meeting 2024 01 25 n315 - JacobPilawa/TriaxSchwarzschild_wiki_5 GitHub Wiki

I've finalized all the final numbers with sources where I got them here, and ran the nDither = 1 grid of models!

Updates from yesterday: I did another pass through of the numbers/kinematics we're giving the models following our conversations yesterday:

• I've masked the single overlapping Mitchell bin by increasing the errors on the moments for this bin to 10000.

• I've added dummy moments for GMOS h9 through h12, and Mitchell h7 through h12. I did this by assigning all moments to 0.0, and adding uncertainties equal to the highest order odd and even moments respectively.

• I obtained the PSF numbers (these were already correct, but I re-verified):

  • GMOS PSF = 0.38" FWHM; in the code, we input sigma of this PSF, which I compute as 0.38"/(2sqrt(2ln(2))) = 0.161
  • Mitchell PSF has a sigma of 0.5, which I am using from N2693/N1453/N57/there are notes on BOX to use this value.

• I updated the kingridtheta numbers (difference between the MGE +x axis and the kinematic axis)

  • The number we input to the code is PA_MGE - PA_kinematics.
  • The MGE has a PA of 44.31 (from Jonelle's paper)/
  • The GMOS +x kinematics is 42.0 (N315_PA_note.txt on BOX)
  • The Mitchell +x kinematics is 42.4 (NGC0315-s2-unfolded-bininfo.txt on BOX)
  • kin1gridtheta = 44.31 - 42 = 2.31
  • kin2gridtheta = 44.31 - 42.4 = 1.91

I've sampled 1000 points with the Latin Hypercube sampling scheme with the following bounds:

• BH = [5e8, 5e9]
• ML = [1,4]
• rho0 = [1e9, 4e9] [in code units which is a factor of 4*pi larger than physical units]
• T = [0.01,0.99]
• Tmaj = [0.01,0.99]
• Tmin = [0.01,0.99]

I've verified that the models were all deprojectable using:

• (1 - T) * Tmin / (1 - T * Tmaj) < QminPrime**2, where QminPrime = 0.664

The resulting point distribution of models are:

Just as a reminder, these are nDither = 1 results!

Given that this is a very coarse grid in 6D, and that these have nDither = 1, I'm happy with how things are looking!

• It doesn't seem that we're missing the minimum in the mass parameters!
• The shape posteriors and 1d panels are a bit funky, but I think this largely comes from the sampling/I am taking these results with a large grain of salt.
• I think, given these results, we're fine to move on to the nDither = 3 results for this same exact grid.

First, just for completeness, here are some stats on the models we've run: