MNRAS 000, 1–13 (2024) Preprint 1 November 2024 Compiled using MNRAS LATEX style file v3.

The VST ATLAS Quasar Survey II: Halo mass profiles of galaxies,
LRGs and galaxy clusters via quasar and CMB lensing

Alice M. Eltvedt,1★ T. Shanks,1 † N. Metcalfe,1 B. Ansarinejad,2 L.F. Barrientos,3

D.N.A.
1
Murphy3,4 and D.M. Alexander1
Centre for Extragalactic Astronomy, Department of Physics, Durham University, South Road, Durham, DH1 3LE, UK
2 School of Physics, University of Melbourne, Parkville, VIC 3010, Australia
3 Instituto de Astrofisica, Facultad de Fisica, Pontificia Universidad Catolica de Chile, Santiago, Chile
arXiv:2410.23341v1 [astro-ph.CO] 30 Oct 2024

4 Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge CB3 0HA, UK

Accepted XXX. Received YYY; in original form ZZZ

ABSTRACT

We cross-correlate a low-contamination subset of the VST ATLAS 𝑔 < 22.5 quasar

catalogue with 𝑔 < 21.5 galaxy clusters, 𝑟 < 21 galaxies and 𝑟 < 19.5 Luminous Red
Galaxies (LRGs) to probe their halo mass profiles via quasar magnification bias caused by
weak lensing. In the case of galaxy clusters we find that at small scales their mass profiles are
well fitted by Navarro, Frenk and White (NFW) models with masses within the expected range.
For the galaxies, we find consistency with previous SDSS-based results for the galaxy-quasar
cross-correlation and the galaxy auto-correlation functions. Disagreement as to whether the
cross-correlation results are in tension with ΛCDM appears due to different assumptions as
to whether galaxies trace mass. We conclude that halo occupation distribution (HOD) models
fit the galaxy - quasar lensing results better than models where galaxies trace the mass. We
further test the cluster and galaxy HOD models in the 2-halo range using the Planck Cosmic
Microwave Background (CMB) lensing map, finding that the cross-correlation with both the
poorest clusters and the galaxies may be marginally over-predicted by the above HOD models.
Finally, we measure the magnification bias of LRGs using both quasar and CMB lensing and
find that the observed quasar lensing amplitude may be ≈ 2× too high and, on larger scales,
the CMB lensing amplitude may be too low to be explained by a standard LRG HOD model.
Key words: quasars: general - galaxies: general - galaxies: clusters: general - dark matter

1 INTRODUCTION The lensing magnification of background objects by large scale

structures can also provide constraints on the cosmological param-
The detection of the accelerated expansion of the universe (e.g.
eters, especially the matter density of the universe (Ω 𝑀 ) and the
Riess et al. 1998; Perlmutter et al. 1999) as well as the existence of
"clumpiness" (𝜎8 ). Galaxy-quasar cross-correlation studies have
dark matter (Zwicky 1933; Rubin et al. 1977), as necessitated by
been conducted since Seldner & Peebles (1979) detected a possible
the currently accepted Λ Cold Dark Matter (ΛCDM) model of the
quasar excess around Lick catalogue galaxies (see also Boyle et al.
universe, has made the undisputed determination and understand-
1988). More recently, works by Myers et al. (2003), Myers et al.
ing of these phenomena a main goal of modern astrophysics. The
(2005) and Mountrichas & Shanks (2007) have used background
existence of dark energy is one possibility to explain the accelerated
2QZ (Croom et al. 2005) quasars to detect the effect of galaxy and
expansion of our universe within the framework of Einstein’s theory
galaxy cluster lensing and Scranton et al. (2005) have performed
of general relativity (e.g. Ratra & Peebles 1988) and dark matter is
such lensing analyses using photo-z selected quasars from the Sloan
needed to explain observations of the clustering of structures (e.g.
Digital Sky Survey (SDSS). Myers et al. (2003) and Mountrichas &
Peebles 1980). In terms of understanding the nature of dark mat-
Shanks (2007) found a higher than expected amplitude of lensing
ter, gravitational lensing analyses are clearly of prime interest (e.g.
magnification bias based on simple Ω𝑚 = 0.3 models that assumed
Narayan & Nityananda 1985; Kaiser & Squires 1993; Kaiser 1998;
galaxies traced the mass, and suggested there may be inconsistency
Myers et al. 2003).
with the standard ΛCDM model. However, Scranton et al. (2005)
argued conversely that their SDSS results were compatible with
★ E-mail: [email protected] the standard ΛCDM model. Ménard et al. (2010) confirmed these
† E-mail: [email protected] findings on the full SDSS imaging catalogue (while also detecting

© 2024 The Authors

2 A. M. Eltvedt et al.
a sub-dominant contribution from galactic dust absorption to the
cross-correlation functions). Here, we perform a weak gravitational
lensing analysis through a cross-correlation of background quasars
and foreground galaxies and galaxy clusters using the VST ATLAS
Quasar Catalogue (see Eltvedt et al. 2023 hereafter Paper I) to pro-
vide independent new data to further address the reasons for this
apparent discrepancy.
The lensing mentioned above is defined as the gravitational
deflection of photons around large masses, which causes a mag-
nification of background sources (e.g. Narayan 1989). This "mag-
nification bias" causes the background objects to appear brighter
than they actually are while reducing the apparent solid angle be-
hind the foreground objects, causing an increase in QSO density
at bright QSO magnitudes where the slope of their number count
is steeper and a decrease at fainter magnitudes where their number
counts are flatter. Here we present our results, their interpretation
and any implications for the cosmological model. We show that an
anti-correlation is detected at faint quasar magnitudes and a positive
correlation at detected at bright magnitudes as predicted by lens-
ing. Through this cross-correlation we will be able to test (HOD)
models and their assumed mass profiles over a wide range of halo
masses. We shall further apply these quasar lensing analyses to
galaxy cluster and LRG samples.
As an independent alternative to quasar lensing we shall also
exploit CMB lensing (e.g. Blanchard & Schneider 1987; Seljak
1996) by cross-correlating our ATLAS galaxy cluster, galaxy and
Figure 1. Maps of our random catalogue in the NGC (upper panel) and SGC
LRG catalogues with the lensing maps of the CMB supplied by (lower panel), covering the same areas as our quasar candidate catalogue.
Planck Collaboration et al. (2018) to measure their halo profiles and The catalogues are split into 8 approximately equal area regions to calcu-
fit HODs as above. Here we shall follow e.g. Blanchard & Schneider late errors from field-to-field variations. We have masked out Tycho stars,
(1987); Seljak (1996) and then more recently e.g. Krolewski et al. globular clusters, nearby dwarf galaxies and areas that are underdense due
(2020, 2021). We note that the resolution of the Planck lensing map to poor observing conditions. These are left as white areas in the maps.
is ≈ 6′ , giving information extending to larger angular scales than
quasar lensing, while still allowing us to make direct comparisons
between these two at intermediate scales in the 1-2 halo regime
at ≈ 1ℎ −1 Mpc. Krolewski et al. (2020, 2021) measure LRG-CMB 2 DATA CATALOGUES
lensing by using unWISE W1 and W2 bands (Schlafly et al. 2019) to
select samples of LRGs at 𝑧 = 0.6, 1, 1, 1.4 to produce cross-power 2.1 Quasar Sample
spectra with the Planck lensing maps. Their main interest is to The VST-ATLAS quasar catalogue described in Paper I has a cer-
measure cosmological parameters and so they confine their studies tain amount of stellar and galaxy contamination, an inevitable con-
to large scales, 60′ − 900′ , whereas we complement the CMB- sequence of requiring high quasar completeness. To perform these
lensing with the small-scale QSO lensing to estimate halo mass weak lensing analyses we use a more conservative, point-source
profiles out to scales of ≈ 0.′ 3−60′ , corresponding to ≈ 0.1−10h −1 only selection of our quasar catalogue to reduce galaxy contamina-
Mpc at our LRG average redshift of 𝑧 ≈ 0.26. In Eltvedt et al (2024) tion as well as possible overlap in the galaxy and quasar catalogues.
(hereafter Paper 3) we shall also use CMB lensing to measure the We use the quasar candidate catalogue with the 𝑢𝑔𝑟𝑖 + 𝑔𝑖𝑊1𝑊2
halo mass profiles of our 𝑧 ≈ 1.7 QSOs themselves, following in cuts described in Section 4 of Paper I. We then further restrict this
particular the work of Geach et al. (2019); Han et al. (2019); Petter point-source candidate selection to 17 < 𝑔 < 22.
et al. (2022, 2023). Following an analysis of preliminary spectroscopically con-
The outline of this paper is as follows. Section 3 and 4 de- firmed QSOs, we also restrict this sample to −0.25 < (𝑔 − 𝑟) < 0.4,
scribes the cross-correlation of ATLAS selected foreground galaxy (𝑢 − 𝑔) < 0.55, (𝑟 − 𝑊1) < 5, and require (𝑊1 − 𝑊2) > 0.4, again
clusters and galaxies respectively with our quasar catalogue and the to reduce the possibility of galaxy contamination in our sample. Of
Planck CMB lensing map. We introduce HOD models in Section 5 this more conservative selection, we only consider quasar candi-
and fit these to the quasar+CMB lensing results for the galaxy clus- dates with photometric redshifts 𝑧 > 1 to prevent overlap in real
ters and the galaxies, and also the galaxy autocorrelation function. space of quasar and galaxy samples, using results from the ANNz2
To address the possibility that the 1-halo term is less well fitted to photometric redshift estimation. We also mask areas around Ty-
the galaxy cross-correlations, in Section 6 we also perform quasar cho stars to 𝑉𝑇 < 12.5 following the method of Ansarinejad et al.
and CMB lensing cross-correlations of foreground LRGs and fit (2023). Also masked are globular clusters and dwarf galaxies as
HOD models that are also tested against the LRG auto-correlation well as a few areas with poor photometry. These selections result in
function. We discuss our results in Section 7. Throughout, we as- a total of 204264 objects giving us a quasar candidate sky density
sume a standard, spatially flat, cosmology with Ω𝑚 = 0.3 and a of 44deg −2 . The QSO distribution can be seen in Fig. 2 of Paper
Hubble constant assumed to be 100 h km s −1 Mpc −1 , with h=0.7 3 and the QSO (and galaxy) masked random catalogue is shown in
unless otherwise stated. Fig. 1 (see Section 3.1).

MNRAS 000, 1–13 (2024)

 VST ATLAS QSO SURVEY II 3
We use the same Tycho stars and globular cluster mask for all of
our catalogues.

2.4 Luminous Red Galaxy Sample

We perform cross correlation with LRGs to test halo occupation
models in Section 6. To do this, we create a catalogue of LRGs based
on the "Cut 1" 𝑧 < 0.4 selection shown in Figure 3 of Eisenstein
et al. (2001), who get an LRG sky density of 14.3 deg −2 . Applying
their selections on our galaxy catalogue as described in 2.3, we get
a sample with a sky density of 9.3 deg −2 . As this is lower than
the 14.3deg −2 sky density, we adjust the selection slightly from
𝑟 𝐾 𝑅𝑂 𝑁 < 19.2 and 𝑟 𝐾 𝑅𝑂 𝑁 < 12.38 + 2.33 ∗ (𝑔 − 𝑟) + 4 ∗ (𝑟 − 𝑖) to
𝑟 𝐾 𝑅𝑂 𝑁 < 19.5 and 𝑟 𝐾 𝑅𝑂𝑁 < 12.68 + 2.33 ∗ (𝑔 − 𝑟) + 4 ∗ (𝑟 − 𝑖)
Figure 2. Sky map of defined 𝑛 ≥ 40 galaxy clusters in a section of the
to increase the density of LRGs we are getting to 16 deg −2 , which
SGC with each point corresponding to a galaxy. is as close to 14.3 deg −2 as can be achieved to 0.1mag accuracy in
the magnitude limit.

2.2 Galaxy Cluster Sample

2.5 Star Control Sample
We use the VST ATLAS Southern Galaxy Cluster Catalogue
(Ansarinejad et al. 2023) to perform the angular cross-correlations We create a subset of stars to check the signal of our cross-
between foreground galaxy clusters and background quasars. The correlations between galaxy clusters and quasars, galaxies and
galaxy groups and clusters in this catalogue were selected using quasars, and finally LRGs and quasars. We select stars away from
VST ATLAS optical photometry in the 𝑔𝑟𝑖𝑧 bands using the the 𝑊1 limit by selecting stars in the same 𝑟 − 𝑊1 range as the
ORCA cluster detection algorithm. The ORCA cluster detection QSOs (i.e. 3 < 𝑟 − 𝑊1 < 5 and 1 < 𝑔 − 𝑟 < 1.4, see Fig. 7 of
algorithm (Murphy et al. 2012) finds similarities in galaxy colours Eltvedt et al. (2023)), as we noticed that stars were being lost due
and regions with a high projected surface density and then uses to potential systematic effects, such as sky subtraction, near the
the friends-of-friends technique to determine galaxy clusters and W1 limit. We also go to the brighter limit of 𝑔 < 21, than the
groups. The selection criteria are described in full by Ansarinejad 𝑔 < 22 limit of our VST ATLAS catalogue to decrease potential
et al. (2023). This cluster catalogue overlaps the full ∼ 4700 deg2 contamination, creating a control sample that is as well positioned
area of our VST-ATLAS quasar survey to a depth of 𝑟 𝐾𝑟 𝑜𝑛 < 21. as possible to check our work.

We introduce the same Tycho stars and globular cluster mask as Previously, when we selected star control samples in the ranges
in our QSO catalogue. This galaxy cluster catalogue is then divided 1 < 𝑟 − 𝑊1 < 3 and 0.3 < 𝑔 − 𝑟 < 1 i.e. including fainter objects in
into clusters with 5 or more members (𝑛 > 5) and clusters with W1 than existed in the QSO sample, we found anomalies where
40 or more members (𝑛 > 40). The resulting 𝑛 ≥ 5 catalogue has cross-correlation of galaxies and stars showed unexpected anti-
𝑁𝑔 = 386268 galaxies, with a galaxy cluster member sky density correlation. This anti-correlation appeared to increase with galaxy
of 82.18deg −2 and a cluster sky density of 6.54deg −2 . The 𝑛 ≥ 40 apparent brightness. We also found that this anti-correlation was
catalogue has 𝑁𝑔 = 60210 galaxies, with a galaxy cluster member more evident in star samples that relied on stars selected at the
sky density of 12.81deg −2 and a cluster sky density of 0.19deg −2 . faintest W1 and W2 NEO7 magnitudes. The effect was reduced,
Fig. 2 shows a patch of sky in the SGC from our 𝑛 ≥ 40 but still not eliminated, when DECALS DR10 "forced" W1 and W2
galaxy group sample. The cross-correlations between the galaxy photometry was used instead of NEO7. We hypothesize that there
cluster and quasar catalogues are performed between quasars and may be a sky subtraction bias in W1 in the vicinity of a bright galaxy
individual members of each galaxy cluster rather than the center of where the sky brightness may be over-estimated. The effect was par-
the clusters. Therefore, the larger clusters are weighted more heavily. ticularly evident in stars selected in 𝑔𝑟𝑊1 to lie at 𝑟 − 𝑊1 < 2. This
selection is otherwise optimal in avoiding galaxy contamination (see
Paper I) but since our QSO samples reach 𝑔 ≈ 𝑟 ∼ 22 this means the
equivalent star sample reaches 𝑊1 ∼ 21 compared to a NEO7 limit
2.3 Galaxy Sample of 𝑊1 ∼ 20 so these samples suffer high incompleteness and will
To perform the cross-correlation analyses of our quasar candidate be more prone to the sky subtraction issue postulated above. When
catalogue and individual galaxies, we also generate galaxy cata- a control star sample with an 𝑟 − 𝑊1 distribution more similar to
logues from the VST ATLAS data using the same star/galaxy sepa- the QSOs was used (i.e. 1 < 𝑔 − 𝑟 < 1.4 and 3 < 𝑟 − 𝑊1 < 5), this
ration as for our QSO sample. To provide an accurate comparison to anti-correlation reduced significantly. We considered the possibility
the work done on SDSS data by Scranton et al. (2005), we require that galaxy contamination in this star sample might also contribute
the galaxies to have detections to 𝑟 𝑠𝑑𝑠𝑠 < 21, using a 0.15 mag to this reduction. However, simple 𝑔 < 22.5 star samples with no
offset to convert from ATLAS Kron 𝑟 to the total 𝑟-band SDSS colour selection also gave no evidence of anti-correlation so we
magnitudes, ie 𝑟 𝑠𝑑𝑠𝑠 = 𝑟 𝐾𝑟 𝑜𝑛 − 0.15, as suggested by previous concluded that the star-galaxy anti-correlation is only serious in star
authors (e.g. Kron 1980; Metcalfe et al. 1991; Shanks et al. 2015).1 samples too close to the 𝑊1 limit. In this case the effect on our QSO
samples will be small. But we shall show the star-galaxy correlation
results alongside the quasar-galaxy correlation results so that the
1 From this point, we shall refer to 𝑟𝑠𝑑𝑠𝑠 as 𝑟𝐾𝑟𝑜𝑛 . size of any possible systematic effect can be judged.

MNRAS 000, 1–13 (2024)

4 A. M. Eltvedt et al.
2.6 CMB Lensing Data Limber’s equation (Limber 1953), we can express the 3-D correla-
tion function (and power spectrum) as 2-D angular correlations.
We use the 2018 release of the Planck lensing convergence base-
To calculate the angular cross-correlation, we need random data
line map, using the CMB-only minimum variance estimates of the
sets with the same input parameters as our quasar + galaxy/galaxy
lensing signal to scales of 𝑙 = 4096 (Planck Collaboration et al.
cluster samples. Therefore, we generate catalogues of uniformly
2018), to perform cross-correlations with our galaxy, galaxy clus-
distributed random points covering the same area as our survey
ter, and LRG samples. Small angular scales correspond to a high
180 deg with typically > 10 times as many sources as the observable data
𝑙 value as 𝜃 ∼ 𝑙 . The Healpix 𝑎 𝑙𝑚 are first smoothed with a sets. These random catalogues are then also masked in the same
Gaussian filter with a FWHM of 15 arcmin. We then convert this manner as our data catalogues (see Fig. 1).
baseline Minimum Variance lensing map from the stored conver-
gence spherical harmonics 𝑎 𝑙𝑚 to a Healpix map (as done by Geach We use the publicly available Correlation Utilities and Two-
et al. 2019) with nside= 2048 and an 𝑙 𝑚𝑎𝑥 = 4096. This then gives point Estimates (CUTE) code (Alonso 2012) to determine the an-
us a list of RA and Dec coordinates of the Healpix pixel centers. gular cross-correlation of our samples. CUTE calculates the cross-
We apply the lensing mask provided by the Planck Collaboration correlation by using the normalized Landy-Szalay estimator for a
et al. (2018) to the CMB data and select two areas that overlap our two-point correlation function, defined as:
∼ 4700deg2 QSO sample.
𝐷 𝐺 𝐷 𝑄 − 𝐷 𝐺 𝑅𝑄 − 𝑅𝐺 𝐷 𝑄 − 𝑅𝐺 𝑅𝑄
𝜔𝐺𝑄 (𝜃) = , (1)
2.7 Possible systematic effects 𝑅𝐺 𝑅𝑄
Contamination of the QSO sample by stars or galaxies will show
different effects on our cross-correlation results. Star contamination
will dilute bright and faint cross-correlations by the fraction of We check the output generated by the Landy-Szalay estimator
stars in the QSO sample. However, the 𝑔𝑟𝑊1 cut we make is by manually checking the 𝐷 𝐺 𝐷 𝑄 , 𝐷 𝐺 𝑅𝑄 , 𝑅𝐺 𝐷 𝑄 , and 𝑅𝐺 𝑅𝑄
very efficient at removing stars at the 𝑔 < 22 magnitude range of outputs which we need to calculate the angular cross-correlation.
our QSO sample. So the main QSO contaminant is likely to be Here 𝐷 𝐺 𝐷 𝑄 denotes the number of data-point pairs drawn from the
galaxies in the same redshift range as the 𝑟 < 21 galaxy sample and galaxy sample and quasar sample with separation 𝜃. For 𝐷 𝐺 𝑅𝑄 the
this will reduce galaxy QSO anti-correlation at faint magnitudes quasar sample is replaced with the sample of randomly distributed
while increasing galaxy-QSO cross-correlation at bright QSO quasar points with the same angular selection function as the data.
magnitudes. However, the restricted version of our quasar sample Similarly, for 𝑅𝐺 𝐷 𝑄 the galaxy sample is replaced with our random
which we are using reduces this contamination (see Section 2.1). galaxy sample. The 𝑅𝐺 𝑅𝑄 output is the number of data-point pairs
We shall see that the level of agreement between the positive drawn from the two random quasar and galaxy samples.
and negative cross-correlations seen at bright and faint QSO To generate error estimates from field-field variations, we di-
magnitudes with a lensing model can be taken as confirming this vide the quasar and galaxy samples into 𝑁 𝑠 = 8 similarly sized
low level of galaxy contamination. ≈ 600𝑑𝑒𝑔 2 regions, 4 in the NGC and 4 in the SGC. These fields
are shown in Fig. 1. Then we estimate the standard errors of the
A similar argument applies to any dust obscuration associated cross-correlation by using the field-field error:
with the foreground galaxy population, since this would increase √︄ Í
the anti-correlation at faint QSO magnitudes while decreasing 𝜎𝑁 −1 (𝜔𝑖 (𝜃) − 𝜔¯𝑖 (𝜃)) 2
the positive signal at bright magnitudes, producing disagreement 𝜎𝜔¯ ( 𝜃 ) = √ 𝑠 = , (2)
𝑁𝑠 𝑁 𝑠2 − 𝑁 𝑠
with the lensing model. Ménard et al. (2011) did find evidence for
dust effects in the SDSS galaxy-QSO cross-correlations but they
were highly sub-dominant with respect to the lensing effect. We
where the sum is over 𝑖 = 1, 𝑁 𝑠 .
tested limiting our QSO sample in the W1 band and compared
the galaxy-QSO cross-correlations to those found in the 𝑔-limited
QSO samples and again found little difference between the two, 3.2 Quasar-Galaxy Cluster Lensing SIS Model
implying that lensing dominates our cross-correlation results.
The lensing of the background objects depends on the mass profiles
The other major systematic was the possible sky subtraction of the foreground objects. For galaxy clusters, we initially assume
issue in W1,W2 in the vicinity of bright galaxies. This evidenced the simplest mass profile of a singular isothermal sphere (SIS). The
itself in a strong anti-correlation between bright galaxies and stars. deflection angle of sources by such foreground lenses is given by:
However, the effect reduced when the star control sample was se-
lected to have r-W1 colours more similar to the QSOs (see Section 4𝐺 𝑀 (< 𝑏) 𝐷𝑠
2.5) and we show these galaxy-star cross-correlations alongside the 𝛼= = (𝜃 − 𝜃 𝑞 ), (3)
𝑏𝑐2 𝐷 𝑙𝑠
galaxy-QSO versions in Figs. 7 and 10, for comparison purposes.

(e.g. Myers et al. 2003) where 𝑏 is the impact parameter, 𝑀 (< 𝑏)

3 QSO - GALAXY CLUSTER LENSING is the mass contained within the radius of the lens, 𝐷 𝑠 is the
angular diameter distance from the observer to the source, 𝐷 𝑙𝑠 is
3.1 Cross-Correlation Method
the angular diameter distance from the source to the lens, 𝜃 is the
We use the data samples described in Section 2 to make a weak angle from the observer’s line of sight to the image, and 𝜃 𝑞 is the
gravitational lensing analysis via a cross-correlation of background angle from the observer’s line of sight to the source quasar.
quasars and foreground galaxies and galaxy clusters. Following

MNRAS 000, 1–13 (2024)

 VST ATLAS QSO SURVEY II 5
We see an increase in apparent brightness/magnitude of the
background object as the surface brightness of the object is con-
served, but spread across a larger surface area. Therefore the flux
received from the object is increased. The magnification, A, of the
object due to a foreground lens can be described as:

𝜃 𝑑𝜃
𝐴= (4)
𝜃 𝑞 𝑑𝜃 𝑞

On the assumption of lensing by a SIS, the mass surface density is:

𝜎2
Σ𝑆𝐼𝑆 = (5)
2𝐺𝑟

where 𝜎 is the velocity dispersion of the SIS and the density goes
𝜎 2 . This can be integrated over a radius of 𝑟 = 0 to
as 𝜌(𝑟) = 2 𝜋𝐺𝑟 2
𝑟 = 𝑏 and combined with Eq. 3 to give the amplification due to a
SIS of a background source at radius 𝜃:

𝜃
𝐴= (6)
− 4𝜋( 𝐷 𝜎 2
𝐷𝑠 ) ( 𝑐 )
𝑙𝑠
𝜃

This amplification factor can also be described as the ratio of the

lensed flux and the unlensed flux (Croom 1997). As the amplifica-
tion affects the relative distribution of background and foreground
objects, we can relate the angular cross-correlation to the amplifi-
cation factor through:

𝜔(𝜃) = 𝐴2.5𝛼−1 − 1 (7)

where 𝛼 is the slope of the cumulative source number count, Figure 3. Results of the cross-correlation of both our bright and faint quasar
𝑑𝑙𝑜𝑔(𝑁)/𝑑𝑚. Zero correlation is predicted at 𝛼 = 0.4 with an candidate catalogues in the 𝑔−band and the VST ATLAS galaxy cluster
anti-correlation at 𝛼 < 0.4, and a positive correlation at 𝛼 > 0.4. catalogue for clusters comprised of n>5 and n>40 galaxies, using the CUTE
code for angular cross-correlation. The SIS model here has a velocity dis-
In our model, we use the flat ΛCDM cosmology, with Ω 𝑀 =
persion of 270𝑘𝑚𝑠 −1 and 460𝑘𝑚𝑠 −1 and the HOD models are using a halo
0.3 and ΩΛ = 0.7. We assume an average foreground galaxy sample mass of 1014 solar masses for the the 𝑛 > 5 galaxy clusters, and a halo mass
and galaxy cluster redshift of 𝑧 = 0.15 and an average quasar sample of 1015 solar masses for the 𝑛 > 40 galaxy clusters. The positive models are
redshift of 𝑧 = 1.5. This gives us an angular diameter distance of for the bright QSO-galaxy cluster cross correlation results and the negative
the quasar sample 𝐷 𝑆 = 1780 Mpc and 𝐷 𝐿𝑆 = 1235 Mpc. We also models are for the faint QSO-gal clust cross correlation results.
use a lensing coefficient of 2.5𝛼 − 1 = −0.37 for the faint QSOs
with 20 < 𝑔 < 21 and 2.5𝛼 − 1 = 0.95 for the bright QSOs with
17 < 𝑔 < 19 taking these and other values from Table 1 of Scranton CHOMP also assumes that halo concentration is a function of halo
et al. (2005) for consistency with their assumptions. mass with the functional form 𝑐(𝑚) ≈ 9(𝑚/𝑚 ∗ ) −0.13 taken from
Bullock et al. (2001). We then compute these projected, lensed
NFW mass profiles by simply isolating the NFW part of the 1-halo
3.3 Quasar-Galaxy Cluster Lensing NFW Model term produced by CHOMP. Full details of the 1- and 2-halo terms
Similar to modelling the cluster lensing via SIS we next model the and their projection and magnification as implemented in CHOMP
clusters using an (Navarro, Frenk & White 1996) NFW model. We are given by Jain et al. (2003) and will be further summarised in
follow this route here because the HOD approach for clusters is Section 5.
less developed than for galaxies. Nevertheless, for computational
convenience we use the Cosmology and HalO Model Python code
3.4 Quasar-Galaxy Cluster Cross-Correlation Results
(CHOMP), which is a halo modelling package written by Morrison,
Scranton, and Schneider to produce the projected, lensed NFW mass We perform the cross-correlation of our 𝑛 > 5 and 𝑛 > 40 galaxy
profile which in 3-D takes the form: cluster catalogues with our 17 < 𝑔 < 19 and 20 < 𝑔 < 21 quasar
samples. We test the robustness of our detections by performing
𝜌0 the cross-correlations with star samples in the same magnitude
𝜌(𝑟) = , (8)
(𝑟/𝑟 𝑠 ) (1 + 𝑟/𝑟 𝑠 ) 2 ranges. We can see in Fig. 3 a clear anti-correlation with the faint,
20 < 𝑔 < 21, quasar samples for both the 𝑛 > 5 and 𝑛 > 40 galaxy
clusters. The cross-correlations with the faint star samples show

MNRAS 000, 1–13 (2024)

6 A. M. Eltvedt et al.

Table 1. Summary of results for galaxy cluster masses. 𝑤𝑐𝑞 denotes the
cluster-QSO cross-correlation, shown for the SIS and NFW cases, and 𝑤𝑐𝜅
is the cluster-CMB cross-correlation. A2023 cluster masses are estimated
by Ansarinejad et al. (2023)

Method n>5 Mass n>5 n>40 Mass n>40

(1013 h −1 𝑀⊙ ) 2
𝜒𝑟𝑒𝑑 (1013 h −1 𝑀⊙ ) 2
𝜒𝑟𝑒𝑑

𝑤𝑐𝑞 SIS 0.57 ± 0.12 1.8 4.7 ± 0.7 2.0

𝑤𝑐𝑞 NFW 10.0 ± 2.1 1.6 100 ± 20 0.5
𝑤𝑐𝜅 NFW 3.2 ± 0.7 3.2 32 ± 12 1.1
A2023 23 ± 8 − 43 ± 27 −

virtually zero correlation in comparison, making a strong argument

for the reality of our detected cluster-quasar cross correlation signals
at both bright and faint QSO magnitudes in Fig. 3.
We perform a 𝜒2 test for both the SIS model and the NFW
based HOD model on the cross-correlation results in order to deter-
mine which model best describes our results in Figs. 3. To do this,
we use the inverse variance weighted mean of the bright and faint
QSO cross-correlation results for both the 𝑛 > 5 and 𝑛 > 40 galaxy
cluster cases.
Figure 4. 𝑤𝑐𝜅 cluster-CMB cross-correlation functions for 𝑛 > 5 and
For the SIS model, we find that the 𝑛 > 5 galaxy cluster -
𝑛 > 40 clusters compared to 𝑤𝑐𝜅 predicted by supplying CHOMP
QSO cross-correlation has a best fit velocity dispersion of 𝜎 =
with the Zheng & Weinberg (2007) HOD parameters of log( 𝑀𝑚𝑖𝑛 ) =
270+50
−65
kms −1 with a reduced 𝜒2 of 1.8 and the 𝑛 > 40 galaxy 12.0, 12.5, 13, 13.5, 14, 14.5, 15, log( 𝑀0 ) = log( 𝑀𝑚𝑖𝑛 ), log( 𝑀1′ ) =
cluster - QSO cross-correlation has a best fit velocity dispersion of log( 𝑀𝑚𝑖𝑛 ) + 1.08, 𝜎𝑀 = 0.4 and 𝛼 = 0.7 with masses in solar mass
2
𝜎 = 460+60−80
kms −1 with a reduced 𝜒2 of 2.0. Using the 𝑀 = 2𝜎𝐺 𝑟 units assuming ℎ = 0.7. The models were integrated over the redshift range
relation appropriate for an SIS model and taking 𝑟 = 0.17h −1 Mpc 0.01 < 𝑧 < 0.36 and 𝜎8 = 0.8 was assumed throughout.
and 𝑟 = 0.48h −1 Mpc for 𝑛 > 5 and 𝑛 > 40 clusters respectively
as empirically estimated from the cluster data themselves.2 These (2007). We again employ the above CHOMP halo modelling pack-
velocity dispersions correspond to masses of 5.7 × 1012 h −1 𝑀⊙ for age and here use it more conventionally, to make 1-halo + 2-halo
𝑛 > 5 clusters and 4.7 × 1013 h −1 𝑀⊙ for 𝑛 > 40 (see Table 1). predictions, with the latter dominant. We assume the following HOD
For the NFW profiles, we similarly perform a 𝜒2 fit to the parameters log(𝑀0 ) = log(𝑀𝑚𝑖𝑛 ), log(𝑀1′ ) = log(𝑀𝑚𝑖𝑛 ) + 1.08,
𝑤 𝑐𝑞 , finding that the 𝑛 > 5 clusters are best fit by a mass of 𝜎𝑀 = 0.4 and 𝛼 = 0.7 with masses in solar mass units assuming
1014±0.09 h −1 𝑀⊙ with a reduced 𝜒2 of 1.6 and the 𝑛 > 40 cluster ℎ = 0.7. These parameters are used for values of log(𝑀𝑚𝑖𝑛 ) =
cross-correlations are best fit by 1015±0.08 h −1 𝑀⊙ with a reduced 12.0, 12.5, 13, 13.5, 14, 14.5, 15 to probe a similar range of masses
𝜒2 of 0.5. studied previously for both the SIS and NFW QSO lensing models.
We conclude that the NFW is a better fit for the galaxy cluster- We assume a flat redshift distribution between 𝑧 = 0.01 and 𝑧 = 0.36
QSO cross-correlation as the SIS generally appears to be too steep at as an approximation for our cluster samples (see Ansarinejad et al.
small scales, while the NFW is better able to fit the dampening of the (2023)) and a flat redshift distribution is also assumed for the CMB
signal at small scales. The implied NFW mass for 𝑛 > 40 clusters between 𝑧 = 1050 and 𝑧 = 1150.
also is more in agreement with the mass estimates of Ansarinejad Shown in Fig. 4, we see the result of cross-correlating the 𝑛 > 5
et al. (2023), based on various calibrations of cluster membership, and 𝑛 > 40 galaxy clusters with the CMB lensing convergence map,
that gave a mean mass of our 𝑛 > 40 galaxy clusters of 4.3 ± 2.7 × along with the various HOD results. There is a potential smoothing
1014 h −1 M ⊙ (see Table 1). at the smallest scales here due to the 6′ resolution of the Planck
CMB lensing convergence data. Therefore, the results in the bin
at the smallest scale may be more systematically uncertain than
3.5 Galaxy Cluster - CMB Lensing Map Cross-Correlation indicated by the field-field error bars.
Galaxy cluster-QSO cross-correlation mainly probes the 1-halo Performing a 𝜒2 fit of the models to the data we find that the
term, whereas cross-correlation of the Planck CMB lensing con- cross-correlation of the 𝑛 > 5 clusters with the CMB lensing conver-
vergence map with the galaxy clusters only constrains the 2-halo gence map is best fit by a HOD with log10 (𝑀𝑚𝑖𝑛 ) = 13.5+0.09−0.11
with
term due to the ≈ 6′ Planck resolution. Nevertheless, we can check a reduced 𝜒2 of 3.2, which is not a good fit. For cross-correlation
if the NFW profiles found to fit our QSO-galaxy cluster cross- of 𝑛 > 40 clusters, we get a best fit model with log10 (𝑀𝑚𝑖𝑛 ) =
correlations give halo masses consistent with the CMB lensing 14.5+0.14
−0.2
with a reduced 𝜒2 of 1.1, with the corresponding NFW
method. We model the CMB lensing by foreground galaxy clusters 1-halo term from QSO lensing giving log(𝑀ℎ ) = 15. In general, the
using the 5-parameter HOD methodology of Zheng & Weinberg cross-correlation of galaxy clusters with the Planck CMB lensing
convergence map seem to agree with the NFW model results from
QSO lensing in the previous section, although the 𝑛 > 40 fit has
2 These cluster radii were empirically estimated from the cluster member- a slightly lower 2-halo mass than the NFW fit for the 1-halo term.
ship, assuming an 8× overdensity on the sky, covering a circular area, and We see a more significant departure in the halo mass predictions of
an average redshift of 𝑧 = 0.2 as suggested by Ansarinejad et al. (2023). the SIS model with the SIS masses being ≈ 10× smaller than the

MNRAS 000, 1–13 (2024)

 VST ATLAS QSO SURVEY II 7
NFW masses, as summarised in Table 1. The average masses of the Σ(𝐷 𝑙 , 𝜃) is the surface mass density of the lens, and Σ𝑐𝑟 (𝐷 𝑙 , 𝐷 𝑠 )
𝑛 > 5 and 𝑛 > 40 clusters as estimated by Ansarinejad et al. (2023) is the critical mass surface density, defined in Myers et al. (2005)
2
are also given in Table 1. We see that for 𝑛 > 40 clusters, our NFW as Σ𝑐𝑟 (𝐷 𝑙 , 𝐷 𝑠 ) = 4 𝑐𝜋𝐺 𝐷𝐷 𝑠
.
𝑙 𝐷𝑙𝑠
lensing masses bracket the estimate of Ansarinejad et al. (2023) and
so are in good agreement. For 𝑛 > 5 clusters the QSO and CMB We can estimate the effective convergence using the relation :
lensing masses are a factor of ≈ 2× smaller than that of Ansarinejad
et al. (2023) and so the agreement is less good here. ∫ 𝑧𝑚𝑎𝑥 (1 + 𝑧) 3 𝑑𝑡 𝑑𝑧
We conclude that for the richer, 𝑛 > 40, galaxy clusters, the 3𝐻02 𝑐 𝑑𝑧
𝜅 𝑒 𝑓 𝑓 (𝜃) = Ω𝑚 (𝛿𝐺 − 1) , (10)
NFW density profile fits significantly better than the SIS profile 8𝜋𝐺 0 Σ𝑐𝑟 (𝑧, 𝑧 𝑠 )
at the small, 1-halo, scales probed by our QSO lensing results.
Generally the SIS profiles are too centrally peaked compared to the
QSO lensing data. At larger scales, the CMB lensing results for these (see Myers et al. 2005; Williams & Irwin 1998). Here, we take
richer clusters also suggest that they are well-fitted by a HOD model 𝑧 = 1.5 as the median redshift of our quasar sample and the galaxy
with a 2-halo term based on a ΛCDM cosmology. The estimated sample peaks at ∼ 0.2, so we integrate to a redshift of 𝑧 𝑚𝑎𝑥 = 0.3
average mass for these richer clusters, assuming NFW/ΛCDM 1+2- where the distribution drops to ∼ 20%. From this calculation, we
halo terms, is in the range 3 × 1014 − 1 × 1015 h −1 M ⊙ , in good find 𝜅¯ = 0.025. The quasar-galaxy cross-correlation can then be
agreement with mass estimates from Ansarinejad et al. (2023) and modelled using the 𝜔𝑔𝑔 and a Taylor expansion of Eq. 7. Therefore
other authors. we predict the galaxy-quasar cross-correlation using:
For the less rich 𝑛 > 5 groups and clusters, the QSO lensing
statistics are poorer and here both the 1-halo NFW and the SIS 2𝜅¯
models provide acceptable fits to these data. The best-fitting NFW 𝜔𝑔𝑞 (𝜃) = (2.5𝛼 − 1) 𝜔𝑔𝑔 (𝜃), (11)
𝑏
model implies a mass of ≈ 1 × 1014 h −1 M ⊙ for this 𝑛 > 5 sample, a
factor of ≈ 2× lower than the estimate of Ansarinejad et al. (2023)
but in agreement within the errors. At larger scales, the CMB lensing (𝜃)
where 𝑏𝜅¯ = ( 𝑒𝛿𝑓 𝑓 −1) . Here 𝑏 represents the linear galaxy bias 𝑏 =
𝜅
signal for this 𝑛 > 5 sample is strongly detected at a level almost as 𝐺
⟨𝛿𝐺 − 1⟩/⟨𝛿 𝑀 − 1⟩. The r.m.s. galaxy fluctuation ⟨𝛿𝐺 ⟩ will be
high as for the 𝑛 > 40 sample. However, in this case, a HOD model estimated via 𝜔𝑔𝑔, here represented by a power law fit to our galaxy
based on a ΛCDM cosmology and where the minimum halo mass sample acf which gives 𝜔𝑔𝑔 = 0.142𝜃 −0.70 in the range 𝜃 < 120′ ,
was allowed to vary in the range 1×1012 < 𝑀min < 3×1015 h −1 M ⊙ as shown in Sec. 5.
could not be found to fit the CMB lensing data when fitted over the In passing, we note the excellent agreement of the ATLAS
full 𝜃 < 300′ range. The reason for this disagreement is currently 17 < 𝑟 < 21 galaxy 𝑤 𝑔𝑔 with the equivalent SDSS 𝑤 𝑔𝑔 of Wang
unclear but will be further investigated in the work on galaxy lensing et al. (2013) also shown in Fig. 7 (a). Given this SDSS-ATLAS acf
following in Sections 4 and 5. agreement extends to 𝜃 = 8deg. or 𝑟 𝑐𝑜𝑚 ≈ 90h −1 Mpc at the average
galaxy redshift of 𝑧 ≈ 0.22, this represents a strong argument for the
accuracy of these two independent results and also for the reliability
4 QSO-GALAXY CROSS-CORRELATION of their parent datasets.
We now turn to estimating foreground galaxy halo masses via the
lensing of background QSOs and the CMB, complemented by con- 4.2 Quasar-galaxy cross-correlation results
straints from the angular autocorrelation function of the same galax-
ies. For the galaxy-QSO cross-correlations, we shall first use a model The results of cross-correlating our ATLAS QSO catalogue in var-
where galaxies trace the mass to connect with the previous studies ious magnitude ranges with our 17 < 𝑟 < 21 mag galaxy catalogue
of, e.g. Myers et al. (2003), before dropping this assumption and is shown in Fig. 5. Also shown is the HOD model from the SDSS
fitting HOD models (such as Scranton et al. 2005; Jain et al. 2003; results of Scranton et al. (2005) in blue, and the WI model de-
Zheng & Weinberg 2007 etc). scribed in the previous section is shown in red. At angular scales
of 𝜃 < 5′ , we see a negative cross-correlation between ATLAS
quasars and foreground galaxies at quasar g-band magnitudes of
4.1 Quasar-Galaxy Cross-Correlation Model 𝑔 > 20 whereas at brighter QSO limits we see a positive correla-
We first use the Williams & Irwin (1998) model, as outlined by tion. These are the same trends as seen by Scranton et al. (2005) and
Myers et al. (2005), to describe the correlation between our quasar by Myers et al. (2003, 2005) previously and they are as expected on
sample and foreground galaxies. Although Myers et al. (2005) uses the basic theoretical lensing model described in Section 4.1.
a galaxy sample to 𝑔 < 20.5, we use a galaxy sample of 𝑟 < 21 To ease model comparisons between Scranton et al. (2005)
in order to match the magnitude limit of the SDSS galaxy sample and ourselves, we use the values for ⟨𝛼𝑆05 − 1⟩=⟨2.5𝛼 − 1⟩ 3 listed
of Scranton et al. (2005). This Williams & Irwin (1998) (from here in Table 1 and Fig. 2 of Scranton et al. (2005). Then, using our
referred to as the WI model) bases predictions for 𝑤 𝑔𝑞 on the auto- 𝑤 𝑔𝑔 = 0.142𝜃 −0.70 fit, with 𝜅¯ = 0.025 and (2.5𝛼 − 1)=-0.37,
correlation, 𝑤 𝑔𝑔 , of the galaxy sample and on the assumption that we see from Fig. 6 that the best fit for the galaxy bias is 𝑏 =
galaxies trace the mass. The lensing convergence 𝜅 is defined as: 0.5+0.13
−0.09
for the ATLAS cross-correlation at 20 < 𝑔 < 21. Here,
the fit based on 9(5) points in the range 𝜃 < 30′ (4′ ), yields low
reduced 𝜒2 = 0.4(0.6). Also in this fit the covariance between the
Σ(𝐷 𝑙 , 𝜃)
𝜅= , (9)
Σ𝑐𝑟 (𝐷 𝑙 , 𝐷 𝑠 )
3 Note that 𝛼𝑆05 refers to a flux limited power law QSO number count,
𝑁 (> 𝑓 ) in the notation of Scranton et al. (2005), whereas in our notation
where 𝐷 𝑙 is again the angular diameter distance of the lens, 𝛼 refers to a magnitude limited power law number count, 𝑁 (< 𝑚).

MNRAS 000, 1–13 (2024)

8 A. M. Eltvedt et al.
𝜔𝑔𝑞 points is ignored since it is usually sub-dominant, due to the
low space density of quasars (Boyle et al. 1988). As previously
noted, these cross-correlation amplitudes are high as measured by
the simple WI model since 𝑏 = 0.5 corresponds to 𝜎8 ≈ 24 when
the usual range is 0.7 < 𝜎8 < 0.8 (e.g. Planck Collaboration et al.
2020; Heymans et al. 2021) i.e. 1.25 < 𝑏 < 1.4.5 Certainly, the
𝑏 = 1.25 (i.e. 𝜎8 ≈ 0.8) model appears to give a poor fit in Fig.
7(c), with the 5 points at 0.′ 5 < 𝜃 < 8′ giving a reduced 𝜒2 = 2.60,
rejecting the model at the 5% significance level. However, dropping
the assumption that galaxies trace mass may mean that models can
be found that are more consistent with ΛCDM.
So as previously suggested by Mountrichas et al. (2009), we
first conclude that there is little disagreement in terms of the ob-
served data between SDSS and our ATLAS results and that the main
disagreement is between these two models. We further conclude that
the Williams & Irwin (1998) assumption that galaxies trace the mass
is unlikely to be correct, given that would imply 𝑏 = 0.5 i.e. 𝜎8 = 2
in contradiction with all observed CMB power spectra. So models
that drop this assumption, like the S05 HOD model, are likely to be
required. However, the S05 HOD model may still underestimate the
lensing signal, particularly at small 𝜃 < 0.′ 5 scales. So in Section 5
we shall look for a HOD model that improves the 𝑤 𝑔𝑞 fit while also
simultaneously fitting the 𝑤 𝑔𝑔 of our 17 < 𝑟 < 21 galaxy sample.

Figure 5. Cross-correlations of our quasar candidate catalogue at 17 < 𝑔 < 5 HOD MODELS VIA QUASAR-GALAXY LENSING AND
19, 19 < 𝑔 < 19.5, 19.5 < 𝑧 < 20, 20.5 < 𝑔 < 21, and 21 < 𝑔 < 22
GALAXY-GALAXY CLUSTERING
and our VST ATLAS galaxy catalogue at 𝑟 < 21, using the CUTE code
for angular cross-correlation across the full sky. We also add the Scranton 5.1 Modelling galaxy-galaxy angular correlations
et al. (2005) HOD model for each of the quasar 𝑔−band magnitude bins.
A bias value of 𝑏 = 0.5 is consistently assumed for our WI model in red. We now make a further check of the Scranton et al. (2005) HOD
The ⟨2.5𝛼 − 1⟩ values for each QSO magnitude range for both our model model using their publicly available code from the CHOMP GitHub
and the Scranton et al. (2005) model are as follows: 0.95 for QSOs in the site written by Morrison, Scranton, and Schneider. The code follows
17 < 𝑔 < 19 range, 0.41 for 19 < 𝑔 < 19.5, 0.07 for 19.5 < 𝑔 < 20, -0.24 Jain et al. (2003) in making predictions for both the angular auto-
for 20 < 𝑔 < 20.5, and -0.5 for 20.5 < 𝑔 < 21. We also assume this -0.5 correlation function 𝑤 𝑔𝑔 and the galaxy-mass cross-correlation
value for the 21 < 𝑔 < 22 range. function 𝑤 𝑔𝜅 based on a mass power-spectrum, 𝑃(𝑘), and a HOD,
with the average number of galaxies per halo of mass 𝑀 being
denoted by < 𝑁 (𝑀) >.
First, we have assumed the simple HOD model < 𝑁 (𝑀) >=
1 + (𝑀/1012.15 ) 1.0 for 𝑀 > 1011.15 𝑀⊙ (with ℎ = 0.7) used by
Scranton et al. (2005) and we use this to predict 𝑤 𝑔𝑔 for the 17 <
𝑟 < 21 galaxy sample used here (see Fig. 7a). We note in passing
that Scranton et al. (2005) did not compare their observed and
predicted 𝑤 𝑔𝑔 . We found that this model with 𝜎8 = 0.8 over-
predicted 𝑤 𝑔𝑔 at 𝜃 < 5′ and under-predicted it at larger, 𝜃 > 5′ ,
scales. This under-prediction of the 2-halo term relative to the 1-halo
term seems a common characteristic of HOD models. Essentially,
the observed 𝑤 𝑔𝑔 seems to show a more exact power-law behaviour
than the HOD models. Mead & Verde (2021) and references therein
suggest that halo models generally underpredict the ΛCDM power-
spectrum in the region between the 1- and 2-halo terms. Indeed,
Peebles (1974, 1980) expressed doubts as to whether a preferred
(halo) scale could ever be produced by the smooth 1/𝑟 2 power-law
behaviour of Newtonian gravity.

Figure 6. Our 𝜔𝑔𝑔 = 0.142𝜃 −0.70 model fit, with 𝜅¯ = 0.025 and ⟨ 𝛼 − 1⟩=- 4
0.37, with bias values of 𝑏 = 0.2, 𝑏 = 0.4, 𝑏 = 0.5, 𝑏 = 0.8, 𝑏 = 1.0, If we assume 𝜎𝑔𝑔,8 ≈ 1 then 𝑏 = 1.25 corresponds to 𝜎8 = 𝜎𝑔𝑔,8 /𝑏 ≈
𝑏 = 1.2, and 𝑏 = 1.4 for our cross-correlation at 20 < 𝑔 < 21. 0.8 whereas 𝑏 = 0.5 corresponds to 𝜎8 ≈ 2.
5 We note that assuming Ω
𝑚 = 1 in eq 10 would also increase the cross-
correlation amplitude and imply a fitted bias value of 𝑏 ≈ 1.7. Although this
value is close to the expected 𝑏 = 2 for this cosmology, this Ω𝑚 = 1 model
is excluded by CMB + 𝐻0 constraints and so we restrict our attention here
to the standard cosmological model with Ω𝑚 = 0.3.

MNRAS 000, 1–13 (2024)

 VST ATLAS QSO SURVEY II 9

Figure 8. HOD models of Scranton et al (2005) (with log(Mmin ) =

11.15(11.0), 𝜎𝑀 = 0.01, log(M0 ) = 0.0, log(M1′ ) = 12.15(12.0),
𝛼 = 1.0) and Zheng et al (2007) 𝑀𝑟 < −20 model (with log(Mmin ) =
12.17(12.02), 𝜎𝑀 = 0.26, log(M0 ) = 11.53(11.38), log(M1′ ) =
13.46(13.31), 𝛼 = 1.06). The LRG model is from Zheng et al (2009)
(with log(Mmin ) = 14.45(14.30), 𝜎𝑀 = 0.71, log(M0 ) = 12.64(12.49),
log( 𝑀1′ ) = 15.10(14.95), 𝛼 = 1.35). All masses assume ℎ = 0.7 (ℎ = 1).

In searching for an improved HOD model, we then considered

the HOD recommended for SDSS galaxies with 𝑀𝑟 < −20 by
Zheng et al. (2007) as an alternative to the simple S05 HOD. The
parameters of this model are given in the caption of Fig. 8. This
model produces slightly improved agreement with the ATLAS 𝑤 𝑔𝑔
at both small and large scales. We also considered the range of HOD
models fitted to SDSS semi-projected correlation functions 𝑤 𝑝 (𝜎)
by Zehavi et al. (2011) (see their Fig. 10 and Table 3) corresponding
to galaxies with absolute magnitudes from 𝑀𝑟 < −18.0 to 𝑀𝑟 <
−22.0 but no better fit to our 17 < 𝑟 < 21 𝑤 𝑔𝑔 was found.
In more general searches within the 5-parameter HOD scheme
of Zheng et al. (2007), we still found it difficult to improve on the
above SDSS 𝑀𝑟 < −20 HOD as a description of the ATLAS 𝑤 𝑔𝑔 .
Given the excellent agreement of the ATLAS 𝑤 𝑔𝑔 and the SDSS
𝑤 𝑔𝑔 of Wang et al. (2013), also shown in Fig. 7a, we have no reason
to believe that this HOD fitting issue stems from the ATLAS data.
So, bearing in mind these residuals at small and large scales, we
shall consider the above two HOD models as reasonable fits and
proceed to test them further using our weak lensing analyses.6

5.2 HOD modelling from galaxy-quasar lensing

We then continue to follow the method of Jain et al. (2003) to
predict the 𝑤 𝑔𝜅 cross-correlations, first assuming the Scranton et al.
Figure 7. (a) 𝑤𝑔𝑔 and 𝑤𝑔𝜅 auto- and cross-correlation functions predicted (2005) HOD. Having multiplied the model 𝑤 𝑔𝜅 ’s in Fig. 7(a) by
by the HOD models of Scranton et al. (2005) and Zheng, Coil & Zehavi (2.5𝛼 − 1) = 0.95, −0.37 for the bright 17 < 𝑔 < 19 and faint
(2007) (with 𝑀𝑟 < −20). Both models assume ℎ = 0.7 and 𝜎8 = 0.8. The 20 < 𝑔 < 21 QSO samples, we compare the Scranton et al. (2005)
𝑤𝑔𝑔 model fitted for the 17 < 𝑟 < 21 galaxies is 𝑤𝑔𝑔 ( 𝜃 ) = 0.142𝜃 −0.70
(red, long dashes). (b) The cross-correlation function, 𝑤𝑔𝑞 ( 𝜃 ), for 17 <
𝑔 < 19 QSO candidates and 17 < 𝑟 < 21 galaxies, compared to the two 6 Fitting a -0.8 power law to our 𝑤𝑔𝑔 at 𝜃 < 60′ and then applying Limber’s
HOD models and the two models of Williams & Irwin (1998) with 𝑏 = 1.25 formula gives a 3-D correlation function scale-length of 𝑟0 = 5h −1 Mpc.
and 𝑏 = 0.75. (c) The same as (b) for the 20 < 𝑔 < 21 limited QSO case.
MNRAS 000, 1–13 (2024)
10 A. M. Eltvedt et al.
Mpc corresponding to 𝜃 ≈ 8′ at the average galaxy redshift of
𝑧 = 0.15. So CMB lensing at Planck resolution is clearly the test of
choice for the 2-halo term while the galaxy QSO cross-correlation
function in Fig. 7 (b),(c), with its scale extending down to ≈ 1′ ,
provides a better test of the 1 halo-term. Here we have seen that
both S05 and Z07 models give reasonable fits to 𝑤 𝑔𝑞 but the S05
HOD fits the CMB lensing data better than the Z07 HOD galaxy at
larger scales, despite both HOD models fitting the 𝑤 𝑔𝑔 equally well
in this range dominated by the 2-halo term. But higher signal-noise
data for QSO lensing and higher resolution data for CMB lensing
should give further interesting tests of both the 1- and 2-halo terms
of these galaxy halo occupation models independently over the full
range of scales.

6 LRG HOD MODELLING

We next attempt to model the VST ATLAS LRG sample that are
assumed to occupy the 0.16 < 𝑧 < 0.36 range with an approximately
Figure 9. 𝑤𝑔−𝐶 𝑀 𝐵 cross-correlation function for 17 < 𝑟 < 21 galaxies flat 𝑛(𝑧) (see Fig. 12 of Eisenstein et al. 2001). From Fig. 10 (a) we
and the Planck (2018) Lensing Map with field-field errors, compared to see that the LRG auto-correlation function 𝑤 𝐿𝑅𝐺−𝐿𝑅𝐺 is ≈ 10×
the HOD models of Scranton et al. (2005) and Zheng et al. (2007) (with higher than the 17 < 𝑟 < 21 galaxy 𝑤 𝑔𝑔 in Fig. 7 (a). The higher
𝑀𝑟 < −20). Both models assume ℎ = 0.7 and 𝜎8 = 0.8. The first bin
amplitude clustering of the LRGs will allow more powerful weak
centre at 6′ corresponds to ≈ 1 h −1 Mpc at the galaxy mean 𝑧 of 𝑧 = 0.15.
lensing tests of the 1- and 2-halo terms for HODs claimed to be
appropriate for LRGs. So we shall now test the LRG HOD model
and Zheng & Weinberg (2007) HOD predictions to our 𝑤 𝑔𝑞 results advocated by Zheng et al. (2009) with 𝑀𝑔 < −21.8 (see Fig. 8) and
in Figs. 7(b, c). In turn, we compare these to the 𝑤 𝑔𝑚 = 𝑤 𝑔𝑔 × 2𝜅/𝑏
¯ first compare it to our LRG 𝑤 𝑔𝑔 (𝜃) in Fig. 10 (a). While reaching
Williams & Irwin (1998) models with 𝑏 = 0.5 and 𝑏 = 1.25. The the amplitude of the observed LRG 𝑤 𝑔𝑔 (𝜃) at 𝜃 ≈ 1′ , we see that
Zheng et al. (2007) HOD model seems to give a better fit than the the HOD predicted 𝑤 𝑔𝑔 again underestimates the observations at
Scranton et al. (2005) model in Figs. 7(b, c). with both models fitting scales of ≈ 10′ , similar to what was found for the 17 < 𝑟 < 21
these data better than the standard 𝑏 = 1.25 (𝜎8 = 0.8) Williams galaxy HOD model of Zheng et al. (2007) in Fig. 7 (a). The fit also
& Irwin (1998) model. Indeed, in Figs. 7 (b),(c) we see that the appears somewhat worse at large scales than found for the SDSS
HOD model of Zheng et al. (2007) gives almost as good a fit as LRG 𝑤 𝑝 (𝜎) by Zheng et al. (2009). Nevertheless, since the HOD
the best fit, 𝑏 = 0.5, Williams & Irwin (1998) model. However, model fits 𝑤 𝑔𝑔 in the range 𝜃 < 5′ we again suggest that it is a
the errors are still large in Figs. 7(b, c) and we remain wary about useful basis to test the HOD model of Zheng et al. (2009) against
the size of the small-scale (𝜃 < 0.′ 5) anti-correlation of the stellar the simpler Williams & Irwin (1998) model using the LRG-QSO
control sample in Fig. 7(c). Another issue is that looking back at cross-correlations as considered in Section 6.1 below.
Fig. 7(a) we note that, at 𝜃 > 1′ , the predicted 𝑤 𝑔𝜅 for the two HOD
models with 𝜎8 = 0.8 lies significantly below the best fit, 𝑏 = 0.5
(or 𝜎8 = 2), Williams & Irwin (1998) model implying that both 6.1 LRG-QSO lensing
sets of models cannot fit the data equally well on these larger scales. As before for galaxies, we investigate the mass distribution around
This motivates a more detailed study of the 1-halo term using LRGs LRGs by analysing their cross-correlation with 17 < 𝑔 < 19
in Section 6 below, while a further test of the 2-halo fit of the HOD and 20 < 𝑔 < 21 ATLAS QSO samples, but based here first on
models is available from the CMB lensing test in Section 5.3 below. the Zheng et al. (2009) HOD model for 𝑤 𝑔𝜅 as shown by the
However, our main conclusion at this point is that we confirm that dashed line in Fig. 10 (a). We see that for the 20 < 𝑔 < 21 QSO
HOD models can be found that simultaneously give reasonable fits case in Fig. 10 (c), a significant anti-correlation signal is seen at
to 𝑤 𝑔𝑔 and 𝑤 𝑔𝑞 at small scales and that these fit 𝑤 𝑔𝑞 significantly 𝜃 < 5′ and particularly at 𝜃 ≈ 0.′ 3 where 𝑤 𝑔𝑞 ≈ −0.17, even
better than simpler models that assume galaxies trace the mass with taking into account that the control star sample also shows a less
bias in the standard ΛCDM 𝑏 ≈ 1.2 − 1.4 (or 𝜎8 ≈ 0.7 − 0.8) range. significant anti-correlation at 𝜃 ≈ 0.′ 3. However, a less strong
signal is seen in the 17 < 𝑔 < 19 case in Fig. 10 (b) where 𝑤 𝑔𝑞
5.3 Further Galaxy-CMB lensing test of HOD models is consistent with zero at all scales; we note that the errors are
larger here. We then checked for the presence of dust by re-doing
In Fig. 9 we show the 17 < 𝑟 < 21 galaxy - Planck CMB Lensing the cross-correlations with the QSO samples limited at bright and
Map cross-correlation function compared to the predictions of the faint W1 magnitudes. The bright cross-correlation is expected to
HOD models of Scranton et al. (2005) and Zheng et al. (2007). Here increase more than the faint cross-correlation in the case of dust
we see that the data is in reasonable agreement with the Scranton due to the steeper QSO 𝑛(𝑔). However, both the bright and faint
et al. (2005) model at all scales (with a reduced 𝜒2 of 2.67) and fits W1 cross-correlations were consistent with the g-limited results
particularly well in the range 10′ < 𝜃 < 60′ with a reduced 𝜒2 of in Figs. 10 (b, c). Inspection of the 𝑤 𝑔𝑞 results in the 8 sub-areas
1.15, whereas the Zheng et al. (2007) model appears to over-predict used for the field-field errors also showed that the anti-correlation
the data at all scales (with a reduced 𝜒2 > 10), despite its good fit to existed in almost all sub-areas.
𝑤 𝑔𝑔 at 𝜃 > 20′ . We also note that the scales probed with the Planck
map are mostly at the scales of the 2-halo term with 𝑟 𝑐𝑜𝑚 ≈ 1h −1 We compare to the model of Zheng et al. (2009) for

MNRAS 000, 1–13 (2024)

 VST ATLAS QSO SURVEY II 11

Figure 11. 𝑤𝑔−𝐶 𝑀 𝐵 cross-correlation function for our LRG sample with
0.16 < 𝑧 < 0.36 and the Planck (2018) Lensing Map with field-field
errors, compared to the HOD model of Zheng et al. (2009) with parameters
𝑙𝑜𝑔 ( 𝑀𝑚𝑖𝑛 ) = 14.45, 𝜎𝑀 = 0.80, 𝑙𝑜𝑔 ( 𝑀0 ) = 12.64, 𝑙𝑜𝑔 ( 𝑀1′ ) = 15.10,
𝛼 = 1.72. The model assumes ℎ = 0.7 and 𝜎8 = 0.8. The first bin centre at
6′ corresponds to ≈ 1 h −1 Mpc at the LRG mean 𝑧 of 𝑧 = 0.26.

𝑀𝑔 < −21.8 LRG’s (solid black line) and see that although it is
consistent with the bright QSO cross-correlation in Fig. 10 (b),
it remains above the less noisy faint QSO result at most scales
below 𝜃 ≈ 5′ in Fig. 10 (c), although the reduced 𝜒2 is still only
1.73 for these 6 points. As in Section 6 above, we have assumed
a flat 𝑛(𝑧) in the range 0.16 < 𝑧 < 0.36 for the LRGs, following
Eisenstein et al. (2001). A Williams & Irwin (1998) model with
𝑏 = 1 based on the LRG-LRG autocorrelation function in Fig.
10 (a) is also shown in Figs. 10 (b, c) assuming the same optical
depth (𝜅 = 0.025) used previously for the 𝑟 < 21 galaxy sample in
Section 4. This model assumes that the LRGs trace the mass and
this model does get closer to the 𝑤 𝑔𝑞 results than the above HOD
model. However, the low point at 𝜃 = 0.′ 3 remains over-estimated
by both. To check if it’s the form or the amplitude of the halo
mass profile that is causing the problem we show the HOD model
multiplied by a factor of 2 as the dashed line in Fig. 10 (c); the fit
improves suggesting that it may be the amplitude rather than the
form of the NFW mass profile that is at fault.

We conclude that the Zheng et al. (2009) LRG HOD that

gives a reasonable fit to the ATLAS LRG 𝑤 𝑔𝑔 at least at small,
𝜃 < 2′ scales may be rejected by 𝑤 𝑔𝑞 in the same angular range.
The problem seems to be that the effective bias produced by the
HOD appears too small and a higher amplitude mass profile may
Figure 10. (a) 𝑤𝑔𝑔 and 𝑤𝑔𝜅 auto- and cross-correlation functions predicted
by the HOD model of Zheng et al. (2009) for 𝑀𝑔 < −21.8 SDSS LRGs be needed to improve the fit. We also note that the LRG HOD also
(with 𝜎8 = 0.8 and ℎ = 0.7), compared to 𝑤𝑔𝑔 ( 𝜃 ) for our LRGs. (b) underestimates the LRG 𝑤 𝑔𝑔 at larger scales and this might only
The cross-correlation function, 𝑤𝑔𝑞 ( 𝜃 ), for 17 < 𝑔 < 19 QSO’s and our be addressed by using a higher value of 𝜎8 >≈ 1 which seems
LRG sample, compared to the HOD model of Zheng et al. (2009) and the 2 another problem for the LRG HOD approach at larger scales to put
models of Williams & Irwin (1998) with 𝑏 = 1 and 𝑏 = 0.6. (c) Same as alongside the lensing magnification problem at smaller scales.
(b) for the QSO magnitude range 20 < 𝑔 < 21.
MNRAS 000, 1–13 (2024)
12 A. M. Eltvedt et al.
6.2 Further LRG-CMB lensing test of HOD model QSO magnification studies were consistent with standard cosmol-
ogy predictions. Generally we agree with the previous conclusions
In Fig. 11 we show the 0.16 < 𝑧 < 0.36 LRG cross-correlation
of Mountrichas & Shanks (2007) that the actual observations are
with the Planck CMB lensing convergence map compared to the
very consistent with each other and that the difference lay in the
prediction of the Zheng et al. (2009) HOD model. Overall, the
models used to interpret these quasar-galaxy cross-correlations.
model fits the data well, giving reduced 𝜒2 = 1.7. Looking in more
Previously, Myers et al. (2005) assumed that galaxies traced the
detail and in the context of the fit of the Z09 model to the LRG
mass up to a linear bias factor and we have again shown on this
𝑤 𝑔𝑔 in Fig. 10(a), we see reasonable agreement between the data
assumption that values of the galaxy bias much smaller than unity
and model here for the 2-halo term at 𝜃 > 10′ although this is the
or equivalently values of 𝜎8 higher than unity are needed for such
range where the model significantly underpredicts the LRG 𝑤 𝑔𝑔 .
models to fit. If instead the HOD approach of Scranton et al. (2005)
At smaller scales where the Z09 model fits the LRG 𝑤 𝑔𝑔 very well,
is followed, then models such as the SDSS 𝑀𝑟 < −20.8 model of
the CMB lensing prediction is too low compared to the observed
Zheng & Weinberg (2007) can be found that at least approximately
result, in agreement with the LRG-QSO cross-correlation result
fit our measured galaxy angular auto-correlation function while si-
seen in Fig. 10(c). Thus the LRG HOD model either fits the LRG
multaneously reasonably fitting the QSO-galaxy cross-correlation
𝑤 𝑔𝑔 while underestimating the QSO and CMB lensing results at
function at the same scales. However, there is a hint that the Zheng
small scales or underpredicts the LRG 𝑤𝑔𝑔 while fitting the CMB
& Weinberg (2007) model that fits 𝑤 𝑔𝑔 is still slightly too low in
lensing result at large scales. This is reasonably consistent with the
lensing magnification amplitude at the smallest scales of 𝑤 𝑔𝑞 . Our
galaxy lensing results in Section 5.3, where at small scales the HOD
strong detection of the ATLAS galaxy- Planck CMB lensing signal
underestimates the galaxy-QSO cross-correlation 𝑤 𝑔𝑞 relative to
was also slightly over-predicted by the Zheng & Weinberg (2007)
the galaxy 𝑤 𝑔𝑔 whereas at large scales the Z07 HOD, at least, fits
HOD model at a similar level as the 2-halo term’s over-prediction
𝑤 𝑔𝑔 while over-predicting the galaxy CMB lensing result. However,
of the group/cluster 𝑛 > 5 sample. But both these deficiencies were
the LRG results are stronger because of their high amplitude and
only marginally detected and this motivated us to look at the lensing
signal-noise. Similar large scale behaviour may also be seen in the
results for the more extreme case of highly clustered LRGs to see if
CMB lensing results for the 𝑛 > 5 groups and clusters sample in
any such problems persisted there.
Fig. 4 and Table 1 of Section 3.5.
We therefore selected a sample of ATLAS 𝑟 < 19.5 LRG’s,
using similar criteria to the SDSS Cut 1 of Eisenstein et al. (2001)
with a 0.16 < 𝑧 < 0.36 redshift range and found an LRG auto-
7 CONCLUSIONS correlation function amplitude ≈ 10× that of the above 17 < 𝑟 < 21
We have detected lensing magnification of background quasars by galaxy sample. We found that the LRG HOD of Zheng et al. (2009)
foreground clusters, galaxies and LRGs. We have used stars as con- again fitted 𝑤 𝑔𝑔 (𝜃) well at small scales but underestimated 𝑤 𝑔𝑔
trol samples and these have suggested there may be a sky-subtraction at larger scales, similar to the galaxy HOD. We then compared the
problem for the NEO7 and DECALS DR10 W1 and W2 magnitudes LRG HOD prediction to the QSO-LRG cross-correlation function
when measured in the vicinity of bright galaxies. We have also in- and found that it under-predicted the amplitude of the LRG anti-
vestigated lensing of the CMB by these VST ATLAS cluster and correlation with 20 < 𝑔 < 21 ATLAS quasars, at a level stronger
galaxy samples and detected strong effects in each case. than the hint in the 17 < 𝑟 < 21 galaxy 𝑤 𝑔𝑞 . Multiplying the
From the lensing of ATLAS quasars by galaxy clusters in the LRG HOD prediction by a factor of 2 significantly improved the fit,
ATLAS catalogue of Ansarinejad et al. (2023) we find that NFW demonstrating the size of the effect. The 17 < 𝑔 < 19 QSO-LRG
profiles with halo masses of ≈ 1 × 1015 M ⊙ fit clusters with 𝑛 > 40 cross-correlation showed less discrepancy with the HOD prediction
members with ≈ 1 × 1014 M ⊙ fitting groups/clusters of 𝑛 > 5 mem- but here the errors are much larger.
bers. The 𝑛 > 40 clusters show the greatest signal but both cluster Overall, we conclude that our QSO-galaxy cross-correlation
samples show a preference for an NFW profile over an SIS at the results are in good agreement with previous authors for clusters
small scales probed by quasar lensing. The larger scales dominated and 17 < 𝑟 < 21 galaxies and that HOD models improve standard
by the 2-halo terms are much better investigated using CMB lens- ΛCDM cosmology fits, in particular in the 17 < 𝑟 < 21 galaxy
ing. Cross-correlation of the Planck CMB lensing convergence map case compared to models where galaxies trace the mass. In the
with the galaxy clusters showed very strong signals for both clus- case of clusters, NFW mass profiles are preferred over SIS profiles,
ter samples and we find cluster masses of ≈ 1 × 1014 M ⊙ for the with NFW mass estimates compatible with previous results for both
𝑛 > 5 clusters and ≈ 3 × 1014 M ⊙ for the 𝑛 > 40 clusters. Overall, clusters and groups. CMB lensing results for groups tended to be
the quasar and CMB lensing mass estimates are in good agreement under-predicted by standard 2-halo models and this was also seen
for both samples. However, the CMB lensing cross-correlation is marginally in the CMB lensing of the 17 < 𝑟 < 21 galaxies. LRGs
less well fitted by the 𝑛 > 5 sample than is the 𝑛 > 40 sample. show the biggest discrepancies with a standard HOD model, where
Also while the quasar and CMB lensing masses bracket the average they under-predict 𝑤 𝑔𝑞 by a factor of ≈ 2× in the fainter QSO
masses quoted for the 𝑛 > 40 clusters, the lensing mass estimates samples, while over-predicting the LRG-CMB lensing result by a
for the 𝑛 > 5 sample are generally lower than those quoted by smaller factor. Further investigation is required to see if improved
Ansarinejad et al. (2023) by a factor of ≈ 3 − 5. HOD models can be found to address these anomalies at large and
For the VST ATLAS 17 < 𝑟 < 21 galaxy sample, we find small scales in the galaxy, group and particularly LRG samples.
that galaxy-galaxy angular auto-correlation and the quasar - galaxy
cross-correlation results are consistent with those for SDSS galax-
ies by respectively Wang et al. (2013) and Scranton et al. (2005)
DATA AVAILABILITY STATEMENT
and both are at similar levels of significance. We then addressed the
question of how e.g. Myers et al. (2005) found too high a level of The ESO VST ATLAS and WISE data we have used are
QSO magnification for compatibility with standard ΛCDM cosmol- all publicly available. The VST ATLAS QSO Catalogue can
ogy compared to Scranton et al. (2005) who found that the SDSS be found at https://astro.dur.ac.uk/cea/vstatlas/qso_

MNRAS 000, 1–13 (2024)

 VST ATLAS QSO SURVEY II 13
catalogue/. All other data relevant to this publication will be Metcalfe N., Shanks T., Fong R., Jones L. R., 1991, MNRAS, 249, 498
supplied on request to the authors. Mountrichas G., Shanks T., 2007, MNRAS, 380, 113
Mountrichas G., Sawangwit U., Shanks T., Croom S. M., Schneider D. P.,
Myers A. D., Pimbblet K., 2009, MNRAS, 394, 2050
Murphy D. N. A., Geach J. E., Bower R. G., 2012, MNRAS, 420, 1861
ACKNOWLEDGEMENTS Myers A. D., Outram P. J., Shanks T., Boyle B. J., Croom S. M., Loaring
N. S., Miller L., Smith R. J., 2003, MNRAS, 342, 467
We acknowledge use of the ESO VLT Survey Telescope (VST) AT-
Myers A. D., Outram P. J., Shanks T., Boyle B. J., Croom S. M., Loaring
LAS. The ATLAS survey is based on data products from observa- N. S., Miller L., Smith R. J., 2005, MNRAS, 359, 741
tions made with ESO Telescopes at the La Silla Paranal Observatory Narayan R., 1989, ApJ, 339, L53
under program ID 177.A-3011(A,B,C,D,E.F,G,H,I,J,K,L,M,N) (see Narayan R., Nityananda R., 1985, in Kapahi V. K., ed., Extragalactic Ener-
Shanks et al. 2015). getic Sources. p. 149
We acknowledge the use of data products from WISE, which is Navarro J. F., Frenk C. S., White S. D. M., 1996, ApJ, 462, 563
a joint project of the University of California, Los Angeles, and the Peebles P. J. E., 1974, ApJ, 189, L51
Jet Propulsion Laboratory (JPL)/California Institute of Technology Peebles P. J. E., 1980, The large-scale structure of the universe. Princeton
(Caltech), funded by the National Aeronautics and Space Admin- University Press
istration (NASA), and from NEOWISE, which is a JPL/Caltech Perlmutter S., et al., 1999, ApJ, 517, 565
Petter G. C., et al., 2022, ApJ, 927, 16
project funded by NASA.
Petter G. C., Hickox R. C., Alexander D. M., Myers A. D., Geach J. E.,
We acknowledge use of SDSS imaging and spectroscopic data. Whalen K. E., Andonie C. P., 2023, ApJ, 946, 27
Funding for SDSS-III has been provided by the Alfred P. Sloan Planck Collaboration et al., 2018, arXiv e-prints, p. arXiv:1807.06210
Foundation, the Participating Institutions, the National Science Planck Collaboration et al., 2020, A&A, 641, A6
Foundation and the US Department of Energy Office of Science. Ratra B., Peebles P. J. E., 1988, Phys. Rev. D, 37, 3406
BA acknowledges support from the Australian Research Coun- Riess A. G., et al., 1998, AJ, 116, 1009
cil’s Discovery Projects scheme (DP200101068). Rubin V. C., Thonnard N., Ford W. K. J., 1977, ApJ, 217, L1
LFB acknowledges support from ANID BASAL project Schlafly E. F., Meisner A. M., Green G. M., 2019, ApJS, 240, 30
FB210003. Scranton R., et al., 2005, ApJ, 633, 589
We finally acknowledge STFC Consolidated Grant Seldner M., Peebles P. J. E., 1979, ApJ, 227, 30
Seljak U., 1996, ApJ, 463, 1
ST/T000244/1 in supporting this research.
Shanks T., et al., 2015, MNRAS, 451, 4238
For the purpose of open access, the author has applied a Cre- Wang Y., Brunner R. J., Dolence J. C., 2013, MNRAS, 432, 1961
ative Commons Attribution (CC BY) licence to any Author Ac- Williams L. L. R., Irwin M., 1998, MNRAS, 298, 378
cepted Manuscript version arising. Zehavi I., et al., 2011, ApJ, 736, 59
We thank the referee for useful comments that have improved Zheng Z., Weinberg D. H., 2007, ApJ, 659, 1
the paper. Zheng Z., Coil A. L., Zehavi I., 2007, ApJ, 667, 760
Zheng Z., Zehavi I., Eisenstein D. J., Weinberg D. H., Jing Y. P., 2009, ApJ,
707, 554
Zwicky F., 1933, Helvetica Physica Acta, 6, 110
REFERENCES
Alonso D., 2012, arXiv e-prints, p. arXiv:1210.1833 This paper has been typeset from a TEX/LATEX file prepared by the author.
Ansarinejad B., Murphy D., Shanks T., Metcalfe N., 2023, MNRAS, 520,
1371
Blanchard A., Schneider J., 1987, A&A, 184, 1
Boyle B. J., Fong R., Shanks T., 1988, MNRAS, 231, 897
Bullock J. S., Kolatt T. S., Sigad Y., Somerville R. S., Kravtsov A. V., Klypin
A. A., Primack J. R., Dekel A., 2001, MNRAS, 321, 559
Croom S. M., 1997, PhD thesis, -
Croom S. M., et al., 2005, MNRAS, 356, 415
Eisenstein D. J., et al., 2001, AJ, 122, 2267
Eltvedt A. M., et al., 2023, MNRAS, 521, 3384
Geach J. E., Peacock J. A., Myers A. D., Hickox R. C., Burchard M. C.,
Jones M. L., 2019, ApJ, 874, 85
Han J., Ferraro S., Giusarma E., Ho S., 2019, MNRAS, 485, 1720
Heymans C., et al., 2021, A&A, 646, A140
Jain B., Scranton R., Sheth R. K., 2003, MNRAS, 345, 62
Kaiser N., 1998, ApJ, 498, 26
Kaiser N., Squires G., 1993, ApJ, 404, 441
Krolewski A., Ferraro S., Schlafly E. F., White M., 2020, J. Cosmology
Astropart. Phys., 2020, 047
Krolewski A., Ferraro S., White M., 2021, J. Cosmology Astropart. Phys.,
2021, 028
Kron R. G., 1980, ApJS, 43, 305
Limber D. N., 1953, ApJ, 117, 134
Mead A. J., Verde L., 2021, MNRAS, 503, 3095
Ménard B., Scranton R., Fukugita M., Richards G., 2010, MNRAS, 405,
1025
Ménard B., Wild V., Nestor D., Quider A., Zibetti S., Rao S., Turnshek D.,
2011, MNRAS, 417, 801

MNRAS 000, 1–13 (2024)