JAIST NOW No.18 Laboratory visit


Catalyst Informatics Project (CREST)

Laboratory introduction (movie)


  • 谷池研究室について
  • 進学希望
  • フォトアルバム
  • お問い合わせ
  • ニュース
  • 最新研究報告

2021.11.01 by 和田 透 Toru WadaThe analysis of Ziegler-Natta (ZN) catalysts is difficult due to its complex and heterogeneous nature. We have attempted to clarify the origin of the active structure of ZN catalysts by applying multifaceted characterization techniques, which track the evolution of the starting material to the final catalyst. The linkage between the polymerization performance and the formation of the active structure, like the nuc leation and growth of the δ-MgCl2 support, the types and abundance of chemical species present on the surfaces, and impurity formation and removal is important in rational design of catalysts with desired performance.

2021.10.28 by Hui YouMetal-organic framework (MOF)-based thin-film nanocomposite (TFN) membranes are promising for addressing the permeability-selectivity tradeoff in desalination. Key challenges have been identified at the dispersion and chemical stability of MOFs as well as at the interfacial bonding with the polymeric matrix. Here, we successfully demonstrated that polydopamine (PDA)-coating of MOF nanoparticles can address all these problems at once. Indeed, it almost doubled the permeability and improved the fouling resistance of the membrane without sacrificing the selectivity.

2021.06.21 by 中野渡 淳 Sunao NakanowatariOxidative coupling of methane (OCM) is a catalytic reaction that directly converts methane to ethylene. The main difficulty of this reaction is based on the fact that methane has a higher chemical stability than ethylene, i.e. strict tradeoff between methane conversion and selectivity to ethylene production. In the past, we acquired catalyst big data comprising of the OCM performance of 300 randomly selected catalysts under 135 reaction conditions. Here, we made the analysis of this big data from two viewpoints, i.e. suppression of ethylene oxidation under harsh conditions, and methane conversion under mild conditions. Thus extracted two kinds of heuristics were successfully combined to develop catalysts which equip both the activity and selectivity.

2021.01.27 by Thanh Nhat NguyenHeterogeneous catalyst development has typically been driven by trial and error, as combinatorial catalyst design that results in high performance is hard to predict. Accumulation of such empirical efforts is regarded as the anthropogenic bias in catalyst informatics, which unpleasantly binds machine learning to known experiences. Here, non-empirical catalyst design is attempted based on the combination of random sampling, high-throughput experimentation (HTE), and data science. 300 quaternary catalysts are randomly sampled from a materials space consisting of 36,540 catalysts, where their performance in relation to oxidative coupling of methane (OCM) is systematically evaluated by HTE. Machine learning is then applied to the resulting bias-free dataset in order to learn the underlying patterns in catalyst performance. The trained machine successfully predicts novel quaternary combinations for OCM with an accuracy of 80%.

2020.11.09 by 高棹 玄徳 Gentoku TakasaoA multisite nature i.e. a distribution in the produced polymer, is an essential feature of the Ziegler-Natta catalyst, which is the mainstream catalyst for industrial polyolefin production. However, the origin of the multisite nature is still unclear. We performed a series of machine learning-aided structure determination for TiCl4-capped MgCl2 nanoplates of different sizes and chemical compositions, and obtained a million structures in the course of the structure determination. Based on the analyses and simulation, a new hypothesis was successfully proposed for the origin of the multisite nature of the Ziegler–Natta catalyst.