Ultrafast pulses from X-ray lasers reveal how atoms move on femtosecond timescales. That’s a quadrillionth of a second. However, it is challenging to measure the properties of the pulses themselves. While determining the maximum strength or ‘amplitude’ of a pulse is simple, the time when the pulse reaches maximum, or ‘phase’, is often obscured. A new study trains neural networks to analyze the pulse to reveal these hidden subcomponents. Physicists also refer to these subcomponents as ‘real’ and ‘imaginary’. Assuming low-resolution measurements, the 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. Knowing the components of each X-ray pulse leads to better, sharper data. This will expand the science possible using ultrafast X-ray lasers, including basic research in chemistry, physics and materials science and applications in areas such as quantum computing. For example, the extra pulse information could enable simpler, higher-resolution experiments over time, reveal new areas of physics and open the door to new research on 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 the properties of an electron beam.
Characterizations of system dynamics are important applications for X-ray free electron lasers (XFELs), but measuring the time domain properties of the X-ray pulses used in those experiments has been a long-term challenge. Diagnosing the properties of each individual XFEL pulse would enable a new class of simpler and potentially higher resolution dynamics experiments. This research by scientists from SLAC National Accelerator Laboratory and the Deutsches Elektronen-Synchrotron is a step towards that goal. The 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 high-resolution X-ray pulses. The model-based ‘physics-informed’ neural network architecture can be trained directly on untagged experimental data and is fast enough for real-time analysis of next-generation megahertz XFELs. Critically, the method also recovers phase, opening the door to coherent control experiments with XFELs, shaping the intricate movement of electrons in condensed matter molecules and systems.
The research was published in Optics Express.
Machine learning paves the way for smarter particle accelerators
D. Ratner et al, Phase and Amplitude Recovery of X-ray FEL Pulses Using Neural Networks and Differentiable Models, Optics Express (2021). DOI: 10.1364/OE.432488
Provided by the US Department of Energy
Quote: Machine learning reveals hidden components of X-ray pulses (2022, Aug 5) retrieved Aug 6, 2022 from https://phys.org/news/2022-08-machine-reveals-hidden-components-x-ray.html
This document is copyrighted. Other than fair dealing for personal study or research, nothing may be reproduced without written permission. The content is provided for informational purposes only.