| 高分子 Vol.71 No.8 |
| >> Japanese | >> English |
特集 ポリマーAI（永愛）
|
| 展望 COVER STORY: Highlight Reviews |
| データ時代の材料研究 Materials Research in Data Era |
出村 雅彦 Masahiko DEMURA |
| ＜要旨＞ 材料は網羅的なビッグデータ構築が困難であり、実験データの再利用を進めつつ、少量のデータを有効に活用していく手法開発が必要となる。加えて、計算シミュレーションによるデータ創出を活用していくことも重要となる。NIMSにおけるデータ駆動研究の取り組みを紹介するとともに、基盤となるデータプラットフォームの今後について俯瞰する。 Keywords: Data-Driven Materials Research / Data Platform |
| ページトップへ▲ |
| データ駆動型高分子材料研究におけるデータ資源の不足：シミュレーションと機械学習の役割 Lack of Data Resources in Data-Driven Polymer Materials Research: The Role of Simulation and Machine Learning |
吉田 亮・林 慶浩 Ryo YOSHIDA, Yoshihiro HAYASHI |
| ＜要旨＞ データ駆動型材料研究における最も大きな壁は、体系的かつ包括的なデータの不足である。材料研究では、データの不足を補う手段の一つとして、機械学習とシミュレーションの融合が重要な役割を担う。本稿は、高分子材料研究の現状を俯瞰した上で、データ駆動型材料研究における機械学習とシミュレーションの活用方法と将来展望を論じる。 Keywords: Polymer Informatics / Machine Learning / Molecular Dynamic Simulation / Open Data |
| ページトップへ▲ |
| 自動設計技術の発展と高分子への応用 Automatic Materials Design and its Application to Polymers |
津田 宏治 Koji TSUDA |
| ＜要旨＞ 機械学習に基づく材料の自動設計技術は、近年、さまざまな分野に応用されている。本稿では、それらの技術がどのように高分子科学に適用されているかについて述べる。それに加えて、好奇心サンプリングによる材料設計法や、人間とAIの連携を進める方法についても述べる。 Keywords: Automatic Materials Design / Black Box Optimization / Curiosity Sampling |
| ページトップへ▲ |
| トピックス COVER STORY: Topics and Products |
| マテリアルズ・インフォマティクスによるポリプロピレン複合材料の組成設計 Materials Informatics Approach to Predictive Models for Elastic Modulus of Polypropylene Composites |
押山 智寛・奥山 倫弘 Tomohiro OSHIYAMA, Michihiro OKUYAMA |
| ＜要旨＞ We applied materials informatics (MI) to existing experimental data for the elastic modulus of polypropylene (PP) composites. Explanatory variables were described by a combination of 0 and 1 representing polypropylene, or by the content ratio of filler and additive, without using materials property data. We constructed a predictive model for the elastic modulus of polypropylene composites using a partial least square regression (PLS) model with dummy variables. The validity of the predicted model was evaluated by comparing observed and predicted elastic moduli for new polypropylene composites. The predictive model is useful for identifying suitable combinations of polypropylene, filler and additive to achieve a desired elastic modulus. Keywords: Materials Informatics / Elastic Modulus / Polypropylene Composite / PLS / Dummy Variables |
| ページトップへ▲ |
| データ駆動的アプローチによる新材料探索 Data-Driven Approach for New Material Exploration |
永田 賢二 Kenji NAGATA |
| ＜要旨＞ This paper describes a framework for a data-driven approach to new materials exploration. As an example, we present a previous study of catalyst exploration for better catalytic reaction efficiency through sparse modeling. Keywords: Data-Driven Approach / Sparse Modeling / Catalyst Exploration |
| ページトップへ▲ |
| 人工知能によるガラスのモデリング Machine Learning for Modeling Glass Materials |
浦田 新吾 Shingo URATA |
| ＜要旨＞ This report introduces several recent applications of machine learning and the related technologies, such as graph theory, for modeling and analyzing oxide glasses. The first example is force-field development by fitting the energy and force evaluated by density functional theory (DFT) to a functional force-field. The Bayesian optimization helps us to find the optimal parameters, automatically, for developing the force-matching potential (FMP). The FMP successfully reproduced the density anomaly of silica glass, and the origin of the anomaly was analyzed using D-measure. The graph theory-based method highlighted an increase of the local structure similar to coesite crystal at high temperature, which would indicate the emergence of the denser structure when the anomaly appears. The final demonstration is application of deep-learning to obtain a more accurate force-field. The neural network potential was able to construct a boroxol ring in a borosilicate glass contrary to the conventional force-field. Keywords: Machine Learning / Graph Theory / Oxide Glass |
| ページトップへ▲ |
| AIを活用した新規MOFの合成条件探索 Machine-Learning-Assisted Synthesis of Novel MOFs |
田中 大輔 Daisuke TANAKA |
| ＜要旨＞ Metal-Organic Frameworks (MOFs) exhibit promising functionalities by utilizing the framework structures. Because MOFs can form many crystal polymorphisms, it is difficult to predict synthesis conditions to realize desired structures. The mechanism of the crystallization process of MOFs is not fully understood, and time consuming exploration has been required to optimize the synthesis conditions. Here, we focused on machine learning techniques, i.e. decision tree analysis, to improve the accuracy of the prediction for the synthesis conditions. In this work, we explored the synthesis conditions of MOFs with sulfide-metal bonds using high throughput screening systems and machine learning techniques. Keywords: Metal-Organic Frameworks / Coordination Polymers / Machine Learning / Crystal Structure / High-Throughput Synthesis |
| ページトップへ▲ |
| グローイングポリマー Polymer Science and I: A Personal Account |
| 知らず知らずに Going with the Flow |
松本 拓也 Takuya MATSUMOTO |
| ＜要旨＞ Seven years have passed since I started to work as a polymer scientist at Kobe University. I am expanding my research field from polymer syntheses to polymer structure and physics. I enjoy this change and my daily research life as a novice researcher. |
| ページトップへ▲ |
| 高分子科学最近の進歩 Front-Line Polymer Science |
| 両親媒性共重合体のアーキテクチャと会合挙動 Chain Architecture and Association Behaviors of Amphiphilic Copolymers in Water |
勝本 之晶 Yukiteru KATSUMOTO |
| ＜要旨＞ The amphiphilic polymer is an important category of the current polymer science. When a molecule has both hydrophobic and hydrophilic moieties, the molecule forms an assembly such as a micelle. For polymeric compounds, there are enormous patterns in the arrangement of hydrophobic and hydrophilic moieties (as monomers or blocks), namely, in the architecture. Various chain architectures have been proposed for amphiphilic copolymers, to control the morphology of unimer and micelles. In this report, a recent progress in the field of amphiphilic copolymer is reviewed. Brush, cyclic, tadpole, and linear multiblock architectures are discussed together with the morphology of micelles. Brush copolymers forms a unimer micelle and undergoes a self-sorting association when two copolymers with a different composition are coexisting in solution. The aggregation number of cyclic, tadpole, and linear alternating multiblock copolymers is smaller than that of the linear analogue. These results clearly show that the chain architecture affects the morphology of the polymer association in solution. Keywords: Amphiphilic Copolymers / Micellization / Association / Chain Architecture / Aqueous Solution |
| ページトップへ▲ |
| Copyright(C) 2022 The Society of Polymer Science, Japan All Rights Reserved. |