Turbulence DNS
Turbulent growth of cloud droplets.
In an NSF-funded project, my PhD work focused on Direct Numerical Simulation (DNS) of cloud droplets in cumulus clouds. This was my entrance to the wonderful world of clouds and I’ve had the chance to work with two of the greatest scientists in the field: Lian-Ping Wang of University of Delaware and Wojtek Grabowski of National Center for Atmospheric Research (NCAR). The main question here is how the warm rain develops from tiny droplets, namely what are the mechanisms causing small droplets to grow?
This problem is challenging because the fluid turbulence (inherently a multi-scale problem) is in the heart of it. So, I developed a massively parallel particle-laden turbulence solver and used it to study the mechanisms leading to droplet growth at 10-60 micron radii.
A brief sketch of the method is depicted in the plot above. The problem consists of simulation of the background air turbulence (vorticity isosurfaces on the left panel) and droplet tracking (right panel) in a 3D cube. This method requires DNS of Navier-Stokes for the fluid and also solution to the particle equation of motion in Lagrangian frame. This algorithm is implemented in a decomposed domain (not shown in this figure) which mandates some sort of a message passing platform. I used MPI for parallelization.
After the fluid field and the particle equation of motion are solved, the next plot is how the middle sections of a rainy cloud look like in the DNS world. Only a single snapshot is shown here. The cones are falling water droplets and the red/blue are vorticity iso-surfaces of the background air. The length of the cones is representative of the droplet velocity. I contributed to the area of particle-turbulence interaction specifically collision statistics, and velocity and acceleration of particles in turbulence.