Suivez

la liste

Starting date: as soon as possible.

Description:

Li-ion batteries are becoming ubiquitous in electronic equipments or vehicles. The industry is looking for safer, denser and environmentally friendlier battery technologies. State-of-the-art Li-ion battery cells recently included silicon in the anode in order to increase the energy density. The next disruptive step will come with solid-state batteries that are believed to be the key to increase thermal stability and safety.

In both cases, difficult mechanical problems arise at the electrode level: silicon exhibits a huge volume variation when the lithium stoichiometry changes. This leads to a breathing of the electrodes that creates various mechanical and electrochemical problems. In all solid-state batteries, even the slightest volume change of materials can cause decohesion of the material and capacity losses.

State of the art modelling of Li-ion batteries involves multi-scale simulations. Before recently, there was a barrier at the electrode scale, which was partly due missing input data about the geometry of the porous layer of particles in the electrodes. Therefore, only homogenized geometrical parameters were used in the models (porosity and tortuosity). Simulations at lower scales used virtual generated microstructures [2]. Thanks to the recent application of artificial intelligence to this field, FIB-SEM-3D and high resolution computed tomography images can be segmented and provide high quality geometrical data to run electrochemical simulations on realistic micro-structures [1].

In order to address those questions, CEA develops direct numerical simulation tools that take into account the complex microstructure of electrodes, which consists in a blend of solid active materials, binders, conductor materials, and liquid or solid electrolyte.

The objectives of this post-doctoral position are:

(1) to push the simulations towards a more accurate representation of the electrode microstructure, by using real (experimental) high-resolution 3D images acquired from 3D FIB-SEM imaging and running HPC simulations on massively parallel computers.
The challenge here consists in finding appropriate linear solvers (among the ones provided by the underlying simulation platform TRUST) for the resolution of the linearized coupled problem, that scale well with the problem size and number of cores, and have a reasonable memory footprint and preparation cost. Indeed the building of the preconditioner often represents a high cost, which not acceptable when the matrix changes at each iteration.

(2) to take into account some mechanical effects in the electrodes, by coupling the electro-chemical simulation with a mechanical modelling software.
A first step will consist in using existing Discrete Elements Model (DEM) simulations to compute the geometry changes of a virtual microstructure during cycling. This data will be used to build the Lagrangian displacement of the 3D mesh of the microstructure and will be fed to the electrochemical simulation (one-way coupling).
Further steps consist in a full bidirectional coupling of the two simulations, and switching from simple and fast DEM models to a more elaborate continuum mechanical modelling that will be able to predict the deformation of a real microstructure acquired with 3D Fib-SEM imaging.

Required skills

  • numerical simulations in general, and linear solver experience in particular ;
  • programming languages : Python and C++, and a good working knowledge of the Linux environment ;
  • solid background in general physics (ideally with a highlight on mechanic)

Work environment

The selected candidate will work at CEA Saclay (ISAS/DM2S) near Paris. Regular trips to CEA Grenoble are to be forseen, since this is a joint project between CEA Saclay and CEA Grenoble (LITEN/LMP). This work will also benefit from an active collaboration with another laboratory at CEA involved in continum and discrete elements simulation of nuclear fuel.
The contract has a one year duration, but can potentially be extended for a second year.
Remuneration depends on the candidate’s previous experience and educational background.

References

1) Machine-learning-revealed statistics of the particle-carbon/binder detachment in lithium-ion battery cathodes, Zhisen Jiang et al, Nature communications, May 2020
2) Electro-chemo-mechanical simulation for lithium ion batteries across the scales, Hofmann T et al, Int. J. of Solids and Structures, Feb 2020. https://doi.org/10.1016/j.ijsolstr.2019.05.002