Exploiting underpotential deposited hydrogen enables energy-efficient nitrate electroreduction to ammonia

Abstract

Electrochemical nitrate reduction in alkaline media offers a sustainable route for ammonia synthesis at rates rivaling those of the Haber–Bosch process. However, its energy efficiency is limited by sluggish nitrate deoxidation and hydrogenation, compounded by challenges in proton supply via H₂O dissociation. Here, we develop an enzyme-like substrate transport channel through a hierarchical arrangement of metallic Ag and Ru nanophases, to enable cascade nitrate-to-ammonia conversion and optimize underpotential-deposited hydrogen utilization. Operando characterization and theoretical calculations reveal that Ag–Ru electronic synergy regulates underpotential-deposited hydrogen coverage at Ru-centered active sites by coupling facilitated H₂O dissociation with *OH-mediated site regeneration, thereby promoting nitrite relay conversion. Our underpotential-deposited hydrogen-assisted nitrate electroreduction system delivers a half-cell ammonia energy efficiency of 53.7% at 0.2 V versus RHE with near-unity Faradaic efficiency across a wide nitrate concentration range, and an ammonia partial current density of 2.2 A cm⁻² at 0 V versus RHE. Pairing cathodic nitrate reduction with anodic H₂ oxidation enables ammonia production costs below $1.15 kg⁻¹ while maintaining sustained energy efficiency over 100 h at 200 mA cm⁻².

Electrochemical nitrate reduction rivals Haber-Bosch ammonia synthesis rates but suffers from poor energy efficiency. Here, the authors show that underpotential deposited hydrogen can be unlocked to accelerate nitrate conversion, offering an pathway to energy-efficient ammonia production.

Introduction

Ammonia (NH₃) is a crucial chemical in agriculture and industry, primarily used in fertilizers and as a precursor to renewable energy vectors and various industrial materials^1,2. The conventional Haber–Bosch (H–B) process for NH₃ synthesis relies heavily on fossil fuels and generates substantial carbon emissions (1.6–2.9 t-CO₂/t-NH₃)³, highlighting the urgent need for more sustainable technologies. The electrochemical reduction of nitrogen oxides (NO_xRR), including NO₃⁻, NO₂⁻, and NO, powered by renewable electricity, presents a compelling alternative for recycling nitrogen (N) contaminants into valuable NH₃^4–7. Current NO_xRR technologies realize high NH₃ yield rates (Y_NH3) (0.1–2 A cm⁻²)^8–12, advancing Y_NH3 towards the realm of industrial viability. The remaining challenge is to achieve such productivities with viable energy efficiency (EE) and cost effectiveness^13–16. Specifically, the EE of the H–B process is approximately 60%^17,18, whereas commercial NH₃ is priced at about $1.15 per kg¹⁹.

Alkaline electrolytes can improve the selectivity of NO_x-to-NH₃ conversion compared to neutral media, while avoiding the complex self-catalytic pathways of NO_x in strongly acidic conditions^20–22. As the predominant form of NO_x species in wastewater, NO₃⁻ ions have been intensively investigated as a nitrogen source for NH₃ production^6,22,23. However, alkaline NO₃RR encounters kinetic challenges due to the initially sluggish deoxygenation of NO₃⁻ via bidentate O adsorption onto catalyst surfaces, followed by hydrogenation of multiple intermediates (e.g., NO) in a single-dentate N-coordination mode²⁴. These elementary steps require distinct catalyst surface properties difficult to be integrated in a single-component catalyst. Recent strategies have therefore sought to decouple the NO₃RR into two steps, NO₃⁻-to-NO₂⁻ and NO₂⁻-to-NH₃, through tandem operation of two distinct active components^13,15,25,26, thereby improving kinetics at low overpotentials. Despite these advancements, proton supply in alkaline media remains hindered by the high energy barriers for H₂O dissociation^27–29, limiting the half-cell EE_NH3 to <42%^13–15,26 and keeping the NH₃ cost to >$1.15 per kg¹⁹ at industrially relevant Y_NH3 (>100 mA cm⁻²).

In an alkaline aqueous solution, the hydrogenation of NO₃RR intermediates can proceed via two potential mechanisms: the Eley–Rideal (E–R) mechanism, involving direct interaction with H₂O, and the Langmuir–Hinshelwood (L–H) mechanism, which requires surface-active hydrogen (*H)^30–32. Mechanistic studies indicate that the L–H pathway tends to dominate in alkaline NO₃RR because steric hindrance arising from cation coordination with oxygen-containing intermediates (e.g., NO) disfavors the E-R route³¹. Efforts to optimize *H generation and spillover focus on tuning catalytic centers or integrating *H-storage materials, such as Co, Ni, Ti, and WO_x-based catalysts^11,33–39. Nevertheless, most reported *H species originate from H₂O dissociation at <0 V versus the reversible hydrogen electrode (RHE) (Fig. 1a), where the overpotential deposited *H (*H_OPD) readily drives hydrogen evolution reaction (HER)^40,41, especially at low NO₃⁻ concentrations, thereby introducing additional energy losses.

Fig. 1. Schematic illustration of *H_UPD-assisted NO₃RR mechanisms.

a Comparative analysis of the peak FE_NH3 and the applied potentials for the NO₃RR on recently reported transition metal-based catalysts at varying NO₃⁻ concentrations (corresponding to the area of the circles). This analysis highlights the goals for an energy-efficient NO₃RR at > 0 V. With low overpotential NO₃RR (at >0 V), coupling cathodic NO₃RR with anodic hydrogen oxidation reaction (HOR) can yield a voltage difference (ΔE) of less than 0 V, enabling energy output. Conversely, a high NO₃RR overpotential demands external energy input. b Mechanism of H_UPD-assisted NO₃RR on the AgRu-15 catalyst, featuring an enzyme-like cascade substrate transport channel. Color code: nitrogen (N) in blue, oxygen (O) in red, and hydrogen (H) in pink. Source data are provided as a Source data file.

In pursuit of a breakthrough in EE, exploiting underpotential deposited *H (*H_UPD) (at >0 V vs. RHE) as a proton source could provide a promising solution (Fig. 1a). The *H_UPD region, typically observed between 0 and 0.35 V (vs. RHE) on platinum-group metals (PGMs, e.g., Pt and Ru)^28,42,43, can eliminate the HER. However, the strong binding of *H_UPD will hinder the competitive adsorption of NO₃⁻ and NO₂^−41,42,44, thereby diminishing NO₃RR activity. Recent studies have shown that the binding strengths of *H_UPD on PGMs can be successfully modulated through alloying and hybridization strategies^27,43. Despite the abundant source of *H_UPD through H₂O dissociation on PGMs under alkaline conditions^28,45, however, balancing the competitive adsorption and conversion rates of *H_UPD, NO₃⁻, and NO₂⁻ at the catalyst surface remains highly challenging. Therefore, successful implementation of *H_UPD-assisted NO₃RR requires precise regulation of catalyst electronic states and an optimal spatial arrangement of active components to coordinate three key elementary steps: NO₃⁻-to-NO₂⁻, NO₂⁻-to-NH₃, and H₂O dissociation to *H_UPD. However, such an integrated catalytic system has not yet been reported.

In this work, we explored a dual-component Ag–Ru hybrid catalyst for *H_UPD-assisted conversion of NO₃⁻ to NH₃. This catalyst features metallic Ru nanocoatings distributed in a gradient on a hierarchical porous Ag matrix, creating a three-dimensional (3D) architecture that mimics enzyme substrate-transport channel with distinct catalytic sites in nanoconfined volumes (Fig. 1b). This architecture enables preferential adsorption and reduction of NO₃⁻ to NO₂⁻ within Ag nanocavities, followed by hydrogenation to NH₃ on Ru surfaces using shuttled *H_UPD. The hybrid catalyst achieves a maximum half-cell EE_NH3 of 53.7% at 0.2 V (vs. RHE) and maintains an EE_NH3 of 47% at an NH₃ current density above 140 mA cm⁻², exceeding previous studies. Techno-economic evaluation in a NO₃RR flow cell, paired with anodic H₂ oxidation, further indicates NH₃ production costs below $1.15 kg⁻¹ across current densities of 100–920 mA cm⁻². Operando characterization and density functional theory (DFT) calculations reveal that Ag–Ru electronic synergy upshifts the d-band center of Ru-centered active sites, thereby accelerating H₂O dissociation kinetics and strengthening *OH adsorption. The former ensures a sustained *H_UPD supply, whereas the latter prevents *H_UPD overaccumulation by consuming *H_UPD to form H₂O, leading to continuous regeneration of Ru-centered active sites. This dynamically balanced *H_UPD coverage underpins competitive NO₃⁻ and NO₂⁻ adsorption, ultimately improving *H_UPD utilization during NO₃RR.

Results

Synthesis and characterization of catalysts

To pursue a high EE of NO₃⁻-to-NH₃ conversion, we developed a porous Ag/Ru-based catalyst featuring a hierarchically spatial distribution of core-shell Ag/Ru metallic phases. This design was intended to emulate the nanoconfined substrate-transport channel of enzymatic systems^46,47, incorporating distinct active sites for cascade catalysis (Fig. 1b). The nanoporous silver (p-Ag) electrode was synthesized via a scalable electrochemical anodization–reduction process, employing Cl⁻ anions as structure-directing agents (Supplementary Figs. S1 and S2). Typical scanning electron microscopy (SEM) images in Fig. 2a, Supplementary Figs. S3 and S4 reveal a coral-like morphology of p-Ag, characterized by interconnected nanoparticles (~100 nm) and hierarchical porosity, which results in a 37.5-fold increase in electrochemically active surface area (ECSA) compared to a smooth Ag matrix (Supplementary Fig. S5 and Table S1). The roughened surface topology of p-Ag is expected to expose an increased fraction of undercoordinated sites, as supported by X-ray diffraction (XRD) and cyclic voltammetry (CV) data, which demonstrate an enhanced contributions from the stable Ag(111) facet and an increased ratio of stepped Ag(220) to flat Ag(100) relative to bulk Ag foil (Supplementary Figs. S6 and S7).

Fig. 2. Structural characterization.

SEM images of p-Ag (a) and AgRu-15 (b). c A TEM image and the selected area electron diffraction (SAED) of AgRu-15. d Aberration-corrected scanning transmission electron microscopy image (STEM) image of AgRu-15. e High-angle annular dark-field (HAADF) STEM and corresponding high-resolution energy-dispersive X-ray spectroscopy (EDX) images over the AgRu-15. Regions 1 and 2 in the EDX maps correspond to zones 1 and 2 in (d), respectively. f EDX line-scan profiles acquired along the orange arrow in the HAADF image of AgRu-15. The y‑axis represents the X‑ray counts for each element. g Cross-section SEM images of AgRu-15 with EDX element mapping. h The corresponding EDX line-scan profiles of a selected area marked by the white arrow in (g). i XPS spectra analysis of Ru 3p for AgRu-15 and Ru/C (Ru supported on carbon paper), as well as Ag 3d for p-Ag and AgRu-15. j The Fourier transforms R-space of Ru k-edge EXAFS spectra and Ag k-edge EXAFS spectra for AgRu-15 and reference samples. k–n Wavelet transforms (WT) for the k²-weighted EXAFS of AgRu-15 and reference samples (Ag foil and Ru foil). In j–n R and α represent uncorrected distance and phase shift, respectively. Source data are provided as a Source data file.

Ruthenium (Ru) catalytic sites were incorporated into the nanoconfined cavities of p-Ag through a combination of galvanic replacement and chronopotentiometric deposition techniques (Supplementary Fig. S1). The galvanic replacement occurred in a 1 M HCl solution, driven by the potential difference between the Ru³⁺/Ru⁰ and AgCl/Ag⁰ redox couples. Based on the subsequent chronopotentiometric deposition time, the optimized hybrid catalyst was denoted AgRu-15, corresponding to a 15-min deposition duration (Supplementary Figs. S8 and S9). As illustrated in Fig. 2b, AgRu-15 maintained the island-like morphology of p-Ag, while the coral-like nanopores were filled owing to volume expansion from Ru incorporation. Transmission electron microscopy (TEM) images of AgRu-15 displayed a hollow structure with an ultrathin wall of approximately 6 nm (Fig. 2c and Supplementary Fig. S10), characteristic of galvanic replacement reactions. Selected area electron diffraction (SAED) patterns confirmed the coexistence of face-centered cubic Ag and hexagonal close-packed Ru with a preferential Ru (002) orientation, consistent with the X-ray diffraction (XRD) results (Supplementary Fig. S11). As a control, pure Ru nanoparticles were deposited on a titanium (Ti) rod under analogous conditions, denoted Ru@Ti (Supplementary Fig. S12).

By integrating aberration-corrected scanning TEM (STEM) with high-resolution energy-dispersive X-ray spectroscopy (EDS), we identified a significant segregation of Ag and Ru nanophases within the ultrathin walls of AgRu-15 (Fig. 2d, e, and Supplementary Fig. S13). The EDS mapping-indicated Ru phase (region 1) exhibited a lattice fringe of 0.214 nm, corresponding to the Ru (002) facet. Notably, the surface lattice fringes of the EDS mapping-indicated Ag-rich phase (region 2) were consistent with those of the Ru (002) facet (0.214 nm). This observation is attributed to the galvanic replacement-induced deposition of Ru nanocoatings over immiscible Ag phases⁴⁸, leading to the formation of metallic Ag/Ru core-shell nanostructures, as confirmed by linear EDS scan profiles (Fig. 2f). Cross-sectional SEM analysis systematically examined the depth distribution profiles of Ag and Ru elements in the AgRu-15, revealing a thickness of approximately 7 μm and confirming the hollow architecture of the hybrid Ag/Ru nanoparticles in a magnified view (Fig. 2g). EDS imaging and line scan profiles further elucidated the Ru-rich surface and the gradient distribution of Ru within the p-Ag substrate of the AgRu-15 catalyst (Fig. 2h). Specifically, the Ag/Ru atomic ratio decreased from 3.36 in the inner layer (5 μm below the surface) to 1.20 at the outer surface (Supplementary Fig. S14).

The surface composition and chemical states of AgRu-15 were analyzed via the deconvolution of X-ray photoelectron spectra (XPS) for Ru 3p and Ag 3d. As illustrated in Fig. 2i, the Ru 3p_3/2 peak at 461.73 eV for metallic Ru⁰ exhibited a ~0.25 eV shift to lower binding energy compared to pure Ru (461.98 eV), indicating electron capture by Ru⁰ from other components of AgRu-15. Additionally, the fitted Ag⁰ peak at 367.9 eV in the Ag 3d_3/2 region shows a ~0.1 eV negative shift relative to Ag foil and p-Ag (368.0 eV) (Fig. 2i and Supplementary Fig. S2). This shift is attributed to Ru-induced charge mixing in Ag, enhancing charge density in the Ag d band while reducing it in the sp band⁴⁹, thereby facilitating electron transfer from Ag to Ru. The XPS-determined atomic ratio of Ag/Ru for the AgRu-15 catalyst is 1.02 (Supplementary Fig. S15), which is in close agreement with the EDS measurement of 1.20 at the outer surface (Supplementary Fig. S14 and Table S2), considering the shallow sampling depth inherent to the XPS technique.

The coordination environments of the Ag and Ru atoms in AgRu-15 were further investigated using extended X-ray absorption fine structure spectroscopy (EXAFS) (Supplementary Fig. S16). As illustrated in Fig. 2j, the Fourier-transformed Ru k²-weighted EXAFS profiles for AgRu-15 revealed a dominant peak at 2.37 Å, aligning with Ru foil and differing from RuO₂, indicative of metallic Ru phases. Fitting the dominant peak indicates a bond length of 2.672 ± 0.005 Å, equivalent to that of the Ru-Ru bond in Ru foil (2.672 ± 0.002 Å). Similarly, the Ag k²-weighted EXAFS profile of AgRu-15 shows a dominant peak at 2.63 Å, corresponding to an Ag–Ag bond length similar to that of Ag foil (Supplementary Tables S3 and S4). These findings indicate that AgRu-15 is comprised of isolated and immiscible metallic Ag and Ru phases rather than an AgRu alloy⁴⁸, further supported by the wavelet transforms of k²-weighted EXAFS data (Fig. 2k–n). Thus, the AgRu-15 catalyst exhibits metallic Ru and Ag active sites located in close proximity within a nanoconfined 3D architecture, which presents an enzyme-like substrate channel conducive to multistep conversion of NO₃⁻ to NH₃ (Fig. 1b).

NO₃RR activity and kinetics over AgRu-15

The electrocatalytic performance of various catalysts for NO₃RR was assessed using a three-electrode H-type cell at pH = 13.96 ± 0.02 (1 M KOH), with concentrations of NO₃⁻, NO₂⁻, and NH₃ monitored via ultraviolet-visible spectrophotometry (Supplementary Fig. S17). All applied potentials were calibrated against the reversible hydrogen electrode (RHE) (Supplementary Fig. S18). As depicted in Fig. 3a, the AgRu-15 catalyst exhibited significantly enhanced NO₃RR activity compared to p-Ag and Ru@Ti. Over the applied potentials from 0.5 to 0 V, the polarization current density of AgRu-15 increased steadily, reaching approximately four times that of p-Ag at potentials below 0.1 V. In contrast, Ru@Ti demonstrated minimal polarization current until the HER occurred below 0 V, indicating its weaker affinity for NO₃⁻. Notably, with an increased NO₃⁻ concentration (0.5 to 1 M), AgRu-15 broke through the concentration-limited current plateau, achieving a current density of 1 A cm⁻² above 0.1 V—a rarity in previous studies. This result highlights the presence of abundant catalytic sites within the nanoconfined 3D architecture of AgRu-15, which facilitates high-rate NO₃RR.

Fig. 3. Electrocatalytic performance.

LSV curves (j, current density) (a) and the corresponding Faradaic efficiency (FE) (b) for the NO₃RR on p-Ag, AgRu-15, and Ru@Ti catalysts in 1 M KOH + 10 mM KNO₃ (pH = 13.96 ± 0.02) at a scan rate of 10 mV s⁻¹. Electrolyte resistance: 1.88 ± 0.05 Ω. Inset in (a) shows LSV curves of AgRu-15 at different nitrate concentrations (50 mM to 1 M). Measurements in (b) were repeated at least three times, and the average FE values are presented with standard deviation error bars. c Onset potentials at a current density of −2 mA cm⁻² and Tafel slopes for NO₃RR and NO₂RR of the catalysts. d Pseudo-first-order kinetic profiles for NO₃RR and NO₂RR at 0.1 V vs. RHE. e H/D kinetic isotopic effect (KIE) for the catalysts in 1 M KOH + 10 mM KNO_3. f NH₃ yield rate (Y_NH3) and Faradaic efficiency (FE_NH3) of AgRu-15 across different nitrate concentrations at 0 V vs. RHE. g Dependence of NO₃⁻, NO₂⁻, and NH₃ concentrations on reaction time for AgRu-15 at 0.1 V vs. RHE. h Heatmaps of the half-cell EE for NH₃ synthesis by AgRu-15 at varying nitrate concentrations and applied potentials. i FE_NH3 and Y_NH3 of the AgRu-15 for NO₃RR in the simulated wastewater electrolyte and real wastewater electrolyte at 0.1 V vs. RHE. See Supplementary Fig. S32 and Table S6 for the detailed concentrations of interfering species. Error bars represent the standard deviation from at least three independent measurements. Source data are provided as a Source data file.

To identify the catalytic nature of the Ag and Ru phases in AgRu-15, we compared its Faradaic efficiencies (FEs) for NH₃ and NO₂⁻ with those of p-Ag and Ru@Ti (Fig. 3b and Supplementary Fig. S19). The AgRu-15 successfully triggered the NO₃RR at 0.3 V, achieving a cumulative FEs of 82.8 ± 8.2% for NH₃ and NO₂⁻ at 0.3 V, and recording a NH₃ yield rate (Y_NH3) of 2.36 ± 1.07 μmol cm⁻² h⁻¹ (Supplementary Fig. S19). As overpotentials increased, the AgRu-15 maintained a high selectivity for NH₃, achieving FEs of 87.9 ± 0.59%, 99.5 ± 1.32%, and 100 ± 0.39% at potentials of 0.2, 0.1, and 0 V, respectively. In contrast, p-Ag exclusively yielded NO₂⁻, while Ru@Ti exhibited a high selectivity for NH₃. Accordingly, the isolated Ag and Ru phases in AgRu-15, which are located in close proximity, may function similarly to p-Ag and Ru@Ti, respectively, and facilitate the conversion of NO₃⁻ to NH₃ via a cascade mechanism.

The kinetics of the AgRu-15 catalyst for NO₃RR were evaluated by comparing its onset potentials, reaction rate orders (ROs), and Tafel slopes with those of p-Ag and Ru@Ti in 10 mM NO₃⁻ and NO₂⁻ solutions, respectively. The AgRu-15 catalyst demonstrated a ~60 mV more positive onset potential for NO₃⁻ reduction than p-Ag and a similar enhancement (~140 mV) for NO₂⁻ reduction compared to Ru@Ti (Fig. 3c and Supplementary Fig. S20). This trend was further supported by the apparent ROs (Fig. 3d and Supplementary Fig. S21). These results, taken together, suggest synergistic effects between isolated Ag and Ru phases within AgRu-15, which enhance the kinetics of the sequential NO₃⁻-to-NO₂⁻ and NO₂⁻-to-NH₃ conversion. As shown in Fig. 3c and Supplementary Fig. S22, the Tafel slope for NO₃⁻ reduction on AgRu-15 was 106 mV dec⁻¹, situated between the values for p-Ag (65.0 mV dec⁻¹) and Ru@Ti (112 mV dec⁻¹), indicating the presence of distinct Ag and Ru-based sites for efficient NO₃⁻ adsorption. Conversely, even if situated between the values for p-Ag (198 mV dec⁻¹) and Ru@Ti (120 mV dec⁻¹), the Tafel slope of AgRu-15 for NO₂⁻ reduction was as high as 182.6 mV dec⁻¹, highlighting a rate-limiting step linked to initial NO₂⁻ adsorption. This result is mainly due to the contribution of the Ag sites; however, it contrasts starkly with the observed optimal onset potential and ROs of AgRu-15 for NO₂⁻ reduction (Fig. 3c, d), indicating that AgRu-15 harbors specific Ru-based active sites that outperform pure Ru sites within Ru@Ti catalysts for NO₂⁻ reduction.

Linear sweep voltammetry (LSV) measurements were conducted for the NO₃RR in H₂O and D₂O to investigate the proton transfer kinetics on AgRu-15, p-Ag, and Ru@Ti catalysts. As shown in Fig. 3e and Supplementary Fig. S23, the H/D kinetic isotope effect (KIE) for p-Ag reached a peak value of 1.9 at 0.26 V, pointing to a significant proton transfer limitation during the deoxygenation of NO₃⁻. Similarly, the Ru@Ti catalyst maintained KIE values consistently exceeding 1.5 across the entire NO₃RR-testing potential range, with a maximal KIE value of 2.66 recorded at 0.085 V. This finding indicates a substantial proton transfer limitation for the pure Ru sites throughout the NH₃ production. In contrast, the KIE value for AgRu-15 displayed a modest increase to 1.3 within the potential range of 0.25 to 0.17 V, subsequently stabilizing at 1.1 until the HER occurred below 0.05 V. These KIE values, which are close to 1.0, indicate that the Ag and Ru phases within AgRu-15 synergistically enhance proton transfer kinetics.

Based on the synergistic tandem catalysis mechanism, the AgRu-15 catalyst sustained high NH₃ FEs across a broad range of NO₃⁻ concentrations. Specifically, it achieved an NH₃ partial current density of 2.2 A cm⁻² in 1 M NO₃⁻ at 0 V, which corresponds to an NH₃ generation rate of 9.8 ± 0.08 mmol h⁻¹ cm⁻² (or 166.6 ± 1.36 mg h⁻¹ cm⁻²) (Fig. 3f and Supplementary Fig. S24). The ECSA of Ru@Ti and the Ru phases within the AgRu-15 catalyst were determined by CO-stripping techniques (Supplementary Fig. S25 and Table S5). When the current density was normalized to the ECSA, AgRu-15 exhibited enhanced activity for the reduction of both NO₃⁻ and NO₂⁻, leading to improved NH₃ generation rates compared to Ru@Ti (Supplementary Fig. S26). This finding highlights the strong intrinsic catalytic performance of the AgRu-15 catalyst.

The AgRu-15 catalyst demonstrated high efficiency in the electrochemical conversion of trace NO₃⁻, achieving an NH₃ FE of 99.86% in 1 mM NO₃⁻ at 0.1 V (Supplementary Fig. S27). Isotope-labeled ¹⁵NO₃⁻ electrolysis, coupled with ¹H nuclear magnetic resonance (NMR) analysis, confirmed that all generated NH₃ originated from the electrocatalytic reduction of NO₃⁻ (Supplementary Fig. S28). A batch conversion test with an initial concentration of 2000 ppm NO₃⁻ demonstrated nearly complete conversion to NH₃ within 1 h, with residual levels of NO₃⁻ and NO₂⁻ remaining below the World Health Organization (WHO) guidelines (Fig. 3g). Moreover, structural analysis revealed that the AgRu-15 catalyst preserved its morphology and surface chemical states after completing 10 consecutive 2-h reduction cycles (Supplementary Figs. S29–31). The heatmap in Fig. 3h illustrates an optimal half-cell energy efficiency (EE) of 53.7% at 0.2 V for the NO₃⁻-to-NH₃ conversion, marking a significant improvement over previous studies. Notably, the AgRu-15 catalyst also exhibits robust anti-interference capability, sustaining an NH₃ FE of approximately 96% even in complex simulated wastewater and real industrial wastewater (Fig. 3i, Supplementary Fig. S32, and Table S6). Comparative literature analysis further underscores the advantages of AgRu-15 for highly energy-efficient NH₃ production (Supplementary Table S7)^{26,39,50–57}.

Mechanisms of enhanced proton transfer kinetics on AgRu-15

To clarify the enhanced proton transfer kinetics of AgRu-15 compared to p-Ag and Ru@Ti, we conducted density functional theory (DFT) calculations to evaluate the reaction barriers (∆G_b) for H₂O dissociation. Slab models of Ru(002) and Ru(002)/Ag(111) were developed to simulate the surfaces of Ru@Ti and AgRu-15, respectively, based on the aforementioned structural analysis (see detailed discussion in Supplementary Fig. S33 and Supplementary Note 1). The Ag(211) slab was selected as a simplified model for the p-Ag surface that exposes a higher fraction of undercoordinated sites (Supplementary Fig. S33)⁵⁸. We note that the DFT calculations are used to identify mechanistic trends rather than to quantitatively reproduce the real catalytic system. As shown in Fig. 4a, Ag(211) exhibits the highest ∆G_b for H₂O dissociation (1.36 eV), followed by Ru(002) (0.77 eV) and Ru(002)/Ag(111) (0.66 eV). The ∆G_b is inversely related to the adsorption energies of (*H +  *OH) species, suggesting the kinetics of active *H generation in the order: Ru(002)/Ag(111) > Ru(002) > Ag(211). The stronger adsorption of (*H + *OH) species on Ru(002)/Ag(111) can be attributed to an upward shift in the Ru d-band center influenced by Ag, as suggested by the projected density of states (Fig. 4b). The differential charge density diagram (inset in Fig. 4b and Supplementary Fig. S34) supports this shift, illustrating interfacial charge redistribution characterized by electron transfer from the Ag substrate to the surface Ru atoms, in alignment with XPS findings (Fig. 2i). Considering that the hydrogenation of NO₃⁻ predominantly follows the L–H mechanism in alkaline media³¹, the active *H species is therefore expected to play an important role in NO₃RR on Ru@Ti and AgRu-15 catalysts.

Fig. 4. Mechanistic roles of *H_upd in NO₃RR.

a Free-energy diagram for H₂O dissociation on Ru(002), Ru(002)/Ag(111), and Ag(211). b The PDOS diagram of the Ru’s d orbitals for Ru(002) and Ru(002)/Ag(111), with an inset illustrating differential charge distributions on Ru(002)/Ag(111), where yellow represents increased charge density and cyan indicates decreased charge density. Color code: Ru-brick red; Ag-gray. c CV profiles of the catalysts at a scan rate of 10 mV s⁻¹ in 1 M KOH solution (without iR-correction). d Plots of C_φ vs. applied potentials (without iR-correction) for the AgRu-15 and Ru@Ti in Ar and H₂-saturated 1 M KOH solutions. C_φ: the *H_UPD adsorption pseudo-capacitance. e Ru-H_atop bands recorded on Ru and AgRu thin film electrodes in Ar-saturated 1 M KOH. f Normalized Ru-H_atop peak area as a function of time at 0 V vs. RHE, with initial spectra in 1 M KOH, followed by KNO₃ injection after 10 min (black arrow) to establish a composition of 1 M KOH + 10 mM KNO₃. g Simulation snapshots of NO₃⁻, NO₂⁻, and H₂O adsorption on the Ru(002)/Ag(111) and Ru(002) surfaces under 1 ML *H_multiple coverage. h Equilibrium coverage densities and adsorption rate constant (k) of NO₃⁻ and NO₂⁻ on the Ru(002)/Ag(111) and Ru(002) surfaces across varying *H_multiple coverages. Source data are provided as a Source data file.

Given the energy-efficient NO₃RR occurring above 0 V, the underpotential deposition of *H (*H_UPD) on various catalysts was assessed through cyclic voltammetry (CV) (Fig. 4c, Supplementary Figs. S35 and 36). As shown in Fig. 4c, the p-Ag catalyst showed negligible *H_UPD-related current, while Ru@Ti and AgRu-15 exhibited notable *H_UPD-related currents between 0 and 0.25 V, consistent with the DFT results (Fig. 4a). The *H_UPD stripping peak for AgRu-15 at 0.113 V, previously denoted to the exchange adsorption between *H_UPD and hydroxyl (*OH)⁵⁹, showed a 28 mV positive shift relative to Ru@Ti, indicating a stronger *H_UPD affinity on the AgRu-15 surface. A plot correlating the voltage scan rate with the *H_UPD/*OH exchange peak (Supplementary Fig. S36) revealed a steeper slope for AgRu-15, suggesting slower *H_UPD desorption rates compared to Ru@Ti⁶⁰. This behavior enables the retention of *H_UPD at elevated potentials, partially explain the more positive onset potentials for NO₂⁻ reduction observed on AgRu-15 compared to Ru@Ti (Fig. 3c and Supplementary Fig. S20e).

The coverages of *H_UPD on the AgRu-15 and Ru@Ti catalysts were estimated using electrochemical impedance spectroscopy (EIS) in Ar and H₂-saturated 1 M KOH^61–63. The utilization of H₂ as a proton source improves proton supply kinetics compared to bulk H₂O⁴², contributing to near-saturation *H_UPD coverages at accessible catalytic sites. Nyquist plot analysis provided insights into the *H_UPD adsorption pseudo-capacitance (C_φ) and the resistance related to *H_UPD formation (R₂) (Supplementary Fig. S37). As shown in Fig. 4d, under an Ar atmosphere, Ru@Ti exhibited *H_UPD adsorption at potentials below 0.175 V, while AgRu-15 began at 0.25 V, indicating the easier formation of *H_UPD on AgRu-15. Integral analysis of C_φ revealed that AgRu-15 had a 1.73-fold higher *H_UPD absorption charge (Q) compared to Ru@Ti; however, its Ru-related ECSA was approximately 7.3 times greater than that of Ru@Ti (Supplementary Fig. S25). When Q is normalized to ECSA to determine the *H_UPD coverage, it indicates that AgRu-15 has only 25% of Ru@Ti’s *H_UPD coverage, suggesting inherently lower *H_UDP coverage on AgRu-15.

The introduction of H₂ caused negligible changes in the Q-values of the Ru@Ti catalyst, indicating its nearly saturated *H_UPD coverage even in Ar (Fig. 4d). In contrast, AgRu-15 exhibited a 25% increase in Q-values, rising from 14.4 to 18.1 mC, which corresponds to increased *H_UPD coverage. However, in the presence of H₂, AgRu-15 showed varying degrees of decline in NO₃RR activities and NH₃ FE, accompanied by a slight increase in NO₂⁻ FE (Supplementary Fig. S38). As such, the enhanced NO₃RR activity of AgRu-15 highly correlates with its inherently low *H_UPD coverage. These findings imply that AgRu-15 and Ru@Ti exhibit different surface adsorption characteristics and *H_UPD utilization mechanisms.

In terms of formation resistance for *H_UPD, AgRu-15 has smaller logR₂ values than Ru@Ti under both Ar and H₂ (Supplementary Fig. S39), which indicates faster kinetics for *H_UPD generation, aligning with DFT calculations (Fig. 4a). The lower slope of logR₂ versus applied potentials for AgRu-15 (8.4 mV dec⁻¹) compared to Ru@Ti (13.6 mV dec⁻¹) highlights the accelerated kinetics of *H_UPD transfer on AgRu-15 (Supplementary Fig. S39a)^63,64. Consequently, the enhanced proton transfer efficiency (low KIEs) for NO₃RR on AgRu-15 can be attributed to the favorable kinetics of *H_UPD generation and transfer, combined with slow *H_UPD desorption rates, despite its inherently low *H_UPD coverage.

Mechanisms of *H_UPD-accelerated NO₃RR on AgRu-15

To investigate the mechanisms of the accelerated NO₃RR on AgRu-15 with inherently low *H_UPD coverage, we utilized attenuated total reflection surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) to monitor the evolution of *H_atop (*H adsorbed on top of Ru atoms) across varying potentials (Fig. 4e). Depending on the applied potentials, the *H_atop can signify both *H_OPD at potentials below 0 V and *H_UPD at potentials above 0 V. While *H_ODP exists only in the form of *H_atop, the *H_UPD can appear in multiple forms based on its concentrations and the applied potentials, such as multiple-coordinated *H (*H_multiple), which remains undetectable by traditional vibrational spectroscopies^27,42,65. Importantly, at potentials above 0 V, *H_multiple, and *H_atop can interconvert dynamically based on applied voltages and active *H concentrations⁴², allowing for their collective classification as *H_UPD in our system.

The *H_atop band for the pure Ru surface at 1845 cm⁻¹ at 0.025 V shifts to 1821 cm⁻¹ at −0.1 V in a NO₃⁻-free electrolyte, yielding a Stark tuning rate of 192 cm⁻¹ V⁻¹ (Fig. 4e). In contrast, the *H_atop band on the AgRu surface was centered at much higher wavenumbers, shifting from 1849 to 1838 cm⁻¹ below 0.05 V, which indicates stronger adsorption but a lower Stark tuning rate of 73.3 cm⁻¹ V⁻¹. This lower Stark tuning rate of the AgRu surface points to its relatively lower *H_atop coverage compared to the pure Ru surface^27,65, aligning with the above CV and EIS analyses. The introduction of NO₃⁻ significantly diminishes the Ru-H_atop band intensity for AgRu catalysts at 0 V, while the effect is minimal for pure Ru (Fig. 4f and Supplementary Fig. S40). This outcome implies that the *H_atop (or *H_UPD) on AgRu catalysts is more active in promoting the NO₃RR. Additionally, increasing the NO₃⁻ concentration to 0.1 M almost suppresses the *H_atop band for AgRu catalysts at 0.015 V, providing evidence that *H_UPD is directly involved in the NO₃RR (Supplementary Fig. S40), supporting the L–H mechanism.

We used DFT calculations to further compare the surface adsorption characteristics of AgRu-15 and Ru@Ti catalysts, focusing on the binding energies of *H_atop and *OH species on Ru(002)/Ag(111) and Ru(002) models at varying *H_multiple coverages (Supplementary Table S8). The DFT results indicate a slightly stronger *H_atop affinity but higher oxophilicity for Ru(002)/Ag(111) compared to Ru(002). For example, at the same 1/3 ML *H_multiple coverage, the calculated binding energies are—0.27 eV for *H_atop on Ru(002)/Ag(111) and –0.19 eV for Ru(002). In contrast, *OH binding energy is calculated to be stronger on Ru(002)/Ag(111) (–0.24 eV) than on Ru(002) (0.06 eV). The weaker *OH adsorption on Ru(002) suggests probable renewal of the Ru@Ti catalyst surface sites through *OH desorption, likely via forming *OH-water-K⁺ adducts⁶⁶. The *OH desorption makes the active sites available for *H_UPD adsorption and may contribute to the observed high *H_UPD coverage on Ru@Ti in the experiment. Correspondingly, for the Ru@Ti surfaces, increasing *H_atop (*H_UPD) coverage results in more occupied active sites, thereby limiting NO₃⁻ adsorption. This interpretation is further supported by the calculated trend that as *H_multiple coverage rises, the binding strengths of NO₃⁻, NO₂⁻, and H₂O decline consistently (Supplementary Table S9).

In contrast, Ru(002)/Ag(111)‘s higher oxophilicity may favor *OH adsorption and promotes the *H_UPD removal via a reaction with *OH to yield H₂O²⁷, which may help explain the inherently lower *H_UPD coverage on AgRu-15. In this context, the low *H_UPD coverage of the AgRu-15 catalyst enables it to provide more available active sites for the competitive NO₃⁻, NO₂⁻, and H₂O adsorption. Moreover, a comparative analysis of the binding energies of NO₃⁻, NO₂⁻, and H₂O on the Ru(002)/Ag(111) surface at varying *H_multiple coverages suggests that Ru(002)/Ag(111) consistently exhibits stronger adsorption of both NO₃⁻ and NO₂⁻ compared to H₂O (Supplementary Fig. S41 and Table S9). Based on these DFT-calculated affinities of NO₃⁻, NO₂⁻, and H₂O under varying *H_multiple coverages, coarse-grained molecular dynamics (CGMD) simulations were conducted to assess the competitive adsorption of these species on the model surfaces (Fig. 4g, h, Supplementary Figs. S42–45, and Table S10–14). The CGMD results suggest that the Ru(002)/Ag(111) surface has larger adsorption rate constants for both NO₃⁻ and NO₂⁻ compared to the Ru(002) surface. However, the adsorption rate constants for H₂O are comparable on both surfaces, likely owing to the high concentration of H₂O in the simulation system (Fig. 4h, Supplementary Tables S11 and 12). Consistent with this trend, Ru(002)/Ag(111) tends to show higher surface coverage of NO₃⁻ and NO₂⁻, together with lower H₂O coverage, than Ru(002) (Fig. 4h, Supplementary Tables S13 and 14). These results suggest that the Ru(002)/Ag(111) surface may be more favorable for competitive adsorption of NO₃⁻ and NO₂⁻, particularly NO₂⁻, thereby facilitating the involvement of *H_UPD in the NO₃RR, as corroborated by the ATR-SEIRAS results. Collectively, the experimental observations and theoretical calculations suggest that the *H_UPD-accelerated NO₃RR on AgRu-15 may arise, at least in part, from the higher oxophilicity of Ru overlayers induced by the Ag matrix. In this picture, Ag–Ru synergy may improve *H_UPD generation and transfer kinetics, as well as competitive NO₃⁻ and NO₂⁻ adsorption, ultimately contributing to energy-efficient NO₃RR.

Identification of NO₃RR pathways

To inquire thoroughly into the NO₃RR pathways, in situ Fourier transform infrared (FT-IR) spectroscopy, and ATR-SEIRAS were employed to monitor changes in reactants and products within a thin-layer electrolyte, as well as adsorbed intermediates on electrodes. Potential-dependent spectra were derived from absorbance variations, where the upward bands indicated the formation of NO₃RR intermediates and the downward bands represented reactant consumption (e.g., NO₃⁻). In thin-layer FT-IR analysis (Fig. 5a–c and Supplementary Fig. S46), downward bands at 1395 and 1354 cm⁻¹ on the p-Ag electrode correspond to the symmetric and asymmetric stretching vibrations of the N–O bond for NO₃⁻¹¹. These peaks intensified at 0.3 V, indicating rapid consumption of NO₃⁻, while an upward band at 1236 cm⁻¹, associated with the antisymmetric stretching vibration of N–O for NO₂⁻, signified its accumulation¹¹. Notably, AgRu-15 exhibited NO₃⁻ consumption at 0.4 V, while Ru@Ti did so at 0.2 V, aligning with their NO₃RR onset potentials. The absence of NO₂⁻ bands for both AgRu-15 and Ru@Ti suggests preferential depletion of NO₂⁻ within their thin-layer electrolytes.

Fig. 5. NO₃RR pathways.

In situ thin-layer electrochemical FT-IR spectra of NO₃RR on p-Ag (a), AgRu-15 (b), and Ru@Ti (c) in 1 M KOH + 100 mM KNO₃. ATR-FTIR spectra of NO₃RR on Ag (d), AgRu (e), and Ru (f) thin film electrodes in 1 M KOH + 100 mM KNO₃. g Reaction free energies for various intermediates of NO₃RR on Ru(002), Ru(002)/Ag(111), and Ag(211) slabs, highlighting the free energy barriers for NO₃⁻ and NO₂⁻ adsorption (ΔG₁ and ΔG₂), as well as for the rate-determining step of NO₂⁻ reduction (ΔG₃). * represents an adsorption site. h Comparative analysis of the energy barriers for NO₂⁻ formation on the Ru(002)/Ag(111) and Ag(211) surfaces. Source data are provided as a Source data file.

ATR-SEIRAS of Ag, AgRu, and Ru thin film electrodes on Au-coated Si prisms revealed a standard band at ~1210 cm⁻¹ attributed to the Si−O−Si stretching vibration (Fig. 5d–f and Supplementary Fig. S47)⁶⁷. Increased intensities at 1285 and 1520 cm⁻¹ on the Ag surface indicated accumulation of NO₂⁻ and *NOH intermediates^68,69, respectively, across the potential range from 0.4 to −0.1 V (Supplementary Table S15). In contrast, AgRu and Ru surfaces showed no adsorbed NO₂⁻ due to their excellent NO₂⁻ reduction activity. Three additional adsorbed intermediates on the AgRu and Ru surfaces were identified: the bending δ_d (H–N–H) mode of NH₃ at 1620 cm⁻¹⁶⁷, the stretching ν (–N–O–) mode of NH₂OH at ~1150 cm^−111,69, and the bending δ (H–N) mode of *NOH at ~1520 cm⁻¹⁶⁹. These findings indicate similar pathways for NO₂⁻ hydrogeneration on AgRu and Ru catalysts.

The NO₃RR pathways on AgRu-15 catalysts—characterized through in situ thin-layer FTIR, ATR-SEIRAS, and DFT calculations—involve sequential deoxygenation (NO₃⁻ → *NO₃ → *NO₂ → *NO), followed by hydrogenation (*NO → *NOH → *NH₂OH → *NH₂ → *NH₃). Adsorption of *NO₃ and its subsequent deoxygenation to *NO₂ are more favorable on Ag(211) surfaces than on Ru(002)/Ag(111) and Ru(002) (Fig. 5g). The Ru(002) exhibits a high Gibbs free energy barrier (ΔG₁ = 1.01 eV) for initial adsorption of NO₃⁻ (*NO₃), which explains the negligible NO₃RR activity of the Ru@Ti catalyst (Fig. 3a). The high free energy barriers for the hydrogenation reactions on Ag(211), particularly for the conversion of *NO to *NOH (ΔG₃ = 1.2 eV), supports the preferential release of NO₂⁻ (Fig. 5h), which aligns with the exclusive selectivity of p-Ag catalysts for NO₂⁻. The gradient distribution of Ag and Ru nanophases in AgRu-15 enables the efficient absorption of NO₂⁻, released from Ag nanocavities, onto Ru nano-coatings, as indicated by the energetically favorable adsorption of *NO₂ (ΔG₂ = −0.31 eV) on Ru(002)/Ag(111) (Fig. 5g). Furthermore, the conversion of *NO to *NOH emerges as the rate-determining step for NO₂⁻ reduction on Ru(002)/Ag(111), with a lower ΔG₃ (0.64 eV) than on Ru(002) (0.97 eV) (Fig. 5g). These findings suggest that the cascade pathways, including the reduction of NO₃⁻ to NO₂⁻ on Ag and the subsequent conversion of NO₂⁻ to NH₃ on Ru, are thermodynamically favorable on AgRu-15 catalysts.

H₂-driven NH₃ production and economic feasibility

Given the similar nature of Pt-group metals (PGM) for *H_UPD generation, the AgRu system was expanded to include a series of Ag-PGM (PGM = Pt, Rh, Pd, and Ir) catalysts. The Ag-PGM catalysts demonstrated higher onset potentials for the NO₃RR compared to p-Ag, as evidenced in Supplementary Figs. S48 and S49. This observation aligns with DFT-derived stronger binding strengths of *NO₃ at the electron-enriched Ag-PGM interface sites (Supplementary Fig. S49), which may serve as active sites for enhanced adsorption and activation of NO₃⁻. Notably, for Ag-PGM catalysts, a decrease in NO₃RR activity is linked to reduced *OH binding strengths on the PGM sites (*Ru, *Rh, *Pt, and *Pd) (Fig. 6a). The NO₃RR activity of AgPt (*Pt) and AgPd (*Pd) is close to that of the p-Ag catalyst, attributed to the lower affinity of *Pt and *Pd sites. Remarkably, the deviation of the AgIr (*Ir) catalyst’s activity from an anticipated trend can be attributed to its insufficient *H_UPD generation capability (Supplementary Fig. S50). The AgRu and AgRh catalysts, characterized by high *OH binding affinity, display improved selectivity for converting NO₃⁻ to NH₃, whereas other counterparts primarily yield NO₂⁻ (Fig. 6b and Supplementary Fig. S51). This result indicates that the NO₂⁻ generated rapidly on the Ag phases of Ag-PGM catalysts cannot be further converted to NH₃ on the PGM phases (e.g., *Pd and *Pt) with low *OH affinity. These findings suggest that the *OH adsorption strength can act as a critical descriptor for designing more efficient *H_UPD-based catalysts, as illustrated by the *H_UPD-accelerated NO₃RR on the AgRu-15 catalyst.

Fig. 6. H₂-driven NH₃ production.

a Correlation of the current density (j) at 0 V vs. RHE with ΔG_*OH for Ag-PGM (PGM = *Ru, *Ir, *Rh, *Pt, and *Pd) catalysts, with a reference line indicating the j of the p-Ag catalyst. The orange tetrahedron denotes the Ru(002) plane of the Ru@Ti catalyst, which exhibits the lowest NO₃RR due to weak NO₃⁻ adsorption. b Heatmaps of FE for NH₃ synthesis across varying potentials using Ag-PGM catalysts. c Polarization curves of alkaline NO₃RR, directly coupled with anodic HOR, collected at 1 M KOH with varying concentrations of NO₃⁻ using AgRu-15 as the cathode and Pt/C as the anode (Scan rate: 10 mV s⁻¹). d Stability assessment of the AgRu-15 cathode in the paired electrolysis system operating in 1 M KOH with 100 mM KNO₃ at 200 mA cm^_2. e Contour plots of the levelized cost as a function of the current density and cell voltage at an NH₃ FE of 81%, utilizing anodic HOR with the H₂ sourced from natural gas without CCUS and the NO₃⁻ from electrodialysis. Source data are provided as a Source data file.

To assess the practical potential of Ag-PGM catalysts for large-scale NO₃⁻-to-NH₃ conversion, we constructed a membrane-free flow cell in a two-electrode configuration for paired alkaline electrolysis (Supplementary Fig. S52). This system features cathodic NO₃RR using the AgRu-15 catalyst, paired with anodic H₂ oxidation reaction (HOR) using commercial Pt/C catalysts. As depicted in Fig. 6c, increasing electrocatalytic current densities resulted in a sudden increase in voltage of the flow cell for NO₃⁻ concentrations of 10, 50, and 100 mM, suggesting mass transfer limitations associated with the porous AgRu-15 catalyst. At NO₃⁻ concentrations of 0.5 and 1.0 M, the paired system achieved current densities of up to 2.0 A cm⁻² at a full-cell voltage of only 0.75 V (Supplementary Fig. S53), demonstrating low energy consumption for large-scale NH₃ production. Notably, under a typical industrially relevant current density of 200 mA cm⁻², the paired NO₃RR-HOR system maintained stable NH₃ production for over 100 h (Fig. 6d), even at full-cell voltages below zero, highlighting the superiority of the AgRu-15 catalyst for low-overpotential NO₃RR (Fig. 1a). Techno-economic evaluations suggest that with an NH₃ FE of 81%, utilizing anodic H₂ (without carbon capture, utilization and storage, CCUS) and cathodic AgRu-15 catalysts can achieve NH₃ production costs below the commercial price of $1.15 per kg across a broad current density range of 100–920 mA cm⁻² (Fig. 6e, see detailed discussions in Supplementary Figs. S54–56 and Table S16). When waste nitrate sources from sewage treatment are utilized as a raw material at no additional costs³⁹, this paired electrolysis system can also produce NH₃ at costs lower than the commercial price, even if blue H₂ with CCUS is used at the anode (Supplementary Fig. S56). This economic feasibility marks a significant advancement in the field of NH₃ electrosynthesis.

Discussion

In summary, this study presents a class of Ag-PGM catalysts, notably AgRu-15, as effective *H_UPD supply systems for energy-efficient NH₃ electrosynthesis from NO₃RR. We developed a 3D hierarchical porous architecture that incorporates nanoconfined Ag and Ru sites to mimic enzymatic substrate channels. Kinetic assessments and DFT calculations confirm significant synergies between isolated Ag and Ru phases, enabling cascade NO₃RR pathways that couple improved onset potentials for NO₃⁻ reduction on Ag with enhanced proton transfer kinetics for NO₂⁻-to-NH₃ conversion on Ru. We elucidate the correlation between *OH binding strength and *H_UPD coverage/accessibility on the Ru surface. Our findings demonstrate that the enhanced oxophilicity of Ru, activated by the Ag matrix, strengthens *OH adsorption, thereby regulating *H_UPD coverage via the consumption of partial *H_UPD to form H₂O. This mechanism enables sustained regeneration of Ru-centered active sites for competitive adsorption of NO₃⁻ and NO₂⁻, ultimately promoting the *H_UPD-assisted NO₃RR. Using the AgRu-15 catalyst, we achieved an NH₃ partial current density of 2.2 A cm⁻² (Y_NH3 = 166.6 ± 1.36 mg h⁻¹ cm⁻²) in 1 M NO₃⁻ at 0 V, a maximum half-cell EE of 53.7% at 0.2 V, and H₂-driven NH₃ production costs below the commercial threshold of $1.15 per kg. Our findings present *OH adsorption strength as a critical descriptor for assessing *H_UPD reactivity, providing insights to design more efficient catalysts for NO_x hydrogenations.

Methods

Chemicals

Ag rods (2 mm diameter, 99.99%), Ag foil (0.1 mm thickness, 99.99%), and Ag mesh (180 mesh, 99.99%) were purchased from Kunshan Shengzhaoduan New Materials Co., Ltd. Titanium rods (2 mm diameter, 99.99%) were obtained from Hebei Aoshuo Metal Materials Co., Ltd. Carbon paper (SEN90L91-0.5) was procured from Suzhou Sinero Technology Co., Ltd. Hydrochloric acid (HCl, 36–38 wt%, AR), nitric acid (HNO₃, 65–68 wt%, AR), potassium hydroxide (KOH, ≥85%, AR), potassium nitrate (KNO₃, ≥99%, AR), potassium chloride (KCl, ≥99.5%, AR), and lead nitrate (Pb(NO₃)₂, ≥99%, AR) were supplied by Xilong Scientific Co., Ltd. Ruthenium(III) chloride hydrate (RuCl₃·xH₂O, Ru content 36–40%, AR) was purchased from Shanghai Macklin Biochemical Co., Ltd. Rhodium(III) chloride hydrate (RhCl₃·xH₂O, Rh ≥38.5%, AR), iridium(III) chloride trihydrate (IrCl₃·3H₂O, ≥98%, AR), and deuterium oxide (D₂O, 99.9 atom% D) were acquired from Beijing Innochem Science and Technology Co., Ltd. Potassium nitrite (KNO₂, ≥97%, AR), sodium tetrachloropalladate(II) (Na₂PdCl₄, ≥98%, AR), hexachloroplatinic acid hexahydrate (H₂PtCl₆·6H₂O, Pt ≥37.5%, AR), sodium hypochlorite solution (NaClO, 5% available chlorine), sulfanilamide (≥99%, AR), N-(1-naphthyl)ethylenediamine dihydrochloride (≥98%, AR), sodium nitroprusside dihydrate (Na₂[Fe(CN)₅NO]·2H₂O, ≥99.5%, AR), phosphoric acid (H₃PO₄, 85 wt%, AR), sodium citrate (C₆H₅Na₃O₇, ≥98%, AR), salicylic acid (C₇H₆O₃, ≥99.5%, AR), ammonium chloride (NH₄Cl, PT), and sulfamic acid (NH₂SO₃H, ≥99.5%, AR) were sourced from Shanghai Aladdin Biochemical Technology Co., Ltd. Platinum on carbon catalyst (20 wt% Pt/C) was obtained from Thermo Fisher Scientific (Alfa Aesar). All chemicals were used as received without further purification.

Pretreatment of Ag rod substrate

The silver (Ag) rod underwent a three-stage surface engineering process: (1) mechanical polishing with graded sandpaper (3000–5000 mesh) to remove native oxides and impurities; (2) electrochemical reduction at −0.576 V vs. RHE in 1 M KOH for 10 min using a Hg/HgO reference electrode and graphite counter electrode to eliminate residual oxides; (3) ultrasonic cleaning in ultrapure water (18.2 MΩ cm, Milli-Q) for 5 min for subsequent experiments.

The Ag rod electrode area was determined using the geometric partition approximation:

$$d=\frac{d_1+d_2+d_3}{3}$$ 1

$$S_1=\frac{{\mathrm{\pi }d}_1^2}{4}$$ 2

$$S_1=\mathrm{\pi }dh$$ 3

$$S=S_1+S_2$$ 4

The Ag rod electrode was geometrically characterized by three defined diameters (base d₁, mid-section d₂, top d₃), corresponding basal area (S₁) and lateral surface area (S₂), and an active length h, with the top portion above h was wrapped with a sealing film to keep it out of the reaction. The electrode area was controlled to ~0.6 cm².

Preparation of p-Ag

The nanoporous silver (p-Ag) was prepared through a scalable electrochemical anodization–reduction process. Initially, anodization was conducted using a two-electrode system, where two silver rods served as the anode and cathode, respectively. A potential of 1.0 V was applied between the two Ag electrodes for 5 min in 1 M HCl at 25 °C, resulting in the formation of a uniform AgCl layer on the anode. Subsequently, the electrochemical reduction of the anodic Ag rod with AgCl coatings was performed in a three-electrode configuration, using a graphite rod as the counter electrode and Hg/HgO as the reference electrode. Cyclic voltammetry was performed (20 cycles, 1.0–0 V vs. RHE, 20 mV s⁻¹ in 1 M KOH) until a stable current was achieved, yielding the p-Ag electrode.

Preparation of AgRu-15

A 25 mM solution of RuCl₃·xH₂O in 1 M HCl was aged for 1 week to prepare the precursor for the AgRu catalysts. The modification of the p-Ag electrode by metallic Ru was conducted using a three-step alternating cycle: (1) chronoamperometric deposition of Ru at −6 mA cm⁻² for 2.5 min in a two-electrode configuration (p-Ag as the cathode and a graphite rod as the anode); (2) immersion of the electrode in the aged electrolyte for 2.5 min under continuous stirring at 600 rpm; and (3) electrochemical reduction of the Ru-modified p-Ag electrode at −1.5 V versus RHE in 1 M KOH for 2.5 min in a three-electrode setup, employing a Hg/HgO reference electrode and a graphite rod as the counter electrode. This three-step cycle was repeated to achieve a cumulative chronoamperometric deposition time for the target sample (e.g., 15 min for AgRu-15). Parallel preparations of AgRu-15 were conducted on Ag foil and Ag mesh substrates. For comparative studies, Ag-PGM (PGM = Rh, Pt, Pd, and Ir) catalysts were synthesized using identical protocols with respective 25 mM precursors (RhCl₃·xH₂O, H₂PtCl₆·6H₂O, Na₂PdCl₄, or IrCl₃·3H₂O) in 1 M HCl. Additionally, Ru@Ti was prepared by modifying Ru on a pretreated titanium (Ti) rod using the same preparation protocol.

Material characterization

Scanning electron microscopy (SEM, JSM-6701) was performed at an accelerating voltage of 10 kV. Transmission electron microscopy (TEM) and high-resolution transmission electron microscopy (HRTEM) were performed on a JEM-2100 microscope equipped with an energy-dispersive X-ray spectroscope (EDX) at an accelerating voltage of 200 kV. High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images were obtained using a Titan 80–300 instrument. X-ray diffraction (XRD) analysis was performed on a D8 Advance diffractometer (Bruker) using Cu-Kα radiation (λ = 1.54 Å). X-ray photoelectron spectroscopy (XPS) measurements were conducted with a Thermo ESCALAB Xi+ (Thermo Fisher). X-ray absorption spectra (XAS) of Ag and Ru K-edges were obtained at the 1W1B beamline of the Beijing Synchrotron Radiation Facility (BSRF), and the data were analyzed using the Athena software package to extract XANES and EXAFS data. ¹H nuclear magnetic resonance (NMR) analyses were performed on a Bruker Avance 600 MHz spectrometer. The Ag and Ru contents were determined by inductively coupled plasma-optical emission spectroscopy (ICP-OES, Agilent 5800).

Electrochemical tests

All electrochemical tests were conducted at ambient conditions using a three-electrode configuration in an H-type cell, separated by a Nafion 117 membrane (thickness: 183 μm), with 1200 rpm magnetic stirring. Pretreatment Protocol for Nafion 117 membrane: The membrane was first treated in 5% H₂O₂ solution at 80 °C for 1 h to remove organic impurities, followed by heating in deionized water at 80 °C for 1 h to remove residual peroxide and promote membrane swelling. It was then treated in 0.5 M H₂SO₄ solution at 80 °C for 1 h for decontamination and protonation and finally heated at 80 °C again in deionized water for 1 h to rinse away residual acid and ensure complete hydration. All electrolyte solutions were freshly prepared prior to use. The AgRu-15 catalyst, serving as a working electrode, was pre-reduced at −0.576 V vs. RHE for 30 min in 1 M KOH to ensure oxide-free surfaces, which, coupled with a Ni foam counter electrode and a Hg/HgO reference electrode (1 M KOH), were connected to a BioLogic (VSP-3e) electrochemical workstation. The electrolyte consisted of 1 M KOH with varying KNO₃ concentrations (1–2000 mM), purged with Ar for 20 min to remove dissolved O₂ and N₂ (pH = 13.96 ± 0.02, Resistance of electrolyte: 1.88 ± 0.05 Ω). A 90% iR correction was applied, and the Hg/HgO electrode was pre-calibrated against a Pt electrode in H₂-saturated electrolyte under 1 M KOH, where the 0 V level of the Pt/H₂ reference corresponds to −0.922 V versus Hg/HgO (Supplementary Fig. S18). All potentials were calibrated to the RHE scale using the equation: E_RHE = E_Hg/HgO + 0.922 V – 90% iR. Cyclic voltammetry (CV) and linear sweep voltammetry (LSV) were performed at a 10 mV s⁻¹ scan rate. Chronopotentiometry (CP) measurements were conducted at various applied potentials in Ar-saturated 1 M KOH with KNO₃ concentrations ranging from 1 to 1000 mM. The kinetic isotope effect (KIE) was assessed using D₂O (99.9 atom% D) as a solvent, calculated using KIE = J_H/J_D, where J_H is the current density in H₂O and J_D is the current density in D₂O. AC impedance spectroscopy (EIS) was recorded over a frequency range of 60 kHz to 10 mHz at an applied potential with an amplitude of 5 mV. The reaction rate constant (k) was determined by monitoring the temporal variation of NO₃⁻ or NO₂⁻ concentrations during electrolysis, based on first-order kinetics with exponential decay, described by C_t = C₀ exp(−kt), where C₀ is the initial concentration of the reactants (NO₃⁻ or NO₂⁻), and C_t is the concentration at time t. Error bars were derived from the relative standard deviations of at least three independent replicates. Note: for electrochemical testing in NO₃⁻ concentrations of 1–100 mM, the electrolyte volume was 30 mL with an electrode area of 0.6 cm²; for 500–2000 mM NO₃⁻, the electrolyte volume was 70 mL with an electrode area of 0.25 cm². The Ru mass loading on the AgRu-15 catalyst was determined to be 0.65 mg cm⁻² by ICP analysis.

Determination of ion concentration

Quantification of nitrogen-containing species was conducted using established spectrometric protocols. All electrolyte samples were diluted to concentrations within the linear detection range of their respective calibration curves.

NH₄⁺ quantification

The concentration of NH₃ was determined via UV–vis spectrophotometry employing the indophenol blue method. Typically, a 2 mL aliquot of the diluted sample was combined with 1 mL of an alkaline reagent (1 M NaOH containing 5 wt% sodium citrate and 5 wt% salicylic acid) and 1 mL of 0.05 M NaClO. Following thorough mixing, 0.2 mL of 1 wt% sodium nitroferricyanide solution was added to initiate the chromogenic reaction. The mixture was incubated at 25 °C for 1 h to produce the indophenol blue complex, with absorbance measured at 655 nm. A calibration curve was established using standard NH₄Cl solutions for NH₃ quantification.

NO₂⁻ quantification

A color reagent was prepared by mixing 4 g of sulfonamide, 0.2 g of N-(1-naphthyl) ethylenediamine dihydrochloride, and 10 mL of phosphoric acid with 50 mL of ultrapure water. For analysis, a 5 mL aliquot of the diluted sample was acidified with 1 mL of 1 M HCl to initiate the acidification. Subsequently, 0.1 mL of the color reagent was added, followed by vigorous shaking to ensure uniformity. The absorption intensity was recorded at 540 nm after 20 min at 25 °C. A calibration curve was generated using standard KNO₂ solutions for NO₂⁻ quantification.

NO₃⁻ quantification

A mixture of 1 mL of 1 M HCl and 0.1 mL of 0.8 wt% sulfamic acid solution was combined with 4 mL of the diluted electrolyte. After thorough mixing, the solution was allowed to stand for 20 min, followed by UV–vis absorbance measurements at wavelengths ranging from 200 to 300 nm. Absorbance intensities at 220 and 275 nm were recorded, and the final absorbance difference was calculated using the formula: A = A_220 nm − 2 * A_275 nm. Quantification was achieved through a calibration curve constructed with standard KNO₃ solutions.

The calculation of FE, yield rate (Y_NH3) and EE_NH3

The Faradaic efficiency (FE) for the NO₃RR to NH₃ and NO₂⁻ at a specified potential was determined using the equation:

$$\mathrm{FE}=\frac{C\times V\times N\times F}{Q}$$ 5

Where C represents the molar concentration of the product (M), V is the volume of the electrolyte (mL), N is the electron transfer stoichiometry (8 for NH₃ and 2 for NO₂⁻), F is the Faraday constant (96485 C mol⁻¹), and Q is the cumulative charge passed during electrolysis (C).

The yield rate of NH₃ (Y_NH3) at the specified potential was calculated using:

$${\mathrm{Y}}_{\mathrm{NH}3}=\frac{C\times V}{S\times T}$$ 6

Where S denotes the area of the catalyst (cm²), and T is the reaction time.

The half-cell energy efficiency of ammonia (EE_NH3) was defined as the ratio of fuel energy to applied electrical power, which was calculated with the following equation:

$${\mathrm{EE}}_{\mathrm{NH}3}=\frac{\left(1.23-E_{NH3}^0\right)\times {\mathrm{FE}}_{\mathrm{NH}3}}{1.23-E}$$ 7

Here, E refers to the applied potential (vs. RHE), FE_NH3 is the Faradaic efficiency for NH₃, and 1.23 V is the equilibrium potential of water oxidation.$E_{\mathrm{NH}3}^0$ is the equilibrium potential of NO₃RR to NH₃ (0.6812 V vs. RHE in 1 M KOH + 10 mM NO₃^⁻), which, under different electrolyte conditions, can be calibrated through the following equation:

$$E_{NH3}^0=E_0^0+\frac{0.0591}{n}\times \log \frac{\left[{\mathrm{NO}}_3^{-}\right]}{{\left[{\mathrm{OH}}^{-}\right]}^9}$$ 8

$E_0^0$ is the theoretical equilibrium potential of NO₃RR to NH₃, which is calculated to be −0.1314 V vs. SHE. $\left[{\text{NO}}_3^{-}\right]$ is the concentration of nitrate (1–1000 mM), $\left[{\text{OH}}^{-}\right]$ is the concentration of OH⁻ (1 M), and n is the electron transfer number of the reaction (n = 8).

¹⁵N isotope-labeling experiment

In this study, a typical procedure for ¹H NMR analysis involved the preparation of a mixture consisting of 125 μL of electrolyte, 125 μL of 15 mM maleic acid in DMSO-d₆, 50 μL of 4 M H₂SO₄ in DMSO-d₆, and 750 μL of DMSO-d₆. Isotopic tracing was carried out using K¹⁵NO₃ (99 atom% ¹⁵N) and K¹⁴NO₃ electrolytes at a concentration of 100 mM in 1 M KOH. The electrocatalytic reduction was performed for 1 h at 0.1 V vs. RHE, followed by ¹H NMR detection of ¹⁵NH₄⁺ and ¹⁴NH₄⁺ to confirm the origin of NH₃.

Paired NO₃RR-HOR systems

The flow cell tests were conducted in a home-made membrane-free electrode assembly at 25 °C. The home-made cell comprised two electrodes: an AgRu-15 electrode (0.5 × 0.5 cm², Ag mesh) for NO₃RR and a Pt/C GDE (2 × 2 cm², 20% Pt/C loading: 0.5 mg cm⁻², carbon paper) for HOR. Linear sweep voltammetry (LSV) were recorded at a scan rate of 10 mV s⁻¹ with a H₂ flow rate of 20 sccm. Long-term stability was evaluated using chronoamperometry (CA) at a current density of 200 mA cm⁻² in flowing electrolyte (1 M KOH + 100 mM KNO₃), maintained with a peristaltic pump at a continuous flow rate of 6 mL min⁻¹. H₂ was supplied to the anodic compartment at a flow rate of 10 sccm. The stability test protocol involved 9 electrolyte renewals over a period of 100 h, with each renewal occurring every 11–12 h.

Electrochemical in situ FTIR reflection analysis

In situ FTIR electrochemical thin-layer spectroscopy measurements were conducted using a Thermo Scientific Nicolet iS50 FTIR spectrometer equipped with a liquid nitrogen-cooled MCT-A detector. The catalyst sample was vertically pressed against a CaF₂ window by a glassy carbon electrode, which was utilized as the working electrode. A platinum sheet and an Hg/HgO electrode were employed as counter and reference electrodes, respectively. The sample spectra were averaged from 200 interference spectra with a resolution of 8 cm⁻¹. The reference spectrum was collected at 0.5 V vs. RHE, with sample spectra obtained from 0.5 to 0 V in 0.1 V increments. The resultant spectra were treated as absorbance, where positive peaks indicate the formation of species, and negative peaks signify the depletion of species.

For the attenuated total reflectance surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) measurements, a gold-plated Si prism with electrochemically deposited catalysts (Ag, AgRu, and Ru thin films) was assembled into a spectral-electrochemical cell, which contained a Platinum sheet as the counter electrode and Hg/HgO electrode as the reference electrode. Typically, for the preparation of Ru and Ag thin films, Ru thin film was electrodeposited on the Au film at a current density of −6 mA cm⁻² for 10 min utilizing a 25 mM RuCl₃·xH₂O + 1 M HCl solution, while silver (Ag) thin films were electrodeposited at −6 mA cm⁻² for 3 min from a 10 mM AgNO₃ + 0.1 M H₂SO₄ solution. The ATR-FTIR spectrum was acquired similarly to that of thin-layer in situ FTIR, with a reference spectrum recorded at 0.4 V vs. RHE. Sample spectra were collected within the potential region from 0.4 to −0.1 V, incremented by 0.05 V. For the in situ recording of Ru-H_atop spectra, sampling spectra were averaged from 32 interference spectra, achieving a resolution of 8 cm⁻¹.

Density functional theory (DFT) calculations

All density functional theory (DFT) calculations were performed using the Vienna Ab Initio Simulation Package (VASP) with projector-augmented wave (PAW) pseudopotentials to describe electron-ion interactions⁷⁰. The exchange-correlation energy was treated within the generalized gradient approximation (GGA) using the Perdew–Burke–Ernzerhof (PBE) functional⁷⁰. A plane-wave basis set with a cutoff energy of 400 eV was employed, and spin polarization was included in all calculations. The Ru(002) and Ag(111) surfaces were modeled using 3 × 3 supercells with four metal layers; Ag(211) surfaces were constructed with 1 × 3 supercells and eight layers. The RuAg surfaces were constructed by vertically stacking two layers of a 3 × 3 Ru(002) slab on top of two layers of a 3 × 3 Ag(111) slab. For hydrogen-covered systems (monolayer H/Ru(002) and H/RuAg), hydrogen atoms were placed at the hollow sites of the Ru surface. Metal clusters (Ru₄, Pt₄, Pd₄, Ir₄, and Rh₄) were deposited on a four-layer Ag(111) substrate to model metal-interface interactions. A vacuum layer of ≥15 Å along the z direction (normal to the surface) was introduced to avoid periodic image interactions, and dipole corrections were applied to eliminate spurious intercell interactions. During geometry optimization, the bottom two layers of all slabs were fixed, while the rest, including adsorbates, metal clusters, and the H layers, were fully relaxed. Van der Waals (vdW) interactions were accounted for using Grimme’s zero-damping DFT-D3 method⁷¹. A 3 × 3 × 1 k-point mesh sampled the surface Brillouin zone. Convergence thresholds were set to 10⁻⁵ eV for electronic self-consistency and 0.03 eV Å⁻¹ for Hellmann–Feynman forces.

The Gibbs free energy change (ΔG) for species involved in the hydrogen evolution reaction (HER) and nitrate reduction (NRA) was computed as:

$$\mathrm{\Delta }G=\mathrm{\Delta }E+\mathrm{\Delta }\mathrm{ZPE}-T\mathrm{\Delta }S$$ 9

Where ΔE is the DFT-calculated energy change, ΔZPE is the zero-point energy correction, and ΔS is the entropy change (evaluated within the harmonic oscillator approximation). All adsorbed-state energies were referenced to the free energies of H₂O(g), H₂(g), HNO₃(g), and the clean slab. For example, ΔG for adsorbed *NO was derived as:

$${\mathrm{\Delta }G}_{*NO}=E_{*NO}+{\mathrm{ZPE}}_{*NO}-{TS}_{*NO}-E_{slab}-G_{HNO3\left(g\right)}+{2G}_{H_{2O\left(g\right)}}-{\frac{3}{2}G}_{H_2\left(g\right)}+0.392$$ 10

The standard free energy change (ΔG₀ = 0.392 eV) for HNO₃(g) formation was obtained by summing the Gibbs free energy of NO₃⁻(aq) → HNO₃(l) (0.317 eV) and the vaporization energy of HNO₃(l) → HNO₃(g) (0.075 eV). These values were derived from the CRC handbook of chemistry and physics. Reaction barriers were determined using the climbing image nudged elastic band (CI-NEB) method (force cutoff: 0.05 eV Å⁻¹)⁷². Transition states (TSs) were confirmed by identifying a single imaginary vibrational frequency in their Hessian matrices. Additional computational details are outlined in Supplementary Note 1 and Supplementary Data 1–8.

Coarse-grained molecular dynamics (CGMD) simulations

The competitive adsorption of NO₃⁻, NO₂⁻, and H₂O on Ru(002)/Ag(111) and Ru(002) surfaces at varying hydrogen coverages (1/3 ML, 2/3 ML, and 1 ML) was quantitatively analyzed through coarse-grained molecular dynamics (CGMD) simulations. Detailed computational methodologies for CGMD are outlined in Supplementary Note 2 and Supplementary Data 1–8.

Author contributions

L.L., W.H. supervised this project. W.H., L.Z. concevied the project, analyzed the data and wrote the manuscript. L.Z. conducted the experiments. H.G. performed the theoretical calculations. R.L., X.L., and Y.F. participated in the discussion for the research.

Peer review

Peer review information

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

Data availability

The data supporting the findings of this study are detailed in the Supporting information/Source data file. 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/s41467-026-71299-x.