MNRAS 000, 1–7 (2020) Preprint 6 May 2020 Compiled using MNRAS LATEX style file v3.

On building a cluster watch-list for identifying strongly

lensed supernovae, gravitational waves and kilonovae

Dan Ryczanowski,1? Graham P. Smith,1 Matteo Bianconi,1 Richard Massey,2

Andrew Robertson,2 Mathilde Jauzac 2,3,4
1 School of Physics and Astronomy, University of Birmingham, Birmingham, B15 2TT, UK
arXiv:2005.02296v1 [astro-ph.GA] 5 May 2020

2 Centre for Extragalactic Astronomy, Durham University, South Road, Durham DH1 3LE, UK
3 Institute for Computational Cosmology, Durham University, South Road, Durham DH1 3LE, UK
4 Astrophysics and Cosmology Research Unit, School of Mathematical Sciences, University of KwaZulu-Natal, Durban 4041, South Africa

Accepted XXX. Received YYY; in original form ZZZ

ABSTRACT
Motivated by discovering strongly-lensed supernovae, gravitational waves, and kilo-
novae in the 2020s, we investigate whether to build a watch-list of clusters based on
observed cluster properties (i.e. lens-plane selection) or on the detectability of strongly-
lensed background galaxies (i.e. source-plane selection). First, we estimate the fraction
of high-redshift transient progenitors that reside in galaxies that are themselves too
faint to be detected as being strongly-lensed. We find ∼ 15 − 50 per cent of transient
progenitors reside in z = 1 − 2 galaxies too faint to be detected in surveys that reach
AB ' 23, such as the Dark Energy Survey. This falls to ∼ <10 per cent at depths that
will be probed by early data releases of LSST (AB ' 25). Second, we estimate a con-
servative lower limit on the fraction of strong lensing clusters that will be missed by
magnitude limited searches for multiply-imaged galaxies and giant arcs due to the
faintness of such images. We find that DES-like surveys will miss ∼ 75 per cent of
1015 M  strong lensing clusters, rising to ∼ 100 per cent of 1014 M  clusters. Deeper
surveys, such as LSST, will miss ∼ 40 per cent at 1015 M  , and ∼ 95 per cent at 1014 M  .
Our results motivate building a cluster watch-list for strongly-lensed transients that
includes those found by lens-plane selection.
Key words: gravitational lensing: strong – gravitational waves – transients: super-
novae

1 INTRODUCTION 2018, 2019a,b; Ng et al. 2018; Li et al. 2018). There is no

clear cut evidence that any of the GW events detected thus
The discovery of Supernova Refsdal marked the dawn of ob-
far are strongly lensed (Hannuksela et al. 2019; Singer et al.
servational studies of strongly-lensed transients (Kelly et al.
2019), due in part to the challenge of detecting a strongly-
2015). A new generation of wide-field optical surveys (e.g.
lensed electromagnetic counterpart and thus localization of
ZTF, Bellm 2014, GOTO, The GOTO Collaboration 2016,
a candidate lensed GW to a lens.
LSST 1 , LSST Science Collaboration et al. 2009) are now
poised to discover large samples of strongly-lensed super- Search strategies for lensed SNe typically involve cross-
novae (SNe) in the 2020s (Goldstein et al. 2019; Goldstein matching a list of newly discovered transients with a watch-
& Nugent 2017; Oguri & Marshall 2010). The prospects for list of luminous red galaxies (e.g. Goldstein et al. 2019).
discovering strongly-lensed gravitational waves (GWs) and Concentrating on individual galaxy lenses is partly driven by
their electromagnetic counterparts has also been discussed the scientific motivation of measuring the Hubble parameter
recently, following the early detections of GWs (Smith et al. with lensed SNe, which benefits from lower systematic uncer-
tainties arising from simpler lens mass distributions (Bonvin
? E-mail: danr@star.sr.bham.ac.uk et al. 2017; Suyu et al. 2017). However, strongly lensed SNe
1 We refer to the Vera Rubin Observatory as LSST throughout and GWs/Kilonovae (KNe) are expected to be dominated
this paper, as the change of name occurred during the time of by high magnification events, i.e. those with lens magnifica-
writing. tion of | µ| > 10, until at least late 2022 when LSST survey

© 2020 The Authors

L2 D. Ryczanowski et al.
operations begin (Smith et al. 2018; Goldstein et al. 2019). watch-list would miss all such events. Secondly, we ask what
Calculations using hydrodynamical simulations were carried fraction of clusters that are capable of strongly-lensing a
out by Robertson et al. (2020) which suggest the mass distri- transient are identifiable as strong lenses by searching for
bution of high-magnification optical depths for lensed point lensed images in wide-field magnitude-limited surveys, i.e.
sources is flat over the range 1012 –1014 M  , highlighting the what fraction are identifiable by source-plane selection? We
prevalence of cluster scale lenses which have typical mass investigate these questions in Sections 2 and 3 respectively,
M200 ' 1014 M  . In addition, the cases of high-magnification before summarising our main results and discussing their
lensing of point sources detected thus far are indeed dom- implications for strongly-lensed transient detection in Sec-
inated by galaxy clusters (e.g. Sharon et al. 2005; Oguri tion 4. All magnitudes quoted are in the AB system, and we
et al. 2013; Kelly et al. 2015; Sharon et al. 2017). This mo- assume a flat cosmology consistent with the recent Planck
tivates constructing cluster-based watch-lists for use along data: H0 = 67.8 kms−1 Mpc−1 , Ω M = 0.308 (Planck Collab-
with wide-field surveys to detect strongly lensed transients. oration et al. 2016).
The search strategy utilising such a watch-list relies on the
detections of transient events from wide-field surveys such as
ZTF and LSST. Detections found near to a cluster included 2 HOSTLESS STRONGLY LENSED
in a watch-list can then be flagged as candidate strongly TRANSIENTS
lensed transients for follow-up observations, which will con-
firm if they are indeed lensed. Transient searches are a core We first consider the first question posed in the introduction,
component of such surveys, hence the focus of this paper on in which transients occur in galaxies that are strongly-lensed
how to prepare for finding transients lensed by clusters in and yet not magnified enough to be detected in wide-field
their alert streams. photometric surveys. In this situation, a watch-list of source-
The detectability of lensed SNe/KNe within the LSST plane selected lenses will not include the clusters responsible
era has been studied by several authors recently. Based on for lensing these galaxies, and the lensed transients would
simulations, Goldstein et al. (2019) predict that many hun- appear to be hostless. To quantify how common this sce-
dreds of strongly lensed supernovae will be detected by LSST nario is, we consider the fraction of the progenitors of GW
per year. Whilst these simulations were carried out with events and SNe that reside in strongly lensed galaxies that
galaxy lenses, the results of Robertson et al. (2020) indicate are too faint to detect in wide-field strong lens searches – i.e.
that clusters should lens a comparable amount. KN counter- the fraction of progenitors that will evolve into apparently
parts to lensed GWs are predicted to be detectable with ded- hostless lensed transients, fhostless . Specifically, we calculate:
icated target of opportunity observations with LSST in red ∫ Llim
optical bands, specifically the z-band, within a few nights of L φ(L) dL
GW detection. A typical source will be located at z ' 1 − 2 fhostless = 0∫ ∞ , (1)
L φ(L) dL
and magnified by | µ| ' 100 (Smith et al. 2019b,c). Tran- 0
sient point sources close to the tangential critical curves of where φ(L) is the galaxy luminosity function, which is
strong-lensing galaxy clusters have also been shown to be weighted in Equation 1 by the luminosity L, and Llim is a
recoverable close to the nominal 5σ detection limit of LSST- nominal limit below which lensed galaxies are not detectable.
like data (Smith et al. 2019a). The prospects for detecting This formulation is based on the assumption that GWs
lensed SNe and KNe in crowded cluster cores, based on wide- and SNe form from stellar remnants, and that the stellar
field survey observations and a watch-list of strong-lensing population is traced by the stellar light in galaxies. We
clusters therefore appear promising. base φ(L) on our Schechter function (Schechter 1976) fits
A key question is: how should clusters be selected in to galaxy number counts from the COSMOS i-band selected
order to construct such a watch-list? Strong-lensing clus- photometric redshift catalogue (Ilbert et al. 2009). The fits
ters can be chosen by searching for visible bright giant arcs are performed for galaxies within particular redshift bins,
and/or multiple images of distant background galaxies (e.g. which are described in Section 3.3.
Marshall et al. 2016; Lenzen et al. 2004) – we refer to this It is important to note that the integrals in Equation 1
as source-plane selection. Alternatively, they can be chosen converge for all of our derived values of the Schechter func-
based on the inferred projected mass density of cluster cores tion faint-end slope parameter, as convergence occurs when
(e.g. Wong et al. 2013; Stapelberg et al. 2019) – a lens-plane α > −2. We adopt z = 1 and z = 2 as two representative red-
selection method. In this letter, we investigate whether rely- shifts as they are typical of the redshifts at which strongly
ing solely on “traditional” source-plane selection is sufficient lensed transients have and will be detected. A single repre-
for creating a watch-list for the discovery of lensed tran- sentative value of the faint-end slope (α = −1.2) is applied to
sients. We concentrate on answering two main questions. both redshifts, rather than their individual fit values. Our
Firstly, what fraction of the progenitors of GW events/SNe results are insensitive to this choice, and is within the un-
can be found within galaxies whose apparent magnitudes are certainties from the COSMOS analysis.
fainter than the limits of optical wide-field surveys? This is Figure 1 shows fhostless as a function of the limiting i-
analogous to estimating the fraction of such transients that band magnitude of observational surveys. We relate this lim-
appear to lack a host galaxy (hereafter, “hostless” lensed iting magnitude to Llim in Equation 1, by assuming that all
transients). A hostless lensed transient (whether GW or elec- images of galaxies that are strongly-lensed by clusters are
tromagnetic radiation) may be the first detectable source magnified by | µ| = 10. The plot shows that a significant frac-
lensed by a particular lens, in which case source-plane se- tion of transient progenitors, fhostless ' 0.15 − 0.5, are housed
lection alone could not have identified this lens prior to the within z ' 1 − 2 galaxies that are fainter than the detection
transient event. Thus a search using a source-plane selected limits of current wide-field surveys, such as the Dark Energy

MNRAS 000, 1–7 (2020)

 Building a strong-lensing cluster watch-list L3
limited searches for multiply-imaged galaxies or bright arcs.
In other words, we estimate the fraction of clusters that
would require lens-plane methods to identify, assuming we
cannot increase the sensitivity of source-plane searches.
The model utilises Poisson statistics to determine the
probability of finding no detectable strongly lensed galax-
ies behind a cluster lens, based on galaxy number densities
from the COSMOS survey. The absence of any detectable
galaxies within a lens’ so-called strong lensing cross-section
implies that no observable bright arcs or multiple images
will be seen, and so would not be identifiable as a strong
lens in a magnitude-limited source-plane search. Our esti-
mates of the fraction of unidentifiable strong-lensing clus-
ters are lower limits for the same reason as discussed in Sec-
tion 2, i.e. that any inefficiency of methods used to search
for multiply-imaged galaxies will act to boost the fraction of
strong lensing clusters that are missed.
Figure 1. Lower limits on the fraction of strongly-lensed SNe
and GWs/KNe that will appear to be hostless. Such transients 3.1 Statistical model
occur in strongly-lensed galaxies that are too faint to be detected
in magnitude-limited searches for strong-lensing clusters based on We estimate f0 , the fraction of unidentifiable strong lensing
the detection of multiple images and arcs. These hostless lensed clusters of mass M200 , from the probability that the number
transients would not initially be identified as lensed, based on of strongly-lensed galaxies behind a cluster that are brighter
strong-lensing cluster watch-lists constructed solely from source- than the photometric depth of the survey, mlimit , is zero:
plane searches. The curves are plotted as a function of the depth Ö
to which strongly-lensed galaxies are identified in wide-field sur- f0 (M200, mlimit ) = P(0|M200, mlimit ) = pi (0|M200, mlimit ), (2)
vey data. Prior to the release of data from the first year of LSST i
(ilim ' 25.5 LSST Science Collaboration et al. 2009), ∼ 15 − 50 per
cent of the strongly-lensed transients are expected to be hostless, where the index i runs over a series of background redshift
i.e. in lensed z = 1 − 2 galaxies beyond the sensitivity limit of bins. The Poisson probability of having no detectable sources
precursor surveys such as DES DR1 (ilim ' 23.5 Abbott et al. in the ith redshift bin within the strong lensing cross-section
2018). of a cluster is given by:

Survey (DES, Abbott et al. 2018). This indicates that with pi (0|M200, mlimit ) = exp[−Ni (mlimit ) σi (M200 )], (3)
present surveys, a significant fraction of lensed transients where Ni is the mean number density of detectable galaxies
will be located in lensed galaxies that are not identifiable in and σi is the strong lensing cross-section of the foreground
magnitude-limited surveys, and thus if the lensed transients lens. It should be noted that σi takes different values de-
are detected they will appear to be hostless. The situation is pending on the redshift of the sources because it depends
less severe for sensitivities comparable with that of LSST af- explicitly on the geometry of the lensing system (Equations
<0.1. However, this is
ter one year of observing, with fhostless ∼ 7 & 8). The overall fraction of unidentifiable strong lensing
little comfort for efforts to detect strongly lensed transients clusters is therefore given by:
within the first year or two of LSST observations, because !
cluster watch-lists for these early years of LSST will be con- Õ
f0 = exp − Ni σi . (4)
structed from pre-LSST data. It is also important to stress
i
that our estimates of fhostless are lower limits, because they
assume that all strongly-lensed galaxies brighter than ilim Equations 3 & 4 assume that galaxies are randomly dis-
are detectable – i.e. source-plane lens detection methods are tributed on the sky, and therefore ignore galaxy cluster-
perfect. ing. Clustering concentrates some of the galaxy population
In summary, our estimates of fhostless motivate consid- into particular regions of space near to each other, mean-
eration of lens-plane search methods for finding strong lens- ing a randomly-positioned aperture of fixed size will find
ing clusters, as they suggest current watch-lists built from zero galaxies within it more frequently than using the above
purely source-plane search methods may not contain all of Poisson-based calculations. Using Equation 4 will therefore
the clusters responsible for lensing a non-negligible number result in a conservative lower limit on f0 .
of hostless transients.
3.2 Strong lensing cross-section
Our model for the strong lensing cross-section, σ, of a lens
3 STRONG-LENSING CLUSTERS IN
relies on an analytic description of the typical mass distribu-
MAGNITUDE LIMITED SURVEYS
tion in galaxy clusters on the scale of the Einstein radius –
To answer the second question posed in the introduction, we i.e. a few tens of arcseconds. The singular isothermal sphere
develop a model to estimate the fraction of strong lensing (SIS) model is commonly used to quantify the density pro-
clusters that will be unidentifiable as lenses in magnitude- file in studies that predict the lensing properties of massive

MNRAS 000, 1–7 (2020)

L4 D. Ryczanowski et al.
galaxies (e.g. Gavazzi et al. 2007). This model has a pro- model, the Einstein radius can be expressed in terms of the
jected density profile of: lens mass as:
σv2
 
2π DLS 2
Σ(R) = , (5) θE = 2 (10 G HL M200 ) 3 , (8)
2GR c DS
where G is the gravitational constant, R is the projected
where HL is the Hubble parameter at the lens redshift, and
radial distance from the centre of the lens and σv is the ve-
DLS and DS are the angular diameter distances between the
locity dispersion. However, the projected density profile of
lens and the source, and the observer and the source re-
clusters in the strong-lensing regime is typically shallower
spectively. M200 is defined as the mass within a radius r200 ,
than the R−1 isothermal profile. For example, the slope of
within which the mean density of the cluster is 200 times
the projected density profile of strong-lensing clusters in the
the critical density of the universe. We adopt a single lens
Local Cluster Substructure Survey (LoCuSS) sample span
plane at a redshift of zL = 0.25, which fits with the binning
exponents in the range −0.11 to −0.87 at the respective Ein-
procedure used in Section 3.3, and is close to the median
stein radii of the clusters for sources at zS = 2, with a median
value from the sample of known strong lensing clusters dis-
of −0.57 (Richard et al. 2010; see also Figure 3 of Umetsu
cussed by Smith et al. (2018). Following these specifications,
et al. 2016).
the strong lensing cross-section within each redshift bin can
A common alternative to the SIS model is the Navarro,
be determined as a function of halo mass.
Frenk and White (NFW, 1997) model. Whilst this model de-
The range of density profile slopes for known strong-
scribes the observed density profile of clusters well on large
lensing clusters noted above implies considerable scatter
scales (e.g. Okabe & Smith 2016), it is too shallow within
in their cross-sections. This could be due to a variety of
the Einstein radius of clusters. For example, the NFW mod-
phenomena such as substructures within cluster cores that
els described in Okabe & Smith (2016) that fit the weak-
are associated with cluster-cluster and cluster-group merg-
lensing constraints on the clusters in the Richard et al.
ers (e.g. Smith et al. 2005; Bradač et al. 2008; Limousin
(2010) strong-lensing sample have exponents in the range
et al. 2012; Jauzac et al. 2019), cluster halo triaxiality (e.g.
−0.22 to −0.41 at the respective Einstein radii of the clus-
Oguri et al. 2005; Corless et al. 2009; Sereno & Zitrin 2012;
ters for a source at zS = 2. Therefore neither the SIS nor the
Sereno et al. 2013, e.g.), or line-of-sight structure (Wambs-
NFW model alone provides a completely faithful description
ganss et al. 2005; Bayliss et al. 2014; D’Aloisio et al. 2014).
of the azimuthally averaged properties of strong-lensing clus-
We investigate this scatter further in the context of our defi-
ters clusters at their Einstein radii – the SIS model is too
nition of the lensing cross-section, in part as a cross-check on
steep and the NFW model is too shallow. We adopt the SIS
our assertion that adopting the SIS model is conservative.
model because its slightly steeper slope will boost the strong-
We consider the fourteen X-ray selected clusters in common
lensing cross section relative to reality and thus ensure that
between the Richard et al. (2010) strong-lensing analysis and
our end values of f0 (Eq. 4) are conservative. After describ-
Okabe & Smith (2016) weak-lensing analysis. We used the
ing our definition of cross section below, we will return to
models of Richard et al. to compute the total source-plane
this point at the end of this Section.
solid angle that is magnified by | µ| ≥ 10, and compared these
For simplicity, we consider a galaxy to be strongly
values with the equivalent SIS cross-sections that are calcu-
lensed if it experiences a gravitational magnification of | µ| ≥
lated using Equation 6 and M200 for each cluster from Ok-
10. We also assume that all strongly lensed galaxies suffer
abe & Smith. The distribution of the ratio of SIS to LoCuSS
the same magnification | µ| = 10 – the smallest value we as-
cross-section is approximately log-normal and is centred at
sociate with strong lensing – and ignore the effect of larger
ln(σSI S /σLoCuSS ) = 0.23, implying that the SIS model typ-
magnifications. This is justified, as the quantity of strong
ically over-estimates the lens cross-sections by 26 per cent
lensing regions with the capability to produce a magnifica-
– i.e. in qualitative agreement with our expectations. The
tion greater than some value | µ| falls off as µ−2 (Blandford &
standard deviation of the distribution implies a factor of
Narayan 1986), and so the population of strong lensing lines
∼ 3.7 scatter around the central value of 0.23, which re-
of sight is dominated by those of lower magnification. Defin-
flects the structural diversity of strong-lensing cluster cores.
ing strong lensing in this way allows us to quantify the strong
Therefore, when discussing cluster mass in later Sections, it
lensing cross-section by finding the region within which the
should be taken to mean “clusters that have cross-sections
magnification exceeds a particular value of | µ|. Following the
comparable with the typical cluster of that mass”. The scat-
equations describing the SIS model in e.g. Dodelson (2017),
ter therefore raises important questions about how to ap-
this condition is found to be satisfied by at least one image
proach the lens-plane selection of strong lensing clusters,
when the source is within the region:
and does not alter the broad conclusions of this letter. We
θE will investigate methods to lens-plane select strong lensing
β< , (6)
| µ| − 1 clusters in future work.
where β is the angular separation from the centre of the A final consideration to be made is of the finite size
lens in the source plane and θ E is the Einstein radius of the of galaxies in the source plane. The typical solid angles of
lens. Using Equation 6, the strong lensing cross-section can optically-selected galaxies from Ferguson et al. (2004) was
be found by taking advantage of the symmetry of the SIS compared to the SIS cross-section, and it was found that
model and using our | µ| = 10 definition of strong lensing: typical z = 2 galaxies subtend a solid angle comparable to
the SIS cross-section of an M200 = 5×1013 M  cluster. There-
πθ E2 fore less massive clusters do not strongly lens a region large
σ = π β2 = , (7)
81 enough to enclose an entire typical galaxy. Because of the
which depends only on the Einstein radius. Within the SIS galaxy number densities (Section 3.3, this caveat does not

MNRAS 000, 1–7 (2020)

 Building a strong-lensing cluster watch-list L5
turn out to affect the conclusions of this study. This is be-
cause the fraction of clusters that lens one or more galaxies
does not become significant until at least M200 ∼ 1014 M 
(depending on observation depth, see Section 3.4), at which
point the cross-section is much larger.

3.3 Galaxy number density

Galaxy number densities, Ni , were estimated using number
counts from the COSMOS catalogue of Ilbert et al. (2009).
Galaxies are binned so that the observable number density
can be quantified as a function of redshift. The catalogue
was split up into eleven redshift bins in the range 0.25 <
z < 5.75, each with a constant width of 0.5. This scheme was
chosen because it spans the majority of the catalogue (which
includes objects up to redshift z = 6) and allowed for the lens
Figure 2. The fraction of strong lensing clusters that are uniden-
to reside at a redshift zL = 0.25, a value consistent with the
tifiable as strong lenses in magnitude-limited surveys, as a func-
current known population of strong lensing clusters (Smith tion of halo mass, M200 , and survey depth, ilim . The coloured lines
et al. 2018). Ultimately, the final three bins centred at z = represent various magnitude depths similar to those of recent and
4.5, 5.0 and 5.5 were excluded from the analysis as they upcoming surveys. The thickness of each line corresponds to the
contained very few galaxies with well-constrained redshifts. uncertainty due to photometric redshift failures. DES-like surveys
The photometric depth of the catalogue from Ilbert will miss ∼ 75 per cent of 1015 M strong-lensing clusters, rising
et al., i = 25, is well matched to the first data release (DR1) to ∼ 100 per cent at ∼ 1014 M . Deeper surveys such as LSST will
of DES, which reaches approximately this magnitude after miss ∼ 40 and ∼ 95 per cent at these masses.
taking into account a lens magnification of | µ| = 10. However,
data from the final release of DES, and other surveys includ-
Section 3. The fractions are calculated for various survey
ing upcoming data from LSST will reach depths up to i ' 28
depths that span ongoing and upcoming surveys. The un-
after including this magnification, and hence will probe in-
certainties in each curve, represented by their width, are due
trinsically fainter galaxies beyond those of the COSMOS
to limitations of the photometric redshifts in the COSMOS
catalogue. Therefore, we fit a Schechter function (Schechter
data set, as outlined in Section 3.3. This plot shows that for
1976) to the number counts in each redshift bin, allowing
the most massive strong lensing galaxy clusters, with masses
the COSMOS number counts to be extrapolated when con-
of ∼ 1015 M  , about 75 per cent would not be identifiable as
sidering deeper surveys.
strong lenses by a survey similar to DES DR1, i.e. ilim ' 23.5
The dominant source of uncertainty in this extrapola-
(Abbott et al. 2018). In deeper surveys, such as the first year
tion is the so-called “catastrophic failure” rate of the pho-
of LSST observations (i ' 25.5), only around 40 per cent of
tometric redshifts. The catastrophic failure rate quantifies
1015 M  will not be identifiable as strong lenses. Lenses of
how often photometric redshift measurements differ signifi-
lower mass are even less likely to be identified. The model
cantly from reliable spectroscopic redshift measurements, as
predicts that even with deeper surveys, no more than 5 per
defined in Ilbert et al. (2009). If the rate is high, then many
cent of the more common 1014 M  strong lensing capable
galaxies will be placed in the wrong redshift bins, skewing
clusters will be identifiable by searching for multiply imaged
the number densities and hence affecting the curve fitting
galaxies or bright arcs. We therefore conclude that the level
and extrapolation. The failure rates provided in the cata-
of incompleteness of source-plane searches will be particu-
logue paper (Ilbert et al. 2009) vary with apparent mag-
larly severe for strong lensing clusters with typical masses
nitude, but are of order 15 to 20 per cent for the faintest
close to the knee of the halo mass function, independent of
sources (which Ilbert et al. categorise as i > 23). We quantify
whether deep future LSST survey data are available.
the overall effect of catastrophic failures on the end result by
considering a worst-case scenario based on the quoted fail-
ure rates, and in this scenario determine how many galax-
ies would appear in the wrong redshift bins. Then, correct 4 SUMMARY AND IMPLICATIONS FOR
for this and re-do the calculation based on the new “cor- STRONGLY-LENSED TRANSIENTS
rected” number counts. The true probability curve will then
Observational results (Sharon et al. 2005; Oguri et al. 2013;
lie somewhere between the original and corrected result, pro-
Kelly et al. 2015; Sharon et al. 2017; Kelly et al. 2018; Rod-
viding a range of uncertainty on the final result. Doing so
ney et al. 2018) and theoretical predictions (Hilbert et al.
provides an error on the final result of < 2 per cent.
2008; Robertson et al. 2020) all point to galaxy clusters
making a significant contribution to the high-magnification
strong lensing optical depth of point sources. This implies
3.4 Predictions of the model
that some of the strongly-lensed transients discoverable by
Figure 2 shows our estimated lower limit on the fraction of ongoing and future optical surveys (e.g. ZTF, GOTO, LSST)
strong-lensing clusters that will be unidentifiable by wide- will be strongly-lensed by galaxy clusters. So far, searches for
field surveys as a function of the lens halo mass and survey transients strongly-lensed (i.e. multiply-imaged) by clusters
magnitude depth, as calculated by the model described in have employed pointed observations of known strong-lensing

MNRAS 000, 1–7 (2020)

L6 D. Ryczanowski et al.
clusters – i.e. clusters that are known to be strong lenses be- ter’ [grant number MR/S017216/1]. This project was also
cause multiply-imaged galaxies have been detected in their supported by the Science and Technology Facilities Council
central regions (e.g. Goobar et al. 2009; Kelly et al. 2015; [grant number ST/L00075X/1]. We thank Johan Richard
Rubin et al. 2018; Smith et al. 2019b), or in other words, for sharing the lens models from Richard et al. (2010) with
clusters selected in the source plane. In this study we have us, and acknowledge helpful discussions with Rahul Biswas,
investigated whether it is beneficial to build a watch-list of Martin Freer, Ariel Goobar, Matteo Maturi, Matt Nicholl,
strong-lensing clusters for wide-field optical surveys such as and Evan Ridley.
LSST, based on only known strong lenses (selected in the
source plane) or on a more extensive (likely more complete
but less pure) list of strong-lensing capable clusters which
includes those selected in the lens plane. REFERENCES
First, in Section 2 and Figure 1, we derived a lower limit
on the fraction of lensed transients whose host galaxies are Abbott T. M. C., et al., 2018, ApJS, 239, 18
fainter than the magnitude limit of wide-field photometric Bayliss M. B., Johnson T., Gladders M. D., Sharon K., Oguri M.,
2014, ApJ, 783, 41
surveys, even after taking lens magnification into account.
Bellm E., 2014, in Wozniak P. R., Graham M. J., Mahabal A. A.,
Such lensed transient point sources would appear to be host- Seaman R., eds, The Third Hot-wiring the Transient Universe
less and would be lensed by clusters that are not identifiable Workshop. pp 27–33 (arXiv:1410.8185)
as strong lenses by a source-plane search. We predict that Blandford R., Narayan R., 1986, ApJ, 310, 568
the fraction of hostless lensed transients is fhostless ' 0.15−0.5 Bonvin V., et al., 2017, MNRAS, 465, 4914
for cluster watch-lists that are based on source-plane selec- Bradač M., et al., 2008, ApJ, 681, 187
tion in data of depth similar to DES DR1. This falls to Corless V. L., King L. J., Clowe D., 2009, MNRAS, 393, 1235
<0.1 for source-plane-based watch-lists derived from
fhostless ∼ D’Aloisio A., Natarajan P., Shapiro P. R., 2014, MNRAS, 445,
LSST year one data. Our estimates of fhostless therefore im- 3581
ply that galaxy clusters that are not identifiable as strong Dodelson S., 2017, Gravitational Lensing. Cambridge Univ. Press
Ferguson H. C., et al., 2004, ApJ, 600, L107
lenses in magnitude limited surveys should be included in
Gavazzi R., Treu T., Rhodes J. D., Koopmans L. V. E., Bolton
watch-lists so that the non-negligible number of lensed host-
A. S., Burles S., Massey R. J., Moustakas L. A., 2007, ApJ,
less transients can be identified, especially if the watch-list 667, 176
is based on relatively shallow data before the LSST survey Goldstein D. A., Nugent P. E., 2017, ApJ, 834, L5
begins. Goldstein D. A., Nugent P. E., Goobar A., 2019, ApJS, 243, 6
Second, in Section 3 and Figure 2, we predict the frac- Goobar A., et al., 2009, A&A, 507, 71
tion of strong lensing capable clusters that are unidentifiable Hannuksela O. A., Haris K., Ng K. K. Y., Kumar S., Mehta A. K.,
due to an absence of detectable multiply-imaged galaxies. Keitel D., Li T. G. F., Ajith P., 2019, ApJ, 874, L2
We predict that the fraction of unidentifiable cluster lenses Hilbert S., White S. D. M., Hartlap J., Schneider P., 2008, MN-
of mass M200 ' 1015 M  is f0 ' 0.75 at depths compara- RAS, 386, 1845
ble to DES DR1, and f0 ' 0.4 at depths comparable to the Ilbert O., et al., 2009, ApJ, 690, 1236
Jauzac M., et al., 2019, MNRAS, 483, 3082
first year of LSST observations. For more abundant 1014 M 
Kelly P. L., et al., 2015, Science, 347, 1123
clusters, we predict that even at LSST one-year depth, the
Kelly P. L., et al., 2018, Nature Astronomy, 2, 334
fraction of unidentifiable strong lensing clusters is f0 ∼ >0.95. LSST Science Collaboration et al., 2009, arXiv e-prints, p.
We emphasize that these predictions are conservative lower arXiv:0912.0201
limits as we assume no galaxy clustering and a perfectly Lenzen F., Schindler S., Scherzer O., 2004, A&A, 416, 391
efficient source-plane cluster strong-lens search algorithm. Li S.-S., Mao S., Zhao Y., Lu Y., 2018, MNRAS, 476, 2220
Taken together, our results on fhostless and f0 both motivate Limousin M., et al., 2012, A&A, 544, A71
building cluster watch-lists for strong-lensing transient dis- Marshall P. J., et al., 2016, MNRAS, 455, 1171
covery based on lens-plane selection. In future work we will Navarro J. F., Frenk C. S., White S. D. M., 1997, ApJ, 490, 493
explore methods for lens-plane selection. Ng K. K. Y., Wong K. W. K., Broadhurst T., Li T. G. F., 2018,
Phys. Rev. D, 97, 023012
Oguri M., Marshall P. J., 2010, MNRAS, 405, 2579
Oguri M., Takada M., Umetsu K., Broadhurst T., 2005, ApJ, 632,
5 ACKNOWLEDGEMENTS 841
Oguri M., et al., 2013, MNRAS, 429, 482
DR (ORCID 0000-0002-4429-3429) acknowledges a PhD stu- Okabe N., Smith G. P., 2016, MNRAS, 461, 3794
dentship from the Science and Technology Facilities Coun- Planck Collaboration et al., 2016, A&A, 594, A13
cil. GPS (ORCID 0000-0003-4494-8277) and MB (ORCID Richard J., et al., 2010, MNRAS, 404, 325
0000-0002-0427-5373) acknowledge support from the Science Robertson A., Smith G. P., Massey R., Eke V., Jauzac M.,
and Technology Facilities Council through grant number Bianconi M., Ryczanowski D., 2020, arXiv e-prints, p.
ST/S000305/1. AR (ORCID 0000-0002-0086-0524) is sup- arXiv:2002.01479
Rodney S. A., et al., 2018, Nature Astronomy, 2, 324
ported by the European Research Council (ERCStG-716532-
Rubin D., et al., 2018, ApJ, 866, 65
PUNCA) and the STFC (ST/N001494/1). RM (ORCID
Schechter P., 1976, ApJ, 203, 297
0000-0002-6085-3780) acknowledges the support of a Royal Sereno M., Zitrin A., 2012, MNRAS, 419, 3280
Society University Research Fellowship. MJ (ORCID 0000- Sereno M., Ettori S., Umetsu K., Baldi A., 2013, MNRAS, 428,
0003-1974-8732) is supported by the United Kingdom Re- 2241
search and Innovation (UKRI) Future Leaders Fellowship Sharon K., et al., 2005, ApJ, 629, L73
’Using Cosmic Beasts to uncover the Nature of Dark Mat- Sharon K., et al., 2017, ApJ, 835, 5

MNRAS 000, 1–7 (2020)

 Building a strong-lensing cluster watch-list L7
Singer L. P., Goldstein D. A., Bloom J. S., 2019, arXiv e-prints,
p. arXiv:1910.03601
Smith G. P., Kneib J.-P., Smail I., Mazzotta P., Ebeling H.,
Czoske O., 2005, MNRAS, 359, 417
Smith G. P., Jauzac M., Veitch J., Farr W. M., Massey R.,
Richard J., 2018, MNRAS, 475, 3823
Smith G. P., Robertson A., Bianconi M., Jauzac M., 2019a, arXiv
e-prints, p. arXiv:1902.05140
Smith G. P., et al., 2019b, MNRAS, 485, 5180
Smith G. P., Nicholl M., Sharon K., Bianconi M., Varricatt W. P.,
Benigni S., Ridley E. J., Gravitationally Wave Gravitational
Lensed Hunters 2019c, GRB Coordinates Network, 26605, 1
Stapelberg S., Carrasco M., Maturi M., 2019, Monthly Notices of
the Royal Astronomical Society, 482, 1824
Suyu S. H., et al., 2017, MNRAS, 468, 2590
The GOTO Collaboration 2016, “The Gravitational-wave Optical
Transient Observer”, https://goto-observatory.org
Umetsu K., Zitrin A., Gruen D., Merten J., Donahue M., Postman
M., 2016, ApJ, 821, 116
Wambsganss J., Bode P., Ostriker J. P., 2005, ApJ, 635, L1
Wong K. C., Zabludoff A. I., Ammons S. M., Keeton C. R., Hogg
D. W., Gonzalez A. H., 2013, ApJ, 769, 52

This paper has been typeset from a TEX/LATEX file prepared by

the author.

MNRAS 000, 1–7 (2020)