DMREF/Collaborative Research: Active Learning-Based Material Discovery for 3D Printed Solids with Locally-Tunable Electrical and Mechanical Properties

设计材料革新和设计我们的未来(DMREF)项目将建立一种多学科的主动学习方法，以发现用于添加剂制造(AM；或3D打印)的新材料化学，其中电气和机械性能可以在本地控制。这是一种集成了先进的合成、表征、模拟和数据科学协议的多学科方法。近年来，AM呈指数级增长。一个新兴领域是将AM用于功能设备。然而，由于缺乏多材料能力，目前的AM技术在制造这些器件方面面临着挑战。数字光处理(DLP)3D打印技术以其速度快、分辨率高等优点成为近年来发展迅速的AM技术。这项研究将开发先进的数据驱动方法，利用实验和计算训练数据来处理多个设计目标，并指导连续几轮的实验。这些方法将被用来发现用于DLP 3D打印的新树脂配方，其中本地材料的性能可以从软到硬，从导电到非导电。这种主动学习方法将极大地加快AM新材料的开发这项研究将产生重大的社会和经济影响，有助于保持和提高美国在AM的领导地位这项研究将传播给本科生、研究生和高中生，并让他们参与研究。这项研究将包括来自代表性不足群体的学生，并促进STEM领域的分歧、公平和包容性。由于存在大量的成分单体，印刷聚合物之间的性能差异很大，以及缺乏一种有效的方法来以适当的比例快速选择这些单体以制备具有满足应用需要的性能的聚合物，阻碍了AM新聚合物的发现。初步工作已经证明了使用主动学习方法来发现具有不同单体组成的新树脂以获得目标机械性能的可行性。这项工作表明，使用共聚墨水设计可以打印出具有局部可变机械性能和导电性的整体部件。这项研究将首先通过实验和模拟建立一套高通量的聚合物性能评估方法。这些措施包括快速合成由不同比例的不同单体组成的聚合物，表征其机械和电学性能，并使用基于密度泛函理论计算数据训练的基于经典和机器学习的力场的分子动力学模拟来预测这些性能。在这些高通量方法的帮助下，这项研究将建立一个先进的多任务主动学习模型，该模型使用来自分子动力学模拟和有限数量的实验的数据来预测聚合物的电学和力学性质。实验和模拟相结合的迭代方法将深入了解聚合物组成和结构对聚合物导电性、力学性能和性能梯度的影响。最后，主动学习模型将被用于指导选择单体来设计和制造3D功能设备。该项目由工程局(ENG)的土木、机械和制造创新部(CMMI)和数学和物理科学局(MPS)的材料研究部(DMR)支持。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。

This Designing Materials to Revolutionize and Engineer our Future (DMREF) project will establish a multi-disciplinary active learning method to discover new materials chemistries for additive manufacturing (AM; or 3D printing) wherein the electrical and mechanical properties can be locally controlled. This is a multi-disciplinary approach with an integration of advanced synthesis, characterization, simulation, and data science protocols. AM has seen exponential growth in recent years. One emerging area is to use AM for functional devices. However, the current AM techniques face challenges in fabricating these devices due to the lack of multi-material capability. Digital light processing (DLP) 3D printing is a rapidly developing AM technique due to its advantage of high speed and high resolution. The research will develop advanced data-driven approaches which utilize both experimental and computational training data to address the multiple design objectives and guide successive rounds of experiments. These approaches will be used to discover new resin formulations for DLP 3D printing where the local material properties can be controlled from soft to stiff and conductive to non-conductive. The active learning method will greatly expedite the development of new materials for AM. The research will have significant societal and economic impacts serving to maintain and enhance the US leadership position in AM. The research will be disseminated to undergraduate, graduate, and high school students and involve them in research. The research will involve students from underrepresented groups and promoting divergence, equity, and inclusion in the STEM field. Discovery of new polymers for AM has been hindered by the existence of a large number of ingredient monomers, large property differences among printed polymers, and the lack of an efficient approach to rapidly select these monomers at proper ratios to make polymers with properties that can meet the application needs. Preliminary work has demonstrated the feasibility to use an active learning approach to discover new resins with different monomer compositions for targeted mechanical properties. This work has shown that it is possible to use the copolymer ink design to print a monolithic part with locally variable mechanical properties and conductivity. The research will first establish a suite of high-throughput methods for polymer property evaluation by experiments and simulations. These include rapidly synthesizing polymers composed of different monomers at different ratios, characterizing their mechanical and electrical properties, and predicting these properties using molecular dynamics simulations with classical and machine learning based force fields trained on density functional theory calculation data. Assisted by these high-throughput methods, the research will establish an advanced multi-task active learning model that uses data from molecular dynamics simulations and from a limited number of experiments to predict polymer electrical and mechanical properties. A combined iterative approach between experiments and simulations will provide insight into effects of polymer composition and structure on polymer conductivity, mechanical properties, and property gradients. Finally, the active learning model will be used to guide the selection of monomers to design and fabricate 3D functional devices.This project is supported by the Division of Civil, Mechanical and Manufacturing Innovation (CMMI) of the Directorate for Engineering (ENG) and the Division of Materials Research (DMR) of the Directorate for Mathematical and Physical Sciences (MPS).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.