1 Introduction

In the last two decades, large metal clusters have become the focus of great attention because of their fascinating structural aesthetics, physical properties, and implication in the active sites of metalloproteins [1–3]. In particular, the discovery that transition-metal clusters can behave as single-molecule magnets has fuelled impressive synthetic efforts aimed at obtaining new spin topologies [1,2]. Many different approaches have thereby been developed, which range from serendipitous assembly [4,5] to rational design [6–13]. Undoubtedly, the former strategy has provided an amazing variety of unpredicted structural types leading at the same time to the observation of new physical phenomena like quantum tunnelling of the magnetization and quantum phase interference [1,14]. Nevertheless, the rational design and realization of complicated structural motifs remains a very attractive, though extremely elusive goal.

The controlled assembly of transition-metal based building blocks can be realized exploiting the coordination bond as a supramolecular tool. Specific architectures can be rapidly assembled in a single step under thermodynamic equilibrium conditions, based on the information content of the reaction system. This includes the preferred coordination geometry of the metal, the favoured coordination mode of the ligand, and the composition of the reaction mixture as given by the ligand-to-metal ratio, the solvent, the presence of counterions or ‘helper’ species (templates), etc. [10,15,16]. The molecular information stored in the building blocks is read at the supramolecular level to yield one or more possible structures. For instance, the metal-ion directed formation of ring structures can be explained on the basis of non-covalent intramolecular interactions [17], template effects [18], or the interplay between metal- and ligand-centred contributions to the turning angle [3,10,19]. Indeed, it is informational complexity that determines the predictability of the final structure(s), thus marking the difference between serendipitous assembly and rational design.

Grids [11,12] and linear arrays [13] are among the most fascinating recent examples of molecular magnetic clusters assembled by fully rational design. Verdaguer, Marvaud and Mallah also synthesized heptanuclear clusters by connecting a first generation of paramagnetic centres to a central hexacyanometallate unit (Fig. 1a) [6–9]. Following a similar strategy, we attempted a controlled expansion of polynuclear rather than mononuclear units. As central cores, we used robust metallasiloxane cages, whose structure comprises two hexaphenylcyclohexasiloxanolate ligands, (PhSiO₂)₆^6– (Fig. 2), coordinated to six divalent metal ions in a sandwich-like fashion, affording an inert and thermodynamically stable torus-like complex [20–23]. The cage often encapsulates a chloride ion exhibiting an unusual μ₆ coordination mode [20–22]. A key feature is the presence of labile solvent molecules that usually occupy six coordination sites available on the external ‘surface’ of the torus. Upon ligand exchange, the cage can thus function as a hexatopic receptor [15] for monodentate units (Fig. 1b), including paramagnetic complex ligands.

In this paper, we report the synthesis, X-ray structure and magnetism of a high-spin hexacopper(II) complex [Cu₆{(PhSiO₂)₆}₂(BuOH)₅]·3 ⁿBuOH (1·3 ⁿBuOH) that can function as a central preformed unit for controlled cluster expansion. A self-assembly reaction with the trigonal-bipyramidal copper(II) mononuclear [Cu(tmpa)CN]PF₆ (2) [24], which contains terminal-cyanide and capping polyamine ligands (tmpa = tris(2-pyridylmethyl)amine), affords a decacopper(II) cluster [Cu₆{(PhSiO₂)₆}₂{NCCu(tmpa)}₄](PF₆)₄ (3), with an unprecedented spin topology (see [25] for a preliminary report).

2 Experimental

3 Synthesis and X-ray structure

The synthesis of cage-like metallaorganosiloxanes containing transition metals is commonly based on the exchange reaction of alkali-metal organosiloxanolates [34] with transition metal chlorides in alcohol-containing media [20–23,35,36].

The resulting cage complexes often display a ‘sandwich-like’ structure supported by two organocyclohexasiloxanolate, (RSiO₂)₆^6–, ligands and encapsulating a μ₆-chloride ion [20–22]. Copper(II)-containing metallasiloxanes are peculiar in that they exhibit two structural types: sandwich-like (with no Cl^– guest) [20,23] and globular-like [37,38], the latter featuring an organocyclododecasiloxanolate, (RSiO₂)₁₂^12–, ligand. The formation of either structural motif can be controlled by carefully tuning the reaction conditions, such as the particular solvent and alkali metal used. For instance, the use of potassium phenylsiloxanolate is known to favour the formation of the sandwich-like cages [37,38]. Although n-butanol as a reaction media generally favours globular-like structure, thus decreasing the yield of sandwich-like one, the resulting cage complex presents a good solubility in organic solvents, which greatly facilitates the subsequent ligand-exchange reaction.

Reaction between CuCl₂ and potassium phenylsiloxanolate, previously synthesized in situ from poly-(phenylsilsequioxane) [22] and potassium hydroxide, in n-butanol affords 1·3 ⁿBuOH as large green–blue crystals. The crystal packing of 1 is shown in Fig. 3. The unit cell contains two [Cu₆{(PhSiO₂)₆}₂(ⁿBuOH)_x] molecules [x = 4 (molecule a) and 6 (molecule b)] packed in an almost orthogonal fashion, the dihedral angle between the mean Cu₆ planes being 87.69(5)°. The overall metal ion topology is analogous to that of the hexacopper(II) siloxanolates [Cu₆{(PhSiO₂)₆}₂(EtOH)₆] (4) [20] and [Cu₆{(PhSiO₂)₆}₂(DMF)₆]·4 DMF [23]. The metal ions are located at the vertices of almost planar hexagons (deviations from the average Cu₆ planes ± 0.0017(13) Å) and are coordinated by four oxygen atoms from two (PhSiO₂)₆^6– ligands (Fig. 4). The most remarkable structural feature of the cage is the presence of 12 six-membered CuO₃Si₂ chelate rings, which exhibit a distorted pseudo-chair conformation. The ⁿBuOH-bound copper(II) ions show a distorted square-pyramidal (SP) coordination, with 0.37- and 0.31-Å mean shifts (for molecule a and b, respectively) from their basal planes toward the apical oxygen donors. By contrast, Cu3 displays a square-planar (S) coordination and a 0.145(6)-Å basal shift. As a consequence the copper–oxygen core of molecule a is compressed along the Cu3^...Cu3′ axis as compared with molecule b and shows a much more pronounced deviation from idealized 6/mmm symmetry. Interestingly, the Cu-O-Cu angles are very similar in both molecular units (mean value 92.8(7) and 93.0(12)° for molecules a and b, respectively) in spite of the presence of copper(II) ions with different coordination geometries.

While the previously reported cluster 4 [20] is only sparingly soluble in organic solvents, complex 1·3 ⁿBuOH can be easily dissolved in chloroform. Reaction with complex ligand 2 gives a decacopper(II) cluster 3 in low yield:

$\left[{\mathrm{Cu}}_6{\left\{{\left({\mathrm{PhSiO}}_2\right)}_6\right\}}_2{\left(\mathrm{B}{}^{𝑛}\mathrm{uOH}\right)}_5\right]\left(\mathbf{1}\right)+4\left[\mathrm{Cu}\left(\mathrm{tmpa}\right)\mathrm{CN}\right]{\mathrm{PF}}_6\left(\mathbf{2}\right)\to \left[{\mathrm{Cu}}_6{\left\{{\left({\mathrm{PhSiO}}_2\right)}_6\right\}}_2{\left\{\mathrm{NCCu}\left(\mathrm{tmpa}\right)\right\}}_4\right]{\left({\mathrm{PF}}_6\right)}_4\left(\mathbf{3}\right)+5^{𝑛}\mathrm{BuOH}$The asymmetric unit of 3·2CHCl₃·2.75MeOH·0.25H₂O comprises one cluster and four hexafluorophosphate ions along with two chloroform and disordered lattice methanol and water molecules. The [Cu₆{(PhSiO₂)₆}₂{NCCu(tmpa)}₄]⁴⁺ cation shown in Figs. 5 and 6 approaches D₂ point-group symmetry. The inner hexagonal Cu₆ array, featuring four SP and two S metal centres, is surrounded by four trigonal-bipyramidal (TB) Cu(tmpa) units linked to the central core through cyanide bridges. The bulky tmpa ligands limit the number Cu(tmpa) units that can be arranged around the Cu₆ core and lead to a cisoid conformation of the Cu–N–C–Cu(tmpa) moieties. Selected structural parameters of the copper–oxygen cores of 1 (molecule a) and 3 are compared in Table 4. Upon substitution of the ⁿBuOH molecules with [Cu(tmpa)CN]⁺ units, the Cu₆ ring becomes slightly less planar (deviation from the mean plane ± 0.0108(6) Å), but it approaches 6/mmm symmetry more closely. The Cu_S^...Cu_S′ distance, in-fact, increases from 5.296(5) in 1 to 5.4164(11) Å in 3, while the Cu_SP^...Cu_SP′ mean distances decrease [5.83(11) vs. 5.731(14) Å in 1 and 3, respectively). Cu_SP-type ions display an increased shift from their oxygen basal plane (av. 0.41 Å) as compared with 1, while the basal shift of four-coordinated Cu_S ions is somewhat reduced (< 0.13 Å). A marked effect can be observed also on the angles at the bridging siloxanolate oxygen atoms where Cu_S–O–Cu_SP and Cu_SP–O–Cu_SP values become very different (av. 91.1(4)° and 97.3(5)°, respectively). The formation of the cyanide bridges has only minor effects on the structure of the [Cu(tmpa)CN]⁺ moieties, affecting mainly the observed departures from trigonal bipyramidal geometry [25].

4 Magnetic properties and HF–EPR spectra

The magnetic properties of 1·3ⁿBuOH are reported in Fig. 7 as a χ T-vs-T plot. The χ T product at 300 K (2.8 emu K mol^–1) is higher than expected for six uncoupled S = 1/2 spins (2.25 emu K mol^–1 with g = 2.00) and increases upon lowering temperature, reaching 6.2 emu K mol^–1 at 3.0 K and decreasing slightly at the lowest temperatures. This behaviour is indicative of moderate ferromagnetic interactions between copper(II) ions, as found in 4 [39]. This conclusion is confirmed by the low temperature χT value, which is expected to approach 6 emu K mol^–1 for an S = 3 ground spin state with g = 2.00. A Heisenberg Hamiltonian with six identical nearest-neighbour coupling constants gave no satisfactory agreement with experimental data (dashed line). We thus attempted to refine the model by using two independent J values, J₁ and J₂ [see Eq. (1) and Fig. 8].

An excellent fit was obtained for T > 10 K with J₁ = –62(1) cm^–1, J₂ = –10.6(2) cm^–1 and g = 2.107(4) (solid line). However, due to (i) the presence of two independent molecules containing a different number of S and SP copper(II) centers, and (ii) the complicated pattern of Cu–O–Cu angles around the ring(s) (Table 2), these best-fit parameters must be taken with great care. The g values are significantly lower than the average Lande’s factor resulting from the analysis of HF-EPR spectra (<g> = 2.172, vide infra), possibly reflecting systematic errors in bulk magnetic measurements. The field dependence of the magnetization at low temperature is shown in the inset in Fig. 7. At the highest fields explored, the magnetization approaches the saturation value 6 N_A μ_B expected for a S = 3 state with g = 2.00. Also, the non-coincidence of the curves recorded at 4.6 and 1.9 K points to the presence of sizeable magnetic anisotropy effects. The experimental magnetization data have been modelled using a second-order spin Hamiltonian with axial symmetry (Eq. (2)):

We now turn to the magnetic properties of 3·2CHCl₃·2.75 MeOH·0.25 H₂O, which are reported in Fig. 10. The χ T product is 4.9 emu K mol^–1 at 300 K, hence larger than expected for ten uncoupled S = 1/2 spins (3.75 emu K mol^–1 with g = 2.00). It increases smoothly upon lowering temperature down to ca. 10 K (6.2 emu K mol^–1) and then decreases rapidly to 4.6 emu K mol^–1 at 2.0 K. A quantitative analysis of χT-vs-T data was based on the Heisenberg Hamiltonian in Eq. (3):

5 Conclusions and outlook

The hexacopper(II) siloxanolate cage 1·3 ⁿBuOH featuring labile ⁿBuOH ligands has been synthesized and characterized by single-crystal X-ray diffraction, magnetic measurements and HF–EPR spectra. The complex has a high-spin S = 3 ground state arising from ferromagnetic exchange interactions among the six S = 1/2 copper(II) centres, as found for the EtOH analogue [39]. However, 1·3 ⁿBuOH is much more soluble in organic solvents and can be used as a robust, preformed unit for the controlled assembly of larger spin clusters via ligand exchange. This approach has led to the isolation of a decacopper(II) complex, [Cu₆{(PhSiO₂)₆}₂{NCCu(tmpa)}₄](PF₆)₄ (3) [25], containing four paramagnetic complex ligands [Cu(tmpa)CN]⁺. Though prevalent ferromagnetic coupling within the Cu₆ core is maintained, the distortions introduced by the new ligands weaken exchange interactions in the ring. In addition, due to unfavourable overlap between magnetic orbitals, magnetic coupling with the peripheral copper(II) ions is weak. Further synthetic studies on similar structures containing transition-metal ions other than copper(II) are in progress.

6 Supplementary material

Crystallographic data (excluding structure factors) for 1·3 ⁿBuOH and 3·2 CHCl₃·2.75 MeOH·0.25 H₂O have been deposited with the Cambridge Crystallographic Data Centre as supplementary publications Nos. CCDC-191357 and CCDC-186933, respectively. These data can be obtained free of charge via www.ccdc.cam.ac.uk/conts/retrieving.html(or from the CCDC, 12 Union Road, Cambridge CB2 1EZ, UK; fax: +44 1223 336033; e–mail: deposit@ccdc.cam.ac.uk).

Acknowledgements

We thank M. Monari and V. Albano (Department of Chemistry ‘G. Ciamician’, University of Bologna, Bologna, Italy) for assistance in X-ray data collection on 3·2 CHCl₃·2.75 MeOH·0.25 H₂O, and H. Weihe (Institute of Chemistry, University of Copenhagen, Copenhagen, Denmark) for providing the programs SIM and SIMSPC used to simulate EPR spectra. This work was financially supported by the University of Modena and Reggio Emilia, by Italian MIUR and CNR, by MOLNANOMAG network and by the ESF Programme ‘Molecular Magnets’.