I am introducing the national project and reviewing the recent results. It started in 2016 FY to carry out the R&D to accelerate the creation and development of innovative functional materials collaborated with National Institute of Advanced Industrial Science and Technology (AIST) and ADMAT. There are 18 companies of chemical, materials and electronics companies. We have successfully created a new technology to largely shorten the trial periods and cycles for the R&D of functionality materials by introducing three techniques (computational science/AI, innovative process technology and advanced characterization) together.

The development of highly active catalysts for the conversion of ethanol to 1,3-butadiene using high-throughput system was described. Assuming a two-stage reactor for converting ethanol to acetaldehyde and an ethanol/acetaldehyde to 1,3-butadiene, we conducted rapid screening for the optimal catalyst for each reaction with high-throughput catalyst preparation and evaluation systems. We found that using Ag/SiO₂ and HfO₂/SiO₂ as catalysts in the first- and second-stage reactors, respectively, resulted in high 1,3-butadiene productivity(99% ethanol conversion and 63% 1,3-butadiene selectivity). The butadiene produced here was collected, purified, and polymerized to butadiene rubber. The rubber was applied to produce PCR tires with sustainable materials.

The relationship between structures and physical properties for rubber blend was investigated by materials informatics approach. SBR/IR/Silica blend rubber, which is commonly used as tire rubber, was targeted in this paper. Model structures were created on a computer, and physical properties were calculated by finite element method (FEM) simulation. Datasets with structural features as explanatory variables and calculated values as objective variables were obtained by three-dimensional structure analysis for model structures. To clarify the contribution of each structural features to physical properties, machine learning was performed using the dataset with the modulus at 50% elongation (M50) as the objective variable. The prediction accuracy of the model was as high as 0.87 in R² value. Furthermore, SHapley Additive exPlanation (SHAP) analysis revealed that the continuity of SBR phase and the shape of the phase separation interface are particularly important for physical characteristics. Finally, we created a structure-property correlation diagram that allow us to narrow down the compositions and structures that satisfied target physical properties without actual experiments. It follows that material development period will be shortened.

Soft Blends Analyzer (SOBA) was developed to link machine learning and computer simulation with Python interface. Using the SOBA, Lenard-Jones Potential particles dynamics were done to generating two dimensional images of the filler dispersion for the education data of Depp Learning (DL). The DL was performed to learn the relationship between the image features and the filler density. The technique was applied to bucky-paper (BP:Non-woven film of CNT). The tiling images of SEM images with different four magnifications were used as DL training data, and accurate classification model was obtained. The model enabled to predict the physical properties of BP not used for learning by the sum of multiplying the probability distribution of classification by the properties of the educated BP. Next, the Generative Adversarial Network (GAN) training was performed with the same SEM images. Finally, GAN's morphing method have made it possible to virtually generate images of BP mixing different CNTs and established a framework of virtual experiment method for predicting its physical properties with the DL model. By the framework, the physical properties of more than 1,700 types of bucky-paper were estimated, which has contributed to reduce the time of material development.

Liquid crystal elastomers (LCEs) are a relatively new class of materials that display soft elasticity, that is, they can be deformed without resistance. Furthermore, LCEs show a rapid and accurate response to external stimuli such as electric, magnetic, and thermal fields. For this reason, it is expected to be applied to actuators or sensors. In order to apply these characteristics to devices, we tried to predict the characteristics of LCE by simulation. First of all, we developed an extended coarse-grained LCE model to enable simulation of systems of various architectures. Our model is a hybrid of Gay-Berne particles and Lennard-Jones particles, based on previously reported LCE modeling techniques. By using molecular dynamics (MD) method, the stress-strain curves as the response to an external force were obtained, and soft elasticity was clearly observed. Then, the regression analysis using machine learning (ML) was conducted on the results of the stress-strain curves of the MD simulations. The results indicated that spacing for a room for mobility of mesogenic units in the design variables of LCE molecules affected elasticity. In addition, the R-squared value of regression curve for stress-strain curves was 0.821, which indicates a strong correlation between the MD data and ML results. Finally, the estimation method of molecular structure from coarse-grained model is discussed.

The overview of “AIST Materials Gate Data Platform” (DPF), which have been developed in the national project “Ultra-High-Throughput Design and Prototyping Technology for Ultra-Advanced Materials Development Project (U2M project)”, is introduced. The U2M project developed five DPFs for different targeted materials, and the detail of DPF for functional polymeric materials are introduced in this article. The DPF contains the data of structures and properties of polymeric materials obtained by the computational simulation and experimental observation. Especially, the higher-order structures such as phase separated structure of polymer blends and dispersion of fillers in polymer matrices, and corresponding properties such as elastic constant and thermal conductivity are focused in the DPF. In April 2022, after the completion of the U2M project, the materials design platform will be operational, and the consortium will be organized to accelerate data-driven materials design. In addition to the introduction of DPFs, we will introduce the plans for these activities to implement the fruits of the project in industry.