Reference stars for flux calibration

[Paolo Berardi]

Sorry, another post about reference stars...

We all observe one or more reference stars in order to calibrate the relative flux of the target spectral profile. I always wondered how reliable are, for this purpose, the spectra which are included in the libraries like Miles, Elodie or Indo-US.

There are cases where spectra of the same star are contained in more than a library, so I tried to compare them. I expected to find more agreement between the spectra, there are indeed several cases in which the continua deviation is not negligible (also considering not or poor reddened starlight - info based on Miles catalog). I don't know why (changes that occur over time?). Follow some examples between Miles and NOAO Indo-US.







The HD114330 profile matches very well above ~4100A, below this value a large deviation becomes evident:


It's known that Indo-US spectra are dereddened (https://www.noao.edu/cflib/continuum.html). So, we should avoid any reddened star with this catalog. An example shows the problem (note this NOAO profile come from the original library, it is not present in the ISIS package):


A comparison adding Elodie 3.1 profile. The Elodie profile deviates a lot below 4000A:


In the HD086986 spectra, Elodie 3.1 profile follows the Miles one until ~4000A, below this value the intensities suddenly rise overcoming all spectra.


What do you think about?

I've done a cross-checks observing some Miles stars (Lhires III and 150 l/mm grating, spectra cropped from ~4100A) and calibrating other observed spectra with a response curve obtained by a Miles star. A comparison with the library spectra appears to confirm a good agreement (except for one case). But it was a small stars sample and I have not done the same for Elodie or Indo-US profiles. Has anyone done a similar pratical experience?

It seems clear that, to follow for example the spectral evolution of a spectroscopic target, the safest thing is to choose a reference star for all observers as soon as possible (before starting any observation is the best way). Anyway I think it would be very useful to know which star/profile of library spectrum is valid for the relative flux calibration purpose. This should lead to have comparable results regardless of the adopted reference star. Is this a dream or are we too much demanding?

I read about professional spectrophotometric standard stars (some are included in the ISIS package). Unfortunately they are few and a lot of them are very faint for amateur equipments (I think).

Have a clear skies!
Paolo

PS: sorry for my bad English...

[Paul Luckas]

Hi Paolo,

Very interesting (and not totally unexpected).

Note also that spectral types are also often different. Eg, HD206165 is listed as a type-B3I in NOAO. In Miles it's a type-B2Ib.

I guess the best we can do is detail all of the sources used in an analysis - particularly if we intend to publish - together with an error analysis. As an example, a search in SIMBAD lists spectral type of a star with a link to the source paper.

I've done a cross-checks observing some Miles stars (Lhires III and 150 l/mm grating, spectra cropped from ~4100A) and calibrating other observed spectra with a response curve obtained by a Miles star. A comparison with the library spectra appears to confirm a good agreement (except for one case). But it was a small stars sample and I have not done the same for Elodie or Indo-US profiles. Has anyone done a similar pratical experience?

I've done a few of these 'tests', and generally get excellent agreement. I can send a link to some examples in another post if you're interested.

I guess combining spectra from multiple contributors using various standards is another can of worms.

Paul

[Paolo Berardi]

Hi Paul, I'm also confident that Miles spectral DB is reliable enough (though it wasn't made for flux calibration). The group of Miles stars I observed for a test (9 stars) is ok. I realized that the strange result I had depends on the star not high in the sky and I didn't meet the parallactic angle. Indeed, profiles deviation begins below a certain wavelength in the blue domain (my guide camera centers on slit the red part of the stellar image).

You are right concerning spectral types, I saw many different classifications in the Vizier (Catalogue of Stellar Spectral Classifications). For example the star HD198001 ranged from A2V to B9.5V (actual Simbad spec_type). The star is also a spectrophotometric standard used by ESO:

https://www.eso.org/sci/observing/tools ... r7950.html

Both Miles and ESO classified it A1V and their continua (relative flux) doesn't seem very different (but not identical):



Anyway the classification is based on spectral lines and, in the absence of effective variations over time, I think we should expect similar continua of the same star if the observed spectra are corrected only for instrumental response and atmospheric transmission.

I've done a few of these 'tests', and generally get excellent agreement. I can send a link to some examples in another post if you're interested.

Thank you for your time Paul, I think it would be very useful to point out any "strange" case, I mean DB stars that it's better to avoid for the flux calibration purpose. If you happen, please let us know (me too).

Paolo

[Robin Leadbeater]

Hi Paolo,

The Indo-US spectra are not reliable as reference stars. The continuum shape is not taken from the original star spectrum. it is synthesised using Pickles spectra. See page 10 of the original paper
https://www.noao.edu/cflib/Valdes.ps

Cheers
Robin

[Robin Leadbeater]

Paul,

Published spectral types are notoriously unreliable and only give a very rough idea of what the spectrum continuum might look like. See Skiff's huge database (>800000 objects) of all published spectral types for examples where the same star has been classified differently.
http://vizier.u-strasbg.fr/viz-bin/VizieR?-source=B/mk

Only accurately flux calibrated spectra are reliable as reference stars. The professionals use reliable specially calibrated flux standards to determine the instrument response and then measure the atmospheric extinction separately using spectrophotometry and a model of the atmosphere.

Robin

[Robin Leadbeater]

My view is we should include in the fits header, the reference star used and a reference to the spectrum used. Provided the star is not variable, spectra reduced using different reference stars can then be compared between observers and corrected if necessary if it is found that a reference spectrum is not correct. This system is used in the BAA variable star database for example (but not in the AAVSO database). I think there have been some discussions about this between Francois Teyssier and Christian Buil. See this post
http://www.spectro-aras.com/forum/viewt ... 1572#p7210

Robin

[Robin Leadbeater]

Paolo,

I agree that ELODIE 3.1 calibration is poor below 4000A. The original ELODIE library stopped at 4100A. I think we should limit the range we use to above this wavelength.
See http://www.spectro-aras.com/forum/viewt ... 1572#p7226

Robin

[Joan Guarro Flo]

Grazie Paolo,

Your complaint has been very clarifying and necessary. We need good and reliable comparison charts of stars.

It is essential that mark a star standard for each object observed.

Un abbraccio, Joan.

[Paolo Berardi]

Thank you all for this interesting discussion. I hope that in future it will be possible to integrate the name of the reference star in the target profile header.

Paolo

[Peter Somogyi]

Just to add my thought, I mostly use 2 reference stars (mostly the same): at begin and end.
I do not trust only 1 reference star, any estimate can easily fail just due to any extra air pollution here (e.g. fossile heating).

Best would be, to add support in ISIS to linearily approximate using 2 ref.stars. I know I'm asking much, but have to think about the ideal future case.
2 ref.stars (especially if they're different stars) also lets to define and/or discover errors. (But a 3rd ref.star would allow to fix such an error.)

I've recently bought a photometric slit, now I say thanks about telling not to use any reference wavelength below 4100A.

- Peter