My research is currently focusing on these main topics.

Convection from a physics point of view:
My interest began with the study of a purely volumetrically heated system, i.e., a layer of fluid with a constant temperature imposed at the top and with an adiabatic condition at the bottom. Since it involves only one top thermal boundary layer, this system is simpler than the classical RayleighBénard system, which makes easier the investigation of the fundamental mechanisms controlling thermal convection.
A first part was conducted with the project TERRAMWH (ANR11ISO40004) that aimed to produce internally heating in laboratory experiments. To do so, Angela Limare used a microwave device developed by INCDTIM team to achieve noncontact internal heating. It represents a milestone for laboratory experiments since it is the first time that an experiment can produce noncontact and homogeneous internal heating. Furthermore, I conducted numerical simulations to validate this innovative prototype.
I then extended the set of numerical simulations in order to provide a precise description of the planform of convection as a function of the RayleighRoberts number (Ra_{H}) and the boundary conditions. Interestingly, we found a dramatic flow reversal. For low Ra_{H} , the planform is composed of hexagonal cells with a center downwelling, while, for larger Ra_{H} the planform is composed of hexagonal cells with a center upwelling. It is to our knowledge the first time that this polarity reversal is documented. We also established scaling laws given the characteristics of cold downwellings such as their average number per unit area, horizontal crosssection, average temperature and vertical velocity as a function of Ra_{H}.
My current work focuses on the development of theoretical scaling laws characterizing the surface heat flux and thermal structure of a convective system as a function of its controlling parameters. I first built a theoretical framework for the volumetrically heated system presented above. Then I extended the framework to mixed heating convection, i.e., with both internal and bottom heating. The next step is to study the effect of viscosity variations with temperature. Existing scaling laws are semiempirical and only work for very strong variations of viscosity with temperature in the socalled stagnantlid regime. I aim to go beyond these limitations and propose a theoretical framework working regardless of the magnitude of the viscosity variations.
Coworkers
 Frédéric Deschamps (IES)
 Cinzia G. Farnetani (IPGP)
 Loïc Fourel (IPGP)
 Erika Di Giuseppe (IPGP)
 Claude Jaupart (IPGP)
 Edouard Kaminski (IPGP)
 Angela Limare (IPGP)
 Camelia Neamtu (INCDTIM)
 Vasile Surducan (INCDTIM)
 Emanoil Surducan (INCDTIM)
 James Badro (IPGP)
 Cinzia G. Farnetani (IPGP)
 Yang Li (CAS)
 SangHeon Shim (ASU)
Spin state transition, implications on mantle convection:
Since the observation by James Badro et al. (2003) of an iron spin state transition in Ferropericlase, there are an increasing number of studies about its implications. There was an apparent issue because dynamics and mineral physics studies found a dramatic effect of these electronic transitions while seismologists cannot see any discontinuities at the depth given by high pressure experiments. We considered a classical pyrolitic composition and we calculated the effect of Fe^{2+ }spin state transition in Ferropericlase on density. The fact to consider a whole mantle composition enabled us to study different compositions. In particular to include the Fe partitioning between Perovskite and Ferropericlase that were found to vary importantly with pressure. Our calculated density agreed with both PREM density and high pressure and temperature experiments. We found a modest effect on dynamics solving the apparent discrepancy between mineral physics and seismological observations.A potential effect of spin state transition is to reduce the stability of LLSVP. Yang Li has therefore included my mineralogical models in a set of spherical numerical simulations aiming to investigate the stability of these primitive reservoirs. We found (again) a modest effect of spin state transition, since it only slightly reduces the stability of the primitive reservoirs. The chemical buoyancy ratio of the primitive material remains the dominant parameter influencing their stability.
We are now investigating the possible compositions for the primitive material producing after 4.5 Gyrs of evolution the LLSVP suggested by seismological observations. In particular, we compare the seismic anomalies, the volume and shape of the reservoirs produced by the numerical simulations with the observations. Our models include for instance the effect of spin state transition, alumina, iron partiioning.