| About this Abstract |
| Meeting |
2024 TMS Annual Meeting & Exhibition
|
| Symposium
|
Structure-Property Relationships of Bulk Metallic Glasses
|
| Presentation Title |
D-40: Machine Learning Based 4D-STEM Analysis of Variation in Structural Heterogeneity in Zr-based Metallic Glasses |
| Author(s) |
Minhazul Islam, Gabriel Calderon, Yuchi Wang, Yuchu Wang, Geun-Hee Yoo, Eun Soo Park, Yue Fan, Yunzhi Wang, Jinwoo Hwang |
| On-Site Speaker (Planned) |
Minhazul Islam |
| Abstract Scope |
We present the machine-learning (ML) based angular correlation (AC) analysis of 4D-STEM nanodiffraction data to understand nanoscale heterogeneity in Zr-Cu and Zr-Cu-Al metallic glasses (MGs). When composition varies from binary, ternary, to off-eutectic, AC analysis reveals the change in the medium range ordering characterized by local structural symmetry. However, due to the high degree of structural heterogeneity, it is necessary to develop new ways to analyze the AC data to establish a meaningful connection between the heterogeneity and properties of MGs. We present an unsupervised ML analysis of the AC data using visual representation learning based on Simple Siamese network, which makes statistically reliable distinctions between different types of AC patterns from different MG volumes. ML analysis reveals more details of how structural heterogeneity varies as a function of composition and temperature, far beyond what averaged AC patterns could reveal previously, leading to better understanding of structure-property relationships in MGs. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Characterization, Machine Learning, |