Suppressing The Contribution of Intrinsic Galaxy Alignments To The Shear Two-Point Correlation Function
Suppressing The Contribution of Intrinsic Galaxy Alignments To The Shear Two-Point Correlation Function
Suppressing The Contribution of Intrinsic Galaxy Alignments To The Shear Two-Point Correlation Function
r
X
i
v
:
a
s
t
r
o
-
p
h
/
0
2
0
8
2
5
6
v
2
2
3
S
e
p
2
0
0
2
Astronomy & Astrophysics manuscript no.
(will be inserted by hand later)
Suppressing the contribution of intrinsic galaxy
alignments to the shear two-point correlation
function
Lindsay King & Peter Schneider
Institut f ur Astrophysik und Extraterrestrische Forschung, Universitat Bonn, Auf
dem H ugel 71, D-53121 Bonn, Germany
Abstract. Cosmological weak lensing gives rise to correlations in the ellipticities of
faint galaxies. This cosmic shear signal depends upon the matter power spectrum,
thus providing a means to constrain cosmological parameters. It has recently been
proposed that intrinsic alignments arising at the epoch of galaxy formation can
also contribute signicantly to the observed correlations, the amplitude increasing
with decreasing survey depth. Here we consider the two-point shear correlation
function, and demonstrate that photometric redshift information can be used to
suppress the intrinsic signal; at the same time Poisson noise is increased, due
to a decrease in the eective number of galaxy pairs. The choice to apply such
a redshift-depending weighting will depend on the characteristics of the survey
in question. In surveys with z 1, although the lensing signal dominates, the
measurement error bars may soon become smaller than the intrinsic alignment
signal; hence, in order not to be dominated by systematics, redshift information
in cosmic shear statistics will become a necessity. We discuss various aspects of
this.
Key words. cosmology dark matter gravitational lensing
1. Introduction
The distortion of distant galaxies by the tidal gravitational eld of intervening matter
inhomogeneities has become known as cosmic shear. Since this lensing signal depends
upon the matter power spectrum, it is an important cosmological tool, as was proposed
in the early 1990s by Blandford et al. (1991), Miralda-Escude (1991) and Kaiser (1992).
This set the scene for further analytic and numerical work (eg. Kaiser 1998; Schneider et
2 King & Schneider: Suppression of intrinsic alignments in cosmic shear surveys
al. 1998; White & Hu 2000), taking into account the non-linear evolution of the power
spectrum which results in increased power on small scales (Hamilton et al. 1991; Peacock
& Dodds 1996).
During 2000, four teams announced the rst observational detections of cosmic shear
(Bacon et al. 2000; Kaiser et al. 2000; van Waerbeke et al. 2000; Wittman et al. 2000;
Maoli et al. 2001), demonstrating the feasibility of its study. Various statistics are used to
quantify cosmic shear, and to compare the observations with predictions for cosmological
models. Here we focus on the shear correlation functions; other measures include the
aperture mass statistic (Schneider et al. 1998) and shear variance (e.g. Kaiser 1992).
Besides its dependence on the large-scale structure in the Universe, the magnitude of
the eect is also sensitive to the redshift distribution of the galaxies used in the analy-
sis. So far, direct redshift estimates have not been obtained for the samples concerned.
Typically the mean redshift has been estimated from surveys of similar depth, and a cor-
responding redshift probability distribution has been assumed (e.g. van Waerbeke et al.
2001; Hoekstra et al. 2002a). Motivated by the recent interest in obtaining photometric
redshifts for cosmic shear surveys, we consider how to make use of redshift information,
and its impact on the constraints permitted by the two-point statistic under considera-
tion.
In weak lensing analyses, it is assumed that the background galaxies are randomly
oriented so that their mean intrinsic ellipticity
s
= 0 and any correlation in observed
ellipticities arises from gravitational lensing. The magnitude of any intrinsic alignment
of the background source population has recently been the subject of many numeri-
cal, analytic and observational studies (e.g. Croft & Metzler 2000; Heavens et al. 2000;
Crittenden et al. 2001 (Cr01); Catelan et al. 2001; Mackey et al. 2002; Brown et al. 2002,
Hui & Zhang 2002). Due to dierences in the assumed origin of intrinsic alignments and
the type of galaxies considered, the numerical and analytic estimates span a couple of
orders of magnitude in amplitude. For example, the analytic model of Cr01 assumes that
any intrinsic signal arises from correlations between the angular momenta of galaxies,
whereas that of Catelan et al. (2001) relies on tidal shear correlations. Nonetheless, all
studies conclude that intrinsic alignments are more important for shallower surveys (be-
coming comparable to the lensing signal for a mean survey redshift z 0.5) and that
the correlation falls o quite rapidly with source separation.
The structure of this paper is as follows. In the next section we outline the rela-
tionships between the matter power spectrum and the lensing correlation functions. The
dependence on the redshift distribution and on cosmology is highlighted. Correlation
functions are derived from observational data, and in Sect. 3 we describe how practical
estimators are used for this purpose. In Sect. 4 a toy model to account for intrinsic align-
ments is developed; the amplitude of the correlation is normalised to that of Cr01. The
introduction of a weighting factor to minimise any contribution of intrinsic alignments to
King & Schneider: Suppression of intrinsic alignments in cosmic shear surveys 3
the shear correlation function is considered in Sect. 5. In Sect. 6 we present the results of
applying such a weighting factor to surveys of dierent mean redshift. We nish in Sect. 7
with a discussion of the results and a future perspective of using redshift information in
cosmic shear analysis.
For a review of cosmological weak lensing and the relevant aspects of cosmology,
see Mellier (1999) and Bartelmann & Schneider (2001; hereafter BS01). In addition, the
present status and outlook for cosmic shear studies is summarised by van Waerbeke et
al. (2002).
2. Power spectra and lensing correlation functions
In this section, the relationships between the power spectra and observable lensing cor-
relation functions are outlined; throughout, we adopt the notation of BS01.
The three-dimensional density uctuation eld (r) at comoving position r is a ho-
mogeneous and isotropic random eld, with a two-point correlation function given by
(r)(r
) C
(|r r
|) . (1)
Such a eld is described by its power spectrum P
(k)
(k
)
_
(2)
3
D
(k k
)P
(|k|) , (2)
where x
m
2c
2
_
wH
0
dw
W(w) f(w)
[f(w), w]
a(w)
, (3)
where H
0
and
m
are the values of the Hubble parameter and the density parameter at
the present epoch, and a(w) is the scale factor at comoving distance w, normalized such
that a(0) = 1 today. The horizon distance is denoted by w
H
, and f(w) is the comoving
angular diameter distance, so that the comoving separation is f(w). The function f(w)
depends on the spatial curvature, K:
f(w) =
_
_
K
1/2
sin(K
1/2
w) (K > 0)
w (K = 0)
(K)
1/2
sinh[(K)
1/2
w] (K < 0)
; (4)
later we assume that K = 0 (i.e.
m
+
W(w)
_
wH
w
dw
p(w
)
f(w
w)
f(w
)
R(w, w
) , (5)
4 King & Schneider: Suppression of intrinsic alignments in cosmic shear surveys
where p(w
)dw
) = f(w
w)/f(w
() (see for example BS01), is related to that of the density uctuations through a
variant of Limbers equation in Fourier space (Kaiser 1998)
P
() =
9H
4
0
2
m
4c
4
_
wH
0
dw
W
2
(w)
a
2
(w)
P
_
f(w)
, w
_
, (6)
where is the angular wave-vector, Fourier space conjugate to .
We see now that P
and P
which in turn
depends on
t
= Re
_
e
2i
= Im
_
e
2i
. (7)
Using the notation of Schneider et al. (2002a), the shear correlation functions are dened
as
() =
t
() =
_
0
d
2
J
0,4
() P
() , (8)
where J
n
are n-th order Bessel functions of the rst kind. P
() = 2
_
0
d [
()J
0,4
()] , (9)
making use of the orthogonality of Bessel functions. From now on we focus on
+
, since
this contains most cosmological information on the scales of interest.
Another quantity that is often determined in cosmological weak lensing studies is the
shear variance inside a circle of radius
c
, which is also related to the convergence power
spectrum
_
| |
2
_
(
c
) =
1
2
_
d P
()
4J
2
1
(
c
)
(
c
)
2
. (10)
This can also be determined directly by placing circular apertures onto the data eld but
is aected by any gaps; these may arise, for example, due to the need to mask regions
King & Schneider: Suppression of intrinsic alignments in cosmic shear surveys 5
containing bright stars and their diraction spikes. However, the shear variance can also
be obtained from
+
:
_
| |
2
_
(
c
) =
_
d
2
c
+
() S
+
_
c
_
, (11)
where van Waerbeke (2000) showed that
S
+
(x) =
_
_
_
1
_
4 cos
1
_
x
2
_
x
4 x
2
(x 2)
0 (x > 2)
. (12)
2.1. Choice of cosmology, power spectrum and source redshift distribution
Unless otherwise stated, our ducial cosmology is a CDM model with
m
= 0.3,
=
0.7 and H
0
= 70 kms
1
Mpc
1
.
A scale-invariant (n = 1, Harrison-Zeldovich) spectrum of primordial uctuations
is assumed. Predicting the shear correlation functions requires a model for the redshift
evolution of the 3-D power spectrum. The tting formula of Bardeen et al. (1986; BBKS)
is used for the transfer function, and the Peacock and Dodds (1996) prescription for the
evolution in the nonlinear regime. The power spectrum normalisation is parameterised
with
8
= 0.9, and the shape parameter = 0.21.
As an aside, the dierences in predicted shear correlation functions using the tting
formula of Eisenstein & Hu (1999) for the transfer function (rather than that of BBKS)
were found to be minimal.
The normalised source redshift distribution is parameterised using the form suggested
by Smail et al. (1995):
p(z)dz =
z
0
(3/)
_
z
z
0
_
2
exp
_
_
z
z
0
_
_
dz , (13)
where
i
=
s
i
+ (
i
) . (14)
We use the denition of ellipticity from Bonnet & Mellier (1995) where for elliptical
isophotes with axis ratio b/a 1, || = (1b/a)(1+b/a)
1
. Further, under this denition
if the unlensed source population is randomly oriented. For simplicity we will use
the notation
i
(
i
). Using (14) it follows that the expectation value
j
_
(
s
i
+
i
)
_
s
j
+
j
__
j
_
+
s
i
s
j
, (15)
King & Schneider: Suppression of intrinsic alignments in cosmic shear surveys 7
0
5e-05
0.0001
0.00015
0.0002
0.00025
0.0003
0.00035
0.0004
0.00045
1 10 100
(arcmin)
+
(
)
Fig. 2. Dependence of
+
on angular scale for z = 0.8 (dotted curve), z = 1.0
(solid curve) and z = 1.2 (short dashed curve). The cosmology for these curves is as
described in the text, with
m
= 0.3. The solid boxes correspond to z = 1.0, now with
m
= 0.38. The crosses represent z = 1.0,
m
= 0.4 and
8
= 0.78.
and in the absence of intrinsic correlations the nal term vanishes. In this case, any
correlation between observed galaxy ellipticities results from weak lensing.
As discussed in Schneider et al. (2002b), the correlation function can be estimated in
bins of width , centred on , and a weight function can be dened such that
() = 1
for /2 < + /2 and is zero otherwise. An estimator for the correlation
function
+
() is given by
+
() =
ij
W
i
W
j
(|
i
j
|)
_
j
_
N
p
()
, N
p
() =
ij
W
i
W
j
(|
i
j
|) , (16)
where W
i
and W
j
are weights (depending on the reliability of the ellipticity measurement,
for example) and N
p
() is the eective number of pairs in that bin. In the absence of
intrinsic correlations, since
j
_
+
(|
i
j
|), it follows that
+
() is an unbiased
estimator of the shear correlation function of the lensing signal
+
().
However, the existence of intrinsic alignments would imply that
s
i
s
j
= 0. The esti-
mator
+
() is no longer an unbiased estimator of
+
(): rather it includes a contribution
from correlated intrinsic (source) ellipticities, for which the two-point correlation func-
tion will be denoted by
int
+
(). In order to minimise the impact of intrinsic correlations
8 King & Schneider: Suppression of intrinsic alignments in cosmic shear surveys
on the interpretation of the weak lensing signal, we need an estimate of the amplitude of
int
+
(), and a scheme to give less weight to galaxy pairs where the intrinsic correlation
is likely to be high.
4. Intrinsic alignments
Correlations between the intrinsic ellipticities of neighbouring galaxies are expected to
arise during their formation, due to similarities in the tidal gravitational eld in which
they form. As noted in the introduction, a number of numerical and analytical stud-
ies have sought to quantify the amplitude of this correlation as a function of angular
separation and redshift. For our purposes, we need a toy model to roughly predict the
amplitude of intrinsic correlations expected. Since there is disagreement on the dominant
mechanism responsible for intrinsic alignments, and on its subsequent enhancement or
washing out, use of a detailed model is not warranted here.
Let us consider a close pair of galaxies with comoving distances w
i
w
j
w and
angular separation . Their comoving separation r is given by
r
2
= (w)
2
+ f
2
(w)()
2
, (17)
where w w
i
w
j
.
Denoting the (3-D) two-point density correlation function by
(r) 1/r behaviour, and because of its analytic simplicity, we also use this form.
Next, this 3-D intrinsic correlation function has to be projected into an angular intrin-
sic ellipticity correlation function, integrating over the distance distribution of galaxies
and taking into account the (observed) galaxy two-point correlation function
gg
(r). Note
that we do not consider the individual tangential and cross components of the correlation
functions of the intrinsic ellipticities. Rather, we consider the sum of these quantities,
which corresponds to
+
in the notation of Schneider et al. (2002a; note that the notation
of Cr01 diers here). We obtain
int
+
()
_
dw
1
p(w
1
)
_
dw
2
p(w
2
)[1 +
gg
(r)] [
(r)]
2
_
dw
1
p(w
1
)
_
dw
2
p(w
2
)[1 +
gg
(r)]
, (18)
where p(w)dw is the galaxy distance distribution, which we take to be of the form given
in Sect. 2.
Given that the scale over which intrinsic alignments operate is much smaller than the
distances to the galaxies, the integral above can be recast as
int
+
()
_
dw p
2
(w)
_
d(w) [1 +
gg
(r)] [
(r)]
2
1 +
_
dw p
2
(w)
_
d(w) [
gg
(r)]
. (19)
King & Schneider: Suppression of intrinsic alignments in cosmic shear surveys 9
The two-point galaxy correlation function is taken to be
gg
(r) = (r/4.3 h
1
Mpc)
1.8
,
constant in comoving separation. Evolution is ignored for both the galaxy and density
correlation functions, in view of the weak evolution that is seen in galaxy samples out to
z 3 (e.g. Porciani & Giavalisco 2002). The normalisation of
int
+
() is set at an angular
scale of 10
and for z = 1, using the value obtained by Cr01; Mackey et al. (2002)
point out that on very small spatial scales ( 1 h
1
Mpc) non-linear clustering eects
may erase any initial alignments, so we perform the normalisation well above this scale.
We end up with a simple model for
int
+
(), which depends upon the redshift distribution
of sources.
5. Inclusion of a redshift dependent weighting factor
An important dierence between ellipticity correlations arising from lensing and those due
to intrinsic alignment is that the latter only has a signicant amplitude over a relatively
small range in redshift. For a given galaxy pair, the lensing correlations depend on the
integrated eect of the matter along the line of sight out to the redshift of the closer
galaxy.
In order to minimise the intrinsic signal, giving less weight to galaxy pairs where
the redshift dierence is small seems like a plausible step. How eective would such a
weighting be, how would the lensing correlation be changed and how would the noise on
our estimate of the correlation function increase?
Comparison of the observed correlation functions with the foregoing expressions re-
quires one to adopt a source redshift distribution; in practice, what will become routinely
available are photometric redshift estimates. The true redshift of galaxy i will be denoted
by z
i
and the photometric redshift estimate by z
i
, with equivalent notation for distances
(i.e. w
i
and w
i
). The probability density to have sources at comoving distance w
i
and to
have photometric distance estimates w
i
will be denoted by p(w
i
) and p( w
i
) respectively,
and p(w
i
, w
j
) is the joint probability to have sources at w
i
and at w
j
. For conditional
probabilities, p(w
i
| w
i
) represents the probability that the true source distance is w
i
given
that the photometric distance estimate is w
i
. Similarly, p( w
i
|w
i
) is the probability that
a photometric distance estimate w
i
will be obtained, given that the true distance is w
i
.
5.1. Estimator
A practical estimator for the lensing correlation function should minimise the relative
signal from intrinsic alignments:
+
() =
ij
W
i
W
j
Z
i,j
(|
i
j
|)
_
j
_
N
p
()
, N
p
() =
ij
W
i
W
j
Z
i,j
(|
i
j
|) ; (20)
Z
i,j
Z( z
i
, z
j
) is a weighting factor which could be of the form
Z( z
i
, z
j
) = 1 exp
_
( z)
2
2
2
Z
_
, (21)
10 King & Schneider: Suppression of intrinsic alignments in cosmic shear surveys
where z is the dierence between the photometric redshift estimates of the galaxies in
a pair, and
Z
is a factor that controls the extent of the region in z space over which
galaxy pairs are down-weighted. This would be chosen to reect photometric uncertainty,
which is typically much larger than the range over which intrinsic correlations operate.
We will neglect the weights W in what follows (since these would be determined from
the data in practice), and see how this estimator performs in Sect. 6.
5.2. Lensing correlation function
We rst consider what eect a redshift dependent weighting factor has on the lensing
induced component of the correlation function,
+
(). The addition of such a weighting
factor modies the eective source distance distribution, and a distance correlation is
introduced. To make this clear, consider the shear correlation function for particular
source distances w
i
and w
j
i.e.
+
(; w
i
, w
j
), which would correspond to sources located
on two sheets (Schneider et al 2002a);
+
(; w
i
, w
j
) =
9H
4
0
2
0
4c
4
_
wi,j
0
dw
a
2
(w)
R(w, w
i
) R(w, w
j
)
_
d
(2)
P
_
f(w)
, w
_
J
0
() , (22)
where the upper limit w
i,j
in the outer integral is the minimum of w
i
and w
j
. What
is important to note is the dependence on the product R(w, w
i
) R(w, w
j
). When (22) is
averaged over the source distance distribution, Schneider et al. (2002a) obtain
+
() =
9H
4
0
2
0
4c
4
_
wH
0
dw
a
2
(w)
_
d
(2)
P
_
f(w)
, w
_
J
0
() R(w, w
i
) R(w, w
j
) , (23)
where the angular brackets in R(w, w
i
) R(w, w
j
) denote averaging over the probability
distribution p(w
i
, w
j
). If p(w
i
, w
j
) is the probability density to have sources at comov-
ing distances w
i
and w
j
, in the absence of intrinsic source correlations one has that
p(w
i
, w
j
) p(w
i
)p(w
j
). Then, (R(w, w
i
)R(w, w
j
)
W
2
(w), as in in (6). The addition
of the weighting factor Z( w
i
, w
j
) changes the probability density of the redshifts of pairs
included in the estimator (20) thus
p(w
i
, w
j
) =
_
wH
0
d w
i
X
_
wH
0
d w
j
Y Z( w
i
, w
j
)
_
wH
0
d w
i
_
wH
0
d w
j
Z( w
i
, w
j
)
_
wH
0
dw
i
X
_
wH
0
dw
j
Y
, (24)
where X p( w
i
) p(w
i
| w
i
) and Y p( w
j
) p(w
j
| w
j
), the inclusion of the denomina-
tor ensuring that the probability distribution is normalised. Using p( w
i
|w
i
) p(w
i
) =
p(w
i
| w
i
) p( w
i
) (and similarly for galaxy j) we have that the expectation value of
R(w, w
i
)R(w, w
j
) =
_
wH
w
dw
i
R(w, w
i
)
_
wH
w
dw
j
R(w, w
j
)
_
wH
0
d w
i
X
_
wH
0
d w
j
Y
_
wH
0
dw
i
_
wH
0
dw
j
_
wH
0
d w
i
X
_
wH
0
d w
j
Y
, (25)
where X
p(w
i
) p( w
i
|w
i
) and Y
p(w
j
) p( w
j
|w
j
)Z( w
i
, w
j
). Substitution of (25)
into (23) gives the lensing two-point correlation
+
(). This depends upon the accuracy
of photometric redshift estimates, and upon ltering using the weighting factor Z.
King & Schneider: Suppression of intrinsic alignments in cosmic shear surveys 11
5.3. Intrinsic correlation function
Now consider how accounting for photometric redshifts and applying the same weighting
factor Z( w
i
, w
j
) to the prescription for the intrinsic correlation function adjusts the
intrinsic correlation function:
int
+
() =
_
wH
0
dw
i
_
wH
0
dw
j
_
wH
0
d w
i
X
_
wH
0
d w
j
Y
[1 +
gg
(r)] [
(r)]
2
_
wH
0
dw
i
_
wH
0
dw
j
_
wH
0
d w
i
X
_
wH
0
d w
j
Y
[1 +
gg
]
, (26)
where X
and Y
i
n
t
+
i
n
t
Fig. 3. The ratio of the intrinsic correlation function
int
+
to the combined lensing and
intrinsic correlation function (
+
+
int
+
) as a function of angular scale . For all cases,
the photometric redshift error is assumed to be Gaussian distributed, with dispersion
phot
= 0.1. The lower set of three curves are for z = 1.0, with the curve marked A
corresponding to the case where there is no redshift dependent weighting factor, Z. The
curve below this is for
Z
= 0.1 and the lowermost curve is for
Z
= 0.4. The upper
set of three curves are for a lower mean redshift z = 0.5, with the curve marked B
being the case where there is no weighting factor. As for the higher mean redshift case,
the other two curves correspond to
Z
= 0.1 and
Z
= 0.4 respectively.
results in a reduction in the eective number of pairs and hence an increase in the noise,
which scales roughly as 1/
_
N
p
. The width of the weighting factor, parameterised by
Z
, controls the fraction of pairs remaining. The photometric accuracy determines the
precision with which galaxies at similar redshifts can be identied, hence the residual
contribution from intrinsic alignments.
The reduction of the eective number of pairs is not the only relevant consideration
for the noise in the determination of the two-point correlation function. Depending on the
survey geometry, cosmic variance can become the dominant contribution to the covariance
of the correlation function for larger angular separations. In particular, for a compact
survey geometry, cosmic variance of the shear eld dominates over shot noise for angular
scales larger than a few arcminutes (e.g., Schneider et al. 2002b). Hence, for those scales
the reduction of the eective number of pairs becomes of little relevance.
King & Schneider: Suppression of intrinsic alignments in cosmic shear surveys 13
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
z
F
Fig. 4. The eective fraction of pairs remaining (F) as a function of
Z
(lter width)
for z = 1.0 (solid curve) and z = 0.5 (dashed curve).
Intrinsic alignments have also been invoked as a possible source of the so-called
B-mode contributions seen in cosmic shear surveys (van Waerbeke et al. 2001, 2002;
Hoekstra et al. 2002b). Such B-modes are not expected from lensing eects and thus
are usually interpreted as a remaining systematics; whereas galaxy correlations can in
principle generate a B-mode contribution (Schneider et al. 2002a), its amplitude is very
small. If the B-mode is due to an intrinsic alignment of galaxies, then the use of the
redshift lter as discussed here should strongly suppress its relative contribution.
Choosing to apply such a weighting factor will be governed by the details of the survey.
Such considerations include that the impact of intrinsic alignments is more dramatic for
surveys of lower mean redshift or depth. Another important factor is the accuracy of
photometric redshift estimates, which depends upon the combination and characteristics
of lters used for the survey - see e.g. Wolf et al. (2001) for a discussion of photometric
redshift estimate performance in the context of lter sets. Further, the size of the error
bars associated with the practical determination of the two-point correlation function will
also play a part in the decision. For deep surveys, with our assumed photometric errors,
in order to decrease the fractional contribution of
int
+
by a few percent, a reduction of
several tens of percent in the number of eective pairs may result. However, removal of the
intrinsic alignment systematic becomes increasingly important as the experimental error
14 King & Schneider: Suppression of intrinsic alignments in cosmic shear surveys
(arcmin)
1e-06
1e-05
0.0001
0.001
1 10
+
(
)
Fig. 5. The lensing correlation function
+
() as a function of , for z = 1 (upper
curves) and for z = 0.5 (lower curves). Within each set of three curves, the uppermost
curve is for no weighting, the middle curve is for
Z
= 0.1 and the lowermost curve is
for
Z
= 0.4.
bars in surveys become smaller. In shallower surveys, the contamination from intrinsic
alignments is much more pronounced, and in this case down-weighting close galaxy pairs
is more eective. Given real data, one might consider a more elaborate weighting scheme
that is a function of not only the dierence in photometric redshifts, but also giving more
weight to higher redshift pairs. Furthermore, the weight factor Z
i,j
need not be chosen
as a function of the estimated photometric redshift only, but can be constructed such
that it depends on the full redshift probability distribution as estimated by photometric
redshift techniques. For example, a natural choice would be to have Z
i,j
depending on
the probability that the two galaxies i, j lie within a narrow redshift interval z, as
estimated from their individual z-distributions. The calculation of the expectation value
of the corresponding shear estimator is then slightly more complicated, but can be done
for each survey at hand.
Throughout we have assumed that photometric redshift estimates will be available
for each of the source galaxies, and that the dispersion in these estimates is independent
of magnitude or redshift. In practice, photometric redshift estimates are reliable down
to a certain limiting magnitude, depending on the survey characteristics and the relative
King & Schneider: Suppression of intrinsic alignments in cosmic shear surveys 15
depths of the lters used. Particularly in a deep survey, this magnitude limit may exclude
many of the fainter and smaller galaxies that are likely to be in the high-redshift tail of
the redshift distribution, and to make a substantial contribution to the lensing signal.
Requiring photometric redshift estimates could in fact lead to an increase in the Poisson
noise of such a survey by reducing the usable number of pairs beyond the eect shown
in Fig. 4, and in addition the reduction in mean redshift could increase the relative
contribution from intrinsic alignments. One possibility to counteract such aects might
be to include fainter galaxies where no photometric redshift estimates are available,
assuming that these are at higher redshifts where intrinsic alignment is less of an issue.
The use of cosmic shear surveys as a tool to place constraints on the 3-D power spec-
trum, so-called power spectrum tomography, would rely on the availability of photometric
redshift estimates for the galaxies involved in the analysis (e.g. Hu 1999; 2002). Croft
& Metzler (2000) demonstrated that the intrinsic alignment signal is more severe when
correlations within a narrow slice in source galaxy redshift space are considered. The ap-
plication of a redshift dependent lter, similar to the one we have described here, would
greatly suppress the intrinsic alignment systematic. Instead of slicing galaxies in redshift,
and calculating the correlation function for galaxies in the same redshift bin which both
increases the relative importance of intrinsic alignments and reduces the total number of
pairs one should correlate galaxies from dierent redshift bins i, j, where the bin width
would be chosen to be of the same order as the redshift uncertainty. This would yield a
set
i,j
() of correlation functions for which the intrinsic signal is strongly suppressed for
i = j. When comparing these correlation functions with predictions from cosmological
models, their covariance must be taken into account, which may turn out to be fairly
complicated (see Schneider et al. 2002b for the case with no redshift information). This
redshift slicing is most straightforwardly done with the correlation function, although for
other estimators of the two-point cosmic shear statistics, such as the shear dispersion or
the aperture mass, similar pair redshift-dependent estimators are easily constructed.
Most of what has been said above also applies to higher-order cosmic shear statistics.
It is well known that the three-point statistics, i.e. the skewness, contains very useful
cosmological information (e.g. van Waerbeke et al. 1999). As true for the 2-point statistics,
the three-point correlation function is the quantity which is most easily derived from a
cosmic shear survey. Bernardeau et al. (2002a) have constructed a statistics based on
the measured three-point correlation function (see Schneider & Lombardi 2002 for the
classication of three-point shear correlators), and Bernardeau et al. (2002b) obtained a
signicant detection in the VIRMOS-DESCART survey data. It is unclear whether, and
by how much, intrinsic ellipticity correlations aect these measurements. Furthermore,
up to now no B-mode estimator in the three-point function has been devised which may
indicate the presence of intrinsic alignment eects. Redshift slicing can of course also be
done for the three-point function which may be the only way to measure these lensing
16 King & Schneider: Suppression of intrinsic alignments in cosmic shear surveys
statistics without the potential inuence of intrinsic eects; one needs to suppress those
triplets of galaxies where the probability of all three being at the same redshift is not
negligibly small (i.e., the estimator is unaected by intrinsic eects if two of the three
are at the same redshift).
Of course, the intrinsic alignment signal that we seek to suppress when performing
a lensing analysis is also interesting in its own right, since it places constraints on the
formation and evolution of galaxies. Suitable catalogues of galaxies to quantify this signal,
and its evolution with redshift, will be a natural by-product of large cosmic shear surveys
with photometric redshift information. For example, using a lter for the determination
of the ellipticity correlation as in (20) with Z
i,j
= 1 Z
i,j
will make this estimate
dominated by the intrinsic alignments, and could thus be used as a rst estimate of it.
More sophisticated techniques would include the consideration of the resulting correlation
function in dependence of the width of the weighting function, and extracting the lensing
and intrinsic signal from this functional form.
If the intrinsic alignment of galaxies indeed occurs, and in particular if the intrinsic
eect is as strong as suggested by some models (e.g. Jing 2002), a deep multi-colour
wide-eld survey with a broad range of lters will be necessary, and rewarding, to remove
this systematic from cosmic shear measurements. Most likely, the near-IR imaging will
present the bottleneck, limiting the magnitude - and thus the eective number density -
of galaxies that can be used for cosmic shear. A wide-eld near-IR camera in space, such
as the PRIME satellite mission, would be an ideal supplement to the planned extensive
ground-based optical cosmic shear surveys.
After we had completed this paper, we became aware of work by Heymans & Heavens
(2002), also addressing how redshift information can be used to reduce the contamination
from intrinsic alignments. They apply their technique to estimate the reduction of the
intrinsic signal in several surveys, including the SDSS photometric and spectroscopic
samples.
Acknowledgements. We would like to thank Doug Clowe, Thomas Erben, Martin Kilbinger,
Patrick Simon, and in particular Marco Lombardi and Ludo van Waerbeke for useful discussions.
We would also like to thank the anonymous referee for very helpful comments. This work was
supported by the Deutsche Forschungsgemeinschaft under the project SCHN 342/31 and by
the TMR Network Gravitational Lensing: New Constraints on Cosmology and the Distribution
of Dark Matter of the EC under contract No. ERBFMRX-CT97-0172.
References
Bacon, D.J., Refregier, A.R., Ellis, R.S. 2000, MNRAS, 318, 625
Bardeen, J.M., Bond, J.R., Kaiser, N., Szalay, A.S. 1986, ApJ, 304, 15 (BBKS)
Bartelmann M., Schneider P., 1999, A&A, 345, 17
Bartelmann M., Schneider P., 2001, Phys. Rep., 340, 291 (BS01)
King & Schneider: Suppression of intrinsic alignments in cosmic shear surveys 17
Bernardeau, F., Mellier, Y., Van Waerbeke, L. 2002a, A&A, 389, L28
Bernardeau, F., Van Waerbeke, L., Mellier, Y. 2002b, A&A in press (also astro-ph/0201029)
Blandford, R.D., Saust, A.B., Brainerd, T.G., Villumsen, J.V. 1991, MNRAS, 251, 600
Bolzonella, M., Miralles, J.-M., Pello, R. 2000, A&A, 363, 476
Bonnet, H., Mellier, Y. 1995, A&A, 303, 331
Brown, M.L., Taylor, A.N., Hambly, N. C., Dye, S. 2002, MNRAS, 333, 501
Catelan, P., Kamionkowski, M., Blandford, R.D. 2001, MNRAS, 320, L7
Crittenden, R.G., Natarajan, P., Pen, U.-L., Theuns, T. 2001, ApJ, 559, 552 (Cr01)
Croft, R., Metzler, C. 2000, ApJ, 545, 561
Eisenstein, D., Hu, W. 1999, ApJ, 511, 5
Hamilton, A.J.S., Matthews, A., Kumar, P., Lu, E. 1991, ApJ, 374, 1
Heavens, A., Refregier, A., Heymans, C. 2000, MNRAS, 319, 649
Heymans, C., Heavens, A. 2002, astro-ph/0208220
Hui, L., Zhang J. 2002, astro-ph/0205512
Hoekstra, H., Yee, H., Gladders, M., Barrientos L., Hall, P., Infante, L. 2002a, ApJ, 572, 55
Hoekstra, H., Yee, H., Gladders, M. 2002b, ApJ in press (also astro-ph/0204295)
Hu, W. 1999, ApJ, 522, L21
Hu, W. 2002, astro-ph/0208093
Jain, B., Seljak, U., White, S.D.M. 2000, ApJ, 530, 547
Jing, Y.P. 2002, astro-ph/0206098
Kaiser, N. 1992, ApJ, 388, 272
Kaiser, N. 1998, ApJ, 498, 26
Kaiser, N., Wilson, G., Luppino, G. 2000, astro-ph/0003338
Mackey, J., White, M., Kamionkowski, M. 2002, MNRAS, 332, 788
Maoli, R., Van Waerbeke, L., Mellier, Y., Schneider, P., Jain, B., Bernardeau, F., Erben, T.,
Fort, B. 2001, A&A 368, 766
Mellier, Y. 1999, ARA&A, 37, 127
Miralda-Escude, J. 1991, ApJ 380, 1
Peacock, J.A., Dodds, S.J. 1996, MNRAS, 280, L19
Porciani, C., Giavalisco, M. 2002, ApJ, 565, 24
Schneider P., van Waerbeke L., Jain B., Kruse G. 1998, MNRAS, 296, 873
Schneider, P., Lombardi, M. 2002, astro-ph/0207454
Schneider, P., van Waerbeke, L., Mellier, Y. 2002a, A&A, 389, 729
Schneider, P., van Waerbeke, L., Kilbinger, M., Mellier, Y., 2002b, A&A in press (also astro-
ph/020618)
Smail, I., Hogg, D., Yan, L., Cohen, J. 1995, ApJ, 449L, 105
Van Waerbeke, L., Bernardeau, F., Mellier, Y. 1999, A&A, 243, 15
Van Waerbeke, L. 2000, MNRAS, 313, 524
Van Waerbeke, L., Mellier, Y., Erben, T. et al. 2000, A&A, 358, 30
Van Waerbeke, L., Mellier, Y., Radovich, M.. et al. 2001, A&A, 374, 757
Van Waerbeke, L., Mellier, Y., Tereno, I. 2002, astro-ph/0206245
White M., Hu W. 2000, ApJ, 537, 1
Wittman, D.M., Tyson, J.A., Kirkman, D., DellAntonio, I., Bernstein, G. 2000, Nat, 405, 143
Wolf, C., Meisenheimer, K., Roser, H.-J. 2001, A&A, 365, 660