Collaborative Research: Data-Driven Microreaction Engineering by Autonomous Robotic Experimentation in Flow

现有的实验策略往往无法全面探索通过多步骤合成过程产生的新化学品和材料的反应宇宙。考虑到为某种化学产品寻找最佳反应物和反应条件的实验搜索的资源有限的性质，由此产生的临时或不知情的实验选择很可能无法发现有价值的反应过程洞察。这一合作研究项目将创建一个科学和工程知识框架，通过在多阶段人工智能(AI)战略指导下的模块化化学合成方法，加速机械反应研究和新兴材料和分子的合成工艺开发。研究团队将产生一种新的数据驱动的科学方法，以加快高性能材料和分子的设计和合成，将开发时间从几年减少到几个月。潜在的应用包括能源和化学技术，对国家的繁荣、健康和安全产生明显的好处。这个跨学科的研究项目涉及反应工程、材料科学和人工智能等多个领域的整合。该项目将在数据驱动的微反应工程和人工智能辅助实验方面培训研究生和本科生。这一合作项目的跨学科性质将加强传统上在STEM相关研究中代表性较低的群体的学生的参与。此外，该项目的成果将通过面向本科生的动手实验模块和YouTube视频教程，免费向公众开放，并基于本研究产生的知识，对现代工程教育产生积极影响。为实现具有多阶段化学的新兴溶液加工材料和分子的数据驱动反应工程概念，需要在人工智能引导的反应空间探索、代理建模和模块化实验方面取得基础进展。本项目旨在通过闭环模块化实验，为数据驱动的微反应工程发展模块化人工智能建模和决策策略的科学基础和理解。这将使得在具有多阶段化学的新兴溶液加工材料和分子的多变量化学合成空间中进行时间和资源高效的导航。模块化的人工智能建模工作将产生新的算法，其中包括特定于问题的结构和决策模式，使自主实验能够超越概念验证演示。具体地说，胶体量子点(QD)的数据驱动微反应工程将成为目标，这一选择的驱动因素是量子点耐人寻味的尺寸和组成可调的光学和光电特性，以及多阶段和工艺敏感的合成。这一合作项目的成果将推进最先进的人工智能指导的化学合成，同时降低人工智能技术的使用门槛，使其在其他科学领域得到广泛应用。此外，多段流反应器系统的模块化代理模型可用于评估、检验和验证纳米晶成核和生长的动力学和力学模型。自主和模块化的流程合成策略将产生一个可转移的计算框架，可以应用于化学科学和工程中的其他问题，包括捕捉多阶段、多目标过程优化的模型，这是整个实验科学中普遍存在的问题。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。

Existing experimental strategies often fail to comprehensively explore the reaction universe of new chemicals and materials created with multi-step synthesis procedures. Given the resource-limited nature of experimental searches to find the best reactants and reaction conditions for a certain chemical product, the resulting ad-hoc or uninformed selection of experiments will likely fail to uncover valuable reaction process insights. This collaborative research project will create a science and engineering knowledge framework for accelerated mechanistic reaction studies and synthesis process development of emerging materials and molecules with multi-stage chemistries through a modular approach to chemical synthesis guided by a multi-stage artificial intelligence (AI) strategy. The research team will produce a new data-driven scientific approach to accelerate design and synthesis of high-performing materials and molecules, reducing development time from years to months. Potential applications include energy and chemical technologies, resulting in clear benefits to the nation's prosperity, health, and security. This interdisciplinary research project involves integration of multiple fields including reaction engineering, materials science, and AI. This project will train graduate and undergraduate students in data-driven microreaction engineering and AI-assisted experimentation. The interdisciplinary nature of this collaborative project will enhance participation of students from groups traditionally underrepresented in STEM-related research. Furthermore, the results of this project will positively impact modern engineering education through hands-on lab modules for undergraduate students and tutorial YouTube videos, free to the public and based on the knowledge generated by this research.Implementation of data-driven reaction engineering concepts for emerging solution-processed materials and molecules with multi-stage chemistries require fundamental advancements of AI-guided reaction space exploration, surrogate modeling, and modular experimentation. This project seeks to develop the science base and understanding of modular AI modeling and decision-making strategies for data-driven microreaction engineering through closed-loop modular experimentation. This will enable time- and resource-efficient navigation through the multivariate chemical synthesis space of emerging solution-processed materials and molecules with multi-stage chemistries. The modular AI modeling effort will result in new algorithms that incorporate problem-specific structure and decision-making modalities, enabling autonomous experimentation to move past proof-of-concept demonstrations. Specifically, data-driven microreaction engineering of colloidal quantum dots (QDs) will be targeted, a choice driven by the intriguing size- and composition-tunable optical and optoelectronic properties of QDs as well as multi-stage and process-sensitive synthesis. The results of this collaborative project will advance the state-of-the-art AI-guided chemical synthesis, while lowering the barrier to the use of AI techniques, enabling their broad application among other scientific domains. Furthermore, the modular surrogate modeling of the multi-stage flow reactor systems can be used for evaluation, testing, and validation of kinetics and mechanistic models of nanocrystal nucleation and growth. The autonomous and modular flow synthesis strategy will result in a transferable computational framework that can be applied to other problems in chemical science and engineering, including the models that capture multi-stage, multi-objective process optimization, a problem ubiquitous throughout experimental sciences.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.