Data corruption in z=0, Illustris-1 mock fits?

Lucas Bignone
  • 10 Feb '16

Hi! I have run into a few mock fits for the z=0 Illustris-1 snapshot that appear to show some kind of data corruption.

For example subhalo 5 shows only noise for all 36 cam0 filters:

5_cam0.png

For comparison the galaxy observatory shows a clear subfind image. I also noticed that it doesn't show any FOF or GZ images.

In other cases I have also run into fits that show off-center structures that do not appear at all in the subfind images of the galaxy observatory. For example subhalo 41170, shows such structure is the content of the cam0, gSDSS filter:

41170_cam0_gSDSS.png

While the corresponding galaxy observatory image shows nothing of the kind.

Is there any problem with the mock fits?

Many thanks in advance!

Dylan Nelson
  • 11 Feb '16

Hi Lucas,

My guess for the first case would be that the FITS image has been mis-centered, so that it didn't actually see any star particles in the FoV.

In the second case, as I recall the FITS images include all stars within the FoV, not just those stars [gravitationally bound] to that particular Subfind subhalo. So this off-center structure is likely another satellite (since 41170 is one of thousands of satellites in FoF 3).

Let me see if I can get firmer answers for you.

Lucas Bignone
  • 12 Feb '16

Thank you Dylan. It is not clear to me if the FITS include other particles beside those belonging to Subfind subhalos, I would certainly appreciate further information.

Dylan Nelson
  • 29 Feb '16

To follow up:

  • The FITS mock data sets include all stars in the FoF group (so, an intermediate case between just particles in the Subhalo and all particles in the FOV, as I surmised above).
  • I've confirmed that you are correct: there is an issue with a small number of satellites in this data set. In particular, whenever the FoF center position and the satellite SubhaloPos are across a periodic boundary from one another. In this case the periodic wrap was incorrect. This is only going to affect some of the most distant satellites of a few of the big clusters. You should add a check for this and throw out these systems: e.g., whenever one or more of X,Y,Z coordinate differences of GroupPos[] and SubhaloPos[] of a satellite is very large.
Lucas Bignone
  • 1 Mar '16

Thank you Dylan, very useful info.

  • Page 1 of 1