17O-EPR determination of the structure and dynamics of copper single-metal sites in zeolites

The bonding of copper ions to lattice oxygens dictates the activity and selectivity of copper exchanged zeolites. By 17O isotopic labelling of the zeolite framework, in conjunction with advanced EPR methodologies and DFT modelling, we determine the local structure of single site CuII species, we quantify the covalency of the metal-framework bond and we assess how this scenario is modified by the presence of solvating H216O or H217O molecules. This enables to follow the migration of CuII species as a function of hydration conditions, providing evidence for a reversible transfer pathway within the zeolite cage as a function of the water pressure. The results presented in this paper establish 17O EPR as a versatile tool for characterizing metal-oxide interactions in open-shell systems.

Supplementary Figure 1. X-band CW-EPR spectra of Cu II in CHA. a Experimental (black) and simulated (red) X-band CW-EPR spectra recorded at 77 K of fully hydrated and dehydrated Cu-CHA. The contribution of each individual species (A -D in Table 1) is shown in blue, green, gold and violet for A, B, C and D, respectively. b Experimental (black) X-band CW-EPR spectra recorded at room temperature at different dehydration stages. The simulations of the spectra of the fully hydrated and dehydrated samples are shown in red. The spectra of the dehydrated sample at RT and 77 K are virtually identical. The simulation of the RT EPR spectrum of the hydrated sample was performed by including a motionally averaged component obtained by imposing a rotational correlation time τ=10 -11 s to the spin-Hamiltonian parameters determined from the rigid-limit spectrum ( Figure S1a). c Relative intensity of Cu II signal (%) as determined by double integration of the first derivative spectra. The intensity of the fully hydrated sample was set to 100 %. d X-band CW-EPR spectra recorded at room temperature of fully hydrated with H2 17 O Cu-CHA and dehydrated at increasing temperatures. The line broadening with respect to the corresponding spectra recorded on the H2 16 O sample is due to the presence of the 17 O isotope (I=5/2). degrees. In the simulations (in red) the 27 Al A-tensor and its relative orientation with respect to gtensor employed are the same as reported in Table 2.

Supplementary
Supplementary Figure 3. a Simulation (in red) of the 1 H HYSCORE spectrum (in black) of the fully hydrated Cu-CHA sample. b Simulation (in red) of the 27 Al HYSCORE spectrum (in black) of the fully dehydrated Cu-CHA sample. Due to the fact that it was impossible to record 27 Al HYSCORE spectrum at the g// position because of the low S/N ratio, it is not possible to obtain a univocal set of parameters. A representative simulation obtained using the spin-Hamiltonian parameters reported in Table 2 is shown which qualitatively reproduces the experimental features. The parameters are consistent with data reported for other Cu-doped zeolites and with the computed values. whereas the blue one corresponds to Cu-CHA sample hydrated with normal water after the isotopic enrichment. The corresponding pictorial representation of the Cu site is reported on the right of each spectrum. The similarity of the two spectra demonstrates that Cu II retains a direct linkage with the framework under hydrating conditions and that the water and framework oxygen equatorial ligands display a similar degree of spin density, in accord with the computed data (see Oe1 and Oe3 in Table  S6). 64 scans were averaged for the 16  The simulated spectra were obtained by using the computed 17 O hyperfine coupling tensors for the three different Al distributions considered in the main text (1Al in brown, 2Al-2NN in blue and 2Al-3NN in red). All four coordinating oxygen donor atoms were considered. We remark that the data refer to geometrically optimized structures at 0 K, while averaged parameters are determined in the experiment.  Figure 6. The labelling of the oxygen atoms refers to the one indicated in Figure 6. Cu-O bond distances are given in nm.  The relative stability of both the sites is evaluated in terms of relative electronic energy per unit cell. DFT computations point out that, in presence of more than one water molecules in the copper coordination sphere, 8MR sites appears to be more stable with respect to 6MR location (Supplementary Figure 8). This difference between the two sites is likely due to the lower coordination number of Cu II cations in 8MR in comparison to 6MR sites when no adsorbates are present. Moreover, the adsorption of water molecules provokes significant change in the local geometry of the copper ions (from distorted square planar to a distorted square pyramidal geometries). Note that the third water molecules in 6MR site does not interact directly with Cu but it is linked to its first coordination sphere through hydrogen bonds. The increasing of the water molecules bound to Cu II leads copper ion shifted upward, above the position of 6MR unit. Such findings are in agreement with the work of Kerkeni (see Reference 71 of the main text) and prove how the location and geometry of copper ions in CHA framework are drastically affected by the presence of adsorbates with oxygen donor atoms like water molecules.

Supplementary
Supplementary Figure 9.   Table  6, evince an exquisite agreement with experimentally obtained tensors. For the sake of clarity, we reported the mean values of aiso and T components of the equatorial and axial protons. While the computed aiso term for axial water protons is always set around 0.1-0.4 MHz, the isotropic hyperfine coupling constants of the equatorial water molecules depend on the orientation of the water molecules with respect to the equatorial plane, as previously discussed by Larsen. 1 Therefore a realistic comparison of the experimental and computed isotropic hfi is difficult to assess because the actual orientation of water ligands is affected by remote water molecules as well as the zeolite framework.      Figure S10 at B3LYP-D3(ABC) level of theory. Hyperfine coupling constants and quadrupole coupling constants are given in MHz.

Supplementary Note 1
From the observation of the hyperfine structure and determination of the isotropic (aiso) and dipolar (T) hyperfine couplings, the electronic spin distribution in a molecular fragment can be obtained. To do so it is important that the hfi interaction with several (preferably all!) nuclei in the molecule are observed. Then, with the knowledge of aiso and T for the atomic species, and assuming that the hfi interaction at a given nucleus is proportional to the electron spin density at that nucleus, one can obtain the spin population in s-type orbitals s, p-type orbitals p.  (4) in Table 3 of the main text). The total spin density on the oxygen ligands obtained as the sum of the two s and p contributions is reported in the main text. These values nicely agree with the Mulliken spin population reported in Supplementary Tables 1 and 2.

Supplementary Note 2
The precise and robust calculation of g-and A-tensors for mononuclear Cu II systems is still a great challenge for quantum chemistry methods. 2,3 While for organic radicals excellent results for g-tensor and hyperfine couplings are already obtained with GGA functionals, for transition metal ions the situation is totally different. On one hand, g-shifts are usually underestimated by standard DFT methods because of a too covalent description of the metal-ligands bonds and a overestimation of the d-d transition energies. 4 On the other hand, the problems in the prediction of hyperfine couplings are related to the spin-orbit coupling (SOC) component (negligible for lighter elements) and to the Fermi use as guessing point Cu complex with octahedral geometry since g-tensor and Cu hyperfine couplings measured for the hydrated state are consistent with such coordination. However, the full optimized structures of (a) and (b) assume a distorted tetragonal and square pyramidal geometry, respectively. On the contrary, model (c) keeps a distorted octahedral geometry. In any case, the Cu-O bond related to the hydroxyl group is slightly shorter with respect to the Cu-O bonds involving water molecules. This is fundamentally due to the stronger negative charge of the oxygen atom in OHgroup with respect to water ligands. The computed hyperfine 1 H and 17 O couplings of the hydroxyl group are completely out of the range of the experimental values (see Supplementary Table  6 and Tables 2 and 3), especially the dipolar part. These findings confirm the proposed assignment for the experimental 1 H and 17 O signals in HYSCORE and ENDOR spectra. Moreover, we also prove that the combination of hyperfine techniques with computational modelling can be exploited to distinguish hydrated [CuOH] + species from aqua Cu II complexes in hydrated copper-exchanged zeolites.