Étude de mécanismes à l’échelle atomique pour les matériaux fonctionnels, apports de la DFT comme un outils en évolution

Résumé :

The purpose to use numerical tools is to treat problems where the analytical solutions are too complex or unknown. In the field of solid-state physics, the advent of the Density Functional Theory (DFT) had a major impact on the studies of materials at the atomic scale. Its numerous implementations and ability to efficiently run on large computing infrastructures, made it popular and it quickly became indispensable in both academic and industrial research, enabling unprecedented insights into the electronic, structural, and chemical properties of complex systems. However, the power of DFT is not solely derived from its theoretical foundations; it is equally dependent on the robustness, efficiency, and adaptability of the computational tools that implement it. ​

This defense will deal with these two aspects: material science through DFT simulations, and maintainance and development of large numerical infrastructures like a DFT code. ​

I will first address how atomistic-level simulations can be used to complement experimental characterisations, through the example of the indirect role played by Se atoms in the efficiency improvement of CdTe solar panels. Revealing such mechanisms is important for material design, driving material engineering by knowledge. ​

A second part will be dedicated to a broad overview on the physics description of a graphite electrode in a Li-ion battery. Starting from a fully charged anode, and following the deintercalation process, I will question what insights we can get from DFT calculations with the existing knowledge obtained from long-passed experiments as from more recent in-operando characterisation results. ​

​Studying the dilute regime in graphite intercalation will lead to open questions about the capability of numerical simulations to address cases where the strong interactions between the cations and the host material, compete with the binding of the layers. To properly address such questions, it is important to have available within DFT, a level of theory capable of treating inhomogeneous systems made of places where covalent bonds are dominant while in other areas van der Waals interactions take the lead. These situations can be commonly found in several classes of materials, from van der Waals heterostructures to hybrid perovskites when out-of-equilibrium processes take place, with defect / impurity diffusion or phase / structural transitions. The recent developments of versatile meta-GGAs associated to dispersion corrections, look promising and have demonstrated their ability to reproduce perfect van der Waals systems. Concurrently, their usage are restricted to some codes, hindered by the complexity in DFT implementations, isolating the diffusion of new ideas. I believe that the advent of code generation through AI is a timely opportunity to help spreading state-of-the-art DFT developments. Such thoughts will be discussed in the last part of the defense.​

_

Contact : alain.farchi@cea.fr