Resonac Turns to AI for Faster Material Development

Resonac Corporation has developed in-house a cheminformatics*1 application program that utilizes artificial intelligence (AI) based on deep learning technology. It has begun using the application to maximize its vast store of physical property data. Mainly, this application significantly reduces the time required to calculate physical properties in developing new chemicals through prediction technology based on information science*2. Moreover, its simple user interface allows even engineers without expertise in cheminformatics to use it. Thus, the cheminformatics application enables the development of new materials in a shorter time than before. This is true, especially in fields such as semiconductor material development, which require rapid development.

In the conventional development of new materials, engineers often estimate the physical properties of new materials in the early stages of experimentation based on literature and experience. Then, the experiments and physical property measurements are repeated. In addition, estimating the physical properties of new materials by quantum chemical calculations*3 or chem-informatics-based simulations usually requires descriptors*4, which are physical property calculations and data input rules. These tasks require expertise and experience in quantum chemical calculations and cheminformatics.

The newly developed cheminformatics application developed by the Research Center for Computational Science and Informatics at Resonac, utilizes deep learning of calculation results and experimental data accumulated in past materials development. Thus, it is possible to estimate the physical properties of new materials several thousand times faster than quantum chemical calculations. Also, the cheminformatics application is a web application with a simple and intuitive user interface. Therefore, even experimenters unfamiliar with cheminformatics can input data using their usual molecular drawing methods. Thus, by using this cheminformatics application, the experimenter can estimate the physical properties of new materials in advance and design new materials.

Resonac will continue to utilize information science and technology to shorten the development time of new materials, especially in fields requiring speedy development, such as semiconductor materials.

*1. Cheminformatics is one of the research fields in chemistry. It estimates unknown compounds’ structure and physical properties by constructing predictive models from accumulated past data. It supports proposals of experimental conditions and to search for molecules that can be used as drugs or raw materials. Cheminformatics has some points in common with materials informatics. However, cheminformatics mainly deals with information on molecules and compounds, while materials informatics deals with information on substances and materials.

*2. Comparison between the newly developed cheminformatics application and quantum chemical calculations using conventional computers.

*3. Quantum chemical calculation is a type of molecular simulation technology. It theoretically analyzes molecules’ structure and physical properties by calculating their electronic states.

*4. The descriptor is a numerical value that expresses the characteristics of a compound’s substructure and physicochemical properties.

*5. The solubility parameter is a numerical value that indicates the solubility behavior of a solvent. It is the square root of the cohesive energy density.

*6. HOMO=Highest Occupied Molecular Orbital. LUMO=Lowest Unoccupied Molecular Orbital. The molecular orbital with the highest energy of electrons is called HOMO, and the molecular orbital with the lowest energy not occupied by electrons is called LUMO.