Job offer

Master Degree or equivalent

Established in 1988, the LEM is a joint CNRS-ONERA research unit comprising approximately 25 people, including permanent staff from ONERA, the CNRS, and the University. Within the CNRS, the unit is affiliated with the Institute of Physics (INP), with a secondary affiliation to the Institute of Chemistry (INC). Within ONERA, the LEM is a unit of the Materials and Structures Department (DMAS). Finally, the LEM operates within the academic framework of Université Paris-Saclay. Its research focuses primarily on studying microstructures that develop within materials, as well as on the modeling and characterization of low-dimensional systems. In addition to external collaborations with academia and industry, the LEM works closely with other ONERA units and departments, aiming to develop and optimize material properties for aeronautics and space applications. The LEM is located at the ONERA site in Châtillon.

For several years, nanoparticles (NPs) have emerged as the foundation for new classes of materials whose properties represent a scientific and technological breakthrough compared to their 3D counterparts. Among all analytical techniques for these infinitesimally small objects, Transmission Electron Microscopy (TEM) is one of the most suitable, as it enables structural and chemical studies with unparalleled precision. However, TEM generates images with degraded signal-to-noise ratio, contrast, and spatio-temporal resolution, which hinder reliable data quantification and interpretation. Furthermore, the extraction of structural information from these images relies on manual acquisition, preventing statistical analysis and introducing human bias during post-processing. Recently, the advent of artificial intelligence (AI) algorithms, such as machine learning and deep learning, has demonstrated exceptional performance in visual classification tasks. These techniques have proven revolutionary in numerous fields, and a similar impact is expected in electron microscopy. Over the past few years, the LEM laboratory has developed research activities in this emerging domain. The primary goal of this thesis is to develop a unified deep learning-based framework for in situ TEM in gaseous and liquid environments, enabling automated, high-throughput, and real-time analysis of TEM image sequences.[1-3] Our approach involves building a dataset through two key steps: atomic-scale simulations to generate structural configurations of NPs using tight-binding formalism, integrated into structural relaxation codes such as Molecular Dynamics or Monte Carlo methods, and TEM image simulation from the atomic configurations, incorporating instrumental noise and optical imperfections of the microscope. This synthetic dataset will then be used to train AI models for analyzing experimental TEM data of NPs in vacuum, gas, or liquid environments. References [1] D. Förster et al., Carbon (2020) [2] R. Moreau et al., npj Comput. Mater. (2026) [3] A. Moncomble et al., Ultramicroscopy (2025)