This page is mainly about my research interests since the beginning of my PhD in Twente. I also have worked on other things during my life.

Convective turbulence

Starting with my PhD thesis, and continued thereafter, I studied turbulence in several canonical systems: Taylor-Couette flow, Rayleigh-Bénard convection and Double diffusive convection using direct numerical simulations.

These systems are used as basic models when trying to understand the behaviour of more complicated natural or man-made flows. For example, Rayleigh-Bénard is useful to understand what is happening inside the Earth's mantle, why the oceanic currents behave like they do, or how is heat transported inside the sun. Likewise, Taylor-Couette can be used as a basic model for many shear flows, going from the astrophysical scale -accretion disks around protostars- to the flow around a jetski's hull. Double diffusive convection is commonly used to model the convection in the ocean. So even if these are easy systems to construct and simulate, they find applicability in many places and not all is known about them.

My PhD thesis was mainly focused on investigating the boundadry layer transitions in the Taylor-Couette system, and it's relation to the analogous Rayleigh-Bénard problem. I think that this problem is now better understood

One of the questions I'm still interested in is on the nature of the large scale structures which form, c.f. the Taylor rolls in this video. Even for very large Reynolds numbers, and in very large computational boxes, these structures seem to be pinned in the same location over time. Up to now, the mechanisms behind this stationarity are not really well understood. Investigating them could lead to the posibility of manipulating their position, and these are topics I'm still interested in.

The main question I focus on is the role the linear instabilities of these systems have on these large-scale structures, by affecting the way small scale fluctuations merge to form these structures. Interestingly enough, structures seem to form in similar ways in non-rotating Rayleigh-Bénard and Taylor-Couette. However, structures for rotating Rayleigh-Bénard look completely different, and look like the cylone/anti-cyclone patterns find in Earth's weather. Thus understanding these structures can lead to more insight on geophysical flows.

AFiD: an open-source code for DNS of turbulent flows

The simulations in the previous research topic are performed using a second order finite difference code [1] with fractional time integration [2]. Other add-ons to the basic simulations are the addition of a Lagrangian second-phase in the flow, and the use of a multiple resolution scheme to decouple the grids where scalars and momentum are discretized [3].

The numerical code used is written in Fortran 90, and has a hybrid MPI/OpenMP parallelization. The MPI domain decomposition is a 1D slab type decomposition, or a 2D decomposition, i.e. in pencils, depending on the version of the code. It uses parallel HDF5 for output, and FFTW/Lapack libraries for some mathematical operations. Simulations ran using this code won in 2012 the Wim van Niewpoort award for most efficient use of HPC computing time in the Netherlands. It has run on up to 10k cores in IBM (Fermi@Cineca), Cray (Hermit@HLRS) and Bull (Curie@TGCC) systems, and was shown to scale up to 64k cores on Curie Thin Nodes.

The 2D version of the code, based on the library 2DECOMP was open sourced under the name of AFiD, and is available for download. Details of the numerical schemes of the code and the computational algorithms are detailed in Ref. [4]. Thanks to help from people at nVidia, a GPU (CUDA) version of the semi-implicit code will soon be available.

Due to the flexibility of second-order finite differences, it is easy to extend the code to include objects in the flow through the immersed boundary method. In our collaboration, we have already simulated Rayleigh-Bénard and Taylor-Couette ith “rough” boundary conditions. Currently, it is possible to include fixed objects, and/or objects with a prescribed motion into the parallel solver, while the serial solver (closed source) is much more flexible, and includes both weak and strong coupled fluid-structure interaction, and an immersed boundary formulation based on moving least squares. We are currently working on extending AFiD with these modules so that it is be able to simulate objects which naturally move in the flow, and which deform.

Ring vortex colissions

Another topic I have recently began working on consists in simulating the collision of two ring vortices, to study the very fast formation of very small (turbulent) structures from relatively smooth initial conditions. Experiments find it hard to control the nature of the perturbations and instabilities of ring vortices, and have results which are reproducible in a general sense, but not really exactly the same, especially due to the chaotic nature of turbulence. In simulations very smooth initial conditions or initial conditions with controlled perturbations can be used, which can allow for better understanding of the vortex-stretching mechanisms behind the formation of these turbulent “clouds”.

Olefaction and fluid mechanics

Smell is a weird sense. In every day life, we tend to underestimate it's importance, because the post-industrial revolution world is focused around sight. However, smell is the only sense not connected to the rest of the brain through the thalamus, but instead it is connected to the limbic system, where most of our emotions rest. Smell is critical for taste, and also for finding a partner. However, it is not very well understood.

It appears that the olefactory tract in humans has a shape that priviliges retro-nasal smell (smelling what's inside your mouth) in contrast to ortho-nasal smell (smelling things outside). This is especially noticeable if you compare the shape of a dog's head, which looks like a long rectangle, and privileges ortho-nasal smell to the shape of a human, which looks more like a tall rectangle. Even accounting for the un-natural size of our brain does not explain these differences. I am interested in the effect of the shape of olefactory tracts on the fluid dynamics of different noses, and understanding how and where in the nasal cavities are things detected by our olefactory sensors.