Automated prediction system of Hamaker constants based on molecular theory and simulations associated with data science

2023 Fiscal Year Final Research Report

• Project/Area Number: 19K05029
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 26030:Composite materials and interfaces-related
• Research Institution: Japan Advanced Institute of Science and Technology
• Principal Investigator: Hongo Kenta 北陸先端科学技術大学院大学, 情報社会基盤研究センター, 准教授 (60405040)
• Project Period (FY): 2019-04-01 – 2024-03-31
• Keywords: 濡れ性 / ハマカー定数 / リフシッツ理論 / 密度汎関数法 / 機械学習 / 分子記述子

Outline of Final Research Achievements

We developed a machine learning model to achieve high-throughput estimation of the Hamaker constant. Since "Direct" regression models taking various molecular fingerprints as explanatory variables and the Hamaker constant as the target variable are black boxes, it is difficult to establish how to improve their performance. Instead of relying on the direct approach, we focus on four spectroscopic parameters constituting the Hamaker constantbased on the Lifshitz theory. These parameters can be evaluated by using a combination of first-principles simulations and several molecular theories.We constructed four machine learning models for these parameters and then integrated them.We have found that the resultant machine learning gives a better prediction than the direct model.

Free Research Field

材料工学

Academic Significance and Societal Importance of the Research Achievements

近年、第一原理計算を活用した「マテリアルズ・インフォマティクス(MI)」研究が急速に進展している。MI研究の主流は機械学習に基づく物性予測・探索だが、対象物性は電子物性とフォノン物性に限定される。本研究は分子理論と第一原理計算を活用して、液体プロセス設計に重要なハマカー定数の算定にMIを導入する初の試みである。従来のMI研究を超え、材料プロセス設計分野にMI研究を拡げるという波及をもたらす。