High accuracy and precision overlay and stitching in electron beam lithography demand rapid and robust detection of alignment marks. Here, we implement an autocorrelation-based adaptive alignment mark detection method that processes one-dimensional backscattered electron signals using first-order differences and autocorrelation to detect marks. Adaptive filter parameters and window sizing are employed to suppress noise while maintaining high computational efficiency, in contrast to conventional two-dimensional approaches. Experiments show that the method achieves sub-2 nm precision (99.99% confidence interval) in both X and Y directions with only 10 line scans at a sampling resolution of 512, corresponding to low positioning uncertainty and a detection time of just 15 ms per mark. These results confirm the method’s suitability for real-time applications (e.g., drift compensation and automatic alignment) and suggest its further integration into key processes in Gaussian beam electron beam lithography, including both linear and nonlinear corrections as well as stitching and overlay processes. A contingency scheme employing robust two-dimensional template matching is proposed to handle potential failures caused by specific conditions (e.g., mark defects or extremely low SNR).
Pan et al. (Mon,) studied this question.