Inverse lithography physics-informed deep neural level set for mask optimization.

Area of Science:

• Integrated Circuit Fabrication
• Computational Lithography
• Deep Learning Applications

Background:

• Optical proximity correction (OPC) is vital for high-resolution integrated circuits.
• Level set-based inverse lithography technology (ILT) offers superior pattern fidelity but suffers from high computational cost.
• Deep learning (DL) can accelerate ILT but lacks domain-specific knowledge for process window enhancement.

Purpose of the Study:

• To develop an efficient mask optimization technique for integrated circuit lithography.
• To enhance both printability and process window (PW) in advanced semiconductor manufacturing.
• To combine the strengths of ILT and DL for improved mask design.

Main Methods:

• Proposed an inverse lithography physics-informed deep neural level set (ILDLS) approach.
• Integrated level set-based ILT as a layer within a deep learning framework.
• Employed iterative mask prediction and correction for optimization.

Main Results:

• ILDLS significantly enhanced printability and process windows compared to pure DL and ILT methods.
• Achieved substantial improvements in computational efficiency over traditional ILT.
• Demonstrated the effectiveness of integrating physics-based knowledge into DL for mask optimization.

Conclusions:

• The ILDLS approach offers a novel and efficient solution for mask optimization in semiconductor lithography.
• Combining DL with inverse lithography physics knowledge overcomes limitations of standalone methods.
• This hybrid approach is promising for advanced integrated circuit manufacturing.