| About this Abstract |
| Meeting |
2024 TMS Annual Meeting & Exhibition
|
| Symposium
|
Novel Strategies for Rapid Acquisition and Processing of Large Datasets from Advanced Characterization Techniques
|
| Presentation Title |
Data-driven Discovery of Dynamics from Time-resolved Coherent Scattering |
| Author(s) |
Nina Andrejevic, Tao Zhou, Qingteng Zhang, Suresh Narayanan, Mathew Cherukara, Maria K. Chan |
| On-Site Speaker (Planned) |
Mathew Cherukara |
| Abstract Scope |
Coherent X-ray scattering (CXS) techniques play a critical role in the investigation of mesoscale phenomena; for example, X-ray photon correlation spectroscopy exploits correlations between scattered intensity fluctuations to derive insights about microscopic sample dynamics, from particle diffusion in colloidal suspensions to fluctuations of magnetic domains. However, the theoretical description of complex dynamics is challenging to obtain or highly approximate. Here, we develop a data-driven framework to uncover mechanistic models directly from time-resolved CXS measurements. We employ neural differential equations to parameterize unknown real-space dynamics and implement a computational forward model to relate real-space predictions to reciprocal-space observations. Our method recovers dynamics of several computational model systems under various simulated conditions of measurement resolution and noise. We further demonstrate the performance of our framework in practice by applying it to discover dynamics in two experimental datasets. Our proposed framework bridges the wide existing gap between approximate models and complex data. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Characterization, |