Effective Point Spread Function Estimation for ADF-STEM Resolution Enhancement Using an Atomically Thin 2D Reference

Abstract

Scanning transmission electron microscopy (STEM) provides atomic-column images for structural and compositional analysis, but resolution is limited by probe broadening, residual aberrations, scan instability, specimen-dependent scattering, and noise. Recent computational approaches for recovering high-spatial-frequency information from experimental images include electron ptychography, machine-learning-based super-resolution, and deconvolution. Deconvolution provides a practical and physically interpretable route for conventional annular dark-field STEM because it directly uses an image-formation model and does not require additional four-dimensional data acquisition or training datasets. However, its reliability is governed by the point spread function used for reconstruction. We present an effective point spread function estimation workflow using an atomically thin two-dimensional reference. A monolayer MoS2 reference was used to reduce thickness-related multiple scattering and specimen-specific broadening, and its preprocessed experimental image was compared with a zero-aberration simulated image through gradient-based optimization of an effective point spread function. The estimated effective point spread function was applied to Si [110] and Nd2Fe14B [100] images acquired under closely matched imaging conditions, extending detectable Fourier information in Si and revealing an intensity feature at the expected Fe-column position between Nd columns. These results demonstrate that the proposed approach can serve as a complementary method for improving the interpretability of conventional STEM imaging.