Machine Learning and Big Data in optical CD metrology for process control

In this technical paper we explore the use of machine learning techniques to enable, and make better, process control that uses optical CD metrology. We focus on showing how the combination of machine learning algorithms that, by their nature, enable automation, with a Big Data infrastructure, allows the automation of recipe creation, recipe monitoring, and recipe control and update. This automation is essential for semiconductor manufacturing, where process stability is of utmost importance and is, however, hard to achieve. We also discuss how this combination of machine-learning algorithms and a Big-Data system improves accuracy, throughput, tool matching and repeatability.