Hello, my name is Arsenios Chatzigeorgiou
and I am going to talk to you about my project in the Summer of HPC called “Visualization
schema for HPC gyrokinetic data”. One problem I had with my project was that
when i was struggling to understand the science behind it i couldn’t say “Come one, you will
get it. It is not nuclear physics” because that’s exactly what it is. But, since most of us don’t know a lot about
nuclear physics (me included), I am not going to try to explain all the variables, the constants
and the equations behind my project. What I am interested in sharing with you is the
importance of the specific field, how the usage of HPC has provided a much needed boosting
in this field, and how my work is actually helping scientists in this field. This is my project’s title, and to help you
understand it I will say that Gyrokinetic is a method of plasma modeling, or specifically
simulating nuclear fusion reactors. Those reactors manipulate small atoms at their plasma
state into steady state plasma fusion, which produces a bigger atom and emits energy. If this one sounded like: Gyro mmmm reactor
mmmm emits energy. Don’t worry, I will clarify everything. So let’s start with the simple
things. It has something to do with generating energy. The small atoms mentioned can be deuterium
or tritium, atoms that are similar to hydrogen. Deuterium has one proton, one neutron, and
one electron, while tritium is similar but has an extra neutron.
The bigger atom mentioned, could be he Hellium. Another thing mentioned was plasma state.
Plasma is a physical state of matter like solid, liquid and gas. Usually by rising temperatures,
atoms are ionized and ions and electrons are floating independently with large energy.
In nuclear fusion, simple atoms in plasma state collide with each other, get binded
and very large amounts of energy are emitted. Scientists try to create nuclear fusion reactors
to generate energy for free, using more or less water molecules, by heating them up using
magnets. It’s like what happening in sun’s core. It
also makes earth happy, since it doesn’t generate nuclear waste. But why would one need a simulation for such
reaction? Scientific research in fusion power started
back in 1940, but is still in development. Problem is we cannot reach steady state plasma
fusion, we cannot make the plasma stay hot for long enough. Scientists tried to change the machine characteristics
to find the ideal for steady state plasma fusion.
Plasma simulations however are much more cheaper and easy to parameterize.
Since, a lot of computational power is needed nuclear physicists use HPCs to do such large
scale simulations. Gyrokinetics as a method of plasma simulation
balances between simplicity and details. Don’t worry, we are almost there. GENE which stands for Gyrokinetic Electromagnetic
Numerical Experiment, is one of those gyrokinetic plasma simulation programs, with good performance
and a lot of parameters to include in simulation. The program output is a couple of binary files
with the calculated results. The major drawback gene has is that the only
way to visualize and analyze those results, is a commercial plug-in based on the IDL language.
Let’s see what an IDL user has to say about it. No it wasn’t easy, I had to see the results.
I pressed on buttons and I couldn’t understand what it is showing. There are a lot of buttons.
It is like a spaceship. Some errors, make us unable to plot some schemes.
Apart from that, the IDL programm is not available for everyone, since a licence is needed in
order to completly analyze the results. And this is where my project comes.
What we managed to do, is to reproduce some of the plotting schemes of the IDL program
into a clean, modern, simple and minimal GUI, in python using PyQt5. We also used some plotting
scripts written for an GENE Tkinter branch. Having everything automated, the user has
to select only the Gene Output folder, and then click the desired plot. Default values
are presented clearly, and all the plots and variables have help message boxes. We managed
to reproduce 1/3 of the IDL program, from the nrg variables variations over time, to
geometric elements of the specific machine, and from momments related plots and animations,
to time steps variation and spatial contour plots. All with the user defined variables
configuration. Yes it is very easy. I have just to click
on the folder of the results, and select the plot I want to see. It’s like seeing the reactor.
The only thing I care now is to make things stay hot in there. That’s what Steve said, and that’s what my
project was. If you want to get more details about the project, check the summer of HPC
site, for my final report. I want to thank Dejan, Gregor and Leon Kos,
and all of the participant of the SoHPC. And that’s all for me.