Asymmetrical electrohydrogenation of CO₂ to ethanol with copper–gold heterojunctions

Significance

Electrochemical CO₂ reduction is a promising strategy to store renewable electrical energy in the chemical bonds of carbon fuels. Currently, high performance has been reported for C₁ products, like CO and formate, which are selectively produced by a number of catalysts through 2e⁻/2H⁺ steps. Due to shared reaction pathways, ethanol is always produced competitively with ethylene. A copper/gold heterojunction catalyst is reported here for the selective production of ethanol through asymmetrical electrohydrogenation of CO₂. It is shown that bimetallic Cu/Au exhibits a 200-fold enhancement in the ratio of ethanol to ethylene compared with a single Cu catalyst.

Abstract

Copper is distinctive in electrocatalyzing reduction of CO₂ into various energy-dense forms, but it often suffers from limited product selectivity including ethanol in competition with ethylene. Here, we describe systematically designed, bimetallic electrocatalysts based on copper/gold heterojunctions with a faradaic efficiency toward ethanol of 60% at currents in excess of 500 mA cm⁻². In the modified catalyst, the ratio of ethanol to ethylene is enhanced by a factor of 200 compared to copper catalysts. Analysis by ATR-IR measurements under operating conditions, and by computational simulations, suggests that reduction of CO₂ at the copper/gold heterojunction is dominated by generation of the intermediate OCCOH*. The latter is a key contributor in the overall, asymmetrical electrohydrogenation of CO₂ giving ethanol rather than ethylene.

An important approach to global warming, and the energy crisis (1), is based on electrochemical reduction of CO₂ with solar, wind, or other renewable energies (2–4). Of the reduction products, the C₂₊ chemicals, ethylene and ethanol, have high energy densities with significant global market values (5–7). High faradaic efficiencies and current densities have already been achieved for CO₂ reduction to 2e⁻ products, like CO and formate based on a number of catalysts (8–11). However, copper-based catalysts provide significant examples of metallic catalysts that generate C₂₊ products (12–14). There are complications in this chemistry from shared reaction pathways. Key intermediates are produced that can lead to ethylene or ethanol competitively (13, 15). Enhanced ethylene production has been achieved by implementing Cu-based catalysts via materials engineering (16–18). Efficient ethanol production by CO₂ reduction still remains a major challenge (19–21).

Recently, new strategies have appeared for improving CO₂ reduction to ethanol. Ren et al. used an oxide-derived Cu_xZn catalyst that enhanced CO₂ electroreduction to ethanol. A two-site mechanism was proposed in which CO, generated at zinc sites, transferred to neighboring copper sites where C–C coupling occurs to form ethanol (15). Similarly, Morales-Guio et al. investigated the synergistic effect of a bimetallic Au/Cu catalyst for promoting ethanol formation (22). In a related study, Ag was also introduced to manipulate Cu surfaces. Addition of Ag enhances the diversity of binding sites which disrupts the coordination environment for intermediates on pathways for ethylene formation (19). However, ethanol selectivity in these studies was < 30%, far lower than industrial requirements.

A recent approach, with high ethanol selectivity, was reported by Xu et al. (23). It was based on an atomically dispersed copper catalyst, but its low current density, 1.23 mA cm⁻², was well below that needed for industrial requirements of > 200 mA cm⁻². In accelerating the rate in aqueous solutions, a gas-phase CO₂ supply could help to overcome slow CO₂ diffusion (24). New approaches are required to improve both ethanol selectivity and current density (25, 26).

The approach taken here was to focus on bimetallic Cu–Au catalysts with the utilization of density functional theory (DFT) to examine possible reaction pathways with an electrostatic self-assembly strategy to fabricate Cu/Au heterojunctions. Comparing single Au, single Cu, CuAu alloys, and isolated Cu–Au mixtures reveals that only Cu/Au heterojunctions display a high faradaic efficiency for ethanol (~60%). With a gas diffusion electrode, current densities as high as ~500 mA cm⁻² were observed that exceed values for practical CO₂ electrolysis. In ethanol synthesis, in situ ATR-IR analysis has detected the key intermediate, OCCOH*, consistent with the results of DFT calculations.

Results

Previous reports have shown that ethylene is the main C₂ product on Cu. CO is the main product on a CuAu alloy (27, 28). There are also trace amounts of ethanol with the bimetallic CuAu catalyst (22). In the current study, Cu/Au heterojunctions are compared with Cu and CuAu alloy for reactivity comparisons.

Reaction pathways that lead to key products from CO₂ on CuAu, Cu/Au, and Cu as catalysts are shown in Fig. 1A and SI Appendix, Figs. S1 and S2. From the data, CO is a key intermediate in the preparation of nonformate products. On the surface of CuAu, CO* binding is relatively weak, while CO* adsorption on both Cu and Cu/Au is more highly favored.

Fig. 1.

(A) Energy diagrams for CO₂ reduction on CuAu, Cu/Au, and Cu. (B) Histograms showing formation energies (eV) for OCCO* and HCO* on CuAu, Cu/Au, and Cu. Free energy diagram for the electrochemical reduction of CO₂ on (C) Cu/Au and (D) Cu. (E) An energy diagram for generating C₂H₅OH from CO₂ that features proton-coupled electron transfer on Cu/Au with Cu, brown; Au, yellow; C, gray; O, red; and H, white.

Reaction pathways involving formation of OCCO* and HCO* from CO are shown in Fig. 1B. The data for Cu and CuAu show that the appearance HCO* as an intermediate is in competition with OCCO* with the relative energies of 1.79 and 1.38 eV. At Cu/Au heterojunctions, direct CO coupling is favored (ΔG = 0.52 eV). C₂H₄ is more easily prepared at Cu by 2HCO* coupling through the intermediate CHOCHO*. The interface, Cu/Au, appears to give C₂H₅OH by asymmetric hydrogenation of OCCO* through the intermediate OCCOH*.

Free energy profiles for the reduction of CO₂ to ethylene and ethanol over Cu/Au and Cu are shown in Fig. 1 C and D. As shown in Fig. 1C, a reaction pathway, that includes as an intermediate OCCOH* for C₂H₅OH production at Cu/Au heterojunctions, occurs in the occurs in the sequence: C​O_2​→C​O_2​*→​CO*​→2C​O*→​OCC​O*→​OCC​OH*​→OC​C*→​OCC​H*→​OCC​H_2​*→​OCC​H_3​*→​OCH​CH_​3*​→HO​CHC​H_3​*→HOCH₂CH₃*→C₂H₅OH.

The pathway leading to the appearance of C₂H₄ on Cu has different properties. In ethylene formation, a key intermediate is CHOCHO* rather than OCCOH*. Once formed, the latter is followed by sequential, symmetric hydrogenation to give C₂H₄ through the reaction sequence: C​O_2​→C​O_2​*→​CO*​→2C​O*→​2HC​O*→​CHO​CHO​*→C​HOC​HOH​*→C​HOH​CHO​H*→​CHC​HOH​*→CH₂CHOH*→C₂H₄ + H₂O (Fig. 1D).

Reactivity at Cu/Au interfaces is shown in Fig. 1E. At the surface, selective formation of C₂H₅OH at Cu/Au originates based on the asymmetrical adsorption geometry of OCCO*. In the intermediate, one carbon atom is bound to Au and the second to Cu with enhanced adsorption energy on Cu. The combined interaction between sites leads to preferential hydrogenation of OCCO* at the carbon atom once it is bound to Cu to give the intermediate OCCOH*.

In the overall reaction pathway, asymmetrical electrohydrogenation leads to an OCCH₃ intermediate. The latter dictates the major product, ethanol. In the scheme, the Cu/Au heterojunctions play a major role in ethanol generation during the reduction of CO₂.

Cu/Au heterojunctions were fabricated through well-designed electrostatic self-assembly of positively charged copper and negatively charged gold nanoparticles, as shown in Fig. 2A (29, 30). High-resolution, transmission electron microscopic (HRTEM) images (SI Appendix, Fig. S3) of Cu/Au electrodes provide evidence for definite crystalline planes with an interplanar spacing of 2.32 Å and 2.39 Å. The latter correspond to the facets of CuO (111) and Au (111). Analysis of the aberration-corrected transmission electron microscopic images led to characterization of the interfacial structure between Cu and Au particles.

Fig. 2.

(A) Schematic illustration of the fabrication of Cu/Au nanoparticle. (B) Aberration-corrected TEM images of Cu/Au heterojunctions with an illustrative, schematic diagram showing a structural schematic diagram in the Inset. HRTEM images and structural schematic diagrams (Inset) for (C) Au nanoparticles, (D) Cu nanoparticles, (E) CuAu alloy, and (F) isolated Cu–Au particles.

As shown in Fig. 2B, the atomic arrangements of Cu and Au particles cross each other at the interface and the further EDX analysis shows a clear interfacial connection between copper and gold (SI Appendix, Fig. S4), consistent with formation of Cu/Au heterojunctions during the synthesis of a tandem Cu/Au interface. TEM images in Fig. 2 C–F show that the synthesis of control samples including Au, Cu, and CuAu alloys and isolated Cu–Au composites, all loaded on carbon substrates.

X-ray diffraction patterns are shown in SI Appendix, Fig. S3. For the Cu/Au sample, three well-defined diffraction peaks at 2θ ≈ 35.5°, 38.7°, and 48.7° were attributable to the (11-1), (111), and (20-2) planes of CuO (PDF: 48-1548), respectively. Diffraction peaks at 2θ ≈ 38.1°, 44.4°, 64.6°, 77.5°, and 81.7° were assignable to the (111), (200), (220), (311), and (222) planes of Au (PDF: 04-0874), respectively. Analysis of the data was consistent with Cu/Au heterojunctions that consist of Cu and Au phases.

Electrocatalytic carbon dioxide reduction was investigated in a H cell under CO₂-saturated, 0.1 M KHCO₃ aqueous solutions. As shown in SI Appendix, Fig. S5, CO was identified as a major product on Au and CuAu alloys and isolated Cu–Au mixtures, and C₂H₄ as the main C₂ product on Cu catalytic surfaces. Ethanol was the main C₂ product on Cu/Au heterojunctions with a maximum ethanol Faradaic efficiency reached at −0.8 V. Only trace amounts of ethanol were detected on Cu and separation Au–Cu samples. Notably, the Faradaic efficiency ratio of ethanol over ethylene on Cu/Au heterojunctions was significantly enhanced by a factor of 200 compared with Cu at −0.8 V (83 vs. 0.41), as shown in Fig. 3A. The latter is consistent with ethanol formation largely promoted on Cu/Au heterojunctions.

Fig. 3.

CO₂RR performances in H cells: (A) Ratios of Faradaic efficiency ratios for ethanol to ethylene on Cu, Cu–Au, and Cu/Au in 0.1 M CO₂-saturated KHCO₃ aqueous solutions at −0.8V vs. RHE. (B) Faradaic efficiencies (%) for the appearance of ethanol at Cu, Au, CuAu, Cu–Au, and Cu/Au in 0.1 M CO₂-saturated KHCO₃ aqueous solutions as a function of applied potential. CO₂RR performances in flow cell: (C) Partial current densities for producing ethanol at Cu/Au heterojunctions and Cu particles in 1 M KOH electrolyte at various IR-corrected potentials. (D) Long-term electrode stability test of the GDE with the best Cu/Au catalysts (at a ratio of 1/2) in 1 M KOH electrolyte at a constant applied potential of −0.75 V vs. RHE.

In previous studies, it was claimed that in tandem Au/Cu catalysis, Au nanoparticles reduce CO₂ to CO with the latter migrating to neighboring Cu surfaces for C–C coupling (22). As shown in SI Appendix, Fig. S4B, with CO as the feed gas, ethanol production on Cu was increased relative to CO₂, but the Faradaic efficiency for ethanol was still low (~10%). The decrease points to CO, not necessarily a key contributor to high ethanol selectivity.

The optimal catalyst loading on carbon substrates was found to be 40% giving an ethanol Faradaic efficiency of 43% shown in SI Appendix, Figs. S6 and S7. The impact of the Cu/Au catalyst was further evaluated by varying the Au to Cu ratio. Decreasing the ratio of Cu to Au leads to an increase in ethanol Faradaic efficiency that reaches 50% at a ratio of 1/2 (SI Appendix, Fig. S7). Based on the results of high-resolution X-ray photoelectron spectroscopic analysis (SI Appendix, Fig. S8), the more intense Cu⁺/Cu⁰ peak that appears in the spectrum, with a decrease in lattice oxygen, shows that oxygen vacancies are generated on the partially reduced sample (31). As shown in Fig. 3B, the Faradaic efficiency for ethanol was improved to 60% at −0.8 V for partially reduced Cu/Au heterojunctions.

In CO₂ reduction at gas diffusion electrodes (GDE), the gaseous CO₂ feed helps to overcome slow CO₂ transport. The latter is inherent in H cells and limits reaction rates (32). To demonstrate the behavior of the Cu/Au heterojunctions under industrial conditions, we evaluated CO₂ reduction in a flow cell coupled with a gas diffusion electrode (SI Appendix, Fig. S9). Gas-diffusion electrodes, that use gaseous reactants such as oxygen or carbon dioxide, can be used with vapor-fed injection in electrolytically wetted porous structures. Alterations in local reaction microenvironments can result in enhanced hydroxide activities at the electrode surface where water activity is lower. In proton-consuming CO₂ reduction, variations in pH can also lead to variations in reaction products.

Our investigation of CO₂ reduction utilized a GDE with a high ethanol Faradaic efficiency of 60% in 1 M KOH at −0.75 V, where liquid products were determined by NMR spectra shown in SI Appendix, Fig. S10. SI Appendix, Figs. S11 and S12 summarize Faradaic efficiencies (%) for all reduction products from CO₂ reduction for flow cell tests in 1 M KOH and 0.1 M KHCO₃, respectively. Equal maximum Faradaic efficiencies of ethanol (~60%) were observed in both electrolytes due to similar local chemical environments (including CO₂ and bicarbonate) created on electrode surface during CO₂RR operation. Partial current densities for ethanol formation, as a function of applied potential, are shown in Fig. 3C. The ethanol partial current density of ~300 mA cm⁻² on Cu/Au, at −0.75V vs. RHE, is a factor of 4 higher than at Cu. Fig. 3D shows a long term of electrode stability for over 90 h with an average current density of ~500 mA cm⁻² at −0.75 V. SI Appendix, Fig. S13 summarizes the state-of-the-art in CO₂ reduction in the production of ethylene, ethanol, carbon monoxide, and formate. As shown in SI Appendix, Table S1, our Cu/Au heterojunctions output the highest ethanol partial current density among the reported catalysts. We further calculated carbon mass balance in flow cells in 1 M KOH (SI Appendix, Fig. S14) and found that 7.5% of carbon was converted to CO₂RR products, which is close to the values (~7.6%) reported in the literature (33). In addition, the cathode energy efficiency was calculated to be 39.8% at an applied potential of −1 V vs. RHE (without considering any IR compensation).

Discussion

Evaluation of the ethanol formation mechanism at Cu/Au heterojunctions utilized potential-dependent, ATR-IR measurements (SI Appendix, Fig. S15) showing evidence for the key intermediate OCCOH* as predicted by the DFT calculations. Fig. 4A shows potential-dependent, absorption spectra for Cu/Au in CO₂-saturated 0.1 M KHCO₃ with bands appearing from 2,250-1,000 cm⁻¹. The corresponding spectra for Au and Cu were given in SI Appendix, Figs. S16 and S17, respectively. A band at 1,640 cm⁻¹ appears arising from a O–H bending mode for H₂O (34). A reversible band at 1,225 cm⁻¹ appears from an O–Si–O stretching vibration in the SiO₂ overlayer on a silicon prism substrate.

Fig. 4.

Potential-dependent, attenuated total reflection infrared (ATR-IR) spectra on Cu/Au heterojunctions. (A) Full-range spectra on Cu/Au heterojunctions in CO₂-saturated 0.1M KHCO₃ from 0.2V vs. RHE to −0.8V vs. RHE. (B) Full-range spectra on Cu/Au heterojunctions in CO₂-saturated 0.1M KHCO₃ (orange) and hydroxypropanone (blue) at −0.4V vs. RHE. (C) Full-range spectra on Cu/Au heterojunctions with H₂O as electrolyte (orange) and with D₂O as electrolyte (dark green) in CO₂-saturated 0.1M KHCO₃ at −0.4V vs. RHE. (D) Full-range spectra on Cu/Au heterojunctions with D₂O as electrolyte in CO₂-saturated 0.1M KHCO₃ from 0V vs. RHE to −0.8V vs. RHE.

Increasing the potential of the electrode negatively past −0.8 V led to the accelerated appearance of gaseous products. Under these conditions, the catalyst layer was detached from the Si surface. Surface measurements showed that loss of the catalyst layer was accompanied by the appearance of surface O–Si–O vibrations (35). A band at 2,110 cm⁻¹ was assignable to a line-bound CO (CO_L). The latter shifts to lower energies at negative scans (SI Appendix, Fig. S18), consistent with the appearance of a Stark effect, as described in earlier reports.

As shown in Fig. 4A, the bands at 1,640 and 1,109 cm⁻¹ are sensitive to changes in scan rates. They appear at negative potentials and require CO₂ for their appearance during CO₂ reduction. As predicted earlier, they may be attributed to OCCOH*. As shown in the DFT calculations in SI Appendix, Fig. S19, in the spectrum of OCCOH*, a C=O stretch appears at 1,593 cm⁻¹ and a C–OH band at 1,117 cm⁻¹. The C–OH band is consistent with the experimental result at 1,109 cm⁻¹, and the C=O band with the experimental result at 1,640 cm⁻¹.

Hydroxypropanone with a carbonyl group adjacent to hydroxyl (OCCOH) was selected as a surrogate to further prove the assignment of the bands cited above to the key intermediate OCCOH*. As shown in Fig. 4B, a band at ~1,099 cm⁻¹ detected on Cu/Au heterojunction during CO₂RR at −0.4V is consistent with the C–OH stretch of hydroxypropanone. Unfortunately, water exhibits a wide O–H bending band in the 1,700-1,550 cm⁻¹ range, which may overlay the signal of C=O vibration. To prove the point, CO₂RR on a Cu/Au heterojunction was performed in heavy water D₂O where O-D bending peak shifts to around 1,150 cm⁻¹. In D₂O solution shown in Fig. 4C, a band around 1,685 cm⁻¹ was still observed, which is consistent with C=O stretching of hydroxypropanone. This suggests that in Fig. 4A, the peak at 1,640 cm⁻¹ indeed originates from the C=O stretch. Moreover, in Fig. 4D, the C=O stretch at ~1,680 cm⁻¹ in D₂O solution, increases under more negative potentials, and shows the same trend with the C–OH band at ~1,109 cm⁻¹ (increasing with more negative potentials in H₂O electrolyte) as shown in Fig. 4A. These results clearly indicate the formation of the intermediate OCCOH* during CO₂ reduction over Cu/Au heterojunctions, while the intermediate was not detected over single Cu and single Au samples.

Based on the DFT analysis, the intermediate OCCOH* adsorbs at Cu/Au heterojunctions with a single carbon atom bound to Au and the other to Cu. The high hydrogenation activity at Cu is caused by the carbon atom bound to Cu (C=O). The latter forms C–OH which undergoes hydrogenation at more negative potentials consistent with a strong interaction between C–OH and Cu. By contrast, the C=O interaction with Au tends to lead to desorption at more negative potentials (~−0.8 V).

Conclusion

We have described an optimized procedure for the electrocatalytic reduction of CO₂ to ethanol based on the rational design and modification of bimetallic catalysts. Electrocatalytic results show that Cu/Au heterojunctions can lead to the selective conversion of CO₂ to ethanol in 60% yield, a result in agreement with the results of DFT calculations. Exploitation of the reactivity toward CO₂ was extended by using a GDE, resulting in a long-term electrode stability of over 90 h at 500 mA cm⁻².

Utilization of in situ ATR-IR spectroscopy reinforces the results of computational calculations with both pointing to the appearance of the intermediate OCCOH* as an important element in the asymmetrical electrohydrogenation of CO₂ to ethanol. The overall results presented here also highlight the fact that utilization of Cu/Au heterojunctions can occur under conditions appropriate for the formation of ethanol under conditions appropriate for industrial production.

Materials and Methods

Full experimental details and procedures for the synthesis of the samples used in the present study are described in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.