Nu Gem at Low Resolution

[AndyWilson]

Hi,

I've revisited Nu Gem but this time at low resolution using the 150 l/mm grating. I think I've done the corrections for instrument response and atmospheric extinction correctly by using the A2V star 26 Gem. I also flat fielded all the spectra.

I'd be interested in any comments good or bad, especially if anyone spots anything that does not look right. I used ISIS for the processing, then created the plot in BASS along with a tweak to the wavelength calibration using the hydrogen Balmer lines.

Clear skies,

Andy

[Robin Leadbeater]

Hi Andy,

Nice spectra with excellent signal/noise and clear H alpha emission in nu Gem. I am not sure about the shape of the Nu Gem continuum though. The intensity looks high at the red end where it seems to be levelling out well above zero. If Nu Gem is typical of a B6 star I would have expected it to have been slightly lower than the A2 star at the red end and slightly higher at the blue end, being a bit hotter. A dark or background subtraction problem somewhere perhaps ? (Small zero errors at the blue and red end where the sensitivity is low can make a big difference after response correction)

Cheers
Robin

[AndyWilson]

Hi Robin,

Thanks for the feedback. The offset was down to me. I shifted the vertical axes of the spectra to avoid any overlap. Normally I remove the y-axis scale but forgot to this time. It could be that my instrument/atmospheric response correction is a little off. This was my first attempt at low resolution correction in ISIS and it took quite a bit of trial and error.

Here is a plot with the offsets and scales identical for both stars. I've not done any flux correction so I don't know how meaningful this is. Both spectra are 600 second integrations. 26 Gem is about a magnitude fainter than Nu Gem though a bigger effect could be how well I positioned each star over the slit. I still find accurate positioning over the slit quite difficult, I just move things back and forth until I think the signal is at maximum.

Cheers,
Andy

[Robin Leadbeater]

Hi Andy,

Yes that looks more sensible. You can really only do a rough absolute flux calibration with a narrow slit as you can never be certain you are getting the same percentage through the slit for each measurement (positioning, tracking, seeing, focus etc). For accurate results, you need to take spectra using a wide slit as well so you catch all the light and then scale the narrow slit results . (You also need photometric skies so everything stays constant between target and reference measurements) Christian runs through the techniques here
http://www.astrosurf.com/buil/calibrati ... ion_en.htm
Alternatively you can use photometric data to scale the spectra as David Boyd does for example.

Fortunately though for most applications it is sufficient just to scale spectra relative to the flux in the continuum at a given wavelength (H alpha for example or some middle wavelength, 5500A say), and unless you are confident in in the quality of your absolute flux calibration, I would say this is probably the best way to go. If you do this for your two spectra you can then compare the continuum shapes directly.

There is also a low resolution spectrum of nu Gem in the BeSS database from 2013 by Francois Cochard using an ALPY 600 which might be interesting to compare with your result.

Cheers
Robin

[AndyWilson]

Hi Robin,

Thanks for the pointers.

I've compared my spectrum with Francois Cochard and mine definitely is growing too steeply to shorter wavelengths. My guess this is something up with my response curve. I've tried a few things without any significant improvement thus far.

I have noticed that my wavelength calibration does not look quite right. Not an overall shift to high or low wavelength, but a small gradual effect to shorter wavelengths. It is pretty much OK at H-Alpha, but gets gradually worse as I move through H-Beta, H-Gamma, etc. I used the neon lamp so that would explain the calibration being good near H-Alpha. Do you think an imperfect wavelength calibration could cause the response correction to go wrong? I am wondering if this is causing the climb of the curve to shorter wavelengths to be too high.

I'll keep at it. It is quite possible I'm not using ISIS quite right for this processing. I followed the tutorial for the LHIRESIII at high resolution. Then I read the LISA tutorial to try to work out how to process low resolution spectra. It wasn't quite applicable but I thought I picked up enough tips to do it right. So I may just be making a basic mistake somewhere in my processing.

Thanks for your help,

Andy

[Robin Leadbeater]

Re wavelength calibration, yes calibrating with just the strong neon lines will give a deteriorating fit towards the blue end. I dont use my LHIRES at lowest resolution (There are people here who do though eg Paolo Berardi who would probably have a solution) but an alternative could be to combine them with the Balmer lines in the reference star to give a better fit at the blue end as is done somewhere in the tutorials for the LISA for example.
(The new update to the LHIRES with the NeAr lamp used in the ALPY solves this problem)

I dont think this would affect the response correction significantly though, particularly if the wavelength error is in both reference and target spectra

There a quite a few potential places where errors can creep in in response correction. (Francois Cochard's spectrum might not even be spot on ) You have covered some eg using a ref star close to the same altitude as the target. Some other subtle ones to watch for are:-

Is the catalogue spectrum of your reference star reliable ? If it is an actual spectrum of the star eg like the MILES stars then it will be but if you are using a generic eg Pickles spectrum then it might not be identical to the actual spectrum of the star used.

If there is any chromatism in your telescope optics then focus shifts between reference and target can change the proportion of light the slit samples different wavelengths and alter the relative shape of the spectra.

For targets at low altitude, atmospheric dispersion, if not in the slit direction, can cause similar problems, though if the target and reference are close in the sky then any effect should cancel

In processing, when smoothing the response take care to make sure the fit is still good at the extremes, particularly at the blue end where the response is generally steep. The test is to re-reduce the spectrum of the reference star using the instrument response you have produced and check it matches the catalogue spectrum (it should of course be identical )

A good way to see if your technique is robust is to take some spectra of a range MILES standards of different spectral types at similar high altitude so the effects of the atmosphere are minimised, and reduce them using the instrument response derived from one of them and see how close they are to the catalogued spectra.

Cheers
Robin

[Peter Somogyi]

When the IR end creeps up back at LoRes, for me it was typically because did not reject some of the 3 big infra telluric holes at ref.star calibration.
In ISIS the wizard for Alpy 600 does it automatically, but when had manual way or LHires work at infra, I have to manually select regions to be cut (Continuum => 2 x double click the interval - not a true reject, you just manually edit it as linear between 2 end points). Sometimes I had to 'Edit' the last pixel value when the telluric hole breaks just at the end of the detector. Some others chose such a high order continuum or filter that (almost) corrects the continuum of the telluric hole (I did see such spectra in the database quite often, question which approach is better).
I don't know how it is done in BASS.

HTH
Peter

[AndyWilson]

Robin & Peter,

Thanks for your useful pointers. It does feel like an instrumental or atmospheric effect that I am not taking into account. Dispersion from the telescope optics or the atmosphere sound likely candidates. Although it also occurred to me whether this is because I am using the LHIRESIII outside of its core design parameters, after all it is designed for high resolution work rather than low. I am mainly interested in high resolution, so not a problem, I am just experimenting with lower resolution to get better familiarity with the spectral sequence and have some fun at the same time!

I have reprocessed the spectra in a number of different ways and gone over various parameters that might affect the result. One thing that might be a big factor is that Nu Gem and 26 Gem were both relatively low at 30-35 degrees altitude with an air mass of 1.7-1.8. So this might well exacerbate any problem. Also, I used the Pickles standard in ISIS as the reference for 26 Gem, so as Robin points out that might give me a problem.

I won't list everything I've tried and checked, but the best result I achieved was by using Castor to calculate the response instead of 26 Gem. I imaged Castor, then Nu Gem, then 26 Gem. The air mass for Castor was 1.2, which is why I initially did not use it. Below shows the result, which is closer to Francois Cochard's and a reference spectrum from Pickles, though still clearly different.

One side benefit is this has given me lots of practice in processing spectra

Cheers,

Andy

[Paolo Berardi]

Hi all,

Re wavelength calibration, yes calibrating with just the strong neon lines will give a deteriorating fit towards the blue end. I dont use my LHIRES at lowest resolution (There are people here who do though eg Paolo Berardi who would probably have a solution)

Robin, with the new calibration module I'd like to find a better solution but, so far, I simply use some Balmer lines of an A0V (or similar) type star observed close to target to avoid shifts for mechanical flexures of spectrograph (just a check for any great radial velocity values involved), plus 6869A and/or 7605A atmospheric absorption lines, to find a 2nd or 3rd order polynom. Usually the RMS is around 0.7A. Then I use the calibration lamp frame to "sync" the spectrum on a isolated known line (ISIS "X coordinate of line at wavelength" function).

Anyway the blue end is "short" with my Lhires III 150... indeed, h-delta is pretty defocused for chromatic aberrations, so I think the spectrum should be cropped from 4100-4200A.

Concerning the problem referred by Andy, I also think the "prime suspects" are the response curve lowered in the NIR region due to telluric absorption not removed and a residual offset in the 2d spectrum used to extract the profiles (it may cause a raising of both blue and red part), as Peter and Robin said.

Andy, I wrote the above before reading your last post. In fact, the far red part of your profiles seems to me still strange. Castor is a very bright star and may be you have a less evident problem because the residual offset is lower in percentage of signal (a zero offset is important for both target and reference star processing).

The behaviour of Lhires III has the same impact on target and reference star, so I think you should have always a correct result with standard data reduction and observing procedures. But wait for more experienced people comments...

I currently have no good weather, else I would have observed nu Gem for a confirmation. I cannot exclude any intrinsec property!

Anyway, you are doing a very useful experience!

Clear sky!
Paolo

[AndyWilson]

Hi Paolo,

Thanks for your reply which contains useful and interesting information. I think you are right about cropping the spectrum at 4100-4200A. I am using an SXVR-H694 and so I get quite a wide spectral coverage, but it does mean the entire spectrum will not be in focus.

I've since thought of a further problem which might explain the spectrum. The skies get dark quite late here now. Though I did wait for darkness, re-checking my planetarium software I think full astronomical darkness was only reached part way through my imaging run. So perhaps the sky background and atmosphere were changing between my target and reference spectra, which would likely affect my result.

I'll have a go at imaging some higher altitude stars and later in the night. Though it is at about the time of year when I won't be able to image and get a full nights sleep in time for work the next day

Thanks for everyone's help, I've learned a great deal from this discussion.

Cheers,

Andy