A Neural Network Model for Wafer-Scale Photoresist Thickness Prediction Using Feature-Screened Process Monitoring Data

Uniform photoresist coating is important to increase the depth of focus and reduce the overlay errors in lithography, whose controlling depends on the accurate photoresist thickness uniformity model. In this study, we devised a one-dimensional convolutional neural network (1D-CNN) model utilizing process monitoring data collected from sensors in the coating equipment during the spin coating process to predict photoresist thickness. Furthermore, we optimized the model by implementing a feature selection process using variance thresholding. Our results demonstrate that the accuracy of the model with screened features significantly surpasses that of the model without screening.