Title:Decoding the relationship between domain structure and functionality in ferroelectrics via hidden latent variables

Abstract:The correlation between local domain structures and switching behavior in ferroelectric materials is explored using the convolutional encoder-decoder networks, enabling image to spectral (im2spec) and spectral to image (spec2im) translations. This approach builds upon the premise that domain structures and polarization dynamics in ferroelectrics are related, and this relationship is ultimately parsimonious, i.e. is based upon a small number of local mechanisms. In im2spec, the sub-images representing local domain structure are compressed via a convolutional network architecture to a small number of latent variables, and the latter are deconvoluted to yield the spectra. In spec2im, the process is reversed. The analysis of latent variables distributions and their real space representations provides insight into the associated mechanisms. This approach provides a workflow to establish the presence of correlation between local spectral responses and local structure, and can be universally applied to spectral imaging techniques such as PFM, scanning tunneling microscopy (STM) and spectroscopy, and electron energy loss spectroscopy (EELS) in scanning transmission electron microscopy (STEM). The interactive Google Colab notebook that goes through all the analysis discussed in the paper is available at this https URL.