Simulations of Galaxy Formation and Large Scale Structure

Galaxy formation is one of the most fascinating topics of modern
cosmology. Since time immemorial, people have desired to understand
the origin, motion and evolution of planets, stars and, more
recently, galaxies and the Universe as a whole. Great advances in
astronomy always have had impact on philosophy and redefined the
self-understanding of mankind within the Universe. The first
milestone on the long road of discoveries was undoubtedly the
formulation of the laws of gravity and mechanics in 1687 by Newton.
Einstein's extension of these laws in the years 1905 and 1913 led
to a revolutionised understanding of space and time. In 1929 Hubble
established the expanding Universe which subsequently led to the
postulation of the hot Big Bang by Lemaitre (1934). Zwicky (1933)
found that most of matter in the Universe is dark. The nature of
this matter, interacting only through gravity and perhaps through
the weak interaction, is still a mystery. Finally, Penzias and
Wilson (1965) discovered the cosmic microwave background radiation,
not only confirming the theory of the Big Bang, but also - as was
observed later - revealing the origin of structure in the Universe.
Today, cosmology and especially galaxy formation are fast paced
exciting scientific fields. Surveys like the Sloan Digital Sky
Survey will soon provide a catalogue of about 500 million galaxies
with an unprecedent wealth of data. Deep observations with 8 or 10
m telescopes or with the Hubble Space telescope allow to observe
objects in their very early evolutionary stages. In addition to
this, the dramatic increase in computer power now allows us to
carry out numerical experiments on galaxies and even on the large
scale structure of the Universe. The latter is possible because of
the extraordinary fact that as a result of microwave background
observations the properties of the Universe some 300,000 years
after the Big Bang are well known. As ordinary matter makes up only
about ten percent of the total matter in the Universe, it can be
neglected in simulations in a first approximation. An initial
density field can then be evolved under the sole influence of
gravity. The result of such simulations may be combined with
semi-analytic models for the baryonic physics associated with
galaxy formation. Gravity is a long range force, and it turns out
that length scales of 100 Mpc or more have to be included in large
scale structure simulations in order to obtain results that are
representative for the Universe as a whole. The sizes of galaxies,
however, are three to four orders of magnitude smaller than this so
that numerical resolution has always been a concern in simulations
which try to include galaxy formation. A clever and powerful trick
alleviates this problem. After a low-resolution simulation has been
performed, a small region of interest is selected and the
simulation is run again, this time concentrating most of the
computational effort on the small region, allowing the resolution
to be increase dramatically without losing tidal influences from
the large cosmological volume. This technique - called resimulation
- is the driving force behind all the simulations that were
performed for this thesis. After having run about 1500
supercomputer jobs it is clear that this technique is extremely
powerful and allows the faithful simulation of objects that are far
into the regime of non-linear evolution while taking into account
the full cosmological context. In the first chapter of this work we
briefly introduce aspects of the observable Universe and discuss
the relevant theoretical background for this thesis. In the second
chapter we use high-resolution simulations of structure formation
to investigate the influence of the local environment of dark
matter haloes on their properties. We run a series of four
re-simulations of a typical, carefully selected representative
region of the Universe so that we can explicitly check for
convergence of the numerical results. In our highest resolution
simulation we are able to resolve dark matter haloes as small as
the one of the large Magellanic cloud. We propose a new method to
estimate the density in the environment of a collapsed object and
find weak correlations of the spin parameter and the concentration
parameter with the local halo density. We find no such correlation
for the halo shapes, the formation time and the last major merging
event. In a second step we produce catalogues of model galaxies
using a semi-analytic model of galaxy formation. We find
correlations between the bulge-to-disk luminosity and the B-V
colour index with the local environment. In chapter three we
compare observations of the internal structure and kinematics of
the eleven known satellites of the Milky Way with simulations of
the formation of its dark halo in a LambdaCDM universe. Earlier
work by Moore et al. 1999 and Klypin et al. 1999 claimed the
cosmological concordance model of the Universe, the LambdaCDM
model, to disagree with observations. The so-called
"substructure-problem" is one of the two major challenges for this
model and has attracted much attention. In order to remove the
discrepancy, changes of the cosmological model have been proposed.
We reinvestigate the substructure-problem using our ultra-high
resolution simulations. For a galaxy-sized dark matter halo, our
mass resolution is the highest resolution ever achieved. In
contrast to the work of Moore et al. 1999 and Klypin et al. 1999,
we find excellent agreement. The observed kinematics are exactly
those predicted for stellar populations with the observed spatial
structure orbiting within the most massive "satellite"
substructures in our simulations. Less massive substructures have
weaker potential wells than those hosting the observed satellites.
If there is a halo substructure "problem", it consists in
understanding why halo substructures have been so inefficient in
making stars. We find that suggested modifications of dark matter
properties (e.g. self-interacting or warm dark matter) may well
spoil the good agreement found for standard Cold Dark Matter. If
the dark matter in the Universe is made of weakly self-interacting
particles, they may self-annihilate and emit gamma-rays. The
detection of the gamma-ray signal would finally, after seventy
years since its discovery, shed light on the nature of the dark
matter. In chapter four we use our ultra-high resolution numerical
simulations to estimate directly the annihilation flux from the
central region of the Milky Way and from dark matter substructures
in its halo. Such estimates remain uncertain because of their
strong dependence on the structure of the densest regions of the
halo. Our numerical experiments suggest, however, that less direct
calculations have typically overestimated the emission from the
centre of the Milky Way and from its halo's substructure. We find
an overall enhancement of at most a factor of a few with respect to
a smooth halo of standard NFW structure. For an observation outside
the region around the galactic centre where the diffuse galactic
gamma-ray background is dominant, GLAST can probe a large region of
possible MSSM models. This result is independent of the exact
structure of the innermost region of the Galaxy. Our analysis shows
that the flux from the inner galaxy exceeds the expected
contribution from the brightest substructure by a large factor.
Nevertheless, for certain MSSM models substructure halos might be
detectable with GLAST.