Fast lithographic source optimization adopting a certain interval contour sampling compressive sensing

Fast source optimization (SO) is an urgent need for 14-5 nm node integrated lithography online technology. Our earlier contour sample-Bayesian compressed sensing SO (CCS-BCS-SO) method has good image fidelity but is slow. In this paper, we propose what we believe to be a new certain interval contour sampling fast iterative shrinkage-thresholding algorithm compressed sensing SO (CICS-FISTA-SO) method. The fast iterative shrinkage-thresholding algorithm (FISTA) is accelerated by using the famous gradient acceleration strategy Nesterov acceleration technology to achieve both fast SO and high-fidelity imaging. The CICS-FISTA-SO method samples the image pixels on the contour region more evenly, preserves the image information more, and significantly reduces the computational complexity. To our best knowledge, FISTA technology is for the first time applied to resolution enhancement technology in a lithography system to ensure fast high-fidelity modes. The proposed CICS-FISTA-SO framework expresses the SO process as a series of reweighted l₁-norm problems, which can obtain a pattern of high image fidelity with fewer sampling measurements. The CICS-FISTA-SO method is 2.23 times faster than the CCS-BCS-SO method. In addition, the average mode error of the CICS-FISTA-SO method is 3.5% lower than that of the CCS-BCS-SO method. The results show that CICS-FISTA-SO can realize fast SO and high-fidelity imaging better than CCS-BCS-SO.