28.08.2015 Views

and Cosmology

Extragalactic Astronomy and Cosmology: An Introduction

Extragalactic Astronomy and Cosmology: An Introduction

SHOW MORE
SHOW LESS
  • No tags were found...

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

7.5 Non-Linear Structure Evolution<br />

<strong>and</strong> the behavior on these small scales is affected by<br />

numerical artefacts.<br />

Computation of the Force Field. The computation of<br />

the force acting on individual particles by summation,<br />

as in (7.36), is not feasible in practice, as can be seen<br />

as follows. Assume the simulation to trace 10 8 particles,<br />

then in total 10 16 terms need to be calculated using<br />

(7.36) – for each time step. Even on the most powerful<br />

computers this is not feasible today. To h<strong>and</strong>le this<br />

problem, one evaluates the force in an approximate way.<br />

One first notes that the force experienced by the i-th particle,<br />

exerted by the j-th particle, is not very sensitive to<br />

small variations in the separation vector r i − r j , as long<br />

as these variations are much smaller than the separation<br />

itself. Except for the nearest particles, the force on<br />

the i-th particle can then be computed by introducing<br />

a grid into the cube <strong>and</strong> shifting the particles in the simulation<br />

to the closest grid point. 7 With this, a discrete<br />

mass distribution on a regular grid is obtained. The force<br />

field of this mass distribution can then be computed by<br />

means of a Fast Fourier Transform (FFT), a fast <strong>and</strong><br />

very efficient algorithm. However, the introduction of<br />

the grid establishes a lower limit to the spatial force<br />

resolution; this is often chosen such that it agrees with<br />

the softening length. Because the size of the grid cells<br />

also defines the spatial resolution of the force field, it<br />

is chosen to be roughly the mean separation between<br />

two particles, so that the number of grid points is typically<br />

of the same order as the number of particles. This is<br />

called the PM (particle–mesh) method. To achieve better<br />

spatial resolution, the interaction of closely neighboring<br />

particles is considered separately. Of course, this<br />

force component first needs to be removed from the<br />

force field as computed by FFT. This kind of calculation<br />

of the force is called the P 3 M (particle–particle<br />

particle–mesh) method.<br />

Initial Conditions <strong>and</strong> Evolution. The initial conditions<br />

for the simulation are set at very high redshift.<br />

The particles are then distributed such that the power<br />

spectrum of the resulting mass distribution resembles<br />

a Gaussian r<strong>and</strong>om field with the theoretical (linear)<br />

power spectrum P(k, z) of the cosmological model. The<br />

7 In practice, the mass of a particle is distributed to all 8 neighboring<br />

grid points, with the relative proportion of the mass depending on the<br />

distance of the particle to each of these grid points.<br />

equations of motion for the particles with the force field<br />

described above are then integrated in time. The choice<br />

of the time step is a critical issue in this integration,<br />

as can be seen from the fact that the force on particles<br />

with relatively close neighbors will change more<br />

quickly than that on rather isolated particles. Hence, the<br />

time step is either chosen such that it is short enough for<br />

the former particles – which requires substantial computation<br />

time – or the time step is varied for different<br />

particles individually, which is clearly the more efficient<br />

strategy. For different times in the evolution, the particle<br />

positions <strong>and</strong> velocities are stored; these results are<br />

then available for subsequent analysis.<br />

Examples of Simulations. The size of the simulations,<br />

measured by the number of particles considered, has<br />

increased enormously in recent years with the corresponding<br />

increase in computing capacities <strong>and</strong> the<br />

development of efficient algorithms. In modern simulations,<br />

512 3 or even more particles are traced. One<br />

example of such a simulation is presented in Fig. 7.10,<br />

where the structure evolution was computed for four different<br />

cosmological models. The parameters for these<br />

simulations <strong>and</strong> the initial conditions (i.e., the initial realization<br />

of the r<strong>and</strong>om field) were chosen such that the<br />

resulting density distributions for the current epoch (at<br />

z = 0) are as similar as possible; by this, the dependence<br />

of the redshift evolution of the density field on<br />

the cosmological parameters can be recognized clearly.<br />

Comparing simulations like these with observations has<br />

contributed substantially to our realizing that the matter<br />

density in our Universe is considerably smaller than<br />

the critical density.<br />

Massive clusters of galaxies have a very low number<br />

density, which can be seen from the fact that the massive<br />

cluster closest to us (Coma) is about 90 Mpc away.<br />

This is directly related to the exponential decrease of<br />

the abundance of dark matter halos with mass, as described<br />

by the Press–Schechter model (see Sect. 7.5.2).<br />

In simulations such as that shown in Fig. 7.10, the simulated<br />

volume is still too small to derive statistically<br />

meaningful results on such sparse mass concentrations.<br />

This difficulty has been one of the reasons for simulating<br />

considerably larger volumes. The Hubble Volume<br />

Simulations (see Fig. 7.11) use a cube with a side length<br />

of 3000h −1 Mpc, not much less than the currently visible<br />

Universe. This simulation is particularly well-suited<br />

295

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!