This thesis describes a large -area 2μm survey undertaken on the U.K. Infrared Telescope in Hawaii, using a new infrared imaging camera. The survey covers 5940' and is complete and uniform to a limit of K = 17. The main aim of the survey was to construct a sample of galaxies, selected by their 2μm flux, for studies of galaxy evolution. The subsidiary aim was to survey the 2μm Universe and search for any new populations of infrared objects, such as protogalaxies and brown dwarf stars.
The first half of this thesis is concerned with the analysis of astronomical data. The detailed methods for constructing the infrared survey are described, including flatfielding, astrometry, mosaicing and photometry. Also described are optical CCD observations which cover the survey area to provide optical- infrared colours of almost all the objects in the sample. The methods of reducing, calibrating and matching the CCD data with the infrared data are detailed. For the study of galaxy evolution the redshifts of 53 K band selected galaxies, with a well defined completeness, were measured using multiple object spectrographs. The reduction of the spectra and the methods used for securing the redshifts, including the cross -correlation technique, are described.
The second half of this thesis is about the scientific results from this survey. Firstly, it appears that no new large populations of objects appear, in particular there appear to be no objects only detected in K. A few extremely red objects objects are found, however their true nature is unclear in the absence of spectra. The best evidence is that a few of them are very cool stars. Secondly the galaxy population appears to have very red R -K colours, although not necessarily so red in B - K, and the K band number -magnitude counts exhibit a much smaller excess over the no- evolution prediction than do the B band optical counts. The K band counts are consistent with only a small amount of luminosity evolution.
It is possible to reconcile the B counts, K counts, and the colours of the galaxies, by introducing a model of galaxy merging in which galaxy numbers per unit volume are no longer conserved with lookback time. This model naturally predicts the surprisingly low redshifts found by other workers in B band selected redshift surveys. This model also predicts that the mean redshift of a K selected survey should be less than the no- evolution prediction. This is indeed found - the redshift distribution for the sample studied here is less than the no- evolution model at the 5% significance level.
Finally the limitations of this survey are reviewed, and the prospects for future deep infrared and spectroscopic surveys discussed.