Machine learning applied to proton radiography of high-energy-density plasmas

Abstract

Proton radiography is a technique extensively used to resolve magnetic field structures in high-energy-density plasmas, revealing a whole variety of interesting phenomena such as magnetic reconnection and collisionless shocks found in astrophysical systems. Existing methods of analyzing proton radiographs give mostly qualitative results or specific quantitative parameters, such as magnetic field strength, and recent work showed that the line-integrated transverse magnetic field can be reconstructed in specific regimes where many simplifying assumptions were needed. Using artificial neural networks, we demonstrate for the first time 3D reconstruction of magnetic fields in the nonlinear regime, an improvement over existing methods, which reconstruct only in 2D and in the linear regime. A proof of concept is presented here, with mean reconstruction errors of less than 5% even after introducing noise. We demonstrate that over the long term, this approach is more computationally efficient compared to other techniques. We also highlight the need for proton tomography because (i) certain field structures cannot be reconstructed from a single radiograph and (ii) errors can be further reduced when reconstruction is performed on radiographs generated by proton beams fired in different directions.

Experimental observations of detached bow shock formation in the interaction of a laser-produced plasma with a magnetized obstacle

“…There are a number of techniques available for reconstructing magnetic field deflections from proton images [38][39][40][41][42] , but most are limited to images in the linear contrast regime, and none of those have been shown to deal with images containing strong caustics. However, because of the inherent symmetry of the magnetic fields in our system primarily deflecting protons radially outward from the wire, we can use the new method of field reconstruction for caustic proton images presented in Levesque and Beesley 20 to estimate the path-integrated magnetic fields of our system.…”

Methods for extremely sparse-angle proton tomography

“…Even in ideal scenarios, this reconstruction pipeline will only produce information about line-integrated fields. Symmetry assumptions may be used to make conclusions about the three-dimensional distribution of fields (for example, using Abel transform inversion if the plasma is known to have an axis of symmetry), and [24] proposed that the additional information available when taking proton radiographs from multiple different probe directions could enable full reconstruction of three-dimensional fields. This poses a vector tomography problem, in the most general case for each of the electric and magnetic fields.…”

“…More recently, [23] developed a statistical approach to compensate for lack of information * benjamin.spiers@physics.ox.ac.uk regarding the transverse profile of the proton beam prior to interaction with the plasma. [24] investigated the application of machine learning methods to the problem of proton radiography inversion, noting the degeneracy involved in interpreting path-integrated measurements, and suggested taking proton radiographs from multiple view angles as a method for resolving field structures spatially. While some experiments-for example those of [25] and more recently [9]-have probed similar interactions along different axes, the first full exploration of the possibility of recovering spatially resolved magnetic field structures from proton radiographs using standard tomography techniques is presented here.…”

Machine Learning Analysis of Quasi-Stationary Magnetic Fields Optically-Driven by Short Laser Pulses

“…It is worth to note its use in image restoration in regular [32] and fluorescent microscopy [33], nonmodel-based bioluminescence tomography reconstruction [34], rapid decoding of the sample image from its hologram over an extended depth of field range [35], wavefront estimation [36] or robust photomask synthesis in inverse lithography technology [37]. Recently it has been demonstrated using synthetic data that ML in principle can be used to analyse proton radiography images and deduce important magnetic field parameters [38]. Here we develop this idea to reconstruct electromagnetic fields in real experimental setup In order to produce radiographs in ballistic simulations and to create synthetic data for the ANN training, model electromagnetic field distributions are required.…”