Online visualization and data discovery for neutron scattering experiments

Neutron scattering makes invaluable contributions to the physical, chemical, and nanostructured materials sciences. Single crystal diffraction experiments collect volumetric scattering data sets representing the internal structure relations by combining datasets of many individual settings at different orientations, times, and sample environment conditions. It is important to be able to access the global picture through simultaneous and interactive visualization and analysis of the data. An online GUI for neutron scattering scientists and visiting researchers will be developed to interactively visualize and post-process multiple instances of wide-ranging data sets. Such data sets include both broad and narrow features in multidimensional space, requiring up to 10,000 x 10,000 x 10,000 voxels for simultaneous inspection of all scales with full context. This challenge will be addressed via multiple techniques, including client-side hardware acceleration with simultaneous server-side processing to enable responsive zooming. Machine learning techniques will be used to identify uninteresting domains, to autodetect regions of interest, and to flag artifacts associated with sample containers or sample environment components. Remote access to large neutron scattering datasets will be demonstrated for a single-crystal beamline with a time-of-flight diffractometer. Machine learning tools will be used to build a neural network for detecting relevant features in the neutron scattering data. Various state-of-the-art technologies for the rapid interactive viewing and exploration of remote data will be tested. Browser-based scientific notebooks will be developed and deployed to domain scientists for testing of the work as it progresses. The software developed for this project will be open source, and novel aspects will be published in the scientific literature. Enabling the inspection and extraction of data as a whole, especially with the possibility of automated detection of features and problems, will yield unprecedented advances in experiment management, big data analysis and scientific discovery. Improved capabilities for exploring large 3D data sets will benefit commercial companies as well as university and lab researchers. The commercialization strategy will include contract R&D and subscription-based sales of our online software.