PtyRANNOSAUR: Ptychography with Robust Artificial Neural Networks Optimized for Subangstrom Accuracy and Ultrafast Reconstruction

Abstract

Abstract We present PtyRANNOSAUR, a data-driven neural network approach that reconstructs atomic-resolution electron ptychography data in seconds, 100–1000 times faster than iterative methods. PtyRANNOSAUR uses convolutional autoencoders to map 4D scanning transmission electron microscopy data to 2D phase images. Each model is trained on a large database of crystal structures and is tailored for a range of experimental parameters, such as accelerating voltage, convergence angle, defocus, and sample thickness. This approach yields high-quality reconstructions without requiring any fine-tuning of hyperparameters, significantly increasing the accessibility and usability of electron ptychography. In addition, the code handles spatial partial coherence, multiple scattering, and scan position errors, which are critical for the highest-quality electron ptychography reconstructions but have not been available without hours-long computations. By testing PtyRANNOSAUR on experimental and simulated data, we show that the neural networks accurately reconstruct atomic structures on a broad range of systems of materials and achieve high spatial resolutions of <0.5 Å, comparable to the best iterative reconstructions of the same data. These advances enable live, easy-to-use, state-of-the-art electron ptychography reconstructions. These features make it possible to conduct hundreds of electron ptychography reconstructions in minutes and should unlock new operational modes for nanocharacterization, such as in situ and high-throughput electron ptychography.

The authors' abstract, as published at the source. Nano Letters, 2026 · DOI ↗