Counts and Sizes of Galaxies in the Hubble Deep Field South: Implications for the Next Generation Space Telescope^*

Science objectives for the Next Generation Space Telescope (NGST) include a large component of galaxy surveys, both imaging and spectroscopy. The Hubble Deep Field data sets include the deepest observations ever made in the ultraviolet, optical, and near-infrared, reaching depths comparable to that expected for NGST spectroscopy. We present the source counts, galaxy sizes, and isophotal filling factors of the Hubble Deep Field South (HDF-S) images. The observed integrated galaxy counts reach over 500 galaxies per square arcminute at magnitudes AB < 30. We extend these counts to fainter levels and further into the infrared using galaxy-count models. It was determined from the HDF (North) and other deep Wide Field Planetary Camera 2 imaging that fainter galaxies are smaller. This trend continues to AB = 29 in the high-resolution HDF-S Space Telescope Imaging Spectrograph (STIS) image, where galaxies have a typical half-light radius of 01. We have run extensive Monte Carlo simulations of the galaxy detection in the HDF-S, and we show that the small measured sizes are not due to selection effects until AB > 29. We compare observed sizes in the optical and near-infrared using the HDF-S Near Infrared Camera and Multi-Object Spectrometer image, showing that after taking into account the different point-spread functions and pixel sizes of the images, galaxies are smaller in the near-infrared than they are in the optical. We analyze the isophotal filling factor of the HDF-S STIS image and show that this image is mostly empty sky even at the limits of galaxy detection, a conclusion we expect to hold true for NGST spectroscopy. At the surface brightness limits expected for NGST imaging, however, about a quarter of the sky is occupied by the outer isophotes of AB < 30 galaxies, requiring deblending to detect the faintest objects. We discuss the implications of these data on several design concepts for the NGST near-infrared spectrography. We compare the effects of resolution and the confusion limit of various designs, as well as the multiplexing advantages of either multiobject or full-field spectroscopy. We argue that the optimal choice for NGST spectroscopy of high-redshift galaxies is a multiobject spectrograph (MOS) with target selection by a microelectromechanical systems (MEMS) device. If this technology does not become available in the next few years, then the second choice would be either a mechanical MOS using movable slits or fibers, or an integral field spectrograph.