Cyclometalated cinchophen ligands on iridium( III ): towards water-soluble complexes with visible luminescence †

Eight cationic heteroleptic iridium( III ) complexes, [Ir(epqc) 2 (N ^ N)] + , were prepared in high yield from a cyclometalated iridium bridged-chloride dimer bearing two ethyl-2-phenylquinoline-4-carboxylate (epqc) ligands. Two X-ray crystallographic studies were undertaken on selected complexes (where the ancillary ligand N^N = 4,4 ’ -dimethyl-2,2 ’ -bipyridine and 4,7-diphenyl-1,10-phenanthroline) each con ﬁ rming the proposed formulations, showing an octahedral coordination at Ir( III ). In general, the complexes are lumi-nescent (620 – 630 nm) with moderately long lifetimes indicative of phosphorescence. Hydrolysis of the ethyl ester moieties of the epqc ligands gave the analogous cinchophen-based complexes, which were water-soluble and visibly luminescent (568 – 631 nm). The spectroscopic and redox characterisation of the complexes was complemented by DFT and TD-DFT calculations, supporting the assignment of dominant 3 MLCT to the emissive character.


Introduction
There has been considerable attention and effort disbursed on iridium(III) complexes with cyclometalated ligands, due to their tuneable and generally efficient photoluminescence properties and subsequent performance in a variety of photophysical and electronic applications. 1 Such complexes are capable of showing intense phosphorescence at room temperature; 2 the heavy atom iridium centre mediating strong spin-orbit coupling and intersystem crossing (ISC), mixing the singlet and triplet excited states and generating high phosphorescent efficiencies. As a consequence, cyclometalated iridium complexes have found many applications in a variety of opto-electronically related applications, such as electrochemical cells, 3 photovoltaics, 4 and luminescence imaging. 5 Their ability to perform in such roles relies upon an understanding of their excited state properties, which can be modulated by altering the cyclometalating and/or ancillary ligand associated with the iridium centre. The ability to tune luminescence emission wavelengths through variation of cyclometalating ligands and ancillary ligands ( predominantly for cationic complexes) renders such complexes very useful, particularly in electrochemiluminescence (ECL). 6 For example, increasing the π-conjugation of phenylpyridine by adding an aromatic ring (to give phenylquinoline) bathochromically shifts the 3 π-π* and triplet metal-to-ligand charge transfer ( 3 MLCT) emission due to lower lying π* orbitals. 7 Consequently, there have been several reports of phenylquinoline derivatives as cyclometalating ligands with various ancillary ligands. 8 However, many applications of Ir(III) complexes in this context, including biologically-related uses, require water solubility. Surprisingly then, there have only been a handful of reports of water-soluble iridium cyclometalated complexes, most commonly where the ancillary diimine ligands are functionalised with solubilising groups such as sugars, 8j,l triazoles, 9 polyethyleneglycol (PEG), 10 bioconjugates 11 and carboxylate groups, 12 as well as the bis-cyclometalated bisaqua complexes; 13 reports of water-solubilising fuctionalisation at the cyclometalating ligand are extremely rare. 14 The purpose of this paper is to present the synthesis and photophysical properties of a class of iridium complex that incorporate cyclometalated cinchophen-based ligands, providing a convenient route towards water-soluble complexes with exploitable photophysical properties; the structural, spectroscopic, electrochemical and photophysical studies are presented together with supporting DFT and TD-DFT calculations on the complexes.
The conversion of complexes 3a-g to their corresponding free acids was achieved by stirring the esterified complexes in an equi-volume mixture of 1 M KOH and acetone under an inert atmosphere. Subsequent neutralisation with 1 M HCl, removal of solvent and extraction with methanol (allowing removal of KCl) led to the isolation of complexes 4a-g as their chloride salts, [Ir( pqca) 2 (N^N)]Cl. However, 1 H NMR spectroscopy and mass spectrometry indicated that it was not possible to isolate 4h by this method, the reasons for which are currently unknown.
All new complexes were characterised using a range of spectroscopic techniques. Firstly, the 1 H NMR spectra of complexes 3a-h are complicated in the aromatic region with overlapping resonances associated with the cyclometalated and diimine ligands; the retention of the ethyl ester functionality was observed in the aliphatic region. The 2-methoxyethyl groups of 3c appeared as broadened triplets at 4.46 and 3.68 ppm together with a singlet at 3.30 ppm. 31 P-{ 1 H} NMR spectroscopy confirmed the presence of the PF 6 − ion with a signature septet ( 1 J PF )a tca. −145 ppm in all cases. Low and high resolution mass spectra were obtained for the complexes 3a-h and each confirmed the identity of the cationic, monomeric species of type [(epqc) 2 Ir(N^N)] + , revealing the parent cations of [M − PF 6 ] + in each case, with the appropriate isotopic distribution. Complex purity was confirmed by elemental analysis.
Upon hydrolysis of the ester groups, the 1 H NMR spectra of the isolated analogues 4a-g all confirmed the absence of the ethyl, or 2-methoxyethyl groups in the case of 4c, yielding 2,2′bipyridyl-4,4′-dicarboxylic acid (bpdc), as well as improved resolution of the aromatic resonances. The absence of a resonance in the 31 P-{ 1 H} NMR spectra indicated the exchange of PF 6 − with Cl − counter ions in the deprotected species. The Scheme 1 Synthetic route to complexes [Ir(epqc)  Solid-state IR spectra were also obtained on all complexes highlighting the conversion from ester (ca. 1720 cm −1 ) to carboxylate (ca. 1580 cm −1 ) via significant low energy shifts in ν(CvO), and the absence of the PF 6 − counter ion stretch (ca. 835 cm −1 ) confirming exchange with chloride.

X-ray crystallography studies
Single crystals of 3b and 3e suitable for X-ray diffraction studies were isolated following vapour diffusion of Et 2 O into concentrated CHCl 3 or MeCN solutions of the complexes, over a period of 48 h at −20°C. The bond lengths and bond angles are reported in Table 1, and the associated data collection parameters are reported in Table S1, ESI. † The structures obtained for the three complexes (  1e,7c,d,20 For 3e the structure showed that the phenyl substituents of the 4,7-diphenyl-1,10phenanthroline ligand are twisted out of planarity from the phenanthroline unit by 55.1°and 50.2°respectively. In addition, as with related phenylquinoxaline complexes, 7d,21 there is a distortion within the quinoline moiety caused by the steric interactions between the chelating ligands ( Fig. 1b and  2b). Some of the inter-ligand C⋯C and C⋯N non-bonding contact distances (Table 1) are shorter than 3.4 Å (i.e. the sum of the van der Waals radii of the atoms). This results in the phenyl groups of the epqc ligand showing deformation angles of 14.1°and 19.1°for 3b and 17.5°and 28.6°for 3e with respect to the quinoline fragment of the ligand.
The bond lengths and angles of 3b and 3e were compared with the optimised values calculated from density functional theory (DFT) studies (also see DFT section and Table 1). In general, a reasonable agreement was obtained between the theoretical and experimentally observed bond lengths, although some small differences were found. The calculated Ir-N bipyridine bond lengths, Ir-N(3) and Ir-N(4), are 0.038 and 0.062 Å longer than the experimental values for 3b. In the case of the cyclometalated Ir-C bonds, calculated values are very similar to the experimentally obtained data for Ir-C(1) in 3b and 3e; similarly the Ir-N quinoline bonds, where the calculated and experimental values are comparable. Again, the calculated structures reveal C⋯C and C⋯N interactions between the ligands. The calculated deformation angles of the quinoline fragment are much lower for 3b and 3e.
Density functional theory (DFT) studies DFT calculations (computed using the B3PW91 hybrid functional) were performed to investigate the frontier orbitals and provide qualitative descriptions of the highest occupied Fig. 1 (a) Ortep representation of [Ir(epqc) 2 (dmbpy)] + 3b (50% probability ellipsoids, solvent molecules, PF 6 − anion and hydrogen atoms have been omitted for clarity) and (b) showing non-bonding contact interactions between chelating ligands.  molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) energy levels.
For the [Ir(epqc) 2 (N^N)] + complexes, the energy levels of the HOMO are sufficiently different (ΔE > 0.2 eV) from the other MOs to be considered independent. In each case the HOMO was located on the metal 5d (Ir) centre and cyclometalated phenyl rings (Fig. S1, ESI †), with little or no coverage of the ancillary diimine ligands. However, varying the diimine ligands does impart a subtle perturbation of the HOMO energy (E HOMO = −7.66 to −7.84 eV). The LUMO for the complexes is, in many cases, close enough in energy to be considered isoenergetic with other close-lying MOs (e.g. LUMO + 1, LUMO + 2). For the complexes where this is not the case (3c, 3f, 3g and 3h), the LUMOs are predominantly delocalised over the diimine ligands as with previous studies of related compounds. 7d For 3a, 3b, 3d and 3e the orbitals show a mixture of diimine and phenylquinoline localisation. The diimine imparts a larger degree of variation in the energy levels of the LUMO energy (E LUMO = −4.71 to −5.47 eV), with 3h showing the lowest LUMO energy level (E LUMO = −5.47 eV) and, therefore, smallest bandgap (E bandgap = 2.29 eV). These results suggest that the lowest energy absorption is predicted to comprise of significant MLCT character and that variation of the diimine ligand could lead to a small degree of tuneable optical properties within this series of complexes. The corresponding cinchophen complexes (4a-h) showed the same general localisation of the frontier orbitals, but revealed a drop in both the HOMO and LUMO energies by an average of 0.12 eV and 0.13 eV, respectively (Fig. S2, ESI †).

Electrochemical studies
The electrochemical characteristics of the [Ir(epqc) 2 (N^N)]PF 6 (3a-h) complexes were studied in de-oxygenated CH 2 Cl 2 . The HOMO energy levels (E HOMO ) were determined from the ionisation potential of the first oxidation (Ir 3+/4+ ) by direct correlation with the redox couple of FeCp 2 0/1+ . The cyclic voltammograms, measured at a platinum disc electrode (scan rate υ = 200 mV s −1 ,1×10 −3 M solutions, 0.1 M [NBu 4 ][PF 6 ]as a supporting electrolyte), each showed one non-fully reversible oxidation (Table S2, ESI †), over the range +1.38 to +1.45 V. The extent of the irreversibility can be ascribed to the contribution of the cyclometalating ligands to the electron density of the HOMO, 22 which in this case is calculated to be ca. 45% from DFT studies. The E HOMO values were determined using the relevant equations 23 and the resultant values fall in the narrow range −5.72 to −5.83 eV (Table S3, ESI †). Each complex also showed one or two partially reversible or irreversible reduction waves, assigned to ligand-centred processes involving both the diimine and phenylquinoline ligands, with complex 3h showing at least four reduction processes, some of which must be associated with the highly reducible anthraquinone fragment.

Electronic properties of the complexes
The UV-vis absorption spectra of complexes 3a-h were obtained as aerated MeCN solutions (5 × 10 −5 M) (Fig. 3, Table 3). Strong absorption bands between 250 and 380 nm were assigned to spin allowed 1 π-π* ligand-centered (LC) transitions arising from both the cyclometalated and diimine ligands within each complex. Weaker bands at 380-480 nm in the visible region were assigned to spin-allowed metal-toligand charge transfer bands ( 1 MLCT) with the possibility of spin-forbidden 3 MLCT transitions contributing to the weaker low-energy shoulder. For 3h it is also likely that ligand-centred transitions associated with the anthraquinone moiety also contribute in this wavelength region; previous studies have shown that such species can possess intra-ligand CT (formally n-π*) character, which is likely to contribute to the lower energy parts of the spectral profile. The variation in coordinated diimine ligand imparts only a very minor variation in the wavelength positioning of the visible absorption bands. As expected the absorption spectra of the corresponding carboxylic acid derivatives 4a-g (5 × 10 −5 M MeOH) share many of the same common features. The principal observation from these spectra is that the visible MLCT-based bands appear as a defined transition ca. 430-455 nm.
TD-DFT calculations ( Table 2) in simulated MeCN suggest the assignment of the lowest lying absorption bands as having substantial MLCT character. For 3a the lowest energy absorption involves excitation from HOMO (Ir-5d + phenyl-π)t o LUMO + 1 (quinoline-π*), predicted to lie at 528 nm (oscillator strength = 0.04 au), and so coincides reasonably closely with the lowest energy shoulder feature seen in Fig. 4. A set of stronger bands centred around 370 nm (387 nm, 0.15 au; 375 nm, 0.17 au; 359 nm, 0.10 au) is also predicted, again in good agreement with Fig. 4. These bands consist of varying combinations of Ir-5d and epqc-π orbitals excited into epqc-π* orbitals, but have no contribution from the ancillary bpy-π* orbitals. In comparison, TD-DFT simulation of 3c in MeCN results in broadly the same pattern of predicted bands at 520 nm (0.04 au), 406 nm (0.06 au), 389 nm (0.07 au) and 384 nm (0.09 au), although the higher energy bands are reduced in intensity relative to the low energy MLCT band. In this complex, two low energy bands involving excitation from HOMO (Ir 5d + phenyl-π) to LUMO and LUMO + 1 (bipy-π* and quinoline-π*, respectively) are predicted at 564 and 522 nm,

Paper
Dalton Transactions but with very low intensity (<0.001 au), suggesting the possibility of MLCT/LLCT character. Deprotection of the ester to yield complex 4a was modelled using both the neutral acid and deprotonated carboxylate form. In the neutral complex, calculations predict a red-shift of the absorption bands discussed above by ca. 15 nm, and alteration of the relative intensities of the absorptions, with predicted bands lying at 544 nm (oscillator strength of 0.04 au), 394 nm (0.12 au), 383 nm (0.15 au) and 365 nm (0.18 au). In contrast, simulation of the deprotonated carboxylate form of 4a indicates a substantial blue-shift of the absorption bands, with the lowest energy absorption with significant oscillator strength coming at 456 nm (0.6 au), and a more intense band at 365 nm (0.12 au). The latter approach is in much better agreement with the experimental data, such that modelling 4a as the bis-carboxylate form seems to be more appropriate.
Steady state luminescence measurements were conducted on aerated MeCN solutions, irradiating the 1 MLCT wavelength (λ ex = 450 nm) absorption (Table 3, Fig. 4). The emission maxima for complexes 3a-g show very little variation (627-630 nm), are broad and featureless, and typically characteristic of MLCT-based transitions. 24 This variation is less than that predicted by TD-DFT (see ESI †), possibly suggesting a diminished diimine contribution to the excited state. Corresponding excitation spectra showed that the complexes could be excited up to a wavelength of ca. 520 nm. Time-resolved emission lifetime measurements revealed that the decays were single-exponential, in each case ca. 200 ns, typical of 3 MLCT character. The complexes each exhibited modest quantum yields (Φ) in aerated MeCN, in line with related species. For 3h, the emission profile was very different, with a higher energy peak at 422 nm (τ < 5 ns), which was assigned to a ligand-centred transition arising from the anthraquinone chromophore, with no evidence of a comparable MLCT transition. The absence of 3 MLCT emission is attributed to the quenching of that state by the anthraquinone-based ancillary ligand.
For ease of comparison the corresponding emission and excitation of the deprotected complexes 4a-g were obtained in MeCN, MeOH and water. With the exception of 4f these species all showed a hypsochromic shift of ca. 50 nm in the 3 MLCT emission maxima (Table 4). Similar measurements in water resulted in an emission peak at ca. 595 nm, revealing the solvent-sensitivity and dipolar nature of the excited state. Emission wavelengths in methanol were typically intermediate between those for water and acetonitrile (for example, Fig. 5). Relative to the esterified analogues (3a-g), the measured

Dalton Transactions Paper
lifetimes were generally extended for complexes 4a-g reflecting the increased energy gap. However, in water the lifetime values also varied greatly as a function of the type of diimine ligand, with 4e (likely to be the most hydrophobic of the diimines in this study) displaying the longest lifetime (τ = 619 ns), suggesting greater shielding of the excited state from the surrounding solvent. The longer emission wavelength of 4f in all solvents appears somewhat anomolous and could be due to a number of factors. With reference to the TD-DFT calculations, the protonation state of the cinchophen ligands influences the emission wavelength. Repeating the measurement in 0.1 M NaOH blue-shifted the emission peak to ca. 590 nm, in accordance with the other complexes in the cinchophen-based series. However, it is difficult to rationalise why only 4f (versus 4g, for example) would retain a protonated carboxylic acid form. An alternative explanation considers the role of protonation at the dppz ligand. The closely related species [Ir( ppy) 2 (dppz)] +25 reveals λ em = 630 nm in MeCN and is thus very comparable to 4f. However, H-bonding interactions with the phenazine nitrogens can dramatically influence the emission properties; in fact, [Ir( ppy) 2 (dppz)] + is non-emissive in water, unlike 4f. The reported photophysics of [Ir( ppy) 2 (dppn)] +25 also compare very well (λ em = 583 nm in MeOH) to 4g, and thus the differences in emission wavelength for 4f could be due to protonation interactions at dppz 26 (of course, these would also be sensitive to the addition of 0.1 M NaOH).

Experimental section
All reactions were performed with the use of vacuum line and Schlenk techniques. Reagents were commercial grade and used without further purification. 1 H and 13 C-{ 1 H} NMR spectra were recorded on an NMR-FT Bruker 400 or 250 MHz and 31 P-{ 1 H} NMR spectra on a Joel Eclipse 300 MHz spectrometer and recorded in CDCl 3 or MeOD solutions. 1 H and 13 C-{ 1 H} NMR chemical shifts (δ) were determined relative to internal tetramethylsilane, Si(CH 3 ) 4 and are given in ppm. Low-resolution mass spectra were obtained by the staff at Cardiff University. High-resolution mass spectra were carried out at the EPSRC National Mass Spectrometry Service at Swansea University, UK. UV-Vis studies were performed on a Jasco V-650 spectrophotometer fitted with a Jasco temperature control unit in MeCN or MeOH solutions (5 × 10 −5 M) at 20°C. Photophysical data were obtained on a JobinYvon-Horiba Fluorolog spectrometer fitted with a JY TBX picoseconds photodetection module in MeCN, MeOH or H 2 O solutions. Emission spectra were uncorrected and excitation spectra were instrument corrected. The pulsed source was a Nano-LED configured for 372 or 459 nm output operating at 500 kHz. Luminescence lifetime profiles were obtained using the JobinYvon-Horiba FluoroHub single photon counting module and the data fits yielded the lifetime values using the provided DAS6 deconvolution software. Electrochemical studies were carried out using a Parstat 2273 potentiostat in conjunction with a three-electrode cell. The auxiliary electrode was a platinum wire and the working electrode a platinum (1.0 mm diameter) disc. The reference was a silver wire separated from the test solution by a fine porosity frit and an agar bridge saturated with KCl. Solutions (10 ml CH 2 Cl 2 ) were 1.0 × 10 −3 mol dm −3 in the test compound and 0.1 mol dm −3 in [NBu n 4 ][PF 6 ] as the supporting electrolyte. Under these conditions, E 0 , for the oneelectron oxidation of [Fe(η-C 5 H 5 ) 2 ] added to the test solutions as an internal calibrant, is +0.46 V in CH 2 Cl 2 . 27 Unless specified, all electrochemical values are at υ = 200 mV s −1 . Microanalyses were performed by London Metropolitan University, UK.

Data collection and processing
Diffraction data for 3b and 3e were collected on a Nonius Kappa-CCD using graphite-monochromated Mo-Kα radiation (λ = 0.71073 Å) at 150 K. Software package Apex 2 (v2.1) was used for the data integration, scaling and absorption correction.

Structure analysis and refinement
The structure was solved by direct methods using SHELXS-97 and was completed by iterative cycles of ΔF-syntheses and fullmatrix least squares refinement. All non-H atoms were refined anisotropically and difference Fourier syntheses were employed in positioning idealised hydrogen atoms and were allowed to ride on their parent C-atoms. All refinements were against F 2 and used SHELX-97. 28 CCDC reference numbers 907280 and 907282 contain the supplementary crystallographic data for this paper.

Paper Dalton Transactions
DFT studies DFT geometry optimisation and orbital calculations were performed on the Gaussian 03 program. 29 Geometry optimisations were carried out without constraints using the B3PW91 functional. The LANL2DZ 30 basis set was used for the Ir centers, and was invoked with pseudo-potentials for the core electrons, a 6-31G(d,p) 31 basis set for all coordinating atoms with a 6-31G 32 basis set for all remaining atoms. All optimisations were followed by frequency calculations to ascertain the nature of the stationary point (minimum or saddle point). TD-DFT studies were performed in Gaussian09 33 using the same functional, but with 6-31G(d) on all non-metal atoms, and also included a simulated MeCN or H 2 O environment using the polarised continuum model (PCM) approach. 34 For prediction of absorption spectra, the geometry used to calculate orbital and other properties was used without modification. For prediction of emission, however, the triplet state was allowed to relax to its optimal geometry using unrestricted B3PW91 in the gas phase, prior to solvated TD-DFT.