Developing and Understanding Thermally Conductive Polymers by Combining Molecular Simulation, Machine Learning and Experiment

大块非晶聚合物通常是导热系数(TC)在0.1-0.5W/mK范围内的绝热材料。然而，由于其机械灵活性和低成本，它们被广泛应用于塑料换热器、电子冷却和电动汽车热管理等换热应用中。然而，由于缺乏物理指导，导热非晶态聚合物的开发一直具有极大的挑战性。该项目旨在将分子模拟、机器学习(ML)和实验相结合，以开发高TC(1.5W/MK)的非晶态聚合物，并了解决定热传输的潜在分子特征。开发的方法和从该项目中了解的物理知识将显著加快和提高开发高TC聚合物的成功率。这些新材料将有助于解决许多热传递应用中的挑战。该项目还将为来自不同背景和科学领域的学生和研究人员提供多学科教育机会。该项目的主题将为美国工业培养未来的劳动力。该项目的目标是显著加快导热聚合物的开发，并了解控制非晶态聚合物TC的潜在结构-性能关系。为了达到这一目标，本项目的目标是：(1)使用高通量分子动力学(MD)模拟建立标准化的聚合物数据库，辅之以现有数据库(例如PolyInfo)和开放文献中的数据；(2)利用独特的ML训练技术--半监督图形不平衡回归框架(SGIR)，开发准确的替代模型，绘制出聚合物化学、结构和TC的结构-性质关系，并使用我们的基于环境的图形合理化(GRIA)技术来识别影响TC的影响分子特征；(3)使用建立的模型预测~100种不同TC的聚合物，并进行详细的MD模拟来计算TC，以验证预测并了解ML确定的结构与性能之间的关系；(4)根据合作聚合物化学家对预测聚合物的可合成性的专家意见，从下选择10种聚合物，合成它们并测量它们的TC。本项目将建立的数据驱动的方法将在导热聚合物领域提供一个有影响力的例子。已建立的协议可用于设计具有其他理想性能的材料，影响超出聚合物或热传输的科学和工程领域。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Bulk amorphous polymers are usually thermal insulators with thermal conductivity (TC) in the range of 0.1-0.5 W/mK. However, they are widely used in heat transfer applications such as plastic heat exchangers, electronic cooling, and electric vehicle thermal management due to their mechanical flexibility and low cost. However, the development of thermally conductive amorphous polymers has been exceptionally challenging due to the lack of physical guidance. This project aims to combine molecular simulations, machine learning (ML), and experiments to develop amorphous polymers with high TC (1.5 W/mK) and understand the underlying molecular features dictating thermal transport. The methodology developed and the physics understood from this project will significantly speed up and increase the success rate of developing high TC polymers. These new materials will contribute to tackling the challenges in many heat transfer applications. This project will also provide multi-disciplinary education opportunities to students and researchers from diverse backgrounds and scientific fields. The topics of this project will cultivate future workforce for the U.S. industry. The goal of this project is to significantly speed up the development of thermally conductive polymer and understand the underlying structure-property relationships governing amorphous polymer TC. To reach this goal, the objectives of this project are: (1) establish a standardized polymer database using high-throughput molecular dynamics (MD) simulations, complemented by data from existing database (e.g., PolyInfo) and open literature; (2) employ a unique ML training technique, semi-supervised framework for graph imbalanced regression (SGIR), to develop accurate surrogate models that map out structure-property relations for polymer chemistry, structure and TC, and use our Graph Rationalization with Environment-based Augmentations (GREA) technique to identify the influential molecular features impacting TC; (3) use the established model to predict ~100 polymers with varying TC and perform detailed MD simulations to calculate the TC to verify the prediction and understand the ML-identified structure-property relationship; (4) down-select 10 polymers with the expert opinion from collaborating polymer chemists on the synthesizability of the predicted polymers, synthesize them and measure their TC. The data-driven approach to be established in this project will provide an impactful example in the field concerning thermally conductive polymers. The established protocol can be followed to design materials with other desirable properties, impacting science and engineering fields beyond polymer or thermal transport.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.