Operando insights into correlating CO coverage and Cu–Au alloying with the selectivity of Au NP-decorated Cu2O nanocubes during the electrocatalytic CO2 reduction

Electrochemical reduction of CO2 (CO2RR) is an attractive technology to reintegrate the anthropogenic CO2 back into the carbon cycle driven by a suitable catalyst. This study employs highly efficient multi-carbon (C2+) producing Cu2O nanocubes (NCs) decorated with CO-selective Au nanoparticles (NPs) to investigate the correlation between a high CO surface concentration microenvironment and the catalytic performance. Structure, morphology and near-surface composition are studied via operando X-ray absorption spectroscopy and surface-enhanced Raman spectroscopy, operando high-energy X-ray diffraction as well as quasi in situ X-ray photoelectron spectroscopy. These operando studies show the continuous evolution of the local structure and chemical environment of our catalysts during reaction conditions. Along with its alloy formation, a CO-rich microenvironment as well as weakened average CO binding on the catalyst surface during CO2RR is detected. Linking these findings to the catalytic function, a complex compositional interplay between Au and Cu is revealed in which higher Au loadings primarily facilitate CO formation. Nonetheless, the strongest improvement in C2+ formation appears for the lowest Au loadings, suggesting a beneficial role of the Au–Cu atomic interaction for the catalytic function in CO2RR. This study highlights the importance of site engineering and operando investigations to unveil the electrocatalyst's adaptations to the reaction conditions, which is a prerequisite to understand its catalytic behavior.

The pH of CO 2 sat.electrolyte is pH 6.4, E 0 of the used Ag/AgCl electrode was 200 mV.
The Faradaic Efficiency of each gas product x ( ) was calculated as with : geometric area of the electrode (cm 2 ), : volume fraction of the product x detected by the GC,

Inductively Coupled Plasma -Mass Spectrometry (ICP-MS)
The catalyst concentration, as well as the atomic compositions of Cu and Au were determined by ICP-MS (Thermo Fisher iCAP RQ).The samples were digested by adding a mixture of acids (1:1:3 H 2 SO 4 :HNO 3 :HCl) into a known amount of the catalyst and heated to 180°C for 30 min using the digestion Microwave Multiwave GO from Anton Paar.Samples from the electrode were digested with the carbon paper, which was discarded afterwards.The stock solutions, samples from the electrode and the electrolyte samples were diluted 3.33, 19 and 4 times in 3% HCl, respectively.

Transmission electron microscopy and energy-dispersive X-ray spectroscopy
The acquisition of scanning transmission electron microscopy (STEM) images and energy-dispersive Xray spectroscopy (EDXS)maps were performed with a FEI Talos F200X microscope equipped with a XFEG field emission gun and operated at an acceleration voltage of 200 kV.STEM images were acquired using a high angle annular dark field (HAADF) detector, while the EDXS data were recorded using the Super-X 4 quadrant silicon drift detector (SDD) system.All data were acquired using Thermo Fisher Scientific Velox software.For atomic resolution imaging and STEM-EDXS elemental mapping, the measurements were performed on a double aberration-corrected FEI Titan 60-300 TEM, operated at 200 kV under HAADF-STEM conditions.EDXS data were similarly acquired using a Super-X system and Velox software.For the quantified elemental and ratio maps, the EDXS map data were binned and processed using inhouse scripts based on the HyperSpy Python library.The quantification was based on k-factors generated with Velox, using the Brown-Powell ionization cross-section model.For the STEM analyses, the as prepared catalysts were drop-casted directly on a Ni lacey carbon grid, while the catalyst after CO2RR was removed from the carbon paper by sonicating it for a short time in 200 µl isopropanol.40 µl of the obtained solution was then drop casted on a Ni lacey carbon grid.

X-ray Diffraction
A Bruker AXS D8 Advance diffractometer in Bragg-Brentano geometry was used for X-ray diffraction (XRD) measurements with Cu K 1+2 radiation and a position sensitive energy dispersive LynxEye XE-T silicon strip detector.XRD patterns were measured in continuous scanning mode in the range between 20 and 100° 2, with an increment of 0.02° and a counting time of 1 s/step.
Operando high-energy XRD experiments were performed at beamline ID31 (ESRF, Grenoble).A homemade three electrode cell was based on the thin film approach with continuous electrolyte used equipped with a leak-free Ag/AgCl reference electrode and a Pt mesh counter electrode.CO 2 -saturated 0.1M KHCO 3 was continuously flown through the spectroelectrochemical cell.An X-ray energy of 67 keV and the working distance of the Dectris Pilatus CdTe detector was calibrated using a CeO 2 reference material.The 2D diffraction pattern were integrated using the pyFAI software package and Rietveld refinement performed using TOPAS (Bruker-AXS, v6).The sample was deposited on highlyoriented pyrolytic graphite electrodes with a loading of ~0.1 mg/cm².The diffraction pattern were recorded in grazing-incidence configuration with the incidence angle optimized for best sample to substrate signal ratio.Rietveld refinements using the software package TOPAS ® (Bruker-AXS) were performed for analysis considering the instrumental broadening of the lab diffractometer, zero error and a sample displacement.

Quasi in situ X-ray photoelectron spectroscopy
Quasi in situ X-ray photoelectron spectroscopy (XPS) measurements were performed in an ultrahigh vacuum (UHV) chamber, geared with a commercial Phoibos100 analyser (SPECS GmbH, E pass = 15 eV) and a XR50 (SPECS GmbH) X-ray source with an Al anode (E Kα =1486.7 eV).The spectra were aligned using Cu 0 (932.67 eV) as reference and fitted using a Shirley-type or a linear background subtraction for X-ray photoelectron or Auger electron spectroscopy, respectively.Quasi in situ XPS experiments were performed in a one compartment cell, which was directly attached to the UHV chamber.After CO 2 RR, the catalyst was washed with Ar-sat.H 2 O to remove residual electrolyte and transferred quickly into UHV under Ar atmosphere to avoid exposure to air and the possible subsequent reoxidation.The electrochemical measurements were carried out using a potentiostat (Autolab PGSTAT 302N, Metrohm).

Operando X-ray absorption fine-structure spectroscopy
Operando X-ray absorption fine-structure spectroscopy (XAFS) measurements were performed at beamlines located at the synchrotron facilities ALBA (CLAESS beamline), SSRL (BL 2-2 beamline) and SOLEIL (SAMBA beamline) as well as the quick X-ray absorption fine structure (QXAFS) beamline (SuperXAS) at SLS synchrotron facility of the Paul Scherrer Institute, respectively.All experiments were conducted in fluorescence mode at the Cu K-edge (8978.9eV) and Au L 3 -edge (11918.7 eV) with corresponding fluorescence detectors (SI).The operando measurements were performed in a threeelectrode electrochemical cell (see Ref. 4 for the schematics of the cell) matching the conditions of the selectivity studies.A leak-free Ag/AgCl was used as a reference electrode, while an Pt mesh was used as a counter electrode.The samples were prepared by drop casting 0.25 mg and 10 mg of catalyst on 0.5 cm 2 area of carbon paper with a microporous layer (GDE, Sigracet 39b).Cu K-edge and Au L 3 -edge data were collected separately for identical fresh samples with different loadings to optimize the absorption edge signal while avoiding self-absorption.The carbon paper with the deposited catalyst served as a working electrode.It was mounted in the electrochemical cell and fixed with Kapton tape, so that the Kapton-covered carbon paper could act as an X-ray window, while the side coated with the catalyst was in contact with the electrolyte.The measurements for both samples were performed exsitu as well as under operando conditions.Energy calibration, background subtraction and normalization of the collected X-ray absorption near-edge (XANES) spectra were performed with a set of home-built Mathematica scripts.The Athena software was used to extract the extended x-ray absorption fine structure (EXAFS). 5The FEFFIT code was used for EXAFS fitting. 6or Au L 3 edge quick XAFS (QXAFS) species were tracked every 1 s and every 100 spectra was averaged to improve the signal quality for Au 2.7 /Cu 2 O NC, while for Au 0.4 /Cu 2 O NC and Au 0.8 /Cu 2 O NC, spectra were collected every 12 min and every two spectra were merged.

Operando surface-enhanced Raman spectroscopy
For operando surface-enhanced Raman spectroscopy (SERS) measurements, a Raman spectrometer (Renishaw, InVia Reflex) equipped with an optical microscope (Leica Microsystems, DM2500M), a motorized stage for sample tracking (Renishaw, MS300 encoded stage), a near-infrared laser (Renishaw, RL785, λ = 785 nm, Pmax = 500 mW), a CCD detector (Renishaw, Centrus) and a water immersion objective (Leica microsystems, 63x, numerical aperture of 0.9), was used.The water immersion objective was covered with a Teflon film (DuPont, 0.013 mm film thickness) to protect it from the electrolyte.A Si(100) wafer (520.5 cm -1 ) was used for calibration.The Raman scattering of the Rayleigh-filtered backscattered light was collected in between 100 -3200 cm -1 with a grating of 1200 lines mm -1 .Electrochemical measurements were performed following a previous report. 7The electrochemical cell was equipped with a Pt counter electrode and a leak-free Ag/AgCl reference electrode; the catalyst was drop-casted on a glassy carbon plate, connected from the back side to the circuit.Measurements were performed with a Biologic SP240 potentiostat.CO 2 -saturated 0.1 M KHCO 3 was used as electrolyte.Spectrum collection was performed with 10 s exposure time.Focus optimization was done by depth scans.Steady-state conditions at the catalyst surface was ensured by waiting at least 10 min before collecting the spectra.Renishaw WiRE 5.2 software was used to baseline-subtract the data with the intelligent spline feature (8 th polynomial order) and to remove cosmic rays.
To gain insight over the stability of the catalysts, we performed long-term measurements over 20 h at -1.03 V and tracked the changes of the FEs, Figure S11.The initial product distribution is comparable with the data in Figure 2 that were determined after 1 h of CO 2 RR.For all catalysts, after the initial activation and stabilization, the gaseous product distributions remain stable over the course of 20h.Nevertheless, a slight decrease in CO formation is observed for Au 0.4 /Cu 2 O NCs and Au 0.8 /Cu 2 O NCs, while CO production remains stable for Au 1.1 /Cu 2 O NCs and increases for Au 2.7 /Cu 2 O NCs, Figure S11c, suggesting a more sluggish catalyst restructuring for low Au loadings.The total liquid products, analyzed after 20 h of CO 2 RR, displays a decrease of the total amount of liquids, Figure S11f.The decrease in the total amount of liquid products is suggested to be observed due to the high polarity of the oxygenates and alcohols during the whole measurement time.
Figure S12 additionally shows a comparison of the liquid products after 1h and after 20h, showing an increase in product distributions for the Cu 2 O NCs, while the product distributions vary upon the addition of Au.In particular, the ethanol and allylalcohol formation is declines by roughly 4 percentage points, while the propanol formation increases slightly.Interestingly the formation of acetaldehyde and propionaldehyde, Figure S12e and f appears mostly during the first hour of CO 2 RR.

Supplementary Note 2: Electrochemical characterization
The catalysts were characterized electrochemically by cyclic voltammetry in CO 2 sat.KHCO 3 and Ar sat.NaOH (Figure S13) after reduction of the catalyst for 1 h.All catalysts show the characteristic peaks for Cu 0  Cu l and Cu 0  Cu ll oxidation as well as the Cu ll  Cu l and Cu l  Cu 0 reduction, respectively.Note that the untypical broad reduction peak from Cu ll  Cu l overlaps with the thick oxide layer that was produced at high oxidizing potentials.The upper limit was chosen to eventually oxidize Au, which we did not observed in our CVs.In CO 2 sat.KHCO 3 , no difference in its redox behavior was found compared to Cu foil.In contradiction, a inhibited oxidation process and a positively shifted Cu oxidation towards higher potentials had been suggested. 8,9 e assign this contradiction to our results to the low amount of alloyed catalysts and to the unordered type of alloy formation.
In Ar sat.NaOH, OH adsorption is observed at 0.63 V vs. RHE for the catalysts and is lacking in the CV for the electropolished Cu foil.Furthermore, the shape of the oxidation peaks shows much broader shapes than the electropolished Cu foil.Double layer capacitance was measured for the catalysts after 1 h of CO 2 RR and compared to the Cu foil to retrieve a roughness factor, Table S6.The catalysts with Au NPs show higher capacitances and roughness factors than the Cu 2 O NCs, which suggests a higher surface area due to the presence of Au.However, metallic Au species and CuAu alloys also contribute to the capacitance, which impedes a clear assignment.S4.             and d) H 2 ; e) current transients and f) liquid products of the catalysts.The gaseous products were analyzed every 15 min using an online GC, the liquid products were analyzed after the 20 h measurement was finished using a liquid GC and a HPLC.The region between 1000 cm -1 and 1700 cm -1 holds vibration peaks of HCO 3 -(1005 cm -1 ) and CO 3 2-(1072 cm -1 ), and from the glassy carbon support (1313 cm -1 and 1600 cm -1 ). -

FigureFigure S3 .
Figure S1.Overview STEM-HAADF micrographs of the catalysts in the as prepared state for a) Cu 2 O NC, b) Au 0.4 /Cu 2 O NC, c) Au 0.8 /Cu 2 O NC, d) Au 1.1 /Cu 2 O NC and e) Au 2.7 /Cu 2 O NC.

Figure S4 .
Figure S4.Atomic resolution, aberration-corrected HAADF STEM micrographs of single grain and multigrain Au NPs of Au 2.7 /Cu 2 O NC.

Figure S5 .
Figure S5.Standard resolution and aberration-corrected atomic resolution STEM-HAADF micrographs of Au 2.7 /Cu 2 O NC a) in the as prepared state and b) after 1 h CO 2 RR at -1.0 V vs. RHE.

Figure S6 .
Figure S6.Standard resolution and aberration-corrected atomic resolution STEM-HAADF images of Au 0.4 /Cu 2 O NC a) the unreacted sample, b) unreacted stable Au nanoparticles, c) reacted sample, d) coalescence of Cu and e) dissolved Cu paralleled with f) Au wetting phenomenon.

Figure S7 .Figure S8 .
Figure S7.STEM-HAADF micrographs and their corresponding binned and quantified EDXS maps forAu, Cu and O as well as the Au/Cu ratio for all catalysts in the as prepared state and after CO 2 RR conditions.

Figure S10 .
Figure S10.Partial current densities, normalized by the electrochemical surface area from DL capacitance measurements at -1.07V vs. RHE in 0.1 M KHCO 3 as a function of the Au NP loading for a) H 2 , CO, CH 4 , C 2 H 4 , b) C 2+ liquids, C 2+ carbonyls and minor liquid productions, c) the total C 2+ products and total C 2+ liquid products

Figure S11 .
Figure S11.Stability measurements over 20 h at -1.03 V. Faradaic efficiencies of a) C 2 H 4 , b) CH 4 , c) COand d) H 2 ; e) current transients and f) liquid products of the catalysts.The gaseous products were analyzed every 15 min using an online GC, the liquid products were analyzed after the 20 h measurement was finished using a liquid GC and a HPLC.

Figure S17 .
Figure S17.Rietveld refinement of HE-XRD pattern acquired at OCP and during CO 2 RR as well as the fitted profiles (pink and red, respectively) of a) Cu 2 O NCs, b) Au 0.4 /Cu 2 O NCs, c) Au 1.1 /Cu 2 O NCs, and d) Cu 2.7 /Cu 2 O NC.The difference between experimental data and the fitted profile are shown in grey below and above the pattern recorded at OCP and during CO 2 RR, respectively.The X-ray energy was set to 67 keV.

Figure S18 .
Figure S18.Fourier-filtered Cu K-edge EXAFS spectra in k-space of the Au x /Cu 2 O NC in a) the asprepared state and b) at -1.0 V. Reference spectra of Cu 2 O and Cu are shown for comparison.

Figure S19 .
Figure S19.Time dependent Cu K-edge XANES spectra of a) Au 0.4 /Cu 2 O NC, b) Au 1.1 /Cu 2 O NC and c) Au 2.7 /Cu 2 O NC during CO 2 RR at -1.0 V vs. RHE and their corresponding time-resolved results of linear combination fitting of the XANES spectra, using spectra for Cu foil, Cu 2 O and CuO as references.The samples were measured at OCP before CO 2 RR conditions were applied at time 0 s.

Figure S20 .
Figure S20.Time dependent Au L 3 -edge XANES spectra of a) Au 0.4 /Cu 2 O NC, b) Au 0.8 /Cu 2 O NC and c) Au 2.7 /Cu 2 O NC during CO 2 RR at -1.0 V vs. RHE.Reference data from a bulk Au-foil are also shown.

Figure S21 .
Figure S21.Operando surface-enhanced Raman spectra of a) Au 0.4 /Cu 2 O NC, b) Au 0.8 /Cu 2 O NC, c) Au 1.1 /Cu 2 O NC and d) Au 2.7 /Cu 2 O NC in CO 2 sat.0.1 M KHCO 3 with stepped potentials from open circuit potential to -1.1V vs. RHE.Key species are identified at 280 cm -1 and 366 cm -1 , which correspond to the restricted rotation of *CO on Cu (CO rot ) and Cu-CO stretching (CO stretch ), respectively, as well as CO stretching band at 2090 cm -1 .

Figure S23 .
Figure S23.CO stretch /CO rot ratios of the respective bands (277 cm -1 and 360 cm -1 ) as a function of applied potential (iR corrected).The results for Cu 2 O were calculated from Ref8 .

Cu l Cu ll Au 2.7 /Cu 2 O NC a) Cu LMM, as prepared Au 0.4 /Cu 2 O NC Au 0.8 /Cu 2 O NC Au 1.1 /Cu 2 O NC Cu 2 O NC Au 2.7 /Cu 2 O NC Intensity (arb. units) Kinetic Energy / eV Cu 0 Cu l b) Cu LMM, after CO 2 RR Au 0.4 /Cu 2 O NC Au 0.8 /Cu 2 O NC Cu 2 O NC Au 1.1 /Cu 2 O NC Figure
S14. Quasi in situ Cu LMM Auger spectra of all catalysts in the a) as prepared state and b) after CO 2 RR in 0.1 M KHCO 3 .

Table S1 .
Composition of Cu and Au of the presented catalyst dispersions, determined by ICP-MS.

Table S2 .
Coherence lengths, lattice parameters and atomic fractions of Cu 2 O and Au extracted from Rietveld refinement of the ex situ XRD pattern of all catalysts in the as prepared state.

Table S3 .
Edge lengths of the Cu NCs and diameters of the Au NP in the as prepared state and after 70 min CO 2 RR obtained from STEM-HAADF micrographs.

Table S4 .
Compositions measured using EDXS comparing the Cu and Au ratios of the catalysts in the as prepared state and after 70 min CO 2 RR.Evaluation based on L lines of the elements.

Table S5 .
ICP-MS analysis of the Au:Cu ratio of the catalysts on 2 cm 2 carbon paper electrode in the asprepared state and after 1 h of CO 2 RR at -1.0 V, and of the electrolyte after reaction.

Table S6 .
Double Layer Capacitances and corresponding roughness factors (normalized to Cu foil) of the catalysts with Nafion and of the reference Cu and Au foils measured after 1 h CO 2 RR at -1.0V vs. RHE.

Table S7 .
XPS composition between Cu, Cu 2 O and CuO obtained by linear combination fitting of the Cu LMM spectra in the as prepared state and after 1 .

Table S8 .
XPS composition of Cu (all chemical states) and Au in the as prepared state and after 1 h under CO 2 RR conditions comparing Cu 3p with Au 4f peaks

Table S9 .
Coherence lengths, lattice parameters and atomic fractions of Cu 2 O and Au extracted from Rietveld refinement of the operando XRD pattern of all catalysts in OCP and under CO 2 RR conditions.

Table S10 .
Structural parameters (coordination number N, interatomic distances R, disorder factors σ 2 ) obtained from fitting the Cu K-edge EXAFS data acquired for Au-decorated Cu2O in the as prepared state and during CO 2 RR at -1.0 V vs. RHE.Correction to photoelectron reference energy ∆E and the obtained R-factor that characterizes fit quality are also reported.Uncertainty of the last digit is reported in parentheses.

Table S11 .
Structural parameters (coordination number N, interatomic distances R, disorder factors σ 2 ) obtained from fitting the experimental Au L 3 -edge EXAFS data acquired in the as prepared state and during CO 2 RR at -1.0 V vs. RHE.Correction to photoelectron reference energy ∆E and the obtained R-factor that characterizes fit quality are also reported.Uncertainty of the last digit is reported in parentheses.

Table S12 .
Register of the most important Au-Cu and Ag-Cu catalysts forming C 2+ products under CO 2 RR.