simulation: SC 2011, Gyrokinetic Toroidal Code (GTC)

Discuss how polywell fusion works; share theoretical questions and answers.

Moderators: tonybarry, MSimon

Post Reply
Jeff Mauldin
Posts: 17
Joined: Thu Feb 21, 2008 8:41 pm

simulation: SC 2011, Gyrokinetic Toroidal Code (GTC)

Post by Jeff Mauldin »

Last week I attended the 2011 SC conference. There were (at least) two papers that dealt with a 'production' supercomputer fusion simulation code known as GTC.

my quick web serach (googling 'Gyrokinetic Toroidal Code'):

http://www.nersc.gov/research-and-devel ... hmarks/gtc

http://w3.pppl.gov/theory/proj_gksim.html

This is a PIC code, and I am wondering (haven't really dug in) how easy it would be to adapt this code to polywell simulation.

I attended one of the papers and found it interesting on several levels. First, as far as I understood, they (reasonably) seemed to be doing the most straightforward and simple thing, which was to conduct the simulation by going through a several (I think they had 4 or 6) step process, where you calculated the fields, calculated the particle motion from the fields, and then went back and calculated the fields again. One interesting detail I saw, if I understood correctly, was that they were simulating the (positive ion) particles as non-point (maybe two point) sources with the size based on a gyroradius.

This being SC 11, most of the focus was on how the simulation was divided up among the processors (one paper dealt with GPU acceleration) and how and how much communication has to occur between the divisions, and optimizations that help. In the paper I attended the toroid was divided up into sections around the circular direction which were distributed to the processors, and the cleverness involved dividing these sections in the radial direction and dividing the particles among the processors and optimizing (minimizing) how much particles had to be exchanged between cells.

I was certainly struck by the fact that the work appears to mostly be straightforward simulation of the particles and the fields. There doesn't seem to be much consideration (that I could glean immediately) of issues like x-ray radiation or plasma currents. This hews to my intuition that the best starting point is to simulate a whole bunch of particles rather than trying to get a sophisticated plasma model.

I have no idea (yet) how the GTC code might be adapted for polywell. I'm pretty sure they are concerned with positive ions rather than electrons, so that might have a bearing on how they chose to go about their simulation. My simple hope is that I could use this code mostly by defining and dividing up the original geometry differently. But even if not, the fact that so much work is going on here validates, for me, the idea that we could get some useful simulation results with fairly straightforward simulation calculations.

Post Reply