Guest inclusion by Borromean weave coordination networks

Abstract Reaction of N,Nʺ-ethylene-1,2-diylbis(3-pyridin-3-ylurea) (L) with AgNO3 in a variety of solvents gives a total of five 2-D Borromean weave coordination polymer networks that adopt three structural types depending on the interactions to the solvent pocket. The Borromean network is of formula [Ag2(L)3](NO3)2 ·solvent where the solvent is either a cluster of water molecules, mixtures of water and acetonitrile, water and methanol or chloroform and methanol. The Borromean structure is a thermodynamic sink and under fast crystallization conditions an alternative 2 + 2 metallomacrocycle forms that can result in metallogel formation. The metallogel structure transforms into the crystalline Borromean network over time. Graphical Abstract


Introduction
The Borromean motif is an example of a prime link, formally 6 3 2 in Alexander-Briggs notation [1,2]. Borromean topologies in chemistry are rare but have been experimentally realized for DNA nanostructures [3], discrete metallosupramolecular assemblies [4,5], 2-D coordination polymer sheets [6][7][8][9][10][11][12], and hydrogen-and halogen-bonded structures [5,[13][14][15]. Mutually entangled topologies in chemical systems are of considerable interest because of changes in properties such as mechanical robustness, framework cooperativity, and dynamical non-trivial behavior that can arise from mechanical bonding. Interest in this topic increased significantly with the award of the 2016 Nobel Prize in chemistry in the area of molecular machines [16][17][18]. While the Borromean topology and topological links in general such as the 2 2 1 (Hopf), 4 2 1 (Solomon), 6 2 1 (Star of David), 6 3 3 (cyclic [3]catenane), and a 12 4 1 [4]catenane have been engendered in chemical systems [19,20], it is more unusual for molecular link systems to display host-guest chemistry. Previously, Barbour reported 2-D Borromean weave coordination polymers based on silver(I) complexes of extended difunctional ligands that are held together with argentophilic interactions [7], while we have reported a related example that includes a discrete water heptamer in a well-defined guest binding pocket [21]. This latter compound is of formula [Ag 2 (L) 3 ](NO 3 ) 2 Á7H 2 O (1) and arises from the reaction of AgNO 3 and N,Nʺ -ethylene-1,2-diylbis(3-pyridin-3-ylurea) (L) in aqueous DMF or aqueous THF. Both Ag I ions in the asymmetric unit act as nodes in a network made up of large hexanuclear rings ( Figure 1) formed from the doubly connected, dipyridyl ligands. The 2-D coordination polymer structure is supported by argentophilic interactions between pairs of stacked Ag I cations (AgÁÁÁAg 3.54 Å). The nitrate anions form infinite stacks, with each anion surrounded by a total of three urea groups from three different ligands, each belonging to a separate sheet ( Figure 1). Each ureaÁÁÁnitrate interaction is an eight-membered hydrogen-bonded ring, and the hydrogen bonding is saturated [22] as each anion is surrounded by six NH hydrogenbond donors. Along with the nitrate cavity, there is an additional cavity that is occupied by a seven-membered water cluster that forms long AgÁÁÁO interactions and strong OHÁÁÁO interactions with the urea carbonyl groups. TGA analysis indicates that the water guest is very tightly bound in this system [21]. We now report a series of related host-guest complexes based on the same Borromean motif as in 1 in which the basic 2-D Borromean weave framework is able to accommodate a variety of guest species.

Results and discussion
In total, N,Nʺ -ethylene-1,2-diylbis(3-pyridin-3-ylurea) (L) forms five Borromean weave inclusion complexes upon slow crystallization in various solvent mixtures with silver(I) nitrate. Within each structure, the basic Borromean weave is similar but the argentophilic interactions and the way in which the 2-D Borromean sheets stack is governed by the included solvent guests. When the nitrate counter anion is replaced by either acetate, tetrafluoroborate, or hexafluorophosphate, several coordination polymers can be prepared, but the Borromean structure is not reproduced and hence, the Borromean topology is likely to arise from the very specific saturated hydrogen bonding interaction to the trigonal nitrate anion. A similar dependency on counter anion  [21]. (Color code: red ¼ oxygen, blue ¼ nitrogen, grey ¼ carbon, white ¼ hydrogen, cyan ¼ silver).
When the crystallization of L in the presence of silver(I) nitrate was performed in a selection of other solvents in addition to the conditions used for 1, four additional Borromean solvate materials were obtained. Their formulas were determined by elemental analysis and/or X-ray crystallography and are shown in Table 1. For 2 and 4, the solvent content is more or less ordered and unambiguous. Solvent proved to be highly disordered for 3 and 5, and in a preliminary report of the structure of 3, solvent was not modeled [21]. The solvent was handled using the solvent masking implementation of the SQUEEZE [24] procedure in Olex2 [25]. This gave an estimate of the number of electrons in the structure arising from solvent. For both complexes, this proved to be lower than the solvent content implied by elemental analysis. It is likely that the complexes can include variable amounts of solvent within the crystal void and that the amount from elemental analysis (which used samples with limited drying in order to preserve solvent content) is an overestimate as a result of surface solvent, while the solvent content from the solvent masking procedure is correct for each crystal studied but likely an underestimate of the total capacity as a result of partial desolvation during crystal handling.
The four Borromean inclusion complexes 2-5 are closely related to 1 (although 3-5 are not isomorphous with 1) with pairs of silver(I) ions linked by argentophilic interactions within each 2-D Borromean sheet. Complex 2 is isomorphous with 1 in P2 1 /c with a similar crystal packing arrangement in which two of the ordered water guests have been replaced by one molecule of methanol. The methanol hydrogen bonds only to water molecules, while the water forms one long coordinate bond to one of the two independent silver ions in the same way as 1; Ag-O 2.79 Å. Complex 3 has a similar-sized unit cell to 1 and 2 but is more symmetrical with a single independent silver(I) ion in space group C2/c. The ordered acetonitrile forms a long coordinate bond to the silver ion Ag-N 2.71 Å, and the remaining solvent occupies a disordered pocket. The lack of specific hydrogen bonding to the carbonyl ligands and guest-guest interactions is likely to be the cause of the higher symmetry. The AgÁÁÁAg distance is reduced to 3.40 Å (from 3.54 Å in 1 and 3.59 Å in 2). A single solvent cluster is able to coordinate to two Ag 2 argentophilically bonded units in adjacent weaves. Thus, channels of nitrate anions are produced and each Ag 2 unit stacks alternately with a solvent cluster. Complex 4 combines elements of 2 and 3. As in 3, the acetonitrile molecules are coordinated to the silver ion Ag-N 2.56 Å allowing a short AgÁÁÁAg interaction of 3.40 Å. The space between the two back-to-back acetonitrile molecules in adjacent Borromean sheets is occupied by a discrete, ordered (H 2 O) 8 water cluster, and hence, the whole (MeCN) 2 (H 2 O) 8 solvent pocket is similar in size to twice that of the (H 2 O) 7 cluster in 1. Creating this "double cavity" results in significantly different crystal packing in which the two Borromean sheets are shifted with respect to one another.
Structure 5 is particularly interesting since it involves the shortest argentophilic interaction of just 3.25 Å. This arises because the silver ions are not coordinated to any solvent molecule on the other side of the AgÁÁÁAg pair and instead, a urea nitrogen atom of a ligand on an adjacent Borromean sheet forms a 3.48 Å long contact to the silver ions. This lack of solvation serves to increase the strength of the argentophilic interactions and give rise to a further different packing type in which the 2-D Borromean sheets are more closely interdigitated. However, a total of four disordered solvent-filled voids per unit cell remain, each of volume 454 Å 3 filled with disordered water and acetonitrile. While the solvent was refined using the solvent mask procedure, the position of the individual Fourier peaks suggests that there is a water molecule in each layer that is hydrogen bonded in between a pair of urea carbonyl groups and in turn interacts with another solvent water molecule, which lies between the layers, an occupant of both.
Within all the structures, the solvent clusters are either at the end of or beside the AgÁÁÁAg argentophilically bonded units and depending on how the layers stack, each Borromean contains a minimum of a four-anion nitrate channel. As with 1, the layers in 2 and 3 are stacked to form infinite nitrate channels and the solvent clusters always have an Ag 2 unit adjacent to them on both sides, as shown in Figure 2.
For 4 and 5, the alternate stacking of Ag 2 units and solvent clusters is not present and a translation between the layers removes the nitrate channels. In 4 where the solvent cluster comprises five molecules of water and one acetonitrile, the layers are arranged so that the Ag 2 -solvent clusters are stacked alternately with a discrete four- anion nitrate channel, as shown in Figure 3, possibly driven by the lack of a second molecule of acetonitrile to allow adjacent layers to interact. This arrangement, as well as the out-of-plane twisting of the pyridyl groups at the Ag I centers, allows adjacent layers to stack via p-p stacking ( Figure 4).
Interestingly, the Borromean topology is not an inevitable outcome of the reaction of AgNO 3 with L. We have previously reported that the reaction of L with AgNO 3 gives an apparently kinetic product from a mixture of acetonitrile, chloroform, and methanol (3:3:4) in which crystals form within five minutes of standing [26]. This solvent mixture is essentially the same as that used to form Borromean 3 but using a more concentrated sample. This material is a [2 þ 2] metallomacrocycle and seems to be the basis for metallogel formation ( Figure 5). Interestingly, crystals of Borromean 1 form upon standing a metallogel sample in aqueous tetrahydrofuran. However, the XRPD pattern of the dried xerogel implies that the gel structure is based on the [2 þ 2] metallomacrocycle, not the Borromean motif [26]. Hence, gels based on metallomacrocycles and Borromean weave coordination networks represent alternative outcomes of the reaction with the former likely to be the kinetic product and the latter the thermodynamic sink.

Conclusion
The isolation of the same core Borromean structure from five different solvent media under slow crystallization conditions suggests that the Borromean weave is a robust structural motif and represents the thermodynamic product of the AgNO 3 /L system. Under fast crystallization conditions, previously reported metallomacrocyclic or metallogel products are isolated [26]. The Borromean weave 2-D network is based on pairs of silver(I) ions linked by argentophilic interactions and can adopt three different structure types. The most common involves a solvent pocket in between silver pairs capable of binding ordered clusters of water or mixtures of water and other solvents. An alternative structure allows two such pockets to merge giving a larger pocket capable of accommodating an (MeCN) 2 (H 2 O) 8 cluster. Ordering is driven by weak interactions to the silver(I) ions and hydrogen bonding to the urea carbonyl oxygen atoms. A third structure type involves significantly stronger argentophilic interactions with d(AgÁÁÁAg) ¼ 3.25 Å with the exposed face of the silver occupied by a long contact to a urea nitrogen. The hexanuclear ring in 1 is approximately threefold symmetric with AgÁÁÁAg separations of 15.75 Å, 16.20 Å, and 16.48 Å and a diameter of ca. 31.5 Å. These dimensions vary by up to 0.3 Å across the various solvates as the Borromean weave motif adjusts to the different included solvents and packing arrangement. The reproducibility of this topologically complex network and its tolerance to a variety of guests highlight the complexity of the crystallization process in these labile systems, and the time needed to engender topological complexity may allow alternative, simpler kinetic products including gels based on a metallomacrocyclic motif to form.