Gilles, M.; Nashed, Y.S.G.; Du, M; Jacobsen, C.; Wild, S.
X-ray ptychography is quickly becoming the standard method for sub-30 nm imaging of thick extended samples. Being a diffraction-based technique, it involves solving an inverse problem to reconstruct the sample structure from a set of measured diffraction patterns. This problem is traditionally solved to recover a 2D representation of a sample, or in some cases a series of 2D slices of the sample; one can then do standard tomography from a set of such images as the specimen is rotated. Here we present a parallelized, optimization-based algorithm to solve the ptychographic phase retrieval problem directly in 3D. Our approach is able to deal with samples that do not fit entirely within the depth of focus of the imaging system. Such samples pose additional challenges to phase retrieval algorithms because of internal diffraction effects within the sample. We demonstrate our approach on a computational sample modeled with 17 million complex variables.