Building inorganic supramolecular architectures using principles adopted from the organic solid state

The possibility of converting crystal engineering principles that originate in the organic solid state into productive strategies for the effective supramolecular assembly of coordination complexes has been established through a range of complementary techniques. Calculated molecular electrostatic potentials, in conjunction with a systematic structural study, demonstrate that the existence and structural importance of the key hydrogen-bond interactions are not disrupted by the presence of metal cations and charge-balancing anions.


Introduction
One of the primary challenges standing in the way of more predictable supramolecular assembly of metal-containing structures with precise metrics and topologies is the lack of transferable and robust guidelines for pre-planned synthesis based on non-covalent interactions (Brammer, 2004;Hosseini, 2005;. In order to make us less reliant on serendipity when targeting a desired supramolecular architecture, we need to map out carefully the structural landscape that describes how different molecules recognize, interact and communicate (in a structural sense), even in the presence of the potentially disruptive counter-ions that regularly appear in metal-organic systems (Aakerö y et al., 2009;Ðaković et al., 2011).
A relatively common approach to crystal engineering comprising coordination complexes employs ligands that are capable of forming reliable coordination-covalent bonds with metal ions, and which at the same time can participate in selfcomplementary non-covalent interactions (typically hydrogen bonds) that promote the directed assembly of discrete complex ions into one, two or three-dimensional architectures (Brammer et al., 2002;Kukovec et al., 2016). Although such strategies are relatively straightforward and logical, and the targets are seemingly attainable and realistic, the reversibility of non-covalent interactions often conspires to produce outcomes that are neither desirable nor expected (Ðaković et al., 2013).
The use of functional groups capable of self-complementary non-covalent interactions for propagating the assembly of discrete building blocks into extended architectures is motivated by the fact that some chemical functionalities can be relied upon to be structurally consistent (Sarma & Desiraju, 2002). The extent to which a particular functional group is likely to form a specific synthon (Desiraju, 1995) can be elucidated through careful analysis of the Cambridge Structural Database (CSD) (Groom et al., 2016;. In this way, the hydrogen-bond preferences of numerous functional groups have been evaluated and quantified in order to develop supramolecular synthetic routes, especially in crystal engineering involving organic molecular solids (Shattock et al., 2008;Aakerö y, Sinha et al., 2013;Vishweshwar et al., 2002).
If we want to maximize our chances of precisely controlling the main structural features of the product of a particular directed-assembly process, it is not sufficient simply to know how likely it is that a given functional group will form selfcomplementary interactions (Desiraju, 2001). In addition, since some functional groups are capable of forming geometrically different homomeric synthons, we also need to establish the relative propensities of these different structural options. For example, carboxylic acids and oximes prefer head-to-head R 2 2 ð8Þ and R 2 2 ð6Þ motifs, respectively, but both groups are also capable of producing one-dimensional catemeric chains, C(4) and C(3), respectively (Figs. 1a and 1b) [For full details of graph-set analysis of hydrogen bonds, see Bernstein et al. (1995)].
Synthon polymorphism (Sreekanth et al., 2007;Gryl et al., 2008;Mukherjee & Desiraju, 2011) can have dramatic structural effects on the exact nature of the overall resulting assembly, which in turn can produce solid forms with unwanted or sub-optimum properties. It is therefore very important, from a supramolecular synthetic perspective, to know (i) what the relative frequency of occurrence is for each option and (ii) what the resulting structural consequences are for each synthon type, because the properties of the bulk are directly connected to the structural consequences of the synthons. Conventional polymorphism highlights in similar ways the critical connection between structure and function and it is well known that manufacturers of high-value organic solid chemicals, such as the pharmaceutical industry, spend enormous resources on making sure that the solid-form landscape for each target is completely mapped out.
In this study, we focus our attention on three molecules that present a carbonyl group adjacent to a heterocyclic nitrogen atom, thereby creating a functionality that can engage in either head-to-head dimers or catemeric chains (Fig. 1c). The three tectons, 2(1H)-pyrazinone (2-pyz), 4(3H)-pyrimidinone (4-pym) and 4(3H)-quinazolinone (4-quz) (Fig. 2), are all multi-functional in the sense that they can act as ligands in transition metal complexes while delivering vectors for supramolecular assembly through hydrogen-bond based synthons.
In order to make reliable inferences from molecular structure to crystal structure it is necessary to minimize the potential for compositional and structural diversity that is often present in metal-containing systems. Therefore, we incorporated a structurally reliable invariant that would act as a robust foundation from which we could directly examine the non-covalent aspects (synthon preferences) of these pyridinone-type ligands. We opted to build the central structural invariant around a Cd II halide framework as this would present a reasonably predictable one-dimensional coordination polymer (Englert, 2010) (Fig. 3).
In addition, the octahedral coordination requirement of Cd II cations offers the necessary anchoring points for two auxiliary ligands in the axial positions (Fig. 4). The supramolecular chemistry of these ligands can then be examined in detail with a view to developing new strategies for reliable and  Hydrogen-bonded dimers and catemers of (a) carboxylic acids, (b) oximes and (c) 2(1H)-pyridinone.   The targeted one-dimensional Cd II -based coordination polymer (M = Cd, X = halide ion and R = 2-pyz, 4-pym or 4-quz.
versatile programmed assembly of metal-containing solidstate architectures.
The goals of this study are to control the details of the coordination chemistry around each metal cation and then to deliver a coordination network of consistent dimensionality and topology through complementary hydrogen bonds. We also want to determine if a relatively simple electrostatic view of the hydrogen bond will allow us to rationalize the observed interactions in metal-containing systems with a number of different (and potentially disruptive) hydrogen-bond acceptor sites.

Materials and methods
All metal salts, precursors and solvents were purchased from commercial suppliers and used without further purification.
Liquid diffusion synthesis. An aqueous solution of the cadmium(II) salt (1 ml) was layered with an ethanol solution of the ligand (2 ml). Solutions were further layered with acetone (1.5 ml) and left standing at room temperature. X-ray quality crystals were harvested after a period of 7-15 d.
Mechanochemical synthesis. Cadmium(II) salts and ligands in an equimolar ratio were ground in a stainless steel jar (10 ml in volume; using two stainless steel grinding balls of 7 mm in diameter) with the assistance of 100 ml of ethanol. A Retch MM200 grinder operating at 25 Hz frequency was used for the grinding experiments.
Full details of the syntheses can be found in the supporting information.

X-ray structure analysis
Single crystals were mounted in a random orientation on a glass fibre using Paratone oil. Data collections were carried out on an Oxford Diffraction Xcalibur four-circle kappa geometry single-crystal diffractometer with a Sapphire 3 CCD detector using graphite monochromated Mo K ( = 0.71073 Å ) radiation and applying the CrysAlisPro Software system (Agilent Technologies Ltd, 2014; Rigaku Oxford Diffraction Ltd, 2015), at 296 (2) K for compounds 2, 3, 5 and 9, 200 (2) K for 1 and 6, or 150 (2) K for 8. Data reduction, including absorption correction, was done using CrysAlisPRO. The structures were solved with SHELXS-2014 (Sheldrick, 2015) and refined on F 2 using SHELXL-2014. Table S1 in the supporting information reports all the crystal data, details of the data collection, and a full summary of the intensity data collection and structure refinement for the crystal structures of 1-3, 5, 6, 8 and 9. Displacement ellipsoid plots were drawn at the 50% probability level (Mercury 3.3; Macrae et al., 2008). CCDC Nos. 1559137-1559143 contain the supplementary crystallographic data. These data can be obtained free of charge from the Cambridge Crystallographic Data Centre via http://www.ccdc.cam.ac.uk/data_request/cif.

Computational details
The geometry of the coordination polymer unit was optimized by employing a one-dimensional periodic boundary condition (PBC) in GAUSSIAN09 (Frisch et al., 2009). The Perdew-Burke-Ernzerhof exchange and correlation functional (PBE) (Perdew et al., 1996(Perdew et al., , 1997 was used, together with the Stuttgart/Dresden effective core potential basis set (SDD). After optimization, the one-dimensional coordination polymer was modelled as a unit with seven octahedrally coordinated cadmium centres to obtain more accurate molecular electrostatic potential (MEP) values. Two monocationic metals (like Na + ) were placed on each of the two ends, ensuring zero total charge of the molecule. With an increasing number of side molecules, the MEP values of the central unit atoms converged to reported values and the effect of the infinite chain termination almost vanished. More details about this will be published in a forthcoming paper. The MEP maps were visualized in GaussView 5.0 (Dennington et al., 2009). The MEP at a specific point on the 0.002 a.u. isodensity surface is given by the electrostatic potential energy (in kJ mol À1 ) that a positive unit charge would experience at that point. A continuous colour spectrum is used to assign different values of electrostatic potential energy values. The most The two synthon types, (left) the R 2 2 ð8Þ dimer and (right) the C(4) catemer, formed between two basic metal-based building units equipped with pyrazinone ligands. The pyrimidinone and quinazolinone ligands are prone to exhibit the same hydrogen-bond motifs.
negative values are coloured red and the most positive values are coloured blue.
The geometries of tetramers and molecular pairs were extracted from experimental crystallographic data after normalization of bond lengths with hydrogen to values obtained from neutron diffraction experiments. Single-point calculations and optimizations of ligands were performed in GAUSSIAN09 using the M06-2X (Zhao & Truhlar, 2008) functional with the def2-TZVP basis set (Feller, 1996;Schuchardt et al., 2007). A quantum theory of atoms in molecules (QTAIM) analysis (Bader, 1990) was performed with the AIMAll programme (Todd, 2016).

Structural studies
A series of cadmium(II) halide complexes with 2(1H)pyrazinone (1-3), 4(3H)-pyrimidinone (4-6) and 4(3H)quinazolinone (7-9) was prepared via two synthetic routes, (i) a liquid diffusion synthesis from a water-ethanol-acetone mixture, and (ii) ethanol-assisted grinding of a mixture of the starting compounds [i.e. cadmium(II) salts and organic ligands] in an equimolar ratio (Fig. 5) (Braga & Grepioni, 2005;Friščić, 2010). Both routes resulted in the same complexes in all nine cases 1-9, emphasizing the highly robust and effective nature of the synthetic protocol. To avoid any potential solvent-solute bias, the same layering procedure was used for all the liquid diffusion experiments (2:4:3 waterethanol-acetone layers). X-ray quality single crystals were harvested for all complexes (apart from 4) from the diffusion experiments after a period of 1-2 weeks.
The structure determinations confirmed the expected onedimensional polymeric chains as primary building units, with the cadmium(II) cations being octahedrally coordinated to four bridging halide ions and two trans-oriented organic ligands (Fig. 6). The ligands are in both cases coordinated via the nitrogen atom distant to the carbonyl oxygen atom, i.e. the nitrogen atom meta to the pyrazinone (1-3) and para to the pyrimidinone (5-6) oxygen atom. The only structural variances within 1-3 and 5-6 arise from the differences in the halide radii that affect the Cd II Á Á ÁCd II distances which, in turn, cause an increased 'tilting' of the aromatic rings upon moving from 1 to 3, and from 5 to 6.
Complex 8 displays essentially the same crystal structure as that previously reported for 7 (refcode NALFEN; Turgunov & Englert, 2010) and comprises targeted one-dimensional polymeric CdBr 2 chains as core units with the expected octahedral geometry around the cadmium(II) cations, including a transarrangement of the two quinazolinone ligands (coordinated via the nitrogen atom para to the quinazolinone oxygen atom) (Fig. 9a).
Finally, the structure determination of 9 reveals an unusual monomeric metal-containing building unit with an unexpected tetrahedral arrangement of two quinazolinone and two iodide ligands (Fig. 9b). Interestingly, the quinazolinone ligands, being coordinated via N1 in 9, display a different coordination mode compared with that observed for 7 and 8. It is the only complex in the series with metal coordination via the nitrogen atom ortho to the carbonyl oxygen atom.
Neighbouring polymeric units in 7 and 8 are linked via two hydrogen bonds of the N-HÁ Á ÁO type forming head-to-head R 2 2 ð8Þ motifs (Fig. 10a). Adjacent monomers of 9 are interconnected via the same supramolecular link, N-HÁ Á ÁO hydrogen bonds, just forming somewhat different supra-molecular motifs in the form of extended chain-like Cð6ÞR 2 2 ð16Þ catemers (Fig. 10b).

4-pym and 4-quz)
is occupied through coordination to the metal centres. Therefore, as it would not have an impact on supramolecular motif formation, it was not included in the search fragments (Fig. 11).
The results showed that the ring motif R 2 2 ð8Þ is more common than the C(4) catemer as it appears in 184 out of 299 times (62%) for the pyridine-type fragment, and in 70 out of 100 times (70%) for the quinoline analogue. The catemeric motif is present in only 25/299 (8%) and in 9/100 (9%) of the pyridine and quinoline fragments, respectively (Fig. 11).

Computational study
To rationalize the supramolecular outcome of these reactions against a backdrop of MEP surfaces, while avoiding computationally expensive optimizations of the entirety of the crystal structure, we optimized the geometry of a onedimensional coordination polymer unit by employing periodic boundary conditions. Then, a finite model of seven octahedrally coordinated cadmium centres with zero total charge was employed to obtain the more accurate MEP values. More details regarding the procedure are provided in the Experimental section. Fig. 12 shows the MEP maps for the polymeric fragments comprising seven cadmium(II) centres, mimicking an infinite coordination polymer chain. By employing three additional units placed on each side of the central cadmium(II) ion, the effect of deliberate termination of the polymeric chain is significantly reduced. The MEP values for the two potential acceptor sites, the coordinated halide anions and the carbonyl oxygen atoms, as well as the two most positive surface potentials for each compound are listed in Table 2. In all nine cases, the carbonyl oxygen atom carries the most negative potential and the halide anion the second most negative one, while the site of the most positive potential is always the hydrogen atom attached to an endocyclic nitrogen atom. The second most positive potential site is located on an aromatic hydrogen atom, the position of which varies depending upon which ligand was used, and is as indicated in Fig. 12.

Discussion
Both mechanochemical methods and liquid diffusion syntheses, starting from cadmium(II) halide salts and the relevant ligands, resulted in one-dimensional coordination polymers as simple metal-based building units containing a central Cd-X structural spine equipped with trans-positioned bi-functional ligands. We succeeded in generating this structural core in eight out of nine synthetic attempts. Only the reaction of cadmium(II) iodide and quinazolinone yielded a monomeric tetrahedral complex, 9. The complex also exhibits a uniquely observed coordination mode of the ligand within the studied series of complexes: it coordinates via the nitrogen atom ortho to the carbonyl group. None of our alternative synthetic  A survey of the CSD revealed a pronounced tendency for R 2 2 ð8Þ synthon formation for the pyridine-and quinoline-type fragments in purely organic solids. Table 2 Calculated MEP values (in kJ mol À1 ) on the two best hydrogen-bond donor and acceptor sites for compounds 1-9, as indicated in Fig. 12

Figure 12
The positions of the atoms with the most positive (blue) and most negative (red) molecular electrostatic potential values for (a) pyrazinone-(in complexes 1-3), (b) pyrimidinone-(in complexes 4-6) and (c) quinazolinone-based units (in complexes 7 and 8, and in the hypothetical model of complex 9) presented for polymer fragments consisting of seven cadmium(II) centres for compounds 2, 5 and 8, respectively. Isodensity surface of 0.002 a.u., colour range À215 (red) to 310 kJ mol À1 (blue). efforts (neither a solvothermal and mechanochemical synthetic procedure, nor supramolecular protection of the undesirable binding site) resulted in a structure with metal coordination via the nitrogen atom positioned meta or para to the carbonyl group, as was observed in all other complexes 1-8. Even though our control of the details of the coordination chemistry in this series was very satisfying [a success rate of 89% (8/9) in preparing one-dimensional metal-containing polymeric chains], at the supramolecular level the desired connectivity was accomplished with a 100% success rate. In all examined compounds the basic building units were interconnected via the targeted primary type of non-covalent interaction, the single-point N-HÁ Á ÁO hydrogen bond. Neither the dimensionality of the building block nor the coordination mode of the ligands prevented the appearance of this hydrogen bond. Moreover, even the halide anions, common counter-ions in metallo-supramolecular systems that are potentially disruptive to synthons which have already been well established in metal-free systems, did not exert any detrimental influence on the supramolecular synthesis. Generally, they carry a substantial negative charge that enables them to compete with hydrogen-bond acceptors that are already present as a part of the supramolecular functionality. In addition, the fact that they are also engaged as metalbridging ligands reduces their overall negative charge to some extent, which should suppress their ability to compete for hydrogen-bond donors in the system. As a result, the required and intended N-HÁ Á ÁO hydrogen bonds remained in place in all nine cases, suggesting that the C O group is the most effective hydrogen-bond acceptor in these compounds.
To give an interpretation of the supramolecular observation, we calculated MEP surfaces (Hunter, 2004) in order to rank potential hydrogen-bond donors and acceptors according to the MEP values (Etter, 1990(Etter, , 1991. In all nine cases, the carbonyl oxygen atom and the hydrogen atom attached to the endocyclic nitrogen atom display the most negative and the most positive electrostatic potential value, respectively. Therefore, we can consider the two sites as the best hydrogenbond acceptor and donor, respectively (Aakerö y & Kanisha, 2014). The fact that the preferred interaction systematically occurs between these two sites suggests that the 'best donorÁ Á Ábest acceptor' guideline (Etter, 1990) can also be applied to carefully tailored metal-organic systems.
According to structural information extracted from the CSD, the supramolecular 'driver' used in this study (formed by having the carbonyl group directly adjacent to a heterocyclic nitrogen atom) is predisposed to form dimers in metal-free systems. However, in our metal-organic systems the cyclic R 2 2 ð8Þ motif occurs in only two out of nine cases, while the catemeric C(4) motif dominates and appears in six out of nine cases. This is, in effect, an example of synthon crossover (Aakerö y et al., 2011), which requires further examination.
To better understand the reasons for this synthon crossover we first performed a QTAIM analysis. While both supramolecular motifs, R 2 2 ð8Þ and C(4), contain the same number of N-HÁ Á ÁO interactions per molecule (polymeric unit), the analysis showed that all crystal structures involving the catemeric C(4) motif (1-6) also display an additional 'second-best donorÁ Á Ásecond-best acceptor' interaction (C-HÁ Á ÁX, Fig. 13), which is not present in structures dominated by the R 2 2 ð8Þ synthon. Instead, the second-best donor in those assemblies participates only in an interaction with the remaining lone pair of the best hydrogen-bond acceptor, namely the carbonyl oxygen atom.
Furthermore, for the three sub-classes of complexes in this series (Fig. 14) we compared the two outcomes, the dimeric and catemeric ones, to try to discern if any other close contacts would favour one particular assembly or synthon formation. If the R 2 2 ð8Þ motif were to occur for the pyrazinone-(1-3) and pyrimidinone-containing (4-6) complexes, it would result in a much closer positioning of the two carbonyl O atoms (shown as red regions in the encircled part of Figs. 14a and 14b) compared with those in the observed structures (Figs. 14d and 14e), and this in turn would lead to increased electrostatic repulsion. In contrast, when the quinazolinone ligands are involved, the added spatial requirements of the second fused aromatic ring ensures that the two carbonyl oxygen atoms are separated when forming the R 2 2 ð8Þ motif (Fig. 14c). In addition, there are also stabilizing short contacts with hydrogen atoms from the 'counter-surface area' (green to pale-blue region, QTAIM analysis for a representative of each sub-class of complexes [CdBr 2 (2-pyz) 2 ] n (2), [CdBr 2 (4-pym) 2 ] n (5) and [CdBr 2 (4-quz) 2 ] n (8), showing the bond-critical points as green dots. Interactions between the best donors and best acceptors are marked with blue ovals, while interactions involving the second-best donors are marked with yellow ovals. QTAIM analysis results for all the complexes can be found in the supporting information (Figs. S16-S19). Fig. 14c and supplementary Fig. S19). In the case of forming the catemeric motif C(4), again unfavourable contacts are to be expected (Fig. 14f).
This detailed inspection of secondary hydrogen-bond interactions and of specific steric requirements clearly indicates that synthon crossover is controlled by a fine balance between close packing and the secondary hydrogen bonds that can form once the primary hydrogen-bond sites have been satisfied.

Conclusions
Within the chemical and materials sciences, considerable efforts and resources are focused on elucidating structurefunction correlations in order to understand the direct relationship between the relative arrangement of atoms and molecules and the properties that they display as a collective. However, much less work has been dedicated to developing reliable supramolecular synthetic pathways for actually organizing molecular building blocks into extended architectures with specific dimensionalities and topologies.
In this comprehensive study we combined synthesis, structural chemistry and cheminformatics with complementary theoretical tools in order to establish if crystal engineering principles that originate in the organic solid state can be adapted to the unique requirements and challenges posed by supramolecular synthesis in the coordination chemistry arena. The transition from organic to inorganic crystal engineering was accomplished through the design and structural implementation of 'bi-functional' ligands that are simultaneously capable of (i) interacting with metal ions in such a way that their coordination number and geometry remain invariant and predictable, and (ii) engaging in self-complementary hydrogen bonds that are highly directional and relatively insensitive to potentially interfering or competing sites. The main supramolecular products that were targeted in this systematic study appeared in each and every one of the nine reported structures, which represents a very satisfactory outcome.
In order to test the durability of the synthetic strategy, we subsequently explored drastically different reaction conditions (solution-phase synthesis versus solvent-assisted grinding), but on no occasion did we observe a structurally different product, which emphasizes the robustness of the synthetic protocol.
The choice of self-complementary and catemeric synthons (responsible for the supramolecular assembly) in this study was guided by a simplified electrostatic view of hydrogenbond interactions, and the results from the calculated MEP values showed that synthon transferability from organic to metal-organic systems is possible, since the relative importance and ranking of the different hydrogen-bond donors/ acceptors remained the same even after the introduction of metal cations and charge-balancing anions. These results further support the hypothesis that a synthetic protocol built upon (i) the 'best donorÁ Á Ábest acceptor' guidelines and (ii) a ranking of potential hydrogen-bonding acceptors/donors on the basis of MEP values (which has found considerable value in organic crystal engineering) can be successfully implemented for directing the effective supramolecular assembly of metal-organic systems. We expect that many more synthons and reproducible molecular recognition events that have initially been identified and explored in the organic solid state can be employed as synthetic vectors for the hierarchical assembly of metal-containing materials of considerable complexity and across a variety of length scales.

Figure 14
The observed (green ticks) and hypothetical (red crosses) models (assembled starting from one-dimensional coordination polymers represented by one metal centre for clarity) of the pyrazinone-(in compounds 1-3), pyrimidinone-(in compounds 4-6) and quinazolinone-based units (in compounds 7 and 8) of the single-coordinated cadmium centre for compounds 2, 5 and 8, respectively.