Feature-specific correlation of structural, optical, and chemical properties in the transmission electron microscope with hypermodal data fusion
Modern TEM instrumentation can probe a wide range of structural, optical, and chemical properties with unprecedented resolution. However, each of these properties must be recorded in independent datasets using different detector modes with no unifying framework currently available for quantitatively mapping their relationships onto chemically distinct features, particularly in complex morphologies. Here, we tackle this challenge by proposing a data acquisition and analysis workflow called "hypermodal data fusion," describing how to directly couple an arbitrary number of highly disparate detector modes including spectroscopy and scanning diffraction and jointly analyze them for correlations. We demonstrate this concept on a random collection of anatase and rutile nanoparticles, first detailing how to use core-loss EELS to unmix the different polymorphs despite three-dimensional overlap along the beam direction and then showing how this can be used to extract polymorph-specific composition, bandgaps, and crystal structure. We conclude with a discussion on the applicability of this workflow for a broad range of materials systems.