meeting 2025 06 24 n57 - JacobPilawa/TriaxSchwarzschild_wiki_6 GitHub Wiki

• Polynomials:
  • We had been looking a bit at the dependence of the moments on the additive/multiplicative polynomials after noticing that the adeg=-1 seemed to be inflated relative to the adeg=0 case. We had been trying this in the first place since the N315 data were run with adeg=-1 (since we had some issues initially with the spectral fitting, in particular with some of the central bins).
  • It seems like adeg=-1 is definitely inflated for N57, and we were looking for a physical reason why this was the case/how we can justify the particular choice of adeg=0/mdeg=3. To that end, I ran some lines of sigma vs. polynomial degree for both cases, and we saw that there seemed to be stability when using adeg=2/3/4 or so (in some cases, adeg=0 still seemed to be a bit high).
  • The dependence on the multiplicative polynomial seemed a bit less extreme, requiring something like mdeg>2 or so ot get stability in the moments. Encouragingly, we
• Sigma Clipping:
  • We had thought that maybe the sigma clipping routine would find additional features that we were not properly masking from the fits, potentially bringing the adeg=-1 and adeg=0 cases a bit more in agremeent, but this doesn't appear to be the case. In fact, in the diagnostics below, it doesn't seem like the sigma clipping is appreciably changing the fits at all, and we are left with either agreement with the "fiducial" case or a still elevated set of moments when using adeg=-1.
• Library Testing:
  • One of the other N57 related questions was why the Barth subset of stars vs. the "trimmed" library from Emily were leading to disagreements in the moments. Prior to the N315 report coming back, we had been trying to further trim away the "trimmed" library using some phsyical reasoning rather than the way I had been doing it (removing top 15 weights iteratively).
  • I tried to further trim the library by finding the spectral types for all stars, and limiting the library to only GKM stars and further to only GKM main sequence stars.

• I think maybe a bit further testing on the library would be worthwhile to see if there are better approaches to the "trimming." I had expected that limiting our results to only GKM stars would have resulted in better agreement, but because the Barth stars are not present in the "trimmed" library, we might be approaching some limit of agreement between the two libraries (the RMS don't appear to be drastically changing).
• For the polynomial testing -- it's a bit difficult for me to think about what to do next. On one hand, we've run all the models with adeg=0/mdeg=3 which appears to be reasonably well behaved for N57. On the other hand, there does appear to be small shifts in the moments if we further increase the adeg to 1/2/3 or so. If we want to do more work, it might be worth testing a subset of models using the kinematics from adeg=1 for example, but this really might just be opening more cans of worms given how far along we are.
• As we turn to the text/Overleaf, I'd like to go back and check that my list of BH>3e9 cases is complete, since I suspect I might be missing a few recent measurements (for the environment plot).

• Note that there are two categories which appear in the plot which are not actually spectral types. There are 10 stars which are not matched on SIMBAD (I can check these out more carefully but wanted to move forward for now), which get labelled as "N" due to being "Not Found" in the database. Theres another star that is listed as a type "kA9hF2mF2(IV)" star, which is the lowercase "k" entry.

• In general we were progressing nicely with N410, and it seemed like the spectra were a bit more robust than other galaxies in some of our testing. Specifically it seemed like the additive polynomial degree didn't seem to matter as much, nor the choice of Barth vs. full "trimmed" library.
• Reproducing Irina's Results:
  • We had briefly fit the asymmetrically binned spectra (on the first N410 page), but paused briefly so ensure I can reproduce Irina's results (using the symmetric scheme).
  • With the symmetric binning scheme, we could quite nicely reproduce Irina's results, and the asymmetric scheme processed with the same settings also appears to agree in the radial moments.
• Polynomial Degree:
  • There appeared to be less of a dependence on the moments vs. additive degree for N410. There still seemed to be a slight "inflation" for adeg=0, again settling to "robust" values for adeg=2/3/4, but the effect is definitely slight. I should probably run an equivalent line of mdeg models but haven't tried for N410 just yet.
• Library Testing
  • It also appeared that the scatter from the different libraries was far less for N410 than for other galaxies. Again, more of a sign that the spectra here are not suffering from whatever issues (maybe masks?) that other galaxies have suffered from.

• N410 I think is more straightforward to make progress on than N57. Given how reasonable the fits are, their agreement with Irina, and the relatively small dependence on the adeg/mdeg/library used, we can probably continue processing this as normal (switching to the asymmetric binning scheme, MGE fitting, and then moving to models).

So given this, I'v fit the symmetric spectra with adeg=0 and adeg=-1, and have some diagnostics below compared to Irina's kinematics:

• Note that Irina's kinematics are not actually on BOX it seems -- again, maybe Emily has this on one of her hard drives -- but I was able to find a version of the kinematics here, in one of the old wikis that I can compare to.
• Note that the two cases below I've been calling Case A sym and Case B sym, where Case A uses adeg=0, and Case B uses adeg=-1.

Here's the spectra for Case A and Case B:

• Note that the Case B spectra are ever-so-slightly worse fit than the Case A spectra:

• I don't have access to Irina's preliminary moments it seems; the only data that are on the Github seem to be just before running models, so these are after doing 100 MCs and taking the median:

• And again as a quick reminder, I can't find a fiducial fit from Irina anywhere so these compare the MC moments (without errors).

• Note that I've also computed the "robust sigma" of the residuals, and find: [5.016, 5.659, 0.011, 0.01, 0.011, 0.014] for each moment.

• There does look to be a systematic "wiggle" though between the two template libraries. At least to me there seem to be ~5 noticeable peaks in the resdiuals here. Potentially this is a sign that we need a higher order polynomial here?