2023.09.16 by Tejkiran Pindi JayakumarIn data-driven materials research, the presence of a large, consistent, and unbiased dataset is essential. We have achieved materials informatics to create such data through high-throughput experimentation. This study corresponds to our next challenge. One of the essentials of solid catalyst design is the combination of multiple components to effectively catalyze chemical reactions comprising many elementary reactions. However, predicting synergistic combinations is highly challenging. In the conversion of ethanol into butadiene, a catalysis of practical importance in the bio-refinery, we successfully achieved a 14-dimensional exploration, optimizing catalyst compositions consisting of 14 different elements, through the combination of high-throughput experimentation and machine learning. This approach allowed us to discover a number of high-performance novel compositions in a short period of time.
2023.06.30 by 筑間 弘樹 Hiroki ChikumaWe previously achieved non-empirical structure determination of catalyst nanostructures by combining a genetic algorithm and DFT calculations. However, this has been hampered by the significant increase in both computational cost and parametric space along with the system size. Here, we utilized previously accumulated DFT calculation results to construct a high-dimensional neural network potential capable of reproducing the DFT results rapidly and accurately. By applying the built potential, we successfully determined the primary particle structure of a Ziegler-Natta catalyst with its size and coverage comparable to those of real catalysts. The obtained structure exhibited structural features consistent with experimental observations.
2022.08.02 by 高棹 玄徳 Gentoku TakasaoA combination of a genetic algorithm and local geometry optimization enables non-empirical structure determination for complex materials such as solid catalysts. However, premature convergence in the genetic algorithm hinders the structure determination for complicated systems. Here, we implemented a distributed genetic algorithm based on migration from a structure database for avoiding the premature convergence, and thus realized the structure determination for TiCl4-capped MgCl2 nanoplates with experimentally consistent sizes. The obtained molecular models are featured with a realistic size and non-ideal surfaces, representing primary particles of Ziegler-Natta catalysts.
2022.05.17 by 瀧本 健 Ken TakimotoSynergistic combinations of stabilizers are essential for imparting durability to polymeric materials, but the low throughput of durability tests and the large number of combinations have hindered systematic research. In this study, we have established a novel protocol for high-throughput experimentation, which is based on solution film casting on microplates and absorbance measurement using a microplate reader. The developed protocol was combined with a genetic algorithm to explore stabilizer formulations for photo-induced yellowing inhibition of transparent plastics. The obtained data, amounting to seven years, were analyzed to successfully derive formulation design guidelines.
2022.03.02 by Dongzhi ZhuDue to its high breakdown voltage and low dielectric loss, biaxially oriented polypropylene (BOPP) is mainly used for thin-film capacitors in automobiles. Here, BOPP nanocomposites with dispersed titanium dioxide nanoparticles were prepared and their dielectric properties were investigated in detail. The nanocomposites exhibited permittivity far beyond the theoretical estimate, and we found out that the key is at the nature of the intermediate layer called interphase. Research on BOPP nanocomposites has been hardly reported due to the difficulty of stretching PP nanocomposites. By using the own reactor granule technology, we overcome the technical challenge, and obtained very important knowledge about thin-film capacitors.