Modeling and interpretation of galaxy spectra: the stellar populations of nearby galaxies

Our current understanding of structure formation in the Universe
seems to be well described by a hierarchical scenario, in which
small units assemble first to produce more massive systems. In
recent years, much observational evidence has been accumulated,
indicating that star formation proceeded instead in an
antihierarchical fashion. Constraining the age and chemical
composition of the stellar populations in galaxies should help shed
light on this apparent dichotomy between mass assembly and star
formation activity. The integrated spectra of galaxies contain
valuable clues about the ages and metallicities of the stars
producing the light. However, at first order, they are affected in
a similar way by age and metallicity. Studies of more refined
spectral diagnostics, such as individual stellar absorption
features, are thus needed to provide more stringent constraints on
these parameters. This method has been limited so far to small
samples of elliptical galaxies, using population synthesis models
with limited spectral resolution and restricted coverage in stellar
effective temperatures. The objective of this thesis is the
interpretation of the optical spectra of large samples of nearby
galaxies in terms of the light-weighted metallicity, age and mass
of their stellar populations. I have developed a new method to
simultaneously derive median-likelihood estimates of each physical
parameter and the associated confidence intervals. The method,
based on a recent highresolution population synthesis code with
full temperature coverage, consists in comparing each observed
spectrum with a comprehensive library of star formation histories.
The constraints are set by the simultaneous fit of an optimally
selected set of spectral absorption features. I have applied this
method to a sample of 200,000 galaxies from the Sloan Digital Sky
Survey, including galaxies with any star formation history, from
quiescent early-type to actively star forming galaxies. Thanks to
the unprecedented statistics, I could give an accurate description
of the galaxy distribution in the full physical parameters space.
The relation between stellar metallicity, age and stellar mass
shows a rapid transition from low-mass, young, metal-poor to
high-mass, old, metal-rich galaxies at a stellar mass of 3×10^10
solar masses, the same characteristic scale of several observed
bi-modalities in galaxy properties. The stellar metallicity-mass
relation is interpreted as a manifestation of galactic winds, which
are more efficient in removing metals from the shallow potential
well of low-mass galaxies. I then explored the implications of the
above relations to re-assess the physical origin of observed
scaling relations of elliptical galaxies, linking their luminous
and dynamical mass to the properties of their stellar populations.
The relations are driven by an increase in metallicity, age and
element abundance ratios with galaxy mass. The scatter is
contributed by a similar amount by both age and metallicity. The
increasing spread towards younger ages at low stellar masses
indicates that low-mass ellipticals either formed their stars later
or have a more extended star formation history. This hints at a
shift in stellar growth towards less massive galaxies in recent
epochs. The large ranges in observational and physical properties
covered by SDSS galaxies make it a representative sample of the
local Universe. I could thus derive the total mass density of
metals and baryons locked up in stars today. I have also studied
how metals and stellar mass are distributed as a function of
various galaxy properties. The galaxies containing the bulk of the
total stellar mass (massive, bulge-dominated galaxies with old
stellar populations) are also those that contribute the largest
fraction of metals, as expected from the mass-metallicity relation.
These quantities set the fundamental constraints at the present
epoch of the cosmic star formation and chemical enrichment
histories. The more detailed knowledge of the relations between
galaxy physical parameters allows a more direct comparison with
predictions from semi-analytic models of galaxy formation and
evolution. Moreover, the more robust constraints represent an
important calibration at redshift zero for similar studies at
higher redshifts.