Exploring Polarization Direction in Potassium Sodium Niobate by 4D STEM, First-Principles Calculations and Machine-Learning Algorithms
Student Contest:
Yes
Affiliation Type:
Academia
Keywords:
Ferroelectrics, Perovskite Materials, Spontaneous Polarization, Defects, 4D STEM, Density Functional Theory (DFT), Machine-Learning Algorithms
Abstract:
Determining spontaneous polarization in lead-free ferroelectric (K₀.₅Na₀.₅)NbO₃ (KNN) is done by using a method which combines 4D STEM with first-principles calculations and machine-learning algorithms. Polarization, crucial for understanding domains and electromechanical properties, is traditionally measured via atomic-resolution STEM, but 4D STEM diffraction patterns provide a reliable alternative. Neural networks trained on simulated diffraction patterns classify Nb atom displacements into eight polarization directions based on KNN's symmetry, confirmed by preliminary cluster analysis using the FastViT AI model. The next step is to apply a combined approach, integrating first-principle DFT calculations with AI-driven analysis to investigate the influence of point defects. Density Functional Theory (DFT) calculations reveal that oxygen vacancies induce significant displacements in the nearest-neighbor Nb ions, which are within the detectable range of 4D STEM. Models derived from DFT calculations will serve as the basis for simulating 4D STEM patterns to train machine-learning algorithms, potentially enabling them to identify and locate defects within the structure.
Track ID:
3.7
Track Name:
ISIF: Characterization, Design, Theory, and Simulation
Secondary Track ID:
1.4
Secondary Track Name:
Isaf: Characterization & Properties Of Ferroelectrics & Related Materials
