Operando Insight into the Correlation between the Structure and Composition of CuZn Nanoparticles and Their Selectivity for the Electrochemical CO2 Reduction

Bimetallic CuZn catalysts have been recently proposed as alternatives in order to achieve selectivity control during the electrochemical reduction of CO2 (CO2RR). However, fundamental understanding of the underlying reaction mechanism and parameters determining the CO2RR performance is still missing. In this study, we have employed size-controlled (∼5 nm) Cu100–xZnx nanoparticles (NPs) supported on carbon to investigate the correlation between their structure and composition and catalytic performance. By tuning the concentration of Zn, a drastic increase in CH4 selectivity [∼70% Faradaic efficiency (F.E.)] could be achieved for Zn contents from 10 to 50, which was accompanied by a suppression of the H2 production. Samples containing a higher Zn concentration displayed significantly lower CH4 production and an abrupt switch in the selectivity to CO. Lack of metal leaching was observed based on quasi in situ X-ray photoelectron spectroscopy (XPS). Operando X-ray absorption fine structure (XAFS) spectroscopy measurements revealed that the alloying of Cu atoms with Zn atoms takes place under reaction conditions and plays a determining role in the product selectivity. Time-dependent XAFS analysis showed that the local structure and chemical environment around the Cu atoms continuously evolve during CO2RR for several hours. In particular, cationic Zn species initially present were found to get reduced as the reaction proceeded, leading to the formation of a CuZn alloy (brass). The evolution of the Cu–Zn interaction with time during CO2RR was found to be responsible for the change in the selectivity from CH4 over Cu-ZnO NPs to CO over CuZn alloy NPs. This study highlights the importance of having access to in depth information on the interplay between the different atomic species in bimetallic NP electrocatalysts under operando reaction conditions in order to understand and ultimately tune their reactivity.


INTRODUCTION
The electrochemical reduction of CO 2 (CO 2 RR) into useful chemicals and fuels has received much attention as a means to build carbon recycling systems. 1,2 However, efficient and inexpensive electrocatalysts are still required to reduce the thermodynamically stable CO 2 molecule while suppressing the H 2 evolution reaction (HER). While various metal surfaces (bulk foils) have been identified to efficiently reduce CO 2 into value-added carbon-based products such as CO (Au, Ag, and Zn), formic acid (Sn, In, and Bi), and hydrocarbons (Cu), 3 their catalytic activity, selectivity, and stability are still insufficient for industrial acceptance. In order to enhance the performance of metal catalysts, several strategies have been proposed, including nanostructuring the metals, 4−11 engineering the metal/electrolyte interface, 12−17 or introducing a secondary metal to create bimetallic structural motifs. 18−27 The utilization of bimetallic catalysts has been considered a promising approach to obtain improved catalytic performance for CO 2 RR. 28,29 CuZn is of particular interest due to its low cost and lack of toxicity. Recent studies have demonstrated enhanced reactivity of CuZn catalysts for CO 2 RR. 30−36 For example, nanoporous CuZn catalysts prepared by annealing and subsequent reduction of commercial CuZn alloy foils showed four and six times higher Faradaic efficiency (F.E.) for CO and HCOOH than those of the untreated CuZn foils. 32 Zn-coated Cu electrodes exhibited higher selectivity for CH 4 (52% F.E.) than bare Cu (23% FE for CH 4 ). 33 Oxide-derived CuZn catalysts were favorable toward the formation of C 2 products (i.e., C 2 H 4 and C 2 H 5 OH), and it was possible to tune the ratio of these products by varying the amount of Zn. 34 This trend in the C 2 selectivity was postulated to be due to the spillover of CO from Zn to Cu sites, which was thought to facilitate the production of C 2 products at the Cu site. 34 Such CO spillover effects were found to be facilitated when a homogeneous distribution of Cu and Zn atoms was formed in the CuZn catalysts. 35 However, despite the former encouraging empirical results, significant discrepancies in the product selectivity of seemingly similar CuZn systems have been reported, and fundamental understanding is still missing regarding the reaction pathways, catalyst structure, and composition leading to a given selectivity trend. Moreover, uncertainties remain concerning the relative importance of the different parameters that may be adjusted in order to tune the materials electrocatalytic performance. For the bulk CuZn catalysts previously investigated, grain boundaries, 37−39 local pH effects, 40−42 or the presence of cationic metal species 6,43−46 have been suggested to contribute to their altered product selectivity. However, the complexity of these material systems hinders our ability to disentangle the specific role of different parameters, which is required in order to understand the intrinsic reactivity of the CuZn system for CO 2 RR.
In this work, morphologically and chemically well-defined ∼5 nm CuZn nanoparticles (NPs) were synthesized via inverse micelle encapsulation and used to investigate the correlation between their structure, composition, and electrocatalytic activity and selectivity. By means of quasi in situ X-ray photoelectron spectroscopy (XPS) and operando X-ray absorption fine-structure spectroscopy (XAFS), the evolution of the structure, chemical state, and composition of the CuZn NPs was investigated under working CO 2 RR conditions. The gradual formation of a Cu−Zn alloy could be observed in the course of the CO 2 RR and correlated to the switch in selectivity from CH 4 to CO.

EXPERIMENTAL SECTION
2.1. Preparation of CuZn NPs. Bimetallic CuZn nanoparticles (NPs) with the same nominal NP size but variable composition were prepared by an inverse micelle encapsulation method. 47 The elemental composition was modified by controlling the molar ratio of the Cu and Zn precursor salts. CuCl 2 and Zn(CH 3 COO) 2 were dissolved in a solution of toluene and tetrahydrofuran, followed by the addition of a second toluene solution with a poly(styrene)-blockpoly(2-vinylpyridine) diblock copolymer (PS(48500)-P2VP(70000), Polymer Source, Inc.). The mixture was then stirred for 2 days. The encapsulated CuZn NPs were deposited on glassy carbon substrates. After deposition, an O 2 -plasma etch treatment (20 W for 10 min, ∼400 mTorr O 2 ) was used to remove the polymeric ligands. The specific synthesis parameters of the seven samples used in this study can be found in Table S1.
2.2. Morphological Characterization. Atomic force microscopy (AFM, Bruker, Multimode 8) was used to determine the NP height (h p ). The CuZn NPs were deposited on silicon wafers for more accurate measurement of h p , since the glassy carbon electrode used for the electrochemical measurements has a rougher surface that makes difficult the background subtraction needed for the NP height determination. The average NP height was used to calculate the average CuZn geometric surface area of the NPs, A CuZn = 4π(h p /2) 2 , assuming spherical micellar NPs. The metal NP surface area was then multiplied by the NP number density to estimate the total surface area of the CuZn NPs in each sample. This parameter was used for the normalization of the current data.
2.3. Electrochemical Characterization. Electrochemical CO 2 reduction experiments were conducted using an Autolab potentiostat (Multi Autolab M204) in an H-type two-compartment electrochemical cell made of polyether ether ketone (PEEK) separated by an anion-exchange membrane (Selemion AMV). A platinum mesh counter electrode and a leak-free Ag/AgCl reference electrode (Innovative Instruments) were used in a three electrode configuration. Purified 0.1 M KHCO 3 solutions were prepared by treating the electrolyte with Chelex 100 Resin (Bio-Rad). The electrolyte was saturated with CO 2 until a pH of 6.8 was achieved. Each data point presented corresponds to an identical freshly prepared sample measured with the chronoamperometric technique at −1.35 V versus RHE.
The gas products were quantified by gas chromatography (GC, Agilent 7890B) equipped with thermal conductivity (TCD) and flame ionization (FID) detectors. The GC was directly connected to the electrochemical cell for online gas analysis. CO 2 gas was bubbled through the electrolyte at an average rate of 20 mL min −1 . The formic acid concentration was analyzed by high-performance liquid chromatography (HPLC, Shimadzu Prominence) equipped with a NUCLEOGEL SUGAR 810 column and a refractive index detector (RID).
2.4. Structural and Chemical Characterization. Quasi in situ X-ray photoelectron spectroscopy (XPS) measurements were performed in an ultrahigh-vacuum setup equipped with a nonmonochromatic Al X-ray source (hv = 1486.6 eV) and a hemispherical electron analyzer (Phoibos 100, SPECS GMbH). The XPS analysis chamber was connected to an electrochemical cell allowing sample transfer without exposure to air after electrochemistry. The measurements were conducted using an analyzer pass energy of E pass = 13 eV and a source power of P = 300 W. All spectra were aligned to the carbon peak (E bin = 284.8 eV) of the glassy carbon substrate. The composition ratio of Cu to Zn was calculated taking into account the relative sensitivity factors (RSF) of the metals (Cu 2p 3/2 : 20.28; Zn 2p 3/2 : 23.93) for an angle of 54°between the X-ray source and the analyzer and the transmission function of the analyzer.
Operando X-ray absorption fine-structure spectroscopy (XAFS) measurements were performed at the SAMBA beamline at SOLEIL synchrotron (France). A homemade electrochemical cell was used to acquire the XAFS spectra. All samples were measured in air and under operando conditions, while the potential was kept constant at −1.35 V versus RHE in a CO 2 -saturated 0.1 M KHCO 3 electrolyte. XAFS spectra for the Cu K-edge (E 0 = 8979 eV) and Zn K-edge (E 0 = 9659 eV) were collected separately. Identically prepared (same NP solution) but different (fresh) samples were used for measurements at either the Cu or the Zn absorption edges. Further details of the XAFS measurements are given in the Supporting Information. Under CO 2 RR conditions, multiple XAFS spectra were acquired, until there

Journal of the American Chemical Society
Article were no visible changes. This took several hours for these samples. Here we use time-dependent X-ray absorption near edge structure (XANES) data to follow the evolution of the sample structure and composition. We also analyzed extended X-ray absorption fine structure (EXAFS) data and compared the data obtained for samples in their as-prepared state in air and in their final state after several hours under applied potential. In addition to the evaluation of the XANES data, EXAFS spectra collected after 1 h CO 2 RR under potential were also analyzed. Note, however, that the acquisition of a single EXAFS spectrum took ca. 12 min, which limits our time resolution. Moreover, due to the low signal-to-noise ratio, analysis of individual EXAFS spectra is problematic, and several spectra needed to be merged. Therefore, the representative EXAFS spectrum for a sample after 1 h of CO 2 RR was obtained after merging 4 spectra collected during 40−80 min of CO 2 RR.
Alignment, background subtraction, and normalization of the XAFS spectra were performed using the conventional approach as implemented in the Athena software. 48 Linear combination fitting was used to process the XANES data. To obtain quantitative information about the distributions of bond lengths, we performed Cu-and Zn K-edge EXAFS data fitting. Details of EXAFS data fitting are given in the Supporting Information. Figures 1 and S1 present AFM images of CuZn NPs with variable composition ratios. The AFM images show that all NPs exhibit spherical shape with uniform coverage across the substrate and narrow size distribution, with an average size of ca. 5 nm. Figure S2 displays the NP height histograms and the average values are included in Table S1. Following our previous work, 46 the metal NP coverage was chosen to be large enough (Table S1) so that the interparticle reactant diffusion and probability of readsorption of intermediates is not hindered.

RESULTS AND DISCUSSION
In order to investigate the surface composition and chemical state of the CuZn NPs in their as-prepared state and after CO 2 RR, quasi in situ XPS measurements were carried out. Figures 2 and S3 show the XPS spectra of the CuZn NPs in their as-prepared state and after CO 2 RR. The surface composition ratio was obtained by integrating the areas of the Cu 2p 3/2 and Zn 2p 3/2 spectra. It was confirmed that the composition of the as-prepared samples was consistent with the starting molar ratio of the precursor salts, and that it did not change after the CO 2 RR ( Figure S4).
As shown in Figure 2a, Cu in the as-prepared samples after O 2 -plasma treatment was found to be in Cu 2+ state, as evidenced by the shakeup satellites. 47,49 After 1 h CO 2 RR, the shakeup features vanished. Simultaneously, a shift of the Cu 2p 3/2 main peak toward lower binding energies was observed (ΔE = 1.0 eV). Nevertheless, a ∼0.8 eV higher binding energy than that of bulk metallic Cu was observed after CO 2 RR. This binding energy shift can be attributed to the initial and final state effects typically observed for small NPs. 49 We also attempted to detect the presence of Cu + species and distinguish them from Cu metal by using the Cu LMM Auger peak. 50 Unfortunately, the former signal for our lowmetal coverage NP samples was beyond the detection limit. The Zn 2p region (Figures 2b and S3) shows that before CO 2 RR the NPs are in the Zn 2+ state (ZnO). The small change of 0.5 eV toward higher binding energies of the Zn 2p 3/2 peak obtained after 1 h CO 2 RR is attributed to the presence of Zn(OH) 2 . 51 Figure 3 shows the current density and F.E. for the CuZn NP samples as a function of their composition. As shown in Figure 3a, the Cu 100 NPs exhibit the highest activity among all investigated samples, and the activity was found to monotonically decrease with increasing Zn content in the CuZn NPs. Interestingly, the F.E. of the CuZn NPs showed different selectivity during CO 2 RR depending on the Cu to Zn ratio ( Figure 3b). Cu 100 and Zn 100 NPs produced CH 4 and CO as the main products, respectively. The three Cu 100−x Zn x NPs (x = 10, 30, and 50) exhibited enhanced CH 4 F.E. compared to pure Cu NPs. Nonetheless, the increased CH 4 selectivity of these three samples was abruptly suppressed when the Zn content was higher than 70%, whereas CO selectivity was found to increase.
In order to further understand the reactivity of the CuZn NPs for CO 2 RR, we also compared the partial current density of the major products as a function of the NP composition, as shown in Figure S5. Interestingly, we found that H 2 production showed a different trend depending on the CH 4 and CO production of the CuZn NPs. The H 2 production was significantly suppressed upon increasing the Zn concentration for the Cu-rich CH 4 -producing CuZn NPs (i.e., Cu 90 Zn 10 , Cu 70 Zn 30 , and Cu 50 Zn 50 ), while no changes in H 2 with Zn concentration were observed when the Zn-rich CO-producing CuZn NPs (i.e., Cu 30 Zn 70 and Cu 10 Zn 90 ) were considered. This result suggests that the protons that should be used to

Journal of the American Chemical Society
Article generate H 2 from water splitting have now a more favorable pathway toward CH 4 formation in the Cu-rich NPs.
In general, the product selectivity of CO 2 reduction is determined by the binding strength of adsorbed *CO and *H on the metal surface. 52,53 According to previous studies, since a bulk polycrystalline Cu electrode has a moderate binding energy for adsorbed *CO, such *CO species can be further hydrogenated to produce CH 4 . 54−56 On the other hand, a relatively weak binding strength of adsorbed *CO is known for polycrystalline Zn electrodes, leading to CO production without further reaction. 34,35 These are also the main products detected for our monometallic Cu 100 and Zn 100 NP samples. However, it should be also considered that the enhanced fraction of low-coordinated atoms in NP samples such as those available here (5 nm) affect the relative binding of *H and *CO with respect to bulk systems of analogous composition, leading to distinct selectivity trends, in particular, enhanced H 2 production, as previously reported. 47,51 Interestingly, we found for our CuZn NPs increased CH 4 production as compared to Cu 100 NPs. However, when the Zn content was higher than 70% in the bimetallic NPs, only CO and H 2 were produced. In the latter case, the proton and adsorbed *CO species that were not participating in the hydrogenation with adsorbed *CO on the Cu site may be released as H 2 and CO. Furthermore, the stability tests of the Cu 50 Zn 50 sample showed that the enhanced CH 4 selectivity started to drop after 1 h, which is accompanied by a slight increase in CO selectivity (Figure 3c). This result indicates that the functionality of the Cu sites has changed over time.
To gain more detailed information about the chemical state and structure of the electrocatalysts during CO 2 RR, XAFS measurements were conducted under operando conditions. Figure 4 shows selected Cu and Zn K-edge XANES spectra for Cu 50 Z 50 NPs in their as-prepared state and under CO 2 RR conditions. Similar plots for Cu 30 Zn 70 and Zn 100 samples are given in Figure S6. Complete sets of time-dependent XANES spectra for these samples are shown in Figure S7. The position of the absorption edge in the XANES spectra of the asprepared samples and the direct comparison with the reference materials (Figure 4 and S8) indicate that the Cu and Zn components are initially completely oxidized and predominantly in the 2+ oxidation state. The local structure around the Cu atoms is similar to that in Cu(OH) 2 , while from the Zn K-edge data we could not assign the observed XANES features to any particular reference material and can only conclude that some disordered oxide (or hydroxide) structure is formed. 51 Under CO 2 RR conditions, both Cu and Zn are reduced, as evidenced by the observed changes in the Cu and Zn K-edge XANES spectra, which become more similar to those of Cu and Zn foils (Figures 4 and S6−S7). Interestingly, the reduction rates appear to be very different for Cu and Zn. We observed that the Cu K-edge XANES spectra did not change much after the prompt initial reduction during the first scan under applied potential. However, at the Zn K-edge some variations in the XANES spectra for samples under potential were observed that had significantly longer characteristic times. To quantify these trends, we performed linear combination analysis using as reference the corresponding XANES spectra of the as-prepared sample (spectrum A in Figures 4a,b and S6), and the final spectrum (spectrum C in Figures 4a,b and S6) obtained after several hours under CO 2 RR conditions. The results for different samples and absorption edges are compared in Figure 4c. It is evident that the reduction of ZnO species takes several hours, while that of the Cu oxide species is completed during the first XAFS scan within several minutes. In agreement with the quasi in situ XPS results, neither the Zn on the NP surface nor that inside the NP core is completely reduced after 1 h of CO 2 RR. The reduction rate of Zn depends, however, on the Cu 100−x Zn x composition, with

Journal of the American Chemical Society
Article ZnO species in pure Zn NPs becoming reduced significantly faster than those in the bimetallic CuZn NPs. Since the Cu oxide species initially present in the small (∼5 nm) NPs were found by XAFS to become reduced very promptly (within several minutes), they can be ruled out as responsible for the much slower change in the selectivity observed.

Journal of the American Chemical Society
Article Even though the chemical state of the Cu species was stable (reduced) during the reaction, small variations in the Cu Kedge XANES spectra can still be observed, which can be linked to gradual changes in the local structure around Cu. To track these changes, linear combination analysis is employed, using in this case as references Cu K-edge XANES spectra corresponding to the final state (spectrum C in Figures 4a and S6b) and the first Cu K-edge XANES spectrum for a reduced sample (obtained after several minutes under CO 2 RR conditions, spectrum B in Figures 4a and S6b). The obtained results, Figure 4d, demonstrate that the local structure around the Cu atoms continues evolving even after Cu is completely reduced. It should be noted that the characteristic times at which changes in the Cu K-edge XANES spectra are observed (Figure 4d) correlate well with the characteristic times of the reduction of the cationic Zn species (Figure 4c). Thus, from this observation, we can infer that the changes observed originate from the gradual enrichment of the samples with reduced Zn species and the interaction between the metallic Cu and metallic Zn species formed during CO 2 RR.
To better understand the nature of the Cu−Zn interactions, EXAFS spectra, which are more sensitive to the geometry of the environment around the absorbing atoms, were analyzed. Fourier transformed (FT) Cu K-edge EXAFS spectra of Cu 50 Zn 50 , Cu 30 Zn 70 , and Zn 100 NPs measured for the asprepared samples, for samples after 1 h of CO 2 RR and in the final state (after 7 h, 2.5 h, and 2 h of CO 2 RR for Cu 50 Zn 50 , Cu 30 Zn 70 , and Zn 100 NPs, correspondingly) are shown in Figure 5a (Cu-K-edge) and Figure S8a (Zn-K-edge). In agreement with the XANES data analysis, the FT-EXAFS spectra of the as-prepared samples reveal the presence of M−O (M = Cu and Zn) structural motifs, as evidenced by a strong peak at ca. 1.5 Å (phase uncorrected). Overall, the FT-EXAFS spectra for the Cu K-edge in the as-prepared samples are close to those of Cu(OH) 2 .
Under CO 2 RR conditions, the M−O peak disappears completely in the Cu K-edge data, while a strong peak at ca. 2.1−2.2 Å appears, which can be associated with the presence of M−M bonds. Thus, complete reduction of cationic Cu species in all CuZn samples is observed, in agreement with XANES data. Note here that since Cu and Zn are neighbors in the Periodic Table we cannot easily distinguish between Cu− Zn and Cu−Cu (or Zn−Zn and Zn−Cu) contributions. By comparing spectra obtained after 1 h of CO 2 RR with the final spectra obtained after several hours of CO 2 RR, small variations can be observed (more pronounced in the Zn K-edge data, Figure S9a), suggesting some gradual structural changes not only in the Zn but also in the Cu atomic environment.
For quantitative analysis, EXAFS data fitting was performed. For samples under CO 2 RR conditions, Cu K-edge EXAFS fitting results are summarized in Figure 5b and in Table S2. The fitting results of the Cu and Zn K-edge EXAFS data of the as-prepared samples are presented in the Figure S9b−d and Tables S3 and S4. The main finding from the Cu K-edge EXAFS data analysis is that the Cu−M interatomic distance is different in Cu 50 Zn 50 (2.58 ± 0.01 Å after 1 h under CO 2 RR conditions) as compared to the Cu 30 Zn 70 (2.617 ± 0.008 Å) NP sample (Table S2 and Figure 5b). Both distances are in between those of a Cu foil (2.56 Å) and a brass CuZn foil (2.62 Å), suggesting alloying of Cu with Zn. The increase of the interatomic distances in fcc-type Cu−Zn alloys with increasing Zn concentration is in agreement with X-ray diffraction data. 57 Moreover, the interatomic distances change with time.  Zn 50 NPs. In both cases the interatomic distances get closer to that in a CuZn brass foil. These time-dependent changes in interatomic distances can explain the observed time-dependent changes in the XANES spectra (Figure 4d) as well as the switch in the selectivity. Note that the sensitivity of the Cu Kedge XANES spectra to interatomic distance changes has been demonstrated in our recent work. 58 This allows us now to track the changes in the catalyst structure with better time-resolution than what was possible based on the EXAFS analysis. By combining insights from EXAFS and XANES data analyses, we conclude that the interaction (alloying) between the Cu and Zn atoms gradually takes place under CO 2 RR conditions, and that it changes the local structure around the Cu atoms, making it more similar to that of bulk CuZn brass. Importantly, such changes took place in both samples (Cu 50 Zn 50 and Cu 30 Zn 70 NPs), but the local structure around Cu in the Znrich sample after 1 h of CO 2 RR was already closer to that in the brass CuZn foil, as suggested by the larger Cu−M distance for this sample. Therefore, this result implies that the ratio of the Cu and Zn elements and their oxidation state is critical in determining both, the alloying degree immediately after the onset of the CO 2 RR, as well as its evolution as a function of time.
Finally, based on the analysis of operando XAFS data we have observed a correlation between the product selectivity of the CuZn NPs and the reduction of the ZnO species and concomitant Cu−Zn alloy formation. It should be noted that there are a number of commonalities between the CO 2 electrochemical reduction and the CO 2 hydrogenation processes also taking place over Cu/ZnO catalysts, where the formation of brass might occur under certain reaction conditions and affect the catalytic selectivity. Since alloying in bimetallic systems induces the change of the geometric (or strain effect) structure, this transformation can lead to the tuning of the product selectivity in CuZn NPs. According to a recent experimental and theoretical study, an expansive strain in the Cu lattice should result in a more favorable stabilization of the reaction intermediates, leading to the formation of products beyond *CO. 49 This expansive Cu strain could explain the enhanced CH 4 selectivity in the Cu-rich CuZn NPs. Alloy formation, however, results also in a ligand effect and modification of the electronic structure, e.g., in a change of the d-band center of the metal atoms. 28,29 When upon reduction of the ZnO species a CuZn alloy is formed, the dband center of Cu is shifted away from the Fermi level. 59 Such shift results in a weakening of the binding strength of *CO due to the occupancy of antibonding states. 18,60 Therefore, product selectivity in the Zn-rich CuZn NPs seems to be controlled by the ligand effect rather than strain effects, leading to the loss of the functionality of the Cu site for further reduction of CO intermediates.
Importantly, our time-dependent XAFS analysis revealed that alloying between Cu and Zn atoms gradually takes place upon reduction of the cationic Zn species. This fact explains why the selectivity of CO 2 RR for CH 4 deteriorates over time. In fact, our durability data acquired for the Cu 50 Zn 50 NPs showed that the concentration of the initially produced CH 4 gradually decreased after 1 h reaction (Figure 3c). In parallel,

Journal of the American Chemical Society
Article the concentration of CO slightly increased. This is assigned to the Cu−Zn alloy formation and the loss of the Cu site ability to hydrogenate adsorbed *CO. This result highlights the critical correlation of the structure and composition of the NP surface and the CO 2 RR selectivity. Furthermore, it also implies that maintaining the metallic Cu-ZnO interface is vital for the stable production of CO 2 products, which can be affected by the reduction of the cationic Zn species in the CuZn alloyed catalysts. In conclusion, the product selectivity of CuZn NPs is determined by a combination of geometric and ligand effects, which become more or less prominent depending on the degree of alloying of the Cu−Zn species. In the first few hours, when the contribution of the Cu−Zn alloy is not significant, CH 4 formation over Cu-ZnO NPs is observed. Selectivity, however, switches to the formation of a CO+H 2 mix when brass is formed under CO 2 RR conditions. 36 Considering that our CuZn NP system was prepared as a model catalyst system, CH 4 and CO are the products to be expected when Cu is in contact with ZnO species within a nanoparticle. It is however known from previous studies that small Cu NPs favor CO production (and H 2 ) over hydrocarbons and alcohols. 61 Therefore, a comparison with other data in the literature on Cu−Zn systems must address two aspects: (i) whether a size effect should be considered and/or (ii) if the investigation has been conducted on a Cu-ZnO system or Cu−Zn alloy. Interestingly, several previous papers have reported the production of C 2 products such as ethylene and ethanol in CuZn catalysts containing Cu/ZnO or CuOx/ ZnO species. 33,34 Others however reported only CO and H 2 production when the samples investigated consisted of Cu−Zn alloys already in the as-prepared state. 36 Our data serve to consolidate the different findings available in the Cu−Zn literature, in particular, to clarify that for bulk-like systems (films and large NPs), Cu−Zn alloys will lead to the exclusive production of CO and H 2 , while dealloyed Cu/ZnO systems might also result in other products such as CH 4 for our NPs or even C 2+ products for larger material systems.

CONCLUSION
In summary, we have demonstrated that the product selectivity of CuZn NPs can be correlated with the initial Cu/Zn ratio as well as with the degree of alloying between the Cu and Zn components, which was found to evolve under CO 2 RR conditions. In particular, time-dependent XAFS data revealed that metallic Cu in close proximity to ZnO leads to the production of CH 4 at the initial stages of the reaction. The progressive reduction of the ZnO species taking place under CO 2 RR conditions and the concomitant enhanced Cu−Zn interaction and brass alloy formation lead to the switch in the selectivity to the exclusive generation of CO and H 2 . Our operando spectroscopy study provides crucial information on the nature of the active species and structure of CuZn NP catalysts. We believe that our findings can help to guide the rational design of bimetallic NP catalysts for CO 2 RR.