You are on page 1of 24

Dies ist Titel der Dissertation

eventuell mit Untertitel


Aniket Agrawal

Munchen 2014

Dies ist Titel der Dissertation


eventuell mit Untertitel
Aniket Agrawal

Masterarbeit
im Studiengang Physik
der LudwigMaximiliansUniversitat
Munchen
vorgelegt von
Aniket Agrawal
Durchgefuhrt am
Max Planck Institut fur Astrophysik
Garching
2013-2014

Supervisor: Prof. Eiichiro Komatsu

Chapter 1
Introduction
The last two decades have seen tremendous improvement in our understanding of the Universe. Space missions such as the COsmic Background Explorer (COBE)[2], the Wilkinson
Microwave Anisotropy Probe (WMAP)[9] and more recently, Planck[16], have revealed extremely precise answers to questions such as how old is the Universe, how big it is, what is
its shape, what is it made of and how it began. All these missions observe light that was
emitted when the Universe was only 380,000 years old (the Cosmic Microwave Background,
or CMB), and so, bereft of much of the complexities associated with age. Therefore, they
provide us with an extremely accurate picture of the young Universe. However, the present
age of the Universe is about 14 billion years and CMB observations carry information about
the late time evolution of the Universe only through very weak effects such the Integrated
Sachs Wolfe effect and the angular diameter distance to the epoch of re-ionization. Also,
perturbations in the energy density of photons are linear[14, chapter 4] and so, CMB can
only give us information on linear (large) scales.
Large scale structure, when complemented with CMB observations, provides us another
powerful probe of the Universe on non-linear (smaller) scales since perturbations in the
matter density field can become non linear. Also, because it happens over a much longer
period in the history of the Universe and it is sensitive to gravitational forces, it has a
stronger dependence on the late-time evolution of the Universe and on the exact nature of
gravity, so providing, a way to probe dark energy and alternative gravity. Indeed, surveys
like the SDSS[1] have revealed a picture of the Universe that is consistent with the one
painted by CMB data. In order to constrain the parameters of our Universe further, and
obtain a better understanding of the physical processes at play in its evolution however,
we must be able to make more precise measurements. Large scale structure surveys suffer
from an inherent imprecision in this regard, to understand which, we must look at how
they actually operate.
Large scale structure surveys, or galaxy redshift surveys, measure the three dimensional
positions of galaxies in the Universe. The co-ordinates of a galaxy are recorded as the
two angles-declination(dec) and right ascension(RA) on the sky, and a third co-ordinate,
redshift(z), which serves as a proxy to the line of sight distance to galaxies(see below).
Using these three dimensional positions, one can construct the correlation function or

1. Introduction

power spectrum of galaxies. These functions are easily predicted from physical theories
and carry information about the cosmology of the Universe. Thus, accurate measurement
of galaxy positions gives us answers to the questions we ask about our Universe.
However, when we observe light from a galaxy we cannot be sure of the position of
the galaxy unless we have information about its energy at the source. And indeed, when
we observe light from galaxies, it is usually the case, that we do not know the intrinsic
luminosities of the objects emitting that light. In such a situation, we cannot directly
estimate the distance to the emitter and we must resort to other means to determine this
distance. Galaxy redshift surveys accomplish this task using the redshifts of galaxies. If
the observed redshift of a galaxy is z one can calculate its speed using Doppler effect as
z

v
c

(1.1)

where c is the speed of light in vacuum and v is the speed of the galaxy with respect to the
observer along the line of sight. We can complement this information with the Hubbles
law [10], ~v = H~r where ~v is the velocity of an object, ~r is the position of the object w.r.t.
the observer, and H is a time dependent factor, which is constant w.r.t. position. Then,
the radial distance of the emitter is obtained as
v
= r.
H

(1.2)

There is, however, an interesting subtlety in this approach. We have assumed that the
galaxies obey Hubbles law strictly. This is not the case in reality. Hubbles law is only
obeyed exactly in a homogeneous and isotropic Universe. In an inhomogeneous Universe
like ours, there are other galaxies and objects around which exert gravitational pull on
a given galaxy over and above the background evolution. The motion of the galaxy due
to these pulls is called peculiar motion and the velocity inferred from the Doppler shift
(1.1) is the sum of the velocity due to Hubble flow and peculiar motion. The presence of
these peculiar velocities disturbs the measurements of three dimensional galaxy locations
in surveys. Since surveys measure redshifts of galaxies and from these, infer their positions
on the sky, we say that these measurements are made in redshift space and the disturbances
produced due to peculiar motion of galaxies are called redshift space distortions. It is clear
that to obtain highly accurate correlation functions or power spectra of galaxies, we must
be able to get rid of these redshift space distortions.
Early attempts to solve the problem of redshift space distortions were made by Peebles[15].
He derived what is now known as the streaming model. Kaiser[11] made significant progress
in solving the problem of redshift space distortions in the linear limit. He calculated the
power spectrum on large scales and his work is still used today to model the linear power
spectrum. Fisher[4] argued that the streaming model fails to asymptote to the Kaiser
limit on large scales because it assumes constant velocity dispersion and independence of
relative velocity w.r.t. scale, and derived an expression for the redshift space correlation
function in terms of the real space correlation function that reproduced the Kaiser limit
on large scales and the streaming model on small scales. Then in 2004, Scoccimarro[18]

3
derived the exact relationship between the correlation functions in real space and in redshift space and showed that assuming a Gaussian probability distribution function(PDF)
for the l.o.s. velocity, even with scale dependent parameters, is not a good approximation.
It does not take into account the non-linearity of the problem fully. Reid and White[17]
made significant progress in modelling redshift space distortions, but they continue to use
a Gaussian PDF for the l.o.s. velocity distribution. We review these approaches in more
detail in chapter 2.
In this thesis, we try to model redshift space distortions using a non Gaussian PDF. We
try to take non linearities into account by using the Zeldovich approximation[20]. We take
a first-principles approach to the problem and show that a better match is obtained to the
l.o.s. velocity PDF over linear theory. Comparing the redshift space correlation function
predicted by our model to results from an N-body simulation shows that our model can
reproduce the monopole of the redshift space correlation function quite well. However,
it fails to reproduce the quadrupole of the correlation function. We also investigate the
cause of this mismatch by looking at different assumptions on the velocity PDFs at high
redshifts. Further work needs to be done to model the PDF well enough to have accurate
predictions for the quadrupole of the redshift space correlation function.

Chapter 2
Theory of Redshift Space Distortions
2.1

Perturbations in the Universe

The standard paradigm of formation of structure in modern cosmology is that all structure
arose out of gravitational instability of small inhomogeneities in the density field. As a
consequence of gravity, regions which were slightly overdense (compared to their environment) grew in density to form the structure that we see today. Complementary to this,
regions which were slightly underdense became more underdense with time and form what
are now called voids. We see that in this paradigm, the physical quantity of interest is the
overdensity defined as
(~x, t) =

(~x, t)
1
(t)

(2.1)

where (~x, t) is the density of the object under consideration at the position ~x at any
is the background density of the Universe at the same time t. From this
time t, and (t)
definition, we see that an overdensity of 0 means the region is at the same density as the
background. Overdense regions are characterised by a positive and underdense regions
by a negative . Of course, it is not possible to describe the dynamics of the Universe
using just a scalar. We must also take into consideration the velocity field of the Universe,
~u(~x, t), where ~u is the peculiar velocity of matter at the point ~x at time t, i.e. if
~r = a~x

(2.2)

~r = a~
x + a~x

(2.3)

then

where ~x is the comoving co-ordinate of a particle. This can be re-written as


~v = H~r + ~u
where ~u = a~x .

(2.4)

2. Theory of Redshift Space Distortions

Using conservation of energy and momentum, it can be shown[3] that these fields obey
the following equations
~
+ {[1 + ] ~u} = 0

(2.5)

~u
e u + ~u ~
~ u =
~ 1 j (ij ) .
+ H~

(2.6)

3
e 2
2 = m H
2

(2.7)

with

These equations are defined in terms of the conformal time instead of the cosmological
time t, with being related to t as
d =

dt
a(t)

(2.8)
da

e is defined as d . The right hand side


and we assume that = 0 at t = 0. The quantity H
a
of the second equation contains the source terms for velocity evolution in the Universe.
The first term corresponds to the gravitational potential generated due to the overdensity
and the second term corresponds to the stress tensor. Note that only the quantities m
e refer to the homogeneous Universe and all other quantities refer to the perturbed
and H
Universe.
The above three equations tell us that given the configuration of the fields (~x, i ) and
~u(~x, i ) at some initial time i , it is possible to calculate the final configuration (~x, f ) and
~u(~x, f ) at any final time f , which might in principle be chosen to be today. We could
then predict the distribution of matter in the Universe that we see today exactly.

2.2
2.2.1

Correlation functions and power spectra


Why correlation functions or power spectra?

The above approach, though promising, is not feasible. Firstly, we do not know the initial
conditions of the fields exactly at any point in time. Since the equations for evolution
of these fields are non-linear, therefore any small errors in the specification of the initial
conditions can lead to quite different predictions at the final instant. Thus, we need
extremely fine-tuned initial conditions, which we do not have. Secondly, even if we set
out to measure the field configuration exactly at some point in time, hoping to predict
the future, the large spatial extent of the Universe means that we see different objects at
different points in time. Thus, we cannot know the configuration of the fields in all of space
at any single instant of time[3]. So we must resort to statistical methods for extracting
information out of observational large scale structure data.

2.2 Correlation functions and power spectra

2.2.2

Some definitions

In this section we will define the physical quantities which will be of interest to us, in
analysing data from large scale structure surveys. We begin with the correlation function,
or the 2-point correlation function, which is the covariance of overdensities at separation
rij = |~ri ~rj |
(~ri , ~rj ) = h(~ri )(~rj )i

(2.9)

where the angular brackets signify ensemble averages over different realisations of the initial
density field. In future, we will always refer to the real space correlation function as simply
and the redshift space correlation function will be denoted with an extra subscript s,
s . In analogy with this definition, if instead of the real space positions ~ri and ~rj we
use the positions in redshift space, ~si and ~sj the correlation function so obtained is the
redshift space correlation function. The assumption of statistical homogeneity implies that
the correlation function is dependent only on the difference ~ri ~rj and not on the exact
location of ~ri and ~rj . Isotropy means that in fact, it is not even dependent on the direction
of ~ri ~rj but just its magnitude |~ri ~rj | r. So, (~ri , ~rj ) = (r).
The power spectrum is the Fourier transform of the correlation function, which we
define as
Z
~
P (k) = (r)eik~r d3 r
(2.10)
and the inverse transform is defined as
Z
(r) =

P (k)e

i~k~
r

d3 k
.
(2)3

(2.11)

Note the extra factors of 2 that appear in the definition of the inverse Fourier transform,
which are put there as a sort of convention. It can be shown[7] that
D

E
~
~
(ki )(kj ) = (2)3 D (~ki + ~kj )P (~ki )

(2.12)

with D denoting the Dirac delta function. The delta function that appears here is a
consequence of assuming statistical homogeneity and isotropy implies that P (~ki ) = P (ki ),
i.e. it only depends on the magnitude of its argument.
At this point it is important to remember that observations only reveal positions in redshift
space, as it is the redshift information of galactic light that we use to calculate its position.
Therefore, all our correlation functions or power spectra are measured in redshift space. Of
course, we would like to obtain these quantities in real space to compare with our models
and so we need to be able to go from one space to another. We will now look at how this
is achieved.

2.3

2. Theory of Redshift Space Distortions

From real space to redshift space

As we remarked in the introduction, the position of a galaxy is assumed to be H~v (1.2),


where ~v is the line of sight velocity of the galaxy, in accordance with Hubbles law. From
the above consideration, we know that ~v = H~r + uz z, where we assume the plane-parallel
approximation, according to which the z-axis is assumed to be aligned to the l.o.s. direction,
so the l.o.s. velocity is just uz . Then the calculated position ~s, where ~s is the position of
the galaxy in redshift space is
~s = ~x +

uz
z.
aH

(2.13)

Note that ~s is the comoving co-ordinate in redshift space for a galaxy at the comoving
co-ordinate ~x in real space. So, ~r = a~x. This equation represents a non-linear mapping
from real space clustering to redshift space clustering. Due to this non-linearity, even the
lowest order statistics in redshift space involve correlations between density and velocity
fields to infinite order. Let us now try to calculate the correlation function in redshift space
and see how it relates to the correlation function in real space[18].
Figure 2.1 shows the effect of peculiar velocities on clustering in redshift space. For
simplicity, we consider a spherical overdensity into which galaxies are falling[6]. When the
infall velocities of the galaxies are not too high, the first case, then the galaxy closer to the
observer appears to be moving away from him at a velocity higher than the Hubble flow,
and thus appears at a higher redshift than in a homogeneous Hubble flow. Similarly, the
farther galaxy appears to be moving at a smaller velocity and so appears a little blueshifted
w.r.t. homogeneous Hubble flow. This results in enhanced clustering in redshift space
along the l.o.s., the squashig effect. In the second case, at turnaround, the velocities are
just high enough that the two galaxies appear to be at the same redshift. The spherical
cluster appears collapsed along the l.o.s. in redshift space. Finally, when the velocities are
even higher, which is the case for random virial motion, the clustering is reduced along the
l.o.s. and enhanced perpendicular to it, the figer-of-god effect. Note also that the galaxies
which lie perpendicular to the l.o.s. are not affected by redshift space distortions. Figure
2.2 shows the effect of redshift space distortions in data obtained in a real survey, the
2dFGRS survey[8]. We clearly see the squashing and finger-of-god effects. The quantity
shown in the figure is the cosmological parameter (see later) that we are interested in
extracting with higher accuracy from redshift surveys.
We will now calculate the amount of enhancement and reduction in these two effects,
mathematically. Let us consider a population of galaxies in a small volume d3 x around
some position ~x in real space. Then, they will occupy a small volume d3 s around the point
~s, given by equation (2.13). Let us also denote the overdensity as in real space and as s
in redshift space. Then, since the number of galaxies within these corresponding volumes
must be the same (and the mean number density of galaxies = is same in both real and
redshift space )
[1 + (~x)]d3 x = [1 + s (~s)]d3 s.

(2.14)

2.3 From real space to redshift space

Figure 2.1: Redshift space distortions induced by peculiar velocities[6]


which gives us
[1 + (~x)]d3 x = [1 + s (~s)]d3 s.

(2.15)

This is the exact relationship between the overdensities in real and redshift space. We can
further write
1 + s (~s) = [1 + (~x)]

d3 x
.
d3 s

(2.16)

Equation (2.13) gives us the relation between co-ordinates in real and redshift space. The
fraction appearing on the right hand side of equation (2.16) is just the inverse of the

10

2. Theory of Redshift Space Distortions

Figure 2.2: Redshift space distortion in observations[8]


Jacobian of this transformation, i.e.
1 + s (~s) =

[1 + (~x)]
J


s
with J = xji . In the plane-parallel approximation, equation (2.13) gives



u
z
z
.
J(~x) = 1 +
aH

(2.17)

(2.18)

We can now use this relationship to study different regimes, which will lead us to the
Kaiser formula and the streaming model.

2.3.1

Kaiser limit

The Kaiser limit[11] of equation (2.16) is obtained when the density and velocity fields and
their first derivatives are relatively small in magnitude. In that case, we can write
z uz
1
'1
J
aH
and we neglect higher powers of z uz . Using this we get
s (~s) = (~x)

z uz
aH

(2.19)

(2.20)

2.3 From real space to redshift space

11

where we have also neglected terms containing products of the density and velocity fields.
The density and velocity fields are coupled to each other through the equations (2.5),
(2.6) and (2.7). In the linear limit they can be reduced to[3]

+ =0

(2.21)

and

e + 3 m H
e 2 = 0
+ H

2
w
~
ew
+H
~ =0

(2.22)
(2.23)

~ ~u and w
~ ~u. The last of these equations tells us that the vorticity of the
with =
~ =
velocity field decays with time, which is expected for an expanding Universe. Consequently,
on large scales, where linear theory applies, the velocity field is irrotational. As such, only
the divergence of the velocity field is sufficient for its description. These equations can be
easily solved analytically to give the solution
(~x, ) =

D( )
(~x, i )
D(i )

(2.24)

where D( ) is the linear growth factor at a time .


Equation (2.21) is easily solved in Fourier space once we introduce the Fourier transformed variables
Z
d3 k e i~k~x
s =
~e
(2.25)
(2)3 s,k
Z
d3 k e i~k~x
=
~ e
(2.26)
(2)3 k
Z
d3 k e i~k~x
~u~ e .
(2.27)
~u =
(2)3 k
e~ ~k and then use
From the fact that the velocity field is irrotational, we can derive ~u
k
the equation (2.21) along with equation (2.24) to get
~
e~ = iaHf k e~ .
~u
k
k2 k

(2.28)

We can also write equation (2.20) in Fourier space as


es,~k = e~k i

kz u
ez,~k
aH

(2.29)

12

2. Theory of Redshift Space Distortions

or


kz2
e
e
s,~k = ~k 1 + f 2 .
k
Then, from the definition of the power spectrum, we get

2
kz2
Ps (k, kz ) = P (k) 1 + f 2
k

(2.30)

(2.31)

which is the Kaiser limit of the non-linear power spectrum. This is the form of the power
spectrum that should be observed on large scales, when density and velocity fields are
small. We can see that even though the real space power spectrum is isotropic (for a
homogeneous and isotropic universe), the redshift space power spectrum is anisotropic. If
we can model these anisotropies, we can extract important information about the quantity
f which is dependent on cosmology and also the nature of gravity.

2.3.2

General relation between real and redshift space correlation functions

Instead of evaluating equation (2.15) in the linear case, we can calculate the more general
case of arbitrary density and velocity fields, as was done by Scoccimarro in[18]. We start
with equation (2.15) and Fourier transform both sides w.r.t. the variable ~s. After using
equation (2.13), we will get
Z
d3 k i~k~x ikz vz
~
~
e
e
[1 + (~x)]
(2.32)
D (k) + s (k) =
(2)3
uz
where we have used vz = aH
.
Using this relation, and after a bit of algebraic manipulation (refer to [18] for more details),
we find the following relation for the correlation functions in real space and in redshift space
Z
1 + s (sk , s ) =
dy [1 + (r)] P (H(sk y), ~r)
(2.33)

where y is the real space separation between a pair of galaxies or objects along the l.o.s., sk
is the l.o.s. separation between a pair of objects in redshift space, and s is the separation
between a pair of objects perpendicular to the l.o.s. which is the same in real and redshift
space (see Figure 2.3).The function P (H(sk y), ~r) is the probability distribution function
of pairs of objects having relative velocity H(sk y) along the l.o.s. and with real space
separations y along the l.o.s. and s perpendicular to the line of sight. Obviously, r2 =
y 2 + s2 .
The physical interpretation of this formula is clear: P maps the pairs at separation
y to separation sk due to relative velocity H(sk y) with probability P (H(sk y), ~r)[18].
The conventional streaming model[15, section 76] is the equation (2.33) with the (1 + )s
replaced by s. Equation (2.33) tells us that in order to calculate the redshift space
correlation function, from the real space one, it is necessary for us to know the l.o.s.
relative velocity distribution of galaxies. We discuss an approach to this later in the thesis.

2.4 Models for Redshift Space Correlation Function

13

Figure 2.3: A schematic of the configuration of pair of objects in real(black) and redshift(red) space. The black arrows denote the peculiar velocities of the objects.

2.4
2.4.1

Models for Redshift Space Correlation Function


The Streaming Model

If we consider the distribution of relative velocities for objects on small scales, then the
distribution of velocities is independent of the separation, i.e. P (H(sk y), ~r) P (H(sk
y)). This is because on these highly non-linear scales, there is essentially no correlation
between the density and velocity fields, and velocity might be considered a totally random
field. In that case, equation (2.33) becomes
Z
Z
1 + s (sk , s ) =
dyP (H(sk y)) +
dy(r)P (H(sk y)).
(2.34)

By a simple change of variables, we may change the integral in the first term on the right
hand side to an integral over the l.o.s. velocity of the probability distribution function of
l.o.s. velocities, which is just 1. This 1 and the 1 on the left hand side cancel, to give us
the conventional streaming model[15]
Z
s (sk , s ) = (r)P (H(sk y))dy.
(2.35)
The streaming model is clearly wrong if we try to extrapolate it to large scales. We
know from our consideration of linear perturbation theory (2.21),(2.22) that the velocity
and density fields are in fact strongly correlated on large scales, which is expected as density
field sources the gravitational field which sources the peculiar velocity field. Indeed, because
the dispersion in velocities depends on scale even down to scales of about 1 Mpc/h, we

14

2. Theory of Redshift Space Distortions

expect the phenomenological parameter p also to carry some scale dependence. However,
it is not clear how to model this dependence given the highly non-linear nature of the
problem. In any case, it is not surprising that the streaming model does not asymptote to
the Kaiser limit. This problem was resolved by Fisher[4].

2.4.2

The Gaussian distribution function

The work of Fisher [4] and Reid and White [17] can be summarized by just saying that
both assume a Gaussian PDF for the l.o.s. relative velocity. Both assume that the density
and velocity fields are Gaussian random fields and so, to specify the PDF, all we need to
do is specify the mean and dispersion of the PDF. Fisher [4] calculates the redshift space
correlation function only on very large scales, and shows that it reproduces the Kaiser
limit. Reid et. al. [17] use the Gaussian PDF to calculate the redshift space correlation
function on all scales and compare it to results from simulations. How one can calculate
the mean and dispersion of velocities in linear theory?
Let us suppose that we want to calculate the mean relative velocity for objects at
locations ~x and ~x + ~r. Let us also denote the velocity of object at ~x = ~u and velocity of
object at ~x + ~r = ~u0 . Then, the mean velocity is given as h(~u ~u0 )(1 + )(1 + 0 )i. Here,
(~x) and 0 (~x + ~r). These factors of (1 + ) account for the weighted average of
velocities, and the (1 + )s are proportional to the number of objects at ~x and ~x + ~r.
h(~u ~u0 )(1 + )(1 + 0 )i = h~u 0 i h~u0 i + O()

(2.36)

where we have neglected higher order terms in the linear limit, h~u ~u0 i = 0 and h~ui =
h~u0 0 i = 0 for a Gaussian field[4]. From linear theory, we also know that
Z
D
dkkj1 (kr)P (k)
h~u i =
r 2
2 HD
0

(2.37)

which is directed along the line of separation of the two pairs, by symmetry. P (k) denotes
the linear matter power spectrum, which is the variance of the density field, extrapolated
to z = 0, using the linear growth factor. Using this we get
f
h(~u ~u )(1 + )(1 + )i = 2
r 2
2
= u12 (r)
r
0

Z
dkkj1 (kr)P (k)

(2.38)
(2.39)

where we have denoted the mean relative velocity along the line of separation as u12 which

is a function of separation and f d(lnD)/d(lna) = D/HD.


Note that the mean velocity
perpendicular to the line of separation is zero. The variances of the velocity fields in linear

2.4 Models for Redshift Space Correlation Function

15

theory are given by[5]


Z
1 H 2f 2
=
dkP (k)
3 2 2
Z
H 2f 2
j1 (kr)
=
dkP (k)
2
2
kr


2 2 Z
H f
j1 (kr)
k =
dkP (k) j0 (kr) 2
2 2
kr
v2

(2.40)
(2.41)
(2.42)

D
E


2
with uk u0k = v2 k k2 and u,1 u0,1 = u,2 u0,2 = v2
, 1, 2 denoting two
directions into which the velocity perpendicular to the line of separation can be resolved.
The quantity v represents the one-point velocity dispersion, i.e. huui.
Now that all the ingredients are assembled, we can construct the full 3 dimensional
velocity PDF, assuming that the 3 components of velocity-parallel and perpendicular to
the line of separation are independent of each other, as they should be. We get a product
of three univariate Gaussians,

(uk u12 (r))2


2 (r)
k

u2
,1
2 (r)

u2
,2
2 (r)

e
e
e
p
p
p(uk , u,1 , u,2 , r) = q
2
2
2 (r) 2
(r)
2k2 (r)

(2.43)

where we explicitly show the dependence of the parameters of the Gaussian on the separation r. We show the scale dependence of the mean velocity and radial dispersion in Figure
2.5.
Note that (2.43) is not the PDF of just the l.o.s. velocities which is what we eventually
seek to calculate redshift space correlation function. However, the PDF of l.o.s. velocities
is obtained from this PDF by a simple calculation.
The line-of-sight velocity PDF
Suppose that we have a pair of objects in a configuration as shown in Figure , i.e. the
radial vector from one object to the other makes an angle with the l.o.s. such that
cos() = . We also resolve the perpendicular velocity into two components, such that one
component lies in the plane containing the line of sight and the line of separation between
the two objects. In the plane parallel approximation, where the l.o.s. is chosen to be the
z direction, this component might just be chosen to be the x or y component.
Let us name the l.o.s. velocity ul and the perpendicular velocity component (in the
plane made by l.o.s. direction unit vector and line of separation unit vector) u,1 . Then,
by simple vector algebra (see Figure 2.4), it can be shown that
ul = uk u,1

p
1 2 .

(2.44)

16

2. Theory of Redshift Space Distortions

Figure 2.4: picture showing line of sight, definition of and perpendicular velocity component
Hence, given the full 3-d PDF p(uk , u,1 , u,2 , r), the l.o.s. velocity PDF is simply the
sum of all pairs, having a 3-d velocity such that the above condition is satisfied, i.e.
Z
p
p(ul , r, ) = p(uk , u,1 , u,2 , r)D (ul uk u,1 1 2 )duk du,1 du,2
(2.45)
with D being the Dirac delta function that imposes the condition (2.44). For the special
case of the Gaussian PDF (2.43), the above integral yields
2

(u u12 )
l
1
2 2
l
p(ul , r, ) = p
e
2
2l

(2.46)

2
where l2 is the dispersion in l.o.s. velocities and l2 = 2[2 k2 + (1 2 )
].
Fisher[4] expands the above PDF further, keeping only terms up to quadratic order in
(ul u12 ), which is a good approximation in the linear limit. Reid and White[17] keep
the full l.o.s. PDF as above, using appropriate mean and dispersions for linearly biased
tracers (this is done by just replacing f in equations (2.40) by f /b where b is the linear
bias factor). Reid and White also compare the results of calculating correlation functions
using the Gaussian l.o.s. velocity PDF to N-body simulations. They find that they are able
to reproduce the real space correlation function and the monopole of the redshift space
correlation function quite well. However, they fail to reproduce the quadrupole of the
redshift space correlation function, which is expected as the monopole is not so sensitive
to the velocity PDF, while the quadrupole is, and results from N-body simulations show
that the velocity PDF is not Gaussian[18]. In fact, it is shown in [18] that the PDF has
skewness and kurtosis even up to scales of the order of 100 Mpc/h. We reproduce their
figure here.

2.4 Models for Redshift Space Correlation Function

17

Figure 2.5: Moments of pairwise velocities parallel to the line connecting the pair as a
function of scale. Top panel: pairwise dispersion 12 (squares) as a function of scale,
c
its connected piece 12
(solid line), and the mean infall v12 (triangles). The dashed line
denotes the predicted 12 in linear dynamics. Middle panel: dimensionless measure of infall
c
(|v12 |/12
, triangles) compared to the skewness s3 of the pairwise velocity PDF (squares);
the skewness dominates at most scales. Bottom panel: kurtosis as a function scale, note
that it does not vanish at large scales and s4 > 1 at all scales[18]

Bibliography
Aubourg, S. Bailey, F. Beutler, V. Bhardwaj, M. Blanton,
[1] L. Anderson, E.
A.S. Bolton, J. Brinkmann, J.R. Brownstein, A. Burden, C.H. Chuang, A.J. Cuesta,
K.S. Dawson, D.J. Eisenstein, S. Escoffier, J.E. Gunn, H. Guo, S. Ho, K. Honscheid, C. Howlett, D. Kirkby, R.H. Lupton, M. Manera, C. Maraston, C.K. McBride,
O. Mena, F. Montesano, R.C. Nichol, S.E. Nuza, M.D. Olmstead, N. Padmanabhan, N. Palanque-Delabrouille, J. Parejko, W.J. Percival, P. Petitjean, F. Prada,
A.M. Price-Whelan, B. Reid, N.A. Roe, A.J. Ross, N.P. Ross, C.G. Sabiu, S. Saito,
L. Samushia, A.G. Sanchez, D.J. Schlegel, D.P. Schneider, C.G. Scoccola, H.J. Seo,
R.A. Skibba, M.A. Strauss, M.E.C. Swanson, D. Thomas, J.L. Tinker, R. Tojeiro,
M.V. Maga
na, L. Verde, D.A. Wake, B.A. Weaver, D.H. Weinberg, M. White, X. Xu,
C. Y`eche, I. Zehavi und G.B. Zhao, Monthly Notices of the Royal Astronomical Society
441 (2014), 24.
[2] C.L. Bennett, A.J. Banday, K.M. Gorski, G. Hinshaw, P. Jackson, P. Keegstra,
A. Kogut, G.F. Smoot, D.T. Wilkinson und E.L. Wright, The Astrophysical Journal Letters 464 (1996), L1.
[3] F. Bernardeau, S. Colombi, E. Gazta
naga und R. Scoccimarro, Physics Reports 367
(2002), 1.
[4] K.B. Fisher, The Astrophysical Journal 448 (1995), 494.
[5] K. Gorski, The Astrophysical Journal Letters 332 (1988), L7.
[6] A.J.S. Hamilton: Linear Redshift Distortions: a Review. Linear Redshift Distortions:
a Review, In The Evolving Universe, herausgegeben von D. Hamilton, Band 231 von
Astrophysics and Space Science Library. (1998) Seite 185.
[7] A.J.S. Hamilton, ArXiv Astrophysics e-prints (2005).
[8] E. Hawkins, S. Maddox, S. Cole, O. Lahav, D.S. Madgwick, P. Norberg, J.A. Peacock, I.K. Baldry, C.M. Baugh, J. Bland-Hawthorn, T. Bridges, R. Cannon, M. Colless, C. Collins, W. Couch, G. Dalton, R. De Propris, S.P. Driver, G. Efstathiou,
R.S. Ellis, C.S. Frenk, K. Glazebrook, C. Jackson, B. Jones, I. Lewis, S. Lumsden,
W. Percival, B.A. Peterson, W. Sutherland und K. Taylor, Monthly Notices of the
Royal Astronomical Society 346 (2003), 78.

20

BIBLIOGRAPHY

[9] G. Hinshaw, D. Larson, E. Komatsu, D.N. Spergel, C.L. Bennett, J. Dunkley,


M.R. Nolta, M. Halpern, R.S. Hill, N. Odegard, L. Page, K.M. Smith, J.L. Weiland, B. Gold, N. Jarosik, A. Kogut, M. Limon, S.S. Meyer, G.S. Tucker, E. Wollack
und E.L. Wright, The Astrophysical Journal Supplement 208 (2013), 19.
[10] E. Hubble, Proceedings of the National Academy of Sciences 15 (1929), 168.
[11] N. Kaiser, Monthly Notices of the Royal Astronomical Society 227 (1987), 1.
[12] T.Y. Lam, T. Nishimichi und N. Yoshida, Monthly Notices of the Royal Astronomical
Society 414 (2011), 289.
[13] T.Y. Lam, F. Schmidt, T. Nishimichi und M. Takada, Phys. Rev. D 88 (2013), 023012.
[14] T. Padmanabhan: Structure Formation in the Universe. Cambridge University Press,
Cambridge, USA, 1995.
[15] P. Peebles: The Large Scale Structure of the Universe. Princeton University Press,
Princeton, New Jersey, USA, 1980.
[16] Planck Collaboration, P.A.R. Ade, N. Aghanim, M. Arnaud, M. Ashdown, J. Aumont, C. Baccigalupi, M. Baker, A. Balbi, A.J. Banday und et al., Astronomy and
Astrophysics 536 (2011), A1.
[17] B.A. Reid und M. White, Monthly Notices of the Royal Astronomical Society 417
(2011), 1913.
[18] R. Scoccimarro, Phys. Rev. D 70 (2004), 083007.
[19] N. Seto und J. Yokoyama, The Astrophysical Journal 492 (1998), 421.
[20] Y.B. Zeldovich, Astronomy and Astrophysics 5 (1970), 84.

You might also like