Development of Collaborative Intelligence System and Application in Energy Materials
Development of Collaborative Intelligence System and Application in Energy Materials
批准号:
22KJ0780
负责人:
ZHANG YUCHENG
金额:
$1.09万
依托单位:
依托单位国家:
日本
项目类别:
Grant-in-Aid for JSPS Fellows
财政年份:
2023
资助国家:
日本
项目状态:
已结题
起止时间:
2023-03-08 至 2024-03-31
中文摘要
以环状透明光学聚合物(CytoP)为例,主要研究成果如下:1.建立并验证了电离势(IP)作为其充电容量、长期稳定性和热稳定性的模拟指标。利用密度泛函理论(DFT)和可极化连续介质模型(PCM)进行溶剂化校正,实现了固体激电效应的快速、准确计算。在精度相当的情况下,计算量是MD-DFT方法的16倍。表面电位衰减测量和热刺激放电(TSD)测量表明,PCM-DFT IP与驻极体的性能具有很好的一致性。采用带密度泛函计算的ChemTS方法进行从头分子生成。对官能团的统计富集性分析表明,羟基(-OH)对提高驻极体的性能具有重要意义。在从所提出的分子中提取可解释的知识之后,开发了一种新的驻极体CTX-A/APDEA。通过实验验证,其性能优于所有商品化的驻极体。在对开源数据库(如PubChem)中的胺进行采样后,建立了一个全新的高质量量子化学数据集。在数据集上训练包括二体相互作用和三体相互作用在内的深度学习模型,并将其用作量子化学性质预测器,用于高通量筛选未知化学空间。
英文摘要
Taking CYTOP (Cyclic Transparent Optical Polymer) as an example of the energy material, the achievements are summarized as follows:1. Ionization potential (IP) is found and validated as the simulation index of its performance, including the charge capacity, long-term stability and thermal stability. Rapid and accurate evaluation of solid-state IP is developed by employing density function theory (DFT) with ‘solvation’ correction by polarizable continuum model (PCM). The computational cost is 16 times smaller than the MD-DFT method with comparable accuracy. Surface potential decay measurement and thermally stimulated discharge (TSD) measurement are conducted, while the PCM-DFT IP shows excellent agreement with the performance of the electrets.2. ChemTS with DFT evaluation is employed for de novo molecule generation. Statistical enrichment analysis of the functional groups shows that hydroxyl group (-OH) is important for improving the performance of the electrets. After extracting the interpretable knowledge from the proposed molecules, a novel electret named CTX-A/APDEA is developed. According to experimental validation, its performance is superior to all commercialized electrets.3. After sampling amines from open-source database (e.g., PubChem), a brand-new high-quality quantum chemical dataset is established. Deep learning models including two-body interaction and three-body interaction are trained on the dataset, and thereafter utilized as quantum chemical property predictors for high-throughput screening of the unknown chemical space.
期刊论文(0)
专著(0)
科研奖励(0)
会议论文
Deep Generative Models for Proposing Novel Amine Molecules in High-Performance Polymer Electret Design
在高性能聚合物驻极体设计中提出新型胺分子的深度生成模型
DOI:
--
发表时间:
2022
期刊:
影响因子:
--
作者:
[Qu Yuchen, Mueller-Cajar Oliver, Yamori Wataru, Eiji YASUHARA, 三條竜平・須貝俊彦, Yucheng Zhang]
通讯作者:
Yucheng Zhang
Property Prediction of Polymer Electret Material with Physics- informed Neural Network
基于物理的神经网络对聚合物驻极体材料的性能进行预测
DOI:
--
发表时间:
2022
期刊:
影响因子:
--
作者:
[Qu Yuchen, Mueller-Cajar Oliver, Yamori Wataru, Eiji YASUHARA, 三條竜平・須貝俊彦, Yucheng Zhang, 三條竜平・須貝俊彦, Yucheng Zhang]
通讯作者:
Yucheng Zhang