Photometric redshifts in the Hubble Deep Fields: evolution of extinction and the star formation rate

Photometric redshifts are studied with a template approach using data from the Hubble Deep Field (HDF) North and South. The problem of aliasing in photometric redshift estimates is investigated in some detail and found not to be a significant problem if at least four photometric bands are available. The performance of the approach presented here appears to exceed that of others in the literature. The rms accuracy of the photometric redshifts is 9.6 per cent in (1 + z), with a 1.5 per cent chance of a significant alias when four or more photometric bands are used. With reasonable restrictions, it is possible to determine the dust extinction as well as the photometric redshift, provided five or more photometric bands are available. An important result is that evolution of (A V (z)) with redshift is seen, with higher values than locally at z = 0.5-1.5, and lower values at z > 2. This is consistent with current models for the star formation history of the Universe. Deconvolving the ultraviolet-to-infrared (UV-to-IR) spectral energy distributions (seds) into an old-star and young-star component allows the determination of M * and M * for each galaxy, as well as z p h o t and A V , provided that infrared photometric bands are available. The expected trend of b = M * /M * t 0 increasing to the past is seen. There is, however, a great deal of scatter in the relation between b and sed type, showing that the recent star formation history is not very well correlated with the long-term history of a galaxy. The 2800-A luminosity function and star formation rate are calculated for a large sample of HDF-N (2490) and HDF-S (28 719) galaxies, using photometric redshifts, for the redshift range 0.2-5. The star formation rates agree reasonably well with those from a variety of other UV, Ha, IR and submillimetre estimates, and with star formation histories used to model optical, IR and submillimetre source counts.