Joint Research Conference

June 24-26, 2014

Abstract:

As advanced technology processes emerge at an exponential rate in modern society, equally advanced analytical procedures must be developed to meet the need. In the field of nanofabrication, engineers are often faced with the challenges of limited resource allocation and tight tolerance requirements. To address these issues, we have developed a novel batch sequential experimentation methodology that adaptively adjusts to process information to guide the design of experiment, aptly named Layers of Experiments with Adaptive Combined Design (LoE/ACD).  LoE is the batch sequential process that utilizes statistical methods to determine the next area of interest and the size of its design space, rapidly honing in on process optima. The ACD objectively balances the characteristics of model-based optimal designs with model-free space-filling designs to adapt to the amount of process knowledge gained in order to improve model parameter estimation while still allowing for exploration of model uncertainties. The synergy of these two components creates a powerful methodology, which we have successfully applied in two real-life scenarios. We determined the optimal temperatures in an elevated-pressure, elevated-temperature deposition process to fabricate 40 nm silver nanoparticles within a tight 5 nm tolerance bound after two layers and twelve data points and in the process to fabricate 20 nm silver nanoparticles after two layers and eight data points, even when the first layer was misspecified due to unknown process mechanisms. Joint work with Sungil Kim, Jye-Chyi Lu, Michael J. Casciato, Martha A. Grover, Heeyoung Kim and Dennis W. Hess (Georgia Institute of Technology)