Stepwise assembly of mixed-metal coordination cages containing both kinetically inert and kinetically labile metal ions: introduction of metal-centred redox and photophysical activity at specific sites.

Stepwise preparation of the heterometallic octanuclear coordination cages [(M a ) 4 (M b ) 4 L 12 ] 16+ is reported, in which M a = Ru or Os and M b = Cd or Co (all in their +2 oxidation state). This requires initial preparation of the kinetically inert mononuclear complexes [(M a )L 3 ] 2+ in which L is a ditopic ligand with two bidentate chelating pyrazolyl-pyridine units: in the complexes [(M a )L 3 ] 2+ one terminus of each ligand is bound to the metal ion, such that the complex has three pendant bidentate sites at which cage assembly can propagate by coordination to additional labile ions M b in a separate step. Thus, combination of four [(M a )L 3 ] 2+ units and four [M b ] 2+ ions results in assembly of the complete cage assemblies [(M a ) 4 (M b ) 4 L 12 ] 16+ in which a metal ion lies at each of the eight vertices, and a bridging ligand spans each of the twelve edges, of a cube. The different types of metal ion necessarily alternate around the periphery with each bridging ligand bound to one metal ion of each type. All four cages have been structurally characterised: in the Ru(II)/Cd(II) cage (reported in a recent communication) the Ru(II) and Cd(II) ions are crystallographically distinct; in the other three cages [Ru(II)/Co(II), Os(II)/Cd(II) and Os(II)/Co(II), reported here] the ions are disordered around the periphery such that every metal site refines as a 50:50 mixture of the two metal atom types. The incorporation of Os(II) units into the cages results in both redox activity [a reversible Os(II)/Os(III) couple for all four metal ions simultaneously, at a modest potential] and luminescence [the Os(II) units have luminescent 3 MLCT excited states which will be good photo-electron donors] being incorporated into the cage superstructure. 2 (0.12g, mmol) in was The to that sample was fully reduced [ i.e. any aerial oxidation was back to see main text], then crystallised by slow diffusion of di-isopropyl ether into the MeNO 2 solution. The crystalline product was collected by filtration and washed with diisopropyl ether, diethyl ether, and cold methanol. The remaining red crystalline precipitate was the pure product. X-Ray quality crystals were grown by slow diffusion of di-isopropyl ether into an MeNO 2 solution. Yield: 0.05 g, 0.005 21%.

Please note that technical editing may introduce minor changes to the text and/or graphics, which may alter content. The journal's standard Terms & Conditions and the Ethical guidelines still apply. In no event shall the Royal Society of Chemistry be held responsible for any errors or omissions in this Accepted Manuscript or any consequences arising from the use of any information it contains.

Dalton Transactions
www.rsc.org/dalton 4 coordination preferences. [8][9][10] Early examples of this come from helicate complexes based on a mixture of octahedral and tetrahedral metal ions which occupy different positions along the helicate axis. 8 A recent example of this principle in a coordination cage is the formation of a cubic cage in which octahedral tris-chelate Fe(II) ions form the corners, and square planar Cu(II) ions with four monodentate ligands occupy the centres of the faces. 9 In this case the octahedral [Fe(NN)6] 2+ units and the planar [CuL4] 2+ units act as orthogonal assembly elements based on two different types of labile metal ion.
Both of these approaches allow the rational design and self-assembly of heterometallic structures with different metal ions at specific sites. However, in all members of our family of cage complexes, 1c all of the metal coordination sites are octahedral tris-chelates with every metal ion being in a tris(pyrazolyl-pyridine) coordination environment. This precludes the two methods outlined above: the equivalence of all metal binding sites means that there is no basis for selecting which ion goes at which position in the cage on the basis of hard/soft considerations or different coordination geometries, so the necessary differentiation between sites does not exist.
Accordingly we have investigated a different approach based on the use of pre-formed, kinetically stable, metal complex fragments with pendant binding sites; combination of these 'complex ligands' with additional labile ions in a second step results in assembly of the pre-formed fragments into a complete cage structure. 11 We note that the use of a combination of 'inert + labile' components to control assembly of heteronuclear complexes with similar coordination sites is known in other contexts, 12 but application of this method to assembly of large cages remains undeveloped.
The approach we have used is shown in Fig. 1 11 In this follow-up paper we extend the strategy to These isomers do not require separating for our purposes, because this 1:3 fac:mer ratio of geometric isomers is precisely what is required for assembly of the complete cages. Other members of our cage family contain varying proportions of fac:mer metal complex units at the vertices, 1c ranging from all fac 15 to all mer, 16 and various ratios in between, as required to facilitate any specific self-assembly. The cubic [M8L12]X16 cages that we use here happen to contain two fac vertices (at opposite ends of the long diagonal of the cube) and six mer vertices. 13 Half of these sites will be occupied by the Os(II) or Ru(II) ions, and the positions of these must strictly alternate with the sites occupied by the labile partner ions, given that the synthetic strategy prevents two ions of the same type from being connected by a single bridging ligand.
Thus the four kinetically inert Ru(II) or Os(II) subcomponents must contain a 1:3 fac:mer isomer ratio, which arises naturally from the synthesis and is exactly what is required for assembly of this cage type -which is one of the reasons why this cage type was chosen for this work. Stepwise assembly of other cages will require either preparation of pure fac or mer metal complex units as building blocks; 17 or will require the isomers to be separated after synthesis of a mixture. This issue is however avoided here, which is helpful because a wide range of chromatographic conditions could not separate the fac and mer isomers of [OsL3](PF6)2.
[OsL3](PF6)2 (mix of isomers) shows a symmetric redox wave, assigned to a chemically reversible Os(II)/Os(III) couple, at +0.46 in MeCN (Fig. 3), which is identical to the Os(II)/Os(III) redox potential of [Os(bipy)3] 2+ . 19 The fac and mer isomers are expected to have very similar redox potentials: however these were not resolved in the voltammetric wave which is symmetric (equal cathodic and anodic peak currents) with ∆Ep = 80 mV, and therefore behaves exactly like a normal one-electron reversible redox process. This redox potential is ca. 0.4 V less positive than the corresponding Ru(II)/Ru(III) couple, 11 which is typical behaviour for isostructural Ru(II) and Os(II) complexes due to the lower ionisation energy of Os(II) compared to Ru(II) in a comparable environment.
The UV/Vis absorption spectrum of [OsL3](PF6)2 ( Fig. 4) shows the usual combination of spin-allowed 1 MLCT absorptions around 400 nm, plus a less intense spin forbidden 3 MLCT absorption manifold which provides a low-energy absorption tail in the 500 -600 nm region, and high-energy ligand-centred transitions in the UV region.
These 1 MLCT and 3 MLCT absorptions are at somewhat higher energy than in [Os(bipy)3] 2+ . Given that the Os-based d(π) orbitals are at similar energy in both cases (as shown by the identical Os(II)/Os(III) redox potentials of [Os(bipy)3] 2+ and [OsL3] 2+ ) it follows that the higher 1 MLCT / 3 MLCT absorption energies in [OsL3](PF6)2 arise from a higher-lying ligand-centred LUMO of the pyrazolyl-pyridine unit compared to a bipy ligand, and this is reflected in the luminescence properties.
[OsL3](PF6)2 shows a broad luminescence spectrum with a maximum at 625 nm in air-equilibrated MeCN (φ = 0.05) which we suggest arises from the 3 MLCT state (Fig.   4, inset). This contrasts with [RuL3](PF6)2 which is non-luminescent in fluid solution at room temperature. This situation arises with Ru(II) tris-diimine complexes when the 3 MLCT and d-d states are sufficiently similar in energy for the d-d state to provide a rapid deactivation pathway. 18 However the greater ligand-field splitting associated with Os(II) compared to Ru(II) in the same coordination environment means that the d-d state is now too high in energy to provide a thermally accessible deactivation pathway, and the lowest-energy 3 MLCT state now shows luminescence. This emission is typical for Os(II) tris-diimines but, consistent with what was observed in the absorption spectrum, is notably higher in energy than that of [Os(bipy)3] 2+ in fluid solution (λem = 743 nm). 19  The 77K emission maximum of [OsL3] 2+ (as its chloride salt to provide solubility in the MeOH/EtOH solvent mixture) shows that the luminescence maximum is sharpened and slightly blue-shifted with the highest energy feature at 620 nm (Fig. 4,inset). This is typical behaviour for 3 MLCT excited states, arising because the lack of solvent repolarisation when the sample is frozen destabilises the excited state. From the highest energy emission feature at 77K we can see that the 3 MLCT energy is 16100 cm -1 , compared to 14100 cm -1 for [Os(bipy)3] 2+ . 19 As the excited-state energy content of photoexcited [OsL3](PF6)2 is 2000 cm -1 higher than that of [Os(bipy)3] 2+ , but the cost of oxidising it to Os(III) is the same, it follows that photo-excited [OsL3](PF6)2 should be a better electron donor than [Os(bipy)3] 2+ by ca. 0.25 eV, which makes it a considerably better excited-state electron donor than the well-known [Ru(bipy)3] 2+ unit. 20

Preparation and structural characterisation of heterometallic cages.
In our recent communication 11  The molecular structure derived from crystallographic data is shown in Fig. 5.
The basic structure of the cage is similar to that of other [M8L12] 16+ cages with the same ligand, having a metal ion at each vertex and a bis-bidentate ligand spanning each of the twelve edges. 13 Extensive inter-ligand π-stacking around the periphery involves alternating arrays of electron-rich (naphthyl) and electron-deficient (coordinated pyrazolyl-pyridine) groups. Metal-metal separations along the cube edges are 11.3 -11.4 Å.
In this case however, unlike with [Ru4Cd4L12](ClO4)16, 11 the metal sites are indistinguishable crystallographically as the cage exhibits disorder over two orientations. If the two alternate sets of four positions in the cube superstructure (cf. In all three cases therefore we can confirm the basic structure crystallographically but disorder of the two types of metal ion prevents unambiguous identification of which metal ion is at which site. However, as mentioned above, the synthetic methods necessarily requires that the two types of metal ion strictly alternate around the periphery, and electrospray mass spectra confirm the formulations of the cage cations with masses and isotope patterns consistent with the expected mixture of four of each type of metal ion (Ru/Co, Os/Cd, Os/Co). Thus, for example, the ES mass The as-synthesised complex containing Os(II) centres is red, but it slowly turns green in solution when exposed to oxygen (Fig. 10); the collapse of the 1 MLCT and 3 MLCT bands in the visible region, and the growth of a weak long-wavelength band which extends into the red region beyond 900 nm [probably LMCT involving Os(III)], are both obvious. We note that the spectra in Fig. 10 could only be recorded from 380 nm at the high-energy end due to the limited solvent window of nitromethane which is the best solvent for this experiment as the oxidised complex precipitates from less polar solvents such as MeCN.
Addition of ascorbic acid reversed the process and regenerates the spectrum of the fully reduced Os(II) form. The clear isosbestic point at ca. 600 nm confirms the chemical reversibility of the process. The stability of the cage in both [Os II 4Cd4L12] 16+ and [Os III 4Cd4L12] 20+ forms is further confirmed by a DOSY spectrum: although the 1 H NMR spectrum of the [Os III 4Cd4L12] 20+ species after oxidation lacked resolution and could not be assigned, the DOSY spectrum shows no change in its diffusion coefficient compared to the starting complex (Fig. 11).
[Os4Cd4L12](ClO4)16 also retains the photophysical properties of the component unit [OsL3](PF6)2. Its luminescence spectrum in solution is essentially identical to that of [OsL3](PF6)2 with a broad maximum at 625 nm in MeCN, with a quantum yield of 2.5% and, again, two lifetime components: 156 ns (minor) and 73 ns (major) which can be ascribed to the mixture of geometric isomers of the Os(II) units. Significantly, the strong naphthalene-based fluorescence characteristic of the free ligand, and which we also saw in the cage complex [Cd8L12](ClO4)16, 13a is completely quenched; which implies the presence of (naphthyl)→Os(II) energy-transfer from the ligand array to the Os(II) ions at the vertices of the cage, presumably assisted by the aromatic π-stacking which brings naphthyl units into close association with Os(II) tris(pyrazolyl-pyridine) termini (see figures of crystal structures). The complex is not sufficiently soluble in solvents that give good low-temperature glasses to get a good 77K luminescence spectrum, but based on the near-identical behaviour of the luminescence from mononuclear [OsL3](PF6)2 and the cage [Os4Cd4L12](ClO4)16 at room temperature, it is reasonable to assume that the 3 MLCT energy of the chromophores in the cage is again 16100 cm -1 and that it will be a good excited-state photo-electron donor to electron-deficient guests that occupy the central cavity.

Conclusions
The stepwise synthetic method of cage assembly for which we reported the first example recently 11 has been extended to complete the preparation of a set of four heterometallic self-assembled cubic cages, in which four kinetically inert ions [Ru(II) or Os(II)] and four kinetically labile ions [Cd(II) or Co(II)] alternate around the periphery of the cage superstructure. There are two particularly important features of the Os(II)containing cages which will be exploited in future work. The first is the reversible redox activity at modest potential, which allows a four-electron redox swing to change the charge on the cage between 16+ and 20+; as guest binding in organic solvents is driven by polar interactions between the guest and the interior surface of the cage, 13 this may provide a mechanism to modulate guest binding for controlled uptake / release. The second is the photophysical activity, with the four Os(II) units -which are good photoelectron donors in their excited states -surrounding the cavity where guests will bind, which opens the door for one-electron or even multi-electron photoinduced interactions between the cage and bound guests.

Preparation of [OsL3](PF6)2.
A mixture of OsCl3•3H2O (0.20 g, 0.57 mmol) and L (1.30 g, 2.85 mmol) in ethylene glycol was heated to reflux under N2 for 12 hours. After cooling to room temperature a saturated aqueous solution of KPF6 was added to precipitate the crude product, which was collected by filtration. The filtrate was washed copiously with water and then desiccated overnight. The crude solid was dissolved in acetonitrile and purified by column chromatography on silica by elution with MeCN/water/saturated aqueous KNO3 (100:4:2, v/v). The main red band was collected and solvent was removed to give a dark red/orange solid. The product was dissolved in water and aqueous KPF6 was added to precipitate pure diffraction data were collected at the EPSRC National Crystallography Service at the University of Southampton, using a Rigaku FR-E+ diffractometer equipped with a Saturn 724+ CCD detector, using high-intensity Mo-Kα radiation from either a rotating anode or a microfocus sealed-tube source. 22 For [Ru4Co4L12](BF4)16•3MeNO2, data were collected on a Bruker Apex-II diffractometer at the University of Sheffield. In each case a crystal was removed from the mother liquor, coated with oil, and transferred rapidly to a stream of cold N2 on the diffractometer to prevent any decomposition due to solvent loss. In all cases, after integration of the raw data, and before merging, an empirical absorption correction was applied (SADABS) 23 based on comparison of multiple symmetry-equivalent measurements. The structures were solved by direct methods and refined by full-matrix least squares on weighted F 2 values for all reflections using the SHELX suite of programs. 24 Pertinent crystallographic data are collected in Table 1.
In all cases crystals exhibited the usual problems of this type of structure, viz.
weak scattering due to a combination of poor crystallinity, solvation, and disorder of anions / solvent molecules. All three structures contained large solvent-accessible voids whose volume was ca. 40% of the total unit cell volume. These showed in the refinement to contain diffuse electron density which could not meaningfully be modelled, ascribed to severely disordered solvent molecules as well as those anions that could not be located. This diffuse electron density was removed from the refinements using the SQUEEZE function in PLATON. 25 As a typical example, in the structure of This helped to keep refinements stable.
As a consequence of this the refinements are of poor quality by normal smallmolecule standards, but are quite typical for large cage structures of this type. We emphasise that in each case the basic structure and connectivity of the complex cation could be unambiguously determined with reasonable precision and we use the structures only for that purpose with no detailed analysis of structural minutiae. Full details are in the individual CIFs. CCDC numbers 1413546 -1413548.    showing the presence of four independent ligand environments consistent with a statistical mixture of fac and mer isomers.