Photometric redshifts and properties of galaxies from the sloan digital sky survey

The determination of photometric redshifts is essential for many
subjects in cosmology and extragalactic astronomy, like the large
scale structure of the Universe, gravitational lensing, or galaxy
evolution. If the spectral energy distribution (SED) of a galaxy is
measured with high enough spectral resolution, the redshift can be
easily derived through the absorption and emission lines which are
created by the elements in the galaxy. However, currently more
telescopes are equipped with large cameras with charged coupled
devices (CCDs) that observe the sky through optical filters. With
these photometric observations it is possible to detect much
fainter astronomical objects than with spectroscopy. Furthermore,
photometric observations are less time consuming and cheaper in
comparison, wherefore they are preferentially used for observations
of statistical meaningful cosmological volumes. Nonetheless,
photometric data, which are often gained by observations through
broadband filters, are not as precisely resolved as spectra.
Therefore one does not have information about the accurate position
in wavelength of spectral lines, but only about the overall shape
of the SED. This is the reason why so-called photometric redshifts
have to be derived by statistical means. One approach to estimate
the redshift through photometry alone are template fitting methods
which compare the fluxes predicted by model spectra with the
observations. After that, a likelihood analysis is performed with
which a probability density function P(z) and the most probable
value of z can be derived. To achieve high accuracies with
photometric redshift template fitting techniques, the model spectra
as well as their corresponding prior probabilities have to be
chosen carefully. In this work I use photometric and spectroscopic
data of luminous red galaxies from the Sloan Digital Sky Survey
(SDSS). I analyze the precision of photometric redshifts estimated
with model SEDs specifically designed to match the set of luminous
red galaxies of SDSS-II at redshifts z ≤ 0.5 in color and I compare
them with published results. These models were created without
information on their properties at wavelengths shorter than the
SDSS u band. However, the galaxy UV characteristics derived from
the model SEDs match those of other observations. Furthermore, I
investigate the SED properties derived from the best fitting models
with respect to spectroscopic data as functions of redshift and
luminosity. At lower redshifts less luminous galaxies from our
sample on average show increased signs of star formation in
comparison to galaxies with higher luminosities. This is supported
by analyses of the line strengths in the spectra. Moreover, star
formation activity increases with increasing redshift which is
caused by the aging of the galaxy population from higher to lower
redshifts. I also generate model spectra for red galaxies from the
SDSS-III located at even higher redshifts 0.45 ≤ z ≤ 0.9. For this
I modify the shape of theoretical spectra to match the data of the
analyzed galaxies to a better extent. The multidimensional space
defined by the colors and the absolute magnitude of the galaxies is
reduced to two dimensions through a self-organizing map. The map is
then partitioned by a k-means algorithm which identifies clusters
in the data. From the cluster cells I select model spectra which
represent the galaxies from within the same cell. A selection of
the models is then used as a template set for photometric redshift
estimation. I find that our models improve the redshift accuracy in
comparison to the results published by SDSS.