CORE TECHNOLOGY
STRATEGIC RESEARCH

MATERIALS INFORMATICS RESEARCH

Conventional materials development requires years and years of repeating try-and-error experiments for enormous number of combinations to find one practical result.
The Materials Informatics Research Domain introduces data analytics to materials development, working with big data learning and multiscale materials simulations to predict candidate materials with the desired performance. Through this approach, we are aiming to greatly shorten resources needed for the materials development, and facilitate the discovery of new materials.

Large-Scale Simulation Technology Using Mechanical Learning Potential

We have developed a machine learning algorithm that identifies interatomic potentials using the results of ab initio calculations as training data. This algorithm enables us to predict the catalytic activity of nanoparticle surfaces (up to 3,000 atoms) and the lithium (Li) diffusion behavior accompanying the structural phase transition of solid Li ion conductors.

Large-Scale Simulation Technology Using Mechanical Learning Potential

Machine Learning Technology for Material Microstructures

Our technology is capable of predicting conductance from electron microscope images of ion conductors using a convolutional neural network (CNN) deep learning model. By visualizing the target areas of the CNN, we identified vacancies as a cause of conductance loss.

Machine Learning Technology for Material Microstructures

SELECTED PAPERS

CORE
TECHNOLOGY

STRATEGIC
RESEARCH

FEATURE

PAGE TOP