Wide Field Infrared Camera for UKIRT

Possibilities for Large Public Surveys with WFCAM

This section is based on a paper submitted to the Sept 1998 meeting of the Ground Based Facilities Committee by the IfA of the University of Edinburgh.

1. Introduction

The science case for the UKIRT wide field facility suggests an operating mode consisting of three overlapping observation types. First, a wide range of exciting problems can be attacked in traditional common-user open-time application mode (i.e. in PATT time). Second, there is a recognised need for Gemini precursor observations and support work. Finally however it gives us an opportunity to perform a large and ambitious public survey. Such a communal survey would (a) be an exciting and world-leading science project in its own right, (b) provide much of the Gemini precursor needs, and a launching point for more detailed UKIRT programmes, and (c) be a cost-effective way to run UKIRT.

The basic argument for a public survey comes in two parts - (i) meeting the science aims of many different large sample projects in a communal way, and (ii) producing an atlas, whose eventual uses cannot be foreseen, but which will be a world resource.

Much of modern astronomy has sprung from the major sky atlases - the 3C catalogue, the Uhuru and Ariel V catalogues, the IRAS catalogue, the COBE sky-maps. Through much of this time the UK Schmidt photographic sky atlas (and its Palomar counterpart) has been the backbone of UK astronomy. As well as providing identifications and saving huge amounts of 4m telescope time, it has been a source of science in its own right (eg Abell clusters, Lynds dark clouds, the APM galaxy catalogue). A large fraction of the science emerging from such sky atlases would have been impossible to predict beforehand. We can expect such productivity from an atlas whenever we move into a new region of parameter space. The first IR sky surveys (DENIS and 2MASS) are just now taking place but these are shallow and only really equivalent to, say, the Shapley Ames catalogue rather than the Palomar/UK Schmidt surveys. The analogy is quite accurate as we are considering going roughly 6 magnitudes deeper with WFCAM compared to DENIS.

2. Survey Forms - Private, Consortia and Public

Many astrophysical questions require large samples to address them, and others require rare objects to be found amongst large populations. These are best addressed by short well focussed experiments appropriate to small allocations in open time. Some problems require considerable amounts of telescope time, which however can still be achieved in open time by the formation of consortia with common aims which can hope to win larger allocations (for example the 2dF redshift survey). However, where the time required becomes a significant fraction of a year on one telescope, and/or where many different scientific projects potentially require overlapping or identical pieces of sky, it makes sense to produce the dataset required in a truly communal way, by a dedicated team.

3. Key Science Drivers for Survey Design

It may be possible eventually to conduct a complete Northern sky survey (and in conjunction with a new southern 4m telescope, perhaps an all-sky survey) but for now it is more realistic to consider a partial survey. We do not present a full science case here, but rather consider selected points that seem likely to drive the optimum design for the survey.

3.1 Generic Atlas Drivers

Any IR atlas must include the Galactic Plane. The longitude limits are set by the UKIRT declination limits of +58 and -40, just including the Galactic Centre. The best latitude limit is not so obvious, but probably +/- 5 is good enough, except that some key nearby open clusters and star formation regions are at mid-latitude, b~15. Several of these are relatively close on the sky around RA = 3-6, suggesting a special mid-latitude survey. The sensible depth of a Galactic Plane survey is probably limited by stellar confusion. Current best estimates are that towards the Galactic Centre at least there is a star every few arcseconds once you reach K=20.

At high latitudes the obvious thing to do is to match the Sloan Digital Sky Survey. The prospect of a quarter of the sky with complete optical and IR colours, and optical spectra, is extremely exciting. The Sloan survey is actually made of several parts. The footprint is approximately b>30, but an additional 2x50 degree strip will be surveyed near the SGP. In the main region, there will be a five band imaging survey complete to R=23 (5 sigma) but it is also intended to get spectra for all galaxies to B=19 and all quasars to B=20. Given the depth that we can reach in the IR in reasonable time, a reasonable aim would be to survey all or most of the Sloan area to a depth which matches the spectroscopic survey, and also to survey a smaller subset which matches the photometric survey. Note that the UKIRT declination limit excludes some of b>30, but the accessible area is still approximately 9000 square degrees. The depth which "matches" emerges from more specific considerations below.

3.2 Galaxies at z<0.1

An IR selected sample gives us the chance to tackle large scale structure problems properly for the first time, selecting by stellar mass rather than a hybrid of mass and activity. Of the order of 10 million galaxies are required, over a large fraction of the sky. This can be achieved by matching the Sloan spectroscopic survey which should have median redshift around 0.1. Galaxy colours are in the range B-K ~2-4 depending on type so for a limit of B=19 we need K=17, but at good signal to noise, so a 5sigma depth of K=19 is needed to give to 30 sigma at K=17 . Many other exciting galaxy projects would be accomplished with such a very large and moderately deep survey. For example one could construct a BVRJHK "main sequence" for galaxies, and see how it varied with density or cluster age. One could search for the "missing dwarfs" predicted by CDM cosmology but not found in optical surveys (they could well have a single very old population and no subsequent star formation).

3.3 Galaxies at z =1-2

Very deep IR surveys are an excellent way to find very high redshift galaxies, and a few hundred objects could address questions such as the history of star formation rate. Such objects will be dense on the sky and very faint - this kind of work is probably better done in open time. However at moderate redshifts (z=0.5 to 2) we can anticipate locating many thousands of objects, and estimating photometric redshifts, in conjunction with the Sloan photometric survey, as the 4000 A break moves through the red optical and into the IR. This opens up the possibility of measuring the growth of clustering, a key prediction of modern cosmologies. Based on the small deep IR surveys carried out so far, a survey to K=19 would be mostly local but with a tail of objects out to Z>1, and would need both optical and IR colours for photometric redshifts. A survey to K=21 would produce mostly galaxies in the range Z=0.5 to 2, and JHK colours alone could produce a crude photometric redshift. The Sloan photometric survey reaches R=23. At low redshifts, this corresponds to B~25 and K=20- 23 depending on galaxy type. However at higher redshifts the K-dimming is much more drastic in the optical than in the IR. For z=1-2 galaxies, a K=20 survey would easily match the Sloan, and K=21 would go beyond it. A depth of K=21 in a subset of the Sloan therefore seems highly desirable. At z=1 a 100 Mpc box is about 5 degrees across. To address the growth of clustering we would need perhaps 10 such contiguous fields. A survey size of around 250 square degrees is therefore needed to K=21.

3.4 Quasars

Combined optical-IR colours are the best way to find quasars, because of the 1micron inflection in the SED. (No other known objects have concave spectra ; see the VJK diagram from Steve Warren in the accompanying paper). Furthermore the K-band insensitivity to reddening would allow us to find the obscured quasar population which may explain the X-ray background and which may outnumber normal quasars by 3 to 1. The DENIS and 2MASS surveys will only get to the tip of the iceberg. To obtain a complete census of local quasars we need to match the intended quasar spectroscopic depth of the Sloan survey, B=20. For normal quasar colours this corresponds to K~17, but good S/N would be wanted so a 5 sigma depth of K=19 would be needed. (Similar to the local galaxy requirements.) Such a survey however is likely to produce large numbers of reddenened quasars not found in the Sloan survey. Spectroscopic follow up of sub- samples will satisfy many quasar addicts for years to come. A final point is that coverage in several passes in successive years would give variability information of unprecedented quality for the quasar population as a whole.

At higher redshifts quasars should also be easy to spot in JHK, as the "Big Blue Bump" moves into the IR; high redshift quasars will be faint and blue in JHK. The completeness of UVX samples at z>2 must be considered suspect, and evolution studies with IR selected samples is a prime aim. However only a few thousand objects are needed, the surface densities should be high, and one wants deep images to reach down the luminosity function, so such programmes are probably best in open time. However, at extremes a large public survey will have an important role to play. It is vital to know if any quasars exist beyond z=5. They may exist but could be extremely sparse on the sky both because of intrinsic rarity and because we can see only the most luminous objects. At z=5 an object like 3C273 will be approximately K=21 if qo=0.5 and K=23 if qo=0; by z=7 it would have K=22 or K=24. A deep (K=23 ?) survey of the order of a square degree would therefore be of extreme interest but beyond the ambition of normal PATT users, taking a few monthsLow mass stars and brown dwarfs.

The subject of brown dwarfs is one which current IR work and surveys such as DENIS are just now turning from fantasy to reality, but the surface is only just being scratched. A prime aim must be to test BD numbers versus cluster age. A Galactic Plane survey would include dozens of open clusters with a variety of ages. However they are typically many times further away than the Pleiades, so that one needs to expose perhaps a hundred times deeper, to around K=20. A second prime aim must be to push from the hydrogen burning limit of (around 0.1 Msun or 100 Jupiter masses) to the putative fragmentation limit (around 10-20 Jupiter masses). There are five obvious very nearby clusters in which this can be achieved. Of these Coma is above the UKIRT declination limit, and the Ursa Major cloud is very near the limit. The other three (Pleaides, Hyades, Taurus- Auriga) are all a few degrees across and whats more are close in the sky, strongly suggesting they could be covered in a single mid- latitude survey. At the distance and age of the Hyades (45pc, 600 Myr) a 20 Jupiter mass BD is expected to have K=20, and a 10 Jupiter mass BD has K=22.3. (The Pleiades is further but younger, so that expected BD fluxes are grossly similar). These numbers strongly suggest that a multi-cluster large area survey to K=20, 20 Jupiter masses, could be carried out in public mode, and deeper selected areas left to open time programmes.

A third prime aim must be to find significant numbers of field brown dwarfs in the disc population. The fact that the best nearby clusters are all at b~15 is not a coincidence but reflects the scale height of the nearby young population. Likewise this suggests that a Galactic Plane survey will miss many of the nearest and brightest BDs. On the other hand a high latitude survey at b>30 will only find extremely nearby objects; there may be very few of these unless the mass function is much steeper than is currently expected. It is essential that a variety of latitudes is sampled. To find field BDs, JHK colours and especially K are essential - because of molecular absorption, such objects are expected to be blue in JHK, not red. However an additional discriminant would be proper motion, as all such objects will be very close. As with quasar variability, this suggests we should accumulate the depth required by several passes in successive years.

3.5 Star Formation Regions

Within obscured star forming regions, an aim would be to reach the main sequence limit in the K-band, which would then be deep enough to satisfy many other aims. As with the BD calculation, to do this for star forming regions to a distance of around 500 pc through the Galactic Plane survey then suggests K~20, similar to the depth suggested by the confusion limit at small longitudes. Outside the Plane, covering the greater Orion star forming region seems an obvious aim. This is only about 30 degrees from the Hyades, and strongly suggests a mid-latitude survey covering all of Pleiades, Hyades, and Orion.

3.6 Galactic Structure

Star counts and kinematics for low mass stellar populations will be a very valuable addition to the studies that will come out of the digitised optical sky surveys. The key requirements are that a good range of longitudes and latitudes are covered, and that the best possible proper motion information is available, suggesting (as with the field brown dwarf survey and quasar variability) constructing all surveys in several passes over consecutive years.

3.7 Preliminary Survey Design

Putting the above considerations together we arrive at a survey design made of several parts of varying sizes and depths. This loses a little simplicity, but meets the various scientific aims with minimum time. The depth has been derived in K-band terms, but nearly all scientific goals need JH and K. For simplicity we assume that J and H together take as much time as K for a "matching" depth although of course what depth matches depends on scientific aim. Hours are converted to nights assuming an average of 10 hours/night. Times calculated assume that 70% of nights will have good weather. The calendar years required will be larger than the used survey time depending on what fraction of UKIRT time is used for the public survey, and what fraction for open user time. For now, we assume 50%.

All the major surveys would be done in several co-added passes in consecutive years, to give variability and proper motion information, as well source reliability confirmation. The following camera performance is assumed.

The minimum imaged area is 0.21 square degrees (all 4 detectors). The 5s limiting magnitude in 1 hour is K=20.8, H=21.6, J=22.4.

The fastest survey speed will likely be 0.75° every 211 seconds (includes overhead estimates), reaching to K=18.4. Attempts to go significantly faster than this will run into overheads and result in low observing efficiency.

3.8 High Latitude Sky Atlas

We aim to match the Sloan spectroscopic survey at b>30, Dec<58, giving approx 9000 square degrees, to K=18.8, J=20, H=19.2. This will take 2.2 years of elapsed time.

3.9 Galactic Plane Atlas

Here we cover two separated strips, l=350 to 90 and l=150 to 250, each covering /b/<5, giving a total area of 2000 sq. degrees. The required depth is K=20. Including J and H, this survey could be done in 5 years.

3.10 Mid Latitude Survey

To include Pleaides, Taurus-Auriga, Hyades, and the greater Orion region, as well as covering the general nearby disc population, we propose surveying a region at l=160 to 215 and b=-12 to -30. This is 920 square degrees. To a depth K=20 and assuming J and H coverage this will take 2.3 years. (If time allowed, extending to b=-5 to join the Galactic Plane survey would obviously be appealing).

3.11 Deep Survey

To study moderate redshift galaxies and the growth of structure (and doubtless many other exciting projects we can't imagine yet) we propose surveying 250 sq. degrees of high latitude sky in JHK to a depth K=21. This would take 3.9 years.

3.12 Very Deep Survey

To search for very high redshift quasars, and other rare objects, and more generally to probe unexplored space, we propose a very deep survey to K=23 covering 10 sq. degree, possibly in the Sloan Southern strip. In practice such a very deep survey could be gradually accumulated as a filler during the other surveys, revisiting the selected field repeatedly. This would take 6.2 years.

3.13 Survey Schedule

The total survey time, assuming all three bands, and 50% of UKIRT time being available for survey mode, is 19 years. This is clearly much longer than desirable. Therefore some scaling back of the area covered in the programmes would be required and/or synergy with VISTA, which could survey at least some of the Southern regions, reducing the load on UKIRT. If VISTA were to take some of the equatorial load, then the complete programme could be completed in perhaps seven years. If approval for the UKIRT wide field facility could be obtained in 1999, it could potentially be delivered in mid-2002 and major survey observations could perhaps start by early 2003.

References

Alcock et al. 1997 ApJ, 486, 697
Barsony et al. 1997 ApJ Supp 112,109
Burrows et al. 1997 ApJ 491,856
Comeron et al. 1993 ApJ 416,185
Dekel A., 1997 astro-ph/9705033
Dunlop J.S. et al., 1996, Nature, 381, 581
Greene & Young 1992 ApJ 395,516
Kenyon et al. 1998 AJ 115,252
Kirkpatrick et al. 1997 AN 476,311
Kochanek 1993 ApJ 419, 12
Kochanek 1995 ApJ 466, 638
Lanzetta K.M., Fernandez-Soto A., Yahil A., 1997 astro-ph/9709166
Larkin et al. 1994 ApJL 420, L9
Madau et al. 1996 MNRAS 283,1388
Maoz & Rix 1993 ApJ 416, 425
Ortiz & Lepine 1993 A&A 279,90
Pei & Fall 1995 ApJ 454, 69
Pettini M. et al., 1997 astro-ph/9707200
Pettini et al. 1997 ApJ 478, 536
Rix 1996 IAU Symp 173, p 1
Ruphy et al. 1997 A&A 326,597
Strom et al. 1995 ApJ 438,813
Tully R.B. et al., 1998 astro-ph/9802247
Unavane et al. 1998 MNRAS 295,119
Warren et al. 1994 ApJ 421, 412
Webster et al. 1995 Nature 375, 469