Copper Pairing in the Mordenite Framework as a Function of the CuI/CuII Speciation

Abstract A series of gas‐phase reactants is used to treat a Cu‐exchanged mordenite zeolite with the aim of studying the influence of the reaction environment on the formation of Cu pairs. The rearrangement of Cu ions to form multimeric sites as a function of their oxidation state was probed by X‐ray absorption spectroscopy (XAS) and also by applying advanced analysis through wavelet transform, a method able to specifically locate Cu–Cu interactions also in the presence of overlapping contributions from other scattering paths. The nature of the Cu‐oxo species formed upon oxidation was further crosschecked by DFT‐assisted fitting of the EXAFS data and by resonant Raman spectroscopy. Altogether, the CuI/CuII speciation clearly correlates with Cu proximity, with metal ion pairs quantitatively forming under an oxidative environment.


Introduction
Within the paradigm of an improved use of fossil resources,s trategies facilitating the exploitation of many small/remote natural gas sources (including biogas) is of utmost importance. [1,2] In particular,c hemical processes able to transform methane (the main constituent of natural gas) into liquid analogues are desirable for as implified handling and transportation. Thepresently implemented syngas-based technologies are energy intensive processes,t hus economically viable only for large-scale applications. [3] As an alternative,t he direct conversion of methane to methanol (DMTM) via partial oxidation represents ap romising route. In analogy to biological systems (e.g.methanotrophic bacteria hosting the pMMO enzymes), [4][5][6] several Cu-based catalysts for DMTM have been developed and studied. [7] Among them, Cu-exchanged zeolites have received much attention after the discovery of their activity in DMTM by Schoonheydtsgroup in 2005. [8] Many different topologies and compositions (in terms of Si/Al and Cu/Al ratios) have been explored, aiming at finding structure-activity relationships. [9] Acombination of experimental and theoretical techniques has been applied to identify the active species responsible for this reactivity,a nd many mechanisms have been proposed to explain their formation and outstanding selectivity.M oreover,d ifferent reaction conditions heavily impact the speciation of Cu inside the zeolite framework, thus influencing DMTM. [10] Overall, the Cu I /Cu II redox cycle and the local coordination environment around Cu sites are the key features for understanding these complex systems.
In this context, opportune gas-phase reactants (e.g.N H 3 ) affect both Cu speciation and framework distribution when used prior to oxidation, as the coordinative nature of NH 3 enhances the Cu cations mobility. [11] Accordingly,t he Cu speciation in the framework changes, [12] as well as the average Cu-Cu distance,int urn conditioning the formation of active species.I ndeed, these are supposedly multinuclear Cu sites bridged by Oa toms;d ifferent Cu-Cu distances could affect oxidation pathways and drive towards different speciation. Thereby,u nderstanding the effect of the redox history of aCu-zeolite on the final speciation of Cu sites (including Cuoxo species) is relevant for optimizing both DMTM catalysts and reaction protocols.A ccordingly,aselection of reducing agents has been used to activate aC u-MOR sample subsequently characterized through XAS,a iming at probing the oxidation state and local environment of the resulting species. Furthermore,t he application of Wavelet Transform (WT) Analysis [13] on these data allowed us correlating the existence of Cu-Cu interactions with the fraction of reduced/oxidized Cu species.T hrough DFT-supported fitting of the EXAFS data, specific Cu-oxo species were identified.

Results and Discussion
In this work, we focus on an ad hoc synthesized Cu-MOR (Si/Al = 8.22 and Cu/Al = 0.27, atomic ratios from EDX), prepared and characterized as described in Section S1 of the Supporting Information (SI). Basic characterization data (powder XRD and N 2 adsorption isotherm at 77 K) are provided in Figure S1 of the SI. We relied on ah ome-made material rather than on ac ommercial one,s ince Cu-MOR samples we previously studied [14] have been proved to contain traces of TiO 2 (anatase polymorph), as shown in Figure S2 of the SI, making it unsuitable for Raman and optical characterization. Figure 1a and cshows the ex situ XAS data collected at RT for the Cu-MOR treated according to Section S1.1. Each sample was subjected to as pecific gas-phase redox treatment. Reduction in NH 3 ,H 2 ,C Oa nd CH 4 was performed at 250 8 8C. Asample deeply reduced in NH 3 at 500 8 8C was also prepared. Regardless the reduction treatment, each sample was subsequently outgassed at 500 8 8Ct oe nsure complete desorption of species possibly coordinating the Cu ions.O xidized samples were produced by exposing the sample,p ossibly pre-reduced in NH 3 at 500 8 8C, to pure O 2 at 500 8 8C. Finally,asample that just underwent vacuum dehydration (thus triggering self-reduction) was considered. Given the edge energy position and the observed XANES features, any contributions from metallic Cu, even in the form of small Cu 0 clusters,c an be safely ruled out over the whole set of investigated samples (see also Figure S3). TheX ANES spectrum in Figure 1a,b elonging to the sample reduced at 500 8 8Cinthe presence of NH 3 ,shows no traces of the pre-edge 1s!3d transition arising at ca. 8978 eV (see the inset in Figure 1a), typical of Cu II ions.Conversely,itischaracterized by ap rominent 1s!4p rising-edge peak located at ca. 8983 eV and by alow intensity in the white line (WL) region, typical of as ite with alow coordination number. In the limit of the energy resolution, these spectral features infer the existence of aq uasi-linear pure Cu I site,i na ccordance with previous studies,a lso involving model compounds. [2,10,[14][15][16][17][18][19] This interpretation is in line with the phase-uncorrected FT-EXAFS spectrum (Figure 1b): the first-shell maximum is observed at 1.5 ,with ashoulder extending within 1.8-2.6 . Thelow intensity of the first-shell peak agrees with atwofold coordinated Cu I center, most likely with framework oxygen atoms.The XANES spectra of the samples treated in NH 3 and in H 2 at 250 8 8Ca re still dominated by aC u I species,b ut the presence am inor fraction of Cu II sites is indicated by the appearance of the Cu II 1s!4p transition at ca. 8987 eV.Subtle modifications are also noted in the pre-edge range,pointing to at race of the Cu II 1s!3d peak. Yet, within the available energy resolution, we cannot conclusively comment about this inherently weak spectral feature in samples 2a nd 3. Consistently,ifcompared to the NH 3 500 8 8Cstate,the EXAFS spectrum exhibits amore intense first-shell peak and astructured second-shell peak, consistent with the presence of Cu II sites interacting with the framework and having ah igher coordination number. TheXANES spectra corresponding to the samples reduced in CO,C H 4 and self-reduced (SR) are almost mutually identical.
From these profiles an increase in the relative contribution from Cu II centers is clear, determining the abatement of the Cu I 1s!4p transition and causing the increase of the Cu II 1s!3d and 1s!4p ones.Overall, XANES indicates an almost equal Cu I /Cu II fraction with ap otential slight preference for Cu II in these samples.T he related EXAFS spectra show the same first-and second-shell features described before,b ut more intense due to the larger fraction of Cu II sites. Finally,i na greement with the literature, [10,14,[19][20][21][22][23] the XANES spectra of the samples oxidized at 500 8 8Caccount for al argely dominant Cu II oxidation state.T he Cu I 1s!4p transition is absent in the XANES of both oxidized samples, inferring apure Cu II state.The corresponding EXAFS spectra are consistent with three/fourfold O-ligated Cu II ,l ocated at well-defined ion-exchange sites in the zeolite framework. [10,12] Thesecond shell appears well-structured and it stems mainly from the Cu-Al/Si and Cu-Cu single scattering (SS) contributions.
Thea mount of Cu I and Cu II species in each XANES spectrum was obtained by linear combination fit (LCF) [24] on the normalized XANES (Figure 1a), in the 8975-9020 eV range.T he number of chemical species was defined through aPrincipal Component Analysis [25][26][27] (see Figure S4 of the SI) and set to two.T he XANES spectra of samples reduced in NH 3 (sample 1) and pre-reduced in NH 3 and then oxidized in O 2 at 500 8 8C(sample 8) were chosen as standards for the leastsquare procedure. [28] TheXANES LCF provided avery small R factor (0.04 %) concerning the reconstruction of the experimental XANES,i ndicating that the selected references are suitable to reproduce each single XANES spectrum of the dataset;adirect comparison between experimental and LCF curves can be found in Figure S5. TheC u I -Cu II fraction, retrieved by this approach, is reported in Figure 1c.T he samples treated with NH 3 and H 2 at 250 8 8Cs how an almost equivalent larger abundance (about 65 %) of Cu I sites. Instead, the self-reduced sample and the ones treated with CO or CH 4 show al ower fraction of Cu I sites (below 40 %), becoming closely nil in the sample directly oxidized at 500 8 8C (< 5%). Due to its sensitivity to the chemical nature of the scatterers surrounding the absorbing atom, we employed the WT approach to provide amore robust description of high R EXAFS features. [13,29] All the WTrepresentations show amain lobe at low R values (Dk:0.0-12.5 À1 and DR:0.5-2.0 ), as observed on the full two dimensional (k,R) maps in Figure S6. As previously described, this feature is due to the SS contributions arising from the (extra-)framework oxygen atoms located in the first coordination shell of the Cu centers.  expected. TheW Tm ap of the sample reduced in NH 3 at 500 8 8Conly shows aweak lobe at k values around 4.0 À1 .In this region, the backscattering amplitude factor terms for Cu-Oa nd Cu-Al/Si SS have their maxima, largely overlapped (see Figure S7). TheW Tm ap confirms that the shoulder appearing in the FT-EXAFS for this sample derives from weak scattering paths involving the farther framework Oand Si/Al atoms of the low coordinated Cu I site.Inthe WT maps of samples reduced at 250 8 8CinNH 3 and H 2 ,the low k-region becomes more structured. Asecond lobe appears at higher k values (Dk:5 .0-7.0 À1 and DR:2 .0-4.0 ), arising from intensified contributions of scattering paths among Si/Al and Cu II sites,s uperimposed to contributions from the lattice O farther from Cu centers.T he presence of as mall Cu-Cu contribution cannot be excluded, albeit overshadowed by signal from the framework atoms.I nf act, aw eak ridge is observed at 7.0 À1 ,w here the backscattering amplitude factor for aC u-Cu SS has its maximum. Thel atter presents increased intensity in the WT maps of samples reduced in CO, CH 4 and SR. Finally,t his feature reaches its maximum intensity for both samples oxidized in O 2 at 500 8 8C, that is, when the content of Cu II sites is the highest. [13,30] At its maximum development, this sub-lobe extends over ar ather broad R-space range,p ointing to ar elatively high level of structural disorder in the Cu-Cu interatomic distance distribution. In order to comparatively assess the presence of Cu-Cu scattering contributions through the investigated states, similarly to some previous reports, [13,30,31] we computed the power density function F R (k)o fe ach WT representation. This quantity was obtained integrating the squared modulus of the WT over the R range within 2.0 and 4.0 ,w hich ensures the inclusion, if present, of whatever Cu-Cu path contribution. Figure 3ashows the results of these calculations. Herein, af irst peak is located for all the states between 0.0-5.5 À1 ,a scribed to the WT low-k sub-lobes,w hich collectively account for the contribution due to O, Si and Al atomic neighbors surrounding the Cu centers.T he second, weaker peak (6.0-8.0 À1 )i sa scribable to Cu-Cu contributions, reflecting the high-k sub-lobe in the WT maps discussed before.T he intensity of this peak, as measured at its maximum at 6.8 À1 for each of the considered states,nicely correlates with the Cu II fraction as previously determined by XANES LCF.S uch ar esult, depicted in Figure 3b,q uantitatively confirms the preference of Cu II toward the formation of multimeric species,w hereas Cu I ions remain preferentially separated. Thea pproach of Cu ions in their oxidized form, coherently observed in an oxidative environment, is most probably accompanied by the formation of Cu-oxo species, supposedly active in DMTM. Thus,t heir exact identification is of utmost interest toward ab etter understanding of the methane oxidation process (despite not explicitly investigated here). Furthermore,the observation of fingerprints of specific Cu-oxo species also supports the qualitative analysis of Cu-Cu distances as inferred by EXAFS data obtained via Fourier or Wavelet transform. Thereby,anE XAFS fitting procedure was carried out. In particular,wefocused our analysis on the two extreme cases:i )att he most reduced sample,t hat is, treated at 500 8 8CinNH 3 (sample 1inFigure 1);and ii)further oxidized in pure O 2 at 500 8 8C( sample 8i nF igure 1). As structural guess we created, on the basis of the recent literature, [10,13,14,17,22,32] four DFT models as described in detail in the Section S5 of SI. TheC u(-oxo) models were hosted in the eight-membered rings of the side pocket of MOR structure,w here the siting of Al atoms (i.e.t he anchoring site for the Cu ions) was selected on the basis of asystematic prescreening of all possible configurations (see Figure S8 for the considered Al substitutional sites and Table S1 for main outcomes). Four Cu(-oxo) structures were considered (graphically represented in Figure S9)  with all the aforementioned Cu-oxo models.N onetheless, only the fit based on the [Cu II -O-Cu II ] 2+ structure was sufficiently in agreement with the experimental data (see Section S6 of the SI) and will be discussed herein. The EXAFS fit results,s ummarized in Figure 4a nd Table S2 of the SI (fixed coordination numbers for the 2[Cu I ] + and the [Cu II -O-Cu II ] 2+ DFT models are given in Table S3 and S4, respectively) and the in Table,p roperly reproduce the experimental spectra (R factor lower than 1% in both the cases) and provided as et of physically reliable optimized parameters.R esults obtained for af it attempt performed with the [Cu II -OO-Cu II ] 2+ sideon structural model are reported in Table S5 (fixed coordination numbers in Table S6), though some physically inconsistent parameters were obtained. Thefitting strategies and parametrization adopted are described in Section S6.1 of the SI, while the individual EXAFS path contributions are shown in Figure S10 and S11. Raw data and best-fit in k-space are shown in Figure S12.
Focusing on the sample reduced in NH 3 at 500 8 8C, the EXAFS first-shell feature mostly originates from the SS paths involving the framework oxygen atoms sited close to the Cu absorbers.The latter were found to be approximately in phase (see Figure S10) with an average Cu-O fw1 distance refined at 1.93 AE 0.01 .T he broadening of the first-shell peak toward longer distances can be ascribed to as econd type of O framework atoms at as lightly longer distance from the absorber (i.e.Cu-O fw2 ). Thecontribution from farther framework atoms (O,A la nd Si)i si nstead weak since,d ue to the high heterogeneity of the Cu-sites,t hese paths are in antiphase to each other, causing the abatement of their EXAFS signals (see Figure S10). Indeed, high R-values features are absent, in accordance with the related WT representation (see Figure 2, panel 1). Finally,the interaction involving the Cu-Cu path was not included in the fit, since the interatomic distance among the two Cu I atoms in the 2[Cu I ] + DFT model is > 5.0 ,that is,outside the limit of detectability of FT-EXAFS analysis for this case.Considering the EXAFS spectrum of Cu-MOR activated in O 2 at 500 8 8Ca fter prereduction in NH 3 at 500 8 8C, the first-shell of the experimental FT-profile is successfully reproduced considering two subshell of Oneighbors in the first-coordination sphere,involving framework (O fw1 )and extraframework (O efw ,that is,involved in the formation of Cu-oxo species) Oa toms.T hese two families of Oa toms contribute in partial antiphase (see Figure S11) at 1.93 AE 0.02 and 1.97 AE 0.02 ,respectively.The second maximum of the FT-EXAFS is effectively modelled by the SS contribution of asingle T fw (Al) atom at adistance of 2.68 AE 0.01 from the Cu absorber.I nt he high-R range, the contribution from farther framework Oand Si atoms (fw) is observed too.F inally,aCu-Cu contribution is refined at 3.28 AE 0.08 (ca. 2.88 in the phase uncorrected FT-EXAFS plot), consistent with previous WT-EXAFS fitting results on oxidized Cu-MOR, [13] as well as conventional EXAFS fitting on Cu-CHA. [10,30] Based on am ore symmetrical mono-m-oxo dicopper site model, Sushkevich et al. [29] reported instead shorter Cu-Cu separations around 2.85 ,yet associated with minority Cu-species,asindicated by coordination numbers of ca. 0.3. Not surprisingly,o ur EXAFS analysis revealed ar elatively high Debye-Waller factor associated with Cu-Cu scattering path, properly reflecting the broad intensity distribution observed for the Cu-Cu sub-lobe in the WT maps ( Figure 2). Thesmall deviation (ca. À0.02 )ofthe fitted R Cu from the DFT distance fully supports the choice of this structure for the EXAFS fitting refinement, in line with the occurrence of such type of [Cu II -O-Cu II ] 2+ cores as adominant, although not exclusive,c onfiguration under the adopted pretreatment conditions.

Conclusion
In summary,asample of Cu-mordenite was systematically treated with abroad set of gas-phase reactants to gain specific information on the influence of different redox-active molecules on Cu pairing.X AS was employed to simultaneously probe the oxidation state and the proximity of the Cu sites as afunction of different redox treatments.Inaddition, the WT approach (augmented by the power density function analysis) was demonstrated being an irreplaceable tool toward the selective assessment of Cu-Cu contributions in the XAS dataset, showing arelation between the oxidation state of the metal center and the proximity of Cu sites.F or the first time, the presence of Cu pairs was quantitatively correlated to the fraction of oxidized Cu II sites present in the sample.D FTassisted EXAFS fitting further allowed identifying the fingerprints of the Cu-oxo species formed after oxidative treatment as due to dimeric Cu-O-Cu sites (also confirmed by resonant Raman spectroscopy,s ee Figure S13). As multiple Cu sites have been proved to work cooperatively through the redox pathways that lead both to activation of the material and reaction with methane,r eliably monitoring the specific interactions between the metal sites is essential to understand the mechanism underlying these processes.I na ddition, exploring how different reactants interact with these materials help in identifying relationships between the activation protocol and the reaction performances.Exploring in arational way the range of variability that characterizes these systems (topology,c omposition, reactants,t emperatures) with tools capable to detect such specific features and trends will be the key to ab etter understanding of selective DMTM process, pointing toward ad hoc engineering of catalysts and reaction protocols that could maximize selectivity and productivity.