Final Project done in Data Analysis and Machine Learning Course in Fall 2023.
This repository explores ferroelectric switching behavior in thin-film materials using Band Excitation Piezoresponse Force Microscopy (BE-PFM) data. Through a combination of machine learning models, visual analysis, and theoretical discussion, the aim is to understand the key physical features that influence polarization switching in Pb(Zr,Ti)O₃.
The repository contains:
- 📘 A whitepaper detailing the methodology and results
- 📊 Static visualizations for initial data exploration
- 🤖 A Jupyter notebook implementing machine learning workflows
- Analyze high-dimensional BE-PFM data collected from a 60×60 pixel grid over 96 voltage steps
- Extract and recalculate physical fit parameters: amplitude (A), phase (ϕ), resonance frequency (ω), and quality factor (Q)
- Apply machine learning to identify trends and features correlated with ferroelectric domain switching
- Present findings through visualizations and a whitepaper
| File/Folder | Description |
|---|---|
BandExcitation_MLl.ipynb |
Main notebook with preprocessing, model training, and evaluation |
BE_Static_Visualizations.ipynb |
Visualizations of raw and fitted data (voltage maps, spatial trends) |
whitepaper.pdf or whitepaper.md |
Explanation of the experimental method, data processing pipeline, and results |
requirements.txt |
Python dependencies list |
@article{agar2020ferroelectric, title={Revealing the Ferroelectric Switching Character Using Deep Recurrent Neural Networks}, author={Agar, J. C. et al.}, journal={[Journal Name]}, year={2020} }