Plasma-Modified Dendritic Cu Catalyst for CO2 Electroreduction

Efficient and active catalysts with high selectivity for hydrocarbons and other valuable chemicals during stable operation are crucial. We have taken advantage of low-pressure oxygen plasmas to modify dendritic Cu catalysts and were able to achieve enhanced selectivity toward C2 and C3 products. Utilizing operando spectroscopic methods such as X-ray absorption fine-structure spectroscopy (XAFS) and quasi in situ X-ray photoelectron spectroscopy (XPS), we observed that the initial presence of oxides in these catalysts before the reaction plays an inferior role in determining their catalytic performance as compared to the overall catalyst morphology. This is assigned to the poor stability of the CuxO species in these materials under the conditions of electrocatalytic conversion of CO2 (CO2RR). Our findings shed light into the strong structure/chemical state-selectivity correlation in CO2RR and open venues for the rational design of more effective catalysts through plasma pretreatments.


■ INTRODUCTION
Due to the rapid advance in technology and simultaneous drastic increase of the overall population, the demand for energy is a concern of general societal interest. A promising strategy toward sustainable energy is the use of electricity derived from renewable energy sources to drive the electrocatalytic conversion of CO 2 (CO 2 RR) into valuable chemicals and fuels 1 that can be subsequently used as feedstock for the chemical industry or as energy storage technology. 2, 3 Among the various catalysts studied so far, 4−6 copper has been shown to have the unique ability to produce hydrocarbons, especially C 2 and C 3 compounds, 4,7 which is due to its moderate binding energy for CO 8 and optimum H binding. 9 The downside is that, unlike other metals such as Ag 10−12 and Au 13,14 or Sn 5,15 and Pb 15 that display high selectivity for CO or HCOO − , respectively, in the case of copper, it is difficult to target the desired products. 1,8,16,17 Various electrode parameters such as composition, 18−20 surface morphology and structure, 21 chemical state, 21−23 prefunctionalization 21,23 or electrolyte characteristics such as its composition 21,23−28 and pH, 29 as well as thermodynamic parameters like temperature 30 and pressure, 30,31 were found to drastically affect the catalysts performance. Nevertheless, a detailed picture of how the reaction takes place in a number of active Cu-based catalysts is missing.
Nanostructured catalysts have proven to outperform bulk materials 21,32 in many aspects of CO 2 RR. For instance, a superior catalytic performance with respect to ethylene production has been observed for Cu nanocubes and attributed to the presence of subsurface oxygen/Cu + species 21,23 and Cu(100) facets. 21,22,33 Additionally, kinetic transport limitations can be overcome by using high surface area nanoporous catalysts such as copper dendrites, which can be operated in a three-phase (gas/solid/liquid) 32,34 flow cell setup with an already demonstrated 63% faradaic efficiency (FE) for ethylene at current densities of up to 750 mA/cm 2 . 35 Other materials tested in similar flow cells were also found to provide excellent C 2+ selectivity at high current densities. 32,35,36 The exceptional C 2 H 4 selectivity previously reported for dendritic catalysts 32,36 was also shared by oxide-derived (OD) catalysts, which also displayed an increased selectivity for multicarbon products while suppressing methane formation. 23,28,37 Although a correlation between the presence of Cu(I) species and subsurface oxygen in Cu-based catalysts and their selectivity for C2−C3 products has been theoretically predicted 32,33 and experimentally demonstrated by several groups, 19,25,34,35 their small content and difficult stabilization makes their detection challenging and still controversial in the literature. It requires a synergistic combination of surface and bulk-sensitive in situ and operando methods able to probe surface and subsurface regions of the catalysts with high sensitivity. Furthermore, the morphology of the catalyst, 19,27,36 initial composition, 19,37,38 pretreatment for the generation of the oxides, 20 Cu particle support, 26,39 and Cu−electrolyte interactions 33,40 are key parameters for the stabilization of Cu(I). While Cu(I) is not very stable on smooth surfaces, it can be stabilized at preoxidized Cu cube/Cu foil interfaces or as CuI in iodine-pretreated surfaces, in boron-doped copper 34 or in copper−nitrate-based catalysts. 35 To date, no detailed in situ or operando study is available on dendritic catalysts to shed light into the possible transformations in their morphology/structure and chemical state under CO 2 RR conditions and to distinguish the possible influence of such parameters in the selectivity trends previously reported.
In this work, we report that, although Cu(I) species have been demonstrated to be advantageous for the increased yield of C2− C3 products, the stabilization of such species requires a delicate interplay among the morphology of the Cu catalysts and its interaction with the environment, such as the support and electrolyte. In the present case, our quasi in situ X-ray photoelectron spectroscopy (XPS) and operando X-ray absorption fine-structure spectroscopy (XAFS) data revealed that Cu(I) species are not stable on the large (micrometer structures) dendritic Cu catalysts supported on Ag and Pt and that the selectivity trends observed can be exclusively assigned to their morphology and the contribution from exposed regions of the respective underlying supports (Ag or Pt). In this case, and contrary to our previous observations for smaller Cu cubes on Cu, Cu(I) species are not stable at the Cu x O/Ag or Pt interfaces, which might be explained on the basis of the lower stability of AgOx and PtOx species and the much lower contact area and interaction with the support of the Cu dendrites as compared, for example, to our former O 2 -plasma treated Cu cubes grown on a Cu foil. 1 Electrochemical Characterization. The reactions were carried out in a customized H-type glass cell using a conventional three electrode setup. A platinum mesh and a leak-free Ag/AgCl electrode (LF-1, Innovative Instruments, Inc.) were used as counter and reference electrode. The electrolyte employed was 0.1 M KHCO 3 (ACS reagent 99.7%, Sigma-Aldrich). To be able to compare the results with the literature, measurements were also conducted in the absence of the O 2 -plasma pretreatment. Prior to usage, the electrolyte was treated with Chelex (Bio-Rad Laboratories Inc.) to remove trace metal impurities. 40,41 Before introducing the working electrode into the cell, the electrolyte was purged with CO 2 gas (99.95%, Air Liquide) for 30 min until a pH of 6.8 was reached. The reaction was investigated at different potentials using the chronoamperometry method. Gas products from the cell were injected online every 16 min into a gas chromatograph (GC, Shimadzu 2014; HayeSepQ + HayeSepR column; flame ionization detector (FID) and thermal conductivity detector (TCD)) while the liquid products were detected after 1.5 h of reaction using a liquid chromatograph (HPLC Shimadzu; NUCLEOGEL SUGAR 810 column; refractive index detector (RID)).
Quasi in Situ X-ray Photoelectron Spectroscopy. Quasi in situ XPS measurements were performed directly after electrochemistry without exposing the sample to air by transferring it from the electrochemical cell (EC-Cell) to the adjacent XPS chamber in ultrahigh vacuum (UHV). Before UHV transfer, the sample was rinsed within the electrochemical cell with 10 mL of deionized water (R = 18.2 MΩ), which was bubbled with N 2 (99.999%, Air Liquide) for 45 min prior to usage. Figure S1 shows the experimental setup. For the XPS measurements, a commercial Phoibos100 analyzer (SPECS GmbH, E pass = 15 eV) and XR50 (SPECS GmbH) X-ray source were employed. For our study, the Al anode (P source = 300 W) was used to measure all but the Cu-Auger spectra. A Mg anode (P source = 250 W) was utilized to avoid an overlap with the Ag 3p region in the case of the Cu samples grown on the Ag substrate. The Cu 2p 3/2 peak corresponding to CuO (E bin = 933.11 eV) 42 was used to align the spectra for the samples containing CuO, and the binding energy of Cu/Cu 2 O (E bin = 932.67 eV) 42

ACS Catalysis
Research Article measurements were carried out using an Autolab potentiostat (PGSTAT 302N).
Scanning Electron Microscopy. Scanning electron microscopy (SEM) measurements were performed using a Quanta 200 FEG microscope from FEI with a field emitter as electron source (10 kV). To acquire the images, a secondary electron detector (Everhart Thornley) was employed. The EDX measurements were performed with a separate liquid-N 2 -cooled detector (10 kV).
X-ray Absorption Fine-Structure Spectroscopy. The XAFS measurements were performed at the undulator beamline P65 of the PETRA III storage ring operating at 6 GeV in top-up mode. For the energy scan, a Si(111) monochromator was used with 60% detuning to reject higher harmonics. The experiments were performed in fluorescence mode at the Cu−K edge (8989 eV) at an angle of 15°using a Passivated Implanted Planar Silicon (PIPS) detector. For each sample, multiple identical spectra were acquired before and under operando conditions and averaged to improve the signal-to-noise ratio. Initial processing of the data was performed in Athena. 43 The spectra have also been corrected for self-absorption using the algorithms implemented within Athena, 44 so that the edges and the edge step size of the merged spectra and our measured reference spectra were aligned. Extended X-ray absorption fine-structure spectroscopy (EXAFS) analysis was conducted in Artemis using the FEFF6 code 43 to extract the coordination numbers (CN), interatomic distances, and disorder parameters. S 0 , the amplitude reduction factor accounting for many body effects, was determined by fitting our measured reference with a fixed coordination number of N = 12 for the Cu−Cu coordination. Linear combination fitting of the X-ray absorption near-edge spectroscopy (XANES) data was done using Athena.
In order to perform the operando measurements, a customized three electrode electrochemical cell including a leak-free Ag/AgCl reference and a Pt mesh counter electrode was used. The electrolyte was an aqueous solution of 0.1 M KHCO 3 .
■ RESULTS AND DISCUSSION Figure 1 shows typical SEM images obtained on our Cu dendrite catalysts deposited on Ag and Pt foils before and after a 5 min O 2 -plasma treatment and subsequent CO 2 RR. It is important to note that, due to the evolving hydrogen bubbles during the Cu dendrite electrodeposition process, parts of the different substrates are not completely covered with copper dendrites after electrodeposition (see Figure S2), and consequently, the substrates may serve as active cocatalyst supports during the CO 2 RR.
As seen in our SEM microscopy images, the low-pressure O 2plasma exposure changes the morphology of the pristine silver substrate even for exposure times as short as 1 min (see Figure  S3 (a−f)), while the clean Pt substrate remains unaffected ( Figure S3 (g−i)). Nevertheless, once the Cu dendrites were electrochemically grown on both substrates, the dendrites themselves were found to be surprisingly resistant to morphological changes after short plasma treatment times up to 1 min (see Figure S4) and to CO 2 RR conditions for up to 1 h at −0.9 V vs RHE. This picture changes for longer plasma treatment times such as 5 min. Here, a decrease in the sharpness of the dendrite tips (Figure 1 (e,g)) is observed, which results from the formation of CuO, as confirmed by our XPS and EDX measurements (Table S1). Upon exposure to the electrochemical conditions, visible roughening and the presence of small particles on top of the dendrites are observed (Figure 1  (f,h)).
To estimate the change in the roughness of the former samples, capacitance measurements were used. 20,23 The data are shown in Figure S5 and are consistent with our SEM analysis. Upon plasma oxidation prior to CO 2 RR, the roughness of both samples increases independently of the substrate used. Nevertheless, the increase is two times higher on the Ag substrate as

ACS Catalysis
Research Article compared to the Pt substrate (see Figure S5). We assigned this difference to the larger changes experienced by the Ag substrate upon O 2 -plasma exposure ( Figure S3).
The combined faradaic efficiency for the targeted C 2 and C 3 hydrocarbons is shown in Figure 2 (a) together with that of CO (Figure 2 (b)), methane (Figure 2 (c)), the parasitic hydrogen evolution reaction (Figure 2 (d)), and the desired multicarbon products, namely, ethanol (Figure 2 (e)) and ethylene (Figure 2 (f)) (for 1-prop, C 2 H 6 , and HCOOH, see Figure S6). It is evident that the O 2 -plasma pretreatment is able to increase the yield of the desired C2−C3 products by up to 20% at −0.9 V vs RHE, while suppressing hydrogen and methane formation, which has not been demonstrated for these types of samples in the absence of the plasma prefunctionalization. Another important new feature of the present samples is that there is an additional promoting effect on the selectivity toward C 2 and C 3 products depending on the substrate used. In this case, and regardless of the plasma pretreatment, the performance of the dendrites is increased when supported on Ag versus Pt. We attribute this difference to the fact that Ag is able to reduce CO 2 toward CO while Pt is known to be active only for the hydrogen evolution reaction (HER) and not for CO 2 RR due to its strong binding to CO that leads to poisoning of the surface. 4 The Ag substrate serves as a source for CO that might be more strongly bound at the rough Cu dendrite/Ag support interface, similar to what has been reported for other gases on these types of interfaces. 34 This might facilitate further CO 2 reduction to C 2 − C 3 products.
To gain further insight into the role of the support on CO 2 RR, the CO faradaic efficiency of the pristine Ag and Pt substrates must be considered, Figures S8 and S9. Our plasma treated Ag substrate is active for CO 2 reduction toward CO in the entire potential range investigated here for Cu dendrites/Ag, while Pt is only shown to be active for HER but not for CO 2 RR. For the Cu/Ag system, 5−10% CO is detected at potentials higher than −0.95 V vs RHE, while no CO is detected for Cu/Pt. This indicates that the CO observed at these high potentials is being produced on exposed Ag substrate areas. 1,23 On the other hand, exposed Pt areas in the Cu/Pt system lead to the observed increase in the hydrogen production above −0.9 V vs RHE.
In Figure 2 (b), it can be seen that the highest CO production at low overpotential (−0.6 V vs RHE) is obtained for the Cu/Ag sample pretreated by a 5 min O 2 -plasma. This is assigned to the role of the Ag support, since the pristine plasma-treated Ag foil is up to 70% more selective toward CO in comparison to the untreated one at the same potential due to its enhanced roughness and special defects created during the plasma pretreatment. 10 Therefore, the improved selectivity toward C 2 and C 3 products observed for the Ag-supported Cu dendrites as compared to those grown on Pt is indeed linked to the amount of CO generated by the substrate which, in a coupled reaction mechanism, is able to become further reduced by the dendritic copper overlayer.
Since in the case of the Cu/Pt system the generated CO can only originate from the copper dendrites, the decreased selectivity toward CO upon oxygen plasma treatment suggests the preference of a different pathway, namely, the one leading to C 2 −C 3 products (18% increase after O 2 -plasma with respect to the as prepared state for potentials as low as −0.7 V vs RHE). In addition, our plasma treatment is able to further reduce the selectivity toward methane for potentials higher than −0.9 V vs RHE in favor of ethylene and ethanol on both substrates. Nevertheless, this additional CH 4 suppression is proportional to the roughness (see Figure S5), which might result in an increased local pH. 34,36 The overall increase in the ethylene and ethanol production can however not be explained by this fact alone, as the increase of up to 20% is already present at potentials as low as −0.9 V vs RHE, while the methane generation is not observed for potentials lower than −0.9 V vs RHE for the as prepared samples on Ag and Pt. Moreover, our plasma treated

ACS Catalysis
Research Article samples were found to suppress the unwanted parasitic hydrogen evolution reaction. Thus, we attribute the overall higher partial current density for hydrocarbons on Ag-supported samples as compared to the Pt-supported ones to the further reduction of the CO generated on the exposed areas of the Ag substrate.
To further understand the role of the chemical state of Cu and the surface roughness on CO 2 RR activity and selectivity, quasi in situ XPS and operando XAFS measurements of our catalysts were conducted. The results are shown in Figures 3 and 4 as well as in Figure S10 and Table S2 for the Ag substrate, which turns out to be metallic during CO 2 RR. As it can be seen from the fitting of the Cu LMM Auger lines with the corresponding reference spectra, all samples contain Cu x O species before CO 2 RR, and the ones O 2 -plasma pretreated were found to have fully oxidized CuO surfaces.
On the Cu dendrite samples, the quasi in situ XPS data ( Figure 3) revealed that, irrespective of the sample pretreatment, the oxides do not survive during CO 2 RR at an applied potential of −0.9 V vs RHE, which is the peak performance potential for ethylene production. Consequently, we do not expect them to be present at even higher potentials (such as −1.0 V or −1.1 V vs RHE), which in fact were found to show a high faradaic efficiency for ethanol and 1-propanol. The improved C 2 −C 3 selectivity of these dendritic structures is therefore not linked to the presence of residual Cu x O during the reaction, as previously discussed in the literature based on ex situ spectroscopic data, 32,45 since they promptly become reduced under reaction conditions, but rather to their rough and defective structure. Additionally, we tested whether Cu x O species are present directly after the in situ dendritic electrochemical growth and before CO 2 RR to shed light into a question raised in the literature 32 on whether the improved C−C coupling observed for these structures could be due to the fact that the dendrites were initially containing Cu x O and that the subsequent oxidereduction process could lead to their improved catalytic performance. We therefore grew the dendrites on a commercially available carbon-based gas diffusion electrode (Freudenberg C2 GDL) within our UHV-compatible electrochemical cell and conducted quasi in situ XPS measurements directly after deposition (following the synthesis described in ref 32) and before air exposure. No Cu x O species were observed directly after deposition ( Figure S11). These XPS data corroborate that the better efficiency for C 2 and C 3 products seen for our plasma-pretreated dendrite samples and also for the pristine dendrite samples presented in the literature is not linked to residual Cu oxides remaining during the reaction.
Regarding the role of the needle-like structure, Klingan et al. 36 were able to show that indeed the local pH at the dendrite surface increases, thus favoring the formation of ethylene over methane. That this increase in the selectivity is indeed linked to the needle like structure and not to an overall roughness of the dendrite surface was demonstrated by Reller et al. 32 who reported a clear drop of the selectivity toward ethylene in correlation to a coarsening of the needle-like structures. Although we managed to stabilize this beneficial dendritic structure even upon plasma pretreatment, our catalysts show a worse performance for C−C hydrocarbons as compared to previously published oxide derived catalysts, 23 which is possibly due to the lack of stabilization of Cu(I) within the copper dendrites supported on Ag and Pt as shown by our quasi in situ XPS and operando XAFS.
Since XPS is a surface-sensitive technique and does not provide bulk chemical or structural information, operando XAFS measurements were carried out to confirm the absence of the Cu oxides also under subsurface reaction conditions, Figure  4. It was hypothesized that such oxides could migrate toward the surface during the reaction due to the concentration gradient across the sample and through their progressive reduction lead to a special defective undercoordinated structure that was stable during the reaction. Linear combination fitting (see Table S3) of the XANES data presented in Figure 4 (a,b) reveals that CuO is the dominant species for our plasma treated samples on both substrates before electrochemistry, which is consistent with our

ACS Catalysis
Research Article XPS analysis (Figure 3), but now, we also demonstrate that it is not only available at the sample surface. Regarding the as prepared Cu dendrites (without plasma exposure) before electrochemistry, a mixture of mainly Cu (53%) and Cu 2 O (46%) for the case of the Pt substrate was found while on Ag, the majority of the Cu dendrites are made of Cu 2 O (86%). We attribute this to the fact that the samples were exposed to air after their synthesis before the operando CO 2 RR XAFS measurements.
Once the potential of −0.9 V vs RHE is applied during CO 2 RR, we see almost immediately a change of the XANES spectrum reflecting only the presence of metallic Cu, independently of the initial oxidation state and the substrate used (see Figure 4), as also determined by linear combination fitting. This change can also be clearly seen when looking at the r-space representation of our EXAFS data (Figure 4 (c,d)), as the characteristic peaks for Cu 2 O (1.45 Å) and CuO (1.55 Å) vanish during CO 2 RR, giving rise to the metallic peak at 2.2 Å, which reflects Cu−Cu coordination. In agreement, our EXAFS fitting analysis shows a Cu−Cu coordination number of N Pt = 10.8−11.3 ± 1.0 for the Pt-supported and N Ag = 11.8 ± 1.5 for the Ag-supported Cu dendrites (see also Table S4). Taking into account the error margins and the fact that the spectra have been corrected for self-absorption, we conclude that our Cu dendrites consist of bulk metallic Cu (N Cu−Cu = 12), which demonstrates that not only the very surface, as probed by XPS, but also the bulk of the sample is metallic during CO 2 RR. Thus, Cu x O species cannot be the active species for the conversion of CO 2 toward hydrocarbons in the dendritic Cu structures. This indicates that the structural properties, especially in the case of the plasma-treated/preoxidized Cu dendrites, are the most important factors determining their CO 2 RR performance.

■ CONCLUSIONS
Our in situ and operando investigation of dendritic Cu electrocatalysts reveals the intricate interplay between structural and chemical characteristics of these materials and their underlying supports with respect to their performance for CO 2 electro-reduction. Here, we were able to improve an already promising CO 2 RR catalyst with respect to its selectivity toward ethylene and ethanol by combining the benefits of a lowpressure O 2 -plasma treatment leading to an increase of the catalyst surface roughness, with the selection of a suitable support, in this case, Ag, that could serve to increase the CO concentration in the proximity of the Cu catalyst. On such structures, a drastic suppression of the parasitic hydrogen production could be achieved.
Moreover, we were able to demonstrate that the ∼45% faradaic efficiency for C 2 −C 3 products obtained is related to structural properties on these materials and not to the presence of residual Cu x O in the vicinity of the surface or within the bulk as was previously postulated for related oxide-derived Cu catalysts based on ex situ spectroscopic data.