BackgroundDue to the limitations of optimization degrees of freedom in traditional optical proximity correction, it cannot meet the mask optimization requirements of advanced technology nodes. Inverse lithography technology (ILT) is considered the most promising resolution enhancement technique, and the trade-off between mask optimization quality and computation time is a challenge.AimThe biggest limitation of ILT is its high computational complexity, which requires exploring an ILT algorithm that can ensure the fidelity of lithography patterns and the process variation (PV) band while also having a short computation time.ApproachWe propose UNeXt-ILT, a deep learning–based ILT technology. The UNeXt model is adopted as the backbone model, and its multi-layer perceptron structure ensures the lightweight of the model while having global context-awareness capability, thus quickly providing a high-quality initial mask and accelerating the overall computation time. In addition, the addition of mask regularization and mask filtering techniques enhances the robustness of gradient descent–based ILT algorithms and further improves the quality of mask optimization.ResultsCompared with the most advanced deep learning–based ILT algorithm, UNeXt-ILT reduces L2 error by 17.83%, reduces PV band by 8.76%, and shortens turnaround time by 34.48%.ConclusionsWe contribute to improving the robustness and computational speed of the ILT algorithm, thereby promoting its wider application.
Lithography, one of the key semiconductor manufacturing processes, requires accurate and robust process window analysis to ensure high-yield chip production. Traditional methods often utilize fixed polynomial functions to fit focus-exposure matrix data, which can fail to account for variations across mask patterns and noise in critical dimension measurements, leading to inaccurate evaluation of the process window and best process condition. This paper introduces a stepwise regression approach that iteratively selects statistically significant terms based on p-values. Evaluation through adjusted R2 and window overlapping before and after noise introduction demonstrates the method’s effectiveness in enhancing both accuracy and noise robustness in process window analysis.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.