Interfacial Chemistry in the Electrocatalytic Hydrogenation of CO2 over C-Supported Cu-Based Systems

Operando soft and hard X-ray spectroscopic techniques were used in combination with plane-wave density functional theory (DFT) simulations to rationalize the enhanced activities of Zn-containing Cu nanostructured electrocatalysts in the electrocatalytic CO2 hydrogenation reaction. We show that at a potential for CO2 hydrogenation, Zn is alloyed with Cu in the bulk of the nanoparticles with no metallic Zn segregated; at the interface, low reducible Cu(I)–O species are consumed. Additional spectroscopic features are observed, which are identified as various surface Cu(I) ligated species; these respond to the potential, revealing characteristic interfacial dynamics. Similar behavior was observed for the Fe–Cu system in its active state, confirming the general validity of this mechanism; however, the performance of this system deteriorates after successive applied cathodic potentials, as the hydrogen evolution reaction then becomes the main reaction pathway. In contrast to an active system, Cu(I)–O is now consumed at cathodic potentials and not reversibly reformed when the voltage is allowed to equilibrate at the open-circuit voltage; rather, only the oxidation to Cu(II) is observed. We show that the Cu–Zn system represents the optimal active ensembles with stabilized Cu(I)–O; DFT simulations rationalize this observation by indicating that Cu–Zn–O neighboring atoms are able to activate CO2, whereas Cu–Cu sites provide the supply of H atoms for the hydrogenation reaction. Our results demonstrate an electronic effect exerted by the heterometal, which depends on its intimate distribution within the Cu phase and confirms the general validity of these mechanistic insights for future electrocatalyst design strategies.

: XPS elemental analysis  Figure S1: Raman Spectroscopy on Cu/G  Figure S2: Scanning electron micrographs  Figure Table S2: Results of Linear combination fits on XANES and EXAFS data for Cu/G and CuZn/G in 0.1 M KHCO 3 2.3.2. K-edge operando spectroscopy study of CuZn/G  Table S3: Evolution of Cu species during CV shown by results of LCF fits on XANES and EXAFS signal for CuZn/G (Cu K-edge).  Table S4: Results of Linear combination fits on XANES and EXAFS data for CuZn/G at Cu and Zn K-edges at constant voltages 2.3.3. K-edge operando spectroscopy study of Cu/G  Figure S7: Operando Cu K edge XANES for Cu/G under potentiostatic control at different potential applied sequentially as indicated.  Table S5: Results of Linear combination fits on XANES and EXAFS data for Cu/G at constant voltages  Figure S8 SEM after operando Study 2.4. Soft X-ray operando spectroscopy 2.4.1. Electrocatalytic performance determined during operando studies  Table S6: Average currents exchanged during the chronoamperometry under steady state at each voltage applied  Figure S9: (a) Cyclic voltammograms recorded in the in situ cell, preceding the in situ spectroscopic measurements for the electrocatalysts; mass spectrum during operando study for: (b) Cu/G; CuZn/G (c) and (CO 2 RR active state) CuFe/G-1 (d).  Figure S10: Mass spectrum recorded during the chronoamperometry study for sample CuFe/G under HER selective state (a), and relative Cu L3-edge spectrum (b). 2.4.2. Cu L-edges and O and C K edges NEXAFS operando studies  Figure S11: Soft X-ray in situ NEXAFS data under a stagnant KHCO 3 electrolyte at different voltages as indicated: Cu L 3 -edge a), O K-edge b) and C K-edge c) spectra for Cu/G.  Figure S12 SEM after operando Study 2.5. Cu L-edges NEXAFS simulation  Figure S13: Comparative plot of experimental and simulated Cu L 3 Absorption edges for reference materials a) Cu metal, b) Cu 2 O and c) CuO. Simulated spectra have been energy-shifted to overlap with experimental ones  Figure S14: Simulated Cu L 3 XANES intensity at the 2p-3d transitions, showing the decrease in intensity of the feature due to the lowest unoccupied d-and s-type electronic states  Figure S15: ( Table S8: Simulated Mulliken population following the introduction of either Zn or Fe (upper and lower panel, respectively) in supercell of Cu(OH) 2 and relaxation of the geometry of the unoccupied orbitals, and consequently the transition probability and spectral intensity 2.6. Computational study 2.6.1. ZnO/Cu model  Figure S16-S39 2.6.2. CuZn alloy model  Figure S40-47  Table S9 1. EXPERIMENTAL PART

1.1.1.XPS and NEXAFS measurements
X-ray photoelectron spectroscopy (XPS) and ambient pressure near edge X-ray absorption fine structure (NEXAFS) measurements in the soft X-ray regime were carried out at the ISISS beamline at Helmholtz-Zentrum Berlin (HZB). The freshly prepared samples were directly exposed to vacuum (10 -7 mbar) in the XPS chamber. XPS measurements were performed applying a suitable excitation energy corresponding to a kinetic energy (KE) of the photo-emitted electrons of 450 eV for the core levels NEXAFS spectra in total electron yield (TEY) were performed using a Faraday cup placed in close proximity to the sample in the APXPS chamber.
The beam-line setting was exit slit (ES) 60 m and fix focus constant (cff) 2.25 (cff=cos/cos). The exit slit value chosen enables an optimal compromise between high photon intensity and good spectral resolution.
The Cu L-edges and O K-edge spectra were processed by subtracting a linear background fitting the pre-edge absorption (between 527 and 530 for Cu L edges and between 525 and 529 eV) and normalizing by the edge height (mean value between 540 and 570 eV for the O K and between 970-980 eV for the Cu L-edges). The C K-edge spectra were processed by subtracting a linear background fitting the pre-edge absorption region (between 283 and 284 eV) and normalizing by the C K-edge height at the peak at 285.7 eV on the assumption that this component is the C=C bond resonance of the graphitic peak of the support, which remains unchanged during the measurements. The energy scale of for each spectrum was calibrated using features of the refocusing mirror drain current recorded simultaneously to the spectrum. No further calibration was performed using references spectra. The Cu L-edges NEXAFS spectra were fitted with a linear combination analysis of reference spectra for CuO, Cu 2 O and Cu using Igor Software.

1.1.2.XAFS measurements and Electrochemical cell for operando study
X-ray absorption experiments (EXAFS and XANES) were performed at the B18 Core EXAFS beamline of Diamond Light Source. The measurements were carried out using the Pt-coated branch of collimating and focusing mirrors, and a Si(111) double-crystal monochromator. A couple of Pt-coated harmonic rejection mirrors were inserted before the first ion chamber and used to filter out photons with higher energy. The size of the beam at the sample position was ca. 1 mm (h) × 1 mm (v). The data were collected in fluorescence mode, by means of a 36-element solid state germanium detector (K max =14), the ion chamber before the sample has been used for measurement of incoming photons (I0 filled with a mixture of 30 mbar of Ar and 1080 mbar of He to optimize sensitivity at 20% efficiency).
Samples were measured both in static and operando conditions. The design of the operando XAFS cell used is this work was reported earlier. 1 The cell is filled with 0.1 M KHCO 3 electrolyte and CO 2 is continuously supplied through an inlet placed though a leak-tight orifice in the lid of the cell and immersed into the liquid electrolyte.
Then, Cu K-edge spectra were measured first while running CV scans and then at fixed constant potentials successively applied in the order indicated for each cases.  Data were normalized using the Athena 2 program with a linear pre-edge and polynomial post-edge background subtracted from the raw data. All XANES data were fitted with linear combination analysis using relevant spectra as reference. A linear combination fit (LCF) was performed on both Cu K XANES and EXAFS regions of the spectra measured during the voltage sweep experiments as well as the experiments at constant voltages. Fits were performed with Athena in the -20 to +80 eV range using relevant recorded spectra as reference, to describe variation in sample composition. Linear combination analysis on EXAFS data was performed in the range (2.5 to 10 Å -1 ).
EXAFS fits were performed using ARTEMIS software. 2 Moreover, the interatomic distances and Debye-Waller factors were optimized by fitting the experimental data.

SEM
Scanning electron microscopy (SEM) and energy dispersive X-ray (EDX) analyses were performed on a Zeiss Ultra SEM operating at acceleration voltages of 1.6 and 15 kV.

Procedure for the electrochemical reduction of CO 2 in liquid phase
A compact-design electrochemical cell was used for the electrochemical reduction of CO 2 in liquid phase. 3 The cell has a three-electrode configuration: the working electrode (about 0.64 cm 2 ) was located at the cathode side, at a small distance from a saturated Ag/AgCl reference electrode to reduce the solution resistance. The electric contact with the working electrode was assured by a Pt wire. Sample were pre-treated before the experiments to eliminate impurities in the electrode and tests to check that the products detected derive from the electroreduction of CO 2 were made, as earlier described. 3 These preliminary experiments include tests with labelled CO 2 . The experiments were carried out at different voltages (-0.5/-2V vs Ag/AgCl range). We find that the optimal behaviour in terms of higher production rates of products from CO 2 conversion was at the applied potential of -1.38 V vs RHE that was then used in the following tests. As earlier commented, we also checked the use of alternative analytical methods such as by NMR (nuclear magnetic resonance) finding preferable the analytical method described above.   XPS spectra in Fig. S3a and b indicate that Cl is the main impurity in Cu/G, whereas S is present more abundantly in CuZn/G and to a lesser extent in CuFe/G.  Table S1 of the supporting information.

Ex-situ Characterization of as prepared electrocatalysts
Accordingly, the surface of Cu/G is composed of approximately 25.8% of Cu atoms on C. The addition of another metal by wet impregnation and the formation of larger particles due to this treatment produces a pronounced decrease of the Cu content within the volume probed by this technique.
The analysis of the Zn L-edges and Fe L-edges in Figure S3d-e and, provides further insights into the chemical speciation of the elements in these samples. We show that Zn in CuZn/G is present as Zn(II) species (sharp pre-edge feature at 1024 eV in the Zn L 3 -edge NEXAFS spectrum). Interestingly, the O K-edge NEXAFS spectrum of CuZn/G in Figure S4b shows several resonances denoted in Figure S4b as A, B and C. Whilst the resonance C is typical for Zn oxides, the resonances at 530 eV (resonance A) and at 532 eV (resonance B) are assigned to Cu(II) species and Cu(I) species, respectively 5, 6 and were reported for Cu-doping of ZnO thin films. 7 Consistently, the pre edge feature of the Cu(II) shifts ( Fig.   S3d) to lower energy indicating an intimate chemical interaction between the Cu(II) and the Zn(II) cations in the starting electrocatalyst. It is possible to infer that despite the Zn being impregnated on the surface of the Cu oxide nanostructures, the Cu-Zn chemical affinity allows for the Cu to be dispersed in the external Zn(II)-overlayer during the thermal annealing step. A closer inspection of the Zn L edges spectrum for CuZn/G enables to identify a high edge jump at 1037 eV, which was also reported for ZnS, consistent with S 2p spectrum (Fig. S3b). 8 It can be therefore assumed that the external overlayer is as a mixture of a Cu-Zn oxide, sulphate and sulphide phases.
In CuFe/G, Fe is present predominantly as a mixture of Fe(II) and Fe(III) species ( Figure S3e). Here we observe an almost quantitative reduction of Cu(II) sites ( Fig. S4a) suggesting for a predominant Fecontaining overlayer on the cuprous oxide nanoparticles in the form of an oxide-sulphate mixed phase.
As shown in the SEM images in Figure S2 c and d, this sample contains a portion of particles detached from the carbon support, which will not be available for electrocatalysis, due to the low contact with the conductive substrate. The electronic effect on the Cu exerted by the heteroatom are presented in in Figure S4a. Interestingly, the Cu(II) and Cu(I) transitions are found at a slightly different excitation energy for the three samples. This can be better evaluated by considering the energy difference between the two transitions, which is circa 2.46 eV for Cu/G and 2.75 eV for CuZn/G. This is due to the different bonding environment of Cu in the three samples.    (Fig. S6). Cu(0) and Cu(I) have no hole in the 3d states, whereas Cu(II) is a d 9 and therefore a weak quadrupole-allowed pre-edge peak is observed (peak A in Fig. S6a). 9 Figure S6: Fluorescent yield Cu K edge XANES for Cu/G (a) and corresponding k 2 -weighted FT EXAFS. Fluorescent yield XANES for CuZn/G at the Cu K edge (b) and Zn K edge (c) and corresponding k 2 -weighted Fourier transform EXAFS. XANES spectra of Cu and Zn standards are included weighted by the percentage found by LCF. Note that the measurements were taken upon immersion of the electrocatalysts in the liquid electrolyte solution and before any electrocatalytic tests were performed.
The Cu(I) is characterized by a dipole allowed 1s → 4p transition, which due to ligand field splits into two resonances B and C. 9 The EXAFS spectra and the linear combination fits (LCF) of the XANES for these samples using three components obtained from reference spectra (Cu metal,  (Fig. S6).
Cu(0) and Cu(I) have no hole in the 3d states, whereas Cu(II) is a d 9 and therefore a weak quadrupoleallowed pre-edge peak is observed (peak A in Fig. S6a). 9 The Cu(I) is characterized by a dipole allowed 1s → 4p transition, which due to ligand field splits into two resonances B and C. 9 The EXAFS spectra and the linear combination fits (LCF) of the XANES for these samples using three components obtained  The LCF is also used for the analysis of the Zn K-edge XAS. Accordingly, the spectrum resembles a mixture of ca. 33% of Zn sulphate distinguishable by an intense white line at 9670 eV and ca. 67% of Zn carbonate characterized by a pre-edge shoulder at 9666 eV and post edge feature at 9674 eV. The former one is a residue of the Zn precursor used, whereas the latter one is formed upon interaction with the bicarbonate solution. We can therefore conclude that in a KHCO 3 solution, under open circuit conditions, the electrodeposited Cu phase undergoes a surface carbonatation reaction, whose extent depends on the exposed surface area. In the case of CuZn/G, we assume that the larger cubo-octahedrons allows a lower Cu exposure and as a consequence the Cu 2 O phase remains mostly unaltered. Moreover, we observe that the Zn(II) sulphates species are converted to carbonates. The EXAFS analysis provides additional information on changes in bond lengths upon immobilization of Zn(II) species. The first shell in (R) Fourier transformed data ( Figure S6b), shows an evident shift of the peak position from 1.43 to 1.51 Å (phase uncorrected), corresponding to Cu-O bond lengths in Cu(I) and Cu(II) species. The second shell, related to metal-metal distances, has a lower intensity but shows an interesting splitting of the peak in the Cu-K edge spectra upon Zn addition: The Cu-Cu contribution in Cu oxide is accompanied by a shoulder at higher distance that could be assigned to the Cu-Zn interaction. This is confirmed by the FT of the Zn K-edge spectrum for the same sample ( Figure S6c) where a similar feature is present at the same interatomic distance, and can be assigned to the Zn-Cu contribution. The observation of an interaction between Cu and Zn atoms confirms the homogeneous distribution of both metals and excludes that the two elements are segregated in isolated phases. The electronic effect of Zn on Cu in the fresh electrocatalyst was also observed in the Cu L edges NEXAFS (Fig. S4a) and OK NEXAFS spectra Fig. S4b. A linear combination fit of the EXAFS k 2 -weighted (k) data (Table S2) confirms the results obtained by LCF of XANES spectra with similar percentages of the different fractions for the two samples analysed. a the spectra were recorded continuously while changing the potential and similar spectra were merged together and herein reported as a voltage range.   Table S3. We observe redox dynamic involving the Cu with the highest abundance of Cu(I) formed at -0.3 V vs Ag/AgCl whereas the maximum reduction of Cu(I) → Cu0 occurs at potential below -0.84 V vs Ag/AgCl.

K-edge operando spectroscopy study of CuZn/G
The fitted Cu K and Zn K edges XANES and EXAFS spectra measured at constant voltages are reported in Figure 4 of the main text. The quantitative analysis is summarized in Table S4. Consistent trends are observed in both the voltage sweep experiments and the constant potentials. We note that the composition of the electrocatalyst at low voltages depends on the preceding voltage applied and the time so the proportion of Cu(I) and Cu(0) changes in different runs at the same potential (Table S3).
This is a direct manifestation of the changes in particles morphology or/and size within the electrodes with time. The quantitative analysis of the spectra in Figure S7 is reported in Table S5. The main phase in Cu/G -2V vs AG/AgCl is a metallic Cu phase, similar to CuZn/G; however, part of the electrocatalyst is still present as Cu 2 O and Cu carbonate. The application of an anodic potential (+ 0.8 V vs Ag/AgCl) induces the oxidation of the metallic Cu to carbonate which is then reduced to Cu 2 O at -0.5 V vs Ag/AgCl, whereas Cu 0 is the dominant phase at -1 V vs Ag/AgCl.   CuFe/G _1 MS in Figure S9d Cu/G MS in Figure S9b - CuZn/G MS in Figure S9c CuFe/G _2 MS in Figure S10a