Introduction

Hydrogen, as an energy carrier, is critical to utilize the renewable energy including wind, solar and hydropower for sustainable development1,2,3,4,5,6. The alkaline hydrogen evolution reaction (HER) with sluggish kinetics limits the efficiency of H2 generation in water electrolysis and chlor-alkali electrolysis due to the additional step of water dissociation7,8,9,10,11,12. It starts from the cleaving of the H−OH bond coupled with an electron to form the adsorbed hydrogen atom (Had, Volmer step). Then, H2 is produced by the combination of two Had (Tafel step) or the interaction of Had and a water molecule (Heyrovsky step)7,13,14,15. Thus, the properties of water dissociation, and hydroxy radical and hydrogen adsorption are crucial for alkaline HER.

To achieve high activity toward these basic steps in alkaline HER, it is necessary to design a functional composite with multiple active centers to facilitate the overall reaction16,17,18,19,20,21. However, such a comprehensive system not only brings difficulty in material synthesis, but also is harmful to the durability of the catalyst due to element or phase segregation22,23,24,25,26, especially in a harsh chlor-alkali solution. The potential-driven structural reconstruction of a pre-catalyst during the working conditions is an efficient method to enhance the electrochemical performance26,27,28,29. The reconstruction of pre-catalysts could obtain the amorphous or defect-rich structure, increased accessible surface area, optimized adsorption properties, and promoted charge transfer30,31,32,33, thus leading to the improvement of activity and stability. For example, Ni2P could be transformed in situ into Ni2P/NiOx core-shell structure during oxygen evolution reaction34. It could accelerate water adsorption and dissociation kinetics due to unique heterostructures and multiple active centers. Besides, the SrIrO3 perovskite electrode experiences self-reconstruction during the alkaline HER because of lattice Sr2+ leaching35. The formed metallic Ir on the surface of perovskite results in the remarkable activity enhancement as well as excellent stability. However, most pre-catalysts undergo the surface-reconstruction, which leads to the low component utilization as the inert internal part is inaccessible to the surface catalysis. Moreover, the reconstruction degree is highly relevant to the reaction environment, and it may be changed with pH, temperature, electrolyte, and applied potential31,36,37,38, which is disadvantageous to the industrial extreme condition. Therefore, the electrocatalysts consisting of single-component nanoparticles prepared via complete in situ reconstruction are highly desired to avoid the above problems during the alkaline HER.

Herein, we develop a bottom-up evolution route to prepare high-activity Rh nanoparticles via in situ electrochemical reduction of new Cs3Rh2I9 perovskite clusters. These electrochemically reduced Rh nanoparticles could be used as a highly efficient HER catalyst in both alkaline and chlor-alkali electrolyte. The new halide-perovskite compound Cs3Rh2I9 with dimer unit [Rh2I9]3− separated by Cs ions as the pre-catalyst of electrochemically synthesized Rh nanoparticles was synthesized by solid state reaction (Fig. 1a). It could be dissolved in N, N-dimethylformamide (DMF). The unique zero-dimensional structure allows it to be downsized into small clusters on a polar nitrogen-doped carbon (NC) support by a simple dissolution-precipitation method (Fig. 1b). The high surface energy of Cs3Rh2I9 clusters on NC (Cs3Rh2I9/NC) could promote an electrochemical self-reduction and bottom-up evolution, leading to the formation of unique Rh nanoparticles with larger lattice spacings and lower atomic coordination number (Fig. 1b). In sharp contrast, such Rh nanoparticles cannot be formed from bulk Cs3Rh2I9 and liquid reduction of RhCl3 by NaBH4 (Fig. 1c, d). The complete reconstructed Cs3Rh2I9/NC-R could significantly reduce the barrier of water dissociation in alkaline HER. Therefore, Cs3Rh2I9/NC-R exhibits high mass activity of 772.1 mA mg−1Rh in a chlorine-alkali electrolyte, which is about 2.5 times and 35.5 times that of liquid-reduced Rh/NC with similar particle size and electrochemically reduced Cs3Rh2I9-R with the larger size, respectively (Fig. 1e, f). And it also shows the negligible activity loss after 50 h durable measurement.

Fig. 1: Schematic illustration of synthesizing perovskite Cs3Rh2I9 bulk crystals and clusters Cs3Rh2I9/NC, and their electrochemical reduction into metallic Rh particles toward alkaline HER.
figure 1

a Synthesis of Cs3Rh2I9 bulk crystal via a solid reaction using I2, Rh, and CsI at 800 oC. The Cs3Rh2I9 crystal could demonstrate a dissolution-precipitation phenomenon in N, N-dimethylformamide (DMF). b Electrochemical reduction of Cs3Rh2I9 clusters (~1.7 nm) supported on NC (Cs3Rh2I9/NC) to form Cs3Rh2I9/NC-R with a little larger size (~2.2 nm) via a bottom-up evolution route. c Electrochemical reduction of bulk Cs3Rh2I9 to form Cs3Rh2I9-R with vert large particle size via a top-down route. d Liquid reduction of RhCl3 in aqueous to form Rh/NC with average particle size of 2.4 nm as the control catalyst. e HER polarization curves of Cs3Rh2I9/NC-R, Cs3Rh2I9-R, and Rh/NC coated on rotating glassy carbon electrode (GCE, 1600 rpm) in a chlorine-alkali electrolyte. f HER mass activity comparison in a chlorine-alkali electrolyte for Cs3Rh2I9/NC-R, Cs3Rh2I9-R, and Rh/NC electrocatalysts at an overpotential of 50 mV.

Full size image

Results

Synthesis and characterization of bulk and cluster Cs3Rh2I9

As a new compound, Cs3Rh2I9 single crystals were synthesized by the solid state reaction using CsI as the flux. Its structure was determined by single-crystal X-ray diffraction (XRD). It belongs to the hexagonal P63/mmc space group with a = b = 7.9648 (14) Å, c = 20.028 (4) Å, V = 1100.3 (4) Å3, Z = 2, and a calculated density of d = 5.272 g cm−3. Details of the atomic coordinates in the compound are shown in Supplementary Table 1, 2. Cs3Rh2I9 with a zero-dimensional structure consists of alternating hexagonal CsI3 layers in ABAA’BA’ stacking sequence (Fig. 2a). Two adjacent [RhI6]3− octahedrons share three I atoms on the plane B to form a [Rh2I9]3− dimer. The powder XRD pattern of Cs3Rh2I9 is consistent with the simulated result (Fig. 2b), confirming its zero-dimensional perovskite structure39. Cs3Rh2I9 shows a semiconductor behavior with an optical band gap of 1.36 V (Fig. 2c)40. Single-crystal Cs3Rh2I9 is diamagnetic due to the d6 electronic configuration of Rh3+ (Supplementary Fig. 2). Energy dispersive X-ray (EDX) spectroscopy shows homogenous distribution of Cs, Rh, and I elements in crystalline Cs3Rh2I9 (Fig. 2d) and EDX results from TEM show Cs, Rh and I with an atomic ratio of ~3:2:9, consistent with the stoichiometry of Cs3Rh2I9 (Supplementary Fig. 3 and Table 3). The high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) confirms the alternating CsI3 layer structure (Fig. 2e). Cs3Rh2I9 also exhibits good stability as its structure does not change in 1.0 M HCl or KOH for 7 days (Supplementary Fig. 5).

Fig. 2: Structure characterization of bulk Cs3Rh2I9 and nanoclusters Cs3Rh2I9/NC.
figure 2

a Atomic structure, and (b) powder XRD pattern and simulated pattern of Cs3Rh2I9 crystal. c Diffuse-reflectance UV–Vis spectrum of Cs3Rh2I9. The inset shows the corresponding Tauc plot. d SEM-EDX mapping images of Cs3Rh2I9 single crystal. e HAADF-STEM image of Cs3Rh2I9 crystal. f TEM image of Cs3Rh2I9 clusters supported on NC (Cs3Rh2I9/NC, Rh content: 5.8 wt.%). The insets show the corresponding HAADF image (scale bar: 5 nm, up) and particle size distribution (down). g TEM-EDX mapping of Cs3Rh2I9/NC. h XPS spectra of Rh 3d for Cs3Rh2I9 and Cs3Rh2I9/NC. i Fourier transform EXAFS of Rh K-edge in R-space for Cs3Rh2I9, Cs3Rh2I9/NC, and Rh foil.

Full size image

Cs3Rh2I9 can be dissolved in N, N-dimethylformamide (DMF) to form a brownish yellow solution due to the polar aprotic property of DMF (Supplementary Fig. 6). When adding water, Cs3Rh2I9 precipitates without any change of structure (Supplementary Fig. 7, 8 and 10). By this means, the use of polar nitrogen-doped carbon (NC) during the precipitation process can significantly reduce the particle size to form Cs3Rh2I9 nanoclusters on NC (Cs3Rh2I9/NC). Figure 2f shows the uniform distribution of Cs3Rh2I9 nanoclusters with a size of ~1.7 nm. No obvious particles were observed in SEM images, further confirming the uniform and ultra-small size of Cs3Rh2I9 (Supplementary Fig. 9). The structure and composition of Cs3Rh2I9/NC is the same as Cs3Rh2I9 (Fig. 2g and Supplementary Fig. 10). The X-ray photoelectron spectroscopy (XPS) spectra show a positive core level shift of about 0.5 eV for Cs3Rh2I9/NC (Fig. 2h), indicating the strong carrier effect of Cs3Rh2I9 clusters on the NC41. The Rh−I coordination was identified by extended X-ray absorption fine structure (EXAFS; Fig. 2i). The shell at 2.40 Å represents the Rh−I scattering path for Cs3Rh2I9 and Cs3Rh2I9/NC, while the fitting results suggest that an additional Rh−Rh scattering path appears in Cs3Rh2I9/NC (Supplementary Fig. 13 and Table 4). This may be caused by partial decomposition of Cs3Rh2I9 nanoclusters under the high-energy measurement (Supplementary Fig. 14).

Electrochemical reduction of Cs3Rh2I9

Under cathodic potentials, it was found that both Cs3Rh2I9 clusters and single crystals are unstable and could be transformed into Rh particles. Especially, the Cs3Rh2I9 nanoclusters on NC can be electrochemically reduced and assembled into Rh nanoparticles with mean particle size of 2.2 nm via a bottom-up evolution route (Fig. 1b). The reduction peak at about −0.035 V versus reversible hydrogen electrode (vs. RHE) in the first CV curve corresponds to the reduction of Rh3+ (Fig. 3a). The in situ Raman experiments indicate that the characteristic structure of Cs3Rh2I9 disappears when the negative potential was applied (Supplementary Fig. 15). To elucidate the process of reconstruction, the reduction was conducted by the potentiostatic measurement at −0.03 V vs. RHE. The in situ electrochemical Quartz Crystal Microbalance (EQCM) experiment indicates that the quality of the electrode continuously decreases and then stabilizes after about 5 min (Fig. 3b). The Inductively Coupled Plasma Mass Spectrometry (ICP-MS) results show the content of Cs+ ions in the electrolyte continuously increases in the initial 5 min while Rh in the electrolyte was barely detected (Fig. 3c). And the XPS spectra shows neither Cs nor I was detected after reduction (Supplementary Fig. 16), which is also confirmed by ICP-MS results (Supplementary Table 5). Rh mass content in Cs3Rh2I9/NC-R was determined by the ICP-MS to be 5.7 ± 0.8 wt.%, which is very close to the initial loading of 5.8 wt.%. This indicates all Cs3Rh2I9 nanoclusters could be electrochemically reduced to metallic Rh0 nanoparticles. And such a reconstruction process is complete. The ex-situ XPS spectra verify the content of Rh3+ decreases while the content of Rh0 increases with the reduction time (Fig. 3d and Supplementary Fig. 17).

Fig. 3: Reconstruction of clusters Cs3Rh2I9/NC into Rh nanoparticles on NC.
figure 3

a The CV curves from 1st to 100th cycle at 100 mV s−1 for Cs3Rh2I9/NC in 1.0 M KOH. b Mass change of the Cs3Rh2I9/NC electrode monitored by in situ EQCM experiment. c ICP-MS of the Cs and Rh contents in the electrolyte at different reduction time. d XPS spectra of Rh 3d for Cs3Rh2I9/NC at different reduction time. The results in (b–d) were obtained at the potentiostatic measurement at −0.03 V vs. RHE.

Full size image

TEM measurements confirm the uniform distribution of Rh particles on the NC after electrochemical reduction (Cs3Rh2I9/NC-R, Fig. 4a and Supplementary Fig. 18) and the Rh particle size (~2.2 nm) is evidently larger than that of Cs3Rh2I9 nanoclusters (~1.7 nm), implying a bottom-up evolution route under cathodic potential. The linear electron energy loss spectroscopy analysis suggests the element in the particle is Rh (Fig. 4b). The bulk Cs3Rh2I9 can also be reconstructed in this way, but it does not reach a stable state within 300 CV cycles due its large size (Supplementary Fig. 19). After reduction (Cs3Rh2I9-R), its edge consists of numerous Rh particles with larger size (4.3 ± 1.2 nm) compared to the product from Cs3Rh2I9/NC (Supplementary Fig. 20, 21). The Rh K-edge X-ray absorption near edge structure spectra indicates the adsorption edge of the reduced Cs3Rh2I9/NC (Cs3Rh2I9/NC-R) shifts to lower energy compared to the initial state (Supplementary Fig. 22). Meanwhile, the shell in EXAFS for the Cs3Rh2I9/NC-R shifts to 2.45 Å (Fig. 4c), similar to that of Rh foil. The fitting results show that its Rh−Rh coordination number is only 8.0 (Fig. 4c and Supplementary Table 6), showing the small particle size. The wavelet transform (WT)-EXAFS analysis shows the first shell of Cs3Rh2I9/NC-R domain at R = 2.45 Å and k = 9.90 Å−1 (Fig. 4d), similar to the Rh foil (R = 2.43 Å and k = 9.80 Å−1) and different from Cs3Rh2I9/NC with Rh−Rh scattering path (R = 2.40 Å and k = 9.50 Å−1).

Fig. 4: Characterization of Cs3Rh2I9/NC-R prepared by electrochemical reduction of Cs3Rh2I9/NC.
figure 4

a TEM image of Cs3Rh2I9/NC-R, showing uniform dispersion of Rh nanoparticles. The inset shows the corresponding HAADF image (scale bar: 5 nm, up) and particle size distribution (down). b The linear electron energy loss spectroscopy of Cs3Rh2I9/NC-R. c Fourier transform EXAFS of the Rh K-edge in R-space for Cs3Rh2I9/NC, Cs3Rh2I9/NC-R and Rh foil. d WT-EXAFS images of Rh for Cs3Rh2I9/NC, Cs3Rh2I9/NC-R and Rh foil. e HAADF-STEM image of Cs3Rh2I9/NC-R, showing a single Rh nanoparticle with larger lattice spacing and twin-GBs. f HRTEM image of Rh nanoparticles (Cs3Rh2I9-R) derived from the top-down electrochemical reduction of bulk Cs3Rh2I9 crystal. g DFT models of Rh with larger lattice spacing and GBs (up panel) and Rh with regular lattice spacing (down panel).

Full size image

Moreover, the HAADF-STEM images indicate the Rh nanoparticles in Cs3Rh2I9/NC-R are rich with grain boundaries (GBs) and large lattice spacings (0.225–0.230 nm) (Fig. 4e and Supplementary Fig. 24). In sharp contrast, the particles in electrochemically reduced Cs3Rh2I9 (Cs3Rh2I9-R) form bulk Cs3Rh2I9 show smaller plane spacings of 0.220 nm and no evident GBs could be observed (Fig. 4f and Supplementary Fig. 21). The formation of twinned Rh with large lattice spacings in Cs3Rh2I9/NC-R may be ascribed to the coupling of smaller Rh clusters with high surface energy during the electrochemical reduction process. Limited by the size of the Cs3Rh2I9 cluster (~1.7 nm), the formed Rh clusters could combine with neighboring clusters into particles (~2.2 nm) to decrease the surface energy (Supplementary Fig. 25). However, the Rh particles formed from bulk Cs3Rh2I9 undergo a top-down process of bulk decomposition. They have particle sizes greater than 2 nm and thus are energetically stable enough to not combine into twin crystals. Furthermore, the twinned Rh in Cs3Rh2I9/NC-R exhibits tensile stress with the enlarged lattice fringe of 0.5–5.5% (Fig. 4e, Supplementary Fig. 24). The corresponding GB was established by density functional theory (DFT) calculation (Fig. 4g) to promote tensile stress of the nearby Rh atoms (from 2.71 to 2.74 Å), suggesting the rich GBs in twinned Rh can stabilize the enlarged (111) lattice. The Rh particle with such tensile stress is considered to facilitate H2O dissociation in alkaline HER2.

HER activity evaluation

The HER activity was first evaluated in 1.0 M KOH electrolyte. Before measurement, the electrode was in situ reduced under the CV between 0.075 and -0.175 vs. RHE at a scan rate of 100 mV s−1 to achieve a stable stage. The Rh/NC with mean size of ~2.4 nm was synthesized by liquid chemical reduction for comparison (Supplementary Fig. 26). The Rh mass content in Rh/NC is 7 wt.%, and the HRTEM implies the lattice spacing of Rh particle is 0.220 nm, similar to that of Cs3Rh2I9-R without tensile stress. As shown in Fig. 5a, the overpotential at 10 mA cm−2 for Cs3Rh2I9/NC-R composed of Rh twin nanoparticles is only 25 mV, evidently lower than those of Cs3Rh2I9-R (123 mV), Pt/C (32 mV) and Rh/NC (41 mV). Moreover, the Tafel slope (30.3 mV dec−1) of Cs3Rh2I9/NC-R is also smaller than that of chemically reduced Rh/NC without GBs and tensile stress (34.8 mV dec−1) (Supplementary Fig. 28a). It implies that the GBs and tensile stress in Rh nanoparticles have a significant effect on boosting alkaline HER. Besides, Cs3Rh2I9/NC-R exhibits high mass activity of 839.8 mA mg−1Rh, 21.6 times that of Cs3Rh2I9-R and 2.4 times that of Rh/NC (Supplementary Fig. 28b). The mass loading of Cs3Rh2I9/NCs was also optimized and compared (Supplementary Fig. 29). The electrochemical surface area (ECSA) is enhanced with the increased mass loading, but the optimal activity is achieved at a Rh mass loading of 5.8 wt.% (Cs3Rh2I9/NC-R; Supplementary Fig. 30-32). In addition, the H2 production faradaic efficiency for twinned Rh (Cs3Rh2I9/NC-R) was confirmed to be nearly 100% through a drainage method (Supplementary Fig. 33). The HER activity of Cs3Rh2I9/NC-R outperforms most of reported Rh-based electrocatalysts (Fig. 5b) and other advanced alkaline electrocatalysts (Supplementary Table 7).

Fig. 5: HER activity in 1.0 M KOH and chlor-alkali electrolyte.
figure 5

a LSV curves of various electrocatalysts including commercial Pt/C (Pt content: 20 wt.%) coated on rotating glassy carbon electrode (GCE, 1600 rpm) in 1.0 M KOH. The catalyst loading amount on GCE is 0.764 mg cm−2. For Rh-based samples, the calculated Rh loading amounts on GCE are 0.045, 0.090 and 0.053 mg cm−2 for Cs3Rh2I9/NC-R, Cs3Rh2I9-R, and Rh/NC, respectively. b Comparison of overpotential at 10 mA cm−2 and Tafel slope for various Rh-based catalysts in 1.0 M KOH. c LSV curves of various electrocatalysts coated on rotating GCE (1600 rpm) in chlorine-alkali electrolyte. d Mass activity normalized to the mass of Rh in chlorine-alkali electrolyte. e Comparison of mass activity and area activity at the overpotential of 50 mV in chlorine-alkali electrolyte. All data show the mean and standard deviation through three repeated measurements. f Stability of Cs3Rh2I9/NC-R at 10 mA cm−2 in a chlorine-alkali electrolyte.

Full size image

The HER activity in the simulated chlorine-alkali electrolyte (3.0 M NaOH + 3.0 M NaCl) was further evaluated in a three-electrode system. As shown in Fig. 5c, the overpotentials for Cs3Rh2I9/NC-R are 21, 65, and 107 mV to reach current densities of 10, 50, and 100 mA cm−2, respectively, significantly lower than those of Cs3Rh2I9-R, Pt/C, and Rh/NC. The mass activity of Cs3Rh2I9/NC-R is 772.1 mA mg−1Rh at −50 mV vs. RHE, outperforming Cs3Rh2I9-R (21.7 mA mg−1Rh) and Rh/NC (307.8 mA mg−1Rh) (Fig. 5d, e). The area activity normalized by ECSA of Cs3Rh2I9/NC-R is 0.067 mA cm−2, manifesting a factor of 2.0 increase than that of Rh/NC (Fig. 5e and Supplementary Fig. 35b). These results confirm the excellent intrinsic activity of Cs3Rh2I9/NC-R enriched with defects. The enhanced activity may be ascribed to the reconstruction process. Cs3Rh2I9/NC-R with unsaturated atom coordination and more accessible area could accelerate water adsorption and dissociation processes, thus leading to the better performance. Furthermore, Cs3Rh2I9/NC-R exhibits good durability with negligible activity loss after 50 h (Fig. 5f). After the durability measurement, the twin Rh nanoparticles show little agglomeration while the chemical state remains unchanged (Supplementary Fig. 37, 38).

To further disclose the origin of high activity of Cs3Rh2I9/NC-R in the alkaline HER process, the DFT calculation was performed. Based on the above characterizations, the Cs3Rh2I9 clusters-derived Rh nanoparticles possess larger lattice spacings and more GBs that liquid-reduced Rh nanoparticles and Rh particles from electro-reduction of bulk Cs3Rh2I9. Thus, different Rh models with larger lattice spacings and more GBs were built to study the HER mechanism in alkaline. As shown in Supplementary Fig. 39, the increase of H−O−H angle and water molecule binding energy agrees well with the degree of distortion and crystal tension, indicating the enhanced ability of water dissociation. Moreover, the largest H−O−H angle of 105.25˚ is found when water molecule is adsorbed on GBs with low coordination number, which further shows GBs would activate the H−OH bond. The water disassociation, as a pre-reaction to form an adsorbed proton, plays a more critical role in alkaline hydrogen evolution8,42. The linear Brønsted–Evans–Polanyi relationship between adsorption energy and dissociative kinetic barrier of H2O allows the use of binding energy of water molecule as the activity descriptor for alkaline HER. The sites at GBs and tensed Rh atoms show high H2O binding energy. The enhanced water dissociation ability in Cs3Rh2I9/NC-R was further verified by in situ Raman spectra. Supplementary Fig. 40a shows obvious interfacial water on Cs3Rh2I9/NC-R surface in 1 M KOH solution over the potential range from 0 to −0.3 V (vs. RHE). The peaks at 1588 and 1632 cm−1 correspond to the G band of carbon substrate and the adsorbed water on Rh, respectively43,44. As the potential decreased, the intensity of H–O–H bending increases sharply. However, such a phenomenon is absent in Rh/NC synthesized by chemical reduction (Supplementary Fig. 40b). Thus, the abundant and multiply active sites on twin crystal Rh are effective for cleaving the H−OH bond, contributing to the high electrochemical performance toward HER under alkaline condition.

Discussion

The defects-rich Rh nanoparticles with average size of ~2.2 nm were prepared by in situ electrochemical reduction of perovskite Cs3Rh2I9 cluster via a bottom-up evolution route. The reduction was investigated by in situ EQCM, ex-situ ICP-MS, and XPS. The as-formed twin crystal Rh nanoparticles with tensile stress exhibits excellent activity and stability in alkaline hydrogen evolution reaction. In 1.0 M KOH, the Cs3Rh2I9/NC-R catalyst showed a low overpotential of 25 mV at the current density of 10 mA cm−2 and a small Tafel slope of 30.3 mV dec−1. Cs3Rh2I9/NC-R exhibits high mass activity of 839.8 mA mg−1Rh, 21.6 times that of Cs3Rh2I9-R with bigger size and 2.4 times that of liquid-reduced Rh/NC. In a chlor-alkali electrolyte, the area activity of Cs3Rh2I9/NC-R (0.067 mA cm−2ECSA at −50 mV vs. RHE) manifests a factor of 4.1 and 2.0 activity increase compared to Cs3Rh2I9-R and Rh/NC, respectively. Moreover, it exhibits good durability with negligible activity loss for 50 h HER measurement. The DFT calculation revealed that Rh nanoparticles of Cs3Rh2I9/NC-R enriched with multiply catalytic sites could accelerate the activation of adsorbed water molecule, thereby smoothing the whole alkaline HER. The study presents new insights into preparing small-sized nanoparticles via in situ electrochemical reconstruction for energy electrocatalysis.

Methods

Synthesis of Cs3Rh2I9 crystal

The Cs3Rh2I9 crystal was prepared via solid state reaction. 50 mg rhodium, 185 mg iodine and 1.5 g CsI were mixed by grinding in Ar glovebox to prevent the influence of water. CsI was served as both raw material and flux. The above powder was annealed at 800 °C for 2000 min in the evacuated quartz tube and cooled to room temperature at the rate of 3 °C min−1. Finally, the product was washed with deionized water for several times and dried in vacuum at room temperature.

Synthesis of NC

NC was prepared by the sol–gel method according to our previous report.41 It started from mixing 1.8 mL formaldehyde and 1 g melamine in 20 mL deionized water under stirring at 50 °C for 1 h. Then, 4.5 g MnCl2 and 6 g PEG were added into the above solution and continuously stirred at room temperature to form a uniform sol. The sol was transferred into a culture dish and dried at 80 °C for 24 h to form a gel precursor. The gel was cut into small slices and pre-carbonized at 400 °C for 2 h in an Ar atmosphere. After grounding the precursor into powder, it was annealed at 900 °C with a heat rate of 3 °C min−1 in the Ar atmosphere. The final product NC was obtained by acid treatment to remove the metal impurities.

Synthesis of nanoclusters Cs3Rh2I9 supported on NC (Cs3Rh2I9/NC)

10 mg Cs3Rh2I9 crystal was added into 40 ml N, N-dimethylformamide (DMF) and stirred at 60 °C to form the brown solution. Then, 10 mg NC was added and the mixture was continually stirred for 1 h. Subsequently, the above mixture was slowly added into 200 ml H2O under fiercely stirring. The induced polar protic solvent with slow SN2 reaction kinetics results in the precipitation of Cs3Rh2I945. The Cs3Rh2I9/NC was collected by suction filtration and washed by water and ethanol for several times. And it was dried in vacuum at 60 °C for 12 h. Cs3Rh2I9/NC with different Rh contents (x wt.%, x = 2.7, 4.3, 5.8, or 7.1) was prepared by changing mass loading of Cs3Rh2I9, and x corresponds to the Rh content. The optimized Cs3Rh2I9/NC (5.8 wt.%) is named as Cs3Rh2I9/NC in the main text and SI if no specific note was mentioned. The re-precipitated Cs3Rh2I9 without adding NC was obtained as the control sample.

Synthesis of Rh/NC by liquid reduction

RhCl3 (7.1 mg) was dissolved in 20 mL N-methyl pyrrolidone (NMP), followed by the addition of 46.5 mg NC. After stirring for 1 h, 0.05 g sodium borohydride in NMP was added into the above solution drop by drop. After stirring for 20 h, Rh/NC was collected by suction filtration and then annealed in 10% H2/Ar at 300 °C for 30 min. The elemental analysis shows the mass content of Rh in Rh/NC is 7.0 wt.%.

Characterization

Single-crystal XRD data was obtained on a Bruker D8 QUEST diffractometer with Mo-Kα (λ = 0.71073 Å) radiation at 300 K. The crystal structure was solved and refined using APEX3 program. Powder XRD was performed on a Bruker D8 Advance diffractometer equipped with mirror-monochromatized source of Cu Kα radiation (λ = 0.15406 nm). The ultraviolet–visible (UV–Vis) light diffuse-reflectance spectra were measured on a UV-4100 spectrophotometer operating from 2000 to 300 nm at room temperature and the BaSO4 powder was used as a 100% reflectance standard. Low-temperature electrical resistivity was measured using a Physical Properties Measurements System (PPMS-Dyna Cool, Quantum Design). SEM was conducted on the JSM-7800F. TEM was conducted on the JEM-2100F and Talos F200X G2. HAADF-STEM and EELS measurements were obtained from aberration-corrected TEM (Hitachi HF5000) and JEOL Triple-C TEM. The chemical states were investigated by X-ray photoelectron spectroscopy (XPS, Thermo Fisher Scientific ESCA Lab 250Xi spectrometer) with focused monochromatic Al Kα radiation (1486.6 eV, 150 W; 500 μm diameter of irradiated area). Ion concentration in electrolyte was determined by the inductively coupled plasma mass spectrometry (XII, Thermo Fisher Scientific). X-ray absorption fine-structure spectroscopy (XAFS) was performed at the Materials Research Collaborative Access Team (MRCAT), Sector 10-BM line at the Advanced Photon Source at Argonne National Laboratory46. Data were processed and fitted using Athena and Artemis for the IFEFFIT suite47,48. All spectra were prepared for Fourier Transform using a Hanning window ranging from 2.0 Å−2 < k < 14 Å−1 with dk = 2 Å−1 and simultaneously fitted in k, k2, and k3 weightings using a Hanning window of 1.8 Å < R < 2.8 Å with dR = 0.2 Å. Wavelet transforms were obtained from processed data using Larch49.

Hydrogen evolution reaction experiments

The electrochemical performance was conducted on the three-electrode system using graphite rod as counter electrode and Hg/HgO as the reference electrode at 25 oC. 5 mg catalyst and 25 uL Nafion (5 wt.%) were added into 475 uL ethanol to form the homogeneous catalyst ink. And the working electrode was prepared by adding 15 uL ink on a rotating GCE (area: 0.1963 cm−2). The catalyst loading amount on GCE is 0.764 mg cm−2. And the calculated Rh loading amounts on GCE are 0.045, 0.090, and 0.053 mg cm−2 for Cs3Rh2I9/NC-R, Cs3Rh2I9-R, and Rh/NC, respectively. Moreover, the commercial Pt/C (Pt content: 20 wt.%) was also used as the control electrocatalyst. The HER measurements were performed in the Ar-saturated electrolyte at the rotating speed of 1600 rpm. Before measurement, the electrode was activated (reduced) under the CV between 0.075 and -0.175 vs. RHE at a scan rate of 100 mV s−1 for 100 cycles. The linear sweep voltammetry (LSV) curves were recorded at the scan rate of 5 mV s−1 with the IR-compensation 90%. The ECSA was obtained from the equation ECSA = Cdl/Cs, where the electrochemical double layer capacitance (Cdl) was obtained from the CV measurement at different scan rates and the specific capacitance (Cs) was 0.4 μF cm−2 in 1.0 M KOH. In chlorine-alkali electrolyte (3.0 M NaOH + 3.0 M NaCl), the Hg/HgO electrode was protected by the salt bridge (1.0 M KOH).

The durable measurement was conducted in a two-compartment cell using the cation exchange membrane as the separator. And the working electrode was prepared by dropping the ink on the carbon cloth (catalyst loading amount: 1 mg cm−2).

Reversible hydrogen electrode (RHE) calibration

The calibration was performed in the H2-saturated electrolyte using Pt foils as counter electrode and working electrode. CV measurements were carried out at the scan rate of 1 mV s−1. The average potentials at the current of zero were set as the thermodynamic potential of RHE.

Electrochemical quartz crystal microbalance experiment

The EQCM experiment was conducted on the QSense Explorer instrument (Biolin Scientific AB, Sweden). The working electrode was prepared by spin-coating the catalyst ink on the Au-coated quartz crystal disk (5 MHz). In situ EQCM was measured simultaneously with potentiostatic measurement at −0.03 V vs. RHE.

Hydrogen production faradic efficiency

The faradic efficiency was determined by the drainage method at 25 °C. A constant current was applied on the electrode, and the electrocatalysis time and the volume of evolved hydrogen were recorded synchronously. Each experiment was repeated three times. The experimental content of produced hydrogen was calculated by the Ideal Gas Law and the theoretical content was determined by the Faraday Law.

Computational methods

All theoretical calculations were performed using DFT, as is implemented in the Vienna ab initio simulation package (VASP). The electron exchange and correlation energy functionals are treated using the generalized gradient approximation, as is captured using the Perdew-Burke-Ernzerh functional (GGA-PBE). Iterative solutions of the Kohn–Sham equations were done using a plane-wave basis set defined using a kinetic energy cutoff of 500 eV. The k-point sampling was obtained from the Monkhorst–Pack scheme with a (6 × 6 × 1) mesh. respectively. Rh (001) surface structures with lattice tensile of 1.5% and 3% were built and optimized to simulate active sites with different crystal strain. Surface structures with Rh (110) GB and standard Rh (001) were chosen to reflect the properties of GB and monocrystal Rh, respectively. Water bonding energy (\({\triangle }{E_{{{{\rm{H}}}_{2}}{{{\rm{O}}}}}} \)) was calculated by the following equation:

$${\triangle }{E_{{{{\rm{H}}}_{2}}{{{\rm{O}}}}}}={E_{{{{\rm{H}}}_{2}}{{{\rm{O}}}}}}+{E_{{{\rm{sub}}}}}-{{E}_{{{\rm{total}}}}}$$

whereas Etotal refers to energy of substrate with corresponding adsorbate, \({\triangle }{E_{{{{\rm{H}}}_{2}}{{{\rm{O}}}}}} \) refer to energy of one H2O molecule, T is selected by room temperature (298 K), and ZPE refers to zero point energy.