Bragg CDI

A new development effort to create an HPC-enabled implementation of the genetic CDI phase retrieval approach. This work is under active development and is not yet deployed for regular use.

Description

A genetic algorithm approach to CDI phase retrieval will improve coherent imaging in two aspects. The first is to enable the recovery of highly reproducible images from a given data set. The second is to render previously impossible to image samples amenable to CDI, opening the door to a greater scientific impact for the method. The basic idea is to do the same phasing process with tens to thousands of random starting points. The diversity of results is then exploited to arrive at a highly reproducible image of the sample. Another aspect of genetic algorithm approaches is in the “fitness” criterion used to evaluate the population of results. This can be tuned to enable phase retrieval of datasets that have previously been impossible to produce images from. It is desired to implement and parallelize software for fast processing by non-expert beamline users.

Current processing time of a 100 MB sample using serial MATLAB code takes 60 minutes using limited parameters. Current data acquisition time for a 100 MB data set is 20 minutes, and will decrease after the completion of the AP-U. Attaining a robust image of a sample in a computation time nearer the data acquisition time will allow nearer real-time feedback into the experimental parameters. The experimenter may begin to do guided, carefully executed experiments. Currently, the vast majority of Bragg CDI users will benefit from semi-real-time phase retrieval for their data. It will also open the instrument up to far less sophisticated CDI users. This technique will be critical to one or more APS-U beamlines.

Distribution

The code for this application is under active development, and is not yet in production use.

https://github.com/AdvancedPhotonSource/cdi

Acknowledgements

The BCDI software is currently being developed by Barbara Frosik of the XSD Scientific Software Engineering & Data Management group (XSD-SDM) in collaboration with Ross Harder of the XSD Microscopy group (XSD-MIC) with funding from U.S. Department of Energy (DOE) Office of Science under Contract No. DE-AC02-06CH11357.

Argonne National Laboratory is supported by the Office of Science of the U.S. Department of Energy. The Office of Science is the single largest supporter of basic research in the physical sciences in the United States, and is working to address some of the most pressing challenges of our time. For more information, please visit science.energy.gov.