R. Robson*
School of Chemistry, University of Melbourne, Parkville, 3050, Victoria, Australia. E-mail: r.robson@unimelb.edu.au; Tel: +61 (0)383446469
First published on 16th July 2008
This article, presented from a personal point of view, is concerned with the design of ligands intended to give specifically either binuclear or tetranuclear metal complexes or coordination polymers. No attempt is made to provide a comprehensive coverage of these topics, the focus being mainly upon results from our laboratory. Some emphasis is placed upon aspects of the historical development of the deliberate construction of coordination polymers (aka MOFs)—materials promising useful applications, the study of which continues to expand exponentially. Some of our recent research is described in which the carbonate ion and the tetracyanoquinodimethane dianion are used as bridging ligands to generate targeted coordination polymers. It is intended that Dalton Perspectives be easily comprehensible to non-specialists in the field; an average second year university chemistry student should be easily able to understand the present contribution.
R. Robson | Richard Robson's roots are in Yorkshire. After a BA and DPhil at Oxford (1955–1962) followed by post-doctoral work at Caltech and Stanford he joined the School of Chemistry at the University of Melbourne in 1966 where he was employed until his retirement in 2004. Since then he has held the honorary position of Professorial Fellow, which has freed him from teaching and administrative duties and allowed him to return to some benchwork. |
Scheme 1 |
Scheme 2 |
We were able to modify I so as to generate binucleating ligands built around a bridging thiophenoxide unit4c (as in IIb and IVb), thereby extending the range of accessible binuclear complexes to include those containing pairs of soft metal centres. In a long sequence of studies with thiolate-based binucleating systems, IIb and IVb, we established: (a) that a wide range of single entities (Z in IIb) could be incorporated as bridging species at the bimetallic site, in several cases in quite unprecedented ways;4 (b) that the bimetallic site could also accommodate two independent species (as in IVb);4h (c) that two initially independent species bound at the bimetallic site could be induced to bond together;4i and (d) that in some cases it was possible to displace the so-formed new bridging species with fresh reactants to establish a cyclic catalytic process.5 Complexes of these binucleating and polynucleating ligands and their descendents continue to be the focus of much current research.7
Two major aims of the binucleating work were to design ligand/metal associations where (a) unusual species may be trapped between the two metal centres and (b) unusual processes may be promoted under the simultaneous influence of the two metal centres, in ways that would be quite impossible in the absence of the special organization deliberately built into the binucleating ligand. Examples of the trapping of unusual species4e are given in Scheme 3 and an example of a catalytic process occurring at the binuclear site is given in Scheme 4.5
Scheme 3 The lines linking Pd centres represent the binucleating ligand. |
Scheme 4 The lines linking Pd centres represent the binucleating ligand. |
Crystallographic studies indicated that the macrocyclic system in V bound four metal centres almost exactly as it was designed to do. However, the central site, occupied by Z in V, was a region of intriguing mystery—what sort of chemistry might occur there under the influence of four closely-spaced metal centres? As indicated elsewhere in this article, design/invention at points such as this merge into exploration/discovery. We discovered for example, a case where a μ4-hydroxo ligand equally bonded to four Ni(II) centres was found at the central site; this was a considerable surprise to us because the hydroxo ligand has available only three pairs of electrons, not four.6 This example illustrates a general point with regard to the theme of design and its limitations, namely that often design works as intended up to a point, beyond which one moves into a zone of exploration. Indeed, this is one of the pleasing and exciting aspects of working with such systems—one deliberately sets up an organised system then asks: “What is going to happen at such-and-such a site when we do so-and-so?”
Fig. 1 Part of the diamond-like {CuI[C(C6H4·CN)4]}+ network. |
Also reported in the same paper8b was one of our earliest attempts at deliberate crystal structure manipulation using cyanide ion, one of the simplest conceivable 2-connecting ligands. Almost half a century earlier Russian workers, using powder diffraction data, had determined the crystal structures of Zn(CN)2 and Cd(CN)2 (amazingly, whilst the Second World War was raging in their part of the world!).9 The Zn and Cd centres acted as 4-connecting tetrahedral nodes and the cyano ligands as linear 2-connecting units to generate diamond-related nets, but the spaces within a single network were so large that they were occupied by a second identical network that interpenetrated the first and vice versa.
I wondered if it might be possible to engineer cyano-bridged systems so as to generate a single diamond net rather than the doubly-interpenetrating arrangement seen in Zn(CN)2. The strategy was based on the fact that the C and N donor ends of the cyanide ligand have different metal ion affinities. The intention was to attempt to selectively replace exactly half the Zn2+ by Cu+, hoping the latter would preferentially bind four cyano carbon donors in a tetrahedral fashion and generate an anionic [CuZn(CN)4]− network with a diamond-like structure. The required counterions would then occupy the spaces where the second framework is found in Zn(CN)2 itself, thereby preventing interpenetration. Models indicated that the tetramethylammonium cation might fit snugly into the adamantane-like cavities formed by the diamond net, with a methyl group directed towards each of the four chair-form cyclohexane-like windows of the cavity. Only half the adamantane-like cavities would need to be so-occupied for charge balance. I found that bringing together the four components Zn2+, Cu+, CN− and NMe4+ in aqueous solution under the simplest imaginable conditions led to the spontaneous assembly of crystalline material with precisely the structure intended. I fully expected only half the adamantane cavities to be occupied by NMe4+ cations (as they were) and expected the other cavities (large enough to enclose an NMe4+ cation!) perhaps to contain a water molecule or two. In fact we were surprised to find, so far as we could see, that these cavities were empty—another example of an unexpected aspect of an otherwise successfully designed system.
The generation of solvated CuI[C(C6H4·CN)4]BF4 and NMe4[CuZn(CN)4], with infinitely extended structures identical in practically all respects to those intended, truly did warrant the description crystal engineering. These results were reported together in a paper in which was proposed a more general approach to the construction of a new and potentially very extensive class of infinite network structures that might provide tailor-made materials of the future.8b A number of general observations or postulates were made in that paper that seemed likely to be of relevance in the (anticipated) future development of the coordination polymer field: (1) Given the limitless range of connecting ligands that could be devised and realistically synthesised (with all sorts of geometries, connectivities and functionalities) and given also the rich geometrical and electronic diversity of metallic elements, it was clear that coordination polymers (if they could be generally obtained in crystalline form as easily as I had obtained CuI[C(C6H4·CN)4]BF4 and NMe4[CuZn(CN)4]) could afford an enormous range of microporous solids resembling zeolites but with a much greater diversity of architectures and properties—properties such as ion exchange, sieving, sorption and heterogeneous catalysis; (2) The polymers envisaged might provide—quote: “materials combining good or even high thermal, chemical and mechanical stability with unusually low density”; (3) Functionalisation of the framework either before or after its construction to provide active sites for heterogeneous catalysis might be straightforward; (4) It might be crucial in the growth of ordered, truly crystalline arrangements that the bond forming step whereby that growth occurs be reversible—thus, if a “wrong step” were taken the system would be able, provided the bond forming step were appropriately reversible, to “backtrack” to eliminate the error and then the orderly build-up could continue (and this was one reason why coordinate bond formation, which is very often rapidly reversible, was so attractive); (5) Bridging ligands bearing chelating sites for metal binding might afford significantly more robust framework structures; (6) Various templates, both neutral and ionic, might be able to “re-direct” the polymerisation to afford a different network. The subsequent vigorous development of the coordination polymer field seems to have borne out all these common-sense postulates, except possibly (3), which remains a possibility very well worth pursuing.
In the late Eighties, having deliberately taken the diamond structure as a target and successfully generated CuI[C(C6H4·CN)4]BF4 and NMe4[CuZn(CN)4] with structures as intended, I sought other structural prototypes as targets: the rutile and the PtS prototypes were immediately attractive because it was easy to conceive geometrically and functionally appropriate building blocks that were also readily available. As mentioned in Section 3 above, the components required for a rutile structure are trigonal 3-connecting units, together with octahedral 6-connecting units, for which roles the tricyanomethanide anion, C(CN)3−, (tcm−), (see VII in Fig. 7), and metal centres capable of adopting an octahedral coordination geometry, respectively, appeared to be the simplest candidates. In 1991 it was reported that bringing Zn2+ and tcm− together under the simplest conditions at room temperature produced single crystals of Zn(tcm)2 which did contain the intended rutile-like networks (see Fig. 2).10 However, the intra-framework spaces were now so large that a second identical but independent network interpenetrated the first and vice versa, a feature that could not possibly have been predicted with any certainty. Eventually we reported a large family of isostructural MII(tcm)2 solids (M = Cr, Mn, Co, Ni, Zn, Cd and Hg) containing two interpenetrating rutile-like networks.10b In collaboration with Professor K. S. Murray we later extended this work to rutile nets in which N(CN)2− rather than C(CN)3− played the role of the 3-connecting node; interestingly, the Co(II) and Ni(II) derivatives of N(CN)2− behave as ferromagnets, providing encouragement for the general notion that coordination polymers may provide solids with new and useful physical properties.11
Fig. 2 A single rutile-like network in Zn(tcm)2. A second identical network, not shown, interpenetrates the first. |
The PtS prototype was a target structure I found particularly attractive in the early stages of our exploration of this area. The essence of the PtS structure is the presence of equal numbers of square planar (or nearly so) 4-connecting centres and tetrahedral (or nearly so) 4-connecting centres, each attached to four of the other type (see Fig. 3). My interest in the PtS structure, from the outset, was stimulated by the exciting possibility of using appropriately functionalised metalloporphyrins to provide the square planar building blocks, but it did seem wise initially, in such an unexplored wilderness, to lower my sights a little and aim for the simplest possible tetrahedral and square planar components, for which roles I chose respectively, Cu(I) (which had served well in both of the above diamond-related nets) and the stable and easily accessible Pt(CN)42− anion (see VIII in Fig. 7). I was able to obtain the compound NMe4{CuI[Pt(CN)4]} in crystalline form very simply from its component ions and my colleagues Dr Bernard Hoskins and Dr Robert Gable showed that the {CuI[Pt(CN)4]}− framework did have the intended PtS-related structure (see Fig. 4); this was reported in 1990.12
Fig. 3 The parent PtS structure (Pt purple, S yellow). |
Fig. 4 The PtS-like {CuI[Pt(CN)4]}− framework. |
This result provided considerable encouragement that the exciting porphyrin possibilities were not unrealistic. For the role of square planar porphyrin component I chose 5,10,15,20-tetra(4-pyridyl)-21H,23H-porphine palladium(II), Pd(tpp), shown as XVI in Fig. 8 (Ma = Pd). Tried and trusted Cu(I) was used as the potential tetrahedral node. In a talk I gave at the 201st National Meeting of the American Chemical Society in Atlanta in April, 1991, I described a compound of composition {CuI[Pd(tpp)]}BF4 that I had made in the form of very small crystals for which we had only unit cell dimensions, not a full structural determination. I reported that the tetragonal cell dimensions agreed very closely with those calculated for the intended PtS-related network. In the subsequently published book of the symposium proceedings,13 in the chapter describing our contribution we omitted specifically to mention this compound, because we still, to our embarrassment, had available only less than conclusive unit cell data and we limited our comments to the following: “The PtS prototype provides a very attractive model for the construction of permeable solids with catalytic potential. If one imagines replacing the planar PtS4 units apparent in Fig. 10a and b (which were two views of the PtS structure itself) with flat, rigid, 4-connecting building blocks larger than the [Pt(CN)4]2− used above, such as porphyrins and phthalocyanins, the potential offered by this net for structures with large channels giving access to banks of catalytic sites can be appreciated.”
Ultimately, we were able to obtain and fully structurally characterise PtS-related networks using 5,10,15,20-tetra(4-pyridyl)-21H,23H-porphine copper(II), Cu(tpp), (see XVI (Ma = CuII) in Fig. 8) and also the somewhat more extended building block 5,10,15,20-tetrakis(4-cyanophenyl)-21H,23H-porphine copper(II), CuII(tcp), (see X (Ma = CuII) in Fig. 7) as the square planar components together with either Cu(I) or Ag(I) as the tetrahedral components—the structures of the PtS-like [CuI(CuIItpp)]+ and [CuI(CuIItcp)]+ frameworks are shown in Fig. 5 and 6 respectively.14
Fig. 5 The PtS-like [CuI(CuIItpp)]+ framework. |
Fig. 6 A single PtS-like [CuI(CuIItcp)]+ framework. A second network, not shown, interpenetrates the first. |
In my talk at the ACS Atlanta meeting in 1991, I proposed a much more general approach to the generation of new types of solids than that proposed in ref. 8. Now, with the deliberate generation of diamond-like, rutile-like and PtS-like networks successfully demonstrated, it was proposed that the construction of networks based on a much wider range of structural prototypes could realistically be contemplated, in which each atom of the prototype net would be replaced by a stereochemically-appropriate molecular building block and each bond of the parent would be replaced by an appropriate molecular connection. In the subsequently published book covering the symposium proceedings we wrote:13 “It is probably wise in early attempts at this framework construction to use the simplest available structural prototypes such as diamond and lonsdaleite (tetrahedral centres), α-polonium (octahedral centres), NbO (square planar centres), PtS (equal numbers of tetrahedral and square planar centres) and rutile (trigonal and octahedral centres in 2 : 1 proportions) but others that exist or that can be envisaged could be put to similar use. Given the wide range of molecular building blocks and connectors that can be conceived for these purposes each prototype net in principle affords a whole family of related frameworks.” These early reports provided encouragement to prospective workers in the field that the approach based on a range of structural prototypes was realistic and that there was a good chance that any framework building attempts they might undertake in the future would generate characterisable crystalline materials rather than intractable amorphous precipitates that might previously have been anticipated.
Our approach to the crystal engineering of coordination polymers couched in net-based terms rather than structure-based terms was first explicitly presented in print in 1994;14 the simple idea was that if we were able to construct building blocks with functionalities and stereochemistries appropriate to a chosen target net, then merely allowing these pre-organised components to react together may, under the correct conditions, lead to the spontaneous assembly of the intended network.
In the early Nineties we initiated a wide-ranging exploratory study of the coordination polymers afforded by a number of carefully chosen 2-, 3-, and 4-connecting bridging ligands that were easily accessible and had reliable internal geometry and reliable metal-binding preferences. One such series of ligands bearing terminal cyano nitrogen donors is shown in Fig. 7, and a similar series with pyridine-like or imidazole-derived N donors is shown in Fig. 8. It seemed crucially important in early work in so uncertain a field to use bridging ligands that were as predictable as possible in their metal binding geometry, and the above two series were carefully chosen with this in mind. Thus, one can be moderately confident, with ligands bearing cyano nitrogen donors as in Fig. 7, that the metal will be bound to nitrogen with the M–N bond more or less co-linear with the CN multiple bond and likewise with ligands bearing pyridine donors, that the metal will be attached more or less in the plane of the pyridine ring and with more or less equal C–N–M angles (give or take a few degrees of flexibility) as in Fig. 8.
Fig. 7 A series of 2-, 3- and 4-connecting cyano-based bridging ligands with various geometries. |
Fig. 8 A series of 2-, 3- and 4-connecting N-heterocycle-based bridging ligands with various geometries. |
When we initiated these exploratory studies, bridging ligands bearing terminal carboxylato donors, in contrast to those with N-donor ligands, appeared much less attractive because of their unpredictability. There was a wealth of evidence from prior coordination chemistry that an RCO2− ligand is capable of binding a metal centre at a wide range of locations relative to the R–C bond, or (complicating the situation still further) is capable of binding more than one metal centre. I was aware that three particular carboxylate arrangements from classical coordination chemistry, seen only under special circumstances, had the potential to provide symmetrical building blocks beautifully pre-arranged for the sorts of coordination polymers we had in mind. These three arrangements were the binuclear copper(II) acetate structure, providing four C–C bonds radiating towards the corners of a square, the tetranuclear basic zinc acetate structure, providing six C–C bonds radiating to the corners of an octahedron and the trinuclear basic chromium acetate structure, providing six C–C bonds radiating to the corners of a trigonal prism. Indeed we did attempt in the early Nineties to construct an α-Po network by the reaction of basic zinc acetate with terephthallic acid, but abandoned this approach after failing to obtain products in the form of single crystals suitable for diffraction studies. More recent admirable work by Yaghi et al. with carboxylate bridging ligands revealed a wide range of exceptionally stable coordination polymers with beautiful structures, showing gas sorption properties that promise useful real-world applications.16 All in all, however, at the outset of our exploratory studies intended to expose the broad sweep of possibilities offered by coordination polymers, bridging ligands binding metal centres at terminal nitrogen donors appeared much more predictable than those based on carboxylate donors.
The series of simple ligands shown in Fig. 7 and 8 yielded many coordination polymers with diverse and unprecedented structures, some of which were anticipated, many of which were not, as described below. In all cases the ligands behaved almost exactly as was intended; the unpredictability, it turned out, arose in the behaviour of the metal centre.
In 1990 we reported the synthesis and structure of [Zn(H2O)2(4,4′-bipyridine)2]SiF6.17 This was the first of literally hundreds of 2D and 3D coordination networks based on 4,4′-bipyridine and related ligands to be reported. Some aspects of the structure were as we expected, but no one on earth, either then or now, I maintain, could have predicted the unprecedented mode of interpenetration adopted. With regard to design considerations, I was aware that it is difficult to assemble six pyridine-derived donors around an octahedral metal centre, on account of the clashes between hydrogen atoms adjacent to nitrogen on neighbouring pyridine nuclei. Although an α-Po (simple cubic) network therefore appeared unlikely, I surmised it was not entirely outside the bounds of possibility in such novel circumstances. I was also aware that an octahedral metal is more likely to take on only four pyridine donors, together with two other trans ligands, and therefore the structure of the sheet polymer seen in Fig. 9 came as no surprise. What was astounding was the way the sheets were packed together. All sheets were equivalent and each one had an infinite number of perpendicular others interpenetrating it. Whether interpenetration will occur or not and the form it takes if it does occur, remain, in general, areas of uncertainty and unpredictability, with regard to the construction of coordination polymers. However, there can be situations where, by design or otherwise, interpenetration is impossible.
Fig. 9 Part of a [Zn(H2O)2(4,4′-bipyridine)2]2+ sheet. |
The general observation that, whilst the ligand behaves as intended the metal centre behaves much less predictably, is illustrated by the variety of structures seen in the coordination polymers of 2,4,6-tri(4-pyridyl)-1,3,5-triazine, tpt, (see XV in Fig. 8), five examples of which follow: (1) With ZnSiF6 the metal centre acts as a 2-connector and generates (10,3)-a nets, eight of which interpenetrate18 (see Fig. 10). For the interested reader not acquainted with the (10,3)-a net, it is considered in more detail in ref. 15 and 18; (2) With Ni(NO3)2 the metal acts as a T-shaped 3-connector and generates the extraordinary (12,3) net19 (see Fig. 11). Further consideration of the (12,3) net and the reasons why it is exceptional can be found in ref. 15 and 19; (3) With CuI(ClO4) the metal acts as a tetrahedral 4-connector and generates cubic boracite-related nets, two of which interpenetrate20 (see Fig. 12); (4) With Hg(ClO4)2 the metal acts as an octahedral 6-connector generating the highly symmetrical cubic net seen in Fig. 13;21 (5) With Zn2+ together with cyanide ion the metal binds two tpt ligands and two cyano ligands to generate an unprecedented doubly interpenetrating cubic network with extraordinarily large sealed off chambers22 (see Fig. 14). In all five of the coordination polymers of tpt just considered, the ligand behaves exactly as intended, binding three metal centres at the corners of an equilateral triangle, or close to it, but the behaviour of the metal is wildly variable and unpredictable.
Fig. 10 (a) The (10,3)-a net. (b) A single Zn/tpt (10,3)-a network in [Zn(tpt)2/3(SiF6)(H2O)2MeOH]. Seven other networks, not shown, interpenetrate. |
Fig. 11 Schematic representation of the (12,3) net in solvated Ni(tpt)(NO3)2. The red T-shaped nodes are provided by Ni. The blue trigonal nodes are located at the centres of the triazine rings. |
Fig. 12 A single boracite-like network in CuI(tpt)4/3(ClO4). A second network, not shown, interpenetrates the first. |
Fig. 13 Schematic representation of the PdF2-like [Hg(tpt)2]2+ network in solvated [Hg(tpt)2](ClO4)2. The blue trigonal nodes represent the centres of the tpt triazine rings and the red 6-connected nodes are located at Hg. |
Fig. 14 Schematic representation of a large sealed-off chamber in solvated Zn(tpt)2/3(CN)(NO3). Selected fragments of two separate, identical and interpenetrating 3D networks are shown here in red and blue. Trigonal nodes (green) represent centres of tpt triazine rings. The yellow centres in the red network and the blue centres in the blue network represent Zn. The edges of the square motifs represent Zn–C–N–Zn connections. |
Another unexpected and unusual structure arose out of our work with the NCAuCN− ligand, intended to bind two metal centres in a more or less linear MNCAuCNM fashion as in VI in Fig. 7. With Zn(II) this gave a coordination polymer in which the ligand behaved as intended but the metal centre took on a tetrahedral coordination environment of four NCAuCN− ligands, which in itself, in the case of Zn(II), was not too surprising, but certainly could not have been confidently predicted. The resulting 4-connected network, however, surprisingly adopted a chiral quartz-like topology and six independent quartz networks interpenetrated (see Fig. 15).23
Fig. 15 A single quartz-like network in Zn[Au(CN)2]2. Five other networks, not shown, interpenetrate the first and each other. |
We explored the coordination polymers formed by the ligand 1,4-bis(imidazole-1-ylmethyl)benzene, bix, (see XIV in Fig. 8), with the intention (in part) of obtaining α-Po-related nets by design. As mentioned earlier, pyridine-derived 2-connecting ligands such as 4,4′-bipyridine seemed inappropriate for the deliberate construction of α-Po-related nets because very rarely is it possible to surround an octahedral metal centre by six pyridine donors, on account of the repulsions between hydrogen atoms adjacent to nitrogen on neighbouring pyridine nuclei. By contrast, octahedral metal centres readily bind six 5-membered N-heterocycles such as imidazoles. We found that bix, with Cd(ClO4)2, gave crystals of solvated Cd(bix)3(ClO4)2 in which the metal centre did take on six imidazole donors to give the intended α-Po-related network—three of them in fact, that interpenetrated24 (see Fig. 16).
Fig. 16 A single α-Po-like network in solvated Cd(bix)3(ClO4)2. Two other networks, not shown, interpenetrate the first and each other. |
With AgNO3, however, bix gives Ag2(bix)3(NO3)2 with a totally unexpected and unprecedented structure (see Fig. 17).25 [Ag2(bix)3]2+ chains are present, consisting of rod and ring components as seen in Fig. 17a. Each individual chain is interlocked with an infinite number of others by multiple rotaxane associations, in which every one of its rod-like segments passes through a ring of another chain and every one of its rings encircles a rod of another chain as seen in Fig. 17b and c. Rotaxanes are associations of a ring and a dumbbell-like component, inescapably locked together in a mechanical, non-bonded fashion as seen in Fig. 17d, whereas catenanes are associations of rings, as in Fig. 17e, linked together in the fashion of an everyday chain. Interpenetrating networks almost invariably generate multiple catenane associations. The structure of Ag2(bix)3(NO3)2 is exceptional in that it is simply devoid of catenane associations and its interpenetration is purely polyrotaxane in character.
Fig. 17 (a) An [Ag2(bix)3]2+ chain in Ag2(bix)3(NO3)2. (b) Four separate chains making four rotaxane associations. (c) Schematic representation of the polyrotaxane associations. (d) A rotaxane. (e) A catenane. |
In the early Nineties we embarked on an exploration of the coordination polymers afforded by dihydroxybenoquinone, dhbqH2, XVII (X = H), and chloranilic acid, canH2, XVII, (X = Cl), that led over a period of years to the construction and structural characterisation of a range of transition metal and lanthanide networks.26 If one were to use ligands of this type deliberately to construct a diamond-like network, one would require 8-coordinate metal centres; we found that in the coordination polymer formed by can2− and Y3+, the metal does take on four bidentate can2− ligands to become an approximately tetrahedral 4-connecting node and the topology is indeed that of diamond.26b
If planar 2-connecting bis-bidentate ligands such as dhbq2− and can2− were to act as bridges between octahedral tris chelated metal centres, the simplest net would be a 2D hexagonal sheet and this would require alternating Δ and Λ metal centres. Alternatively, if all metal centres were of one hand, a 3D (10,3)-a net would result. We found that dhbq2− affords an isostructural series of compounds of composition M2(dhbq)3·24H2O (M = Y, La, Ce, Gd, Yb or Lu).26a,b Each metal centre is 9-coordinate with three bidentate dhbq2− ligands and three cis aqua ligands. The metal centres therefore act as 3-connecting nodes and generate the 2D hexagonal grid sheet structure seen in Fig. 18. In continuation of this work we are presently investigating a newly discovered series of dhbq2−-derived coordination polymers with Mn2+, Fe2+, Co2+ and Zn2+ that have the (10,3)-a topology.
We found that the ligand hexaazatriphenylene, XVIII, abbreviated hat, upon reaction with AgClO4 or AgBF4 in nitromethane gave crystals of composition Ag(hat)X·2CH3NO2 (X = ClO4 or BF4) in which every metal centre was chelated by three hat ligands and every hat ligand was chelated to three Ag centres.27 A 3-connected 3D network with the (10,3)-a topology was thereby generated (see Fig. 19) yet another result that was hoped for but could not possibly have been guaranteed. This was the first structurally characterised example of a coordination polymer derived from a bridging ligand providing three bidentate metal binding sites. Molecules of nitromethane were arranged in a helical fashion in the chiral micropores making very close O⋯N contacts with each other. A tetragonal single crystal of Ag(hat)ClO4·2CH3NO2, upon exposure to the atmosphere underwent exchange of lattice nitromethane for water, thereby becoming cubic; of some significance in our opinion was the fact that not only was the (10,3)-a connectivity retained during this process but also single crystal character was retained—testament to the effectiveness of the chelating tactic for reinforcing networks.
Fig. 19 Part of the (10,3)-a [Ag(hat)]+ network in Ag(hat)(ClO4)·2CH3NO2. |
This form of hydrated Ag(hat)(ClO4), obtained by replacing the nitromethane in Ag(hat)ClO4·2CH3NO2 by water, differs in an interesting manner from the form obtained by crystallising Ag(hat)(ClO4) from aqueous solution. Again, in this latter case, every metal centre binds three hat units and every hat binds three metal centres to give a 3D 3-connected net, but now the topology is that of the (8,3)-b net.15 These examples serve to illustrate the general point made earlier that outcomes of attempts to construct coordination polymers appear to be determined often by a very fine balance of complex and subtle effects, such as relatively weak interactions between the framework and solvate molecules.
Hat is by no means the ideal triply-chelating ligand for network construction because it is not entirely reliable as a 3-connecting building block, sometimes, with divalent metal cations, managing to bind only two of them. We surmise that the reason for this is that, as divalent metal cations are added to successive chelating sites of this electrically neutral ligand, thereby increasing the positive charge on the system, the affinity for cations at the remaining sites decreases. We consider therefore that triply chelating ligands with an overall negative charge may be much better able than hat to secure three metal cations and may provide more stable 3-connected coordination polymers. We have a number of such ligands under current investigation.
We found that if the guanidinium cation is added to the deep blue aqueous solution formed when CuII is dissolved in excess CO32−/HCO3−, a dark blue crystalline solid of composition [C(NH2)3]2Cu(CO3)2 separates, in which the 3D [Cu(CO3)2]2− coordination network has the diamond topology.28
We noticed, accompanying the dark blue [C(NH2)3]2Cu(CO3)2 crystals, a very minor proportion of very small, but paler blue crystals. These are cubic and have the composition Cu6(CO3)12[C(NH2)3]8·K4·8H2O. The [Cu(CO3)2]2− network in this minor product, like that in the major [C(NH2)3]2Cu(CO3)2 product, is 4-connected, but now the topology is analogous to that seen in sodalite.28 The parent sodalite net is shown in Fig. 20. A sodalite cage within the [Cu(CO3)2]2− network is shown in Fig. 21. In each of the hexagonal faces of the sodalite cages are found not one guanidinium cation, but pairs of them in surprisingly close contact (C⋯C, 3.19 Å)—surprising because substantial repulsion would be expected between two formally 1+ charged cations in close contact. These pairs of guanidinium cations are highly complementary to the Cu6(CO3)6 rings that surround them, making no less than twelve equivalent hydrogen bonds to carbonate oxygen atoms as shown in Fig. 21b. The composition per cage is {Cu6(CO3)12[C(NH2)3]8}4− and four K+ cations together with eight water molecules are located inside each sodalite cage.
Fig. 20 (a) A truncated octahedral sodalite cage. (b) The 3D sodalite network. Each cage shares its hexagonal faces with eight neighbouring cages and shares its square faces with six neighbouring cages as can be seen here by focusing on the central cage highlighted in blue. |
Fig. 21 (a) The Cu2+ and CO32− components of a cage within the sodalite-like network in Cu6(CO3)12[C(NH2)3]8·K4·8H2O. (b) A pair of guanidinium cations making twelve H-bonds to carbonate oxygen centres of a Cu6(CO3)6 hexagonal face of a sodalite cage. (c) A Cu4(CO3)4 square face of a sodalite cage. |
One of the square faces of the sodalite cage is shown in Fig. 21c. Four non-coordinated oxygen atoms of carbonate can be seen located at the corners of a square, the distance from any oxygen to the centre of the square being 2.04 Å. The K+ cation in Cu6(CO3)12[C(NH2)3]8·K4·8H2O is too large to be accommodated inside this “O4 hole” and the four K+ cations per sodalite cage are disordered over six equivalent sites in the interior of the cage.
We reasoned that a smaller cation, such as Li+ or Na+, may be able to fit in-plane within the O4 hole and that this may be energetically very favourable. If every O4 hole in the sodalite network were to be occupied by M+ (M = Li or Na) in this way the Cu2+/[C(NH2)3]+/CO32−/M+ contents per cage would be [Cu6[C(NH2)3]8(CO3)12M3]−, and would carry an overall minus one charge per cage. Models suggested that the NMe4+ cation might fit nicely at the centre of each of these hypothetical sodalite cages so as to achieve electrical neutrality. We, therefore, wondered if a combination of Li+ plus NMe4+ (or Na+ plus NMe4+) might induce a mixture containing Cu2+, CO32− and [C(NH2)3]+ to form a network with the sodalite topology in preference to [C(NH2)3]2Cu(CO3)2 with its diamond topology. We found that crystals of composition Cu6(CO3)12[C(NH2)3]8M3·NMe4·xH2O (M = Li or Na) are readily obtained when aqueous solutions containing [C(NH2)3]+, NMe4+ and either Li+ or Na+ are added to a solution of CuII in excess K2CO3 + KHCO3. The [Cu(CO3)2]2− network does have the intended sodalite topology, the guanidinium cations do appear in pairs at the centres of the Cu6(CO3)6 hexagonal rings, the Li+ (or Na+) do occupy the O4 holes and the tetramethylammonium cation is located at the centre of the sodalite cage, all of which features are almost exactly as intended. It appears that the targeted structure is strongly preferred, because even though K+ is present in large excess this cation is not incorporated. What we have here is the specific assembly of five separate species, Cu2+, CO32−, Li+ (or Na+), [C(NH2)3]+ and NMe4+ in aqueous solution at room temperature into a complex but highly symmetrical 3D structure that was fully intended. It is not an overstatement to describe this as crystal construction by design, but it has to be admitted that the outcome could not have been guaranteed. Aspects of the structures that were not and could not have been predicted were the numbers of water molecules included per cage and their disposition within the cage; one disordered water molecule per cage was observed in the Na derivative and five disordered water molecules per cage in the Li derivative.
We have found that similar sodalite-related structures are formed when numerous other divalent cations (Mg2+, Mn2+, Fe2+, Co2+, Ni2+, Zn2+, Cd2+) are used in place of Cu2+, together with [C(NH2)3]+, NMe4+ and Na+, all three of which were intended to perform and do perform essentially the same roles as described above.
Very different outcomes are observed when two side-by-side aqueous reaction mixtures are set up, both containing the same concentrations of Cu2+, [C(NH2)3]+, K2CO3 and KHCO3, differing only in that Gd(NO3)3 is present in one and absent in the other. Deep blue crystals of [C(NH2)3]2Cu(CO3)2 with the diamond topology, as described above and as expected, separate from the Gd-free mixture, whereas pale blue crystals of likely composition {Cu6(CO3)12[C(NH2)3]8}·Gd2·K2·(OH)4·H2O containing a sodalite-like Cu(CO3)22− network structure separate from the other. The Gd3+ cation is clearly somehow or other re-directing the assembly of the other components, but the exact nature of the contents of the sodalite cages, which are highly disordered, remains a mystery at this stage. Some of the other smaller trivalent lanthanide cations (Nd3+, Sm3+, Eu3+, Gd3+, Tb3+, Dy3+, Ho3+, Er3+, Yb3+ and also Y3+) are able to play the same re-directing role as Gd3+ but the larger La3+, Ce3+ and Pr3+ fail to give sodalite networks. Possibly there simply is not enough room inside the sodalite cages to accommodate these larger cations together with the K+ and the OH− (or O2−) presumed also to be located there.
So far, we have isolated and characterised thirty or so different sodalite-related structures but we are confident many more examples could be obtained, because the number of possible combinations of (a) metal centre in the framework, (b) cations contained in the O4 holes and (c) cations and other species inside the sodalite cages is so large.
We have recently isolated crystalline compounds in which divalent cations, rather than the monovalent Li+ and Na+ in the examples above, are incorporated into the O4 holes—the divalent metal cations in the O4 holes may be the same as those forming part of the sodalite network itself or they may be different. We regard this as a major advance, opening up numerous possibilities for the observation of new types of cooperative magnetic and electronic behaviour as indicated below. In some of these recent examples with divalent cations in the O4 holes there are pairs of guanidinium cations in the hexagonal faces of the sodalite cages, as in all the examples considered above, but in others only one guanidinium cation is associated with each hexagonal ring of the network and yet essentially the same MII(CO3)22− sodalite topology is present—a new sub-class promising many new possibilities. When we have divalent Ma cations in the O4 holes and divalent Mb cations in the framework, it is possible to trace out continuous 3D networks of Ma–O–Mb connections, i.e. continuous 3D networks in which Ma and Mb are separated by only a single oxygen atom of carbonate. Interesting cooperative magnetic and electronic behaviour may therefore arise from this close electronic Ma/Mb communication, either when Ma and Mb are both paramagnetic or when they are both capable of accepting or donating an electron—add to this variety the capability of incorporating paramagnetic lanthanide ions into the sodalite cages and the wide range of possibilities on offer can be appreciated.
This already extensive series of high symmetry solids with a common sodalite framework type, but with widely variable combinations of components, promises to be one of the most flexible systems known with regard to the sheer numbers of different compounds that are potentially accessible. The role of the symmetrical hydrogen bonding guanidinium cation, highly complementary to the M6(CO3)6 rings of the anionic network that surround it, is no doubt crucial in providing this structural control.
Although the properties of TCNQ/metal associations have been very extensively investigated, relatively few X-ray structural studies of coordination polymers of this type have been reported, possibly because of the difficulty of obtaining these compounds in a pure and suitably crystalline form. The TCNQ/metal field nevertheless shows great promise for the generation of solids with interesting and possibly useful electronic/magnetic properties. For example, thin films of Cu(TCNQ) and Ag(TCNQ), when subjected to an electric field, switch at a certain threshold potential from a high resistance state to a low resistance state,32 and an interesting family of M(TCNQ)n magnets (M = Mn, Fe, Co, Ni; Tc 8–60 K) has been reported;33 unfortunately, however, these magnetic materials were ill-defined, with compositions varying according to the starting material used and the structures could not be determined. There is a pressing need for reliable and simple synthetic approaches that provide pure, crystalline TCNQ products amenable to single crystal X-ray studies.
In my own work at the bench I have recently returned to the TCNQ area, with the thought in mind that the reduced species (NC)2CH·C6H4·CH(CN)2, structure XX,
I found that the very simple reaction between TCNQH2, Cd(NO3)2, (Ph3PMe)Br and Li(CH3CO2) in MeOH–DMF at room temperature in air yields an air-stable, diamagnetic, crystalline product of composition (Ph3PMe)2[Cd2(TCNQ)3].34 In contrast to derivatives of TCNQ−˙ which are very dark in colour or almost black, this compound is pale yellow–green, consistent with the presence of TCNQ in the expected dianionic oxidation state. The crystal structure was determined by Dr Tim Hudson. Every TCNQ2− unit binds four Cd2+ centres at the corners of a quadrilateral, that is almost a square, and every Cd2+ is in an octahedral environment of six TCNQ2− nitrogen donors, whereby an anionic 3D {[Cd2(TCNQ)3]2−}∞ network is generated. The disposition of the cadmium centres is only very slightly distorted from primitive cubic, as can be seen in the representation of the anionic 3D {[Cd2(TCNQ)3]2−}∞ network shown in Fig. 22. If the pseudo cubic arrangement seen in Fig. 22 is imagined divided into eight “octants”, a triphenylmethylphosphonium cation occupies every octant with its phosphorus atom close to the centre.
Fig. 22 The Cd2(TCNQ)32− network in (Ph3PMe)2[Cd2(TCNQ)3] in which TCNQ is formally present as the dianion. |
Dr Hudson has subsequently made and determined the structures of several other TCNQ2−-based coordination networks using the TCNQH2 approach. One of these contains a [CuI(TCNQ)]− network with the PtS-related topology (Fig. 23), in which the TCNQ2− acts as a planar 4-connecting node and the CuI centre acts as a tetrahedral 4-connecting node, as it had in several other networks referred to above. This was a pleasing addition to the collection of previously generated PtS-related networks, viz.[CuI[Pt(CN)4]}− (Fig. 4),12 [CuI(CuIItpp)]+ (Fig. 5),14 and [CuI(CuIItcp)]+ (Fig. 6),14 as well as that identified in AgTCNQ.31
Fig. 23 The PtS-like [CuI(TCNQ)]− network in which TCNQ is present formally as the dianion. |
N,N′-Dicyanoquinodiimine, XXI, abbreviated DCNQI, and its many substituted variants, give coordination polymers with interesting electronic/magnetic properties (e.g. metal-like electrical conductivity and even superconductivity) arising from the capacity of these ligands to exist either as the spin-paired neutral molecule, DCNQI0, or as the radical anion, DCNQI−˙.35 The vast majority of the metal/DCNQI compounds studied have the formal composition (MI)(DCNQI−˙)(DCNQI0) in which MI is a d0 or d10 cation such as an alkali metal cation, CuI, AgI or TlI. The reduced form of DCNQI, viz., DCNQIH2, XXII, like TCNQH2, is readily available and air-stable. We considered that this might provide a new class of DCNQI2−-based metal derivatives in the same way TCNQH2 afforded metal derivatives of TCNQ2−. We have very recently isolated and structurally characterised the first coordination polymer of DCNQI2−.
The enormously fruitful research area known as coordination chemistry stretches back well over a century to Werner and beyond and the framework building exercises above in fact constitute nothing more than applied coordination chemistry.31 It is clear that any element of deliberate design that may be possible in the coordination framework construction area, has to be largely based on the huge bank of information about the way ligands and metal centres associate that traditional coordination chemistry provides. It seemed perfectly natural, therefore, that we should continue to call these networks coordination polymers.
Large numbers of research groups are now involved in constructing and studying coordination polymers, foremost amongst which is the group of Professor Omar Yaghi who has generated many novel coordination polymers with structures of outstanding beauty, which he chooses to call metal–organic frameworks or MOFs. Some of these are unusually stable and are very effective as materials for the storage of gases, including, most excitingly, dihydrogen.
Faced by the strong tidal surge of international acceptance of the MOF terminology, I feel like King Canute36 in expressing my opinion that this label is ill-chosen and unnecessary. All inorganic chemists know immediately what is meant by the term coordination complex and generally no distinction between “organic” ligands and “inorganic” ligands is made or required—why should coordination complexes that happen to be polymeric be treated differently? Classical coordination complexes such as CoIII(NH3)63+ and CoIII[NH2(CH2)2NH2]33+ have been regarded for decades as very closely-related species despite the fact one contains organic ligands and the other inorganic ligands. There is a large class of non-polymeric coordination complexes and coordination oligomers whose ligands are organic—in this new re-naming system are they to be described as metal–organic ions? or, if they happen to carry no charge, metal–organic molecules? Muddying the waters is the long-established term “organometallic”, which refers to complexes containing metal–carbon bonds. The MOF terminology is unnecessarily restrictive: why should we set aside as a special category called MOFs, those coordination polymers that happen to make use of organic bridging ligands (or some even more narrowly defined sub-group thereof) and relegate networks formed from perfectly respectable “inorganic” bridging species to some limbo? As far as I have been able to ascertain, one of the earliest uses of the term metal–organic framework by the Yaghi group was to describe hydrated CuI(4,4′-bipyridine)1.5(NO3)—clearly a coordination polymer.37 Of course researchers are free to label the systems they work with as they see fit, but to assert that MOFs are not coordination polymers and constitute a new and special class ignores the unquestionable fact that they are simply polymeric systems held together by coordinate bonds.
In a News Feature in Nature magazine recently there appeared a photograph carrying the caption “Chemical creations: Omar Yaghi was the first to design a metal–organic framework”.38 One has to admit that this statement, given the origin of the term metal–organic framework, is not (in a strictly legalistic sense) incorrect. It appears that large numbers of workers in the coordination polymer field are under the same misapprehension apparent in this News Feature, because certain Yaghi papers are highly cited as the initiators of a new era of “reticular synthesis” of MOFs.
In a review of 2003 entitled “Reticular synthesis and the design of new materials”39 the core design principles are set forth in a section entitled “Logic of synthesis by design” containing the following passage: “On the conceptual level, the underlying process for logical synthesis of MOFs starts with the designer choosing a specific target network, which is then deconstructed into its component geometric units. Molecular building blocks with the same geometry as those of the units are then assembled into a MOF structure that has the target network topology.” The authors gave no indication in that article (nor elsewhere to the best of my knowledge) that they were aware of the following passage appearing in the same journal nine years earlier:14 “One approach to crystal engineering that we have been developing is first to choose as a geometrical/topological model one of a number of simple 3D nets such as diamond (all centres tetrahedral), α-Po (all centres octahedral), PtS (equal numbers of square planar and tetrahedral centres), rutile (octahedral and trigonal centres in 1 : 2 proportions) and so on, and then try to devise ways of chemically linking together molecular building blocks with a functionality and a stereochemistry appropriate to the chosen net.”
With regard to design as applied to constructing coordination polymers, the best my group has achieved is to obtain—sometimes!—structures that were intended but could not have been guaranteed. I suspect this is generally true of work in this area. The ligands, in our experience, as indicated above, generally bind metal centres exactly as intended and, therefore, possibly could be described legitimately as designed ligands, but often the metal centres behave unpredictably. The tactic of putting the chelate effect to use in framework construction has hardly started to be exploited; many opportunities are available here for the creative design of connecting ligands capable of binding metal centres strongly and predictably at chelating sites, thereby affording improved structural control in the assembly process and also mechanically and chemically more robust networks.
The extensive series of carbonate-based coordination networks with the sodalite topology, described in Section 4.5, was deliberately designed on the basis of an initial serendipitous observation, but, again, the outcome in any particular case could not have been guaranteed. This series illustrates the value of well-designed templates—the symmetrical hydrogen bonding guanidinium template was crucial to the reliable assembly of Cu2+ and CO32− to form sodalite-related Cu(CO3)22− coordination networks.
The construction of a chemical framework structure where the outcome is beyond all doubt, as in the conversion of the engineer's plans into a working internal combustion engine, perfect and as intended in all respects, remains an objective for the future. Any such absolutely guaranteed framework construction would necessarily be mathematically based and the level of theory required would be way beyond anything presently available.
Our approach to building infinite frameworks based on simple structural/topological prototypes such as diamond, rutile, PtS, α-Po etc., and on the use of sensibly designed bridging ligands, has proved to be illuminating and useful and will no doubt continue to be so. Unintended structures, which in our experience have been many and frequent, have often been of exceptional interest. The value of this sort of work (ours and that of many other groups), designed or not, appears to me to be beyond doubt—it has been extraordinarily fruitful, revealing large numbers of unprecedented and interesting structures as well as potentially useful materials. Whether or not we and all the other researchers producing coordination polymers are deluding ourselves regarding design, as suggested by Jansen and Schon, the results speak for themselves.
In the meantime working in coordination chemistry, with its limitless variety, continues to be great fun and, for myself, I am happy to be labelled a mere explorer. Opportunities for exciting advances, of both intellectual and practical value, are still afforded by the great sweep of chemistry that is at our disposal for creative use in planned framework construction, embracing not only coordination chemistry, but also main group chemistry and even organic chemistry.41
This journal is © The Royal Society of Chemistry 2008 |