Você está na página 1de 8

Impact of Photometric Redshift Modeling for Cosmology with

Weak Lensing
Abhass Kumar
Supervisor: Prof. Subhabrata Majumdar
Abstract
Weak lensing studies started in the later part of the 20
th
century and since then it has contin-
ued to provide improved accuracy on the various cosmological parameters slowly but steadily. The
information that it provides come from statistical studies of the shear produced by it. A power
spectrum for shear is obtained from that data and it can be analyzed to give information about the
structure formations. Here the shear power spectrum is calculated taking in mind the systematics of
the photometric redshift of the source galaxies.
1 Introduction
Weak lensing, as the name suggests is the very weak bending of light when it passes along a region of
diused matter. Concentrated masses produce much greater bending and it is called strong gravitational
lensing. Weak lensing can only be studied across a statistically large sample. The eects of weak lensing
can be used to probe the matter distribution between us and an ensemble of distant sources (so that the
randomness in individual shears cancels out).
Lensing in general depends only on the mass distribution and not on the physical nature of it. Hence
weak lensing provides a very good method of investigating dark matter. In the recent years, weak lensing
observations have been giving increasingly better results in terms of constraints on various cosmological
parameters like the matter density
m
, the Hubble parameter H
0
, the dark energy density

etc. More
details on weak lensing can be found in [5] [6].
To analyze weak lensing, we need to know the distances involved, the density of the matter intervening
the source and the observation point and the distribution of the source of light. The distances used
here are dened in terms of the redshift z = 1

o

e
(
o
is the observed wavelength,
e
is the emitted
wavelength) of the object. This is possible because in an expanding universe, light coming from all the
objects is shifted towards the red end of the spectrum and more so for distant objects than nearer ones.
As light has a nite speed, the more distant object that we look at, the further back in time we are
looking. The relation between redshift and comoving time is given by the dierential equation
dt
dz
=
1
H(z)
(1)
1
H(z) is the Hubble parameter at a redshift z. In terms of the Hubble parameter at the earth H
0
i.e.
redshift of 0, H(z) = H
0
E(z) where E(z) is given by
E(z) =
_

m
(1 + z)
3
+ (1
m
)(1 + z)
3(1+w
0
+w
a
)
(2)
for a at universe. In the equation above,
m
is the matter density and

= 1
m
is the dark energy
density. w
0
and w
a
dene the dark energy equation of state parameters w(a) = w
0
+ w
a
z
1+z
.
2 Distances in cosmology
We can now dene the distance in terms of these parameters. The comoving distance [2] is dened in
such a way that it factors out the expansion of the universe s.t. the distances remain constant over time.
It is equal to the proper distance in the present time. Mathematically, it is
(z) =
c
H
0
_
z
0
dz

E(z

)
(3)
We can also dene the angular diameter distance D
A
which is the ratio of the transverse physical
size of an object to its angular size. It is used to get actual distances from angular ones measured in
telescopes.
D
A
=

1 + z
=
c
H
0
(1 + z)
_
z
0
dz

E(z

)
(4)
Figure 1: Distances in cosmology. The y-axis unit is in Mpc
3 Matter Power Spectrum
The other thing required for weak lensing analysis is the density of the matter intervening the source
and the observer. It is better to work with density perturbations (x) =
(x)

instead of the density
itself because the mean of the perturbations is zero and we need to consider only the variance. In fact
2
the power spectrum of weak lensing involves the variance and not the mean of the densities. Also the
matter in the universe is gravitationally bound so that regions of over-density tend to grow while those
of under-density shrink in other words, density perturbations give rise to large scale structure in the
universe. The analysis of perturbations is done using independent plane waves. We dene (

k) as the
Fourier transform of (x). The variance is dened as [4]

2
=

k
(

k)

k) =

k
(

k)(

k) (5)
The last term follows because the perturbations are real. The perturbations do not depend on the direction
of

k because of the assumption of homogeneity and isotropy. For a large number of k (=|

k|)values, the
summation goes over to an integral.

2
=
1
(2)
3
_

0

2
k
d

k =
1
2
2
_

0
P(k) k
2
dk (6)
Here P(k) =
2
k
is the matter power spectrum. It is usual to assume a powerlaw form for it (P(k) =
Ak
n
, n is spectral index). However, such a form is not convergent. So we need to lter the perturbation.
We do so by a window function W(k, R). Also, the information about the uctuations does not arrive unal-
tered which forces the need for a transfer function T(k) =
ln(1+2.34q)
2.34q
_
1 + 3.89q + (1.61q)
2
+ (5.46q)
3
+ (6.71q)
4
_
1/4
where q =
k
h
with

= (
m
h) exp
_

b


b

m
_
to be multiplied to the uctuation . Thus the matter
power spectrum is
P(k) = Ak
n
T(q)
2
(7)
Figure 2: The matter power spectrum P(k)
This denition of the power spectrum is for a redshift of zero. To get the power spectrum at any given
z, we multiply it by the growth function D(z) [7] shown in Fig. 3. so that P(k, z) = Ak
n
T(q)
2
D(z).
The amplitude A is found by normalizing the variance to the value of variance at a volume with radius
R = 8h
1
Mpc denoted by
8
.

The Hubble parameter H


0
is usually taken as h 100 Km s
1
Mpc
1
. Hence if H
0
= 70 Km s
1
Mpc
1
then h = 0.7
3
Figure 3: The growth function D(z) =
(z)
(0)
4 The source galaxy distribution
Next, we look at the source galaxy distribution, which are lensed. The shear produced by weak lensing
is very small and hence necessitates statistical treatment. We need to observe many galaxies together
which is almost impossible a task using spectroscopy. Hence we use photometry. However, it introduces a
scatter in the true redshifts of the galaxies. There is a probability distribution of measuring a photometric
redshift given the true redshift p(z
ph
|z) [3] which has to be multiplied to the actual redshift distribution
of galaxies n(z) and integrated over the range of the photometric redshift bins to give the distribution
in any given bin. The binning done here is a ve bin tomography with the rst bin being z
ph
= (0, 0.6).
The second bin is (0.6, 1.2), the third is (1.2, 1.8), the fourth is (1.8, 2.4) and the fth is (2.4, 3.0).
n
i
(z) =
_
z
i+1
ph
z
i
ph
dz
ph
p(z
ph
|z) n(z) (8)
5 Weak Lensing analysis
With the distances, matter densities and the source distribution at hand, we can dene the weak lensing
shear and its power spectrum.
The weak lensing shear is the weighted integral of the density perturbation eld (equal to the convergence
eld in this limit). The average convergence from sources in the i

th bin in the direction



is [6] [1]

i
(

) =
1
H
0
_

0
dz
E(z)
W
i
(z)(z, z

) (9)
The weight function W
i
(z) is given by the relation [1]
W
i
(z) =
3
2

m
H
2
0
c
2
(1 + z)(z)
_

z
dz

n
i
(z

)
(z, z

)
(z

)
(10)
4
(z) is the comoving distance of the lens, (z

) is the distance of the source and (z, z

) is the distance
between the source and the lens. The upper limit of the integral is cut o by the source distance. For
numerical computations, the upper limit can safely be put to 5 as after that n
i
(z) becomes zero.
The shear power spectrum C

is the spherical harmonic transform [6] of the convergence eld in the


i

th and j

th bins and is given by (under the Limber approximation) [1]


C
ij

=
c
H
0
_

0
dz
1
(z)
2
E(z)
W
i
(z)W
j
(z)P
k
_
k =
l
(z)
, z
_
(11)
To calculate C

, we need explicit forms for the galaxy distribution. Observations suggest an exponential
distribution of galaxies in true redshift which is given as follows:
n(z) =
z
2
2z
3
0
exp
_

z
z
0
_
(12)
Two dierent probability distributions one Gaussian and one Gaussian with perturbative additions will
be used for the photometric redshift distribution of galaxies. The Gaussian distribution is
p
G
(z
ph
|z) =
1

2
z
exp
_

(z z
ph
z
bias
)
2
2
2
z
_
(13)
Correspondingly the distribution of galaxies in the i

th bin is (putting z
bias
equal to zero)
n
i
(z) =
1
2
n(z)
_
erf
_
z
i+1
ph
z

2
z
_
erf
_
z
i
ph
z

2
z
__
(14)
Figure 4: The source galaxy distribution assuming Gaussian distribution for photometric redshifts
For a very large sample, the central limit theorem tells us that the distribution will be Gaussian but for
any sub-sample, there will be some distortions in the distribution which make it non-Gaussian. What we
observe is just a sub-sample and this makes skewness () and kurtosis () necessary for better observation
of photometric and spectroscopic redshift of a sub-sample of galaxies.
5
For the non-Gaussian probability distribution, Edgeworth expansion was used. If the assumed actual
distribution has a skewness and kurtosis which are small, then it can be approximated as follows:
p
NG
(z
ph
|z)

p
G
(z
ph
|z)

6
d
3
p
G
dz
3
ph
+

24
d
4
p
G
dz
4
ph
+

2
72
d
6
p
G
dz
6
ph
(15)
with p
G
given by eq. 13. The comparison of the Gaussian and the non-Gaussian probability distribution
is given in Fig. 4 for = = 0.1.
Figure 5: Comparison of the Gaussian and Non-Gaussian photometric redshift distribution
With this photometric redshift distribution, the galaxy distribution in the i

th bin changes to
n
i
z
=
1
2
n(z)
_
erf
_
z
i+1
ph
z

2
z
_
erf
_
z
i
ph
z

2
z
__

n(z)
6
exp(
(zz
i+1
ph
)
2
2
2
Z
)(z z
i+1
ph
+
z
)(z + z
i+1
ph
+
z
) exp(
(zz
i
ph
)
2
2
2
z
)(z z
i
ph
+
z
)(z + z
i
ph
+
z
)

2
5
z
+
n(z)
24
e

(zz
i+1
ph
)
2
2
2
z
_
3
2
z
+ (z z
i+1
ph
)
2
_
(z z
i+1
ph
) e

(zz
i
ph
)
2
2
2
z
_
3
2
z
+ (z z
i
ph
)
2
_
(z z
i
ph
)

2
7
z
+

2
n(z)
72
e

(zz
i+1
ph
)
2
2
2
z
_
15(z z
i+1
ph
) +
(zz
i+1
ph
)
5

4
z
+
10(z+z
i+1
ph
)
3

2
z
_

2
7
z

2
n(z)
72
e

(zz
i
ph
)
2
2
2
z
_
15(z z
i
ph
) +
(zz
i
ph
)
5

4
z
+
10(z+z
i
ph
)
3

2
z
_

2
7
z
(16)
The parameter values used in my project work are given in Table 1.
6
Table 1: The various parameter values used in this work
Parameter
m

b
h
z
Median redshift z
med
for z
0

8
baryon matter density galaxy distribution
Values 0.266 0.0443 0.7 0.03(1 + z) 1 z
med
/2.674 0.8
Figure 6: Comparison of C
l
for dierent photometric redshift distributions
6 Conclusion and Further work
It was seen that the curves for C

do not dier much for smaller values of and as should be the case
but as their values are increased, there is a marked dierence in their values. The dierence is more so
with an increase in while it is comparatively much less for just an increase in as can be seen from
Fig. 6. It can be seen that the curve with a Gaussian distribution for photo redshift is at the bottom
followed by the curve with both = = 0.005. At the top is the curve with = 0.01 and = 0.005
followed by the one with both = = 0.01. In the middle lies the one with = 0.005 and = 0.01.
We can use the C

thus obtained for Fisher matrix analysis and nd constraints on the cosmology
parameters like the matter density
m
, the dark energy parameters w
0
and w
a
. Since a non-Gaussian
probability distribution was assumed for the photometric redshifts, the skewness and kurtosis form extra
parameters in the Fisher analysis. Their values can be guessed at from the information that the Fisher
matrix will provide thus giving us a better estimate of the actual probability. The constraints on the
various parameters may come out to be the same as by previous results or may be dierent. They may
become even better or even worse. In any case, that will something interesting to work and nd out.
Better results will be always welcome but worse results will lead to a better understanding.
7
References
[1] Sudeep Das; Roland de Putter; Eric V. Linder; Reiko Nakajima. Weak lensing science, surveys and
systematics. arXiv:1102.5090v1, 2011.
[2] David Hogg. Distance measures in cosmology. arXiv:astro-ph/9905116v4, 2004.
[3] Ma; Hu; Huterer. Eects of photometric redshift uncertainties on weak-lensing tomography. The
Astrophysical Journal, 2006.
[4] Coles; Lucchin. Cosmology: The Origin and Evolution of Cosmic Structure.
[5] Bartelmann; Schneider. Weak lensing. Phys. Rep., 2001.
[6] Dipak Munshi; Patrick Valegeas; Ludovic van Waerbeke; Alan Heavens. Cosmology with weak lensing
surveys. Phys. Rep., 2008.
[7] Battye; Weller. Constraining cosmological parameters using sunyaev-zeldovich cluster surveys. Phys-
ical Review D, 2003.
Acknowledgment
I worked under Prof. Subhabrata Majumdar as a VSRP and I sincerely thank him for all the help and
time that he gave to me. He was always ready to answer all my questions.
Also I would like to thank Satej Khedekar. I had many discussions with him and also had some help
with the coding.
I would also like to thank the Department of Theoretical Physics which is so friendly and helpful.
8

Você também pode gostar