In the MEGACAM project, some specifications are still a matter of debate. This is especially the case for the grens option, for which no decisions are fixed yet, neither the final design of the dispersing element. In addition, there have been some discussions recently about the scientific interest of the grens mode at faint magnitude levels, because of the limited performances in deep fields, the crowding effect and the contamination by bright stars.
In order to quantify these effects and to test the feasability of such observations for specific programmes, I have simulated some realistic CCD images that can be obtained with MEGACAM and a grens. I have also extracted some spectra and measured the resulting S/N ratio at different magnitudes and tested the influence of the seeing on the spectral resolution. These simulations are discussed in the context of the search for well defined samples of quasars.
The wavelength range is chosen to cover the full optical range in two exposures, so the B simulation corresponds to a bandpass of 3700-6300 Å and the I to 6300-9000 Å. These ranges are arbitrarily chosen and could in fact be better optimised because the sky background, which is the main limitation in terms of detectivity and S/N, strongly depends on the width of the bandpass. The sky background is then computed for a simulated image of 1 hour exposure, including a photon noise contribution. The background corresponds to approximately 35,000 electrons per pixel in B and 95,000 electrons in I. These numbers are of course quite high, and could be reduced by segmenting the exposure if the full well capacity of the CCD becomes a limitation in the dynamics of the observation. Performances are computed for a 1 hour exposure and are obviously limited by the sky background. This means that they are increased by 1 magnitude when the integration time is multiplied by 6. The spectral dispersion was chosen as 40 Å/pixel, a typical value for very low resolution spectroscopy, but it remains to be discussed in details for a final choice for MEGACAM. Note also that in slitless spectroscopy, the resolution strongly depends on the seeing which fixes the width of the aperture, especially for unresolved objects or faint and small ones. In order to be more realistic, the 0-order and the 2nd-order contributions are added, with a transmission relative to the 1st-order of about 15% for the 0-order (the re-scaled direct image) and 3.5% for the 2nd-order (i.e. an attenuation of 3.5 magnitudes, and a dispersion of 20 Å/pixel). Finally the full image is convolved with a PSF corresponding to a seeing of 0.5''. All objects with B < 25 or I < 22.5 are associated with a spectrum. It is not necessary to go significantly deeper because the bright sky background limits the efficiency of grens observations. This is also the reason why crowding effects are not severe, at least in empty fields. An example of the simulated images is shown in Figures 6 and 7.
The extraction of the spectra is rather easy in the case of grens observations, as the position of the 0-order image fixes the origin in wavelength. For prism observations, this would be more difficult. A solution would be to split the exposure in
two and to turn the prism of 180 degrees in order to have a reversed dispersion, as it was the case for prism-objective plates. S/N=10 is obtained in 1 hour for B = 22.5 or I = 20.5 on the continuum of the extracted spectra.
Application to a survey of QSOs and AGNs at I < 23
One of the scientific programmes of the MEGACAM consortium is to build a sample of quasars in order to test their evolution at high redshift and the geometry of the Universe through their 3D spatial distribution (see section 2.11 in the MEGACAM document for a detailed scientific justification). The idea is to select a sample of unresolved objects and use multicolour
UBVRI photometry to separate quasars from stars. This procedure is rather efficient except for red objects where confusion exists between red stars (generally faint and cold M stars) and distant quasars. This point is critical because the aim is to built a complete sample of quasars with no redshift cut-off and the efficiency of detecting quasars from UBVRI colors only is known to drop in the redshift range 2.5 - 3.3. Two solutions are proposed, one being to add 3 narrow band filters at wavelengths corresponding to special features in the spectra of M stars (essentially TiO absorption features at 7000-7500 Å ), the other one is to extract the low resolution spectra obtained with the grens. I have tested the feasability of such observation by analysing the separation between M stars and quasars spectra at various magnitudes and various redshifts. The simulations are similar to the previous ones, except that the spatial distribution was simply a regular grid in the CCD frame. The input template for the quasars is a synthetic spectrum generated from a power law continuum combined with the contribution of wide emission lines characteristics of active nuclei. The magnitude is in all cases computed in the I-band, even for B simulations.
The main result is that up to I = 23 it seems possible to separate the QSO and M star spectra with a S/N of about 4 on the continuum in 6 hours in I. In order to optimise this survey of quasars selected from their I magnitude, a possibility would be to integrate during 6 hours in I and 2 hours in B, as quasars are very blue objects in B-I, whatever the redshift. With a full coverage of the optical wavelength range and some overlap between the different ranges in grens observations, this would make this kind of programme quite feasible, at least up to I = 22.5 in a reasonable observing time (detection becomes more marginal and seeing dependent at fainter magnitudes). But if for example one drops the redshift range [4,5], the B exposure will be enough to search for the Ly line, with a slight extension of the spectral range. I-selected quasars being much brighter in the blue, this would allow to obtain a sample of quasars complete to I = 23 in 2 to 3 hours only. Other solutions would be to reduce the wavelength interval of each exposure to about 2000 Å and to increase a little the dispersion up to 20 Å/pixel or 30 Å/pixel. The gain is double as the sky background is reduced and the detectivity increased, and emission lines are more easily identified when the dispersion is lowered. The final compromise between these parameters for a given astrophysical programme remains to be explored in detail, once one is convinced of the interest in observing with a grens and MEGACAM.