Bragg coherent diffraction x-ray imaging of nanoparticles with unknown orientation

Bragg coherent diffraction x-ray imaging (BCDI) allows 3D imaging of the internal structure of nanocrystals and single grains at nanoscale resolution but fundamentally relies on precise knowledge of the angular orientation of the sample, which prevents serial acquisition schemes. Here, we show that completely unknown rocking angles can be recovered, both for experimental and simulated data, using a deep learning-based strategy. Importantly, no prior knowledge of the phase is required, and only an approximate shape is needed. Systematic simulations demonstrate that phase pattern variability, signal-to-noise ratio, and shape variability strongly influence prediction performance. The successful assembly of a single diffraction volume from three similar particles marks a first step toward realizing serial BCDI.