Revealing Ferroelectric Switching Character Using Deep Recurrent Neural Networks

The ability to manipulate domains and domain walls underpins function in a range of next-generation applications of ferroelectrics. While there have been demonstrations of controlled nanoscale manipulation of domain structures to drive emergent properties, such approaches lack an internal feedback loop required for automation. Here, using a deep sequence-to-sequence autoencoder we automate the extraction of features of nanoscale ferroelectric switching from multichannel hyperspectral band-excitation piezoresponse force microscopy of tensile-strained PbZr0.2Ti0.8O3 with a hierarchical domain structure. Using this approach, we identify characteristic behavior in the piezoresponse and cantilever resonance hysteresis loops, which allows for the classification and quantification of nanoscale-switching mechanisms. Specifically, we are able to identify elastic hardening events which are associated with the nucleation and growth of charged domain walls. This work demonstrates the efficacy of unsupervised neural networks in learning features of the physical response of a material from nanoscale multichannel hyperspectral imagery and provides new capabilities in leveraging multimodal in operando spectroscopies and automated control for the manipulation of nanoscale structures in materials.