Cage-like structures can self-assemble from suitable metal ions and organic linkers, but the size of the assemblies was limited. The surprise discovery of a new series of cages opens up fresh horizons for self-assembly. See Letter p.563
Nature has inspired an efficient way to synthesize large molecular aggregates: self-assembly. In the natural process, multiple copies of subunits (proteins, for example) spontaneously agglomerate into complex, hierarchical architectures such as virus shells (capsids). On page 563, Fujita et al.1 report their use of self-assembly to prepare the largest known, artificially synthesized cage-type object that has a precise atomic composition — an almost spherical shell that assembles from 144 components.
Construction in the macroscopic world is usually associated with building sites, where many builders simultaneously assemble bricks to give shape to an architect's design. But although synthetic chemists can be thought of as molecular architects2, construction of molecules at the nanometre scale works quite differently. Conventional synthetic protocols allow target compounds to be constructed only through a linear sequence of steps, and the intermediates formed after each step must be purified in a time-consuming and yield-reducing way. This limits both the size and complexity of target molecules, so that the most complicated structures synthesized so far consist of no more than several hundred atoms and are only a few nanometres in length (see ref. 3, for example).
Chemists have therefore used self-assembly extensively to make molecular superstructures on different length scales and of diverse shapes and structures. In particular, metallosupramolecular chemistry involves the self-assembly of bifunctional organic ligands such as bipyridyl molecules (which possess two binding sites for metal ions) with ions of metals such as palladium (Pd2+). The structures that emerge can be polyhedra, in which the metal ions act as vertices that are connected by edges formed by the organic ligands.
If individual ligands in these compounds can be exchanged easily for other ligands, then the resulting systems can rapidly adjust and rearrange in solution to reach the energetically most stable structure as the sole or major product. Under such dynamic conditions, the structure and topology of the final assembly are mainly governed by three factors: the preferential formation of closed-shell structures that maximize the number of metal-occupied binding sites4, rather than polymeric products; the second law of thermodynamics, which maximizes entropy by favouring the formation of many small cages at the expense of larger assemblies; and the preferential formation of 'isotropic' structures that have indistinguishable subcomponents to minimize surface energy and distribute local strain equally throughout the assembly. It therefore follows that the most favourable structures are highly symmetrical objects in the shape of Platonic solids (regular convex polyhedra such as cubes or octahedra) or Archimedean solids (semiregular polyhedra composed of different regular polygons that converge at identical vertices).
Additional design constraints also apply, such as the need for palladium ions to bind to ligands in a square-planar geometry — which implies the convergence of exactly four edges at each vertex. Taken together, the constraints on the self-assembly of palladium ions with bifunctional ligands reduce the number of potential target structures to five cages of formula PdnL2n, in which n can be 6, 12, 24, 30 or 60, and L is the ligand5 (Fig. 1a).
Over the past two decades, Makoto Fujita's research group has pioneered self-assembly by synthesizing representatives of the first four members of the series: the Pd6L12 octahedron6; the Pd12L24 cuboctahedron7; the Pd24L48 rhombicuboctahedron8; and the Pd30L60 icosidodecahedron9. The type of structure that forms depends on the ligand design. For example, subtle changes in the angle formed between the two pyridyl units in a bipyridyl ligand can induce Pd24L48 stoichiometry to form, rather than Pd12L24 (ref. 8).
In the present work, Fujita et al. targeted the elusive Pd60L120 rhombicosidodecahedron, the last representative of the series. However, the authors serendipitously discovered the formation of a Pd30L60 cage whose single-crystal X-ray structure clearly differed from the previously reported one9, and the topology of which did not correspond to any of the Platonic and Archimedean solids. The authors therefore postulated the existence of a new series of polyhedra, which had been reported as theoretical possibilities by mathematicians10 but never observed in any natural or artificial assemblies: closed-shell frameworks in which eight equally distributed triangles are incorporated into a system of squares (Fig. 1b). These structures are reminiscent of Goldberg polyhedra, which consist of 12 pentagons connected by hexagons, and which are ubiquitous in natural and biological systems — such as fullerene structures and virus capsids.
Impressively, the authors also predicted that the next homologue of the series, Pd48L96, would be more stable than the isolated compound, and were able to manually separate individual crystals of this larger assembly from the products of their reaction. This cage is by far the most complex molecular structure of precise atomic composition to have been synthesized until now, and is constructed from 144 components through 192 individual metal–ligand interactions.
What is the largest cage structure that could be self-assembled? The extended Goldberg series of polyhedra provides an indefinite number of ever-greater structures, so in principle there is no intrinsic limit to size. However, it will be increasingly difficult to overcome the entropic penalties associated with the self-assembly of large cages, and to avoid the unwanted but faster formation of smaller cages.
Are there likely to be any practical applications for these giant cages? Investigations into the chemistry and properties of such assemblies might be severely hampered by the difficulties in synthesizing them, especially in bulk quantities rather than just as individual crystals. The structural integrity of the cages, both in solution and in the solid state, is also an unknown crucial issue for any applications. Nevertheless, these huge metal–organic assemblies might encapsulate giant biomolecules such as proteins by forming host–guest interactions, thus stabilizing the biomolecules and potentially allowing control of their structures in unnatural conditions.
Apart from their value as benchmarks for artificial self-assembly processes, Fujita and co-workers' structures might also inspire interest from other scientific areas. For instance, mathematicians could seek more-exotic topologies as targets for self-assembly, and biologists might search for previously unsuspected topologies in virus capsids or other large biological assemblies. And only time will tell whether Fujita and colleagues' synthetic masterpiece will be the starting point for further journeys into yet-unexplored chemical territory.