News

An article written by employees of Mitsubishi Materials Corporation on technology for predicting the properties of materials utilizing a machine learning model usable in the development of new materials has been published in the British general scientific journal "Scientific Reports."
The machine learning model developed in this article is a technology that can build a model that predicts the complex stability constants (*3) of metal complexes (*2) based on the data in the database (*1) of the U.S. National Institute of Standards and Technology (NIST), that can also be used in the development of new materials.

(*1) Data of 19,810 metal complexes for 57 ions
(*2) Substance consisting of one or more molecules surrounding and bonded to a metal ion
(*3) Important property index of metal complexes used in surface treatment technology such as plating, separation and refinement technology for extracting pure substances, molecular design in the pharmaceutical and other fields, and analytical chemistry, etc.

In the field of materials development, in order to quickly respond to diversifying demands for various materials, the use of materials informatics (MI) utilizing data science-based machine learning and artificial intelligence in place of traditional methods that rely on trial and error is expected to lead to shorter development periods and the discovery and development of new materials, and its utilization is advancing.
We believe that the results of this study successfully shows a methodology for predicting properties based on MI while extracting features (parameters) that are easily understood as material properties.

Mitsubishi Materials will continue to contribute to the building of a prosperous society through the development of new materials and the development and provision of high-value-added products.

Source: Scientific Reports

< Journal that publishes the article >

The results of this research were submitted in April 2022 to the international journal "Scientific Reports" (online), a general scientific journal of the Springer Nature Group, and were officially accepted in June.

Title of journal

Scientific Reports (online)

Title of article

Machine learning-based analysis of overall stability constants of metal-ligand complexes

Authors

Kaito Kanahashi, Makoto Urushihara, and Kenji Yamaguchi

DOI Number

10.1038/s41598-022-15300-9

URL

https://www.nature.com/articles/s41598-022-15300-9