Ammonia electrosynthesis from nitrate using a stable amorphous/crystalline dual-phase Cu catalyst

Abstract

Renewable energy-driven electrocatalytic nitrate reduction reaction presents a low-carbon and sustainable route for ammonia synthesis under mild conditions. Yet, the practical application of this process is currently hindered by unsatisfactory electrocatalytic activity and long-term stability. Herein we achieve high-rate ammonia electrosynthesis using a stable amorphous/crystalline dual-phase Cu catalyst. The ammonia partial current density and formation rate reach 3.33 ± 0.005 A cm⁻² and 15.5 ± 0.02 mmol h⁻¹ cm⁻² at a low cell voltage of 2.6 ± 0.01 V, respectively. Remarkably, the dual-phase Cu catalyst can maintain stable ammonia production with a Faradaic efficiency of around 90% at a high current density of 1.5 A cm⁻² for up to 300 h. A scale-up demonstration with an electrode size of 100 cm² achieves an ammonia formation rate as high as 11.9 ± 0.5 g h⁻¹ at a total current of 160 A. The impressive electrocatalytic performance is ascribed to the presence of stable amorphous Cu domains which promote the adsorption and hydrogenation of nitrogen-containing intermediates, thus improving reaction kinetics for ammonia formation. This work underscores the importance of stabilizing metastable amorphous structures for improving electrocatalytic reactivity and long-term stability.

The authors develop an efficient and stable amorphous/crystalline dual-phase Cu catalyst towards electrocatalytic nitrate reduction reaction, with an ammonia Faradaic efficiency of around 90% at a high current density of 1.5 A cm⁻² for up to 300 h.

Introduction

Ammonia (NH₃) is currently produced by the energy-intensive Haber–Bosch process which converts nitrogen (N₂) and hydrogen (H₂) to NH₃ at high temperatures (400−500 °C) and high pressures (10–30 MPa). The Haber–Bosch process accounts for 1–2% of global energy consumption and roughly 1% of global CO₂ emission^1,2. Alternative routes for green NH₃ synthesis under mild conditions are highly desired. The renewable energy-driven electrocatalytic nitrate reduction reaction (NO₃⁻RR), with NO₃⁻ (with concentrations from hundreds of ppm to >1 M) from wastewater being as a nitrogen source and water being as a hydrogen source, provides a low-carbon route for green NH₃ synthesis under ambient conditions, with promising energy and environmental sustainability^3,4.

The NH₃ production via NO₃⁻RR involves multiple proton and electron transfer steps (NO₃⁻+ 6H₂O + 8e⁻ → NH₃ + 9OH⁻), leading to sluggish reaction kinetics⁵. The NO₃⁻RR rate is governed by the adsorption of nitrogen species present during NO₃⁻RR and the formation of active hydrogen species from water dissociation^6,7. An efficient catalyst for NO₃⁻RR should exhibit balanced adsorption of the above intermediates, thus directing the reaction pathway to NH₃ at a high reaction rate while minimizing the formation of side products such as nitrite (NO₂⁻) and H₂^8,9. The adsorption of these intermediates is highly dependent on the geometric and electronic structures of metal catalysts, as demonstrated by remarkably distinct NO₃⁻RR performances over noble metal (e.g., Pt, Ru, Pd)^6,10,11, non-noble metal (e.g., Fe, Co, Ni, Cu)^7,12–18, and bimetallic (e.g., CuPd, CuNi) catalysts^19–23. Cu has been recognized as one of the most active non-noble metals, owing to its prominent ability for the conversion of NO₃⁻ to NO₂^−4,24. To date, while Cu-based catalysts have exhibited high NH₃ Faradaic efficiency²⁵, their activity and stability are still unsatisfactory for practical application.

Amorphous catalytic materials, with extremely disordered atomic arrangement in long range, possess high density of low-coordinated sites which are essential for the adsorption and activation of reactants and intermediates^26,27. However, the amorphous materials are metastable and prone to suffering from crystallization during reaction, especially under highly reducing electrochemical conditions^28–30. Here we report a stable amorphous/crystalline dual-phase Cu (a/c-Cu) catalyst for high-rate NH₃ electrosynthesis via NO₃⁻RR. The a/c-Cu catalyst exhibits a NH₃ partial current density of 3.33 ± 0.005 A cm⁻² and a NH₃ formation rate of 15.5 ± 0.02 mmol h⁻¹ cm⁻² at a low cell voltage of 2.6 ± 0.01 V in an alkaline membrane electrode assembly (MEA) electrolyzer. Remarkably, the NH₃ Faradaic efficiency maintains around 90% at an applied current density of 1.5 A cm⁻² for 300 h. The impressive NO₃⁻RR performance is rationalized by the presence of stable amorphous Cu domains which promote NH₃ generation by optimizing the adsorption of N-containing intermediates and facilitating the formation of active hydrogen species from water dissociation.

Results

NO₃⁻RR performance of a/c-Cu catalyst

The a/c-Cu catalyst was prepared by annealing a commercially available Cu foam in air at 600 °C followed by in situ electrochemical reduction during NO₃⁻RR (Supplementary Fig. 1). The NO₃⁻RR performance of the a/c-Cu catalyst was measured in a home-made alkaline MEA electrolyzer (Supplementary Fig. 2). The anolyte and catholyte were 1 M KOH and 1 M KOH + 0.2 M KNO₃ solutions, respectively. The NO₃⁻RR measurements were performed in the galvanostatic mode. The gas product (H₂) and liquid products (NH₃ and NO₂⁻) were analyzed and quantified using an on-line gas chromatography and an ultraviolet–visible (UV–Vis) spectrophotometer, respectively (Supplementary Figs. 3–5). The quantification of NH₃ was further validated by proton nuclear magnetic resonance (¹H NMR) analysis (Supplementary Figs. 6–8). Herein, the annealed Cu foam is directly used as a working electrode. Such an integrated electrode is featured with facile preparation and shows significantly improved electrocatalytic performance compared to a conventional electrode with drop-casted catalyst powder (Supplementary Fig. 9). The three-dimensional and highly porous foam structure can remarkably facilitate mass transport especially at high current densities^31,32. Figure 1a shows the Faradaic efficiencies and cell voltage as a function of applied current density over the a/c-Cu electrode. The a/c-Cu electrode achieves a peak NH₃ Faradaic efficiency of 92 ± 0.87% at an applied current density of 1.5 A cm⁻². The NH₃ Faradaic efficiency still maintains over 83% at an applied current density up to 4.0 A cm⁻², which translates into a NH₃ partial current density of 3.33 ± 0.005 A cm⁻² and a NH₃ formation rate of 15.5 ± 0.02 mmol h⁻¹ cm⁻² at a low cell voltage of 2.6 ± 0.01 V (Supplementary Fig. 10). The effect of annealing temperature on the NO₃⁻RR performance was also investigated by further measuring another two Cu foam samples annealed in air at 300 and 900 °C, respectively. The sample annealed at 600 °C (i.e., the a/c-Cu sample) shows the highest NO₃⁻RR performance (Supplementary Fig. 11). The NH₃ partial current density over our a/c-Cu electrode measured in 0.2 M KNO₃ outperforms a raw Cu foam electrode (Supplementary Fig. 12) as well as previously reported values achieved even with a higher KNO₃ concentration up to 1 M (Fig. 1b and Supplementary Table 1)^{6–8,12,17–20}. The peak full-cell energy efficiency for NH₃ production reaches 26 ± 1% at 1.0 A cm⁻² and outperforms most of previous studies (Fig. 1c, Supplementary Fig. 9 and Table 2)^{14,23,33–35}. The stability test of the a/c-Cu electrode was performed at an applied current density of 1.5 A cm⁻². In a course of 300 h, the cell voltage is almost stable and the NH₃ Faradaic efficiency maintains around 90% (Fig. 1d, Supplementary Fig. 13). We tried to scale up the NO₃⁻RR process by increasing the geometric electrode area from 1 to 100 cm² (Supplementary Fig. 14). As shown in Fig. 1e, the NH₃ Faradaic efficiency reaches 94 ± 3.9% at an applied current of 160 A (the used direct current power supply has an upper current limit of 170 A) and a cell voltage of 2.23 ± 0.005 V. The highest NH₃ formation rate reaches up to 11.9 ± 0.5 g h⁻¹ (Supplementary Fig. 15a) and the corresponding energy consumption for NH₃ production is as low as 108.3 ± 0.47 kJ g⁻¹_NH3 at 160 A (Supplementary Fig. 15b). The reported NO₃⁻RR performance in terms of NH₃ formation rate and energy consumption is superior to state-of-the-art cases in the literature (Fig. 1f)^33–36. Overall, by turning a commercially available Cu foam into an active a/c-Cu catalyst through facile annealing treatment, we have achieved very promising selectivity, reaction rate, and stability towards industrial ammonia synthesis via NO₃⁻RR.

Fig. 1. NO₃⁻RR performance of a/c-Cu catalyst measured in MEA electrolyzers.

a Faradaic efficiencies and cell voltage as a function of applied current density. Comparisons of b performance and c full-cell energy efficiency reported in this work measured in a 1-cm² electrolyzer with literature^{6–8,12,14,17–20,23,33–35}. d Stability test at 1.5 A cm⁻². e Scale-up performance with a 100-cm² electrolyzer and f performance comparison with literature^33–36. The error bars represent standard error of the mean and are made based on three fully separate and identical measurements. Source data are provided as a Source Data file.

Structure characterizations

To reveal structure-reactivity correlations of the a/c-Cu catalyst for the impressive NO₃⁻RR performance, a series of structural characterizations and control experiments were conducted. The annealed Cu foam in air at 600 °C was firstly characterized by X-ray diffraction (XRD) and scanning electron microscopy (SEM). The characteristic XRD pattern of a pure CuO phase indicates that the presence of a thick CuO layer on the raw Cu foam after annealing treatment in air (Supplementary Fig. 16). Meanwhile, the original flat surface of the raw Cu foam becomes very rough with irregularly shaped grains (Supplementary Fig. 17). After NO₃⁻RR, these large grains disappear and form a rough and porous layer (Supplementary Fig. 18), along with the reduction of CuO revealed by the XRD pattern of the post-reaction sample (Fig. 2a). The operando Cu K-edge X-ray absorption near edge structure (XANES) spectra of the a/c-Cu catalyst further confirm that CuO is fully reduced to metallic Cu during NO₃⁻RR (Fig. 2b and Supplementary Fig. 19). This is consistent with the results obtained from the Cu LMM Auger spectrum of the quasi in situ X-ray photoelectron spectroscopy (XPS) analysis (Supplementary Fig. 20)³⁷. The Fourier transformation of the extended X-ray absorption fine-structure (EXAFS) spectra show decreased peak intensity of the Cu-Cu coordination at 2.3 Å (Fig. 2c), and the average Cu coordination number estimated by fitting results is around 6.7 ± 0.4 (Supplementary Figs. 21, 22 and Supplementary Table 3). These results indicate the presence of abundant low-coordinated Cu sites that are in situ generated during NO₃⁻RR (Supplementary Fig. 23)^38,39. High-resolution transmission electron microscopy (HRTEM) images show that the a/c-Cu catalyst consists of abundant amorphous domains which are separated by crystalline regions (Fig. 2d and Supplementary Fig. 24). The amorphous/crystalline dual-phase structure is further confirmed by the corresponding fast Fourier transform (FFT) patterns acquired from amorphous and crystalline regions (Fig. 2e). High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images and energy-dispersive X-ray spectroscopy (EDS) elemental mapping analysis of the a/c Cu catalyst clearly show the co-existence of disordered Cu domains and crystalline Cu domains (Fig. 2f and Supplementary Fig. 25).

Fig. 2. Structure characterizations.

a XRD patterns of a/c-Cu and a/c-Cu-H₂ catalysts after NO₃⁻RR. Operando b normalized XANES and c Fourier-transformed EXAFS spectra of a/c-Cu catalyst at a working potential of –0.8 V (vs. RHE). d HRTEM images of a/c-Cu catalyst and e corresponding FFT patterns acquired from different regions. f HAADF-STEM image and corresponding EDS element map of a/c-Cu catalyst. g Ratios of amorphous Cu domains. The error bars represent standard error of the mean and are made based on ten fully separate and identical measurements. h DSC curves. i CV curves of Pb stripping. j NH₃ partial current density at a cell voltage of 2.35 V over a/c-Cu and a/c-Cu-H₂ catalysts. The error bars represent standard error of the mean and are made based on three fully separate and identical measurements. Source data are provided as a Source Data file.

To reveal whether the impressive NO₃⁻RR performance correlates with the presence of amorphous domains, we further treated the Cu foam annealed in air at 600 °C by annealing in H₂ at 350 °C prior to NO₃⁻RR and the control sample was denoted as a/c-Cu-H₂ (Supplementary Figs. 17c and 18c). As indicated by the much sharper XRD characteristic Cu patterns in Fig. 2a, the H₂ annealing treatment results in a significant increase in the crystallinity of the a/c-Cu-H₂ catalyst, compared with the a/c-Cu catalyst. Accordingly, HRTEM images of the a/c-Cu-H₂ catalyst indicate that the amorphous domains reduce drastically (Supplementary Fig. 26). Quantitative and statistical analysis based on ten typical HRTEM images for each sample demonstrates that the ratio of the amorphous domains decreases from 24.1 ± 3.3% over a/c-Cu to 5.5 ± 2.7% over a/c-Cu-H₂ (Fig. 2g). The increased crystallinity (thus, decreased amorphization) induced by the H₂ annealing treatment is further evidenced by differential scanning calorimetry (DSC) measurements (Fig. 2h). The exothermic peak observed at around 133 °C is assigned to the crystallization of the amorphous Cu domains⁴⁰. The larger exothermic peak suggests a higher ratio of the amorphous Cu domains in the a/c-Cu catalyst. It is widely accepted that amorphous materials possess abundant low-coordinated sites owing to their long-range atomic disorder²⁶. The Pb underpotential deposition (Pb-UPD) was used to characterize the surfaces of the a/c-Cu and a/c-Cu-H₂ catalysts. For the a/c-Cu catalyst, the Pb stripping peak at −0.15 V vs. Ag/AgCl which is assigned to low-coordinated sites is more prominent (Fig. 2i)⁴¹, consistent with its higher ratio of the amorphous domains (Fig. 2g). The NO₃⁻RR performance of the a/c-Cu-H₂ catalyst was also measured under identical reaction conditions (Supplementary Fig. 27). As shown in Fig. 2j, the NH₃ current density over the a/c-Cu-H₂ catalyst is much lower than that over the a/c-Cu catalyst at a cell voltage of 2.35 V. These control experiments indicate that the improved ammonia synthesis is positively correlated with the amorphous Cu domains.

It has been reported that amorphous materials are prone to suffering from atomic rearrangement processes such as crystallization under electrochemical reaction conditions^29,42. We further conducted Pb-UPD measurements over the a/c-Cu catalyst after different reaction durations (0.5, 1.0, and 2.0 h). As shown in Supplementary Figs. 28 and 29, the ratios of main facets and the amounts of low-coordinated sites (thus, amorphous domains) are almost constant over time, indicative of good structural stability of the dual-phase a/c-Cu catalyst during NO₃⁻RR. In contrast, a pure amorphous Cu catalyst (a-Cu) quickly turns crystalline under NO₃⁻RR conditions (Supplementary Fig. 30). Therefore, it is speculated that the amorphous Cu domains in the a/c Cu catalyst are likely kinetically stabilized by adjacent crystalline Cu domains^43,44. The a/c-Cu catalyst after reaction at 1.5 A cm⁻² for over 50 h is further characterized by HAADF-STEM (Supplementary Figs. 31 and 32), and the amorphous Cu domains can be still observed, further indicating the good structural stability. Moreover, as evidenced by the electrochemical impedance spectra (EIS) in Supplementary Fig. 33, the poor conductivity of amorphous catalysts⁴⁵ has also been largely improved owing to the presence of adjacent crystalline Cu domains, resulting in low Ohmic resistance and thus low energy consumption for NH₃ production (Fig. 1c, f). Nevertheless, the stable amorphous/crystalline dual-phase structure of the a/c-Cu catalyst, is responsible for the high current density and long-term stability towards NH₃ synthesis.

Spectroscopic investigations

The NH₃ formation from NO₃⁻RR involves multiple deoxygenation and hydrogenation steps which are closely associated with the adsorption of NO₃⁻ and N-containing intermediates as well as the generation of active hydrogen species from water dissociation^6,20. To figure out how the dual-phase a/c-Cu catalyst promotes NO₃⁻RR to NH₃, thorough spectroscopic characterizations and mechanistic investigation experiments were conducted (Supplementary Figs. 34 and 35). In situ Raman spectroscopy measurements were performed in a home-made flow cell to observe the adsorption and reduction of NO₃⁻ on the surface of the a/c-Cu catalyst. As shown in Fig. 3a, upon applying potentials, the NO₃⁻ species in the solution (1050 cm⁻¹)⁴⁶ begin to adsorb onto catalyst surface, as evidenced by the characteristic Raman peaks at 1005 and 1370 cm⁻¹ which correspond to the NO stretching vibration⁴⁷ and the NO₂ antisymmetric stretching vibration⁴⁸ of the adsorbed NO₃⁻ species, respectively. The Raman peak at 1098 cm⁻¹, assigned to copper-nitrate complexes, also suggests an interaction between copper and nitrate⁴⁷. The Raman peaks at 1120 and 1259 cm⁻¹ are typically assigned to the symmetric and antisymmetric stretching vibration of the important adsorbed NO₂⁻ intermediate in a nitro configuration^47,48. Furthermore, the presence of the chelating nitrito configuration (*ONO*⁻, at 1301 cm⁻¹) and the bridging nitro configuration (*N*OO⁻, at 1435 cm⁻¹) further demonstrating the strong adsorption capacity for NO_x intermediates on the a/c-Cu catalyst^47,48. The broad Raman peak at 1534 cm⁻¹ is likely be assigned to the ν(N = O) of HNO* (1530 cm⁻¹) and the antisymmetric bending vibration of the HNH of NH₃ (1550 cm⁻¹)^47,48. However, when the applied potential becomes more negative, the intensities of NO_x peaks decrease significantly while that of the Cu-NH₂ (387 and 446 cm⁻¹)⁴⁹ and the N–H bending mode of NH₂ (1654 cm⁻¹)^50 increase, indicative of the fast conversion of NO_x species. On the other hand, these important intermediates are almost invisible over the a/c-Cu-H₂ catalyst (Supplementary Fig. 36). This is probably associated with the presence of the amorphous Cu domains with abundant low-coordinated sites that are considered to be active for NO₃⁻RR^51,52. The presence of the adsorbed species over the a/c-Cu catalyst was further validated by in situ attenuated total reflectance surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) measurements (Fig. 3c). The downward and upward IR bands at 1340 and 1256 cm⁻¹ confirm the consumption and generation of NO₃⁻ and *NO₂, respectively⁵³. At the same time, the N–H bending band at 1442 cm⁻¹ and the NH₂ wagging band at 1294 cm⁻¹ are also detected owing to NH₃ formation^54,55.

Fig. 3. Spectroscopic investigations.

a In situ Raman spectra over a/c-Cu catalyst at different potentials (vs. RHE). b EPR spectra of outlet electrolytes with or without KNO₃ using DMPO as radical trapping reagent. c In situ ATR-SEIRAS spectra over a/c-Cu catalyst at different potentials (vs. RHE) and d KIEs over a/c-Cu and a/c-Cu-H₂ catalysts. Source data are provided as a Source Data file.

Apart from the adsorbed species, potential-dependent surface structure changes are also observed by in situ Raman measurements. The disappearance of the Raman peak at 298, 343, and 609 cm⁻¹ assigned to CuO at open circuit potential (OCP) indicates the reduction of CuO to metallic Cu during NO₃⁻RR (Fig. 3a), consistent with XRD, quasi in situ XPS, and XANES results. Meanwhile, the two emerging Raman peaks at 539 and 716 cm⁻¹ can be assigned to the symmetric stretching of Cu−OH^56,57 and the bending mode of Cu−OH^58,59, respectively. The surface-adsorbed –OH is considered to be produced from water dissociation during NO₃⁻RR (Fig. 3a). In contrast, only minor symmetric stretching of Cu−OH Raman peak can be observed over the a/c-Cu-H₂ catalyst (Supplementary Fig. 36), indicating the more favorable water dissociation owing to the presence of the amorphous Cu domains. The improved water disassociation is further verified by directly detecting the other dissociation product, adsorbed hydrogen (*H) species, at an applied current density of 2 A cm⁻². The *H detection was performed through being trapped with 5,5-dimethyl-1-pyrroline-N-oxide (DMPO) followed by electron paramagnetic resonance (EPR) measurements. Upon electrolysis in the absence of NO₃⁻ at 2 A cm⁻², the DMPO-H intensity of the a/c-Cu catalyst is much higher than that of the a/c-Cu-H₂, indicating the a/c-Cu has stronger ability to drive the H₂O dissociation and generate abundant *H species⁶⁰. After adding NO₃⁻, the DMPO-H intensity of both the a/c-Cu and a/c-Cu-H₂ catalyst decrease significantly due to the *H consumption by NO₃⁻ reduction. However, the stronger DMPO-H peaks over the a/c-Cu catalyst indicate the more favorable *H generation compared with the a/c-Cu-H₂ catalyst⁶¹, in agreement with the Raman results. As hydrogenation steps are significantly affected by proton transfer rate, the kinetic isotope effect (KIE) of H/D is determined. When D₂O is used instead of H₂O in the electrolyte, the current density decreases (Supplementary Fig. 37). The KIE over the a/c-Cu catalyst is ~1.2, much lower than that over the a/c-Cu-H₂ catalyst (Fig. 3e), further highlighting the facilitated water disassociation³⁴.

Therefore, we have experimentally shown that the dual-phase a/c-Cu catalyst with rich amorphous Cu domains also plays dual roles in NO₃⁻RR to NH₃, namely, improving the adsorption of N-containing intermediates and facilitating the formation of active hydrogen species from water dissociation.

Density functional theory (DFT) calculations

The dual roles of the amorphous Cu domains present in the a/c-Cu catalyst were further validated by DFT calculations. As Cu(100) was the main facet in the a/c-Cu catalyst (Supplementary Fig. 28), the amorphous/crystalline dual-phase Cu structure was modelled by ab initio molecular dynamics (AIMD) simulations using a slab with four Cu(100) layers, of which the top two layers were relaxed and the bottom two layers were fixed⁶². For the AIMD simulations, a temperature of 800 K was employed to accelerate structural sampling, facilitating the generation of an amorphous Cu surface. The free energy profiles of NO₃⁻RR over Cu(100) and the amorphous phase of the a/c-Cu model were first studied, as shown in Fig. 4a and Supplementary Figs. 38 and 39. Firstly, *NO₃ adsorbs on the Cu(100) and a/c-Cu surfaces with energies of −0.08 and −0.01 eV, respectively. On the Cu(100) surface, the subsequent *NO₃ hydrogenation to *NO₃H undergoes an uphill with an energy of 0.10 eV and the potential-determining step (PDS) is *NO hydrogenation to *NHO with an energy of 0.35 eV (Fig. 4a). In contrast, the a/c-Cu surface significantly improves *NO₃ hydrogenation with a distinct downhill trend. In this case, the PDS shifts to the formation of *NO₂ from *NO₃H with a lower energy of 0.27 eV.

Fig. 4. DFT calculations.

a Reaction pathways of NO₃⁻RR on Cu(100) and a/c-Cu. The inserts are atomic models for Cu(100) and a/c-Cu. b Projected density of states of *NO₃ adsorbed on Cu(100) and a/c-Cu. The asterisks represent the sum of other d-orbitals. c Charge density difference between *NO and Cu(100) (c₁ and c₂), *NO and a/c-Cu (c₃ and c₄). Isosurface = 2 × 10⁻³ e/ų. d Reaction pathway for water dissociation on Cu(100) and a/c-Cu. Source data are provided as a Source Data file.

As the adsorption and activation of NO₃⁻ is the initial reaction step^63,64, we further estimated the projected density of states (PDOS) of *NO₃ (Fig. 4b). For *NO₃ adsorbed on Cu(100), the p_y orbital of O and the d_x2-y2 of Cu form a hybrid orbital in the range from −3.89 to −5.21 eV below the Fermi level (The detailed splitting d-orbitals of Cu and p-orbitals of O are displayed in Supplementary Fig. 40). For a/c-Cu, the p_x+p_y orbital of O and the d_x2-y2 + d_xz of Cu form a hybrid orbital in the range from −3.30 to −3.65 eV below the Fermi level (Supplementary Fig. 41). The hybridization region on a/c-Cu is closer to the Fermi level than that on Cu(100), which rationalizes the more favorable *NO₃ hydrogenation. The *NO hydrogenation process has been generally considered as the PDS in an alkaline environment^63,65,66, but in this work the PDS shifts from *NO hydrogenation on Cu(100) to the formation of *NO₂ on a/c-Cu. Thus, we estimated the charge density difference for the *NO intermediate. As shown in Fig. 4c, there is obvious charge transfer between *NO and Cu atoms on both Cu(100) (Fig. 4c₁, c₂) and a/c-Cu (Fig. 4c₃, c₄). However, the charge density is substantially discrete for *NO adsorbed on a/c-Cu. This is largely related to the disordered arrangement of the amorphous phase of a/c-Cu, which leads to drastic d-orbital electron splitting and facilitates electron transfer to *NO. Thus, the *NO hydrogenation to *NHO is remarkably improved and becomes an exothermic process on a/c-Cu.

The *H formation from water dissociation was further investigated on Cu(100) and a/c-Cu. As shown in Fig. 4d and Supplementary Fig. 42, the dissociation of H–OH on Cu(100) requires an energy input of 0.34 eV, much more difficult than the downhill process on a/c-Cu, consistent with the experimentally KIE results (Fig. 3e)³⁸. Overall, the above theoretical investigations further demonstrate that the amorphous Cu structure present in the amorphous/crystalline dual-phase Cu catalyst promotes NO₃⁻RR to NH₃ by optimizing the adsorption of N-containing intermediates and facilitating water dissociation.

Discussion

In summary, we have developed an a/c-Cu catalyst with stable amorphous/crystalline dual-phase structure, by annealing a commercially available Cu foam. The a/c-Cu catalyst exhibits impressive NO₃⁻RR performance in an alkaline MEA electrolyzer, with a NH₃ partial current density of 3.33 ± 0.005 A cm⁻² and a NH₃ formation rate of 15.5 ± 0.02 mmol h⁻¹ cm⁻² at a low cell voltage of 2.6 ± 0.01 V. Moreover, the NH₃ Faradaic efficiency maintains around 90% at an applied current density of 1.5 A cm⁻² for 300 h. The scale-up demonstration with an electrode size of 100 cm² achieves a maximum NH₃ formation rate up to 11.9 ± 0.5 g h⁻¹ and an energy consumption of 108.3 ± 0.47 kJ g⁻¹_NH3 at an applied current of 160 A. The high NH₃ formation rate is ascribed to the amorphous Cu domains present in the dual-phase a/c-Cu catalyst, which promote NO₃⁻RR by optimizing the adsorption of N-containing intermediates and facilitating the formation of active hydrogen species from water dissociation. This work highlights the importance of stabilizing metastable amorphous structures for improving electrocatalytic reactivity and long-term stability.

Methods

Chemicals and materials

Cu foams (>99.9%) were purchased from Keshenghe Suzhou. Copper(II) chloride dihydrate (CuCl₂·2H₂O, >99%) was purchased from Tianjin Kemiou Chemical Reagent Co., Ltd. Sodium hydroxide (NaOH, >96%), potassium nitrate (KNO₃, >99%), ammonium chloride (NH₄Cl, >99.8%), trisodium citrate dihydrate (>99%), phosphoric acid (H₃PO₄, >85%), hydrochloric acid (HCl, >36%) and ethylene glycol (>99%) were purchased from Sinopharm Chemical Reagent Co., Ltd. Potassium hydroxide (KOH, >95%), Nafion ionomer (5%), salicylic acid (>99.5%), sulfonamide (>99.8%), lead(II) perchlorate hydrate (Pb(ClO₄)₂·3H₂O, >95%) and tannic acid (>99%) were purchased from Macklin. 5,5-dimethyl-1-pyrroline-N-oxide (DMPO, >97%) and potassium nitrite (KNO₂, >97%) were purchased from Aladdin. N-(1-naphthyl) ethylenediamine dihydrochloride (>98%) was purchased from Sigma-Aldrich (Shanghai) Trading Co., Ltd. Sodium nitroferricyanide (C₅FeN₆Na₂O, >99%) was purchased from Shanghai Chemical Reagent Co. Ir black catalyst was purchased from Johnson Matthey Corp. Ultrapure water (18.2 MΩ) was used in all experiments. All the chemicals were used without further purification.

Catalyst preparation

A commercial Cu foam with a geometric area of 1 cm² and a thickness of 0.2 mm (roughly 30 mg) was firstly annealed in air at 600 °C for 3 h. Subsequently, the annealed Cu foam was in situ electrochemically reduced during NO₃⁻RR to form the a/c-Cu catalyst. The a/c-Cu-H₂ catalyst was prepared by treating the above air-annealed Cu foam in 5% H₂/Ar (50 mL min⁻¹) at 350 °C for 3 h. Control samples (a/c-Cu-300 and a/c-Cu-900) were also prepared with the same method, but the annealing temperatures were 300 and 900 °C, respectively. An amorphous Cu catalyst was prepared with a method from previous literature⁶⁷. A homogenous solution was formed by dissolving 0.2 g of CuCl₂·2H₂O in 40 mL of ethylene glycol and stirring for 30 min. After adding 160 mg of tannic acid, the mixture was stirred for another 30 min. Then, 2 mL of 1.0 M NaOH was gradually added the above solution, followed by stirring for another 10 min. The amorphous Cu catalyst was washed three times with water and acetone, respectively, and then dried in vacuum overnight.

Materials characterizations

XRD patterns were obtained using a PANalytical X’pert PPR diffractometer equipped with a Cu Kα radiation source (λ = 1.5418 Å). SEM images were captured using a JSM-7900F Field-Emission SEM. HRTEM images were obtained using a JEM-2100 with an accelerating voltage of 200 kV. HAADF-STEM images were acquired with a JEM-ARM200F (JEOL, Japan). The ratio of amorphous domains was quantified from HRTEM images using Gatan Microscopy Suite (GMS-3) software and ImageJ software, and was defined as the total amorphous area divided by the overall catalyst area in each image. Error bars were made based on ten independent measurements.

Operando XAS measurements were carried out at the BL11B, BL14W1, and BL20U beamlines of the Shanghai Synchrotron Radiation Facility (SSRF). Energy calibration was performed using the absorption edge of a Cu foil as a reference. A home-made flow cell was used for XAS measurements in fluorescence mode (Supplementary Fig. 19), with a catalyst-coated gas diffusion electrode as working electrode. The anolyte and catholyte were 1 M KOH solution (5 mL min⁻¹) and 1 M KOH + 0.2 M KNO₃ solution (5 mL min⁻¹), respectively. The XAS data was analyzed using the ATHENA and ARTEMIS software.

In situ Raman spectroscopy measurements were carried out using a Renishaw inVia Raman microscope with a 785 nm near-infrared laser (Supplementary Fig. 34). The laser power was set to 1% and the exposure time was 50 s. A home-made flow cell with three electrodes was used for the Raman characterization. The cell was equipped with catalyst-coated gas diffusion electrode as the working electrode, a Pt wire as counter electrode, and an Ag/AgCl reference electrode. The cell had a quartz optical window, approximately 6 mm away from the working electrode, and the space in between was filled with catholyte. The anolyte and catholyte were 1 M KOH solution (5 mL min⁻¹) and 1 M KOH + 0.2 M KNO₃ solution (5 mL min⁻¹), respectively. The spectra were collected at different potentials after electrolysis for at least 2 min.

In situ ATR-SEIRAS measurement was conducted using an INVENIO S FTIR spectrometer with MCT detector (Supplementary Fig. 35). A home-made H cell was used for ATR-SEIRAS measurement, with Pt wire as a counter electrode and Ag/AgCl electrode as a reference electrode. For working electrode, an Au film was sputtered onto a silicon prism, followed by drop-casting a catalyst ink onto the Au film. All spectra were presented as relative change in absorbance, referenced to background spectra collected at OCP.

Quasi in situ XPS and Auger electron spectroscopy (AES) measurements were conducted using a Thermo Scientific ESCALAB 250Xi spectrometer with an Al Kα X-ray source. The NO₃⁻RR measurements were carried out in a MEA electrolyzer inside a glovebox, maintaining an O₂ concentration below 0.01 ppm. After electrolysis for at least 20 min, the electrodes were transferred into the XPS analysis chamber using a mobile transfer chamber, without air exposure during the transfer process.

EPR spectroscopy measurements were conducted using a Bruker A200 spectrometer. A DMPO solution (10 mg mL⁻¹) as a radical trapping reagent was added to the catholyte. After NO₃⁻RR in a MEA electrolyzer, the catholyte was collected, rapidly frozen in liquid nitrogen to prevent degradation, and then thawed for EPR analysis.

DSC measurements was carried out using NETZSCH DSC300 Select with a heat rate of 5 °C min⁻¹ under N₂ atmosphere. Both the a/c-Cu and a/c-Cu-H₂ powder catalysts were used for measurements.

The hydroxide (OH^–) electrosorption experiments were performed through cyclic voltammetry (CV) measurements in an H-cell, using Ar-saturated 1 M KOH aqueous solution. The potential window was set from –0.8 to –0.523 V vs. Ag/AgCl, with a scan rate of 10 mV s^–1. Before CV measurements, the electrode was firstly pre-reduced under NO₃⁻RR conditions by applying a potential of –2.0 V vs. Ag/AgCl for 15 min.

The Pb UPD measurements for characterizing low-coordinated Cu sites were referred to in the literature⁴¹. Following NO₃⁻RR at an applied potential of –2.0 V vs. Ag/AgCl for 15 min, the catalysts were measured by CV in a solution containing 0.1 M NaClO₄, 10 mM HClO₄, and 3 mM Pb(ClO₄)₂. The potential range was set from –0.4 to –0.1 V vs. Ag/AgCl, with a scan rate of 10 mV s^–1. The characteristic peak at –0.15 V vs. Ag/AgCl was identified as low-coordinated Cu sites.

The H/D KIE experiments were conducted using the linear sweep voltammetry (LSV) in H-cell at a scan rate of 10 mV s^–1, with iR correction. The electrode was firstly pre-reduced under NO₃⁻RR conditions at an applied potential of –2.0 V vs. Ag/AgCl for 15 min. Subsequently, LSV measurements was carried out in the electrolyte, which consisted of 0.2 M KNO₃ and 1.0 M KOH, with either H₂O or D₂O as the solvent. The KIE values were calculated based on the ratio of current density measured in H₂O to that measured in D₂O.

NO₃⁻RR measurements

NO₃⁻RR performance was assessed using an alkaline MEA electrolyzer at ambient temperature (20–25 °C)⁶⁸. The electrolyzer was assembled using a graphite flow field plate for catholyte and Pt-coated titanium flow field plates for anolyte (Supplementary Fig. 2). A quaternary ammonia poly(N-methyl-piperidine-co-p-terphenyl) (QAPPT) membrane was used as an anion exchange membrane. The anode was prepared by coating an ink of QAPPT-impregnated Ir black catalyst (1.5 mg cm⁻²) onto a porous foam. The annealed Cu foam with a geometric area of 1 cm² was directly used as an integrated porous electrode, if not stated otherwise. For comparison, the annealed foam was also ground into powder and drop-casted onto a carbon paper to form a conventional gas diffusion electrode. The NO₃⁻RR measurements were performed in the galvanostatic mode using an Ivium electrochemical workstation. The anolyte and catholyte were 1 M KOH solution (5 mL min⁻¹) and 1 M KOH + 0.2 M KNO₃ solution (3 mL min⁻¹), respectively. In the scale-up experiments, an annealed Cu foam with a geometric area of 100 cm² was used and the catholyte was 1 M KOH + 0.5 M KNO₃ (50 mL min⁻¹). In all the performance measurements including 300-h stability test, all the electrolytes were not recirculated and fresh electrolytes were always used. The NO₃⁻RR measurements were performed in the galvanostatic mode using a Keysight N8940A autoranging system direct current (DC) power supply (0–80 V/0–170 A, 5000 W). The electrolysis durations at each applied current were 15 and 10 min for 1-cm² and 100-cm² electrolyzers, respectively. Both catholyte and anolyte were collected at each applied current for subsequent liquid product analysis.

Products analysis

Gas products (i.e., H₂ in this work) were quantified using an on-line gas chromatography (Agilent, GC 8860), equipped with a thermal conductivity detector (TCD). Liquid products (NO₂⁻ and NH₃) were analyzed by UV–vis spectrophotometer (Shimadzu, Uv−2700i).

For NH₃ quantification, the indophenol blue method was conducted. Specifically, 2 mL of 1 M NaOH solution containing 5.0 wt% salicylic acid and 5.0 wt% sodium citrate, followed by 1 mL of 0.05 M NaClO, and 0.2 mL of 1 wt% C₅FeN₆Na₂O (sodium nitroferricyanide), were successively introduced into 2 mL of the diluted electrolyte, kept in the dark for 2 h. The NH₃ concentration was determined based on the absorbance at 655 nm. Additionally, ¹H-NMR spectroscopy was also used to quantify the NH₃ (NH₄⁺) production using a JEOL JNM-ECZL400S NMR spectrometer. In a typical procedure, 0.5 mL of standard solution/electrolyte was adjusted to pH 2 by adding 0.6 mL of 1.0 M HCl. Next, 0.5 mL of the above solution was mixed with the 0.1 mL DMSO-d6 (containing 2.5 mg mL^–1 C₄H₄O₄ as an internal standard). The quantification of NH₄⁺ was calculated with the peak area ratio of NH₄⁺ versus C₄H₄O₄.

For NO₂^– quantification, the Griess method was employed. Specifically, after neutralizing 3 mL of the diluted electrolyte with 3 mL of 1.0 M HCl, 0.1 mL of Griess color reagent, containing 4 g of sulfonamide, 10 mL of H₃PO₄ (85%), and 0.2 g of N-(1-naphthyl) ethylenediamine dihydrochloride in 100 mL of water, was added. The mixture was then kept in the dark for 20 min. The NO₂^– concentration was determined based on the absorbance at 540 nm.

The Faradaic efficiency of product i is calculated as follows:

$${\mathrm{\epsilon }}_{\mathrm{Faradaic},\mathrm{i}}={\mathrm{Q}}_{\mathrm{i}}/{\mathrm{Q}}_{\mathrm{total}}\times 100=\left({\mathrm{N}}_{\mathrm{i}}\times {\mathrm{n}}_{\mathrm{i}}\times \mathrm{F}\right)/{\mathrm{Q}}_{\mathrm{total}}\times 100$$ 1

Where, ε_Faradaic,i: the Faradaic efficiency of product i, Q_total: the consumed charge, C; N_i: the amount of the product i, mol; n_i: the number of transferred charge in the product i; F: Faraday constant, 96485 C mol^–1.

The NH₃ partial current density is calculated as follows:

$${\mathrm{j}}_{\mathrm{NH}3}={\mathrm{j}}_{\mathrm{total}}\times {\mathrm{\epsilon }}_{\mathrm{Faradaic},\mathrm{NH}3}$$ 2

The full-cell energy efficiency (EE) of NH₃ is defined as follows:

$${\mathrm{EE}}_{\mathrm{NH}3}=\left(1.23-{{\mathrm{E}}^0}_{\mathrm{NH}3}\right)\times {\mathrm{\epsilon }}_{\mathrm{Faradaic},\mathrm{NH}3}/\mathrm{U}$$ 3

Where, EE_NH3: the full-cell energy efficiency (EE) of NH₃, %; 1.23: the equilibrium potential of oxygen evolution reaction, V vs. RHE; E⁰_NH3: the equilibrium potential of NO₃^– to NH₃, which is 0.69 V vs. RHE¹⁹; U: cell voltage, V.

The energy consumption for NH₃ production is defined as follows³³:

$$\mathrm{Energy\; consumption}=\mathrm{I}\times \mathrm{U}/\left({\mathrm{C}}_{\mathrm{NH}3}\times \mathrm{v}\times 17\right)$$ 4

Where, energy consumption: kJ g^–1_NH3; I: the applied current, A; C_NH3: the molar concentration of NH₃, mol L^–1; v: the flow rate of the catholyte, L s^–1; 17: the molecular mass of NH₃, g mol^–1.

Computational details

All simulations were carried out using the Vienna ab initio simulation package (VASP 6.2.0)⁶⁹. Electron-ion interactions were described by the projector-augmented wave (PAW) pseudopotentials with a cutoff energy of 450 eV⁷⁰. For electron exchange and correlation energies, the Perdew–Burke–Ernzerh of generalized gradient approximation (GGA-PBE) was represented⁷¹. DFT-D3 with Becke–Jonson (BJ) damping method was used to correct the long-range van der Waals interaction⁷². The exchange-correlation functional with a Gaussian smearing width term of 0.05 eV was used. The convergence criteria for force and electronic self-consistent iteration were set to 0.01 eV Å^–1 and 1 × 10^–5 eV, respectively. In addition, all calculations are spin-polarized. For all surface optimization, the sampling of the Brillouin zone was performed using a Monkhorst–Pack scheme of (2 × 2 × 1). For all DOS calculation, the sampling of the Brillouin zone was performed using a Monkhorst–Pack scheme of (5 × 5 × 1)⁷³. The vacuum layer of 15 Å were selected to prevent their periodic images between adjacent layers. The VASPKIT code was used for postprocessing computational data obtained from VASP⁷⁴.

For the adsorption of *NO₃, the energy of HNO₃ in the liquid phase is converted to *NO₃ proton by the change of entropy as a correction factor, as shown below:

The adsorption energy of NO₃⁻ (ΔG*_NO3) is approximately expressed as:

$${\mathrm{\Delta }\mathrm{G}}_{*\mathrm{NO}3}={\mathrm{G}}_{*\mathrm{NO}3}-{\mathrm{G}}_{*}-{\mathrm{G}}_{\mathrm{HNO}3}+0.5{\mathrm{G}}_{\mathrm{H}2}+{\mathrm{\Delta }\mathrm{G}}_{\mathrm{correct}}$$ 5

$${\mathrm{\Delta }\mathrm{G}}_{\mathrm{correct}}=-\mathrm{\Delta }{\mathrm{G}}_{\mathrm{S}1}-\mathrm{\Delta }{\mathrm{G}}_{\mathrm{S}2}$$ 6

where G_*NO3, G_*, G_HNO3, and G_H2 are the Gibbs free energy of NO₃ adsorbed on Cu substrates, Cu substrates, HNO₃, and H₂ molecules in the gas phase, respectively. According to CRC handbook of chemistry and physics⁷⁵, ΔG_S1 = − 0.075 eV and ΔG_S2 = − 0.317 eV. Therefore, ΔG_correct is set to 0.075 + 0.317 = 0.392 eV.

The Cu(100) model was constructed using a 4 × 4 × 1 supercell comprising four layers and a total of 112 Cu atoms. For the a/c-Cu model, a slab with four Cu(100) layers was generated, with the bottom two layers fixed and the top two layers relaxed during AIMD simulations⁶². A time step of 2 fs was used, initially under constant-pressure conditions as the temperature was ramped from 300 K to 800 K. Subsequently, a canonical ensemble (NVT) was employed with a Nose–Hoover thermostat set at 800 K to accelerate structural sampling and facilitate the formation of an amorphous Cu surface. The first 15 ps of AIMD served as the equilibration period, followed by a 5 ps production run for data collection and analysis.

Author contributions

Conceptualization: G.W. and D.G.; methodology: Y.W., S.W., Y.F., J.S., P.W., R.L., and D.G.; investigation: Y.W. and S.W.; visualization: Y.W., S.W., and D.G.; funding acquisition: G.W., D.G., and S.W.; supervision: G.W. and D.G.; writing—original draft: Y.W., S.W., and D.G.; writing—review and editing: D.G., G.W., and X.B.

Peer review

Peer review information

Nature Communications thanks Dong-Hee Lim and the other, anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data that support the findings of this study are available within the paper and the Supplementary Information. Other relevant data are available from the corresponding authors on reasonable request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-55889-9.