Wide Field Infrared Camera for UKIRT

Science Programmes with WFCAM

1. Introduction

This case was compiled from various contributors and presentations to various committees including the GBFC, Astronomy Committee and UKIRT Board.

2. Widefield Infrared Galaxy Surveys

2.1 Local Galaxies

For local galaxies, the IR comes into its own for highly extinguished objects, either with internal dust, or galaxies near to the plane of the Milky Way. The extinction is a strong function of wavelength: AJ = 0.27 AV, AH = 0.17 AV, AK = 0.10 AV.. Galaxies with AV around 3-4 will therefore be visible as red objects in H-K, but will be virtually undetectable at J and shorter. The longest wavelength employed determines how close to the galactic plane it is possible to work.

Extragalactic surveys at low | b | are not simply masochism, but are driven by the fact that some of the richest local superclusters, responsible for a good fraction of the local dipole velocity, are located behind the plane (e.g. Dekel 1997). IRAS is of some help in mapping these systems, but velocity studies based on Tully-Fisher methods require optical/IR magnitudes that are corrected properly for the effects of extinction. This requirement favours data at the longest feasible wavelength, which is normally taken to be K (e.g. Tully et al. 1998). Note that accurate Tully-Fisher work in fact benefits greatly from IR data, even at high latitudes, because of the need to understand internal extinction. UKIRT can access a significant fraction of the Southern sky, so there is a possibility here of fruitful overlap with the AAO 6dF survey, which will concentrate on Tully-Fisher velocity surveys in its second phase.

An extremely exciting data set will be the ugrizJHK imaging which will be available for local galaxies from combining SLOAN/WFCAM data. This will provide stellar population and dust information as well as multi-band morphologies on a huge number of galaxies at z<0.1 to provide the first large multi-band database to compare low-z and high-z populations.

The WFCAM survey will also allow the construction of JHK luminosity functions from the SLOAN/2df redshift survey samples (currently based on r/B selected samples) - for a less dust-sensitive estimate of the local galaxy population, as well as providing all-sky samples for Tully-Fisher and Fundamental plane work.

It should also be possible to use the survey to search for red 'evolved' LSB galaxies by suitable filtering of the images. This will be the first red-selected LSB catalogue - current surveys are blue-selected and hence typically find blue (young) LSBs. The existence of an equivalent population of red (old) LSBs has yet to be determined. The volume density of such galaxies means that they can only be reasonably detected in a very wide field relatively deep near-IR survey.

2.2 Galaxies at High Redshift

IR data become of increasing importance in the study of distant galaxies. Optical magnitudes probe the rest-frame UV, measuring mainly the current activity in unextinguished young stars. IR data are vital in three respects.

(i) diagnosing the presence of an old population

(ii) allowing accurate photometric redshifts, especially in ultradeep surveys (K>20)

(iii) estimating extinction in the starburst component

Considering points (i) & (ii) first, what matters is the placing of the 4000-angstrom break. Figure 1 shows the spectrum of an old elliptical at z = 1.55 (Dunlop et al. 1996). The JHK magnitudes fix the old-population baseline, and the large break between I and J clearly indicates both great age (rather than extinction), and also allows a good estimate of the redshift. Since the HDF, photometric redshifts for all classes of galaxy have been an active industry (e.g. Lanzetta et al. 1998); these estimates depend for their success on identifying the 4000-angstrom break, so that IR data are crucial if the method is to work at high redshift. Consider figure 1 again: for z > 1.8, the break passes through the J band, so that J-H becomes the main redshift diagnostic, with H-K taking over for z > 2.8. However, even for z > 1.8, K data are really required to establish that the IR colours are normal, and that, for example, the galaxy is not heavily extinguished. Extinction in high-redshift galaxies must be understood in order to estimate the total star-formation rate. Pettini et al. (1997) have compared G-R colours of high-z starbursts to theoretical models; the observed colours are too red, from which Pettini et al. infer extinction, so that the star-formation rates are scaled upwards by factors of up to 10. However,

Figure 1. Spectrum of an old elliptical at Z=1.55

this foreground-screen modelling must be too simplistic; some parts of the galaxy will be more deeply embedded than others, and IR data are the only robust way of investigating the overall output of these galaxies. K-selected samples to z=3 are selected at rest-frame V band and so provide a more reliable view of the galaxy populations at these epochs free from the worst effects of dust obscuration.

Figure 2. Redshift of galaxies in the HDF as a function of K magnitude

To put this discussion in context, it may be useful to see expected redshift content as a function of IR depth. Figure 2 is a plot of redshift against K magnitude for the HDF; the HST data allow very good redshift estimates when combined with the KPNO JHK images (available via the HDF home page). The lesson of this picture is that galaxies in the critical z > 1.8 regime start to appear in substantial numbers for K > 19. Galaxy surveys will therefore automatically divide into two regimes: (i) 'all-sky', where the interest will be in high-precision studies of stellar populations and internal extinction, plus the possibility of finding rare luminous galaxies at high redshift; (ii) 'selected areas', where the focus will be on the high-redshift universe. From the point of view of this latter aspect, a 4-m survey based on UKIRT will be uniquely powerful, not only because it can provide K data, but also because it will have the sensitivity to reach the depths where high-z galaxies dominate.

2.3 Clusters

Deep wide-field surveys in the IR are a powerful tool for searching for high redshift (z=1-2) clusters compared to other techniques (optical surveys or x-rays). This is due to the early-type galaxies which dominate the cluster populations having only modest K-corrections in the near-IR compared to the optical. The number of massive clusters (M>1015 Mo) at z>1 is a sensitive test of the cosmological curvature - declining by 3-4 orders of magnitude between W=0.1 and W=1. The difficulty in applying this test is the small number of clusters expected ~ one per square degree and hence the large area which must be surveyed to obtain a reliable result. We estimate that a massive cluster at z~1 could be detected in a K=20 survey at a S/N of 4-6 with a 1 Mpc (3 arcminute) diameter search cell. Applying colour selection based on J-H/H-K colours to the sample would extend this technique to lower masses (1014 Mo) or higher redshifts z=2-3. Clearly, a widefield JHK survey to K=20 will uncover thousands of rich clusters at z=1-2 if we live in a low density universe, and will provide a strong test of the cosmological curvature as well as a substantial catalogue of clusters for detailed study of the influence of the environment on the properties of galaxies in the early universe and the origin of the morphology-density relation.

3. The UKIRT Wide Field and the use of near-IR selected samples of quasars for cosmology

3.1 Introduction

Samples of bright quasars have proved extremely valuable resources for the observational study of the formation of galaxies using damped Lya absorbers (DLAs), and for cosmology from the analysis of the number of gravitational lenses. The study by Pei and Fall (1995) of the statistics of DLAs (detected in the spectra of bright quasars) is an approach to the universal history of star formation that is entirely independent of the faint galaxy imaging work of Steidel and others. Remarkably their calculated curve of the history of star formation was obtained earlier than the curve of Madau et al (1996) and the results are quite similar. However their analysis relies on simplifying assumptions about the dust-to-gas ratio in the DLAs, in order to quantify the bias. Larger quasar samples selected at longer wavelengths would remove this uncertainty and improve the statistics. The situation for gravitational lensing is similar. The studies of Kochanek (1993, 1995) and Maoz and Rix (1993) of the HST Snapshot Survey (bright optically selected quasars) currently provide one of the best upper limits on the value of the cosmological constant. The great uncertainty in this work (e.g. Rix 1996) is again the effects of dust. If elliptical galaxies at redshifts 0.5 < z < 1.0 contain substantially more dust than today, several lensed quasars may have been missed and the limits on A,, will be severely weakened. A number of radio-selected gravitational lenses do show evidence for reddening.

In this note we consider how to remove the uncertainties in these important results using quasar samples flux-limited at near-infrared wavelengths to minimise biases due to dust. Such a sample of quasars can only be achieved by including imaging at K. We also discuss the general question of incompleteness of optical samples of quasars.

3.2 Near Infrared KX selection of Quasars and Number Counts

Because of their power-law spectra and the thermal spectra of stars, quasars are bluer than stars at short wavelengths and redder than stars at long wavelengths. This points to a K-excess (KX) selection analogous to the ultra-violet excess (UVX) selection that is very effective for redshifts z < 2.2. The suggested method is illustrated in Fig. 3 which compares the VJK colours of bright quasars (mostly from the LBQS) and stars. This method appears to be even more effective than the UVX method since reddened quasars will also be detectable by this technique. The reddening vector depends on the redshift, but is either parallel to the stellar sequence or pointing slightly away at all redshifts of interest. Apparently the KX method would produce the most unbiased samples of quasars yet. The method will work to redshifts at least until Lya is well into the V band, i.e. to z - 3.5, and it might be possible to extend it to higher redshifts using redder optical wavelengths, e.g. R, instead. In the corresponding VJH diagram the quasar distribution overlaps the stellar locus. The VJK diagram is the best two-colour diagram from combinations of V, J, H, K, for separating quasars from stars.

This table provides an estimate of the cumulative numbers of quasars per square degree brighter than the tabulated limiting magnitudes. These numbers have been computed from the B-band number counts assuming an average quasar spectral energy distribution. The tabulated numbers are expected to be quite accurate down to K = 19, H = 19.5, J = 20, but rather uncertain at fainter magnitudes.

Figure3. Colour-colour diagram for quasars showing their separation from stars

Table 1: Cumulative numbers of quasars deg-2 brighter than specified mag. These numbers are conservative as based mostly on UVX surveys which are very incomplete for z > 2.2.

3.3 Damped Lyman Alpha Galaxies and the History of Star Formation

The damped Lya galaxies (DLAs) seen in the spectra of distant quasars contain virtually all the neutral hydrogen (HI) in the Universe. The cosmic density of HI at z = 3 is similar to the density of stars today and the DLAs at z = 3 are of low metallicity, less than 1/10 solar (Pettini et al 1997). This suggests that the high-redshift DLAs are the gas reservoirs from which today's stars formed. Therefore an analysis of the decline in the density of HI with cosmic time can provide the history of star formation in the Universe as long as the obscuring effects of dust can be accounted for. A quasar lying behind a dusty DLA will suffer extinction and may drop out of the samples of bright quasars used to survey for DLAs, so the density of HI is underestimated. Pei and Fall (1995) correct for this bias, accounting in a self-consistent way for the increasing obscuration as star formation progresses. In this way they compute the history of star formation and chemical evolution in the DLAs.

The analysis of Pei and Fall nevertheless rests on a number of assumptions, in particular that the dust-to-gas ratio in the DLAs is a constant of the population at any redshift. The effects of dust are dramatic at these redshifts since we are observing the restframe ultra-violet. For example a quasar behind a z = 2.5 DLA with E(B - V) of only 0.2 would suffer 2.8 mag extinction in the B band. So DLAs with this much dust would almost certainly not be found. Therefore if there is a large dispersion in the dust-to-gas ratio it would be difficult to identify in the sample used. This raises the possibility that Pei and Fall's assumption is incorrect and there exist high-redshift dusty DLAs that have been completely missed by optical surveys so far. A spectroscopic survey for DLAs using a magnitude limited sample selected at near-ir wavelengths, ideally in the K band, would reduce the uncertainty due to dust enormously. For the above example the extinction would be lowered from 2.8 mag to only 0.5 mag. Bright (to allow high S/N spectroscopy) high-redshift (as DLAs can only be found at z > 1.8 by optical means) quasars are required.

Based on the quasar luminosity functions of Warren, Hewett, and Osmer (1994) we estimate that a survey covering 1500 sq degs to K = 16.5 (at S/N=30) would produce some 1700 quasars in the redshift interval 2.2 < z < 3.5, and some 300 DLAs. The last number should be compared against the total number of DLAs so far discovered of 85, after more than 10 years of surveys. These predictions are expected to be quite accurate. Bright quasars are required not only for detection of the DLAs but also for high-resolution measurements of metal lines to measure the metallicities. A typical quasar at K = 16.5 has R = 18.5. This means that even quite red quasars in the sample will be bright enough' for detailed spectroscopy with an 8m class telescope.

It is worth noting that radio-selected samples, although unaffected by dust, are not well suited to the problem. For any survey it is necessary to obtain spectra of all the quasars in the sample. Radio samples, however, have a large spread in optical magnitudes and include quasars that are fainter than B = 22, which would be too faint for the spectroscopy necessary to detect the DLAs and the high-resolution observations necessary to measure their metallicity.

3.4 Dusty Lenses and the Value of the Cosmological Constant

One of the best limits on the cosmological constant, and one that is in conflict with the recent results from high-redshift supernovae, comes from the statistics of gravitational lensing, Lambda. < 0.7 at 90% confidence (e.g. Maoz and Rix 1993). (Note: Chiba and Yoshii (astro-ph/9808321) draw different conclusions. However, the fundamental point is that the number of garvitational lenses, and their image separations, can be used to measure the cosmological parameters, but the current samples may be substantially incomplete due to dust in the lenses.) The limits are derived by computing the expected fraction of the quasar population that will be gravitaionally lensed given the known properties of the local galaxy population. The optical depth to lensing for a given lens population is 4-5 times higher in a flat universe with Lambda,, = 0.7 than in one in which Lambda,, = 0. The best survey to date for placing limits on Lambda,, is the HST Gravitational Lens Snapshot Survey of 500 bright z > I quasars, which includes 5 lenses. Bright quasars are best for this purpose because the steep luminosity function at bright magnitudes brings in proportionally more quasars if amplified by lensing ("amplification bias") i.e. bright samples contain a larger fraction of lenses.

For an isothermal sphere the lensing cross section is proportional to the stellar central velocity dispersion to the fourth power. Because of this elliptical/SO galaxies dominate the lensing optical depth. The amount of dust in local E/SO galaxies is small so one might expect that dust would not bias the estimate of Lambda,, even though derived from an optically-selected sample of quasars. Nevertheless a number of radio-selected gravitational lenses are very red (e.g. Larkin et al 1994). Although at present it is not clear whether the reddening is in the source or in the lens, these observations raise the possibility that at the typical redshifts of the lenses 0.5 < z < 1.0 E/SO galaxies contain signifcant amounts of dust. Alternatively the optical depth due to spiral galaxies has been underestimated. Either way the limits put on the value of A,, would be severely weakened.

The problem is very similar to the difficulty in correcting for biases in surveys for DLAs, and again a wide-field survey for bright quasars would provide the definitive answer. A VJK survey to K = 16 (S/N=30) over 5000 square degrees would produce a sample of 5000 quasars. Follow-up snapshot imaging of each quasar to K = 18.5 in excellent seeing and using high-order adaptive optics could detect any secondary images 2.5 mag fainter than the primary, with separation greater than - 0.2 arcsec. These search parameters are similar to those of the HST Snapshot Survey, but the sample is ten times larger and the extinction is greatly reduced. This survey should produce some 50 gravitational lenses and place extremely tight limits on the value of Lambda.

The ambitious CLASS+JVAS radio survey for gravitational lenses is expected to produce in total some 20 lenses. The difficulty with this work is in establishing the redshift distribution of the unlensed source population, which is essential in order to interpret the results. In fact in many cases of radio gravitational lenses the source redshift has proved impossible to measure despite the boosted flux from lensing. Measurement of the redshift distribution of the bright K-selected quasar sample proposed here will be straightforward as their typical optical magnitudes will be R < 18. Note also that it is not required to measure redshifts for all 5000 quasars. Spectra would be obtained of the lensed quasars and sufficient of the unlensed quasars to measure accurately the bright end of the quasar luminosity function. In addition it would be necessary to measure the quasar luminosity to several magnitudes fainter than K = 16 i.e. to the faintest unlensed magnitude of the lensed quasars. This is covered by the next subsection.

3.5 Red Quasars

Webster et al (1995, Nature, P75, 469), on the basis of the measured B - K colours. of radio-selected quasars, argue that approximately 80% of quasars are missed in optical surveys due to the obscuring effects of dust. This result is currently disputed and other interpretations than the effects of dust are possible: for example that the red B - K colours of many quasars in their sample are due to the stellar population in the host galaxy, or high-frequency synchrotron emission The issue of the bolometric output of the quasar population is an important one, however, and one that can only be addressed by making surveys for quasars at two different wavelengths. Since radio-loud quasars may be a different population to radio-quiet quasars probably the best wavelength to complement optical surveys is the K band as it maximises the wavelength separation while allowing faint limiting magnitudes to be reached. A useful project therefore would be the measurement of the quasar luminosity function in a K-band limited survey. This requires a wide-field survey at bright magnitudes and a more narrow survey to faint limits. The projects listed above will more than satisfy the former requirement. A faint survey to find 1000 quasars brighter than K=20 (S/N=20) would need to cover about 5 square degrees, and would satisfy the latter requirement. The measured luminosity function is also needed for the interpretation of the lensing survey, as noted above.

4. The UKIRT Wide Field In Galactic Astronomy

4.1 Introduction

The following discussion of the science potential for Galactic Astronomy assumes that the UKIRT wide field is used to carry out a survey of the Galactic Plane, producing signal-to-noise of ~5 on sources with K=20. This represents a factor of 100 improvement over the sensitivity of the 2Mass survey. Sensitivity would be approximately one magnitude better than this at J and H.

4.2 Brown Dwarfs and Very Low Mass Stars

While the presence of brown dwarfs in binaries can be inferred from radial velocity measurements, isolated brown dwarfs and low-mass stars can only be detected by photometric surveys which extend into the infrared. Early work on such objects employed infrared photographic plates, but it is now clear that true infrared surveys with large apertures are the only productive way ahead. Genuine Brown Dwarfs are, of course, extremely faint. Even for relatively luminous objects such as GL229 (MK=15), the 2Mass survey samples a volume extending to less than l0pc from the Sun, and is expected to detect only some 200 such sources (Kirkpatrick et al. 1997). Since the volume sampled reaches much less than one scale height from the galactic plane,.the z-distribution of these objects will remain undetermined.

A wide field on UKIRT offers two possibilities: to push our knowledge of the brown-dwarf mass function some hundred times deeper in the same volume as that sampled by 2Mass, and to sample a much larger volume to the same luminosity limit (and hence bring the statistics out of the small-number regime). The mass function is still rising into the brown-dwarf region; the turnover probably occurs between 0.04 and 0.01 M(sun), which will not be sampled by any existing survey in any volume. For a 0.02 M(sun) BD, MK=19.6 (Burrows et al. 1997). A deep survey to K=20 would detect, for example, 10 BDs at this mass limit in both the Hyades and Pleiades clusters. Secondly, it is likely that the brown dwarfs currently being studied represent only the youngest such objects; a deeper survey will sample a greater range of ages (this is why the mass detection limit in the Pleiades is the same as that in the Hyades, despite the former's greater distance). In the field, for a K-band sensitivity of 20, volumes of 100 and 100,000 cubic parsecs can be sampled for 0.02 M(sun) BDs of age 5 and 1 Gyr respectively. Among the younger objects of this mass alone, this guarantees between 30 and 3000 detections (these calculations involve extrapolation from the field star IMF; these numbers assume that the power-law index lies between +1 and -1. In the cluster calculation above, we assumed that the IMF was flat).

Fluxes in the K-band are extremely sensitive to photospheric methane absorption; thus, the K band is a sensitive discriminant of objects with T(eff)<1000K. Models by Burrows et al. (1997) predict that such objects have extraordinarily BLUE J-K colour, for example, while J-H remains typical of hydrogen-burning M dwarfs. For [M/H]=0.0 and T(eff)=500K, J-H=+0.6, while H-K=-2.0; thus even an upper limit of J-K<-0.5 is sufficient to distinguish between very cool BDs and allother stellar types.

4.3 The Cool White Dwarfs

The coolest white dwarfs are the oldest stars in the galaxy.An understanding of the population of very cool white dwarfs provides constraints on the age of the galaxy. Furthermore, white dwarfs have been proposed as a predominant component of the dark matter halo inferred by measurements of the rotation curve of the galaxy. Recent microlensing results (Alcock et al 1997) have added weight to this hypothesis with the detection of a significant number of 0.5 MO objects in the halo of the galaxy. In spite of searches for very cool and faint white dwrafs, this is a population which remains relatively unknown.

Figure 4. I-K, V-I colour-colour diagram for the Bergeron, Ruiz & Leggett (1997) sample of cool white dwarfs (solid squares) showing the location of WD0346+246 (open circle with error bars). CIA-affected colours from models are shown as curves for various helium/hydrogen ratios.

The luminosity funtion for cool white dwarfs compiled from the Luyten Half-Second catalogue shows a dramatic turnover at MV=16. However, potential incompleteness in this survey could explain the severity of the decline in the luminosity function. Additionally, Isern et al (1998) argue that the "pencil beam" HST surveys cover such a small area that they do not strongly constrain the numbers of halo white dwarfs.

The discovery of cool white dwarfs from their proper motions is a laborious and slow process. A much better method would be to find them using broadband colours from surveys. Unfortunately, in the optical their colours are similar to those of red dwarfs and are difficult to separate from the vast numbers of background objects found in large-area surveys. However, the remarkable effects of collisionally-induced H2 absorption (CIA) in the high-pressure atmospheres result in very anomolous colours in the infrared. This has been highlighted by Hodgkin et al (1999) who show that the V-I, I-K colours of WD0346+246 are completely unlike those of any other object. The CIA results is flux which drops off rapidly in the IR - far more rapidly than would be expected from a blackbody at the optical colour temperature (3850 K). So V-I = 1.5 while I-K= -1.7 ! In fact, such objects could probably even be found from their IR colours alone since J-K= -1.6 and the bluest colour of a normal object (Rayleigh Jeans tail) is -0.21. An IR survey would therefore be capable of finding large numbers of such objects from J-K colour alone, allowing follow-up studies, temperature and age determinations.

Their space distribution in the galactic plane could be studied. If we assumed that any object with J-K<-0.2 was a candidate CIA-distorted cool dwarf spectrum then a survey with limiting magnitudes of 20.8 at J and 21 at K could detect objects like WD0346+246 at a distance of 122 pc, allowing their distribution orthogonal to the galactic plane to be studied.

4.4 Stellar Populations and Galactic Structure

Surveys such as DENIS are begiming to have an impact on studies of Galactic structure (see, e.g., Unavane et al. 1998). However, the dominant contributor to star counts at the sensitivities typical of DENIS and 2Mass are K and M giants (Ruphy et al. 1997). Late-type dwarfs are sampled to less than 1kpc from the Sun. Spatial variations in the giant/dwarf ratio, which reflect the star-formation process and initial mass function, are not determined by the current surveys. The UK IR deep survey detects sources 100 times fainter, allowing dwarfs to be traced much further across the galactic disk, and giants and supergiants to be detected on the far side of the galaxy.

4.5 Star Formation and the Initial Luminosity Function

Surveys of dark clouds reveal the initial luminosity function, which can in principle be compared with the field star function. This is currently the subject of some controversy, with apparent differences between results from different environments. For example, Comeron et al. (1993) suggest that the luminosity function increases monotonically below the H burning limit, while Strom et al. (1995) suggest that field brown dwarfs are rare. Even within the well-studied rho Oph complex, the most sensitive current surveys are complete only to K<14 (Barsony et al. 1997). For an age of a few million years, this permits detection of PMS objects only down to a few tenths of a solar mass in a cloud such as rho Oph. As with the brown dwarfs, the UKIRT widefield would probe the IMF down to objects of order a hundredth of a solar mass. It is also possible that the initial luminosity function is spatially variable even within a single star-forming complex (as suggested for rho Oph by Greene & Young 1992), so extensive surveys are required. Currently only a few local clouds (Cha 1, rho Oph, Taurus, Orion, Perseus) have been surveyed, and the variation of the luminosity function with position in the Galaxy remains essentially unexplored.

Figure 5. Shows theHR diagram with pre-main sequence evolutionary tracks.Stars above (grey dots) and below (dark dots) the hydrogen burning limit at an age typical of the Rho Ophiuchi cluster. Horizontal bars show the luminosities corresponding to limiting K magnitudes of 15 and 20 at the distance of the Rho Oph cluster. Insert shows the histogram of sources detected in the Barsony et al survey - completeness limit is around 15.

Surveys of dark clouds reveal the initial luminosity function, which can in principle be compared with the field star function. This is currently the subject of some controversy, with apparent differences between results from different environments. For example, Comeron et al. (1993) suggest that the luminosity function increases monotonically below the H burning limit, while Strom et al. (1995) suggest that field brown dwarfs are rare. Even within the well-studied rho Oph complex, the most sensitive current surveys are complete only to K<14 (Barsony et al. 1997). For an age of a few million years, this permits detection of PMS objects only down to a few tenths of a solar mass in a cloud such as rho Oph. As with the brown dwarfs, the UKIRT widefield would probe the IMF down to objects of order a hundredth of a solar mass. It is also possible that the initial luminosity function is spatially variable even within a single star-forming complex (as suggested for rho Oph by Greene & Young 1992), so extensive surveys are required. Currently only a few local clouds (Cha 1, rho Oph, Taurus, Orion, Perseus) have been surveyed, and the variation of the luminosity function with position in the Galaxy remains essentially unexplored.

4.6 Dust Properties in Star-Forming Regions

Star formation in molecular clouds is regulated by the properties of dust grains in these regions. It is well known that the extinction law in dark clouds is different than that in the diffuse ISM, in a sense which is consistent with grain growth in the denser regions. However, there is considerable scatter in the slope of the extinction law as measured by different observers. This is a clear problem: even the fraction of cluster members considered to have infrared excesses is a strong function of the reddening law adopted. Kenyon et al. (1998) emphasized the importance of standard filter bandpasses and good calibration when determining the wavelength dependence of extinction. This in turn depends on accurate calibration of survey photometry, and comparison of large numbers of field stars behind the target cloud with similar numbers of field stars in a nearby control area. All of these requirements will be fulfilled by the UKIRT wide field. The techniques developed by Kenyon et al. permit the use of infrared observations alone (unaccompanied by optical spectroscopy) to determine the infrared extinction law. Thus the UK IR deep survey will be able to determine the variation of extinction in regions too heavily reddened to be accessible to scrutiny at visual wavelengths.

The ability to identify field stars suffering extreme extinction (Av>20m) will also provide sources with clean intrinsic spectra behind the densest regions of dark clouds, which can be used in infrared spectroscopy of the grain constituents in such regions. Only one such object is known at present (Elias 16, situated almost directly behind the dark core TMC-1). Due to the fact that it is not embedded in the core and thus samples unmodified quiescent cloud material, this source has become a benchmark object against which the effect of depletion upon dark-cloud chemistry is tested (due to the detection of solid H20, solid CO, solid C02).

4.7 Source Confusion

Star-count models (e.g. Ortiz & Lepine 1993) predict approximately 104 stars per square degree in Baade's window at a limiting K magnitude of 10; extrapolation to K-20 implies one star (cumulative) per 10 square arcseconds. Thus even in the relatively transparent Baade window, useful survey data may be obtained. Closer to the Galactic Centre, it is likely that survey fields as deep as 20th magnitude would be confusion-limited, but useful observations will be possible at the larger Galactic longitudes where much of the dark-cloud survey work will take place.

4.8 The J-H, H-K colour plane

Observations of distant regions of the Galactic plane are, with a small number of exceptions such as the Baade windows onto the bulge, subject to severe dust extinction at visible wavelengths. The infrared extension of this extinction is reasonably well understood, following an apparently universal power-law decline through the J, H and K bands. The J-H, H-K diagram therefore provides a fundamental tool for discriminating between objects whose red colours indicate only a high degree of reddening, and those with a significant amount of intrinsic thermal emission due to heated dust. The K band is crucial: since dust grains sublime at around 1000K, colours measured in the J and H bands do not reflect the presence of local dust. An infrared survey which does not provide good sensitivity in the K band therefore also dispenses with the only broad-band diagnostic between embedded objects in star-forming regions, and reddened stars in the background field. The fact that the current 2Mass survey telescopes are instrumented to produce simultaneous data in all three near-infrared bands is a recognition of the importance of being able (i) to identify variable sources and (ii) to ensure that colour indices are measured co-temporally.