Collaborative Research: DMREF: AI-enabled Automated design of ultrastrong and ultraelastic metallic alloys

发现新合金的传统试错法已经变得越来越昂贵和耗时。设计材料革新和设计我们的未来(DMREF)项目旨在利用人工智能的力量，实现金属合金的快速和自动化设计，这些合金能够在永久塑性变形之前承受极端应力和可恢复的弹性变形。潜在的候选合金是由多种高浓度化学元素组成的复杂的浓缩合金。这些合金包含金属晶体中化学元素和原子位置的复杂波动。这些波动中的巨大自由度阻碍了对具有峰值强度和峰值弹性变形极限的合金的有效寻找。为了克服这一障碍，研究小组将使用人工智能、计算建模和实验工具来设计、合成和测试超强和超弹性金属合金。将基于人工智能技术和物理模型的集成，构建一个独特的两阶段自动化研究工作流程，从数据驱动的方法过渡到基于物理的方法。这种集成将增强对复杂材料中变形机制的理解，使其能够在结构和功能应用中使用。这个具有不同背景的研究团队将为本科生和研究生提供学习材料科学和人工智能的机会。此外，该项目致力于促进研究和教育的多样性、公平性和包容性。研究小组将通过教育和外联活动，积极让代表性不足的少数族裔学生参与研究项目。通过这项研究开发的创新战略，在人工智能的支持下，不仅将对金属合金设计产生革命性的影响，而且将对多功能材料和制造工艺的发展产生革命性的影响。研究团队致力于开发一种基于人工智能的自动化研究工作流程，以彻底改变超强和超弹性金属合金的设计和制造工艺，这两种合金同时具有极高的屈服强度和弹性极限。一般的策略是操纵和精确地定制局部晶格扭曲和化学浓度波动，以阻止复杂的浓缩合金中的变形缺陷运动。为了实现这一目标，自动化研究工作流程将在物理原理和人工智能的帮助下无缝地整合材料设计的每个步骤。具体地说，迭代设计步骤将涉及变形缺陷的原子模拟，使用自动共溅射和现场表征反馈沉积难熔金属基复杂浓缩金属合金薄膜，然后使用先进的纳米机械测量、光谱技术和尖端电子显微镜进行全面的机械和结构表征。通过利用低阶矩阵/张量分解和自动编码神经网络，将从模拟、沉积参数、机械行为、光谱和化学/结构表征结果中提取材料结构和缺陷属性的关键特征。这些关键特征促进了两阶段自动化研究工作流程的构建，该工作流程从数据驱动的方法过渡到基于物理的方法来设计和验证候选合金。该项目旨在促进对极端载荷条件下变形机理的科学理解，以及复杂浓缩合金和其他化学复杂材料的制造技术。研究团队为不同背景的学生提供了广泛的教育机会，包括材料科学、计算机科学和机械工程专业的学生。此外，该项目通过存档和共享公共材料研究的公共存储库和协作平台上的代码和数据来促进协作和创新。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。

The traditional trial-and-error approach for discovering new alloys has become increasingly expensive and time-consuming. This Designing Materials to Revolutionize and Engineer our Future (DMREF) project aims to leverage the power of artificial intelligence to enable the rapid and automated design of metallic alloys capable of withstanding both extreme stress and recoverable elastic deformation before permanent plastic deformation. The potential candidate alloys are complex concentrated alloys that are consisted of multiple high-concentration chemical elements. These alloys contain intricate fluctuations of both chemical elements and atomic positions within metallic crystals. The tremendous degrees of freedom in these fluctuations obstruct the efficient search for alloys with peak strength and peak elastic deformation limit. To overcome this barrier, the research team will employ artificial intelligence, computational modeling, and experimental tools to design, synthesize, and test ultrastrong and ultraelastic metallic alloys. A unique two-stage automated research workflow that transits from a data-driven approach to a physics-based approach will be constructed based on integrations of artificial intelligence techniques and physical models. Such integrations will enhance the understanding of deformation mechanisms in complex materials, enabling their use in structural and functional applications. This research team with diverse backgrounds will provide incorporative opportunities for undergraduate and graduate students to learn both materials science and artificial intelligence. Moreover, this project is committed to promoting diversity, equity, and inclusion in research and education. The research team will actively engage underrepresented minority students in research projects through education and outreach activities. The innovative strategies developed through this research, enabled by artificial intelligence, will have transformative impacts not only on metallic alloy design but also on the development of multifunctional materials and manufacturing processes.The research team is devoted to developing an artificial intelligence-enabled automated research workflow to revolutionize the design and manufacturing processes of ultrastrong and ultraelastic metallic alloys, which have extremely high yield strengths and elastic limits simultaneously. The general strategy is to manipulate and precisely tailor the local lattice distortions and chemical concentration fluctuations for impeding deformation defect motions in complex concentrated alloys. To achieve this goal, the automated research workflow will seamlessly integrate each step of material design aided by physical principles and artificial intelligence. Specifically, iterative design steps will involve atomistic simulations of deformation defects, depositing thin films of refractory metals-based complex concentrated metallic alloys using automated co-sputtering and in-situ characterization feedback, followed by comprehensive mechanical and structural characterizations using advanced nanomechanical measurements, spectroscopic techniques, and cutting-edge electron microscopy. By leveraging low-rank matrix/tensor factorization and autoencoder neural networks, key features of material structures and defect properties will be extracted from simulations, deposition parameters, mechanical behaviors, spectra, and chemical/structural characterization results. These key features facilitate the construction of a two-stage automated research workflow that transitions from a data-driven approach to a physics-based approach for designing and validating alloy candidates. This project aims to advance both the scientific understanding of deformation mechanisms under extreme loading conditions and manufacturing technologies of complex concentrated alloys and other chemically complex materials. The research team provides broad education opportunities for students with diverse backgrounds, including those in materials science, computer science, and mechanical engineering majors. Also, this project promotes collaboration and innovation through the archiving and sharing of codes and data on Materials Commons, a public repository and collaboration platform for materials studies.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.