The Morphology of Galaxies in the Baryon Oscillation Spectroscopic Survey

Abstract

We study the morphology of luminous and massive galaxies at 0.3<z<0.7
targeted in the Baryon Oscillation Spectroscopic Survey (BOSS) using publicly
available Hubble Space Telescope imaging from COSMOS. Our sample (240 objects)
provides a unique opportunity to check the visual morphology of these galaxies
which were targeted based solely on stellar population modelling. We find that
the majority (74+/-6%) possess an early-type morphology (elliptical or S0),
while the remainder have a late-type morphology. This is as expected from the
goals of the BOSS target selection which aimed to predominantly select slowly
evolving galaxies, for use as cosmological probes, while still obtaining a fair
fraction of actively star forming galaxies for galaxy evolution studies. We
show that a colour cut of (g-i)>2.35 selects a sub-sample of BOSS galaxies with
90% early-type morphology - more comparable to the earlier Luminous Red Galaxy
(LRG) samples of SDSS-I/II. The remaining 10% of galaxies above this cut have a
late-type morphology and may be analogous to the "passive spirals" found at
lower redshift. We find that 23+/-4% of the early-type galaxies are unresolved
multiple systems in the SDSS imaging. We estimate that at least 50% of these
are real associations (not projection effects) and may represent a significant
"dry merger" fraction. We study the SDSS pipeline sizes of BOSS galaxies which
we find to be systematically larger (by 40%) than those measured from HST
images, and provide a statistical correction for the difference. These details
of the BOSS galaxies will help users of the data fine-tune their selection
criteria, dependent on their science applications. For example, the main goal
of BOSS is to measure the cosmic distance scale and expansion rate of the
Universe to percent-level precision - a point where systematic effects due to
the details of target selection may become important.

Full text

Page 1

arXiv:1106.3331v2 [astro-ph.CO] 4 Aug 2011
Mon. Not. R. Astron. Soc. 000, 000–000 (0000) Printed 5 August 2011(MN LATEX style file v2.2)
The Morphology of Galaxies in the Baryon Oscillation
Spectroscopic Survey
Karen L. Masters1,2, Claudia Maraston1, Robert C. Nichol1,2, Daniel Thomas1,2,
Alessandra Beifiori1, Kevin Bundy3, Edward M. Edmondson1, Tim D. Higgs1,
Alexie Leauthaud4,5, Rachel Mandelbaum6, Janine Pforr1, Ashley J. Ross1,
Nicholas P. Ross4, Donald P. Schneider7, Ramin Skibba8, Jeremy Tinker9,
Rita Tojeiro1, David A. Wake10, Jon Brinkmann11, Benjamin A. Weaver9
1Institute for Cosmology and Gravitation, University of Portsmouth, Portsmouth, PO1 3FX, UK
2SEPnet, South East Physics Network, (www.sepnet.ac.uk)
3Department of Astronomy, University of California, Berkeley, CA, USA
4Lawrence Berkeley National Lab, 1 Cyclotron Road, Berkeley, CA 94720, USA
5Berkeley Center for Cosmological Physics, University of California, Berkeley, CA 94720, USA
6Department of Astrophysical Sciences, Princeton University, Peyton Hall, Princeton, NJ 08544, USA
7Department of Astronomy and Astrophysics, The Pennsylvania State University, 525 Davey Lab, University Park, PA 16802, USA
8Steward Observatory, University of Arizona, 933 N. Cherry Ave. Tucson, AZ 85721, USA
9Center for Cosmology and Particle Physics, New York University, New York, NY 10003, USA
10Department of Astronomy, Yale University, New Haven, CT 06520, USA
11Apache Point Observatory, Apache Point Road, P.O. Box 59, Sunspot, NM 88349, USA
E-mail: karen.masters@port.ac.uk
5 August 2011
ABSTRACT
We study the morphology and size of the luminous and massive galaxies at 0.3 < z <
0.7 targeted in the Baryon Oscillation Spectroscopic Survey (BOSS) using publicly
available Hubble Space Telescope (HST) imaging, and catalogues, from the COSMic
Origins Survey (COSMOS). Our sample (240 objects) provides a unique opportunity
to check the visual morphology of these galaxies which were targeted based solely on
stellar population modeling. We find that the majority of BOSS galaxies (74±6%)
possess an early-type morphology (elliptical or lenticular), while the remainder have
a late-type (spiral disc) morphology. This is as expected from the goals of the BOSS
target selection which aimed to predominantly select slowly evolving galaxies, for use
as cosmological probes, while still obtaining a fair fraction of actively star forming
galaxies for galaxy evolution studies. We show that a colour cut of (g − i) > 2.35
is able to select a sub-sample of BOSS galaxies with ≥ 90% early-type morphology
and thus more comparable to the earlier Luminous Red Galaxy (LRG) samples of
SDSS-I/II. The remaining ≃ 10% of galaxies above this (g − i) cut have a late-type
morphology and may be analogous to the “passive spirals” found at lower redshift.
We find that 23±4% of the early-type BOSS galaxies are unresolved multiple systems
in the SDSS imaging. We estimate that at least 50% of these multiples are likely real
associations and not projection effects and may represent a significant “dry merger”
fraction. We study the SDSS pipeline sizes of BOSS galaxies which we find to be
systematically larger (by 40%) than those measured from HST images, and provide a
statistical correction for the difference. These details of the BOSS galaxies will help
users of the BOSS data fine-tune their selection criteria, dependent on their science
applications. For example, the main goal of BOSS is to measure the cosmic distance
scale and expansion rate of the Universe to percent-level precision – a point where
systematic effects due to the details of target selection may become important.
Key words: galaxies: ellipticals - galaxies: spirals - galaxies: morphology - galaxies:
photometry - surveys
c ? 0000 RAS

Page 2

2
K.L. Masters et al.
1 INTRODUCTION
The Baryon Oscillation Spectroscopic Survey (BOSS) is one
of four surveys being carried out as part of the Sloan Digital
Sky Survey III (SDSS-III; Eisenstein et al. 2011). The main
goal of BOSS is to measure the cosmic distance scale and
expansion rate of the Universe with percent-level precision
at z < 0.7 and z ∼ 2.5 using the Baryon Acoustic Oscilla-
tion (BAO) scale. To achieve this goal, BOSS is performing
a redshift survey of 1.5 million massive luminous galaxies
(between 0.3 < z < 0.7) and 150,000 quasars at z > 2.5
selected from the SDSS imaging data published as part of
Data Release Eight (DR8; Aihara et al. 2011).
Previous cosmological surveys of distant luminous
galaxies have focused on the reddest galaxies because they
are efficient tracers of the underlying dark matter distri-
bution, e.g. Luminous Red Galaxies (LRGs; Eisenstein et
al. 2001, Cannon et al. 2006). However, BOSS is now in-
cluding bluer galaxies than these previous samples partly
to increase the sky density of sources observed (to improve
the BAO measurements), as well as to sample intrinsically
bluer galaxies to provide a more representative census of
the galaxy population for comparison with models of galaxy
evolution (e.g. Kauffmann et al. 1993).
With the goal of percent-level precision on the cosmo-
logical parameters, BOSS is entering the regime where the
details of galaxy evolution, and sample selection, play an
important role in understanding the systematic errors as-
sociated with accurate measurements of the clustering of
galaxies on large scales. Different types of galaxies populate
their dark matter halos differently (e.g. Hogg et al 2003,
Zehavi et al. 2005, Ross & Brunner 2009), and especially
evolve on different timescales so a detailed understanding of
the galaxy types in the BOSS sample is important both for
cosmology and studies of galaxy evolution.
Galaxies are selected for inclusion in BOSS (and many
other modern surveys) using stellar population models
which predict the expected colours as a function of the
galaxy star formation history, dust content and redshift.
This method is effective as the stellar evolutionary status
of a galaxy imprints onto its spectral energy distribution,
hence colours, with the well-known effect that galaxies dom-
inated by old stars, or more generally having a small scatter
in age - are redder than galaxies hosting star formation.
To a large extent this distinction is mirrored in the galaxy
morphology, with older and redder galaxies typically having
early-type morphologies, while star-forming galaxies usually
possess discs. However, since a galaxy’s morphology is also
affected by gas accretion and merging histories which effect
the dynamics of the stars, the colour and morphology of a
galaxy can become decoupled.
In recent years, there has been a resurgence of interest
in the morphology of galaxies, especially at low redshift. For
example, Schawinski et al (2007) and Thomas et al (2010)
constructed a sample of 50,000 visually–inspected early-type
galaxies from SDSS, and showed that morphology was an
important factor in the interpretation of the transition re-
gion between the “blue cloud” (late–type galaxies) and the
“red sequence” (early–type galaxies). These studies discov-
ered and discussed so-called “blue ellipticals”, i.e. galaxies
with an elliptical morphology and blue colours (also con-
firmed by Schawinski et al. 2009; Kannappan et al. 2009).
The Galaxy Zoo project (Lintott et al. 2008), with the
help of citizen scientists have now provided visual classi-
fications for all galaxies in the original (z < 0.3) SDSS-
I/II data (Lintott et al. 2011). This work has confirmed
the strong correlation between colour and morphology of
galaxies, but in addition hi-lighted objects in which the
colour and morphology have become decoupled, such as the
blue ellipticals described above, and also a population of
“red spirals” (Bamford et al. 2009; Masters et al. 2010b),
which have also been found elsewhere (e.g. Wolf et al. 2009;
Bundy et al. 2010) and represent a significant fraction of the
massive spiral population, as well as possibly an important
evolutionary stepping stone for all massive late-types mov-
ing to the red sequence. Moreover, such work has re-defined
the classic morphology-density relation (Dressler et al. 1980)
at low redshift as predominantly a colour-density relation,
i.e., at a fixed galaxy colour, there is a weak morphology–
density relationship, while at a fixed morphology, there re-
mains a significant range of colours with density (see Blanton
et al. 2005, Ball, Loveday & Brunner 2008; Bamford et al.
2009; Skibba et al. 2009).
In this paper we shall study the visual morphology, and
catalogued sizes, of galaxies in the BOSS survey which have
been targeted using colours predicted by population models.
We shall accomplish the double aim of checking the effective-
ness of a purely colour-based selection in targeting mostly
passive early-type galaxies as well as determining the rela-
tion between colour and morphology at these redshifts. With
a median seeing of 1.1′′for the SDSS DR8 imaging data, the
BOSS target galaxies are barely resolved (see Figures 2, 3
& 5) thus making it difficult to obtain reliable visual mor-
phologies or sizes from these data. Therefore, higher angular
resolution (and deeper) imaging is needed to determine the
detailed shapes of BOSS galaxies, and the best facility cur-
rently available for studying such high-redshift galaxy pop-
ulations is the Hubble Space Telescope (HST).
We construct a sample of BOSS targets detected in the
publicly available HST imaging of the COSMic Origins Sur-
vey (COSMOS; Scoville et al. 2007). Although small (240
objects), such an unbiased match between COSMOS and
BOSS provides an important sample of objects for statisti-
cal studies and is unlikely to grow in size or depth over the
next few years.
Where appropriate, we assume a cosmological model of
a flat, ΛCDM Universe with H0 = 70 km s−1Mpc−1
ΩM
= 0.3 consistent with recent WMAP observations
(Komatsu et al. 2011).
and
2 SAMPLE AND DATA
The SDSS-III DR8 imaging data covers 15,000 deg2of high
galactic latitude sky comprising of the SDSS-I/II Legacy
survey and 3000 deg2of new imaging concentrated in
the Southern Hemisphere (York et al. 2000; Aihara et al.
2011). Images in five filters (ugriz, Fukugita et al. 1996) are
obtained with a CCD camera (Gunn et al. 1998) mounted
on the SDSS Telescope (Gunn et al. 2006).
Spectroscopic targets for BOSS are selected from the
high Galactic latitude regions of this imaging (around 10,000
deg2in total) using colour cuts designed to select luminous,
massive galaxies with an approximately uniform distribution
c ? 0000 RAS, MNRAS 000, 000–000

Page 3

The Morphology of BOSS Galaxies
3
in stellar mass over the redshift range 0.3 < z < 0.7. The
details of target selection are given in Eisenstein et al. (2011)
and summarised in Section 2.3 below. The BOSS galaxy
sample is designed to have a total angular source density
of approximately 150 galaxies deg−2which is an order of
magnitude greater than the SDSS-I/II LRG sample.
2.1COSMOS HST Imaging
COSMOS is the largest HST survey yet undertaken and was
specifically designed to survey all types of galaxy environ-
ments at z > 0.5 (Scoville et al. 2007). The HST imag-
ing for COSMOS (Koekemoer et al. 2007; Leauthaud et al.
2007) is publicly available1over a 1.64 deg2field centred at
α = 10h, δ = 2.2◦(J2000). The COSMOS field is covered
by the SDSS-I/II Legacy imaging data and in total, 240
BOSS targets are found in common between the two sets
of images, consistent with the expected source density of
BOSS targets (i.e.. 240 targets over 1.68 deg2= 143 galax-
ies deg−2). This COSMOS-BOSS sample can be considered
largely representative of the entire BOSS galaxy sample,
particularly in the higher redshift CMASS sub-sample (see
Section 2.3), but we note that in the lower redshift LOZ
subset (0.2 < z < 0.4), the presence of a significant over-
density in the COSMOS field at z = 0.35 (e.g. Kovaˇ c et al.
2010) may bias the morphological mix towards red, early
types through the morphology-density relation.
We show the sky distribution of these 240 BOSS tar-
gets compared to all COSMOS identified galaxies in Figure
1. For these 240 BOSS targets, we have visually inspected
images taken by the Advanced Camera for Surveys (ACS)
instrument on HST in the I-band (F814W) filter, which at
the typical redshift for BOSS, corresponds to a rest–frame
V-band filter. We also inspect I- and V-band colour com-
posite images provided by Griffith et al. (in prep.)2and a
different set of colour composite images made for the Galaxy
Zoo: Hubble project (Lintott et al. in prep).
We note that ACS images of other BOSS galaxies will
be available in the HST public archive in images taken for
other observations. A preliminary search of the HST ACS
public archive finds 795 images to the same depth as the
COSMOS data in the F814W filter. Of these, only 563 were
unique (i.e., the centres of the pointings are separated by
more than 10 arcseconds), and 398 of these 563 images are
within the approximate SDSS-III BOSS imaging area. This
is less than the 567 ACS images which make up COSMOS,
and is roughly 70% of the area of COSMOS so would add a
likely ≃ 170 more galaxies to our comparison sample (assum-
ing the same surface density of BOSS galaxies). While this
figure could increase the sample size of our study, we note
that these additional ACS fields, and associated galaxies,
could be potentially biased in their morphological mix, e.g.,
if found in HST imaging of the cores of groups and clusters
of galaxies, whereas the COSMOS field would specifically
avoid such bias. We also make use of the publicly available
Zurich Estimator of Structural Types (ZEST) measurements
(Scarlata et al. 2007) for our BOSS galaxies in COSMOS
and such measurements have not been made for the entire
1See http://irsa.ipac.caltech.edu
2http://www.ugastro.berkeley.edu/∼rgriffit/Morphologies/
149.4 149.6 149.8 150.0 150.2 150.4 150.6 150.8
RA (J2000 degrees)
1.6
1.8
2.0
2.2
2.4
2.6
2.8
Dec (J2000 degrees)
Figure 1. The sky distribution of BOSS target galaxies in the
COSMOS field is shown as coloured points over plotted on the
positions of all galaxies detected by COSMOS (small grey points).
We identify BOSS targets by their sub-sample (see Section 2.3);
CMASS objects are shown in red, while LOZ objects are shown
in blue. Two of the targets fit both target criteria and are shown
as purple.
HST archive. Hence, we decide to rely only on BOSS galax-
ies imaged by HST in the COSMOS area.
2.2BOSS Imaging and Spectra
Throughout this paper, we use the SDSS-III DR8 pipeline
photometric measurements as described in Aihara et al.
(2011). Information on the SDSS pipeline and algorithms
is also given in Stoughton et al. (2002).
For the colours of BOSS galaxies, we use the model mag-
nitudes. In the SDSS pipeline (Stoughton et al. 2002), both
a de Vaucouleurs and an exponential profile fit is performed
on each galaxy, and the best fit model is a linear combina-
tion of the two profiles. Model magnitudes are calculated
using the profile shape fit in r-band and applied to each
of the five SDSS bands (allowing for a variable amplitude).
For apparent luminosities, we use cmodel magnitudes where
the model is fit individually in each SDSS band. All magni-
tudes are corrected for Galactic extinction using reddening
estimates from the DIRBE maps by Schlegel, Finkbeiner &
Davis (1998).
For SDSS measured galaxy sizes, we use the effective ra-
dius from the de Vaucouleur fit. This is a seeing–corrected
quantity, albeit with some approximations made, so seeing
may have a residual effect as we will discuss further in Sec-
tion 4, and especially because the typical sizes of BOSS
galaxies are comparable to the size of the seeing disk.
The SDSS seeing in the COSMOS field is typical for the
whole DR8 imaging, e.g., the median seeing for the whole
SDSS is 1.08′′with an interquartile range of 0.95-1.22′′(Ross
et al. 2011)3. The median seeing for our subset of BOSS
targets in COSMOS is 1.10′′, with an interquartile range of
3The improvement in seeing over previous SDSS imaging releases
is due to the large fraction of repeat images that have been taken.
In such cases, the best seeing is selected for release, thus reducing
the median seeing level significantly over earlier data releases.
c ? 0000 RAS, MNRAS 000, 000–000

Page 4

4
K.L. Masters et al.
1.02-1.16′′(we are quoting the parameter PSF FWHM in i-band
from DR8).
An important caveat to this work is that the SDSS
imaging in the COSMOS field has a higher sky noise level
than most of the other DR8 imaging. The median sky noise
level in the whole DR8 is 20.27 mags (Ross et al. 2011)4,
with an upper quartile of 20.11 mags. Almost all (97%) of
the BOSS targets in the COSMOS imaging have SDSS sky
noise values in this upper quartile, which of course trans-
lates to noisier photometry, and shallower images than are
typical for DR8. This may have an effect, in particular, on
the SDSS measured sizes.
BOSS spectra were taken for 171 of the BOSS galaxy
targets on 10th March 2011 (note that the entire COSMOS
area fits within a single SDSS plate). These observations re-
sulted in 166 BOSS galaxy redshifts. The observations which
did not result in galaxy redshifts are made up of 2 spectra
which had unreliable fits, or zwarning = 4; and 3 spectra of
stars in our Galaxy. An additional 58 of the targets in our
sample had redshift measurements which were taken as part
of the SDSSI/II program. The remaining 11 targets could
not be observed due to problems with fiber collisions. In
total this adds up to 224 of the BOSS targets (93%) hav-
ing reliable galaxy redshifts. Where appropriate, we use the
redshift measurement for these galaxies, but do not use any
additional information from the BOSS (or SDSS) spectra.
We defer such an analysis for future work.
2.3BOSS Target Selection
The BOSS galaxy target selection is based on colour and ap-
parent magnitude cuts defined in the observed-frame g,r,i
photometry which have been derived via population synthe-
sis models. It also includes cuts based on PSF magnitudes
which are used to reduce stellar contamination. The exact
cuts are given in Eisenstein et al. (2011). For the scope of this
paper, it suffices to recall that the BOSS target selections
are split into two sub-samples which are aimed at selecting
galaxies in the different redshift ranges of BOSS, namely:
(1) a Low Redshift Sample (LOZ), aimed at selecting lumi-
nous massive galaxies at 0.2 < z < 0.4 (see the area below
the dashed line in Figure 4); and (2) a higher redshift Con-
stant Mass Sample (CMASS), selecting luminous and mas-
sive galaxies at z > 0.4 (area above the dashed line in Figure
4). Note that the CMASS, high-z cut is a new BOSS cut,
while the LOZ cut is similar to the original Luminous Red
Galaxy (LRG) Cut I selection from SDSS-I/II (Eisenstein
et al. 2001).
The BOSS target selection cuts are based on the ex-
pected track of a passively–evolving, constant stellar mass
galaxy based on a population model for luminous red galax-
ies by Maraston et al. (2009). As has been discussed pre-
viously in both Eisenstein et al. (2011) and White et al.
(2011), the CMASS colour selection closely tracks the red-
shifted colours of luminous and massive galaxies at z ≃ 0.5,
while the (colour dependent) i-band magnitude constraint in
4Ross et al. (2011) report this value expressed in “nanomaggies”,
which are a linear unit of flux used in the SDSS collaboration.
We use the conversion from a flux, f, in nanomaggies to, m in
magnitudes of m = 22.5 mag − 2.5logf
CMASS is aimed at selecting objects with M⋆ ≃ 1011M⊙,
independent of their intrinsic rest-frame colours. Figure 1
of White et al. (2011) demonstrates that this set of cuts
successfully isolates galaxies at z > 0.4.
What is most relevant to our study is that: i) the main
colour cut for CMASS (the selection of d⊥= (r − i) − (g −
r)/8.0 > 0.55) is basically a selection on the observed-frame
colour (r−i) (i.e. the line is almost horizontal in Figure 4).
All galaxies in the redshift range of CMASS (0.4 < z < 0.7)
have their 4000˚ A break feature somewhere in the r-band
since at z = 0.4 the break is exactly between g and r at
5600˚ A, and at z = 0.75 it moves into i-band (at 7000˚ A).
This means that all galaxies at these redshifts are red in
observed frame (r − i), regardless of stellar content; ii) no
strong cut was made in the observed-frame (g − r), which
would select approximately in intrinsic (u − g) colour at
these redshifts – a colour which is very sensitive to stellar
content; iii) a (slightly colour dependent) i-band magnitude
based cuts is applied to require that (only) massive galaxies
are selected.
The stellar population model based target selection for
BOSS described above, implies that we should find within
the sample mostly early-type galaxies – since this type of
galaxy dominates the massive galaxy population. But, given
the absence of a red colour cut in (g − r), massive blue
galaxies could also be found. We shall see in the next Section
(Figure 4) that the real morphologies of the BOSS galaxies
in our sample reflect well this expectation.
3VISUAL MORPHOLOGY
We visually inspected both the COSMOS and SDSS im-
ages of all 240 BOSS target galaxies in the COSMOS area.
The COSMOS images were used to first classify galaxies as
either “early” or “late-type” morphology. We then further
separated these broad classes into possibly lenticular (S0)
early-type galaxies, disturbed galaxies, and/or multiple sys-
tems (including multiple cores). For the late-type galaxies,
we also recorded if the galaxy was viewed as “edge-on” disc
or possessed a bar through the middle of the galaxy.
Table 1 shows the morphological mix of our sample,
and we present all of our visual classifications in Appendix
A. We find that the majority of BOSS galaxies in our COS-
MOS sample are early-type galaxies (74±6%; we quote the
Poisson error on fractions throughout, but note that this is
strictly only valid for large sample sizes and fractions neither
close to 0% or 100%. The error we quote will therefore be
an underestimate of the true confidence region particularly
for the small fractions and subsets; see (Cameron 2011) for
more details), but there is a substantial fraction of late-type
galaxies present (24±3%). Representative images from
each morphological class of object are provided in Figure
2 for the CMASS sample and in Figure 3 for the LOZ
sample, and all 240 galaxies are available for inspection at
http://icg.port.ac.uk/∼mastersk/BOSSmorphologies/
which shows both the SDSS gri colour composites alongside
the COSMOS HST I-band and colour composite images.
c ? 0000 RAS, MNRAS 000, 000–000

Page 5

The Morphology of BOSS Galaxies
5
!"#$%&'%()*&+,-./&
0"1)&'%()*&+23./&
456&
78*19#:);&
<9$=($)&
41"#*&+2./&
>"##);&
!;?)@AB&
78*19#:);&
<9$=($)&
CD5E,FG2&
CD5EG5HF&
CD5EGIH3&
CD5E3G5J&
CD5EG,GH&
CD5EF,I2&
CD5E3-53&
CKL4<L4D5EF-I3&
CD5E3I,G&
Figure 2. Examples of SDSS composite colour gri images and ACS I-band images (shown as black on white) for BOSS CMASS galaxies
with different morphological types. Shown are 10 examples out of a total sample of 129 objects. All images are 15′′square, and the BOSS
2′′diameter fibre is illustrated by a green circle. The redshifts of these objects are shown in the figure. The barred late–type was not
observed by BOSS due to fibre collisions, but has a redshift from zCOSMOS (Lilly et al. 2007). The double point source system has also
been observed spectroscopically by BOSS and is confirmed to be made up of stars in our Galaxy (i.e. z = 0).
c ? 0000 RAS, MNRAS 000, 000–000