The SDSS Discovery of a Strongly Lensed Post-Starburst Galaxy at z=0.766

Abstract

We present the first result of a survey for strong galaxy-galaxy lenses in Sloan Digital Sky Survey (SDSS) images. SDSS J082728.70+223256.4 was selected as a lensing candidate using selection criteria based on the color and positions of objects in the SDSS photometric catalog. Follow-up imaging and spectroscopy showed this object to be a lensing system. The lensing galaxy is an elliptical at z = 0.349 in a galaxy cluster. The lensed galaxy has the spectrum of a post-starburst galaxy at z = 0.766. The lensing galaxy has an estimated mass of $\sim 1.2 \times 10^{12} M_{\odot}$ and the corresponding mass to light ratio in the B-band is $\sim 26 M_{\odot}/L_{\odot}$ inside 1.1 effective radii of the lensing galaxy. Our study shows how catalogs drawn from multi-band surveys can be used to find strong galaxy-galaxy lenses having multiple lens images. Our strong lensing candidate selection based on photometry-only catalogs will be useful in future multi-band imaging surveys such as SNAP and LSST. Comment: Accepted for publication in AJ

Full text

Page 1

arXiv:0804.1487v1 [astro-ph] 9 Apr 2008
THE SDSS DISCOVERY OF A STRONGLY LENSED
POST-STARBURST GALAXY AT z=0.766
Min-Su Shin1, Michael A. Strauss1, Masamune Oguri2, Naohisa Inada3, Emilio E. Falco4,
Tom Broadhurst5, James E. Gunn1
ABSTRACT
We present the first result of a survey for strong galaxy-galaxy lenses in Sloan
Digital Sky Survey (SDSS) images. SDSS J082728.70+223256.4 was selected as
a lensing candidate using selection criteria based on the color and positions of
objects in the SDSS photometric catalog. Follow-up imaging and spectroscopy
showed this object to be a lensing system. The lensing galaxy is an elliptical at
z = 0.349 in a galaxy cluster. The lensed galaxy has the spectrum of a post-
starburst galaxy at z = 0.766. The lensing galaxy has an estimated mass of
∼ 1.2× 1012M⊙and the corresponding mass to light ratio in the B-band is ∼
26 M⊙/L⊙inside 1.1 effective radii of the lensing galaxy. Our study shows how
catalogs drawn from multi-band surveys can be used to find strong galaxy-galaxy
lenses having multiple lens images. Our strong lensing candidate selection based
on photometry-only catalogs will be useful in future multi-band imaging surveys
such as SNAP and LSST.
Subject headings: gravitational lensing — galaxies
1.INTRODUCTION
Gravitational lensing is a highly useful phenomenon in astrophysics. It allows us to
study faint galaxies that have been magnified, and can be used for a variety of cosmological
1Princeton University Observatory, Peyton Hall, Ivy Lane, Princeton, NJ 08544
2Kavli Institute for Particle Astrophysics and Cosmology, Stanford University, 2575 Sand Hill Road,
Menlo Park, CA 94025
3Cosmic Radiation Laboratory, RIKEN (The Physical and Chemical Research Organization), 2-1 Hiro-
sawa, Wako, Saitama 351-0198, Japan
4Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138
5School of Physics and Astronomy, Tel Aviv University, Tel Aviv 69978, Israel.

Page 2

– 2 –
constraints. For example, a strong lens forming multiple images of a background source
allows us to determine the mass distribution of the lensing object (e.g. Bolton et al. 2006).
Moreover, the discovery of a substantial number of lens systems makes it possible to study
the potential wells of lensing galaxies in a statistical way (e.g. Koopmans et al. 2006). The
lensed galaxy itself is magnified, allowing us to study its properties in detail (e.g. Seitz et al.
1998; Richard et al. 2006). When a background quasar is strongly lensed, this magnification
sometimes makes the host galaxy visible (Peng et al. 2006). The statistics of strong lenses
allow constraints on the cosmological parameters (e.g. Kochanek 2003; Oguri 2007), including
models of dark energy (Fedeli & Bartelmann 2007; Oguri et al. 2008).
Because of the importance of gravitational lensing systems, there have been many sys-
tematic studies of them and surveys for them. In particular, strong gravitational lensing of
quasars has been studied systematically by the CASTLES collaboration (Mu˜ noz et al. 1998),
the CLASS (Myers et al. 2001), and the Sloan Digital Sky Survey (SDSS) (Inada et al. 2007).
While various surveys of galaxy lensing have been carried out (e.g., Fassnacht et al. 2004;
Moustakas et al. 2007), many discoveries of lensed galaxies have been serendipitous, such as
the discovery of the lensed galaxy cB58 at z=2.7 (Yee et al. 1996; Seitz et al. 1998). Sys-
tematic searches for strongly lensed galaxies have been also made in the fields of massive
clusters (e.g. Pell´ o et al. 2005). For example, HST WFPC2 images were used to find giant
arcs around galaxy clusters (Sand et al. 2005). Hennawi et al. (2008) obtained deep images
of 240 rich galaxy clusters selected from SDSS, discovering 16 new giant arc systems. These
giant galaxy cluster lenses allow us to determine the mass distribution on scales of several
hundred kpc.
The large volume spectroscopic survey of the SDSS (York et al. 2000) allows system-
atic searches for galaxy-galaxy lenses. Bolton et al. (2004) and Willis et al. (2006) searched
for objects whose spectra show features of both low-redshift ellipticals and higher-redshift
emission lines. Deep HST imaging confirms that the foreground elliptical is lensing the back-
ground galaxy. This method has discovered enough strong lensing cases to allow a statistical
study of lensing galaxies (e.g. Bolton et al. 2006, 2007).
Using SDSS imaging data, Allam et al. (2007) serendipitously discovered a lensed Ly-
man break galaxy at z = 2.73 around a lensing galaxy at z = 0.38. The lensing galaxy of
this lens system has an SDSS spectrum, which does not show any features of the lensed
background galaxy because the separation between the lensing galaxy and the lensed image
is larger than the 1.′′5 radius of the spectrograph fiber. This lens therefore did not enter the
Bolton et al. (2004) sample.
These discoveries suggest an automated search for galaxy lenses in wide-field imaging
data. Several recent studies have used automatic image analysis techniques to find strong

Page 3

– 3 –
lenses in deep images (Lenzen et al. 2004; Horesh et al. 2005; Alard 2006; Seidel & Bartelmann
2006; Cabanac et al. 2007). Estrada et al. (2007) have also applied a similar method to SDSS
images to find giant arcs around clusters of galaxies. These methods use the characteristic
arc shape of lens images by calculating the ellipticity or the length-to-width ratios of objects
in the imaging data. These arc-shape detection programs can find a hidden compact ring
around a lensing galaxy, as shown by Cabanac et al. (2007).
However, strong lensing detection based on the shape of objects has several difficulties.
First, this kind of survey depends on several parameters that are tuned for an optimal
detection rate. For example, when the properties of the images such as signal-to-noise ratio
and seeing change, the parameters have to be retuned before the codes are applied to the new
images. Second, it does not work well when the images are not deep enough to show a bright
arc shape, even if the lensed galaxy is quite visible. Third, it is computationally demanding
to process massive quantities of wide field image data with shape detection codes when we
want to search all lensing possibilities around all elliptical galaxies. A simpler method to
search for strong galaxy-galaxy lensing would be valuable.
We find that a lensing search based only on a photometric catalog can be used to detect
strong galaxy-galaxy lenses in SDSS images, even though they are not deep enough to show
the full morphology of lensed arcs. All SDSS imaging data are processed by an automated
pipeline (Lupton et al. 2002). Measured parameters of detected objects including multi-band
magnitudes and correlations with neighboring objects are archived to a database from the
pipeline. Indeed, the photometric catalog has enough information to allow us to separate
the rare lenses from other objects.
In this paper we show the first results of our search for strong galaxy-galaxy lenses in
SDSS. We explain the candidate selection strategy in the following section. In §3, we present
follow-up observational data of the first confirmed lensing object, SDSS J082728.70+223256.4.
A possible lens model is given in §4. The properties of the lensing galaxy are described in
§5. Discussion and conclusions follow in §6. In this paper we focus on demonstrating the
use of a multi-band photometry catalog to find lensing systems. We are still refining our
selection algorithm, so we will quantify the completeness of our sample and give details of
our final candidate selection method in a subsequent paper.
2. CANDIDATE SELECTION
The SDSS produces a photometric catalog in five bands using a 2.5m telescope equipped
with 30 CCDs (Gunn et al. 1998, 2006) as well as a spectroscopic catalog of objects selected

Page 4

– 4 –
from the imaging data. The properties of all objects detected above 5 σ in at least one band
over the ∼ 104deg2imaged to date are measured by a dedicated pipeline (Lupton et al.
2002) and are calibrated photometrically (Tucker et al. 2006) and astrometrically (Pier et al.
2003). We use the fifth public data release (Adelman-McCarthy et al. 2007) in selecting lens
candidates. Since most possible lensing galaxies are early-type galaxies because of their
large velocity dispersions (Turner et al. 1984; Fukugita & Turner 1991), we assume that
lensing galaxies are red galaxies in the SDSS images. Because only the most luminous of
those possible lensing galaxies have redshifts above 0.2 (Strauss et al. 2002; Eisenstein et al.
2001), we do not restrict our analysis to the sample of objects with redshifts.
Our candidate selection method is still being developed, so we only outline the criteria
that were used in finding the reported lens system here. First, we identify all red galaxies in
the photometric catalog that have a model magnitude in g-band < 23.0 without correction
for Galactic extinction. Red galaxies are defined using color cuts suggested by Goto et al.
(2002):
− 0.2 < (r − i) − 0.25(g − r) − 0.18 < 0.2, (1)
−0.2 < (g − r) < 2.0,(2)
where we use model magnitudes throughout. This color cut selects intrinsically red galaxies
up to redshift 0.4. We next select those red galaxies that have at least two photometric
companions between 2.′′4 and 3.′′6 away, a typical radius of strongly lensed systems, and a
large enough separation that the image deblender separates the lens images from the red
galaxy.
In order to discriminate a lensing system from a galaxy group or cluster, we require
that the u − g colors of the possible lens images differ by more than two magnitudes from
that of the lensing galaxy. This color difference is much larger than the observed color
range of cluster ellipticals (e.g. Goto et al. 2002). This color difference cut was adopted
to test our preliminary selection method and to reduce possible contamination of neighbor
galaxies around red galaxies. While cluster ellipticals are red and tend to be undetected in
u, background lensed galaxies are often quite blue and have very different u − g colors. We
discuss refinement of this color cut in §5.
In most strong quadruple lensing systems we expect to find at least two bright knots
on one side of the lensing galaxy1. We therefore require that the angle between the two
companion objects with respect to the possible lensing galaxy be less than 90◦. Even though
the surface brightness of the arc may not be high enough to appear in the SDSS images,
1The case of double images is discussed in §5.

Page 5

– 5 –
the positions of the bright knots are precise enough to check whether the positions of the
photometric objects are consistent with a lensing system.
Finally, we eye-inspect the objects that satisfy our criteria. We found 38154 galaxies
that satisfy the color cut and have at least one companion with u−g color difference of more
than two magnitudes in the separation bin of 2.′′4 and 3.′′6. Among the galaxies found, 1644
objects have more than one companion. But only 38 objects have an angle of less than 90◦
between the two lens images; these we examine by eye.
From its appearance in the SDSS images (Figure 1), SDSS J082728.83+223253.9 (G)
was the most promising lensing galaxy in our search. The basic SDSS photometry data are
given in Table 1. The model magnitude u − g color of this elliptical galaxy is 4.48 without
Galactic extinction correction. The lensing galaxy is undetected in u, as expected for a
red elliptical. Therefore, the u − g color of the lensing galaxy should be interpreted as a
lower limit. Meanwhile, the lens images are deblended as three photometric objects: SDSS
J082728.87+223256.5 (A), SDSS J082728.70+223256.4 (B), and SDSS J082728.86+223252.6
(C). The first two objects are classified as galaxies by the SDSS software while the last one is
star-like. The distances of objects A and B from the lensing galaxy are ∼ 3.′′0, and satisfy our
selection criteria. Although image C is also close to the lensing galaxy, its model magnitude
in the g-band is fainter than 23 mag, and therefore was not selected by our criteria. The
angle between images A and B is ∼ 40◦, i.e. less than 90◦. Images A and B have u − g
= 0.36 ± 0.65 and −0.01 ± 0.08 without Galactic extinction correction, respectively. The
larger color error for A simply reflects the fact that it is 2 magnitudes fainter than B. The
difference of u − g between the lensing galaxy and the lens images is larger than the color
error itself. The fact that two objects with different u−g colors from the lensing galaxy are
on the same side of the galaxy strongly suggests that they are images of a gravitationally
lensed background galaxy.
3. FOLLOW-UP OBSERVATIONS
3.1.APO 3.5m Spectroscopy
We observed the brightest object A using the Double Imaging Spectrograph (DIS) of
the Astrophysical Research Consortium 3.5m telescope at Apache Point Observatory (APO)
on 2006 December 17 (UT). We set the two gratings to cover the wavelength ranges from
∼ 3700˚ A to ∼ 5570˚ A and from ∼ 5300˚ A to ∼ 9000˚ A, with dispersions of 1.83˚ A per pixel
and 2.31˚ A per pixel for the blue and red channels of DIS, respectively. We exposed for 2400
seconds using a 1.′′5 slit. The spectrum given in Figure 2 has signal-to-noise ratio of ∼ 7.5 per