Public thesis defense Victor Trinquet - IMCN
sst |
Non-linear optical properties of materials: a combined first-principles and machine learning approach
Thursday October 15th, 2026, 4pm - Auditorium BARB91 - Place Sainte-Barbe, 1 - 1348 Louvain-la-Neuve
Nonlinear optical materials are at the heart of modern photonics, enabling technologies such as frequency conversion, ultra-fast switching, and advanced laser systems. However, the discovery of high-performance nonlinear optical crystals is hindered by the stringent and often contradictory requirements they must satisfy, as well as the vastness of chemical space. This thesis addresses this challenge by presenting a combined first-principles and machine learning approach to accelerate the rational design and discovery of nonlinear optical materials. In the first part of this work, we establish the theoretical foundations of nonlinear optics, with a focus on second-harmonic generation, and detail the computational methods rooted in density functional theory and its extensions for predicting linear and nonlinear optical properties.
We then apply these methods to both synthesized and hypothetical materials, providing insights into the origins of their nonlinear responses and validating our computational approaches against experimental data.
This leads us to revisit the sum-over-states approach and derive two new expressions for the SHG components as well as a refined band-resolved analysis to investigate the interactions between the states.
The second part of the work introduces a materials informatics framework that integrates machine learning, federated databases, and high-throughput screening. By leveraging active learning strategies and creating comprehensive datasets of computed refractive indices and second-harmonic generation tensors, we demonstrate how to efficiently explore vast materials spaces and identify promising candidates. The integration of first-principles calculations via modular workflows with machine learning not only accelerates the discovery process but also helps researchers in the understanding of the structure-property relationships governing nonlinear optical behavior. This combined approach paves the way for the development of next-generation nonlinear optical materials tailored to meet the demanding requirements of modern applications across the electromagnetic spectrum.
Jury members
Prof. Gian-Marco Rignanese (UCLouvain), Supervisor
Prof. Xavier Urbain (UCLouvain), Chairperson
Prof. Xavier Gonze (UCLouvain), Secretary
Prof. Christophe De Vleeschouwer (UCLouvain)
Prof. Silvana Botti (Ruhr University Bochum, Germany)
Dr. Alexander Ganose (Imperial College London, UK)