Synthesis of Ni1−xFexOyHz. Ni1−xFexOyHz was synthesized via a modified hydrothermal procedure38. Desired ratios of Ni and Fe were obtained by mixing different stochiometries of Ni(NO3)2 and Fe(NO3)3·9H2O in 50 mL deionized water to reach a total concentration of 20 mM. 35 mM urea was added into the solution to assist hydrolysis. Then, trisodium citrate (0.25 mM) was added into the above solution under stirring. All chemicals were used as purchased. After thoroughly mixing, the solution was sealed in a Teflon-lined stainless-steel autoclave and hydrothermally treated at 150 °C for 48 h. The obtained powder was washed by deionized water for two times and ethanol for one time, and finally dried at 50 °С under air condition.
Synthesis of LaNi1−xFexO3. LaNi1−xFexO3 was synthesized via a sol-gel method39. Desired ratios of La, Ni and Fe were obtained by mixing different stochiometries of La(NO3)3·9H2O, Ni(NO3)2 and Fe(NO3)3·9H2O in 20 mL deionized water. Then, citric acid and acrylamide were added into the above solution under stirring. The molar ratio of total metal ions, citric acid and acrylamide was controlled to be 1:3:9. After thoroughly stirring, the mixture was heated to 110 °С to generate gel, which was subsequently dried in air at 150 °С for 12 h. Finally, after calcination at 800 °C for 6 h, the LaNi1−xFexO3 powder with different Ni/Fe stoichiometries were obtained.
Synthesis of NiFe2O4. NiFe2O4 was synthesized via a similar method as LaNi1−xFexO3. Ni(NO3)2 and Fe(NO3)3·9H2O (molar Ni: Fe=1:2) were dissolved in 20 mL of deionized water. Then, citric acid and acrylamide were added into the above solution under stirring. The molar ratio of total metal ions, citric acid and acrylamide was controlled to be 1:3:9. After thoroughly stirring, the mixture was heated to 110 °С to generate gel. The obtained gel was dried in air at 150 °С for 12 h. Finally, after calcination at 800 °C for 6 h, the NiFe2O4 powder was obtained.
Synthesis of ultra-thin MFeOxHy. Ultra-thin MFeOxHy (M= Mn, Fe, Co, Ni and Cu) were synthesized using formamide as an inhibitor of layer growth30,40, in order to fully expose the lattice oxygen and ensure the total exchange of 16O by isotope labelled 18O in solution. Typically, a 10.0 mL nitrate solution containing 37.5 mM of total metal ions (molar M:Fe=3:1) was mixed with 20.0 mL of 10 mM NaNO3 containing 23 vol% formamide. Under stirring, the solution was heated up to 80 °C. Then, the pH of solution was adjusted to ~10 by dropping 0.25 M KOH. After cooling down to the room temperature, the product was washed with ethanol and deionized water (1:1 vol%) for three times. Finally, the obtained ultra-thin MFeOxHy was dispersed and stored in deionized water.
Off-line physical characterizations.
The XRD patterns were recorded on a Bruker D8 Advance diffractometer with Cu Ka radiation (λ = 1.5418 Å) and a Lynx Eye detector. The specific surface area was determined by BET N2 adsorption isotherms on an ASAP Tristar II 3020 after 20 h outgassing at 80 °C. The N2 adsorption was measured at 77 K in a relative pressure range p/p0 of 0.05-0.3. The ICP-OES tests were performed on Thermo Fisher Scientific iCAP 7400.
OER electrochemical measurements.
The electrochemical measurement was performed using a Biologic VSP-300 potentiostat. All CV tests were measured using a rotation disk electrode (RDE) with 1600 rpm using a PINE MSR rotator in a 1 M purified Fe free-KOH aqueous solution (the purification procedure was rigorously following a previously reported method26). All catalysts were loaded on glassy carbon electrodes (as working electrodes) and mixed with a certain amount of carbon black to ensure the conductivity. A platinum wire and Hg/HgO (CHI instruments) were used as the counter and reference electrode, respectively. The potential of Hg/HgO was routinely calibrated against a reversible hydrogen electrode (RHE, eDAQ Inc.). All potentials in electrochemical tests were converted to the RHE scale with resistance compensation, which was determined by electrochemical impedance measurement.
The catalyst inks were prepared by mixing 2 mg of the catalyst, 2 mL deionized water, 2 mL ethanol, 0.6 mg carbon black and 20 μL Nafion (5 wt %). All inks were ultrasonicated for at least 30 min to ensure the dispersity. Then, 20 μL of the as-prepared ink was dropped onto a glassy carbon electrode (0.196 cm2) and dried in air naturally. The mass loading of catalyst for OER test is calculated to be ~51 μg/cmgeo2.
Operando Raman measurement.
All Raman spectra were recorded by a Horiba JY XploRA Raman spectrometer under a 638 nm laser excitation. The Raman shift was calibrated using the silicon peak at 521 cm−1 before measurements. We customized a three-electrode cell to probe the structural evolution of catalysts (see the schematic illustration in Supplementary Fig. 29). Catalysts were loaded on an electrochemically roughened Au electrode to obtain the surface-enhanced Raman signal (SERS)41. To prepare a SERS substrate, 10 nm Ti/100 nm Au electrode was thermally evaporated onto a silicon substrate. The roughened Au can be obtained by ~20 oxidation-reduction CV cycles between −0.3 V (30 s hold) and 1.3 V (1.3 s hold) vs. Ag/AgCl in a 0.1 M KCl aqueous solution30.
Operando XANES and EXAFS measurement.
The XAS measurements were carried out at BL11B beamline of Shanghai Synchrotron Radiation Facility (SSRF). For all XAS tests, monochromatized X-ray beam was provided by a double-crystal Si (111) monochromator, and the photon energies were calibrated to the first inflection point of the K-edge from Ni foil at 8333 eV and Fe foil at 7112 eV, respectively. The reference spectra were recorded in a transmission mode, while all the spectra of catalysts were measured in a fluorescence mode with a Lytle detector filled with Argon gas. All operando measurements were carried out in a three-electrode cell (see Supplementary Fig. 30). The catalysts were loaded on a carbon paper (~1 mg/cm2) as the working electrode. A platinum wire and Hg/HgO were used as the counter and reference electrode, respectively, and the electrolyte was 1 M purified KOH. Ni K-edge spectra was recorded from 8133 eV to 9128 eV and Fe K-edge in the range of 6912 eV to 7912 eV. The data collected were normalized and processed in the ATHENA program integrated with IFEFFIT software package.
EXAFS analysis and fittings.
The acquired EXAFS data were processed according to standard procedures using the ATHENA module implemented in the IFEFFIT software packages42. The k3-weighted EXAFS spectra were obtained by subtracting the post-edge background from the overall absorption and then normalizing with respect to the edge-jump step. Subsequently, k3-weighted χ (k) data of Ni K-edge and Fe K-edge were Fourier transformed to real (R) space using a hanning windows (dk = 1.0 Å−1) to separate the EXAFS contributions from different coordination shells.
To obtain the quantitative structural parameters around central atoms, least-squares curve parameter fitting was performed using the ARTEMIS module of IFEFFIT software packages42, which is based on the following equation:
Nj denotes the number of neighboring atoms in jth shell at a distance of Rj from the central atom. S02 is an amplitude reduction factor. is the effective curved-wave backscattering amplitude. Rj is the distance between the X-ray absorbing central atom and the atoms in the jth atomic shell. σj2 is the Debye-Waller parameter of the jth atomic shell. λj(k) is the electron mean free path. φij(k) is the ab initio phase function for shell jth.
For as-synthesized Ni0.63Fe0.37OxHy, LaNi0.61Fe0.39O3 and NiFe2O4, both M–O and M–M shells were fitted to distinguish local structures of hydroxide, perovskite and spinel. The coordination numbers (CN) were fixed as the nominal values and K range used for the simulations were 2.5-12 Å−1. For operando XAS, only M–O shell was fitted to monitor M–O local structure change during electrooxidation process. S02 was fixed to 0.70. The K range used in operando data fitting was 2.5-10.5 Å−1 to exclude the noise in high K range. All fitting results are listed in Supplementary Fig. 31, Tables 3 and 6-8.
The TOFFe (assuming all Fe ions are active sites) was calculated by the following equation:
Where i, NA, F and NFe represent the current (A) at an overpotential of 300 mV, Avogadro number (6.022 × 1023 O2 molecules per mol O2), Faraday constant (96,485 C/mol) and the number of all Fe in catalysts, respectively.
The TOFsurface-Fe (assuming Fe ions locating at surface are active) was calculated by the following equation:
To determine the number of surface Fe ions (Nsurface-Fe), we calculated the surface density of 3d metal atoms (ρ) of three structures based on their atomic arrangements, and either XAS or XRD data (see Supplementary Note 2 and Fig. 32). Therefore, we calculated Nsurface-Fe by the following equation:
Where S and Fe atom% represent the surface area determined by BET (see Supplementary Fig. 33 and Table 9) and Fe content in 3d metal atoms. This approach was regarded as a reasonable way to evaluate the intrinsic activities of powdery catalysts as previously reported43,44.
The stochastic surface walking is based on the global neural network potential (SSW-NN) method, as implemented in LASP code (http://www.lasphub.com)45,46, in which the SSW-NN achieves fast global potential energy surface (PES) exploration to resolve the stable phase of Ni-Fe superoxo-hydroxide during the reaction. The Fe-Ni-O-H quaternary element G-NN potential was developed by self-learning the DFT global PES data set, which was generated from the SSW global PES exploration for systems with different Fe-Ni-O-H compositions or structures (see Supplementary Note 3 and Table 11). There are six steps: (1) Generating the global dataset from the SSW global optimization trajectories and computing the dataset using DFT calculation; (2) Training the NN potential with dataset; (3) Benchmarking the accuracy between the current NN potential and DFT calculation for selected structures from SSW trajectories and retraining the NN potential by adding new dataset; (4) Iteratively performing (1–3) steps until the PES deviation is low enough. The accuracy of G-NN potential is typically 5-10 meV/atom for RMSE of energy and 0.1-0.2 eV/Å for RMSE of force; (5) Performing the SSW global optimization on the NN PES for target problem. (6) Recomputing the energy of key structures with DFT calculations. Zero-point energy (ZPE) was calculated from the phono spectra using the density functional perturbation theory to obtain the convex hull diagram.
All DFT calculations were performed using Vienna Ab initio Simulation Package (VASP)47 with projected augmented wave (PAW) pseudo-potentials and the Perdew-Burke-Ernzerhof (PBE) functional with Hubbard term correction48,49. The effective Hubbard term (Ueff) was set at 3.3 eV for Fe and 5.5 eV for Ni in accordance with the linear response approach50,51. The plane-wave cutoff energy was set to 500 eV. The Monkhorst-Pack scheme with k-point meshes of (2×4×2) and (2×4×1) are utilized for bulk and slab systems, respectively. The limited memory Broyden-Fletcher-Goldfarb-Shanno (LBFGS) method is used for geometry relaxation until the maximal force is less than 0.08 eV/Å. The composition of surface with surface superoxide is Ni24Fe8O72H32 in each unit cell with dimensions of 11.84×16.40 Å2. The solvation effect due to the long-range electrostatic interaction was modeled by a periodic continuum solvation model with modified Poisson-Boltzmann equation (CM-MPB)52,53. The computational hydrogen electrode (CHE) approach was utilized to evaluate the thermodynamics of OER on all Ni0.75Fe0.25OyHz surfaces. All the reported free energy changes in the OER refer to the electrode potential (U) at 1.53VRHE.
40. Yu, J., Martin, B. R., Clearfield, A., Luo, Z. & Sun, L. One-step direct synthesis of layered double hydroxide single-layer nanosheets. Nanoscale7, 9448-9451 (2015).
41. Gao, P., Patterson, M. L., Tadayyoni, M. A. & Weaver, M. J. Gold as a ubiquitous substrate for intense surface-enhanced Raman scattering. Langmuir1, 173-176 (1985).
42. Ravel, B. & Newville, M. ATHENA, ARTEMIS, HEPHAESTUS: data analysis for X-ray absorption spectroscopy using IFEFFIT. J. Synchrotron Rad.12, 537-541 (2005).
43. Wei, C. et al. Approaches for measuring the surface areas of metal oxide electrocatalysts for determining their intrinsic electrocatalytic activity. Chem. Soc. Rev.48, 2518-2534 (2019).
44. Sun, Y. et al. Covalency competition dominates the water oxidation structure–activity relationship on spinel oxides. Nat. Catal.3, 554-563 (2020).
45. Huang, S. D., Shang, C., Zhang, X. J. & Liu, Z. P. Material discovery by combining stochastic surface walking global optimization with a neural network. Chem. Sci.8, 6327-6337 (2017).
46. Huang, S. D., Shang, C., Kang, P. L., Zhang, X. J. & Liu, Z. P. LASP: Fast global potential energy surface exploration. WIREs Comput Mol Sci.9, e1415 (2019).
47. Kresse, G. & Furthmuller, J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys. Rev. B54, 11169-11186 (1996).
48. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple. Phys. Rev. Lett.77, 3865-3868 (1996).
49. Cococcioni, M. & de Gironcoli, S. Linear response approach to the calculation of the effective interaction parameters in the LDA + U method. Phys. Rev. B71, 035105 (2005).
50. Li, Y. F. & Selloni, A. Mechanism and activity of water oxidation on selected surfaces of pure and Fe-doped NiOx. ACS Catal.4, 1148-1153 (2014).
51. Li, Y. F. & Liu, Z. P. Structure and water oxidation activity of 3d metal oxides. WIREs. Comput. Mol. Sci.6, 47-64 (2016).
52. Fattebert, J. L. & Gygi, F. Linear-scaling first-principles molecular dynamics with plane-waves accuracy. Phys. Rev. B73, 115124 (2006).
53. Walter, M. G. et al. Solar water splitting cells. Chem. Rev.110, 6446-6473 (2010).