Multielectron transportation of polyoxometalate-grafted metalloporphyrin coordination frameworks for selective CO2-to-CH4 photoconversion

Abstract Photocatalytic CO2 reduction into energy carriers is of utmost importance due to the rising concentrations of CO2 and the depleting energy resource. However, the highly selective generation of desirable hydrocarbon fuel, such as methane (CH4), from CO2 remains extremely challenging. Herein, we present two stable polyoxometalate-grafted metalloporphyrin coordination frameworks (POMCFs), which are constructed with reductive Zn-ϵ-Keggin clusters and photosensitive tetrakis(4-carboxylphenyl)porphyrin (H2TCPP) linkers, exhibiting high selectivity (>96%) for CH4 formation in a photocatalytic CO2-reduction system. To our knowledge, the high CH4 selectivity of POMCFs has surpassed all of the reported coordination-framework-based heterogeneous photocatalysts for CO2-to-CH4 conversion. Significantly, the introduction of a Zn-ϵ-keggin cluster with strong reducing ability is the important origin for POMCFs to obtain high photocatalytic selectivity for CH4 formation, considering that eight MoV atoms can theoretically donate eight electrons to fulfill the multielectron reduction process of CO2-to-CH4 transformation.


INTRODUCTION
Excessive CO 2 discharge derived from the continuous burning of fossil fuels has caused global warming and environmental issues [1,2]. Artificial conversion of excess CO 2 into a serviceable energy product is an important pathway to achieve sustainable development [3][4][5][6]. Solar-driven photocatalytic reduction of CO 2 to carbon-neutral fuels (CO, CH 4 ) and/or value-added chemicals (HCOOH, CH 3 OH) affords a feasible strategy for the aforesaid conversion [7][8][9][10][11]. The implementation of this reaction can mitigate the greenhouse effect and energy crisis simultaneously. However, the structural activation process of the CO 2 molecule is particularly difficult because of its intrinsically chemical inertness and high C=O bond-cleavage enthalpy [12]. In order to circumvent the highly negative equilibrium potential versus normal hydrogen electrode (NHE) for thermodynamically unfavorable CO 2 •-intermediate, proton-assisted multiple-electron reductive products including chemicals and/or hydrocarbon are commonly obtained so as to lower the activation energy of photocatalytic CO 2 conversion [10]. Even so, the formation of high-order protonand electron-transferring products still needs to surmount a considerable kinetic barrier, and the competitive H 2 evolution further increases the difficulty for getting the aimed product selectively [13,14]. For instance, the photosynthesis of CH 4 -one kind of the most desirable and valuable hydrocarbon fuels in the photoreaction system-has been a great challenge [15], since the accomplishment of the eight-electron-transport process requires the photocatalyst to offer both strong reducing capability and sufficient electrons theoretically.
At present, the reported photocatalysts for CH 4 formation mainly fall into two categories: (i) heterogeneous nanoscale semiconductors [16][17][18], which RESEARCH ARTICLE exhibit relatively low activity and selectivity for CH 4 despite suitable band gaps and good recyclability; and (ii) homogeneous molecular metal complexes, with identifiable active centers, very few of which show high activity and selectivity for CH 4 [19]. Although their crystallography essence is propitious to the clarification of the photocatalytic mechanism [20][21][22][23][24][25][26], the structural dissolubility makes them hard to separate from reductive products. In view of the practical application, constructing a heterogeneous photocatalyst with a well-defined structure for selective CH 4 generation may be a good chance. Recently, a handful of moderate-or high-valence metal or metal cluster-based coordination frameworks (MCFs) have been investigated in the field of CO 2 photoconversion [27][28][29][30][31][32][33][34][35][36][37], but the majority of reductive products primarily focus on twoelectron transferred CO/HCOOH. Only the individual porphyrin-based MCF family (MOF-525) presents a low selectivity for CH 4 generation [28], which is probably related to the insufficient redox and electron-donating abilities of catalytically active units. In addition, the structural collapsibility of ordinary MCFs in aqueous phase enables the photocatalytic reaction to occur exclusively in an organic medium [29,30] that no doubt exerts adverse effects on the environment.
Based on the aforementioned considerations, we conceived that polyoxometalate (POM)-based coordination frameworks (POMCFs), with wellknown structural stability and favorable catalytic performance [38,39], are probably more beneficial to execute the photocatalytic reduction of CO 2 due to the synergistic effect originated from the integration of POM and MCF [40][41][42][43]. In particular, the Zn-ε-Keggin cluster of the PMo 12 'electron sponges' family [44,45], including eight Mo V atoms, can behave as a strong reductive component and contribute eight electrons theoretically. In addition, the Zn-ε-Keggin, a tetrahedral node, is formed by four-trapped Zn (II) locating in ε-Keggin (PMo 12 ). Compared with most anionic POMs, the ε-Keggin modified with metal Zn becomes a cationic cluster, which is favorable for coordination with organic ligands. Consequently, if the reductive POM cluster and porphyrin derivative can be employed to fabricate POMCF, having both the visible-light-harvesting and photo-excited electron migration, which would be a good strategy towards selectively photoreducing CO 2 to multielectronreductive products.
Herein, we report on two structural analogous porphyrin-based POMCFs, [ Both of them show expected architectural stability and preeminent performance on the photocatalytic reduction of CO 2 . The whole photocatalytic reaction was performed in aqueous solution, without any participation of a photosensitizer and preciousmetal co-catalyst. It is worth noting that the strong reducing ability of the Zn-ε-Keggin entity, combined with the remarkable optical and electrical properties of the TCPP linker, successfully endows these compounds with superior photocatalytic selectivity for CH 4 production. To our knowledge, the high CH 4 selectivity of 96.6% for NNU-13 has surpassed all of the reported heterogeneous coordination framework-based photocatalysts [46], while NNU-14 displays a slightly low CH 4 selectivity of 96.2% by reason of its subtle structural deformation.

RESULTS AND DISCUSSION
Crystalline NNU-13 and NNU-14, prepared through similar hydrothermal synthesis protocols (Supplementary Fig. 1), show nearly identical host frameworks constructed from Zn-ε-Keggin nodes and four-connected TCPP linkers. Single-crystal X-ray diffraction analysis demonstrated that these two compounds crystallize in the orthorhombic Fmmm (NNU-13) and monoclinic C2/m (NNU-14) space groups (Supplementary Table 1), respectively. We have been working on the polyoxometalate-grafted photosensitive/electrosensitive ligands, such as NNU-13 and Co-PMOF [47]. Compared to the synthesis method of NNU-13, the pH value of the solution was different and tris-(4-pyridyl)triazine (TPT) was additionally added when NNU-14 was synthesized. The TPT acts as a templating agent in the synthesis of NNU-14. In each Zn-ε-Keggin cluster, two Zn1 sites along the b axis coordinate with two different TCPP ligands, while the remaining two Zn2 sites along the a axis oppositely bridge with two adjacent Zn-ε-Keggin motifs, under the symmetry operation of the 2 1 screw axis, forming a wavy POM chain in the end ( Fig. 1a and b). In comparison, the octahedral Zn3 atom is captured by the 'pocket' of the TCPP center with two axial coordinated H 2 O molecules. Four carboxyl groups of each TCPP ligand concurrently are linked to four Zn1 atoms from different Zn-ε-Keggin clusters of four POM chains (Fig. 1c). Subsequently, a 3D network is established by connecting Zn-TCPP linkers with Zn1 sites of the POM chain alternately (Supplementary Figs 2 and 3), leading to an approximate dihedral angle of 105.6 • between the POM chain and Zn3-TCPP metalloligand along the crystallographic a axis (Fig.  1d). Ultimately, NNU-13 shows a two-fold interpenetrated mog topological structure ( Supplementary  Fig. 4a) as a result of multiple existing symmetry elements (rotation and screw axes together with mirror and glide planes). Likewise, NNU-14 also has the same mog topology ( Supplementary Fig. 4b), but its comparatively low spatial symmetry gives rise to approximate dihedral angles of 113.3 • and 97.7 • between the POM chain and Zn-TCPP linker along the crystallographic a axis (Fig. 1d). Notably, this structural distortion leads to the divergence on lattice parameters ( Supplementary Fig. 5) and the shortest Zn-Zn distances (between the active Zn center in the TCPP pocket and the nearest Zn atom on the Zn-ε-Keggin cluster) of 10.7 (NNU-13) and 10.1Å (NNU-14), respectively ( Fig. 1e and Supplementary Table 2b).
The phase purity and chemical stability of NNU-13 and NNU-14 compounds were confirmed by powder X-ray diffraction (PXRD) patterns ( Supplementary Fig. 6). The two POMCF samples were separately immersed into aqueous solutions with various pH values from 5 to 12 for at least 12 h. Noticeably, the PXRD patterns of all treated crystals remain intact, indicating that no phase transition or structural collapse occurred, even under the condition of the conventionally artificial photosynthesis-reaction system (Supplementary Fig. 6) [48]. The two POMCFs are relatively stable, mainly due to the high-temperature hydrothermal synthesis and small pore sizes. There are many other strategies that can be used for the synthesis of stable Zn-MOF such as the introduction of hydrophobic methyl groups [49]. NNU-13 can adsorb CO 2 with the maximum uptake of 34.1 cm 3 /g at 298 K, while NNU-14 can adsorb CO 2 with the maximum uptake of 14.3 cm 3 /g at 298 K ( Supplementary Fig. 7). Both NNU-13 and NNU-14 have the ability to capture CO 2 , which is conducive to CO 2 photoreduction. Besides, in order to validate the visible-light-enrichment capability, the corresponding diffuse reflectance UV-vis spectra were characterized. Both the compounds NNU-13 and NNU-14 exhibit broad adsorption covering the whole UV-vis region, as well as a red shift of the Soret band, along with a slight increase in the band intensity ( Supplementary Fig. 8a). Moreover, the decreased number of adsorption peaks at the Q band compared with the free H 2 TCPP ligand, associated with increased molecular symmetry for metalloporphyrin, indicates the occurrence of charge transfer between the zinc ion and the TCPP ligand. Based on these results, the band gaps are ∼1.59 (NNU-13) RESEARCH ARTICLE and 1.64 eV (NNU-14)-slightly higher than the H 2 TCPP ligand of 1.53 eV ( Supplementary  Fig. 8b)-unveiling the potential for these POM-CFs being as semiconducting photocatalysts. To elucidate the semiconductor character of title compounds and their possibilities for the subsequent photocatalytic conversion of CO 2 , Mott−Schottky measurements were performed at different frequencies (Supplementary Fig. 8c and d). From the C −2 values versus the applied potentials function relationship, the positive slopes of the curves are consistent with that of the typical n-type semiconductors, and the flat-band potentials determined from the intersection point are approximately −1.18 V versus Ag/AgCl (NNU-13) and −1.30 V versus Ag/AgCl (NNU-14). Since the bottom of the conduction band (LUMO) in the n-type semiconductors is commonly close to the flat-band potential [50], the LUMO of NNU-13 and NNU-14 can be estimated to be −0.98 V (versus NHE) and −1.10 V (versus NHE), respectively (Supplementary Fig.  8c and d). In view of such negative potentials for LUMO in POMCFs, the CO 2 molecule theoretically can be converted into CO (−0.53 V versus NHE) and CH 4 (−0.24 V versus NHE) upon visible-light irradiation. The CV curves of NNU-13 and NNU-14 were carried out in the reaction solution and four pairs of redox peaks appeared in the potential range from −0.6 to 0.5 V versus Ag/AgCl, attributed to the four-single-electron transfer of the POM secondary building block in the alkaline solution ( Supplementary Fig. 9).
The visible-light-driven CO 2 reductions of NNU-13 and NNU-14 were conducted under a pure CO 2 atmosphere in aqueous solution with triethanolamine (TEOA) as a sacrificial agent, in the absence of any photosensitizer and precious-metal co-catalysts. Gaseous CH 4 and CO are the main reactive products detected by gas chromatography and no competitive H 2 evolution was observed in the whole photoreaction process ( Fig. 2 and Supplementary Figs 10 and 11). Such low-photocatalytic H 2 -production activities of NNU-13 and NNU-14 can be further supported by replacing CO 2 with N 2 (Supplementary Table 3). After 6 h, only 0.055 (NNU-13) and 0.035 (NNU-14) μmol H 2 were detected under identical reaction conditions, without any CO generated. As indicated by Fig. 2a and b, the output of CH 4 increases steadily with a prolonged time of light irradiation, while the CO yield has a negligible growth. With the reaction going on, the amount of CH 4 for NNU-13 reaches a maximum of 3.52 μmol (i.e. 704 μmol g −1 ) after 6 h, whereas NNU-14 performs a CH 4 generation of 1.56 μmol (i.e. 312 μmol g −1 ) after 7 h. By contrast, the CO amounts determined after the reaction peaked at 4.2 μmol g −1 h −1 (NNU-13) and 1.2 μmol g −1 h −1 (NNU-14). Moreover, both NNU-13 and NNU-14 exhibit high selectivity (CH 4 over CO) of 96.6% and 96.2% (Fig. 2c), due to the similar connection strategy between the POM and TCPP. It is worth noting that both the selectivity and the activity of CH 4 for these POMCFs are the highest among the reported heterogeneous coordination framework-based photocatalysts applied in the photocatalytic reduction of CO 2 (Supplementary Table 4) [28,34]. There are no Co and Fe in NNU-13 and NNU-14, which reveals that the generation of CH 4 eliminates the interference of Fe and Co (Supplementary Table  5). The relevant parameters of this photoreduction system, including TONs, TOFs and CH 4 , are summarized in Supplementary Table 6. The overall CH 4 -TON of NNU-13 was estimated to be 3.56, which is higher than that of NNU-14 (1.57), suggesting that both of these complexes are indeed catalytic for CO 2 -to-CH 4 conversion. The apparent quantum yield of these optimized POMCFs-promoted CO 2photoreduction systems were estimated to be 0.04% (NNU-13) and 0.02% (NNU-14) at a monochromatic irradiation of λ = 550 nm. Obviously, the compound NNU-13 has higher photocatalytic CO 2reduction activity than NNU-14, as demonstrated by evidently distinguishing the transient photocurrent responses and electrochemical-impedance spectra ( Supplementary Figs 12 and 13). Although both of them display obvious photocurrent responses to visible-light irradiation, NNU-13 has a much higher photocurrent than that of NNU-14, which means a better separation efficiency of the photo-induced electron-hole pairs for NNU-13. At the same time, the semicircle size of the Nyquist plot for NNU-13 is much smaller, reflecting an acceleration of the interfacial charge-transfer process, in accordance with the above photocurrent-response results. These differences in the charge separation and the kinetics of charge transfer can be attributed to the structural distortion on the dihedral angles between the POM chain and the Zn-TCPP linker in NNU-13 and NNU-14.
Additionally, in a series of reference experiments that were conducted in the absence of POMCF catalysts, CO 2 or light illumination, no detectable products were observed in the reaction system (Supplementary Table 3). Moreover, the influence of different POMCF qualities on the activity of the photoconversion CO 2 reaction was also investigated [31], for which the high selectivity of CH 4 over CO was still retained (Supplementary Fig. 14). The photocatalytic stability of these POMCFs was evaluated using recycling tests. From the time-course plots of CH 4 evolution, the photocatalysts maintain excellent activities even after three cycles (Supplementary   15). The slight decay in the CH 4 -evolution rate in each successive run is probably attributed to the mass loss in the recovery process of used samples. There has been no noticeable alteration on the infrared spectra and PXRD patterns performed before and after three cycles of the photocatalytic reaction that confirmed the structural robustness of NNU-13 and NNU-14 ( Supplementary Figs 16 and 17). Moreover, the filtrate reaction was further executed to confirm the heterogeneous nature of these POM-CFs. Once the catalysts were removed from the reaction system after several hours of photocatalytic reduction, the production of CH 4 and CO stopped, which clearly precluded the influence of potential intermediate or decomposed active components on the product generation ( Supplementary Fig. 18). From these results, both NNU-13 and NNU-14 possess high durability toward the CO 2 photocatalytic reduction under visible-light irradiation. After the photocatalytic reaction, extremely trace amounts of formic acid was produced in aqueous solution as detected by ion chromatography. An isotopic experiment using 13 CO 2 as substrate was performed under identical photocatalytic reaction conditions to validate the carbon source of the generated gases [18,33,51] and the products were analysed by gas chromatography and mass spectra. As shown in Fig. 2d and Supplementary Figs 19 and 20, the peaks at m/z = 17/16/15/14 and m/z = 29 were assigned to fragments of 13 CH 4 and 13 CO, respectively, providing solid proof that these POMCFs are indeed active and capable of selectively converting CO 2 to CH 4 under visible-light irradiation. In addition, taking NNU-13, for example, the experiment under the CO atmosphere was carried out and the result verified that CO is an intermediate for further CH 4 generation ( Supplementary Fig. 21).
To disclose the origin of and difference in the photocatalytic performances of NNU-13 and NNU-14, the roles of the structural components including the POM subunit (Zn-ε-Keggin) and TCPP organic linker were considered. On the one hand, it is well known that all the previously reported TCPP-based MOF photocatalysts in the photocatalytic CO 2reduction system, involving mono/multi-metal active center either in a moderate-or high-valence state, tend to produce high selectivity and activity, for CO over CH 4 , such as the MOF-525 family and Zr-MOFs [28,34]. Although high-valent metal clusters could serve as a multielectron acceptor in some cases, its poor reducing ability makes it hard to contribute enough electrons to satisfy the RESEARCH ARTICLE eight-electron process of CH 4 . Therefore, based on the reported porphyrin-MOFs for CO 2 photoreduction, two-electron transferred CO as a frequently predominant gaseous product is probably related to the reduction intensity of the active center. To confirm the photocatalytic activity of NNU-13 in the production of CH 4 , a series of control experiments were carried out, which involved Zn-ε-Keggin (Supplementary Figs 22 and 23), Zn-TCPP and the mixture of Zn-ε-Keggin and Zn-TCPP as independent photocatalysts (Supplementary Table 7). It is obvious that both the Zn-ε-keggin cluster and the Zn-TCPP linker produced CO (370 and 16 μmol g −1 , respectively), while the mixtures of Zn-ε-Keggin (POM) and Zn-TCPP (sensitizer) produced CO and CH 4 with a lower yield and selectivity for CH 4 . By contrast, NNU-13 assembled with POM and porphyrin showed a higher yield and selectivity for CH 4 . Moreover, the generated CO from Zn-ε-Keggin was mainly used for CH 4 formation. Therefore, the integration of photosensitive TCPP with reductive Zn-ε-Keggin within NNU-13 and NNU-14 played an indispensable role in obtaining high CH 4 selectivity. On the other hand, in order to eliminate the potential impact of the TCPP linker on high CH 4 selectivity and the activity of NNU-13 and NNU-14, similar Zn-ε-Keggin-based POMCF (NNU-12) ( Supplementary Figs 24 and 25) [52], with H 2 BCPT instead of TCPP as the bridging ligand and light-harvesting unit, was synthesized to execute a comparable CO 2 photocatalytic reaction. We found that NNU-12 still maintained a high CH 4 over CO selectivity of 81.3% (Supplementary Table 8), which again corroborated the significance of the POM component with a strong reducing ability and electron-transferring ability (i.e. redox ability). Because of the eight Mo V atoms in the reductive Zn-ε-Keggin cluster, it can theoretically easily offer eight electrons to complete the eight-electron reduction process from CO 2 to CH 4 . Of course, the relatively poor photocatalytic activity of CH 4 in NNU-12 also indicated the important roles of the excellent visible-light-harvesting and electron-transport capabilities of the TCPP linker, in particular for NNU-13 and NNU-14, whose high photocatalytic properties are primarily realized by the synergistic incorporation of the advantages of the Zn-ε-Keggin node and TCPP linker. In other words, the photo-generated electrons transferred from the TCPP linker flowing to the POM cluster are indispensable to the efficient photocatalytic CO 2 reduction. Additionally, the relative location and distance between the active center in the TCPP pocket and the Zn-ε-Keggin cluster (Fig. 1e) result in different levels of competition on the CO 2 molecule absorbed, as well as different numbers of active sites in the crystal cells (Supple-mentary Fig. 5) of NNU-13 and NNU-14, both of which are probably responsible for the discrepant photocatalytic activities for these POMCFs.
To obtain a deeper understanding on the photoexcited charge-carrier separation mechanism in these porphyrin-based POMCF catalytic systems, theoretical calculation was carried out based on density functional theory (DFT). The total density of states (TDOS), partial density of states (PDOS) and band structures are shown in Supplementary  Figs 26 and 27 and 3a-c, and the Fermi levels were taken as the energy zero. From the TDOS, we could see that the alpha-DOS (DOS: density of state) is different from the beta-DOS, which suggested that the MCF was with a single electron and a magnetic property (Supplementary Fig. 26). From the PDOS, we could see that the tops of the valence band (VB) were composed of the states of all elements, which means the top of the VB was distribution by both POMs and TCPP. The bottom of the conduction band (CB) was composed of the O 2p states, Zn 3d states and Mo 4d states. We could see that the Zn 3d states that composed the bottom of the CB are those in the POMs. It suggested that the bottom of the CB was distributed substantially by the POMs. This composition of the DOS suggested that both the POMs and the TCPP could act as electron donors while only POMs act as electron acceptors. In this case, once the light irradiates the MCF, the electrons in the VB are exited. Despite the charge transfer in the POMs, the electrons in the TCPP will be excited and transferred to the POMs, which matches our experimental prediction very well. In order to show the distribution clearly, the partialcharge-density maps of the VB and CB are shown in Fig. 3d and e. The partial-charge-density maps were intuitional figures involving the predominant distribution and transferring direction of CB and VB electrons. According to the partial-charge-density maps, we could see that the VB was distributed by the TCPP while the CB was distributed by POMs. Despite the charge transfer in POMs, the TCPP may act as an electron donor and POMs may act as electron acceptors, which matched well with our analysis of DOS. Based on the above calculations, the mechanism could be explained as follows: the light irradiates on the MCF and the electrons in the VB are first excited. Subsequently, the electrons in the VB (TCPP) transfer to the CB (POMs), which can be further utilized for the activation and reduction of absorbed CO 2 .
In light of the combination of the experimental results and theoretical calculations, a mechanism with respect to the superior photocatalytic CO 2reduction performances of TCPP-based POMCFs and possible photo-generated electron-transport  pathway in the system was proposed (Fig. 4). For most of the porphyrin-based MCFs, the catalytic active sites generally situated at the metal-modified porphyrin center and/or relatively high-valent metal/metal cluster nodes [28,34]. The porphyrin segments absorb the photos to generate electrons and allow the procreant electron transfer to the central metal ions/cluster, leading to the formation of CO as the primary gas product. However, for the TCPP-based POMCFs with strong reducing ability (NNU -13 and NNU-14), the predominant product is CH 4 . As the calculations revealed, the electrons of the tops of the VB were mainly concentrated on the TCPP ligands (Fig. 3d). The TCPP linkers absorb the light and move the photo-generated electrons on the VB toward the reductive Zn-ε-Keggin fragments. Zn-ε-Keggin and Zn-TCPP have been used as catalysts to perform CO 2 photoreduction reactions, respectively, and the results are listed in Supplementary Table 7. Obviously, these results illustrate that the synergistic effect between the Zn-ε-Keggin and the Zn-TCPP in NNU-13 can

RESEARCH ARTICLE
further reduce the produced CO to CH 4 , and CO is indeed the intermediate for CO 2 -to-CH 4 conversion. Ultimately, the enrichment of electrons on the reductive Zn-ε-Keggin fragments (based on the eight Mo V atoms) probably makes it preferable to provide enough electrons to CO 2 , allowing the further combination with the protons from the water to produce eight-electron-transferred CH 4 .
In the meantime, trace amounts of two-electrontransferred CO are still more likely to be generated on the Zn-TCPP sites relying on the ligand-to-metal charge transition (LMCT) effect. Therefore, CH 4 was the predominant reduced product with high selectivity in the gas production of the TCPP-based POMCF catalytic system. Of course, the Mo VI ions also receive the generated electrons and then deliver them, but with lower activity than Mo V atoms. The TEOA as the sacrificial reagent consumed the electron holes of the VB. The protons mainly come from the water due to its being the reaction solvent. An additional verification experiment using NNU-13 and acetonitrile (aprotic solvent including a little water) as photocatalyst and reagent was performed to prove this consideration, only generating very small amounts of CO (∼10 μmol g −1 ).

CONCLUSION
In summary, the heterogeneous photocatalytic reduction of CO 2 to CH 4 in the aqueous phase was realized for the first time over POM-containing MCFs, NNU-13 and NNU-14, fabricated with reductive Zn-ε-Keggin cluster and visible-light-responsive TCPP linker. Theoretical calculations revealed that the photo-generated carriers of the VB and the CB are mostly distributed on the TCPP group and Zn-ε-Keggin cluster, respectively. Consequently, the photo-excited electrons more easily flow to the POM port by efficient intercoupling between the reductive Zn-ε-Keggin unit and the TCPP linker. Therefore, these POMCFs exhibit high photocatalytic CH 4 selectivity (>96%) and activity that have far surpassed all of the reported MCF-based photocatalysts. Note that the introduction of POM building blocks with potent reducing ability not only endows NNU-13 and NNU-14 with favorable structural rigidity, but also it indeed facilitates the photocatalytic selectivity of CH 4 by theoretically delivering adequate electrons to accomplish the eight-electron reduction of the CO 2 molecule. We expect that such a feasible approach, assembling a strong reducing component into visible-light-sensitized photocatalyst architecture, can ignite research enthusiasm towards the construction of efficient POMCF photocatalysts for the highly selective reduction of CO 2 to CH 4 or other high-valued hydrocarbons.

Single-crystal X-ray crystallography
Single-crystal X-ray diffraction (XRD) data of NNU-13 and NNU-14 have been measured on a Bruker APEXII CCD diffractometer with graphitemonochromated Mo Kα radiation (λ = 0.71073Å) at 296 K. These two structures were solved by a direct method and the SHELXT program, and refined on Olex 2 software by the SHELXL program. A multi-scan technique was used for Absorption corrections. The TBA + cations could not be detected in the structure, and some other guest molecules are disordered. Thus, the data were further corrected with SQUEEZE on the PLATON software to eliminate some guest molecules and then check the space group of these two crystals. The detailed crystallographic information and selected bond lengths are listed in Supplementary Tables 1 and 2a.

Photochemical measurements
The photocatalytic CO 2 -reduction experiments were probed on an evaluation system (CEL-SPH2N, CEAULIGHT, China) in a 100-mL quartz container, and the two round openings of the container were sealed with rubber mats. A xenon arc lamp (CEL-HXF300/CEL-HXUV300, 200 mW/cm 2 ) with a UV-cutoff filter (420-800 nm) was utilized as irradiation source. The as-prepared photocatalyst (5 mg) was evacuated in mixed solutions (30 mL) with analytical reagent (AR) TEOA (2 mL) and deionized water (28 mL) and pre-degassed with CO 2 (99.999%) for 30 min to remove air before irradiation. The pH values were determined to be 10.5, 10.5 and 7.3 in the absence of the catalyst with the TEOA, after suspension of the catalyst and after sparging with CO 2 (30 min). The sealed reaction system (CO 2 pressure of 1 atm) was positioned about 10 cm away from visible-light source, and stirring was continued constantly to ensure that the photocatalyst particles were in suspension. The reactor was connected using a circulating cooling water system to maintain the temperature of the solution at around 20 • C. Gaseous product was measured using gas chromatography (GC-7900, CEAULIGHT, China, column type: TDX-1) equipped with a flame ionization detector (FID) and a thermal-conductivity detector (TCD). The isotope-labeled experiments were recorded using gas chromatography-mass spectrometry (GC-MS, 7890A and 5875C, Agilent). GC-MS column (19091P-Q04PT), gas (He as carrier), flow rate (1.4 mL/min, 5.2027 psi), oven ramping parameters (injector temperature: 250 • C, column temperature is constant temperatures: 30 • C).

Photoelectrochemical studies
Five milligrams of crystal samples were ground and dispersed in 1 mL mixed solution of ethanol and water under sonication for 1 h. Then, 10 μL above the turbid liquid was daubed on indium-tin oxide (ITO) glass with a fixed circular area (diameter 6 cm) and the electrode was dried for about 24 h in a desiccator. The electrochemical experiments were performed on an electrochemical workstation (CHI 660e) in a standard three-electrode system. The Mott-Schottky tests were performed in 0.2 M Na 2 SO 4 solution, using a catalyst/ITO electrode as the working electrode, an Ag/AgCl electrode as the reference electrode and a Pt plate as the counter electrode at different frequencies of 500, 1000 and 1500 Hz. The photocurrent tests were carried out in tris-HCl electrolyte (0.1 M) at a bias of 0.0 V using an Xe lamp as the light source. Electrochemicalimpedance spectroscopy measurement was carried out on the electrochemical workstation (Bio-Logic, VSP) by applying an AC voltage with 10-mV amplitude in a frequency range from 1000 kHz to 100 mHz.

SUPPLEMENTARY DATA
Supplementary data are available at NSR online.