Catalytic dioxygen reduction mediated by a tetranuclear cobalt complex supported on a stannoxane core †

The synthesis, spectroscopic characterization (infrared, electron paramagnetic resonance and X-ray absorption spectroscopies) and density functional theoretical calculations of a tetranuclear cobalt complex Co 4 L1 involving a nonheme ligand system, L1 , supported on a stannoxane core are reported. Co 4 L1 , similar to the previously reported hexanuclear cobalt complex Co 6 L2 , shows a unique ability to catalyze dioxygen (O 2 ) reduction, where product selectivity can be changed from a preferential 4e − /4H + dioxygen-reduction (to water) to a 2e − /2H + process (to hydrogen peroxide) only by increasing the temperature from − 50 to 30 °C. Detailed mechanistic insights were obtained on the basis of kinetic studies on the overall catalytic reaction as well as by low-temperature spectroscopic (UV-Vis, resonance Raman and X-ray absorption spectroscopies) trapping of the end-on μ -1,2-peroxodicobalt( III ) intermediate 1 . The Co 4 L1 - and Co 6 L2 -mediated O 2 -reduction reactions exhibit di ﬀ erent reaction kinetics, and yield di ﬀ erent ratios of the 2e − /2H + and 4e − /4H + products at − 50 °C, which can be attributed to the di ﬀ erent stabilities of the μ -1,2-peroxodicobalt( III ) intermediates formed upon dioxygen activation in the two cases. The deep mechanistic insights into the transition-metal mediated dioxygen reduction process that are obtained from the present study should serve as useful and broadly applicable principles for future design of more e ﬃ cient catalysts in fuel cells.


Introduction
Significant, attention has been focused in recent years on the synthesis of transition metal based dendrimer structures owing to their diverse applications in various fields. 1 In particular, these dendrimers, in many cases, allow synergistic interactions between the individual transition metal centers in carrying out a variety of important transformations. The organooxotin clusters are in particular attractive because of the diversity of arrangements that they adopt, such as ladder, O-capped, cube, butterfly, drum, one, two and three-dimensional structures, (1D, 2D, and 3D). 2-6 Furthermore, incorporation of redox-active transition-metal centers into the stannox-ane clusters has previously led to the demonstration of important reactivity patterns. 7, 8 For example, an extensive cooperative effect between the Cu centers was observed during the cleavage of supercoiled DNA catalyzed by a hexanuclear Cuporphyrin complex, supported on a stannoxane core. 7a In our group we have previously demonstrated the ability of a nonheme stannoxane based hexanuclear ligand system to undergo O-O bond formation 7b and O-O bond cleavage reactions, 8 when bound to iron(II) and cobalt(II) centers, respectively. In the present manuscript we report the synthesis, characterization and X-ray structure of a tetranuclear stannoxane based non-heme ligand system (L1), and a detailed kinetic study of the catalytic dioxygen reduction reaction mediated by the corresponding cobalt complex Co 4 L1. Notably, catalytic reductions of O 2 to water or H 2 O 2 have tremendous technological significance. 9-12 However, in contrast to biology, where cheap and readily available transition-metal complexes of Fe, and Cu are employed for O 2 reduction, 13 high loadings of a precious metal like platinum is warranted for achieving appreciable reactivity during abiological O 2 -reduction reactions. 12d Thus the present study is relevant to the ongoing research activities that are being dedicated towards the development of O 2 reduction catalysts based on nonprecious metals. 14 Furthermore, it provides deep mechanistic insights into the factors that control two-vs. four-electron reductions of O 2 , thereby providing useful and broadly applicable principles for the future design of more efficient O 2 reduction catalysts.

Results and disscussion
Synthesis and characterisation of L1 The condensation reaction (Schemes S1 and S2 †) of equimolar amounts of di-n-butyltin oxide and 4-(1,3-bis(2-pyridylmethyl)-2-imidazolidinyl)benzoic acid in toluene afforded L1 as a pale yellow solid. The molecular structure of L1 shows that a planar Sn 4 O 2 core supports the four metal-binding sites ( Fig. 1: top). This is in contrast to the situation reported earlier for the hexanuclear non-heme ligand system L2, where six metal-binding sites were located in a wheel-like arrangement around a central Sn 6 O 6 prismane core ( Fig. 1: bottom). The stannoxane core in Ligand L1 adopts a ladder framework with two central and two terminal tin atoms. The tetranuclear structure of L1 is maintained in solution. 119 Sn NMR spectrum of L1 exhibits two sharp singlets of equal intensity at −210.82 ppm and −213.81 ppm (Fig. S1 †), which is the characteristic signature for a planar Sn 4 O 2 core. 2-6 The infrared spectrum shows four vibrations at 1622 cm −1 , 1591 cm −1 , 1569 cm −1 , and 1545 cm −1 for the carboxyl absorptions (ν COO ), and one strong band at 682 cm −1 assigned to ν Sn−O for the Sn 4 O 2 core (Fig. S2 †).

Synthesis and characterization of Co 4 L1
The reaction of L1 with 4 equiv. of Co(CF 3 SO 3 ) 2 in acetone yields Co 4 L1 as a dark yellow powder in 70% yield (Scheme 1). The C, H, and N content of Co 4 L1, determined by elemental analysis, established the presence of four cobalt atoms per tetrameric ligand, with two triflates associated with each cobalt (see ESI †).
The infrared spectrum of Co 4 L1 depicts the characteristic vibrations of the Sn 4 O 2 core at 1625 cm −1 , 1593 cm −1 , 1572 cm −1 , 1549 cm −1 , and 682 cm −1 (Fig. S2 †). These vibrations are only slightly shifted relative to that of L1, which reveals that the tetranuclear arrangement is also maintained in Co 4 L1.
Electronic and structural information of Co 4 L1 were obtained from X-ray absorption spectroscopy (XAS) in conjunction with density functional theory (DFT) calculations. The near edge structure (XANES) was used for determination of the oxidation states, whereas the extended fine structure (EXAFS) unraveled the local site geometries around the Co atoms. The spectra are displayed in Fig. 2, the corresponding fit values are collected in Table 1. The XANES spectrum of Co 4 L1 (blue trace) is displayed together with spectra from Co reference compounds of known oxidation states (Co 2+ , Co 2.66+ , Co 3+ ), see Fig. 2a; it is consistent with a Co 2+ oxidation state in Co 4 L1 Fig. 1 Comparison of the distances between the metal binding sites in L1 (top) and L2 (bottom). X-ray crystal structure of L1 and L2 with 30% ellipsoid probability of the atoms. Hydrogen atoms and the n-butyl ( n Bu-) groups on the tin atoms have been omitted for clarity. Color code: nitrogen-blue; carbon-grey; oxygen-red; tin-green.
Scheme 1 Synthesis of the tetra-nuclear cobalt(II) complex (Co 4 L1) from the tetra-nuclear stannoxane ligand (L1) and the formation of the cobalt(III)-peroxo complex (1). (Fig. 2b). The EXAFS of Co 4 L1 could be well fitted by four shells, with one shorter N-shell with coordination number (N) of 5, a longer N/O-shell with N = 1, and two C-shells with N = 3 and 2 ( Fig. 2c and d). Attempts to fit Co 4 L1 with a sum of N = 5 in the first two shells (instead of 6) significantly worsen the fit parameters. In principle, there are up to nine C-atoms within a radius of 3.5 Å around the Co-atom, however, due to the pronounced inhomogeneity of the Co-C distances, these shells may partially cancel each other out. The average Co-O/N distance is found to be 2.17 Å.
Since there are no X-ray diffraction (XRD) structures available for Co 4 L1 and the Co atoms are mainly surrounded by O, N and C atoms with similar scattering properties, the EXAFS fits may suffer from non-uniqueness and misinterpretations. In order to reduce this problem as well as to obtain suitable phase functions for the fits, DFT calculations were conducted in the experimentally observed (from EPR; Fig. S3 †) S = 3/2 spin state for a series of potential structural variants of the monomeric subsection of the organic ligands, starting from the modified XRD structure of the tetrameric stannoxane ligand (see Fig. 3 and S4 †). This approach is justified as there are no intra-molecular electronic interactions detectable between adjacent Co(II) sites, as evident from the X-band EPR spectrum of Co 4 L1, which exhibits a major axial signal with effective g′ ⊥ = 4.01 and g′ ∥ ≈ 2.0 corresponding to the S = 3/2 ground state (Fig. S3 †). Structural variants include the   coordination of triflate (OTf ) and/or solvent acetone molecules in cis-or trans-orientations, and with or without inherent molecular symmetry, (see legend of Fig. S4 †). Four of the six DFT models show hexa-coordinated Co(II) (no. 1, 2, 5, and 6), and the other two show penta-coordination (no. 3 and 4; see Table S1 †). Since unrestrained EXAFS fits of Co 4 L1 clearly indicate hexa-coordination, the corresponding DFT models are considered to be closer to the actual structure of Co 4 L1. The average Co-X (X = N, O) bond distances, however, vary in the narrow range of 2.09 to 2.19 Å in all six models, which can hardly be discriminated by EXAFS, but all of them are close to the 2.17 Å obtained from the experiments. The lowest molecular energy is obtained for the hexa-coordinated model 5, followed by penta-coordinated model 3, which is only 5 kJ mol −1 higher in energy. Fine structural details are visible in the EXAFS wave, which promises more insight into the real structure than geometric and energetic considerations alone. Accordingly, EXAFS was calculated for all six small DFT models and compared to the experimental spectra of Co 4 L1, see Fig. S5 in the ESI. † Here again, model 5 apparently gives the best match to the spectrum of Co 4 L1, followed by the other hexa-coordinate models 2, 6 and 1, whereas the two penta-coordinate models 3 and 4 give the worst match. Taking all results into accountgeometry, energy and EXAFS -DFT model 5 seems to be the closest representative for the structure underlying in the experimental data of Co 4 L1 we have so far. In this model, the six Oand N-atoms bound to the central Co-atom are aligned in a low symmetrical fashion, which might be described as quadratic-pyramidal (O 2 N 3 ) with one (extra) N-atom below but close to the quadratic plane, (see Fig. 3). However, these results seemingly are in contradiction with the XANES spectrum of Co 4 L1, which looks like typical octahedral or trigonal-bipyramidal (i.e. high local coordination symmetry) compounds, e.g. the hexa-aqua Co(II) compound used for reference (black line in Fig. 2a).
The answer to this riddle might be the potentially underrated electronic and steric effect of the agostic proton in DFT model 5, which is part of the carboxylated phenyl group, see Fig. 3. Since the organic ligand system has only limited flexibility, an unoccupied coordination site can be filled by a C-H⋯Co contact, which changes the picture. Taking the H-atom into account, the coordination geometry is rather an N-capped O 2 N 3 H-octahedron, see green lines in Fig. 3, than a square pyramid. In order to understand this in more detail, the electronic situation of model 5 was thoroughly analyzed by means of Real-Space Bonding Indicators (RSBIs) extracted from the computed electron density (ED, Fig. S6 †). Fig. S6b † shows the spin-density, the majority of which is localized at the Co-atom, as expected, and with minor contributions at all six non-H-atoms. Bond topological analysis of the ED according to the Atoms-In-Molecules (AIM) 15 theory, however, finds bond critical points (bcp) and thus bond paths to all seven O-, N-and H-atoms, see Fig. S6c. † AIM theory also provides atomic basins. Mapping the ED distribution on them discovers bonding regions and strength of chemical interactions. The AIM atomic Co basin has the basic shape of a cube (typical for octahedral ligand sphere) with one edge cropped by the capping N-atom, see Fig. S6d. † The more interesting point, however, is that the shape of the basin is also flat along the Co⋯H axis, although the agnostic interaction is quite weak (only little ED accumulation on the respective cube face). This "regular shape" of the Co-atom is also visible applying the Non-Covalent interactions Index (NCI), 16 which uncovers noncovalent bonding aspects of strong medium and even very weak atom-atom contacts, see Fig. S6e. † Ring-shaped bluecolored NCI basins indicate dominating covalent bonding aspects (one O, one N), whereas disc-shaped blue-colored NCI basins indicate dominating non-covalent bonding aspects (one O, three N). The agostic Co⋯H contact is represented by a flat and extended greenish-blue colored NCI basin, being typical for weak non-covalent interactions, such as H⋯H or metallophilic contacts. AIM and NCI are complemented by the Electron Localizability Indicator (ELI-D), 17 which dissects realspace into regions/basins of (non-) bonding electron pairs, resembling in a way the Lewis-picture of chemical bonding. An iso-surface representation is shown in Fig. S6f. † Highlighted (solid, green) are the six non-bonding d-electron ELI-D basins of the Co-atom, which altogether form a regular polyhedron in order to minimize electron-electron repulsion to the electron pairs from the electron donating ligand atoms, according to the well-known "key-lock" arrangement in transition metal chemistry.

Co 4 L1 catalyzed dioxygen reduction reaction
The evaluation of the catalytic activity of Co 4 L1 towards oxygen reduction was carried out using the Fukuzumi and Guilard's method; 18 decamethyl-ferrocene was employed as a one electron donor, triflic (TfOH) or fluoroboric (HBF 4 ) acids were used as proton source, and, in their presence, O 2 was set to react with a catalytic amount of Co 4 L1 in acetone. The occurrence of the oxygen reduction reaction was proved by the formation of decamethylferrocenium ion (Fc* + ) with a characteristic absorption band at 780 nm ( Fig. 4; ε 780 nm = 520 M −1 cm −1 ). 19,20 Notably, the rate and yield of formation of Fc* + is not significantly affected by the nature of the proton source (TfOH or HBF 4 Fig. S7d †), thereby suggesting that the conjugate bases (OTf − or BF 4 − ) play no major role in controlling the efficiency of the O 2 -reduction reactions. However, the concentration of Fc* + formed in the complex Co 4 L1-catalyzed reduction of O 2 by Fc* is dependent on the temperature at which the reactions were performed (Fig. 4 bottom, S7a-c †). At 30°C 0.35 mM of Fc* + ion is generated in the reaction, which corresponds approximately twice that of the O 2 concentration (0.18 mM). Thus, only two-electron reduction of O 2 occurs at 30°C. With decreasing temperature, the amount of Fc* + generated from O 2 reduction increases, presumably because of the increasing contribution of the four-electron reduction of O 2 . At 25°C the amount of Fc* + formed is 0.44 mM, which is 2.5 times that of the O 2 concentration. The mechanism shifts predominantly to a four-electron reduction process at −50°C; the amount of Fc* + gener-  (Fig. S8 †).

Reaction of Co 4 L1 with dioxygen to form 1
An acetone solution of Co 4 L1, when treated with O 2 saturated acetone at −50°C, results in the formation of an orange species 1 with an intense absorption maximum λ max (ε max , M −1 cm −1 ) centered at 464 nm (12 200 M −1 cm −1 ). As the temperature is increased, the absorption band at 464 nm due to 1 is decreased (Fig. 6: top one). This process is reversible in the temperature range −50 to 30°C. The resonance Raman (rR) spectrum ( Fig. 6: bottom one) of 1 in acetone-d 6 displays two isotopically sensitive vibrational bands at 862 (O-O stretching mode of a peroxo ligand) and 595 cm −1 (Co-O stretching mode), which are downshifted to 808 and 561 cm −1 , respectively, in 18 O 2 prepared samples. XAS studies were also performed to probe the oxidation state and the coordination environment of Co in 1. The XANES spectra of 1 when compared with that of Co 4 L1 and other reference compounds reveals an almost complete oxidation from Co 2+ to Co 3+ during the transformation of Co 4 L1 to 1. Additionally, the edge shape of 1 shows minor altera-

Dalton Transactions Paper
This Similarly, no reaction of 1 with TfOH was observed in the absence of Fc*. However, in the presence of both TFA and Fc* 1 underwent fast decay, presumably by a proton coupled electron transfer (PCET) mechanism to form water as the major product (Fig. 7 top). At 25°C in the absence of TFA, no reduction of 1 by Fc* was observed, very similar to our findings at −50°C. However, in presence of TFA, even in the absence of Fc*, fast decay of 1 was observed (Fig. 7 bottom), with the release of H 2 O 2 by a proton transfer (PT) mechanism.
The temperature dependence of the PT and PCET processes will be the controlling factor in determining the temperature dependence of the 4e − /4H + vs. 2e − /2H + reductions of dioxygen mediated by 1. We therefore compared the temperature-dependence of the PCET and PT processes of 1 at various temperatures (Fig. 7, 8 (Fig. 7 bottom). PT is found to vary with temperature at a much more drastic rate relative to that of PCET, and it becomes the predominant mechanism for the reduction of 1 at temperatures >11°C (Fig. 8a).

Conclusions
In our previous study 8 we reported the synthesis and characterization of a hexanuclear cobalt complex Co 6 L2 involving a nonheme ligand system, L2, supported on a Sn 6 O 6 stannoxane core ( Fig. 1: bottom), whose cobalt complex acts as a unique catalyst for dioxygen reduction, whose selectivity can be changed from a preferential 4e − /4H + dioxygen-reduction (to water) to a 2e − /2H + process (to hydrogen peroxide) only by increasing the temperature from −50 to 25°C. Herein, we report the synthesis and characterization of a tetranuclear Co 4 L1 complex, supported on the stannoxane core, and compare its dioxygen reduction ability with that of Co 6 L2 (Fig. 8). The temperature dependence of the product selectivity of the catalytic dioxygen reduction is still observed in Co 4 L1; however, some subtle differences are noted relative to Co 6 L2, which can be attributed to the different nuclearity and Co-Co distances in the two cases (Fig. 1).
In L2 all the six plausible metal binding sites are equidistant from each other at 11.365 Å. This is in contrast to L1, where only two of the four metal binding sites are in close proximity to each other. The more symmetric nature of L2 ensures an efficient cooperative dioxygen binding in Co 6 L2 relative to Co 4 L1, which results in the lower stability of 1 compared to that of the corresponding {[L2(Co III (O 2 )Co III ) 3 ]} 12+ complex 2 that is formed upon dioxygen activation of Co 6 L2.
The faster self-decay rate (1 × 10 −4 s −1 for 2 vs. 2 × 10 −3 for 1 at 25°C), as well as the 16 cm −1 downshift in the Co-O vibration energy (ν(Co-O) for 1 is 595 cm −1 and 611 cm −1 for 2) 8 in 1 relative to 2, is consistent with the lower stability of 1. Accordingly, as previously observed, the high enthalpic stability of 2 makes its formation at −50°C highly favored that leads to the complete oxygenation of Co 6 L2. Complex 2 then undergoes O-O bond cleavage via a PCET mechanism to yield water as the sole product under catalytic turnover conditions. The rate constant of the reaction was found to be independent of the O 2 concentration; the kinetic equation at −50°C for Co 6 L2 is where "k cat " is the third-order rate constant for the catalytic 4e − -reduction of O 2 by Fc* at −50°C and k obs is the pseudo first-order rate constant. In contrast, an equilibrium binding of O 2 occurs for Co 4 L1, even at −50°C, so that the rate of the catalytic reaction shows a linear dependence on the O 2 concentration (Fig. S10 †). The rate equation for Co 4 L1 is where "k cat " is the fourth-order rate constant for the catalytic 4e − -reduction of O 2 by Fc* at −50°C and k obs is the pseudo first-order rate constant. Furthermore, Co 4 L1 catalysed O 2 reduction yields 15-30% H 2 O 2 at −50°C, in contrast to Co 6 L2 for which no H 2 O 2 production could be detected at this temperature. However the rate constant of the two-electron O 2 reduction at +25°C is a fourth-order process for both Co 6 L2 and Co 4 L1 (Fig. S11 †).
The constraints imposed by the stannoxane core ensure entropic instability of both 1 and 2. This is mainly because of the large reduction in the Co-Co distances that is associated with their formation. Although experimental determination of the Co-Co distances in Co 4 L1, Co 6 L2, 1 and 2 was not possible, approximate shortening of ∼2.4 Å (from a distance of 6.82 Å in L1 to the DFT calculated distance of 4.48 Å in 1) and ∼7 Å (from a distance of 11.36 Å in L2 to the DFT calculated distance of 4.48 Å in 2) can be predicted for dioxygen binding at Co 4 L1 and Co 6 L2 complexes, respectively. This would impose a large strain on the μ-1,2-peroxo-dicobalt(III) cores in 1 and 2, which would attribute to their instability at higher temperatures upon protonation leading to the formation of H 2 O 2 as the major product. Thus for both Co 4 L1 and Co 6 L2, an equilibrium binding of O 2 will take place at 25-30°C, such that only a small portion of Co 4 L1 and Co 6 L2 will be converted to 1 and 2, respectively. This would also explain the experimentally observed direct correlation of the reaction rates to oxygen concentration at 25-30°C in both cases.
In summary, the Co 4 L1 complex like the previously reported Co 6 L2 complex is a unique catalyst for dioxygen-reduction reaction, whereby the product selectivity can be changed from

Dalton Transactions Paper
a predominant 4e/4H + reduction process (to water) at −50°C to a 2e − /2H + process at 25-30°C. μ-1,2-peroxo-dicobalt(III) complexes 1 and 2 are proposed as plausible reactive intermediates, which are reduced to H 2 O by a PCET mechanism at −50°C, or to H 2 O 2 by a proton transfer mechanism at 25-30°C. For both 1 and 2, the PT rates are found to vary drastically with temperature relative to the PCET rates, and PT becomes the predominant mechanism at 11°C for 1 and at 19.5°C for 2. The ∼10°C reduction in the transition temperature for 1 can be attributed to its reduced stability relative to 2, as also evident from the faster self-decay rate and lower ν(Co-O) vibration energy in 1 relative to 2. This study, therefore, underlines the importance of subtle electronic and steric changes in the reactivity of the biologically relevant metal-dioxygen intermediates, and how they can control the 2e − /2H + vs. 4e − /4H + product selectivity in catalytic dioxygen reductions.

Conflicts of interest
There are no conflicts to declare.