Recent Advances in Photocatalytic CO 2 Utilisation Over Multifunctional Metal–Organic Frameworks

: The e ﬃ cient conversion of carbon dioxide (CO 2 ) to high-value chemicals using renewable solar energy is a highly attractive but very challenging process that is used to address ever-growing energy demands and environmental issues. In recent years, metal–organic frameworks (MOFs) have received signiﬁcant research attention owing to their tuneability in terms of their composition, structure, and multifunctional characteristics. The functionalisation of MOFs by metal nanoparticles (NPs) is a promising approach used to enhance their light absorption and photocatalytic activity. The e ﬃ cient charge separation and strong CO 2 binding a ﬃ nity of hybrid MOF-based photocatalysts facilitate the CO 2 conversion process. This review summarises the latest advancements involving noble metal, non-noble-metal, and miscellaneous species functionalised MOF-based hybrid photocatalysts for the reduction of CO 2 to carbon monoxide (CO) and other value-added chemicals. The novel synthetic strategies and their corresponding structure–property relationships have also been discussed for solar-to-chemical energy conversion. Furthermore, the current challenges and prospects in practical applications are also highlighted for sustainable energy production.


Introduction
The continuous increase in the greenhouse gas (CO 2 ) concentration in the atmosphere has become a global environmental concern [1,2]. To mitigate this environmental issue, significant research efforts have been directed towards CO 2 capture and storage (CCS) to decrease the level of emissions [3][4][5]. However, the high costs involved in CO 2 capture and transportation restrict the practical application of this process. Direct CO 2 conversion into value-added chemicals such as CO, methanol, and other hydrocarbons [6] is a promising route to overcome the dependency on non-renewable fossil fuels [7][8][9]. However, a significant amount of energy is required to break the highly stable molecules of CO 2 , which have a bond enthalpy of 805 kJ mol −1 . Therefore, the use of abundant and reliable solar energy to carry out the photocatlytic CO 2 reduction is appealing and has received tremendous research attention for the sustainable conversion of solar energy to chemical energy by mimicking the natural photosynthetic process [10][11][12][13][14][15][16]. The CO 2 photoreduction product formation is dependent on the number of electrons and protons transferred in the reaction, as summarised in Equations (1)- (7) in Table 1, along with their reduction potentials compared to normal hydrogen electrodes (NHE) at pH 7.0 [17]. The photocatalytic CO 2 reduction process involves three main steps: (i) visible light absorption by the photocatalyst, (ii) generation of charge carriers, and (iii) efficient adsorption and activation of CO 2 on the surface of photocatalyst [18]. In the past few decades, research efforts have been focused on expanding the visible light absorption by incorporating the multifunctional properties of photocatalysts to promote the conversion efficiency [19][20][21][22][23][24][25]. However, the current development concerning the photocatalytic efficiency for CO 2 reduction is far from satisfactory because of the fast recombination of charge carriers, Table 1. Multielectron reactions of CO 2 reduction and their reduction potentials at pH 7.0 with reference to a normal hydrogen electrode (NHE) at 25 • C and 1 atm.

Equation Number
Reaction Reduction Potential (V) vs. NHE Metal-organic frameworks (MOFs) are a new class of porous solids prepared by linking multi-dentate organic ligands with a metal or metal clusters, leading to the development of single-site solid catalysts [32][33][34][35][36]. They have emerged as promising materials for a range of applications, including acid-based catalysis [37][38][39][40], redox reactions [41][42][43], drug delivery [44][45][46][47][48], asymmetric reactions [49][50][51][52][53][54][55], gas sorption [56][57][58][59], and photocatalysis [60][61][62][63][64], because of their large surface areas and high levels of porosity. In particular, MOFs are promising for photocatalysis due to the flexible ways in which they can be tailored toward a specific property. A number of structural parameters, such as the pore size, shape, topology, dimensionality, and chemical environment, can be altered through careful selection of the metal clusters and organic linkers that constitute their building blocks [65]. The structural arrangement assists in the integration of active sites and photosensitisers in a single material through metal nodes and organic linkers, allowing the development of multifunctional MOFs. Their hierarchical organisation allows the incorporation of both light absorption and catalytic components to achieve efficient solar utilisation in artificial photosynthesis. MOF-5 was amongst the first photoactive MOF materials studied for its charge separation efficiency under light irradiation, which led to the further development of various semiconducting MOFs for photocatalytic applications [66,67]. The functionalisation of MOFs can be carried out in situ or by post-synthetic modification (PSM). In situ functionalisation can be achieved using various ligands or metals during the synthesis of MOFs, while the post-synthetic approach involves the modification of the synthesised parent MOF by introducing organic functionalities, metal nanoparticles (NPs), oxides, carbon heterostructures, amines, and quantum dots [68].
MOFs are promising materials for photocatalytic CO 2 reduction for the following reasons: (i) their light absorption properties can be tuned, leading to improved charge separation efficiency; (ii) they possess the capability to capture CO 2 ; (iii) the constituent metal and organic linkers can participate in the catalytic reaction; (iv) the structural porosity allows access of guest molecules to active sites for efficient photocatalytic reduction [69,70]. Figure 1a illustrates the process of photocatalytic CO 2 reduction over MOFs under solar light irradiation. Although there are several published review papers summarising the synthesis, strategies, and applications in photocatalytic reactions, it is essential to provide a timely progress of the recent research advances in this field [17,18,[71][72][73][74][75][76][77][78][79][80][81][82]. Figure 1b displays the continuous increase in the thousands of publications reported every year based on MOFs and CO 2 reduction reactions, showing the popularity of this field amongst scientists around the world. In this comprehensive review article, we present the recent advancements on the synthesis and catalytic application of functionalised MOFs. The article is divided into three different sections focusing on the visible light responsive MOFs functionalised by (i) noble metal NPs, (ii) non-noble metal NPs, and (iii) Figure 1. (a) Schematic illustration depicting photocatalytic CO 2 reduction into CO and value-added chemicals over metal-organic frameworks (MOFs). (b) Number of publications in the last 20 years along with the averages calculated from Scopus, Web of Science, and SciFinder for the entries "MOFs" and "CO 2 reduction" as of September 2020. Figure 1a reproduced with permission from [81]. Copyright 2020, Elsevier.

Noble-Metal-Based MOFs
Within the photocatalytic literature, the use of noble metals is ubiquitous. Owing to the ability to harness localised surface plasmonic resonance (LSPR), it is possible to tune the excitation frequencies of noble metal NPs to the visible spectral range through variations in particle size, shape, and dielectric environment [83][84][85][86][87][88]. The excited plasmons can undergo decay via radiative [89] and non-radiative processes [90]. Radiative decay involves the re-emission of photons, whereas non-radiative decay takes place via the generation of electron-hole pairs (e − -h + ), known as Landau damping (LD) [83]. LD is a quantum mechanical phenomenon involving the transfer of plasmons into e − -h + pairs within femtoseconds. Additionally, the large-field enhancement effects are also generated inside and near the metal NPs due to the collective electronic oscillations [91,92]. The field enhancement effects have been widely applied in various applications, such as in surface-enhanced Raman scattering (SERS), tip-enhanced Raman scattering (TERS), sensing, and nanophotonics [83]. Plasmonic heating is another possible non-radiative energy dissipation pathway leading to temperature increments of the catalytic system [93,94]. The photo-induced formation of electron-hole pairs makes these NPs ideal candidates for photocatalytic processes, such as the reduction of CO 2 to high-value, renewable fuels such as hydrogen (H 2 ), methanol, and formic acid [95]. The heterogenisation of MOFs with other catalytically active metal such as Pt and Rh has also been discussed as an effective strategy for the preparation of hybrid photocatalysts for CO 2 utilisation. Within this section, the activities of noble-metal-based photocatalysts will be examined, which are summarised in Table 2 below.  [103] Catalysts 2020, 10, 1176 5 of 32

Ag-Based Functionalised MOFs
The introduction of NPs onto MOF architectures can be facilitated through a variety of synthetic strategies, and largely depends on the stability of the MOF towards NP formation conditions. Consideration of the intended site for NPs is important, and timely characterisation is often reported to help ascertain the locations of NPs in terms of whether they reside within the porous frameworks or decorate the outer surfaces. Isolation of NPs within the porous framework is usually achieved through solvothermal methods, by mixing pre-synthesised NPs into the precursor solution containing the metal salts and organic linker. This method "grows" the framework around the NPs, encompassing them within its porous system. Conversely, in situ reduction of NP precursors or the direct deposition of NPs onto MOFs often isolates NPs on the surfaces. The latter of these two strategies provide a less synthetically demanding route for NPs@MOF catalysts, as the former relies heavily on inherent MOF stability towards reductive conditions using borohydrides or H 2 .
Min Choi et al. attached a rhenium-based photocentre (Re I (CO) 3 Cl) to the same bipyridyl dicarboxylic acid linker (ReTC) for mounting within UiO-67 [15]. In the synthesis of the Re n -framework, a solvothermal protocol was established, in which varying stoichiometric ratios of the bipyridyl and photoactive linker were added to the precursor solution to yield frameworks with varying Re contents ( Figure 2a). Under visible light illumination using trimethylamine as a sacrificial electron donor, it was found that the framework consisting of 3 photoactive ReTC linkers per unit cell provided the highest activity for conversion of CO 2 to CO. Structural analysis of the Re 3 -MOF showed that the ReTC linkers covalently bound within the octahedral cavities of the framework and adopted a fac arrangement about the Re-centre, as confirmed by Fourier difference maps, by IR and NMR imaging, and by the presence of a metal-to-ligand charge-transfer (MLCT) band at 400 nm.
Catalysts 2020, 10, x FOR PEER REVIEW 5 of 30 To further improve the photocatalytic activity, the Re3-MOF was coupled to the Ag nanocubes to produce a bifunctional catalyst, as shown in Figure 2b. Following the same solvothermal method used to prepare the Re3-MOF but with the addition of pre-formed Ag nanocubes, the Re3-MOF was crystallised around the nanocubes to produce Ag⊂Ren-MOF, the thickness of which was controlled through the synthesis time. Figure 2c,d display the transmission electron spectroscopy (TEM) image of the Ag⊂Re3-MOF, in which the Re3-MOF was constructed on the surface of Ag nanocubes. The maximum absorption in the ultraviolet-visible (UV-vis) spectra of Ag⊂Re3-MOF compared to the Re3-MOF and Ag nanocube is also shown in Figure 2e. In a similar manner to NPs, the Ag nanocubes can spatially localise the photoactive Re centres to intensified electric fields at their plasmon resonance frequency, which can be orders of magnitudes higher than the incident electromagnetic field. Under identical photocatalytic conditions, the Ag⊂Re3-MOF exhibited a 5-fold enhancement of activity over the Re3-MOF. Notably, Ag nanocubes encapsulated in a thinner MOF layer (16 nm) provided a 7-fold enhancement of the photocatalytic activity (ca. TON = 0.09/hr). This observation was attributed to the exponential decay of the near field from LSPR with increasing distance from the surface of the nanocube. Control experiments using MOF layers with no ReTC (Ag⊂Re0-MOF) showed no activity, ruling out the possibility of Ag being responsible for CO production. Furthermore, upon coating a similar loading of ReTC onto similar sized Cu NPs, which do not observe LSPR characteristics that match the absorption features of ReTC, activity enhancement was not observed. Using a photodeposition method to form NPs, Chen et al. decorated cobalt based-zeolitic imidazolate framework (Co-ZIF-9) samples with varying quantities of Ag NPs in order to enhance the previously reported catalytic potential of bare Co-ZIF-9 [104], as well as its capability as a cocatalyst within a cadmium sulfide(CdS)-bypyridine photocatalytic system [96,105]. In the presence of a photosensitiser ((Ru(bpy)3)Cl2•6H2O) (bpy = 2' ,2-bipyridine) and triethanolamine (TEOA), Co-ZIF-9 provided 13.3 and 19.3 μmol of CO and H2, respectively. Owing to the Co 2+ centres within the framework, the deep-purple appearance of Co-ZIF-9 provides strong adsorption in the visible spectral range above 500 nm. With the addition of Ag NPs, the absorption range of the hybrid catalyst To further improve the photocatalytic activity, the Re 3 -MOF was coupled to the Ag nanocubes to produce a bifunctional catalyst, as shown in Figure 2b. Following the same solvothermal method used to prepare the Re 3 -MOF but with the addition of pre-formed Ag nanocubes, the Re 3 -MOF was crystallised Catalysts 2020, 10, 1176 6 of 32 around the nanocubes to produce Ag⊂Re n -MOF, the thickness of which was controlled through the synthesis time. Figure 2c,d display the transmission electron spectroscopy (TEM) image of the Ag⊂Re 3 -MOF, in which the Re 3 -MOF was constructed on the surface of Ag nanocubes. The maximum absorption in the ultraviolet-visible (UV-vis) spectra of Ag⊂Re 3 -MOF compared to the Re 3 -MOF and Ag nanocube is also shown in Figure 2e. In a similar manner to NPs, the Ag nanocubes can spatially localise the photoactive Re centres to intensified electric fields at their plasmon resonance frequency, which can be orders of magnitudes higher than the incident electromagnetic field. Under identical photocatalytic conditions, the Ag⊂Re 3 -MOF exhibited a 5-fold enhancement of activity over the Re 3 -MOF. Notably, Ag nanocubes encapsulated in a thinner MOF layer (16 nm) provided a 7-fold enhancement of the photocatalytic activity (ca. TON = 0.09/hr). This observation was attributed to the exponential decay of the near field from LSPR with increasing distance from the surface of the nanocube. Control experiments using MOF layers with no ReTC (Ag⊂Re 0 -MOF) showed no activity, ruling out the possibility of Ag being responsible for CO production. Furthermore, upon coating a similar loading of ReTC onto similar sized Cu NPs, which do not observe LSPR characteristics that match the absorption features of ReTC, activity enhancement was not observed.
Using a photodeposition method to form NPs, Chen et al. decorated cobalt based-zeolitic imidazolate framework (Co-ZIF-9) samples with varying quantities of Ag NPs in order to enhance the previously reported catalytic potential of bare Co-ZIF-9 [104], as well as its capability as a co-catalyst within a cadmium sulfide(CdS)-bypyridine photocatalytic system [96,105]. In the presence of a photosensitiser ((Ru(bpy) 3 )Cl 2 ·6H 2 O) (bpy = 2',2-bipyridine) and triethanolamine (TEOA), Co-ZIF-9 provided 13.3 and 19.3 µmol of CO and H 2 , respectively. Owing to the Co 2+ centres within the framework, the deep-purple appearance of Co-ZIF-9 provides strong adsorption in the visible spectral range above 500 nm. With the addition of Ag NPs, the absorption range of the hybrid catalyst expanded to include the higher energy wavelengths in the 350-500 nm region. With 2.07 wt % of Ag NPs, an immediate change in selectivity was observed; CO generation increased to 17.2 µmol and H 2 evolution decreased to 14.8 µmol. The optimum wt % of Ag within the framework was found to be 5 wt %. To elucidate the source of the reduction product, Chen et al. [96] conducted an isotopic experiment using carbon-13-labelled CO 2 ( 13 CO 2 ). Under the same reaction conditions, mass spectrometry reported peaks at m/z 29 and 45 only, with no signal detected at m/z 28. This confirmed that the produced CO was the result of successful reduction of the CO 2 feedstock rather than decomposition of other organics within the photocatalytic system.
Through the reduction of an AgNO 3 precursor using NaBH 4 , Guo et al. decorated the external surfaces of pre-formed chromium terephthalate metal-organic framework, MIL-101(Cr) (Matérial Institut Lavoisier) nanocrystals with~8 nm Ag NPs [97]. Displaying the characteristic deep-green colour of Cr 3+ , bare MIL-101(Cr) strongly absorbs in the visible region. The MOF decorated with Ag NPs expanded the absorption corresponding to the LSPRs of Ag in the absorption wavelength range into the near-IR region (400-1200 nm), providing a greater coverage of the solar spectrum. By efficiently engineering different sizes of MIL-101(Cr) nanocrystals through varying the MOFprecursor concentrations, Guo et al. effectively demonstrated the crystal-size dependency of the photocatalytic system in the photoreduction of CO 2 to CO, H 2 , and CH 4 products. Across a sample range encompassing crystal sizes from 80 to 800 nm, it was shown that the Ag-decorated MIL-101(Cr) catalysts with the smallest crystallite size provided superior catalytic activity in TEAOH, yielding rates of 808.2, 427.5, and 82.1 µmolg −1 h −1 for CO, H 2 , and CH 4 , respectively. The photoreduction rates using 800 nm crystallites were 23, 18, and 35 times lower than the 80 nm system, respectively. Without the inclusion of Ag NPs, the 80 nm catalysts provided the highest rates for bare MIL-101(Cr) but showed vastly decreased performance when compared to those decorated with Ag NPs. Furthermore, under identical conditions and NP loading, Ag NPs dispersed onto Al 2 O 3 provided a very low rate of CO production (8.8 µmolg −1 h −1 ). Using a simplified model, the size-dependent activity was attributed to the densities of unit cells on the edges and corners as percentages of the total surface unit cells. For the 80 nm catalyst, Catalysts 2020, 10, 1176 7 of 32 63.6% of surface unit cells within the crystal lie along edges and corners. This percentage significantly decreases with increasing crystal size, with the 800 nm particle displaying only 6.6% of unit cells along edges and corners. This gives rise to favourable photo-responsive properties for the smaller crystallites, providing superior photocurrents under illumination, and concomitantly the lowest electron transfer resistance. Together, the deposition of Ag NPs provides enhanced light adsorption, while size-engineered MIL-101(Cr) facilitates enhanced charge-separation.
Utilising another member of the ZIF-family, Deng et al. utilised a solvothermal treatment of a preformed, Ag-NP-decorated Co-ZIF-67 framework with an appropriate linker (2,5-dihydroxyterephthalic acid, H 4 DOBDC) in order to induce a transformation to an Ag-NP-functionalised, hollow Co-MOF-74 [98]. Through this novel strategy, Co-MOF-74 particles with a double-layer hollow shell structure were prepared without any remarkable change in the morphology in comparison to original ZIF-67 particles.
Various experiments were conducted by changing the molar ratios of H 4 DOBDC to ZIF-67 (n(L)/n(M)) and the concentration of H 4 DOBDC (c(L)) to yield hollow shells, as shown in Figure 3. Owing to the cobalt density difference between Co-MOF-74 and Co-ZIF-67 and because H 4 DOBDC molecules are bigger than the ZIF-67 channel, the transformation of a ZIF-67 particle to MOF-74 occurs from outside to inside. The low number of Co atoms available to fill in the spaces of the original ZIF-67 particles facilitates the formation of MOF-74's hollow structure. In the instance of higher n(L)/n(M) ratios but low c(L), single-layered hollow structures will be formed. Compared to traditionally synthesised Co-MOF-74 (MOF-74-C), the transformation of ZIF-67 resulted in monodisperse 300 nm particles of Co-MOF-74 (MOF-74-T) composed of ca. 30 nm crystallites. Although providing lower total N 2 adsorption capacity, MOF-74-T gave a much larger external surface area and increased pore volume, which was attributed to the hollow nature of the MOF architecture. To demonstrate the structural advantages of MOF-74-T and AgNPs@MOF-74 compared to the MOF-74-C, the efficiency levels of the co-catalysts were tested in the photocatalytic conversion of CO 2 to CO in the presence of Ru(bpy) 3 Cl 2 and TEOA. It was found that all catalysts displayed H 2 generation ability, however superior performance was shown by the AgNPs@MOF-74 catalyst (CO = ca. 10 µmol/h, H 2 = ca. 17 µmol/h), which was attributed to the LSPR effect of Ag NPs. MOF-74-T provided decreased activity without decoration of NPs, however still provided better performance than the conventionally crystallised MOF by a factor of 1.8 (MOF-74-T: CO = ca. 4.4 µmol/h, H 2 = ca. 8 µmol/h). Deng et al. [98] concluded that the larger external surface area of the hollow structure leads to the exposure of more co-catalytic active sites, while the thin shell allows diffusion of the reactants and bulky Ru(bpy) 3 [98] concluded that the larger external surface area of the hollow structure leads to the exposure of more co-catalytic active sites, while the thin shell allows diffusion of the reactants and bulky Ru(bpy)3Cl2. Reproduced with permission from [98]. Copyright 2019 Wiley-VCH.

Au-Based Functionalised MOFs
In order to replicate the intriguing catalytic performance of ultra-thin materials, such as carbonbased sheets, metals, and metal dichalcogenides, compared to other heterogeneous morphologies, Chen et al. employed a Co 2+ -based porphyrin paddle wheel framework (PPF) to host Au NPs [99].  Reproduced with permission from [98]. Copyright 2019 Wiley-VCH.

Au-Based Functionalised MOFs
In order to replicate the intriguing catalytic performance of ultra-thin materials, such as carbonbased sheets, metals, and metal dichalcogenides, compared to other heterogeneous morphologies, Chen et al. employed a Co 2+ -based porphyrin paddle wheel framework (PPF) to host Au NPs [99]. Figure 4a-f display the scanning electron microscopy (SEM), TEM, and atomic force microscopy (AFM) images of PPF and Au-PPF catalysts with different magnifications. To achieve thin sheets, polyvinylpyrrolidone (PVP) was employed as a structure-directing agent (SDA) to restrict growth along the c-axis, resulting in square sheets of 1 µm width and 20 nm thickness. Similar synthesis in the absence of SDA resulted in thicker nanosheets (120 nm). Reproduced with permission from [98]. Copyright 2019 Wiley-VCH.

Au-Based Functionalised MOFs
In order to replicate the intriguing catalytic performance of ultra-thin materials, such as carbonbased sheets, metals, and metal dichalcogenides, compared to other heterogeneous morphologies, Chen et al. employed a Co 2+ -based porphyrin paddle wheel framework (PPF) to host Au NPs [99]. Figure 4a-f display the scanning electron microscopy (SEM), TEM, and atomic force microscopy (AFM) images of PPF and Au-PPF catalysts with different magnifications. To achieve thin sheets, polyvinylpyrrolidone (PVP) was employed as a structure-directing agent (SDA) to restrict growth along the c-axis, resulting in square sheets of 1 μm width and 20 nm thickness. Similar synthesis in the absence of SDA resulted in thicker nanosheets (120 nm).  To decorate the sheets with pre-synthesised Au NPs, Chen et al. [99] added citrate-modified NPs to a dispersion of PPF-3 nanosheets. In order to achieve adequate adhesion between NPs and the PPF-3 sheets, thermal treatment of the PPF-3 in DMF was required. Before washing, it was found that the negative zeta potential of the NPs (−19.4 mV) made it difficult for them to bind to the surface of the PPF-3, which also produced a negative value (−5.9 mV) arising from the surface-bound PVP. Treatment and removal of the SDA shifted the PPF-3 zeta potential to a positive value (+4.9 mV), allowing favourable electrostatic interaction and uniform dispersion across the external surfaces of the PPF-3 sheets. This observation was also partly attributed to the migration of Co 2+ ions within the PPF-3 sheets.
Using an acetonitrile-ethanol mixture (v/v = 4:1), the photocatalysts produced formic acid (HCOOH) as the sole product. Both Au-NP-decorated PPF-3 sheets provided higher HCOOH production rates compared to their parent frameworks, with the thin-sheet hybrid providing a superior rate of 42.7 µmolg −1 h −1 , representing a 5-fold increase in activity compared to the framework without Au NPs (8.3 µmolg −1 h −1 ). Identical catalysis on thicker sheets without NPs provided further inferior activity, clearly demonstrating the importance of the framework thickness to intrinsic electron transfer and mass transport properties. The substitution of ethanol for other aliphatic alcohols such as n-propanol and n-butanol decreased the rates of HCOOH production to 18.5 and 12.1 µmolg −1 h −1 , respectively.
Following electro-and photochemical analyses, Chen et al. [99] attributed the activity of the Au NP/PPF-3 system to two factors: (i) upon visible light illumination, electrons from the PPF-3 highest occupied molecular orbital (HOMO) transition to the lowest unoccupied molecular orbital (LUMO) and are transferred to the Co 2+ centres, generating Co + ions; (ii) deposition of Au NPs amplifies the local electric fields induced by LSPR, resulting in the absorption of more photons and leading to a higher concentration of Co + ions.
Beyond its function as a porous host, Khaletskaya et al. [100] showed amine functionalised-MIL-125 (NH 2 -MIL-125(Ti)) to be suitable as a sacrificial precursor for the synthesis of TiO 2 . Motivated by the challenges associated with the controllable production of nanoscale TiO 2 particles and the discernible advantages this can bring to the field of photocatalysis, the pyrolytic conversion of MOF precursors offers routes to unusual metal oxide morphologies, as their porosity provides a host matrix for the loading of additional components [100]. Utilising a solvothermal encapsulation synthesis protocol, Khaletskaya et al. [100] placed pre-formed, monodisperse Au NPs (2.1 ± 0.8 nm) within the precursor MOF solution and heated the resulting mixture to induce crystallisation of the framework. This resulted in a hybrid material for which much larger Au NPs (3.8-20 nm) were identified as residing primarily on the outer surface. Following successful crystallisation, the hybrid MOF was heated in dry O 2 at 450 • C for 2 h to convert the decorated framework into an Au/TiO 2 hybrid. Through this synthetic pathway, 300-450 nm nanocrystals retaining the size and morphology of the host framework were produced, validating the synthetic rationale. A similar process was performed using bare NH 2 -MIL-125(Ti) to produce a TiO 2 reference. Notably, analysis by powder X-ray diffraction (PXRD) and Fourier transform (FT) Raman spectroscopy revealed that the presence of Au NPs induced selective crystallisation into the rutile phase, whereas the TiO 2 reference contained traces of anatase [100].
In the photocatalytic formation of CH 4 from a CO 2 /H 2 O mixture (1.5 and 0.6%, respectively) diluted in He, the Au NP/TiO 2 outperformed the TiO 2 reference catalyst, producing ca. 50 ppm CH 4 across a 6 h reaction. Interestingly, to establish the stability and reusability across subsequent cycles, Khaletskaya et al. [100] observed significant increase in activity for the TiO 2 reference catalyst in the 2nd cycle. Citing similar peculiarities to the work of Strunk et al. [106] on Au/Ti/SBA-15, this observation was attributed to the availability of a residual "carbon pool" on the TiO 2 reference catalyst surface at the beginning of the 2nd cycle [106]. These surface species were generated in the 1st cycle and could not be removed through evacuation between cycles. Consistent CH 4 production was observed for the Au NP/TiO 2 hybrid, which clearly outperformed commercial P25 and AUROlite (commercial Au/TiO 2 ) [106].

Pt-and Rh-Based Functionalised MOFs
Employing a different member of the MIL family, Sun et al. compared the activity of Pt and Au NPs on NH 2 -MIL-125(Ti) for the formation of hydrogen and formate (HCOO − ) [101]. MIL-125(Ti) and its amino-based derivative, NH 2 -MIL-125(Ti), have been explored extensively for their photocatalytic potential. Reported first by Férey and coworkers, MIL-125(Ti) absorbs strongly in the UV region and has been shown to provide photochromic behaviour when irradiated in the presence of alcohols [107]. To extend the light absorption into the visible region, the terephthalic acid linker was replaced with the amino derivative, 2-aminoterephthalic acid, by Li and co-workers to prepare NH 2 -MIL-125(Ti) [108]. In carrying out this substitution, it was found that not only had the absorption properties of the framework been extended into the visible region ( Figure 5), the CO 2 sorption capacity had also improved compared to the parent framework.  [101]. MIL-125(Ti) and its amino-based derivative, NH2-MIL-125(Ti), have been explored extensively for their photocatalytic potential. Reported first by Férey and coworkers, MIL-125(Ti) absorbs strongly in the UV region and has been shown to provide photochromic behaviour when irradiated in the presence of alcohols [107]. To extend the light absorption into the visible region, the terephthalic acid linker was replaced with the amino derivative, 2-aminoterephthalic acid, by Li and co-workers to prepare NH2-MIL-125(Ti) [108]. In carrying out this substitution, it was found that not only had the absorption properties of the framework been extended into the visible region ( Figure 5), the CO2 sorption capacity had also improved compared to the parent framework. Under visible light irradiation, NH2-MIL-125(Ti) was shown to be active in the photoreduction of CO2 to provide HCOO -, whereas the non-substituted framework provided no activity, signifying the importance of the amino group in enhancing the adsorption range of the framework [109]. In the formation of NPs via in situ reduction with H2, Sun et al. demonstrated that Pt-and Au-NP-deposited NH2-MIL-125(Ti) samples provide 235 and 40 μmol of H2, respectively, as the major products over 8 Under visible light irradiation, NH 2 -MIL-125(Ti) was shown to be active in the photoreduction of CO 2 to provide HCOO − , whereas the non-substituted framework provided no activity, signifying the importance of the amino group in enhancing the adsorption range of the framework [109]. In the formation of NPs via in situ reduction with H 2 , Sun et al. demonstrated that Pt-and Au-NP-deposited NH 2 -MIL-125(Ti) samples provide 235 and 40 µmol of H 2 , respectively, as the major products over 8 h, whereas no detectable H 2 was observed with NH 2 -MIL-125(Ti). This was attributed to the ability of the noble metal NPs to trap excited electrons from the parent framework and provide redox reaction sites for hydrogen evolution. A similar system was demonstrated by Horiuchi et al. using Pt NPs, providing 33 µmol of H 2 after 9 h [110]. Despite this activity, with reference to the previous work by Li and coworkers [108], Sun et al. anticipated a reduction in HCOO − production upon deposition of noble metal NPs [101]. In the presence of sacrificial electron donor, TEOA, the organic linker within NH 2 -MIL-125(Ti), acts as an antenna, transitioning into a long-lived excited charge separation state upon visible light absorption into the ligand-to-metal charge transfer (LMCT) band. This is followed by the transfer of an electron from the organic linker to the metal node. Within these nodes, octomeric Ti-O units (Ti 8 O 8 (OH) 4 ) are considered as mimics of isolated titanium oxide quantum dots and undergo reduction to the active Ti 3+ photocatalytic species, producing HCOO − . It was suggested that the noble metal NPs would reduce the possibility of electron transfer to the metal nodes, diminishing the generation of the active Ti 3+ species. As predicted, under identical photocatalytic conditions, the framework with Au NPs provided a lower HCOO − (9.06 µmol) yield than the bare NH 2 -MIL-125(Ti) (10.75 µmol). The same was not observed for the framework with Pt NPs, however, which provided a 21% increase in activity (12.96 µmol). The use of electron spin resonance (ESR) spectroscopy and DFT calculations revealed that the disparity in activity was attributed to H 2 spillover from the Pt NPs, resulting in the migration of H to corner-sharing oxygen atoms between Ti octahedra, forming Ti 3+ species. The thermodynamic barrier for this process was calculated to be lower in the case of Pt compared to Au (0.53 vs. 0.94 eV), providing the observed activity [101].
Using an amino derivative of UiO-68(Zr), NH 2 -UiO-68, Guo et al. demonstrated the dependence of CO 2 photoreduction performance on the locations of NPs and the lengths of linkers within the MOF hybrids [102]. Contrasting step-wise and one-pot synthesis strategies, Guo et al. deposited Pt NPs (2 wt %) onto the surface (Pt(2)/NH 2 -UiO-68) by adding the pre-synthesised MOF into a Pt NP suspension or embedded them within the MOF (Pt(2)@NH 2 -UiO-68) via a solvothermal method. Using surface-sensitive, high-resolution X-ray photoelectron spectroscopy (XPS), contrasting Pt peak areas were observed, providing NP locations for catalysts containing equal Pt wt % values, with increased areas observed for surface NPs. The effect of the NP location on the ability of the catalysts to trap electrons was clearly evident during photo and electrochemical measurements, with Pt(2)@NH 2 -UiO-68 displaying a 3-fold increase in transient photocurrent response, a significantly smaller capacitance, and an almost fully quenched photoluminescence emission intensity compared to Pt(2)/NH 2 -UiO-68. Such observations imply that the position of the Pt NPs has a great influence on the charge separation efficiency in the charge utilisation processes.
In the photoreduction of CO 2 under visible light in the presence of TEOA, CO was produced as the sole product using these catalyst systems. Compared to the bare framework, all Pt NPs containing catalysts provided increased CO production rates. In particular, Pt(2)@NH 2 -UiO-68 provided the best performance, displaying a 3-fold enhancement in activity (400.2 µmolg −1 ) compared to the Pt(2)/NH 2 -UiO-68 catalyst with external NPs. This 2 wt % catalyst further outperformed identically synthesised catalysts with loadings of 1 and 4 wt %, as well as a physical mixture of 2 wt % Pt NPs and NH 2 -UiO-68 (37.8 µmolg −1 ), signifying the effect of the NP location on the photocatalytic performance. From these observations and associated Mott-Schottky measurements, Guo et al. [102] suggested that the n-type semiconductor behaviour provided by the NH 2 -UiO-68 framework with a 2.87 eV band gap facilitated electron transfer from the framework to the Pt NPs, effectively promoting rapid charge transfer, as illustrated in Figure 6. The location of LUMO of the MOF framework at −0.60 V and the energy level of Pt at 1.15 V facilitates the efficient electron transfer under visible light irradiation. The electron-rich Pt species exhibit much higher photocatalytic activity in the CO 2 reduction reaction. 10 68 (37.8 μmolg −1 ), signifying the effect of the NP location on the photocatalytic performance. From these observations and associated Mott-Schottky measurements, Guo et al. [102] suggested that the n-type semiconductor behaviour provided by the NH2-UiO-68 framework with a 2.87 eV band gap facilitated electron transfer from the framework to the Pt NPs, effectively promoting rapid charge transfer, as illustrated in Figure 6. The location of LUMO of the MOF framework at −0.60 V and the energy level of Pt at 1.15 V facilitates the efficient electron transfer under visible light irradiation. The electron-rich Pt species exhibit much higher photocatalytic activity in the CO2 reduction reaction. To differentiate the effect of the linker length, an identical catalyst with embedded NPs was synthesised using NH2-UiO-66 (analogous framework by varying the length of the linker) and tested under identical photo-reductive conditions, providing a reduced yield of CO (102.6 μmolg −1 ). This observation was attributed to the ability of the NH2-UiO-68-based framework to accommodate more CO2 guest molecules.
Aside from the decoration of NPs, another synthetic route for imparting noble-metal-based photocatalytic properties upon a MOF is to mount photoactive moieties onto the organic linker, most commonly performed via PSM procedures. Using UiO-67 as the host framework, Chambers et al. demonstrated this synthetic strategy using a rhodium-based linker modelled upon the biphenyl To differentiate the effect of the linker length, an identical catalyst with embedded NPs was synthesised using NH 2 -UiO-66 (analogous framework by varying the length of the linker) and tested under identical photo-reductive conditions, providing a reduced yield of CO (102.6 µmolg −1 ). This observation was attributed to the ability of the NH 2 -UiO-68-based framework to accommodate more CO 2 guest molecules.
Aside from the decoration of NPs, another synthetic route for imparting noble-metal-based photocatalytic properties upon a MOF is to mount photoactive moieties onto the organic linker, most commonly performed via PSM procedures. Using UiO-67 as the host framework, Chambers et al. demonstrated this synthetic strategy using a rhodium-based linker modelled upon the biphenyl dicarboxylic acid linker of the original MOF, as illustrated in Scheme 1 [103]. Using a bipyridyl dicarboxylic acid analogue, a photoactive Cp*RhCl 2 group (Cp* = pentamethylcyclopendadiene) was bound to the pyridyl nitrogen atoms (CpRh(bpydc)). Owing to the large accommodating pores and inherent stability of the crystallised UiO-67 framework towards moisture and temperature, the framework could undergo post-synthetic ligand exchange, a common synthetic protocol used to insert linkers that would not otherwise crystallise into the framework through traditional solvothermal methods [111]. This process exchanged bound biphenyl linkers with the Cp*Rh linker, heterogenising the photoactive rhodium centre and yielding a series of functionalised, isoreticular MOFs, namely, 5%, 10%, 20%, and 35%-Cp*Rh@UiO-67. Notably, confirming the successful linker exchange, the NMR analysis showed peaks of the displaced biphenyl linker in equivalent stoichiometric quantities. An identical process using a bipyridyl analogue, a linker without dicarboxylic acid groups (CpRh(byr)), provided no such peaks. Although uptake of the CpRh(byr) was observed, it was concluded that it resided within the pores of the framework and could easily be washed out through sequential washing.

11
MOFs, namely, 5%, 10%, 20%, and 35%-Cp*Rh@UiO-67. Notably, confirming the successful linker exchange, the NMR analysis showed peaks of the displaced biphenyl linker in equivalent stoichiometric quantities. An identical process using a bipyridyl analogue, a linker without dicarboxylic acid groups (CpRh(byr)), provided no such peaks. Although uptake of the CpRh(byr) was observed, it was concluded that it resided within the pores of the framework and could easily be washed out through sequential washing. In establishing the photocatalytic behaviour of the Cp*Rh@UiO-67 in the presence of TEOA and the photosensitiser, Ru(bpy)3Cl2, the rates of HCOOand H2 generation were compared to that of the homogeneous CpRh(bpydc) linker, as well as the bipyridyl analogue (CpRh(byr)). Other than formate, no other carbon-containing products were detected. In the homogenous phase, CpRh(byr) outperformed CpRh(bpydc) in both HCOOand H2 production ( . Both systems showed comparable stability over several hours, with a decrease in performance attributed to the breakdown of the photosensitiser rather than the photocatalyst. In particular, the longevity and reusability of the Cp*Rh@UiO-67 was confirmed over 6 cycles, and a cumulative photoactivity in excess of 4 days with only a 20% loss in activity was observed, which was ascribed to continual handling of the catalyst (1.4 mg) and perspective mass losses. Studying the effect of Rh % on catalytic activity, it was found that 10% loading was optimal in the heterogeneous system, beyond which formate production decreased rapidly with concomitant increase in H2 formation. Above 10% loading in the homogenous system, all activity significantly decreased, suggesting that the heterogenisation procedure provided a stabilisation effect, reducing the possibility of bimetallic deactivation pathways. Table 3 summarises the MOF hybrid photocatalysts discussed in this section by corelating the specific catalytic descriptors with the observed photocatalytic activity trend, including the crystallite size, nature of the organic linker, surface area, thickness of the MOF framework, location of the NPs, length of the organic linker, and effect of introducing bifunctionality within the MOF framework. Both systems showed comparable stability over several hours, with a decrease in performance attributed to the breakdown of the photosensitiser rather than the photocatalyst. In particular, the longevity and reusability of the Cp*Rh@UiO-67 was confirmed over 6 cycles, and a cumulative photoactivity in excess of 4 days with only a 20% loss in activity was observed, which was ascribed to continual handling of the catalyst (1.4 mg) and perspective mass losses. Studying the effect of Rh % on catalytic activity, it was found that 10% loading was optimal in the heterogeneous system, beyond which formate production decreased rapidly with concomitant increase in H 2 formation. Above 10% loading in the homogenous system, all activity significantly decreased, suggesting that the heterogenisation procedure provided a stabilisation effect, reducing the possibility of bimetallic deactivation pathways. Table 3 summarises the MOF hybrid photocatalysts discussed in this section by corelating the specific catalytic descriptors with the observed photocatalytic activity trend, including the crystallite size, nature of the organic linker, surface area, thickness of the MOF framework, location of the NPs, length of the organic linker, and effect of introducing bifunctionality within the MOF framework.

Non-Noble-Metal-Based MOFs
The use of less expensive non-noble-metals [112][113][114] such as Co, Ni, Fe, and Cu in MOF photocatalysts with high catalytic activity and selectivity is still a challenge when developing low-cost materials [15,[96][97][98]. It is highly desired to develop cost-effective materials with high performance and durability [115][116][117], and research efforts have focussed on developing non-noble-metal MOF analogues that rival the catalytic performance of the noble-metal-based counterparts. Much attention has been paid to Fe-containing MOFs, since iron is an earth-abundant element and iron-containing complexes are commonly used in photocatalysis. Further, Fe-based MOF materials are reported to be directly excited by visible light due to the existence of iron-oxo clusters. In particular, terephthalateand porphyrinic-based Fe-MOFs have seen considerable attention in recent years [118][119][120][121]. Table 4 summarises the case studies of Fe-based MOF hybrid photocatalysts discussed in this section.  In-Fe x TCPP-MOF Solvothermal 300 W Xe ≥400 nm CO 3469 µmolg −1 (24 h) [120] Acknowledging the difficulties in establishing the photocatalytic processes within NH 2 -MIL-125(Ti), Wang et al. reported the photocatalytic performance over three typical Fe-based MOFs and their amino-functionalised derivatives; MIL-101(Fe), MIL-53(Fe), and MIL-88B(Fe) [118]. All three Fe-based MOFs are based upon the terephthalic acid linker bound to Fe-octahedra and the selected MIL frameworks vary only in their topology. MIL-53 is composed of infinite M-OH chains of corner-sharing octahedra bound to four neighbouring chains and is most readily recognised for its associated "breathing" ability. MIL-88B and MIL-101 both share the same trimeric Fe 3 O node, within which three metal octahedra are bound by a single bridging µ 3 oxygen molecule and are connected to a further 6 nodes through the terephthalic linker. Similar to MIL-53, MIL-88B shares similar breathing properties and arranges in staggered layers, forming expandable channels upon host incorporation. Conversely, MIL-101 forms a rigid structure with two differently sizes pores composed of hexagonal and pentagonal windows. Through functionalisation with the amino linker derivative, Wang et al. [118] enhanced the absorption of the MOFs within the visible region beyond 700 nm, as well as the adsorption capacity towards CO 2 compared to the unfunctionalised MOFs.
In the presence of TEOA, all functionalised and non-functionalised MOFs reduced CO 2 to HCOO − , with superior activity provided by NH 2 -MIL-101(Fe), producing 178 µmol in 8 h. Across the MOF series, Wang et al. [118] demonstrated a linear trend between formate production and CO 2 adsorption, with all amino-functionalised MOFs providing enhanced production compared with their parent framework. Further, using the MIL-101-based frameworks, the reduction process was shown to be photocatalytic in nature, as when a cut-off filter was employed to restrict irradiation wavelengths to above those adsorbed by the non-functionalised framework, HCOO − was produced, whereas 5.6 µmol of HCOO − was produced over the amino-derivative. From these observations, Wang et al. concluded that the amino derivative provides an additional excitation pathway for unfunctionalised MOFs, while visible light irradiation results in an excited charge separation state by transferring an electron from the O 2− to Fe 3+ within the metal nodes, generating Fe 2+ , which is capable of reducing CO 2 . For the amino-functionalised derivatives, excitation of the NH 2 functionality followed by an electron transfer from the excited organic linker to the metal centre also generates Fe 2+ , as shown in Scheme 2. This dual-excitation pathway is similar to that observed for the NH 2 -MIL-125(Ti) and NH 2 -Uio-66(Zr) frameworks described above [108,122]. In-FexTCPP-MOF Solvothermal 300 W Xe ≥400 nm CO 3469 μmolg −1 (24 h) [120] Acknowledging the difficulties in establishing the photocatalytic processes within NH2-MIL-125(Ti), Wang et al. reported the photocatalytic performance over three typical Fe-based MOFs and their amino-functionalised derivatives; MIL-101(Fe), MIL-53(Fe), and MIL-88B(Fe) [118]. All three Febased MOFs are based upon the terephthalic acid linker bound to Fe-octahedra and the selected MIL frameworks vary only in their topology. MIL-53 is composed of infinite M-OH chains of cornersharing octahedra bound to four neighbouring chains and is most readily recognised for its associated "breathing" ability. MIL-88B and MIL-101 both share the same trimeric Fe3O node, within which three metal octahedra are bound by a single bridging μ3 oxygen molecule and are connected to a further 6 nodes through the terephthalic linker. Similar to MIL-53, MIL-88B shares similar breathing properties and arranges in staggered layers, forming expandable channels upon host incorporation. Conversely, MIL-101 forms a rigid structure with two differently sizes pores composed of hexagonal and pentagonal windows. Through functionalisation with the amino linker derivative, Wang et al. [118] enhanced the absorption of the MOFs within the visible region beyond 700 nm, as well as the adsorption capacity towards CO2 compared to the unfunctionalised MOFs.
In the presence of TEOA, all functionalised and non-functionalised MOFs reduced CO2 to HCOO -, with superior activity provided by NH2-MIL-101(Fe), producing 178 μmol in 8 h. Across the MOF series, Wang et al. [118] demonstrated a linear trend between formate production and CO2 adsorption, with all amino-functionalised MOFs providing enhanced production compared with their parent framework. Further, using the MIL-101-based frameworks, the reduction process was shown to be photocatalytic in nature, as when a cut-off filter was employed to restrict irradiation wavelengths to above those adsorbed by the non-functionalised framework, HCOO − was produced, whereas 5.6 μmol of HCOO − was produced over the amino-derivative. From these observations, Wang et al. concluded that the amino derivative provides an additional excitation pathway for unfunctionalised MOFs, while visible light irradiation results in an excited charge separation state by transferring an electron from the O 2− to Fe 3+ within the metal nodes, generating Fe 2+ , which is capable of reducing CO2. For the amino-functionalised derivatives, excitation of the NH2 functionality followed by an electron transfer from the excited organic linker to the metal centre also generates Fe 2+ , as shown in Scheme 2. This dual-excitation pathway is similar to that observed for the NH2-MIL-125(Ti) and NH2-Uio-66(Zr) frameworks described above [108,122]. Scheme 2. Scheme for dual-excitation pathways for CO2 reduction in amine-functionalised, Fecontaining MOFs. Reproduced with permission from [118]. Copyright 2014, American Chemical Society.
Most recently, Dao et al. demonstrated the tuning of selectivity using the same aminofunctionalised Fe-MOFs to produce CO by designing a solvent-free system, removing the organic media that are traditionally utilised in photocatalytic reductions, as depicted in Scheme 3 [119]. Scheme 2. Scheme for dual-excitation pathways for CO 2 reduction in amine-functionalised, Fe-containing MOFs. Reproduced with permission from [118]. Copyright 2014, American Chemical Society.
Most recently, Dao et al. demonstrated the tuning of selectivity using the same amino-functionalised Fe-MOFs to produce CO by designing a solvent-free system, removing the organic media that are traditionally utilised in photocatalytic reductions, as depicted in Scheme 3 [119]. Within this system, the MOF photocatalyst and TEOA were uniformly dispersed into a glass fibre film. CO 2 gas was pumped into the system and circulated around a pipeline to increase the contact between the catalyst and CO 2 , generating a gas−solid interface. Compared with a conventional gas−liquid−solid reaction route, the solvent-free gas−solid interface removes the organic medium, providing a more environmentally friendly catalysis route and increasing CO 2 adsorption. Within this system, the MOF photocatalyst and TEOA were uniformly dispersed into a glass fibre film. CO2 gas was pumped into the system and circulated around a pipeline to increase the contact between the catalyst and CO2, generating a gas−solid interface. Compared with a conventional gas−liquid−solid reaction route, the solvent-free gas−solid interface removes the organic medium, providing a more environmentally friendly catalysis route and increasing CO2 adsorption. Under photocatalytic conditions, all MOFs provided non-linear CO production curves, which were attributed to the high boiling point of TEOA. This facet resulted in slow volatilisation of the TEOA, and therefore little CO generation at the beginning of the cycle. For all experiments, no H2 or CH4 gas was evolved and no HCOOH was observed as a liquid product, as determined through NMR. Of the MOFs, NH2-MIL-101(Fe) provided the best catalytic performance over 5 h, with a CO formation rate of 87.6 μmolg −1 , 5.6 times higher than that of the worst performing NH2-MIL-53(Fe) (15.7 μmolg −1 ). As previous identified by Wang et al. [118], the superior performance of the MIL-101 framework was attributed to the enhanced CO2 adsorption capability, as well as coordinatively unsaturated Fe sites within the metal node of the MOF, which become available after removal of pendant H2O ligands through thermal treatment.
Photoluminescence spectra of the Fe MOFs displayed remarkably decreased fluorescence compared to the amino-functionalised linker, which exhibits an emission at 485 nm. This quenching is attributed to the effective LMCT upon the incorporation of the amino group, and the degree of quenching matches the catalytic trend. Further, using photocurrent measurements, Dao et al. [119] showed that NH2-MIL-101(Fe) possesses the highest photocurrent, and concomitantly the smallest resistance amongst the tested MOFs, suggesting that the higher catalytic capacity of NH2-MIL-101(Fe) is attributed to its more efficient charge transfer ability.
Within porphyrinic-based MOF, PCN-221(Fe), Wu et al. encapsulated lead halide perovskite (LHP) quantum dots (QDs) to construct a series of composite photocatalysts of MAPbI3@PCN-221(Fex) (x = 0 -1), as shown in Figure 7 [121]. Based upon 4,4′,4",4"'-(porphyrin-5,10,15,20-tetrayl) tetrabenzoate] (TCPP) linkers connected by zirconium nodes, the PCN-221 framework provides a unique binding site at the centre of the porphyrin linker to localise a single metal ion. By altering the ratio of bare linkers (H2TCPP) to metal-containing linkers (TCPP(Fe)), a series of catalysts with varying Fe contents can be synthesised easily. Driven by the improvement in catalytic performance of LHP QDs for converting CO2 into CO and CH4 when loaded onto graphene oxide or g-C3N4, the strategy employed by Wu et al. [121] was founded upon three considerations: (i) QDs with ultrasmall size and homogeneous distribution can be generated in the pores of the MOF; (ii) the MOF Under photocatalytic conditions, all MOFs provided non-linear CO production curves, which were attributed to the high boiling point of TEOA. This facet resulted in slow volatilisation of the TEOA, and therefore little CO generation at the beginning of the cycle. For all experiments, no H 2 or CH 4 gas was evolved and no HCOOH was observed as a liquid product, as determined through NMR. Of the MOFs, NH 2 -MIL-101(Fe) provided the best catalytic performance over 5 h, with a CO formation rate of 87.6 µmolg −1 , 5.6 times higher than that of the worst performing NH 2 -MIL-53(Fe) (15.7 µmolg −1 ). As previous identified by Wang et al. [118], the superior performance of the MIL-101 framework was attributed to the enhanced CO 2 adsorption capability, as well as coordinatively unsaturated Fe sites within the metal node of the MOF, which become available after removal of pendant H 2 O ligands through thermal treatment.
Photoluminescence spectra of the Fe MOFs displayed remarkably decreased fluorescence compared to the amino-functionalised linker, which exhibits an emission at 485 nm. This quenching is attributed to the effective LMCT upon the incorporation of the amino group, and the degree of quenching matches the catalytic trend. Further, using photocurrent measurements, Dao et al. [119] showed that NH 2 -MIL-101(Fe) possesses the highest photocurrent, and concomitantly the smallest resistance amongst the tested MOFs, suggesting that the higher catalytic capacity of NH 2 -MIL-101(Fe) is attributed to its more efficient charge transfer ability.
Within porphyrinic-based MOF, PCN-221(Fe), Wu et al. encapsulated lead halide perovskite (LHP) quantum dots (QDs) to construct a series of composite photocatalysts of MAPbI 3 @PCN-221(Fe x ) (x = 0 -1), as shown in Figure 7 [121]. Based upon 4,4 ,4",4"'-(porphyrin-5,10,15,20-tetrayl) tetrabenzoate] (TCPP) linkers connected by zirconium nodes, the PCN-221 framework provides a unique binding site at the centre of the porphyrin linker to localise a single metal ion. By altering the ratio of bare linkers (H 2 TCPP) to metal-containing linkers (TCPP(Fe)), a series of catalysts with varying Fe contents can be synthesised easily. Driven by the improvement in catalytic performance of LHP QDs for converting CO 2 into CO and CH 4 when loaded onto graphene oxide or g-C 3 N 4 , the strategy employed by Wu et al. [121] was founded upon three considerations: (i) QDs with ultra-small size and homogeneous distribution can be generated in the pores of the MOF; (ii) the MOF framework can provide improved stability for the LHP QDs, as well as increased catalytic performance; (iii) localisation of the LHP QDs close to the MOFs catalytic centre. The metal ion residing in the TCPP linker shortens the charge transfer distance, enhancing the charge separation efficiency of LHP QDs under visible light irradiation and improving the catalytic activity of the MOFs.  In a CO2-saturated ethyl acetate solution containing a small amount of water as the sacrificial agent, both QD-decorated and parent framework photocatalysts were tested for the formation of CO and CH4. Without QDs, the frameworks displayed selectivity for CH4, whereas CO was the primary product when QDs were encapsulated or frameworks did not include any Fe. No H2 or liquid products such as HCOOH and CH3OH were detected for any photocatalytic systems. Without QDs, the yields of CO and CH4 were 13 and 38 μmolg −1 , respectively, after 25 h of irradiation, after which the frameworks were observed to decompose. Under the same conditions, a framework containing QDs but no Fe displayed little activity. With both QDs and Fe (x < 0) incorporated into the catalysts, all exhibited significant increases in CO2 photoreduction activity. Most notably, the photocatalyst containing 20% TCPP(Fe) linker exhibited the highest photocatalytic activity, achieving 104 μmolg −1 of CO and 325 μmolg −1 of CH4. Furthermore, Wu et al. demonstrated that the encapsulation of LHP QDs provides greater stability to the pristine PCN-221(Fex) photocatalyst series, resulting in superior stability of MAPbI3 QDs compared to those reported on other substrates such a ZIFs [16,114,116]. As such, the MAPbI3@PCN-221(Fex) photocatalysts display significantly enhanced yields for CO2 reduction over extended time periods (1559 μmolg −1 over 80 h, CO (34%) and CH4 (66%)) that are 25-38 times higher than the parent PCN-221(Fex) frameworks.
During 13 CO2/H2 18 O isotope trace experiments, mass spectrum signals for 13 CO (m/z = 29) and 13 CH4 (m/z = 17) were observed, indicating that both CO and CH4 originated from the reduction of CO2 rather than photo-oxidation of ethyl acetate. Similarly, the evolution of 18 O2 (m/z = 36) was observed, demonstrating that the water was the reducing agent rather than ethyl acetate. This observation was replicated upon substitution of the ethyl acetate with acetonitrile, for which photocatalytic cycles produced similar amounts of CO, CH4, and O2. On the basis of these experiments, Wu et al. highlighted the role this catalytic system can play in the practical application of artificial photosynthesis using water as an electron source.
Most recently, Wang et al. explored a similar methodology to achieve a 2-fold interpenetrated MOF framework containing 4 connected indium nodes and metal-doped porphyrin linkers [120]. The identity of the metal at the centre of the porphyrin linkers, as well as the overall metal content within the linkers, was varied to achieve a series of photocatalysts containing indium (In-InTCPP-MOF), To encapsulate QDs into the pores of the MOF, as-prepared PCN-221(Fe x ) was immersed in a PbI 2 solution and then rinsed to remove residual surface PbI 2 . The resulting PbI 2 @PCN-221(Fe x ) was then immersed in a methyl ammonium iodide (MAI) solution to generate MAPbI 3 @PCN-221(Fe x ) composite photocatalysts, with QDs of ca. 1.8 nm within the 2 nm pores.
In a CO 2 -saturated ethyl acetate solution containing a small amount of water as the sacrificial agent, both QD-decorated and parent framework photocatalysts were tested for the formation of CO and CH 4 . Without QDs, the frameworks displayed selectivity for CH 4 , whereas CO was the primary product when QDs were encapsulated or frameworks did not include any Fe. No H 2 or liquid products such as HCOOH and CH 3 OH were detected for any photocatalytic systems. Without QDs, the yields of CO and CH 4 were 13 and 38 µmolg −1 , respectively, after 25 h of irradiation, after which the frameworks were observed to decompose. Under the same conditions, a framework containing QDs but no Fe displayed little activity. With both QDs and Fe (x < 0) incorporated into the catalysts, all exhibited significant increases in CO 2 photoreduction activity. Most notably, the photocatalyst containing 20% TCPP(Fe) linker exhibited the highest photocatalytic activity, achieving 104 µmolg −1 of CO and 325 µmolg −1 of CH 4 . Furthermore, Wu et al. demonstrated that the encapsulation of LHP QDs provides greater stability to the pristine PCN-221(Fe x ) photocatalyst series, resulting in superior stability of MAPbI 3 QDs compared to those reported on other substrates such a ZIFs [16,114,116]. As such, the MAPbI 3 @PCN-221(Fe x ) photocatalysts display significantly enhanced yields for CO 2 reduction over extended time periods (1559 µmolg −1 over 80 h, CO (34%) and CH 4 (66%)) that are 25-38 times higher than the parent PCN-221(Fe x ) frameworks.
During 13 CO 2 /H 2 18 O isotope trace experiments, mass spectrum signals for 13 CO (m/z = 29) and 13 CH 4 (m/z = 17) were observed, indicating that both CO and CH 4 originated from the reduction of CO 2 rather than photo-oxidation of ethyl acetate. Similarly, the evolution of 18 O 2 (m/z = 36) was observed, demonstrating that the water was the reducing agent rather than ethyl acetate. This observation was replicated upon substitution of the ethyl acetate with acetonitrile, for which photocatalytic cycles produced similar amounts of CO, CH 4 , and O 2 . On the basis of these experiments, Wu et al. highlighted the role this catalytic system can play in the practical application of artificial photosynthesis using water as an electron source.
Most recently, Wang et al. explored a similar methodology to achieve a 2-fold interpenetrated MOF framework containing 4 connected indium nodes and metal-doped porphyrin linkers [120].
The identity of the metal at the centre of the porphyrin linkers, as well as the overall metal content within the linkers, was varied to achieve a series of photocatalysts containing indium (In-InTCPP-MOF), cobalt (In-Co 1.71 TCPP-MOF), or iron (In-Fe x TCPP-MOF, x = 0.75/1.91/3.56). Utilising XPS measurements, the oxidation states of the Fe and Co ions within the porphyrins were probed. Through identification of the characteristic Fe 2p 3/2 and Fe 2p 1/2 peaks at binding energies of 711.0 and 724.4 eV, respectively, as well as satellite peaks at 719.1 and 728.7 eV, it was concluded that Fe-containing catalysts possessed Fe residing in a Fe(III) oxidation state. For the Co-containing MOF, binding energies of Co 2p 3/2 and Co 2p 1/2 appearing at 780.3 and 795.8 eV, respectively, indicated a Co(II) species in the porphyrin centres. Across all frameworks, In 3d peaks also confirmed the presence of In(III) species.
Within a CO 2 -saturated ethyl acetate solution using L-ascorbgyl palmitate (L-AP) as the electron sacrificial agent, the Fe-containing frameworks were found to provide the highest catalytic activity, providing CO as the main product, with only trace amounts of CH 4 (ca. 0.5%) being evolved. After 24 h, no H 2 was detected. From the Fe series, In-Fe 1.91 TCPP-MOF was shown to display the highest catalytic efficiency, with a total CO yield of 3469 µmolg −1 in 24 h. The obtained rate of CO formation indicates the superior activity amongst most noble-metal-free MOF photocatalytic systems. Despite being similar in metal content, the In-Co 1.71 TCPP-MOF provided far less CO, yet outperformed the indium-based framework. The dependence on metal identity was further illustrated by time-resolved PL spectroscopy, as the lifetime was shown to decrease from 0.65 ns for the indium-based framework to 0.19 and 0.38 ns for the Fe-and Co-substituted frameworks, respectively, suggesting rapid electron consumption in the Fe-based system.
Within this system, Wang et al. proposed a mechanistic pathway describing electron transfer from the photoexcited porphyrin linker to the single Fe 3+ centres, achieving CO 2 reduction. To complete the photocatalytic cycle, the resulting positively charged electron holes oxidize L-AP. This mechanism was supported by calculation of the conduction band (CB, −0.51 V vs. reversible hydrogen electrode (RHE)) potential, which was found to be more negative than that of the standard reduction potential for CO 2 to CO (−0.12 V vs. RHE), thus facilitating CO 2 reduction under visible light. Further, DFT calculations of the CO 2 reduction steps using Fe− and Co−porphyrin analogues showed that the formation of [Co II (Por 2− )(COOH − )] − from [Co 0 (Por)] 2− is approximately thermoneutral (∆G = −0.25 kcal mol −1 ), whereas the same process within Fe analogues was calculated to be more favourable (∆G = −4.10 kcal mol −1 ), likely due to the higher stability of [Fe II (Por 2− )(COOH − )] − compared to the corresponding Co analogues.
A summary of catalytic descriptors, including the topology, excitation pathway, presence of unsaturated Fe sites, oxidation state, and proximity of QDs to MOF, has been provided for non-noble-metal functionalised MOFs in Table 5 by linking them with the photocatalytic response observed in the CO 2 utilisation.

MOFs Incorporated with Miscellaneous Species
The incorporation of transition metals within the MOF architectures has been found to be effective in terms of decreasing the rate of charge recombination and enhancing the charge separation efficiency. However, the high cost and scarcity of some important metals such as Au, Ag, and Ru limits their industrial application. This section covers the miscellaneous species involving carbon-based heterostructures, such as graphene oxide, quantum dots, and carbon nitride, used to develop non-toxic catalysts with enhanced light-harvesting ability. Table 6 enlists the unique catalytic systems that have been developed recently to achieve excellent photocatalytic performances.  The rational design of multicomponent MOF catalysts by using a carbon-based matrix for enhanced charge transfer was investigated for the first time by Peng et al. [123]. The integration of reduced graphene oxide (rGO) along with a two-dimensional (2D)-MOF (Ni 3 HITP 2 , HITP = 2,3,6,7,10,11-hexaiminotriphenylene) heterostructure was synthesised via Coulombic and π-π interactions for the photocatalytic CO 2 reduction [123]. The photocatalyst was prepared by dispersing and ultrasonicating the Ni 3 HITP 2 MOF in distilled water followed by the addition of rGO under continuous stirring. After vacuum filtration, the composite catalyst, Ni 3 HITP 2 /rGO, was dried and labelled as NHPG-x, where x = 0.5, 2, and 4, representing the specific amounts of MOF used during synthesis. The fabrication of the heterostructure with rGO assisted in the following ways: (i) the high surface area and conductivity of rGO assisted in the improved charge harvesting; (ii) rGO bridged the efficient charge transfer between catalytic active and light absorption components; (iii) the generation of electronic states enhanced CO 2 activation. The change in the electronic states was studied using XPS and X-ray absorption near edge structure (XANES) spectroscopic analysis. The presence of a shoulder peak at 855.6 eV in the Ni L-edge XANES spectrum for NHPG confirmed the intimate contact between the two components.
The photocatalytic CO 2 reduction reaction was carried out using [Ru(bpy) 3 ] 2+ , TEOA, and NHPG as the photosensitiser, electron donor, and heterogeneous catalyst, respectively. A superior CO yield of 10.5 µmolh −1 (TOF = 0.46 min −1 ) over NHPG in comparison to 1.0 µmolh −1 for pure rGO was obtained, along with a high selectivity value of 91.7%. Several control experiments in the absence of light photosensitiser were carried out and CO and H 2 formation was not detected. Moreover, the 13 C isotopic studies revealed the formation of 13 CO, confirming that the source of CO was from CO 2 and not from other carbon sources, such as TEOA and CH 3 CN. The photocatalytic mechanistic pathway for the enhanced CO 2 reduction to CO using NHPG composite films can be summarised in three simple steps: (i) photosensitiser excitation under light irradiation; (ii) photoelectron transfer from rGO to metal active site in the MOF; (iii) CO 2 reduction to CO at Ni-Ni 4 sites. This work is an example of promoting the CO 2 conversion synergistically via improved electron transport and light harvesting through the integration of MOF with the rGO heterostructure.
A similar approach of encapsulating carbon dots (CDs) with the MOF was explored by Li et al. CDs are special nanomaterials with sizes below 10 nm, which also contain various polyaromatic functional groups. They can exhibit strong UV-Vis absorption due to π-π* and n-π* transitions. The NH 2 -UiO-66 MOF was integrated with CDs in order to improve the efficiency of the photocatalytic CO 2 reduction [124]. The authors have compared the influence of the encapsulation of CDs within the pores (CD@NH 2 -UiO-66) and on the surface of MOFs (CD/NH 2 -UiO-66), as illustrated in Scheme 4. The presence of CDs leads to improved charge transfer due to the generation of heterojunctions within the MOFs. This strategy to develop novel MOF hybrid photocatalysts was investigated in the photocatalytic reduction of CO 2 to CO under visible light irradiation conditions. The rate of CO generation with pristine NH 2 -UiO-66 was 3.5 µmol g cat −1 h −1 and for surface-decorated CD/NH 2 -UiO-66 was 4.0 µmol g cat −1 h −1 . A significantly superior catalytic yield of 16.6 µmol g cat −1 h −1 was obtained with embedded CDs in the MOF, which was 4.5 times higher than the pristine NH 2 -UiO-66, highlighting the effect of the CD location on MOF photocatalysts. The photocurrent measurements indicated the highest photocurrent value for the embedded samples, highlighting the efficient charge carrier separation in the MOF. The migration of photogenerated electrons from Zr-oxo clusters of MOFs to CDs is faster in the CD@NH 2 -UiO-66 catalyst in comparison to the movement of electrons on the surface of the MOF to reach decorated CDs in CD/NH 2 -UiO-66. The photocatalyst could be reused ten times without any loss of activity and while maintaining its morphology, as confirmed by SEM and XRD measurements. The incorporation of CDs as electron receptors as well as photosenitisers is an effective strategy for designing and developing photocatalysts for CO 2 utilisation.
carrier separation in the MOF. The migration of photogenerated electrons from Zr-oxo clusters of MOFs to CDs is faster in the CD@NH2-UiO-66 catalyst in comparison to the movement of electrons on the surface of the MOF to reach decorated CDs in CD/NH2-UiO-66. The photocatalyst could be reused ten times without any loss of activity and while maintaining its morphology, as confirmed by SEM and XRD measurements. The incorporation of CDs as electron receptors as well as photosenitisers is an effective strategy for designing and developing photocatalysts for CO2 utilisation. Reproduced with permission from [124]. Copyright 2020, the Royal Society of Chemistry.
Graphitic carbon nitride (g-C3N4) is one of the most studied photocatalysts for CO2 reduction under visible light irradiation, but its limited CO2 adsorption capacity and inferior charge separation ability restrict its use in the design of active photocatalysts. Recently, Liu et al. [125] have explored the cooperative effects of integrating semiconductors with ZIF-8 MOFs to enhance the photocatalytic redox ability, charge separation efficiency, and CO2 adsorption ability. The decoration of tubular g-Scheme 4. Schematic illustration of two synthetic routes toward carbon dot (CD)-decorated NH 2 -UiO-66 particles and CD-embedded NH 2 -UiO-66 particles, respectively. The decoration strategy is based on different surface charges, while the encapsulation strategy is based on using CDs as seeds. Reproduced with permission from [124]. Copyright 2020, the Royal Society of Chemistry.
Graphitic carbon nitride (g-C 3 N 4 ) is one of the most studied photocatalysts for CO 2 reduction under visible light irradiation, but its limited CO 2 adsorption capacity and inferior charge separation ability restrict its use in the design of active photocatalysts. Recently, Liu et al. [125] have explored the cooperative effects of integrating semiconductors with ZIF-8 MOFs to enhance the photocatalytic redox ability, charge separation efficiency, and CO 2 adsorption ability. The decoration of tubular g-C 3 N 4 (TCN) with a zeolitic imidazolate framework-8 (ZIF-8) was explored in the CO 2 /H 2 O vapor to CH 3 OH conversion process [125]. The fabrication of photocatalyst was carried out using an in situ deposition method, in which ZIF-8 nanoclusters were grown on the g-C 3 N 4 at room temperature. The negatively charged g-C 3 N 4 nanotubes have a great affinity for zinc ions, and hence assist in the growth of ZIF-8 nanoclusters without any surface functionalisation. The TEM image of TCN before and after ZIF-8 deposition is shown in Figure 8a,b. To understand the kinetics of the CO 2 reduction reaction, it is essential to study the CO 2 adsorption ability on the surface of the photocatalyst. The linear relationship between the adsorption capacity and relative pressure evidences the interaction between CO 2 and g-C 3 N 4 /ZIF-8. The photocatalytic CH 3 OH production takes place via formation of carbonate, formate, formaldehyde, and methoxy groups, as evidenced by in situ FTIR spectroscopy. Hence, the mechanism for CH 3 OH generation was proposed by 2e − and 2H + pathways: Catalysts 2020, 10, x FOR PEER REVIEW 19 of 30 C3N4 (TCN) with a zeolitic imidazolate framework-8 (ZIF-8) was explored in the CO2/H2O vapor to CH3OH conversion process [125]. The fabrication of photocatalyst was carried out using an in situ deposition method, in which ZIF-8 nanoclusters were grown on the g-C3N4 at room temperature. The negatively charged g-C3N4 nanotubes have a great affinity for zinc ions, and hence assist in the growth of ZIF-8 nanoclusters without any surface functionalisation. The TEM image of TCN before and after ZIF-8 deposition is shown in Figure 8a,b. To understand the kinetics of the CO2 reduction reaction, it is essential to study the CO2 adsorption ability on the surface of the photocatalyst. The linear relationship between the adsorption capacity and relative pressure evidences the interaction between CO2 and g-C3N4/ZIF-8. The photocatalytic CH3OH production takes place via formation of carbonate, formate, formaldehyde, and methoxy groups, as evidenced by in situ FTIR spectroscopy. Hence, the mechanism for CH3OH generation was proposed by 2e − and 2H + pathways: → → → Figure 8c displays the obtained catalytic performances for methanol production using various catalysts, including bulk C3N4 (BCN), tubular C3N4 (TCN), and varied amounts of ZIF-8 on C3N4 (TCNZ). The level of methanol generation using the optimal deposition of ZIF-8 on C3N4 was three times higher than with the BCN photocatalyst. However, it can be seen that the ZIF-8 had bilateral effects on the CH3OH generation by enhancing the CO2 adsorption ability, but the charge transfer was also inhibited due to the poor electrical conductivity of ZIF-8. Hence, it has been demonstrated that reasonable combination and surface functionalisation of TCN photocatalysts can lead to synergetic tuning of photocatalytic processes in the improved conversion of CO2 to CH3OH. Sun et al. studied the effects of different morphologies of ZIF-67 synthesised using the solventinduced method with different volume ratios of methanol and water in the photocatalytic reduction of CO2 to CO [126]. The three different morphologies, namely rhombic dodecahedral, pitaya-like, and 2D leaf-like, were labelled as ZIF-67_1, ZIF-67_2, and ZIF-67_3, respectively. The optical responses of the MOF materials were measured together with the photosensitiser [Ru(bpy)3]Cl2•6H2O to understand the electronic properties. The absorption within the range of 400-700 nm was associated with the LMCT, depicting the light harvesting ability from blue to red regions of the spectrum. The CO2 adsorption capability was also investigated for the morphology-controlled ZIF-67 materials. ZIF- Figure 8. TEM image of (a) tubular g-C 3 N 4 (TCN), (b) tubular g-C 3 N 4 /ZIF-8 (TCNZ8), and (c) photocatalytic generation of CH 3 OH over prepared catalysts. Adapted with permission from [125]. Copyright 2017, Elsevier. Figure 8c displays the obtained catalytic performances for methanol production using various catalysts, including bulk C 3 N 4 (BCN), tubular C 3 N 4 (TCN), and varied amounts of ZIF-8 on C 3 N 4 (TCNZ). The level of methanol generation using the optimal deposition of ZIF-8 on C 3 N 4 was three times higher than with the BCN photocatalyst. However, it can be seen that the ZIF-8 had bilateral effects on the CH 3 OH generation by enhancing the CO 2 adsorption ability, but the charge transfer was also inhibited due to the poor electrical conductivity of ZIF-8. Hence, it has been demonstrated that reasonable combination and surface functionalisation of TCN photocatalysts can lead to synergetic tuning of photocatalytic processes in the improved conversion of CO 2 to CH 3 OH.
Sun et al. studied the effects of different morphologies of ZIF-67 synthesised using the solventinduced method with different volume ratios of methanol and water in the photocatalytic reduction of CO 2 to CO [126]. The three different morphologies, namely rhombic dodecahedral, pitaya-like, and 2D leaf-like, were labelled as ZIF-67_1, ZIF-67_2, and ZIF-67_3, respectively. The optical responses of the MOF materials were measured together with the photosensitiser [Ru(bpy) 3 ]Cl 2 ·6H 2 O to understand the electronic properties. The absorption within the range of 400-700 nm was associated with the LMCT, depicting the light harvesting ability from blue to red regions of the spectrum. The CO 2 adsorption capability was also investigated for the morphology-controlled ZIF-67 materials. ZIF-67_3 had the smallest surface area and exhibited the maximum CO 2 adsorption value of 1.16 mmol g −1 . This was attributed to the (i) presence of flexible cavities between the 2D layers and (ii) strong interactions between the CO 2 and the linker. The photocatalytic CO 2 reduction to CO was performed in the presence of the photosensitiser and with TEOA as the electron donor. Again, ZIF-67_3, having a leaf-like morphology, displayed superior catalytic performance of 3.89 µmol mg −1 h −1 under visible light irradiation when compared to the values of 3.75 µmol mg −1 h −1 for ZIF-67_1 and 3.06 µmol mg −1 h −1 for ZIF-67_2.
A plausible reaction mechanism was proposed, as shown in Figure 9, which was further supported by various operating techniques, including photoluminescence and femtosecond transient absorption  . Possible reaction mechanism for the ZIF-67_3 co-catalysed photoreduction of CO2 by employing [Ru(bpy)3] 2+ and TEOA as the photosensitiser and electron donor, respectively. Reproduced with permission from [126]. Copyright 2018, the Royal Society of Chemistry.
The incorporation of single atoms within the framework is still a challenge because of their tendency to aggregate under reaction conditions. Ye et al. explored the incorporation of single Co atoms implanted within the MOF network and investigated the charge separation efficiency in the porphyrin unit of MOF-525 [80]. In this report, MOF-525, elaborated as Zr6O4(OH)4(TCPP-H2)3, was selected as the framework to incorporate Co single-atom species in the porphyrin unit to form MOF-525-Co. The existence of mononuclear centres of Co was confirmed by the extended X-ray absorption fine structure (EXAFS) spectroscopy. The photocatalytic performance was evaluated under visible light irradiation in the presence of TEOA as an electron donor for CO2 reduction to CO and CH4. The superior catalytic performance of MOF-525-Co with CO and CH4 evolution rates of 200.6 and 36.76 . Possible reaction mechanism for the ZIF-67_3 co-catalysed photoreduction of CO 2 by employing [Ru(bpy) 3 ] 2+ and TEOA as the photosensitiser and electron donor, respectively. Reproduced with permission from [126]. Copyright 2018, the Royal Society of Chemistry.
The incorporation of single atoms within the framework is still a challenge because of their tendency to aggregate under reaction conditions. Ye et al. explored the incorporation of single Co atoms implanted within the MOF network and investigated the charge separation efficiency in the porphyrin unit of MOF-525 [80]. In this report, MOF-525, elaborated as Zr 6 O 4 (OH) 4 (TCPP-H 2 ) 3 , was selected as the framework to incorporate Co single-atom species in the porphyrin unit to form MOF-525-Co. The existence of mononuclear centres of Co was confirmed by the extended X-ray absorption fine structure (EXAFS) spectroscopy. The photocatalytic performance was evaluated under visible light irradiation in the presence of TEOA as an electron donor for CO 2 reduction to CO and CH 4 . The superior catalytic performance of MOF-525-Co with CO and CH 4 evolution rates of 200.6 and 36.76 µmol g −1 h −1 , respectively, was compared with MOF-525-Zn, bare MOF-525, and the H 6 TCPP ligand. Approximately 3-and 6-fold enhancements of the CO and CH 4 evolution values were obtained for MOF-525-Co compared to MOF-525. This was attributed to a higher CO 2 uptake capacity because of an increase in the number of affinity sites after metallisation of MOF-525. It was also demonstrated by experimental investigation and theoretical simulations that the photogenerated electrons could be easily transferred to the active site in MOF-525-Co, leading to the improved charge separation efficiency. Furthermore, Co-incorporated MOF-525 displayed excellent stability at 3 catalytic cycles without any significant loss of activity.
Apart from implanting various species within the MOFs [129,130], many research groups have developed novel materials derived from several monometallic and bimetallic metal-organic frameworks [122,131]. For example, Zhang et al. synthesised cobalt tetraoxide (Co 3 O 4 ) hierarchical nanosheets (HNSs) via two-step synthesis, involving the preparation of Co-MOF using 1,4-naphthalenedicarboxylic acid as the organic linker followed by calcination (Scheme 5) [127]. Interestingly, when the isophthalic acid (H 2 IPA) was used as the organic linker, nanoballs of Co 3 O 4 were formed, which ultimately formed NPs after thermal treatment. The 2D morphology of HNSs was confirmed by SEM images, as shown in Figure 10a was confirmed by SEM images, as shown in Figure 10a,b. The TEM images revealed the existence of smaller particles sized 10-20 nm, as shown in Figure 10c-e.

Scheme 5. Schematic illustration of the synthesis process of Co3O4 hierarchical nanosheets and Co3O4
NPs from the Co MOF. Reproduced with permission from [127]. Copyright 2020, the Royal Society of Chemistry.
The photocatalytic performance and selectivity of CO generation using Co3O4 HNSs was compared with Co3O4 NPs and the commercial Co3O4, as summarised in Figure 10f. The most frequently used components, [Ru(bpy)3]Cl2•6H2O and TEOA, were employed as the photosenitiser and the sacrificial agent for the reaction, respectively. As shown, the Co3O4 HNSs displayed superior activity and selectivity of 39.7 μmol h −1 and 77%, respectively, under visible light irradiation. The apparent quantum efficiency (AQE) for CO generation was found to be 0.42% at light irradiation value of 450 nm. Several control experiments were carried out, including 13 C isotopic studies to confirm the origin of CO as being from CO2 and not from other carbon sources. No CO or H2 formation was observed in the absence of a photosensitiser, TEOA, or visible light irradiation, suggesting their significance in obtaining superior photocatalytic performance. One of the major drawbacks of MOFs is their lack of reusability in practical applications. Herein, the Co3O4 HNSs derived from Co-MOF was found to be stable without any loss of catalytic activity, even after 4 runs. The authors explained the contribution of HNSs to the excellent catalytic performance as being due to two aspects: (i) the formation of HNSs derived from an MOF with higher surface area values, a mesoporous structure, and better CO2 adsorption capacity; (ii) the formation of small NPs with narrow size distribution and fewer oxygen vacancies, leading to improved charge separation and migration efficiency under visible light irradiation. Scheme 5. Schematic illustration of the synthesis process of Co 3 O 4 hierarchical nanosheets and Co 3 O 4 NPs from the Co MOF. Reproduced with permission from [127]. Copyright 2020, the Royal Society of Chemistry. was confirmed by SEM images, as shown in Figure 10a,b. The TEM images revealed the existence of smaller particles sized 10-20 nm, as shown in Figure 10c-e.

Scheme 5. Schematic illustration of the synthesis process of Co3O4 hierarchical nanosheets and Co3O4
NPs from the Co MOF. Reproduced with permission from [127]. Copyright 2020, the Royal Society of Chemistry.
The photocatalytic performance and selectivity of CO generation using Co3O4 HNSs was compared with Co3O4 NPs and the commercial Co3O4, as summarised in Figure 10f. The most frequently used components, [Ru(bpy)3]Cl2•6H2O and TEOA, were employed as the photosenitiser and the sacrificial agent for the reaction, respectively. As shown, the Co3O4 HNSs displayed superior activity and selectivity of 39.7 μmol h −1 and 77%, respectively, under visible light irradiation. The apparent quantum efficiency (AQE) for CO generation was found to be 0.42% at light irradiation value of 450 nm. Several control experiments were carried out, including 13 C isotopic studies to confirm the origin of CO as being from CO2 and not from other carbon sources. No CO or H2 formation was observed in the absence of a photosensitiser, TEOA, or visible light irradiation, suggesting their significance in obtaining superior photocatalytic performance. One of the major drawbacks of MOFs is their lack of reusability in practical applications. Herein, the Co3O4 HNSs derived from Co-MOF was found to be stable without any loss of catalytic activity, even after 4 runs. The authors explained the contribution of HNSs to the excellent catalytic performance as being due to two aspects: (i) the formation of HNSs derived from an MOF with higher surface area values, a mesoporous structure, and better CO2 adsorption capacity; (ii) the formation of small NPs with narrow size distribution and fewer oxygen vacancies, leading to improved charge separation and migration efficiency under visible light irradiation. Figure 10. (a,b) SEM, (c-e) TEM, and (f) photocatalytic reduction of CO2 to CO and H2 over different catalysts. Note: Co3O4-C stands for commercial Co3O4 and Co3O4-N2 refers to the N2 annealing product Co MOF NSs. Adapted with permission from [127]. Copyright 2020, the Royal Society of Chemistry.
Similarly, Chen and co-workers developed an MOF-derived photocatalyst for improved photocatalytic performance for CO2 conversion to CH3OH. They synthesised ZnO/NiO porous hollow spheres through the thermal treatment of Ni-Zn bimetallic organic frameworks (MOFs) [128].  The photocatalytic performance and selectivity of CO generation using Co 3 O 4 HNSs was compared with Co 3 O 4 NPs and the commercial Co 3 O 4 , as summarised in Figure 10f. The most frequently used components, [Ru(bpy) 3 ]Cl 2 ·6H 2 O and TEOA, were employed as the photosenitiser and the sacrificial agent for the reaction, respectively. As shown, the Co 3 O 4 HNSs displayed superior activity and selectivity of 39.7 µmol h −1 and 77%, respectively, under visible light irradiation. The apparent quantum efficiency (AQE) for CO generation was found to be 0.42% at light irradiation value of 450 nm. Several control experiments were carried out, including 13 C isotopic studies to confirm the origin of CO as being from CO 2 and not from other carbon sources. No CO or H 2 formation was observed in the absence of a photosensitiser, TEOA, or visible light irradiation, suggesting their significance in obtaining superior photocatalytic performance. One of the major drawbacks of MOFs is their lack of reusability in practical applications. Herein, the Co 3 O 4 HNSs derived from Co-MOF was found to be stable without any loss of catalytic activity, even after 4 runs. The authors explained the contribution of HNSs to the excellent catalytic performance as being due to two aspects: (i) the formation of HNSs derived from an MOF with higher surface area values, a mesoporous structure, and better CO 2 adsorption capacity; (ii) the formation of small NPs with narrow size distribution and fewer oxygen vacancies, leading to improved charge separation and migration efficiency under visible light irradiation.
Similarly, Chen and co-workers developed an MOF-derived photocatalyst for improved photocatalytic performance for CO 2 conversion to CH 3 OH. They synthesised ZnO/NiO porous hollow spheres through the thermal treatment of Ni-Zn bimetallic organic frameworks (MOFs) [128]. It has been reported that hollow structures can enhance the light absorption ability because of the multiscattering effect and that the porous structure can lead to a high surface area, increased CO 2 adsorption capacity, and improved photocatalytic performance. The decomposition of bimetallic MOFs leads to the homogeneous mixing of n-type ZnO and p-type NiO, leading to the formation of p-n heterojunctions within the hybrid material. The obtained hollow spheres were labeled as ZN-10, ZN-30, and ZN-50 based on the differences in the amounts of precursor solution used in the synthesis. The XPS studies revealed the transfer of electrons from ZnO to NiO in the composite, further assisting in improving the charge separation efficiency. The photocatalytic CO 2 reduction to CH 3 OH over ZN-X hollow spheres was compared with pristine ZnO and NiO metal oxides. The relatively low amounts of CH 3 OH generation on ZnO (0.6 µmol h −1 g −1 ) and NiO (negligible) suggested that pure oxides are not active under the given experimental conditions. However, all ZnO/NiO composites were found be active and a superior catalytic performance was shown by ZN-30 (1.57 µmol h −1 g −1 ), which was three times greater than results obtained over ZnO. The excess amounts of NiO in the catalyst can restrict the light absorption ability and may also have acted as the charge recombination centre that led to the decrease in the photocatalytic activity of ZN-50. Table 7 lists information on the catalytic descriptors discussed in this section, including the surface area, number of affinity sites, organic linker, effect of morphology, and location of CDs for miscellaneous species incorporated into MOFs in photocatalytic CO 2 utilisation. Table 7. Various catalytic descriptors summarising the photocatalytic trends in the miscellaneous species incorporated into MOFs for CO 2 utilisation.

Descriptor Photocatalyst Photocatalytic Trends Ref
Surface area rGO-Ni 3 HITP 2 rGO heterostructure creating high surface area and enhanced conductivity [123] Location of CDs (carbon dots) CD@NH 2 -UiO-66 An approximately 4 times higher catalytic yield with embedded CDs than surface decoration [124] Affinity sites g-C 3 N 4 /ZIF-8 Negatively charged g-C 3 N 4 provided affinity to Zn ions for the growth of ZIF-8 nanoclusters [125] Single atom/Co-MOF Co metallisation increasing the number of affinity sites in MOF-525 for CO 2 adsorption [80] Synergistic interaction g-C 3 N 4 /ZIF-8 Improved CO 2 conversion due to synergistic tuning between g-C 3 N 4 and ZIF-8 [125] Morphology vs. surface area Morphology-controlled ZIF-67 Smallest surface area for 2D leaf-like ZIF-67 with maximum CO 2 adsorption because of flexible cavities absent in rhombic and pitaya-like morphologies [126] Organic linker vs.

Summary and Outlook
In summary, we have compared and contrasted the photocatalytic CO 2 reduction of three different types of MOF-based hybrid materials functionalised by (i) noble metal NPs (Ag, Au, Pd, Ru), (ii) non-noble-metal NPs (Fe, Co), and (iii) miscellaneous species, including carbon dots, r-graphene oxide, and g-C 3 N 4 . These MOF-based hybrid systems, along with their photocatalytic performances, have been listed in Tables 2, 4 and 6. Moreover, we have linked specific catalytic descriptors to the observed photocatalytic trends at the end of each section (Tables 3, 5 and 7). For example, the size of MOF crystallites can significantly impact the photocatalytic performance. The choice and length of an appropriate organic linker, the location of NPs, the surface area, and the oxidation state have been discussed, corelating the trends observed in the photocatalytic CO 2 reduction. The higher surface area of MOFs and their ability to tune the porosity, metal centres, and organic ligands are the reasons for their unique flexibility in various applications, such as in drug delivery, gas sorption, CO 2 capture, and sensing. These versatile structural properties enable photocatalytic applications in hydrogen generation, CO 2 reduction, pollutant degradation, and several organic transformation reactions. The functionalisation of MOFs by metal substitution and the integration of photo-sensitive components or semiconductors assist in improving the visible light harvesting and CO 2 conversion efficiency. We also discussed the preparative techniques and factors affecting the introduction of multifunctionalities within the MOF framework, such as carbon dots, single atoms, quantum dots, and g-C 3 N 4 semiconductors, in enhancing the overall conversion efficiency. MOFs are emerging heterogeneous photocatalysts that are used for CO 2 reduction, especially because of their excellent CO 2 capture capacities.
Despite great progress being made in the design and development of MOF-based photocatalysts for CO 2 reduction, there are still some challenges and limitations in achieving artificial photosynthesis in practical applications. Owing to the organic nature of the linkers, MOFs have much lower stability toward temperature, pressure, moisture, and pH than other zeo-type materials. Many MOFs breakdown at temperatures above 350 • C, dramatically reducing the number of applications. Due to this, MOFs must be considered for more niche, lower temperature syntheses, such as fine chemical synthesis, which can utilise their shape-specific topologies and high metal contents. MOFs have various limitations, including their fast recombination of charge carriers, limited absorption in the visible light region, and complex photocatalytic mechanisms involved in the reduction reaction. The main concern regarding the use of MOFs in the photocatalytic reduction of CO 2 is the small amounts of products formed, which restricts their industrial application. The photocatalytic stability is another important issue faced by materials scientists in maintaining the number of recycling test runs. Many research reports have demonstrated the superior photocatalytic performance of MOFs when compared with inorganic semiconductors; however, continuous decreases in activity during recycling tests have hindered their large-scale use. The reports summarised here include the use of sacrificial reagents to carry out photocatalytic reduction, which are expensive and detrimental to the environment. Research efforts are required to overcome the use of such additional sacrificial reagents. The investigation of photocatalytic CO 2 reduction is still at an early stage and we predict that this article will further assist in providing timely progress and guidelines for the design of photoactive MOF-based catalysts for CO 2 reduction. We look forward to major advancements in this field for practical applications.
Author Contributions: Conceptualisation, P.V. and R.R.; writing-P.V. and D.J.S.; writing-review and editing, P.V., D.J.S., and R.R.; supervision, R.R. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.