Inverse lithography source and mask optimization via Bayesian compressive sensing.

Area of Science:

• Semiconductor manufacturing
• Computational lithography
• Advanced patterning techniques

Background:

• Source and Mask Optimization (SMO) is crucial for lithographic fidelity in 14-5 nm nodes.
• Balancing lithography fidelity and computational efficiency in SMO remains a significant challenge.
• Previous compressive sensing SMO (CS-SMO) methods accelerated SMO but had limitations in imaging fidelity.

Purpose of the Study:

• To introduce a novel Bayesian compressive sensing source and mask optimization (BCS-SMO) method.
• To achieve both fast SMO procedures and high-fidelity patterns simultaneously.
• To improve upon the imaging fidelity of existing CS-SMO methods.

Main Methods:

• The proposed BCS-SMO method formulates the SMO procedure as a series of re-weighted l1-norm reconstruction problems.
• Weights are iteratively updated within the reconstruction process.
• Bayesian inference principles are integrated into the compressive sensing framework.

Main Results:

• BCS-SMO significantly improves lithographic fidelity compared to the current CS-SMO method.
• The method achieves this improvement with similar computational efficiency to existing techniques.
• Demonstrated superior pattern fidelity in simulation or experimental results.

Conclusions:

• BCS-SMO offers a promising approach for high-fidelity and efficient source and mask optimization.
• This technique addresses the critical trade-off between speed and accuracy in advanced lithography.
• The novel Bayesian framework provides enhanced performance for next-generation semiconductor manufacturing.