Because of their outstanding thermo-mechanical properties, ceramic materials are essential building blocks in applications with harsh operating conditions. In this work we consider the case of ceramic materials submitted to radiative environments, where the question of the preservation of the structural integrity of the ceramic components, consecutive to the development of radiation-induced strain, is a critical issue. The characterization of the irradiation induced strain in these materials is therefore of utmost importance.
X-ray diffraction is probably the most well established characterization technique to determine the lattice strain in materials. Combined with numerical modelling methods and least-square fitting algorithms, it allows to retrieve the spatial distribution of the lattice strain, i.e. the strain profile [1,2,3,4,5]. While the success of this method is undeniable, it remains largely semi-automated, in the sense that the simulation is performed under the supervision of a human expert in order to guide the fitting algorithm. Because of this, the time required to analyze the data can range between several minutes to several hours depending on the complexity of the problem.
With the development of modern characterization instruments, especially at synchrotron facilities, but also, though to a lesser extent, with laboratory diffractometers, this semi-automated approach becomes problematic when several GB, or even TB, of data can be generated in the course of a few days. In this context, machine learning algorithms and, more specifically, deep learning algorithms based on deep neural networks, are expected to play a crucial role. In this work we demonstrate that deep convolutional neural networks can be used to automatically retrieve strain profiles in ion-irradiated materials on the sole basis of XRD data, without human intervention and with analysis times << 0.1 seconds. The potential of the method will be illustrated with ZrO₂ crystals irradiated with Cs ions at difference fluences [6].

References:
[1] A. Boulle, O. Masson, R. Guinebretière, A. Dauger, "A new method for the determination of strain profiles in epitaxic thin films using X-ray diffraction", J. Appl. Cryst. 36 (2003) 1424-1431.

[2] A. Boulle, A. Debelle, “Strain profile determination in ion implanted single crystals using generalized simulated annealing”, J. Appl. Cryst 43 (2010) 1046-1052.

[3] H. Palancher, P. Goudeau, A. Boulle, F. Rieutord, V. Favre-Nicolin, N. Blanc, J. Fouet, C. Onofri, “Strain profiles in ion implanted ceramic polycrystals: a new approach based on reciprocal-space crystal selection”, Appl. Phys. Lett. 108 (2016) 031903.

[4] M. Souilah, A. Boulle, A. Debelle, “RaDMaX: a graphical program for the determination of strain and damage profiles in irradiated crystals”, J. Appl. Cryst 49 (2016) 311-316.

[5] A. Boulle, V. Mergnac, “RaDMaX online: a web-based program for the determination of strain and damage profiles in irradiated crystals using X-ray diffraction”, J. Appl. Cryst 53 (2020) 587-593.

[6] Boulle A, Debelle A, “Convolutional neural network analysis of X-ray diffraction data: strain profile retrieval in ion beam modified crystals”, Machine Learning : Sci. Tech.. https://doi.org/10.1088/2632-2153/acab4c