Materials informatics (MI) refers to an attempt to dramatically accelerate the research and development of materials science by making full use of artificial intelligence such as machine learning. In order to realize MI, sufficient "material big data" is required in terms of both quality and scale, but the data accumulated in the form of scientific papers strongly reflects the experimental methods and interests of researchers and also has performance. It is not suitable for machine learning because it does not contain low material data.
This time, a joint research group of Japan Advanced Institute of Science and Technology, Hokkaido University, and Kumamoto University tried to use MI for catalyst development.We have developed a high-throughput catalyst evaluation device that can automatically acquire as many as 4000 catalyst data per day in the oxidation coupling reaction that synthesizes ethane and ethylene from methane using a catalyst.As a result, we succeeded in acquiring 30 catalyst big data, which exceeds the number of data accumulated in the past 12000 years by an order of magnitude, in just three days.
Furthermore, by analyzing the obtained catalyst big data by machine learning and improving the solid catalyst and reaction process based on the results, we succeeded in greatly improving the methane oxidation coupling reaction yield.
In this way, it can be said that "catalytic informatics" based on high-throughput experiments, material big data, and data science proved that 30 years of research can be carried out in a short period of less than one month of actual work.In the future, it is expected that the same methodology will accelerate research and development in various material fields and produce materials that contribute to the sustainable development of human society.
Paper information:[ACS Catalysis] High-Throughput Experimentation and Catalyst Informatics for Oxidative Coupling of Methane