meeting 2025 07 07 n57 - JacobPilawa/TriaxSchwarzschild_wiki_6 GitHub Wiki
Context
- Following up on some minor loose ends from last time, including some more superspectra testing.
- Takeaways:
- I think the most major takeaway from the testing below is that the superspectrum test (after appropriately removing some of the "strange" spectra) appears to be quite consistent with the fiducial Barth results, with many of the systematic differences being greatly mitigated or non-existent.
- The results for other superspectra templates (such as including only F/G/K templates alone) appears to support what we've seen before, and using the set of FGK stars alone also leads to somewhat decent agreement with the Barth results.
- Moving forward:
- I think we can probably use the superspectrum results to justify/support using the Barth templates alone (since I think this is what was done for H15 or M87, where the superspectrum tempaltes were used to motivate the final set of template chocies).
Superspectrum Testing
-
I've processed the superspectrum with a few different libraries and have diagnostics for this below. I then tried to refit the N57 spectra with the non-zero templates from each of the cases. The various cases I've processed are:
- Full Library = All 706 stars in CaT Library
- Barth only = 15 stars from Table 2 of Barth
- 384 Trimmed + 3 Missing = Set of "cleaned" templates Emily had passed me + 3 Barth stars NOT in that set
- F only = only considering stars of spectral type "F" from alldata.txt
- G only = only considering stars of spectral type "G" from alldata.txt
- K only = only considering stars of spectral type "K" from alldata.txt
- GK/FK/GF Only = Only using the GK/FK/GF spectral types from alldata.txt
- FGK Only = Only using the FGK spectral types from alldata.txt (the archived data file associated with CaT library)
-
First, here's the resulting fits to the superspectrum for each of the cases:
| Full Library | Barth Only | 384 + 3 | F Only | G Only | K Only | GK | FK | GF | FGK |
|---|---|---|---|---|---|---|---|---|---|
| [images/250707/full_library_pPXF_plot_00000-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/barth_only_pPXF_plot_00000-1.png) | [images/250707/trimmed384_plus_3missing_pPXF_plot_00000-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/F_only_pPXF_plot_00000-1.png) | [images/250707/G_only_pPXF_plot_00000-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/K_only_pPXF_plot_00000-1.png) | [images/250707/GK_pPXF_plot_00000-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/FK_pPXF_plot_00000-1.png) | [images/250707/GF_pPXF_plot_00000-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/FGK_pPXF_plot_00000-1.png) |
Weights
- Fitting the superspectrum with the various libraries defined above picks out the following templates for each case:
| Full Library | Barth Only | 384 + 3 | F Only | G Only | K Only | GK | FK | GF | FGK |
|---|---|---|---|---|---|---|---|---|---|
| [images/250707/full_library_prelim.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/barth_only_prelim.png) | [images/250707/trimmed384_plus_3missing_prelim.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/F_only_prelim.png) | [images/250707/G_only_prelim.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/K_only_prelim.png) | [images/250707/GK_prelim.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/FK_prelim.png) | [images/250707/GF_prelim.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/FGK_prelim.png) |
- And if I refit the N57 spectra with these templates only, I get the following set of 1-to-1 plots/differences:
| Full Library | Barth Only | 384 + 3 | F Only | G Only | K Only | GK | FK | GF | FGK | |
|---|---|---|---|---|---|---|---|---|---|---|
| One-to-One | [images/250707/250707_barth_vs_superspectrum_full_library.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/250707_barth_vs_superspectrum_barth_only.png) | [images/250707/250707_barth_vs_superspectrum_384_plus_3.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/250707_barth_vs_superspectrum_F_only.png) | [images/250707/250707_barth_vs_superspectrum_G_only.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/250707_barth_vs_superspectrum_K_only.png) | [images/250707/250707_barth_vs_superspectrum_GK_only.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/250707_barth_vs_superspectrum_FK_only.png) | [images/250707/250707_barth_vs_superspectrum_GF_only.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/250707_barth_vs_superspectrum_FGK_only.png) | |||||
| Differences | [images/250707/250707_diff_barth_vs_superspectrum_full_library.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/250707_diff_barth_vs_superspectrum_barth_only.png) | [images/250707/250707_diff_barth_vs_superspectrum_384_plus_3.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/250707_diff_barth_vs_superspectrum_F_only.png) | [images/250707/250707_diff_barth_vs_superspectrum_G_only.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/250707_diff_barth_vs_superspectrum_K_only.png) | [images/250707/250707_diff_barth_vs_superspectrum_GK_only.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/250707_diff_barth_vs_superspectrum_FK_only.png) | [images/250707/250707_diff_barth_vs_superspectrum_GF_only.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/250707_diff_barth_vs_superspectrum_FGK_only.png) |
pre-meeting question from Chung-Pei
- Just prior to the meeting, CP had asked if I had tried removing the strange templates from the top 15/384 superspectrum case and rerun it for each bin. I've put together that test here, but I think it's slightly different (but similar in spirit) to what CP was asking. As a quick reminder, the superspectra tests I've been doing have involved fitting the superspectrum with some large number of templates, finding which of those templates receieve any weight, and refitting the N57 bins with only the templates that have recevied weight.
- In the last meeting, I fit the superspectrum with the set of 384 cleaned from Emily + 3 missing Barth templates, and this resulted in 11 templates being given weight:
- (filename, weights) --> Spectral Type
- (scan423.fits, 0.185) --> Spectral Type = K
- (scan554.fits, 0.146) --> Spectral Type = K
- (scan670.fits, 0.118) --> Spectral Type = F
- (scan638.fits, 0.103) --> Spectral Type = K
- (scan488.fits, 0.081) --> Spectral Type = K
- (scan262.fits, 0.079) --> Spectral Type = F
- (scan671.fits, 0.066) --> Spectral Type = G
- (scan470.fits, 0.050) --> Spectral Type = M
- (scan387.fits, 0.034) --> Spectral Type = M
- (scan628.fits, 0.028) --> Spectral Type = B
- (scan685.fits, 0.000) --> Spectral Type = M
- I would argue that scan685, scan628, scan387, and scan470 all look quite "strange," so I've removed them from the set of 11 templates and refit the N57 bins:
- In the last meeting, I fit the superspectrum with the set of 384 cleaned from Emily + 3 missing Barth templates, and this resulted in 11 templates being given weight:
| One-to-One | Histogram of Differences | Apoly distribution vs. Barth |
|---|---|---|
| [images/250707/250707_barth_vs_superspectrum_7_of_11.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/250707_diff_barth_vs_superspectrum_7_of_11.png) | images/250707/test250708_7_of_11_superspec.png |
- One interesting thing is that there are now two bins that have extremely large sigma values (somewhat hidden behind the panel above). The spectral fits for the 7/11 and fiducial barth case are here:
| Case | 00023 | 00146 |
|---|---|---|
| Barth | [images/250707/bpPXF_plot_00023-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/bpPXF_plot_00146-1.png) | |
| 7/11 Superspec | [images/250707/pPXF_plot_00023-1.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/pPXF_plot_00146-1.png) |
- It seems like, to me, this is our "best case" that we've seen so far. It appears as though the Barth results are pretty much consistent with the superspectrum results, assuming we trim out the M and B stellar spectra (which already receive the lowest weight from the superspectrum fit).
Some of the more deviant cases
- I tried to spot check a few of the cases that had large polynomial degrees/stood out from some of the comparison plots. Here are some notes from these cases:
- 00183_prelim.p --> in the 384/384+3 libraries, this has the most extreme additive polynomial of ~-0.75, but notably this can bounce around a bit. For example when using the 12 overlapping templates, the additive polynomial flips positive to ~0.183.
- 00146_prelim.p --> in the 7/11 superspectrum test just above here, this is the lowest adeg, again at ~-0.75 or so. There does not appear to be anything peculiar about this spectrum to my eye.
Difference in Apoly vs. Difference in Moments
- As a slightly different test, one thing we wanted to see was how the adeg and moments are trading off for cases where we have adeg=0 for both sets of kinematics. This is slightly different from what we had been doing, in which we were looking at difference in apoly vs. difference in moments for adeg=0 vs. adeg=-1.
- So for this, I went back to the full fits (rather than using just the subset from the superspectra), and have some diagnostics below:
| Barth vs. Barth Reprocessed | Barth vs. 209 Trimmed | Barth vs. 384 Trimmed | Barth vs. 12 Overlapping | Barth vs. 384 Trimmed + 3 Missing | Barth vs. 3 Missing |
|---|---|---|---|---|---|
| [images/250707/TODAY_250630_barth_reprocess_diffs.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/TODAY_250522_adeg0_trimmed_diffs.png) | [images/250707/TODAY_250630_384trimmed_diffs.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/TODAY_250630_12overlapping_diffs.png) | [images/250707/TODAY_250630_384trimmed_plus_3barth_diffs.png]]](/JacobPilawa/TriaxSchwarzschild_wiki_6/wiki/[[images/250707/TODAY_250630_3missing_diffs.png) |