Research interests

My interest lies in the performance and quality of numerical simulations. Such simulations are used more and more everyday, not only as numerical experiments helping understand physics, but also as input for decision-makers. It is therefore equally important that simulations meet the performance level required by their use in industrial contexts, but also that the results are precise enough to get confidence in the decisions that are taken based on them.

Neutron transport

I’m interested in the broad field of simulation for the physics of (pressurized water) nuclear reactors, in particular as far as neutronics is concerned.

More precisely, I’m interested in all numerical techniques helping depart from the classical, industrial two-stage methods in which the neutron transport equation is only solved at the scale of a fuel assembly, and whole core calculations are performed using the approximate diffusion model.

I work on the definition and implementation of numerical methods allowing the solution of the neutron transport equation at the scale of whole 3D nuclear reactor cores. In particular, I specialize in numerical techniques involving the Method of Characteristics (MoC) and fusion-like methods (2D-1D iterations), such as implemented in the MICADO solver.

Numerical verification

As far as results quality is involved, I’m interested in the process of numerical verification, in which one checks the consistency between the obtained result and the mathematical problem it is supposed to be a solution of. More precisely, I try to investigate the effect of using floating-point arithmetics to implement algorithms that are usually designed to work with real numbers.

I am one of the original developers of Verrou, a tool relying on monte-carlo arithmetics and random rounding to help diagnose and fix numerical instabilities in industrial scientific computing codes.


Last updated 01/09/2019