irradiation is obtained, which is significantly higher than the reference system (complex without cage; TOF = 1 h −1 and TON = 6).

CO 2 Reduction.
Wisser and co-workers prepared Rh(II)-paddlewheel cages C8a and C8b for light-driven CO 2 reduction with dodecoxybenzene-1,3-dicarboxylic acid (11a) and isophthalic acid (11b) as ligands, respectively (see Figure  12a). 131 Uniquely, C8a and C8b can be functionalized with axial ligands on both the endohedral rhodium and the exohedral rhodium center. Thus, ditopic ligands can be used to cross-link the spherical cages to form supramolecular polymers ( Figure  12b). While stepwise addition of ditopic imidazole ligand bix (12) led to spherical coordination polymer particles (C 12 Rh-CPP, C9a, and HRh-CPP, C9b), direct addition 12 equiv of 12 to C8a induced formation of a supramolecular aerogel (C 12 Rh-SAG, C10). In both cases, the axial ligand 12 coordinates to the exohedral site of the Rh(II) dimer nodes, leaving the endohedral Rh(II) open as catalytic center. Photochemical CO 2 reduction with the different species was performed, using Ru(bpy) 3 Cl 2 3a as PS and TEOA as sacrificial reductant.
Despite the difference in size and porosity, C 12 -containing polymers C9a and C10 featured similar photocatalytic activity, obtaining a TOF of 59 h −1 for the conversion of CO 2 exclusively into HCOOH. In the case of the single sphere C8a, a slightly lower activity was observed (TOF = 52 h −1 ). The effect of the axial ligand was tested by addition of monodentate 1benzylimidazole, which coordinated to the exohedral Rh(II) center. A TOF of 58 h −1 was observed, clearly demonstrating the positive influence of axial coordination. As shown previously, the axial ligand on the exohedral rhodium center leads to increased Figure 11. Cubic porphyrin 8 based cage [Fe 8 8 6 ] C7 that may encapsulate Pt II (dmb)Cl 2 catalyst 9 for the production of hydrogen with Ru(bpy) 3 Cl 2 3a as photosensitizer and methyl viologen 10 as redox mediator. electron density on the endohedral rhodium center, which results in improved CO 2 reduction. 132 The same effect is anticipated to lead to better performance of polymers C9a and C10. Single sphere C8b displayed slightly lower activity (TOF = 43 h −1 ), likely due to the lack of alkyl chains which contribute to increasing the electron density. Recycling experiments using C9a revealed that the polymeric catalyst features similar initial activity for the first 2 h within 4 cycles. The activity rapidly decreased after 2 h, which was explained by photosensitizer degradation. Apparently, the cages remain stable even after 4 runs of 8 h. Additionally, TEM experiments showed that no Rh nanoparticles were formed during catalysis, confirming the stability of the supramolecular polymeric network and the cages themselves.

Encapsulated Photosensitizer
Supramolecular cages can host guest molecule by employing attractive forces such as electrostatic interactions, or using hydrophobic effects. 24 These strategies have also been used to encapsulate molecular photosensitizers. In the following, we will discuss recent work on encapsulated photosensitizers. In these examples, the metal nodes of the cages act as catalytic centers.
Duan and co-workers designed tetrahedron [Co 4 13 4 ] (C11), which can be prepared via self-assembly of triphenylamine ligands substituted with thiosemicarbazone handles that can coordinate to Co II cornerstones as the catalytic centers (see Figure 13a). 133 Binding studies revealed that one fluorescein molecule 15 can bind in the cavity of C11. Optimal conditions for photocatalytic H 2 evolution were found to be a H 2 O:EtOH (1:1) mixture, using triethylamine (Et 3 N) as sacrificial electron donor. The initial TOF for [15⊂C11] was calculated at 750 h −1 , and the value for the TON was 11000 over a period of 24 h, which was the highest activity for a cobalt−fluorescein system at the time. For comparison, a smaller cage [Co 4 14 4 ] (C12) was synthesized with comparable redox potentials and coordination structure yet with a smaller cavity compared to C11 and unable to bind 15. Under similar photocatalytic proton reduction conditions, a decrease in initial TOF (450 h −1 ) and TON (4500) were obtained. Fluorescence quenching studies revealed that both cages follow a different mechanistic pathway. As a result of preorganization, the mechanistic pathway of C11 is dominated by PET from 15* to C11 (oxidative quenching), inducing immediate H 2 evolution. The mechanism in C12 is initiated by reductive quenching of 15* to 15 − , which then has to diffuse to the cobalt center for H 2 evolution.
He and co-workers synthesized electron-deficient redox cage [Ni 4 16 6 ] (C13, see Figure 13b), containing Ni II as the catalytic center, with similar binding properties of 15 as in C11. 134 Proton reduction was facilitated in a H 2 O:EtOH (1:1) mixture with 12% Et 3 N as the sacrificial electron donor. The calculated TON obtained was up to 1200 mol of hydrogen per mole of catalyst under optimal conditions. Using smaller bowl-shaped [Ni17 3 ] (C14) that cannot bind 15 in its cavity resulted in lower H 2 evolution up to 0.5 mL over a period of 24 h (vs >1.2 mL for C13).
In a competition experiment, adenosine triphosphate (ATP) was used to block the cavity as it binds more strongly than 15. Upon addition of ATP to the host−guest complex [15⊂C13], all bound 15 is substituted by ATP. Expulsion of 15 from the cage led to a 65% decrease in H 2 evolution activity compared to the original system. Addition of ATP to C14 only lowered the  activity slightly, not having as much effect on the electron transfer as compared to the inhibition in C13. Similar differences in mechanistic pathways are assumed as for C11 and C12.
Duan and co-workers continued to synthesize box-like cages [Fe 8 18 6 ] (C15) and [Fe 8 19 6 ] (C16), bearing Fe II (bpy) 3 and Fe II (2-pyCH 2 �N) 3 cornerstones and tetraphenylethylene ligands 18 and 19 as backbones, respectively ( Figure 14). 135 The cavity of C16 is large enough to bind two fluorescein molecules 15, leading to stronger quenching of the fluorescence compared to C15, which can bind only one molecule of 15. The binding constant for two 15 in the cavity of C16 is larger than for only one (positive cooperativity). As a result of this binding, C16 produces more H 2 under optimal conditions (MeCN: H 2 O, pH = 11). Similar to C11, fluorescence quenching and lifetime studies indicate a PET from 15* to both C15 and C16.
Dihydropyridine amido moieties (DHPA) are important in nature, as they play a role in electron transfer, and it is the key molecular structure in nicotinamide adenine dinucleotide (NADH) ( Figure 15). 136 Addition of ATP to the catalytic system stopped the H 2 production completely, inhibiting the photocatalytic reaction. This confirms that photoinduced H 2 production occurs in the cavity rather than via a diffusion dependent homogeneous system.
He, Guo, and co-workers synthesized [Ni 6 21 6 ] 12+ (C18), in which the ligand 21 contains three binding sites ( Figure 16). 139 Because C18 is cationic, photosensitizer Ru(dcbpy) 3 4− 3c present in its anionic form in alkaline media binds through electrostatic interactions in the pocket of C18. Encapsulation of the photosensitizer was confirmed by MS and 1 H NMR. Under optimal conditions, an initial TOF of 1100 mol hydrogen per mole of catalyst per hour and TON of 1600 mol hydrogen per mole catalyst was achieved, using TEOA (15%) as a sacrificial reductant in a H 2 O:EtOH (1:1) mixture (0.26 mL H 2 in 5 h). As expected from previous systems, an oxidative quenching pathway was established by transient absorption spectroscopy (TA). A peak was observed at λ = 420 nm, in line with the formation of Ru III complexes expected to form by oxidative     Figure 17). 140 Figure 17) was synthesized, bearing two hydroxymethyl groups per ligand. Interestingly, the photocatalytic H 2 evolution activity of C19b was very similar under the same conditions for C19a and did not decrease significantly in a 1:4 EtOH:H 2 O mixture, providing strategies for the development of highly efficient water-soluble homogeneous proton reduction catalysts.
Zhao, Duan and co-workers prepared a negatively charged supramolecular square [Co 2 25 2 ] 4− (C21) and studied supramolecular complex formation with Ru(bpy) 3 2+ 3a. 141 The crystal structure of the host−guest complex indicated binding of two 3a molecules to square C21 (Figure 18a), which is stabilized by electrostatic interactions and hydrogen bonding while substituting the counterion. The host−guest binding of C21 and two 3a complexes was also confirmed by isothermal titration calorimetry (ITC). Based on fluorescence quenching experiments, the photocatalytic pathway was found to be similar as previous host−guest systems. The excited state of the photosensitizer is directly oxidatively quenched by the Co II cornerstones. In this system, optimal conditions were obtained with AA as sacrificial electron donor, with up to 400 μL of H 2 in 9 h with a corresponding initial TOF of 40 h −1 . H 2 evolution shows a linear relation with the concentration of C21, which is similar to previous systems. 135 A mononuclear variant [Co26 2 ] 2− (C22) was synthesized (Figure 18b), which under the same conditions only yielded trace amounts of H 2 . In addition, a triangular prismatic cage [Co 3 27 2 ] 6− (C23) was synthesized (Figure 18c). ITC assays confirmed the host−guest formation of [3a 3 ⊂C23], replacing all Et 4 N + counterions. H 2 evolution with C23 was investigated, which under similar conditions as for C21 yielded 350 μL of H 2 , which is slightly lower than C21.

H 2 S Splitting.
Reek and Duan prepared an octahedron Ni II -cage [Ni 6 28 4 ] (C24) (Figure 19) for photochemical splitting of H 2 S into H 2 and elemental sulfur. 142 C24 bears tritopic triphenylamine ligands 28 with hydrazinecarbothioamide coordinating moieties, which are self-assembled in situ during cage formation. The photosensitizer fluorescein 15 can bind in the cavity, once more preassembling both PS and the catalytic center. 143 Photochemical H 2 evolution was performed using TEA as a sacrificial electron donor in a H 2 O:EtOH (1:1) mixture at pH = 12.6, obtaining an initial TOF of 1250 mol of hydrogen per mole of catalyst per hour and a TON of 25000 per mole of C24. As the 15* follows an oxidative quenching pathway because of the close proximity to the catalyst, the oxidized dye 15 + can oxidize S 2− to elemental sulfur, and indeed formation of a yellow powder is observed during the reaction. Mononuclear complex [Ni29 2 ] (C25) displays significantly lower yields in photochemical H 2 evolution, and no elemental sulfur was formed, indicating that the S 2− was not oxidized under these conditions. In another experiment where glucosamine was added as competitive binder for the cavity in C24, also no sulfur was produced, demonstrating the importance of preorganization of PS and the catalyst.
Recently, Jing, Duan, and co-workers synthesized helical capsule [Fe 2 30 3 ] (C26) for simultaneous H 2 S splitting and hydrogenation of nitrobenzene ( Figure 20). 144 Both capsules contain Fe(bpy) 3 2+ units as metal nodes, representing the active catalyst. In the cavities, one molecule of fluorescein (15) can be encapsulated as PS. The reaction follows an oxidative quenching pathway. Photochemical H 2 evolution was observed in a MeCN:H 2 O (1:1) solution at pH = 11.5. A TON of 276.7 and a TOF up to 19.7 h −1 were observed. A yellow powder was also obtained from the reaction, which could be characterized as elemental sulfur. Nitrobenzene is able to bind in the cavity of C26, resulting in hydrogenation to aniline in 99% yield under photocatalytic conditions. Larger substrates were also explored, e.g., 1-nitronaphthalene, 9-nitroanthracene, and 1-nitropyrene, requiring longer reaction time to reach 99% yield as the size of the substrate increases (up to 420 min for 1-nitropyrene). The longer reaction time is also in line with kinetic studies that observe an initial rate of 1.87 mM min −1 for nitrobenzene, compared to 0.28 mM min −1 for 1-nitropyrene. A larger substrate, 2,4,6-triphenylnitrobenzene, that is larger than the windows of the cavity of C26, could not be hydrogenated to the aniline product and only a small amount of hydrogen gas was observed.
The authors also further investigated the influence of the hydrazide groups as hydrogen bond donors for preorganization of the nitrobenzene substrates. Another capsule bearing a pdivinylbenzene unit [Fe 2 31 3 ] (C27) was synthesized containing no hydrogen bonding motifs, yet a slightly larger opening (4.9 Å vs 3.8 Å for C27) and similar electrochemical properties. Interestingly, higher photocatalytic H 2 production was obtained    (Figure 21). 145 Binding of PS 15 in the cavity of C28 enables direct PET from 15* to C28, similarly to the systems discussed previously in this section. Optimal conditions were found using TEA as a sacrificial reductant in 1:1 MeCN:H 2 O solvent mixture at pH = 11. The initial TOF for H 2 formation showed to be 160 h −1 , with a total TON at 1250 per mole of catalyst over a period of 12 h. Under similar conditions in CO 2 saturated solvent, HCOO − production was observed in amounts increasing to 0.46 μmol (initial TOF = 0.8 mol of HCOO − per mole of the catalyst per hour, TON = 9.3 per mole of catalyst over 12 h). Only trace amounts of H 2 were formed during the reaction and no other byproducts, indicating the dominating pathway of CO 2 reduction over H 2 evolution.
Interestingly, the mononuclear complex [Ni33 2 ] (C29) did undergo photocatalytic H 2 evolution (0.93 mL over a period of 12 h) under similar conditions as C28, yet no HCOO − was observed in the presence of CO 2 . The photocatalytic mechanism is anticipated to be the same as for C28, as the 15 molecules can bind via π−π stacking to the ligands of C29 (as indicated by 1 H NMR and NOESY experiments). With a difference in dihedral angle (23.8°for C28 and 17.0°for C29), the authors concluded that the strained coordination in C28 results in a more distorted coordination at the Ni centers which enables the activation of CO 2 .

Photoactive Cages with Encapsulated or Incorporated Catalysts
In this section, we discuss cages containing photoactive ligands with encapsulated catalysts.

Encapsulated Catalysts.
Reek and co-workers utilized Nitschke-type cage [Fe 4 34 6 ] (C30) with Zn(II) porphyrin ligands as host for a [FeFe]-hydrogenase mimic 35 containing pyridyl−phosphole ligands (Figure 22). 146 The pyridyl units coordinate to the Zn(II) of the porphyrins and thereby enable catalyst binding inside of the cage. According to fluorescence quenching titration experiments, one catalyst molecule binds strongly to the cage. In the presence of TFA as a proton source and 4-mercaptobenzoic acid as a sacrificial   The system is limited to low acid concentrations in order to prevent pyridine protonation, which leads to catalyst dissociation. Importantly, encapsulation lowered the overpotential required for catalysis as indicated by electrochemical measurements. In addition, time-resolved spectroscopy showed fast PET from the host to the encapsulated catalyst at 0.5 ps, whereas charge recombination occurs in around 37 ps. Single-electron reduction of 35 was confirmed by time-resolved IR spectroscopy, with an overall quantum yield of 1%.
A second example using a [FeFe]-hydrogenase mimic 36 as catalyst was demonstrated by Duan and co-workers. Cage [Ce 4 37 6 ] (C31) was synthesized using a carbazole photosensitizer in the ligand structure ( Figure 23). 147 Host−guest studies revealed that the proton reduction catalyst 36 binds to the cavity of C31 in a 2:1 ratio. As seen previously in other host− guest systems, photoluminescence studies showed direct quenching of the photosensitizer, indicating PET from PS to 36, facilitated by preorganization. Light-driven H 2 evolution was carried out in a MeCN:DMF:H 2 O (8:1:1) mixture using N i Pr 2 EtH•OAc as a sacrificial electron donor, obtaining a TON of 30 in 4 h with a TOF of 11 h −1 in the first hour. The free ligand 37 was used as reference, leading to trace amount of H 2 production. Additionally, occupation of the cavity by ATP led to complete deactivation of photocatalytic H 2 evolution, clearly demonstrating the necessity of preorganization. The authors did not mention if the Ce IV ions could play a role in the photocatalytic H 2 production and did not investigate their redox potential.

Metal Nodes As Catalysts.
Finally, it is possible to incorporate both light-harvesting and catalytic centers as building blocks in one cage. This strategy allows to preorganize both functions and at the same time leave the cavity available for substrate or cofactor binding.

Proton Reduction.
Similarly to C5 (Figure 9), an aniline derivative [(Zr 3 O(OH) 3 Cp 3 ) 4 (38) 6 )]Cl 4 (C32) was synthesized by Su and co-workers, exhibiting photocatalytic proton reduction using H 2 O as proton source (Figure 24). 148 Analogous photoactive metal−organic framework (MOF) UiO-66-NH 2 was used as a reference to compare the activity of C32 to the corresponding MOF. Interestingly, H 2 evolution using C32 (510 μmol g −1 h −1 ) is more than 20 times higher than using that of UiO-66-NH 2 (25 μmol g −1 h −1 ) in a 1:3 MeCN:H 2 O mixture. Photoluminescence studies revealed that the emission intensity in C32 is significantly lower than in UiO-66-NH 2 , proposing more efficient charge transfer from the excited phenylamino group to the Zr-cluster in C32. H 2 evolution could be improved further by supporting Pt NPs on the surface of C32 as cocatalyst (Pt/C32), increasing the H 2 evolution to an optimal 1058 μmol g −1 h −1 at 17 wt % Pt NP. Supporting Pt NPs on UiO-66-NH 2 increased the H 2 evolution to 62 μmol g −1 h −1 , which is still ca. 17 times lower than the cage.
In a follow-up study, Su and co-workers investigated the effect of how Pt NPs are immobilized on the Zr-cage. 149 In the previous study, it was reported that the NPs are immobilized on the surface after formation of the cage, while in this study the focus is on in situ immobilization during the formation of the   Chemical Reviews pubs.acs.org/CR Review cage (Pt⊂C32). H 2 evolution appeared to be much higher for the in situ immobilized cage, having an H 2 evolution rate of 10.8 × 10 3 μmol g −1 h −1 , compared to 1058 μmol g −1 h −1 of Pt/C32. The higher activity is ascribed to the shorter charge transfer distance to the internal Pt NPs in Pt⊂C32 as they are more likely in the cavity, rather than the NPs being immobilized on the face of the cage in Pt/C32. Su and co-workers synthesized the bimetallic molecular cage [Pd 6 39 8 ] 28+ (C33, see Figure 25a), with the ruthenium photosensitizer being used as ligand in the cage. 150 Here the Pd(II) nodes of the cage serve as the active catalyst for proton reduction. Photocatalytic experiments were performed in a 9:1 DMSO:H 2 O mixture using TEOA as the sacrificial electron donor. In the first 3 h cycle, the H 2 evolution rate of 380 μmol h −1 was obtained, declining gradually to ca. 150 μmol h −1 in the 16th 3 h cycle. The TOF decreases from 30 h −1 in the first cycle to 11 h −1 in the 16th cycle. The activity of C33 in photo-driven proton reduction was seemingly higher by a factor of ca. 2.7 compared to previously reported Ru−Pd assemblies, indicating higher stability in the cage C33 during proton reduction. This is confirmed by prolonged irradiation of the C33, where nanoparticles formed after ca. 100 h of irradiation while using other Ru−Pd assemblies, typically NPs were observed within 48 h of irradiation.
The authors used ultrafast TA to follow the rate of intramolecular electron transfer from the photosensitizer to the catalytic center in DMSO. The reaction is initiated by excitation of the Ru complex to the singlet metal-to-ligand charge transfer ( 1 MLCT) state, which within 0.3 ps undergoes intersystem crossing to the 3 MLCT state (Figure 25a). From here, two pathways are possible, either decay to the ground state, which results in phosphorescence, or intraligand charge transfer (ILCT) from the phenanthroline to the benzimidazole moiety, which occurs in 7 ps. Finally, the electron is transferred to the catalytic site by ligand-to-metal charge transfer (LMCT), which was observed to happen in 122 ps. This shows that electron transfer through covalently bound and coordinatively bound groups occurs fast enough to outcompete BET in the supramolecular sphere.
The same group also investigated the influence of adding electron mediator tetrathiafulvalene (40), 151 which was encapsulated in the cavity of C33 as a result of the hydrophobic effects (Figure 25b). 152,153 A significant difference in photocatalytic activity was observed, generating up to 2680 μmol H 2 with a corresponding TON of 1015 in 47 h by addition of 20 equiv of 40, compared to 1597 μmol H 2 (TON = 605) in 47 h when no 40 was added. Lowering the concentration of 40 to 10 equiv displays similar activity for the first 14 h, after which the activity readily declines, indicating that a sufficient amount of 40 not only enhances H 2 formation but also improves the stability of the cage. Increasing the concentration of 40 to 40 equiv, however, decreases the initial activity in photocatalytic H 2 formation due to competing or disturbing electron transfer relay from nonencapsulated 40 guests out of the cavity of the cage. The effect of preorganization was then investigated by using Pd(py) 4 2+ as proton reduction catalyst and Ru(bpy) 3 2+ 3a as PS in addition with 40, which only obtained half of the activity that C33 featured for the first 6 h, after which the activity depleted.     (43)(44) 5 )]Cl 4 (C35) ( Figure  27). 154 Photocatalytic CO 2 reduction with the cage was performed in MeCN using TEA as the sacrificial electron donor, reaching TON values per catalytic site of 20, with 96% selectivity toward CO. In comparison, a TON of 3.74 was obtained for molecular catalyst 43 under the same conditions. 43 shows little activity after 1.5 h, while C35 retains 64% of its activity after 3 cycles of 5 h.

CO 2 Reduction. Choi and co-workers incorporated a Re(I) CO 2 reduction catalyst [Re(dcbpy)(CO) 3 ]Cl
It was shown that in MeCN C35 forms aggregates. Single C35 cages can be obtained by dispersing the material in MeOH. Particle sizes were determined by dynamic light scattering (DLS) measurements to be 5.5 nm, which corresponds well to the value of 5.7 nm found by TEM. The CO generation using single C35 increased the activity by 3.4-fold compared to the bulk C35, reaching a TON and TOF of 59 and 120 h −1 , respectively. In order to gain insight into the role of the NH 2 side groups of 44, DFT calculations were performed. It was found that hydrogen bonding in the cage structure between the O of the ligating carboxylate and H of the NH 2 stabilized the transition state, which does not occur in the mononuclear reference system due to the lack of surrounding ligands. This indicates that not only the stabilization of the catalyst enhances CO 2 reduction but also the ligand scaffold of the cage promotes higher activity.
In the examples discussed in this section, the light-harvesting unit was installed as linker, and the metal node served as catalytic center. An alternative to this is the co-incorporation of both functions in the form of linkers. Current examples are mostly based on symmetric cages, however, heteroleptic coordination cages exhibiting multifunctionality are nowadays also accessible. 69,155,156 Several strategies to prepare these cages have been reported; the most efficient include: (i) the use of shapecomplementary ligands, (ii) donor-site engineering, or (iii) the hierarchical buildup of such cages. This type of more complicated structures allow new designs to combine lightharvesting units with catalytic centers.

PHOTOREDOX CATALYSIS FOR THE SYNTHESIS OF COMPLEX MOLECULES
The use of light in the synthesis of complex molecules has taken considerable interest because it can create new reaction pathways. 157 167 polymerizations, 168 and technological development of flow chemistry. 169 In this context, a plethora of new PSs as photoredox catalysts have been developed. 58,162,170 In addition, new dual-type photoredox methodologies have been developed, combining photoredox events with for example organocatalysis, 171 transition metal catalysis, 54 electrocatalysis, 172 and others. 173,174 These dual-type strategies are based on generating highly reactive radical-type intermediates using light as a reagent, of which the reactivity is controlled by the introduction of the second component. Supramolecular cages have been demonstrated to provide an interesting tool to control selectivity and activity in photoredox catalysis. Preorganization of the photoredox catalyst and the reactive molecule that quenches the excited state can be achieved inside the cage and should prevent the diffusion dependence of the PET. However, the same preorganization could also facilitate BET. In our discussion, we will focus on four strategies that have been applied to perform photoredox catalysis in the cavity of cages: (i) encapsulation of lightabsorbing guests, (ii) formation of a host−guest charge transfer (CT) complex, (iii) incorporation of a PS in the linker, and (iv) incorporation of a PS in the metal node.

Encapsulation of Light-Absorbing Guests
The unique microenvironment in the cavity of supramolecular cages has been extensively used to encapsulate various guests by means of electrostatic, hydrophobic, and van der Waals interactions. 24,175,176 Similarly, light-absorbing molecules can   (Figure 29a). 177 They found that two guest molecules of α-diketone 46 in H 2 O could be encapsulated in C36a ( Figure  29b). 178 After filtration of excess substrate, the newly formed host−guest complex was irradiated with a mercury lamp, forming three different reaction products: cyclized products 47a and 47b and OH-substituted 48. However, in the absence of C36a, mainly products resulting from homolytic cleavage, such as benzaldehyde, were present in the mixture of compounds that formed. Therefore, the authors demonstrated that the cavity of C36a induces a spatial constraint by preorganization of the substrates that suppresses the reaction pathways that are dominant in bulk solution and enables different reactions. In addition, the same authors found that ortho-quinone 49a and padamantyl toluene (50) in water could be selectively coencapsulated in C36a in a 1:1:1 ratio. 179 Remarkably, when only 49a was exposed to C36a, barely any encapsulation was observed, indicating a strong positive cooperativity in binding. Upon photoirradiation (mercury lamp) of the host−guest complex, an hydrogen atom abstraction was proposed to lead to the formation of a benzylic radical species. Due to the encapsulation in the cage, this radical species favored the selective formation of the O-coupled 1,4-adduct 51a ( Figure  29c). The reaction in absence of C36a resulted in a complex mixture of products. Both of these studies show that the reactivity of the photoinduced reactions can be altered by the microenvironment created by the cavity of cages. However, because these substrates can also undergo reactions in absence of C36a, stochiometric amounts of cage are required for proper control of the reaction. However, using quinone 49b instead of 49a remarkably allowed lower C36a loading (50 mol %) due to rapid equilibration of the inclusion complexes to give the resulting coupled product 51b in 80% yield under the same conditions. Yet the competition between reactions in and outside of the cavity still limit the catalytic application of this strategy.
The group of Duan recently reported the self-assembly of [Zn 8 52 6 ] C37, which contains triarylamine moieties in the cage walls (Figure 30a). 180 The photocatalytic reactivity of the C37 and anthraquinone 53 mixture was evaluated during the reduction of chlorobenzene derivatives 54a−g (Figure 30b). Under optimized conditions, dehalogenation reactions proceed fast (30 min) in excellent to good yields of 55a−g (97% to 71%). However, electron rich chlorobenzenes with higher reduction potentials could not be converted by this procedure.
Control experiments showed a significant drop of activity when C37 was absent or replaced by either building block 52 or the Zn-salt. Na 2 CO 3 accelerated the reaction, as was reported for similar photochemical dehalogenation reactions. 181 Interestingly, the power of the light source showed a quadratic relation to the yield of the reaction after 20 min. This second-order dependence of photons in the rate law indicates that a twophoton process is occurring as the rate-determining step.
An ideal model reaction for photochemical cage catalysis is the thermally forbidden [2+2] cycloaddition. By making use of the excited state, the reaction becomes symmetrically allowed and can proceed in high conversions. 182 Using cage C36a and similar bowl Pd 6 56 4 C38 (56 = 2,4,6-tri(pyridine-3-yl)-1,3,5triazine), the Fujita group explored several [2+2] additions. 183 Generally, an excess of alkene guests was added to the hosts in water, followed by filtration to remove unbound guests. The resulting solution was then irradiated with a mercury lamp to give the products encapsulated in the hosts, which were removed by extraction. In their first report they showed that the cavity of C36a gave quantitatively rise to only syn and head-to-tail isomers after the [2+2] additions of 57a,b and 58a,b ( Figure 31) in both homo-and heterodimerization reactions (59−61). 183 In the homodimerization of quinones 58a and 58b, the cavity of the bowl shaped C38 appeared to position the substrates such that the (head-to-tail) syn products were formed in quantitative fashion. Later, they also reported the selective heteroadditions of usually photochemically inert maleimide 62a with 57a to give the corresponding product 63a in high yields (Figure 32a). 184 Host−guest interactions proved critical. Maleimide 62b is poorly bound inside the cavity of C36a, and as a result homodimer 59 is also formed. Chiral cages, accessible by using Pd cornerstones based on chiral cyclohexane diamine (C36b), provide the opportunity to explore enantioselective photoconversion. Indeed, performing the [2+2] addition between maleimide 64 and fluoranthene 65a or 65b (Figure 32b) in the cavity of C36b provides a degree of stereo control over the photoinduced reaction (40 and 50% ee for 65a and 65b, respectively). 185 Interestingly, the capping group based on chiral cyclohexane diamine only induces a small conformational change to the cavity with respect to the parent cage, yet it was already sufficient to convert the substrate with moderate ee, highlighting the potential of host induced (stereo) selectivity.
In this context, cyclodextrins are also frequently used for the [2+2] additions. Due to their natural chirality, they often provide excellent stereoselectivity when reactions are carried out inside these hollow structures, which have been used for various photochemical reactions. These examples are beyond the current scope but are nicely reviewed elsewhere. 186

Host−Guest Charge Transfer Complexes
Electron−donor−acceptor (EDA) complexes may result from the interaction between electron-rich donors and electron-poor acceptors. The formation of such complexes gives rise to a new set of HOMO and LUMO orbitals resulting from their interaction. Due to the stabilizing effect of the noncovalent interaction, this may result in new charge transfer (CT) absorption bands that lie within the visible light energy. This can often be readily observed by a color change of the solution upon addition of guest to host. Irradiation of these CT bands may result in SET from the donor to acceptor, creating a radical pair. Typically, these radical pairs recombine within a few picoseconds, hampering the use in diffusion dependent chemistry. 187 However, intercepting the radical pair with a fast  fluoranthene derivates 65a and 65b in C36b, providing chiral products with moderate ee's. 184,185 chemical reaction from a preorganized substrate could provide a new paradigm for photochemistry. 188 The group of Dasgupta has investigated CT phenomena inside supramolecular cages in detail and provided evidence that such radical pairs can be intercepted. Using the Pd 6 L 4 C36a cage, they investigated the incarceration of electron-rich aromatic guests 9-methylanthracene (67), 1-methylnaphtalene (68), and toluene (69). 189 The resulting host−guest complexes showed CT bands in the visible range, with 67⊂C36a showing the most red-shifted absorption maximum (λ = 475 nm). Using TA measurements, the authors found that a radical cation is generated for all three guests, which decays in the picosecond range to a neutral radical species due to proton transfer with the aqueous solvent. The resulting benzylic radical then has a relatively long lifetime and, in the presence of O 2 , yields various oxidized products (Figure 33). For substrate 67, this reaction with O 2 can lead to substitutions in the central ring (70−73, see Figure 33a). However, for 68, this addition is not possible, resulting in a smaller range of oxidized products (74 and 75, see Figure 33b). Remarkably, the photocatalyzed aerobic oxidation of toluene with 5 mol % C36a resulted in the selective formation of benzaldehyde (76) in 93% yield (Figure 33c). Using TA spectroscopy, the authors were able to study the proton transfer of the guests to the solvent. They found that in the case of 68 and 69, the radical cation state decays with single time constants of 354 and 46 ps, respectively. However, in the case of 67, they found two corresponding time constants for the deprotonation, namely 14 and 47 ps. The authors attributed these different time constants to two different conformations of the encapsulated guest molecules with different solvent exposure, as confirmed by temperature dependent 1 H NMR studies. Therefore, they highlight that the specific preorganization of the substrate may be crucial for the reaction progress and to prevent BET processes.
The proposed mechanism of the photoinduced oxidation of toluene is shown in Figure 34. Upon host−guest complex formation, a CT band forms, which can be excited by 400 nm LEDs. This generates the corresponding radical pair, with the radical cation of toluene and radical anion of C36a. Before BET reproduces the ground-state system, proton transfer from the activated toluene to the aqueous solvent occurs fast (46 ps) to give the neutral benzylic radical. This radical species is long-lived and may be oxidized to the aldehyde by oxygen. The affinity for benzaldehyde in C36a is much smaller compared to toluene due to the increased polarity, which results in guest exchange. The decreased binding affinity in the product is crucial for catalysis in cages with low catalytic loading, which is exemplified in this oxidation.
The group of Sun developed Pd 4 77 2 C39, containing a more spacious cavity than C36a, which showed promising photochromic and redox properties ( Figure 35). 190 Host−guest complexation between C39 and spiro-epoxy naphthalenone 78 was investigated with 1 H NMR, DOSY, and molecular mechanics computations, and the results suggested the encapsulation of up to four guests in C39. Interestingly, the  UV−vis spectrum of the host−guest complex showed an additional CT band around 375 nm. Irradiation of the aqueous [78 4 ⊂C39] with blue LEDs (λ max = 450 nm, 6 W) under inert atmosphere for 8 h afforded the ring-opened aldehyde 79 in 90% yield. In the absence of C30, product 79 is formed in significantly lower yields. However, irradiation of 78 in chloroform with purple LEDs (λ max = 390 nm) quantitatively yielded 79, indicating that the presence of C39 allows for the use of lower energy light due to the formation of a CT complex. Interestingly, when aqueous (78) 4 ⊂C39 is heated to 50°C under air in the dark, a Wacker-like process occurs that provides 80 quantitatively in 2 h. Therefore, the host−guest (78) 4 ⊂C39 complex allows for a condition-controlled environment that provides either 79 under irradiation or 80 thermally.
The same group replaced the central benzene group in 77 by anthracene (81) in C40 to increase the absorption properties of the cage. Interestingly, they found that exposure of flat aromatic guests such as 4-nitrothioanisole (82a) induce a structural change from Pd 6 81 3 capsule C40 to Pd 4 81 2 bowl C41, as determined by 1 H NMR studies and single-crystal X-ray diffraction measurements (Figure 36). 191 Reversibility of this system was also demonstrated by removing the guests, resulting in the reformation of capsule C40. In addition to the structural change of C40 to C41, the photophysical properties also changed upon guest encapsulation. A color change from yellow to red occurred upon exposure of C40 to a solution of 10 equiv of 4-methoxythioanisole (82b) in D 2 O, which was shown by UV−vis studies to be an enhanced blue-light absorption (λ > 425 nm). These absorption changes were attributed to the formation of a CT band between the electron-rich guest and the electron-poor triazine moiety of the host.
The group of Sun performed photocatalytic oxidations with oxygen from the air by irradiating a solution of thioanisoles (82a−e) and capsule C40 with blue LEDs (Figure 37). This efficiently yielded the corresponding sulfoxides (83a−e), whereas overoxidation to the sulfone barely occurred (≤1%).
Control experiments revealed that the reaction does not proceed in absence of cage, air, or light.
To shine light on the nature of the oxidant, the authors added NaN 3 ( 1 O 2 quencher) and benzoquinone (O 2 .− capturing agent) to the reaction mixture. They found that NaN 3 does not inhibit catalysis, whereas the addition of benzoquinone results in significantly reduced catalytic activity. Based on these experiments, the authors concluded that the superoxide anion is responsible for the oxidation of the thioanisoles. Therefore, authors proposed a mechanism for this photocatalytic oxidation (Figure 38), which first involves binding of the guest in capsule C40, inducing structural change to bowl C41, subsequently, visible light excitation results in effective electron transfer by the host−guest complex to facilitate the oxidation of sulfides to sulfoxides by the superoxide anion. Then, due to the increased hydrophilicity and poor shape complementarity of the sulfoxides, the product is replaced by a new substrate. Finally, after full conversion of the sulfides, capsule C40 can be recovered due to the poor binding of sulfoxides in either capsule C40 or bowl C41. The authors thus presented an elegant enzyme-mimicking photocatalytic system that undergoes structural and photophysical changes upon guest binding, however, the exact function of the CT complex is not yet fully clarified.  In a recent study, the group of Sun studied the self-assembly of flexible cage Pd 2 84 2 C42, based on macrocycle 84. 192 Selfassembly gave rise to a mixture of complexes in absence of templating guests, but the presence of a template such as decatungstate 85 resulted in the clean formation of [85⊂C42] self-assemblies (Figure 39a). Interestingly, by adding 85 ( Figure  39b) as a template, quantitative formation of [85⊂C42] was observed based on 1 H NMR, ESI-MS, and XRD analyses. According to the crystal structures, the cavity of C42 adapts its size, depending on the nature of the guest. Because 84 is flexible due to presence of the CH 2 -linkers, there is still space inside the cavity of [85⊂C42] to bind more guest molecules. Moreover, a CT interaction between host C42 and guest 85 was evident from the UV−vis spectrum, which allows for the excitation at longer wavelength light (λ = 465 nm) than is possible for free 85. The authors found that toluene derivatives 86a−g undergo oxidation by air to the respective aldehydes 87a−g in good yield and with high selectivity (Figure 39c). Even the di-and trimethyl substituted derivatives (87e and 87f) mainly underwent a single oxidation (87% and 83%, respectively), although the di-and trialdehydes were also present in the product mixtures. Control experiments revealed that only 84, C42, or 85 cannot perform photocatalysis under the used conditions because they do not absorb the used 465 nm light. However, when 84 and 85 are exposed together to the photocatalytic conditions, the aldehyde is produced (36% yield), indicating that a charge transfer interaction is occurring. Thus, the design of flexible cage C42 allows for the encapsulation of large guests, such as polyoxometalates (e.g., 85), which opens new opportunities toward the encapsulation of these interesting materials that are used frequently in photocatalysis and as redox mediators. 193−195

Photoactive Cages
The design of supramolecular coordination cages that possess light-absorbing functionalities is interesting for molecular sensing and photocatalysis. Toward these ends, several cages have been synthesized that either contain light-absorbing bridging ligands or metals ions. 37,39 With these self-assembled structures in hand, photocatalysis should be feasible in a controlled and efficient manner due to the spacial constraints presented by the cages. Additionally, the photophysical properties undergo changes when the PSs are placed in close and constraint proximity, which have recently been reviewed. 196 In this section, our discussion will focus on the application of these systems in organic photoredox catalysis.

Photosensitizer Installed on the Linkers of the Cage.
One way of producing a light-absorbing supramolecular cage is by selecting desired organic linker groups. The cavity presented by the cage can be used to control substrate concentration as well as provide preorganization during photochemical reactions performed with such cages ( Figure  40).
In 2004, the group of Fujita were the first to show that photoactive cages allowed light-driven chemical reactions after sensitization. Their C36a cage showed absorbance in de UVregion (λ<370 nm) and was able to encapsulate photo inert adamantane (88) in D 2 O. Irradiation of the resulting host− guest [88 4 ⊂C36a] under aerobic conditions yielded oxidized products 89 and 90 in a total of 24% yield (Figure 41a). This regioselective oxidation thus showed one turnover per host molecule, indicating that catalytic turnover was not achieved. Interestingly, irradiating the host−guest complex in absence of O 2 resulted in the formation of 89 and a long-lived blue solution, likely resulting from the one-electron-reduced host. This was substantiated by the EPR spectrum of the blue solution, which contained a radical species (g = 2.002) at cryogenic and room temperatures, indicative of an organic radical. To elucidate the nature of the oxidant, the authors performed 18 O-labeling experiments with either H 2 O or O 2 , which both resulted in the incorporation of 18 O in the products. The authors thus concluded that the mechanism must include the electron transfer from 88 to C36a after excitation, generating the radical pair. However, the radical cation of 88 may then react with either O 2 or H 2 O to give rise to 90 or 89, respectively, whereas C36a .− is only regenerated in the presence of O 2 . However, a clear conclusion on the reason for the single turnover was not reported.
Similarly, cage C36a was used by the same group for photochemically induced anti-Markovnikov hydration of aryl alkynes. 198 They found that aryl alkyne 91 in D 2 O formed a host−guest complex [91 2 ⊂C36a] that undergoes a lightinduced hydration (Figure 41b). Similarly to the previously discussed approach, cage C36a is excited by irradiating UV light (λ < 370 nm), after which a PET occurs with the substrate. According to the authors, C36a accepts an electron from 91, which is then attacked at the anti-Markovnikov position by H 2 O from the solvent. The anti-Markovnikov selectivity was high, as the generated radical at the benzylic position is more stable. BET from C36a to the benzylic radical intermediate, paired with a protonation, resulted in formation of the benzyl ketone 92. It is noteworthy that benzyl ketones are unstable under UV conditions, often resulting in α-cleavage of the carbonyl group. However, C36a acted as a protective shell, preventing undesired further reactivity of the product.
The group of Fujita showed that also triquinacene (93) could undergo selective oxidation in the presence of C36a under similar conditions (Figure 41c). 199 The authors found 94 as the sole reaction product after UV irradiation of the host−guest complex [93 4 ⊂C36a] for 1 h in aerobic environments. The authors found that the reaction does not proceed under anaerobic conditions, thus nominating dissolved O 2 as the oxygen source. The proposed mechanism is similar to the previously discussed photoreactions, starting with excitation of  C36a, resulting in PET from 93 to C36a. The radical cation is then intercepted by O 2 to give the peroxy-radical, which is degraded to the more stable alcohol in 94. Whereas the previous reactions showed no more than one turnover per cage, a yield of 35% was reported in this case, showing promise for catalytic conversion.
In 2019, the Fujita group demonstrated that using C36a also resulted in higher turnovers. 200 They found that aryl-or vinylsubstituted cyclopropanes such as steroid 95 could be demethylenated in the cavity of C36a (100 mol %) under UV light irradiation in high yields (Figure 41d). After full conversion of the starting material, demethylenated 96 was found to be present in 95% 1 H NMR yield and was isolated in 86% yield. Besides 96, also hydrated formaldehyde CH 2 (OD) 2 was found to be present in the reaction mixture. The authors expanded the scope to two vinyl-or aryl-substituted cyclopropanes, which also underwent the desired transformation in good yields. Mechanistic studies showed that a similar mechanism is operating as previously described. However, the authors showed that NO 2 − was produced by performing Griess' method, 201,202 whereas the presence of O 2 was not required for the reaction to occur. Therefore, they proposed that the cyclopropane ring opening is assisted by a NO 3 − counteranion, which is subsequently reduced by the cage to the NO 2 − anion. The authors showed that the resulting NO 2 − counterion containing C36a could be used a total of 10 times by adding more substrate each time. After consumption of 10 NO 3 − counterions, no more turnovers could occur due to product inhibition. Therefore, TON = 10 could be achieved due to the stoichiometric consumption of the counterions.
The group of Raymond have used a similar strategy to carry out photoreactions within their [K 12 Ga 4 96 6 ] (C43) cage ( Figure 42). 203 The cage absorbs UV-A light (325 nm), which could induce photo processes with a suitable substrate. Because C43 is anionic, it efficiently encapsulates cationic species such as cinnamyl ammonium ions (97a−g). Instead of the hypothesized [2+2]-cycloaddition of model substrate 97a, the authors observed a 1,3-rearrangement upon irradiation with UV-A light under reaction conditions to give 98a. This rearrangement does not occur when the cavity of C43 is occupied if the cage is absent or if no light is present, indicating that a photoinduced reaction occurs inside the cavity of C43. The authors observed that the binding affinity plays a key role: when log(K int ) > 2 (97a−c), no side products are observed, however, when log(K int ) < 2 (97d−g), cinnamyl alcohol and tertiary amines (99d−g) are formed. The authors showed with use of TA, UV− vis, and fluorescence measurements that indeed C43 is excited by UV-A-light, followed by PET to the encapsulated substrate. Based on this, the authors postulated that this PET occurs in I from host to guest (Figure 43), generating the corresponding radical pair in II. Then, the amine functionality is released, forming a stabilized allyl radical in III. Then the amine can either stay encapsulated or escape the cavity, depending on the binding affinity. The allylic radical in C43 performs BET, leaving an allyl cation IV, which is either attacked by the amine to yield the corresponding product 98d⊂C41. Interestingly, if the amine is released from the cavity, solvent water may attack the allyl cation but leads to the formation of cinnamyl alcohol, i.e., the nucleophilic attack on the outer carbon.
More recently, pillar [5]arenes were functionalized with a Nphenyl-phenothiazine (100) photoredox catalyst by Schmidt and Esser (Figure 44). 204 Deprotecting a single methoxy group to give a phenol moiety opened up 100 functionalization on the pillar [5]arene to give cavitand C44. This functionalized cavitand showed increased photoredox activity during reductive dehalogenations of alkyl bromides. 205 With Stern−Volmer fluorescence quenching studies, the authors showed that the excited state of C44 is quenched more efficiently by 101a and 101b, in comparison with 100. In addition, the authors also performed catalytic experiments with substrates that could not bind in the pillar [5]arene cavity. In that case, no significant yield difference was observed between C44 and 100, indicating that the cavity induces substrate preorganization, which accelerates the reaction. Similarly, the pillar [5]arenes were also used in polymer-based heterogeneous photoredox catalysis and provided substrate selectivity for substrates that bound in the cavity. 206 Therefore, it seems that the cavity of pillar [5]arenes does not change the reaction outcome of the bound substrates, but it does show increased efficiency for the photochemical conversion.
The group of Su has shown that photoreactions can be carried out with stereoselectivity in the cavity of cages. 207 They designed ruthenium based C33, which can be made enantiopure (Λ-C33 and Δ-C33) depending on the absolute conformation of the Ru(II) centers, which has been shown to promote [2+2] cycloadditions and H 2 formation. 150,208,209 The authors found that naphthol derivatives (103a and 103b) could be encapsulated by C33, as was evident from the 1 H NMR (titration) experiments which suggest that up to ten guests can be encapsulated by C33. Upon irradiation of a [103a⊂C33] solution in H 2 O under aerobic conditions with blue LEDs, oxidative dimerization to 104a occurred quantitatively ( Figure  45a). The authors showed that C33 could be recycled after extraction of the product from the cavity. Dimerization with free ligand 39 in MeOH or with 3a in water were inefficient (<23% yield), indicating that the encapsulation of 103a promotes the dimerization to 104a. However, 104a rapidly racemized in solution and, therefore, bromo-substituted 103b was selected to determine the stereoselectivity of the reaction. Due to poor solubility of 103b in water, a mixture of MeCN:H 2 O (1:1 v/v) solution was selected as the reaction medium. Photocatalysis with Δ-C33 preferentially formed the S-104b, whereas Λ-C33 gave mainly R-104b in modest yields and ee, which varied with the loading of C33 (Figure 45b). A lower loading (5 mol %) of C33 afforded better enantioselectivity because it ensures that a sufficient amount of substrates 103b are bound in the cavity. However, the overall conversion decreased due to a lower loading because there are more free substrates 103b that react inefficiently. Clearly, the chirality induced by homochiral Λ-C33 or Δ-C33 can be transferred, likely due to preorganization of the substrates inside the homochiral cavity. Therefore, the authors established the first photoactive chiral cage that is able to promote regio-and stereoselectivity in photoredox conversions of encapsulated guests.
Cage C33 was also shown to promote the selectivity of [2+2] cycloaddition reactions. 209 As a model reaction, the dimerization of acenaphthylenes 57a and 57c was studied ( Figure 46). Besides high diastereoselective control using a racemic mixture,  with enantiopure Δ-C33 or Λ-C33 enantioselective control was also achieved to form the desired anti-head-to-head stereoisomers 105a and 105b in up to 88% ee. Kinetic studies revealed that the rate of the photochemical reaction is enhanced by an order of magnitude, when compared to a molecular catalyst. The enhancement may be attributed to the pre-organization effect of the molecular cage. Additionally, TA spectroscopy was used to reveal three processes in the picosecond range: (i) ISC to the triplet state in 0.28 ps, (ii) intraligand charge-transfer (ILCT) in 8.54 ps, and (iii) EnT from C33* to 56a in 145 ps. This EnT rate again proved to be an order of magnitude faster than that of free Ru-based ligand 39, which supports the cage effect on rate enhancement. The group of Su then explored the scope of the [2+2] cycloaddition in the cavity of C33. 208 Homocoupling [2+2] additions showed a high tolerance for α,β-unsaturated ketones and esters containing aromatic substituents on the β position. Generally, the transformations occurred preferably to form the syn-isomer in high diastereoselective fashion (up to >20:1) and with isolated yields up to 97% (Figure 47a). Heterocoupling also occurred for similar substrates in similar diastereoselectivity (up to 20:1) in moderate to good yields (40−85%) (Figure 47b). However, the reported scope was limited and may be attributed to two factors: (i) the inherent molecular suitability of the reactants to undergo heterocoupling, and (ii) the coencapsulation efficiency of the two reactants in the supramolecular host. Therefore, the steric and electronic properties of the substrates play a crucial role in the outcome of the photocatalyzed [2+2] heterocoupling.
By making use of a cobalt-based molecular cluster with sulfonylcalix [4]arenes, the group of Cui incorporated lightabsorbing anthraquinone derivatives in the backbone of selfassembled cages. 210 These cages were also able to promote [2+2] cycloadditions of several chalcones with a preference for the head-to-head syn-diastereomer. Although the degree of diastereoselective control was moderate (up to 13:1), it showed significant improvement over the control experiments (7.8:1). This work shows that organic photosensitizers can also be utilized for effective [2+2] cycloadditions inside the cavity of a supramolecular sphere.

Photosensitizers Installed on the Metal Nodes of the Cage.
Transition-metal-based photoredox catalysts are often used, and their incorporation in cages has been established. 39 Subsequently, supramolecular cages have been   Chemical Reviews pubs.acs.org/CR Review developed that contain these photoredox catalysts as their metal nodes. Performing photocatalytic conversions in their cavities makes use of the emergent properties which may result in increased activity, new reactivity, and induced selectivity ( Figure  48). The groups of He and Duan developed heterometallic Ir 2 Co 3 cage (C45) based on common photoredox catalyst fac-Ir III (ppy) 3 (Figure 49a). 211 Self-assembly with 2 equiv of the NH 2 substituted fac-Ir III (ppy) 3 , 6 equiv of 2-formylpyridine and 3 equiv of Co II (ClO 4 ) 2 in acetonitrile afforded C45. Two labile MeCN ligands surround each Co(II) atom in C45, which can be easily exchanged to afford vacant sites and give C45 dual functionality character. Upon addition of CO 3 2− to a solution of C45, the MeCN ligands are released and carbonate is encapsulated to form [CO 3 2− ⊂C45]. Photocatalytic trichloromethylation of 2-acylpyridines 107a− f to the respective products 108a−f could be achieved with C45 in high yields (Figure 49b). 211 Control experiments showed that the individual components did not result in significant conversion of model substate 107a. Based on this, the authors propose that the Co II sites should be in close proximity to the photosensitizers for efficient turnover, highlighting the dualcharacter of C45. Interestingly, CO 3 ⊂C45 did not result in significant catalytic activity (6% yield of 108a) under the same conditions. The authors therefore designed a photoactive cage with dual-functional character that is responsive to external CO 3 2− . The group of Fujita also developed a cage using the Ir III (ppy) 3 catalyst as metal node. Self-assembly of 3 equiv of enantiomerically pure Ir-precursor 109 with 2 equiv of tripodal ligand 110 afforded Λor Δ-C46 (Figure 50a). Although a cavity is formed in the center, it appears to be too narrow to bind a guest. Instead, three binding pockets near the ppy ligands are formed at the edges of the complex, which bind anions such as PF 6 − . These pockets were used to perform a photoinduced E−Z isomerization of E-trifluoro styryl borate 111 to the Z-isomer 112 Within 5 min, the photo stationary state (E:Z = 44:56) was reached. Substrate selectivity was also demonstrated by using the anionic 111 and the neutral analogue 113 (which produces 114, see Figure 50b). It is, however, still unclear what the effect of C46 on the stereoselectivity is of transformations in the small (see Figure 50c) binding pockets.  Figure 51). 213 All three cages (C47−C49) were characterized by 1 H-, 31  Rhodamine B (6) was then selected as acceptor dye as it shows absorption in the same wavelengths as the emission maxima of C48 and C49. The addition of 6 to either C48 or C49 resulted in the unaltered UV−vis spectrum of the cages with the added absorption spectrum of 6. Therefore, the authors suggest that the components have no interactions in the ground state, and thus the cages do not encapsulate 6. Fluorescence studies and time-correlated single photon counting techniques (TCSPC) showed that FRET occurred efficiently between C48 or C49 and 6 (77% or 58%, respectively). Therefore, catalytic transformations were investigated using LH system C48:6 (5:1) for a photoinduced radical cascade cyclization between N,N-dimethylaniline 118 and N-substituted maleimides 119a−f using white LEDs at rt for 12 h (Figure 52). The reaction is initiated by the oxidation of the amine by excited 6, to form radical cation I, which loses a proton to form carboncentered radical II. The maleimide reacts with this radical to form a C−C bond (III), which then ring closes to make dearomatized IV. After sequential oxidation by superoxide from air and deprotonation, this gives products 120a−f in high yield (74−99%). The same reaction conditions in absence of C48 with 6, 1 or 5 mol %, resulted in poor conversion (5% or 14%, respectively). Accordingly, the authors conclude that C48 acts as LH system that collects higher energy photons and performs FRET to 6, increasing the activity of the photocatalytic transformation.
To summarize, four general strategies have been successfully exploited to promote photochemical transformations inside the cavity of cages. The group of Fujita has performed the pioneering work with C36a by using UV light to promote photochemical transformations of various encapsulated substrates. The unique interaction between host and guest has been exploited to excite the emerging CT complexes that give rise to radical reactivity of the guest by the groups of Dasgupta and Sun. Increased interest of the chemical community in photoredox catalysis has stimulated the development of cages that contain photoredox moieties either in the linker or in the node. These hosts can subsequently be used to promote photoredox-type transformations of encapsulated guests, giving rise to unique reactivity. Finally, the group of Su has shown that stereoselectivity can be induced in the cavity of homochiral C33, encouraging future research to pursue different kinds of

TOWARD IMPLEMENTATION IN DEVICES
Up to now, light-driven catalysis utilizing coordination cages has been almost exclusively taking place in homogeneous solution or at least in suspension. As mentioned in the introduction, the ultimate goal of artificial photosynthesis is to combine a lightdriven oxidation reaction (e.g., water oxidation) with lightdriven reduction (e.g., proton reduction) in one material or device. Likewise, for organic synthesis, it would be beneficial to be able to recycle the cage catalysts or use them in a continuous setup. 217 Therefore, immobilization of cages on a heterogeneous support might be useful, as it would allow easier separation of products from the cage catalysts. In this final part, we summarize strategies for immobilization of cages on (electrode) surfaces. The strategies discussed may be used both for the construction of cage-based DS-PECs, and (more generally) for integration of cage catalysts in flow reactors and for developing catalyst recycling strategies.
In recent literature, several methods have been developed to immobilize coordination cages on different heterogeneous supports such as gold, 218,219 alumina, 220 and polymers. 221 Depending on the purpose and the nature of the support, different strategies are used as depicted in Figure 53.
The most simple approach of immobilization is to drop-cast a suspension of the cage on a solid material, e.g., an electrode. Similar strategies have been used to drop cast MOFs for electrochemical studies. 222 Qi et al. have drop-casted C35 dispersed in 500 μL water/0.5 wt % naphthol aqueous solution/ ethanol mixed solvent (V/V/V = 1:2:2) on glassy carbon electrodes in order to perform cyclic voltammetry to investigate electrochemical properties of the cage catalyst. 154 Although this strategy may be useful for electrochemical analysis, such films are typically unstable and will not remain functional for long-term operation.
Another, more synthetically demanding, strategy is the construction of cages on a surface. Au surfaces can generally be functionalized with small molecules by immobilization of thiols and dithiols in the form of self-assembled monolayers (SAM). 223 Reinhoudt and co-workers utilized this strategy to construct coordination cages based on various cavitands and Pd(II) or Pt(II) as connecting nodes on Au(111) surfaces. 218,219 First, Au(111) was prefunctionalized with a SAM of a cavitand containing long alkylchains terminated with thiols by soaking the substrate in a solution of the cavitand (0.1 mM) in EtOH:CHCl 3 (3:1). In a second step, the cages are constructed via metal mediated self-assembly with metal complex M(dppp)-(OTf) 2 (M = Pd(II) or Pt(II), dppp = diphenylphosphinepropyl) and a second (different) cavitand (Figure 53d). Here, the bidentate capping ligand dppp ensures the formation of discrete cages instead of coordination polymers. Analysis of the cage-functionalized Au surface by atomic force microscopy (AFM) revealed two distinct species in ratio 1:1 of 2 and 4 nm size, respectively. While the latter was attributed to the selfassembled coordination cages, the smaller species most likely is the SAM only. Apparently, only 50% of the pre-immobilized SAM was accessible for the construction of surface cages in this method.
As an alternative strategy, the same group prepared homoleptic, pre-assembled cages from a cavitand containing alkylthiolate chains and M(dppp) (Figure 53c). Making use of hydrophobic interactions of the long alkyl chains, these cages could be immobilized on Au surfaces that were prefunctionalized with a SAM of simple alkylthiolate molecules. This strategy allows the immobilization of preformed cages and can in principle also be applied to heteroleptic cages with one type of ligand decorated with alkyl chains. Dalcanale and co-workers followed a similar strategy to assemble and immobilize cages on Si surfaces. 224,225 A different approach is making use of electrostatic interactions between cage and support (Figure 53b). Raymond and coworkers utilized electrostatic interactions to immobilize anionic cage C41 on a cross-linked polymer with cationic functionalities (Figure 54). 221 The cage-functionalized polymer was employed as heterogeneous catalyst for aza-Prins and aza-Cope reactions in a continuous-flow setup. Interestingly, the heterogenized cage catalyst displayed enhanced activity and significantly increased robustness (>60 h) compared to the corresponding homogeneous system in solution.
Nitschke and co-workers developed a general protocol to immobilize cages on acidic or basic alumina. The cavities of the adsorbed cages were still accessible for guest encapsulation. As proof-of-principle, the group demonstrated the separation of Diels−Alder reagents by two immobilized cages. 220  In summary, the immobilization of coordination cages on various heterogeneous support materials has been reported. Raymond and Nitschke have demonstrated the ability of immobilized cages to act as catalysts under continuous flow conditions, in the case of Raymond even with higher performance than the homogeneous analogues. Despite the high potential, the use of such immobilized cages in catalytic applications is very limited to date. Further research in this direction, including electrochemical applications, is required to bridge the gap to application. For cage catalyst recycling, a wider scope of approaches may be available, for example, the formation of polymers with cages as building blocks. As discussed in section 3.2.3, Wisser and co-workers utilized the polymer strategy to connect Rh spheres with ditopic imidazole ligands to form polymer particles. This approach not only resulted in stabilization of the Rh catalysts but also allowed better recycling of the heterogenized catalyst.

CONCLUSION AND OUTLOOK
We have reviewed the application of supramolecular coordination cages in light-driven catalysis and have discussed various strategies. By focusing on artificial photosynthesis and photocatalysis for synthesis in separate sections, a broad overview of the opportunities provided by supramolecular coordination cages is presented. The differences in goals between artificial photosynthesis and photo(redox)catalysis implies that the supramolecular cage strategies are rather different. The literature examples in both sections are organized based on the position of the PS within the supramolecular construct.
Supramolecular cages provide an interesting platform for artificial photosynthesis to organize different required components in close proximity to optimize for function. As highlighted in Figure 4, several strategies are possible, which in general work well for homogeneous reactions. The redox potentials of PS and catalyst can be fine-tuned by supramolecular interactions between host and guest, which can also lower the overpotentials for the desired half-reaction. The same interactions also stabilize the catalytic intermediates, which allows for a more efficient process as the time difference of photophysical processes and catalysis can be bridged more easily. For the overall functioning of the assembly within a device, directional electron transfer becomes crucial. Supramolecular cages with encapsulated catalysts that allow directional electron transfer are therefore promising, especially if such a light-absorbing cage can be immobilized on an electrode surface.
In organic photocatalysis, supramolecular cages can be used to induce selectivity, increase the rate, and change the outcome of the photochemical reactions. This is achieved by preorganization effects (leading to high local concentration) and spatial constraints induced by the supramolecular cavity. Additionally, due to unique host−guest CT interactions, lower energy light may be used, offering multiple advantages. Moreover, a chiral cavity is able to transfer the chirality to the reaction products, which allows for enantioselective photoredox catalysis. However, it is key to understand the ground-state host−guest equilibria and their interactions to achieve the maximum impact from the supramolecular strategy.
Herein, we have shown that supramolecular coordination cages are attractive reaction vessels for photocatalytic methodologies. Although the design and synthesis of new (photoactive) cages is ongoing work and much progress has already been achieved, the use of these novel cages for photocatalysis often still remains underexplored. Moreover, fundamental studies aiming at a more detailed understanding of (photoinduced) electron and energy transfer processes in confined space should provide more insight into the fundamental mechanistic steps. Until now, pump/probe spectroscopic techniques have shown to be very insightful when it comes to characterizing short-lived intermediates and following the electron transfer pathway, thus providing the desired insight. Eventually, both the ground-state properties of the host−guest system as well as the excited-state reactivity need to be well understood to allow efficient excitedstate transformations by design in a rational way.
Interestingly, photoredox catalysis and artificial photosynthesis have recently received enormous interest from the chemical community, yet these communities operate rather independently from one another. Although the goals are largely different, there may be some general lessons to learn, and mutual inspiration is likely. Charge directionality, preventing charge recombination and stabilization of instable intermediates are such common goals. Some of these challenges can be met by execution with and/or inside the cavities of supramolecular hosts. This will expand the scope of photochemically active cages and in turn will help to understand and design even more efficient strategies.
Most of the cages presented in this review are homoleptic cages with only one type of ligand. In some cases, ligand mixtures have been used in a statistical mixture. The application of heteroleptic cages in which different ligand building blocks are used, potentially each with distinct positions and functionality, provides a next level of organization. Pioneering work on construction of such heteroleptic structures has been performed by the groups of Stang, Fujita, and Clever, and has been summarized in a recent review. 69 Finally, the immobilization of supramolecular cages on a heterogeneous support like alumina or polymers may lead to recyclable and long-term stable photocatalysts based on coordination cages. Immobilizing cages on electrode surfaces or semiconductors will allow their implementation on (photo)-electrochemical devices. These steps are without doubt required to bring this field a step closer to application. In addition, he has a research program on catalysis for green energy applications, aiming at solar to fuel devices based on molecular components, also including supramolecular approaches.