Machine learning in electron beam lithography to boost photoresist formulation design for high-resolution patterning

Area of Science:

• Materials Science
• Nanotechnology
• Artificial Intelligence in Manufacturing

Background:

• Critical dimension (CD) reduction enhances chip resolution and performance.
• Electron Beam Lithography (EBL) enables sub-10 nm patterning, but resolution depends heavily on photoresist formulation.
• Optimizing photoresist formulations for smaller CDs is time-consuming and costly, with limited focus on formulation itself.

Purpose of the Study:

• To develop an efficient photoresist formulation optimization technique for high-resolution patterning.
• To leverage machine learning (LSTM) to predict critical dimensions (CDs) based on photoresist formulations.
• To accelerate the screening of suitable photoresist formulations for electron beam lithography (EBL).

Main Methods:

• Combined electron beam lithography (EBL) experiments with a machine learning long short-term memory (LSTM) network.
• Developed a CD photoresist evaluation model using the LSTM network.
• Utilized the CD model to create a photoresist formulation optimizer, achieving a line width of 26 nm.

Main Results:

• The LSTM network accurately predicted CDs, showing consistency with experimental EBL results.
• The developed optimizer successfully screened photoresist formulations meeting specific CD requirements.
• Demonstrated the feasibility of AI in accelerating photoresist design for sub-10 nm patterning.

Conclusions:

• The AI-driven approach significantly enhances photoresist formulation design for high-resolution patterning.
• The LSTM network provides a reliable tool for predicting CD performance and screening formulations.
• This work offers a novel perspective and practical guidance for utilizing artificial intelligence in EBL experiments and photoresist development.