Postdoctoral Fellow (M/F) - Representation of dislocation networks for machine learning of atomistic simulations

Job Description

Missions

Scientific context

Crystalline metal alloys are ideal structural materials due to their unique combination of ductility and strength, which allows them to bend rather than crack under heavy external loads. Metals are also very sustainable because they are lightweight and easily recyclable, thus conserving raw materials. Plastic deformation of metals is made possible by the collective movement of dislocations, linear plastic displacement defects that are generally constrained to move on low-index crystal planes. Under repeated loading, dislocations form dense, entangled networks within the microstructure of a metal, causing work hardening, loss of ductility, and ultimately fracture. Predicting how dislocation networks evolve and lead to component failure is a major open challenge in engineering. Theoretical models of dislocation plasticity are essential, as experimental observations are only indirect or destructive. Despite decades of effort to obtain a closed equation for dislocation microstructure evolution, current methods are physics-inspired but manually tuned, lacking data-driven representations essential for leveraging machine learning tools that have shown great capability in analysis and prediction. Very similar problems were encountered during the construction of atomic potentials for molecular simulations; the solution was the development of high-dimensional “descriptor” functions to represent atomic data sets.

Job description

This postdoctoral fellowship, which is part of the ANR DaPreDis project (see below), will enable the development of a data-driven framework for representing dislocation networks, drawing on recent advances in machine learning for atomic systems. The first part of the work will focus on generating 3D dislocation microstructures using state-of-the-art simulations at the meso and atomic scales (dislocation dynamics and molecular dynamics), followed by the design of a representation for comparing dislocation data from both types of simulations. The proposed data representation must respect the symmetries and invariances of dislocation microstructures and comply with known physical laws. In the second part of the work, data from atomistic simulations and ML predictions will be used to reveal new physical mechanisms or correlations in an unbiased, data-driven manner. The data representations developed will be used to build new data-driven models to advance our understanding of critical open problems in metal plasticity, in collaboration with experimental colleagues.

Activity

Several activities will be carried out in this postdoctoral work.
- Set up and run large-scale atomistic simulations,
- Develop, code, and compare different descriptors adapted to dislocation networks,
- Predict the temporal evolution of systems beyond the simulated times and compare with results at other scales
- Write up the results in scientific papers,
- Communicate with other project members and carry out assignments in their teams.

Your Profil

Skills

We are looking for motivated candidates who hold (or will soon be defending) a PhD in materials science. Experience in atomistic simulations for crystal defects and machine learning is essential, as well as a working knowledge of written and spoken English. Serious candidates should send a cover letter and CV with the contact details of at least two references.

Your Work Environment

Presentation of the Consortium

The ANR DaPreDis project (2024-2028) is a collaborative effort between Prof. Sylvain Queyreau, LSPM, Sorbonne Paris Nord University (FR), and Prof. Thomas Swinburne, CNRS and CINaM, Aix-Marseille University (FR) & University of Michigan Ann Arbor (USA). The project has already recruited a doctoral researcher in mid-2025 to work on complementary aspects of the project. In addition to regular team meetings between Paris, Marseille, and Chicago, the project will support travel to international conferences and research visits to collaborators in the United States and Europe.

LSPM: The Laboratory of Process and Materials Science (LSPM) is composed of researchers from the fields of process engineering, mechanics, physics, and chemistry, conducting research in the broad field of materials science and processing. The LSPM is particularly renowned for its expertise in multiscale simulations and original experiments on the deformation and microstructuring of metals.

CINaM: Located on the magnificent Luminy campus, on the edge of the Calanques National Park, the Marseille Interdisciplinary Center for Nanosciences (CINaM) conducts research on matter at the nanoscale, a vast field covering the growth and microstructural properties of crystalline solids, surface chemistry, catalysis, and the dynamics of living systems.

This position is located in a sector covered by the protection of scientific and technical potential (PPST) and therefore requires, in accordance with the regulations, that your arrival be authorized by the competent authority of the MESR.

Compensation and benefits

Compensation

Starting at €3,131 gross per month

Annual leave and RTT

44 jours

Remote Working practice and compensation

Pratique et indemnisation du TT

Transport

Prise en charge à 75% du coût et forfait mobilité durable jusqu’à 300€

About the offer

About the CNRS

The CNRS is a major player in fundamental research on a global scale. The CNRS is the only French organization active in all scientific fields. Its unique position as a multi-specialist allows it to bring together different disciplines to address the most important challenges of the contemporary world, in connection with the actors of change.

CNRS

The research professions