Correction model for photoresist shrinkage in critical dimension scanning electron microscopy based on improved spider wasp optimizer-optimized support vector regression

<p dir="ltr">With the continuous advancement of semiconductor manufacturing, the critical dimensions (CD) of integrated circuits have significantly decreased, imposing more stringent requirements on photolithography metrology precision. Critical dimension scanning electron microscopy (CD-SEM) is widely used for dimensional measurement, but electron beam irradiation during the process causes shrinkage in photoresist materials, resulting in measurement deviations. To accurately predict the photoresist shrinkage (ΔCD) induced by CD-SEM, this study proposes a prediction model based on Support Vector Regression (SVR), with hyperparameters are optimized using an Improved Spider Wasp Optimizer (ISWO). The proposed ISWO integrates cyclic population reduction, good point set initialization, and a dynamic tradeoff strategy to enhance global search capability and convergence efficiency. The ISWO-SVR model’s performance is evaluated against other optimization-based models using MSE, MAE, MAPE, and R². Experimental results demonstrate that the ISWO-SVR model achieves superior prediction accuracy and robustness. This work provides an effective and practical modeling approach to mitigate the impact of photoresist shrinkage in CD-SEM dimensional measurements.</p>