The Nuclear Fuel Department (DEC) conducts research and development on nuclear reactor fuels, leveraging unique expertise in characterizing fuel behavior under both normal and accident conditions. This is achieved through hot laboratory tests and integrated experiments in research reactors, closely linked with numerical simulations using multiscale, multiphysics, and multi-reactor models.
The scientific computing tools essential for studying fuel behavior are developed within the Fuel Behavior Studies and Simulation Service (SESC) and are integrated into the PLEIADES digital platform. These tools incorporate simplified laws and models that enable reasonable computation times.
The primary missions of the Fuel Behavior Modeling Laboratory (LM2C) include: providing justification for the simplifications used in PLEIADES applications; integrating a multiscale approach from atomistic calculations to mesoscopic representations in microstructure-scale analyses; simulating fuel materials from atomic to grain scales to determine their properties; designing analytical experiments and interpreting the results in collaboration with laboratories conducting the experiments; and developing the thermochemical approaches necessary for understanding fuel behavior.
Materials, solid state physics
Postdoc
Postdoc: Generative AI application to the calculation of physical properties of nuclear fuels
Development of cutting-edge machine learning tools to study atomic transport properties in uranium-plutonium mixed oxides, a type of nuclear fuel with significant implications for nuclear fuel efficiency and waste reduction.
12
By joining our team, you will have the opportunity to leverage your skills and enthusiasm in impactful projects that benefit society.
Machine learning (ML) is now commonly used in materials science to enhance the predictive capabilities of physical models. ML interatomic potentials (MLIP) trained on electronic-structure calculations have become standard tools for conducting efficient yet physically accurate molecular dynamics simulations.
More recently, generative deep-learning models are being explored to learn hidden property distributions in an unsupervised manner, and generate new atomic structures according to these distributions.
The goal of this project is to combine MLIPs and generative methods to address atomic transport properties in uranium-plutonium mixed oxides (MOX).
You will use the ML generative tools developed by our team to generate representative atomic configurations and build an ab initio database.
You will then utilize this training database to develop a new MLIP for MOX, leveraging the experience gained from developing analogous MLIPs for the corresponding pure oxides.
Finally, you will apply the new MLIP to calculate atomic diffusion coefficients, which are crucial for predicting irradiation-induced microstructure evolution and the in-reactor behavior of MOX fuels.
The work will be conducted at the Nuclear Fuel Department of CEA, within a scientific environment characterized by a high level of expertise in materials modelling, and in close collaboration with other CEA teams in the Paris region specialized in ML methods.
The results will be disseminated through scientific publications and participation in international conferences.
Machine learning, Interatomic Potentials, LAMMPS
You have a PhD in solid-state physics/chemistry, materials science, computational physics, or a related field.
You have some expertise in machine learning.
Experience in atomic-scale modeling is a valuable additional asset.
Submit your application to join our team and contribute to CEA's groundbreaking and innovative projects!
In line with CEA's commitment to the inclusion of people with disabilities, this position is open to everyone. CEA offers worplace adaptations and/or flexible work arrangements."
Cadarache
France, Provence-Côte d'Azur, Bouches du Rhône (13)
Autre
Solid State Physics, Materials Science, Computational Physics, Machine Learning
04/11/2024