Predictive Analysis of Thermoelectric Materials with Graph Neural Network

Niveau : Master 2

Période : spring 2025

Context:

The Jean Rouxel Institute of Materials in Nantes (IMN) and the Laboratory of Digital Sciences (LS2N) are joining forces to propose a six-month internship in the field of researching new materials for energy (Thermoelectric materials). This project aims to combine the expertise of IMN in machine learning for
the discovery of new materials with the recognized skills of LS2N in the field of deep networks, including Graph Neural Networks (GNN).

Subject:

The project will leverage the advantages of Graph Neural Networks (GNN), machine learning algorithms capable of analyzing complex network structures, such as the atomic structures of materials. Collecting data on the properties of different materials are used to train a GNN model to predict their properties based on their atomic structure. Specifically, each atom is represented as a node in the graph, and the connections between atoms are represented by edges. This graphical representation of atomic structures is used as input for the GNN model. The network’s output will be the prediction of material properties associated with these atomic structures. After optimization, this innovative approach will enable the rapid identification of promising candidates thermoelectric then could be applied to other applications (photovoltaic, catalysis …).

Application:

Send your CV and motivation letter to david.berthebaud@cnrs-imn.fr, romain.gautier@cnrs-imn.fr
and harold.mouchere@univ-nantes.fr