Machine Learning-Based Diffractive Image Analysis with Subwavelength Resolution

Abstract

Far-field analysis of small objects is severely constrained by the diffraction limit. Existing tools achieving subdiffraction resolution often utilize point-by-point image reconstruction via scanning or labeling. Here, we present a new technique capable of fast and accurate characterization of two-dimensional structures with at least λ0/25 theoretical resolution, based on a single far-field intensity measurement. Experimentally, we realized this technique resolving 180 nm-scale features, the smallest available to us, with 845 nm laser light, reaching a resolution of λ0/5. A comprehensive analysis of machine learning algorithms was performed to gain insight into the learning process and to understand the flow of subwavelength information through the system. Image parametrization, suitable for diffractive configurations and highly tolerant to random noise, was developed. The proposed technique can be applied to new optical characterization tools with high spatial resolution, fast data acquisition, and artificial intelligence, such as high-speed nanoscale metrology and quality control, and can be further developed to high-resolution spectroscopy.