Modeling and optimization of thermal-flow lithography process using a neural-genetic approach

The lithography process is the most critical step in fabricating nanostructure for integrated circuit manufacturing. The most important variable in lithography process is the line width or critical dimension (CD), which perhaps is one of the most direct impact variables on the device performance and speed. This study presents a framework combining Taguchi orthogonal experiments, artificial neural network (ANN) modeling technique and genetic algorithm for sub-100 nm contact holes fabrication in lithography process. The Taguchi method utilizes S/N ratio and ANOVA to analyze the significant process parameters related to the CD, whereas the ANN establishes the relationship between controllable parameters and quality responses. The proposed Neural-Genetic approach not only can find optimal or close-to-optimal solutions but also can obtain both better and more robust results than the ANN algorithm. The confirmation results clearly demonstrated both the smaller-the-better CD and nominal-is-best CD (target 50 nm) that the proposed procedure was effective and practicable from a production perspective.