SUMMARY

Ferroelectrics are a class of functional materials exhibiting a spontaneous polarization, switchable through a sufficiently large external field. Their large piezoelectric response, or induced mechanical strain to applied fields, makes them attractive candidates for miniaturized devices, such as sensors and actuators; however, their performance depends on several factors (e.g., lattice strain, size and motion of ferroelectric domains). Thus, there is great interest in studying ferroelectric phenomena at the nanoscale, and piezoresponse force microscopy (PFM) is the leading method for characterizing these materials' functional response. PFM is a contact mode, atomic force microscopy-based characterization technique, where surface displacements are measured as a function of a bias applied through the conductive tip. However, signal interpretation can be challenging as it is sensitive to many phenomena, such as charge injection, ionic migration, surface roughness, as well as short- and long-range electrostatic interactions between the microcantilever and the surface. The goal of this work is to demonstrate machine learning (ML) based methodologies for fingerprinting and, eventually, quantifying the impact of various signal contributors. The first objective entails the development of data curation approaches to enable the correlation of information stored in multiple PFM response channels, thus maximizing the information density for subsequent analysis. The second objective aims to reduce user bias in analysis by leveraging domain knowledge to guide ML models towards identifying signal contributors of interest. The third objective will investigate a select number dimensionality reduction approaches and evaluate their efficacy in separating and quantifying various signal contributors, with an emphasis on electro-chemo-mechanical effects – prominent non-ferroelectric contributors to the PFM signal in switching experiments. The final objective is to develop an experimental approach to minimize the influence of instrument-based contributions to the response, through deliberate choice of the excitation waveform, supported with computational models of the cantilever-sample interactions.

Meeting link: https://gatech.zoom.us/j/91801409592?pwd=UWNvUENjVkUvTDY3dDVJblQ0Ny94dz09