Structure and Reactivity of Oxygen-Bridged Diamino Dicopper(II) Complexes in Cu-Ion-Exchanged Chabazite Catalyst for NH3-Mediated Selective Catalytic Reduction

The NH3-mediated selective catalytic reduction (NH3-SCR) of NOx over Cu-ion-exchanged chabazite (Cu-CHA) catalysts is the basis of the technology for abatement of NOx from diesel vehicles. A crucial step in this reaction is the activation of oxygen. Under conditions for low-temperature NH3-SCR, oxygen only reacts with CuI ions, which are present as mobile CuI diamine complexes [CuI(NH3)2]+. To determine the structure and reactivity of the species formed by oxidation of these CuI diamine complexes with oxygen at 200 °C, we have followed this reaction, using a Cu-CHA catalyst with a Si/Al ratio of 15 and 2.6 wt% Cu, by X-ray absorption spectroscopies (XANES and EXAFS) and diffuse reflectance UV-Vis spectroscopy, with the support of DFT calculations and advanced EXAFS wavelet transform analysis. The results provide unprecedented direct evidence for the formation of a [Cu2(NH3)4O2]2+ mobile complex with a side-on μ-η2,η2-peroxo diamino dicopper(II) structure, accounting for 80–90% of the total Cu content. These [Cu2(NH3)4O2]2+ are completely reduced to [CuI(NH3)2]+ at 200 °C in a mixture of NO and NH3. Some N2 is formed as well, which suggests the role of the dimeric complexes in the low-temperature NH3-SCR reaction. The reaction of [Cu2(NH3)4O2]2+ complexes with NH3 leads to a partial reduction of the Cu without any formation of N2. The reaction with NO results in an almost complete reduction to CuI, under the formation of N2. This indicates that the low-temperature NH3-SCR reaction proceeds via a reaction of these complexes with NO.


INTRODUCTION
The selective catalytic reduction of NO x by ammonia (NH 3 -SCR) to nitrogen and water is the basis for the current technology for NO x abatement in the exhaust of lean-burn heavy-duty and passenger vehicles. This technology has already resulted in significant improvements of exhaust gas emissions from diesel vehicles. Catalysts based on Cu-ion exchanged chabazite (Cu-CHA) are very effective for this reaction and are commonly applied today. These catalysts feature a high activity around 200°C, a good selectivity for N 2 formation, and excellent thermal stability in the harsh conditions of exhaust after-treatment systems. 1,2 The NH 3 -SCR reaction is a redox reaction following the equation 4NO + 4NH 3 + O 2 → 4N 2 + 6H 2 O. The NH 3 -SCR activity of Cu-CHA catalysts is due to the capability of the Cu ions to reversibly change the oxidation state between Cu I and Cu II . 3−5 In the NH 3 -SCR reaction cycle, Cu II is reduced to Cu I , followed by a reoxidation of the Cu I to restore the Cu II . The reaction cycle can be performed stepwise, by alternating a reduction in a mixture of NO and NH 3 , and an oxidation in a mixture of NO and O 2 . 4,6−9 For the reduction half-cycle, there is converging evidence that the reduction of Cu II by NO and NH 3 at around 200°C results in the formation of linear [Cu I (NH 3 ) 2 ] + complexes. These complexes are weakly bound to the zeolite, and therefore mobile. 3,6,10,11 In the oxidation half cycle, the Cu I species reacts with O 2 to form a Cu II species, and at low temperatures, O 2 exclusively reacts with the Cu I species. Therefore, the reaction of O 2 with the linear [Cu I (NH 3 ) 2 ] + complexes is an essential step in the NH 3 -SCR reaction cycle at low temperature. 12 To complete the activation and dissociation of the O 2 molecule, four electrons are required. As a single Cu I is capable of delivering only one electron (no evidence for Cu III formation has ever been reported in the numerous studies on NH 3 -SCR), this means that other electron sources are required, which can be other Cu I ions, NO or other reaction intermediates. [4][5][6]13,14 Following these thoughts, it has been shown that the dissociation of O 2 becomes easier when a single O 2 molecule interacts with two Cu I ions simultaneously to form Cupairs. 3,4,7,15,16 Combining this with the mobility of the linear [Cu I (NH 3 ) 2 ] + complexes, a reaction mechanism has been worked out where Cu-pair formation is facilitated by diffusion of the [Cu I (NH 3 ) 2 ] + complexes inside the zeolite. 4,13 Such a mechanism involving the formation of Cu-pairs is supported by the observation that the NH 3 -SCR rate at low temperature is, for low Cu contents, proportional to the square of the Cu content in the catalyst. 4,12,17,18 At higher temperatures, the [Cu I (NH 3 ) 2 ] + complexes decompose, 10,19 and the Cu I is then expected to lose its mobility. As a result, the formation of Cupairs, and therefore also the activation of O 2 , becomes more difficult. This seems to be the reason for the often observed decrease in NO x conversion with increasing temperatures around 300°C, which separates the low-and high-temperature regimes for Cu-CHA catalysts. 4 At high temperatures, the reaction may occur on isolated ZCu I sites (where Z indicates coordination to zeolite oxygens in the proximity of an Al exchange site), possibly mediated by the formation of Cunitrate and Cu-nitrite species, 2,6,15,16,20 which then further reacts with ammonia to yield N 2 and H 2 O.
In a model where the formation of Cu-pairs is facilitated by diffusion of [Cu I (NH 3 ) 2 ] + complexes, the actual active center for the activation O 2 is not directly associated with a specific site or location of the Cu in the zeolite. In a fresh Cu-CHA material, the positive Cu ions balance the negative charges in the zeolite framework induced by the Al substitution; a Cu II ion is anchored by either a single framework Al atom, as a Z[Cu II (OH)] species, or by two framework Al atoms, to yield a Z 2 Cu II species. These Cu species are then located either in a double 6-membered ring or an 8-membered ring. Upon exposure to a reaction gas for NH 3 -SCR, which contains NO, NH 3 , O 2 and H 2 O, these Z[Cu II (OH)] and Z 2 Cu II species become solvated by NH 3 leading to the formation of the mobile [Cu I (NH 3 ) 2 ] + complexes. These complexes are able to diffuse to about 9 Å away from their anchor point in the time scale of catalytic turnover, which enables the Cu-pair formation necessary for the activation of O 2 . 4,17 This means that the original location of the Cu ions does not immediately affect the reactivity of the Cu, but the different local environments of the Cu ions may affect the solvation of the Cu by ammonia. Furthermore, Density Functional Theory (DFT) calculations have shown that the formation of Cu pairs from two [Cu I (NH 3 ) 2 ] + complexes becomes more difficult in areas with two Al atoms located close to each other. 21 The structures that are formed after the reaction of an O 2 molecule and two [Cu I (NH 3 ) 2 ] + complexes are Cu complexes containing two Cu-centers bridged by oxygen; the general formula of these complexes is [Cu 2 (NH 3 ) 4 O 2 ] 2+ . Efforts to determine the structure of this complex have mainly been based on DFT calculations. Figure 1 shows three possible structures of the [Cu 2 (NH 3 ) 4 O 2 ] 2+ complexes, which differ in the way the oxygen molecule is bound to the Cu, and whether dissociation of the O−O bond takes place or not. The stability of these structures calculated with DFT depends on the functional chosen in the calculation. 22 Calculations on ammonia-ligated Cu 2 O 2 cores in the gas phase have shown how the predicted stability of different structures depends on the method (DFT vs post-Hartree−Fock), due to the differences in the description of the electron correlation contribution to the core conformation. 23,24 More recent DFT calculations on O 2 activation and dissociation by [Cu I (NH 3 ) 2 ] + complexes in CHA also showed a strong dependence of the calculated stabilities and structural parameters on the selected functional. DFT calculations using a PBE functional, with or without van der Waals corrections, often result in the bis-μ-oxo diamino dicopper(III) complex 17,22 (Figure 1c), which implies that the reaction of O 2 with [Cu I (NH 3 ) 2 ] + complexes results in the dissociation of the O−O bond. With a HSE06 hybrid functional, one finds a μ-   22 Including a Hubbard-U term of 6 eV in DFT (DFT+U) with a PBE functional, and van der Waals corrections results in a correct prediction of the dissociation of the O−O bond in Cu 2 O 2 complexes in enzymes with a structure similar to those expected in the Cu-CHA zeolite. 22 Calculations of the [Cu 2 (NH 3 ) 4 O 2 ] 2+ complex with that method pointed to the formation of the μ-η 2 ,η 2 -peroxo diamino dicopper(II) complex (Figure 1b), 22 17 The concept that dissociation of O 2 requires two solvated Cu I ions also opens the question whether Cu pairs play a role beyond O 2 activation. Therefore, we also investigate the reactivity of the [Cu 2 (NH 3 ) 4 O 2 ] 2+ complexes toward NO and NH 3 . These goals have been pursued by applying X-ray absorption spectroscopy (XAS), both near-edge (XANES) and extended range (EXAFS), with a well-established operando setup, 6,10,25 and diffuse reflectance ultraviolet−visible−nearinfrared (DR UV-Vis-NIR) spectroscopy 19,26 using a Cu-CHA catalyst. DFT calculations were used for optimization of the [Cu 2 (NH 3 ) 4 O 2 ] 2+ structures illustrated in Figure 1, which were used as input for the interpretation of the EXAFS data. To enhance the sensitivity of EXAFS to multicopper moieties, we applied wavelet transform (WT) analysis, giving unprecedented insights in the formation and separation of the solvated Cu-pairs in the NH 3 -SCR reaction cycle over Cu-CHA catalysts.

EXPERIMENTAL SECTION
The catalyst used in this study was a Cu-CHA material with Si/Al = 15 with a Cu content of 2.6 wt%, corresponding to Cu/Al = 0.5, a Cu density of around 0.4 Cu/1000 Å 3 (0.3 Cu per chabazite cage) and a theoretical mean Cu−Cu distance of 16.6 Å. At this Cu density, about 90% of Cu I ions can be oxidized by O 2 , forming the [Cu 2 (NH 3 ) 4 O 2 ] 2+ complexes. 17 The steps depicted in Scheme 1 have been used to form these complexes and to study the reactivity toward NO and NH 3 experimentally. These steps are as follow: The steps in this protocol were followed independently by XAS coupled to an online mass spectrometer for a qualitative effluent gas analysis (BM23 beamline of the European Synchrotron Radiation Facility), 28 and by DR UV-Vis-NIR. In this Article, we present the development of spectra with time under isothermal conditions, until a steady-state was observed. The DR UV-Vis-NIR spectra are reported as relative reflectance (R%), to avoid artifacts due to the use of the Kubelka−Munk function. 26 We refer to the Supporting Information (SI), section 1, for more experimental details and description of the equipment used.   This evidence points to the presence of different, NH 3solvated Cu II species with respect to the frameworkcoordinated Cu II ions (fw-Cu II ) known to be present in the catalyst after pretreatment in O 2 . 7,25,29−31 To strengthen such indications, we initially analyzed by linear combination fit (LCF) the XANES after oxidation in O 2 (step 3) using the spectra obtained at step 1 (pretreated in O 2 ) and at step 2 (reduced in NO/NH 3 ) as references for Cu II and Cu I components, respectively (see SI, Figure S11). Overall, estimates for Cu I and Cu II percentages are in reasonable agreement with qualitative analysis. It is also evident that the rising edge peak at 8982.5 eV diagnostic of Cu I species is excellently reproduced, indicating that the Cu I component, i.e., linear [Cu I (NH 3 ) 2 ] + , is the same at both steps 2 and 3. However, significant discrepancies between experimental and LCF curve are found when considering the rising-edge region where Cu II 1s→4p transitions typically occurs, as well as the shape and energy position of the white-line peak. These discrepancies translate into a structured residual function, with well-defined maxima and minima well above the noise level, further indicating that two spectroscopically distinguishable Cu II species are present at steps 1 and 3.
In the UV-Vis-NIR spectra, the oxidation of the linear [Cu I (NH 3 ) 2 ] + complex is visible as follows. Due to the d 10 closed-shell configuration of Cu I , the typical ligand-field d-d transitions are absent, and the spectrum is dominated by a ligand-to-metal charge transfer (LMCT) transition observed in the range 30 000−45 000 cm −1 (blue curve in Figure 2c). 26,32 The oxidation of Cu I to Cu II by O 2 is reflected in the development of an intense d-d absorption centered at 13 850 cm −1 and a red-shift of the LMCT transitions, from ca. 35 000 to 25 000 cm −1 (arbitrarily measured at R = 60%). The features at 6515 and 4970 cm −1 in the NIR region are due to the overtones and combination modes of NH 3 and NH 4 + , confirming that the Cu II species after oxidation still contains the NH 3 ligands. The spectrum of the Cu-CHA catalyst pretreated in O 2 (dark gray dashed curve) is reported for comparison, showing that the coordination geometry of the Cu II species formed by oxidation of [Cu I (NH 3 ) 2 ] + is different with respect to that of the variety of framework-coordinated monomeric/multimeric ions (fw-Cu II , such as Z[Cu II (OH)]/ Z[Cu II (OO*)] etc.) responsible for the typical Cu-CHA "quadruplet" (complex absorption in the d-d region with components at 20 000, 16 350, 13 300, and 10 600 (sh) cm −1 ). 26 Figure 3a shows these data without phase correction. The firstshell peak is almost doubled in intensity after O 2 interaction (from blue to red), indicating that the coordination number of Cu in the Cu II complexes is higher than in the linear [Cu I (NH 3 ) 2 ] + complex. In the next shell, the unstructured feature observed for the mobile [Cu I (NH 3 ) 2 ] + complex evolves toward a broad scattering feature peaking at ca. 2.4 Å. Even though this feature is close to the second-shell EXAFS signature of framework-coordinated Cu II ions after pretreatment in O 2 (dashed gray line), it can be clearly distinguished, indicating that, at least a substantial part of the Cu II complex is still mobile in the cage. Finally, a third peak around 3.2 Å develops, which could correspond to contributions from a Cu−Cu scattering in the Cu II complexes shown in Figure 1. All these features in the FT-EXAFS are consistent with the formation of [Cu 2 (NH 3 ) 4 O 2 ] 2+ complexes. However, a more detailed analysis is required to rule out re-coordination of Cu II ions to zeolite framework and, once this possibility is excluded, to identify the precise [Cu 2 (NH 3 ) 4 O 2 ] 2+ structure, based on the three possibilities shown in Figure 1.
To this aim, we have first performed quantitative EXAFS analysis of the spectra obtained at step 1 (gray dashed line in Figure 3). The adopted fitting model is based on structural characteristics conserved for most fw-Cu II species previously proposed to form in Cu-CHA upon pretreatment in O 2 . 7,25,29−31,34 Importantly, it accounts for a distinctive scattering contribution of charge-balancing Al atoms (Al fw ) located at ca. 2.75 Å range from the Cu II center, while maintaining a certain degree of flexibility to account for fractional contribution from multicopper species (additional information can be found in SI, section 2.1). As detailed in the SI, the fit resulted in a good level of reproduction of experimental EXAFS spectrum for the pretreated catalyst at step 1, with physically meaningful values for all the refined parameters, as well as for their fitting errors (see SI, Figure S2 and Table S2).
A second stage of our analysis consisted in a test EXAFS fit of the experimental spectrum of the [Cu(NH 3 ) 2 ] + complexes oxidized in O 2 (step 3, red solid line in Figure 3), using exactly the same model based on fw-Cu II , which guaranteed a successful fit for the EXAFS of the pretreated catalyst (see SI, section 2.3). While the numerical agreement between best-fit and experimental curve was formally satisfactory, this was achieved at the expense of the physical meaning of the optimized parameters (e.g., Debye−Waller factors as high as 0.1 Å 2 ) as well as of their accuracy (unphysically high fitting errors). Such inconsistencies most severely affect the Cu−Al fw coordination shell, representing a diagnostic contribution for the large majority of fw-coordinated Cu II species proposed so far in the CHA framework. Consistently with the LCF XANES results in SI, Figure S11, the failure of this test EXAFS fit strongly supports the spectroscopically detectable diversity of Cu II species formed in Cu-CHA at steps 1 and 3. The state of the catalyst at step 3 is clearly not consistent with fwcoordinated Cu II species, thus paving the way to deeper structural analysis considering mobile [Cu II 2 (NH 3 ) 4 O 2 ] 2+ complexes.
To determine the precise structure of the [Cu 2 (NH 3 ) 4 O 2 ] 2+ complexes, first we have optimized the structures in the gas phase (that is not including the zeolite effect in the calculations) for the three complexes shown in Figure 1 by DFT, including the three main structures proposed in the literature by different authors: NH 3 -solvated side-on 22 and end-on 4 peroxides as well as the bis(μ-oxo)-dicopper core. 17 Then, these structures were used as input to fit the observed EXAFS features. A detailed description of the DFT and fitting procedures and results can be found in the SI, section 2.4. Note that, due to the difficulty in determining the relative stabilities of the [Cu 2 (NH 3 ) 4  Fitting the EXAFS data with the (end-on) trans μ-1,2peroxo diamino dicopper(II) or bis-μ-oxo diamino dicopper-(III) models (Figure 1a 4,17 A conclusive assignment of the features observed in conventional Fourier transform (FT) EXAFS spectra, however, is hindered by the simultaneous presence of various types of atomic neighbors surrounding the Cu absorber, especially in the high-R region. If two or more types of elemental neighbors and/or scattering interactions are localized at close distances around the absorber, their contributions in the direct space R overlap and often become indistinguishable. For the Cu-CHA zeolite studied here, these potentially include single scattering paths from framework Al/Si/O in zeolite-coordinated Cu moieties as well as Cu in multicopper species, which can be coordinated to the zeolite or mobile. 29 The intense multiple scattering paths involving first-shell O/N neighbors in the proposed μ-η 2 ,η 2 - Journal of the American Chemical Society pubs.acs.org/JACS Article peroxo diamino dicopper(II) moiety mentioned above are also expected to fall in this R-space range.
To resolve this, it is possible to exploit the fact that the contributions from different elemental neighbors appear at different locations in k-space, because the backscattering amplitude factor F(k) strongly depends on the atomic number Z. Figure 4a shows the F(k) curves associated with the elements relevant in this work, namely O, N, Al, Si, and Cu. It is clear that signals produced by heavier atoms, such as Cu, are localized at higher k values with respect to lighter atoms. On this background, a WT analysis allows for a better discrimination of the nature of the scattering contributions around the absorber, compared to the classical FT analysis. 30,43−47 The WT analysis results in a 2D representation of the EXAFS, simultaneously revealing the signal features in both R-and k-space. Then, one can visually resolve the scattering contributions originating from atomic neighbors having enough Z-contrast in their F(k) functions. A more detailed description of the WT analysis technique is given in the SI, section 3.1. In the relevant R-space range, EXAFS WT for the spectrum collected in O 2 at 200°C after pretreatment in O 2 at 400°C (Figure 4b), clearly splits in two lobes. The first sub-lobe, localized in the k range 1−5 Å −1 and R range 2−2.8 Å is associated with the framework atoms: O, Si and Al. The second one, localized at higher k values (i.e., 6−8 Å −1 ), is principally related to Cu−Cu contributions in oxygen-bridged Cu dimers or, more in general, multicopper moieties. WT analysis further validate the previously mentioned EXAFS fitting results for step 1 based on a prototypical fw-Cu II model (see SI, section 2.1), confirming the simultaneous presence of Cu−Al, Cu−O/Si, and Cu−Cu scattering contributions in the high-R EXAFS range for the pretreated catalyst. The presence of Cu-oxo dimeric/polymeric cores in oxygen activated Cu-CHA catalysts represents a novelty with respect to previous literature in the context of the NH 3 -SCR reaction. 25 This aspect has been recently established by different authors studying the nature of Cu-oxo species in Cu-CHA for the direct methane to methanol conversion. 26,29,30,34,35,43,47 While from the EXAFS fit in SI, section 2.1 an average Cu−Cu coordination number N Cu = 0.5 ± 0.3 is estimated, a more precise identification of the nature and amount of these dimeric/polymeric structures in the oxygen activated Cu-CHA is outside the scope of this manuscript and probably beyond the possibilities of the technique. However, the contrast between F(k) function for Cu and for the rest of relevant elements in the system is sufficient to reveal the presence of (a fraction) of multicopper moieties.
Indeed, the EXAFS backscattering amplitude factors in Figure 4a show that the F(k) functions of lighter elements, such as O/N and Si/Al, have maxima at around 3−4 Å −1 , while for Cu, the position of the main peak significantly shifts to a kvalue of around 7 Å −1 . These differences lead to the observed lobe splitting, enabling an unambiguous, visual discrimination of contributions stemming from Cu or framework atomic contributions. Due to the substantial overlap of the related backscattering amplitude functions, it is not possible to discriminate by means of WT among O/N and Si/Al contributions.
A WT analysis of the EXAFS spectrum for the mobile [Cu I (NH 3 ) 2 ] + complexes (Figure 4c), which is obtained after exposure of the catalyst pretreated in O 2 to the NO/NH 3 gas mixture at 200°C (step 2), shows a complete reduction and mobilization of the Cu ions in the system. 6,10,17,19,48 The second-shell peak in the conventional FT-EXAFS disappears (Figure 3a and SI, Figure S6) and all the high-R features are substantially decreased in the corresponding WT map in Figure 4c. Even though the sub-lobe at k = 7 Å −1 associated with the Cu−Cu signal is completely lost at this step, a lowintensity feature is still visible in the k-space range 1−5 Å −1 , which is most likely due to multiple scattering paths involving the first-shell N ligands in the linear [Cu I (NH 3 ) 2 ] + moieties.
The crucial step for the activation of oxygen is the isothermal oxidation of the [Cu I (NH 3 ) 2 ] + complexes, and the corresponding WT analysis is shown in Figure 4d Figure 4d), the WT intensity is rather localized in Rspace. It peaks at ca. 2.8 Å in the phase-uncorrected R-axis, pointing to a uniform Cu−Cu interatomic distance around 3.5 Å. This indicates, that the reaction of [Cu I (NH 3 ) 2 ] + with O 2 results in a well-defined structure, compatible with the side-on μ-η 2 ,η 2 -peroxo diamino dicopper(II) complex shown in Figure  1b and Figure 3c. In contrast, after heating in O 2 at 400°C and subsequent cooling to 200°C in O 2 (step 1, Figure 4b), a broader intensity distribution in R-space is observed in the kspace region characteristic of Cu−Cu scattering, in agreement with the presence of more heterogeneous multicopper species in the pretreated Cu-CHA catalyst. 29,34,35 Figure 4 also reports the EXAFS-WTs related to the reactivity of the formed μ-η 2 ,η 2 -peroxo diamino dicopper(II) complexes with the key SCR reactants, NO (step 4, Figure 4e) and NH 3 (step 4′, Figure 4f). In both cases, the sub-lobe at k = 7 Å −1 is clearly lost, providing direct structural evidence for the cleavage of dicopper cores upon separate exposure to NO or NH 3 at 200°C. A moderately intense sub-lobe is instead still visible in the low-k range, characteristic of low-Z scatterers. As argued before, this feature most likely stems from multiple scattering contributions involving N and O atoms.
Finally, to comparatively assess the presence of Cu−Cu scattering contributions throughout the investigated reaction steps, we computed the power density function Φ R of the WT representation. 45 This quantity was obtained integrating the square of the modulus of the WT over the R-range 2−4 Å, that should contain the Cu−Cu signal contribution. Φ R is given by the following expression: (1) where R min = 2 Å, R max = 4, and W ψ (k,R) is the wavelet transform representation of the EXAFS signal depending on the mother function used (see SI, section 3.1 for details). Figure 5 presents the results of these calculations, summarizing the above observations about EXAFS WTs. A common first peak, for all the steps, is localized in the 0.0−5.5 Å −1 range: it corresponds to the WT low-k sub-lobe, collectively accounting for the contributions from O, N, Si and Al atoms. The second main peak is present only in catalyst pretreated in O 2 (curve 1) and after [Cu I (NH 3 ) 2 ] + reaction with O 2 (curve 3). The position of this peak exactly corresponds to the maximum of the Cu backscattering amplitude function shown in Figure 4a, clearly indicating the formation of [Cu 2 (NH 3 ) 4 (O 2 )] 2+ complexes.

Reactivity of [Cu 2 (NH 3 ) 4 (O 2 )] 2+ Species toward NH 3 and NO.
To determine the reactivity of μ-η 2 ,η 2 -peroxo diamino dicopper(II) complexes toward NO and NH 3 , which are the main reactants in NH 3 -SCR, we have exposed them to NO and NH 3 separately, and in a 1:1 mixture. Exposure to a mixture of NO and NH 3 results in a complete restoration of a Cu I oxidation state as [Cu I (NH 3 ) 2 ] + with formation of the N 2 product, confirming the NH 3 -SCR reaction and the reversibility of the oxidation of the [Cu I (NH 3 ) 2 ] + species (see SI, Figure S12 and qualitative mass spectrometry analysis in SI, Figure S16). This observation provides experimental evidence that it is possible to close the NH 3 -SCR reaction cycle between the above-discussed and identified "homogeneous-like" μη 2 ,η 2 -peroxo diamino dicopper(II) and [Cu(NH 3 ) 2 ] + complexes.
3.2.1. Reactivity toward NH 3 : Structural Changes. As shown by EXAFS-WT analysis above (Figure 4f), exposure of the μ-η 2 ,η 2 -peroxo diamino dicopper(II) complexes to NH 3 results in the separation of the copper centers. No significant Journal of the American Chemical Society pubs.acs.org/JACS Article N 2 evolution is observed during this transformation (see SI, Figure S17). In this section, we focus on the structural changes of the resulting Cu complexes. Figure 6 reports the XAS and UV-Vis-NIR spectra for the μη 2 ,η 2 -peroxo diamino dicopper(II) complexes during exposure to NH 3 at 200°C. The characteristic Cu I transition at ∼8982.5 eV for the linear [Cu I (NH 3 ) 2 ] + appears, but does not reach the intensity observed in the fully reduced catalyst (dashed darkblue curve in Figure 6a). The intensity of the 1s→3d pre-edge peak (Figure 6c) decreases, corroborating that some reduction of the Cu takes place. In the corresponding UV-Vis-NIR spectra, the intense d-d transition, which appears at around 14 000 cm −1 (Figure 6d) becomes less intense and shifts from 13 800 to 14 400 cm −1 , without substantial change in the peak shape. The fact that this peak does not disappear completely, indicates that a part of the Cu remains in the Cu II state. The shift is in agreement with a change in the ligands bonding the Cu II ions, affecting the d-d orbital splitting. 19,32,34,49 This is also testified by the progressive consumption of the LMCT absorption between 27 000 and 31 000 cm −1 , which is consistent with the disappearance of the peroxo group in the diamino dicopper(II) complex. 26,34,35 Thus, exposure of the μη 2 ,η 2 -peroxo diamino dicopper(II) complexes to NH 3 leads to a change in the ligands of the Cu and to a partial reduction to a Cu I species.
The UV-Vis features described above only indicate the changes in the coordination sphere of Cu II ions, Cu I ions being essentially silent. On the other hand, XANES and EXAFS give average information on all copper species formed in this reaction step. A decrease in the average number of ligands surrounding Cu ions is indicated by the reduced intensity of the EXAFS first shell peak (Figure 6b Figure S13) indicates that the fractions of [Cu I (NH 3 ) 2 ] + and [Cu II (NH 3 ) 3 (X)] + at equilibrium are 65% and 35%, respectively. Considering that we have 16% Cu I and 84% Cu II before exposure to NH 3 (see section 3.1), this means that approximately 58% of the amount of the copper ions in the μ-η 2 ,η 2 -peroxo diamino dicopper(II) complexes are reduced to Cu I .
3.2.2. Reactivity toward NH 3 : Interpretation. The results described above indicate that NH 3 is reacting with the μ-η 2 ,η 2peroxo diamino dicopper(II) complexes, breaking the copper dimer and reducing a consistent fraction of the Cu II ions to Cu I . No N 2 is observed during this reaction, in agreement with the fact that direct oxidation of ammonia (eq 2), only occurs on Cu-CHA at higher temperature. 50 The XANES and UV-Vis spectra (light blue curves in Figure  6a,d) are consistent with the presence of Cu II ions in a pseudosquare-planar geometry, similar to the [Cu II (NH 3 ) 3 (OH)] + species predicted by Paolucci et al. and experimentally observed by Borfecchia et al. 7,19 We could thus hypothesize that the corresponding Cu II ions are in the form of a superoxo amino [Cu II (NH 3 ) 3 (OO*)] + complex, as depicted in Figure  6e. This geometry is consistent with the relatively high intensity of the low-k sub-lobe in EXAFS WT data (Figure 4f and curve 4′ in Figure 5), related to multiple scattering contributions from N/O ligand atoms. The superoxo [Cu II (NH 3 ) 3 (OO*)] + complex could be formed by a one-electron transfer from the bridged peroxo group in the μ-η 2 ,η 2 -peroxo diamino dicopper(II) complexes to one of the Cu II ions, with consequent formation of Cu I ions and of the superoxo ligand. The resulting Cu I ions are thus stabilized as [Cu I (NH 3 ) 2 ] + . This could be rationalized with eq 3, which should result in the reduction of 50% of the Cu II ions in the dimer to Cu I .
Our XANES linear combination fitting indicates a higher efficiency of the Cu II -to-Cu I reduction with NH 3 (ca. 58% of Cu II reduced) with respect to what is expected on the basis of eq 3. This could be related to a further reduction of the superoxo [Cu II (NH 3 ) 3 (OO*)] + complexes by the available NH 3 present in the system, with formation of [Cu I (NH 3 ) 2 ] + . Interestingly, DFT simulations using the M06-HF-D3 functional predict that reaction 3 is exothermic at the experimental conditions. The computed variation of internal energy is −4.6 kJ/mol; see SI for further details. Despite the clear indications provided by the linear combination fitting and DFT simulations, the associated uncertainties are too high to use them as an ultimate proof for the reaction 3, so this may be investigated in further works. Interestingly, the reaction tentatively proposed in (3) could provide clues as to the origin of the observed NH 3 -inhibition effect and negative apparent NH 3 rate orders, observed by different authors. 12,51,52 3.2.3. Reactivity toward NO: Experimental Evidence. The reaction of the μ-η 2 ,η 2 -peroxo diamino dicopper(II) complexes with NO results in the separation of the copper centers, as shown by EXAFS-WT analysis (see above, Figure 4e). This is accompanied by some formation of N 2 , as monitored by online mass spectrometry (see SI, Figure S18). Due to the used experimental setup, the acquired mass spectrometry data are not accurate enough to be used for quantitative considerations. This section provides information about the dynamics of the reaction and summarizes the main experimental findings. Figure 7 reports the XAS and UV-Vis-NIR spectra collected when contacting the μ-η 2 ,η 2 -peroxo diamino dicopper(II) complexes formed in step 3 with NO at 200°C. The changes in the pre-edge features at the Cu K-edge in XANES reveals a fast and effective Cu II -to-Cu I transformation. The characteristic 1s→4p transition at ∼8982.5 eV reappears in the spectrum (Figure 7a), and the weak 1s→3d transition at 8977.3 eV, indicative of a Cu II species, disappears (Figure 7c). To illustrate the differences in the formation of Cu I species in the reactions of the μ-η 2 ,η 2 -peroxo diamino dicopper(II) complexes with NO and NH 3 , we have compared the temporal evolution of the 1s→4p transition at ∼8982.5 eV in the two cases ( Figure 7f). These data show that the reduction of the μη 2 ,η 2 -peroxo diamino dicopper(II) complex is faster and more efficient with NO than with NH 3 , since the Cu I peak reaches about 76% of the intensity observed for the fully reduced reference state, i.e., [Cu I (NH 3 ) 2 ] + , as formed in step 2 (dark blue dashed line in Figure 7a) after only 5 min, stabilizing at about 84% after 35 min. In the case of reduction in NH 3 , the intensity is about 43% after 5 min, and reaches about 62% after 35 min. We note that the intensity of the Cu I rising-edge peak can be affected by the geometry of Cu complexes. 43 However, even though these data cannot be used to obtain precise kinetics of the reaction, they clearly show a difference in the reduction behavior in the two cases.
In the DR UV-Vis-NIR spectra, the reaction with NO is visible in the symmetrical d-d absorption at around 13850 cm −1 , corresponding to the Cu II sites in a pseudo-square-planar geometry in the μ-η 2 ,η 2 -peroxo diamino dicopper(II) complex. In the first minutes of NO exposure, the intensity reaches a minimum (Figure 7d, from red to orange curve), followed by the development of a complex absorption with maxima at 20 000, 16 350, 13 300, and 10 600 cm −1 (shoulder), associated with Cu II in a different local environment ( Figure  7e, from orange to light blue). These new features resemble the typical "quadruplet", as often observed in Cu-CHA samples after pretreatment in O 2 (see Figure 2 and SI, Figure S14), which are assigned to the formation of a variety of monomeric and multimeric framework-coordinated Cu II ions, such as Z[Cu II (OH)]/Z[Cu II (OO*)] sites etc. 26,33,34 This indicates the formation of some oxidic Cu II species during the reaction. The LMCT absorption between 27 000 and 31 000 cm −1 (Figure 7d), related to the bridged peroxo groups in the μη 2 ,η 2 -peroxo diamino dicopper(II) complex, also shows a rapid decrease in intensity in the reaction with NO. The subsequent change in the geometry of remaining Cu II sites is reflected in a small blue-shift in the LMCT position (Figure 7e). We note that the XANES data do not reveal the presence of a Cu II species after reaction with NO, suggesting that the observed Cu II fraction remains below the XAS detection limit under our experimental conditions, which is estimated at around 10% of the total Cu content. This would then indicate that DR UV-Vis is very sensitive to the formation of this oxidic Cu II species, due to the strong influence of the local geometry on the corresponding extinction coefficient.
The structure of the Cu I ions formed during the reaction of μ-η 2 ,η 2 -peroxo diamino dicopper(II) complex with NO is different from those observed after reduction in NO/NH 3 at 200°C or in the reaction with NH 3 . The shape of the XANES rising-edge and white-line peaks clearly differ from those of the linear [Cu(NH 3 ) 2 ] + complexes, as indicated by the light blue and dotted dark blue curves in Figure 7a, and those of ligandfree, framework-coordinated ZCu I . 25 Overall, the Cu K-edge XANES resembles that observed after desorption of NH 3 (see SI, Figure S15), which has been assigned to frameworkcoordinated linear Cu I amino complexes, Z[Cu I (NH 3 )]. 19 This assignment is supported by the NH 3 /NH 4 + vibrational modes still present in the NIR region (Figure 7d and e) and by the decrease observed in the first-shell peak in EXAFS (Figure  7b), indicating a change from a four-to a two-fold coordination of Cu (Figures 4e and 7g). The broadening of the second and third shell regions can be moreover connected to the relatively high degree of freedom (in terms of bond length and angles) of the proposed Z[Cu I (NH 3 )] entities with respect to [Cu(NH 3 ) 2 ] + or "bare" ZCu I species. 6,10, 19 3.2.4. Reactivity toward NO: Interpretation. The results in Figure 7d show that NO reaction causes a fast disaggregation of the side-on μ-η 2 ,η 2 -peroxo diamino dicopper(II) complexes. The Cu II species are almost completely reduced to Cu I , while the bridging peroxo groups are consumed and N 2 is formed. The formation of N 2 is fast (see SI, Figure S18), indicating that μ-η 2 ,η 2 -peroxo diamino dicopper(II) complexes are very reactive toward NO. These results provide experimental support for the conclusion from DFT calculations, that NO facilitates the dissociation of the O−O bond in oxygen. 4,13,15,16 According to our interpretation of the XANES results (Figure 7a), the Cu I species after reaction with NO consists of framework-coordinated Z[Cu I (NH 3 )] moieties, implying that each Cu ion in the dimer loses only one amino ligand during the reduction. Starting from the μ-η 2 ,η 2 -peroxo diamino dicopper(II) complexes, the NH 3 -SCR reaction requires stoichiometrically two NH 3 molecules per Cu for a complete reduction of the Cu II , 8 and therefore the second NH 3 molecule is a nonligand NH 3 . In a recent DFT study, it is proposed that the NH 3 -SCR reaction proceeds via the decomposition of HONO and H 2 NNO intermediates to N 2 and H 2 O, over Brønsted NH 4 + (or H + /NH 3 (g)) sites. 13 This would be a way to include a second nonligand NH 3 molecule in the SCR cycle, but such a role of Brønsted sites in the NH 3 -SCR cycle still needs experimental verification. 53 Even though the XANES results point to a complete reduction of the Cu II , a minor fraction of Cu II is still present after reaction with NO, which remains below the detection limit of XANES under our experimental conditions. The presence of these Cu II moieties in the XAS experiment is indicated by the small amount of N 2 that is formed upon adding NH 3 to the NO feed after step 4, to restore the fully reduced state consisting of [Cu(NH 3 ) 2 ] + complexes (SI, Figure S19). The formation of N 2 indicates that some reduction of Cu II takes place, thus proving that the reduction with NO alone was not complete. We expect the amount of this residual Cu II fraction to depend on the Cu content and Si/ Al ratio of the Cu-CHA material.
The features in the UV-Vis observed at 20 000, 16 350, 13 300, and 10 600 (sh) cm −1 after the reaction with NO is completed resemble the UV−Vis "quadruplet", that is often observed for Cu II in Cu-CHA and other small pore zeolites. 26,[33][34][35][36]54 These features have been assigned to Cu II species attached to the zeolite framework, such as Z-[Cu II (OH)], Z[Cu II (OO*)], framework-coordinated Cu dimers with O/OH bridging moieties, or even larger Cu clusters. 34 This suggests that the Cu II species that remains after reaction of the μ-η 2 ,η 2 -peroxo diamino dicopper(II) complex with NO is attached to the zeolite framework as well. We also note that the high intensity of these features is comparable to that observed in fully oxidized Cu-CHA samples (see SI, Figure S14), despite the low fraction of Cu II . This puzzling finding could be related to the fact that this spectroscopic feature is the result of a variety of Cu II ions with similar but not identical local environments affecting the d-splitting, 55,56 as recently predicted by Li et al. 34

CONCLUSIONS
We have studied the activation of oxygen over the mobile linear [Cu(NH 3 ) 2 ] + complexes in a Cu-CHA catalyst for NH 3 -SCR (Si/Al ratio = 15 and 2.6 wt% Cu), which is a crucial step in the NH 3 -SCR reaction, by combining X-ray absorption spectroscopy, and diffuse reflectance UV-Vis-NIR spectroscopy.