Immobilization of molecular catalysts for artificial photosynthesis

Artificial photosynthesis offers a way of producing fuels or high-value chemicals using a limitless energy source of sunlight and abundant resources such as water, CO2, and/or O2. Inspired by the strategies in natural photosynthesis, researchers have developed a number of homogeneous molecular systems for photocatalytic, photoelectrocatalytic, and electrocatalytic artificial photosynthesis. However, their photochemical instability in homogeneous solution are hurdles for scaled application in real life. Immobilization of molecular catalysts in solid supports support provides a fine blueprint to tackle this issue. This review highlights the recent developments in (i) techniques for immobilizing molecular catalysts in solid supports and (ii) catalytic water splitting, CO2 reduction, and O2 reduction with the support-immobilized molecular catalysts. Remaining challenges for molecular catalyst-based devices for artificial photosynthesis are discussed in the end of this review.


Introduction
Natural photosynthesis provides over 99% of nutrients to the living organisms on earth. At the beginning of food chains, photosynthetic autotrophs, i.e. plants, algae, and bacteria, convert inorganic substances into energyintensive food by using sunlight as an energy source. Through the course of 3.4 billion years [1] nature has selected molecular systems to carry out the photochemical processes. Molecular catalysts have abundant advantages: (i) well defined molecular structure and active center (ii) feasible mechanistic study with the help of various spectroscopic techniques, (iii) maximized active sites in homogeneous solutions, (iv) high tunability of their chemical/physical properties by rational molecular design, etc. [2]. These features inspired the development of molecular catalysts for solar energy transduction in artificial photosynthesis.
The concept of storing the solar energy into higherenergy chemicals is the basis of artificial photosynthesis, which is dedicated to producing fuels/high-value chemicals such as H 2 , methane, methanol, or hydrogen peroxide. A number of molecular catalysts have been developed for water splitting [3][4][5][6][7][8][9][10][11][12], CO 2 reduction [13][14][15][16][17][18][19][20][21], and O 2 reduction [22,23]. However, the molecular catalysts in homogeneous solution often suffer from low stability in redox conditions. In addition, debates have been made whether the organometallic catalysts are true homogeneous catalysts or just a precursor of heterogeneous catalysis [24]. One important implication of natural photosynthetic system is that the molecular cofactors are optimally positioned in a protein matrix, so the photochemical processes are carried out in an efficient and robust way [25][26][27][28]. Although replicating the complicated configuration of natural photosystem is not that straightforward, more simple chemistry was addressed to immobilize molecular catalysts in solid matrices, i.e., covalent bonds and non-covalent interactions. This review revisits the techniques for immobilizing molecules in solid supports and summarizes recent advances in artificial photosynthesis using the techniques.

Covalent immobilization
Covalent bond between a catalyst and its support can result in stable immobilization. Several synthetic methods have been proposed for covalent attachment of molecular catalysts onto carbon-based surfaces [29][30][31][32]. A widely used strategy to for direct C-C bond formation was pioneered by Pinson and Savéant is based on electrochemical grafting based on reduction of aryl diazonium molecules. The diazonium salts are reduced on the surface of carbon-based electrode to form aryl radicals, which reacts with the carbon surface to form stable C-C bonds [31][32][33]. This method was successfully adopted for the preparation of quinone decorated electrodes showing electrocatalytic activities for catalytic oxygen reduction (Scheme 1a) [34][35][36][37][38][39].
An alternative approach based on electrochemical oxidation of aryl acetates was reported [40][41][42]. The detailed mechanism is more complicated compared to the diazonium pathway and still under debate [42,43]. González group proposed one possible mechanism, where anodic oxidation of carboxylates results in removal of CO 2 , thus yields arylmethyl radical. The radical can be further oxidized to form methyl carbocation, which forms C-C bond with carbon surface (Scheme 1b) [42].
Another example is the Cu I -catalyzed azide-alkyne cycloaddition (CuAAC) "click" reaction. The click reaction usually uses mild reaction condition and gives high conversion yield. Azide-modification of graphitic carbon surface can be simply carried out by either immersing them in acetonitrile solution of IN 3 or exposing them to IN 3 gas [55][56][57]. Further modification of carbon surface then can be carried out with CuAAC reaction (Scheme 2).
Metal-organic frameworks (MOFs) are emerging platforms for immobilization of molecular catalysts. Development of synthetic strategies of metal-organic frameworks (MOFs) provides rational design of molecular catalysts embedded in them. Two major constitutional components of MOF are metal ions/clusters and linking ligands, providing strategies to realize MOF catalysis based on (i) metal ions/clusters as catalytic sites or (ii) modification of organic ligands with molecular catalysts, respectively [58].

Non-covalent immobilization
Non-covalent immobilization of molecular catalysts allows a facile mix-and-go strategy for catalyst preparation [59,60]. A figure of merit for non-covalent interaction over covalent bonding is that due to the rather weak electronic interplay, the physical/chemical properties of Scheme 1 Covalent attachment of a aryl diazonium salts and b aryl acetates onto carbon surface both anchoring molecules and supporting matrices can be maintained. A number of non-covalent functionalization strategies, using π-π interaction, hydrogen bonding, electrostatic interaction, physical encapsulation, and mechanical interlocking, etc., have been reported so far.
Electrostatic interaction can also play a role to immobilize molecular catalysts. Oxidized CNTs or graphenes can afford strong ionic conjugates. Liang et al. demonstrated functionalization of GO and reduced graphene oxide (rGO) with ammonium salts [72]. Electrostatic functionalization of carbon support can be further expanded to end-functional polymers. Kim group reported stable dispersion of rGO in organic solvents by incorporating amine-terminated polystyrene [73]. Hydrophobicity/hydrophilicity of a molecule-support hybrid system can be effectively tuned by this technique, thus opens wider range of solvent selection.

Photocatalysis
Photocatalysis is a process where light energy is directly converted into chemical energy. It starts with absorption of photons by photosensitizers, followed by cascade charge/energy transfer to catalysts which convert substrates into products. Artificial photosynthesis schemes are multi-electronic processes, which implies multiple charges should be directed and accumulated at the catalytic center to run the photochemical cycles (Scheme 3a). It is a challenging subject, since photon absorption is generally a monoelectronic process, results in single charge production. In this context, it is required to accumulate electron/holes by multiple rounds of single electronic absorption before exciton decay, charge recombination, or back charge transfer. The best example is photosystem II (PSII) in chlorophyll, which is responsible for O 2 evolution in natural photosynthesis. The active units in PSII are optimally positioned in a protein matrix so the charge transfer reactions are kinetically controlled to extract four electrons from a Mn 4 Ca cluster to run water oxidation (Scheme 3b) [74]. It is not straightforward to directly replicate the complex protein matrix of nature, however, efforts have been made to borrow the strategies in more simple chemical bonds or supramolecular chemistry [75][76][77][78][79].
Ni-based MOFs are reported as visible-light driven photocatalysts for CO 2 reduction. Simple ligand modification for PCN-222 and PCN-601 resulted in major products of formate anion and CH 4 , respectively [90,91]. The result implies the importance of ligand selection, which affects the charge separation/transfer kinetics and the reaction sphere morphology. Li group introduced amino groups into the bridging ligands, resulting in increased catalytic activities for CO 2 reduction compared to the MOFs without amino groups [92,93]. Systematic studies on Ti-based MOFs, namely MIL-125(Ti) for amine-free MOF and NH 2 -MIL-125(Ti) for aminofunctionalized MOF, showed the amine moiety on the ligands enhances light absorption in the visible range and increases adsorption of CO 2 (Fig. 3).
Photocatalytic activities can be further enhanced by incorporating photosensitizers and/or catalysts into the MOFs. The UiO MOFs have been extensively studied as immobilizing matrix for molecular catalysts due to their facile ligand tunability and high chemical/photochemical stabilities. The UiO family has a configuration where Zr 6 (OH) 4 O 4 12+ clusters are 12-connected by organic dicarboxylate linkers to form a MOF framework with fcc structure. Various organometallic complexes can replace the linking ligands by filling the well-defined cavities while maintaining the structure of parent UiO MOF.
Kim et al. selected a bipyridine-embedded UiO-67 MOF, namely BUiO, as a self-healing platform for molecular Pt(II) catalysts and Ir(III) photosensitizers. A series of BUiO-based MOFs, namely, Pt n _Ir_BUiO, comprising 2,2′-bipyridine-5,5′-dicarboxylate (L) as a self-healing site, Pt(II)(L)Cl 2 as a H 2 -evolving catalyst, and Ir(III) (ppy) 2 (L) as a photosensitizer were synthesized and tested for photocatalytic H 2 production [95]. When the metal-diimine bonds were cleaved during photolysis, abundant free diamine ligands promoted re-coordination of the cleaved metal ions to recover the molecular catalytic activity (Fig. 5a). Accordingly, the catalysis with Pt 0.1 _Ir_BUiO prolonged more than 6.5 days without significant decrease in its activity, while the control MOF without a free diimine ligand lost molecular catalytic activity after 7.5 h by leaching of Pt and forming colloids (Fig. 5b).

Dye-sensitized inorganic/organometallic hybrids
In inorganic photocatalysis, metal oxides are typically studied due to their stability and facile synthesis. More importantly, multiple charges can be stored in a nanocrystal, which can be further utilized for multielectronic photocatalytic cycles. Depending on the size of a nanocrystal, different maximum number of charges can be stored. Mayer and co-workers have shown a ZnO nanocrystal with 3.5 nm radius can take up to 120 electrons per particle [96]. Most of them work well with UV light and only a few visible-light-responsive photocatalytic systems are reported so far [97,98]. Non-oxide nanocrystals such as CdS or CdSe have suitable bandgaps with strong visible-light absorption, which is suitable for artificial photosynthesis. However, the toxicity of Cd limits their usage in real life. As an analogue to dye-sensitized solar cells, hybridizing inorganic nanocrystals with organic/organometallic dyes offers visible-light driven photocatalysis. Scheme 4 illustrates electronic processes in a Ru(II)-TiO 2 /Pt system for photocatalytic H 2 production: photoexcitation of a Ru(II) photosensitizer (Ru II L x ), followed by (1) charge transfer from the photoexcited states of the Ru(II) dye (Ru II *L x ) to TiO 2 nanocrystal, (ii) back electron transfer from TiO 2 conduction band to the dye, (iii) electron transfer to Pt catalyst, (iv) water reduction to produce H 2 , and (v) regeneration of the Ru II L x ground state by an electron donor (D).

Dye-sensitized photoelectrochemical (DS-PEC) cells
In the early 1970s, Honda and Fujishima reported a photoelectrochemical (PEC) water splitting cell comprising TiO 2 photoanode and Pt counter electrode (Fig. 6) [107,108]. By illumination of the photoanode (λ < 415 nm), O 2 and H 2 were generated at the photoanode and counter electrode, respectively. Since the pioneering works by Honda and Fujishima, significant progress has been made in pursuit of PEC water splitting systems using TiO 2 photoanode or NiO photocathode. However, the large bandgaps of TiO 2 and NiO limits the PEC cells operating with visible light. To tackle this issue, visible-light-absorbing semiconductors (VLA-SCs) have been studied, however, the catalytic activities of the VLA-SCs are rather lower than those of TiO 2 or NiO [109,110]. One promising strategy is sensitizing the large bandgap inorganic semiconductors with molecular photosensitizers. The concept of dye-sensitized solar cell (DSSC) was first proposed by O'Regan and Grätzel in 1991 [111]. Ru(II) complexes were adsorbed on the surface of TiO 2 Scheme 5 a Illustration of non-covalent incorporation of TPPH on the basal plane of rGO surfaces through Strong π-π interaction. b Formation of DPyP-Cr 3+ -GO (I and II) and DPyP-GO. Copyright permissions from ACS [105,106] nanocrystals, which facilitates visible-light absorption by the Ru(II) dyes, followed by electrons from the photoexcited states injected into the TiO 2 nanocrystals. The DSSC was first designed to generate electricity; same strategy can be applied to DS-PEC cells for catalytic reactions. Figure 5 depicts the working principle of a water splitting DS-PEC with dye-sensitized photoanode: (i) light absorption by dyes anchored to n-type semiconductor nanoparticles, (ii) injection of photoexcited electrons into the conduction band of semiconductor, (iii) extraction of electrons from water oxidation catalysts (WOCs), (iv) flow of electrons to counter electrode of Pt through outer circuit, and (v) catalysis on the surface of each electordes with supplied holes/electrons (Scheme 6). As an analogue to the dye-sensitized photoanode, dye-sensitized photocathode is also feasible, by incorporating dyes with p-type semiconductor nanoparticles.
Reductive catalysis such as CO 2 reduction or O 2 reduction requires development of photocathodes with proper catalysts assembly. Inoue and co-workers reported a reduction of CO 2 with dye-sensitized photocathode where a dinuclear Zn(II)-Re(I) complex is anchored to the surface of NiO nanoparticles (Fig. 7b) [128]. Although   Fig. 6 Illustration of a PEC cell reported by Honda and Fujishima: 1 is a TiO 2 photoanode, 2 is a Pt counter electrode, 3 is a sintered glass diaphragm, 4 is an external load, and 5 is a voltameter. Copyright permission from The Chemical Society of Japan [107] Scheme 6 Schematic working principles of a water splitting DS-PEC with dye-sensitized DS-PEC cell. Copyright permission from RSC [112] Fig. 7 DS-PEC cells with a a dye-sensitized photoanode for water splitting and b a dye-sensitized photocathode for CO 2 reduction. Copyright permissions from ACS and Elsevier [113,128] efficient hole injection efficiency from the dye to the NiO was observed, rather low Faradaic efficiency of 6.2% and turnover number of 10 were recorded due to decomposition of the sensitizing Zn(II) porphyrin part. With the same strategy, Ishitani group reported a dye-sensitized photocathode where Ru(II)-Re(I) dyads are anchored to NiO nanoparticles [129]. Enhanced Faradaic efficiency of 71% and turnover number of 32 were reported with the DS-PEC cell. Further improvement of Faradaic efficiency to 93% was achieved by replacing the NiO nanoparticles to CuGaO 2 [130].

Electrocatalysis
In a broader context, electrocatalytic fuel producing systems wired to external power sources, e.g. photovoltaic cells, is also categorized as artificial photosynthesis. Immobilization of molecular catalysts on electrodes requires few requirements for successful electrocatalysis: (i) charge transfer between electrode surface and immobilizing matrix shoud be assured, (ii) the immobilizing matrix should act as a charge transfer channel, and (iii) substrates and products can freely diffuse in the bulk of immobilizing matrix. By fulfilling these requirements, a number of electrocatalytic systems for water splitting, CO 2 reduction, or O 2 reduction have been reported.
A number of studies have been made in the development of molecular catalysts for homogeneous water splitting and they are well reviewed previously [21,131,132]. The huge library of water splitting catalysts provides facile way to directly immobilize them on the electrode surfaces for electrocatalytic systems. Garrido-Barros et al. demonstrated electrocatalytic water oxidation with Cu(II) complexes strongly bound on the basal plane of graphene sheets (Fig. 8a) [133]. Comparative study with the graphene-immobilized system (G-2 2− ) and a homogeneous system (2 2− ) as a control showed remarkable increase in catalytic activity: G-2 2− and 2 2− showed maximum turnover frequencies of 540 s − 1 and 128 s − 1 , respectively. CO 2 reduction in protic media inevitably competes with proton production, resulting in production of H 2 as a byproduct. Interestingly, immobilization of electrocatalysts in solid matrices often enhances their selectivity to specific products by suppressing proton reduction. Wang and co-workers demonstrated highly selective CO 2  Electrocatalytic O 2 reduction with a mechanically interlocked CNT with an anthraquinone macrocycle. Copyright permission from ACS [143] electroreduction over H 2 evolution with a Cu(I) complex adsorbed on graphene electrode [134]. It was proposed that the mesostructure of graphene favors diffusion of CO 2 and limits mass transport of protons. More recently, Zhu and co-workers reported electrocatalytic CO 2 reduction by immobilizing pyrrolidinonyl Ni(II) phthalocyanine (PyNiPc) on CNTs (Fig. 8b) [135]. The PyNiPc/ CNT catalyst suppressed H2 evolution and selectively promoted CO 2 reduction; a high CO/H 2 value of 650 was recorded.
Anthraquinone and its derivatives and its derivatives have been explored as molecular electrocatalysts for O 2 reduction, since anthraquinone is low-cost, has suitable chemical functionality, and high selectivity for H 2 O 2 [136][137][138][139]. Attempts to immobilize anthraquinones on electrodes comprise: (i) directly linking to glassy carbon electrodes through covalent bonds [140], (ii) incorporating in insoluble polymer matrix [141], (iii) and covalently linking to CNTs [142]. Wielend et al.. demonstrated a new approach of "mechanically interlocking" anthraquinone-based catalysts around CNTs [143] The rotaxane architecture prevents dissolution problems of physically adsorbed organic molecules upon electrochemical reduction, while retains the electrochemical properties of the pristine molecule (Fig. 9).

Summary and outlook
The purpose of this review has been to understand the techniques to immobilize molecular catalysts in matrix-supports and incorporating them into artificial photosynthesis. Enhancement of catalytic activity and robustness was achieved with a simple chemistry using covalent bonds or non-covalent interactions. In addition, selectivity for specific product can be also enhanced by anchoring them in well-defined mesoporous structure. Synergetic effects were shown by hybridizing complementary molecular catalysts and solid supports.
The results provide a good blueprint for artificial photosynthesis; the researches so far are in a lab-scale and require next step forward for practical application of artificial photosynthesis in real life. Especially, proper device design is required to address the subjects listed below.

Scalability
The global production of H 2 reached 790 Mton in 2018 (https ://www.iea.org) and the H 2 O 2 market valued 2.49 billion USD in 2019, (https ://www.grand viewr esear ch.com). H 2 and H 2 O 2 are exclusively produced by natural gas reforming and anthraquinone process, respectively. Both of the methods require high energy input; renewable method, such as artificial photosynthesis, is desired to substitute the production. In this regard, scaled artificial photosynthesis with proper device design is required to comply with the large numbers. Atmospheric CO 2 concentration of 420 ppm is generally not high enough to directly run the catalytic reactions. The low CO 2 solubility in water (ca. 1.5 g L − 1 at the standard condition) also limits the scalability of CO 2 reduction. Techniques such as gas-phase electrolysis coupled with flow cells are under development to circumvent the issue.

Products separation
Artificial photosynthesis comprises two half-reactions, resulting in at least two different products. In terms of monetary expenses, simple and cost-effective separation methods are required to separate a desired product from the admixture of reagents, substrates, and crude products. In water splitting, H 2 and O 2 are produced in a same phase of gas. Gas separation membranes are widely used to selectively collect H 2 from the water splitting products. CO 2 reduction is much more complicated due to a variety of possible products, vide supra. The gaseous mixture CO and H 2 bubbles out from the reaction solution; it can be further separated or directly used as a syngas. The liquid phase separation requires rather high energy and complicated techniques.
Artificial photosynthesis is a promising, at the same time, a challenging subject. Multidisciplinary collaboration in chemistry, material science, biology, physics is required for successful realization of artificial photosynthesis.