Early-type galaxies in the Hubble deep field: The star formation history

We have investigated the properties of a complete K-band selected sample of 35 elliptical and S0 galaxies brighter than K = 20.15 mag in the Hubble deep field, as representative of the field galaxy population. This sample has been derived from deep K-band image by the KPNO-IRIM camera, by applying a rigorous morphological classification scheme based on quantitative analyses of the surface brightness profiles. The completeness of the sample is proved by a careful evaluation of all biasing effects inherent in the automated selection procedure. Fifteen objects have spectroscopic redshifts, while for the remaining 20 a photometric redshift is estimated from a seven-color broadband spectrum (including four HST and three near-IR bands). This data set, based on public archives from HST and from deep observations at Kitt Peak and Hawaii, is unique for its morphological information and for its photometric and spectroscopic coverage. The broadband spectra of the sample galaxies, together with a few basic assumptions about the IMF and the stellar evolutionary paths, allow us to date their dominant stellar populations. The majority of bright early-type galaxies in this held are found at redshifts z less than or similar to 1.3 to display colors indicative of a fairly wide range of ages (typically 1.5 to 3 Gyr). Because of the different cosmological timescales, the star formation history depends to some extent on the assumed value for the cosmological deceleration parameter: we find that the major episodes of star formation (SF) building up typical M-star galaxies have taken place during a wide redshift interval 1 < z < 4 for q(0) = 0.5, which becomes 1 < z < 3 for q(0) = 0.15. There seems to be a tendency for lower mass (M < 5 x 10(10) M.) systems to have the bulk of their SF protracted to lower redshifts. Our estimated galactic masses, for a Salpeter IMF, are found in the range from a few similar to 10(9) M. to a few 10(11) M. already at z similar or equal to 1. So the bright end of the E/S0 population is mostly in place by that cosmic epoch, with space densities, masses, and luminosities consistent with those of the local field E/S0 population. We argue that the strong decrease of the comoving mass density of early-type galaxies found by some authors already by z similar or equal to 1 might be due to improper color classification, since these objects are usually found to display blue young populations mixed with old red stars. Instead, what distinguishes the present sample is a remarkable absence of objects at z > 1.3, which should be detectable during the luminous star formation phase expected to happen at these redshifts. Obvious solutions are (1) that the merging events triggering the SF imply strongly perturbed morphologies that prevent selecting them by our morphological classification filter, or (2) that a dust-polluted interstellar medium (ISM) obscures the (either continuous or episodic) events of star formation, after which gas consumption (or a galactic wind) cleans up the galaxy. We conclude that the likely solution is a combination thereof, i.e., a set of dust-enshrouded merging-driven starbursts occurring during the first few gigayears of the galaxy's lifetime. While our main conclusions are moderately dependent on the assumed value of q(0), an open universe is favored in our analysis by the match of the K-band local luminosity functions with the observed numbers of faint distant galaxies. Two sources of uncertainty in our analysis, i.e. , the possible presence of a background cluster or group at z similar to 1 in the HDF (possibly contaminating the z distribution) and the lack of a complete spectroscopic identification, are shown unlikely to affect our main results. In any case, they will be reduced soon by new observations in the southern HDF and by deep spectroscopic surveys with large telescopes