Ultralow overpotential nitrate reduction to ammonia via a three-step relay mechanism

Ammonia plays a substantial role in agriculture and the next generation of carbon-free energy supply. Electrocatalytic nitrate reduction to NH3 is attractive for nitrate removal and NH3 production under ambient conditions. However, the energy efficiency is limited by the high reaction overpotential. Here we propose a three-step relay mechanism composed of a spontaneous redox reaction, electrochemical reduction and electrocatalytic reduction to overcome this issue. RuxCoy alloys were designed and adopted as model catalysts. Ru15Co85 exhibits an onset potential of +0.4 V versus reversible hydrogen electrode, and an energy efficiency of 42 ± 2%. The high performance results in a low production cost of US$0.49 ± 0.02 per kilogram of ammonia. The high nitrate reduction performances on Ru15Fe85 and Ru15Ni85 also highlight the promising potential of the relay mechanism. Electrocatalytic nitrate reduction represents an opportunity to generate ammonia under ambient conditions, yet the efficiency has been limited by the large overpotential required. Here, a Ru–Co alloy demonstrates a three-step relay mechanism involving a spontaneous redox step that reduces the overpotential for the process.

Ammonia plays a substantial role in agriculture and the next generation of carbon-free energy supply. Electrocatalytic nitrate reduction to NH 3 is attractive for nitrate removal and NH 3 production under ambient conditions. However, the energy efficiency is limited by the high reaction overpotential. Here we propose a three-step relay mechanism composed of a spontaneous redox reaction, electrochemical reduction and electrocatalytic reduction to overcome this issue. Ru x Co y alloys were designed and adopted as model catalysts. Ru 15 Co 85 exhibits an onset potential of +0.4 V versus reversible hydrogen electrode, and an energy efficiency of 42 ± 2%. The high performance results in a low production cost of US$0.49 ± 0.02 per kilogram of ammonia. The high nitrate reduction performances on Ru 15 Fe 85 and Ru 15 Ni 85 also highlight the promising potential of the relay mechanism. Ammonia (NH 3 )-one of the most common industrial chemicals-is essential for the production of nitrogenous fertilizers and shows great promise as a next-generation hydrogen-rich fuel [1][2][3] . NH 3 , with an annual global market of 150 million metric tonnes, is presently produced by the reaction of fossil-fuel-derived hydrogen with nitrogen under high-temperature high-pressure conditions (the Haber-Bosch method). Alternative routes for sustainable NH 3 synthesis are desirable. Excessive nitrate from overfertilization and industrial sewage is currently discharged into surface water and underground aquifers, threatening human health 4 . The electrocatalytic nitrate reduction reaction (NO 3 − RR) is thus a viable approach to NH 3 production that has advantages for environmental restoration and energy economics, under ambient conditions. Moreover, the NO 3 − RR can be coupled with plasma/photocatalysis-driven nitrogen oxidation techniques to achieve conversion from air to NH 3 (refs. 5-8). Great progress has been made regarding the NO 3 − RR in terms of fundamental research and practical applications [9][10][11][12] . Very recently, an industry-relevant current density of 1.0 A cm −2 for NH 3 synthesis was reported, with an onset potential of +0.22 V versus reversible hydrogen electrode (RHE) 9 . Notably, the theoretical potential of the NO 3 − RR is +0.69 V versus RHE under alkaline conditions (pH 14) 13 , much higher than the reported optimal onset potential 14 . Large overpotentials will not only lead to energy loss, but also promote the competitive hydrogen evolution reaction (HER) and reduce the Faradaic efficiency (FE) of NH 3 , resulting in an unsatisfactory energy efficiency (EE) for the NO 3 − RR. The high overpotential is caused by the sluggish rate-determining step of nitrate electroreduction to nitrite 15 . The rational design and construction of advanced electrocatalysts to overcome this issue based on the existing reaction mechanism remains challenging.
In this Article, a series of Ru x Co y hollow nanododecahedrons (HNDs) are synthesized to carry out NO 3 − electroreduction to NH 3 . Ru 15 Co 85 HNDs exhibit the highest performance, with an onset potential of +0.4 V versus RHE. Under optimal conditions, NH 3 is generated at a rate as high as 3.2 ± 0.17 mol g cat −1 h −1 , with a corresponding FE of 97 ± 5%. The energy consumption and production cost per kilo of NH 3 are calculated as 16.33 ± 0.68 kWh and US$0.49 ± 0.02 at +0.3 V versus RHE. The high performance can be maintained under an industrially relevant current density of 1.0 A cm −2 and a wide nitrate concentration range (10-2,000 mM), demonstrating promising application potential. Furthermore, the combined results of in situ electrochemical isotope-labelled attenuated total reflection Fourier-transformed infrared spectroscopy (ATR-FTIR), X-ray absorption fine structure (XAFS),  Table 1). Moreover, the surface content analysis of Ru and Co based on X-ray photoelectron spectroscopy (XPS; Supplementary Figs. 6 and 7 and Supplementary Tables 1 and 2) shows similar results. The uniform distribution of high Ru content in Co originates from the confinement effect of the ZIF-67 porous structure, which leads to the Ru 3+ and Co 2+ ions being evenly stored 28 . Locally enlarged XPS analysis Fig. 2f,g) shows that the binding energies of Ru 0 and Co 0 in Ru x Co y HNDs exhibit positive (Δ max, Ru = 0.46 eV) and negative (Δ max, Co = −0.32 eV) shifts compared to pure Ru and Co, respectively, indicating electron transfer from Ru to Co 29 . This phenomenon was also confirmed by X-ray absorption fine structure (XAFS) analysis. The three-dimensional wavelet transforms of the K-edge extended XAFS results (Fig. 2h,i) exhibit a contoured feature with maximum intensities of metallic Ru and Co in Ru 15 Co 85 HNDs. The slight reverse shift of the absorption threshold of Ru and Co in Ru 15 Co 85 HNDs (Supplementary Raman spectroscopy, X-ray diffraction (XRD) and theoretical simulations reveal the function of the three-step relay mechanism in decreasing the overpotential. The as-proposed three-step relay mechanism can also be extended to Ru x Fe y and Ru x Ni y for efficient NO 3 − RR to NH 3 .

Catalyst design for the electrocatalytic nitrate reduction reaction
The sluggish rate-determining step in nitrate electroreduction to nitrite (NO 2 − ) leads to a high overpotential (Fig. 1a) 15 . Theoretically, Co can undergo a spontaneous redox reaction with NO 3 − to produce Co(OH) 2 and NO 2 − with a Gibbs free energy change of −303.01 kJ mol −1 (Fig. 1b, step 1) 16 . In this way, the rate-determining step from nitrate to nitrite can be easily completed. Subsequently, Co(OH) 2 and NO 2 − can be reduced into Co and NH 3 through electrochemical and electrocatalytic processes, respectively, with the participation of active hydrogen (Fig. 1b, steps 2 and 3). Thus, the production of active hydrogen species is crucial for this three-step relay mechanism (spontaneous redox between Co and NO 3 − , Co(OH) 2 electroreduction to Co, and electrocatalytic conversion of NO 2 − to NH 3 ). According to the classic scaling relation, the adsorption energy of a hydrogen atom can be regarded as the descriptor for active hydrogen formation 17 . Ru, Rh, Pd, Ir and Pt possess moderate adsorption energies for hydrogen atoms (Fig. 1c) [18][19][20][21] . Among them, the cost of Ru is the lowest. Moreover, hollow nanostructures have been suggested to be conducive to mass transfer and atomic utilization in the electrocatalytic process 22 .Thus, Ru x Co y alloy hollow nanostructures are promising catalyst candidates for nitrate-to-ammonia electroreduction.

Preparation and characterization of Ru x Co y HNDs
The Ru x Co y HND catalysts were prepared via a two-step chemical conversion method, as shown in Fig. 2a Step 1 Step 3 Step 2 The hydrogen reversible reaction was first used to calibrate the reference electrode ( Fig. 3a and Supplementary Fig. 9) 30 . As illustrated in the linear sweep voltammetry (LSV) curves in Fig. 3b and Supplementary   Fig. 10, the j of all catalysts shows notable increases after adding nitrate, indicating that the NO 3 − RR is proceeding over the Ru x Co y HNDs. With increasing Ru content, the j values first increase and then decrease. Ru 15 Co 85 HNDs exhibit the highest j and the lowest Tafel slope (Fig. 3c), implying the fastest electron-transfer frequency for the NO 3 − RR. Following chronoamperometry measurements ( Supplementary Fig. 11), the nitrite and NH 3 products were analysed and quantified by colouration   j This work Ref. 10 Ref. 13 Ref. 14 Ref. 34 Ref. 35 Ref   13 . Under +0 V versus RHE, NH 3 is generated at a high FE of 97 ± 5% with a corresponding yield rate of 3.21 ± 0.17 mol g cat −1 h −1 (ref. 14). The EE for NH 3 synthesis shows an optimal value (42 ± 2%) at +0.3 V versus RHE for Ru 15 Co 85 HNDs ( Fig. 3f and Supplementary  Fig. 18). For comparison, Co and Ru 15 Co 85 nanoparticles were prepared via a co-precipitation method. The results show that the HND structure can increase the number of active sites and promote the intrinsic activity of active sites for the NO (refs. 11,12), the cost of as-produced NH 3 is substantial. On the basis of the price of renewable electricity alone (US$0.03 kWh −1 ) 32 , the energy consumption and production cost per kilo of NH 3 over Ru 15 Co 85 HNDs are calculated as 16.33 ± 0.68 kWh and US$0.49 ± 0.02, lower than the current commercial price (approximately US$1.0-1.5) 33 . Note that this is a simple cost accounting based on electricity price, without considering capital costs and Ohmic losses. Taking the environmental benefit of nitrate contaminant removal into account, the reported Ru 15 Co 85 HNDs for the NO 3 − RR are very appealing. The key performance parameters are summarized and compared with recently reported reports in Fig. 3g and Supplementary Table 3 13,14,[34][35][36] . It is worth noting that the high performance was maintained well for 30 continuous cycles of chronoamperometry tests ( Supplementary Fig. 21). Indeed, after the chronoamperometry tests, the hollow nanododecahedron-like morphology, lattice spacing and Ru content were almost the same as those of the sample before the electrochemical test ( Supplementary  Fig. 22). These results indicate the high durability of Ru 15 Co 85 HNDs for the NO 3 − RR. To further highlight their industrial production potential, a series of experiments were performed. First, electrolytes with different NO 3 − concentrations (0.01-2 M) were studied, and the high performance was found to be maintained well over a wide range (Fig.  3h). A larger electrolytic cell equipped with a peristaltic pump to circulate the electrolyte was then used for the NO 3 − RR test (details are provided in the Methods and Supplementary Fig. 23a). The chronopotentiometry measurements operated over a wide range of current densities (50-1,000 mA cm −2 , Supplementary Fig. 23b and Fig. 3i). The highest NH 3 yield rate was 3.83 ± 0.08 mmol cm −2 h −1 at 1,000 mA cm −2 , and over 100 h of long-term stability was achieved at 200 mA cm −2 (Fig. 3j). These results demonstrate the promising application potential of Ru 15 Fig. 25), suggests the spontaneous formation of Co(OH) 2 (equation (1)). Meanwhile, the absence of Raman and XRD signals for Ru HNDs excludes their participation in the spontaneous redox process. Electrochemical in situ Raman spectra of Ru 15 Co 85 HNDs reveal the disappearance of the Co-O characteristic peak (688 cm −1 ) below 0 V versus RHE (Fig. 4b), which is more positive than the same feature for Co HNDs (−0.5 V versus RHE; Supplementary Figs. 26 and 27) 30 . These results demonstrate that redox-derived Co(OH) 2 species can be electroreduced in situ to metallic Co 0 (equation (2)), and the presence of Ru can promote the electrochemical reduction process. Electrochemical in situ XRD characterization confirmed the electroreduction of Co(OH) 2 into Co 0 (Fig. 4d). The first LSV curve of Ru 15 Co 85 HNDs after soaking in NO 3 − electrolyte shows that the Co(OH) 2 initial reduction potential starts from +0.4 V versus RHE (Fig. 4e), which corresponds to the onset potential of the NO 3 − RR over Ru 15 Co 85 HNDs (+0.4 V versus RHE; Fig. 3e and Supplementary Fig. 15b). Interestingly, this potential is more positive than the theoretical equilibrium potential of the Co 0 /Co(OH) 2 redox couple. One possible explanation is that the kinetic effect induced by the coupling of a chemical reaction to an electron-transfer process causes this potential shift. Under constant electroreduction conditions, the characteristic Raman peak of Co(OH) 2 (688 cm −1 ) intermittently appears and disappears with the addition of a small amount of nitrate and after the chronoamperometry reaction ( Fig. 4f and Supplementary Fig. 28), indicating that the redox reaction (equation (1)) and electroreduction reaction (equation (2)) are proceeding. To confirm this, an electrochemical in situ XANES analysis of the Ru 15 Co 85 HNDs (Fig. 4g) (1)) and the electroreduction of Co(OH) 2 to Co (equation (2)) proceed simultaneously during the NO 3 − RR, and a dynamic Co valence cycle is achieved: The nitrite electroreduction reaction (NO 2 − RR) to NH 3 (equation (3)) was then studied by means of a mixed-isotopic labelling experiment (Fig. 4i). When equimolar 14 Fig. 30). These results demonstrate the advantage of the aforementioned three-step relay mechanism in decreasing the overpotential for the NO 3 − RR (Fig. 1). Moreover, this mechanism can also be extended to Ru 15 Ni 85 and Ru 15 Fe 85 HNDs, which both exhibit high performance for nitrate reduction to NH 3 ( Supplementary Figs. 31 and 32).
The yield rate of NH 3 in an alkaline environment (pH ≥ 7) is insensitive to changes in the pH value, indicating a concerted proton-electron transfer (CPET) pathway for the hydrogenation process ( Fig. 5a and Supplementary Fig. 33) 38 . Ru 15 Co 85 HNDs and Ru HNDs possess similar underpotential H deposition zones, whereas Co HNDs show no hydrogen zone in the present test range (Supplementary Fig. 34). Moreover, electrochemical quasi in situ electron paramagnetic resonance (EPR) reveals that the introduction of Ru is conducive to the formation of hydrogen radicals (Fig. 5b), implying a positive effect of Ru on active hydrogen formation. We next used electrochemical isotope labelling in situ ATR-FTIR spectra (Fig. 5c,d and Supplementary Figs. 35 and 36) and online differential electrochemical mass spectrometry Article https://doi.org/10.1038/s41929-023-00951-2 (DEMS) (Fig. 5e) Fig. 36). Electrochemical online DEMS tests show the following m/z signals: NO (30), NH 3 (17), N 2 (28), HNO (31) and NH 2 OH (33) (Fig. 5e) 41 . On the basis of these results, the possible reaction pathways for the NO 2 − RR over Ru 15 Co 85 HNDs, including dissociative, distal-O associative, distal-N associative and alternating-N associative pathways, were deduced; these are summarized in Supplementary Fig. 37 and Supplementary Note 5.

Theoretical simulations
Assuming that the complete conversion from NO 3 − to NH 3 occurs exclusively through CPET pathways, a minimum applied potential (U) of −0.3 V versus RHE is required to generate NH 3 thermodynamically, as marked by the vertical dashed line in Fig. 6a Table 4). The transformation of *NO to *N is the potential-determining step. The relationship among the applied potential, CPET steps and Ru ratio is presented in Fig. 6b (surface models are displayed in Supplementary Fig. 39). For almost all the Ru x Co y samples (Supplementary Tables 5-8 and Supplementary Note 8), the transformation of *NO 3 to *N demands negative applied potentials, especially for the *NO/*N couple (where the area is dark red, with U between approximately −0.37 and −0.21 V versus RHE), which implies that nitrate reduction via the purely electrochemical pathway is most likely restricted thermodynamically by the breaking of N−O bonds, which demand negative applied potentials as low as −0.37 V versus RHE. However, these results contradict the experimental findings in Fig. 3e, where NH 3 is already generated at U > 0 V versus RHE. These findings indicate the existence of pathways other than the electrochemical method, especially for the N-O bond-breaking process. Figure 6c presents the as-proposed three-step relay pathway for NO 3 − (aq) reduction to NH 3 . The reduction of NO 3 − (aq) to NO 2 − (aq) occurs through a redox reaction with Co 0 , which is experimentally supported and matches well with the well-known findings that active metals reduce nitrate to nitrite in basic media 42 . The chemical cleavage of the N-O bond in *NO 2 and *NO is thermodynamically highly favoured, with reaction Gibbs energies of −2.0 eV and −2.6 eV, respectively. The corresponding activation energies on the typical Ru 15 Co 85 are 0.67 eV and 0.19 eV, respectively. Images of the various adsorbates and the related transition states (TSs) are shown in Fig. 6c. Interestingly, volcano-shaped curves are observed for the adsorption of *NH 2 and *NO on various surfaces (Fig. 6d), with Ru 7 Co 93 and Ru 15 Co 85 located near the summit regions, which coincides with the experimental findings ( Fig. 3a-f) that these two alloys show better activity than the others.

Discussion
In summary, we propose a three-step relay mechanism to decrease the reaction overpotential for the NO 3 − RR. A series of Ru x Co y HND catalysts have been designed and prepared. Ru 15 Co 85 HNDs exhibit optimal catalytic performance (onset potential of +0.4 V versus RHE; EE of 42 ± 2%; NH 3 cost of US$0.49 ± 0.02 kg −1 ). The electrochemical in situ ATR-FTIR, XAFS, Raman, XRD and DFT calculations reveal that the high NO 3 − RR performance originates from the enhanced three-step relay processes on Ru 15 Co 85 HNDs (spontaneous redox between Co and NO 3 − , Co(OH) 2 electroreduction to Co, and electrocatalytic conversion of NO 2 − to NH 3 ). The introduction of Ru can promote the electrocatalytic reduction of NO 2 − to NH 3 and the electroreduction of Co(OH) 2 to Co due to its excellent hydrogen supply capacity. Moreover, the efficient NO 3 − RR performance on Ru 15 Fe 85 and Ru 15 Ni 85 confirms the universality of the relay mechanism. This work provides a reaction pathway to improve the energy efficiency of the NO 3 − RR, and will inspire the design of efficient catalysts for other electrocatalytic processes.

Materials
All chemicals were analytical grade and used as received without further purification. Deionized water was used in all experiments.

Material characterization
Scanning electron microscopy images were taken with a Hitachi S-4800 scanning electron microscope. TEM images were obtained using a JEOL JEM-200 microscope. XRD patterns were recorded with a Bruker D8 Focus Diffraction System by using a Cu Kα source (λ = 0.154178 nm). XPS spectra were collected on a Thermo Fisher Scientific K-Alpha+

XRD analysis
The lattice spacing was calculated from the characteristic peaks of XRD based on the Bragg equation: where d represents the lattice spacing and θ is the angle between the incident ray and the crystal plane, corresponding to the characteristic peak angle of XRD. Notably, the abscissa of XRD (Fig. 2e) is 2θ. n is a natural number (here, we take n = 1) and λ is the wavelength of the X-ray (here, λ = 0.1544178 nm).
We calculated the content of Ru in the alloy by Vegard's law: x Ru + x Co = 1 (6) where d, d Ru and d Co represent the lattice spacings of the alloy, Ru and Co, respectively. Among them, d Ru and d Co can be obtained by the Bragg equation of our catalysts (pure Ru and Co samples). x Ru and x Co are the molar contents of Ru and Co in the alloy, respectively. The particle size can be calculated by the full-width at half-maximum (FWHM) of XRD based on the Debye-Scherrer equation: Article https://doi.org/10.1038/s41929-023-00951-2 where D represents the particle size, K is the Scherrer constant (here, we take K = 0.89), B represents the FWHM of the XRD characteristic peak, θ is the Bragg diffraction angle, and λ is the wavelength of the X-ray (here λ = 0.1544178 nm). For the calculation of FWHM and particle size, we used the relevant functions in 'Jade 6' software for integration and calculation.

Preparation of ZIF-67
Co(NO 3 ) 2 ·6H 2 O (30 mg) and 2-methylimidazole (0.5 g) were added into a mixed solution of 10 ml of H 2 O and 10 ml of methanol under continuous stirring. After 30 min at room temperature, ZIF-67 with a dodecahedral structure was obtained by centrifugation and then washed with water and methanol three times. Finally, the ZIF-67 powder was dispersed in 3 ml of methanol to prepare the ZIF-67 suspension.

Preparation of Ru x Co y HNDs
For Ru x Co y HNDs, set amounts of RuCl 3 and CoCl 2 (total 72 μmol) were dissolved in a mixed solution of 3 ml of H 2 O and 3 ml of methanol. The ZIF-67 suspension (0.8 ml) was transferred to the mixed solution under continuous stirring. After 2 h at 60 °C, Ru x Co y O z HNDs were obtained by centrifugation and then washed with water and methanol three times. After overnight drying, the powder samples were annealed at 360 °C for 2 h in 3% H 2 /Ar (hydrogen content, 3%) to produce Ru x Co y HNDs. The mole fractions of Ru and Co in Ru x Co y HNDs were adjusted by changing the amounts of RuCl 3 and CoCl 2 added, and established by an inductively coupled plasma test. x and y in Ru x Co y HNDs represent their molar fractions.
For Ru HNDs 28  For Co HNDs 44 , 13.2 mg CoCl 2 and 0.8 ml ZIF-67 suspension were added to 6 ml of methanol under continuous stirring. The above solution was transferred to a 25 ml Teflon-lined autoclave and maintained at 120 °C for 2 h. After cooling to room temperature naturally, Co x O y HNDs were obtained by centrifugation and then washed with water and methanol three times.

Preparation of Ru 15 Co 85 nanoparticles
RuCl 3 (20 mg) and Co(NO 3 ) 2 ·6H 2 O (159 mg) were added to 40 ml of deionized water. After full dissolution, the pH of the solution was rapidly adjusted to exceed 11 with 1 M KOH solution. After half an hour of water heating at 60 °C under continuous stirring, Ru 15 Co 85 O z nanoparticles were obtained by centrifugation and then washed with water and methanol three times. After overnight drying, the powder samples were annealed at 360 °C for 2 h in 3% H 2 /Ar (hydrogen content, 3%) to produce Ru 15 Co 85 nanoparticles.

RR measurements
Electrochemical NO 3 − RR measurements were performed with an Ivium-n-Stat electrochemical workstation (Ivium Technologies). A typical three-electrode H-cell was used, with a working electrode, an Ag/AgCl electrode (saturated KCl solution) as the reference electrode, and a carbon rod counter electrode in 0.1 M KOH or 0.1 M KNO 3 + 0.1 M KOH electrolyte; this was separated into a cathode cell (25 ml) and an anode cell (30 ml) by an anion membrane (Alkymer AE-115). For the catalytic potential, we did not use iR correction, except under special instructions. For the chronoamperometry test, carbon paper (0.5 × 0.5 cm 2 ) decorated with 0.15-mg catalysts was used as the working electrode. In a typical procedure, an evenly distributed catalyst suspension was prepared by ultrasonically mixing 20 mg of catalyst into 8 ml of H 2 O, 2 ml of isopropyl alcohol and 50 μl of Nafion. The 75-μl suspension covered the carbon paper surface (0.25 cm 2 ). The current density was normalized by the catalyst mass, the geometric area of the electrode, and the ECSA. All electrochemical data (except stability testing) were repeated more than three times, with error bars representing the standard deviation of the data. All potentials were calibrated to the RHE by the following equation: (8) where E Ag/AgCl represents the experimental applied potential. Notably, the correction term (0.0591 pH + φ reference ) is calibrated by a hydrogen reversible reaction (Fig. 3a, HER and HOR).
For the ECSA, we used the double-layer capacitance method in 0.1 M KOH solution in the non-Faradic potential range with different scan rates from 10 to 100 mV s −1 . The ECSA of the working electrodes was calculated according to the following equations: where I c represents the charging current at different scan rates, ν is the scan rate, C dl is the double-layer capacitance, and C s is the specific capacitance for a flat metallic surface, which is generally in the range 20-60 μF cm −2 (according to reports, we assume this is 40 μF cm −2 ) [45][46][47] .
The current-density conversion formula from mass-normalized to geometric area-normalized is given by 11) where j geo represents the current density normalized by the geometric area, and j mass is the current density normalized by mass. m is the mass of the supported catalyst (0.15 mg), and S is the carbon paper geometric area of the supported catalyst (0.25 cm −2 ). Linear voltammetry profile measurements were conducted under a flow of N 2 using a rotating disk electrode (RDE) deposited with the catalysts (40 μg) as the working electrode at a rotation rate of ~100-1,600 r.p.m. and a sweep rate of 2 mV s −1 . All polarization curves were subjected to 80% iR correction. Before this test, we conducted ten cycles of cyclic voltammetry measurements to clean the catalyst surface. Unless otherwise stated, all linear voltammetry curves were recorded after five prescans to achieve stabilization.
The NH 3 Faradaic efficiency was calculated according to 12) where Q represents the applied overall coulomb quantity (C), Q NH3 is the coulomb required to produce NH 3 , n is the electron-transfer number (for 1 mol NH 3 , it is 8), V is the volume of the catholyte of the cathode chamber (25 ml), c NH3 is the concentration of NH 3 produced, and F is the Faraday constant (96,485 C mol −1 ). The EE was defined as the ratio of fuel energy to applied electrical power, which was calculated by (13) where E θ NH3 represents the equilibrium potential of nitrate electroreduction to NH 3 (0.69 V versus RHE under alkaline conditions), E θ OER Article https://doi.org/10.1038/s41929-023-00951-2 is the equilibrium potential of the oxygen evolution reaction (OER) (1.23 V versus RHE), FE NH3 is the Faradaic efficiency for NH 3 , and E OER and E NH3 are the applied potentials (the overpotential of OER refers to the recently reported literature 48 ).
For the amplified NH 3 production process, we used chronopotentiometry to demonstrate the industrial application potential. The composition and type of electrolytic cell are the same as those of the previous chronoamperometry test, except that the volume of the electrolytic cell is 80 ml. For preparation of the electrode, carbon felt (0.5 × 0.5 cm 2 ) decorated with 0.6 mg of catalysts was used as the working electrode, and titanium mesh (1.5 × 2 cm 2 ) was used as the counter electrode. An Ag/AgCl electrode (saturated KCl solution) was used as the reference electrode. We also used a peristaltic pump to promote mass transfer, and the liquid circulation speed was 200 ml min −1 .

Electrochemical in situ Raman tests
The in situ Raman measurements were carried out by the aforementioned Raman microscope and electrochemical workstation. The cell was made of Teflon with a quartz window between the sample and the objective. The working electrode was immersed in electrolyte through the wall of the cell, and the electrode plane was kept perpendicular to the laser. A platinum wire and Ag/AgCl served as the counter and reference electrodes, respectively. LSV curves were obtained from 0.5 to −0.6 V versus RHE with a scan rate of 2 mV s −1 . Electrochemical intermittent in situ Raman measurements were carried out in 0.1 M KOH solution under +0.3 V versus RHE. After collecting the first Raman spectrum, we added 1 ml of 0.1 M KNO 3 solution to the electrolyte. After 1 min, we collected the second Raman spectrum. After 20 min, the nitrate was totally reduced, then the Raman spectrum was collected again. We repeated this cycle test four times.

Electrochemical in situ ATR-FTIR tests
Electrochemical in situ ATR-FTIR measurements were performed on a Linglu Instruments ECIR-II cell mounted on a Pike Veemax III ATR with a single bounce silicon crystal covered with an Au membrane in internal reflection mode. Spectra were recorded on a Thermo Nicolet Nexus 670 spectrometer. The electrolyte was degassed by bubbling N 2 for 30 min before the measurement. The single-bounce silicon crystal covered with a Au membrane was prepared by the following procedure.

Electrochemical online DEMS test
Electrolyte (0.1 M KOH + 0.1 M KNO 3 ) was flowed into a homemade electrochemical cell by a peristaltic pump. Glassy carbon electrodes coated with Ru 15 Co 85 HNDs catalyst, Pt wire and Ag/AgCl electrodes were used as the working electrode, counter electrode and reference electrode, respectively. The applied voltage (−0.2 V versus RHE) was applied alternately, with an interval of 2 min. After the electrochemical test was over and the mass signal had returned to baseline, the next cycle started, using the same conditions to avoid accidental error. After seven cycles, the experiment was ended.

Electrochemical in situ XAS test
Electrochemical in situ XAS measurements at the Co K-edge were carried out at the 1W1B beamline of the BSRF. The electrolytic cell was made in-house with Teflon containing 0.1 M KOH or 0.1 M KOH + 0.1 M KNO 3 electrolyte, with a graphite rod and Ag/AgCl electrode used as the counter and reference electrodes, respectively. Carbon paper (2 × 2 cm 2 ) loaded with an electrocatalyst (3 mg) was used as the working electrode, and the catalyst was concentrated in the centre of the carbon paper with an effective area of more than 1 cm 2 . The in situ XAS signal was collected in fluorescence mode for a chronoamperometry measurement at 0 V or −0.5 V versus RHE. No pretreatment was required before the electrochemical chronoamperometry test.

Electrochemical in situ XRD test
Electrochemical in situ XRD patterns were measured on a Rigaku Smart-lab9KW Diffraction System using a Cu Kα source (λ = 0.15406 nm). The electrolytic cell was made up of Teflon with Pt wire as the counter electrode and a Hg/HgO electrode as the reference electrode. Carbon paper (0.5 × 0.5 cm 2 ) loaded with catalyst (0.6 mg) was used as the working electrode, and the patterns were collected via chronoamperometry measurements under different potentials. Before collecting the data, we ran the chronoamperometry test under the applied potential for 5 min. The pattern was collected in the 2θ range from 20° to 70° under an applied potential from −0.4 V to 0.5 V versus RHE. Each diffraction pattern was collected for 5 min for statistical analysis.

Determination and quantitation of ammonia using UV-vis
We used the sodium salicylate method to detect the concentration of NH 3 after chronoamperometry measurements with different potentials. A certain amount of electrolyte was removed from the electrolytic cell and diluted to 5 ml to the detection range, then 0.5 ml of sodium salicylate solution (50 g l −1 sodium salicylate, 50 g l −1 potassium sodium tartrate and 20 g l −1 NaOH), 0.05 ml of sodium nitroprusside solution (10 g l −1 sodium nitroprusside) and 0.05 ml of sodium hypophosphite solution (40 ml l −1 , 13 wt% NaClO and 30 g l −1 NaOH) were added to the aforementioned solution. After 20 min, the absorption spectrum was tested using a UV-vis spectrophotometer, and the absorption intensities at a wavelength of 650 nm were recorded. The concentration−absorbance curve was calibrated using a series of standard ammonium chloride solutions (0, 0.02, 0.04, 0.06, 0.08, 0.1 mM), and the ammonium chloride crystal was dried at ~105-110 °C for 2 h in advance.

Determination and quantitation of ammonia using 1 H NMR
After the electrochemical chronoamperometry measurements, we first adjusted the pH of the electrolyte to neutrality (pH 7) and then added maleic acid (the content of the final mixed solution was 0.4 mg ml −1 ). Subsequently, 0.5 ml of mixed solution, 50 μl of H 2 SO 4 (4 M) and 50 μl of DMSO-d 6 were transferred to the NMR tube. The concentration-peak area curve from the NMR spectrum was calibrated using a series of standard ammonium chloride solutions (5, 10, 15, 20 and 25 mM), and the ammonium chloride crystal was dried at ~105-110 °C for 2 h in advance.

Determination and quantitation of nitrite using UV-vis
A certain amount of electrolyte was removed from the electrolytic cell and diluted to 5 ml to the detection range. Then, 0.1 ml of nitrite colour reagent was added to the aforementioned solution. After 20 min, the absorption spectrum was tested using a UV-vis spectrophotometer, and the absorption intensities at a wavelength of 540 nm were recorded. The nitrite colour reagent consisted of 40 g l −1 p-aminobenzenesulfonamide, 100 ml l −1 phosphoric acid and 2 g l −1 N-(1-naphthyl)-ethylenediamine dihydrochloride. The concentrationabsorbance curve was calibrated using a series of standard potassium nitrite solutions (0, 0.005, 0.01, 0.015, 0.02 and 0.03 mM), and the potassium nitrite crystal was dried at ~105-110 °C for 2 h in advance.

Computational methods
Vienna Ab initio Simulation Package (VASP) code was used to perform the DFT calculations, with the Perdew-Burke−Ernzerhof (PBE)