THE HUBBLE DEEP FIELD AND THE DISAPPEARING DWARF GALAXIES

Several independent lines of reasoning, both theoretical and observati onal, suggest that the very faint (B greater than or equal to 24) gala xies seen in deep images of the sky are small low-mass galaxies that h ave experienced a short epoch of star formation at redshifts 0.5 less than or equal to z less than or equal to 1 and have since faded into l ow-luminosity, low-surface-brightness objects. Such a scenario, which arises naturally if star formation in dwarf galaxies is delayed by pho toionizaton resulting from the metagalactic UV radiation field, provid es an attractive way to reconcile the Einstein-de Sitter (Omega = 1; L ambda = 0) cosmological model with the steeply rising galaxy counts ob served at blue wavelengths. Babul & Ferguson constructed a specific re alization of this model, deriving the dwarf galaxy mass function from the cold dark matter (CDM) power spectrum, and arguing that the gas in dwarf galaxy haloes will recombine at z similar to 1. The Hubble Deep Field (HDF) images provide a stringent test of this model. We compare the model with the data by constructing simulated images that reprodu ce the spatial resolution and noise properties of the real data, and b y carrying out source detection and photometry for the simulations in the same way they were carried out for the real data. The selection bi ases and systematic errors that are inevitable in dealing with faint g alaxies are thus incorporated directly into the model. We compare the model predictions for the counts, sizes, and colours of galaxies obser ved in the HDF to observations and to the predictions from a low-q(0) pure-luminosity-evolution (PLE) model. Both models fail to reproduce t he observations. The low-q(0) model predicts far more Lyman-break 'dro pouts' than are seen in the data. The fading dwarf model predicts too many remnants: faded dwarf galaxies in the redshift range 0.2 < z < 0. 5 that should be detectable in the HDF as low-surface-brightness red o bjects but are not seen. If fading dwarf galaxies are to reconcile the Einstein-de Sitter geometry with the counts, then the dwarf populatio n must (i) form earlier than z similar to 1, with a higher initial lum inosity; (ii) have an initial-mass function more heavily weighted towa rd massive stars than the Salpeter IMF; or (iii) expand much more than assumed during the supernova wind phase.