Tuning a ffi nity and reversibility for O 2 binding in dinuclear Co ( II ) complexes †

and 700 × 10 atm at 40 °C by varying the number of H and Cl atoms in the bridging acetato ligands of [Co2(bpbp)(CH(3−n)ClnCO2)(CH3CN)2] , where bpbp = 2,6-bis(N,N-bis(2-pyridylmethyl)aminomethyl)-4tert-butylphenolate and n = {0, 1, 2, 3}. O2 binds most strongly to the deoxy complex containing the acetato bridge and the O2 affinity decreases linearly as the number of Cl atoms is increased from 0 to 3 in [Co2(bpbp)(O2)(CH3CO2)] , [Co2(bpbp)(O2)(CH2ClCO2)] , [Co2(bpbp)(O2)(CHCl2CO2)] 2+ and [Co2(bpbp)(O2)(CCl3CO2)] . The O2 affinities can be qualitatively correlated with both the pKa value of the parent acetic or chloroacetic acid and the redox potential of the O2 /O2 ̇ − couple measured for the peroxidebridged complexes. The redox potential varies between 510 mV (vs. Fc) for the acetato-bridged complex to 696 mV for the trichloroacetato-bridged system. Despite the clear difference in reactivity in solution, there are no clear trends which can be correlated to O2 affinity in the O–O bond lengths in the X-ray crystal structures at 180 K (1.415(4)–1.424(2) Å) or in the frequencies of the peroxido O–O stretch in the solid-state resonance Raman spectra at 298 K (830–836 cm). Using density functional theory calculations, we conclude that the Co(II) atoms of the deoxy complexes coordinate solvent molecules as auxiliary ligands and that a conformation change of the ligand is involved in the reversible O2 binding process. The alternative of five coordination in the deoxy Co(II) complexes is therefore seen as less likely. The crystal structure and p(O2)50% are also reported for the 1-naphthoato-bridged oxy complex [Co2(bpbp)(O2)(C10H7O2)] , and the O2 binding affinity in that case is also qualitatively consistent with the expectation from the pKa of the parent 1-naphthoic acid.


Introduction
Reversible dioxygen binding is a life-supporting process for respiring organisms carried out by three classes of metalloproteins: hemoglobin, hemerythrin, and hemocyanin, the latter two having dimetallic active sites.Although cobalt has not been found in any naturally occurring dioxygen carrier, it was shown over four decades ago that cobalt-substituted hemoglobin (coboglobin), is able to bind O 2 reversibly, albeit with lower affinity than the native iron protein.Oxycoboglobin is formulated as Co(III) with a coordinated superoxide. 1 Mononuclear cobalt complexes based on salen 2 and corrole 3 ligands can also achieve reversible O 2 binding, although this property can be quenched by formation of peroxido-bridged dimers. 2,48][9] Suzuki and coworkers [10][11][12] have reported that reversible O 2 binding can also occur for dicobalt(II) systems based on alkoxido and phenolato-hinged dinucleating ligands.In these, both cobalt atoms are oxidised from Co(II) to Co(III) by the oxidative addition of dioxygen, thus facilitating the reduction of O 2 to peroxide rather than the highly reactive superoxide as is the case for the mononuclear systems and which are prone to irreversible selfannihilation reactions.
Materials displaying fully reversible O 2 binding properties have potential industrial and medical uses where storage, concentration gradients akin to that shown by the life-sustaining complementary functions of hemoglobin and myoglobin, and release of oxygen need to be finely controlled. 13,14For example, incorporation of complexes able to bind oxygen reversibly into membranes has been used to increase oxygen permeability, thereby facilitating concentration of oxygen from air. 15,16We report here a systematic study of the electronic effects of the supporting bridging carboxylato co-ligand on the O 2 affinity of the deoxy forms of the series of complexes which include those shown in Scheme 1.This work demonstrates a relatively facile method for tuning O 2 affinity, a property that could be used in the design of metal-organic materials for applications in creating gradients of O 2 concentration.(5-oxy) which contains the cobridging naphthalene-1-carboxylato group was prepared analogously to 1-oxy and 2-oxy.The dicobalt(III) oxy complexes are dark brown and the deoxy complexes are pink.The deoxy complexes of the Cl-substituted carboxylates were on some occasions obtained as very air-sensitive pink powders by working under an inert atmosphere.However, single crystals suitable for X-ray structural analysis were not forthcoming.These powders turn brown rapidly on exposure to air.The reversible binding process is observed easily in methanol or acetonitrile solution by warming the oxy complexes in order to effect the dark brown to pink colour change.On cooling, O 2 is absorbed and bound again and the dark brown colour returns.This process can be repeated many times in organic solvent solution.Evacuation and purging with N 2 also removes O 2 .

Synthesis
In accordance with expected ligand-field stabilization effects, the auxiliary bridging groups will be relatively labile in the deoxy dicobalt(II) forms.In this situation, solvent coordination can compete with the carboxylato ligands, especially those that are less basic.In water, the consequence is the scrambling reaction depicted in eqn (1).We have shown previously that the more acidic carboxylic acids favour formation of the dicarboxylato-bridged complexes which do not bind O 2 . 17The generation of [Co 2 (bpbp)(RCOO) 2 ] + species was substantiated by ESI mass spectrometry and was seen to be promoted if a significant amount of water was present.When water ligands fill all of the auxiliary coordination sites, an outer-sphere oxidation of the complex in the presence of air produces the "met" hydroxido-aqua complexes irreversibly, 18 according to eqn (2).Both of these reactions will compete with reversible O 2 binding and in the presence of large concentrations of water and/or when the parent carboxylic acid is relatively acidic, water inhibits O 2 binding.
Resonance Raman spectra performed in the solid state show clearly the band due to ν(O-O) which are observed at 830.63, 837.76, 836.41 and 837.76 cm −1 for 1-oxy, 2-oxy, 3-oxy, and 4-oxy respectively (measurements averaged over several crystal orientations, ESI Fig. S2 †).These values are consistent with the peroxido formulation. 19,20No correlation is evident between the positions of the bands and the electron-withdrawing effect of the carboxylato ligand.For all four complexes, the intensity of the ν(O-O) band decreased over time as the spectra were recorded, and as the colour of the sample changed from dark brown to light pink.This indicated that deoxygenation of the complexes was caused by the laser, either directly by photoactivation or by local heating of the sample.
The deoxygenated complexes could not be structurally characterized in the solid state.Thus it remains unclear if the cobalt ions are five-coordinated or, for example, solvent or counteranion-derived ligands occupy the sixth coordination site in the absence of peroxide.Suzuki and co-workers proposed that closely related and structurally uncharacterized deoxy dicobalt(II) complexes are five coordinated on the basis of solution electronic spectra and solid-state magnetic susceptibility data. 12,21-24

X-ray crystal structures
The coordination mode of bpbp − is identical in all of the oxy complexes (Fig. 1).The phenolato O atom bridges between the two Co atoms and each dipyridylmethylamine groups are fac coordinated to one Co(II) metal centre.The coordination spheres are completed by the RCOO − and O 2 2− groups bridging between the Co(II) metal centres.One O atom of the peroxido ligand (O2) lies trans to an amine donor (N1) and the other (O3) lies trans to a pyridyl donor (N5) of bpbp − .The same applies to the bridging carboxylato O donors (O4 trans to N2 and O5 trans to N4).The bond lengths around the Co atoms are listed in Table 1.The Co-ligand and O-O bond lengths might be expected to decrease stepwise from 1-oxy to 4-oxy as the bridging carboxylato ligand becomes more electron withdrawing.However, there are very few significant differences in the Co-ligand bonds.Likewise, there are no significant differences in the O-O bond lengths.Complex 5-oxy, which has the relatively bulky naphthoato ligand bridging between the two Co atoms, also has an essentially identical coordination sphere.The O-O bond length does appear to be significantly shorter in 5-oxy compared to 1-oxy and 2-oxy, but this does not provide any consistent indication of the p(O 2 ) 50% values, as discussed below.

Oxygen affinity in solution
The oxygen affinities of the deoxy complexes were evaluated in solution using UV-vis spectroscopy.The oxygenation-deoxygenation process is fully reversible and can be repeated several times in acetonitrile, acetone, ethanol and methanol solutions.Following the method described by Drago and co-workers, 25,26 the equilibrium constant for the uptake of O 2 was extracted from a plot of the partial pressure p(O 2 ) against p(O 2 )/(A − A 0 ), where A is the absorbance at the given partial pressure of O 2 and A 0 is the absorbance for p(O 2 ) = 0.A linear least-squares fit through the data points crosses the secondary axis at −1/K, where K is the equilibrium constant of eqn (3 The equilibrium constants at different temperatures were used to construct van't Hoff plots (Fig. 2), from which the thermodynamic data were derived.
It is evident that the increasing number of electron-withdrawing Cl atoms on the auxiliary bridging carboxylato ligand serve to decrease the oxygen affinity in the complexes in going from 1-deoxy to 4-deoxy and at 40 °C, p(O 2 ) 50% = 2.3 ± 0.9 × 10 −3 atm, 23 ± 4 × 10 −3 atm, 132 ± 21 × 10 −3 atm and 700 ± 90 × 10 −3 atm for 1-deoxy, 2-deoxy, 3-deoxy and 4-deoxy, respectively.For comparison, p(O 2 ) 50% for hemoglobin under physiological conditions has been found to be 35.0 × 10 −3 atm, and for myoglobin at 20 °C, p(O 2 ) 50% = 0.76 × 10 −3 atm. 27Modification of the bridging carboxylato ligand provides a facile approach to tailor the oxygen affinity of dicobalt complexes of this type.A good measure for the electron-withdrawing ability of the carboxylato ligand is the pK a value of the parent acid,  a Two complexes in the asymmetric crystallographic unit.

Cyclic voltammetry
Cyclic voltammograms were obtained for complexes 1-oxy-4oxy in acetonitrile solution.At positive potentials versus Fc 0/+ , the bridging peroxido can be reversibly oxidised to a stable superoxido group. 18With a more electron-withdrawing bridging carboxylato ligand, the oxidation of the peroxido ligand becomes energetically more demanding and the reversible redox wave is shifted to higher potentials (Table 2).A linear correlation is observed between Δ r G for the oxygenation reaction and the oxidation potential of the oxy complexes which is associated with a O 2

2−
/O 2 ˙− redox couple (Fig. 3).The trend illustrates that it becomes progressingly harder to remove an electron from the peroxido group of 4-oxy compared to that of 1-oxy.

Electronic structure analysis
The experimental data present a conundrum, in that the progressively weaker binding of O 2 in going from 1-deoxy to 4-deoxy, is not reflected in any of the solid state properties of the oxy complexes.The O-O distances in 1-oxy-5-oxy are identical within experimental error, and there is no apparent trend in the resonance Raman O-O stretching frequencies.The oxygen affinity is, however, a function of both the oxy and deoxy forms, so we now consider possible structures for the latter.Attempts at in situ deoxygenation of single crystals invariably resulted in loss of crystallinity.In the crystal structure of complex 1-oxy at room temperature, the O-O bond length (1.409(4) Å) is shorter than at 180 K (the crystallographic difference compared to 1.423(2) Å being just significant), and this apparent increase in the O-O bond order is probably a precursor to the deoxygenation process.Heating above room temperature causes crystal decomposition, and we have not been able to obtain any definitive information about the structure of the deoxy form.‡   To address the question of the structural nature of the deoxygenated species, we turned to density functional theory.The deoxygenated complexes could contain five-coordinate Co(II), as suggested by Suzuki, 12,[21][22][23][24] or six-coordinate Co(II) through solvent and/or counteranion coordination.The former seems likely in the solid state, while the latter seems more probable in the presence of a large excess of acetonitrile (the conditions under which we have measured p(O 2 ) 50% ).The weakly coordinating ClO 4 − or BF 4 − counteranions should not compete strongly for the empty coordination sites, so this possibility was not considered for the calculations.
We have considered the displacement reaction shown in eqn ( 4), where O 2 replaces either 2, 1 or 0 CH 3 CN ligands in the coordination sphere.Δ r G values for the twelve reactions (n = 0-3) are collected in Table 3, where they are compared to the experimental values.Note that the difficulties associated with computing accurately the entropies of dissociation of CH 3 CN into bulk solvent preclude an accurate determination of the absolute value of Δ r G but we expect these errors to be constant across the series of chemically related compounds, 1-deoxy-4-deoxy, and so we expect to reproduce the observed trend towards stronger binding of O 2 for the complexes with the less electron withdrawing bridging carboxylato ligands.
If solvent is assumed not to coordinate to the deoxy species (i.e.m = 0) then the structures obtained by removing the O 2 unit and reoptimising with two high-spin (S = 3/2) Co(II) centres fail to reproduce the observed tendency towards stronger binding for 1-deoxy relative to 4-deoxy.The two limiting forms (n = 0 and n = 3) have very similar computed Δ r G while the middle two members are slightly more negative.The computed Δ r G values, for the optimised structures where one acetonitrile molecule is coordinating to a single Co(II) centre predict oxygen binding to become stepwise less favourable from 2-deoxy to 4-deoxy, but the trend does not include 1-deoxy.The structures where both Co(II) centres are coordinatively saturated by CH 3 CN placed in the binding pocket of O 2 are quite strained as the removal of the O 2 unit leaves a cavity that is a bit too small to accommodate two CH 3 CN molecules in close proximity to each other.Alternative minima that are 5-14 kJ mol −1 lower in energy are found if the tridentate dipyridylmethylamine moiety with its tertiary amine trans to a carboxylato O is allowed to rearrange from a fac conformation in the oxy form to a mer conformation in the deoxy form.This opens up a coordination site for one acetonitrile molecule trans to the phenolato O atom, directed away from the second acetonitrile ligand (Fig. 4).Whilst this configuration is not the most common for dinuclear complexes of bpbp − it has precedence in the X-ray crystal structures of [V 2 (bpbp)- 28 [Mn 2 (bpbp)(ClO 4 ) 2 (THF)] 2+ 29 and [FeCuF 2 -(bpbp)(H 2 O)] 2+ . 30It is interesting to note that all of these structures contain different first coordination spheres at the two metal ions, in contrast to the majority of complexes of this type.Using the more stable conformation for the coordinatively saturated species we reproduce the trend in Δ r G well, predicting increases in steps of 5 kJ mol −1 per Cl atom substituted into the bridging carboxylato ligand similar to the effect observed in the experimental measurements (Table 3).Based on these results we suggest that 1-, 2-, 3and 4-deoxy have two solvent molecules coordinated to complete an octahedral coordination environment at each cobalt atom.The O 2 binding pathway could conceivably proceed via formation of the transient superoxido intermediate, [Co II (CH 3 CN)Co III (O 2 )-(bpbp)(CH 3 COO)] 2+ , formed through acetonitrile substitution by O 2 at one Co site (that in which the acetonitrile is trans to the amine in Fig. 4).The distal O atom of the resultant superoxide then attacks the second Co site and the cis pyridine on that Co moves around, thereby expelling the remaining acetonitrile.

Conclusions
We have demonstrated that dicobalt(II) complexes of the bpbp − ligand can be easily tuned with respect to oxygen affinity and reversible chemisorption by changing the electronic properties of the bridging carboxylato ligand.The oxy forms of these complexes have been structurally characterised.DFT calculations suggest that the deoxygenated complexes have a pseudo octahedral coordination environment completed by the coordination of two solvent molecules.

Dalton Transactions Paper
Experimental section

Physical measurements
IR spectra were measured as KBr discs using a Hitachi 270-30 IR spectrometer.UV-visible absorption spectra were recorded on a Shimadzu UV-3100 spectrophotometer with an inbuilt Shimadzu CPS-240A thermo-electrical temperature controlled cell holder.Spectrophotometric measurements of 1-5 in acetonitrile, acetone, ethanol and methanol were carried out at 40 °C under various O 2 partial pressures p(O 2 ).Argon and dioxygen gas were mixed using two Hastings HFC-202 mass flow controllers and passed into ∼0.1 mM solutions (gas flow 30 mL min −1 ) of the complex in a quartz cell (1 cm path length).The gas mixture was bubbled through the solution for 5 min before the measurement to ensure that equilibrium had been reached; longer bubbling times gave identical results.Absorption shoulders, absent in the deoxy form, that grow on oxygenation at 380 and 480 nm were monitored.Up to 5% of the solvent evaporated during the collection of a data series and the weight of the cuvette plus solution was measured between each data collection so that the measured absorptions could be corrected for the change in concentration.The binding affinity is given as p(O 2 ) 50% = K −1 and represents the partial pressure of O 2 needed to oxygenate 50% of the complex.The strong binding affinity resulted in saturation effects at high p(O 2 ) and measurements at these partial pressures could not be used in determination of K. Gas mixtures with p(O 2 ) lower than ∼1.5% could not be obtained without exceeding the calibration range of the flow controllers.Elemental analyses were performed at the Chemistry Department II, Copenhagen University, Denmark and Atlantic Microlab, Inc., Norcross, Georgia, USA.Cyclic Voltammetry (CV) was recorded in acetonitrile solution under dry anaerobic conditions using an Autolab system (Eco Chemie, The Netherlands), controlled by the GPES software.The working electrode was a platinum disk, the auxiliary electrode a platinum wire and the reference electrode a Ag/Ag + (0.01 M AgNO 3 ).TBAClO 4 (TBA = t-butylammonium) 0.1 M was used as electrolyte and all potentials are given versus the ferrocene/ferrocenium (Fc 0/+ ) redox couple (E 1/2 = 88 mV vs. Ag/Ag + , ΔE = 75-80 mV).Resonance Raman spectra were measured with a commercially available confocal scanning Raman microscope (Alpha300R) using linearly polarized illumination at the wavelength of 532 nm, 600 lines per mm diffraction grating, and ×100 objective (N.A. = 0.90).CAUTION!Although we encountered no problems during preparation of the perchlorate salts, perchlorate salts of metal complexes are potentially explosive and should be handled with caution in small quantities.Synthesis 2,6-Bis[bis(2-pyridylmethyl)aminomethyl]-4-tert-butylphenol (Hbpbp) was prepared as described previously. 31 peroxide O-O bonds significantly.The perchlorate anions are rotationally disordered and the anion geometry was restrained to be a regular tetrahedron.In 4-oxy, relatively large residual electron density remains in the vicinity of one perchlorate anion, indicative of residual rotational disorder not accounted for by the model.In 5-oxy, there are two complexes in the crystallographic asymmetric unit, related by local inversion pseudosymmetry.

Computational details
Density functional theory (DFT) calculations were performed with the Gaussian09 software package and the functional TPSSh by Tao, Perdew, Staroverov and Scuseria. 34,35The SDD basis set 36 (Stuttgart/Dresden ECP) was used for Co and the TZVP 37 basis set on the peroxido ligand and for the calculations on molecular dioxygen.For all other atoms, the splitvalence basis set of Ahlrich and co-workers 38 was applied with polarisation functions taken from the turbomole basis set library. 39 All structures were optimised in the gas phase and were confirmed to be minima on their potential energy surfaces by calculation of their vibrational frequencies.Solvent effects were estimated through single point calculations on the optimised gas phase structures with a continues polarisation model 40 as implemented in Gaussian09.

Table 2
Measured p(O 2 ) 50% values and reduction potentials of the O 2