Demand-driven dynamic control points insertion for high-fidelity and efficient curvilinear optical proximity correction

Curvilinear optical proximity correction is critical for advanced technology nodes. However, conventional methods rely on a uniform or static control points distribution. Such a strategy neglects the correction demands of different regions and the influence of adjacent regions, leading to inefficient resource allocation and sub-optimal imaging fidelity. A demand-driven dynamic control point insertion framework is introduced to overcome the control points distribution inefficiency in curvilinear optical proximity correction. Our method begins with a sparse initial set of control points and iteratively identifies critical regions. By leveraging the local support property of B-spline curves, our method dynamically inserts a new control point near the critical region and adjusts adjacent points to shift the critical region. By bypassing redundant uniform initialization, this method achieves high-fidelity and efficient curvilinear optical proximity correction while resolving inherent resource misallocation. This paper presents a novel dynamic control point insertion framework designed to address the fundamental flaws in the initial control point distribution for a curvilinear mask. Simulations demonstrate the proposed method's capability to establish rational control point distributions, achieving lower pattern error with up to 20% fewer control points compared to uniform sampling, demonstrating superior efficiency and fidelity.