The subject of this thesis is quasars with redshifts greater than seven. We first consider the z = 7.084 quasar ULAS J1120+0641, the most distant quasar discovered to date. We combine new and existing photometric observations of the source to create the multiwavelength SED for ULAS J1120+0641. We measure a bolometric luminosity
$L_bol = 2.6 ± 0.6 × 10 ^ 47 erg s^−1$ for the source, and estimate a star formation rate in the range $60 - 270 M_sol yr ^−1$ from the [CII] line luminosity and the 158 μm continuum luminosity. We find that at the time observed the black hole was growing in mass more than 100 times faster than the stellar bulge, relative to the mass ratio measured in the local Universe.
We present a deep X-shooter spectrum of the Lyman series forest towards ULAS J1120+0641. We detect seven narrow flux transmission spikes in the Lyα forest. We consider the possible detection of a 4.5σ spike in the Lyβ forest. We also present revised Hubble Space Telescope F814W photometry of the source, paying close attention to the issue of charge transfer efficiency losses. We combine the observations to constrain the evolution of the Lyα effective optical depth $(τ _GP^eff)$ with redshift. We find $τ_GP^eff ∝ (1 + z) ^ ξ$ where $ξ = 11.2 +0.4
−0.6$, for z > 5.5. The data can only weakly constrain the hydrogen IGM neutral fraction at $z ∼ 6.5, x_HI > 10^−4$, similar to limits at redshift z ∼ 6. We discuss
difficulties associated with extending Lyα forest analyses to redshifts much greater than six.
The second half of the thesis begins a search for $z >~ 8$ quasars in the UKIDSS Large Area Survey and VIKING. We extend existing Bayesian model comparison techniques to higher redshifts. We assess the plausibility of such a search by simulating realistically sized populations of contaminating galactic L and T dwarf stars, and elliptical galaxies. We also assess the quasar selection function in both surveys. We predict ∼ 2 quasars will be detectable across the two surveys, although the redshift evolution of the luminosity
function is uncertain. Finally, we begin the search for z ∼ 8 quasar candidates using VIKING list-driven
photometry, using $∼ 25 deg ^ 2$ of sky. We use cuts including morphological information to reduce an initial sample of $∼ 1.1 × 10 ^ 6 objects$ to ∼ 8000. Applying our Bayesian model comparison technique to the remaining sources, we find two candidates with $P_q > 0.1$, both of which are quickly ruled out as real quasars. We are nevertheless encouraged to find zero candidates for follow-up observations, as the implied number of good candidates will be manageable over the full area of VIKING.