Direct Observation of Transition Metal Ions Evolving into Single Atoms: Formation and Transformation of Nanoparticle Intermediates

Abstract Understanding the dynamical evolution from metal ions to single atoms is of great importance to the rational development of synthesis strategies for single atom catalysts (SACs) against metal sintering during pyrolysis. Herein, an in situ observation is disclosed that the formation of SACs is ascertained as a two‐step process. There is initially metal sintering into nanoparticles (NPs) (500–600 °C), followed by the conversion of NPs into metal single atoms (Fe, Co, Ni, Cu SAs) at higher temperature (700–800 °C). Theoretical calculations together with control experiments based on Cu unveil that the ion‐to‐NP conversion can arise from the carbon reduction, and NP‐to‐SA conversion being steered by generating more thermodynamically stable Cu‐N4 configuration instead of Cu NPs. Based on the evidenced mechanism, a two‐step pyrolysis strategy to access Cu SACs is developed, which exhibits excellent ORR performance.

Electron microscopy characterization. Scanning electron microscopy (SEM) was carried out using a JEOL JSM-6390A instrument. Transmission electron microscopy (TEM), highresolution TEM (HRTEM) images, the corresponding energy-dispersive X-ray spectroscopy (EDS) and selected area electron diffraction (SAED) were taken on a FEI Talos F200X.
Aberration-corrected HAADF-STEM images were performed on a JEOL JEMARM200F TEM/STEM system. In-situ ETEM experiments were carried out on Thermo Scientific Themis G 3 ETEM equipped with DENS solutions Climate holder and Gas Supply System, using the primary electron energy of 300 keV. The sample were dispersed in alcohol and loaded in the MEMS-based nanoreactor, which was then mounted in a Climate holder. The environment inside was 1 bar He, heating by the MEMS heater.
XAFS measurements. The X-ray absorption fine structure spectra were collected at the Beijing Synchrotron Radiation Facility (BSRF) in China. The storage rings of BSRF were operated at 2.5 GeV with an average current of 250 mA. Using Si (111) double crystal monochromator, the data collection was carried out in transmission/fluorescence mode using ionization chamber.
All spectra were collected in ambient conditions. The acquired EXAFS data were processed according to the standard procedures using the ATHENA module implemented in the IFEFFIT software packages. The k 3 -weighted EXAFS spectra were obtained by subtracting the post-edge background from the overall absorption and the normalizing with respect to the edge-jump step. Subsequently, k 3 -weight χ(k) data of Kedge 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 parameter around central atoms, least-squares curve parameter fitting was performed using the ARTEMIS module of IFEFFIT software packages.

Electrochemical measurements:
All electrochemical measurements for ORR were conducted in a conventional standard threeelectrode system at room temperature on an electrochemical workstation (CHI 760E, CH Instrument, Shanghai, China). A rotating-disk electrode (RDE) equipment was connected to the three-electrode system to test the ORR activity. A glassy carbon (GC) disk with diameter of 5 mm was used as working electrode. Graphite rod and Ag/AgCl (3M KCl) were applied as the counter and reference electrodes, respectively. In this work, catalyst inks were prepared as follows: Cu SAs/NC and Pt/C were prepared by ultrasonically dispersing 5 mg of catalyst in 1 mL solution (containing 490 μL of ethanol, 490 μL of water, and 20 μL of 5 wt% Nafion). And, the working electrode was prepared by coating 5 μL catalyst ink on glassy carbon RDE for 4 times. Before the ORR tests, the electrolyte (0.1 M KOH solution) was filled with O2 flow for 30 min to achieve the O2 saturated solution. All potential values were normalized to the reversible hydrogen electrode (RHE) according to the Nernst equation: The RDE tests were carried out at a sweep rate of 10 mV/s with different rotating speeds ranging from 900 to 2500 rpm. The electron transfer number was determined by Koutecky-Levich equation: Where J is the measured current density, JL and JK are diffusion-limiting and kinetic densities, respectively, ω indicates the angular velocity of the disk, n denotes the transferred electron number, F is the Faraday constant (85484 C mol -1 ), C0 is the bulk concentration of O2 (1.2 × 10 -6 mol cm -3 ), D0 is the diffusion coefficient of O2 (1.9 × 10 -5 cm 2 s -1 ) and V is the kinematic viscosity of the electrolyte (0.01 cm 2 s -1 ). The stability tests were performed by current vs. time (i-t) chronoamperometric response.
The rotating ring-disk electrode (RRDE) test was conducted by LSV in the same potential range at a scan rate of 10 mV/s at 1600 rpm, while the ring disk was set to 1.2 V vs RHE. The hydrogen peroxide yield (H2O2 %) and the electron transfer number (n) were calculated using the following equations: Where ID is the disk current, IR is the ring current, N is the H2O2 collection coefficient at the ring and N = 0.4.

Computational Method：
We have employed the first-principles [1,2] to perform density functional theory (DFT) calculations within the generalized gradient approximation (GGA) using the Perdew-Burke-Ernzerhof (PBE) [3] formulation. The Brillouin zone integration was performed using 3×3×1 Monkhorst-Pack k point sampling for structures. The models were set in a 8×7 supercell and the thickness of vacuum layer was set as 15 Å to get rid of the influence from the virtual interlayer interaction. We have chosen the projected augmented wave (PAW) potentials [4,5] to describe the ionic cores and take valence electrons into account using a plane wave basis set with a kinetic energy cutoff of 450 eV. Partial occupancies of the Kohn−Sham orbitals were allowed using the Gaussian smearing method and a width of 0.03 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 0.05 eV Å −1 .
Grimme's DFT-D3 methodology [6] was used to describe the dispersion interactions among all the atoms in adsorption models. The adsorption energies (Eads) were calculated as Eads= Ead/sub -Ead -Esub, where Ead/sub, Ead, and Esub are the total energies of the optimized adsorbate/substrate system, the adsorbate in the structure, and the clean substrate, respectively.
What's more, M ions energies had been evaluated using the climbing nudged elastic band (CI-NEB) methods.

Statistical Analysis：
The data with a symbol of ± presented as mean ± standard deviation calculated on a minimum of three independent samples. Table   Figure                         Cu-O bond situated at 529.6 eV [7] was not observed in the O 1s XPS spectrum of Cu SACs/NC, confirming that Cu single atoms are mainly coordinated with nitrogen.    a) coordination number; b) bond distance; c) Debye-waller factors; d) the inner potential correction; e) goodness of fit. Ѕ0 2 was set as 0.84 for Cu-N, which was obtained from the experimental EXAFS fit of reference CuPc by fixing CN as the known crystallographic value and was fixed to all the samples.