Mon. Not. R. Astron. Soc. 000, 1–12 (2006) Printed 13 March 2018 (MN LATEX style file v2.

2)

Potential sources of contamination to weak lensing measurements:

constraints from N-body simulations

Catherine Heymans1⋆ , Martin White2,3, Alan Heavens4, Chris Vale5,2 &

Ludovic Van Waerbeke1
arXiv:astro-ph/0604001v2 16 Jun 2006

1 Department of Physics and Astronomy, 6224 Agricultural Road, University of British Columbia, Vancouver, BC, V6T 1Z1, Canada.
2 Department of Physics and Astronomy, 601 Campbell Hall, University of California Berkeley, CA 94720, USA.
3 Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, CA 94720, USA.
4 SUPA†, Institute for Astronomy, University of Edinburgh, Blackford Hill, Edinburgh, EH9 3HJ, UK.
5 Theoretical Astrophysics, Fermi National Accelerator Laboratory, Batavia, IL 60510, USA.

13 March 2018

ABSTRACT

We investigate the expected correlation between the weak gravitational shear of distant
galaxies and the orientation of foreground galaxies, through the use of numerical simulations.
This shear-ellipticity correlation can mimic a cosmological weak lensing signal, and is poten-
tially the limiting physical systematic effect for cosmology with future high-precision weak
lensing surveys. We find that, if uncorrected, the shear-ellipticity correlation could contribute
up to 10% of the weak lensing signal on scales up to 20 arcminutes, for lensing surveys with a
median depth zm = 1. The most massive foreground galaxies are expected to cause the largest
correlations, a result also seen in the Sloan Digital Sky Survey. We find that the redshift de-
pendence of the effect is proportional to the lensing efficiency of the foreground, and this
offers prospects for removal to high precision, although with some model dependence. The
contamination is characterised by a weakly negative B-mode, which can be used as a diag-
nostic of systematic errors. We also provide more accurate predictions for a second potential
source of error, the intrinsic alignment of nearby galaxies. This source of contamination is
less important, however, as it can be easily removed with distance information.
Key words: cosmology: observations - gravitational lensing - large-scale structure.

1 INTRODUCTION multi-colour surveys, weak lensing has the exciting prospect of pro-
viding constraints to distinguish between a cosmological constant
Weak gravitational lensing by large-scale structure provides a
(w = −1) and quintessence models (w(z) > −1).
test of cosmology that is based on the well understood physics
Lensing by large-scale structure weakly distorts images of
of gravity. The lensing strength is directly related to the large-
background galaxies such that the observed galaxy ellipticity
scale distribution of matter and thus enables constraints to be
placed on cosmological parameters such as the matter density eobs ≃ es + γ , (1)
parameter Ωm and the amplitude of the matter power spectrum
σ8 (Van Waerbeke et al. 2001; Hoekstra et al. 2002; Bacon et al. where es is the intrinsic galaxy ellipticity, or source ellipticity, and
2003; Jarvis et al. 2003; Brown et al. 2003; Hamana et al. 2003; γ is the weak lensing shear distortion. Lensing by large-scale struc-
Massey et al. 2005; Rhodes et al. 2004; Van Waerbeke et al. 2005; ture is coherent and hence it induces weak correlations in the ob-
Heymans et al. 2005; Semboloni et al. 2006; Hoekstra et al. 2006). served ellipticities of galaxies such that heaobs ebobs iθ > 0, where the
Combined with redshift information, measurements of weak lens- average is taken over all galaxy pairs (a, b) separated by angle θ.
ing are also able to constrain the equation of state of the dark energy From equation 1,
w(z) as lensing can probe the effect dark energy has on both the heaobs ebobs i = heas ebs i + hγ a ebs i + hγ b eas i + hγ a γ b i . (2)
evolution of the matter power spectrum and the redshift-distance
relationship. With the next generation of deep-field and wide-field In order to extract the weak cosmological signal hγ a γ b i from
the observed galaxy ellipticity correlation heaobs ebobs i, all intrin-
sic terms including es are typically assumed to be negligible.
⋆ [email protected] There are however potential physical limitations to this assump-
† Scottish Universities Physics Alliance tion related to correlations between galaxy shape and the den-
2 Heymans et al.
sity field. This could arise, for example, through tidal torques tigated observationally by Mandelbaum et al. (2006) who measure
operating during galaxy formation (Hoyle 1949; Peebles 1969; the correlation between the intrinsic ellipticities of nearby galaxies
Heavens & Peacock 1988). Such a correlation would introduce an and the density field that surrounds them in the Sloan Digital Sky
intrinsic alignment between nearby galaxies mimicking the weak Survey (SDSS). As the density field is responsible for the gravi-
lensing effect, i.e if galaxies a and b were physically close heas ebs i =
6 tational lensing of background galaxies, Mandelbaum et al. (2006)
0. Investigated observationally (Brown et al. (2002); Lee et al. are able to estimate the shear-ellipticity correlation for any lensing
(2002); Heymans et al. (2004), hereafter H04, Mandelbaum et al. survey. They find that for most galaxy types, the shear-ellipticity
(2006)), analytically (Catelan et al. 2001; Crittenden et al. 2001; correlation is consistent with no signal. The shapes of the most lu-
Lee & Pen 2001; Porciani et al. 2002; Mackey et al. 2002) and minous red galaxies are however found to be be strongly correlated
through numerical simulations (Heavens et al. (2000), hereafter with the density field. For a lensing survey with a median source
HRH, Croft & Metzler (2000); Jing (2002), H04), it is currently redshift zm ∼ 1 the most luminous galaxies would then introduce
thought that this intrinsic alignment effect is weak, at the level of a shear-ellipticity correlation signal that is found to be at a similar
a few percent contamination to the lensing signal from a deep sur- magnitude to the pure lensing correlation signal from a ΛCDM cos-
vey of median redshift zm ∼ 1. Division of a redshift survey for mology. Averaging over the full galaxy sample, Mandelbaum et al.
lensing tomography (see for example Hu 1999), however, signifi- (2006) suggest that there would be a ∼ 5% contamination to the
cantly increases the contamination in the lower redshift bins (H04). pure lensing signal. These results are however subject to noise (the
As shown by Heymans & Heavens (2003) and King & Schneider results for the full galaxy sample are also consistent with zero con-
(2002) this contamination can be effectively removed at high ac- tamination), and are based on measurements taken at very low red-
curacy by optimally downweighting pairs of nearby galaxies in the shift and projected to the higher redshifts that are probed by lens-
lensing analysis (see H04 for an implementation of this method). ing surveys. In this paper we therefore aim to address this issue
Hence this physical systematic need not limit the accuracy of fu- through the analysis of extensive numerical dark matter simula-
ture multi-colour weak lensing surveys. tions (Vale & White 2003), following the philosophy of HRH and
Correlations between a galaxy’s shape and its surrounding H04 whereby galaxy ellipticity is defined from the properties of the
density field may also introduce another more subtle effect, first parent dark matter halos.
discussed by Hirata & Seljak (2004). Consider the image of a dis- This paper is organised as follows. In section 2 we describe
tant galaxy lensed by the dark matter density surrounding a fore- the N-body simulations that we use in this analysis and the different
ground galaxy. If the shape of the foreground galaxy is correlated models used to populate dark matter halos with luminous galaxies.
with the tidal gravitational field, this could lead to a non-zero cor- We update the results of HRH and H04 in section 3 by present-
relation1 between intrinsic foreground galaxy ellipticity es and the ing higher signal-to-noise measurements of the E and B-mode in-
lensing shear distortion experienced by the distant galaxy γ, i.e trinsic alignment correlation heas ebs iθ from our N-body simulations
if galaxy a is always more distant than galaxy b, hγ a ebs i 6= 0. and we compare these results to the recent observational measure-
Hirata & Seljak (2004) explore two analytical models to estimate ments of Mandelbaum et al. (2006). We also investigate the effect
the magnitude of this effect finding that when galaxy ellipticity is of nearby voids on galaxy shape and compare our results to the ob-
quadratically related to the tidal or density field, as suggested by servational measurements of Trujillo et al. (2006). In section 4 we
tidal torque theory, no shear-ellipticity correlation is found. When present measurements of the E and B-mode shear-ellipticity corre-
galaxy ellipticity is linearly related to the tidal or density field, how- lation hγ a ebs i and measure the redshift dependence of this signal.
ever, the shear-ellipticity correlation can be similar to the magni- This allows us to provide fitting formulae for the shear-ellipticity
tude of the weak lensing shear correlation. In this model, galaxies correlation measured from different galaxy models. We discuss our
are assumed initially to be spherical and acquire ellipticity by re- findings in section 5 and conclude in section 6.
sponding to the linear tidal field. For a spherically symmetric over-
density, the tidal field compresses the spherical galaxy along the di-
rection to the over-density centre. Around a spherically symmetric
proto-void, the linear theory would stretch the galaxy along the di- 2 N-BODY SIMULATIONS
rection to the void centre. There is some observational evidence to The basis of our calculation is a large N-body simulation of a
support the linear model from Trujillo et al. (2006), who find that ΛCDM cosmology. The simulation used 5123 particles in a pe-
the rotation axes of galaxies located near voids are preferentially riodic cubic box 300h−1 Mpc on a side. This represents a large
aligned with the void surface. enough cosmological volume to ensure a fair sample of the Uni-
If a shear-ellipticity correlation does exist then its strength verse, while maintaining enough mass resolution to identify galac-
is expected to be dependent on the lensing efficiency of the lens- tic mass halos. The cosmological model is chosen to provide a
source pair (Hirata & Seljak 2004). King (2005) has developed reasonable fit to a wide range of observations with Ωm = 0.3,
this idea to show that using correlation function tomography and ΩΛ = 0.7, H0 = 100 h kms−1 Mpc−1 with h = 0.7, ΩB h2 =
knowledge of the redshift dependence of each source of contamina- 0.02 and a scale-invariant spectrum with σ8 = 0.8. The simu-
tion, one could separate the true cosmological lensing signal from lation was started at z = 60 and evolved to z = 0 using the
both the intrinsic galaxy alignment contamination and the shear- TreePM code of White (2002). The full phase space distribution
ellipticity correlation contamination. was recorded every 100h−1 Mpc from z ≃ 2 to z = 0. The grav-
The shear-ellipticity correlation effect has recently been inves- itational force softening was of a spline form, with a “Plummer-
equivalent” softening length of 20 h−1 kpc comoving. The particle
mass is 1.7 × 1010 h−1 M⊙ allowing us to find bound halos with
1 We hereafter refer to this effect as a shear-ellipticity correlation. In the masses several times 1011 h−1 M⊙ .
literature Hirata & Seljak (2004) and King (2005) refer to this effect as in- For each output we determine the halo catalogue by running
trinsic alignment-lensing interference and Mandelbaum et al. (2006) use the a “friends-of-friends” (FoF) group finder (e.g. Davis et al. (1985))
term GI contamination. with a linking length b = 0.15 in units of the mean inter-particle
 Potential sources of contamination to weak lensing 3
spacing. This procedure partitions the particles into equivalence in H04. Note that for completeness, we also present the results,
classes, by linking together all particle pairs separated by less than in section 3, for a spiral model where the disk is aligned with the
a distance b. This means that FoF halos are bounded by a surface of angular momentum vector of its parent halo, as in HRH. The re-
density roughly 140 times the background density. For each halo sults in the shear-ellipticity correlation analysis that follow in sec-
we computed the angular momentum and the moment of inertia tion 4, however, are unaffected by whether the disk is misaligned or
tensor, using the centre of mass of the particles in the FoF group as not, and we therefore only present the shear-ellipticity correlation
the centre of the halo. results from our misaligned spiral model. Our mix model popu-
A past light cone was constructed by stacking the boxes back lates dark matter halos with the same elliptical or misaligned spiral
along the line of sight, with boxes shifted by random amounts in galaxies according to the early-type and late-type CLFs of Cooray
the x- and y-directions and viewed down randomly chosen axes to (2006). These CLFs are constrained such that they reproduce the lu-
avoid artificial periodicity. The full projected mass distribution was minosity functions and clustering properties of red and blue galax-
used to create convergence maps of κ at a source redshift plane of ies. Our simple galaxy ellipticity models will clearly break down
zs ∼ 1 and zs ∼ 0.5, where κ is defined to be the dimensionless in the case of parent halos which host more than one galaxy. We
surface mass density. The gravitational shear γ = (γ1 , γ2 ) that a therefore limit our study to singly occupied halos which satellite
galaxy at zs ∼ 1, or zs ∼ 0.5 would experience is then created galaxies are unlikely to inhabit. For our mix model, this criterion
from the κ maps (Kaiser & Squires 1993). The halo information results with roughly one third of the halos populated with elliptical
was transformed into the same coordinate system to create a light galaxies at all redshifts.
cone halo distribution. 12 different realizations of the 20482 pixel In this paper we determine the ellipticity parameters εi for
maps were made in all, each with a 25 square degree field-of-view. each galaxy, where εi for a perfect ellipse with axial ratio β at po-
sition angle θ, is given by,
1 − β2
   
2.1 Galaxy Models ε1 cos 2θ
= . (3)
ε2 1 + β2 sin 2θ
From a catalogue of dark matter halos, there are two steps required
to simulate luminous galaxies with ellipticities. Firstly we need to In order to study the effect of these intrinsic galaxy shapes on weak
populate the halos with galaxies, for which, in this paper, we use the lensing measurements we then define and use the alternative ellip-
conditional luminosity function (CLF) of Cooray & Milosavljević ticity parameters ei given by
(2005), and secondly we need to assign ellipticities. We follow εi
HRH and H04 in relating the ellipticities e of the galaxies to the ei = , (4)
2R
ellipticity of the halo (the ‘elliptical’ model) or via the angular mo-
where the responsivity R is a correction to account for the response
mentum of the halo (the ‘spiral’ model). In this section we first
of an ensemble of galaxies to weak lensing shear (Rhodes et al.
describe the CLF halo occupation model, and then discuss the el-
2000; Bernstein & Jarvis 2002). Without this responsivity correc-
liptical, spiral and galaxy mix models that we use in the analysis
tion to our ellipticity measurements, equation 1 would not apply.
that follows.
For the following unweighted analysis, R = 1 − hε2i i = 0.89 for
The CLF, denoted by Φ(L|Mh , z) gives the average number
all galaxy models (although note that for simplicity in HRH and
of galaxies with luminosities between L and L + dL that reside
H04 the responsivity R was set equal to 1). The resulting ei ellip-
in halos of mass Mh at a redshift of z. The CLF model is sepa-
ticity distributions have a mean hei i = 0 and a width σe = 0.33
rated into terms associated with central and satellite galaxies with
which is very similar to what is measured from space-based data
a further division into early-type and late-type galaxies. The free
(Heymans et al. 2006).
model parameters are then constrained in Cooray (2006) by obser-
vations of the galaxy luminosity function and galaxy correlation
function over a wide range of redshift samples (for example the
SDSS at z < 0.1 (Zehavi et al. 2004) and the COMBO-17 survey 3 INTRINSIC GALAXY ALIGNMENTS
at 0.4 < z < 0.8 (Wolf et al. 2003)). The results of this analy-
In this section we update the analysis of HRH and H04 by re-
sis show a very convincing match between CLF model predictions
measuring the intrinsic alignment correlation heas ebs i from simu-
and observational results at all redshifts and thus provide an ade-
lations. The N-body simulation used in this analysis is at a signifi-
quate model with which to populate simulated dark matter halos
cantly higher resolution than the VIRGO simulations (Jenkins et al.
with luminous galaxies. The average number of galaxies that re-
1998) used by HRH allowing for more reliable measurements of
R in a halo of mass Mh at redshift z is given by N (Mh , z) =
side
halo angular momentum. It also covers a larger cosmological vol-
dL Φ(L|Mh , z) where the luminosity ranges from the faintest
ume and thus yields higher signal-to-noise measurements. The evo-
observable magnitude (which we set to a typical ground-based sur-
lution of the large-scale structure is now preserved, permitting a
vey depth of r < 25.5), to an absolute brightest galaxy magnitude
direct measurement of the E and B modes of the angular intrinsic
Mr > −23. We use the best fit Cooray (2006) CLF parameters
alignment correlation. As a result we do not have to rely on pro-
or the chosen fiducial values in the cases where the parameters are
jecting a measured real space correlation function into an angular
poorly constrained. This results in a realistic redshift distribution of
correlation function using weakly constrained redshift dependent
the foreground galaxies with roughly 15 galaxies per square arcmin
galaxy clustering models. The analyses of the simulations differ
with z < 1.0.
slightly in the fact that we now populate halos according to the
We use three simple galaxy models which we term, ellipticals,
CLF of Cooray (2006) instead of introducing low and high mass
spirals and mix. Ellipticals are defined to have the same ellipticity
cuts to remove halos less likely to harbour a single galaxy. Our
as their parent halos. Spirals are modelled as a thick disk randomly
galaxy models are however identical, with the new addition of the
misaligned with the angular momentum vector is its parent halo,
mix model. The intrinsic ellipticity correlation η(rab ), is given by
with a mean misalignment angle of 20◦ as suggested by the re-
sults of van den Bosch et al. (2002) and discussed in more detail η(rab ) = het (ra )et (rb )i + her (ra )er (rb )i , (5)
4 Heymans et al.
Model Parameter het et i her er i η(rab )

Ellipticals A/10−3 1.47+0.05

−0.02 0.38+0.18
−0.08 1.75+0.08
−0.02

B 2.94+0.07
−0.04 0.96+0.40
−0.30 2.77+0.09
−0.07

Mix A/10−3 0.45+0.10

−0.05 0.07+.08
−0.05 0.47+0.13
−0.07

Spirals A/10−3 0.30+0.05

−0.02 −0.02+0.07
−0.05 0.28+0.10
−0.05

Misaligned A/10−3 0.15+0.08

−0.02 0.00+0.05
−0.02 0.07+0.10
−0.05

HRH* A/10−3 - - 0.35 ± 0.16

SDSS A/10−3 - - 0.57 ± 0.72

Table 1. The intrinsic alignment amplitude A and scale length B as defined

in equation 6 for the tangential het et i, and rotated her er i ellipticity corre-
lation functions, as well as the total ellipticity correlation η(rab ). For the
elliptical model, both A and B are free parameters. For all other models
B is constrained to be B = 1 h−1 Mpc. The best fits to the results pre-
sented in this paper can be compared to the result from H04 (HRH*) and
Mandelbaum et al. (2006) (SDSS), where we have included a scaling factor
of 4R2 to account for the shear responsivity (equation 4) that we include
in this analysis. The quoted errors are all 1σ, except in the case of SDSS
which corresponds to a 3σ constraint.

Figure 1. The intrinsic galaxy ellipticity correlation η(rab ) as a function of

comoving galaxy separation rab as measured from the simulation assum- we have included a scaling factor of 4R2 to account for the shear
ing three different galaxy models; ellipticals, spirals (placed perpendicular responsivity (equation 4) that we include in this analysis. It is in-
to the angular momentum vector as in HRH, lower left, and misaligned, as teresting to note that the rotated component of the ellipticity cor-
in H04 and the rest of this paper, lower right) and mix. The shaded region
relation function her er i, is found to be weaker than the tangential
shows the 95% confidence region constrained from measurements of in-
component het et i, which is in contrast to previous assumptions that
trinsic galaxy alignments in the SDSS Mandelbaum et al. (2006), where the
best-fit is over-plotted (solid). The best fit model from the misaligned spiral het et i ≃ her er i .
galaxies measured in H04 is shown dotted. Taking all halos between a redshift of 0 < z < 1 we mea-
sure the angular galaxy ellipticity correlation hes (θa )es (θb )i, di-
rectly from the simulations. Crittenden et al. (2002) show that the
correlation functions can be decomposed into the following E- and
where the comoving galaxy separation rab = |ra − rb |, and het et i
B-type correlators,
and her er i are the tangential and rotated correlation functions (see
for example H04). The measured η(rab ) is shown in Figure 1 for ξ+ (θ) + ξ ′ (θ) ξ+ (θ) − ξ ′ (θ)
ξ E (θ) = ξ B (θ) = (7)
each galaxy model, where for completeness we have also included 2 2
the HRH spiral model where the disk is placed perpendicular to where, in this case,
the angular momentum vector of the halo. The shaded region in
each plot is the 1σ confidence region for η(rab ) as constrained ξ± (θ) = het et iθ ± her er iθ , (8)
from measurements of intrinsic galaxy alignments in the SDSS and
(Mandelbaum et al. 2006). The strength of the signal for the ellip- Z ∞ Z ∞
dϑ dϑ
tical model at separations r > 3h−1 Mpc is inconsistent with the ξ ′ (θ) = ξ− (θ) + 4 ξ− (ϑ) − 12θ2 ξ− (ϑ). (9)
θ ϑ θ ϑ3
SDSS results at the 6σ level and we can thus reject this model at
99.9% confidence. The spiral and mix models are consistent with Following the method detailed in Pen et al. (2002), we have
the SDSS constraints. All results are consistent with the previous decomposed the ellipticity correlation functions into their E and
results from HRH and H04, and now supersede those constraints. B modes ξ E/B , assuming that hea eb iθ = 0 for θ > 20 arcmin.
Following H04, we provide parameter fits to these results us- Note that we treat each realization of the N-body simulations as an
ing the following model independent pointing and plot the mean and co-variance between
the results from each realisation. The results are shown in Fig-
A
hei (ra )ei (rb )i = , (6) ure 2 where the measured E-mode can be compared to the expected
1 + (rab /B)2
ΛCDM shear correlation from a weak lensing survey with median
where the amplitude A is chosen to be a free parameter. For the el- redshift zm ∼ 0.7. This redshift corresponds to the median red-
liptical model, B is also allowed to be a free parameter. For all other shift of the galaxies, or the populated halos, in the simulations. For
models B constrained to B = 1 h−1 Mpc. Table 1 compares the both the spiral and mix models, the signal is consistent with zero
best fitting parameters for the tangential het et i and rotated her er i on all scales, although see H04 for a noise-free E/B mode analy-
and rotated her er i ellipticity correlation functions, as well as the sis of the expected angular correlation signal from the best-fitting
total ellipticity correlation η(rab ). For the elliptical model, both A model to η(rab ), which is directly applicable to the spiral and mix
and B are free parameters. For all other models B is constrained models analysed in this work. The elliptical model exhibits strong
to be B = 1 h−1 Mpc. The best fits to the results presented in E and B modes out to θ < 10 arcmin, although the small angle
this paper can be compared to the result from H04 (referred to as B-mode are consistent with zero. Figure 3 shows the E/B mode de-
HRH* in that paper) and Mandelbaum et al. (2006) (SDSS), where composition of the elliptical model signal when we have repeated
 Potential sources of contamination to weak lensing 5
the analysis removing galaxy pairs that have a physical separation
of rab < 10h−1 Mpc, the resulting E and B mode signal is consis-
tent with zero on all scales, as expected.

3.1 Intrinsic Galaxy Alignment with Voids

We follow the analysis of Trujillo et al. (2006) to determine the de-
gree of alignment between the rotation axis of galaxies located on
the shells of cosmic voids, and the surface of those voids. We use
a section of the simulation at fixed output time z ∼ 0.325 and
take all halos with mass Mh > 8.5 × 1011 h−1 M⊙ yielding a sim-
ilar number density of the 2dFGRS and SDSS surveys analysed
by Trujillo et al. (2006). Voids are selected using the same method
of Patiri et al. (2006) and for each galaxy within rvoid < r <
rvoid + 4h−1 Mpc of the void centre (which has radius rvoid ) we
calculate cos(θ) = r.L/(|r||L|) where L is the angular momen-
tum of the parent halo. Figure 4 shows the resulting distribution of
galaxies residing on the surface of voids with 0 6 θ 6 90 degrees.
The errors come from a bootstrap analysis, and the solid line is the
average result when the halo angular momenta are randomised. We
find a distribution that is consistent with the null hypothesis that the
angular momentum vector of halos near voids is randomly oriented
with respect to the centre of voids. For our spiral galaxy model,
where galaxy orientation is determined by the angular momenta of
the parent dark matter halo, this result implies that no significant
correlation exists between the orientation of galaxies and the cen-
tre of nearby voids in our simulations. This is in contrast to the
Trujillo et al. (2006) detection results, which are over-plotted on
figure 4 that can reject this null hypothesis at the 99.7% confidence
level. These results are well fitted by the analytic prediction of Lee
(2004), (shown dot-dashed). We discuss this difference further in
section 6.

4 SHEAR-ELLIPTICITY CORRELATIONS
In this section we present shear-ellipticity correlation functions and
their E/B mode decomposition for the three different galaxy mod-
els, as described in section 2.1. Figure 5 shows the measured cor-
relation hγei between foreground galaxy ellipticity e (where the
foreground galaxies have redshifts 0 < z < 1) and the gravita-
tional shear γ experienced by a set of background source galaxies
at zs ∼ 1 where
hγeiθ = hγt (θa )et (θb )i + hγr (θa )er (θb )i . (10)
Note that we treat each realization of the N-body simulations as
an independent pointing and plot the mean and variance between
the results from each realisation. This result represents the level of
contamination that one would expect from each galaxy model for a
deep lensing survey with a median redshift zm ∼ 1. These results
show an anti-correlation, as predicted by Hirata & Seljak (2004)
and measured by Mandelbaum et al. (2006). The strongest signal is Figure 2. The E mode (circle) and B mode (cross) decomposition of the
found from the elliptical model, where all occupied halos contain angular intrinsic galaxy ellipticity correlation hea b
s es iθ as a function of an-
a galaxy whose shape is the same as its parent dark matter halo. gular galaxy separation on the sky, θ. This is measured from the simulation
The spiral model shows a correlation that is consistent with zero assuming three different galaxy models; ellipticals (upper panel), spirals
on all scales, and the mix model shows a weak but significant anti- (middle panel) and mix (lower panel). The resulting E-mode can be com-
correlation on scales θ < 20 arcmin. For reference, the solid theory pared to the expected shear correlation (solid line) from a weak lensing
curve shows the the shear-shear correlation function (multiplied by survey with median redshift zm = 0.7, which corresponds to the median
−1) for a deep zs = 1 survey using the adopted ΛCDM cosmology. redshift of the galaxies in the simulations. For clarity the B-mode points are
slightly offset to the right.
Note that the shear-ellipticity correlation for the mix model has a
similar magnitude to the models tested in King (2005). In contrast
however, we find signal that has power out to angular scales that
6 Heymans et al.

Figure 3. The measured angular intrinsic galaxy ellipticity correlation

hea b
s es iθ as a function of angular galaxy separation on the sky, θ when
galaxy pairs that have a physical separation of rab < 10h−1 Mpc are re-
moved from the analysis. This is measured from the simulation assuming
the elliptical galaxy model (right) and mix galaxy model (left). The shear
correlation from a weak lensing survey with median redshift zm = 0.7,
which corresponds to the median redshift of the galaxies in the simulations,
is shown for reference (solid line).

Figure 5. The shear-ellipticity correlation function hγeiθ for the elliptical

(circles), spiral (triangle) and mix (filled) galaxy models. In this result all
source galaxies (that experience a lensing shear γ) are placed at a source
redshift zs = 1. The foreground galaxies, with intrinsic ellipticity e, pop-
ulate the redshift space 0 < z < 1. These results are tabulated in table 2
along with the results for a survey were all source galaxies are placed at
zs = 0.5. For reference, the solid theory curve shows the shear-shear cor-
relation function (multiplied by −1) for a deep zs = 1 survey using the
adopted ΛCDM cosmology.

zm ∼ 1 θ ∼ 0.5 θ∼1 θ ∼ 10

Elliptical −19 ± 1 % −20 ± 1 % −31 ± 2 %

Spirals 1±1% 1±1% 2±2%
Figure 4. The distribution of galaxies residing near to a void as a function of Mix −6 ± 1 % −6 ± 1 % −10 ± 2 %
their degree of alignment between the rotation axis of the galaxies and the zm ∼ 0.5 θ ∼ 0.5 θ∼1 θ ∼ 10
surface of the void (circles). For a perfectly aligned set of galaxies we would
expect to see a peak at θ = 90◦ . Instead our result is consistent with the Elliptical −67 ± 5 % −43 ± 5 % −82 ± 8 %
null hypothesis that galaxies are randomly oriented with all values of cos(θ) Spirals 8±5% 3±5% −10 ± 8 %
equally likely. The errors come from a bootstrap analysis, and the solid Mix −24 ± 4 % −14 ± 5 % −31 ± 8 %
line is the average result when the halo angular momenta are randomised.
Our results can be compared with the observational results of Trujillo et al. Table 2. The percentage of contamination to a measurement of the shear-
(2006) (crosses) and the best fitting analytical prediction to these results shear correlation function by a shear-ellipticity correlation from each galaxy
from Lee (2004), shown dot-dashed. model at θ = 0.5, 1.0, 10.0 arcmin. Results are given for both a deep zs ∼
1 lensing survey (upper rows), whose results are plotted in figure 5, and
medium deep lensing zs ∼ 0.5 survey (lower rows).
are significantly larger than the 1 arcmin cut-off considered in King
(2005).
Table 2 details the level of contamination to a weak lensing to the mass of the lens. It is also reminiscent of what in seen in
measurement by a shear-ellipticity correlation from each galaxy Mandelbaum et al. (2006), where the strongest signal is measured
model, for both a deep zs ∼ 1 survey, whose results are plotted from the most luminous and hence the most massive galaxy sam-
in figure 5, and medium deep zs ∼ 0.5 survey. The mix model ple. This result allows us to briefly discuss the impact of multiply-
shows a level of shear-ellipticity correlation which is −6 ± 1% of occupied halos, which we have removed from our analysis as a
the lensing signal at an angular scale θ = 1 arcmin for the deep result of our uncertainty in how to connect the properties of the par-
zs ∼ 1 survey. ent dark matter halo to the luminous galaxy members. One could
Figure 6 shows the shear-ellipticity correlation function hγei consider the case where the central galaxy is represented by our el-
for the mix galaxy model for two samples of galaxies split by liptical model and the satellite galaxies are randomly oriented. The
mass. The strongest shear-ellipticity correlation is seen for the most multiply occupied halos are typically the more massive halos, and
massive galaxy sample where Mh > 1.0 × 1012 h−1 M⊙ . This we would therefore expect to see a strong signal from the central
is expected, as the strength of the lensing shear is proportional galaxies, as seen with the most massive galaxy sample shown in fig-
 Potential sources of contamination to weak lensing 7

Figure 6. The shear-ellipticity correlation function hγeiθ for the mix

galaxy model for two samples of galaxies split by mass; where Mh >
1.0 × 1012 h−1 M⊙ (open) and where Mh < 1.0 × 1012 h−1 M⊙ (filled).
Note that all source galaxies (that experience a lensing shear γ) are placed
at a source redshift zs = 1 and that for reference, the solid theory curve
shows the the shear-shear correlation function (multiplied by −1) for a deep
zs = 1 survey using the adopted ΛCDM cosmology.

ure 6. Randomly oriented satellite galaxies would not contribute to

the shear-ellipticity correlation signal and would thus dilute effect
of the central galaxy. The reality of this scenario is however diffi-
cult to quantify, and indeed there is evidence that suggests satellite
galaxies are not randomly oriented with a tendency to align with the
major axis of the central galaxy (Brainerd 2005; Yang et al. 2006).
We therefore limit our discussion, in the rest of this paper to singly-
occupied halos, deferring a quantitative analysis that includes satel-
lite galaxies for a future paper analysing higher resolution N-body
simulations.
Following Crittenden et al. (2002) and the method detailed in
Pen et al. (2002), we decompose the measured shear-ellipticity cor-
relation functions into their E and B modes, equation 7, assuming
that hγeiθ = 0 for θ > 20 arcmin. Figure 7 shows the result-
ing E and B mode signals for all three galaxy models. As expected
the E and B modes from the spiral model are consistent with zero.
For the mix and elliptical models, the B-modes are negative but
weaker than the corresponding E-modes. The B-modes are how-
ever fairly constant on scales θ < 15 arcmin and would therefore
be detectable if all B-modes from other non-lensing sources, such Figure 7. The E mode (circle) and B mode (cross) of the shear-ellipticity
correlation function for the elliptical (upper panel), spiral (middle panel)
as residual point-spread function distortions, are negligible.
and mix (lower panel) galaxy models. For clarity the B mode points are
slightly offset to the right.

4.1 Redshift dependence

In this section we measure the dependence of the shear-ellipticity although at lower significance, and the shear-ellipticity correlation
correlation signal as a function of the foreground galaxy redshift. for the spiral model remains consistent with zero for all redshift
We present the results for the elliptical model as the strength of the slices.
signal allows us effectively to probe the redshift dependence in the Figure 8 shows the redshift dependent shear-ellipticity corre-
presence of increased noise from the redshift binning. The trends lation for 5 different galaxy redshift slices; 0.0 − 0.4, 0.4 − 0.65,
seen with the elliptical model are also seen with the mix model, 0.65 − 0.75, 0.75 − 0.85, 0.85 − 0.95. This result demonstrates
8 Heymans et al.

Figure 9. 2σ constraints on the amplitude A and scale dependence θ0

(equation 12) of the shear-ellipticity correlation assuming the elliptical
galaxy model. Parameter constraints are determined from the 5 different
galaxy redshift slices shown in figure 8 where γ is measured at zs = 1.0,
plus 2 additional low redshift slices where γ is measured at zs = 0.5.

where the distances D are all comoving angular diameter distances

to the lens Dl , to the source Ds , and between the lens and source
Dls . We might expect the shear-ellipticity correlation signal to
scale with the lensing efficiency and to test this hypothesis, the
lower panel of figure 8 shows the shear-ellipticity correlation sig-
nal with the lensing efficiency factored out; we divide by E for the
median lens redshift in each redshift slice, and multiply by a nor-
malising factor N set equal to the lensing efficiency E of the mean
redshift lens for the whole halo sample where zl = 0.68. The result
shows that the redshift dependent shear-ellipticity correlation mea-
surements are very consistent with one another once the lensing
efficiency is factored out. This then motivates the following param-
eterisation of the redshift evolution of the shear-ellipticity correla-
tion,
A
hγ(zs ) e(zl )iθ = E , (12)
θ + θ0
where the amplitude A and scale dependence θ0 are free parame-
ters. E is the lensing efficiency of the lens-source (γe) pair (equa-
tion 11), and zl is set equal to the median redshift of each galaxy
redshift slice. Note that this parameterisation is similar to the ex-
Figure 8. The redshift dependence of the shear-ellipticity correlation as ponential parameterisation proposed by King (2005), but this func-
a function of galaxy redshift slice zl . The upper panel shows the mea- tional form is found to produce a better fit to the results. Figure 9
sured signal for the following galaxy redshifts; 0.0 < zl < 0.4 (circle), shows the two parameter 2σ constraints on the amplitude A and
0.4 < zl < 0.65 (solid), 0.65 < zl < 0.75 (square), 0.75 < zl < 0.85 scale dependence θ0 as measured independently from the 5 differ-
(hexagon), 0.85 < zl < 0.95 (triangle) . The dashed line shows the ellip- ent galaxy redshift slices shown in figure 8 where γ is measured at
tical measurement in Figure 5 from the full galaxy sample 0.0 < zl < 1.0. zs = 1.0, plus 2 additional low redshift slices where γ is measured
The lower panel shows the same signal with the lensing efficiency E fac-
at zs = 0.5. The consistent results for each galaxy redshift slice
tored out and normalized to the lensing efficiency N of the mean galaxy
suggest that the evolution of the shear-ellipticity correlation can be
redshift of the full galaxy sample where zl = 0.68
safely modelled through the lensing efficiency of the lens-source
pair.
Using this parameterisation (equation 12) we place constraints
that the most significant anti-correlation is found for the low red- on the amplitude A and scale dependence θ0 of the shear-ellipticity
shift (z < 0.65) galaxy samples, which are the most efficient for correlation measured jointly from the redshift dependent results.
lensing source galaxies at zs = 1. The lensing efficiency E of the Figure 10 shows the resulting two-parameter 1 and 2σ constraints
lens-source pair is defined to be for the elliptical model (upper) and mix model (lower). The best
Dl Dls fitting results for the elliptical model are A = (−5.60 ± 0.18) ×
E= . (11) 10−7 h Mpc−1 and θ0 = 1.83 ± 0.13 arcmin. For the mix model
Ds
 Potential sources of contamination to weak lensing 9
are negligible. We must therefore seek to derive and apply optimal
methods to remove these sources of contamination in the analyses
of the next generation of weak lensing surveys. In this paper we
have sought to answer the question, what level of contamination
might one expect, and our conclusions are clearly dependent on
our choice of galaxy model; ellipticals, spirals or mix.
A wealth of observational evidence allows us immediately
to rule out the elliptical model on several grounds. The mea-
surement of the intrinsic alignment signal η(r) from the SDSS
(Mandelbaum et al. 2006) is inconsistent with the elliptical model
signal at the 99.9% level, as shown in Figure 1. Similar constraints
could also be placed from the intrinsic alignment signal measured
in the SuperCOSMOS survey (Brown et al. 2003). Figure 2 shows
the expected angular ellipticity correlation in the absence of weak
lensing for a survey with median redshift zm = 0.7. This medium
depth survey is similar to the Red sequence Cluster Survey (RCS)
Hoekstra et al. (2002) and the CTIO survey (Jarvis et al. 2003).
Weak lensing analyses of these two surveys constrain the amplitude
of the matter power spectrum to be σ8 ∼ 0.8 ± 0.15 for Ωm = 0.3,
which is consistent with the results from the analysis of the cos-
mic microwave background (Spergel et al. 2006). If more than half
of the RCS and CTIO measured lensing signal came from intrinsic
alignments, as suggested by the results of the elliptical model, the
‘true’ value of σ8 would be then be very low and inconsistent with
other observational constraints. The COMBO-17 survey has been
analysed using redshift weighting to remove the intrinsic contribu-
tion from close galaxy pairs (H04) and only a small difference was
noted in the resulting signal, placing constraints on the intrinsic
alignment contamination that would also reject the elliptical model
at the 99% confidence level. Lastly, if the shear-ellipticity correla-
tions were as strongly anti-correlated as expected from an ellipti-
cal model for median-deep surveys (see Table 2) we would expect
to measure no weak lensing signal in the RCS and CTIO surveys,
which is not the case. All of the evidence therefore suggests that
not every galaxy has the same shape as its parent halo, and we can
Figure 10. 1 and 2σ constraints on the amplitude A and scale dependence
therefore rule out the 100% elliptical model.
θ0 (equation 12) of the shear-ellipticity correlation hγei , assuming the el-
liptical galaxy model (upper) and mix galaxy model (lower). For clarity, The results from the spiral model are consistent with the
only the 1σ constraints are shown for the tangential hγt et i and rotated Mandelbaum et al. (2006) SDSS constraints on intrinsic galaxy
hγr er i shear-ellipticity correlations, where hγr er i is the weakest signal. alignment. It predicts a very low intrinsic alignment correlation and
The best-fit parameters are marked with a star. low shear-ellipticity correlation that is consistent with zero on all
scales. It is important however to take note of the SDSS results
for their ‘L6’ massive elliptical galaxy sample (Mandelbaum et al.
A = (−1.29±0.16)×10−7 h Mpc−1 and θ0 = 0.93+0.38 −0.29 arcmin. 2006). This sample of massive ellipticals exhibit a strong correla-
The 1σ errors on A are marginalised over θ0 and vice versa. The tion between their galaxy shape and the surrounding density field.
constraints measured for the tangential hγt et i and rotated hγr er i This observation has been propagated to high redshift to predict a
shear-ellipticity correlations are also shown in figure 10, but for measured shear-ellipticity anti-correlation that is ∼ 100% ± 50%
clarity only the 1σ constraints are presented. Note that the hγr er i of the magnitude of the expected shear-shear correlation signal for
shear-ellipticity correlation is always weaker than the hγt et i shear- a survey with zm = 1. The amplitude of the ‘L6’ shear-ellipticity
ellipticity correlation. correlation is similar to the measured shear-ellipticity correlation
signal from our elliptical model, but is inconsistent at the 4σ level
with the measured shear-ellipticity correlation signal from our spi-
ral model. In order to reproduce the strong SDSS signal from a sub-
5 DISCUSSION
sample of galaxies in our simulations we therefore require a small
The results that have been presented in this paper and others (e.g population of elliptical galaxies and can thus rule out the 100%
H04, Hirata & Seljak (2004); Mandelbaum et al. (2006)), demon- spiral model.
strate that if a relation exists between the shape of galaxies and the We have modelled a mixed population of spiral and elliptical
properties of their parent dark matter halos then there is the poten- galaxies using the early-type/late-type halo occupation distribution
tial for contamination to the measurement of weak gravitational of Cooray (2006) which reproduces the observed luminosity func-
lensing by large-scale structure. The conclusion that one should tions and clustering properties of red and blue galaxies. The results
draw from these works is that it is no longer acceptable to assume from this mix model are consistent with the all the observational
that all the intrinsic contributions to the observed galaxy elliptic- results of Mandelbaum et al. (2006) It predicts low intrinsic align-
ity correlation function (i.e. terms that include es in equation 2) ment correlation, weak, though significant, shear-ellipticity corre-
10 Heymans et al.
lations, and negative B-modes. B-modes from systematics associ- fore recommend using the hγei model (equation 12) that we have
ated with the observations (for example residual distortions from provided as a guide to the level of contamination, which should be
the optical point-spread function) would typically produce positive self-consistently determined from the lensing data using the result
B-modes and hence one would not expect to be able to detect the that the contamination scales with lensing efficiency. As proposed
weakly negative B-mode signature of the mix model in current data. by Hirata & Seljak (2004), measurements of cross-correlation to-
The effect of the shear-ellipticity correlation is to reduce the mography hγα γβ i (see for example Takada & White (2004)) will
measured shear-shear correlation below the amplitude of the true be a useful diagnostic tool, as the true cosmological lensing sig-
cosmological lensing signal. If uncorrected for, this would result in nal scales differently with redshift with the lensing efficiency in-
an underestimate of σ8 from weak lensing measurements. The de- tegrated over all foreground lenses. One would therefore able to
gree of underestimation would be survey depth dependent as from project out the different redshift dependences of the two signals
Table 2 we expect a medium-depth survey to suffer ∼ 20% con- as demonstrated by King (2005) and simultaneously constrain the
tamination and a deep survey to suffer ∼ 10% contamination. A cosmological lensing signal and the contaminating shear ellipticity
comparison of the results from different depth surveys should there- correlations.
fore, in principle, be able to constrain models of shear-ellipticity
correlations as one would expect to measure lower values of σ8
from shallower surveys. Whilst it is tempting to discuss the trend
in the literature of somewhat higher σ8 measurements from deeper 6 CONCLUSION
lensing surveys, the uncertainties in the source redshift distribu- We have investigated potential sources of contamination to the
tion (Van Waerbeke et al. 2006), sampling variance and shear cal- measurement of weak lensing by large-scale structure from the in-
ibration errors (Heymans et al. 2006) means that we are unable to trinsic alignment of galaxies, in section 3, and from the presence of
make any firm conclusions with current lensing data to confirm or shear-ellipticity correlations between foreground galaxy shapes and
rule out the mix galaxy model. the weak lensing shear experienced by background galaxies, in sec-
tion 4. These two sources of physical contamination behave rather
differently. The intrinsic alignment contamination arises from close
5.1 Diagnostics for shear-ellipticity contamination
galaxy pairs, and can thus be eliminated by downweighting those
For the next generation of lensing surveys, the technical sys- galaxy pairs which are at close separations (Heymans & Heavens
tematics associated with the detection and measurement of weak 2003; King & Schneider 2002), or by measuring cross-correlation
lensing have a promising chance of being negligible. A signifi- tomography between different redshift bins. Intrinsic ellipticity-
cant detection of negative B-modes would then suggest contam- ellipticity correlations are thus not a problem if photometric red-
ination from shear-ellipticity correlations. In addition, the results shifts are available. In contrast, the shear-ellipticity correlation
of Mandelbaum et al. (2006) and the comparison of the different arises from galaxy pairs which are at large physical separations,
models used in this analysis demonstrate that shear-ellipticity cor- with the strongest correlations arising for lens-source pairs where
relation is likely to be galaxy morphology dependent. For future the lensing efficiency E (equation 11) is at its strongest. This could
multi-colour surveys with photometric redshifts one will be able potentially be directly eliminated through morphological galaxy se-
effectively to separate red-sequence elliptical galaxies from the lection, or from measuring weak lensing shear correlation within
galaxy sample using a simple colour cut (see for example Bell et al. thin redshift slices (King 2005). All possible techniques to remove
(2004)). With large area surveys one could then analyse the data these sources of contamination will require good photometric red-
with and without the elliptical galaxy population. A difference be- shift measurements and acquiring deep multi-colour data should be
tween the two measurements would suggest contamination from made a top priority for all future lensing surveys.
shear-ellipticity correlations. If that is the case, a zero B-mode mea- As discussed in section 5 the conclusions are dependent on the
surement from an analysis where all elliptical-type galaxies have choice of model that we use to connect galaxy shape to the proper-
been rejected might well serve as a good indication that the con- ties of its parent dark matter halo. General trends are that for deeper
tamination has been removed. surveys, where the cosmological lensing signal is at its strongest,
In section 4.1 we investigated the redshift dependence of the the degree of contamination is decreased. Eliminating the elliptical
shear-ellipticity correlation hγα eβ i where zα < zβ . The results, model through comparison to several observation results allows us
shown in figure 8 demonstrated that the evolution of the signal is conclude that for medium deep surveys zm > 0.5, intrinsic align-
proportional to the lensing efficiency E (equation 11) of the lens- ment contamination contributes only to the small angular scales
source pair, where, for the shear-ellipticity correlation hγα eβ i the θ < 1.5 arcmin, and we refer the reader to H04 for a more de-
source is at zα and the lens is at zβ . The model that provides the tailed discussion on the implications of intrinsic galaxy alignments
most realistic description appears to be the mix model for which we for future lensing surveys. The shear-ellipticity correlation is con-
have constrained the amplitude of the hγei signal to ∼ 10% accu- sistent with zero for our spiral model. To reproduce the results of
racy. We have calculated the level of contamination arising from a Mandelbaum et al. (2006) however it is necessary to include a pop-
mixed galaxy population for a deep zm ∼ 1 survey analysed using ulation of elliptical galaxies which introduce contamination to the
the shear correlation function hγγi, and found it to be at the level shear lensing signal at the ∼ 10% level for a deep zm = 1 survey
of −6 ± 1% and −10 ± 2% at angular scales 1 and 10 arcmin re- and ∼ 20% level for a medium deep zm = 0.5 survey (see table 2).
spectively. Using the hγei model (equation 12) with the best fitting The galaxy models that we have applied to the simulations
parameters to our results, one could now potentially remove this are indeed clear simplifications of reality as they consider only the
contaminating signal to an accuracy of better than 1 % of the lens- properties of the parent halo, not the halo environment or the tidal
ing signal. We would urge caution with applying such a method density field directly. The non-detection of any alignment between
however, as our results in section 3.1 clearly indicate the limita- a galaxy’s spin axis and the surface of voids in our simulations,
tions of the simple models that we have used to link the properties in section 3.1, is in some tension with the analysis of the 2dF and
of galaxies with their parent dark matter halos. We would there- SDSS surveys (Trujillo et al. 2006). If the observation is robust, it
 Potential sources of contamination to weak lensing 11
is therefore likely that it is not only the properties of the parent Brainerd T. G., 2005, ApJL, 628, L101
halo that impact on galaxy shape, but also the surrounding envi- Brown M., Taylor A., Bacon D., Gray M., Dye S., Meisenheimer
ronment to some extent. In the rare cases of galaxies near voids or K., Wolf C., 2003, MNRAS, 341, 100
clusters, we might well expect the environment to have a stronger Brown M., Taylor A., Hambly N., Dye S., 2002, MNRAS, 333,
effect on galaxy shape than its parent dark matter halo and it will 501
be important to investigate this further with full hydro-dynamical Catelan P., Kamionkowski M., Blandford R. D., 2001, MNRAS,
simulations on cosmological scales (see for example Navarro et al. 320, L7
2004). Cooray A., 2006, MNRAS, 365, 842
The important point to take away from this analysis is that our Cooray A., Milosavljević M., 2005, ApJL, 627, L89
numerical study supports the notion that the contamination from Crittenden R., Natarajan R., Pen U., Theuns T., 2001, ApJ, 559,
shear-ellipticity correlation scales with lensing efficiency. This sim- 552
ple scaling with redshift offers possibilities of determining the cos- Crittenden R., Natarajan R., Pen U., Theuns T., 2002, ApJ, 568,
mic shear signal and the contamination simultaneously, either by 20
using the models constrained in section 4.1 for the behaviour of the Croft R. A. C., Metzler C. A., 2000, ApJ, 545, 561
contamination (cf King 2005), or by making fewer assumptions and Davis M., Efstathiou G., Frenk C. S., White S. D. M., 1985, ApJ,
parameterising the contamination. Because of the redshift scaling, 292, 371
only a single function needs to be parameterised. Cosmological pa- Hamana T., Miyazaki S., Shimasaku K., Furusawa H., Doi M.,
rameter estimation would then proceed by marginalising over the Hamabe M., Imi K., Kimura M., Komiyama Y., Nakata F., Okada
nuisance parameters describing the contamination. Within the con- N., Okamura S., Ouchi M., Sekiguchi M., Yagi M., Yasuda N.,
text of the model considered here, the hγei contamination can be 2003, ApJ, 597, 98
estimated to an accuracy of 3% and 10% for the elliptical and mix Heavens A., Peacock J., 1988, MNRAS, 232, 339
models respectively. Heavens A., Refregier A., Heymans C., 2000, MNRAS, 319, 649
A more detailed study of the redshift dependence of shear- Heymans C., Bell E. F., Rix H.-W., Barden M., Borch A., Cald-
ellipticity correlation will, however, be required for the more am- well J. A. R., McIntosh D. H., Meisenheimer K., Peng C. Y.,
bitious future lensing experiments which aim to determine w(z). Wolf C., Beckwith S. V. W., Häußler B., Jahnke K., Jogee S.,
Such studies will need to include the satellite galaxy population Sanchez S. F., Somerville R. S., Wisotzki L., 2006, MNRAS Let-
that is missing from this analysis in addition to an improved mixed ters accepted
galaxy population model that takes into account an observation- Heymans C., Brown M., Heavens A., Meisenheimer K., Taylor
ally constrained redshift evolution in the ratio of elliptical to spiral A., Wolf C., 2004, MNRAS, 347, 895
galaxies for different halo masses. This will require the develop- Heymans C., Brown M. L., Barden M., Caldwell J. A. R., Jahnke
ment of better models for galaxy formation in dark matter simu- K., Rix H.-W., Taylor A. N., Beckwith S. V. W., Bell E. F., Borch
lations to improve our understanding of the connection between A., Häußler B., Jogee S., McIntosh D. H., Meisenheimer K.,
galaxies and their dark matter halos and the resulting impact on Peng C. Y., Sanchez S. F., Somerville R. S., Wisotzki L., Wolf
weak lensing studies. In the meanwhile we should rely on prob- C., 2005, MNRAS, 160
ing this phenomenon with existing data sets, in order to develop Heymans C., Heavens A., 2003, MNRAS, 339, 711
suitable diagnostics and analysis tools that will not subject weak Heymans et al. C., 2006, MNRAS, 368, 1323
lensing measurements to the kinds of contamination investigated in Hirata C. M., Seljak U., 2004, Phys. Rev. D, 70
this paper thus allowing weak lensing to reach its full potential as a Hoekstra H., Mellier Y., Waerbeke L. V., Fu E. S. L., Hudson
high precision dark matter and dark energy probe. M. J., Maoli R., Parker L., Tereno I., Benabed K., 2006, A&A,
accepted, astroph/0511089
Hoekstra H., Yee H., Gladders M., 2002, ApJ, 577, 595
Hoyle F., 1949, Problems of Cosmical Aerodynamics, p. 195
7 ACKNOWLEDGMENTS
Hu W., 1999, ApJ, 552, L21
We thank Rachel Mandelbaum and Chris Hirata for useful discus- Jarvis M., Bernstein G., Jain B., Fischer P., Smith D., Tyson J.,
sions on this project, Asantha Cooray for making code based on Wittman D., 2003, ApJ, 125, 1014
his CLF publicly available, and Ignacio Trujillo for providing us Jenkins A., Frenk C., Thomas P., Colberg J., White S., Couchman
with his results and the analytical predictions of Lee (2004). We H., Peacock J., Efstathiou G., Nelson A., 1998, ApJ, 499, 20
would also like to thank the referee for his/her helpful comments. Jing Y. P., 2002, MNRAS, 335, L89
The simulations used in this analysis were performed on the IBM- Kaiser N., Squires G., 1993, ApJ, 404, 441
SP at NERSC. CH is supported by a CITA National fellowship and, King L., Schneider P., 2002, A&A, 396, 411
along with LVW, acknowledges support from NSERC and CIAR. King L. J., 2005, A&A, 441, 47
MW was supported in part by NASA and the NSF. Lee J., 2004, ApJ, 614, L1
Lee J., Pen U., 2001, ApJ, 555, 106
Lee J., Pen U., Seljak U., 2002, ApJ, 543, L107
Mackey J., White M., Kamionkowski M., 2002, MNRAS, 332,
REFERENCES
788
Bacon D., Massey R., Refregier A., Ellis R., 2003, MNRAS, 344, Mandelbaum R., Hirata C. M., Ishak M., Seljak U., Brinkmann J.,
673 2006, MNRAS, 367, 611
Bell E. F., Wolf C., Meisenheimer K., Rix H., Borch A., Dye S., Massey R., Refregier A., Bacon D. J., Ellis R., Brown M. L., 2005,
Kleinheinrich M., Wisotzki L., McIntosh D. H., 2004, ApJ, 608, MNRAS, 359, 1277
752 Navarro J. F., Abadi M. G., Steinmetz M., 2004, ApJL, 613, L41
Bernstein G. M., Jarvis M., 2002, AJ, 123, 583 Patiri S. G., Betancort-Rijo J. E., Prada F., Klypin A., Gottlöber
12 Heymans et al.
S., 2006, MNRAS, 369, 335
Peebles J., 1969, ApJ, 155, 393
Pen U.-L., Van Waerbeke L., Mellier Y., 2002, ApJ, 567, 31
Porciani C., Dekel A., Hoffman Y., 2002, MNRAS, 332, 325
Rhodes J., Refregier A., Collins N. R., Gardner J. P., Groth E. J.,
Hill R. S., 2004, ApJ, 605, 29
Rhodes J., Refregier A., Groth E. J., 2000, ApJ, 536, 79
Semboloni E., Mellier Y., Waerbeke L. V., Hoekstra H., Tereno I.,
Benabed K., Gwyn S., Fu L., Hudson M. J., Maoli R., Parker L.,
2006, A&A, 452, 51
Spergel et al. D., 2006, ApJ, submitted, astroph/0603449
Takada M., White M., 2004, ApJL, 601, L1
Trujillo I., Carretero C., Patiri S. G., 2006, ApJL, 640, L111
Vale C., White M., 2003, ApJ, 592, 699
van den Bosch F. C., Abel T., Croft R. A. C., Hernquist L., White
S. D. M., 2002, ApJ, 576, 21
Van Waerbeke L., Mellier Y., Hoekstra H., 2005, A&A, 429, 75
Van Waerbeke L., Mellier Y., Radovich M., Bertin E., Dantel-Fort
M., McCracken H., Fèvre O. L., Foucaud S., Cuillandre J., Erben
T., Jain B., Schneider P., Bernardeau F., Fort B., 2001, A&A,
374, 757
Van Waerbeke L., White M., Hoekstra H., Heymans C., 2006, As-
troparticle Physics accepted astroph/0603696
White M., 2002, ApJS, 143, 241
Wolf C., Meisenheimer K., Rix H.-W., Borch A., Dye S., Klein-
heinrich M., 2003, A&A, 401, 73
Yang X., van den Bosch F. C., Mo H. J., Mao S., Kang X., Wein-
mann S. M., Guo Y., Jing Y. P., 2006, MNRAS, pp 528–+
Zehavi et al. I., 2004, ApJ, 608, 16