Manipulating the Coordination Structure of Molecular Cobalt Sites in Periodic Mesoporous Organosilica for CO2 Photoreduction

Photocatalytic CO2 reduction, including reaction rate, product selectivity, and longevity, is highly sensitive to the coordination structure of the catalytic active sites, and the precise design of the active site remains a challenge in heterogeneous catalysts. Herein, we report on the modulation of the coordination structure of MNx-type active sites (M = Co or Ni; x = 4 or 5) anchored on a periodic mesoporous organosilica (PMO) support to improve photocatalytic CO2 reduction. The PMO was functionalized with pendant 3,6-di(2′-pyridyl)pyridazine (dppz) groups to allow immobilization of molecular Co and Ni complexes with polypyridine ligands. A comparative analysis of CO2 photoreduction in the presence of an organic photosensitizer (4CzIPN, 1,2,3,5-tetrakis(carbazol-9-yl)-4,6-dicyanobenzene) and a conventional [Ru(bpy)3]Cl2 sensitizer revealed strong influence of the coordination environment on the catalytic performance. CoN5-PMO demonstrated a superior CO2 photoreduction activity than the other materials and displayed a cobalt-based turnover number (TONCO) of 92 for CO evolution at ∼75% selectivity after 3 h irradiation in the presence of 4CzIPN. The hybrid CoN5-PMO catalyst exhibited better activity than its homogeneous [CoN5] counterpart, indicating that the heterogenization promotes the formation of isolated active sites with improved longevity and faster catalytic rate.


■ INTRODUCTION
The photocatalytic valorization of CO 2 , in particular through its transformation into CO or syngas, represents an appealing CO 2 mitigation strategy that has the potential to significantly contribute toward net-zero carbon economy.The use of homogeneous or semiheterogeneous colloidal photosystems, combining a photosensitizer with a CO 2 reduction cocatalyst, has emerged as a simple and effective approach in solar-driven CO 2 conversion technology.However, these hybrid photosystems often have several shortcomings such as expensive noble-metal-based sensitizers and cocatalysts, low photostability of sensitizers and/or cocatalysts, poor recyclability, and limited visible light absorption, restricting their implementation in practical systems.From a sustainability perspective, it is important to design photosystems that use organic dyes or non-noble metal-based sensitizers for harvesting the solar energy.However, in comparison to the Ru-and Ir-based molecular sensitizers, which are ubiquitous in photocatalytic CO 2 reduction literature, organic dyes are relatively underutilized as photosensitizers.A few recent reports on precious-metal-free photosystems have demonstrated the viability of organic dyes, such as purpurin, 1,2 phenoxazine, 3 anthraquinone derivatives, 4 and thermally activated delayed fluorescence (TADF) compounds, 5−7 but generally, the scope of the catalysts that can be combined with the organic dyes is quite narrow.Therefore, it is important to expand the library of precious-metal-free photosystems by developing new catalyst/dye combinations for CO 2 photoreduction.
−11 While molecular catalysts offer distinct benefits such as better atom efficiency, high product selectivity, tunability, and scope for mechanistic understanding, they often suffer from poor stability and are nonrecyclable when used under homogeneous conditions.However, these shortcomings can be mitigated via heterogenization of the molecular complexes on solid supports, 12−15 which bridges between homogeneous and heterogeneous systems and provides an opportunity to precisely regulate the structure of the active site on solid support at atomic levels.The primary coordination environment around the central metal-active site plays a crucial role in controlling the catalytic reaction.−19 MN x -type active sites are frequently reported in single-atom catalysts on various support materials including carbon nitride, graphene, and metal oxides, 18−24 but fine-tuning the coordination structure of the integrated MN x sites is often challenging.In this context, we report a strategy for controlling the molecular structure of metal-active sites grafted on functionalized periodic mesoporous organosilica (PMO) support to optimize CO 2 photoreduction.Notably, the porous support plays an important role by allowing diffusion of the reactants and reagents to access the molecular active sites and by regulating the local chemical environment during catalysis. 25MOs, synthesized from organo-bridged alkoxysilane precursors by a directing surfactant self-assembly approach, are one of the most representative organic−inorganic hybrid materials with the organic functionalities homogeneously distributed within the pores and the walls of the silica framework. 26Due to their unique characteristics such as ordered mesostructures with adjustable pore size, high surface areas, and tunable hydrophobicity/hydrophilicity, PMOs have been demonstrated to be promising scaffolds to construct heterogeneous photocatalytic systems for CO 2 reduction. 27,28−31 The surface bipyridine groups serve as chelating ligands for the formation of metal complexes on the surface pores, thus resulting in the formation of isolated active sites.−31 Recently, catalytic performance was improved by using Bpy-PMO with a tubular structure that facilitated the diffusion of the reactants through the large pores, leading to faster photocatalysis and enhanced quantum yields. 32As an alternative to the Bpy-PMO platform to immobilize metal complexes, recently our group reported the synthesis of a novel dipyridyl-pyridazine triethoxysilane precursor (Ndppz) through a facile and rapid approach based on an inverse electron demand Diels−Alder (iEDDA) reaction. 33By co-condensation reactions of the Ndppz precursor with 1,2-bis(triethoxysilyl)ethane (BTEE), we achieved a novel PMO material with surface-attached pendant N-chelating heterocyclic ligands, 3,6-di(2′-pyridyl)pyridazine (dppz).Ndppz-PMO was successfully utilized as a solid chelating ligand to immobilize Ru and Ir complexes and generate light-harvesting materials for the photocatalytic hydrogen evolution reaction.Very recently, we have also reported the synthesis of a dppz-functionalized mesoporous SBA material and its use as a heterogeneous water oxidation catalyst after complexation with IrCp*Cl. 34These results demonstrate the great potential of heterogenized N-chelating heterocyclic ligands as a novel platform for the construction of dppz-based solid MN x catalysts.
Systemic analysis to rationally design the optimum coordination environment of MN x -type active sites toward CO 2 photoreduction in heterogeneous colloidal suspensions has rarely been reported. 35Herein, the dppz-functionalized PMO (NdppzPMO) has been used as a versatile scaffold for installing MN x -type units on the surface and investigate the effect of changing the primary coordination sphere on photocatalytic CO 2 reduction.Four single-site MN x -PMO catalysts (M = Co or Ni, x = 4, 5, x representing the number of pyridine N atoms coordinated with the metal center) were synthesized by anchoring [M(tpy)] 2+ or [M(bpy)] 2+ moieties (tpy = 2,2′:6′,2″-terpyridine and bpy = 2,2′-bipyridine).In CO 2 photoreduction studies with a traditional Ru(bpy) 3 Cl 2 photosensitizer (denoted as [Ru-PS]) and an organic TADF dye (4CzIPN), MN x -PMO catalysts displayed distinct CO evolution activity with CoN 5 -PMO exhibiting superior performance than the other three materials.−7,37−40 The coordination environment around Co in CoN 5 -PMO was confirmed by X-ray absorption spectroscopy (XAS), and the origin of the activity at the CoN 5 -active site was supported by mechanistic studies on a molecular CoN 5 analogue.

■ RESULTS AND DISCUSSION
Synthesis and Characterization.The PMO with pendant 3,6-di(2′-pyridyl)pyridazine (dppz) binding motifs was synthesized by co-condensation of the Ndppz trialkoxysilane-functionalized precursor (denoted as Ndppz, 30 mol %) and a conventional bis-silane precursor (BTEE, 70 mol %), in the presence of a cationic surfactant (Figure 1a). 33uccessful synthesis of Ndppz-functionalized PMO (NdppzPMO) was confirmed by powder X-ray diffraction and Raman spectroscopy (Figure 2).Molecular loading of 0.56 mmol dppz g −1 was determined from elemental analysis (3.15 wt % of N content).The assembly of different MN x -isolated single sites on NdppzPMO was accomplished by the coordination of the [M(tpy)] and [M(bpy)] units (tpy = 2,2':6′,2″-terpyridine and bpy = 2,2′-bipyridine) onto the surface N-chelating motifs (Figure 1b).The hybrid PMOs are defined as MN x -PMO.The postsynthetic metalation of NdppzPMO with the molecular precursors (M(tpy)Cl 2 and M(bpy)Cl 2 ) promoted a change of color of the solid from pale pink to light brown, suggesting successful formation of MN x -PMO.The metal loading in the materials was determined by inductively coupled plasma-optical emission spectroscopy (ICP-OES) (Table S1).As a control material, CoN 2 -PMO was synthesized by metalation of NdppzPMO with CoCl 2 , producing cobalt centers coordinated to the dppz units in PMO.
Preliminary screening of the four MN x -PMOs as photocatalysts for CO 2 reduction in the presence of [Ru-PS] or 4CzIPN as a photosensitizer and triethanolamine (TEOA) as a sacrificial electron donor demonstrated superior performance of CoN 5 -PMO (vide infra).Subsequently, the in-depth characterization of the hybrid material was focused on CoN 5 -PMO.ICP-OES analysis of CoN 5 -PMO showed a cobalt loading of 0.041 μmol Co mg −1 , which was consistent with the 0.035 μmol tpy mg −1 estimated by 1 H NMR of aciddigested CoN 5 -PMO (Figure S1).This data indicates that only ∼7% of the available dppz ligands were coordinated with Co(tpy) units, suggesting that the remaining uncoordinated dppz ligands have lower accessibility.The preservation of the periodic mesoporous structure after the formation of surface CoN 5 sites was confirmed by lowangle X-ray diffraction, transmission electron microscopy, and nitrogen adsorption−desorption isotherms.Powder X-ray diffraction pattern of CoN 5 -PMO showed characteristic lattice planes of a P6 mm hexagonal arrangement structure with a main reflection at 2θ value of 1.90°(an interplanar spacing (d 100 ) of 4.6 nm) analogous to the pristine material (Figure 2a). 41However, differences in the scattering contrasts within the pores were observed due to the decrease in the intensity of the diffractogram obtained after Co(tpy) coordination. 34XRD results were supported by TEM images, which showed a highly ordered structure with the pore channel axis oriented parallel to the electron beam (Figure 2b).The porosity of CoN 5 -PMO was evaluated by nitrogen physisorption measurements, obtaining a nitrogen adsorption−desorption isotherm with capillary condensation in the P/P 0 range from 0.3 to 0.7 (Figure 2c).The isotherm exhibited largely type IV behavior and a hysteresis loop of type H2, typical of ordered mesoporous materials.In addition, the pore size distribution (PSD) also indicated the presence of microporosity (Figure 2c�inset). 42Successful immobilization of the Co complex within PMO cavities was confirmed by a decrease in Brunauer−Emmett−Teller surface area (S BET ) from 901 to 750 m 2 g −1 and pore volume (V P ) from 0.72 to 0.60 cm 3 g −1 .However, the pore diameter (D P ) remained unchanged, with an average value of 3.2 nm.
Surface functionalization of NdppzPMO with CoN 5 sites was corroborated by UV−vis diffuse reflectance spectroscopy (UV−vis DRS) and Raman spectroscopy, while local electronic and atomic structures around the cobalt center were assessed by XAS and X-ray photoelectron spectroscopy (XPS).As shown in Figure 2d, the UV−vis DRS of CoN 5 -PMO exhibited a characteristic π−π* band at 290 nm originating from the dppz and the terpyridine ligands 43,44 and an n−π* band at 540 nm for the pyridazine units.In addition, coordination of the metal complex led to the appearance of a broad shoulder at around 450 nm ascribed to metal-to-ligand charge transfer (MLCT).This finding was corroborated with similar absorptions observed in the UV−vis spectra of the homogeneous analogue (Figure 2d) and Co(tpy)-based complexes. 45Immobilization of Co(tpy) in CoN 5 -PMO was further confirmed by Raman spectroscopy, which showed characteristic Si−O stretching vibration at 995 cm −1 , intense C�N vibrations at 1590 cm −1 , and additional signals at 1549, 1300, and 1245 cm −1 originating from various skeletal vibrations of pyridine/pyridazine heterocycles (Figure 2e). 33ew signals were observed at 1601, 1471, 1330, and 1026 cm −1 that can be attributed to the anchored Co(tpy) moiety. 46,47Furthermore, the absence of symmetric ν O−O stretching vibration for μ-O 2 species at 821 cm −1 confirms the formation of monomeric CoN 5 species.Thermogravimetric analysis (TGA) of CoN 5 -PMO showed two clear weight loss steps: (1) ∼ 8% loss below 100 °C from evaporation of solvent(s) and ( 2) ∼ 7% loss at 390−450 °C likely due to thermal degradation of the organic content (Figure S2).The weight loss profile of CoN 5 -PMO was similar to that for NdppzPMO, consistent with the unchanged underlying structure of the PMO support.
XPS measurements were performed to probe the surface composition of PMO and the electronic structure of the CoN 5 sites (Figures 2f and S3).The survey spectrum of CoN 5 -PMO showed the presence of Si, O, C, N, and Co (Figure S3a).In the Co 2p region, two peaks were observed at 780.9 eV (Co 2p 3/2 ) and 796.8 eV (Co 2p 1/2 ) (Figure 2f), which was in good agreement to the peaks observed for the reference molecular dimeric Co(III) complex, [{Co(tpy)(bpy)} 2 (μ-O 2 )](PF 6 ) 4 (abbreviated as CoN 5 bpy dimer), demonstrating the +3 oxidation state of Co.The weak shakeup satellite peaks for Co are also consistent with a predominantly diamagnetic Co 3+ species.The N 1s region of the XPS for CoN 5 bpy dimer features a peak at 399.6 eV, which is similar to that observed for CoN 5 -PMO and previously reported systems based on NdppzPMO. 33Deconvolution of the N 1s peak showed two signals at 399.9 and 399.1 eV (Figure S3d), which can be attributed to Co−N bonds and noncoordinated pyridine/ pyridazine N, respectively.To further resolve the local coordination environment around the cobalt center, steadystate XAS was performed at the Diamond Light Source (Figures 3, S4 and S5).The X-ray near-edge spectroscopy (XANES) spectra for CoN 5 -PMO at Co K-edge showed a very similar position and shape of the absorption edge to that for the CoN 5 bpy dimer, 48 indicating that the cobalt centers in PMO are in an octahedral geometry with five pyridinic ligands and a coordinated solvent molecule.The first derivative of XANES spectra showed excellent agreement between the edge energy (E 0 ) of CoN 5 -PMO and CoN 5 bpy dimer, while the two Co(II) reference compounds displayed edge position at lower energy, which is consistent with a +3 oxidation state of the Co centers in PMO (Figure S4b).Fourier transform-extended X-ray absorption fine structure (FT-EXAFS) spectrum of CoN 5 -PMO in the R-space showed a dominant peak at ∼1.34 Å and two weaker peaks at 2.30 and 3.31 Å (Figure 3b).The first peak at a shorter radial distance originates from the singlescattering Co−N/O pathways in the first coordination shell.The scattering paths visible at higher R values (R = 2.30 and 3.31 Å) originate from the second and third coordination shell of tpy and dppz ligands and include Co−C/N single-scattering and Co−N−C/Co−C−C multiple-scattering paths (Figure 3b,3c).The close similarity between the R-space EXAFS data for CoN 5 -PMO and CoN 5 bpy dimer suggests an analogous coordination environment around the Co center in PMO with five N donors from tpy and dppz ligands and one O/N donor from a solvent molecule (H 2 O or MeCN).In comparison, the major peak for the first coordination shell of the two Co(II) reference compounds (Co(bpy)Cl 2 and CoCl 2 ) appeared at longer radial distances, consistent with the +2 oxidation state of Co.To gain further details of the molecular structure of the Co site, EXAFS fittings were performed using the FEFF input model built from the crystal structure of the CoN 5 bpy dimer, by changing the bipyridine ligand to dppz.The best fit in the Rspace and k-space for the first and second coordination shells is shown in Figures 3c and S5, respectively.The fitting parameters are listed in Table S2, and the structural model is shown in Figure S6.The best fit indicates that the first coordination shell consists of four Co−N bonds with a 1.97 Å bond distance, another Co−N bond at 1.80 Å, and a coordinated water molecule (or a hydroxide ligand) with a 1.82 Å long Co−O bond.The shorter Co−N distance corresponds to the bond between Co and the central nitrogen atom of the tpy ligand.The Co−N/O bond distances are consistent with a Co(III) center and are within the range of distances reported in the literature, 48,49 validating our fitting model.The second EXAFS peak at 2.30 Å was fitted with a second shell of single-scattering paths originating from the tpy and dppz ligands, which consists of six Co−C single-scattering paths at 2.86 Å, a Co−N path at 2.99 Å, and another Co−C path at 3.08 Å. Importantly, the fit suggests the absence of Co−Co scattering paths, confirming the molecular nature of single Co(III) sites in the hybrid PMO material (Figure 3c).Using similar fitting procedures, the coordination structure of the Co sites in CoN 4 -PMO was also evaluated based on the EXAFS fit with a structural model of Co(bpy)(dppz)(OH 2 ) 2 (Figures S4 and S5), and the results are summarized in Table S3.
To corroborate the CO 2 reduction capability of the CoN 5active sites on PMO, the molecular analogue, CoN 5 bpy dimer, was probed in solution by using cyclic voltammetry (Figure S7).All potentials are reported against the ferrocene couple (Fc +/0 ).In N 2 -purged acetonitrile solution, the dimeric Co complex dissociates to form mononuclear [Co(tpy)(bpy)] 2+ species (denoted as [CoN 5 bpy ] 2+ ), presenting close structural model of the active site in CoN 5 -PMO. 48Cyclic voltammogram of [CoN 5 bpy ] 2+ under N 2 showed two reversible reductions at E 1/2 = −0.12 and −1.17 V, attributable to Co III/II and Co II/I processes. 50At further reducing potentials, two overlapping reductions were observed with an irreversible process at −1.95 V (peak potential) and a quasi-reversible process at −2.05 V (E 1/2 ), which can be assigned to ligandcentered reductions and/or Co I/0 reduction.Under CO 2 saturation, current enhancement was observed near the third reduction process (onset potential −1.83 V and midwave potential −1.95 V), indicative of electrocatalytic CO 2 reduction.In the presence of 10% (v/v) triethanolamine  (TEOA) as a proton source, the onset of catalysis undergoes an anodic shift to −1.70 V, indicating that TEOA facilitates CO 2 at the Co-active sites.To further investigate the impact of replacing bpy with dppz ligand on CO 2 reduction, a second molecular cobalt complex was synthesized bearing tpy and dppz ligands, [Co(tpy)(dppz)Cl](PF 6 ) 2 (denoted as CoN 5 dppz ). Cyclic voltammograms of CoN 5 dppz recorded in the CO 2 -saturated MeCN/TEOA (10% v/v) electrolyte showed an earlier onset of catalysis at −1.51 V, while the reduction potential for the two Co-centered process remained unchanged at −0.11 V (Co III/II ) and −1.17 V (Co II/I ) (Figures S8 and S9).This suggests that dppz ligand-centered reduction processes are involved in the catalytic reduction of CO 2 , and more electron-deficient nature of dppz enables an earlier onset of catalysis compared to the CoN 5 bpy analogue.Photocatalytic CO 2 Reduction.Since polypyridyl complexes of Co and Ni are well-known CO 2 reduction catalysts, 51−56 the four MN x -PMOs with distinct isolated single sites (CoN 4 -PMO, CoN 5 -PMO, NiN 4 -PMO, and NiN 5 -PMO) were initially screened for photocatalytic CO 2 reduction in the presence of [Ru-PS] and TEOA donor.In a typical experiment 1 mg of MN x -PMO was dispersed in a CH 3 CN/ TEOA (9:1) mixture containing the photosensitizer, purged with CO 2 to saturate the colloidal suspension, and irradiated under UV-filtered simulated solar light (λ > 400 nm, 100 mW cm −2 , AM 1.5G).Gas chromatography of the headspace showed that all the four materials produced CO and H 2 (Table 1, entries 1−4).No liquid product was observed by 1 H NMR spectroscopy or ion chromatography.Among the four catalysts, CoN 5 -PMO exhibited the best activity with a CO evolution of 2.13 μmol mg −1 , a CO selectivity of 64.2%, and a corresponding turnover number (TON CO ) of 52 after 1 h based on cobalt loading.Enhanced CO evolution and better CO selectivity were observed for CoN x -PMOs when BIH was introduced in the system as a two-electron donor, consistent with previous reports (Table 1, entries 5−8). 57,58In contrast, NiN x -PMOs displayed lower CO evolution in the presence of BIH.It should be noted that in the presence of BIH donor, TEOA mainly functions as a proton acceptor from BIH +• .TEOA is also known to form a zwitterionic alkylcarbonate adduct with CO 2 , which is a good proton donor. 59The preliminary results with PMO materials showed promising activity toward CO 2 photoreduction, but the use of a noblemetal-based photosensitizer was a drawback, and metal-free organic photosensitizers were explored as potential alternatives for photocatalysis.
Three different organic dyes (9-cyanoanthracene, purpurin or 4CzIPN) were subsequently screened for photocatalytic activity, but CO evolution was observed only in the presence of 4CzIPN.Under this condition, CoN 5 -PMO outperformed the other catalysts with a CO evolution of 2.54 ± 0.17 μmol mg −1 , a CO selectivity of 71%, and a TON CO of 62 after 1 h irradiation (Table 1, entries 9−13).CoN 2 and CoN 4 -PMO produced less CO at lower selectivity compared to CoN 5 -PMO.This result confirms that the Co centers coordinated with five N atoms provide the optimum active site structure for effective CO 2 photoreduction.In comparison, when the mononuclear molecular cobalt complex (CoN 5 dppz ) was employed as a homogeneous catalyst under identical conditions, a considerably slower CO evolution was observed with a TON CO of 20.9 after 1 h, highlighting the benefits of heterogenizing CoN 5 sites on PMO support toward facilitating CO 2 photoreduction (Table 1, entry 18).Notably, the catalytic activity dropped to only 0.50 ± 0.03 μmol of CO mg −1 when BIH was used as the donor with 4CzIPN (Table 1, entry 15), indicating that BIH is not effective at quenching photoexcited 4CzIPN.Control experiments demonstrate that all components including CoN 5 -PMO, 4CzIPN, TEOA, visible light, and CO 2 are required for photocatalysis, and a negligible amount of CO was evolved in the absence of any single component (Table S4).Interestingly, the CoN 5 -PMO/4CzIPN/TEOA photosystem produced 6.23 μmol H 2 mg −1 under N 2 , corresponding to a turnover number (TON H2 ) of ∼152.This indicates that the CoN 5 -active site is capable of catalyzing H 2 evolution, but the product selectivity is tuned toward CO under CO 2 -saturated conditions.
Photocatalysis experiments with CoN 5 -PMO over longer duration showed that the CO evolution started to plateau after 2 h for both [Ru-PS]/TEOA and 4CzIPN/TEOA combinations (Figure 4).The amounts of CO produced at the plateau were 2.35 ± 0.11 and 3.78 ± 0.32 μmol mg −1 for [Ru-PS] and 4CzIPN, respectively, corresponding to Co-based TON CO of 57 and 92.For [Ru-PS], the CO selectivity peaked at 81% after 15 min, followed by gradual decrease over 2 h due to the enhanced H 2 evolution catalyzed by [Ru-PS] photodegradation products. 60,61In contrast, the CO selectivity remained largely constant at 70−75% when 4CzIPN was employed, suggesting that CoN 5 -PMO retained the CO 2 reduction activity, and the longevity of the photocatalytic process was limited by the stability of photosensitizers.This hypothesis was confirmed by the addition of fresh 4CzIPN into the reaction mixture after CO evolution had ceased, which reactivated the system and resumed the CO evolution (Figure 5a).Notably, marked improvement in performance was observed when BIH was added as a donor to the [Ru-PS]/TEOA system, leading to evolution of 5.76 ± 0.47 μmol of CO mg −1 at 87% CO selectivity after 2 h (Figure S10).However, significant CO evolution was observed after 2 h in the control experiment with a [Ru-PS]/TEOA/BIH mixture without CoN 5 -PMO, caused by the Ru-based decomposition products.Therefore, we focused on the metal-free 4CzIPN/TEOA combination for subsequent studies due to its robust activity and noncatalytic photodegradation products.
The photocatalytic activity was optimized by varying the catalyst/photosensitizer ratio.At fixed concentrations of TEOA (10% v/v) and CoN 5 -PMO (1 mg), increasing the 4CzIPN concentration enhanced the overall rate of CO production (Figure S11a).Early deactivation of the system was observed at lower concentration of 4CzIPN, as demonstrated by bleaching of the yellow color and plateauing of CO evolution.The amount of CO evolved during the initial 15 min irradiation showed a linear correlation with concentration of 4CzIPN, suggesting a pseudo first-order dependence (Figure S11b).When the 4CzIPN concentration was fixed at 0.5 mM and the amount of CoN 5 -PMO was varied from 1 to 3 mg, a similar rate of CO evolution was observed with the highest TON CO obtained at low catalyst loading (Figure S12).This suggests that the activity of the system is limited by electron transfer to the catalyst.Moreover, light scattering in concentrated colloidal suspension can also impede the photocatalytic activity. 62The apparent quantum yield (AQY) for CO evolution by the CoN 5 -PMO/4CzIPN/TEOA photosystem was determined to be 0.51% at monochromatic irradiation of 467 nm, using ferrioxalate as a chemical actinometer (Figures S13 and S14). 63he heterogeneous nature of the photocatalytic system was investigated by six 1 h recycling experiments using a higher loading of CoN 5 -PMO (5 mg) to minimize the impact of material loss during workups between cycles.The CO evolution gradually increased with the consecutive recycling runs, with the activity reaching a maximum during the fourth cycle, followed by a slow decrease in the subsequent runs.The CO selectivity increased from first to third cycle and remained largely unchanged afterward at ∼80% (Figure 5c).A similar slow activation behavior was also reported in photocatalytic CO 2 reduction by cobalt phthalocyanine-modified mesoporous carbon nitride. 64For CoN 5 -PMO, the cumulative amount of CO produced over six recycling runs showed slightly slower rate during 0−2 h and a near-linear CO evolution rate during the subsequent 4 h photocatalysis (Figure 5d).Linear fit of the gas evolved during second to sixth runs yields a CO evolution rate of 3.82 μmol h −1 and a turnover frequency (TOF CO ) of 19.3 h −1 for 5 mg of CoN 5 -PMO (Figure S15).It should be noted that TOF CO is determined based on the initial Co loading and does not account for any loss of Co due to leaching.This result confirms the heterogeneous nature of the catalysis and suggests that the longevity of the system is likely limited by the photodegradation of 4CzIPN.Since the photosensitizer was replenished after each recycling run, sustained CO 2 photoreduction was observed with CoN 5 -PMO.Interestingly, ICP-OES analysis of CoN 5 -PMO after the recycling experiment showed a Co loading of 0.018 μmol mg −1 , indicating a loss of 56% Co during the six recycling runs.To corroborate this data, a leaching test was performed by filtering the reaction mixture after 30 min of irradiation and retesting the filtrate for photocatalytic activity.Irradiation of the CO 2 -saturated filtrate showed lower CO evolution rate, suggesting that contributions from both homogeneous and heterogeneous catalysts were present in the photosystem (Figure 5b).The homogeneous contribution likely came from the [Co(tpy)] fragments in solution that decoordinated from the dppz ligands on the PMO support.This result was supported by the photocatalytic activity displayed by Co(tpy)-Cl 2 under homogeneous conditions in the presence of 4CzIPN (1.28 μmol CO after 1 h, TON CO = 31, 90% CO; Table S4, entry 11).Addition of unfunctionalized nondppz PMO to the homogeneous photosystem had minimal effect on CO evolution activity (Table S4, entry 12).In solution, Co(tpy)Cl 2 can disproportionate to form [Co(tpy) 2 ] 2+ , 65 and therefore, the photolysate likely contained a mixture of both species that contribute toward CO evolution.However, the steady CO evolution rate observed during recycling experiments suggests that the contribution from heterogeneous catalysis is dominant.We hypothesize that the fraction of [CoN 5 ]-active sites buried deeply within the mesopores can potentially become more accessible with longer exposure time, which can compensate for the Co loss from leaching and lead to a steady CO evolution rate.Another possibility is that all [CoN 5 ] sites did not engage in catalysis due to the electron transfer step from photosensitizer to [CoN 5 ] being rate limiting.Since a higher catalyst (5 mg)-to-sensitizer (0.5 mM) ratio was used in the recycling experiments, it is likely that only the more accessible [CoN 5 ] sites engaged in catalysis during the initial runs.During the subsequent recycling runs, when some of these [CoN 5 ] sites underwent deactivation due to loss of Co(tpy) fragments, the reaction rate was compensated by the pristine and intact [CoN 5 ] units that did not engage in catalysis previously.
To evaluate structural changes during the photocatalytic reactions, the irradiated CoN 5 -PMO catalyst was characterized by low-angle XRD and TEM (Figure S16a,b), which demonstrated retention of the mesoporous structure after catalysis.The integrity of the catalytic centers was examined by Co 2p XPS surface analysis of the postirradiation catalyst, showing unaffected binding energies for Co 2p 3/2 and Co 2p 1/2 peaks at 780.9 and 796.8 eV, respectively, which suggested the preservation of the initial oxidation state for the remaining Co centers (Figure S16c).In addition, the partial loss of Co species during photocatalysis was confirmed with the decrease in the intensity of the Co 2p contributions in XPS and the reduction of both π−π* and MLCT transitions in comparison with the nonirradiated CoN 5 -PMO (Figure S16).
To gain mechanistic insights, the photosystem was probed fluorescence and in situ UV−vis spectroscopy.The molecular complex, CoN 5 bpy dimer, was used as a proxy for CoN 5 -PMO to perform these experiments under homogeneous conditions.Quenching experiments of a 4CzIPN solution excited at 400 nm demonstrated a clear decay of the emission at 560 nm in the presence of TEOA (Figure S17a).The reductive quenching rate constant (k q ) was determined to be 9.6 × 10 9 M −1 s −1 based on the Stern− Volmer equation, close to the diffusion limit (Figure S17b).In contrast, the oxidative quenching of photoexcited 4CzIPN* by CoN 5 bpy dimer was minimal (Figure S18).Furthermore, the concentration of TEOA in the photocatalytic experiments was significantly higher than the catalyst concentration, and consequently, the reductive quenching of 4CzIPN* to form [4CzIPN] •− is expected to be the first electron-transfer step.The reduced sensitizer, [4CzIPN] , is a stronger reductant than the photoexcited sensitizer (E 1/2 (PS*/PS + ) = −1.04V). 7,36 Based on the electrochemical data for [CoN 5 bpy ] 2+ and [CoN 5 dppz ] 2+ , [4CzIPN] •− is sufficiently reducing to generate the reduced catalyst and prompt CO 2 reduction in the presence of TEOA (onset potential −1.51 V).
In situ UV−vis spectroscopy was employed to probe the "model" reaction mixture containing [CoN 5 bpy ] 2+ , 4CzIPN, and TEOA under irradiation with visible light.As shown in Figure 6a, the preirradiation spectrum of the solution displayed peaks at 504 and 554 nm corresponding to [CoN 5 bpy ] 2+ and at 433 nm for 4CzIPN, suggesting no electron transfer in the dark.Under irradiation, the color of the CO 2 -saturated reaction mixture changed from yellow to dark brown with the appearance of new absorbance bands at 420 and 550 nm (Figure 6b).The intensity of the bands gradually increased with the exposure time and plateaued at ∼30 min, fitting a sigmoidal process (Figure 6c).In contrast, when 4CzIPN was irradiated in the presence of TEOA, a decrease in absorbance was observed for the characteristic bands at 363 and 432 nm with a concomitant increase at 319 and 328 nm (Figure S19), consistent with the reductive quenching of 4CzIPN* by TEOA.This indicates that the new absorption bands observed during photocatalysis at 420 and 550 nm can be attributed to the reduced [CoN 5 bpy ] species.Upon removing the light source, the absorption spectrum of the photolysis solution returned to the preirradiation state, albeit with an additional band at ∼630 nm, suggesting that the catalyst did not undergo photodegradation (Figure 6b�inset).The additional peak at 630 nm could be related to coordination of a different ligand to Co in the sixth position when the catalyst returned to its resting state, [CoN 5 bpy ] 2+ .The spectral change during photocatalysis was corroborated by spectroelectrochemical (SEC) analysis of the CoN 5 bpy dimer (Figure 6d).species.We speculate that the slightly lower activity observed during the first cycle of recycling tests could be related to the hypothesis that CoN 5 -PMO serves as a precatalyst and requires a reduction step to form the active Co II catalyst.Furthermore, use of 5 times higher catalyst loading in the recycling experiment could also contribute to a slow reduction process as light scattering in a concentrated suspension becomes a limiting factor.

■ CONCLUSIONS
In summary, we have successfully demonstrated modulation of coordination structures of heterogenized molecular catalysts to control CO 2 photoreduction activity in colloidal suspension.A dppz-functionalized PMO was used as a scaffold for immobilization of [M(tpy)] and [M(bpy)] moieties (M = Co or Ni) to construct four different materials consisting of or Ni-active sites with different numbers of coordinated N atoms coordinated with the metal center.Among the series, CoN 5 -PMO displayed superior activity toward photoreduction of CO 2 , demonstrating that both the nature of metal center and the primary coordination structure strongly influence the catalytic performance.The precious metal-free photosystem containing CoN 5 -PMO reached a TON CO value of 92 after 3 h at ∼75% CO selectivity, and the deactivation of the system was caused by the photodegradation of the sensitizer.Recycling experiments demonstrated that the CO evolution remained nearly constant over six cycles with an average TOF CO value of 19.3 h −1 .To the best of our knowledge, this work presents a rare example of utilizing mesoporous organosilica-based catalysts in a precious metal-free photosystem for CO 2 reduction.Notably, the heterogeneous catalyst (CoN 5 -PMO) outperformed the analogous molecular Co complex (CoN 5 dppz ) in solution, highlighting the advantage of catalyst immobilization.The coordination structure of the Co-active site in CoN 5 -PMO was confirmed by XAS analysis.Electrochemical studies on CoN 5 dppz under homogeneous conditions indicated that the dppz ligand likely plays an important role in catalysis by enabling the reduction of the active site at milder potential.This work develops a versatile strategy for regulating the molecular structure of catalytic metal sites grafted on a mesoporous support and highlights the interplay between the coordination environment and catalytic CO 2 reduction.
CoN 5 -PMO.NdppzPMO (30 mg) was suspended in a methanolic solution (10 mL) of Co(tpy)Cl 2 (6 mg, 0.017 mmol), and the mixture was refluxed overnight under a nitrogen atmosphere.The resulting solid was collected by filtration, washed with methanol to remove any unreacted Co(tpy)Cl 2 complex, and dried under vacuum to give CoN 5 -PMO (30 mg).
CoN 2 -PMO.NdppzPMO (30 mg) was suspended in a THF solution (10 mL) of CoCl 2 •6H 2 O (4 mg, 0.017 mmol), and the mixture was refluxed overnight under a nitrogen atmosphere.The resulting solid was collected by filtration, washed with THF and methanol to remove any unreacted CoCl 2 •6H 2 O, and dried under vacuum to give CoN 2 -PMO (30 mg).Co loading from ICP: 0.324 mmol of Co g −1 .
Photocatalytic CO 2 Reduction.Photocatalytic CO 2 reduction experiments were performed in 10 mL clear glass screw vials (Thermo Fisher, catalogue number 10-SV) sealed with rubber septa.In a typical photocatalytic reaction, PMO catalyst (1 mg) was dispersed in 9:1 (v/ v) MeCN/TEOA (4 mL) mixture containing 0.5 mM photosensitizer (Ru(bpy) 3 2+ or 4CzIPN).For selected samples, 10 mM BIH was added to the reaction mixture as an electron donor.The mixture was purged with CO 2 for 15 min and irradiated under 1 sun illumination (100 mW cm −2 ) using a SciSun-LP-150 solar simulator.For the homogeneous counterpart experiments, the molecular catalyst (0.04 μmol) was added to the reaction mixture (4 mL) instead of the heterogeneous catalyst, followed by purging with CO 2 .Control experiments were conducted under similar conditions by suppressing one component of the photocatalytic system (i.e., visible light, heterogeneous catalyst, photosensitizer, sacrificial electron donor, and CO 2 ) to assess the influence of each parameter individually.The temperature of the photocatalysis mixture was maintained at ∼27 °C.
The catalytic activity was expressed in terms of TON CO , determined as moles of CO produced per mole of cobalt or nickel, while the selectivity of the system toward CO was estimated according to following equation: Recycling Experiments.The PMO catalyst (5 mg) was subjected to multiple photocatalytic cycles of 1 h irradiation under identical conditions.After each photocatalytic CO 2 reduction cycle, the catalyst was collected by centrifugation and washed three times with acetonitrile to remove physisorbed species.The resulting solid was dried under vacuum and used for subsequent recycling experiments by adding fresh photosensitizer and MeCN/TEOA solution.

Figure 3 .
Figure 3. X-ray absorption data: (a) Co K-edge XANES spectra and (b) Fourier transform R-space EXAFS data of CoN 5 -PMO, CoN 5 bpy dimer, Co(bpy)Cl 2 , and CoCl 2 ; and (c) EXAFS fitting of CoN 5 -PMO with the data shown as open circles and the FEFF fit as a solid red line.
Co or Ni, x = 4 or 5).e 0.041 μmol of CoN 5 dppz in 4 mL of MeCN/TEOA (9:1 v/v) containing 0.5 mM 4CzIPN was used for photocatalysis, and the total amount of CO and H 2 evolved after 1 h irradiation is reported in the table.

Figure 5 .
Figure 5. (a) Reactivation of photocatalysis mixture by addition of fresh 0.5 mM 4CzIPN (red trace).Black dashed line demonstrates plateauing of the activity after 2 h of irradiation.(b) Filtration experiment for CoN 5 -PMO to distinguish the activity of the solid catalyst and active species in the solution.Solid line shows the normal course of the reaction using colloidal CoN 5 -PMO.The dashed line shows the activity of the filtrate collected after 30 min irradiation.(c) Recycling experiment for CoN 5 -PMO (5 mg) showing CO evolution and CO selectivity during six 1 h photocatalysis experiments.(d) Cumulative amount of CO and H 2 evolved over the course of 6 h recycling experiments.
Electrochemically reduced [CoN 5 bpy ] solution at −2.0 V exhibited two absorption bands at 418 and 555 nm, and reoxidation of [CoN 5 bpy ] at +0.2 V regenerated the original spectrum with an additional band at 630 nm.The SEC data are consistent with the spectral changes observed during photocatalysis, indicating that the electron transfer from [4CzIPN] •− to [CoN 5 bpy ] 2+ generated the reduced Co species to initiate CO 2 binding and its subsequent photoreduction to CO.Based on literature reports, 5,54,66−68 and supported by our mechanistic investigations, a tentative photocatalytic cycle for the CoN 5 -PMO/4CzIPN/TEOA photosystem is proposed in Scheme 1.The photoexcited 4CzIPN* is reductively quenched by TEOA to form [4CzIPN] •− , which has sufficient reducing power to reduce the [CoN 5 ]-active sites in PMO and initiate the CO 2 reduction catalytic cycle.The [CoN 5 ] center undergoes two metal-centered reduction by [4CzIPN] •− , and a further ligand-centered reduction to yield [Co I (tpy)-(dppz •− )(S)] species anchored to PMO (S = solvent).We hypothesize that partial delocalization of the electron within the dppz ligand stabilizes the reduced species and prevents its degradation.Subsequent dissociation of the solvent and binding of CO 2 forms an intermediate species, [Co II (tpy)-(dppz •− )(CO 2 − )].Protonation of this intermediate by TEOA leads to the formation of [Co II (tpy)(dppz •− )(CO 2 H)] + which is subsequently converted to [Co II (tpy)(dppz)(CO)] 2+ through C−O bond cleavage and the release of water.Finally, the catalyst is regenerated by the dissociation of the Co−CO bond and release of CO, restoring [Co II (tpy)(dppz)(S)]

Figure 6 .
Figure 6.(a) UV−vis absorption spectra of 4CzIPN (0.1 mM), CoN 5 bpy (0.11 mM), and their mixture in MeCN/TEOA (9:1 v/v).(b) Evolution of the UV−vis spectra of the photocatalysis mixture containing 4CzIPN (0.1 mM) and CoN 5 bpy (0.11 mM) in CO 2saturated MeCN/TEOA (9:1 v/v) during visible light irradiation for 30 min, followed by 1 h storage in the dark.(c) Kinetic trace for the formation of photogenerated reduced species, demonstrated by the change in absorbance of the 550 nm band with irradiation time.The red line shows the sigmoidal fitted curve.(d) In situ UV−vis spectroelectrochemistry of the CoN 5 bpy dimer (∼5 mM) in MeCN.The platinum working electrode was held at each potential for 2 min before recording the spectrum.The pre-electrolysis spectrum recorded at open-circuit voltage is shown in dark gray.