MPC model sensitivity analysis: model ambit vs. correction runtime

Mask Process Correction (MPC) is becoming increasingly relevant as the industry moves toward more challenging technology nodes. Because running MPC on large layouts can be extremely resource intensive, it is important to strike a balance between the quality of the correction and the total turnaround time (TAT). This paper describes the results of applying a geometry-based MPC solution to a mask lithography process created at Toppan where the model is calibrated from AEI metrology data of patterns that accounts for beam blur, etch, and proximity effects present in the etched mask up to ~1 um. In this solution, the model calibration can result in different but equivalent predictors, i.e., the model parameters can differ while the overall error residuals (model RMS) can be nearly identical. The following sections probe a possible trade-off between correction quality and speed by testing how an MPC software based on edge movement behaves as the effective range of an enhanced multi-Gaussian mask model template is constrained.