Change page style: 

Atmospheric Differential Refraction in GMOS data

This page highlights the detrimental effect that atmospheric refraction has on GMOS data, with certain configurations being more susceptible than others. Strategies are given which can help the user to minimize the impact on data when preparing their observations.

Background Information

GMOS was designed to take advantage of the anticipated sub-arcsec seeing of the Gemini Telescope. The original instrument design has provisions for an atmospheric dispersion compensator (ADC) to improve the performance of GMOS in both spectroscopic and imaging modes. However because there was an expectation that GMOS would be used mostly in the red, the installation of an ADC was postponed. As a result, neither GMOS is equipped with an ADC and because Gemini is investing most of its resources in the commissioning of new instruments,  there are no plans to install such devices in the near future.

The absence of an ADC on either GMOS has several implications on short-wavelength imaging and, especially, on spectral data taken with the instruments. In this document we examine the effects produced by the absence of ADC, based in optical modeling of the effect of differential refraction on GMOS images and spectra, as a function of airmass. We also discuss which GMOS configurations are potentially affected and what we can do to minimize the effect on the data. Further details can be found in Kathy Roth's presentation to the Gemini staff and NGOs on December 10, 2008.

Differential Atmospheric Refraction on Mauna Kea (Model)

The differential atmospheric refraction relative to 500 nm (in arcsec) for different wavelengths as a function of the airmass is shown in Figure 1. This model is specific for Mauna Kea; for Cerro Pachon the effect is more pronounced. The model was derived using the IDL code written by Enrico Marchetti, ESO, that is available from the ESO public webpages. The parameters appropriate for Mauna Kea that were used are Temperature = 0 deg C, Relative Humidity = 14.5 %, Barometric Pressure = 615 mbar. From this figure one can see that at even the modest airmass of 1.2 the extreme red and blue wavelengths of GMOS sensitivity are displaced with respect to one another by over an arcsec. The blue wavelengths has a more pronounced displacement with airmass than the red.

Diff. Atm. Ref. at MK
Figure 1.

Image Quality

Differential atmospheric refraction effects are more pronounced at shorter wavelengths and under good seeing conditions. Figure 2 below shows the effect of differential atmospheric refraction (relative to 500 nm) on Mauna Kea (model) for a point source image with a disk seeing of 0.3" and at three different airmasses (1.05, 1.5 and 2.0) using the g' filter (wavelength range 398-552 nm). As can be seen in this figure, the model star looks strongly elongated, particularly at airmasses 1.5 and 2. In fact, under such good seeing conditions, g-band images are substantially elongated even at airmasses lower than 1.5.


Diff. Atm. Ref. at MK
Figure 2. 


At poorer seeing conditions the effect of differential refraction is somewhat masked. That can be seen in Figure 3 below, where model images computed assuming a seeing of 0.6" are shown. One can see that images look less elongated in this case, even though the effect is pronounced at airmasses 1.5 and above, and is still important even at airmasses below 1.5.


Diff. Atm. Ref. at MK
Figure 3. 


At red wavelengths the impact of differential atmospheric refraction is less important. This is shown in Figure 4 below for a model i'-band image of a point source with 0.3" seeing. This figure shows that in the far red the image elongation effects due to differential atmospheric refraction are negligible even at high airmass and in very good seeing conditions.


Diff. Atm. Ref. at MK
Figure 4. 


Spectroscopic Slit Losses

Differential atmospheric refraction can also have a strong impact on spectroscopic data taken with GMOS, particularly at short wavelenghts. This is illustrated in Figures 5 through 8, which show model estimates of slit loss as a function of airmass, slit width and orientation, and seeing. Figure 5 shows plots of fraction of light loss to due to differential atmospheric refraction as a function of wavelength, assuming that the star is centered on the slit at 500 nm. In each panel, different curves represent the fractional light loss for observations taken at different airmasses. The conclusion to be drawn is that light loss is obviously worse at higher airmass. Different panels in Figure 5 through 8 show the effect of varying the slit orientation relative to the parallactic angle. The farther the slit is oriented from parallactic angle, the worse is the loss of light, culminating of course when the slit is perpendicular to the paralactic angle.


Diff. Atm. Ref. at MK
Figure 5. 


Slit losses are obviously ameliorated when wider slits are used. The effect can be assessed by comparing the plots in Figures 5 and 6, where models for slits with 0.5 and 0.95 arcsec widths, respectively, are shown.


Diff. Atm. Ref. at MK
Figure 6 


Finally, the effect of seeing can be estimated by comparing Figures 7 and 8, where calculations adopting seeings of 0.3 and 0.9 arcsec are shown, leading to the conclusion that light losses are worse in better seeing conditions.


Diff. Atm. Ref. at MK
Figure 7. 



Diff. Atm. Ref. at MK
Figure 8. 

Another concern for spectroscopic observations is the effect of differential atmospheric refraction on MOS spectroscopy when the pre-imaging is obtained at a very different airmass than the spectroscopy. In Figure 9, we show the magnitude of differential refraction over the 5.5 arcmin field of view of GMOS for different wavelengths. As can be seen, the amount of differential refraction over the field varies significantly as a function of airmass. The wavelength of light also has an effect, being worse in shorter wavelengths, but this is a second order effect. For any wavelength, the differential atmospheric refraction across the GMOS field of view changes by about 0.2 arcsec between airmasses of 1 to 2. MOS masks which use the full field of view and which are designed with 0.5 arcsec or smaller slits from pre-imaging observed at an airmass ~2 and which are subsequently observed close to an airmass of 1 may incur significant slit losses. Any mask with slit widths < 1 arcsec may be affected.

Diff. Atm. Ref. at MK

Figure 9.

Observing Strategies to Mitigate the Effect of Differential Atmospheric Refraction

General principles:


  • For imaging, the only way of minimizing the effects is by observing at low airmass. Effects are worst under IQ20 conditions.
  • For spectroscopy, observations blueward of 450 nm are affected the most. Observers are advised to use the widest slit that science can tolerate, and to choose their central wavelength (effective guide wavelength) with care in order to minimize slit losses in the spectral regions of interest.
  • Long-slit observations can be collected with the slit oriented along the parallactic angle, but that may not be possible due to lack of guide stars.
  • Observers are encouraged to request elevation constraints appropriate for their chosen PA or alternatively design their longslit observations for mean parallactic angle. Please, check example in the GMOS libraries for how to design a mean parallactic angle observation. Any elevation constraints must be approved by the head of science operations at the appropriate site.
  • For MOS, slit position angles are fixed once the mask is designed. Observers should consider the parallactic angle when chosing position angle. They may need to resort to tilted slits combined with airmass constraints. Similar airmass constraints may be desired for the pre-imaging as well.


Default Position Angle Dependence onTarget Declination

As targets rise above the horizon their parallactic angle is aligned along or close to 90 deg (with a dependence on target declination and how much that differs from the latitude of the telescope position). As all targets transit their parallactic angle rotates through 0 deg. For targets which transit high in the sky (~ 25-30 deg from zenith) the slit losses due to atmospheric refraction are minimal if the position angle is set to 90 deg. This is because while the targets are above airmass 1.2 it is not as critical to have the position angle aligned along the parallactic angle since the displacement due to atmospheric refraction is small for most configurations, and when the target is below airmass 1.2 the parallactic angle has moved more than 45 deg away from 0 deg. This effect is demonstrated in Figure 10 below, where the dependence of parallactic angle on both hour angle and airmass are shown for several different target declinations. This plot is specific to Mauna Kea; however the effect is similar for Cerro Pachon where the main quantity to consider is the difference between the observatory latitude and the target declination. For reference, the latitude of Cerro Pachon is -30.24 deg, and that of Mauna Kea is +19.82 deg.

Parallactic Angle as a function of Hour Angle and Airmass for different Target Declinations (Mauna Kea)

Figure 10.

For this reason, the default position angle of all templates in the GMOS-N and GMOS-S Libraries have been changed recently (as of December 2010) to 90 deg. At a minimum, in order to limit the effects of atmospheric particularly on spectral data, observers should use PA = 90 deg (or equivalently 270 deg) if guide star selection allows for targets with a declination within 25-30 deg of zenith at the appropriate telescope (for GMOS-N, -10 < Dec < 50; for GMOS-S, -60 < Dec < 0). For targets with declinations outside these ranges it is very important to at least change the default position angle from 90 deg to either 0 deg or 180 deg. For those configurations most susceptible to the effects of atmospheric refraction (very blue spectral configurations or very narrow slit widths) it is recommended to follow the more detailed Observing Strategies described above.

Relevant Bibliography:

Fillipenko (1982, PASP, 94, 715)

Arribas et al. (1999, A&AS, 136, 189)