Probing CO2 Reduction Pathways for Copper Catalysis Using an Ionic Liquid as a Chemical Trapping Agent

Abstract The key to fully leveraging the potential of the electrochemical CO2 reduction reaction (CO2RR) to achieve a sustainable solar‐power‐based economy is the development of high‐performance electrocatalysts. The development process relies heavily on trial and error methods due to poor mechanistic understanding of the reaction. Demonstrated here is that ionic liquids (ILs) can be employed as a chemical trapping agent to probe CO2RR mechanistic pathways. This method is implemented by introducing a small amount of an IL ([BMIm][NTf2]) to a copper foam catalyst, on which a wide range of CO2RR products, including formate, CO, alcohols, and hydrocarbons, can be produced. The IL can selectively suppress the formation of ethylene, ethanol and n‐propanol while having little impact on others. Thus, reaction networks leading to various products can be disentangled. The results shed new light on the mechanistic understanding of the CO2RR, and provide guidelines for modulating the CO2RR properties. Chemical trapping using an IL adds to the toolbox to deduce the mechanistic understanding of electrocatalysis and could be applied to other reactions as well.


Introduction
Thee lectrochemical CO 2 reduction reaction (CO2RR) provides ap romising solution to offset the increased atmospheric CO 2 concentration, and also represents an excellent option for storing intermittent renewable electricity (e.g. solar,w ind energy) by producing value-added chemicals. [1] However,p oor energy conversion efficiencya nd broad product spectrum are major barriers to achieving economic viability of the CO2RR. Intensive effort has been spent searching for high performance electrocatalysts. [2] Copper (Cu) is identified as the only metal that produces hydrocarbons and alcohols with appreciable Faradaic efficiency (FE), [3] due to its moderate binding strength with key intermediate species. [4] Despite its unique catalytic properties, mechanistic understanding of the reaction pathways which provides the basis of steering the CO2RR toward desired products,remains controversial. Although the adsorbed *CO species is well-accepted as ak ey intermediate leading to various C 2+ products,i tr emains an open challenge to elucidate the mechanistic pathways from *CO to C 2+ products on Cu. Especially,the formation mechanisms of ethylene and ethanol have long been the subject under debate in both experimental and theoretical studies. [5] Mechanistic understanding of the CO2RR are derived almost exclusively through in situ/operando spectroscopic techniques (e.g., IR, Raman). [6] Early in situ spectroscopic studies of Cu electrodes suggest that hydrogenation of *CO to *CH 2 would be the precursor to ethylene and ethanol, [6c,7] while others suggest that formation of these C 2 species would mainly proceed through forming a*CO dimer (*C 2 O 2 À )which is subsequently protonated to *CO-COH. [8] These discrepancies may stem from the inherent limitations of spectroscopic techniques.T he limitations include the interference from the solvent or spectator species, [8b, 9] limited temporal/ spatial resolution due to the low coverage and short residual time of key intermediates, [6c,d] and ill-defined background signals that are sensitive to electrode pretreatment history and cell configurations, [6d,10] and all these may add to the uncertainty of the measurement and make interpretation of resultant spectra an on-trivial task. [6d] Complementary ways of analyzing the CO2RR mechanism are highly desirable.
Chemical trapping is regarded as an effective way to study reaction mechanisms.I to riginated in organic chemistry and was widely applied in catalysis. [11] Thereaction mechanism is deduced using ac ompound (trapping agent) that reacts specifically with one or more reaction intermediate(s) to form as table product(s). Thet rapping agent stops/decelerates specific reactions,a nd reaction mechanisms can then be deduced by examining the products.Bell et al. demonstrated in their exemplary works that the production of hydrocarbons from CO hydrogenation involved adsorbed methylene species as akey intermediate,asshown by the suppressed formation of hydrocarbons in presence of methylene scavengers. [12] This chemical trapping method has not yet been applied to electrocatalysis,l argely due to the lack of suitable chemical trapping agents that can selectively interact with specific intermediates without being oxidized/reduced under electrochemical conditions.I nspired by previous works where ionic liquids (ILs) were employed as surface modifiers to modulate the catalytic properties of avariety of electrocatalysts,anILis used here as achemical trapping agent to analyze the CO2RR pathways in Cu catalysts.This idea is realized by analyzing the IL-induced perturbation in the product spectrum. Therationales for choosing ILs also include their coordination ability with CO2RR intermediates and good stability over aw ide potential window. [13] ILs have been used as either pure electrolyte or electrolyte additive to change the CO2RR properties in various metal catalysts (e.g.,Ag, Pb). [14] ILs are reported to lower the overpotential and explicitly favor the formation of CO,p resumably through coordinating with reduction intermediates (e.g., CO 2 À C)byeither stabilizing the intermediates or preventing their spatial approach. [13c, 15] In the current study,the IL is introduced by immobilizing as mall amount of 1-butyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide ([BMIm][NTf 2 ]) on aC u-Foam catalyst (see Figure S1 in the Supporting Information). This method follows the concept of "solid catalyst with ionic liquid layer (SCILL)", which was first invented in heterogeneous catalysis, [16] and was soon successfully transferred to electrocatalysis particularly in improving electrocatalysts for the oxygen reduction reaction (ORR). [17] Thehydrophobic nature of the IL and capillary force ensure the confinement of the IL within the catalysts even in aqueous electrolytes. [17d,e] We demonstrate that IL can act as ac hemical trapping agent in the CO2RR. Its presence significantly alters the product spectrum by selectively suppressing the formation of ethylene,ethanol, and n-propanol, without disturbing either FE or partial current density of the others.T hese findings demonstrate selective interactions between the IL and one or more reaction intermediate(s), while the altered product distribution provides au nique perspective to track the CO2RR pathways.T his work may represent as imple approach to gaining mechanistic insights into the CO2RR, and also paves anew way in modulating the CO2RR activity and selectivity.

Results and Discussion
Cu foams were chosen because of the unique catalytic property of Cu and the abundance of porous structure which is beneficial to IL immobilization. Cu-Foams were prepared using ah ydrogen evolution reaction (HER) assisted electrodeposition method, [18] with aC up late as the substrate and copper sulfate as the precursor ( Figure S2). [BMIm][NTf 2 ] was chosen because of its hydrophobic nature and ability to coordinate with CO 2 and/or its anion radical. [15,19] Figure 1a Figure S7). Both samples exhibit amajor XPS peak at abinding energy (BE) of 932.5 eV,which associates with Cu 0 /Cu + .The Cu LMM Auger spectra confirm that the surface Cu on both samples mainly exists as Cu + (i.e. Cu 2 O), [20] which is not surprising since the oxidation of Cu to Cu 2 Oo ccurs immediately upon air exposure. [21] Am inor shoulder peak at aB Eo f9 34.7 eV,w hich associates with Cu(OH) 2 , [20] can also be observed on pristine Cu-Foam, indicating that asmall portion of Cu 2 OinCu-Foam are prone to further oxidation to form Cu(OH) 2 .T his consequent oxidation process was also reported by Tannenbaum et al. when studying the initial oxidation behavior of Cu in air. [21] Intriguingly,t his shoulder peak is absent on Cu-Foam-IL, implying that the IL can help suppress surface oxidation, which is in line with our previous study on Pt-based catalysts. [17e-g] Notwithstanding this difference,c onsidering the well documented readiness of copper oxide reduction under CO2RR conditions, [10,22] the presence of asmall portion of Cu(OH) 2 species on initial Cu-Foam is not expected to play as ignificant role in altering the product distribution. The CO2RR performance on Cu is sensitive to surface facets of Cu. [23] To find out whether the IL can change the Cu surface by selectively blocking certain facets,w ep erformed Pb UPD stripping experiments on both samples ( Figure S8). The comparable integrated areas of Pb UPD stripping peaks verify that (selective) blocking of Cu facets by the IL can be ruled out.
TheC O 2 electrolysis experiments were performed in ag as-tight electrochemical cell with anode and cathode separated by an anion exchange membrane ( Figure S9).   Figure S10). Despite the fluctuation, the electrolysis current densities are more or less comparable at the beginning and end of the electrolysis on both samples. This result indicates that Cu foams are stable during the electrolysis regardless of IL modification, which is also evidenced by the intact dendritic structures of both Cu foams after the electrolysis ( Figure S11). Thes tability of the IL on Cu foams during the CO2RR was also probed by performing post-reaction analyses of Cu-Foam-IL using both XPS and diffuse reflectance infrared Fourier transform spectroscopy techniques.T he characteristic signals of IL can be clearly resolved using both techniques after the electrolysis (Figure S12), implying that the IL can be well-maintained within the Cu foams during electrolysis.The overall current densities are comparable between these two samples,d espite as light current increase in Cu-Foam-IL at potentials of À0.7 and À0.8 Vv ersus reversible hydrogen electrode (vs.R HE). These results verify that the presence of the IL has not induced any change in mass transport properties of reactant molecules (CO 2 )f rom bulk electrolyte to Cu-Foam surfaces, and also imply that the majority of the CO 2 molecules may approach the catalyst surface in afree form instead of an ILcoordinated form. Figure 2b compares the FEs of various products on both samples at À0.7 V. Av ariety of products, including CO,f ormate,e thylene glycol (EG), ethylene, ethane,e thanol, n-propanol, methane,a nd acetate can be detected, with CO,f ormate,a nd EG identified as the major products (in addition to hydrogen). Va rious CO2RR products can usually be observed in Cu foams,while the major product depends on their morphology,a ctive surface area, and foam thickness. [2b,18, 24] Intriguingly,h erein we observe that EG, which is usually identified as aminor product in Cu catalysts, is produced with impressively high FEs ( % 20 %) on both samples.These results showcase that Cu foams are aversatile platform in producing value-added CO2RR products.
Thep otential dependent FEs of various products on pristine and IL-modified Cu-Foams are compared in Figure 3. TheH ER, am ajor competing reaction of the CO2RR, still dominates the product spectra on both catalysts.Asurge in H 2 production is observed at electrode potentials lower than À0.7 V, relating to the liberation of surface sites from adsorbed *CO. [25] Meanwhile,t he HER is promoted by IL  modification. This may stem from the inherent acidity and superior proton transfer capability of the IL being used, which offers greater proton availability for the HER. [14a, 26] TheH 2 production rates on both samples converge at lower electrode potentials (< À0.85 V), indicating that at higher reaction rate, the HER is mainly limited by the diffusion of proton (or proton source) from bulk solution to the catalyst surface,and the influence of IL modification is not pronounced. Similar potential-dependent FEs for major CO2RR products,including formate,E G, CO,e thylene,a nd ethane,c an be observed on both samples despite some minor difference in FEs for EG and formate at around À0.7 V, due to the liberation of strongly adsorbed *CO intermediate from Cu surfaces. Different from other studies of the CO2RR on Cu catalysts, on which methane is am ajor product, in the current work, methane is produced with ar ather low FE (< 1%)o nb oth catalysts.S imilarly,B roekmann et al. also observed that C 1 pathway to methane was almost completely suppressed on Cu foams. [18] Themorphology or surface faceting of Cu catalysts plays ac rucial role in determining the product selectivity of CO2RR. [27] Fori nstance,C u(100) facets favor the formation of ethylene while Cu(111) facets facilitate the formation of methane. [27b] This structure sensitive behavior of the CO2RR on Cu catalysts originates from the differences in binding energy of *CO and/or energetic barrier for the CÀCcoupling or hydrogenation step between different Cu facets. [8b, 28] Herein, the low FEs of methane on both catalysts imply that the Cu foams after the initial reduction of surface Cu 2 O species during the CO2RR might be enclosed by abundant Cu(100) facets as suggested by Broekmann et al. [18,24] The comparable FEs and onset potentials for major products such as CO and formate on Cu-Foam and Cu-Foam-IL also verify that the presence of IL has not induced any fundamental structural change on the Cu foam itself,and at the same time, the possible blockage or surface rearrangement of specific faceting by IL molecules during the CO2RR can be excluded. Them ost striking effect induced by IL modification is that ethanol and n-propanol, giving amaximum FE of 7%and 5% on pristine Cu-Foam, respectively,a re completely absent on Cu-Foam-IL (Figures 3h and i).M eanwhile,t he FE of ethylene is solely suppressed in the high overpotential region (< À0.7 V), with the highest FE decreasing from 10.2 %t o 5.2 %after IL modification (Figure 3d), while little difference can be observed in the low overpotential region (i.e.f rom À0.6 to À0.7 V). Thes ame conclusion can also be drawn by comparing the partial current densities of CO2RR products ( Figure S13). TheI Lh as selectively slowed down the production rate of ethylene in the high overpotential region and ceased the production of ethanol and n-propanol. These results demonstrate that the feasibility of the IL as achemical trapping agent, which provides the basis for deducing the CO2RR pathways by analyzing the altered product spectrum in presence of the IL.
Despite understandings of reaction pathways on Cu catalysts are rife with controversy,s ome consensus has been reached, which enables discussion of the observed chemical trapping results.T ransferring the first electron to CO 2 to form CO 2 À C anion is considered as the rate-determining step for CO 2 activation because of the high reorganization energy needed to activate alinear CO 2 molecule to form CO 2 À C anion with bent coordination geometry. [8d, 15, 29] Moreover,C Oi s identified as akey intermediate during the reduction of CO 2 to various C 2+ products,since CO is the only C 1 molecule that gives similar product spectrum as CO 2 on aC uc atalyst. [3a,d] However,i tr emains elusive how the adsorbed CO intermediate is further converted into various products.Intrigued by the altered product spectrum after IL modification, we clarify several elusive reduction pathways by referring to the widely reported yet controversial mechanism in literature,as summarized in Figure 4.
Among various products,e thylene shows the most interesting response to IL modification. Its formation is only suppressed at high overpotentials,while at low overpotentials both FE (Figure 3d)a nd the partial current density of ethylene ( Figure S13d) are almost the same regardless of IL modification. This result strongly suggests that ethylene could form by two separate pathways at high and low overpotentials.Adual pathway mechanism for ethylene production was proposed by Koper et al. when studying CO reduction on Cu. [28b] One pathway (Pathway II) involves the dimerization of two adjacent CO at low overpotentials,w hich is later reduced and protonated to form ethylene.T he dimerization would proceed by forming ah ydrogenated CO dimer (*CO-COH) as confirmed by spectroscopic and theoretical studies. [8b] On the other pathway (Pathway I), CO is converted into either *CHO [4a, 28b] or *COH [30] at high overpotentials, which is then reduced to carbene-like *CH 2 species,followed by either C À Ccoupling between two *CH 2 ,orCOinsertion as in the Fischer-Tropsch process,t op roduce ethylene. [30] The dual pathway mechanism may also hold its validity for ethylene production in Cu-Foams.T he IL could selectively quench one or more intermediates in Pathway I, which eventually suppresses the formation of ethylene at high overpotentials,w hile it appears that Pathway II, which starts at relatively low overpotentials and involves the C À C coupling through CO dimerization, is undisturbed by the IL.
Ethane is not atypical CO2RR product on Cu catalysts. [31] Thep roduction of ethane with as ignificant FE is explicitly observed on nanostructured porous Cu catalysts. [31,32] Ethane can be seen as areduction product of ethylene after two more protonation steps.The porous structure of Cu catalysts seems to increase the retention time of pre-formed products in aconfined space.Therefore,for along time,the formation of ethane has been attributed to the re-adsorption and reduction of pre-formed ethylene on Cu catalysts (Pathways Ia nd IIA). [31,32] However,b oth FE and partial current density of ethane are actually insensitive to IL modification (Figures 3e and S13e). Thee ntirely different responses of FEs for ethylene and ethane to IL modification imply that ethane is formed via an independent pathway.Recent works report that ethane is produced by the CO dimerization pathway involving ethoxy intermediate, [33] which reconciles with our observation that the pathway involving CO dimerization is undisturbed by the IL. These findings suggest that production of ethane would mainly proceed through Pathway IIB (Figure 4).
Ethanol is considered to share the similar formation mechanism as ethylene. [3d,8d] Twor eaction pathways,w hich involve either formation of carbene intermediate (*CH 2 ) (Pathway I) or dimerization of two adjacent CO (Pathway II), are usually proposed (Figure 4). We found that formation of ethanol is completely suppressed on Cu-Foam-IL, which suggests that IL traps the key intermediate(s) leading to the formation of ethanol. Similarly,n -propanol is not produced on Cu-Foam-IL. It is generally accepted that the formation of n-propanol undergoes intramolecular C À Ccoupling between CO and hydrogenated carbon (e.g.,carbene *CH 2 ), followed by proton/electron transfer to form propionaldehyde,a n intermediate that is further reduced to n-propanol (Figure 4). [3d, 8d, 34] It can be seen that the formation of both ethanol and n-propanol involves carbene species (*CH 2 ), which is also the intermediate to produce ethylene at high overpotentials. TheI L-induced suppression of ethanol, n-propanol, and ethylene (at high overpotentials,Pathway I) infers that these products likely share one or more common intermediate(s) selectively trapped by the IL.
Regarding the other CO2RR products including CO and formate,their differences in FEs and partial current densities are quite minor or within measurement error between two catalysts,d etermining their pathways conclusively becomes challenging.Nevertheless,some inspiring information can be deduced. Fori nstance,t he formation of CO and formate is insensitive to IL modification, indicating that starting from the adsorption of CO 2 on Cu surfaces to the formation of adsorbed CO,t he IL seems to play an egligible role,o ri n other words,the IL does not take effect through coordinating with CO 2 molecules which are more likely transported to the catalysts surface in af ree form. Moreover,E Gi su sually detected as am inor product of the CO2RR. [31,35] However, herein both Cu foam catalysts exhibit fairly high FE of EG: up to 25 %a nd 19 %o nC u-Foam and Cu-Foam-IL, respectively.T he formation of EG is double-checked by analyzing the liquid products using GC-MS ( Figure S14). Consensus on the reaction pathway to EG has not yet been reached, although it is inferred that EG formation might proceed through aC Od imerization mechanism. [31,35] Herein, EG formation is always accompanied by formate,a nd their FEs exhibit similar potential-dependent behavior,t hat is,h igher FEs obtained at lower overpotentials and maximum FEs obtained at around À0.7 V. These results imply that these two products probably share the same intermediate,for example, *CO 2 À ,w hich has been experimentally confirmed as ak ey intermediate to produce formate. [36] Brennecke et al. suggested that CÀCc oupling could also take place between two adsorbed CO 2 À to form oxalate species. [19a] Theh ypothesis here is that EG is produced via dimerization of two adsorbed *CO 2 À species,i nstead of *CO,f ollowed by multistep reduction and protonation to give EG (Figure 4). The predominant product at À0.7 Vs witches from EG on Cu-Foam to formate on Cu-Foam-IL. TheI Lm ay inhibit the dimerization process of the co-adsorbed CO 2 À C species by preventing their close approach. [19a] It is also intriguing to observe that IL modification exhibits little impact on the methane formation. Tw or eaction pathways are usually proposed for the methane formation. One pathway involves carbene (*CH 2 )a sa ni ntermediate,w hich is further reduced to *CH 3 and finally to CH 4 .T he other pathway is through hydrogenation of *CO to form *CHO,followed by amultiple electron-proton transfer process to produce CH 4 (Fig-Figure 4. Proposed reaction roadmaps of CO2RR on Cu catalysts. Selected intermediates are presented for clarity.U nfeasible pathways are marked by red crosses. ure S15). Considering that Pathway I( carbene pathway) has been significantly suppressed by the IL, herein, comparable FEs of methane on both Cu foams leads us to hypothesize that methane is mainly produced through the latter pathway ( Figure 4).
Analyzing the IL-induced change in CO2RR product distribution provides au nique perspective to gain some unprecedented mechanistic insights into the Cu catalyzed CO2RR which actually bypasses the necessity of explicit understandings about the chemical identity of surface intermediate(s). Based on the above results,asimplified overview of the reaction pathways that lead to varied CO2RR products is summarized in Figure 5, where the IL suppressed products and pathways are highlighted in yellow.I ti s intriguing to observe that the bifurcation of intermediates leading to the suppressed products starts right after the formation of adsorbed carbene (i.e.* CH 2 ) ( Figure 4). This hints that the key intermediate(s) are either *CH 2 or other species (e.g., *COH, *CHO,* C, *CH) that can further be converted to *CH 2 ( Figure S16). Another key question is how the IL molecules can trap the surface intermediate(s). IL molecules are reported to adopt ac harge-separated layered structure with alternating cation-/anion-rich layer at electrified surfaces. [17f,37] Accordingly,[ BMIm] + cations should be enriched at the innermost (Stern) layer of the electrodeelectrolyte interface when the electrode is negatively polarized (i.e.t he CO 2 electrolysis conditions). Therefore,u nderstanding of how [BMIm] + cations can possibly interact with other species would be crucial to extrapolate the role of ILs during the CO2RR. It is well documented that an imidazolium cation can easily be deprotonated at its C2-site,t hus converting the C2-site into ar eactive center due to its nucleophilicity. [38] Accordingly,t oc larify whether [BMIm] + interacts with surface intermediate(s) via its C2-site,a n imidazolium-based IL on which the C2-site at the imidazolium cation ring is "neutralized" by amethyl group (denoted as [BMMIm] + ,F igure S17a), was used for modifying Cu foams.Itturns out that the chemical trapping effect of the IL is not pronounced. Both ethanol and propanol can be detected, and the formation of ethylene at high overpotentials is not suppressed ( Figure S18). Furthermore,a nother IL, [HMIm][NTf 2 ]w hich shares structural similarity with [BMIm][NTf 2 ]b ut features al onger cationic chain, was also tested. Although both ethanol and propanol can still be detected, their FEs are much lower than those on unmodified counterpart, and ethylene formation is also suppressed (Figure S18). Twomore common ILs (i.e.[MTBD][NTf 2 ], [P 1444 ]-[NTf 2 ]) were also tested for comparison. Not surprisingly,no pronounced chemical trapping effect can be identified using either IL ( Figure S18). Their product spectra are comparable to that of the unmodified Cu-Foam, except for as lightly higher FEs of H 2 on Cu-Foam modified with [MTBD][NTf 2 ], probably due to the protonic nature of this IL. These results lead us to hypothesize that the IL traps the surface key intermediates through bonding with carbene (or other hydrogenated carbon species) on Cu surfaces.T his process may involve deprotonation and following alkylation reactions at the C2-site of the imidazolium ring. [39] Conclusion This work outlines an ew strategy to probe CO2RR pathways.T he IL alters the product spectrum during the CO2RR on Cu foams.A nalyzing the responses of CO2RR products to IL modification is au nique way to gain new insights into CO2RR pathways:1 )Ethanol and n-propanol form explicitly through a" carbene" mechanism, while formation of ethylene could proceed through two independent pathways which involve carbene and dimerized CO as key intermediates at high and low overpotentials,r espectively; 2) Thep resence of the IL can selectively suppress the formation of those products involving carbene intermediates, likely by forming stable imidazolium-carbene compound(s); 3) Ethane,w hich has long been considered ar eduction product of re-adsorbed ethylene during CO2RR, is identified as proceeding with an independent pathway that involves CO dimerization process.C onsidering the great structural flexibility in ILs,i dentification of reaction pathways for CO 2 products by carefully designing task-specific ILs to selectively interact with intermediate species may be feasible.T he success of this will bring IL modification closer to being ag eneric strategy for analyzing complicated CO 2 reduction pathways.T his approach is transferable to other electrocatalytic reactions and materials.This work demonstrates the possibility of moderating the CO2RR product spectrum by rationally leveraging the IL modification effect, which can be key to finely tuning the catalytic properties of aC O 2 reduction catalyst at am olecular level.