CAREER: Modeling the Roll-to-Roll Soft Lithography Printing Process Through Deep Learning and Real-time Sensing

该学院早期职业发展（Career）基金通过建立一种基于学习的建模方法来指导连续微接触印刷过程的设计和控制，并研究连续图案形成机制，重点关注卷对卷软光刻技术的进步。微接触软光刻是一种极具吸引力的成本效益的方法，通过在柔性基材上使用邮票进行选择性机械接触，形成微米和纳米尺度特征的平方米区域。将微接触印刷适应于连续的卷对卷平台，促进了柔性电子和可穿戴设备等应用。在软光刻中，转移图案的保真度依赖于成功的机械接触和在印基界面上对材料转移的控制。然而，在卷对卷微接触印刷中，通常制造结构的潜在微特征演变尚未完全理解，以指导印刷过程的成功设计和控制。本项目研究了一种新的微接触印刷图案形成的建模方法，通过实时学习图案过程，将图案形成机制与印刷过程变化和缺陷联系起来，以进行质量控制。这项研究还得到了多学科课程的发展的补充，其中包括卷对卷印刷的研究和创建独立的动手教育包，以鼓励不同教育水平的年轻学生从事制造业的职业。本研究的目的是通过深度学习了解微接触印刷的基本机制，为卷对卷软光刻技术奠定科学基础。为了实现这一目标，本研究的目标是：(1)研究用于设计和控制卷对卷微接触印刷的接触区域的物理力学模型；(2)建立在线视觉-力-变形感知网络，用于评估打印几何形状；(3)通过深度学习建立实时时间戳微特征模式-压力-变形行为模型。将打印的图案图像系统地分割并转换为一个单一的几何变量，该几何变量与力和位移数据同步并集成，用于描述打印状态。深度学习架构通过测量打印变量和接触几何形状来模拟打印图案几何形状的保真度。该项目寻求下列问题的答案：(i)印刷变量与接触几何之间的精确可解释关系是什么；（ii）接触几何形状、材料特性和卷对卷印刷参数在图案几何形状形成中的作用。研究的重点是深入了解微接触印模的变形行为和印模几何形状的形成机制。该项目丰富了软光刻建模、实时传感、深度学习、卷对卷印刷过程设计和控制的知识库，有助于智能制造的进步。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This Faculty Early Career Development (CAREER) grant focuses on advances in roll-to-roll soft lithography by establishing a learning-based modeling method that guides the design and control of continuous microcontact printing processes and investigates continuous pattern formation mechanisms. Microcontact soft lithography is an attractive cost-effective method of patterning meter square areas of micro- and nano-scale features via selective mechanical contact on flexible substrates using stamps. Adapting microcontact printing to continuous, roll-to-roll platforms facilitates applications such as flexible electronics and wearables. The fidelity of the transferred pattern in soft lithography is dependent on successful mechanical contact and control of material transfer at the stamp-substrate interface. However, the underlying microfeature evolution in commonly fabricated structures in roll-to-roll microcontact printing is not yet fully understood to guide the successful design and control of the printing process. This project studies a novel modeling approach for microcontact print pattern formation with real-time learning from the patterning processes while linking the pattern formation mechanisms with print process variations and defects for quality control. The research is complemented by the development of a multi-disciplinary curriculum combined with research in roll-to-roll printing and the creation of self-contained hands-on educational kits to encourage young students at various educational levels to pursue careers in manufacturing.The goal of this research is to understand the fundamental mechanics of microcontact printing through deep learning and establishing a scientific basis for roll-to-roll soft lithography. Towards this goal, the research objectives are: (1) Investigate a physical mechanics model of contact regions for the design and control of roll-to-roll microcontact printing; (2) Establish an in-line vision-force-deformation sensing network for assessing print geometry; and (3) Model real time stamp microfeature pattern-pressure-deformation behavior through deep learning. The printed pattern images are systematically segmented and converted to a single geometric variable which is synchronized and integrated with the force and displacement data for describing the state of print. A deep learning architecture models the fidelity of a print pattern geometry by measurement of the print variables and contact geometries. The project seeks answers to the following questions: (i) What is the precise interpretable relationship between print variables and contact geometry; and (ii) What are the roles of contact geometries, material properties, and roll-to-roll printing parameters in the formation of pattern geometry. The overarching focus is to achieve a deep understanding of the deformation behavior of microcontact stamp and the formation mechanisms of print geometry. The project enriches the knowledge base for soft lithography modeling, real-time sensing, deep learning, and design and control of roll-to-roll print process and contributes to the advancements in intelligent manufacturing.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.