(a) Cu and O2 fragment oxidation states
Among the various experimental data that may be used to understand the electronic structure of the 1:1 Cu/O2 adducts, the O-O stretching frequency (ν(O-O)) and the edge features in X-ray absorption spectra (XAS)13,15 are especially informative for ascertaining the O-O bond order and the Cu oxidation state, respectively (Table 1). The ν(O-O) values for 1a,b16 and 217 are similar to those usual for superoxide complexes (∼1075-1295 cm-1),11,18 whereas the values for 3a,b7,19 and 49 are significantly lower, albeit above the region typical for peroxides (∼750-930 cm-1). Such intermediate values have been noted previously for metal/O2 adducts,11,18 and the existence of a “more or less continuous range of values”11c that span 700-1300 cm-1 for such species has been noted.20 To better place the ν(O-O) data for 1-4 into perspective, we previously presented a correlation between these values and the associated O-O distances that was applicable to a range of side-on metal/O2 adducts characterized both via theory and experiment, and included simple oxygen species such as O2, O2-, and O22-.12 This correlation is better expressed by recourse to Badger’s rule (eq. 1),21 an
empirical relation between an equilibrium internuclear distance (re) and the associated stretching frequency (ν) that has been applied with success recently in analyses of Fe-O and S-S bonding.22,23 Here we apply it to a greater range of compounds than analyzed in ref. 12, yet limited to side-on metal/O2 adducts excepting 2 and some simple oxygen species uncoordinated to any metal. The data (listed in supporting information Table S1) are plotted as O-O distance vs. 1/ν2/3 in Figure 3, and are in good agreement with a linear fit to eq. 1 giving C = 70.7 and d = 0.671 with R2 = 0.96. This fit excludes the experimental data point for 1a, which is deemed an outlier due to its unreasonably short O-O distance equivalent to that of free O2.
In further studies aimed at defining the electronic structures of the similarly side-on bound adducts 1b and 3a, a direct experimental comparison of their Cu oxidation states was accomplished through X-ray absorption spectroscopy (Table 1).13 Both the K- and L3-pre-edges for 1b were ∼2 eV lower than those for 3a, supporting their respective assignments as Cu(II) and Cu(III) complexes. Accentuating this point, the pre-edge 1s → 3d energy for 1b falls into the range observed for a set of bona fide Cu(II) complexes (8978.8 × 0.4 eV),15,24 whereas that for 3a is similar to those of other Cu(III) compounds (8981 ± 0.5 eV; cf. data for [(Me3tacn)2Cu2(μ-OH)2]2+ and [(Me3tacn)2Cu2(μ-O)2]2+ in Table 1).15 Similar comparisons of L3-edges (2p→ 3d transitions) are even more striking due to the high resolution of L-edge data (Figure 4).25 In general, differences in XAS edge energies may be traced to different charges at the absorbing atom (Q) and/or ligand fields (LF).26 On the basis of calculations and correlations to XPS data, LF contributions were deemed to dominate the XAS edge energy disparities between 1b and 3a, with the Q for the Cu(III) site in 3a being essentially the same as the Cu(II) center in 1b because of compensation by the strongly electron donating β-diketiminate ligand.13
Computational studies on Cu- and other metal-O2 complexes have been particularly useful in defining geometrical and electronic structural features having a bearing on oxidation state. As depicted in Figure 3, the theoretical data (all of which derive from density functional calculations with the mPWPW91 functional27 and basis sets of polarized double- to triple-ζ quality) follow Badger’s rule particularly well and are in excellent agreement with experiment in most instances. In cases of disagreement, challenges associated with librational motion and structural disorder may render the experimental data less reliable than the theoretical (cf. 1a, as noted above).12 The theoretical and experimental data taken together comprise a collection of bond lengths and vibrational frequencies that smoothly span the range from superoxide-like to peroxide-like, and this smooth progression suggests that the assignment of standard “integer” metal and O2-fragment oxidation states is not necessarily a straightforward procedure—covalent character in the metal-O2 bonding leads to species that may be regarded as valence-bond hybrids of the limiting superoxide and peroxide extremes.
The covalent communication between the metal and O2 fragment and its impact on apparent oxidation state has been analyzed in detail for a simplified model of 3 as illustrated in Figure 5.28 The 11b2 and 12b2 orbitals are bonding and antibonding combinations of the in-plane O2 π* orbital and the Cu dxz orbital (taking the z axis as the C2v symmetry axis and the x axis as parallel to the O-O bond vector). In the singlet state, the 3a2 out-of-plane O2 π* orbital is empty (it is singly occupied in the higher energy triplet state), and multiconfigurational treatments predict that the singlet ground state is well represented by
where c1 and c2 are configuration weights that indicate the relative importance of each determinant and ensure normalization. When the diketiminate ligand is rendered poorly electron donating by fractionally increasing the atomic charge of the nitrogen atoms (or by changing a backbone carbon nucleus to nitrogen; data not shown), c1 and c2 are found to be roughly equal in magnitude such that the 11b2 and 12b2 orbitals may be considered to be similarly populated. Algebraically, this situation is equivalent to a 2-electron in 2-orbital open-shell singlet, where the 2 orbitals are formed from positive and negative linear combinations of 11b2 and 12b2. In this case, those linear combinations correspond to an isolated Cu dxz orbital and an isolated O2 π* orbital, which would be the ionic valence bond picture associated with a classic Cu(II) superoxide species. Consistent with this analysis, the O2 bond length is predicted to be in the shorter range typical of a superoxide. As the electron donating character of the ligand is increased by fractionally decreasing the nitrogen nuclear charge (or changing a backbone carbon nucleus to boron; data not shown), the opposite effect is observed. The ratio of c1 to c2 becomes large and the 11b2 orbital simultaneously localizes more heavily on the O2 fragment. The resulting closed-shell wave function corresponds to a Cu(III)-peroxide species and consistent with this picture the O2 bond length is predicted to be correspondingly long. Variation between these two extremes proceeds smoothly with variation in nitrogen nuclear charge (Figure 5), illustrating the degree to which covalent communication mediates charge flow and the ligand influences the Cu-O2 interaction. It is noteworthy that the distance between the Cu and O2 fragment is not particularly sensitive to variation in the ligand’s electron donating capabilities, in contrast to a prior suggestion.29