Pixelated source-mask joint optimization (SMO) plays a crucial role in improving the resolution and image fidelity of optical lithography process. However, the ever growing integration density of semiconductor devices incurs the explosion of data throughput, and poses a considerable challenge on the computational efficiency of pixelated SMO algorithms. This paper proposes to use compressive sensing (CS) to effectively reduce the computational complexity of the SMO algorithm. The proposed SMO algorithm is sequential, where the lithographic source and mask patterns are optimized alternatively. Based on the lithography imaging model, source optimization is transformed to a linear CS reconstruction problem with a nonnegative constraint on the illumination intensity. On the other hand, the problem of mask optimization is solved by a nonlinear CS method aided by sparsity and low-rank regularizations. The dimensionality of the objective function is reduced by downsampling the layout pattern to be printed. The improvement in computational efficiency is verified, and a set of simulations are presented to assess the proposed SMO algorithms. Speedup by several-fold is attained over traditional gradient-based SMO method. Improvement in lithography image fidelity and mask manufacturability is obtained compared to the traditional SMO method and state-of-the-art CS-based SMO method.

中文翻译：

---

基于压缩感知的快速像素化光刻源和掩模联合优化

像素化源-掩模联合优化 (SMO) 在提高光学光刻工艺的分辨率和图像保真度方面起着至关重要的作用。然而，半导体器件不断增长的集成密度导致数据吞吐量的爆炸式增长，并对像素化 SMO 算法的计算效率提出了相当大的挑战。本文提出使用压缩感知（CS）来有效降低SMO算法的计算复杂度。所提出的 SMO 算法是顺序的，其中光刻源和掩模图案交替优化。基于光刻成像模型，光源优化被转换为对光照强度具有非负约束的线性 CS 重建问题。另一方面，掩码优化问题通过由稀疏和低秩正则化辅助的非线性 CS 方法解决。通过对要打印的布局图案进行下采样来降低目标函数的维数。验证了计算效率的提高，并提供了一组模拟来评估所提出的 SMO 算法。与传统的基于梯度的 SMO 方法相比，速度提高了几倍。与传统的 SMO 方法和最先进的基于 CS 的 SMO 方法相比，获得了光刻图像保真度和掩模可制造性的改进。并提供了一组模拟来评估所提出的 SMO 算法。与传统的基于梯度的 SMO 方法相比，速度提高了几倍。与传统的 SMO 方法和最先进的基于 CS 的 SMO 方法相比，获得了光刻图像保真度和掩模可制造性的改进。并提供了一组模拟来评估所提出的 SMO 算法。与传统的基于梯度的 SMO 方法相比，速度提高了几倍。与传统的 SMO 方法和最先进的基于 CS 的 SMO 方法相比，获得了光刻图像保真度和掩模可制造性的改进。