Future Directions in Machine Learning–Driven Nanodevice Fabrication

Abstract

Translate Article

We now stand on the cusp of a similar scenario in nanodevice fabrication. As we move into the far sub–50-nm scale, traditional fabrication methods start to falter. There is a transition needed to solve this scaling problem, and the proposed solution lies in machine learning. Over the past few years, there have been several successful and significant attempts to train a learning machine to generate patterns for nanodevices. These strategies overcome traditional limitations, such as grid-based design processes or the time-intensive nature of design, by simulating fabrication or material properties. This opens up new design spaces that save significant time and money. Although this prospect itself is an exciting demonstration of the power of machine learning, our goal is twofold. On the one hand, we wish to overview the current landscape of machine learning for nanodevice fabrication, spread over various applications and techniques. On the other, this chapter pushes the conversation beyond the current state of affairs and explores possible future directions for machine learning–based nanodevice fabrication. We have organized this essay where we start with an illustration of machine learning for different stages of the nanofabrication process. In that process, we also summarize the different methods used in several related experimental studies, which helped researchers achieve rapid prototyping of a number of nanofabrication strategies, as well as rapid quantitative discovery of new and accurate metric space estimators. We then summarize the recent related reviews, which include both commercial hardware and experimental studies. Finally, we discuss various open problems, as well as the future directions in this research space within a broader context.