Polycatenated 2D Hydrogen-Bonded Binary Supramolecular Organic Frameworks (SOFs) with Enhanced Gas Adsorption and Selectivity

Controlled assembly of two-dimensional (2D) supramolecular organic frameworks (SOFs) has been demonstrated through a binary strategy in which 1,4-bis-(4-(3,5-dicyano-2,6-dipyridyl)pyridyl)naphthalene (2), generated in situ by oxidative dehydrogenation of 1,4-bis-(4-(3,5-dicyano-2,6-dipyridyl)dihydropyridyl)naphthalene (1), is coupled in a 1:1 ratio with terphenyl-3,3′,4,4′-tetracarboxylic acid (3; to form SOF-8), 5,5′-(anthracene-9,10-diyl)diisophthalic acid (4; to form SOF-9), or 5,5′-bis-(azanediyl)-oxalyl-diisophthalic acid (5; to form SOF-10). Complementary O–H···N hydrogen bonds assemble 2D 63-hcb (honeycomb) subunits that pack as layers in SOF-8 to give a three-dimensional (3D) supramolecular network with parallel channels hosting guest DMF (DMF = N,N′-dimethylformamide) molecules. SOF-9 and SOF-10 feature supramolecular networks of 2D → 3D inclined polycatenation of similar hcb layers as those in SOF-8. Although SOF-8 suffers framework collapse upon guest removal, the polycatenated frameworks of SOF-9 and SOF-10 exhibit excellent chemical and thermal stability, solvent/moisture durability, and permanent porosity. Moreover, their corresponding desolvated (activated) samples SOF-9a and SOF-10a display enhanced adsorption and selectivity for CO2 over N2 and CH4. The structures of these activated compounds are well described by quantum chemistry calculations, which have allowed us to determine their mechanical properties, as well as identify their soft deformation modes and a large number of low-energy vibration modes. These results not only demonstrate an effective synthetic platform for porous organic molecular materials stabilized solely by primary hydrogen bonds but also suggest a viable means to build robust SOF materials with enhanced gas uptake capacity and selectivity.


■ INTRODUCTION
Porous organic molecular solids have been proposed as a new platform of chemical materials. They tend to show low framework density resulting from being composed of light elements (typically limited to H, C, N, O, and B), and their synthesis can be achieved via controlled assembly using concepts from crystal engineering. 1−5 Various applications of porous organic molecular solids, including in catalysis, medicine, molecular sensing, storage, and separation, 6−10 depend on achieving designed architectures with predictable pore structures and the necessary functionality. However, the construction of porous organic molecular solids presents challenges in the delivery of required properties, including stable and permanent porosity, tailored pore structure, and functionality. Supramolecular organic frameworks (SOFs), 11−18 which are structurally mimetic of the well-known metal organic frameworks (MOFs), 19−23 define porous crystalline molecular solids built from organic constituents assembled through supramolecular interactions. SOFs have shown promising applications in gas storage and separation. 24,25 The prevailing strategy for the construction of SOFs involves crystallization of a single organic constituent from an appropriate solvent system by exploiting supramolecular interactions such as hydrogen bonds, halogen bonds, and/or π···π interactions. 26−30 Given the "soft" and "flexible" nature of supramolecular interactions, as well as the numerous donating/accepting groups present, prediction of structures and controlled assembly of hydrogen-bonded SOFs in such single-component systems can be challenging. 30,31 We have now developed a binary design strategy, 25 which requires two types of organic constituents possessing separate complementary hydrogen bond donating and acceptor groups. The organic constituents should be chemically and geometrically adaptable to give intermolecular interactions that are optimized and balanced for assembly of binary SOF materials. 32 The importance of dimensionality and well-defined in-plane crystallinity of two-dimensional (2D) materials is now well established, 28,29 and structural diversity of 2D molecular networks comes not only from the molecular assembly but also from the different packing arrangements possible for these nets. 29,33 In contrast to 3D SOFs, 12−18 2D SOFs assembled solely through hydrogen bonds rarely exhibit both structural robustness and significant adsorption capacity in the solid state. With this in mind, we sought to construct thermally and chemically stable 2D SOF materials for selective adsorption of carbon dioxide (CO 2 ) using a new binary strategy. 1,4-Bis-(4-(3,5-dicyano-2,6-dipyridyl)dihydropyridyl)naphthalene (1) has been designed as a precursor for the generation of 1,4-bis-(4-(3,5-dicyano-2,6-dipyridyl) pyridyl)naphthalene (2) via in situ oxidative dehydrogenation. 25,34 We report that the combination of 2 with the tetracarboxylic acids terphenyl-3,3′,4,4′tetracarboxylic acid, 3, 5,5′-(anthracene-9,10-diyl)diisophthalic acid, 4, or 5,5′-bis-(azanediyl)-oxalyl-diisophthalic acid, 5, leads to the assembly of 2D binary SOF materials (Scheme 1). The exo-pyridyl (acceptors) and carboxyl (donors) groups on two separate organic constituents provide complementary and directional hydrogen bonding interactions for the assembly of porous structures with decorating phenyl, anthryl, or oxamide groups. Quantum chemistry calculations are used to determine the energetic and mechanical stability of the various frameworks synthesized here, as well as to analyze their elastic moduli and vibrational properties.

■ EXPERIMENTAL SECTION
Chemicals and General Methods. Commercially available reagents and organic solvents were used as received without further purification. 3-Amino-3-(4-pyridinyl)-propionitrile, naphthalene-1,4dicarbaldehyde, terphenyl-3,3′,4,4′-tetracarboxylic acid, 5,5′-bis-(azanediyl)-oxalyl-diisophthalic acid, and 5,5′-(anthracene-9,10-diyl)diisophthalic acid were prepared according to the reaction procedures previously described in the literature; naphthalene-1,4-dicarbaldehyde was synthesized using an adapted method reported previously. 35,36 Elemental analyses (C, H, and N) were performed on a CE-440 elemental analyzer. Infrared (IR) spectra were recorded with a PerkinElmer Spectrum One instrument as KBr pellets in the range 400−4000 cm −1 , or on a Nicolet is5 FT-IR spectrophotometer in the range 550−4000 cm −1 using attenuated total reflectance (ATR) mode. 1 H NMR spectra were recorded on a Bruker DPX-400 spectrometer. Thermal gravimetric analyses (TGA) were performed under a flow of N 2 (20 mL min −1 ) with a heating rate of 10°C min −1 using a TA SDT-600 thermogravimetric analyzer. Powder X-ray diffraction (PXRD) measurements were carried out at room temperature on a PANalytical X'Pert PRO diffractometer using Cu Kα radiation (λ = 1.5418 Å) and generator settings of 40 kV, 40 mA, at a scan speed of 0.02°/s and a step size of 0.005°in 2θ. CO 2 , N 2 , and CH 4 isotherms were recorded using an IGA gravimetric adsorption apparatus (Hiden) at the University of Nottingham in a clean ultra-high-vacuum system with a diaphragm and turbo pumping system. Before measurement, about 60 mg of solvent-exchanged sample was loaded into the sample basket within the adsorption instrument and then degassed under dynamic vacuum at 100°C for 24 h to obtain fully desolvated samples.
Crystallography. Single crystal X-ray data were collected on an Rigaku Oxford Diffraction SuperNovaII Atlas X-ray diffractometer at the University of Nottingham. Details of the data collection are included in the CIF files. Structures of SOF 8−10 were solved by direct methods and developed by difference Fourier techniques, using the SHELXL software package (Table S1). 38 Hydrogen atoms of the ligands were placed geometrically and refined using a riding model. The unit cell volume of SOF 8−10 included a large region of disordered solvent which could not be modeled as discrete atomic sites. We therefore employed PLATON/SQUEEZE 37 to calculate the contribution of the solvent region to the diffraction and thereby produced a set of solvent-free diffraction intensities. The solvent molecules of SOF 8−10 are included in the unit cell contents and in all parameters derived from these. Despite multiple attempts, the limited crystal size and stability of SOF-8 restricted the quality of the resulting data set, but this did not preclude the identification of the framework structure. Topological analysis and polycatenation identification of SOF-9 and SOF-10 were performed using ToposPro software. 39 ■ RESULTS AND DISCUSSION Structural Description. In a typical procedure, reaction of 1 and 3 in a 1:1 molar ratio in dimethylformamide (DMF) at 90°C resulted in the formation of light-yellow prismatic crystals of SOF-8 after 72 h. Single crystal X-ray diffraction reveals that SOF-8 crystallizes in the monoclinic space group P2 1 /m (Table S1), and features a 2D hydrogen-bonded network structure in which 1 has undergone a structural transformation into 2 through oxidative dehydrogenation (Scheme 1). The exo-pyridyl and carboxyl groups on 2 and 3 contribute to the complementary O−H···N hydrogen bonds that direct the self-assembly process. Each molecule of 2 interacts with four neighboring molecules of 3 through strong primary hydrogen bonds (O carboxyl H···N pyridyl , 2.562/2.568 Å; Table S2) to form a 2D honeycomb layer (6 3 -hcb) supramolecular organic framework with windows of ca. 10.1 Å × 15.9 Å (Figure 1, left). The 2D 6 3 -hcb layers stack in a parallel fashion and give rise to an overall 3D supramolecular structure with one-dimensional cyano-decorated channels ( Figure S5, Supporting Information). It should be noted that the naphthalene moieties on 2 are disordered in symmetric positions, resulting in them pointing up and down into the windows of the 2D layers. In this way, neighboring 2D layers in SOF-8 interact with each other via π···π interactions.
Single crystal X-ray analysis confirms that SOF-9 crystallizes as a 1:1 mixture of 2 and 4 in the orthorhombic space group Pba2 (Table S1). The 2D honeycomb layer subunit in SOF-9 is built by primary O−H···N hydrogen bonds between exo-pyridyl and carboxyl groups on 2 and 4 ( Figure 1, middle), hydrogen bonds characterized by O···N distances in the range 2.548− 2.633 Å, and approximately linear O−H···N angles (Table S2).  In comparison with SOF-8, windows of the 2D layers in SOF-9 are more regular, with dimensions of ca. 12.0 Å × 12.3 Å. Packing of the 2D layers generates two sets of hcb layers, 40−43 resulting in a 3D inclined polycatenated framework ( Figure  2a,b). The dihedral angle between these two sets of layers is 79.6°. Each window of a layer is polycatenated with four other windows from two different layers via Hopf links. 39,44 Alternatively, it can be described that each window encircles two edges passing through it (Figure 2b, inset).
X-ray crystallography confirms that SOF-10 crystallizes in the orthorhombic space group Pbc2 1 (Table S1) Table S1, Supporting Information). Packing of the 2D layers results in a 3D inclined polycatenated framework similar to that of SOF-9, consisting of two sets of equivalent hcb layers (Figure 2a,c). The dihedral angle between these two sets of layers is 59.5°. Each window of a layer is polycatenated with six windows from three different layers via Hopf links. Alternatively, each window can be described as encircling the three edges passing through it (Figure 2c, inset). It should be noted that the O−H···N hydrogen bonds in SOF 8−10 are considerably shorter than the ordinary H-bond (2.70−3.00 Å), and they are thus considered as double charge-assisted H-bonds, comprised of strong 1/2− D···H + ···A 1/2− interactions (D = H-bond donor; A = H-bond acceptor). 45,46 Adsorption and Selectivity. The total solvent-accessible volume of SOF 8−10 after removal of guest DMF molecules was estimated to be ca. 39.2%, 34.2%, and 29.9%, respectively, as calculated using the PLATON/VOID routine. 37 It is not surprising that nonentangled SOF-8 has the largest potential molecular voids, and the decrease of molecular voids for SOF-9 and SOF-10 is attributable to framework interpenetration. However, structural collapse occurs upon removal of guest DMF molecules from SOF-8 ( Figures S1 and S2, Supporting Information). In contrast, the polycatenated SOF-9 and SOF-10 display excellent thermal stability, as evaluated by thermogravimetric analysis (TGA) and powder X-ray diffractions (PXRD) (Figures S1, S3, and S4, Supporting Information), and they retain their structural integrity and crystallinity upon solvent exchange as well as upon the removal of guest molecules.
Quantum Chemical Modeling. In order to gain more insight into these new 2D SOF materials, and to understand why SOF-9 and SOF-10 are, like SOF-7, 25 stable upon guest removal while SOF-8 is not, we analyzed these structures by means of quantum chemistry calculations, at the density functional theory (DFT) level. Starting from the guest-free crystallographic structures, geometry optimization was performed first to check that these materials, together with SOF-7, are well described at this level of theory, and then focused on their energetic and mechanical stabilities, as well as on vibrational properties. The relaxed structures are in excellent agreement with those obtained by single crystal X-ray crystallography, with deviations below 2% (even below 1% for SOF-8 and SOF-10) in either volume or individual cell parameters, as shown in Table S5. Intermolecular binding via O carboxyl H···N pyridyl bonds is well reproduced, with O···H···N distances between 2.52 and 2.57 Å, although slightly shorter than observed in the experimental structures for SOF-9 and SOF-10.
For each of the relaxed structures, the binding energy (or formation enthalpy) has been estimated by subtracting from the total energy (appropriately weighted) that of its constituting tectons. These binding energies have been found to be all fairly large, between −429 and −438 kJ mol −1 (per carboxylate-based constituent). It should be noted that these refer to the formation of SOF materials in the absence of solvents; we expect the formation enthalpies of these materials in DMF to be significantly smaller, since a rather large binding energy of DMF to carboxylate-based constituents (calculated binding energy as −65 to −80 kJ mol −1 , depending on the constituent) is essentially lost when this constituent binds to an exo-pyridyl group.
The mechanical properties of these SOF materials have also been investigated by means of computing their second-order elastic stiffness tensors (also known as elastic constants). This has proven to be a reliable indicator of mechanical stability and potential for crystal-to-crystal phase transitions in metal− organic frameworks and other framework materials. 55−57 The resulting bulk, Young's, and shear moduli are listed in Table S6. All materials have reasonably large bulk moduli, between 9.0 and 9.5 GPa. Their response to isostatic pressure is however unusual, with (small) negative linear compressibility in specific directions. Whereas spatially averaged Young's moduli E, ranging from 5.8 GPa in SOF-7 to 9.0 GPa in SOF-10, do not differ much from one material to another, their response to directional strains is instructive. Some materials are particularly soft with respect to specific deformation modes, especially the shear moduli which can be very low, G min = 0.21 GPa for SOF-7 and 0.41 GPa for SOF-8. Both materials are also sensitive to specific compression modes with E min of the order of 1 GPa or below. For instance, the soft shearing mode of SOF-8 corresponds to sliding of a 2D honeycomb layer with respect to neighboring layers. Its counterpart in SOF-7 is similar, implicating relative motions of neighboring constituents, although the situation is slightly more complex since neighboring coordination layers are intertwined. 25 The higher stability of SOF-8 over SOF-7 to shearing deformations, from these results, seems inconsistent with higher stability of SOF-7 to desolvation, which might induce some local deformations. An important difference, however, is that SOF-7 consists of intertwined networks (as do SOF-9 and SOF-10), which may result in a higher stability to large deformations (beyond the elastic regime), compared to the essentially layered SOF-8. A similar behavior has been observed, for example, in the family of MUF-8 and MUF-9 metal−organic frameworks with controlled partial interpenetration. 58 Vibration modes of SOF-7 and SOF-8 have been further analyzed. For the latter, a large number of low-frequency modes are found: eight modes with frequencies ω < 30 cm −1 (not including three purely translational modes), and about 50 modes spread quite regularly over the range of 30−100 cm −1 . An analysis of the former eight gives the following trends: (a) In two of these modes, constituents vibrate as (almost) rigid units, and their motion involves significant deformations in the O−H···N units (shortening/ elongation of O−H or N···H bonds). (b) In the two other modes, constituents from the same coordination layer vibrate in phase, but such that neighboring layers slide with respect to each other. This induces local deformations, e.g., bending of C−C− C angles at tertiary carbons, and relative motions of neighboring aromatic cycles coupled by π···π interactions. (c) The remaining four modes have more complex structures, involving significant bending of C−C−C angles at tertiary carbons and/or of dihedrals centered on single C−C bonds between aromatic rings. Similarly, SOF-7 has a large number of low-frequency vibration modes, but fewer (in comparison to SOF-8) in the ω < 20 cm −1 range. Moreover, modes of the type (c), connected to the polycatenated nature of the framework, are again absent for SOF-7. This qualitative difference, as well as the larger population of vibration modes in SOF-8 at ambient temperature, may explain its instability upon desolvation.

Article
Finally, vibrational mode calculations at finite (positive and negative) pressures have been performed in order to determine the thermal expansion coefficients α V = V −1 (dV/dT). We obtained the values α V = 15.7 MK −1 for SOF-7, and a remarkably low α V = 2.2 MK −1 for SOF-8. The latter value is due to the presence, among low-energy vibration modes, of several modes that tend to soften (decrease in frequency) under compression; i.e., these modes alone would result in negative thermal expansion (NTE). This makes it likely that NTE will be encountered in further SOF materials, which opens further perspectives for applications of these materials.

■ CONCLUSION
In summary, we have demonstrated a rational design strategy to prepare 2D binary supramolecular organic framework (SOF) materials assembled via complementary hydrogen bonding interactions. The SOF assembly is tunable through rational design of organic constituents with hydrogen bond donating and accepting groups. The porous supramolecular organic frameworks SOF 8−10 display similar 2D hydrogen-bonded 6 3 -hcb layers with various packing modes: SOF-8 features parallel packing of 2D layers, whereas SOF-9 and SOF-10 possess 2D → 3D inclined polycatenation of 6 3 -hcb layers. The higher chemical and thermal stability for SOF-9 and SOF-10, which are stable up to 300°C, have been achieved through structural polycatenation. Further enhancement of adsorption capacity in activated samples of SOF-10a depends greatly on the introduction of favorable binding sites (the oxamide group in this case) for guest CO 2 molecules. Therefore, appropriate functionalization of the organic constituents in SOF materials favors not only the stability of the materials but also their gas adsorption capacity and selectivity. Quantum chemistry calculations suggest a connection between this stability and the existence of soft deformation modes (especially shearing modes) and low-energy vibration modes. The latter may in particular account for the instability of SOF-8 upon guest removal, and also for unusual thermal expansion properties. These results further confirm that our binary design strategy is effective for the synthesis of highly stable and robust functional porous supramolecular organic framework materials with potentials for storage and separation. Moreover, enhanced CO 2 adsorption and selectivity has been achieved by structural regulation via controlled assembly of 2D SOF materials into polycatenated structures.

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.cgd.8b00153.
Gas adsorption isotherms, gas uptake data and analysis, and quantum chemistry calculations, tables, and figures (PDF)