−60 °C solution synthesis of atomically dispersed cobalt electrocatalyst with superior performance

Temperature can govern morphologies, structures and properties of products from synthesis in solution. A reaction in solution at low temperature may result in different materials than at higher temperature due to thermodynamics and kinetics of nuclei formation. Here, we report a low-temperature solution synthesis of atomically dispersed cobalt in a catalyst with superior performance. By using a water/alcohol mixed solvent with low freezing point, liquid-phase reduction of a cobalt precursor with hydrazine hydrate is realized at −60 °C. A higher energy barrier and a sluggish nucleation rate are achieved to suppress nuclei formation; thus atomically dispersed cobalt is successfully obtained in a catalyst for oxygen reduction with electrochemical performance superior to that of a Pt/C catalyst. Furthermore, the atomically dispersed cobalt catalyst is applied in a microbial fuel cell to obtain a high maximum power density (2550 ± 60 mW m−2) and no current drop upon operation for 820 h.

Supplementary Note 1

Thermodynamic calculation of solution reduction reaction
The fundamental criterion of reaction spontaneity (thermodynamic feasibility) ∆G can be described as follows: ∆G = ∆H -T∆S (1) where S is the entropy, H is the enthalpy, and T is the solution temperature. Solution reactions occur spontaneously if ∆G is still negative at a relatively low temperature.

Temperature-dependent reduction rate
Recently, many groups have devoted extensive research efforts to search for a quantitative knob that can be adjusted to precisely manipulate the nucleation and growth of nanocrystals in a predictable manner. Some results have clearly demonstrated that the reduction rate of salt precursor not only plays an essential role in determining the outcome of a synthesis, but also serves as a quantitative knob for controlling the products of synthesis. In most concerned cases by supplying the reducing agent in great excess relative to the salt precursor, the reduction rate can be simplified to that of a pseudo-first-order reaction as follows: where k is the rate constant, [M n+ ] corresponds to the concentrations of the salt precursor, A is the frequency factor, R is the universal gas constant, and T is the reaction temperature 1-5. Obviously, the reduction rate decreases dramatically with the decrease in the reaction temperature, provided all other conditions remain unchanged. This indicates that the reaction temperature can be regarded as a key parameter to regulate and control the solution reduction process of metal precursors.

Temperature-dependent nucleation rate
A rate of nucleation of N particles during time t can be described by using an Arrhenius type equation (7) as follows: where A is a pre-exponential factor, γ is the increase in the free energy per unit surface area of the nucleus, ν is the molar volume of the nucleus, kB is the Boltzmann constant, T is the temperature,  The raw data analysis was performed using IFEFFIT software package according to the standard data analysis procedures 8 . The spectra were calibrated, averaged, pre-edge background subtracted, and post-edge normalized using Athena program in IFEFFIT software package. The Fourier transformation of the k 3 -weighted EXAFS oscillations, k 3 · χ(k), from k space to R space was performed over a range of 3.0-11.5 Å -1 (3.0-14.2 for Co foil) to obtain a radial distribution function.
And data fitting was done by Artemis program in IFEFFIT.

Air-cathode material and its fabrication
Air cathodes contain three layers including a catalyst layer, a current collector, and a diffusion layer. Carbon black powder, PTFE dispersion, and ethanol were mixed to obtain a diffusion layer onto a stainless steel mesh (60 × 60) by a press process according to previous study 9 . For Co/NMC-LT900, Co/NMC-RT900, and NMC-900, 60 mg samples were used as catalysts for air cathodes (11 cm 2 total area and 7 cm 2 projected area), leading to a final catalyst loading of 5.45 mg cm -2 . The catalyst layer was fabricated by coating the mixture of catalyst (60 mg), deionized water (388 μL), and PTFE dispersion (70 μL) onto the other side of stainless steel mesh in diffusion layer. Then another stainless steel mesh (60 × 60) facing the catalyst layer was pressed together at 10 MPa for 10 min and dried at 80 o C to assemble the final cathodes prior to use. Pt-based aircathode was prepared by following the same procedure using 10% Pt on Vulcan XC-72 powder as a benchmark.

Electrochemical measurements
The electrochemical performance of air-cathodes was evaluated using a potentiostat (PGSTAT where J is the measured current density, JK and JL are the kinetic and limiting current densities, respectively, ω is the rotation speed (rpm) of the disk, n is the electron transfer number, F is the

Computational Method
Calculations were performed by using the density functional theory (DFT) with Vienna ab initio package (VASP) 12,13 . The general gradient approximation of Perdew-Burke-Ernzerhof (GGA-PBE) functional was used to describe the exchange-correlation interactions between electrons 14,15 .
The energy cutoff was set at 500 eV for plane wave functions. In this study, the vacuum layers were set at 12 Å, and the van der Waals (vdW) interaction was calculated by the DFT-D3 method.
The reciprocal space was sampled using a 2 × 2 × 1 point grid by using Monkhorst-Pack K-points scheme 16,17 . The structures were relaxed until the residual force on each atom was less than 0.01 eV Å -1 . In the calculations, a 10 × 10 × 1 graphene supercell was used. The pyridinic-N and graphitic-N are two typical NMC in the experiment; therefore, they were mainly considered in this study.
The adsorption energy (Ea) was calculated by using the following equation: E a = E substrate+adsorbate -E substrate -E adsorbate (11) where Esubstrate+adsorbate is the total energy of adsorbate on substrate, Esubstrate is the total energy of substrate, and Eadsorbate is the energy of the adsorbate. As for the Co adsorption on graphene, the total energy of single Co atom was used for the adsorbate during calculating the adsorption energy.
Based on this definition, the smaller adsorption energy indicates the more stable adsorption on substrates. The ORR in alkaline medium is a complete 4eprocess, where the changes of free energy for each step were calculated to reflect the ORR activity. For every one electron transfer step, the free energy change (∆G) can be expressed as follows: The ∆E, ∆ZPE, T∆S, e, and U represent the energy changes, zero-point energy correction, entropic energy, the elementary charge, and the potential used during the ORR, respectively. The ∆E and ∆ZPE values were obtained from DFT calculations and the T∆S was obtained from the standard thermodynamic data. The ORR process in this study includes the following four steps: * + O2 + H + + e -→ *OOH, *OOH + H + + e -→ *O + H2O, *O + H + + e -→ *OH, and *OH + H + + e -→ H2O. Here, the * and *OOH (*O and *OH) represent an adsorbed site of substrate and adsorbed OOH (O and OH), respectively. As shown in the main context, the OH desorption is the key step.
In our calculations, we considered the ORR processes with one and two OH adsorption on the Co.
The results show that the two OH adsorption on the Co exhibits the smaller energy barrier for ORR, as reported in the main context.
In order to investigate the stability of atomically dispersed Co in solution, the first-principle molecular dynamics (FPMD) simulation was carried with NVT ensemble along with a Nose-Hoover thermostat using CP2K/QUICKSTEP package 17 , at the target temperatures of -60 o C. The