General synthesis of high-entropy single-atom nanocages for electrosynthesis of ammonia from nitrate

Abstract

Given the growing emphasis on energy efficiency, environmental sustainability, and agricultural demand, there’s a pressing need for decentralized and scalable ammonia production. Converting nitrate ions electrochemically, which are commonly found in industrial wastewater and polluted groundwater, into ammonia offers a viable approach for both wastewater treatment and ammonia production yet limited by low producibility and scalability. Here we report a versatile and scalable solution-phase synthesis of high-entropy single-atom nanocages (HESA NCs) in which Fe and other five metals-Co, Cu, Zn, Cd, and In-are isolated via cyano-bridges and coordinated with C and N, respectively. Incorporating and isolating the five metals into the matrix of Fe resulted in Fe-C₅ active sites with a minimized symmetry of lattice as well as facilitated water dissociation and thus hydrogenation process. As a result, the Fe-HESA NCs exhibited a high selectivity toward NH₃ from the electrocatalytic reduction of nitrate with a Faradaic efficiency of 93.4% while maintaining a high yield rate of 81.4 mg h⁻¹ mg⁻¹.

Introduction

The nitrogen cycle stands as a pivotal biogeochemical process on the Earth, yet the overabundance of nitrogen compounds in water poses a significant threat to ecological systems¹. Nitrate (NO₃⁻), a primary nitrogen pollutant, pervades groundwater, rivers, and lakes, highlighting the pressing need to address the imbalance in the nitrogen cycle via effectively mitigating nitrate contamination^2,3,4. Conventional nitrate removal methods, including anaerobic bacteria and physicochemical approaches, often entail substantial capital investment and maintenance costs⁵. Electrochemical reduction emerges as a promising, eco-friendly solution, offering efficient nitrate removal and ammonia synthesis with minimal installation footprint and low operational expenses^2,6. This technology powered by electricity harnesses renewable energy sources like wind and solar power to drive the process, reducing overall energy consumption^{7,8,9,10,11,12,13,14,15}. However, the simultaneous 9 proton and 8e⁻ process of electrocatalytic nitrate reduction to ammonia demands high-performance catalysts and meticulous control over reaction parameters, electrolyte concentration, and potential¹⁶, underlining the need for continued research into efficient catalyst design and process optimization to advance sustainable nitrogen management practices.

Recently, incorporation of high-entropy into manifold material categories has emerged as an effective strategy to stabilize their crystal structure and swell their chemical and structural multiplicity via enlarged configurational entropy. To date, an extensive array of high-entropy materials (HEM), starting with high-entropy alloys (HEAs)¹⁷ and high-entropy ceramics (HECs)^18,19, followed by high-entropy oxides (HEOs)^20,21 and high-entropy MXenes (HEMXs)²², have found applications across a broad range of fields, spanning electrochemical conversion, energy storage, and catalysis. The emerging methodology of advanced materials design, harnessing configurational entropy to prompt diverse interactions among incorporated elements (known as the “cocktail effect”), often imparts these materials with unparalleled properties^17,18,20,23. Integrating a considerable number of distinct elements, in usually equimolar proportions within single-phase electrocatalysts endows them with elevated configurational entropy, allowing for assorted surface interaction combinations to substantially enhance both activity and stability towards targeted reactions²⁴. However, the high compositional complexity of HEMs often frustrates the effort for manipulating their morphology and surface structures without compromising their versatility and scalability.

To maximize the atom efficiency of metals in an electrocatalyst, single-atom catalysts (SACs) has been well-developed wherein metal atoms are individually dispersed and anchored on a host material^25,26 As electrocatalytic processes require optimum number of active sites to proceed effectively, an essential challenge in designing effective SACs is striking a balance between augmenting single-atom loading and preventing atom from aggregation^27,28. In contrast to conventional SACs, a single-atom lattice (SAL) can be delivered by cyanogels in which metal atoms are atomically dispersed in a host lattice that thermodynamically stabilize isolated atom centers with tunable electronic properties¹⁵. Moreover, cyanogels-derived SAL represents a structural motif with a high density of isolated atoms in which the first nearest neighbor of a single atom can be C or N in the host lattice¹⁵. For example, as one type of cyanogels, Prussian blue analogues (PBAs) feature a SAL in which each metal atom is linked by six cyanide bonds²⁹. Owing to cyanide bridge separation, each metal atom is naturally isolated, overcoming the challenge of atom aggregation in many SACs. Furthermore, the moderate strength of coordination between the cyanide group and metal atoms endows PBAs with high morphological and compositional tunability^30,31. Specifically, hollow nanostructures of PBAs have been synthesized via templating methods in which the PBA templates were etched with different agents such as acids³², bases^33,34, and metal ions^35,36,37. In addition, several recent studies have shown that the high-entropy concept can be introduced into PBAs to significantly improve their cycling performance for reversible energy storage^38,39,40. Hence, one can envision that PBA-based cyanogels with both high-entropy SAL and well-defined hollow structure can be obtained via a templating method and represent a unique structural concept for diverse catalytic applications. Nonetheless, achieving a feasible synthesis of such a well-defined structure with precisely tailored nanoscopic architecture remains a formidable challenge.

Here we present a proof-of-concept introduction of a solution-phase synthesis of cyanogel-based high-entropy single-atom nanocages (HESA NCs) via a two-step approach with versatility and scalability. Via cyano-reaction and 3D polymerization in an aqueous solution under ambient condition, we first obtained cyanogel-based nanocubes with a medium-entropy single-atom lattice (MESA nanocubes) as templates for subsequential etching. Different from HEAs nanoparticles (NPs) in which metal atoms are connected each other via metallic bonds, the metal atoms/ions are isolated and covalently coordinated with N or C in the lattice (Fig. 1). Therefore, although both HEAs and MESA nanocubes can deliver face-centered cubic (fcc) lattice of metal atoms with high or medium entropy, their distinct nature of bonds results in differential behavior under chemical etching. Specifically, random surface etching is thermodynamically favorable for HEAs NPs, delivering arbitrary distribution of lattice vacancies and thus NPs with an irregular surface (iNPs). On the contrary, spatially selective etching of the MESA nanocubes can be achieved via confined generation of lattice vacancies, removing the internal region of nanocubes and thus resulting in HESA NCs. We systematically investigated the generation mechanism of the HESA NCs as well as their high-entropy effects on the electrocatalytic nitrate reduction reaction (NO₃RR). Our Fe-HESA NCs catalyst exhibited a high selectivity toward NH₃ from NO₃RR with a Faradaic efficiency of 93.4% and an exceedingly high yield rate of 81.4 mg h⁻¹ mg⁻¹ at −0.6 V versus the reversible hydrogen electrode (RHE). Furthermore, Fe-HESA NCs demonstrated catalytic stability over 20 consecutive electrolysis cycles. Operando synchrotron-radiation Fourier transform infrared spectroscopy (SR-FTIR) suggests that the high activity of the Fe-HESA NCs catalyst can be attributed to the boosted water dissociation at high voltage, benefiting to the hydrogenation process for the generation of NH₃.

Fig. 1: Schematic illustration of the HEA iNPs and HESA NCs synthesis.

a Colloidal synthesis of HEA iNPs starts with homogeneous nucleation of metal atoms connected by metallic bonds, followed by the growth of nuclei into NPs. After an oxidative etching process, the HEA NPs are supposed to deliver an irregular shape due to the random etching of surface atoms. b In the case of HESA NCs, the MESA nanocubes are firstly formed via ligand exchange and cyano polymerization, followed by selective etching of the internal space, generating HESA NCs.

Results

Synthesis and characterization of MESA nanocubes and HESA NCs

A typical synthesis process for PBA-based cyanogel with HESA and cubic hollow structure consists of two steps: (1) synthesis of cyanogel nanocubes, and (2) subsequent spatially selective etching with ammonia solution, as illustrated schematically in Fig. 1b. First, five d- and p-block metal cations: Co(II), Cu(II), Zn(II), Cd(II), and In(III) from metal chlorides or nitrates, were selected as metal sources and mixed with K₃[Fe(CN)₆] to fabricate MESA nanocubes by a simple room temperature ligand exchange/substitution method. The ligand exchange reactions between metal ions and Fe-cyano molecules resulted in the generation of new covalent bonds and linear chains of M–N≡C–Fe–C≡N–M (M = Co, Cu, Zn, Cd, In) as constitutional unit. After 3D polymerization, the lattice was composed of single metal atoms coordinated with six C or N atoms, in which medium-entropy single atoms with a fcc lattice were anchored and stabilized in the PBA lattice. In the synthesis process, the molar ratio between K₃[Fe(CN)₆] and all other metal salts was fixed at 1:2 to ensure the formation of cyanogel with a cubic shape. In the second step, the interior compositions could be selectively etched by breaking the bonds between M and –N≡C–Fe(III) and thus generating lattice vacancies of Mⁿ⁺ (V_M) and [Fe(III)(CN)₆]³⁻ enclosed by V_M in a controlled manner^34,35, resulting in the formation of cyanogel NCs with HESA and extraction of K⁺ ions. In addition, a redox reaction may also be involved in the etching process for the reduction of Fe(III) to Fe(II)^34,35.

The morphology of the MESA nanocubes were characterized by scanning electron microscopy (SEM) and bright-field transmission electron microscopy (BF-TEM) imaging, as shown in Supplementary Fig. 1a and Fig. 2a. The as-synthesized MESA particles exhibited a well-defined cubic shape with a slightly concave surface and particle sizes from 50 to 150 nm. The high-resolution TEM (HRTEM) garnered on the corner of a MESA nanocube shows a profile around 90° (Supplementary Fig. 2d), indicating the dominance of {100} facet exposed on the surface of the nanocube with a cubic crystal structure. It is noteworthy that the zoomed gel nanocubes always went through slight morphology and crystallinity changes under intensive electron beam due to their limited beam stability. Powder X-ray diffraction (PXRD) was performed to investigate the crystal structure and lattice parameters of MESA nanocubes. The PXRD pattern exhibits a set of characteristic diffraction peaks of PBA lattice without excessive peaks for other phases, demonstrating a single-phase PBA structure of the product (Fig. 2k)¹⁵. The slight shift of diffraction peaks to higher degree relative to FeCo PBA ( JCPDS: #46-0907) implies that the introduction of multi-metals would result in a certain degree of lattice expansion²³. In addition, the composition of MESA nanocubes were mensurated by inductively-coupled plasma mass spectrometry (ICP-MS) and energy dispersive X-ray (EDX) spectrometry, as shown in Supplementary Table 1 and Supplementary Fig. 2e, demonstrating a Fe:Co:Cu:Zn:Cd:In atomic ratio of 39:6:9:10:26:10 and thus a mixed configurational entropy ($\Delta {S}_{{{{{\rm{mix}}}}}}^{{{{{\rm{conf}}}}}}$) of 1.46R²⁶. The and high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image and corresponding element mapping analysis of an individual nanocube demonstrate the homogeneous distribution of C, N, and Fe throughout the nanocube (Fig. 2b, i), in line with the presence of –[M–N≡C–Fe–C≡N]– constitutional unit composing the fcc framework of the lattice. Remarkably, heterogeneous and localized distribution of Co, Cu, Cd, and In were observed. Specifically, the Cd and In are mainly confined in the interior region of the nanocube, while Co and Cu are preferentially located in the exterior and surface region of the nanocube (Supplementary Fig. 3). The inhomogeneous elemental distributions were also confirmed by EDX line scans (Fig. 2d), which is crucial for subsequent spatially selective etching with changes over morphology, composition, and configurational entropy of the final product. According to previous studies, the insertion of K⁺ ions leads to the coexistence of Fe(III) and Fe(II) in the lattice, resulting in differential electron donating ability of the N atom in the –N≡C–Fe and thus distinct bond strength between M and –N≡C–Fe^19,20,27. To confirm the localization of Fe(III) and Fe(II) lattice points, we conducted Fe-L-edge electron energy-loss spectroscopy (EELS) across the nanocube. As shown in Fig. 2c, the intensive EELS profiles apparently reveal a core@shell distribution of Fe(III)@Fe(II), indicating different reactivity and etching resistivity across the nanocube. In addition to the K⁺ ions induced charge, the presence of Fe(II) may also be ascribed to involvement of In(III) which can disrupt the charge balance of the system. The crystalline characteristics of the nanocubes were also investigated by selected-area electron diffraction (SAED) pattern collected using cryo-TEM at 80 kV to increase their beam stability and thus maintain their crystallinity. The hexagonal symmetry pattern was observed with clear diffraction spots of (220) and (02-2) lattice planes of PBA (Fig. 2a), which can be ascribed to [111] zone axis diffraction of fcc framework.

Fig. 2: Synthesis and characterizations of MESA nanocubes and HESA NCs.

a BF-TEM image and SAED pattern of MESA nanocubes. Scale bar is 2 1/nm. b HAADF-STEM of an individual MESA nanocube. c Fe-L-edge EELS throughout the MESA nanocube. d EDS line scans of an individual MESA nanocube. e BF-TEM image and SAED pattern of HESA NCs. Scale bar is 2 1/nm. f HAADF-STEM of an individual HESA NC. g zoom-in HAADF-STEM image and localized Fe-L-edge EELS on the corner and wall of the HESA NC, respectively. h EDS line scans of an individual HESA NC. EDX elemental maps of an individual MESA nanocube (i) and HESA NC (j). Scale bars are 50 nm for all images. k PXRD patterns of MESA nanocubes (black line) and HESA NCs (red line). Red dashed lines: FeCo PBA (JCPDS: #46-0907).

The as-obtained MESA nanocubes serve as a precursor to HESA NCs via a spatially selective etching process. Supplementary Fig. 1b and Fig. 2e show the SEM and BF-TEM images of the HESA NCs, revealing a cubic shape as well as high hollowness and enhanced surface flatness of the particles with a size range from 50 to 150 nm. The rectangular profile observed using HRTEM at the corner of the NC demonstrate a well-defined cubic hollow structure with the preservation of {100} facet on the surface and a wall thickness around 12 nm (Supplementary Fig. 2i). PXRD pattern of HESA NCs shows diminished crystallinity with reduced number of characteristic diffraction peaks, indicating the increased symmetry of the lattice (Fig. 2k). The diffraction peaks shifted to higher degree relative to those of MESA nanocubes, suggesting a lattice contraction after etching due to the extraction of K⁺ ions from the nanocubes (Supplementary Table 1)^24,25. The composition of HESA NCs was also confirmed ICP-MS and EDX spectrometry, (Supplementary Table 1 and Supplementary Fig. 2j), indicating a Fe:Co:Cu:Zn:Cd:In atomic ratio of 29:17:15:10:18:11 and thus a $\Delta {S}_{{{{{\rm{mix}}}}}}^{{{{{\rm{conf}}}}}}$ of 1.59R. The increase of compositional equality and thus $\Delta {S}_{{{{{\rm{mix}}}}}}^{{{{{\rm{conf}}}}}}$ from 1.46R to 1.59R resulted in the change from MESAL (∆S < 1.5R) to HESAL (∆S > 1.5R) with an increased thermodynamic stability (Supplementary Fig. 4). The presence and homogeneous distribution of all elements were also verified by HAADF-STEM image and corresponding element mapping and line scan analysis of an individual HESA NC (Fig. 2h, j). Fe-L-EELS profile collected on the wall of the NC indicates the dominance of Fe(II) lattice points, as shown in Fig. 2g. Figure 2e shows the SAED pattern of the HESA NCs along [111] zone axis, demonstrating a lower crystallinity with a similar hexagonal symmetry pattern including diffraction spots of (220) and (02-2) lattice planes of PBA but additional diffraction spots and slight amorphous characteristics.

The local coordination environment and chemical state of different metal elements before and after the etching process were studied by X-ray photoelectron spectroscopy (XPS) and transmission X-ray absorption fine structure (XAFS). The core-level Fe 2p spectra validate that there are corresponding divalent and trivalent species of Fe in both MESA nanocubes and HESA NCs (Fig. 3a)³⁸. Specifically, the fitted spectra of the Fe 2p region exhibit two sets of peak doublets at binding energies of 708.5 (Fe 2p_3/2)/721.6 (Fe 2p_1/2) eV and 710.0 (Fe 2p_3/2)/723.5 (Fe 2p_1/2) eV for Fe²⁺ and Fe³⁺, respectively. Notably, the decreased peak area ratio of Fe³⁺/Fe²⁺ after etching indicates the reduced amount of Fe³⁺, which is in agreement with Fe-L-EELS analysis. X-ray absorption near edge structure (XANES) spectra at the Fe K-edge for MESA nanocubes (before etching) and HESA NCs (after etching) are shown in Fig. 3b, with Fe foil, FeO and Fe₂O₃ as references. The curve tendencies of gel samples demonstrate that these materials feature different near-edge absorptions from those of Fe foil, FeO and Fe₂O₃. The XANES profiles before and after etching are nearly identical, indicating that the etching process barely changed the coordination environment adopted by Fe moieties in gel samples. Remarkably, the absorption edge position of Fe has apparently shifted to lower energy level after etching, which implies that the etching process resulted in the decrease of oxidation state caused by the preferential removal of Fe(III) from the lattice, in line with the Fe-L-EELS result⁴¹. XANES spectra of other doped metals indicate similar valence states originating from their metal precursors (Supplementary Fig. 5). The Fe, Co, Cu, Zn, and In K-edge k²-weighted Fourier transformed (FT) extended X-ray absorption fine structure (EXAFS) spectra for gel samples before and after etching demonstrates evident Fe–C, Co–N, Cu–N, Zn–N, Cd–N, and In–N (Fig. 3c and Supplementary Fig. 6) scattering path, further signifying that Fe, Co, Cu, Zn, Cd, and In are successfully introduced and atomically dispersed into the PBA framework^38,41. The Fe, Co, Cu, Zn, Cd, and In K-edge XANES and EXAFS spectra indicate that both MESA nanocubes and HESA NCs feature similar near-edge absorptions, which indicates isolated metal atoms states coordinated with C or N in both MESA nanocubes and HESA NCs.

Fig. 3: Structural analysis and formation mechanistic study of HESA NCs.

a XPS spectra of Fe 2p of gel samples before and after the etching process. b Fe K-edge XANES spectra of Fe foil, FeO, Fe₂O₃, and gel samples before and after the etching process. c FT k²-weighted χ(k) function of EXAFS spectra for the Fe K-edge. d Atomic percentage of metal elements and corresponding ∆S after etching MESA nanocubes by ammonia with concentration of 0 (E0), 2 (E1), 5 (E2), 10 (E3), 15 (E4), and 20 (E5) mmol. ATR-FTIR localized spectra (e) and PXRD patterns (f) of the etched products. g Schematic illustration of the lattice evolution during the etching process with localized dissolution of [Fe(III)(CN)₆]³⁻ and increasing ∆S via the ligand substitution mechanism.

The XPS was also employed to unravel the surface chemical compositions and valence state of different elements of cyanogels before and after etching. (Supplementary Fig. 7). The Co 2p spectra verify the coexistence of divalent and trivalent species of Co, while the Cu 2p spectra demonstrate the presence of monovalent and divalent Cu species in both MESA nanocubes and HESA NCs^40,42. In addition, the Zn 2p, Cd 3d, and In 3d spectra validate that the Zn, Cd, In cations mainly exist in the form of Zn²⁺, Cd²⁺, and In⁺, respectively^38,40,43. Specifically, the fitted Co 2p spectra show two sets of peak doublets at binding energies of 782.1 (Co 2p_3/2)/797.4 (Co 2p_1/2) eV and 784.2 (Co 2p_3/2)/799.4 (Co 2p_1/2) eV for Co³⁺ and Co²⁺, respectively, while no substantial changes of peak area ratio of Co³⁺/Co²⁺ were observed after etching (Supplementary Fig. 7b). The fitted spectra of the Cu 2p_3/2 region were found to display peaks at 931.6 and 934.7 eV that can be assigned to Cu⁺ and Cu²⁺, respectively, together with their shake-up satellites (Supplementary Fig. 7c)^38,40,42. The increase of Cu⁺ after etching signifies an internal redox reaction from Cu²⁺ to Cu⁺ occurred during the etching process. A major Zn 2p peak doublet was detected at 1022.0 eV (Zn 2p_3/2) and 1045.1 eV (Zn 2p_1/2), revealing the existence of Zn²⁺ before and after etching (Supplementary Fig. 7d). A single Cd 3d peak doublet at 404.8 eV (Cd 3d_5/2) and 411.5 eV (Cd 3d_3/2) (Supplementary Fig. 7e). These binding energies are typical of Cd in its preferred valence state of 2+⁴⁰. The In 3d eak located at 445.5 eV (In 2p_5/2)/453.1 (In 2p_3/2) eV and 446.8 eV (In 2p_5/2)/454.4 (In 2p_3/2) eV can be assigned to In⁺ and In³⁺, respectively, verifying the dominance of In⁺ in both gel samples and suggesting an internal redox reaction from In³⁺ to In⁺ in the formation of MESA nanocubes (Supplementary Fig. 7f). The influence of etching on the surface valence band of the cyanogels was also studied, revealing the shift of the valence band away from the Fermi level (E_F) after the etching process (Supplementary Fig. 8). The chemical structure change of the cyanogels caused by etching was also investigated by Fourier transform infrared (FTIR) spectroscopy analysis. As shown in Supplementary Fig. 9a, the sharp peak at 1608 cm⁻¹ and the broadband at ~3400 cm⁻¹ in both samples originate from the O–H bending vibration and H–O–H stretching vibration, respectively, of water molecules existing in the samples⁴⁴. The typical C≡N group stretching vibration was detected from the adsorption bands in the range of 2000–2200 cm^−1 45. Specifically, the v (C≡N) peak at 2155 cm⁻¹ (I) for Fe(III)–C≡N–M exists in the MESA nanocubes but disappears in the HESA NCs, whereas the v (C≡N) peak at 2100 cm⁻¹ (II) for Fe(II)–C≡N–M is present in both samples (Supplementary Fig. 9b)⁴⁵. In addition, an additional v (N-H) peak at 1414 cm⁻¹ of NH₄⁺ was detected after etching³⁴. These FTIR analysis can be summarized into three major changes of chemical structure during etching: (1) the water molecules adsorbed in the lattice of the cyanogels were partially removed by ammonia solution; (2) the linear chains of M–N≡C–Fe(III)–C≡N–M were substantially eliminated, which is attributed to the preferential bond breaking between M and –N≡C–Fe(III) in the internal region of the nanocubes; (3) occurrence of a redox reaction during the ammonia treatment for the reduction of Fe(III) to Fe(II) and generation of NH₄^+ 34,35.

To further evaluate the versatility of the synthetic strategy of HESA NCs shown in Fig. 1b, we designed more control experiments and found that the hollowness of NCs can be governed by simply altering the feed ratio of metal precursors in the reaction mixture (see Supplementary Table 2 for details), resulting in NCs with wall-thickness of 40, 35, 12, and 6 nm after etching of gel nanocubes with different elemental compositions (Supplementary Fig. 10). In addition, it was found that variation of feed ratio of metal precursors (see Supplementary Table 3 for details) could also lead to gel nanocubes with lower surface flatness, which could serve as a precursor to NCs with enlarged cavities on the walls (Supplementary Fig. 11). The tunable hollowness and openness of HESA NCs with well-controlled parameters validate the versatility of the synthetic strategy to produce gel nanostructures with controlled parameters. Furthermore, to demonstrate the potential synthetic scalability, we attempted our synthesis up to 100 times just by proportionally increasing the volumes of reaction solutions. Noticeably, the stirring rate was increased from 400 rpm to 600 rpm to promote mass transfer and thus uniform dispersal of the metal precursors to enable homogeneous nucleation via the ligand exchange cyano reactions. Supplementary Fig. 12 shows SEM and BF-TEM images of the HESA NCs synthesized using the scale-up protocols in which the NCs feature a well-defined cubic shape and hollowness with a size in the range of 50–150 nm, consistent with that of the product from a standard synthesis. According to the ICP-MS analysis and EDX spectra shown in Supplementary Fig. 13 and Table 3, the HESA NCs delivered a Fe:Co:Cu:Zn:Cd:In atomic ratio of 27:16:17:9:19:12 close to the standard product. These results confirm the scalability of the synthetic strategy for HESA NCs.

Mechanistic understanding of the spatially selective etching

To fundamentally understand the formation mechanism of the HESA NCs via spatially selective etching, systematic characterizations of intermediates at different reaction stages were performed. The functional ions derived from ammonia solution for etching were firstly figured out by replacing ammonia solution with the same amount of KOH or NH₄Cl as etchants, respectively, while other experimental parameters were kept the same. As shown in Supplementary Fig. 14, the barely changed products after etched by KOH eliminates the possibility of OH⁻ ions as the etchant, while the obvious hollowness of NCs etched by NH₄Cl demonstrates the key role played by NH₃ or NH₄⁺ ions during the etching process. To trace and quantify the changes of MESA nanocubes in terms of morphology, composition, chemical structure, and lattice parameters during etching, we designed a series of ex-situ experiments to simulate the time-dependent in-situ characterizations. Since the etching-induced changes occurred instantly upon contacted with ammonia, typically within a few seconds, it was almost impossible to conduct ex-situ time-dependent experiments to collect corresponding intermediates. As such, we attempted to collect the “pseudo time-dependent products” by using ammonia solutions with a range of reduced concentrations (Supplementary Table 4). Using this method, we have successfully captured intermediate samples (denoted as E0 to E5) during the fast etching process. Supplementary Fig. 15a–f shows the structure and morphology of the intermediate products collected at different reaction stages. As shown in Supplementary Fig. 15b, initial reaction with ammonia resulted in the generation of a yolk@shell structure likely due to the contact of outmost lattice points of Mⁿ⁺ coordinated with –N≡C–Fe(III) in the core region with ammonia. As the reaction continues, the internal region of the nanocube became visibly void, generating a double shell structure (Supplementary Fig. 15c). This special structure might be caused by the redox reaction between terminal [Fe(III)(CN)₆] and ammonia, resulting in newly generated M–N≡C–Fe(II) with a higher etching resistibility inside the nanocube^34,35. As the etching proceeded, the inner shell was completely removed while the outer shell with a larger thickness and high internal surface roughness was observed (Supplementary Fig. 15d), which was considered as the involvement of dissolution and recrystallization processes of Fe(II)/Fe(III) lattice points³³. Finally, a hollow nanobox with flat internal and external surfaces was obtained with increased extent of etching (Supplementary Fig. 15e). Further etching of the nanobox increased the porosity of the walls, producing a nanocage with observable cavities on the walls (Supplementary Fig. 15f). This result indicates the existence of M–N≡C–Fe(III) on the walls of the nanobox which could be etched away after sufficient etching time or under a robust etching condition. The summarized schematic illustration of the spatially selective etching process of intermediates are shown in Supplementary Fig. 15g.

The compositional evolution during etching was evaluated by ICP-MS and EDX spectra of intermediate samples. As shown in Supplementary Fig. 16, different variation tendencies of elements can be distinguished when normalized to the amount of Fe in the samples. For example, increased amount of Cd is exhibited from E0 to E3 followed by a decrease from E3 to E5, while a continuous increase of Co is shown from E0 to E5. The quantified change of atomic percentage of each element obtained by ICP-MS analysis is shown in Fig. 3d, demonstrating the substantial decrease in Fe and Cd from 39% and 26% to 29% and 18%, respectively. Meanwhile, an apparent increase in Co from 6% to 17% was also observed. This difference can be attributed to the heterogeneous distribution of Fe, Co, and Cd in the nanocubes, in which Fe(III) and Cd are confined in the inner core while Co is mainly located in the outer shell. As such, the spatially selective etching process preferentially eliminated a greater number of elements that distributed in the interior region of the nanocubes with a higher atomic percentage, resulting in nanocages with improved compositional equality and thus elevated $\Delta {S}_{{{{{\rm{mix}}}}}}^{{{{{\rm{conf}}}}}}$. In addition to the compositional evolution, the chemical structure info was further studied by FTIR (Supplementary Fig. 17). As shown in Fig. 3e, the peak at 2155 cm⁻¹ disappeared from E0 to E1, indicating the fast chains scission of M–N≡C–Fe(III)–C≡N–M and transformation of lattice points from Fe(III) to Fe(II) via redox reaction upon etching. The emergence of v (N-H) peak for NH₄⁺ at 1414 cm⁻¹ from E0 to E1 (Supplementary Fig. 17) further verify the occurrence of the redox reaction. Noticeably, there is a shift of the stretching v (C≡N) band at 2100 cm⁻¹ to a lower wavenumber of 2089 cm⁻¹. During the etching process, simultaneous dissolution and recrystallization reactions inside the nanocubes were anticipated to prompt dynamic changes of compositions on the inner surface of the shell, resulting in the change of metal ions linked to the C≡N bonds. The stretching v (C≡N) band gradually shifted back to the original location from E1 to E5, suggesting the discontinued variation of metal ions linked to the C≡N bonds and the termination of the etching process.

To further investigate the crystal structure evolution throughout the etching process, PXRD patterns of intermediate samples were collected (Fig. 3f). From E0 to E3, the diffraction peaks of (111) and (200) were shifted to higher degree, indicating lattice contraction and internal lattice strain due to the extraction of K⁺ ions^39,40. In addition, the (220) peak split into two peaks, together with new peaks present around 20° and 30°, indicating reduced lattice symmetry in the initial stage of etching. In the subsequent etching process as represented from E3 to E5, the peak merging of (220) and the gradual disappearance of characteristic peaks demonstrate the increased symmetry of the lattice^39,40,45. On the basis of the above observations, we propose an etching mechanism based on a ligand substitution reaction and a scheme of lattice evolution pathway towards the HESA NCs. As illustrated in Supplementary Fig. 4, the presence of NH₃ could act as coordination ligands with M_6/n[Fe(III)(CN)₆]₂ for the generation of [M(NH₃)₆]ⁿ⁺ and evacuation of gel nanocubes^34,35. The atomic level illustration of the etching mechanism is shown in Fig. 3g, in which the generation of lattice vacancies can be divided into three major steps: (1) rapid dissolution of Mⁿ⁺ derived from M–N≡C–Fe(III) via ligand exchange; (2) release of K⁺ ions from the lattice; (3) generation of [Fe(III)(CN)₆]³⁻ vacancies in the center of the lattice and occurrence of redox reaction producing Fe(II).

Catalytic performance and mechanism insights into NO₃RR over Fe-HESA NCs

It is recently recognized that the high-entropy-based electrocatalysts may be advantageous in multistep reactions requiring multiple adsorption sides due to the large combination of elements and other entropy-induced effects benefiting the electron transfer process⁴⁶. As such, we chose NO₃RR as a model reaction and evaluated performance of HESA NCs by loading them onto carbon black (Vulcan XC-72R) (Supplementary Fig. 18). As a comparison, Fe-Co gel NCs with a size of ~100 nm were synthesized using the same method except for the involvement of only Co precursor for ligand exchange (Supplementary Figs. 19–21). Since both gel catalysts are composed of Fe-based PBA frameworks, we rename them as Fe-HESA NCs and Fe-Co NCs, respectively, to emphasize their similarity and difference hereinafter. The polarization curves for NO₃RR were acquired in a gas-tight H-cell at room temperature in a 0.5 M Na₂SO₄ solution containing 100 mM NaNO₃ (Supplementary Fig. 22). As shown in Fig. 4a, the Fe-HESA NCs exhibited an onset potential of −0.08 V versus reversible hydrogen electrode (RHE), much more positive than that of Fe-Co NCs (−0.23 V) as a widely accepted descriptor of promoted reaction kinetics. In the absence of NO₃⁻, the barely observed current density of polarization curve reveals a substantially diminished hydrogen evolution reaction (HER) activity of Fe-HESA NCs. In contrast, Fe-HESA NCs delivered a remarkable enhancement in current density (90.4 mA cm⁻² at −0.8 V) in the presence of NO₃⁻, while the increase of current density on Fe-Co NCs was very limited (16.8 mA cm⁻² at −0.8 V). The mass activity (MA) obtained by the partial current density of NH₃ normalized by the mass of Fe on Fe-HESA NCs was also higher than that on Fe-Co NCs over a range of potentials (Fig. 4b). Additionally, the corresponding turnover frequency (TOF) for NO₃⁻-to-NH₃ conversion per Fe site directly indicates at least 5 times enhancement of rate on individual Fe sites of Fe-HESA NCs relative to that of Fe-Co NCs, further distinguishing their intrinsic activities towards NO₃RR.

Fig. 4: Electrocatalytic NO₃RR performance.

a Linear scan voltammetry curves of Fe-HESA NCs and Fe-Co NCs normalized to the geometric area. b MA and TOF for NH₃ production at various potentials. c FE of NH₃ over Fe-HESA NCs (blank pattern) and Fe-Co NCs (slash pattern) at different potentials. d YR of NH₃ over Fe-HESA NCs and Fe-Co NCs at different potentials. All potentials are not iR corrected. e ¹H NMR spectrum of the products generated during the electrocatalytic NO₃RR over Fe-HESA NCs in 0.1 M Na¹⁵NO₃ or 0.1 M Na¹⁴NO₃ at −0.6 V_RHE. f Quantification of NH₃ via UV-Vis and NMR measurements at −0.6 V vs. RHE. g Long-term chronoamperometry test for 150 h and the cycling test (inset) at −0.6 V_RHE. Error bars denote the standard deviations calculated from three independent measurements.

Chronoamperometry measurements of the catalysts were performed at diverse potentials for subsequential qualification of products (Supplementary Fig. 23). The indophenol blue spectrophotometric method and the Griess test were used to quantify the produced NH₃ and NO₂⁻, respectively (Supplementary Fig. 24). Faradaic efficiency (FE) and yield rate (YR) of NH₃ over the catalysts are shown in Fig. 4c, d, respectively. The Fe-HESA NCs showed high selectivity towards NH₃ from the NO₃RR with a FE of 93.4% and a YR of 81.4 mg h⁻¹ mg⁻¹ at −0.6 V versus RHE. As a main byproduct of the NO₃RR on Fe-HESA NCs, the FE of NO₂⁻ was over 25% at −0.3 V versus RHE while gradually reduced with increased potentials (Supplementary Fig. 25). As a comparison, Fe-Co NCs displayed FE and YR of NH₃ lower than those of Fe-HESA NCs over different potentials, indicating that the enhanced electrocatalytic performance of NO₃RR is potentially originated from the HESA. Meanwhile, no H₂ can be detected at considered potentials on both Fe-HESA and Fe-Co NCs (Supplementary Fig. 26), indicating the complete inhibition of HER over the catalysts. Control experiments were carried out at −0.6 V versus RHE in 0.5 M Na₂SO₄ solution without NaNO₃. As shown in Supplementary Figs. 27a, b and 28a, b, NH₃ was barely detected in the electrolyte, excluding the N contaminants from other sources including the catalyst, electrolyte, and environment. ¹⁵N isotope labeling experiments were conducted to further verify that the produced NH₃ was resulting from the feeding NO₃⁻ electrolyte (Supplementary Figs. 27c, d and 28c, d). After electrolysis at −0.6 V versus RHE, triple coupling and doublet peaks corresponding to ¹⁴NH₄⁺ and ¹⁵NH₄⁺ were detected in the ¹H NMR spectra of the electrolytes containing ¹⁴NO₃⁻ and ¹⁵NO₃⁻, respectively (Fig. 4e), proving that the produced NH₃ was derived from the NO₃RR. The FE and YR were very close to those determined by colorimetric methods (Fig. 4f), while the decreased current density, FE, and YR imply a potentially negative effect of ¹⁵N on NH₃ production.

The outstanding NO₃RR performance of Fe-HESA NCs underline the consequence of HESA. To clearly demonstrate this, we compared the NO₃RR performance of Fe-HESA NCs, Fe-Co NCs, and other state-of-art Fe-based catalysts. As summarized in Supplementary Table 5, the Fe-HESA NCs delivered a superiority relative to other catalysts. This demonstrates the key role of HESA in boosting NO₃RR performance, endowing Fe-HESA NCs as one of the best reported neutral NO₃RR electrocatalysts (Supplementary Table 6). The stability of NO₃RR performance was also evaluated on Fe-HESA NCs by conducting successive 20 electrolysis cycles at a fixed potential (Supplementary Figs. 29 and 30). As shown in Fig. 4g, the FE and YR of NH₃ exhibited slight fluctuations around 90.8% and 84.1 mg h⁻¹ mg⁻¹, respectively, over the consecutive electrolysis cycles. After the stability test, the catalysts were detached from the glassy carbon electrode by sonication and characterized by BF-TEM, XRD, and XPS. As shown in Supplementary Fig. 31, both the cubic and hollow morphologies were preserved after the long-term stability test. Besides, the crystal structure of Fe-HESA NCs was also well-preserved (Supplementary Fig. 32). Moreover, XPS analysis demonstrates the absence of Fe³⁺ while no apparent alteration was observed in the core levels spectra of other elements after the stability test, suggesting that the Fe sites were electrochemically reduced during NO₃RR as active sites while no chemical states changed for other metals after the stability test (Supplementary Fig. 33). The long-term stability of the catalyst was investigated by a chronoamperometry test for 150 h at −0.6 V versus RHE. As demonstrated in Fig. 4g, the current density went through a slight decrease in the first 22 h, after which it became quite stable over days, demonstrating the reasonable current stability realized by the catalyst.

To investigate the reaction process and mechanism of NO₃RR on the Fe-HESA NCs, Operando synchrotron-radiation Fourier transform infrared spectroscopy (SR-FTIR) was conducted (Supplementary Fig. 34). As shown in Fig. 5a, b, infrared signals in the range of 4000–2800 cm⁻¹ and 2200–1000 cm⁻¹ under applied potential between −0.1 and −0.6 V versus RHE were collected and analyzed. In the SR-FTIR operation results of 3450–3150 cm⁻¹ (Fig. 5a, b), the IR intensity corresponding to v (N-H) of NH_x species increased as the voltage decreased from −0.1 to −0.6 V, indicating an increase in NH₃ yield⁴⁷. Within the range of 2200–1000 cm⁻¹, peaks observed at 1229 and 1118 cm⁻¹ were ascribed to the NO₂⁻ and *NH₂OH intermediates, respectively^48,49. It is apparent that both the production rates of NO₂⁻ and *NH₂OH have nearly linear correlations with the applied potential decreased from −0.1 to −0.5 V versus RHE (Fig. 5b). In addition, the Co-H_ad and H_ad signals were also observed at 2110 and 2060 cm⁻¹, respectively, starting from −0.2 V versus RHE⁵⁰. As shown in Fig. 5c, the fast increase of H_ad signals indicates the enhanced water dissociation at higher voltage, which is beneficial to the hydrogenation process of NO₃RR for the generation of NH₃.

Fig. 5: Mechanistic investigation of the NO₃RR on HESA NCs.

Operando SR-FTIR signals in the range of 4000–2800 cm⁻¹ (a) and 2200–1000 cm⁻¹ (b). c IR peak intensity versus the applied potential for the NO₃RR process. d Conceptional illustration of the effects of configurational entropy and lattice symmetry on electronic property. The electron localization increases with increasing entropy and decreasing symmetry.

Intriguingly, a previously developed survey of the symmetry evolution in carbon catalysts indicates a negative correlation between electrocatalytic activity and basal-plane symmetry⁴¹. This trend extends from the highest symmetry D_6h observed in pristine graphene layers to D_2h of nitrogen-doped single atom materials and C_s of high-entropy single atom materials. This phenomenon implies a positive correlation between catalyst activity and symmetry reduction, which helps to explain the origin of the catalytic activity of the HESA from the perspective of lattice symmetry and electron localization. As summarized in Fig. 5d, the electron localization is supposed to generate at carbon and nitrogen atoms due to the coordination bond of Fe-C₅ and M–N₅, respectively, and increase with symmetry reduction of lattice⁴¹. Compared with unitary and binary SA with a lattice symmetry of D_3h, HESA displays a symmetry minimization in the system, and the spatial distribution of electrons is more localized and chaotic for delivering favorable electron transfer properties during electrocatalysis of NO₃RR. And the maximized ∆S of HESA is anticipated to feature the highest entropic stabilization induced by incorporation of a high number of elements. This thermodynamically benefits the HESA phase and thus leads to additional enhancement of electrochemical stability of Fe-HESA NCs under the long-term stability test of NO₃RR.

Discussion

We have developed the scalable solution-processed HESA NCs as an effective electrocatalyst for ammonia synthesis from nitrate. The key to the synthetic strategy is the introduction of heterogeneous distributions of metallic elements into the lattice of the cyanogel nanocubes in which metal ions coordinate with –N≡C–Fe(III) and –N≡C–Fe(II) in the core and shell, respectively. Upon reacting with ammonia, the bonding between metal ions and –N≡C–Fe(III) can be readily broken via a ligand exchange reaction, resulting in a spatially selective etching process for the transformation of MESA nanocubes to HESA NCs with increased ∆S. Further, we utilized our Fe-HESA NCs as a model high-entropy catalytic platform for NO₃RR to illuminate their structure-property-performance relationships via benchmarking with the Fe-Co NCs counterparts. The entropy-induced symmetry reduction of single-atom lattice combined with the enhanced water dissociation and hydrogenation process of NO₃RR leads to high activity, selectivity, and durability of the catalyst, delivering both high FE (93.4%) and YR (81.4 mg h⁻¹ mg⁻¹) of NH₃ over continuous 20 cycles of electrolysis. The findings of this study provide an insightful guideline for fabricating scalable yet controllable hollow nanostructures with single-atom lattice and high entropy for sustainable electrocatalytic reactions.

Methods

Chemicals and materials

All the chemical reagents were used as received from Sigma-Aldrich unless specified. These include potassium hexacyanoferrate(III) (K₃Fe(CN)₆, ≥99.0%), cobalt nitrate hexahydrate (Co(NO₃)₂ · 6H₂O ≥ 98%), copper chloride dihydrate (CuCl₂ · 2H₂O, 99%), zinc chloride (ZnCl₂, ≥97%), cadmium chloride (CdCl₂, 99.99%), indium chloride (InCl₃, 98%), trisodium citrate dihydrate (Na₃C₆H₅O₇·2H₂O, ≥99.0%), ammonia hydroxide (NH₄OH, 28–30%), sodium sulfate (Na₂SO₄, ≥99.0%), sodium nitrate (NaNO₃, ≥99.0%). All aqueous solutions were prepared using deionized (DI) water with a resistivity of 18.2 MΩ cm at room temperature.

Synthesis of MESA nanocubes

Topically, 3.3 mM of Co(NO₃)₂ · 6H₂O, 3.3 mM of CuCl₂ · 2H₂O, 10.2 mM of ZnCl₂, 10.2 mM of CdCl₂, 13.0 mM of InCl₃, and 40 mM of Na₃C₆H₅O₇·2H₂O were dissolved in 2.5 mL DI water, denoted as solution A. Solution B was prepared by dissolving 20 mM of K₃Fe(CN)₆ in 2.5 mL DI water. Subsequentially, solution A was rapidly injected into solution B at a stirring rate of 400 rpm, and the mixture was continuously stirring for 15 min. After reaction the product was precipitated by ethanol and collected via centrifugation, followed by washing twice with isopropanol and water. The product was then re-dispersed in 2.5 mL ethanol for further use.

Synthesis of HESA NCs

In a standard synthesis, 2.5 mL NH₄OH (28–30%) solution was firstly diluted with 10 mL DI water. A suspension of nanocubes dispersed in 2.5 mL ethanol was rapidly injected into 6.25 mL of diluted NH₄OH solution at a stirring rate of 500 rpm, and the mixture was left to stir for 15 min. After that, the etching solution was centrifuged and washed twice with ethanol and water to remove excessive NH₄OH, followed by drying in an oven at 60 °C for further use.

Scaling up the synthesis by 100 times

In a typical synthesis, solution A was prepared by dissolving 3.3 mM of Co(NO₃)₂ · 6H₂O, 3.3 mM of CuCl₂, 10.2 mM of ZnCl₂, 10.2 mM of CdCl₂, 13.0 mM of InCl₃, and 40 mM of Na₃C₆H₅O₇·2H₂O were dissolved in 250 mL DI water. Solution B was prepared by dissolving 20 mM of K₃Fe(CN)₆ in 250 mL DI water. Subsequentially, solution A was rapidly injected into solution B at a stirring rate of 600 rpm, and the mixture was continuously stirring for 15 min. After reaction the product was precipitated by ethanol and collected via centrifugation, followed by washing twice with isopropanol and water. After that, the product was re-dispersed in 250 mL ethanol. The suspension was then rapidly injected into 625 mL of diluted NH₄OH solution at a stirring rate of 1000 rpm, and the mixture was left to stir for 15 min. Finally, the etching solution was centrifuged and washed twice with ethanol and water to remove excessive NH₄OH, followed by drying in an oven at 60 °C for further use.

Calculation of the entropy (∆S) of a material system

In general, the entropy of a material system (ΔS_mix) includes four contributions: configurational entropy (${\Delta S}_{{{{{\rm{mix}}}}}}^{{{{{\rm{conf}}}}}}$), vibration entropy (${\Delta S}_{{{{{\rm{mix}}}}}}^{{{{{\rm{vib}}}}}}$), electronic entropy (${\Delta S}_{{{{{\rm{mix}}}}}}^{{{{{\rm{elec}}}}}}$), and magnetic entropy (${\Delta S}_{{{{{\rm{mix}}}}}}^{{{{{\rm{mag}}}}}}$). Their quantitative relationship can be expressed as:

$${\Delta S}_{{{{{\rm{mix}}}}}}={\Delta S}_{{{{{\rm{mix}}}}}}^{{{{{\rm{conf}}}}}}+{\Delta S}_{{{{{\rm{mix}}}}}}^{{{{{\rm{vib}}}}}}+{\Delta S}_{{{{{\rm{mix}}}}}}^{{{{{\rm{elec}}}}}}+{\Delta S}_{{{{{\rm{mix}}}}}}^{{{{{\rm{mag}}}}}}$$

(1)

In a solid solution, the configurational entropy contributes predominately to the entropy of the system⁵¹. Therefore, we use configurational entropy to denote the mixing entropy of the system, to avoid complex calculations or simulations, which can be calculated using⁵²:

$${\Delta S}_{{{{{\rm{mix}}}}}}={\Delta S}_{{{{{\rm{mix}}}}}}^{{{{{\rm{conf}}}}}}=-R\left[{c}_{i}{ln}{c}_{i}+\ldots+{c}_{n}{ln}{c}_{n}\right]=-R{\sum}_{i=1}^{n}{c}_{i}{ln}{c}_{i}$$

(2)

where R represents the gas constant, c_i is the mole percentage of the ith component, and n is the number of elements. Noting that since the lattice site of Fe is fixed in the lattice of HESA, it is considered that Fe does not change the configuration of the system, and thus it is excluded in the calculation of ${\Delta S}_{{{{{\rm{mix}}}}}}^{{{{{\rm{conf}}}}}}$.

Preparation of carbon-supported HESA NCs

Five mg of dried HESA NCs was dispersed in 10 mL DI water and 5 mg carbon black was dispersed in 10 mL ethanol, both of which were sonicated for 20 min. The HESA NCs suspension was then added into carbon black suspension under sonication. The mixture was sonicated for another 1 h before centrifugation. Finally, the mixture was dried in an oven at 60 °C for 30 min.

Characterizations

BF-TEM images were taken using a Hitachi S5500 microscope operated at 30 kV by drop casting and drying the dispersions on carbon-coated copper grids under ambient conditions. HRTEM, HAADF-STEM, and EDX mapping analyses were performed using a JEOL NEOARM probe-corrected transmission electron microscope with aberration correction. The metal contents in the as-obtained catalysts were determined using a THERMO VG PQ ExCell quadrupole-based ICP-MS. For obtaining accurate metal compositions, specimens, which were diluted by dissolving powder samples in 2% HNO₃, were collected with an ion concentration below 200 ppb. XRD patterns were obtained using a Rigaku MiniFlex 600 Diffractometer operating at 40 mA and 45 kV using Cu Kα1 radiation (λ = 1.541 Å) with 5° min⁻¹ from 5 to 80°. XPS measurements were carried out on a VersaProbe4 X-ray Photoelectron Spectrometer. An Infrared Spectrometer-Infinity Gold FTIR was used in the FTIR test. Transmission XAS was performed partially using a laboratory Rowland-circle based instrument (easyXAFS300+, easy XAFS LLC, Renton, WA, USA) and partially at 7BM at National Synchrotron Light Source II (NSLS-II). The XAS data were analyzed using the software package Athena.

Electrode preparation

To prepare the working electrode, the suspension of catalysts (1 mg mL⁻¹ Fe) was obtained by dispersing the catalyst in a mixture of 490 μL DI water, 490 μL isopropanol, and 20 μL Nafion solution (5 wt%) under ultrasonication for 30 min. The concentration of Fe in the final dispersion was determined by ICP-MS. The suspension was deposited on a pre-cleaned glassy carbon electrode (ALS Co., 1.5 cm × 1 cm) and allowed to dry under ambient conditions to achieve a Fe mass loading of ~50 ug cm^-2.

Electrochemical measurements

Electrochemical measurements were carried out on a BioLogic potentiostat with VMP-3 model. All measurements were performed in a gas-tight H-cell using a three-electrode system and a bipolar membrane at room temperature. Reversible hydrogen electrode (RHE, Mini-HydroFlex, Gasketel) served as a reference electrode and a platinum wire were used as the reference and counter electrodes, respectively. Before testing, the Nafion 212 membrane (size: 3 cm × 3 cm, thickness: 50 μm) was pretreated by sequentially heating it in 5% H₂O₂ aqueous solution at 80 °C (1 h) and deionized water at 80 °C (1 h), followed by treatment in 0.5 M H₂SO₄ (1 h) and deionized water (1 h). Before each recording, the working electrode was cycled at a scanning rate of 50 mV s⁻¹ in the potential range for linear sweep voltammetry (LSV) measurements to clean and stabilize the catalyst surface until reaching a steady state. Current densities were normalized to the geometric area of the working electrode (~0.5 cm²). For electrocatalytic NO₃RR, cathode compartment was filled with 45 mL solution of 100 mM NaNO₃ and 0.5 M Na₂SO₄ (pH = 6.93 ± 0.1), while anode compartment was filled with 45 mL solution of 0.5 M Na₂SO₄ (pH = 7.02 ± 0.1). Before the electrochemical measurement, the catholyte was purged with Ar for at least 30 min. The LSV curves were obtained by scanning the potential from 0.5 to −0.8 V_RHE at a rate of 10 mV s⁻¹. Chronoamperometry measurements were conducted at various potentials with an Ar flow rate of 20 sccm in the cathodic compartment.

Quantification of NO₂ ⁻ concentration using the Griess test

The concentration of NO₂⁻ in electrolyte was quantified by Griess test after chronoamperometry tests^15,31. The Griess agent was prepared by dissolving 800 mg of N-(1-Naphthyl)ethylenediamine dihydrochloride, 40 mg of sulfonamide, and 2 mL of H₃PO₄ (85%) into 10 ml of DI water. Prior to analysis, the NO₂⁻-containing electrolyte was diluted to an appropriate detection level, and 2 mL of diluted solutions were mixed with 40 μL of Griess agent and rested for 10 min at room temperature. Absorbance at ca. 540 nm of UV-Vis spectrophotometry was used to quantify the concentration of NO₂⁻. The calibration curve was acquired using a variety of NaNO₂ aqueous solutions as standard samples with the same operation (Supplementary Fig. 24c, d).

Quantification of NH₃ concentration using the indophenol blue method

The concentration of produced NH₃ was quantified by a modified indophenol blue method^15,31. After chronoamperometry tests, 1 mL of electrolyte was extracted from the electrochemical reaction vessel and diluted to an appropriate detection level. Subsequently, 1 mL of a 1.0 M NaOH solution containing 5 wt % salicylic acid and 5 wt % sodium citrate, 500 μL of 0.05 M NaClO, and 100 μL of 1 wt% C₅FeN₆Na₂O (Sodium nitroferricyanide) aqueous solution were successively added into the diluted electrolyte. After 2 h of chromogenic reaction at room temperature, the formation of indophenol blue was determined at a wavelength of 655 nm in UV-Vis spectrophotometry. The calibration curve was acquired using a variety of (NH₄)₂SO₄ aqueous solutions as standard samples with the same operation (Supplementary Fig. 24a, b).

Quantification of NH₃ concentration using the ¹H NMR spectroscopy

An isotope-labeled tracer experiment was conducted to further quantify the produced NH₃ using the ¹H NMR spectroscopy with 100 mM ¹⁵NO₃⁻ solution as the nitrogen source^15,31. After chronoamperometry test, 200 μL of electrolyte were extracted from the electrochemical reaction vessel and then mixed with 10 μL of maleic acid (3.6 mM, as the internal standard) aqueous solution, 10 μL of H₂SO₄ aqueous solution (4 M), and 380 μL of DMSO-d₆. After adding tetramethylsilane as the reference, the mixture was sealed into an NMR tube (5 mm in diameter, 400 MHz). The water peak was suppressed using a pre-saturation method. The concentration of produced NH₃ can be quantified by calculating the integral area (I) of the vinylic singlets for maleic acid and the typical triplet for ammonium based on the followed equation:

$${C}_{{{{{{\rm{NH}}}}}}_{4}^{+}}=\frac{{I}_{{{{{{\rm{NH}}}}}}_{4}^{+}}/{H}_{{{{{{\rm{NH}}}}}}_{4}^{+}}}{{I}_{s}/{H}_{s}}\times {C}_{s}$$

(3)

where ${C}_{{{{{{\rm{NH}}}}}}_{4}^{+}}$ and ${C}_{s}$ are the concentrations of NH₃ and maleic acid in NMR tubes; ${H}_{{{{{{\rm{NH}}}}}}_{4}^{+}}$ and ${H}_{s}$ are the number of protons for ammonium and maleic acid; ${I}_{{{{{{\rm{NH}}}}}}_{4}^{+}}$ and ${I}_{s}$ are the integrals of ¹H NMR peaks for NH₃ and maleic acid. The isotopic labeling experiment was performed to verify the origin of NH₃ using 0.5 M Na₂SO₄ containing Na¹⁵NO₃ as the electrolyte in the same chronoamperometry test as described above.

Determination of H₂ yield using gas chromatography

The H₂ yield were analyzed by an online gas chromatograph equipped with a thermal conductivity detector and a flame ionization detector (Shanghai Ramiin GC 2060).

Calculation of the mass activity for NH₃ production

The mass activity for NH₃ synthesis is defined as the partial current density of NH₃ normalized by the mass of Fe, which was calculated with the following equation:

$${{{\rm{MA}}}}({{{{\rm{NH}}}}}_{3})=\frac{i\times {{{\rm{FE}}}}({{{{\rm{NH}}}}}_{3})}{m}$$

(4)

where i is the total current, ${{{\rm{FE}}}}({{{{\rm{NH}}}}}_{3})$ is the Faradaic efficiency for ammonia. m is the mass of Fe.

Calculation of the TOF of NO₃RR

The TOF for NH₃ synthesis is defined as the number of ammonia molecules produced per unit time and per Fe sites. Thus, TOF values can be derived from YR(NH₃) with the following equation:

$${{{\rm{TOF}}}}=\frac{{{{\rm{YR}}}}({{{{\rm{NH}}}}}_{3})\times 56\times {{{N}}}_{{{A}}}}{17\times 3600\times {{{N}}}_{{{A}}}}$$

(5)

where N_A is the Avogadro constant (6.02 × 10²³).

Calculation of the Faradaic efficiency (FE) and NH₃ yield rate (YR)

FEs and mass-normalized yield rates were calculated:

$${{{\rm{FE}}}}({{{{\rm{NH}}}}}_{3})=(8F\times C\times V\times n)/Q$$

(6)

$${{{\rm{FE}}}}({{{{{\rm{NO}}}}}_{2}}^{-})=(2F\times C\times V\times n)/Q$$

(7)

$${{{\rm{YR}}}}({{{{\rm{NH}}}}}_{3})=(C\times V\times n)/(t\times m)$$

(8)

where F is the Faraday constant (96,485 C/mol), C is the measured NH₃ concentration (mg/mL), V is the volume of electrolyte, Q is the total charge passed through the electrode, n is the dilution factor, m is the mass of Fe and t is the time of electrolysis.

Operando SR-FTIR measurements

The operando SR-FTIR measurements were conducted using a home-made spectro-electrochemical cell⁵³. A ~60-nm-thick SR-FTIR-active Au film was pre-deposited on the reflection plane of a semicircular cylinder Si prism, after which the catalyst was drop-casted on the film. An optical system embedded in the spectroscopic chamber was used to carry out the operando SR-FTIR measurements. To reach a high signal/noise ratio, the spectrum was acquired by averaging 256 interferograms with an incident angle of the IR beam set to 55°. The absorbance spectra were shown at various applied potentials.