Selective electroreduction of acetylene to 1,3-butadiene on iodide-induced Cu^δ+–Cu⁰ sites

Electrifying chemical processes that produce multicarbon molecules is a pivotal step towards the much-needed decarbonization of chemical industries¹. Towards this end, those chemical processes that most impact our global energy–carbon nexus must be first identified. One such process is the steam cracking of naphtha to ethylene, which also yields 1,3-butadiene (1,3-BD) as a by-product². During naphtha steam cracking, 12–16 kg of 1,3-BD are obtained alongside every 100 kg of ethylene^2,3. 1,3-BD is a vital feedstock used in the global manufacturing of synthetic rubber and engineering plastics⁴. In recent years, however, the production of ethylene from the dehydrogenation of ethane in shale gas has proven more cost effective⁵. This has resulted in a dramatic decrease in naphtha cracking. Although 1,3-BD can also be formed from the ethane feedstock, only 1–2 kg of it are now produced per 100 kg of ethylene^6,7. 1,3-BD can also be synthesized by the oxidative dehydrogenation of butane or the dehydrogenation and condensation of ethanol over silicon and magnesium oxide catalysts^8,9,10. All the above methods are energy intensive and require high temperatures and pressures to operate. From this perspective, a more environmentally benign strategy to synthesize 1,3-BD is highly desirable.

In recent years, acetylene has been considered as a precursor to synthesize value-added chemicals and polymers¹¹. We describe here an approach to synthesize 1,3-BD by the electrochemical reductive coupling of acetylene (2C₂H₂(g) + 2H₂O + 2e⁻ → C₄H₆(g) + 2OH⁻, E⁰ = +0.98 V versus standard hydrogen electrode (SHE) at pH 7.0; Supplementary Note 1). C₂H₂ electroreduction (e-C₂H₂R) is potentially more sustainable because it can be conducted under relatively mild conditions, uses water as the hydrogen source for reduction, and can be powered by renewable electricity. Additionally, the acetylene feedstock¹¹ can be synthesized from carbon dioxide or methane^2,12,13, the top two greenhouse gases in the atmosphere¹⁴. Thus far, previous works have been devoted to e-C₂H₂R to ethylene (C₂H₂(g) + 2H₂O + 2e⁻ → C₂H₄(g) + 2OH⁻, E⁰ = +0.40 V versus SHE at pH 7.0)^{15,16,17,18,19,20,21,22}. In the studies that showed 1,3-BD formation, typically on copper catalysts, it was formed as a by-product with current densities smaller than −4 mA cm⁻², which are insufficient for industrial adoption^{23,24,25,26,27}. Synthesizing 1,3-BD with high yields is particularly challenging because at larger overpotentials *C₂H₃, a key intermediate for 1,3-BD formation, hydrogenates to *C₂H₄ and desorbs from the catalyst surface as C₂H₄(g), instead of undergoing coupling to form *C₄H₆ (ref. ²⁶).

In this work, we reveal a strategy to electrolyse C₂H₂ to 1,3-BD, forming it as a major product with high selectivity. Our approach is to introduce Cu^δ+–Cu⁰ sites on a copper catalyst, the former having been found to promote the coupling of CO intermediates formed during the electroreduction of CO₂ (CO₂RR) to C₂₊ products^28,29,30. Using in situ Raman spectroscopy and X-ray absorption spectroscopy (XAS), we show that the formation of stable Cu^δ+–Cu⁰ active sites can be promoted using potassium iodide (KI) electrolyte^31,32,33. On these sites, acetylene was reduced to 1,3-BD with a maximum Faradaic efficiency (FE_1,3-BD) of 93% at −0.85 V versus SHE and a partial current density (j_1,3-BD) of −75.0 mA cm⁻² at −1.0 V versus SHE. These figures of merit are substantially higher than those obtained during e-C₂H₂R in KCl, KBr and K₂SO₄ electrolytes. Using density functional theory (DFT) calculations, we show that the presence of iodide helps preserve the partially oxidized copper sites. In addition, coadsorbed iodide species on the partially oxidized copper sites lead to favourable thermodynamics and surmountable coupling barriers for acetylene reduction to 1,3-BD.

Characterization of Cu₂O-nanocube-derived catalyst

We prepared Cu₂O nanocubes (Cu₂O_NC) using a previously reported method³⁴. Scanning electron microscopy (SEM) of the freshly prepared catalyst showed that it consisted of ~30 nm cubes (Fig. 1a). This catalyst was then used for e-C₂H₂R in 1 M KI electrolyte at a representative potential of −1.0 V versus SHE (all potentials stated hereafter are referenced to the SHE). After use for e-C₂H₂R, the particles’ cubic morphology was largely lost and ~60 nm agglomerates were observed instead (Fig. 1b). The morphologies of the catalyst particles after they were used for e-C₂H₂R in K₂SO₄, KCl and KBr electrolytes were also similar (Fig. 1c–e). However, double-layer capacitance measurements indicate that the roughness factors of the particles in KI were ~1.1–1.2× higher than those in K₂SO₄, KCl or KBr electrolytes (Fig. 1f and Supplementary Table 1). This is in line with previous works showing the roughening of copper in the presence of I⁻ (refs. ^35,36,37).

a–e, SEM images of Cu₂O_NC catalyst (a), before (green background) and after e-C₂H₂R (pink background) in KI (b), K₂SO₄ (c), KCl (d) and KBr (e) electrolyte. f, Double-layer capacitance (C_dl) measurements of Cu₂O_NC-derived catalysts after e-C₂H₂R. The plots denote the capacitive current density versus scan rate. The C_dl was measured in 0.5 M K₂SO₄. Each error bar represents the standard deviation (s.d.) of three independent measurements. g, X-ray diffractograms of Cu₂O_NC catalyst before and after e-C₂H₂R at −1.0 V versus SHE in KI and K₂SO₄. The X-ray diffractograms were assigned in accordance with PDF 00-003-0892 for Cu₂O and PDF 00-001-1242 for copper. h, In situ Raman spectra of Cu₂O_NC catalyst at OCP with C₂H₂, after 10 min of electrolysis at −1.0 V versus SHE in KI and K₂SO₄, and after electrolysis. Inset: enlarged region of the Cu₂O Raman peaks under OCP conditions; the Raman spectra (i) and (ii) indicate the absence and presence of oxide peaks, respectively, during e-C₂H₂R in KI. A graphite disc was used as a substrate for characterization.

Full size image

X-ray diffraction (XRD) of Cu₂O_NC revealed signals corresponding to those of Cu₂O (Fig. 1g). After Cu₂O_NC was used as an e-C₂H₂R catalyst in 1 M KI electrolyte at −1.0 V, two additional XRD peaks were observed at 43.5° and 50.7°, which we ascribe respectively to the (111) and (200) diffractions of metallic copper. Intense Cu₂O diffractions were nonetheless still measured, which could be due to the Cu₂O_NC not being fully reduced at −1.0 V. This observation is consistent with a previous work, in which Cu₂O XRD peaks from Cu₂O_NC were still present after their use for CO₂RR at −1.55 V versus SHE in KHCO₃ (pH 6.8)³⁸. Cu₂O and copper diffraction peaks were similarly observed on the Cu₂O_NC-derived catalysts after they were used for e-C₂H₂R in K₂SO₄, KCl and KBr electrolytes (Fig. 1g and Supplementary Fig. 1). We note that the surface copper atoms could have also oxidized to Cu₂O during sample transfer for XRD analysis. Therefore, we resorted to in situ Raman and XAS to determine how the Cu₂O_NC catalyst evolves during electrolysis.

In situ Raman spectroscopy was conducted on the Cu₂O_NC catalyst during C₂H₂ reduction in 1 M KI electrolyte (Fig. 1h and Supplementary Table 2). At open circuit potential (OCP), Cu₂O was identified, as evident by its Raman peaks at 148, 218, 522 and 622 cm⁻¹ (ref. ³⁹). When a potential of −1.0 V was applied, these peaks were largely attenuated, although still detectable. Specifically, the 522 and 622 cm⁻¹ peaks could be observed in spectrum (ii) in Fig. 1h. This indicates that Cu₂O was not fully reduced to metallic copper. Bands at 1,134, 1,530 and 1,695 cm⁻¹ were also measured. The 1,695 cm⁻¹ signal could be ascribed to the ν(C≡C) stretching vibration of adsorbed C₂H₂ (ref. ²⁶), while those at 1,134 and 1,530 cm⁻¹ could be attributed to polyacetylene⁴⁰. No peaks corresponding to CuI, adsorbed 1,3-BD, *C₂H₃ or other reduced species from e-C₂H₂R were observed, which could be due to their short lifetimes, small surface population or the insufficient limits of detection afforded by our spectrometer (Supplementary Fig. 2).

In situ Raman measurements during e-C₂H₂R in 0.5 M K₂SO₄ electrolyte were also conducted at −1.0 V (Fig. 1h). While we detected the ν(C≡C) stretching signal of C₂H₂ at 1,695 cm⁻¹ and polyacetylene Raman peaks at 1,134 and 1,530 cm⁻¹ (refs. ^26,40), signals that could be attributed to Cu₂O were not detected. Similar observations were obtained from C₂H₂ electrolysis in 1 M KCl and KBr electrolytes (Supplementary Fig. 3). This indicates that the use of KI electrolyte aids in preserving the Cu₂O species during electrolysis. Indeed, based on predictions from the Pourbaix diagrams of Cu–H₂O–X systems (X = I⁻, Cl⁻, Br⁻ or SO₄²⁻), the reduction potential of Cu⁺ to Cu⁰ in the presence of I⁻ is at least 0.15 V more negative compared with Cl⁻, Br⁻ and SO₄²⁻ (refs. ^41,42,43).

When the applied negative potentials were removed, peaks at 218, 522 and 622 cm⁻¹ were observed within 2 min on all the surfaces (Fig. 1h and Supplementary Fig. 3). This signifies that metallic copper reoxidized to Cu₂O. The signals of polyacetylene at 1,020, 1,134, 1,280 and 1,530 cm⁻¹ were also observed post electrolysis after the C₂H₂ gas flow was removed^25,40, indicating that the formed polyacetylene remained on the catalyst surface during electrolysis.

The oxidation state of the Cu₂O_NC catalyst during e-C₂H₂R was monitored using in situ XAS in fluorescence mode (Fig. 2 and Supplementary Fig. 4). The Cu K-edge X-ray near-edge absorption spectrum (XANES) of a freshly prepared Cu₂O_NC catalyst corresponds to that of a Cu₂O standard (Fig. 2a). During e-C₂H₂R in 1 M KI electrolyte at −1.0 V, the rising edge spectrum of the Cu₂O_NC-derived catalyst shifted towards the position observed for a metallic copper foil standard, and the absorption edge shape resembled that of metallic copper³⁶. However, the rising edge energies still reside between those of pristine copper and Cu₂O, indicating that the oxidation state of the copper catalyst lies between those of Cu⁰ and Cu⁺. Interestingly, when we conducted in situ XAS of the Cu₂O_NC-derived catalyst in KCl, KBr and K₂SO₄ electrolytes, the rising edge energies of copper were notably closer to that of copper foil. A linear combination fitting of the XANES spectra further showed that the amount of Cu⁺ in the catalyst is the highest in KI (31%) compared with those observed in KCl (8%), KBr (8%) and K₂SO₄ (6%) electrolytes (Supplementary Table 3).

a,b, In situ XANES (a) and EXAFS (b) spectra acquired on Cu₂O_NC catalyst at OCP during e-C₂H₂R at −1.0 V versus SHE in 1 M KI, KCl, KBr and 0.5 M K₂SO₄ electrolytes. Copper foil and Cu₂O were used as standards for comparison. Inset, a: enlarged pre-edge region from 8,970 to 8,979 eV.

Full size image

In situ extended X-ray absorption fine structure (EXAFS) analysis was used to determine changes in the local chemical environment of copper during e-C₂H₂R. Peaks corresponding to Cu–O and metallic Cu–Cu bonds were observed at 1.4 and 2.2 Å, respectively (Fig. 2b)⁴⁴. Strikingly, the in situ EXAFS spectra measured in KCl, KBr and K₂SO₄ electrolytes showed Cu–O peak intensities to be at least 50% less intense than in KI. Instead, the Cu–Cu peak was the more dominant feature observed, indicating that metallic Cu⁰ is the main species present in KCl, KBr or K₂SO₄ electrolytes.

Overall, both operando Raman spectroscopy and XAS indicate that the presence of I⁻ preserved a larger amount of oxidized copper species in the copper catalyst during e-C₂H₂R. This finding is in good agreement with a previous XANES study showing that iodide-modified copper contained a higher concentration of Cu⁺ sites compared with other halide-modified copper materials³⁶.

Electrocatalytic activity of Cu₂O_NC for e-C₂H₂R

The Cu₂O_NC catalyst was loaded (5 mg cm⁻²) onto a carbon gas diffusion layer and used as an electrode for e-C₂H₂R in a flow cell (Supplementary Fig. 5). Unless specified, all current densities stated were normalized to the geometric surface area of the exposed electrode, which was 1 cm². Nickel foam and an Ag/AgCl electrode were used as the counterelectrode and the reference electrode, respectively. Linear sweep voltammetry (LSV) was first conducted on the Cu₂O_NC electrode in a 1 M KI catholyte and a 1 M KOH anolyte (Fig. 3a). In the presence of C₂H₂, larger currents were measured starting from −0.85 V, as compared to that in N₂ gas. This suggests that the Cu₂O_NC is catalytically active for e-C₂H₂R.

a, LSV measurements of Cu₂O_NC catalyst in N₂ or 50% C₂H₂ gas feedstock. Scan rate, 50 mV s⁻¹. b, FEs and partial current densities for 1,3-BD, C₂H₄, butenes and H₂ from e-C₂H₂R as a function of the applied potential. Each electrolysis measurement lasted 50 min. Each error bar represents the s.d. of three independent measurements. c, Comparison of FEs and partial current densities for 1,3-BD with those of previously reported catalysts^{23,24,25,26,27}. The potentials were taken with reference to the SHE. d, FEs and partial current densities from an 11 h stability experiment. A constant current of −12 mA cm⁻² was applied over the course of the experiment. A 2 min oxidation pulse of +0.4 V versus SHE was introduced at the 9 h mark to remove polyacetylene from the catalyst surface. The electrolyte used in these measurements was 1 M KI.

Full size image

We next investigate the effects of various C₂H₂ gas concentrations on the e-C₂H₂R performance at −1.0 V in a 1 M KI catholyte (Supplementary Table 4). The gas feedstock consisted of C₂H₂ balanced with N₂ gas, with a total flow rate of 50 sccm into the electrolytic cell. The gaseous products were analysed using gas chromatography (Supplementary Fig. 6). When the C₂H₂ concentration increased from 5% to 50%, the FE_1,3-BD increased from 58% to 72%, while the FE_C2H4 decreased from 41% to 28%. We attribute the increased selectivity for 1,3-BD at larger C₂H₂ concentrations to the higher coverage of C₂H₂ on the catalyst surface, which promotes C–C coupling to form more 1,3-BD. However, no further increase in FE_1,3-BD was observed in 100% C₂H₂. We also conducted e-C₂H₂R at different catalyst mass loadings of 1.0, 2.5 and 5.0 mg cm⁻² and found that the j_1,3-BD was 3–5 times higher with a 5.0 mg cm⁻² catalyst loading (Supplementary Table 5). Since the optimal performance was attained at a C₂H₂ concentration of 50% and a catalyst loading of 5.0 mg cm⁻², we utilized these conditions in all subsequent experiments.

The activity and selectivity of e-C₂H₂R on the Cu₂O_NC-derived catalyst in KI electrolyte were quantified at constant potentials from −0.85 to −1.4 V. From −0.85 to −1.3 V, the major products formed were 1,3-BD and C₂H₄ (Fig. 3b and Supplementary Table 6). We found that at −0.85 V, the FE_1,3-BD was as high as 93%, with a j_1,3-BD of −8 mA cm⁻²; the remaining FE is due to C₂H₄. At more negative potentials, FE_1,3-BD decreased while FE_C2H4 increased, although the partial current densities for both compounds continued to increase. The maximum j_1,3-BD attained was −75 mA cm⁻² at −1.0 V. The FE_1,3-BD at −0.85 V and the j_1,3-BD at −1.0 V in KI electrolyte appear to be the highest reported for the formation of 1,3-BD via e-C₂H₂R (Fig. 3c). It is noteworthy that the j_1,3-BD obtained in this work was at least 20-fold higher than those in previous works^{23,24,25,26,27}. Minor products such as 1-butene, trans-2-butene and cis-2-butene were also detected, with the highest FE of 3.9% and j of ∼−13 mA cm⁻² for these butenes occurring at −1.3 V, suggesting that applying a more negative potential further reduces 1,3-BD to these C₄ olefins. We note that although n-butane and isobutane are possible products, we did not detect these even at a negative applied potential of −1.3 V (Supplementary Fig. 8). Trace amounts of ethane (FE, 0.2–0.3%) were detected starting from −1.2 V. Due to the relatively positive potentials applied, H₂ gas was only produced from the parasitic hydrogen evolution reaction at −1.3 V (ref. ⁴⁵). Overall, the total FEs of the products ranged from 97% to 105%, indicating that all the electrochemically formed products have been accounted for. We note that although acetylene undergoes polymerization in the presence of copper to form polyacetylene^46,47, there is no associated electron transfer for that product. Thus, polyacetylene formation cannot be suitably incorporated in our FE analysis.

To ascertain that the aforementioned products were formed from e-C₂H₂R, we performed electrolysis at −1.0 V using a N₂ feedstock. We also held the working electrode at OCP in 50% C₂H₂ gas (Supplementary Fig. 6). Both measurements did not yield any carbon-containing products, which show that all the products measured in the preceding e-C₂H₂R measurements (excluding polyacetylene and H₂) were electrochemically formed from acetylene.

The catalytic stability for e-C₂H₂R in KI electrolyte was assessed using constant current electrolysis at −12 mA cm⁻² (Fig. 3d). The FE_1,3-BD was maintained at ~90 % over 8 h, while the j_1,3-BD was constant at ∼−11 mA cm⁻². The FE_1,3-BD decreased to ~85% at the 9 h mark, which we attribute to the formation of polyacetylene and the coking of the catalyst surface. To remediate this, we introduced an oxidation pulse of +0.4 V for 2 min to remove the carbonaceous compounds from the electrode surface⁴⁸. With this oxidation pulse, an FE_1,3-BD of ~90% could be sustained for over 11 h. We also assessed the stability at a higher current density of −25 mA cm⁻² and found that an FE_1,3-BD of 72% could be maintained for 6 h (Supplementary Fig. 9).

We then determined if utilizing another anolyte can affect the selectivity for 1,3-BD. When 1 M KI was used as both the anolyte and catholyte for e-C₂H₂R at −1.0 V (Supplementary Table 7), the FE_1,3-BD was 72% and the j_1,3-BD was −61 mA cm⁻², which were similar to the results obtained with a 1 M KOH anolyte. This indicates that the anolyte identity does not affect the production of 1,3-BD. We also found that the catholyte pH increased to 12.7 after electrolysis with a KI anolyte (Supplementary Table 8), which is comparable to the increase in pH observed after e-C₂H₂R with a KOH anolyte (catholyte pH increased from 7.0 to 12.6). This suggests that the increase in pH observed in non-buffering KI can be largely attributed to the production of OH⁻ species during e-C₂H₂R. We further conducted e-C₂H₂R in a two-electrode flow cell to evaluate the application prospects for 1,3-BD synthesis (Supplementary Table 9 and Supplementary Fig. 10). An FE_1,3-BD of 91% can be achieved at a cell voltage of −2.05 V. These two figures of merit are respectively similar to the highest FE_1,3-BD obtained in a three-electrode cell at −0.85 V (Fig. 3b) and our calculated cell voltage (Supplementary Note 2).

Importance of iodide for 1,3-BD production from e-C₂H₂R

We then systematically investigated the influence of iodide on 1,3-BD production. A representative potential of −1.0 V was applied for all experiments, unless otherwise mentioned. We first varied the concentration of iodide in the catholyte to be 0.25, 0.5 or 1 M. K₂SO₄ was added to the catholyte to keep the concentration of K⁺ at 1 M. We found that the FE_1,3-BD decreases from 72% in 1 M KI to 66% in 0.25 M KI. When only K₂SO₄ was used, the FE_1,3-BD was 56% (Fig. 4a and Supplementary Table 10). This signifies that the quantity of I⁻ present influences 1,3-BD production.

a, FEs for 1,3-BD and C₂H₄ during e-C₂H₂R at −1.0 V versus SHE in different concentrations of KI, b,c, Cation (b) and anion (c) electrolytes. In all experiments, the cation concentration was maintained at 1 M. In experiments without KI, 0.5 M K₂SO₄ was used instead. d, ECSA-normalized current densities of 1,3-BD at −1.0 V versus SHE in K₂SO₄, KCl, KBr and KI electrolytes. Each error bar represents the s.d. of three independent measurements.

Full size image

The type of cation and anion of the electrolyte has been found to impact the selectivity for coupled products during CO₂RR^34,35,49. Therefore, we studied the effects of changing the electrolyte cation on 1,3-BD production during e-C₂H₂R (Fig. 4b and Supplementary Table 11). The FE_1,3-BD and j_1,3-BD attained in NaI (74%, −71 mA cm⁻²) and CsI (74%, −74 mA cm⁻²) electrolytes were similar to those observed in KI electrolyte (72%, −75 mA cm⁻²), which indicates that the identity of the cation does not notably affect the formation of 1,3-BD. However, varying the type of anion yielded different selectivity and activity for 1,3-BD formation (Fig. 4c and Supplementary Tables 12–14). In K₂SO₄, KCl and KBr electrolytes, the FE_1,3-BD were remarkably similar at 56%, 54% and 55%, respectively, while the j_1,3-BD ranged from −45 to −47 mA cm⁻². Furthermore, the maximum j_1,3-BD in K₂SO₄, KCl and KBr electrolytes was only –60 mA cm⁻², which was attained at a more negative potential of –1.05 V and with a substantially lower FE_1,3-BD of 41–45%. We also conducted e-C₂H₂R at −1.0 V in KF electrolyte and found that the FE_1,3-BD (56%) and j_1,3-BD (−44 mA cm⁻²) were lower than those in KI (Supplementary Table 15). Overall, these results indicate that the presence of I⁻ species in the catholyte boosts 1,3-BD production, and that the overpotential to obtain a high FE_1,3-BD and j_1,3-BD is 0.05 V less negative. Interestingly, the FEs and j for butenes (1-butene, trans-2-butene and cis-2-butene) were ~2–6 times higher in K₂SO₄, KCl and KBr compared with in KI, indicating that the formation of these C₄ olefins is more favourable in the former electrolytes.

We now assessed whether improvements in 1,3-BD production were due to differences in the surface areas of the Cu₂O_NC-derived catalysts immersed in the various electrolytes. To do this, we normalized the j_1,3-BD with the electrochemically active surface areas (ECSAs) of the catalysts after they were used for e-C₂H₂R (Fig. 4d). The ECSA-normalized current density of 1,3-BD (j_{1,3-BD, ECSA}) produced from e-C₂H₂R in KI electrolyte at –1.0 V was −0.080 mA cm⁻², which is ~45–55% higher than the j_{1,3-BD, ECSA} in K₂SO₄ (–0.055 mA cm⁻²), KCl (–0.055 mA cm⁻²) and KBr (–0.051 mA cm⁻²) electrolytes. The higher j_{1,3-BD, ECSA} in KI indicates that the ECSA is not the crucial factor in improving 1,3-BD production, and that the Cu₂O_NC-derived catalyst is intrinsically more active for 1,3-BD production during e-C₂H₂R with I⁻. Our SEM analysis also showed that the catalysts have no clear differences in their morphologies regardless of the electrolytes used (Fig. 1a–e). Overall, we rule out differences in particle surface areas and morphology as contributing factors in 1,3-BD production.

Based on these results, we assert that enhancing the amount of Cu^δ+–Cu⁰ species is important in promoting 1,3-BD production from e-C₂H₂R. This is in line with previous studies which showed an increase in C₂₊ products formed by C–C coupling during CO₂ electrolysis in KI electrolyte^32,36,50,51. In these works, the coupled products formed were found to be correlated to the higher amount of Cu^δ+–Cu⁰ species in KI. We postulate that the adsorption of I⁻ on Cu₂O, which is the strongest among the halides^52,53, prevented the Cu^δ+–Cu⁰ species on the catalyst surface from being fully reduced, even under a negative potential. We note that it is experimentally difficult to accurately determine if Cu^δ+–Cu⁰ ensembles are the exact active sites for 1,3-BD formation. This is because such species may only be viewed by scanning tunnelling microscopy, which further requires well-ordered substrates, and ultrahigh-vacuum and low-temperature conditions for atomic-scale imaging⁵⁴. However, our e-C₂H₂R was conducted on powder catalysts, under ambient conditions and in an aqueous electrolyte. We further highlight, for the same aforementioned reasons, that the type of CO₂RR active sites on copper catalysts has also not been experimentally proven, even after years of extensive studies⁵⁵. Thus, we will now investigate, using theoretical calculations, the presence and impact of the Cu^δ+–Cu⁰ species and I⁻ on 1,3-BD production from e-C₂H₂R.

Computational modelling of 1,3-BD production

We performed DFT calculations at the level of the generalized gradient approximation to gain insight into the acetylene electroreduction pathway to 1,3-BD (Fig. 5 and Supplementary Fig. 11). Two acetylene molecules are sequentially adsorbed and hydrogenated to form a pair of *C₂H₃ moieties with water molecules from the environment acting as proton donors. These two moieties then couple to produce *C₄H₆, which desorbs as 1,3-BD (Fig. 5a). Since the pK_a of water is 11 orders of magnitude smaller than that of acetylene (14 versus 25), we did not consider the latter as a proton source. We also determined the effect of halides on fully metallic, partially oxidized and fully oxidized copper active sites. Cu(100) with a square-symmetry stripe of copper atoms on top was used to represent the fully metallic phase (Fig. 5b). To simulate a partially oxidized and defective electrode, we utilized the stripe with adsorbed oxygen, denoted as stripe + *O (Fig. 5d). This choice is justified by the high activity of (100) facets for making C–C bonds⁵⁶, and because oxide-derived copper active sites tend to be rough^57,58,59. For the Cu₂O phase, we considered a copper-terminated Cu₂O(111) slab, which has copper sites in the +1 oxidation state (Fig. 5h). Such a slab was simulated because the copper-terminated Cu₂O(111) surface has been previously shown to be more stable than the oxygen-terminated one at potentials under CO₂RR conditions⁶⁰. In all three electrodes, the presence of coadsorbed *I and/or *Cl moieties was considered (Fig. 5c,e,f,i,j; further details of the computational methodology can be found in Methods and in Supplementary Fig. 11 and Supplementary Tables 16–18). To observe clear trends made of sufficient data points, we also included calculations for stripes on Ni(100) and Ag(100) with and without *O and *I.

a, Schematic representation of the reaction pathway of acetylene reduction to 1,3-BD with water as the proton source. Carbon and hydrogen atoms appear in brown and grey, copper is shown in beige. b–j, The active sites considered in the computational models are a square-symmetry stripe on top of a Cu(100) terrace (b), stripe + *I (c), stripe + *O (d), stripe + *O + *I (e), stripe + *O + *Cl (f), stripe + *O + *HSO₄ (g), Cu₂O(111) (h), Cu₂O(111) + *I (i) and Cu₂O(111) + *Cl (j).

Full size image

Using Cu₂O(111), we first modelled how the removal of surface *O (*O + 2e⁻ + H₂O → * + 2OH⁻) is influenced by the presence of iodide and chloride. The potential to dissolve an oxygen atom from Cu₂O(111) was lowered by ~100 mV when iodide is near the *O species, as compared to chloride (Supplementary Table 19). This indicates that iodide prevents the reduction of Cu₂O to copper, in agreement with our experimental observations (Figs. 1h and 2).

We then determined the impact of *I and *Cl on 1,3-BD formation. The calculated activity of all active sites in Fig. 5 and their analogues on nickel and silver can be observed in Fig. 6 as a Sabatier-type volcano plot, where the activity for e-C₂H₂R is assessed in terms of free energy (see 'Computational methods') and plotted as a function of the free energy of adsorption of acetylene (\(\Delta {G}_{\mathrm{{C}_{2}{H}_{2}}}\)) at pH 7 and −0.85 V versus SHE, which is the least negative potential reported in our experiments. We choose \(\Delta {G}_{\mathrm{{C}_{2}{H}_{2}}}\) as a descriptor because acetylene is the main reactant and, according to the Sabatier principle, C₂H₂ adsorption should be neither too weak nor too strong for optimal electrocatalytic activity⁶¹. For convenience, the free energies in Fig. 6 for all materials are listed in Supplementary Tables 20–22, the acetylene binding distances are in Supplementary Table 23, and the corresponding free-energy diagrams for all the copper-based active sites are depicted in Supplementary Figs. 12 and 13.

The computed reaction free energies are plotted against the adsorption free energies of acetylene (\(\Delta {G}_{\mathrm{{C}_{2}{H}_{2}}}\)) on various electrodes. The active sites included are stripe, stripe + *O + *I, stripe + *O + *Cl, stripe + *O + *HSO₄, stripe + *I and stripe + *O on copper, nickel and silver electrodes at −0.85 V versus SHE. The initial hydrogenation step is limiting on the right side of the volcano (green), and the coupling step is limiting on the left side (orange). The colour of the points indicates their limiting step, and the active sites are shown in Fig. 5. Inset: summary of the active sites of each metal. The data points represent the average activities and adsorption energies of each catalyst with different coadsorbed species, while the error bars represent the s.d. of the data in the main panel for each catalyst. Average data for Cu₂O(111) are also provided for comparison (see also Supplementary Tables 20–22). The vertical line marks the energy range where acetylene is preferentially adsorbed (left-hand side) or in the gas phase (right-hand side).

Full size image

Apart from delaying the reduction of the oxide, we found that the anions also modify the binding energies of acetylene differently. To evince this, consider the data points for the Cu(100) stripe (denoted as stripe) in Fig. 6, which are all located on the strong binding side of the volcano plot. With respect to the clean stripe, *O + *Cl strengthens \(\Delta {G}_{\mathrm{{C}_{2}{H}_{2}}}\) and decreases the e-C₂H₂R activity, while *O + *I weakens \(\Delta {G}_{\mathrm{{C}_{2}{H}_{2}}}\) and increases the activity. We also studied the effect of *O + *HSO₄ on the Cu(100) stripe (Fig. 6) and found that it unsuitably strengthens \(\Delta {G}_{\mathrm{{C}_{2}{H}_{2}}}\). Similarly, the combination of *O + *I is beneficial for nickel stripes, but we note that only copper stripe sites are close to the ideal \(\Delta {G}_{\mathrm{{C}_{2}{H}_{2}}}\approx -0.34\,{\mathrm{eV}}\) at −0.85 V versus SHE because nickel binds acetylene and the reaction intermediates too strongly. Conversely, the mild adsorption energies of the copper stripes guarantee, in principle, that the surface will not be poisoned by strongly adsorbed species. In turn, silver sites bind acetylene and the other species too weakly. In fact, silver sites lie to the right of the vertical line in Fig. 6, which indicates that acetylene adsorption might be insufficient on those sites. A detailed description showing the changes in the volcano plot shape and the activities of the surface sites with three different potentials (0, −0.40 and −0.85 V versus SHE) are shown in Supplementary Figs. 14 and 15.

We further considered the nature of the active sites for 1,3-BD formation by means of the atomic partial charges on the stripe + *O + *I using Bader analysis (Supplementary Fig. 16)⁶². We found that the ensemble of copper stripe atoms in contact with *O and *I have in total one electron fewer than in a fully metallic surface. In addition, Supplementary Fig. 16 shows that the intermediates *C₂H₂ and *C₂H₃ adsorb on opposing bridge sites of the stripe edges where one of the copper atoms is partially positively charged and the other is neutral, suggesting that the active sites are indeed Cu^δ+–Cu⁰.

Since the limiting step for 1,3-BD production on copper stripes at −0.85 V versus SHE is the coupling of two *C₂H₃ moieties, which is a non-electrochemical step, we calculated the transition state for such a coupling on the active site closest to the volcano summit, namely, the stripe + *O + *I using the nudged elastic band method⁶³. To form the C–C bond, a diffusion barrier of 0.56 eV is needed for one *C₂H₃ moiety to approach the other. Subsequently, C–C coupling takes place with a barrier of 0.47 eV (Supplementary Figs. 17 and 18). As both barriers are surmountable at room temperature⁶⁴, we conclude that undercoordinated stripe sites with *O and *I are both thermodynamically and kinetically poised to reduce acetylene to 1,3-BD.

The inset in Fig. 6 provides a summary of the trends and shows that various active sites on the partially oxidized copper stripes are close to the top of the volcano, whereas Cu₂O(111) is not as active, which suggests that a certain degree of surface reduction is necessary for copper sites to be active for 1,3-BD production. To close the computational analysis, we would like to note that our present model may be supplemented in future studies by means of constant-potential simulations^65,66.

Overall, this work has illustrated the feasibility of using e-C₂H₂R to selectively produce 1,3-BD under mild conditions and with water as the proton source. We have used acetylene, a feedstock that can be derived from two major greenhouse gases in the atmosphere, namely, CO₂ and methane. Our collective experimental and computational findings have shown that the formation of Cu^δ+–Cu⁰ sites in the presence of I⁻ results in a more efficient production of 1,3-BD from e-C₂H₂R. The 1,3-BD that is formed can be extracted from the product gas stream using pressure swing adsorption or extractive distillation^3,67. We note that the state-of-the-art thermocatalytic conversion of C₂H₂ to 1,3-BD has a selectivity of only ~20% (ref. ⁶⁸). In contrast, we have shown the synthesis of 1,3-BD from e-C₂H₂R with an FE of 93%. Thus, e-C₂H₂R is probably a more viable option for the synthesis of 1,3-BD. To further determine if e-C₂H₂R can be practically implemented, we estimate the energy consumption needed to synthesize 1,3-BD from e-C₂H₂R when starting from a methane feedstock. By first converting methane to acetylene by electric arc plasma treatment¹², and then utilizing the acetylene for e-C₂H₂R to 1,3-BD, we estimate that the overall energy required to synthesize 1,3-BD is ∼30–85% lesser than from the conventional methods of naphtha, ethane or crude oil cracking (Supplementary Note 2). However, we note that this is only a preliminary estimate that may change in the future as the technology improves. The energy efficiency of our approach can be further improved, either by microwave plasma treatment to synthesize the acetylene feedstock⁶⁹, or by utilizing the thermodynamically more favourable methanol oxidation as the anodic reaction in the electrolytic cell⁷⁰.

We showed here that a Cu₂O_NC-derived electrode with KI electrolyte is an efficient catalyst for the electroreduction of acetylene to 1,3-BD. Remarkably, by employing this set-up, we were able to enhance the formation of 1,3-BD, such that it is now formed as a primary product, with an FE_1,3-BD of 93% at −0.85 V versus SHE and a j_1,3-BD of −75 mA cm⁻² at −1.0 V versus SHE. Operando Raman and XAS spectroscopic analysis revealed that I⁻ helped preserve the oxidized state of Cu₂O, resulting in the formation of Cu^δ+–Cu⁰ species on the catalyst under negative applied potentials, which we correlate to the larger amount of 1,3-BD produced. Conversely, e-C₂H₂R in Cl⁻-, Br⁻-and SO₄²⁻-based electrolytes produced a lower amount of 1,3-BD due to the substantially higher amount of metallic Cu⁰ present. We showed using DFT that the ensemble of undercoordinated copper sites with coadsorbed *I and *O results in a favourable adsorption of acetylene and its subsequent hydrogenation to the key intermediate *C₂H₃. The simulations further show that these sites display surmountable C–C coupling barriers for *C₂H₃ moieties, and thus enable the selective production of 1,3-BD, which corroborates our experimental findings.

In perspective, this work highlights a potentially more sustainable pathway for 1,3-BD production compared to the conventional thermocatalysis route. The diverse chemistry arising from acetylene electrolysis creates viable alternatives in the synthesis of value-added molecules, which is crucial for the decarbonization of chemical industries. However, a j_1,3-BD of −75 mA cm⁻² is still insufficient to replace current industrial methods. Thus, future works can focus on enhancing the j_1,3-BD towards commercially relevant production rates. Further studies can also be undertaken to improve the selectivity for other C₄ compounds such as 1-butene by developing tandem catalysts for e-C₂H₂R in an iodide electrolyte, or by co-feeding other feedstocks such as CO₂ to synthesize other multicarbon compounds.

Catalyst preparation

The Cu₂O_NC were prepared using a previously reported method³⁴. In brief, 4 ml of 0.1 M CuSO₄·5H₂O solution was first added into 360 ml of ultrapure water (18.2 MΩ) under vigorous stirring. Then, 10 ml of 1 M NaOH was added, followed by the addition of 16 ml of 0.2 M ascorbic acid. The solution was stirred for another 30 min. The Cu₂O_NC were extracted using centrifugation, washed with ultrapure water and ethanol, and then dried at 70 °C in an oven.

For the preparation of the working electrodes, 5 mg of Cu₂O_NC was mixed with 500 μl of isopropanol and 25 μl of Nafion solution, and the mixture was sonicated. The catalyst ink was then dropcast onto a gas diffusion layer (Sigracet 28BC + 20 wt% FEPD 121 wet proofing) and dried on a hot plate. The catalyst mass loading was 5 mg cm⁻².

Electrochemical reduction of C₂H₂

A three-electrode flow cell set-up was used (Supplementary Fig. 5). Ag/AgCl (saturated KCl, Pine) and nickel foam (Xiamen Lith Machine) were used as the reference electrode and the counter electrode, respectively. Potassium iodide (Meryer, 99.99%), potassium bromide (Meryer, 99.99%), potassium chloride (Meryer, 99.99%), potassium sulfate (Meryer, 99.99%), potassium fluoride (Merck, 99.0%), sodium iodide (Sigma Aldrich, 99.5%) and caesium iodide (Sigma Aldrich, 99.99%) were used without further purification for making catholytes. The cation concentration for all electrolyte mixtures was 1 M. 1 M KOH (Meryer, 99.99%) was used as the anolyte because the oxygen evolution activity is favourable under basic conditions. The cathodic compartment (10 ml of catholyte) and anodic compartment (10 ml of anolyte) were separated by an anion-exchange membrane (Selemion AMV, AGC Asahi Glass). A mixture of C₂H₂ (Chem-gas, 99.8%) and N₂ gas (Chem-gas, 99.99%) was used as the gas feedstock. The total gas flow rate is 50 sccm.

A potentiostat (Gamry Reference 600) was used to control and measure the electrochemical activity. The current interrupt mode was used to compensate for the iR drop during chronoamperometry. Each constant potential electrolysis was carried out for 50 min. The potentials in this work are referenced to the SHE using the average potential of the Ag/AgCl electrode (E_Ag/AgCl) given by equation (1):

The gaseous products (C₂H₄, 1,3-BD, 1-butene, trans-2-butene, cis-2-butene) were detected by an online gas chromatograph (GC, Shimadzu GC2014) with a flame ionization detector. H₂ is detected with a thermal conductivity detector. The measurements were made over the course of three GC injections (with intervals of 1,140 s between consecutive injections). Details of our chromatography set-up have been previously described⁷¹. A second gas chromatograph (Wasson-ECE GCMS) with a flame ionization detector was used to further confirm the gaseous products formed during e-C₂H₂R.

Characterization of electrodes and electrolytes

For our SEM, ECSA, XRD and Raman spectroscopy characterization, the Cu₂O_NC catalyst was dropcast onto a graphite disc (Ted Pella). For our XANES characterization, Cu₂O_NC catalyst coated on a gas diffusion layer was used.

The surface morphologies of the catalysts were characterized by SEM (JEOL JSM-6701F, 5 keV). The pH of the electrolytes was measured using a HI-2002 Edge pH meter (Hanna Instruments).

The ECSAs of the electrodes were estimated using their double-layer capacitances (C_dl). The catalysts were first washed using ultrapure water after e-C₂H₂R. Their C_dl were then measured in 0.5 M K₂SO₄ electrolyte under N₂ saturation. A three-electrode H-cell (Ag/AgCl reference electrode and a graphite counterelectrode) was used. The scan rates applied were 10, 20, 30, 40 and 50 mV s⁻¹.

XRD analysis was used to determine the composition of the catalyst. The system used was a Bruker D8 Advance (Cu Kα 40 kV, 40 mA) with a 2D Lynxeye XE PSD counter detector, with the incoming signal fixed at 5°.

Raman spectroscopy was performed using a confocal Raman microscope (Horiba Jobin Yvon) in epi-illumination mode (top-down). A helium–neon laser with 633 nm wavelength (Pacific Lasertec, 05-LHP-845, 17 mW) was used as the excitation source. A water-immersion objective lens (LOMO, APO 70×; numerical aperture, 1.23) covered with a 0.013 mm thin Teflon film (American Durafilm) was used for focusing and collecting the incident and scattered laser light⁷². The backscattered light was filtered through a 633 nm edge filter, before being directed into a spectrometer (LabRAM HR evolution)/charge-coupled device detector (Syncerity OE). The acquisition time for each spectrum was 30 s.

XAS measurements at the Cu K-edge were performed at the XAFCA facility of the Singapore Synchrotron Light Source. The XAS spectra of the Cu₂O_NC catalysts were acquired under e-C₂H₂R reaction conditions using a home-built electrochemical cell with a platinum-wire counterelectrode and an Ag/AgCl reference (Supplementary Fig. 4). The XAS and EXAFS data analysis was performed using ATHENA and ARTEMIS, respectively, which are part of the Demeter software package⁷³.

Computational methods

Three different copper-based electrode models were considered. We selected Cu(100) with a square-symmetry stripe on top of it to represent rough Cu⁰ sites. To represent oxidized copper sites, we used a copper-terminated Cu₂O(111) facet. In addition, to simulate a partially oxidized and defective electrode, we utilized the stripe with adsorbed O, denoted as stripe + *O. We also considered the coadsorption of either *I or *Cl on all surface models. In total, seven different electrode models have been considered (Supplementary Fig. 11). We also considered *O + *HSO₄ on the copper stripes.

Periodic slabs were constructed that had four atomic layers for the stripe and four copper atomic layers for Cu₂O(111). During structural optimization, the two bottommost layers remained fixed to provide a bulk environment for the surface layers. All slab models were built from the converged Perdew–Burke-Ernzerhof (PBE) lattice constants, which were found to be a = 3.64 Å for metallic copper and a = 4.31 Å for Cu₂O. The PBE lattice parameter of Cu₂O was taken from the literature⁷⁴. For the Cu(100) stripe, 5 × 4 supercells were used; in turn, a 2 × 2 supercell was used for Cu₂O(111). Supplementary Fig. 11 provides visual representations of the slab models.

All energy values and corresponding minimum energy structures were obtained from DFT calculations carried out with the VASP code⁷⁵, using the PBE exchange-correlation functional⁷⁶, which provides accurate representations of transition metals in the bulk^77,78 and their low Miller index surfaces⁷⁹. In all model systems, the Cu(3d) shell is filled, such that no Hubbard-like (U–J) terms were necessary. The valence electron density was expanded in a plane-wave basis set with a kinetic energy cut-off of 450 eV. The projector augmented wave method⁸⁰, as implemented in VASP⁸¹, was used to incorporate the effect of core electrons on the valence electron density. To facilitate convergence of the self-consistent field procedure, electronic temperature smearing at the Fermi level was used following the Methfessel–Paxton approach⁸² with a value of 0.2 eV for metal slabs, while a Gaussian smearing of 0.1 eV was used for the oxide. Upon convergence, all total energies were extrapolated to 0 K. The numerical integration in the reciprocal space was carried out using Monkhorst–Pack⁸³ grids of 2 × 3 × 1 special k-points for the Cu(100) stripes and 2 × 2 × 1 for Cu₂O(111). These k-point grids ensured convergence of the adsorption energies within ±0.05 eV. To prevent spurious electrostatic interactions, the periodically repeated images in the vertical direction were separated by more than 15 Å of vacuum and dipole corrections were added. Geometry optimizations utilized the conjugate-gradient optimization algorithm until the maximum residual force on all atoms was below 0.05 eV Å⁻¹. The calculations on nickel stripes were spin unrestricted and made with a lattice constant of 3.52 Å. In turn, the calculations on silver stripes were spin restricted and made with a lattice constant of 4.16 Å. In both cases, the geometry optimizations started from analogous configurations as those found for copper stripes. Furthermore, boxes of 9 × 10 × 11 ų were used to calculate the energy of H₂, H₂O, acetylene (C₂H₂) and 1,3-BD (C₄H₆), considering the Γ-point only, using Gaussian smearing and an electronic temperature of 0.001 eV, further extrapolated to 0 K.

The reaction free energies were approximated as ∆G ≈ ∆E_DFT + ∆ZPE − T∆S, where ∆E_DFT is the reaction energy calculated with DFT, ∆ZPE is the zero-point energy change between reactants and products, and T∆S is the entropy correction at 298.15 K. The ZPEs of all species and vibrational S of the adsorbed species were computed from the vibrational frequencies obtained with DFT at T = 298.15 K, within the harmonic oscillator approximation. For free molecules, the values of S were calculated from data in thermodynamic tables⁸⁴ at 298.15 K (Supplementary Table 16). To remove the intrinsic error of the PBE functional in the energy of gas-phase molecules, the computed total energy values of C₂H₂ and C₄H₆ were corrected by 0.12 and 0.67 eV, respectively. These corrections were obtained using the semiempirical method described by Granda-Marulanda et al.⁸⁵, based on the difference between DFT-calculated and experimental formation energies. A liquid-phase correction of −0.09 eV was used for water⁸⁶.

The computational hydrogen electrode was used to assess the free energy of proton–electron pairs⁸⁶. The additive inverse of the activity used to make the volcano plots (−∆G) is defined as the most positive reaction free energy among the electrochemical steps at a given applied potential: ∆G = max(∆G₁,∆G₂,∆G₃). The free energies of the electrochemical steps at pH 7 and various applied potentials were computed as: ∆G_i(U) = ∆G_i(0) + eU − 0.059pH.

We did not account for water solvation in our calculations because acetylene, butadiene and the reaction intermediates do not have polar functional groups able to make strong hydrogen bonds with water. To illustrate this, we computed the solvation stabilization for *C₂H₂ and *C₂H₂ + *C₂H₃ adsorbed on the copper stripe in two ways: using the VASPsol implicit solvation method implemented in VASP^87,88 and using an iterative microsolvation method^89,90. Supplementary Table 17 shows that water does not appreciably contribute to the stabilization of the adsorbed intermediates, as the solvation energies do not exceed 0.05 eV per adsorbate in any case. In contrast, we observe for *C₂H₂ and *C₂H₂ + *C₂H₃ on the stripe + *O a clear stabilization in the presence of *I and *Cl, as shown in Supplementary Table 18.