Machine learning reveals hidden components of X-ray pulses

Machine learning reveals hidden components of X-ray pulses

The X-ray pulse (white line) is constructed from ‘real’ and ‘imaginary’ components (red and blue dashes) that define the quantum effects. The neural network analyzes the low-resolution measurements (black shadow) to detect the high-resolution pulse and its components. Credit: SLAC National Accelerator Laboratory

Ultrafast pulses of X-ray lasers reveal how atoms are moving on femtosecond time scales. That’s a millionth of a second. However, measuring the properties of the pulses themselves is challenging. While determining the maximum pulse strength, or “amplitude”, it is apparent, and often the time when the pulse reaches its maximum, or “phase” is hidden. A new study trains neural networks to analyze impulses to reveal these hidden subcomponents. Physicists also call these subcomponents “real” and “imaginary.” Starting with low-resolution measurements, neural networks reveal finer details with each pulse, and can analyze pulses millions of times faster than previous methods.

The new analysis method is up to three times more accurate and millions of times faster than existing methods. Identify the components of each ray to throb It leads to better and clearer data. This will expand the range of science possible with ultrafast X-ray lasers, including basic research in chemistry, physics and Materials science and applications in areas such as quantum computing. For example, the additional pulse information could enable simpler and higher-resolution experiments of time-resolution, reveal new areas in physics, and open the door to new investigations in quantum mechanics. The neural network approach used here could also have broad applications in x-ray and accelerator science, including learning the shape of proteins or electron beam properties.

Characterizations of system dynamics are important applications of X-ray free electron lasers (XFELs), but measuring the time-domain properties of the X-ray pulses used in those experiments is a long-standing challenge. Diagnosing the characteristics of each individual XFEL pulse could enable a new class of simpler and possibly higher-resolution dynamics experiments. This research, conducted by scientists from SLAC National Accelerator Laboratory and Deutsches Elektronen-Synchrotron, is a step toward that goal. new approach trains neural networks, a form of machine learning, to combine low-resolution measurements in both the time and frequency domains and restore the properties of X-ray pulses with high accuracy. This ‘physics-informed’ model-based neural network architecture can be trained directly on unlabeled experimental data and is fast enough for real-time analysis on new generation Megahertz XFELs. Crucially, the method also recovers the phase, opening the door to coherent control experiments with XFELs, shaping the complex motion of electrons in molecules and condensed matter systems.

The search was published in Optix Express.

Machine learning paves the way for smarter particle accelerators

more information:
Rattner et al., Phase and amplitude recovery of FEL X-ray pulses using neural networks and differentiable models, Optix Express (2021). doi: 10.1364/OE.432488

Introduction of
US Department of Energy

the quote: Machine learning reveals hidden components of X-ray pulses (2022, Aug 5) retrieved Aug 6, 2022 from

This document is subject to copyright. Notwithstanding any fair dealing for the purpose of private study or research, no part may be reproduced without written permission. The content is provided for informational purposes only.