Optimization of lithography source illumination arrays using diffraction subspaces.

Area of Science:

• Photolithography
• Computational Imaging
• Semiconductor Manufacturing

Background:

• Traditional lithography illumination optimization (ILO) methods face challenges in computational efficiency and robustness.
• Existing adaptive compressive sensing (CS) approaches for ILO require optimizing numerous variables, limiting speed and performance.

Purpose of the Study:

• To develop a more efficient and robust lithography illumination optimization (ILO) approach.
• To improve computational speed and imaging performance compared to existing ILO methods.

Main Methods:

• Developed an ILO approach based on subspace compressive sensing (CS).
• Optimized only critical source pixels within a defined subspace, reducing optimization variables.
• Formulated ILO as an lp-norm (0 < p < 1) inverse reconstruction problem with sparse source pattern representation.
• Employed an lp-norm reconstruction algorithm for enhanced robustness over l1-norm methods.

Main Results:

• Simulations at 45nm and 14nm technology nodes demonstrated significant improvements.
• The subspace CS method drastically reduced the number of optimization variables, enhancing computation speed.
• The lp-norm reconstruction proved more robust than traditional l1-norm algorithms.
• Overall improvements in computational efficiency, robustness, and imaging performance were achieved.

Conclusions:

• The proposed subspace CS and lp-norm based ILO method offers superior computational efficiency and robustness.
• This approach significantly enhances the imaging performance of lithography processes.
• The method represents a substantial advancement for optimizing illumination in semiconductor manufacturing.