Topology- and wavelength-governed CO2 reduction photocatalysis in molecular catalyst-metal–organic framework assemblies

Optimising catalyst materials for visible light-driven fuel production requires understanding complex and intertwined processes including light absorption and catalyst stability, as well as mass, charge, and energy transport. These phenomena can be uniquely combined (and ideally controlled) in porous host–guest systems. Towards this goal we designed model systems consisting of molecular complexes as catalysts and porphyrin metal–organic frameworks (MOFs) as light-harvesting and hosting porous matrices. Two MOF-rhenium molecule hybrids with identical building units but differing topologies (PCN-222 and PCN-224) were prepared including photosensitiser-catalyst dyad-like systems integrated via self-assembled molecular recognition. This allowed us to investigate the impact of MOF topology on solar fuel production, with PCN-222 assemblies yielding a 9-fold turnover number enhancement for solar CO2-to-CO reduction over PCN-224 hybrids as well as a 10-fold increase compared to the homogeneous catalyst-porphyrin dyad. Catalytic, spectroscopic and computational investigations identified larger pores and efficient exciton hopping as performance boosters, and further unveiled a MOF-specific, wavelength-dependent catalytic behaviour. Accordingly, CO2 reduction product selectivity is governed by selective activation of two independent, circumscribed or delocalised, energy/electron transfer channels from the porphyrin excited state to either formate-producing MOF nodes or the CO-producing molecular catalysts.


Introduction
Against the backdrop of increasing global energy requirements, photocatalysis has the potential to reduce the necessary energy for feedstock production and to transform waste into valuable feedstocks. 1 Solar fuel production, merging solar energy harvesting with subsequent chemical energy conversion, is a pertinent route towards such articial photosynthesis. 2 The latter motivates bio-inspired molecular catalysts, ne-tuned to maximise selectivity and atom efficiency while minimising activation energy barriers. 3,4 However, these are oen mired by limited stability and device integrability, as well as modest lightharvesting efficiency for photocatalytic systems operating under broad-spectrum irradiation. 1 Accordingly, current efforts synergise material and catalyst co-design to enable energy-intense reactions, such as selective CO 2 reduction, in lightabsorbing hybrids. 3,4 Here, metal-organic framework (MOF) materials show promise due to their highly controllable chemical and (photo) physical properties enabled by their modular assembly. 5 This unlocks a breadth of differing porosities, topologies, pore sizes, guest inclusion opportunities, optoelectronic properties, and more, which intertwine and deeply condition reaction environments. 6-8 Thus, combining molecular catalysts and MOFs to heterogeneous hybrids enables catalyst stabilisation and recyclability while also gaining control over its proximal and global environment. 7,9,10 While previous reports on heterogeneous thermal MOF catalysis demonstrated that nanoreactors built by pore walls affect reaction activity, selectivity, and substrate diffusion to active sites, 8,11 these ndings are not generally transferable to systems with light-driven catalysis, cascade electroreductions or dened molecular complexes as the active species. Accordingly, challenges remain in the conceptual understanding of the intrinsic effects of a MOF pore environment on such molecular MOF hybrids towards rational design of effective materials for photocatalysis. Although a few studies have investigated pore size variation, topology-based substrate orientation, and active site distance optimisation, these parameters remain poorly understood and systematic studies natural light harvesting antennae protein scaffold, this approach enables the introduction of matching dyadic systems into two different MOF topologies towards monitoring the topology-catalysis relationships. Applying these Re-PCN colloids for photocatalytic CO 2 reduction in acetonitrile (MeCN) with a SED showed a 9-fold TON CO enhancement for PCN-222 over PCN-224 hybrids as well as a 10-fold increase compared to the homogeneous catalyst-porphyrin dyad. Characterisations by powder X-ray diffraction (PXRD), solution and solid-state UV-vis, and attenuated total reectance infrared (ATR-IR) spectroscopy; inductively coupled plasma mass spectrometry (ICP-MS); N 2 adsorption; density functional theory (DFT) calculations and difference envelope density (DED) analysis uncovered mechanistic insights controlling reaction rate (via topology-guided mass and exciton transport) and selectivity (via wavelength-enabled electron/energy channel switching).
The unaltered PXRD data of both MOFs upon Zn metalation and Re catalyst incorporation revealed preserved phase purity and crystallinity while being in accordance with the simulated patterns (Fig. 2a). The ATR-IR spectra of both Re-222 and Re-224 display symmetric and asymmetric CO vibrations at 1926 and 2025 cm −1 and 1927 and 2024 cm −1 , respectively ( Fig. 2b and S4 †), characteristic of 1, and suggesting successful entrapment as well as unaffected molecular structure aer immobilisation in the MOFs. 36 Furthermore, the shi of Zn-N bands at $398 cm −1 aer loading of 1 hints at a comprehensive coordination of the Zn porphyrins (Fig. S5 †). 37 Solid-state UV-vis spectra showed comprehensive absorption in the visible-light region for both metalated MOFs, highlighting the suitability of these materials for broad range visible light harvesting ( Fig. S6 and S7 †). While 222 and 224 display broadened absorption features in line with 2's characteristic Soret band at 420 nm and the Q-bands in the range of 510-650 nm (Fig. S8 †), hallmark symmetry-restoring changes were observed upon Zn metalation potentially indicating quantitative Zn metalation. In contrast, incorporation of 1 does not affect the absorption spectra suggesting that visible light irradiation will mainly result in MOF host excitation (Fig. S1, S6 and S7 †).
ICP-MS performed aer MOF digestion of 222-Zn and 224-Zn revealed Zr : Zn ratios of 1.8 and 1.9, respectively, indicating that every porphyrin is likely metalated (as Zr : Zn theoretical ¼ 3 and 4, respectively, assuming reaction yield ¼ 100%) and that additional Zn is physisorbed in the MOF pores (Table S1 †). Together with the Zn-TCPP, these may act as additional anchoring sites for 1. Nevertheless, the Re content in Re-222 and Re-224 was found similar, $82 and $75 nmol Re mg MOF −1 , respectively, further conrming the successful incorporation of 1 in 222-Zn and 224-Zn ($0.1 catalyst per linker in both cases, see Table S1 †). Loading values are comparable to other molecule-MOF hybrids which oen use carboxylic or phosphonic acids for node-or linker-anchoring. 9,12,19,38 Accordingly, our approach provides a synthetically straightforward alternative anchoring design to traditional acid groups (increasing the design space and assembly exibility) while still affording stable tethering and similar loadings. Nitrogen adsorption measurements showed a reduction in the N 2 quantity adsorbed for both 222-Zn and 224-Zn compared to the corresponding pristine MOFs. The Brunauer-Emmett-Teller (BET) surface area was reduced from $1990 to $1550 m 2 g −1 for 222 and 222-Zn, and from $1980 to $1750 m 2 g −1 for 224 and 224-Zn, respectively ( Fig. S9 and S10 †). Such a decrease upon metalation is common and has been observed for other PCN analogues. 8,34 A further decrease of the N 2 quantity adsorbed upon incorporating 1 is observable for both MOF topologies ( Fig. 2c and S10 †). The N 2 quantity adsorbed in Re-224 is more dramatically reduced compared to Re-222, indicating a higher extent of pore blocking in the former case. This is further illustrated by the BET surface areas and calculated pore size distributions of $1219 m 2 g −1 for Re-222, while a BET surface area of merely $618 m 2 g −1 is obtained for Re-224. Scanning electron microscopy (SEM) images visualised rod-shaped $10 Â 1 mm 222 crystals and cube-type $0.5 Â 0.5 mm 224 crystals, in line with literature ( Fig. S11 and S12 †). 14 Previous studies have reported on the measurements of conduction and valence band levels of PCN MOFs. 39 Alternatively, a recent study on a PCN-molecular catalyst hybrid showed that complete isolation of the organic porphyrin linkers in a rigid structure with insulating metal nodes prevents an electronic coupling, concluding that excited energy levels of PCN MOFs can be regarded as those of the free linker. 19 The oxidation potential of 2-Zn* (S 1 ) was previously calculated at −1.24 V vs. saturated calomel electrode (V SCE ). 19 With E(1/1 − ) ¼ −0.64 V SCE , photoinduced electron transfer from 2-Zn* to 1 in the PCN assemblies is thermodynamically possible, and was reported as the preferred rst electron transfer step for Re-porphyrin dyads, as opposed to reductive quenching by the SED. 24,25,31 In photocatalysis conditions, the corresponding oxidised PS at E ¼ 0.88 V SCE will react with 1,3-dimethyl-2-phenyl-2,3-dihydro-1H-benzo [d]imidazole (BIH) employed as a benign SED to regenerate the PS ground state with E(BIH + /BIH) ¼ 0.33 V SCE (Fig. S13 †). 39,40 In summary, assembly characterisation suggests similar catalyst loading and light-absorbing behaviour for both topologies, however, differing in permanent porosity and, accordingly, likely differing in substrate-catalyst accessibility.

Photocatalysis under broadband irradiation
Photocatalysis experiments were conducted under heterogeneous colloidal conditions, namely suspended MOF assembly (1.5 mg) in wet organic media (4 mL MeCN and 0.12 mL deionised H 2 O) and BIH (225 mg, 1 mmol) as the SED. The resulting suspension was saturated by bubbling CO 2 . Irradiation was performed with a Xenon Light Source (300 W, l ¼ 430-740 nm, $10 mW cm −2 ). CO and H 2 evolution were analysed by gas chromatography of the reaction headspace, while formate formation was investigated by 1 H NMR of the reaction solution.
First, we examined the activity of the homogeneous dyadic PS-catalyst system as a benchmark for the Re-PCN assemblies towards CO 2 reduction. To verify the dyadic system in solution, UV-vis spectra of 0.05 mM 2-Zn in MeCN before and aer the addition of equimolar quantities of 1 were recorded (Fig. S14 †). These displayed a 5 nm bathochromic shi of the Soret band from 414 to 419 nm indicating self-assembly of 1 and 2-Zn in solution. 41 As ReX(CO) 3 (bpy)-based catalysts can photoreduce CO 2 without an external photosensitiser, 22,42 irradiation of 1 in the catalytic media (identical to heterogeneous conditions above) gave selective CO formation with a maximal TON CO of $5 aer 2 h. Repeating the experiments in presence of an equimolar amount of 2-Zn yielded TON CO $10 ( Fig. 3a and Table S2, see ESI † for TON calculation). While low stability under photocatalytic conditions is a common issue of Re catalysts, such a low activity compared to other reports on ReX(CO) 3 (bpy)based catalysts and porphyrin-rhenium-catalyst dyads (with TONs in the low 100s) highlights the inuence of the (pyridine) linking units. 22,[26][27][28]43 Remarkably, irradiating Re-222 colloids in photocatalytic media for 80 h conversely produced $27 mmol HCOO − g −1 h −1 (TON HCOO − per node $5) and subcatalytic amounts of CO ($4 mmol CO g −1 h −1 ; TON CO per Re < 1) (Fig. 3a, b, Tables S3 and S4 †). Incident photon conversion efficiency measurements and apparent quantum yield (AQY) calculations (details in ESI, Tables S5 and S6 †) gave values of $0.01% (CO) and $0.10% (formate) at l ¼ 450 nm. AQYs in the range of 0.1-3.0% are common for colloidal visible light-driven CO 2 reduction for comparable systems. 19,44,45 Control photocatalysis experiments using Re-free 222-Zn showed no CO evolution but a comparable formate production rate, while experiments run without BIH or light yielded in all cases no detectable CO, H 2 or formate (Table S3 †). Product generation can be ascribed to light absorption by 2 followed by electron transfers to the CO-selective Re catalysts or to the Zr-oxo node-catalysts for formate production. The latter is in line with previous reports, and suggest two competing CO 2 reduction mechanisms concomitantly occurring in Re-222. 17,18,39,46 The corresponding Re-224-based experiments showed overall lower TONs and rates than for Re-222 but similar reactivity trends favouring formate over CO production, i.e., $2 mmol CO g −1 h −1 and $9 mmol HCOO − g −1 h −1 (Fig. 3a, b and Table S3 †).
While interfacing the Re catalyst amid the MOF signicantly increases its lifetime compared to when employed in homogenous conditions, its catalytic involvement is limited as formate evolution prevails. Together with impeded CO evolution, differing numerical performance between the two hybrid MOFs underlines a key inuence of the host on the photocatalytic activity.
Post-catalysis PXRD analysis of Re-222 and Re-224 showed crystalline frameworks, while the Re(CO) 3 moiety was no longer visible in the corresponding ATR-IR spectrum ( Fig. S15 and S16 †). This suggests that while the host is virtually stable under catalysis conditions, the catalyst is not, in line with its homogeneous instability. 9,43,47 To examine the reasons for the differing product evolution rates in Re-222 and Re-224 samples, as well as the dramatic change of product selectivity between homogeneous and heterogeneous systems, the photocatalytic mechanism was investigated through spectroscopic and catalytic experiments.

Catalytic mechanism investigation and photocatalysis under selective irradiation
The topology of PCN-222 and -224 differs signicantly. While 222 has hexagonal channels of $35Å in size and trigonal channels of $10Å, 224 has cubic pores/impeded channels with $20Å diameter ( Fig. 1c and S10 †). 8,13 The geometric structure of the catalyst optimised by DFT calculations yielded a maximum van-der-Waals sphere diameter of $13.5Å, smaller than the MOFs' pore apertures (Fig. 1b, atomic coordinates in ESI †). Despite similar molecular loadings of 1, partial pore blocking is likely more prominent in Re-224 than Re-222 as shown by the former's lower N 2 uptake ( Fig. 2c and Table S1 †). DED analysis was thus performed to map the electron density introduced by 1. Briey, measured PXRD patterns were Le Bail-rened and the resulting structure factors were used to create structure envelopes for pristine and loaded MOFs, which were subtracted from each other (Fig. S17-S20, details in ESI †). 48 For Re-222, 1 was found directed into the hexagonal channels leaving inner space for substrate diffusion (Fig. 4). In contrast, for Re-224, 1 lls almost the complete space of every second pore. These observations are supported by DFT calculations of 1 in a pore fragment of each MOF, conrming the catalyst positions suggested by DED analysis (Fig. 4a, b, d and e; details and atomic coordinates in ESI †). Specically, for Re-224, this implies pore blocking potentially further rationalised by DFT calculations that also revealed the possibility of anchoring 1 via two coordinating N-Zn bonds to neighbouring Zn-porphyrins in Re-224. Together with ATR-IR spectra (shi of Zn-N bands aer 1 loading), steric pore restrictions (particularly for 224-Zn), and comparable catalyst loadings per linker for Re-222/Re-224, the DED analysis supports that anchoring of 1 occurs predominantly at the TCPP-bound Zn (Fig. 4, S5 and Table S1 †). While not directly observed, partial coordination to physisorbed Zn cannot be ruled out.
Overall, these analyses depict a more hindered nanoenvironment in Re-224 than in Re-222 potentially limiting reactant (e.g., BIH max. diameter ¼ 10.6Å), 21 and product diffusion in line with a lower catalytic activity for the former.
Besides mass transport, transferring conclusions from recent reports on MOF-based exciton migration suggest that charge separation (CS) probability is signicantly favoured for Re-222 over Re-224, due to topology-induced higher exciton hopping rates and a lower Förster radius R between chromophores (Förster resonance energy transfer scaling with R −6 ). 20,49 Here, the translational hopping distance along the c-axis is $17 A for 222, and $39Å for 224 ( Fig. 4c and f and non-translational distances in Fig. S21 †). Additionally, 222's triangular motif with cofacial conformation translating along the lateral direction of the crystallographic c-axis further improves excited state delocalisation and hopping rates, 50 due to head-to-tail coupling. 51 Steady-state luminescence measurements were conducted towards further understanding of photoinduced processes in our systems. Luminescence experiments on a 0.05 mM solution of 2-Zn in MeCN showed two prominent emission bands at l ¼ 649 and 714 nm upon excitation at either 415 or 515 nm (Soret or Q band, respectively) ( Fig. 5a and b), matching literature reports ascribed to the de-excitation of the S 1 states. 52 The latter is produced by: internal conversion of the S 2 state, via Soret region irradiation, or directly, via Q-band irradiation. Adding an equimolar amount of 1 to the solution to replicate homogeneous catalysis system resulted in a quenching of the luminescence intensity, comparable for both excitation wavelengths, and ascribed to photoinduced intramolecular electron transfers from 2-Zn to 1. Incomplete quenching possibly stems from dynamic self-assembly and unpaired 2-Zn as well as inefficient electron transfer. With uorescence experiments on colloid MOFs being challenging due to high scattering and sedimentation propensity, fully soluble Zr-oxo-based assemblies were prepared as small-scale counterparts of the hybrid PCN systems to enable properties replication and photophysical investigations. Discrete Zr 6 -oxo-nodes (Zr 6 ), featuring methacrylate ligands as capping agents to prevent aggregation, were synthesised and isolated following a literature procedure (see ESI †). 53 Associating these with 2-Zn yielded the fully soluble CO 2 H-tethered composites Zr 6 j2-Zn (see ESI † for assembly conditions) and allowed solution uorescence measurements without affecting 2-Zn's absorption properties (Fig. S22 †). In stark contrast to the Zr 6 -free homogeneous systems, a lower luminescence intensity was recorded aer excitation at 415 nm (Fig. 5a) whereas excitation at 515 nm afforded a spectrum comparable to the one of 2-Zn (Fig. 5b). These behaviours were also reproduced in the corresponding action spectra (l em ¼ 720 nm) as Zr 6 j2-Zn displayed unchanged Q-band contributions compared to 2-Zn, and adding 1 resulted in lesser contributions (Fig. S23 †).
Such wavelength-dependent luminescence behaviour hints at different energy relaxation pathways. Here, excitation into the Soret band in the Zr 6 j2-Zn assembly results in a high-energy S 2 state (zero-zero transition energy, E 00 z 2.9 eV) and likely induces fast electron transfer to a Zr-oxo cluster. 18 As no uorescence quenching was observed upon Q-band irradiation, this pathway is disabled from the S 1 state possibly due to its lower energy level (E 00 z 2.0 eV). This also implies that the kinetics of charge separation from S 2 are signicantly faster than the S 2 / S 1 internal conversion. While Zn porphyrins present various and complex relaxation mechanisms, this interpretation is coherent with the documented slow S 2 / S 1 kinetics in Zn porphyrins, and in line with hot electron transfers from S 2 as previously observed in molecular dyads and in heterogeneous assemblies. 25,54,55 As inferred by the wavelength-independent quenching of uorescence in the 2-Zn with 1 system (Fig. 5), reduction of 1 from the S 1 excited state of 2-Zn is anticipated and consistent with the signicant associated exergonicity (change in Gibbs free energy value, DG CS ¼ −0.6 eV).
To investigate whether this behaviour is reected in the photocatalysis performance, we performed a series of wavelength-dependent investigations using lower energycentred wavelengths l ¼ 490-740 nm ($8 mW cm −2 ), thus omitting Soret band excitation and corresponding S 2 excited state production.
For Re-222-based experiments, a dramatically increased reactivity, and a reversed selectivity from HCOO − to CO were observed with $370 mmol CO g −1 h −1 and <1 mmol HCOO − g −1 h −1 (Fig. 6a and Table S3 †). Correspondingly, TON CO $100 were reached aer 80 h compared to below 1 obtained with l > 430 nm and a higher absolute light intensity (Fig. 3a, 6b, Tables S3 and S4 †). Re-224-based experiments gave a similar trend with $24 mmol CO g −1 h −1 (TONs CO $ 12) and <1 mmol HCOO − g −1 h −1 aer $60 h, albeit with the absolute performance remaining signicantly inferior to Re-222 due to its topological limitations (Fig. 6a, b and Table S3 †). With selective irradiation predominantly yielding formate or CO (Table S3 †), electronic communication between nodes and Re catalysts within the pores is unlikely. Aer 80 h of >490 nm irradiation PXRD analysis of Re-222 and Re-224 showed crystalline framework retention, while the Re(CO) 3 moiety was no longer visible in ATR-IR spectra, comparable to >430 nm irradiation ( Fig. S15 and S16 †). For Re-222 and Re-224, post-catalysis ICP-MS measurements revealed that 27 and 17% of immobilised Re leached into the supernatant, respectively (Table S7 †). This is consistent with other reported photodegradations of Re catalysts and described complex decomposition and subsequent loss of activity. 9,43,47,56 In addition, corresponding dynamic light scattering measurements revealed no Re nanoparticle formation (Fig. S24 †).
Control experiments with 13 C-labelled CO 2 for both Re-PCN hybrids produced only 13 CO, conrming that CO 2 was the sole source of CO (Fig. S25 †). Further control experiments conducted with the CO 2 H-analogue of 1, i.e., fac-ReBr(CO) 3 (4,4 ′dicarboxy-2,2 ′ -bipyridine), node-anchored in PCN-222 and PCN-224 (synthesis in ESI, Fig. S26-S28 and Table S1 †) showed comparable wavelength-dependent performance trends to Re-222 and Re-224 under identical reaction conditions (Table S3 †). This is consistent with a previously reported PCN-222-Re assembly displaying selective CO 2 -to-CO evolution upon >500 nm irradiation 19 and thus excludes the catalyst and dyadic motif as the source of the observed trends.
These wavelength-dependent results suggest that excited state (S 2 vs. S 1 )-specic, selectivity-dictating catalytic channels are occurring in the MOF assemblies, with electron delivery to the CO evolution molecular catalyst or the formate producing node being selectively activated by low or high energy wavelength, respectively. This was conrmed by AQY measurements at l ¼ 520 nm, giving $0.35% (CO) and $0.01% (formate), showing a reverse trend compared to 450 nm values (Tables S5  and S6 †).
Interestingly, the correlations between product selectivity and irradiation wavelength were found in both Re-PCN assemblies as well as in the homogeneous Zr 6 j2-Znj1 system. The latter showed limited CO evolution (TONs $ 1) and traces of formate using full irradiation, while >490 nm irradiation experiments yielded TONs $10 (Table S2 †). By contrast, homogeneous Zr 6 -free molecular 2-Zn with 1 delivered wavelength-independent performance. This further highlights the critical impact of the assembly/framework (MOF vs. molecular assemblies) on photocatalysis (Fig. 3a, b, 6a, b, Tables S2 and S3 †).
Luminescence and catalytic experiments together provide consistent evidence that irradiation with l > 430 nm during photocatalysis mainly results in excitation of the Soret-band (S 0 -S 2 transition, Fig. S6-S8 †) in PCN assemblies upon which the S 2 state is rapidly quenched by electron transfer to the Zr 6 -oxo-node (Fig. 6c). 17,18,39 Here, bimolecular reaction with the diffusing SED is likely to proceed subsequently to regenerate the ground state (Fig. S13 †) or detrimental charge recombination occurs. 19 Selective CO 2 -to-formate reduction proceeds via reduced Zr(III) centres upon collecting another electron. 17,46 With each linker connected to four nodes and charge separation occurring at the light harvesting centre, this constitutes a localised electron channel (Fig. 6d). Shiing to irradiation with l > 490 nm mainly results in excitation of the Q bands (S 0 -S 1 transition), 55 affording a lower energy excited state unable to trigger node-quenching but able to reduce 1. The latter effects selective CO 2 -to-CO reduction upon accumulating two electrons. 9,19 Alternatively, in absence of a Re catalyst in direct vicinity, directional exciton migration proceeds within the MOF structure. 19 As an exciton can visit 100+ linkers in its lifetime, 19,51 this energy funnelling is highly delocalised and provides a long-range, antennae-like catalytic channel (Fig. 6d).
The minute, deactivated CO evolution when using full irradiation is noteworthy considering that Soret and Q bands are concomitantly excited. The rationale is likely intricated and may involve substrate diffusion, competing photophysical processes and slower CO catalysis kinetics. Tentatively, the high molar absorption of S 2 together with an excess of nodes to catalysts (statistically each linker neighbours four Zr 6 nodes and $0.1 Re catalysts, Table S1 †) could result in a kinetic preference for available CO 2 (with affinity to coordinate to Lewis-acidic Zr nodes) to be converted to formate, depleting local substrate concentrations. 57 While future in-depth spectroscopic studies are needed to elucidate this phenomenon, this nding may hold important repercussions as most literature research focuses on full spectrum broadband irradiation.
Overall, the results show that the linker's excited states and quenching kinetics with the molecular catalyst vs. the node are critical for understanding and tailoring reactivity. As a perspective, the observed wavelength-dependency provides an edge over similar systems as these usually show a reverse trend, i.e., high-energy (near) UV irradiation achieves higher performance. 45,58 Here, lower-energy visible light provides selective molecular CO 2 -to-CO reduction, allowing unusual control over product selectivity based on irradiation wavelength.

Experimental
For a comprehensive description of analytical methods and experimental procedures the reader is referred to the ESI. † An overview of the central syntheses is provided here.
This intermediate (214 mg, 235 mmol, 1.0 eq.) was dissolved in THF/MeOH 1 : 1 (15 mL). KOH (691 mg, 12.3 mmol, 52 eq.) was dissolved in H 2 O (6 mL) and added to the prior solution. The reaction mixture was reuxed for 5 h under continuous stirring. Aer the mixture was cooled to room temperature, the organic solvents were removed by rotary evaporation. The crude product was dissolved in H 2 O (32 mL) and heated at 90 C for 10 min. Aer cooling to room temperature, the aqueous mixture was acidied with 1 M HCl solution (21 mL) and 37% HCl (1 mL) in respective order. The resulting dark green precipitate was isolated by centrifugation and washed with H 2 O (8 Â 30 mL). The dark green product was dried in vacuo overnight (137 mg, 161 mmol, 68% yield, see ESI †).
PCN-222. The synthesis was adapted from a literature known procedure. 14,33 In a 20 mL screw cap vial, 2 (12.5 mg, 0.016 mmol, 1.00 eq.) and ZrOCl 2 $8H 2 O (23.5 mg, 0.073 mmol, 4.56 eq.) were dissolved in N,N-diethylformamide (DEF, 3 mL). Aer addition of 4-tert-butylbenzoic acid (1350 mg, 7.57 mmol, 473 eq.), the mixture was ultrasonicated for 10 min. The mixture was heated at 120 C for 12 h in an oven. The solid was separated by centrifugation and soaked with DMF (3 Â 6 mL) and dried in vacuo. The solid (12.2 mg) was dispersed in DMF (8.1 mL) and acidied with 8 M HCl (0.3 mL). The mixture was heated in an oven at 120 C for 12 h. The purple solid was collected by centrifugation and soaked in DMF (3 Â 6 mL) and acetone (3 Â 6 mL) and dried in vacuo to yield the product, whose characterisations matched literature reports (see ESI †).
Metalation of 222 or 224 with Zn. The synthesis was adapted from a literature known procedure. 34,35 ZnCl 2 (59.9 mg for 222 and 22.9 mg for 224) was dissolved in DMF (3 mL). Aer addition of the MOF powder (30 mg), the suspension was heated in an oven at 100 C for 24 C. The product was isolated by centrifugation and washed with DMF (3 Â 6 mL) and acetone (3 Â 6 mL). Drying in vacuo yielded the product, whose characterisations matched literature reports (see ESI †).
Molecular catalyst loading in 222-Zn or 224-Zn. A 0.1 mM solution of 1 in MeCN (15 mL) was added to powder samples of the respective activated MOF sample (10.0 mg) in VWR® TraceClean® 20 mL vials. Aer 24 h in the dark, the suspension was centrifuged, and the supernatant removed. The resulting powder was washed with fresh MeCN (3 Â 7 mL) for 2 h per wash cycle and then dried overnight in vacuo.
Photocatalytic CO 2 reduction. Photocatalytic tests were performed in air-tight 50 mL Schlenk asks with the respective MOF material (1.5 mg), a stir bar, MeCN (4 mL), deionised H 2 O (0.12 mL, 3 vol%) and BIH (225 mg, 1 mmol). For CO 2 reduction the reaction suspension and headspace were fully saturated with CO 2 for seven minutes and aer sealing a CO 2 overpressure was applied, totalling a pressure of 1.45 bar.
The irradiation source was a heat-free white light generator Asahi Spectra MAX-303 Compact Xenon Light Source 300 W, with either a XVL0430-Longpass 430 nm lter (430-740 nm irradiation, $10 mW cm −2 ) or a XVL0490-Longpass 490 nm lter (490-740 nm irradiation, $8 mW cm −2 ). Reaction products were determined through headspace analysis by gas chromatography and 1 H NMR of the reaction solution.

Conclusions
We successfully synthesised and characterised two different hybrids of fac-ReBr(CO) 3 (qtpy) in porphyrinic MOFs (PCN-222 and PCN-224) by post-synthetic self-assembly, allowing straightforward topology-independent catalyst anchoring. Corresponding smaller scale, molecular and supramolecular assemblies were also prepared for comparison and benchmarking purposes. Here, we demonstrated that material interfacing at various scales, i.e., molecule-molecule (1 + 2-Zn), molecule-cluster (Zr 6 j2-Znj1), and molecule-particle (Re-PCNs) produces composites displaying unique and complex behaviours that are concomitantly reminiscent, diverging and unseen in their individual building blocks.
Mainly, assembling Re catalyst-PCN composites afforded photocatalysts bearing two catalytic channels that are selectively activated by specic incident irradiation wavelengths and thus, ultimately govern the catalysis product selectivity. Shorter wavelengths resulted in localised charge transfers from the excited porphyrin linkers to the Zr 6 -oxo-nodes, producing formate. Alternatively, lower energy irradiation can be used to activate the CO-selective Re catalysts with exciton migration enabling long-range, delocalised catalysis. By contrast, the corresponding molecular and supramolecular systems yielded product selective and wavelength-independent catalysis.
While molecule-like localised S 2 and S 1 states more accurately describe the observed photoinduced processes in Re-PCNs over typical band-like description of MOFs, the hybrids also display material-like properties elusive to molecular species. Thus, a topology effect was shown with the PCN-222 hybrids displaying a higher photocatalytic activity over PCN-224, as small pore sizes hinder mass diffusion and short coupling distances promote antennae-effect. Further the nanoparticular hybrid porous systems enabled a 10-fold increase in CO 2 reduction activity compared to their discrete counterparts, ascribed to stabilising pore connement.
Although preliminary, this work constitutes the seminal investigation of the likely considerable inuence of MOF topology on solar fuel production. Future efforts on controlled distances between catalyst and photosensitiser, understanding of underlying photoinduced processes, varied anchor principles, as well as optimised energy and mass transport will allow orthogonal component assembly and minimise energy loss for efficient solar fuel production. Finally, we show that while broadband light is oen considered in its entirety in the broader literature, a ner differentiation may be needed to understand complex hybrid materials' behaviour and performance.

Data availability
The source data is available from the corresponding author upon reasonable request.

Conflicts of interest
There are no conicts to declare.
Catalysis Research Center, especially from Jürgen Kudermann, is gratefully acknowledged. Additional thanks to Silva Kronawitter and Johanna Haimerl for proof reading. This work was supported by the German Research Foundation (DFG) Priority Program 1928 'Coordination Networks: Building Blocks for Functional Systems', the research project MOFMOX (grant number: FI 502/43-1), and by the Excellence Cluster 2089 'econversion' (Fundamentals of Energy Conversion Processes).