Crystallographic investigation into the self-assembly, guest binding, and flexibility of urea functionalised metal-organic frameworks

Abstract Introduction of hydrogen bond functionality into metal-organic frameworks can enhance guest binding and activation, but a combination of linker flexibility and interligand hydrogen bonding often results in the generation of unwanted structures where the functionality is masked. Herein, we describe the self-assembly of three materials, where Cd2+, Ca2+, and Zn2+ are linked by N,Nʹ-bis(4-carboxyphenyl)urea, and examine the effect of the urea units on structure formation, the generation of unusual secondary building units, structural flexibility, and guest binding. The flexibility of the Zn MOF is probed through single-crystal to single-crystal transformations upon exchange of DMF guests for CS2, showing that the lability of the [Zn4O(RCO2)6] cluster towards solvation enables the urea linkers to adopt distorted conformations as the MOF breathes, even facilitating rotation from the trans/trans to the trans/cis conformation without compromising the overall topology. The results have significant implications in the mechanistic understanding of the hydrolytic stability of MOFs, and in preparing heterogeneous organocatalysts.


OPEN ACCESS
Supramolecular chemiStry 2018, Vol. 30,No. 9,[732][733][734][735][736][737][738][739][740][741] trans/cis, cis/cis -with the trans/trans conformation (also known as the syn, syn conformation) expected to be optimal for organocatalysis and guest binding (38,39). Herein, we describe the direct self-assembly and solid-state structures of three coordination polymers of LH 2 and different metal cations, Cd 2+ , Ca 2+ and Zn 2+ , discussing the effect of urea incorporation on structure, formation of novel inorganic SBUs, guest binding, and flexibility. We probe the guest binding properties of the Zn MOF with CS 2 as a mimic for CO 2 , and show that incorporation of the guest induces single-crystal to single-crystal (SCSC) transformations with notable changes in both linker conformation and SBU coordination chemistry. These results have significant implications in the understanding of activation, guest binding and hydrolysis of H-bond functionalised MOFs.

General
All chemicals and solvents were purchased from Alfa Aesar, Fisher Scientific, VWR, and Sigma Aldrich, and used as received.
The structures were solved using ShelxT (45) and refined against F 2 using Shelx2015 (46) within Olex2 (47). [Cd(L) DMF 3 ] n was treated as a two component twin related by a 2-fold rotation about the 100 direction. The twin component fractions refined to 0.421(6)/0.579 (6) giving significant improvement, however, probably due to some unaccounted for twinning, only the Cd atom was refined with anisotropic atomic displacement parameters (adps).
Direct synthesis of urea-containing MOFs, by introduction of urea moieties into organic linkers to target specific topologies in an isoreticular synthetic approach, is complicated by the possibility of structure-directing hydrogen bonding between the linkers inducing unexpected or unwanted MOF structures. For example, we have previously reported that the use of N,Nʹ-bis(4-pyridyl)urea instead of 4,4ʹ-bipyridine in solvothermal syntheses with Zn 2+ sources and dicarboxylic acids results in MOFs with interpenetrated diamondoid topology, rather than the expected pillared primitive cubic topology, as a consequence of hydrogen bonding between the urea groups of the pillar and the carboxylate groups of the ligand directing structure formation (Figure 1(b)) (37). The formation of hydrogen bonds between different parts of the MOF structure could mask potential catalytic sites, and so an understanding of how it can be avoided when using ligands such as N,Nʹ-bis(4-carboxyphenyl)urea (LH 2 , Figure  1(c)) is essential. Additionally, hydrogen bonding moieties often introduce structural flexibility; urea units have rotational freedom around the C-N-C moieties and can adopt (Figure 1(d)) three different conformations -trans/trans, For this structure SQUEEZE (48) was only used to calculate the solvent accessible volume (32 Å 3 ) however for all other structures reported herein SQUEEZE was used to calculate and account for the electron density within the solvent accessible void; details are given in Table 1. Disorder was present and modelled as two 0.5 occupied sites in one linker for both [Ca 5 (L) 5 (DMF) 3 (H 2 O) 2 ] n and [Zn 4 (L) 3 (DMF) (H 2 O)] n . Distance restraints were used in the case of the disordered fragments and for the solvent (DMF and CS 2 ) geometry. For the samples treated with CS 2 all the structures showed residual electron density on the Zn sites possibly due to unaccounted for twinning however treatment as a two component crystal was not satisfactory and was not used.
Synthesis LH 2 was synthesised by a modified literature procedure (49).
[Cd(L)(DMF) 3 ] n . Cadmium nitrate tetrahydrate (0.010 g, 0.032 mmol), LH 2 (0.009 g, 0.030 mmol) and N,Ndimethylformamide (DMF, 5 ml) were added to a 25 ml Pyrex reagent bottle and sonicated. The resulting solution was placed in the oven at 100 °C for 48 h. The bottle was removed from the oven after this period and allowed to cool to room temperature. The crystals were left to stand in their mother solution. [Ca 5 (L) 5 (DMF) 3 (H 2 O) 2 ] n . Calcium nitrate tetrahydrate (0.008 g, 0.033 mmol), LH 2 (0.010 g, 0.033 mmol) and DMF (10 ml) were added to a 50 ml Pyrex reagent bottle and sonicated. The resulting solution was placed in the oven at 100 °C for 48 h. The bottle was removed from the oven after this period and allowed to cool to room temperature. The crystals were left to stand in their mother solution.
Crystal Data for [Ca 5 (L) 5  Approximately 30% of the cell volume is not occupied by the framework and contains diffuse and disordered solvent molecules. This electron density was accounted for using SQUEEZE within PLATON (48) which calculated a solvent accessible volume of 2565 Å 3 containing 681 electrons (the equivalent of ~17 molecules of DMF) per unit cell. Crystal structure data are available from the CCDC, deposition number 1558144.
[Zn 4 O(L) 3 (DMF) 2 ] n -I. Zinc nitrate hexahydrate (0.050 g, 0.168 mmol), LH 2 (0.050 g, 0.167 mmol) and DMF (20 ml) were added to a 100 ml reagent bottle and sonicated. The resulting solution was placed in the oven at 90 °C for 24 h. The bottle was removed from the oven after this period and allowed to cool to room temperature. The crystals were left to stand in their mother solution. Crystals of [Zn 4 (L) 3 (DMF) 2 ] n -II were also isolated from this synthesis in the same container. As the two solvates are identical in connectivity and topology they were not separated, and referred to as [Zn 4 (L) 3  Approximately 33% of the cell volume is not occupied by the framework and contains diffuse and disordered solvent molecules. This electron density was accounted for using SQUEEZE within PLATON (48) which calculated a solvent accessible volume of 3164 Å 3 containing 905 electrons (the equivalent of ~22.6 molecules of DMF) per unit cell. Crystal structure data are available from the CCDC, deposition number 1558146.
A small amount of [Zn 4 (L) 3 (DMF) 2 ] n was removed from its mother solution by pipette (ca. 2 ml) and added to a scintillation vial containing CS 2 (ca. 5 ml). The CS 2 was exchanged for fresh CS 2 multiple times to remove as much DMF as possible. The CS 2 was replenished daily and single crystal X-ray diffraction data of daughter products, namely [Zn 4 O(L) 3  Approximately 10% of the cell volume is not occupied by the framework and contains diffuse and disordered solvent molecules. This electron density was accounted for using SQUEEZE within PLATON (48) (14) 18.3487 (7), 29.8698 (11), 16.0422 (7) 18.6280 (12) The seven coordinate SBU has only been observed in one other Cd 2+ coordination polymer -a related material with tetrabromoterephthalate linkers (50) -but derivatives where the DMF ligands are replaced by water to form coordination polymers are also known (51,52). The urea units of L 2clearly direct the formation of this structure, and would be unavailable for guest binding.
When LH 2 is combined with Ca(NO 3 ) 2 ·4H 2 O in DMF and heated to 100 °C for 48 h, a very small quantity of crystals of [Ca 5 (L) 5 (DMF) 3 (H 2 O) 2 ] n results. The structure contains infinite chains of calcium cations linked by carboxylate units of L 2-. There are five crystallographically independent Ca 2+ cations in the chains, but each has the same overall seven-coordinate, distorted pentagonal bipyramidal geometry (Figure 3(a)).
All crystallographic data are summarised and compared in Table 1. CCDC 1558143-1558149 contain the supplementary crystallographic data. These data can be obtained free of charge from The Cambridge Crystallographic Data Centre; see https://www.ccdc.cam.ac.uk/.

Results and discussion
The urea-dicarboxylate ligand, LH 2 , was prepared on the gram scale according to a modification of a literature procedure in two steps from t-butyl 4-aminobenzoate and carbonyldiimidazole (49). Attempts were subsequently made to prepare MOFs containing the ligand with free urea units to examine guest binding. Solvothermal synthesis with Cd(NO 3 ) 2 ·4H 2 O in DMF at 100 °C for 48 h resulted in the isolation of the one-dimensional coordination polymer [Cd(L)(DMF) 3 ] n . The material consists of chains of L 2molecules connected by seven-coordinate Cd 2+ cations with distorted pentagonal bipyramidal geometry (Figure 2(a)). The Cd 2+ centres coordinate to both oxygen atoms of the carboxylate units of two molecules of L 2-, linking them in a trans manner, with three DMF molecules occupying the remainder of the coordination sphere. The 1D chains that result have the urea N-H units projecting only in one direction for any given chain, and these form bifurcated H-bond interactions (Figure 2   an analogous ligand with a central amido group rather than a urea moiety (59). [Zn 4 O(L) 3 (DMF) 2 ] n -I is triply interpenetrated, with three identical nets nested within one another (Figure 4(b)), but surprisingly there are no netnet hydrogen bonding interactions between the urea and carboxylate functionalities. Instead, in the crystal structure each of the three crystallographically independent urea units binds a DMF molecule through bifurcated hydrogen bonding to the formamide oxygens (N1···O40T = 2.696, N2···O40T = 2.942 Å; N1A···O11S = 2.848, N2A···O11S = 2.863 Å; N1B···O21S = 2.967, N2B···O21S = 2.850 Å).
Interestingly, a second set of crystals was observed after the solvothermal synthesis with different morphologyagglomerates of plates (Figure 4(c)) -which single crystal X-ray diffraction revealed to be a closely related material, [Zn 4 O(L) 3 (DMF) 2 ] n -II, with identical connectivity and topology to [Zn 4 O(L) 3 (DMF) 2 ] n -I but a slightly different unit cell, presumably due to differing levels of solvation. Isolation of this additional phase is likely a consequence of the One carboxylate unit of L 2chelates with both its oxygen atoms to the Ca 2+ cation, with each oxygen coordinating to an adjacent calcium centre to overall bridge three Ca 2+ cations in a (η 2 :η 2 :μ 3 ) fashion. Two further carboxylate units of L 2molecules bridge from the Ca 2+ cation to adjacent Ca 2+ centres above and below, both in a (η 1 :η 1 :μ 2 ) motif. Each Ca 2+ therefore coordinates to six carboxylate oxygen atoms from five different ligands, with a final DMF or water molecule making up the coordination sphere (Figure 3(b)). Each molecule of L 2links the infinite 1D chains of Ca 2+ ions through a (η 2 :η 2 :μ 3 ) motif at one carboxylate and a (η 1 :η 1 :μ 2 ) motif at the other. The chains run down the crystallographic a axis, and are linked into an approximately hexagonal array by molecules of L 2-.
There are five crystallographically independent linker molecules, all of which experience some distortion from an idealised planar structure. Small triangular pores run down the crystallographic a axis (Figure 3(c)) and these spaces are filled in the crystal structure with a large number of water and DMF solvent molecules, with two of the five independent urea units binding water guests through a bifurcated H-bonding motif (N1C···O5 W = 3.381, N2C···O5 W = 2.940 Å; N1D···O4 W = 2.827, N2D···O4 W = 3.005 Å) and a further two urea units binding DMF molecules (N1···O61S = 2.832, N2···O61S = 2.801 Å; N1A···O56S = 3.377, N2A···O56S = 2.809 Å) in a similar manner (Figure 3(d)). There are also small voids between the chains perpendicular to the channels (Figure 3(e)). The 1D chains of Ca 2+ cations units in [Ca 5 (L) 5 (DMF) 3 (H 2 O) 2 ] n enforce a topology in which the urea moieties are not involved in any inter-ligand hydrogen bonding. Unfortunately we were unable to find suitable synthetic conditions to access any more than a few crystals per reaction, and so alternative MOFs were sought for study.
Solvothermal reaction of LH 2 with Zn(NO 3 ) 2 ·6H 2 O in DMF for 24 h at 90 °C yielded block-shaped crystals of [Zn 4 O(L) 3 (DMF) 2 ] n -I, which has the well-known IRMOF topology first reported by Yaghi in 1999 for MOF-5 (also known as IRMOF-1), [Zn 4 O(bdc) 3 ] n where bdc = 1,4-benzenedicarboxylate (53). The material has a slightly different SBU than the parent IRMOF structure; instead of four tetrahedral Zn 2+ cations linked by a μ 4 -O 2and six carboxylates, there are three tetrahedral Zn 2+ centres and one octahedral Zn 2+ cation with two additional coordinated DMF ligands (Figure 4(a)). This solvated SBU is rarely seen in IRMOF structures, but is known to occur when IRMOFs are prepared from linkers that are flexible (54,55) or deviate from ideal linear geometry (56,57). Whilst the L 2linkers lie along linear edge positions connecting adjacent Zn 4 O SBUs in the primitive cubic topology, the carboxyphenyl groups are disposed at approximately 145° rather than 180°, and the linkers also bow in and out of the plane, indicating significant flexibility (58). Indeed, [Zn 4 O(L) 3 (DMF) 2 ] n -I is structurally similar to NJU-Bai2, a Zn 2+ IRMOF prepared from  Figure 5(b)). It has been shown previously that DMF can dynamically bind to the metal cluster in the archetypal material MOF-5 (68), and these results clearly demonstrate that, even in MOFs where the DMF interacts with the SBU sufficiently to be located crystallographically, it can still be exchanged under mild conditions for alternative ligands. This partially hydrated cluster could also be considered as an intermediate species in the hydrolysis of the MOF, and indeed a model for all Zn MOFs that contain the basic zinc acetate SBU, which are known to be particularly susceptible to hydrolysis (69). Whilst the topology of the MOF remains unchanged, the flexibility is apparent when comparing [Zn 4 O(L) 3 (DMF)(H 2 O)] n to the parent structure. The ability of the ligand to distort, combined with the labile coordination chemistry of the SBU, allows the MOF to deviate from an approximately cubic arrangement ( Figure 5(a)) to a flatter structure reminiscent of a rhombohedron ( Figure 5(b)).
Leaving the same batch of crystals in CS 2 for a further 15 days, with daily replenishment of the CS 2 , resulted in a further SCSC transformation to [Zn 4 O(L) 3 ] n , where all coordinated solvents had been removed and the SBU has the conventional Zn 4 O(RCO 2 ) 6 composition ( Figure 5(c)). Additionally, guest exchange occurred, with significant quantities of CS 2 now located and ordered within the pores of the MOF, replacing weakly-bound DMF molecules. Tellingly, the only DMF that remains within the pores is hydrogen bonded to the urea units (N1···O6D = 2.893, N2···O6D = 2.903 Å; N1A···O1D = 2.860, N2A···O1D = 2.843 Å; N1B···O11D = 2.863, N2B···O11D = 2.847 Å), indicating the strength of the interaction, with the CS 2 located in the pores. While CS 2 may be a good geometric mimic of CO 2 , it is a weaker hydrogen bond acceptor, likely due to it having an opposite quadrupole moment and thus not being flexibility of the MOF allowing for crystallisation of closely related solvates from one synthesis. As the topology and structure of the two are identical, and because mechanical manipulation would be the only possible method of separation, we refer to the combined material as simply [Zn 4 O(L) 3 (DMF) 2 ] n in the remainder of the study.
We first reported the isolation of this material in our previous study of urea incorporation into MOFs (37) but did not fully describe the structure. During the preparation of this manuscript, a report detailing the synthesis of [Zn 4 O(L) 3 (DMF) 2 ] n and its catalytic activity in the Friedel-Crafts reaction between indole and β-nitrostyrene was published (29). The fact that the urea groups are not involved in 'host-host' hydrogen bonding facilitates this reported catalytic activity once the bound DMF molecules are removed, further highlighting the importance of limiting these interactions by topological control during the design of organocatalytic MOFs.
As part of our attempts to understand the availability of the urea groups for guest binding within the pores of this MOF, we attempted to replace the DMF solvents of crystallisation with CS 2 by soaking crystals of the MOF in this liquid CO 2 mimic. This approach has been successfully used to examine guest binding in coordination cages (60), but reported structures of MOFs with bound CS 2 remain relatively rare (61)(62)(63)(64)(65)(66)(67). As with the recently published study (29), [Zn 4 O(L) 3 (DMF) 2 ] n was found to be relatively unstable to solvent removal, but CS 2 is a volatile nonpolar solvent that may allow for efficient activation, and samples maintained crystallinity on solvent exchange from DMF to CS 2 . Single crystal X-ray diffraction analysis on a sample that had been soaked in CS 2 at room temperature for 4 days revealed that rather than exchanging bound DMF guests for CS 2   The trans/cis linkers are, however, significantly shorter than the trans/trans linkers: the SBU-SBU distances measured between central μ 4 -O 2ligands is 17.55 Å for the trans/ cis linker, as opposed to 19.68 Å and 19.94 Å for the two crystallographically independent trans/trans linkers, distances which are similar to those in the other Zn MOFs. The structure therefore is deviating from the cubic arrangement by pulling in two vertices on one side of the cube. We believe that the aforementioned flexibility of the material, derived from the L 2− ligands, allows it to endure such a significant change in linker length and geometry without breaking connectivity or altering interpenetration. This flexibility, combined with the labile coordination chemistry of the [Zn 4 O(RCO 2 ) 6 ] SBU, may, however, be responsible for the eventual framework collapse on drying -we were unable to successfully activate the porosity [Zn 4 O(L) 3 (DMF) 2 ] n to ascertain if the urea units enhanced CO 2 uptake.
The structure still has a rhombohedral-like arrangement, but seems visually less distorted than [Zn 4 O(L) 3 (DMF) (H 2 O)] n . To attempt to quantify the flexibility-induced distortion, the angles between adjacent SBUs were measured for each MOF, taking the bridging μ 4 -O 2units of the Zn 4 O SBUs as their centres, and collated in Table 2. For a perfect cubic structure, twelve angles of 90° would be expected, but the non-linear geometry of L 2− , coupled with its ability to bow out of the plane, means a range of angles are observed experimentally. The standard deviation was calculated as a metric to describe the magnitude of distortion, with larger standard deviations indicating more significant distortion from the idealised cubic arrangement. The two solvates of [Zn 4 O(L) 3 (DMF) 2 ] n have similar values, while [Zn 4 O(L) 3 (DMF)(H 2 O)] n has the largest standard deviation, with angles ranging from ~60° to 120°, which is obvious from inspection of the structure ( Figure 5(b)). [Zn 4 O(L) 3 ] n lies between the two extremes.
On leaving the batch of crystals for 6 more days in CS 2 , a more dramatic structural rearrangement occurs, again in a SCSC manner, to form [Zn 4 O(L) 3 (DMF)] n ( Figure 5(d)). One of the three crystallographically independent urea ligands undergoes rotation from the trans/trans conformation to the trans/cis arrangement, dramatically changing both the length and geometry of the linker (as well as its ability to bind guests or act as an organocatalyst) without altering the connectivity or interpenetration of the MOF (70). To accommodate the distortion, the SBU picks up a DMF molecule, leaving one of the Zn 2+ centres in a 5-coordinate trigonal bipyramidal geometry ( Figure  5(d)), very rarely seen in MOFs with the zinc acetate SBU, while the rest remain tetrahedral. This coordinated DMF molecule is likely one of those previously bound at the urea groups -only one of the three urea moieties now H-bonds a DMF guest -but exchange with the outer solvent cannot be excluded as a possibility. CS 2 molecules fill the pores (Figure 6(a)).  and two solvates of [Zn 4 O(L) 3 (DMF) 2 ] n , which has the prototypical IRMOF topology. In concordance with previous work, the urea moieties in [Cd(L)(DMF) 3 ] n direct structure formation through interligand H-bonding, but the topological restrictions placed on the urea units in the other two materials leave them free to bind guest solvent molecules, and an independent report published during our study described [Zn 4 O(L) 3 (DMF) 2 ] n to be an active organocatalyst as a result. CS 2 was used as a mimic to probe guest binding in [Zn 4 O(L) 3 (DMF) 2 ] n , resulting in a number of single-crystal to single-crystal transformations that illustrated (i) the lability of the [Zn 4 O(RCO 2 ) 6 ] SBU, capturing crystallographic snapshots in unusual states of solvation, and (ii) the flexibility of the framework, enabled by the facile coordination chemistry of the cluster and the flexibility of the organic ligand. Three additional crystal structures showed the MOF breathing as it exchanged DMF pore solvent for CS 2 , eventually resulting in a dramatic configurational rearrangement of some of the urea units from the trans/trans to the trans/cis conformation, which are not capable of binding guests for organocatalysis. These results provide valuable structural insights into the dynamic behaviour of MOFs with flexible linkers upon activation, which have significant implications for the development of heterogeneous organocatalysts, and also give evidence for potential hydrolysis mechanisms. Additionally, we anticipate that CS 2 may be considered a mild alternative solvent for activating MOFs in future; whilst [Zn 4 O(L) 3 (DMF) 2 ] n did collapse on solvent removal, CS 2 was the only solvent which penetrated its pores without damaging single crystals.