AI summaryⓘ
The authors studied how to use different machine learning models to automatically find the direction of tiny electrical polarizations in a special material from complex electron microscope images. They found that models trained on perfect fake data work well on similar fake images but struggle with real experimental data because of differences between them. By using a custom training method and data processing, the authors could reduce this problem somewhat. They also noticed that when the models made mistakes, it was often because some images lacked clear information, and these mistakes sometimes pointed to defects in the material. This research helps improve ways to analyze microscopic images using AI.
4D-STEMpolarization directionferroelectricsmachine learningResNetVGGconvolutional neural networkPCAk-Nearest Neighborsdata augmentation
Authors
Matej Martinc, Goran Dražič, Anton Kokalj, Katarina Žiberna, Janina Roknić, Matic Poberžnik, Sašo Džeroski, Andreja Benčan Golob
Abstract
Four-dimensional scanning transmission electron microscopy (4D-STEM) provides rich, atomic-scale insights into materials structures. However, extracting specific physical properties - such as polarization directions essential for understanding functional properties of ferroelectrics - remains a significant challenge. In this study, we systematically benchmark multiple machine learning models, namely ResNet, VGG, a custom convolutional neural network, and PCA-informed k-Nearest Neighbors, to automate the detection of polarization directions from 4D-STEM diffraction patterns in ferroelectric potassium sodium niobate. While models trained on synthetic data achieve high accuracy on idealized synthetic diffraction patterns of equivalent thickness, the domain gap between simulation and experiment remains a critical barrier to real-world deployment. In this context, a custom made prototype representation training regime and PCA-based methods, combined with data augmentation and filtering, can better bridge this gap. Error analysis reveals periodic missclassification patterns, indicating that not all diffraction patterns carry enough information for a successful classification. Additionally, our qualitative analysis demonstrates that irregularities in the model's prediction patterns correlate with defects in the crystal structure, suggesting that supervised models could be used for detecting structural defects. These findings guide the development of robust, transferable machine learning tools for electron microscopy analysis.