Seeing Galaxy Formation in a Cold Light


The SCUBA instrument consists of two sets of bolometers which simultaneously observe at a long wavelength (here 850µ) and a short wavelength (here 450µ). The long array contains 37 bolometers and the short 91, and since both arrays undersample the field the arrays are rastered to form a final properly sampled image. To cancel system and sky variations the image is rapidly chopped, and slightly more slowly nodded. For the present images these cancellations were made with a fixed offset of 40" in the right ascension direction. Each object therefore appears as a positive image with two negative images at +/- 40" in the E-W direction, each of half the source strength. The diffraction limit is 14" at 850µ and 7.5" at 450µ so there are approximately 100 independent beams in the 850µ image and about 350 in the 450µ image; each of the images is roughly 2.3 arc minutes in diameter.
Deep field images of random points on the sky are intended to yield the 'typical' properties of the galaxy population, and in particular to determine the number of sources at each flux level. This kind of information tells us what types of objects are producing the light which is seen in the submillimeter, and, by identifying individual objects, allows us to find out their properties at other wavelengths and to determine, if possible, their redshifts.
Sources that are bright enough may be found directly on the images, and the number counts (the number above a given flux in a given area) determined. For these 50-hour exposures, the 1 sigma limits for a direct source detection in the centers of the field are between about 0.4 and 0.7 mJy, depending on the methodology used to define the noise. Source recovery may be simulated in various ways to check this number. Combining the two fields in Barger et al. with the single field in Hughes et al. we find 5 sources above 3mJy and 1 source above 6 mJy in just over 16 square arcminutes of sky. At these flux levels all sources added to the fields are easily recovered so these numbers directly determine the counts at these fluxes, which are shown by the crosses in the plot below. There is still substantial small-number uncertainty (the error bars are the ± 1 sigma errors, based on the observed numbers of objects) but the counts are roughly detemined at useful levels.
The red curves show model counts from Guiderdoni et al. The red solid line shows one of their models (A) in which the submillimeter counts arise from dust in optical galaxies. This substantially underpredicts the counts. Alternative models with a large population of ultraluminous submillimeter galaxies (E), such as the dashed line can match the counts and the submillimeter background. Such models predict that the bulk of the galaxies selected at 850µ will lie at large redshift.
Counts at fainter magnitudes can also be constrained by looking at the fluctuations they produce in the image. If the faint end sources rise too rapidly, as is the case in the simulated images below (where N(>S) = 18000 (S/mJy)^(-1) counts per square degree), then the images will be too structured. The four panels on the right show Monte Carlo simulations of fields with this source count, which can be compared with the actual SCUBA images of the two Hawaii fields, on the right. The purple shaded region in the plot of the counts shows the 95% confidence region of the counts at fainter fluxes, obtained from a fluctuation P(D) analysis on the two Hawaii fields.