Unraveling the electronegativity-dominated intermediate adsorption on high-entropy alloy electrocatalysts — TEI-tagged view

Jiace HaopersName, Zechao ZhuangpersName, Kecheng CaopersName, Guohua GaopersName, Chan WangpersName, Feili LaipersName, Shuanglong LupersName, Piming MapersName, Weifu DongpersName, Tianxi LiupersName, Mingliang DupersName, Han ZhupersName (2022date). “Unraveling the electronegativity-dominated intermediate adsorption on high-entropy alloy electrocatalysts.” Nature Communicationstitle·j.

Object of study — FeCoNiXRu high-entropy alloyterm·dom nanoparticles as electrocatalyststerm·key with multiple active sitesterm·key

Research question — how multiple active sitesterm·key in HEAsterm·dom relate to reaction intermediates

Contribution — an understanding of the correlation between specific active sitesterm·key and intermediates for overcoming scaling relationsterm·dom

Full text — tagged

The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.

Unraveling the electronegativity-dominated intermediate adsorption on high-entropy alloyterm·dom electrocatalyststerm·key Abstract High-entropy alloysterm·key have received considerable attention in the field of catalysis due to their exceptional properties. However, few studies hitherto focus on the origin of their outstanding performance and the accurate identification of active centers. Herein, we report a conceptual and experimental approach to overcome the limitations of single-element catalysts by designing a FeCoNiXRu (X: Cu, Cr, and Mnterm) High-entropy alloysterm·key system with various active sitesterm·key that have different adsorption capacities for multiple intermediates. The electronegativityterm·dom differences between mixed elements in HEAterm·dom induce significant charge redistributionterm·dom and create highly active Co and Ru sitesterm·key with optimized energy barriersterm·dom for simultaneously stabilizing OH * and H * intermediates, which greatly enhances the efficiency of water dissociationterm·key in alkaline conditions. This work provides an in-depth understanding of the interactions between specific active sitesterm·key and intermediates, which opens up a fascinating direction for breaking scaling relationterm·dom issues for multistep reactions.

Introduction Electrocatalytic reactions hold the key to the efficiency of many energy conversion and storage systems, which highly require effective and stable electrocatalyststerm·key to tune the rates (selectivity) of reaction pathways during catalytic process. A great challenge facing materials scientists arises from the ongoing need for advanced electrocatalyststerm·key with reasonable activities that meet the needs of rapid development 1 – 3 . Recently, high-entropy alloyterm·dom (HEAterm·dom) catalysts with near-equiatomic proportions, have attracted wide attention due to their unprecedented properties caused by high configuration entropy, lattice distortion, sluggish diffusion, and cocktail effectsterm·dom 4 – 8 . A series of nanoscale HEA catalyststerm·dom have been applied for ammonia oxidation, water splitting, methanol oxidation, CO 2 reduction, and dye degradation, and they deliver superior performance than conventional alloysterm·dom 7 – 13 . Infinite elemental combinations and unconventional compositions offer many possibilities for regulating catalytic performance and overcoming the limitations of single-element catalysts, making HEAsterm·dom an intriguing option in the field of electrocatalysis 11 .

The catalytic activities of monometallic catalysts depend substantially on the adsorption of reactants on surface-active sites 14 . Scaling relationterm·dom issues applicable to many multistep reactions require the stabilization of multiple intermediates during the reaction, but the limited number of active sitesterm·key with monometallic catalysts cannot achieve simultaneous stabilization of all intermediates on the available active sitesterm·key with optimal binding energies 15 , 16 . Nano-catalysts have been rapidly developed in recent years with effectively explored strategies (including alloying, nanostructuring, defect, heterostructure, and lattice strain) to reveal the active sitesterm·key and cooperation with each other (Fig. 1a ) 17 – 23 . Single atom catalyststerm·dom (SACsterm·dom) have attracted tremendous effort to exploit effective routes with good control over coordination environment, composition (dimer), metal loading, and substrate, which are needed for the discovery of the correlations between compositional engineering and optimization of catalytic behavior (Fig. 1b ) 24 . Numerous studies have been devoted to focusing on active sitesterm·key and mechanisms of nano catalysts and SACsterm·dom, and however, as an equally important catalyst system, accurate identification of active centers and their activity origins in HEA catalyststerm·dom are greatly neglected. Furthermore, the relationships between multiple active sitesterm·key in HEAsterm·dom and reaction intermediates are still obscure with the lack of guidelines for designing reasonable active sitesterm·key. Fig. 1 Illustration of the concepts for designing nano catalysts, SACsterm·dom and HEA catalyststerm·dom. Designed strategies for creating active sitesterm·key in ( a ) nano catalysts and ( b ) SACsterm·dom. c Schematic illustration of HEAterm·dom electrocatalyststerm·key with identified electronegativity-dependent preferences for active siteterm·key adsorption of the intermediates OH * and H * during H 2 O dissociation and H 2 production steps. d Adjustments of the HERterm·dom activities of HEAterm·dom electrocatalyststerm·key by tailoring the electronegativityterm·dom of the composition. e Identifying active sitesterm·key of HEAterm·dom for stabilization of intermediates by operando electrochemical Raman spectra.

Herein we propose a conceptual and experimental approach to overcome the limitations of single-element catalysts by designing a FeCoNiXRu (X: Cu, Cr, and Mnterm) HEAterm·dom system with two kinds of active sitesterm·key that have different adsorption capacities for multiple intermediates. HEAterm·dom NPs are synthesized in electrospun carbon nanofibersterm (CNFsterm) and displayed a thermodynamically driven phase transition, as revealed by in situ characterization. We employ the HEAsterm·dom as electrocatalyststerm·key for the alkaline hydrogen evolution reactionterm·dom (HERterm·dom) and successfully identify electronegativity-dependent preferences for active siteterm·key adsorption of the intermediates OH * and H * during H 2 O dissociation and H 2 production steps (Fig. 1c ). In the FeCoNiMnRu NP, the Co sitesterm·dom are the most active centers with the lowest energy barrierterm·dom (0.34 eV) for water dissociationterm·key. During the subsequent H * adsorption/desorption process, H * leaves Co and is absorbed on Ru sitesterm·key, which exhibit the lowest Δ G H* of −0.07 eV. Adjustments of the HERterm·dom activities of HEA catalyststerm·dom are shown experimentally and theoretically by tailoring the electronegativityterm·dom of the composition (Fig. 1d and e ). Constructing controllable multifunctional active sitesterm·key on HEAterm·dom surfaces allows for different interactions with related intermediates, which is a powerful option for solving this scaling relationterm·dom challenge.

Results Synthesis and materials characterization The model electrocatalyststerm·key, single-phase FeCoNiXRu solid solution HEAterm·dom NPs (X = Cr, Mnterm, and Cu), were synthesized in carbon nanofibersterm (FeCoNiXRu/CNFsterm) through a polymer nanofiber reactor strategy by combining the electrospinning technology and graphitization process. The schematic illustration of the synthetic procedure for FeCoNiXRu/CNFsterm (X = Cr, Mnterm, and Cu) are shown in Supplementary Fig. 1 and the details are described in the Experimental section.

As revealed by the field emission scanning electron microscopy (FE-SEM) image in Fig. 2a , large amounts of FeCoNiMnRu NPs were densely and uniformly anchored in CNFsterm, and intertwined CNFsterm with diameters ranging from 100 to 200 nm exhibited porous three-dimensional (3D) networks. The transmission electron microscopy (TEM) image (Fig. 2b ) also displays the uniform distribution of FeCoNiMnRu NPs with an average diameter of approximately 14.2 ± 9.1 nm (Supplementary Fig. 2 ). Figure 2c and d display distinctly visible lattice fringes with interplanar crystal spacings of 2.1 and 1.8 Å, corresponding to the (111) and (200) facets. All interplanar spacings were determined by measuring the total distances of 20 successive corresponding planes (Supplementary Fig. 3 ). In addition, the fast Fourier transform (FFT) pattern (inset in Fig. 2c ) further reveals the face-centered cubic ( fcc) crystal structures of FeCoNiMnRu HEAterm·dom NPs and exhibits the presence of typical (111), (200), and (220) planes. High-angle annular dark-field STEM (HAADF-STEM) and STEM energy dispersive X-ray (STEM-EDX) elemental mapping images (Fig. 2e ) show the homogeneous distribution of elements Fe, Co, Ni, Mnterm, and Ru in a single FeCoNiMnRu HEAterm·dom NP. Additionally, the line-scan EDX spectra (Supplementary Fig. 4 ) also reveal the distribution of Mnterm, Fe, Co, Ni, and Ru elements throughout the whole HEAterm·dom NP, further demonstrating the formation of homogeneous structures in FeCoNiMnRu HEAterm·dom. The mapping area contains 10 HEAterm·dom NPs (Supplementary Fig. 5 ) and the line-scan STEM-EDX of 3 HEAterm·dom NPs (Supplementary Fig. 6 ) have also been conducted. The results exhibit the uniform distribution of Fe, Co, Ni, Mnterm, and Ru elements among all the HEAterm·dom NPs, further suggesting the repeatability of HEAterm·dom NPs with uniform composition distribution. The content of each element in FeCoNiMnRu/CNFsterm was measured by inductively coupled plasma-optical emission spectrometry (ICP-OES). The composition of HEAterm·dom was calculated to be Fe 0.23 Co 0.22 Ni 0.22 Mnterm 0.14 Ru 0.19 (Supplementary Table 1 ). A calculated mixing entropy of ΔS > 1.59 R was determined from ICP-OES results, suggesting the intrinsic nature of HEAsterm·dom without phase separation. The FeCoNiXRu/CNFsterm (X = Cr and Cu) (Supplementary Fig. 7 and Supplementary Tables 2 , 3 ) with similar Ru contents and control samples of FeCoNi/CNFsterm, FeCoNiMn/CNFsterm, and FeCoNiRu/CNFsterm (Supplementary Fig. 8 ) were prepared using the same approach and they all exhibited morphologies similar to that of FeCoNiMnRu/CNFsterm. Fig. 2 Morphological and structural characterization of FeCoNiMnRu/CNFsterm. a FE-SEM, b TEM and c , d HRTEM images of FeCoNiMnRu/CNFsterm. The inset in ( c ) is the corresponding FFT pattern of a FeCoNiMnRu HEAterm·dom NP. e HAADF-STEM and the corresponding STEM-EDX mapping images of a FeCoNiMnRu HEAterm·dom NP supported on CNFsterm. f XRD patterns of FeCoNiMnRu/CNFsterm with detailed Rietveld refinements. g Real-time in situ XRD patterns for FeCoNiMnRu/CNFsterm with temperatures ranging from 25 to 1000 °C. h The corresponding enlarged in situ XRD patterns of FeCoNiMnRu/CNFsterm. The heating rate was kept at 30 °C min -1 during the whole process. i The thermodynamically driven phase transition of FeCoNiMnRu NPs in CNFterm nanofiber reactors.

The crystalline structures of FeCoNiMnRu/CNFsterm (Fig. 2f ) prepared at 1000 °C under 3 h treatment and the corresponding control samples (Supplementary Fig. 9 ) were investigated with X-ray diffraction (XRD) patterns. As shown in Fig. 2f , the FeCoNiMnRu/CNFsterm exhibits three main diffraction peaks at 2θ = 43°, 50° and 74°, which can be indexed to the (111), (200), and (220) planes of the fcc phases (PDF#47-1417), respectively. No separated XRD peaks from Fe, Co, Ni, Mnterm, Ru, or metal oxides were observed, suggesting the formation of a single-phase HEAterm·dom. In addition, the corresponding XRD detailed Rietveld refinements also confirm the single phase HEAterm·dom structure of FeCoNiMnRu NPs. Compared with the standard line patterns of the FeNi alloy (PDF#47-1417), the fcc diffraction peaks of the FeCoNiMnRu HEAterm·dom shifted slightly to lower angles due to the lattice distortions caused by the incorporation of Ru, Mnterm and Fe atoms and the resulted high entropy 4 .

All of the XRD patterns of FeCoNi/CNFsterm, FeCoNiMn/CNFsterm and FeCoNiRu/CNFsterm (Supplementary Fig. 9 ) exhibit the characteristic peaks for the (111), (200), and (220) planes of the fcc phase, suggesting that the fcc crystal structures can be well maintained after changing in the numbers of elements. Furthermore, with the incorporation of Mnterm and Ru atoms along with Fe, Co and Ni atoms, the peak positions of the (111) planes for FeCoNiMn/CNFsterm, FeCoNiRu/CNFsterm, and FeCoNiMnRu/CNFsterm gradually move to lower angles, suggesting a strong high−entropy effects 6 . As shown in Supplementary Fig. 10 , the XRD patterns of FeCoNiXRu/CNFsterm (X = Cr, Mnterm and Cu) confirm the fcc structures, and the diffraction peaks exhibit slight differences caused by the different compositions.

The temperature-dependent in situ XRD patterns are illustrated in Fig. 2g with temperatures range from 25 to 1000 °C. As shown in Fig. 2g , only taenite (FeNi alloy, PDF#47−1417) is observed after treated after 600 °C, with (111) planes at 2θ = 43.8°. The FeCoNiMnRu HEAterm·dom fcc phase and Mnterm 3 Co 7 phase (marked as #, PDF#18–0407) coexisted between 800 and 1000 °C, where the HEAterm·dom becomes the dominant phase. The fraction of the HEAterm·dom fcc phase increases with the increase of equilibration temperature, while the fraction of the Mnterm 3 Co 7 phase decreased from 800 to 1000 °C, suggesting that more Mnterm and Co atoms diffused into the HEAterm·dom fcc crystal lattice to produce a near-equimolar mixture of component by way of an effect driven by thermodynamics. During annealing at 1000 °C for 3 h, full conversion to single-phase HEAterm·dom occurred without observation of additional peaks, suggesting the complete formation of FeCoNiMnRu HEAterm·dom. Figure 2h clearly shows that the peak positions of (111) planes for HEAterm·dom initially shifted to lower 2θ angles between 800 and 1000 °C and the asymmetry of diffraction peak strengthened with increased temperatures, suggesting the generation of a larger lattice distortion caused by differences in the atomic radii 10 . Then, the prolonged annealing treatment can reduce the lattice distortion of the HEAterm·dom crystal, as evidenced by the positively shifted peak position and the enhanced symmetry of diffraction peaks. In regard of (200) and (220) planes (Supplementary Fig. 11 ), the peak shifts display the same trend as that of (111) planes.

We proposed a possible growth process of HEAterm·dom NPs and the thermodynamically driven phase transition of FeCoNiMnRu NPs in CNFterm nanofiber reactors is illustrated in Fig. 2i . During graphitization, the Fe/Co/Ni/Mnterm/Ru mixed metal precursors decomposed first, and then the reduced metal clusters were bonded and confined within the PAN-derived CNFsterm. At relative low temperature 600–800 °C, the metal elements with small atom radii differences prefer to form alloy phase and insufficient heating energy at low temperature cause the slightly atom diffusion, which make both of HEAterm·dom and Mnterm 3 Co 7 phases co-exist. At high temperature 1000 °C, sufficient dynamic energy caused the metal atoms to diffuse dramatically, leading to homogeneous formation of single-phase HEAterm·dom alloy. It is concluded that the high temperature coupled with prolonged heating treatment provided the activation energy that drove complete mixing of multiple metal element atoms. The XRD patterns of FeCoNiMnRu/CNFsterm synthesized at 800, 900, and 1000 °C under prolonged heat treatment for 3 h were also performed. As shown in Supplementary Fig. 12 , compared with the in situ XRD patterns of FeCoNiMnRu/CNFsterm without prolonged heat treatment, all the diffraction peaks for fcc HEAterm·dom NPs ((111), (200), (220) planes) exhibit positively shifts to high values, suggesting the reduced lattice parameters and lattice distortion. It is indicated that after the prolonged heat treatment for 3 h, all of the diffraction peaks for Mnterm 3 Co 7 phase vanished, suggesting the complete formation of FeCoNiMnRu HEAterm·dom. Figure 2f and Supplementary Fig. 13 show the XRD patterns of FeCoNiMnRu/CNFs-800-3h and FeCoNiMnRu/CNFs-1000-3h with detailed Rietveld refinements. Both of the FeCoNiMnRu HEAterm·dom NPs obtained at 800 °C and 1000 °C under 3 h prolonged treatment was single phase structure with similar component. Therefore, negatively shifted peak position through in situ XRD results from 800 to 1000 °C suggest the growth process of single-phase HEAterm·dom (Fig. 2g ). The XRD results of FeCoNiMnRu/CNFsterm with prolonged heat treatment from 1000 °C to 1000-3 h exhibit a positively shifted peaks position, further show a structural symmetry optimization by reducing the lattice distortion in HEAterm·dom NPs.

The X-ray absorption spectroscopyterm·mth (XASterm·mth) (Fig. 3a–d ) was performed to investigate the chemical states of FeCoNiMnRu/CNFsterm. The X-ray absorption near-edge structure (XANES) results (Fig. 3a and c ) indicate that the pre-absorption edge features for Co and the absorption edge for Ru both are metallicity by comparing with reference metal Co, CoO and Co 3 O 4 foils, demonstrating that the Co and Ru elements in HEAterm·dom NPs are in metallic state. The post-edge for Co and Ru in HEAterm·dom exhibits slight deviation in the shape and intensity when compared with the reference metal Co and Ru foils. These features indicate the alloy formation rather than elemental segregation into pure metals, which would show the same length as metal Co and Ru foils. The extended X-ray absorption fine structure (EXAFS) of Co and Ru were determined through the fitting of the Fourier transform (FT) spectra. As shown in Fig. 3b and d , the FT-EXAFS spectra indicate that the average bond length of Ru and Co in HEAterm·dom NPs is quite different from the metallic bond in bulk Co and Ru references, suggesting that the Co and Ru elements are surrounded by different metallic species (Fe, Mnterm and Ni). The bond structures of Co and Ru in HEAterm·dom reveal the similar average bond length without any oxidation when compared CoO and Co 3 O 4 foils, further confirming the metallic states of Co and Ru in HEAterm·dom after stability test. According to the EXAFS fitting (Supplementary Fig. 14 ), the bond length (R) and coordination numbers of each bond type in the HEAterm·dom were summarized in Supplementary Table 4 . The reliability of the fitting method is supported by smaller R factors. Fig. 3 Surface chemical states characterized by X-ray absorption spectroscopyterm·mth (XASterm·mth) and X-ray photoelectron spectroscopyterm·mth (XPSterm·mth). a The Co K-edge XANES spectra and b FT-EXAFS spectra of FeCoNiMnRu/CNFsterm, Co foil, CoO foil, and Co 3 O 4 foil. c The Ru K-edge XANES spectra and d FT-EXAFS spectra of FeCoNiMnRu/CNFsterm, Ru foil, and RuO 2 foil. High-resolution XPSterm·mth spectra of as-prepared FeCoNiMnRu/CNFsterm: e Fe 2 p , f Co 2 p , g Ni 2 p , h Mnterm 2 p , and i Ru 3 p .

X-ray photoelectron spectroscopyterm·mth (XPSterm·mth) was utilized to investigate the surface compositions and electronic effects of the as-prepared FeCoNiMnRu/CNFsterm. The XPSterm·mth survey spectrum of the as-prepared FeCoNiMnRu/CNFsterm is shown in Supplementary Fig. 15 . Figure 3e shows the presence of Fe 0 , Fe 2+ and Fe 3+ species, and the BEs at 707.1, 711.5, 724.8 eV, 714.3 and 727.2 eV are ascribed to Fe 0 2 p 3/2 , Fe 2+ 2 p 3/2 , Fe 2+ 2 p 1/2 , Fe 3+ 2 p 3/2 and Fe 3+ 2 p 1/2 , respectively. Another peak with a BE at 720.6 eV is attributed to the satellite peaks of Fe 2 p 25 . For the Co 2 p spectrum (Fig. 3f ), the main doublets at 777.8, 792.7, 779.7, 795.1, 782.2 and 796.6 eV correspond to Co 0 2 p 3/2 , Co 0 2 p 1/2 , Co 3+ 2 p 3/2 , Co 3+ 2 p 1/2 , Co 2+ 2 p 3/2 and Co 2+ 2 p 1/2 , respectively 26 . The Ni 2 p spectrum (Fig. 3g ) displays the coexistence of metallic Ni 0 (852.6 and 869.9 eV for Ni 0 2 p 3/2 and Ni 0 2 p 1/2 ), Ni 2+ (854.5 and 871.8 eV for Ni 2+ 2 p 3/2 and Ni 2+ 2 p 1/2 ), and Ni 3+ (856.7 and 874.2 eV for Ni 3+ 2 p 3/2 and Ni 3+ 2 p 1/2 ) 27 . The high-resolution Mnterm 2 p spectrum (Fig. 3h ) suggests the coexistence of Mnterm 2+ 2 p 3/2 (641.6 eV), Mnterm 2+ 2 p 1/2 (653.1 eV), Mnterm 3+ 2 p 3/2 (643.4 eV) and Mnterm 3+ 2 p 1/2 (654.6 eV). Two satellite peaks (marked as “Sat.”) appear at binding energies (BEs) of 648.5 and 659.7 eV 28 .

The Ru 3 p XPSterm·mth spectrum of FeCoNiMnRu/CNFsterm (Fig. 3i ) exhibit only metallic Ru states (462.0 and 484.1 eV for Ru 0 3 p 3/2 and Ru 0 3 p 1/2 ) 29 . The O 1 s spectrum (Supplementary Fig. 16 ) reveals peaks at 531.8 and 532.8 eV, which are ascribed to hydroxyl groups and residual oxygen-containing groups on the surface of FeCoNiMnRu/CNFsterm, respectively 30 . There are no peaks observed for lattice O 2− of metal oxides, and the strong adsorption of hydroxyl groups would be beneficial for water splitting. The Fe 2 p , Co 2 p , Ni 2 p , Mnterm 2 p and Ru 2p XPSterm·mth spectra of the controlled samples including Ru/CNFsterm, FeCoNi/CNFsterm, FeCoNiMn/CNFsterm, and FeCoNiRu/CNFsterm were performed to provide more electron interaction information among the metal elements in HEAterm·dom (Supplementary Fig. 17 ). The corresponding BE information were summarized in Supplementary Table 5 . The BEs for Co, Ni, Fe and Mnterm in FeCoNiMnRu/CNFsterm show negative shifts when compared with FeCoNi/CNFsterm, FeCoNiRu/CNFsterm and FeCoNiMn/CNFsterm, respectively (Supplementary Fig. 17a–d ). As shown in Supplementary Fig. 17e , the BEs for Ru in FeCoNiMnRu/CNFsterm exhibit positive shift when compared with the FeCoNiRu/CNFsterm and Ru/CNFsterm, suggesting that the Ru in HEAterm·dom NPs served as the electron acceptor. The results strongly demonstrate the electron transfers from Fe, Co, Ni and Mnterm atoms to Ru atoms in HEAterm·dom NPs that is due to the higher electronegativityterm·dom of Ru (2.20) than those of Ni (1.91), Co (1.88), Fe (1.83), and Mnterm (1.55) 31 .

Evaluation of electrochemical performance To evaluate the electrochemical performance, all of the as-prepared electrocatalyststerm·key were used as the working electrodes in a typical three-electrode system. A saturated calomel electrode (SCE) electrode and graphite rod were used as reference and counter electrodes, respectively. The HERterm·dom, oxygen evolution reactionterm·dom (OERterm·dom) and overall water splittingterm·dom (OWSterm·dom) reaction were performed in Ar-saturated 1.0 M KOH electrolyte, and all potentials were calibrated with a reversible hydrogen electrodeterm·mth (RHEterm·mth). As shown in Fig. 4a , the FeCoNiMnRu/CNFsterm achieves the lowest overpotentials of 71 mV to produce a current density of 100 mA cm −2 and a Tafel slope of 67.4 mV dec −1 (Supplementary Fig. 18 ), which are much more excellent than the indicated values for Ru/CNFsterm (421 mV and 337.1 mV dec −1 ), FeCoNi/CNFsterm (410 mV and 180.9 mV dec −1 ), FeCoNiMn/CNFsterm (329 mV and 162.3 mV dec −1 ), and FeCoNiRu/CNFsterm (122 mV and 120.9 mV dec −1 ). The commercial Pt/C cannot support a current density of 100 mA cm −2 below −0.2 V, suggesting the excellent HERterm·dom activity of HEAsterm·dom under alkaline conditions. The low Tafel slope of FeCoNiMnRu/CNFsterm (67.4 mV dec −1 ) indicates the operation of the Volmer–Heyrovsky mechanismterm·dom 22 . The overpotentials at 100 mA cm –2 and Tafel slopes of the as-prepared electrocatalyststerm·key are summarized in Fig. 4b . The remarkable electrocatalytic activity is further supported by electrochemical impedance spectroscopyterm·mth (EISterm·mth) performed at an overpotential of 50 mV. As shown in Supplementary Fig. 19 , the Nyquist plots of the as-prepared electrodes exhibit the characteristic semicircles. The FeCoNiMnRu/CNFsterm presents the smallest charge transfer resistance (R ct ) value of 11.4 Ω compared to the FeCoNi/CNFsterm (637.9 Ω), FeCoNiMn (950.1 Ω) and FeCoNiRu/CNFsterm (18.6 Ω), which served to accelerate the sluggish reaction kinetics. Fig. 4 Electrochemical performance of HEAterm·dom electrocatalyststerm·key. a HERterm·dom polarization curves, b the corresponding histogram for overpotentials at 100 mA cm –2 and Tafel slopes obtained for Ru/CNFsterm, FeCoNi/CNFsterm, FeCoNiMn/CNFsterm, FeCoNiRu/CNFsterm, FeCoNiMnRu/CNFsterm, and Pt/C in 1.0 M KOH electrolyte. c ECSA-normalized HERterm·dom polarization curves of the as-prepared electrocatalyststerm·key. d OERterm·dom polarization curves for the as-prepared electrocatalyststerm·key, commercial RuO 2 and IrO 2 in 1.0 M KOH electrolyte. e Comparison of HERterm·dom and OERterm·dom overpotentials at 10 mA cm –2 in 1.0 M KOH for different catalysts. Those recently reported literatures for HERterm·dom and OERterm·dom electrocatalyststerm·key were shown in Supplementary Tables 6 and 7 . f Polarization curves for full water splitting by the as-prepared electrocatalyststerm·key in a two-electrode configuration at a scan rate of 2 mV s –1 . g Long-term stability measurement of the FeCoNiMnRu/CNFsterm electrode at −1.16 V vs. RHEterm·mth in 1 M KOH electrolyte for 600 h. The insets in ( g ) are the XRD patterns (left) and STEM-EDS mapping images (right) of FeCoNiMnRu/CNFsterm after the long-term stability test.

To evaluate the active sitesterm·key and the intrinsic activities of the as-prepared electrocatalyststerm·key, electrochemical surface areas (ECSAs) were measured by a double-layer capacitance (C dl ) method 32 . As shown in Supplementary Fig. 20 , the C dl values for FeCoNi/CNFsterm, FeCoNiMn/CNFsterm, FeCoNiRu/CNFsterm, and FeCoNiMnRu/CNFsterm are 16.4, 62.1, 110.5, and 104.6 mF. The ECSA–normalized LSV curves (Fig. 4c ) show that the intrinsic activity of FeCoNiMnRu/CNFsterm at 0.01 mA cm −2 (ECSA) is 85 mV, which is remarkably enhanced relative to the values of FeCoNi/CNFsterm (341 mV), FeCoNiMn/CNFsterm (309 mV), and FeCoNiRu/CNFsterm (146 mV), suggesting the high intrinsic HERterm·dom activity of HEAterm·dom. Figure 4d shows the OERterm·dom LSV curves of the as-prepared electrodes. The FeCoNiMnRu/CNFsterm requires the lowest overpotential of 308 mV to reach 100 mA cm −2 , as shown by the indicated values for Ru/CNFsterm (564 mV), FeCoNi/CNFsterm (382 mV), FeCoNiMn/CNFsterm (344 mV), and FeCoNiRu/CNFsterm (318 mV). The remarkably enhanced OERterm·dom kinetics of the FeCoNiMnRu HEAterm·dom are reflected by the low Tafel slope of 61.3 mV dec −1 (Supplementary Fig. 21 ).

Comparisons of the overpotentials and Tafel slopes for different electrocatalyststerm·key are summarized in Supplementary Fig. 22 . The OERterm·dom results exhibit a trend similar to that for HERterm·dom activity in that HEAterm·dom shows superior activity. As shown in Supplementary Fig. 23 , the HERterm·dom and OERterm·dom LSV curves normalized by geometric area and mass loading of noble metal indicate that the FeCoNiMnRu/CNFsterm exhibit outstanding electrocatalytic activities and even prominently surpass the state-of-art Pt/C (20 wt%), IrO 2 and RuO 2 electrocatalyststerm·key at high current density. At overpotential of 300 mV for HERterm·dom the FeCoNiMnRu/CNFsterm show higher mass activity of 2666 than that of Pt/C (1688 mA mg −1 Pt ). In addition, at overpotential of 450 mV for OERterm·dom, the FeCoNiMnRu/CNFsterm obtain the highest mass activity of 487 mA mg −1 Ru , which is significantly higher than RuO 2 (146 mA mg −1 Ru ), and IrO 2 (49 mA mg −1 Ir ). Furthermore, the overpotentials required for a current density of 10 mA cm −2 are compared with those recently reported HERterm·dom, OERterm·dom, and OWSterm·dom electrocatalyststerm·key in alkaline electrolytes (Fig. 4e , Supplementary Tables 6 – 8 ), which demonstrate the excellent activity of FeCoNiMnRu/CNFsterm.

In addition, the electrocatalytic activities of Ru-containing electrocatalyststerm·key have been further evaluated by comparing a series of FeCoNiMnRu/CNFsterm with different Ru contents. The FeCoNiMnRu/CNFsterm with different Ru contents were denoted as FeCoNiMnRu 0.5 /CNFsterm, FeCoNiMnRu/CNFsterm, and FeCoNiMnRu 2 /CNFsterm, and the corresponding Ru contents were 2.41, 4.23 and 7.13 wt%, respectively, which were measured by ICP-OES. As shown in Supplementary Fig. 24 , it is shown that the FeCoNiMnRu/CNFsterm and FeCoNiMnRu 2 /CNFsterm display very similar values of overpotentials and Tafel slopes at 100 mA cm −2 , while FeCoNiMnRu 0.5 /CNFsterm show obviously inferior activity. The ECSA values of FeCoNiMnRu/CNFsterm, FeCoNiMnRu 0.5 /CNFsterm and FeCoNiMnRu 2 /CNFsterm were calculated to be 2614, 2818, and 2093 cm 2 for FeCoNiMnRu/CNFsterm, FeCoNiMnRu 0.5 /CNFsterm, and FeCoNiMnRu 2 /CNFsterm, respectively. The R ct values of FeCoNiMnRu/CNFsterm, FeCoNiMnRu 0.5 /CNFsterm, and FeCoNiMnRu 2 /CNFsterm were measured to be 11.4, 16.3, and 5.1 Ω, respectively. The OERterm·dom activities were also evaluated by LSV curves. The overpotentials at 100 mA cm −2 of FeCoNiMnRu/CNFsterm, FeCoNiMnRu 0.5 /CNFsterm, and FeCoNiMnRu 2 /CNFsterm were determined to be 308, 324, and 299 mV, respectively. The results indicate that FeCoNiMnRu/CNFsterm achieved higher current density than that of FeCoNiMnRu 2 /CNFsterm at high overpotential (>300 mV). Therefore, at relative low Ru content range (2–4 wt%), high Ru contents would lead to enhanced activity for water splitting. When the Ru contents increased to very high values (>4 wt%), the Ru contents could pose negligible or negative effects on improving the electrocatalytic activity. Inspired by the above results, we further fabricated an alkaline electrolyzer by employing FeCoNiMnRu/CNFsterm as both the anode and cathode to explore practical electrolytic applications (Fig. 4f ). Interestingly, the FeCoNiMnRu/CNFsterm||FeCoNiMnRu/CNFsterm system requires only a low voltage of 1.65 V at 100 mA cm −2 , which is lower than those of FeCoNiMn/CNFsterm||FeCoNiMn/CNFsterm (1.93 V), FeCoNiRu/CNFsterm||FeCoNiRu/CNFsterm (1.71 V), and Pt/C||RuO 2 (Supplementary Fig. 25 ).

The electrochemical durability of FeCoNiMnRu/CNFsterm was characterized by LSV, CV cycles, and chronoamperometry measurements. Supplementary Fig. 26 shows the LSV curves of FeCoNiMnRu/CNFsterm before and after 10000 CV cycles; they nearly overlap with each other, suggesting the superior stability of FeCoNiMnRu/CNFsterm. The chronoamperometric curve (Fig. 4g ) for FeCoNiMnRu/CNFsterm was measured at −1.16 V vs. RHEterm·mth for more than 600 h in 1.0 M KOH. The current density at 1 A cm −2 displays no evident changes, also suggesting its remarkable stability. This is ascribed to the wondrous corrosion resistance of HEAterm·dom structures. The OERterm·dom stability of FeCiNiMnRu/CNFsterm was conducted at 1.55 V vs. RHEterm·mth for 10 h (Supplementary Fig. 27 ). The current density showed no obvious decay during the test, and furthermore, the LSV curves before and after stability display negligible change. In addition, as shown in Supplementary Fig. 28 , at 60 °C, the FeCoNiMnRu/CNFsterm can maintain a high current density of 1000 mA cm −2 at −2.22 V vs. RHEterm·mth for 100 h, suggesting no obvious degradation in current density. In 10 M KOH, the FeCoNiMnRu/CNFsterm also can afford a high current density of 1000 mA cm −2 at −0.77 V vs. RHEterm·mth for 100 h without current density degradation (Supplementary Fig. 29 ).

As illustrated in the FE-SEM, TEM and HRTEM images of the FeCoNiMnRu/CNFsterm electrode after the long-term stability test (Supplementary Fig. 30 ), the electrode can well maintain its initial nanoparticle morphology comprising HEAterm·dom and 3D nanofiber networks. XRD patterns (inset in Fig. 4g ) of FeCoNiMnRu/CNFsterm obtained before and after the stability test show that the HEAterm·dom can also retain the same fcc structure seen for the initial structure without any newly formed phases, suggesting ultrastable HERterm·dom performance with an alkaline electrolyte. STEM-EDS element mapping images (inset in Fig. 4g ) confirm the lack of phase separation and the homogeneous distribution of Fe, Co, Ni, Mnterm, and Ru in HEAterm·dom NPs after the stability test.

We further used the XASterm·mth to investigate the chemical states of FeCoNiMnRu/CNFsterm before and after the HERterm·dom stability test. As shown in Fig. 5a–f , the pre-absorption edge features for Co and the absorption edge for Ru both are metallicity by comparing with reference metal Co and Ru foils, demonstrating that the Co and Ru elements in HEAterm·dom NPs are in metallic state. Meanwhile, after the stability test, the XANES of Co and Ru in HEAterm·dom NPs also keep metallic state, suggesting the excellent oxidation resistance during the long-term HERterm·dom stability test. The FT-EXAFS spectra (Fig. 5b and e ) indicate that the bond structure of Co and Ru in HEAterm·dom before and after stability test reveal the similar average bond length without any oxidation when compared CoO and Co 3 O 4 foils (Fig. 3a–d ), further confirming the metallic states of Co and Ru in HEAterm·dom after stability test. According to the EXAFS fitting (Fig. 5c and f ), the bond length (R) and coordination numbers of each bond type in the HEAterm·dom before and after stability test were summarized in Supplementary Table 4 . The FT-EXAFS and WT-EXAFS spectra (Fig. 5g–i ) of Co and Ru in HEAterm·dom before and after stability tests have negligible mismatch, which means that the FeCoNiMnRu HEAterm·dom keep metallic states in long-term stability tests, exhibiting extraordinary durability. The reliability of the fitting method is supported by smaller R factors. Fig. 5 Stability characterization by XASterm·mth. a The Co K-edge XANES spectra and b FT-EXAFS spectra of FeCoNiMnRu/CNFsterm before and after stability test, and Co foil. c The corresponding FT-EXAFS fitting curves of FeCoNiMnRu/CNFsterm before and after stability test. d The Ru K-edge XANES spectra and e FT-EXAFS spectra of FeCoNiMnRu/CNFsterm before and after stability test, and Ru foil. f The corresponding FT-EXAFS fitting curves of FeCoNiMnRu/CNFsterm before and after stability test. The WT-EXAFS spectra of Co and Ru in ( g ) metal foils and h , i FeCoNiMnRu/CNFsterm before and after stability tests.

The relationship between active sitesterm·key and intermediates These discriminating enhancements of HERterm·dom, OERterm·dom, and OWSterm·dom activities imply key roles for multiple metals serving as active centers, and we further used density functional theoryterm·mth (DFTterm·mth) calculations to determine the cooperation of multiple metal active sitesterm·key in the alkaline HERterm·dom. The Tafel slope of the FeCoNiMnRu/CNFsterm (67.4 mV dec −1 ) suggested the Volmer–Heyrovsky reaction pathway. Figure 6a illustrates the atomic configurations at catalytic sites of FeCoNiMnRu HEAterm·dom in the four different stages. The H 2 O * molecule absorbed on the HEAterm·dom surface (stage 1) is destabilized at the H−OH bond (stage 2), which is then dissociated to generate coadsorption of H * and OH * intermediates (stage 3). The H * intermediate will be detached from the surfaces after combining with another H * to give H 2 production (stage 4). The water dissociationterm·key into H * and OH * and adsorption of the H * are potential-determining stepsterm·dom (PDSterm·dom), which determine the water dissociationterm·key rates. The chemical structures of the FeCoNiMnRu HEAterm·dom were shown in Supplementary Fig. 31 . Figure 6b shows the energy profile for water dissociationterm·key on Fe, Co, Ni, and Ru sitesterm·key of the FeCoNiMnRu HEAterm·dom surface at four states. In addition, the chemical structures and atomic configurations of Fe, Co, Ni, and Ru sitesterm·key of FeCoNiMnRu HEAterm·dom during H 2 O dissociation are shown in Supplementary Figs. 32 – 35 . Interestingly, the energy barrierterm·dom for breaking the H−OH bond (stage 1 → stage 2) on Co siteterm·dom of HEAterm·dom is the lowest as 0.34 eV in comparison with the Fe site (0.70 eV), Ni site (0.63 eV), and Ru siteterm·dom (0.67 eV). These results suggest that the H 2 O adsorption and dissociation are more favorable at Co siteterm·dom, which is beneficial to accelerating water dissociationterm·key for the generation of H * intermediates. Fig. 6 Theoretical calculation and in situ electrochemical-Raman characterization. a The atomic configurations on catalytic sites of FeCoNiMnRu HEAterm·dom at the four stages during H 2 O dissociation. b Reaction energy profile for water dissociationterm·key on various catalytic sites of the FeCoNiMnRu HEAterm·dom surface. c Gibbs free energy (Δ G H* ) profiles on various catalytic sites of the FeCoNiMnRu HEAterm·dom surface. d – f Operando electrochemical-Raman spectra collected for FeCoNiMnRu/CNFsterm during the HERterm·dom process in 1.0 M KOH electrolyte.

We further calculated the Gibbs free energies of atomic hydrogen adsorbed (Δ G H* ) (Fig. 6c ) at four catalytic sites of HEAterm·dom to reveal the influence of different metal sites on hydrogen adsorption. Atomic configurations of the FeCoNiMnRu HEAterm·dom at the H * adsorption stage on Fe, Co, Ni, and Ru sitesterm·key are shown in Supplementary Fig. 36 . The DFTterm·mth results indicated that the Ru sitesterm·key achieve the most appealing Δ G H* of −0.07 eV, as compared to those of Fe (−0.13 eV), Ni (−0.27 eV), and Co (−0.43 eV), which suggests that H * is preferentially stabilized at the Ru sitesterm·key. Therefore, during the whole electrocatalytic water splitting process, the Co and Ru sitesterm·key function to simultaneously accelerate the H 2 O disassociation and H * adsorption with the lowest energies, and this remarkable cooperation avoids active siteterm·key blocking and accelerates the whole water dissociationterm·key process.

The active sitesterm·key in HEAterm·dom for the stabilization of intermediates were further investigated by operando electrochemical Raman spectra. As shown in Fig. 6d , the Raman spectrum of FeCoNiMnRu/CNFsterm determined at 0 V displays three peaks at 1316, 1590, and 2616 cm –1 , corresponding to the D band, G band, and 2D band of CNFsterm, respectively 33 . When a potential of 60 mV was applied, new Raman peaks corresponding to Fe–O bonds were observed at 215 and 290 cm –1 , while Raman peaks at 585 and 704 cm –1 were ascribed to Co-O bonds 34 – 36 . The Raman peaks located at approximately 446 and 530 cm −1 were attributed to the emergence of Ni-O bonds 37 . The newly formed Fe–O, Ni–O, and Co–O bonds suggest the generation of Fe-OH * , Ni-OH * , and Co-OH * intermediates, which originate from the cleavage of H 2 O molecules. Interestingly, two sharp Raman peaks that emerged at 2069 and 2092 cm −1 correspond to the Ru-H bonds, strongly suggesting the formation of Ru-H * intermediates 38 , 39 . With increasing applied potentials ranging from 60 to 180 mV, the intensities of all characteristic Raman peaks continuously increased, suggesting enhanced HERterm·dom activity. We further obtained operando electrochemical Raman spectra of FeCoNiMn/CNFsterm during the HERterm·dom process (Supplementary Fig. 37 ). Without the Ru metal, no Raman peaks for Ru–H bonds were observed in the vicinity of 2000–2200 cm −1 , directly confirming the ability of Ru to absorb H * . The Raman results provided direct evidence that Co and Ru active sitesterm·key in the FeCoNiMnRu HEAterm·dom stabilize different intermediates. In the typical FeCoNiMnRu HEAterm·dom, the Co sitesterm·dom facilitate H 2 O dissociation, and the Ru sitesterm·key simultaneously accelerate the combination of H * to H 2 . Therefore, the stabilization of multiple intermediates on various active sitesterm·key in HEAterm·dom was verified experimentally and theoretically.

The above results indicated that the Fe, Co, Ni, and Ru sitesterm·key in HEAsterm·dom play different roles in the HERterm·dom, and we further present a series of FeCoNiXRu (X = Mnterm, Cr, Cu) HEAsterm·dom by using fifth elements (X) with different electronegativities to understand the relationship between electronegativityterm·dom and HERterm·dom activity. Considering that the difference in atomic configurations may change the adsorption energy, 8 randomly selected configurations of FeCoNiXRu (X = Mnterm, Cr, Cu) HEAterm·dom were analyzed by DFTterm·mth calculation. The chemical structures of FeCoNiMnRu, FeCoNiCrRu and FeCoNiCuRu HEAterm·dom with 8 randomly selected configurations are shown in Supplementary Figs. 38 – 40 . As shown in Fig. 7a and b , the overpotentials at 100 mA cm −2 for FeCoNiCuRu/CNFsterm (245 mV), FeCoNiCrRu/CNFsterm (126 mV), and FeCoNiMnRu/CNFsterm (71 mV) suggest a strong relationship with the electronegativityterm·dom of the fifth metal in HEAterm·dom. For the OERterm·dom (Supplementary Fig. 41 ), the low electronegativityterm·dom of Mnterm (1.55) compared with those of Cr (1.66) and Cu (1.90) gives FeCoNiMnRu/CNFsterm the best OERterm·dom activity with the lowest overpotential of 308 mV at current density of 100 mA cm −2 . The d-band center of each metal sites in FeCoNiMnRu and the three HEAterm·dom structures were calculated and shown in Fig. 7c and d . The d-band orbitals with large spin polarization can be divided into spin up and spin down. As shown in Fig. 7d , a smaller number of spin states occupy the spin down orbitals, which is likely to be a spin-polarized catalytic active center and generally participate in catalytic reactions. As shown in Supplementary Figs. 42 – 44 , there are much more electrons can be observed on Ru after the charge distributions in all three HEAterm·dom structures. The results indicate the charge transfers from Fe, Co, Ni and Mnterm metals to Ru metal. Fig. 7 Relationship between metal electronegativityterm·dom and electrochemical performance. a HERterm·dom polarization curves obtained on FeCoNiXRu/CNFsterm (X = Cr, Mnterm, and Cu) in 1.0 M KOH electrolyte. b Correlation between the HERterm·dom overpotentials at 100 mA cm −2 , Tafel slopes, and the electronegativities of metals X (Cr, Mnterm, and Cu). c The d-orbital projected density of states (PDOS) of Fe, Co, Ni, Mnterm, Ru, and FeCoNiMnRu. d Comparison of PDOS of FeCoNiXRu (X = Mnterm, Cr, Cu) HEAterm·dom. e Reaction energy profile for water dissociationterm·key at Co sitesterm·dom of FeCoNiMnRu, FeCoNiCrRu, and FeCoNiCuRu HEAterm·dom surfaces. f Correlation between the energy barrierterm·dom for H 2 O dissociation at different metal sites and the electronegativities of metals (Cr, Mnterm, and Cu). g Gibbs free energy (Δ G H* ) profiles at Ru sitesterm·key on the FeCoNiMnRu, FeCoNiCrRu, and FeCoNiCuRu HEAterm·dom surfaces. h Correlation between Δ G H* at different metal sites and the electronegativities of metals X (Cr, Mnterm, and Cu).

We further calculated the energy profile for water dissociationterm·key on Fe (Supplementary Figs. 45 , 46 ), Co (Fig. 7e ), Ni (Supplementary Figs. 47 , 48 ), and Ru (Supplementary Fig. 49 ) sites on different FeCoNiMnRu, FeCoNiCrRu, and FeCoNiCuRu HEAterm·dom surfaces. As shown in Fig. 7e , the energy profiles for the water dissociationterm·key on Co siteterm·dom of the three HEAterm·dom surfaces demonstrate that the Co sitesterm·dom in FeCoNiMnRu HEAterm·dom has the lowest energy barrierterm·dom (0.34 eV) for dissociation of water into H * and OH * when compared with those of FeCoNiCrRu (0.47 eV) and FeCoNiCuRu (0.66 eV) HEAsterm·dom. Figure 7f strongly confirms that the electronegativityterm·dom of the fifth metal in HEAterm·dom can regulate the energy barrierterm·dom for water dissociationterm·key at each metal site. In particular, energy barriersterm·dom for water dissociationterm·key on Fe, Co, and Ru sitesterm·key are reduced by introducing the less-electronegative Mnterm, and the Co siteterm·dom still exhibits the lowest value of 0.34 eV, suggesting Co sitesterm·dom are the preferred locations for water disassociation.

Additionally, the calculated Δ G H* values for Ru, Fe, Co, Ni sites in FeCoNiMnRu, FeCoNiCrRu and FeCoNiCuRu are shown in Fig. 7g and Supplementary Figs. 50 – 52 . The atomic configurations for H * adsorption at the Fe, Co, Ni, and Ru catalytic sites of FeCoNiCrRu, and FeCoNiCuRu HEAterm·dom are shown in Supplementary Figs. 53 and 54 . The results also demonstrate that optimized Δ G H* values can be realized by introducing less-electronegative metals (Fig. 7h ). Furthermore, all Ru sitesterm·key in FeCoNiMnRu, FeCoNiCrRu, and FeCoNiCuRu showed the most appealing Δ G H* compared with those of Fe, Co, and Ni sites, indicating that Ru sitesterm·key are still the preferred sites for H * adsorption. The Δ G H* on Fe, Co, Ni, and Ru sitesterm·key in FeCoNiXRu (X = Mnterm, Cr, Cu) with 8 randomly selected configurations were analyzed by DFTterm·mth calculation, and the values of Δ G H* were summarized in Supplementary Fig. 55 and Tables 9 – 11 . DFTterm·mth Results suggest that all of the Ru sitesterm·key in FeCoNiXRu (X = Mnterm, Cr, Cu) display lower values than Fe, Co, and Ni sites, demonstrating that the atomic configurations may not affect the Δ G H* order on Fe, Co, and Ni. In addition, the Ru sitesterm·key in FeCoNiMnRu also exhibit the lowest Δ G H* than those Ru sitesterm·key in FeCoNiCrRu, and FeCoNiCuMn HEAterm·dom, further indicating that the five metal with low electronegativityterm·dom in HEAterm·dom could reduce the Δ G H* . According to the bond length and coordination numbers of Co and Ru in HEAterm·dom, as investigated by XASterm·mth (Fig. 3 ), we have chosen the atomic configuration 4 accordingly with XASterm·mth results as a representative sample to show the relationship between the Δ G H* and electronegativityterm·dom, unveiling the electrocatalytic mechanism of HERterm·dom on FeCoNiXRu (X = Mnterm, Cr, Cu) HEAterm·dom.

Based on the above results, the relationships between metal electronegativities in HEAsterm·dom and the energy barriersterm·dom for water dissociationterm·key and H * adsorption at various active metal sites have been established. Charge transfer between different surface atoms occurs in HEAsterm·dom containing five principal elements with different electronegativities, further leading to significant charge redistributionterm·dom on the surfaces of alloys (Supplementary Figs. 42 – 44 ). In the FeCoNiMnRu HEAterm·dom, the electronegativityterm·dom differences in Fe (1.83), Co (1.88), Ni (1.91), Mnterm (1.55), and Ru (2.20) induce significant charge redistributionterm·dom and create the most active Co and Ru sitesterm·key with optimized energy barriersterm·dom for simultaneously stabilizing OH * and H * intermediates, greatly promoting the HERterm·dom efficiency in alkaline solution. We prepared a series of HEAsterm·dom by fixing Fe, Co, Ni, and Ru metals and varying the fifth metal among Mnterm, Cr, and Cu. The decrease in electronegativities from Cu (1.90) to Mnterm (1.55) leads to the reduced energy barriersterm·dom for water dissociationterm·key and H * adsorption. Changing in the fifth metal in a HEAterm·dom did not affect the adsorption energy order on active siteterm·key in HEAterm·dom; Co siteterm·dom was the most active sitesterm·key for OH * adsorption, while Ru siteterm·dom was the most active sitesterm·key for H * adsorption. In a FeCoNiMnRu NP, the Co siteterm·dom was the preferred active siteterm·key with the lowest energy barriersterm·dom (0.34 eV) for water dissociationterm·key when compared with Fe, Ni, and Ru sitesterm·key. During the subsequent H * adsorption/desorption process, H * leaves Co and is absorbed on Ru sitesterm·key due to its lowest ΔG H* of −0.07 eV. Adjustments of the HERterm·dom activities of HEA catalyststerm·dom were shown experimentally and theoretically by tailoring the electronegativities of the compositions.

Discussion In summary, FeCoNiMnRu HEAterm·dom NPs were designed and synthesized in electrospun CNFsterm by combining the electrospinning technique and graphitization process. The transformation from a multiphase to a single-phase HEAterm·dom was evident from the real-time in situ XRD results and demonstrated a thermodynamically driven phase transition. The FeCoNiMnRu/CNFsterm achieved low overpotentials of 71 and 308 mV to drive a current density of 100 mA cm −2 for the HERterm·dom and OERterm·dom, respectively. It required only 1.65 V to achieve two-electrode overall water splittingterm·dom, and the chronoamperometric curve exhibited a stable current density of 1 A cm −2 during continuous electrolysis for more than 600 h in 1.0 M KOH, suggesting its remarkable stability. In the FeCoNiMnRu HEAterm·dom, the Co siteterm·dom facilitated H 2 O dissociation, and the Ru sitesterm·key simultaneously accelerated the combination of H * to H 2 . The ability to stabilize multiple intermediates on various active sitesterm·key in HEAsterm·dom was verified experimentally and theoretically. Adjustments of the HERterm·dom activities of HEA catalyststerm·dom were shown experimentally and theoretically by tailoring the electronegativities of the composition. This work provides an in-depth understanding of the correlation between specific active sitesterm·key and intermediates, opening up a fascinating approach for overcoming the scaling relationterm·dom issues seen for multistep reactions.

Methods Materials Anhydrous ferric trichloride (FeCl 3 , ≥97.0%), nickel chloride hexahydrate (NiCl 2 ·6H 2 O, AR), and N,N-dimethylformamide (DMF, ≥99.0%) were acquired from Sinopharm Chemical Reagent Co., Ltd. Manganese chloride tetrahydrate (MnCl 2 ·4H 2 O, 99.99%), cobalt chloride hexahydrate (CoCl 2 ·6H 2 O, AR), anhydrous copper chloride (CuCl 2 , 98%), commercial Pt/C (20 wt%), ruthenium oxide (RuO 2 ), iridium dioxide (IrO 2 ), and potassium hydroxide (KOH, GR, 95%) were purchased from Shanghai Macklin Biochemical Co., Ltd. Chromium chloride hexahydrate (CrCl 3 ·6H 2 O, 99%) was acquired from Beijing Innochem Science & Technology Co., Ltd. Ruthenium chloride trihydrate (RuCl 3 ·3H 2 O, 95%) was provided by Bide Pharmatech Ltd. Polyacrylonitrile (PAN, M w = 1.49 × 10 5 , copolymerized with 10 wt% acrylate) was supplied by Sinopec Shanghai Petrochemical Co., Ltd. Nafion117 solution (5 wt%) was obtained from Shanghai Aladdin Biochemical Technology Co., Ltd.

Synthesis of FeCoNiMnRu HEAterm·dom nanoparticles supported on CNFsterm Typically, 0.5 mmol of MnCl 2 ·4H 2 O, 0.5 mmol of FeCl 3 , 0.5 mmol of CoCl 2 ·6H 2 O, 0.5 mmol of NiCl 2 ·6H 2 O, 0.5 mmol of RuCl 3 ·3H 2 O and 1.5 g of PAN were dissolved in 22 g of dimethyl formamide (DMF). Homogeneous metal salts/PAN polymer solution was acquired after stirred by magnetic stirring apparatus for 8 h at room temperature. Afterwards, the syringes with stainless needles were filled with the as-prepared precursor solution and assembled to the electrospinning machine (YFSP-T, Tianjin Yunfan Technology Co., Ltd.) with an anode voltage of 20 kV, an injection rate of 0.3 mL/h, and a distance between the collector and needle of 18 cm. The as-prepared MnFeCoNiRu/PAN precursor nanofibers were put into the heating section of the home-built chemical vapor deposition (CVD) furnace. For the pre-oxidation process, the membranes were heated to 230 °C in air with a heating rate of 2 °C min −1 and maintained for 3 h. Then, the furnace was heated to 1000 °C under Ar atmosphere with a heating rate of 2 °C min −1 and maintained for 3 h. Finally, the as-synthesized FeCoNiMnRu HEAterm·dom/CNFsterm was obtained after the furnace cooling down to room temperature under Ar atmosphere.

Synthesis of metal alloy and HEAterm·dom nanoparticles supported on CNFsterm Ru/CNFsterm, FeCoNi/CNFsterm, FeCoNiMn/CNFsterm, FeCoNiRu/CNFsterm, FeCoNiCrRu/CNFsterm, and FeCoNiCuRu/CNFsterm were also synthesized through the same processes with those of FeCoNiMnRu/CNFsterm. The precursor solutions of above control samples contained 0.5 mmol of each Fe, Co, Ni, Cu, Mnterm, Ru, Cr metal salts, 1.5 g PAN, and 22 g DMF. FeCoNiMnRu 0.5 /CNFsterm and FeCoNiMnRu 2 /CNFsterm were prepared by halving and doubling the amount of RuCl 3 ·3H 2 O. FeCoNiMnRu/CNFsterm synthesized at 800 and 900 °C were also synthesized to unveil the thermodynamically driven phase transition process of HEAterm·dom NPs.

Materials characterizations The field-emission scanning electron microscope (FE-SEM, HITACHI S-4800) at an acceleration voltage of 3 kV was applied to collect the FE-SEM images. The transmission electron microscope (TEM, JEM-2100 plus) at an acceleration voltage of 200 kV was used to record the TEM images. Bright field and high-angle annular dark field scanning transmission electron microscopy (STEM) images, energy dispersive X-ray spectroscopy (EDX) mapping images, and line-scan EDX spectra were characterized by a Tecnai G2 F30S-Twin, Philips-FEI at an acceleration voltage of 300 kV. Data of inductively coupled plasma-optical emission spectrometry (ICP-OES) were acquired by Agilent 720ES. X-ray diffraction (XRD) patterns were taken via a Smartlab 9kw advance powder X-ray Cu K α radiation diffractometer (λ = 1.5406 Å) in the 2θ range of 20~80° at the scanning rate of 0.5 or 10° min −1 , with the Cu K α source operating at 40 kV and 40 mA. X-ray photoelectron (XPSterm·mth) spectra were acquired by Thermo Scientific K-Alpha with the Al (mono) K α source (1486.6 eV) operating at 12 kV and 6 mA. The binding energies were calibrated with C 1 s (284.8 eV) as the standard. The BL08U1-A at the Shanghai Synchrotron Radiation Facility (SSRF) operated at 500 eV with injection currents of 100 mA was used to obtain the Co K-edge and Ru K-edge X-ray absorption near edge structure (XANES) spectra, with radiation monochromatized by a Si (111) double-crystal monochromator. Metal Co, Ru, CoO, Co 3 O 4 , and RuO 2 were taken as controls. The operando electrochemical Raman test was carried out in a round home-made electrolyzer. Ag/AgCl electrode and Pt wire served as the RE and CE, respectively. The prepared electrocatalyststerm·key were dispersed on the glass carbon electrode (GCE). The electrochemical processes were conducted in 1.0 M KOH saturated with Ar and controlled by a CHI660E electrochemical workstation. Raman spectrometer (inVia) used the laser wavelength of 785 nm.

Electrochemical Characterization Electrochemical measurements were all conducted in a typical three-electrode system at 25 °C in 1.0 M KOH with a Autolab electrochemical workstation. Saturated calomel electrode (SCE) and graphite rod were used as reference electrode (RE) and count electrode (CE), respectively. The SCE was calibrated before each test. The self-supported CNFs-based materials were cut into 1 × 1 cm −2 and served as working electrode (WE). Potentials were converted to the reversible hydrogen electrodeterm·mth (RHEterm·mth) by the equation E RHEterm·mth = E SCE + 0.244 + 0.059 × pH. Pt/C (20 wt%), RuO 2 and IrO 2 powder were taken as controls and deposited on glassy carbon electrode (GCE) with diameter of 3 mm for measurement. To prepare the electrocatalystterm·key ink, 3 mg of electrocatalystterm·key was dispersed into 1 mL mixed solvent with a volume ratio of V isopropanol : V water = 3:1. After 30 min of ultrasonication, 25 μL Nafion117 solution was added. After another 30 min for ultrasonication, 5 μL electrocatalystterm·key ink was casted on GCE and dried in the air naturally. All linear sweep voltammetry (LSV) curves were obtained at a scan rate of 2 mV s –1 with 95% iR-compensation. Tafel plots were gained according to the Tafel equation: 1 η = a + b log j where η , b , and j represent the overpotential, Tafel slope, and current density, respectively. Electrochemical double layer capacitances (C dl ) were measured by analyzing the cyclic voltammetry (CV) curves at scan rate of 10, 20, 30, 40, and 50 mV s –1 in the range of –0.78 to –1.00 V vs . SCE. Plotting the Δ i /2 (Δ i = i a – i c , where i a and i c represent the positive and negative current, respectively) at –0.89 V vs . SCE against the scan rate ( v ), the C dl can be calculated by the equation: C dl = Δ i /2 v . Electrochemical active surface area (ECSA) was estimated by the equation: ECSA = C dl /C s , where the specific capacitance value (C s ) was taken 0.04 mF cm –2 . The ECSA-normalized LSV curves were acquired by the equation: j ECSA = i /ECSA, where j ECSA and i is the current density normalized to ECSA and current of the working electrode, respectively. 10,000 CV cycles in the range of –0.4 to 0.05 V vs . RHEterm·mth at a scan rate of 100 mV s –1 and chronoamperometry were performed to evaluate the durability of samples. In the course of the chronoamperometry test, 10 min pauses were inset into the electrolysis process with intervals of about 10 h operating. Meanwhile, the electrolyte was renewed and the RE was calibrated during the pauses. For the comparison of mass activities and eliminating the difference in electrode structure, Pt/C, RuO 2 , IrO 2 , and FeCoNiMnRu/CNFsterm powder were dropped onto GCE, and the calculation process is shown as follow: J mass = J geo × 0.07069 cm 2 /m, m = 3 mg × ω × (5 μL/1000 μL), where m and ω are the mass loading and mass fraction of noble metal (Ru, Pt, or Ir).

Computational methods We have employed the first-principles 40 , 41 to perform all Spin-polarization density functional theoryterm·mth (DFTterm·mth) calculations within the generalized gradient approximation (GGA) using the Perdew-Burke-Ernzerhof (PBE) 42 formulation. vdW corrections was added. We have chosen the projected augmented wave (PAW) potentials 43 , 44 to describe the ionic cores and take valence electrons into account using a plane wave basis set with a kinetic energy cutoff of 400 eV. Partial occupancies of the Kohn−Sham orbitals were allowed using the Gaussian smearing method and a width of 0.05 eV. The electronic energy was considered self-consistent when the energy change was smaller than 10 −6 eV. A geometry optimization was considered convergent when the energy change was smaller than 0.04 eV Å −1 . The vacuum spacing in a direction perpendicular to the plane of the structure is 15 Å. The Brillouin zone integration is performed using 3 × 3 × 1 Monkhorst-Pack k-point sampling for a structure. Finally, the adsorption energies (E ads ) were calculated as Eq. ( 2 ): 2 E ads = E ad / sub − E ad − E sub where E ad/sub , E ad , and E sub are the total energies of the optimized adsorbate/substrate system, the adsorbate in the structure, and the clean substrate, respectively. The free energy was calculated using the Eq. ( 3 ): 3 G = E + ZPE − TS where G, E, ZPE, and TS are the free energy, total energy from DFTterm·mth calculations, zero point energy, and entropic contributions, respectively. The ZPE has been calculation using the 6 × 2 × 2 Monkhorst-Pack k-point and DFTterm·mth + U correction has been used in our systems. In our alloy structure, the FCC structure has been obtained with the 80 atoms.

Supplementary information Supplementary Information Peer Review File