1. In situ X-ray absorption spectroscopy. In this study we determined the presence of CoIV=O intermediate state using in situ XANES and EXAFS of electrochemical WOR of (1) at 1.4 V vs. Ag/AgCl potential in a basic water solution (pH = 12) for prolonged time (8 hours) (Fig. 2a, b, Table S1). XAS measurements were performed for initial (1), in situ bulk electrolysis (BE) of (1) at pH=12, and (1) oxidized with cerium (IV) ammonium nitrate (CeIV) at pH = 1 (measured at 20K) (Fig. 2a, b). Due to high activity of this catalyst, pH of the electrochemical solution of (1) during the bulk electrolysis shifted towards increasing acidity; pH monitoring and basification occurred prior to each X-ray scan, as necessary. Formation of Co species in high oxidation states is reflected in the shift to the higher energy and shape change of the Co K-edge (Fig. 2a). At pH = 1, XAS of the (1) oxidized with 20 equiv of CeIV also revealed K-edge shift to the higher energy (Fig. S1).
Compound (1) contains four CoIII ions, while (1)+ has a single CoIV. According to DFT calculations, further oxidation of (1)+ to the [CoIII2CoIV2O4]2+ is possible (Tables S2, S3) in non-protic solvents. Co K-edge XANES measured during BE of (1) in water at pH = 12 (Fig. 2a) also reflects a formation of the [CoIII2CoIV2] intermediate as reported earlier for electrolysis in CH3CN ( ~1 eV K-edge shift compared to solution before the BE).21 In addition, the top of the edge shape becomes pointier (Fig. 2a). This change indicates a potential geometric change in the cluster and change in the form of the cluster frontier’s orbitals. EPR analysis of the samples quickly frozen from the in situ cell during prolonged BE under applied potential indicates mostly EPR silent species, agreeing with [CoIII2CoIV2] oxidation state assignment (Fig. 3a, b).
EXAFS analysis indicates considerable changes in the structure of (1) at pH=12 and the applied potential (Fig. 2b). EXAFS fits of the initial (1) and (1)+ formed by CeIV oxidation agree with known structures of these species (Table S4, Fig. S2). In situ experiment at pH=12 was conducted at two different beamtimes, ensuring EXAFS reproducibility (Table S1). First coordination sphere of Co is comprised of oxygens as m3-oxo bridges, expected Co-O a ~1.85 Å, and single nitrogen ligand from pyridine with Co-N at ~1.90 Å. However, to achieve satisfactory EXAFS fits of the in situ data, addition of short ~1.65-1.70 Å Co-O vector is required (Table S1). This distance agrees well with DFT model of [CoIII2O4CoIV=O/CoIV(CH3CO2)mono]0 (Table 1, S5, Fig. 1d) and is similar to previously-reported Co=O bond length of 1.72Å.27 For Co-O at ~1.67 Å, both N=0.25 and N=0.5 vectors give similar fit results. Thus, EXAFS fits cannot distinguish whether the intermediate has one or two CoIV=O units, while higher number of units is unlikely. Our DFT analysis (below) shows that one CoIV=O fragment per [Co4O5]n+ may enable O-O bond formation. DFT of [CoIII2O4CoIV=O/CoIV(CH3CO2)mono]0 model also predicts two groups of Co-Co distances: one shorter (~2.73 Å) and one longer (~2.86-2.91 Å). EXAFS fits improve if Co-Co vector is split into two shells (Table S1).
2. In situ EPR. Oxidation of (1) by one electron results in well-known paramagnetic (S=1/2) cubium signal (1)+.14 This signal can be produced by oxidation with CeIV at acidic pH or by bulk electrolysis at neutral pH~7 where the catalyst is inactive (Fig. S6). EPR samples from bulk electrolysis in pH=12 solution at 1.4 V vs. Ag/AgCl electrode were quickly frozen and analyzed by X-band EPR at 20K (Fig. 3a, b). Such samples are largely EPR silent, except for the background due to frozen oxygen evolved under these BE conditions. Loss of EPR signal is consistent with further sample oxidation to (1)2+ state, which can have [CoIII2O4CoIV=O/CoIV(CH3CO2)mono]0 form, as discussed below based on XAS analysis and DFT. Traces of cubium (1)+ signal are detected upon sample melting (data not shown). New, short-lived S=1/2 signal is detected when spectra are closely analyzed (Fig. 3, b). The signal peaks at g ~2.08 and has gzz component centered around g~1.80. Hyperfine splitting ~50-60 G is noticeable around gzz component and to the left of the g ~2.08 peak. It is assigned to the 59Co I=7/2 hfs. The g-tensor values near g~2 indicate a significant electron localization on the ligand. Low g-tensor values were previously reported for CoIII-O2- superoxo com-
Table 1. DFT analysis of O-O bond formation by CoIII2CoIV2 state of the Co4O4 - cubane
Reactions of O-O bond formation
|
DG, eV
|
BP86
|
TCoIII2O4CoIV=O/CoIV(CH3CO2)mono + H2O = SCoIII2O4CoIII-OOH/CoIII(CH3CO2H)mono
|
+0.92
|
TCoIII2O4CoIV=O/CoIV(CH3CO2)mono / H2O = SCoIII2O4CoIII-OOH/CoIII(CH3CO2H)mono
|
+1.12
|
TCoIII2O4CoIV=O/CoIV(CH3CO2)mono = SCoIII3O4CoIII(OOC(O)CH3)bridging
|
+0.56
|
BP86 with 15% Hartree Fock
|
TCoIII2O4CoIV=O/CoIV(CH3CO2)mono + H2O = SCoIII2O4CoIII-OOH/CoIII(CH3CO2H)mono
|
-0.55
|
TCoIII2O4CoIV=O/CoIV(CH3CO2)mono / H2O = SCoIII2O4CoIII-OOH/CoIII(CH3CO2H)mono
|
-0.19
|
Additional oxidation reactions
|
Redox potential, E, eV
|
TCoIII2O4CoIV=O/CoIV(CH3CO2)mono =DCoIIIO4CoIV=O/CoIV,CoIV(CH3CO2)mono + e-
|
+0.86
|
SCoIII2O4CoIII-OOH/CoIII(CH3CO2H)mono = DCoIII2O4CoIV-OOH/CoIII(CH3CO2H)mono + e-
|
+0.12
|
plexes.33, 34, 35 Below we compare DFT computed 59Co hfs and conclude that an assignment of this new signal is possible as CoIV-OOH terminal peroxo species.
3. DFT analysis. Multiple potential mechanisms for O-O formation by [Co4O4]n+ cubane have been discussed (Figure 1b, c). 13, 14, 15, 16, 17, 21, 22, 24 The experimental consensus that bridging oxygens do not exchange under catalytic conditions suggests that O-O bond formation involving bridging oxygens should not be considered. It is experimentally established that CoIII2CoIV2 is a catalytically competent state.
One observation that [Co4O4(bpy)4Ac2]2+ is not an active catalyst called for a specific mechanism with side-on peroxide formation from Co(OH)2 corner (Figure 1b).14 The redox potential for (1) to form CoIII2CoIV2 in acetonitrile solvent21 is in Table S2. Cyclic voltammogram data for (1) in acetonitrile has confirmed the earlier result21 (Figure S3), while in the same electrochemical window the bpy analog reversibly oxidizes at ~+0.71 V.6, 36 However, no second redox couple was observed. Hence, oxidation of the bpy derivative requires greater energy input, likely due to its positive charge. DFT calculations confirm this effect (Table S3). Note that protic solvents are known to form hydrogen bonds to the cubane μ3-oxos, reducing donation to cobalt and raising the potentials in aqueous solution.17 This effect causes (1) and the bpy analog to reach the CoIII3CoIV level at identical potential in water. Nonetheless, the CoIII2CoIV2 level was observed for (1) and remain undetected for the bpy analog, potentially explaining lack of its catalytic activity.
The [CoIII2CoIV2]2+ cubane can be in a triplet (S=1) or a singlet state (S=0). Both these states are energetically equal.21 We interpret in situ XANES results as an observation of CoIII2CoIV2 intermediate, agreeing with the earlier studies.21 EXAFS data indicate that at pH=12 the intermediate contains at least one CoIV=O group. The minimally modified structure where this group can be achieved is [CoIII2O4CoIV=O/CoIV(CH3CO2)mono]0 where the single bridging acetate becomes a terminal ligand (Figure 1d). For this structure, the triplet state is ~0.7 eV lower than the singlet (Table S5). DFT-computed CoIV=O in this intermediate (~1.675 Å) matches well the Co-O distance (~1.67 Å) detected in EXAFS (Table S1). Co-Co distances that are separated into two groups, the shorter (~2.73 Å) and the longer (~2.85 Å), also match the DFT model (Table S1).
Recent time-resolved rapid-scan IR observation of photo-oxidized cubane at pH=12.3 indicates no free acetate at reaction conditions, suggesting that all acetates remain at least as the terminal ligands to the cubane.29 However, this observation contradicts the paper’s following interpretations: assigning ~833 cm-1 band to Co-OO-Co peroxo species. Thus, we explored [CoIII2O4CoIV=O/CoIV(CH3CO2)mono]0 reaction with water via WNA mechanism which has negative (-0.27 eV) DG in the calculations done with 15% of Hartree Fock exchange (Table 1, Fig. 1d). Multiple
Table 2. DFT computed key spectroscopic characteristics of the reactive intermediates in BE of (1) in water solution at pH = 12 at 1.4V vs Ag/AgCl applied potential.
Intermediate/
59Co hfs in gauss for S=1/2 species
|
Vibration (parentheses is HF)
|
Isotope shift*
|
TCoIII2O4CoIV=O/CoIV(CH3CO2)mono
|
Co=O: 744 cm-1
(911 cm-1)
(1.61 angstroms)
|
Co=18O: 713 cm-1
|
SCoIII2O4CoIII-OOH/CoIII(CH3CO2H)mono
|
-O-OH: 762 cm-1
(985 cm-1)
(Co-O: 1.86A)
(O-O: 1.41A)
|
-O-OD: 761 cm-1
-18O-OH: 738 cm-1
-18O-18OH: 718 cm-1
-O-18OH: 742 cm-1
|
SCoIII2O4CoIII-OOH/CoIII(CH3CO2)mono
|
-O-OH: 821 cm-1
|
|
SCoIII2O4CoIII-OO/CoIII(CH3CO2)mono
|
-O-OH: 932 cm-1
|
|
CoIII2O4CoIV-OOH/CoIII(CH3CO2H)mono
59Co hfs: Axx=34; Ayy=58; Azz=78
|
-O-OH: 807 cm-1
|
-O-OD: 809 cm-1
-18O-OH: 782 cm-1
-18O-18OH: 762 cm-1
-O-18OH: 786 cm-1
|
CoIII2O4CoIV-OOH/CoIII(CH3CO2)mono
59Co hfs: Axx=30; Ayy=52; Azz=75
|
-O-OH: 875 cm-1
|
-O-OD: 873 cm-1
-18O-OH: 849 cm-1
-18O-18OH: 826 cm-1
-O-18OH: 852 cm-1
|
CoIII2O4CoIV-OO/CoIII(CH3CO2)mono
59Co hfs: Axx=16; Ayy=15; Azz=22
|
-O-OH: 1090 cm-1
|
|
SCoIII3O4CoIII(OOC(O)CH3)bridging
|
Co-O-O-C
725 cm-1
|
Co-18O-O-C
720 cm-1
|
CoIII2O4CoIII-OO-CoIII
|
O-O 718 cm-1
|
|
CoIII2O4CoIV-OO-CoIII
59Co hfs: Axx=14; Ayy=11; Azz=23
59Co hfs: Axx=13; Ayy=11; Azz=23
|
O-O 905 cm-1
|
|
* Isotope shifts were computed for most promising intermediates.
research groups noted that inclusion of variable % of Hartree Fock exchange stabilizes the peroxide species in PS II OEC.37, 38, 39 In the O-O bond formation the terminal acetate acts as a proton acceptor group. A barrier of ~ +0.9 eV is estimated for the computed WNA process (Movie S1, see Supporting Information for details, Fig. S7). Additional driving force can be provided by proton removal at pH=12 and oxidation of formed CoIII3O4CoIII-OOH to CoIII3O4CoIV(OOH). This should require a moderate potential (Table 1) while shifting an equilibrium towards a peroxo form.
Direct involvement of oxygen-containing ligands such as N-oxides and carboxylates into O-O bond formation was proposed for some water oxidation catalysts.40, 41 Currently, experimental confirmation of the direct carboxylate ligand involvement in O-O bond formation is lacking. Nevertheless, for complete analysis of all possible pathways, S[CoIII3O4CoIII(OOC(O)CH3)bridging]0 product of O-O bond formation was computed (Table 1). Energetically, this pathway is similar in energy to WNA but was not further investigated pending an experimental evidence.
DFT-computed spectroscopic properties of proposed intermediates are summarized in Table 2. The O-OH vibration in S[CoIII2O4CoIII-OOH/CoIII(CH3CO2H)mono]0 peroxide is ~762 cm-1 and is insensitive to O-OH to O-OD exchange (Table 2). Thus, this type of end-on peroxide also agrees with reported FT-IR29. O-OH vibration (~807 cm-1) of the oxidized [CoIII2O4CoIV-OOH/CoIII(CH3CO2H)mono]+ form is highly plausible under applied potential and better agrees with the FT-IR experiment. Deprotonation of carboxylate ligand, plausible at pH=12 (Table S8), shifts computed O-O vibration further to ~875 cm-1 (Table 2). Both [CoIII2O4CoIV-OOH/CoIII(CH3CO2H)mono]0 and [CoIII2O4CoIV-OOH/CoIII(CH3CO2)mono]+ result in O-O vibrational frequencies very close (taking into account ~50 cm-1 uncertainty of calculations) to experimentally reported ~833 cm-1 frequency29.
DFT-computed 59Co hyperfine splitting of both [CoIII2O4CoIV-OOH/CoIII(CH3CO2H)mono]+ and [CoIII2O4CoIV-OOH/CoIII(CH3CO2)mono]0 species is in agreement with EPR results (Table 2). Considering pH and applied potential, we conclude that [CoIII2O4CoIV-OOH/CoIII(CH3CO2)mono]0 is likely a peroxo intermediate with spectroscopic properties matching FT-IR and EPR. DFT predicts weak driving force for its deprotonation forming [CoIII2O4CoIV-OO/CoIII(CH3CO2)mono]- (Table S8). [CoIII2O4CoIV-OO/CoIII(CH3CO2)mono]- is predicted to have smaller 59Co hfs on the order of 15-22 G (Table 2) due to larger spin density localization on –OO superoxo fragment. This is an end-on peroxo intermediate while the attempts to produce side-on peroxide lead towards end-on state.
We computed bridging peroxide (CoIII2O4CoIII-OO-CoIII) that was proposed to explain FT-IR results29 but obtained ~718 cm-1. This O-O bond frequency significantly deviates from the experimental ~833 cm-1 value. The O-O vibration in oxidized bridging peroxo- intermediate CoIII2O4CoIV-OO-CoIII was computed to be ~905 cm-1. Generation of CoIII2O4CoIV=O/CoIV=O (Fig. 1c) and its precursor of CoIII2O4CoIII-OO-CoIII bridging peroxide, carries additional ~+0.6 eV per removed proton penalty when compared with the generation of CoIII2O4CoIV=O/CoIV(CH3CO2)mono.
(Table 1). We did not computationally investigate pathway in Fig. 1b since it requires de-coordination of two ace-tate ligands from the same Co center. We suggest that statistically it is less probable than de-coordination of just one acetate ligand.
A strong motivation to study the Co4O4 cubane was its resemblance to the CaMn4O5 cluster of PS II photosynthetic enzyme.14, 42 In 2015, we proposed MnIV=O as an intermediate of the S3-state of the OEC able to satisfy spectroscopic properties and be positioned to form O-O bond early in the S3 to S0 transition (Fig. 4a). 37, 43, 44 Time-resolved XRD analysis of the S3 state45, 46, 47 resulted in data that might be interpreted as a formation of at least three possible structures: MnIV-OH, Mn-OO-Mn, or MnIV=O, with the last one favored in the most recent study45. Interestingly, [CoIII2O4CoIV=O/CoIV(CH3CO2)mono]0 displays strong radicaloid character for the oxygen in the CoIV=O group (see spin density, Fig. 1d). Similar radicaloid character of the RuV=O fragment oxygen was established via direct spin density mapping by EPR for water oxidation intermediate in the “blue dimer” catalyst.48 In the PS II, the O-O bond was proposed to form via radical coupling between the oxygens in MnIV=O and bridging Mn-O-Mn.37, 49 Mechanism observed for the [CoIII2O4CoIV=O/CoIV(CH3CO2)mono]0 with involvement of the carboxylate and bridging oxygen as proton acceptors in the O-O bond formation (Fig. 1d), prompted analysis of similar configurations for the OEC (Fig. 4). The S3 state model allows WNA of the Ca2+-bound H2O on the MnIV=O. The closest carboxylate ligand able to accept a proton is E189, while two oxo bridges, the left (~+0.2 eV) and the right one (~+1.5 eV), of the MnIV=O fragment may form hydrogen bonds with the peroxides formed (Fig. 4b). Such peroxides are energetically plausible for formation, even at S3 state. Note that formation of a peroxide via radical coupling will better agree with the TR-XRD data of the S3 state, as this does not require additional oxygen atoms except already detected in TR-XRD.45, 46, 47 To account for peroxides formed via WNA on the MnIV=O, additional oxygen density should be present in the S3 state (Fig. 4a). Thus, such a peroxide can only be possible as a minor configuration. One electron oxidation of the OEC forming the S4 state results in similar energetics with (~+1.8 eV) for the left and (~+0.2 eV) for the right (Fig. 4b, Table S9). Future detection of electron densities in the S4 state (formed ~200-1000 msec in the S3 to S0 transition) will help to discriminate between radical coupling and WNA mechanisms.