Abstract:
The X-ray free-electron laser (XFEL) has opened up numerous scientific opportunities for ultrafast dynamics and particle imaging studies. However, as the complexity of investigated systems increases and demand for XFEL facilities grows, obtaining meaningful measurements within the limited beam time has become increasingly challenging. In this talk, we present two recent efforts to address this issue by incorporating machine learning techniques into data collection and analysis in XFEL. In the first study, we integrate neural network models and Bayesian experimental design algorithms for experiment steering. We demonstrate the effectiveness of this approach in measuring magnetic excitations with X-ray Photon Fluctuation Spectroscopy (XPFS) through numerical simulations. Our benchmarks reveal that the proposed method improves online estimations of model parameters and yields more meaningful measurements. In the second study, we introduce a neural network-based fast Single Particle Imaging (SPI) algorithm capable of reconstructing entire 3D structures of particles from individual diffraction patterns. Employing equivariant neural networks and contrastive learning, we demonstrate that the trained model can reliably distinguish different slices from the reciprocal space volume and accurately recover the corresponding real space volume. We believe that this algorithm will significantly accelerate conventional methods and lay the groundwork for real-time SPI applications in the future.
Bio:
Zhantao Chen recently earned his PhD from the Massachusetts Institute of Technology (MIT) and joined the SLAC National Accelerator Laboratory as a Research Associate. His research interests encompass the application of machine learning (ML) techniques to materials science and condensed matter physics, and innovative integrations of ML methods with computational physics for enhanced data collection and analysis.
Microsoft Teams meeting
Join on your computer, mobile app or room device
Click here to join the meeting
Meeting ID: 267 408 499 014
Passcode: 98Z3Cz
Download Teams | Join on the web
Or call in (audio only)
+1 630-556-7958,,478435753# United States, Big Rock
Phone Conference ID: 478 435 753#
Find a local number | Reset PIN
Learn More | Meeting options