Syntheses and Reductions of C -Dimesitylboryl-1,2-dicarba-closo - dodecaboranes

: Two C -dimesitylboryl-1,2-dicarba-closo -dodecaboranes, 1-(BMes 2 )-2-R-1,2-C 2 B 10 H 10 ( 1 , R = H, 2 , R = Ph), were synthesised by lithiation of 1,2-dicarba-closo -dodecaborane and 1-phenyl-1,2-dicarba-closo -dodecaborane, respectively, with n -butyllithium and subsequent reaction with fluorodimesitylborane. These novel compounds were structurally characterised by X-ray crystallography. Compounds 1 and 2 are hydrolysed on prolonged exposure to air to give mesitylene and boronic acids 1-(B(OH) 2 )-2-R-1,2-C 2 B 10 H 10 ( 3 , R = H, 4 , R = Ph respectively). Addition of fluoride anions to 1 and 2 resulted in boryl-carborane bond cleavage to give dimesitylborinic acid HOBMes 2 . UV absorption bands at 318-333 nm were observed for 1 and 2 corresponding to local π-π*-transitions within the dimesitylboryl


Introduction
Tri-coordinate boron compounds have been intensely investigated in the past two decades in view of potential applications as functional materials. 1 The most widely employed functional moiety containing a tri-coordinate boron atom is the dimesitylboryl group (BMes 2 ; Mes = 2,4,6-Me 3 C 6 H 2 ) in which the unsaturated boron centre is kinetically stabilised by steric shielding of the mesityl groups. The empty p z -orbital at the boron atom can interact with the π-system of attached organic skeletons which leads to a narrowing of the HOMO-LUMO-gap (HLG) by lowering the LUMO energy. Indeed, the π-acceptor strength of the BMes 2 group is similar to those of cyano-2 and nitro-3 groups. These electronic characteristics have led to organic materials containing BMes 2 units finding application as electron-transporting materials in opto-electronic devices. 4,5 Compounds containing BMes 2 can be strongly fluorescent and thus have been used in organic light emitting diodes (OLEDs). 5,6 Moreover, the ability of the boron atom to form selectively covalent adducts with small anions has led to applications of these compounds as colorimetric and luminescent sensors for fluoride 7-9 and cyanide. 10,11 Other boron-containing compounds that have gained considerable interest in the last seven years in the field of opto-electronic materials are derivatives of the dicarba-closododecaborane isomers (1,2-, 1,7-and 1,12-C 2 B 10 H 12 which are ortho-, meta-and paracarborane, respectively). 12,13 Due to their delocalised σ-electron systems ('3D aromaticity'), these clusters possess high thermal and chemical stabilities. 14 The ortho-carborane is a unique electron-acceptor when connected to a donor at one or both cluster carbon atoms (at C1 and/or C2 in 1,2-C 2 B 10 H 12 ) due to the elasticity of the cluster C1-C2 bond. [15][16][17] The ortho-carborane unit thus can play an active role as the acceptor in donor-acceptor molecules (dyads).
Photoexcitation of such dyads induce a charge transfer from an organic scaffold to the carborane cluster which either led to luminescence quenching 18 or to charge transfer (CT) emissions or both depending on the solvents used [19][20][21] and whether the materials were investigated as solids. [22][23][24][25][26][27][28] Compounds with both BMes 2 and ortho-carboranyl groups are known 29,30 with para-phenylene bridges linking both units. In these systems, the orthocarboranyl group served as a strongly inductive electron-withdrawing group as the cluster increased the Lewis acidity of the triarylborane as found by fluoride ion titrations compared to the triarylborane without the cluster attached. 29 Another compound containing both BMes 2 and ortho-carboranyl group was reported with an ethenylene bridge linking both units. 31 Of the few ortho-carboranes with tri-coordinate boron substituents at their carbon atoms reported, [24][25][26][32][33][34][35][36] only C-benzodiazaborolyl-ortho-carboranes have been investigated regarding their photophysical, electrochemical and spectroelectrochemical properties. 25 However, the benzodiazaborolyl group generally acts as a π-donor and is therefore electronically quite distinct from the BMes 2 moiety. In order to better understand the photophysical and electrochemical properties of ortho-carboranes with boryl groups, studies with compounds containing a tri-coordinate boron π-acceptor would be appealing and allow one to determine more precisely the electronic interplay between the carborane and tricoordinate boron electron-withdrawing units. Therefore we present herein, the syntheses and crystal structures of two ortho-carborane derivatives 1 and 2 with a BMes 2 group at one of the cage carbon atoms and their photophysical and electrochemical properties (Figure 1). The geometry of the reduced species of 2 was also determined by a combination of 11

Syntheses and characterisation of 1 and 2
Compounds 1 and 2 were synthesised by reaction of fluorodimesitylborane with the corresponding C-lithiocarborane, generated in situ by metallation of ortho-carborane and 1phenyl-ortho-carborane, in boiling toluene ( Figure 1). The elevated temperatures proved to be essential as no conversion was observed at ambient temperature. Purification was achieved by aqueous work-up and the target compounds were isolated in moderate yields by crystallisation from n-hexane/dichloromethane mixtures. The elemental analytical result for 1 was significantly lower (0.8% for carbon) than the calculated value. Similar discrepancies have been noted elsewhere for related compounds. 33 It is possible that the formation of boron carbide during the combustion analysis may adversely affect the values obtained. Signals in the 11 B{ 1 H} NMR spectra of 1 and 2 between 3.7 and -12.9 ppm confirm the presence of the ortho-carborane clusters. The 11 B peaks at 78.9 ppm (1) and 80.4 ppm (2) are assigned to the BMes 2 groups and are very broad compared to the peaks corresponding to the cluster. The 11 B chemical shifts of the Mes 2 B groups in 1 and 2 are virtually identical to trimesitylborane (79.2 ppm) and phenyldimesitylborane (79.3 ppm). 37 Thus, the orthocarboranyl groups influence the chemical shift of the three-coordinate boron atom of the BMes 2 unit in the same manner as a phenyl or mesityl substituent.

Hydrolyses of 1 and 2
Since aqueous work-ups were used in the preparation of 1 and 2 these C-boryl-orthocarboranes are considered to be water-stable -unlike many reported C-boryl-ortho-carboranes that could not be isolated pure due to facile hydrolysis. 32 The 1 H and 11 B NMR spectra for 1 and 2 in deuterated chloroform solutions containing excess water also showed no changes.
The sterics of the mesityl and the carboranyl groups appear to prevent facile hydrolysis of the boron atom in 1 and 2.
Solids of 1 and 2 do, however, hydrolyse on prolonged exposure to air (complete conversion after three weeks for 1 and eighteen months for 2) to give mesitylene and the new carboranylboronic acids, 3 and 4 ( Figure 2). While these acids could not be obtained pure, they were identified by multinuclear NMR spectroscopies and mass spectrometries. The observed cleavage of the B-C(mesityl) bond in the process is not without precedent, it has been shown elsewhere that mesitylene is formed from the reaction of dimesitylborinic acid, Mes 2 BOH, with trimethylaluminium. 38 The initial steps in these air-induced hydrolyses of 1 and 2 probably involve cleavage of the B-C(mesityl) bonds by oxygen (as in the B-C(phenyl) bond cleavage reaction of triphenylboron by oxygen 39 ) followed by hydrolysis with traces of water present in air. As many organic dimesitylboranes have been explored as fluoride sensors, 7,29 the reactivity of 1 and 2 towards fluoride ions was of interest. Chloroform solutions of 1 and 2 were treated with an excess of tetra-n-butylammonium fluoride hydrate (TBAFH) while acetonitrile solutions of 1 and 2 were added with potassium fluoride (KF) and 18-crown-6 to obtain the desired fluoride adducts [1·F]and [2·F] -, respectively, where the fluoride ion is bound to the boryl atom. Hydrolysis took place instead in all cases to give Mes 2 BOH, and the corresponding unsubstituted carborane, 1,2-C 2 B 10 H 12 or 1-Ph-1,2-C 2 B 10 H 11 , as detected by 1 H, 11 B, 13 C and 19 F NMR spectroscopies on the reaction mixtures ( Figure 2). These reactions were complicated by fluoride-ion deboronation processes on the unsubstituted carboranes to give 11 B and 19 F NMR peaks corresponding to fluoroborates of the boron atom initially removed from the cluster. 40 It is possible that fluorodimesitylborane, Mes 2 BF, is initially formed in the reaction as Mes 2 BF is known to be easily hydrolysed to Mes 2 BOH. 39 The reactions of potassium hydroxide (KOH) and 18-crown-6 with 1 and 2 in acetonitrile gave Mes 2 BOH and the corresponding unsubstituted carborane. Deboronation products were also present in the latter reactions as the combination of KOH and 18-crown-6 is a strong deboronating agent. 42
The BMes 2 groups are linked to the cage carbon atoms by B-C single bonds (C1-B1) with lengths of 1.635(3) Å in 1 and 1.629(2) Å in 2, which is at the upper edge of the range determined for other C-boryl-ortho-carboranes (1.607(4) -1.649(12) Å). [24][25][26]33,36 The B-C bond lengths between the mesityl rings and the boryl-boron atoms (B1-C3/C12 1.581(3) Å -1.602(2) Å) are typical for dimesitylboranes. As a consequence of the three-dimensional shape of the cluster in both structures, the interplanar angle enclosed by the mesityl ring pointing towards the second cage carbon atom and the plane defined by the boryl-boron atoms and the three neighbouring carbon atoms (77.9(1)° (1), 78.6(1)° (2)) is larger than in most reported structures of BMes 2 compounds. 44 A virtually perpendicular orientation of the phenyl substituent in 2 with respect to the C1-C2 axis (torsion angles = C1-C2-C21-C22 94.3(2)°, C1-C2-C21-C26 -91.5(2)°) corresponds to the situation in other disubstituted phenyl-orthocarboranes and is preferred due to sterics. 43,45 Photophysics Photophysical data for 1 and 2 are listed in Table 2. The absorption maxima of both Cdimesitylboryl-ortho-carboranes ( Figure S1) in solvents of different polarity do not display any significant solvatochromism. The lack of solvatochromism points to very similar dipole moments in the electronic ground state and the initial excited state indicating that local transitions within the dimesitylboryl unit give rise to the absorption bands observed. The presence of the phenyl ring at C2 in 2 appears to lower the HOMO-LUMO energy gap as the absorption maxima of 2 (330 nm -333 nm) are bathochromically shifted by approximately 12 -14 nm compared to 1 (318 -319 nm) in all solvents used. The energy difference in the absorption maxima between 1 (329 nm) and 2 (332 nm) in the solid state is smaller.
Emission maxima of both compounds in cyclohexane are virtually identical at 541 nm for 1 and 544 nm for 2 with large Stokes shifts of 12300 cm -1 for 1 and 11920 cm -1 for 2. The luminescence spectra ( Figure 5) reveal positive solvatochromism with emission maxima in the more polar solvent dichloromethane in the red emission region at 664 nm for 1 and 643 nm for 2. By using the Lippert-Mataga method 46,47 with an Onsager radius of 4.00 Å estimated from the molecular structures, the calculated transition dipole moments are 10.4 D (1) and 9.2 D (2) similar to transition dipole moments of C-benzodiazaborolyl-orthocarboranes (6.9 -10.9 D). [24][25][26] The results show that the carborane cluster plays a part in the emission process acting as the electron acceptor of the charge transfer process after excitation.
The solid-state emissions occur at longer wavelengths than in cyclohexane ( Figure 6). This bathochromic shift is more pronounced for 1 (567 nm) than for 2 (550 nm) and is presumably caused by fluorophore-fluorophore interactions. The quantum yields (Φ) are very low in all solvents (< 1%) with higher values of 2% (1) and 7% (2) found in the solid state.
In addition to these low-energy emissions, compound 1 displays a weaker emission band at the violet edge of the visible spectrum in polar solvents. The high-energy emissions with smaller Stokes shifts of 4710 -6600 cm -1 probably originate from local transitions at the BMes 2 group. 48 [a] in L mol -1 cm -1 .

Electrochemistry
The electrochemical properties of both C-dimesitylboryl-ortho-carboranes 1 and 2 were investigated by cyclic voltammetry (CV, Figure 7). The peak potentials measured in acetonitrile and dichloromethane solutions, with platinum and glassy carbon working electrodes are listed in Table S1. The traces resemble reported CV data on reductions of carboranes elsewhere 20,21,[25][26][27][49][50][51][52][53][54] and, by inference, reductions take place at the carborane clusters in 1 and 2. CV traces for reductions of the BMes 2 group would involve a simple oneelectron reversible wave. In contrast, the one-electron reduction wave associated with PhBMes 2 occurs at -2.30 V (vs the ferrocenium/ferrocene couple at 0.0 V) 2 and hence the BMes 2 localised reduction in 1 and 2 likely falls outside the electrochemical window examined here. A CV trace for ortho-carborane or a C-monosubstituted-ortho-carborane usually shows a two-electron cathodic wave and an anodic wave that is not of the same current intensity as the cathodic wave. 25,49,50 Often the peak-peak separation between the cathodic and anodic peaks of the wave can be several hundred millivolts due to the structural rearrangement of the dianion on the CV timescale. Decomposition and proton-coupled electron transfer (PCET) processes can also complicate the electrochemical response. 49,55 The CV of 1 in acetonitrile with a glassy carbon working electrode shows a twoelectron cathodic wave at -1.85 V and two anodic waves at -1.40 V and -1.28 V with values referenced to the ferrocenium/ferrocene redox couple at 0 V ( Figure 7). The cathodic wave value of -1.85 V for 1 means that 1 is more easily reduced than C-monophenyl-orthocarborane at -2.25 V 50 reflecting the substantial electron-withdrawing effect of the BMes 2 group. Similar CV traces are observed for 1 with a platinum working electrode and with DCM as solvent ( Figure S2 and Table S1). The non-equivalent current intensities between the forward and reverse waves for 1 suggest that the dianion of [1] 2is not stable and would be difficult to isolate.
A CV trace for ortho-carborane with aryl substituents at both cluster carbon atoms generally shows a reversible wave (or two) on reduction. 20, [25][26][27]49,[51][52][53][54] In several cases, a stepwise reduction involving two separated one-electron reduction steps has been found, with the initial reduction process giving rise to a radical anion with an unusual 2n+3 skeletal electron (SE) count. One example is diphenyl-ortho-carborane 5 where the radical anion has been shown to be stable enough to be observed spectroscopically at ambient temperature in solution ( Figure 8). 51 The first one-electron reduction process on the CV timescale is usually slow due to the rearrangement of the cluster and is often immediately followed by a second one-electron process giving an apparently two-electron cathodic wave. The latter wave is usually evident in DCM for these carboranes. 26 However, on back oxidation two separate anodic waves (with the combined current intensities similar to the current intensity of the cathodic wave) are evident corresponding to two one-electron processes. 25,26 Figure 8. Two one-electron reductions for 1,2-diphenyl-ortho-carborane 5.
The CV of 2 in acetonitrile here shows two very closely spaced and barely resolved one-electron waves (Figure 7). These waves are separated by less than 80 mV based on gaussian analyses of square-wave potential traces ( Figure S3). The two half-wave potentials are estimated to be -1.31 and -1.39 V which indicate that 2 is reduced more readily than diphenyl-ortho-carborane 3 at -1.57 and -1.72 V under the same CV conditions. Thus, compound 2 does form a radical anion with a 2n+3 skeletal electron count. However, the very low comproportionation constant (K c ) associated with the intermediate [2] .means that spectroscopic observation is challenging, and the monoanion could not be isolated in any appreciable concentration in the comproportionated mixture obtained following one-electron reduction ( Figure S4). The CV of 2 in DCM shows the expected CV pattern with a one twoelectron cathodic wave and two one-electron anodic waves ( Figure S2 and Table S1).

Computations
Calculated bond lengths and angles from geometries of 1 and 2 optimised at B3LYP/6-31G* are in good agreement with the experimental values (Table 1). The lengthening of the C1-B6 bonds compared to the C1-B3 bonds is attributed to steric repulsion between an ortho-methyl group of one of the mesityl rings and the B6-H unit. Computed energy barriers for the rotation around the C1-B1 bonds are 6.1 kcal·mol -1 in 1 and 18.6 kcal·mol -1 in 2. Comparison between computed GIAO-NMR and experimental 11 B NMR chemical shifts for 1-4 are in very good agreement (Table S2). The HOMO is a combination of π-orbitals at the mesityl groups (π(Mes)) and the LUMO consists mainly of the empty p-orbital of the boryl boron atom (p(B)) ( Figure 9). Antibonding orbitals with significant cluster contributions have much higher energies (> -0.16 eV) and thus the influence of the clusters on the absorption process is merely inductive in both cases.
According to TD-DFT calculations π(Mes)-p(B) transitions with oscillator strengths (f) of 0.0065 to 0.0698 can be assigned to the absorption bands of both compounds (Tables S3 and   S4). Therefore, the electron density in the initially formed excited state is shifted within the BMes 2 -unit only which is not expected to entail strong changes in the overall dipole moment.
This is in agreement with the lack of solvatochromism in the absorption spectra. Weak transitions between the π-orbitals of the phenyl group of 2 and the p(B) orbital as well as π-π* transitions within the phenyl ring occur at considerably higher energy far in the UV region.
The HOMO-LUMO gap energy of 2 is 0.10 eV smaller than that in 1 which agrees well with the observed bathochromic shift of the absorption maximum of 2 compared to 1.

Geometries of the dianions [2] 2and [5] 2-
While closo-dicarbadodecaboranes all adopt the pseudo-icosahedral geometry, several different geometries of nido-dicarbadodecaborane dianions have been determined crystallographically ( Figure 11). 56    Geometry optimisations on [5] 2reveal that the bowl-shaped geometry C is more stable than A and B by 5.5 and 11.0 kcalmol -1 respectively. More importantly, the computed GIAO 11 B NMR chemical shifts of the bowl-shaped geometry fit well with observed shifts when fluctionality between the two mirror-image geometries of C takes place in solution (Table 4). The geometry D found in [9] 2could not be located for [5] 2where the initial geometry D rearranged to C on optimisation.  Reduction of 2 with sodium metal in 1,2-dimethoxyethane (DME) yielded a dark red solid identified as [Na(DME) n ] 2 [2] by 1 H, 11 B and 13 C NMR spectroscopy. Purification of this extremely air-sensitive salt by crystallization was not successful. Salts of dicarbadodecaborane dianions are known to be extremely air-and moisture-sensitive. 57 The 11 B{ 1 H} NMR spectrum recorded in CD 3 CN revealed a 2:1:1:2:2:2 pattern for the cluster atoms and a very broad signal at 67.5 ppm corresponding to the boryl boron atom (Figure 14). The latter peak is considerably shifted to higher field by about 13 ppm compared to the neutral starting material.    25.46 ( 13 C)]. The 13 C NMR peaks were assigned with the aid of observed 13 C DEPT spectra and computed 13 C NMR shifts. Electron Ionisation (EI) and Atmospheric pressure Solids Analysis Probe (ASAP) mass spectra were recorded with a VG Autospec sector field (Micromass) and Xevo QTOF (Waters) mass spectrometers respectively.

1-Dimesitylboryl-1,2-dicarba-closo-dodecaborane (1):
A solution of n-butyllithium (1.6 M in n-hexane, 2.37 mL, 3.79 mmol) was added to 1,2dicarba-closo-dodecaborane (0.52 g, 3.61 mmol) in toluene (35 mL) at 0 °C. After 16 h stirring at ambient temperature a solution of fluorodimesitylborane (0.96 g, 3.58 mmol) in toluene (6 mL) was added to the resulting suspension. The mixture was heated at reflux temperature for 5 h and washed with water (2 × 10 mL) and saturated sodium chloride solution (10 mL) subsequently. The combined aqueous layers were extracted with toluene (10 mL) and the combined organic phases were dried over sodium sulfate and freed from volatiles in vacuo. The crude product was recrystallised from a mixture of n-hexane (30 mL) and dichloromethane (2 mL
Subsequently it was washed with water (2 × 15 mL) and saturated sodium chloride solution (15 mL). The combined organic phases were dried over sodium sulfate and the solvent was removed in vacuo. Impurities were sublimed from the residue at 80 °C in vacuo and the remaining solid was recrystallised from a mixture of n-hexane (80 mL) and dichloromethane (5 mL). The product 2 was obtained as colourless crystals. Yield: 0.97 g (51 %

Hydrolyses of 1 and 2
A drop of water (excess) was mixed with a solution of 1 or 2 in deuterated chloroform in an NMR tube and the mixture was analysed by NMR spectroscopy. No change in the spectra was observed after a week.

Reductions of 2 and 5
Method 1: A piece of excess sodium metal was added to a solution of 1-dimesitylboryl-2phenyl-1,2-dicarba-closo-dodecaborane 2 (0.07 g, 0.14 mmol) in 1,2-dimethoxyethane (0.5 mL  Method 2: Finely-cut alkali metal pieces were added to a solution of 2 (0.07 g, 0.14 mmol) in tetrahydrofuran (0.5 mL). A purple colour occurred immediately at the metal surface followed by a clear dark red solution after 2 h. The reaction mixture was then analysed by 11 B NMR spectroscopy and in many experiments the desired dianion was present as the carborane compound (Table S5). 1 H and 13 C NMR spectra were also obtained for Na 2 [2] when deuterated THF was used in place of THF. Na 2 [2]. 1 H{ 11  the peak corresponding to ortho-CH 3 groups is hidden within the d 8 -THF peak and the peaks for the cage carbons were not detected above the noise levels, see Figures S14 and S15 for 1 H{ 11 B} and 13 C NMR spectra. Exposing the dark red solution containing Na 2 [2] slowly to air gave a light yellow solution identified by NMR spectroscopy to contain a mixture of the starting material 2 and 1-phenyl-ortho-carborane in a 9:1 ratio. Method 2 was also used in the reductions of 5 with alkali metals (Li, Na, K) but with deuterated THF in all cases and NMR data of M 2 [5] (M = Li, Na, K) are listed in Table S6.

Photophysical measurements
For all solution state measurements, samples were contained in quartz cuvettes of 10 × 10 mm (Hellma type 111-QS, suprasil, optical precision). Cyclohexane was used as received from commercial sources (p. a. quality), the other solvents were dried by standard methods prior to use. Concentrations varied from 20 to 100 µM in order to get analysable emission spectra due to the low quantum yields. Effects of the concentration on the shape of the observed emission spectra were excluded in this concentration range. Solid samples were prepared by vacuum sublimation on quartz plates (35 × 10 × 1 mm) using standard Schlenk equipment and conditions. Each plate was laid in a 100 mL round bottom flask and a crystal of the sample substance placed below it was sublimed. Absorption was measured with a UV/VIS double-beam spectrometer (Shimadzu UV-2550), using the solvent as a reference.
The output of a continuous Xe-lamp (75 W, LOT Oriel) was wavelength-separated by a first monochromator (Spectra Pro ARC-175, 1800 l/mm grating, Blaze 250 nm) and then used to irradiate a sample. The fluorescence was collected by mirror optics at right angles and imaged on the entrance slit of a second spectrometer while compensating astigmatism at the same time. The signal was detected by a back-thinned CCD camera (RoperScientific, 1024 \ 256 pixels) in the exit plane of the spectrometer. The resulting images were spatially and spectrally resolved. As the next step, one averaged fluorescence spectrum was calculated from the raw images and stored in the computer. This process was repeated for different excitation wavelengths. The result is a two-dimensional fluorescence pattern with the y-axis corresponding to the excitation, and the x-axis to the emission wavelength. The wavelength range is λ ex = 230-430 nm (in 1 nm increments) for the UV light and λ em = 305-894 nm for the detector. The time to acquire a complete EES is typically less than 15 min. Postprocessing of the EES includes subtraction of the dark current background, conversion of pixel to wavelength scales, and multiplication with a reference file to take the varying lamp intensity as well as grating and detection efficiency into account. Stokes shifts were calculated from excitation and emission maxima, which were extracted from spectra that were converted from wavelength to wavenumbers beforehand. The quantum yields in solution were determined against POPOP (p-bis-5-phenyl-oxazolyl(2)-benzene) (Φ F = 0.93) as the standard.
The solid-state fluorescence was measured by addition of an integrating sphere (Labsphere, coated with Spectralon, Ø 12.5 cm) to the existing experimental setup. At the exit slit of the first monochromator the exciting light was transferred into a quartz fibre (LOT Oriel, LLB592). It passed a condensor lens and illuminated a 1 cm 2 area on the sample in the centre of the sphere. The emission and exciting light was imaged by a second quartz fibre on the entrance slit of the detection monochromator. Post-processing of the spectra was done as described above. The measurement and calculation of quantum yields was performed according to the method described by Mello. 71

Electrochemistry
Cyclic voltammetric measurements were carried out using an EcoChemie Autolab PG-

Crystallographic studies
Single crystals were coated with a layer of hydrocarbon oil and attached to a glass fiber. Crystallographic data were collected with a Bruker AXS X8 Prospector Ultra APEX II diffractometer with Cu-Kα radiation (graphite monochromator, λ = 1.54178 Å) at 100 K.
Crystallographic programs used for structure solution and refinement were from SHELX-97. 73 The structures were solved by direct methods and were refined by using full-matrix least squares on F 2 of all unique reflections with anisotropic thermal parameters for all nonhydrogen atoms. The hydrogen atoms bonded to the carborane units were refined isotropically, all other hydrogen atoms were refined using a riding model with U(H) = 1.5 U eq for CH 3 groups and U(H) = 1.2 U eq for all others. Crystallographic data for the compounds are listed in Table S7. CCDC-1048027 (1) and CCDC-1048028 (2) contain the supplementary crystallographic data for this paper. These data can be obtained free of charge from the Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif

Computational details
All computations were carried out with the Gaussian 09 package. 74 The model geometries were fully optimised with the B3LYP functional 75 with no symmetry constraints using the 6-31G* basis set 76 for all atoms. Frequency calculations on all optimised geometries revealed no imaginary frequencies. Computed absorption data were obtained from TD-DFT 77 calculations on S 0 geometries whereas computed emission data were from the S 1 geometries.