Star formation in high-density environments in the early Universe

After struggling with the first batch of SCUBA's photom.t data - a painful effort to convince ourselves that a coadded flux for 8C1435+635 that bounced from a big, fat zero mJy one night to 20mJy the next night could in fact be trusted to the required sub-mJy accuracy (Ivison et al. 1998, ApJ, 494, 211) - we became converts to map64.t. Even if it did take umpteen times longer to reach the required flux levels, seeing a beam-sized source appear at the expected position gave tremendous faith in a source's reality and the potential for serendipitous detections was too tempting to pass up, particularly at a time when submm cosmology could boast a measly dozen submm-selected galaxies.

Subsequent efforts to nail high-redshift radio galaxies with SCUBA involved both photom.t (Archibald et al. 2000, astro-ph/0002083) and map64.t (Ivison et al. 2000, astro-ph/0005234). In the mapping project, serendipitous detections came thick and fast; the rate of detections seemed to far exceed those of blank-field surveys, and it quickly became apparent that there might be an over-density of submm sources in the fields of distant radio galaxies (e.g. Figure 1).

Figure 1: a) Full 850-micron image of the z = 3.8 radio galaxy, 4C41.17; b) 450-micron image of 4C41.17, smoothed to a resolution of 10'' FWHM; c) Central, cleaned portion of the 850-micron image.

Indeed, biased galaxy-formation theories predict that massive galaxies at high redshifts should act as signposts to high-density environments which subsequently evolve into the cores of the richest clusters seen at the present day. These regions are expected to be characterised by over-densities of young galaxies, probably including a population of dusty, interaction-driven starbursts - the progenitors of massive cluster ellipticals.

It seems quite plausible that by mapping high-redshift AGN we have stumbled upon this population of clustered submm galaxies. We have undertaken searches in the fields of half-a-dozen radio galaxies and quasars at z ~ 4, with the same number of fields slated for 2000/01. Our maps are extremely deep - down at the mJy rms level at 850 microns - and they reveal an order-of-magnitude over-density of luminous submm galaxies compared to typical fields. The likelihood of finding such an over-density in random fields is <0.002.

If we believe that the redshifts are the same as the signpost galaxies - an aspect of the study that we are working hard on via deep imaging with OVRO, IRAM, VLA, WHT, Gemini and Keck - then they have bolometric luminosities, >10^13 L_o, which implies star-formation rates (SFRs) consistent with those required to form a massive galaxy in a fraction of a Gyr.

Our target fields also appear to exhibit over-densities of extremely red objects (EROs), some of which may be associated with the submm sources (Figure 2), and Lyman-break galaxies. In a paper detailing the acquisition and analysis of our data in the 4C41.17 field (Ivison et al. 2000, astro-ph/0005234) we have proposed that the over-densities of both submm and ERO sources in these fields represent young dusty, starburst galaxies forming within proto-clusters centred on the high-redshift radio galaxies and quasars, clusters which are also traced by a less-obscured population of Lyman-break galaxies.

Figure 2: KPNO 2.1-m K'-band imaging and HST F702W images of 25x25'' regions around the submm sources near 4C41.17. For HzRG850.1, the optical data are from Keck II. EROs are marked with squares; 850-micron positions are marked with 6''- or 8''-diameter black circles (an ellipse for HzRG850.2); the 450-micron position of HzRG850.1 is marked with a white circle.

Figure 3: K'-band images of regions around 4C41.17. 850-micron data are shown as contours. Red solid circles denote EROs (R-K > 6). Two EROs are probably associated with the blended sub-mm galaxy, HzRG850.2. For the other sub-mm galaxies (HzRG850.1, 4C41.17), green dashed circles denote the likeliest counterparts (faint and red, in the case of HzRG850.1, though a bona fide ERO could also be responsible for the sub-mm emission).

We have introduced a nomenclature for their classification, analogous to that used for proto-stars (Andre, Ward-Thompson & Barsony 1994, ApJ, 406, 122) and building upon the evolutionary scheme for ULIRGs proposed by Sanders et al. (1988, ApJ, 325, 74). For operational purposes we based this scheme on the typical depths achieved in follow-up observations (I ~ 26, K ~ 21) and the properties of typical submm galaxies (Ivison et al. 2000, MNRAS, 315, 209).

We defined the following classes: Class 0, very-highly obscured sources, where there is no plausible optical or near-IR counterpart; Class I, highly obscured sources, where only a near-IR counterpart exists (often EROs); and Class II, where an obvious optical counterpart is seen (IIa: pure starburst; IIb: type-II AGN; IIc: type-I AGN). The latter class may overlap with the most massive examples of Lyman-break objects.

Figure 1 demonstrates that the 4C41.17 field contains several extremely luminous submm galaxies, possibly in a structure associated with the radio galaxy. The density of sources brighter than 8mJy is 1220 +/- 860/deg^2, well above the weighted mean 8-mJy blank-field count (134 +/- 57/deg^2). The standard probabilistic methodology used in studies of clustering and confusion suggests that on average we would have to observe more than 600 typical blank fields before finding a configuration of the type seen towards 4C41.17.

Turning to the optical and near-IR imaging of this field, there are several possible counterparts to HzRG850.1, the bright submm galaxy to the south-east of the map centre (Figures 2 and 3). The bluest is 850.1.K1 (Class II), a source detected in U and thus probably a low-redshift, blue galaxy, unlikely to be associated with the SCUBA source. Another two galaxies, 850.1.K2 and 850.1.K3 (both Class I) have similar very red colours, V-K ~ 7. The similar colours of 850.1.K2 and 850.1.K3, combined with the fact that HzRG850.1 is spatially extended, suggests that both these galaxies may be jointly responsible for the submm emission; however, we also note some very faint emission in V and K to the west of 850.1.K2 that underlies the submm emission, and in V alone to the south.

For HzRG850.2, to the north-east of the map centre, three faint galaxies are visible in K. The bluer of these, 850.2.K1, lies just within the error ellipse, while the others, 850.2.K2 and 850.2.K3, are much redder, R-K > 6.4, and lie on the edge of the nominal error ellipse. In a deep HST F702W image, both 850.2.K1 and 850.2.K2 are each resolved into two components, while 850.2.K3 cannot be seen (R > 26). For 850.2.K1 the sub-components both appear to be galaxies, while in 850.2.K2 they may either be two galaxies or a single system crossed by a dust lane. As with HzRG850.1, it is extremely plausible that both of these counterparts may be contributing to the submm emission from HzRG850.2, given its morphology.

The remaining SCUBA sources lie in comparatively shallow regions of the near-IR and optical images and do not show unambiguous identifications.

In summary, HzRG850.1 and HzRG850.2 are likely to be associated with EROs (Class I sources). It is interesting that our success in obtaining candidate identifications, in particular the association with EROs, correlates with the observed flux density of the submm emission. Is it possible that the brightest submm sources are typically associated with Class I or II counterparts (e.g. HR10, SMM J09429+4658, SMM J02399-0136, SMM J14011+0252, SMM J00266+1708 as well as the brightest sources found by the UK's 8-mJy survey, the Dutch cluster survey and the latest 1.3-mm survey with MAMBO at IRAM) whereas fainter submm sources typically have Class 0 counterparts.

Can these submm sources can be unambiguously placed at high redshifts, strengthening the case for their association with 4C41.17? We are reliant on redshift-sensitive parameters: the 450- to 850-micron flux ratio and the 850-micron to 1.4-GHz flux ratio. The observed 450- to 850-micron flux ratios for 4C41.17 and HzRG850.1 are 3.2 +/- 0.8 and 2.2 +/- 0.6, consistent with their being at the same redshift, with best-fit values z ~ 3.8 and z ~ 4.8. These limits are supported by constraints from the 850-micron to 1.4-GHz flux ratio which yields a robust limit of z > 1.2 for HzRG850.1. The limits for HzRG850.2 and HzRG850.3 are slightly weaker, but they are certainly not local galaxies.

The extreme colours, V-K ~ 7, of the probable galaxy counterparts to HzRG850.1 and HzRG850.2 also suggests that these systems lie at high redshift. If these galaxies do lie at z = 3.8 then they are very luminous, in excess of 50 L*, although we note that luminosities close to this are found for the confirmed Class II counterparts SMM J02399-0136 (z = 2.8) and SMM J14011+0252 (z = 2.6).

We conclude that the available constraints suggest that the over-density of bright submm galaxies, at least in the 4C41.17 field, is likely to lie at z > 2.8 and is thus consistent with being associated with the radio galaxy (though spectroscopic confirmation is clearly a top priority). This suggests that in addition to the over-density of Lyman-break galaxies identified around 4C41.17, we should also add a comparably numerous population of highly-obscured and very luminous submm galaxies.