A Heterometallic Three-Dimensional Metal−Organic Framework Bearing an Unprecedented One-Dimensional Branched-Chain Secondary Building Unit

A heterometallic metal−organic framework (MOF) of [Cd6Ca4(BTB)6(HCOO)2(DEF)2(H2O)12]∙DEF∙xSol (1, H3BTB = benzene-1,3,5-tribenzoic acid; DEF = N,N′-diethylformamide; xSol. = undefined solvates within the pore) was prepared by solvothermal reaction of Cd(NO3)2·4H2O, CaO and H3BTB in a mixed solvent of DEF/H2O/HNO3. The compatibility of these two divalent cations from different blocks of the periodic table results in a solid-state structure consisting of an unusual combination of a discrete V-shaped heptanuclear cluster of [Cd2Ca]2Ca′ and an infinite one-dimensional (1D) chain of [Cd2CaCa′]n that are orthogonally linked via a corner-shared Ca2+ ion (denoted as Ca′), giving rise to an unprecedented branched-chain secondary building unit (SBU). These SBUs propagate via tridentate BTB to yield a three-dimensional (3D) structure featuring a corner-truncated P41 helix in MOF 1. This outcome highlights the unique topologies possible via the combination of carefully chosen s- and d-block metal ions with polydentate ligands.

The presence of a second metal ion in the one-pot assembly may play a structure-directing role in obtaining MOF topologies inaccessible from the individual metal sources [30][31][32][33][34]51]. By comparison, the post-synthetic method is better for obtaining MOFs with predictable topology, and the targeted metal ion can diffuse into the pores to trigger the cation incorporation/exchange without disrupting the overall framework connectivity [49,[52][53][54][55][56][57][58][59]. The one-pot reaction faces the risk of forming separate MOFs from dependent metal ions, whereas the post-synthesis method has the challenge of incomplete conversion, requiring repeated experiments to obtain a desirable conversion efficiency.
The one-pot assembly of heterometallic MOFs is facilitated by metal ions of similar ionic radius and charge. Mixtures of Zn 2+ (0.070 ± 0.007 nm) and Mg 2+ (0.070 ± 0.004 nm) or Cd 2+ (0.0910 ± 0.003 nm) and Ca 2+ (0.103 ± 0.005 nm) may have promising advantages in this respect [60]. However, there are only limited reports of heterometallic MOFs of these elements, presumably because of the relatively few exploitable properties of Mg 2+ and Ca 2+ .
Incorporating s-block metal ions, such as Li + , Na + , Mg 2+ , and Ca 2+ , into MOF assemblies may impart several advantages. For example, the flexible coordination geometries of these ions and the ionic forces between the metal ions and ligands (relative to dative bonds between a typical d-block metal ion and the ligand) may permit novel framework topologies. Some of these s-block metal ions (e.g., Li + ) have additional advantages of being inexpensive and relatively non-toxic [12,50,61].
In this work, we report a three-dimensional (3D) MOF of [Cd 6 Ca 4 (BTB) 6 (HCOO) 2 (DEF) 2 (H 2 O) 12 ]·DEF·xSol. (1, H 3 BTB = benzene-1,3,5-tribenzoic acid; DEF = N,N -diethylformamide; xSol. = undefined solvates within the pore) sustained by an unprecedented one-dimensional (1D) branched SBU, which, in turn, consists of a discrete V-shaped heptanuclear cluster of [Cd 2 Ca] 2 Ca and an infinite 1D chain of [Cd 2 CaCa ] n linked via a corner-shared Ca 2+ ion (denoted as Ca ). Such a complicated MOF generated via the mixed use of a Cd 2+ and Ca 2+ features the first case of a framework structure containing both Ca 2+ and BTB ligands. In addition, the structure of MOF 1 is dramatically different from, and in fact much more delicate than, all the literature examples that are exclusively from the assembly of Cd 2+ with BTB ligands [62][63][64][65].

Synthesis and Material Characterization of MOF 1
The motivation for introducing Ca 2+ into a Cd 2+ /H 3 BTB assembly originated from our quest to incorporate Ca 2+ into the linear Cd 3 -based two-dimensional (2D) MOF of [Cd 3 (BTB) 2 (DEF) 4 ]·2(DEF) 0.5 [62,63], as we and others have demonstrated that the heterometallic combinations of Cd/Zn, Cd/Co, and Zn/Co are readily tolerated in linear trimetallic cluster SBUs [26,66] [65]. The [Et 2 NH 2 ] + cations likely arose from the acid-catalyzed hydrolysis of DEF [67][68][69][70], and subsequently served as a template for the framework formation. In the present work, we employed CaO as a replacement of CaSO 4 to consume part of the acid, and the new neutral 3D MOF 1 was thus obtained in a high yield of 87%. It is interesting to note that DEF also decomposed in the present case, with the anionic part of HCOO − acting as a bridging ligand for structure propagation.
MOF 1 contains Ca 2+ as a key component, is stable under aerobic conditions, and is stable and insoluble in MeOH, EtOH, DMF, DEF, CH 2 Cl 2 , CHCl 3 , and MeCN. The powder X-ray diffraction (PXRD) pattern of the bulk crystalline sample of MOF 1 agreed well with that simulated from the single-crystal diffraction data, indicating its bulk phase purity ( Figure 1). The thermogravimetric analysis (TGA) of MOF 1 indicated that the framework is stable up until ca. 100 • C (Supplementary Figure S1), followed by a continuous weight loss up to ca. 650 • C. At this stage, we were not able to provide a clear compositional analysis based on the TGA result due to the large pore cavity of MOF 1 (as will be discussed below). It is also unfortunate that there was no obvious plateau found upon initial solvate loss. The thermally unstable nature of MOF 1 prevented us from studying its surface area via gas adsorption, and our preliminary experiments suggested that MOF 1 exhibits no adsorption of N 2 at 77 K (Supplementary Figure S2), presumably due to the collapse of the 3D framework. The Fourier-transform infrared spectrum (FT-IR) of MOF 1 contained absorptions at 2975 cm −1 , 2930 cm −1 , and 1652 cm −1 corresponding to the-CH 3 , -CH 2 -, and -C=O bonds of te DEF solvate [71,72]. In addition, peaks at 1610 cm −1 , 1536 cm −1 , and 1397 cm −1 were assignable as the asymmetrical and symmetrical stretching vibrations of the carboxylate [66,73].  [71,72]. In addition, peaks at 1610 cm −1 , 1536 cm −1 , and 1397 cm −1 were assignable as the asymmetrical and symmetrical stretching vibrations of the carboxylate [66,73].
The powder X-ray diffraction patterns of MOF 1, showing a good consistency between the experimental (red) and simulated (black) patterns and thus a high phase purity of the bulk material. The slight inconsistency in the intensity of some peaks might have been due to the different orientation of the crystallites in the sample, as the metal-organic framework (MOF) is nonmerohedrally twinned.

Crystal Structure Analysis of MOF 1
The presence of Cd 2+ and Ca 2+ and their exact atomic sites in MOF 1 could be unambiguously inferred from the X-ray crystallographic studies. This benefited from the large discrepancy of the atomic mass between Cd 2+ (112.441) and Ca 2+ (40.078), as any incorrect assignment of the atomic sites will lead to strikingly unrealistic thermal ellipsoids and other related issues during the structure refinement. Such MOF assemblies from a mixture of Cd 2+ and Ca 2+ sources thus have an advantage for structure determination by X-ray crystallography, in stark contrast to those heterometallic MOFs bearing mixed metals with similar atomic mass, such as Zn/Co and Mn/Co [23,66]. MOF 1 crystallizes in the tetragonal crystal system space group I41/a (Table 1), and its asymmetrical unit contains three Cd 2+ , three Ca 2+ , three BTB and one HCOO bridging ligands, together with one coordinated DEF (to Cd3) and six coordinated H2O solvates (two to Cd1, one to Cd2, one to Ca1, and two to Ca2). There is also one DEF solvate with half occupancy residing within the cavity of MOF 1. As shown in Figure 2a cluster, Cd1 (6-coordinate) is coordinated by three distinctive carboxylates, i.e., one monodentate (η 1 ), one bridging (μ2-η 1 :η 1 ), and one chelating-bridging (μ2-η 1 :η 2 ), in addition to two H2O solvates that terminate the propagation of the cluster. Cd2 (7-coordinate) is associated with one bridging (μ2-η 1 :η 1 ) and a pair of chelating-bridging (μ2-η 1 :η 2 ) carboxylates, one chelating-bridging HCOO (μ3-η 1 :η 1 :η 2 ), and one additional coordinated H2O. In addition, the central Ca1 (6-coordinate) is associated with two bridging (μ2-η 1 :η 1 ) and three chelating-bridging (μ2-η 1 :η 2 ) carboxylates, in addition to one terminally coordinated H2O. On the other hand, the central Ca2 (Ca', 8-connected) is associated with a pair of chelating-bridging HCOO (μ3-η 1 :η 1 :η 2 ) and four terminal H2O solvates. The [Cd2Ca]-Ca'-[Cd2Ca] angle (θ) is roughly 132° (Figure 2b). The powder X-ray diffraction patterns of MOF 1, showing a good consistency between the experimental (red) and simulated (black) patterns and thus a high phase purity of the bulk material. The slight inconsistency in the intensity of some peaks might have been due to the different orientation of the crystallites in the sample, as the metal-organic framework (MOF) is non-merohedrally twinned.
The intersection of the discrete V-shaped heptanuclear [Cd2Ca]2Ca' subunits and the rod-shaped [Cd2CaCa']n chain through Ca' (Ca2) gives rise to a 1D branched SBU (Figure 3a,b). Such an unprecedentedly broad SBU features isolated voids which are filled with the bulky BTB ligands. It may be that these BTB ligands serve as the template for the arrangement of this unique SBU. These sizable SBUs are further associated with BTB ligands to give a 3D framework structure featuring densely packed BTB ligands when looking along the crystallographic c direction (Figure 4a). When looking along the crystallographic a or b direction, 1D corner-sharing channels with rhombohedral apertures are observed (Figure 4b). A Platon void calculation indicates a total solvent-accessible volume (including the free DEF) for each cell unit amounting to 22,138.0 Å 3 or 43.3% of the total cell volume (51,134.0 Å 3 ) [74].  (Figure 2a,b) as discussed. The [Cd 2 Ca] cluster unit herein comprises a pair of Cd centers (Cd3) symmetrically related via the central Ca3, in which Cd3 (6-coordinate) is associated by a pair of bridging (µ 2 -η 1 :η 1 ) and one chelating-bridging (µ 2 -η 1 :η 2 ) carboxylates, one chelating-bridging HCOO (µ 3 -η 1 :η 1 :η 2 ), and a terminally coordinated DEF solvate. Meanwhile, the central Ca3 (6-coordinate) is bonded by six carboxylates, of which four are in bridging (µ 2 -η 1 :η 1 ) and two in chelating-bridging (µ 2 -η 1 :η 2 ) coordination fashions.
The intersection of the discrete V-shaped heptanuclear [Cd 2 Ca] 2 Ca subunits and the rod-shaped [Cd 2 CaCa ] n chain through Ca (Ca2) gives rise to a 1D branched SBU (Figure 3a,b). Such an unprecedentedly broad SBU features isolated voids which are filled with the bulky BTB ligands. It may be that these BTB ligands serve as the template for the arrangement of this unique SBU. These sizable SBUs are further associated with BTB ligands to give a 3D framework structure featuring densely packed BTB ligands when looking along the crystallographic c direction (Figure 4a). When looking along the crystallographic a or b direction, 1D corner-sharing channels with rhombohedral apertures are observed (Figure 4b       It is interesting to note that the two types of [Cd2Ca] cluster units (Cd1-Ca1-Cd2 denoted as type A node, and Cd3-Ca3-Cd3 denoted as type B node; ratio type A:type B = 2:1) propagate in the c direction in a helical sequence around the 41 axis ( Figure 5). In such a helical chain, the Ca' (Ca2, denoted as type C node) is critical and serves as the corner mediating the turn between type A and type B nodes. However, an additional corner to mediate the turn between two sequential type A nodes is missing (we herein use node □ to represent such an imaginary corner). These four types   It is interesting to note that the two types of [Cd 2 Ca] cluster units (Cd1-Ca1-Cd2 denoted as type A node, and Cd3-Ca3-Cd3 denoted as type B node; ratio type A:type B = 2:1) propagate in the c direction in a helical sequence around the 4 1 axis ( Figure 5). In such a helical chain, the Ca (Ca2, denoted as type C node) is critical and serves as the corner mediating the turn between type A and type B nodes. However, an additional corner to mediate the turn between two sequential type A nodes is missing (we herein use node to represent such an imaginary corner). These four types   It is interesting to note that the two types of [Cd2Ca] cluster units (Cd1-Ca1-Cd2 denoted as type A node, and Cd3-Ca3-Cd3 denoted as type B node; ratio type A:type B = 2:1) propagate in the c direction in a helical sequence around the 41 axis ( Figure 5). In such a helical chain, the Ca' (Ca2, denoted as type C node) is critical and serves as the corner mediating the turn between type A and type B nodes. However, an additional corner to mediate the turn between two sequential type A nodes is missing (we herein use node □ to represent such an imaginary corner). These four types   From a topological perspective, we may consider Ca (Ca2) as a tetrahedral 4-connecting node (type C node, presented as cyan in Figure 6) that is linked to a pair of neighboring equivalents in the crystallographic b direction (within the rod-shaped chain, Figure 6c,d) and a pair of [Cd 2 Ca] (type A node presented as orange in Figure 6). The four angles around the Ca (Ca2) are θ1 = 136 • , θ2 = θ3 = 99 • , and θ4 = 152 • (Figure 6b). Four type A nodes are arranged in a head-to-head fashion to generate a small A 4 motif. This A 4 motif also functions as a tetrahedral node and extends to four type C nodes, and ultimately gives rise to a diamond-type 3D network. From a topological perspective, we may consider Ca' (Ca2) as a tetrahedral 4-connecting node (type C node, presented as cyan in Figure 6) that is linked to a pair of neighboring equivalents in the crystallographic b direction (within the rod-shaped chain, Figure 6c,d) and a pair of [Cd2Ca] (type A node presented as orange in Figure 6). The four angles around the Ca' (Ca2) are θ1 = 136°, θ2 = θ3 = 99°, and θ4 = 152° (Figure 6b). Four type A nodes are arranged in a head-to-head fashion to generate a small A4 motif. This A4 motif also functions as a tetrahedral node and extends to four type C nodes, and ultimately gives rise to a diamond-type 3D network.  Figure (c,d) show the diamondoid network looking along the crystallographic b axis, with (c) also elucidating connections among these clusters.
It should be noted that the MOF 1 structure reported herein is drastically different from, and as a matter of fact more complicated than, several BTB-based MOFs of Cd, such as 2D MOF of [Cd 3 (BTB) 2 (DEF) 4 ]·2(DEF) 0.5 [62][63][64]

General
Cd(NO 3 ) 2 ·4H 2 O (≥ 99.99%, Macklin), benzene-1,3,5-tribenzoic acid (H 3 BTB, > 98.0%, TCI), N,N -diethylformamide (DEF, > 99.0%, TCI), CaO (98.0%, Macklin), and HNO 3 (65.0−68.0%, Enox) were obtained from commercial sources and used as received. Fourier-transform infrared (FT-IR) spectrum was measured on a Varian 1000 FT-IR spectrometer (Varian, Inc., Palo Alto, CA, USA) as KBr disks (400−4000 cm −1 ). Elemental analyses for C, H, and N were conducted on a Carlo-Erba CHNO-S microanalyzer (Carlo Erba, Waltham, MA, USA), with the sample first immersed in CHCl 3 to replace the encapsulated species with volatile CHCl 3 and air-dried before analysis. The thermogravimetric analysis (TGA) was performed on a TA instrument Q500 (TA instruments, New Castle, DE, USA) from room temperature to 800 • C at a heating rate of 10 • C min −1 under an N 2 gas flow in an Al 2 O 3 pan. The powder X-ray diffraction (PXRD) pattern was recorded on a Bruker D8 GADDS (General Area Detector Diffraction System) micro-diffractometer (Bruker AXS GmbH, Karlsruhe, Germany) equipped with a VANTEC-2000 area detector (Bruker AXS GmbH, Karlsruhe, Germany) using the Φ rotation method. Nitrogen adsorption isotherms were recorded using a BELSORP-max (MicrotracBEL Corp., Osaka, Japan). The encapsulated solvent in the sample was exchanged with CHCl 3 twice and air-dried in the fumehood, before being transferred to the instrument for activation at room temperature under a vacuum of 10 −2 kPa. The evacuated sample tube was weighed again after 36 h and the sample mass was determined by subtracting the original mass. The nitrogen isotherms were measured using a liquid nitrogen bath (77 K).

X-Ray Crystallography for MOF 1
The single-crystal structure of MOF 1 was analyzed on a Bruker D8 Quest CCD X-ray diffractometer with graphite monochromated Mo Kα (λ = 0.71073 Å) radiation. Refinement and reduction of the collected data were achieved using the Bruker SAINT program and applied to all complexes with absorption correction (multi-scan) [75]. The crystal structure was solved by direct methods and refined on F 2 by full-matrix least-squares techniques with the SHELXTL-2013 program [76]. In MOF 1, the two coordinated waters (O25 and O26) display positional disorder with relative ratios of 0.54/0.46 refined for the two components. The occupancy factors for the atoms of the dissociated DEF solvate were fixed at 0.5 to obtain reasonable thermal factors. The hydrogen atoms on the water molecules (O21, O23, and O24) were located from the difference Fourier map, while those on the water molecule (O22) were generated by considering their possible hydrogen-bonding interactions with atoms nearby. The O-H distances and thermal parameters were subsequently constrained to O-H = 0.83 Å and U iso (H) = 1.2U eq (O). The hydrogen atoms on the disordered water solvates (O25, O26/O25a, O26a) were not located. The structure adopted non-merohedral twinning about the (−1 −1 0) and the twin law [0 −1 0 −1 0 0 0 0 −1] was included during the refinement. A large amount of spatially delocalized electron density in the lattice was also found (3247 electrons in 20,271 Å 3 solvent-accessible volume), but acceptable refinement could not be obtained for this electron density. The solvent contribution was then modeled using SQUEEZE in the Platon program suite [77]. Crystallographic data were deposited with the Cambridge Crystallographic Data Center as supplementary publication number CCDC 1990807. These data can be obtained free of charge from the Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif. A summary of the key crystallographic data is listed in Table 1, and selected bond lengths and angles are depicted in Supplementary Table S1.

Conclusions
The isolation of MOF 1 exemplified the usefulness of s-block metal ions for the MOF assembly. The three independent Ca 2+ ions in MOF 1 exhibit coordination numbers of either six (Ca1 and Ca3) or eight (Ca2), with the latter seldom achieved for a typical first or second-row d-block metal complex. The structure of MOF 1 is complex and dramatically different from MOFs exclusively based on Cd-BTB. Cd 2+ and Ca 2+ , although similar in size, are quite different in atomic mass, which made them easy to distinguish during the X-ray crystal structure analysis. MOF 1 was insufficiently stable for complete removal of solvate from the large pore, which prevented us from obtaining meaningful porosity data via gas adsorption studies. Nevertheless, we believe that careful selection of s-block metals to mix with d-block metal ions (not limited to two) could be a powerful strategy for preparing MOFs with interesting topologies and unique properties.
Supplementary Materials: Figure S1: The TGA curve of MOF 1 showing that the framework is stable up to ca. 100 • C, followed by continuous weight loss and framework decomposition until ca. 650 • C; Figure S2: The N 2 (77 K) sorption isotherms of MOF 1 with the black squares and red circles representing adsorption and desorption. P 0 is the saturated vapor pressure of the adsorbates at the measurement temperatures; Table S1: Selected bond distances in MOF 1 involving the Cd and Ca centers.

Conflicts of Interest:
The authors declare no conflict of interest.