Electrochemical Reduction of Carbon Dioxide to 1‐Butanol on Oxide‐Derived Copper

Abstract The electroreduction of carbon dioxide using renewable electricity is an appealing strategy for the sustainable synthesis of chemicals and fuels. Extensive research has focused on the production of ethylene, ethanol and n‐propanol, but more complex C4 molecules have been scarcely reported. Herein, we report the first direct electroreduction of CO2 to 1‐butanol in alkaline electrolyte on Cu gas diffusion electrodes (Faradaic efficiency=0.056 %, j 1‐Butanol=−0.080 mA cm−2 at −0.48 V vs. RHE) and elucidate its formation mechanism. Electrolysis of possible molecular intermediates, coupled with density functional theory, led us to propose that CO2 first electroreduces to acetaldehyde‐a key C2 intermediate to 1‐butanol. Acetaldehyde then undergoes a base‐catalyzed aldol condensation to give crotonaldehyde via electrochemical promotion by the catalyst surface. Crotonaldehyde is subsequently electroreduced to butanal, and then to 1‐butanol. In a broad context, our results point to the relevance of coupling chemical and electrochemical processes for the synthesis of higher molecular weight products from CO2.


Introduction
Thee lectrochemical carbon dioxide reduction reaction (CO 2 RR) to fuels and chemicals,when powered by renewable electricity,i sapotentially sustainable way to alleviate our pressing global energy demands and to avert climate change. [1] Copper-based materials are the only family of catalysts that can reduce CO 2 to multi-carbon molecules with significant Faradaic efficiencies (FE)a nd current densities (j). [2,3] Among the multi-carbon products,C 2 molecules such as ethylene and ethanol can be facilely formed (FE = 30-50 %). [2,4] Themain C 3 product reported is n-propanol (FE = 10-13 %), [5,6] alongside small quantities of propionaldehyde, allyl alcohol, acetone,p ropylene and propane (total FE < 3%). [4,5] Reports on the formation of C 4 molecules,m any of which have much higher commercial value,a re scarce and have been limited to hydrocarbons showing FE < 1%. [7] The drastic decrease in the selectivity of aproduct as the number of carbon atoms in it increases suggests that the coupling mechanism to form am ulti-carbon product follows a" polymerization" scheme of *CO that obeys the Flory-Schulz distribution.
Interestingly,t he direct production of 1-butanol (CH 3 CH 2 CH 2 CH 2 OH) from electrochemical CO 2 reduction has not been reported. This oxygenate,w hich has ah igh volumetric energy density of 29.2 MJ L À1 and is less hygroscopic and corrosive than ethanol, has been suggested for direct use as af uel or in diesel-blends. [8] Schmid and coworkers have utilized bacteria to convert CO (generated from CO 2 electrolysis) to 1-butanol. [9] More recently,amechanistic study of CO 2 reduction to n-propanol revealed that minor and yet-to-be-quantified amounts of 1-butanol can be co-produced from the electrochemical reduction of acetaldehyde and CO in 0.1m KOHonoxide-derived Cu electrodes. [10] Still, no strategy to successfully electrosynthesize 1-butanol or any C 4 oxygenates from CO 2 has been conceived. To tackle this challenge,i ti sc rucial to understand and map out the mechanism and kinetics for its formation.
Herein, we report and quantify for the first time the formation of C 4 oxygenates from alkaline electrolysis of CO 2 using CuO-derived Cu gas diffusion electrodes (GDE) in aflow cell. Thepredominant C 4 product was 1-butanol (FE = 0.056 %, j 1-Butanol = À0.080 mA cm À2 )a tÀ0.48 Vv s. RHE (Reversible Hydrogen Electrode;from here on, all potentials are referenced to the RHE and all currents are normalized to the exposed geometric surface area of the electrodes,u nless otherwise stated). We then elucidate the reaction mechanism by combining analyses of reaction products from electrolyses of possible intermediates and density functional theory (DFT) investigations.T he formation of the critical C 4 intermediate,c rotonaldehyde (CH 3 CH = CHCHO), was traced to the aldol condensation (CÀCb ond formation) of two acetaldehyde (CH 3 CHO) molecules generated from CO 2 electroreduction (Figure 1a). Thealdol reaction is promoted by both OH À ions in the electrolyte and the electrocatalyst surface.C rotonaldehyde then undergoes at wo-step electroreduction to 1-butanol. We also unveil the critical role of pH at different stages of the reaction mechanism, pointing towards new strategies for increasing the performance of electro-assisted conversion of CO 2 into 1-butanol. Figure 1. a) Simplified reaction scheme for CO 2 reduction to 1-butanol. Further details are provided in Section S3. b) Diagram of CO 2 flow cell electrolysiss et up:characterizationo fC uO-derived Cu cathodes by (i)S EM, (ii)X RD and (iii)XPS (B.E. refers to binding energy). We note that the detection of Cu 2 Ointhe XRD and CuO + Cu(OH) 2 signals in the XPS data are likely due to surface oxidation of the electrode from its exposure to air (see Section S2.1 for amore detailed discussion);( iv) large CO 2 reduction current densities( in the order of À100 mA cm À2 )w hich gave asufficient rate of product formation to allow detection of minor products;(v) sensitive analytical techniques like headspace GC can quantify minor products down to the mm-scale. c) Faradaic efficiencies (FE,c olored bars) and partial current densities( j, &)o fC 2 ,C 3 and C 4 products from electrolysiso fCO 2 on CuO-derived Cu GDE in 1.0 m KOH. The major C 2 ,C 3 and C 4 products are ethylene, n-propanola nd 1-butanol, respectively. Other detected products are shown in Table S1.

Results and Discussion
We electroreduced CO 2 at various potentials using CuOderived Cu GDE cathodes in af low cell (Figure 1b, Sections S1,S2). Thec atalyst was electrodeposited onto the GDE using ap reviously-published procedure. [11] Aqueous 1.0 m KOHw as used as the electrolyte.T he high CO 2 RR current densities from the flow cell electrolysis (Figure 1b-(iv)), which circumvents mass transport limitations,combined with the use of highly sensitive headspace gas chromatography (Figure 1b(v)) improves the detection and quantification of liquid products with low FEs and current densities. This allows us to detect CO 2 reduction products that have,todate,never been observed.
Thet otal FEs of carbonaceous products were 68-69 %a t À0.48 and À0.58 V ( Figure 1c,T able S1). Them ajor multicarbon products are C 2 molecules,n amely ethylene and ethanol, which are typically formed on oxide-derived copper catalysts. [2,12] Theh ighest FE C 2 of 48 %w as observed at À0.58 V, with acorresponding j C 2 of À210 mA cm À2 .Minor C 2 products (FE 0.1 %) such as acetaldehyde and ethane were also detected. In the case of acetaldehyde,t he low FE is ar esult of the chemical and/or electrochemical transformations it readily undergoes,aswewill discuss below.C 3 species, mainly n-propanol, were also detected, with amaximum FE C 3 of 6.5 %a nd j C 3 of À18.5 mA cm À2 obtained at À0.48 and À0.58 V, respectively.O verall, the catalytic activities toward C 2 and C 3 molecules from CO 2 reduction are comparable or higher than values previously reported for Cu catalysts loaded onto carbon GDEs (Table S3). Interestingly,w ed etected C 4 oxygenates such as 1-butanol and butanal (maximum total FE C 4 = 0.060 %atthe onset potential of À0.48 V), in contrast to previous studies which only identified hydrocarbons for the C 4 fraction. [7,13] Thedominant product was 1-butanol, which is in line with the fact that aldehydes can be easily electroreduced to their corresponding alcohols,a sd emonstrated in the cases of formaldehyde and acetaldehyde. [14] No carboncontaining products were found in control experiments performed without applied potentials.
As 1-butanol is the sole C 4 alcohol product, it could not come from C À Cc oupling of four individual C 1 adsorbates such as *CO in aF lory-Schulz distribution, since 2-butanol was not detected. This led us to postulate that the formation of 1-butanol could occur through ac ombination of electrochemical and chemical steps.S pecifically,t he aldol condensation of two C 2 intermediates,s uch as acetaldehyde,g ives rise to the C 4 backbone of crotonaldehyde,w hich is further reduced to C 4 terminal oxygenates like butanal and 1-butanol. While mechanisms for the formation of major C 2 and C 3 products,including acetaldehyde,have been widely discussed in the literature, [6,10,[15][16][17][18] pathways for producing C 4 products are rarely mentioned. Herein, we focus on acetaldehyde reactivity as it is much less explored mechanistically.N onetheless,wealso present the two steps preceding acetaldehyde formation, which involve the key ethenyloxy intermediate, CH 2 CHO*, and discuss the lateral pathways for CO 2 reduction to C 1 -C 3 products in Section S3.
To test our hypothesis for 1-butanol formation via acetaldehyde,w ee lectroreduced 50 mm acetaldehyde in 0.1m KOHo nC uO-derived Cu electrodes (Section S4). The product distribution at an optimized potential of À0.44 Vi s summarized in Figure 2a (see also Table S5 for product distributions at other applied potentials). Thep roduct with the highest selectivity was 1-butanol (FE = 9.6 %, j 1-Butanol = À1.06 mA cm À2 ), consistent with expectations from ab asecatalyzed aldol condensation CÀCcoupling step.The remaining electrolysis products were other C 4 oxygenates such as butanal (CH 3 CH 2 CH 2 CHO) and crotyl alcohol (CH 3 CH = CHCH 2 OH), as well as ethanol. During the electrolysis,w eo bserved some coloration of the anion-exchange membrane due to its exposure to the alkaline acetaldehydecontaining electrolyte,b ut control experiments excluded its interference with our electrolysis results ( Figures S8,9).
Detection of crotonaldehyde (3.3 mm,e quivalent to 13.2 %a cetaldehyde conversion) after electrolysis suggests that the C 4 backbone of 1-butanol could be formed via abasecatalyzed aldol condensation (Figure 2a,i nset). We utilized the increase in local pH close to the electrode under electroreduction conditions to test this hypothesis and performed fixed-current electrolyses of 50 mm acetaldehyde in 0.01, 0.1 and 1.0 m potassium phosphate buffer (pH 7, Figure 2b,S ection S4.3). Higher buffer concentrations mitigate the local pH increase during electroreduction, thus lowering the overall local pH. Our results reveal that electrolysis in 0.01m potassium phosphate buffer gave the highest selectivity toward C 4 products (FE = 0.4 %, with 1butanol as the main product) and percentage of acetaldehyde converted to crotonaldehyde (1.5 mm,e quivalent to 6.2 % acetaldehyde conversion). In contrast, neither C 4 products nor crotonaldehyde were detected from experiments performed in 1.0 m buffer (Figure 2b inset). These observations of an alkaline local pH promoting the production of 1-butanol on copper directly support its formation via the aldol condensation of two acetaldehyde molecules and suggest an enhanced reaction rate close to the catalytic surface under electrolysis conditions. We further investigated, using DFT,t he base-catalyzed aldol condensation mechanism to form the C 4 backbone. High-level methods including solvent and potential effects have been employed to study electrochemical networks up to C 2 species,i ncluding the formation of acetaldehyde from CO 2 . [16] However, the multiple conformations of C 4 molecules and the complexity of the reaction network with chemical (bulk solvent and interface) and electrochemical steps limits us to the use of DFT coupled to the Computational Hydrogen Electrode (CHE). Thef ormation of the C 4 backbone comprises two steps:t he C À Cc oupling between two acetaldehyde molecules to form 3-hydroxybutanal (CH 3 CH(OH)CH 2 CHO), and the subsequent dehydration of the latter to crotonaldehyde (Figure 2c). In solution, the aldol condensation starts with the stripping of an a-hydrogen from acetaldehyde by OH À to form an ethenyloxy anion (CH 2 CHO À ,F igure 2c(i)). The a-carbon of CH 2 CHO À then attacks the carbonyl group of as econd acetaldehyde molecule,w hich is subsequently protonated to 3-hydroxybutanal (Figure 2c(ii)). Then, 3-hydroxybutanal loses an a-hydrogen as ap roton, to generate ac arbene species (Figure 2c(iii)), which forms crotonaldehyde by hydroxyl elimination. Con- show the percentage of acetaldehyde that was converted to crotonaldehyde in the electrolyte in each case. c) Potential energy diagram of acetaldehyde condensation to crotonaldehydeinsolution (dark blue) and mediated by the surface (black). The final OH À removal, which can be assisted by one electron donated from the surface, is promoted by reductive potentials (dark red). Water molecules were omitted for clarity.d )F aradaic efficiencies of products from alkaline electrolyses of 50 mm crotonaldehyde and 50 mm butanal on CuO-derived Cu. e) Potential energy diagramofcrotonaldehydereduction to 1-butanolvia butanal (blue). Under negative potentials, butanal can be adsorbed via ao ne-electron transfer (dark red), which promotes its further reaction instead of desorption. Dashed lines represent adsorptions/desorptions. Dotted lines represent proton-coupled electron transfers( PCET). Additional (electro)chemical routes are shown in SectionsS 3and S6. sistent with findings from the literature, [19] the latter is the rate-determining step in solution. We further note that for the case of CO 2 reduction, adsorbed ethenyloxy species is formed as aprecursor of acetaldehyde and ethanol, [16,20] and thus can also readily react with an acetaldehyde molecule in solution and be subsequently hydrogenated to form 3-hydroxybutanal (Section S3).
Theamount of acetaldehyde converted to crotonaldehyde during electrolysis (13.2 %) is larger than the 8.8 %c onversion (or 2.2 mm crotonaldehyde) when 50 mm acetaldehyde was aged in 0.1m KOHfor 1hwithout applied potential (inset in Figure 2a). This observation suggests that the Cu surface can promote the aldol condensation at cathodic potentials. DFT analysis reveals that this alternative pathway starts with 3-hydroxybutanal adsorbing exothermically on Cu and losing an a-hydrogen as an adsorbed H(dark red in Figure 2c). The hydroxyl group is then eliminated, in parallel with an electron transfer, to give crotonaldehyde.Onthe Cu surface,this step is promoted by negative applied potentials.Overall, the DFT investigation reveals that the alkaline electrolyte promotes the initial CÀCc oupling step between two acetaldehyde molecules,w hile the Cu surface promotes the subsequent dehydration step to crotonaldehyde.
To elucidate the fate of C 4 species after the aldol condensation step,w ee lectrolyzed 50 mm crotonaldehyde on CuO-derived Cu in 0.1m KOHa tÀ0.44 V ( Figure 2d, Section S5). Them ajor carbonaceous product was 1-butanol (FE = 14.8 %), while small amounts of butanal (FE = 0.3 %) and crotyl alcohol (FE = 1.1 %) were also detected. The Faradaic selectivity of 1-butanol (FE 1-Butanol normalized by the FE of all the C 4 products) from crotonaldehyde electrolysis was 91.4 %, which is similar to the case of acetaldehyde (94.7 %, Table S10). This reinforces the role of crotonaldehyde as the main intermediate in the electrosynthesis of acetaldehyde to C 4 oxygenates.T he presence of ethanol (FE = 4.5 %) was attributed to the reduction of acetaldehyde present due to the hydroxide-catalyzed retro-aldol reaction of crotonaldehyde,w hich is known to occur at room temperature. [21] This observation highlights the complexity of crotonaldehyde chemistry under aqueous alkaline conditions,a nd leads us to infer that the low total Faradaic efficiencyo f 72.2 %i saconsequence of undetected products from other side reactions of crotonaldehyde in the alkaline electrolyte ( Figure S10).
Butanal and crotyl alcohol are known intermediates in the gas-phase hydrogenation of crotonaldehyde to 1-butanol, [22] and their presence during crotonaldehyde electrolysis suggests that they are potential electrochemically-active intermediates to 1-butanol. Therefore,w ee lectrolyzed butanal and crotyl alcohol under the same conditions (Figure 2d, Table S9). 1-Butanol was the sole product from the electroreduction of butanal (FE = 17.3 %; the balance product is H 2 ). Only hydrogen was detected during the electrolysis of crotyl alcohol, which indicates that the latter is electrochemically inert, in good agreement with theoretical calculations in Figure S11.
Theoretical analysis of crotonaldehyde reduction reveals that butanal is formed by sequential hydrogenation of the band a-carbons of crotonaldehyde (Figure 2e). Once formed, butanal tends to desorb rather than further react. However,at potentials more reductive than À1.02 Vv s. SHE (standard hydrogen electrode), butanal receives an electron from the cathode surface to form the CH 3 CH 2 CH 2 C*HO À anion (I4 in Figure 2e). As this adsorption does not involve proton transfers,i ti si ndependent of the electrolyte pH in the SHE scale.C H 3 CH 2 CH 2 C*HO À is subsequently protonated to yield 1-hydroxybutyl (I5 in Figure 2e), which is further hydrogenated in ac hemical step (E a = 0.39 eV) to produce 1-butanol. These theoretical findings are corroborated by our results from crotonaldehyde electrolysis performed at pH 7 and pH 13 (Table S9). At À1.20 Vvs. SHE, 1-butanol was the most selective product in both electrolytes,consistent with the pH-independent adsorption of butanal. However, at À0.90 V vs.S HE, butanal was the most selective product, indicating that this potential was insufficient for its further reduction to 1-butanol. Alternative chemical routes from butanal to 1butanol are shown in Figure S12.
Aldehydes can be hydrated to geminal diols in aqueous alkaline solution. Signals belonging to hydrated crotonaldehyde (CH 3 CH=CHCH(OH) 2 )w ere observed in nuclear magnetic resonance (NMR) spectroscopic analyses of 50 mm crotonaldehyde or acetaldehyde dissolved in 0.1m KOH (Figure 3a,F igure S13). We therefore considered the possibility of hydrated crotonaldehyde as an intermediate to 1butanol. DFT suggests that hydrated crotonaldehyde cannot be further electrochemically reduced due to al arger activation barrier (+ 0.41 eV) compared to its desorption energy (+ 0.26 eV), as shown in Figure 3b.T his result was corroborated by the FE 1-Butanol of 46-47 %o bserved from the crotonaldehyde and butanal electrolyses in 0.1m potassium phosphate buffer (pH 7) at À0.79 V ( Figure 3c,T able S9), which was almost three times more than the values from electrolyses in 0.1m KOH. Figure 3a shows that in an eutral medium, only peaks corresponding to crotonaldehyde were observed. Then eutral buffer electrolyte therefore likely suppressed the hydration process and increased the availability of unhydrated crotonaldehyde for reduction.
To find ab etter electrocatalyst to enhance the FE of 1butanol from acetaldehyde and crotonaldehyde reduction, we proceeded to identify an activity descriptor.T ot his end, we electrolyzed acetaldehyde in 0.1m KOHa nd crotonaldehyde in 0.1m potassium phosphate buffer on different metal discs (Section S8). Since we have demonstrated that both the formation of C 4 oxygenates via the aldol condensation pathway,a nd the reactivity of crotonaldehyde are affected by the (local) pH, we employed aconstant-current electrolysis at À10 mA cm À2 to identify the different reactivities of the metals.F or acetaldehyde electrolysis,w ef ound that the selectivity of Cu, Fe,C o, Ni, Ag, and Au towards 1-butanol (Figure 4a)a nd all C 4 products ( Figure S14a) can be correlated to the cathode metal-oxygen bond strength, with Fe showing the highest selectivity towards 1-butanol (FE = 4.0 %). As imilar trend was observed for the six abovementioned metals and also Pd and Pt for crotonaldehyde reduction to 1-butanol (Figure 4b), with Fe also showing the highest FE of 26.3 %. Thel inear-scaling relationships end sharply in aselectivity cliff. [23] Theorigin of the discontinuity is likely due to ap hase transformation. According to their Pourbaix diagrams, [24] Zn, Ti,C r, and Mo may have surface oxide layers under the working cathodic potentials and thus have poor yields towards 1-butanol and other C 4 products (Section S9). Incidentally,t hese metals were more selective for reducing acetaldehyde to crotyl alcohol (Table S11), and crotonaldehyde to butanal (Table S12), probably because the dominating linear-scaling relationships differ from those of the pure metals.B ased on these results,w ep ut forward that the C 4 product selectivity is influenced by the affinity of the catalyst to oxygen.
Collectively,our results gave the reasons for the low FE of 1-butanol during the electroreduction of CO 2 on oxidederived copper.T he first factor is the low activity for acetaldehyde production on copper materials (FE CH 3 CHO 0.1 %, j CH 3 CHO À 0.41 mA cm À2 in Table S1, FE CH 3 CHO < 2.1 %, j CH 3 CHO < 1mAcm À2 in the literature [4,25] ). Thes econd factor is that the conversion of acetaldehyde to ethanol on copper is kinetically facile and strongly competing (FE ethanol = 36.5 %, while FE 1-Butanol = 9.6 %a tÀ0.44 Vv s. RHE, Figure 2a). [17,26] Thet hird factor is that the formation of the C 4 backbone and its subsequent reduction to 1-butanol are promoted by conflicting experimental conditions.W hile an alkaline environment facilitates aldol condensation between acetaldehyde molecules to crotonaldehyde,its hydration to an unreactive form is also promoted.

Conclusion
In summary,wed etected for the first time C 4 oxygenates from CO 2 electroreduction on CuO-derived Cu, with 1butanol being the most favored product amongst them (FE = 0.056 %, j = À0.080 mA cm À2 at À0.48 V). Theq uantification was made possible by ac ombination of the high-rate electrolysis,a chieved by using aG DE in af low cell, and sensitive analytical techniques.W eruled out the formation of 1-butanol from the C À Cc oupling of four individual C 1 adsorbates,s uch as *CO.I nstead, the combination of experimental and theoretical studies established ar ich reaction mechanism that combines chemical and electrochemical steps,w here the electrocatalytically generated acetaldehyde plays aprominent role.Its base-catalyzed aldol condensation, promoted by high local pH and the catalyst surface,produces crotonaldehyde,w hich is subsequently electroreduced to 1butanol.
This study further highlights the challenges associated with the one-pot approach to converting al ow molecular weight feedstock like CO 2 into complex functionalized molecules.W ed iscover that contrasting catalysts and conditions are required to maximize the yield of each step.I n addition to operating under highly alkaline conditions,w e also note that as ingle electrocatalytic surface can hardly optimize all the required steps.For example,among the metal discs tested, Fe was identified as the most selective catalyst for acetaldehyde reduction to 1-butanol, but it is not active per se for CO 2 reduction. Therefore,designing aone-pot reactor for the electrosynthesis of large molecules would inevitably be associated with low performance.I nstead, we propose that amore viable synthetic strategy would be to deconvolute the multi-step process into sequential operation units.T herein, chemical or electrochemical reactions with different process conditions could be independently optimized. Thes eparate stepwise-optimized reactors could then be placed in tandem for the efficient conversion of each intermediate,l eading to increased yield of the desired product. . Faradaic efficiencies of 1-butanolf rom À10 mA cm À2 constant-current electrolysiso fa)acetaldehyde and b) crotonaldehydeon selected metals as afunction of the DFT-computed adsorbed oxygen stability on these metals with respect to water and hydrogen. Metals that are typically oxides at 0Vvs. RHE at the pH of the supporting electrolyte are shown as hollow symbols. 1-Butanol was not detected from acetaldehyde electrolysiso nTi, Cr,Mo, Pd, and Pt.