Maximized atom utilization in a high-entropy metallene via single atom alloying for boosted nitrate electroreduction to ammonia

Abstract

High-entropy alloys, with their unique structural characteristics and intrinsic properties, have evolved to be one of the most popular catalysts for energy-related applications. However, the geometry of the traditional nanoparticle morphology confines the majority of active atoms to the particle core, deeming them ineffective. In this study, we present a class of two-dimensional high-entropy alloys, namely, high-entropy metallenes, constructed by alloying various single-atom metals in atomically thin layers and reveal their great feasibility for electrocatalytic nitrate reduction to ammonia. Through multimetal interactions, various active centres are formed and sufficiently exposed over the metallene. Each element performs its own duties and jointly lowers the energy barrier of the rate-determining step. As expected, the proof-of-concept PdCuNiCoZn high-entropy metallene delivers satisfactory catalytic performance across wide pH ranges. In particular, in a strongly alkaline electrolyte, a maximum ammonia yield rate of 447 mg h⁻¹ mg⁻¹ and a high Faradaic efficiency of 99.0% are achieved.

The conventional nanoparticle morphology in high-entropy alloys confines most active atoms to the particle core, making them inaccessible. Here, two-dimensional high entropy metallenes are reported, achieving maximized atom utilization and showing great feasibility for nitrate reduction.

Introduction

Ammonia (NH₃) is an important type of clean energy that can not only serve as an alternative to gasoline, diesel, and other fossil fuels to directly provide clean fuel for automobile engines but also produce hydrogen via catalytic decomposition to provide safe hydrogen for vehicle fuel cells^1–3. To date, ammonia has been industrially synthesized through the Haber‒Bosch process, which requires high temperatures (400−500 °C) and high pressures (100‒200 kPa)^4,5. This highly energy-consuming process places a heavy burden on the sustainable development of human beings. Recently, the electrocatalytic nitrogen reduction reaction (NRR) has been regarded as an alternative route for ammonia production^6–8. However, owing to the strong bond energy of N≡N (941 kJ mol⁻¹) and poor N₂ solubility, most NRRs reported in the literature have a low NH₃ yield rate ( < 200 μg h⁻¹ mg⁻¹) and low Faradaic efficiency ( < 30%)^9,10. This bottleneck persists despite numerous efforts to resolve it.

Alternatively, nitrate, a significant water pollutant, is considered a good nitrogen source due to its lower bond energy (204 kJ mol⁻¹) and high water solubility^11,12. Using nitrate as a nitrogen source could remove water contaminants and produce ammonia simultaneously, which has become a popular topic of research¹³. Many studies have reported the production of ammonia via nitrate reduction (NO₃RR), and a much higher NH₃ yield rate and Faradaic efficiency than those of the NRR have been achieved. For example, single-atom catalysts were designed to maximize the use of active sites and inhibit the competing hydrogen evolution reaction (HER)^14,15. However, owing to the low loading of single atoms (usually <3 wt%), the number of active sites on single-atom catalysts is quite small. Moreover, the element types in single-atom catalysts are relatively simple, and optimizing the activity of the active centre is challenging, which hinders further enhancement of the NH₃ yield rate. In this regard, developing a catalyst that possesses abundant active sites with high intrinsic activity is imperative.

Recently, high-entropy alloys (HEAs), emerging materials that are composed of five or more elements with concentrations of 5–35 atom % for each constituent in the form of a single-phase solid solution, have been greatly favoured in the field of heterogeneous catalysis^16–18. Within HEAs, different atomic species are confined to the same lattice in a disorderly manner. This high degree of heterogeneity provides diverse sites for molecular adsorption and reactions. Additionally, owing to the difference in atomic size, severe lattice distortion occurs, which could facilitate the activation and transport of reaction species. Moreover, the multicomponent characteristics render HEAs tuneable in terms of composition and structure. Compared with single metal, binary, or ternary alloy catalysts, it is more convenient to modulate the d-band centre of active metals within HEAs and obtain appropriate adsorption energy towards key intermediates, thus improving their intrinsic activity¹⁹. However, bottlenecks in the exertion of theoretical performance remain. In terms of morphology, HEA catalysts are usually synthesized as nanoparticles, and in a particle-structure catalyst, only atoms at its surface can act as active sites, and the inner atoms remain underutilized. In this context, designing HEAs with metallene structures to increase atom utilization and expose abundant active sites may be a better choice^20–22. High-entropy metallenes formed via single-atom alloying at the atomic level confine several elements into 2D metallenes with a uniform distribution, thus realizing the atomic dispersion of each element with a high loading density and maximized exposure. Simultaneously, the intrinsic performance of active centres can be greatly improved through multimetal interactions in HEAs. This kind of high-entropy materials with both high-quality and high-quantity active sites is expected to be an ideal NO₃RR electrocatalyst.

Inspired by this finding, in this work, we designed a class of 2D HEA materials, namely, high-entropy metallenes, as NO₃RR catalysts. These materials are constructed by alloying multiple single-atom metals in atomically thin layers, which perform their own duties and jointly benefit the whole reaction, thus achieving maximized atom utilization. Through multimetal interactions, various active centres are formed and sufficiently exposed over the metallene. The proof-of-concept PdCuNiCoZn high-entropy metallene delivers satisfactory NO₃RR performance in both neutral and alkaline electrolytes. In particular, in a strongly alkaline electrolyte, a maximum ammonia yield rate of 447 mg h⁻¹ mg⁻¹ and a high Faradaic efficiency of 99.0% are achieved in 0.1 M KNO₃. An in-depth mechanistic study reported that the strong interactions between each metal site push the d-band centre closer to the Fermi level, increasing the activity of the electrons on PdCuNiCoZn to promote the adsorption and activation of some key intermediates. As a result, the energy barrier for *NO hydrogenation as the rate-determining step (RDS) is significantly reduced at each active centre of the high-entropy metallene.

Results

Synthesis and characterization of PdCuNiCoZn high-entropy metallenes

Considering their decent NO₃RR performance and high compatibility, Pd^23,24, Co^25,26, Ni²⁷, Cu²⁸, and Zn²⁹ were selected as the proof-of-concept elements. The high-entropy metallene was synthesized through a facile wet‒chemical approach. Metal acetylacetonate salts and a certain amount of ascorbic acid serving as a reducing agent were dissolved, followed by ultrasonication and heating. The solution gradually turned dark during the reaction, indicating the reduction and coordination fusion process. Through the effect of a reducing agent, a condensation reaction occurred between acetylacetone molecules to form cross-linked metal bonds³⁰. The selection of metal acetylacetonates is also necessary due to the strong metal‒acetylacetonate interaction, which can facilitate coprecipitation by slowing the rate of precipitation^31,32. A PdCuNiCoZn high-entropy metallene was obtained, which exhibited a typical graphene-like structure, as shown by scanning electron microscopy (SEM) and scanning transmission electron microscopy (STEM) images (Fig. 1a, b), with a thickness of 1.98 nm, as measured via atomic force microscopy (AFM) (Supplementary Fig. 1). Elemental mapping and energy dispersive X-ray spectroscopy (EDX) line scanning confirmed the uniform distributions of Pd, Cu, Ni, Co, and Zn (Supplementary Fig. 2). The atomic ratio of each element in the PdCuNiCoZn high-entropy metallene was further determined to be Pd/Co/Ni/Cu/Zn= 23.9/24.7/20.1/24.3/7.0 via inductively coupled plasma‒optical emission spectrometry (ICP‒OES).

Fig. 1. Physical characterization of the PdCuNiCoZn high-entropy metallene.

a Scanning electron microscopy (SEM), b scanning transmission electron microscopy (STEM), and c high-resolution transmission electron microscopy (HRTEM) images of the PdCuNiCoZn high-entropy metallene. d X-ray diffraction (XRD) patterns of different samples. The abbreviation “arb. units” refers to “arbitrary units”. e, f aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC-HAADF-STEM) images of the PdCuNiCoZn high-entropy metallene. The inset in (f) shows the fast Fourier transformation (FFT) pattern of the PdCuNiCoZn high-entropy metallene.

The high-resolution transmission electron microscopy (HRTEM) image revealed that the interplanar spacing of 0.215 nm could be attributed to the (111) plane of the face-centred cubic (fcc) structure (Fig. 1c)³³. Upon further characterization of the crystal structure via X-ray diffraction (XRD) (Fig. 1d), the diffraction peak at approximately 41° could be attributed to the (111) plane of the fcc lattice, which matches JCPDS No. 96-101-1113 and the above HRTEM results. For a comparison, Pd metallene (Supplementary Figs. 3‒6), PdCu bimetallene (Supplementary Figs. 7‒10), PdCuNi ternary metallene (Supplementary Figs. 11‒14), and PdCuNiCo quaternary metallene (Supplementary Figs. 15‒18) were also prepared as comparative samples and characterized carefully. Notably, the diffraction peak of the PdCuNiCoZn high-entropy metallene attributed to the (111) plane shows a positive shift compared with that of the other counterparts, indicating the confinement of four transition metals with smaller radii into the fcc lattice. Moreover, these peaks exhibit lower intensities, indicating the existence of distorted structures caused by the incorporation of other atoms with various radii, which is a typical characteristic of high-entropy materials and is also consistent with the lattice distortion observed in the HRTEM image (Fig. 1c)^34,35. These characterizations confirmed the successful synthesis of PdCuNiCoZn high-entropy metallenes with a single-phase solid solution structure, in which Pd, Cu, Ni, Co, and Zn atoms randomly occupy the Pd-based fcc lattice³⁶.

To clearly observe the existence form of each element and analyse the distribution of metal atoms on the PdCuNiCoZn high-entropy metallene, aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (AC-HAADF-STEM) images of the high-entropy metallene from the randomly selected crystalline regions are provided (Fig. 1e and f, Supplementary Fig. 19). The fast Fourier transformation (FFT) image along the [011] zone axis fully confirms the typical fcc structure. Surprisingly, a high density of single dots appeared, which can be attributed to the possible existence of single atoms anchored in the crystalline region of the PdCuNiCoZn high-entropy metallene. To confirm the above deduction and further investigate the chemical state and coordination environment of the metals in the PdCuNiCoZn high-entropy metallene at the atomic level, synchrotron radiation X-ray absorption spectroscopy (XAS) characterization, including X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) spectroscopy, was performed. The K-edge adsorption energy of Pd is similar to that of Pd foil (Fig. 2a), indicating that metallic Pd is dominant in the PdCuNiCoZn metallene. In contrast, the K-edge adsorption energies of Cu, Ni, Co, and Zn are positive compared with those of their metal foil standard samples (Fig. 2b‒e), confirming their positive valence states. The adsorption edge of Cu is between the Cu foil and CuO, and the adsorption edges of Ni, Co, and Zn are near their divalent oxides. The EXAFS spectra (R space) after Fourier transformations of the original data are shown in Fig. 2f‒j. For the EXAFS spectra of Cu, the peak at 2.46 Å could be assigned to the Pd‒Cu bond, confirming the strong interaction between Cu and Pd³⁷. Another peak at approximately 1.51 Å in the R space could be ascribed to the first-shell Cu‒O bond^38,39. In addition, no Cu‒Cu coordination peaks belonging to Cu or CuO are detected, further confirming the atomic dispersion of Cu on the Pd-based metallene substrate. Similarly, in the EXAFS spectra of Ni, Co, and Zn, the chemical bonds of Pd‒Ni, Pd‒Co, and Pd‒Zn are clearly observed, and no Ni‒Ni, Co‒Co, or Zn‒Zn coordination peaks can be observed, confirming the existence of single Ni, Co, and Zn atoms. Conclusively, the above FT-EXAFS spectra verify the atomic-level dispersion of Cu, Ni, Co, and Zn on Pd-based metallenes via single-atom alloying. The EXAFS fitting results and the fitting parameters also confirm that the isolated metal atoms are coordinated with Pd and O atoms (Supplementary Fig. 20 and Supplementary Table 1). Moreover, in the K-edge wavelet transform (WT) EXAFS data, any metal‒metal coordination signal, with the exception of Pd‒Pd, is absent in PdCuNiCoZn, whereas M‒O and M‒Pd bonds (M = Cu, Ni, Co, and Zn) can be clearly observed (Fig. 2k‒o), further identifying the single-atom-scale multielement alloying structure. This structure would induce a synergistic effect and stimulate electron redistribution. The electronic state of each element in the PdCuNiCoZn high-entropy metallene was subsequently explored via X-ray photoelectron spectroscopy (XPS) (Supplementary Fig. 21). The high-resolution Cu/Ni/Co/Zn 2p XPS spectra can be deconvoluted into two doublets assigned to Cu/Ni/Co/Zn 2p_3/2 and Cu/Ni/Co/Zn 2p_1/2. In the high-resolution XPS spectra of Pd 3d, the peaks at approximately 335.0 and 340.3 eV could be attributed to Pd⁰ 3d_5/2 and Pd⁰ 3d_3/2, and the peaks at 336.6 and 341.9 eV could be assigned to Pd²⁺ 3d_5/2 and Pd²⁺ 3d_3/2⁴⁰. Notably, compared with that of the other samples, the binding energy of Pd in PdCuNiCoZn obviously negatively shifted by 0.3 eV (Supplementary Fig. 22), indicating electron transfer from the transition metals to the Pd site. The strong interaction between multiple components in high-entropy metallenes could effectively regulate the electronic state of each site, holding great potential for optimizing the adsorption energy of intermediates and greatly lowering the reaction energy barrier of the NO₃RR.

Fig. 2. XAS characterization of the PdCuNiCoZn high-entropy metallene.

a–e X-ray absorption near edge structure (XANES) spectra of the K-edges of Pd, Cu, Ni, Co, and Zn in the PdCuNiCoZn high-entropy metallene. f–j FT k³-weighted extended X-ray absorption fine structure (EXAFS) spectra of Pd, Cu, Ni, Co, and Zn in the PdCuNiCoZn high-entropy metallene. k–o Wavelet transform (WT) EXAFS data of Pd, Cu, Ni, Co, and Zn in the PdCuNiCoZn high-entropy metallene. The abbreviation “arb. units” refers to “arbitrary units”.

Electrocatalytic nitrate reduction performance

A certain amount of PdCuNiCoZn high-entropy metallene was loaded onto carbon cloth with an optimized loading content to evaluate its NO₃RR activity. First, linear sweep voltammetry (LSV) was performed in 0.5 M K₂SO₄ with different concentrations of NO₃^- (Fig. 3a). With increasing NO₃^- concentration from 0 to 0.1 M, the current density gradually increases, and there is a wide gap in the current density with and without NO₃^-, indicating that the NO₃RR is the primary process in the presence of NO₃^-. Additionally, the onset potential for nitrate reduction presents a positive shift with increasing NO₃^- concentration. In contrast, the current densities and current density gaps of Pd metallene, PdCu bimetallene, PdCuNi ternary metallene, and PdCuNiCo quaternary metallene in the presence of NO₃^- are smaller than those of PdCuNiCoZn high-entropy metallene (Supplementary Fig. 23), highlighting the advantages of high-entropy metallene design.

Fig. 3. Electrochemical NO₃RR performance of the PdCuNiCoZn high-entropy metallene.

a Linear sweep voltammetry (LSV) curves of PdCuNiCoZn high-entropy metallenes in 0.5 M K₂SO₄ with different concentrations of NO₃^-. b, c Comparison of mass-normalized NH₃ yield rates and corresponding Faradaic efficiencies under different applied potentials in Ar-saturated 0.1 M KNO₃ + 0.5 M K₂SO₄ (pH = 7 ± 0.3) at room temperature of 25 °C with Ar flow rate of 10 sccm. d NH₃ Faradaic efficiencies and mass-normalized yield rates at −0.6 V versus RHE of PdCuNiCoZn high-entropy metallenes in 0.5 M K₂SO₄ with different KNO₃ concentrations. e LSV curves of PdCuNiCoZn high-entropy metallene in 0.1 M KNO₃ + 1.0 M KOH. f NH₃ Faradaic efficiencies and mass-normalized yield rates of PdCuNiCoZn in Ar-saturated 0.1 M KNO₃ + 1.0 M KOH (pH = 14 ± 0.2) at room temperature of 25 °C with Ar flow rate of 10 sccm. The error bars correspond to the standard deviations of measurements of three separately prepared samples under the same conditions. g ¹H-nuclear magnetic resonance (¹H-NMR) spectra of the electrolytes produced from the NO₃RR using ¹⁴NO₃^- and ¹⁵NO₃^- as the isotopic NO₃^- source. h Consecutive recycling test over PdCuNiCoZn high-entropy metallene in 0.1 M KNO₃ + 0.5 M K₂SO₄ at −0.9 V versus RHE. The electrolytic cell resistance was 4.3 ± 0.2 Ω, and all the potentials were not iR-corrected.

Second, amperometric i-t measurements were conducted in 0.1 M KNO₃/0.5 M K₂SO₄ to quantitatively evaluate the NO₃RR activity of PdCuNiCoZn high-entropy metallenes in a neutral electrolyte (Supplementary Figs. 24‒29). Upon tuning the applied potential from −0.3 V to −1.2 V versus the reversible hydrogen electrode (RHE), the NH₃ yield rate over the PdCuNiCoZn high-entropy metallene continues to increase, whereas its Faradaic efficiency exhibits a typical volcano trend (Fig. 3b, c, Supplementary Table 2). The highest NH₃ yield rate was observed at −1.2 V versus RHE, which exceeded 275 mg h⁻¹ mg⁻¹, and a maximum NH₃ Faradaic efficiency of 95.4% was achieved at −0.6 V versus RHE. Notably, PdCuNiCoZn high-entropy metallenes also show high Faradaic efficiency ( > 90%) over a wide operating potential window from −0.6 V to −1.0 V versus RHE, indicating the conversion of NO₃^- to NH₃ with low energy consumption, which is beneficial for industrial-scale applications. In contrast, the Pd metallene, PdCu bimetallene, PdCuNi ternary metallene, and PdCuNiCo quaternary metallene deliver inferior NO₃RR activity. In particular, at low overpotentials, the NH₃ yield rate and Faradaic efficiency of PdCuNiCoZn high-entropy metallenes are an order of magnitude greater. Moreover, this performance is competitive among reported catalysts in the literature under similar electrolyte environments (Supplementary Table 3). The production of other byproducts, such as nitrite (NO₂^-) and hydrazine (N₂H₄), was subsequently quantified (Supplementary Figs. 30‒33). As shown, the Faradaic efficiency of NO₂^- was approximately 40% at −0.3 V versus RHE but quickly fell below 5% at most of the applied potentials, varying from −0.6 V to −1.2 V versus RHE, indicating its high selectivity towards NH₃. Moreover, almost no N₂H₄ could be detected in the electrolyte after electrolysis.

Because different NO₃^- concentrations exist in different sources, further catalytic performance tests of PdCuNiCoZn high-entropy metallenes were conducted in 0.5 M K₂SO₄ with different NO₃^- concentrations ranging from 0.01 M to 0.1 M at −0.6 V versus RHE (Supplementary Fig. 34). Both the current density and the NH₃ yield rate increased with increasing NO₃^- concentration, whereas the NH₃ Faradaic efficiency gradually increased (Fig. 3d), which could be attributed to the NO₃^- mass diffusion limit⁴¹. Moreover, the effect of pH on the NO₃RR was also investigated. To rule out the effect of different additional ions, a certain amount of H₂SO₄ or KOH was used to adjust the pH of the K₂SO₄. As shown in the LSV curves of PdCuNiCoZn high-entropy metallenes in acidic electrolytes (pH adjusted to 2), the onset potential for the NO₃RR (−0.7 V versus RHE) seems much more negative than that in neutral electrolytes (−0.3 V versus RHE), and the corresponding current density was also lower (Supplementary Fig. 35). This result could be attributed to the competitive HER in acidic environments, which would suppress nitrate reduction⁴². In contrast, under weakly alkaline conditions in electrolytes (pH adjusted to 12), a wide gap exists between the current density curves with or without NO₃^-, and a much positive onset potential for the NO₃RR could be achieved (0.0 V versus RHE), which is even lower than that in neutral electrolytes (Supplementary Fig. 36a), indicating that the NO₃RR on PdCuNiCoZn high-entropy metallene was also favourable in alkaline electrolytes. In the following NO₃RR activity test conducted from −0.1 V to −1.0 V versus RHE, a maximum NH₃ Faradaic efficiency of 95.6% was achieved at −0.4 V versus RHE, and an NH₃ yield rate greater than 376 mg h⁻¹ mg⁻¹ (22.1 mol g⁻¹ h⁻¹) was recorded at −1.0 V versus RHE (Supplementary Fig. 36b‒d), which is also competitive across most reported catalysts in similar electrolyte environments (Supplementary Table 3). Based on this finding, the NO₃RR activity of PdCuNiCoZn was further investigated in a strongly alkaline electrolyte (0.1 M KNO₃/1.0 M KOH, pH = 14 ± 0.2) (Fig. 3e, Supplementary Figs. 37 and 38). The NH₃ Faradic efficiency exhibited a typical volcano trend, and a maximum NH₃ Faradaic efficiency greater than 99.0% was achieved at −0.5 V versus RHE (Fig. 3f). The highest NH₃ yield rate is observed at −0.8 V versus RHE, which is greater than 447 mg h⁻¹ mg⁻¹ (26.3 mol g⁻¹ h⁻¹). The above exploration expands our recognition of the use of PdCuNiCoZn high-entropy metallenes in the NO₃RR under different electrolyte environments and indicates their wide applicability in both neutral and alkaline environments with different NO₃^- concentrations.

Several additional control experiments were performed to eliminate contaminants from the environment and catalysts (Supplementary Fig. 39). The NH₃ produced in the NO₃^--free electrolyte was almost negligible compared with the reaction conducted in the electrolyte with NO₃^-. Moreover, pure carbon cloth and Ketjen black could hardly produce NH₃ in the KNO₃ electrolyte, confirming that the activity originated from the loaded catalysts. To further verify that the as-produced NH₃ originates from the reduction of NO₃^-, ¹⁵N isotope labelling experiments were performed. The ¹H-nuclear magnetic resonance (¹H-NMR) spectrum of the electrolyte subjected to K¹⁵NO₃ reduction shows two peaks representing ¹⁵NH₄⁺ at 7.02 and 6.84 ppm, whereas the spectrum of the electrolyte with K¹⁴NO₃ as the reactant shows three peaks attributed to ¹⁴NH₄⁺ at 7.06, 6.93, and 6.80 ppm (Fig. 3g). Moreover, quantitative calculations of the ammonia concentration using maleic acid as an external standard were also performed (Supplementary Figs. 40 and 41). The concentrations of both ¹⁴NH₄⁺ and ¹⁵NH₄⁺ are quite close to the results obtained via the indophenol blue method, confirming the accuracy of the different quantitative methods and that the generation of NH₃ indeed originates from the electroreduction of NO₃^-. Additionally, we checked the stability of the PdCuNiCoZn high-entropy metallene (Supplementary Fig. 42). During the cycling stability test at −0.9 V versus RHE in 0.5 M K₂SO₄/0.1 M KNO₃, the NH₃ yield rate remained above 120 mg h⁻¹ mg⁻¹ and the Faradaic efficiency remained above 85% after 50 h of running (Fig. 3h). According to the characterization via TEM, XRD, and XPS, the morphology, crystal structure, and chemical state of the high-entropy metallene remained almost unchanged after electrolysis (Supplementary Figs. 43–45), thus confirming its high stability towards the NO₃RR.

Mechanistic investigation

The reasons behind the NH₃ yield rate and Faradaic efficiency of PdCuNiCoZn high-entropy metallenes were explored both experimentally and theoretically. The electrochemically active surface area (ECSA) of each sample was estimated by the double-layer capacitance (C_dl). Cyclic voltammetry (CV) at different scan rates was performed in the region where no Faradaic current occurred. Compared with Pd metallene (C_dl = 2.48 mF cm⁻², A_ECSA = 62 cm²_ECSA), PdCu metallene (C_dl = 4.75 mF cm^-2, A_ECSA = 118.75 cm²_ECSA) and PdCuNi ternary metallene (C_dl = 6.45 mF cm⁻², A_ECSA = 161.13 cm²_ECSA), PdCuNiCo quaternary metallene (C_dl = 11.53 mF cm⁻², A_ECSA = 288.13 cm²_ECSA) and PdCuNiCoZn high-entropy metallene (C_dl = 11.54 mF cm⁻², A_ECSA = 288.38 cm²_ECSA) possess much higher C_dl and ECSA values (Supplementary Fig. 46). This finding indicates that the introduction of “high-entropy” into metallene materials results in abundant active sites⁴³, which might originate from the high degree of heterogeneity in high-entropy materials⁴⁴.

To determine the dominant NO₃RR pathway, in situ Raman, in situ attenuated total reflectance surface enhanced infrared absorption spectroscopy (ATR-SEIRAS), and online differential electrochemical mass spectrometry (DEMS) were conducted (Supplementary Fig. 47). First, in situ Raman characterization as a function of electrolysis time was conducted at −0.8 V versus RHE. The signals of -NH at 1530 cm⁻¹ and -NH₂ at 1154 cm⁻¹ gradually appeared during electrolysis (Fig. 4a), confirming the generation of ammonia^45,46. Moreover, when the peak of SO₄^2- at 980 cm^-1 is used as a reference, the peak of NO₃^- at 1049 cm^-1 gradually weakens, implying its consumption. To obtain more details to differentiate the nitrate signal at the electrode surface (adsorbed species) from that in the bulk electrolyte, we amplified the Raman signals from 800 cm⁻¹ to 1300 cm⁻¹ (Supplementary Fig. 48). When electrolysis started, peaks of the adsorbed NO₃^- (NO₃^-_ads) at 1014 cm⁻¹ and 1019 cm⁻¹ as well as the N = O stretching vibration from the unidentate nitrate near 990 cm⁻¹ gradually appeared, indicating that the NO₃^- species began to adsorb on the surface of the catalysts. As time progressed, the intensity of these adsorbed nitrate species gradually weakened, indicating the conversion of NO₃^- to NH₃^47,48. Further insights can be obtained from the in situ ATR-SEIRAS results (Fig. 4b). The peak at 1630 cm⁻¹ is attributed to the H–O–H bending mode of the interfacial water molecules⁴⁹. A negative band at ~1310 cm⁻¹ is assigned to NO₃^-, which shows a negative increase owing to its continuous consumption^50,51. Moreover, positive bands appeared, including a deoxidation intermediate (*NO₂ at 1220 cm⁻¹)^52,53 and a hydrogenation intermediate (*NH₂ at 1160 cm⁻¹)^54,55. The signals of these key intermediates imply the effective activation of NO₃^- via a hydrogenation/deoxidation process to form NH₃. Second, DEMS was applied to collect the m/z signals of the reaction intermediates during the NO₃RR at a given voltage. The m/z signals of 46, 30, 14, 15, 16, and 17, corresponding to NO₂, NO, N, NH, NH₂, and NH₃, respectively, were captured (Fig. 4c). The DEMS signal strengths of other possible intermediates, such as NH₂OH, N₂, and N₂O, are almost zero or several orders of magnitude lower (Supplementary Fig. 49). Based on the combination of the above in situ characterizations, the NO₃RR pathway over PdCuNiCoZn high-entropy metallene is assumed to predominantly follow NO₃^- → *NO₃ → *HNO₃ → *NO₂ → *HNO₂ → *NO → *NOH → *N → *NH → *NH₂ → NH₃(g)⁵⁶.

Fig. 4. Analysis of the NO₃RR pathway over the PdCuNiCoZn high-entropy metallene.

a In situ Raman spectra of PdCuNiCoZn high-entropy metallenes as a function of time. The abbreviation “arb. units” refers to “arbitrary units”. b In situ attenuated total reflectance surface enhanced infrared absorption spectroscopy (ATR-SEIRAS) spectra of PdCuNiCoZn high-entropy metallenes at different potentials during the NO₃RR. c Online differential electrochemical mass spectrometry (DEMS) spectra of PdCuNiCoZn high-entropy metallenes during the NO₃RR. The abbreviation “m/z” refers to “mass-to-charge ratio”. The (d) Gibbs free energy diagrams for the NO₃RR at different active centres of PdCuNiCoZn high-entropy metallenes. e Gibbs free energy diagrams for the NO₃RR on PdCuNiCoZn, PdCu, and Pd. The abbreviation “RDS” refers to “rate-determining step”.

DFT calculations were subsequently conducted to explore the in-depth mechanism involved. The most exposed lowest-energy Pd(111) facet with a random element distribution was simplified as a model for calculations, and structural models of PdCuNiCoZn high-entropy metallene, Pd metallene, and PdCu bimetallene were built accordingly (Supplementary Fig. 50). After optimization of the structural model, the adsorption energy of each intermediate on the surface of the catalyst was calculated. After screening, four kinds of active centres are determined to be potential reaction sites (Supplementary Figs. 51–54). Notably, the adsorbates generally do not interact with only one atom at the site. Neighbouring atoms continue to participate and will affect the adsorbate binding properties, either by affecting the electronic state of the “atom site” or by directly binding to the adsorbate⁵⁷. Therefore, these active centres are denoted as Pd‒Pd sites, Pd‒Cu sites, Pd‒Co sites, and Co‒Ni sites on the basis of specific binding sites. The Gibbs free energy diagrams for the NO₃RR at different sites are presented in Fig. 4d. The energy profile at U = 0 V shows that the hydrogenation of *NO has the highest energy barrier among all the steps of the NO₃RR for all the cases (0.45 eV for the Pd‒Pd site, 0.73 eV for the Pd‒Cu site, 0.91 eV for the Pd‒Co site, and 0.45 eV for the Co‒Ni site), implying that the hydrogenation of *NO is the RDS. To confirm this RDS experimentally, electrochemical reduction of nitric oxide was conducted (Supplementary Fig. 55). The results show that the polarization of NO reduction has a similar onset potential to that of nitrate reduction, thus it is a key reaction in the NO₃RR. By comparison, the RDS on pristine Pd metallene, which is also the hydrogenation of *NO, reaches 1.02 eV (Supplementary Fig. 56), thus highlighting the advantages of high-entropy metallene design. Furthermore, to reflect the actual reaction process, the applied potential was also considered in the DFT calculations. After a potential of −0.245 V was applied, where the reaction initiates, the energy barrier for most steps sharply decreases (Fig. 4e and Supplementary Figs. 57 and 58). The energy barrier for *NO hydrogenation at −0.245 V continues to follow PdCuNiCoZn (0.21 eV) < Pd (0.78 eV). PdCu bimetallene not only delivers a relatively larger energy barrier than PdCuNiCoZn high-entropy metallene does (0.55 eV under U = 0 V and 0.30 eV under U = -0.245 V) (Supplementary Figs. 59 and 60) but is also more prone to contribute to the HER due to the low energy barrier (Supplementary Fig. 61). Regarding the possible reaction pathway, the overall reaction energy of *NO → *NOH → *N → *NH → *NH₂ → NH₃(g) is more exothermic than that of *NO → *NOH → *NHOH → *NH₂OH → *NH₂ → NH₃(g) (Supplementary Fig. 62), indicating that the NO₃RR is more inclined to follow the N pathway, which is in accordance with the in situ characterization results.

To gain insight into the catalytic effect, the partial projected density of states (PDOSs) of each element in the PdCuNiCoZn high-entropy metallene were carefully analysed. Owing to the spin polarization effect, the d-band orbitals can be divided into spin-up orbitals and spin-down orbitals. Spin states occupying spin–down orbitals are likely to be the active centers⁵⁸. From the PDOS pattern, we can observe that most of the orbitals of each element in high-entropy metallenes strongly overlap (Fig. 5a), which facilitates site-to-site electron transfer between different elements during the NO₃RR⁵⁹. After the adsorption of NO₃^-, the p-orbitals were observed to partially overlap with the d-orbitals of high-entropy metallenes (Fig. 5b). Interestingly, the orbitals of NO₃^- changed from sharp and localized to wide and delocalized after its activation on the surface of the high-entropy metallene, indicating the strong interaction between NO₃^- and the high-entropy metallene, which would benefit the subsequent hydrogenation/deoxidation process and promote the whole reaction. During the NO₃RR process, each element performs its own duties and jointly benefits multielectron reactions. Specifically, the orbital of Zn is quite localized and occupies the deepest position away from Fermi level (E_F) to play the role of an electron reservoir⁶⁰. Ni and Co at positions near the E_F act as electron depletion centres and strongly facilitate the adsorption of key intermediates⁶¹. Pd and Cu, which are electron rich and have broad orbital characteristics, can accelerate electron transfer and increase the likelihood of bonding³⁶. As a result, maximized atom utilization is achieved. Moreover, owing to the regulated electronic state originating from the interaction between each component, compared with Pd (−1.55 eV) and PdCu (−1.25 eV), the d-band centre of PdCuNiCoZn (-0.89 eV) is clearly closer to the Fermi surface (Fig. 5c and d). This finding indicates that the electrons on PdCuNiCoZn are more active than those on PdCuNiCoZn, which could promote the adsorption and activation of some key intermediates, thus increasing the catalytic activity of high-entropy metallenes.

Fig. 5. Mechanistic investigation of the NO₃RR over the PdCuNiCoZn high-entropy metallene.

a Projected density of states (PDOSs) of all the elements in the PdCuNiCoZn high-entropy metallene. The abbreviation “E_F” refers to “Fermi level”. b PDOSs for NO₃^- adsorption on PdCuNiCoZn. c, d PDOSs and d-band centres of PdCuNiCoZn, PdCu, and Pd.

Discussion

A class of 2D HEA materials, namely, high-entropy metallenes, is successfully constructed via the alloying of multiple single-atom metals in atomically thin layers. Each metal performs its own duties and jointly benefits the whole reaction, thus achieving maximized atom utilization. Through multimetal interactions, various active centres are formed and sufficiently exposed over the metallene. When it is applied as a NO₃RR catalyst, the proof-of-concept PdCuNiCoZn high-entropy metallene delivers satisfactory performance across wide pH ranges, especially in strongly alkaline electrolytes (a maximum ammonia yield rate of 447 mg h⁻¹ mg⁻¹ and a high Faradaic efficiency of 99.0%). Through the combination of a series of in situ characterizations and theoretical calculations, the NO₃RR process over PdCuNiCoZn high-entropy metallenes is shown to predominantly follow the N pathway, with the hydrogenation of *NO as the RDS. Because the strong interaction between each metal site on the high-entropy metallene pushes its d-band centre closer to the Fermi level, the energy barrier for *NO hydrogenation is reduced at each active centre of the high-entropy metallene during the NO₃RR. Our study not only highlights the great potential of high-entropy metallenes in the NO₃RR but also facilitates the design of high-entropy materials with multiple sites and high intrinsic activity as ideal electrocatalysts. High-entropy metallenes expand the domain of “high-entropy” material families, which is also expected to promote research on other electrocatalytic reactions.

Methods

Reagents

Palladium(II) acetylacetonate (Pd(acac)₂) (99%), cobalt(II) acetylacetonate (Co(acac)₂) (98%), nickel(II) acetylacetonate (Ni(acac)₂) (95%), copper(Ⅱ) acetylacetonate (Cu(acac)₂) (97%), zinc(Ⅱ) acetylacetonate (Zn(acac)₂) (97%), N,N-dimethylformamide (DMF) (99.5%), acetic acid (99.8%), ascorbic acid (AA) (99%), tungsten hexacarbonyl (W(CO)₆) (97%), ammonium chloride (NH₄Cl) (99%), ammonium chloride-¹⁵N (¹⁵NH₄Cl) (99 atom%), potassium nitrate-¹⁵N (K¹⁵NO₃) (99 atom%), salicylic acid (C₇H₆O₃) (99.5%), sodium citrate dihydrate (C₆H₅Na₃O₇) (99%), sodium hypochlorite (NaClO) (active chlorine basis > 5.0%), sodium hydroxide (NaOH) (99.9%), sodium nitroferricyanide (III) dihydrate (Na₂Fe(CN)₅NO·2H₂O) (99.0%), p-aminobenzenesulfonamide (C₆H₇NO₃S) (99.8%), and p-dimethylamino benzaldehyde ((CH₃)₂NC₆H₄CHO) (99%) were purchased from Aladdin. Potassium sulfate (K₂SO₄) (99.0%) was purchased from Sigma‒Aldrich. Potassium nitrate (KNO₃) (99%), Nafion solution (5 wt%) and N-(1-naphthyl) ethylenediamine dihydrochloride (C₁₂H₁₄N₂·2HCl) (96%) were purchased from Alfa-Aesar. Dimethyl sulfoxide (DMSO-d₆) was purchased from Macklin. Potassium nitrite (KNO₂) (97%), hydrochloric acid (HCl) (37%) and hydrazine hydrate (N₂H₄) (98%) were purchased from Sinopharm Chemical Reagent Co., Ltd. (China). Absolute ethanol (C₂H₅OH) (99.5%) was purchased from Shanghai Lingfeng Chemical Reagent Co., Ltd. All the chemicals were commercially available and used without any further purification. Deionized water (resistance of 18 MΩ cm⁻¹) was used during the experimental process. All the chemicals used were commercially available and used without any further purification.

Synthesis of PdCuNiCoZn high-entropy metallene

First, 10 mg of Pd(acac)₂, 10 mg of Co(acac)₂, 8 mg of Ni(acac)₂, 10 mg of Cu(acac)₂, 6 mg of Zn(acac)₂, 80 mg of ascorbic acid and 60 mg of W(CO)₆ were dissolved in 16 mL of DMF and 4 mL of acetic acid, followed by ultrasonication for 1 h. Second, the transparent solution was sealed in a glass bottle and transferred to an oil bath at 90 °C for 12 h. After it was naturally cooled to room temperature, the colloidal product was washed with a mixture of ethanol and cyclohexane six times and then dried at 60 °C for future use.

Synthesis of PdCuNiCo quaternary metallenes

First, 10 mg of Pd(acac)₂, 10 mg of Co(acac)₂, 8 mg of Ni(acac)₂, 10 mg of Cu(acac)₂, 60 mg of ascorbic acid and 60 mg of W(CO)₆ were dissolved in 16 mL of DMF and 4 mL of acetic acid, followed by ultrasonication for 1 h. Second, the transparent solution was sealed in a glass bottle and transferred to an oil bath at 80 °C for 12 h. After it was naturally cooled to room temperature, the colloidal product was washed with a mixture of ethanol and cyclohexane six times and then dried at 60 °C for future use.

Synthesis of the PdCuNi ternary metallene

First, 10 mg of Pd(acac)₂, 10 mg of Co(acac)₂, 8 mg of Ni(acac)₂, 40 mg of ascorbic acid, and 60 mg of W(CO)₆ were dissolved in 16 mL of DMF and 4 mL of acetic acid, followed by ultrasonication for 1 h. Second, the transparent solution was sealed in a glass bottle and transferred to an oil bath at 80 °C for 12 h. After naturally cooling to room temperature, the colloidal product was washed with a mixture of ethanol and cyclohexane several times and then dried at 60 °C for future use.

Synthesis of PdCu bimetallene

First, 12 mg of Pd(acac)₂, 8 mg of Cu(acac)₂, and 40 mg of W(CO)₆ were dispersed in 16 mL of DMF and 4 mL of acetic acid, followed by ultrasonication for 15 min and then heating to 70 °C for 2 h. The black colloidal product was washed with ethanol six times and dried at 60 °C for future use.

Synthesis of Pd metallene

First, 12 mg of Pd(acac)₂ and 40 mg of W(CO)₆ were dispersed in 16 mL of DMF and 4 mL of acetic acid, followed by ultrasonication for 15 min and then heating to 70 °C for 2 h. The black colloidal product was washed with ethanol six times and dried at 60 °C for future use.

Physical characterizations

A morphology study of the catalysts was conducted via SEM (Regulus 8230). AC-HAADF-STEM was performed with an FEI Theims Z. AFM was performed with a Bruker Dimension Icon. ICP‒OES was performed on a Vista-MPX. XRD measurements were performed with a Bruker D8 Advance diffractometer. HRTEM, EDX mapping, and line scan analysis were performed with an FEI Tecnai F-20. XAS was performed at the Shanghai Synchrotron Radiation Facility. XPS was conducted on an ESCALAB 250Xi.

Cathode preparation

Carbon cloth (1 cm × 1 cm, treated with oxygen plasma) was used as the substrate of the working electrode. First, 1 mg of catalyst was mixed with Ketjen black at a mass ratio of 1:1, which was then ultrasonicated in 2 mL of ethanol supplemented with 10 μL of 5 wt% Nafion solution for 2 h. Second, 100 μL of the catalyst ink was uniformly cast onto the carbon cloth, which was then dried at room temperature overnight. The mass loading is 0.05 mg cm^-2. All the Pd, PdCu, PdCuNi, PdCuNiCo, and PdCuNiCoZn electrodes were prepared in this way.

Electrochemical NO₃RR measurements

The electrochemical NO₃RR measurements were conducted with a CHI-630 electrochemical workstation (Shanghai Chenhua Instrument Co., Ltd.) in an H-type electrolytic cell (resistance of 4.3 ± 0.2 Ω), which was separated by a Nafion-117 membrane (3 cm × 3 cm, thickness of 183 μm, DuPont). First, the membrane was boiled in a 5% H₂O₂ aqueous solution at 80 °C for 1 h and then soaked in ultrapure water for 30 min. Second, the membrane was subsequently boiled in a 5% H₂SO₄ aqueous solution at 80 °C for 1 h and then boiled in ultrapure water for 30 min to complete the pretreatment. Pt sheets (1 cm × 1 cm) and Ag/AgCl electrodes (saturated KCl solution) were used as the counter and reference electrodes, respectively. The reference electrode (RE) was calibrated using the electrochemical workstation by recording the potential difference between it and another RE which is solely stored and used for calibration only in a saturated potassium chloride solution. The as-prepared carbon cloths coated with the catalyst were used as the working electrode. For the electrochemical NO₃RR measurements, both the cathode and anode chambers were filled with 30 mL of 0.5 M K₂SO₄ + 0.1 M KNO₃ (pH = 7 ± 0.3) electrolyte or 1.0 M KOH + 0.1 M KNO₃ (pH = 14 ± 0.2) electrolyte. For 0.5 M K₂SO₄ + 0.1 M KNO₃, 43.5 g of K₂SO₄ and 5.05 g of KNO₃ were dissolved in deionized water and diluted to a volume of 500 mL. For 1.0 M KOH + 0.1 M KNO₃, 28 g of KOH and 5.05 g of KNO₃ were dissolved in deionized water and diluted to a volume of 500 mL. The cathodic electrolyte was purged with high-purity argon for 30 min before use. The electrolysis process was recorded under Ar atmosphere (10 sccm) with the electrolyte constant stirring of 400 rpm. In the LSV and chronoamperometric measurements, all potentials were not iR-corrected and were converted to RHE according to the following formula:

$$E\left(\mathrm{V}\mathrm{versus\; RHE}\right)=E\left(\mathrm{V}\mathrm{versus\; Ag}/\mathrm{AgCl}\right)+0.0591\times \mathrm{pH}+0.197$$ 1

The current density was normalized to the geometric electrode area (1 cm²). After the test was complete, the as-obtained electrolyte and absorption solution were diluted to a suitable concentration for ultraviolet-visible (UV‒vis) spectrophotometry and ¹H‒NMR tests.

NO₃RR product analysis

The concentration of ammonia was determined via the indophenol blue method with slight modifications. First, 4 g of NaOH, 5 g of salicylic acid, and 6.58 g of sodium citrate were dissolved in 100 mL of ultrapure water and named A. Second, 3.1 mL of NaClO was diluted to 100 mL and named B. Sodium nitroferricyanide (0.505 g) was dissolved in 50 mL of ultrapure water and named C. Last, 2 mL of the diluted electrolyte was mixed with 2 mL of A, 1 mL of B and 0.2 mL of C. After 2 h in the dark, the absorption spectrum was measured via UV‒vis, and the absorbance at 655 nm was recorded. The ammonia concentration–absorbance curve was generated beforehand via a series of (NH₄)₂SO₄/K₂SO₄ and (NH₄)₂SO₄/KOH with known ammonia concentrations (0, 0.1, 0.2, 0.4, 0.8, 1.2, and 1.6 ppm).

Nitrite was detected via the Greiss spectrophotometric method with slight modifications. First, 0.5 g of p-aminobenzenesulfonamide, 5 mL of acetic acid, and 5 mg of N-(1-naphthyl) ethylenediamine dihydrochloride were diluted to 100 mL in ultrapure water and used as chromogenic agents. Second, 1 mL of the diluted electrolyte was mixed with 4 mL of the chromogenic agent. After 15 min in the dark, the absorption spectrum was measured via UV–vis, and the absorbance at 540 nm was recorded. The nitrite concentration–absorbance curve was generated beforehand using a series of KNO₂/K₂SO₄ solutions with known nitrite concentrations (0, 0.2, 0.4, 0.8, 1.2, 1.6, and 2.0 ppm).

Hydrazine was detected via the Watt‒Chrisp method with slight modifications. A mixture of para-(dimethylamino) benzaldehyde (5.99 g), HCl (concentrated, 30 mL), and ethanol (300 mL) was used as a colour reagent. The chromogenic agent (5 mL) was added to the electrolyte (5 mL) and stirred for 10 min at room temperature. After 15 min in the dark, the absorption spectrum was measured via UV–vis, and the absorbance at 455 nm was recorded. The hydrazine concentration–absorbance curve was generated beforehand using a series of N₂H₄/K₂SO₄ solutions with known hydrazine concentrations (0, 0.2, 0.4, 0.6, 0.8, 1.2, and 2.0 ppm).

¹H-NMR determination of ammonia

In addition to the colorimetric method, the concentration of ammonia was also measured via ¹H-NMR using DMSO-d₆ as the solvent and maleic acid as the internal standard. The calibration curves were plotted via the following method:

A series of (¹⁴NH₄)₂SO₄ or (¹⁵NH₄)₂SO₄ with certain concentrations in 0.1 M K₂SO₄ (depositions would form if 0.5 M K₂SO₄ was mixed with DMSO-d₆) were used as standards. First, 5 mL of the (¹⁴NH₄)₂SO₄ or (¹⁵NH₄)₂SO₄ standard solution was mixed with 25 μL of 1 M H₂SO₄ containing 0.04 mg of maleic acid. Second, 0.5 mL of the standard solution was mixed with 0.1 mL of DMSO-d_6, and the mixture was tested with a 400 MHz Liquid Bruker Avance NMR spectrometer. Lastly, the calibration curve was obtained from the peak area ratio between NH₄⁺ and maleic acid.

K¹⁴NO₃ and K¹⁵NO₃ were subsequently used as nitrate sources in the labelling experiment. The electrolyte was collected after NO₃^- reduction at −1.2 V versus RHE for 1 h. First, the electrolyte was diluted 5 times, and its pH was also adjusted to ~7. Second, 5 mL of the electrolyte was mixed with 25 μL of 1 M H₂SO₄ containing 0.04 mg of maleic acid. Third, 0.5 mL of the above mixture was mixed with 0.1 mL of DMSO-d₆, and the mixture was tested with a 400 MHz Liquid Bruker Avance NMR spectrometer. The ¹⁴NH₄⁺ and ¹⁵NH₄⁺ concentrations were calculated according to the as-obtained calibration curves mentioned above.

Calculation of the product yield rate and Faradaic efficiency

$$\mathrm{Yield}\mathrm{rate}\left({\mathrm{NH}}_3\right)=\left[c\left({\mathrm{NH}}_3\right)\times V\right]/\left(t\times m\right)$$ 2

$$\mathrm{Faradaic}\mathrm{efficiency}\left({\mathrm{NH}}_3\right)=\left[c\left({\mathrm{NH}}_3\right)\times V\times 8F\right]/\left(17Q\right)$$ 3

$$\mathrm{Faradaic}\mathrm{efficiency}\left({\mathrm{NO}}_{2^{-}}\right)=\left[c\left({\mathrm{NO}}_{2^{-}}\right)\times V\times 2F\right]/\left(46Q\right)$$ 4

$$\mathrm{Faradaic}\mathrm{efficiency}\left({\mathrm{N}}_2{\mathrm{H}}_4\right)=\left[c\left({\mathrm{N}}_2{\mathrm{H}}_4\right)\times V\times 7F\right]/\left(32Q\right)$$ 5

where F represents the Faraday constant (96485 C mol⁻¹), c(NH₃) denotes the mass concentration of ammonia in the electrolyte, c(NO₂^-) indicates the mass concentration of nitrite in the electrolyte, c(N₂H₄) represents the mass concentration of hydrazine in the electrolyte, V denotes the volume of the cathodic electrolyte, Q indicates the total charge passing through the electrode, t represents the reaction time, and m denotes the loading mass of the catalysts. The error bars correspond to the standard deviations of measurements of three separately prepared samples under the same conditions.

ECSA analysis

The electrochemical double-layer capacitance (C_dl) was determined by analyzing cyclic voltammetry (CV) curves recorded at varying scan rates (50–100 mV s⁻¹) within the non-Faradaic potential window. C_dl was calculated as half the slope of the linear fit obtained by plotting the current density difference (mA cm⁻²) between the anodic sweep and the cathodic sweep at the centre of the scan interval against the scan rate.

$$j_c\left(\mathrm{non}-\mathrm{Faradaic}\mathrm{current}\mathrm{density}\right)={\mathrm{C}}_{\mathrm{dl}}\left(\mathrm{d}\phi \right)/\left(\mathrm{d}t\right)=\mathrm{const}$$ 6

The electrochemically active surface area was determined by normalizing the C_dl with respect to the specific capacitance of a flat reference material with a real surface area of 1 cm² (usually, the capacitance value of an ideal flat surface is 40 μF cm⁻²). The ECSA can be calculated as follows:

$${\mathrm{A}}_{\mathrm{ECSA}}={\mathrm{C}}_{\mathrm{dl}}\left(\mathrm{mF}{\mathrm{cm}}^{-2}\right)/40\mathrm{\mu }\mathrm{F}{\mathrm{cm}}^{-2}\mathrm{per}{\mathrm{cm}}^2\mathrm{ECSA}$$ 7

In situ Raman, ATR-SEIRAS, and DEMS characterizations

In situ Raman experiments were conducted via a Horiba Jobin Yvon HR evolution instrument (France) with a 532 nm laser source (Ventus LP 532) and a 50 × objective. The applied potential was controlled by an electrochemical workstation (CHI 660E). An in situ electrochemical Raman experiment was performed in a customized cell. The catalyst ink was dropped onto carbon paper. Simultaneously, the working electrode was perpendicular to the incident laser. A platinum wire and an Ag/AgCl electrode (filled with saturated KCl) were used as the counter and reference electrodes, respectively. A mixed solution (0.5 M K₂SO₄ + 0.1 M KNO₃) was used as the electrolyte. The Raman signals were collected as a function of time.

In situ ATR-SEIRAS experiments were conducted using a Thermo Nicolet 8700 spectrometer equipped with an MCT detector cooled by liquid nitrogen. The Si prism deposited with Au was soaked in a piranha solution (7:3 volumetric ratios of 98% H₂SO₄ and 30% H₂O₂) for 2 hours. Subsequently, 30 μl of catalyst ink was deposited and dried on the Au-film working electrode. The ink-coated prism was subsequently assembled into a custom-made spectroelectrochemical cell containing 1.0 M KOH + 0.1 M KNO₃ as the working electrode. Ag/AgCl was used as a reference and was introduced near the working electrode via a Luggin capillary. A Pt mesh (1 cm × 1 cm) served as the counter electrode. The spectral resolution was 4 cm⁻¹ for all the measurements. The signals were collected as a function of given potentials.

Online DEMS studies were conducted via the Shanghai Linglu QAS 100. The device was constructed by a gas interface between the electrolytic reaction chamber and the mass spectrometry analysis system through a hydrophobic polytetrafluoroethylene (PTFE) membrane. The electrolysis reaction occurred on one side of the PTFE membrane, and the product gas was quickly extracted by a vacuum pump and transported to the other side of the membrane for real-time analysis by a mass spectrometry system. During the test, the chronoamperometry mode was used. The potential of the working electrode was maintained at −0.7 V versus RHE in 0.5 M K₂SO₄ + 0.1 M KNO₃ for a specific duration to monitor the mass spectrometry signal at the corresponding m/z ratio in real time. After each test, the signal is allowed to fully return to the baseline value to ensure that the system fully recovers to its initial state, and then the same test conditions are repeated for the next round of experiments.

DFT computational method and model

A four-layer Pd(111) surface slab model was used to simulate Pd metallene. Based on the Pd metallene, the atoms in the uppermost layer were randomly substituted to simulate PdCu and PdCoNiCuZn (Supplementary data 1). The Vienna ab initio simulation package (VASP) was employed for DFT computations. Electron-electron interactions were treated using the Perdew–Burke–Ernzerhof functional within the generalized gradient approximation framework. For the plane-wave basis set, a 450 eV cut-off energy was applied, while a 3 × 3 × 1 Gamma-centered k-point mesh ensured proper convergence. Structural optimization criteria included an electronic energy convergence threshold of 10⁻⁵ eV and a maximum force tolerance of 0.02 eV Å⁻¹ during ionic relaxation. A 15 Å vacuum layer along the z-direction prevented artificial periodic interactions, and the Grimme’s DFT-D3 method with zero damping accounted for van der Waals dispersion effects. Moreover, to model aqueous environments, the VASPsol implicit solvent model was implemented. The adsorption energy ΔE of the A group on the surface of the substrate was defined as:

$$E_{\mathrm{ads}}={E*}_{+\mathrm{intermediate}}-E*-E_{\mathrm{molecule}}$$ 8

where E_molecule is the molecule energy in gas phase, while E_{* + intermediate} and E_* are the energies of the surface with/without adsorbed molecules. The Gibbs free energy change (ΔG) was defined as:

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

where ΔE represents the energy difference between products and reactants; T and ΔS indicate room temperature and the entropy change, respectively; ΔZPE denotes the zero-point energy correction; and ΔG_pH indicates the contribution of H⁺ and is expressed as –k_BT×ln(10)×pH, where k_B is the Boltzmann constant.

Author contributions

M.W., C.Y. and J.L. conceived and designed this work. Y.Z. prepared and characterized the material and performed the experiments. L.Z. conducted the theoretical calculations. X.S., Z.Z. and T.Q. discussed the results and commented on the manuscript. Y.Z. and L.Z. wrote the paper; M.W., C.Y. and J.L. revised the paper.

Peer review

Peer review information

Nature Communications thanks Chang Hyuck Choi, Jingjing Duan, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data supporting the findings of this study are available within the article and the Supplementary Information 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/s41467-025-63317-1.