June 10, 2019 at 9:33 pm #3235
I am reducing the recently observed GMOS-S B600 IFU data in one-slit mode with Hamamatsu CCDs, and I noticed that the sky lines are shifted by a significant amount in the final data cube without sky subtraction. Specifically, [OI]5577 has redshifted by ~3.9 A or ~200 km/s on average, [OI]6300 has redshifted by ~4.65 A or ~219 km/s on average, [OI]6364 has redshifted by ~4.73 A or ~223 km/s on average.
I checked the wavelength calibration many times and don’t see anything wrong. I played with the different parameters of gswavelength and identified the lines interactively, but still nothing changed. The CuAr arc was taken 13 days after the science observations, and I am now worried this may be the cause of the shift I see in the sky lines.
So I wonder what I should do to get the right wavelength-calibrated data I need. Has anyone met similar problems before? Suggestions are more than appreciated! 🙂June 12, 2019 at 9:27 pm #3237
An arc from 13 days later should still be usable, but you cannot expect its zero point to match the science data precisely if it was not taken as part of your night-time observation (even on the same day). Even when pointing continuously at the same target, you are likely to see noticeable flexure within a couple of hours.
The right thing is to refer to the sky lines for zero-point correction, as you have done. Often you can just go through the reduction process as far as “gftransform” and then edit the CRVAL1 keyword in the SCI header for each exposure (eg. using “hedit”) so that the final WCS matches your expected line wavelengths.
Of course you may additionally want to make some heliocentric / barycentric correction to those observed wavelengths. If your observation happens to be split over multiple epochs, you might need to do that before you can stack all your data accurately (in which case you won’t be able to use sky lines to verify the final alignment).
The errors you mention do seem a bit larger than I’d expect. My recollection is that the grating re-alignment is supposed to happen to an accuracy of <=6 pixels and that flexure on the telescope is of the order of a couple of pixels, so perhaps it wasn't quite as good as normal, but it sounds like that arc should still be usable. Do double check, however, that it's really the best matching one available and you aren't overlooking a closer one (often there are several). Cheers, James.June 12, 2019 at 9:33 pm #3238
You might also like to refer to these tutorials:
The second one discusses a bit about flexure correction.June 13, 2019 at 7:57 pm #3239
Thanks a lot for your quick response in detail!
Yes, the shifts in the sky lines are quite significant. Also the shifts are wavelength dependent and fiber dependent, which makes it not just a pure zero point shift. I planned to do a linear fit to get the wavelength dependence of the shifts, but I was worried since I only have those three sky lines.
Do you feel that I need to check if it is really a zero point shift in the physical pixel? If you think so, how should I do that?
You must be logged in to reply to this topic.