How supernova explosions power galactic winds
by: Creasey, Peter; Theuns,Tom; Bower, Richard G.
[ADS:2013MNRAS.429.1922C; pdf; First author; second; third]I thought that this was an interesting paper. Because it is very similar to what I am doing it gave me a number of ideas about how to set up specific tests and a way of organizing my data to pick out the important details. This paper is a subset of Peter Creasey's PhD work so the general approach and specific questions that he is exploring are different from mine, but on some level we are doing similar work.
He is using the MHD code FLASH, but only in the hydro configuration (like I am doing with Athena). Even though FLASH has the great advantage of being a AMR code they ran into some problems where the AMR was trying to refine the simulation beyond what was practical so they turned off the AMR component of FLASH and used it as a fixed grid solver (which automatically removed the major advantage of using FLASH, and basically turned FLASH into a FORTRAN version of Athena).
When I read that in the paper I instantly knew why they had done that. Because of the increased resolution from the AMR the time step in the simulation would drop to something incredibly small. Thus to do a 5 Myr simulation like they are doing, it would take a VERY LONG TIME to do a single simulation. I would estimate that a single run may take several months on 128+ processors. By turning off the AMR they could force a coarser resolution and thus take larger time steps, and thus run a number of simulations in less time (I think they did 61 simulations if I remember correctly).
Their resolution for their simulations ranged from 32x32x160 to 256x256x1280 over a spacial range of 200x200x1000 pcs. Thus their highest resolution runs had about 50% fewer cells than my highest resolution runs, but my simulations are spread out over a box of 1000 pc3, so I have slightly lower spacial resolution.
They are looking at how supernova feedback affects mass loading which they define as β≡M˙wind/M˙⋆ (those are supposed to be M dots, as in change of Mass). This is a measure of how much mass is flowing out of the grid over how much gets converted into new stars (oh and as a note, for this major important equation they reference Stringer et al. (2011), but there is no reference to Stringer et al. (2011) in the bibliography. There is a Stringer et al. (2012), but even though that paper is the intended paper it does not have the β in it as advertised. Stringer may use another symbol but they don't use β in any of their equations. Anyway minor thing.) Basically they are looking at how much gas escapes the galaxy based on global properties of the galaxy which can then be matched to larger simulations where galactic outflow of individual galaxies is important.
Below is Figure 8 from the paper with the original caption:
 |
| Figure 8. Matrix view of simulations varying gas surface density (Σg) and gas fraction (fg), each panel showing a time-averaged vertical velocity for the upper half plane of each simulation (i.e. the disc is at the base of each panel). Gas surface density increases from left to right, gas fraction increases from bottom to top. There appears to be a strong trend in wind velocity towards the lower right-hand panels, i.e. a disc with low gas fraction but high gas surface density tends to generate a faster wind. |
