We compile the first comprehensive census of hot luminous stars in the 30 Doradus (30 Dor) star forming region of the LMC. The census extends to a radius of 10arcmin (150pc) from the central cluster, R136. Stars were selected photometrically and combined with the latest spectral types. 1145 candidate hot luminous stars were identified of which >700 were considered genuine early type stars that contribute to feedback. We assess the spectroscopic completeness to be 85% in outer regions (>5pc) but fall to 35% in the vicinity of R136, giving a total of 500 hot luminous stars with spectroscopy. Stellar calibrations and models were used to obtain their physical parameters before integrated values were compared to global observations and the population synthesis code, Starburst99. The 31 W-R and Of/WN stars made large contributions to the total ionising and wind luminosities of ~40% and ~50%, respectively. Stars with Minit>100Msun also showed high contributions to the global feedback, ~25% in both cases. Such massive stars are not accounted for by the current Starburst99 code, which underestimated the ionising and wind luminosities of R136 by factors of ~2 and ~9, respectively. The census inferred a SFR of 0.073+/-0.04Msun/yr for 30 Dor, typically higher than results from popular SFR calibrations. However, it remained consistent with a far-UV luminosity tracer and a combined Halpha and mid-infrared tracer, but only after correcting for Halpha extinction. The global ionising output exceeded measurements from the associated gas and dust, suggesting ~6(+55/-6)% of ionising photons escape the region. When studying very luminous star forming regions, it is therefore essential to include the most massive stars to ensure a reliable energy budget. If 30 Dor is typical of other large star forming regions, estimates of the SFR will be underpredicted if this escape fraction is not accounted for.(abridged)

We used cosmological simulation with self-consistent radiative transfer to investigate the physical nature of Lyman Limit systems at z=4. In agreement with previous studies, we find that most of Lyman Limit systems are ionized by the cosmological background, while higher column density systems seem to be illuminated by the local sources of radiation. In addition, we find that most of Lyman limit systems in our simulations are located within the virial radii of galaxies with a wide range of masses, and are physically associated with them (``bits and pieces'' of galaxy formation). While the finite resolution of our simulations cannot exclude an existence of a second population of self-shielded, neutral gas clouds located in low mass dark matter halos (``minihalos''), our simulations are not consistent with ``minihalos'' dominating the total abundance of Lyman limit systems.

We use cosmological simulations to explore the large-scale effects of reionization. Since reionization is a process that involves a large dynamic range - from galaxies to rare bright quasars - we need to be able to cover a significant volume of the universe in our simulation without losing hte important small scale effects from galaxies. Here we have taken an approach that uses clumping factors derived from small scale simulations to approximate the radiative transfer on the sub-cell scales. Using this technique, we can cover a simulation size up to $1280 h^{-1} Mpc$ with $10 h^{-1} Mpc$ cells. This allows us to construct synthetic spectra of quasars similar to observed spectra of SDSS quasars at high reshifts and compare them to the observational data. These spectra can then be analyzed for HII region sizes, the presence of the Gunn-Peterson trough and the Lyman-$\alpha$ forest.

We use cosmological simulations of reionization to predict the effect of large HII regions around individual high-redshift quasars on the possible signal from the redshifted 21cm line of neutral hydrogen in the pre-reionization era. We show that these HII regions appear as ``spectral dips'' in frequency space with equivalent widths in excess of 3 mK MHz or depths in excess of about 1.5 mK, and that they are by far the most prominent cosmological signals in the redshifted HI distribution. These spectral dips are expected to be present in almost every line of sight. If the spectral dips of a large enough sample of HII regions are well resolved in frequency space, the distribution of line depth and equivalent width in frequency with a known observing beamsize can be used to infer the HII region size distribution and the mean difference in neutral hydrogen density between the \HII regions (which may contain self-shielded neutral gas clumps) and the surrounding medium, providing a powerful test for models of reionization.