Presentation Information

[22a-12N-4]Development of Chemically Amplified Resist Assisted by Deep Learning

〇Chen Tang1, Tanaka Toshiaki2, Sekiguchi Atsushi2, Hirai Yoshihiko2, Yasuda Masaaki2 (1.Osaka Pref. Univ., 2.Osaka Metropolitan Univ.)

Keywords:

resist synthesis,deep learning,monomer concentration ratio

In the development of new photoresists, thousands of preliminary synthesis experiments are required by changing the ratios of monomers and other components. In order to improve the efficiency of resist development methods, we newly propose a method to suggest the optimal synthesis ratio using deep learning (DL) based on the limited testable pre-experimental results. Only 5 exposure-development properties of preliminary test resists with different monomer composition ratios are learned to predict the exposure-development properties for the monomer composition ratio. A two-layer neural network with a back-propagation scheme was used to predict the properties. From the predicted results assisted by DL, γ and Eth values were extracted to predict the optimal monomer composition ratio. A validation experiment was conducted to verify that the γ and Eth values were as predicted. This dramatically reduced the preliminary experiments.