post doctoral position (M/F) \"Machine Learning for d- and f-Metal Catalysis\"

CNRS; France; Sur place; EUR 42 000 - 54 000; Plein temps

Résumé du poste

CNRS, Laboratoire de Physique et Chimie des Nano-Objets (LPCNO) at INSA Toulouse, invites applications for a 18-month research position. The project focuses on machine-learning strategies to optimise d-/f-metal catalysts, transforming legacy DFT data into a curated database and training equivariant potentials.

Collaboration with experimental groups is expected. Expected work includes harmonising DFT data, developing surrogate models, and disseminating results through publications and open-source

Description du poste

Organisation/Company CNRS Department Laboratoire de physique et chimie des nano-objets Research Field Chemistry Chemistry » Computational chemistry Researcher Profile Recognised Researcher (R2) Application Deadline 19 Oct 2026 - 23:59 (UTC) Country France Type of Contract Temporary Job Status Full-time Hours Per Week 35 Offer Starting Date 15 Nov 2026 Is the job funded through the EU Research Framework Programme? Not funded by a EU programme Is the Job related to staff position within a Research Infrastructure? No

Offer Description

Within the MAD-CAT project (NanoX funding, 18 months), the recruited researcher will develop machine-learning strategies for the optimisation and discovery of d- and f-metal catalysts, towards sustainable (de)polymerization. He/she will turn the legacy of DFT calculations accumulated by the MPC team (several hundred reaction intermediates and transition states) into a curated, open-access database, and use it to train equivariant interatomic potentials (MACE) and interpretable surrogate models. These tools will be applied to CO₂ reduction and stereoselective polymerization.

• Extract, curate and harmonise DFT data (geometries, energies, forces) on d- and f-metal complexes, and build an open, documented database
• Perform complementary DFT calculations (data augmentation with perturbed geometries, on-the-fly refinement)
• Fine-tune a MACE equivariant interatomic potential and set up an active-learning loop
• Couple this potential with the AFIR method and automated transition-state search techniques (NEB) to explore complete reaction networks
• Develop low-data surrogate models based on physico-chemical descriptors, interpreted with xAI (SHAP)
• Contribute to the inverse design of ligands and catalysts (VAE coupled to a neural network)
• Apply these tools to CO₂ reduction and stereoselective polymerization (successive insertions, tacticity control)
• Disseminate results: publications, conference talks, open-access release of codes and datasets (pyPhysChem platform)
• Take part in team life and, occasionally, in the co-supervision of PhD students and interns

The position is based at the Laboratoire de Physique et Chimie des Nano-Objets (LPCNO, UMR 5215 INSA Toulouse – CNRS – Université de Toulouse), on the INSA Toulouse campus. LPCNO carries out multidisciplinary research on nano-objects, from their synthesis to their physical and chemical properties, within five teams: Physical and Chemical Modelling, Nanomagnetism, Nanostructures and Organometallic Chemistry, Nanotech, and Quantum Optoelectronics.
The recruited researcher will join the Physical and Chemical Modelling team (MPC), under the supervision of Profs. Laurent Maron and Romuald Poteau. The team models and analyses molecular systems, materials, complex media and nanomaterials. Its common foundation is strong expertise in DFT electronic-structure calculations, complemented by molecular dynamics, force-field optimisation and GW calculations. It addresses electronic structure, thermodynamic stability, kinetics and spectroscopic properties in a coherent way, in close collaboration with experimental groups; this theory–experiment link is at the heart of its identity. The team has recognised expertise in reaction mechanisms and d- and f-metal complexes (about 600 publications).
Artificial intelligence applied to chemistry is now a strategic focus of the team: property prediction, classification, in silico design and experimental data analysis, with an almost systematic use of explainable AI (xAI). The team develops the open-source pyPhysChem platform and co-supervises several PhD projects at the chemistry–data science interface. It benefits from about 2 million CPU/GPU hours per year at the CALMIP regional supercomputing centre.

• PhD in theoretical or computational chemistry, computational materials science or a closely related field
• Priority expertise: machine-learning interatomic potentials (MACE or other equivariant potentials) and/or the AFIR method
• Solid background in DFT and reaction‑mechanism studies; experience with transition‑metal, lanthanide or actinide chemistry is an asset
• Proficiency in Python and scientific machine learning tools (PyTorch, generative models, chemical descriptors)
• Experience with high-performance computing (CPU/GPU)
• Autonomy, team spirit and interest in dialogue with experimentalists
• Good written and oral scientific communication in English