PhD defense: Laure Cazals

Title

Optimal X-ray speciation imaging of radiation-sensitive materials.

Abstract

State-of-the-art spectral imaging using high-brightness sources offers unique capabilities for the speciation of chemical elements in materials science. Spatial scanning enables the imaging of large areas of objects, whilst the acquisition of a spectrum at each point provides chemical information at the pixel level. The main limitations lie in the resulting long experiment durations and localised dose deposition, which can ultimately alter the sample. Damage is a major concern for data reliability, as well as for the physical integrity of samples, particularly for cultural heritage studies. It is common practice to target key spectral features in representative spectra -acquired from the literature or on reference materials- and to collect a pile of images at these diagnostic energies. Although this empirical approach allows certain chemical phases to be distinguished, we propose a statistical approach that minimises the collection effort whilst maximising information. We demonstrate that the Cramér-Rao bound can be used as a criterion for selecting energy points for a given time budget. Using a Frank-Wolfe algorithm, we identify the energy grid that optimises the solution to an inverse problem and, consequently, the determination of the spatial distribution of pure species. To extend the optimisation process to other experimental parameters, we develop a digital twin specifically designed for the 3D X-ray Raman scattering imaging of paint stratigraphies composed of organic materials. It models a spectral imaging experiment in which the acquisition time, the number of energy points, and the energy grid can be modified synthetically. By simulating a wide range of experimental conditions, we demonstrate that this digital twin enables optimal use of the deposited dose. We show that this method can be applied to further materials and to additional experimental modalities. We provide optimisation protocols for X-ray absorption imaging using various a priori assumptions. Beyond X-ray probes, we demonstrate that it is possible to determine the thickness of a varnish layer based on an optimal selection of excitation and emission wavelengths in UV-visible photoluminescence, thereby providing a clear and usable signal for “low-tech” examination. We discuss the possibility that the approaches developed in this work could be considered routine for future materials analysis to prevent radiation-induced damage, including damage that has not yet been identified.

PhD supervision

• Director: Dr. Loïc Bertrand, Research Director, ENS Paris-Saclay
• Co-director: Dr. Agnès Desolneux, Research Director, CNRS

Members of the jury

• Ludovic Duponchel, Professor, Université de Lille, Reviewer

• Amélie Juhin, Research Director, Sorbonne Université, Reviewer

• Alexandre Dazzi, Professor, Université Paris-Saclay, Examiner

• Erwan le Pennec, Professor, Ecole Polytechnique, Examiner

• Francesca Rosi, Research Director, Università degli Studi di Perugia, Examiner