Stephen Gwyn's Ph.D. Page

The Hubble Deep Fields (HDF's) are the deepest optical images yet obtained, so deep in fact that most of the galaxies in the fields are too faint to be observed spectroscopically with current telescopes. Only the brightest 10 percent of the galaxies have spectroscopic redshifts. To learn more about the remaining 90 percent, one must use photometric redshifts.

Rather than observing narrow spectral features of galaxy spectra, such as lines, the photometric redshift technique concentrates on broad features, such as the 4000 Angstrom break. In this method, the photometry of observed galaxies is converted into low resolution spectral energy distributions (SED's). Redshifts are determined by comparing these SED's to redshifted template galaxy spectra. This method may be used in the HDF's to down to a magnitude limit of I=28. Although photometric redshifts are not as precise as their spectroscopic counterparts, the large sample (over 2000 galaxies) and the unprecedented depth of the Hubble Deep Fields allow one to trace the evolution of several properties of galaxies from z=5 to the present in a statistical manner. My thesis studied four such aspects:

1. The clustering of galaxies was examined by measuring the projected spatial correlation function. The clustering signal is rather weak in the small area of the Hubble Deep Fields. It was found that there is a slight increase in clustering around z=0.5 in the HDF-North relative to the HDF-South. When the redshift distributions of the HDFN and the HDFS are compared, one finds a significantly greater number of galaxies at this same redshift. This suggests the presence of a structure (a very weak cluster or a very strong group) in the HDF-North.
2. The star formation rate density (SFRD) was determined by measuring the UV-luminosity density. After correcting for dust extinction, the star formation rate was found to decrease exponentially with time with e-folding rate of about 4 Gyr. This result, based on a single method of determining the SFRD on a homogeneous sample extending from z=0 to z=4.5, is consistent with the various results in the literature, which are based on a wide variety of methods and samples at different redshifts. As an aside, three low-redshift samples of galaxies, were used to study the dependence of the amount of dust extinction with galaxy type. Here "type" is measured either by colour or by visual morphology. It was found that, although there is considerable scatter, the average extinction is the same for all galaxy types.
3. The B-band galaxy number and luminosity densities were studied simultaneously to examine the merging history of the Universe. The number density measures the total number of galaxies there are in the universe, while the luminosity density effectively measures the total number of stars. When two galaxies merge into a single object, the total number of stars is conserved. It was found that, while the total B-band luminosity density of universe decreases only slightly with time since z=4.5, the number density of galaxies drops considerably more. This difference in rates of declines makes it possible to quantify the merger rate.
4. The morphology of galaxies was quantified using a "lumpiness" parameter: L. This parameter measures the number of local maxima in the image of a galaxy. A smooth object, such as an elliptical has L=1 while a active star-forming galaxy with many HII regions might have a value of L=30. Rest-frame B-band images were made of both HDF's by K-correcting each pixel of each galaxy in the frames using the photometric redshifts of the parent galaxies. It was found that L increases with increasing apparent brightness, increasing absolute brightness and increasing redshift, albeit only slightly. The first correlation is an observational effect while the second two are physically meaningful. While the brightest high redshift galaxies have disturbed morphologies, the more typical object is more uniform and compact.

My thesis is available in:

• Postscript format
• PDF format