2006MNRAS.370.1057S

Since the advent of Submillimetre Common-User Bolometer Array (SCUBA) on the James Clerk Maxwell Telescope (JCMT), a series of complementary surveys has resolved the bulk of the far-infrared extragalactic background into discrete sources. This has revealed a population of heavily dust-obscured sources at high redshift (z > 1) undergoing an intense period of massive star-forming activity with inferred star formation rates of several hundred to several thousand solar masses per year. Taken together, these existing surveys cover a total area of 460 arcmin² to a range of depths, but combining the results has hitherto been complicated by the fact that different survey groups have used different methods of data reduction and source extraction. In this paper, we re-reduce and re-analyse all of the blank-field surveys to date in an almost identical manner to that employed in the `SCUBA 8-mJy Survey'. Comparative source catalogues are given which include a number of new significant source detections as well as failing to confirm some of those objects previously published. These new source catalogues have been combined to produce the most accurate number counts to date from 2 to 12.5mJy. We find N(>4mJy) = 620⁺¹¹⁰_–190deg^–2,N(>6mJy) = 310⁺⁶⁰_–100deg^–2,N(>8mJy) = 150⁺⁴⁰_–60deg^–2 and N(>10mJy) = 40⁺²⁰_–20deg^–2 after correcting for the effects of incompleteness, flux-density boosting and contamination from spurious/confused detections. Furthermore, the cumulative number counts appear to steepen beyond S₈₅₀> 8mJy, which could indicate an intrinsic turnover in the underlying luminosity function placing an upper limit on the luminosity (and hence mass) of a high-redshift galaxy. We have also investigated the clustering properties of the bright S₈₅₀> 5mJy SCUBA population by means of two-point angular correlation functions. We find a ≃3.5σ excess of pairs within the first 100 arcsec over that expected from a Poisson distribution. Fits of a standard power law of the form w(θ) = A θ^–δ to the angular correlation functions for S₈₅₀ > 5mJy are limited in accuracy by the small number of source detections but appear to be broadly consistent with those measured for extremely red objects. Nearest-neighbour analyses further support strong clustering on arcmin scales, rejecting the null hypothesis that the distribution of the submillimetre sources is random at the 95 per cent confidence level for S₈₅₀> 5mJy, and at the 99 per cent confidence level for S₈₅₀> 7mJy.