Introduction

The renewable and nonpolluting nature of hydrogen energy has aroused the enthusiasm of researchers owing to the limited fossil energy available on earth and the increasing demand for renewable energy sources1,2,3. Electrochemical water decomposition is regarded as an ideal approach that is essential for future carbon-neutral processes and sustainable energy systems4,5. A typical water spitting system has an anodic oxygen evolution reaction (OER: 4OH− → O2 + 2H2O + 4e−) and a cathodic hydrogen evolution reaction (HER: 2H2O + 2e−→H2 + 2OH−) in an alkaline medium with a dynamic potential of 1.23 V6. However, the sluggish reaction kinetics often result in a high overpotential and a highly efficient and durable electrocatalyst is especially needed to improve the energy conversion efficiencies7,8,9,10,11,12.

Until now, the most effective electrocatalysts are still noble metal-based RuO2, IrO2 and Pt/C13, but their high costs, scarcities and instabilities have greatly hindered their use in a wide range of applications14,15,16,17,18,19. Considering these challenges, the use of nonnoble metals would reduce the dosages of the noble metals and enhance the catalytic activity owing to a synergistic electronic effect or generate defect effects. Surprisingly, these defects could alter the local electronic structure and serve as “docking” sites to trap atomic metal species and form new coordination structures for the active sites20. Consequently, it is essential to obtain in-depth insight into the roles of defective sites in engineering the electrocatalytic energy conversion process, which would be particularly important in guiding the designs of ideal catalysts21. Among the three noble metals used, Ru-based materials are especially promising catalysts for hydrogen production from water splitting because their binding strengths with hydrogen are similar to those of Pt, and they exhibit high HER activity22. Previous strategies for enhancing the HER in alkaline media were focused on facilitating the sluggish water dissociation reactions by incorporating a specific component (e.g., a transition metal hydroxide) onto the catalytically active species (e.g., Ru) or inducing surface reconstruction to generate more activity23. Yttrium (Y) is a rare earth element, and it exhibits appreciable electrical conductivity and a unique 4 f electron layer, which improves the performance in hydrogen production from electrolytic water splitting when combined with Ru active sites24,25. Furthermore, the ionic radius of Y is smaller than that of Ru, which can promote oxygen vacancies as well as structural changes26.

Although noble metal-based electrocatalysts such as Pt are ideal materials for the HER, alloyed materials27,28,29 have been used to provide improved activity30. Among the metal-based catalysts used as HER catalysts, carbon is utilized as the catalyst support31,32 owing to its good conductivity and to avoid agglomeration of the active catalysts for long-term use33. Great effort has been expended to utilize nonmetal catalysts, including carbon-related materials such as graphene-doped graphenes34,35,36 and graphitic carbon nitride (g-C3N4)37,38,39. Alloying is an important way to improve the performance of a catalyst and is commonly used to prepare precious metal-based catalysts40. In this study, we present an effective strategy for remarkably enhancing the H coverage on a catalyst for the HER in alkaline media. Specifically, ruthenium-yttrium nanocomposites (RuYO2−x) were loaded onto a graphene system rich in oxygen vacancies, and the protons inserted into HxYO2−x were transferred to RuYO2−x/C during the HER and greatly increased the hydrogen coverage on RuYO2−x/C and improved the HER performance. Meticulous characterization, including electron paramagnetic resonance (EPR) studies, confirmed robust coupling of the Ru sites with oxygens near the yttrium vacancies. The Ru atomic content was 8.84 at%, which greatly decreased the cost of the noble-metal content compared with those of commercial Ru/C and Pt/C. DFT calculations revealed that the d-band center of Ru might be adjusted closer to the Fermi level due to the synergistic effects of the Y-O-Ru bonds and Vo, which lowered the energy barrier for water dissociation and facilitated the adsorption of water.

Results and discussion

Morphologies and structures

Tunable bimetallic nanocomposites RuYO2−x were loaded on graphene (RuYO2−x/C) via a solvothermal method, as shown in Fig. 1, which utilized two metal precursors: a ruthenium salt (RuCl3) and a yttrium precursor (HxYO2−x) grown on the two-dimensional graphene nanosheets. In particular, the unique low-temperature solution-processed hydrogen yttrium bronze (HMOs) HxYO2−x was synthesized through a controllable hydrogen insertion method. The yttrium powder was oxidized by hydrogen peroxide in the presence of ethanol. With the reaction rate controlled by ethanol, HxYO2−x was readily obtained (Supplementary Fig. S1). Subsequently, intermetallic RuY anchored on two-dimensional graphene was prepared by a one-pot solvothermal method. A mixed solution of RuCl3 and HxYO2−x was then added to graphene to form RuYO2−x/C nanocomposites. The detailed structural model of RuYO2−x/C on the [222] planes was enlarged to determine the association and arrangement of atoms. The obtained RuYO2−x/C electrocatalyst with a small number of oxygen vacancies (Vo) exhibited desirable electrocatalytic properties. For the sake of comparison, we also utilized yttrium powder (Y) to synthesize the corresponding RuYOx/C directly, which might show the difference between the Y powder and the HxYO2−x bronzes (the RuYO2−x/C was prepared from Y powder, while the RuYO2−x/C was prepared from the as-prepared HxYO2−x bronze) from the same synthetic process.

Fig. 1
figure 1

Schematic illustration for the preparation of RuYO2−x/C.

Full size image

The crystal structures of the Y powder and the prepared HxYO2−x, RuYO2−x/C, and RuYOx/C electrocatalysts were characterized by X-ray diffraction (XRD), as shown in Fig. 2a. The XRD patterns of the as-prepared materials showed their well-defined crystalline structures. The peaks for HxYO2−x appeared at 29.16°, 32°, 34.5°, 48.14°, and 59.09°, which were indexed to the (222), (003), (112), (313), and (136) lattice planes of Y2O3 (PDF#43–1036 and PDF#44–0399), respectively. There was also a peak for Y2O3 in the XRD pattern for the Y powder, probably because it was oxidized in air. Compared to those for the Y powder, the peaks for HxYO2−x were moved to higher angles, indicating lattice shrinkage, which might have been caused by oxygen vacancies. RuYO2−x/C and RuYOx/C had sharp peaks at 43.7° and 43.419°, respectively, which corresponded to the (511) lattice plane of Y2Ru2O7 (PDF#28–1456) and indicated successful binding of Ru and Y. Compared with RuYOx/C, the peak for RuYO2−x/C moved to a higher angle, indicating that the lattice of RuYO2−x/C shrank due to the presence of oxygen vacancies41. In addition, RuYO2−x/C showed a small peak at 29.19°, which was attributed to the (222) crystal face of Y2O3 (PDF#43–1036). To investigate whether RuO2, as well as Ru, was formed in the RuYO2−x/C, the standard data for RuO2 (PDF#43–1027) and Ru (PDF#06–0663) were added to the XRD pattern. The results showed that there was no peak for RuO2 in the RuYO2−x/C and RuYOx/C, which might have been caused by the low content of RuO2 formed via Ru oxidation. In RuYO2−x/C and RuYOx/C, a small peak at 44.06° corresponded to the (101) lattice place of Ru, indicating the presence of a small amount of Ru.

Fig. 2: Morphological characterization.
figure 2

a XRD patterns of RuYO2−x/C. b SEM images of RuYO2−x/C and c RuYO2−x/C. d Magnified HRTEM image of RuYO2−x/C. e–i High-angle annular dark-field image of RuYO2−x/C and EDS elemental maps (C: Yellow, O: Olive, Ru: Red, Y: Green).

Full size image

Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) measurements were used to probe the morphological characteristics. The surface morphologies of the RuYO2−x/C nanocomposites are shown in Fig. 2b and Supplementary Fig. S2a, b, which indicated that RuYO2−x/C was composed of densely packed particles grown onto the graphene sheets. Figure 2c and Supplementary Fig. S3 show the TEM images of RuYO2−x/C, further demonstrating that the as-prepared nanomaterials were spherical particles loaded on graphene. To investigate the interior lattice, high-resolution TEM (HRTEM) images of RuYO2−x/C are shown in Fig. 2d and marked with yellow and orange rectangles. Figure 2d displays the HRTEM images of the RuYO2−x/C sample, which were accompanied by FFT patterns and IFFT patterns of the selected areas. From Supplementary Fig. S3c, d, 7 and 4 stripes were calculated, respectively. Finally, the lattice spacings of RuYO2−x/C were determined to be 0.3 nm and 2.94 nm, which corresponded to the (401) lattice planes of Y2O3 (PDF#44–0399; d = 2.992 Å) and the (222) lattice planes of Ru2Y2O7 (PDF#28–1456; d = 2.929 Å), respectively. In addition, the lattice fringes of RuYO2−x/C at the edge were blurred, which might have been due to the formation of oxygen vacancies induced by hydrogen reduction42. The high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image in Fig. 2e–i revealed the distributions of O, Ru, and Y elements in the catalyst. Energy dispersive spectroscopy (EDS) spectroscopy showed that the atomic ratio of Y and Ru in the as-prepared RuYO2−x/C sample was 13.08:8.84 (Supplementary Fig. S2f), which corresponded to an approximate composition ratio of 0.12:0.88 shown for the Y and Ru in RuYO2−x/C by ICP‒AES (Supplementary Tables S2 and S1). In addition, we captured TEM images of HxYO2−x with the same method, and Supplementary Fig. S4a shows that HxYO2−x formed multilayered stacked sheets, and the crystal plane spacing of HxYO2−x was 0.17 nm in the high-resolution HRTEM images (Supplementary Fig. S4b, c), corresponding to the (7 1–2) crystal faces of Y2O3 (PDF#44–0399). Supplementary Fig. S4d–f showed that Y and O were evenly distributed, and combined with the XRD analysis of HxYO2−x, it was clear that the Y powder had been oxidized to HxYO2−x. The atomic ratio of Y to O in the HxYO2−x was 67.66:32.34, as shown by the energy dispersive analysis (EDS) in Supplementary Fig. S4g.

Chemical valence states and band structures

X-ray photoelectron spectroscopy (XPS) was used to investigate the electronic structures and valence states of the RuYO2−x/C and RuYOx/C electrocatalysts and HxYO2−x. The XPS survey spectrum showed the presence of Ru, Y and O in RuYO2−x/C and RuYOx/C and Y and O in HxYO2−x, RuYO2−x/C and RuYOx/C (Supplementary Fig. S5). The atomic ratios of the elements in the RuYO2−x/C and RuYOx/C catalysts were estimated with XPS, as shown in Supplementary Table S2. The ratio of Ru/Y atoms measured with XPS (2.6:0.5) was much larger than that measured with EDS (8.84:13.08) because XPS only detected the surface atoms at depths of a few nanometers, which revealed the presence of a Ru surface in RuYO2−x/C43. The Ru 3d spectra of RuYO2−x/C and RuYOx/C are shown in Fig. 3a. In RuYO2−x/C, the peaks at 280.1 eV and 284.4 eV were the Ru 3d5/2 and Ru 3d3/2 binding energies for Ru0, and those at 281.18 eV and 286.1 eV were the Ru 3d5/2 and Ru 3d1/2 peaks for Ru4+, respectively42,44. The four Ru 3d peaks seen for RuYOx/C were roughly the same as those for RuYO2−x/C (Ru0: 280.1 eV and 284.4 eV; Ru4+: 282.18 eV and 286.1 eV). The Ru 3p spectra for RuYO2−x/C and RuYOx/C are shown in Fig. 3b. The peaks at 461.8 eV and 483.43 eV for RuYO2−x/C were attributed to the Ru 3p5/2 and Ru 3p3/2 states of Ru0, confirming the presence of Ru0. The peaks at 486 eV and 463.46 eV were those of Ru4+, which might have been caused by oxidation45. The peak for Ru0 in RuYOx/C was divided into peaks at 484.14 eV (Ru 3p1/2) and 461.88 eV (Ru 3p3/2), and the peaks at 463.66 eV and 487.14 eV were assigned to the 3p3/2 and 3p3/2 states of Ru4+. The Ru 3p peaks for RuYO2−x/C showed positive displacements compared with those for RuYOx/C. Combined with the peak areas for RuYO2−x/C and RuYOx/C in Supplementary Table S3, the dominant species in RuYOx/C was Ru0. The Ru0 reduced the dissociation energy of H2O and optimized the free energy for hydrogen adsorption46. When Ru was bound to HxYO2−x, the ratio of Ru4+ to Ru0 was increased, which meant that the ability of Ru to lose electrons was enhanced, and the valence state was increased.

Fig. 3: Structural characterizations.
figure 3

RuYO2−x/C, RuYOx/C and HxYO2−x XPS fine spectra; a Ru 3d, b Ru 3p, c Y 3d, and d O 1s. e EPR spectra of HxYO2−x and RuYO2−x/C. f Cutoff levels (Ecutoff) and Fermi levels (EFermi) of UPS of HxYO2−x and RuYO2−x/C.

Full size image

As shown in Fig. 3c, we compared the Y 3d spectra for RuYO2−x/C, RuYOx/C and HxYO2−x. Two peaks at 157.5 eV and 159.43 eV for RuYOx/C were attributed to the Y 3d5/2 and Y 3d3/2 states, respectively; compared to those for RuYO2−x/C (Y 3d5/2: 157.53 eV; Y 3d3/2: 159.6 eV) and HxYO2−x (Y 3d5/2: 157.7 eV; Y 3d3/2: 159.7 eV), the peaks for RuYOx/C had obvious negative displacements, indicating that the extents of Y oxidation in RuYO2−x/C and HxYO2−x were higher than that in RuYOx/C47. Compared with those for HxYO2−x, the Y 3d states for RuYO2−x/C exhibited lower binding energies, revealing that electron accumulation in Y optimized the hydrogen adsorption process. Combined with the XPS results for Ru, this showed that some electrons in Ru were transferred to Y, and doping with Ru adjusted the interfacial electronic structure of the catalyst. The Ru and Y synergistically improved the catalytic performance.

Oxygen vacancies (Vo) are common in metal oxides, and they often enhance surface electrocatalysis and promote electron transfer41. In general, the O 1s XPS peaks could be divided into three regions for metal-O, Vo, and adsorbed water molecules48,49. The O 1s XPS spectra of RuYO2−x/C, RuYOx/C, and HxYO2−x are shown in Fig. 3d (RuYO2−x/C (Ru/Y-O: 529 eV; Vo: 530.53 eV; adsorbed water molecules: 531.7 eV)), RuYOx/C (Ru/Y-O: 531.36 eV; Vo: 531 eV; adsorbed water molecules: 533 eV) and HxYO2−x (Y-O: 529.48 eV; Vo: 531 eV; adsorbed water molecules: 532.52 eV). It is worth noting that the proportion of Vo peak area in RuYO2−x/C was 37.06%, which was higher than those in RuYOx/C (23.81%) and HxYO2−x (21.13%) (Supplementary Table S4), indicating that there were abundant oxygen vacancies in RuYO2−x/C50. The low Vo peak area ratio of HxYO2−x revealed that the electronic interactions between Ru and Y were the main reason for the oxygen vacancies in the RuYO2−x/C catalyst. Electron transfer from Ru to Y led to a decrease in the metal-O content, which led to oxygen escape and more oxygen vacancies in RuYO2−x/C51.

Electron paramagnetic resonance (EPR) measurements confirmed the formation of oxygen vacancies in RuYO2−x/C, as shown in Fig. 3e. As expected, the Ru cations were stabilized by the oxygen vacancies, which was attributed to an electronic coupling mechanism52. RuYO2−x/C, RuYOx/C, and HxYO2−x all exhibited symmetric EPR peaks at g = 2.004, which were caused by unpaired electrons at the oxygen vacancies53. The EPR signal strength was related to the number of oxygen vacancies, and the higher signal strength of the RuYO2−x/C adhesive indicated that there were abundant oxygen vacancies in the RuYO2−x/C catalyst, which was consistent with the results of the XPS analyses. After Ru was combined with HxYO2−x, the electronic structure changed and the lattice distortion caused by the oxygen vacancies increased the conductivity of the catalyst, facilitated the adsorption of H2O and OH− groups on the surface of the catalyst and improved the electrocatalytic performance54,55,56.

In addition, to confirm the preparation of the catalytic materials via electron escape, ultraviolet photoelectron spectroscopy (UPS) studies and work function (WF) calculations were performed for HxYO2−x and RuYO2−x/C to determine the electron modulation mechanisms. As shown in Fig. 3f, the UPS results showed that the secondary electronic cutoff edges of the catalytic materials were closely related to the doping levels of Ru and Y57. The WF is usually approximately half the ionization energy of a metal-free atom. The WF of a metal is expressed as the minimum energy required for an electron with an initial energy equal to the Fermi level (EF) to escape from the interior of the metal into a vacuum58. The value of the WF is related to the surface condition, and the WF also shows periodic changes as the atomic number increases. In semiconductors, the conduction band’s minimum energy and the valence band’s maximum energy are generally lower than the minimum electron escape energy of the metal. For an electron to escape from a semiconductor, it requires the corresponding amount of energy. This indicated that the EF of the catalyst in this study was related to metal doping59. The value of EF indicates how strongly the electrons are bound to the metal. A smaller EF means it is easier for electrons to leave the metal60,61. The WF value can be calculated from the formula in Fig. 3f. The slope of the VB-XPS plot shown in Supplementary Fig. S6a, b showed that the valence band maximum (VBM) values of HxYO2−x and RuYO2−x/C were −0.92 eV. The band diagram parameters calculated from the UPS data, including the vacuum level (EVac), EF and the valence band (VB), are shown in Fig. 3f. HxYO2−x (EF = 7.95 eV, WF = 6.11 eV) had a higher Ef energy level and a lower WF than RuYO2−x/C (EF = 7.43 eV, WF = 6.38 eV). This indicated that HxYO2−x had better electrical conductivity and could transfer electrons to RuYO2−x/C62. After compounding with Ru atoms, RuYO2−x/C presented a higher WF, which indicated a greater ability to resist electron loss and withstand the anode potential, which increased the positive working potential63. The UPS spectra of RuYO2−x/C with different graphene contents and the corresponding VB-XPS spectra of RuYO2−x are shown in Supplementary Fig. S6c, d.

HER electrocatalytic performance analyses

RuYO2−x/C exhibited electrical conductivity and oxygen vacancies. We investigated the HER catalytic capabilities of the synthesized catalysts in a typical three-electrode system with a 1.0 M KOH solution used as the electrolyte. For comparison, a commercial RuO2 catalyst was tested as a reference. All potentials were referenced to that of the reversible hydrogen electrode (RHE). Figure 4a shows the linear scanning voltammetry (LSV) curves for the Ru-YO2−x/C and Ru-YOx/C series synthesized from HxYO2−x and yttrium powder with different ratios. Undoubtedly, commercial RuO2 exhibited the highest catalytic activity, with an overpotential η10 = 46 mV at a current density of 10 mA cm−2. Impressively, RuYO2−x/C(0.2:0.1) had an η10 of only 56 mV, indicating excellent electrocatalytic performance and an overpotential that was only 10 mV higher than that of RuO2. The measured η10 values were 166 mV (RuYOx/C(0.2:0.1)), 226 mV (RuYO2−x/C(0.2:0.4)), and 246 mV (RuYOx/C(0.2:0.4)). HxYO2−x showed almost no HER catalytic performance. As shown in Supplementary Fig. S7, an appropriate Ru/Y molar ratio slightly affected the catalytic performance of the prepared RuYO2−x/C and RuYOx/C in the HER. The LSV results revealed that with increasing Y concentration, the performance of the synthesized catalyst decreased, which might have occurred because trace Y coordinated and cooperated with the Ru to improve the catalytic performance. Under the same conditions, we also compared the electrocatalytic performance of RuYO2−x/C(0.2:0.1) with different graphene contents (Fig. 4b). The HER performance of RuYO2−x/C with a 3 mg graphene content (0.2:0.1) still showed excellent electrochemical activity, while the catalytic activity of RuYO2−x with 0 mg graphene content was minimal (146 mV). Previous studies showed that most electrocatalysts are prone to corrosion and passivation in extreme environments. Due to the surface corrosion resistance and excellent electrical conductivities of carbon materials, the electrocatalytic activities and stabilities of metal nanoparticles are effectively enhanced after coupling with carbon64,65. The overpotentials for other RuYO2−x/C(0.2:0.1) catalysts with different graphene contents were 116 mV (RuYO2−x/C (1 mg)), 86 mV (RuYO2−x/C (1.5 mg)), and 66 mV (RuYO2−x/C (1 mg)).

Fig. 4: HER performance in 1.0 M KOH.
figure 4

a, b HER polarization curves. c Cdl values from the current density differences plotted against scan rates for RuO2, RuYO2−x/C(0.2:0.1), RuYOx/C(0.2:0.1), RuYO2−x/C(0.2:0.4), and RuYOx/C(0.2:0.4). d EIS Nyquist plots for RuO2, RuYO2−x/C(0.2:0.1), RuYOx/C(0.2:0.1), RuYO2−x/C(0.2:0.4), and RuYOx/C(0.2:0.4). e Tafel plots for RuO2, RuYO2−x/C(0.2:0.1), RuYOx/C(0.2:0.1), RuYO2−x/C(0.2:0.4), RuYOx/C(0.2:0.4) and HxYO2−x. f Stability tests for RuYO2−x/C run at a constant potential of 56 mV (vs. RHE) for 24 h.

Full size image

The electrochemically active surface area (ECSA) is a significant parameter revealing the intrinsic electrocatalytic activity of a catalyst, and it is determined from the double-layer capacitance (Cdl) measured with cyclic voltammetry (CV)66. The electrochemical active surface area (ECSA) was estimated based on ECSA = Cdl/Cs, where Cdl corresponds to the double-layer charging current versus the scan rate and Cs corresponds to a specific capacitance67. To probe the electrocatalytic activity of RuYO2−x/C(0.2:0.1) in more depth, we calculated the electrochemically active surface area (ECSA) by measuring the Cdl values from the CV curves (Fig. 4c). The Cdl value for RuYO2−x/C(0.2:0.1) was 8.7 mF cm−2, which was higher than those of RuYOx/C(0.2:0.1) (7.84 mF cm−2), RuYO2−x/C(0.2:0.4) (3.37 mF cm−2), and RuYOx/C(0.2:0.4) (1.52 mF cm−2), which indicated that the RuYO2−x/C(0.2:0.1) electrode had more active sites. The Cdl values of RuYO2−x/C(0.2:0.1) with different graphene contents are shown in Supplementary Fig. S9 and Fig. 9. Electrochemical impedance spectroscopy (EIS) was also employed to investigate the HER kinetics. The impedance spectrum (Fig. 4d, Supplementary Fig. S10) showed that the charge transfer resistance of the RuYO2−x/C(0.2:0.1) catalyst was the smallest among all of the materials, thus indicating that the RuYO2−x/C(0.2:0.1) electrode exhibited faster HER kinetics and accelerated faradaic processes. A Tafel slope analysis was carried out to investigate the reaction kinetics of the HER. As shown in Fig. 4e and Supplementary Fig. S11, RuYO2−x/C(0.2:0.1) exhibited an apparently low Tafel slope of 63.18 mV dec−1, which corresponded to Tafel step-limited HER kinetics. A linear fit for the Tafel data showed that the Tafel slope for RuYO2−x/C(0.2:0.1) was only 5.67 mV dec−1 higher than that for RuO2, and RuYO2−x/C(0.2:0.1) exhibited the most efficient HER reaction kinetics among the other synthesized materials. Figure 4f indicates that no significant current decay was observed after 24 h of continuous RuYO2−x/C(0.2:0.1) testing with a current density of 10 mA cm−2. We performed TEM tests on the RuYO2−x/C(0.2:0.1) catalysts before and after HER electrocatalysis. The chronoamperometry curve shown in Fig. 4f for RuYO2−x/C(0.2:0.1) indicated that the current density was maintained with negligible loss during a 24 h stability test, which revealed the excellent long-term electrochemical stability of RuYO2−x/C(0.2:0.1). The stability of the morphology was the main reason for the stable performance of the RuYO2−x/C(0.2:0.1) catalysts. The inset in Fig. 4f displays HRTEM images of the RuYO2−x/C(0.2:0.1) sample after stability testing, which was accompanied by the FFT patterns and IFFT patterns of the selected areas in Supplementary Fig. S12b. Supplementary Fig. S12c reveals that the lattice fringe spacing was 0.301 nm, which corresponded to the (401) crystal surface of Y2O3. The distributions of elements in RuYO2−x/C(0.2:0.1) after the stability test are shown in Supplementary Fig. S12d, i. The composition and distributions of the elements in RuYO2−x/C(0.2:0.1) barely changed after the stability test. In addition, we determined the effects of different loadings of RuYO2−x/C(0.2:0.1) catalyst inks on the glassy carbon electrode. The best HER performance of 56 mV was reached when the ink concentration on the glassy carbon electrode was 0.42 μg cm−2. As shown in Supplementary Fig. S13, the HER performance gradually stabilized as the catalyst load on the glassy carbon electrode was increased. Supplementary Fig. S14 shows photographs of the nickel foam used for stability testing. A comprehensive consideration of the hydrogen evolution potential, Tafel slope, electric double-layer capacitance, and impedance of RuYO2−x/C(0.2:0.1) indicated that the nanomaterials constructed from Ru, Y and graphene accelerated charge transfer and thus improved the hydrogen evolution performance of the materials. Subsequently, as shown in Table 146,47,68,69,70,71,72,73,74,75, we also compared the performance with those of other reported electrocatalysts, and the performance of our synthesized RuYO2−x/C catalyst was superior. First, the capacities of Ru-based catalysts to bind hydrogen are similar to those of Pt, and they have excellent catalytic ability for the HER. Although it is also a precious metal, Ru is relatively inexpensive and more abundant than Pt, making it the most promising substitute for Pt. We combined Ru with the rare earth element Y. Because Y itself has good electrical conductivity and a smaller ionic radius than Ru, it formed oxygen vacancies and changed the structure. In addition, the EDS and XPS quantitative analyses showed that the contents of Ru in the RuYO2−x/C catalysts were low, and appropriate proportions of Ru and Y synergistically enhanced the electrocatalytic activity, so RuYO2−x/C exhibited excellent HER catalysis.

Table 1 Comparison of reported electrocatalysts on basic HER activity.
Full size table

Electrocatalytic enhancement mechanism

The alkaline HER follows the Volmer-Heyrovsky or Volmer-Tafel mechanistic pathways (Volmer: H2O + M + e−→M-H* + OH−; Heyrovsky: H2O + M-H*+e−→H2 + OH− + M; Tafel: 2 M-H* → H2 + 2 M). To gain more insight into the excellent electrocatalytic hydrogen evolution performance of RuYO2−x/C in alkaline media, we simulated the RuYO2−x bimetallic nanoparticles and HxYO2−x to determine their effects on the electronic structure at the atomic level (Fig. 5a, b)76. Herein, we selected the (222) surface of RuYO2−x/C and the corresponding TEM and XRD results. A highly active catalyst enabling hydrogen production from water splitting effectively promotes hydrogen release and hydrogen adsorption. The EF value of HxYO2−x was much larger than that of RuYO2−x after the combination of Ru and HxYO2−x. This indicated that electron transfer occurred during the combination of Ru and HxYO2−x, thus improving the HER performance77. In alkaline media, the overall HER reaction pathway includes the dissociation of H2O and the formation of adsorbed hydrogen intermediates, as well as ultimate hydrogen generation78,79,80. Therefore, superior alkaline HER electrocatalysts should simultaneously exhibit moderate H binding energy and a relatively low H2O dissociation barrier. Figure 5c, d displays the H adsorption energies (ΔEH) calculated for the RuYO2−x and HxYO2−x surfaces, and the corresponding H adsorption structures are shown in Supplementary Fig. S15. Subsequently, we also calculated the hydrogen adsorption free energies (ΔGH*) for RuYO2−x (222) and HxYO2−x to evaluate their HER activities; higher ΔGH* values indicate weaker hydrogen adsorption, and vice versa81. From a thermodynamic perspective, the ideal Gibbs free energy for the adsorbed hydrogen atoms should be close to zero82.

Fig. 5: DFT-calculated density of states (DOS) with corresponding optimized structures.
figure 5

a, b Molecular models of RuYO2−x (222) and HxYO2−x. c, d PDOS calculations for RuYO2−x and HxYO2−x.

Full size image

An electron density difference (EDD) analysis (Fig. 6a, b) revealed the electronic structure of RuYO2−x and confirmed that the Y sites had accumulated electron density, while the Ru centers lost electron density and were considered to be the superior activation sites for OH adsorption, which was consistent with the XPS data. Hydrogen adsorption was extremely weak on the external surfaces of HxYO2−x, but it was much stronger on the inner surfaces, suggesting that the interior of HxYO2−x was more favorable for hydrogen adsorption. In conjunction with the previous UPS analysis, a reasonable explanation for hydrogen spillover from HxYO2−x to Ru is as follows: the difference in WF values for RuYO2−x and HxYO2−x led to electron accumulation at the subsurface of HxYO2−x, which enhanced hydrogen adsorption and moved internal protons to the external surface. The Gibbs free energy for adsorption at standard atmospheric pressure (Gads) was defined as Gads=Eads + ΔG(T), where ΔG(T) is the sum of Gibbs free energy corrections at temperature T. Hence, we concluded that ΔG(T) =ΔZPE + ΔH(T) − TΔS(T), where ZPE, ΔH(T), and ΔS(T) represented the zero-vibration energy change, the enthalpy change, and the entropy change after adsorption. Figure 6c and Supplementary Fig. S16 indicate that the ΔGH* of RuYO2−x was closer to the optimal value for the HER (ΔGH* = 0 eV) than those of the other substrates, which supported the experimental results indicating that RuYO2−x showed better HER activity than HxYO2−x. Overall, the RuYO2−x catalyst with more Ru-O-Y bonds and VO significantly reduced the energy barrier for breaking the H-OH bonds and accelerated water dissociation. In addition, the strong metal-support interactions resulted in the optimum energy for hydrogen adsorption and desorption. These phenomena synergistically rationalized the enhanced activity and favorable kinetics seen for RuYO2−x in catalytic hydrogen evolution in alkaline electrolytes. The combination of Ru and HxYO2−x was a key factor determining the HER activity.

Fig. 6: DFT-calculated reaction energies of the HER with charge difference maps.
figure 6

a, b Charge density distributions of RuYO2−x (222) and HxYO2−x. c Calculated free energy diagrams for the HER run on RuYO2−x (222) and HxYO2−x.

Full size image

A mechanism was proposed to explain the greatly enhanced HER activity of RuYO2−x/C in alkaline media. Under an applied cathodic potential, protons in the electrolyte were inserted into the oxygen-deficient HxYO2−x and then combined with Ru on a graphene substrate to produce RuYO2−x/C. The abundance of oxygen vacancies in RuYO2−x/C markedly increased the proton storage capacity and enhanced charge transfer. As a result, the oxygen-deficient HxYO2−x acted as a proton reservoir to supply protons to the Ru surface, which recombined to evolve molecular hydrogen. As the overpotentials were increased, the HxYO2−x dissociated water to generate protons, which also spilled over to Ru. The hydrogen spillover from the HxYO2−x to Ru changed the rate-limiting step of the HER on Ru in a neutral alkaline solution from water dissociation to hydrogen recombination, which greatly improved the HER kinetics.

Conclusions

In summary, we designed a high-density RuYO2−x/C bimetallic nanocomposite with low Ru loadings and grown on two-dimensional graphene sheets. The interactions between the Ru atoms and HxYO2−x were key to the catalytic performance of RuYO2−x/C in hydrogen production via hydrolysis. To demonstrate the high catalytic activity of Ru combined with HxYO2−x, a number of RuYOx/C catalysts with Ru atoms combined with yttrium monomers were also prepared for comparison. Additionally, the presence of Vo was crucial in promoting the hydrogen evolution reaction. With the same ratio of bimetallic atoms, it is noteworthy that RuYO2−x/C exhibited high HER activity in alkaline media as well as long-term stability. Therefore, RuYO2−x/C exhibits potential for application as a water-splitting catalyst.

Methods

Materials

Ruthenium(III) trichloride hydrate (RuCl3·xH2O, 99%), yttrium (Y, 99.9%) graphene (C, >95%), N,N-dimethylformamide (DMF, 99.5%), Nafion solution (5wt.%, contain 15–20% water), potassium hydroxide (KOH, 90%) and ethanol (C2H5OH, 99.9%) were purchased from Aladin Reagent. Hydrogen peroxide (H2O2, ≥30% reagent) was purchased from Greagent. All materials were used directly without any purification. The 18 MΩ cm−1 deionized water was prepared with an ultra-pure purifying device.

Synthesis of HxYO2 − x

Yttrium power, 0.1 g, was weighed and added to 10 mL of anhydrous ethanol, then 0.4 mL of hydrogen peroxide solution was added to the mixed solution with a pipette; the solution was stirred vigorously for 18 h, placed into an oven at 60 °C to dry for 1 h and then removed.

Synthesis of Ru-YO2 − x/C

First, 3 mg of graphene was weighed into a centrifuge tube with 10 mL of deionized water. Then, 1 mL of anhydrous ethanol and DMF was added, and the mixture was sonicated for 30 min. RuCl3‧xH2O (0.2 mmol) and HxYO2−x (0.1 mmol) were added to the sonicated graphene mixture and sonicated for 30 min.

The mixed liquid was transferred to a 20 mL stainless steel autoclave in a PTFE-lined bottle and placed in an oven at 120 °C for 16 h. After the reaction was completed and the autoclave had cooled naturally to room temperature, the product was washed by centrifugation with anhydrous ethanol three times and placed in an oven at 60 °C until the product was completely dry. The product was removed and ground to powder.

The Ru:Y molar ratios for the Ru-YO2−x/C catalysts prepared in this study were 0.2:0.1, 0.2:0.2, and 0.2:0.4.

Synthesis of Ru-YOx/C

The Ru-YOx/C catalysts were prepared in the same way as the Ru-YO2−x/C catalysts, except that the HxYO2−x was replaced with Y powder.

Preparation of the supported catalysts

Four milligrams of catalyst was weighed into a centrifuge tube, and then 30 μL of Nafion solution and 1 mL of a mixture of water and ethanol (3:1) were added. The tube was sonicated for 40 min, and then 14 μL of liquid was dripped onto a glassy carbon electrode with an area of 0.196 cm2 and allowed to dry naturally until a film was formed on the electrode surface.

Electrochemical measurements

The electrochemical properties of the materials was tested in 1.0 mol L−1 KOH solutions at room temperature with an electrochemical workstation (CHI760E). The working electrode was prepared by depositing the catalyst on a glassy carbon (GC) electrode with an area of 0.19625 cm2. A saturated calomel electrode (SCE) and platinum flake electrode were utilized as the reference electrode and counter electrode, respectively. The polarization curves were acquired from linear sweep voltammetry (LSV) at a sweep rate of 5 mV·s−1. The LSV curve was plotted as the logarithm of the standard potential and current density to obtain the Tafel diagram. Electrochemical impedance spectroscopy (EIS) was tested from 100 kHz to 1 Hz. The electrochemical double-layer capacitance (Cdl) was determined by cyclic voltammetry (CV), and the potential scan rates were 20, 40, 60, 80 and 100 mV·s−1. The measured potentials vs. SCE were referenced to the reversible hydrogen electrode (RHE) with the following equation:

$$E({RHE})=E({SCE})+0.231+0.0591{pH}$$
(1)