meeting 2025 03 18 n57 - JacobPilawa/TriaxSchwarzschild_wiki_5 GitHub Wiki
Context
-
Still working on some refits to the N57 spectra. In this bullet, I just wanted to get a decent set of fits using the old pPXF fitting code since the new version was giving me some issues yesterday that I still need to figure out.
-
As a quick reminder, there are two major changes between these new sets of fits and the old sets of fits -- the old fits had been using the off-by-one set of Barth stars and were also using bias = 0.0.
- The two sets of fits below correct this, and instead use the correct Barth stars library and a bias of 0.2.
- The only difference between the two fits below are the choice of degree for the additive polynomial. "Case H" has adeg=-1, whereas Case I has adeg=0. The original N57 fits used adeg=0, and you can tell in the results below that Case I reproduces the original kinematics much better, despite the spectral fits not being appreciably different between Case H and Case I.
-
Takeaways:
- I think we can move forward with Case I (correct Barth stars + bias=0.2 + adeg=0) since this seems to provide really great spectral fits while still mostly agreeing with the original kinematics which used the incorrect template library.
- As a quick reminder, we had lost the N57 orbit libraries a couple of years ago since they had been in the scratch directory which was getting purged (rather than the projects direcotry), so we'll have to rerun some models to reminimize this new set of kinematics.
- Here's a bullet which has a summary of the various cubes/chi2 panels we had up to now. Since I'm not entirely sure how things will shake out, I think it might be worth starting by rerunning Cube A1 to ensure things roughly track each other. Depending on how those chi2 values look, we might be able to jump right to a later cube or simply add Cube A2 + scalings on both.
- Expanse's lustre filesystem is currently down, but I'll start setting things up to run models + minimize once everything is back. They estimate it will be ready for use by tonight or tomorrow morning.
Diagnostics
- First, here's a comparison of the old kinematics (without any point symmetrization) against the two cases presented at the bottom of the page. As I had mentioned above, the adeg=0 case is in much better agreement (and is more consistent in method) with the original kinematics than the adeg=-1 case.
- One quick note -- I did median subtract the v's for both sets of kinematics for comparison here, but none of the actual point-symmetrization has been done.
| Origianl vs. Case H (adeg=-1) | Original vs. Case I (adeg=0) |
|---|---|
| [images/250318/quick_check_CaseH.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250318/quick_check_CaseI.png) |
A few select spectra with polynomials plotted
- Just a quick update following some email exchanges -- here are 10 random spectra (consistent across both cases) where I have the additive and multiplicative polynomials shown here for comparison.
| Case H (adeg=-1) | Case I (adeg=0) |
|---|---|
| [images/250319/CaseH/pPXF_plot_0-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250319/CaseI/pPXF_plot_0-1.png) | |
| [images/250319/CaseH/pPXF_plot_1-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250319/CaseI/pPXF_plot_1-1.png) | |
| [images/250319/CaseH/pPXF_plot_2-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250319/CaseI/pPXF_plot_2-1.png) | |
| [images/250319/CaseH/pPXF_plot_3-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250319/CaseI/pPXF_plot_3-1.png) | |
| [images/250319/CaseH/pPXF_plot_4-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250319/CaseI/pPXF_plot_4-1.png) | |
| [images/250319/CaseH/pPXF_plot_5-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250319/CaseI/pPXF_plot_5-1.png) | |
| [images/250319/CaseH/pPXF_plot_6-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250319/CaseI/pPXF_plot_6-1.png) | |
| [images/250319/CaseH/pPXF_plot_7-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250319/CaseI/pPXF_plot_7-1.png) | |
| [images/250319/CaseH/pPXF_plot_8-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250319/CaseI/pPXF_plot_8-1.png) | |
| [images/250319/CaseH/pPXF_plot_9-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250319/CaseI/pPXF_plot_9-1.png) |
All Spectra
- And here's side by side comparison of the new spectra fits themselves. Both sets of fits look very nice, with none of the MC realizations as being flagged as being in disagreement with the mean as they get processed. As you'll see in the comparisons below, the adeg=-1 has quite elevated sigma relative to the old set of fits despite the fits looking normal, so I think it'd make more sense to stick with adeg=0 of Case I.
- As a quick reminder from above, these fits use:
- Corrected Barth Stars
- Adeg = -1 (Case H) and Adeg=0 (Case I)
- Bias = 0.2
| Case H (adeg=-1) Spectral Fits | Case I (adeg=0) Spectral Fits |
|---|---|
| [images/250318/spectra_CaseH-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_5/wiki/[[images/250318/spectra_CaseI-1.png) |