What produces the ionizing background at large redshift?

The absence of a Gunn-Peterson trough in the spectra of high-redshift quasars requires the intergalactic medium to have been highly ionized at an early epoch. Photoionization from quasars detected at high redshift in optical surveys is insufficient to ionize the medium to the required level (Shapiro and Giroux). While the Gunn-Peterson test might be alternatively satisfied by assuming that the medium has been shock-heated by explosions (e. g. , supernova outbursts in galaxies), the proximity effect (Bajtlik, Duncan, and Ostriker) gives an independent estimate of the intensity of the ionizing background. We show that the observed quasars also fail to emit sufficient ionizing radiation to satisfy this constraint. Thus, even if the IGM was shock-heated, other sources of ionizing radiation are required at high redshift. We consider two candidates for these additional sources: main-sequence, high-mass stars in primordial galaxies, and other quasars which are not seen because of dust obscuration by intervening galaxies. We show that both models can satisfy all the observational constraints with reasonable parameters and indicate how future observations can distinguish between the two possibilities. If the intensity of the ionizing background is not higher than the value indicated by the proximity effect, models in which photoionization is the only energy input into the intergalactic medium have to be stretched to the limits not to violate the Gunn-Peterson test from the high-redshift quasars, and it may soon become evident that shock heating of the IGM is required whatever the source of the photon field. We show that the spectrum of the ionizing background at large redshift should have a break at the He II ionization edge due to absorption of radiation by Lyα clouds. We propose that observations of helium lines from Lyα clouds or an analysis of line ratios in metal line absorption systems may be able to tell us what sources are producing the ionizing background. If OMEGAb_ is as given by light-element nucleosynthesis, then models for galaxy formation in which the bulk of the star formation occurs later than z ~ 2. 5 are inconsistent with the spectra of high-redshift quasars, because of the lack of sufficient sources of ionizing radiation at high redshift.