meeting 2025 06 26 n410 - JacobPilawa/TriaxSchwarzschild_wiki_6 GitHub Wiki
Context
Note this is still a bit of a work in progress as of 6/26 afternoon. Adding some more tests/thoughts to what is below.
Adding some tests following the meeting for N410, including tests on the template library, varying the bias, varying the additive and multiplicative polynomial, and running the sigma clipping routine on the data.
Templates:
The new set of 348 templates does not seem to appreciably "fix" the template issues we had been seeing before. In fact, limiting the template library to the 348 templates from earlier this week gives worse RMS agreement when compared to the old "trimmed" library of ~209 stars Emily had passed before.
Despite this, the RMS difference when swapping out the templates is smaller than it was for N57 and does not appear to have the same magntiude of systematic biases (such as in h4) as we had seen in other galaxies. I think this is generally consistent with how robust the spectra have been to other changes (like the bias or sigma clipping that I have diagnotics for below).
I've included a version where I plot the kinematics from the 209 vs. 348 templates against one another, and at least here, it appears like the moments are in very nice agreement, with the exception that the sigma for the 209 stars seems to be ever so slightly inflated relative to the 384 stars. This goes in the right direction that the 209 stars are more consistent with the Barth subset alone (compared to the 348 tempaltes).
Polynomials:
Multiplicative Degree: it seems like this multiplicative degree again needs to be something like at least 2 or above for the results to be reasonable. In many of the bins, there is hardly any variation beyond mdeg=3 (for adeg=0).
Additive degree: I'm seeing a similar trend that we saw with the other galaxies for N410, where adeg=-1 gives elevated sigma values relative to the other adegrees. If we continue to increase adeg from adeg=0 to adeg=4 or so, the results are somewhat robust (there is a bit of variation here), but nothing too extreme I think.
At least from what I'm seeing, without anthing "better" to go off of, I think fixing adeg=0 should be alright for N410, too. It seems like no matter which polynomial order we are to pick, we are going to have some disagreement for a subset of the bins, so I don't think there will be a perfect solution here.
Sigma Clipping:
The sigma clipping (for reasonable sigma = 2.0, 3.0) seem to have very little effect on the resulting kinematics compared to our fiducial case. In particular, from a quick glance, it seems like the spectra are of high enough quality that the sigma=3.0 case is hardly picking up new areas to mask. Even if we go with the a more conservative sigma=2.0, the kinematics are still in very nice agreement with the "base" set of kinematics.
Bias:
The bias similarly has essentially 0 effect on the kinematics. Note that I'm using the base settings (adeg=0, mdeg=3, Barth templates, etc)., and bias=0.0 and bias=0.2 are essentially exactly the same.
Diagnostics
Polynomial Degress
Varying Adeg (mdeg=3)
Varying dmeg (adeg=0)
I quickly made a version of the mdeg test zoomed to the same y-range as the "varying adeg" test for our reference:
Varying Mdeg Zoomed
Template Testing
I'm find that the 209 vs. 348 set of templates seem to be about the same in terms of ability to recover the Barth results (if you're really trusting the RMS, the 209 stars actually appear to agree a bit better than the 348 templates). With that said, this seems to be in general a better agreement than what we see with N57.
Old "Trimmed" Library (209 stars)
New "Trimmed" Library (348 stars)
I also have a version where the y-axes here are plotted against one another:
209 vs. 348 Trimmed Libraries
Sigma Clipping
Note that the y-axes here are remnants from creating some of the other plots on the page. The cases here use identical fiducial settings, the only thing being changed is whether or not I'm sigma clipping.
It appears like the sigma clipping has little impact on the kinematics.
Sigma Clipped = 2.0
Sigma Clipped = 3.0
And here are the fits to the spectra themselves so you can get a sense of how many new spots are "masked" with the sigma clipping routine.
No Sigma Clip
Sigma Clipped = 2.0
Sigma Clipped = 3.0
Bias
I've also processed our fiducial set of kinematics with both bias = 0.2 (default) and bias = 0.0, and there is virtually no difference in the kinematics for N410:
