Real-time machine-learning-driven control system of a deformable mirror for achieving aberration-free X-ray wavefronts

Aug 6, 2026

Upload by Patrick

A neural-network machine learning model is developed to control a bimorph adaptive mirror to achieve and preserve aberration-free coherent X-ray wavefronts at synchrotron radiation and free electron laser beamlines. The controller is trained using a mirror actuator response directly measured at a beamline with a real-time single-shot wavefront sensor based on a coded mask and wavelet-transform analysis. Testing on a bimorph deformable mirror at the 28-ID IDEA beamline achieves few-second response time and sub-wavelength accuracy at 20 keV, outperforming a linear model and generalizing across different mirror bending mechanisms.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages17

Reading Time43 min

AccessFree to download

Introduction

Adaptive optics challenge in coherent X-ray beamlines

Limitations of linear, open-loop deformable mirror control

Methods and approach

Neural-network controller for bimorph adaptive mirrors

Real-time single-shot wavefront sensing and training data

Experimental validation

Beamline testing results at the 28-ID IDEA beamline

Response time, wavefront fidelity, and comparison to linear models

Lawrence Berkeley National Laboratory LBL Publications Title Real-time machine-learning-driven control system of a deformable mirror for achieving aberration-free X-ray wavefronts. Permalink [https://escholarship.org/uc/item/6fx7w42m](https://escholarship.org/uc/item/6fx7w42m) Journal Optics Express, 31(13) ISSN 1094-4087 Authors Rebuffi, Luca Shi, Xianbo Qiao, Zhi et al. Publication Date 2023-06-19 DOI 10.1364/oe.488189 Copyright Information This work is made available under the terms of a Creative Commons AttributionNonCommercial License, available at [https://creativecommons.org/licenses/by-nc/4.0/](https://creativecommons.org/licenses/by-nc/4.0/) Peer reviewed [eScholarship.org](eScholarship.org) Powered by the California Digital Library University of California Research Article Vol. 31, No. 13/19 Jun 2023/Optics Express 21264 Real-time machine-learning-driven control system of a deformable mirror for achieving aberration-free X-ray wavefronts LUCA REBUFFI , 1,* XIANBO SHI , 1 ZHI QIAO, 1 MATTHEW J. HIGHLAND, 1 MATTHEW G. FRITH , 1 ANTOINE WOJDYLA , 2 KENNETH A. GOLDBERG , 2 AND LAHSEN ASSOUFID1 1 Argonne National Laboratory, 9700 S CassAve, Lemont, IL 60439, USA 2 Lawrence Berkeley National Laboratory, 1 Cyclotron Rd, Berkeley, CA 94720, USA * [lrebuffi@anl.gov](lrebuffi@anl.gov) Abstract: A neural-network machine learning model is developed to control a bimorph adaptive mirror to achieve and preserve aberration-free coherent X-ray wavefronts at synchrotron radiation and free electron laser beamlines. The controller is trained on a mirror actuator response directly measured at a beamline with a real-time single-shot wavefront sensor, which uses a coded mask and wavelet-transform analysis. The system has been successfully tested on a bimorph deformable mirror at the 28-ID IDEA beamline of the Advanced Photon Source at Argonne National Laboratory. It achieved a response time of a few seconds and maintained desired wavefront shapes (e.g., a spherical wavefront) with sub-wavelength accuracy at 20 keV of X-ray energy. This result is significantly better than what can be obtained using a linear model of the mirror’s response. The developed system has not been tailored to a specific mirror and can be applied, in principle, to different kinds of bending mechanisms and actuators. © 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement 1. Introduction Achieving and maintaining high-intensity focused X-ray beams with near-perfect wavefront quality and high stability is the main challenge and prerequisite for experiments at 4th-generation synchrotron radiation and free electron laser (FEL) beamlines. This represents the main challenge for X-ray optical elements, which will necessarily have much more demanding specifications than those for other applications because of the shorter wavelength and the ultra-small emittance of the radiation beams generated by these sources. When using coherent photons from diffractionlimited light sources, it is critical to maintain a well-controlled wavefront and suppress unwanted static distortions and dynamic disturbance [1] . Not only can the degradation of the wavefront be detrimental for phase-sensitive imaging techniques like tomography [2], but wavefront uniformity is of particular importance for coherent X-ray scattering experiments using X-ray Photon Correlation Spectroscopy, Coherent Surface Scattering Imaging, and Coherent X-ray Diffraction Imaging techniques. Wavefront distortions degrade the sample speckle contrast, which can hinder data interpretation [3] and lead to the failure of the sample phase retrieval and reconstruction. Therefore, the optical elements must i) be manufactured with a surface figure closely following an ideal mathematical shape, ii) automatically and repeatably align and focus the beam to match different sample and experiment requirements, and iii) provide real-t

FAQs

What problem does the proposed control system address?

It targets the challenge of achieving and maintaining high-quality, stable, aberration-free coherent X-ray wavefronts for synchrotron and FEL experiments.

How is the neural-network controller trained?

It is trained on beamline-measured actuator response data using a real-time single-shot wavefront sensor that combines a coded mask with wavelet-transform analysis.

What performance was demonstrated in the beamline experiment?

At the 28-ID IDEA beamline, the system achieved a response time of a few seconds and maintained desired wavefront shapes with sub-wavelength accuracy at 20 keV, outperforming a linear response model.

Aug 6, 2026

Upload by Patrick