Optical proximity correction by using unsupervised learning and the patch loss function.

Area of Science:

• Semiconductor Manufacturing
• Computational Lithography
• Machine Learning Applications

Background:

• Miniaturization of semiconductor nodes presents significant manufacturing challenges.
• Traditional optical proximity correction (OPC) relies on slow and costly iterative optimization.
• Existing OPC methods struggle to meet the demands of advanced semiconductor fabrication.

Purpose of the Study:

• To develop a faster and more cost-effective approach for optical proximity correction (OPC).
• To address the limitations of traditional iterative optimization in semiconductor manufacturing.
• To leverage unsupervised learning for efficient OPC implementation.

Main Methods:

• A novel framework utilizing patch loss and a generative adversarial network (GAN) was developed.
• The proposed method employs unsupervised learning, reducing reliance on external OPC tools.
• The target pattern is used directly as model input, simplifying the workflow.

Main Results:

• The framework enables swift realization of optical proximity correction (OPC).
• Unsupervised learning approach provides a fast alternative to iterative optimization.
• The model effectively generates corrected patterns without dependence on proprietary OPC software.

Conclusions:

• The proposed GAN-based framework offers a rapid and efficient solution for OPC.
• This approach significantly reduces the time and cost associated with semiconductor manufacturing.
• The method demonstrates the potential of unsupervised learning in addressing complex lithography challenges.