The Canada-United kingdom deep submillimeter survey. II. First identifications, redshifts, and implications for galaxy evolution

Identifications are sought for 12 submillimeter sources detected in a deep submillimeter survey. Six are securely identified, two have probable identi fications, and four remain unidentified with I-AB > 25. Spectroscopic and e stimated photometric redshifts indicate that four of the sources have z < 1 and four have 1 < z < 3, with the remaining four empty-field sources proba bly lying at z > 3. The spectral energy distributions of the identification s, as defined by measurements or upper limits to the flux densities at 8000 Angstrom, at 15, 450, 850 mu m, and at 6 cm, are consistent with the spect ral energy distributions of high-extinction starbursts such as Arp 220. The far-IR luminosities of the sources at z > 0.5 are of order 3 x 10(12) h(50 )(-2) L-., i.e., slightly larger than that of Arp 220. As with local ultral uminous infrared galaxies, the optical luminosities of the identified galax ies are comparable to present-day L*, and the optical morphologies of many of the galaxies show evidence for mergers or highly disruptive interactions . Based on this small sample, the cumulative bolometric luminosity function shows strong evolution to z similar to 1, but weaker or possibly even nega tive evolution beyond. The redshift dependence of the far-IR luminosity den sity does not appear, at this early stage, to be inconsistent with that see n in the ultraviolet luminosity density. Although the computation of bolome tric luminosities is quite uncertain, the population of very luminous galax ies that is detected in the surveys at z>1 is already matching, in the far- IR, the bolometric output in the ultraviolet of the whole optically selecte d population. Assuming that the energy source in the far-IR is massive star s, this suggests that the total luminous output from star formation in the universe will be dominated by the far-IR emission once the lower luminosity sources, below the current far-IR detection threshold, are included. Furth ermore, the detected systems have individual star formation rates (exceedin g 300 h(50)(-2) M-. yr(-1)) that are much higher than seen in the ultraviol et-selected samples and that are sufficient to form substantial stellar pop ulations on dynamical timescales of 10(8) yr. The association with mergerli ke morphologies and the obvious presence of dust makes it attractive to ide ntify these systems as forming the metal-rich spheroid population, in which case we would infer that much of this activity has occurred relatively rec ently, at z similar to 2.