Spectroelectrochemical insight into copper cobalt catalysts for CO₂ and nitrite co-electroreduction to urea

Abstract

Electrochemical CO₂ and nitrite co-reduction provides a sustainable urea synthesis route but remains limited by low selectivity and an undecided C–N coupling mechanism. Here, we report co-sputtered bimetallic Cu–Co catalysts that facilitate urea formation via a tandem relay mechanism. The optimal Cu:Co ratio of 1:1 achieves a urea yield rate of 61 ± 6 mmol h⁻¹g_cat⁻¹ at –1.2 V vs. RHE under neutral pH, emphasizing the importance of proton balance in sustaining proton-coupled electron transfer. In situ synchrotron-based infrared and Raman spectroscopy monitor the dynamic evolution of *CO, *NH₂, and C‒N intermediates. In situ X-ray absorption spectroscopy indicates the structural stability of metallic Cu and Co active sites. Density functional theory calculations suggest that *COOH + *NH₂ coupling initiates urea pathway, with *NH₂CO formation as the potential-determining step. This study integrates rational catalyst design and in situ spectroelectrochemical analysis to advance understanding of electrochemical C–N coupling for urea synthesis.

Introduction

Urea (CO(NH₂)₂) is the most utilized fertilizer globally with market size around $129 billion in 2023, which it is projected to grow to around $161 billion by 2032¹. Currently, the primary method for commercial urea production is the Bosch-Meiser process, which involves the reaction of ammonia (NH₃) and CO₂ under elevated temperature (150–200 °C) and pressure (100–200 bar)². NH₃ is typically derived from the Haber-Bosch process, which relies on hydrogen sourced from fossil fuels, making upstream hydrogen production the main contributor to energy consumption and carbon emissions². As a result, the overall urea production remains carbon-intensive, generating ~910 kg of CO₂ for every ton of urea produced^3,4.

Electrochemical co-reduction of CO₂ and NO_x⁻, such as NO₂⁻ or NO₃⁻, offers a promising, more sustainable alternative. These NO_x⁻ species can be sourced from exhaust flue gas, chemical production by-product, wastewater streams, or synthesized via cold plasma oxidation of N₂^5,6,7,8,9,10. Such approaches could enable decentralized and potentially carbon-neutral urea production. However, urea electrosynthesis from CO₂ and NO_x⁻ is a complex multistep process involving multiple electron transfer (Tables S1, S2, Supplementary Data 1) and competing with mechanistically simpler reduction processes, viz. the hydrogen evolution reaction (HER), CO₂ reduction reactions (CO₂RR), and NO₂⁻/NO₃⁻ reduction reactions (NO₂⁻RR/NO₃⁻RR or NO_x⁻RR)^2,11. Notable challenges include limited CO₂ adsorption capacity¹², sluggish dissociation of the C=O bond^12,13, and difficulty in balancing adsorption and activation of both carbon- and nitrogen-derived species^2,14.

At the core of these challenges lies the C–N coupling step, whose mechanism remains poorly resolved due to interplay of multiple surface-bound species and dynamic changes in catalyst structure under reaction conditions. Proposed C–N coupling pathways involve intermediates such as *CO, *NO, *NH₂, and *NH, with key steps often limited by proton-coupled electron transfer (PCET) barriers^11,15,16,17. Elucidating these pathways—and understanding how the catalyst surface evolves during operation—requires in situ spectroelectrochemical characterization techniques, such as Raman, infrared, and X-ray absorption spectroscopy (XAS), which can directly capture transient intermediates and structural changes^12,13. Achieving the most energetically favorable pathway requires precise tuning of intermediate adsorption energies, a task that may be inherently challenging for a monometallic catalyst. This has led to growing interest in multimetallic designs, where distinct active sites can preferentially stabilize different intermediates and facilitate sequential steps. Indeed, such strategy has been successfully applied in the oxygen evolution reaction (OER)^18,19 and CO₂RR^20,21 fields, offering a blueprint for tackling the complexity of urea electrosynthesis.

Copper has been widely studied due to its ability to facilitate CO₂ and NO_x⁻ activation, moderate hydrogen binding, and tendency to form key intermediates such as *CO and *NH₂^{22,23,24,25,26,27}. However, Cu suffers from limited urea selectivity and deactivation caused by strong adsorption of nitrogen intermediates^{28,29,30,31,32}. Cobalt, on the other hand, offers high conductivity, favorable *CO binding, and the ability to stabilize *H intermediates, thereby promoting downstream protonation or hydrogenation steps^{31,33,34,35,36}.

These complementary characteristics led us to a working hypothesis that combining Cu and Co could couple their strengths—enhancing NO_x⁻ reduction to reactive nitrogen species on Cu, while stabilizing *CO and *H on Co—to accelerate C–N bond formation^31,37,38. Supporting this concept, recent work on amorphous Cu–Co boride (a-Cu_0.1CoB_x) demonstrated alleviation of strong N-based and *CO intermediates binding on the catalyst surface and achieved a faradaic efficiency (FE) of ca 28% with a urea yield rate of ca 310 μg h⁻¹ mg_cat⁻¹ (5.2 mmol h⁻¹ g_cat⁻¹) from CO₂ and NO₃⁻³⁹.

In parallel, we also considered an alternative, cross-assigned tandem hypothesis—that Cu predominantly improves the CO₂ adsorption and activation, stabilizing *CO/*COOH^40,41, whereas Co primarily promotes NO₂⁻ activation and stabilizes N-intermediates^42,43. To address the ambiguity and resolve the roles of the two metals, we explored tandem Cu‒Co catalysts prepared via co-sputtering. This approach enables atomic-level mixing of two metals, providing different but complementary active sites for key reaction steps. We anticipated that such atomic-scale integration would enhance PCET, promote the formation of key intermediates, thereby optimizing the C‒N coupling process^44,45,46.

To test these hypotheses, we first performed systematic electrochemical studies under varying electrolyte conditions—including pH and nitrogen-source species—to evaluate urea formation and identify optimal conditions. Guided by these results, we employed in situ spectroelectrochemical techniques—in situ Raman, synchrotron radiation Fourier-transform infrared spectroscopy (SR-FTIR), and XAS—to monitor the formation and evolution of reaction intermediates directly on the catalyst surface under operating conditions. Finally, density functional theory (DFT) calculations were performed under the optimized reaction conditions to provide atomic-scale insights into the roles of Cu and Co.

Together, these results support the alternative cross-assigned mechanism, rationalizing the observed selectivity where Cu predominantly stabilizes CO₂-derived intermediates (*CO/*COOH), and Co drives *NO₂⁻ reduction and stabilizing *NH₂ for C‒N coupling. This integrated experimental–theoretical approach aims to unravel the C–N coupling mechanism and guide the rational design of efficient catalysts for sustainable urea electrosynthesis.

Results

Catalyst characterization

We varied Cu and Co stoichiometry within the synthesized electrodes via co-sputtering metals to form Cu₁Co₂, Cu₁Co₁, and Cu₂Co₁ onto carbon fiber paper (CFP) substrates (collectively referred to as Cu_(n)Co_(1-n)) and characterized as-prepared materials using a range of physiochemical characterization methods. Inductively coupled plasma mass spectrometry (ICP-MS) analysis confirmed that metals were deposited close to expected ratios (Table S3, Supplementary Data 2).

Scanning electron microscopy (SEM) coupled to energy-dispersive X-ray spectroscopic (EDS) analysis confirms uniform distribution of the Cu and Co over the surface of carbon fibers (Fig. 1a–c and Figs. S1, S2, S4). EDS analysis under low-magnification transmission electron microscopy (TEM) conditions indicates homogenous mixing of Cu and Co within the sputtered materials with no observable segregation within detection limit of individual metallic phases (Fig. 1d, e and Figs. S3–S5). Higher resolution TEM of the as-prepared catalyst indicates the presence of well-defined crystal facets of a face-centered cubic (fcc) structure with an interlayer spacing of ~1.9 and 2.1 Å. This is consistent with the fcc (200) and fcc (111) facets, respectively, of both copper and cobalt metal phases (Fig. 1f and Figs. S6)⁴⁷. The selected area electron diffraction (SAED) patterns further validate the presence of these phases (Fig. 1g and Fig. S6). Moreover, these crystal facets remain observable even after electrochemical testing, as evidenced by the post-reaction TEM analysis (Fig. S7). X-ray diffraction (XRD) patterns of all samples exhibit a prominent peak at 43.6°, which is consistent with the (111) plane of the fcc Cu and Co metals (ICDD 00-004-0836 and 01-071-4238) (Fig. S8). A peak at 39.6° is most likely associated with the 200 signals of the monoclinic CuO phase (ICDD 01-070-6827).

Fig. 1: Surface morphology of the Cu₁Co₁ catalyst.

a–c SEM images of an as-deposited catalyst at different magnifications; image in (c) is manually colorized to support the discrimination between the catalyst (red) and carbon fiber paper (gray). Scanning transmission electron microscopic (STEM) image of an as-deposited catalyst (d), and corresponding EDS mapping of Cu (red), Co (green), and O (blue) (e). f High resolution TEM images of an as-deposited catalyst at different magnifications. Yellow lines highlight the Cu(111) and Co(111) facets with ~2.1 Å spacing. The inset shows the corresponding fast Fourier transform (FFT) pattern (scale bar: 10 nm⁻¹), confirming the crystalline nature of the observed lattice fringes. g Selected area electron diffraction (SAED) pattern of an as-deposited catalyst showing diffraction ring corresponding to face-centered cubic (fcc) (111) and (200) facets.

X-ray photoelectron spectroscopy (XPS) was employed to elucidate the surface chemical composition of the as-prepared Cu_(n)Co_(1-n) catalysts (Fig. 2a, b and Figs. S9–S14), with monometallic sputtered Cu and Co serving as references. The survey spectrum reveals distinct peaks corresponding to C 1s, O 1s, Cu 2p and Co 2p, with relative intensities of Cu 2p and Co 2p peaks following the bulk composition (Fig. S9). Fitting of the high-resolution Cu 2p and Co 2p spectra of as-synthesized catalysts using the parameters reported by Biesinger et al.^48,49 and our previous reports^50,51, as a guide displays the presence of Cu(0), Cu(I), Cu(II), Co(0), and Co(II) species, indicating oxidation of surface Cu and Co metals to Cu(I), Cu(II) and Co(II) due to air exposure (Fig. 2a, b, Figs. S11–S12, Table S4, Supplementary Data 3)^52,53. High-resolution Cu LMM Auger spectra further confirms the presence of Cu(0), Cu(I), and Cu(II) species (Fig. S13). Analysis of the high-resolution O 1s spectra of Cu_(n)Co_(1-n) catalysts identifies lattice (529.3 eV) and non-lattice oxygen (~531.5 and 532.5 eV) (Fig. S14)⁵⁴. Post-reaction analysis of Cu_(n)Co_(1-n) catalyst reveals changes in surface structure after 12 h electrochemical testing (vide infra) (Fig. 2a, b and Figs. S9). In addition to Cu 2p, Co 2p, C 1s, and O 1s peaks, survey spectra detect K 2p and K 2s peaks, indicating adsorption of K⁺ from the electrolyte on the catalyst surface. Analysis of Co 2p, Cu 2p, and Cu LMM Auger spectra indicate further oxidation of Cu(0) and Co(0) to Cu(I), Cu(II) and Co(II) induced by the electrocatalytic testing (Figs. S11–S13)⁵⁵. Consequently, the oxidation leads to the increased peak intensity of lattice oxygen, as seen in high-resolution O 1s spectra (Fig. S14). XAS measurements of Cu_(n)Co_(1-n) catalysts were recorded at Cu and Co K-edges. The X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) spectra (Fig. 2c–h and Figs. S15–S18) confirm that the as-prepared catalysts are dominated by metallic Cu(0) and Co(0), consistent with the XRD and HRTEM results. Fourier transformation (FT-EXAFS) and k³-weighted EXAFS spectra further support this assignment (Fig. S16). Linear combination fitting (LCF) of both XANES (Figs. S19) and EXAFS (Figs. S20) indicate that the materials being composites primarily consisting of metallic Cu and Co, without evidence of intermetallic phases. In addition to the dominant metallic phases, minor Cu₂O/Cu(OH)₂ and Co₃O₄/Co(OH)₂ species are also present as detectable oxide/hydroxide phase. After electroreduction, XANES spectra (Figs. S15, S17–S18) show slight oxidation, evidenced by an increase in edge intensity, consistent with XPS results. Nevertheless, FT-EXAFS at both Cu and Co K-edges confirms that Cu–Cu and Co–Co metallic interactions remain predominant (Fig. S17–S19), with LCF analysis indicating that the post-reaction catalysts are still primarily metallic, with fractions of oxidized Cu and Co species present (Fig. S19, S20).

Fig. 2: Physical characterization of the Cu₁Co₁ catalyst.

High-resolution Cu 2p (a) and Co 2p (b) of as-deposited catalyst. The XANES at Cu K-edge (c) and Co K-edge (d). The k³-weighted EXAFS spectra at Cu K-edge (e) and Co K-edge (f). The normalized FT-EXAFS of Cu₁Co₁ at Cu K-edge (g) and Co K-edge (h) of Cu₁Co₁ catalyst. In (a, b) circles denote the measured counts per second (CPS), while black dashed line represents the fitted background.

Raman spectroscopy was also conducted to characterize the components of the Cu_(n)Co_(1-n) catalysts (Fig. S21). At 50% laser power, the as-prepared Cu₁Co₁ catalyst exhibits peaks indicative of Co(OH)₂ at ~530 and 600 cm⁻¹, Cu₂O at ~147 and 218 cm⁻¹, and CuO at ~294 cm⁻^{1 56,57,58,59,60}. These oxide/hydroxide features are attributed to the oxidation of metallic Cu and Co resulting from air exposure^61,62. Under power-dependent Raman acquisition, additional CuO and Co₃O₄ modes emerge only at higher laser power (>50%) that may indicate a laser-induced oxidation component^63,64. Meanwhile, LCF of EXAFS spectra (Fig. S20) place an upper bound of ≤8% Co₃O₄ in Cu_(n)Co_(1-n) catalysts, implying that any pre-existing Co₃O₄ is minor near-surface fraction below bulk XAS detectability. Taken together, power-dependent Raman and LCF EXAFS spectra support that the Co₃O₄ bands observed by Raman arise primarily from laser-induced oxidation, possibly superimposed on a very thin surface Co₃O₄ layer.

Comprehensive characterization confirmed that as-prepared catalysts predominantly comprise metallic Cu(0) and Co(0) states. Upon electrochemical testing, partial surface oxidations to Cu(I), Cu(II) and Co(II) were observed, with the bulk metal phases largely retained, demonstrating structural stability under reaction conditions.

Electrocatalytic performance

CO₂ and KNO₂ were selected as the carbon and nitrogen sources for our experiments to enhance C‒N coupling based on our previous finding and the above discussions⁶⁵. The Cu₁Co₁ catalyst was initially evaluated for its activity towards the CO₂RR, NO₂⁻RR, and electroreduction of CO₂ and NO₂⁻ in the CO₂-saturated 0.2 M KHCO₃ solution (pH 7), Ar-saturated 0.2 M KHCO₃ + 0.02 M KNO₂ solution (pH 8), and CO₂-saturated 0.2 M KHCO₃ + 0.02 M KNO₂ solution (pH 7), respectively. To maintain saturation in these catholytes, CO₂ or Ar gas was continuously injected.

Cyclic voltammetry (CV) measurements (Fig. 3a and Figs. S22a–S25) reveal the current response of Cu₁Co₁ catalyst under each reaction condition. Notably, distinct oxidation and reduction peaks are observed between −0.1 and 0.85 V vs. RHE in all cases (Fig. S23), indicating the formation of Cu and Co oxides/hydroxides and their subsequent reduction. Notably, Cu₁Co₁ exhibits quasi-reversible redox behavior, in contrast to the reversible transition of pure Cu and the irreversible features of pure Co (Fig. S24)^66,67,68,69. The irreversible nature of pure Co is likely linked to Co²⁺/Co³⁺ dissolution into the electrolyte^70,71. Together, these findings suggest that Cu–Co bimetallic occupy an intermediate regime, likely due to electronic and structural stabilization at the interface⁷².

Fig. 3: Electrochemical performance of Cu₁Co₁.

a The cyclic voltammetry of the Cu₁Co₁ catalyst for the CO₂RR, NO₂⁻RR, and electroreduction of CO₂ and NO₂⁻ (scan rate 20 mV s⁻¹, 5th scan shown). Currents are normalized to the geometric surface area of the electrode. b The FE_NH4⁺ and yield rate from the NO₂⁻RR. c The FE of the CO₂RR products. d The FE_CO(NH2)2 and yield rate from the electroreduction of CO₂ and NO₂⁻. e Representative ¹H NMR spectrum showing CO(NH₂)₂ produced at −1.2 V vs. RHE. f Urea detection by ¹H-NMR at t = 0, 5, and 12 h of the experiment at −1.2 V vs. RHE. The CO₂RR, NO₂⁻RR, and electroreduction of CO₂ and NO₂⁻ were carried out in CO₂-saturated 0.2 M KHCO₃, Ar-saturated 0.2 M KHCO₃ and 0.02 M KNO₂, and CO₂-saturated 0.2 M KHCO₃ and 0.02 M KNO₂ solutions, respectively. The error bars represent the standard deviation for at least three independent measurements.

Furthermore, the CV profiles show that CO₂RR exhibits lower current density compared to NO₂⁻RR and electroreduction of CO₂ and NO₂⁻, suggesting relatively lower activity towards CO₂ reduction (Fig. 3a). Chronoamperometry (i-t) measurements for the CO₂RR, NO₂⁻RR, and electroreduction of CO₂ and NO₂⁻ indicate that the Cu₁Co₁ catalyst achieves j of approximately −44, −80, and −64 mA cm⁻² (−0.22, −0.41, −0.33 mA cm_ECSA⁻²) at −1.2 V vs. RHE, respectively (Fig. S28-S31). NO₂⁻RR testing of the Cu₁Co₁ catalyst over the potential range −0.6 to −1.2 V vs. RHE showed that the highest ammonium (NH₄⁺) selectivity was achieved at −0.8 V vs. RHE with FE and yield rate of 96 ± 4% and 340 ± 9 mmol h⁻¹ g_cat⁻¹ (0.24 ± 0.006 nmol s⁻¹ cm_ECSA⁻²) (Fig. 3b). These values are comparable to other Cu- and Co-based catalysts in non-alkaline conditions^73,74. CO₂RR was also conducted in the same potentials range as NO₂⁻RR (Fig. 3c). At −0.6 V vs. RHE, the reaction was highly selective to formate (HCOO⁻) with a small contribution of acetate (CH₃COO⁻) and methane (CH₄). As the potential becomes more negative, the FE for formate progressively decreases while the amounts of other products, including methanol (CH₃OH⁻), grows. At −1.0 V vs. RHE, the FE for CH₄, CH₃OH, and CH₃COO⁻ reaches 21 ± 3%, 16 ± 4%, and 12 ± 2%, respectively. The use of even more negative potential did not improve product selectivity due to the enhanced HER.

Urea electrosynthesis was conducted via co-electroreduction of CO₂ and NO₂⁻. During 12 h experiments, urea was consistently detected at −1.0 V vs. RHE and more negative potentials using both ¹H NMR spectroscopy and gas chromatography with mass-spectrometry detector (GC-MS) (Fig. 3d,e and Fig. S32). Furthermore, ¹⁵N isotope-tracing experiments confirm that the produced urea originates from the NO_x⁻ feed (Fig. S32 and Fig. S42). The highest performance was observed at −1.2 V vs. RHE, with a maximum FE of 11 ± 2% and a yield rate of 61 ± 6 mmol h^‒1 g_cat^‒1 (8.5 ± 0.8 nmol s⁻¹ cm⁻² per geometric surface area of the electrode or 0.043 ± 0.005 nmol s⁻¹ cm_ECSA⁻²).

Although CO₂RR exhibits high H₂ formation at this potential (Fig. 3c), introducing NO₂⁻ suppresses HER through two effects: (i) NO₂⁻-driven electron–proton sinking and site blocking—NO₂⁻ might occupy H-binding sites and lower HER^75,76. This is consistent with the additional redox couple that appears in CV curves for Cu₁Co₁ upon NO₂⁻ addition (Figs. S23–S25), indicative of NO₂⁻-metal interaction on the catalyst surface⁷⁷. (ii) Tandem perimeter kinetics of Cu‒Co that co-generate *CO/*COOH and *NH₂ in close proximity (Fig. S33), as confirmed through in situ SR-FTIR (Fig. S40), thereby enabling high urea rates despite the H₂-favoring bias⁷⁵.

Time-dependent studies revealed that urea was first detectable after 5 h, with an FE and yield rate of 10 ± 1% and 56 ± 3 mmol h^‒1 g_cat^‒1, respectively (Fig. 3f). The stability test further demonstrated that the Cu₁Co₁ catalyst can steadily produce urea over 48 h (Fig. S30, S31). In addition to urea, C₂ and other C‒N products, such as ethanol (C₂H₅OH), acetate, and methylamine (CH₃NH₂) were detected during the experiments at −1.0 V vs. RHE, although at lower FE and yield rates (Fig. S30). The maximum FEs for methylamine, ethanol, and acetate were observed at −1.3 V vs. RHE, measuring at 1.2 ± 0.1%, 3.0 ± 0.7%, and 1.0 ± 0.1 %, respectively (Fig. S29).

When NO₃⁻ was used as a nitrogen source instead of NO₂⁻, similar products were formed but at lower faradaic efficiencies and yield rates (Fig. 4a and Fig. S30). Urea was detected only at more negative potentials (≤−1.1 V vs. RHE) and achieved maximum values of 8 ± 1% and 26 ± 4 mmol h⁻¹ g_cat⁻¹ (0.018 ± 0.003 nmol s⁻¹ cm_ECSA⁻²) at −1.3 V vs. RHE (Fig. S31). This difference is consistent with our previous findings, where the activation and reduction of NO₂⁻ require lower overpotentials compared to NO₃⁻, enabling the preferential reduction of prior reactant⁶⁵. Supporting evidence was provided by in situ SR-FTIR spectra (Fig. S42), in which characteristic urea-related vibrations (*NCO, *CO carbonyl, and *CN) were observed more weakly in CO₂ + NO₃⁻ than in CO₂ + NO₂⁻ experiments^14,16,78.

Fig. 4: Electrochemical performance comparison of Cu_(n)Co_(1-n) catalysts and effect of electrolyte.

a Faradaic efficiency and yield rate of C‒N products from different nitrogen sources. The concentration of nitrogen sources was kept at 0.02 M in the 0.2 M KHCO₃ solution. The applied potential for the test was ˗1.2 V vs. RHE. b Cyclic voltammetry curves of Cu_(n)Co_(1-n) catalysts during electroreduction of CO₂ and NO₂⁻. c Faradaic efficiency of C‒N products using Cu_(n)Co_(1-n) catalysts after electroreduction of CO₂ and NO₂⁻ for 12 h. d The electroreduction of CO₂ and NO₂⁻ in various catholyte and anolyte at −1.2 V vs. RHE. For (a) electroreduction of CO₂ and NO₂⁻ were carried out in CO₂-saturated 0.2 M KHCO₃ with addition of 0.02 M KNO₃ or 0.02 M KNO₂ or 0.02 M NH₂OH. Catholyte for (b, c) is CO₂-saturated 0.2 M KHCO₃ + 0.02 M KNO₂. The concentrations of anolyte in (d) are 0.1 M KOH (pH 13), 0.2 M K₂SO₄ + H₂SO₄ (pH 5), and 0.2 M KHCO₃ (pH 8). For catholyte in (d), the buffer concentrations are identical to those of the anolyte but include 0.02 M KNO₂. All testing was performed using a continues CO₂ gas flow. The error bars represent the standard deviation for at least three independent measurements.

To further probe nitrogen source effects, hydroxylamine (NH₂OH) was tested. Instead of urea, the reaction yielded ammonium carbamate ([NH₄][NH₂COO]) with FE of 16 ± 3% and yield rate of 90 ± 5 mmol h^‒1 g_cat⁻¹ (0.064 ± 0.004 nmol s⁻¹ cm_ECSA⁻²) at −1.2 V vs. RHE (Fig. 4a). This suggests that the dominant process is electrochemical reduction of NH₂OH to NH₄⁺, while the carbamate anion forms chemically through reaction of hydroxylamine with CO₂ ^79,80. The absence of ammonium carbamate in NO₂⁻ and NO₃⁻ experiments implies that hydroxylamine is not an intermediate in the co-reduction pathway, or if transiently formed, is rapidly converted into other species. Overall, these results establish that NO₂⁻ is a more effective nitrogen source than NO₃⁻ or NH₂OH for urea electrosynthesis. Compared with reported catalysts (Table S5, Supplementary Data 4), the Cu₁Co₁ system achieves competitive urea yield rates under prolonged operation, demonstrating its effectiveness as a tandem Cu–Co catalyst for selective C–N coupling.

We also investigated electroreduction of CO₂ and NO₂^– at −1.2 V vs. RHE using monometallic Cu and Co catalysts (Fig. S30 and Fig. S33) as control references. The sputtered Cu catalyst produced no detectable C–N products, instead yielding formate, ethanol, and propanol (Fig. S29, S47, S48). This outcome contrasts with many reports describing strong C–N coupling on Cu and underscores a more nuanced picture^15,81. In the literature, high urea selectivity is typically associated with oxide-derived Cu, where mixed Cu(0)/Cu(I) and defect-rich surfaces stabilize polar intermediates such as *CO, *NCO and *NH₂CO^27,82. By comparison, sputtered Cu typically generates thin, relatively smooth films with fewer defect sites, thereby limiting the stabilization of key intermediates required for C‒N coupling⁸³. Consequently, metallic Cu favors CO₂ adsorption and *CO/*COOH formation⁴⁰, consistent with the results observed in the in situ SR-FTIR spectra (Fig. S42). For pure Co, the main products were ammonium and H₂ (Fig. S33), with no evidence of C–N products, also corroborated by in situ SR-FTIR (Fig. S42).

These controls underscore that neither Cu nor Co alone is sufficient for effective C–N coupling. Instead, the tandem interface of Cu–Co is essential: Cu sites efficiently adsorb and activate CO₂ to *CO, while Co sites accelerate the stepwise reduction of NO₂⁻ to *NH₂/*NH₃. The spatial co-generation and proximity of *CO and *NH₂ at the Cu–Co perimeter enhances the likelihood of C–N bond formation, leading to higher urea activity and selectivity than either metal alone.

Electrochemical CV measurements further highlight this synergy, as the Cu–Co bimetallic catalyst exhibits intermediate quasi-reversible behavior between pure Cu and Co (Fig. S24). This suggests that Cu suppresses Co dissolution by stabilizing its redox transitions, while Co mitigates Cu over-oxidation^78,84,85. Such mutual stabilization likely arises from electronic and structural interactions at the Cu–Co interface, further reinforcing the unique cooperative role of the tandem system. Consistently, electrochemical impedance spectroscopy (EIS) reveals that Cu₁Co₁ has a lower charge transfer resistance compared to pure Cu and Co (Fig. S27), underscoring its enhanced electronic synergy that facilitates CO₂ and NO₂⁻ and underpins the C–N coupling performance of the bimetallic catalyst. This effect was not reproduced by simple physical mixing: a drop-cast Cu/Co composite produced only ammonium, formate, H₂, CO, and C₂₊ alcohols (Fig. S33), with no detectable urea. This finding demonstrates that atomic-level proximity achieved by co-sputtering is critical for cooperative interactions and the relay-type mechanism required for urea synthesis.

Given the importance of Cu and Co mixing, we also explored the effects of the Co concentration/ratio in the Cu_(n)/Co_(1-n) matrix on the FE and yield rate of urea. As compared to Cu₁Co₁, the Cu₂Co₁ composition produces lower j at potentials more negative than −1.0 V vs. RHE (Fig. 4b). Methylamine is detected as the main C‒N product, albeit at a low FE of less than 1%, within the potential range of −1.0 to −1.2 V vs. RHE (Fig. 4c). The increased amount of Cu in the catalyst favored the C‒C coupling, producing C₂₊ products such as ethanol and n-propanol. At −1.2 V vs. RHE, the FEs for ethanol and n-propanol achieve 22 ± 4% and 2.1 ± 0.4 %, respectively. At −1.3 V vs. RHE, the FE of both products improved to 25 ± 2% and 2.4 ± 0.7%, accordingly. Conversely, the use of the Co-rich Cu₁Co₂ catalyst enhances j for the electroreduction of CO₂ and NO₂⁻ at potentials more negative than −1.0 V vs. RHE as compared to Cu₁Co₁ (Fig. 4b). However, neither urea nor methylamine were produced. The predominant products from the Cu₁Co₂ catalyst are ammonium, formate, and H₂. These results highlight the importance of a specific Cu:Co ratio in balancing the adsorption and reduction of reactants, thus promoting urea generation without overshadowing the process with excessive side reactions (HER, CO₂RR, and NO₂⁻RR). Moreover, all Cu_(n)Co_(1-n) catalysts demonstrated good stability over the course of 12 h chronoamperometric experiments for the co-reduction of CO₂ and nitrite, as confirmed by ICP-MS analysis (Fig. S35), showing less than 10 wt% loss of Cu and Co. This loss is considered low compared to the literature^30,86.

Beyond tuning the Cu:Co ratio, we also explored the effects of the ion-exchange membrane and pH of the catholyte and anolyte on the effectiveness of the C‒N coupling for the urea generation. Control experiments were performed by adjusting the pH of both catholyte and anolyte to acidic, neutral, and alkaline conditions under continuous CO₂-gas flow (Table S6, Supplementary Data 5). No urea was detected after co-reduction of CO₂ and NO₂⁻ at −1.2 V vs. RHE under either acidic (pH 3 and 5, K₂SO₄, Nafion membrane) or alkaline conditions (pH 10 and 13, KOH, anion exchange membrane (AEM)). Under alkaline conditions, formate was the dominant product, acidic electrolytes suppressed C‒N coupling and steered the reaction toward H₂ and ammonium generation (Fig. S36). Urea formation was consistently observed only when KHCO₃ was used as buffer in both catholyte and anolyte, together with Nafion as membrane (Fig. 4d and Fig. S31). Notably, replacing Nafion membrane with AEM under identical conditions eliminated urea production (Table S6, Supplementary Data 5). These results indicate that controlled proton availability is critical for regulating surface protonation steps, consistent with a PCET pathway rather than mediation by surface-bound hydrogen (*H)^87,88,89. This conclusion is further supported by in situ SR-FTIR measurements (Fig. S41), which show the absence of broad vibrational signatures characteristic of *H-related species in the 3300–3400 cm⁻¹ region^90,91.

To rationalize these trends, we consider the electrolyte-dependent mechanistic factors. Neutral KHCO₃ maintains relatively high dissolved CO₂ concentrations, in contrast to alkaline conditions where CO₂ is converted to HCO₃⁻/CO₃²⁻, thereby enabling effective *CO/*COOH formation⁹². In acidic electrolytes, *CO can form but HER dominates surface sites and mass transport, strongly inhibiting C–N coupling⁹³. For NO₂⁻ reduction, neutral pH provides sufficient proton activity to drive *NO hydrogenation to *NH₂ without excessive HER, whereas limited proton availability in alkaline electrolytes slows hydrogenation steps and reduces *NH₂ coverage^94,95,96. Furthermore, neutral KHCO₃ mitigates cathodic alkalinization, stabilizing intermediates such as *NCO and *NH₂CO without severe *OH blocking that occurs in strongly alkaline media^97,98,99.

Anolyte variation further highlighted the importance of proton flux (Table S6, Supplementary Data 5). Crucially, while the anolyte pH was varied to modulate charge transport, the catholyte bulk pH was maintained as intended, confirmed by negligible shifts (≤ ± 0.2 pH units) before and after testing. Maintaining the catholyte at neutral KHCO₃ but lowering the anolyte to pH 5 enabled detectable urea (FE 2 ± 0.2%, yield rate 14 ± 2 mmol h^‒1 g_cat^‒1; Fig. 4F and Fig. S37), whereas excessive proton concentration (pH 3) diminished CO₂ availability and enhanced competing H₂/NH₄⁺ pathways. This delicate balance in H⁺ concentration reflects the interplay between CO₂ activation and NO₂⁻ hydrogenation, in which optimal PCET-driven protonation enables co-coverage of *CO/*COOH and *NH₂^{23,88,100,101}. Consistent with this interpretation, in situ SR-FTIR detected their co-existence alongside urea vibrational bands under neutral conditions (Fig. S41-S42). These results indicate that urea formation depends on both an optimal Cu–Co surface and controlled proton management across both membrane and electrolyte compartments.

The role of electrolyte pH on catalyst stability is further reflected in CVs recorded in the non-catalytic potential window (Fig. S34). Acidic electrolytes promoted rapid Cu/Co redox cycling, consistent with accelerated dissolution^102,103, whereas alkaline electrolytes induced broad, hysteretic oxidation waves characteristic of irreversible oxide/hydroxide formation⁹⁹. By contrast, neutral KHCO₃ displayed only weak and broadened transitions, indicating moderate surface dynamics. These CV signatures demonstrate that acidic and alkaline conditions drive aggressive redox activity and oxide formation, compromising structural integrity, whereas neutral conditions preserve a stable Cu–Co interface.

The instability was corroborated by ICP analysis of Cu and Co dissolution under strongly acidic and alkaline environments (Fig. S35). In acidic media (pH 3), Cu and Co dissolution achieved 8 ± 2 wt% and 17 ± 2 wt%, respectively. In alkaline media (pH 13), dissolution was lower (3 ± 1 wt% for Cu and 8 ± 2 wt% for Co). Such losses reduce the long-term stability and performance of the catalysts. Finally, we observed that urea is prone to decomposition into ammonium and formate/formic acid under highly acidic (pH 3, K₂SO₄) and alkaline (pH 13, KOH) conditions (Fig. S38), especially at low concentrations.

In situ spectroelectrochemical studies

To correlate surface intermediates with catalyst structure, we combined in situ Raman and SR-FTIR with XAS. In situ Raman spectroscopy (900–3400 cm⁻¹) tracked the evolution of intermediates during CO₂ and NO₂⁻ electroreduction (Fig. 5a and Fig. S39). Under open circuit potential (OCP), bands from bidentate nitrite (νNO₂, ~1360 cm⁻¹) and buffer solution (HCO₃⁻, ~1062 cm⁻¹) were observed and persisted at −1.2 V vs. RHE^104,105,106. After 20 min at −1.2 V vs. RHE, new features at ~730 (NO bending) and ~2116 cm⁻¹ (CO stretching) appeared, consistent with the reduction of NO₂⁻ and CO₂, respectively^107,108,109. Urea related vibrations at 1002 cm⁻¹ (C‒N) and 575 cm⁻¹ (amide OCN) emerged after 30 min, but were absent in the control CO₂RR or NO₂⁻RR (Fig. S38), indicating its specific association with C–N coupling¹⁰⁷. Similar intermediates were identified under different applied potentials (Fig.S39), with C‒N bond and OCN bending identified from −0.9 V vs. RHE. Nevertheless, the amine group intermediate signals (e.g., *NH₂) resulted from further reduction of NO₂⁻ and NO are likely undetectable in Raman spectroscopy due to weaker NH₂ vibrations and low laser power¹¹⁰.

Fig. 5: In situ spectroelectrochemical investigation of Cu₁Co₁ catalyst.

(a) In situ Raman spectra of Cu₁Co₁ catalyst at −1.2 V vs. RHE. In situ SR-FTIR spectra of Cu₁Co₁ catalyst in the range of 900–2200 cm⁻¹ (b) and 2800–3400 cm⁻¹ (c). In situ XANES spectra (d), k³-weighted EXAFS spectra (e), and FT-EXAFS spectra (f) of Cu₁Co₁ catalyst at Cu K-edge. In situ XANES spectra (g), k³-weighted EXAFS spectra (h), and FT-EXAFS spectra (i) of Cu₁Co₁ catalyst at Co K-edge. All experiments were conducted in a CO₂-saturated 0.2 M KHCO₃ and 0.02 M KNO₂ solution.

In situ SR-FTIR analysis (900–3400 cm⁻¹) provided complementary evidence of intermediate formation at various applied potentials (Fig. 5b, c). Bands assigned to *NH₂ bending mode (~3115 cm⁻¹), *CO stretching (~2065 cm⁻¹), *NH₂ rocking vibration (~1150 cm⁻¹)_, and *NH₃ rocking/wagging mode (~930 cm⁻¹) appeared from −0.8 V vs. RHE, indicative of CO₂ and NO₂⁻ reduction^{14,82,111,112,113}. An amide-related *NCO vibrations (e.g., *NHCO, *NH₂CO, or *NH₂COOH), emerged around 1694 cm⁻¹ from −0.9 V vs. RHE, suggesting the onset of C‒N coupling formation^14,16,78. Vibrational features assigned to the carbonyl (*CO) and *CN bonds of urea were observed at ~1590 cm⁻¹ and 1455 cm⁻¹, respectively, over the potential range of −0.9 V to −1.3 V vs. RHE, align closely with the C=O and C‒N stretching frequencies of the urea reference (1589 cm⁻¹ and 1457 cm⁻¹, respectively; Fig. S40). In contrast, control CO₂RR and NO₂⁻RR experiments display only *CO/*COOH and *NH₂/*NH₃ features (Fig. S41), without any C–N features, thereby indicating that *CO/*COOH and *NH₂ likely act as precursors for C‒N coupling^14,16,17,78.

Catalyst comparison highlighted the role of the Cu‒Co interface. Urea-associated carbonyl and *CN vibrational features were exclusively observed for Cu₁Co₁ (Fig. S42). Pure Cu exhibited mainly *CO (~2065 cm⁻¹) and *COOH ( ~ 1720 cm⁻¹) bands associated with gaseous CO and formate formation, along with a C‒O stretching band (*CO) at ~1030 cm⁻¹ corresponding to methanol¹¹⁴. These results indicate that CO₂RR dominates on pure Cu, consistent with earlier observations. In contrast, pure Co displayed no carbonyl or *CN bands, yielding mainly formate and ammonia. These results indicate that the Cu–Co interface promotes the C–N coupling steps leading to urea.

Isotope labeling with ¹⁵NO₃⁻ confirmed the nitrogen origin of the *CN band (Fig. S43). Spectra for ¹⁴NO₂⁻ and ¹⁴NO₃⁻ were nearly identical, apart from an additional *NO₃⁻ feature at ~1340 cm⁻^{1 113}. Substitution with ¹⁵NO₃⁻ shifted nitrogen-containing vibrational modes to lower wavenumbers, with the *C¹⁵N feature of urea detected at 1411 cm⁻¹, while the carbonyl peak remained unchanged¹¹⁵. This isotopic shift confirms that the *CN signal originates from NO_x⁻-derived nitrogen.

Catalyst structure and dynamics were probed by in situ Raman and XAS. In situ Raman spectra in the 300–1000 cm⁻¹ region showed no oxides or hydroxides signals under negative potentials (Fig. S39), suggesting that the active sites during the electroreduction are in the metallic forms of Cu and Co. Only at OCP, surface hydroxides (Co(OH)₂ and Cu(OH)₂) were detected at ~502.5 cm⁻¹, implying the interaction between Cu/Co and the electrolyte^31,58,116.

The results from in situ Raman are corroborated by in situ XAS analysis at Cu and Co K-edges. The in situ XANES of Cu₁Co₁ catalyst at Cu K-edge (Fig. 5d and Fig. S44) reveals a characteristic XANES consistent with Cu(0) and a 3d¹⁰ electron configuration^25,117. A slight shift to lower effective nuclear charge in the rising edge between 8980 and 8990 eV was observed under increasingly applied potentials, suggesting some surface copper oxides is converted to Cu(0)¹¹⁸. Additionally, the main peak (white line) during in situ conditions is shifted to lower energies (Fig. S44), consistent with a transient reduction associated with adsorbed reactants during electrocatalytic reactions¹¹⁹. The k³-weighted EXAFS (Fig. 5e) and normalized FT-EXAFS (Fig. 5f) at Cu K-edge confirm that the majority Cu retains characteristics of metallic Cu under negative potentials, indicating that Cu returns to its metallic state. OCP measurements before and after reaction show that Cu₁Co₁ exhibits Cu–O and Cu–Cu interactions at apparent distances of ~1.4 and ~2.2 Å, respectively. Additionally, LCF of XANES and EXAFS spectra (Fig. S49–S50) confirms that Cu(0) is the dominant component under cathodic potentials, with copper oxides and hydroxides (i.e., Cu(OH)₂) each accounting for <10%. These minor oxides/hydroxides species re-emerge once the potential is removed, driven by re-oxidation upon contact with the electrolyte^65,120. Similar interaction patterns are also observed during the CO₂RR and NO₂⁻RR using Cu₁Co₁ and Cu (Fig. S45-S51).

In situ XAS measurements at the Co K-edge, examining both near-edge and extended-edge structures, reveal a gradual decline in the oxidation state of the catalyst to metallic Co with increasing cathodic potential (Fig. 5g–i and Fig. S52). This change is marked by a subtle shift towards a lower effective nuclear charge between 7710 and 7718 eV, accompanied by a decrease in white-line intensity, both indicative of a reduction in the oxidation state of Co (Fig. 5g and Fig. S52). Unlike copper, which tends to form oxides under OCP condition, the k³-weighted EXAFS (Fig. 5H) and normalized FT-EXAFS (Fig. 5i) at Co K-edge demonstrate that Co predominantly retains the properties of metallic Co under both OCP conditions and negative potentials. This is likely attributed to the dissolution of Co²⁺/Co³⁺ species into the catholyte, rendering them undetectable by XAS. The FT-EXAFS analysis further shows a consistent increase in the intensity of Co–Co coordination at an apparent distance of 2.1 Å with the application of negative potential (Fig. S52). Comparable interaction patterns are also observed for both CO₂RR and NO₂⁻RR using Cu₁Co₁ and Co (Fig. S53–S55). LCF of the XANES and EXAFS spectra further verifies that metallic Co is the principal phase under both OCP and negative potential, with trace of Co₃O₄ and Co(OH)₂ during cathodic potential (Fig. S56–S58).

Together, these results indicate a strong structure-activity correlation. Minor copper oxides/hydroxides detected by LCF EXAFS/XANES may, in principle, stabilize polar intermediates such as *NCO^121,122,123. Similarly, cobalt oxides/hydroxides have been reported to enhance NO₂⁻ activation and hydrogenation^31,124. However, these species are present only in trace amounts, and their limited surface coverage or transient lifetime appears insufficient under electroreduction conditions, as depicted in CV measurements (Fig. S23). This is consistent with the in situ SR-FTIR spectra of pure Cu, which show only weak *CN features (Fig. S42). It should also be noted that LCF provides bulk-averaged phase fractions and cannot resolve whether these minority species are surface-localized, catalytically accessible, or stable under bias; hence, their role is considered indicative rather than definitive for the active surface. Overall, the evidence indicates that Cu alone enhances CO₂ adsorption and stabilizes *CO/*COOH intermediates but lacks sufficient *NH₂ coverage; Co alone reduces NO₂⁻ to *NH₂/*NH₃ but provides inadequate *CO (Fig. S42). In contrast, at the Cu–Co interface, CO₂-derived *CO/*COOH from Cu and NO₂⁻-derived *NH₂ from Co are generated in close proximity, which enhances the probability of C–N bond formation. Based on these mechanistic observations, we then performed DFT modeling of the reaction pathways.

DFT calculations

Experimental results, including ex situ XAS and XPS, alongside in situ measurements of Raman and XAS, have indicated both Cu and Co metals as the dominant phases. DFT calculations are subsequently used to further understand the function of Cu and Co in the catalyst. Both Cu and Co stabilize *H intermediates on their surfaces (∆G = −0.54 and −0.62 eV) (Fig. 6a), which suppresses H₂ evolution but might not necessarily lead to significant *H-mediated hydrogenation. Rather, this stabilization maintains the catalyst surface in a state conducive to PCET, whereby proton transfer from the electrolyte, coupled with electron transfer, drives the hydrogenation of *CO/*COOH and *NO₂⁻-derived intermediates¹²⁵. This interpretation aligns with our experimental observations that urea formation is sensitive to proton flux across the membrane, consistent with a PCET-controlled mechanism.

Fig. 6: The DFT calculations for CO₂RR, NO₂⁻RR, and electrochemical C‒N coupling.

The DFT calculations for HER (a), CO₂ reduction to CO (b), and NO₂⁻ reduction to NH₃ (c). d Species configuration of C‒N coupling between *NH + *COOH, *NH + *CO, and *NH₂ + *COOH, on Cu₁Co₁ catalyst. e The DFT calculations of electrochemical C‒N coupling. f Schematic illustration of Cu₁Co₁ catalyst that shows favorability of *NH₂ coupling with *COOH. The dash lines in (c) show possible DFT-calculated reaction pathways.

DFT calculations indicate that Cu and Co exhibit distinct but complementary functions within the tandem system: (i) Cu enhances CO₂ adsorption, while Co facilitates *COOH and *CO formation during CO₂RR (Fig. 6b); and (ii) Co promotes NO₂⁻ adsorption, whereas Cu favors *NH₂ generation during NO₂⁻RR (Fig. 6c and Fig. S59). This synergistic interaction aligns with in situ spectroscopic observation of *CO/*COOH and *NH₂ co-appearance under proton-accessible conditions. Calculations of C–N coupling pathways (Fig. 6d) further indicate that *NH₂ + *COOH coupling is the most thermodynamically favorable route toward urea, with the formation of *NH₂CO identified as the potential-determining step (ΔG = 0.78, 0.22, and 0.57 eV on Cu, Co, and Cu–Co interface, respectively; Fig. 6e). Once *NH₂CO is formed, subsequent PCET steps toward *NH₂CONH and *NH₂CONH₂ proceed spontaneously (Fig. 6f).

We acknowledge that our present models likely underestimate the contribution of defect-rich or amorphous sites, which could further lower activation barriers and stabilize key intermediates¹²⁶. Future studies employing large-scale simulations or machine-learning-assisted approaches will be needed to better capture the structural complexity of sputtered catalysts¹²⁷. It should also be noted that the DFT calculations are based on idealized surface models under vacuum conditions and do not fully capture solvent effects, dynamic surface reconstructions, or potential-dependent processes that occur during catalysis.

Collectively, the DFT results are qualitatively consistent with the experimental observations, suggesting that the Cu–Co interface likely balances C–N intermediate adsorption, lowers the barrier for *NH₂CO formation, and facilitates proton-driven hydrogenation steps. This proposed synergy offers a plausible explanation for the experimentally observed dependence of urea formation on proton flux and electrolyte environment. Furthermore, the calculations imply that Co may mitigate the overly strong adsorption of C–N intermediates on pure Cu, while the spatial proximity of Cu and Co sites promotes the overall catalytic activity.

Discussion

In summary, we report a strategy for the electroreduction of CO₂ and NO₂⁻ into urea using a co-sputtered Cu₁Co₁ catalyst. This catalyst exhibits an atomic-level mixing of Cu and Co, and shows lower charge transfer resistance compared to pure Cu and Co catalysts. The Cu₁Co₁ catalyst demonstrates C–N coupling capabilities, achieving a maximum urea FE and yield rate of 11 ± 2 % and 61 ± 6 mmol h^‒1 g_cat^‒1 at −1.2 V vs. RHE (j = −64 mA cm⁻²). The results indicate that both the Cu:Co ratio and electrolyte environment strongly influence C–N coupling. A 1:1 Cu–Co ratio provides an optimal balance between adsorption and reduction of intermediates, promoting urea formation while suppressing competing HER, CO₂RR, and NO₂⁻RR. Urea formation is favored under neutral KHCO₃, in which proton availability is appropriately balanced to sustain PCET without diverting the pathway toward competing H₂ evolution. In situ Raman, SR-FTIR, and XAS indicate that C–N coupling proceeds through relay mechanism at the Cu–Co interface: Cu promotes CO₂ adsorption and NO₂⁻ activation, while Co facilitates NO₂⁻ adsorption and *CO/COOH formation. DFT calculations further support that urea generation arises from the coupling of *NH₂ and *COOH.

Methods

Materials

Catalyst precursors used to produce the co-sputtered catalyst were Cu (AJA International USA, ≥99.99%) and Co (AJA International USA, ≥99.9%). Carbon-based electrodes were using carbon fiber papers (Fuel cell store, 5% PTFE, 0.28 mm thickness, 0.44 g cm⁻³ bulk density, and 78% porosity).

Electrolyte solutions were prepared using KHCO₃ (Sigma-Aldrich, ≥99.7%), KNO₂ (Sigma-Aldrich, ≥99.6%), KOH (Sigma-Aldrich, ≥90%, flakes), K₂SO₄ (Sigma-Aldrich, ≥90%, powder), H₂SO₄ (Sigma-Aldrich, 95–98%) and deionized water (Satorius™ Arium®, 18.2 MΩ cm⁻¹ at 23 ± 2 °C). The gases for electrochemical testing used CO₂ (Coregas, grade 2.5) and Ar (Coregas, grade 5).

Ammonium identification used Berthelot method employing NH₄Cl (Sigma-Aldrich, ≥99.99%), NaOH (Sigma-Aldrich, ≥98%), salicylic acid (Sigma-Aldrich, ≥99%), sodium citrates (Sigma-Aldrich, ≥99%), sodium hypochlorite solution (Sigma-Aldrich, 4.00–4.99%), and sodium nitroferricyanide solution (Sigma-Aldrich, ≥99%). The ¹H-NMR for CO₂RR liquid product and urea identification used DMSO-d₆ (Cambridge Isotope Laboratories, D-99.9%) as a solvent and sodium acetate (Sigma-Aldrich, ≥99%) as an external standard.

Identification of urea via GC-MS was employed using chemicals including acetonitrile (Sigma Aldrich, 99.8%), N-methyl-N-(t-butyldimethylsilyl)trifluoroacetamide (MtBSTFA) (Sigma Aldrich, >97%), and toluene (Sigma Aldrich, ≥99.8%). The urea identification via NMR isotope tracing was applied using K¹⁵NO3 (Sigma Aldrich, 10 atom % ¹⁵N).

Cu₂O (Sigma Aldrich, ≥99.999%, anhydrous), CuO (Sigma Aldrich, 99.99%), Cu(OH)₂ (Sigma Aldrich, ≥95%), CoO (Sigma Aldrich, ≥99.5%), Co(OH)₂ (Sigma Aldrich, ≥95%), and Co₃O₄ (Sigma Aldrich, ≥99.5%) were applied as reference materials for XAS characterization of the catalyst.

Catalyst preparation

Cobalt and copper were deposited onto the carbon fiber paper substrate by simultaneous co-sputtering DC magnetron guns (1000 W) in 1200 s to obtain Cu‒Co samples. Argon was used during the deposition in the sputter coater system (AJA Intl. – ATC-2400V). The stoichiometric ratios of Co and Cu were controlled through tuning the power of gun sources. The catalyst loading is 0.2 ± 0.02 mg cm⁻².

Electrochemical measurements

Electrosynthesis was performed in a three-electrode customized H-cell system using a Nafion 117 membrane to separate the compartments within a gas-tight cell. A platinum (Pt) plate and saturated calomel electrode (SCE) served as the counter and reference electrode, respectively. The reference electrode was calibrated against the reversible hydrogen electrode (RHE) in the same electrolyte prior to all measurements. The calibration was performed using a clean Pt plate as the working and counter electrodes in a H₂-saturated electrolyte. The open-circuit potential was recorded until stabilization. This calibration procedure was repeated regularly to ensure consistency across measurements. The electrolyte consisted of Ar-saturated 0.2 M KHCO₃, Ar-saturated 0.2 M KHCO₃ + 0.02 M KNO₂, and CO₂-saturated 0.2 M KHCO₃ were utilized as the catholyte for HER, NO₂⁻RR and CO₂RR, respectively. The catholyte for C−N coupling electrosynthesis used CO₂-saturated 0.2 M KHCO₃ + 0.02 M KNO₂ solution. The anolyte for all electrochemical measurements used 0.2 M KHCO₃ solution but without gas saturation. All electrolytes were prepared freshly on the day of electrolysis to ensure solution stability and reproducibility. Before experiments, the cathodic compartment of the cell was purged with the relevant gas for 10 min, achieving pH levels of 8 ± 0.2, 7 ± 0.2, and 7 ± 0.2 for NO₂⁻RR, CO₂RR and C–N coupling electrosynthesis, respectively. The Ar or CO₂ gas flowrate was maintained at 10 mL min⁻¹ (9.4 standard liter per minute (SLPM)) during the HER, NO₂⁻RR, CO₂RR and C–N coupling reactions. The gas flow rate was quantified using a flowmeter (Dwyer RM Rate-Master) integrated into the gas inlet line of the cell. Electrochemical testing was conducted with an Autolab M204 (Metrohm Autolab) electrochemical workstation. The tests for NO₂⁻RR and CO₂RR are conducted over a duration of 1 h, while electroreduction of CO₂ and NO₂⁻ is carried out between 5 and 12 h. Fixed potential electrolysis was then carried out in triplicate. Following electrolysis, all potential were manually iR corrected, where i is the measured current and R is the electrolyte resistance (Fig. S27). The corrected potentials were then converted to the RHE reference for benchmarking with the literature using the following equation: E_RHE (V) = E_SCE (V) + 0.245 + 0.059 × pH for acidic/neutral conditions or E_RHE (V) = E_Hg/HgO (V) + 0.098 + 0.059 × pH for alkaline conditions. During the adjustment of the catholyte and anolyte to acidic and alkaline conditions, as mentioned in the main text, a 0.2 M K₂SO₄ solution served as the buffer for the acidic electrolyte, with the pH being adjusted using H₂SO₄. Conversely, the alkaline electrolyte utilized a KOH solution, the concentration of which varied to achieve the desired pH levels. A Hg/HgO electrode was employed as the reference electrode in alkaline electrolytes.

The cyclic voltammetry (CV) for the electrochemically active surface area (ECSA) calculations is measured with certain potential ranges (referred to Fig. S26) at scan rate of 5, 10, 20, 40, 60, 80, and 100 mV s⁻¹. The difference between the cathodic and anodic sweeps in the CV curves plotted versus the scan rate of the CV, then fit linearly to obtain the slope that represents the double-layer capacitance (C_dl). The ECSA calculation follows the equation: ECSA = C_dl/C_s, where C_dl is shown in (Fig. S26), and C_s is the capacitance of an atomically planar surface (C_s is estimated to be 40 μF cm⁻²)^128,129. The ECSA for each electrode size was then calculated by the following equation: ECSA = (cm²_ECSA per cm²_GEOMETRIC) × (cm²_GEOMETRIC)¹³⁰. The electrochemical impedance spectroscopy (EIS) was measured under −1.2 V vs. RHE in a CO₂-saturated 0.2 M KHCO₃ and 0.02 M KNO₂ solution with the frequency from 100 kHz to 0.1 Hz. The EIS fitting was conducted using The ZFit function in ECLab software version 11.42.

Product quantification

A Shimadzu gas chromatograph (Model 2010 Plus) equipped with both a thermal conductivity detector (TCD) and a flame ionization detector (FID) was employed to detect gas products such as H₂, CO, and CH₄, with calibration data provided in Fig. S60. For the quantification of liquid products from CO₂RR, ¹H NMR spectroscopy was used. Specifically, ~400 µL of catholyte was combined with 200 µL of DMSO-d₆ and analyzed using a 600 MHz ¹H 1D liquid NMR spectrometer equipped with a cryoprobe over 128 scans. An external standard method utilizing 10 mmol mL⁻¹ of sodium acetate facilitated quantification.

Ammonium levels were measured using ~0.5 mL of catholyte, employing the Berthelot reaction. The reaction involved mixing the catholyte with reagent 1 (comprising 0.4 mL of 1 M sodium hydroxide, 5 wt% salicylic acid, and 5 wt% sodium citrates). This mixture was then combined with 0.1 mL of reagent 2 (0.05 M sodium hypochlorite solution) and 30 µL of reagent 3 (1 wt% sodium nitroferricyanide solution). Following sonication and a 2-h incubation in the dark at room temperature, the ammonium concentration was quantified using a Shimadzu UV-3600 UV-vis-NIR spectrophotometer and 1 cm quartz cuvettes, measuring absorbance between 550 and 800 nm wavelengths. Calibration curve data (Fig. S61) from electrolyte solutions with known NH₄Cl concentrations were used for this purpose.

Furthermore, identification and quantification of urea and other liquid products in C–N coupling electrosynthesis were performed using the ¹H NMR method as described by Huang et al.¹³¹. Approximately 400 µL of liquid sample was mixed with 200 µL of DMSO-d₆ solution (D-99.9%, Cambridge Isotope Laboratories) and analyzed in a 600 MHz ¹H 1D liquid NMR spectrometer (Bruker Advance) equipped with a cryoprobe over 512 scans¹³¹. The ¹H spectrum was recorded using the water suppression excitation sculpting technique, and quantification of urea was determined using the calibration curve of various urea concentration detected in the ¹H-NMR method (Fig. S61). For isotope tracing of urea using the ¹H NMR method, K¹⁵NO3 was employed as nitrogen sources.

Other than NMR technique, GC-MS is also used to identify urea. Approximately 100 µL liquid sample was dried using nitrogen evaporator needle system (~10 min per 50 µL). After drying, sample was derivatized to its t-butyldimethylsilyl ester form at 60 °C for 1 h by adding 10 µL acetonitrile and 40 µL N-methyl-N-(t-butyldimethylsilyl)trifluoroacetamide (MtBSTFA) that consisted of 1% t-butyldimethylchlorosilane (tBDMCS). Derivatives were reconstituted in 40 µL toluene, then injected (1 µL splitless = 0.25–1.25 µL media equivalent on column) into GC-MS for analysis. If sample consists of urea, it will contain urea derivatives, Urea, N, N’-bis(tert-butyldimethylsilyl).

The faradaic efficiency (FE) and yield rate of liquid products were calculated to evaluate the catalytic performance using the following equations:

$${FE}(\%)=\frac{{n}\times {F}\times {{V}}_{{aq}}\times {{C}}_{{{\rm{Product}}}}}{{Q}}\times 100\%$$

(1)

where n is number of electron transfers, F is the Faraday constant (96485.33 C mol^–1), V_aq is the volume of catholyte solution (L), C_Product is the concentration of product (M), and Q is passed charge (C).

$${{Yield \, rate}}_{{{\rm{Product}}}}({mmol}{{h}}^{-1}{{g}}_{{cat}}^{-1})=\frac{\,{{C}}_{{{\rm{Product}}}}\times {{V}}_{{aq}}\times {MW}}{{t}\times {A}\times {\varGamma }}$$

(2)

where MW is the molecular weight of product, t is duration of the electrolysis (h), A is the geometric surface area of the electrode (cm²), and Γ is the catalyst loading per geometric surface area of the electrode (g cm^–2); C_Product for this calculation is expressed in mM.

The faradaic efficiency of the gaseous products was calculated using the following equation:

$${FE}(\%)=\frac{{n}\times {F}\times {V}_{{{{\rm{gas}}}}}\times {x}}{{Q}}\times 100\%$$

(3)

where V_gas is the volume of headspace (L), and ${\mbox{x}}$ is the concentration of the product in the headspace (mol L_gas^–1).

Material characterization

XRD patterns were obtained using a D2 Phaser Bruker instrument equipped with Cu Kα radiation (λ = 1.54 Å) across a scan range of 10° to 80° at a rate of 1.17 degrees per minute. The catalyst on CFP was positioned on a zero-diffraction plate made of silicon crystal. XPS was performed using a Thermo ESCALAB250Xi spectrometer, utilizing a monochromated Al Kα source (energy 1486.68 eV), operated at 120 W (13.8 kV × 8.7 mA), with a hemispherical analyzer in fixed analyzer transmission mode. The vacuum chamber maintained a pressure typically between 10⁻⁹ and 10⁻⁸ mbar. Survey spectra were recorded at a pass energy of 100 eV, and high-resolution spectra for individual peaks were acquired at 20 eV pass energy. The photoelectron take-off angle was set at 90°, with a spot size of ~500 µm. Data analysis was facilitated using ThermoScientific Avantage software. Elemental identification and quantification were achieved through survey spectra, with atomic concentrations determined using integral peak intensities and sensitivity factors provided by the manufacturer. Binding energies were calibrated to the C 1s peak at 284.8 eV, representing aliphatic hydrocarbons. The accuracy of quantitative XPS is estimated at 10–15%, with a precision often exceeding 5%. The peak in the XPS spectrum was deconvoluted using the CasaXPS software.

Morphological analysis was performed using a Zeiss Ultraplus field-emission scanning electron microscope (FESEM) operating at 3 kV and a JEOL 2100F high-resolution TEM at 200 kV. For SEM, electrode samples were mounted on specimen stubs using carbon tape. TEM samples were prepared by scraping material off the electrode surface and dispersing it onto holey carbon-coated nickel grids. Elemental mapping was conducted using a JEOL SDD energy-dispersive X-ray spectrometer coupled with the JEOL 2100F. The HRTEM and SAED analysis were performed with a JEOL JEM-F200. Sample was place on in-situ holders. This instrument achieves a lattice resolution of 0.16 nm in annular dark-field scanning STEM (ADF-STEM) mode and 0.1 nm in TEM mode at 200 kV. Additionally, the F200 features a JEOL 100 mm² energy-dispersive X-ray (EDX) detector that offers an energy resolution of 130 eV (Mn K) and a collection angle of 0.9 steradian. Ex situ Raman measurements were performed using an inVia Raman spectrometer (Horiba JOBIN YVON US/HR800 UV) with a 633 nm laser wavelength. X-ray absorption spectroscopy (XAS) of the Cu K-edge and Co K-edge was carried out at the Australian Synchrotron (ANSTO), in the XAS and Medium Energy X-ray Absorption (MEX-1) spectroscopy beamlines. Fluorescence detection was used for powder data, with Cu foil, Cu₂O, Cu(OH)₂, CuO, Co foil, CoO, Co(OH)₂ and Co₃O₄ serving as reference samples. XAS data were analyzed using Athena software and the fitting using linear combination function. MS Excel and Origin were used for the final presentations of data and fine calibrations.

For post-reaction characterization, the catalysts were subjected to 12 h of electroreduction in CO₂-saturated 0.2 M KHCO₃ + 0.02 M KNO₂ solution. After electrolysis, the samples were collected, rinsed with deionized water, and stored in a vacuum chamber until further analysis.

In situ measurements

In situ SR-FTIR measurements were conducted at the Australian Synchrotron (ANSTO), Melbourne using an SR-FTIR setup that incorporated a ZnSe crystal as the infrared transmission window (cut-off energy 625 cm^–1). This setup was paired with an infrared microscope (Bruker Hyperion 3000 with a 20x objective) and an FTIR spectrometer (Bruker 70 v/s) featuring a KBr beam splitter and a liquid-nitrogen-cooled mercury cadmium telluride detector. The catalyst electrode was positioned closely against the ZnSe crystal window with a micrometer-scale gap to reduce infrared light loss. The system was operated in reflection mode with vertical infrared light incidence to maintain the integrity of the SR-FTIR spectra. Each infrared absorption spectrum was derived by averaging 256 scans at a resolution of 2 cm^–1. Before conducting systematic measurements, a background spectrum of the catalyst electrode was recorded at open-circuit voltage. The potential settings for the electroreduction reaction were established at −0.8, −0.9, −1.0, −1.1, −1.2, and −1.3 V vs. RHE using CO₂-saturated 0.2 M KHCO₃ and 0.02 M KNO₂ as catholyte. Fig. S62 illustrates the in situ electrochemical cell and setup configuration.

In situ Raman measurements were carried out using an inVia Raman spectrometer system (Horiba JOBIN YVON US/HR800 UV) with a spectro-electrochemical Raman cell. The Raman cell was operated using a 633 nm argon ion laser and was equipped with an Ag/AgCl reference electrode and a Pt wire counter electrode. The cell contained an electrolyte solution of CO₂-saturated 0.2 M KHCO₃ + 0.02 M KNO₂ and was connected to an Autolab potentiostat (PGSTAT204) for simulating electrocatalytic C–N coupling.

In situ XAS measurements was carried out based on previous experiment¹³². In situ Cu K-edge and Co K-edge XAS data were gathered on the multipole wiggler XAS beamline (12-ID) at the ANSTO, Melbourne. The beamline operated with an electron beam energy of 3.0 GeV and a beam current of 200 mA (top-up mode), using “mode 1” and “mode 2” for Co K-edge and Cu K-edge, respectively, with a Si(111) monochromator and focusing optics, achieving a beam flux of ~2 × 10¹¹ photons per second and a beam size of 0.5 mm × 0.3 mm. Data collection was conducted in fluorescence mode with a solid-state 18-element detector. The in situ cell (Fig. S63) was mounted on the beamline sample table, positioning the electrode at a 45° angle to the fluorescence detector. An OKEN ionization chamber measured the incident radiation, and all measurements were taken at ambient temperature (23 ± 1 °C).

To minimize air absorption of fluorescent X-rays, the sample environment was maintained in a helium atmosphere by connecting a flight tube between the beamline and the cell and applying a helium overpressure between the cell and detector. Each X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS, k = 12) spectrum required ~15 and 30 min, respectively. Data was collected at multiple sample spots to confirm homogeneity. The raw data was processed in Athena.

DFT calculations

First-principles density functional theory (DFT) calculations were performed using the Vienna Ab initio Simulation Package (VASP)^133,134. The electronic wave functions were treated using the Projector Augmented Wave (PAW) approach^135,136. The exchange correlation functionals were approximated by the generalized gradient approximation (GGA) as parameterized by the Perdew-Burke-Ernzerhof (PBE)¹³⁷. The plane-wave cut-off energy of 500 eV and 3 × 3 × 1 Gamma k-point were used for the electronic self-consistent calculation, until it was converged to 1 × 10⁻⁵ eV. The ionic relaxation steps were performed using the conjugate-gradient method (IBRION = 2) and continued until the total force on each atom dropped below a tolerance of –0.05 eV/Å.

Surface investigations were studied using slab methods, for the hybrid Co (111) and Cu (111) surfaces. A vacuum region with the thickness of 20 Å was introduced perpendicular to the surface (in the z-direction). To model the bulk-surface interface, atoms in the two outermost layers were allowed to relax, while the coordinates of atoms in other layers were fixed to their bulk positions. The free energy diagrams were developed by taking into accounts the zero-point energy and entropy contributions. Additionally, long-range van der Waals interactions were considered using DFT D-3 method¹³⁸.