Research

Material design by machine learning

Machine learning is a means of gaining knowledge for materials development from a data set consisting of data on a large number of materials. Using data sciece algorithms based on Python and other programming languages, we aim to discover knowledge from the data sets we have accumulated so far and develop novel materials. For example, an organic molecule can be represented in two dimensions by a structural formula, and its molecular structure can be vectorized or graphed by various methods. Mathematical representations of molecules can be analyzed by machine learning (ML) and lead to the design of molecular structures with the desired physical properties.

Functionality modeling by statistical regression

The function of a material often depends on the experimental conditions, and the output varies depending on the external conditions (temperature, light intensity, wavelength, etc.) and the size and shape of the material itself. Therefore, if we try to obtain the relationship of how much output is obtained under what conditions by experiment alone, we need to conduct a huge number of experiments under different conditions. To solve this problem, it is effective to use data science to express the relationship between experimental conditions and output in a mathematical manner. So far, we have succeeded in mathematically representing the bending behavior of a crystal that bends when exposed to light (photomechanical crystal) by using regression analysis based on machine learning.(CrystEngComm, 2021)。

Mechanically responsive molecular crystals

Crystalline materials are generally thought to be hard, but mechanical crystals that can be significantly deformed by stimuli such as light and heat have been discovered and are expected to be used as soft actuator materials. In my doctoral research, I have created mechanical functions using new phase transition mechanisms such as photoisomerization, thermal phase transition, and photo-triggered phase transition (see below). Currently, I am working on the evaluation of mechanical properties of mechanical crystals, analysis of the dynamic behavior of photo-triggered phase transitions, and creation of new crystals, while also utilizing data science.