Introduction

It is a central premise of reticular chemistry that structures of the highest symmetry (i.e., minimal transitivity) are the most likely to result from linking simple, symmetric building units1,2,3,4. This notion holds true for metal-organic frameworks (MOFs)5 whose structures are simple in that they contain one kind of multi-metallic unit and one kind of organic linker. However, the premise of minimal transitivity is not as widely observed for more complex MOFs whose compositions extend beyond binary building units6,7,8,9,10. This is because the ratios of linkers incorporated within such MOF structures and their respective length ratios must be a perfect match. Indeed, this becomes increasingly difficult to control when moving toward multiple building units. Given the desire to increase the complexity of MOFs (i.e., increase the number of building units)11, the following questions arise: Can the transitivity principle be effectively extended? If so, what are these structures and what are their properties? For the first two questions, we know that complex MOFs constructed from multiple building units rarely follow the minimal transitivity principle (MOF-205 is a notable exception).

In this contribution, we describe a complex ternary MOF system, termed MOF-907, which is constructed from trigonal prismatic Fe3O(–CO2)6 clusters with triangular 4,4′,4”-benzene-1,3,5-triyl-tris(benzoic acid) (H3BTB) and linear 2,6-naphthalenedicarboxylic acid (H2NDC) linkers. For this system, the ideal linker length ratio and linker incorporation is met for achieving a minimal transitive nha net. In addressing the third question, we elucidate the relationship between high symmetry and the favorable geometry of the pore structure within such systems, leading to this material′s enhanced catalytic properties.

Results

Single crystal X-ray diffraction studies

Single crystals of MOF-907 (Fig. 1 and Supplementary Figs. 1 and 2) were obtained upon heating a reaction mixture containing H3BTB, H2NDC, and Fe(NO3)3·9H2O in N,N-dimethylformamide (DMF) at 120 °C for 24 h. Acetic acid, acting as a modulator to adjust the acidity, was also added to the reaction mixture (Supplementary Methods). Single crystal X-ray diffraction analysis revealed that MOF-907 crystallized in the body-centered cubic space group, Im-3m (No. 229), with a lattice parameter of a = 44.4781(14) Å (Supplementary Data 1 and Supplementary Table 1).

Fig. 1
figure 1

Structural representation of MOF-907. (a, b) Trigonal prismatic Fe3O(–CO2)6 clusters linked together by triangular 4,4′,4”-benzene-1,3,5-triyl-tris(benzoate) (BTB3−) and linear 2,6-naphthalenedicarboxylate (NDC2−) linkers, corresponding to 6-connected, 3-connected, and 2-connected (-c) points of extension, respectively, yield a MOF with a previously unseen net, nha (depicted in augmented (-a) form). (c) The single crystal structure of MOF-907 is presented. Atom colors: Fe: blue polyhedra. C: black, and O: red. All H atoms are omitted for clarity. The yellow and orange spheres indicate the free space in the cages

Full size image

The structure of MOF-907 is made up of four hierarchical levels. These levels can be reduced down to the assembly of a primary cage comprised of six trigonal prismatic Fe3O(–CO2)6 clusters acting as vertices to form a distorted octahedral-shaped cage with a 12 Å diameter (Fig. 2a). In this primary cage, two tritopic BTB3− linkers connect the vertices along opposing equatorial edges with the remaining coordination occurring across the faces to both of the axial vertices. Completing the primary cage are six NDC2− linkers, two of which fill in the remaining equatorial edges and the other four connecting the equatorial vertices with those of the axial sites (on opposing axial edges from the BTB3− axial face). These primary cages then assemble into the second level of the structure, in which each trigonal prismatic vertex of the primary octahedral cage is shared with another primary octahedral cage (Fig. 2a). The second level is created upon assembling eight primary cages into a cuboctahedral-shaped cage with a diameter of 19 Å (cuboctahedral cage 1, Fig. 2a). This gives rise to a square window with a diameter of ~13 Å. A third level is then built up from six cuboctahedral cage 1 units of 19 Å in diameter (cuboctahedral cage 2, Fig. 2a). Finally, the overall structure is realized through combining all three levels: one cuboctahedral cage 2, six cuboctahedral cage 1, and 24 total primary cages generating two types of 1-D interconnected pore channels due to the cubic lattice structure.

Fig. 2
figure 2

Conceptual basis for the realization of MOF-907 adopting the minimal transitive nha net. This structure is built up from four hierarchical levels. a At the fundamental level, a primary distorted octahedral cage consisting of six Fe3O(–CO2)6 clusters connected by two BTB3− and six NDC2− linkers, is formed. A cuboctahedral cage 1 then assemblies from eight vertex-sharing primary cages. This leads to the assembly of six of these cuboctahedral cages to form another cuboctahedral cage (cage 2). The overall structure then assembles through a combination of one cuboctahedral cage 2, six cuboctahedral cage 1, and 24 total primary cages. b The natural tiling of the nha topology is shown for each of the cages. Atom colors: Fe: blue polyhedral, C: black, and O: red. BTB3− linkers are shown in filled pink to aid in distinguishing them from the NDC2− linkers (filled green). H atoms are omitted for clarity

Full size image

Topological analysis of MOF-907

Topological deconstruction of MOF-907 was performed using ToposPro software12. For topological analysis, the Fe3O(–CO2)6 secondary building unit (SBU) and BTB3− and NDC2− linkers were first reduced into simplified nodes, resulting in 6-connected and 3-connected points of extension for Fe3O(–CO2)6 and BTB3−, respectively. It is noted that NDC was represented as an edge when connecting the 6-c point of extension for the Fe3O(–CO2)6 nodes. Through this deconstruction, we uncovered the first case of a net, termed nha, with minimal transitivity based on trigonal prismatic SBUs linked together by triangular and linear units (Supplementary Fig. 3). Furthermore, this is only the second complex MOF structure, constructed from mixed linkers, to have adopted a minimal transitive net7. The transitivity of the nha tiling (face-to-face packing of cages) is described in terms of pqrs (where p is the number of vertices, q is the number of edges, r is the number of faces, and s is the type of tiles). For nha, pqrs = 2244. Further analysis of the tiling reveals three cages described as 4[46] + [86] + 9[44.82], respectively (where the symbol […mn…] denotes that there are n faces with m-membered rings) (Fig. 2b).

When considering the reason for realizing the nha net, it was evident that the ratio of the tritopic:ditopic linker lengths (LT:LD; where LT = length of the tritopic linker and LD = length of the ditopic linker) played a critical role. Minimal transitive nets are achieved with the highest frequency for MOFs constructed with only one linker because they are the most symmetric. However, for the case of MOFs constructed from two (or more) linkers, the relative ratio of linker lengths must be a perfect fit to achieve maximal symmetry. Based on crystal structure evaluation and analyses, we observed that the ratio of linker lengths for MOF-907 (LT:LD = 0.90) achieves this in contrast to other MOFs with similar constituent building units: MIL-142A (LT:LD = 1.29), PCN-280 (LT:LD = 0.71), and PCN-285 (LT:LD = 0.59) (Fig. 3a and Supplementary Fig. 4). Upon further investigation into the reasoning behind MOF-907 adopting a minimal transitive net, we identified a second structural phenomenon: the ratio of the tritopic:ditopic linkers (NT:ND; where NT = the number of tritopic linkers and ND = the number of ditopic linkers) connected to the Fe3O SBU was variable. In fact, we found that there were three possible ratios when considering this structural phenomenon in Fe-based MOFs with the same constituent building units: (i) 4 tritopic and 2 ditopic linkers (NT:ND = 2; ex. PCN-285, PCN-280, and MIL-142A-E); (ii) 3 tritopic and 3 ditopic linkers (NT:ND = 1); and (iii) 2 tritopic and 4 ditopic linkers (NT:ND = 0.5; ex. MOF-907) (Fig. 3). As such, we found that MOF-907 had perfectly-sized tritopic and ditopic linkers (LT:LD = 0.90) and combination thereof incorporated within the SBU (NT:ND = 0.5), which led to the realization of the highest symmetry minimal transitive network for MOFs comprised of these building units (Fig. 3). It is noted that an optimal input ratio of starting materials (molar ratio of BTB:NDC = 0.5–0.75) together with an exact amount of modulator (AcOH, 0.2 mL) were required to produce the resulting structure during crystal formation.

Fig. 3
figure 3

Principles for achieving minimal transitivity in MOF-907. a Two structural features played pivotal roles in producing MOF-907, which adopts a highly symmetric, minimal transitive net: The ratio of tritopic:ditopic linker lengths (LT:LD); and the proportion of tritopic to ditopic linker incorporated within the SBU (NT:ND). b These structural features are guiding principles for achieving the highest symmetric material based on Fe3O SBUs and mixed tritopic and ditopic linkers

Full size image

After solvent-exchange and full activation, the activated sample was measured powder X-ray diffraction (PXRD) which is in agreement with the as-synthesized and simulated pattern (Supplementary Fig. 5). The formulation of the chemical structure was achieved through a combination of digested 1H nuclear magnetic resonance spectroscopy (NMR), elemental microanalysis (EA), thermal gravimetric analysis (TGA), Mössbauer spectroscopy, and Fourier transform infrared spectroscopy (FT-IR) measurements (Supplementary Figs. 6–8). The digested 1H NMR spectrum confirmed the ratio of BTB:NDC to be 1:3 (Supplementary Fig. 11). Following this, EA was employed to resolve MOF-907′s chemical formula as FeIII3O(OH)(BTB)0.67(NDC)2. TGA was then utilized to provide further support for this formulation. After heating activated MOF-907 in air up to 600 °C, the observed residual mass (22.3%) was consistent with the expected mass of Fe2O3 (23.4%) after framework decomposition. The presence of one capping OH− ligand (necessary for charge balancing) was proven by FT-IR, where an O-H stretching frequency vibration at 520 cm–1 was clearly observed (Supplementary Fig. 7)13, 14. To ensure the oxidation state of Fe in MOF-907 was correct, Mössbauer spectroscopy was employed to confirm the presence of only FeIII within the structure (Supplementary Fig. 8). Finally, the permanent porosity of MOF-907 was proven by N2 adsorption isotherms at 77 K. MOF-907 exhibited a Brunauer−Emmett−Teller (Langmuir) surface area of 1600 (1800) m2 g–1 (Supplementary Figs. 9 and 10), which is slightly less than the theoretical geometrical surface area determined from the crystal structure (2000 m2 g–1).

Catalytic properties of MOF-907

Motivated by the high porosity and appropriately-sized 1-D interconnected pore channels of MOF-907 in conjunction with the presence of strong Lewis acidity via potential coordinatively unsaturated metal sites in the Fe3O SBUs15,16,17, we sought to assess whether such a structure would have, as a natural outcome, a high performance for catalytic applications. As such, MOF-907 was employed as a catalyst for the microwave-assisted radical polymerization of methyl methacrylate (MMA; Fig. 4)–a model reaction that requires a catalyst with strong Lewis acidity18, 19. Initial attention was placed on optimizing the catalysis conditions, including the amount of MOF-907 catalyst, co-initiator, organic solvent, and reaction time. As such, two different co-initiators, ethyl α-bromophenylacetate (EBPA) and ethyl α-bromoisobutyrate (EBiB), were tested (Supplementary Fig. 12). For the reactions using EBiB, polyMMA was produced in a yield of 68% using MOF-907 catalyst (0.06 mol% loading), with an average Mn = 15,550 g mol–1 and a polydispersity index (PDI) of 1.31. When EBPA was used as the co-initiator, polyMMA formed in lower yield (43%) and with an overall lower Mn (14,230 g mol–1) and higher PDI (1.43) (Table 1 and Supplementary Table 2 and Supplementary Fig. 13).

Fig. 4
figure 4

Polymerization reaction. General polymerization reaction of MMA

Full size image
Table 1 Microwave-assisted radical polymerization of methyl methacrylate
Full size table

Control experiments in which the MOF-907 catalyst or microwave irradiation was absent demonstrated no observable product formation. Without microwave irradiation, radical generation and propagation cannot occur, which would prohibit the polymerization process. The effect of monomer concentration on the quality of product was subsequently investigated by reducing the amount of solvent used in the reaction. Specifically, halving the amount of DMF added to the mixture, in effect, doubled the monomer concentration, increased the Mn value from 15,550 to 20,680 g mol–1, and lowered the PDI to 1.23 with a nearly quantitative yield of 98% (Table 1 and Supplementary Table 3).

In order to ensure that the catalytic activity was derived from the Lewis acidic SBUs in MOF-907, we employed other MOF catalysts, specifically Ni-VNU-74(II)20 and HKUST-121, which contain SBUs with open Lewis acidic active sites. As expected, both Ni-VNU-74(II) and HKUST-1 were active; however, MOF-907 outperformed both of these materials under identical reaction conditions in producing high Mn polyMMA product, in higher yields, and with a lower PDI (Table 1). This is attributed to the strong Lewis acidic sites of Fe-based MOFs as compared to Ni-based or Cu-based MOFs15,16,17. Additionally, the pore size of MOF-907 is more suitable for the polymerization reaction than those of Ni-VNU-74(II) (23.6 Å 1-D mesopores) and HKUST-1 (8 Å 3-D arrangement pores). It is noted that the MIL-142 series would be ideal comparative catalysts due to their identical building unit composition;8 however, after exhaustive attempts, we were unable to reproduce these materials experimentally (Note: no crystallographic information files were reported or found in the Cambridge Crystallographic Data Center (CCDC) for this series). Our attention then turned to comparing the performance of MOF-907 with PCN-2809 and MIL-101(Fe)22–both of which are constructed from the same Fe3O SBU active site.

As shown in Table 1, polyMMA products with much lower Mn (15,080 and 13,860 g mol–1) and higher PDI (1.35 and 1.32) values were obtained when using MIL-101(Fe) and PCN-280, respectively. Furthermore, MOF-907 outperformed these catalysts in terms of yield (98% v. 20% and 35% for MOF-907 v. MIL-101(Fe) and PCN-280, respectively). Since identical molar amounts of catalyst were used, the number of FeIII active sites in MOF-907 is equal to MIL-101(Fe) and PCN-28023, 24. Hence, we concluded that structural features, imparted by the topological network of MOF-907, played a pivotal role in enhancing the catalytic performance. Specifically, the monomers (substrates) have to be well-distributed throughout the pores in order to fully interact with the Fe3O SBUs. If this occurs, then one can take full advantage of the intrinsic catalytic active sites of the MOF material25. Indeed, MOF-907 possesses 1-D interconnected channels, 19 Å in size, with an accessible pore window of 13 Å. These appropriately-sized 1-D channels are better suited to disperse the substrates for polymerization than the small 3-D pore windows found in PCN-280 (12 Å) and MIL-101(Fe) (14 Å).

The scanning electron microscope (SEM) image of bulk phase (large-scale synthesis) of MOF-907 was also measured to confirm the phase purity and uniformity of the sample (Supplementary Fig. 14). To ensure that the catalysis is in fact taking place within the internal pore environment as opposed to the external surface, we then carried out control experiments using MOF-907 crystals of varying size. The yield of polyMMA experienced little change when using 29 μm-sized MOF-907 particles (98% yield) or 92 μm-sized MOF-907 particles (91% yield) (Supplementary Fig. 15). This is an important control experiment as smaller-sized crystals typically offer higher catalytic activity if the catalysis is taking place on the external surface26. Since there was little change in polyMMA yield when using different sized particles of MOF-907, it is effectively concluded that the reaction took place mostly within the pores as a result of the unique structural features of MOF-90726. It is noted that for the comparative studies, as shown in Table 1, the particle sizes of MIL-101(Fe) (6 μm), HKUST-1 (18 μm), PCN-280 (37 μm), and Ni-VNU-74(II) (28 μm) were all relatively smaller than the particle size of MOF-907 (Supplementary Fig. 16). This confirms that the structural features of MOF-907 are the most important factor for achieving exceptional catalytic performance.

In order to validate the effect of the catalyst’s pore size on the polymerization reaction, mesoporous PCN-285 (22 Å)9, constructed from BTB and 4,4′-azoBDC linkers, was employed as a catalyst in the polymerization of MMA. The molecular weight of polyMMA produced by PCN-285 (17,100 g mol−1) was smaller than that of the MMA polymer produced by MOF-907 (20,680 g mol−1). The PDI value of polyMMA, obtained by MOF-907, was also lower than PCN-285 (1.23 compared to 1.41, respectively). The catalytic activity of MOF-907 was further investigated by carrying out the polymerization reaction using a larger-sized monomer, benzyl methacrylate (BMA) (Supplementary Methods). Under identical conditions to those used in the polyMMA synthesis, polyBMA product was achieved with a relatively high Mn (14,100 g mol–1), high yield (76%), and low PDI (1.59) when using EBPA as the initiator (Supplementary Fig. 17). MOF-907 again outperforms comparative catalysts for this reaction (Supplementary Table 4 and 5).

Discussion

While conventional27, 28 and photocatalyzed radical polymerization26, 29,30,31 is performed over one day, the microwave-assisted polymerization of MMA by MOF-907 was achieved in just 30 min (Table 1)26,27,28,29,30,31,32,33,34. A detailed kinetic study of this process was performed by monitoring the decrease in MMA concentration as a function of reaction time. Consequentially, the polymerization was a zero-order reaction, thus exhibiting independence from the reaction rate and concentration of MMA (Supplementary Fig. 18). MOF-907 also offers the advantage of recyclability. Over the course of three consecutive reactions, the crystallinity and structural integrity remained intact as evidenced by powder X-ray diffraction analysis (Supplementary Fig. 19). The turnover number was found to be 493 with no noticeable loss in catalytic activity observed as indicated by GPC and NMR analysis (Supplementary Figs. 20–22 and Supplementary Tables 6 and 7). Given the performance of MOF-907, it is likely that the strong Lewis acidic sites, derived from the Fe3O SBUs15,16,17, contributed to fast activation and stabilization of free radicals and, thus, induced chain-growth propagation35,36,37,38,39,40. Indeed, these factors were critical in producing monodisperse, high molecular weight polyMMA.

Methods

Synthesis and characterization of MOF-907

H3BTB (0.012 g, 0.027 mmol) and H2NDC (0.01 g, 0.046 mmol) were dissolved in 3 mL DMF. A mixture of 1 mL DMF and acetic acid (0.20 mL, 3.49 mmol) containing iron(III) nitrate nonahydrate (0.02 g, 0.050 mmol) was then added to the solution containing H3BTB and H2NDC. The reacted solution was subsequently placed in a glass 8 mL vial and heated at 120 °C for 24 h. After that time elapsed, orange cube-shaped crystals were then collected by decantation.

Large-scale synthesis of MOF-907

H3BTB (0.36 g, 0.81 mmol) and H2NDC (0.30 g, 1.38 mmol) were dissolved in 90 mL of DMF. A mixture of 30 mL DMF and acetic acid (6.0 mL, 104.7 mmol) containing iron(III) nitrate nonahydrate (0.60 g, 1.5 mmol) was then added to the solution H3BTB and H2NDC. The reacted solution was subsequently placed in a 200 mL glass bottle and heated at 120 °C for 24 h. The orange powder was cooled down to room temperature and collected by decantation. The product was then washed with DMF (3 × 60 mL, each day) to remove the starting reagents. MOF-907 was immersed in anhydrous acetone for 3 days (3 × 60 mL, each day). A 100 mg activated sample (21% yield based on iron(III) nitrate nonahydrate) was collected after evacuating at 90 °C for 24 h. EA of activated sample: Calcd. for C42H23O17Fe3.0.75C3H7NO (C3H7NO = DMF) = FeIII3O(OH)(BTB)2/3(NDC)2.0.75DMF: C, 52.00; H, 2.79; N, 1.03%. Found: C, 54.16; H, 3.32; N, 0.96%. Noted: DMF is the guest molecules inside the pores. Characterization details are presented in the Supplementary Methods.

Single crystal X-ray diffraction studies

Data was collected at beamline 11.3.1 at the Advanced Light Source, Lawrence Berkeley National Lab at 0.77490 Å. A single crystal of MOF-907 was isolated from the mother liquor of the reaction, mounted on a MiTeGen® kapton loop and placed in a 100(2) K nitrogen cold stream.

The data was processed with the Bruker APEX2 software package, integrated using SAINT v8.37 A and corrected for absorption by SADABS 2016/2 routines (no correction was made for extinction or decay). The structures were solved by intrinsic phasing (SHELXT 2014/4) and refined by full-matrix least squares on F2 (SHELXL-2014-7). Once the framework atoms were located in the difference Fourier maps, the SQUEEZE routine in PLATON was performed to remove scattering from disordered guest molecules residing in the pores. The details of the single crystal refinement are found in the Supplementary Table 1.

Polymerization of methyl methacrylate

A 3.0 mg of activated MOF-907 (0.0032 mmol) was loaded into a 10 mL glass vial for use in the microwave system. 1.0 mL of DMF (13 mmol), 301 µL of methyl methacrylate (MMA, 2.81 mmol), and 1.9 µL of ethyl α-bromoisobutyrate (EBiB, 1.29 × 10–5 mmol) or 2.3 µL ethyl α-bromophenylacetate (EBPA, 1.29 × 10–5 mmol) were then added into the vial. The vial was sealed with a microwave cap. The reaction system was stirred for 30 min at room temperature (covered with Al foil to prevent light interference) before loading into the microwave system. The synthesis condition for the polymerization reaction was set up at 100 °C and 30 min with 200 W of power under a closed system. When the reaction was completed, MOF-907 was collected by centrifugation and washed with dicholoromethane 3–5 times. The polyMMA product was crystallized in a solution of 30 mL MeOH and washed with methanol 3 times. The polymer was dried at room temperature for 2 days or under vacuum for 2 h. The following conditions were optimized: amount of initiator, different concentration of the reaction, temperature, power, and the reaction time. The reported conditions above are optimized. The same procedure was applied using other catalysts (HKUST-1, Ni-VNU-74(II), and FeCl3).