phaser: An all-in-one package for (multislice) electron ptychography

Abstract: Electron ptychography is a groundbreaking technique for the advanced characterization of materials. Successful ptychography relies on robust implementations of reconstruction algorithms with process raw data into phase images. Current software has enabled deep sub-angstrom multislice electron ptychographic reconstructions, but reconstructions remain challenging and computationally expensive. In addition, current software generally lacks the modularity to act as a platform for the development and testing of new solver components and algorithms. For instance, algorithms based on gradient descent and autodifferentiation promise faster reconstructions with higher quality, but it is difficult to implement these algorithms in existing software. Here, we present \texttt{phaser}, an open source Python package which provides a unified interface to both conventional and gradient descent based ptychographic algorithms. Reconstructions are specified in a declarative format, and can be run from a command line, Jupyter notebook, or web interface. Multiple computational backends are supported to provide maximum flexibility. With the JAX computational backend, a 6x improvement in iteration speed is achieved over a state-of-the-art package, \texttt{fold_slice}. Additionally, \texttt{phaser}'s gradient descent solver achieves improved reconstruction quality.