*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Title: Physical and Chemical Insights into Complex Ferroelectric Oxides Through Machine Learning Approaches
Committee:
Dr. Nazanin Bassiri-Gharb, ME, Chair , Advisor
Dr. Asif Khan, Co-Advisor
Dr. Levent Degetrekin, ECE
Dr. Todd Sulchek, ME
Dr. Matthieu Bloch, ECE
Dr. Mark Davenport, ECE
Abstract: Ferroelectric materials are extensively used for electromechanical applications, ranging from medical ultrasound and underwater transducers, to speakers, micro and nano-positioners, etc.. However, much about the interplay of physical and chemical phenomena resulting in such high electromechanical response remains under debate. In recent years, machine learning (ML) approaches have been used to address said challenges, particularly through local probing of properties via piezoresponse force microscopy techniques. But, the shortcomings of ML, such as their ``interpretability" and lack of physical constraints, have become increasingly apparent. This work addresses the local probing of the electromechanical response of ferroelectrics, and the effectiveness of ML techniques applied to this field. First, a method for imposing physical and chemical constraints and allowing for correlation of the various measurement parameters in ML was developed, enabling a multi-parameter study. Expounding upon this work, a scientifically consistent procedure for applying ML to analysis of PFM studies was demonstrated, highlighting challenges in literatures reports. In a final study on the impact of ferroelectric domain walls, out-of-the-box ML algorithms were outperformed by conventional statistical approaches, demonstrating the limitations of these techniques. These combined studies inform on a range of physical and/or chemical phenomena, which contribute to the electromechanical response of ferroelectrics, including the impact of local heterogeneities and extrinsic effects from ferroelectric domain walls. These studies were further enabled by the creation of an open-source characterization approach based upon wide bandwidth piezoresponse force microscopy, developed to expand the capabilities of the technique and its user community.
