Electrolyzed water is an efficient and sustainable way to generate hydrogen, and is considered an effective method for the production, storage and use of renewable energy in the future
Oxygen evolution reaction (OER), as the cathodic reaction of water electrolysis, is the key reaction to improve hydrogen production by electrolysis
At present, single-atom catalysts are considered to be the most effective OER catalysts because of their 100% utilization, uniformity and selectivity.
Therefore, it is of great significance to prepare stable, high-efficiency single-atom catalysts
However, due to insufficient exploration of the formation mechanism of different metal single atoms, catalytic performance and other issues, simple and low-cost preparation of stable single-atom catalysts has become a major challenge in the current catalysis field
The research group of Professor Li Teng from the University of Maryland proposed a single-atom catalyst design scheme based on first-principles calculations and machine learning of topology
This program not only provides theoretical guidance for rational selection and preparation of transition metal element single-atom catalysts, but also evaluates the performance of single-atom catalysts 130,000 times faster than simply by DFT calculations
This ultra-efficient design scheme makes it possible to comprehensively predict and evaluate the electrocatalytic activity of all transition metal single atoms, thereby providing unprecedented rational guidance for the design of single-atom catalysts
Related work entitled "Machine learning accelerated prediction of overpotential of oxygen evolution reaction of single atom catalysts" was published online in "iScience" and was selected as the cover of the journal in May 2021
OER catalytic performance of single metal atoms.
The research team used the DFT method to calculate the OER performance of 15 metal atoms at single-vacancy and double-vacancy defects.
According to the OER reaction process in Figure 1A, the adsorption energy of the reaction intermediates, O, OH, and OOH on the metal is calculated, and the overpotential is calculated, and the catalytic performance is judged according to the magnitude of the overpotential (Figure 1B-E)
Figure 1 OER catalytic overpotentials of 15 metal atoms at single-vacancy and double-vacancy defects, respectively
Machine learning method based on topological structure
In order to overcome the problems of "long calculation period" and "huge computing resource occupation" of DFT method, Professor Li Teng's research group innovatively proposed an OER catalytic activity prediction method based on topological structure machine learning
This method is mainly based on the topological structure learning algorithm.
By analyzing the topological structure around the metal atoms, the node information and the link information between the metal atoms and the substrate are extracted and a small amount of (13 single-vacancy structure and double-vacancy structure metal atoms) DFT calculation data are combined.
Train the prediction model and predict the OER catalytic performance of other transition metals on carbon substrates with different structures (Figure 2)
Figure 2 Machine learning method based on topology
Machine learning method based on topology to predict single-atom OER catalytic performance.
this paper, the machine learning method based on topology structure not only takes structural information around metal atoms as input data, but also introduces five parameters related to OER catalytic performance (the quality of metal elements, Radius, d-orbit ebook, electronegativity, electron affinity and first ionization energy, Figure 3B) predict the model, and the prediction error accuracy is 6.
49% (Figure 3A)
The single-atom catalyst prediction method based on the machine learning method proposed in this study can greatly increase the speed of catalyst screening
Specifically, it takes about 36 hours to determine the ORR catalytic activity of each element by performing DFT calculation on a 40-core supercomputer
The method based on machine learning, including training and prediction of a total of 14 elements, takes about 40 seconds, and can be carried out on a home laptop
The screening rate of the catalyst was increased by approximately 130,000 times (Figure 3C)
Figure 3 Topological structure-based machine learning method accelerates the prediction of metal OER catalytic performance
Relationship between single-atom OER catalytic performance and element properties
In order to more effectively select single-atom catalysts, the research team proposed a volcanic curve description method and characterized the OER catalytic activity of all single-atom catalysts (Figure 4b, c)
Figure 4 Volcanic relationship between single-atom catalytic performance and element properties
Paper cited information:
Machine learning-accelerated prediction of overpotential of oxygen evolution reaction of single-atom catalysts, iScience, Vol.
24, Issue 5, (2021) 102398.
About the author:
Professor Li Teng's team at the University of Maryland (http://lit .
edu/) Focus on the design and development of high-performance sustainable materials, soft materials, low-dimensional nanomaterials, atomic scale catalysts, energy storage materials, etc.
The relevant research results are published in Nature, Science, Nature Review Materials, Nature Nanotechnology , Nature Sustainability, PNAS, PRL, JACS, Advanced Materials, Materials Today, Advanced Energy Materials and other top international journals.
In 2018, it won the R&D100 award known as the "International Invention and Creation Oscar" and the University of Maryland Invention of the Year Award in 2019 (Physical science field)
Professor Li Teng is currently the director of the Advanced Sustainable Materials and Technology Laboratory at the University of Maryland and the associate editor of Extreme Mechanics Letters.
He was awarded the Young Scientist Medal of the International Society of Engineering Sciences (2016)
Professor Li Teng and Professor Suo Zhigang of Harvard University jointly initiated the creation of iMechanica.
org in 2006, and it has become the network resource platform with the most users in the field of international mechanics
Professor Li Teng’s WeChat video account was launched on July 5, 2020 (WeChat video account: Professor Li Teng), which aims to share scientific research experience and help young scholars take off in scientific research
The short videos released every day received widespread attention and were interviewed by the media, and quickly became the head of the WeChat video account for scientific research.
Professor Li Teng will broadcast live on the WeChat video account on Wednesday and Saturday at 9 pm Beijing time
At present, it has provided dry content to more than 120,000 scientific researchers through WeChat video account live broadcast
He launched the [Scientific Paper Expert Clinic] series of live broadcasts on the video account, the first live broadcast of revised papers
In the first month of the program’s opening, the total number of live views in the live broadcast room exceeded 42,000, the highest single live view was 12,000, and the total number of likes exceeded 150,000.
It quickly became a hot live broadcast for researchers
Add Professor Li Teng's WeChat to receive the latest live broadcast information of [Scientific Takeoff Expert Clinic] as soon as possible