Roadmap and science questions for interactive spectral analysis

[gully]

We now have a working interactive spectral analysis tool intuition.py built on bokeh. This tool will allow us to get "by-eye" best fit parameters and build an intuition for the spectra, either in-place-of or in-advance-of a full-on spectral inference procedure. Here are the science questions I brainstormed with a coarse classification for how easy/hard the task is.

🟢 = Easy
🟡 = Takes some work
🔴 = Research Project

1. 🟢 Do the raw spectra show variations conspicuous to the human eye? If so what are the wavelength locations of the largest such perturbations?
2. 🟢 What is the overall performance of the custom cloudy Sonora models?
3. 🟡 What is the best fit temperature and surface gravity? To what extent do these values vary with spectral order (i.e. wavelength/species)?
4. 🟡 What is the overall character of the spectral variations? Is there evidence for line profile perturbations?
5. 🟡 With what filling factor of clouds is each component most consistent?
6. 🟡 To what extent are the pairs of parameters degenerate, for example resolution and surface gravity, or spectral resolution and filling factor, or resolution and filling factor? Are there discrete spectral features or groups of features that break these degeneracies?
7. 🔴 Can the heritage of spectral perturbations be convincingly assigned to astrophysical variation and not, say, telluric correction artifacts or instrumental/wavelength-dependent slit-loss?
8. 🔴 Is there conspicuous evidence for two distinct spectral components (cloud and cloud-free) coexisting in the single composite spectrum?
9. 🔴 How do the observed spectral changes relate to the instantaneous TESS flux (available April 9)?
10. 🔴 Does the overall pattern of best fit values comport with theoretical expectations? Overall what do we learn from the spectra?

[gully]

1. Do the raw spectra show variations conspicuous to the human eye?

Yes, the spectra show variations perceivable to the human eye. Conspicuous is relative here, and the real goal should be to report the variability quantitatively--- percentage departures from the mean spectrum, or a mean spectrum and variance spectrum.

I may make an annotated table of the most conspicuous perturbations directly in a LaTeX table in the paper draft. We may-or-may-not keep that table in the paper, but it seems like an adequate holding place while we work out the shape of the results.

[gully]

2. 🟢 What is the overall performance of the custom cloudy Sonora models?

Overall the cloudy models fit much better than the cloud-free models. The fit is better for both components A & B, and essentially all echelle orders in K-band.

[gully]

3. 🟡 What is the best fit temperature and surface gravity? To what extent do these values vary with spectral order (i.e. wavelength/species)?

I tabulated the best fit temperature and surface gravity for all K-band echelle orders from 2.0 to 2.44 micron.

Component B shows Teff of 1600 K for all but two orders: 1300 K at 2.1 micron and 1700 at 2.23 micron. All orders were consistent with 4.5 log g. All orders were consistent with 100% cloudy, except for CO at 2.3 micron, which was consistent with a range of filling factors from 0.69 and up.

Component A was Teff of 1600 K, log g of 4.5, and 100% filling factor of clouds across the board.

The best fit resolution bounced around a lot, probably due more to the inadequacy of the models and smoothing kernel than actual physics, but maybe there is something there.

The grid of Teffs only include 1000, 1300, 1600, 1700 and up, so the convergence to 1600 may be artificial, with unrealized better fits in between 1300 and 1600.

[gully]

4. 🟡 What is the overall character of the spectral variations? Is there evidence for line profile perturbations?

Overall the spectra changes are very subtle. They are mostly line profile distortions, or maybe ever-so-slight line depth changes. There is evidence for line profile perturbations, but we have not quantified them rigorously.

[gully]

5. 🟡 With what filling factor of clouds is each component most consistent?

💯

[gully]

6. 🟡 To what extent are the pairs of parameters degenerate, for example resolution and surface gravity, or spectral resolution and filling factor, or resolution and filling factor? Are there discrete spectral features or groups of features that break these degeneracies?

Generally the parameters are not too degenerate. Surface gravity and filling factor of clouds are somewhat degenerate, but the observed spectra require both high gravity and high coverage of clouds, so we max out both log g and f.

It would be nice to have finer Teff resolution to answer this question more carefully.

[gully]

8. 🔴 Is there conspicuous evidence for two distinct spectral components (cloud and cloud-free) coexisting in the single composite spectrum?

No, not really. The spectra are almost all maxed out at 100% cloudy. The difference between 98% and 100% cloudy is nearly indistinguishable by eye, so we would want to build some metrics to get to cloud coverage differences at this level.

[gully]

7. 🔴 Can the heritage of spectral perturbations be convincingly assigned to astrophysical variation and not, say, telluric correction artifacts or instrumental/wavelength-dependent slit-loss?

Hmm, this is a hard question. I am not certain at the moment. I think there are genuine perturbations to the line profiles, but telluric residual artifacts are common and difficult to remove.

One step forward could be the following seemingly pathological strategy: Rather than dividing by an A0V calibrator, simply divide component B by component A. They share the same airmass and instrumental flexure, so they should work as much better calibrators for each other. Then either multiply back by a pre-trained model of A, or simply doctor the model comparison in the same way: with ratios of two models of different temperature and cloud coverage and vsini. The current problem is that our Teff resolution is too coarse for models, so maybe just a data-driven approach would suffice? There's the problem that you don't know the absolute normalization between the two components. You would know this value if you did careful echellogram modeling with the binary axis aligned with the PA.