Icosahedral metallacarborane/carborane species derived from 1,1 ’ -bis( o -carborane) †‡

,

In the 1 H spectrum of a freshly-prepared CDCl 3 solution of 2 are CH cage resonances at δ 4.03 and 3.91 but these are too close to each other to speculate which is due to the carborane and which is due to the ruthenacarborane. The 1 H NMR spectrum of 2 also confirms overall molecular asymmetry with two integral-3 doublets (and not one integral-6 doublet) for the CH(CH 3 ) 2 protons of the p-cymene ligand. The 11 B{ 1 H} NMR spectrum of 2 consists of ten resonances between δ 2.8 and −17.2 with relative integrals 1 : 1 : 2 : 1 : 2 : 2 : 4 : 2 : 1 : 3 from high frequency to low frequency.
With time, solutions of 2 show clear evidence for a slow transformation of 2 into an isomer 3, a compound which was originally isolated along with 2 from the initial reaction. A THF solution of 2 heated to reflux for two hours reveals its complete conversion to 3, with 58% of the compound being recovered following work-up involving thin layer chromatography (TLC). In 3 there is a significantly greater separation of the CH cage resonances, which now appear at δ 3.64 and 2.63. Since only the ruthenacarborane part of 2 has changed in its isomerisation into 3 we tentatively assign the lower frequency resonance, δ 2.63, as arising from CH cage in the {RuC 2 B 9 } portion of 3.
For compounds 2, 4 and 5 we attempted to identify which 11 B resonances were due to which part of the molecule ({MC 2 B 9 } or {C 2 B 10 }) from 11 B{ 1 H}-11 B{ 1 H} COSY spectra but, as was the case with [1] − , it proved impossible to do this unambiguously. In Table 1 we list the weighted average 11 B chemical shifts, <δ( 11 B)>, of the conjoined species 1,1′-bis(o-carborane), 2, 4 and 5 along with those of their "components", [1,2-closo-C 2 B 10  Although the spectra of 1,1′-bis(o-carborane) 3,6 and all these "components" have been reported previously we have remeasured some of them here in CDCl 3 for internal consistency. Note that we have not included compound 3 in this These data show that when 1,1′-bis(o-carborane) and the metallacarborane-carborane species 2, 4 and 5 are "constructed" from their constituent parts the <δ( 11 B)> value for the "product" lies to high frequency of the (weighted) average of that of the two "components". For 5 the <δ( 11 B)> value is very close (and slightly to low frequency of ) to that for the metallacarborane component, whilst for 1,1′-bis(o-carborane), 2 and 4 the <δ( 11 B)> value is actually to high frequency of that of both components.
Perspective views of single molecules of the 3,1,2-RuC 2 B 9 -1′,2′-C 2 B 10 species 2, its 2,1,8-RuC 2 B 9 -1′,2′-C 2 B 10 analogue 3, and the equivalent cobalt species 4 and 5 are presented in Fig. 3 Perspective view of the anion in the salt [BTMA] [1] and atom numbering scheme. In the unprimed cage there is partial disorder of B3, part of which appears as the 12 th atom of an icosahedron (not shown for clarity); partial occupancies are B3 0.548(10) and B12 0.452 (10). The H atom bridging on the open face of the unprimed cage was not located. Displacement ellipsoids are drawn at the 40% probability level except for H atoms.     atom having a misfit >0.018 Å. The {C 2 B 10 } fragments fit better with their reference molecule, the overall misfit here being 0.02-0.03 Å, and it is always C1′ or C2′ that has the largest individual misfit, typically 0.05-0.06 Å. It is clear from Fig. 4 and 6 that a consistent feature of the 3,1,2-MC 2 B 9 -1′,2′-C 2 B 10 structures is a pronounced bend-back of the arene or Cp ligand in a direction away from the C 2 B 10 substituent on C1. This structural feature is undoubtedly the result of intramolecular steric crowding, which also likely contributes to the relatively large misfit values of the metal atoms in 2 and 4. The ligand bend-back is conveniently quantified by θ, the dihedral angle between the plane of the ligand C atoms (arene or Cp) and the plane defined by B5B6B11B12B9 (the lower pentagonal belt usually taken as the reference plane in 3,1,2-MC 2 B 9 icosahedra). 29 For 2 θ is 16.08 (9) In the 2,1,8-MC 2 B 9 -1′,2′-C 2 B 10 compounds 3 and 5 significant intramolecular steric crowding is removed since the C 2 B 10 substituent to the MC 2 B 9 cage is now at position 8 and so not adjacent to the metal atom. Consequently the arene or Cp ring plane lies effectively parallel to the lower pentagonal belt, now the C8B4B5B10B12 plane [θ is only 0.27(5)°in 3 and 2.19(7)°i n 5], and the C8-C1′ distances are 1.5294(17) and 1.5329(16)Å, respectively, slightly shorter than or identical to the intercage C-C distance in 1,1′-bis(o-carborane). 1 The gross similarities between the structures of 2 and 4 (similar ligand bend-back angles, similar C-C1′ distances) imply that, to a first approximation, they are equally sterically crowded. However, whilst 2 is relatively easily isomerised to 3 by gentle heating, even prolonged heating to reflux of 4 in toluene does not convert it into 5; rather 4 has to be reduced to the anion [4] − which then isomerises ( presumably to [5] − ) at room temperature, affording 5 on aerial oxidation. This reduction-induced isomerisation of metallacarboranes has precedent in the literature. 30 Thus, as already has been noted, it appears that the basicity of the metal fragment, and not simply the steric crowding it affords, is important in determining the ease of 3,1,2-MC 2 B 9 to 2,1,8-MC 2 B 9 isomerisation in these species. Given that it is generally accepted that cobaltacarboranes are more susceptible to isomerisation than ruthenacarboranes, at least for 13-vertex species, 31 this is an interesting observation and one that we will address more fully in future contributions. 32

Conclusions
Examples of 12-vertex metallacarborane/carborane compounds, MC 2 B 9 -C 2 B 10 , derived from single deboronation and    **A trace amount of an orange spot (R f = 0.60) identified as 4 was also observed and its identity confirmed via 1 H NMR spectroscopy.
Attempted thermal isomerisation of 4. Compound 4 (0.038 g, 0.10 mmol) was dissolved in toluene (20 mL) and the solution heated at reflux for 5 h. The solvent was removed and the crude residue was submitted for 1 H and 11 B NMR spectroscopies, however there was no evidence that 4 had converted to 5. Preparative TLC using an eluent of DCM-petroleum ether, 30 : 70, led to the recovery of 4 (0.020 g, 53%).
Redox isomerisation of 4. To a solution of 4 (0.012 g, 0.030 mmol) in dry degassed THF (10 mL) was added a solution of sodium naphthalenide (1 mL of a 0.031 M solution in THF, 0.031 mmol). The reaction was allowed to stir under nitrogen for 1 h, oxidised using a water aspirator for 30 min, and solvent was removed in vacuo. Only compound 5 was identified by 1 H and 11 B NMR spectroscopies.

Crystallography
Diffraction-quality crystals of salt [BTMA] [1] and compounds 2, 3, 4 and 5 were afforded by slow diffusion of a CH 2 Cl 2 solution of the appropriate species and 40-60 petroleum ether at −30°C. Intensity data for all except 4 were collected on a Bruker X8 APEXII diffractometer using Mo-K α X-radiation, with crystals mounted in inert oil on a cryoloop and cooled to 100 K by an Oxford Cryosystems Cryostream. Compound 4 afforded crystals too small for our in-house system and consequently data were collected at the National Crystallographic Service at the University of Southampton at 100 K on a Rigaku AFC12 diffractometer operating with Mo-K α X-radiation. Indexing, data collection and absorption correction were performed using the APEXII suite of programs. 36 Structures were solved by direct methods (SHELXS 37 or OLEX2 38 ) and refined by fullmatrix least-squares (SHELXL). 37 Cage C atoms not involved in the intercage link were identified by a combination of (i) the examination of refined (as B) isotropic thermal parameters, (ii) the lengths of cage connectivities, (iii) the Vertex-Centroid Distance Method 39 and (iv) the Boron-H Distance Method, 40 with all four methods affording excellent mutual agreement. The anion in [BTMA] [1] is partially disordered. The C 2 B 10 cage is fully ordered but the C 2 B 9 cage has one B atom disordered between two sites, B3 and B12, with SOFs 0.548 (10) and 0.452(10) respectively. Atoms B3 and B12 were refined with an isotropic thermal parameter fixed at 0.03 Å 2 . There is also partial disorder in 3 between atoms C2′ and B3′ (C 2 B 10 cage), successfully modelled with vertex 2 being 0.446(19)C + 0.554(19)B, with complementary SOFs at vertex 3.
In [BTMA] [1] it was not possible to locate the (disordered) bridging H atom associated with the open face of the nido cage and final refinement with constrained BH and C cage H atoms (B-H = C cage -H = 1.12 Å) afforded better agreement than that with these H atoms allowed to refine. In the BTMA cation the H atoms were constrained to C phenyl -H = 0.95 Å, C secondary -H = 0.99 Å, C methyl -H = 0.98 Å. For all other structures BH and C cage H atoms were allowed to refine positionally whilst other H atoms were constrained to idealised geometries; C aromatic -H = 1.00 Å, C Cp -H = 1.00 Å, C tertiary -H = 1.00 Å, C methyl -H = 0.98 Å. All H displacement parameters, U iso , were constrained to be 1.2 × U eq (bound B or C) except Me H atoms [U iso (H) = 1.5 × U eq C(Me)]. Table 4 contains further experimental details.

Calculations
All geometries were optimised without constraints using Gaussian 03, Revision D.01 41 employing the BP86 functional 42 and 6-31G** basis sets for B, C and H atoms. 43 Analytical frequency calculations were used to confirm geometries as minima or transition states. The transition state was further characterised through IRC calculations. 44