Inverse Reticle Optimization with Quantum Annealing and Hybrid Solvers

Reticle optimization is a computationally demanding task in optical microlithography for advanced semiconductor fabrication. In this study, we explore the effectiveness of D-Wave's quantum annealing (QA) and hybrid steepest descent (SD) solvers in solving pixelated binary reticle optimization problems. We show that the energy derived from the objective function depends on annealing time and inter-sample correlation. Specifically, longer annealing times and reduced inter-sample correlations result in lower energy. Moreover, introducing efficient pausing strategies in forward annealing could reduce the QA runtime by approximately 100-fold while achieving similar results to long annealing times. Finally, reticles with increased variables lead to widespread irregular values in default sorted QA energies due to quantum chain breakages, which could potentially limit the probability of attaining the optimal solution. A hybrid approach that applies the classical SD algorithm to the QA results increases the probability of locating the global minimum solution and reduces runtime to about one-third compared to the classical SD solver. These findings facilitate our comprehension of quantum computing for accelerating computational lithography in semiconductor manufacturing.