基于微流控液滴技术的自动化高通量材料筛选系统的研制

Due to the problems of long cycle and high cost, the traditional model of material research and development, "trial and error method", is difficult to meet the urgent demand for new materials, and has become the bottleneck of new material research and development. Previous studies of the applicant have found that microfluidic droplet technology can achieve high-throughput preparation of materials. Therefore, this project aims to study high-throughput screening of functional metal-organic frameworks (MOFs), and develop an automatic and high-throughput screening system that integrated with high-throughput preparation, rapid characterization and artificial intelligence. Firstly, through numerical and experimental analysis, the microfluidic chip with chaotic mixed channel is designed and fabricated. Then, in combination with the programmable mobile platform, a liquid phase synthesis system with continuous gradient changes, such as solution concentration, pH value and temperature, is developed. The collaborative optimization between the array arrangement of droplet synthesis and the fast characterization are further studied. Then, based on convolution neural network and migration learning, a deep learning model is established to automatically identify characterization datum of MOFs and obtain the MOFs with new structures and their preparation conditions. Finally, the effectiveness of the high-throughput screening system is verified by magnifying experiments, and the high efficiency and low-cost preparation of new materials are realized. The successful implementation of this project is helpful to speed up the research and development of new functional materials, which high efficiency experimental technology for materials genome initiative.

传统的材料研发模式-“试错法”，受制于周期长和成本高等问题，难以满足目前对新材料的急迫需求，已成为新材料研发的瓶颈。申请人前期研究发现，微流控液滴技术可实现材料的高通量制备。在此研究基础上，本项目以筛选功能MOFs材料为代表，研制一个集成高通量制备、快速表征和人工智能的自动化、高通量功能材料筛选系统。首先，通过仿真和实验分析，设计并加工具有混沌混合通道的微流控芯片；然后，结合可编程移动平台，开发溶液浓度、pH值和温度等梯度连续变化的材料液相合成系统，进一步研究二维空间液滴合成阵列排布和快速表征系统之间的协同优化控制；随后，基于卷积神经网络和迁移学习建立深度学习模型，自动筛选材料表征数据图片集，获得新型MOFs材料及其制备条件；最后，通过放大实验，验证高通量筛选系统的有效性，实现新材料的高效低成本制备。本项目的成功实施，有助于加快新材料的研发，为国家材料基因组计划提供先进的高效实验技术。

本项目针对传统材料研发模式存在周期长和成本高等问题，以自动化高通量筛选材料系统为主要研究对象，选择功能MOFs材料为代表性产品，微流控合成液滴阵列和深度学习为关键技术，研制出高效率低成本的微混合S型阵列液滴材料筛选的实验装置，为材料研发提供了新模式和思路。.在项目实施期间，项目组设计并制备了新型微混合+S型阵列微流控芯片；加工了9乘以9梯度变化的微坑阵列；搭建了一个微流控液滴高通量筛选系统；实现了液滴制备和快速表征一体化控制；优化了Faster R-CNN和YOLO算法在晶体形貌上的识别，共发表SCI相关学术论文 5 篇，申请发明专利 1 项。综上，本项目圆满完成立项时制定的研究目标。