Xian Du, mechanical and industrial engineering, has received a two-year $498,764 grant from the National Science Foundation to support his research into a new sensing and control technology for roll-to-roll (R2R) printing.
This research project is meant to establish a technological base for the development of a multiscale in-line measurement, or metrology platform, that, according to Du, could promote “both the invention and manufacturing of revolutionary new flexible electronics products, giving the U.S. a competitive edge in the global economy.”
Du says R2R printing of flexible electronics involves fabricating thin electronic structures ranging in size from nanometer to millimeter along a continuously moving flexible film or substrate at speeds of meters per minute. The film moves between two spools the way film moves through a camera.
“The roll-to-roll printing technique offers the potential to radically shift the cost structure for large-area nanostructured devices and enables versatile applications of flexible functional systems,” says Du. “However, a limitation of present continuous printing processes is that in-line metrology is unavailable for process monitoring and control.”
Du adds that “In this study, ultra-thin print patterns along a continuously moving flexible web are imaged, registered and measured in real-time. This process control system can be adapted for different roll-to-roll printing processes for a variety of applications such as industrial internet-of-things and infrastructure health-monitoring.”
However, Du says, “numerous research gaps must be met for these printing processes to be scaled up to industrial scale.”
The research gaps include invisibility of the ultra-thin patterns in a normal optical imaging environment, loss of pattern registration, optical limits on field-of-view and resolution and inability of conventional control methods to capture high-order dynamics and nonlinearity in these printing processes.
“To meet these research gaps,” says Du, “this project develops in-line metrology for print pattern quality monitoring of nano-thin monolayer print processes, investigates high-resolution imaging and registration of large-area nano- and micron-scale patterns, and explores the deep-learning-based predictive control of R2R printing processes by integrating in-line multiscale metrology and process modeling.”
According to Du, the in-line monolayer pattern is imaged using real-time water vapor condensation figures and synchronous image processing. The predictive model is a recurrent conditional deep predictive neural network that incorporates short-term and long-term nonlinearly dynamic print input-output responses to optimize prediction errors.