Machine learning is a groundbreaking technology that helps researchers unlock and understand the vast spectrum of data available to us. Recently, researchers at the Institute of Industrial Science, University of Tokyo, utilized machine-learning algorithms to predict the density of states in organic molecules – molecules composed primarily of carbon – an invaluable tool in predicting its properties.
Core-loss spectroscopy is a technique used to measure the electronic structure of molecules, determining the amount of energy levels an electron can hold in a given molecule. Core-loss spectroscopy combines two methods, energy loss near-edge spectroscopy (ELNES) and X-ray absorption near-edge structure (XANES) by irradiating the material with a beam of electrons or X-rays, which then scatters the material and absorbs energy to determine the overall number of energy states.
To improve the interpretation of this spectroscopy data, researchers at the Institute of Industrial Science, UTokyo, developed a machine-learning algorithm to predict the density of electronic states. A database was constructed with the densities and corresponding core-loss spectra from over 22,000 molecules, and the algorithm was optimized to predict the ground state of both occupied and unoccupied states. Excluding tiny molecules and adding specific noise to the data was found to improve predictions. Overall, this helps researchers understand the material’s properties and expedites the design of functional molecules.
The Institute of Industrial Science is a large university-attached research institute of the University of Tokyo. With its over 120 laboratories, 1,200 members, and many disciplines of engineering, the institute attempts to bridge the large gaps between academic and real-world applications.
Po-Yen Chen was the lead author of this study. He is a specialist in using machine learning to facilitate predictions regarding the densities of electronic states. Through his research, he has been able to assist other physicists and chemists in analyzing the structure of organic molecules.
Teruyasu Mizoguchi is a senior author of the study who has been researching organic molecules for many years. He acts as an advisor to Po-Yen Chen and the team and his expertise has enabled the study to be successful. His research has particular implications for the medical and pharmaceutical sciences.
