Asymmetric Coordination Regulating D-Orbital Spin-Electron Filling in Single-Atom Iron Catalyst for Efficient Oxygen Reduction.

The single-atom Fe-N-C catalyst has shown great promise for the oxygen reduction reaction (ORR), yet the intrinsic activity is not satisfactory. There is a pressing need to gain a deeper understanding of the charge configuration of the Fe-N-C catalyst and to develop rational modulation strategies. Herein, we have prepared a single-atom Fe catalyst with the co-coordination of N and O (denoted as Fe-N/O-C) and adjacent defect, proposing a strategy to optimize the d-orbital spin-electron filling of Fe sites by fine-tuning the first coordination shell. The Fe-N/O-C exhibits significantly better ORR activity compared to its Fe-N-C counterpart and commercial Pt/C, with a much more positive half-wave potential (0.927 V) and higher kinetic current density. Moreover, using the Fe-N/O-C catalyst, the Zn-air battery and proton exchange membrane fuel cell achieve peak power densities of up to 490 and 1179 mW cm-2, respectively. Theoretical studies and in situ electrochemical Raman spectroscopy reveal that Fe-N/O-C undergoes charge redistribution and negative shifting of the d-band center compared to Fe-N-C, thus optimizing the adsorption free energy of ORR intermediates. This work demonstrates the feasibility of introducing an asymmetric first coordination shell for single-atom catalysts and provides a new optimization direction for their practical application.

Current Status and Perspectives of Dual-Atom Catalysts Towards Sustainable Energy Utilization.

The exploration of sustainable energy utilization requires the implementation of advanced electrochemical devices for efficient energy conversion and storage, which are enabled by the usage of cost-effective, high-performance electrocatalysts. Currently, heterogeneous atomically dispersed catalysts are considered as potential candidates for a wide range of applications. Compared to conventional catalysts, atomically dispersed metal atoms in carbon-based catalysts have more unsaturated coordination sites, quantum size effect, and strong metal-support interactions, resulting in exceptional catalytic activity. Of these, dual-atomic catalysts (DACs) have attracted extensive attention due to the additional synergistic effect between two adjacent metal atoms. DACs have the advantages of full active site exposure, high selectivity, theoretical 100% atom utilization, and the ability to break the scaling relationship of adsorption free energy on active sites. In this review, we summarize recent research advancement of DACs, which includes (1) the comprehensive understanding of the synergy between atomic pairs; (2) the synthesis of DACs; (3) characterization methods, especially aberration-corrected scanning transmission electron microscopy and synchrotron spectroscopy; and (4) electrochemical energy-related applications. The last part focuses on great potential for the electrochemical catalysis of energy-related small molecules, such as oxygen reduction reaction, CO2 reduction reaction, hydrogen evolution reaction, and N2 reduction reaction. The future research challenges and opportunities are also raised in prospective section.

Non-convex Learning via Replica Exchange Stochastic Gradient MCMC.

Replica exchange Monte Carlo (reMC), also known as parallel tempering, is an important technique for accelerating the convergence of the conventional Markov Chain Monte Carlo (MCMC) algorithms. However, such a method requires the evaluation of the energy function based on the full dataset and is not scalable to big data. The naïve implementation of reMC in mini-batch settings introduces large biases, which cannot be directly extended to the stochastic gradient MCMC (SGMCMC), the standard sampling method for simulating from deep neural networks (DNNs). In this paper, we propose an adaptive replica exchange SGMCMC (reSGMCMC) to automatically correct the bias and study the corresponding properties. The analysis implies an acceleration-accuracy trade-off in the numerical discretization of a Markov jump process in a stochastic environment. Empirically, we test the algorithm through extensive experiments on various setups and obtain the state-of-the-art results on CIFAR10, CIFAR100, and SVHN in both supervised learning and semi-supervised learning tasks.