Unlocking new possibilities in ionic thermoelectric materials: a machine learning perspective

作     者：Yidan Wu Dongxing Song Meng An Cheng Chi Chunyu Zhao Bing Yao Weigang Ma Xing Zhang Yidan Wu;Dongxing Song;Meng An;Cheng Chi;Chunyu Zhao;Bing Yao;Weigang Ma;Xing Zhang

作者机构：Key Laboratory for Thermal Science and Power Engineering of Ministry of Education Department of Engineering Mechanics Tsinghua University Key Laboratory of Process Heat Transfer and Energy Saving of Henan Province School of Mechanics and Safety Engineering Zhengzhou University College of Mechanical and Electrical Engineering Shaanxi University of Science and Technology Key Laboratory of Power Station Energy Transfer Conversion and System of Ministry of Education School of Energy Power and Mechanical Engineering North China Electric Power University School of Materials and Chemical Engineering Xuzhou University of Technology 

出 版 物：《National Science Review》 (国家科学评论(英文版))

年 卷 期：2025年第12卷第1期

页      面：195-206页

核心收录：

学科分类：12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 081104[工学-模式识别与智能系统] 08[工学] 080501[工学-材料物理与化学] 0805[工学-材料科学与工程（可授工学、理学学位）] 0835[工学-软件工程] 0811[工学-控制科学与工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：supported by the Tsinghua-Toyota Joint Research Fund the National Natural Science Foundation of China(52 176 078 and 52 250 273) the Tsinghua University Initiative Scientific Research Program 

主　　题：thermoelectric conversion ionic thermoelectric materials machine learning interpretable analysis 

摘      要：The high thermopower of ionic thermoelectric(i-TE) materials holds promise for miniaturized waste-heat recovery devices and thermal sensors. However, progress is hampered by laborious trial-and-error experimentations, which lack theoretical underpinning. Herein, by introducing the simplified molecular-input line-entry system, we have addressed the challenge posed by the inconsistency of i-TE material types, and present a machine learning model that evaluates the Seebeck coefficient with an R2of0.98 on the test dataset. Using this tool, we experimentally identify a waterborne polyurethane/potassium iodide ionogel with a Seebeck coefficient of 41.39 m V/K. Furthermore, interpretable analysis reveals that the number of rotatable bonds and the octanol-water partition coefficient of ions negatively affect Seebeck coefficients, which is corroborated by molecular dynamics simulations. This machine learning-assisted framework represents a pioneering effort in the i-TE field, offering significant promise for accelerating the discovery and development of high-performance i-TE materials.