CFHT Widefield IR Imager

Some of the scientific projects and roles which a CFHT widefield near-IR imager could fulfil are explored. To achieve the scientific goals it is argued that the field size should be >20 and the array should be at least 6K $\times$ 6K. The imager should incorporate tip/tilt compensation and automatic focus controls to be competitive with the performance of other 4 and 8m class telescopes.

Introduction

The scientific cases for a widefield IR imager are so numerous and well-known that they scarcely need be repeated. In fact, the difficulty (and cost) of producing large format near-IR detectors is the only reason why there aren't several large arrays already in use around the world. The case for such an imager on CFHT is perhaps even more compelling than for other telescopes due to the excellent seeing and low water vapour content of the site, and also due to the relatively low thermal emssion of the CFH telescope itself. At a time when our communities are eagerly awaiting access to the VLT and Gemini telescopes, a widefield near-IR camera is even more desirous, since it will be able to provide strong support and complementary observations for a wide variety of projects.

Context

The CFH community will soon have access to 8m telescopes that will have powerful IR capabilities. Gemini N, in fact, is being optimized to deliver exceptional IR performance, with silver-coated mirrors and low overall emissivity (goal: 2%). The near-IR camera, NIRI, is being optimized to take maximum advantage of the expected Gemini image quality with and without adaptive optics, so that even its widest field is relatively small, 110 $\times$ 110. Similarly, the field of ISAAC, one of the first VLT instruments, will be 25 $\times$ 25. On the other hand, the field of the VLT second generation near-IR imaging spectrograph, NIRMOS, will be very large, four cameras of 6 $\times$ 8 each (utilizing four 2048 x 2048 Hg:Cd:Te Rockwell devices). However, the instrument will be uncooled so that its performance in H and K will be significantly reduced compared to a fully cryogenic instrument. With the recent news that the VLT is actually achieving superb image quality (and so, presumably, will Gemini), CFHT will not be in a very competitive position except with a very wide-field, fully cryogenic, camera delivering excellent images. In fact, other 4m class telescopes have plans for, or are considering, 4K $\times$ 4K mosaics of near-IR detectors, so CFHT should try to obtain the biggest possible imager in order to be really competitive and to take advantage of the CFHT site/excellent telescope performance.

Types of Projects

Pre-planning

Time on the 8m telescopes will be at a premium and also the trend is moving towards more queue-scheduled and service observations where careful planning is essential, if not critical. This will be particularly true for spectroscopic and adaptive optic observations on these telescopes where precise coordinates and offsets will be required (see below). The same will be even more true in preparing for Next Generation Space Telescope (NGST) observations.

Population Studies

Naturally, there are many classical studies utilizing some form of colour magnitude diagrams that would benefit greatly from deep, widefield near-IR surveys. These range from detailed studies of the stellar populations associated with molecular clouds, large nearby star-forming regions, the halo of the Milky Way, the Bulge of the Milky Way, etc., metallicity distributions and star-forming histories of nearby galaxies, etc., etc. The study of even the closest of these requires surveys to significant depths. For example, one must go fainter than K $\sim$ 19 to reach a good sample of Giant Molecular Clouds. Here-to-for, most such studies have been severely hampered by the small size of existing detectors (which are usually much smaller than the targets and, for example, are much smaller than the corresponding visible or X-ray surveys/detectors which provide complementary information).

Detecting rare or unusual objects

The near-IR offers very significant advantages in the selection of various classes of objects for statistical studies or for further investigation. Some of the obvious ones range from nearby small solar system objects, brown dwarfs and low-mass stars to very high redshift QSOs and star-forming galaxies and obscured starburst galaxies.

Photometric redshifts

Recently, the power of the photometric redshift technique has been demonstrated by many authors (e.g., see Hogg et al. 1998). Figure 1 kindly provided by S. Gwyn, shows that a combination of UBRI colours gives good redshift estimates for z < 1 but not for higher redshifts, whereas UBRIJHK colours give excellent redshift estimates for all redshifts that have thus far been probed in the Hubble Deep Field. Now that our understanding of the galaxy population with z < 1 is reasonably well established, it is becoming increasingly important to detect higher redshift galaxies for further study. This is especially true since there is increasing evidence that star formation peaked near z $\sim$ 2 (Madau 1997), as did the the space density of quasars (Hook et al. 1998).

**Figure 1:** Left: Redshifts derived from *UBRI* photometry of galaxies in the HDF compared with spectroscopically-determined redshifts. Right: Similar comparison for redshifts derived from *UBRIJHK* photometry of galaxies in the HDF (Gwyn 1998).
$\begin{figure} \plottwo{cramptonIfig1.ps} {cramptonIfig2.ps} \end{figure}$

Large-scale structure

Recent studies at high redshifts indicate that the clustering of galaxies at z $\sim$ 3 is comparable to that seen locally (Steidel et al. 1998). It will be interesting to follow this to even higher redshifts through near-IR surveys since the amount of large-scale power in the angular correlation function at these redshifts provides significant constraints on cosmological models. Again, deep and wide-area surveys are required because even by z $\sim$ 2 a L* galaxy has K_AB $\sim$ 22 and a 100 $\times$ 100 Mpc area subtends $\sim$ 3 $\times$ 3 (for H₀ = 65 km s^-1 kpc ^-1 and q₀ = 0.1).

Identification of sources

As studies in other wavebands probe to fainter and fainter limits over increasingly large areas, the requirements for precise identification of faint sources in the near-IR are increasing dramatically. This is especially true of surveys like the ISO and SCUBA surveys that can detect obscured or very high redshift galaxies but have poor angular resolution and hence positional accuracy. At very faint limits there are frequently many visible sources within the error boxes, and images or colours in the near-IR that are much less affected by reddening are needed to determine secure identifications. Similarly, the very high redshift counterparts of radio sources or X-ray sources can frequently be recognized through a combination of optical and near-IR colours. Ideally, a near-IR imager should have a field comparable to that of forthcoming instruments such as XMM, AXAF, MegaCam, SIRTF, VLA/FIRST, LSA/MMA, etc. In other words, the field should be of the order of one degree.

Desired Characteristics

Field size

As for all imaging survey instruments, the bigger the field size the better. This is partly because many of the targets have characteristic sizes of at least a degree and partly because otherwise the time required to complete major surveys such as those planned for MegaCam becomes prohibitive. There are also minmum dimensions in order that such a major instrument (and, hence, CFHT) be competitive. As mentioned above, 4K mosaic imagers are in the planning stages for other 4m telescopes already. CIRSI, the Cambridge Imager, will be a mosaic of 4 $\times$ 1K Hg:Cd:Te detectors that can be placed at the foci of various telescopes (and hence the foreoptics/filter will be uncooled and operation is planned only in $J \& H$ ), offering fields up to 16 $\times$ 16. In imaging mode, the VLT NIRMOS will also be a mosaic operating at $J \& H$ with a similar field of view. Thus, to be competitive at all in $J \& H$ the minimum size should be 4K $\times$ 4K with a field of view of >16 $\times$ 16. Obviously, to give CFHT a significant advantage, the field should be minimally >20. This probably implies a $\geq$ 6K $\times$ 6K mosaic assuming a reasonable pixel scale (see next section).

Pixel scale

Particularly in the near-IR, one should expect images to have FWHM $\leq$ 0.4'' with CFHT during a substantial fraction of the time. Indeed, images approaching 0.25'' have been reported on Mauna Kea with tip/tilt correction with UKIRT and with the UH 88-inch telescopes. Since many of the scientific projects mentioned above require significant depth as well as a large field and significant gains in depth can be realized with improved image quality (especially with the bright near-IR sky), then tip/tilt compensation is essential. Since part of CFHT's reputation has been built upon delivering superb images, one presumably would like to adequately sample the FWHM, implying pixel scales of 0.2'' or smaller.

A study should be carried out to investigate the tradeoffs between pixel size and field size for the anticipated median image quality. The pixel size obviously has implications on the detectability and depth achieved in a given time, and obviously on the size of field surveyed per unit time. Simulations and analyses carried out by Elston (1996) in a similar type of study for a CTIO imager demonstrates that intuitive answers aren't always correct and that, surprisingly, a relatively large pixel scale can bring substantial gains in pure survey mode.

Arguments can thus be made for two pixel scales, one to provide the largest field and one to critically sample the best possible images (which might have FWHM 0.25'').

Astrometric Requirements

Although related to the previous two sections, requirements related to the determination of precise source positions merit separate mention. For example, preparation for observing runs with instruments on 8m class telescopes or NGST will frequently require extremely accurate relative positions. For example, GNIRS, the Gemini Near-IR Spectrograph, will use 0.1'' slits and have 0.05'' pixels so that accurate placement of objects within the slits (or on integral field units) will be much more of a challenge than we're used to. Similarly, the NGST goal is 8m diffraction-limited performance in the 1-5 $\micron$ region, i.e., the order of 0.04'' images at J. Thus, to avoid significant target positioning overheads, relative astrometry to an accuracy better than 0.02'' is very desirable. It has been demonstrated that 1/50 pixel astrometric accuracy is achievable for point sources, but I suspect that most of us would be much more comfortable with trying to attain 1/10 pixel positional accuracy for, say, faint galaxies with typically low S/N. This implies a maximum pixel scale of 0.2''.

As telescopes and, in particular, interferometric and adaptive optic systems provide better and better image quality, so, too, do the requirements increase for absolute astrometry. Images with FWHM $\sim$ 0.1'' or better can now be routinely obtained, but it is often extremely difficult and frustrating to try to register them with equally good or better resolution radio maps, for example. The relatively small fields of these high resolution images often exacerbate the problem because there simply aren't any known reference objects within the field. Thus the registration and identification of sources observed or detected at other wavelengths will be an increasingly important role for a large near-IR imager. Since there is only about one astrometric standard (with m < 18) per 7 arcmin^-2 at the galactic poles, a substantial field is required to attain appropriate accuracy. Considering that at least 9 stars are required for reliable astrometry, the field must be larger than 8 $\times$ 8.

Naturally, a requirement that the imager be capable of providing accurate astrometric positions also implies that considerable attention must be paid to flexure and stability in the design of the instrument.

Summary

A widefield near-IR imager will offer the CFH community many opportunities to carry out very competitive, very exciting science as well as providing extremely important complementary observations for MegaCam, the 8-10m telescopes, and space missions. In order to achieve many of the science goals and in order that it represent an advance on projects currently underway, it should:

The last two requirements imply a minimum of 6K $\times$ 6K pixels. In this review I have deliberately considered only the scientific and competitive aspects of a widefield near-IR imager for CFHT, and not considered what is feasible or reasonable to build or finance. Such a large imager operating in the near-IR will not be simple or cheap. However, Morbey (1993) has designed visible-light correctors for the CFHT cassegrain focus with fields of >20 and superb image quality and, in some ways, it is easier to achieve this in the near-IR. Hodapp (1998) also has a design for a 4K mosaic imager with a field of 17 $\times$ 17. To illuminate a 20 $\times$ 20 field at the cassegrain focus, very large lenses will be required, but even exotic glasses like CaF₂ are now available in diameters up to 350mm, so that restrictions are financial rather technical.

Acknowledgements: Richard Elston's excellent summary of wide field IR imaging was an extremely useful resource. I would also like to thank Jean-Luc Beuzit, Tim Davidge, René Doyon, Klaus Hodapp and Olivier LeFèvre for very useful discussions and material. Stephen Gwyn provided his latest photometric redshift results to demonstrate the great potential of that method and, in particular, of his variant of the method.

$\begin{references} % latex2html id marker 43 \reference{} Elston, R. 1996, The ... ...et al. 1998, in The Birth of Galaxies, Xth Rencontres de Blois \end{references}$