Andy Connolly (JHU), and a group of collaborators including Mark Dickinson (JHU), Matt Bershady (Penn State), Peter Eisenhardt (JPL), Richard Elston (Florida), and Adam Stanford (LLNL) have used KPNO 4-m IRIM observations of the Hubble Deep Field (HDF) to show that star formation in the universe peaked at z = 1.5, about when the universe was ~ 40% of its present age. Although redshifts estimated from the original HDF observations, themselves, had suggested that the maximal star-forming phase of the universe had occured between 1 < z < 2, the IR observations are critical for actually probing the star-formation rates over this interval.
The Hubble Deep Field has provided a unique probe of galaxy evolution. As valuable as the HST observations are, however, reading the cosmological history sampled by the HDF has also benefitted from extensive follow-on projects, such as the KPNO near-IR observations. Over the course of ten days from 27 April-6 May 1996 Matt Bershady, Mark Dickinson, Peter Eisenhardt, Richard Elston, and Adam Stanford observed the HDF using the IRIM camera on the KPNO Mayall 4-m Telescope. A total of 11.0, 11.3 and 22.9 hours of data were collected for the HDF in the J, H and Ks passbands. The formal magnitude limits for a 2'' aperture and a signal-to-noise ratio of five are 23.45, 22.29, and 21.92 in the J, H and Ks filter respectively. The IRIM field-of-view of 2.56' is well matched to that of the WFPC2 images. Figure 1 shows the composite IRIM image of the HDF. The near-infrared data contain approximately 300 sources and are available from http://www.stsci.edu/ftp/science/hdf/clearinghouse/irim/irim_hdf.html.
Using Keck spectroscopic redshifts as a training set, a correlation was
determined between the optical and near-infrared photometric observations
and the spectroscopic redshifts. The importance of including the near-IR
data as opposed to just the optical HST data can be understood as follows.
For redshifts z < 1 the dominant feature that provides the photometric
redshift is the break in the galaxy continuum at ~ 4000 Å as it passes
through the WFPC2 F450W and F606W filters. For redshifts greater than 1,
this break moves out of the optical spectral region and into the
near-infrared. For star-forming galaxies, the ultraviolet continuum from
Lyman (1215 Å) to ~ 3000 Å is relatively devoid of strong features, and
consequently there is little information in optical photometry from which to
estimate redshifts at 1 < z < 2. At redshifts greater than 2.2, the 912 Å
Lyman limit enters the U band, and redshifts can again be estimated in an
analogous manner to those from the 4000 Å break (Steidel et al. 1996, Madau
et al. 1996). The near-IR photometry bridges this gap by allowing
observation of the 4000 Å break over the 1 < z < 2 region. With the addition
of deep J-band data alone, one can detect the 4000 Å break out to z = 2.1.
From the photometric redshift sample, Connolly et al. construct the comoving
luminosity density of galaxies, at 2800 Å, for three redshift intervals:
0.5 < z < 1, 1 < z < 1.5, and 1.5 < z < 2. After correcting for
incompleteness in the sample, assuming that the galaxy distribution is well
matched by a Schechter luminosity function with a slope
of = 1.3, the luminosity densities were converted to metal enrichment
rates (MER) using the spectral synthesis models of Bruzual and Charlot.
Figure 2 shows the MER for the redshift range 0 < z < 4 using data from
Gallego et al. (1995), Lilly et al. (1996), Madau (1996) and the new
Connolly et al. photometric redshift sample.
The spectroscopic, photometric and Lyman break galaxy samples provide a remarkably consistent picture of the evolution of the star formation history as a function of redshift. There is a rapid rise in the star formation rate as we look back in cosmic time to a peak at a redshift of 1.5. From there it then falls by a factor of two out to a redshift of 3. While this rapid evolution is well matched by the number-magnitude relation for galaxies it remains a challenge for the theoretical modeling of galaxy formation and evolution.