meeting 2025 02 26 n315 - JacobPilawa/TriaxSchwarzschild_wiki_5 GitHub Wiki

I started by trying to recreate the maps in Paper XIII. These should be the symmetrically binned N315 GMOS kinematics.

• The files I'm using for this are in GMOS_kinematics/N0315/orbit_model_inputs. In this directory there are two tar files which contain the relevant data.
• The "ppxf" tar file contains the kinematics. There are three files for each kinematic bin denoted like 4050_bootstrap.dat, 4050_kin.dat, 4050_montecarlo.dat.
• I'm assuming that the _kin.dat files contain the actual kinematics and that the uncertainties are from the bootstrap files which is what is stated in Paper XIII.
• One thing to note: there seem to be 6 bins which have missing data from this folder. These bins are 1054, 1059, 1064, 3044, 3050, 3052, and they simply don't have a corresponding _kind.dat file in this directory, but they are included in the file containing the bin information. For these data points, I'm currently setting their values to be NaN.

The resulting maps are just below, and they look more or less consistent with what's in Paper XIII (though there are some strange bits that I am trying to sort out; currently can't tell if it's a colormap/color scale artificat or something else but I'll return to this below).

• One other interesting note -- it doesn't seem like the 1d panels on BOX are consistent with the 1d panels in Paper XIII. Still trying to figure out why/where this comes from. For example on BOX, sigma has 3 points > 400km/s in the 1d panel, but these points are not in the Paper XIII panels. Neither the plot on BOX nor the plot in Paper XIII has the ~> 500 km/s point which is definitely on box for the symmetric binning scheme.

• One thing that stands out to me -- there is a point "1000_kin.dat" near the very center which has sigma ~520 or so, but this does not appear in Paper XIII's appendix plot/it's above the y-axis scale. I'm not sure where this point went in the published version.

The radial profiles seemed to show me that the errors we were obtainging were a good deal smaller than the errors that are on box (again, these are computed as the standard deviation over the 100 bootstrap realizations listed on BOX). Notably, there is a statement comparing this approach to a standard Monte Carlo approach, but I'm using our normal pPXF routine so they two methods should be the same.

Here's a histogram of both the values and the uncertainties for the two sets of data (unsymmetrized for both):

• As you can see, there appears to be a factor of ~1.5 to 2 difference in the size of the uncertainty between my kinematics and those on BOX. I think that that's a decent chance this is the cause of the large chi2 we've been finding.
• This error bar difference I think virtually entirely makes up the difference in our chi2. Inflating the errors on average by ~1.75 will decrease the total chi2 by a factor of ~3, bringing us down to a reduced chi2 of ~0.42 which is inline with our other galaxies.

• While there are minor differences per bin/per moment, the overall trends between the two chi2s are quite consistent with one another. There's nothing obviously wrong or discrepant between the NNLS and kinem chi2s for N315, it seems.