Main

Reticular chemistry—the construction of materials based on extended linkages between molecular building blocks—provides a powerful toolbox for developing solid-state materials with well-defined crystal structures, designed chemical functionalities and tunable porosities1,2,3,4. Among reticular chemicals, metal–organic frameworks (MOFs) constructed by metal–organic coordination linkages5,6,7, and covalent organic frameworks (COFs) constructed by reversible covalent organic linkages8,9,10 stand out as they show promise for applications including gas storage and separation, catalysis, drug delivery, sensing and ion transport11. However, MOFs and COFs each possess distinct advantages and drawbacks, posing a dilemma in the quest for optimal reticular materials. MOFs typically exhibit high crystallinity due to the high reversibility and rigid geometry of metal–organic linkages12, thus ensuring well-defined structures, which can be solved by single-crystal X-ray diffraction (SCXRD). Such crystallinity is difficult to achieve in COFs, and only a few COFs have been analysed by SCXRD13,14,15,16,17. On the other hand, COFs provide easier access to high chemical stability due to their strong covalent bonds12 which can be further reinforced by postsynthetic locking reactions18. MOFs require specific constituents (for example, Zr(IV) carboxylates such as UiO-66 (ref. 19) or Ni(II) pyrazolates such as PCN-601 (ref. 20)) to achieve stability under aqueous conditions or against acids or bases.

Given these complementary attributes, we envisioned using both metal–organic and reversible covalent organic linkages to obtain materials with balanced properties. So far, such a combination of linkages has been limited to cases where either type of linkage creates finite (that is, non-extended) building units (Fig. 1b and Supplementary Fig. 1). For example, some frameworks such as woven COFs are synthesized using building units containing isolated metal–organic linkages, while the framework itself is extended by covalent organic linkages only (hence they are usually called (metal-)‘COFs’)21,22,23,24,25,26. Conversely, in some cases, covalent organic linkages are used to build finite building units, and extended metal–organic linkages reticulate these into frameworks27,28,29,30,31. Other frameworks have been synthesized by the alternation of finite building units based on metal–organic and covalent organic linkages32,33,34,35,36,37,38,39,40. Such coexistence of metal–organic and covalent organic linkages in one framework has substantially expanded the toolbox of reticular chemistry41,42. However, there does not appear to be any report in which metal–organic and covalent organic linkages create extended substructures for each type of linkage, which are both reticulated into a continuous framework. Such a ‘double’ extension scheme would thus constitute a new reticular approach enabling a close interplay between MOF and COF chemistry.

Fig. 1: MOCOFs.
figure 1

a, Reaction scheme for the synthesis of MOCOF-1. Dmt, 4,4′-dimethoxytrityl group; X, variable counteranion. b, Schematic representation of MOCOFs by the double extension of metal–organic and covalent organic linkages, compared with conventional MOFs, COFs and previous MOF–COF fusion examples with single extension. c, SCXRD structure of MOCOF-1 (298 K). Hydrogen atoms and solvent molecules are omitted for clarity.

Full size image

Here we report the synthesis of crystalline porous frameworks by double extension of metal–organic and covalent organic linkages, resulting in a class of materials that we name metal–organic–covalent–organic frameworks (MOCOFs) (Fig. 1b). The first example, MOCOF-1, is obtained by the combination of coordinative bonds and imine linkages, each forming a three-dimensional (3D) continuous framework (Fig. 1a). The framework comprises an extended coordination network and infinite imine-linked organic rods (Fig. 1c). MOCOF-1 shows higher crystallinity than similar imine COFs due to its metal–organic linkages. On the other hand, MOCOF-1 also has higher chemical stability than similar MOFs, supported by covalent organic linkages. Furthermore, the structure of MOCOF-1 is tunable by isoreticular chemistry with different substituents and linker lengths. Thus, MOCOFs open up uncharted territory in reticular materials that combine the properties of MOFs and COFs.

Results and discussion

Structure of MOCOF-1

MOCOF-1 was synthesized by a solvothermal reaction between [5,10,15,20-tetrakis(4-((4,4′-dimethoxytrityl)amino)phenyl)porphinato]cobalt(II) (Co(tdpp)), terephthalaldehyde, water and Brønsted acids (Fig. 1a), where Co(tdpp) is a surrogate of [5,10,15,20-tetrakis(4-aminophenyl)porphinato]cobalt (Co(tapp)) with amino groups protected by 4,4′-dimethoxytrityl groups.

SCXRD analysis revealed that MOCOF-1 consists of deprotected Co(tapp) units linked by self-coordination of two amino groups with cobalt and imine condensation of the other two amino groups with terephthalaldehyde (Fig. 1c, Supplementary Figs. 2–6 and Supplementary Table 2). This desymmetrization of Co(tapp) is a result of the unbalanced self-coordination of the Co(tapp) metalloligand, which bears four amino groups but can accommodate only two ligands on each cobalt site, similar to a reported cobalt–porphyrin self-coordination MOF43. In MOCOF-1, both coordination and imine linkages extend to form a porous crystalline framework in the cubic space group I213 with a unit cell parameter a = 31.9425(1) Å.

The net topology of MOCOF-1 was analysed using the ToposPro program44 dividing the net by each type of linkage, following the merged net theory developed for heterolinker MOFs45,46 (Supplementary Table 4). Regarding the coordination linkages, the Co(tapp) unit acts as a four-coordinated node via two axial coordination bonds on cobalt and two outward coordination bonds with amino groups (Fig. 2a–c), which extends to the lcv net, also known as the hyperkagome lattice (Fig. 2d). Regarding the imine linkages, the Co(tapp) unit is two-coordinated with terephthalaldehyde linkers and extends into a zigzag chain (Fig. 2a–c). These chains are arranged three-dimensionally in the three-way Π invariant rod packing47 without entanglements (Fig. 2d). Merging the lcv net and the Π invariant rod packing of zigzag chains gives an experimentally unreported net, which was theoretically proposed as thp-x-6-I213 (ref. 48) and is now registered as lvz in the Reticular Chemistry Structure Resource (RCSR) database49 (Fig. 2e and Supplementary Fig. 8). The lvz net features transitivity of 1231, complying with the minimal transitivity principle50 for one type of vertex (Co(tapp)) and two types of edges (cobalt–aniline coordination and terephthalaldehyde–diimine). MOCOF-1 also exhibits the maximum symmetry of the lvz net with Co(tapp) nodes (site symmetry, 2), terephthalaldehyde–diimine links (site symmetry, 2), asymmetric cobalt–aniline coordination links (site symmetry, 1), and the I213 space group (Supplementary Fig. 9). This topological analysis manifests the double extension structure of MOCOF-1, which is distinct from previous MOF–COF combination chemistry with single or alternate extension structures (Supplementary Fig. 1).

Fig. 2: Topological structure of MOCOF-1.
figure 2

a, Chemical structure of the topological nodes. b, Chemical structures of the topological links. c, Geometry of the nodes with each type of the links. d, Topological embeddings regarding each type of the links. e, Net embedding including both types of edges.

Full size image

Further topological analysis revealed the pore structures of the lvz net and MOCOF-1 (Supplementary Fig. 9). The faces of the lvz net are one type of an equilateral triangle and two types of twisted pentagons (Supplementary Fig. 9c). In MOCOF-1, the triangular face corresponds to a sterically closed three-membered ring based on Co(tapp) self-coordination, and the pentagonal faces correspond to five-membered rings based on both coordination and imine linkages with ∼10 Å apertures. The lvz net exhibits a single type of [32.56] tiles, which corresponds to irregular trigonal–bipyramidal cages in MOCOF-1 with the largest pore size of ∼19 Å (Supplementary Fig. 9d). These cages are interconnected in the pcu net topology through five-membered ring apertures in MOCOF-1, giving a continuous pore network (Supplementary Fig. 9e). This pore structure of MOCOF-1 agrees with the N2 sorption experiment results (see below).

MOCOF-1 is chiral because of the intrinsically chiral topology of the lvz net, in line with the topological chirality of the coordination lcv net and the organic Π invariant rod packing. The analysis of one single crystal gave a Flack parameter of 0.004(6), indicating the enantiopurity of the crystal. In the literature, topologically chiral reticular materials are often reported in the field of MOFs51 but rarely achieved by COFs because of the difficulty in designing a suitable building block52. MOCOF-1 shows that a chiral topology involving covalent organic linkages can be achieved in MOCOFs via structural guidance by metal–organic linkages.

The structure of bulk samples synthesized as microcrystals is confirmed to be the same as SCXRD by powder X-ray diffraction (PXRD), 1H NMR with acid digestion, 13C, 15N, 59Co solid-state (ss) NMR and infrared (IR) absorbance measurements. The PXRD pattern shows a good match with the simulated pattern (Supplementary Fig. 10). The 13C ssNMR spectrum shows at least 13 aromatic signals resolved, which can be explained by the desymmetrization of C4v-symmetric Co(tapp) into local C2 symmetry as in the SCXRD structure (Supplementary Fig. 11). The 15N ssNMR spectrum shows three signals which can be assigned to imine (−49.8 ppm), porphyrin (−271 ppm)53 and amine nitrogens upfield-shifted by the porphyrin ring current (−394 ppm) (Supplementary Fig. 12). The 59Co ssNMR spectrum shows a single Co(III) site with chemical shift and quadrupolar parameters (\(\delta_{\mathrm{cs}}^{\mathrm{iso}}=8{,}869\,{\mathrm{ppm}}\), \(|{\delta}_{\mathrm{cs}}^{zz}-{\delta}_{\mathrm{cs}}^{\mathrm{iso}}|=1{,}763\,{\mathrm{ppm}}\), e2qQ/h = 3 MHz from the static spectrum) that match well with those of typical hexacoordinate cobalt porphyrins (\(\delta_{\mathrm{cs}}^{\mathrm{iso}}=8{,}900\,{\mathrm{ppm}}\), \(|{\delta}_{\mathrm{cs}}^{zz}-{\delta}_{\mathrm{cs}}^{\mathrm{iso}}|=1{,}750\,{\mathrm{ppm}}\), e2qQ/h ~ 4 MHz)54 (Supplementary Fig. 13). The IR absorbance spectrum shows νN–H and νC=N bands with an intensity ratio between those of Co(tapp) and COF-366-Co (see below), in agreement with the coexistence of amine and imine moieties (Supplementary Figs. 14 and 15).

The oxidation state of cobalt in MOCOF-1 was analysed by magnetometry, SCXRD and 1H digestion NMR. Superconducting quantum interference device (SQUID) magnetometry shows Curie–Weiss paramagnetism with a Curie constant of C = 0.043(11) emu K mol−1, which corresponds to only 9% of a typical low-spin Co(II) porphyrin (0.49 emu K mol−1)55 and hence in line with a majority of diamagnetic Co(III) states (Supplementary Fig. 16). The SCXRD structure shows the axial Co–N distances to be 2.0417(14) Å, which corresponds to a typical Co(III) porphyrin–amine complex (2.060(3) Å)56 rather than its Co(II) analogue (2.436(2) Å)57. The 1H digestion NMR spectrum of a sample without supercritical CO2 (scCO2) treatment shows the presence of 0.70 equiv. of the Brønsted acid used, 2-chloro-4-nitrophenol, indicating a Co(III) state with a deprotonated 2-chloro-4-nitrophenolate counteranion (Supplementary Fig. 17). These results indicate that the majority of Co(II) centres undergo in situ oxidation into Co(III) during MOCOF-1 formation, although the synthesis was conducted under an inert atmosphere free of aerobic O2. We identified the oxidant as 4,4′-dimethoxytritylium cation formed from the deprotection by-product 4,4′-dimethoxytrityl alcohol and a Brønsted acid, which is detected by an ultraviolet–visible (UV–vis) absorbance peak at 507 nm (Supplementary Fig. 18). 4,4′-Dimethoxytritylium has a redox potential of −0.44 V (versus ferrocene) estimated from the literature58, which is positive enough to oxidize an octahedral Co(II) porphyrin complex (−0.52 V)59 (Supplementary Table 5). Although in situ oxidation of Co(tdpp) is used in this research for the ease of synthesis, MOCOF-1 forms also from a preoxidized precursor [CoIII(tdpp)(pba)2](OTf) (pba, p-bromoaniline; OTf, trifluoromethanesulfonate) (Supplementary Fig. 19). The cationic charge from the Co(III) state is balanced by either a counteranion or deprotonation of the coordinated aniline group, as discussed further below. Meanwhile, the origin of residual paramagnetism was analysed by electron paramagnetic resonance (EPR) spectroscopy. The EPR spectrum reveals the presence of two S = 1/2 species coupled with 59Co (I = 7/2) at 15 K (Supplementary Fig. 20 and Supplementary Table 6): the major species (80%, g = [1.973(2) 2.175(4) 2.240(2)], A = [80(3) 308(6) 101(3)] MHz) can be attributed to a bis(amine)-coordinated Co(II) porphyrin60 with a rhombic distortion, while the minor species (20%, g = [1.995(5) 1.995(5) 2.045(6)], A = [23(1) 23(1) 35(2)] MHz) can be assigned to a mono(amine)-coordinated superoxide Co(III) porphyrin60. Therefore, the paramagnetism found in the SQUID measurement (9% of all cobalt) can be explained by CoII(tapp) incorporated into the product without in situ oxidation by 4,4′-dimethoxytritylium. The majority of these Co(II) ions reside in the crystallographic bis(amine)-coordinated site of MOCOF-1, while the minority is suggested to be present at the surface or defect sites coordinated with oxygen as CoIII–O2·−.

Formation of MOCOF-1

The formation of MOCOF-1 is based on the dual usage of amino groups for coordination and imine linkages, in contrast to the previous heterolinkage frameworks based on orthogonal reactions, such as metal–carboxylate and amine–aldehyde pairs61. The dual role of amino groups enables complex structure evolution from simple building blocks, but the two types of reactions need to be balanced. Indeed, depending on the reaction conditions, three side phases, COF-366-Co, [CoIII(tapp)]nXn (X = counteranion) and CoII(tapp) besides MOCOF-1 are obtained from Co(tdpp) and terephthalaldehyde based on differences in the formation of each type of linkages (Fig. 3a).

Fig. 3: Formation of MOCOF-1 and side phases COF-366-Co, [CoIII(tapp)]nXn and CoII(tapp) by different extensions of metal–organic and covalent organic linkages.
figure 3

a, Structures based on the SCXRD result of MOCOF-1 and the PXRD result of COF-366-Co. b, Reaction scheme of coordination linkage formation promoted by 4,4′-dimethoxytritylium (Dmt+). c, Reaction scheme of imine linkage formation equilibrium dependent on H2O. X, variable counteranion.

Full size image

The side phases were prepared separately and characterized as references (Supplementary Table 7). First, COF-366-Co, a reported COF with a stacked-layer square lattice structure constructed by four-fold imine condensation of Co(tapp) with 2 equiv. of terephthalaldehyde62, was prepared according to our previous work63. This phase is discerned by its characteristic PXRD pattern, aggregated platelet morphology observed by scanning electron microscopy (SEM) and a terephthalaldehyde/Co(tapp) ratio of 2:1 in 1H digestion NMR (Supplementary Figs. 21–23). Second, [CoIII(tapp)]nXn, a coordination polymer without imine linkages, was prepared under the same conditions as MOCOF-1, except that terephthalaldehyde was not added. Its chemical structure is similar to the coordination network of MOCOF-1, as shown by the similarity of the 59Co ssNMR parameters (\(\delta_{\mathrm{cs}}^{\mathrm{iso}}=8{,}814\,{\mathrm{ppm}}\), \(|{\delta}_{\mathrm{cs}}^{zz}-{\delta}_{\mathrm{cs}}^{\mathrm{iso}}|=1{,}951\,{\mathrm{ppm}}\); Supplementary Figs. 24) and 13C and 15N ssNMR spectra except for the imine-related peaks (Supplementary Figs. 25 and 26). However, the PXRD pattern shows no peak (Supplementary Fig. 27), indicating that an amorphous coordination network forms in the absence of imine linkages. This phase can be discerned by the absence of terephthalaldehyde in 1H digestion NMR and the non-faceted morphology observed by SEM (Supplementary Fig. 28). Third, CoII(tapp) is formed as a solution under the same conditions as [CoIII(tapp)]nXn, except for lower acidity and lower polarity of the reaction medium. Under these conditions, the yield of solid products was very low and the filtrate was dark red, indicating CoII(tapp) was not oxidized and stayed dissolved in the solution as monomeric or oligomeric species because of weak Co(II)–amine coordination. The oxidation is hindered, presumably because the lower acidity and polarity suppress the formation of 4,4′-dimethoxytritylium from 4,4′-dimethoxytrityl alcohol, which is necessary for the oxidation of Co(II).

It is worth noting that the partial topologies of MOCOF-1, namely the lcv coordination net and the Π imine rod packing, are different from those of the corresponding side phases, that is, the amorphous coordination network of [CoIII(tapp)]nXn and the sql imine net of COF-366-Co. This difference implies that during MOCOF-1 formation, coordination and imine linkages are simultaneously extended, acting as templates for each other, and guiding the formation of the lcv and Π topologies, respectively.

The quantitative and selective synthesis of MOCOF-1 avoiding these side products was achieved by optimizing the conditions (Supplementary Table 7 and Supplementary Figs. 29 and 30). With our initial conditions, which resemble typical COF synthesis, COF-366-Co was obtained as the major product, indicating that imine condensation is preferred over cobalt–amine coordination and consumes all the amino groups. Using less terephthalaldehyde and more water promotes the formation of MOCOF-1 over COF-366-Co, by shifting the equilibrium of imine condensation to the amine side (Fig. 3c)64. However, the suppression of imine condensation also leads to the formation of [CoIII(tapp)]nXn and CoII(tapp). Alternatively, promoting Co(II) oxidation and coordination by higher acidity or solvent polarity also leads to the formation of MOCOF-1 (Fig. 3b). The balance between cobalt–amine coordination and imine condensation does not have to be exactly adjusted because the self-coordination of Co(tapp) consumes only two amino groups per molecule for coordination on two open coordination sites on cobalt and leaves two amino groups for imine condensation. Eventually, tuning the terephthalaldehyde amount, water amount, acidity and solvent polarity led to the selective and quantitative formation of MOCOF-1.

The phase purity of bulk microcrystalline samples is confirmed by PXRD, SEM, 1H digestion NMR, 59Co ssNMR and IR absorbance. PXRD confirms the absence of crystalline impurities such as COF-366-Co (Supplementary Fig. 10). SEM shows uniform polyhedral particles distinct from platelet particles of COF-366-Co and amorphous particles of [CoIII(tapp)]nXn (Supplementary Fig. 31). The 1H digestion NMR spectrum shows the terephthalaldehyde/Co(tapp) molar ratio to be 1:1.00, which agrees with the MOCOF-1 structure (Supplementary Fig. 32). The 59Co ssNMR spectrum shows a single Co(III) species, excluding other octahedral Co(III) porphyrin impurities (Supplementary Fig. 13).

The 4,4′-dimethoxytrityl protection group of Co(tdpp) is important for the synthesis of MOCOF-1 because of the aforementioned oxidation of Co(II) by 4,4′-dimethoxytritylium. The deprotection of 4,4′-dimethoxytritylamines readily proceeds in the presence of water and acid, releasing Co(tapp) and 4,4′-dimethoxytrityl alcohol in the initial stage of the reaction63. The removal of 4,4′-dimethoxytrityl groups during MOCOF-1 synthesis is confirmed by 1H digestion NMR and IR (Supplementary Figs. 33 and 34). The unprotected precursor Co(tapp) also gives MOCOF-1 due to the partial oxidation of Co(tapp) under air63, but the product contains COF-366-Co because the oxidation is incomplete (Supplementary Fig. 35).

Properties of MOCOF-1

The crystal size of MOCOF-1 can reach 100 µm within 5 days, as confirmed by SEM and SCXRD (Fig. 4a and Supplementary Fig. 36). Such a large crystal size in a relatively short reaction time has rarely been achieved with imine COFs17. Indeed, when a mixture of MOCOF-1 and COF-366-Co is obtained under the aforementioned unoptimized conditions, MOCOF-1 formed ~4 µm faceted crystals whereas COF-366-Co formed ~50 nm crystallites, achieving two orders of magnitude larger crystal sizes from the same building blocks (Supplementary Fig. 37). This crystal size difference suggests that extended covalent organic linkages can be efficiently arranged into a crystalline structure with the support of highly reversible metal–organic linkages.

Fig. 4: Properties of MOCOF-1.
figure 4

a, SEM image of single crystals. b, N2 sorption isotherms of a pristine sample and samples soaked in water, 10 mM NaOH aq. or pyridine for 17 h at r.t. c, N2 sorption isotherm and BET surface area (77 K) of a pristine sample compared with [Co(tapp)]nXn and COF-366-Co. d, 1H digestion NMR spectra (20 vol% (35 wt% DCl/D2O)/DMSO-d6, 400 MHz) before and after scCO2 treatment. The numbers of acid molecules are calculated by integration. e, Reaction scheme of desorption and adsorption of an acid molecule HX. f, 1H digestion NMR spectra after soaking in a 1,4-dioxane solution of 3,5-dinitrobenzoic acid or 2,4-dinitrophenol.

Full size image

The chemical stability of MOCOF-1 was evaluated by soaking MOCOF-1 in several types of media for 17 h at room temperature. The N2 sorption isotherms and corresponding Brunauer–Emmett–Teller (BET) surface areas, PXRD patterns and 1H digestion NMR spectra of the treated samples are similar to those of a pristine sample, indicating that the crystallinity and composition are mostly retained in water, 10 mM NaOH aq. and pyridine (Fig. 4b and Supplementary Figs. 38–42). The N2 uptakes and the corresponding BET surface areas even show slight increases, which are attributable to the extraction of unknown impurities or guest molecules. These results suggest that MOCOF-1 is stable in coordinating media, in contrast to a typical cobalt–amine MOF which rapidly decomposes on exposure to moisture via the exchange of amine ligands65. However, 10 mM HCl aq. caused a drastic decrease in crystallinity as evidenced by PXRD (Fig. 4b and Supplementary Fig. 43) and a partial loss of terephthalaldehyde shown by 1H digestion NMR (Supplementary Fig. 44), which indicates disintegration induced by acid-catalysed hydrolysis of imine linkages. The stability in coordinating media and the instability against imine hydrolysis indicate that the structural stability of MOCOF-1 is both supported and limited by the strength of imine linkages, which act as a scaffold for the weaker cobalt–amine linkages.

The thermal stability of MOCOF-1 was evaluated by thermogravimetric analysis (TGA) and variable-temperature PXRD. The TGA curve in air shows decomposition by combustion around 350 °C, similar to COF-366-Co (ref. 62) (Supplementary Fig. 45). Although no notable mass loss is observed at this temperature, PXRD patterns in air show complete amorphization over 100 °C (Supplementary Fig. 46), indicating pore collapse at this temperature. Under an argon atmosphere, the pore collapse was slower but started at a lower temperature (Supplementary Figs. 45 and 46). It is concluded that MOCOF-1 is thermally stable up to 100 °C under air.

The gas sorption ability of MOCOF-1 was analysed by N2 and CO2 sorption after activation by scCO2 and vacuum. The N2 sorption isotherm shows a BET surface area of 2,836 m2 g−1 (Fig. 4c and Supplementary Figs. 47–49), a pore size distribution peak at 1.9 nm and a total pore volume of 1.24 cm3 g−1 (Supplementary Fig. 50). The pore size distribution agrees with the aforementioned result of SCXRD analysis (Supplementary Fig. 9). The BET surface area is 1.4 times larger than that of COF-366-Co (1,998 m2 g−1)63, reflecting the 3D topology of MOCOF-1 by coordination linkages without π stacking. It is also 17 times larger than that of [CoIII(tapp)n]Xn (169 m2 g−1, Supplementary Figs. 51–53), indicating that the separation of Co(tapp) units by imine linkages in MOCOF-1 is important for porosity. Likewise, the structural pore volume of MOCOF-1 (1.03 cm3 g−1 for pore widths <3 nm) is larger than those of COF-366-Co (0.76 cm3 g−1) and [CoIII(tapp)]nXn (0 cm3 g−1) (Supplementary Fig. 54). Both the BET surface area and the total pore volume of MOCOF-1 exceed those of previously reported frameworks with hierarchical metal–organic and covalent organic linkages42. MOCOF-1 also shows a CO2 uptake capacity of 44.1 cm3 g−1 at 298 K, 1 bar, with a relatively low adsorption enthalpy of −19.7(6) kJ mol−1 (Supplementary Figs. 55–57). These results suggest that MOCOF-1 is a highly porous material capable of gas sorption.

The adsorption and desorption of acid molecules on MOCOF-1 were studied by 1H digestion NMR. The NMR spectra show a decrease in the number of 2-chloro-4-nitrophenolate counteranions from 0.70 equiv. to 0.15 equiv. per Co(III) during scCO2 treatment (Fig. 4d and Supplementary Fig. 17). This result suggests that 2-chloro-4-nitrophenol is desorbed from MOCOF-1 as free acid in scCO2 flow, forming an anilido structure similar to a reported anilido-coordinated zinc MOF66 (Fig. 4e). This process depends on the acidity of the desorbing acid: while weak Brønsted acids such as 2-chloro-4-nitrophenol or p-nitrophenol are desorbed from MOCOF-1, stronger acids such as 3,5-dinitrobenzoic acid and 2,4-dinitrophenol remain after scCO2 treatment (Supplementary Figs. 58 and 59). Although scCO2 is used, no carbamate or (bi)carbonate structure is observed in deprotonated MOCOF-1 by 13C ssNMR (no carbonyl signal beyond 158 ppm; Supplementary Fig. 11), IR (no C=O stretch band in the region of 1,800–1,632 cm−1; Supplementary Fig. 14) or SCXRD. Likewise, the chemisorption of CO2 gas is not observed, reflected in the aforementioned low adsorption enthalpy. Prompted by the desorption results, we tested the adsorption of acid molecules from solutions by the anilido form of MOCOF-1 by soaking it in a 1,4-dioxane solution of 3,5-dinitrobenzoic acid or 2,4-dinitrophenol. The 1H digestion NMR spectra after soaking evidence incorporation of the acid molecules in a ratio of ∼1:1 to Co(III), indicating that MOCOF-1 can adsorb acid molecules at the Brønsted-basic Co(III)–anilido sites (Fig. 4f and Supplementary Fig. 60).

The optical properties of MOCOF-1 were investigated by UV–vis absorption, electronic circular dichroism (ECD) and second harmonic generation (SHG). The UV–vis absorption spectrum of MOCOF-1 shows three major absorption bands which can be assigned to the Soret and Q bands of the cobalt–porphyrin chromophore (Supplementary Fig. 61). The ECD spectrum of bulk nanocrystalline powder exhibits negligible signals around the absorption bands, indicating a racemic conglomerate of enantiomorphic crystals (Supplementary Fig. 61). Nonetheless, efficient SHG is observed from individual submicrometre particles, supporting the presence of non-centrosymmetric, enantioenriched particles (Supplementary Fig. 62). This result agrees with the aforementioned SCXRD analysis of the large-crystal sample, which showed the enantiopurity of a single crystal.

Synthesis of isoreticular analogues

To gauge the tunability of the MOCOF-1 structure and to demonstrate the generality of our synthetic approach, we synthesized isoreticular analogues of the archetypical structure MOCOF-1. MOCOF-1-R2 with various functional groups (R = OH, OMe, F, Br) were obtained using difunctionalized terephthalaldehyde derivatives (Fig. 5a). The formation of isoreticular structures is confirmed by PXRD, which shows that all compounds have similar patterns that can be indexed with similar unit cells and the same cubic I213 space group (Fig. 5b and Supplementary Figs. 63–66). Furthermore, expanded analogues MOCOF-2 and MOCOF-3 were obtained using longer dialdehydes, 2,6-naphthalenedicarboxyaldehyde and thieno[3,2-b]thiophene-2,5-dicarbaldehyde (Fig. 5a), giving a similar PXRD pattern with slightly larger unit cells in the same space group (Fig. 5b and Supplementary Figs. 67 and 68). The purity of the products was confirmed by PXRD, SEM and 1H digestion NMR, with results similar to the case of MOCOF-1 (Supplementary Figs. 63–80).

Fig. 5: Isoreticular analogues of MOCOF-1: MOCOF-1-R2 (R = OH, OMe, F, Br), MOCOF-2 and MOCOF-3.
figure 5

a, Chemical structures. b, PXRD patterns (Cu Kα1). c, SCXRD structure of MOCOF-2 (100 K). Hydrogen atoms and solvent molecules are omitted for clarity. d, Comparison of the molecular geometries of the Co(tapp) nodes in MOCOF-1 and MOCOF-2 based on SCXRD. e, Pore size distributions and their maximum positions of MOCOF-1 and MOCOF-2 based on N2 sorption (77 K).

Full size image

The expanded crystal structure of MOCOF-2 was determined by SCXRD analysis (Fig. 5c, Supplementary Figs. 3, 5, 7 and 81 and Supplementary Table 3). Notably, the lengths of the coordination linkages are unchanged compared to MOCOF-1 (Co–Co distance, 10.3 Å → 10.1 Å), while the organic linkages are extended (25.2 Å → 27.2 Å). These geometric changes are due to the rotation of cobalt–amine coordination bonds and the slight deformation of the meso carbon atoms of porphyrin rings (Fig. 5e).

The isoreticular expansion allows the fine-tuning of pore sizes. N2 sorption analysis shows that the pore size distribution of MOCOF-2 (maximum at 1.98 nm) is slightly shifted to larger sizes compared with MOCOF-1 (maximum at 1.91 nm) (Fig. 5f and Supplementary Figs. 82 and 83), while the BET surface area (2,818 m2 g−1; Supplementary Figs. 84 and 85) stays similar to that of MOCOF-1.

These results suggest that structural modulation by isoreticular chemistry similar to that in conventional MOFs applies to MOCOFs, thus pointing towards versatile design and synthesis strategies in this class of reticular materials.

Conclusions

In this work, we report MOCOFs, a class of reticular materials based on the fusion of MOFs and COFs, featuring double extension of metal–organic and covalent organic linkages. The first example, MOCOF-1, is constructed from a Co(III) tetra-aminoporphyrin node and a dialdehyde linker, which form extended cobalt–amine coordination linkages and imine linkages in the chiral lvz net topology. The synthesis of MOCOF-1 is based on the double usage of amino groups for metal coordination and imine condensation, which can be adjusted by Co(II) oxidation and water content, respectively. MOCOF-1 shows higher crystallinity than its COF analogue and topological chirality due to the presence of metal–organic linkages. Meanwhile, MOCOF-1 exhibits higher stability against coordinating media compared with MOFs with similar motifs, due to the presence of covalent organic linkages. These properties suggest that MOCOFs can inherit the merits of MOFs and COFs via the extension of each type of linkage throughout a framework. MOCOF-1 also has a high BET surface area of 2,836 m2 g−1 and is capable of adsorbing acid molecules after activation. Furthermore, its structure can be modulated by isoreticular chemistry to rationally introduce different functional groups and pore sizes. Therefore, MOCOFs show promise as a class of chemically tunable porous materials. MOCOF-1 not only represents the first representative of MOCOFs constructed by double linkage extension, but also showcases that MOF and COF chemistry can be synergistically combined to achieve tailored properties in a yet uncharted territory of reticular materials.

Methods

Gas sorption

Gas sorption isotherms were recorded on a Quantachrome Instruments Autosorb iQ 3 sorptometer at 77 K (N2) and given temperatures (CO2). The samples were activated under a high vacuum at 30 °C for 12 h before measurements. For CO2 sorption, the samples were additionally activated by one adsorption–desorption cycle of CO2 at 273 K.

BET analyses were conducted using a program written in our group by A. M. Pütz (https://github.com/AlexanderPuetz) based on the software BETSI by Osterrieth and Fairen-Jimenez67. The fitting ranges were selected to meet all four Rouquerol criteria, to contain the maximum (knee) in the Rouquerol plot, and to minimize the error under the fourth Rouquerol criterium, that is, between the experimental and theoretical partial monolayer loading.

The pore size distributions, cumulative pore volumes and total pore volumes were calculated with a kernel of N2 at 77 K on carbon (quenched solid density functional theory, adsorption branch) using relative pressures from 10−7 to 1 in AsiQwin software v.3.01.

NMR spectroscopy

Digestion of samples was conducted with 20 vol% (35 wt% DCl/D2O)/DMSO-d6 for 2 days at room temperature unless otherwise noted. Solution 1H NMR spectra were recorded on a JEOL ECZ 400S 400 MHz spectrometer at 25 °C. Chemical shifts (δ) were referenced relative to TMS (0 ppm). The spectra were processed using the software iNMR v.6.4.5 with exponential weighting of 0.5 Hz, Gaussian weighting of 0.5 Hz, manual phase correction and automatic baseline correction with segmented straight lines.

PXRD

PXRD patterns were recorded on a Stoe Stadi P diffractometer with Cu or Co Kα1 radiation from a primary Ge(111)-Johann-type monochromator and a Mythen detector (Dectris) in Debye–Scherrer geometry with sample spinning at room temperature. The samples were either sealed in 1.0-mm-diameter Hilgenberg glass no. 14 capillaries or compressed into a 2.5-mm-diameter disc and sealed between two pieces of Kapton tape.

SEM

SEM images were recorded on a Zeiss Merlin microscope under an electron high-tension voltage of 1.5 kV. Before measurements, samples were dispersed on carbon tape with a spatula and coated using a Leica EM ACE600 sputter-coater with 6.0 nm of iridium from a 30° angle unless otherwise noted.

Syntheses

scCO2

scCO2 rinsing and drying were performed with a Leica EM CPD300 critical point dryer. The sample container was initially half-filled with methanol at 13 °C. After CO2 infusion, the fluid was stirred for 15 min. CO2 was removed at 40 °C.

Co(tdpp)

This compound was synthesized similarly to a trityl analogue63 with some modifications. A 500 ml two-neck flask was equipped with a magnetic stir bar, a reflux condenser, a rubber septum and a paraffin bubbler, and was dried under vacuum with a heat gun. The flask was charged with TAPP (801 mg, 1,187 µmol, 1 equiv.), 4,4′-dimethoxytrityl chloride (2,051 mg, 5.11 equiv.), Et3N (1.98 ml, 12.0 equiv.) and anhydrous tetrahydrofuran (192 ml) under an argon atmosphere. The mixture was stirred for 1 h. To the mixture were added Et3N (5.78 ml, 35.0 equiv.) and CoCl2 (1,539 mg, 10.0 equiv.). The set-up was briefly purged with argon and the mixture was bubbled with argon for 30 min. The mixture was heated to reflux under 1.15 atm for 2.5 h, cooled to room temperature, and chromatographed through a pad of neutral Al2O3 (121 g, 4.2 cm (height) × 6.1 cm (diameter)) with tetrahydrofuran until the eluate became pale pink. The eluate was dried by rotary evaporation and then under vacuum. The solids were dissolved in CHCl3 (100 ml). Reprecipitation was induced with n-hexane (500 ml). The precipitate was collected by suction filtration, rinsed with CHCl3/n-hexane (1:5, v/v) and n-hexane, and dried under vacuum to yield the title compound as red-brown solids (61% yield). 1H NMR (CDCl3, 400 MHz): δ (ppm) = 15.80 (br, 8H), 12.79 (br, 8H), 8.94 (br, 8H), 8.60 (d, J = 7.7 Hz, 8H), 8.50 (d, J = 8.4 Hz, 16H), 8.10 (t, J = 7.5 Hz, 8H), 7.90 (t, J = 7.2 Hz, 4H), 7.64 (d, J = 8.3 Hz, 16H), 6.79 (s, 4H), 4.30 (s, 24H). Electrospray ionization mass spectrometry: m/z = 1,940.8 (M+, calcd. 1,940.7). Inductively coupled plasma optical emission spectroscopy: Co 2.787(3)% (w/w, calcd. 3.04%).

MOCOF-1 (microcrystals)

An 8 ml Schlenk bomb was charged with Co(tdpp) (25.8 mg, 13.3 µmol, 1 equiv.), terephthalaldehyde (3.57 mg, 26.6 µmol, 2.00 equiv.), water (21.6 µl, 2.39 mmol, 90.0 equiv.) and m-dinitrobenzene/nitrobenzene/1,4-dioxane (1:3.2:9.5 v/v/v, 667 µl). The mixture was sonicated until the solids were dissolved. The solution was frozen with a liquid N2 bath for 3 min. To the frozen mixture was added 2-chloro-4-nitrophenol (138 mg, 798 µmol, 60.0 equiv.). The inner atmosphere was evacuated to <0.01 mbar for 1 min, the tap was closed and the mixture was thawed with a water bath at room temperature. This freeze–pump–thaw cycle was repeated two more times. The thawed mixture was manually stirred, sonicated, heated at 120 °C in a convection oven for 3 day and cooled to room temperature. To the mixture was added dimethylformamide (DMF, 1 ml). The solids were collected onto a filter paper by suction filtration and rinsed with DMF, CHCl3 and methanol, maintaining a wet state. The filter paper containing the solids was soaked in methanol and treated with scCO2, giving MOCOF-1 as a purple solid (quantitative yield).

MOCOF-1 (large crystals)

Large single crystals were grown in conditions modified from the microcrystal preparation, replacing water 90.0 equiv. with 180 equiv., 2-chloro-4-nitrophenol with p-nitrophenol (30.0 equiv.), m-dinitrobenzene/nitrobenzene/1,4-dioxane with nitrobenzene/1,4-dioxane (1:3 v/v, 667 µl), and 3 days of heating with 5 days. p-Nitrophenol was added to the solution immediately before freezing. A dark purple solid is obtained in 18% yield. PXRD and SEM analyses indicated large single crystals of MOCOF-1 up to 100 µm along with COF-366-Co and [CoIII(tapp)]nXn impurities.

MOCOF-1-(OH)2

This compound was synthesized in the same manner as MOCOF-1, replacing terephthalaldehyde with 2,5-dihydroxyterephthalaldehyde (1.00 equiv.) and water 90.0 equiv. with 135 equiv., to give a dark violet solid (quantitative yield).

MOCOF-1-(OMe)2

This compound was synthesized in the same manner as MOCOF-1, replacing terephthalaldehyde with 2,5-dimethoxyterephthalaldehyde (2.00 equiv.), water 90.0 equiv. with 170 equiv., 2-chloro-4-nitrophenol with p-nitrophenol (30.0 equiv.), and 3 days of heating with 6 days, to obtain a dark violet solid (76% yield).

MOCOF-1-F2

This compound was synthesized in the same manner as MOCOF-1, replacing terephthalaldehyde with 2,5-difluoroterephthalaldehyde (1.00 equiv.), to obtain a dark purple solid (35% yield).

MOCOF-1-Br2

This compound was synthesized in the same manner as MOCOF-1, replacing terephthalaldehyde with 2,5-dibromoterephthalaldehyde (2.00 equiv.) and DMF (1 ml) with methanol (667 µl) and 1 M NaCl aq. (333 µl), skipping the rinse with DMF and CHCl3, to obtain a dark purple solid (quantitative yield).

MOCOF-2 (microcrystals)

This compound was synthesized in the same manner as MOCOF-1, replacing terephthalaldehyde with 2,6-naphthalenedicarboxyaldehyde (2.00 equiv.), to obtain a dark purple solid (quantitative yield).

MOCOF-2 (large crystals)

Large single crystals were grown in conditions modified from the microcrystal preparation, replacing water 90.0 equiv. with 180 equiv., 2-chloro-4-nitrophenol with p-nitrophenol (30.0 equiv.), m-dinitrobenzene/nitrobenzene/1,4-dioxane with nitrobenzene/1,4-dioxane (1:3 v/v, 667 µl), and 3 days of heating with 13 days. Dark purple solids are obtained in 18% yield. PXRD and SEM analyses indicated large single crystals of MOCOF-2 up to 80 µm along with COF and [CoIII(tapp)]nXn impurities.

MOCOF-3

This compound was synthesized in the same manner as MOCOF-1, replacing terephthalaldehyde with thieno[3,2-b]thiophene-2,5-dicarbaldehyde (1.00 equiv.), to obtain a dark purple solid (quantitative yield).

Other experiments

Stability test of MOCOF-1

A 6 ml vial was charged with MOCOF-1 (∼12 mg) and medium to be tested (5 ml of pyridine, 10 mM NaOH aq., water, 1 mM HCl aq., or nothing) and left for 17 h. The solids were collected onto a filter paper by suction filtration and rinsed with methanol, maintaining a wet state. The filter paper containing the solids was soaked in methanol, treated with scCO2 and subjected to analysis.

Acid adsorption from solution by MOCOF-1

A 2 ml tube was charged with MOCOF-1 (1 µmol, 0.83 mg), a stir bar and a 1,4-dioxane solution of 3,5-dinitrobenzoic acid or 2,4-dinitrophenol (1 equiv., 1,000 µl). The mixture was sonicated briefly, stirred at room temperature for 24 days and centrifuged. The supernatant was removed, and the sediment was rinsed with 1,4-dioxane (1 ml) twice, dried under vacuum and subjected to 1H digestion NMR analysis.